linux/fs/btrfs/extent-tree.c

5939 lines
154 KiB
C
Raw Normal View History

/*
* Copyright (C) 2007 Oracle. All rights reserved.
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public
* License v2 as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* General Public License for more details.
*
* You should have received a copy of the GNU General Public
* License along with this program; if not, write to the
* Free Software Foundation, Inc., 59 Temple Place - Suite 330,
* Boston, MA 021110-1307, USA.
*/
#include <linux/sched.h>
#include <linux/pagemap.h>
#include <linux/writeback.h>
#include <linux/blkdev.h>
#include <linux/version.h>
#include "compat.h"
#include "hash.h"
#include "crc32c.h"
#include "ctree.h"
#include "disk-io.h"
#include "print-tree.h"
#include "transaction.h"
#include "volumes.h"
#include "locking.h"
#include "ref-cache.h"
#include "compat.h"
#define PENDING_EXTENT_INSERT 0
#define PENDING_EXTENT_DELETE 1
#define PENDING_BACKREF_UPDATE 2
struct pending_extent_op {
int type;
u64 bytenr;
u64 num_bytes;
u64 parent;
u64 orig_parent;
u64 generation;
u64 orig_generation;
int level;
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
struct list_head list;
int del;
};
static int finish_current_insert(struct btrfs_trans_handle *trans, struct
btrfs_root *extent_root, int all);
static int del_pending_extents(struct btrfs_trans_handle *trans, struct
btrfs_root *extent_root, int all);
static struct btrfs_block_group_cache *
__btrfs_find_block_group(struct btrfs_root *root,
struct btrfs_block_group_cache *hint,
u64 search_start, int data, int owner);
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
static int pin_down_bytes(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
u64 bytenr, u64 num_bytes, int is_data);
static int update_block_group(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
u64 bytenr, u64 num_bytes, int alloc,
int mark_free);
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
static int block_group_bits(struct btrfs_block_group_cache *cache, u64 bits)
{
return (cache->flags & bits) == bits;
}
/*
* this adds the block group to the fs_info rb tree for the block group
* cache
*/
int btrfs_add_block_group_cache(struct btrfs_fs_info *info,
struct btrfs_block_group_cache *block_group)
{
struct rb_node **p;
struct rb_node *parent = NULL;
struct btrfs_block_group_cache *cache;
spin_lock(&info->block_group_cache_lock);
p = &info->block_group_cache_tree.rb_node;
while (*p) {
parent = *p;
cache = rb_entry(parent, struct btrfs_block_group_cache,
cache_node);
if (block_group->key.objectid < cache->key.objectid) {
p = &(*p)->rb_left;
} else if (block_group->key.objectid > cache->key.objectid) {
p = &(*p)->rb_right;
} else {
spin_unlock(&info->block_group_cache_lock);
return -EEXIST;
}
}
rb_link_node(&block_group->cache_node, parent, p);
rb_insert_color(&block_group->cache_node,
&info->block_group_cache_tree);
spin_unlock(&info->block_group_cache_lock);
return 0;
}
/*
* This will return the block group at or after bytenr if contains is 0, else
* it will return the block group that contains the bytenr
*/
static struct btrfs_block_group_cache *
block_group_cache_tree_search(struct btrfs_fs_info *info, u64 bytenr,
int contains)
{
struct btrfs_block_group_cache *cache, *ret = NULL;
struct rb_node *n;
u64 end, start;
spin_lock(&info->block_group_cache_lock);
n = info->block_group_cache_tree.rb_node;
while (n) {
cache = rb_entry(n, struct btrfs_block_group_cache,
cache_node);
end = cache->key.objectid + cache->key.offset - 1;
start = cache->key.objectid;
if (bytenr < start) {
if (!contains && (!ret || start < ret->key.objectid))
ret = cache;
n = n->rb_left;
} else if (bytenr > start) {
if (contains && bytenr <= end) {
ret = cache;
break;
}
n = n->rb_right;
} else {
ret = cache;
break;
}
}
spin_unlock(&info->block_group_cache_lock);
return ret;
}
/*
* this is only called by cache_block_group, since we could have freed extents
* we need to check the pinned_extents for any extents that can't be used yet
* since their free space will be released as soon as the transaction commits.
*/
static int add_new_free_space(struct btrfs_block_group_cache *block_group,
struct btrfs_fs_info *info, u64 start, u64 end)
{
u64 extent_start, extent_end, size;
int ret;
mutex_lock(&info->pinned_mutex);
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
while (start < end) {
ret = find_first_extent_bit(&info->pinned_extents, start,
&extent_start, &extent_end,
EXTENT_DIRTY);
if (ret)
break;
if (extent_start == start) {
start = extent_end + 1;
} else if (extent_start > start && extent_start < end) {
size = extent_start - start;
ret = btrfs_add_free_space(block_group, start,
size);
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
BUG_ON(ret);
start = extent_end + 1;
} else {
break;
}
}
if (start < end) {
size = end - start;
ret = btrfs_add_free_space(block_group, start, size);
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
BUG_ON(ret);
}
mutex_unlock(&info->pinned_mutex);
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
return 0;
}
static int cache_block_group(struct btrfs_root *root,
struct btrfs_block_group_cache *block_group)
{
struct btrfs_path *path;
int ret = 0;
struct btrfs_key key;
struct extent_buffer *leaf;
int slot;
u64 last = 0;
u64 first_free;
int found = 0;
if (!block_group)
return 0;
root = root->fs_info->extent_root;
if (block_group->cached)
return 0;
path = btrfs_alloc_path();
if (!path)
return -ENOMEM;
path->reada = 2;
/*
* we get into deadlocks with paths held by callers of this function.
* since the alloc_mutex is protecting things right now, just
* skip the locking here
*/
path->skip_locking = 1;
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
first_free = max_t(u64, block_group->key.objectid,
BTRFS_SUPER_INFO_OFFSET + BTRFS_SUPER_INFO_SIZE);
key.objectid = block_group->key.objectid;
key.offset = 0;
btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
if (ret < 0)
goto err;
ret = btrfs_previous_item(root, path, 0, BTRFS_EXTENT_ITEM_KEY);
if (ret < 0)
goto err;
if (ret == 0) {
leaf = path->nodes[0];
btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
if (key.objectid + key.offset > first_free)
first_free = key.objectid + key.offset;
}
while(1) {
leaf = path->nodes[0];
slot = path->slots[0];
if (slot >= btrfs_header_nritems(leaf)) {
ret = btrfs_next_leaf(root, path);
if (ret < 0)
goto err;
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
if (ret == 0)
continue;
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
else
break;
}
btrfs_item_key_to_cpu(leaf, &key, slot);
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
if (key.objectid < block_group->key.objectid)
goto next;
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
if (key.objectid >= block_group->key.objectid +
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
block_group->key.offset)
break;
if (btrfs_key_type(&key) == BTRFS_EXTENT_ITEM_KEY) {
if (!found) {
last = first_free;
found = 1;
}
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
add_new_free_space(block_group, root->fs_info, last,
key.objectid);
last = key.objectid + key.offset;
}
next:
path->slots[0]++;
}
if (!found)
last = first_free;
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
add_new_free_space(block_group, root->fs_info, last,
block_group->key.objectid +
block_group->key.offset);
block_group->cached = 1;
ret = 0;
err:
btrfs_free_path(path);
return ret;
}
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
/*
* return the block group that starts at or after bytenr
*/
struct btrfs_block_group_cache *btrfs_lookup_first_block_group(struct
btrfs_fs_info *info,
u64 bytenr)
{
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
struct btrfs_block_group_cache *cache;
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
cache = block_group_cache_tree_search(info, bytenr, 0);
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
return cache;
}
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
/*
* return the block group that contains teh given bytenr
*/
struct btrfs_block_group_cache *btrfs_lookup_block_group(struct
btrfs_fs_info *info,
u64 bytenr)
{
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
struct btrfs_block_group_cache *cache;
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
cache = block_group_cache_tree_search(info, bytenr, 1);
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
return cache;
}
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
static struct btrfs_space_info *__find_space_info(struct btrfs_fs_info *info,
u64 flags)
{
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
struct list_head *head = &info->space_info;
struct list_head *cur;
struct btrfs_space_info *found;
list_for_each(cur, head) {
found = list_entry(cur, struct btrfs_space_info, list);
if (found->flags == flags)
return found;
}
return NULL;
}
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
static u64 div_factor(u64 num, int factor)
{
if (factor == 10)
return num;
num *= factor;
do_div(num, 10);
return num;
}
static struct btrfs_block_group_cache *
__btrfs_find_block_group(struct btrfs_root *root,
struct btrfs_block_group_cache *hint,
u64 search_start, int data, int owner)
{
struct btrfs_block_group_cache *cache;
struct btrfs_block_group_cache *found_group = NULL;
struct btrfs_fs_info *info = root->fs_info;
u64 used;
u64 last = 0;
u64 free_check;
int full_search = 0;
int factor = 10;
int wrapped = 0;
if (data & BTRFS_BLOCK_GROUP_METADATA)
factor = 9;
if (search_start) {
struct btrfs_block_group_cache *shint;
shint = btrfs_lookup_first_block_group(info, search_start);
if (shint && block_group_bits(shint, data)) {
spin_lock(&shint->lock);
used = btrfs_block_group_used(&shint->item);
if (used + shint->pinned + shint->reserved <
div_factor(shint->key.offset, factor)) {
spin_unlock(&shint->lock);
return shint;
}
spin_unlock(&shint->lock);
}
}
if (hint && block_group_bits(hint, data)) {
spin_lock(&hint->lock);
used = btrfs_block_group_used(&hint->item);
if (used + hint->pinned + hint->reserved <
div_factor(hint->key.offset, factor)) {
spin_unlock(&hint->lock);
return hint;
}
spin_unlock(&hint->lock);
last = hint->key.objectid + hint->key.offset;
} else {
if (hint)
last = max(hint->key.objectid, search_start);
else
last = search_start;
}
again:
while (1) {
cache = btrfs_lookup_first_block_group(root->fs_info, last);
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
if (!cache)
break;
spin_lock(&cache->lock);
last = cache->key.objectid + cache->key.offset;
used = btrfs_block_group_used(&cache->item);
if (block_group_bits(cache, data)) {
free_check = div_factor(cache->key.offset, factor);
if (used + cache->pinned + cache->reserved <
free_check) {
found_group = cache;
spin_unlock(&cache->lock);
goto found;
}
}
spin_unlock(&cache->lock);
cond_resched();
}
if (!wrapped) {
last = search_start;
wrapped = 1;
goto again;
}
if (!full_search && factor < 10) {
last = search_start;
full_search = 1;
factor = 10;
goto again;
}
found:
return found_group;
}
struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root,
struct btrfs_block_group_cache
*hint, u64 search_start,
int data, int owner)
{
struct btrfs_block_group_cache *ret;
ret = __btrfs_find_block_group(root, hint, search_start, data, owner);
return ret;
}
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
/* simple helper to search for an existing extent at a given offset */
int btrfs_lookup_extent(struct btrfs_root *root, u64 start, u64 len)
{
int ret;
struct btrfs_key key;
struct btrfs_path *path;
path = btrfs_alloc_path();
BUG_ON(!path);
key.objectid = start;
key.offset = len;
btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
ret = btrfs_search_slot(NULL, root->fs_info->extent_root, &key, path,
0, 0);
btrfs_free_path(path);
return ret;
}
/*
* Back reference rules. Back refs have three main goals:
*
* 1) differentiate between all holders of references to an extent so that
* when a reference is dropped we can make sure it was a valid reference
* before freeing the extent.
*
* 2) Provide enough information to quickly find the holders of an extent
* if we notice a given block is corrupted or bad.
*
* 3) Make it easy to migrate blocks for FS shrinking or storage pool
* maintenance. This is actually the same as #2, but with a slightly
* different use case.
*
* File extents can be referenced by:
*
* - multiple snapshots, subvolumes, or different generations in one subvol
* - different files inside a single subvolume
* - different offsets inside a file (bookend extents in file.c)
*
* The extent ref structure has fields for:
*
* - Objectid of the subvolume root
* - Generation number of the tree holding the reference
* - objectid of the file holding the reference
* - number of references holding by parent node (alway 1 for tree blocks)
*
* Btree leaf may hold multiple references to a file extent. In most cases,
* these references are from same file and the corresponding offsets inside
* the file are close together.
*
* When a file extent is allocated the fields are filled in:
* (root_key.objectid, trans->transid, inode objectid, 1)
*
* When a leaf is cow'd new references are added for every file extent found
* in the leaf. It looks similar to the create case, but trans->transid will
* be different when the block is cow'd.
*
* (root_key.objectid, trans->transid, inode objectid,
* number of references in the leaf)
*
* When a file extent is removed either during snapshot deletion or
* file truncation, we find the corresponding back reference and check
* the following fields:
*
* (btrfs_header_owner(leaf), btrfs_header_generation(leaf),
* inode objectid)
*
* Btree extents can be referenced by:
*
* - Different subvolumes
* - Different generations of the same subvolume
*
* When a tree block is created, back references are inserted:
*
* (root->root_key.objectid, trans->transid, level, 1)
*
* When a tree block is cow'd, new back references are added for all the
* blocks it points to. If the tree block isn't in reference counted root,
* the old back references are removed. These new back references are of
* the form (trans->transid will have increased since creation):
*
* (root->root_key.objectid, trans->transid, level, 1)
*
* When a backref is in deleting, the following fields are checked:
*
* if backref was for a tree root:
* (btrfs_header_owner(itself), btrfs_header_generation(itself), level)
* else
* (btrfs_header_owner(parent), btrfs_header_generation(parent), level)
*
* Back Reference Key composing:
*
* The key objectid corresponds to the first byte in the extent, the key
* type is set to BTRFS_EXTENT_REF_KEY, and the key offset is the first
* byte of parent extent. If a extent is tree root, the key offset is set
* to the key objectid.
*/
static int noinline lookup_extent_backref(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct btrfs_path *path,
u64 bytenr, u64 parent,
u64 ref_root, u64 ref_generation,
u64 owner_objectid, int del)
{
struct btrfs_key key;
struct btrfs_extent_ref *ref;
struct extent_buffer *leaf;
u64 ref_objectid;
int ret;
key.objectid = bytenr;
key.type = BTRFS_EXTENT_REF_KEY;
key.offset = parent;
ret = btrfs_search_slot(trans, root, &key, path, del ? -1 : 0, 1);
if (ret < 0)
goto out;
if (ret > 0) {
ret = -ENOENT;
goto out;
}
leaf = path->nodes[0];
ref = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_ref);
ref_objectid = btrfs_ref_objectid(leaf, ref);
if (btrfs_ref_root(leaf, ref) != ref_root ||
btrfs_ref_generation(leaf, ref) != ref_generation ||
(ref_objectid != owner_objectid &&
ref_objectid != BTRFS_MULTIPLE_OBJECTIDS)) {
ret = -EIO;
WARN_ON(1);
goto out;
}
ret = 0;
out:
return ret;
}
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
/*
* updates all the backrefs that are pending on update_list for the
* extent_root
*/
static int noinline update_backrefs(struct btrfs_trans_handle *trans,
struct btrfs_root *extent_root,
struct btrfs_path *path,
struct list_head *update_list)
{
struct btrfs_key key;
struct btrfs_extent_ref *ref;
struct btrfs_fs_info *info = extent_root->fs_info;
struct pending_extent_op *op;
struct extent_buffer *leaf;
int ret = 0;
struct list_head *cur = update_list->next;
u64 ref_objectid;
u64 ref_root = extent_root->root_key.objectid;
op = list_entry(cur, struct pending_extent_op, list);
search:
key.objectid = op->bytenr;
key.type = BTRFS_EXTENT_REF_KEY;
key.offset = op->orig_parent;
ret = btrfs_search_slot(trans, extent_root, &key, path, 0, 1);
BUG_ON(ret);
leaf = path->nodes[0];
loop:
ref = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_ref);
ref_objectid = btrfs_ref_objectid(leaf, ref);
if (btrfs_ref_root(leaf, ref) != ref_root ||
btrfs_ref_generation(leaf, ref) != op->orig_generation ||
(ref_objectid != op->level &&
ref_objectid != BTRFS_MULTIPLE_OBJECTIDS)) {
printk(KERN_ERR "couldn't find %Lu, parent %Lu, root %Lu, "
"owner %u\n", op->bytenr, op->orig_parent,
ref_root, op->level);
btrfs_print_leaf(extent_root, leaf);
BUG();
}
key.objectid = op->bytenr;
key.offset = op->parent;
key.type = BTRFS_EXTENT_REF_KEY;
ret = btrfs_set_item_key_safe(trans, extent_root, path, &key);
BUG_ON(ret);
ref = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_ref);
btrfs_set_ref_generation(leaf, ref, op->generation);
cur = cur->next;
list_del_init(&op->list);
unlock_extent(&info->extent_ins, op->bytenr,
op->bytenr + op->num_bytes - 1, GFP_NOFS);
kfree(op);
if (cur == update_list) {
btrfs_mark_buffer_dirty(path->nodes[0]);
btrfs_release_path(extent_root, path);
goto out;
}
op = list_entry(cur, struct pending_extent_op, list);
path->slots[0]++;
while (path->slots[0] < btrfs_header_nritems(leaf)) {
btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
if (key.objectid == op->bytenr &&
key.type == BTRFS_EXTENT_REF_KEY)
goto loop;
path->slots[0]++;
}
btrfs_mark_buffer_dirty(path->nodes[0]);
btrfs_release_path(extent_root, path);
goto search;
out:
return 0;
}
static int noinline insert_extents(struct btrfs_trans_handle *trans,
struct btrfs_root *extent_root,
struct btrfs_path *path,
struct list_head *insert_list, int nr)
{
struct btrfs_key *keys;
u32 *data_size;
struct pending_extent_op *op;
struct extent_buffer *leaf;
struct list_head *cur = insert_list->next;
struct btrfs_fs_info *info = extent_root->fs_info;
u64 ref_root = extent_root->root_key.objectid;
int i = 0, last = 0, ret;
int total = nr * 2;
if (!nr)
return 0;
keys = kzalloc(total * sizeof(struct btrfs_key), GFP_NOFS);
if (!keys)
return -ENOMEM;
data_size = kzalloc(total * sizeof(u32), GFP_NOFS);
if (!data_size) {
kfree(keys);
return -ENOMEM;
}
list_for_each_entry(op, insert_list, list) {
keys[i].objectid = op->bytenr;
keys[i].offset = op->num_bytes;
keys[i].type = BTRFS_EXTENT_ITEM_KEY;
data_size[i] = sizeof(struct btrfs_extent_item);
i++;
keys[i].objectid = op->bytenr;
keys[i].offset = op->parent;
keys[i].type = BTRFS_EXTENT_REF_KEY;
data_size[i] = sizeof(struct btrfs_extent_ref);
i++;
}
op = list_entry(cur, struct pending_extent_op, list);
i = 0;
while (i < total) {
int c;
ret = btrfs_insert_some_items(trans, extent_root, path,
keys+i, data_size+i, total-i);
BUG_ON(ret < 0);
if (last && ret > 1)
BUG();
leaf = path->nodes[0];
for (c = 0; c < ret; c++) {
int ref_first = keys[i].type == BTRFS_EXTENT_REF_KEY;
/*
* if the first item we inserted was a backref, then
* the EXTENT_ITEM will be the odd c's, else it will
* be the even c's
*/
if ((ref_first && (c % 2)) ||
(!ref_first && !(c % 2))) {
struct btrfs_extent_item *itm;
itm = btrfs_item_ptr(leaf, path->slots[0] + c,
struct btrfs_extent_item);
btrfs_set_extent_refs(path->nodes[0], itm, 1);
op->del++;
} else {
struct btrfs_extent_ref *ref;
ref = btrfs_item_ptr(leaf, path->slots[0] + c,
struct btrfs_extent_ref);
btrfs_set_ref_root(leaf, ref, ref_root);
btrfs_set_ref_generation(leaf, ref,
op->generation);
btrfs_set_ref_objectid(leaf, ref, op->level);
btrfs_set_ref_num_refs(leaf, ref, 1);
op->del++;
}
/*
* using del to see when its ok to free up the
* pending_extent_op. In the case where we insert the
* last item on the list in order to help do batching
* we need to not free the extent op until we actually
* insert the extent_item
*/
if (op->del == 2) {
unlock_extent(&info->extent_ins, op->bytenr,
op->bytenr + op->num_bytes - 1,
GFP_NOFS);
cur = cur->next;
list_del_init(&op->list);
kfree(op);
if (cur != insert_list)
op = list_entry(cur,
struct pending_extent_op,
list);
}
}
btrfs_mark_buffer_dirty(leaf);
btrfs_release_path(extent_root, path);
/*
* Ok backref's and items usually go right next to eachother,
* but if we could only insert 1 item that means that we
* inserted on the end of a leaf, and we have no idea what may
* be on the next leaf so we just play it safe. In order to
* try and help this case we insert the last thing on our
* insert list so hopefully it will end up being the last
* thing on the leaf and everything else will be before it,
* which will let us insert a whole bunch of items at the same
* time.
*/
if (ret == 1 && !last && (i + ret < total)) {
/*
* last: where we will pick up the next time around
* i: our current key to insert, will be total - 1
* cur: the current op we are screwing with
* op: duh
*/
last = i + ret;
i = total - 1;
cur = insert_list->prev;
op = list_entry(cur, struct pending_extent_op, list);
} else if (last) {
/*
* ok we successfully inserted the last item on the
* list, lets reset everything
*
* i: our current key to insert, so where we left off
* last time
* last: done with this
* cur: the op we are messing with
* op: duh
* total: since we inserted the last key, we need to
* decrement total so we dont overflow
*/
i = last;
last = 0;
total--;
if (i < total) {
cur = insert_list->next;
op = list_entry(cur, struct pending_extent_op,
list);
}
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
} else {
i += ret;
}
cond_resched();
}
ret = 0;
kfree(keys);
kfree(data_size);
return ret;
}
static int noinline insert_extent_backref(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct btrfs_path *path,
u64 bytenr, u64 parent,
u64 ref_root, u64 ref_generation,
u64 owner_objectid)
{
struct btrfs_key key;
struct extent_buffer *leaf;
struct btrfs_extent_ref *ref;
u32 num_refs;
int ret;
key.objectid = bytenr;
key.type = BTRFS_EXTENT_REF_KEY;
key.offset = parent;
ret = btrfs_insert_empty_item(trans, root, path, &key, sizeof(*ref));
if (ret == 0) {
leaf = path->nodes[0];
ref = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_extent_ref);
btrfs_set_ref_root(leaf, ref, ref_root);
btrfs_set_ref_generation(leaf, ref, ref_generation);
btrfs_set_ref_objectid(leaf, ref, owner_objectid);
btrfs_set_ref_num_refs(leaf, ref, 1);
} else if (ret == -EEXIST) {
u64 existing_owner;
BUG_ON(owner_objectid < BTRFS_FIRST_FREE_OBJECTID);
leaf = path->nodes[0];
ref = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_extent_ref);
if (btrfs_ref_root(leaf, ref) != ref_root ||
btrfs_ref_generation(leaf, ref) != ref_generation) {
ret = -EIO;
WARN_ON(1);
goto out;
}
num_refs = btrfs_ref_num_refs(leaf, ref);
BUG_ON(num_refs == 0);
btrfs_set_ref_num_refs(leaf, ref, num_refs + 1);
existing_owner = btrfs_ref_objectid(leaf, ref);
if (existing_owner != owner_objectid &&
existing_owner != BTRFS_MULTIPLE_OBJECTIDS) {
btrfs_set_ref_objectid(leaf, ref,
BTRFS_MULTIPLE_OBJECTIDS);
}
ret = 0;
} else {
goto out;
}
btrfs_mark_buffer_dirty(path->nodes[0]);
out:
btrfs_release_path(root, path);
return ret;
}
static int noinline remove_extent_backref(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct btrfs_path *path)
{
struct extent_buffer *leaf;
struct btrfs_extent_ref *ref;
u32 num_refs;
int ret = 0;
leaf = path->nodes[0];
ref = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_ref);
num_refs = btrfs_ref_num_refs(leaf, ref);
BUG_ON(num_refs == 0);
num_refs -= 1;
if (num_refs == 0) {
ret = btrfs_del_item(trans, root, path);
} else {
btrfs_set_ref_num_refs(leaf, ref, num_refs);
btrfs_mark_buffer_dirty(leaf);
}
btrfs_release_path(root, path);
return ret;
}
#ifdef BIO_RW_DISCARD
static void btrfs_issue_discard(struct block_device *bdev,
u64 start, u64 len)
{
#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,28)
blkdev_issue_discard(bdev, start >> 9, len >> 9, GFP_KERNEL);
#else
blkdev_issue_discard(bdev, start >> 9, len >> 9);
#endif
}
#endif
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
static int noinline free_extents(struct btrfs_trans_handle *trans,
struct btrfs_root *extent_root,
struct list_head *del_list)
{
struct btrfs_fs_info *info = extent_root->fs_info;
struct btrfs_path *path;
struct btrfs_key key, found_key;
struct extent_buffer *leaf;
struct list_head *cur;
struct pending_extent_op *op;
struct btrfs_extent_item *ei;
int ret, num_to_del, extent_slot = 0, found_extent = 0;
u32 refs;
u64 bytes_freed = 0;
path = btrfs_alloc_path();
if (!path)
return -ENOMEM;
path->reada = 1;
search:
/* search for the backref for the current ref we want to delete */
cur = del_list->next;
op = list_entry(cur, struct pending_extent_op, list);
ret = lookup_extent_backref(trans, extent_root, path, op->bytenr,
op->orig_parent,
extent_root->root_key.objectid,
op->orig_generation, op->level, 1);
if (ret) {
printk("Unable to find backref byte nr %Lu root %Lu gen %Lu "
"owner %u\n", op->bytenr,
extent_root->root_key.objectid, op->orig_generation,
op->level);
btrfs_print_leaf(extent_root, path->nodes[0]);
WARN_ON(1);
goto out;
}
extent_slot = path->slots[0];
num_to_del = 1;
found_extent = 0;
/*
* if we aren't the first item on the leaf we can move back one and see
* if our ref is right next to our extent item
*/
if (likely(extent_slot)) {
extent_slot--;
btrfs_item_key_to_cpu(path->nodes[0], &found_key,
extent_slot);
if (found_key.objectid == op->bytenr &&
found_key.type == BTRFS_EXTENT_ITEM_KEY &&
found_key.offset == op->num_bytes) {
num_to_del++;
found_extent = 1;
}
}
/*
* if we didn't find the extent we need to delete the backref and then
* search for the extent item key so we can update its ref count
*/
if (!found_extent) {
key.objectid = op->bytenr;
key.type = BTRFS_EXTENT_ITEM_KEY;
key.offset = op->num_bytes;
ret = remove_extent_backref(trans, extent_root, path);
BUG_ON(ret);
btrfs_release_path(extent_root, path);
ret = btrfs_search_slot(trans, extent_root, &key, path, -1, 1);
BUG_ON(ret);
extent_slot = path->slots[0];
}
/* this is where we update the ref count for the extent */
leaf = path->nodes[0];
ei = btrfs_item_ptr(leaf, extent_slot, struct btrfs_extent_item);
refs = btrfs_extent_refs(leaf, ei);
BUG_ON(refs == 0);
refs--;
btrfs_set_extent_refs(leaf, ei, refs);
btrfs_mark_buffer_dirty(leaf);
/*
* This extent needs deleting. The reason cur_slot is extent_slot +
* num_to_del is because extent_slot points to the slot where the extent
* is, and if the backref was not right next to the extent we will be
* deleting at least 1 item, and will want to start searching at the
* slot directly next to extent_slot. However if we did find the
* backref next to the extent item them we will be deleting at least 2
* items and will want to start searching directly after the ref slot
*/
if (!refs) {
struct list_head *pos, *n, *end;
int cur_slot = extent_slot+num_to_del;
u64 super_used;
u64 root_used;
path->slots[0] = extent_slot;
bytes_freed = op->num_bytes;
mutex_lock(&info->pinned_mutex);
ret = pin_down_bytes(trans, extent_root, op->bytenr,
op->num_bytes, op->level >=
BTRFS_FIRST_FREE_OBJECTID);
mutex_unlock(&info->pinned_mutex);
BUG_ON(ret < 0);
op->del = ret;
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
/*
* we need to see if we can delete multiple things at once, so
* start looping through the list of extents we are wanting to
* delete and see if their extent/backref's are right next to
* eachother and the extents only have 1 ref
*/
for (pos = cur->next; pos != del_list; pos = pos->next) {
struct pending_extent_op *tmp;
tmp = list_entry(pos, struct pending_extent_op, list);
/* we only want to delete extent+ref at this stage */
if (cur_slot >= btrfs_header_nritems(leaf) - 1)
break;
btrfs_item_key_to_cpu(leaf, &found_key, cur_slot);
if (found_key.objectid != tmp->bytenr ||
found_key.type != BTRFS_EXTENT_ITEM_KEY ||
found_key.offset != tmp->num_bytes)
break;
/* check to make sure this extent only has one ref */
ei = btrfs_item_ptr(leaf, cur_slot,
struct btrfs_extent_item);
if (btrfs_extent_refs(leaf, ei) != 1)
break;
btrfs_item_key_to_cpu(leaf, &found_key, cur_slot+1);
if (found_key.objectid != tmp->bytenr ||
found_key.type != BTRFS_EXTENT_REF_KEY ||
found_key.offset != tmp->orig_parent)
break;
/*
* the ref is right next to the extent, we can set the
* ref count to 0 since we will delete them both now
*/
btrfs_set_extent_refs(leaf, ei, 0);
/* pin down the bytes for this extent */
mutex_lock(&info->pinned_mutex);
ret = pin_down_bytes(trans, extent_root, tmp->bytenr,
tmp->num_bytes, tmp->level >=
BTRFS_FIRST_FREE_OBJECTID);
mutex_unlock(&info->pinned_mutex);
BUG_ON(ret < 0);
/*
* use the del field to tell if we need to go ahead and
* free up the extent when we delete the item or not.
*/
tmp->del = ret;
bytes_freed += tmp->num_bytes;
num_to_del += 2;
cur_slot += 2;
}
end = pos;
/* update the free space counters */
spin_lock_irq(&info->delalloc_lock);
super_used = btrfs_super_bytes_used(&info->super_copy);
btrfs_set_super_bytes_used(&info->super_copy,
super_used - bytes_freed);
spin_unlock_irq(&info->delalloc_lock);
root_used = btrfs_root_used(&extent_root->root_item);
btrfs_set_root_used(&extent_root->root_item,
root_used - bytes_freed);
/* delete the items */
ret = btrfs_del_items(trans, extent_root, path,
path->slots[0], num_to_del);
BUG_ON(ret);
/*
* loop through the extents we deleted and do the cleanup work
* on them
*/
for (pos = cur, n = pos->next; pos != end;
pos = n, n = pos->next) {
struct pending_extent_op *tmp;
#ifdef BIO_RW_DISCARD
u64 map_length;
struct btrfs_multi_bio *multi = NULL;
#endif
tmp = list_entry(pos, struct pending_extent_op, list);
/*
* remember tmp->del tells us wether or not we pinned
* down the extent
*/
ret = update_block_group(trans, extent_root,
tmp->bytenr, tmp->num_bytes, 0,
tmp->del);
BUG_ON(ret);
#ifdef BIO_RW_DISCARD
map_length = tmp->num_bytes;
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
ret = btrfs_map_block(&info->mapping_tree, READ,
tmp->bytenr, &map_length, &multi,
0);
if (!ret) {
struct btrfs_bio_stripe *stripe;
int i;
stripe = multi->stripes;
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
if (map_length > tmp->num_bytes)
map_length = tmp->num_bytes;
for (i = 0; i < multi->num_stripes;
i++, stripe++)
btrfs_issue_discard(stripe->dev->bdev,
stripe->physical,
map_length);
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
kfree(multi);
}
#endif
list_del_init(&tmp->list);
unlock_extent(&info->extent_ins, tmp->bytenr,
tmp->bytenr + tmp->num_bytes - 1,
GFP_NOFS);
kfree(tmp);
}
} else if (refs && found_extent) {
/*
* the ref and extent were right next to eachother, but the
* extent still has a ref, so just free the backref and keep
* going
*/
ret = remove_extent_backref(trans, extent_root, path);
BUG_ON(ret);
list_del_init(&op->list);
unlock_extent(&info->extent_ins, op->bytenr,
op->bytenr + op->num_bytes - 1, GFP_NOFS);
kfree(op);
} else {
/*
* the extent has multiple refs and the backref we were looking
* for was not right next to it, so just unlock and go next,
* we're good to go
*/
list_del_init(&op->list);
unlock_extent(&info->extent_ins, op->bytenr,
op->bytenr + op->num_bytes - 1, GFP_NOFS);
kfree(op);
}
btrfs_release_path(extent_root, path);
if (!list_empty(del_list))
goto search;
out:
btrfs_free_path(path);
return ret;
}
static int __btrfs_update_extent_ref(struct btrfs_trans_handle *trans,
struct btrfs_root *root, u64 bytenr,
u64 orig_parent, u64 parent,
u64 orig_root, u64 ref_root,
u64 orig_generation, u64 ref_generation,
u64 owner_objectid)
{
int ret;
struct btrfs_root *extent_root = root->fs_info->extent_root;
struct btrfs_path *path;
if (root == root->fs_info->extent_root) {
struct pending_extent_op *extent_op;
u64 num_bytes;
BUG_ON(owner_objectid >= BTRFS_MAX_LEVEL);
num_bytes = btrfs_level_size(root, (int)owner_objectid);
mutex_lock(&root->fs_info->extent_ins_mutex);
if (test_range_bit(&root->fs_info->extent_ins, bytenr,
bytenr + num_bytes - 1, EXTENT_WRITEBACK, 0)) {
u64 priv;
ret = get_state_private(&root->fs_info->extent_ins,
bytenr, &priv);
BUG_ON(ret);
extent_op = (struct pending_extent_op *)
(unsigned long)priv;
BUG_ON(extent_op->parent != orig_parent);
BUG_ON(extent_op->generation != orig_generation);
extent_op->parent = parent;
extent_op->generation = ref_generation;
} else {
extent_op = kmalloc(sizeof(*extent_op), GFP_NOFS);
BUG_ON(!extent_op);
extent_op->type = PENDING_BACKREF_UPDATE;
extent_op->bytenr = bytenr;
extent_op->num_bytes = num_bytes;
extent_op->parent = parent;
extent_op->orig_parent = orig_parent;
extent_op->generation = ref_generation;
extent_op->orig_generation = orig_generation;
extent_op->level = (int)owner_objectid;
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
INIT_LIST_HEAD(&extent_op->list);
extent_op->del = 0;
set_extent_bits(&root->fs_info->extent_ins,
bytenr, bytenr + num_bytes - 1,
EXTENT_WRITEBACK, GFP_NOFS);
set_state_private(&root->fs_info->extent_ins,
bytenr, (unsigned long)extent_op);
}
mutex_unlock(&root->fs_info->extent_ins_mutex);
return 0;
}
path = btrfs_alloc_path();
if (!path)
return -ENOMEM;
ret = lookup_extent_backref(trans, extent_root, path,
bytenr, orig_parent, orig_root,
orig_generation, owner_objectid, 1);
if (ret)
goto out;
ret = remove_extent_backref(trans, extent_root, path);
if (ret)
goto out;
ret = insert_extent_backref(trans, extent_root, path, bytenr,
parent, ref_root, ref_generation,
owner_objectid);
BUG_ON(ret);
finish_current_insert(trans, extent_root, 0);
del_pending_extents(trans, extent_root, 0);
out:
btrfs_free_path(path);
return ret;
}
int btrfs_update_extent_ref(struct btrfs_trans_handle *trans,
struct btrfs_root *root, u64 bytenr,
u64 orig_parent, u64 parent,
u64 ref_root, u64 ref_generation,
u64 owner_objectid)
{
int ret;
if (ref_root == BTRFS_TREE_LOG_OBJECTID &&
owner_objectid < BTRFS_FIRST_FREE_OBJECTID)
return 0;
ret = __btrfs_update_extent_ref(trans, root, bytenr, orig_parent,
parent, ref_root, ref_root,
ref_generation, ref_generation,
owner_objectid);
return ret;
}
static int __btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
struct btrfs_root *root, u64 bytenr,
u64 orig_parent, u64 parent,
u64 orig_root, u64 ref_root,
u64 orig_generation, u64 ref_generation,
u64 owner_objectid)
{
struct btrfs_path *path;
int ret;
struct btrfs_key key;
struct extent_buffer *l;
struct btrfs_extent_item *item;
u32 refs;
path = btrfs_alloc_path();
if (!path)
return -ENOMEM;
path->reada = 1;
key.objectid = bytenr;
key.type = BTRFS_EXTENT_ITEM_KEY;
key.offset = (u64)-1;
ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path,
0, 1);
if (ret < 0)
return ret;
BUG_ON(ret == 0 || path->slots[0] == 0);
path->slots[0]--;
l = path->nodes[0];
btrfs_item_key_to_cpu(l, &key, path->slots[0]);
if (key.objectid != bytenr) {
btrfs_print_leaf(root->fs_info->extent_root, path->nodes[0]);
printk("wanted %Lu found %Lu\n", bytenr, key.objectid);
BUG();
}
BUG_ON(key.type != BTRFS_EXTENT_ITEM_KEY);
item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item);
refs = btrfs_extent_refs(l, item);
btrfs_set_extent_refs(l, item, refs + 1);
btrfs_mark_buffer_dirty(path->nodes[0]);
btrfs_release_path(root->fs_info->extent_root, path);
path->reada = 1;
ret = insert_extent_backref(trans, root->fs_info->extent_root,
path, bytenr, parent,
ref_root, ref_generation,
owner_objectid);
BUG_ON(ret);
finish_current_insert(trans, root->fs_info->extent_root, 0);
del_pending_extents(trans, root->fs_info->extent_root, 0);
btrfs_free_path(path);
return 0;
}
int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
u64 bytenr, u64 num_bytes, u64 parent,
u64 ref_root, u64 ref_generation,
u64 owner_objectid)
{
int ret;
if (ref_root == BTRFS_TREE_LOG_OBJECTID &&
owner_objectid < BTRFS_FIRST_FREE_OBJECTID)
return 0;
ret = __btrfs_inc_extent_ref(trans, root, bytenr, 0, parent,
0, ref_root, 0, ref_generation,
owner_objectid);
return ret;
}
int btrfs_extent_post_op(struct btrfs_trans_handle *trans,
struct btrfs_root *root)
{
finish_current_insert(trans, root->fs_info->extent_root, 1);
del_pending_extents(trans, root->fs_info->extent_root, 1);
return 0;
}
int btrfs_lookup_extent_ref(struct btrfs_trans_handle *trans,
struct btrfs_root *root, u64 bytenr,
u64 num_bytes, u32 *refs)
{
struct btrfs_path *path;
int ret;
struct btrfs_key key;
struct extent_buffer *l;
struct btrfs_extent_item *item;
WARN_ON(num_bytes < root->sectorsize);
path = btrfs_alloc_path();
path->reada = 1;
key.objectid = bytenr;
key.offset = num_bytes;
btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path,
0, 0);
if (ret < 0)
goto out;
if (ret != 0) {
btrfs_print_leaf(root, path->nodes[0]);
printk("failed to find block number %Lu\n", bytenr);
BUG();
}
l = path->nodes[0];
item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item);
*refs = btrfs_extent_refs(l, item);
out:
btrfs_free_path(path);
return 0;
}
int btrfs_cross_ref_exist(struct btrfs_trans_handle *trans,
struct btrfs_root *root, u64 bytenr)
{
struct btrfs_root *extent_root = root->fs_info->extent_root;
struct btrfs_path *path;
struct extent_buffer *leaf;
struct btrfs_extent_ref *ref_item;
struct btrfs_key key;
struct btrfs_key found_key;
u64 ref_root;
u64 last_snapshot;
u32 nritems;
int ret;
key.objectid = bytenr;
key.offset = (u64)-1;
key.type = BTRFS_EXTENT_ITEM_KEY;
path = btrfs_alloc_path();
ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0);
if (ret < 0)
goto out;
BUG_ON(ret == 0);
ret = -ENOENT;
if (path->slots[0] == 0)
goto out;
path->slots[0]--;
leaf = path->nodes[0];
btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
if (found_key.objectid != bytenr ||
found_key.type != BTRFS_EXTENT_ITEM_KEY)
goto out;
last_snapshot = btrfs_root_last_snapshot(&root->root_item);
while (1) {
leaf = path->nodes[0];
nritems = btrfs_header_nritems(leaf);
if (path->slots[0] >= nritems) {
ret = btrfs_next_leaf(extent_root, path);
if (ret < 0)
goto out;
if (ret == 0)
continue;
break;
}
btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
if (found_key.objectid != bytenr)
break;
if (found_key.type != BTRFS_EXTENT_REF_KEY) {
path->slots[0]++;
continue;
}
ref_item = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_extent_ref);
ref_root = btrfs_ref_root(leaf, ref_item);
if (ref_root != root->root_key.objectid &&
ref_root != BTRFS_TREE_LOG_OBJECTID) {
ret = 1;
goto out;
}
if (btrfs_ref_generation(leaf, ref_item) <= last_snapshot) {
ret = 1;
goto out;
}
path->slots[0]++;
}
ret = 0;
out:
btrfs_free_path(path);
return ret;
}
int btrfs_cache_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
struct extent_buffer *buf, u32 nr_extents)
{
struct btrfs_key key;
struct btrfs_file_extent_item *fi;
u64 root_gen;
u32 nritems;
int i;
int level;
int ret = 0;
int shared = 0;
if (!root->ref_cows)
return 0;
if (root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID) {
shared = 0;
root_gen = root->root_key.offset;
} else {
shared = 1;
root_gen = trans->transid - 1;
}
level = btrfs_header_level(buf);
nritems = btrfs_header_nritems(buf);
if (level == 0) {
struct btrfs_leaf_ref *ref;
struct btrfs_extent_info *info;
ref = btrfs_alloc_leaf_ref(root, nr_extents);
if (!ref) {
ret = -ENOMEM;
goto out;
}
ref->root_gen = root_gen;
ref->bytenr = buf->start;
ref->owner = btrfs_header_owner(buf);
ref->generation = btrfs_header_generation(buf);
ref->nritems = nr_extents;
info = ref->extents;
for (i = 0; nr_extents > 0 && i < nritems; i++) {
u64 disk_bytenr;
btrfs_item_key_to_cpu(buf, &key, i);
if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
continue;
fi = btrfs_item_ptr(buf, i,
struct btrfs_file_extent_item);
if (btrfs_file_extent_type(buf, fi) ==
BTRFS_FILE_EXTENT_INLINE)
continue;
disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
if (disk_bytenr == 0)
continue;
info->bytenr = disk_bytenr;
info->num_bytes =
btrfs_file_extent_disk_num_bytes(buf, fi);
info->objectid = key.objectid;
info->offset = key.offset;
info++;
}
ret = btrfs_add_leaf_ref(root, ref, shared);
if (ret == -EEXIST && shared) {
struct btrfs_leaf_ref *old;
old = btrfs_lookup_leaf_ref(root, ref->bytenr);
BUG_ON(!old);
btrfs_remove_leaf_ref(root, old);
btrfs_free_leaf_ref(root, old);
ret = btrfs_add_leaf_ref(root, ref, shared);
}
WARN_ON(ret);
btrfs_free_leaf_ref(root, ref);
}
out:
return ret;
}
int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
struct extent_buffer *orig_buf, struct extent_buffer *buf,
u32 *nr_extents)
{
u64 bytenr;
u64 ref_root;
u64 orig_root;
u64 ref_generation;
u64 orig_generation;
u32 nritems;
u32 nr_file_extents = 0;
struct btrfs_key key;
struct btrfs_file_extent_item *fi;
int i;
int level;
int ret = 0;
int faili = 0;
int (*process_func)(struct btrfs_trans_handle *, struct btrfs_root *,
u64, u64, u64, u64, u64, u64, u64, u64);
ref_root = btrfs_header_owner(buf);
ref_generation = btrfs_header_generation(buf);
orig_root = btrfs_header_owner(orig_buf);
orig_generation = btrfs_header_generation(orig_buf);
nritems = btrfs_header_nritems(buf);
level = btrfs_header_level(buf);
if (root->ref_cows) {
process_func = __btrfs_inc_extent_ref;
} else {
if (level == 0 &&
root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID)
goto out;
if (level != 0 &&
root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID)
goto out;
process_func = __btrfs_update_extent_ref;
}
for (i = 0; i < nritems; i++) {
cond_resched();
if (level == 0) {
btrfs_item_key_to_cpu(buf, &key, i);
if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
continue;
fi = btrfs_item_ptr(buf, i,
struct btrfs_file_extent_item);
if (btrfs_file_extent_type(buf, fi) ==
BTRFS_FILE_EXTENT_INLINE)
continue;
bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
if (bytenr == 0)
continue;
nr_file_extents++;
ret = process_func(trans, root, bytenr,
orig_buf->start, buf->start,
orig_root, ref_root,
orig_generation, ref_generation,
key.objectid);
if (ret) {
faili = i;
WARN_ON(1);
goto fail;
}
} else {
bytenr = btrfs_node_blockptr(buf, i);
ret = process_func(trans, root, bytenr,
orig_buf->start, buf->start,
orig_root, ref_root,
orig_generation, ref_generation,
level - 1);
if (ret) {
faili = i;
WARN_ON(1);
goto fail;
}
}
}
out:
if (nr_extents) {
if (level == 0)
*nr_extents = nr_file_extents;
else
*nr_extents = nritems;
}
return 0;
fail:
WARN_ON(1);
return ret;
}
int btrfs_update_ref(struct btrfs_trans_handle *trans,
struct btrfs_root *root, struct extent_buffer *orig_buf,
struct extent_buffer *buf, int start_slot, int nr)
{
u64 bytenr;
u64 ref_root;
u64 orig_root;
u64 ref_generation;
u64 orig_generation;
struct btrfs_key key;
struct btrfs_file_extent_item *fi;
int i;
int ret;
int slot;
int level;
BUG_ON(start_slot < 0);
BUG_ON(start_slot + nr > btrfs_header_nritems(buf));
ref_root = btrfs_header_owner(buf);
ref_generation = btrfs_header_generation(buf);
orig_root = btrfs_header_owner(orig_buf);
orig_generation = btrfs_header_generation(orig_buf);
level = btrfs_header_level(buf);
if (!root->ref_cows) {
if (level == 0 &&
root->root_key.objectid != BTRFS_TREE_LOG_OBJECTID)
return 0;
if (level != 0 &&
root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID)
return 0;
}
for (i = 0, slot = start_slot; i < nr; i++, slot++) {
cond_resched();
if (level == 0) {
btrfs_item_key_to_cpu(buf, &key, slot);
if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
continue;
fi = btrfs_item_ptr(buf, slot,
struct btrfs_file_extent_item);
if (btrfs_file_extent_type(buf, fi) ==
BTRFS_FILE_EXTENT_INLINE)
continue;
bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
if (bytenr == 0)
continue;
ret = __btrfs_update_extent_ref(trans, root, bytenr,
orig_buf->start, buf->start,
orig_root, ref_root,
orig_generation, ref_generation,
key.objectid);
if (ret)
goto fail;
} else {
bytenr = btrfs_node_blockptr(buf, slot);
ret = __btrfs_update_extent_ref(trans, root, bytenr,
orig_buf->start, buf->start,
orig_root, ref_root,
orig_generation, ref_generation,
level - 1);
if (ret)
goto fail;
}
}
return 0;
fail:
WARN_ON(1);
return -1;
}
static int write_one_cache_group(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct btrfs_path *path,
struct btrfs_block_group_cache *cache)
{
int ret;
int pending_ret;
struct btrfs_root *extent_root = root->fs_info->extent_root;
unsigned long bi;
struct extent_buffer *leaf;
ret = btrfs_search_slot(trans, extent_root, &cache->key, path, 0, 1);
if (ret < 0)
goto fail;
BUG_ON(ret);
leaf = path->nodes[0];
bi = btrfs_item_ptr_offset(leaf, path->slots[0]);
write_extent_buffer(leaf, &cache->item, bi, sizeof(cache->item));
btrfs_mark_buffer_dirty(leaf);
btrfs_release_path(extent_root, path);
fail:
finish_current_insert(trans, extent_root, 0);
pending_ret = del_pending_extents(trans, extent_root, 0);
if (ret)
return ret;
if (pending_ret)
return pending_ret;
return 0;
}
int btrfs_write_dirty_block_groups(struct btrfs_trans_handle *trans,
struct btrfs_root *root)
{
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
struct btrfs_block_group_cache *cache, *entry;
struct rb_node *n;
int err = 0;
int werr = 0;
struct btrfs_path *path;
u64 last = 0;
path = btrfs_alloc_path();
if (!path)
return -ENOMEM;
while(1) {
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
cache = NULL;
spin_lock(&root->fs_info->block_group_cache_lock);
for (n = rb_first(&root->fs_info->block_group_cache_tree);
n; n = rb_next(n)) {
entry = rb_entry(n, struct btrfs_block_group_cache,
cache_node);
if (entry->dirty) {
cache = entry;
break;
}
}
spin_unlock(&root->fs_info->block_group_cache_lock);
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
if (!cache)
break;
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
cache->dirty = 0;
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
last += cache->key.offset;
err = write_one_cache_group(trans, root,
path, cache);
/*
* if we fail to write the cache group, we want
* to keep it marked dirty in hopes that a later
* write will work
*/
if (err) {
werr = err;
continue;
}
}
btrfs_free_path(path);
return werr;
}
static int update_space_info(struct btrfs_fs_info *info, u64 flags,
u64 total_bytes, u64 bytes_used,
struct btrfs_space_info **space_info)
{
struct btrfs_space_info *found;
found = __find_space_info(info, flags);
if (found) {
spin_lock(&found->lock);
found->total_bytes += total_bytes;
found->bytes_used += bytes_used;
found->full = 0;
spin_unlock(&found->lock);
*space_info = found;
return 0;
}
found = kzalloc(sizeof(*found), GFP_NOFS);
if (!found)
return -ENOMEM;
list_add(&found->list, &info->space_info);
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
INIT_LIST_HEAD(&found->block_groups);
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
init_rwsem(&found->groups_sem);
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
spin_lock_init(&found->lock);
found->flags = flags;
found->total_bytes = total_bytes;
found->bytes_used = bytes_used;
found->bytes_pinned = 0;
found->bytes_reserved = 0;
found->bytes_readonly = 0;
found->full = 0;
found->force_alloc = 0;
*space_info = found;
return 0;
}
static void set_avail_alloc_bits(struct btrfs_fs_info *fs_info, u64 flags)
{
u64 extra_flags = flags & (BTRFS_BLOCK_GROUP_RAID0 |
BTRFS_BLOCK_GROUP_RAID1 |
BTRFS_BLOCK_GROUP_RAID10 |
BTRFS_BLOCK_GROUP_DUP);
if (extra_flags) {
if (flags & BTRFS_BLOCK_GROUP_DATA)
fs_info->avail_data_alloc_bits |= extra_flags;
if (flags & BTRFS_BLOCK_GROUP_METADATA)
fs_info->avail_metadata_alloc_bits |= extra_flags;
if (flags & BTRFS_BLOCK_GROUP_SYSTEM)
fs_info->avail_system_alloc_bits |= extra_flags;
}
}
static void set_block_group_readonly(struct btrfs_block_group_cache *cache)
{
spin_lock(&cache->space_info->lock);
spin_lock(&cache->lock);
if (!cache->ro) {
cache->space_info->bytes_readonly += cache->key.offset -
btrfs_block_group_used(&cache->item);
cache->ro = 1;
}
spin_unlock(&cache->lock);
spin_unlock(&cache->space_info->lock);
}
u64 btrfs_reduce_alloc_profile(struct btrfs_root *root, u64 flags)
{
u64 num_devices = root->fs_info->fs_devices->rw_devices;
if (num_devices == 1)
flags &= ~(BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID0);
if (num_devices < 4)
flags &= ~BTRFS_BLOCK_GROUP_RAID10;
if ((flags & BTRFS_BLOCK_GROUP_DUP) &&
(flags & (BTRFS_BLOCK_GROUP_RAID1 |
BTRFS_BLOCK_GROUP_RAID10))) {
flags &= ~BTRFS_BLOCK_GROUP_DUP;
}
if ((flags & BTRFS_BLOCK_GROUP_RAID1) &&
(flags & BTRFS_BLOCK_GROUP_RAID10)) {
flags &= ~BTRFS_BLOCK_GROUP_RAID1;
}
if ((flags & BTRFS_BLOCK_GROUP_RAID0) &&
((flags & BTRFS_BLOCK_GROUP_RAID1) |
(flags & BTRFS_BLOCK_GROUP_RAID10) |
(flags & BTRFS_BLOCK_GROUP_DUP)))
flags &= ~BTRFS_BLOCK_GROUP_RAID0;
return flags;
}
static int do_chunk_alloc(struct btrfs_trans_handle *trans,
struct btrfs_root *extent_root, u64 alloc_bytes,
u64 flags, int force)
{
struct btrfs_space_info *space_info;
u64 thresh;
int ret = 0;
mutex_lock(&extent_root->fs_info->chunk_mutex);
flags = btrfs_reduce_alloc_profile(extent_root, flags);
space_info = __find_space_info(extent_root->fs_info, flags);
if (!space_info) {
ret = update_space_info(extent_root->fs_info, flags,
0, 0, &space_info);
BUG_ON(ret);
}
BUG_ON(!space_info);
spin_lock(&space_info->lock);
if (space_info->force_alloc) {
force = 1;
space_info->force_alloc = 0;
}
if (space_info->full) {
spin_unlock(&space_info->lock);
goto out;
}
thresh = space_info->total_bytes - space_info->bytes_readonly;
thresh = div_factor(thresh, 6);
if (!force &&
(space_info->bytes_used + space_info->bytes_pinned +
space_info->bytes_reserved + alloc_bytes) < thresh) {
spin_unlock(&space_info->lock);
goto out;
}
spin_unlock(&space_info->lock);
ret = btrfs_alloc_chunk(trans, extent_root, flags);
if (ret) {
printk("space info full %Lu\n", flags);
space_info->full = 1;
}
out:
mutex_unlock(&extent_root->fs_info->chunk_mutex);
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
return ret;
}
static int update_block_group(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
u64 bytenr, u64 num_bytes, int alloc,
int mark_free)
{
struct btrfs_block_group_cache *cache;
struct btrfs_fs_info *info = root->fs_info;
u64 total = num_bytes;
u64 old_val;
u64 byte_in_group;
while(total) {
cache = btrfs_lookup_block_group(info, bytenr);
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
if (!cache)
return -1;
byte_in_group = bytenr - cache->key.objectid;
WARN_ON(byte_in_group > cache->key.offset);
spin_lock(&cache->space_info->lock);
spin_lock(&cache->lock);
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
cache->dirty = 1;
old_val = btrfs_block_group_used(&cache->item);
num_bytes = min(total, cache->key.offset - byte_in_group);
if (alloc) {
old_val += num_bytes;
cache->space_info->bytes_used += num_bytes;
if (cache->ro) {
cache->space_info->bytes_readonly -= num_bytes;
WARN_ON(1);
}
btrfs_set_block_group_used(&cache->item, old_val);
spin_unlock(&cache->lock);
spin_unlock(&cache->space_info->lock);
} else {
old_val -= num_bytes;
cache->space_info->bytes_used -= num_bytes;
if (cache->ro)
cache->space_info->bytes_readonly += num_bytes;
btrfs_set_block_group_used(&cache->item, old_val);
spin_unlock(&cache->lock);
spin_unlock(&cache->space_info->lock);
if (mark_free) {
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
int ret;
ret = btrfs_add_free_space(cache, bytenr,
num_bytes);
if (ret)
return -1;
}
}
total -= num_bytes;
bytenr += num_bytes;
}
return 0;
}
static u64 first_logical_byte(struct btrfs_root *root, u64 search_start)
{
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
struct btrfs_block_group_cache *cache;
cache = btrfs_lookup_first_block_group(root->fs_info, search_start);
if (!cache)
return 0;
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
return cache->key.objectid;
}
int btrfs_update_pinned_extents(struct btrfs_root *root,
u64 bytenr, u64 num, int pin)
{
u64 len;
struct btrfs_block_group_cache *cache;
struct btrfs_fs_info *fs_info = root->fs_info;
WARN_ON(!mutex_is_locked(&root->fs_info->pinned_mutex));
if (pin) {
set_extent_dirty(&fs_info->pinned_extents,
bytenr, bytenr + num - 1, GFP_NOFS);
} else {
clear_extent_dirty(&fs_info->pinned_extents,
bytenr, bytenr + num - 1, GFP_NOFS);
}
while (num > 0) {
cache = btrfs_lookup_block_group(fs_info, bytenr);
BUG_ON(!cache);
len = min(num, cache->key.offset -
(bytenr - cache->key.objectid));
if (pin) {
spin_lock(&cache->space_info->lock);
spin_lock(&cache->lock);
cache->pinned += len;
cache->space_info->bytes_pinned += len;
spin_unlock(&cache->lock);
spin_unlock(&cache->space_info->lock);
fs_info->total_pinned += len;
} else {
spin_lock(&cache->space_info->lock);
spin_lock(&cache->lock);
cache->pinned -= len;
cache->space_info->bytes_pinned -= len;
spin_unlock(&cache->lock);
spin_unlock(&cache->space_info->lock);
fs_info->total_pinned -= len;
if (cache->cached)
btrfs_add_free_space(cache, bytenr, len);
}
bytenr += len;
num -= len;
}
return 0;
}
static int update_reserved_extents(struct btrfs_root *root,
u64 bytenr, u64 num, int reserve)
{
u64 len;
struct btrfs_block_group_cache *cache;
struct btrfs_fs_info *fs_info = root->fs_info;
while (num > 0) {
cache = btrfs_lookup_block_group(fs_info, bytenr);
BUG_ON(!cache);
len = min(num, cache->key.offset -
(bytenr - cache->key.objectid));
spin_lock(&cache->space_info->lock);
spin_lock(&cache->lock);
if (reserve) {
cache->reserved += len;
cache->space_info->bytes_reserved += len;
} else {
cache->reserved -= len;
cache->space_info->bytes_reserved -= len;
}
spin_unlock(&cache->lock);
spin_unlock(&cache->space_info->lock);
bytenr += len;
num -= len;
}
return 0;
}
int btrfs_copy_pinned(struct btrfs_root *root, struct extent_io_tree *copy)
{
u64 last = 0;
u64 start;
u64 end;
struct extent_io_tree *pinned_extents = &root->fs_info->pinned_extents;
int ret;
mutex_lock(&root->fs_info->pinned_mutex);
while(1) {
ret = find_first_extent_bit(pinned_extents, last,
&start, &end, EXTENT_DIRTY);
if (ret)
break;
set_extent_dirty(copy, start, end, GFP_NOFS);
last = end + 1;
}
mutex_unlock(&root->fs_info->pinned_mutex);
return 0;
}
int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct extent_io_tree *unpin)
{
u64 start;
u64 end;
int ret;
mutex_lock(&root->fs_info->pinned_mutex);
while(1) {
ret = find_first_extent_bit(unpin, 0, &start, &end,
EXTENT_DIRTY);
if (ret)
break;
btrfs_update_pinned_extents(root, start, end + 1 - start, 0);
clear_extent_dirty(unpin, start, end, GFP_NOFS);
if (need_resched()) {
mutex_unlock(&root->fs_info->pinned_mutex);
cond_resched();
mutex_lock(&root->fs_info->pinned_mutex);
}
}
mutex_unlock(&root->fs_info->pinned_mutex);
return 0;
}
static int finish_current_insert(struct btrfs_trans_handle *trans,
struct btrfs_root *extent_root, int all)
{
u64 start;
u64 end;
u64 priv;
u64 search = 0;
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
u64 skipped = 0;
struct btrfs_fs_info *info = extent_root->fs_info;
struct btrfs_path *path;
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
struct pending_extent_op *extent_op, *tmp;
struct list_head insert_list, update_list;
int ret;
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
int num_inserts = 0, max_inserts;
path = btrfs_alloc_path();
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
INIT_LIST_HEAD(&insert_list);
INIT_LIST_HEAD(&update_list);
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
max_inserts = extent_root->leafsize /
(2 * sizeof(struct btrfs_key) + 2 * sizeof(struct btrfs_item) +
sizeof(struct btrfs_extent_ref) +
sizeof(struct btrfs_extent_item));
again:
mutex_lock(&info->extent_ins_mutex);
while (1) {
ret = find_first_extent_bit(&info->extent_ins, search, &start,
&end, EXTENT_WRITEBACK);
if (ret) {
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
if (skipped && all && !num_inserts) {
skipped = 0;
search = 0;
continue;
}
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
mutex_unlock(&info->extent_ins_mutex);
break;
}
ret = try_lock_extent(&info->extent_ins, start, end, GFP_NOFS);
if (!ret) {
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
skipped = 1;
search = end + 1;
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
if (need_resched()) {
mutex_unlock(&info->extent_ins_mutex);
cond_resched();
mutex_lock(&info->extent_ins_mutex);
}
continue;
}
ret = get_state_private(&info->extent_ins, start, &priv);
BUG_ON(ret);
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
extent_op = (struct pending_extent_op *)(unsigned long) priv;
if (extent_op->type == PENDING_EXTENT_INSERT) {
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
num_inserts++;
list_add_tail(&extent_op->list, &insert_list);
search = end + 1;
if (num_inserts == max_inserts) {
mutex_unlock(&info->extent_ins_mutex);
break;
}
} else if (extent_op->type == PENDING_BACKREF_UPDATE) {
list_add_tail(&extent_op->list, &update_list);
search = end + 1;
} else {
BUG();
}
}
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
/*
* process the update list, clear the writeback bit for it, and if
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
* somebody marked this thing for deletion then just unlock it and be
* done, the free_extents will handle it
*/
mutex_lock(&info->extent_ins_mutex);
list_for_each_entry_safe(extent_op, tmp, &update_list, list) {
clear_extent_bits(&info->extent_ins, extent_op->bytenr,
extent_op->bytenr + extent_op->num_bytes - 1,
EXTENT_WRITEBACK, GFP_NOFS);
if (extent_op->del) {
list_del_init(&extent_op->list);
unlock_extent(&info->extent_ins, extent_op->bytenr,
extent_op->bytenr + extent_op->num_bytes
- 1, GFP_NOFS);
kfree(extent_op);
}
}
mutex_unlock(&info->extent_ins_mutex);
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
/*
* still have things left on the update list, go ahead an update
* everything
*/
if (!list_empty(&update_list)) {
ret = update_backrefs(trans, extent_root, path, &update_list);
BUG_ON(ret);
}
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
/*
* if no inserts need to be done, but we skipped some extents and we
* need to make sure everything is cleaned then reset everything and
* go back to the beginning
*/
if (!num_inserts && all && skipped) {
search = 0;
skipped = 0;
INIT_LIST_HEAD(&update_list);
INIT_LIST_HEAD(&insert_list);
goto again;
} else if (!num_inserts) {
goto out;
}
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
/*
* process the insert extents list. Again if we are deleting this
* extent, then just unlock it, pin down the bytes if need be, and be
* done with it. Saves us from having to actually insert the extent
* into the tree and then subsequently come along and delete it
*/
mutex_lock(&info->extent_ins_mutex);
list_for_each_entry_safe(extent_op, tmp, &insert_list, list) {
clear_extent_bits(&info->extent_ins, extent_op->bytenr,
extent_op->bytenr + extent_op->num_bytes - 1,
EXTENT_WRITEBACK, GFP_NOFS);
if (extent_op->del) {
list_del_init(&extent_op->list);
unlock_extent(&info->extent_ins, extent_op->bytenr,
extent_op->bytenr + extent_op->num_bytes
- 1, GFP_NOFS);
mutex_lock(&extent_root->fs_info->pinned_mutex);
ret = pin_down_bytes(trans, extent_root,
extent_op->bytenr,
extent_op->num_bytes, 0);
mutex_unlock(&extent_root->fs_info->pinned_mutex);
ret = update_block_group(trans, extent_root,
extent_op->bytenr,
extent_op->num_bytes,
0, ret > 0);
BUG_ON(ret);
kfree(extent_op);
num_inserts--;
}
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
}
mutex_unlock(&info->extent_ins_mutex);
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
ret = insert_extents(trans, extent_root, path, &insert_list,
num_inserts);
BUG_ON(ret);
/*
* if we broke out of the loop in order to insert stuff because we hit
* the maximum number of inserts at a time we can handle, then loop
* back and pick up where we left off
*/
if (num_inserts == max_inserts) {
INIT_LIST_HEAD(&insert_list);
INIT_LIST_HEAD(&update_list);
num_inserts = 0;
goto again;
}
/*
* again, if we need to make absolutely sure there are no more pending
* extent operations left and we know that we skipped some, go back to
* the beginning and do it all again
*/
if (all && skipped) {
INIT_LIST_HEAD(&insert_list);
INIT_LIST_HEAD(&update_list);
search = 0;
skipped = 0;
num_inserts = 0;
goto again;
}
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
out:
btrfs_free_path(path);
return 0;
}
static int pin_down_bytes(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
u64 bytenr, u64 num_bytes, int is_data)
{
int err = 0;
struct extent_buffer *buf;
if (is_data)
goto pinit;
buf = btrfs_find_tree_block(root, bytenr, num_bytes);
if (!buf)
goto pinit;
/* we can reuse a block if it hasn't been written
* and it is from this transaction. We can't
* reuse anything from the tree log root because
* it has tiny sub-transactions.
*/
if (btrfs_buffer_uptodate(buf, 0) &&
btrfs_try_tree_lock(buf)) {
u64 header_owner = btrfs_header_owner(buf);
u64 header_transid = btrfs_header_generation(buf);
if (header_owner != BTRFS_TREE_LOG_OBJECTID &&
2008-09-26 14:09:34 +00:00
header_owner != BTRFS_TREE_RELOC_OBJECTID &&
header_transid == trans->transid &&
!btrfs_header_flag(buf, BTRFS_HEADER_FLAG_WRITTEN)) {
clean_tree_block(NULL, root, buf);
btrfs_tree_unlock(buf);
free_extent_buffer(buf);
return 1;
}
btrfs_tree_unlock(buf);
}
free_extent_buffer(buf);
pinit:
btrfs_update_pinned_extents(root, bytenr, num_bytes, 1);
BUG_ON(err < 0);
return 0;
}
/*
* remove an extent from the root, returns 0 on success
*/
static int __free_extent(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
u64 bytenr, u64 num_bytes, u64 parent,
u64 root_objectid, u64 ref_generation,
u64 owner_objectid, int pin, int mark_free)
{
struct btrfs_path *path;
struct btrfs_key key;
struct btrfs_fs_info *info = root->fs_info;
struct btrfs_root *extent_root = info->extent_root;
struct extent_buffer *leaf;
int ret;
int extent_slot = 0;
int found_extent = 0;
int num_to_del = 1;
struct btrfs_extent_item *ei;
u32 refs;
key.objectid = bytenr;
btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
key.offset = num_bytes;
path = btrfs_alloc_path();
if (!path)
return -ENOMEM;
path->reada = 1;
ret = lookup_extent_backref(trans, extent_root, path,
bytenr, parent, root_objectid,
ref_generation, owner_objectid, 1);
if (ret == 0) {
struct btrfs_key found_key;
extent_slot = path->slots[0];
while(extent_slot > 0) {
extent_slot--;
btrfs_item_key_to_cpu(path->nodes[0], &found_key,
extent_slot);
if (found_key.objectid != bytenr)
break;
if (found_key.type == BTRFS_EXTENT_ITEM_KEY &&
found_key.offset == num_bytes) {
found_extent = 1;
break;
}
if (path->slots[0] - extent_slot > 5)
break;
}
if (!found_extent) {
ret = remove_extent_backref(trans, extent_root, path);
BUG_ON(ret);
btrfs_release_path(extent_root, path);
ret = btrfs_search_slot(trans, extent_root,
&key, path, -1, 1);
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
if (ret) {
printk(KERN_ERR "umm, got %d back from search"
", was looking for %Lu\n", ret,
bytenr);
btrfs_print_leaf(extent_root, path->nodes[0]);
}
BUG_ON(ret);
extent_slot = path->slots[0];
}
} else {
btrfs_print_leaf(extent_root, path->nodes[0]);
WARN_ON(1);
printk("Unable to find ref byte nr %Lu root %Lu "
"gen %Lu owner %Lu\n", bytenr,
root_objectid, ref_generation, owner_objectid);
}
leaf = path->nodes[0];
ei = btrfs_item_ptr(leaf, extent_slot,
struct btrfs_extent_item);
refs = btrfs_extent_refs(leaf, ei);
BUG_ON(refs == 0);
refs -= 1;
btrfs_set_extent_refs(leaf, ei, refs);
btrfs_mark_buffer_dirty(leaf);
if (refs == 0 && found_extent && path->slots[0] == extent_slot + 1) {
struct btrfs_extent_ref *ref;
ref = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_extent_ref);
BUG_ON(btrfs_ref_num_refs(leaf, ref) != 1);
/* if the back ref and the extent are next to each other
* they get deleted below in one shot
*/
path->slots[0] = extent_slot;
num_to_del = 2;
} else if (found_extent) {
/* otherwise delete the extent back ref */
ret = remove_extent_backref(trans, extent_root, path);
BUG_ON(ret);
/* if refs are 0, we need to setup the path for deletion */
if (refs == 0) {
btrfs_release_path(extent_root, path);
ret = btrfs_search_slot(trans, extent_root, &key, path,
-1, 1);
BUG_ON(ret);
}
}
if (refs == 0) {
u64 super_used;
u64 root_used;
#ifdef BIO_RW_DISCARD
u64 map_length = num_bytes;
struct btrfs_multi_bio *multi = NULL;
#endif
if (pin) {
mutex_lock(&root->fs_info->pinned_mutex);
ret = pin_down_bytes(trans, root, bytenr, num_bytes,
owner_objectid >= BTRFS_FIRST_FREE_OBJECTID);
mutex_unlock(&root->fs_info->pinned_mutex);
if (ret > 0)
mark_free = 1;
BUG_ON(ret < 0);
}
/* block accounting for super block */
spin_lock_irq(&info->delalloc_lock);
super_used = btrfs_super_bytes_used(&info->super_copy);
btrfs_set_super_bytes_used(&info->super_copy,
super_used - num_bytes);
spin_unlock_irq(&info->delalloc_lock);
/* block accounting for root item */
root_used = btrfs_root_used(&root->root_item);
btrfs_set_root_used(&root->root_item,
root_used - num_bytes);
ret = btrfs_del_items(trans, extent_root, path, path->slots[0],
num_to_del);
BUG_ON(ret);
btrfs_release_path(extent_root, path);
ret = update_block_group(trans, root, bytenr, num_bytes, 0,
mark_free);
BUG_ON(ret);
#ifdef BIO_RW_DISCARD
/* Tell the block device(s) that the sectors can be discarded */
ret = btrfs_map_block(&root->fs_info->mapping_tree, READ,
bytenr, &map_length, &multi, 0);
if (!ret) {
struct btrfs_bio_stripe *stripe = multi->stripes;
int i;
if (map_length > num_bytes)
map_length = num_bytes;
for (i = 0; i < multi->num_stripes; i++, stripe++) {
btrfs_issue_discard(stripe->dev->bdev,
stripe->physical,
map_length);
}
kfree(multi);
}
#endif
}
btrfs_free_path(path);
finish_current_insert(trans, extent_root, 0);
return ret;
}
/*
* find all the blocks marked as pending in the radix tree and remove
* them from the extent map
*/
static int del_pending_extents(struct btrfs_trans_handle *trans, struct
btrfs_root *extent_root, int all)
{
int ret;
int err = 0;
u64 start;
u64 end;
u64 priv;
u64 search = 0;
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
int nr = 0, skipped = 0;
struct extent_io_tree *pending_del;
struct extent_io_tree *extent_ins;
struct pending_extent_op *extent_op;
struct btrfs_fs_info *info = extent_root->fs_info;
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
struct list_head delete_list;
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
INIT_LIST_HEAD(&delete_list);
extent_ins = &extent_root->fs_info->extent_ins;
pending_del = &extent_root->fs_info->pending_del;
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
again:
mutex_lock(&info->extent_ins_mutex);
while(1) {
ret = find_first_extent_bit(pending_del, search, &start, &end,
EXTENT_WRITEBACK);
if (ret) {
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
if (all && skipped && !nr) {
search = 0;
continue;
}
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
mutex_unlock(&info->extent_ins_mutex);
break;
}
ret = try_lock_extent(extent_ins, start, end, GFP_NOFS);
if (!ret) {
search = end+1;
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
skipped = 1;
if (need_resched()) {
mutex_unlock(&info->extent_ins_mutex);
cond_resched();
mutex_lock(&info->extent_ins_mutex);
}
continue;
}
BUG_ON(ret < 0);
ret = get_state_private(pending_del, start, &priv);
BUG_ON(ret);
extent_op = (struct pending_extent_op *)(unsigned long)priv;
clear_extent_bits(pending_del, start, end, EXTENT_WRITEBACK,
GFP_NOFS);
if (!test_range_bit(extent_ins, start, end,
EXTENT_WRITEBACK, 0)) {
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
list_add_tail(&extent_op->list, &delete_list);
nr++;
} else {
kfree(extent_op);
ret = get_state_private(&info->extent_ins, start,
&priv);
BUG_ON(ret);
extent_op = (struct pending_extent_op *)
(unsigned long)priv;
clear_extent_bits(&info->extent_ins, start, end,
EXTENT_WRITEBACK, GFP_NOFS);
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
if (extent_op->type == PENDING_BACKREF_UPDATE) {
list_add_tail(&extent_op->list, &delete_list);
search = end + 1;
nr++;
continue;
}
mutex_lock(&extent_root->fs_info->pinned_mutex);
ret = pin_down_bytes(trans, extent_root, start,
end + 1 - start, 0);
mutex_unlock(&extent_root->fs_info->pinned_mutex);
ret = update_block_group(trans, extent_root, start,
end + 1 - start, 0, ret > 0);
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
unlock_extent(extent_ins, start, end, GFP_NOFS);
BUG_ON(ret);
kfree(extent_op);
}
if (ret)
err = ret;
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
search = end + 1;
if (need_resched()) {
mutex_unlock(&info->extent_ins_mutex);
cond_resched();
mutex_lock(&info->extent_ins_mutex);
}
}
if (nr) {
ret = free_extents(trans, extent_root, &delete_list);
BUG_ON(ret);
}
if (all && skipped) {
INIT_LIST_HEAD(&delete_list);
search = 0;
nr = 0;
goto again;
}
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
return err;
}
/*
* remove an extent from the root, returns 0 on success
*/
static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
u64 bytenr, u64 num_bytes, u64 parent,
u64 root_objectid, u64 ref_generation,
u64 owner_objectid, int pin)
{
struct btrfs_root *extent_root = root->fs_info->extent_root;
int pending_ret;
int ret;
WARN_ON(num_bytes < root->sectorsize);
if (root == extent_root) {
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
struct pending_extent_op *extent_op = NULL;
mutex_lock(&root->fs_info->extent_ins_mutex);
if (test_range_bit(&root->fs_info->extent_ins, bytenr,
bytenr + num_bytes - 1, EXTENT_WRITEBACK, 0)) {
u64 priv;
ret = get_state_private(&root->fs_info->extent_ins,
bytenr, &priv);
BUG_ON(ret);
extent_op = (struct pending_extent_op *)
(unsigned long)priv;
extent_op->del = 1;
if (extent_op->type == PENDING_EXTENT_INSERT) {
mutex_unlock(&root->fs_info->extent_ins_mutex);
return 0;
}
}
if (extent_op) {
ref_generation = extent_op->orig_generation;
parent = extent_op->orig_parent;
}
extent_op = kmalloc(sizeof(*extent_op), GFP_NOFS);
BUG_ON(!extent_op);
extent_op->type = PENDING_EXTENT_DELETE;
extent_op->bytenr = bytenr;
extent_op->num_bytes = num_bytes;
extent_op->parent = parent;
extent_op->orig_parent = parent;
extent_op->generation = ref_generation;
extent_op->orig_generation = ref_generation;
extent_op->level = (int)owner_objectid;
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
INIT_LIST_HEAD(&extent_op->list);
extent_op->del = 0;
set_extent_bits(&root->fs_info->pending_del,
bytenr, bytenr + num_bytes - 1,
EXTENT_WRITEBACK, GFP_NOFS);
set_state_private(&root->fs_info->pending_del,
bytenr, (unsigned long)extent_op);
mutex_unlock(&root->fs_info->extent_ins_mutex);
return 0;
}
/* if metadata always pin */
if (owner_objectid < BTRFS_FIRST_FREE_OBJECTID) {
if (root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID) {
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
struct btrfs_block_group_cache *cache;
/* btrfs_free_reserved_extent */
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
cache = btrfs_lookup_block_group(root->fs_info, bytenr);
BUG_ON(!cache);
btrfs_add_free_space(cache, bytenr, num_bytes);
update_reserved_extents(root, bytenr, num_bytes, 0);
return 0;
}
pin = 1;
}
/* if data pin when any transaction has committed this */
if (ref_generation != trans->transid)
pin = 1;
ret = __free_extent(trans, root, bytenr, num_bytes, parent,
root_objectid, ref_generation,
owner_objectid, pin, pin == 0);
finish_current_insert(trans, root->fs_info->extent_root, 0);
pending_ret = del_pending_extents(trans, root->fs_info->extent_root, 0);
return ret ? ret : pending_ret;
}
int btrfs_free_extent(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
u64 bytenr, u64 num_bytes, u64 parent,
u64 root_objectid, u64 ref_generation,
u64 owner_objectid, int pin)
{
int ret;
ret = __btrfs_free_extent(trans, root, bytenr, num_bytes, parent,
root_objectid, ref_generation,
owner_objectid, pin);
return ret;
}
static u64 stripe_align(struct btrfs_root *root, u64 val)
{
u64 mask = ((u64)root->stripesize - 1);
u64 ret = (val + mask) & ~mask;
return ret;
}
/*
* walks the btree of allocated extents and find a hole of a given size.
* The key ins is changed to record the hole:
* ins->objectid == block start
* ins->flags = BTRFS_EXTENT_ITEM_KEY
* ins->offset == number of blocks
* Any available blocks before search_start are skipped.
*/
static int noinline find_free_extent(struct btrfs_trans_handle *trans,
struct btrfs_root *orig_root,
u64 num_bytes, u64 empty_size,
u64 search_start, u64 search_end,
u64 hint_byte, struct btrfs_key *ins,
u64 exclude_start, u64 exclude_nr,
int data)
{
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
int ret = 0;
struct btrfs_root * root = orig_root->fs_info->extent_root;
u64 total_needed = num_bytes;
u64 *last_ptr = NULL;
u64 last_wanted = 0;
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
struct btrfs_block_group_cache *block_group = NULL;
int chunk_alloc_done = 0;
int empty_cluster = 2 * 1024 * 1024;
int allowed_chunk_alloc = 0;
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
struct list_head *head = NULL, *cur = NULL;
int loop = 0;
int extra_loop = 0;
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
struct btrfs_space_info *space_info;
WARN_ON(num_bytes < root->sectorsize);
btrfs_set_key_type(ins, BTRFS_EXTENT_ITEM_KEY);
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
ins->objectid = 0;
ins->offset = 0;
if (orig_root->ref_cows || empty_size)
allowed_chunk_alloc = 1;
if (data & BTRFS_BLOCK_GROUP_METADATA) {
last_ptr = &root->fs_info->last_alloc;
empty_cluster = 64 * 1024;
}
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
if ((data & BTRFS_BLOCK_GROUP_DATA) && btrfs_test_opt(root, SSD))
last_ptr = &root->fs_info->last_data_alloc;
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
if (last_ptr) {
if (*last_ptr) {
hint_byte = *last_ptr;
last_wanted = *last_ptr;
} else
empty_size += empty_cluster;
} else {
empty_cluster = 0;
}
search_start = max(search_start, first_logical_byte(root, 0));
search_start = max(search_start, hint_byte);
if (last_wanted && search_start != last_wanted) {
last_wanted = 0;
empty_size += empty_cluster;
}
total_needed += empty_size;
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
block_group = btrfs_lookup_block_group(root->fs_info, search_start);
if (!block_group)
block_group = btrfs_lookup_first_block_group(root->fs_info,
search_start);
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
space_info = __find_space_info(root->fs_info, data);
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
down_read(&space_info->groups_sem);
while (1) {
struct btrfs_free_space *free_space;
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
/*
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
* the only way this happens if our hint points to a block
* group thats not of the proper type, while looping this
* should never happen
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
*/
if (empty_size)
extra_loop = 1;
if (!block_group)
goto new_group_no_lock;
if (unlikely(!block_group->cached)) {
mutex_lock(&block_group->cache_mutex);
ret = cache_block_group(root, block_group);
mutex_unlock(&block_group->cache_mutex);
if (ret)
break;
}
mutex_lock(&block_group->alloc_mutex);
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
if (unlikely(!block_group_bits(block_group, data)))
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
goto new_group;
if (unlikely(block_group->ro))
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
goto new_group;
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
free_space = btrfs_find_free_space(block_group, search_start,
total_needed);
if (free_space) {
u64 start = block_group->key.objectid;
u64 end = block_group->key.objectid +
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
block_group->key.offset;
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
search_start = stripe_align(root, free_space->offset);
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
/* move on to the next group */
if (search_start + num_bytes >= search_end)
goto new_group;
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
/* move on to the next group */
if (search_start + num_bytes > end)
goto new_group;
if (last_wanted && search_start != last_wanted) {
total_needed += empty_cluster;
empty_size += empty_cluster;
last_wanted = 0;
/*
* if search_start is still in this block group
* then we just re-search this block group
*/
if (search_start >= start &&
search_start < end) {
mutex_unlock(&block_group->alloc_mutex);
continue;
}
/* else we go to the next block group */
goto new_group;
}
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
if (exclude_nr > 0 &&
(search_start + num_bytes > exclude_start &&
search_start < exclude_start + exclude_nr)) {
search_start = exclude_start + exclude_nr;
/*
* if search_start is still in this block group
* then we just re-search this block group
*/
if (search_start >= start &&
search_start < end) {
mutex_unlock(&block_group->alloc_mutex);
last_wanted = 0;
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
continue;
}
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
/* else we go to the next block group */
goto new_group;
}
ins->objectid = search_start;
ins->offset = num_bytes;
btrfs_remove_free_space_lock(block_group, search_start,
num_bytes);
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
/* we are all good, lets return */
mutex_unlock(&block_group->alloc_mutex);
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
break;
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
}
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
new_group:
mutex_unlock(&block_group->alloc_mutex);
new_group_no_lock:
/* don't try to compare new allocations against the
* last allocation any more
*/
last_wanted = 0;
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
/*
* Here's how this works.
* loop == 0: we were searching a block group via a hint
* and didn't find anything, so we start at
* the head of the block groups and keep searching
* loop == 1: we're searching through all of the block groups
* if we hit the head again we have searched
* all of the block groups for this space and we
* need to try and allocate, if we cant error out.
* loop == 2: we allocated more space and are looping through
* all of the block groups again.
*/
if (loop == 0) {
head = &space_info->block_groups;
cur = head->next;
loop++;
} else if (loop == 1 && cur == head) {
int keep_going;
/* at this point we give up on the empty_size
* allocations and just try to allocate the min
* space.
*
* The extra_loop field was set if an empty_size
* allocation was attempted above, and if this
* is try we need to try the loop again without
* the additional empty_size.
*/
total_needed -= empty_size;
empty_size = 0;
keep_going = extra_loop;
loop++;
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
if (allowed_chunk_alloc && !chunk_alloc_done) {
up_read(&space_info->groups_sem);
ret = do_chunk_alloc(trans, root, num_bytes +
2 * 1024 * 1024, data, 1);
down_read(&space_info->groups_sem);
if (ret < 0)
goto loop_check;
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
head = &space_info->block_groups;
/*
* we've allocated a new chunk, keep
* trying
*/
keep_going = 1;
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
chunk_alloc_done = 1;
} else if (!allowed_chunk_alloc) {
space_info->force_alloc = 1;
}
loop_check:
if (keep_going) {
cur = head->next;
extra_loop = 0;
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
} else {
break;
}
} else if (cur == head) {
break;
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
}
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
block_group = list_entry(cur, struct btrfs_block_group_cache,
list);
search_start = block_group->key.objectid;
cur = cur->next;
}
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
/* we found what we needed */
if (ins->objectid) {
if (!(data & BTRFS_BLOCK_GROUP_DATA))
trans->block_group = block_group;
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
if (last_ptr)
*last_ptr = ins->objectid + ins->offset;
ret = 0;
} else if (!ret) {
printk(KERN_ERR "we were searching for %Lu bytes, num_bytes %Lu,"
" loop %d, allowed_alloc %d\n", total_needed, num_bytes,
loop, allowed_chunk_alloc);
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
ret = -ENOSPC;
}
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
up_read(&space_info->groups_sem);
return ret;
}
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
static void dump_space_info(struct btrfs_space_info *info, u64 bytes)
{
struct btrfs_block_group_cache *cache;
struct list_head *l;
printk(KERN_INFO "space_info has %Lu free, is %sfull\n",
info->total_bytes - info->bytes_used - info->bytes_pinned -
info->bytes_reserved, (info->full) ? "" : "not ");
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
down_read(&info->groups_sem);
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
list_for_each(l, &info->block_groups) {
cache = list_entry(l, struct btrfs_block_group_cache, list);
spin_lock(&cache->lock);
printk(KERN_INFO "block group %Lu has %Lu bytes, %Lu used "
"%Lu pinned %Lu reserved\n",
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
cache->key.objectid, cache->key.offset,
btrfs_block_group_used(&cache->item),
cache->pinned, cache->reserved);
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
btrfs_dump_free_space(cache, bytes);
spin_unlock(&cache->lock);
}
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
up_read(&info->groups_sem);
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
}
static int __btrfs_reserve_extent(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
u64 num_bytes, u64 min_alloc_size,
u64 empty_size, u64 hint_byte,
u64 search_end, struct btrfs_key *ins,
u64 data)
{
int ret;
u64 search_start = 0;
u64 alloc_profile;
struct btrfs_fs_info *info = root->fs_info;
if (data) {
alloc_profile = info->avail_data_alloc_bits &
info->data_alloc_profile;
data = BTRFS_BLOCK_GROUP_DATA | alloc_profile;
} else if (root == root->fs_info->chunk_root) {
alloc_profile = info->avail_system_alloc_bits &
info->system_alloc_profile;
data = BTRFS_BLOCK_GROUP_SYSTEM | alloc_profile;
} else {
alloc_profile = info->avail_metadata_alloc_bits &
info->metadata_alloc_profile;
data = BTRFS_BLOCK_GROUP_METADATA | alloc_profile;
}
again:
data = btrfs_reduce_alloc_profile(root, data);
/*
* the only place that sets empty_size is btrfs_realloc_node, which
* is not called recursively on allocations
*/
if (empty_size || root->ref_cows) {
if (!(data & BTRFS_BLOCK_GROUP_METADATA)) {
ret = do_chunk_alloc(trans, root->fs_info->extent_root,
2 * 1024 * 1024,
BTRFS_BLOCK_GROUP_METADATA |
(info->metadata_alloc_profile &
info->avail_metadata_alloc_bits), 0);
}
ret = do_chunk_alloc(trans, root->fs_info->extent_root,
num_bytes + 2 * 1024 * 1024, data, 0);
}
WARN_ON(num_bytes < root->sectorsize);
ret = find_free_extent(trans, root, num_bytes, empty_size,
search_start, search_end, hint_byte, ins,
trans->alloc_exclude_start,
trans->alloc_exclude_nr, data);
if (ret == -ENOSPC && num_bytes > min_alloc_size) {
num_bytes = num_bytes >> 1;
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
num_bytes = num_bytes & ~(root->sectorsize - 1);
num_bytes = max(num_bytes, min_alloc_size);
do_chunk_alloc(trans, root->fs_info->extent_root,
num_bytes, data, 1);
goto again;
}
if (ret) {
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
struct btrfs_space_info *sinfo;
sinfo = __find_space_info(root->fs_info, data);
printk("allocation failed flags %Lu, wanted %Lu\n",
data, num_bytes);
dump_space_info(sinfo, num_bytes);
BUG();
}
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
return ret;
}
int btrfs_free_reserved_extent(struct btrfs_root *root, u64 start, u64 len)
{
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
struct btrfs_block_group_cache *cache;
cache = btrfs_lookup_block_group(root->fs_info, start);
if (!cache) {
printk(KERN_ERR "Unable to find block group for %Lu\n", start);
return -ENOSPC;
}
btrfs_add_free_space(cache, start, len);
2008-09-26 14:09:34 +00:00
update_reserved_extents(root, start, len, 0);
return 0;
}
int btrfs_reserve_extent(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
u64 num_bytes, u64 min_alloc_size,
u64 empty_size, u64 hint_byte,
u64 search_end, struct btrfs_key *ins,
u64 data)
{
int ret;
ret = __btrfs_reserve_extent(trans, root, num_bytes, min_alloc_size,
empty_size, hint_byte, search_end, ins,
data);
update_reserved_extents(root, ins->objectid, ins->offset, 1);
return ret;
}
static int __btrfs_alloc_reserved_extent(struct btrfs_trans_handle *trans,
struct btrfs_root *root, u64 parent,
u64 root_objectid, u64 ref_generation,
u64 owner, struct btrfs_key *ins)
{
int ret;
int pending_ret;
u64 super_used;
u64 root_used;
u64 num_bytes = ins->offset;
u32 sizes[2];
struct btrfs_fs_info *info = root->fs_info;
struct btrfs_root *extent_root = info->extent_root;
struct btrfs_extent_item *extent_item;
struct btrfs_extent_ref *ref;
struct btrfs_path *path;
struct btrfs_key keys[2];
if (parent == 0)
parent = ins->objectid;
/* block accounting for super block */
spin_lock_irq(&info->delalloc_lock);
super_used = btrfs_super_bytes_used(&info->super_copy);
btrfs_set_super_bytes_used(&info->super_copy, super_used + num_bytes);
spin_unlock_irq(&info->delalloc_lock);
/* block accounting for root item */
root_used = btrfs_root_used(&root->root_item);
btrfs_set_root_used(&root->root_item, root_used + num_bytes);
if (root == extent_root) {
struct pending_extent_op *extent_op;
extent_op = kmalloc(sizeof(*extent_op), GFP_NOFS);
BUG_ON(!extent_op);
extent_op->type = PENDING_EXTENT_INSERT;
extent_op->bytenr = ins->objectid;
extent_op->num_bytes = ins->offset;
extent_op->parent = parent;
extent_op->orig_parent = 0;
extent_op->generation = ref_generation;
extent_op->orig_generation = 0;
extent_op->level = (int)owner;
Btrfs: batch extent inserts/updates/deletions on the extent root While profiling the allocator I noticed a good amount of time was being spent in finish_current_insert and del_pending_extents, and as the filesystem filled up more and more time was being spent in those functions. This patch aims to try and reduce that problem. This happens two ways 1) track if we tried to delete an extent that we are going to update or insert. Once we get into finish_current_insert we discard any of the extents that were marked for deletion. This saves us from doing unnecessary work almost every time finish_current_insert runs. 2) Batch insertion/updates/deletions. Instead of doing a btrfs_search_slot for each individual extent and doing the needed operation, we instead keep the leaf around and see if there is anything else we can do on that leaf. On the insert case I introduced a btrfs_insert_some_items, which will take an array of keys with an array of data_sizes and try and squeeze in as many of those keys as possible, and then return how many keys it was able to insert. In the update case we search for an extent ref, update the ref and then loop through the leaf to see if any of the other refs we are looking to update are on that leaf, and then once we are done we release the path and search for the next ref we need to update. And finally for the deletion we try and delete the extent+ref in pairs, so we will try to find extent+ref pairs next to the extent we are trying to free and free them in bulk if possible. This along with the other cluster fix that Chris pushed out a bit ago helps make the allocator preform more uniformly as it fills up the disk. There is still a slight drop as we fill up the disk since we start having to stick new blocks in odd places which results in more COW's than on a empty fs, but the drop is not nearly as severe as it was before. Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-11-12 19:19:50 +00:00
INIT_LIST_HEAD(&extent_op->list);
extent_op->del = 0;
mutex_lock(&root->fs_info->extent_ins_mutex);
set_extent_bits(&root->fs_info->extent_ins, ins->objectid,
ins->objectid + ins->offset - 1,
EXTENT_WRITEBACK, GFP_NOFS);
set_state_private(&root->fs_info->extent_ins,
ins->objectid, (unsigned long)extent_op);
mutex_unlock(&root->fs_info->extent_ins_mutex);
goto update_block;
}
memcpy(&keys[0], ins, sizeof(*ins));
keys[1].objectid = ins->objectid;
keys[1].type = BTRFS_EXTENT_REF_KEY;
keys[1].offset = parent;
sizes[0] = sizeof(*extent_item);
sizes[1] = sizeof(*ref);
path = btrfs_alloc_path();
BUG_ON(!path);
ret = btrfs_insert_empty_items(trans, extent_root, path, keys,
sizes, 2);
BUG_ON(ret);
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
extent_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
struct btrfs_extent_item);
btrfs_set_extent_refs(path->nodes[0], extent_item, 1);
ref = btrfs_item_ptr(path->nodes[0], path->slots[0] + 1,
struct btrfs_extent_ref);
btrfs_set_ref_root(path->nodes[0], ref, root_objectid);
btrfs_set_ref_generation(path->nodes[0], ref, ref_generation);
btrfs_set_ref_objectid(path->nodes[0], ref, owner);
btrfs_set_ref_num_refs(path->nodes[0], ref, 1);
btrfs_mark_buffer_dirty(path->nodes[0]);
trans->alloc_exclude_start = 0;
trans->alloc_exclude_nr = 0;
btrfs_free_path(path);
finish_current_insert(trans, extent_root, 0);
pending_ret = del_pending_extents(trans, extent_root, 0);
if (ret)
goto out;
if (pending_ret) {
ret = pending_ret;
goto out;
}
update_block:
ret = update_block_group(trans, root, ins->objectid, ins->offset, 1, 0);
if (ret) {
printk("update block group failed for %Lu %Lu\n",
ins->objectid, ins->offset);
BUG();
}
out:
return ret;
}
int btrfs_alloc_reserved_extent(struct btrfs_trans_handle *trans,
struct btrfs_root *root, u64 parent,
u64 root_objectid, u64 ref_generation,
u64 owner, struct btrfs_key *ins)
{
int ret;
if (root_objectid == BTRFS_TREE_LOG_OBJECTID)
return 0;
ret = __btrfs_alloc_reserved_extent(trans, root, parent, root_objectid,
ref_generation, owner, ins);
update_reserved_extents(root, ins->objectid, ins->offset, 0);
return ret;
}
/*
* this is used by the tree logging recovery code. It records that
* an extent has been allocated and makes sure to clear the free
* space cache bits as well
*/
int btrfs_alloc_logged_extent(struct btrfs_trans_handle *trans,
struct btrfs_root *root, u64 parent,
u64 root_objectid, u64 ref_generation,
u64 owner, struct btrfs_key *ins)
{
int ret;
struct btrfs_block_group_cache *block_group;
block_group = btrfs_lookup_block_group(root->fs_info, ins->objectid);
mutex_lock(&block_group->cache_mutex);
cache_block_group(root, block_group);
mutex_unlock(&block_group->cache_mutex);
ret = btrfs_remove_free_space(block_group, ins->objectid,
ins->offset);
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
BUG_ON(ret);
ret = __btrfs_alloc_reserved_extent(trans, root, parent, root_objectid,
ref_generation, owner, ins);
return ret;
}
/*
* finds a free extent and does all the dirty work required for allocation
* returns the key for the extent through ins, and a tree buffer for
* the first block of the extent through buf.
*
* returns 0 if everything worked, non-zero otherwise.
*/
int btrfs_alloc_extent(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
u64 num_bytes, u64 parent, u64 min_alloc_size,
u64 root_objectid, u64 ref_generation,
u64 owner_objectid, u64 empty_size, u64 hint_byte,
u64 search_end, struct btrfs_key *ins, u64 data)
{
int ret;
ret = __btrfs_reserve_extent(trans, root, num_bytes,
min_alloc_size, empty_size, hint_byte,
search_end, ins, data);
BUG_ON(ret);
if (root_objectid != BTRFS_TREE_LOG_OBJECTID) {
ret = __btrfs_alloc_reserved_extent(trans, root, parent,
root_objectid, ref_generation,
owner_objectid, ins);
BUG_ON(ret);
} else {
update_reserved_extents(root, ins->objectid, ins->offset, 1);
}
return ret;
}
struct extent_buffer *btrfs_init_new_buffer(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
u64 bytenr, u32 blocksize)
{
struct extent_buffer *buf;
buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
if (!buf)
return ERR_PTR(-ENOMEM);
btrfs_set_header_generation(buf, trans->transid);
btrfs_tree_lock(buf);
clean_tree_block(trans, root, buf);
btrfs_set_buffer_uptodate(buf);
if (root->root_key.objectid == BTRFS_TREE_LOG_OBJECTID) {
set_extent_dirty(&root->dirty_log_pages, buf->start,
buf->start + buf->len - 1, GFP_NOFS);
} else {
set_extent_dirty(&trans->transaction->dirty_pages, buf->start,
buf->start + buf->len - 1, GFP_NOFS);
}
trans->blocks_used++;
return buf;
}
/*
* helper function to allocate a block for a given tree
* returns the tree buffer or NULL.
*/
struct extent_buffer *btrfs_alloc_free_block(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
u32 blocksize, u64 parent,
u64 root_objectid,
u64 ref_generation,
int level,
u64 hint,
u64 empty_size)
{
struct btrfs_key ins;
int ret;
struct extent_buffer *buf;
ret = btrfs_alloc_extent(trans, root, blocksize, parent, blocksize,
root_objectid, ref_generation, level,
empty_size, hint, (u64)-1, &ins, 0);
if (ret) {
BUG_ON(ret > 0);
return ERR_PTR(ret);
}
buf = btrfs_init_new_buffer(trans, root, ins.objectid, blocksize);
return buf;
}
int btrfs_drop_leaf_ref(struct btrfs_trans_handle *trans,
struct btrfs_root *root, struct extent_buffer *leaf)
{
u64 leaf_owner;
u64 leaf_generation;
struct btrfs_key key;
struct btrfs_file_extent_item *fi;
int i;
int nritems;
int ret;
BUG_ON(!btrfs_is_leaf(leaf));
nritems = btrfs_header_nritems(leaf);
leaf_owner = btrfs_header_owner(leaf);
leaf_generation = btrfs_header_generation(leaf);
for (i = 0; i < nritems; i++) {
u64 disk_bytenr;
cond_resched();
btrfs_item_key_to_cpu(leaf, &key, i);
if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
continue;
fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
if (btrfs_file_extent_type(leaf, fi) ==
BTRFS_FILE_EXTENT_INLINE)
continue;
/*
* FIXME make sure to insert a trans record that
* repeats the snapshot del on crash
*/
disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
if (disk_bytenr == 0)
continue;
ret = __btrfs_free_extent(trans, root, disk_bytenr,
btrfs_file_extent_disk_num_bytes(leaf, fi),
leaf->start, leaf_owner, leaf_generation,
key.objectid, 0);
BUG_ON(ret);
atomic_inc(&root->fs_info->throttle_gen);
wake_up(&root->fs_info->transaction_throttle);
cond_resched();
}
return 0;
}
static int noinline cache_drop_leaf_ref(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct btrfs_leaf_ref *ref)
{
int i;
int ret;
struct btrfs_extent_info *info = ref->extents;
for (i = 0; i < ref->nritems; i++) {
ret = __btrfs_free_extent(trans, root, info->bytenr,
info->num_bytes, ref->bytenr,
ref->owner, ref->generation,
info->objectid, 0);
atomic_inc(&root->fs_info->throttle_gen);
wake_up(&root->fs_info->transaction_throttle);
cond_resched();
BUG_ON(ret);
info++;
}
return 0;
}
int drop_snap_lookup_refcount(struct btrfs_root *root, u64 start, u64 len,
u32 *refs)
{
int ret;
ret = btrfs_lookup_extent_ref(NULL, root, start, len, refs);
BUG_ON(ret);
#if 0 // some debugging code in case we see problems here
/* if the refs count is one, it won't get increased again. But
* if the ref count is > 1, someone may be decreasing it at
* the same time we are.
*/
if (*refs != 1) {
struct extent_buffer *eb = NULL;
eb = btrfs_find_create_tree_block(root, start, len);
if (eb)
btrfs_tree_lock(eb);
mutex_lock(&root->fs_info->alloc_mutex);
ret = lookup_extent_ref(NULL, root, start, len, refs);
BUG_ON(ret);
mutex_unlock(&root->fs_info->alloc_mutex);
if (eb) {
btrfs_tree_unlock(eb);
free_extent_buffer(eb);
}
if (*refs == 1) {
printk("block %llu went down to one during drop_snap\n",
(unsigned long long)start);
}
}
#endif
cond_resched();
return ret;
}
/*
* helper function for drop_snapshot, this walks down the tree dropping ref
* counts as it goes.
*/
static int noinline walk_down_tree(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct btrfs_path *path, int *level)
{
u64 root_owner;
u64 root_gen;
u64 bytenr;
u64 ptr_gen;
struct extent_buffer *next;
struct extent_buffer *cur;
struct extent_buffer *parent;
struct btrfs_leaf_ref *ref;
u32 blocksize;
int ret;
u32 refs;
WARN_ON(*level < 0);
WARN_ON(*level >= BTRFS_MAX_LEVEL);
ret = drop_snap_lookup_refcount(root, path->nodes[*level]->start,
path->nodes[*level]->len, &refs);
BUG_ON(ret);
if (refs > 1)
goto out;
/*
* walk down to the last node level and free all the leaves
*/
while(*level >= 0) {
WARN_ON(*level < 0);
WARN_ON(*level >= BTRFS_MAX_LEVEL);
cur = path->nodes[*level];
if (btrfs_header_level(cur) != *level)
WARN_ON(1);
if (path->slots[*level] >=
btrfs_header_nritems(cur))
break;
if (*level == 0) {
ret = btrfs_drop_leaf_ref(trans, root, cur);
BUG_ON(ret);
break;
}
bytenr = btrfs_node_blockptr(cur, path->slots[*level]);
ptr_gen = btrfs_node_ptr_generation(cur, path->slots[*level]);
blocksize = btrfs_level_size(root, *level - 1);
ret = drop_snap_lookup_refcount(root, bytenr, blocksize, &refs);
BUG_ON(ret);
if (refs != 1) {
parent = path->nodes[*level];
root_owner = btrfs_header_owner(parent);
root_gen = btrfs_header_generation(parent);
path->slots[*level]++;
ret = __btrfs_free_extent(trans, root, bytenr,
blocksize, parent->start,
root_owner, root_gen,
*level - 1, 1);
BUG_ON(ret);
atomic_inc(&root->fs_info->throttle_gen);
wake_up(&root->fs_info->transaction_throttle);
cond_resched();
continue;
}
/*
* at this point, we have a single ref, and since the
* only place referencing this extent is a dead root
* the reference count should never go higher.
* So, we don't need to check it again
*/
if (*level == 1) {
ref = btrfs_lookup_leaf_ref(root, bytenr);
2008-09-26 14:09:34 +00:00
if (ref && ref->generation != ptr_gen) {
btrfs_free_leaf_ref(root, ref);
ref = NULL;
}
if (ref) {
ret = cache_drop_leaf_ref(trans, root, ref);
BUG_ON(ret);
btrfs_remove_leaf_ref(root, ref);
btrfs_free_leaf_ref(root, ref);
*level = 0;
break;
}
if (printk_ratelimit()) {
printk("leaf ref miss for bytenr %llu\n",
(unsigned long long)bytenr);
}
}
next = btrfs_find_tree_block(root, bytenr, blocksize);
if (!next || !btrfs_buffer_uptodate(next, ptr_gen)) {
free_extent_buffer(next);
next = read_tree_block(root, bytenr, blocksize,
ptr_gen);
cond_resched();
#if 0
/*
* this is a debugging check and can go away
* the ref should never go all the way down to 1
* at this point
*/
ret = lookup_extent_ref(NULL, root, bytenr, blocksize,
&refs);
BUG_ON(ret);
WARN_ON(refs != 1);
#endif
}
WARN_ON(*level <= 0);
if (path->nodes[*level-1])
free_extent_buffer(path->nodes[*level-1]);
path->nodes[*level-1] = next;
*level = btrfs_header_level(next);
path->slots[*level] = 0;
cond_resched();
}
out:
WARN_ON(*level < 0);
WARN_ON(*level >= BTRFS_MAX_LEVEL);
if (path->nodes[*level] == root->node) {
parent = path->nodes[*level];
bytenr = path->nodes[*level]->start;
} else {
parent = path->nodes[*level + 1];
bytenr = btrfs_node_blockptr(parent, path->slots[*level + 1]);
}
blocksize = btrfs_level_size(root, *level);
root_owner = btrfs_header_owner(parent);
root_gen = btrfs_header_generation(parent);
ret = __btrfs_free_extent(trans, root, bytenr, blocksize,
parent->start, root_owner, root_gen,
*level, 1);
free_extent_buffer(path->nodes[*level]);
path->nodes[*level] = NULL;
*level += 1;
BUG_ON(ret);
cond_resched();
return 0;
}
/*
* helper function for drop_subtree, this function is similar to
* walk_down_tree. The main difference is that it checks reference
* counts while tree blocks are locked.
*/
static int noinline walk_down_subtree(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct btrfs_path *path, int *level)
{
struct extent_buffer *next;
struct extent_buffer *cur;
struct extent_buffer *parent;
u64 bytenr;
u64 ptr_gen;
u32 blocksize;
u32 refs;
int ret;
cur = path->nodes[*level];
ret = btrfs_lookup_extent_ref(trans, root, cur->start, cur->len,
&refs);
BUG_ON(ret);
if (refs > 1)
goto out;
while (*level >= 0) {
cur = path->nodes[*level];
if (*level == 0) {
ret = btrfs_drop_leaf_ref(trans, root, cur);
BUG_ON(ret);
clean_tree_block(trans, root, cur);
break;
}
if (path->slots[*level] >= btrfs_header_nritems(cur)) {
clean_tree_block(trans, root, cur);
break;
}
bytenr = btrfs_node_blockptr(cur, path->slots[*level]);
blocksize = btrfs_level_size(root, *level - 1);
ptr_gen = btrfs_node_ptr_generation(cur, path->slots[*level]);
next = read_tree_block(root, bytenr, blocksize, ptr_gen);
btrfs_tree_lock(next);
ret = btrfs_lookup_extent_ref(trans, root, bytenr, blocksize,
&refs);
BUG_ON(ret);
if (refs > 1) {
parent = path->nodes[*level];
ret = btrfs_free_extent(trans, root, bytenr,
blocksize, parent->start,
btrfs_header_owner(parent),
btrfs_header_generation(parent),
*level - 1, 1);
BUG_ON(ret);
path->slots[*level]++;
btrfs_tree_unlock(next);
free_extent_buffer(next);
continue;
}
*level = btrfs_header_level(next);
path->nodes[*level] = next;
path->slots[*level] = 0;
path->locks[*level] = 1;
cond_resched();
}
out:
parent = path->nodes[*level + 1];
bytenr = path->nodes[*level]->start;
blocksize = path->nodes[*level]->len;
ret = btrfs_free_extent(trans, root, bytenr, blocksize,
parent->start, btrfs_header_owner(parent),
btrfs_header_generation(parent), *level, 1);
BUG_ON(ret);
if (path->locks[*level]) {
btrfs_tree_unlock(path->nodes[*level]);
path->locks[*level] = 0;
}
free_extent_buffer(path->nodes[*level]);
path->nodes[*level] = NULL;
*level += 1;
cond_resched();
return 0;
}
/*
* helper for dropping snapshots. This walks back up the tree in the path
* to find the first node higher up where we haven't yet gone through
* all the slots
*/
static int noinline walk_up_tree(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct btrfs_path *path,
int *level, int max_level)
{
u64 root_owner;
u64 root_gen;
struct btrfs_root_item *root_item = &root->root_item;
int i;
int slot;
int ret;
for (i = *level; i < max_level && path->nodes[i]; i++) {
slot = path->slots[i];
if (slot < btrfs_header_nritems(path->nodes[i]) - 1) {
struct extent_buffer *node;
struct btrfs_disk_key disk_key;
node = path->nodes[i];
path->slots[i]++;
*level = i;
WARN_ON(*level == 0);
btrfs_node_key(node, &disk_key, path->slots[i]);
memcpy(&root_item->drop_progress,
&disk_key, sizeof(disk_key));
root_item->drop_level = i;
return 0;
} else {
struct extent_buffer *parent;
if (path->nodes[*level] == root->node)
parent = path->nodes[*level];
else
parent = path->nodes[*level + 1];
root_owner = btrfs_header_owner(parent);
root_gen = btrfs_header_generation(parent);
clean_tree_block(trans, root, path->nodes[*level]);
ret = btrfs_free_extent(trans, root,
path->nodes[*level]->start,
path->nodes[*level]->len,
parent->start, root_owner,
root_gen, *level, 1);
BUG_ON(ret);
if (path->locks[*level]) {
btrfs_tree_unlock(path->nodes[*level]);
path->locks[*level] = 0;
}
free_extent_buffer(path->nodes[*level]);
path->nodes[*level] = NULL;
*level = i + 1;
}
}
return 1;
}
/*
* drop the reference count on the tree rooted at 'snap'. This traverses
* the tree freeing any blocks that have a ref count of zero after being
* decremented.
*/
int btrfs_drop_snapshot(struct btrfs_trans_handle *trans, struct btrfs_root
*root)
{
int ret = 0;
int wret;
int level;
struct btrfs_path *path;
int i;
int orig_level;
struct btrfs_root_item *root_item = &root->root_item;
WARN_ON(!mutex_is_locked(&root->fs_info->drop_mutex));
path = btrfs_alloc_path();
BUG_ON(!path);
level = btrfs_header_level(root->node);
orig_level = level;
if (btrfs_disk_key_objectid(&root_item->drop_progress) == 0) {
path->nodes[level] = root->node;
extent_buffer_get(root->node);
path->slots[level] = 0;
} else {
struct btrfs_key key;
struct btrfs_disk_key found_key;
struct extent_buffer *node;
btrfs_disk_key_to_cpu(&key, &root_item->drop_progress);
level = root_item->drop_level;
path->lowest_level = level;
wret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
if (wret < 0) {
ret = wret;
goto out;
}
node = path->nodes[level];
btrfs_node_key(node, &found_key, path->slots[level]);
WARN_ON(memcmp(&found_key, &root_item->drop_progress,
sizeof(found_key)));
/*
* unlock our path, this is safe because only this
* function is allowed to delete this snapshot
*/
for (i = 0; i < BTRFS_MAX_LEVEL; i++) {
if (path->nodes[i] && path->locks[i]) {
path->locks[i] = 0;
btrfs_tree_unlock(path->nodes[i]);
}
}
}
while(1) {
wret = walk_down_tree(trans, root, path, &level);
if (wret > 0)
break;
if (wret < 0)
ret = wret;
wret = walk_up_tree(trans, root, path, &level,
BTRFS_MAX_LEVEL);
if (wret > 0)
break;
if (wret < 0)
ret = wret;
if (trans->transaction->in_commit) {
ret = -EAGAIN;
break;
}
atomic_inc(&root->fs_info->throttle_gen);
wake_up(&root->fs_info->transaction_throttle);
}
for (i = 0; i <= orig_level; i++) {
if (path->nodes[i]) {
free_extent_buffer(path->nodes[i]);
path->nodes[i] = NULL;
}
}
out:
btrfs_free_path(path);
return ret;
}
int btrfs_drop_subtree(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct extent_buffer *node,
struct extent_buffer *parent)
{
struct btrfs_path *path;
int level;
int parent_level;
int ret = 0;
int wret;
path = btrfs_alloc_path();
BUG_ON(!path);
BUG_ON(!btrfs_tree_locked(parent));
parent_level = btrfs_header_level(parent);
extent_buffer_get(parent);
path->nodes[parent_level] = parent;
path->slots[parent_level] = btrfs_header_nritems(parent);
BUG_ON(!btrfs_tree_locked(node));
level = btrfs_header_level(node);
extent_buffer_get(node);
path->nodes[level] = node;
path->slots[level] = 0;
while (1) {
wret = walk_down_subtree(trans, root, path, &level);
if (wret < 0)
ret = wret;
if (wret != 0)
break;
wret = walk_up_tree(trans, root, path, &level, parent_level);
if (wret < 0)
ret = wret;
if (wret != 0)
break;
}
btrfs_free_path(path);
return ret;
}
static unsigned long calc_ra(unsigned long start, unsigned long last,
unsigned long nr)
{
return min(last, start + nr - 1);
}
static int noinline relocate_inode_pages(struct inode *inode, u64 start,
u64 len)
{
u64 page_start;
u64 page_end;
2008-09-26 14:09:34 +00:00
unsigned long first_index;
unsigned long last_index;
unsigned long i;
struct page *page;
struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
struct file_ra_state *ra;
struct btrfs_ordered_extent *ordered;
2008-09-26 14:09:34 +00:00
unsigned int total_read = 0;
unsigned int total_dirty = 0;
int ret = 0;
ra = kzalloc(sizeof(*ra), GFP_NOFS);
mutex_lock(&inode->i_mutex);
2008-09-26 14:09:34 +00:00
first_index = start >> PAGE_CACHE_SHIFT;
last_index = (start + len - 1) >> PAGE_CACHE_SHIFT;
2008-09-26 14:09:34 +00:00
/* make sure the dirty trick played by the caller work */
ret = invalidate_inode_pages2_range(inode->i_mapping,
first_index, last_index);
if (ret)
goto out_unlock;
file_ra_state_init(ra, inode->i_mapping);
2008-09-26 14:09:34 +00:00
for (i = first_index ; i <= last_index; i++) {
if (total_read % ra->ra_pages == 0) {
btrfs_force_ra(inode->i_mapping, ra, NULL, i,
2008-09-26 14:09:34 +00:00
calc_ra(i, last_index, ra->ra_pages));
}
total_read++;
again:
if (((u64)i << PAGE_CACHE_SHIFT) > i_size_read(inode))
2008-09-26 14:09:34 +00:00
BUG_ON(1);
page = grab_cache_page(inode->i_mapping, i);
if (!page) {
2008-09-26 14:09:34 +00:00
ret = -ENOMEM;
goto out_unlock;
}
if (!PageUptodate(page)) {
btrfs_readpage(NULL, page);
lock_page(page);
if (!PageUptodate(page)) {
unlock_page(page);
page_cache_release(page);
2008-09-26 14:09:34 +00:00
ret = -EIO;
goto out_unlock;
}
}
wait_on_page_writeback(page);
page_start = (u64)page->index << PAGE_CACHE_SHIFT;
page_end = page_start + PAGE_CACHE_SIZE - 1;
lock_extent(io_tree, page_start, page_end, GFP_NOFS);
ordered = btrfs_lookup_ordered_extent(inode, page_start);
if (ordered) {
unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
unlock_page(page);
page_cache_release(page);
btrfs_start_ordered_extent(inode, ordered, 1);
btrfs_put_ordered_extent(ordered);
goto again;
}
set_page_extent_mapped(page);
btrfs_set_extent_delalloc(inode, page_start, page_end);
2008-09-26 14:09:34 +00:00
if (i == first_index)
set_extent_bits(io_tree, page_start, page_end,
EXTENT_BOUNDARY, GFP_NOFS);
set_page_dirty(page);
2008-09-26 14:09:34 +00:00
total_dirty++;
unlock_extent(io_tree, page_start, page_end, GFP_NOFS);
unlock_page(page);
page_cache_release(page);
}
out_unlock:
kfree(ra);
mutex_unlock(&inode->i_mutex);
2008-09-26 14:09:34 +00:00
balance_dirty_pages_ratelimited_nr(inode->i_mapping, total_dirty);
return ret;
}
2008-09-26 14:09:34 +00:00
static int noinline relocate_data_extent(struct inode *reloc_inode,
struct btrfs_key *extent_key,
u64 offset)
{
struct btrfs_root *root = BTRFS_I(reloc_inode)->root;
struct extent_map_tree *em_tree = &BTRFS_I(reloc_inode)->extent_tree;
struct extent_map *em;
u64 start = extent_key->objectid - offset;
u64 end = start + extent_key->offset - 1;
2008-09-26 14:09:34 +00:00
em = alloc_extent_map(GFP_NOFS);
BUG_ON(!em || IS_ERR(em));
em->start = start;
2008-09-26 14:09:34 +00:00
em->len = extent_key->offset;
Btrfs: Add zlib compression support This is a large change for adding compression on reading and writing, both for inline and regular extents. It does some fairly large surgery to the writeback paths. Compression is off by default and enabled by mount -o compress. Even when the -o compress mount option is not used, it is possible to read compressed extents off the disk. If compression for a given set of pages fails to make them smaller, the file is flagged to avoid future compression attempts later. * While finding delalloc extents, the pages are locked before being sent down to the delalloc handler. This allows the delalloc handler to do complex things such as cleaning the pages, marking them writeback and starting IO on their behalf. * Inline extents are inserted at delalloc time now. This allows us to compress the data before inserting the inline extent, and it allows us to insert an inline extent that spans multiple pages. * All of the in-memory extent representations (extent_map.c, ordered-data.c etc) are changed to record both an in-memory size and an on disk size, as well as a flag for compression. From a disk format point of view, the extent pointers in the file are changed to record the on disk size of a given extent and some encoding flags. Space in the disk format is allocated for compression encoding, as well as encryption and a generic 'other' field. Neither the encryption or the 'other' field are currently used. In order to limit the amount of data read for a single random read in the file, the size of a compressed extent is limited to 128k. This is a software only limit, the disk format supports u64 sized compressed extents. In order to limit the ram consumed while processing extents, the uncompressed size of a compressed extent is limited to 256k. This is a software only limit and will be subject to tuning later. Checksumming is still done on compressed extents, and it is done on the uncompressed version of the data. This way additional encodings can be layered on without having to figure out which encoding to checksum. Compression happens at delalloc time, which is basically singled threaded because it is usually done by a single pdflush thread. This makes it tricky to spread the compression load across all the cpus on the box. We'll have to look at parallel pdflush walks of dirty inodes at a later time. Decompression is hooked into readpages and it does spread across CPUs nicely. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-29 18:49:59 +00:00
em->block_len = extent_key->offset;
2008-09-26 14:09:34 +00:00
em->block_start = extent_key->objectid;
em->bdev = root->fs_info->fs_devices->latest_bdev;
set_bit(EXTENT_FLAG_PINNED, &em->flags);
/* setup extent map to cheat btrfs_readpage */
lock_extent(&BTRFS_I(reloc_inode)->io_tree, start, end, GFP_NOFS);
2008-09-26 14:09:34 +00:00
while (1) {
int ret;
spin_lock(&em_tree->lock);
ret = add_extent_mapping(em_tree, em);
spin_unlock(&em_tree->lock);
if (ret != -EEXIST) {
free_extent_map(em);
break;
}
btrfs_drop_extent_cache(reloc_inode, start, end, 0);
}
unlock_extent(&BTRFS_I(reloc_inode)->io_tree, start, end, GFP_NOFS);
return relocate_inode_pages(reloc_inode, start, extent_key->offset);
2008-09-26 14:09:34 +00:00
}
2008-09-26 14:09:34 +00:00
struct btrfs_ref_path {
u64 extent_start;
u64 nodes[BTRFS_MAX_LEVEL];
u64 root_objectid;
u64 root_generation;
u64 owner_objectid;
u32 num_refs;
int lowest_level;
int current_level;
int shared_level;
struct btrfs_key node_keys[BTRFS_MAX_LEVEL];
u64 new_nodes[BTRFS_MAX_LEVEL];
2008-09-26 14:09:34 +00:00
};
2008-09-26 14:09:34 +00:00
struct disk_extent {
Btrfs: Add zlib compression support This is a large change for adding compression on reading and writing, both for inline and regular extents. It does some fairly large surgery to the writeback paths. Compression is off by default and enabled by mount -o compress. Even when the -o compress mount option is not used, it is possible to read compressed extents off the disk. If compression for a given set of pages fails to make them smaller, the file is flagged to avoid future compression attempts later. * While finding delalloc extents, the pages are locked before being sent down to the delalloc handler. This allows the delalloc handler to do complex things such as cleaning the pages, marking them writeback and starting IO on their behalf. * Inline extents are inserted at delalloc time now. This allows us to compress the data before inserting the inline extent, and it allows us to insert an inline extent that spans multiple pages. * All of the in-memory extent representations (extent_map.c, ordered-data.c etc) are changed to record both an in-memory size and an on disk size, as well as a flag for compression. From a disk format point of view, the extent pointers in the file are changed to record the on disk size of a given extent and some encoding flags. Space in the disk format is allocated for compression encoding, as well as encryption and a generic 'other' field. Neither the encryption or the 'other' field are currently used. In order to limit the amount of data read for a single random read in the file, the size of a compressed extent is limited to 128k. This is a software only limit, the disk format supports u64 sized compressed extents. In order to limit the ram consumed while processing extents, the uncompressed size of a compressed extent is limited to 256k. This is a software only limit and will be subject to tuning later. Checksumming is still done on compressed extents, and it is done on the uncompressed version of the data. This way additional encodings can be layered on without having to figure out which encoding to checksum. Compression happens at delalloc time, which is basically singled threaded because it is usually done by a single pdflush thread. This makes it tricky to spread the compression load across all the cpus on the box. We'll have to look at parallel pdflush walks of dirty inodes at a later time. Decompression is hooked into readpages and it does spread across CPUs nicely. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-29 18:49:59 +00:00
u64 ram_bytes;
2008-09-26 14:09:34 +00:00
u64 disk_bytenr;
u64 disk_num_bytes;
u64 offset;
u64 num_bytes;
Btrfs: Add zlib compression support This is a large change for adding compression on reading and writing, both for inline and regular extents. It does some fairly large surgery to the writeback paths. Compression is off by default and enabled by mount -o compress. Even when the -o compress mount option is not used, it is possible to read compressed extents off the disk. If compression for a given set of pages fails to make them smaller, the file is flagged to avoid future compression attempts later. * While finding delalloc extents, the pages are locked before being sent down to the delalloc handler. This allows the delalloc handler to do complex things such as cleaning the pages, marking them writeback and starting IO on their behalf. * Inline extents are inserted at delalloc time now. This allows us to compress the data before inserting the inline extent, and it allows us to insert an inline extent that spans multiple pages. * All of the in-memory extent representations (extent_map.c, ordered-data.c etc) are changed to record both an in-memory size and an on disk size, as well as a flag for compression. From a disk format point of view, the extent pointers in the file are changed to record the on disk size of a given extent and some encoding flags. Space in the disk format is allocated for compression encoding, as well as encryption and a generic 'other' field. Neither the encryption or the 'other' field are currently used. In order to limit the amount of data read for a single random read in the file, the size of a compressed extent is limited to 128k. This is a software only limit, the disk format supports u64 sized compressed extents. In order to limit the ram consumed while processing extents, the uncompressed size of a compressed extent is limited to 256k. This is a software only limit and will be subject to tuning later. Checksumming is still done on compressed extents, and it is done on the uncompressed version of the data. This way additional encodings can be layered on without having to figure out which encoding to checksum. Compression happens at delalloc time, which is basically singled threaded because it is usually done by a single pdflush thread. This makes it tricky to spread the compression load across all the cpus on the box. We'll have to look at parallel pdflush walks of dirty inodes at a later time. Decompression is hooked into readpages and it does spread across CPUs nicely. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-29 18:49:59 +00:00
u8 compression;
u8 encryption;
u16 other_encoding;
2008-09-26 14:09:34 +00:00
};
2008-09-26 14:09:34 +00:00
static int is_cowonly_root(u64 root_objectid)
{
if (root_objectid == BTRFS_ROOT_TREE_OBJECTID ||
root_objectid == BTRFS_EXTENT_TREE_OBJECTID ||
root_objectid == BTRFS_CHUNK_TREE_OBJECTID ||
root_objectid == BTRFS_DEV_TREE_OBJECTID ||
root_objectid == BTRFS_TREE_LOG_OBJECTID)
return 1;
return 0;
}
2008-09-26 14:09:34 +00:00
static int noinline __next_ref_path(struct btrfs_trans_handle *trans,
struct btrfs_root *extent_root,
struct btrfs_ref_path *ref_path,
int first_time)
{
struct extent_buffer *leaf;
struct btrfs_path *path;
struct btrfs_extent_ref *ref;
struct btrfs_key key;
struct btrfs_key found_key;
u64 bytenr;
u32 nritems;
int level;
int ret = 1;
2008-09-26 14:09:34 +00:00
path = btrfs_alloc_path();
if (!path)
return -ENOMEM;
2008-09-26 14:09:34 +00:00
if (first_time) {
ref_path->lowest_level = -1;
ref_path->current_level = -1;
ref_path->shared_level = -1;
2008-09-26 14:09:34 +00:00
goto walk_up;
}
walk_down:
level = ref_path->current_level - 1;
while (level >= -1) {
u64 parent;
if (level < ref_path->lowest_level)
break;
2008-09-26 14:09:34 +00:00
if (level >= 0) {
bytenr = ref_path->nodes[level];
} else {
bytenr = ref_path->extent_start;
}
BUG_ON(bytenr == 0);
2008-09-26 14:09:34 +00:00
parent = ref_path->nodes[level + 1];
ref_path->nodes[level + 1] = 0;
ref_path->current_level = level;
BUG_ON(parent == 0);
2008-09-26 14:09:34 +00:00
key.objectid = bytenr;
key.offset = parent + 1;
key.type = BTRFS_EXTENT_REF_KEY;
2008-09-26 14:09:34 +00:00
ret = btrfs_search_slot(trans, extent_root, &key, path, 0, 0);
if (ret < 0)
goto out;
2008-09-26 14:09:34 +00:00
BUG_ON(ret == 0);
2008-09-26 14:09:34 +00:00
leaf = path->nodes[0];
nritems = btrfs_header_nritems(leaf);
if (path->slots[0] >= nritems) {
ret = btrfs_next_leaf(extent_root, path);
if (ret < 0)
goto out;
if (ret > 0)
goto next;
leaf = path->nodes[0];
}
2008-09-26 14:09:34 +00:00
btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
if (found_key.objectid == bytenr &&
found_key.type == BTRFS_EXTENT_REF_KEY) {
if (level < ref_path->shared_level)
ref_path->shared_level = level;
2008-09-26 14:09:34 +00:00
goto found;
}
2008-09-26 14:09:34 +00:00
next:
level--;
btrfs_release_path(extent_root, path);
cond_resched();
2008-09-26 14:09:34 +00:00
}
/* reached lowest level */
ret = 1;
goto out;
walk_up:
level = ref_path->current_level;
while (level < BTRFS_MAX_LEVEL - 1) {
u64 ref_objectid;
if (level >= 0) {
bytenr = ref_path->nodes[level];
} else {
bytenr = ref_path->extent_start;
}
BUG_ON(bytenr == 0);
2008-09-26 14:09:34 +00:00
key.objectid = bytenr;
key.offset = 0;
key.type = BTRFS_EXTENT_REF_KEY;
2008-09-26 14:09:34 +00:00
ret = btrfs_search_slot(trans, extent_root, &key, path, 0, 0);
if (ret < 0)
goto out;
2008-09-26 14:09:34 +00:00
leaf = path->nodes[0];
nritems = btrfs_header_nritems(leaf);
if (path->slots[0] >= nritems) {
ret = btrfs_next_leaf(extent_root, path);
if (ret < 0)
goto out;
if (ret > 0) {
/* the extent was freed by someone */
if (ref_path->lowest_level == level)
goto out;
btrfs_release_path(extent_root, path);
goto walk_down;
}
leaf = path->nodes[0];
}
2008-09-26 14:09:34 +00:00
btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
if (found_key.objectid != bytenr ||
found_key.type != BTRFS_EXTENT_REF_KEY) {
/* the extent was freed by someone */
if (ref_path->lowest_level == level) {
ret = 1;
goto out;
}
btrfs_release_path(extent_root, path);
goto walk_down;
}
found:
ref = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_extent_ref);
ref_objectid = btrfs_ref_objectid(leaf, ref);
if (ref_objectid < BTRFS_FIRST_FREE_OBJECTID) {
if (first_time) {
level = (int)ref_objectid;
BUG_ON(level >= BTRFS_MAX_LEVEL);
ref_path->lowest_level = level;
ref_path->current_level = level;
ref_path->nodes[level] = bytenr;
} else {
WARN_ON(ref_objectid != level);
}
} else {
WARN_ON(level != -1);
}
first_time = 0;
2008-09-26 14:09:34 +00:00
if (ref_path->lowest_level == level) {
ref_path->owner_objectid = ref_objectid;
ref_path->num_refs = btrfs_ref_num_refs(leaf, ref);
}
/*
2008-09-26 14:09:34 +00:00
* the block is tree root or the block isn't in reference
* counted tree.
*/
2008-09-26 14:09:34 +00:00
if (found_key.objectid == found_key.offset ||
is_cowonly_root(btrfs_ref_root(leaf, ref))) {
ref_path->root_objectid = btrfs_ref_root(leaf, ref);
ref_path->root_generation =
btrfs_ref_generation(leaf, ref);
if (level < 0) {
/* special reference from the tree log */
ref_path->nodes[0] = found_key.offset;
ref_path->current_level = 0;
}
ret = 0;
goto out;
}
2008-09-26 14:09:34 +00:00
level++;
BUG_ON(ref_path->nodes[level] != 0);
ref_path->nodes[level] = found_key.offset;
ref_path->current_level = level;
2008-09-26 14:09:34 +00:00
/*
* the reference was created in the running transaction,
* no need to continue walking up.
*/
if (btrfs_ref_generation(leaf, ref) == trans->transid) {
ref_path->root_objectid = btrfs_ref_root(leaf, ref);
ref_path->root_generation =
btrfs_ref_generation(leaf, ref);
ret = 0;
goto out;
}
2008-09-26 14:09:34 +00:00
btrfs_release_path(extent_root, path);
cond_resched();
}
2008-09-26 14:09:34 +00:00
/* reached max tree level, but no tree root found. */
BUG();
out:
2008-09-26 14:09:34 +00:00
btrfs_free_path(path);
return ret;
}
2008-09-26 14:09:34 +00:00
static int btrfs_first_ref_path(struct btrfs_trans_handle *trans,
struct btrfs_root *extent_root,
struct btrfs_ref_path *ref_path,
u64 extent_start)
{
2008-09-26 14:09:34 +00:00
memset(ref_path, 0, sizeof(*ref_path));
ref_path->extent_start = extent_start;
2008-09-26 14:09:34 +00:00
return __next_ref_path(trans, extent_root, ref_path, 1);
}
2008-09-26 14:09:34 +00:00
static int btrfs_next_ref_path(struct btrfs_trans_handle *trans,
struct btrfs_root *extent_root,
struct btrfs_ref_path *ref_path)
{
2008-09-26 14:09:34 +00:00
return __next_ref_path(trans, extent_root, ref_path, 0);
}
static int noinline get_new_locations(struct inode *reloc_inode,
struct btrfs_key *extent_key,
u64 offset, int no_fragment,
struct disk_extent **extents,
int *nr_extents)
{
struct btrfs_root *root = BTRFS_I(reloc_inode)->root;
struct btrfs_path *path;
struct btrfs_file_extent_item *fi;
struct extent_buffer *leaf;
2008-09-26 14:09:34 +00:00
struct disk_extent *exts = *extents;
struct btrfs_key found_key;
u64 cur_pos;
u64 last_byte;
u32 nritems;
2008-09-26 14:09:34 +00:00
int nr = 0;
int max = *nr_extents;
int ret;
2008-09-26 14:09:34 +00:00
WARN_ON(!no_fragment && *extents);
if (!exts) {
max = 1;
exts = kmalloc(sizeof(*exts) * max, GFP_NOFS);
if (!exts)
return -ENOMEM;
}
2008-09-26 14:09:34 +00:00
path = btrfs_alloc_path();
BUG_ON(!path);
2008-09-26 14:09:34 +00:00
cur_pos = extent_key->objectid - offset;
last_byte = extent_key->objectid + extent_key->offset;
ret = btrfs_lookup_file_extent(NULL, root, path, reloc_inode->i_ino,
cur_pos, 0);
if (ret < 0)
goto out;
if (ret > 0) {
ret = -ENOENT;
goto out;
}
2008-09-26 14:09:34 +00:00
while (1) {
leaf = path->nodes[0];
nritems = btrfs_header_nritems(leaf);
2008-09-26 14:09:34 +00:00
if (path->slots[0] >= nritems) {
ret = btrfs_next_leaf(root, path);
if (ret < 0)
goto out;
2008-09-26 14:09:34 +00:00
if (ret > 0)
break;
leaf = path->nodes[0];
}
btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
2008-09-26 14:09:34 +00:00
if (found_key.offset != cur_pos ||
found_key.type != BTRFS_EXTENT_DATA_KEY ||
found_key.objectid != reloc_inode->i_ino)
break;
2008-09-26 14:09:34 +00:00
fi = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_file_extent_item);
if (btrfs_file_extent_type(leaf, fi) !=
BTRFS_FILE_EXTENT_REG ||
btrfs_file_extent_disk_bytenr(leaf, fi) == 0)
break;
2008-09-26 14:09:34 +00:00
if (nr == max) {
struct disk_extent *old = exts;
max *= 2;
exts = kzalloc(sizeof(*exts) * max, GFP_NOFS);
memcpy(exts, old, sizeof(*exts) * nr);
if (old != *extents)
kfree(old);
}
2008-09-26 14:09:34 +00:00
exts[nr].disk_bytenr =
btrfs_file_extent_disk_bytenr(leaf, fi);
exts[nr].disk_num_bytes =
btrfs_file_extent_disk_num_bytes(leaf, fi);
exts[nr].offset = btrfs_file_extent_offset(leaf, fi);
exts[nr].num_bytes = btrfs_file_extent_num_bytes(leaf, fi);
Btrfs: Add zlib compression support This is a large change for adding compression on reading and writing, both for inline and regular extents. It does some fairly large surgery to the writeback paths. Compression is off by default and enabled by mount -o compress. Even when the -o compress mount option is not used, it is possible to read compressed extents off the disk. If compression for a given set of pages fails to make them smaller, the file is flagged to avoid future compression attempts later. * While finding delalloc extents, the pages are locked before being sent down to the delalloc handler. This allows the delalloc handler to do complex things such as cleaning the pages, marking them writeback and starting IO on their behalf. * Inline extents are inserted at delalloc time now. This allows us to compress the data before inserting the inline extent, and it allows us to insert an inline extent that spans multiple pages. * All of the in-memory extent representations (extent_map.c, ordered-data.c etc) are changed to record both an in-memory size and an on disk size, as well as a flag for compression. From a disk format point of view, the extent pointers in the file are changed to record the on disk size of a given extent and some encoding flags. Space in the disk format is allocated for compression encoding, as well as encryption and a generic 'other' field. Neither the encryption or the 'other' field are currently used. In order to limit the amount of data read for a single random read in the file, the size of a compressed extent is limited to 128k. This is a software only limit, the disk format supports u64 sized compressed extents. In order to limit the ram consumed while processing extents, the uncompressed size of a compressed extent is limited to 256k. This is a software only limit and will be subject to tuning later. Checksumming is still done on compressed extents, and it is done on the uncompressed version of the data. This way additional encodings can be layered on without having to figure out which encoding to checksum. Compression happens at delalloc time, which is basically singled threaded because it is usually done by a single pdflush thread. This makes it tricky to spread the compression load across all the cpus on the box. We'll have to look at parallel pdflush walks of dirty inodes at a later time. Decompression is hooked into readpages and it does spread across CPUs nicely. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-29 18:49:59 +00:00
exts[nr].ram_bytes = btrfs_file_extent_ram_bytes(leaf, fi);
exts[nr].compression = btrfs_file_extent_compression(leaf, fi);
exts[nr].encryption = btrfs_file_extent_encryption(leaf, fi);
exts[nr].other_encoding = btrfs_file_extent_other_encoding(leaf,
fi);
BUG_ON(exts[nr].offset > 0);
BUG_ON(exts[nr].compression || exts[nr].encryption);
BUG_ON(exts[nr].num_bytes != exts[nr].disk_num_bytes);
2008-09-26 14:09:34 +00:00
cur_pos += exts[nr].num_bytes;
nr++;
if (cur_pos + offset >= last_byte)
break;
if (no_fragment) {
ret = 1;
goto out;
2008-09-26 14:09:34 +00:00
}
path->slots[0]++;
}
WARN_ON(cur_pos + offset > last_byte);
if (cur_pos + offset < last_byte) {
ret = -ENOENT;
goto out;
}
ret = 0;
out:
2008-09-26 14:09:34 +00:00
btrfs_free_path(path);
if (ret) {
if (exts != *extents)
kfree(exts);
} else {
*extents = exts;
*nr_extents = nr;
}
return ret;
}
static int noinline replace_one_extent(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct btrfs_path *path,
struct btrfs_key *extent_key,
struct btrfs_key *leaf_key,
struct btrfs_ref_path *ref_path,
struct disk_extent *new_extents,
int nr_extents)
{
struct extent_buffer *leaf;
struct btrfs_file_extent_item *fi;
struct inode *inode = NULL;
struct btrfs_key key;
u64 lock_start = 0;
u64 lock_end = 0;
u64 num_bytes;
u64 ext_offset;
u64 first_pos;
u32 nritems;
int nr_scaned = 0;
2008-09-26 14:09:34 +00:00
int extent_locked = 0;
int extent_type;
2008-09-26 14:09:34 +00:00
int ret;
memcpy(&key, leaf_key, sizeof(key));
first_pos = INT_LIMIT(loff_t) - extent_key->offset;
2008-09-26 14:09:34 +00:00
if (ref_path->owner_objectid != BTRFS_MULTIPLE_OBJECTIDS) {
if (key.objectid < ref_path->owner_objectid ||
(key.objectid == ref_path->owner_objectid &&
key.type < BTRFS_EXTENT_DATA_KEY)) {
key.objectid = ref_path->owner_objectid;
key.type = BTRFS_EXTENT_DATA_KEY;
key.offset = 0;
}
2008-09-26 14:09:34 +00:00
}
while (1) {
ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
if (ret < 0)
goto out;
leaf = path->nodes[0];
nritems = btrfs_header_nritems(leaf);
next:
if (extent_locked && ret > 0) {
/*
* the file extent item was modified by someone
* before the extent got locked.
*/
unlock_extent(&BTRFS_I(inode)->io_tree, lock_start,
lock_end, GFP_NOFS);
extent_locked = 0;
}
if (path->slots[0] >= nritems) {
if (++nr_scaned > 2)
2008-09-26 14:09:34 +00:00
break;
BUG_ON(extent_locked);
ret = btrfs_next_leaf(root, path);
if (ret < 0)
goto out;
if (ret > 0)
break;
leaf = path->nodes[0];
nritems = btrfs_header_nritems(leaf);
}
btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
if (ref_path->owner_objectid != BTRFS_MULTIPLE_OBJECTIDS) {
if ((key.objectid > ref_path->owner_objectid) ||
(key.objectid == ref_path->owner_objectid &&
key.type > BTRFS_EXTENT_DATA_KEY) ||
(key.offset >= first_pos + extent_key->offset))
break;
}
if (inode && key.objectid != inode->i_ino) {
BUG_ON(extent_locked);
btrfs_release_path(root, path);
mutex_unlock(&inode->i_mutex);
iput(inode);
inode = NULL;
continue;
}
if (key.type != BTRFS_EXTENT_DATA_KEY) {
path->slots[0]++;
ret = 1;
goto next;
}
fi = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_file_extent_item);
extent_type = btrfs_file_extent_type(leaf, fi);
if ((extent_type != BTRFS_FILE_EXTENT_REG &&
extent_type != BTRFS_FILE_EXTENT_PREALLOC) ||
2008-09-26 14:09:34 +00:00
(btrfs_file_extent_disk_bytenr(leaf, fi) !=
extent_key->objectid)) {
path->slots[0]++;
ret = 1;
goto next;
}
num_bytes = btrfs_file_extent_num_bytes(leaf, fi);
ext_offset = btrfs_file_extent_offset(leaf, fi);
if (first_pos > key.offset - ext_offset)
first_pos = key.offset - ext_offset;
if (!extent_locked) {
lock_start = key.offset;
lock_end = lock_start + num_bytes - 1;
} else {
if (lock_start > key.offset ||
lock_end + 1 < key.offset + num_bytes) {
unlock_extent(&BTRFS_I(inode)->io_tree,
lock_start, lock_end, GFP_NOFS);
extent_locked = 0;
}
2008-09-26 14:09:34 +00:00
}
if (!inode) {
btrfs_release_path(root, path);
inode = btrfs_iget_locked(root->fs_info->sb,
key.objectid, root);
if (inode->i_state & I_NEW) {
BTRFS_I(inode)->root = root;
BTRFS_I(inode)->location.objectid =
key.objectid;
BTRFS_I(inode)->location.type =
BTRFS_INODE_ITEM_KEY;
BTRFS_I(inode)->location.offset = 0;
btrfs_read_locked_inode(inode);
unlock_new_inode(inode);
}
/*
* some code call btrfs_commit_transaction while
* holding the i_mutex, so we can't use mutex_lock
* here.
*/
if (is_bad_inode(inode) ||
!mutex_trylock(&inode->i_mutex)) {
iput(inode);
inode = NULL;
key.offset = (u64)-1;
goto skip;
}
}
if (!extent_locked) {
struct btrfs_ordered_extent *ordered;
btrfs_release_path(root, path);
lock_extent(&BTRFS_I(inode)->io_tree, lock_start,
lock_end, GFP_NOFS);
ordered = btrfs_lookup_first_ordered_extent(inode,
lock_end);
if (ordered &&
ordered->file_offset <= lock_end &&
ordered->file_offset + ordered->len > lock_start) {
unlock_extent(&BTRFS_I(inode)->io_tree,
lock_start, lock_end, GFP_NOFS);
btrfs_start_ordered_extent(inode, ordered, 1);
btrfs_put_ordered_extent(ordered);
key.offset += num_bytes;
goto skip;
}
if (ordered)
btrfs_put_ordered_extent(ordered);
extent_locked = 1;
continue;
}
if (nr_extents == 1) {
/* update extent pointer in place */
btrfs_set_file_extent_disk_bytenr(leaf, fi,
new_extents[0].disk_bytenr);
btrfs_set_file_extent_disk_num_bytes(leaf, fi,
new_extents[0].disk_num_bytes);
btrfs_mark_buffer_dirty(leaf);
btrfs_drop_extent_cache(inode, key.offset,
key.offset + num_bytes - 1, 0);
ret = btrfs_inc_extent_ref(trans, root,
new_extents[0].disk_bytenr,
new_extents[0].disk_num_bytes,
leaf->start,
root->root_key.objectid,
trans->transid,
key.objectid);
2008-09-26 14:09:34 +00:00
BUG_ON(ret);
ret = btrfs_free_extent(trans, root,
extent_key->objectid,
extent_key->offset,
leaf->start,
btrfs_header_owner(leaf),
btrfs_header_generation(leaf),
key.objectid, 0);
2008-09-26 14:09:34 +00:00
BUG_ON(ret);
btrfs_release_path(root, path);
key.offset += num_bytes;
} else {
BUG_ON(1);
#if 0
2008-09-26 14:09:34 +00:00
u64 alloc_hint;
u64 extent_len;
int i;
/*
* drop old extent pointer at first, then insert the
* new pointers one bye one
*/
btrfs_release_path(root, path);
ret = btrfs_drop_extents(trans, root, inode, key.offset,
key.offset + num_bytes,
key.offset, &alloc_hint);
BUG_ON(ret);
for (i = 0; i < nr_extents; i++) {
if (ext_offset >= new_extents[i].num_bytes) {
ext_offset -= new_extents[i].num_bytes;
continue;
}
extent_len = min(new_extents[i].num_bytes -
ext_offset, num_bytes);
ret = btrfs_insert_empty_item(trans, root,
path, &key,
sizeof(*fi));
BUG_ON(ret);
leaf = path->nodes[0];
fi = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_file_extent_item);
btrfs_set_file_extent_generation(leaf, fi,
trans->transid);
btrfs_set_file_extent_type(leaf, fi,
BTRFS_FILE_EXTENT_REG);
btrfs_set_file_extent_disk_bytenr(leaf, fi,
new_extents[i].disk_bytenr);
btrfs_set_file_extent_disk_num_bytes(leaf, fi,
new_extents[i].disk_num_bytes);
Btrfs: Add zlib compression support This is a large change for adding compression on reading and writing, both for inline and regular extents. It does some fairly large surgery to the writeback paths. Compression is off by default and enabled by mount -o compress. Even when the -o compress mount option is not used, it is possible to read compressed extents off the disk. If compression for a given set of pages fails to make them smaller, the file is flagged to avoid future compression attempts later. * While finding delalloc extents, the pages are locked before being sent down to the delalloc handler. This allows the delalloc handler to do complex things such as cleaning the pages, marking them writeback and starting IO on their behalf. * Inline extents are inserted at delalloc time now. This allows us to compress the data before inserting the inline extent, and it allows us to insert an inline extent that spans multiple pages. * All of the in-memory extent representations (extent_map.c, ordered-data.c etc) are changed to record both an in-memory size and an on disk size, as well as a flag for compression. From a disk format point of view, the extent pointers in the file are changed to record the on disk size of a given extent and some encoding flags. Space in the disk format is allocated for compression encoding, as well as encryption and a generic 'other' field. Neither the encryption or the 'other' field are currently used. In order to limit the amount of data read for a single random read in the file, the size of a compressed extent is limited to 128k. This is a software only limit, the disk format supports u64 sized compressed extents. In order to limit the ram consumed while processing extents, the uncompressed size of a compressed extent is limited to 256k. This is a software only limit and will be subject to tuning later. Checksumming is still done on compressed extents, and it is done on the uncompressed version of the data. This way additional encodings can be layered on without having to figure out which encoding to checksum. Compression happens at delalloc time, which is basically singled threaded because it is usually done by a single pdflush thread. This makes it tricky to spread the compression load across all the cpus on the box. We'll have to look at parallel pdflush walks of dirty inodes at a later time. Decompression is hooked into readpages and it does spread across CPUs nicely. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-29 18:49:59 +00:00
btrfs_set_file_extent_ram_bytes(leaf, fi,
new_extents[i].ram_bytes);
btrfs_set_file_extent_compression(leaf, fi,
new_extents[i].compression);
btrfs_set_file_extent_encryption(leaf, fi,
new_extents[i].encryption);
btrfs_set_file_extent_other_encoding(leaf, fi,
new_extents[i].other_encoding);
2008-09-26 14:09:34 +00:00
btrfs_set_file_extent_num_bytes(leaf, fi,
extent_len);
ext_offset += new_extents[i].offset;
btrfs_set_file_extent_offset(leaf, fi,
ext_offset);
btrfs_mark_buffer_dirty(leaf);
btrfs_drop_extent_cache(inode, key.offset,
key.offset + extent_len - 1, 0);
ret = btrfs_inc_extent_ref(trans, root,
new_extents[i].disk_bytenr,
new_extents[i].disk_num_bytes,
leaf->start,
root->root_key.objectid,
trans->transid, key.objectid);
2008-09-26 14:09:34 +00:00
BUG_ON(ret);
btrfs_release_path(root, path);
inode_add_bytes(inode, extent_len);
2008-09-26 14:09:34 +00:00
ext_offset = 0;
num_bytes -= extent_len;
key.offset += extent_len;
if (num_bytes == 0)
break;
}
BUG_ON(i >= nr_extents);
#endif
2008-09-26 14:09:34 +00:00
}
if (extent_locked) {
unlock_extent(&BTRFS_I(inode)->io_tree, lock_start,
lock_end, GFP_NOFS);
extent_locked = 0;
}
skip:
if (ref_path->owner_objectid != BTRFS_MULTIPLE_OBJECTIDS &&
key.offset >= first_pos + extent_key->offset)
break;
cond_resched();
}
ret = 0;
out:
btrfs_release_path(root, path);
if (inode) {
mutex_unlock(&inode->i_mutex);
if (extent_locked) {
unlock_extent(&BTRFS_I(inode)->io_tree, lock_start,
lock_end, GFP_NOFS);
}
iput(inode);
}
return ret;
}
int btrfs_reloc_tree_cache_ref(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct extent_buffer *buf, u64 orig_start)
{
int level;
int ret;
BUG_ON(btrfs_header_generation(buf) != trans->transid);
BUG_ON(root->root_key.objectid != BTRFS_TREE_RELOC_OBJECTID);
level = btrfs_header_level(buf);
if (level == 0) {
struct btrfs_leaf_ref *ref;
struct btrfs_leaf_ref *orig_ref;
orig_ref = btrfs_lookup_leaf_ref(root, orig_start);
if (!orig_ref)
return -ENOENT;
ref = btrfs_alloc_leaf_ref(root, orig_ref->nritems);
if (!ref) {
btrfs_free_leaf_ref(root, orig_ref);
return -ENOMEM;
}
ref->nritems = orig_ref->nritems;
memcpy(ref->extents, orig_ref->extents,
sizeof(ref->extents[0]) * ref->nritems);
btrfs_free_leaf_ref(root, orig_ref);
ref->root_gen = trans->transid;
ref->bytenr = buf->start;
ref->owner = btrfs_header_owner(buf);
ref->generation = btrfs_header_generation(buf);
ret = btrfs_add_leaf_ref(root, ref, 0);
WARN_ON(ret);
btrfs_free_leaf_ref(root, ref);
}
return 0;
}
static int noinline invalidate_extent_cache(struct btrfs_root *root,
struct extent_buffer *leaf,
struct btrfs_block_group_cache *group,
struct btrfs_root *target_root)
{
struct btrfs_key key;
struct inode *inode = NULL;
struct btrfs_file_extent_item *fi;
u64 num_bytes;
u64 skip_objectid = 0;
u32 nritems;
u32 i;
nritems = btrfs_header_nritems(leaf);
for (i = 0; i < nritems; i++) {
btrfs_item_key_to_cpu(leaf, &key, i);
if (key.objectid == skip_objectid ||
key.type != BTRFS_EXTENT_DATA_KEY)
continue;
fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
if (btrfs_file_extent_type(leaf, fi) ==
BTRFS_FILE_EXTENT_INLINE)
continue;
if (btrfs_file_extent_disk_bytenr(leaf, fi) == 0)
continue;
if (!inode || inode->i_ino != key.objectid) {
iput(inode);
inode = btrfs_ilookup(target_root->fs_info->sb,
key.objectid, target_root, 1);
}
if (!inode) {
skip_objectid = key.objectid;
continue;
}
num_bytes = btrfs_file_extent_num_bytes(leaf, fi);
lock_extent(&BTRFS_I(inode)->io_tree, key.offset,
key.offset + num_bytes - 1, GFP_NOFS);
btrfs_drop_extent_cache(inode, key.offset,
key.offset + num_bytes - 1, 1);
unlock_extent(&BTRFS_I(inode)->io_tree, key.offset,
key.offset + num_bytes - 1, GFP_NOFS);
cond_resched();
}
iput(inode);
return 0;
}
static int noinline replace_extents_in_leaf(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct extent_buffer *leaf,
struct btrfs_block_group_cache *group,
struct inode *reloc_inode)
{
struct btrfs_key key;
struct btrfs_key extent_key;
struct btrfs_file_extent_item *fi;
struct btrfs_leaf_ref *ref;
struct disk_extent *new_extent;
u64 bytenr;
u64 num_bytes;
u32 nritems;
u32 i;
int ext_index;
int nr_extent;
int ret;
new_extent = kmalloc(sizeof(*new_extent), GFP_NOFS);
BUG_ON(!new_extent);
ref = btrfs_lookup_leaf_ref(root, leaf->start);
BUG_ON(!ref);
ext_index = -1;
nritems = btrfs_header_nritems(leaf);
for (i = 0; i < nritems; i++) {
btrfs_item_key_to_cpu(leaf, &key, i);
if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
continue;
fi = btrfs_item_ptr(leaf, i, struct btrfs_file_extent_item);
if (btrfs_file_extent_type(leaf, fi) ==
BTRFS_FILE_EXTENT_INLINE)
continue;
bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
num_bytes = btrfs_file_extent_disk_num_bytes(leaf, fi);
if (bytenr == 0)
continue;
ext_index++;
if (bytenr >= group->key.objectid + group->key.offset ||
bytenr + num_bytes <= group->key.objectid)
continue;
extent_key.objectid = bytenr;
extent_key.offset = num_bytes;
extent_key.type = BTRFS_EXTENT_ITEM_KEY;
nr_extent = 1;
ret = get_new_locations(reloc_inode, &extent_key,
group->key.objectid, 1,
&new_extent, &nr_extent);
if (ret > 0)
continue;
BUG_ON(ret < 0);
BUG_ON(ref->extents[ext_index].bytenr != bytenr);
BUG_ON(ref->extents[ext_index].num_bytes != num_bytes);
ref->extents[ext_index].bytenr = new_extent->disk_bytenr;
ref->extents[ext_index].num_bytes = new_extent->disk_num_bytes;
btrfs_set_file_extent_disk_bytenr(leaf, fi,
new_extent->disk_bytenr);
btrfs_set_file_extent_disk_num_bytes(leaf, fi,
new_extent->disk_num_bytes);
btrfs_mark_buffer_dirty(leaf);
ret = btrfs_inc_extent_ref(trans, root,
new_extent->disk_bytenr,
new_extent->disk_num_bytes,
leaf->start,
root->root_key.objectid,
trans->transid, key.objectid);
2008-09-26 14:09:34 +00:00
BUG_ON(ret);
ret = btrfs_free_extent(trans, root,
bytenr, num_bytes, leaf->start,
btrfs_header_owner(leaf),
btrfs_header_generation(leaf),
key.objectid, 0);
2008-09-26 14:09:34 +00:00
BUG_ON(ret);
cond_resched();
}
kfree(new_extent);
BUG_ON(ext_index + 1 != ref->nritems);
btrfs_free_leaf_ref(root, ref);
return 0;
}
int btrfs_free_reloc_root(struct btrfs_trans_handle *trans,
struct btrfs_root *root)
2008-09-26 14:09:34 +00:00
{
struct btrfs_root *reloc_root;
int ret;
2008-09-26 14:09:34 +00:00
if (root->reloc_root) {
reloc_root = root->reloc_root;
root->reloc_root = NULL;
list_add(&reloc_root->dead_list,
&root->fs_info->dead_reloc_roots);
btrfs_set_root_bytenr(&reloc_root->root_item,
reloc_root->node->start);
btrfs_set_root_level(&root->root_item,
btrfs_header_level(reloc_root->node));
memset(&reloc_root->root_item.drop_progress, 0,
sizeof(struct btrfs_disk_key));
reloc_root->root_item.drop_level = 0;
ret = btrfs_update_root(trans, root->fs_info->tree_root,
&reloc_root->root_key,
&reloc_root->root_item);
BUG_ON(ret);
2008-09-26 14:09:34 +00:00
}
return 0;
}
int btrfs_drop_dead_reloc_roots(struct btrfs_root *root)
{
struct btrfs_trans_handle *trans;
struct btrfs_root *reloc_root;
struct btrfs_root *prev_root = NULL;
struct list_head dead_roots;
int ret;
unsigned long nr;
INIT_LIST_HEAD(&dead_roots);
list_splice_init(&root->fs_info->dead_reloc_roots, &dead_roots);
while (!list_empty(&dead_roots)) {
reloc_root = list_entry(dead_roots.prev,
struct btrfs_root, dead_list);
list_del_init(&reloc_root->dead_list);
BUG_ON(reloc_root->commit_root != NULL);
while (1) {
trans = btrfs_join_transaction(root, 1);
BUG_ON(!trans);
mutex_lock(&root->fs_info->drop_mutex);
ret = btrfs_drop_snapshot(trans, reloc_root);
if (ret != -EAGAIN)
break;
mutex_unlock(&root->fs_info->drop_mutex);
nr = trans->blocks_used;
ret = btrfs_end_transaction(trans, root);
BUG_ON(ret);
btrfs_btree_balance_dirty(root, nr);
}
free_extent_buffer(reloc_root->node);
ret = btrfs_del_root(trans, root->fs_info->tree_root,
&reloc_root->root_key);
BUG_ON(ret);
mutex_unlock(&root->fs_info->drop_mutex);
nr = trans->blocks_used;
ret = btrfs_end_transaction(trans, root);
BUG_ON(ret);
btrfs_btree_balance_dirty(root, nr);
kfree(prev_root);
prev_root = reloc_root;
}
if (prev_root) {
btrfs_remove_leaf_refs(prev_root, (u64)-1, 0);
kfree(prev_root);
}
return 0;
}
int btrfs_add_dead_reloc_root(struct btrfs_root *root)
{
list_add(&root->dead_list, &root->fs_info->dead_reloc_roots);
return 0;
}
int btrfs_cleanup_reloc_trees(struct btrfs_root *root)
{
struct btrfs_root *reloc_root;
struct btrfs_trans_handle *trans;
struct btrfs_key location;
int found;
int ret;
mutex_lock(&root->fs_info->tree_reloc_mutex);
ret = btrfs_find_dead_roots(root, BTRFS_TREE_RELOC_OBJECTID, NULL);
BUG_ON(ret);
found = !list_empty(&root->fs_info->dead_reloc_roots);
mutex_unlock(&root->fs_info->tree_reloc_mutex);
if (found) {
trans = btrfs_start_transaction(root, 1);
BUG_ON(!trans);
ret = btrfs_commit_transaction(trans, root);
BUG_ON(ret);
}
location.objectid = BTRFS_DATA_RELOC_TREE_OBJECTID;
location.offset = (u64)-1;
location.type = BTRFS_ROOT_ITEM_KEY;
reloc_root = btrfs_read_fs_root_no_name(root->fs_info, &location);
BUG_ON(!reloc_root);
btrfs_orphan_cleanup(reloc_root);
return 0;
}
static int noinline init_reloc_tree(struct btrfs_trans_handle *trans,
struct btrfs_root *root)
{
struct btrfs_root *reloc_root;
struct extent_buffer *eb;
struct btrfs_root_item *root_item;
struct btrfs_key root_key;
int ret;
BUG_ON(!root->ref_cows);
if (root->reloc_root)
return 0;
root_item = kmalloc(sizeof(*root_item), GFP_NOFS);
BUG_ON(!root_item);
ret = btrfs_copy_root(trans, root, root->commit_root,
&eb, BTRFS_TREE_RELOC_OBJECTID);
BUG_ON(ret);
root_key.objectid = BTRFS_TREE_RELOC_OBJECTID;
root_key.offset = root->root_key.objectid;
root_key.type = BTRFS_ROOT_ITEM_KEY;
memcpy(root_item, &root->root_item, sizeof(root_item));
btrfs_set_root_refs(root_item, 0);
btrfs_set_root_bytenr(root_item, eb->start);
btrfs_set_root_level(root_item, btrfs_header_level(eb));
btrfs_set_root_generation(root_item, trans->transid);
2008-09-26 14:09:34 +00:00
btrfs_tree_unlock(eb);
free_extent_buffer(eb);
ret = btrfs_insert_root(trans, root->fs_info->tree_root,
&root_key, root_item);
BUG_ON(ret);
kfree(root_item);
reloc_root = btrfs_read_fs_root_no_radix(root->fs_info->tree_root,
&root_key);
BUG_ON(!reloc_root);
reloc_root->last_trans = trans->transid;
reloc_root->commit_root = NULL;
reloc_root->ref_tree = &root->fs_info->reloc_ref_tree;
root->reloc_root = reloc_root;
return 0;
}
/*
* Core function of space balance.
*
* The idea is using reloc trees to relocate tree blocks in reference
* counted roots. There is one reloc tree for each subvol, and all
* reloc trees share same root key objectid. Reloc trees are snapshots
* of the latest committed roots of subvols (root->commit_root).
*
* To relocate a tree block referenced by a subvol, there are two steps.
* COW the block through subvol's reloc tree, then update block pointer
* in the subvol to point to the new block. Since all reloc trees share
* same root key objectid, doing special handing for tree blocks owned
* by them is easy. Once a tree block has been COWed in one reloc tree,
* we can use the resulting new block directly when the same block is
* required to COW again through other reloc trees. By this way, relocated
* tree blocks are shared between reloc trees, so they are also shared
* between subvols.
2008-09-26 14:09:34 +00:00
*/
static int noinline relocate_one_path(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct btrfs_path *path,
struct btrfs_key *first_key,
struct btrfs_ref_path *ref_path,
struct btrfs_block_group_cache *group,
struct inode *reloc_inode)
{
struct btrfs_root *reloc_root;
struct extent_buffer *eb = NULL;
struct btrfs_key *keys;
u64 *nodes;
int level;
int shared_level;
2008-09-26 14:09:34 +00:00
int lowest_level = 0;
int ret;
if (ref_path->owner_objectid < BTRFS_FIRST_FREE_OBJECTID)
lowest_level = ref_path->owner_objectid;
if (!root->ref_cows) {
2008-09-26 14:09:34 +00:00
path->lowest_level = lowest_level;
ret = btrfs_search_slot(trans, root, first_key, path, 0, 1);
BUG_ON(ret < 0);
path->lowest_level = 0;
btrfs_release_path(root, path);
return 0;
}
mutex_lock(&root->fs_info->tree_reloc_mutex);
ret = init_reloc_tree(trans, root);
BUG_ON(ret);
reloc_root = root->reloc_root;
shared_level = ref_path->shared_level;
ref_path->shared_level = BTRFS_MAX_LEVEL - 1;
2008-09-26 14:09:34 +00:00
keys = ref_path->node_keys;
nodes = ref_path->new_nodes;
memset(&keys[shared_level + 1], 0,
sizeof(*keys) * (BTRFS_MAX_LEVEL - shared_level - 1));
memset(&nodes[shared_level + 1], 0,
sizeof(*nodes) * (BTRFS_MAX_LEVEL - shared_level - 1));
2008-09-26 14:09:34 +00:00
if (nodes[lowest_level] == 0) {
path->lowest_level = lowest_level;
ret = btrfs_search_slot(trans, reloc_root, first_key, path,
0, 1);
BUG_ON(ret);
for (level = lowest_level; level < BTRFS_MAX_LEVEL; level++) {
eb = path->nodes[level];
if (!eb || eb == reloc_root->node)
break;
nodes[level] = eb->start;
if (level == 0)
btrfs_item_key_to_cpu(eb, &keys[level], 0);
else
btrfs_node_key_to_cpu(eb, &keys[level], 0);
}
if (nodes[0] &&
ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
eb = path->nodes[0];
ret = replace_extents_in_leaf(trans, reloc_root, eb,
group, reloc_inode);
BUG_ON(ret);
}
btrfs_release_path(reloc_root, path);
} else {
2008-09-26 14:09:34 +00:00
ret = btrfs_merge_path(trans, reloc_root, keys, nodes,
lowest_level);
2008-09-26 14:09:34 +00:00
BUG_ON(ret);
}
/*
* replace tree blocks in the fs tree with tree blocks in
* the reloc tree.
*/
ret = btrfs_merge_path(trans, root, keys, nodes, lowest_level);
BUG_ON(ret < 0);
if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
ret = btrfs_search_slot(trans, reloc_root, first_key, path,
0, 0);
BUG_ON(ret);
extent_buffer_get(path->nodes[0]);
eb = path->nodes[0];
btrfs_release_path(reloc_root, path);
2008-09-26 14:09:34 +00:00
ret = invalidate_extent_cache(reloc_root, eb, group, root);
BUG_ON(ret);
free_extent_buffer(eb);
}
mutex_unlock(&root->fs_info->tree_reloc_mutex);
2008-09-26 14:09:34 +00:00
path->lowest_level = 0;
return 0;
}
static int noinline relocate_tree_block(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
struct btrfs_path *path,
struct btrfs_key *first_key,
struct btrfs_ref_path *ref_path)
{
int ret;
ret = relocate_one_path(trans, root, path, first_key,
ref_path, NULL, NULL);
BUG_ON(ret);
if (root == root->fs_info->extent_root)
btrfs_extent_post_op(trans, root);
return 0;
}
static int noinline del_extent_zero(struct btrfs_trans_handle *trans,
struct btrfs_root *extent_root,
struct btrfs_path *path,
struct btrfs_key *extent_key)
{
int ret;
ret = btrfs_search_slot(trans, extent_root, extent_key, path, -1, 1);
if (ret)
goto out;
ret = btrfs_del_item(trans, extent_root, path);
out:
btrfs_release_path(extent_root, path);
return ret;
}
static struct btrfs_root noinline *read_ref_root(struct btrfs_fs_info *fs_info,
struct btrfs_ref_path *ref_path)
{
struct btrfs_key root_key;
root_key.objectid = ref_path->root_objectid;
root_key.type = BTRFS_ROOT_ITEM_KEY;
if (is_cowonly_root(ref_path->root_objectid))
root_key.offset = 0;
else
root_key.offset = (u64)-1;
return btrfs_read_fs_root_no_name(fs_info, &root_key);
}
static int noinline relocate_one_extent(struct btrfs_root *extent_root,
struct btrfs_path *path,
struct btrfs_key *extent_key,
struct btrfs_block_group_cache *group,
struct inode *reloc_inode, int pass)
{
struct btrfs_trans_handle *trans;
struct btrfs_root *found_root;
struct btrfs_ref_path *ref_path = NULL;
struct disk_extent *new_extents = NULL;
int nr_extents = 0;
int loops;
int ret;
int level;
struct btrfs_key first_key;
u64 prev_block = 0;
trans = btrfs_start_transaction(extent_root, 1);
BUG_ON(!trans);
if (extent_key->objectid == 0) {
ret = del_extent_zero(trans, extent_root, path, extent_key);
goto out;
}
ref_path = kmalloc(sizeof(*ref_path), GFP_NOFS);
if (!ref_path) {
ret = -ENOMEM;
goto out;
}
for (loops = 0; ; loops++) {
if (loops == 0) {
ret = btrfs_first_ref_path(trans, extent_root, ref_path,
extent_key->objectid);
} else {
ret = btrfs_next_ref_path(trans, extent_root, ref_path);
}
if (ret < 0)
goto out;
if (ret > 0)
break;
if (ref_path->root_objectid == BTRFS_TREE_LOG_OBJECTID ||
ref_path->root_objectid == BTRFS_TREE_RELOC_OBJECTID)
continue;
found_root = read_ref_root(extent_root->fs_info, ref_path);
BUG_ON(!found_root);
/*
* for reference counted tree, only process reference paths
* rooted at the latest committed root.
*/
if (found_root->ref_cows &&
ref_path->root_generation != found_root->root_key.offset)
continue;
if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID) {
if (pass == 0) {
/*
* copy data extents to new locations
*/
u64 group_start = group->key.objectid;
ret = relocate_data_extent(reloc_inode,
extent_key,
group_start);
if (ret < 0)
goto out;
break;
}
level = 0;
} else {
level = ref_path->owner_objectid;
}
if (prev_block != ref_path->nodes[level]) {
struct extent_buffer *eb;
u64 block_start = ref_path->nodes[level];
u64 block_size = btrfs_level_size(found_root, level);
eb = read_tree_block(found_root, block_start,
block_size, 0);
btrfs_tree_lock(eb);
BUG_ON(level != btrfs_header_level(eb));
if (level == 0)
btrfs_item_key_to_cpu(eb, &first_key, 0);
else
btrfs_node_key_to_cpu(eb, &first_key, 0);
btrfs_tree_unlock(eb);
free_extent_buffer(eb);
prev_block = block_start;
}
if (ref_path->owner_objectid >= BTRFS_FIRST_FREE_OBJECTID &&
pass >= 2) {
/*
* use fallback method to process the remaining
* references.
*/
if (!new_extents) {
u64 group_start = group->key.objectid;
new_extents = kmalloc(sizeof(*new_extents),
GFP_NOFS);
nr_extents = 1;
2008-09-26 14:09:34 +00:00
ret = get_new_locations(reloc_inode,
extent_key,
group_start, 1,
2008-09-26 14:09:34 +00:00
&new_extents,
&nr_extents);
if (ret)
2008-09-26 14:09:34 +00:00
goto out;
}
btrfs_record_root_in_trans(found_root);
ret = replace_one_extent(trans, found_root,
path, extent_key,
&first_key, ref_path,
new_extents, nr_extents);
if (ret < 0)
goto out;
continue;
}
btrfs_record_root_in_trans(found_root);
if (ref_path->owner_objectid < BTRFS_FIRST_FREE_OBJECTID) {
ret = relocate_tree_block(trans, found_root, path,
&first_key, ref_path);
} else {
/*
* try to update data extent references while
* keeping metadata shared between snapshots.
*/
ret = relocate_one_path(trans, found_root, path,
&first_key, ref_path,
group, reloc_inode);
}
if (ret < 0)
goto out;
}
ret = 0;
out:
btrfs_end_transaction(trans, extent_root);
kfree(new_extents);
kfree(ref_path);
return ret;
}
static u64 update_block_group_flags(struct btrfs_root *root, u64 flags)
{
u64 num_devices;
u64 stripped = BTRFS_BLOCK_GROUP_RAID0 |
BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID10;
num_devices = root->fs_info->fs_devices->rw_devices;
if (num_devices == 1) {
stripped |= BTRFS_BLOCK_GROUP_DUP;
stripped = flags & ~stripped;
/* turn raid0 into single device chunks */
if (flags & BTRFS_BLOCK_GROUP_RAID0)
return stripped;
/* turn mirroring into duplication */
if (flags & (BTRFS_BLOCK_GROUP_RAID1 |
BTRFS_BLOCK_GROUP_RAID10))
return stripped | BTRFS_BLOCK_GROUP_DUP;
return flags;
} else {
/* they already had raid on here, just return */
if (flags & stripped)
return flags;
stripped |= BTRFS_BLOCK_GROUP_DUP;
stripped = flags & ~stripped;
/* switch duplicated blocks with raid1 */
if (flags & BTRFS_BLOCK_GROUP_DUP)
return stripped | BTRFS_BLOCK_GROUP_RAID1;
/* turn single device chunks into raid0 */
return stripped | BTRFS_BLOCK_GROUP_RAID0;
}
return flags;
}
int __alloc_chunk_for_shrink(struct btrfs_root *root,
struct btrfs_block_group_cache *shrink_block_group,
int force)
{
struct btrfs_trans_handle *trans;
u64 new_alloc_flags;
u64 calc;
spin_lock(&shrink_block_group->lock);
if (btrfs_block_group_used(&shrink_block_group->item) > 0) {
spin_unlock(&shrink_block_group->lock);
trans = btrfs_start_transaction(root, 1);
spin_lock(&shrink_block_group->lock);
new_alloc_flags = update_block_group_flags(root,
shrink_block_group->flags);
if (new_alloc_flags != shrink_block_group->flags) {
calc =
btrfs_block_group_used(&shrink_block_group->item);
} else {
calc = shrink_block_group->key.offset;
}
spin_unlock(&shrink_block_group->lock);
do_chunk_alloc(trans, root->fs_info->extent_root,
calc + 2 * 1024 * 1024, new_alloc_flags, force);
btrfs_end_transaction(trans, root);
} else
spin_unlock(&shrink_block_group->lock);
return 0;
}
2008-09-26 14:09:34 +00:00
static int __insert_orphan_inode(struct btrfs_trans_handle *trans,
struct btrfs_root *root,
u64 objectid, u64 size)
{
struct btrfs_path *path;
struct btrfs_inode_item *item;
struct extent_buffer *leaf;
int ret;
path = btrfs_alloc_path();
if (!path)
return -ENOMEM;
ret = btrfs_insert_empty_inode(trans, root, path, objectid);
if (ret)
goto out;
leaf = path->nodes[0];
item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_inode_item);
memset_extent_buffer(leaf, 0, (unsigned long)item, sizeof(*item));
btrfs_set_inode_generation(leaf, item, 1);
btrfs_set_inode_size(leaf, item, size);
btrfs_set_inode_mode(leaf, item, S_IFREG | 0600);
btrfs_set_inode_flags(leaf, item, BTRFS_INODE_NODATASUM |
BTRFS_INODE_NOCOMPRESS);
2008-09-26 14:09:34 +00:00
btrfs_mark_buffer_dirty(leaf);
btrfs_release_path(root, path);
out:
btrfs_free_path(path);
return ret;
}
static struct inode noinline *create_reloc_inode(struct btrfs_fs_info *fs_info,
struct btrfs_block_group_cache *group)
{
struct inode *inode = NULL;
struct btrfs_trans_handle *trans;
struct btrfs_root *root;
struct btrfs_key root_key;
u64 objectid = BTRFS_FIRST_FREE_OBJECTID;
int err = 0;
root_key.objectid = BTRFS_DATA_RELOC_TREE_OBJECTID;
root_key.type = BTRFS_ROOT_ITEM_KEY;
root_key.offset = (u64)-1;
root = btrfs_read_fs_root_no_name(fs_info, &root_key);
if (IS_ERR(root))
return ERR_CAST(root);
trans = btrfs_start_transaction(root, 1);
BUG_ON(!trans);
err = btrfs_find_free_objectid(trans, root, objectid, &objectid);
if (err)
goto out;
err = __insert_orphan_inode(trans, root, objectid, group->key.offset);
BUG_ON(err);
err = btrfs_insert_file_extent(trans, root, objectid, 0, 0, 0,
Btrfs: Add zlib compression support This is a large change for adding compression on reading and writing, both for inline and regular extents. It does some fairly large surgery to the writeback paths. Compression is off by default and enabled by mount -o compress. Even when the -o compress mount option is not used, it is possible to read compressed extents off the disk. If compression for a given set of pages fails to make them smaller, the file is flagged to avoid future compression attempts later. * While finding delalloc extents, the pages are locked before being sent down to the delalloc handler. This allows the delalloc handler to do complex things such as cleaning the pages, marking them writeback and starting IO on their behalf. * Inline extents are inserted at delalloc time now. This allows us to compress the data before inserting the inline extent, and it allows us to insert an inline extent that spans multiple pages. * All of the in-memory extent representations (extent_map.c, ordered-data.c etc) are changed to record both an in-memory size and an on disk size, as well as a flag for compression. From a disk format point of view, the extent pointers in the file are changed to record the on disk size of a given extent and some encoding flags. Space in the disk format is allocated for compression encoding, as well as encryption and a generic 'other' field. Neither the encryption or the 'other' field are currently used. In order to limit the amount of data read for a single random read in the file, the size of a compressed extent is limited to 128k. This is a software only limit, the disk format supports u64 sized compressed extents. In order to limit the ram consumed while processing extents, the uncompressed size of a compressed extent is limited to 256k. This is a software only limit and will be subject to tuning later. Checksumming is still done on compressed extents, and it is done on the uncompressed version of the data. This way additional encodings can be layered on without having to figure out which encoding to checksum. Compression happens at delalloc time, which is basically singled threaded because it is usually done by a single pdflush thread. This makes it tricky to spread the compression load across all the cpus on the box. We'll have to look at parallel pdflush walks of dirty inodes at a later time. Decompression is hooked into readpages and it does spread across CPUs nicely. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-10-29 18:49:59 +00:00
group->key.offset, 0, group->key.offset,
0, 0, 0);
2008-09-26 14:09:34 +00:00
BUG_ON(err);
inode = btrfs_iget_locked(root->fs_info->sb, objectid, root);
if (inode->i_state & I_NEW) {
BTRFS_I(inode)->root = root;
BTRFS_I(inode)->location.objectid = objectid;
BTRFS_I(inode)->location.type = BTRFS_INODE_ITEM_KEY;
BTRFS_I(inode)->location.offset = 0;
btrfs_read_locked_inode(inode);
unlock_new_inode(inode);
BUG_ON(is_bad_inode(inode));
} else {
BUG_ON(1);
}
err = btrfs_orphan_add(trans, inode);
out:
btrfs_end_transaction(trans, root);
if (err) {
if (inode)
iput(inode);
inode = ERR_PTR(err);
}
return inode;
}
int btrfs_relocate_block_group(struct btrfs_root *root, u64 group_start)
{
struct btrfs_trans_handle *trans;
struct btrfs_path *path;
2008-09-26 14:09:34 +00:00
struct btrfs_fs_info *info = root->fs_info;
struct extent_buffer *leaf;
struct inode *reloc_inode;
struct btrfs_block_group_cache *block_group;
struct btrfs_key key;
u64 skipped;
u64 cur_byte;
u64 total_found;
u32 nritems;
int ret;
int progress;
2008-09-26 14:09:34 +00:00
int pass = 0;
2008-09-26 14:09:34 +00:00
root = root->fs_info->extent_root;
block_group = btrfs_lookup_block_group(info, group_start);
BUG_ON(!block_group);
2008-09-26 14:09:34 +00:00
printk("btrfs relocating block group %llu flags %llu\n",
(unsigned long long)block_group->key.objectid,
(unsigned long long)block_group->flags);
path = btrfs_alloc_path();
2008-09-26 14:09:34 +00:00
BUG_ON(!path);
2008-09-26 14:09:34 +00:00
reloc_inode = create_reloc_inode(info, block_group);
BUG_ON(IS_ERR(reloc_inode));
2008-09-26 14:09:34 +00:00
__alloc_chunk_for_shrink(root, block_group, 1);
set_block_group_readonly(block_group);
2008-09-26 14:09:34 +00:00
btrfs_start_delalloc_inodes(info->tree_root);
btrfs_wait_ordered_extents(info->tree_root, 0);
again:
skipped = 0;
total_found = 0;
progress = 0;
2008-09-26 14:09:34 +00:00
key.objectid = block_group->key.objectid;
key.offset = 0;
key.type = 0;
cur_byte = key.objectid;
2008-09-26 14:09:34 +00:00
trans = btrfs_start_transaction(info->tree_root, 1);
btrfs_commit_transaction(trans, info->tree_root);
2008-09-26 14:09:34 +00:00
mutex_lock(&root->fs_info->cleaner_mutex);
btrfs_clean_old_snapshots(info->tree_root);
btrfs_remove_leaf_refs(info->tree_root, (u64)-1, 1);
mutex_unlock(&root->fs_info->cleaner_mutex);
while(1) {
ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
if (ret < 0)
goto out;
next:
leaf = path->nodes[0];
nritems = btrfs_header_nritems(leaf);
if (path->slots[0] >= nritems) {
ret = btrfs_next_leaf(root, path);
if (ret < 0)
goto out;
if (ret == 1) {
ret = 0;
break;
}
leaf = path->nodes[0];
nritems = btrfs_header_nritems(leaf);
}
2008-09-26 14:09:34 +00:00
btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
2008-09-26 14:09:34 +00:00
if (key.objectid >= block_group->key.objectid +
block_group->key.offset)
break;
if (progress && need_resched()) {
btrfs_release_path(root, path);
2008-09-26 14:09:34 +00:00
cond_resched();
progress = 0;
2008-09-26 14:09:34 +00:00
continue;
}
progress = 1;
2008-09-26 14:09:34 +00:00
if (btrfs_key_type(&key) != BTRFS_EXTENT_ITEM_KEY ||
key.objectid + key.offset <= cur_byte) {
path->slots[0]++;
goto next;
}
total_found++;
2008-09-26 14:09:34 +00:00
cur_byte = key.objectid + key.offset;
btrfs_release_path(root, path);
2008-09-26 14:09:34 +00:00
__alloc_chunk_for_shrink(root, block_group, 0);
ret = relocate_one_extent(root, path, &key, block_group,
reloc_inode, pass);
BUG_ON(ret < 0);
if (ret > 0)
skipped++;
2008-09-26 14:09:34 +00:00
key.objectid = cur_byte;
key.type = 0;
key.offset = 0;
}
2008-09-26 14:09:34 +00:00
btrfs_release_path(root, path);
2008-09-26 14:09:34 +00:00
if (pass == 0) {
btrfs_wait_ordered_range(reloc_inode, 0, (u64)-1);
invalidate_mapping_pages(reloc_inode->i_mapping, 0, -1);
WARN_ON(reloc_inode->i_mapping->nrpages);
}
2008-09-26 14:09:34 +00:00
if (total_found > 0) {
printk("btrfs found %llu extents in pass %d\n",
(unsigned long long)total_found, pass);
pass++;
if (total_found == skipped && pass > 2) {
iput(reloc_inode);
reloc_inode = create_reloc_inode(info, block_group);
pass = 0;
}
2008-09-26 14:09:34 +00:00
goto again;
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
}
2008-09-26 14:09:34 +00:00
/* delete reloc_inode */
iput(reloc_inode);
/* unpin extents in this range */
trans = btrfs_start_transaction(info->tree_root, 1);
btrfs_commit_transaction(trans, info->tree_root);
2008-09-26 14:09:34 +00:00
spin_lock(&block_group->lock);
WARN_ON(block_group->pinned > 0);
WARN_ON(block_group->reserved > 0);
WARN_ON(btrfs_block_group_used(&block_group->item) > 0);
spin_unlock(&block_group->lock);
ret = 0;
out:
2008-09-26 14:09:34 +00:00
btrfs_free_path(path);
return ret;
}
int find_first_block_group(struct btrfs_root *root, struct btrfs_path *path,
struct btrfs_key *key)
{
int ret = 0;
struct btrfs_key found_key;
struct extent_buffer *leaf;
int slot;
ret = btrfs_search_slot(NULL, root, key, path, 0, 0);
if (ret < 0)
goto out;
while(1) {
slot = path->slots[0];
leaf = path->nodes[0];
if (slot >= btrfs_header_nritems(leaf)) {
ret = btrfs_next_leaf(root, path);
if (ret == 0)
continue;
if (ret < 0)
goto out;
break;
}
btrfs_item_key_to_cpu(leaf, &found_key, slot);
if (found_key.objectid >= key->objectid &&
found_key.type == BTRFS_BLOCK_GROUP_ITEM_KEY) {
ret = 0;
goto out;
}
path->slots[0]++;
}
ret = -ENOENT;
out:
return ret;
}
2008-09-26 14:09:34 +00:00
int btrfs_free_block_groups(struct btrfs_fs_info *info)
{
struct btrfs_block_group_cache *block_group;
struct rb_node *n;
spin_lock(&info->block_group_cache_lock);
while ((n = rb_last(&info->block_group_cache_tree)) != NULL) {
block_group = rb_entry(n, struct btrfs_block_group_cache,
cache_node);
rb_erase(&block_group->cache_node,
&info->block_group_cache_tree);
spin_unlock(&info->block_group_cache_lock);
btrfs_remove_free_space_cache(block_group);
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
down_write(&block_group->space_info->groups_sem);
2008-09-26 14:09:34 +00:00
list_del(&block_group->list);
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
up_write(&block_group->space_info->groups_sem);
2008-09-26 14:09:34 +00:00
kfree(block_group);
spin_lock(&info->block_group_cache_lock);
2008-09-26 14:09:34 +00:00
}
spin_unlock(&info->block_group_cache_lock);
return 0;
}
int btrfs_read_block_groups(struct btrfs_root *root)
{
struct btrfs_path *path;
int ret;
struct btrfs_block_group_cache *cache;
struct btrfs_fs_info *info = root->fs_info;
struct btrfs_space_info *space_info;
struct btrfs_key key;
struct btrfs_key found_key;
struct extent_buffer *leaf;
root = info->extent_root;
key.objectid = 0;
key.offset = 0;
btrfs_set_key_type(&key, BTRFS_BLOCK_GROUP_ITEM_KEY);
path = btrfs_alloc_path();
if (!path)
return -ENOMEM;
while(1) {
ret = find_first_block_group(root, path, &key);
if (ret > 0) {
ret = 0;
goto error;
}
if (ret != 0)
goto error;
leaf = path->nodes[0];
btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
cache = kzalloc(sizeof(*cache), GFP_NOFS);
if (!cache) {
ret = -ENOMEM;
break;
}
spin_lock_init(&cache->lock);
mutex_init(&cache->alloc_mutex);
mutex_init(&cache->cache_mutex);
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
INIT_LIST_HEAD(&cache->list);
read_extent_buffer(leaf, &cache->item,
btrfs_item_ptr_offset(leaf, path->slots[0]),
sizeof(cache->item));
memcpy(&cache->key, &found_key, sizeof(found_key));
key.objectid = found_key.objectid + found_key.offset;
btrfs_release_path(root, path);
cache->flags = btrfs_block_group_flags(&cache->item);
ret = update_space_info(info, cache->flags, found_key.offset,
btrfs_block_group_used(&cache->item),
&space_info);
BUG_ON(ret);
cache->space_info = space_info;
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
down_write(&space_info->groups_sem);
list_add_tail(&cache->list, &space_info->block_groups);
up_write(&space_info->groups_sem);
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
ret = btrfs_add_block_group_cache(root->fs_info, cache);
BUG_ON(ret);
set_avail_alloc_bits(root->fs_info, cache->flags);
if (btrfs_chunk_readonly(root, cache->key.objectid))
set_block_group_readonly(cache);
}
ret = 0;
error:
btrfs_free_path(path);
return ret;
}
int btrfs_make_block_group(struct btrfs_trans_handle *trans,
struct btrfs_root *root, u64 bytes_used,
u64 type, u64 chunk_objectid, u64 chunk_offset,
u64 size)
{
int ret;
struct btrfs_root *extent_root;
struct btrfs_block_group_cache *cache;
extent_root = root->fs_info->extent_root;
root->fs_info->last_trans_new_blockgroup = trans->transid;
cache = kzalloc(sizeof(*cache), GFP_NOFS);
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
if (!cache)
return -ENOMEM;
cache->key.objectid = chunk_offset;
cache->key.offset = size;
spin_lock_init(&cache->lock);
mutex_init(&cache->alloc_mutex);
mutex_init(&cache->cache_mutex);
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
INIT_LIST_HEAD(&cache->list);
btrfs_set_key_type(&cache->key, BTRFS_BLOCK_GROUP_ITEM_KEY);
btrfs_set_block_group_used(&cache->item, bytes_used);
btrfs_set_block_group_chunk_objectid(&cache->item, chunk_objectid);
cache->flags = type;
btrfs_set_block_group_flags(&cache->item, type);
ret = update_space_info(root->fs_info, cache->flags, size, bytes_used,
&cache->space_info);
BUG_ON(ret);
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
down_write(&cache->space_info->groups_sem);
list_add_tail(&cache->list, &cache->space_info->block_groups);
up_write(&cache->space_info->groups_sem);
Btrfs: free space accounting redo 1) replace the per fs_info extent_io_tree that tracked free space with two rb-trees per block group to track free space areas via offset and size. The reason to do this is because most allocations come with a hint byte where to start, so we can usually find a chunk of free space at that hint byte to satisfy the allocation and get good space packing. If we cannot find free space at or after the given offset we fall back on looking for a chunk of the given size as close to that given offset as possible. When we fall back on the size search we also try to find a slot as close to the size we want as possible, to avoid breaking small chunks off of huge areas if possible. 2) remove the extent_io_tree that tracked the block group cache from fs_info and replaced it with an rb-tree thats tracks block group cache via offset. also added a per space_info list that tracks the block group cache for the particular space so we can lookup related block groups easily. 3) cleaned up the allocation code to make it a little easier to read and a little less complicated. Basically there are 3 steps, first look from our provided hint. If we couldn't find from that given hint, start back at our original search start and look for space from there. If that fails try to allocate space if we can and start looking again. If not we're screwed and need to start over again. 4) small fixes. there were some issues in volumes.c where we wouldn't allocate the rest of the disk. fixed cow_file_range to actually pass the alloc_hint, which has helped a good bit in making the fs_mark test I run have semi-normal results as we run out of space. Generally with data allocations we don't track where we last allocated from, so everytime we did a data allocation we'd search through every block group that we have looking for free space. Now searching a block group with no free space isn't terribly time consuming, it was causing a slight degradation as we got more data block groups. The alloc_hint has fixed this slight degredation and made things semi-normal. There is still one nagging problem I'm working on where we will get ENOSPC when there is definitely plenty of space. This only happens with metadata allocations, and only when we are almost full. So you generally hit the 85% mark first, but sometimes you'll hit the BUG before you hit the 85% wall. I'm still tracking it down, but until then this seems to be pretty stable and make a significant performance gain. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2008-09-23 17:14:11 +00:00
ret = btrfs_add_block_group_cache(root->fs_info, cache);
BUG_ON(ret);
ret = btrfs_insert_item(trans, extent_root, &cache->key, &cache->item,
sizeof(cache->item));
BUG_ON(ret);
finish_current_insert(trans, extent_root, 0);
ret = del_pending_extents(trans, extent_root, 0);
BUG_ON(ret);
set_avail_alloc_bits(extent_root->fs_info, type);
return 0;
}
2008-09-26 14:09:34 +00:00
int btrfs_remove_block_group(struct btrfs_trans_handle *trans,
struct btrfs_root *root, u64 group_start)
{
struct btrfs_path *path;
struct btrfs_block_group_cache *block_group;
struct btrfs_key key;
int ret;
root = root->fs_info->extent_root;
block_group = btrfs_lookup_block_group(root->fs_info, group_start);
BUG_ON(!block_group);
BUG_ON(!block_group->ro);
2008-09-26 14:09:34 +00:00
memcpy(&key, &block_group->key, sizeof(key));
path = btrfs_alloc_path();
BUG_ON(!path);
btrfs_remove_free_space_cache(block_group);
rb_erase(&block_group->cache_node,
&root->fs_info->block_group_cache_tree);
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
down_write(&block_group->space_info->groups_sem);
2008-09-26 14:09:34 +00:00
list_del(&block_group->list);
Btrfs: fix enospc when there is plenty of space So there is an odd case where we can possibly return -ENOSPC when there is in fact space to be had. It only happens with Metadata writes, and happens _very_ infrequently. What has to happen is we have to allocate have allocated out of the first logical byte on the disk, which would set last_alloc to first_logical_byte(root, 0), so search_start == orig_search_start. We then need to allocate for normal metadata, so BTRFS_BLOCK_GROUP_METADATA | BTRFS_BLOCK_GROUP_DUP. We will do a block lookup for the given search_start, block_group_bits() won't match and we'll go to choose another block group. However because search_start matches orig_search_start we go to see if we can allocate a chunk. If we are in the situation that we cannot allocate a chunk, we fail and ENOSPC. This is kind of a big flaw of the way find_free_extent works, as it along with find_free_space loop through _all_ of the block groups, not just the ones that we want to allocate out of. This patch completely kills find_free_space and rolls it into find_free_extent. I've introduced a sort of state machine into this, which will make it easier to get cache miss information out of the allocator, and will work well with my locking changes. The basic flow is this: We have the variable loop which is 0, meaning we are in the hint phase. We lookup the block group for the hint, and lookup the space_info for what we want to allocate out of. If the block group we were pointed at by the hint either isn't of the correct type, or just doesn't have the space we need, we set head to space_info->block_groups, so we start at the beginning of the block groups for this particular space info, and loop through. This is also where we add the empty_cluster to total_needed. At this point loop is set to 1 and we just loop through all of the block groups for this particular space_info looking for the space we need, just as find_free_space would have done, except we only hit the block groups we want and not _all_ of the block groups. If we come full circle we see if we can allocate a chunk. If we cannot of course we exit with -ENOSPC and we are good. If not we start over at space_info->block_groups and loop through again, with loop == 2. If we come full circle and haven't found what we need then we exit with -ENOSPC. I've been running this for a couple of days now and it seems stable, and I haven't yet hit a -ENOSPC when there was plenty of space left. Also I've made a groups_sem to handle the group list for the space_info. This is part of my locking changes, but is relatively safe and seems better than holding the space_info spinlock over that entire search time. Thanks, Signed-off-by: Josef Bacik <jbacik@redhat.com>
2008-10-29 18:49:05 +00:00
up_write(&block_group->space_info->groups_sem);
2008-09-26 14:09:34 +00:00
spin_lock(&block_group->space_info->lock);
block_group->space_info->total_bytes -= block_group->key.offset;
block_group->space_info->bytes_readonly -= block_group->key.offset;
spin_unlock(&block_group->space_info->lock);
block_group->space_info->full = 0;
2008-09-26 14:09:34 +00:00
/*
memset(shrink_block_group, 0, sizeof(*shrink_block_group));
kfree(shrink_block_group);
*/
ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
if (ret > 0)
ret = -EIO;
if (ret < 0)
goto out;
ret = btrfs_del_item(trans, root, path);
out:
btrfs_free_path(path);
return ret;
}