- 13 Apr, 2023 9 commits
-
-
Chao Yu authored
In __replace_atomic_write_block(), we missed to check return value of inc_valid_block_count(), for extreme testcase that f2fs image is run out of space, it may cause inconsistent status in between SIT table and total valid block count. Cc: Daeho Jeong <daehojeong@google.com> Fixes: 3db1de0e ("f2fs: change the current atomic write way") Signed-off-by:
Chao Yu <chao@kernel.org> Signed-off-by:
Jaegeuk Kim <jaegeuk@kernel.org>
-
Chao Yu authored
Otherwise, if truncation on cow_inode failed, remained data may pollute current transaction of atomic write. Cc: Daeho Jeong <daehojeong@google.com> Fixes: a46bebd5 ("f2fs: synchronize atomic write aborts") Signed-off-by:
-
Daeho Jeong authored
We should not pass relative address in a zone to __f2fs_issue_discard_zone(). Signed-off-by:
Daeho Jeong <daehojeong@google.com> Reviewed-by:
-
Jaegeuk Kim authored
F2FS has the same issue in ext4_rename causing crash revealed by xfstests/generic/707. See also commit 0813299c ("ext4: Fix possible corruption when moving a directory") CC: stable@vger.kernel.org Signed-off-by:
-
Yohan Joung authored
To prevent excessive increase in preemption count add radix_tree_preload_end in retry Signed-off-by:
Yohan Joung <yohan.joung@sk.com> Signed-off-by:
-
Chao Yu authored
With -O quota mkfs option, xfstests generic/417 fails due to fsck detects data corruption on quota inodes. [ASSERT] (fsck_chk_quota_files:2051) --> Quota file is missing or invalid quota file content found. The root cause is there is a hole f2fs doesn't hold quota inodes, so all recovered quota data will be dropped due to SBI_POR_DOING flag was set. - f2fs_fill_super - f2fs_recover_orphan_inodes - f2fs_enable_quota_files - f2fs_quota_off_umount <--- quota inodes were dropped ---> - f2fs_recover_fsync_data - f2fs_enable_quota_files - f2fs_quota_off_umount This patch tries to eliminate the hole by holding quota inodes during entire recovery flow as below: - f2fs_fill_super - f2fs_recover_quota_begin - f2fs_recover_orphan_inodes - f2fs_recover_fsync_data - f2fs_recover_quota_end Then, recovered quota data can be persisted after SBI_POR_DOING is cleared. Signed-off-by:
-
Chao Yu authored
With below case, it can mount multi-device image w/ rw option, however one of secondary device is set as ro, later update will cause panic, so let's introduce f2fs_dev_is_readonly(), and check multi-devices rw status in f2fs_remount() w/ it in order to avoid such inconsistent mount status. mkfs.f2fs -c /dev/zram1 /dev/zram0 -f blockdev --setro /dev/zram1 mount -t f2fs dev/zram0 /mnt/f2fs mount: /mnt/f2fs: WARNING: source write-protected, mounted read-only. mount -t f2fs -o remount,rw mnt/f2fs dd if=/dev/zero of=/mnt/f2fs/file bs=1M count=8192 kernel BUG at fs/f2fs/inline.c:258! RIP: 0010:f2fs_write_inline_data+0x23e/0x2d0 [f2fs] Call Trace: f2fs_write_single_data_page+0x26b/0x9f0 [f2fs] f2fs_write_cache_pages+0x389/0xa60 [f2fs] __f2fs_write_data_pages+0x26b/0x2d0 [f2fs] f2fs_write_data_pages+0x2e/0x40 [f2fs] do_writepages+0xd3/0x1b0 __writeback_single_inode+0x5b/0x420 writeback_sb_inodes+0x236/0x5a0 __writeback_inodes_wb+0x56/0xf0 wb_writeback+0x2a3/0x490 wb_do_writeback+0x2b2/0x330 wb_workfn+0x6a/0x260 process_one_work+0x270/0x5e0 worker_thread+0x52/0x3e0 kthread+0xf4/0x120 ret_from_fork+0x29/0x50 Signed-off-by:
-
Wang Han authored
This should be 'disable' rather than 'disabled'. Reported-by:
LoveSy <shana@zju.edu.cn> Signed-off-by:
Wang Han <wanghan1995315@gmail.com> Signed-off-by:
-
Chao Yu authored
i_gc_rwsem[WRITE] and i_gc_rwsem[READ] lock order is reversed in gc_data_segment() and f2fs_dio_write_iter(), fix to keep consistent lock order as below: 1. lock i_gc_rwsem[WRITE] 2. lock i_gc_rwsem[READ] Signed-off-by:
-
- 10 Apr, 2023 10 commits
-
-
Chao Yu authored
xfstest generic/361 reports a bug as below: f2fs_bug_on(sbi, sbi->fsync_node_num); kernel BUG at fs/f2fs/super.c:1627! RIP: 0010:f2fs_put_super+0x3a8/0x3b0 Call Trace: generic_shutdown_super+0x8c/0x1b0 kill_block_super+0x2b/0x60 kill_f2fs_super+0x87/0x110 deactivate_locked_super+0x39/0x80 deactivate_super+0x46/0x50 cleanup_mnt+0x109/0x170 __cleanup_mnt+0x16/0x20 task_work_run+0x65/0xa0 exit_to_user_mode_prepare+0x175/0x190 syscall_exit_to_user_mode+0x25/0x50 do_syscall_64+0x4c/0x90 entry_SYSCALL_64_after_hwframe+0x72/0xdc During umount(), if cp_error is set, f2fs_wait_on_all_pages() should not stop waiting all F2FS_WB_CP_DATA pages to be writebacked, otherwise, fsync_node_num can be non-zero after f2fs_wait_on_all_pages() causing this bug. In this case, to avoid deadloop in f2fs_wait_on_all_pages(), it needs to drop all dirty pages rather than redirtying them. Signed-off-by:
-
Chao Yu authored
xfstest generic/019 reports a bug: kernel BUG at mm/filemap.c:1619! RIP: 0010:folio_end_writeback+0x8a/0x90 Call Trace: end_page_writeback+0x1c/0x60 f2fs_write_end_io+0x199/0x420 bio_endio+0x104/0x180 submit_bio_noacct+0xa5/0x510 submit_bio+0x48/0x80 f2fs_submit_write_bio+0x35/0x300 f2fs_submit_merged_ipu_write+0x2a0/0x2b0 f2fs_write_single_data_page+0x838/0x8b0 f2fs_write_cache_pages+0x379/0xa30 f2fs_write_data_pages+0x30c/0x340 do_writepages+0xd8/0x1b0 __writeback_single_inode+0x44/0x370 writeback_sb_inodes+0x233/0x4d0 __writeback_inodes_wb+0x56/0xf0 wb_writeback+0x1dd/0x2d0 wb_workfn+0x367/0x4a0 process_one_work+0x21d/0x430 worker_thread+0x4e/0x3c0 kthread+0x103/0x130 ret_from_fork+0x2c/0x50 The root cause is: after cp_error is set, f2fs_submit_merged_ipu_write() in f2fs_write_single_data_page() tries to flush IPU bio in cache, however f2fs_submit_merged_ipu_write() missed to check validity of @bio parameter, result in submitting random cached bio which belong to other IO context, then it will cause use-after-free issue, fix it by adding additional validity check. Fixes: 0b20fcec ("f2fs: cache global IPU bio") Signed-off-by:
-
Chao Yu authored
i_crtime will never change after inode creation, so we don't need to copy it into f2fs_inode_info.i_disk_time[3], and monitor its change to decide whether updating inode page, remove related stuff. Signed-off-by:
-
Chao Yu authored
f2fs has supported multi-device feature, to check devices' rw status, it should use f2fs_hw_is_readonly() rather than bdev_read_only(), fix it. Meanwhile, it removes f2fs_hw_is_readonly() check condition in: - f2fs_write_checkpoint() - f2fs_convert_inline_inode() As it has checked f2fs_readonly() condition, and if f2fs' devices were readonly, f2fs_readonly() must be true. Signed-off-by:
-
Weizhao Ouyang authored
Use common implementation of file type conversion helpers. Signed-off-by:
Weizhao Ouyang <o451686892@gmail.com> Reviewed-by:
-
Yangtao Li authored
Remove unnecessary lz4hc_compress_pages(). Signed-off-by:
Yangtao Li <frank.li@vivo.com> Reviewed-by:
-
Yangtao Li authored
Let's use sysfs_emit. Signed-off-by:
-
Yangtao Li authored
If the compress feature is not enabled, there is no need to set compress-related parameters. Signed-off-by:
-
Yonggil Song authored
When f2fs tries to checkpoint during foreground gc in LFS mode, system crash occurs due to lack of free space if the amount of dirty node and dentry pages generated by data migration exceeds free space. The reproduction sequence is as follows. - 20GiB capacity block device (null_blk) - format and mount with LFS mode - create a file and write 20,000MiB - 4k random write on full range of the file RIP: 0010:new_curseg+0x48a/0x510 [f2fs] Code: 55 e7 f5 89 c0 48 0f af c3 48 8b 5d c0 48 c1 e8 20 83 c0 01 89 43 6c 48 83 c4 28 5b 41 5c 41 5d 41 5e 41 5f 5d c3 cc cc cc cc <0f> 0b f0 41 80 4f 48 04 45 85 f6 0f 84 ba fd ff ff e9 ef fe ff ff RSP: 0018:ffff977bc397b218 EFLAGS: 00010246 RAX: 00000000000027b9 RBX: 0000000000000000 RCX: 00000000000027c0 RDX: 0000000000000000 RSI: 00000000000027b9 RDI: ffff8c25ab4e74f8 RBP: ffff977bc397b268 R08: 00000000000027b9 R09: ffff8c29e4a34b40 R10: 0000000000000001 R11: ffff977bc397b0d8 R12: 0000000000000000 R13: ffff8c25b4dd81a0 R14: 0000000000000000 R15: ffff8c2f667f9000 FS: 0000000000000000(0000) GS:ffff8c344ec80000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 000000c00055d000 CR3: 0000000e30810003 CR4: 00000000003706e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 Call Trace: <TASK> allocate_segment_by_default+0x9c/0x110 [f2fs] f2fs_allocate_data_block+0x243/0xa30 [f2fs] ? __mod_lruvec_page_state+0xa0/0x150 do_write_page+0x80/0x160 [f2fs] f2fs_do_write_node_page+0x32/0x50 [f2fs] __write_node_page+0x339/0x730 [f2fs] f2fs_sync_node_pages+0x5a6/0x780 [f2fs] block_operations+0x257/0x340 [f2fs] f2fs_write_checkpoint+0x102/0x1050 [f2fs] f2fs_gc+0x27c/0x630 [f2fs] ? folio_mark_dirty+0x36/0x70 f2fs_balance_fs+0x16f/0x180 [f2fs] This patch adds checking whether free sections are enough before checkpoint during gc. Signed-off-by:
Yonggil Song <yonggil.song@samsung.com> [Jaegeuk Kim: code clean-up] Reviewed-by:
-
Yangtao Li authored
There is only single instance of these ops, and Jaegeuk point out that: Originally this was intended to give a chance to provide other allocation option. Anyway, it seems quit hard to do it anymore. So remove the indirection and call f2fs_get_victim() directly. Signed-off-by:
-
- 04 Apr, 2023 5 commits
-
-
Jaegeuk Kim authored
We got a kernel panic if old_addr is NULL. https://bugzilla.kernel.org/show_bug.cgi?id=217266 BUG: kernel NULL pointer dereference, address: 0000000000000000 Call Trace: <TASK> f2fs_commit_atomic_write+0x619/0x990 [f2fs a1b985b80f5babd6f3ea778384908880812bfa43] __f2fs_ioctl+0xd8e/0x4080 [f2fs a1b985b80f5babd6f3ea778384908880812bfa43] ? vfs_write+0x2ae/0x3f0 ? vfs_write+0x2ae/0x3f0 __x64_sys_ioctl+0x91/0xd0 do_syscall_64+0x5c/0x90 entry_SYSCALL_64_after_hwframe+0x72/0xdc RIP: 0033:0x7f69095fe53f Fixes: 2f3a9ae9 ("f2fs: introduce trace_f2fs_replace_atomic_write_block") Cc: <stable@vger.kernel.org> Reviewed-by:
-
Qilin Tan authored
Made iostat lock irq safe to avoid potentinal deadlock. Deadlock scenario: f2fs_attr_store -> f2fs_sbi_store -> _sbi_store -> spin_lock(sbi->iostat_lock) <interrupt request> -> scsi_end_request -> bio_endio -> f2fs_dio_read_end_io -> f2fs_update_iostat -> spin_lock_irqsave(sbi->iostat_lock) ===> Dead lock here Fixes: 61803e98 ("f2fs: fix iostat related lock protection") Fixes: a1e09b03 ("f2fs: use iomap for direct I/O") Signed-off-by:Qilin Tan <qilin.tan@mediatek.com> Reviewed-by:
-
Yohan Joung authored
Fix alignment check to be correct in npo2 as well Signed-off-by:
-
Yangtao Li authored
Opt_compress_chksum, Opt_compress_mode and Opt_compress_cache lack the necessary check to see if the image supports compression, let's add it. Signed-off-by:
-
Yangtao Li authored
is_extension_exist() only return two values, 0 or 1. So there is no need to use int type. Signed-off-by:
-
- 29 Mar, 2023 16 commits
-
-
Jaegeuk Kim authored
[ 16.945668][ C0] Call trace: [ 16.945678][ C0] dump_backtrace+0x110/0x204 [ 16.945706][ C0] dump_stack_lvl+0x84/0xbc [ 16.945735][ C0] __schedule_bug+0xb8/0x1ac [ 16.945756][ C0] __schedule+0x724/0xbdc [ 16.945778][ C0] schedule+0x154/0x258 [ 16.945793][ C0] bit_wait_io+0x48/0xa4 [ 16.945808][ C0] out_of_line_wait_on_bit+0x114/0x198 [ 16.945824][ C0] __sync_dirty_buffer+0x1f8/0x2e8 [ 16.945853][ C0] __f2fs_commit_super+0x140/0x1f4 [ 16.945881][ C0] f2fs_commit_super+0x110/0x28c [ 16.945898][ C0] f2fs_handle_error+0x1f4/0x2f4 [ 16.945917][ C0] f2fs_decompress_cluster+0xc4/0x450 [ 16.945942][ C0] f2fs_end_read_compressed_page+0xc0/0xfc [ 16.945959][ C0] f2fs_handle_step_decompress+0x118/0x1cc [ 16.945978][ C0] f2fs_read_end_io+0x168/0x2b0 [ 16.945993][ C0] bio_endio+0x25c/0x2c8 [ 16.946015][ C0] dm_io_dec_pending+0x3e8/0x57c [ 16.946052][ C0] clone_endio+0x134/0x254 [ 16.946069][ C0] bio_endio+0x25c/0x2c8 [ 16.946084][ C0] blk_update_request+0x1d4/0x478 [ 16.946103][ C0] scsi_end_request+0x38/0x4cc [ 16.946129][ C0] scsi_io_completion+0x94/0x184 [ 16.946147][ C0] scsi_finish_command+0xe8/0x154 [ 16.946164][ C0] scsi_complete+0x90/0x1d8 [ 16.946181][ C0] blk_done_softirq+0xa4/0x11c [ 16.946198][ C0] _stext+0x184/0x614 [ 16.946214][ C0] __irq_exit_rcu+0x78/0x144 [ 16.946234][ C0] handle_domain_irq+0xd4/0x154 [ 16.946260][ C0] gic_handle_irq.33881+0x5c/0x27c [ 16.946281][ C0] call_on_irq_stack+0x40/0x70 [ 16.946298][ C0] do_interrupt_handler+0x48/0xa4 [ 16.946313][ C0] el1_interrupt+0x38/0x68 [ 16.946346][ C0] el1h_64_irq_handler+0x20/0x30 [ 16.946362][ C0] el1h_64_irq+0x78/0x7c [ 16.946377][ C0] finish_task_switch+0xc8/0x3d8 [ 16.946394][ C0] __schedule+0x600/0xbdc [ 16.946408][ C0] preempt_schedule_common+0x34/0x5c [ 16.946423][ C0] preempt_schedule+0x44/0x48 [ 16.946438][ C0] process_one_work+0x30c/0x550 [ 16.946456][ C0] worker_thread+0x414/0x8bc [ 16.946472][ C0] kthread+0x16c/0x1e0 [ 16.946486][ C0] ret_from_fork+0x10/0x20 Fixes: bff139b4 ("f2fs: handle decompress only post processing in softirq") Fixes: 95fa90c9 ("f2fs: support recording errors into superblock") Reviewed-by:
-
Hans Holmberg authored
In some cases, e.g. for zoned block devices, direct writes are forced into buffered writes that will populate the page cache and be written out just like buffered io. Direct reads, on the other hand, is supported for the zoned block device case. This has the effect that applications built for direct io will fill up the page cache with data that will never be read, and that is a waste of resources. If we agree that this is a problem, how do we fix it? A) Supporting proper direct writes for zoned block devices would be the best, but it is currently not supported (probably for a good but non-obvious reason). Would it be feasible to implement proper direct IO? B) Avoid the cost of keeping unwanted data by syncing and throwing out the cached pages for buffered O_DIRECT writes before completion. This patch implements B) by reusing the code for how partial block writes are flushed out on the "normal" direct write path. Note that this changes the performance characteristics of f2fs quite a bit. Direct IO performance for zoned block devices is lower for small writes after this patch, but this should be expected with direct IO and in line with how f2fs behaves on top of conventional block devices. Another open question is if the flushing should be done for all cases where buffered writes are forced. Signed-off-by:
Hans Holmberg <hans.holmberg@wdc.com> Reviewed-by:
-
Yangtao Li authored
BUG_ON() will be triggered when writing files concurrently, because the same page is writtenback multiple times. 1597 void folio_end_writeback(struct folio *folio) 1598 { ...... 1618 if (!__folio_end_writeback(folio)) 1619 BUG(); ...... 1625 } kernel BUG at mm/filemap.c:1619! Call Trace: <TASK> f2fs_write_end_io+0x1a0/0x370 blk_update_request+0x6c/0x410 blk_mq_end_request+0x15/0x130 blk_complete_reqs+0x3c/0x50 __do_softirq+0xb8/0x29b ? sort_range+0x20/0x20 run_ksoftirqd+0x19/0x20 smpboot_thread_fn+0x10b/0x1d0 kthread+0xde/0x110 ? kthread_complete_and_exit+0x20/0x20 ret_from_fork+0x22/0x30 </TASK> Below is the concurrency scenario: [Process A] [Process B] [Process C] f2fs_write_raw_pages() - redirty_page_for_writepage() - unlock page() f2fs_do_write_data_page() - lock_page() - clear_page_dirty_for_io() - set_page_writeback() [1st writeback] ..... - unlock page() generic_perform_write() - f2fs_write_begin() - wait_for_stable_page() - f2fs_write_end() - set_page_dirty() - lock_page() - f2fs_do_write_data_page() - set_page_writeback() [2st writeback] This problem was introduced by the previous commit 7377e853 ("f2fs: compress: fix potential deadlock of compress file"). All pagelocks were released in f2fs_write_raw_pages(), but whether the page was in the writeback state was ignored in the subsequent writing process. Let's fix it by waiting for the page to writeback before writing. Cc: Christoph Hellwig <hch@lst.de> Fixes: 4c8ff709 ("f2fs: support data compression") Fixes: 7377e853 ("f2fs: compress: fix potential deadlock of compress file") Signed-off-by:Qi Han <hanqi@vivo.com> Signed-off-by:
-
Yangtao Li authored
As Christoph Hellwig point out: Please avoid the else by doing the goto in the branch. Signed-off-by:
-
Jaegeuk Kim authored
If we manage the zone capacity per zone type, it'll break the GC assumption. And, the current logic complains valid block count mismatch. Let's apply zone capacity to all zone type, if specified. Fixes: de881df9 ("f2fs: support zone capacity less than zone size") Reviewed-by:
-
Yangtao Li authored
It seems inappropriate that the current logic does not handle filemap_fdatawrite() errors, so let's fix it. Signed-off-by:
-
Yangtao Li authored
BIW reduce the s_flag array size and make s_flag constant. Signed-off-by:
-
Yonggil Song authored
When using f2fs on a zoned block device with 2MiB zone size, IO errors occurs because f2fs tries to write data to a zone that has not been reset. The cause is that f2fs tries to discard multiple zones at once. This is caused by a condition in f2fs_clear_prefree_segments that does not check for zoned block devices when setting the discard range. This leads to invalid reset commands and write pointer mismatches. This patch fixes the zoned block device with 2MiB zone size to reset one zone at a time. Signed-off-by:
-
Jaegeuk Kim authored
This is a last part to remove the memory sharing for rb_tree in extent_cache. This should also fix arm32 memory alignment issue. [struct extent_node] [struct rb_entry] [0] struct rb_node rb_node; [0] struct rb_node rb_node; union { union { struct { struct { [16] unsigned int fofs; [12] unsigned int ofs; unsigned int len; unsigned int len; }; unsigned long long key; } __packed; Cc: <stable@vger.kernel.org> Fixes: 13054c54 ("f2fs: introduce infra macro and data structure of rb-tree extent cache") Reviewed-by: -
Jaegeuk Kim authored
This is a second part to remove the mixed use of rb_tree in discard_cmd from extent_cache. This should also fix arm32 memory alignment issue caused by shared rb_entry. [struct discard_cmd] [struct rb_entry] [0] struct rb_node rb_node; [0] struct rb_node rb_node; union { union { struct { struct { [16] block_t lstart; [12] unsigned int ofs; block_t len; unsigned int len; }; unsigned long long key; } __packed; Cc: <stable@vger.kernel.org> Fixes: 004b6862 ("f2fs: use rb-tree to track pending discard commands") Signed-off-by: -
Jaegeuk Kim authored
Let's reduce the complexity of mixed use of rb_tree in victim_entry from extent_cache and discard_cmd. This should fix arm32 memory alignment issue caused by shared rb_entry. [struct victim_entry] [struct rb_entry] [0] struct rb_node rb_node; [0] struct rb_node rb_node; union { struct { unsigned int ofs; unsigned int len; }; [16] unsigned long long mtime; [12] unsigned long long key; } __packed; Cc: <stable@vger.kernel.org> Fixes: 093749e2 ("f2fs: support age threshold based garbage collection") Reviewed-by: -
Yonggil Song authored
When f2fs skipped a gc round during victim migration, there was a bug which would skip all upcoming gc rounds unconditionally because skipped_gc_rwsem was not initialized. It fixes the bug by correctly initializing the skipped_gc_rwsem inside the gc loop. Fixes: 6f8d4455 ("f2fs: avoid fi->i_gc_rwsem[WRITE] lock in f2fs_gc") Signed-off-by:
-
Yangtao Li authored
We should set the error code when dqget() failed. Fixes: 2c1d0305 ("f2fs: support F2FS_IOC_FS{GET,SET}XATTR") Signed-off-by:
-
Yangtao Li authored
Let's use BIT() and GENMASK() instead of open it. Signed-off-by:
-
Yangtao Li authored
This patch export below sysfs entries for better control cached compress page count. /sys/fs/f2fs/<disk>/compress_watermark /sys/fs/f2fs/<disk>/compress_percent Signed-off-by:
-
Li Zetao authored
After commit 26b5a079 ("f2fs: cleanup dirty pages if recover failed"), f2fs_sync_inode_meta() is only used in checkpoint.c, so f2fs_sync_inode_meta() should only be visible inside. Delete the declaration in the header file and change f2fs_sync_inode_meta() to static. Signed-off-by:
Li Zetao <lizetao1@huawei.com> Reviewed-by:
-
