| Age | Commit message (Collapse) | Author |
|---|
|
git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs
Pull iomap updates from Christian Brauner:
"FUSE iomap Support for Buffered Reads:
This adds iomap support for FUSE buffered reads and readahead. This
enables granular uptodate tracking with large folios so only
non-uptodate portions need to be read. Also fixes a race condition
with large folios + writeback cache that could cause data corruption
on partial writes followed by reads.
- Refactored iomap read/readahead bio logic into helpers
- Added caller-provided callbacks for read operations
- Moved buffered IO bio logic into new file
- FUSE now uses iomap for read_folio and readahead
Zero Range Folio Batch Support:
Add folio batch support for iomap_zero_range() to handle dirty
folios over unwritten mappings. Fix raciness issues where dirty data
could be lost during zero range operations.
- filemap_get_folios_tag_range() helper for dirty folio lookup
- Optional zero range dirty folio processing
- XFS fills dirty folios on zero range of unwritten mappings
- Removed old partial EOF zeroing optimization
DIO Write Completions from Interrupt Context:
Restore pre-iomap behavior where pure overwrite completions run
inline rather than being deferred to workqueue. Reduces context
switches for high-performance workloads like ScyllaDB.
- Removed unused IOCB_DIO_CALLER_COMP code
- Error completions always run in user context (fixes zonefs)
- Reworked REQ_FUA selection logic
- Inverted IOMAP_DIO_INLINE_COMP to IOMAP_DIO_OFFLOAD_COMP
Buffered IO Cleanups:
Some performance and code clarity improvements:
- Replace manual bitmap scanning with find_next_bit()
- Simplify read skip logic for writes
- Optimize pending async writeback accounting
- Better variable naming
- Documentation for iomap_finish_folio_write() requirements
Misaligned Vectors for Zoned XFS:
Enables sub-block aligned vectors in XFS always-COW mode for zoned
devices via new IOMAP_DIO_FSBLOCK_ALIGNED flag.
Bug Fixes:
- Allocate s_dio_done_wq for async reads (fixes syzbot report after
error completion changes)
- Fix iomap_read_end() for already uptodate folios (regression fix)"
* tag 'vfs-6.19-rc1.iomap' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs: (40 commits)
iomap: allocate s_dio_done_wq for async reads as well
iomap: fix iomap_read_end() for already uptodate folios
iomap: invert the polarity of IOMAP_DIO_INLINE_COMP
iomap: support write completions from interrupt context
iomap: rework REQ_FUA selection
iomap: always run error completions in user context
fs, iomap: remove IOCB_DIO_CALLER_COMP
iomap: use find_next_bit() for uptodate bitmap scanning
iomap: use find_next_bit() for dirty bitmap scanning
iomap: simplify when reads can be skipped for writes
iomap: simplify ->read_folio_range() error handling for reads
iomap: optimize pending async writeback accounting
docs: document iomap writeback's iomap_finish_folio_write() requirement
iomap: account for unaligned end offsets when truncating read range
iomap: rename bytes_pending/bytes_accounted to bytes_submitted/bytes_not_submitted
xfs: support sub-block aligned vectors in always COW mode
iomap: add IOMAP_DIO_FSBLOCK_ALIGNED flag
xfs: error tag to force zeroing on debug kernels
iomap: remove old partial eof zeroing optimization
xfs: fill dirty folios on zero range of unwritten mappings
...
|
|
I think there are several things wrong with this function:
A) xfs_bmapi_write can return a much larger unwritten mapping than what
the caller asked for. We convert part of that range to written, but
return the entire written mapping to iomap even though that's
inaccurate.
B) The arguments to xfs_reflink_convert_cow_locked are wrong -- an
unwritten mapping could be *smaller* than the write range (or even
the hole range). In this case, we convert too much file range to
written state because we then return a smaller mapping to iomap.
C) It doesn't handle delalloc mappings. This I covered in the patch
that I already sent to the list.
D) Reassigning count_fsb to handle the hole means that if the second
cmap lookup attempt succeeds (due to racing with someone else) we
trim the mapping more than is strictly necessary. The changing
meaning of count_fsb makes this harder to notice.
E) The tracepoint is kinda wrong because @length is mutated. That makes
it harder to chase the data flows through this function because you
can't just grep on the pos/bytecount strings.
F) We don't actually check that the br_state = XFS_EXT_NORM assignment
is accurate, i.e that the cow fork actually contains a written
mapping for the range we're interested in
G) Somewhat inadequate documentation of why we need to xfs_trim_extent
so aggressively in this function.
H) Not sure why xfs_iomap_end_fsb is used here, the vfs already clamped
the write range to s_maxbytes.
Fix these issues, and then the atomic writes regressions in generic/760,
generic/617, generic/091, generic/263, and generic/521 all go away for
me.
Cc: stable@vger.kernel.org # v6.16
Fixes: bd1d2c21d5d249 ("xfs: add xfs_atomic_write_cow_iomap_begin()")
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: John Garry <john.g.garry@oracle.com>
Signed-off-by: Carlos Maiolino <cem@kernel.org>
|
|
With the 20 Oct 2025 release of fstests, generic/521 fails for me on
regular (aka non-block-atomic-writes) storage:
QA output created by 521
dowrite: write: Input/output error
LOG DUMP (8553 total operations):
1( 1 mod 256): SKIPPED (no operation)
2( 2 mod 256): WRITE 0x7e000 thru 0x8dfff (0x10000 bytes) HOLE
3( 3 mod 256): READ 0x69000 thru 0x79fff (0x11000 bytes)
4( 4 mod 256): FALLOC 0x53c38 thru 0x5e853 (0xac1b bytes) INTERIOR
5( 5 mod 256): COPY 0x55000 thru 0x59fff (0x5000 bytes) to 0x25000 thru 0x29fff
6( 6 mod 256): WRITE 0x74000 thru 0x88fff (0x15000 bytes)
7( 7 mod 256): ZERO 0xedb1 thru 0x11693 (0x28e3 bytes)
with a warning in dmesg from iomap about XFS trying to give it a
delalloc mapping for a directio write. Fix the software atomic write
iomap_begin code to convert the reservation into a written mapping.
This doesn't fix the data corruption problems reported by generic/760,
but it's a start.
Cc: stable@vger.kernel.org # v6.16
Fixes: bd1d2c21d5d249 ("xfs: add xfs_atomic_write_cow_iomap_begin()")
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: John Garry <john.g.garry@oracle.com>
Signed-off-by: Carlos Maiolino <cem@kernel.org>
|
|
Use the iomap folio batch mechanism to select folios to zero on zero
range of unwritten mappings. Trim the resulting mapping if the batch
is filled (unlikely for current use cases) to distinguish between a
range to skip and one that requires another iteration due to a full
batch.
Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
Refactor and tweak the IOMAP_ZERO logic in preparation to support
filling the folio batch for unwritten mappings. Drop the superfluous
imap offset check since the hole case has already been filtered out.
Split the the delalloc case handling into a sub-branch, and always
trim the imap to the requested offset/count so it can be more easily
used to bound the range to lookup in pagecache.
Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
Jan Kara reported that the shared ILOCK held across the journal
flush during fdatasync operations slows down O_DSYNC DIO on
unwritten extents significantly. The underlying issue is that
unwritten extent conversion needs the ILOCK exclusive, whilst the
datasync operation after the extent conversion holds it shared.
Hence we cannot be flushing the journal for one IO completion whilst
at the same time doing unwritten extent conversion on another IO
completion on the same inode. This means that IO completions
lock-step, and IO performance is dependent on the journal flush
latency.
Jan demonstrated that reducing the ifdatasync lock hold time can
improve O_DSYNC DIO to unwritten extents performance by 2.5x.
Discussion on that patch found issues with the method, and we
came to the conclusion that separately tracking datasync flush
sequences was the best approach to solving the problem.
The fsync code uses the ILOCK to serialise against concurrent
modifications in the transaction commit phase. In a transaction
commit, there are several disjoint updates to inode log item state
that need to be considered atomically by the fsync code. These
operations are all done under ILOCK_EXCL context:
1. ili_fsync_flags is updated in ->iop_precommit
2. i_pincount is updated in ->iop_pin before it is added to the CIL
3. ili_commit_seq is updated in ->iop_committing, after it has been
added to the CIL
In fsync, we need to:
1. check that the inode is dirty in the journal (ipincount)
2. check that ili_fsync_flags is set
3. grab the ili_commit_seq if a journal flush is needed
4. clear the ili_fsync_flags to ensure that new modifications that
require fsync are tracked in ->iop_precommit correctly
The serialisation of ipincount/ili_commit_seq is needed
to ensure that we don't try to unnecessarily flush the journal.
The serialisation of ili_fsync_flags being set in
->iop_precommit and cleared in fsync post journal flush is
required for correctness.
Hence holding the ILOCK_SHARED in xfs_file_fsync() performs all this
serialisation for us. Ideally, we want to remove the need to hold
the ILOCK_SHARED in xfs_file_fsync() for best performance.
We start with the observation that fsync/fdatasync() only need to
wait for operations that have been completed. Hence operations that
are still being committed have not completed and datasync operations
do not need to wait for them.
This means we can use a single point in time in the commit process
to signal "this modification is complete". This is what
->iop_committing is supposed to provide - it is the point at which
the object is unlocked after the modification has been recorded in
the CIL. Hence we could use ili_commit_seq to determine if we should
flush the journal.
In theory, we can already do this. However, in practice this will
expose an internal global CIL lock to the IO path. The ipincount()
checks optimise away the need to take this lock - if the inode is
not pinned, then it is not in the CIL and we don't need to check if
a journal flush at ili_commit_seq needs to be performed.
The reason this is needed is that the ili_commit_seq is never
cleared. Once it is set, it remains set even once the journal has
been committed and the object has been unpinned. Hence we have to
look that journal internal commit sequence state to determine if
ili_commit_seq needs to be acted on or not.
We can solve this by clearing ili_commit_seq when the inode is
unpinned. If we clear it atomically with the last unpin going away,
then we are guaranteed that new modifications will order correctly
as they add a new pin counts and we won't clear a sequence number
for an active modification in the CIL.
Further, we can then allow the per-transaction flag state to
propagate into ->iop_committing (instead of clearing it in
->iop_precommit) and that will allow us to determine if the
modification needs a full fsync or just a datasync, and so we can
record a separate datasync sequence number (Jan's idea!) and then
use that in the fdatasync path instead of the full fsync sequence
number.
With this infrastructure in place, we no longer need the
ILOCK_SHARED in the fsync path. All serialisation is done against
the commit sequence numbers - if the sequence number is set, then we
have to flush the journal. If it is not set, then we have nothing to
do. This greatly simplifies the fsync implementation....
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Tested-by: Jan Kara <jack@suse.cz>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Carlos Maiolino <cem@kernel.org>
|
|
Don't pass expr to XFS_TEST_ERROR. Most calls pass a constant false,
and the places that do pass an expression become cleaner by moving it
out.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Carlos Maiolino <cem@kernel.org>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs
Pull vfs iomap updates from Christian Brauner:
- Refactor the iomap writeback code and split the generic and ioend/bio
based writeback code.
There are two methods that define the split between the generic
writeback code, and the implemementation of it, and all knowledge of
ioends and bios now sits below that layer.
- Add fuse iomap support for buffered writes and dirty folio writeback.
This is needed so that granular uptodate and dirty tracking can be
used in fuse when large folios are enabled. This has two big
advantages. For writes, instead of the entire folio needing to be
read into the page cache, only the relevant portions need to be. For
writeback, only the dirty portions need to be written back instead of
the entire folio.
* tag 'vfs-6.17-rc1.iomap' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs:
fuse: refactor writeback to use iomap_writepage_ctx inode
fuse: hook into iomap for invalidating and checking partial uptodateness
fuse: use iomap for folio laundering
fuse: use iomap for writeback
fuse: use iomap for buffered writes
iomap: build the writeback code without CONFIG_BLOCK
iomap: add read_folio_range() handler for buffered writes
iomap: improve argument passing to iomap_read_folio_sync
iomap: replace iomap_folio_ops with iomap_write_ops
iomap: export iomap_writeback_folio
iomap: move folio_unlock out of iomap_writeback_folio
iomap: rename iomap_writepage_map to iomap_writeback_folio
iomap: move all ioend handling to ioend.c
iomap: add public helpers for uptodate state manipulation
iomap: hide ioends from the generic writeback code
iomap: refactor the writeback interface
iomap: cleanup the pending writeback tracking in iomap_writepage_map_blocks
iomap: pass more arguments using the iomap writeback context
iomap: header diet
|
|
The iomap_folio_ops are only used for buffered writes, including the zero
and unshare variants. Rename them to iomap_write_ops to better describe
the usage, and pass them through the call chain like the other operation
specific methods instead of through the iomap.
xfs_iomap_valid grows a IOMAP_HOLE check to keep the existing behavior
that never attached the folio_ops to a iomap representing a hole.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/20250710133343.399917-12-hch@lst.de
Acked-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
The extra bdev_ is weird, so drop it. Also improve the comment to make
it clear these are the hardware limits.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: John Garry <john.g.garry@oracle.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Carlos Maiolino <cem@kernel.org>
|
|
For when large atomic writes (> 1x FS block) are supported, there will be
various occasions when HW offload may not be possible.
Such instances include:
- unaligned extent mapping wrt write length
- extent mappings which do not cover the full write, e.g. the write spans
sparse or mixed-mapping extents
- the write length is greater than HW offload can support
- no hardware support at all
In those cases, we need to fallback to the CoW-based atomic write mode. For
this, report special code -ENOPROTOOPT to inform the caller that HW
offload-based method is not possible.
In addition to the occasions mentioned, if the write covers an unallocated
range, we again judge that we need to rely on the CoW-based method when we
would need to allocate anything more than 1x block. This is because if we
allocate less blocks that is required for the write, then again HW
offload-based method would not be possible. So we are taking a pessimistic
approach to writes covering unallocated space.
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
[djwong: various cleanups]
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: John Garry <john.g.garry@oracle.com>
|
|
For CoW-based atomic writes, reuse the infrastructure for reflink CoW fork
support.
Add ->iomap_begin() callback xfs_atomic_write_cow_iomap_begin() to create
staging mappings in the CoW fork for atomic write updates.
The general steps in the function are as follows:
- find extent mapping in the CoW fork for the FS block range being written
- if part or full extent is found, proceed to process found extent
- if no extent found, map in new blocks to the CoW fork
- convert unwritten blocks in extent if required
- update iomap extent mapping and return
The bulk of this function is quite similar to the processing in
xfs_reflink_allocate_cow(), where we try to find an extent mapping; if
none exists, then allocate a new extent in the CoW fork, convert unwritten
blocks, and return a mapping.
Performance testing has shown the XFS_ILOCK_EXCL locking to be quite
a bottleneck, so this is an area which could be optimised in future.
Christoph Hellwig contributed almost all of the code in
xfs_atomic_write_cow_iomap_begin().
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
[djwong: add a new xfs_can_sw_atomic_write to convey intent better]
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: John Garry <john.g.garry@oracle.com>
|
|
Pull xfs updates from Carlos Maiolino:
- XFS zoned allocator: Enables XFS to support zoned devices using its
real-time allocator
- Use folios/vmalloc for buffer cache backing memory
- Some code cleanups and bug fixes
* tag 'xfs-6.15-merge' of git://git.kernel.org/pub/scm/fs/xfs/xfs-linux: (70 commits)
xfs: remove the flags argument to xfs_buf_get_uncached
xfs: remove the flags argument to xfs_buf_read_uncached
xfs: remove xfs_buf_free_maps
xfs: remove xfs_buf_get_maps
xfs: call xfs_buf_alloc_backing_mem from _xfs_buf_alloc
xfs: remove unnecessary NULL check before kvfree()
xfs: don't wake zone space waiters without m_zone_info
xfs: don't increment m_generation for all errors in xfs_growfs_data
xfs: fix a missing unlock in xfs_growfs_data
xfs: Remove duplicate xfs_rtbitmap.h header
xfs: trigger zone GC when out of available rt blocks
xfs: trace what memory backs a buffer
xfs: cleanup mapping tmpfs folios into the buffer cache
xfs: use vmalloc instead of vm_map_area for buffer backing memory
xfs: buffer items don't straddle pages anymore
xfs: kill XBF_UNMAPPED
xfs: convert buffer cache to use high order folios
xfs: remove the kmalloc to page allocator fallback
xfs: refactor backing memory allocations for buffers
xfs: remove xfs_buf_is_vmapped
...
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs
Pull vfs iomap updates from Christian Brauner:
- Allow the filesystem to submit the writeback bios.
- Allow the filsystem to track completions on a per-bio bases
instead of the entire I/O.
- Change writeback_ops so that ->submit_bio can be done by the
filesystem.
- A new ANON_WRITE flag for writes that don't have a block number
assigned to them at the iomap level leaving the filesystem to do
that work in the submission handler.
- Incremental iterator advance
The folio_batch support for zero range where the filesystem provides
a batch of folios to process that might not be logically continguous
requires more flexibility than the current offset based iteration
currently offers.
Update all iomap operations to advance the iterator within the
operation and thus remove the need to advance from the core iomap
iterator.
- Make buffered writes work with RWF_DONTCACHE
If RWF_DONTCACHE is set for a write, mark the folios being written as
uncached. On writeback completion the pages will be dropped.
- Introduce infrastructure for large atomic writes
This will eventually be used by xfs and ext4.
* tag 'vfs-6.15-rc1.iomap' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs: (42 commits)
iomap: rework IOMAP atomic flags
iomap: comment on atomic write checks in iomap_dio_bio_iter()
iomap: inline iomap_dio_bio_opflags()
iomap: fix inline data on buffered read
iomap: Lift blocksize restriction on atomic writes
iomap: Support SW-based atomic writes
iomap: Rename IOMAP_ATOMIC -> IOMAP_ATOMIC_HW
xfs: flag as supporting FOP_DONTCACHE
iomap: make buffered writes work with RWF_DONTCACHE
iomap: introduce a full map advance helper
iomap: rename iomap_iter processed field to status
iomap: remove unnecessary advance from iomap_iter()
dax: advance the iomap_iter on pte and pmd faults
dax: advance the iomap_iter on dedupe range
dax: advance the iomap_iter on unshare range
dax: advance the iomap_iter on zero range
dax: push advance down into dax_iomap_iter() for read and write
dax: advance the iomap_iter in the read/write path
iomap: convert misc simple ops to incremental advance
iomap: advance the iter on direct I/O
...
|
|
Flag IOMAP_ATOMIC_SW is not really required. The idea of having this flag
is that the FS ->iomap_begin callback could check if this flag is set to
decide whether to do a SW (FS-based) atomic write. But the FS can set
which ->iomap_begin callback it wants when deciding to do a FS-based
atomic write.
Furthermore, it was thought that IOMAP_ATOMIC_HW is not a proper name, as
the block driver can use SW-methods to emulate an atomic write. So change
back to IOMAP_ATOMIC.
The ->iomap_begin callback needs though to indicate to iomap core that
REQ_ATOMIC needs to be set, so add IOMAP_F_ATOMIC_BIO for that.
These changes were suggested by Christoph Hellwig and Dave Chinner.
Signed-off-by: John Garry <john.g.garry@oracle.com>
Link: https://lore.kernel.org/r/20250320120250.4087011-4-john.g.garry@oracle.com
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
Direct writes to zoned RT devices are extremely simple. After taking the
block reservation before acquiring the iolock, the iomap direct I/O calls
into ->iomap_begin which will return a "fake" iomap for the entire
requested range. The actual block allocation is then done from the
submit_io handler using code shared with the buffered I/O path.
The iomap_dio_ops set the bio_set to the (iomap) ioend one and initialize
the embedded ioend, which allows reusing the existing ioend based buffered
I/O completion path.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: "Darrick J. Wong" <djwong@kernel.org>
|
|
Implement buffered writes including page faults and block zeroing for
zoned RT devices. Buffered writes to zoned RT devices are split into
three phases:
1) a reservation for the worst case data block usage is taken before
acquiring the iolock. When there are enough free blocks but not
enough available one, garbage collection is kicked off to free the
space before continuing with the write. If there isn't enough
freeable space, the block reservation is reduced and a short write
will happen as expected by normal Linux write semantics.
2) with the iolock held, the generic iomap buffered write code is
called, which through the iomap_begin operation usually just inserts
delalloc extents for the range in a single iteration. Only for
overwrites of existing data that are not block aligned, or zeroing
operations the existing extent mapping is read to fill out the srcmap
and to figure out if zeroing is required.
3) the ->map_blocks callback to the generic iomap writeback code
calls into the zoned space allocator to actually allocate on-disk
space for the range before kicking of the writeback.
Note that because all writes are out of place, truncate or hole punches
that are not aligned to block size boundaries need to allocate space.
For block zeroing from truncate, ->setattr is called with the iolock
(aka i_rwsem) already held, so a hacky deviation from the above
scheme is needed. In this case the space reservations is called with
the iolock held, but is required not to block and can dip into the
reserved block pool. This can lead to -ENOSPC when truncating a
file, which is unfortunate. But fixing the calling conventions in
the VFS is probably much easier with code requiring it already in
mainline.
Similarly because all writes are out place, the zoned allocator can't
support unwritten extents and thus the FALLOC_FL_ALLOCATE_RANGE range
mode of fallocate. Other fallocate modes that would reserved space
but don't need to to provide proper semantics do work but do not
reserve space.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: "Darrick J. Wong" <djwong@kernel.org>
|
|
Zone file systems reuse the basic RT group enabled XFS file system
structure to support a mode where each RT group is always written from
start to end and then reset for reuse (after moving out any remaining
data). There are few minor but important changes, which are indicated
by a new incompat flag:
1) there are no bitmap and summary inodes, thus the
/rtgroups/{rgno}.{bitmap,summary} metadir files do not exist and the
sb_rbmblocks superblock field must be cleared to zero.
2) there is a new superblock field that specifies the start of an
internal RT section. This allows supporting SMR HDDs that have random
writable space at the beginning which is used for the XFS data device
(which really is the metadata device for this configuration), directly
followed by a RT device on the same block device. While something
similar could be achieved using dm-linear just having a single device
directly consumed by XFS makes handling the file systems a lot easier.
3) Another superblock field that tracks the amount of reserved space (or
overprovisioning) that is never used for user capacity, but allows GC
to run more smoothly.
4) an overlay of the cowextsize field for the rtrmap inode so that we
can persistently track the total amount of rtblocks currently used in
a RT group. There is no data structure other than the rmap that
tracks used space in an RT group, and this counter is used to decide
when a RT group has been entirely emptied, and to select one that
is relatively empty if garbage collection needs to be performed.
While this counter could be tracked entirely in memory and rebuilt
from the rmap at mount time, that would lead to very long mount times
with the large number of RT groups implied by the number of hardware
zones especially on SMR hard drives with 256MB zone sizes.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: "Darrick J. Wong" <djwong@kernel.org>
|
|
Delalloc reservations are not supported in userspace, and thus it doesn't
make sense to share this helper with xfsprogs.c. Move it to xfs_iomap.c
toward the two callers.
Note that there rest of the delalloc handling should probably eventually
also move out of xfs_bmap.c, but that will require a bit more surgery.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: "Darrick J. Wong" <djwong@kernel.org>
|
|
xfs_{add,dec}_freecounter already handles the block and RT extent
percpu counters, but it currently hardcodes the passed in counter.
Add a freecounter abstraction that uses an enum to designate the counter
and add wrappers that hide the actual percpu_counters. This will allow
expanding the reserved block handling to the RT extent counter in the
next step, and also prepares for adding yet another such counter that
can share the code. Both these additions will be needed for the zoned
allocator.
Also switch the flooring of the frextents counter to 0 in statfs for the
rthinherit case to a manual min_t call to match the handling of the
fdblocks counter for normal file systems.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: "Darrick J. Wong" <djwong@kernel.org>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs into xfs-6.15-merge
|
|
Allow the file system to pass private data which can be used by the
iomap_begin and iomap_end methods through the private pointer in the
iomap_iter structure.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20250206064035.2323428-12-hch@lst.de
Reviewed-by: "Darrick J. Wong" <djwong@kernel.org>
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
Allow the file system to pass private data which can be used by the
iomap_begin and iomap_end methods through the private pointer in the
iomap_iter structure.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20250206064035.2323428-11-hch@lst.de
Reviewed-by: "Darrick J. Wong" <djwong@kernel.org>
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
xfs_dax_write_iomap_end
In xfs_dax_write_iomap_end(), directly return the result of
xfs_reflink_cancel_cow_range() when !written, ensuring proper
error propagation and improving code robustness.
Fixes: ea6c49b784f0 ("xfs: support CoW in fsdax mode")
Cc: stable@vger.kernel.org # v6.0
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Wentao Liang <vulab@iscas.ac.cn>
Signed-off-by: Carlos Maiolino <cem@kernel.org>
|
|
https://git.kernel.org/pub/scm/linux/kernel/git/djwong/xfs-linux into staging-merge
xfs: enable quota for realtime volumes [v5.5 08/10]
At some point, I realized that I've refactored enough of the quota code
in XFS that I should evaluate whether or not quota actually works on
realtime volumes. It turns out that it nearly works: the only broken
pieces are chown and delayed allocation, and reporting of project
quotas in the statvfs output for projinherit+rtinherit directories.
Fix these things and we can have realtime quotas again after 20 years.
With a bit of luck, this should all go splendidly.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
|
|
https://git.kernel.org/pub/scm/linux/kernel/git/djwong/xfs-linux into staging-merge
xfs: shard the realtime section [v5.5 06/10]
Right now, the realtime section uses a single pair of metadata inodes to
store the free space information. This presents a scalability problem
since every thread trying to allocate or free rt extents have to lock
these files. Solve this problem by sharding the realtime section into
separate realtime allocation groups.
While we're at it, define a superblock to be stamped into the start of
the rt section. This enables utilities such as blkid to identify block
devices containing realtime sections, and avoids the situation where
anything written into block 0 of the realtime extent can be
misinterpreted as file data.
The best advantage for rtgroups will become evident later when we get to
adding rmap and reflink to the realtime volume, since the geometry
constraints are the same for rt groups and AGs. Hence we can reuse all
that code directly.
This is a very large patchset, but it catches us up with 20 years of
technical debt that have accumulated.
With a bit of luck, this should all go splendidly.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
|
|
https://git.kernel.org/pub/scm/linux/kernel/git/djwong/xfs-linux into staging-merge
xfs: create incore rt allocation groups [v5.5 04/10]
Add in-memory data structures for sharding the realtime volume into
independent allocation groups. For existing filesystems, the entire rt
volume is modelled as having a single large group, with (potentially) a
number of rt extents exceeding 2^32 blocks, though these are not likely
to exist because the codebase has been a bit broken for decades. The
next series fills in the ondisk format and other supporting structures.
With a bit of luck, this should all go splendidly.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
|
|
Refactor the quota preallocation watermarking code so that it'll work
for realtime quota too. Convert the do_div calls into div_u64 for
compactness.
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
|
|
Unlike AGs, RTGs don't always have metadata in their first blocks, and
thus we don't get automatic protection from merging I/O completions
across RTG boundaries. Add code to set the IOMAP_F_BOUNDARY flag for
ioends that start at the first block of a RTG so that they never get
merged into the previous ioend.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
|
|
To prepare for adding per-rtgroup bitmap files, make the xfs_rtxnum_t
type encode the RT extent number relative to the rtgroup. The biggest
part of this to clearly distinguish between the relative extent number
that gets masked when converting from a global block number and length
values that just have a factor applied to them when converting from
file system blocks.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
|
|
Shared the regular buffered write iomap_ops with the page fault path
and just check for the IOMAP_FAULT flag to skip delalloc punching.
This keeps the delalloc punching checks in one place, and will make it
easier to convert iomap to an iter model where the begin and end
handlers are merged into a single callback.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Carlos Maiolino <cem@kernel.org>
|
|
gitolite.kernel.org:pub/scm/linux/kernel/git/vfs/vfs
Pull iomap fixes from Christian Brauner:
"Fixes for iomap to prevent data corruption bugs in the fallocate
unshare range implementation of fsdax and a small cleanup to turn
iomap_want_unshare_iter() into an inline function"
* tag 'vfs-6.12-rc6.iomap' of gitolite.kernel.org:pub/scm/linux/kernel/git/vfs/vfs:
iomap: turn iomap_want_unshare_iter into an inline function
fsdax: dax_unshare_iter needs to copy entire blocks
fsdax: remove zeroing code from dax_unshare_iter
iomap: share iomap_unshare_iter predicate code with fsdax
xfs: don't allocate COW extents when unsharing a hole
|
|
When ->iomap_end is called on a short write to the COW fork it needs to
punch stale delalloc data from the COW fork and not the data fork.
Ensure that IOMAP_F_NEW is set for new COW fork allocations in
xfs_buffered_write_iomap_begin, and then use the IOMAP_F_SHARED flag
in xfs_buffered_write_delalloc_punch to decide which fork to punch.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Carlos Maiolino <cem@kernel.org>
|
|
Change to always set xfs_buffered_write_iomap_begin for COW fork
allocations even if they don't overlap existing data fork extents,
which will allow the iomap_end callback to detect if it has to punch
stale delalloc blocks from the COW fork instead of the data fork. It
also means we sample the sequence counter for both the data and the COW
fork when writing to the COW fork, which ensures we properly revalidate
when only COW fork changes happens.
This is essentially a revert of commit 72a048c1056a ("xfs: only set
IOMAP_F_SHARED when providing a srcmap to a write"). This is fine because
the problem that the commit fixed has now been dealt with in iomap by
only looking at the actual srcmap and not the fallback to the write
iomap.
Note that the direct I/O path was never changed and has always set
IOMAP_F_SHARED for all COW fork allocations.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Carlos Maiolino <cem@kernel.org>
|
|
Introduce a local iomap_flags variable so that the code allocating new
delalloc blocks in the data fork can fall through to the found_imap
label and reuse the code to unlock and fill the iomap.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Carlos Maiolino <cem@kernel.org>
|
|
xfs_buffered_write_iomap_begin can also create delallocate reservations
that need cleaning up, prepare for that by adding support for the COW
fork in xfs_bmap_punch_delalloc_range.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Carlos Maiolino <cem@kernel.org>
|
|
All XFS callers of iomap_zero_range and iomap_file_unshare already hold
invalidate_lock, so we can't take it again in
iomap_file_buffered_write_punch_delalloc.
Use the passed in flags argument to detect if we're called from a zero
or unshare operation and don't take the lock again in this case.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Carlos Maiolino <cem@kernel.org>
|
|
xfs_file_write_zero_eof is the only caller of xfs_zero_range that does
not take XFS_MMAPLOCK_EXCL (aka the invalidate lock). Currently that
is actually the right thing, as an error in the iomap zeroing code will
also take the invalidate_lock to clean up, but to fix that deadlock we
need a consistent locking pattern first.
The only extra thing that XFS_MMAPLOCK_EXCL will lock out are read
pagefaults, which isn't really needed here, but also not actively
harmful.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Carlos Maiolino <cem@kernel.org>
|
|
XFS (which currently is the only user of iomap_write_delalloc_release)
already holds invalidate_lock for most zeroing operations. To be able
to avoid a deadlock it needs to stop taking the lock, but doing so
in iomap would leak XFS locking details into iomap.
To avoid this require the caller to hold invalidate_lock when calling
iomap_write_delalloc_release instead of taking it there.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Carlos Maiolino <cem@kernel.org>
|
|
Currently iomap_file_buffered_write_punch_delalloc can be called from
XFS either with the invalidate lock held or not. To fix this while
keeping the locking in the file system and not the iomap library
code we'll need to life the locking up into the file system.
To prepare for that, open code iomap_file_buffered_write_punch_delalloc
in the only caller, and instead export iomap_write_delalloc_release.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Carlos Maiolino <cem@kernel.org>
|
|
It doesn't make sense to allocate a COW extent when unsharing a hole
because holes cannot be shared.
Fixes: 1f1397b7218d7 ("xfs: don't allocate into the data fork for an unshare request")
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Link: https://lore.kernel.org/r/172796813277.1131942.5486112889531210260.stgit@frogsfrogsfrogs
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
iomap_file_buffered_write_punch_delalloc can only return errors if either
the ->punch callback returned an error, or if someone changed the API of
mapping_seek_hole_data to return a negative error code that is not
-ENXIO.
As the only instance of ->punch never returns an error, an such an error
would be fatal anyway remove the entire error propagation and don't
return an error code from iomap_file_buffered_write_punch_delalloc.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20240910043949.3481298-6-hch@lst.de
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
XFS will need to look at the flags in the iomap structure, so pass it
down all the way to the callback.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20240910043949.3481298-5-hch@lst.de
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
To fix short write error handling, We'll need to figure out what operation
iomap_file_buffered_write_punch_delalloc is called for. Pass the flags
argument on to it, and reorder the argument list to match that of
->iomap_end so that the compiler only has to add the new punch argument
to the end of it instead of reshuffling the registers.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20240910043949.3481298-4-hch@lst.de
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
About half of xfs_ilock_for_iomap deals with a special case for direct
I/O writes to COW files that need to take the ilock exclusively. Move
this code into the one callers that cares and simplify
xfs_ilock_for_iomap.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Chandan Babu R <chandanbabu@kernel.org>
|
|
xfs/205 produces the following failure when always_cow is enabled:
--- a/tests/xfs/205.out 2024-02-28 16:20:24.437887970 -0800
+++ b/tests/xfs/205.out.bad 2024-06-03 21:13:40.584000000 -0700
@@ -1,4 +1,5 @@
QA output created by 205
*** one file
+ !!! disk full (expected)
*** one file, a few bytes at a time
*** done
This is the result of overly aggressive attempts to align cow fork
delalloc reservations to the CoW extent size hint. Looking at the trace
data, we're trying to append a single fsblock to the "fred" file.
Trying to create a speculative post-eof reservation fails because
there's not enough space.
We then set @prealloc_blocks to zero and try again, but the cowextsz
alignment code triggers, which expands our request for a 1-fsblock
reservation into a 39-block reservation. There's not enough space for
that, so the whole write fails with ENOSPC even though there's
sufficient space in the filesystem to allocate the single block that we
need to land the write.
There are two things wrong here -- first, we shouldn't be attempting
speculative preallocations beyond what was requested when we're low on
space. Second, if we've already computed a posteof preallocation, we
shouldn't bother trying to align that to the cowextsize hint.
Fix both of these problems by adding a flag that only enables the
expansion of the delalloc reservation to the cowextsize if we're doing a
non-extending write, and only if we're not doing an ENOSPC retry. This
requires us to move the ENOSPC retry logic to xfs_bmapi_reserve_delalloc.
I probably should have caught this six years ago when 6ca30729c206d was
being reviewed, but oh well. Update the comments to reflect what the
code does now.
Fixes: 6ca30729c206d ("xfs: bmap code cleanup")
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Chandan Babu R <chandanbabu@kernel.org>
|
|
Currently the calls to xfs_iext_count_may_overflow and
xfs_iext_count_upgrade are always paired. Merge them into a single
function to simplify the callers and the actual check and upgrade
logic itself.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: "Darrick J. Wong" <djwong@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Chandan Babu R <chandanbabu@kernel.org>
|
|
Unreserving quotas can't fail due to quota limits, and we'll notice a
shut down file system a bit later in all the callers anyway. Return
void and remove the error checking and propagation in the callers.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: "Darrick J. Wong" <djwong@kernel.org>
Signed-off-by: Chandan Babu R <chandanbabu@kernel.org>
|
|
xfs_bmapi_write can return 0 without actually returning a mapping in
mval in two different cases:
1) when there is absolutely no space available to do an allocation
2) when converting delalloc space, and the allocation is so small
that it only covers parts of the delalloc extent before the
range requested by the caller
Callers at best can handle one of these cases, but in many cases can't
cope with either one. Switch xfs_bmapi_write to always return a
mapping or return an error code instead. For case 1) above ENOSPC is
the obvious choice which is very much what the callers expect anyway.
For case 2) there is no really good error code, so pick a funky one
from the SysV streams portfolio.
This fixes the reproducer here:
https://lore.kernel.org/linux-xfs/CAEJPjCvT3Uag-pMTYuigEjWZHn1sGMZ0GCjVVCv29tNHK76Cgg@mail.gmail.com0/
which uses reserved blocks to create file systems that are gravely
out of space and thus cause at least xfs_file_alloc_space to hang
and trigger the lack of ENOSPC handling in xfs_dquot_disk_alloc.
Note that this patch does not actually make any caller but
xfs_alloc_file_space deal intelligently with case 2) above.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reported-by: 刘通 <lyutoon@gmail.com>
Reviewed-by: "Darrick J. Wong" <djwong@kernel.org>
Signed-off-by: Chandan Babu R <chandanbabu@kernel.org>
|
|
Current clone operation could be non-atomic if the destination of a file
is beyond EOF, user could get a file with corrupted (zeroed) data on
crash.
The problem is about preallocations. If you write some data into a file:
[A...B)
and XFS decides to preallocate some post-eof blocks, then it can create
a delayed allocation reservation:
[A.........D)
The writeback path tries to convert delayed extents to real ones by
allocating blocks. If there aren't enough contiguous free space, we can
end up with two extents, the first real and the second still delalloc:
[A....C)[C.D)
After that, both the in-memory and the on-disk file sizes are still B.
If we clone into the range [E...F) from another file:
[A....C)[C.D) [E...F)
then xfs_reflink_zero_posteof() calls iomap_zero_range() to zero out the
range [B, E) beyond EOF and flush it. Since [C, D) is still a delalloc
extent, its pagecache will be zeroed and both the in-memory and on-disk
size will be updated to D after flushing but before cloning. This is
wrong, because the user can see the size change and read the zeroes
while the clone operation is ongoing.
We need to keep the in-memory and on-disk size before the clone
operation starts, so instead of writing zeroes through the page cache
for delayed ranges beyond EOF, we convert these ranges to unwritten and
invalidate any cached data over that range beyond EOF.
Suggested-by: Dave Chinner <david@fromorbit.com>
Signed-off-by: Zhang Yi <yi.zhang@huawei.com>
Reviewed-by: "Darrick J. Wong" <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Chandan Babu R <chandanbabu@kernel.org>
|
