| Age | Commit message (Collapse) | Author |
|---|
|
Instead of using __clear_extent_bit() we can use clear_extent_bit() since
we pass a NULL value for the changeset argument.
Reviewed-by: Boris Burkov <boris@bur.io>
Reviewed-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
|
|
[BUG WITH EXPERIMENTAL LARGE FOLIOS]
When testing the experimental large data folio support with compression,
there are several ASSERT()s triggered from btrfs_decompress_buf2page()
when running fsstress with compress=zstd mount option:
- ASSERT(copy_len) from btrfs_decompress_buf2page()
- VM_BUG_ON(offset + len > PAGE_SIZE) from memcpy_to_page()
[CAUSE]
Inside btrfs_decompress_buf2page(), we need to grab the file offset from
the current bvec.bv_page, to check if we even need to copy data into the
bio.
And since we're using single page bvec, and no large folio, every page
inside the folio should have its index properly setup.
But when large folios are involved, only the first page (aka, the head
page) of a large folio has its index properly initialized.
The other pages inside the large folio will not have their indexes
properly initialized.
Thus the page_offset() call inside btrfs_decompress_buf2page() will
result garbage, and completely screw up the @copy_len calculation.
[FIX]
Instead of using page->index directly, go with page_pgoff(), which can
handle non-head pages correctly.
So introduce a helper, file_offset_from_bvec(), to get the file offset
from a single page bio_vec, so the copy_len calculation can be done
correctly.
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: Sweet Tea Dorminy <sweettea-kernel@dorminy.me>
Reviewed-by: Boris Burkov <boris@bur.io>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
|
|
Simplify conditionally reading an rb_entry(), there's the
rb_entry_safe() helper that checks the node pointer for NULL so we don't
have to write it explicitly.
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: David Sterba <dsterba@suse.com>
|
|
Allow get_range_bits() to take an extent state pointer to pointer argument
so that we can cache the first extent state record in the target range, so
that a caller can use it for subsequent operations without doing a full
tree search. Currently the only user is try_release_extent_state(), which
then does a call to __clear_extent_bit() which can use such a cached state
record.
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
|
|
When the release_folio callback (from struct address_space_operations) is
invoked we don't allow the folio to be released if its range is currently
locked in the inode's io_tree, as it may indicate the folio may be needed
by the task that locked the range.
However if the range is locked because an ordered extent is finishing,
then we can safely allow the folio to be released because ordered extent
completion doesn't need to use the folio at all.
When we are under memory pressure, the kernel starts writeback of dirty
pages (folios) with the goal of releasing the pages from the page cache
after writeback completes, however this often is not possible on btrfs
because:
* Once the writeback completes we queue the ordered extent completion;
* Once the ordered extent completion starts, we lock the range in the
inode's io_tree (at btrfs_finish_one_ordered());
* If the release_folio callback is called while the folio's range is
locked in the inode's io_tree, we don't allow the folio to be
released, so the kernel has to try to release memory elsewhere,
which may result in triggering more writeback or releasing other
pages from the page cache which may be more useful to have around
for applications.
In contrast, when the release_folio callback is invoked after writeback
finishes and before ordered extent completion starts or locks the range,
we allow the folio to be released, as well as when the release_folio
callback is invoked after ordered extent completion unlocks the range.
Improve on this by detecting if the range is locked for ordered extent
completion and if it is, allow the folio to be released. This detection
is achieved by adding a new extent flag in the io_tree that is set when
the range is locked during ordered extent completion.
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
|
|
Drop reference to pages from the comment since the function is fully folio
aware and works regardless of how many pages are in the folio. Also while
at it, capitalize the first word and make it more explicit that
release_folio is a callback from struct address_space_operations.
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
|
|
The function btrfs_punch_hole_lock_range() needs to make sure there is
no other folio in the range, thus it goes with filemap_range_has_page(),
which works pretty fine.
But if we have large folios, under the following case
filemap_range_has_page() will always return true, forcing
btrfs_punch_hole_lock_range() to do a very time consuming busy loop:
start end
| |
|//|//|//|//| | | | | | | | |//|//|
\ / \ /
Folio A Folio B
In the above case, folio A and B contain our start/end indexes, and there
are no other folios in the range. Thus we do not need to retry inside
btrfs_punch_hole_lock_range().
To prepare for large data folios, introduce a helper,
check_range_has_page(), which will:
- Shrink the search range towards page boundaries
If the rounded down end (exclusive, otherwise it can underflow when @end
is inside the folio at file offset 0) is no larger than the rounded up
start, it means the range contains no other pages other than the ones
covering @start and @end.
Can return false directly in that case.
- Grab all the folios inside the range
- Skip any large folios that cover the start and end indexes
- If any other folios are found return true
- Otherwise return false
This new helper is going to handle both large folios and regular ones.
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
|
|
This involves the following modifications:
- Set the order flags for __filemap_get_folio() inside
prepare_one_folio()
This will allow __filemap_get_folio() to create a large folio if the
address space supports it.
- Limit the initial @write_bytes inside copy_one_range()
If the largest folio boundary splits the initial write range, there is
no way we can write beyond the largest folio boundary.
This is done by a simple helper calc_write_bytes().
- Release exceeding reserved space if the folio is smaller than expected
Which is doing the same handling when short copy happens.
All the preparations should not change the behavior when the largest
folio order is 0.
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
|
|
There are several things not ideal in copy_one_range():
- Unnecessary temporary variables
* block_offset
* reserve_bytes
* dirty_blocks
* num_blocks
* release_bytes
These are utilized to handle short-copy cases.
- Inconsistent handling of btrfs_delalloc_release_extents()
There is a hidden behavior that, after reserving metadata for X bytes
of data write, we have to call btrfs_delalloc_release_extents() with X
once and only once.
Calling btrfs_delalloc_release_extents(X - 4K) and
btrfs_delalloc_release_extents(4K) will cause outstanding extents
accounting to go wrong.
This is because the outstanding extents mechanism is not designed to
handle shrinking of reserved space.
Improve above situations by:
- Use a single @reserved_start and @reserved_len pair
Now we reserve space for the initial range, and if a short copy
happened and we need to shrink the reserved space, we can easily
calculate the new length, and update @reserved_len.
- Introduce helpers to shrink reserved data and metadata space
This is done by two new helpers, shrink_reserved_space() and
btrfs_delalloc_shrink_extents().
The later will do a better calculation if we need to modify the
outstanding extents, and the first one will be utilized inside
copy_one_range().
- Manually unlock, release reserved space and return if no byte is
copied
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
|
|
The EXTENT_UPTODATE io tree flag is now used only to mark ranges in the
fs_info->excluded_extents as used by super blocks and not available for
extent allocation (to prevent adding those ranges as free space in the
in memory space caches). As we can use any flag for that purpose, and
we are using EXTENT_DIRTY for the pinned extents io tree for example,
remove the EXTENT_UPTODATE flag and use instead EXTENT_DIRTY for the
excluded extents io tree.
Reviewed-by: Boris Burkov <boris@bur.io>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
|
|
At btrfs_add_new_free_space() we keep searching for ranges in the excluded
extents io tree that have the EXTENT_DIRTY bit set, however we never ever
set that bit for ranges in that tree. That is a leftover from when that
function used the global freed extents trees (fs_info->freed_extents[2]),
where we used both the EXTENT_DIRTY and EXTENT_UPTODATE bits, but those
trees are gone with commit fe119a6eeb67 ("btrfs: switch to per-transaction
pinned extents"), which introduced the fs_info->excluded_extents io tree,
where only EXTENT_UPTODATE is set.
So remove the EXTENT_DIRTY bit search at btrfs_add_new_free_space().
Reviewed-by: Boris Burkov <boris@bur.io>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
|
|
After commit 52b029f42751 ("btrfs: remove unnecessary EXTENT_UPTODATE
state in buffered I/O path") we never set EXTENT_UPTODATE in an inode's
io_tree anymore, but we still have some code attempting to clear that
bit from an inode's io_tree. Remove that code as it doesn't do anything
anymore. The sole use of the EXTENT_UPTODATE bit is for the excluded
extents io_tree (fs_info->excluded_extents), which is used to track the
locations of super blocks, so that their ranges are never marked as free,
making them unavailable for extent allocation.
Reviewed-by: Boris Burkov <boris@bur.io>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
|
|
If we fsync a file (or directory) that has no more hard links, because
while a process had a file descriptor open on it, the file's last hard
link was removed and then the process did an fsync against the file
descriptor, after a power failure or crash the file still exists after
replaying the log.
This behaviour is incorrect since once an inode has no more hard links
it's not accessible anymore and we insert an orphan item into its
subvolume's tree so that the deletion of all its items is not missed in
case of a power failure or crash.
So after log replay the file shouldn't exist anymore, which is also the
behaviour on ext4, xfs, f2fs and other filesystems.
Fix this by not ignoring inodes with zero hard links at
btrfs_log_inode_parent() and by committing an inode's delayed inode when
we are not doing a fast fsync (either BTRFS_INODE_COPY_EVERYTHING or
BTRFS_INODE_NEEDS_FULL_SYNC is set in the inode's runtime flags). This
last step is necessary because when removing the last hard link we don't
delete the corresponding ref (or extref) item, instead we record the
change in the inode's delayed inode with the BTRFS_DELAYED_NODE_DEL_IREF
flag, so that when the delayed inode is committed we delete the ref/extref
item from the inode's subvolume tree - otherwise the logging code will log
the last hard link and therefore upon log replay the inode is not deleted.
The base code for a fstests test case that reproduces this bug is the
following:
. ./common/dmflakey
_require_scratch
_require_dm_target flakey
_require_mknod
_scratch_mkfs >>$seqres.full 2>&1 || _fail "mkfs failed"
_require_metadata_journaling $SCRATCH_DEV
_init_flakey
_mount_flakey
touch $SCRATCH_MNT/foo
# Commit the current transaction and persist the file.
_scratch_sync
# A fifo to communicate with a background xfs_io process that will
# fsync the file after we deleted its hard link while it's open by
# xfs_io.
mkfifo $SCRATCH_MNT/fifo
tail -f $SCRATCH_MNT/fifo | \
$XFS_IO_PROG $SCRATCH_MNT/foo >>$seqres.full &
XFS_IO_PID=$!
# Give some time for the xfs_io process to open a file descriptor for
# the file.
sleep 1
# Now while the file is open by the xfs_io process, delete its only
# hard link.
rm -f $SCRATCH_MNT/foo
# Now that it has no more hard links, make the xfs_io process fsync it.
echo "fsync" > $SCRATCH_MNT/fifo
# Terminate the xfs_io process so that we can unmount.
echo "quit" > $SCRATCH_MNT/fifo
wait $XFS_IO_PID
unset XFS_IO_PID
# Simulate a power failure and then mount again the filesystem to
# replay the journal/log.
_flakey_drop_and_remount
# We don't expect the file to exist anymore, since it was fsynced when
# it had no more hard links.
[ -f $SCRATCH_MNT/foo ] && echo "file foo still exists"
_unmount_flakey
# success, all done
echo "Silence is golden"
status=0
exit
A test case for fstests will be submitted soon.
Reviewed-by: Boris Burkov <boris@bur.io>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
|
|
There's an explanation of how space info works at the top of
fs/btrfs/space-info.c, which makes reference to a variable called
bytes_may_reserve. There's nothing called that in the code, and wasn't
at time the comment was written; as far I can tell this is a typo, and
it should actually be bytes_may_use.
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Mark Harmstone <maharmstone@fb.com>
Signed-off-by: David Sterba <dsterba@suse.com>
|
|
This flag is set after inserting the eb to the buffer tree and cleared
on it's removal. It was added in commit 34b41acec1ccc0 ("Btrfs: use a
bit to track if we're in the radix tree") and wanted to make use of it,
faa2dbf004e89e ("Btrfs: add sanity tests for new qgroup accounting
code"). Both are 10+ years old, we can remove the flag.
Signed-off-by: Daniel Vacek <neelx@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
|
|
This flag is no longer being used. It was added by commit a826d6dcb32d
("Btrfs: check items for correctness as we search") but it's no longer
being used after commit f26c92386028 ("btrfs: remove reada
infrastructure").
Reviewed-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: Daniel Vacek <neelx@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
|
|
This flag is no longer being used. It was added by commit ab0fff03055d
("btrfs: add READAHEAD extent buffer flag") and used in commits:
79fb65a1f6d9 ("Btrfs: don't call readahead hook until we have read the entire eb")
78e62c02abb9 ("btrfs: Remove extent_io_ops::readpage_io_failed_hook")
371cdc0700c7 ("btrfs: introduce subpage metadata validation check")
Finally all the code using it was removed by commit f26c92386028 ("btrfs: remove
reada infrastructure").
Reviewed-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: Daniel Vacek <neelx@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
|
|
This flag was added by commit 656f30dba7ab ("Btrfs: be aware of btree
inode write errors to avoid fs corruption") but it stopped being used
after commit 046b562b20a5 ("btrfs: use a separate end_io handler for
read_extent_buffer").
Reviewed-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: Daniel Vacek <neelx@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
|
|
Inside the main loop of btrfs_buffered_write() we are doing a lot of
heavy lifting inside a while() loop.
This makes it pretty hard to read, factor out the content into a helper,
copy_one_range() to do the work.
This has no functional change, but with some minor variable renames,
e.g. rename all "sector" into "block".
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
|
|
Inside the main loop of btrfs_buffered_write(), we have a complex data
and metadata space reservation code, which tries to reserve space for
a COW write, if failed then fallback to check if we can do a NOCOW
write.
Factor out that part of code into a dedicated helper, reserve_space(),
to make the main loop a little easier to read.
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
|
|
Inside the main loop of btrfs_buffered_write(), if something wrong
happened, there is a out-of-loop cleanup path to release the reserved
space.
This behavior saves some code lines, but makes it much harder to read,
as we need to check release_bytes to make sure when we need to do the
cleanup.
Factor out the cleanup part into a helper, release_reserved_space(), to
do the cleanup inside the main loop, so that we can move @release_bytes
inside the loop.
This will make later refactoring of the main loop much easier.
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
|
|
Commit c87c299776e4 ("btrfs: make buffered write to copy one page a
time") changed how the variable @force_page_uptodate was updated.
Before that commit the variable was only initialized to false at the
beginning of the function, and after hitting a short copy, the next
retry on the same folio would force the folio to be read from the disk.
But after the commit, the variable is always initialized to false at the
beginning of the loop's scope, causing prepare_one_folio() never to get a
true value passed in.
The change in behavior is not a huge deal, it only makes a difference
on how we handle short copies:
Old: Allow the buffer to be split
The first short copy will be rejected, that's the same for both
cases.
But for the next retry, we require the folio to be read from disk.
Then even if we hit a short copy again, since the folio is already
uptodate, we do not need to handle partial uptodate range, and can
continue, marking the short copied range as dirty and continue.
This will split the buffer write into the folio as two buffered
writes.
New: Do not allow the buffer to be split
The first short copy will be rejected, that's the same for both
cases.
For the next retry, we do nothing special, thus if the short copy
happened again, we reject it again, until either the short copy is
gone, or we failed to fault in the buffer.
This will mean the buffer write into the folio will either fail or
succeed, no splitting will happen.
To me, either solution is fine, but the new one makes it simpler and
requires no special handling, so I prefer that solution.
And since @force_page_uptodate is always false when passed into
prepare_one_folio(), we can just remove the variable.
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
|
|
Commit 1d2fbb7f1f9e ("btrfs: allow compression even if the range is not
page aligned") introduced the block perfect compression for block size <
page size cases.
Before that commit, if the fs block size is smaller than page size (aka
subpage cases), compressed write is only enabled if the dirty range is
fully page aligned.
This block perfect compression support was introduced in v6.13, and has
been tested for two kernel releases.
I believe it's time to move it out of experimental features so that we
can get more tests in the real world.
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
|
|
All callers now have a folio. Pass it to page_put_link(), saving a
hidden call to compound_head(). Also add kernel-doc for page_get_link()
and page_put_link().
Signed-off-by: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Link: https://lore.kernel.org/20250514171316.3002934-4-willy@infradead.org
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
Mirror the changes to __page_get_link() by retrieving a folio from
the page cache instead of a page. Removes two hidden calls to
compound_head().
Signed-off-by: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Link: https://lore.kernel.org/20250514171316.3002934-3-willy@infradead.org
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
Retrieve a folio from the page cache instead of a page and operate
on it. Removes two hidden calls to compound_head().
Signed-off-by: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Link: https://lore.kernel.org/20250514171316.3002934-2-willy@infradead.org
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
Allows killing processes that are waiting for the inode lock.
Signed-off-by: Max Kellermann <max.kellermann@ionos.com>
Link: https://lore.kernel.org/20250513150327.1373061-4-max.kellermann@ionos.com
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
Allows killing processes that are waiting for the inode lock.
Signed-off-by: Max Kellermann <max.kellermann@ionos.com>
Link: https://lore.kernel.org/20250513150327.1373061-3-max.kellermann@ionos.com
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
Allows killing processes that are waiting for the inode lock.
Signed-off-by: Max Kellermann <max.kellermann@ionos.com>
Link: https://lore.kernel.org/20250513150327.1373061-2-max.kellermann@ionos.com
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
This is a preparation for large readdir buffers in fuse.
Simply setting the fuse buffer size to the userspace buffer size should
work, the record sizes are similar (fuse's is slightly larger than libc's,
so no overflow should ever happen).
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
Signed-off-by: Jaco Kroon <jaco@uls.co.za>
Link: https://lore.kernel.org/20250513151012.1476536-1-mszeredi@redhat.com
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
On our HDFS servers with 12 HDDs per server, a HDFS datanode[0] startup
involves scanning all files and caching their metadata (including dentries
and inodes) in memory. Each HDD contains approximately 2 million files,
resulting in a total of ~20 million cached dentries after initialization.
To minimize dentry reclamation, we set vfs_cache_pressure to 1. Despite
this configuration, memory pressure conditions can still trigger
reclamation of up to 50% of cached dentries, reducing the cache from 20
million to approximately 10 million entries. During the subsequent cache
rebuild period, any HDFS datanode restart operation incurs substantial
latency penalties until full cache recovery completes.
To maintain service stability, we need to preserve more dentries during
memory reclamation. The current minimum reclaim ratio (1/100 of total
dentries) remains too aggressive for our workload. This patch introduces
vfs_cache_pressure_denom for more granular cache pressure control. The
configuration [vfs_cache_pressure=1, vfs_cache_pressure_denom=10000]
effectively maintains the full 20 million dentry cache under memory
pressure, preventing datanode restart performance degradation.
Link: https://hadoop.apache.org/docs/r1.2.1/hdfs_design.html#NameNode+and+DataNodes [0]
Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
Link: https://lore.kernel.org/20250511083624.9305-1-laoar.shao@gmail.com
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
When getting the directory contents, the entries are first fetched to a
kernel buffer, then they are copied to userspace with dir_emit(). This
second phase is non-blocking as long as the userspace buffer is not paged
out, making it interruptible makes zero sense.
Overload d_type as flags, since it only uses 4 bits from 32.
Reviewed-by: Bernd Schubert <bschubert@ddn.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
Link: https://lore.kernel.org/20250513112335.1473177-1-mszeredi@redhat.com
Signed-off-by: Christian Brauner <brauner@kernel.org>
|
|
For multiple devices, both primary and extra devices should be the
same type. `erofs_init_device` has already guaranteed that if the
primary is a file-backed device, extra devices should also be
regular files.
However, if the primary is a block device while the extra device
is a file-backed device, `erofs_init_device` will get an ENOTBLK,
which is not treated as an error in `erofs_fc_get_tree`, and that
leads to an UAF:
erofs_fc_get_tree
get_tree_bdev_flags(erofs_fc_fill_super)
erofs_read_superblock
erofs_init_device // sbi->dif0 is not inited yet,
// return -ENOTBLK
deactivate_locked_super
free(sbi)
if (err is -ENOTBLK)
sbi->dif0.file = filp_open() // sbi UAF
So if -ENOTBLK is hitted in `erofs_init_device`, it means the
primary device must be a block device, and the extra device
is not a block device. The error can be converted to -EINVAL.
Fixes: fb176750266a ("erofs: add file-backed mount support")
Signed-off-by: Sheng Yong <shengyong1@xiaomi.com>
Reviewed-by: Gao Xiang <hsiangkao@linux.alibaba.com>
Reviewed-by: Hongbo Li <lihongbo22@huawei.com>
Link: https://lore.kernel.org/r/20250515014837.3315886-1-shengyong1@xiaomi.com
Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
|
|
Convert ufs to the new mount API
Removed a BUG_ON from show_options(); worst case, it prints
a null string where the BUG_ON would have fired.
Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
|
|
Have ufs_parse_options() refuse to change an already set flavour,
and move the "can't change the flavour" logics on remount there,
while we are at it - the only difference is that flavour is
already set.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
|
|
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
|
|
EROFS uses NID to indicate the on-disk inode offset, which can
exceed 32 bits. However, the default encode_fh uses the ino32,
thus it doesn't work if the image is larger than 128GiB.
Let's introduce our own helpers to encode file handles.
It's easy to reproduce:
1. prepare an erofs image with nid bigger than U32_MAX
2. mount -t erofs foo.img /mnt/erofs
3. set exportfs with configuration: /mnt/erofs *(rw,sync,
no_root_squash)
4. mount -t nfs $IP:/mnt/erofs /mnt/nfs
5. md5sum /mnt/nfs/foo # foo is the file which nid bigger
than U32_MAX. # you will get ESTALE error.
In the case of overlayfs, the underlying filesystem's file
handle is encoded in ovl_fb.fid, which is similar to NFS's
case. If the NID of file is larger than U32_MAX, the overlay
will get -ESTALE error when calls exportfs_decode_fh.
Fixes: 3e917cc305c6 ("erofs: make filesystem exportable")
Signed-off-by: Hongbo Li <lihongbo22@huawei.com>
Reviewed-by: Gao Xiang <hsiangkao@linux.alibaba.com>
Link: https://lore.kernel.org/r/20250507094015.14007-1-lihongbo22@huawei.com
Signed-off-by: Gao Xiang <hsiangkao@linux.alibaba.com>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux
Pull btrfs fixes from David Sterba:
- fix potential endless loop when discarding a block group when
disabling discard
- reinstate message when setting a large value of mount option 'commit'
- fix a folio leak when async extent submission fails
* tag 'for-6.15-rc6-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux:
btrfs: add back warning for mount option commit values exceeding 300
btrfs: fix folio leak in submit_one_async_extent()
btrfs: fix discard worker infinite loop after disabling discard
|
|
cifs_prepare_read() might be called with a disconnected channel, where
TCP_Server_Info::max_read is set to zero due to reconnect, so calling
->negotiate_rize() will set @rsize to default min IO size (64KiB) and
then logging
CIFS: VFS: SMB: Zero rsize calculated, using minimum value
65536
If the reconnect happens in cifsd thread, cifs_renegotiate_iosize()
will end up being called and then @rsize set to the expected value.
Since we can't rely on the value of @server->max_read by the time we
call cifs_prepare_read(), try to ->negotiate_rize() only if
@cifs_sb->ctx->rsize is zero.
Reported-by: Steve French <stfrench@microsoft.com>
Fixes: c59f7c9661b9 ("smb: client: ensure aligned IO sizes")
Signed-off-by: Paulo Alcantara (Red Hat) <pc@manguebit.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
|
|
The response buffer for the CREATE request handled by smb311_posix_mkdir()
is leaked on the error path (goto err_free_rsp_buf) because the structure
pointer *rsp passed to free_rsp_buf() is not assigned until *after* the
error condition is checked.
As *rsp is initialised to NULL, free_rsp_buf() becomes a no-op and the leak
is instead reported by __kmem_cache_shutdown() upon subsequent rmmod of
cifs.ko if (and only if) the error path has been hit.
Pass rsp_iov.iov_base to free_rsp_buf() instead, similar to the code in
other functions in smb2pdu.c for which *rsp is assigned late.
Cc: stable@vger.kernel.org
Signed-off-by: Jethro Donaldson <devel@jro.nz>
Signed-off-by: Steve French <stfrench@microsoft.com>
|
|
fsck_err() needs the btree transaction passed to it if there is one - so
that it can unlock/relock around prompting userspace for fixing the
error.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
Convert the orangefs filesystem to the new internal mount API as the old
one will be obsoleted and removed. This allows greater flexibility in
communication of mount parameters between userspace, the VFS and the
filesystem.
See Documentation/filesystems/mount_api.txt for more information.
[sandeen: forward-port older patch, fix SB_POSIXACL handling]
Signed-off-by: David Howells <dhowells@redhat.com>
Co-developed-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Eric Sandeen <sandeen@redhat.com>
cc: Mike Marshall <hubcap@omnibond.com>
cc: Martin Brandenburg <martin@omnibond.com>
cc: devel@lists.orangefs.org
Signed-off-by: Mike Marshall <hubcap@omnibond.com>
|
|
Fsck wants to do transaction commits from an outer context; it may have
other repair to do (i.e. duplicate backpointers).
But when calling backpointer_not_found() from runtime code, i.e. runtime
self healing, we should be doing the commit - the outer context expects
to just be doing lookups.
This fixes bugs where we get stuck spinning, reported as "RCU lock hold
time warnings.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
Since bch2_seek_pagecache_data() searches for dirty data, we only want
to call it for holes in the extents btree - otherwise we have an
accidental O(n^2), as we repeatedly search the same range.
Reported-by: Marcin Mirosław <marcin@mejor.pl>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
set_should_be_locked() needs to be called before peek_key_cache(), which
traverses other paths and may do a trans unlock/relock.
This fixes an assertion pop in path_peek_slot(), when the path we're
using is unexpectedly not uptodate.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
Before invoking bch2_accounting_mem_mod_locked in
bch2_gc_accounting_done, we already write locked mark_lock,
in bch2_accounting_mem_insert, we lock mark_lock again.
Signed-off-by: Alan Huang <mmpgouride@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
Prevent jobs that do lots of scanning (i.e. evacuatee, scrub) from
causing OOMs.
The shrinker code seems to be having issues when it doesn't do any
freeing because it's just flipping off the acccessed bit - and the
accessed bit shouldn't be set on first use anyways.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
When the journal is low on space, we might do discards from
journal_res_get() -> journal_entry_open().
Make sure we set j->can_discard correctly, so that if we're low on space
but not because discards aren't keeping up we don't livelock.
Fixes: 8e4d28036c29 ("bcachefs: Don't aggressively discard the journal")
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
This fixes btree locking assert pops users were seeing during evacuate:
https://github.com/koverstreet/bcachefs/issues/878
May 09 22:45:02 sharon kernel: bcachefs (68116e25-fa2d-4c6f-86c7-e8b431d792ae): bch2_btree_insert_node(): node not locked at level 1
May 09 22:45:02 sharon kernel: bch2_btree_node_rewrite [bcachefs]: watermark=btree no_check_rw alloc l=0-1 mode=none nodes_written=0 cl.remaining=2 journal_seq=0
May 09 22:45:02 sharon kernel: path: idx 1 ref 1:0 S B btree=alloc level=0 pos 0:3699637:0 0:3698012:1-0:3699637:0 bch2_move_btree.isra.0+0x1db/0x490 [bcachefs] uptodate 0 locks_want 2
May 09 22:45:02 sharon kernel: l=0 locks intent seq 4 node ffff8bd700c93600
May 09 22:45:02 sharon kernel: l=1 locks unlocked seq 1712 node ffff8bd6fd5e7a00
May 09 22:45:02 sharon kernel: l=2 locks unlocked seq 2295 node ffff8bd6cc725400
May 09 22:45:02 sharon kernel: l=3 locks unlocked seq 0 node 0000000000000000
Evacuate walks btree nodes with bch2_btree_iter_next_node() and rewrites
them, bch2_btree_update_start() upgrades the path to take intent locks
as far as it needs to.
But next_node() does low level unlock/relock calls on individual nodes,
and didn't handle the case where a path is supposed to be holding
multiple intent locks. If a path has locks_want > 1, it needs to be
either holding locks on all the btree nodes (at each level) requested,
or none of them.
Fix this with a bch2_btree_path_downgrade().
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
|
Fix bch2_bkey_clear_needs_rebalance(): indirect extents are never
supposed to have bch_extent_rebalance stripped off, because that's how
we get the IO path options when we don't have the original inode it
belonged to.
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
|
