linux/fs/btrfs/transaction.c

1151 lines
30 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/fs.h>
#include <linux/sched.h>
#include <linux/writeback.h>
#include <linux/pagemap.h>
#include <linux/blkdev.h>
#include "ctree.h"
#include "disk-io.h"
#include "transaction.h"
#include "locking.h"
#include "tree-log.h"
#define BTRFS_ROOT_TRANS_TAG 0
static noinline void put_transaction(struct btrfs_transaction *transaction)
{
WARN_ON(transaction->use_count == 0);
transaction->use_count--;
if (transaction->use_count == 0) {
list_del_init(&transaction->list);
memset(transaction, 0, sizeof(*transaction));
kmem_cache_free(btrfs_transaction_cachep, transaction);
}
}
Btrfs: async block group caching This patch moves the caching of the block group off to a kthread in order to allow people to allocate sooner. Instead of blocking up behind the caching mutex, we instead kick of the caching kthread, and then attempt to make an allocation. If we cannot, we wait on the block groups caching waitqueue, which the caching kthread will wake the waiting threads up everytime it finds 2 meg worth of space, and then again when its finished caching. This is how I tested the speedup from this mkfs the disk mount the disk fill the disk up with fs_mark unmount the disk mount the disk time touch /mnt/foo Without my changes this took 11 seconds on my box, with these changes it now takes 1 second. Another change thats been put in place is we lock the super mirror's in the pinned extent map in order to keep us from adding that stuff as free space when caching the block group. This doesn't really change anything else as far as the pinned extent map is concerned, since for actual pinned extents we use EXTENT_DIRTY, but it does mean that when we unmount we have to go in and unlock those extents to keep from leaking memory. I've also added a check where when we are reading block groups from disk, if the amount of space used == the size of the block group, we go ahead and mark the block group as cached. This drastically reduces the amount of time it takes to cache the block groups. Using the same test as above, except doing a dd to a file and then unmounting, it used to take 33 seconds to umount, now it takes 3 seconds. This version uses the commit_root in the caching kthread, and then keeps track of how many async caching threads are running at any given time so if one of the async threads is still running as we cross transactions we can wait until its finished before handling the pinned extents. Thank you, Signed-off-by: Josef Bacik <jbacik@redhat.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-07-14 01:29:25 +00:00
static noinline void switch_commit_root(struct btrfs_root *root)
{
free_extent_buffer(root->commit_root);
root->commit_root = btrfs_root_node(root);
}
/*
* either allocate a new transaction or hop into the existing one
*/
static noinline int join_transaction(struct btrfs_root *root)
{
struct btrfs_transaction *cur_trans;
cur_trans = root->fs_info->running_transaction;
if (!cur_trans) {
cur_trans = kmem_cache_alloc(btrfs_transaction_cachep,
GFP_NOFS);
BUG_ON(!cur_trans);
root->fs_info->generation++;
cur_trans->num_writers = 1;
cur_trans->num_joined = 0;
cur_trans->transid = root->fs_info->generation;
init_waitqueue_head(&cur_trans->writer_wait);
init_waitqueue_head(&cur_trans->commit_wait);
cur_trans->in_commit = 0;
cur_trans->blocked = 0;
cur_trans->use_count = 1;
cur_trans->commit_done = 0;
cur_trans->start_time = get_seconds();
Btrfs: do extent allocation and reference count updates in the background The extent allocation tree maintains a reference count and full back reference information for every extent allocated in the filesystem. For subvolume and snapshot trees, every time a block goes through COW, the new copy of the block adds a reference on every block it points to. If a btree node points to 150 leaves, then the COW code needs to go and add backrefs on 150 different extents, which might be spread all over the extent allocation tree. These updates currently happen during btrfs_cow_block, and most COWs happen during btrfs_search_slot. btrfs_search_slot has locks held on both the parent and the node we are COWing, and so we really want to avoid IO during the COW if we can. This commit adds an rbtree of pending reference count updates and extent allocations. The tree is ordered by byte number of the extent and byte number of the parent for the back reference. The tree allows us to: 1) Modify back references in something close to disk order, reducing seeks 2) Significantly reduce the number of modifications made as block pointers are balanced around 3) Do all of the extent insertion and back reference modifications outside of the performance critical btrfs_search_slot code. #3 has the added benefit of greatly reducing the btrfs stack footprint. The extent allocation tree modifications are done without the deep (and somewhat recursive) call chains used in the past. These delayed back reference updates must be done before the transaction commits, and so the rbtree is tied to the transaction. Throttling is implemented to help keep the queue of backrefs at a reasonable size. Since there was a similar mechanism in place for the extent tree extents, that is removed and replaced by the delayed reference tree. Yan Zheng <yan.zheng@oracle.com> helped review and fixup this code. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-03-13 14:10:06 +00:00
cur_trans->delayed_refs.root = RB_ROOT;
Btrfs: do extent allocation and reference count updates in the background The extent allocation tree maintains a reference count and full back reference information for every extent allocated in the filesystem. For subvolume and snapshot trees, every time a block goes through COW, the new copy of the block adds a reference on every block it points to. If a btree node points to 150 leaves, then the COW code needs to go and add backrefs on 150 different extents, which might be spread all over the extent allocation tree. These updates currently happen during btrfs_cow_block, and most COWs happen during btrfs_search_slot. btrfs_search_slot has locks held on both the parent and the node we are COWing, and so we really want to avoid IO during the COW if we can. This commit adds an rbtree of pending reference count updates and extent allocations. The tree is ordered by byte number of the extent and byte number of the parent for the back reference. The tree allows us to: 1) Modify back references in something close to disk order, reducing seeks 2) Significantly reduce the number of modifications made as block pointers are balanced around 3) Do all of the extent insertion and back reference modifications outside of the performance critical btrfs_search_slot code. #3 has the added benefit of greatly reducing the btrfs stack footprint. The extent allocation tree modifications are done without the deep (and somewhat recursive) call chains used in the past. These delayed back reference updates must be done before the transaction commits, and so the rbtree is tied to the transaction. Throttling is implemented to help keep the queue of backrefs at a reasonable size. Since there was a similar mechanism in place for the extent tree extents, that is removed and replaced by the delayed reference tree. Yan Zheng <yan.zheng@oracle.com> helped review and fixup this code. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-03-13 14:10:06 +00:00
cur_trans->delayed_refs.num_entries = 0;
cur_trans->delayed_refs.num_heads_ready = 0;
cur_trans->delayed_refs.num_heads = 0;
Btrfs: do extent allocation and reference count updates in the background The extent allocation tree maintains a reference count and full back reference information for every extent allocated in the filesystem. For subvolume and snapshot trees, every time a block goes through COW, the new copy of the block adds a reference on every block it points to. If a btree node points to 150 leaves, then the COW code needs to go and add backrefs on 150 different extents, which might be spread all over the extent allocation tree. These updates currently happen during btrfs_cow_block, and most COWs happen during btrfs_search_slot. btrfs_search_slot has locks held on both the parent and the node we are COWing, and so we really want to avoid IO during the COW if we can. This commit adds an rbtree of pending reference count updates and extent allocations. The tree is ordered by byte number of the extent and byte number of the parent for the back reference. The tree allows us to: 1) Modify back references in something close to disk order, reducing seeks 2) Significantly reduce the number of modifications made as block pointers are balanced around 3) Do all of the extent insertion and back reference modifications outside of the performance critical btrfs_search_slot code. #3 has the added benefit of greatly reducing the btrfs stack footprint. The extent allocation tree modifications are done without the deep (and somewhat recursive) call chains used in the past. These delayed back reference updates must be done before the transaction commits, and so the rbtree is tied to the transaction. Throttling is implemented to help keep the queue of backrefs at a reasonable size. Since there was a similar mechanism in place for the extent tree extents, that is removed and replaced by the delayed reference tree. Yan Zheng <yan.zheng@oracle.com> helped review and fixup this code. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-03-13 14:10:06 +00:00
cur_trans->delayed_refs.flushing = 0;
cur_trans->delayed_refs.run_delayed_start = 0;
Btrfs: do extent allocation and reference count updates in the background The extent allocation tree maintains a reference count and full back reference information for every extent allocated in the filesystem. For subvolume and snapshot trees, every time a block goes through COW, the new copy of the block adds a reference on every block it points to. If a btree node points to 150 leaves, then the COW code needs to go and add backrefs on 150 different extents, which might be spread all over the extent allocation tree. These updates currently happen during btrfs_cow_block, and most COWs happen during btrfs_search_slot. btrfs_search_slot has locks held on both the parent and the node we are COWing, and so we really want to avoid IO during the COW if we can. This commit adds an rbtree of pending reference count updates and extent allocations. The tree is ordered by byte number of the extent and byte number of the parent for the back reference. The tree allows us to: 1) Modify back references in something close to disk order, reducing seeks 2) Significantly reduce the number of modifications made as block pointers are balanced around 3) Do all of the extent insertion and back reference modifications outside of the performance critical btrfs_search_slot code. #3 has the added benefit of greatly reducing the btrfs stack footprint. The extent allocation tree modifications are done without the deep (and somewhat recursive) call chains used in the past. These delayed back reference updates must be done before the transaction commits, and so the rbtree is tied to the transaction. Throttling is implemented to help keep the queue of backrefs at a reasonable size. Since there was a similar mechanism in place for the extent tree extents, that is removed and replaced by the delayed reference tree. Yan Zheng <yan.zheng@oracle.com> helped review and fixup this code. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-03-13 14:10:06 +00:00
spin_lock_init(&cur_trans->delayed_refs.lock);
INIT_LIST_HEAD(&cur_trans->pending_snapshots);
list_add_tail(&cur_trans->list, &root->fs_info->trans_list);
extent_io_tree_init(&cur_trans->dirty_pages,
root->fs_info->btree_inode->i_mapping,
GFP_NOFS);
spin_lock(&root->fs_info->new_trans_lock);
root->fs_info->running_transaction = cur_trans;
spin_unlock(&root->fs_info->new_trans_lock);
} else {
cur_trans->num_writers++;
cur_trans->num_joined++;
}
return 0;
}
/*
* this does all the record keeping required to make sure that a reference
* counted root is properly recorded in a given transaction. This is required
* to make sure the old root from before we joined the transaction is deleted
* when the transaction commits
*/
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
static noinline int record_root_in_trans(struct btrfs_trans_handle *trans,
struct btrfs_root *root)
{
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
if (root->ref_cows && root->last_trans < trans->transid) {
WARN_ON(root == root->fs_info->extent_root);
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
WARN_ON(root->commit_root != root->node);
radix_tree_tag_set(&root->fs_info->fs_roots_radix,
(unsigned long)root->root_key.objectid,
BTRFS_ROOT_TRANS_TAG);
root->last_trans = trans->transid;
btrfs_init_reloc_root(trans, root);
}
return 0;
}
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
int btrfs_record_root_in_trans(struct btrfs_trans_handle *trans,
struct btrfs_root *root)
{
if (!root->ref_cows)
return 0;
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
mutex_lock(&root->fs_info->trans_mutex);
if (root->last_trans == trans->transid) {
mutex_unlock(&root->fs_info->trans_mutex);
return 0;
}
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
record_root_in_trans(trans, root);
mutex_unlock(&root->fs_info->trans_mutex);
return 0;
}
/* wait for commit against the current transaction to become unblocked
* when this is done, it is safe to start a new transaction, but the current
* transaction might not be fully on disk.
*/
static void wait_current_trans(struct btrfs_root *root)
{
struct btrfs_transaction *cur_trans;
cur_trans = root->fs_info->running_transaction;
if (cur_trans && cur_trans->blocked) {
DEFINE_WAIT(wait);
cur_trans->use_count++;
while (1) {
prepare_to_wait(&root->fs_info->transaction_wait, &wait,
TASK_UNINTERRUPTIBLE);
if (cur_trans->blocked) {
mutex_unlock(&root->fs_info->trans_mutex);
schedule();
mutex_lock(&root->fs_info->trans_mutex);
finish_wait(&root->fs_info->transaction_wait,
&wait);
} else {
finish_wait(&root->fs_info->transaction_wait,
&wait);
break;
}
}
put_transaction(cur_trans);
}
}
enum btrfs_trans_type {
TRANS_START,
TRANS_JOIN,
TRANS_USERSPACE,
};
static struct btrfs_trans_handle *start_transaction(struct btrfs_root *root,
int num_blocks, int type)
{
struct btrfs_trans_handle *h =
kmem_cache_alloc(btrfs_trans_handle_cachep, GFP_NOFS);
int ret;
mutex_lock(&root->fs_info->trans_mutex);
if (!root->fs_info->log_root_recovering &&
((type == TRANS_START && !root->fs_info->open_ioctl_trans) ||
type == TRANS_USERSPACE))
wait_current_trans(root);
ret = join_transaction(root);
BUG_ON(ret);
h->transid = root->fs_info->running_transaction->transid;
h->transaction = root->fs_info->running_transaction;
h->blocks_reserved = num_blocks;
h->blocks_used = 0;
h->block_group = 0;
h->alloc_exclude_nr = 0;
h->alloc_exclude_start = 0;
Btrfs: do extent allocation and reference count updates in the background The extent allocation tree maintains a reference count and full back reference information for every extent allocated in the filesystem. For subvolume and snapshot trees, every time a block goes through COW, the new copy of the block adds a reference on every block it points to. If a btree node points to 150 leaves, then the COW code needs to go and add backrefs on 150 different extents, which might be spread all over the extent allocation tree. These updates currently happen during btrfs_cow_block, and most COWs happen during btrfs_search_slot. btrfs_search_slot has locks held on both the parent and the node we are COWing, and so we really want to avoid IO during the COW if we can. This commit adds an rbtree of pending reference count updates and extent allocations. The tree is ordered by byte number of the extent and byte number of the parent for the back reference. The tree allows us to: 1) Modify back references in something close to disk order, reducing seeks 2) Significantly reduce the number of modifications made as block pointers are balanced around 3) Do all of the extent insertion and back reference modifications outside of the performance critical btrfs_search_slot code. #3 has the added benefit of greatly reducing the btrfs stack footprint. The extent allocation tree modifications are done without the deep (and somewhat recursive) call chains used in the past. These delayed back reference updates must be done before the transaction commits, and so the rbtree is tied to the transaction. Throttling is implemented to help keep the queue of backrefs at a reasonable size. Since there was a similar mechanism in place for the extent tree extents, that is removed and replaced by the delayed reference tree. Yan Zheng <yan.zheng@oracle.com> helped review and fixup this code. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-03-13 14:10:06 +00:00
h->delayed_ref_updates = 0;
if (!current->journal_info && type != TRANS_USERSPACE)
Btrfs: proper -ENOSPC handling At the start of a transaction we do a btrfs_reserve_metadata_space() and specify how many items we plan on modifying. Then once we've done our modifications and such, just call btrfs_unreserve_metadata_space() for the same number of items we reserved. For keeping track of metadata needed for data I've had to add an extent_io op for when we merge extents. This lets us track space properly when we are doing sequential writes, so we don't end up reserving way more metadata space than what we need. The only place where the metadata space accounting is not done is in the relocation code. This is because Yan is going to be reworking that code in the near future, so running btrfs-vol -b could still possibly result in a ENOSPC related panic. This patch also turns off the metadata_ratio stuff in order to allow users to more efficiently use their disk space. This patch makes it so we track how much metadata we need for an inode's delayed allocation extents by tracking how many extents are currently waiting for allocation. It introduces two new callbacks for the extent_io tree's, merge_extent_hook and split_extent_hook. These help us keep track of when we merge delalloc extents together and split them up. Reservations are handled prior to any actually dirty'ing occurs, and then we unreserve after we dirty. btrfs_unreserve_metadata_for_delalloc() will make the appropriate unreservations as needed based on the number of reservations we currently have and the number of extents we currently have. Doing the reservation outside of doing any of the actual dirty'ing lets us do things like filemap_flush() the inode to try and force delalloc to happen, or as a last resort actually start allocation on all delalloc inodes in the fs. This has survived dbench, fs_mark and an fsx torture test. Signed-off-by: Josef Bacik <jbacik@redhat.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-09-11 20:12:44 +00:00
current->journal_info = h;
root->fs_info->running_transaction->use_count++;
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
record_root_in_trans(h, root);
mutex_unlock(&root->fs_info->trans_mutex);
return h;
}
struct btrfs_trans_handle *btrfs_start_transaction(struct btrfs_root *root,
int num_blocks)
{
return start_transaction(root, num_blocks, TRANS_START);
}
struct btrfs_trans_handle *btrfs_join_transaction(struct btrfs_root *root,
int num_blocks)
{
return start_transaction(root, num_blocks, TRANS_JOIN);
}
struct btrfs_trans_handle *btrfs_start_ioctl_transaction(struct btrfs_root *r,
int num_blocks)
{
return start_transaction(r, num_blocks, TRANS_USERSPACE);
}
/* wait for a transaction commit to be fully complete */
static noinline int wait_for_commit(struct btrfs_root *root,
struct btrfs_transaction *commit)
{
DEFINE_WAIT(wait);
mutex_lock(&root->fs_info->trans_mutex);
while (!commit->commit_done) {
prepare_to_wait(&commit->commit_wait, &wait,
TASK_UNINTERRUPTIBLE);
if (commit->commit_done)
break;
mutex_unlock(&root->fs_info->trans_mutex);
schedule();
mutex_lock(&root->fs_info->trans_mutex);
}
mutex_unlock(&root->fs_info->trans_mutex);
finish_wait(&commit->commit_wait, &wait);
return 0;
}
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
#if 0
/*
* rate limit against the drop_snapshot code. This helps to slow down new
* operations if the drop_snapshot code isn't able to keep up.
*/
static void throttle_on_drops(struct btrfs_root *root)
{
struct btrfs_fs_info *info = root->fs_info;
int harder_count = 0;
harder:
if (atomic_read(&info->throttles)) {
DEFINE_WAIT(wait);
int thr;
thr = atomic_read(&info->throttle_gen);
do {
prepare_to_wait(&info->transaction_throttle,
&wait, TASK_UNINTERRUPTIBLE);
if (!atomic_read(&info->throttles)) {
finish_wait(&info->transaction_throttle, &wait);
break;
}
schedule();
finish_wait(&info->transaction_throttle, &wait);
} while (thr == atomic_read(&info->throttle_gen));
harder_count++;
if (root->fs_info->total_ref_cache_size > 1 * 1024 * 1024 &&
harder_count < 2)
goto harder;
if (root->fs_info->total_ref_cache_size > 5 * 1024 * 1024 &&
harder_count < 10)
goto harder;
if (root->fs_info->total_ref_cache_size > 10 * 1024 * 1024 &&
harder_count < 20)
goto harder;
}
}
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
#endif
void btrfs_throttle(struct btrfs_root *root)
{
mutex_lock(&root->fs_info->trans_mutex);
if (!root->fs_info->open_ioctl_trans)
wait_current_trans(root);
mutex_unlock(&root->fs_info->trans_mutex);
}
static int __btrfs_end_transaction(struct btrfs_trans_handle *trans,
struct btrfs_root *root, int throttle)
{
struct btrfs_transaction *cur_trans;
struct btrfs_fs_info *info = root->fs_info;
int count = 0;
while (count < 4) {
unsigned long cur = trans->delayed_ref_updates;
trans->delayed_ref_updates = 0;
if (cur &&
trans->transaction->delayed_refs.num_heads_ready > 64) {
trans->delayed_ref_updates = 0;
/*
* do a full flush if the transaction is trying
* to close
*/
if (trans->transaction->delayed_refs.flushing)
cur = 0;
btrfs_run_delayed_refs(trans, root, cur);
} else {
break;
}
count++;
Btrfs: do extent allocation and reference count updates in the background The extent allocation tree maintains a reference count and full back reference information for every extent allocated in the filesystem. For subvolume and snapshot trees, every time a block goes through COW, the new copy of the block adds a reference on every block it points to. If a btree node points to 150 leaves, then the COW code needs to go and add backrefs on 150 different extents, which might be spread all over the extent allocation tree. These updates currently happen during btrfs_cow_block, and most COWs happen during btrfs_search_slot. btrfs_search_slot has locks held on both the parent and the node we are COWing, and so we really want to avoid IO during the COW if we can. This commit adds an rbtree of pending reference count updates and extent allocations. The tree is ordered by byte number of the extent and byte number of the parent for the back reference. The tree allows us to: 1) Modify back references in something close to disk order, reducing seeks 2) Significantly reduce the number of modifications made as block pointers are balanced around 3) Do all of the extent insertion and back reference modifications outside of the performance critical btrfs_search_slot code. #3 has the added benefit of greatly reducing the btrfs stack footprint. The extent allocation tree modifications are done without the deep (and somewhat recursive) call chains used in the past. These delayed back reference updates must be done before the transaction commits, and so the rbtree is tied to the transaction. Throttling is implemented to help keep the queue of backrefs at a reasonable size. Since there was a similar mechanism in place for the extent tree extents, that is removed and replaced by the delayed reference tree. Yan Zheng <yan.zheng@oracle.com> helped review and fixup this code. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-03-13 14:10:06 +00:00
}
mutex_lock(&info->trans_mutex);
cur_trans = info->running_transaction;
WARN_ON(cur_trans != trans->transaction);
WARN_ON(cur_trans->num_writers < 1);
cur_trans->num_writers--;
if (waitqueue_active(&cur_trans->writer_wait))
wake_up(&cur_trans->writer_wait);
put_transaction(cur_trans);
mutex_unlock(&info->trans_mutex);
Btrfs: proper -ENOSPC handling At the start of a transaction we do a btrfs_reserve_metadata_space() and specify how many items we plan on modifying. Then once we've done our modifications and such, just call btrfs_unreserve_metadata_space() for the same number of items we reserved. For keeping track of metadata needed for data I've had to add an extent_io op for when we merge extents. This lets us track space properly when we are doing sequential writes, so we don't end up reserving way more metadata space than what we need. The only place where the metadata space accounting is not done is in the relocation code. This is because Yan is going to be reworking that code in the near future, so running btrfs-vol -b could still possibly result in a ENOSPC related panic. This patch also turns off the metadata_ratio stuff in order to allow users to more efficiently use their disk space. This patch makes it so we track how much metadata we need for an inode's delayed allocation extents by tracking how many extents are currently waiting for allocation. It introduces two new callbacks for the extent_io tree's, merge_extent_hook and split_extent_hook. These help us keep track of when we merge delalloc extents together and split them up. Reservations are handled prior to any actually dirty'ing occurs, and then we unreserve after we dirty. btrfs_unreserve_metadata_for_delalloc() will make the appropriate unreservations as needed based on the number of reservations we currently have and the number of extents we currently have. Doing the reservation outside of doing any of the actual dirty'ing lets us do things like filemap_flush() the inode to try and force delalloc to happen, or as a last resort actually start allocation on all delalloc inodes in the fs. This has survived dbench, fs_mark and an fsx torture test. Signed-off-by: Josef Bacik <jbacik@redhat.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-09-11 20:12:44 +00:00
if (current->journal_info == trans)
current->journal_info = NULL;
memset(trans, 0, sizeof(*trans));
kmem_cache_free(btrfs_trans_handle_cachep, trans);
if (throttle)
btrfs_run_delayed_iputs(root);
return 0;
}
int btrfs_end_transaction(struct btrfs_trans_handle *trans,
struct btrfs_root *root)
{
return __btrfs_end_transaction(trans, root, 0);
}
int btrfs_end_transaction_throttle(struct btrfs_trans_handle *trans,
struct btrfs_root *root)
{
return __btrfs_end_transaction(trans, root, 1);
}
/*
* when btree blocks are allocated, they have some corresponding bits set for
* them in one of two extent_io trees. This is used to make sure all of
* those extents are sent to disk but does not wait on them
*/
int btrfs_write_marked_extents(struct btrfs_root *root,
struct extent_io_tree *dirty_pages, int mark)
{
int ret;
int err = 0;
int werr = 0;
struct page *page;
struct inode *btree_inode = root->fs_info->btree_inode;
u64 start = 0;
u64 end;
unsigned long index;
while (1) {
ret = find_first_extent_bit(dirty_pages, start, &start, &end,
mark);
if (ret)
break;
while (start <= end) {
cond_resched();
index = start >> PAGE_CACHE_SHIFT;
start = (u64)(index + 1) << PAGE_CACHE_SHIFT;
page = find_get_page(btree_inode->i_mapping, index);
if (!page)
continue;
btree_lock_page_hook(page);
if (!page->mapping) {
unlock_page(page);
page_cache_release(page);
continue;
}
if (PageWriteback(page)) {
if (PageDirty(page))
wait_on_page_writeback(page);
else {
unlock_page(page);
page_cache_release(page);
continue;
}
}
err = write_one_page(page, 0);
if (err)
werr = err;
page_cache_release(page);
}
}
if (err)
werr = err;
return werr;
}
/*
* when btree blocks are allocated, they have some corresponding bits set for
* them in one of two extent_io trees. This is used to make sure all of
* those extents are on disk for transaction or log commit. We wait
* on all the pages and clear them from the dirty pages state tree
*/
int btrfs_wait_marked_extents(struct btrfs_root *root,
struct extent_io_tree *dirty_pages, int mark)
{
int ret;
int err = 0;
int werr = 0;
struct page *page;
struct inode *btree_inode = root->fs_info->btree_inode;
u64 start = 0;
u64 end;
unsigned long index;
while (1) {
ret = find_first_extent_bit(dirty_pages, start, &start, &end,
mark);
if (ret)
break;
clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS);
while (start <= end) {
index = start >> PAGE_CACHE_SHIFT;
start = (u64)(index + 1) << PAGE_CACHE_SHIFT;
page = find_get_page(btree_inode->i_mapping, index);
if (!page)
continue;
if (PageDirty(page)) {
btree_lock_page_hook(page);
wait_on_page_writeback(page);
err = write_one_page(page, 0);
if (err)
werr = err;
}
wait_on_page_writeback(page);
page_cache_release(page);
cond_resched();
}
}
if (err)
werr = err;
return werr;
}
/*
* when btree blocks are allocated, they have some corresponding bits set for
* them in one of two extent_io trees. This is used to make sure all of
* those extents are on disk for transaction or log commit
*/
int btrfs_write_and_wait_marked_extents(struct btrfs_root *root,
struct extent_io_tree *dirty_pages, int mark)
{
int ret;
int ret2;
ret = btrfs_write_marked_extents(root, dirty_pages, mark);
ret2 = btrfs_wait_marked_extents(root, dirty_pages, mark);
return ret || ret2;
}
int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans,
struct btrfs_root *root)
{
if (!trans || !trans->transaction) {
struct inode *btree_inode;
btree_inode = root->fs_info->btree_inode;
return filemap_write_and_wait(btree_inode->i_mapping);
}
return btrfs_write_and_wait_marked_extents(root,
&trans->transaction->dirty_pages,
EXTENT_DIRTY);
}
/*
* this is used to update the root pointer in the tree of tree roots.
*
* But, in the case of the extent allocation tree, updating the root
* pointer may allocate blocks which may change the root of the extent
* allocation tree.
*
* So, this loops and repeats and makes sure the cowonly root didn't
* change while the root pointer was being updated in the metadata.
*/
static int update_cowonly_root(struct btrfs_trans_handle *trans,
struct btrfs_root *root)
{
int ret;
u64 old_root_bytenr;
u64 old_root_used;
struct btrfs_root *tree_root = root->fs_info->tree_root;
old_root_used = btrfs_root_used(&root->root_item);
btrfs_write_dirty_block_groups(trans, root);
Btrfs: do extent allocation and reference count updates in the background The extent allocation tree maintains a reference count and full back reference information for every extent allocated in the filesystem. For subvolume and snapshot trees, every time a block goes through COW, the new copy of the block adds a reference on every block it points to. If a btree node points to 150 leaves, then the COW code needs to go and add backrefs on 150 different extents, which might be spread all over the extent allocation tree. These updates currently happen during btrfs_cow_block, and most COWs happen during btrfs_search_slot. btrfs_search_slot has locks held on both the parent and the node we are COWing, and so we really want to avoid IO during the COW if we can. This commit adds an rbtree of pending reference count updates and extent allocations. The tree is ordered by byte number of the extent and byte number of the parent for the back reference. The tree allows us to: 1) Modify back references in something close to disk order, reducing seeks 2) Significantly reduce the number of modifications made as block pointers are balanced around 3) Do all of the extent insertion and back reference modifications outside of the performance critical btrfs_search_slot code. #3 has the added benefit of greatly reducing the btrfs stack footprint. The extent allocation tree modifications are done without the deep (and somewhat recursive) call chains used in the past. These delayed back reference updates must be done before the transaction commits, and so the rbtree is tied to the transaction. Throttling is implemented to help keep the queue of backrefs at a reasonable size. Since there was a similar mechanism in place for the extent tree extents, that is removed and replaced by the delayed reference tree. Yan Zheng <yan.zheng@oracle.com> helped review and fixup this code. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-03-13 14:10:06 +00:00
while (1) {
old_root_bytenr = btrfs_root_bytenr(&root->root_item);
if (old_root_bytenr == root->node->start &&
old_root_used == btrfs_root_used(&root->root_item))
break;
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
btrfs_set_root_node(&root->root_item, root->node);
ret = btrfs_update_root(trans, tree_root,
&root->root_key,
&root->root_item);
BUG_ON(ret);
Btrfs: do extent allocation and reference count updates in the background The extent allocation tree maintains a reference count and full back reference information for every extent allocated in the filesystem. For subvolume and snapshot trees, every time a block goes through COW, the new copy of the block adds a reference on every block it points to. If a btree node points to 150 leaves, then the COW code needs to go and add backrefs on 150 different extents, which might be spread all over the extent allocation tree. These updates currently happen during btrfs_cow_block, and most COWs happen during btrfs_search_slot. btrfs_search_slot has locks held on both the parent and the node we are COWing, and so we really want to avoid IO during the COW if we can. This commit adds an rbtree of pending reference count updates and extent allocations. The tree is ordered by byte number of the extent and byte number of the parent for the back reference. The tree allows us to: 1) Modify back references in something close to disk order, reducing seeks 2) Significantly reduce the number of modifications made as block pointers are balanced around 3) Do all of the extent insertion and back reference modifications outside of the performance critical btrfs_search_slot code. #3 has the added benefit of greatly reducing the btrfs stack footprint. The extent allocation tree modifications are done without the deep (and somewhat recursive) call chains used in the past. These delayed back reference updates must be done before the transaction commits, and so the rbtree is tied to the transaction. Throttling is implemented to help keep the queue of backrefs at a reasonable size. Since there was a similar mechanism in place for the extent tree extents, that is removed and replaced by the delayed reference tree. Yan Zheng <yan.zheng@oracle.com> helped review and fixup this code. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-03-13 14:10:06 +00:00
old_root_used = btrfs_root_used(&root->root_item);
ret = btrfs_write_dirty_block_groups(trans, root);
Btrfs: do extent allocation and reference count updates in the background The extent allocation tree maintains a reference count and full back reference information for every extent allocated in the filesystem. For subvolume and snapshot trees, every time a block goes through COW, the new copy of the block adds a reference on every block it points to. If a btree node points to 150 leaves, then the COW code needs to go and add backrefs on 150 different extents, which might be spread all over the extent allocation tree. These updates currently happen during btrfs_cow_block, and most COWs happen during btrfs_search_slot. btrfs_search_slot has locks held on both the parent and the node we are COWing, and so we really want to avoid IO during the COW if we can. This commit adds an rbtree of pending reference count updates and extent allocations. The tree is ordered by byte number of the extent and byte number of the parent for the back reference. The tree allows us to: 1) Modify back references in something close to disk order, reducing seeks 2) Significantly reduce the number of modifications made as block pointers are balanced around 3) Do all of the extent insertion and back reference modifications outside of the performance critical btrfs_search_slot code. #3 has the added benefit of greatly reducing the btrfs stack footprint. The extent allocation tree modifications are done without the deep (and somewhat recursive) call chains used in the past. These delayed back reference updates must be done before the transaction commits, and so the rbtree is tied to the transaction. Throttling is implemented to help keep the queue of backrefs at a reasonable size. Since there was a similar mechanism in place for the extent tree extents, that is removed and replaced by the delayed reference tree. Yan Zheng <yan.zheng@oracle.com> helped review and fixup this code. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-03-13 14:10:06 +00:00
BUG_ON(ret);
}
if (root != root->fs_info->extent_root)
switch_commit_root(root);
return 0;
}
/*
* update all the cowonly tree roots on disk
*/
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
static noinline int commit_cowonly_roots(struct btrfs_trans_handle *trans,
struct btrfs_root *root)
{
struct btrfs_fs_info *fs_info = root->fs_info;
struct list_head *next;
struct extent_buffer *eb;
Btrfs: do extent allocation and reference count updates in the background The extent allocation tree maintains a reference count and full back reference information for every extent allocated in the filesystem. For subvolume and snapshot trees, every time a block goes through COW, the new copy of the block adds a reference on every block it points to. If a btree node points to 150 leaves, then the COW code needs to go and add backrefs on 150 different extents, which might be spread all over the extent allocation tree. These updates currently happen during btrfs_cow_block, and most COWs happen during btrfs_search_slot. btrfs_search_slot has locks held on both the parent and the node we are COWing, and so we really want to avoid IO during the COW if we can. This commit adds an rbtree of pending reference count updates and extent allocations. The tree is ordered by byte number of the extent and byte number of the parent for the back reference. The tree allows us to: 1) Modify back references in something close to disk order, reducing seeks 2) Significantly reduce the number of modifications made as block pointers are balanced around 3) Do all of the extent insertion and back reference modifications outside of the performance critical btrfs_search_slot code. #3 has the added benefit of greatly reducing the btrfs stack footprint. The extent allocation tree modifications are done without the deep (and somewhat recursive) call chains used in the past. These delayed back reference updates must be done before the transaction commits, and so the rbtree is tied to the transaction. Throttling is implemented to help keep the queue of backrefs at a reasonable size. Since there was a similar mechanism in place for the extent tree extents, that is removed and replaced by the delayed reference tree. Yan Zheng <yan.zheng@oracle.com> helped review and fixup this code. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-03-13 14:10:06 +00:00
int ret;
Btrfs: do extent allocation and reference count updates in the background The extent allocation tree maintains a reference count and full back reference information for every extent allocated in the filesystem. For subvolume and snapshot trees, every time a block goes through COW, the new copy of the block adds a reference on every block it points to. If a btree node points to 150 leaves, then the COW code needs to go and add backrefs on 150 different extents, which might be spread all over the extent allocation tree. These updates currently happen during btrfs_cow_block, and most COWs happen during btrfs_search_slot. btrfs_search_slot has locks held on both the parent and the node we are COWing, and so we really want to avoid IO during the COW if we can. This commit adds an rbtree of pending reference count updates and extent allocations. The tree is ordered by byte number of the extent and byte number of the parent for the back reference. The tree allows us to: 1) Modify back references in something close to disk order, reducing seeks 2) Significantly reduce the number of modifications made as block pointers are balanced around 3) Do all of the extent insertion and back reference modifications outside of the performance critical btrfs_search_slot code. #3 has the added benefit of greatly reducing the btrfs stack footprint. The extent allocation tree modifications are done without the deep (and somewhat recursive) call chains used in the past. These delayed back reference updates must be done before the transaction commits, and so the rbtree is tied to the transaction. Throttling is implemented to help keep the queue of backrefs at a reasonable size. Since there was a similar mechanism in place for the extent tree extents, that is removed and replaced by the delayed reference tree. Yan Zheng <yan.zheng@oracle.com> helped review and fixup this code. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-03-13 14:10:06 +00:00
ret = btrfs_run_delayed_refs(trans, root, (unsigned long)-1);
BUG_ON(ret);
eb = btrfs_lock_root_node(fs_info->tree_root);
btrfs_cow_block(trans, fs_info->tree_root, eb, NULL, 0, &eb);
btrfs_tree_unlock(eb);
free_extent_buffer(eb);
Btrfs: do extent allocation and reference count updates in the background The extent allocation tree maintains a reference count and full back reference information for every extent allocated in the filesystem. For subvolume and snapshot trees, every time a block goes through COW, the new copy of the block adds a reference on every block it points to. If a btree node points to 150 leaves, then the COW code needs to go and add backrefs on 150 different extents, which might be spread all over the extent allocation tree. These updates currently happen during btrfs_cow_block, and most COWs happen during btrfs_search_slot. btrfs_search_slot has locks held on both the parent and the node we are COWing, and so we really want to avoid IO during the COW if we can. This commit adds an rbtree of pending reference count updates and extent allocations. The tree is ordered by byte number of the extent and byte number of the parent for the back reference. The tree allows us to: 1) Modify back references in something close to disk order, reducing seeks 2) Significantly reduce the number of modifications made as block pointers are balanced around 3) Do all of the extent insertion and back reference modifications outside of the performance critical btrfs_search_slot code. #3 has the added benefit of greatly reducing the btrfs stack footprint. The extent allocation tree modifications are done without the deep (and somewhat recursive) call chains used in the past. These delayed back reference updates must be done before the transaction commits, and so the rbtree is tied to the transaction. Throttling is implemented to help keep the queue of backrefs at a reasonable size. Since there was a similar mechanism in place for the extent tree extents, that is removed and replaced by the delayed reference tree. Yan Zheng <yan.zheng@oracle.com> helped review and fixup this code. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-03-13 14:10:06 +00:00
ret = btrfs_run_delayed_refs(trans, root, (unsigned long)-1);
BUG_ON(ret);
while (!list_empty(&fs_info->dirty_cowonly_roots)) {
next = fs_info->dirty_cowonly_roots.next;
list_del_init(next);
root = list_entry(next, struct btrfs_root, dirty_list);
update_cowonly_root(trans, root);
}
down_write(&fs_info->extent_commit_sem);
switch_commit_root(fs_info->extent_root);
up_write(&fs_info->extent_commit_sem);
return 0;
}
/*
* dead roots are old snapshots that need to be deleted. This allocates
* a dirty root struct and adds it into the list of dead roots that need to
* be deleted
*/
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
int btrfs_add_dead_root(struct btrfs_root *root)
{
mutex_lock(&root->fs_info->trans_mutex);
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
list_add(&root->root_list, &root->fs_info->dead_roots);
mutex_unlock(&root->fs_info->trans_mutex);
return 0;
}
/*
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
* update all the cowonly tree roots on disk
*/
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
static noinline int commit_fs_roots(struct btrfs_trans_handle *trans,
struct btrfs_root *root)
{
struct btrfs_root *gang[8];
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
struct btrfs_fs_info *fs_info = root->fs_info;
int i;
int ret;
int err = 0;
while (1) {
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
ret = radix_tree_gang_lookup_tag(&fs_info->fs_roots_radix,
(void **)gang, 0,
ARRAY_SIZE(gang),
BTRFS_ROOT_TRANS_TAG);
if (ret == 0)
break;
for (i = 0; i < ret; i++) {
root = gang[i];
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
radix_tree_tag_clear(&fs_info->fs_roots_radix,
(unsigned long)root->root_key.objectid,
BTRFS_ROOT_TRANS_TAG);
btrfs_free_log(trans, root);
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
btrfs_update_reloc_root(trans, root);
if (root->commit_root != root->node) {
Btrfs: async block group caching This patch moves the caching of the block group off to a kthread in order to allow people to allocate sooner. Instead of blocking up behind the caching mutex, we instead kick of the caching kthread, and then attempt to make an allocation. If we cannot, we wait on the block groups caching waitqueue, which the caching kthread will wake the waiting threads up everytime it finds 2 meg worth of space, and then again when its finished caching. This is how I tested the speedup from this mkfs the disk mount the disk fill the disk up with fs_mark unmount the disk mount the disk time touch /mnt/foo Without my changes this took 11 seconds on my box, with these changes it now takes 1 second. Another change thats been put in place is we lock the super mirror's in the pinned extent map in order to keep us from adding that stuff as free space when caching the block group. This doesn't really change anything else as far as the pinned extent map is concerned, since for actual pinned extents we use EXTENT_DIRTY, but it does mean that when we unmount we have to go in and unlock those extents to keep from leaking memory. I've also added a check where when we are reading block groups from disk, if the amount of space used == the size of the block group, we go ahead and mark the block group as cached. This drastically reduces the amount of time it takes to cache the block groups. Using the same test as above, except doing a dd to a file and then unmounting, it used to take 33 seconds to umount, now it takes 3 seconds. This version uses the commit_root in the caching kthread, and then keeps track of how many async caching threads are running at any given time so if one of the async threads is still running as we cross transactions we can wait until its finished before handling the pinned extents. Thank you, Signed-off-by: Josef Bacik <jbacik@redhat.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-07-14 01:29:25 +00:00
switch_commit_root(root);
btrfs_set_root_node(&root->root_item,
root->node);
}
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
err = btrfs_update_root(trans, fs_info->tree_root,
&root->root_key,
&root->root_item);
if (err)
break;
}
}
return err;
}
/*
* defrag a given btree. If cacheonly == 1, this won't read from the disk,
* otherwise every leaf in the btree is read and defragged.
*/
int btrfs_defrag_root(struct btrfs_root *root, int cacheonly)
{
struct btrfs_fs_info *info = root->fs_info;
int ret;
struct btrfs_trans_handle *trans;
unsigned long nr;
smp_mb();
if (root->defrag_running)
return 0;
trans = btrfs_start_transaction(root, 1);
while (1) {
root->defrag_running = 1;
ret = btrfs_defrag_leaves(trans, root, cacheonly);
nr = trans->blocks_used;
btrfs_end_transaction(trans, root);
btrfs_btree_balance_dirty(info->tree_root, nr);
cond_resched();
trans = btrfs_start_transaction(root, 1);
if (root->fs_info->closing || ret != -EAGAIN)
break;
}
root->defrag_running = 0;
smp_mb();
btrfs_end_transaction(trans, root);
return 0;
}
#if 0
/*
* when dropping snapshots, we generate a ton of delayed refs, and it makes
* sense not to join the transaction while it is trying to flush the current
* queue of delayed refs out.
*
* This is used by the drop snapshot code only
*/
static noinline int wait_transaction_pre_flush(struct btrfs_fs_info *info)
{
DEFINE_WAIT(wait);
mutex_lock(&info->trans_mutex);
while (info->running_transaction &&
info->running_transaction->delayed_refs.flushing) {
prepare_to_wait(&info->transaction_wait, &wait,
TASK_UNINTERRUPTIBLE);
mutex_unlock(&info->trans_mutex);
schedule();
mutex_lock(&info->trans_mutex);
finish_wait(&info->transaction_wait, &wait);
}
mutex_unlock(&info->trans_mutex);
return 0;
}
/*
* Given a list of roots that need to be deleted, call btrfs_drop_snapshot on
* all of them
*/
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
int btrfs_drop_dead_root(struct btrfs_root *root)
{
struct btrfs_trans_handle *trans;
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
struct btrfs_root *tree_root = root->fs_info->tree_root;
unsigned long nr;
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
int ret;
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
while (1) {
/*
* we don't want to jump in and create a bunch of
* delayed refs if the transaction is starting to close
*/
wait_transaction_pre_flush(tree_root->fs_info);
trans = btrfs_start_transaction(tree_root, 1);
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
/*
* we've joined a transaction, make sure it isn't
* closing right now
*/
if (trans->transaction->delayed_refs.flushing) {
btrfs_end_transaction(trans, tree_root);
continue;
}
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
ret = btrfs_drop_snapshot(trans, root);
if (ret != -EAGAIN)
break;
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
ret = btrfs_update_root(trans, tree_root,
&root->root_key,
&root->root_item);
if (ret)
break;
nr = trans->blocks_used;
ret = btrfs_end_transaction(trans, tree_root);
BUG_ON(ret);
btrfs_btree_balance_dirty(tree_root, nr);
cond_resched();
}
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
BUG_ON(ret);
ret = btrfs_del_root(trans, tree_root, &root->root_key);
BUG_ON(ret);
nr = trans->blocks_used;
ret = btrfs_end_transaction(trans, tree_root);
BUG_ON(ret);
free_extent_buffer(root->node);
free_extent_buffer(root->commit_root);
kfree(root);
btrfs_btree_balance_dirty(tree_root, nr);
return ret;
}
#endif
/*
* new snapshots need to be created at a very specific time in the
* transaction commit. This does the actual creation
*/
static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans,
struct btrfs_fs_info *fs_info,
struct btrfs_pending_snapshot *pending)
{
struct btrfs_key key;
struct btrfs_root_item *new_root_item;
struct btrfs_root *tree_root = fs_info->tree_root;
struct btrfs_root *root = pending->root;
struct extent_buffer *tmp;
struct extent_buffer *old;
int ret;
u64 objectid;
new_root_item = kmalloc(sizeof(*new_root_item), GFP_NOFS);
if (!new_root_item) {
ret = -ENOMEM;
goto fail;
}
ret = btrfs_find_free_objectid(trans, tree_root, 0, &objectid);
if (ret)
goto fail;
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
record_root_in_trans(trans, root);
btrfs_set_root_last_snapshot(&root->root_item, trans->transid);
memcpy(new_root_item, &root->root_item, sizeof(*new_root_item));
key.objectid = objectid;
/* record when the snapshot was created in key.offset */
key.offset = trans->transid;
btrfs_set_key_type(&key, BTRFS_ROOT_ITEM_KEY);
old = btrfs_lock_root_node(root);
btrfs_cow_block(trans, root, old, NULL, 0, &old);
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
btrfs_set_lock_blocking(old);
btrfs_copy_root(trans, root, old, &tmp, objectid);
btrfs_tree_unlock(old);
free_extent_buffer(old);
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
btrfs_set_root_node(new_root_item, tmp);
ret = btrfs_insert_root(trans, root->fs_info->tree_root, &key,
new_root_item);
btrfs_tree_unlock(tmp);
free_extent_buffer(tmp);
if (ret)
goto fail;
key.offset = (u64)-1;
memcpy(&pending->root_key, &key, sizeof(key));
fail:
kfree(new_root_item);
return ret;
}
static noinline int finish_pending_snapshot(struct btrfs_fs_info *fs_info,
struct btrfs_pending_snapshot *pending)
{
int ret;
int namelen;
u64 index = 0;
struct btrfs_trans_handle *trans;
struct inode *parent_inode;
struct btrfs_root *parent_root;
parent_inode = pending->dentry->d_parent->d_inode;
parent_root = BTRFS_I(parent_inode)->root;
trans = btrfs_join_transaction(parent_root, 1);
/*
* insert the directory item
*/
namelen = strlen(pending->name);
ret = btrfs_set_inode_index(parent_inode, &index);
ret = btrfs_insert_dir_item(trans, parent_root,
pending->name, namelen,
parent_inode->i_ino,
&pending->root_key, BTRFS_FT_DIR, index);
if (ret)
goto fail;
btrfs_i_size_write(parent_inode, parent_inode->i_size + namelen * 2);
ret = btrfs_update_inode(trans, parent_root, parent_inode);
BUG_ON(ret);
ret = btrfs_add_root_ref(trans, parent_root->fs_info->tree_root,
pending->root_key.objectid,
parent_root->root_key.objectid,
parent_inode->i_ino, index, pending->name,
namelen);
BUG_ON(ret);
fail:
btrfs_end_transaction(trans, fs_info->fs_root);
return ret;
}
/*
* create all the snapshots we've scheduled for creation
*/
static noinline int create_pending_snapshots(struct btrfs_trans_handle *trans,
struct btrfs_fs_info *fs_info)
{
struct btrfs_pending_snapshot *pending;
struct list_head *head = &trans->transaction->pending_snapshots;
int ret;
list_for_each_entry(pending, head, list) {
ret = create_pending_snapshot(trans, fs_info, pending);
BUG_ON(ret);
}
return 0;
}
static noinline int finish_pending_snapshots(struct btrfs_trans_handle *trans,
struct btrfs_fs_info *fs_info)
{
struct btrfs_pending_snapshot *pending;
struct list_head *head = &trans->transaction->pending_snapshots;
int ret;
while (!list_empty(head)) {
pending = list_entry(head->next,
struct btrfs_pending_snapshot, list);
ret = finish_pending_snapshot(fs_info, pending);
BUG_ON(ret);
list_del(&pending->list);
kfree(pending->name);
kfree(pending);
}
return 0;
}
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
static void update_super_roots(struct btrfs_root *root)
{
struct btrfs_root_item *root_item;
struct btrfs_super_block *super;
super = &root->fs_info->super_copy;
root_item = &root->fs_info->chunk_root->root_item;
super->chunk_root = root_item->bytenr;
super->chunk_root_generation = root_item->generation;
super->chunk_root_level = root_item->level;
root_item = &root->fs_info->tree_root->root_item;
super->root = root_item->bytenr;
super->generation = root_item->generation;
super->root_level = root_item->level;
}
int btrfs_transaction_in_commit(struct btrfs_fs_info *info)
{
int ret = 0;
spin_lock(&info->new_trans_lock);
if (info->running_transaction)
ret = info->running_transaction->in_commit;
spin_unlock(&info->new_trans_lock);
return ret;
}
int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
struct btrfs_root *root)
{
unsigned long joined = 0;
unsigned long timeout = 1;
struct btrfs_transaction *cur_trans;
struct btrfs_transaction *prev_trans = NULL;
DEFINE_WAIT(wait);
int ret;
int should_grow = 0;
unsigned long now = get_seconds();
int flush_on_commit = btrfs_test_opt(root, FLUSHONCOMMIT);
Btrfs: add extra flushing for renames and truncates Renames and truncates are both common ways to replace old data with new data. The filesystem can make an effort to make sure the new data is on disk before actually replacing the old data. This is especially important for rename, which many application use as though it were atomic for both the data and the metadata involved. The current btrfs code will happily replace a file that is fully on disk with one that was just created and still has pending IO. If we crash after transaction commit but before the IO is done, we'll end up replacing a good file with a zero length file. The solution used here is to create a list of inodes that need special ordering and force them to disk before the commit is done. This is similar to the ext3 style data=ordering, except it is only done on selected files. Btrfs is able to get away with this because it does not wait on commits very often, even for fsync (which use a sub-commit). For renames, we order the file when it wasn't already on disk and when it is replacing an existing file. Larger files are sent to filemap_flush right away (before the transaction handle is opened). For truncates, we order if the file goes from non-zero size down to zero size. This is a little different, because at the time of the truncate the file has no dirty bytes to order. But, we flag the inode so that it is added to the ordered list on close (via release method). We also immediately add it to the ordered list of the current transaction so that we can try to flush down any writes the application sneaks in before commit. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-03-31 17:27:11 +00:00
btrfs_run_ordered_operations(root, 0);
Btrfs: do extent allocation and reference count updates in the background The extent allocation tree maintains a reference count and full back reference information for every extent allocated in the filesystem. For subvolume and snapshot trees, every time a block goes through COW, the new copy of the block adds a reference on every block it points to. If a btree node points to 150 leaves, then the COW code needs to go and add backrefs on 150 different extents, which might be spread all over the extent allocation tree. These updates currently happen during btrfs_cow_block, and most COWs happen during btrfs_search_slot. btrfs_search_slot has locks held on both the parent and the node we are COWing, and so we really want to avoid IO during the COW if we can. This commit adds an rbtree of pending reference count updates and extent allocations. The tree is ordered by byte number of the extent and byte number of the parent for the back reference. The tree allows us to: 1) Modify back references in something close to disk order, reducing seeks 2) Significantly reduce the number of modifications made as block pointers are balanced around 3) Do all of the extent insertion and back reference modifications outside of the performance critical btrfs_search_slot code. #3 has the added benefit of greatly reducing the btrfs stack footprint. The extent allocation tree modifications are done without the deep (and somewhat recursive) call chains used in the past. These delayed back reference updates must be done before the transaction commits, and so the rbtree is tied to the transaction. Throttling is implemented to help keep the queue of backrefs at a reasonable size. Since there was a similar mechanism in place for the extent tree extents, that is removed and replaced by the delayed reference tree. Yan Zheng <yan.zheng@oracle.com> helped review and fixup this code. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-03-13 14:10:06 +00:00
/* make a pass through all the delayed refs we have so far
* any runnings procs may add more while we are here
*/
ret = btrfs_run_delayed_refs(trans, root, 0);
BUG_ON(ret);
cur_trans = trans->transaction;
Btrfs: do extent allocation and reference count updates in the background The extent allocation tree maintains a reference count and full back reference information for every extent allocated in the filesystem. For subvolume and snapshot trees, every time a block goes through COW, the new copy of the block adds a reference on every block it points to. If a btree node points to 150 leaves, then the COW code needs to go and add backrefs on 150 different extents, which might be spread all over the extent allocation tree. These updates currently happen during btrfs_cow_block, and most COWs happen during btrfs_search_slot. btrfs_search_slot has locks held on both the parent and the node we are COWing, and so we really want to avoid IO during the COW if we can. This commit adds an rbtree of pending reference count updates and extent allocations. The tree is ordered by byte number of the extent and byte number of the parent for the back reference. The tree allows us to: 1) Modify back references in something close to disk order, reducing seeks 2) Significantly reduce the number of modifications made as block pointers are balanced around 3) Do all of the extent insertion and back reference modifications outside of the performance critical btrfs_search_slot code. #3 has the added benefit of greatly reducing the btrfs stack footprint. The extent allocation tree modifications are done without the deep (and somewhat recursive) call chains used in the past. These delayed back reference updates must be done before the transaction commits, and so the rbtree is tied to the transaction. Throttling is implemented to help keep the queue of backrefs at a reasonable size. Since there was a similar mechanism in place for the extent tree extents, that is removed and replaced by the delayed reference tree. Yan Zheng <yan.zheng@oracle.com> helped review and fixup this code. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-03-13 14:10:06 +00:00
/*
* set the flushing flag so procs in this transaction have to
* start sending their work down.
*/
cur_trans->delayed_refs.flushing = 1;
Btrfs: do extent allocation and reference count updates in the background The extent allocation tree maintains a reference count and full back reference information for every extent allocated in the filesystem. For subvolume and snapshot trees, every time a block goes through COW, the new copy of the block adds a reference on every block it points to. If a btree node points to 150 leaves, then the COW code needs to go and add backrefs on 150 different extents, which might be spread all over the extent allocation tree. These updates currently happen during btrfs_cow_block, and most COWs happen during btrfs_search_slot. btrfs_search_slot has locks held on both the parent and the node we are COWing, and so we really want to avoid IO during the COW if we can. This commit adds an rbtree of pending reference count updates and extent allocations. The tree is ordered by byte number of the extent and byte number of the parent for the back reference. The tree allows us to: 1) Modify back references in something close to disk order, reducing seeks 2) Significantly reduce the number of modifications made as block pointers are balanced around 3) Do all of the extent insertion and back reference modifications outside of the performance critical btrfs_search_slot code. #3 has the added benefit of greatly reducing the btrfs stack footprint. The extent allocation tree modifications are done without the deep (and somewhat recursive) call chains used in the past. These delayed back reference updates must be done before the transaction commits, and so the rbtree is tied to the transaction. Throttling is implemented to help keep the queue of backrefs at a reasonable size. Since there was a similar mechanism in place for the extent tree extents, that is removed and replaced by the delayed reference tree. Yan Zheng <yan.zheng@oracle.com> helped review and fixup this code. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-03-13 14:10:06 +00:00
ret = btrfs_run_delayed_refs(trans, root, 0);
Btrfs: do extent allocation and reference count updates in the background The extent allocation tree maintains a reference count and full back reference information for every extent allocated in the filesystem. For subvolume and snapshot trees, every time a block goes through COW, the new copy of the block adds a reference on every block it points to. If a btree node points to 150 leaves, then the COW code needs to go and add backrefs on 150 different extents, which might be spread all over the extent allocation tree. These updates currently happen during btrfs_cow_block, and most COWs happen during btrfs_search_slot. btrfs_search_slot has locks held on both the parent and the node we are COWing, and so we really want to avoid IO during the COW if we can. This commit adds an rbtree of pending reference count updates and extent allocations. The tree is ordered by byte number of the extent and byte number of the parent for the back reference. The tree allows us to: 1) Modify back references in something close to disk order, reducing seeks 2) Significantly reduce the number of modifications made as block pointers are balanced around 3) Do all of the extent insertion and back reference modifications outside of the performance critical btrfs_search_slot code. #3 has the added benefit of greatly reducing the btrfs stack footprint. The extent allocation tree modifications are done without the deep (and somewhat recursive) call chains used in the past. These delayed back reference updates must be done before the transaction commits, and so the rbtree is tied to the transaction. Throttling is implemented to help keep the queue of backrefs at a reasonable size. Since there was a similar mechanism in place for the extent tree extents, that is removed and replaced by the delayed reference tree. Yan Zheng <yan.zheng@oracle.com> helped review and fixup this code. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-03-13 14:10:06 +00:00
BUG_ON(ret);
mutex_lock(&root->fs_info->trans_mutex);
if (cur_trans->in_commit) {
cur_trans->use_count++;
mutex_unlock(&root->fs_info->trans_mutex);
btrfs_end_transaction(trans, root);
ret = wait_for_commit(root, cur_trans);
BUG_ON(ret);
mutex_lock(&root->fs_info->trans_mutex);
put_transaction(cur_trans);
mutex_unlock(&root->fs_info->trans_mutex);
return 0;
}
trans->transaction->in_commit = 1;
trans->transaction->blocked = 1;
if (cur_trans->list.prev != &root->fs_info->trans_list) {
prev_trans = list_entry(cur_trans->list.prev,
struct btrfs_transaction, list);
if (!prev_trans->commit_done) {
prev_trans->use_count++;
mutex_unlock(&root->fs_info->trans_mutex);
wait_for_commit(root, prev_trans);
mutex_lock(&root->fs_info->trans_mutex);
put_transaction(prev_trans);
}
}
if (now < cur_trans->start_time || now - cur_trans->start_time < 1)
should_grow = 1;
do {
int snap_pending = 0;
joined = cur_trans->num_joined;
if (!list_empty(&trans->transaction->pending_snapshots))
snap_pending = 1;
WARN_ON(cur_trans != trans->transaction);
prepare_to_wait(&cur_trans->writer_wait, &wait,
TASK_UNINTERRUPTIBLE);
if (cur_trans->num_writers > 1)
timeout = MAX_SCHEDULE_TIMEOUT;
else if (should_grow)
timeout = 1;
mutex_unlock(&root->fs_info->trans_mutex);
if (flush_on_commit || snap_pending) {
btrfs_start_delalloc_inodes(root, 1);
ret = btrfs_wait_ordered_extents(root, 0, 1);
BUG_ON(ret);
}
Btrfs: add extra flushing for renames and truncates Renames and truncates are both common ways to replace old data with new data. The filesystem can make an effort to make sure the new data is on disk before actually replacing the old data. This is especially important for rename, which many application use as though it were atomic for both the data and the metadata involved. The current btrfs code will happily replace a file that is fully on disk with one that was just created and still has pending IO. If we crash after transaction commit but before the IO is done, we'll end up replacing a good file with a zero length file. The solution used here is to create a list of inodes that need special ordering and force them to disk before the commit is done. This is similar to the ext3 style data=ordering, except it is only done on selected files. Btrfs is able to get away with this because it does not wait on commits very often, even for fsync (which use a sub-commit). For renames, we order the file when it wasn't already on disk and when it is replacing an existing file. Larger files are sent to filemap_flush right away (before the transaction handle is opened). For truncates, we order if the file goes from non-zero size down to zero size. This is a little different, because at the time of the truncate the file has no dirty bytes to order. But, we flag the inode so that it is added to the ordered list on close (via release method). We also immediately add it to the ordered list of the current transaction so that we can try to flush down any writes the application sneaks in before commit. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-03-31 17:27:11 +00:00
/*
* rename don't use btrfs_join_transaction, so, once we
* set the transaction to blocked above, we aren't going
* to get any new ordered operations. We can safely run
* it here and no for sure that nothing new will be added
* to the list
*/
btrfs_run_ordered_operations(root, 1);
smp_mb();
if (cur_trans->num_writers > 1 || should_grow)
schedule_timeout(timeout);
mutex_lock(&root->fs_info->trans_mutex);
finish_wait(&cur_trans->writer_wait, &wait);
} while (cur_trans->num_writers > 1 ||
(should_grow && cur_trans->num_joined != joined));
ret = create_pending_snapshots(trans, root->fs_info);
BUG_ON(ret);
Btrfs: do extent allocation and reference count updates in the background The extent allocation tree maintains a reference count and full back reference information for every extent allocated in the filesystem. For subvolume and snapshot trees, every time a block goes through COW, the new copy of the block adds a reference on every block it points to. If a btree node points to 150 leaves, then the COW code needs to go and add backrefs on 150 different extents, which might be spread all over the extent allocation tree. These updates currently happen during btrfs_cow_block, and most COWs happen during btrfs_search_slot. btrfs_search_slot has locks held on both the parent and the node we are COWing, and so we really want to avoid IO during the COW if we can. This commit adds an rbtree of pending reference count updates and extent allocations. The tree is ordered by byte number of the extent and byte number of the parent for the back reference. The tree allows us to: 1) Modify back references in something close to disk order, reducing seeks 2) Significantly reduce the number of modifications made as block pointers are balanced around 3) Do all of the extent insertion and back reference modifications outside of the performance critical btrfs_search_slot code. #3 has the added benefit of greatly reducing the btrfs stack footprint. The extent allocation tree modifications are done without the deep (and somewhat recursive) call chains used in the past. These delayed back reference updates must be done before the transaction commits, and so the rbtree is tied to the transaction. Throttling is implemented to help keep the queue of backrefs at a reasonable size. Since there was a similar mechanism in place for the extent tree extents, that is removed and replaced by the delayed reference tree. Yan Zheng <yan.zheng@oracle.com> helped review and fixup this code. Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-03-13 14:10:06 +00:00
ret = btrfs_run_delayed_refs(trans, root, (unsigned long)-1);
BUG_ON(ret);
WARN_ON(cur_trans != trans->transaction);
/* btrfs_commit_tree_roots is responsible for getting the
* various roots consistent with each other. Every pointer
* in the tree of tree roots has to point to the most up to date
* root for every subvolume and other tree. So, we have to keep
* the tree logging code from jumping in and changing any
* of the trees.
*
* At this point in the commit, there can't be any tree-log
* writers, but a little lower down we drop the trans mutex
* and let new people in. By holding the tree_log_mutex
* from now until after the super is written, we avoid races
* with the tree-log code.
*/
mutex_lock(&root->fs_info->tree_log_mutex);
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
ret = commit_fs_roots(trans, root);
BUG_ON(ret);
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
/* commit_fs_roots gets rid of all the tree log roots, it is now
* safe to free the root of tree log roots
*/
btrfs_free_log_root_tree(trans, root->fs_info);
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
ret = commit_cowonly_roots(trans, root);
BUG_ON(ret);
btrfs_prepare_extent_commit(trans, root);
cur_trans = root->fs_info->running_transaction;
spin_lock(&root->fs_info->new_trans_lock);
root->fs_info->running_transaction = NULL;
spin_unlock(&root->fs_info->new_trans_lock);
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
btrfs_set_root_node(&root->fs_info->tree_root->root_item,
root->fs_info->tree_root->node);
Btrfs: async block group caching This patch moves the caching of the block group off to a kthread in order to allow people to allocate sooner. Instead of blocking up behind the caching mutex, we instead kick of the caching kthread, and then attempt to make an allocation. If we cannot, we wait on the block groups caching waitqueue, which the caching kthread will wake the waiting threads up everytime it finds 2 meg worth of space, and then again when its finished caching. This is how I tested the speedup from this mkfs the disk mount the disk fill the disk up with fs_mark unmount the disk mount the disk time touch /mnt/foo Without my changes this took 11 seconds on my box, with these changes it now takes 1 second. Another change thats been put in place is we lock the super mirror's in the pinned extent map in order to keep us from adding that stuff as free space when caching the block group. This doesn't really change anything else as far as the pinned extent map is concerned, since for actual pinned extents we use EXTENT_DIRTY, but it does mean that when we unmount we have to go in and unlock those extents to keep from leaking memory. I've also added a check where when we are reading block groups from disk, if the amount of space used == the size of the block group, we go ahead and mark the block group as cached. This drastically reduces the amount of time it takes to cache the block groups. Using the same test as above, except doing a dd to a file and then unmounting, it used to take 33 seconds to umount, now it takes 3 seconds. This version uses the commit_root in the caching kthread, and then keeps track of how many async caching threads are running at any given time so if one of the async threads is still running as we cross transactions we can wait until its finished before handling the pinned extents. Thank you, Signed-off-by: Josef Bacik <jbacik@redhat.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-07-14 01:29:25 +00:00
switch_commit_root(root->fs_info->tree_root);
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
btrfs_set_root_node(&root->fs_info->chunk_root->root_item,
root->fs_info->chunk_root->node);
Btrfs: async block group caching This patch moves the caching of the block group off to a kthread in order to allow people to allocate sooner. Instead of blocking up behind the caching mutex, we instead kick of the caching kthread, and then attempt to make an allocation. If we cannot, we wait on the block groups caching waitqueue, which the caching kthread will wake the waiting threads up everytime it finds 2 meg worth of space, and then again when its finished caching. This is how I tested the speedup from this mkfs the disk mount the disk fill the disk up with fs_mark unmount the disk mount the disk time touch /mnt/foo Without my changes this took 11 seconds on my box, with these changes it now takes 1 second. Another change thats been put in place is we lock the super mirror's in the pinned extent map in order to keep us from adding that stuff as free space when caching the block group. This doesn't really change anything else as far as the pinned extent map is concerned, since for actual pinned extents we use EXTENT_DIRTY, but it does mean that when we unmount we have to go in and unlock those extents to keep from leaking memory. I've also added a check where when we are reading block groups from disk, if the amount of space used == the size of the block group, we go ahead and mark the block group as cached. This drastically reduces the amount of time it takes to cache the block groups. Using the same test as above, except doing a dd to a file and then unmounting, it used to take 33 seconds to umount, now it takes 3 seconds. This version uses the commit_root in the caching kthread, and then keeps track of how many async caching threads are running at any given time so if one of the async threads is still running as we cross transactions we can wait until its finished before handling the pinned extents. Thank you, Signed-off-by: Josef Bacik <jbacik@redhat.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-07-14 01:29:25 +00:00
switch_commit_root(root->fs_info->chunk_root);
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
update_super_roots(root);
if (!root->fs_info->log_root_recovering) {
btrfs_set_super_log_root(&root->fs_info->super_copy, 0);
btrfs_set_super_log_root_level(&root->fs_info->super_copy, 0);
}
memcpy(&root->fs_info->super_for_commit, &root->fs_info->super_copy,
sizeof(root->fs_info->super_copy));
trans->transaction->blocked = 0;
wake_up(&root->fs_info->transaction_wait);
mutex_unlock(&root->fs_info->trans_mutex);
ret = btrfs_write_and_wait_transaction(trans, root);
BUG_ON(ret);
write_ctree_super(trans, root, 0);
/*
* the super is written, we can safely allow the tree-loggers
* to go about their business
*/
mutex_unlock(&root->fs_info->tree_log_mutex);
btrfs_finish_extent_commit(trans, root);
/* do the directory inserts of any pending snapshot creations */
finish_pending_snapshots(trans, root->fs_info);
2008-09-26 14:09:34 +00:00
mutex_lock(&root->fs_info->trans_mutex);
cur_trans->commit_done = 1;
root->fs_info->last_trans_committed = cur_trans->transid;
Btrfs: async block group caching This patch moves the caching of the block group off to a kthread in order to allow people to allocate sooner. Instead of blocking up behind the caching mutex, we instead kick of the caching kthread, and then attempt to make an allocation. If we cannot, we wait on the block groups caching waitqueue, which the caching kthread will wake the waiting threads up everytime it finds 2 meg worth of space, and then again when its finished caching. This is how I tested the speedup from this mkfs the disk mount the disk fill the disk up with fs_mark unmount the disk mount the disk time touch /mnt/foo Without my changes this took 11 seconds on my box, with these changes it now takes 1 second. Another change thats been put in place is we lock the super mirror's in the pinned extent map in order to keep us from adding that stuff as free space when caching the block group. This doesn't really change anything else as far as the pinned extent map is concerned, since for actual pinned extents we use EXTENT_DIRTY, but it does mean that when we unmount we have to go in and unlock those extents to keep from leaking memory. I've also added a check where when we are reading block groups from disk, if the amount of space used == the size of the block group, we go ahead and mark the block group as cached. This drastically reduces the amount of time it takes to cache the block groups. Using the same test as above, except doing a dd to a file and then unmounting, it used to take 33 seconds to umount, now it takes 3 seconds. This version uses the commit_root in the caching kthread, and then keeps track of how many async caching threads are running at any given time so if one of the async threads is still running as we cross transactions we can wait until its finished before handling the pinned extents. Thank you, Signed-off-by: Josef Bacik <jbacik@redhat.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-07-14 01:29:25 +00:00
wake_up(&cur_trans->commit_wait);
put_transaction(cur_trans);
put_transaction(cur_trans);
mutex_unlock(&root->fs_info->trans_mutex);
Btrfs: proper -ENOSPC handling At the start of a transaction we do a btrfs_reserve_metadata_space() and specify how many items we plan on modifying. Then once we've done our modifications and such, just call btrfs_unreserve_metadata_space() for the same number of items we reserved. For keeping track of metadata needed for data I've had to add an extent_io op for when we merge extents. This lets us track space properly when we are doing sequential writes, so we don't end up reserving way more metadata space than what we need. The only place where the metadata space accounting is not done is in the relocation code. This is because Yan is going to be reworking that code in the near future, so running btrfs-vol -b could still possibly result in a ENOSPC related panic. This patch also turns off the metadata_ratio stuff in order to allow users to more efficiently use their disk space. This patch makes it so we track how much metadata we need for an inode's delayed allocation extents by tracking how many extents are currently waiting for allocation. It introduces two new callbacks for the extent_io tree's, merge_extent_hook and split_extent_hook. These help us keep track of when we merge delalloc extents together and split them up. Reservations are handled prior to any actually dirty'ing occurs, and then we unreserve after we dirty. btrfs_unreserve_metadata_for_delalloc() will make the appropriate unreservations as needed based on the number of reservations we currently have and the number of extents we currently have. Doing the reservation outside of doing any of the actual dirty'ing lets us do things like filemap_flush() the inode to try and force delalloc to happen, or as a last resort actually start allocation on all delalloc inodes in the fs. This has survived dbench, fs_mark and an fsx torture test. Signed-off-by: Josef Bacik <jbacik@redhat.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-09-11 20:12:44 +00:00
if (current->journal_info == trans)
current->journal_info = NULL;
kmem_cache_free(btrfs_trans_handle_cachep, trans);
if (current != root->fs_info->transaction_kthread)
btrfs_run_delayed_iputs(root);
return ret;
}
/*
* interface function to delete all the snapshots we have scheduled for deletion
*/
int btrfs_clean_old_snapshots(struct btrfs_root *root)
{
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
LIST_HEAD(list);
struct btrfs_fs_info *fs_info = root->fs_info;
mutex_lock(&fs_info->trans_mutex);
list_splice_init(&fs_info->dead_roots, &list);
mutex_unlock(&fs_info->trans_mutex);
Btrfs: Mixed back reference (FORWARD ROLLING FORMAT CHANGE) This commit introduces a new kind of back reference for btrfs metadata. Once a filesystem has been mounted with this commit, IT WILL NO LONGER BE MOUNTABLE BY OLDER KERNELS. When a tree block in subvolume tree is cow'd, the reference counts of all extents it points to are increased by one. At transaction commit time, the old root of the subvolume is recorded in a "dead root" data structure, and the btree it points to is later walked, dropping reference counts and freeing any blocks where the reference count goes to 0. The increments done during cow and decrements done after commit cancel out, and the walk is a very expensive way to go about freeing the blocks that are no longer referenced by the new btree root. This commit reduces the transaction overhead by avoiding the need for dead root records. When a non-shared tree block is cow'd, we free the old block at once, and the new block inherits old block's references. When a tree block with reference count > 1 is cow'd, we increase the reference counts of all extents the new block points to by one, and decrease the old block's reference count by one. This dead tree avoidance code removes the need to modify the reference counts of lower level extents when a non-shared tree block is cow'd. But we still need to update back ref for all pointers in the block. This is because the location of the block is recorded in the back ref item. We can solve this by introducing a new type of back ref. The new back ref provides information about pointer's key, level and in which tree the pointer lives. This information allow us to find the pointer by searching the tree. The shortcoming of the new back ref is that it only works for pointers in tree blocks referenced by their owner trees. This is mostly a problem for snapshots, where resolving one of these fuzzy back references would be O(number_of_snapshots) and quite slow. The solution used here is to use the fuzzy back references in the common case where a given tree block is only referenced by one root, and use the full back references when multiple roots have a reference on a given block. This commit adds per subvolume red-black tree to keep trace of cached inodes. The red-black tree helps the balancing code to find cached inodes whose inode numbers within a given range. This commit improves the balancing code by introducing several data structures to keep the state of balancing. The most important one is the back ref cache. It caches how the upper level tree blocks are referenced. This greatly reduce the overhead of checking back ref. The improved balancing code scales significantly better with a large number of snapshots. This is a very large commit and was written in a number of pieces. But, they depend heavily on the disk format change and were squashed together to make sure git bisect didn't end up in a bad state wrt space balancing or the format change. Signed-off-by: Yan Zheng <zheng.yan@oracle.com> Signed-off-by: Chris Mason <chris.mason@oracle.com>
2009-06-10 14:45:14 +00:00
while (!list_empty(&list)) {
root = list_entry(list.next, struct btrfs_root, root_list);
list_del(&root->root_list);
if (btrfs_header_backref_rev(root->node) <
BTRFS_MIXED_BACKREF_REV)
btrfs_drop_snapshot(root, 0);
else
btrfs_drop_snapshot(root, 1);
}
return 0;
}