linux/fs/buffer.c, branch v6.16

Merge tag 'vfs-6.16-rc1.writepage' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs

2025-05-26T15:23:09Z

Pull final writepage conversion from Christian Brauner: "This converts vboxfs from ->writepage() to ->writepages(). This was the last user of the ->writepage() method. So remove ->writepage() completely and all references to it" * tag 'vfs-6.16-rc1.writepage' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs: fs: Remove aops->writepage mm: Remove swap_writepage() and shmem_writepage() ttm: Call shmem_writeout() from ttm_backup_backup_page() i915: Use writeback_iter() shmem: Add shmem_writeout() writeback: Remove writeback_use_writepage() migrate: Remove call to ->writepage vboxsf: Convert to writepages 9p: Add a migrate_folio method

fs/buffer: optimize discard_buffer()

2025-05-21T07:34:29Z

While invalidating, the clearing of the bits in discard_buffer() is done in one fully ordered CAS operation. In the past this was done via individual clear_bit(), until e7470ee89f0 (fs: buffer: do not use unnecessary atomic operations when discarding buffers). This implies that there were never strong ordering requirements outside of being serialized by the buffer lock. As such relax the ordering for archs that can benefit. Further, the implied ordering in buffer_unlock() makes current cmpxchg implied barrier redundant due to release semantics. And while in theory the unlock could be part of the bulk clearing, it is best to leave it explicit, but without the double barriers. Signed-off-by: Davidlohr Bueso Link: https://lore.kernel.org/20250515173925.147823-5-dave@stgolabs.net Reviewed-by: Jan Kara Signed-off-by: Christian Brauner

fs/buffer: remove superfluous statements

2025-05-21T07:34:29Z

Get rid of those unnecessary return statements. Signed-off-by: Davidlohr Bueso Link: https://lore.kernel.org/20250515173925.147823-4-dave@stgolabs.net Reviewed-by: Jan Kara Signed-off-by: Christian Brauner

fs/buffer: avoid redundant lookup in getblk slowpath

2025-05-21T07:34:29Z

__getblk_slow() already implies failing a first lookup as the fastpath, so try to create the buffers immediately and avoid the redundant lookup. This saves 5-10% of the total cost/latency of the slowpath. Signed-off-by: Davidlohr Bueso Link: https://lore.kernel.org/20250515173925.147823-3-dave@stgolabs.net Reviewed-by: Jan Kara Signed-off-by: Christian Brauner

fs/buffer: use sleeping lookup in __getblk_slowpath()

2025-05-21T07:34:28Z

Just as with the fast path, call the lookup variant depending on the gfp flags. Signed-off-by: Davidlohr Bueso Link: https://lore.kernel.org/20250515173925.147823-2-dave@stgolabs.net Reviewed-by: Jan Kara Signed-off-by: Christian Brauner

fs: Remove redundant errseq_set call in mark_buffer_write_io_error.

2025-05-09T10:31:57Z

mark_buffer_write_io_error sets sb->s_wb_err to -EIO twice. Once in mapping_set_error and once in errseq_set. Only mapping_set_error checks if bh->b_assoc_map->host is NULL. Discovered during null pointer dereference during writeback to a failing device: [] ? mark_buffer_write_io_error+0x98/0xc0 [] ? mark_buffer_write_io_error+0x8e/0xc0 [] end_buffer_async_write+0x90/0xd0 [] end_bio_bh_io_sync+0x2b/0x40 [] blk_update_request+0x1b6/0x480 [] blk_mq_end_request+0x18/0x30 [] blk_mq_dispatch_rq_list+0x4da/0x8e0 [] __blk_mq_sched_dispatch_requests+0x218/0x6a0 [] blk_mq_sched_dispatch_requests+0x3a/0x80 [] blk_mq_run_hw_queue+0x108/0x330 [] blk_mq_flush_plug_list+0x178/0x5f0 [] __blk_flush_plug+0x41/0x120 [] blk_finish_plug+0x22/0x40 [] wb_writeback+0x150/0x280 [] ? set_worker_desc+0x9f/0xc0 [] wb_workfn+0x24e/0x4a0 Fixes: 485e9605c0573 ("fs/buffer.c: record blockdev write errors in super_block that it backs") Signed-off-by: Jeremy Bongio Link: https://lore.kernel.org/20250507123010.1228243-1-jbongio@google.com Reviewed-by: Christoph Hellwig Signed-off-by: Christian Brauner

mm/migrate: fix sleep in atomic for large folios and buffer heads

2025-04-22T16:16:08Z

The large folio + buffer head noref migration scenarios are being naughty and blocking while holding a spinlock. As a consequence of the pagecache lookup path taking the folio lock this serializes against migration paths, so they can wait for each other. For the private_lock atomic case, a new BH_Migrate flag is introduced which enables the lookup to bail. This allows the critical region of the private_lock on the migration path to be reduced to the way it was before ebdf4de5642fb6 ("mm: migrate: fix reference check race between __find_get_block() and migration"), that is covering the count checks. The scope is always noref migration. Reported-by: kernel test robot Reported-by: syzbot+f3c6fda1297c748a7076@syzkaller.appspotmail.com Closes: https://lore.kernel.org/oe-lkp/202503101536.27099c77-lkp@intel.com Fixes: 3c20917120ce61 ("block/bdev: enable large folio support for large logical block sizes") Reviewed-by: Jan Kara Co-developed-by: Luis Chamberlain Signed-off-by: Davidlohr Bueso Link: https://kdevops.org/ext4/v6.15-rc2.html # [0] Link: https://lore.kernel.org/all/aAAEvcrmREWa1SKF@bombadil.infradead.org/ # [1] Link: https://lore.kernel.org/20250418015921.132400-8-dave@stgolabs.net Tested-by: kdevops@lists.linux.dev # [0] [1] Reviewed-by: Luis Chamberlain Signed-off-by: Christian Brauner

fs/buffer: use sleeping version of __find_get_block()

2025-04-22T16:16:08Z

Convert to the new nonatomic flavor to benefit from potential performance benefits and adapt in the future vs migration such that semantics are kept. Convert write_boundary_block() which already takes the buffer lock as well as bdev_getblk() depending on the respective gpf flags. There are no changes in semantics. Suggested-by: Jan Kara Reviewed-by: Jan Kara Signed-off-by: Davidlohr Bueso Link: https://kdevops.org/ext4/v6.15-rc2.html # [0] Link: https://lore.kernel.org/all/aAAEvcrmREWa1SKF@bombadil.infradead.org/ # [1] Link: https://lore.kernel.org/20250418015921.132400-4-dave@stgolabs.net Tested-by: kdevops@lists.linux.dev # [0] [1] Reviewed-by: Luis Chamberlain Signed-off-by: Christian Brauner

fs/buffer: introduce sleeping flavors for pagecache lookups

2025-04-22T16:16:08Z

Add __find_get_block_nonatomic() and sb_find_get_block_nonatomic() calls for which users will be converted where safe. These versions will take the folio lock instead of the mapping's private_lock. Reviewed-by: Jan Kara Signed-off-by: Davidlohr Bueso Link: https://kdevops.org/ext4/v6.15-rc2.html # [0] Link: https://lore.kernel.org/all/aAAEvcrmREWa1SKF@bombadil.infradead.org/ # [1] Link: https://lore.kernel.org/20250418015921.132400-3-dave@stgolabs.net Tested-by: kdevops@lists.linux.dev Reviewed-by: Luis Chamberlain Signed-off-by: Christian Brauner

fs/buffer: split locking for pagecache lookups

2025-04-22T16:16:07Z

Callers of __find_get_block() may or may not allow for blocking semantics, and is currently assumed that it will not. Layout two paths based on this. The the private_lock scheme will continued to be used for atomic contexts. Otherwise take the folio lock instead, which protects the buffers, such as vs migration and try_to_free_buffers(). Per the "hack idea", the latter can alleviate contention on the private_lock for bdev mappings. For reasons of determinism and avoid making bugs hard to reproduce, the trylocking is not attempted. No change in semantics. All lookup users still take the spinlock. Reviewed-by: Jan Kara Signed-off-by: Davidlohr Bueso Link: https://kdevops.org/ext4/v6.15-rc2.html # [0] Link: https://lore.kernel.org/all/aAAEvcrmREWa1SKF@bombadil.infradead.org/ # [1] Link: https://lore.kernel.org/20250418015921.132400-2-dave@stgolabs.net Tested-by: kdevops@lists.linux.dev Reviewed-by: Luis Chamberlain Signed-off-by: Christian Brauner