diff options
| author | Christian Brauner <brauner@kernel.org> | 2026-09-04 12:54:52 +0200 |
|---|---|---|
| committer | Christian Brauner <brauner@kernel.org> | 2026-09-10 09:27:29 +0200 |
| commit | 7c5bb35718a6ecc2001e2f8f6abc068ab926b583 (patch) | |
| tree | 73b38691bf46dcd45f3746630ba71535124811ea | |
| parent | 00db7c70bf7c6a79a548f484c76af60563a7736e (diff) | |
| parent | a21153de5aa936900aa98fd81643895c7fd2a267 (diff) | |
| download | linux-next-7c5bb35718a6ecc2001e2f8f6abc068ab926b583.tar.gz linux-next-7c5bb35718a6ecc2001e2f8f6abc068ab926b583.zip | |
Merge patch series "Buffer head documentation updates"
Matthew Wilcox (Oracle) <willy@infradead.org> says:
Mostly just tidying up some out of date documentation, but I found the
time to document some functions I thought were worth documenting.
* patches from https://patch.msgid.link/20260828205429.3678204-1-willy@infradead.org:
buffer: Rename label
buffer: Refer to folios in comments
buffer: Document mmb_mark_buffer_dirty()
buffer: Document sb_bread(), sb_bread_unmovable() and sb_breadahead()
fs: Remove references to mark_buffer_dirty_inode()
buffer: Remove reference to unmap_buffer()
Link: https://patch.msgid.link/20260828205429.3678204-1-willy@infradead.org
Signed-off-by: Christian Brauner (Amutable) <brauner@kernel.org>
| -rw-r--r-- | Documentation/filesystems/porting.rst | 2 | ||||
| -rw-r--r-- | fs/buffer.c | 91 | ||||
| -rw-r--r-- | include/linux/buffer_head.h | 50 |
3 files changed, 93 insertions, 50 deletions
diff --git a/Documentation/filesystems/porting.rst b/Documentation/filesystems/porting.rst index 60880eb0c49d..de96b6745600 100644 --- a/Documentation/filesystems/porting.rst +++ b/Documentation/filesystems/porting.rst @@ -348,7 +348,7 @@ simply of return 1. Note that all actual eviction work is done by caller after As before, clear_inode() must be called exactly once on each call of ->evict_inode() (as it used to be for each call of ->delete_inode()). Unlike before, if you are using inode-associated metadata buffers (i.e. -mark_buffer_dirty_inode()), it's your responsibility to call +mmb_mark_buffer_dirty()), it's your responsibility to call invalidate_inode_buffers() before clear_inode(). NOTE: checking i_nlink in the beginning of ->write_inode() and bailing out diff --git a/fs/buffer.c b/fs/buffer.c index a43f3c9ecaa0..1dc933ba6925 100644 --- a/fs/buffer.c +++ b/fs/buffer.c @@ -264,7 +264,7 @@ __find_get_block_slow(struct block_device *bdev, sector_t block, bool atomic) bh = bh->b_this_page; } while (bh != head); - /* we might be here because some of the buffers on this page are + /* we might be here because some of the buffers on this folio are * not mapped. This is due to various races between * file io on the block device and getblk. It gets dealt with * elsewhere, don't buffer_error if we had some unmapped buffers @@ -310,7 +310,7 @@ static void end_buffer_async_read(struct buffer_head *bh, int uptodate) /* * Be _very_ careful from here on. Bad things can happen if * two buffer heads end IO at almost the same time and both - * decide that the page is now completely done. + * decide that the folio is now completely done. */ first = folio_buffers(folio); spin_lock_irqsave(&first->b_uptodate_lock, flags); @@ -518,8 +518,8 @@ EXPORT_SYMBOL_GPL(mmb_has_buffers); * * Do this in two main stages: first we copy dirty buffers to a * temporary inode list, queueing the writes as we go. Then we clean - * up, waiting for those writes to complete. mark_buffer_dirty_inode() - * doesn't touch b_assoc_buffers list if b_mmb is not NULL so we are sure the + * up, waiting for those writes to complete. mmb_mark_buffer_dirty() + * doesn't touch b_assoc_buffers list if b_mmb is set so we are sure the * buffer stays on our list until IO completes (at which point it can be * reaped). */ @@ -540,7 +540,7 @@ int mmb_sync(struct mapping_metadata_bhs *mmb) bh = BH_ENTRY(mmb->list.next); WARN_ON_ONCE(bh->b_mmb != mmb); __remove_assoc_queue(mmb, bh); - /* Avoid race with mark_buffer_dirty_inode() which does + /* Avoid race with mmb_mark_buffer_dirty() which does * a lockless check and we rely on seeing the dirty bit */ smp_mb(); if (buffer_dirty(bh) || buffer_locked(bh)) { @@ -578,7 +578,7 @@ int mmb_sync(struct mapping_metadata_bhs *mmb) bh = BH_ENTRY(tmp.prev); get_bh(bh); __remove_assoc_queue(mmb, bh); - /* Avoid race with mark_buffer_dirty_inode() which does + /* Avoid race with mmb_mark_buffer_dirty() which does * a lockless check and we rely on seeing the dirty bit */ smp_mb(); if (buffer_dirty(bh)) { @@ -616,6 +616,14 @@ void write_boundary_block(struct block_device *bdev, } } +/** + * mmb_mark_buffer_dirty - Mark a metadata buffer dirty. + * @bh: The buffer to mark dirty. + * @mmb: The list of buffers to add the buffer to. + * + * Mark the buffer dirty and add it to the list if it is not already on + * a list. + */ void mmb_mark_buffer_dirty(struct buffer_head *bh, struct mapping_metadata_bhs *mmb) { @@ -684,7 +692,7 @@ bool block_dirty_folio(struct address_space *mapping, struct folio *folio) } while (bh != head); } /* - * Lock out page's memcg migration to keep PageDirty + * Lock out folio's memcg migration to keep folio dirty flag * synchronized with per-memcg dirty page counters. */ newly_dirty = !folio_test_set_dirty(folio); @@ -950,23 +958,23 @@ __getblk_slow(struct block_device *bdev, sector_t block, } /* - * The relationship between dirty buffers and dirty pages: + * The relationship between dirty buffers and dirty folios: * - * Whenever a page has any dirty buffers, the page's dirty bit is set, and - * the page is tagged dirty in the page cache. + * Whenever a folio has any dirty buffers, the folio's dirty flag is set, and + * the folio is tagged dirty in the page cache. * * At all times, the dirtiness of the buffers represents the dirtiness of - * subsections of the page. If the page has buffers, the page dirty bit is + * subsections of the folio. If the folio has buffers, the folio dirty flag is * merely a hint about the true dirty state. * - * When a page is set dirty in its entirety, all its buffers are marked dirty - * (if the page has buffers). + * When a folio is set dirty in its entirety, all its buffers are marked dirty + * (if the folio has buffers). * - * When a buffer is marked dirty, its page is dirtied, but the page's other + * When a buffer is marked dirty, its folio is dirtied, but the folio's other * buffers are not. * * Also. When blockdev buffers are explicitly read with bread(), they - * individually become uptodate. But their backing page remains not + * individually become uptodate. But their backing folio remains not * uptodate - even if all of its buffers are uptodate. A subsequent * block_read_full_folio() against that folio will discover all the uptodate * buffers, will set the folio uptodate and will perform no I/O. @@ -977,7 +985,7 @@ __getblk_slow(struct block_device *bdev, sector_t block, * @bh: the buffer_head to mark dirty * * mark_buffer_dirty() will set the dirty bit against the buffer, then set - * its backing page dirty, then tag the page as dirty in the page cache + * its backing folio dirty, then tag the folio as dirty in the page cache * and then attach the address_space's inode to its superblock's dirty * inode list. * @@ -1220,7 +1228,7 @@ static void bh_lru_install(struct buffer_head *bh) /* * the refcount of buffer_head in bh_lru prevents dropping the - * attached page(i.e., try_to_free_buffers) so it could cause + * attached folio (i.e., try_to_free_buffers) so it could cause * failing page migration. * Skip putting upcoming bh into bh_lru until migration is done. */ @@ -1284,7 +1292,7 @@ lookup_bh_lru(struct block_device *bdev, sector_t block, unsigned size) * Perform a pagecache lookup for the matching buffer. If it's there, refresh * it in the LRU and mark it as accessed. If it is not present then return * NULL. Atomic context callers may also return NULL if the buffer is being - * migrated; similarly the page is not marked accessed either. + * migrated; similarly the folio is not marked accessed either. */ static struct buffer_head * find_get_block_common(struct block_device *bdev, sector_t block, @@ -1293,7 +1301,7 @@ find_get_block_common(struct block_device *bdev, sector_t block, struct buffer_head *bh = lookup_bh_lru(bdev, block, size); if (bh == NULL) { - /* __find_get_block_slow will mark the page accessed */ + /* __find_get_block_slow will mark the folio accessed */ bh = __find_get_block_slow(bdev, block, atomic); if (bh) bh_lru_install(bh); @@ -1479,15 +1487,14 @@ void folio_set_bh(struct buffer_head *bh, struct folio *folio, } EXPORT_SYMBOL(folio_set_bh); -/* - * Called when truncating a buffer on a page completely. - */ - /* Bits that are cleared during an invalidate */ #define BUFFER_FLAGS_DISCARD \ (1 << BH_Mapped | 1 << BH_New | 1 << BH_Req | \ 1 << BH_Delay | 1 << BH_Unwritten | 1 << BH_Write_EIO) +/* + * Called when truncating a buffer on a folio completely. + */ static void discard_buffer(struct buffer_head * bh) { unsigned long b_state; @@ -1615,9 +1622,7 @@ EXPORT_SYMBOL(create_empty_buffers); * moment when something will explicitly mark the buffer dirty (hopefully that * will not happen until we will free that block ;-) We don't even need to mark * it not-uptodate - nobody can expect anything from a newly allocated buffer - * anyway. We used to use unmap_buffer() for such invalidation, but that was - * wrong. We definitely don't want to mark the alias unmapped, for example - it - * would confuse anyone who might pick it with bread() afterwards... + * anyway. * * Also.. Note that bforget() doesn't lock the buffer. So there can be * writeout I/O going on against recently-freed buffers. We don't wait on that @@ -1653,7 +1658,7 @@ void clean_bdev_aliases(struct block_device *bdev, sector_t block, sector_t len) /* Recheck when the folio is locked which pins bhs */ head = folio_buffers(folio); if (!head) - goto unlock_page; + goto unlock_folio; bh = head; do { if (!buffer_mapped(bh) || (bh->b_blocknr < block)) @@ -1666,7 +1671,7 @@ void clean_bdev_aliases(struct block_device *bdev, sector_t block, sector_t len) next: bh = bh->b_this_page; } while (bh != head); -unlock_page: +unlock_folio: folio_unlock(folio); } folio_batch_release(&fbatch); @@ -1714,7 +1719,7 @@ static struct buffer_head *folio_create_buffers(struct folio *folio, * * If block_write_full_folio() is called for regular writeback * (wbc->sync_mode == WB_SYNC_NONE) then it will redirty a folio which - * has a locked buffer. This only can happen if someone has written + * has a locked buffer. This can only happen if someone has written * the buffer directly, with bh_submit(). At the address_space level * the folio writeback flag prevents this contention from occurring. * @@ -2217,9 +2222,9 @@ int generic_write_end(const struct kiocb *iocb, struct address_space *mapping, if (old_size < pos) pagecache_isize_extended(inode, old_size, pos); /* - * Don't mark the inode dirty under page lock. First, it unnecessarily - * makes the holding time of page lock longer. Second, it forces lock - * ordering of page lock and transaction start for journaling + * Don't mark the inode dirty under folio lock. First, it unnecessarily + * makes the holding time of folio lock longer. Second, it forces lock + * ordering of folio lock and transaction start for journaling * filesystems. */ if (i_size_changed) @@ -2345,7 +2350,7 @@ int block_read_full_folio(struct folio *folio, get_block_t *get_block) * BH_Async_Read tells end_buffer_async_read() that this * buffer is not under async I/O. * - * The folio comes unlocked when it has no locked + * The folio is unlocked when it has no locked * buffer_async buffers left. * * The folio lock prevents anyone starting new async @@ -2455,7 +2460,7 @@ static int cont_expand_zero(const struct kiocb *iocb, } } - /* page covers the boundary, find the boundary offset */ + /* folio crosses the boundary, find the boundary offset */ if (index == curidx) { zerofrom = curpos & ~PAGE_MASK; /* if we will expand the thing last block will be filled */ @@ -2513,18 +2518,18 @@ EXPORT_SYMBOL(cont_write_begin); /* * block_page_mkwrite() is not allowed to change the file size as it gets - * called from a page fault handler when a page is first dirtied. Hence we must - * be careful to check for EOF conditions here. We set the page up correctly - * for a written page which means we get ENOSPC checking when writing into + * called from a page fault handler when a folio is first dirtied. Hence we must + * be careful to check for EOF conditions here. We set the folio up correctly + * for a written folio which means we get ENOSPC checking when writing into * holes and correct delalloc and unwritten extent mapping on filesystems that * support these features. * * We are not allowed to take the i_rwsem here so we have to play games to - * protect against truncate races as the page could now be beyond EOF. Because - * truncate writes the inode size before removing pages, once we have the - * page lock we can determine safely if the page is beyond EOF. If it is not - * beyond EOF, then the page is guaranteed safe against truncation until we - * unlock the page. + * protect against truncate races as the folio could now be beyond EOF. Because + * truncate writes the inode size before removing folios, once we have the + * folio lock we can determine safely if the folio is beyond EOF. If it is not + * beyond EOF, then the folio is guaranteed safe against truncation until we + * unlock the folio. * * Direct callers of this function should protect against filesystem freezing * using sb_start_pagefault() - sb_end_pagefault() functions. @@ -2542,7 +2547,7 @@ int block_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf, size = i_size_read(inode); if ((folio->mapping != inode->i_mapping) || (folio_pos(folio) >= size)) { - /* We overload EFAULT to mean page got truncated */ + /* We overload EFAULT to mean folio got truncated */ ret = -EFAULT; goto out_unlock; } diff --git a/include/linux/buffer_head.h b/include/linux/buffer_head.h index 20b8fca1abfa..e7a701ce029d 100644 --- a/include/linux/buffer_head.h +++ b/include/linux/buffer_head.h @@ -364,20 +364,58 @@ static inline void bforget(struct buffer_head *bh) __bforget(bh); } -static inline struct buffer_head * -sb_bread(struct super_block *sb, sector_t block) +/** + * sb_bread - Read a block. + * @sb: The superblock to read from. + * @block: Block number in units of block size. + * + * Read a specified block, and return the buffer head that refers + * to it. The memory is allocated from the movable area so that it can + * be migrated. The returned buffer head has its refcount increased. + * The caller should call brelse() when it has finished with the buffer. + * + * Context: May sleep waiting for I/O. + * Return: NULL if the block was unreadable. + */ +static inline +struct buffer_head *sb_bread(struct super_block *sb, sector_t block) { return __bread_gfp(sb->s_bdev, block, sb->s_blocksize, __GFP_MOVABLE); } -static inline struct buffer_head * -sb_bread_unmovable(struct super_block *sb, sector_t block) +/** + * sb_bread_unmovable - Read a block. + * @sb: The superblock to read from. + * @block: Block number in units of block size. + * + * Read a specified block, and return the buffer head that refers to it. + * The memory is allocated from the unmovable area so that pointers into + * it remain valid after compaction runs. The returned buffer head has + * its refcount increased. The caller should call brelse() when it has + * finished with the buffer. + * + * Context: May sleep waiting for I/O. + * Return: NULL if the block was unreadable. + */ +static inline +struct buffer_head *sb_bread_unmovable(struct super_block *sb, sector_t block) { return __bread_gfp(sb->s_bdev, block, sb->s_blocksize, 0); } -static inline void -sb_breadahead(struct super_block *sb, sector_t block) +/** + * sb_breadahead - Start readahead. + * @sb: Superblock identifying the block device. + * @block: The block to read. + * + * Read this block. The I/O will be flagged as being readahead rather + * than immediate read, but (unlike the page cache), surrounding blocks + * will not be read. + * + * Context: May sleep in order to allocate memory. + */ +static inline +void sb_breadahead(struct super_block *sb, sector_t block) { __breadahead(sb->s_bdev, block, sb->s_blocksize); } |
