linux_dsm_epyc7002/fs/ext4/move_extent.c

700 lines
21 KiB
C
Raw Normal View History

// SPDX-License-Identifier: LGPL-2.1
/*
* Copyright (c) 2008,2009 NEC Software Tohoku, Ltd.
* Written by Takashi Sato <t-sato@yk.jp.nec.com>
* Akira Fujita <a-fujita@rs.jp.nec.com>
*/
#include <linux/fs.h>
#include <linux/quotaops.h>
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h percpu.h is included by sched.h and module.h and thus ends up being included when building most .c files. percpu.h includes slab.h which in turn includes gfp.h making everything defined by the two files universally available and complicating inclusion dependencies. percpu.h -> slab.h dependency is about to be removed. Prepare for this change by updating users of gfp and slab facilities include those headers directly instead of assuming availability. As this conversion needs to touch large number of source files, the following script is used as the basis of conversion. http://userweb.kernel.org/~tj/misc/slabh-sweep.py The script does the followings. * Scan files for gfp and slab usages and update includes such that only the necessary includes are there. ie. if only gfp is used, gfp.h, if slab is used, slab.h. * When the script inserts a new include, it looks at the include blocks and try to put the new include such that its order conforms to its surrounding. It's put in the include block which contains core kernel includes, in the same order that the rest are ordered - alphabetical, Christmas tree, rev-Xmas-tree or at the end if there doesn't seem to be any matching order. * If the script can't find a place to put a new include (mostly because the file doesn't have fitting include block), it prints out an error message indicating which .h file needs to be added to the file. The conversion was done in the following steps. 1. The initial automatic conversion of all .c files updated slightly over 4000 files, deleting around 700 includes and adding ~480 gfp.h and ~3000 slab.h inclusions. The script emitted errors for ~400 files. 2. Each error was manually checked. Some didn't need the inclusion, some needed manual addition while adding it to implementation .h or embedding .c file was more appropriate for others. This step added inclusions to around 150 files. 3. The script was run again and the output was compared to the edits from #2 to make sure no file was left behind. 4. Several build tests were done and a couple of problems were fixed. e.g. lib/decompress_*.c used malloc/free() wrappers around slab APIs requiring slab.h to be added manually. 5. The script was run on all .h files but without automatically editing them as sprinkling gfp.h and slab.h inclusions around .h files could easily lead to inclusion dependency hell. Most gfp.h inclusion directives were ignored as stuff from gfp.h was usually wildly available and often used in preprocessor macros. Each slab.h inclusion directive was examined and added manually as necessary. 6. percpu.h was updated not to include slab.h. 7. Build test were done on the following configurations and failures were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my distributed build env didn't work with gcov compiles) and a few more options had to be turned off depending on archs to make things build (like ipr on powerpc/64 which failed due to missing writeq). * x86 and x86_64 UP and SMP allmodconfig and a custom test config. * powerpc and powerpc64 SMP allmodconfig * sparc and sparc64 SMP allmodconfig * ia64 SMP allmodconfig * s390 SMP allmodconfig * alpha SMP allmodconfig * um on x86_64 SMP allmodconfig 8. percpu.h modifications were reverted so that it could be applied as a separate patch and serve as bisection point. Given the fact that I had only a couple of failures from tests on step 6, I'm fairly confident about the coverage of this conversion patch. If there is a breakage, it's likely to be something in one of the arch headers which should be easily discoverable easily on most builds of the specific arch. Signed-off-by: Tejun Heo <tj@kernel.org> Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
2010-03-24 15:04:11 +07:00
#include <linux/slab.h>
#include "ext4_jbd2.h"
#include "ext4.h"
#include "ext4_extents.h"
/**
* get_ext_path() - Find an extent path for designated logical block number.
* @inode: inode to be searched
* @lblock: logical block number to find an extent path
* @ppath: pointer to an extent path pointer (for output)
*
* ext4_find_extent wrapper. Return 0 on success, or a negative error value
* on failure.
*/
static inline int
get_ext_path(struct inode *inode, ext4_lblk_t lblock,
struct ext4_ext_path **ppath)
{
struct ext4_ext_path *path;
path = ext4_find_extent(inode, lblock, ppath, EXT4_EX_NOCACHE);
if (IS_ERR(path))
return PTR_ERR(path);
if (path[ext_depth(inode)].p_ext == NULL) {
ext4_ext_drop_refs(path);
kfree(path);
*ppath = NULL;
return -ENODATA;
}
*ppath = path;
return 0;
}
/**
* ext4_double_down_write_data_sem() - write lock two inodes's i_data_sem
* @first: inode to be locked
* @second: inode to be locked
*
* Acquire write lock of i_data_sem of the two inodes
*/
void
ext4_double_down_write_data_sem(struct inode *first, struct inode *second)
{
if (first < second) {
down_write(&EXT4_I(first)->i_data_sem);
down_write_nested(&EXT4_I(second)->i_data_sem, I_DATA_SEM_OTHER);
} else {
down_write(&EXT4_I(second)->i_data_sem);
down_write_nested(&EXT4_I(first)->i_data_sem, I_DATA_SEM_OTHER);
}
}
/**
* ext4_double_up_write_data_sem - Release two inodes' write lock of i_data_sem
*
* @orig_inode: original inode structure to be released its lock first
* @donor_inode: donor inode structure to be released its lock second
* Release write lock of i_data_sem of two inodes (orig and donor).
*/
void
ext4_double_up_write_data_sem(struct inode *orig_inode,
struct inode *donor_inode)
{
up_write(&EXT4_I(orig_inode)->i_data_sem);
up_write(&EXT4_I(donor_inode)->i_data_sem);
}
/**
* mext_check_coverage - Check that all extents in range has the same type
*
* @inode: inode in question
* @from: block offset of inode
* @count: block count to be checked
* @unwritten: extents expected to be unwritten
* @err: pointer to save error value
*
* Return 1 if all extents in range has expected type, and zero otherwise.
*/
static int
mext_check_coverage(struct inode *inode, ext4_lblk_t from, ext4_lblk_t count,
int unwritten, int *err)
{
struct ext4_ext_path *path = NULL;
struct ext4_extent *ext;
int ret = 0;
ext4_lblk_t last = from + count;
while (from < last) {
*err = get_ext_path(inode, from, &path);
if (*err)
goto out;
ext = path[ext_depth(inode)].p_ext;
if (unwritten != ext4_ext_is_unwritten(ext))
goto out;
from += ext4_ext_get_actual_len(ext);
ext4_ext_drop_refs(path);
}
ret = 1;
out:
ext4_ext_drop_refs(path);
kfree(path);
return ret;
}
ext4: clean up online defrag bugs in move_extent_per_page() Non-full list of bugs: 1) uninitialized extent optimization does not hold page's lock, and simply replace brunches after that writeback code goes crazy because block mapping changed under it's feets kernel BUG at fs/ext4/inode.c:1434! ( 288'th xfstress) 2) uninitialized extent may became initialized right after we drop i_data_sem, so extent state must be rechecked 3) Locked pages goes uptodate via following sequence: ->readpage(page); lock_page(page); use_that_page(page) But after readpage() one may invalidate it because it is uptodate and unlocked (reclaimer does that) As result kernel bug at include/linux/buffer_head.c:133! 4) We call write_begin() with already opened stansaction which result in following deadlock: ->move_extent_per_page() ->ext4_journal_start()-> hold journal transaction ->write_begin() ->ext4_da_write_begin() ->ext4_nonda_switch() ->writeback_inodes_sb_if_idle() --> will wait for journal_stop() 5) try_to_release_page() may fail and it does fail if one of page's bh was pinned by journal 6) If we about to change page's mapping we MUST hold it's lock during entire remapping procedure, this is true for both pages(original and donor one) Fixes: - Avoid (1) and (2) simply by temproraly drop uninitialized extent handling optimization, this will be reimplemented later. - Fix (3) by manually forcing page to uptodate state w/o dropping it's lock - Fix (4) by rearranging existing locking: from: journal_start(); ->write_begin to: write_begin(); journal_extend() - Fix (5) simply by checking retvalue - Fix (6) by locking both (original and donor one) pages during extent swap with help of mext_page_double_lock() Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2012-09-26 23:52:07 +07:00
/**
* mext_page_double_lock - Grab and lock pages on both @inode1 and @inode2
*
* @inode1: the inode structure
* @inode2: the inode structure
* @index1: page index
* @index2: page index
ext4: clean up online defrag bugs in move_extent_per_page() Non-full list of bugs: 1) uninitialized extent optimization does not hold page's lock, and simply replace brunches after that writeback code goes crazy because block mapping changed under it's feets kernel BUG at fs/ext4/inode.c:1434! ( 288'th xfstress) 2) uninitialized extent may became initialized right after we drop i_data_sem, so extent state must be rechecked 3) Locked pages goes uptodate via following sequence: ->readpage(page); lock_page(page); use_that_page(page) But after readpage() one may invalidate it because it is uptodate and unlocked (reclaimer does that) As result kernel bug at include/linux/buffer_head.c:133! 4) We call write_begin() with already opened stansaction which result in following deadlock: ->move_extent_per_page() ->ext4_journal_start()-> hold journal transaction ->write_begin() ->ext4_da_write_begin() ->ext4_nonda_switch() ->writeback_inodes_sb_if_idle() --> will wait for journal_stop() 5) try_to_release_page() may fail and it does fail if one of page's bh was pinned by journal 6) If we about to change page's mapping we MUST hold it's lock during entire remapping procedure, this is true for both pages(original and donor one) Fixes: - Avoid (1) and (2) simply by temproraly drop uninitialized extent handling optimization, this will be reimplemented later. - Fix (3) by manually forcing page to uptodate state w/o dropping it's lock - Fix (4) by rearranging existing locking: from: journal_start(); ->write_begin to: write_begin(); journal_extend() - Fix (5) simply by checking retvalue - Fix (6) by locking both (original and donor one) pages during extent swap with help of mext_page_double_lock() Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2012-09-26 23:52:07 +07:00
* @page: result page vector
*
* Grab two locked pages for inode's by inode order
*/
static int
mext_page_double_lock(struct inode *inode1, struct inode *inode2,
pgoff_t index1, pgoff_t index2, struct page *page[2])
ext4: clean up online defrag bugs in move_extent_per_page() Non-full list of bugs: 1) uninitialized extent optimization does not hold page's lock, and simply replace brunches after that writeback code goes crazy because block mapping changed under it's feets kernel BUG at fs/ext4/inode.c:1434! ( 288'th xfstress) 2) uninitialized extent may became initialized right after we drop i_data_sem, so extent state must be rechecked 3) Locked pages goes uptodate via following sequence: ->readpage(page); lock_page(page); use_that_page(page) But after readpage() one may invalidate it because it is uptodate and unlocked (reclaimer does that) As result kernel bug at include/linux/buffer_head.c:133! 4) We call write_begin() with already opened stansaction which result in following deadlock: ->move_extent_per_page() ->ext4_journal_start()-> hold journal transaction ->write_begin() ->ext4_da_write_begin() ->ext4_nonda_switch() ->writeback_inodes_sb_if_idle() --> will wait for journal_stop() 5) try_to_release_page() may fail and it does fail if one of page's bh was pinned by journal 6) If we about to change page's mapping we MUST hold it's lock during entire remapping procedure, this is true for both pages(original and donor one) Fixes: - Avoid (1) and (2) simply by temproraly drop uninitialized extent handling optimization, this will be reimplemented later. - Fix (3) by manually forcing page to uptodate state w/o dropping it's lock - Fix (4) by rearranging existing locking: from: journal_start(); ->write_begin to: write_begin(); journal_extend() - Fix (5) simply by checking retvalue - Fix (6) by locking both (original and donor one) pages during extent swap with help of mext_page_double_lock() Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2012-09-26 23:52:07 +07:00
{
struct address_space *mapping[2];
unsigned fl = AOP_FLAG_NOFS;
BUG_ON(!inode1 || !inode2);
if (inode1 < inode2) {
mapping[0] = inode1->i_mapping;
mapping[1] = inode2->i_mapping;
} else {
swap(index1, index2);
ext4: clean up online defrag bugs in move_extent_per_page() Non-full list of bugs: 1) uninitialized extent optimization does not hold page's lock, and simply replace brunches after that writeback code goes crazy because block mapping changed under it's feets kernel BUG at fs/ext4/inode.c:1434! ( 288'th xfstress) 2) uninitialized extent may became initialized right after we drop i_data_sem, so extent state must be rechecked 3) Locked pages goes uptodate via following sequence: ->readpage(page); lock_page(page); use_that_page(page) But after readpage() one may invalidate it because it is uptodate and unlocked (reclaimer does that) As result kernel bug at include/linux/buffer_head.c:133! 4) We call write_begin() with already opened stansaction which result in following deadlock: ->move_extent_per_page() ->ext4_journal_start()-> hold journal transaction ->write_begin() ->ext4_da_write_begin() ->ext4_nonda_switch() ->writeback_inodes_sb_if_idle() --> will wait for journal_stop() 5) try_to_release_page() may fail and it does fail if one of page's bh was pinned by journal 6) If we about to change page's mapping we MUST hold it's lock during entire remapping procedure, this is true for both pages(original and donor one) Fixes: - Avoid (1) and (2) simply by temproraly drop uninitialized extent handling optimization, this will be reimplemented later. - Fix (3) by manually forcing page to uptodate state w/o dropping it's lock - Fix (4) by rearranging existing locking: from: journal_start(); ->write_begin to: write_begin(); journal_extend() - Fix (5) simply by checking retvalue - Fix (6) by locking both (original and donor one) pages during extent swap with help of mext_page_double_lock() Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2012-09-26 23:52:07 +07:00
mapping[0] = inode2->i_mapping;
mapping[1] = inode1->i_mapping;
}
page[0] = grab_cache_page_write_begin(mapping[0], index1, fl);
ext4: clean up online defrag bugs in move_extent_per_page() Non-full list of bugs: 1) uninitialized extent optimization does not hold page's lock, and simply replace brunches after that writeback code goes crazy because block mapping changed under it's feets kernel BUG at fs/ext4/inode.c:1434! ( 288'th xfstress) 2) uninitialized extent may became initialized right after we drop i_data_sem, so extent state must be rechecked 3) Locked pages goes uptodate via following sequence: ->readpage(page); lock_page(page); use_that_page(page) But after readpage() one may invalidate it because it is uptodate and unlocked (reclaimer does that) As result kernel bug at include/linux/buffer_head.c:133! 4) We call write_begin() with already opened stansaction which result in following deadlock: ->move_extent_per_page() ->ext4_journal_start()-> hold journal transaction ->write_begin() ->ext4_da_write_begin() ->ext4_nonda_switch() ->writeback_inodes_sb_if_idle() --> will wait for journal_stop() 5) try_to_release_page() may fail and it does fail if one of page's bh was pinned by journal 6) If we about to change page's mapping we MUST hold it's lock during entire remapping procedure, this is true for both pages(original and donor one) Fixes: - Avoid (1) and (2) simply by temproraly drop uninitialized extent handling optimization, this will be reimplemented later. - Fix (3) by manually forcing page to uptodate state w/o dropping it's lock - Fix (4) by rearranging existing locking: from: journal_start(); ->write_begin to: write_begin(); journal_extend() - Fix (5) simply by checking retvalue - Fix (6) by locking both (original and donor one) pages during extent swap with help of mext_page_double_lock() Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2012-09-26 23:52:07 +07:00
if (!page[0])
return -ENOMEM;
page[1] = grab_cache_page_write_begin(mapping[1], index2, fl);
ext4: clean up online defrag bugs in move_extent_per_page() Non-full list of bugs: 1) uninitialized extent optimization does not hold page's lock, and simply replace brunches after that writeback code goes crazy because block mapping changed under it's feets kernel BUG at fs/ext4/inode.c:1434! ( 288'th xfstress) 2) uninitialized extent may became initialized right after we drop i_data_sem, so extent state must be rechecked 3) Locked pages goes uptodate via following sequence: ->readpage(page); lock_page(page); use_that_page(page) But after readpage() one may invalidate it because it is uptodate and unlocked (reclaimer does that) As result kernel bug at include/linux/buffer_head.c:133! 4) We call write_begin() with already opened stansaction which result in following deadlock: ->move_extent_per_page() ->ext4_journal_start()-> hold journal transaction ->write_begin() ->ext4_da_write_begin() ->ext4_nonda_switch() ->writeback_inodes_sb_if_idle() --> will wait for journal_stop() 5) try_to_release_page() may fail and it does fail if one of page's bh was pinned by journal 6) If we about to change page's mapping we MUST hold it's lock during entire remapping procedure, this is true for both pages(original and donor one) Fixes: - Avoid (1) and (2) simply by temproraly drop uninitialized extent handling optimization, this will be reimplemented later. - Fix (3) by manually forcing page to uptodate state w/o dropping it's lock - Fix (4) by rearranging existing locking: from: journal_start(); ->write_begin to: write_begin(); journal_extend() - Fix (5) simply by checking retvalue - Fix (6) by locking both (original and donor one) pages during extent swap with help of mext_page_double_lock() Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2012-09-26 23:52:07 +07:00
if (!page[1]) {
unlock_page(page[0]);
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time ago with promise that one day it will be possible to implement page cache with bigger chunks than PAGE_SIZE. This promise never materialized. And unlikely will. We have many places where PAGE_CACHE_SIZE assumed to be equal to PAGE_SIZE. And it's constant source of confusion on whether PAGE_CACHE_* or PAGE_* constant should be used in a particular case, especially on the border between fs and mm. Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much breakage to be doable. Let's stop pretending that pages in page cache are special. They are not. The changes are pretty straight-forward: - <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN}; - page_cache_get() -> get_page(); - page_cache_release() -> put_page(); This patch contains automated changes generated with coccinelle using script below. For some reason, coccinelle doesn't patch header files. I've called spatch for them manually. The only adjustment after coccinelle is revert of changes to PAGE_CAHCE_ALIGN definition: we are going to drop it later. There are few places in the code where coccinelle didn't reach. I'll fix them manually in a separate patch. Comments and documentation also will be addressed with the separate patch. virtual patch @@ expression E; @@ - E << (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ expression E; @@ - E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ @@ - PAGE_CACHE_SHIFT + PAGE_SHIFT @@ @@ - PAGE_CACHE_SIZE + PAGE_SIZE @@ @@ - PAGE_CACHE_MASK + PAGE_MASK @@ expression E; @@ - PAGE_CACHE_ALIGN(E) + PAGE_ALIGN(E) @@ expression E; @@ - page_cache_get(E) + get_page(E) @@ expression E; @@ - page_cache_release(E) + put_page(E) Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 19:29:47 +07:00
put_page(page[0]);
ext4: clean up online defrag bugs in move_extent_per_page() Non-full list of bugs: 1) uninitialized extent optimization does not hold page's lock, and simply replace brunches after that writeback code goes crazy because block mapping changed under it's feets kernel BUG at fs/ext4/inode.c:1434! ( 288'th xfstress) 2) uninitialized extent may became initialized right after we drop i_data_sem, so extent state must be rechecked 3) Locked pages goes uptodate via following sequence: ->readpage(page); lock_page(page); use_that_page(page) But after readpage() one may invalidate it because it is uptodate and unlocked (reclaimer does that) As result kernel bug at include/linux/buffer_head.c:133! 4) We call write_begin() with already opened stansaction which result in following deadlock: ->move_extent_per_page() ->ext4_journal_start()-> hold journal transaction ->write_begin() ->ext4_da_write_begin() ->ext4_nonda_switch() ->writeback_inodes_sb_if_idle() --> will wait for journal_stop() 5) try_to_release_page() may fail and it does fail if one of page's bh was pinned by journal 6) If we about to change page's mapping we MUST hold it's lock during entire remapping procedure, this is true for both pages(original and donor one) Fixes: - Avoid (1) and (2) simply by temproraly drop uninitialized extent handling optimization, this will be reimplemented later. - Fix (3) by manually forcing page to uptodate state w/o dropping it's lock - Fix (4) by rearranging existing locking: from: journal_start(); ->write_begin to: write_begin(); journal_extend() - Fix (5) simply by checking retvalue - Fix (6) by locking both (original and donor one) pages during extent swap with help of mext_page_double_lock() Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2012-09-26 23:52:07 +07:00
return -ENOMEM;
}
/*
* grab_cache_page_write_begin() may not wait on page's writeback if
* BDI not demand that. But it is reasonable to be very conservative
* here and explicitly wait on page's writeback
*/
wait_on_page_writeback(page[0]);
wait_on_page_writeback(page[1]);
if (inode1 > inode2)
swap(page[0], page[1]);
ext4: clean up online defrag bugs in move_extent_per_page() Non-full list of bugs: 1) uninitialized extent optimization does not hold page's lock, and simply replace brunches after that writeback code goes crazy because block mapping changed under it's feets kernel BUG at fs/ext4/inode.c:1434! ( 288'th xfstress) 2) uninitialized extent may became initialized right after we drop i_data_sem, so extent state must be rechecked 3) Locked pages goes uptodate via following sequence: ->readpage(page); lock_page(page); use_that_page(page) But after readpage() one may invalidate it because it is uptodate and unlocked (reclaimer does that) As result kernel bug at include/linux/buffer_head.c:133! 4) We call write_begin() with already opened stansaction which result in following deadlock: ->move_extent_per_page() ->ext4_journal_start()-> hold journal transaction ->write_begin() ->ext4_da_write_begin() ->ext4_nonda_switch() ->writeback_inodes_sb_if_idle() --> will wait for journal_stop() 5) try_to_release_page() may fail and it does fail if one of page's bh was pinned by journal 6) If we about to change page's mapping we MUST hold it's lock during entire remapping procedure, this is true for both pages(original and donor one) Fixes: - Avoid (1) and (2) simply by temproraly drop uninitialized extent handling optimization, this will be reimplemented later. - Fix (3) by manually forcing page to uptodate state w/o dropping it's lock - Fix (4) by rearranging existing locking: from: journal_start(); ->write_begin to: write_begin(); journal_extend() - Fix (5) simply by checking retvalue - Fix (6) by locking both (original and donor one) pages during extent swap with help of mext_page_double_lock() Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2012-09-26 23:52:07 +07:00
return 0;
}
/* Force page buffers uptodate w/o dropping page's lock */
static int
mext_page_mkuptodate(struct page *page, unsigned from, unsigned to)
{
struct inode *inode = page->mapping->host;
sector_t block;
struct buffer_head *bh, *head, *arr[MAX_BUF_PER_PAGE];
unsigned int blocksize, block_start, block_end;
int i, err, nr = 0, partial = 0;
BUG_ON(!PageLocked(page));
BUG_ON(PageWriteback(page));
if (PageUptodate(page))
return 0;
blocksize = i_blocksize(inode);
ext4: clean up online defrag bugs in move_extent_per_page() Non-full list of bugs: 1) uninitialized extent optimization does not hold page's lock, and simply replace brunches after that writeback code goes crazy because block mapping changed under it's feets kernel BUG at fs/ext4/inode.c:1434! ( 288'th xfstress) 2) uninitialized extent may became initialized right after we drop i_data_sem, so extent state must be rechecked 3) Locked pages goes uptodate via following sequence: ->readpage(page); lock_page(page); use_that_page(page) But after readpage() one may invalidate it because it is uptodate and unlocked (reclaimer does that) As result kernel bug at include/linux/buffer_head.c:133! 4) We call write_begin() with already opened stansaction which result in following deadlock: ->move_extent_per_page() ->ext4_journal_start()-> hold journal transaction ->write_begin() ->ext4_da_write_begin() ->ext4_nonda_switch() ->writeback_inodes_sb_if_idle() --> will wait for journal_stop() 5) try_to_release_page() may fail and it does fail if one of page's bh was pinned by journal 6) If we about to change page's mapping we MUST hold it's lock during entire remapping procedure, this is true for both pages(original and donor one) Fixes: - Avoid (1) and (2) simply by temproraly drop uninitialized extent handling optimization, this will be reimplemented later. - Fix (3) by manually forcing page to uptodate state w/o dropping it's lock - Fix (4) by rearranging existing locking: from: journal_start(); ->write_begin to: write_begin(); journal_extend() - Fix (5) simply by checking retvalue - Fix (6) by locking both (original and donor one) pages during extent swap with help of mext_page_double_lock() Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2012-09-26 23:52:07 +07:00
if (!page_has_buffers(page))
create_empty_buffers(page, blocksize, 0);
head = page_buffers(page);
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time ago with promise that one day it will be possible to implement page cache with bigger chunks than PAGE_SIZE. This promise never materialized. And unlikely will. We have many places where PAGE_CACHE_SIZE assumed to be equal to PAGE_SIZE. And it's constant source of confusion on whether PAGE_CACHE_* or PAGE_* constant should be used in a particular case, especially on the border between fs and mm. Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much breakage to be doable. Let's stop pretending that pages in page cache are special. They are not. The changes are pretty straight-forward: - <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN}; - page_cache_get() -> get_page(); - page_cache_release() -> put_page(); This patch contains automated changes generated with coccinelle using script below. For some reason, coccinelle doesn't patch header files. I've called spatch for them manually. The only adjustment after coccinelle is revert of changes to PAGE_CAHCE_ALIGN definition: we are going to drop it later. There are few places in the code where coccinelle didn't reach. I'll fix them manually in a separate patch. Comments and documentation also will be addressed with the separate patch. virtual patch @@ expression E; @@ - E << (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ expression E; @@ - E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ @@ - PAGE_CACHE_SHIFT + PAGE_SHIFT @@ @@ - PAGE_CACHE_SIZE + PAGE_SIZE @@ @@ - PAGE_CACHE_MASK + PAGE_MASK @@ expression E; @@ - PAGE_CACHE_ALIGN(E) + PAGE_ALIGN(E) @@ expression E; @@ - page_cache_get(E) + get_page(E) @@ expression E; @@ - page_cache_release(E) + put_page(E) Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 19:29:47 +07:00
block = (sector_t)page->index << (PAGE_SHIFT - inode->i_blkbits);
ext4: clean up online defrag bugs in move_extent_per_page() Non-full list of bugs: 1) uninitialized extent optimization does not hold page's lock, and simply replace brunches after that writeback code goes crazy because block mapping changed under it's feets kernel BUG at fs/ext4/inode.c:1434! ( 288'th xfstress) 2) uninitialized extent may became initialized right after we drop i_data_sem, so extent state must be rechecked 3) Locked pages goes uptodate via following sequence: ->readpage(page); lock_page(page); use_that_page(page) But after readpage() one may invalidate it because it is uptodate and unlocked (reclaimer does that) As result kernel bug at include/linux/buffer_head.c:133! 4) We call write_begin() with already opened stansaction which result in following deadlock: ->move_extent_per_page() ->ext4_journal_start()-> hold journal transaction ->write_begin() ->ext4_da_write_begin() ->ext4_nonda_switch() ->writeback_inodes_sb_if_idle() --> will wait for journal_stop() 5) try_to_release_page() may fail and it does fail if one of page's bh was pinned by journal 6) If we about to change page's mapping we MUST hold it's lock during entire remapping procedure, this is true for both pages(original and donor one) Fixes: - Avoid (1) and (2) simply by temproraly drop uninitialized extent handling optimization, this will be reimplemented later. - Fix (3) by manually forcing page to uptodate state w/o dropping it's lock - Fix (4) by rearranging existing locking: from: journal_start(); ->write_begin to: write_begin(); journal_extend() - Fix (5) simply by checking retvalue - Fix (6) by locking both (original and donor one) pages during extent swap with help of mext_page_double_lock() Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2012-09-26 23:52:07 +07:00
for (bh = head, block_start = 0; bh != head || !block_start;
block++, block_start = block_end, bh = bh->b_this_page) {
block_end = block_start + blocksize;
if (block_end <= from || block_start >= to) {
if (!buffer_uptodate(bh))
partial = 1;
continue;
}
if (buffer_uptodate(bh))
continue;
if (!buffer_mapped(bh)) {
err = ext4_get_block(inode, block, bh, 0);
if (err) {
SetPageError(page);
return err;
}
if (!buffer_mapped(bh)) {
zero_user(page, block_start, blocksize);
set_buffer_uptodate(bh);
ext4: clean up online defrag bugs in move_extent_per_page() Non-full list of bugs: 1) uninitialized extent optimization does not hold page's lock, and simply replace brunches after that writeback code goes crazy because block mapping changed under it's feets kernel BUG at fs/ext4/inode.c:1434! ( 288'th xfstress) 2) uninitialized extent may became initialized right after we drop i_data_sem, so extent state must be rechecked 3) Locked pages goes uptodate via following sequence: ->readpage(page); lock_page(page); use_that_page(page) But after readpage() one may invalidate it because it is uptodate and unlocked (reclaimer does that) As result kernel bug at include/linux/buffer_head.c:133! 4) We call write_begin() with already opened stansaction which result in following deadlock: ->move_extent_per_page() ->ext4_journal_start()-> hold journal transaction ->write_begin() ->ext4_da_write_begin() ->ext4_nonda_switch() ->writeback_inodes_sb_if_idle() --> will wait for journal_stop() 5) try_to_release_page() may fail and it does fail if one of page's bh was pinned by journal 6) If we about to change page's mapping we MUST hold it's lock during entire remapping procedure, this is true for both pages(original and donor one) Fixes: - Avoid (1) and (2) simply by temproraly drop uninitialized extent handling optimization, this will be reimplemented later. - Fix (3) by manually forcing page to uptodate state w/o dropping it's lock - Fix (4) by rearranging existing locking: from: journal_start(); ->write_begin to: write_begin(); journal_extend() - Fix (5) simply by checking retvalue - Fix (6) by locking both (original and donor one) pages during extent swap with help of mext_page_double_lock() Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2012-09-26 23:52:07 +07:00
continue;
}
}
BUG_ON(nr >= MAX_BUF_PER_PAGE);
arr[nr++] = bh;
}
/* No io required */
if (!nr)
goto out;
for (i = 0; i < nr; i++) {
bh = arr[i];
if (!bh_uptodate_or_lock(bh)) {
err = bh_submit_read(bh);
if (err)
return err;
}
}
out:
if (!partial)
SetPageUptodate(page);
return 0;
}
/**
* move_extent_per_page - Move extent data per page
*
* @o_filp: file structure of original file
* @donor_inode: donor inode
* @orig_page_offset: page index on original file
* @donor_page_offset: page index on donor file
* @data_offset_in_page: block index where data swapping starts
* @block_len_in_page: the number of blocks to be swapped
* @unwritten: orig extent is unwritten or not
* @err: pointer to save return value
*
* Save the data in original inode blocks and replace original inode extents
* with donor inode extents by calling ext4_swap_extents().
* Finally, write out the saved data in new original inode blocks. Return
* replaced block count.
*/
static int
move_extent_per_page(struct file *o_filp, struct inode *donor_inode,
pgoff_t orig_page_offset, pgoff_t donor_page_offset,
int data_offset_in_page,
int block_len_in_page, int unwritten, int *err)
{
struct inode *orig_inode = file_inode(o_filp);
ext4: clean up online defrag bugs in move_extent_per_page() Non-full list of bugs: 1) uninitialized extent optimization does not hold page's lock, and simply replace brunches after that writeback code goes crazy because block mapping changed under it's feets kernel BUG at fs/ext4/inode.c:1434! ( 288'th xfstress) 2) uninitialized extent may became initialized right after we drop i_data_sem, so extent state must be rechecked 3) Locked pages goes uptodate via following sequence: ->readpage(page); lock_page(page); use_that_page(page) But after readpage() one may invalidate it because it is uptodate and unlocked (reclaimer does that) As result kernel bug at include/linux/buffer_head.c:133! 4) We call write_begin() with already opened stansaction which result in following deadlock: ->move_extent_per_page() ->ext4_journal_start()-> hold journal transaction ->write_begin() ->ext4_da_write_begin() ->ext4_nonda_switch() ->writeback_inodes_sb_if_idle() --> will wait for journal_stop() 5) try_to_release_page() may fail and it does fail if one of page's bh was pinned by journal 6) If we about to change page's mapping we MUST hold it's lock during entire remapping procedure, this is true for both pages(original and donor one) Fixes: - Avoid (1) and (2) simply by temproraly drop uninitialized extent handling optimization, this will be reimplemented later. - Fix (3) by manually forcing page to uptodate state w/o dropping it's lock - Fix (4) by rearranging existing locking: from: journal_start(); ->write_begin to: write_begin(); journal_extend() - Fix (5) simply by checking retvalue - Fix (6) by locking both (original and donor one) pages during extent swap with help of mext_page_double_lock() Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2012-09-26 23:52:07 +07:00
struct page *pagep[2] = {NULL, NULL};
handle_t *handle;
ext4_lblk_t orig_blk_offset, donor_blk_offset;
unsigned long blocksize = orig_inode->i_sb->s_blocksize;
unsigned int tmp_data_size, data_size, replaced_size;
ext4: don't read blocks from disk after extents being swapped I notice ext4/307 fails occasionally on ppc64 host, reporting md5 checksum mismatch after moving data from original file to donor file. The reason is that move_extent_per_page() calls __block_write_begin() and block_commit_write() to write saved data from original inode blocks to donor inode blocks, but __block_write_begin() not only maps buffer heads but also reads block content from disk if the size is not block size aligned. At this time the physical block number in mapped buffer head is pointing to the donor file not the original file, and that results in reading wrong data to page, which get written to disk in following block_commit_write call. This also can be reproduced by the following script on 1k block size ext4 on x86_64 host: mnt=/mnt/ext4 donorfile=$mnt/donor testfile=$mnt/testfile e4compact=~/xfstests/src/e4compact rm -f $donorfile $testfile # reserve space for donor file, written by 0xaa and sync to disk to # avoid EBUSY on EXT4_IOC_MOVE_EXT xfs_io -fc "pwrite -S 0xaa 0 1m" -c "fsync" $donorfile # create test file written by 0xbb xfs_io -fc "pwrite -S 0xbb 0 1023" -c "fsync" $testfile # compute initial md5sum md5sum $testfile | tee md5sum.txt # drop cache, force e4compact to read data from disk echo 3 > /proc/sys/vm/drop_caches # test defrag echo "$testfile" | $e4compact -i -v -f $donorfile # check md5sum md5sum -c md5sum.txt Fix it by creating & mapping buffer heads only but not reading blocks from disk, because all the data in page is guaranteed to be up-to-date in mext_page_mkuptodate(). Cc: stable@vger.kernel.org Signed-off-by: Eryu Guan <guaneryu@gmail.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2016-02-12 13:20:43 +07:00
int i, err2, jblocks, retries = 0;
int replaced_count = 0;
ext4: clean up online defrag bugs in move_extent_per_page() Non-full list of bugs: 1) uninitialized extent optimization does not hold page's lock, and simply replace brunches after that writeback code goes crazy because block mapping changed under it's feets kernel BUG at fs/ext4/inode.c:1434! ( 288'th xfstress) 2) uninitialized extent may became initialized right after we drop i_data_sem, so extent state must be rechecked 3) Locked pages goes uptodate via following sequence: ->readpage(page); lock_page(page); use_that_page(page) But after readpage() one may invalidate it because it is uptodate and unlocked (reclaimer does that) As result kernel bug at include/linux/buffer_head.c:133! 4) We call write_begin() with already opened stansaction which result in following deadlock: ->move_extent_per_page() ->ext4_journal_start()-> hold journal transaction ->write_begin() ->ext4_da_write_begin() ->ext4_nonda_switch() ->writeback_inodes_sb_if_idle() --> will wait for journal_stop() 5) try_to_release_page() may fail and it does fail if one of page's bh was pinned by journal 6) If we about to change page's mapping we MUST hold it's lock during entire remapping procedure, this is true for both pages(original and donor one) Fixes: - Avoid (1) and (2) simply by temproraly drop uninitialized extent handling optimization, this will be reimplemented later. - Fix (3) by manually forcing page to uptodate state w/o dropping it's lock - Fix (4) by rearranging existing locking: from: journal_start(); ->write_begin to: write_begin(); journal_extend() - Fix (5) simply by checking retvalue - Fix (6) by locking both (original and donor one) pages during extent swap with help of mext_page_double_lock() Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2012-09-26 23:52:07 +07:00
int from = data_offset_in_page << orig_inode->i_blkbits;
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time ago with promise that one day it will be possible to implement page cache with bigger chunks than PAGE_SIZE. This promise never materialized. And unlikely will. We have many places where PAGE_CACHE_SIZE assumed to be equal to PAGE_SIZE. And it's constant source of confusion on whether PAGE_CACHE_* or PAGE_* constant should be used in a particular case, especially on the border between fs and mm. Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much breakage to be doable. Let's stop pretending that pages in page cache are special. They are not. The changes are pretty straight-forward: - <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN}; - page_cache_get() -> get_page(); - page_cache_release() -> put_page(); This patch contains automated changes generated with coccinelle using script below. For some reason, coccinelle doesn't patch header files. I've called spatch for them manually. The only adjustment after coccinelle is revert of changes to PAGE_CAHCE_ALIGN definition: we are going to drop it later. There are few places in the code where coccinelle didn't reach. I'll fix them manually in a separate patch. Comments and documentation also will be addressed with the separate patch. virtual patch @@ expression E; @@ - E << (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ expression E; @@ - E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ @@ - PAGE_CACHE_SHIFT + PAGE_SHIFT @@ @@ - PAGE_CACHE_SIZE + PAGE_SIZE @@ @@ - PAGE_CACHE_MASK + PAGE_MASK @@ expression E; @@ - PAGE_CACHE_ALIGN(E) + PAGE_ALIGN(E) @@ expression E; @@ - page_cache_get(E) + get_page(E) @@ expression E; @@ - page_cache_release(E) + put_page(E) Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 19:29:47 +07:00
int blocks_per_page = PAGE_SIZE >> orig_inode->i_blkbits;
struct super_block *sb = orig_inode->i_sb;
ext4: don't read blocks from disk after extents being swapped I notice ext4/307 fails occasionally on ppc64 host, reporting md5 checksum mismatch after moving data from original file to donor file. The reason is that move_extent_per_page() calls __block_write_begin() and block_commit_write() to write saved data from original inode blocks to donor inode blocks, but __block_write_begin() not only maps buffer heads but also reads block content from disk if the size is not block size aligned. At this time the physical block number in mapped buffer head is pointing to the donor file not the original file, and that results in reading wrong data to page, which get written to disk in following block_commit_write call. This also can be reproduced by the following script on 1k block size ext4 on x86_64 host: mnt=/mnt/ext4 donorfile=$mnt/donor testfile=$mnt/testfile e4compact=~/xfstests/src/e4compact rm -f $donorfile $testfile # reserve space for donor file, written by 0xaa and sync to disk to # avoid EBUSY on EXT4_IOC_MOVE_EXT xfs_io -fc "pwrite -S 0xaa 0 1m" -c "fsync" $donorfile # create test file written by 0xbb xfs_io -fc "pwrite -S 0xbb 0 1023" -c "fsync" $testfile # compute initial md5sum md5sum $testfile | tee md5sum.txt # drop cache, force e4compact to read data from disk echo 3 > /proc/sys/vm/drop_caches # test defrag echo "$testfile" | $e4compact -i -v -f $donorfile # check md5sum md5sum -c md5sum.txt Fix it by creating & mapping buffer heads only but not reading blocks from disk, because all the data in page is guaranteed to be up-to-date in mext_page_mkuptodate(). Cc: stable@vger.kernel.org Signed-off-by: Eryu Guan <guaneryu@gmail.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2016-02-12 13:20:43 +07:00
struct buffer_head *bh = NULL;
/*
* It needs twice the amount of ordinary journal buffers because
* inode and donor_inode may change each different metadata blocks.
*/
ext4: clean up online defrag bugs in move_extent_per_page() Non-full list of bugs: 1) uninitialized extent optimization does not hold page's lock, and simply replace brunches after that writeback code goes crazy because block mapping changed under it's feets kernel BUG at fs/ext4/inode.c:1434! ( 288'th xfstress) 2) uninitialized extent may became initialized right after we drop i_data_sem, so extent state must be rechecked 3) Locked pages goes uptodate via following sequence: ->readpage(page); lock_page(page); use_that_page(page) But after readpage() one may invalidate it because it is uptodate and unlocked (reclaimer does that) As result kernel bug at include/linux/buffer_head.c:133! 4) We call write_begin() with already opened stansaction which result in following deadlock: ->move_extent_per_page() ->ext4_journal_start()-> hold journal transaction ->write_begin() ->ext4_da_write_begin() ->ext4_nonda_switch() ->writeback_inodes_sb_if_idle() --> will wait for journal_stop() 5) try_to_release_page() may fail and it does fail if one of page's bh was pinned by journal 6) If we about to change page's mapping we MUST hold it's lock during entire remapping procedure, this is true for both pages(original and donor one) Fixes: - Avoid (1) and (2) simply by temproraly drop uninitialized extent handling optimization, this will be reimplemented later. - Fix (3) by manually forcing page to uptodate state w/o dropping it's lock - Fix (4) by rearranging existing locking: from: journal_start(); ->write_begin to: write_begin(); journal_extend() - Fix (5) simply by checking retvalue - Fix (6) by locking both (original and donor one) pages during extent swap with help of mext_page_double_lock() Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2012-09-26 23:52:07 +07:00
again:
*err = 0;
jblocks = ext4_writepage_trans_blocks(orig_inode) * 2;
handle = ext4_journal_start(orig_inode, EXT4_HT_MOVE_EXTENTS, jblocks);
if (IS_ERR(handle)) {
*err = PTR_ERR(handle);
return 0;
}
orig_blk_offset = orig_page_offset * blocks_per_page +
data_offset_in_page;
donor_blk_offset = donor_page_offset * blocks_per_page +
data_offset_in_page;
/* Calculate data_size */
if ((orig_blk_offset + block_len_in_page - 1) ==
((orig_inode->i_size - 1) >> orig_inode->i_blkbits)) {
/* Replace the last block */
tmp_data_size = orig_inode->i_size & (blocksize - 1);
/*
* If data_size equal zero, it shows data_size is multiples of
* blocksize. So we set appropriate value.
*/
if (tmp_data_size == 0)
tmp_data_size = blocksize;
data_size = tmp_data_size +
((block_len_in_page - 1) << orig_inode->i_blkbits);
} else
data_size = block_len_in_page << orig_inode->i_blkbits;
replaced_size = data_size;
ext4: clean up online defrag bugs in move_extent_per_page() Non-full list of bugs: 1) uninitialized extent optimization does not hold page's lock, and simply replace brunches after that writeback code goes crazy because block mapping changed under it's feets kernel BUG at fs/ext4/inode.c:1434! ( 288'th xfstress) 2) uninitialized extent may became initialized right after we drop i_data_sem, so extent state must be rechecked 3) Locked pages goes uptodate via following sequence: ->readpage(page); lock_page(page); use_that_page(page) But after readpage() one may invalidate it because it is uptodate and unlocked (reclaimer does that) As result kernel bug at include/linux/buffer_head.c:133! 4) We call write_begin() with already opened stansaction which result in following deadlock: ->move_extent_per_page() ->ext4_journal_start()-> hold journal transaction ->write_begin() ->ext4_da_write_begin() ->ext4_nonda_switch() ->writeback_inodes_sb_if_idle() --> will wait for journal_stop() 5) try_to_release_page() may fail and it does fail if one of page's bh was pinned by journal 6) If we about to change page's mapping we MUST hold it's lock during entire remapping procedure, this is true for both pages(original and donor one) Fixes: - Avoid (1) and (2) simply by temproraly drop uninitialized extent handling optimization, this will be reimplemented later. - Fix (3) by manually forcing page to uptodate state w/o dropping it's lock - Fix (4) by rearranging existing locking: from: journal_start(); ->write_begin to: write_begin(); journal_extend() - Fix (5) simply by checking retvalue - Fix (6) by locking both (original and donor one) pages during extent swap with help of mext_page_double_lock() Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2012-09-26 23:52:07 +07:00
*err = mext_page_double_lock(orig_inode, donor_inode, orig_page_offset,
donor_page_offset, pagep);
if (unlikely(*err < 0))
ext4: clean up online defrag bugs in move_extent_per_page() Non-full list of bugs: 1) uninitialized extent optimization does not hold page's lock, and simply replace brunches after that writeback code goes crazy because block mapping changed under it's feets kernel BUG at fs/ext4/inode.c:1434! ( 288'th xfstress) 2) uninitialized extent may became initialized right after we drop i_data_sem, so extent state must be rechecked 3) Locked pages goes uptodate via following sequence: ->readpage(page); lock_page(page); use_that_page(page) But after readpage() one may invalidate it because it is uptodate and unlocked (reclaimer does that) As result kernel bug at include/linux/buffer_head.c:133! 4) We call write_begin() with already opened stansaction which result in following deadlock: ->move_extent_per_page() ->ext4_journal_start()-> hold journal transaction ->write_begin() ->ext4_da_write_begin() ->ext4_nonda_switch() ->writeback_inodes_sb_if_idle() --> will wait for journal_stop() 5) try_to_release_page() may fail and it does fail if one of page's bh was pinned by journal 6) If we about to change page's mapping we MUST hold it's lock during entire remapping procedure, this is true for both pages(original and donor one) Fixes: - Avoid (1) and (2) simply by temproraly drop uninitialized extent handling optimization, this will be reimplemented later. - Fix (3) by manually forcing page to uptodate state w/o dropping it's lock - Fix (4) by rearranging existing locking: from: journal_start(); ->write_begin to: write_begin(); journal_extend() - Fix (5) simply by checking retvalue - Fix (6) by locking both (original and donor one) pages during extent swap with help of mext_page_double_lock() Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2012-09-26 23:52:07 +07:00
goto stop_journal;
/*
* If orig extent was unwritten it can become initialized
* at any time after i_data_sem was dropped, in order to
* serialize with delalloc we have recheck extent while we
* hold page's lock, if it is still the case data copy is not
* necessary, just swap data blocks between orig and donor.
*/
if (unwritten) {
ext4_double_down_write_data_sem(orig_inode, donor_inode);
/* If any of extents in range became initialized we have to
* fallback to data copying */
unwritten = mext_check_coverage(orig_inode, orig_blk_offset,
block_len_in_page, 1, err);
if (*err)
goto drop_data_sem;
unwritten &= mext_check_coverage(donor_inode, donor_blk_offset,
block_len_in_page, 1, err);
if (*err)
goto drop_data_sem;
if (!unwritten) {
ext4_double_up_write_data_sem(orig_inode, donor_inode);
goto data_copy;
}
if ((page_has_private(pagep[0]) &&
!try_to_release_page(pagep[0], 0)) ||
(page_has_private(pagep[1]) &&
!try_to_release_page(pagep[1], 0))) {
*err = -EBUSY;
goto drop_data_sem;
}
replaced_count = ext4_swap_extents(handle, orig_inode,
donor_inode, orig_blk_offset,
donor_blk_offset,
block_len_in_page, 1, err);
drop_data_sem:
ext4_double_up_write_data_sem(orig_inode, donor_inode);
goto unlock_pages;
}
data_copy:
ext4: clean up online defrag bugs in move_extent_per_page() Non-full list of bugs: 1) uninitialized extent optimization does not hold page's lock, and simply replace brunches after that writeback code goes crazy because block mapping changed under it's feets kernel BUG at fs/ext4/inode.c:1434! ( 288'th xfstress) 2) uninitialized extent may became initialized right after we drop i_data_sem, so extent state must be rechecked 3) Locked pages goes uptodate via following sequence: ->readpage(page); lock_page(page); use_that_page(page) But after readpage() one may invalidate it because it is uptodate and unlocked (reclaimer does that) As result kernel bug at include/linux/buffer_head.c:133! 4) We call write_begin() with already opened stansaction which result in following deadlock: ->move_extent_per_page() ->ext4_journal_start()-> hold journal transaction ->write_begin() ->ext4_da_write_begin() ->ext4_nonda_switch() ->writeback_inodes_sb_if_idle() --> will wait for journal_stop() 5) try_to_release_page() may fail and it does fail if one of page's bh was pinned by journal 6) If we about to change page's mapping we MUST hold it's lock during entire remapping procedure, this is true for both pages(original and donor one) Fixes: - Avoid (1) and (2) simply by temproraly drop uninitialized extent handling optimization, this will be reimplemented later. - Fix (3) by manually forcing page to uptodate state w/o dropping it's lock - Fix (4) by rearranging existing locking: from: journal_start(); ->write_begin to: write_begin(); journal_extend() - Fix (5) simply by checking retvalue - Fix (6) by locking both (original and donor one) pages during extent swap with help of mext_page_double_lock() Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2012-09-26 23:52:07 +07:00
*err = mext_page_mkuptodate(pagep[0], from, from + replaced_size);
if (*err)
goto unlock_pages;
/* At this point all buffers in range are uptodate, old mapping layout
* is no longer required, try to drop it now. */
if ((page_has_private(pagep[0]) && !try_to_release_page(pagep[0], 0)) ||
(page_has_private(pagep[1]) && !try_to_release_page(pagep[1], 0))) {
*err = -EBUSY;
goto unlock_pages;
}
ext4_double_down_write_data_sem(orig_inode, donor_inode);
replaced_count = ext4_swap_extents(handle, orig_inode, donor_inode,
orig_blk_offset, donor_blk_offset,
block_len_in_page, 1, err);
ext4_double_up_write_data_sem(orig_inode, donor_inode);
ext4: clean up online defrag bugs in move_extent_per_page() Non-full list of bugs: 1) uninitialized extent optimization does not hold page's lock, and simply replace brunches after that writeback code goes crazy because block mapping changed under it's feets kernel BUG at fs/ext4/inode.c:1434! ( 288'th xfstress) 2) uninitialized extent may became initialized right after we drop i_data_sem, so extent state must be rechecked 3) Locked pages goes uptodate via following sequence: ->readpage(page); lock_page(page); use_that_page(page) But after readpage() one may invalidate it because it is uptodate and unlocked (reclaimer does that) As result kernel bug at include/linux/buffer_head.c:133! 4) We call write_begin() with already opened stansaction which result in following deadlock: ->move_extent_per_page() ->ext4_journal_start()-> hold journal transaction ->write_begin() ->ext4_da_write_begin() ->ext4_nonda_switch() ->writeback_inodes_sb_if_idle() --> will wait for journal_stop() 5) try_to_release_page() may fail and it does fail if one of page's bh was pinned by journal 6) If we about to change page's mapping we MUST hold it's lock during entire remapping procedure, this is true for both pages(original and donor one) Fixes: - Avoid (1) and (2) simply by temproraly drop uninitialized extent handling optimization, this will be reimplemented later. - Fix (3) by manually forcing page to uptodate state w/o dropping it's lock - Fix (4) by rearranging existing locking: from: journal_start(); ->write_begin to: write_begin(); journal_extend() - Fix (5) simply by checking retvalue - Fix (6) by locking both (original and donor one) pages during extent swap with help of mext_page_double_lock() Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2012-09-26 23:52:07 +07:00
if (*err) {
if (replaced_count) {
block_len_in_page = replaced_count;
replaced_size =
block_len_in_page << orig_inode->i_blkbits;
} else
ext4: clean up online defrag bugs in move_extent_per_page() Non-full list of bugs: 1) uninitialized extent optimization does not hold page's lock, and simply replace brunches after that writeback code goes crazy because block mapping changed under it's feets kernel BUG at fs/ext4/inode.c:1434! ( 288'th xfstress) 2) uninitialized extent may became initialized right after we drop i_data_sem, so extent state must be rechecked 3) Locked pages goes uptodate via following sequence: ->readpage(page); lock_page(page); use_that_page(page) But after readpage() one may invalidate it because it is uptodate and unlocked (reclaimer does that) As result kernel bug at include/linux/buffer_head.c:133! 4) We call write_begin() with already opened stansaction which result in following deadlock: ->move_extent_per_page() ->ext4_journal_start()-> hold journal transaction ->write_begin() ->ext4_da_write_begin() ->ext4_nonda_switch() ->writeback_inodes_sb_if_idle() --> will wait for journal_stop() 5) try_to_release_page() may fail and it does fail if one of page's bh was pinned by journal 6) If we about to change page's mapping we MUST hold it's lock during entire remapping procedure, this is true for both pages(original and donor one) Fixes: - Avoid (1) and (2) simply by temproraly drop uninitialized extent handling optimization, this will be reimplemented later. - Fix (3) by manually forcing page to uptodate state w/o dropping it's lock - Fix (4) by rearranging existing locking: from: journal_start(); ->write_begin to: write_begin(); journal_extend() - Fix (5) simply by checking retvalue - Fix (6) by locking both (original and donor one) pages during extent swap with help of mext_page_double_lock() Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2012-09-26 23:52:07 +07:00
goto unlock_pages;
}
ext4: clean up online defrag bugs in move_extent_per_page() Non-full list of bugs: 1) uninitialized extent optimization does not hold page's lock, and simply replace brunches after that writeback code goes crazy because block mapping changed under it's feets kernel BUG at fs/ext4/inode.c:1434! ( 288'th xfstress) 2) uninitialized extent may became initialized right after we drop i_data_sem, so extent state must be rechecked 3) Locked pages goes uptodate via following sequence: ->readpage(page); lock_page(page); use_that_page(page) But after readpage() one may invalidate it because it is uptodate and unlocked (reclaimer does that) As result kernel bug at include/linux/buffer_head.c:133! 4) We call write_begin() with already opened stansaction which result in following deadlock: ->move_extent_per_page() ->ext4_journal_start()-> hold journal transaction ->write_begin() ->ext4_da_write_begin() ->ext4_nonda_switch() ->writeback_inodes_sb_if_idle() --> will wait for journal_stop() 5) try_to_release_page() may fail and it does fail if one of page's bh was pinned by journal 6) If we about to change page's mapping we MUST hold it's lock during entire remapping procedure, this is true for both pages(original and donor one) Fixes: - Avoid (1) and (2) simply by temproraly drop uninitialized extent handling optimization, this will be reimplemented later. - Fix (3) by manually forcing page to uptodate state w/o dropping it's lock - Fix (4) by rearranging existing locking: from: journal_start(); ->write_begin to: write_begin(); journal_extend() - Fix (5) simply by checking retvalue - Fix (6) by locking both (original and donor one) pages during extent swap with help of mext_page_double_lock() Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2012-09-26 23:52:07 +07:00
/* Perform all necessary steps similar write_begin()/write_end()
* but keeping in mind that i_size will not change */
ext4: don't read blocks from disk after extents being swapped I notice ext4/307 fails occasionally on ppc64 host, reporting md5 checksum mismatch after moving data from original file to donor file. The reason is that move_extent_per_page() calls __block_write_begin() and block_commit_write() to write saved data from original inode blocks to donor inode blocks, but __block_write_begin() not only maps buffer heads but also reads block content from disk if the size is not block size aligned. At this time the physical block number in mapped buffer head is pointing to the donor file not the original file, and that results in reading wrong data to page, which get written to disk in following block_commit_write call. This also can be reproduced by the following script on 1k block size ext4 on x86_64 host: mnt=/mnt/ext4 donorfile=$mnt/donor testfile=$mnt/testfile e4compact=~/xfstests/src/e4compact rm -f $donorfile $testfile # reserve space for donor file, written by 0xaa and sync to disk to # avoid EBUSY on EXT4_IOC_MOVE_EXT xfs_io -fc "pwrite -S 0xaa 0 1m" -c "fsync" $donorfile # create test file written by 0xbb xfs_io -fc "pwrite -S 0xbb 0 1023" -c "fsync" $testfile # compute initial md5sum md5sum $testfile | tee md5sum.txt # drop cache, force e4compact to read data from disk echo 3 > /proc/sys/vm/drop_caches # test defrag echo "$testfile" | $e4compact -i -v -f $donorfile # check md5sum md5sum -c md5sum.txt Fix it by creating & mapping buffer heads only but not reading blocks from disk, because all the data in page is guaranteed to be up-to-date in mext_page_mkuptodate(). Cc: stable@vger.kernel.org Signed-off-by: Eryu Guan <guaneryu@gmail.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2016-02-12 13:20:43 +07:00
if (!page_has_buffers(pagep[0]))
create_empty_buffers(pagep[0], 1 << orig_inode->i_blkbits, 0);
bh = page_buffers(pagep[0]);
for (i = 0; i < data_offset_in_page; i++)
bh = bh->b_this_page;
for (i = 0; i < block_len_in_page; i++) {
*err = ext4_get_block(orig_inode, orig_blk_offset + i, bh, 0);
if (*err < 0)
break;
bh = bh->b_this_page;
ext4: don't read blocks from disk after extents being swapped I notice ext4/307 fails occasionally on ppc64 host, reporting md5 checksum mismatch after moving data from original file to donor file. The reason is that move_extent_per_page() calls __block_write_begin() and block_commit_write() to write saved data from original inode blocks to donor inode blocks, but __block_write_begin() not only maps buffer heads but also reads block content from disk if the size is not block size aligned. At this time the physical block number in mapped buffer head is pointing to the donor file not the original file, and that results in reading wrong data to page, which get written to disk in following block_commit_write call. This also can be reproduced by the following script on 1k block size ext4 on x86_64 host: mnt=/mnt/ext4 donorfile=$mnt/donor testfile=$mnt/testfile e4compact=~/xfstests/src/e4compact rm -f $donorfile $testfile # reserve space for donor file, written by 0xaa and sync to disk to # avoid EBUSY on EXT4_IOC_MOVE_EXT xfs_io -fc "pwrite -S 0xaa 0 1m" -c "fsync" $donorfile # create test file written by 0xbb xfs_io -fc "pwrite -S 0xbb 0 1023" -c "fsync" $testfile # compute initial md5sum md5sum $testfile | tee md5sum.txt # drop cache, force e4compact to read data from disk echo 3 > /proc/sys/vm/drop_caches # test defrag echo "$testfile" | $e4compact -i -v -f $donorfile # check md5sum md5sum -c md5sum.txt Fix it by creating & mapping buffer heads only but not reading blocks from disk, because all the data in page is guaranteed to be up-to-date in mext_page_mkuptodate(). Cc: stable@vger.kernel.org Signed-off-by: Eryu Guan <guaneryu@gmail.com> Signed-off-by: Theodore Ts'o <tytso@mit.edu>
2016-02-12 13:20:43 +07:00
}
ext4: clean up online defrag bugs in move_extent_per_page() Non-full list of bugs: 1) uninitialized extent optimization does not hold page's lock, and simply replace brunches after that writeback code goes crazy because block mapping changed under it's feets kernel BUG at fs/ext4/inode.c:1434! ( 288'th xfstress) 2) uninitialized extent may became initialized right after we drop i_data_sem, so extent state must be rechecked 3) Locked pages goes uptodate via following sequence: ->readpage(page); lock_page(page); use_that_page(page) But after readpage() one may invalidate it because it is uptodate and unlocked (reclaimer does that) As result kernel bug at include/linux/buffer_head.c:133! 4) We call write_begin() with already opened stansaction which result in following deadlock: ->move_extent_per_page() ->ext4_journal_start()-> hold journal transaction ->write_begin() ->ext4_da_write_begin() ->ext4_nonda_switch() ->writeback_inodes_sb_if_idle() --> will wait for journal_stop() 5) try_to_release_page() may fail and it does fail if one of page's bh was pinned by journal 6) If we about to change page's mapping we MUST hold it's lock during entire remapping procedure, this is true for both pages(original and donor one) Fixes: - Avoid (1) and (2) simply by temproraly drop uninitialized extent handling optimization, this will be reimplemented later. - Fix (3) by manually forcing page to uptodate state w/o dropping it's lock - Fix (4) by rearranging existing locking: from: journal_start(); ->write_begin to: write_begin(); journal_extend() - Fix (5) simply by checking retvalue - Fix (6) by locking both (original and donor one) pages during extent swap with help of mext_page_double_lock() Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2012-09-26 23:52:07 +07:00
if (!*err)
*err = block_commit_write(pagep[0], from, from + replaced_size);
ext4: clean up online defrag bugs in move_extent_per_page() Non-full list of bugs: 1) uninitialized extent optimization does not hold page's lock, and simply replace brunches after that writeback code goes crazy because block mapping changed under it's feets kernel BUG at fs/ext4/inode.c:1434! ( 288'th xfstress) 2) uninitialized extent may became initialized right after we drop i_data_sem, so extent state must be rechecked 3) Locked pages goes uptodate via following sequence: ->readpage(page); lock_page(page); use_that_page(page) But after readpage() one may invalidate it because it is uptodate and unlocked (reclaimer does that) As result kernel bug at include/linux/buffer_head.c:133! 4) We call write_begin() with already opened stansaction which result in following deadlock: ->move_extent_per_page() ->ext4_journal_start()-> hold journal transaction ->write_begin() ->ext4_da_write_begin() ->ext4_nonda_switch() ->writeback_inodes_sb_if_idle() --> will wait for journal_stop() 5) try_to_release_page() may fail and it does fail if one of page's bh was pinned by journal 6) If we about to change page's mapping we MUST hold it's lock during entire remapping procedure, this is true for both pages(original and donor one) Fixes: - Avoid (1) and (2) simply by temproraly drop uninitialized extent handling optimization, this will be reimplemented later. - Fix (3) by manually forcing page to uptodate state w/o dropping it's lock - Fix (4) by rearranging existing locking: from: journal_start(); ->write_begin to: write_begin(); journal_extend() - Fix (5) simply by checking retvalue - Fix (6) by locking both (original and donor one) pages during extent swap with help of mext_page_double_lock() Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2012-09-26 23:52:07 +07:00
if (unlikely(*err < 0))
goto repair_branches;
/* Even in case of data=writeback it is reasonable to pin
* inode to transaction, to prevent unexpected data loss */
*err = ext4_jbd2_inode_add_write(handle, orig_inode,
(loff_t)orig_page_offset << PAGE_SHIFT, replaced_size);
ext4: clean up online defrag bugs in move_extent_per_page() Non-full list of bugs: 1) uninitialized extent optimization does not hold page's lock, and simply replace brunches after that writeback code goes crazy because block mapping changed under it's feets kernel BUG at fs/ext4/inode.c:1434! ( 288'th xfstress) 2) uninitialized extent may became initialized right after we drop i_data_sem, so extent state must be rechecked 3) Locked pages goes uptodate via following sequence: ->readpage(page); lock_page(page); use_that_page(page) But after readpage() one may invalidate it because it is uptodate and unlocked (reclaimer does that) As result kernel bug at include/linux/buffer_head.c:133! 4) We call write_begin() with already opened stansaction which result in following deadlock: ->move_extent_per_page() ->ext4_journal_start()-> hold journal transaction ->write_begin() ->ext4_da_write_begin() ->ext4_nonda_switch() ->writeback_inodes_sb_if_idle() --> will wait for journal_stop() 5) try_to_release_page() may fail and it does fail if one of page's bh was pinned by journal 6) If we about to change page's mapping we MUST hold it's lock during entire remapping procedure, this is true for both pages(original and donor one) Fixes: - Avoid (1) and (2) simply by temproraly drop uninitialized extent handling optimization, this will be reimplemented later. - Fix (3) by manually forcing page to uptodate state w/o dropping it's lock - Fix (4) by rearranging existing locking: from: journal_start(); ->write_begin to: write_begin(); journal_extend() - Fix (5) simply by checking retvalue - Fix (6) by locking both (original and donor one) pages during extent swap with help of mext_page_double_lock() Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2012-09-26 23:52:07 +07:00
unlock_pages:
unlock_page(pagep[0]);
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time ago with promise that one day it will be possible to implement page cache with bigger chunks than PAGE_SIZE. This promise never materialized. And unlikely will. We have many places where PAGE_CACHE_SIZE assumed to be equal to PAGE_SIZE. And it's constant source of confusion on whether PAGE_CACHE_* or PAGE_* constant should be used in a particular case, especially on the border between fs and mm. Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much breakage to be doable. Let's stop pretending that pages in page cache are special. They are not. The changes are pretty straight-forward: - <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN}; - page_cache_get() -> get_page(); - page_cache_release() -> put_page(); This patch contains automated changes generated with coccinelle using script below. For some reason, coccinelle doesn't patch header files. I've called spatch for them manually. The only adjustment after coccinelle is revert of changes to PAGE_CAHCE_ALIGN definition: we are going to drop it later. There are few places in the code where coccinelle didn't reach. I'll fix them manually in a separate patch. Comments and documentation also will be addressed with the separate patch. virtual patch @@ expression E; @@ - E << (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ expression E; @@ - E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ @@ - PAGE_CACHE_SHIFT + PAGE_SHIFT @@ @@ - PAGE_CACHE_SIZE + PAGE_SIZE @@ @@ - PAGE_CACHE_MASK + PAGE_MASK @@ expression E; @@ - PAGE_CACHE_ALIGN(E) + PAGE_ALIGN(E) @@ expression E; @@ - page_cache_get(E) + get_page(E) @@ expression E; @@ - page_cache_release(E) + put_page(E) Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 19:29:47 +07:00
put_page(pagep[0]);
ext4: clean up online defrag bugs in move_extent_per_page() Non-full list of bugs: 1) uninitialized extent optimization does not hold page's lock, and simply replace brunches after that writeback code goes crazy because block mapping changed under it's feets kernel BUG at fs/ext4/inode.c:1434! ( 288'th xfstress) 2) uninitialized extent may became initialized right after we drop i_data_sem, so extent state must be rechecked 3) Locked pages goes uptodate via following sequence: ->readpage(page); lock_page(page); use_that_page(page) But after readpage() one may invalidate it because it is uptodate and unlocked (reclaimer does that) As result kernel bug at include/linux/buffer_head.c:133! 4) We call write_begin() with already opened stansaction which result in following deadlock: ->move_extent_per_page() ->ext4_journal_start()-> hold journal transaction ->write_begin() ->ext4_da_write_begin() ->ext4_nonda_switch() ->writeback_inodes_sb_if_idle() --> will wait for journal_stop() 5) try_to_release_page() may fail and it does fail if one of page's bh was pinned by journal 6) If we about to change page's mapping we MUST hold it's lock during entire remapping procedure, this is true for both pages(original and donor one) Fixes: - Avoid (1) and (2) simply by temproraly drop uninitialized extent handling optimization, this will be reimplemented later. - Fix (3) by manually forcing page to uptodate state w/o dropping it's lock - Fix (4) by rearranging existing locking: from: journal_start(); ->write_begin to: write_begin(); journal_extend() - Fix (5) simply by checking retvalue - Fix (6) by locking both (original and donor one) pages during extent swap with help of mext_page_double_lock() Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2012-09-26 23:52:07 +07:00
unlock_page(pagep[1]);
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time ago with promise that one day it will be possible to implement page cache with bigger chunks than PAGE_SIZE. This promise never materialized. And unlikely will. We have many places where PAGE_CACHE_SIZE assumed to be equal to PAGE_SIZE. And it's constant source of confusion on whether PAGE_CACHE_* or PAGE_* constant should be used in a particular case, especially on the border between fs and mm. Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much breakage to be doable. Let's stop pretending that pages in page cache are special. They are not. The changes are pretty straight-forward: - <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN}; - page_cache_get() -> get_page(); - page_cache_release() -> put_page(); This patch contains automated changes generated with coccinelle using script below. For some reason, coccinelle doesn't patch header files. I've called spatch for them manually. The only adjustment after coccinelle is revert of changes to PAGE_CAHCE_ALIGN definition: we are going to drop it later. There are few places in the code where coccinelle didn't reach. I'll fix them manually in a separate patch. Comments and documentation also will be addressed with the separate patch. virtual patch @@ expression E; @@ - E << (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ expression E; @@ - E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ @@ - PAGE_CACHE_SHIFT + PAGE_SHIFT @@ @@ - PAGE_CACHE_SIZE + PAGE_SIZE @@ @@ - PAGE_CACHE_MASK + PAGE_MASK @@ expression E; @@ - PAGE_CACHE_ALIGN(E) + PAGE_ALIGN(E) @@ expression E; @@ - page_cache_get(E) + get_page(E) @@ expression E; @@ - page_cache_release(E) + put_page(E) Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 19:29:47 +07:00
put_page(pagep[1]);
ext4: clean up online defrag bugs in move_extent_per_page() Non-full list of bugs: 1) uninitialized extent optimization does not hold page's lock, and simply replace brunches after that writeback code goes crazy because block mapping changed under it's feets kernel BUG at fs/ext4/inode.c:1434! ( 288'th xfstress) 2) uninitialized extent may became initialized right after we drop i_data_sem, so extent state must be rechecked 3) Locked pages goes uptodate via following sequence: ->readpage(page); lock_page(page); use_that_page(page) But after readpage() one may invalidate it because it is uptodate and unlocked (reclaimer does that) As result kernel bug at include/linux/buffer_head.c:133! 4) We call write_begin() with already opened stansaction which result in following deadlock: ->move_extent_per_page() ->ext4_journal_start()-> hold journal transaction ->write_begin() ->ext4_da_write_begin() ->ext4_nonda_switch() ->writeback_inodes_sb_if_idle() --> will wait for journal_stop() 5) try_to_release_page() may fail and it does fail if one of page's bh was pinned by journal 6) If we about to change page's mapping we MUST hold it's lock during entire remapping procedure, this is true for both pages(original and donor one) Fixes: - Avoid (1) and (2) simply by temproraly drop uninitialized extent handling optimization, this will be reimplemented later. - Fix (3) by manually forcing page to uptodate state w/o dropping it's lock - Fix (4) by rearranging existing locking: from: journal_start(); ->write_begin to: write_begin(); journal_extend() - Fix (5) simply by checking retvalue - Fix (6) by locking both (original and donor one) pages during extent swap with help of mext_page_double_lock() Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2012-09-26 23:52:07 +07:00
stop_journal:
ext4_journal_stop(handle);
if (*err == -ENOSPC &&
ext4_should_retry_alloc(sb, &retries))
goto again;
ext4: clean up online defrag bugs in move_extent_per_page() Non-full list of bugs: 1) uninitialized extent optimization does not hold page's lock, and simply replace brunches after that writeback code goes crazy because block mapping changed under it's feets kernel BUG at fs/ext4/inode.c:1434! ( 288'th xfstress) 2) uninitialized extent may became initialized right after we drop i_data_sem, so extent state must be rechecked 3) Locked pages goes uptodate via following sequence: ->readpage(page); lock_page(page); use_that_page(page) But after readpage() one may invalidate it because it is uptodate and unlocked (reclaimer does that) As result kernel bug at include/linux/buffer_head.c:133! 4) We call write_begin() with already opened stansaction which result in following deadlock: ->move_extent_per_page() ->ext4_journal_start()-> hold journal transaction ->write_begin() ->ext4_da_write_begin() ->ext4_nonda_switch() ->writeback_inodes_sb_if_idle() --> will wait for journal_stop() 5) try_to_release_page() may fail and it does fail if one of page's bh was pinned by journal 6) If we about to change page's mapping we MUST hold it's lock during entire remapping procedure, this is true for both pages(original and donor one) Fixes: - Avoid (1) and (2) simply by temproraly drop uninitialized extent handling optimization, this will be reimplemented later. - Fix (3) by manually forcing page to uptodate state w/o dropping it's lock - Fix (4) by rearranging existing locking: from: journal_start(); ->write_begin to: write_begin(); journal_extend() - Fix (5) simply by checking retvalue - Fix (6) by locking both (original and donor one) pages during extent swap with help of mext_page_double_lock() Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2012-09-26 23:52:07 +07:00
/* Buffer was busy because probably is pinned to journal transaction,
* force transaction commit may help to free it. */
if (*err == -EBUSY && retries++ < 4 && EXT4_SB(sb)->s_journal &&
jbd2_journal_force_commit_nested(EXT4_SB(sb)->s_journal))
ext4: clean up online defrag bugs in move_extent_per_page() Non-full list of bugs: 1) uninitialized extent optimization does not hold page's lock, and simply replace brunches after that writeback code goes crazy because block mapping changed under it's feets kernel BUG at fs/ext4/inode.c:1434! ( 288'th xfstress) 2) uninitialized extent may became initialized right after we drop i_data_sem, so extent state must be rechecked 3) Locked pages goes uptodate via following sequence: ->readpage(page); lock_page(page); use_that_page(page) But after readpage() one may invalidate it because it is uptodate and unlocked (reclaimer does that) As result kernel bug at include/linux/buffer_head.c:133! 4) We call write_begin() with already opened stansaction which result in following deadlock: ->move_extent_per_page() ->ext4_journal_start()-> hold journal transaction ->write_begin() ->ext4_da_write_begin() ->ext4_nonda_switch() ->writeback_inodes_sb_if_idle() --> will wait for journal_stop() 5) try_to_release_page() may fail and it does fail if one of page's bh was pinned by journal 6) If we about to change page's mapping we MUST hold it's lock during entire remapping procedure, this is true for both pages(original and donor one) Fixes: - Avoid (1) and (2) simply by temproraly drop uninitialized extent handling optimization, this will be reimplemented later. - Fix (3) by manually forcing page to uptodate state w/o dropping it's lock - Fix (4) by rearranging existing locking: from: journal_start(); ->write_begin to: write_begin(); journal_extend() - Fix (5) simply by checking retvalue - Fix (6) by locking both (original and donor one) pages during extent swap with help of mext_page_double_lock() Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2012-09-26 23:52:07 +07:00
goto again;
return replaced_count;
ext4: clean up online defrag bugs in move_extent_per_page() Non-full list of bugs: 1) uninitialized extent optimization does not hold page's lock, and simply replace brunches after that writeback code goes crazy because block mapping changed under it's feets kernel BUG at fs/ext4/inode.c:1434! ( 288'th xfstress) 2) uninitialized extent may became initialized right after we drop i_data_sem, so extent state must be rechecked 3) Locked pages goes uptodate via following sequence: ->readpage(page); lock_page(page); use_that_page(page) But after readpage() one may invalidate it because it is uptodate and unlocked (reclaimer does that) As result kernel bug at include/linux/buffer_head.c:133! 4) We call write_begin() with already opened stansaction which result in following deadlock: ->move_extent_per_page() ->ext4_journal_start()-> hold journal transaction ->write_begin() ->ext4_da_write_begin() ->ext4_nonda_switch() ->writeback_inodes_sb_if_idle() --> will wait for journal_stop() 5) try_to_release_page() may fail and it does fail if one of page's bh was pinned by journal 6) If we about to change page's mapping we MUST hold it's lock during entire remapping procedure, this is true for both pages(original and donor one) Fixes: - Avoid (1) and (2) simply by temproraly drop uninitialized extent handling optimization, this will be reimplemented later. - Fix (3) by manually forcing page to uptodate state w/o dropping it's lock - Fix (4) by rearranging existing locking: from: journal_start(); ->write_begin to: write_begin(); journal_extend() - Fix (5) simply by checking retvalue - Fix (6) by locking both (original and donor one) pages during extent swap with help of mext_page_double_lock() Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2012-09-26 23:52:07 +07:00
repair_branches:
/*
* This should never ever happen!
* Extents are swapped already, but we are not able to copy data.
* Try to swap extents to it's original places
*/
ext4_double_down_write_data_sem(orig_inode, donor_inode);
replaced_count = ext4_swap_extents(handle, donor_inode, orig_inode,
orig_blk_offset, donor_blk_offset,
block_len_in_page, 0, &err2);
ext4_double_up_write_data_sem(orig_inode, donor_inode);
ext4: clean up online defrag bugs in move_extent_per_page() Non-full list of bugs: 1) uninitialized extent optimization does not hold page's lock, and simply replace brunches after that writeback code goes crazy because block mapping changed under it's feets kernel BUG at fs/ext4/inode.c:1434! ( 288'th xfstress) 2) uninitialized extent may became initialized right after we drop i_data_sem, so extent state must be rechecked 3) Locked pages goes uptodate via following sequence: ->readpage(page); lock_page(page); use_that_page(page) But after readpage() one may invalidate it because it is uptodate and unlocked (reclaimer does that) As result kernel bug at include/linux/buffer_head.c:133! 4) We call write_begin() with already opened stansaction which result in following deadlock: ->move_extent_per_page() ->ext4_journal_start()-> hold journal transaction ->write_begin() ->ext4_da_write_begin() ->ext4_nonda_switch() ->writeback_inodes_sb_if_idle() --> will wait for journal_stop() 5) try_to_release_page() may fail and it does fail if one of page's bh was pinned by journal 6) If we about to change page's mapping we MUST hold it's lock during entire remapping procedure, this is true for both pages(original and donor one) Fixes: - Avoid (1) and (2) simply by temproraly drop uninitialized extent handling optimization, this will be reimplemented later. - Fix (3) by manually forcing page to uptodate state w/o dropping it's lock - Fix (4) by rearranging existing locking: from: journal_start(); ->write_begin to: write_begin(); journal_extend() - Fix (5) simply by checking retvalue - Fix (6) by locking both (original and donor one) pages during extent swap with help of mext_page_double_lock() Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2012-09-26 23:52:07 +07:00
if (replaced_count != block_len_in_page) {
ext4_error_inode_block(orig_inode, (sector_t)(orig_blk_offset),
EIO, "Unable to copy data block,"
ext4: clean up online defrag bugs in move_extent_per_page() Non-full list of bugs: 1) uninitialized extent optimization does not hold page's lock, and simply replace brunches after that writeback code goes crazy because block mapping changed under it's feets kernel BUG at fs/ext4/inode.c:1434! ( 288'th xfstress) 2) uninitialized extent may became initialized right after we drop i_data_sem, so extent state must be rechecked 3) Locked pages goes uptodate via following sequence: ->readpage(page); lock_page(page); use_that_page(page) But after readpage() one may invalidate it because it is uptodate and unlocked (reclaimer does that) As result kernel bug at include/linux/buffer_head.c:133! 4) We call write_begin() with already opened stansaction which result in following deadlock: ->move_extent_per_page() ->ext4_journal_start()-> hold journal transaction ->write_begin() ->ext4_da_write_begin() ->ext4_nonda_switch() ->writeback_inodes_sb_if_idle() --> will wait for journal_stop() 5) try_to_release_page() may fail and it does fail if one of page's bh was pinned by journal 6) If we about to change page's mapping we MUST hold it's lock during entire remapping procedure, this is true for both pages(original and donor one) Fixes: - Avoid (1) and (2) simply by temproraly drop uninitialized extent handling optimization, this will be reimplemented later. - Fix (3) by manually forcing page to uptodate state w/o dropping it's lock - Fix (4) by rearranging existing locking: from: journal_start(); ->write_begin to: write_begin(); journal_extend() - Fix (5) simply by checking retvalue - Fix (6) by locking both (original and donor one) pages during extent swap with help of mext_page_double_lock() Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2012-09-26 23:52:07 +07:00
" data will be lost.");
*err = -EIO;
}
replaced_count = 0;
goto unlock_pages;
}
/**
* mext_check_arguments - Check whether move extent can be done
*
* @orig_inode: original inode
* @donor_inode: donor inode
* @orig_start: logical start offset in block for orig
* @donor_start: logical start offset in block for donor
* @len: the number of blocks to be moved
*
* Check the arguments of ext4_move_extents() whether the files can be
* exchanged with each other.
* Return 0 on success, or a negative error value on failure.
*/
static int
mext_check_arguments(struct inode *orig_inode,
struct inode *donor_inode, __u64 orig_start,
__u64 donor_start, __u64 *len)
{
__u64 orig_eof, donor_eof;
unsigned int blkbits = orig_inode->i_blkbits;
unsigned int blocksize = 1 << blkbits;
orig_eof = (i_size_read(orig_inode) + blocksize - 1) >> blkbits;
donor_eof = (i_size_read(donor_inode) + blocksize - 1) >> blkbits;
if (donor_inode->i_mode & (S_ISUID|S_ISGID)) {
ext4_debug("ext4 move extent: suid or sgid is set"
" to donor file [ino:orig %lu, donor %lu]\n",
orig_inode->i_ino, donor_inode->i_ino);
return -EINVAL;
}
if (IS_IMMUTABLE(donor_inode) || IS_APPEND(donor_inode))
return -EPERM;
/* Ext4 move extent does not support swapfile */
if (IS_SWAPFILE(orig_inode) || IS_SWAPFILE(donor_inode)) {
ext4_debug("ext4 move extent: The argument files should "
"not be swapfile [ino:orig %lu, donor %lu]\n",
orig_inode->i_ino, donor_inode->i_ino);
return -EBUSY;
}
if (ext4_is_quota_file(orig_inode) && ext4_is_quota_file(donor_inode)) {
ext4_debug("ext4 move extent: The argument files should "
"not be quota files [ino:orig %lu, donor %lu]\n",
orig_inode->i_ino, donor_inode->i_ino);
return -EBUSY;
}
/* Ext4 move extent supports only extent based file */
if (!(ext4_test_inode_flag(orig_inode, EXT4_INODE_EXTENTS))) {
ext4_debug("ext4 move extent: orig file is not extents "
"based file [ino:orig %lu]\n", orig_inode->i_ino);
return -EOPNOTSUPP;
} else if (!(ext4_test_inode_flag(donor_inode, EXT4_INODE_EXTENTS))) {
ext4_debug("ext4 move extent: donor file is not extents "
"based file [ino:donor %lu]\n", donor_inode->i_ino);
return -EOPNOTSUPP;
}
if ((!orig_inode->i_size) || (!donor_inode->i_size)) {
ext4_debug("ext4 move extent: File size is 0 byte\n");
return -EINVAL;
}
/* Start offset should be same */
if ((orig_start & ~(PAGE_MASK >> orig_inode->i_blkbits)) !=
(donor_start & ~(PAGE_MASK >> orig_inode->i_blkbits))) {
ext4_debug("ext4 move extent: orig and donor's start "
"offsets are not aligned [ino:orig %lu, donor %lu]\n",
orig_inode->i_ino, donor_inode->i_ino);
return -EINVAL;
}
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=$((2**32-1)) Reported-by: Kazuya Mio <k-mio@sx.jp.nec.com> Signed-off-by: Lukas Czerner <lczerner@redhat.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-06-06 11:05:17 +07:00
if ((orig_start >= EXT_MAX_BLOCKS) ||
(donor_start >= EXT_MAX_BLOCKS) ||
ext4: Fix max file size and logical block counting of extent format file Kazuya Mio reported that he was able to hit BUG_ON(next == lblock) in ext4_ext_put_gap_in_cache() while creating a sparse file in extent format and fill the tail of file up to its end. We will hit the BUG_ON when we write the last block (2^32-1) into the sparse file. The root cause of the problem lies in the fact that we specifically set s_maxbytes so that block at s_maxbytes fit into on-disk extent format, which is 32 bit long. However, we are not storing start and end block number, but rather start block number and length in blocks. It means that in order to cover extent from 0 to EXT_MAX_BLOCK we need EXT_MAX_BLOCK+1 to fit into len (because we counting block 0 as well) - and it does not. The only way to fix it without changing the meaning of the struct ext4_extent members is, as Kazuya Mio suggested, to lower s_maxbytes by one fs block so we can cover the whole extent we can get by the on-disk extent format. Also in many places EXT_MAX_BLOCK is used as length instead of maximum logical block number as the name suggests, it is all a bit messy. So this commit renames it to EXT_MAX_BLOCKS and change its usage in some places to actually be maximum number of blocks in the extent. The bug which this commit fixes can be reproduced as follows: dd if=/dev/zero of=/mnt/mp1/file bs=<blocksize> count=1 seek=$((2**32-2)) sync dd if=/dev/zero of=/mnt/mp1/file bs=<blocksize> count=1 seek=$((2**32-1)) Reported-by: Kazuya Mio <k-mio@sx.jp.nec.com> Signed-off-by: Lukas Czerner <lczerner@redhat.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-06-06 11:05:17 +07:00
(*len > EXT_MAX_BLOCKS) ||
(donor_start + *len >= EXT_MAX_BLOCKS) ||
ext4: Fix max file size and logical block counting of extent format file Kazuya Mio reported that he was able to hit BUG_ON(next == lblock) in ext4_ext_put_gap_in_cache() while creating a sparse file in extent format and fill the tail of file up to its end. We will hit the BUG_ON when we write the last block (2^32-1) into the sparse file. The root cause of the problem lies in the fact that we specifically set s_maxbytes so that block at s_maxbytes fit into on-disk extent format, which is 32 bit long. However, we are not storing start and end block number, but rather start block number and length in blocks. It means that in order to cover extent from 0 to EXT_MAX_BLOCK we need EXT_MAX_BLOCK+1 to fit into len (because we counting block 0 as well) - and it does not. The only way to fix it without changing the meaning of the struct ext4_extent members is, as Kazuya Mio suggested, to lower s_maxbytes by one fs block so we can cover the whole extent we can get by the on-disk extent format. Also in many places EXT_MAX_BLOCK is used as length instead of maximum logical block number as the name suggests, it is all a bit messy. So this commit renames it to EXT_MAX_BLOCKS and change its usage in some places to actually be maximum number of blocks in the extent. The bug which this commit fixes can be reproduced as follows: dd if=/dev/zero of=/mnt/mp1/file bs=<blocksize> count=1 seek=$((2**32-2)) sync dd if=/dev/zero of=/mnt/mp1/file bs=<blocksize> count=1 seek=$((2**32-1)) Reported-by: Kazuya Mio <k-mio@sx.jp.nec.com> Signed-off-by: Lukas Czerner <lczerner@redhat.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2011-06-06 11:05:17 +07:00
(orig_start + *len >= EXT_MAX_BLOCKS)) {
ext4_debug("ext4 move extent: Can't handle over [%u] blocks "
ext4: Fix max file size and logical block counting of extent format file Kazuya Mio reported that he was able to hit BUG_ON(next == lblock) in ext4_ext_put_gap_in_cache() while creating a sparse file in extent format and fill the tail of file up to its end. We will hit the BUG_ON when we write the last block (2^32-1) into the sparse file. The root cause of the problem lies in the fact that we specifically set s_maxbytes so that block at s_maxbytes fit into on-disk extent format, which is 32 bit long. However, we are not storing start and end block number, but rather start block number and length in blocks. It means that in order to cover extent from 0 to EXT_MAX_BLOCK we need EXT_MAX_BLOCK+1 to fit into len (because we counting block 0 as well) - and it does not. The only way to fix it without changing the meaning of the struct ext4_extent members is, as Kazuya Mio suggested, to lower s_maxbytes by one fs block so we can cover the whole extent we can get by the on-disk extent format. Also in many places EXT_MAX_BLOCK is used as length instead of maximum logical block number as the name suggests, it is all a bit messy. So this commit renames it to EXT_MAX_BLOCKS and change its usage in some places to actually be maximum number of blocks in the extent. The bug which this commit fixes can be reproduced as follows: dd if=/dev/zero of=/mnt/mp1/file bs=<blocksize> count=1 seek=$((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 11:05:17 +07:00
"[ino:orig %lu, donor %lu]\n", EXT_MAX_BLOCKS,
orig_inode->i_ino, donor_inode->i_ino);
return -EINVAL;
}
if (orig_eof <= orig_start)
*len = 0;
else if (orig_eof < orig_start + *len - 1)
*len = orig_eof - orig_start;
if (donor_eof <= donor_start)
*len = 0;
else if (donor_eof < donor_start + *len - 1)
*len = donor_eof - donor_start;
if (!*len) {
ext4_debug("ext4 move extent: len should not be 0 "
"[ino:orig %lu, donor %lu]\n", orig_inode->i_ino,
donor_inode->i_ino);
return -EINVAL;
}
return 0;
}
/**
* ext4_move_extents - Exchange the specified range of a file
*
* @o_filp: file structure of the original file
* @d_filp: file structure of the donor file
* @orig_blk: start offset in block for orig
* @donor_blk: start offset in block for donor
* @len: the number of blocks to be moved
* @moved_len: moved block length
*
* This function returns 0 and moved block length is set in moved_len
* if succeed, otherwise returns error value.
*
*/
int
ext4_move_extents(struct file *o_filp, struct file *d_filp, __u64 orig_blk,
__u64 donor_blk, __u64 len, __u64 *moved_len)
{
struct inode *orig_inode = file_inode(o_filp);
struct inode *donor_inode = file_inode(d_filp);
struct ext4_ext_path *path = NULL;
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time ago with promise that one day it will be possible to implement page cache with bigger chunks than PAGE_SIZE. This promise never materialized. And unlikely will. We have many places where PAGE_CACHE_SIZE assumed to be equal to PAGE_SIZE. And it's constant source of confusion on whether PAGE_CACHE_* or PAGE_* constant should be used in a particular case, especially on the border between fs and mm. Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much breakage to be doable. Let's stop pretending that pages in page cache are special. They are not. The changes are pretty straight-forward: - <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN}; - page_cache_get() -> get_page(); - page_cache_release() -> put_page(); This patch contains automated changes generated with coccinelle using script below. For some reason, coccinelle doesn't patch header files. I've called spatch for them manually. The only adjustment after coccinelle is revert of changes to PAGE_CAHCE_ALIGN definition: we are going to drop it later. There are few places in the code where coccinelle didn't reach. I'll fix them manually in a separate patch. Comments and documentation also will be addressed with the separate patch. virtual patch @@ expression E; @@ - E << (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ expression E; @@ - E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ @@ - PAGE_CACHE_SHIFT + PAGE_SHIFT @@ @@ - PAGE_CACHE_SIZE + PAGE_SIZE @@ @@ - PAGE_CACHE_MASK + PAGE_MASK @@ expression E; @@ - PAGE_CACHE_ALIGN(E) + PAGE_ALIGN(E) @@ expression E; @@ - page_cache_get(E) + get_page(E) @@ expression E; @@ - page_cache_release(E) + put_page(E) Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 19:29:47 +07:00
int blocks_per_page = PAGE_SIZE >> orig_inode->i_blkbits;
ext4_lblk_t o_end, o_start = orig_blk;
ext4_lblk_t d_start = donor_blk;
int ret;
if (orig_inode->i_sb != donor_inode->i_sb) {
ext4_debug("ext4 move extent: The argument files "
"should be in same FS [ino:orig %lu, donor %lu]\n",
orig_inode->i_ino, donor_inode->i_ino);
return -EINVAL;
}
/* orig and donor should be different inodes */
if (orig_inode == donor_inode) {
ext4_debug("ext4 move extent: The argument files should not "
"be same inode [ino:orig %lu, donor %lu]\n",
orig_inode->i_ino, donor_inode->i_ino);
return -EINVAL;
}
/* Regular file check */
if (!S_ISREG(orig_inode->i_mode) || !S_ISREG(donor_inode->i_mode)) {
ext4_debug("ext4 move extent: The argument files should be "
"regular file [ino:orig %lu, donor %lu]\n",
orig_inode->i_ino, donor_inode->i_ino);
return -EINVAL;
}
/* TODO: it's not obvious how to swap blocks for inodes with full
journaling enabled */
if (ext4_should_journal_data(orig_inode) ||
ext4_should_journal_data(donor_inode)) {
ext4_msg(orig_inode->i_sb, KERN_ERR,
"Online defrag not supported with data journaling");
return -EOPNOTSUPP;
}
if (IS_ENCRYPTED(orig_inode) || IS_ENCRYPTED(donor_inode)) {
ext4_msg(orig_inode->i_sb, KERN_ERR,
"Online defrag not supported for encrypted files");
return -EOPNOTSUPP;
}
/* Protect orig and donor inodes against a truncate */
lock_two_nondirectories(orig_inode, donor_inode);
/* Wait for all existing dio workers */
inode_dio_wait(orig_inode);
inode_dio_wait(donor_inode);
/* Protect extent tree against block allocations via delalloc */
ext4_double_down_write_data_sem(orig_inode, donor_inode);
/* Check the filesystem environment whether move_extent can be done */
ret = mext_check_arguments(orig_inode, donor_inode, orig_blk,
donor_blk, &len);
if (ret)
goto out;
o_end = o_start + len;
while (o_start < o_end) {
struct ext4_extent *ex;
ext4_lblk_t cur_blk, next_blk;
pgoff_t orig_page_index, donor_page_index;
int offset_in_page;
int unwritten, cur_len;
ret = get_ext_path(orig_inode, o_start, &path);
if (ret)
goto out;
ex = path[path->p_depth].p_ext;
next_blk = ext4_ext_next_allocated_block(path);
cur_blk = le32_to_cpu(ex->ee_block);
cur_len = ext4_ext_get_actual_len(ex);
/* Check hole before the start pos */
if (cur_blk + cur_len - 1 < o_start) {
if (next_blk == EXT_MAX_BLOCKS) {
o_start = o_end;
ret = -ENODATA;
goto out;
}
d_start += next_blk - o_start;
o_start = next_blk;
continue;
/* Check hole after the start pos */
} else if (cur_blk > o_start) {
/* Skip hole */
d_start += cur_blk - o_start;
o_start = cur_blk;
/* Extent inside requested range ?*/
if (cur_blk >= o_end)
goto out;
} else { /* in_range(o_start, o_blk, o_len) */
cur_len += cur_blk - o_start;
}
unwritten = ext4_ext_is_unwritten(ex);
if (o_end - o_start < cur_len)
cur_len = o_end - o_start;
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time ago with promise that one day it will be possible to implement page cache with bigger chunks than PAGE_SIZE. This promise never materialized. And unlikely will. We have many places where PAGE_CACHE_SIZE assumed to be equal to PAGE_SIZE. And it's constant source of confusion on whether PAGE_CACHE_* or PAGE_* constant should be used in a particular case, especially on the border between fs and mm. Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much breakage to be doable. Let's stop pretending that pages in page cache are special. They are not. The changes are pretty straight-forward: - <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN}; - page_cache_get() -> get_page(); - page_cache_release() -> put_page(); This patch contains automated changes generated with coccinelle using script below. For some reason, coccinelle doesn't patch header files. I've called spatch for them manually. The only adjustment after coccinelle is revert of changes to PAGE_CAHCE_ALIGN definition: we are going to drop it later. There are few places in the code where coccinelle didn't reach. I'll fix them manually in a separate patch. Comments and documentation also will be addressed with the separate patch. virtual patch @@ expression E; @@ - E << (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ expression E; @@ - E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ @@ - PAGE_CACHE_SHIFT + PAGE_SHIFT @@ @@ - PAGE_CACHE_SIZE + PAGE_SIZE @@ @@ - PAGE_CACHE_MASK + PAGE_MASK @@ expression E; @@ - PAGE_CACHE_ALIGN(E) + PAGE_ALIGN(E) @@ expression E; @@ - page_cache_get(E) + get_page(E) @@ expression E; @@ - page_cache_release(E) + put_page(E) Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 19:29:47 +07:00
orig_page_index = o_start >> (PAGE_SHIFT -
orig_inode->i_blkbits);
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time ago with promise that one day it will be possible to implement page cache with bigger chunks than PAGE_SIZE. This promise never materialized. And unlikely will. We have many places where PAGE_CACHE_SIZE assumed to be equal to PAGE_SIZE. And it's constant source of confusion on whether PAGE_CACHE_* or PAGE_* constant should be used in a particular case, especially on the border between fs and mm. Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much breakage to be doable. Let's stop pretending that pages in page cache are special. They are not. The changes are pretty straight-forward: - <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN}; - page_cache_get() -> get_page(); - page_cache_release() -> put_page(); This patch contains automated changes generated with coccinelle using script below. For some reason, coccinelle doesn't patch header files. I've called spatch for them manually. The only adjustment after coccinelle is revert of changes to PAGE_CAHCE_ALIGN definition: we are going to drop it later. There are few places in the code where coccinelle didn't reach. I'll fix them manually in a separate patch. Comments and documentation also will be addressed with the separate patch. virtual patch @@ expression E; @@ - E << (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ expression E; @@ - E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ @@ - PAGE_CACHE_SHIFT + PAGE_SHIFT @@ @@ - PAGE_CACHE_SIZE + PAGE_SIZE @@ @@ - PAGE_CACHE_MASK + PAGE_MASK @@ expression E; @@ - PAGE_CACHE_ALIGN(E) + PAGE_ALIGN(E) @@ expression E; @@ - page_cache_get(E) + get_page(E) @@ expression E; @@ - page_cache_release(E) + put_page(E) Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 19:29:47 +07:00
donor_page_index = d_start >> (PAGE_SHIFT -
donor_inode->i_blkbits);
offset_in_page = o_start % blocks_per_page;
if (cur_len > blocks_per_page- offset_in_page)
cur_len = blocks_per_page - offset_in_page;
/*
* Up semaphore to avoid following problems:
* a. transaction deadlock among ext4_journal_start,
* ->write_begin via pagefault, and jbd2_journal_commit
* b. racing with ->readpage, ->write_begin, and ext4_get_block
* in move_extent_per_page
*/
ext4_double_up_write_data_sem(orig_inode, donor_inode);
/* Swap original branches with new branches */
move_extent_per_page(o_filp, donor_inode,
orig_page_index, donor_page_index,
offset_in_page, cur_len,
unwritten, &ret);
ext4_double_down_write_data_sem(orig_inode, donor_inode);
if (ret < 0)
break;
o_start += cur_len;
d_start += cur_len;
}
*moved_len = o_start - orig_blk;
if (*moved_len > len)
*moved_len = len;
out:
ext4: Fix double-free of blocks with EXT4_IOC_MOVE_EXT At the beginning of ext4_move_extent(), we call ext4_discard_preallocations() to discard inode PAs of orig and donor inodes. But in the following case, blocks can be double freed, so move ext4_discard_preallocations() to the end of ext4_move_extents(). 1. Discard inode PAs of orig and donor inodes with ext4_discard_preallocations() in ext4_move_extents(). orig : [ DATA1 ] donor: [ DATA2 ] 2. While data blocks are exchanging between orig and donor inodes, new inode PAs is created to orig by other process's block allocation. (Since there are semaphore gaps in ext4_move_extents().) And new inode PAs is used partially (2-1). 2-1 Create new inode PAs to orig inode orig : [ DATA1 | used PA1 | free PA1 ] donor: [ DATA2 ] 3. Donor inode which has old orig inode's blocks is deleted after EXT4_IOC_MOVE_EXT finished (3-1, 3-2). So the block bitmap corresponds to old orig inode's blocks are freed. 3-1 After EXT4_IOC_MOVE_EXT finished orig : [ DATA2 | free PA1 ] donor: [ DATA1 | used PA1 ] 3-2 Delete donor inode orig : [ DATA2 | free PA1 ] donor: [ FREE SPACE(DATA1) | FREE SPACE(used PA1) ] 4. The double-free of blocks is occurred, when close() is called to orig inode. Because ext4_discard_preallocations() for orig inode frees used PA1 and free PA1, though used PA1 is already freed in 3. 4-1 Double-free of blocks is occurred orig : [ DATA2 | FREE SPACE(free PA1) ] donor: [ FREE SPACE(DATA1) | DOUBLE FREE(used PA1) ] Signed-off-by: Akira Fujita <a-fujita@rs.jp.nec.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
2009-11-24 22:19:57 +07:00
if (*moved_len) {
ext4_discard_preallocations(orig_inode);
ext4_discard_preallocations(donor_inode);
}
ext4_ext_drop_refs(path);
kfree(path);
ext4_double_up_write_data_sem(orig_inode, donor_inode);
unlock_two_nondirectories(orig_inode, donor_inode);
return ret;
}