The root cause is after commit 84b5bb8bf0f6 ("f2fs: modify
f2fs_is_checkpoint_ready logic to allow more data to be written with the
CP disable"), we will get chance to allow f2fs_is_checkpoint_ready() to
return true once below conditions are all true:
1. checkpoint is disabled
2. there are not enough free segments
3. there are enough free blocks
Then it will cause f2fs_balance_fs() to trigger foreground GC.
void f2fs_balance_fs(struct f2fs_sb_info *sbi, bool need)
...
if (!f2fs_is_checkpoint_ready(sbi))
return;
And the testcase mounts f2fs image w/ gc_merge,checkpoint=disable, so deadloop
will happen through below race condition:
This patch fix to add check condition in f2fs_balance_fs(), so that if
checkpoint is disabled, we will just skip trigger foreground GC to
avoid such deadloop issue.
Meanwhile let's remove f2fs_is_checkpoint_ready() check condition in
f2fs_balance_fs(), since it's redundant, due to the main logic in the
function is to check:
a) whether checkpoint is disabled
b) there is enough free segments
f2fs_balance_fs() still has all logics after f2fs_is_checkpoint_ready()'s
removal.
Reported-by: syzbot+aa5bb5f6860e08a60450@syzkaller.appspotmail.com Closes: https://lore.kernel.org/linux-f2fs-devel/682d743a.a00a0220.29bc26.0289.GAE@google.com Fixes: 84b5bb8bf0f6 ("f2fs: modify f2fs_is_checkpoint_ready logic to allow more data to be written with the CP disable") Cc: Qi Han <hanqi@vivo.com> Signed-off-by: Chao Yu <chao@kernel.org> Reviewed-by: Zhiguo Niu <zhiguo.niu@unisoc.com> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
yohan.joung [Mon, 12 May 2025 07:36:10 +0000 (16:36 +0900)]
f2fs: add ckpt_valid_blocks to the section entry
when performing buffered writes in a large section,
overhead is incurred due to the iteration through
ckpt_valid_blocks within the section.
when SEGS_PER_SEC is 128, this overhead accounts for 20% within
the f2fs_write_single_data_page routine.
as the size of the section increases, the overhead also grows.
to handle this problem ckpt_valid_blocks is
added within the section entries.
yohan.joung [Mon, 12 May 2025 07:36:09 +0000 (16:36 +0900)]
f2fs: add a method for calculating the remaining blocks in the current segment in LFS mode.
In LFS mode, the previous segment cannot use invalid blocks,
so the remaining blocks from the next_blkoff of the current segment
to the end of the section are calculated.
Chao Yu [Tue, 13 May 2025 05:57:20 +0000 (13:57 +0800)]
f2fs: use vmalloc instead of kvmalloc in .init_{,de}compress_ctx
.init_{,de}compress_ctx uses kvmalloc() to alloc memory, it will try
to allocate physically continuous page first, it may cause more memory
allocation pressure, let's use vmalloc instead to mitigate it.
[Test]
cd /data/local/tmp
touch file
f2fs_io setflags compression file
f2fs_io getflags file
for i in $(seq 1 10); do sync; echo 3 > /proc/sys/vm/drop_caches;\
time f2fs_io write 512 0 4096 zero osync file; truncate -s 0 file;\
done
[Result]
Before After Delta
21.243 21.694 -2.12%
For compression, we recommend to use ioctl to compress file data in
background for workaround.
For decompression, only zstd will be affected.
Signed-off-by: Chao Yu <chao@kernel.org> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Chao Yu [Tue, 13 May 2025 11:25:38 +0000 (19:25 +0800)]
f2fs: don't over-report free space or inodes in statvfs
This fixes an analogus bug that was fixed in modern filesystems:
a) xfs in commit 4b8d867ca6e2 ("xfs: don't over-report free space or
inodes in statvfs")
b) ext4 in commit f87d3af74193 ("ext4: don't over-report free space
or inodes in statvfs")
where statfs can report misleading / incorrect information where
project quota is enabled, and the free space is less than the
remaining quota.
This commit will resolve a test failure in generic/762 which tests
for this bug.
generic/762 - output mismatch (see /share/git/fstests/results//generic/762.out.bad)
--- tests/generic/762.out 2025-04-15 10:21:53.371067071 +0800
+++ /share/git/fstests/results//generic/762.out.bad 2025-05-13 16:13:37.000000000 +0800
@@ -6,8 +6,10 @@
root blocks2 is in range
dir blocks2 is in range
root bavail2 is in range
-dir bavail2 is in range
+dir bavail2 has value of 1539066
+dir bavail2 is NOT in range 304734.87 .. 310891.13
root blocks3 is in range
...
(Run 'diff -u /share/git/fstests/tests/generic/762.out /share/git/fstests/results//generic/762.out.bad' to see the entire diff)
HINT: You _MAY_ be missing kernel fix:
XXXXXXXXXXXXXX xfs: don't over-report free space or inodes in statvfs
f2fs: simplify return value handling in f2fs_fsync_node_pages
Always assign ret where the error happens, and jump to out instead
of multiple loop exit conditions to prepare for changes in the
__write_node_folio calling convention.
Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Chao Yu <chao@kernel.org> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
f2fs: always unlock the page in f2fs_write_single_data_page
Consolidate the code to unlock the page in f2fs_write_single_data_page
instead of leaving it to the callers for the AOP_WRITEPAGE_ACTIVATE case.
Replace AOP_WRITEPAGE_ACTIVATE with a positive return of 1 as this case
now doesn't match the historic ->writepage special return code that is
on it's way out now that ->writepage has been removed.
Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Chao Yu <chao@kernel.org> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Since commits 7ff0104a8052 ("f2fs: Remove f2fs_write_node_page()") and 3b47398d9861 ("f2fs: Remove f2fs_write_meta_page()'), f2fs can't be
called from reclaim context any more. Remove all code keyed of the
wbc->for_reclaim flag, which is now only set for writing out swap or
shmem pages inside the swap code, but never passed to file systems.
Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Chao Yu <chao@kernel.org> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Support to inject a timeout fault into function, currently it only
support to inject timeout to commit_atomic_write flow to reproduce
inconsistent bug, like the bug fixed by commit f098aeba04c9 ("f2fs:
fix to avoid atomicity corruption of atomic file").
By default, the new type fault will inject 1000ms timeout, and the
timeout process can be interrupted by SIGKILL.
Signed-off-by: Chao Yu <chao@kernel.org> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
------------[ cut here ]------------
WARNING: CPU: 3 PID: 579 at fs/f2fs/segment.c:2832 new_curseg+0x5e8/0x6dc
pc : new_curseg+0x5e8/0x6dc
Call trace:
new_curseg+0x5e8/0x6dc
f2fs_allocate_data_block+0xa54/0xe28
do_write_page+0x6c/0x194
f2fs_do_write_node_page+0x38/0x78
__write_node_page+0x248/0x6d4
f2fs_sync_node_pages+0x524/0x72c
f2fs_write_checkpoint+0x4bc/0x9b0
__checkpoint_and_complete_reqs+0x80/0x244
issue_checkpoint_thread+0x8c/0xec
kthread+0x114/0x1bc
ret_from_fork+0x10/0x20
get_new_segment() detects inconsistent status in between free_segmap
and free_secmap, let's record such error into super block, and bail
out get_new_segment() instead of continue using the segment.
Signed-off-by: Chao Yu <chao@kernel.org> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Chao Yu [Tue, 6 May 2025 07:47:25 +0000 (15:47 +0800)]
f2fs: sysfs: add encoding_flags entry
This patch adds a new sysfs entry /sys/fs/f2fs/<disk>/encoding_flags,
it is a read-only entry to show the value of sb.s_encoding_flags, the
value is hexadecimal.
f2fs: zone: fix to calculate first_zoned_segno correctly
A zoned device can has both conventional zones and sequential zones,
so we should not treat first segment of zoned device as first_zoned_segno,
instead, we need to check zone type for each zone during traversing zoned
device to find first_zoned_segno.
Otherwise, for below case, first_zoned_segno will be 0, which could be
wrong.
w/ below testcase, resize will generate a corrupted image which
contains inconsistent metadata, so when mounting such image, it
will trigger kernel panic:
sit_i->bitmap_size is 192, so size of sit bitmap is 192*8=1536, at maximum
there are 1536 sit blocks, however MAIN_SEGS is 261893, so that sit_blk_cnt
is 4762, build_sit_entries() -> current_sit_addr() tries to access
out-of-boundary in sit_bitmap at offset from [1536, 4762), once sit_bitmap
and sit_bitmap_mirror is not the same, it will trigger f2fs_bug_on().
Let's add sanity check in f2fs_sanity_check_ckpt() to avoid panic.
Cc: stable@vger.kernel.org Signed-off-by: Chao Yu <chao@kernel.org> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
After 4961acdd65c9 ("f2fs: fix to tag gcing flag on page during block
migration"), we will tag page w/ gcing flag for raw page of cluster
during its migration.
However, if the inode is both encrypted and compressed, during
ioc_decompress(), it will tag page w/ gcing flag, and it increase
F2FS_WB_DATA reference count:
- f2fs_write_multi_page
- f2fs_write_raw_page
- f2fs_write_single_page
- do_write_page
- f2fs_submit_page_write
- WB_DATA_TYPE(bio_page, fio->compressed_page)
: bio_page is encrypted, so mapping is NULL, and fio->compressed_page
is NULL, it returns F2FS_WB_DATA
- inc_page_count(.., F2FS_WB_DATA)
Then, during end_io(), it decrease F2FS_WB_CP_DATA reference count:
- f2fs_write_end_io
- f2fs_compress_write_end_io
- fscrypt_pagecache_folio
: get raw page from encrypted page
- WB_DATA_TYPE(&folio->page, false)
: raw page has gcing flag, it returns F2FS_WB_CP_DATA
- dec_page_count(.., F2FS_WB_CP_DATA)
In order to fix this issue, we need to detect gcing flag in raw page
in f2fs_is_cp_guaranteed().
Fixes: 4961acdd65c9 ("f2fs: fix to tag gcing flag on page during block migration") Reported-by: Jan Prusakowski <jprusakowski@google.com> Signed-off-by: Chao Yu <chao@kernel.org> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
All callers except __get_inode_rdev() and __set_inode_rdev() now have a
folio, but the only callers of those two functions do have a folio, so
pass the folio to them and then into get_dnode_addr().
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org> Reviewed-by: Chao Yu <chao@kernel.org> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
f2fs: Convert dnode_of_data->node_page to node_folio
All assignments to this struct member are conversions from a folio
so convert it to be a folio and convert all users. At the same time,
convert data_blkaddr() to take a folio as all callers now have a folio.
Remove eight calls to compound_head().
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org> Reviewed-by: Chao Yu <chao@kernel.org> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
f2fs: Convert f2fs_load_compressed_page() to f2fs_load_compressed_folio()
The only caller already has a folio, so pass it in. Copy the entire
size of the folio to support large block sizes. Remove two calls to
compound_head().
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org> Reviewed-by: Chao Yu <chao@kernel.org> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Look up a folio instead of a page, and if that fails, allocate a folio.
Removes five calls to compound_head(), one of the last few references to
add_to_page_cache_lru() and honours the cpuset_do_page_mem_spread()
setting.
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org> Reviewed-by: Chao Yu <chao@kernel.org> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Convert all callers to pass in a pointer to a folio instead of a page.
Also convert f2fs_inode_by_name() to take a folio pointer. Removes
six calls to compound_head().
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org> Reviewed-by: Chao Yu <chao@kernel.org> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
The one caller of __f2fs_get_acl() which passes a non-NULL page already
has a folio, so pass it in, then into f2fs_getxattr(), which lets us
pass it to lookup_all_xattrs().
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org> Reviewed-by: Chao Yu <chao@kernel.org> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
The one caller already has folios, so pass them in, and further pass
them to __f2fs_set_acl() and f2fs_acl_create(). There should be
no change to the generated code for this commit.
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org> Reviewed-by: Chao Yu <chao@kernel.org> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
projects / thirdparty / kernel / linux.git / log
