linux/fs/ext4/extents.c

4351 lines
113 KiB
C
Raw Normal View History

/*
* Copyright (c) 2003-2006, Cluster File Systems, Inc, info@clusterfs.com
* Written by Alex Tomas <alex@clusterfs.com>
*
* Architecture independence:
* Copyright (c) 2005, Bull S.A.
* Written by Pierre Peiffer <pierre.peiffer@bull.net>
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License version 2 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 Licens
* along with this program; if not, write to the Free Software
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-
*/
/*
* Extents support for EXT4
*
* TODO:
* - ext4*_error() should be used in some situations
* - analyze all BUG()/BUG_ON(), use -EIO where appropriate
* - smart tree reduction
*/
#include <linux/module.h>
#include <linux/fs.h>
#include <linux/time.h>
#include <linux/jbd2.h>
#include <linux/highuid.h>
#include <linux/pagemap.h>
#include <linux/quotaops.h>
#include <linux/string.h>
#include <linux/slab.h>
#include <linux/falloc.h>
#include <asm/uaccess.h>
#include <linux/fiemap.h>
#include "ext4_jbd2.h"
#include "ext4_extents.h"
#include <trace/events/ext4.h>
static int ext4_split_extent(handle_t *handle,
struct inode *inode,
struct ext4_ext_path *path,
struct ext4_map_blocks *map,
int split_flag,
int flags);
static int ext4_ext_truncate_extend_restart(handle_t *handle,
struct inode *inode,
int needed)
{
int err;
if (!ext4_handle_valid(handle))
return 0;
if (handle->h_buffer_credits > needed)
return 0;
err = ext4_journal_extend(handle, needed);
if (err <= 0)
return err;
err = ext4_truncate_restart_trans(handle, inode, needed);
if (err == 0)
err = -EAGAIN;
return err;
}
/*
* could return:
* - EROFS
* - ENOMEM
*/
static int ext4_ext_get_access(handle_t *handle, struct inode *inode,
struct ext4_ext_path *path)
{
if (path->p_bh) {
/* path points to block */
return ext4_journal_get_write_access(handle, path->p_bh);
}
/* path points to leaf/index in inode body */
/* we use in-core data, no need to protect them */
return 0;
}
/*
* could return:
* - EROFS
* - ENOMEM
* - EIO
*/
static int ext4_ext_dirty(handle_t *handle, struct inode *inode,
struct ext4_ext_path *path)
{
int err;
if (path->p_bh) {
/* path points to block */
err = ext4_handle_dirty_metadata(handle, inode, path->p_bh);
} else {
/* path points to leaf/index in inode body */
err = ext4_mark_inode_dirty(handle, inode);
}
return err;
}
static ext4_fsblk_t ext4_ext_find_goal(struct inode *inode,
struct ext4_ext_path *path,
ext4_lblk_t block)
{
int depth;
if (path) {
struct ext4_extent *ex;
depth = path->p_depth;
ext4: fix 32bit overflow in ext4_ext_find_goal() ext4_ext_find_goal() returns an ideal physical block number that the block allocator tries to allocate first. However, if a required file offset is smaller than the existing extent's one, ext4_ext_find_goal() returns a wrong block number because it may overflow at "block - le32_to_cpu(ex->ee_block)". This patch fixes the problem. ext4_ext_find_goal() will also return a wrong block number in case a file offset of the existing extent is too big. In this case, the ideal physical block number is fixed in ext4_mb_initialize_context(), so it's no problem. reproduce: # dd if=/dev/zero of=/mnt/mp1/tmp bs=127M count=1 oflag=sync # dd if=/dev/zero of=/mnt/mp1/file bs=512K count=1 seek=1 oflag=sync # filefrag -v /mnt/mp1/file Filesystem type is: ef53 File size of /mnt/mp1/file is 1048576 (256 blocks, blocksize 4096) ext logical physical expected length flags 0 128 67456 128 eof /mnt/mp1/file: 2 extents found # rm -rf /mnt/mp1/tmp # echo $((512*4096)) > /sys/fs/ext4/loop0/mb_stream_req # dd if=/dev/zero of=/mnt/mp1/file bs=512K count=1 oflag=sync conv=notrunc result (linux-2.6.37-rc2 + ext4 patch queue): # filefrag -v /mnt/mp1/file Filesystem type is: ef53 File size of /mnt/mp1/file is 1048576 (256 blocks, blocksize 4096) ext logical physical expected length flags 0 0 33280 128 1 128 67456 33407 128 eof /mnt/mp1/file: 2 extents found result(apply this patch): # filefrag -v /mnt/mp1/file Filesystem type is: ef53 File size of /mnt/mp1/file is 1048576 (256 blocks, blocksize 4096) ext logical physical expected length flags 0 0 66560 128 1 128 67456 66687 128 eof /mnt/mp1/file: 2 extents found Signed-off-by: Kazuya Mio <k-mio@sx.jp.nec.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-01-10 17:12:28 +00:00
/*
* Try to predict block placement assuming that we are
* filling in a file which will eventually be
* non-sparse --- i.e., in the case of libbfd writing
* an ELF object sections out-of-order but in a way
* the eventually results in a contiguous object or
* executable file, or some database extending a table
* space file. However, this is actually somewhat
* non-ideal if we are writing a sparse file such as
* qemu or KVM writing a raw image file that is going
* to stay fairly sparse, since it will end up
* fragmenting the file system's free space. Maybe we
* should have some hueristics or some way to allow
* userspace to pass a hint to file system,
* especially if the latter case turns out to be
ext4: fix 32bit overflow in ext4_ext_find_goal() ext4_ext_find_goal() returns an ideal physical block number that the block allocator tries to allocate first. However, if a required file offset is smaller than the existing extent's one, ext4_ext_find_goal() returns a wrong block number because it may overflow at "block - le32_to_cpu(ex->ee_block)". This patch fixes the problem. ext4_ext_find_goal() will also return a wrong block number in case a file offset of the existing extent is too big. In this case, the ideal physical block number is fixed in ext4_mb_initialize_context(), so it's no problem. reproduce: # dd if=/dev/zero of=/mnt/mp1/tmp bs=127M count=1 oflag=sync # dd if=/dev/zero of=/mnt/mp1/file bs=512K count=1 seek=1 oflag=sync # filefrag -v /mnt/mp1/file Filesystem type is: ef53 File size of /mnt/mp1/file is 1048576 (256 blocks, blocksize 4096) ext logical physical expected length flags 0 128 67456 128 eof /mnt/mp1/file: 2 extents found # rm -rf /mnt/mp1/tmp # echo $((512*4096)) > /sys/fs/ext4/loop0/mb_stream_req # dd if=/dev/zero of=/mnt/mp1/file bs=512K count=1 oflag=sync conv=notrunc result (linux-2.6.37-rc2 + ext4 patch queue): # filefrag -v /mnt/mp1/file Filesystem type is: ef53 File size of /mnt/mp1/file is 1048576 (256 blocks, blocksize 4096) ext logical physical expected length flags 0 0 33280 128 1 128 67456 33407 128 eof /mnt/mp1/file: 2 extents found result(apply this patch): # filefrag -v /mnt/mp1/file Filesystem type is: ef53 File size of /mnt/mp1/file is 1048576 (256 blocks, blocksize 4096) ext logical physical expected length flags 0 0 66560 128 1 128 67456 66687 128 eof /mnt/mp1/file: 2 extents found Signed-off-by: Kazuya Mio <k-mio@sx.jp.nec.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-01-10 17:12:28 +00:00
* common.
*/
ex = path[depth].p_ext;
ext4: fix 32bit overflow in ext4_ext_find_goal() ext4_ext_find_goal() returns an ideal physical block number that the block allocator tries to allocate first. However, if a required file offset is smaller than the existing extent's one, ext4_ext_find_goal() returns a wrong block number because it may overflow at "block - le32_to_cpu(ex->ee_block)". This patch fixes the problem. ext4_ext_find_goal() will also return a wrong block number in case a file offset of the existing extent is too big. In this case, the ideal physical block number is fixed in ext4_mb_initialize_context(), so it's no problem. reproduce: # dd if=/dev/zero of=/mnt/mp1/tmp bs=127M count=1 oflag=sync # dd if=/dev/zero of=/mnt/mp1/file bs=512K count=1 seek=1 oflag=sync # filefrag -v /mnt/mp1/file Filesystem type is: ef53 File size of /mnt/mp1/file is 1048576 (256 blocks, blocksize 4096) ext logical physical expected length flags 0 128 67456 128 eof /mnt/mp1/file: 2 extents found # rm -rf /mnt/mp1/tmp # echo $((512*4096)) > /sys/fs/ext4/loop0/mb_stream_req # dd if=/dev/zero of=/mnt/mp1/file bs=512K count=1 oflag=sync conv=notrunc result (linux-2.6.37-rc2 + ext4 patch queue): # filefrag -v /mnt/mp1/file Filesystem type is: ef53 File size of /mnt/mp1/file is 1048576 (256 blocks, blocksize 4096) ext logical physical expected length flags 0 0 33280 128 1 128 67456 33407 128 eof /mnt/mp1/file: 2 extents found result(apply this patch): # filefrag -v /mnt/mp1/file Filesystem type is: ef53 File size of /mnt/mp1/file is 1048576 (256 blocks, blocksize 4096) ext logical physical expected length flags 0 0 66560 128 1 128 67456 66687 128 eof /mnt/mp1/file: 2 extents found Signed-off-by: Kazuya Mio <k-mio@sx.jp.nec.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-01-10 17:12:28 +00:00
if (ex) {
ext4_fsblk_t ext_pblk = ext4_ext_pblock(ex);
ext4_lblk_t ext_block = le32_to_cpu(ex->ee_block);
if (block > ext_block)
return ext_pblk + (block - ext_block);
else
return ext_pblk - (ext_block - block);
}
/* it looks like index is empty;
* try to find starting block from index itself */
if (path[depth].p_bh)
return path[depth].p_bh->b_blocknr;
}
/* OK. use inode's group */
return ext4_inode_to_goal_block(inode);
}
/*
* Allocation for a meta data block
*/
static ext4_fsblk_t
ext4_ext_new_meta_block(handle_t *handle, struct inode *inode,
struct ext4_ext_path *path,
struct ext4_extent *ex, int *err, unsigned int flags)
{
ext4_fsblk_t goal, newblock;
goal = ext4_ext_find_goal(inode, path, le32_to_cpu(ex->ee_block));
newblock = ext4_new_meta_blocks(handle, inode, goal, flags,
NULL, err);
return newblock;
}
static inline int ext4_ext_space_block(struct inode *inode, int check)
{
int size;
size = (inode->i_sb->s_blocksize - sizeof(struct ext4_extent_header))
/ sizeof(struct ext4_extent);
if (!check) {
#ifdef AGGRESSIVE_TEST
if (size > 6)
size = 6;
#endif
}
return size;
}
static inline int ext4_ext_space_block_idx(struct inode *inode, int check)
{
int size;
size = (inode->i_sb->s_blocksize - sizeof(struct ext4_extent_header))
/ sizeof(struct ext4_extent_idx);
if (!check) {
#ifdef AGGRESSIVE_TEST
if (size > 5)
size = 5;
#endif
}
return size;
}
static inline int ext4_ext_space_root(struct inode *inode, int check)
{
int size;
size = sizeof(EXT4_I(inode)->i_data);
size -= sizeof(struct ext4_extent_header);
size /= sizeof(struct ext4_extent);
if (!check) {
#ifdef AGGRESSIVE_TEST
if (size > 3)
size = 3;
#endif
}
return size;
}
static inline int ext4_ext_space_root_idx(struct inode *inode, int check)
{
int size;
size = sizeof(EXT4_I(inode)->i_data);
size -= sizeof(struct ext4_extent_header);
size /= sizeof(struct ext4_extent_idx);
if (!check) {
#ifdef AGGRESSIVE_TEST
if (size > 4)
size = 4;
#endif
}
return size;
}
/*
* Calculate the number of metadata blocks needed
* to allocate @blocks
* Worse case is one block per extent
*/
int ext4_ext_calc_metadata_amount(struct inode *inode, ext4_lblk_t lblock)
{
struct ext4_inode_info *ei = EXT4_I(inode);
int idxs, num = 0;
idxs = ((inode->i_sb->s_blocksize - sizeof(struct ext4_extent_header))
/ sizeof(struct ext4_extent_idx));
/*
* If the new delayed allocation block is contiguous with the
* previous da block, it can share index blocks with the
* previous block, so we only need to allocate a new index
* block every idxs leaf blocks. At ldxs**2 blocks, we need
* an additional index block, and at ldxs**3 blocks, yet
* another index blocks.
*/
if (ei->i_da_metadata_calc_len &&
ei->i_da_metadata_calc_last_lblock+1 == lblock) {
if ((ei->i_da_metadata_calc_len % idxs) == 0)
num++;
if ((ei->i_da_metadata_calc_len % (idxs*idxs)) == 0)
num++;
if ((ei->i_da_metadata_calc_len % (idxs*idxs*idxs)) == 0) {
num++;
ei->i_da_metadata_calc_len = 0;
} else
ei->i_da_metadata_calc_len++;
ei->i_da_metadata_calc_last_lblock++;
return num;
}
/*
* In the worst case we need a new set of index blocks at
* every level of the inode's extent tree.
*/
ei->i_da_metadata_calc_len = 1;
ei->i_da_metadata_calc_last_lblock = lblock;
return ext_depth(inode) + 1;
}
static int
ext4_ext_max_entries(struct inode *inode, int depth)
{
int max;
if (depth == ext_depth(inode)) {
if (depth == 0)
max = ext4_ext_space_root(inode, 1);
else
max = ext4_ext_space_root_idx(inode, 1);
} else {
if (depth == 0)
max = ext4_ext_space_block(inode, 1);
else
max = ext4_ext_space_block_idx(inode, 1);
}
return max;
}
static int ext4_valid_extent(struct inode *inode, struct ext4_extent *ext)
{
ext4_fsblk_t block = ext4_ext_pblock(ext);
int len = ext4_ext_get_actual_len(ext);
return ext4_data_block_valid(EXT4_SB(inode->i_sb), block, len);
}
static int ext4_valid_extent_idx(struct inode *inode,
struct ext4_extent_idx *ext_idx)
{
ext4_fsblk_t block = ext4_idx_pblock(ext_idx);
return ext4_data_block_valid(EXT4_SB(inode->i_sb), block, 1);
}
static int ext4_valid_extent_entries(struct inode *inode,
struct ext4_extent_header *eh,
int depth)
{
struct ext4_extent *ext;
struct ext4_extent_idx *ext_idx;
unsigned short entries;
if (eh->eh_entries == 0)
return 1;
entries = le16_to_cpu(eh->eh_entries);
if (depth == 0) {
/* leaf entries */
ext = EXT_FIRST_EXTENT(eh);
while (entries) {
if (!ext4_valid_extent(inode, ext))
return 0;
ext++;
entries--;
}
} else {
ext_idx = EXT_FIRST_INDEX(eh);
while (entries) {
if (!ext4_valid_extent_idx(inode, ext_idx))
return 0;
ext_idx++;
entries--;
}
}
return 1;
}
static int __ext4_ext_check(const char *function, unsigned int line,
struct inode *inode, struct ext4_extent_header *eh,
int depth)
{
const char *error_msg;
int max = 0;
if (unlikely(eh->eh_magic != EXT4_EXT_MAGIC)) {
error_msg = "invalid magic";
goto corrupted;
}
if (unlikely(le16_to_cpu(eh->eh_depth) != depth)) {
error_msg = "unexpected eh_depth";
goto corrupted;
}
if (unlikely(eh->eh_max == 0)) {
error_msg = "invalid eh_max";
goto corrupted;
}
max = ext4_ext_max_entries(inode, depth);
if (unlikely(le16_to_cpu(eh->eh_max) > max)) {
error_msg = "too large eh_max";
goto corrupted;
}
if (unlikely(le16_to_cpu(eh->eh_entries) > le16_to_cpu(eh->eh_max))) {
error_msg = "invalid eh_entries";
goto corrupted;
}
if (!ext4_valid_extent_entries(inode, eh, depth)) {
error_msg = "invalid extent entries";
goto corrupted;
}
return 0;
corrupted:
ext4_error_inode(inode, function, line, 0,
"bad header/extent: %s - magic %x, "
"entries %u, max %u(%u), depth %u(%u)",
error_msg, le16_to_cpu(eh->eh_magic),
le16_to_cpu(eh->eh_entries), le16_to_cpu(eh->eh_max),
max, le16_to_cpu(eh->eh_depth), depth);
return -EIO;
}
#define ext4_ext_check(inode, eh, depth) \
__ext4_ext_check(__func__, __LINE__, inode, eh, depth)
int ext4_ext_check_inode(struct inode *inode)
{
return ext4_ext_check(inode, ext_inode_hdr(inode), ext_depth(inode));
}
#ifdef EXT_DEBUG
static void ext4_ext_show_path(struct inode *inode, struct ext4_ext_path *path)
{
int k, l = path->p_depth;
ext_debug("path:");
for (k = 0; k <= l; k++, path++) {
if (path->p_idx) {
ext_debug(" %d->%llu", le32_to_cpu(path->p_idx->ei_block),
ext4_idx_pblock(path->p_idx));
} else if (path->p_ext) {
ext_debug(" %d:[%d]%d:%llu ",
le32_to_cpu(path->p_ext->ee_block),
ext4_ext_is_uninitialized(path->p_ext),
ext4_ext_get_actual_len(path->p_ext),
ext4_ext_pblock(path->p_ext));
} else
ext_debug(" []");
}
ext_debug("\n");
}
static void ext4_ext_show_leaf(struct inode *inode, struct ext4_ext_path *path)
{
int depth = ext_depth(inode);
struct ext4_extent_header *eh;
struct ext4_extent *ex;
int i;
if (!path)
return;
eh = path[depth].p_hdr;
ex = EXT_FIRST_EXTENT(eh);
ext_debug("Displaying leaf extents for inode %lu\n", inode->i_ino);
for (i = 0; i < le16_to_cpu(eh->eh_entries); i++, ex++) {
ext_debug("%d:[%d]%d:%llu ", le32_to_cpu(ex->ee_block),
ext4_ext_is_uninitialized(ex),
ext4_ext_get_actual_len(ex), ext4_ext_pblock(ex));
}
ext_debug("\n");
}
static void ext4_ext_show_move(struct inode *inode, struct ext4_ext_path *path,
ext4_fsblk_t newblock, int level)
{
int depth = ext_depth(inode);
struct ext4_extent *ex;
if (depth != level) {
struct ext4_extent_idx *idx;
idx = path[level].p_idx;
while (idx <= EXT_MAX_INDEX(path[level].p_hdr)) {
ext_debug("%d: move %d:%llu in new index %llu\n", level,
le32_to_cpu(idx->ei_block),
ext4_idx_pblock(idx),
newblock);
idx++;
}
return;
}
ex = path[depth].p_ext;
while (ex <= EXT_MAX_EXTENT(path[depth].p_hdr)) {
ext_debug("move %d:%llu:[%d]%d in new leaf %llu\n",
le32_to_cpu(ex->ee_block),
ext4_ext_pblock(ex),
ext4_ext_is_uninitialized(ex),
ext4_ext_get_actual_len(ex),
newblock);
ex++;
}
}
#else
#define ext4_ext_show_path(inode, path)
#define ext4_ext_show_leaf(inode, path)
#define ext4_ext_show_move(inode, path, newblock, level)
#endif
void ext4_ext_drop_refs(struct ext4_ext_path *path)
{
int depth = path->p_depth;
int i;
for (i = 0; i <= depth; i++, path++)
if (path->p_bh) {
brelse(path->p_bh);
path->p_bh = NULL;
}
}
/*
* ext4_ext_binsearch_idx:
* binary search for the closest index of the given block
* the header must be checked before calling this
*/
static void
ext4_ext_binsearch_idx(struct inode *inode,
struct ext4_ext_path *path, ext4_lblk_t block)
{
struct ext4_extent_header *eh = path->p_hdr;
struct ext4_extent_idx *r, *l, *m;
ext_debug("binsearch for %u(idx): ", block);
l = EXT_FIRST_INDEX(eh) + 1;
r = EXT_LAST_INDEX(eh);
while (l <= r) {
m = l + (r - l) / 2;
if (block < le32_to_cpu(m->ei_block))
r = m - 1;
else
l = m + 1;
ext_debug("%p(%u):%p(%u):%p(%u) ", l, le32_to_cpu(l->ei_block),
m, le32_to_cpu(m->ei_block),
r, le32_to_cpu(r->ei_block));
}
path->p_idx = l - 1;
ext_debug(" -> %d->%lld ", le32_to_cpu(path->p_idx->ei_block),
ext4_idx_pblock(path->p_idx));
#ifdef CHECK_BINSEARCH
{
struct ext4_extent_idx *chix, *ix;
int k;
chix = ix = EXT_FIRST_INDEX(eh);
for (k = 0; k < le16_to_cpu(eh->eh_entries); k++, ix++) {
if (k != 0 &&
le32_to_cpu(ix->ei_block) <= le32_to_cpu(ix[-1].ei_block)) {
printk(KERN_DEBUG "k=%d, ix=0x%p, "
"first=0x%p\n", k,
ix, EXT_FIRST_INDEX(eh));
printk(KERN_DEBUG "%u <= %u\n",
le32_to_cpu(ix->ei_block),
le32_to_cpu(ix[-1].ei_block));
}
BUG_ON(k && le32_to_cpu(ix->ei_block)
<= le32_to_cpu(ix[-1].ei_block));
if (block < le32_to_cpu(ix->ei_block))
break;
chix = ix;
}
BUG_ON(chix != path->p_idx);
}
#endif
}
/*
* ext4_ext_binsearch:
* binary search for closest extent of the given block
* the header must be checked before calling this
*/
static void
ext4_ext_binsearch(struct inode *inode,
struct ext4_ext_path *path, ext4_lblk_t block)
{
struct ext4_extent_header *eh = path->p_hdr;
struct ext4_extent *r, *l, *m;
if (eh->eh_entries == 0) {
/*
* this leaf is empty:
* we get such a leaf in split/add case
*/
return;
}
ext_debug("binsearch for %u: ", block);
l = EXT_FIRST_EXTENT(eh) + 1;
r = EXT_LAST_EXTENT(eh);
while (l <= r) {
m = l + (r - l) / 2;
if (block < le32_to_cpu(m->ee_block))
r = m - 1;
else
l = m + 1;
ext_debug("%p(%u):%p(%u):%p(%u) ", l, le32_to_cpu(l->ee_block),
m, le32_to_cpu(m->ee_block),
r, le32_to_cpu(r->ee_block));
}
path->p_ext = l - 1;
ext_debug(" -> %d:%llu:[%d]%d ",
le32_to_cpu(path->p_ext->ee_block),
ext4_ext_pblock(path->p_ext),
ext4_ext_is_uninitialized(path->p_ext),
ext4_ext_get_actual_len(path->p_ext));
#ifdef CHECK_BINSEARCH
{
struct ext4_extent *chex, *ex;
int k;
chex = ex = EXT_FIRST_EXTENT(eh);
for (k = 0; k < le16_to_cpu(eh->eh_entries); k++, ex++) {
BUG_ON(k && le32_to_cpu(ex->ee_block)
<= le32_to_cpu(ex[-1].ee_block));
if (block < le32_to_cpu(ex->ee_block))
break;
chex = ex;
}
BUG_ON(chex != path->p_ext);
}
#endif
}
int ext4_ext_tree_init(handle_t *handle, struct inode *inode)
{
struct ext4_extent_header *eh;
eh = ext_inode_hdr(inode);
eh->eh_depth = 0;
eh->eh_entries = 0;
eh->eh_magic = EXT4_EXT_MAGIC;
eh->eh_max = cpu_to_le16(ext4_ext_space_root(inode, 0));
ext4_mark_inode_dirty(handle, inode);
ext4_ext_invalidate_cache(inode);
return 0;
}
struct ext4_ext_path *
ext4_ext_find_extent(struct inode *inode, ext4_lblk_t block,
struct ext4_ext_path *path)
{
struct ext4_extent_header *eh;
struct buffer_head *bh;
short int depth, i, ppos = 0, alloc = 0;
eh = ext_inode_hdr(inode);
depth = ext_depth(inode);
/* account possible depth increase */
if (!path) {
path = kzalloc(sizeof(struct ext4_ext_path) * (depth + 2),
GFP_NOFS);
if (!path)
return ERR_PTR(-ENOMEM);
alloc = 1;
}
path[0].p_hdr = eh;
path[0].p_bh = NULL;
i = depth;
/* walk through the tree */
while (i) {
int need_to_validate = 0;
ext_debug("depth %d: num %d, max %d\n",
ppos, le16_to_cpu(eh->eh_entries), le16_to_cpu(eh->eh_max));
ext4_ext_binsearch_idx(inode, path + ppos, block);
path[ppos].p_block = ext4_idx_pblock(path[ppos].p_idx);
path[ppos].p_depth = i;
path[ppos].p_ext = NULL;
bh = sb_getblk(inode->i_sb, path[ppos].p_block);
if (unlikely(!bh))
goto err;
if (!bh_uptodate_or_lock(bh)) {
trace_ext4_ext_load_extent(inode, block,
path[ppos].p_block);
if (bh_submit_read(bh) < 0) {
put_bh(bh);
goto err;
}
/* validate the extent entries */
need_to_validate = 1;
}
eh = ext_block_hdr(bh);
ppos++;
if (unlikely(ppos > depth)) {
put_bh(bh);
EXT4_ERROR_INODE(inode,
"ppos %d > depth %d", ppos, depth);
goto err;
}
path[ppos].p_bh = bh;
path[ppos].p_hdr = eh;
i--;
if (need_to_validate && ext4_ext_check(inode, eh, i))
goto err;
}
path[ppos].p_depth = i;
path[ppos].p_ext = NULL;
path[ppos].p_idx = NULL;
/* find extent */
ext4_ext_binsearch(inode, path + ppos, block);
/* if not an empty leaf */
if (path[ppos].p_ext)
path[ppos].p_block = ext4_ext_pblock(path[ppos].p_ext);
ext4_ext_show_path(inode, path);
return path;
err:
ext4_ext_drop_refs(path);
if (alloc)
kfree(path);
return ERR_PTR(-EIO);
}
/*
* ext4_ext_insert_index:
* insert new index [@logical;@ptr] into the block at @curp;
* check where to insert: before @curp or after @curp
*/
static int ext4_ext_insert_index(handle_t *handle, struct inode *inode,
struct ext4_ext_path *curp,
int logical, ext4_fsblk_t ptr)
{
struct ext4_extent_idx *ix;
int len, err;
err = ext4_ext_get_access(handle, inode, curp);
if (err)
return err;
if (unlikely(logical == le32_to_cpu(curp->p_idx->ei_block))) {
EXT4_ERROR_INODE(inode,
"logical %d == ei_block %d!",
logical, le32_to_cpu(curp->p_idx->ei_block));
return -EIO;
}
if (unlikely(le16_to_cpu(curp->p_hdr->eh_entries)
>= le16_to_cpu(curp->p_hdr->eh_max))) {
EXT4_ERROR_INODE(inode,
"eh_entries %d >= eh_max %d!",
le16_to_cpu(curp->p_hdr->eh_entries),
le16_to_cpu(curp->p_hdr->eh_max));
return -EIO;
}
len = EXT_MAX_INDEX(curp->p_hdr) - curp->p_idx;
if (logical > le32_to_cpu(curp->p_idx->ei_block)) {
/* insert after */
if (curp->p_idx != EXT_LAST_INDEX(curp->p_hdr)) {
len = (len - 1) * sizeof(struct ext4_extent_idx);
len = len < 0 ? 0 : len;
ext_debug("insert new index %d after: %llu. "
"move %d from 0x%p to 0x%p\n",
logical, ptr, len,
(curp->p_idx + 1), (curp->p_idx + 2));
memmove(curp->p_idx + 2, curp->p_idx + 1, len);
}
ix = curp->p_idx + 1;
} else {
/* insert before */
len = len * sizeof(struct ext4_extent_idx);
len = len < 0 ? 0 : len;
ext_debug("insert new index %d before: %llu. "
"move %d from 0x%p to 0x%p\n",
logical, ptr, len,
curp->p_idx, (curp->p_idx + 1));
memmove(curp->p_idx + 1, curp->p_idx, len);
ix = curp->p_idx;
}
ix->ei_block = cpu_to_le32(logical);
ext4_idx_store_pblock(ix, ptr);
le16_add_cpu(&curp->p_hdr->eh_entries, 1);
if (unlikely(ix > EXT_LAST_INDEX(curp->p_hdr))) {
EXT4_ERROR_INODE(inode, "ix > EXT_LAST_INDEX!");
return -EIO;
}
err = ext4_ext_dirty(handle, inode, curp);
ext4_std_error(inode->i_sb, err);
return err;
}
/*
* ext4_ext_split:
* inserts new subtree into the path, using free index entry
* at depth @at:
* - allocates all needed blocks (new leaf and all intermediate index blocks)
* - makes decision where to split
* - moves remaining extents and index entries (right to the split point)
* into the newly allocated blocks
* - initializes subtree
*/
static int ext4_ext_split(handle_t *handle, struct inode *inode,
unsigned int flags,
struct ext4_ext_path *path,
struct ext4_extent *newext, int at)
{
struct buffer_head *bh = NULL;
int depth = ext_depth(inode);
struct ext4_extent_header *neh;
struct ext4_extent_idx *fidx;
int i = at, k, m, a;
ext4_fsblk_t newblock, oldblock;
__le32 border;
ext4_fsblk_t *ablocks = NULL; /* array of allocated blocks */
int err = 0;
/* make decision: where to split? */
/* FIXME: now decision is simplest: at current extent */
/* if current leaf will be split, then we should use
* border from split point */
if (unlikely(path[depth].p_ext > EXT_MAX_EXTENT(path[depth].p_hdr))) {
EXT4_ERROR_INODE(inode, "p_ext > EXT_MAX_EXTENT!");
return -EIO;
}
if (path[depth].p_ext != EXT_MAX_EXTENT(path[depth].p_hdr)) {
border = path[depth].p_ext[1].ee_block;
ext_debug("leaf will be split."
" next leaf starts at %d\n",
le32_to_cpu(border));
} else {
border = newext->ee_block;
ext_debug("leaf will be added."
" next leaf starts at %d\n",
le32_to_cpu(border));
}
/*
* If error occurs, then we break processing
* and mark filesystem read-only. index won't
* be inserted and tree will be in consistent
* state. Next mount will repair buffers too.
*/
/*
* Get array to track all allocated blocks.
* We need this to handle errors and free blocks
* upon them.
*/
ablocks = kzalloc(sizeof(ext4_fsblk_t) * depth, GFP_NOFS);
if (!ablocks)
return -ENOMEM;
/* allocate all needed blocks */
ext_debug("allocate %d blocks for indexes/leaf\n", depth - at);
for (a = 0; a < depth - at; a++) {
newblock = ext4_ext_new_meta_block(handle, inode, path,
newext, &err, flags);
if (newblock == 0)
goto cleanup;
ablocks[a] = newblock;
}
/* initialize new leaf */
newblock = ablocks[--a];
if (unlikely(newblock == 0)) {
EXT4_ERROR_INODE(inode, "newblock == 0!");
err = -EIO;
goto cleanup;
}
bh = sb_getblk(inode->i_sb, newblock);
if (!bh) {
err = -EIO;
goto cleanup;
}
lock_buffer(bh);
err = ext4_journal_get_create_access(handle, bh);
if (err)
goto cleanup;
neh = ext_block_hdr(bh);
neh->eh_entries = 0;
neh->eh_max = cpu_to_le16(ext4_ext_space_block(inode, 0));
neh->eh_magic = EXT4_EXT_MAGIC;
neh->eh_depth = 0;
/* move remainder of path[depth] to the new leaf */
if (unlikely(path[depth].p_hdr->eh_entries !=
path[depth].p_hdr->eh_max)) {
EXT4_ERROR_INODE(inode, "eh_entries %d != eh_max %d!",
path[depth].p_hdr->eh_entries,
path[depth].p_hdr->eh_max);
err = -EIO;
goto cleanup;
}
/* start copy from next extent */
m = EXT_MAX_EXTENT(path[depth].p_hdr) - path[depth].p_ext++;
ext4_ext_show_move(inode, path, newblock, depth);
if (m) {
struct ext4_extent *ex;
ex = EXT_FIRST_EXTENT(neh);
memmove(ex, path[depth].p_ext, sizeof(struct ext4_extent) * m);
le16_add_cpu(&neh->eh_entries, m);
}
set_buffer_uptodate(bh);
unlock_buffer(bh);
err = ext4_handle_dirty_metadata(handle, inode, bh);
if (err)
goto cleanup;
brelse(bh);
bh = NULL;
/* correct old leaf */
if (m) {
err = ext4_ext_get_access(handle, inode, path + depth);
if (err)
goto cleanup;
le16_add_cpu(&path[depth].p_hdr->eh_entries, -m);
err = ext4_ext_dirty(handle, inode, path + depth);
if (err)
goto cleanup;
}
/* create intermediate indexes */
k = depth - at - 1;
if (unlikely(k < 0)) {
EXT4_ERROR_INODE(inode, "k %d < 0!", k);
err = -EIO;
goto cleanup;
}
if (k)
ext_debug("create %d intermediate indices\n", k);
/* insert new index into current index block */
/* current depth stored in i var */
i = depth - 1;
while (k--) {
oldblock = newblock;
newblock = ablocks[--a];
bh = sb_getblk(inode->i_sb, newblock);
if (!bh) {
err = -EIO;
goto cleanup;
}
lock_buffer(bh);
err = ext4_journal_get_create_access(handle, bh);
if (err)
goto cleanup;
neh = ext_block_hdr(bh);
neh->eh_entries = cpu_to_le16(1);
neh->eh_magic = EXT4_EXT_MAGIC;
neh->eh_max = cpu_to_le16(ext4_ext_space_block_idx(inode, 0));
neh->eh_depth = cpu_to_le16(depth - i);
fidx = EXT_FIRST_INDEX(neh);
fidx->ei_block = border;
ext4_idx_store_pblock(fidx, oldblock);
ext_debug("int.index at %d (block %llu): %u -> %llu\n",
i, newblock, le32_to_cpu(border), oldblock);
/* move remainder of path[i] to the new index block */
if (unlikely(EXT_MAX_INDEX(path[i].p_hdr) !=
EXT_LAST_INDEX(path[i].p_hdr))) {
EXT4_ERROR_INODE(inode,
"EXT_MAX_INDEX != EXT_LAST_INDEX ee_block %d!",
le32_to_cpu(path[i].p_ext->ee_block));
err = -EIO;
goto cleanup;
}
/* start copy indexes */
m = EXT_MAX_INDEX(path[i].p_hdr) - path[i].p_idx++;
ext_debug("cur 0x%p, last 0x%p\n", path[i].p_idx,
EXT_MAX_INDEX(path[i].p_hdr));
ext4_ext_show_move(inode, path, newblock, i);
if (m) {
memmove(++fidx, path[i].p_idx,
sizeof(struct ext4_extent_idx) * m);
le16_add_cpu(&neh->eh_entries, m);
}
set_buffer_uptodate(bh);
unlock_buffer(bh);
err = ext4_handle_dirty_metadata(handle, inode, bh);
if (err)
goto cleanup;
brelse(bh);
bh = NULL;
/* correct old index */
if (m) {
err = ext4_ext_get_access(handle, inode, path + i);
if (err)
goto cleanup;
le16_add_cpu(&path[i].p_hdr->eh_entries, -m);
err = ext4_ext_dirty(handle, inode, path + i);
if (err)
goto cleanup;
}
i--;
}
/* insert new index */
err = ext4_ext_insert_index(handle, inode, path + at,
le32_to_cpu(border), newblock);
cleanup:
if (bh) {
if (buffer_locked(bh))
unlock_buffer(bh);
brelse(bh);
}
if (err) {
/* free all allocated blocks in error case */
for (i = 0; i < depth; i++) {
if (!ablocks[i])
continue;
ext4_free_blocks(handle, inode, NULL, ablocks[i], 1,
EXT4_FREE_BLOCKS_METADATA);
}
}
kfree(ablocks);
return err;
}
/*
* ext4_ext_grow_indepth:
* implements tree growing procedure:
* - allocates new block
* - moves top-level data (index block or leaf) into the new block
* - initializes new top-level, creating index that points to the
* just created block
*/
static int ext4_ext_grow_indepth(handle_t *handle, struct inode *inode,
unsigned int flags,
struct ext4_ext_path *path,
struct ext4_extent *newext)
{
struct ext4_ext_path *curp = path;
struct ext4_extent_header *neh;
struct buffer_head *bh;
ext4_fsblk_t newblock;
int err = 0;
newblock = ext4_ext_new_meta_block(handle, inode, path,
newext, &err, flags);
if (newblock == 0)
return err;
bh = sb_getblk(inode->i_sb, newblock);
if (!bh) {
err = -EIO;
ext4_std_error(inode->i_sb, err);
return err;
}
lock_buffer(bh);
err = ext4_journal_get_create_access(handle, bh);
if (err) {
unlock_buffer(bh);
goto out;
}
/* move top-level index/leaf into new block */
memmove(bh->b_data, curp->p_hdr, sizeof(EXT4_I(inode)->i_data));
/* set size of new block */
neh = ext_block_hdr(bh);
/* old root could have indexes or leaves
* so calculate e_max right way */
if (ext_depth(inode))
neh->eh_max = cpu_to_le16(ext4_ext_space_block_idx(inode, 0));
else
neh->eh_max = cpu_to_le16(ext4_ext_space_block(inode, 0));
neh->eh_magic = EXT4_EXT_MAGIC;
set_buffer_uptodate(bh);
unlock_buffer(bh);
err = ext4_handle_dirty_metadata(handle, inode, bh);
if (err)
goto out;
/* create index in new top-level index: num,max,pointer */
err = ext4_ext_get_access(handle, inode, curp);
if (err)
goto out;
curp->p_hdr->eh_magic = EXT4_EXT_MAGIC;
curp->p_hdr->eh_max = cpu_to_le16(ext4_ext_space_root_idx(inode, 0));
curp->p_hdr->eh_entries = cpu_to_le16(1);
curp->p_idx = EXT_FIRST_INDEX(curp->p_hdr);
if (path[0].p_hdr->eh_depth)
curp->p_idx->ei_block =
EXT_FIRST_INDEX(path[0].p_hdr)->ei_block;
else
curp->p_idx->ei_block =
EXT_FIRST_EXTENT(path[0].p_hdr)->ee_block;
ext4_idx_store_pblock(curp->p_idx, newblock);
neh = ext_inode_hdr(inode);
ext_debug("new root: num %d(%d), lblock %d, ptr %llu\n",
le16_to_cpu(neh->eh_entries), le16_to_cpu(neh->eh_max),
le32_to_cpu(EXT_FIRST_INDEX(neh)->ei_block),
ext4_idx_pblock(EXT_FIRST_INDEX(neh)));
neh->eh_depth = cpu_to_le16(path->p_depth + 1);
err = ext4_ext_dirty(handle, inode, curp);
out:
brelse(bh);
return err;
}
/*
* ext4_ext_create_new_leaf:
* finds empty index and adds new leaf.
* if no free index is found, then it requests in-depth growing.
*/
static int ext4_ext_create_new_leaf(handle_t *handle, struct inode *inode,
unsigned int flags,
struct ext4_ext_path *path,
struct ext4_extent *newext)
{
struct ext4_ext_path *curp;
int depth, i, err = 0;
repeat:
i = depth = ext_depth(inode);
/* walk up to the tree and look for free index entry */
curp = path + depth;
while (i > 0 && !EXT_HAS_FREE_INDEX(curp)) {
i--;
curp--;
}
/* we use already allocated block for index block,
* so subsequent data blocks should be contiguous */
if (EXT_HAS_FREE_INDEX(curp)) {
/* if we found index with free entry, then use that
* entry: create all needed subtree and add new leaf */
err = ext4_ext_split(handle, inode, flags, path, newext, i);
if (err)
goto out;
/* refill path */
ext4_ext_drop_refs(path);
path = ext4_ext_find_extent(inode,
(ext4_lblk_t)le32_to_cpu(newext->ee_block),
path);
if (IS_ERR(path))
err = PTR_ERR(path);
} else {
/* tree is full, time to grow in depth */
err = ext4_ext_grow_indepth(handle, inode, flags,
path, newext);
if (err)
goto out;
/* refill path */
ext4_ext_drop_refs(path);
path = ext4_ext_find_extent(inode,
(ext4_lblk_t)le32_to_cpu(newext->ee_block),
path);
if (IS_ERR(path)) {
err = PTR_ERR(path);
goto out;
}
/*
* only first (depth 0 -> 1) produces free space;
* in all other cases we have to split the grown tree
*/
depth = ext_depth(inode);
if (path[depth].p_hdr->eh_entries == path[depth].p_hdr->eh_max) {
/* now we need to split */
goto repeat;
}
}
out:
return err;
}
/*
* search the closest allocated block to the left for *logical
* and returns it at @logical + it's physical address at @phys
* if *logical is the smallest allocated block, the function
* returns 0 at @phys
* return value contains 0 (success) or error code
*/
static int ext4_ext_search_left(struct inode *inode,
struct ext4_ext_path *path,
ext4_lblk_t *logical, ext4_fsblk_t *phys)
{
struct ext4_extent_idx *ix;
struct ext4_extent *ex;
int depth, ee_len;
if (unlikely(path == NULL)) {
EXT4_ERROR_INODE(inode, "path == NULL *logical %d!", *logical);
return -EIO;
}
depth = path->p_depth;
*phys = 0;
if (depth == 0 && path->p_ext == NULL)
return 0;
/* usually extent in the path covers blocks smaller
* then *logical, but it can be that extent is the
* first one in the file */
ex = path[depth].p_ext;
ee_len = ext4_ext_get_actual_len(ex);
if (*logical < le32_to_cpu(ex->ee_block)) {
if (unlikely(EXT_FIRST_EXTENT(path[depth].p_hdr) != ex)) {
EXT4_ERROR_INODE(inode,
"EXT_FIRST_EXTENT != ex *logical %d ee_block %d!",
*logical, le32_to_cpu(ex->ee_block));
return -EIO;
}
while (--depth >= 0) {
ix = path[depth].p_idx;
if (unlikely(ix != EXT_FIRST_INDEX(path[depth].p_hdr))) {
EXT4_ERROR_INODE(inode,
"ix (%d) != EXT_FIRST_INDEX (%d) (depth %d)!",
ix != NULL ? ix->ei_block : 0,
EXT_FIRST_INDEX(path[depth].p_hdr) != NULL ?
EXT_FIRST_INDEX(path[depth].p_hdr)->ei_block : 0,
depth);
return -EIO;
}
}
return 0;
}
if (unlikely(*logical < (le32_to_cpu(ex->ee_block) + ee_len))) {
EXT4_ERROR_INODE(inode,
"logical %d < ee_block %d + ee_len %d!",
*logical, le32_to_cpu(ex->ee_block), ee_len);
return -EIO;
}
*logical = le32_to_cpu(ex->ee_block) + ee_len - 1;
*phys = ext4_ext_pblock(ex) + ee_len - 1;
return 0;
}
/*
* search the closest allocated block to the right for *logical
* and returns it at @logical + it's physical address at @phys
* if *logical is the smallest allocated block, the function
* returns 0 at @phys
* return value contains 0 (success) or error code
*/
static int ext4_ext_search_right(struct inode *inode,
struct ext4_ext_path *path,
ext4_lblk_t *logical, ext4_fsblk_t *phys)
{
struct buffer_head *bh = NULL;
struct ext4_extent_header *eh;
struct ext4_extent_idx *ix;
struct ext4_extent *ex;
ext4_fsblk_t block;
int depth; /* Note, NOT eh_depth; depth from top of tree */
int ee_len;
if (unlikely(path == NULL)) {
EXT4_ERROR_INODE(inode, "path == NULL *logical %d!", *logical);
return -EIO;
}
depth = path->p_depth;
*phys = 0;
if (depth == 0 && path->p_ext == NULL)
return 0;
/* usually extent in the path covers blocks smaller
* then *logical, but it can be that extent is the
* first one in the file */
ex = path[depth].p_ext;
ee_len = ext4_ext_get_actual_len(ex);
if (*logical < le32_to_cpu(ex->ee_block)) {
if (unlikely(EXT_FIRST_EXTENT(path[depth].p_hdr) != ex)) {
EXT4_ERROR_INODE(inode,
"first_extent(path[%d].p_hdr) != ex",
depth);
return -EIO;
}
while (--depth >= 0) {
ix = path[depth].p_idx;
if (unlikely(ix != EXT_FIRST_INDEX(path[depth].p_hdr))) {
EXT4_ERROR_INODE(inode,
"ix != EXT_FIRST_INDEX *logical %d!",
*logical);
return -EIO;
}
}
*logical = le32_to_cpu(ex->ee_block);
*phys = ext4_ext_pblock(ex);
return 0;
}
if (unlikely(*logical < (le32_to_cpu(ex->ee_block) + ee_len))) {
EXT4_ERROR_INODE(inode,
"logical %d < ee_block %d + ee_len %d!",
*logical, le32_to_cpu(ex->ee_block), ee_len);
return -EIO;
}
if (ex != EXT_LAST_EXTENT(path[depth].p_hdr)) {
/* next allocated block in this leaf */
ex++;
*logical = le32_to_cpu(ex->ee_block);
*phys = ext4_ext_pblock(ex);
return 0;
}
/* go up and search for index to the right */
while (--depth >= 0) {
ix = path[depth].p_idx;
if (ix != EXT_LAST_INDEX(path[depth].p_hdr))
goto got_index;
}
/* we've gone up to the root and found no index to the right */
return 0;
got_index:
/* we've found index to the right, let's
* follow it and find the closest allocated
* block to the right */
ix++;
block = ext4_idx_pblock(ix);
while (++depth < path->p_depth) {
bh = sb_bread(inode->i_sb, block);
if (bh == NULL)
return -EIO;
eh = ext_block_hdr(bh);
/* subtract from p_depth to get proper eh_depth */
if (ext4_ext_check(inode, eh, path->p_depth - depth)) {
put_bh(bh);
return -EIO;
}
ix = EXT_FIRST_INDEX(eh);
block = ext4_idx_pblock(ix);
put_bh(bh);
}
bh = sb_bread(inode->i_sb, block);
if (bh == NULL)
return -EIO;
eh = ext_block_hdr(bh);
if (ext4_ext_check(inode, eh, path->p_depth - depth)) {
put_bh(bh);
return -EIO;
}
ex = EXT_FIRST_EXTENT(eh);
*logical = le32_to_cpu(ex->ee_block);
*phys = ext4_ext_pblock(ex);
put_bh(bh);
return 0;
}
/*
* ext4_ext_next_allocated_block:
ext4: Fix max file size and logical block counting of extent format file Kazuya Mio reported that he was able to hit BUG_ON(next == lblock) in ext4_ext_put_gap_in_cache() while creating a sparse file in extent format and fill the tail of file up to its end. We will hit the BUG_ON when we write the last block (2^32-1) into the sparse file. The root cause of the problem lies in the fact that we specifically set s_maxbytes so that block at s_maxbytes fit into on-disk extent format, which is 32 bit long. However, we are not storing start and end block number, but rather start block number and length in blocks. It means that in order to cover extent from 0 to EXT_MAX_BLOCK we need EXT_MAX_BLOCK+1 to fit into len (because we counting block 0 as well) - and it does not. The only way to fix it without changing the meaning of the struct ext4_extent members is, as Kazuya Mio suggested, to lower s_maxbytes by one fs block so we can cover the whole extent we can get by the on-disk extent format. Also in many places EXT_MAX_BLOCK is used as length instead of maximum logical block number as the name suggests, it is all a bit messy. So this commit renames it to EXT_MAX_BLOCKS and change its usage in some places to actually be maximum number of blocks in the extent. The bug which this commit fixes can be reproduced as follows: dd if=/dev/zero of=/mnt/mp1/file bs=<blocksize> count=1 seek=$((2**32-2)) sync dd if=/dev/zero of=/mnt/mp1/file bs=<blocksize> count=1 seek=$((2**32-1)) Reported-by: Kazuya Mio <k-mio@sx.jp.nec.com> Signed-off-by: Lukas Czerner <lczerner@redhat.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-06-06 04:05:17 +00:00
* returns allocated block in subsequent extent or EXT_MAX_BLOCKS.
* NOTE: it considers block number from index entry as
* allocated block. Thus, index entries have to be consistent
* with leaves.
*/
static ext4_lblk_t
ext4_ext_next_allocated_block(struct ext4_ext_path *path)
{
int depth;
BUG_ON(path == NULL);
depth = path->p_depth;
if (depth == 0 && path->p_ext == NULL)
ext4: Fix max file size and logical block counting of extent format file Kazuya Mio reported that he was able to hit BUG_ON(next == lblock) in ext4_ext_put_gap_in_cache() while creating a sparse file in extent format and fill the tail of file up to its end. We will hit the BUG_ON when we write the last block (2^32-1) into the sparse file. The root cause of the problem lies in the fact that we specifically set s_maxbytes so that block at s_maxbytes fit into on-disk extent format, which is 32 bit long. However, we are not storing start and end block number, but rather start block number and length in blocks. It means that in order to cover extent from 0 to EXT_MAX_BLOCK we need EXT_MAX_BLOCK+1 to fit into len (because we counting block 0 as well) - and it does not. The only way to fix it without changing the meaning of the struct ext4_extent members is, as Kazuya Mio suggested, to lower s_maxbytes by one fs block so we can cover the whole extent we can get by the on-disk extent format. Also in many places EXT_MAX_BLOCK is used as length instead of maximum logical block number as the name suggests, it is all a bit messy. So this commit renames it to EXT_MAX_BLOCKS and change its usage in some places to actually be maximum number of blocks in the extent. The bug which this commit fixes can be reproduced as follows: dd if=/dev/zero of=/mnt/mp1/file bs=<blocksize> count=1 seek=$((2**32-2)) sync dd if=/dev/zero of=/mnt/mp1/file bs=<blocksize> count=1 seek=$((2**32-1)) Reported-by: Kazuya Mio <k-mio@sx.jp.nec.com> Signed-off-by: Lukas Czerner <lczerner@redhat.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-06-06 04:05:17 +00:00
return EXT_MAX_BLOCKS;
while (depth >= 0) {
if (depth == path->p_depth) {
/* leaf */
if (path[depth].p_ext !=
EXT_LAST_EXTENT(path[depth].p_hdr))
return le32_to_cpu(path[depth].p_ext[1].ee_block);
} else {
/* index */
if (path[depth].p_idx !=
EXT_LAST_INDEX(path[depth].p_hdr))
return le32_to_cpu(path[depth].p_idx[1].ei_block);
}
depth--;
}
ext4: Fix max file size and logical block counting of extent format file Kazuya Mio reported that he was able to hit BUG_ON(next == lblock) in ext4_ext_put_gap_in_cache() while creating a sparse file in extent format and fill the tail of file up to its end. We will hit the BUG_ON when we write the last block (2^32-1) into the sparse file. The root cause of the problem lies in the fact that we specifically set s_maxbytes so that block at s_maxbytes fit into on-disk extent format, which is 32 bit long. However, we are not storing start and end block number, but rather start block number and length in blocks. It means that in order to cover extent from 0 to EXT_MAX_BLOCK we need EXT_MAX_BLOCK+1 to fit into len (because we counting block 0 as well) - and it does not. The only way to fix it without changing the meaning of the struct ext4_extent members is, as Kazuya Mio suggested, to lower s_maxbytes by one fs block so we can cover the whole extent we can get by the on-disk extent format. Also in many places EXT_MAX_BLOCK is used as length instead of maximum logical block number as the name suggests, it is all a bit messy. So this commit renames it to EXT_MAX_BLOCKS and change its usage in some places to actually be maximum number of blocks in the extent. The bug which this commit fixes can be reproduced as follows: dd if=/dev/zero of=/mnt/mp1/file bs=<blocksize> count=1 seek=$((2**32-2)) sync dd if=/dev/zero of=/mnt/mp1/file bs=<blocksize> count=1 seek=$((2**32-1)) Reported-by: Kazuya Mio <k-mio@sx.jp.nec.com> Signed-off-by: Lukas Czerner <lczerner@redhat.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-06-06 04:05:17 +00:00
return EXT_MAX_BLOCKS;
}
/*
* ext4_ext_next_leaf_block:
ext4: Fix max file size and logical block counting of extent format file Kazuya Mio reported that he was able to hit BUG_ON(next == lblock) in ext4_ext_put_gap_in_cache() while creating a sparse file in extent format and fill the tail of file up to its end. We will hit the BUG_ON when we write the last block (2^32-1) into the sparse file. The root cause of the problem lies in the fact that we specifically set s_maxbytes so that block at s_maxbytes fit into on-disk extent format, which is 32 bit long. However, we are not storing start and end block number, but rather start block number and length in blocks. It means that in order to cover extent from 0 to EXT_MAX_BLOCK we need EXT_MAX_BLOCK+1 to fit into len (because we counting block 0 as well) - and it does not. The only way to fix it without changing the meaning of the struct ext4_extent members is, as Kazuya Mio suggested, to lower s_maxbytes by one fs block so we can cover the whole extent we can get by the on-disk extent format. Also in many places EXT_MAX_BLOCK is used as length instead of maximum logical block number as the name suggests, it is all a bit messy. So this commit renames it to EXT_MAX_BLOCKS and change its usage in some places to actually be maximum number of blocks in the extent. The bug which this commit fixes can be reproduced as follows: dd if=/dev/zero of=/mnt/mp1/file bs=<blocksize> count=1 seek=$((2**32-2)) sync dd if=/dev/zero of=/mnt/mp1/file bs=<blocksize> count=1 seek=$((2**32-1)) Reported-by: Kazuya Mio <k-mio@sx.jp.nec.com> Signed-off-by: Lukas Czerner <lczerner@redhat.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-06-06 04:05:17 +00:00
* returns first allocated block from next leaf or EXT_MAX_BLOCKS
*/
static ext4_lblk_t ext4_ext_next_leaf_block(struct ext4_ext_path *path)
{
int depth;
BUG_ON(path == NULL);
depth = path->p_depth;
/* zero-tree has no leaf blocks at all */
if (depth == 0)
ext4: Fix max file size and logical block counting of extent format file Kazuya Mio reported that he was able to hit BUG_ON(next == lblock) in ext4_ext_put_gap_in_cache() while creating a sparse file in extent format and fill the tail of file up to its end. We will hit the BUG_ON when we write the last block (2^32-1) into the sparse file. The root cause of the problem lies in the fact that we specifically set s_maxbytes so that block at s_maxbytes fit into on-disk extent format, which is 32 bit long. However, we are not storing start and end block number, but rather start block number and length in blocks. It means that in order to cover extent from 0 to EXT_MAX_BLOCK we need EXT_MAX_BLOCK+1 to fit into len (because we counting block 0 as well) - and it does not. The only way to fix it without changing the meaning of the struct ext4_extent members is, as Kazuya Mio suggested, to lower s_maxbytes by one fs block so we can cover the whole extent we can get by the on-disk extent format. Also in many places EXT_MAX_BLOCK is used as length instead of maximum logical block number as the name suggests, it is all a bit messy. So this commit renames it to EXT_MAX_BLOCKS and change its usage in some places to actually be maximum number of blocks in the extent. The bug which this commit fixes can be reproduced as follows: dd if=/dev/zero of=/mnt/mp1/file bs=<blocksize> count=1 seek=$((2**32-2)) sync dd if=/dev/zero of=/mnt/mp1/file bs=<blocksize> count=1 seek=$((2**32-1)) Reported-by: Kazuya Mio <k-mio@sx.jp.nec.com> Signed-off-by: Lukas Czerner <lczerner@redhat.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-06-06 04:05:17 +00:00
return EXT_MAX_BLOCKS;
/* go to index block */
depth--;
while (depth >= 0) {
if (path[depth].p_idx !=
EXT_LAST_INDEX(path[depth].p_hdr))
return (ext4_lblk_t)
le32_to_cpu(path[depth].p_idx[1].ei_block);
depth--;
}
ext4: Fix max file size and logical block counting of extent format file Kazuya Mio reported that he was able to hit BUG_ON(next == lblock) in ext4_ext_put_gap_in_cache() while creating a sparse file in extent format and fill the tail of file up to its end. We will hit the BUG_ON when we write the last block (2^32-1) into the sparse file. The root cause of the problem lies in the fact that we specifically set s_maxbytes so that block at s_maxbytes fit into on-disk extent format, which is 32 bit long. However, we are not storing start and end block number, but rather start block number and length in blocks. It means that in order to cover extent from 0 to EXT_MAX_BLOCK we need EXT_MAX_BLOCK+1 to fit into len (because we counting block 0 as well) - and it does not. The only way to fix it without changing the meaning of the struct ext4_extent members is, as Kazuya Mio suggested, to lower s_maxbytes by one fs block so we can cover the whole extent we can get by the on-disk extent format. Also in many places EXT_MAX_BLOCK is used as length instead of maximum logical block number as the name suggests, it is all a bit messy. So this commit renames it to EXT_MAX_BLOCKS and change its usage in some places to actually be maximum number of blocks in the extent. The bug which this commit fixes can be reproduced as follows: dd if=/dev/zero of=/mnt/mp1/file bs=<blocksize> count=1 seek=$((2**32-2)) sync dd if=/dev/zero of=/mnt/mp1/file bs=<blocksize> count=1 seek=$((2**32-1)) Reported-by: Kazuya Mio <k-mio@sx.jp.nec.com> Signed-off-by: Lukas Czerner <lczerner@redhat.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-06-06 04:05:17 +00:00
return EXT_MAX_BLOCKS;
}
/*
* ext4_ext_correct_indexes:
* if leaf gets modified and modified extent is first in the leaf,
* then we have to correct all indexes above.
* TODO: do we need to correct tree in all cases?
*/
static int ext4_ext_correct_indexes(handle_t *handle, struct inode *inode,
struct ext4_ext_path *path)
{
struct ext4_extent_header *eh;
int depth = ext_depth(inode);
struct ext4_extent *ex;
__le32 border;
int k, err = 0;
eh = path[depth].p_hdr;
ex = path[depth].p_ext;
if (unlikely(ex == NULL || eh == NULL)) {
EXT4_ERROR_INODE(inode,
"ex %p == NULL or eh %p == NULL", ex, eh);
return -EIO;
}
if (depth == 0) {
/* there is no tree at all */
return 0;
}
if (ex != EXT_FIRST_EXTENT(eh)) {
/* we correct tree if first leaf got modified only */
return 0;
}
/*
* TODO: we need correction if border is smaller than current one
*/
k = depth - 1;
border = path[depth].p_ext->ee_block;
err = ext4_ext_get_access(handle, inode, path + k);
if (err)
return err;
path[k].p_idx->ei_block = border;
err = ext4_ext_dirty(handle, inode, path + k);
if (err)
return err;
while (k--) {
/* change all left-side indexes */
if (path[k+1].p_idx != EXT_FIRST_INDEX(path[k+1].p_hdr))
break;
err = ext4_ext_get_access(handle, inode, path + k);
if (err)
break;
path[k].p_idx->ei_block = border;
err = ext4_ext_dirty(handle, inode, path + k);
if (err)
break;
}
return err;
}
int
ext4_can_extents_be_merged(struct inode *inode, struct ext4_extent *ex1,
struct ext4_extent *ex2)
{
unsigned short ext1_ee_len, ext2_ee_len, max_len;
/*
* Make sure that either both extents are uninitialized, or
* both are _not_.
*/
if (ext4_ext_is_uninitialized(ex1) ^ ext4_ext_is_uninitialized(ex2))
return 0;
if (ext4_ext_is_uninitialized(ex1))
max_len = EXT_UNINIT_MAX_LEN;
else
max_len = EXT_INIT_MAX_LEN;
ext1_ee_len = ext4_ext_get_actual_len(ex1);
ext2_ee_len = ext4_ext_get_actual_len(ex2);
if (le32_to_cpu(ex1->ee_block) + ext1_ee_len !=
le32_to_cpu(ex2->ee_block))
return 0;
/*
* To allow future support for preallocated extents to be added
* as an RO_COMPAT feature, refuse to merge to extents if
* this can result in the top bit of ee_len being set.
*/
if (ext1_ee_len + ext2_ee_len > max_len)
return 0;
#ifdef AGGRESSIVE_TEST
if (ext1_ee_len >= 4)
return 0;
#endif
if (ext4_ext_pblock(ex1) + ext1_ee_len == ext4_ext_pblock(ex2))
return 1;
return 0;
}
/*
* This function tries to merge the "ex" extent to the next extent in the tree.
* It always tries to merge towards right. If you want to merge towards
* left, pass "ex - 1" as argument instead of "ex".
* Returns 0 if the extents (ex and ex+1) were _not_ merged and returns
* 1 if they got merged.
*/
static int ext4_ext_try_to_merge_right(struct inode *inode,
struct ext4_ext_path *path,
struct ext4_extent *ex)
{
struct ext4_extent_header *eh;
unsigned int depth, len;
int merge_done = 0;
int uninitialized = 0;
depth = ext_depth(inode);
BUG_ON(path[depth].p_hdr == NULL);
eh = path[depth].p_hdr;
while (ex < EXT_LAST_EXTENT(eh)) {
if (!ext4_can_extents_be_merged(inode, ex, ex + 1))
break;
/* merge with next extent! */
if (ext4_ext_is_uninitialized(ex))
uninitialized = 1;
ex->ee_len = cpu_to_le16(ext4_ext_get_actual_len(ex)
+ ext4_ext_get_actual_len(ex + 1));
if (uninitialized)
ext4_ext_mark_uninitialized(ex);
if (ex + 1 < EXT_LAST_EXTENT(eh)) {
len = (EXT_LAST_EXTENT(eh) - ex - 1)
* sizeof(struct ext4_extent);
memmove(ex + 1, ex + 2, len);
}
le16_add_cpu(&eh->eh_entries, -1);
merge_done = 1;
WARN_ON(eh->eh_entries == 0);
if (!eh->eh_entries)
EXT4_ERROR_INODE(inode, "eh->eh_entries = 0!");
}
return merge_done;
}
/*
* This function tries to merge the @ex extent to neighbours in the tree.
* return 1 if merge left else 0.
*/
static int ext4_ext_try_to_merge(struct inode *inode,
struct ext4_ext_path *path,
struct ext4_extent *ex) {
struct ext4_extent_header *eh;
unsigned int depth;
int merge_done = 0;
int ret = 0;
depth = ext_depth(inode);
BUG_ON(path[depth].p_hdr == NULL);
eh = path[depth].p_hdr;
if (ex > EXT_FIRST_EXTENT(eh))
merge_done = ext4_ext_try_to_merge_right(inode, path, ex - 1);
if (!merge_done)
ret = ext4_ext_try_to_merge_right(inode, path, ex);
return ret;
}
/*
* check if a portion of the "newext" extent overlaps with an
* existing extent.
*
* If there is an overlap discovered, it updates the length of the newext
* such that there will be no overlap, and then returns 1.
* If there is no overlap found, it returns 0.
*/
static unsigned int ext4_ext_check_overlap(struct inode *inode,
struct ext4_extent *newext,
struct ext4_ext_path *path)
{
ext4_lblk_t b1, b2;
unsigned int depth, len1;
unsigned int ret = 0;
b1 = le32_to_cpu(newext->ee_block);
len1 = ext4_ext_get_actual_len(newext);
depth = ext_depth(inode);
if (!path[depth].p_ext)
goto out;
b2 = le32_to_cpu(path[depth].p_ext->ee_block);
/*
* get the next allocated block if the extent in the path
* is before the requested block(s)
*/
if (b2 < b1) {
b2 = ext4_ext_next_allocated_block(path);
ext4: Fix max file size and logical block counting of extent format file Kazuya Mio reported that he was able to hit BUG_ON(next == lblock) in ext4_ext_put_gap_in_cache() while creating a sparse file in extent format and fill the tail of file up to its end. We will hit the BUG_ON when we write the last block (2^32-1) into the sparse file. The root cause of the problem lies in the fact that we specifically set s_maxbytes so that block at s_maxbytes fit into on-disk extent format, which is 32 bit long. However, we are not storing start and end block number, but rather start block number and length in blocks. It means that in order to cover extent from 0 to EXT_MAX_BLOCK we need EXT_MAX_BLOCK+1 to fit into len (because we counting block 0 as well) - and it does not. The only way to fix it without changing the meaning of the struct ext4_extent members is, as Kazuya Mio suggested, to lower s_maxbytes by one fs block so we can cover the whole extent we can get by the on-disk extent format. Also in many places EXT_MAX_BLOCK is used as length instead of maximum logical block number as the name suggests, it is all a bit messy. So this commit renames it to EXT_MAX_BLOCKS and change its usage in some places to actually be maximum number of blocks in the extent. The bug which this commit fixes can be reproduced as follows: dd if=/dev/zero of=/mnt/mp1/file bs=<blocksize> count=1 seek=$((2**32-2)) sync dd if=/dev/zero of=/mnt/mp1/file bs=<blocksize> count=1 seek=$((2**32-1)) Reported-by: Kazuya Mio <k-mio@sx.jp.nec.com> Signed-off-by: Lukas Czerner <lczerner@redhat.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-06-06 04:05:17 +00:00
if (b2 == EXT_MAX_BLOCKS)
goto out;
}
/* check for wrap through zero on extent logical start block*/
if (b1 + len1 < b1) {
ext4: Fix max file size and logical block counting of extent format file Kazuya Mio reported that he was able to hit BUG_ON(next == lblock) in ext4_ext_put_gap_in_cache() while creating a sparse file in extent format and fill the tail of file up to its end. We will hit the BUG_ON when we write the last block (2^32-1) into the sparse file. The root cause of the problem lies in the fact that we specifically set s_maxbytes so that block at s_maxbytes fit into on-disk extent format, which is 32 bit long. However, we are not storing start and end block number, but rather start block number and length in blocks. It means that in order to cover extent from 0 to EXT_MAX_BLOCK we need EXT_MAX_BLOCK+1 to fit into len (because we counting block 0 as well) - and it does not. The only way to fix it without changing the meaning of the struct ext4_extent members is, as Kazuya Mio suggested, to lower s_maxbytes by one fs block so we can cover the whole extent we can get by the on-disk extent format. Also in many places EXT_MAX_BLOCK is used as length instead of maximum logical block number as the name suggests, it is all a bit messy. So this commit renames it to EXT_MAX_BLOCKS and change its usage in some places to actually be maximum number of blocks in the extent. The bug which this commit fixes can be reproduced as follows: dd if=/dev/zero of=/mnt/mp1/file bs=<blocksize> count=1 seek=$((2**32-2)) sync dd if=/dev/zero of=/mnt/mp1/file bs=<blocksize> count=1 seek=$((2**32-1)) Reported-by: Kazuya Mio <k-mio@sx.jp.nec.com> Signed-off-by: Lukas Czerner <lczerner@redhat.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-06-06 04:05:17 +00:00
len1 = EXT_MAX_BLOCKS - b1;
newext->ee_len = cpu_to_le16(len1);
ret = 1;
}
/* check for overlap */
if (b1 + len1 > b2) {
newext->ee_len = cpu_to_le16(b2 - b1);
ret = 1;
}
out:
return ret;
}
/*
* ext4_ext_insert_extent:
* tries to merge requsted extent into the existing extent or
* inserts requested extent as new one into the tree,
* creating new leaf in the no-space case.
*/
int ext4_ext_insert_extent(handle_t *handle, struct inode *inode,
struct ext4_ext_path *path,
struct ext4_extent *newext, int flag)
{
struct ext4_extent_header *eh;
struct ext4_extent *ex, *fex;
struct ext4_extent *nearex; /* nearest extent */
struct ext4_ext_path *npath = NULL;
int depth, len, err;
ext4_lblk_t next;
unsigned uninitialized = 0;
int flags = 0;
if (unlikely(ext4_ext_get_actual_len(newext) == 0)) {
EXT4_ERROR_INODE(inode, "ext4_ext_get_actual_len(newext) == 0");
return -EIO;
}
depth = ext_depth(inode);
ex = path[depth].p_ext;
if (unlikely(path[depth].p_hdr == NULL)) {
EXT4_ERROR_INODE(inode, "path[%d].p_hdr == NULL", depth);
return -EIO;
}
/* try to insert block into found extent and return */
if (ex && !(flag & EXT4_GET_BLOCKS_PRE_IO)
&& ext4_can_extents_be_merged(inode, ex, newext)) {
ext_debug("append [%d]%d block to %d:[%d]%d (from %llu)\n",
ext4_ext_is_uninitialized(newext),
ext4_ext_get_actual_len(newext),
le32_to_cpu(ex->ee_block),
ext4_ext_is_uninitialized(ex),
ext4_ext_get_actual_len(ex),
ext4_ext_pblock(ex));
err = ext4_ext_get_access(handle, inode, path + depth);
if (err)
return err;
/*
* ext4_can_extents_be_merged should have checked that either
* both extents are uninitialized, or both aren't. Thus we
* need to check only one of them here.
*/
if (ext4_ext_is_uninitialized(ex))
uninitialized = 1;
ex->ee_len = cpu_to_le16(ext4_ext_get_actual_len(ex)
+ ext4_ext_get_actual_len(newext));
if (uninitialized)
ext4_ext_mark_uninitialized(ex);
eh = path[depth].p_hdr;
nearex = ex;
goto merge;
}
depth = ext_depth(inode);
eh = path[depth].p_hdr;
if (le16_to_cpu(eh->eh_entries) < le16_to_cpu(eh->eh_max))
goto has_space;
/* probably next leaf has space for us? */
fex = EXT_LAST_EXTENT(eh);
next = EXT_MAX_BLOCKS;
if (le32_to_cpu(newext->ee_block) > le32_to_cpu(fex->ee_block))
next = ext4_ext_next_leaf_block(path);
if (next != EXT_MAX_BLOCKS) {
ext_debug("next leaf block - %d\n", next);
BUG_ON(npath != NULL);
npath = ext4_ext_find_extent(inode, next, NULL);
if (IS_ERR(npath))
return PTR_ERR(npath);
BUG_ON(npath->p_depth != path->p_depth);
eh = npath[depth].p_hdr;
if (le16_to_cpu(eh->eh_entries) < le16_to_cpu(eh->eh_max)) {
ext_debug("next leaf isn't full(%d)\n",
le16_to_cpu(eh->eh_entries));
path = npath;
goto has_space;
}
ext_debug("next leaf has no free space(%d,%d)\n",
le16_to_cpu(eh->eh_entries), le16_to_cpu(eh->eh_max));
}
/*
* There is no free space in the found leaf.
* We're gonna add a new leaf in the tree.
*/
if (flag & EXT4_GET_BLOCKS_PUNCH_OUT_EXT)
flags = EXT4_MB_USE_ROOT_BLOCKS;
err = ext4_ext_create_new_leaf(handle, inode, flags, path, newext);
if (err)
goto cleanup;
depth = ext_depth(inode);
eh = path[depth].p_hdr;
has_space:
nearex = path[depth].p_ext;
err = ext4_ext_get_access(handle, inode, path + depth);
if (err)
goto cleanup;
if (!nearex) {
/* there is no extent in this leaf, create first one */
ext_debug("first extent in the leaf: %d:%llu:[%d]%d\n",
le32_to_cpu(newext->ee_block),
ext4_ext_pblock(newext),
ext4_ext_is_uninitialized(newext),
ext4_ext_get_actual_len(newext));
path[depth].p_ext = EXT_FIRST_EXTENT(eh);
} else if (le32_to_cpu(newext->ee_block)
> le32_to_cpu(nearex->ee_block)) {
/* BUG_ON(newext->ee_block == nearex->ee_block); */
if (nearex != EXT_LAST_EXTENT(eh)) {
len = EXT_MAX_EXTENT(eh) - nearex;
len = (len - 1) * sizeof(struct ext4_extent);
len = len < 0 ? 0 : len;
ext_debug("insert %d:%llu:[%d]%d after: nearest 0x%p, "
"move %d from 0x%p to 0x%p\n",
le32_to_cpu(newext->ee_block),
ext4_ext_pblock(newext),
ext4_ext_is_uninitialized(newext),
ext4_ext_get_actual_len(newext),
nearex, len, nearex + 1, nearex + 2);
memmove(nearex + 2, nearex + 1, len);
}
path[depth].p_ext = nearex + 1;
} else {
BUG_ON(newext->ee_block == nearex->ee_block);
len = (EXT_MAX_EXTENT(eh) - nearex) * sizeof(struct ext4_extent);
len = len < 0 ? 0 : len;
ext_debug("insert %d:%llu:[%d]%d before: nearest 0x%p, "
"move %d from 0x%p to 0x%p\n",
le32_to_cpu(newext->ee_block),
ext4_ext_pblock(newext),
ext4_ext_is_uninitialized(newext),
ext4_ext_get_actual_len(newext),
nearex, len, nearex, nearex + 1);
memmove(nearex + 1, nearex, len);
path[depth].p_ext = nearex;
}
le16_add_cpu(&eh->eh_entries, 1);
nearex = path[depth].p_ext;
nearex->ee_block = newext->ee_block;
ext4_ext_store_pblock(nearex, ext4_ext_pblock(newext));
nearex->ee_len = newext->ee_len;
merge:
/* try to merge extents to the right */
if (!(flag & EXT4_GET_BLOCKS_PRE_IO))
ext4_ext_try_to_merge(inode, path, nearex);
/* try to merge extents to the left */
/* time to correct all indexes above */
err = ext4_ext_correct_indexes(handle, inode, path);
if (err)
goto cleanup;
err = ext4_ext_dirty(handle, inode, path + depth);
cleanup:
if (npath) {
ext4_ext_drop_refs(npath);
kfree(npath);
}
ext4_ext_invalidate_cache(inode);
return err;
}
static int ext4_ext_walk_space(struct inode *inode, ext4_lblk_t block,
ext4_lblk_t num, ext_prepare_callback func,
void *cbdata)
{
struct ext4_ext_path *path = NULL;
struct ext4_ext_cache cbex;
struct ext4_extent *ex;
ext4_lblk_t next, start = 0, end = 0;
ext4_lblk_t last = block + num;
int depth, exists, err = 0;
BUG_ON(func == NULL);
BUG_ON(inode == NULL);
ext4: Fix max file size and logical block counting of extent format file Kazuya Mio reported that he was able to hit BUG_ON(next == lblock) in ext4_ext_put_gap_in_cache() while creating a sparse file in extent format and fill the tail of file up to its end. We will hit the BUG_ON when we write the last block (2^32-1) into the sparse file. The root cause of the problem lies in the fact that we specifically set s_maxbytes so that block at s_maxbytes fit into on-disk extent format, which is 32 bit long. However, we are not storing start and end block number, but rather start block number and length in blocks. It means that in order to cover extent from 0 to EXT_MAX_BLOCK we need EXT_MAX_BLOCK+1 to fit into len (because we counting block 0 as well) - and it does not. The only way to fix it without changing the meaning of the struct ext4_extent members is, as Kazuya Mio suggested, to lower s_maxbytes by one fs block so we can cover the whole extent we can get by the on-disk extent format. Also in many places EXT_MAX_BLOCK is used as length instead of maximum logical block number as the name suggests, it is all a bit messy. So this commit renames it to EXT_MAX_BLOCKS and change its usage in some places to actually be maximum number of blocks in the extent. The bug which this commit fixes can be reproduced as follows: dd if=/dev/zero of=/mnt/mp1/file bs=<blocksize> count=1 seek=$((2**32-2)) sync dd if=/dev/zero of=/mnt/mp1/file bs=<blocksize> count=1 seek=$((2**32-1)) Reported-by: Kazuya Mio <k-mio@sx.jp.nec.com> Signed-off-by: Lukas Czerner <lczerner@redhat.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-06-06 04:05:17 +00:00
while (block < last && block != EXT_MAX_BLOCKS) {
num = last - block;
/* find extent for this block */
ext4: Fix potential fiemap deadlock (mmap_sem vs. i_data_sem) Fix the following potential circular locking dependency between mm->mmap_sem and ei->i_data_sem: ======================================================= [ INFO: possible circular locking dependency detected ] 2.6.32-04115-gec044c5 #37 ------------------------------------------------------- ureadahead/1855 is trying to acquire lock: (&mm->mmap_sem){++++++}, at: [<ffffffff81107224>] might_fault+0x5c/0xac but task is already holding lock: (&ei->i_data_sem){++++..}, at: [<ffffffff811be1fd>] ext4_fiemap+0x11b/0x159 which lock already depends on the new lock. the existing dependency chain (in reverse order) is: -> #1 (&ei->i_data_sem){++++..}: [<ffffffff81099bfa>] __lock_acquire+0xb67/0xd0f [<ffffffff81099e7e>] lock_acquire+0xdc/0x102 [<ffffffff81516633>] down_read+0x51/0x84 [<ffffffff811a2414>] ext4_get_blocks+0x50/0x2a5 [<ffffffff811a3453>] ext4_get_block+0xab/0xef [<ffffffff81154f39>] do_mpage_readpage+0x198/0x48d [<ffffffff81155360>] mpage_readpages+0xd0/0x114 [<ffffffff811a104b>] ext4_readpages+0x1d/0x1f [<ffffffff810f8644>] __do_page_cache_readahead+0x12f/0x1bc [<ffffffff810f86f2>] ra_submit+0x21/0x25 [<ffffffff810f0cfd>] filemap_fault+0x19f/0x32c [<ffffffff81107b97>] __do_fault+0x55/0x3a2 [<ffffffff81109db0>] handle_mm_fault+0x327/0x734 [<ffffffff8151aaa9>] do_page_fault+0x292/0x2aa [<ffffffff81518205>] page_fault+0x25/0x30 [<ffffffff812a34d8>] clear_user+0x38/0x3c [<ffffffff81167e16>] padzero+0x20/0x31 [<ffffffff81168b47>] load_elf_binary+0x8bc/0x17ed [<ffffffff81130e95>] search_binary_handler+0xc2/0x259 [<ffffffff81166d64>] load_script+0x1b8/0x1cc [<ffffffff81130e95>] search_binary_handler+0xc2/0x259 [<ffffffff8113255f>] do_execve+0x1ce/0x2cf [<ffffffff81027494>] sys_execve+0x43/0x5a [<ffffffff8102918a>] stub_execve+0x6a/0xc0 -> #0 (&mm->mmap_sem){++++++}: [<ffffffff81099aa4>] __lock_acquire+0xa11/0xd0f [<ffffffff81099e7e>] lock_acquire+0xdc/0x102 [<ffffffff81107251>] might_fault+0x89/0xac [<ffffffff81139382>] fiemap_fill_next_extent+0x95/0xda [<ffffffff811bcb43>] ext4_ext_fiemap_cb+0x138/0x157 [<ffffffff811be069>] ext4_ext_walk_space+0x178/0x1f1 [<ffffffff811be21e>] ext4_fiemap+0x13c/0x159 [<ffffffff811390e6>] do_vfs_ioctl+0x348/0x4d6 [<ffffffff811392ca>] sys_ioctl+0x56/0x79 [<ffffffff81028cb2>] system_call_fastpath+0x16/0x1b other info that might help us debug this: 1 lock held by ureadahead/1855: #0: (&ei->i_data_sem){++++..}, at: [<ffffffff811be1fd>] ext4_fiemap+0x11b/0x159 stack backtrace: Pid: 1855, comm: ureadahead Not tainted 2.6.32-04115-gec044c5 #37 Call Trace: [<ffffffff81098c70>] print_circular_bug+0xa8/0xb7 [<ffffffff81099aa4>] __lock_acquire+0xa11/0xd0f [<ffffffff8102f229>] ? sched_clock+0x9/0xd [<ffffffff81099e7e>] lock_acquire+0xdc/0x102 [<ffffffff81107224>] ? might_fault+0x5c/0xac [<ffffffff81107251>] might_fault+0x89/0xac [<ffffffff81107224>] ? might_fault+0x5c/0xac [<ffffffff81124b44>] ? __kmalloc+0x13b/0x18c [<ffffffff81139382>] fiemap_fill_next_extent+0x95/0xda [<ffffffff811bcb43>] ext4_ext_fiemap_cb+0x138/0x157 [<ffffffff811bca0b>] ? ext4_ext_fiemap_cb+0x0/0x157 [<ffffffff811be069>] ext4_ext_walk_space+0x178/0x1f1 [<ffffffff811be21e>] ext4_fiemap+0x13c/0x159 [<ffffffff81107224>] ? might_fault+0x5c/0xac [<ffffffff811390e6>] do_vfs_ioctl+0x348/0x4d6 [<ffffffff8129f6d0>] ? __up_read+0x8d/0x95 [<ffffffff81517fb5>] ? retint_swapgs+0x13/0x1b [<ffffffff811392ca>] sys_ioctl+0x56/0x79 [<ffffffff81028cb2>] system_call_fastpath+0x16/0x1b Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2009-12-10 02:30:02 +00:00
down_read(&EXT4_I(inode)->i_data_sem);
path = ext4_ext_find_extent(inode, block, path);
ext4: Fix potential fiemap deadlock (mmap_sem vs. i_data_sem) Fix the following potential circular locking dependency between mm->mmap_sem and ei->i_data_sem: ======================================================= [ INFO: possible circular locking dependency detected ] 2.6.32-04115-gec044c5 #37 ------------------------------------------------------- ureadahead/1855 is trying to acquire lock: (&mm->mmap_sem){++++++}, at: [<ffffffff81107224>] might_fault+0x5c/0xac but task is already holding lock: (&ei->i_data_sem){++++..}, at: [<ffffffff811be1fd>] ext4_fiemap+0x11b/0x159 which lock already depends on the new lock. the existing dependency chain (in reverse order) is: -> #1 (&ei->i_data_sem){++++..}: [<ffffffff81099bfa>] __lock_acquire+0xb67/0xd0f [<ffffffff81099e7e>] lock_acquire+0xdc/0x102 [<ffffffff81516633>] down_read+0x51/0x84 [<ffffffff811a2414>] ext4_get_blocks+0x50/0x2a5 [<ffffffff811a3453>] ext4_get_block+0xab/0xef [<ffffffff81154f39>] do_mpage_readpage+0x198/0x48d [<ffffffff81155360>] mpage_readpages+0xd0/0x114 [<ffffffff811a104b>] ext4_readpages+0x1d/0x1f [<ffffffff810f8644>] __do_page_cache_readahead+0x12f/0x1bc [<ffffffff810f86f2>] ra_submit+0x21/0x25 [<ffffffff810f0cfd>] filemap_fault+0x19f/0x32c [<ffffffff81107b97>] __do_fault+0x55/0x3a2 [<ffffffff81109db0>] handle_mm_fault+0x327/0x734 [<ffffffff8151aaa9>] do_page_fault+0x292/0x2aa [<ffffffff81518205>] page_fault+0x25/0x30 [<ffffffff812a34d8>] clear_user+0x38/0x3c [<ffffffff81167e16>] padzero+0x20/0x31 [<ffffffff81168b47>] load_elf_binary+0x8bc/0x17ed [<ffffffff81130e95>] search_binary_handler+0xc2/0x259 [<ffffffff81166d64>] load_script+0x1b8/0x1cc [<ffffffff81130e95>] search_binary_handler+0xc2/0x259 [<ffffffff8113255f>] do_execve+0x1ce/0x2cf [<ffffffff81027494>] sys_execve+0x43/0x5a [<ffffffff8102918a>] stub_execve+0x6a/0xc0 -> #0 (&mm->mmap_sem){++++++}: [<ffffffff81099aa4>] __lock_acquire+0xa11/0xd0f [<ffffffff81099e7e>] lock_acquire+0xdc/0x102 [<ffffffff81107251>] might_fault+0x89/0xac [<ffffffff81139382>] fiemap_fill_next_extent+0x95/0xda [<ffffffff811bcb43>] ext4_ext_fiemap_cb+0x138/0x157 [<ffffffff811be069>] ext4_ext_walk_space+0x178/0x1f1 [<ffffffff811be21e>] ext4_fiemap+0x13c/0x159 [<ffffffff811390e6>] do_vfs_ioctl+0x348/0x4d6 [<ffffffff811392ca>] sys_ioctl+0x56/0x79 [<ffffffff81028cb2>] system_call_fastpath+0x16/0x1b other info that might help us debug this: 1 lock held by ureadahead/1855: #0: (&ei->i_data_sem){++++..}, at: [<ffffffff811be1fd>] ext4_fiemap+0x11b/0x159 stack backtrace: Pid: 1855, comm: ureadahead Not tainted 2.6.32-04115-gec044c5 #37 Call Trace: [<ffffffff81098c70>] print_circular_bug+0xa8/0xb7 [<ffffffff81099aa4>] __lock_acquire+0xa11/0xd0f [<ffffffff8102f229>] ? sched_clock+0x9/0xd [<ffffffff81099e7e>] lock_acquire+0xdc/0x102 [<ffffffff81107224>] ? might_fault+0x5c/0xac [<ffffffff81107251>] might_fault+0x89/0xac [<ffffffff81107224>] ? might_fault+0x5c/0xac [<ffffffff81124b44>] ? __kmalloc+0x13b/0x18c [<ffffffff81139382>] fiemap_fill_next_extent+0x95/0xda [<ffffffff811bcb43>] ext4_ext_fiemap_cb+0x138/0x157 [<ffffffff811bca0b>] ? ext4_ext_fiemap_cb+0x0/0x157 [<ffffffff811be069>] ext4_ext_walk_space+0x178/0x1f1 [<ffffffff811be21e>] ext4_fiemap+0x13c/0x159 [<ffffffff81107224>] ? might_fault+0x5c/0xac [<ffffffff811390e6>] do_vfs_ioctl+0x348/0x4d6 [<ffffffff8129f6d0>] ? __up_read+0x8d/0x95 [<ffffffff81517fb5>] ? retint_swapgs+0x13/0x1b [<ffffffff811392ca>] sys_ioctl+0x56/0x79 [<ffffffff81028cb2>] system_call_fastpath+0x16/0x1b Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2009-12-10 02:30:02 +00:00
up_read(&EXT4_I(inode)->i_data_sem);
if (IS_ERR(path)) {
err = PTR_ERR(path);
path = NULL;
break;
}
depth = ext_depth(inode);
if (unlikely(path[depth].p_hdr == NULL)) {
EXT4_ERROR_INODE(inode, "path[%d].p_hdr == NULL", depth);
err = -EIO;
break;
}
ex = path[depth].p_ext;
next = ext4_ext_next_allocated_block(path);
exists = 0;
if (!ex) {
/* there is no extent yet, so try to allocate
* all requested space */
start = block;
end = block + num;
} else if (le32_to_cpu(ex->ee_block) > block) {
/* need to allocate space before found extent */
start = block;
end = le32_to_cpu(ex->ee_block);
if (block + num < end)
end = block + num;
} else if (block >= le32_to_cpu(ex->ee_block)
+ ext4_ext_get_actual_len(ex)) {
/* need to allocate space after found extent */
start = block;
end = block + num;
if (end >= next)
end = next;
} else if (block >= le32_to_cpu(ex->ee_block)) {
/*
* some part of requested space is covered
* by found extent
*/
start = block;
end = le32_to_cpu(ex->ee_block)
+ ext4_ext_get_actual_len(ex);
if (block + num < end)
end = block + num;
exists = 1;
} else {
BUG();
}
BUG_ON(end <= start);
if (!exists) {
cbex.ec_block = start;
cbex.ec_len = end - start;
cbex.ec_start = 0;
} else {
cbex.ec_block = le32_to_cpu(ex->ee_block);
cbex.ec_len = ext4_ext_get_actual_len(ex);
cbex.ec_start = ext4_ext_pblock(ex);
}
if (unlikely(cbex.ec_len == 0)) {
EXT4_ERROR_INODE(inode, "cbex.ec_len == 0");
err = -EIO;
break;
}
err = func(inode, next, &cbex, ex, cbdata);
ext4_ext_drop_refs(path);
if (err < 0)
break;
if (err == EXT_REPEAT)
continue;
else if (err == EXT_BREAK) {
err = 0;
break;
}
if (ext_depth(inode) != depth) {
/* depth was changed. we have to realloc path */
kfree(path);
path = NULL;
}
block = cbex.ec_block + cbex.ec_len;
}
if (path) {
ext4_ext_drop_refs(path);
kfree(path);
}
return err;
}
static void
ext4_ext_put_in_cache(struct inode *inode, ext4_lblk_t block,
__u32 len, ext4_fsblk_t start)
{
struct ext4_ext_cache *cex;
BUG_ON(len == 0);
spin_lock(&EXT4_I(inode)->i_block_reservation_lock);
cex = &EXT4_I(inode)->i_cached_extent;
cex->ec_block = block;
cex->ec_len = len;
cex->ec_start = start;
spin_unlock(&EXT4_I(inode)->i_block_reservation_lock);
}
/*
* ext4_ext_put_gap_in_cache:
* calculate boundaries of the gap that the requested block fits into
* and cache this gap
*/
static void
ext4_ext_put_gap_in_cache(struct inode *inode, struct ext4_ext_path *path,
ext4_lblk_t block)
{
int depth = ext_depth(inode);
unsigned long len;
ext4_lblk_t lblock;
struct ext4_extent *ex;
ex = path[depth].p_ext;
if (ex == NULL) {
/* there is no extent yet, so gap is [0;-] */
lblock = 0;
ext4: Fix max file size and logical block counting of extent format file Kazuya Mio reported that he was able to hit BUG_ON(next == lblock) in ext4_ext_put_gap_in_cache() while creating a sparse file in extent format and fill the tail of file up to its end. We will hit the BUG_ON when we write the last block (2^32-1) into the sparse file. The root cause of the problem lies in the fact that we specifically set s_maxbytes so that block at s_maxbytes fit into on-disk extent format, which is 32 bit long. However, we are not storing start and end block number, but rather start block number and length in blocks. It means that in order to cover extent from 0 to EXT_MAX_BLOCK we need EXT_MAX_BLOCK+1 to fit into len (because we counting block 0 as well) - and it does not. The only way to fix it without changing the meaning of the struct ext4_extent members is, as Kazuya Mio suggested, to lower s_maxbytes by one fs block so we can cover the whole extent we can get by the on-disk extent format. Also in many places EXT_MAX_BLOCK is used as length instead of maximum logical block number as the name suggests, it is all a bit messy. So this commit renames it to EXT_MAX_BLOCKS and change its usage in some places to actually be maximum number of blocks in the extent. The bug which this commit fixes can be reproduced as follows: dd if=/dev/zero of=/mnt/mp1/file bs=<blocksize> count=1 seek=$((2**32-2)) sync dd if=/dev/zero of=/mnt/mp1/file bs=<blocksize> count=1 seek=$((2**32-1)) Reported-by: Kazuya Mio <k-mio@sx.jp.nec.com> Signed-off-by: Lukas Czerner <lczerner@redhat.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-06-06 04:05:17 +00:00
len = EXT_MAX_BLOCKS;
ext_debug("cache gap(whole file):");
} else if (block < le32_to_cpu(ex->ee_block)) {
lblock = block;
len = le32_to_cpu(ex->ee_block) - block;
ext_debug("cache gap(before): %u [%u:%u]",
block,
le32_to_cpu(ex->ee_block),
ext4_ext_get_actual_len(ex));
} else if (block >= le32_to_cpu(ex->ee_block)
+ ext4_ext_get_actual_len(ex)) {
ext4_lblk_t next;
lblock = le32_to_cpu(ex->ee_block)
+ ext4_ext_get_actual_len(ex);
next = ext4_ext_next_allocated_block(path);
ext_debug("cache gap(after): [%u:%u] %u",
le32_to_cpu(ex->ee_block),
ext4_ext_get_actual_len(ex),
block);
BUG_ON(next == lblock);
len = next - lblock;
} else {
lblock = len = 0;
BUG();
}
ext_debug(" -> %u:%lu\n", lblock, len);
ext4_ext_put_in_cache(inode, lblock, len, 0);
}
/*
* ext4_ext_check_cache()
* Checks to see if the given block is in the cache.
* If it is, the cached extent is stored in the given
* cache extent pointer. If the cached extent is a hole,
* this routine should be used instead of
* ext4_ext_in_cache if the calling function needs to
* know the size of the hole.
*
* @inode: The files inode
* @block: The block to look for in the cache
* @ex: Pointer where the cached extent will be stored
* if it contains block
*
* Return 0 if cache is invalid; 1 if the cache is valid
*/
static int ext4_ext_check_cache(struct inode *inode, ext4_lblk_t block,
struct ext4_ext_cache *ex){
struct ext4_ext_cache *cex;
struct ext4_sb_info *sbi;
int ret = 0;
/*
* We borrow i_block_reservation_lock to protect i_cached_extent
*/
spin_lock(&EXT4_I(inode)->i_block_reservation_lock);
cex = &EXT4_I(inode)->i_cached_extent;
sbi = EXT4_SB(inode->i_sb);
/* has cache valid data? */
if (cex->ec_len == 0)
goto errout;
if (in_range(block, cex->ec_block, cex->ec_len)) {
memcpy(ex, cex, sizeof(struct ext4_ext_cache));
ext_debug("%u cached by %u:%u:%llu\n",
block,
cex->ec_block, cex->ec_len, cex->ec_start);
ret = 1;
}
errout:
if (!ret)
sbi->extent_cache_misses++;
else
sbi->extent_cache_hits++;
spin_unlock(&EXT4_I(inode)->i_block_reservation_lock);
return ret;
}
/*
* ext4_ext_in_cache()
* Checks to see if the given block is in the cache.
* If it is, the cached extent is stored in the given
* extent pointer.
*
* @inode: The files inode
* @block: The block to look for in the cache
* @ex: Pointer where the cached extent will be stored
* if it contains block
*
* Return 0 if cache is invalid; 1 if the cache is valid
*/
static int
ext4_ext_in_cache(struct inode *inode, ext4_lblk_t block,
struct ext4_extent *ex)
{
struct ext4_ext_cache cex;
int ret = 0;
if (ext4_ext_check_cache(inode, block, &cex)) {
ex->ee_block = cpu_to_le32(cex.ec_block);
ext4_ext_store_pblock(ex, cex.ec_start);
ex->ee_len = cpu_to_le16(cex.ec_len);
ret = 1;
}
return ret;
}
/*
* ext4_ext_rm_idx:
* removes index from the index block.
*/
static int ext4_ext_rm_idx(handle_t *handle, struct inode *inode,
struct ext4_ext_path *path)
{
int err;
ext4_fsblk_t leaf;
/* free index block */
path--;
leaf = ext4_idx_pblock(path->p_idx);
if (unlikely(path->p_hdr->eh_entries == 0)) {
EXT4_ERROR_INODE(inode, "path->p_hdr->eh_entries == 0");
return -EIO;
}
err = ext4_ext_get_access(handle, inode, path);
if (err)
return err;
if (path->p_idx != EXT_LAST_INDEX(path->p_hdr)) {
int len = EXT_LAST_INDEX(path->p_hdr) - path->p_idx;
len *= sizeof(struct ext4_extent_idx);
memmove(path->p_idx, path->p_idx + 1, len);
}
le16_add_cpu(&path->p_hdr->eh_entries, -1);
err = ext4_ext_dirty(handle, inode, path);
if (err)
return err;
ext_debug("index is empty, remove it, free block %llu\n", leaf);
ext4_free_blocks(handle, inode, NULL, leaf, 1,
EXT4_FREE_BLOCKS_METADATA | EXT4_FREE_BLOCKS_FORGET);
return err;
}
/*
* ext4_ext_calc_credits_for_single_extent:
* This routine returns max. credits that needed to insert an extent
* to the extent tree.
* When pass the actual path, the caller should calculate credits
* under i_data_sem.
*/
int ext4_ext_calc_credits_for_single_extent(struct inode *inode, int nrblocks,
struct ext4_ext_path *path)
{
if (path) {
int depth = ext_depth(inode);
int ret = 0;
/* probably there is space in leaf? */
if (le16_to_cpu(path[depth].p_hdr->eh_entries)
< le16_to_cpu(path[depth].p_hdr->eh_max)) {
/*
* There are some space in the leaf tree, no
* need to account for leaf block credit
*
* bitmaps and block group descriptor blocks
* and other metadat blocks still need to be
* accounted.
*/
/* 1 bitmap, 1 block group descriptor */
ret = 2 + EXT4_META_TRANS_BLOCKS(inode->i_sb);
return ret;
}
}
return ext4_chunk_trans_blocks(inode, nrblocks);
}
/*
* How many index/leaf blocks need to change/allocate to modify nrblocks?
*
* if nrblocks are fit in a single extent (chunk flag is 1), then
* in the worse case, each tree level index/leaf need to be changed
* if the tree split due to insert a new extent, then the old tree
* index/leaf need to be updated too
*
* If the nrblocks are discontiguous, they could cause
* the whole tree split more than once, but this is really rare.
*/
int ext4_ext_index_trans_blocks(struct inode *inode, int nrblocks, int chunk)
{
int index;
int depth = ext_depth(inode);
if (chunk)
index = depth * 2;
else
index = depth * 3;
return index;
}
static int ext4_remove_blocks(handle_t *handle, struct inode *inode,
struct ext4_extent *ex,
ext4_lblk_t from, ext4_lblk_t to)
{
unsigned short ee_len = ext4_ext_get_actual_len(ex);
int flags = EXT4_FREE_BLOCKS_FORGET;
if (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode))
flags |= EXT4_FREE_BLOCKS_METADATA;
#ifdef EXTENTS_STATS
{
struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
spin_lock(&sbi->s_ext_stats_lock);
sbi->s_ext_blocks += ee_len;
sbi->s_ext_extents++;
if (ee_len < sbi->s_ext_min)
sbi->s_ext_min = ee_len;
if (ee_len > sbi->s_ext_max)
sbi->s_ext_max = ee_len;
if (ext_depth(inode) > sbi->s_depth_max)
sbi->s_depth_max = ext_depth(inode);
spin_unlock(&sbi->s_ext_stats_lock);
}
#endif
if (from >= le32_to_cpu(ex->ee_block)
&& to == le32_to_cpu(ex->ee_block) + ee_len - 1) {
/* tail removal */
ext4_lblk_t num;
ext4_fsblk_t start;
num = le32_to_cpu(ex->ee_block) + ee_len - from;
start = ext4_ext_pblock(ex) + ee_len - num;
ext_debug("free last %u blocks starting %llu\n", num, start);
ext4_free_blocks(handle, inode, NULL, start, num, flags);
} else if (from == le32_to_cpu(ex->ee_block)
&& to <= le32_to_cpu(ex->ee_block) + ee_len - 1) {
/* head removal */
ext4_lblk_t num;
ext4_fsblk_t start;
num = to - from;
start = ext4_ext_pblock(ex);
ext_debug("free first %u blocks starting %llu\n", num, start);
ext4_free_blocks(handle, inode, 0, start, num, flags);
} else {
printk(KERN_INFO "strange request: removal(2) "
"%u-%u from %u:%u\n",
from, to, le32_to_cpu(ex->ee_block), ee_len);
}
return 0;
}
/*
* ext4_ext_rm_leaf() Removes the extents associated with the
* blocks appearing between "start" and "end", and splits the extents
* if "start" and "end" appear in the same extent
*
* @handle: The journal handle
* @inode: The files inode
* @path: The path to the leaf
* @start: The first block to remove
* @end: The last block to remove
*/
static int
ext4_ext_rm_leaf(handle_t *handle, struct inode *inode,
struct ext4_ext_path *path, ext4_lblk_t start,
ext4_lblk_t end)
{
int err = 0, correct_index = 0;
int depth = ext_depth(inode), credits;
struct ext4_extent_header *eh;
ext4_lblk_t a, b, block;
unsigned num;
ext4_lblk_t ex_ee_block;
unsigned short ex_ee_len;
unsigned uninitialized = 0;
struct ext4_extent *ex;
struct ext4_map_blocks map;
/* the header must be checked already in ext4_ext_remove_space() */
ext_debug("truncate since %u in leaf\n", start);
if (!path[depth].p_hdr)
path[depth].p_hdr = ext_block_hdr(path[depth].p_bh);
eh = path[depth].p_hdr;
if (unlikely(path[depth].p_hdr == NULL)) {
EXT4_ERROR_INODE(inode, "path[%d].p_hdr == NULL", depth);
return -EIO;
}
/* find where to start removing */
ex = EXT_LAST_EXTENT(eh);
ex_ee_block = le32_to_cpu(ex->ee_block);
ex_ee_len = ext4_ext_get_actual_len(ex);
while (ex >= EXT_FIRST_EXTENT(eh) &&
ex_ee_block + ex_ee_len > start) {
if (ext4_ext_is_uninitialized(ex))
uninitialized = 1;
else
uninitialized = 0;
ext_debug("remove ext %u:[%d]%d\n", ex_ee_block,
uninitialized, ex_ee_len);
path[depth].p_ext = ex;
a = ex_ee_block > start ? ex_ee_block : start;
b = ex_ee_block+ex_ee_len - 1 < end ?
ex_ee_block+ex_ee_len - 1 : end;
ext_debug(" border %u:%u\n", a, b);
/* If this extent is beyond the end of the hole, skip it */
if (end <= ex_ee_block) {
ex--;
ex_ee_block = le32_to_cpu(ex->ee_block);
ex_ee_len = ext4_ext_get_actual_len(ex);
continue;
} else if (a != ex_ee_block &&
b != ex_ee_block + ex_ee_len - 1) {
/*
* If this is a truncate, then this condition should
* never happen because at least one of the end points
* needs to be on the edge of the extent.
*/
ext4: Fix max file size and logical block counting of extent format file Kazuya Mio reported that he was able to hit BUG_ON(next == lblock) in ext4_ext_put_gap_in_cache() while creating a sparse file in extent format and fill the tail of file up to its end. We will hit the BUG_ON when we write the last block (2^32-1) into the sparse file. The root cause of the problem lies in the fact that we specifically set s_maxbytes so that block at s_maxbytes fit into on-disk extent format, which is 32 bit long. However, we are not storing start and end block number, but rather start block number and length in blocks. It means that in order to cover extent from 0 to EXT_MAX_BLOCK we need EXT_MAX_BLOCK+1 to fit into len (because we counting block 0 as well) - and it does not. The only way to fix it without changing the meaning of the struct ext4_extent members is, as Kazuya Mio suggested, to lower s_maxbytes by one fs block so we can cover the whole extent we can get by the on-disk extent format. Also in many places EXT_MAX_BLOCK is used as length instead of maximum logical block number as the name suggests, it is all a bit messy. So this commit renames it to EXT_MAX_BLOCKS and change its usage in some places to actually be maximum number of blocks in the extent. The bug which this commit fixes can be reproduced as follows: dd if=/dev/zero of=/mnt/mp1/file bs=<blocksize> count=1 seek=$((2**32-2)) sync dd if=/dev/zero of=/mnt/mp1/file bs=<blocksize> count=1 seek=$((2**32-1)) Reported-by: Kazuya Mio <k-mio@sx.jp.nec.com> Signed-off-by: Lukas Czerner <lczerner@redhat.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-06-06 04:05:17 +00:00
if (end == EXT_MAX_BLOCKS - 1) {
ext_debug(" bad truncate %u:%u\n",
start, end);
block = 0;
num = 0;
err = -EIO;
goto out;
}
/*
* else this is a hole punch, so the extent needs to
* be split since neither edge of the hole is on the
* extent edge
*/
else{
map.m_pblk = ext4_ext_pblock(ex);
map.m_lblk = ex_ee_block;
map.m_len = b - ex_ee_block;
err = ext4_split_extent(handle,
inode, path, &map, 0,
EXT4_GET_BLOCKS_PUNCH_OUT_EXT |
EXT4_GET_BLOCKS_PRE_IO);
if (err < 0)
goto out;
ex_ee_len = ext4_ext_get_actual_len(ex);
b = ex_ee_block+ex_ee_len - 1 < end ?
ex_ee_block+ex_ee_len - 1 : end;
/* Then remove tail of this extent */
block = ex_ee_block;
num = a - block;
}
} else if (a != ex_ee_block) {
/* remove tail of the extent */
block = ex_ee_block;
num = a - block;
} else if (b != ex_ee_block + ex_ee_len - 1) {
/* remove head of the extent */
block = b;
num = ex_ee_block + ex_ee_len - b;
/*
* If this is a truncate, this condition
* should never happen
*/
ext4: Fix max file size and logical block counting of extent format file Kazuya Mio reported that he was able to hit BUG_ON(next == lblock) in ext4_ext_put_gap_in_cache() while creating a sparse file in extent format and fill the tail of file up to its end. We will hit the BUG_ON when we write the last block (2^32-1) into the sparse file. The root cause of the problem lies in the fact that we specifically set s_maxbytes so that block at s_maxbytes fit into on-disk extent format, which is 32 bit long. However, we are not storing start and end block number, but rather start block number and length in blocks. It means that in order to cover extent from 0 to EXT_MAX_BLOCK we need EXT_MAX_BLOCK+1 to fit into len (because we counting block 0 as well) - and it does not. The only way to fix it without changing the meaning of the struct ext4_extent members is, as Kazuya Mio suggested, to lower s_maxbytes by one fs block so we can cover the whole extent we can get by the on-disk extent format. Also in many places EXT_MAX_BLOCK is used as length instead of maximum logical block number as the name suggests, it is all a bit messy. So this commit renames it to EXT_MAX_BLOCKS and change its usage in some places to actually be maximum number of blocks in the extent. The bug which this commit fixes can be reproduced as follows: dd if=/dev/zero of=/mnt/mp1/file bs=<blocksize> count=1 seek=$((2**32-2)) sync dd if=/dev/zero of=/mnt/mp1/file bs=<blocksize> count=1 seek=$((2**32-1)) Reported-by: Kazuya Mio <k-mio@sx.jp.nec.com> Signed-off-by: Lukas Czerner <lczerner@redhat.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-06-06 04:05:17 +00:00
if (end == EXT_MAX_BLOCKS - 1) {
ext_debug(" bad truncate %u:%u\n",
start, end);
err = -EIO;
goto out;
}
} else {
/* remove whole extent: excellent! */
block = ex_ee_block;
num = 0;
if (a != ex_ee_block) {
ext_debug(" bad truncate %u:%u\n",
start, end);
err = -EIO;
goto out;
}
if (b != ex_ee_block + ex_ee_len - 1) {
ext_debug(" bad truncate %u:%u\n",
start, end);
err = -EIO;
goto out;
}
}
/*
* 3 for leaf, sb, and inode plus 2 (bmap and group
* descriptor) for each block group; assume two block
* groups plus ex_ee_len/blocks_per_block_group for
* the worst case
*/
credits = 7 + 2*(ex_ee_len/EXT4_BLOCKS_PER_GROUP(inode->i_sb));
if (ex == EXT_FIRST_EXTENT(eh)) {
correct_index = 1;
credits += (ext_depth(inode)) + 1;
}
credits += EXT4_MAXQUOTAS_TRANS_BLOCKS(inode->i_sb);
err = ext4_ext_truncate_extend_restart(handle, inode, credits);
if (err)
goto out;
err = ext4_ext_get_access(handle, inode, path + depth);
if (err)
goto out;
err = ext4_remove_blocks(handle, inode, ex, a, b);
if (err)
goto out;
if (num == 0) {
/* this extent is removed; mark slot entirely unused */
ext4_ext_store_pblock(ex, 0);
} else if (block != ex_ee_block) {
/*
* If this was a head removal, then we need to update
* the physical block since it is now at a different
* location
*/
ext4_ext_store_pblock(ex, ext4_ext_pblock(ex) + (b-a));
}
ex->ee_block = cpu_to_le32(block);
ex->ee_len = cpu_to_le16(num);
/*
* Do not mark uninitialized if all the blocks in the
* extent have been removed.
*/
if (uninitialized && num)
ext4_ext_mark_uninitialized(ex);
err = ext4_ext_dirty(handle, inode, path + depth);
if (err)
goto out;
/*
* If the extent was completely released,
* we need to remove it from the leaf
*/
if (num == 0) {
ext4: Fix max file size and logical block counting of extent format file Kazuya Mio reported that he was able to hit BUG_ON(next == lblock) in ext4_ext_put_gap_in_cache() while creating a sparse file in extent format and fill the tail of file up to its end. We will hit the BUG_ON when we write the last block (2^32-1) into the sparse file. The root cause of the problem lies in the fact that we specifically set s_maxbytes so that block at s_maxbytes fit into on-disk extent format, which is 32 bit long. However, we are not storing start and end block number, but rather start block number and length in blocks. It means that in order to cover extent from 0 to EXT_MAX_BLOCK we need EXT_MAX_BLOCK+1 to fit into len (because we counting block 0 as well) - and it does not. The only way to fix it without changing the meaning of the struct ext4_extent members is, as Kazuya Mio suggested, to lower s_maxbytes by one fs block so we can cover the whole extent we can get by the on-disk extent format. Also in many places EXT_MAX_BLOCK is used as length instead of maximum logical block number as the name suggests, it is all a bit messy. So this commit renames it to EXT_MAX_BLOCKS and change its usage in some places to actually be maximum number of blocks in the extent. The bug which this commit fixes can be reproduced as follows: dd if=/dev/zero of=/mnt/mp1/file bs=<blocksize> count=1 seek=$((2**32-2)) sync dd if=/dev/zero of=/mnt/mp1/file bs=<blocksize> count=1 seek=$((2**32-1)) Reported-by: Kazuya Mio <k-mio@sx.jp.nec.com> Signed-off-by: Lukas Czerner <lczerner@redhat.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-06-06 04:05:17 +00:00
if (end != EXT_MAX_BLOCKS - 1) {
/*
* For hole punching, we need to scoot all the
* extents up when an extent is removed so that
* we dont have blank extents in the middle
*/
memmove(ex, ex+1, (EXT_LAST_EXTENT(eh) - ex) *
sizeof(struct ext4_extent));
/* Now get rid of the one at the end */
memset(EXT_LAST_EXTENT(eh), 0,
sizeof(struct ext4_extent));
}
le16_add_cpu(&eh->eh_entries, -1);
}
ext_debug("new extent: %u:%u:%llu\n", block, num,
ext4_ext_pblock(ex));
ex--;
ex_ee_block = le32_to_cpu(ex->ee_block);
ex_ee_len = ext4_ext_get_actual_len(ex);
}
if (correct_index && eh->eh_entries)
err = ext4_ext_correct_indexes(handle, inode, path);
/* if this leaf is free, then we should
* remove it from index block above */
if (err == 0 && eh->eh_entries == 0 && path[depth].p_bh != NULL)
err = ext4_ext_rm_idx(handle, inode, path + depth);
out:
return err;
}
/*
* ext4_ext_more_to_rm:
* returns 1 if current index has to be freed (even partial)
*/
static int
ext4_ext_more_to_rm(struct ext4_ext_path *path)
{
BUG_ON(path->p_idx == NULL);
if (path->p_idx < EXT_FIRST_INDEX(path->p_hdr))
return 0;
/*
* if truncate on deeper level happened, it wasn't partial,
* so we have to consider current index for truncation
*/
if (le16_to_cpu(path->p_hdr->eh_entries) == path->p_block)
return 0;
return 1;
}
static int ext4_ext_remove_space(struct inode *inode, ext4_lblk_t start)
{
struct super_block *sb = inode->i_sb;
int depth = ext_depth(inode);
struct ext4_ext_path *path;
handle_t *handle;
int i, err;
ext_debug("truncate since %u\n", start);
/* probably first extent we're gonna free will be last in block */
handle = ext4_journal_start(inode, depth + 1);
if (IS_ERR(handle))
return PTR_ERR(handle);
again:
ext4_ext_invalidate_cache(inode);
/*
* We start scanning from right side, freeing all the blocks
* after i_size and walking into the tree depth-wise.
*/
depth = ext_depth(inode);
path = kzalloc(sizeof(struct ext4_ext_path) * (depth + 1), GFP_NOFS);
if (path == NULL) {
ext4_journal_stop(handle);
return -ENOMEM;
}
path[0].p_depth = depth;
path[0].p_hdr = ext_inode_hdr(inode);
if (ext4_ext_check(inode, path[0].p_hdr, depth)) {
err = -EIO;
goto out;
}
i = err = 0;
while (i >= 0 && err == 0) {
if (i == depth) {
/* this is leaf block */
err = ext4_ext_rm_leaf(handle, inode, path,
start, EXT_MAX_BLOCKS - 1);
/* root level has p_bh == NULL, brelse() eats this */
brelse(path[i].p_bh);
path[i].p_bh = NULL;
i--;
continue;
}
/* this is index block */
if (!path[i].p_hdr) {
ext_debug("initialize header\n");
path[i].p_hdr = ext_block_hdr(path[i].p_bh);
}
if (!path[i].p_idx) {
/* this level hasn't been touched yet */
path[i].p_idx = EXT_LAST_INDEX(path[i].p_hdr);
path[i].p_block = le16_to_cpu(path[i].p_hdr->eh_entries)+1;
ext_debug("init index ptr: hdr 0x%p, num %d\n",
path[i].p_hdr,
le16_to_cpu(path[i].p_hdr->eh_entries));
} else {
/* we were already here, see at next index */
path[i].p_idx--;
}
ext_debug("level %d - index, first 0x%p, cur 0x%p\n",
i, EXT_FIRST_INDEX(path[i].p_hdr),
path[i].p_idx);
if (ext4_ext_more_to_rm(path + i)) {
struct buffer_head *bh;
/* go to the next level */
ext_debug("move to level %d (block %llu)\n",
i + 1, ext4_idx_pblock(path[i].p_idx));
memset(path + i + 1, 0, sizeof(*path));
bh = sb_bread(sb, ext4_idx_pblock(path[i].p_idx));
if (!bh) {
/* should we reset i_size? */
err = -EIO;
break;
}
if (WARN_ON(i + 1 > depth)) {
err = -EIO;
break;
}
if (ext4_ext_check(inode, ext_block_hdr(bh),
depth - i - 1)) {
err = -EIO;
break;
}
path[i + 1].p_bh = bh;
/* save actual number of indexes since this
* number is changed at the next iteration */
path[i].p_block = le16_to_cpu(path[i].p_hdr->eh_entries);
i++;
} else {
/* we finished processing this index, go up */
if (path[i].p_hdr->eh_entries == 0 && i > 0) {
/* index is empty, remove it;
* handle must be already prepared by the
* truncatei_leaf() */
err = ext4_ext_rm_idx(handle, inode, path + i);
}
/* root level has p_bh == NULL, brelse() eats this */
brelse(path[i].p_bh);
path[i].p_bh = NULL;
i--;
ext_debug("return to level %d\n", i);
}
}
/* TODO: flexible tree reduction should be here */
if (path->p_hdr->eh_entries == 0) {
/*
* truncate to zero freed all the tree,
* so we need to correct eh_depth
*/
err = ext4_ext_get_access(handle, inode, path);
if (err == 0) {
ext_inode_hdr(inode)->eh_depth = 0;
ext_inode_hdr(inode)->eh_max =
cpu_to_le16(ext4_ext_space_root(inode, 0));
err = ext4_ext_dirty(handle, inode, path);
}
}
out:
ext4_ext_drop_refs(path);
kfree(path);
if (err == -EAGAIN)
goto again;
ext4_journal_stop(handle);
return err;
}
/*
* called at mount time
*/
void ext4_ext_init(struct super_block *sb)
{
/*
* possible initialization would be here
*/
if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
#if defined(AGGRESSIVE_TEST) || defined(CHECK_BINSEARCH) || defined(EXTENTS_STATS)
printk(KERN_INFO "EXT4-fs: file extents enabled");
#ifdef AGGRESSIVE_TEST
printk(", aggressive tests");
#endif
#ifdef CHECK_BINSEARCH
printk(", check binsearch");
#endif
#ifdef EXTENTS_STATS
printk(", stats");
#endif
printk("\n");
#endif
#ifdef EXTENTS_STATS
spin_lock_init(&EXT4_SB(sb)->s_ext_stats_lock);
EXT4_SB(sb)->s_ext_min = 1 << 30;
EXT4_SB(sb)->s_ext_max = 0;
#endif
}
}
/*
* called at umount time
*/
void ext4_ext_release(struct super_block *sb)
{
if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
return;
#ifdef EXTENTS_STATS
if (EXT4_SB(sb)->s_ext_blocks && EXT4_SB(sb)->s_ext_extents) {
struct ext4_sb_info *sbi = EXT4_SB(sb);
printk(KERN_ERR "EXT4-fs: %lu blocks in %lu extents (%lu ave)\n",
sbi->s_ext_blocks, sbi->s_ext_extents,
sbi->s_ext_blocks / sbi->s_ext_extents);
printk(KERN_ERR "EXT4-fs: extents: %lu min, %lu max, max depth %lu\n",
sbi->s_ext_min, sbi->s_ext_max, sbi->s_depth_max);
}
#endif
}
/* FIXME!! we need to try to merge to left or right after zero-out */
static int ext4_ext_zeroout(struct inode *inode, struct ext4_extent *ex)
{
ext4_fsblk_t ee_pblock;
unsigned int ee_len;
int ret;
ee_len = ext4_ext_get_actual_len(ex);
ee_pblock = ext4_ext_pblock(ex);
ret = sb_issue_zeroout(inode->i_sb, ee_pblock, ee_len, GFP_NOFS);
if (ret > 0)
ret = 0;
return ret;
}
/*
* used by extent splitting.
*/
#define EXT4_EXT_MAY_ZEROOUT 0x1 /* safe to zeroout if split fails \
due to ENOSPC */
#define EXT4_EXT_MARK_UNINIT1 0x2 /* mark first half uninitialized */
#define EXT4_EXT_MARK_UNINIT2 0x4 /* mark second half uninitialized */
/*
* ext4_split_extent_at() splits an extent at given block.
*
* @handle: the journal handle
* @inode: the file inode
* @path: the path to the extent
* @split: the logical block where the extent is splitted.
* @split_flags: indicates if the extent could be zeroout if split fails, and
* the states(init or uninit) of new extents.
* @flags: flags used to insert new extent to extent tree.
*
*
* Splits extent [a, b] into two extents [a, @split) and [@split, b], states
* of which are deterimined by split_flag.
*
* There are two cases:
* a> the extent are splitted into two extent.
* b> split is not needed, and just mark the extent.
*
* return 0 on success.
*/
static int ext4_split_extent_at(handle_t *handle,
struct inode *inode,
struct ext4_ext_path *path,
ext4_lblk_t split,
int split_flag,
int flags)
{
ext4_fsblk_t newblock;
ext4_lblk_t ee_block;
struct ext4_extent *ex, newex, orig_ex;
struct ext4_extent *ex2 = NULL;
unsigned int ee_len, depth;
int err = 0;
ext_debug("ext4_split_extents_at: inode %lu, logical"
"block %llu\n", inode->i_ino, (unsigned long long)split);
ext4_ext_show_leaf(inode, path);
depth = ext_depth(inode);
ex = path[depth].p_ext;
ee_block = le32_to_cpu(ex->ee_block);
ee_len = ext4_ext_get_actual_len(ex);
newblock = split - ee_block + ext4_ext_pblock(ex);
BUG_ON(split < ee_block || split >= (ee_block + ee_len));
err = ext4_ext_get_access(handle, inode, path + depth);
if (err)
goto out;
if (split == ee_block) {
/*
* case b: block @split is the block that the extent begins with
* then we just change the state of the extent, and splitting
* is not needed.
*/
if (split_flag & EXT4_EXT_MARK_UNINIT2)
ext4_ext_mark_uninitialized(ex);
else
ext4_ext_mark_initialized(ex);
if (!(flags & EXT4_GET_BLOCKS_PRE_IO))
ext4_ext_try_to_merge(inode, path, ex);
err = ext4_ext_dirty(handle, inode, path + depth);
goto out;
}
/* case a */
memcpy(&orig_ex, ex, sizeof(orig_ex));
ex->ee_len = cpu_to_le16(split - ee_block);
if (split_flag & EXT4_EXT_MARK_UNINIT1)
ext4_ext_mark_uninitialized(ex);
/*
* path may lead to new leaf, not to original leaf any more
* after ext4_ext_insert_extent() returns,
*/
err = ext4_ext_dirty(handle, inode, path + depth);
if (err)
goto fix_extent_len;
ex2 = &newex;
ex2->ee_block = cpu_to_le32(split);
ex2->ee_len = cpu_to_le16(ee_len - (split - ee_block));
ext4_ext_store_pblock(ex2, newblock);
if (split_flag & EXT4_EXT_MARK_UNINIT2)
ext4_ext_mark_uninitialized(ex2);
err = ext4_ext_insert_extent(handle, inode, path, &newex, flags);
if (err == -ENOSPC && (EXT4_EXT_MAY_ZEROOUT & split_flag)) {
err = ext4_ext_zeroout(inode, &orig_ex);
if (err)
goto fix_extent_len;
/* update the extent length and mark as initialized */
ex->ee_len = cpu_to_le32(ee_len);
ext4_ext_try_to_merge(inode, path, ex);
err = ext4_ext_dirty(handle, inode, path + depth);
goto out;
} else if (err)
goto fix_extent_len;
out:
ext4_ext_show_leaf(inode, path);
return err;
fix_extent_len:
ex->ee_len = orig_ex.ee_len;
ext4_ext_dirty(handle, inode, path + depth);
return err;
}
/*
* ext4_split_extents() splits an extent and mark extent which is covered
* by @map as split_flags indicates
*
* It may result in splitting the extent into multiple extents (upto three)
* There are three possibilities:
* a> There is no split required
* b> Splits in two extents: Split is happening at either end of the extent
* c> Splits in three extents: Somone is splitting in middle of the extent
*
*/
static int ext4_split_extent(handle_t *handle,
struct inode *inode,
struct ext4_ext_path *path,
struct ext4_map_blocks *map,
int split_flag,
int flags)
{
ext4_lblk_t ee_block;
struct ext4_extent *ex;
unsigned int ee_len, depth;
int err = 0;
int uninitialized;
int split_flag1, flags1;
depth = ext_depth(inode);
ex = path[depth].p_ext;
ee_block = le32_to_cpu(ex->ee_block);
ee_len = ext4_ext_get_actual_len(ex);
uninitialized = ext4_ext_is_uninitialized(ex);
if (map->m_lblk + map->m_len < ee_block + ee_len) {
split_flag1 = split_flag & EXT4_EXT_MAY_ZEROOUT ?
EXT4_EXT_MAY_ZEROOUT : 0;
flags1 = flags | EXT4_GET_BLOCKS_PRE_IO;
if (uninitialized)
split_flag1 |= EXT4_EXT_MARK_UNINIT1 |
EXT4_EXT_MARK_UNINIT2;
err = ext4_split_extent_at(handle, inode, path,
map->m_lblk + map->m_len, split_flag1, flags1);
if (err)
goto out;
}
ext4_ext_drop_refs(path);
path = ext4_ext_find_extent(inode, map->m_lblk, path);
if (IS_ERR(path))
return PTR_ERR(path);
if (map->m_lblk >= ee_block) {
split_flag1 = split_flag & EXT4_EXT_MAY_ZEROOUT ?
EXT4_EXT_MAY_ZEROOUT : 0;
if (uninitialized)
split_flag1 |= EXT4_EXT_MARK_UNINIT1;
if (split_flag & EXT4_EXT_MARK_UNINIT2)
split_flag1 |= EXT4_EXT_MARK_UNINIT2;
err = ext4_split_extent_at(handle, inode, path,
map->m_lblk, split_flag1, flags);
if (err)
goto out;
}
ext4_ext_show_leaf(inode, path);
out:
return err ? err : map->m_len;
}
#define EXT4_EXT_ZERO_LEN 7
/*
* This function is called by ext4_ext_map_blocks() if someone tries to write
* to an uninitialized extent. It may result in splitting the uninitialized
* extent into multiple extents (up to three - one initialized and two
* uninitialized).
* There are three possibilities:
* a> There is no split required: Entire extent should be initialized
* b> Splits in two extents: Write is happening at either end of the extent
* c> Splits in three extents: Somone is writing in middle of the extent
*/
static int ext4_ext_convert_to_initialized(handle_t *handle,
struct inode *inode,
struct ext4_map_blocks *map,
struct ext4_ext_path *path)
{
struct ext4_map_blocks split_map;
struct ext4_extent zero_ex;
struct ext4_extent *ex;
ext4_lblk_t ee_block, eof_block;
unsigned int allocated, ee_len, depth;
int err = 0;
int split_flag = 0;
ext_debug("ext4_ext_convert_to_initialized: inode %lu, logical"
"block %llu, max_blocks %u\n", inode->i_ino,
(unsigned long long)map->m_lblk, map->m_len);
eof_block = (inode->i_size + inode->i_sb->s_blocksize - 1) >>
inode->i_sb->s_blocksize_bits;
if (eof_block < map->m_lblk + map->m_len)
eof_block = map->m_lblk + map->m_len;
depth = ext_depth(inode);
ex = path[depth].p_ext;
ee_block = le32_to_cpu(ex->ee_block);
ee_len = ext4_ext_get_actual_len(ex);
allocated = ee_len - (map->m_lblk - ee_block);
WARN_ON(map->m_lblk < ee_block);
/*
* It is safe to convert extent to initialized via explicit
* zeroout only if extent is fully insde i_size or new_size.
*/
split_flag |= ee_block + ee_len <= eof_block ? EXT4_EXT_MAY_ZEROOUT : 0;
/* If extent has less than 2*EXT4_EXT_ZERO_LEN zerout directly */
if (ee_len <= 2*EXT4_EXT_ZERO_LEN &&
(EXT4_EXT_MAY_ZEROOUT & split_flag)) {
err = ext4_ext_zeroout(inode, ex);
if (err)
goto out;
err = ext4_ext_get_access(handle, inode, path + depth);
if (err)
goto out;
ext4_ext_mark_initialized(ex);
ext4_ext_try_to_merge(inode, path, ex);
err = ext4_ext_dirty(handle, inode, path + depth);
goto out;
}
/*
* four cases:
* 1. split the extent into three extents.
* 2. split the extent into two extents, zeroout the first half.
* 3. split the extent into two extents, zeroout the second half.
* 4. split the extent into two extents with out zeroout.
*/
split_map.m_lblk = map->m_lblk;
split_map.m_len = map->m_len;
if (allocated > map->m_len) {
if (allocated <= EXT4_EXT_ZERO_LEN &&
(EXT4_EXT_MAY_ZEROOUT & split_flag)) {
/* case 3 */
zero_ex.ee_block =
cpu_to_le32(map->m_lblk);
zero_ex.ee_len = cpu_to_le16(allocated);
ext4_ext_store_pblock(&zero_ex,
ext4_ext_pblock(ex) + map->m_lblk - ee_block);
err = ext4_ext_zeroout(inode, &zero_ex);
if (err)
goto out;
split_map.m_lblk = map->m_lblk;
split_map.m_len = allocated;
} else if ((map->m_lblk - ee_block + map->m_len <
EXT4_EXT_ZERO_LEN) &&
(EXT4_EXT_MAY_ZEROOUT & split_flag)) {
/* case 2 */
if (map->m_lblk != ee_block) {
zero_ex.ee_block = ex->ee_block;
zero_ex.ee_len = cpu_to_le16(map->m_lblk -
ee_block);
ext4_ext_store_pblock(&zero_ex,
ext4_ext_pblock(ex));
err = ext4_ext_zeroout(inode, &zero_ex);
if (err)
goto out;
}
split_map.m_lblk = ee_block;
split_map.m_len = map->m_lblk - ee_block + map->m_len;
allocated = map->m_len;
}
}
allocated = ext4_split_extent(handle, inode, path,
&split_map, split_flag, 0);
if (allocated < 0)
err = allocated;
out:
return err ? err : allocated;
}
/*
* This function is called by ext4_ext_map_blocks() from
* ext4_get_blocks_dio_write() when DIO to write
* to an uninitialized extent.
*
* Writing to an uninitialized extent may result in splitting the uninitialized
* extent into multiple /initialized uninitialized extents (up to three)
* There are three possibilities:
* a> There is no split required: Entire extent should be uninitialized
* b> Splits in two extents: Write is happening at either end of the extent
* c> Splits in three extents: Somone is writing in middle of the extent
*
* One of more index blocks maybe needed if the extent tree grow after
* the uninitialized extent split. To prevent ENOSPC occur at the IO
* complete, we need to split the uninitialized extent before DIO submit
* the IO. The uninitialized extent called at this time will be split
* into three uninitialized extent(at most). After IO complete, the part
* being filled will be convert to initialized by the end_io callback function
* via ext4_convert_unwritten_extents().
*
* Returns the size of uninitialized extent to be written on success.
*/
static int ext4_split_unwritten_extents(handle_t *handle,
struct inode *inode,
struct ext4_map_blocks *map,
struct ext4_ext_path *path,
int flags)
{
ext4_lblk_t eof_block;
ext4_lblk_t ee_block;
struct ext4_extent *ex;
unsigned int ee_len;
int split_flag = 0, depth;
ext_debug("ext4_split_unwritten_extents: inode %lu, logical"
"block %llu, max_blocks %u\n", inode->i_ino,
(unsigned long long)map->m_lblk, map->m_len);
eof_block = (inode->i_size + inode->i_sb->s_blocksize - 1) >>
inode->i_sb->s_blocksize_bits;
if (eof_block < map->m_lblk + map->m_len)
eof_block = map->m_lblk + map->m_len;
/*
* It is safe to convert extent to initialized via explicit
* zeroout only if extent is fully insde i_size or new_size.
*/
depth = ext_depth(inode);
ex = path[depth].p_ext;
ee_block = le32_to_cpu(ex->ee_block);
ee_len = ext4_ext_get_actual_len(ex);
split_flag |= ee_block + ee_len <= eof_block ? EXT4_EXT_MAY_ZEROOUT : 0;
split_flag |= EXT4_EXT_MARK_UNINIT2;
flags |= EXT4_GET_BLOCKS_PRE_IO;
return ext4_split_extent(handle, inode, path, map, split_flag, flags);
}
static int ext4_convert_unwritten_extents_endio(handle_t *handle,
struct inode *inode,
struct ext4_ext_path *path)
{
struct ext4_extent *ex;
int depth;
int err = 0;
depth = ext_depth(inode);
ex = path[depth].p_ext;
ext_debug("ext4_convert_unwritten_extents_endio: inode %lu, logical"
"block %llu, max_blocks %u\n", inode->i_ino,
(unsigned long long)le32_to_cpu(ex->ee_block),
ext4_ext_get_actual_len(ex));
err = ext4_ext_get_access(handle, inode, path + depth);
if (err)
goto out;
/* first mark the extent as initialized */
ext4_ext_mark_initialized(ex);
/* note: ext4_ext_correct_indexes() isn't needed here because
* borders are not changed
*/
ext4_ext_try_to_merge(inode, path, ex);
/* Mark modified extent as dirty */
err = ext4_ext_dirty(handle, inode, path + depth);
out:
ext4_ext_show_leaf(inode, path);
return err;
}
static void unmap_underlying_metadata_blocks(struct block_device *bdev,
sector_t block, int count)
{
int i;
for (i = 0; i < count; i++)
unmap_underlying_metadata(bdev, block + i);
}
/*
* Handle EOFBLOCKS_FL flag, clearing it if necessary
*/
static int check_eofblocks_fl(handle_t *handle, struct inode *inode,
ext4_lblk_t lblk,
struct ext4_ext_path *path,
unsigned int len)
{
int i, depth;
struct ext4_extent_header *eh;
struct ext4_extent *last_ex;
if (!ext4_test_inode_flag(inode, EXT4_INODE_EOFBLOCKS))
return 0;
depth = ext_depth(inode);
eh = path[depth].p_hdr;
if (unlikely(!eh->eh_entries)) {
EXT4_ERROR_INODE(inode, "eh->eh_entries == 0 and "
"EOFBLOCKS_FL set");
return -EIO;
}
last_ex = EXT_LAST_EXTENT(eh);
/*
* We should clear the EOFBLOCKS_FL flag if we are writing the
* last block in the last extent in the file. We test this by
* first checking to see if the caller to
* ext4_ext_get_blocks() was interested in the last block (or
* a block beyond the last block) in the current extent. If
* this turns out to be false, we can bail out from this
* function immediately.
*/
if (lblk + len < le32_to_cpu(last_ex->ee_block) +
ext4_ext_get_actual_len(last_ex))
return 0;
/*
* If the caller does appear to be planning to write at or
* beyond the end of the current extent, we then test to see
* if the current extent is the last extent in the file, by
* checking to make sure it was reached via the rightmost node
* at each level of the tree.
*/
for (i = depth-1; i >= 0; i--)
if (path[i].p_idx != EXT_LAST_INDEX(path[i].p_hdr))
return 0;
ext4_clear_inode_flag(inode, EXT4_INODE_EOFBLOCKS);
return ext4_mark_inode_dirty(handle, inode);
}
static int
ext4_ext_handle_uninitialized_extents(handle_t *handle, struct inode *inode,
struct ext4_map_blocks *map,
struct ext4_ext_path *path, int flags,
unsigned int allocated, ext4_fsblk_t newblock)
{
int ret = 0;
int err = 0;
ext4_io_end_t *io = EXT4_I(inode)->cur_aio_dio;
ext_debug("ext4_ext_handle_uninitialized_extents: inode %lu, logical"
"block %llu, max_blocks %u, flags %d, allocated %u",
inode->i_ino, (unsigned long long)map->m_lblk, map->m_len,
flags, allocated);
ext4_ext_show_leaf(inode, path);
/* get_block() before submit the IO, split the extent */
if ((flags & EXT4_GET_BLOCKS_PRE_IO)) {
ret = ext4_split_unwritten_extents(handle, inode, map,
path, flags);
/*
* Flag the inode(non aio case) or end_io struct (aio case)
* that this IO needs to conversion to written when IO is
* completed
*/
ext4: serialize unaligned asynchronous DIO ext4 has a data corruption case when doing non-block-aligned asynchronous direct IO into a sparse file, as demonstrated by xfstest 240. The root cause is that while ext4 preallocates space in the hole, mappings of that space still look "new" and dio_zero_block() will zero out the unwritten portions. When more than one AIO thread is going, they both find this "new" block and race to zero out their portion; this is uncoordinated and causes data corruption. Dave Chinner fixed this for xfs by simply serializing all unaligned asynchronous direct IO. I've done the same here. The difference is that we only wait on conversions, not all IO. This is a very big hammer, and I'm not very pleased with stuffing this into ext4_file_write(). But since ext4 is DIO_LOCKING, we need to serialize it at this high level. I tried to move this into ext4_ext_direct_IO, but by then we have the i_mutex already, and we will wait on the work queue to do conversions - which must also take the i_mutex. So that won't work. This was originally exposed by qemu-kvm installing to a raw disk image with a normal sector-63 alignment. I've tested a backport of this patch with qemu, and it does avoid the corruption. It is also quite a lot slower (14 min for package installs, vs. 8 min for well-aligned) but I'll take slow correctness over fast corruption any day. Mingming suggested that we can track outstanding conversions, and wait on those so that non-sparse files won't be affected, and I've implemented that here; unaligned AIO to nonsparse files won't take a perf hit. [tytso@mit.edu: Keep the mutex as a hashed array instead of bloating the ext4 inode] [tytso@mit.edu: Fix up namespace issues so that global variables are protected with an "ext4_" prefix.] Signed-off-by: Eric Sandeen <sandeen@redhat.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-02-12 13:17:34 +00:00
if (io && !(io->flag & EXT4_IO_END_UNWRITTEN)) {
io->flag = EXT4_IO_END_UNWRITTEN;
ext4: serialize unaligned asynchronous DIO ext4 has a data corruption case when doing non-block-aligned asynchronous direct IO into a sparse file, as demonstrated by xfstest 240. The root cause is that while ext4 preallocates space in the hole, mappings of that space still look "new" and dio_zero_block() will zero out the unwritten portions. When more than one AIO thread is going, they both find this "new" block and race to zero out their portion; this is uncoordinated and causes data corruption. Dave Chinner fixed this for xfs by simply serializing all unaligned asynchronous direct IO. I've done the same here. The difference is that we only wait on conversions, not all IO. This is a very big hammer, and I'm not very pleased with stuffing this into ext4_file_write(). But since ext4 is DIO_LOCKING, we need to serialize it at this high level. I tried to move this into ext4_ext_direct_IO, but by then we have the i_mutex already, and we will wait on the work queue to do conversions - which must also take the i_mutex. So that won't work. This was originally exposed by qemu-kvm installing to a raw disk image with a normal sector-63 alignment. I've tested a backport of this patch with qemu, and it does avoid the corruption. It is also quite a lot slower (14 min for package installs, vs. 8 min for well-aligned) but I'll take slow correctness over fast corruption any day. Mingming suggested that we can track outstanding conversions, and wait on those so that non-sparse files won't be affected, and I've implemented that here; unaligned AIO to nonsparse files won't take a perf hit. [tytso@mit.edu: Keep the mutex as a hashed array instead of bloating the ext4 inode] [tytso@mit.edu: Fix up namespace issues so that global variables are protected with an "ext4_" prefix.] Signed-off-by: Eric Sandeen <sandeen@redhat.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-02-12 13:17:34 +00:00
atomic_inc(&EXT4_I(inode)->i_aiodio_unwritten);
} else
ext4_set_inode_state(inode, EXT4_STATE_DIO_UNWRITTEN);
if (ext4_should_dioread_nolock(inode))
map->m_flags |= EXT4_MAP_UNINIT;
goto out;
}
/* IO end_io complete, convert the filled extent to written */
if ((flags & EXT4_GET_BLOCKS_CONVERT)) {
ret = ext4_convert_unwritten_extents_endio(handle, inode,
path);
if (ret >= 0) {
ext4_update_inode_fsync_trans(handle, inode, 1);
err = check_eofblocks_fl(handle, inode, map->m_lblk,
path, map->m_len);
} else
err = ret;
goto out2;
}
/* buffered IO case */
/*
* repeat fallocate creation request
* we already have an unwritten extent
*/
if (flags & EXT4_GET_BLOCKS_UNINIT_EXT)
goto map_out;
/* buffered READ or buffered write_begin() lookup */
if ((flags & EXT4_GET_BLOCKS_CREATE) == 0) {
/*
* We have blocks reserved already. We
* return allocated blocks so that delalloc
* won't do block reservation for us. But
* the buffer head will be unmapped so that
* a read from the block returns 0s.
*/
map->m_flags |= EXT4_MAP_UNWRITTEN;
goto out1;
}
/* buffered write, writepage time, convert*/
ret = ext4_ext_convert_to_initialized(handle, inode, map, path);
if (ret >= 0) {
ext4_update_inode_fsync_trans(handle, inode, 1);
err = check_eofblocks_fl(handle, inode, map->m_lblk, path,
map->m_len);
if (err < 0)
goto out2;
}
out:
if (ret <= 0) {
err = ret;
goto out2;
} else
allocated = ret;
map->m_flags |= EXT4_MAP_NEW;
/*
* if we allocated more blocks than requested
* we need to make sure we unmap the extra block
* allocated. The actual needed block will get
* unmapped later when we find the buffer_head marked
* new.
*/
if (allocated > map->m_len) {
unmap_underlying_metadata_blocks(inode->i_sb->s_bdev,
newblock + map->m_len,
allocated - map->m_len);
allocated = map->m_len;
}
/*
* If we have done fallocate with the offset that is already
* delayed allocated, we would have block reservation
* and quota reservation done in the delayed write path.
* But fallocate would have already updated quota and block
* count for this offset. So cancel these reservation
*/
if (flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE)
ext4_da_update_reserve_space(inode, allocated, 0);
map_out:
map->m_flags |= EXT4_MAP_MAPPED;
out1:
if (allocated > map->m_len)
allocated = map->m_len;
ext4_ext_show_leaf(inode, path);
map->m_pblk = newblock;
map->m_len = allocated;
out2:
if (path) {
ext4_ext_drop_refs(path);
kfree(path);
}
return err ? err : allocated;
}
/*
* Block allocation/map/preallocation routine for extents based files
*
*
* Need to be called with
* down_read(&EXT4_I(inode)->i_data_sem) if not allocating file system block
* (ie, create is zero). Otherwise down_write(&EXT4_I(inode)->i_data_sem)
*
* return > 0, number of of blocks already mapped/allocated
* if create == 0 and these are pre-allocated blocks
* buffer head is unmapped
* otherwise blocks are mapped
*
* return = 0, if plain look up failed (blocks have not been allocated)
* buffer head is unmapped
*
* return < 0, error case.
*/
int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
struct ext4_map_blocks *map, int flags)
{
struct ext4_ext_path *path = NULL;
struct ext4_extent newex, *ex;
ext4_fsblk_t newblock = 0;
int err = 0, depth, ret;
unsigned int allocated = 0;
unsigned int punched_out = 0;
unsigned int result = 0;
struct ext4_allocation_request ar;
ext4_io_end_t *io = EXT4_I(inode)->cur_aio_dio;
struct ext4_map_blocks punch_map;
ext_debug("blocks %u/%u requested for inode %lu\n",
map->m_lblk, map->m_len, inode->i_ino);
trace_ext4_ext_map_blocks_enter(inode, map->m_lblk, map->m_len, flags);
/* check in cache */
if (!(flags & EXT4_GET_BLOCKS_PUNCH_OUT_EXT) &&
ext4_ext_in_cache(inode, map->m_lblk, &newex)) {
if (!newex.ee_start_lo && !newex.ee_start_hi) {
if ((flags & EXT4_GET_BLOCKS_CREATE) == 0) {
/*
* block isn't allocated yet and
* user doesn't want to allocate it
*/
goto out2;
}
/* we should allocate requested block */
} else {
/* block is already allocated */
newblock = map->m_lblk
- le32_to_cpu(newex.ee_block)
+ ext4_ext_pblock(&newex);
/* number of remaining blocks in the extent */
allocated = ext4_ext_get_actual_len(&newex) -
(map->m_lblk - le32_to_cpu(newex.ee_block));
goto out;
}
}
/* find extent for this block */
path = ext4_ext_find_extent(inode, map->m_lblk, NULL);
if (IS_ERR(path)) {
err = PTR_ERR(path);
path = NULL;
goto out2;
}
depth = ext_depth(inode);
/*
* consistent leaf must not be empty;
* this situation is possible, though, _during_ tree modification;
* this is why assert can't be put in ext4_ext_find_extent()
*/
if (unlikely(path[depth].p_ext == NULL && depth != 0)) {
EXT4_ERROR_INODE(inode, "bad extent address "
"lblock: %lu, depth: %d pblock %lld",
(unsigned long) map->m_lblk, depth,
path[depth].p_block);
err = -EIO;
goto out2;
}
ex = path[depth].p_ext;
if (ex) {
ext4_lblk_t ee_block = le32_to_cpu(ex->ee_block);
ext4_fsblk_t ee_start = ext4_ext_pblock(ex);
unsigned short ee_len;
/*
* Uninitialized extents are treated as holes, except that
* we split out initialized portions during a write.
*/
ee_len = ext4_ext_get_actual_len(ex);
/* if found extent covers block, simply return it */
if (in_range(map->m_lblk, ee_block, ee_len)) {
newblock = map->m_lblk - ee_block + ee_start;
/* number of remaining blocks in the extent */
allocated = ee_len - (map->m_lblk - ee_block);
ext_debug("%u fit into %u:%d -> %llu\n", map->m_lblk,
ee_block, ee_len, newblock);
if ((flags & EXT4_GET_BLOCKS_PUNCH_OUT_EXT) == 0) {
/*
* Do not put uninitialized extent
* in the cache
*/
if (!ext4_ext_is_uninitialized(ex)) {
ext4_ext_put_in_cache(inode, ee_block,
ee_len, ee_start);
goto out;
}
ret = ext4_ext_handle_uninitialized_extents(
handle, inode, map, path, flags,
allocated, newblock);
return ret;
}
/*
* Punch out the map length, but only to the
* end of the extent
*/
punched_out = allocated < map->m_len ?
allocated : map->m_len;
/*
* Sense extents need to be converted to
* uninitialized, they must fit in an
* uninitialized extent
*/
if (punched_out > EXT_UNINIT_MAX_LEN)
punched_out = EXT_UNINIT_MAX_LEN;
punch_map.m_lblk = map->m_lblk;
punch_map.m_pblk = newblock;
punch_map.m_len = punched_out;
punch_map.m_flags = 0;
/* Check to see if the extent needs to be split */
if (punch_map.m_len != ee_len ||
punch_map.m_lblk != ee_block) {
ret = ext4_split_extent(handle, inode,
path, &punch_map, 0,
EXT4_GET_BLOCKS_PUNCH_OUT_EXT |
EXT4_GET_BLOCKS_PRE_IO);
if (ret < 0) {
err = ret;
goto out2;
}
/*
* find extent for the block at
* the start of the hole
*/
ext4_ext_drop_refs(path);
kfree(path);
path = ext4_ext_find_extent(inode,
map->m_lblk, NULL);
if (IS_ERR(path)) {
err = PTR_ERR(path);
path = NULL;
goto out2;
}
depth = ext_depth(inode);
ex = path[depth].p_ext;
ee_len = ext4_ext_get_actual_len(ex);
ee_block = le32_to_cpu(ex->ee_block);
ee_start = ext4_ext_pblock(ex);
}
ext4_ext_mark_uninitialized(ex);
ext4_ext_invalidate_cache(inode);
err = ext4_ext_rm_leaf(handle, inode, path,
map->m_lblk, map->m_lblk + punched_out);
if (!err && path->p_hdr->eh_entries == 0) {
/*
* Punch hole freed all of this sub tree,
* so we need to correct eh_depth
*/
err = ext4_ext_get_access(handle, inode, path);
if (err == 0) {
ext_inode_hdr(inode)->eh_depth = 0;
ext_inode_hdr(inode)->eh_max =
cpu_to_le16(ext4_ext_space_root(
inode, 0));
err = ext4_ext_dirty(
handle, inode, path);
}
}
goto out2;
}
}
/*
* requested block isn't allocated yet;
* we couldn't try to create block if create flag is zero
*/
if ((flags & EXT4_GET_BLOCKS_CREATE) == 0) {
/*
* put just found gap into cache to speed up
* subsequent requests
*/
ext4_ext_put_gap_in_cache(inode, path, map->m_lblk);
goto out2;
}
/*
* Okay, we need to do block allocation.
*/
/* find neighbour allocated blocks */
ar.lleft = map->m_lblk;
err = ext4_ext_search_left(inode, path, &ar.lleft, &ar.pleft);
if (err)
goto out2;
ar.lright = map->m_lblk;
err = ext4_ext_search_right(inode, path, &ar.lright, &ar.pright);
if (err)
goto out2;
/*
* See if request is beyond maximum number of blocks we can have in
* a single extent. For an initialized extent this limit is
* EXT_INIT_MAX_LEN and for an uninitialized extent this limit is
* EXT_UNINIT_MAX_LEN.
*/
if (map->m_len > EXT_INIT_MAX_LEN &&
!(flags & EXT4_GET_BLOCKS_UNINIT_EXT))
map->m_len = EXT_INIT_MAX_LEN;
else if (map->m_len > EXT_UNINIT_MAX_LEN &&
(flags & EXT4_GET_BLOCKS_UNINIT_EXT))
map->m_len = EXT_UNINIT_MAX_LEN;
/* Check if we can really insert (m_lblk)::(m_lblk + m_len) extent */
newex.ee_block = cpu_to_le32(map->m_lblk);
newex.ee_len = cpu_to_le16(map->m_len);
err = ext4_ext_check_overlap(inode, &newex, path);
if (err)
allocated = ext4_ext_get_actual_len(&newex);
else
allocated = map->m_len;
/* allocate new block */
ar.inode = inode;
ar.goal = ext4_ext_find_goal(inode, path, map->m_lblk);
ar.logical = map->m_lblk;
ar.len = allocated;
if (S_ISREG(inode->i_mode))
ar.flags = EXT4_MB_HINT_DATA;
else
/* disable in-core preallocation for non-regular files */
ar.flags = 0;
if (flags & EXT4_GET_BLOCKS_NO_NORMALIZE)
ar.flags |= EXT4_MB_HINT_NOPREALLOC;
newblock = ext4_mb_new_blocks(handle, &ar, &err);
if (!newblock)
goto out2;
ext_debug("allocate new block: goal %llu, found %llu/%u\n",
ar.goal, newblock, allocated);
/* try to insert new extent into found leaf and return */
ext4_ext_store_pblock(&newex, newblock);
newex.ee_len = cpu_to_le16(ar.len);
/* Mark uninitialized */
if (flags & EXT4_GET_BLOCKS_UNINIT_EXT){
ext4_ext_mark_uninitialized(&newex);
/*
* io_end structure was created for every IO write to an
* uninitialized extent. To avoid unnecessary conversion,
* here we flag the IO that really needs the conversion.
* For non asycn direct IO case, flag the inode state
* that we need to perform conversion when IO is done.
*/
if ((flags & EXT4_GET_BLOCKS_PRE_IO)) {
ext4: serialize unaligned asynchronous DIO ext4 has a data corruption case when doing non-block-aligned asynchronous direct IO into a sparse file, as demonstrated by xfstest 240. The root cause is that while ext4 preallocates space in the hole, mappings of that space still look "new" and dio_zero_block() will zero out the unwritten portions. When more than one AIO thread is going, they both find this "new" block and race to zero out their portion; this is uncoordinated and causes data corruption. Dave Chinner fixed this for xfs by simply serializing all unaligned asynchronous direct IO. I've done the same here. The difference is that we only wait on conversions, not all IO. This is a very big hammer, and I'm not very pleased with stuffing this into ext4_file_write(). But since ext4 is DIO_LOCKING, we need to serialize it at this high level. I tried to move this into ext4_ext_direct_IO, but by then we have the i_mutex already, and we will wait on the work queue to do conversions - which must also take the i_mutex. So that won't work. This was originally exposed by qemu-kvm installing to a raw disk image with a normal sector-63 alignment. I've tested a backport of this patch with qemu, and it does avoid the corruption. It is also quite a lot slower (14 min for package installs, vs. 8 min for well-aligned) but I'll take slow correctness over fast corruption any day. Mingming suggested that we can track outstanding conversions, and wait on those so that non-sparse files won't be affected, and I've implemented that here; unaligned AIO to nonsparse files won't take a perf hit. [tytso@mit.edu: Keep the mutex as a hashed array instead of bloating the ext4 inode] [tytso@mit.edu: Fix up namespace issues so that global variables are protected with an "ext4_" prefix.] Signed-off-by: Eric Sandeen <sandeen@redhat.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-02-12 13:17:34 +00:00
if (io && !(io->flag & EXT4_IO_END_UNWRITTEN)) {
io->flag = EXT4_IO_END_UNWRITTEN;
ext4: serialize unaligned asynchronous DIO ext4 has a data corruption case when doing non-block-aligned asynchronous direct IO into a sparse file, as demonstrated by xfstest 240. The root cause is that while ext4 preallocates space in the hole, mappings of that space still look "new" and dio_zero_block() will zero out the unwritten portions. When more than one AIO thread is going, they both find this "new" block and race to zero out their portion; this is uncoordinated and causes data corruption. Dave Chinner fixed this for xfs by simply serializing all unaligned asynchronous direct IO. I've done the same here. The difference is that we only wait on conversions, not all IO. This is a very big hammer, and I'm not very pleased with stuffing this into ext4_file_write(). But since ext4 is DIO_LOCKING, we need to serialize it at this high level. I tried to move this into ext4_ext_direct_IO, but by then we have the i_mutex already, and we will wait on the work queue to do conversions - which must also take the i_mutex. So that won't work. This was originally exposed by qemu-kvm installing to a raw disk image with a normal sector-63 alignment. I've tested a backport of this patch with qemu, and it does avoid the corruption. It is also quite a lot slower (14 min for package installs, vs. 8 min for well-aligned) but I'll take slow correctness over fast corruption any day. Mingming suggested that we can track outstanding conversions, and wait on those so that non-sparse files won't be affected, and I've implemented that here; unaligned AIO to nonsparse files won't take a perf hit. [tytso@mit.edu: Keep the mutex as a hashed array instead of bloating the ext4 inode] [tytso@mit.edu: Fix up namespace issues so that global variables are protected with an "ext4_" prefix.] Signed-off-by: Eric Sandeen <sandeen@redhat.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-02-12 13:17:34 +00:00
atomic_inc(&EXT4_I(inode)->i_aiodio_unwritten);
} else
ext4_set_inode_state(inode,
EXT4_STATE_DIO_UNWRITTEN);
}
if (ext4_should_dioread_nolock(inode))
map->m_flags |= EXT4_MAP_UNINIT;
}
err = check_eofblocks_fl(handle, inode, map->m_lblk, path, ar.len);
if (!err)
err = ext4_ext_insert_extent(handle, inode, path,
&newex, flags);
if (err) {
int fb_flags = flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE ?
EXT4_FREE_BLOCKS_NO_QUOT_UPDATE : 0;
/* free data blocks we just allocated */
/* not a good idea to call discard here directly,
* but otherwise we'd need to call it every free() */
ext4_discard_preallocations(inode);
ext4_free_blocks(handle, inode, NULL, ext4_ext_pblock(&newex),
ext4_ext_get_actual_len(&newex), fb_flags);
goto out2;
}
/* previous routine could use block we allocated */
newblock = ext4_ext_pblock(&newex);
allocated = ext4_ext_get_actual_len(&newex);
if (allocated > map->m_len)
allocated = map->m_len;
map->m_flags |= EXT4_MAP_NEW;
/*
* Update reserved blocks/metadata blocks after successful
* block allocation which had been deferred till now.
*/
if (flags & EXT4_GET_BLOCKS_DELALLOC_RESERVE)
ext4_da_update_reserve_space(inode, allocated, 1);
/*
* Cache the extent and update transaction to commit on fdatasync only
* when it is _not_ an uninitialized extent.
*/
if ((flags & EXT4_GET_BLOCKS_UNINIT_EXT) == 0) {
ext4_ext_put_in_cache(inode, map->m_lblk, allocated, newblock);
ext4_update_inode_fsync_trans(handle, inode, 1);
} else
ext4_update_inode_fsync_trans(handle, inode, 0);
out:
if (allocated > map->m_len)
allocated = map->m_len;
ext4_ext_show_leaf(inode, path);
map->m_flags |= EXT4_MAP_MAPPED;
map->m_pblk = newblock;
map->m_len = allocated;
out2:
if (path) {
ext4_ext_drop_refs(path);
kfree(path);
}
trace_ext4_ext_map_blocks_exit(inode, map->m_lblk,
newblock, map->m_len, err ? err : allocated);
result = (flags & EXT4_GET_BLOCKS_PUNCH_OUT_EXT) ?
punched_out : allocated;
return err ? err : result;
}
void ext4_ext_truncate(struct inode *inode)
{
struct address_space *mapping = inode->i_mapping;
struct super_block *sb = inode->i_sb;
ext4_lblk_t last_block;
handle_t *handle;
int err = 0;
ext4: flush the i_completed_io_list during ext4_truncate Ted first found the bug when running 2.6.36 kernel with dioread_nolock mount option that xfstests #13 complained about wrong file size during fsck. However, the bug exists in the older kernels as well although it is somehow harder to trigger. The problem is that ext4_end_io_work() can happen after we have truncated an inode to a smaller size. Then when ext4_end_io_work() calls ext4_convert_unwritten_extents(), we may reallocate some blocks that have been truncated, so the inode size becomes inconsistent with the allocated blocks. The following patch flushes the i_completed_io_list during truncate to reduce the risk that some pending end_io requests are executed later and convert already truncated blocks to initialized. Note that although the fix helps reduce the problem a lot there may still be a race window between vmtruncate() and ext4_end_io_work(). The fundamental problem is that if vmtruncate() is called without either i_mutex or i_alloc_sem held, it can race with an ongoing write request so that the io_end request is processed later when the corresponding blocks have been truncated. Ted and I have discussed the problem offline and we saw a few ways to fix the race completely: a) We guarantee that i_mutex lock and i_alloc_sem write lock are both hold whenever vmtruncate() is called. The i_mutex lock prevents any new write requests from entering writeback and the i_alloc_sem prevents the race from ext4_page_mkwrite(). Currently we hold both locks if vmtruncate() is called from do_truncate(), which is probably the most common case. However, there are places where we may call vmtruncate() without holding either i_mutex or i_alloc_sem. I would like to ask for other people's opinions on what locks are expected to be held before calling vmtruncate(). There seems a disagreement among the callers of that function. b) We change the ext4 write path so that we change the extent tree to contain the newly allocated blocks and update i_size both at the same time --- when the write of the data blocks is completed. c) We add some additional locking to synchronize vmtruncate() and ext4_end_io_work(). This approach may have performance implications so we need to be careful. All of the above proposals may require more substantial changes, so we may consider to take the following patch as a bandaid. Signed-off-by: Jiaying Zhang <jiayingz@google.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-01-10 17:47:05 +00:00
/*
* finish any pending end_io work so we won't run the risk of
* converting any truncated blocks to initialized later
*/
ext4_flush_completed_IO(inode);
/*
* probably first extent we're gonna free will be last in block
*/
err = ext4_writepage_trans_blocks(inode);
handle = ext4_journal_start(inode, err);
if (IS_ERR(handle))
return;
if (inode->i_size & (sb->s_blocksize - 1))
ext4_block_truncate_page(handle, mapping, inode->i_size);
if (ext4_orphan_add(handle, inode))
goto out_stop;
down_write(&EXT4_I(inode)->i_data_sem);
ext4_ext_invalidate_cache(inode);
ext4_discard_preallocations(inode);
/*
* TODO: optimization is possible here.
* Probably we need not scan at all,
* because page truncation is enough.
*/
/* we have to know where to truncate from in crash case */
EXT4_I(inode)->i_disksize = inode->i_size;
ext4_mark_inode_dirty(handle, inode);
last_block = (inode->i_size + sb->s_blocksize - 1)
>> EXT4_BLOCK_SIZE_BITS(sb);
err = ext4_ext_remove_space(inode, last_block);
/* In a multi-transaction truncate, we only make the final
* transaction synchronous.
*/
if (IS_SYNC(inode))
ext4_handle_sync(handle);
up_write(&EXT4_I(inode)->i_data_sem);
out_stop:
/*
* If this was a simple ftruncate() and the file will remain alive,
* then we need to clear up the orphan record which we created above.
* However, if this was a real unlink then we were called by
* ext4_delete_inode(), and we allow that function to clean up the
* orphan info for us.
*/
if (inode->i_nlink)
ext4_orphan_del(handle, inode);
inode->i_mtime = inode->i_ctime = ext4_current_time(inode);
ext4_mark_inode_dirty(handle, inode);
ext4_journal_stop(handle);
}
static void ext4_falloc_update_inode(struct inode *inode,
int mode, loff_t new_size, int update_ctime)
{
struct timespec now;
if (update_ctime) {
now = current_fs_time(inode->i_sb);
if (!timespec_equal(&inode->i_ctime, &now))
inode->i_ctime = now;
}
/*
* Update only when preallocation was requested beyond
* the file size.
*/
if (!(mode & FALLOC_FL_KEEP_SIZE)) {
if (new_size > i_size_read(inode))
i_size_write(inode, new_size);
if (new_size > EXT4_I(inode)->i_disksize)
ext4_update_i_disksize(inode, new_size);
} else {
/*
* Mark that we allocate beyond EOF so the subsequent truncate
* can proceed even if the new size is the same as i_size.
*/
if (new_size > i_size_read(inode))
ext4_set_inode_flag(inode, EXT4_INODE_EOFBLOCKS);
}
}
/*
* preallocate space for a file. This implements ext4's fallocate file
* operation, which gets called from sys_fallocate system call.
* For block-mapped files, posix_fallocate should fall back to the method
* of writing zeroes to the required new blocks (the same behavior which is
* expected for file systems which do not support fallocate() system call).
*/
long ext4_fallocate(struct file *file, int mode, loff_t offset, loff_t len)
{
struct inode *inode = file->f_path.dentry->d_inode;
handle_t *handle;
loff_t new_size;
unsigned int max_blocks;
int ret = 0;
int ret2 = 0;
int retries = 0;
struct ext4_map_blocks map;
unsigned int credits, blkbits = inode->i_blkbits;
/*
* currently supporting (pre)allocate mode for extent-based
* files _only_
*/
if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
return -EOPNOTSUPP;
/* Return error if mode is not supported */
if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
return -EOPNOTSUPP;
if (mode & FALLOC_FL_PUNCH_HOLE)
return ext4_punch_hole(file, offset, len);
trace_ext4_fallocate_enter(inode, offset, len, mode);
map.m_lblk = offset >> blkbits;
/*
* We can't just convert len to max_blocks because
* If blocksize = 4096 offset = 3072 and len = 2048
*/
max_blocks = (EXT4_BLOCK_ALIGN(len + offset, blkbits) >> blkbits)
- map.m_lblk;
/*
* credits to insert 1 extent into extent tree
*/
credits = ext4_chunk_trans_blocks(inode, max_blocks);
mutex_lock(&inode->i_mutex);
ret = inode_newsize_ok(inode, (len + offset));
if (ret) {
mutex_unlock(&inode->i_mutex);
trace_ext4_fallocate_exit(inode, offset, max_blocks, ret);
return ret;
}
retry:
while (ret >= 0 && ret < max_blocks) {
map.m_lblk = map.m_lblk + ret;
map.m_len = max_blocks = max_blocks - ret;
handle = ext4_journal_start(inode, credits);
if (IS_ERR(handle)) {
ret = PTR_ERR(handle);
break;
}
ret = ext4_map_blocks(handle, inode, &map,
EXT4_GET_BLOCKS_CREATE_UNINIT_EXT |
EXT4_GET_BLOCKS_NO_NORMALIZE);
if (ret <= 0) {
#ifdef EXT4FS_DEBUG
WARN_ON(ret <= 0);
printk(KERN_ERR "%s: ext4_ext_map_blocks "
"returned error inode#%lu, block=%u, "
"max_blocks=%u", __func__,
inode->i_ino, map.m_lblk, max_blocks);
#endif
ext4_mark_inode_dirty(handle, inode);
ret2 = ext4_journal_stop(handle);
break;
}
if ((map.m_lblk + ret) >= (EXT4_BLOCK_ALIGN(offset + len,
blkbits) >> blkbits))
new_size = offset + len;
else
new_size = ((loff_t) map.m_lblk + ret) << blkbits;
ext4_falloc_update_inode(inode, mode, new_size,
(map.m_flags & EXT4_MAP_NEW));
ext4_mark_inode_dirty(handle, inode);
ret2 = ext4_journal_stop(handle);
if (ret2)
break;
}
if (ret == -ENOSPC &&
ext4_should_retry_alloc(inode->i_sb, &retries)) {
ret = 0;
goto retry;
}
mutex_unlock(&inode->i_mutex);
trace_ext4_fallocate_exit(inode, offset, max_blocks,
ret > 0 ? ret2 : ret);
return ret > 0 ? ret2 : ret;
}
/*
* This function convert a range of blocks to written extents
* The caller of this function will pass the start offset and the size.
* all unwritten extents within this range will be converted to
* written extents.
*
* This function is called from the direct IO end io call back
* function, to convert the fallocated extents after IO is completed.
* Returns 0 on success.
*/
int ext4_convert_unwritten_extents(struct inode *inode, loff_t offset,
ssize_t len)
{
handle_t *handle;
unsigned int max_blocks;
int ret = 0;
int ret2 = 0;
struct ext4_map_blocks map;
unsigned int credits, blkbits = inode->i_blkbits;
map.m_lblk = offset >> blkbits;
/*
* We can't just convert len to max_blocks because
* If blocksize = 4096 offset = 3072 and len = 2048
*/
max_blocks = ((EXT4_BLOCK_ALIGN(len + offset, blkbits) >> blkbits) -
map.m_lblk);
/*
* credits to insert 1 extent into extent tree
*/
credits = ext4_chunk_trans_blocks(inode, max_blocks);
while (ret >= 0 && ret < max_blocks) {
map.m_lblk += ret;
map.m_len = (max_blocks -= ret);
handle = ext4_journal_start(inode, credits);
if (IS_ERR(handle)) {
ret = PTR_ERR(handle);
break;
}
ret = ext4_map_blocks(handle, inode, &map,
EXT4_GET_BLOCKS_IO_CONVERT_EXT);
if (ret <= 0) {
WARN_ON(ret <= 0);
printk(KERN_ERR "%s: ext4_ext_map_blocks "
"returned error inode#%lu, block=%u, "
"max_blocks=%u", __func__,
inode->i_ino, map.m_lblk, map.m_len);
}
ext4_mark_inode_dirty(handle, inode);
ret2 = ext4_journal_stop(handle);
if (ret <= 0 || ret2 )
break;
}
return ret > 0 ? ret2 : ret;
}
ext4: make FIEMAP and delayed allocation play well together Fix the FIEMAP ioctl so that it returns all of the page ranges which are still subject to delayed allocation. We were missing some cases if the file was sparse. Reported by Chris Mason <chris.mason@oracle.com>: >We've had reports on btrfs that cp is giving us files full of zeros >instead of actually copying them. It was tracked down to a bug with >the btrfs fiemap implementation where it was returning holes for >delalloc ranges. > >Newer versions of cp are trusting fiemap to tell it where the holes >are, which does seem like a pretty neat trick. > >I decided to give xfs and ext4 a shot with a few tests cases too, xfs >passed with all the ones btrfs was getting wrong, and ext4 got the basic >delalloc case right. >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 >$ fiemap-test foo >ext: 0 logical: [ 0.. 255] phys: 0.. 255 >flags: 0x007 tot: 256 > >Horray! But once we throw a hole in, things go bad: >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 >$ fiemap-test foo >< no output > > >We've got a delalloc extent after the hole and ext4 fiemap didn't find >it. If I run sync to kick the delalloc out: >$sync >$ fiemap-test foo >ext: 0 logical: [ 256.. 511] phys: 34048.. 34303 >flags: 0x001 tot: 256 > >fiemap-test is sitting in my /usr/local/bin, and I have no idea how it >got there. It's full of pretty comments so I know it isn't mine, but >you can grab it here: > >http://oss.oracle.com/~mason/fiemap-test.c > >xfsqa has a fiemap program too. After Fix, test results are as follows: ext: 0 logical: [ 256.. 511] phys: 0.. 255 flags: 0x007 tot: 256 ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x001 tot: 256 $ mkfs.ext4 /dev/xxx $ mount /dev/xxx /mnt $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 $ sync $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=3 $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=5 $ fiemap-test foo ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x000 tot: 256 ext: 1 logical: [ 768.. 1023] phys: 0.. 255 flags: 0x006 tot: 256 ext: 2 logical: [ 1280.. 1535] phys: 0.. 255 flags: 0x007 tot: 256 Tested-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Andreas Dilger <adilger@dilger.ca> Signed-off-by: Yongqiang Yang <xiaoqiangnk@gmail.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-02-27 22:25:47 +00:00
/*
* Callback function called for each extent to gather FIEMAP information.
*/
static int ext4_ext_fiemap_cb(struct inode *inode, ext4_lblk_t next,
struct ext4_ext_cache *newex, struct ext4_extent *ex,
void *data)
{
__u64 logical;
__u64 physical;
__u64 length;
__u32 flags = 0;
ext4: make FIEMAP and delayed allocation play well together Fix the FIEMAP ioctl so that it returns all of the page ranges which are still subject to delayed allocation. We were missing some cases if the file was sparse. Reported by Chris Mason <chris.mason@oracle.com>: >We've had reports on btrfs that cp is giving us files full of zeros >instead of actually copying them. It was tracked down to a bug with >the btrfs fiemap implementation where it was returning holes for >delalloc ranges. > >Newer versions of cp are trusting fiemap to tell it where the holes >are, which does seem like a pretty neat trick. > >I decided to give xfs and ext4 a shot with a few tests cases too, xfs >passed with all the ones btrfs was getting wrong, and ext4 got the basic >delalloc case right. >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 >$ fiemap-test foo >ext: 0 logical: [ 0.. 255] phys: 0.. 255 >flags: 0x007 tot: 256 > >Horray! But once we throw a hole in, things go bad: >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 >$ fiemap-test foo >< no output > > >We've got a delalloc extent after the hole and ext4 fiemap didn't find >it. If I run sync to kick the delalloc out: >$sync >$ fiemap-test foo >ext: 0 logical: [ 256.. 511] phys: 34048.. 34303 >flags: 0x001 tot: 256 > >fiemap-test is sitting in my /usr/local/bin, and I have no idea how it >got there. It's full of pretty comments so I know it isn't mine, but >you can grab it here: > >http://oss.oracle.com/~mason/fiemap-test.c > >xfsqa has a fiemap program too. After Fix, test results are as follows: ext: 0 logical: [ 256.. 511] phys: 0.. 255 flags: 0x007 tot: 256 ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x001 tot: 256 $ mkfs.ext4 /dev/xxx $ mount /dev/xxx /mnt $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 $ sync $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=3 $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=5 $ fiemap-test foo ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x000 tot: 256 ext: 1 logical: [ 768.. 1023] phys: 0.. 255 flags: 0x006 tot: 256 ext: 2 logical: [ 1280.. 1535] phys: 0.. 255 flags: 0x007 tot: 256 Tested-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Andreas Dilger <adilger@dilger.ca> Signed-off-by: Yongqiang Yang <xiaoqiangnk@gmail.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-02-27 22:25:47 +00:00
int ret = 0;
struct fiemap_extent_info *fieinfo = data;
unsigned char blksize_bits;
ext4: make FIEMAP and delayed allocation play well together Fix the FIEMAP ioctl so that it returns all of the page ranges which are still subject to delayed allocation. We were missing some cases if the file was sparse. Reported by Chris Mason <chris.mason@oracle.com>: >We've had reports on btrfs that cp is giving us files full of zeros >instead of actually copying them. It was tracked down to a bug with >the btrfs fiemap implementation where it was returning holes for >delalloc ranges. > >Newer versions of cp are trusting fiemap to tell it where the holes >are, which does seem like a pretty neat trick. > >I decided to give xfs and ext4 a shot with a few tests cases too, xfs >passed with all the ones btrfs was getting wrong, and ext4 got the basic >delalloc case right. >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 >$ fiemap-test foo >ext: 0 logical: [ 0.. 255] phys: 0.. 255 >flags: 0x007 tot: 256 > >Horray! But once we throw a hole in, things go bad: >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 >$ fiemap-test foo >< no output > > >We've got a delalloc extent after the hole and ext4 fiemap didn't find >it. If I run sync to kick the delalloc out: >$sync >$ fiemap-test foo >ext: 0 logical: [ 256.. 511] phys: 34048.. 34303 >flags: 0x001 tot: 256 > >fiemap-test is sitting in my /usr/local/bin, and I have no idea how it >got there. It's full of pretty comments so I know it isn't mine, but >you can grab it here: > >http://oss.oracle.com/~mason/fiemap-test.c > >xfsqa has a fiemap program too. After Fix, test results are as follows: ext: 0 logical: [ 256.. 511] phys: 0.. 255 flags: 0x007 tot: 256 ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x001 tot: 256 $ mkfs.ext4 /dev/xxx $ mount /dev/xxx /mnt $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 $ sync $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=3 $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=5 $ fiemap-test foo ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x000 tot: 256 ext: 1 logical: [ 768.. 1023] phys: 0.. 255 flags: 0x006 tot: 256 ext: 2 logical: [ 1280.. 1535] phys: 0.. 255 flags: 0x007 tot: 256 Tested-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Andreas Dilger <adilger@dilger.ca> Signed-off-by: Yongqiang Yang <xiaoqiangnk@gmail.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-02-27 22:25:47 +00:00
blksize_bits = inode->i_sb->s_blocksize_bits;
logical = (__u64)newex->ec_block << blksize_bits;
if (newex->ec_start == 0) {
ext4: make FIEMAP and delayed allocation play well together Fix the FIEMAP ioctl so that it returns all of the page ranges which are still subject to delayed allocation. We were missing some cases if the file was sparse. Reported by Chris Mason <chris.mason@oracle.com>: >We've had reports on btrfs that cp is giving us files full of zeros >instead of actually copying them. It was tracked down to a bug with >the btrfs fiemap implementation where it was returning holes for >delalloc ranges. > >Newer versions of cp are trusting fiemap to tell it where the holes >are, which does seem like a pretty neat trick. > >I decided to give xfs and ext4 a shot with a few tests cases too, xfs >passed with all the ones btrfs was getting wrong, and ext4 got the basic >delalloc case right. >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 >$ fiemap-test foo >ext: 0 logical: [ 0.. 255] phys: 0.. 255 >flags: 0x007 tot: 256 > >Horray! But once we throw a hole in, things go bad: >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 >$ fiemap-test foo >< no output > > >We've got a delalloc extent after the hole and ext4 fiemap didn't find >it. If I run sync to kick the delalloc out: >$sync >$ fiemap-test foo >ext: 0 logical: [ 256.. 511] phys: 34048.. 34303 >flags: 0x001 tot: 256 > >fiemap-test is sitting in my /usr/local/bin, and I have no idea how it >got there. It's full of pretty comments so I know it isn't mine, but >you can grab it here: > >http://oss.oracle.com/~mason/fiemap-test.c > >xfsqa has a fiemap program too. After Fix, test results are as follows: ext: 0 logical: [ 256.. 511] phys: 0.. 255 flags: 0x007 tot: 256 ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x001 tot: 256 $ mkfs.ext4 /dev/xxx $ mount /dev/xxx /mnt $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 $ sync $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=3 $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=5 $ fiemap-test foo ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x000 tot: 256 ext: 1 logical: [ 768.. 1023] phys: 0.. 255 flags: 0x006 tot: 256 ext: 2 logical: [ 1280.. 1535] phys: 0.. 255 flags: 0x007 tot: 256 Tested-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Andreas Dilger <adilger@dilger.ca> Signed-off-by: Yongqiang Yang <xiaoqiangnk@gmail.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-02-27 22:25:47 +00:00
/*
* No extent in extent-tree contains block @newex->ec_start,
* then the block may stay in 1)a hole or 2)delayed-extent.
*
* Holes or delayed-extents are processed as follows.
* 1. lookup dirty pages with specified range in pagecache.
* If no page is got, then there is no delayed-extent and
* return with EXT_CONTINUE.
* 2. find the 1st mapped buffer,
* 3. check if the mapped buffer is both in the request range
* and a delayed buffer. If not, there is no delayed-extent,
* then return.
* 4. a delayed-extent is found, the extent will be collected.
*/
ext4_lblk_t end = 0;
pgoff_t last_offset;
pgoff_t offset;
pgoff_t index;
pgoff_t start_index = 0;
ext4: make FIEMAP and delayed allocation play well together Fix the FIEMAP ioctl so that it returns all of the page ranges which are still subject to delayed allocation. We were missing some cases if the file was sparse. Reported by Chris Mason <chris.mason@oracle.com>: >We've had reports on btrfs that cp is giving us files full of zeros >instead of actually copying them. It was tracked down to a bug with >the btrfs fiemap implementation where it was returning holes for >delalloc ranges. > >Newer versions of cp are trusting fiemap to tell it where the holes >are, which does seem like a pretty neat trick. > >I decided to give xfs and ext4 a shot with a few tests cases too, xfs >passed with all the ones btrfs was getting wrong, and ext4 got the basic >delalloc case right. >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 >$ fiemap-test foo >ext: 0 logical: [ 0.. 255] phys: 0.. 255 >flags: 0x007 tot: 256 > >Horray! But once we throw a hole in, things go bad: >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 >$ fiemap-test foo >< no output > > >We've got a delalloc extent after the hole and ext4 fiemap didn't find >it. If I run sync to kick the delalloc out: >$sync >$ fiemap-test foo >ext: 0 logical: [ 256.. 511] phys: 34048.. 34303 >flags: 0x001 tot: 256 > >fiemap-test is sitting in my /usr/local/bin, and I have no idea how it >got there. It's full of pretty comments so I know it isn't mine, but >you can grab it here: > >http://oss.oracle.com/~mason/fiemap-test.c > >xfsqa has a fiemap program too. After Fix, test results are as follows: ext: 0 logical: [ 256.. 511] phys: 0.. 255 flags: 0x007 tot: 256 ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x001 tot: 256 $ mkfs.ext4 /dev/xxx $ mount /dev/xxx /mnt $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 $ sync $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=3 $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=5 $ fiemap-test foo ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x000 tot: 256 ext: 1 logical: [ 768.. 1023] phys: 0.. 255 flags: 0x006 tot: 256 ext: 2 logical: [ 1280.. 1535] phys: 0.. 255 flags: 0x007 tot: 256 Tested-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Andreas Dilger <adilger@dilger.ca> Signed-off-by: Yongqiang Yang <xiaoqiangnk@gmail.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-02-27 22:25:47 +00:00
struct page **pages = NULL;
struct buffer_head *bh = NULL;
ext4: make FIEMAP and delayed allocation play well together Fix the FIEMAP ioctl so that it returns all of the page ranges which are still subject to delayed allocation. We were missing some cases if the file was sparse. Reported by Chris Mason <chris.mason@oracle.com>: >We've had reports on btrfs that cp is giving us files full of zeros >instead of actually copying them. It was tracked down to a bug with >the btrfs fiemap implementation where it was returning holes for >delalloc ranges. > >Newer versions of cp are trusting fiemap to tell it where the holes >are, which does seem like a pretty neat trick. > >I decided to give xfs and ext4 a shot with a few tests cases too, xfs >passed with all the ones btrfs was getting wrong, and ext4 got the basic >delalloc case right. >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 >$ fiemap-test foo >ext: 0 logical: [ 0.. 255] phys: 0.. 255 >flags: 0x007 tot: 256 > >Horray! But once we throw a hole in, things go bad: >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 >$ fiemap-test foo >< no output > > >We've got a delalloc extent after the hole and ext4 fiemap didn't find >it. If I run sync to kick the delalloc out: >$sync >$ fiemap-test foo >ext: 0 logical: [ 256.. 511] phys: 34048.. 34303 >flags: 0x001 tot: 256 > >fiemap-test is sitting in my /usr/local/bin, and I have no idea how it >got there. It's full of pretty comments so I know it isn't mine, but >you can grab it here: > >http://oss.oracle.com/~mason/fiemap-test.c > >xfsqa has a fiemap program too. After Fix, test results are as follows: ext: 0 logical: [ 256.. 511] phys: 0.. 255 flags: 0x007 tot: 256 ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x001 tot: 256 $ mkfs.ext4 /dev/xxx $ mount /dev/xxx /mnt $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 $ sync $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=3 $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=5 $ fiemap-test foo ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x000 tot: 256 ext: 1 logical: [ 768.. 1023] phys: 0.. 255 flags: 0x006 tot: 256 ext: 2 logical: [ 1280.. 1535] phys: 0.. 255 flags: 0x007 tot: 256 Tested-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Andreas Dilger <adilger@dilger.ca> Signed-off-by: Yongqiang Yang <xiaoqiangnk@gmail.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-02-27 22:25:47 +00:00
struct buffer_head *head = NULL;
unsigned int nr_pages = PAGE_SIZE / sizeof(struct page *);
pages = kmalloc(PAGE_SIZE, GFP_KERNEL);
if (pages == NULL)
return -ENOMEM;
offset = logical >> PAGE_SHIFT;
ext4: make FIEMAP and delayed allocation play well together Fix the FIEMAP ioctl so that it returns all of the page ranges which are still subject to delayed allocation. We were missing some cases if the file was sparse. Reported by Chris Mason <chris.mason@oracle.com>: >We've had reports on btrfs that cp is giving us files full of zeros >instead of actually copying them. It was tracked down to a bug with >the btrfs fiemap implementation where it was returning holes for >delalloc ranges. > >Newer versions of cp are trusting fiemap to tell it where the holes >are, which does seem like a pretty neat trick. > >I decided to give xfs and ext4 a shot with a few tests cases too, xfs >passed with all the ones btrfs was getting wrong, and ext4 got the basic >delalloc case right. >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 >$ fiemap-test foo >ext: 0 logical: [ 0.. 255] phys: 0.. 255 >flags: 0x007 tot: 256 > >Horray! But once we throw a hole in, things go bad: >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 >$ fiemap-test foo >< no output > > >We've got a delalloc extent after the hole and ext4 fiemap didn't find >it. If I run sync to kick the delalloc out: >$sync >$ fiemap-test foo >ext: 0 logical: [ 256.. 511] phys: 34048.. 34303 >flags: 0x001 tot: 256 > >fiemap-test is sitting in my /usr/local/bin, and I have no idea how it >got there. It's full of pretty comments so I know it isn't mine, but >you can grab it here: > >http://oss.oracle.com/~mason/fiemap-test.c > >xfsqa has a fiemap program too. After Fix, test results are as follows: ext: 0 logical: [ 256.. 511] phys: 0.. 255 flags: 0x007 tot: 256 ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x001 tot: 256 $ mkfs.ext4 /dev/xxx $ mount /dev/xxx /mnt $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 $ sync $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=3 $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=5 $ fiemap-test foo ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x000 tot: 256 ext: 1 logical: [ 768.. 1023] phys: 0.. 255 flags: 0x006 tot: 256 ext: 2 logical: [ 1280.. 1535] phys: 0.. 255 flags: 0x007 tot: 256 Tested-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Andreas Dilger <adilger@dilger.ca> Signed-off-by: Yongqiang Yang <xiaoqiangnk@gmail.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-02-27 22:25:47 +00:00
repeat:
last_offset = offset;
head = NULL;
ret = find_get_pages_tag(inode->i_mapping, &offset,
PAGECACHE_TAG_DIRTY, nr_pages, pages);
if (!(flags & FIEMAP_EXTENT_DELALLOC)) {
/* First time, try to find a mapped buffer. */
if (ret == 0) {
out:
for (index = 0; index < ret; index++)
page_cache_release(pages[index]);
/* just a hole. */
kfree(pages);
return EXT_CONTINUE;
}
index = 0;
next_page:
ext4: make FIEMAP and delayed allocation play well together Fix the FIEMAP ioctl so that it returns all of the page ranges which are still subject to delayed allocation. We were missing some cases if the file was sparse. Reported by Chris Mason <chris.mason@oracle.com>: >We've had reports on btrfs that cp is giving us files full of zeros >instead of actually copying them. It was tracked down to a bug with >the btrfs fiemap implementation where it was returning holes for >delalloc ranges. > >Newer versions of cp are trusting fiemap to tell it where the holes >are, which does seem like a pretty neat trick. > >I decided to give xfs and ext4 a shot with a few tests cases too, xfs >passed with all the ones btrfs was getting wrong, and ext4 got the basic >delalloc case right. >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 >$ fiemap-test foo >ext: 0 logical: [ 0.. 255] phys: 0.. 255 >flags: 0x007 tot: 256 > >Horray! But once we throw a hole in, things go bad: >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 >$ fiemap-test foo >< no output > > >We've got a delalloc extent after the hole and ext4 fiemap didn't find >it. If I run sync to kick the delalloc out: >$sync >$ fiemap-test foo >ext: 0 logical: [ 256.. 511] phys: 34048.. 34303 >flags: 0x001 tot: 256 > >fiemap-test is sitting in my /usr/local/bin, and I have no idea how it >got there. It's full of pretty comments so I know it isn't mine, but >you can grab it here: > >http://oss.oracle.com/~mason/fiemap-test.c > >xfsqa has a fiemap program too. After Fix, test results are as follows: ext: 0 logical: [ 256.. 511] phys: 0.. 255 flags: 0x007 tot: 256 ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x001 tot: 256 $ mkfs.ext4 /dev/xxx $ mount /dev/xxx /mnt $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 $ sync $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=3 $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=5 $ fiemap-test foo ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x000 tot: 256 ext: 1 logical: [ 768.. 1023] phys: 0.. 255 flags: 0x006 tot: 256 ext: 2 logical: [ 1280.. 1535] phys: 0.. 255 flags: 0x007 tot: 256 Tested-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Andreas Dilger <adilger@dilger.ca> Signed-off-by: Yongqiang Yang <xiaoqiangnk@gmail.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-02-27 22:25:47 +00:00
/* Try to find the 1st mapped buffer. */
end = ((__u64)pages[index]->index << PAGE_SHIFT) >>
ext4: make FIEMAP and delayed allocation play well together Fix the FIEMAP ioctl so that it returns all of the page ranges which are still subject to delayed allocation. We were missing some cases if the file was sparse. Reported by Chris Mason <chris.mason@oracle.com>: >We've had reports on btrfs that cp is giving us files full of zeros >instead of actually copying them. It was tracked down to a bug with >the btrfs fiemap implementation where it was returning holes for >delalloc ranges. > >Newer versions of cp are trusting fiemap to tell it where the holes >are, which does seem like a pretty neat trick. > >I decided to give xfs and ext4 a shot with a few tests cases too, xfs >passed with all the ones btrfs was getting wrong, and ext4 got the basic >delalloc case right. >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 >$ fiemap-test foo >ext: 0 logical: [ 0.. 255] phys: 0.. 255 >flags: 0x007 tot: 256 > >Horray! But once we throw a hole in, things go bad: >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 >$ fiemap-test foo >< no output > > >We've got a delalloc extent after the hole and ext4 fiemap didn't find >it. If I run sync to kick the delalloc out: >$sync >$ fiemap-test foo >ext: 0 logical: [ 256.. 511] phys: 34048.. 34303 >flags: 0x001 tot: 256 > >fiemap-test is sitting in my /usr/local/bin, and I have no idea how it >got there. It's full of pretty comments so I know it isn't mine, but >you can grab it here: > >http://oss.oracle.com/~mason/fiemap-test.c > >xfsqa has a fiemap program too. After Fix, test results are as follows: ext: 0 logical: [ 256.. 511] phys: 0.. 255 flags: 0x007 tot: 256 ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x001 tot: 256 $ mkfs.ext4 /dev/xxx $ mount /dev/xxx /mnt $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 $ sync $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=3 $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=5 $ fiemap-test foo ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x000 tot: 256 ext: 1 logical: [ 768.. 1023] phys: 0.. 255 flags: 0x006 tot: 256 ext: 2 logical: [ 1280.. 1535] phys: 0.. 255 flags: 0x007 tot: 256 Tested-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Andreas Dilger <adilger@dilger.ca> Signed-off-by: Yongqiang Yang <xiaoqiangnk@gmail.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-02-27 22:25:47 +00:00
blksize_bits;
if (!page_has_buffers(pages[index]))
ext4: make FIEMAP and delayed allocation play well together Fix the FIEMAP ioctl so that it returns all of the page ranges which are still subject to delayed allocation. We were missing some cases if the file was sparse. Reported by Chris Mason <chris.mason@oracle.com>: >We've had reports on btrfs that cp is giving us files full of zeros >instead of actually copying them. It was tracked down to a bug with >the btrfs fiemap implementation where it was returning holes for >delalloc ranges. > >Newer versions of cp are trusting fiemap to tell it where the holes >are, which does seem like a pretty neat trick. > >I decided to give xfs and ext4 a shot with a few tests cases too, xfs >passed with all the ones btrfs was getting wrong, and ext4 got the basic >delalloc case right. >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 >$ fiemap-test foo >ext: 0 logical: [ 0.. 255] phys: 0.. 255 >flags: 0x007 tot: 256 > >Horray! But once we throw a hole in, things go bad: >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 >$ fiemap-test foo >< no output > > >We've got a delalloc extent after the hole and ext4 fiemap didn't find >it. If I run sync to kick the delalloc out: >$sync >$ fiemap-test foo >ext: 0 logical: [ 256.. 511] phys: 34048.. 34303 >flags: 0x001 tot: 256 > >fiemap-test is sitting in my /usr/local/bin, and I have no idea how it >got there. It's full of pretty comments so I know it isn't mine, but >you can grab it here: > >http://oss.oracle.com/~mason/fiemap-test.c > >xfsqa has a fiemap program too. After Fix, test results are as follows: ext: 0 logical: [ 256.. 511] phys: 0.. 255 flags: 0x007 tot: 256 ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x001 tot: 256 $ mkfs.ext4 /dev/xxx $ mount /dev/xxx /mnt $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 $ sync $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=3 $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=5 $ fiemap-test foo ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x000 tot: 256 ext: 1 logical: [ 768.. 1023] phys: 0.. 255 flags: 0x006 tot: 256 ext: 2 logical: [ 1280.. 1535] phys: 0.. 255 flags: 0x007 tot: 256 Tested-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Andreas Dilger <adilger@dilger.ca> Signed-off-by: Yongqiang Yang <xiaoqiangnk@gmail.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-02-27 22:25:47 +00:00
goto out;
head = page_buffers(pages[index]);
ext4: make FIEMAP and delayed allocation play well together Fix the FIEMAP ioctl so that it returns all of the page ranges which are still subject to delayed allocation. We were missing some cases if the file was sparse. Reported by Chris Mason <chris.mason@oracle.com>: >We've had reports on btrfs that cp is giving us files full of zeros >instead of actually copying them. It was tracked down to a bug with >the btrfs fiemap implementation where it was returning holes for >delalloc ranges. > >Newer versions of cp are trusting fiemap to tell it where the holes >are, which does seem like a pretty neat trick. > >I decided to give xfs and ext4 a shot with a few tests cases too, xfs >passed with all the ones btrfs was getting wrong, and ext4 got the basic >delalloc case right. >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 >$ fiemap-test foo >ext: 0 logical: [ 0.. 255] phys: 0.. 255 >flags: 0x007 tot: 256 > >Horray! But once we throw a hole in, things go bad: >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 >$ fiemap-test foo >< no output > > >We've got a delalloc extent after the hole and ext4 fiemap didn't find >it. If I run sync to kick the delalloc out: >$sync >$ fiemap-test foo >ext: 0 logical: [ 256.. 511] phys: 34048.. 34303 >flags: 0x001 tot: 256 > >fiemap-test is sitting in my /usr/local/bin, and I have no idea how it >got there. It's full of pretty comments so I know it isn't mine, but >you can grab it here: > >http://oss.oracle.com/~mason/fiemap-test.c > >xfsqa has a fiemap program too. After Fix, test results are as follows: ext: 0 logical: [ 256.. 511] phys: 0.. 255 flags: 0x007 tot: 256 ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x001 tot: 256 $ mkfs.ext4 /dev/xxx $ mount /dev/xxx /mnt $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 $ sync $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=3 $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=5 $ fiemap-test foo ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x000 tot: 256 ext: 1 logical: [ 768.. 1023] phys: 0.. 255 flags: 0x006 tot: 256 ext: 2 logical: [ 1280.. 1535] phys: 0.. 255 flags: 0x007 tot: 256 Tested-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Andreas Dilger <adilger@dilger.ca> Signed-off-by: Yongqiang Yang <xiaoqiangnk@gmail.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-02-27 22:25:47 +00:00
if (!head)
goto out;
index++;
ext4: make FIEMAP and delayed allocation play well together Fix the FIEMAP ioctl so that it returns all of the page ranges which are still subject to delayed allocation. We were missing some cases if the file was sparse. Reported by Chris Mason <chris.mason@oracle.com>: >We've had reports on btrfs that cp is giving us files full of zeros >instead of actually copying them. It was tracked down to a bug with >the btrfs fiemap implementation where it was returning holes for >delalloc ranges. > >Newer versions of cp are trusting fiemap to tell it where the holes >are, which does seem like a pretty neat trick. > >I decided to give xfs and ext4 a shot with a few tests cases too, xfs >passed with all the ones btrfs was getting wrong, and ext4 got the basic >delalloc case right. >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 >$ fiemap-test foo >ext: 0 logical: [ 0.. 255] phys: 0.. 255 >flags: 0x007 tot: 256 > >Horray! But once we throw a hole in, things go bad: >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 >$ fiemap-test foo >< no output > > >We've got a delalloc extent after the hole and ext4 fiemap didn't find >it. If I run sync to kick the delalloc out: >$sync >$ fiemap-test foo >ext: 0 logical: [ 256.. 511] phys: 34048.. 34303 >flags: 0x001 tot: 256 > >fiemap-test is sitting in my /usr/local/bin, and I have no idea how it >got there. It's full of pretty comments so I know it isn't mine, but >you can grab it here: > >http://oss.oracle.com/~mason/fiemap-test.c > >xfsqa has a fiemap program too. After Fix, test results are as follows: ext: 0 logical: [ 256.. 511] phys: 0.. 255 flags: 0x007 tot: 256 ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x001 tot: 256 $ mkfs.ext4 /dev/xxx $ mount /dev/xxx /mnt $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 $ sync $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=3 $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=5 $ fiemap-test foo ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x000 tot: 256 ext: 1 logical: [ 768.. 1023] phys: 0.. 255 flags: 0x006 tot: 256 ext: 2 logical: [ 1280.. 1535] phys: 0.. 255 flags: 0x007 tot: 256 Tested-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Andreas Dilger <adilger@dilger.ca> Signed-off-by: Yongqiang Yang <xiaoqiangnk@gmail.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-02-27 22:25:47 +00:00
bh = head;
do {
if (end >= newex->ec_block +
newex->ec_len)
/* The buffer is out of
* the request range.
*/
goto out;
if (buffer_mapped(bh) &&
end >= newex->ec_block) {
start_index = index - 1;
ext4: make FIEMAP and delayed allocation play well together Fix the FIEMAP ioctl so that it returns all of the page ranges which are still subject to delayed allocation. We were missing some cases if the file was sparse. Reported by Chris Mason <chris.mason@oracle.com>: >We've had reports on btrfs that cp is giving us files full of zeros >instead of actually copying them. It was tracked down to a bug with >the btrfs fiemap implementation where it was returning holes for >delalloc ranges. > >Newer versions of cp are trusting fiemap to tell it where the holes >are, which does seem like a pretty neat trick. > >I decided to give xfs and ext4 a shot with a few tests cases too, xfs >passed with all the ones btrfs was getting wrong, and ext4 got the basic >delalloc case right. >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 >$ fiemap-test foo >ext: 0 logical: [ 0.. 255] phys: 0.. 255 >flags: 0x007 tot: 256 > >Horray! But once we throw a hole in, things go bad: >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 >$ fiemap-test foo >< no output > > >We've got a delalloc extent after the hole and ext4 fiemap didn't find >it. If I run sync to kick the delalloc out: >$sync >$ fiemap-test foo >ext: 0 logical: [ 256.. 511] phys: 34048.. 34303 >flags: 0x001 tot: 256 > >fiemap-test is sitting in my /usr/local/bin, and I have no idea how it >got there. It's full of pretty comments so I know it isn't mine, but >you can grab it here: > >http://oss.oracle.com/~mason/fiemap-test.c > >xfsqa has a fiemap program too. After Fix, test results are as follows: ext: 0 logical: [ 256.. 511] phys: 0.. 255 flags: 0x007 tot: 256 ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x001 tot: 256 $ mkfs.ext4 /dev/xxx $ mount /dev/xxx /mnt $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 $ sync $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=3 $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=5 $ fiemap-test foo ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x000 tot: 256 ext: 1 logical: [ 768.. 1023] phys: 0.. 255 flags: 0x006 tot: 256 ext: 2 logical: [ 1280.. 1535] phys: 0.. 255 flags: 0x007 tot: 256 Tested-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Andreas Dilger <adilger@dilger.ca> Signed-off-by: Yongqiang Yang <xiaoqiangnk@gmail.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-02-27 22:25:47 +00:00
/* get the 1st mapped buffer. */
goto found_mapped_buffer;
}
ext4: make FIEMAP and delayed allocation play well together Fix the FIEMAP ioctl so that it returns all of the page ranges which are still subject to delayed allocation. We were missing some cases if the file was sparse. Reported by Chris Mason <chris.mason@oracle.com>: >We've had reports on btrfs that cp is giving us files full of zeros >instead of actually copying them. It was tracked down to a bug with >the btrfs fiemap implementation where it was returning holes for >delalloc ranges. > >Newer versions of cp are trusting fiemap to tell it where the holes >are, which does seem like a pretty neat trick. > >I decided to give xfs and ext4 a shot with a few tests cases too, xfs >passed with all the ones btrfs was getting wrong, and ext4 got the basic >delalloc case right. >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 >$ fiemap-test foo >ext: 0 logical: [ 0.. 255] phys: 0.. 255 >flags: 0x007 tot: 256 > >Horray! But once we throw a hole in, things go bad: >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 >$ fiemap-test foo >< no output > > >We've got a delalloc extent after the hole and ext4 fiemap didn't find >it. If I run sync to kick the delalloc out: >$sync >$ fiemap-test foo >ext: 0 logical: [ 256.. 511] phys: 34048.. 34303 >flags: 0x001 tot: 256 > >fiemap-test is sitting in my /usr/local/bin, and I have no idea how it >got there. It's full of pretty comments so I know it isn't mine, but >you can grab it here: > >http://oss.oracle.com/~mason/fiemap-test.c > >xfsqa has a fiemap program too. After Fix, test results are as follows: ext: 0 logical: [ 256.. 511] phys: 0.. 255 flags: 0x007 tot: 256 ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x001 tot: 256 $ mkfs.ext4 /dev/xxx $ mount /dev/xxx /mnt $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 $ sync $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=3 $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=5 $ fiemap-test foo ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x000 tot: 256 ext: 1 logical: [ 768.. 1023] phys: 0.. 255 flags: 0x006 tot: 256 ext: 2 logical: [ 1280.. 1535] phys: 0.. 255 flags: 0x007 tot: 256 Tested-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Andreas Dilger <adilger@dilger.ca> Signed-off-by: Yongqiang Yang <xiaoqiangnk@gmail.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-02-27 22:25:47 +00:00
bh = bh->b_this_page;
end++;
} while (bh != head);
/* No mapped buffer in the range found in this page,
* We need to look up next page.
*/
if (index >= ret) {
/* There is no page left, but we need to limit
* newex->ec_len.
*/
newex->ec_len = end - newex->ec_block;
goto out;
}
goto next_page;
} else {
ext4: make FIEMAP and delayed allocation play well together Fix the FIEMAP ioctl so that it returns all of the page ranges which are still subject to delayed allocation. We were missing some cases if the file was sparse. Reported by Chris Mason <chris.mason@oracle.com>: >We've had reports on btrfs that cp is giving us files full of zeros >instead of actually copying them. It was tracked down to a bug with >the btrfs fiemap implementation where it was returning holes for >delalloc ranges. > >Newer versions of cp are trusting fiemap to tell it where the holes >are, which does seem like a pretty neat trick. > >I decided to give xfs and ext4 a shot with a few tests cases too, xfs >passed with all the ones btrfs was getting wrong, and ext4 got the basic >delalloc case right. >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 >$ fiemap-test foo >ext: 0 logical: [ 0.. 255] phys: 0.. 255 >flags: 0x007 tot: 256 > >Horray! But once we throw a hole in, things go bad: >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 >$ fiemap-test foo >< no output > > >We've got a delalloc extent after the hole and ext4 fiemap didn't find >it. If I run sync to kick the delalloc out: >$sync >$ fiemap-test foo >ext: 0 logical: [ 256.. 511] phys: 34048.. 34303 >flags: 0x001 tot: 256 > >fiemap-test is sitting in my /usr/local/bin, and I have no idea how it >got there. It's full of pretty comments so I know it isn't mine, but >you can grab it here: > >http://oss.oracle.com/~mason/fiemap-test.c > >xfsqa has a fiemap program too. After Fix, test results are as follows: ext: 0 logical: [ 256.. 511] phys: 0.. 255 flags: 0x007 tot: 256 ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x001 tot: 256 $ mkfs.ext4 /dev/xxx $ mount /dev/xxx /mnt $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 $ sync $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=3 $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=5 $ fiemap-test foo ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x000 tot: 256 ext: 1 logical: [ 768.. 1023] phys: 0.. 255 flags: 0x006 tot: 256 ext: 2 logical: [ 1280.. 1535] phys: 0.. 255 flags: 0x007 tot: 256 Tested-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Andreas Dilger <adilger@dilger.ca> Signed-off-by: Yongqiang Yang <xiaoqiangnk@gmail.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-02-27 22:25:47 +00:00
/*Find contiguous delayed buffers. */
if (ret > 0 && pages[0]->index == last_offset)
head = page_buffers(pages[0]);
bh = head;
index = 1;
start_index = 0;
}
ext4: make FIEMAP and delayed allocation play well together Fix the FIEMAP ioctl so that it returns all of the page ranges which are still subject to delayed allocation. We were missing some cases if the file was sparse. Reported by Chris Mason <chris.mason@oracle.com>: >We've had reports on btrfs that cp is giving us files full of zeros >instead of actually copying them. It was tracked down to a bug with >the btrfs fiemap implementation where it was returning holes for >delalloc ranges. > >Newer versions of cp are trusting fiemap to tell it where the holes >are, which does seem like a pretty neat trick. > >I decided to give xfs and ext4 a shot with a few tests cases too, xfs >passed with all the ones btrfs was getting wrong, and ext4 got the basic >delalloc case right. >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 >$ fiemap-test foo >ext: 0 logical: [ 0.. 255] phys: 0.. 255 >flags: 0x007 tot: 256 > >Horray! But once we throw a hole in, things go bad: >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 >$ fiemap-test foo >< no output > > >We've got a delalloc extent after the hole and ext4 fiemap didn't find >it. If I run sync to kick the delalloc out: >$sync >$ fiemap-test foo >ext: 0 logical: [ 256.. 511] phys: 34048.. 34303 >flags: 0x001 tot: 256 > >fiemap-test is sitting in my /usr/local/bin, and I have no idea how it >got there. It's full of pretty comments so I know it isn't mine, but >you can grab it here: > >http://oss.oracle.com/~mason/fiemap-test.c > >xfsqa has a fiemap program too. After Fix, test results are as follows: ext: 0 logical: [ 256.. 511] phys: 0.. 255 flags: 0x007 tot: 256 ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x001 tot: 256 $ mkfs.ext4 /dev/xxx $ mount /dev/xxx /mnt $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 $ sync $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=3 $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=5 $ fiemap-test foo ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x000 tot: 256 ext: 1 logical: [ 768.. 1023] phys: 0.. 255 flags: 0x006 tot: 256 ext: 2 logical: [ 1280.. 1535] phys: 0.. 255 flags: 0x007 tot: 256 Tested-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Andreas Dilger <adilger@dilger.ca> Signed-off-by: Yongqiang Yang <xiaoqiangnk@gmail.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-02-27 22:25:47 +00:00
found_mapped_buffer:
if (bh != NULL && buffer_delay(bh)) {
/* 1st or contiguous delayed buffer found. */
if (!(flags & FIEMAP_EXTENT_DELALLOC)) {
/*
* 1st delayed buffer found, record
* the start of extent.
*/
flags |= FIEMAP_EXTENT_DELALLOC;
newex->ec_block = end;
logical = (__u64)end << blksize_bits;
}
/* Find contiguous delayed buffers. */
do {
if (!buffer_delay(bh))
goto found_delayed_extent;
bh = bh->b_this_page;
end++;
} while (bh != head);
for (; index < ret; index++) {
ext4: make FIEMAP and delayed allocation play well together Fix the FIEMAP ioctl so that it returns all of the page ranges which are still subject to delayed allocation. We were missing some cases if the file was sparse. Reported by Chris Mason <chris.mason@oracle.com>: >We've had reports on btrfs that cp is giving us files full of zeros >instead of actually copying them. It was tracked down to a bug with >the btrfs fiemap implementation where it was returning holes for >delalloc ranges. > >Newer versions of cp are trusting fiemap to tell it where the holes >are, which does seem like a pretty neat trick. > >I decided to give xfs and ext4 a shot with a few tests cases too, xfs >passed with all the ones btrfs was getting wrong, and ext4 got the basic >delalloc case right. >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 >$ fiemap-test foo >ext: 0 logical: [ 0.. 255] phys: 0.. 255 >flags: 0x007 tot: 256 > >Horray! But once we throw a hole in, things go bad: >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 >$ fiemap-test foo >< no output > > >We've got a delalloc extent after the hole and ext4 fiemap didn't find >it. If I run sync to kick the delalloc out: >$sync >$ fiemap-test foo >ext: 0 logical: [ 256.. 511] phys: 34048.. 34303 >flags: 0x001 tot: 256 > >fiemap-test is sitting in my /usr/local/bin, and I have no idea how it >got there. It's full of pretty comments so I know it isn't mine, but >you can grab it here: > >http://oss.oracle.com/~mason/fiemap-test.c > >xfsqa has a fiemap program too. After Fix, test results are as follows: ext: 0 logical: [ 256.. 511] phys: 0.. 255 flags: 0x007 tot: 256 ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x001 tot: 256 $ mkfs.ext4 /dev/xxx $ mount /dev/xxx /mnt $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 $ sync $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=3 $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=5 $ fiemap-test foo ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x000 tot: 256 ext: 1 logical: [ 768.. 1023] phys: 0.. 255 flags: 0x006 tot: 256 ext: 2 logical: [ 1280.. 1535] phys: 0.. 255 flags: 0x007 tot: 256 Tested-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Andreas Dilger <adilger@dilger.ca> Signed-off-by: Yongqiang Yang <xiaoqiangnk@gmail.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-02-27 22:25:47 +00:00
if (!page_has_buffers(pages[index])) {
bh = NULL;
break;
}
head = page_buffers(pages[index]);
if (!head) {
bh = NULL;
break;
}
ext4: make FIEMAP and delayed allocation play well together Fix the FIEMAP ioctl so that it returns all of the page ranges which are still subject to delayed allocation. We were missing some cases if the file was sparse. Reported by Chris Mason <chris.mason@oracle.com>: >We've had reports on btrfs that cp is giving us files full of zeros >instead of actually copying them. It was tracked down to a bug with >the btrfs fiemap implementation where it was returning holes for >delalloc ranges. > >Newer versions of cp are trusting fiemap to tell it where the holes >are, which does seem like a pretty neat trick. > >I decided to give xfs and ext4 a shot with a few tests cases too, xfs >passed with all the ones btrfs was getting wrong, and ext4 got the basic >delalloc case right. >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 >$ fiemap-test foo >ext: 0 logical: [ 0.. 255] phys: 0.. 255 >flags: 0x007 tot: 256 > >Horray! But once we throw a hole in, things go bad: >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 >$ fiemap-test foo >< no output > > >We've got a delalloc extent after the hole and ext4 fiemap didn't find >it. If I run sync to kick the delalloc out: >$sync >$ fiemap-test foo >ext: 0 logical: [ 256.. 511] phys: 34048.. 34303 >flags: 0x001 tot: 256 > >fiemap-test is sitting in my /usr/local/bin, and I have no idea how it >got there. It's full of pretty comments so I know it isn't mine, but >you can grab it here: > >http://oss.oracle.com/~mason/fiemap-test.c > >xfsqa has a fiemap program too. After Fix, test results are as follows: ext: 0 logical: [ 256.. 511] phys: 0.. 255 flags: 0x007 tot: 256 ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x001 tot: 256 $ mkfs.ext4 /dev/xxx $ mount /dev/xxx /mnt $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 $ sync $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=3 $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=5 $ fiemap-test foo ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x000 tot: 256 ext: 1 logical: [ 768.. 1023] phys: 0.. 255 flags: 0x006 tot: 256 ext: 2 logical: [ 1280.. 1535] phys: 0.. 255 flags: 0x007 tot: 256 Tested-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Andreas Dilger <adilger@dilger.ca> Signed-off-by: Yongqiang Yang <xiaoqiangnk@gmail.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-02-27 22:25:47 +00:00
if (pages[index]->index !=
pages[start_index]->index + index
- start_index) {
ext4: make FIEMAP and delayed allocation play well together Fix the FIEMAP ioctl so that it returns all of the page ranges which are still subject to delayed allocation. We were missing some cases if the file was sparse. Reported by Chris Mason <chris.mason@oracle.com>: >We've had reports on btrfs that cp is giving us files full of zeros >instead of actually copying them. It was tracked down to a bug with >the btrfs fiemap implementation where it was returning holes for >delalloc ranges. > >Newer versions of cp are trusting fiemap to tell it where the holes >are, which does seem like a pretty neat trick. > >I decided to give xfs and ext4 a shot with a few tests cases too, xfs >passed with all the ones btrfs was getting wrong, and ext4 got the basic >delalloc case right. >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 >$ fiemap-test foo >ext: 0 logical: [ 0.. 255] phys: 0.. 255 >flags: 0x007 tot: 256 > >Horray! But once we throw a hole in, things go bad: >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 >$ fiemap-test foo >< no output > > >We've got a delalloc extent after the hole and ext4 fiemap didn't find >it. If I run sync to kick the delalloc out: >$sync >$ fiemap-test foo >ext: 0 logical: [ 256.. 511] phys: 34048.. 34303 >flags: 0x001 tot: 256 > >fiemap-test is sitting in my /usr/local/bin, and I have no idea how it >got there. It's full of pretty comments so I know it isn't mine, but >you can grab it here: > >http://oss.oracle.com/~mason/fiemap-test.c > >xfsqa has a fiemap program too. After Fix, test results are as follows: ext: 0 logical: [ 256.. 511] phys: 0.. 255 flags: 0x007 tot: 256 ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x001 tot: 256 $ mkfs.ext4 /dev/xxx $ mount /dev/xxx /mnt $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 $ sync $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=3 $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=5 $ fiemap-test foo ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x000 tot: 256 ext: 1 logical: [ 768.. 1023] phys: 0.. 255 flags: 0x006 tot: 256 ext: 2 logical: [ 1280.. 1535] phys: 0.. 255 flags: 0x007 tot: 256 Tested-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Andreas Dilger <adilger@dilger.ca> Signed-off-by: Yongqiang Yang <xiaoqiangnk@gmail.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-02-27 22:25:47 +00:00
/* Blocks are not contiguous. */
bh = NULL;
break;
}
bh = head;
do {
if (!buffer_delay(bh))
/* Delayed-extent ends. */
goto found_delayed_extent;
bh = bh->b_this_page;
end++;
} while (bh != head);
}
} else if (!(flags & FIEMAP_EXTENT_DELALLOC))
/* a hole found. */
goto out;
found_delayed_extent:
newex->ec_len = min(end - newex->ec_block,
(ext4_lblk_t)EXT_INIT_MAX_LEN);
if (ret == nr_pages && bh != NULL &&
newex->ec_len < EXT_INIT_MAX_LEN &&
buffer_delay(bh)) {
/* Have not collected an extent and continue. */
for (index = 0; index < ret; index++)
page_cache_release(pages[index]);
goto repeat;
}
ext4: make FIEMAP and delayed allocation play well together Fix the FIEMAP ioctl so that it returns all of the page ranges which are still subject to delayed allocation. We were missing some cases if the file was sparse. Reported by Chris Mason <chris.mason@oracle.com>: >We've had reports on btrfs that cp is giving us files full of zeros >instead of actually copying them. It was tracked down to a bug with >the btrfs fiemap implementation where it was returning holes for >delalloc ranges. > >Newer versions of cp are trusting fiemap to tell it where the holes >are, which does seem like a pretty neat trick. > >I decided to give xfs and ext4 a shot with a few tests cases too, xfs >passed with all the ones btrfs was getting wrong, and ext4 got the basic >delalloc case right. >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 >$ fiemap-test foo >ext: 0 logical: [ 0.. 255] phys: 0.. 255 >flags: 0x007 tot: 256 > >Horray! But once we throw a hole in, things go bad: >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 >$ fiemap-test foo >< no output > > >We've got a delalloc extent after the hole and ext4 fiemap didn't find >it. If I run sync to kick the delalloc out: >$sync >$ fiemap-test foo >ext: 0 logical: [ 256.. 511] phys: 34048.. 34303 >flags: 0x001 tot: 256 > >fiemap-test is sitting in my /usr/local/bin, and I have no idea how it >got there. It's full of pretty comments so I know it isn't mine, but >you can grab it here: > >http://oss.oracle.com/~mason/fiemap-test.c > >xfsqa has a fiemap program too. After Fix, test results are as follows: ext: 0 logical: [ 256.. 511] phys: 0.. 255 flags: 0x007 tot: 256 ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x001 tot: 256 $ mkfs.ext4 /dev/xxx $ mount /dev/xxx /mnt $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 $ sync $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=3 $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=5 $ fiemap-test foo ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x000 tot: 256 ext: 1 logical: [ 768.. 1023] phys: 0.. 255 flags: 0x006 tot: 256 ext: 2 logical: [ 1280.. 1535] phys: 0.. 255 flags: 0x007 tot: 256 Tested-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Andreas Dilger <adilger@dilger.ca> Signed-off-by: Yongqiang Yang <xiaoqiangnk@gmail.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-02-27 22:25:47 +00:00
for (index = 0; index < ret; index++)
page_cache_release(pages[index]);
kfree(pages);
}
physical = (__u64)newex->ec_start << blksize_bits;
length = (__u64)newex->ec_len << blksize_bits;
if (ex && ext4_ext_is_uninitialized(ex))
flags |= FIEMAP_EXTENT_UNWRITTEN;
if (next == EXT_MAX_BLOCKS)
flags |= FIEMAP_EXTENT_LAST;
ext4: make FIEMAP and delayed allocation play well together Fix the FIEMAP ioctl so that it returns all of the page ranges which are still subject to delayed allocation. We were missing some cases if the file was sparse. Reported by Chris Mason <chris.mason@oracle.com>: >We've had reports on btrfs that cp is giving us files full of zeros >instead of actually copying them. It was tracked down to a bug with >the btrfs fiemap implementation where it was returning holes for >delalloc ranges. > >Newer versions of cp are trusting fiemap to tell it where the holes >are, which does seem like a pretty neat trick. > >I decided to give xfs and ext4 a shot with a few tests cases too, xfs >passed with all the ones btrfs was getting wrong, and ext4 got the basic >delalloc case right. >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 >$ fiemap-test foo >ext: 0 logical: [ 0.. 255] phys: 0.. 255 >flags: 0x007 tot: 256 > >Horray! But once we throw a hole in, things go bad: >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 >$ fiemap-test foo >< no output > > >We've got a delalloc extent after the hole and ext4 fiemap didn't find >it. If I run sync to kick the delalloc out: >$sync >$ fiemap-test foo >ext: 0 logical: [ 256.. 511] phys: 34048.. 34303 >flags: 0x001 tot: 256 > >fiemap-test is sitting in my /usr/local/bin, and I have no idea how it >got there. It's full of pretty comments so I know it isn't mine, but >you can grab it here: > >http://oss.oracle.com/~mason/fiemap-test.c > >xfsqa has a fiemap program too. After Fix, test results are as follows: ext: 0 logical: [ 256.. 511] phys: 0.. 255 flags: 0x007 tot: 256 ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x001 tot: 256 $ mkfs.ext4 /dev/xxx $ mount /dev/xxx /mnt $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 $ sync $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=3 $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=5 $ fiemap-test foo ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x000 tot: 256 ext: 1 logical: [ 768.. 1023] phys: 0.. 255 flags: 0x006 tot: 256 ext: 2 logical: [ 1280.. 1535] phys: 0.. 255 flags: 0x007 tot: 256 Tested-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Andreas Dilger <adilger@dilger.ca> Signed-off-by: Yongqiang Yang <xiaoqiangnk@gmail.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-02-27 22:25:47 +00:00
ret = fiemap_fill_next_extent(fieinfo, logical, physical,
length, flags);
ext4: make FIEMAP and delayed allocation play well together Fix the FIEMAP ioctl so that it returns all of the page ranges which are still subject to delayed allocation. We were missing some cases if the file was sparse. Reported by Chris Mason <chris.mason@oracle.com>: >We've had reports on btrfs that cp is giving us files full of zeros >instead of actually copying them. It was tracked down to a bug with >the btrfs fiemap implementation where it was returning holes for >delalloc ranges. > >Newer versions of cp are trusting fiemap to tell it where the holes >are, which does seem like a pretty neat trick. > >I decided to give xfs and ext4 a shot with a few tests cases too, xfs >passed with all the ones btrfs was getting wrong, and ext4 got the basic >delalloc case right. >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 >$ fiemap-test foo >ext: 0 logical: [ 0.. 255] phys: 0.. 255 >flags: 0x007 tot: 256 > >Horray! But once we throw a hole in, things go bad: >$ mkfs.ext4 /dev/xxx >$ mount /dev/xxx /mnt >$ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 >$ fiemap-test foo >< no output > > >We've got a delalloc extent after the hole and ext4 fiemap didn't find >it. If I run sync to kick the delalloc out: >$sync >$ fiemap-test foo >ext: 0 logical: [ 256.. 511] phys: 34048.. 34303 >flags: 0x001 tot: 256 > >fiemap-test is sitting in my /usr/local/bin, and I have no idea how it >got there. It's full of pretty comments so I know it isn't mine, but >you can grab it here: > >http://oss.oracle.com/~mason/fiemap-test.c > >xfsqa has a fiemap program too. After Fix, test results are as follows: ext: 0 logical: [ 256.. 511] phys: 0.. 255 flags: 0x007 tot: 256 ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x001 tot: 256 $ mkfs.ext4 /dev/xxx $ mount /dev/xxx /mnt $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=1 $ sync $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=3 $ dd if=/dev/zero of=/mnt/foo bs=1M count=1 seek=5 $ fiemap-test foo ext: 0 logical: [ 256.. 511] phys: 33280.. 33535 flags: 0x000 tot: 256 ext: 1 logical: [ 768.. 1023] phys: 0.. 255 flags: 0x006 tot: 256 ext: 2 logical: [ 1280.. 1535] phys: 0.. 255 flags: 0x007 tot: 256 Tested-by: Eric Sandeen <sandeen@redhat.com> Reviewed-by: Andreas Dilger <adilger@dilger.ca> Signed-off-by: Yongqiang Yang <xiaoqiangnk@gmail.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-02-27 22:25:47 +00:00
if (ret < 0)
return ret;
if (ret == 1)
return EXT_BREAK;
return EXT_CONTINUE;
}
/* fiemap flags we can handle specified here */
#define EXT4_FIEMAP_FLAGS (FIEMAP_FLAG_SYNC|FIEMAP_FLAG_XATTR)
static int ext4_xattr_fiemap(struct inode *inode,
struct fiemap_extent_info *fieinfo)
{
__u64 physical = 0;
__u64 length;
__u32 flags = FIEMAP_EXTENT_LAST;
int blockbits = inode->i_sb->s_blocksize_bits;
int error = 0;
/* in-inode? */
if (ext4_test_inode_state(inode, EXT4_STATE_XATTR)) {
struct ext4_iloc iloc;
int offset; /* offset of xattr in inode */
error = ext4_get_inode_loc(inode, &iloc);
if (error)
return error;
physical = iloc.bh->b_blocknr << blockbits;
offset = EXT4_GOOD_OLD_INODE_SIZE +
EXT4_I(inode)->i_extra_isize;
physical += offset;
length = EXT4_SB(inode->i_sb)->s_inode_size - offset;
flags |= FIEMAP_EXTENT_DATA_INLINE;
brelse(iloc.bh);
} else { /* external block */
physical = EXT4_I(inode)->i_file_acl << blockbits;
length = inode->i_sb->s_blocksize;
}
if (physical)
error = fiemap_fill_next_extent(fieinfo, 0, physical,
length, flags);
return (error < 0 ? error : 0);
}
/*
* ext4_ext_punch_hole
*
* Punches a hole of "length" bytes in a file starting
* at byte "offset"
*
* @inode: The inode of the file to punch a hole in
* @offset: The starting byte offset of the hole
* @length: The length of the hole
*
* Returns the number of blocks removed or negative on err
*/
int ext4_ext_punch_hole(struct file *file, loff_t offset, loff_t length)
{
struct inode *inode = file->f_path.dentry->d_inode;
struct super_block *sb = inode->i_sb;
struct ext4_ext_cache cache_ex;
ext4_lblk_t first_block, last_block, num_blocks, iblock, max_blocks;
struct address_space *mapping = inode->i_mapping;
struct ext4_map_blocks map;
handle_t *handle;
loff_t first_block_offset, last_block_offset, block_len;
loff_t first_page, last_page, first_page_offset, last_page_offset;
int ret, credits, blocks_released, err = 0;
first_block = (offset + sb->s_blocksize - 1) >>
EXT4_BLOCK_SIZE_BITS(sb);
last_block = (offset + length) >> EXT4_BLOCK_SIZE_BITS(sb);
first_block_offset = first_block << EXT4_BLOCK_SIZE_BITS(sb);
last_block_offset = last_block << EXT4_BLOCK_SIZE_BITS(sb);
first_page = (offset + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
last_page = (offset + length) >> PAGE_CACHE_SHIFT;
first_page_offset = first_page << PAGE_CACHE_SHIFT;
last_page_offset = last_page << PAGE_CACHE_SHIFT;
/*
* Write out all dirty pages to avoid race conditions
* Then release them.
*/
if (mapping->nrpages && mapping_tagged(mapping, PAGECACHE_TAG_DIRTY)) {
err = filemap_write_and_wait_range(mapping,
first_page_offset == 0 ? 0 : first_page_offset-1,
last_page_offset);
if (err)
return err;
}
/* Now release the pages */
if (last_page_offset > first_page_offset) {
truncate_inode_pages_range(mapping, first_page_offset,
last_page_offset-1);
}
/* finish any pending end_io work */
ext4_flush_completed_IO(inode);
credits = ext4_writepage_trans_blocks(inode);
handle = ext4_journal_start(inode, credits);
if (IS_ERR(handle))
return PTR_ERR(handle);
err = ext4_orphan_add(handle, inode);
if (err)
goto out;
/*
* Now we need to zero out the un block aligned data.
* If the file is smaller than a block, just
* zero out the middle
*/
if (first_block > last_block)
ext4_block_zero_page_range(handle, mapping, offset, length);
else {
/* zero out the head of the hole before the first block */
block_len = first_block_offset - offset;
if (block_len > 0)
ext4_block_zero_page_range(handle, mapping,
offset, block_len);
/* zero out the tail of the hole after the last block */
block_len = offset + length - last_block_offset;
if (block_len > 0) {
ext4_block_zero_page_range(handle, mapping,
last_block_offset, block_len);
}
}
/* If there are no blocks to remove, return now */
if (first_block >= last_block)
goto out;
down_write(&EXT4_I(inode)->i_data_sem);
ext4_ext_invalidate_cache(inode);
ext4_discard_preallocations(inode);
/*
* Loop over all the blocks and identify blocks
* that need to be punched out
*/
iblock = first_block;
blocks_released = 0;
while (iblock < last_block) {
max_blocks = last_block - iblock;
num_blocks = 1;
memset(&map, 0, sizeof(map));
map.m_lblk = iblock;
map.m_len = max_blocks;
ret = ext4_ext_map_blocks(handle, inode, &map,
EXT4_GET_BLOCKS_PUNCH_OUT_EXT);
if (ret > 0) {
blocks_released += ret;
num_blocks = ret;
} else if (ret == 0) {
/*
* If map blocks could not find the block,
* then it is in a hole. If the hole was
* not already cached, then map blocks should
* put it in the cache. So we can get the hole
* out of the cache
*/
memset(&cache_ex, 0, sizeof(cache_ex));
if ((ext4_ext_check_cache(inode, iblock, &cache_ex)) &&
!cache_ex.ec_start) {
/* The hole is cached */
num_blocks = cache_ex.ec_block +
cache_ex.ec_len - iblock;
} else {
/* The block could not be identified */
err = -EIO;
break;
}
} else {
/* Map blocks error */
err = ret;
break;
}
if (num_blocks == 0) {
/* This condition should never happen */
ext_debug("Block lookup failed");
err = -EIO;
break;
}
iblock += num_blocks;
}
if (blocks_released > 0) {
ext4_ext_invalidate_cache(inode);
ext4_discard_preallocations(inode);
}
if (IS_SYNC(inode))
ext4_handle_sync(handle);
up_write(&EXT4_I(inode)->i_data_sem);
out:
ext4_orphan_del(handle, inode);
inode->i_mtime = inode->i_ctime = ext4_current_time(inode);
ext4_mark_inode_dirty(handle, inode);
ext4_journal_stop(handle);
return err;
}
int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
__u64 start, __u64 len)
{
ext4_lblk_t start_blk;
int error = 0;
/* fallback to generic here if not in extents fmt */
if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
return generic_block_fiemap(inode, fieinfo, start, len,
ext4_get_block);
if (fiemap_check_flags(fieinfo, EXT4_FIEMAP_FLAGS))
return -EBADR;
if (fieinfo->fi_flags & FIEMAP_FLAG_XATTR) {
error = ext4_xattr_fiemap(inode, fieinfo);
} else {
ext4_lblk_t len_blks;
__u64 last_blk;
start_blk = start >> inode->i_sb->s_blocksize_bits;
last_blk = (start + len - 1) >> inode->i_sb->s_blocksize_bits;
ext4: Fix max file size and logical block counting of extent format file Kazuya Mio reported that he was able to hit BUG_ON(next == lblock) in ext4_ext_put_gap_in_cache() while creating a sparse file in extent format and fill the tail of file up to its end. We will hit the BUG_ON when we write the last block (2^32-1) into the sparse file. The root cause of the problem lies in the fact that we specifically set s_maxbytes so that block at s_maxbytes fit into on-disk extent format, which is 32 bit long. However, we are not storing start and end block number, but rather start block number and length in blocks. It means that in order to cover extent from 0 to EXT_MAX_BLOCK we need EXT_MAX_BLOCK+1 to fit into len (because we counting block 0 as well) - and it does not. The only way to fix it without changing the meaning of the struct ext4_extent members is, as Kazuya Mio suggested, to lower s_maxbytes by one fs block so we can cover the whole extent we can get by the on-disk extent format. Also in many places EXT_MAX_BLOCK is used as length instead of maximum logical block number as the name suggests, it is all a bit messy. So this commit renames it to EXT_MAX_BLOCKS and change its usage in some places to actually be maximum number of blocks in the extent. The bug which this commit fixes can be reproduced as follows: dd if=/dev/zero of=/mnt/mp1/file bs=<blocksize> count=1 seek=$((2**32-2)) sync dd if=/dev/zero of=/mnt/mp1/file bs=<blocksize> count=1 seek=$((2**32-1)) Reported-by: Kazuya Mio <k-mio@sx.jp.nec.com> Signed-off-by: Lukas Czerner <lczerner@redhat.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-06-06 04:05:17 +00:00
if (last_blk >= EXT_MAX_BLOCKS)
last_blk = EXT_MAX_BLOCKS-1;
len_blks = ((ext4_lblk_t) last_blk) - start_blk + 1;
/*
* Walk the extent tree gathering extent information.
* ext4_ext_fiemap_cb will push extents back to user.
*/
error = ext4_ext_walk_space(inode, start_blk, len_blks,
ext4_ext_fiemap_cb, fieinfo);
}
return error;
}