linux/fs/xfs/xfs_reflink.c

1726 lines
48 KiB
C
Raw Normal View History

// SPDX-License-Identifier: GPL-2.0+
/*
* Copyright (C) 2016 Oracle. All Rights Reserved.
* Author: Darrick J. Wong <darrick.wong@oracle.com>
*/
#include "xfs.h"
#include "xfs_fs.h"
#include "xfs_shared.h"
#include "xfs_format.h"
#include "xfs_log_format.h"
#include "xfs_trans_resv.h"
#include "xfs_mount.h"
#include "xfs_defer.h"
#include "xfs_inode.h"
#include "xfs_trans.h"
#include "xfs_bmap.h"
#include "xfs_bmap_util.h"
#include "xfs_trace.h"
#include "xfs_icache.h"
#include "xfs_btree.h"
#include "xfs_refcount_btree.h"
#include "xfs_refcount.h"
#include "xfs_bmap_btree.h"
#include "xfs_trans_space.h"
#include "xfs_bit.h"
#include "xfs_alloc.h"
#include "xfs_quota.h"
#include "xfs_reflink.h"
#include "xfs_iomap.h"
#include "xfs_ag.h"
#include "xfs_ag_resv.h"
/*
* Copy on Write of Shared Blocks
*
* XFS must preserve "the usual" file semantics even when two files share
* the same physical blocks. This means that a write to one file must not
* alter the blocks in a different file; the way that we'll do that is
* through the use of a copy-on-write mechanism. At a high level, that
* means that when we want to write to a shared block, we allocate a new
* block, write the data to the new block, and if that succeeds we map the
* new block into the file.
*
* XFS provides a "delayed allocation" mechanism that defers the allocation
* of disk blocks to dirty-but-not-yet-mapped file blocks as long as
* possible. This reduces fragmentation by enabling the filesystem to ask
* for bigger chunks less often, which is exactly what we want for CoW.
*
* The delalloc mechanism begins when the kernel wants to make a block
* writable (write_begin or page_mkwrite). If the offset is not mapped, we
* create a delalloc mapping, which is a regular in-core extent, but without
* a real startblock. (For delalloc mappings, the startblock encodes both
* a flag that this is a delalloc mapping, and a worst-case estimate of how
* many blocks might be required to put the mapping into the BMBT.) delalloc
* mappings are a reservation against the free space in the filesystem;
* adjacent mappings can also be combined into fewer larger mappings.
*
xfs: mark speculative prealloc CoW fork extents unwritten Christoph Hellwig pointed out that there's a potentially nasty race when performing simultaneous nearby directio cow writes: "Thread 1 writes a range from B to c " B --------- C p "a little later thread 2 writes from A to B " A --------- B p [editor's note: the 'p' denote cowextsize boundaries, which I added to make this more clear] "but the code preallocates beyond B into the range where thread "1 has just written, but ->end_io hasn't been called yet. "But once ->end_io is called thread 2 has already allocated "up to the extent size hint into the write range of thread 1, "so the end_io handler will splice the unintialized blocks from "that preallocation back into the file right after B." We can avoid this race by ensuring that thread 1 cannot accidentally remap the blocks that thread 2 allocated (as part of speculative preallocation) as part of t2's write preparation in t1's end_io handler. The way we make this happen is by taking advantage of the unwritten extent flag as an intermediate step. Recall that when we begin the process of writing data to shared blocks, we create a delayed allocation extent in the CoW fork: D: --RRRRRRSSSRRRRRRRR--- C: ------DDDDDDD--------- When a thread prepares to CoW some dirty data out to disk, it will now convert the delalloc reservation into an /unwritten/ allocated extent in the cow fork. The da conversion code tries to opportunistically allocate as much of a (speculatively prealloc'd) extent as possible, so we may end up allocating a larger extent than we're actually writing out: D: --RRRRRRSSSRRRRRRRR--- U: ------UUUUUUU--------- Next, we convert only the part of the extent that we're actively planning to write to normal (i.e. not unwritten) status: D: --RRRRRRSSSRRRRRRRR--- U: ------UURRUUU--------- If the write succeeds, the end_cow function will now scan the relevant range of the CoW fork for real extents and remap only the real extents into the data fork: D: --RRRRRRRRSRRRRRRRR--- U: ------UU--UUU--------- This ensures that we never obliterate valid data fork extents with unwritten blocks from the CoW fork. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Christoph Hellwig <hch@lst.de>
2017-02-03 07:14:02 +08:00
* As an optimization, the CoW extent size hint (cowextsz) creates
* outsized aligned delalloc reservations in the hope of landing out of
* order nearby CoW writes in a single extent on disk, thereby reducing
* fragmentation and improving future performance.
*
* D: --RRRRRRSSSRRRRRRRR--- (data fork)
* C: ------DDDDDDD--------- (CoW fork)
*
* When dirty pages are being written out (typically in writepage), the
xfs: mark speculative prealloc CoW fork extents unwritten Christoph Hellwig pointed out that there's a potentially nasty race when performing simultaneous nearby directio cow writes: "Thread 1 writes a range from B to c " B --------- C p "a little later thread 2 writes from A to B " A --------- B p [editor's note: the 'p' denote cowextsize boundaries, which I added to make this more clear] "but the code preallocates beyond B into the range where thread "1 has just written, but ->end_io hasn't been called yet. "But once ->end_io is called thread 2 has already allocated "up to the extent size hint into the write range of thread 1, "so the end_io handler will splice the unintialized blocks from "that preallocation back into the file right after B." We can avoid this race by ensuring that thread 1 cannot accidentally remap the blocks that thread 2 allocated (as part of speculative preallocation) as part of t2's write preparation in t1's end_io handler. The way we make this happen is by taking advantage of the unwritten extent flag as an intermediate step. Recall that when we begin the process of writing data to shared blocks, we create a delayed allocation extent in the CoW fork: D: --RRRRRRSSSRRRRRRRR--- C: ------DDDDDDD--------- When a thread prepares to CoW some dirty data out to disk, it will now convert the delalloc reservation into an /unwritten/ allocated extent in the cow fork. The da conversion code tries to opportunistically allocate as much of a (speculatively prealloc'd) extent as possible, so we may end up allocating a larger extent than we're actually writing out: D: --RRRRRRSSSRRRRRRRR--- U: ------UUUUUUU--------- Next, we convert only the part of the extent that we're actively planning to write to normal (i.e. not unwritten) status: D: --RRRRRRSSSRRRRRRRR--- U: ------UURRUUU--------- If the write succeeds, the end_cow function will now scan the relevant range of the CoW fork for real extents and remap only the real extents into the data fork: D: --RRRRRRRRSRRRRRRRR--- U: ------UU--UUU--------- This ensures that we never obliterate valid data fork extents with unwritten blocks from the CoW fork. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Christoph Hellwig <hch@lst.de>
2017-02-03 07:14:02 +08:00
* delalloc reservations are converted into unwritten mappings by
* allocating blocks and replacing the delalloc mapping with real ones.
* A delalloc mapping can be replaced by several unwritten ones if the
* free space is fragmented.
*
* D: --RRRRRRSSSRRRRRRRR---
* C: ------UUUUUUU---------
*
* We want to adapt the delalloc mechanism for copy-on-write, since the
* write paths are similar. The first two steps (creating the reservation
* and allocating the blocks) are exactly the same as delalloc except that
* the mappings must be stored in a separate CoW fork because we do not want
* to disturb the mapping in the data fork until we're sure that the write
* succeeded. IO completion in this case is the process of removing the old
* mapping from the data fork and moving the new mapping from the CoW fork to
* the data fork. This will be discussed shortly.
*
* For now, unaligned directio writes will be bounced back to the page cache.
* Block-aligned directio writes will use the same mechanism as buffered
* writes.
*
xfs: mark speculative prealloc CoW fork extents unwritten Christoph Hellwig pointed out that there's a potentially nasty race when performing simultaneous nearby directio cow writes: "Thread 1 writes a range from B to c " B --------- C p "a little later thread 2 writes from A to B " A --------- B p [editor's note: the 'p' denote cowextsize boundaries, which I added to make this more clear] "but the code preallocates beyond B into the range where thread "1 has just written, but ->end_io hasn't been called yet. "But once ->end_io is called thread 2 has already allocated "up to the extent size hint into the write range of thread 1, "so the end_io handler will splice the unintialized blocks from "that preallocation back into the file right after B." We can avoid this race by ensuring that thread 1 cannot accidentally remap the blocks that thread 2 allocated (as part of speculative preallocation) as part of t2's write preparation in t1's end_io handler. The way we make this happen is by taking advantage of the unwritten extent flag as an intermediate step. Recall that when we begin the process of writing data to shared blocks, we create a delayed allocation extent in the CoW fork: D: --RRRRRRSSSRRRRRRRR--- C: ------DDDDDDD--------- When a thread prepares to CoW some dirty data out to disk, it will now convert the delalloc reservation into an /unwritten/ allocated extent in the cow fork. The da conversion code tries to opportunistically allocate as much of a (speculatively prealloc'd) extent as possible, so we may end up allocating a larger extent than we're actually writing out: D: --RRRRRRSSSRRRRRRRR--- U: ------UUUUUUU--------- Next, we convert only the part of the extent that we're actively planning to write to normal (i.e. not unwritten) status: D: --RRRRRRSSSRRRRRRRR--- U: ------UURRUUU--------- If the write succeeds, the end_cow function will now scan the relevant range of the CoW fork for real extents and remap only the real extents into the data fork: D: --RRRRRRRRSRRRRRRRR--- U: ------UU--UUU--------- This ensures that we never obliterate valid data fork extents with unwritten blocks from the CoW fork. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Christoph Hellwig <hch@lst.de>
2017-02-03 07:14:02 +08:00
* Just prior to submitting the actual disk write requests, we convert
* the extents representing the range of the file actually being written
* (as opposed to extra pieces created for the cowextsize hint) to real
* extents. This will become important in the next step:
*
* D: --RRRRRRSSSRRRRRRRR---
* C: ------UUrrUUU---------
*
* CoW remapping must be done after the data block write completes,
* because we don't want to destroy the old data fork map until we're sure
* the new block has been written. Since the new mappings are kept in a
* separate fork, we can simply iterate these mappings to find the ones
* that cover the file blocks that we just CoW'd. For each extent, simply
* unmap the corresponding range in the data fork, map the new range into
xfs: mark speculative prealloc CoW fork extents unwritten Christoph Hellwig pointed out that there's a potentially nasty race when performing simultaneous nearby directio cow writes: "Thread 1 writes a range from B to c " B --------- C p "a little later thread 2 writes from A to B " A --------- B p [editor's note: the 'p' denote cowextsize boundaries, which I added to make this more clear] "but the code preallocates beyond B into the range where thread "1 has just written, but ->end_io hasn't been called yet. "But once ->end_io is called thread 2 has already allocated "up to the extent size hint into the write range of thread 1, "so the end_io handler will splice the unintialized blocks from "that preallocation back into the file right after B." We can avoid this race by ensuring that thread 1 cannot accidentally remap the blocks that thread 2 allocated (as part of speculative preallocation) as part of t2's write preparation in t1's end_io handler. The way we make this happen is by taking advantage of the unwritten extent flag as an intermediate step. Recall that when we begin the process of writing data to shared blocks, we create a delayed allocation extent in the CoW fork: D: --RRRRRRSSSRRRRRRRR--- C: ------DDDDDDD--------- When a thread prepares to CoW some dirty data out to disk, it will now convert the delalloc reservation into an /unwritten/ allocated extent in the cow fork. The da conversion code tries to opportunistically allocate as much of a (speculatively prealloc'd) extent as possible, so we may end up allocating a larger extent than we're actually writing out: D: --RRRRRRSSSRRRRRRRR--- U: ------UUUUUUU--------- Next, we convert only the part of the extent that we're actively planning to write to normal (i.e. not unwritten) status: D: --RRRRRRSSSRRRRRRRR--- U: ------UURRUUU--------- If the write succeeds, the end_cow function will now scan the relevant range of the CoW fork for real extents and remap only the real extents into the data fork: D: --RRRRRRRRSRRRRRRRR--- U: ------UU--UUU--------- This ensures that we never obliterate valid data fork extents with unwritten blocks from the CoW fork. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Christoph Hellwig <hch@lst.de>
2017-02-03 07:14:02 +08:00
* the data fork, and remove the extent from the CoW fork. Because of
* the presence of the cowextsize hint, however, we must be careful
* only to remap the blocks that we've actually written out -- we must
* never remap delalloc reservations nor CoW staging blocks that have
* yet to be written. This corresponds exactly to the real extents in
* the CoW fork:
*
* D: --RRRRRRrrSRRRRRRRR---
* C: ------UU--UUU---------
*
* Since the remapping operation can be applied to an arbitrary file
* range, we record the need for the remap step as a flag in the ioend
* instead of declaring a new IO type. This is required for direct io
* because we only have ioend for the whole dio, and we have to be able to
* remember the presence of unwritten blocks and CoW blocks with a single
* ioend structure. Better yet, the more ground we can cover with one
* ioend, the better.
*/
/*
* Given an AG extent, find the lowest-numbered run of shared blocks
* within that range and return the range in fbno/flen. If
* find_end_of_shared is true, return the longest contiguous extent of
* shared blocks. If there are no shared extents, fbno and flen will
* be set to NULLAGBLOCK and 0, respectively.
*/
static int
xfs_reflink_find_shared(
struct xfs_perag *pag,
struct xfs_trans *tp,
xfs_agblock_t agbno,
xfs_extlen_t aglen,
xfs_agblock_t *fbno,
xfs_extlen_t *flen,
bool find_end_of_shared)
{
struct xfs_buf *agbp;
struct xfs_btree_cur *cur;
int error;
error = xfs_alloc_read_agf(pag, tp, 0, &agbp);
if (error)
return error;
cur = xfs_refcountbt_init_cursor(pag->pag_mount, tp, agbp, pag);
error = xfs_refcount_find_shared(cur, agbno, aglen, fbno, flen,
find_end_of_shared);
xfs_btree_del_cursor(cur, error);
xfs_trans_brelse(tp, agbp);
return error;
}
/*
* Trim the mapping to the next block where there's a change in the
* shared/unshared status. More specifically, this means that we
* find the lowest-numbered extent of shared blocks that coincides with
* the given block mapping. If the shared extent overlaps the start of
* the mapping, trim the mapping to the end of the shared extent. If
* the shared region intersects the mapping, trim the mapping to the
* start of the shared extent. If there are no shared regions that
* overlap, just return the original extent.
*/
int
xfs_reflink_trim_around_shared(
struct xfs_inode *ip,
struct xfs_bmbt_irec *irec,
bool *shared)
{
struct xfs_mount *mp = ip->i_mount;
struct xfs_perag *pag;
xfs_agblock_t agbno;
xfs_extlen_t aglen;
xfs_agblock_t fbno;
xfs_extlen_t flen;
int error = 0;
/* Holes, unwritten, and delalloc extents cannot be shared */
if (!xfs_is_cow_inode(ip) || !xfs_bmap_is_written_extent(irec)) {
*shared = false;
return 0;
}
trace_xfs_reflink_trim_around_shared(ip, irec);
pag = xfs_perag_get(mp, XFS_FSB_TO_AGNO(mp, irec->br_startblock));
agbno = XFS_FSB_TO_AGBNO(mp, irec->br_startblock);
aglen = irec->br_blockcount;
error = xfs_reflink_find_shared(pag, NULL, agbno, aglen, &fbno, &flen,
true);
xfs_perag_put(pag);
if (error)
return error;
*shared = false;
if (fbno == NULLAGBLOCK) {
/* No shared blocks at all. */
return 0;
}
if (fbno == agbno) {
/*
* The start of this extent is shared. Truncate the
* mapping at the end of the shared region so that a
* subsequent iteration starts at the start of the
* unshared region.
*/
irec->br_blockcount = flen;
*shared = true;
return 0;
}
/*
* There's a shared extent midway through this extent.
* Truncate the mapping at the start of the shared
* extent so that a subsequent iteration starts at the
* start of the shared region.
*/
irec->br_blockcount = fbno - agbno;
return 0;
}
int
xfs_bmap_trim_cow(
xfs: introduce an always_cow mode Add a mode where XFS never overwrites existing blocks in place. This is to aid debugging our COW code, and also put infatructure in place for things like possible future support for zoned block devices, which can't support overwrites. This mode is enabled globally by doing a: echo 1 > /sys/fs/xfs/debug/always_cow Note that the parameter is global to allow running all tests in xfstests easily in this mode, which would not easily be possible with a per-fs sysfs file. In always_cow mode persistent preallocations are disabled, and fallocate will fail when called with a 0 mode (with our without FALLOC_FL_KEEP_SIZE), and not create unwritten extent for zeroed space when called with FALLOC_FL_ZERO_RANGE or FALLOC_FL_UNSHARE_RANGE. There are a few interesting xfstests failures when run in always_cow mode: - generic/392 fails because the bytes used in the file used to test hole punch recovery are less after the log replay. This is because the blocks written and then punched out are only freed with a delay due to the logging mechanism. - xfs/170 will fail as the already fragile file streams mechanism doesn't seem to interact well with the COW allocator - xfs/180 xfs/182 xfs/192 xfs/198 xfs/204 and xfs/208 will claim the file system is badly fragmented, but there is not much we can do to avoid that when always writing out of place - xfs/205 fails because overwriting a file in always_cow mode will require new space allocation and the assumption in the test thus don't work anymore. - xfs/326 fails to modify the file at all in always_cow mode after injecting the refcount error, leading to an unexpected md5sum after the remount, but that again is expected Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
2019-02-19 01:38:49 +08:00
struct xfs_inode *ip,
struct xfs_bmbt_irec *imap,
bool *shared)
{
/* We can't update any real extents in always COW mode. */
if (xfs_is_always_cow_inode(ip) &&
!isnullstartblock(imap->br_startblock)) {
*shared = true;
return 0;
}
/* Trim the mapping to the nearest shared extent boundary. */
return xfs_reflink_trim_around_shared(ip, imap, shared);
}
static int
xfs_reflink_convert_cow_locked(
struct xfs_inode *ip,
xfs_fileoff_t offset_fsb,
xfs_filblks_t count_fsb)
xfs: mark speculative prealloc CoW fork extents unwritten Christoph Hellwig pointed out that there's a potentially nasty race when performing simultaneous nearby directio cow writes: "Thread 1 writes a range from B to c " B --------- C p "a little later thread 2 writes from A to B " A --------- B p [editor's note: the 'p' denote cowextsize boundaries, which I added to make this more clear] "but the code preallocates beyond B into the range where thread "1 has just written, but ->end_io hasn't been called yet. "But once ->end_io is called thread 2 has already allocated "up to the extent size hint into the write range of thread 1, "so the end_io handler will splice the unintialized blocks from "that preallocation back into the file right after B." We can avoid this race by ensuring that thread 1 cannot accidentally remap the blocks that thread 2 allocated (as part of speculative preallocation) as part of t2's write preparation in t1's end_io handler. The way we make this happen is by taking advantage of the unwritten extent flag as an intermediate step. Recall that when we begin the process of writing data to shared blocks, we create a delayed allocation extent in the CoW fork: D: --RRRRRRSSSRRRRRRRR--- C: ------DDDDDDD--------- When a thread prepares to CoW some dirty data out to disk, it will now convert the delalloc reservation into an /unwritten/ allocated extent in the cow fork. The da conversion code tries to opportunistically allocate as much of a (speculatively prealloc'd) extent as possible, so we may end up allocating a larger extent than we're actually writing out: D: --RRRRRRSSSRRRRRRRR--- U: ------UUUUUUU--------- Next, we convert only the part of the extent that we're actively planning to write to normal (i.e. not unwritten) status: D: --RRRRRRSSSRRRRRRRR--- U: ------UURRUUU--------- If the write succeeds, the end_cow function will now scan the relevant range of the CoW fork for real extents and remap only the real extents into the data fork: D: --RRRRRRRRSRRRRRRRR--- U: ------UU--UUU--------- This ensures that we never obliterate valid data fork extents with unwritten blocks from the CoW fork. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Christoph Hellwig <hch@lst.de>
2017-02-03 07:14:02 +08:00
{
struct xfs_iext_cursor icur;
struct xfs_bmbt_irec got;
struct xfs_btree_cur *dummy_cur = NULL;
int dummy_logflags;
int error = 0;
xfs: mark speculative prealloc CoW fork extents unwritten Christoph Hellwig pointed out that there's a potentially nasty race when performing simultaneous nearby directio cow writes: "Thread 1 writes a range from B to c " B --------- C p "a little later thread 2 writes from A to B " A --------- B p [editor's note: the 'p' denote cowextsize boundaries, which I added to make this more clear] "but the code preallocates beyond B into the range where thread "1 has just written, but ->end_io hasn't been called yet. "But once ->end_io is called thread 2 has already allocated "up to the extent size hint into the write range of thread 1, "so the end_io handler will splice the unintialized blocks from "that preallocation back into the file right after B." We can avoid this race by ensuring that thread 1 cannot accidentally remap the blocks that thread 2 allocated (as part of speculative preallocation) as part of t2's write preparation in t1's end_io handler. The way we make this happen is by taking advantage of the unwritten extent flag as an intermediate step. Recall that when we begin the process of writing data to shared blocks, we create a delayed allocation extent in the CoW fork: D: --RRRRRRSSSRRRRRRRR--- C: ------DDDDDDD--------- When a thread prepares to CoW some dirty data out to disk, it will now convert the delalloc reservation into an /unwritten/ allocated extent in the cow fork. The da conversion code tries to opportunistically allocate as much of a (speculatively prealloc'd) extent as possible, so we may end up allocating a larger extent than we're actually writing out: D: --RRRRRRSSSRRRRRRRR--- U: ------UUUUUUU--------- Next, we convert only the part of the extent that we're actively planning to write to normal (i.e. not unwritten) status: D: --RRRRRRSSSRRRRRRRR--- U: ------UURRUUU--------- If the write succeeds, the end_cow function will now scan the relevant range of the CoW fork for real extents and remap only the real extents into the data fork: D: --RRRRRRRRSRRRRRRRR--- U: ------UU--UUU--------- This ensures that we never obliterate valid data fork extents with unwritten blocks from the CoW fork. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Christoph Hellwig <hch@lst.de>
2017-02-03 07:14:02 +08:00
if (!xfs_iext_lookup_extent(ip, ip->i_cowfp, offset_fsb, &icur, &got))
xfs: mark speculative prealloc CoW fork extents unwritten Christoph Hellwig pointed out that there's a potentially nasty race when performing simultaneous nearby directio cow writes: "Thread 1 writes a range from B to c " B --------- C p "a little later thread 2 writes from A to B " A --------- B p [editor's note: the 'p' denote cowextsize boundaries, which I added to make this more clear] "but the code preallocates beyond B into the range where thread "1 has just written, but ->end_io hasn't been called yet. "But once ->end_io is called thread 2 has already allocated "up to the extent size hint into the write range of thread 1, "so the end_io handler will splice the unintialized blocks from "that preallocation back into the file right after B." We can avoid this race by ensuring that thread 1 cannot accidentally remap the blocks that thread 2 allocated (as part of speculative preallocation) as part of t2's write preparation in t1's end_io handler. The way we make this happen is by taking advantage of the unwritten extent flag as an intermediate step. Recall that when we begin the process of writing data to shared blocks, we create a delayed allocation extent in the CoW fork: D: --RRRRRRSSSRRRRRRRR--- C: ------DDDDDDD--------- When a thread prepares to CoW some dirty data out to disk, it will now convert the delalloc reservation into an /unwritten/ allocated extent in the cow fork. The da conversion code tries to opportunistically allocate as much of a (speculatively prealloc'd) extent as possible, so we may end up allocating a larger extent than we're actually writing out: D: --RRRRRRSSSRRRRRRRR--- U: ------UUUUUUU--------- Next, we convert only the part of the extent that we're actively planning to write to normal (i.e. not unwritten) status: D: --RRRRRRSSSRRRRRRRR--- U: ------UURRUUU--------- If the write succeeds, the end_cow function will now scan the relevant range of the CoW fork for real extents and remap only the real extents into the data fork: D: --RRRRRRRRSRRRRRRRR--- U: ------UU--UUU--------- This ensures that we never obliterate valid data fork extents with unwritten blocks from the CoW fork. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Christoph Hellwig <hch@lst.de>
2017-02-03 07:14:02 +08:00
return 0;
do {
if (got.br_startoff >= offset_fsb + count_fsb)
break;
if (got.br_state == XFS_EXT_NORM)
continue;
if (WARN_ON_ONCE(isnullstartblock(got.br_startblock)))
return -EIO;
xfs_trim_extent(&got, offset_fsb, count_fsb);
if (!got.br_blockcount)
continue;
got.br_state = XFS_EXT_NORM;
error = xfs_bmap_add_extent_unwritten_real(NULL, ip,
XFS_COW_FORK, &icur, &dummy_cur, &got,
&dummy_logflags);
if (error)
return error;
} while (xfs_iext_next_extent(ip->i_cowfp, &icur, &got));
return error;
xfs: mark speculative prealloc CoW fork extents unwritten Christoph Hellwig pointed out that there's a potentially nasty race when performing simultaneous nearby directio cow writes: "Thread 1 writes a range from B to c " B --------- C p "a little later thread 2 writes from A to B " A --------- B p [editor's note: the 'p' denote cowextsize boundaries, which I added to make this more clear] "but the code preallocates beyond B into the range where thread "1 has just written, but ->end_io hasn't been called yet. "But once ->end_io is called thread 2 has already allocated "up to the extent size hint into the write range of thread 1, "so the end_io handler will splice the unintialized blocks from "that preallocation back into the file right after B." We can avoid this race by ensuring that thread 1 cannot accidentally remap the blocks that thread 2 allocated (as part of speculative preallocation) as part of t2's write preparation in t1's end_io handler. The way we make this happen is by taking advantage of the unwritten extent flag as an intermediate step. Recall that when we begin the process of writing data to shared blocks, we create a delayed allocation extent in the CoW fork: D: --RRRRRRSSSRRRRRRRR--- C: ------DDDDDDD--------- When a thread prepares to CoW some dirty data out to disk, it will now convert the delalloc reservation into an /unwritten/ allocated extent in the cow fork. The da conversion code tries to opportunistically allocate as much of a (speculatively prealloc'd) extent as possible, so we may end up allocating a larger extent than we're actually writing out: D: --RRRRRRSSSRRRRRRRR--- U: ------UUUUUUU--------- Next, we convert only the part of the extent that we're actively planning to write to normal (i.e. not unwritten) status: D: --RRRRRRSSSRRRRRRRR--- U: ------UURRUUU--------- If the write succeeds, the end_cow function will now scan the relevant range of the CoW fork for real extents and remap only the real extents into the data fork: D: --RRRRRRRRSRRRRRRRR--- U: ------UU--UUU--------- This ensures that we never obliterate valid data fork extents with unwritten blocks from the CoW fork. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Christoph Hellwig <hch@lst.de>
2017-02-03 07:14:02 +08:00
}
/* Convert all of the unwritten CoW extents in a file's range to real ones. */
int
xfs_reflink_convert_cow(
struct xfs_inode *ip,
xfs_off_t offset,
xfs_off_t count)
{
struct xfs_mount *mp = ip->i_mount;
xfs_fileoff_t offset_fsb = XFS_B_TO_FSBT(mp, offset);
xfs_fileoff_t end_fsb = XFS_B_TO_FSB(mp, offset + count);
xfs_filblks_t count_fsb = end_fsb - offset_fsb;
int error;
xfs: mark speculative prealloc CoW fork extents unwritten Christoph Hellwig pointed out that there's a potentially nasty race when performing simultaneous nearby directio cow writes: "Thread 1 writes a range from B to c " B --------- C p "a little later thread 2 writes from A to B " A --------- B p [editor's note: the 'p' denote cowextsize boundaries, which I added to make this more clear] "but the code preallocates beyond B into the range where thread "1 has just written, but ->end_io hasn't been called yet. "But once ->end_io is called thread 2 has already allocated "up to the extent size hint into the write range of thread 1, "so the end_io handler will splice the unintialized blocks from "that preallocation back into the file right after B." We can avoid this race by ensuring that thread 1 cannot accidentally remap the blocks that thread 2 allocated (as part of speculative preallocation) as part of t2's write preparation in t1's end_io handler. The way we make this happen is by taking advantage of the unwritten extent flag as an intermediate step. Recall that when we begin the process of writing data to shared blocks, we create a delayed allocation extent in the CoW fork: D: --RRRRRRSSSRRRRRRRR--- C: ------DDDDDDD--------- When a thread prepares to CoW some dirty data out to disk, it will now convert the delalloc reservation into an /unwritten/ allocated extent in the cow fork. The da conversion code tries to opportunistically allocate as much of a (speculatively prealloc'd) extent as possible, so we may end up allocating a larger extent than we're actually writing out: D: --RRRRRRSSSRRRRRRRR--- U: ------UUUUUUU--------- Next, we convert only the part of the extent that we're actively planning to write to normal (i.e. not unwritten) status: D: --RRRRRRSSSRRRRRRRR--- U: ------UURRUUU--------- If the write succeeds, the end_cow function will now scan the relevant range of the CoW fork for real extents and remap only the real extents into the data fork: D: --RRRRRRRRSRRRRRRRR--- U: ------UU--UUU--------- This ensures that we never obliterate valid data fork extents with unwritten blocks from the CoW fork. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Christoph Hellwig <hch@lst.de>
2017-02-03 07:14:02 +08:00
ASSERT(count != 0);
xfs: mark speculative prealloc CoW fork extents unwritten Christoph Hellwig pointed out that there's a potentially nasty race when performing simultaneous nearby directio cow writes: "Thread 1 writes a range from B to c " B --------- C p "a little later thread 2 writes from A to B " A --------- B p [editor's note: the 'p' denote cowextsize boundaries, which I added to make this more clear] "but the code preallocates beyond B into the range where thread "1 has just written, but ->end_io hasn't been called yet. "But once ->end_io is called thread 2 has already allocated "up to the extent size hint into the write range of thread 1, "so the end_io handler will splice the unintialized blocks from "that preallocation back into the file right after B." We can avoid this race by ensuring that thread 1 cannot accidentally remap the blocks that thread 2 allocated (as part of speculative preallocation) as part of t2's write preparation in t1's end_io handler. The way we make this happen is by taking advantage of the unwritten extent flag as an intermediate step. Recall that when we begin the process of writing data to shared blocks, we create a delayed allocation extent in the CoW fork: D: --RRRRRRSSSRRRRRRRR--- C: ------DDDDDDD--------- When a thread prepares to CoW some dirty data out to disk, it will now convert the delalloc reservation into an /unwritten/ allocated extent in the cow fork. The da conversion code tries to opportunistically allocate as much of a (speculatively prealloc'd) extent as possible, so we may end up allocating a larger extent than we're actually writing out: D: --RRRRRRSSSRRRRRRRR--- U: ------UUUUUUU--------- Next, we convert only the part of the extent that we're actively planning to write to normal (i.e. not unwritten) status: D: --RRRRRRSSSRRRRRRRR--- U: ------UURRUUU--------- If the write succeeds, the end_cow function will now scan the relevant range of the CoW fork for real extents and remap only the real extents into the data fork: D: --RRRRRRRRSRRRRRRRR--- U: ------UU--UUU--------- This ensures that we never obliterate valid data fork extents with unwritten blocks from the CoW fork. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Christoph Hellwig <hch@lst.de>
2017-02-03 07:14:02 +08:00
xfs_ilock(ip, XFS_ILOCK_EXCL);
error = xfs_reflink_convert_cow_locked(ip, offset_fsb, count_fsb);
xfs: mark speculative prealloc CoW fork extents unwritten Christoph Hellwig pointed out that there's a potentially nasty race when performing simultaneous nearby directio cow writes: "Thread 1 writes a range from B to c " B --------- C p "a little later thread 2 writes from A to B " A --------- B p [editor's note: the 'p' denote cowextsize boundaries, which I added to make this more clear] "but the code preallocates beyond B into the range where thread "1 has just written, but ->end_io hasn't been called yet. "But once ->end_io is called thread 2 has already allocated "up to the extent size hint into the write range of thread 1, "so the end_io handler will splice the unintialized blocks from "that preallocation back into the file right after B." We can avoid this race by ensuring that thread 1 cannot accidentally remap the blocks that thread 2 allocated (as part of speculative preallocation) as part of t2's write preparation in t1's end_io handler. The way we make this happen is by taking advantage of the unwritten extent flag as an intermediate step. Recall that when we begin the process of writing data to shared blocks, we create a delayed allocation extent in the CoW fork: D: --RRRRRRSSSRRRRRRRR--- C: ------DDDDDDD--------- When a thread prepares to CoW some dirty data out to disk, it will now convert the delalloc reservation into an /unwritten/ allocated extent in the cow fork. The da conversion code tries to opportunistically allocate as much of a (speculatively prealloc'd) extent as possible, so we may end up allocating a larger extent than we're actually writing out: D: --RRRRRRSSSRRRRRRRR--- U: ------UUUUUUU--------- Next, we convert only the part of the extent that we're actively planning to write to normal (i.e. not unwritten) status: D: --RRRRRRSSSRRRRRRRR--- U: ------UURRUUU--------- If the write succeeds, the end_cow function will now scan the relevant range of the CoW fork for real extents and remap only the real extents into the data fork: D: --RRRRRRRRSRRRRRRRR--- U: ------UU--UUU--------- This ensures that we never obliterate valid data fork extents with unwritten blocks from the CoW fork. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Christoph Hellwig <hch@lst.de>
2017-02-03 07:14:02 +08:00
xfs_iunlock(ip, XFS_ILOCK_EXCL);
return error;
}
xfs: fix transaction leak in xfs_reflink_allocate_cow() When xfs_reflink_allocate_cow() allocates a transaction, it drops the ILOCK to perform the operation. This Introduces a race condition where another thread modifying the file can perform the COW allocation operation underneath us. This result in the retry loop finding an allocated block and jumping straight to the conversion code. It does not, however, cancel the transaction it holds and so this gets leaked. This results in a lockdep warning: ================================================ WARNING: lock held when returning to user space! 4.18.5 #1 Not tainted ------------------------------------------------ worker/6123 is leaving the kernel with locks still held! 1 lock held by worker/6123: #0: 000000009eab4f1b (sb_internal#2){.+.+}, at: xfs_trans_alloc+0x17c/0x220 And eventually the filesystem deadlocks because it runs out of log space that is reserved by the leaked transaction and never gets released. The logic flow in xfs_reflink_allocate_cow() is a convoluted mess of gotos - it's no surprise that it has bug where the flow through several goto jumps then fails to clean up context from a non-obvious logic path. CLean up the logic flow and make sure every path does the right thing. Reported-by: Alexander Y. Fomichev <git.user@gmail.com> Tested-by: Alexander Y. Fomichev <git.user@gmail.com> Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=200981 Signed-off-by: Dave Chinner <dchinner@redhat.com> [hch: slight refactor] Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Dave Chinner <david@fromorbit.com>
2018-09-29 11:47:15 +08:00
/*
* Find the extent that maps the given range in the COW fork. Even if the extent
* is not shared we might have a preallocation for it in the COW fork. If so we
* use it that rather than trigger a new allocation.
*/
static int
xfs_find_trim_cow_extent(
struct xfs_inode *ip,
struct xfs_bmbt_irec *imap,
struct xfs_bmbt_irec *cmap,
xfs: fix transaction leak in xfs_reflink_allocate_cow() When xfs_reflink_allocate_cow() allocates a transaction, it drops the ILOCK to perform the operation. This Introduces a race condition where another thread modifying the file can perform the COW allocation operation underneath us. This result in the retry loop finding an allocated block and jumping straight to the conversion code. It does not, however, cancel the transaction it holds and so this gets leaked. This results in a lockdep warning: ================================================ WARNING: lock held when returning to user space! 4.18.5 #1 Not tainted ------------------------------------------------ worker/6123 is leaving the kernel with locks still held! 1 lock held by worker/6123: #0: 000000009eab4f1b (sb_internal#2){.+.+}, at: xfs_trans_alloc+0x17c/0x220 And eventually the filesystem deadlocks because it runs out of log space that is reserved by the leaked transaction and never gets released. The logic flow in xfs_reflink_allocate_cow() is a convoluted mess of gotos - it's no surprise that it has bug where the flow through several goto jumps then fails to clean up context from a non-obvious logic path. CLean up the logic flow and make sure every path does the right thing. Reported-by: Alexander Y. Fomichev <git.user@gmail.com> Tested-by: Alexander Y. Fomichev <git.user@gmail.com> Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=200981 Signed-off-by: Dave Chinner <dchinner@redhat.com> [hch: slight refactor] Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Dave Chinner <david@fromorbit.com>
2018-09-29 11:47:15 +08:00
bool *shared,
bool *found)
{
xfs_fileoff_t offset_fsb = imap->br_startoff;
xfs_filblks_t count_fsb = imap->br_blockcount;
struct xfs_iext_cursor icur;
*found = false;
/*
* If we don't find an overlapping extent, trim the range we need to
* allocate to fit the hole we found.
*/
if (!xfs_iext_lookup_extent(ip, ip->i_cowfp, offset_fsb, &icur, cmap))
cmap->br_startoff = offset_fsb + count_fsb;
if (cmap->br_startoff > offset_fsb) {
xfs_trim_extent(imap, imap->br_startoff,
cmap->br_startoff - imap->br_startoff);
return xfs_bmap_trim_cow(ip, imap, shared);
}
xfs: fix transaction leak in xfs_reflink_allocate_cow() When xfs_reflink_allocate_cow() allocates a transaction, it drops the ILOCK to perform the operation. This Introduces a race condition where another thread modifying the file can perform the COW allocation operation underneath us. This result in the retry loop finding an allocated block and jumping straight to the conversion code. It does not, however, cancel the transaction it holds and so this gets leaked. This results in a lockdep warning: ================================================ WARNING: lock held when returning to user space! 4.18.5 #1 Not tainted ------------------------------------------------ worker/6123 is leaving the kernel with locks still held! 1 lock held by worker/6123: #0: 000000009eab4f1b (sb_internal#2){.+.+}, at: xfs_trans_alloc+0x17c/0x220 And eventually the filesystem deadlocks because it runs out of log space that is reserved by the leaked transaction and never gets released. The logic flow in xfs_reflink_allocate_cow() is a convoluted mess of gotos - it's no surprise that it has bug where the flow through several goto jumps then fails to clean up context from a non-obvious logic path. CLean up the logic flow and make sure every path does the right thing. Reported-by: Alexander Y. Fomichev <git.user@gmail.com> Tested-by: Alexander Y. Fomichev <git.user@gmail.com> Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=200981 Signed-off-by: Dave Chinner <dchinner@redhat.com> [hch: slight refactor] Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Dave Chinner <david@fromorbit.com>
2018-09-29 11:47:15 +08:00
*shared = true;
if (isnullstartblock(cmap->br_startblock)) {
xfs_trim_extent(imap, cmap->br_startoff, cmap->br_blockcount);
xfs: fix transaction leak in xfs_reflink_allocate_cow() When xfs_reflink_allocate_cow() allocates a transaction, it drops the ILOCK to perform the operation. This Introduces a race condition where another thread modifying the file can perform the COW allocation operation underneath us. This result in the retry loop finding an allocated block and jumping straight to the conversion code. It does not, however, cancel the transaction it holds and so this gets leaked. This results in a lockdep warning: ================================================ WARNING: lock held when returning to user space! 4.18.5 #1 Not tainted ------------------------------------------------ worker/6123 is leaving the kernel with locks still held! 1 lock held by worker/6123: #0: 000000009eab4f1b (sb_internal#2){.+.+}, at: xfs_trans_alloc+0x17c/0x220 And eventually the filesystem deadlocks because it runs out of log space that is reserved by the leaked transaction and never gets released. The logic flow in xfs_reflink_allocate_cow() is a convoluted mess of gotos - it's no surprise that it has bug where the flow through several goto jumps then fails to clean up context from a non-obvious logic path. CLean up the logic flow and make sure every path does the right thing. Reported-by: Alexander Y. Fomichev <git.user@gmail.com> Tested-by: Alexander Y. Fomichev <git.user@gmail.com> Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=200981 Signed-off-by: Dave Chinner <dchinner@redhat.com> [hch: slight refactor] Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Dave Chinner <david@fromorbit.com>
2018-09-29 11:47:15 +08:00
return 0;
}
/* real extent found - no need to allocate */
xfs_trim_extent(cmap, offset_fsb, count_fsb);
xfs: fix transaction leak in xfs_reflink_allocate_cow() When xfs_reflink_allocate_cow() allocates a transaction, it drops the ILOCK to perform the operation. This Introduces a race condition where another thread modifying the file can perform the COW allocation operation underneath us. This result in the retry loop finding an allocated block and jumping straight to the conversion code. It does not, however, cancel the transaction it holds and so this gets leaked. This results in a lockdep warning: ================================================ WARNING: lock held when returning to user space! 4.18.5 #1 Not tainted ------------------------------------------------ worker/6123 is leaving the kernel with locks still held! 1 lock held by worker/6123: #0: 000000009eab4f1b (sb_internal#2){.+.+}, at: xfs_trans_alloc+0x17c/0x220 And eventually the filesystem deadlocks because it runs out of log space that is reserved by the leaked transaction and never gets released. The logic flow in xfs_reflink_allocate_cow() is a convoluted mess of gotos - it's no surprise that it has bug where the flow through several goto jumps then fails to clean up context from a non-obvious logic path. CLean up the logic flow and make sure every path does the right thing. Reported-by: Alexander Y. Fomichev <git.user@gmail.com> Tested-by: Alexander Y. Fomichev <git.user@gmail.com> Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=200981 Signed-off-by: Dave Chinner <dchinner@redhat.com> [hch: slight refactor] Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Dave Chinner <david@fromorbit.com>
2018-09-29 11:47:15 +08:00
*found = true;
return 0;
}
xfs: Fix false ENOSPC when performing direct write on a delalloc extent in cow fork On a higly fragmented filesystem a Direct IO write can fail with -ENOSPC error even though the filesystem has sufficient number of free blocks. This occurs if the file offset range on which the write operation is being performed has a delalloc extent in the cow fork and this delalloc extent begins much before the Direct IO range. In such a scenario, xfs_reflink_allocate_cow() invokes xfs_bmapi_write() to allocate the blocks mapped by the delalloc extent. The extent thus allocated may not cover the beginning of file offset range on which the Direct IO write was issued. Hence xfs_reflink_allocate_cow() ends up returning -ENOSPC. The following script reliably recreates the bug described above. #!/usr/bin/bash device=/dev/loop0 shortdev=$(basename $device) mntpnt=/mnt/ file1=${mntpnt}/file1 file2=${mntpnt}/file2 fragmentedfile=${mntpnt}/fragmentedfile punchprog=/root/repos/xfstests-dev/src/punch-alternating errortag=/sys/fs/xfs/${shortdev}/errortag/bmap_alloc_minlen_extent umount $device > /dev/null 2>&1 echo "Create FS" mkfs.xfs -f -m reflink=1 $device > /dev/null 2>&1 if [[ $? != 0 ]]; then echo "mkfs failed." exit 1 fi echo "Mount FS" mount $device $mntpnt > /dev/null 2>&1 if [[ $? != 0 ]]; then echo "mount failed." exit 1 fi echo "Create source file" xfs_io -f -c "pwrite 0 32M" $file1 > /dev/null 2>&1 sync echo "Create Reflinked file" xfs_io -f -c "reflink $file1" $file2 &>/dev/null echo "Set cowextsize" xfs_io -c "cowextsize 16M" $file1 > /dev/null 2>&1 echo "Fragment FS" xfs_io -f -c "pwrite 0 64M" $fragmentedfile > /dev/null 2>&1 sync $punchprog $fragmentedfile echo "Allocate block sized extent from now onwards" echo -n 1 > $errortag echo "Create 16MiB delalloc extent in CoW fork" xfs_io -c "pwrite 0 4k" $file1 > /dev/null 2>&1 sync echo "Direct I/O write at offset 12k" xfs_io -d -c "pwrite 12k 8k" $file1 This commit fixes the bug by invoking xfs_bmapi_write() in a loop until disk blocks are allocated for atleast the starting file offset of the Direct IO write range. Fixes: 3c68d44a2b49 ("xfs: allocate direct I/O COW blocks in iomap_begin") Reported-and-Root-caused-by: Wengang Wang <wen.gang.wang@oracle.com> Signed-off-by: Chandan Babu R <chandan.babu@oracle.com> Reviewed-by: Darrick J. Wong <djwong@kernel.org> [djwong: slight editing to make the locking less grody, and fix some style things] Signed-off-by: Darrick J. Wong <djwong@kernel.org>
2022-08-04 23:59:27 +08:00
static int
xfs_reflink_convert_unwritten(
struct xfs_inode *ip,
struct xfs_bmbt_irec *imap,
struct xfs_bmbt_irec *cmap,
bool convert_now)
{
xfs_fileoff_t offset_fsb = imap->br_startoff;
xfs_filblks_t count_fsb = imap->br_blockcount;
int error;
/*
* cmap might larger than imap due to cowextsize hint.
*/
xfs_trim_extent(cmap, offset_fsb, count_fsb);
/*
* COW fork extents are supposed to remain unwritten until we're ready
* to initiate a disk write. For direct I/O we are going to write the
* data and need the conversion, but for buffered writes we're done.
*/
if (!convert_now || cmap->br_state == XFS_EXT_NORM)
return 0;
trace_xfs_reflink_convert_cow(ip, cmap);
error = xfs_reflink_convert_cow_locked(ip, offset_fsb, count_fsb);
if (!error)
cmap->br_state = XFS_EXT_NORM;
return error;
}
static int
xfs_reflink_fill_cow_hole(
struct xfs_inode *ip,
struct xfs_bmbt_irec *imap,
struct xfs_bmbt_irec *cmap,
bool *shared,
uint *lockmode,
bool convert_now)
{
struct xfs_mount *mp = ip->i_mount;
xfs: fix transaction leak in xfs_reflink_allocate_cow() When xfs_reflink_allocate_cow() allocates a transaction, it drops the ILOCK to perform the operation. This Introduces a race condition where another thread modifying the file can perform the COW allocation operation underneath us. This result in the retry loop finding an allocated block and jumping straight to the conversion code. It does not, however, cancel the transaction it holds and so this gets leaked. This results in a lockdep warning: ================================================ WARNING: lock held when returning to user space! 4.18.5 #1 Not tainted ------------------------------------------------ worker/6123 is leaving the kernel with locks still held! 1 lock held by worker/6123: #0: 000000009eab4f1b (sb_internal#2){.+.+}, at: xfs_trans_alloc+0x17c/0x220 And eventually the filesystem deadlocks because it runs out of log space that is reserved by the leaked transaction and never gets released. The logic flow in xfs_reflink_allocate_cow() is a convoluted mess of gotos - it's no surprise that it has bug where the flow through several goto jumps then fails to clean up context from a non-obvious logic path. CLean up the logic flow and make sure every path does the right thing. Reported-by: Alexander Y. Fomichev <git.user@gmail.com> Tested-by: Alexander Y. Fomichev <git.user@gmail.com> Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=200981 Signed-off-by: Dave Chinner <dchinner@redhat.com> [hch: slight refactor] Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Dave Chinner <david@fromorbit.com>
2018-09-29 11:47:15 +08:00
struct xfs_trans *tp;
xfs_filblks_t resaligned;
xfs: Fix false ENOSPC when performing direct write on a delalloc extent in cow fork On a higly fragmented filesystem a Direct IO write can fail with -ENOSPC error even though the filesystem has sufficient number of free blocks. This occurs if the file offset range on which the write operation is being performed has a delalloc extent in the cow fork and this delalloc extent begins much before the Direct IO range. In such a scenario, xfs_reflink_allocate_cow() invokes xfs_bmapi_write() to allocate the blocks mapped by the delalloc extent. The extent thus allocated may not cover the beginning of file offset range on which the Direct IO write was issued. Hence xfs_reflink_allocate_cow() ends up returning -ENOSPC. The following script reliably recreates the bug described above. #!/usr/bin/bash device=/dev/loop0 shortdev=$(basename $device) mntpnt=/mnt/ file1=${mntpnt}/file1 file2=${mntpnt}/file2 fragmentedfile=${mntpnt}/fragmentedfile punchprog=/root/repos/xfstests-dev/src/punch-alternating errortag=/sys/fs/xfs/${shortdev}/errortag/bmap_alloc_minlen_extent umount $device > /dev/null 2>&1 echo "Create FS" mkfs.xfs -f -m reflink=1 $device > /dev/null 2>&1 if [[ $? != 0 ]]; then echo "mkfs failed." exit 1 fi echo "Mount FS" mount $device $mntpnt > /dev/null 2>&1 if [[ $? != 0 ]]; then echo "mount failed." exit 1 fi echo "Create source file" xfs_io -f -c "pwrite 0 32M" $file1 > /dev/null 2>&1 sync echo "Create Reflinked file" xfs_io -f -c "reflink $file1" $file2 &>/dev/null echo "Set cowextsize" xfs_io -c "cowextsize 16M" $file1 > /dev/null 2>&1 echo "Fragment FS" xfs_io -f -c "pwrite 0 64M" $fragmentedfile > /dev/null 2>&1 sync $punchprog $fragmentedfile echo "Allocate block sized extent from now onwards" echo -n 1 > $errortag echo "Create 16MiB delalloc extent in CoW fork" xfs_io -c "pwrite 0 4k" $file1 > /dev/null 2>&1 sync echo "Direct I/O write at offset 12k" xfs_io -d -c "pwrite 12k 8k" $file1 This commit fixes the bug by invoking xfs_bmapi_write() in a loop until disk blocks are allocated for atleast the starting file offset of the Direct IO write range. Fixes: 3c68d44a2b49 ("xfs: allocate direct I/O COW blocks in iomap_begin") Reported-and-Root-caused-by: Wengang Wang <wen.gang.wang@oracle.com> Signed-off-by: Chandan Babu R <chandan.babu@oracle.com> Reviewed-by: Darrick J. Wong <djwong@kernel.org> [djwong: slight editing to make the locking less grody, and fix some style things] Signed-off-by: Darrick J. Wong <djwong@kernel.org>
2022-08-04 23:59:27 +08:00
xfs_extlen_t resblks;
int nimaps;
int error;
bool found;
xfs: fix transaction leak in xfs_reflink_allocate_cow() When xfs_reflink_allocate_cow() allocates a transaction, it drops the ILOCK to perform the operation. This Introduces a race condition where another thread modifying the file can perform the COW allocation operation underneath us. This result in the retry loop finding an allocated block and jumping straight to the conversion code. It does not, however, cancel the transaction it holds and so this gets leaked. This results in a lockdep warning: ================================================ WARNING: lock held when returning to user space! 4.18.5 #1 Not tainted ------------------------------------------------ worker/6123 is leaving the kernel with locks still held! 1 lock held by worker/6123: #0: 000000009eab4f1b (sb_internal#2){.+.+}, at: xfs_trans_alloc+0x17c/0x220 And eventually the filesystem deadlocks because it runs out of log space that is reserved by the leaked transaction and never gets released. The logic flow in xfs_reflink_allocate_cow() is a convoluted mess of gotos - it's no surprise that it has bug where the flow through several goto jumps then fails to clean up context from a non-obvious logic path. CLean up the logic flow and make sure every path does the right thing. Reported-by: Alexander Y. Fomichev <git.user@gmail.com> Tested-by: Alexander Y. Fomichev <git.user@gmail.com> Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=200981 Signed-off-by: Dave Chinner <dchinner@redhat.com> [hch: slight refactor] Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Dave Chinner <david@fromorbit.com>
2018-09-29 11:47:15 +08:00
resaligned = xfs_aligned_fsb_count(imap->br_startoff,
imap->br_blockcount, xfs_get_cowextsz_hint(ip));
resblks = XFS_DIOSTRAT_SPACE_RES(mp, resaligned);
xfs: fix transaction leak in xfs_reflink_allocate_cow() When xfs_reflink_allocate_cow() allocates a transaction, it drops the ILOCK to perform the operation. This Introduces a race condition where another thread modifying the file can perform the COW allocation operation underneath us. This result in the retry loop finding an allocated block and jumping straight to the conversion code. It does not, however, cancel the transaction it holds and so this gets leaked. This results in a lockdep warning: ================================================ WARNING: lock held when returning to user space! 4.18.5 #1 Not tainted ------------------------------------------------ worker/6123 is leaving the kernel with locks still held! 1 lock held by worker/6123: #0: 000000009eab4f1b (sb_internal#2){.+.+}, at: xfs_trans_alloc+0x17c/0x220 And eventually the filesystem deadlocks because it runs out of log space that is reserved by the leaked transaction and never gets released. The logic flow in xfs_reflink_allocate_cow() is a convoluted mess of gotos - it's no surprise that it has bug where the flow through several goto jumps then fails to clean up context from a non-obvious logic path. CLean up the logic flow and make sure every path does the right thing. Reported-by: Alexander Y. Fomichev <git.user@gmail.com> Tested-by: Alexander Y. Fomichev <git.user@gmail.com> Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=200981 Signed-off-by: Dave Chinner <dchinner@redhat.com> [hch: slight refactor] Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Dave Chinner <david@fromorbit.com>
2018-09-29 11:47:15 +08:00
xfs_iunlock(ip, *lockmode);
*lockmode = 0;
error = xfs_trans_alloc_inode(ip, &M_RES(mp)->tr_write, resblks, 0,
false, &tp);
xfs: fix transaction leak in xfs_reflink_allocate_cow() When xfs_reflink_allocate_cow() allocates a transaction, it drops the ILOCK to perform the operation. This Introduces a race condition where another thread modifying the file can perform the COW allocation operation underneath us. This result in the retry loop finding an allocated block and jumping straight to the conversion code. It does not, however, cancel the transaction it holds and so this gets leaked. This results in a lockdep warning: ================================================ WARNING: lock held when returning to user space! 4.18.5 #1 Not tainted ------------------------------------------------ worker/6123 is leaving the kernel with locks still held! 1 lock held by worker/6123: #0: 000000009eab4f1b (sb_internal#2){.+.+}, at: xfs_trans_alloc+0x17c/0x220 And eventually the filesystem deadlocks because it runs out of log space that is reserved by the leaked transaction and never gets released. The logic flow in xfs_reflink_allocate_cow() is a convoluted mess of gotos - it's no surprise that it has bug where the flow through several goto jumps then fails to clean up context from a non-obvious logic path. CLean up the logic flow and make sure every path does the right thing. Reported-by: Alexander Y. Fomichev <git.user@gmail.com> Tested-by: Alexander Y. Fomichev <git.user@gmail.com> Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=200981 Signed-off-by: Dave Chinner <dchinner@redhat.com> [hch: slight refactor] Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Dave Chinner <david@fromorbit.com>
2018-09-29 11:47:15 +08:00
if (error)
return error;
*lockmode = XFS_ILOCK_EXCL;
error = xfs_find_trim_cow_extent(ip, imap, cmap, shared, &found);
xfs: fix transaction leak in xfs_reflink_allocate_cow() When xfs_reflink_allocate_cow() allocates a transaction, it drops the ILOCK to perform the operation. This Introduces a race condition where another thread modifying the file can perform the COW allocation operation underneath us. This result in the retry loop finding an allocated block and jumping straight to the conversion code. It does not, however, cancel the transaction it holds and so this gets leaked. This results in a lockdep warning: ================================================ WARNING: lock held when returning to user space! 4.18.5 #1 Not tainted ------------------------------------------------ worker/6123 is leaving the kernel with locks still held! 1 lock held by worker/6123: #0: 000000009eab4f1b (sb_internal#2){.+.+}, at: xfs_trans_alloc+0x17c/0x220 And eventually the filesystem deadlocks because it runs out of log space that is reserved by the leaked transaction and never gets released. The logic flow in xfs_reflink_allocate_cow() is a convoluted mess of gotos - it's no surprise that it has bug where the flow through several goto jumps then fails to clean up context from a non-obvious logic path. CLean up the logic flow and make sure every path does the right thing. Reported-by: Alexander Y. Fomichev <git.user@gmail.com> Tested-by: Alexander Y. Fomichev <git.user@gmail.com> Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=200981 Signed-off-by: Dave Chinner <dchinner@redhat.com> [hch: slight refactor] Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Dave Chinner <david@fromorbit.com>
2018-09-29 11:47:15 +08:00
if (error || !*shared)
goto out_trans_cancel;
xfs: Fix false ENOSPC when performing direct write on a delalloc extent in cow fork On a higly fragmented filesystem a Direct IO write can fail with -ENOSPC error even though the filesystem has sufficient number of free blocks. This occurs if the file offset range on which the write operation is being performed has a delalloc extent in the cow fork and this delalloc extent begins much before the Direct IO range. In such a scenario, xfs_reflink_allocate_cow() invokes xfs_bmapi_write() to allocate the blocks mapped by the delalloc extent. The extent thus allocated may not cover the beginning of file offset range on which the Direct IO write was issued. Hence xfs_reflink_allocate_cow() ends up returning -ENOSPC. The following script reliably recreates the bug described above. #!/usr/bin/bash device=/dev/loop0 shortdev=$(basename $device) mntpnt=/mnt/ file1=${mntpnt}/file1 file2=${mntpnt}/file2 fragmentedfile=${mntpnt}/fragmentedfile punchprog=/root/repos/xfstests-dev/src/punch-alternating errortag=/sys/fs/xfs/${shortdev}/errortag/bmap_alloc_minlen_extent umount $device > /dev/null 2>&1 echo "Create FS" mkfs.xfs -f -m reflink=1 $device > /dev/null 2>&1 if [[ $? != 0 ]]; then echo "mkfs failed." exit 1 fi echo "Mount FS" mount $device $mntpnt > /dev/null 2>&1 if [[ $? != 0 ]]; then echo "mount failed." exit 1 fi echo "Create source file" xfs_io -f -c "pwrite 0 32M" $file1 > /dev/null 2>&1 sync echo "Create Reflinked file" xfs_io -f -c "reflink $file1" $file2 &>/dev/null echo "Set cowextsize" xfs_io -c "cowextsize 16M" $file1 > /dev/null 2>&1 echo "Fragment FS" xfs_io -f -c "pwrite 0 64M" $fragmentedfile > /dev/null 2>&1 sync $punchprog $fragmentedfile echo "Allocate block sized extent from now onwards" echo -n 1 > $errortag echo "Create 16MiB delalloc extent in CoW fork" xfs_io -c "pwrite 0 4k" $file1 > /dev/null 2>&1 sync echo "Direct I/O write at offset 12k" xfs_io -d -c "pwrite 12k 8k" $file1 This commit fixes the bug by invoking xfs_bmapi_write() in a loop until disk blocks are allocated for atleast the starting file offset of the Direct IO write range. Fixes: 3c68d44a2b49 ("xfs: allocate direct I/O COW blocks in iomap_begin") Reported-and-Root-caused-by: Wengang Wang <wen.gang.wang@oracle.com> Signed-off-by: Chandan Babu R <chandan.babu@oracle.com> Reviewed-by: Darrick J. Wong <djwong@kernel.org> [djwong: slight editing to make the locking less grody, and fix some style things] Signed-off-by: Darrick J. Wong <djwong@kernel.org>
2022-08-04 23:59:27 +08:00
xfs: fix transaction leak in xfs_reflink_allocate_cow() When xfs_reflink_allocate_cow() allocates a transaction, it drops the ILOCK to perform the operation. This Introduces a race condition where another thread modifying the file can perform the COW allocation operation underneath us. This result in the retry loop finding an allocated block and jumping straight to the conversion code. It does not, however, cancel the transaction it holds and so this gets leaked. This results in a lockdep warning: ================================================ WARNING: lock held when returning to user space! 4.18.5 #1 Not tainted ------------------------------------------------ worker/6123 is leaving the kernel with locks still held! 1 lock held by worker/6123: #0: 000000009eab4f1b (sb_internal#2){.+.+}, at: xfs_trans_alloc+0x17c/0x220 And eventually the filesystem deadlocks because it runs out of log space that is reserved by the leaked transaction and never gets released. The logic flow in xfs_reflink_allocate_cow() is a convoluted mess of gotos - it's no surprise that it has bug where the flow through several goto jumps then fails to clean up context from a non-obvious logic path. CLean up the logic flow and make sure every path does the right thing. Reported-by: Alexander Y. Fomichev <git.user@gmail.com> Tested-by: Alexander Y. Fomichev <git.user@gmail.com> Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=200981 Signed-off-by: Dave Chinner <dchinner@redhat.com> [hch: slight refactor] Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Dave Chinner <david@fromorbit.com>
2018-09-29 11:47:15 +08:00
if (found) {
xfs_trans_cancel(tp);
goto convert;
}
xfs: mark speculative prealloc CoW fork extents unwritten Christoph Hellwig pointed out that there's a potentially nasty race when performing simultaneous nearby directio cow writes: "Thread 1 writes a range from B to c " B --------- C p "a little later thread 2 writes from A to B " A --------- B p [editor's note: the 'p' denote cowextsize boundaries, which I added to make this more clear] "but the code preallocates beyond B into the range where thread "1 has just written, but ->end_io hasn't been called yet. "But once ->end_io is called thread 2 has already allocated "up to the extent size hint into the write range of thread 1, "so the end_io handler will splice the unintialized blocks from "that preallocation back into the file right after B." We can avoid this race by ensuring that thread 1 cannot accidentally remap the blocks that thread 2 allocated (as part of speculative preallocation) as part of t2's write preparation in t1's end_io handler. The way we make this happen is by taking advantage of the unwritten extent flag as an intermediate step. Recall that when we begin the process of writing data to shared blocks, we create a delayed allocation extent in the CoW fork: D: --RRRRRRSSSRRRRRRRR--- C: ------DDDDDDD--------- When a thread prepares to CoW some dirty data out to disk, it will now convert the delalloc reservation into an /unwritten/ allocated extent in the cow fork. The da conversion code tries to opportunistically allocate as much of a (speculatively prealloc'd) extent as possible, so we may end up allocating a larger extent than we're actually writing out: D: --RRRRRRSSSRRRRRRRR--- U: ------UUUUUUU--------- Next, we convert only the part of the extent that we're actively planning to write to normal (i.e. not unwritten) status: D: --RRRRRRSSSRRRRRRRR--- U: ------UURRUUU--------- If the write succeeds, the end_cow function will now scan the relevant range of the CoW fork for real extents and remap only the real extents into the data fork: D: --RRRRRRRRSRRRRRRRR--- U: ------UU--UUU--------- This ensures that we never obliterate valid data fork extents with unwritten blocks from the CoW fork. Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Christoph Hellwig <hch@lst.de>
2017-02-03 07:14:02 +08:00
/* Allocate the entire reservation as unwritten blocks. */
xfs: fix transaction leak in xfs_reflink_allocate_cow() When xfs_reflink_allocate_cow() allocates a transaction, it drops the ILOCK to perform the operation. This Introduces a race condition where another thread modifying the file can perform the COW allocation operation underneath us. This result in the retry loop finding an allocated block and jumping straight to the conversion code. It does not, however, cancel the transaction it holds and so this gets leaked. This results in a lockdep warning: ================================================ WARNING: lock held when returning to user space! 4.18.5 #1 Not tainted ------------------------------------------------ worker/6123 is leaving the kernel with locks still held! 1 lock held by worker/6123: #0: 000000009eab4f1b (sb_internal#2){.+.+}, at: xfs_trans_alloc+0x17c/0x220 And eventually the filesystem deadlocks because it runs out of log space that is reserved by the leaked transaction and never gets released. The logic flow in xfs_reflink_allocate_cow() is a convoluted mess of gotos - it's no surprise that it has bug where the flow through several goto jumps then fails to clean up context from a non-obvious logic path. CLean up the logic flow and make sure every path does the right thing. Reported-by: Alexander Y. Fomichev <git.user@gmail.com> Tested-by: Alexander Y. Fomichev <git.user@gmail.com> Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=200981 Signed-off-by: Dave Chinner <dchinner@redhat.com> [hch: slight refactor] Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Dave Chinner <david@fromorbit.com>
2018-09-29 11:47:15 +08:00
nimaps = 1;
error = xfs_bmapi_write(tp, ip, imap->br_startoff, imap->br_blockcount,
xfs: don't set bmapi total block req where minleft is xfs_bmapi_write() takes a total block requirement parameter that is passed down to the block allocation code and is used to specify the total block requirement of the associated transaction. This is used to try and select an AG that can not only satisfy the requested extent allocation, but can also accommodate subsequent allocations that might be required to complete the transaction. For example, additional bmbt block allocations may be required on insertion of the resulting extent to an inode data fork. While it's important for callers to calculate and reserve such extra blocks in the transaction, it is not necessary to pass the total value to xfs_bmapi_write() in all cases. The latter automatically sets minleft to ensure that sufficient free blocks remain after the allocation attempt to expand the format of the associated inode (i.e., such as extent to btree conversion, btree splits, etc). Therefore, any callers that pass a total block requirement of the bmap mapping length plus worst case bmbt expansion essentially specify the additional reservation requirement twice. These callers can pass a total of zero to rely on the bmapi minleft policy. Beyond being superfluous, the primary motivation for this change is that the total reservation logic in the bmbt code is dubious in scenarios where minlen < maxlen and a maxlen extent cannot be allocated (which is more common for data extent allocations where contiguity is not required). The total value is based on maxlen in the xfs_bmapi_write() caller. If the bmbt code falls back to an allocation between minlen and maxlen, that allocation will not succeed until total is reset to minlen, which essentially throws away any additional reservation included in total by the caller. In addition, the total value is not reset until after alignment is dropped, which means that such callers drop alignment far too aggressively than necessary. Update all callers of xfs_bmapi_write() that pass a total block value of the mapping length plus bmbt reservation to instead pass zero and rely on xfs_bmapi_minleft() to enforce the bmbt reservation requirement. This trades off slightly less conservative AG selection for the ability to preserve alignment in more scenarios. xfs_bmapi_write() callers that incorporate unrelated or additional reservations in total beyond what is already included in minleft must continue to use the former. Signed-off-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
2019-10-22 00:26:48 +08:00
XFS_BMAPI_COWFORK | XFS_BMAPI_PREALLOC, 0, cmap,
&nimaps);
if (error)
goto out_trans_cancel;
xfs_inode_set_cowblocks_tag(ip);
error = xfs_trans_commit(tp);
if (error)
return error;
/*
* Allocation succeeded but the requested range was not even partially
* satisfied? Bail out!
*/
if (nimaps == 0)
return -ENOSPC;
xfs: Fix false ENOSPC when performing direct write on a delalloc extent in cow fork On a higly fragmented filesystem a Direct IO write can fail with -ENOSPC error even though the filesystem has sufficient number of free blocks. This occurs if the file offset range on which the write operation is being performed has a delalloc extent in the cow fork and this delalloc extent begins much before the Direct IO range. In such a scenario, xfs_reflink_allocate_cow() invokes xfs_bmapi_write() to allocate the blocks mapped by the delalloc extent. The extent thus allocated may not cover the beginning of file offset range on which the Direct IO write was issued. Hence xfs_reflink_allocate_cow() ends up returning -ENOSPC. The following script reliably recreates the bug described above. #!/usr/bin/bash device=/dev/loop0 shortdev=$(basename $device) mntpnt=/mnt/ file1=${mntpnt}/file1 file2=${mntpnt}/file2 fragmentedfile=${mntpnt}/fragmentedfile punchprog=/root/repos/xfstests-dev/src/punch-alternating errortag=/sys/fs/xfs/${shortdev}/errortag/bmap_alloc_minlen_extent umount $device > /dev/null 2>&1 echo "Create FS" mkfs.xfs -f -m reflink=1 $device > /dev/null 2>&1 if [[ $? != 0 ]]; then echo "mkfs failed." exit 1 fi echo "Mount FS" mount $device $mntpnt > /dev/null 2>&1 if [[ $? != 0 ]]; then echo "mount failed." exit 1 fi echo "Create source file" xfs_io -f -c "pwrite 0 32M" $file1 > /dev/null 2>&1 sync echo "Create Reflinked file" xfs_io -f -c "reflink $file1" $file2 &>/dev/null echo "Set cowextsize" xfs_io -c "cowextsize 16M" $file1 > /dev/null 2>&1 echo "Fragment FS" xfs_io -f -c "pwrite 0 64M" $fragmentedfile > /dev/null 2>&1 sync $punchprog $fragmentedfile echo "Allocate block sized extent from now onwards" echo -n 1 > $errortag echo "Create 16MiB delalloc extent in CoW fork" xfs_io -c "pwrite 0 4k" $file1 > /dev/null 2>&1 sync echo "Direct I/O write at offset 12k" xfs_io -d -c "pwrite 12k 8k" $file1 This commit fixes the bug by invoking xfs_bmapi_write() in a loop until disk blocks are allocated for atleast the starting file offset of the Direct IO write range. Fixes: 3c68d44a2b49 ("xfs: allocate direct I/O COW blocks in iomap_begin") Reported-and-Root-caused-by: Wengang Wang <wen.gang.wang@oracle.com> Signed-off-by: Chandan Babu R <chandan.babu@oracle.com> Reviewed-by: Darrick J. Wong <djwong@kernel.org> [djwong: slight editing to make the locking less grody, and fix some style things] Signed-off-by: Darrick J. Wong <djwong@kernel.org>
2022-08-04 23:59:27 +08:00
convert:
xfs: Fix false ENOSPC when performing direct write on a delalloc extent in cow fork On a higly fragmented filesystem a Direct IO write can fail with -ENOSPC error even though the filesystem has sufficient number of free blocks. This occurs if the file offset range on which the write operation is being performed has a delalloc extent in the cow fork and this delalloc extent begins much before the Direct IO range. In such a scenario, xfs_reflink_allocate_cow() invokes xfs_bmapi_write() to allocate the blocks mapped by the delalloc extent. The extent thus allocated may not cover the beginning of file offset range on which the Direct IO write was issued. Hence xfs_reflink_allocate_cow() ends up returning -ENOSPC. The following script reliably recreates the bug described above. #!/usr/bin/bash device=/dev/loop0 shortdev=$(basename $device) mntpnt=/mnt/ file1=${mntpnt}/file1 file2=${mntpnt}/file2 fragmentedfile=${mntpnt}/fragmentedfile punchprog=/root/repos/xfstests-dev/src/punch-alternating errortag=/sys/fs/xfs/${shortdev}/errortag/bmap_alloc_minlen_extent umount $device > /dev/null 2>&1 echo "Create FS" mkfs.xfs -f -m reflink=1 $device > /dev/null 2>&1 if [[ $? != 0 ]]; then echo "mkfs failed." exit 1 fi echo "Mount FS" mount $device $mntpnt > /dev/null 2>&1 if [[ $? != 0 ]]; then echo "mount failed." exit 1 fi echo "Create source file" xfs_io -f -c "pwrite 0 32M" $file1 > /dev/null 2>&1 sync echo "Create Reflinked file" xfs_io -f -c "reflink $file1" $file2 &>/dev/null echo "Set cowextsize" xfs_io -c "cowextsize 16M" $file1 > /dev/null 2>&1 echo "Fragment FS" xfs_io -f -c "pwrite 0 64M" $fragmentedfile > /dev/null 2>&1 sync $punchprog $fragmentedfile echo "Allocate block sized extent from now onwards" echo -n 1 > $errortag echo "Create 16MiB delalloc extent in CoW fork" xfs_io -c "pwrite 0 4k" $file1 > /dev/null 2>&1 sync echo "Direct I/O write at offset 12k" xfs_io -d -c "pwrite 12k 8k" $file1 This commit fixes the bug by invoking xfs_bmapi_write() in a loop until disk blocks are allocated for atleast the starting file offset of the Direct IO write range. Fixes: 3c68d44a2b49 ("xfs: allocate direct I/O COW blocks in iomap_begin") Reported-and-Root-caused-by: Wengang Wang <wen.gang.wang@oracle.com> Signed-off-by: Chandan Babu R <chandan.babu@oracle.com> Reviewed-by: Darrick J. Wong <djwong@kernel.org> [djwong: slight editing to make the locking less grody, and fix some style things] Signed-off-by: Darrick J. Wong <djwong@kernel.org>
2022-08-04 23:59:27 +08:00
return xfs_reflink_convert_unwritten(ip, imap, cmap, convert_now);
out_trans_cancel:
xfs_trans_cancel(tp);
return error;
xfs: Fix false ENOSPC when performing direct write on a delalloc extent in cow fork On a higly fragmented filesystem a Direct IO write can fail with -ENOSPC error even though the filesystem has sufficient number of free blocks. This occurs if the file offset range on which the write operation is being performed has a delalloc extent in the cow fork and this delalloc extent begins much before the Direct IO range. In such a scenario, xfs_reflink_allocate_cow() invokes xfs_bmapi_write() to allocate the blocks mapped by the delalloc extent. The extent thus allocated may not cover the beginning of file offset range on which the Direct IO write was issued. Hence xfs_reflink_allocate_cow() ends up returning -ENOSPC. The following script reliably recreates the bug described above. #!/usr/bin/bash device=/dev/loop0 shortdev=$(basename $device) mntpnt=/mnt/ file1=${mntpnt}/file1 file2=${mntpnt}/file2 fragmentedfile=${mntpnt}/fragmentedfile punchprog=/root/repos/xfstests-dev/src/punch-alternating errortag=/sys/fs/xfs/${shortdev}/errortag/bmap_alloc_minlen_extent umount $device > /dev/null 2>&1 echo "Create FS" mkfs.xfs -f -m reflink=1 $device > /dev/null 2>&1 if [[ $? != 0 ]]; then echo "mkfs failed." exit 1 fi echo "Mount FS" mount $device $mntpnt > /dev/null 2>&1 if [[ $? != 0 ]]; then echo "mount failed." exit 1 fi echo "Create source file" xfs_io -f -c "pwrite 0 32M" $file1 > /dev/null 2>&1 sync echo "Create Reflinked file" xfs_io -f -c "reflink $file1" $file2 &>/dev/null echo "Set cowextsize" xfs_io -c "cowextsize 16M" $file1 > /dev/null 2>&1 echo "Fragment FS" xfs_io -f -c "pwrite 0 64M" $fragmentedfile > /dev/null 2>&1 sync $punchprog $fragmentedfile echo "Allocate block sized extent from now onwards" echo -n 1 > $errortag echo "Create 16MiB delalloc extent in CoW fork" xfs_io -c "pwrite 0 4k" $file1 > /dev/null 2>&1 sync echo "Direct I/O write at offset 12k" xfs_io -d -c "pwrite 12k 8k" $file1 This commit fixes the bug by invoking xfs_bmapi_write() in a loop until disk blocks are allocated for atleast the starting file offset of the Direct IO write range. Fixes: 3c68d44a2b49 ("xfs: allocate direct I/O COW blocks in iomap_begin") Reported-and-Root-caused-by: Wengang Wang <wen.gang.wang@oracle.com> Signed-off-by: Chandan Babu R <chandan.babu@oracle.com> Reviewed-by: Darrick J. Wong <djwong@kernel.org> [djwong: slight editing to make the locking less grody, and fix some style things] Signed-off-by: Darrick J. Wong <djwong@kernel.org>
2022-08-04 23:59:27 +08:00
}
static int
xfs_reflink_fill_delalloc(
struct xfs_inode *ip,
struct xfs_bmbt_irec *imap,
struct xfs_bmbt_irec *cmap,
bool *shared,
uint *lockmode,
bool convert_now)
{
struct xfs_mount *mp = ip->i_mount;
struct xfs_trans *tp;
int nimaps;
int error;
bool found;
do {
xfs_iunlock(ip, *lockmode);
*lockmode = 0;
error = xfs_trans_alloc_inode(ip, &M_RES(mp)->tr_write, 0, 0,
false, &tp);
if (error)
return error;
*lockmode = XFS_ILOCK_EXCL;
error = xfs_find_trim_cow_extent(ip, imap, cmap, shared,
&found);
if (error || !*shared)
goto out_trans_cancel;
if (found) {
xfs_trans_cancel(tp);
break;
}
ASSERT(isnullstartblock(cmap->br_startblock) ||
cmap->br_startblock == DELAYSTARTBLOCK);
/*
* Replace delalloc reservation with an unwritten extent.
*/
nimaps = 1;
error = xfs_bmapi_write(tp, ip, cmap->br_startoff,
cmap->br_blockcount,
XFS_BMAPI_COWFORK | XFS_BMAPI_PREALLOC, 0,
cmap, &nimaps);
if (error)
goto out_trans_cancel;
xfs_inode_set_cowblocks_tag(ip);
error = xfs_trans_commit(tp);
if (error)
return error;
/*
* Allocation succeeded but the requested range was not even
* partially satisfied? Bail out!
*/
if (nimaps == 0)
return -ENOSPC;
} while (cmap->br_startoff + cmap->br_blockcount <= imap->br_startoff);
return xfs_reflink_convert_unwritten(ip, imap, cmap, convert_now);
xfs: fix transaction leak in xfs_reflink_allocate_cow() When xfs_reflink_allocate_cow() allocates a transaction, it drops the ILOCK to perform the operation. This Introduces a race condition where another thread modifying the file can perform the COW allocation operation underneath us. This result in the retry loop finding an allocated block and jumping straight to the conversion code. It does not, however, cancel the transaction it holds and so this gets leaked. This results in a lockdep warning: ================================================ WARNING: lock held when returning to user space! 4.18.5 #1 Not tainted ------------------------------------------------ worker/6123 is leaving the kernel with locks still held! 1 lock held by worker/6123: #0: 000000009eab4f1b (sb_internal#2){.+.+}, at: xfs_trans_alloc+0x17c/0x220 And eventually the filesystem deadlocks because it runs out of log space that is reserved by the leaked transaction and never gets released. The logic flow in xfs_reflink_allocate_cow() is a convoluted mess of gotos - it's no surprise that it has bug where the flow through several goto jumps then fails to clean up context from a non-obvious logic path. CLean up the logic flow and make sure every path does the right thing. Reported-by: Alexander Y. Fomichev <git.user@gmail.com> Tested-by: Alexander Y. Fomichev <git.user@gmail.com> Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=200981 Signed-off-by: Dave Chinner <dchinner@redhat.com> [hch: slight refactor] Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Dave Chinner <david@fromorbit.com>
2018-09-29 11:47:15 +08:00
out_trans_cancel:
xfs_trans_cancel(tp);
return error;
}
xfs: Fix false ENOSPC when performing direct write on a delalloc extent in cow fork On a higly fragmented filesystem a Direct IO write can fail with -ENOSPC error even though the filesystem has sufficient number of free blocks. This occurs if the file offset range on which the write operation is being performed has a delalloc extent in the cow fork and this delalloc extent begins much before the Direct IO range. In such a scenario, xfs_reflink_allocate_cow() invokes xfs_bmapi_write() to allocate the blocks mapped by the delalloc extent. The extent thus allocated may not cover the beginning of file offset range on which the Direct IO write was issued. Hence xfs_reflink_allocate_cow() ends up returning -ENOSPC. The following script reliably recreates the bug described above. #!/usr/bin/bash device=/dev/loop0 shortdev=$(basename $device) mntpnt=/mnt/ file1=${mntpnt}/file1 file2=${mntpnt}/file2 fragmentedfile=${mntpnt}/fragmentedfile punchprog=/root/repos/xfstests-dev/src/punch-alternating errortag=/sys/fs/xfs/${shortdev}/errortag/bmap_alloc_minlen_extent umount $device > /dev/null 2>&1 echo "Create FS" mkfs.xfs -f -m reflink=1 $device > /dev/null 2>&1 if [[ $? != 0 ]]; then echo "mkfs failed." exit 1 fi echo "Mount FS" mount $device $mntpnt > /dev/null 2>&1 if [[ $? != 0 ]]; then echo "mount failed." exit 1 fi echo "Create source file" xfs_io -f -c "pwrite 0 32M" $file1 > /dev/null 2>&1 sync echo "Create Reflinked file" xfs_io -f -c "reflink $file1" $file2 &>/dev/null echo "Set cowextsize" xfs_io -c "cowextsize 16M" $file1 > /dev/null 2>&1 echo "Fragment FS" xfs_io -f -c "pwrite 0 64M" $fragmentedfile > /dev/null 2>&1 sync $punchprog $fragmentedfile echo "Allocate block sized extent from now onwards" echo -n 1 > $errortag echo "Create 16MiB delalloc extent in CoW fork" xfs_io -c "pwrite 0 4k" $file1 > /dev/null 2>&1 sync echo "Direct I/O write at offset 12k" xfs_io -d -c "pwrite 12k 8k" $file1 This commit fixes the bug by invoking xfs_bmapi_write() in a loop until disk blocks are allocated for atleast the starting file offset of the Direct IO write range. Fixes: 3c68d44a2b49 ("xfs: allocate direct I/O COW blocks in iomap_begin") Reported-and-Root-caused-by: Wengang Wang <wen.gang.wang@oracle.com> Signed-off-by: Chandan Babu R <chandan.babu@oracle.com> Reviewed-by: Darrick J. Wong <djwong@kernel.org> [djwong: slight editing to make the locking less grody, and fix some style things] Signed-off-by: Darrick J. Wong <djwong@kernel.org>
2022-08-04 23:59:27 +08:00
/* Allocate all CoW reservations covering a range of blocks in a file. */
int
xfs_reflink_allocate_cow(
struct xfs_inode *ip,
struct xfs_bmbt_irec *imap,
struct xfs_bmbt_irec *cmap,
bool *shared,
uint *lockmode,
bool convert_now)
{
int error;
bool found;
ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));
if (!ip->i_cowfp) {
ASSERT(!xfs_is_reflink_inode(ip));
xfs_ifork_init_cow(ip);
}
error = xfs_find_trim_cow_extent(ip, imap, cmap, shared, &found);
if (error || !*shared)
return error;
/* CoW fork has a real extent */
if (found)
return xfs_reflink_convert_unwritten(ip, imap, cmap,
convert_now);
/*
* CoW fork does not have an extent and data extent is shared.
* Allocate a real extent in the CoW fork.
*/
if (cmap->br_startoff > imap->br_startoff)
return xfs_reflink_fill_cow_hole(ip, imap, cmap, shared,
lockmode, convert_now);
/*
* CoW fork has a delalloc reservation. Replace it with a real extent.
* There may or may not be a data fork mapping.
*/
if (isnullstartblock(cmap->br_startblock) ||
cmap->br_startblock == DELAYSTARTBLOCK)
return xfs_reflink_fill_delalloc(ip, imap, cmap, shared,
lockmode, convert_now);
/* Shouldn't get here. */
ASSERT(0);
return -EFSCORRUPTED;
}
/*
* Cancel CoW reservations for some block range of an inode.
*
* If cancel_real is true this function cancels all COW fork extents for the
* inode; if cancel_real is false, real extents are not cleared.
*
* Caller must have already joined the inode to the current transaction. The
* inode will be joined to the transaction returned to the caller.
*/
int
xfs_reflink_cancel_cow_blocks(
struct xfs_inode *ip,
struct xfs_trans **tpp,
xfs_fileoff_t offset_fsb,
xfs_fileoff_t end_fsb,
bool cancel_real)
{
struct xfs_ifork *ifp = xfs_ifork_ptr(ip, XFS_COW_FORK);
struct xfs_bmbt_irec got, del;
struct xfs_iext_cursor icur;
int error = 0;
if (!xfs_inode_has_cow_data(ip))
return 0;
if (!xfs_iext_lookup_extent_before(ip, ifp, &end_fsb, &icur, &got))
return 0;
/* Walk backwards until we're out of the I/O range... */
while (got.br_startoff + got.br_blockcount > offset_fsb) {
del = got;
xfs_trim_extent(&del, offset_fsb, end_fsb - offset_fsb);
/* Extent delete may have bumped ext forward */
if (!del.br_blockcount) {
xfs_iext_prev(ifp, &icur);
goto next_extent;
}
trace_xfs_reflink_cancel_cow(ip, &del);
if (isnullstartblock(del.br_startblock)) {
error = xfs_bmap_del_extent_delay(ip, XFS_COW_FORK,
&icur, &got, &del);
if (error)
break;
} else if (del.br_state == XFS_EXT_UNWRITTEN || cancel_real) {
ASSERT((*tpp)->t_highest_agno == NULLAGNUMBER);
/* Free the CoW orphan record. */
xfs_refcount_free_cow_extent(*tpp, del.br_startblock,
del.br_blockcount);
error = xfs_free_extent_later(*tpp, del.br_startblock,
xfs: use deferred frees for btree block freeing Btrees that aren't freespace management trees use the normal extent allocation and freeing routines for their blocks. Hence when a btree block is freed, a direct call to xfs_free_extent() is made and the extent is immediately freed. This puts the entire free space management btrees under this path, so we are stacking btrees on btrees in the call stack. The inobt, finobt and refcount btrees all do this. However, the bmap btree does not do this - it calls xfs_free_extent_later() to defer the extent free operation via an XEFI and hence it gets processed in deferred operation processing during the commit of the primary transaction (i.e. via intent chaining). We need to change xfs_free_extent() to behave in a non-blocking manner so that we can avoid deadlocks with busy extents near ENOSPC in transactions that free multiple extents. Inserting or removing a record from a btree can cause a multi-level tree merge operation and that will free multiple blocks from the btree in a single transaction. i.e. we can call xfs_free_extent() multiple times, and hence the btree manipulation transaction is vulnerable to this busy extent deadlock vector. To fix this, convert all the remaining callers of xfs_free_extent() to use xfs_free_extent_later() to queue XEFIs and hence defer processing of the extent frees to a context that can be safely restarted if a deadlock condition is detected. Signed-off-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Darrick J. Wong <djwong@kernel.org> Signed-off-by: Darrick J. Wong <djwong@kernel.org> Reviewed-by: Chandan Babu R <chandan.babu@oracle.com>
2023-06-29 02:04:32 +08:00
del.br_blockcount, NULL,
XFS_AG_RESV_NONE, false);
if (error)
break;
/* Roll the transaction */
error = xfs_defer_finish(tpp);
if (error)
break;
/* Remove the mapping from the CoW fork. */
xfs_bmap_del_extent_cow(ip, &icur, &got, &del);
/* Remove the quota reservation */
error = xfs_quota_unreserve_blkres(ip,
del.br_blockcount);
if (error)
break;
} else {
/* Didn't do anything, push cursor back. */
xfs_iext_prev(ifp, &icur);
}
next_extent:
if (!xfs_iext_get_extent(ifp, &icur, &got))
break;
}
/* clear tag if cow fork is emptied */
if (!ifp->if_bytes)
xfs_inode_clear_cowblocks_tag(ip);
return error;
}
/*
* Cancel CoW reservations for some byte range of an inode.
*
* If cancel_real is true this function cancels all COW fork extents for the
* inode; if cancel_real is false, real extents are not cleared.
*/
int
xfs_reflink_cancel_cow_range(
struct xfs_inode *ip,
xfs_off_t offset,
xfs_off_t count,
bool cancel_real)
{
struct xfs_trans *tp;
xfs_fileoff_t offset_fsb;
xfs_fileoff_t end_fsb;
int error;
trace_xfs_reflink_cancel_cow_range(ip, offset, count);
xfs: introduce an always_cow mode Add a mode where XFS never overwrites existing blocks in place. This is to aid debugging our COW code, and also put infatructure in place for things like possible future support for zoned block devices, which can't support overwrites. This mode is enabled globally by doing a: echo 1 > /sys/fs/xfs/debug/always_cow Note that the parameter is global to allow running all tests in xfstests easily in this mode, which would not easily be possible with a per-fs sysfs file. In always_cow mode persistent preallocations are disabled, and fallocate will fail when called with a 0 mode (with our without FALLOC_FL_KEEP_SIZE), and not create unwritten extent for zeroed space when called with FALLOC_FL_ZERO_RANGE or FALLOC_FL_UNSHARE_RANGE. There are a few interesting xfstests failures when run in always_cow mode: - generic/392 fails because the bytes used in the file used to test hole punch recovery are less after the log replay. This is because the blocks written and then punched out are only freed with a delay due to the logging mechanism. - xfs/170 will fail as the already fragile file streams mechanism doesn't seem to interact well with the COW allocator - xfs/180 xfs/182 xfs/192 xfs/198 xfs/204 and xfs/208 will claim the file system is badly fragmented, but there is not much we can do to avoid that when always writing out of place - xfs/205 fails because overwriting a file in always_cow mode will require new space allocation and the assumption in the test thus don't work anymore. - xfs/326 fails to modify the file at all in always_cow mode after injecting the refcount error, leading to an unexpected md5sum after the remount, but that again is expected Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
2019-02-19 01:38:49 +08:00
ASSERT(ip->i_cowfp);
offset_fsb = XFS_B_TO_FSBT(ip->i_mount, offset);
if (count == NULLFILEOFF)
end_fsb = NULLFILEOFF;
else
end_fsb = XFS_B_TO_FSB(ip->i_mount, offset + count);
/* Start a rolling transaction to remove the mappings */
error = xfs_trans_alloc(ip->i_mount, &M_RES(ip->i_mount)->tr_write,
0, 0, 0, &tp);
if (error)
goto out;
xfs_ilock(ip, XFS_ILOCK_EXCL);
xfs_trans_ijoin(tp, ip, 0);
/* Scrape out the old CoW reservations */
error = xfs_reflink_cancel_cow_blocks(ip, &tp, offset_fsb, end_fsb,
cancel_real);
if (error)
goto out_cancel;
error = xfs_trans_commit(tp);
xfs_iunlock(ip, XFS_ILOCK_EXCL);
return error;
out_cancel:
xfs_trans_cancel(tp);
xfs_iunlock(ip, XFS_ILOCK_EXCL);
out:
trace_xfs_reflink_cancel_cow_range_error(ip, error, _RET_IP_);
return error;
}
/*
xfs: split up the xfs_reflink_end_cow work into smaller transactions In xfs_reflink_end_cow, we allocate a single transaction for the entire end_cow operation and then loop the CoW fork mappings to move them to the data fork. This design fails on a heavily fragmented filesystem where an inode's data fork has exactly one more extent than would fit in an extents-format fork, because the unmap can collapse the data fork into extents format (freeing the bmbt block) but the remap can expand the data fork back into a (newly allocated) bmbt block. If the number of extents we end up remapping is large, we can overflow the block reservation because we reserved blocks assuming that we were adding mappings into an already-cleared area of the data fork. Let's say we have 8 extents in the data fork, 8 extents in the CoW fork, and the data fork can hold at most 7 extents before needing to convert to btree format; and that blocks A-P are discontiguous single-block extents: 0......7 D: ABCDEFGH C: IJKLMNOP When a write to file blocks 0-7 completes, we must remap I-P into the data fork. We start by removing H from the btree-format data fork. Now we have 7 extents, so we convert the fork to extents format, freeing the bmbt block. We then move P into the data fork and it now has 8 extents again. We must convert the data fork back to btree format, requiring a block allocation. If we repeat this sequence for blocks 6-5-4-3-2-1-0, we'll need a total of 8 block allocations to remap all 8 blocks. We reserved only enough blocks to handle one btree split (5 blocks on a 4k block filesystem), which means we overflow the block reservation. To fix this issue, create a separate helper function to remap a single extent, and change _reflink_end_cow to call it in a tight loop over the entire range we're completing. As a side effect this also removes the size restrictions on how many extents we can end_cow at a time, though nobody ever hit that. It is not reasonable to reserve N blocks to remap N blocks. Note that this can be reproduced after ~320 million fsx ops while running generic/938 (long soak directio fsx exerciser): XFS: Assertion failed: tp->t_blk_res >= tp->t_blk_res_used, file: fs/xfs/xfs_trans.c, line: 116 <machine registers snipped> Call Trace: xfs_trans_dup+0x211/0x250 [xfs] xfs_trans_roll+0x6d/0x180 [xfs] xfs_defer_trans_roll+0x10c/0x3b0 [xfs] xfs_defer_finish_noroll+0xdf/0x740 [xfs] xfs_defer_finish+0x13/0x70 [xfs] xfs_reflink_end_cow+0x2c6/0x680 [xfs] xfs_dio_write_end_io+0x115/0x220 [xfs] iomap_dio_complete+0x3f/0x130 iomap_dio_rw+0x3c3/0x420 xfs_file_dio_aio_write+0x132/0x3c0 [xfs] xfs_file_write_iter+0x8b/0xc0 [xfs] __vfs_write+0x193/0x1f0 vfs_write+0xba/0x1c0 ksys_write+0x52/0xc0 do_syscall_64+0x50/0x160 entry_SYSCALL_64_after_hwframe+0x49/0xbe Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Brian Foster <bfoster@redhat.com>
2018-12-13 00:46:19 +08:00
* Remap part of the CoW fork into the data fork.
*
* We aim to remap the range starting at @offset_fsb and ending at @end_fsb
* into the data fork; this function will remap what it can (at the end of the
* range) and update @end_fsb appropriately. Each remap gets its own
* transaction because we can end up merging and splitting bmbt blocks for
* every remap operation and we'd like to keep the block reservation
* requirements as low as possible.
*/
xfs: split up the xfs_reflink_end_cow work into smaller transactions In xfs_reflink_end_cow, we allocate a single transaction for the entire end_cow operation and then loop the CoW fork mappings to move them to the data fork. This design fails on a heavily fragmented filesystem where an inode's data fork has exactly one more extent than would fit in an extents-format fork, because the unmap can collapse the data fork into extents format (freeing the bmbt block) but the remap can expand the data fork back into a (newly allocated) bmbt block. If the number of extents we end up remapping is large, we can overflow the block reservation because we reserved blocks assuming that we were adding mappings into an already-cleared area of the data fork. Let's say we have 8 extents in the data fork, 8 extents in the CoW fork, and the data fork can hold at most 7 extents before needing to convert to btree format; and that blocks A-P are discontiguous single-block extents: 0......7 D: ABCDEFGH C: IJKLMNOP When a write to file blocks 0-7 completes, we must remap I-P into the data fork. We start by removing H from the btree-format data fork. Now we have 7 extents, so we convert the fork to extents format, freeing the bmbt block. We then move P into the data fork and it now has 8 extents again. We must convert the data fork back to btree format, requiring a block allocation. If we repeat this sequence for blocks 6-5-4-3-2-1-0, we'll need a total of 8 block allocations to remap all 8 blocks. We reserved only enough blocks to handle one btree split (5 blocks on a 4k block filesystem), which means we overflow the block reservation. To fix this issue, create a separate helper function to remap a single extent, and change _reflink_end_cow to call it in a tight loop over the entire range we're completing. As a side effect this also removes the size restrictions on how many extents we can end_cow at a time, though nobody ever hit that. It is not reasonable to reserve N blocks to remap N blocks. Note that this can be reproduced after ~320 million fsx ops while running generic/938 (long soak directio fsx exerciser): XFS: Assertion failed: tp->t_blk_res >= tp->t_blk_res_used, file: fs/xfs/xfs_trans.c, line: 116 <machine registers snipped> Call Trace: xfs_trans_dup+0x211/0x250 [xfs] xfs_trans_roll+0x6d/0x180 [xfs] xfs_defer_trans_roll+0x10c/0x3b0 [xfs] xfs_defer_finish_noroll+0xdf/0x740 [xfs] xfs_defer_finish+0x13/0x70 [xfs] xfs_reflink_end_cow+0x2c6/0x680 [xfs] xfs_dio_write_end_io+0x115/0x220 [xfs] iomap_dio_complete+0x3f/0x130 iomap_dio_rw+0x3c3/0x420 xfs_file_dio_aio_write+0x132/0x3c0 [xfs] xfs_file_write_iter+0x8b/0xc0 [xfs] __vfs_write+0x193/0x1f0 vfs_write+0xba/0x1c0 ksys_write+0x52/0xc0 do_syscall_64+0x50/0x160 entry_SYSCALL_64_after_hwframe+0x49/0xbe Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Brian Foster <bfoster@redhat.com>
2018-12-13 00:46:19 +08:00
STATIC int
xfs_reflink_end_cow_extent(
struct xfs_inode *ip,
xfs_fileoff_t *offset_fsb,
xfs_fileoff_t end_fsb)
{
xfs: split up the xfs_reflink_end_cow work into smaller transactions In xfs_reflink_end_cow, we allocate a single transaction for the entire end_cow operation and then loop the CoW fork mappings to move them to the data fork. This design fails on a heavily fragmented filesystem where an inode's data fork has exactly one more extent than would fit in an extents-format fork, because the unmap can collapse the data fork into extents format (freeing the bmbt block) but the remap can expand the data fork back into a (newly allocated) bmbt block. If the number of extents we end up remapping is large, we can overflow the block reservation because we reserved blocks assuming that we were adding mappings into an already-cleared area of the data fork. Let's say we have 8 extents in the data fork, 8 extents in the CoW fork, and the data fork can hold at most 7 extents before needing to convert to btree format; and that blocks A-P are discontiguous single-block extents: 0......7 D: ABCDEFGH C: IJKLMNOP When a write to file blocks 0-7 completes, we must remap I-P into the data fork. We start by removing H from the btree-format data fork. Now we have 7 extents, so we convert the fork to extents format, freeing the bmbt block. We then move P into the data fork and it now has 8 extents again. We must convert the data fork back to btree format, requiring a block allocation. If we repeat this sequence for blocks 6-5-4-3-2-1-0, we'll need a total of 8 block allocations to remap all 8 blocks. We reserved only enough blocks to handle one btree split (5 blocks on a 4k block filesystem), which means we overflow the block reservation. To fix this issue, create a separate helper function to remap a single extent, and change _reflink_end_cow to call it in a tight loop over the entire range we're completing. As a side effect this also removes the size restrictions on how many extents we can end_cow at a time, though nobody ever hit that. It is not reasonable to reserve N blocks to remap N blocks. Note that this can be reproduced after ~320 million fsx ops while running generic/938 (long soak directio fsx exerciser): XFS: Assertion failed: tp->t_blk_res >= tp->t_blk_res_used, file: fs/xfs/xfs_trans.c, line: 116 <machine registers snipped> Call Trace: xfs_trans_dup+0x211/0x250 [xfs] xfs_trans_roll+0x6d/0x180 [xfs] xfs_defer_trans_roll+0x10c/0x3b0 [xfs] xfs_defer_finish_noroll+0xdf/0x740 [xfs] xfs_defer_finish+0x13/0x70 [xfs] xfs_reflink_end_cow+0x2c6/0x680 [xfs] xfs_dio_write_end_io+0x115/0x220 [xfs] iomap_dio_complete+0x3f/0x130 iomap_dio_rw+0x3c3/0x420 xfs_file_dio_aio_write+0x132/0x3c0 [xfs] xfs_file_write_iter+0x8b/0xc0 [xfs] __vfs_write+0x193/0x1f0 vfs_write+0xba/0x1c0 ksys_write+0x52/0xc0 do_syscall_64+0x50/0x160 entry_SYSCALL_64_after_hwframe+0x49/0xbe Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Brian Foster <bfoster@redhat.com>
2018-12-13 00:46:19 +08:00
struct xfs_iext_cursor icur;
struct xfs_bmbt_irec got, del, data;
xfs: split up the xfs_reflink_end_cow work into smaller transactions In xfs_reflink_end_cow, we allocate a single transaction for the entire end_cow operation and then loop the CoW fork mappings to move them to the data fork. This design fails on a heavily fragmented filesystem where an inode's data fork has exactly one more extent than would fit in an extents-format fork, because the unmap can collapse the data fork into extents format (freeing the bmbt block) but the remap can expand the data fork back into a (newly allocated) bmbt block. If the number of extents we end up remapping is large, we can overflow the block reservation because we reserved blocks assuming that we were adding mappings into an already-cleared area of the data fork. Let's say we have 8 extents in the data fork, 8 extents in the CoW fork, and the data fork can hold at most 7 extents before needing to convert to btree format; and that blocks A-P are discontiguous single-block extents: 0......7 D: ABCDEFGH C: IJKLMNOP When a write to file blocks 0-7 completes, we must remap I-P into the data fork. We start by removing H from the btree-format data fork. Now we have 7 extents, so we convert the fork to extents format, freeing the bmbt block. We then move P into the data fork and it now has 8 extents again. We must convert the data fork back to btree format, requiring a block allocation. If we repeat this sequence for blocks 6-5-4-3-2-1-0, we'll need a total of 8 block allocations to remap all 8 blocks. We reserved only enough blocks to handle one btree split (5 blocks on a 4k block filesystem), which means we overflow the block reservation. To fix this issue, create a separate helper function to remap a single extent, and change _reflink_end_cow to call it in a tight loop over the entire range we're completing. As a side effect this also removes the size restrictions on how many extents we can end_cow at a time, though nobody ever hit that. It is not reasonable to reserve N blocks to remap N blocks. Note that this can be reproduced after ~320 million fsx ops while running generic/938 (long soak directio fsx exerciser): XFS: Assertion failed: tp->t_blk_res >= tp->t_blk_res_used, file: fs/xfs/xfs_trans.c, line: 116 <machine registers snipped> Call Trace: xfs_trans_dup+0x211/0x250 [xfs] xfs_trans_roll+0x6d/0x180 [xfs] xfs_defer_trans_roll+0x10c/0x3b0 [xfs] xfs_defer_finish_noroll+0xdf/0x740 [xfs] xfs_defer_finish+0x13/0x70 [xfs] xfs_reflink_end_cow+0x2c6/0x680 [xfs] xfs_dio_write_end_io+0x115/0x220 [xfs] iomap_dio_complete+0x3f/0x130 iomap_dio_rw+0x3c3/0x420 xfs_file_dio_aio_write+0x132/0x3c0 [xfs] xfs_file_write_iter+0x8b/0xc0 [xfs] __vfs_write+0x193/0x1f0 vfs_write+0xba/0x1c0 ksys_write+0x52/0xc0 do_syscall_64+0x50/0x160 entry_SYSCALL_64_after_hwframe+0x49/0xbe Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Brian Foster <bfoster@redhat.com>
2018-12-13 00:46:19 +08:00
struct xfs_mount *mp = ip->i_mount;
struct xfs_trans *tp;
struct xfs_ifork *ifp = xfs_ifork_ptr(ip, XFS_COW_FORK);
xfs: split up the xfs_reflink_end_cow work into smaller transactions In xfs_reflink_end_cow, we allocate a single transaction for the entire end_cow operation and then loop the CoW fork mappings to move them to the data fork. This design fails on a heavily fragmented filesystem where an inode's data fork has exactly one more extent than would fit in an extents-format fork, because the unmap can collapse the data fork into extents format (freeing the bmbt block) but the remap can expand the data fork back into a (newly allocated) bmbt block. If the number of extents we end up remapping is large, we can overflow the block reservation because we reserved blocks assuming that we were adding mappings into an already-cleared area of the data fork. Let's say we have 8 extents in the data fork, 8 extents in the CoW fork, and the data fork can hold at most 7 extents before needing to convert to btree format; and that blocks A-P are discontiguous single-block extents: 0......7 D: ABCDEFGH C: IJKLMNOP When a write to file blocks 0-7 completes, we must remap I-P into the data fork. We start by removing H from the btree-format data fork. Now we have 7 extents, so we convert the fork to extents format, freeing the bmbt block. We then move P into the data fork and it now has 8 extents again. We must convert the data fork back to btree format, requiring a block allocation. If we repeat this sequence for blocks 6-5-4-3-2-1-0, we'll need a total of 8 block allocations to remap all 8 blocks. We reserved only enough blocks to handle one btree split (5 blocks on a 4k block filesystem), which means we overflow the block reservation. To fix this issue, create a separate helper function to remap a single extent, and change _reflink_end_cow to call it in a tight loop over the entire range we're completing. As a side effect this also removes the size restrictions on how many extents we can end_cow at a time, though nobody ever hit that. It is not reasonable to reserve N blocks to remap N blocks. Note that this can be reproduced after ~320 million fsx ops while running generic/938 (long soak directio fsx exerciser): XFS: Assertion failed: tp->t_blk_res >= tp->t_blk_res_used, file: fs/xfs/xfs_trans.c, line: 116 <machine registers snipped> Call Trace: xfs_trans_dup+0x211/0x250 [xfs] xfs_trans_roll+0x6d/0x180 [xfs] xfs_defer_trans_roll+0x10c/0x3b0 [xfs] xfs_defer_finish_noroll+0xdf/0x740 [xfs] xfs_defer_finish+0x13/0x70 [xfs] xfs_reflink_end_cow+0x2c6/0x680 [xfs] xfs_dio_write_end_io+0x115/0x220 [xfs] iomap_dio_complete+0x3f/0x130 iomap_dio_rw+0x3c3/0x420 xfs_file_dio_aio_write+0x132/0x3c0 [xfs] xfs_file_write_iter+0x8b/0xc0 [xfs] __vfs_write+0x193/0x1f0 vfs_write+0xba/0x1c0 ksys_write+0x52/0xc0 do_syscall_64+0x50/0x160 entry_SYSCALL_64_after_hwframe+0x49/0xbe Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Brian Foster <bfoster@redhat.com>
2018-12-13 00:46:19 +08:00
unsigned int resblks;
int nmaps;
xfs: split up the xfs_reflink_end_cow work into smaller transactions In xfs_reflink_end_cow, we allocate a single transaction for the entire end_cow operation and then loop the CoW fork mappings to move them to the data fork. This design fails on a heavily fragmented filesystem where an inode's data fork has exactly one more extent than would fit in an extents-format fork, because the unmap can collapse the data fork into extents format (freeing the bmbt block) but the remap can expand the data fork back into a (newly allocated) bmbt block. If the number of extents we end up remapping is large, we can overflow the block reservation because we reserved blocks assuming that we were adding mappings into an already-cleared area of the data fork. Let's say we have 8 extents in the data fork, 8 extents in the CoW fork, and the data fork can hold at most 7 extents before needing to convert to btree format; and that blocks A-P are discontiguous single-block extents: 0......7 D: ABCDEFGH C: IJKLMNOP When a write to file blocks 0-7 completes, we must remap I-P into the data fork. We start by removing H from the btree-format data fork. Now we have 7 extents, so we convert the fork to extents format, freeing the bmbt block. We then move P into the data fork and it now has 8 extents again. We must convert the data fork back to btree format, requiring a block allocation. If we repeat this sequence for blocks 6-5-4-3-2-1-0, we'll need a total of 8 block allocations to remap all 8 blocks. We reserved only enough blocks to handle one btree split (5 blocks on a 4k block filesystem), which means we overflow the block reservation. To fix this issue, create a separate helper function to remap a single extent, and change _reflink_end_cow to call it in a tight loop over the entire range we're completing. As a side effect this also removes the size restrictions on how many extents we can end_cow at a time, though nobody ever hit that. It is not reasonable to reserve N blocks to remap N blocks. Note that this can be reproduced after ~320 million fsx ops while running generic/938 (long soak directio fsx exerciser): XFS: Assertion failed: tp->t_blk_res >= tp->t_blk_res_used, file: fs/xfs/xfs_trans.c, line: 116 <machine registers snipped> Call Trace: xfs_trans_dup+0x211/0x250 [xfs] xfs_trans_roll+0x6d/0x180 [xfs] xfs_defer_trans_roll+0x10c/0x3b0 [xfs] xfs_defer_finish_noroll+0xdf/0x740 [xfs] xfs_defer_finish+0x13/0x70 [xfs] xfs_reflink_end_cow+0x2c6/0x680 [xfs] xfs_dio_write_end_io+0x115/0x220 [xfs] iomap_dio_complete+0x3f/0x130 iomap_dio_rw+0x3c3/0x420 xfs_file_dio_aio_write+0x132/0x3c0 [xfs] xfs_file_write_iter+0x8b/0xc0 [xfs] __vfs_write+0x193/0x1f0 vfs_write+0xba/0x1c0 ksys_write+0x52/0xc0 do_syscall_64+0x50/0x160 entry_SYSCALL_64_after_hwframe+0x49/0xbe Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Brian Foster <bfoster@redhat.com>
2018-12-13 00:46:19 +08:00
int error;
/* No COW extents? That's easy! */
xfs: split up the xfs_reflink_end_cow work into smaller transactions In xfs_reflink_end_cow, we allocate a single transaction for the entire end_cow operation and then loop the CoW fork mappings to move them to the data fork. This design fails on a heavily fragmented filesystem where an inode's data fork has exactly one more extent than would fit in an extents-format fork, because the unmap can collapse the data fork into extents format (freeing the bmbt block) but the remap can expand the data fork back into a (newly allocated) bmbt block. If the number of extents we end up remapping is large, we can overflow the block reservation because we reserved blocks assuming that we were adding mappings into an already-cleared area of the data fork. Let's say we have 8 extents in the data fork, 8 extents in the CoW fork, and the data fork can hold at most 7 extents before needing to convert to btree format; and that blocks A-P are discontiguous single-block extents: 0......7 D: ABCDEFGH C: IJKLMNOP When a write to file blocks 0-7 completes, we must remap I-P into the data fork. We start by removing H from the btree-format data fork. Now we have 7 extents, so we convert the fork to extents format, freeing the bmbt block. We then move P into the data fork and it now has 8 extents again. We must convert the data fork back to btree format, requiring a block allocation. If we repeat this sequence for blocks 6-5-4-3-2-1-0, we'll need a total of 8 block allocations to remap all 8 blocks. We reserved only enough blocks to handle one btree split (5 blocks on a 4k block filesystem), which means we overflow the block reservation. To fix this issue, create a separate helper function to remap a single extent, and change _reflink_end_cow to call it in a tight loop over the entire range we're completing. As a side effect this also removes the size restrictions on how many extents we can end_cow at a time, though nobody ever hit that. It is not reasonable to reserve N blocks to remap N blocks. Note that this can be reproduced after ~320 million fsx ops while running generic/938 (long soak directio fsx exerciser): XFS: Assertion failed: tp->t_blk_res >= tp->t_blk_res_used, file: fs/xfs/xfs_trans.c, line: 116 <machine registers snipped> Call Trace: xfs_trans_dup+0x211/0x250 [xfs] xfs_trans_roll+0x6d/0x180 [xfs] xfs_defer_trans_roll+0x10c/0x3b0 [xfs] xfs_defer_finish_noroll+0xdf/0x740 [xfs] xfs_defer_finish+0x13/0x70 [xfs] xfs_reflink_end_cow+0x2c6/0x680 [xfs] xfs_dio_write_end_io+0x115/0x220 [xfs] iomap_dio_complete+0x3f/0x130 iomap_dio_rw+0x3c3/0x420 xfs_file_dio_aio_write+0x132/0x3c0 [xfs] xfs_file_write_iter+0x8b/0xc0 [xfs] __vfs_write+0x193/0x1f0 vfs_write+0xba/0x1c0 ksys_write+0x52/0xc0 do_syscall_64+0x50/0x160 entry_SYSCALL_64_after_hwframe+0x49/0xbe Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Brian Foster <bfoster@redhat.com>
2018-12-13 00:46:19 +08:00
if (ifp->if_bytes == 0) {
*offset_fsb = end_fsb;
return 0;
xfs: split up the xfs_reflink_end_cow work into smaller transactions In xfs_reflink_end_cow, we allocate a single transaction for the entire end_cow operation and then loop the CoW fork mappings to move them to the data fork. This design fails on a heavily fragmented filesystem where an inode's data fork has exactly one more extent than would fit in an extents-format fork, because the unmap can collapse the data fork into extents format (freeing the bmbt block) but the remap can expand the data fork back into a (newly allocated) bmbt block. If the number of extents we end up remapping is large, we can overflow the block reservation because we reserved blocks assuming that we were adding mappings into an already-cleared area of the data fork. Let's say we have 8 extents in the data fork, 8 extents in the CoW fork, and the data fork can hold at most 7 extents before needing to convert to btree format; and that blocks A-P are discontiguous single-block extents: 0......7 D: ABCDEFGH C: IJKLMNOP When a write to file blocks 0-7 completes, we must remap I-P into the data fork. We start by removing H from the btree-format data fork. Now we have 7 extents, so we convert the fork to extents format, freeing the bmbt block. We then move P into the data fork and it now has 8 extents again. We must convert the data fork back to btree format, requiring a block allocation. If we repeat this sequence for blocks 6-5-4-3-2-1-0, we'll need a total of 8 block allocations to remap all 8 blocks. We reserved only enough blocks to handle one btree split (5 blocks on a 4k block filesystem), which means we overflow the block reservation. To fix this issue, create a separate helper function to remap a single extent, and change _reflink_end_cow to call it in a tight loop over the entire range we're completing. As a side effect this also removes the size restrictions on how many extents we can end_cow at a time, though nobody ever hit that. It is not reasonable to reserve N blocks to remap N blocks. Note that this can be reproduced after ~320 million fsx ops while running generic/938 (long soak directio fsx exerciser): XFS: Assertion failed: tp->t_blk_res >= tp->t_blk_res_used, file: fs/xfs/xfs_trans.c, line: 116 <machine registers snipped> Call Trace: xfs_trans_dup+0x211/0x250 [xfs] xfs_trans_roll+0x6d/0x180 [xfs] xfs_defer_trans_roll+0x10c/0x3b0 [xfs] xfs_defer_finish_noroll+0xdf/0x740 [xfs] xfs_defer_finish+0x13/0x70 [xfs] xfs_reflink_end_cow+0x2c6/0x680 [xfs] xfs_dio_write_end_io+0x115/0x220 [xfs] iomap_dio_complete+0x3f/0x130 iomap_dio_rw+0x3c3/0x420 xfs_file_dio_aio_write+0x132/0x3c0 [xfs] xfs_file_write_iter+0x8b/0xc0 [xfs] __vfs_write+0x193/0x1f0 vfs_write+0xba/0x1c0 ksys_write+0x52/0xc0 do_syscall_64+0x50/0x160 entry_SYSCALL_64_after_hwframe+0x49/0xbe Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Brian Foster <bfoster@redhat.com>
2018-12-13 00:46:19 +08:00
}
xfs: split up the xfs_reflink_end_cow work into smaller transactions In xfs_reflink_end_cow, we allocate a single transaction for the entire end_cow operation and then loop the CoW fork mappings to move them to the data fork. This design fails on a heavily fragmented filesystem where an inode's data fork has exactly one more extent than would fit in an extents-format fork, because the unmap can collapse the data fork into extents format (freeing the bmbt block) but the remap can expand the data fork back into a (newly allocated) bmbt block. If the number of extents we end up remapping is large, we can overflow the block reservation because we reserved blocks assuming that we were adding mappings into an already-cleared area of the data fork. Let's say we have 8 extents in the data fork, 8 extents in the CoW fork, and the data fork can hold at most 7 extents before needing to convert to btree format; and that blocks A-P are discontiguous single-block extents: 0......7 D: ABCDEFGH C: IJKLMNOP When a write to file blocks 0-7 completes, we must remap I-P into the data fork. We start by removing H from the btree-format data fork. Now we have 7 extents, so we convert the fork to extents format, freeing the bmbt block. We then move P into the data fork and it now has 8 extents again. We must convert the data fork back to btree format, requiring a block allocation. If we repeat this sequence for blocks 6-5-4-3-2-1-0, we'll need a total of 8 block allocations to remap all 8 blocks. We reserved only enough blocks to handle one btree split (5 blocks on a 4k block filesystem), which means we overflow the block reservation. To fix this issue, create a separate helper function to remap a single extent, and change _reflink_end_cow to call it in a tight loop over the entire range we're completing. As a side effect this also removes the size restrictions on how many extents we can end_cow at a time, though nobody ever hit that. It is not reasonable to reserve N blocks to remap N blocks. Note that this can be reproduced after ~320 million fsx ops while running generic/938 (long soak directio fsx exerciser): XFS: Assertion failed: tp->t_blk_res >= tp->t_blk_res_used, file: fs/xfs/xfs_trans.c, line: 116 <machine registers snipped> Call Trace: xfs_trans_dup+0x211/0x250 [xfs] xfs_trans_roll+0x6d/0x180 [xfs] xfs_defer_trans_roll+0x10c/0x3b0 [xfs] xfs_defer_finish_noroll+0xdf/0x740 [xfs] xfs_defer_finish+0x13/0x70 [xfs] xfs_reflink_end_cow+0x2c6/0x680 [xfs] xfs_dio_write_end_io+0x115/0x220 [xfs] iomap_dio_complete+0x3f/0x130 iomap_dio_rw+0x3c3/0x420 xfs_file_dio_aio_write+0x132/0x3c0 [xfs] xfs_file_write_iter+0x8b/0xc0 [xfs] __vfs_write+0x193/0x1f0 vfs_write+0xba/0x1c0 ksys_write+0x52/0xc0 do_syscall_64+0x50/0x160 entry_SYSCALL_64_after_hwframe+0x49/0xbe Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Brian Foster <bfoster@redhat.com>
2018-12-13 00:46:19 +08:00
resblks = XFS_EXTENTADD_SPACE_RES(mp, XFS_DATA_FORK);
error = xfs_trans_alloc(mp, &M_RES(mp)->tr_write, resblks, 0,
XFS_TRANS_RESERVE, &tp);
xfs: split up the xfs_reflink_end_cow work into smaller transactions In xfs_reflink_end_cow, we allocate a single transaction for the entire end_cow operation and then loop the CoW fork mappings to move them to the data fork. This design fails on a heavily fragmented filesystem where an inode's data fork has exactly one more extent than would fit in an extents-format fork, because the unmap can collapse the data fork into extents format (freeing the bmbt block) but the remap can expand the data fork back into a (newly allocated) bmbt block. If the number of extents we end up remapping is large, we can overflow the block reservation because we reserved blocks assuming that we were adding mappings into an already-cleared area of the data fork. Let's say we have 8 extents in the data fork, 8 extents in the CoW fork, and the data fork can hold at most 7 extents before needing to convert to btree format; and that blocks A-P are discontiguous single-block extents: 0......7 D: ABCDEFGH C: IJKLMNOP When a write to file blocks 0-7 completes, we must remap I-P into the data fork. We start by removing H from the btree-format data fork. Now we have 7 extents, so we convert the fork to extents format, freeing the bmbt block. We then move P into the data fork and it now has 8 extents again. We must convert the data fork back to btree format, requiring a block allocation. If we repeat this sequence for blocks 6-5-4-3-2-1-0, we'll need a total of 8 block allocations to remap all 8 blocks. We reserved only enough blocks to handle one btree split (5 blocks on a 4k block filesystem), which means we overflow the block reservation. To fix this issue, create a separate helper function to remap a single extent, and change _reflink_end_cow to call it in a tight loop over the entire range we're completing. As a side effect this also removes the size restrictions on how many extents we can end_cow at a time, though nobody ever hit that. It is not reasonable to reserve N blocks to remap N blocks. Note that this can be reproduced after ~320 million fsx ops while running generic/938 (long soak directio fsx exerciser): XFS: Assertion failed: tp->t_blk_res >= tp->t_blk_res_used, file: fs/xfs/xfs_trans.c, line: 116 <machine registers snipped> Call Trace: xfs_trans_dup+0x211/0x250 [xfs] xfs_trans_roll+0x6d/0x180 [xfs] xfs_defer_trans_roll+0x10c/0x3b0 [xfs] xfs_defer_finish_noroll+0xdf/0x740 [xfs] xfs_defer_finish+0x13/0x70 [xfs] xfs_reflink_end_cow+0x2c6/0x680 [xfs] xfs_dio_write_end_io+0x115/0x220 [xfs] iomap_dio_complete+0x3f/0x130 iomap_dio_rw+0x3c3/0x420 xfs_file_dio_aio_write+0x132/0x3c0 [xfs] xfs_file_write_iter+0x8b/0xc0 [xfs] __vfs_write+0x193/0x1f0 vfs_write+0xba/0x1c0 ksys_write+0x52/0xc0 do_syscall_64+0x50/0x160 entry_SYSCALL_64_after_hwframe+0x49/0xbe Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Brian Foster <bfoster@redhat.com>
2018-12-13 00:46:19 +08:00
if (error)
return error;
/*
xfs: split up the xfs_reflink_end_cow work into smaller transactions In xfs_reflink_end_cow, we allocate a single transaction for the entire end_cow operation and then loop the CoW fork mappings to move them to the data fork. This design fails on a heavily fragmented filesystem where an inode's data fork has exactly one more extent than would fit in an extents-format fork, because the unmap can collapse the data fork into extents format (freeing the bmbt block) but the remap can expand the data fork back into a (newly allocated) bmbt block. If the number of extents we end up remapping is large, we can overflow the block reservation because we reserved blocks assuming that we were adding mappings into an already-cleared area of the data fork. Let's say we have 8 extents in the data fork, 8 extents in the CoW fork, and the data fork can hold at most 7 extents before needing to convert to btree format; and that blocks A-P are discontiguous single-block extents: 0......7 D: ABCDEFGH C: IJKLMNOP When a write to file blocks 0-7 completes, we must remap I-P into the data fork. We start by removing H from the btree-format data fork. Now we have 7 extents, so we convert the fork to extents format, freeing the bmbt block. We then move P into the data fork and it now has 8 extents again. We must convert the data fork back to btree format, requiring a block allocation. If we repeat this sequence for blocks 6-5-4-3-2-1-0, we'll need a total of 8 block allocations to remap all 8 blocks. We reserved only enough blocks to handle one btree split (5 blocks on a 4k block filesystem), which means we overflow the block reservation. To fix this issue, create a separate helper function to remap a single extent, and change _reflink_end_cow to call it in a tight loop over the entire range we're completing. As a side effect this also removes the size restrictions on how many extents we can end_cow at a time, though nobody ever hit that. It is not reasonable to reserve N blocks to remap N blocks. Note that this can be reproduced after ~320 million fsx ops while running generic/938 (long soak directio fsx exerciser): XFS: Assertion failed: tp->t_blk_res >= tp->t_blk_res_used, file: fs/xfs/xfs_trans.c, line: 116 <machine registers snipped> Call Trace: xfs_trans_dup+0x211/0x250 [xfs] xfs_trans_roll+0x6d/0x180 [xfs] xfs_defer_trans_roll+0x10c/0x3b0 [xfs] xfs_defer_finish_noroll+0xdf/0x740 [xfs] xfs_defer_finish+0x13/0x70 [xfs] xfs_reflink_end_cow+0x2c6/0x680 [xfs] xfs_dio_write_end_io+0x115/0x220 [xfs] iomap_dio_complete+0x3f/0x130 iomap_dio_rw+0x3c3/0x420 xfs_file_dio_aio_write+0x132/0x3c0 [xfs] xfs_file_write_iter+0x8b/0xc0 [xfs] __vfs_write+0x193/0x1f0 vfs_write+0xba/0x1c0 ksys_write+0x52/0xc0 do_syscall_64+0x50/0x160 entry_SYSCALL_64_after_hwframe+0x49/0xbe Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Brian Foster <bfoster@redhat.com>
2018-12-13 00:46:19 +08:00
* Lock the inode. We have to ijoin without automatic unlock because
* the lead transaction is the refcountbt record deletion; the data
* fork update follows as a deferred log item.
*/
xfs_ilock(ip, XFS_ILOCK_EXCL);
xfs_trans_ijoin(tp, ip, 0);
error = xfs_iext_count_may_overflow(ip, XFS_DATA_FORK,
XFS_IEXT_REFLINK_END_COW_CNT);
if (error == -EFBIG)
error = xfs_iext_count_upgrade(tp, ip,
XFS_IEXT_REFLINK_END_COW_CNT);
if (error)
goto out_cancel;
/*
* In case of racing, overlapping AIO writes no COW extents might be
* left by the time I/O completes for the loser of the race. In that
* case we are done.
*/
if (!xfs_iext_lookup_extent(ip, ifp, *offset_fsb, &icur, &got) ||
got.br_startoff >= end_fsb) {
*offset_fsb = end_fsb;
goto out_cancel;
xfs: split up the xfs_reflink_end_cow work into smaller transactions In xfs_reflink_end_cow, we allocate a single transaction for the entire end_cow operation and then loop the CoW fork mappings to move them to the data fork. This design fails on a heavily fragmented filesystem where an inode's data fork has exactly one more extent than would fit in an extents-format fork, because the unmap can collapse the data fork into extents format (freeing the bmbt block) but the remap can expand the data fork back into a (newly allocated) bmbt block. If the number of extents we end up remapping is large, we can overflow the block reservation because we reserved blocks assuming that we were adding mappings into an already-cleared area of the data fork. Let's say we have 8 extents in the data fork, 8 extents in the CoW fork, and the data fork can hold at most 7 extents before needing to convert to btree format; and that blocks A-P are discontiguous single-block extents: 0......7 D: ABCDEFGH C: IJKLMNOP When a write to file blocks 0-7 completes, we must remap I-P into the data fork. We start by removing H from the btree-format data fork. Now we have 7 extents, so we convert the fork to extents format, freeing the bmbt block. We then move P into the data fork and it now has 8 extents again. We must convert the data fork back to btree format, requiring a block allocation. If we repeat this sequence for blocks 6-5-4-3-2-1-0, we'll need a total of 8 block allocations to remap all 8 blocks. We reserved only enough blocks to handle one btree split (5 blocks on a 4k block filesystem), which means we overflow the block reservation. To fix this issue, create a separate helper function to remap a single extent, and change _reflink_end_cow to call it in a tight loop over the entire range we're completing. As a side effect this also removes the size restrictions on how many extents we can end_cow at a time, though nobody ever hit that. It is not reasonable to reserve N blocks to remap N blocks. Note that this can be reproduced after ~320 million fsx ops while running generic/938 (long soak directio fsx exerciser): XFS: Assertion failed: tp->t_blk_res >= tp->t_blk_res_used, file: fs/xfs/xfs_trans.c, line: 116 <machine registers snipped> Call Trace: xfs_trans_dup+0x211/0x250 [xfs] xfs_trans_roll+0x6d/0x180 [xfs] xfs_defer_trans_roll+0x10c/0x3b0 [xfs] xfs_defer_finish_noroll+0xdf/0x740 [xfs] xfs_defer_finish+0x13/0x70 [xfs] xfs_reflink_end_cow+0x2c6/0x680 [xfs] xfs_dio_write_end_io+0x115/0x220 [xfs] iomap_dio_complete+0x3f/0x130 iomap_dio_rw+0x3c3/0x420 xfs_file_dio_aio_write+0x132/0x3c0 [xfs] xfs_file_write_iter+0x8b/0xc0 [xfs] __vfs_write+0x193/0x1f0 vfs_write+0xba/0x1c0 ksys_write+0x52/0xc0 do_syscall_64+0x50/0x160 entry_SYSCALL_64_after_hwframe+0x49/0xbe Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Brian Foster <bfoster@redhat.com>
2018-12-13 00:46:19 +08:00
}
xfs: split up the xfs_reflink_end_cow work into smaller transactions In xfs_reflink_end_cow, we allocate a single transaction for the entire end_cow operation and then loop the CoW fork mappings to move them to the data fork. This design fails on a heavily fragmented filesystem where an inode's data fork has exactly one more extent than would fit in an extents-format fork, because the unmap can collapse the data fork into extents format (freeing the bmbt block) but the remap can expand the data fork back into a (newly allocated) bmbt block. If the number of extents we end up remapping is large, we can overflow the block reservation because we reserved blocks assuming that we were adding mappings into an already-cleared area of the data fork. Let's say we have 8 extents in the data fork, 8 extents in the CoW fork, and the data fork can hold at most 7 extents before needing to convert to btree format; and that blocks A-P are discontiguous single-block extents: 0......7 D: ABCDEFGH C: IJKLMNOP When a write to file blocks 0-7 completes, we must remap I-P into the data fork. We start by removing H from the btree-format data fork. Now we have 7 extents, so we convert the fork to extents format, freeing the bmbt block. We then move P into the data fork and it now has 8 extents again. We must convert the data fork back to btree format, requiring a block allocation. If we repeat this sequence for blocks 6-5-4-3-2-1-0, we'll need a total of 8 block allocations to remap all 8 blocks. We reserved only enough blocks to handle one btree split (5 blocks on a 4k block filesystem), which means we overflow the block reservation. To fix this issue, create a separate helper function to remap a single extent, and change _reflink_end_cow to call it in a tight loop over the entire range we're completing. As a side effect this also removes the size restrictions on how many extents we can end_cow at a time, though nobody ever hit that. It is not reasonable to reserve N blocks to remap N blocks. Note that this can be reproduced after ~320 million fsx ops while running generic/938 (long soak directio fsx exerciser): XFS: Assertion failed: tp->t_blk_res >= tp->t_blk_res_used, file: fs/xfs/xfs_trans.c, line: 116 <machine registers snipped> Call Trace: xfs_trans_dup+0x211/0x250 [xfs] xfs_trans_roll+0x6d/0x180 [xfs] xfs_defer_trans_roll+0x10c/0x3b0 [xfs] xfs_defer_finish_noroll+0xdf/0x740 [xfs] xfs_defer_finish+0x13/0x70 [xfs] xfs_reflink_end_cow+0x2c6/0x680 [xfs] xfs_dio_write_end_io+0x115/0x220 [xfs] iomap_dio_complete+0x3f/0x130 iomap_dio_rw+0x3c3/0x420 xfs_file_dio_aio_write+0x132/0x3c0 [xfs] xfs_file_write_iter+0x8b/0xc0 [xfs] __vfs_write+0x193/0x1f0 vfs_write+0xba/0x1c0 ksys_write+0x52/0xc0 do_syscall_64+0x50/0x160 entry_SYSCALL_64_after_hwframe+0x49/0xbe Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Brian Foster <bfoster@redhat.com>
2018-12-13 00:46:19 +08:00
/*
* Only remap real extents that contain data. With AIO, speculative
* preallocations can leak into the range we are called upon, and we
* need to skip them. Preserve @got for the eventual CoW fork
* deletion; from now on @del represents the mapping that we're
* actually remapping.
xfs: split up the xfs_reflink_end_cow work into smaller transactions In xfs_reflink_end_cow, we allocate a single transaction for the entire end_cow operation and then loop the CoW fork mappings to move them to the data fork. This design fails on a heavily fragmented filesystem where an inode's data fork has exactly one more extent than would fit in an extents-format fork, because the unmap can collapse the data fork into extents format (freeing the bmbt block) but the remap can expand the data fork back into a (newly allocated) bmbt block. If the number of extents we end up remapping is large, we can overflow the block reservation because we reserved blocks assuming that we were adding mappings into an already-cleared area of the data fork. Let's say we have 8 extents in the data fork, 8 extents in the CoW fork, and the data fork can hold at most 7 extents before needing to convert to btree format; and that blocks A-P are discontiguous single-block extents: 0......7 D: ABCDEFGH C: IJKLMNOP When a write to file blocks 0-7 completes, we must remap I-P into the data fork. We start by removing H from the btree-format data fork. Now we have 7 extents, so we convert the fork to extents format, freeing the bmbt block. We then move P into the data fork and it now has 8 extents again. We must convert the data fork back to btree format, requiring a block allocation. If we repeat this sequence for blocks 6-5-4-3-2-1-0, we'll need a total of 8 block allocations to remap all 8 blocks. We reserved only enough blocks to handle one btree split (5 blocks on a 4k block filesystem), which means we overflow the block reservation. To fix this issue, create a separate helper function to remap a single extent, and change _reflink_end_cow to call it in a tight loop over the entire range we're completing. As a side effect this also removes the size restrictions on how many extents we can end_cow at a time, though nobody ever hit that. It is not reasonable to reserve N blocks to remap N blocks. Note that this can be reproduced after ~320 million fsx ops while running generic/938 (long soak directio fsx exerciser): XFS: Assertion failed: tp->t_blk_res >= tp->t_blk_res_used, file: fs/xfs/xfs_trans.c, line: 116 <machine registers snipped> Call Trace: xfs_trans_dup+0x211/0x250 [xfs] xfs_trans_roll+0x6d/0x180 [xfs] xfs_defer_trans_roll+0x10c/0x3b0 [xfs] xfs_defer_finish_noroll+0xdf/0x740 [xfs] xfs_defer_finish+0x13/0x70 [xfs] xfs_reflink_end_cow+0x2c6/0x680 [xfs] xfs_dio_write_end_io+0x115/0x220 [xfs] iomap_dio_complete+0x3f/0x130 iomap_dio_rw+0x3c3/0x420 xfs_file_dio_aio_write+0x132/0x3c0 [xfs] xfs_file_write_iter+0x8b/0xc0 [xfs] __vfs_write+0x193/0x1f0 vfs_write+0xba/0x1c0 ksys_write+0x52/0xc0 do_syscall_64+0x50/0x160 entry_SYSCALL_64_after_hwframe+0x49/0xbe Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Brian Foster <bfoster@redhat.com>
2018-12-13 00:46:19 +08:00
*/
while (!xfs_bmap_is_written_extent(&got)) {
if (!xfs_iext_next_extent(ifp, &icur, &got) ||
got.br_startoff >= end_fsb) {
*offset_fsb = end_fsb;
goto out_cancel;
}
xfs: split up the xfs_reflink_end_cow work into smaller transactions In xfs_reflink_end_cow, we allocate a single transaction for the entire end_cow operation and then loop the CoW fork mappings to move them to the data fork. This design fails on a heavily fragmented filesystem where an inode's data fork has exactly one more extent than would fit in an extents-format fork, because the unmap can collapse the data fork into extents format (freeing the bmbt block) but the remap can expand the data fork back into a (newly allocated) bmbt block. If the number of extents we end up remapping is large, we can overflow the block reservation because we reserved blocks assuming that we were adding mappings into an already-cleared area of the data fork. Let's say we have 8 extents in the data fork, 8 extents in the CoW fork, and the data fork can hold at most 7 extents before needing to convert to btree format; and that blocks A-P are discontiguous single-block extents: 0......7 D: ABCDEFGH C: IJKLMNOP When a write to file blocks 0-7 completes, we must remap I-P into the data fork. We start by removing H from the btree-format data fork. Now we have 7 extents, so we convert the fork to extents format, freeing the bmbt block. We then move P into the data fork and it now has 8 extents again. We must convert the data fork back to btree format, requiring a block allocation. If we repeat this sequence for blocks 6-5-4-3-2-1-0, we'll need a total of 8 block allocations to remap all 8 blocks. We reserved only enough blocks to handle one btree split (5 blocks on a 4k block filesystem), which means we overflow the block reservation. To fix this issue, create a separate helper function to remap a single extent, and change _reflink_end_cow to call it in a tight loop over the entire range we're completing. As a side effect this also removes the size restrictions on how many extents we can end_cow at a time, though nobody ever hit that. It is not reasonable to reserve N blocks to remap N blocks. Note that this can be reproduced after ~320 million fsx ops while running generic/938 (long soak directio fsx exerciser): XFS: Assertion failed: tp->t_blk_res >= tp->t_blk_res_used, file: fs/xfs/xfs_trans.c, line: 116 <machine registers snipped> Call Trace: xfs_trans_dup+0x211/0x250 [xfs] xfs_trans_roll+0x6d/0x180 [xfs] xfs_defer_trans_roll+0x10c/0x3b0 [xfs] xfs_defer_finish_noroll+0xdf/0x740 [xfs] xfs_defer_finish+0x13/0x70 [xfs] xfs_reflink_end_cow+0x2c6/0x680 [xfs] xfs_dio_write_end_io+0x115/0x220 [xfs] iomap_dio_complete+0x3f/0x130 iomap_dio_rw+0x3c3/0x420 xfs_file_dio_aio_write+0x132/0x3c0 [xfs] xfs_file_write_iter+0x8b/0xc0 [xfs] __vfs_write+0x193/0x1f0 vfs_write+0xba/0x1c0 ksys_write+0x52/0xc0 do_syscall_64+0x50/0x160 entry_SYSCALL_64_after_hwframe+0x49/0xbe Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Brian Foster <bfoster@redhat.com>
2018-12-13 00:46:19 +08:00
}
del = got;
xfs_trim_extent(&del, *offset_fsb, end_fsb - *offset_fsb);
/* Grab the corresponding mapping in the data fork. */
nmaps = 1;
error = xfs_bmapi_read(ip, del.br_startoff, del.br_blockcount, &data,
&nmaps, 0);
xfs: split up the xfs_reflink_end_cow work into smaller transactions In xfs_reflink_end_cow, we allocate a single transaction for the entire end_cow operation and then loop the CoW fork mappings to move them to the data fork. This design fails on a heavily fragmented filesystem where an inode's data fork has exactly one more extent than would fit in an extents-format fork, because the unmap can collapse the data fork into extents format (freeing the bmbt block) but the remap can expand the data fork back into a (newly allocated) bmbt block. If the number of extents we end up remapping is large, we can overflow the block reservation because we reserved blocks assuming that we were adding mappings into an already-cleared area of the data fork. Let's say we have 8 extents in the data fork, 8 extents in the CoW fork, and the data fork can hold at most 7 extents before needing to convert to btree format; and that blocks A-P are discontiguous single-block extents: 0......7 D: ABCDEFGH C: IJKLMNOP When a write to file blocks 0-7 completes, we must remap I-P into the data fork. We start by removing H from the btree-format data fork. Now we have 7 extents, so we convert the fork to extents format, freeing the bmbt block. We then move P into the data fork and it now has 8 extents again. We must convert the data fork back to btree format, requiring a block allocation. If we repeat this sequence for blocks 6-5-4-3-2-1-0, we'll need a total of 8 block allocations to remap all 8 blocks. We reserved only enough blocks to handle one btree split (5 blocks on a 4k block filesystem), which means we overflow the block reservation. To fix this issue, create a separate helper function to remap a single extent, and change _reflink_end_cow to call it in a tight loop over the entire range we're completing. As a side effect this also removes the size restrictions on how many extents we can end_cow at a time, though nobody ever hit that. It is not reasonable to reserve N blocks to remap N blocks. Note that this can be reproduced after ~320 million fsx ops while running generic/938 (long soak directio fsx exerciser): XFS: Assertion failed: tp->t_blk_res >= tp->t_blk_res_used, file: fs/xfs/xfs_trans.c, line: 116 <machine registers snipped> Call Trace: xfs_trans_dup+0x211/0x250 [xfs] xfs_trans_roll+0x6d/0x180 [xfs] xfs_defer_trans_roll+0x10c/0x3b0 [xfs] xfs_defer_finish_noroll+0xdf/0x740 [xfs] xfs_defer_finish+0x13/0x70 [xfs] xfs_reflink_end_cow+0x2c6/0x680 [xfs] xfs_dio_write_end_io+0x115/0x220 [xfs] iomap_dio_complete+0x3f/0x130 iomap_dio_rw+0x3c3/0x420 xfs_file_dio_aio_write+0x132/0x3c0 [xfs] xfs_file_write_iter+0x8b/0xc0 [xfs] __vfs_write+0x193/0x1f0 vfs_write+0xba/0x1c0 ksys_write+0x52/0xc0 do_syscall_64+0x50/0x160 entry_SYSCALL_64_after_hwframe+0x49/0xbe Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Brian Foster <bfoster@redhat.com>
2018-12-13 00:46:19 +08:00
if (error)
goto out_cancel;
/* We can only remap the smaller of the two extent sizes. */
data.br_blockcount = min(data.br_blockcount, del.br_blockcount);
del.br_blockcount = data.br_blockcount;
trace_xfs_reflink_cow_remap_from(ip, &del);
trace_xfs_reflink_cow_remap_to(ip, &data);
if (xfs_bmap_is_real_extent(&data)) {
/*
* If the extent we're remapping is backed by storage (written
* or not), unmap the extent and drop its refcount.
*/
xfs_bmap_unmap_extent(tp, ip, &data);
xfs_refcount_decrease_extent(tp, &data);
xfs_trans_mod_dquot_byino(tp, ip, XFS_TRANS_DQ_BCOUNT,
-data.br_blockcount);
} else if (data.br_startblock == DELAYSTARTBLOCK) {
int done;
/*
* If the extent we're remapping is a delalloc reservation,
* we can use the regular bunmapi function to release the
* incore state. Dropping the delalloc reservation takes care
* of the quota reservation for us.
*/
error = xfs_bunmapi(NULL, ip, data.br_startoff,
data.br_blockcount, 0, 1, &done);
if (error)
goto out_cancel;
ASSERT(done);
}
xfs: split up the xfs_reflink_end_cow work into smaller transactions In xfs_reflink_end_cow, we allocate a single transaction for the entire end_cow operation and then loop the CoW fork mappings to move them to the data fork. This design fails on a heavily fragmented filesystem where an inode's data fork has exactly one more extent than would fit in an extents-format fork, because the unmap can collapse the data fork into extents format (freeing the bmbt block) but the remap can expand the data fork back into a (newly allocated) bmbt block. If the number of extents we end up remapping is large, we can overflow the block reservation because we reserved blocks assuming that we were adding mappings into an already-cleared area of the data fork. Let's say we have 8 extents in the data fork, 8 extents in the CoW fork, and the data fork can hold at most 7 extents before needing to convert to btree format; and that blocks A-P are discontiguous single-block extents: 0......7 D: ABCDEFGH C: IJKLMNOP When a write to file blocks 0-7 completes, we must remap I-P into the data fork. We start by removing H from the btree-format data fork. Now we have 7 extents, so we convert the fork to extents format, freeing the bmbt block. We then move P into the data fork and it now has 8 extents again. We must convert the data fork back to btree format, requiring a block allocation. If we repeat this sequence for blocks 6-5-4-3-2-1-0, we'll need a total of 8 block allocations to remap all 8 blocks. We reserved only enough blocks to handle one btree split (5 blocks on a 4k block filesystem), which means we overflow the block reservation. To fix this issue, create a separate helper function to remap a single extent, and change _reflink_end_cow to call it in a tight loop over the entire range we're completing. As a side effect this also removes the size restrictions on how many extents we can end_cow at a time, though nobody ever hit that. It is not reasonable to reserve N blocks to remap N blocks. Note that this can be reproduced after ~320 million fsx ops while running generic/938 (long soak directio fsx exerciser): XFS: Assertion failed: tp->t_blk_res >= tp->t_blk_res_used, file: fs/xfs/xfs_trans.c, line: 116 <machine registers snipped> Call Trace: xfs_trans_dup+0x211/0x250 [xfs] xfs_trans_roll+0x6d/0x180 [xfs] xfs_defer_trans_roll+0x10c/0x3b0 [xfs] xfs_defer_finish_noroll+0xdf/0x740 [xfs] xfs_defer_finish+0x13/0x70 [xfs] xfs_reflink_end_cow+0x2c6/0x680 [xfs] xfs_dio_write_end_io+0x115/0x220 [xfs] iomap_dio_complete+0x3f/0x130 iomap_dio_rw+0x3c3/0x420 xfs_file_dio_aio_write+0x132/0x3c0 [xfs] xfs_file_write_iter+0x8b/0xc0 [xfs] __vfs_write+0x193/0x1f0 vfs_write+0xba/0x1c0 ksys_write+0x52/0xc0 do_syscall_64+0x50/0x160 entry_SYSCALL_64_after_hwframe+0x49/0xbe Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Brian Foster <bfoster@redhat.com>
2018-12-13 00:46:19 +08:00
/* Free the CoW orphan record. */
xfs_refcount_free_cow_extent(tp, del.br_startblock, del.br_blockcount);
xfs: split up the xfs_reflink_end_cow work into smaller transactions In xfs_reflink_end_cow, we allocate a single transaction for the entire end_cow operation and then loop the CoW fork mappings to move them to the data fork. This design fails on a heavily fragmented filesystem where an inode's data fork has exactly one more extent than would fit in an extents-format fork, because the unmap can collapse the data fork into extents format (freeing the bmbt block) but the remap can expand the data fork back into a (newly allocated) bmbt block. If the number of extents we end up remapping is large, we can overflow the block reservation because we reserved blocks assuming that we were adding mappings into an already-cleared area of the data fork. Let's say we have 8 extents in the data fork, 8 extents in the CoW fork, and the data fork can hold at most 7 extents before needing to convert to btree format; and that blocks A-P are discontiguous single-block extents: 0......7 D: ABCDEFGH C: IJKLMNOP When a write to file blocks 0-7 completes, we must remap I-P into the data fork. We start by removing H from the btree-format data fork. Now we have 7 extents, so we convert the fork to extents format, freeing the bmbt block. We then move P into the data fork and it now has 8 extents again. We must convert the data fork back to btree format, requiring a block allocation. If we repeat this sequence for blocks 6-5-4-3-2-1-0, we'll need a total of 8 block allocations to remap all 8 blocks. We reserved only enough blocks to handle one btree split (5 blocks on a 4k block filesystem), which means we overflow the block reservation. To fix this issue, create a separate helper function to remap a single extent, and change _reflink_end_cow to call it in a tight loop over the entire range we're completing. As a side effect this also removes the size restrictions on how many extents we can end_cow at a time, though nobody ever hit that. It is not reasonable to reserve N blocks to remap N blocks. Note that this can be reproduced after ~320 million fsx ops while running generic/938 (long soak directio fsx exerciser): XFS: Assertion failed: tp->t_blk_res >= tp->t_blk_res_used, file: fs/xfs/xfs_trans.c, line: 116 <machine registers snipped> Call Trace: xfs_trans_dup+0x211/0x250 [xfs] xfs_trans_roll+0x6d/0x180 [xfs] xfs_defer_trans_roll+0x10c/0x3b0 [xfs] xfs_defer_finish_noroll+0xdf/0x740 [xfs] xfs_defer_finish+0x13/0x70 [xfs] xfs_reflink_end_cow+0x2c6/0x680 [xfs] xfs_dio_write_end_io+0x115/0x220 [xfs] iomap_dio_complete+0x3f/0x130 iomap_dio_rw+0x3c3/0x420 xfs_file_dio_aio_write+0x132/0x3c0 [xfs] xfs_file_write_iter+0x8b/0xc0 [xfs] __vfs_write+0x193/0x1f0 vfs_write+0xba/0x1c0 ksys_write+0x52/0xc0 do_syscall_64+0x50/0x160 entry_SYSCALL_64_after_hwframe+0x49/0xbe Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Brian Foster <bfoster@redhat.com>
2018-12-13 00:46:19 +08:00
/* Map the new blocks into the data fork. */
xfs_bmap_map_extent(tp, ip, &del);
xfs: split up the xfs_reflink_end_cow work into smaller transactions In xfs_reflink_end_cow, we allocate a single transaction for the entire end_cow operation and then loop the CoW fork mappings to move them to the data fork. This design fails on a heavily fragmented filesystem where an inode's data fork has exactly one more extent than would fit in an extents-format fork, because the unmap can collapse the data fork into extents format (freeing the bmbt block) but the remap can expand the data fork back into a (newly allocated) bmbt block. If the number of extents we end up remapping is large, we can overflow the block reservation because we reserved blocks assuming that we were adding mappings into an already-cleared area of the data fork. Let's say we have 8 extents in the data fork, 8 extents in the CoW fork, and the data fork can hold at most 7 extents before needing to convert to btree format; and that blocks A-P are discontiguous single-block extents: 0......7 D: ABCDEFGH C: IJKLMNOP When a write to file blocks 0-7 completes, we must remap I-P into the data fork. We start by removing H from the btree-format data fork. Now we have 7 extents, so we convert the fork to extents format, freeing the bmbt block. We then move P into the data fork and it now has 8 extents again. We must convert the data fork back to btree format, requiring a block allocation. If we repeat this sequence for blocks 6-5-4-3-2-1-0, we'll need a total of 8 block allocations to remap all 8 blocks. We reserved only enough blocks to handle one btree split (5 blocks on a 4k block filesystem), which means we overflow the block reservation. To fix this issue, create a separate helper function to remap a single extent, and change _reflink_end_cow to call it in a tight loop over the entire range we're completing. As a side effect this also removes the size restrictions on how many extents we can end_cow at a time, though nobody ever hit that. It is not reasonable to reserve N blocks to remap N blocks. Note that this can be reproduced after ~320 million fsx ops while running generic/938 (long soak directio fsx exerciser): XFS: Assertion failed: tp->t_blk_res >= tp->t_blk_res_used, file: fs/xfs/xfs_trans.c, line: 116 <machine registers snipped> Call Trace: xfs_trans_dup+0x211/0x250 [xfs] xfs_trans_roll+0x6d/0x180 [xfs] xfs_defer_trans_roll+0x10c/0x3b0 [xfs] xfs_defer_finish_noroll+0xdf/0x740 [xfs] xfs_defer_finish+0x13/0x70 [xfs] xfs_reflink_end_cow+0x2c6/0x680 [xfs] xfs_dio_write_end_io+0x115/0x220 [xfs] iomap_dio_complete+0x3f/0x130 iomap_dio_rw+0x3c3/0x420 xfs_file_dio_aio_write+0x132/0x3c0 [xfs] xfs_file_write_iter+0x8b/0xc0 [xfs] __vfs_write+0x193/0x1f0 vfs_write+0xba/0x1c0 ksys_write+0x52/0xc0 do_syscall_64+0x50/0x160 entry_SYSCALL_64_after_hwframe+0x49/0xbe Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Brian Foster <bfoster@redhat.com>
2018-12-13 00:46:19 +08:00
/* Charge this new data fork mapping to the on-disk quota. */
xfs_trans_mod_dquot_byino(tp, ip, XFS_TRANS_DQ_DELBCOUNT,
(long)del.br_blockcount);
xfs: split up the xfs_reflink_end_cow work into smaller transactions In xfs_reflink_end_cow, we allocate a single transaction for the entire end_cow operation and then loop the CoW fork mappings to move them to the data fork. This design fails on a heavily fragmented filesystem where an inode's data fork has exactly one more extent than would fit in an extents-format fork, because the unmap can collapse the data fork into extents format (freeing the bmbt block) but the remap can expand the data fork back into a (newly allocated) bmbt block. If the number of extents we end up remapping is large, we can overflow the block reservation because we reserved blocks assuming that we were adding mappings into an already-cleared area of the data fork. Let's say we have 8 extents in the data fork, 8 extents in the CoW fork, and the data fork can hold at most 7 extents before needing to convert to btree format; and that blocks A-P are discontiguous single-block extents: 0......7 D: ABCDEFGH C: IJKLMNOP When a write to file blocks 0-7 completes, we must remap I-P into the data fork. We start by removing H from the btree-format data fork. Now we have 7 extents, so we convert the fork to extents format, freeing the bmbt block. We then move P into the data fork and it now has 8 extents again. We must convert the data fork back to btree format, requiring a block allocation. If we repeat this sequence for blocks 6-5-4-3-2-1-0, we'll need a total of 8 block allocations to remap all 8 blocks. We reserved only enough blocks to handle one btree split (5 blocks on a 4k block filesystem), which means we overflow the block reservation. To fix this issue, create a separate helper function to remap a single extent, and change _reflink_end_cow to call it in a tight loop over the entire range we're completing. As a side effect this also removes the size restrictions on how many extents we can end_cow at a time, though nobody ever hit that. It is not reasonable to reserve N blocks to remap N blocks. Note that this can be reproduced after ~320 million fsx ops while running generic/938 (long soak directio fsx exerciser): XFS: Assertion failed: tp->t_blk_res >= tp->t_blk_res_used, file: fs/xfs/xfs_trans.c, line: 116 <machine registers snipped> Call Trace: xfs_trans_dup+0x211/0x250 [xfs] xfs_trans_roll+0x6d/0x180 [xfs] xfs_defer_trans_roll+0x10c/0x3b0 [xfs] xfs_defer_finish_noroll+0xdf/0x740 [xfs] xfs_defer_finish+0x13/0x70 [xfs] xfs_reflink_end_cow+0x2c6/0x680 [xfs] xfs_dio_write_end_io+0x115/0x220 [xfs] iomap_dio_complete+0x3f/0x130 iomap_dio_rw+0x3c3/0x420 xfs_file_dio_aio_write+0x132/0x3c0 [xfs] xfs_file_write_iter+0x8b/0xc0 [xfs] __vfs_write+0x193/0x1f0 vfs_write+0xba/0x1c0 ksys_write+0x52/0xc0 do_syscall_64+0x50/0x160 entry_SYSCALL_64_after_hwframe+0x49/0xbe Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Brian Foster <bfoster@redhat.com>
2018-12-13 00:46:19 +08:00
/* Remove the mapping from the CoW fork. */
xfs_bmap_del_extent_cow(ip, &icur, &got, &del);
error = xfs_trans_commit(tp);
xfs_iunlock(ip, XFS_ILOCK_EXCL);
if (error)
xfs: split up the xfs_reflink_end_cow work into smaller transactions In xfs_reflink_end_cow, we allocate a single transaction for the entire end_cow operation and then loop the CoW fork mappings to move them to the data fork. This design fails on a heavily fragmented filesystem where an inode's data fork has exactly one more extent than would fit in an extents-format fork, because the unmap can collapse the data fork into extents format (freeing the bmbt block) but the remap can expand the data fork back into a (newly allocated) bmbt block. If the number of extents we end up remapping is large, we can overflow the block reservation because we reserved blocks assuming that we were adding mappings into an already-cleared area of the data fork. Let's say we have 8 extents in the data fork, 8 extents in the CoW fork, and the data fork can hold at most 7 extents before needing to convert to btree format; and that blocks A-P are discontiguous single-block extents: 0......7 D: ABCDEFGH C: IJKLMNOP When a write to file blocks 0-7 completes, we must remap I-P into the data fork. We start by removing H from the btree-format data fork. Now we have 7 extents, so we convert the fork to extents format, freeing the bmbt block. We then move P into the data fork and it now has 8 extents again. We must convert the data fork back to btree format, requiring a block allocation. If we repeat this sequence for blocks 6-5-4-3-2-1-0, we'll need a total of 8 block allocations to remap all 8 blocks. We reserved only enough blocks to handle one btree split (5 blocks on a 4k block filesystem), which means we overflow the block reservation. To fix this issue, create a separate helper function to remap a single extent, and change _reflink_end_cow to call it in a tight loop over the entire range we're completing. As a side effect this also removes the size restrictions on how many extents we can end_cow at a time, though nobody ever hit that. It is not reasonable to reserve N blocks to remap N blocks. Note that this can be reproduced after ~320 million fsx ops while running generic/938 (long soak directio fsx exerciser): XFS: Assertion failed: tp->t_blk_res >= tp->t_blk_res_used, file: fs/xfs/xfs_trans.c, line: 116 <machine registers snipped> Call Trace: xfs_trans_dup+0x211/0x250 [xfs] xfs_trans_roll+0x6d/0x180 [xfs] xfs_defer_trans_roll+0x10c/0x3b0 [xfs] xfs_defer_finish_noroll+0xdf/0x740 [xfs] xfs_defer_finish+0x13/0x70 [xfs] xfs_reflink_end_cow+0x2c6/0x680 [xfs] xfs_dio_write_end_io+0x115/0x220 [xfs] iomap_dio_complete+0x3f/0x130 iomap_dio_rw+0x3c3/0x420 xfs_file_dio_aio_write+0x132/0x3c0 [xfs] xfs_file_write_iter+0x8b/0xc0 [xfs] __vfs_write+0x193/0x1f0 vfs_write+0xba/0x1c0 ksys_write+0x52/0xc0 do_syscall_64+0x50/0x160 entry_SYSCALL_64_after_hwframe+0x49/0xbe Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Brian Foster <bfoster@redhat.com>
2018-12-13 00:46:19 +08:00
return error;
/* Update the caller about how much progress we made. */
*offset_fsb = del.br_startoff + del.br_blockcount;
return 0;
out_cancel:
xfs_trans_cancel(tp);
xfs_iunlock(ip, XFS_ILOCK_EXCL);
xfs: split up the xfs_reflink_end_cow work into smaller transactions In xfs_reflink_end_cow, we allocate a single transaction for the entire end_cow operation and then loop the CoW fork mappings to move them to the data fork. This design fails on a heavily fragmented filesystem where an inode's data fork has exactly one more extent than would fit in an extents-format fork, because the unmap can collapse the data fork into extents format (freeing the bmbt block) but the remap can expand the data fork back into a (newly allocated) bmbt block. If the number of extents we end up remapping is large, we can overflow the block reservation because we reserved blocks assuming that we were adding mappings into an already-cleared area of the data fork. Let's say we have 8 extents in the data fork, 8 extents in the CoW fork, and the data fork can hold at most 7 extents before needing to convert to btree format; and that blocks A-P are discontiguous single-block extents: 0......7 D: ABCDEFGH C: IJKLMNOP When a write to file blocks 0-7 completes, we must remap I-P into the data fork. We start by removing H from the btree-format data fork. Now we have 7 extents, so we convert the fork to extents format, freeing the bmbt block. We then move P into the data fork and it now has 8 extents again. We must convert the data fork back to btree format, requiring a block allocation. If we repeat this sequence for blocks 6-5-4-3-2-1-0, we'll need a total of 8 block allocations to remap all 8 blocks. We reserved only enough blocks to handle one btree split (5 blocks on a 4k block filesystem), which means we overflow the block reservation. To fix this issue, create a separate helper function to remap a single extent, and change _reflink_end_cow to call it in a tight loop over the entire range we're completing. As a side effect this also removes the size restrictions on how many extents we can end_cow at a time, though nobody ever hit that. It is not reasonable to reserve N blocks to remap N blocks. Note that this can be reproduced after ~320 million fsx ops while running generic/938 (long soak directio fsx exerciser): XFS: Assertion failed: tp->t_blk_res >= tp->t_blk_res_used, file: fs/xfs/xfs_trans.c, line: 116 <machine registers snipped> Call Trace: xfs_trans_dup+0x211/0x250 [xfs] xfs_trans_roll+0x6d/0x180 [xfs] xfs_defer_trans_roll+0x10c/0x3b0 [xfs] xfs_defer_finish_noroll+0xdf/0x740 [xfs] xfs_defer_finish+0x13/0x70 [xfs] xfs_reflink_end_cow+0x2c6/0x680 [xfs] xfs_dio_write_end_io+0x115/0x220 [xfs] iomap_dio_complete+0x3f/0x130 iomap_dio_rw+0x3c3/0x420 xfs_file_dio_aio_write+0x132/0x3c0 [xfs] xfs_file_write_iter+0x8b/0xc0 [xfs] __vfs_write+0x193/0x1f0 vfs_write+0xba/0x1c0 ksys_write+0x52/0xc0 do_syscall_64+0x50/0x160 entry_SYSCALL_64_after_hwframe+0x49/0xbe Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Brian Foster <bfoster@redhat.com>
2018-12-13 00:46:19 +08:00
return error;
}
/*
* Remap parts of a file's data fork after a successful CoW.
*/
int
xfs_reflink_end_cow(
struct xfs_inode *ip,
xfs_off_t offset,
xfs_off_t count)
{
xfs_fileoff_t offset_fsb;
xfs_fileoff_t end_fsb;
int error = 0;
trace_xfs_reflink_end_cow(ip, offset, count);
offset_fsb = XFS_B_TO_FSBT(ip->i_mount, offset);
end_fsb = XFS_B_TO_FSB(ip->i_mount, offset + count);
/*
* Walk forwards until we've remapped the I/O range. The loop function
xfs: split up the xfs_reflink_end_cow work into smaller transactions In xfs_reflink_end_cow, we allocate a single transaction for the entire end_cow operation and then loop the CoW fork mappings to move them to the data fork. This design fails on a heavily fragmented filesystem where an inode's data fork has exactly one more extent than would fit in an extents-format fork, because the unmap can collapse the data fork into extents format (freeing the bmbt block) but the remap can expand the data fork back into a (newly allocated) bmbt block. If the number of extents we end up remapping is large, we can overflow the block reservation because we reserved blocks assuming that we were adding mappings into an already-cleared area of the data fork. Let's say we have 8 extents in the data fork, 8 extents in the CoW fork, and the data fork can hold at most 7 extents before needing to convert to btree format; and that blocks A-P are discontiguous single-block extents: 0......7 D: ABCDEFGH C: IJKLMNOP When a write to file blocks 0-7 completes, we must remap I-P into the data fork. We start by removing H from the btree-format data fork. Now we have 7 extents, so we convert the fork to extents format, freeing the bmbt block. We then move P into the data fork and it now has 8 extents again. We must convert the data fork back to btree format, requiring a block allocation. If we repeat this sequence for blocks 6-5-4-3-2-1-0, we'll need a total of 8 block allocations to remap all 8 blocks. We reserved only enough blocks to handle one btree split (5 blocks on a 4k block filesystem), which means we overflow the block reservation. To fix this issue, create a separate helper function to remap a single extent, and change _reflink_end_cow to call it in a tight loop over the entire range we're completing. As a side effect this also removes the size restrictions on how many extents we can end_cow at a time, though nobody ever hit that. It is not reasonable to reserve N blocks to remap N blocks. Note that this can be reproduced after ~320 million fsx ops while running generic/938 (long soak directio fsx exerciser): XFS: Assertion failed: tp->t_blk_res >= tp->t_blk_res_used, file: fs/xfs/xfs_trans.c, line: 116 <machine registers snipped> Call Trace: xfs_trans_dup+0x211/0x250 [xfs] xfs_trans_roll+0x6d/0x180 [xfs] xfs_defer_trans_roll+0x10c/0x3b0 [xfs] xfs_defer_finish_noroll+0xdf/0x740 [xfs] xfs_defer_finish+0x13/0x70 [xfs] xfs_reflink_end_cow+0x2c6/0x680 [xfs] xfs_dio_write_end_io+0x115/0x220 [xfs] iomap_dio_complete+0x3f/0x130 iomap_dio_rw+0x3c3/0x420 xfs_file_dio_aio_write+0x132/0x3c0 [xfs] xfs_file_write_iter+0x8b/0xc0 [xfs] __vfs_write+0x193/0x1f0 vfs_write+0xba/0x1c0 ksys_write+0x52/0xc0 do_syscall_64+0x50/0x160 entry_SYSCALL_64_after_hwframe+0x49/0xbe Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Brian Foster <bfoster@redhat.com>
2018-12-13 00:46:19 +08:00
* repeatedly cycles the ILOCK to allocate one transaction per remapped
* extent.
*
* If we're being called by writeback then the pages will still
xfs: split up the xfs_reflink_end_cow work into smaller transactions In xfs_reflink_end_cow, we allocate a single transaction for the entire end_cow operation and then loop the CoW fork mappings to move them to the data fork. This design fails on a heavily fragmented filesystem where an inode's data fork has exactly one more extent than would fit in an extents-format fork, because the unmap can collapse the data fork into extents format (freeing the bmbt block) but the remap can expand the data fork back into a (newly allocated) bmbt block. If the number of extents we end up remapping is large, we can overflow the block reservation because we reserved blocks assuming that we were adding mappings into an already-cleared area of the data fork. Let's say we have 8 extents in the data fork, 8 extents in the CoW fork, and the data fork can hold at most 7 extents before needing to convert to btree format; and that blocks A-P are discontiguous single-block extents: 0......7 D: ABCDEFGH C: IJKLMNOP When a write to file blocks 0-7 completes, we must remap I-P into the data fork. We start by removing H from the btree-format data fork. Now we have 7 extents, so we convert the fork to extents format, freeing the bmbt block. We then move P into the data fork and it now has 8 extents again. We must convert the data fork back to btree format, requiring a block allocation. If we repeat this sequence for blocks 6-5-4-3-2-1-0, we'll need a total of 8 block allocations to remap all 8 blocks. We reserved only enough blocks to handle one btree split (5 blocks on a 4k block filesystem), which means we overflow the block reservation. To fix this issue, create a separate helper function to remap a single extent, and change _reflink_end_cow to call it in a tight loop over the entire range we're completing. As a side effect this also removes the size restrictions on how many extents we can end_cow at a time, though nobody ever hit that. It is not reasonable to reserve N blocks to remap N blocks. Note that this can be reproduced after ~320 million fsx ops while running generic/938 (long soak directio fsx exerciser): XFS: Assertion failed: tp->t_blk_res >= tp->t_blk_res_used, file: fs/xfs/xfs_trans.c, line: 116 <machine registers snipped> Call Trace: xfs_trans_dup+0x211/0x250 [xfs] xfs_trans_roll+0x6d/0x180 [xfs] xfs_defer_trans_roll+0x10c/0x3b0 [xfs] xfs_defer_finish_noroll+0xdf/0x740 [xfs] xfs_defer_finish+0x13/0x70 [xfs] xfs_reflink_end_cow+0x2c6/0x680 [xfs] xfs_dio_write_end_io+0x115/0x220 [xfs] iomap_dio_complete+0x3f/0x130 iomap_dio_rw+0x3c3/0x420 xfs_file_dio_aio_write+0x132/0x3c0 [xfs] xfs_file_write_iter+0x8b/0xc0 [xfs] __vfs_write+0x193/0x1f0 vfs_write+0xba/0x1c0 ksys_write+0x52/0xc0 do_syscall_64+0x50/0x160 entry_SYSCALL_64_after_hwframe+0x49/0xbe Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Brian Foster <bfoster@redhat.com>
2018-12-13 00:46:19 +08:00
* have PageWriteback set, which prevents races with reflink remapping
* and truncate. Reflink remapping prevents races with writeback by
* taking the iolock and mmaplock before flushing the pages and
* remapping, which means there won't be any further writeback or page
* cache dirtying until the reflink completes.
*
* We should never have two threads issuing writeback for the same file
* region. There are also have post-eof checks in the writeback
* preparation code so that we don't bother writing out pages that are
* about to be truncated.
*
* If we're being called as part of directio write completion, the dio
* count is still elevated, which reflink and truncate will wait for.
* Reflink remapping takes the iolock and mmaplock and waits for
* pending dio to finish, which should prevent any directio until the
* remap completes. Multiple concurrent directio writes to the same
* region are handled by end_cow processing only occurring for the
* threads which succeed; the outcome of multiple overlapping direct
* writes is not well defined anyway.
*
* It's possible that a buffered write and a direct write could collide
* here (the buffered write stumbles in after the dio flushes and
* invalidates the page cache and immediately queues writeback), but we
* have never supported this 100%. If either disk write succeeds the
* blocks will be remapped.
*/
while (end_fsb > offset_fsb && !error)
error = xfs_reflink_end_cow_extent(ip, &offset_fsb, end_fsb);
xfs: split up the xfs_reflink_end_cow work into smaller transactions In xfs_reflink_end_cow, we allocate a single transaction for the entire end_cow operation and then loop the CoW fork mappings to move them to the data fork. This design fails on a heavily fragmented filesystem where an inode's data fork has exactly one more extent than would fit in an extents-format fork, because the unmap can collapse the data fork into extents format (freeing the bmbt block) but the remap can expand the data fork back into a (newly allocated) bmbt block. If the number of extents we end up remapping is large, we can overflow the block reservation because we reserved blocks assuming that we were adding mappings into an already-cleared area of the data fork. Let's say we have 8 extents in the data fork, 8 extents in the CoW fork, and the data fork can hold at most 7 extents before needing to convert to btree format; and that blocks A-P are discontiguous single-block extents: 0......7 D: ABCDEFGH C: IJKLMNOP When a write to file blocks 0-7 completes, we must remap I-P into the data fork. We start by removing H from the btree-format data fork. Now we have 7 extents, so we convert the fork to extents format, freeing the bmbt block. We then move P into the data fork and it now has 8 extents again. We must convert the data fork back to btree format, requiring a block allocation. If we repeat this sequence for blocks 6-5-4-3-2-1-0, we'll need a total of 8 block allocations to remap all 8 blocks. We reserved only enough blocks to handle one btree split (5 blocks on a 4k block filesystem), which means we overflow the block reservation. To fix this issue, create a separate helper function to remap a single extent, and change _reflink_end_cow to call it in a tight loop over the entire range we're completing. As a side effect this also removes the size restrictions on how many extents we can end_cow at a time, though nobody ever hit that. It is not reasonable to reserve N blocks to remap N blocks. Note that this can be reproduced after ~320 million fsx ops while running generic/938 (long soak directio fsx exerciser): XFS: Assertion failed: tp->t_blk_res >= tp->t_blk_res_used, file: fs/xfs/xfs_trans.c, line: 116 <machine registers snipped> Call Trace: xfs_trans_dup+0x211/0x250 [xfs] xfs_trans_roll+0x6d/0x180 [xfs] xfs_defer_trans_roll+0x10c/0x3b0 [xfs] xfs_defer_finish_noroll+0xdf/0x740 [xfs] xfs_defer_finish+0x13/0x70 [xfs] xfs_reflink_end_cow+0x2c6/0x680 [xfs] xfs_dio_write_end_io+0x115/0x220 [xfs] iomap_dio_complete+0x3f/0x130 iomap_dio_rw+0x3c3/0x420 xfs_file_dio_aio_write+0x132/0x3c0 [xfs] xfs_file_write_iter+0x8b/0xc0 [xfs] __vfs_write+0x193/0x1f0 vfs_write+0xba/0x1c0 ksys_write+0x52/0xc0 do_syscall_64+0x50/0x160 entry_SYSCALL_64_after_hwframe+0x49/0xbe Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com> Reviewed-by: Brian Foster <bfoster@redhat.com>
2018-12-13 00:46:19 +08:00
if (error)
trace_xfs_reflink_end_cow_error(ip, error, _RET_IP_);
return error;
}
/*
xfs: only run COW extent recovery when there are no live extents As part of multiple customer escalations due to file data corruption after copy on write operations, I wrote some fstests that use fsstress to hammer on COW to shake things loose. Regrettably, I caught some filesystem shutdowns due to incorrect rmap operations with the following loop: mount <filesystem> # (0) fsstress <run only readonly ops> & # (1) while true; do fsstress <run all ops> mount -o remount,ro # (2) fsstress <run only readonly ops> mount -o remount,rw # (3) done When (2) happens, notice that (1) is still running. xfs_remount_ro will call xfs_blockgc_stop to walk the inode cache to free all the COW extents, but the blockgc mechanism races with (1)'s reader threads to take IOLOCKs and loses, which means that it doesn't clean them all out. Call such a file (A). When (3) happens, xfs_remount_rw calls xfs_reflink_recover_cow, which walks the ondisk refcount btree and frees any COW extent that it finds. This function does not check the inode cache, which means that incore COW forks of inode (A) is now inconsistent with the ondisk metadata. If one of those former COW extents are allocated and mapped into another file (B) and someone triggers a COW to the stale reservation in (A), A's dirty data will be written into (B) and once that's done, those blocks will be transferred to (A)'s data fork without bumping the refcount. The results are catastrophic -- file (B) and the refcount btree are now corrupt. In the first patch, we fixed the race condition in (2) so that (A) will always flush the COW fork. In this second patch, we move the _recover_cow call to the initial mount call in (0) for safety. As mentioned previously, xfs_reflink_recover_cow walks the refcount btree looking for COW staging extents, and frees them. This was intended to be run at mount time (when we know there are no live inodes) to clean up any leftover staging events that may have been left behind during an unclean shutdown. As a time "optimization" for readonly mounts, we deferred this to the ro->rw transition, not realizing that any failure to clean all COW forks during a rw->ro transition would result in catastrophic corruption. Therefore, remove this optimization and only run the recovery routine when we're guaranteed not to have any COW staging extents anywhere, which means we always run this at mount time. While we're at it, move the callsite to xfs_log_mount_finish because any refcount btree expansion (however unlikely given that we're removing records from the right side of the index) must be fed by a per-AG reservation, which doesn't exist in its current location. Fixes: 174edb0e46e5 ("xfs: store in-progress CoW allocations in the refcount btree") Signed-off-by: Darrick J. Wong <djwong@kernel.org> Reviewed-by: Chandan Babu R <chandan.babu@oracle.com> Reviewed-by: Dave Chinner <dchinner@redhat.com>
2021-12-16 03:52:23 +08:00
* Free all CoW staging blocks that are still referenced by the ondisk refcount
* metadata. The ondisk metadata does not track which inode created the
* staging extent, so callers must ensure that there are no cached inodes with
* live CoW staging extents.
*/
int
xfs_reflink_recover_cow(
struct xfs_mount *mp)
{
struct xfs_perag *pag;
xfs_agnumber_t agno;
int error = 0;
if (!xfs_has_reflink(mp))
return 0;
for_each_perag(mp, agno, pag) {
error = xfs_refcount_recover_cow_leftovers(mp, pag);
if (error) {
xfs: active perag reference counting We need to be able to dynamically remove instantiated AGs from memory safely, either for shrinking the filesystem or paging AG state in and out of memory (e.g. supporting millions of AGs). This means we need to be able to safely exclude operations from accessing perags while dynamic removal is in progress. To do this, introduce the concept of active and passive references. Active references are required for high level operations that make use of an AG for a given operation (e.g. allocation) and pin the perag in memory for the duration of the operation that is operating on the perag (e.g. transaction scope). This means we can fail to get an active reference to an AG, hence callers of the new active reference API must be able to handle lookup failure gracefully. Passive references are used in low level code, where we might need to access the perag structure for the purposes of completing high level operations. For example, buffers need to use passive references because: - we need to be able to do metadata IO during operations like grow and shrink transactions where high level active references to the AG have already been blocked - buffers need to pin the perag until they are reclaimed from memory, something that high level code has no direct control over. - unused cached buffers should not prevent a shrink from being started. Hence we have active references that will form exclusion barriers for operations to be performed on an AG, and passive references that will prevent reclaim of the perag until all objects with passive references have been reclaimed themselves. This patch introduce xfs_perag_grab()/xfs_perag_rele() as the API for active AG reference functionality. We also need to convert the for_each_perag*() iterators to use active references, which will start the process of converting high level code over to using active references. Conversion of non-iterator based code to active references will be done in followup patches. Note that the implementation using reference counting is really just a development vehicle for the API to ensure we don't have any leaks in the callers. Once we need to remove perag structures from memory dyanmically, we will need a much more robust per-ag state transition mechanism for preventing new references from being taken while we wait for existing references to drain before removal from memory can occur.... Signed-off-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Allison Henderson <allison.henderson@oracle.com> Reviewed-by: Darrick J. Wong <djwong@kernel.org>
2023-02-13 06:14:42 +08:00
xfs_perag_rele(pag);
break;
}
}
return error;
}
/*
* Reflinking (Block) Ranges of Two Files Together
*
* First, ensure that the reflink flag is set on both inodes. The flag is an
* optimization to avoid unnecessary refcount btree lookups in the write path.
*
* Now we can iteratively remap the range of extents (and holes) in src to the
* corresponding ranges in dest. Let drange and srange denote the ranges of
* logical blocks in dest and src touched by the reflink operation.
*
* While the length of drange is greater than zero,
* - Read src's bmbt at the start of srange ("imap")
* - If imap doesn't exist, make imap appear to start at the end of srange
* with zero length.
* - If imap starts before srange, advance imap to start at srange.
* - If imap goes beyond srange, truncate imap to end at the end of srange.
* - Punch (imap start - srange start + imap len) blocks from dest at
* offset (drange start).
* - If imap points to a real range of pblks,
* > Increase the refcount of the imap's pblks
* > Map imap's pblks into dest at the offset
* (drange start + imap start - srange start)
* - Advance drange and srange by (imap start - srange start + imap len)
*
* Finally, if the reflink made dest longer, update both the in-core and
* on-disk file sizes.
*
* ASCII Art Demonstration:
*
* Let's say we want to reflink this source file:
*
* ----SSSSSSS-SSSSS----SSSSSS (src file)
* <-------------------->
*
* into this destination file:
*
* --DDDDDDDDDDDDDDDDDDD--DDD (dest file)
* <-------------------->
* '-' means a hole, and 'S' and 'D' are written blocks in the src and dest.
* Observe that the range has different logical offsets in either file.
*
* Consider that the first extent in the source file doesn't line up with our
* reflink range. Unmapping and remapping are separate operations, so we can
* unmap more blocks from the destination file than we remap.
*
* ----SSSSSSS-SSSSS----SSSSSS
* <------->
* --DDDDD---------DDDDD--DDD
* <------->
*
* Now remap the source extent into the destination file:
*
* ----SSSSSSS-SSSSS----SSSSSS
* <------->
* --DDDDD--SSSSSSSDDDDD--DDD
* <------->
*
* Do likewise with the second hole and extent in our range. Holes in the
* unmap range don't affect our operation.
*
* ----SSSSSSS-SSSSS----SSSSSS
* <---->
* --DDDDD--SSSSSSS-SSSSS-DDD
* <---->
*
* Finally, unmap and remap part of the third extent. This will increase the
* size of the destination file.
*
* ----SSSSSSS-SSSSS----SSSSSS
* <----->
* --DDDDD--SSSSSSS-SSSSS----SSS
* <----->
*
* Once we update the destination file's i_size, we're done.
*/
/*
* Ensure the reflink bit is set in both inodes.
*/
STATIC int
xfs_reflink_set_inode_flag(
struct xfs_inode *src,
struct xfs_inode *dest)
{
struct xfs_mount *mp = src->i_mount;
int error;
struct xfs_trans *tp;
if (xfs_is_reflink_inode(src) && xfs_is_reflink_inode(dest))
return 0;
error = xfs_trans_alloc(mp, &M_RES(mp)->tr_ichange, 0, 0, 0, &tp);
if (error)
goto out_error;
/* Lock both files against IO */
if (src->i_ino == dest->i_ino)
xfs_ilock(src, XFS_ILOCK_EXCL);
else
xfs_lock_two_inodes(src, XFS_ILOCK_EXCL, dest, XFS_ILOCK_EXCL);
if (!xfs_is_reflink_inode(src)) {
trace_xfs_reflink_set_inode_flag(src);
xfs_trans_ijoin(tp, src, XFS_ILOCK_EXCL);
src->i_diflags2 |= XFS_DIFLAG2_REFLINK;
xfs_trans_log_inode(tp, src, XFS_ILOG_CORE);
xfs_ifork_init_cow(src);
} else
xfs_iunlock(src, XFS_ILOCK_EXCL);
if (src->i_ino == dest->i_ino)
goto commit_flags;
if (!xfs_is_reflink_inode(dest)) {
trace_xfs_reflink_set_inode_flag(dest);
xfs_trans_ijoin(tp, dest, XFS_ILOCK_EXCL);
dest->i_diflags2 |= XFS_DIFLAG2_REFLINK;
xfs_trans_log_inode(tp, dest, XFS_ILOG_CORE);
xfs_ifork_init_cow(dest);
} else
xfs_iunlock(dest, XFS_ILOCK_EXCL);
commit_flags:
error = xfs_trans_commit(tp);
if (error)
goto out_error;
return error;
out_error:
trace_xfs_reflink_set_inode_flag_error(dest, error, _RET_IP_);
return error;
}
/*
* Update destination inode size & cowextsize hint, if necessary.
*/
int
xfs_reflink_update_dest(
struct xfs_inode *dest,
xfs_off_t newlen,
xfs_extlen_t cowextsize,
unsigned int remap_flags)
{
struct xfs_mount *mp = dest->i_mount;
struct xfs_trans *tp;
int error;
if (newlen <= i_size_read(VFS_I(dest)) && cowextsize == 0)
return 0;
error = xfs_trans_alloc(mp, &M_RES(mp)->tr_ichange, 0, 0, 0, &tp);
if (error)
goto out_error;
xfs_ilock(dest, XFS_ILOCK_EXCL);
xfs_trans_ijoin(tp, dest, XFS_ILOCK_EXCL);
if (newlen > i_size_read(VFS_I(dest))) {
trace_xfs_reflink_update_inode_size(dest, newlen);
i_size_write(VFS_I(dest), newlen);
dest->i_disk_size = newlen;
}
if (cowextsize) {
dest->i_cowextsize = cowextsize;
dest->i_diflags2 |= XFS_DIFLAG2_COWEXTSIZE;
}
xfs_trans_log_inode(tp, dest, XFS_ILOG_CORE);
error = xfs_trans_commit(tp);
if (error)
goto out_error;
return error;
out_error:
trace_xfs_reflink_update_inode_size_error(dest, error, _RET_IP_);
return error;
}
/*
* Do we have enough reserve in this AG to handle a reflink? The refcount
* btree already reserved all the space it needs, but the rmap btree can grow
* infinitely, so we won't allow more reflinks when the AG is down to the
* btree reserves.
*/
static int
xfs_reflink_ag_has_free_space(
struct xfs_mount *mp,
xfs_agnumber_t agno)
{
struct xfs_perag *pag;
int error = 0;
if (!xfs_has_rmapbt(mp))
return 0;
pag = xfs_perag_get(mp, agno);
if (xfs_ag_resv_critical(pag, XFS_AG_RESV_RMAPBT) ||
xfs_ag_resv_critical(pag, XFS_AG_RESV_METADATA))
error = -ENOSPC;
xfs_perag_put(pag);
return error;
}
/*
* Remap the given extent into the file. The dmap blockcount will be set to
* the number of blocks that were actually remapped.
*/
STATIC int
xfs_reflink_remap_extent(
struct xfs_inode *ip,
struct xfs_bmbt_irec *dmap,
xfs_off_t new_isize)
{
struct xfs_bmbt_irec smap;
struct xfs_mount *mp = ip->i_mount;
struct xfs_trans *tp;
xfs_off_t newlen;
int64_t qdelta = 0;
unsigned int resblks;
xfs: try worst case space reservation upfront in xfs_reflink_remap_extent Now that we've converted xfs_reflink_remap_extent to use the new xfs_trans_alloc_inode API, we can focus on its slightly unusual behavior with regard to quota reservations. Since it's valid to remap written blocks into a hole, we must be able to increase the quota count by the number of blocks in the mapping. However, the incore space reservation process requires us to supply an asymptotic guess before we can gain exclusive access to resources. We'd like to reserve all the quota we need up front, but we also don't want to fail a written -> allocated remap operation unnecessarily. The solution is to make the remap_extents function call the transaction allocation function twice. The first time we ask to reserve enough space and quota to handle the absolute worst case situation, but if that fails, we can fall back to the old strategy: ask for the bare minimum space reservation upfront and increase the quota reservation later if we need to. Later in this patchset we change the transaction and quota code to try to reclaim space if we cannot reserve free space or quota. Restructuring the remap_extent function in this manner means that if the fallback increase fails, we can pass that back to the caller knowing that the transaction allocation already tried freeing space. Signed-off-by: Darrick J. Wong <djwong@kernel.org> Reviewed-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de>
2021-01-28 02:40:00 +08:00
bool quota_reserved = true;
bool smap_real;
bool dmap_written = xfs_bmap_is_written_extent(dmap);
int iext_delta = 0;
int nimaps;
int error;
/*
* Start a rolling transaction to switch the mappings.
*
* Adding a written extent to the extent map can cause a bmbt split,
* and removing a mapped extent from the extent can cause a bmbt split.
* The two operations cannot both cause a split since they operate on
* the same index in the bmap btree, so we only need a reservation for
* one bmbt split if either thing is happening. However, we haven't
* locked the inode yet, so we reserve assuming this is the case.
xfs: try worst case space reservation upfront in xfs_reflink_remap_extent Now that we've converted xfs_reflink_remap_extent to use the new xfs_trans_alloc_inode API, we can focus on its slightly unusual behavior with regard to quota reservations. Since it's valid to remap written blocks into a hole, we must be able to increase the quota count by the number of blocks in the mapping. However, the incore space reservation process requires us to supply an asymptotic guess before we can gain exclusive access to resources. We'd like to reserve all the quota we need up front, but we also don't want to fail a written -> allocated remap operation unnecessarily. The solution is to make the remap_extents function call the transaction allocation function twice. The first time we ask to reserve enough space and quota to handle the absolute worst case situation, but if that fails, we can fall back to the old strategy: ask for the bare minimum space reservation upfront and increase the quota reservation later if we need to. Later in this patchset we change the transaction and quota code to try to reclaim space if we cannot reserve free space or quota. Restructuring the remap_extent function in this manner means that if the fallback increase fails, we can pass that back to the caller knowing that the transaction allocation already tried freeing space. Signed-off-by: Darrick J. Wong <djwong@kernel.org> Reviewed-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de>
2021-01-28 02:40:00 +08:00
*
* The first allocation call tries to reserve enough space to handle
* mapping dmap into a sparse part of the file plus the bmbt split. We
* haven't locked the inode or read the existing mapping yet, so we do
* not know for sure that we need the space. This should succeed most
* of the time.
*
* If the first attempt fails, try again but reserving only enough
* space to handle a bmbt split. This is the hard minimum requirement,
* and we revisit quota reservations later when we know more about what
* we're remapping.
*/
resblks = XFS_EXTENTADD_SPACE_RES(mp, XFS_DATA_FORK);
xfs: try worst case space reservation upfront in xfs_reflink_remap_extent Now that we've converted xfs_reflink_remap_extent to use the new xfs_trans_alloc_inode API, we can focus on its slightly unusual behavior with regard to quota reservations. Since it's valid to remap written blocks into a hole, we must be able to increase the quota count by the number of blocks in the mapping. However, the incore space reservation process requires us to supply an asymptotic guess before we can gain exclusive access to resources. We'd like to reserve all the quota we need up front, but we also don't want to fail a written -> allocated remap operation unnecessarily. The solution is to make the remap_extents function call the transaction allocation function twice. The first time we ask to reserve enough space and quota to handle the absolute worst case situation, but if that fails, we can fall back to the old strategy: ask for the bare minimum space reservation upfront and increase the quota reservation later if we need to. Later in this patchset we change the transaction and quota code to try to reclaim space if we cannot reserve free space or quota. Restructuring the remap_extent function in this manner means that if the fallback increase fails, we can pass that back to the caller knowing that the transaction allocation already tried freeing space. Signed-off-by: Darrick J. Wong <djwong@kernel.org> Reviewed-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de>
2021-01-28 02:40:00 +08:00
error = xfs_trans_alloc_inode(ip, &M_RES(mp)->tr_write,
resblks + dmap->br_blockcount, 0, false, &tp);
if (error == -EDQUOT || error == -ENOSPC) {
quota_reserved = false;
error = xfs_trans_alloc_inode(ip, &M_RES(mp)->tr_write,
resblks, 0, false, &tp);
}
if (error)
goto out;
/*
* Read what's currently mapped in the destination file into smap.
* If smap isn't a hole, we will have to remove it before we can add
* dmap to the destination file.
*/
nimaps = 1;
error = xfs_bmapi_read(ip, dmap->br_startoff, dmap->br_blockcount,
&smap, &nimaps, 0);
if (error)
goto out_cancel;
ASSERT(nimaps == 1 && smap.br_startoff == dmap->br_startoff);
smap_real = xfs_bmap_is_real_extent(&smap);
/*
* We can only remap as many blocks as the smaller of the two extent
* maps, because we can only remap one extent at a time.
*/
dmap->br_blockcount = min(dmap->br_blockcount, smap.br_blockcount);
ASSERT(dmap->br_blockcount == smap.br_blockcount);
trace_xfs_reflink_remap_extent_dest(ip, &smap);
/*
* Two extents mapped to the same physical block must not have
* different states; that's filesystem corruption. Move on to the next
* extent if they're both holes or both the same physical extent.
*/
if (dmap->br_startblock == smap.br_startblock) {
if (dmap->br_state != smap.br_state)
error = -EFSCORRUPTED;
goto out_cancel;
}
/* If both extents are unwritten, leave them alone. */
if (dmap->br_state == XFS_EXT_UNWRITTEN &&
smap.br_state == XFS_EXT_UNWRITTEN)
goto out_cancel;
/* No reflinking if the AG of the dest mapping is low on space. */
if (dmap_written) {
error = xfs_reflink_ag_has_free_space(mp,
XFS_FSB_TO_AGNO(mp, dmap->br_startblock));
if (error)
goto out_cancel;
}
/*
* Increase quota reservation if we think the quota block counter for
* this file could increase.
*
* If we are mapping a written extent into the file, we need to have
* enough quota block count reservation to handle the blocks in that
* extent. We log only the delta to the quota block counts, so if the
* extent we're unmapping also has blocks allocated to it, we don't
* need a quota reservation for the extent itself.
*
* Note that if we're replacing a delalloc reservation with a written
* extent, we have to take the full quota reservation because removing
* the delalloc reservation gives the block count back to the quota
* count. This is suboptimal, but the VFS flushed the dest range
* before we started. That should have removed all the delalloc
* reservations, but we code defensively.
*
* xfs_trans_alloc_inode above already tried to grab an even larger
* quota reservation, and kicked off a blockgc scan if it couldn't.
* If we can't get a potentially smaller quota reservation now, we're
* done.
*/
xfs: try worst case space reservation upfront in xfs_reflink_remap_extent Now that we've converted xfs_reflink_remap_extent to use the new xfs_trans_alloc_inode API, we can focus on its slightly unusual behavior with regard to quota reservations. Since it's valid to remap written blocks into a hole, we must be able to increase the quota count by the number of blocks in the mapping. However, the incore space reservation process requires us to supply an asymptotic guess before we can gain exclusive access to resources. We'd like to reserve all the quota we need up front, but we also don't want to fail a written -> allocated remap operation unnecessarily. The solution is to make the remap_extents function call the transaction allocation function twice. The first time we ask to reserve enough space and quota to handle the absolute worst case situation, but if that fails, we can fall back to the old strategy: ask for the bare minimum space reservation upfront and increase the quota reservation later if we need to. Later in this patchset we change the transaction and quota code to try to reclaim space if we cannot reserve free space or quota. Restructuring the remap_extent function in this manner means that if the fallback increase fails, we can pass that back to the caller knowing that the transaction allocation already tried freeing space. Signed-off-by: Darrick J. Wong <djwong@kernel.org> Reviewed-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de>
2021-01-28 02:40:00 +08:00
if (!quota_reserved && !smap_real && dmap_written) {
error = xfs_trans_reserve_quota_nblks(tp, ip,
dmap->br_blockcount, 0, false);
if (error)
goto out_cancel;
}
if (smap_real)
++iext_delta;
if (dmap_written)
++iext_delta;
error = xfs_iext_count_may_overflow(ip, XFS_DATA_FORK, iext_delta);
if (error == -EFBIG)
error = xfs_iext_count_upgrade(tp, ip, iext_delta);
if (error)
goto out_cancel;
if (smap_real) {
/*
* If the extent we're unmapping is backed by storage (written
* or not), unmap the extent and drop its refcount.
*/
xfs_bmap_unmap_extent(tp, ip, &smap);
xfs_refcount_decrease_extent(tp, &smap);
qdelta -= smap.br_blockcount;
} else if (smap.br_startblock == DELAYSTARTBLOCK) {
int done;
/*
* If the extent we're unmapping is a delalloc reservation,
* we can use the regular bunmapi function to release the
* incore state. Dropping the delalloc reservation takes care
* of the quota reservation for us.
*/
error = xfs_bunmapi(NULL, ip, smap.br_startoff,
smap.br_blockcount, 0, 1, &done);
if (error)
goto out_cancel;
ASSERT(done);
}
/*
* If the extent we're sharing is backed by written storage, increase
* its refcount and map it into the file.
*/
if (dmap_written) {
xfs_refcount_increase_extent(tp, dmap);
xfs_bmap_map_extent(tp, ip, dmap);
qdelta += dmap->br_blockcount;
}
xfs_trans_mod_dquot_byino(tp, ip, XFS_TRANS_DQ_BCOUNT, qdelta);
/* Update dest isize if needed. */
newlen = XFS_FSB_TO_B(mp, dmap->br_startoff + dmap->br_blockcount);
newlen = min_t(xfs_off_t, newlen, new_isize);
if (newlen > i_size_read(VFS_I(ip))) {
trace_xfs_reflink_update_inode_size(ip, newlen);
i_size_write(VFS_I(ip), newlen);
ip->i_disk_size = newlen;
xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
}
/* Commit everything and unlock. */
error = xfs_trans_commit(tp);
goto out_unlock;
out_cancel:
xfs_trans_cancel(tp);
out_unlock:
xfs_iunlock(ip, XFS_ILOCK_EXCL);
out:
if (error)
trace_xfs_reflink_remap_extent_error(ip, error, _RET_IP_);
return error;
}
/* Remap a range of one file to the other. */
int
xfs_reflink_remap_blocks(
struct xfs_inode *src,
loff_t pos_in,
struct xfs_inode *dest,
loff_t pos_out,
loff_t remap_len,
loff_t *remapped)
{
struct xfs_bmbt_irec imap;
struct xfs_mount *mp = src->i_mount;
xfs_fileoff_t srcoff = XFS_B_TO_FSBT(mp, pos_in);
xfs_fileoff_t destoff = XFS_B_TO_FSBT(mp, pos_out);
xfs_filblks_t len;
xfs_filblks_t remapped_len = 0;
xfs_off_t new_isize = pos_out + remap_len;
int nimaps;
int error = 0;
len = min_t(xfs_filblks_t, XFS_B_TO_FSB(mp, remap_len),
XFS_MAX_FILEOFF);
trace_xfs_reflink_remap_blocks(src, srcoff, len, dest, destoff);
while (len > 0) {
unsigned int lock_mode;
/* Read extent from the source file */
nimaps = 1;
lock_mode = xfs_ilock_data_map_shared(src);
error = xfs_bmapi_read(src, srcoff, len, &imap, &nimaps, 0);
xfs_iunlock(src, lock_mode);
if (error)
break;
/*
* The caller supposedly flushed all dirty pages in the source
* file range, which means that writeback should have allocated
* or deleted all delalloc reservations in that range. If we
* find one, that's a good sign that something is seriously
* wrong here.
*/
ASSERT(nimaps == 1 && imap.br_startoff == srcoff);
if (imap.br_startblock == DELAYSTARTBLOCK) {
ASSERT(imap.br_startblock != DELAYSTARTBLOCK);
error = -EFSCORRUPTED;
break;
}
trace_xfs_reflink_remap_extent_src(src, &imap);
/* Remap into the destination file at the given offset. */
imap.br_startoff = destoff;
error = xfs_reflink_remap_extent(dest, &imap, new_isize);
if (error)
break;
if (fatal_signal_pending(current)) {
error = -EINTR;
break;
}
/* Advance drange/srange */
srcoff += imap.br_blockcount;
destoff += imap.br_blockcount;
len -= imap.br_blockcount;
remapped_len += imap.br_blockcount;
}
if (error)
trace_xfs_reflink_remap_blocks_error(dest, error, _RET_IP_);
*remapped = min_t(loff_t, remap_len,
XFS_FSB_TO_B(src->i_mount, remapped_len));
return error;
}
/*
* If we're reflinking to a point past the destination file's EOF, we must
* zero any speculative post-EOF preallocations that sit between the old EOF
* and the destination file offset.
*/
static int
xfs_reflink_zero_posteof(
struct xfs_inode *ip,
loff_t pos)
{
loff_t isize = i_size_read(VFS_I(ip));
if (pos <= isize)
return 0;
trace_xfs_zero_eof(ip, isize, pos - isize);
return xfs_zero_range(ip, isize, pos - isize, NULL);
}
/*
* Prepare two files for range cloning. Upon a successful return both inodes
* will have the iolock and mmaplock held, the page cache of the out file will
* be truncated, and any leases on the out file will have been broken. This
* function borrows heavily from xfs_file_aio_write_checks.
*
* The VFS allows partial EOF blocks to "match" for dedupe even though it hasn't
* checked that the bytes beyond EOF physically match. Hence we cannot use the
* EOF block in the source dedupe range because it's not a complete block match,
* hence can introduce a corruption into the file that has it's block replaced.
*
* In similar fashion, the VFS file cloning also allows partial EOF blocks to be
* "block aligned" for the purposes of cloning entire files. However, if the
* source file range includes the EOF block and it lands within the existing EOF
* of the destination file, then we can expose stale data from beyond the source
* file EOF in the destination file.
*
* XFS doesn't support partial block sharing, so in both cases we have check
* these cases ourselves. For dedupe, we can simply round the length to dedupe
* down to the previous whole block and ignore the partial EOF block. While this
* means we can't dedupe the last block of a file, this is an acceptible
* tradeoff for simplicity on implementation.
*
* For cloning, we want to share the partial EOF block if it is also the new EOF
* block of the destination file. If the partial EOF block lies inside the
* existing destination EOF, then we have to abort the clone to avoid exposing
* stale data in the destination file. Hence we reject these clone attempts with
* -EINVAL in this case.
*/
int
xfs_reflink_remap_prep(
struct file *file_in,
loff_t pos_in,
struct file *file_out,
loff_t pos_out,
loff_t *len,
unsigned int remap_flags)
{
struct inode *inode_in = file_inode(file_in);
struct xfs_inode *src = XFS_I(inode_in);
struct inode *inode_out = file_inode(file_out);
struct xfs_inode *dest = XFS_I(inode_out);
int ret;
/* Lock both files against IO */
ret = xfs_ilock2_io_mmap(src, dest);
if (ret)
return ret;
/* Check file eligibility and prepare for block sharing. */
ret = -EINVAL;
/* Don't reflink realtime inodes */
if (XFS_IS_REALTIME_INODE(src) || XFS_IS_REALTIME_INODE(dest))
goto out_unlock;
/* Don't share DAX file data with non-DAX file. */
if (IS_DAX(inode_in) != IS_DAX(inode_out))
goto out_unlock;
if (!IS_DAX(inode_in))
ret = generic_remap_file_range_prep(file_in, pos_in, file_out,
pos_out, len, remap_flags);
else
ret = dax_remap_file_range_prep(file_in, pos_in, file_out,
pos_out, len, remap_flags, &xfs_read_iomap_ops);
if (ret || *len == 0)
goto out_unlock;
/* Attach dquots to dest inode before changing block map */
ret = xfs_qm_dqattach(dest);
if (ret)
goto out_unlock;
/*
* Zero existing post-eof speculative preallocations in the destination
* file.
*/
ret = xfs_reflink_zero_posteof(dest, pos_out);
if (ret)
goto out_unlock;
/* Set flags and remap blocks. */
ret = xfs_reflink_set_inode_flag(src, dest);
if (ret)
goto out_unlock;
xfs: flush removing page cache in xfs_reflink_remap_prep On a sub-page block size filesystem, fsx is failing with a data corruption after a series of operations involving copying a file with the destination offset beyond EOF of the destination of the file: 8093(157 mod 256): TRUNCATE DOWN from 0x7a120 to 0x50000 ******WWWW 8094(158 mod 256): INSERT 0x25000 thru 0x25fff (0x1000 bytes) 8095(159 mod 256): COPY 0x18000 thru 0x1afff (0x3000 bytes) to 0x2f400 8096(160 mod 256): WRITE 0x5da00 thru 0x651ff (0x7800 bytes) HOLE 8097(161 mod 256): COPY 0x2000 thru 0x5fff (0x4000 bytes) to 0x6fc00 The second copy here is beyond EOF, and it is to sub-page (4k) but block aligned (1k) offset. The clone runs the EOF zeroing, landing in a pre-existing post-eof delalloc extent. This zeroes the post-eof extents in the page cache just fine, dirtying the pages correctly. The problem is that xfs_reflink_remap_prep() now truncates the page cache over the range that it is copying it to, and rounds that down to cover the entire start page. This removes the dirty page over the delalloc extent from the page cache without having written it back. Hence later, when the page cache is flushed, the page at offset 0x6f000 has not been written back and hence exposes stale data, which fsx trips over less than 10 operations later. Fix this by changing xfs_reflink_remap_prep() to use xfs_flush_unmap_range(). Signed-off-by: Dave Chinner <dchinner@redhat.com> Reviewed-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
2018-11-20 05:31:10 +08:00
/*
* If pos_out > EOF, we may have dirtied blocks between EOF and
* pos_out. In that case, we need to extend the flush and unmap to cover
* from EOF to the end of the copy length.
*/
if (pos_out > XFS_ISIZE(dest)) {
loff_t flen = *len + (pos_out - XFS_ISIZE(dest));
ret = xfs_flush_unmap_range(dest, XFS_ISIZE(dest), flen);
} else {
ret = xfs_flush_unmap_range(dest, pos_out, *len);
}
if (ret)
goto out_unlock;
2023-10-18 04:12:08 +08:00
xfs_iflags_set(src, XFS_IREMAPPING);
if (inode_in != inode_out)
xfs_ilock_demote(src, XFS_IOLOCK_EXCL | XFS_MMAPLOCK_EXCL);
return 0;
out_unlock:
xfs_iunlock2_io_mmap(src, dest);
return ret;
}
/* Does this inode need the reflink flag? */
int
xfs_reflink_inode_has_shared_extents(
struct xfs_trans *tp,
struct xfs_inode *ip,
bool *has_shared)
{
struct xfs_bmbt_irec got;
struct xfs_mount *mp = ip->i_mount;
struct xfs_ifork *ifp;
struct xfs_iext_cursor icur;
bool found;
int error;
ifp = xfs_ifork_ptr(ip, XFS_DATA_FORK);
error = xfs_iread_extents(tp, ip, XFS_DATA_FORK);
if (error)
return error;
*has_shared = false;
found = xfs_iext_lookup_extent(ip, ifp, 0, &icur, &got);
while (found) {
struct xfs_perag *pag;
xfs_agblock_t agbno;
xfs_extlen_t aglen;
xfs_agblock_t rbno;
xfs_extlen_t rlen;
if (isnullstartblock(got.br_startblock) ||
got.br_state != XFS_EXT_NORM)
goto next;
pag = xfs_perag_get(mp, XFS_FSB_TO_AGNO(mp, got.br_startblock));
agbno = XFS_FSB_TO_AGBNO(mp, got.br_startblock);
aglen = got.br_blockcount;
error = xfs_reflink_find_shared(pag, tp, agbno, aglen,
&rbno, &rlen, false);
xfs_perag_put(pag);
if (error)
return error;
/* Is there still a shared block here? */
if (rbno != NULLAGBLOCK) {
*has_shared = true;
return 0;
}
next:
found = xfs_iext_next_extent(ifp, &icur, &got);
}
return 0;
}
/*
* Clear the inode reflink flag if there are no shared extents.
*
* The caller is responsible for joining the inode to the transaction passed in.
* The inode will be joined to the transaction that is returned to the caller.
*/
int
xfs_reflink_clear_inode_flag(
struct xfs_inode *ip,
struct xfs_trans **tpp)
{
bool needs_flag;
int error = 0;
ASSERT(xfs_is_reflink_inode(ip));
error = xfs_reflink_inode_has_shared_extents(*tpp, ip, &needs_flag);
if (error || needs_flag)
return error;
/*
* We didn't find any shared blocks so turn off the reflink flag.
* First, get rid of any leftover CoW mappings.
*/
error = xfs_reflink_cancel_cow_blocks(ip, tpp, 0, XFS_MAX_FILEOFF,
true);
if (error)
return error;
/* Clear the inode flag. */
trace_xfs_reflink_unset_inode_flag(ip);
ip->i_diflags2 &= ~XFS_DIFLAG2_REFLINK;
xfs_inode_clear_cowblocks_tag(ip);
xfs_trans_log_inode(*tpp, ip, XFS_ILOG_CORE);
return error;
}
/*
* Clear the inode reflink flag if there are no shared extents and the size
* hasn't changed.
*/
STATIC int
xfs_reflink_try_clear_inode_flag(
struct xfs_inode *ip)
{
struct xfs_mount *mp = ip->i_mount;
struct xfs_trans *tp;
int error = 0;
/* Start a rolling transaction to remove the mappings */
error = xfs_trans_alloc(mp, &M_RES(mp)->tr_write, 0, 0, 0, &tp);
if (error)
return error;
xfs_ilock(ip, XFS_ILOCK_EXCL);
xfs_trans_ijoin(tp, ip, 0);
error = xfs_reflink_clear_inode_flag(ip, &tp);
if (error)
goto cancel;
error = xfs_trans_commit(tp);
if (error)
goto out;
xfs_iunlock(ip, XFS_ILOCK_EXCL);
return 0;
cancel:
xfs_trans_cancel(tp);
out:
xfs_iunlock(ip, XFS_ILOCK_EXCL);
return error;
}
/*
* Pre-COW all shared blocks within a given byte range of a file and turn off
* the reflink flag if we unshare all of the file's blocks.
*/
int
xfs_reflink_unshare(
struct xfs_inode *ip,
xfs_off_t offset,
xfs_off_t len)
{
struct inode *inode = VFS_I(ip);
int error;
if (!xfs_is_reflink_inode(ip))
return 0;
trace_xfs_reflink_unshare(ip, offset, len);
inode_dio_wait(inode);
if (IS_DAX(inode))
error = dax_file_unshare(inode, offset, len,
&xfs_dax_write_iomap_ops);
else
error = iomap_file_unshare(inode, offset, len,
&xfs_buffered_write_iomap_ops);
if (error)
goto out;
error = filemap_write_and_wait_range(inode->i_mapping, offset,
offset + len - 1);
if (error)
goto out;
/* Turn off the reflink flag if possible. */
error = xfs_reflink_try_clear_inode_flag(ip);
if (error)
goto out;
return 0;
out:
trace_xfs_reflink_unshare_error(ip, error, _RET_IP_);
return error;
}