From 3be78c73179c9347bdc0a92b2898063bd2300ff7 Mon Sep 17 00:00:00 2001 From: Theodore Ts'o Date: Fri, 16 Aug 2013 21:22:41 -0400 Subject: ext4: use unsigned int for es_status values Don't use an unsigned long long for the es_status flags; this requires that we pass 64-bit values around which is painful on 32-bit systems. Instead pass the extent status flags around using the low 4 bits of an unsigned int, and shift them into place when we are reading or writing es_pblk. Signed-off-by: "Theodore Ts'o" Reviewed-by: Zheng Liu --- fs/ext4/inode.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'fs/ext4/inode.c') diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c index c2ca04e67a4..0569c745475 100644 --- a/fs/ext4/inode.c +++ b/fs/ext4/inode.c @@ -553,7 +553,7 @@ int ext4_map_blocks(handle_t *handle, struct inode *inode, } if (retval > 0) { int ret; - unsigned long long status; + unsigned int status; if (unlikely(retval != map->m_len)) { ext4_warning(inode->i_sb, @@ -653,7 +653,7 @@ found: if (retval > 0) { int ret; - unsigned long long status; + unsigned int status; if (unlikely(retval != map->m_len)) { ext4_warning(inode->i_sb, @@ -1633,7 +1633,7 @@ add_delayed: set_buffer_delay(bh); } else if (retval > 0) { int ret; - unsigned long long status; + unsigned int status; if (unlikely(retval != map->m_len)) { ext4_warning(inode->i_sb, -- cgit v1.2.3-70-g09d2 From 09930042a2e94cf8ee79d22943915612c1e4ba51 Mon Sep 17 00:00:00 2001 From: Jan Kara Date: Sat, 17 Aug 2013 09:57:56 -0400 Subject: ext4: move test whether extent to map can be extended to one place Currently the logic whether the current buffer can be added to an extent of buffers to map is split between mpage_add_bh_to_extent() and add_page_bufs_to_extent(). Move the whole logic to mpage_add_bh_to_extent() which makes things a bit more straightforward and make following i_size fixes easier. Signed-off-by: Jan Kara Signed-off-by: "Theodore Ts'o" Cc: stable@vger.kernel.org --- fs/ext4/inode.c | 49 +++++++++++++++++++++++++++++-------------------- 1 file changed, 29 insertions(+), 20 deletions(-) (limited to 'fs/ext4/inode.c') diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c index 0569c745475..787497d536b 100644 --- a/fs/ext4/inode.c +++ b/fs/ext4/inode.c @@ -1904,34 +1904,48 @@ static int ext4_writepage(struct page *page, * * @mpd - extent of blocks * @lblk - logical number of the block in the file - * @b_state - b_state of the buffer head added + * @bh - buffer head we want to add to the extent * - * the function is used to collect contig. blocks in same state + * The function is used to collect contig. blocks in the same state. If the + * buffer doesn't require mapping for writeback and we haven't started the + * extent of buffers to map yet, the function returns 'true' immediately - the + * caller can write the buffer right away. Otherwise the function returns true + * if the block has been added to the extent, false if the block couldn't be + * added. */ -static int mpage_add_bh_to_extent(struct mpage_da_data *mpd, ext4_lblk_t lblk, - unsigned long b_state) +static bool mpage_add_bh_to_extent(struct mpage_da_data *mpd, ext4_lblk_t lblk, + struct buffer_head *bh) { struct ext4_map_blocks *map = &mpd->map; - /* Don't go larger than mballoc is willing to allocate */ - if (map->m_len >= MAX_WRITEPAGES_EXTENT_LEN) - return 0; + /* Buffer that doesn't need mapping for writeback? */ + if (!buffer_dirty(bh) || !buffer_mapped(bh) || + (!buffer_delay(bh) && !buffer_unwritten(bh))) { + /* So far no extent to map => we write the buffer right away */ + if (map->m_len == 0) + return true; + return false; + } /* First block in the extent? */ if (map->m_len == 0) { map->m_lblk = lblk; map->m_len = 1; - map->m_flags = b_state & BH_FLAGS; - return 1; + map->m_flags = bh->b_state & BH_FLAGS; + return true; } + /* Don't go larger than mballoc is willing to allocate */ + if (map->m_len >= MAX_WRITEPAGES_EXTENT_LEN) + return false; + /* Can we merge the block to our big extent? */ if (lblk == map->m_lblk + map->m_len && - (b_state & BH_FLAGS) == map->m_flags) { + (bh->b_state & BH_FLAGS) == map->m_flags) { map->m_len++; - return 1; + return true; } - return 0; + return false; } static bool add_page_bufs_to_extent(struct mpage_da_data *mpd, @@ -1946,18 +1960,13 @@ static bool add_page_bufs_to_extent(struct mpage_da_data *mpd, do { BUG_ON(buffer_locked(bh)); - if (!buffer_dirty(bh) || !buffer_mapped(bh) || - (!buffer_delay(bh) && !buffer_unwritten(bh)) || - lblk >= blocks) { + if (lblk >= blocks || !mpage_add_bh_to_extent(mpd, lblk, bh)) { /* Found extent to map? */ if (mpd->map.m_len) return false; - if (lblk >= blocks) - return true; - continue; + /* Everything mapped so far and we hit EOF */ + return true; } - if (!mpage_add_bh_to_extent(mpd, lblk, bh->b_state)) - return false; } while (lblk++, (bh = bh->b_this_page) != head); return true; } -- cgit v1.2.3-70-g09d2 From 5f1132b2ba8c873f25982cf45917e8455fb6c962 Mon Sep 17 00:00:00 2001 From: Jan Kara Date: Sat, 17 Aug 2013 10:02:33 -0400 Subject: ext4: fix ext4_writepages() in presence of truncate Inode size can arbitrarily change while writeback is in progress. When ext4_writepages() has prepared a long extent for mapping and truncate then reduces i_size, mpage_map_and_submit_buffers() will always map just one buffer in a page instead of all of them due to lblk < blocks check. So we end up not using all blocks we've allocated (thus leaking them) and also delalloc accounting goes wrong manifesting as a warning like: ext4_da_release_space:1333: ext4_da_release_space: ino 12, to_free 1 with only 0 reserved data blocks Note that the problem can happen only when blocksize < pagesize because otherwise we have only a single buffer in the page. Fix the problem by removing the size check from the mapping loop. We have an extent allocated so we have to use it all before checking for i_size. We also rename add_page_bufs_to_extent() to mpage_process_page_bufs() and make that function submit the page for IO if all buffers (upto EOF) in it are mapped. Reported-by: Dave Jones Reported-by: Zheng Liu Signed-off-by: Jan Kara Signed-off-by: "Theodore Ts'o" Cc: stable@vger.kernel.org --- fs/ext4/inode.c | 107 ++++++++++++++++++++++++++++++++++---------------------- 1 file changed, 66 insertions(+), 41 deletions(-) (limited to 'fs/ext4/inode.c') diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c index 787497d536b..19fa2e07627 100644 --- a/fs/ext4/inode.c +++ b/fs/ext4/inode.c @@ -1890,6 +1890,26 @@ static int ext4_writepage(struct page *page, return ret; } +static int mpage_submit_page(struct mpage_da_data *mpd, struct page *page) +{ + int len; + loff_t size = i_size_read(mpd->inode); + int err; + + BUG_ON(page->index != mpd->first_page); + if (page->index == size >> PAGE_CACHE_SHIFT) + len = size & ~PAGE_CACHE_MASK; + else + len = PAGE_CACHE_SIZE; + clear_page_dirty_for_io(page); + err = ext4_bio_write_page(&mpd->io_submit, page, len, mpd->wbc); + if (!err) + mpd->wbc->nr_to_write--; + mpd->first_page++; + + return err; +} + #define BH_FLAGS ((1 << BH_Unwritten) | (1 << BH_Delay)) /* @@ -1948,12 +1968,29 @@ static bool mpage_add_bh_to_extent(struct mpage_da_data *mpd, ext4_lblk_t lblk, return false; } -static bool add_page_bufs_to_extent(struct mpage_da_data *mpd, - struct buffer_head *head, - struct buffer_head *bh, - ext4_lblk_t lblk) +/* + * mpage_process_page_bufs - submit page buffers for IO or add them to extent + * + * @mpd - extent of blocks for mapping + * @head - the first buffer in the page + * @bh - buffer we should start processing from + * @lblk - logical number of the block in the file corresponding to @bh + * + * Walk through page buffers from @bh upto @head (exclusive) and either submit + * the page for IO if all buffers in this page were mapped and there's no + * accumulated extent of buffers to map or add buffers in the page to the + * extent of buffers to map. The function returns 1 if the caller can continue + * by processing the next page, 0 if it should stop adding buffers to the + * extent to map because we cannot extend it anymore. It can also return value + * < 0 in case of error during IO submission. + */ +static int mpage_process_page_bufs(struct mpage_da_data *mpd, + struct buffer_head *head, + struct buffer_head *bh, + ext4_lblk_t lblk) { struct inode *inode = mpd->inode; + int err; ext4_lblk_t blocks = (i_size_read(inode) + (1 << inode->i_blkbits) - 1) >> inode->i_blkbits; @@ -1963,32 +2000,18 @@ static bool add_page_bufs_to_extent(struct mpage_da_data *mpd, if (lblk >= blocks || !mpage_add_bh_to_extent(mpd, lblk, bh)) { /* Found extent to map? */ if (mpd->map.m_len) - return false; + return 0; /* Everything mapped so far and we hit EOF */ - return true; + break; } } while (lblk++, (bh = bh->b_this_page) != head); - return true; -} - -static int mpage_submit_page(struct mpage_da_data *mpd, struct page *page) -{ - int len; - loff_t size = i_size_read(mpd->inode); - int err; - - BUG_ON(page->index != mpd->first_page); - if (page->index == size >> PAGE_CACHE_SHIFT) - len = size & ~PAGE_CACHE_MASK; - else - len = PAGE_CACHE_SIZE; - clear_page_dirty_for_io(page); - err = ext4_bio_write_page(&mpd->io_submit, page, len, mpd->wbc); - if (!err) - mpd->wbc->nr_to_write--; - mpd->first_page++; - - return err; + /* So far everything mapped? Submit the page for IO. */ + if (mpd->map.m_len == 0) { + err = mpage_submit_page(mpd, head->b_page); + if (err < 0) + return err; + } + return lblk < blocks; } /* @@ -2012,8 +2035,6 @@ static int mpage_map_and_submit_buffers(struct mpage_da_data *mpd) struct inode *inode = mpd->inode; struct buffer_head *head, *bh; int bpp_bits = PAGE_CACHE_SHIFT - inode->i_blkbits; - ext4_lblk_t blocks = (i_size_read(inode) + (1 << inode->i_blkbits) - 1) - >> inode->i_blkbits; pgoff_t start, end; ext4_lblk_t lblk; sector_t pblock; @@ -2048,18 +2069,26 @@ static int mpage_map_and_submit_buffers(struct mpage_da_data *mpd) */ mpd->map.m_len = 0; mpd->map.m_flags = 0; - add_page_bufs_to_extent(mpd, head, bh, - lblk); + /* + * FIXME: If dioread_nolock supports + * blocksize < pagesize, we need to make + * sure we add size mapped so far to + * io_end->size as the following call + * can submit the page for IO. + */ + err = mpage_process_page_bufs(mpd, head, + bh, lblk); pagevec_release(&pvec); - return 0; + if (err > 0) + err = 0; + return err; } if (buffer_delay(bh)) { clear_buffer_delay(bh); bh->b_blocknr = pblock++; } clear_buffer_unwritten(bh); - } while (++lblk < blocks && - (bh = bh->b_this_page) != head); + } while (lblk++, (bh = bh->b_this_page) != head); /* * FIXME: This is going to break if dioread_nolock @@ -2328,14 +2357,10 @@ static int mpage_prepare_extent_to_map(struct mpage_da_data *mpd) lblk = ((ext4_lblk_t)page->index) << (PAGE_CACHE_SHIFT - blkbits); head = page_buffers(page); - if (!add_page_bufs_to_extent(mpd, head, head, lblk)) + err = mpage_process_page_bufs(mpd, head, head, lblk); + if (err <= 0) goto out; - /* So far everything mapped? Submit the page for IO. */ - if (mpd->map.m_len == 0) { - err = mpage_submit_page(mpd, page); - if (err < 0) - goto out; - } + err = 0; /* * Accumulated enough dirty pages? This doesn't apply -- cgit v1.2.3-70-g09d2 From 5208386c501276df18fee464e21d3c58d2d79517 Mon Sep 17 00:00:00 2001 From: Jan Kara Date: Sat, 17 Aug 2013 10:07:17 -0400 Subject: ext4: simplify truncation code in ext4_setattr() Merge conditions in ext4_setattr() handling inode size changes, also move ext4_begin_ordered_truncate() call somewhat earlier because it simplifies error recovery in case of failure. Also add error handling in case i_disksize update fails. Signed-off-by: Jan Kara Signed-off-by: "Theodore Ts'o" Cc: stable@vger.kernel.org --- fs/ext4/inode.c | 109 +++++++++++++++++++++++++------------------------------- 1 file changed, 49 insertions(+), 60 deletions(-) (limited to 'fs/ext4/inode.c') diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c index 19fa2e07627..38f430119fe 100644 --- a/fs/ext4/inode.c +++ b/fs/ext4/inode.c @@ -4600,7 +4600,9 @@ int ext4_setattr(struct dentry *dentry, struct iattr *attr) ext4_journal_stop(handle); } - if (attr->ia_valid & ATTR_SIZE) { + if (attr->ia_valid & ATTR_SIZE && attr->ia_size != inode->i_size) { + handle_t *handle; + loff_t oldsize = inode->i_size; if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))) { struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); @@ -4608,73 +4610,60 @@ int ext4_setattr(struct dentry *dentry, struct iattr *attr) if (attr->ia_size > sbi->s_bitmap_maxbytes) return -EFBIG; } - } - - if (S_ISREG(inode->i_mode) && - attr->ia_valid & ATTR_SIZE && - (attr->ia_size < inode->i_size)) { - handle_t *handle; - - handle = ext4_journal_start(inode, EXT4_HT_INODE, 3); - if (IS_ERR(handle)) { - error = PTR_ERR(handle); - goto err_out; - } - if (ext4_handle_valid(handle)) { - error = ext4_orphan_add(handle, inode); - orphan = 1; - } - EXT4_I(inode)->i_disksize = attr->ia_size; - rc = ext4_mark_inode_dirty(handle, inode); - if (!error) - error = rc; - ext4_journal_stop(handle); - - if (ext4_should_order_data(inode)) { - error = ext4_begin_ordered_truncate(inode, + if (S_ISREG(inode->i_mode) && + (attr->ia_size < inode->i_size)) { + if (ext4_should_order_data(inode)) { + error = ext4_begin_ordered_truncate(inode, attr->ia_size); - if (error) { - /* Do as much error cleanup as possible */ - handle = ext4_journal_start(inode, - EXT4_HT_INODE, 3); - if (IS_ERR(handle)) { - ext4_orphan_del(NULL, inode); + if (error) goto err_out; - } - ext4_orphan_del(handle, inode); - orphan = 0; - ext4_journal_stop(handle); + } + handle = ext4_journal_start(inode, EXT4_HT_INODE, 3); + if (IS_ERR(handle)) { + error = PTR_ERR(handle); + goto err_out; + } + if (ext4_handle_valid(handle)) { + error = ext4_orphan_add(handle, inode); + orphan = 1; + } + EXT4_I(inode)->i_disksize = attr->ia_size; + rc = ext4_mark_inode_dirty(handle, inode); + if (!error) + error = rc; + ext4_journal_stop(handle); + if (error) { + ext4_orphan_del(NULL, inode); goto err_out; } } - } - - if (attr->ia_valid & ATTR_SIZE) { - if (attr->ia_size != inode->i_size) { - loff_t oldsize = inode->i_size; - i_size_write(inode, attr->ia_size); - /* - * Blocks are going to be removed from the inode. Wait - * for dio in flight. Temporarily disable - * dioread_nolock to prevent livelock. - */ - if (orphan) { - if (!ext4_should_journal_data(inode)) { - ext4_inode_block_unlocked_dio(inode); - inode_dio_wait(inode); - ext4_inode_resume_unlocked_dio(inode); - } else - ext4_wait_for_tail_page_commit(inode); - } - /* - * Truncate pagecache after we've waited for commit - * in data=journal mode to make pages freeable. - */ - truncate_pagecache(inode, oldsize, inode->i_size); + i_size_write(inode, attr->ia_size); + /* + * Blocks are going to be removed from the inode. Wait + * for dio in flight. Temporarily disable + * dioread_nolock to prevent livelock. + */ + if (orphan) { + if (!ext4_should_journal_data(inode)) { + ext4_inode_block_unlocked_dio(inode); + inode_dio_wait(inode); + ext4_inode_resume_unlocked_dio(inode); + } else + ext4_wait_for_tail_page_commit(inode); } - ext4_truncate(inode); + /* + * Truncate pagecache after we've waited for commit + * in data=journal mode to make pages freeable. + */ + truncate_pagecache(inode, oldsize, inode->i_size); } + /* + * We want to call ext4_truncate() even if attr->ia_size == + * inode->i_size for cases like truncation of fallocated space + */ + if (attr->ia_valid & ATTR_SIZE) + ext4_truncate(inode); if (!rc) { setattr_copy(inode, attr); -- cgit v1.2.3-70-g09d2 From 90e775b71ac4e685898c7995756fe58c135adaa6 Mon Sep 17 00:00:00 2001 From: Jan Kara Date: Sat, 17 Aug 2013 10:09:31 -0400 Subject: ext4: fix lost truncate due to race with writeback The following race can lead to a loss of i_disksize update from truncate thus resulting in a wrong inode size if the inode size isn't updated again before inode is reclaimed: ext4_setattr() mpage_map_and_submit_extent() EXT4_I(inode)->i_disksize = attr->ia_size; ... ... disksize = ((loff_t)mpd->first_page) << PAGE_CACHE_SHIFT /* False because i_size isn't * updated yet */ if (disksize > i_size_read(inode)) /* True, because i_disksize is * already truncated */ if (disksize > EXT4_I(inode)->i_disksize) /* Overwrite i_disksize * update from truncate */ ext4_update_i_disksize() i_size_write(inode, attr->ia_size); For other places updating i_disksize such race cannot happen because i_mutex prevents these races. Writeback is the only place where we do not hold i_mutex and we cannot grab it there because of lock ordering. We fix the race by doing both i_disksize and i_size update in truncate atomically under i_data_sem and in mpage_map_and_submit_extent() we move the check against i_size under i_data_sem as well. Signed-off-by: Jan Kara Signed-off-by: "Theodore Ts'o" Cc: stable@vger.kernel.org --- fs/ext4/ext4.h | 24 ++++++++++++++++++++---- fs/ext4/inode.c | 17 ++++++++++++----- 2 files changed, 32 insertions(+), 9 deletions(-) (limited to 'fs/ext4/inode.c') diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h index 58dede76f75..3dbc56eb484 100644 --- a/fs/ext4/ext4.h +++ b/fs/ext4/ext4.h @@ -2432,16 +2432,32 @@ do { \ #define EXT4_FREECLUSTERS_WATERMARK 0 #endif +/* Update i_disksize. Requires i_mutex to avoid races with truncate */ static inline void ext4_update_i_disksize(struct inode *inode, loff_t newsize) { - /* - * XXX: replace with spinlock if seen contended -bzzz - */ + WARN_ON_ONCE(S_ISREG(inode->i_mode) && + !mutex_is_locked(&inode->i_mutex)); + down_write(&EXT4_I(inode)->i_data_sem); + if (newsize > EXT4_I(inode)->i_disksize) + EXT4_I(inode)->i_disksize = newsize; + up_write(&EXT4_I(inode)->i_data_sem); +} + +/* + * Update i_disksize after writeback has been started. Races with truncate + * are avoided by checking i_size under i_data_sem. + */ +static inline void ext4_wb_update_i_disksize(struct inode *inode, loff_t newsize) +{ + loff_t i_size; + down_write(&EXT4_I(inode)->i_data_sem); + i_size = i_size_read(inode); + if (newsize > i_size) + newsize = i_size; if (newsize > EXT4_I(inode)->i_disksize) EXT4_I(inode)->i_disksize = newsize; up_write(&EXT4_I(inode)->i_data_sem); - return ; } struct ext4_group_info { diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c index 38f430119fe..fc4051eb4e0 100644 --- a/fs/ext4/inode.c +++ b/fs/ext4/inode.c @@ -2237,12 +2237,10 @@ static int mpage_map_and_submit_extent(handle_t *handle, /* Update on-disk size after IO is submitted */ disksize = ((loff_t)mpd->first_page) << PAGE_CACHE_SHIFT; - if (disksize > i_size_read(inode)) - disksize = i_size_read(inode); if (disksize > EXT4_I(inode)->i_disksize) { int err2; - ext4_update_i_disksize(inode, disksize); + ext4_wb_update_i_disksize(inode, disksize); err2 = ext4_mark_inode_dirty(handle, inode); if (err2) ext4_error(inode->i_sb, @@ -4627,18 +4625,27 @@ int ext4_setattr(struct dentry *dentry, struct iattr *attr) error = ext4_orphan_add(handle, inode); orphan = 1; } + down_write(&EXT4_I(inode)->i_data_sem); EXT4_I(inode)->i_disksize = attr->ia_size; rc = ext4_mark_inode_dirty(handle, inode); if (!error) error = rc; + /* + * We have to update i_size under i_data_sem together + * with i_disksize to avoid races with writeback code + * running ext4_wb_update_i_disksize(). + */ + if (!error) + i_size_write(inode, attr->ia_size); + up_write(&EXT4_I(inode)->i_data_sem); ext4_journal_stop(handle); if (error) { ext4_orphan_del(NULL, inode); goto err_out; } - } + } else + i_size_write(inode, attr->ia_size); - i_size_write(inode, attr->ia_size); /* * Blocks are going to be removed from the inode. Wait * for dio in flight. Temporarily disable -- cgit v1.2.3-70-g09d2 From 7afe5aa59ed3da7b6161617e7f157c7c680dc41e Mon Sep 17 00:00:00 2001 From: Dmitry Monakhov Date: Wed, 28 Aug 2013 14:30:47 -0400 Subject: ext4: convert write_begin methods to stable_page_writes semantics Use wait_for_stable_page() instead of wait_on_page_writeback() Signed-off-by: Dmitry Monakhov Signed-off-by: "Theodore Ts'o" Reviewed-by: Jan Kara --- fs/ext4/inode.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'fs/ext4/inode.c') diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c index fc4051eb4e0..47c8e465437 100644 --- a/fs/ext4/inode.c +++ b/fs/ext4/inode.c @@ -969,7 +969,8 @@ retry_journal: ext4_journal_stop(handle); goto retry_grab; } - wait_on_page_writeback(page); + /* In case writeback began while the page was unlocked */ + wait_for_stable_page(page); if (ext4_should_dioread_nolock(inode)) ret = __block_write_begin(page, pos, len, ext4_get_block_write); @@ -2678,7 +2679,7 @@ retry_journal: goto retry_grab; } /* In case writeback began while the page was unlocked */ - wait_on_page_writeback(page); + wait_for_stable_page(page); ret = __block_write_begin(page, pos, len, ext4_da_get_block_prep); if (ret < 0) { -- cgit v1.2.3-70-g09d2 From 70261f568f3c08552f034742e3d5cb78c3877766 Mon Sep 17 00:00:00 2001 From: Anatol Pomozov Date: Wed, 28 Aug 2013 14:40:12 -0400 Subject: ext4: Fix misspellings using 'codespell' tool Signed-off-by: Anatol Pomozov Signed-off-by: "Theodore Ts'o" --- fs/ext3/dir.c | 2 +- fs/ext4/dir.c | 2 +- fs/ext4/ext4_jbd2.h | 2 +- fs/ext4/extents.c | 2 +- fs/ext4/inode.c | 8 ++++---- fs/ext4/migrate.c | 2 +- 6 files changed, 9 insertions(+), 9 deletions(-) (limited to 'fs/ext4/inode.c') diff --git a/fs/ext3/dir.c b/fs/ext3/dir.c index f522425aaa2..bafdd48eefd 100644 --- a/fs/ext3/dir.c +++ b/fs/ext3/dir.c @@ -41,7 +41,7 @@ static unsigned char get_dtype(struct super_block *sb, int filetype) /** * Check if the given dir-inode refers to an htree-indexed directory - * (or a directory which chould potentially get coverted to use htree + * (or a directory which could potentially get converted to use htree * indexing). * * Return 1 if it is a dx dir, 0 if not diff --git a/fs/ext4/dir.c b/fs/ext4/dir.c index 3c7d288ae94..680bb338891 100644 --- a/fs/ext4/dir.c +++ b/fs/ext4/dir.c @@ -33,7 +33,7 @@ static int ext4_dx_readdir(struct file *, struct dir_context *); /** * Check if the given dir-inode refers to an htree-indexed directory - * (or a directory which chould potentially get coverted to use htree + * (or a directory which could potentially get converted to use htree * indexing). * * Return 1 if it is a dx dir, 0 if not diff --git a/fs/ext4/ext4_jbd2.h b/fs/ext4/ext4_jbd2.h index 2877258d949..81cfefa9dc0 100644 --- a/fs/ext4/ext4_jbd2.h +++ b/fs/ext4/ext4_jbd2.h @@ -197,7 +197,7 @@ static inline void ext4_journal_callback_add(handle_t *handle, * ext4_journal_callback_del: delete a registered callback * @handle: active journal transaction handle on which callback was registered * @jce: registered journal callback entry to unregister - * Return true if object was sucessfully removed + * Return true if object was successfully removed */ static inline bool ext4_journal_callback_try_del(handle_t *handle, struct ext4_journal_cb_entry *jce) diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c index e7580aeb6fb..916e88498c8 100644 --- a/fs/ext4/extents.c +++ b/fs/ext4/extents.c @@ -3220,7 +3220,7 @@ fix_extent_len: * ext4_split_extents() splits an extent and mark extent which is covered * by @map as split_flags indicates * - * It may result in splitting the extent into multiple extents (upto three) + * It may result in splitting the extent into multiple extents (up to three) * There are three possibilities: * a> There is no split required * b> Splits in two extents: Split is happening at either end of the extent diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c index 47c8e465437..9115f280751 100644 --- a/fs/ext4/inode.c +++ b/fs/ext4/inode.c @@ -1916,7 +1916,7 @@ static int mpage_submit_page(struct mpage_da_data *mpd, struct page *page) /* * mballoc gives us at most this number of blocks... * XXX: That seems to be only a limitation of ext4_mb_normalize_request(). - * The rest of mballoc seems to handle chunks upto full group size. + * The rest of mballoc seems to handle chunks up to full group size. */ #define MAX_WRITEPAGES_EXTENT_LEN 2048 @@ -2057,7 +2057,7 @@ static int mpage_map_and_submit_buffers(struct mpage_da_data *mpd) if (page->index > end) break; - /* Upto 'end' pages must be contiguous */ + /* Up to 'end' pages must be contiguous */ BUG_ON(page->index != start); bh = head = page_buffers(page); do { @@ -2256,7 +2256,7 @@ static int mpage_map_and_submit_extent(handle_t *handle, /* * Calculate the total number of credits to reserve for one writepages * iteration. This is called from ext4_writepages(). We map an extent of - * upto MAX_WRITEPAGES_EXTENT_LEN blocks and then we go on and finish mapping + * up to MAX_WRITEPAGES_EXTENT_LEN blocks and then we go on and finish mapping * the last partial page. So in total we can map MAX_WRITEPAGES_EXTENT_LEN + * bpp - 1 blocks in bpp different extents. */ @@ -2443,7 +2443,7 @@ static int ext4_writepages(struct address_space *mapping, if (ext4_should_dioread_nolock(inode)) { /* - * We may need to convert upto one extent per block in + * We may need to convert up to one extent per block in * the page and we may dirty the inode. */ rsv_blocks = 1 + (PAGE_CACHE_SIZE >> inode->i_blkbits); diff --git a/fs/ext4/migrate.c b/fs/ext4/migrate.c index f99bdb8548b..2ae73a80c19 100644 --- a/fs/ext4/migrate.c +++ b/fs/ext4/migrate.c @@ -494,7 +494,7 @@ int ext4_ext_migrate(struct inode *inode) * superblock modification. * * For the tmp_inode we already have committed the - * trascation that created the inode. Later as and + * transaction that created the inode. Later as and * when we add extents we extent the journal */ /* -- cgit v1.2.3-70-g09d2 From 7b7a8665edd8db733980389b098530f9e4f630b2 Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Wed, 4 Sep 2013 15:04:39 +0200 Subject: direct-io: Implement generic deferred AIO completions Add support to the core direct-io code to defer AIO completions to user context using a workqueue. This replaces opencoded and less efficient code in XFS and ext4 (we save a memory allocation for each direct IO) and will be needed to properly support O_(D)SYNC for AIO. The communication between the filesystem and the direct I/O code requires a new buffer head flag, which is a bit ugly but not avoidable until the direct I/O code stops abusing the buffer_head structure for communicating with the filesystems. Currently this creates a per-superblock unbound workqueue for these completions, which is taken from an earlier patch by Jan Kara. I'm not really convinced about this use and would prefer a "normal" global workqueue with a high concurrency limit, but this needs further discussion. JK: Fixed ext4 part, dynamic allocation of the workqueue. Signed-off-by: Christoph Hellwig Signed-off-by: Jan Kara Signed-off-by: Al Viro --- fs/direct-io.c | 85 ++++++++++++++++++++++++++++++++++++--------- fs/ext4/ext4.h | 11 ------ fs/ext4/inode.c | 28 ++++----------- fs/ext4/page-io.c | 30 ++++------------ fs/ext4/super.c | 16 --------- fs/ocfs2/aops.c | 8 +---- fs/super.c | 18 +++++----- fs/xfs/xfs_aops.c | 28 +++------------ fs/xfs/xfs_aops.h | 3 -- include/linux/buffer_head.h | 2 ++ include/linux/fs.h | 7 ++-- 11 files changed, 105 insertions(+), 131 deletions(-) (limited to 'fs/ext4/inode.c') diff --git a/fs/direct-io.c b/fs/direct-io.c index 7ab90f5081e..8b31b9f449f 100644 --- a/fs/direct-io.c +++ b/fs/direct-io.c @@ -127,6 +127,7 @@ struct dio { spinlock_t bio_lock; /* protects BIO fields below */ int page_errors; /* errno from get_user_pages() */ int is_async; /* is IO async ? */ + bool defer_completion; /* defer AIO completion to workqueue? */ int io_error; /* IO error in completion path */ unsigned long refcount; /* direct_io_worker() and bios */ struct bio *bio_list; /* singly linked via bi_private */ @@ -141,7 +142,10 @@ struct dio { * allocation time. Don't add new fields after pages[] unless you * wish that they not be zeroed. */ - struct page *pages[DIO_PAGES]; /* page buffer */ + union { + struct page *pages[DIO_PAGES]; /* page buffer */ + struct work_struct complete_work;/* deferred AIO completion */ + }; } ____cacheline_aligned_in_smp; static struct kmem_cache *dio_cache __read_mostly; @@ -221,16 +225,16 @@ static inline struct page *dio_get_page(struct dio *dio, * dio_complete() - called when all DIO BIO I/O has been completed * @offset: the byte offset in the file of the completed operation * - * This releases locks as dictated by the locking type, lets interested parties - * know that a DIO operation has completed, and calculates the resulting return - * code for the operation. + * This drops i_dio_count, lets interested parties know that a DIO operation + * has completed, and calculates the resulting return code for the operation. * * It lets the filesystem know if it registered an interest earlier via * get_block. Pass the private field of the map buffer_head so that * filesystems can use it to hold additional state between get_block calls and * dio_complete. */ -static ssize_t dio_complete(struct dio *dio, loff_t offset, ssize_t ret, bool is_async) +static ssize_t dio_complete(struct dio *dio, loff_t offset, ssize_t ret, + bool is_async) { ssize_t transferred = 0; @@ -258,19 +262,26 @@ static ssize_t dio_complete(struct dio *dio, loff_t offset, ssize_t ret, bool is if (ret == 0) ret = transferred; - if (dio->end_io && dio->result) { - dio->end_io(dio->iocb, offset, transferred, - dio->private, ret, is_async); - } else { - inode_dio_done(dio->inode); - if (is_async) - aio_complete(dio->iocb, ret, 0); - } + if (dio->end_io && dio->result) + dio->end_io(dio->iocb, offset, transferred, dio->private); + + inode_dio_done(dio->inode); + if (is_async) + aio_complete(dio->iocb, ret, 0); + kmem_cache_free(dio_cache, dio); return ret; } +static void dio_aio_complete_work(struct work_struct *work) +{ + struct dio *dio = container_of(work, struct dio, complete_work); + + dio_complete(dio, dio->iocb->ki_pos, 0, true); +} + static int dio_bio_complete(struct dio *dio, struct bio *bio); + /* * Asynchronous IO callback. */ @@ -290,8 +301,13 @@ static void dio_bio_end_aio(struct bio *bio, int error) spin_unlock_irqrestore(&dio->bio_lock, flags); if (remaining == 0) { - dio_complete(dio, dio->iocb->ki_pos, 0, true); - kmem_cache_free(dio_cache, dio); + if (dio->result && dio->defer_completion) { + INIT_WORK(&dio->complete_work, dio_aio_complete_work); + queue_work(dio->inode->i_sb->s_dio_done_wq, + &dio->complete_work); + } else { + dio_complete(dio, dio->iocb->ki_pos, 0, true); + } } } @@ -510,6 +526,41 @@ static inline int dio_bio_reap(struct dio *dio, struct dio_submit *sdio) return ret; } +/* + * Create workqueue for deferred direct IO completions. We allocate the + * workqueue when it's first needed. This avoids creating workqueue for + * filesystems that don't need it and also allows us to create the workqueue + * late enough so the we can include s_id in the name of the workqueue. + */ +static int sb_init_dio_done_wq(struct super_block *sb) +{ + struct workqueue_struct *wq = alloc_workqueue("dio/%s", + WQ_MEM_RECLAIM, 0, + sb->s_id); + if (!wq) + return -ENOMEM; + /* + * This has to be atomic as more DIOs can race to create the workqueue + */ + cmpxchg(&sb->s_dio_done_wq, NULL, wq); + /* Someone created workqueue before us? Free ours... */ + if (wq != sb->s_dio_done_wq) + destroy_workqueue(wq); + return 0; +} + +static int dio_set_defer_completion(struct dio *dio) +{ + struct super_block *sb = dio->inode->i_sb; + + if (dio->defer_completion) + return 0; + dio->defer_completion = true; + if (!sb->s_dio_done_wq) + return sb_init_dio_done_wq(sb); + return 0; +} + /* * Call into the fs to map some more disk blocks. We record the current number * of available blocks at sdio->blocks_available. These are in units of the @@ -581,6 +632,9 @@ static int get_more_blocks(struct dio *dio, struct dio_submit *sdio, /* Store for completion */ dio->private = map_bh->b_private; + + if (ret == 0 && buffer_defer_completion(map_bh)) + ret = dio_set_defer_completion(dio); } return ret; } @@ -1269,7 +1323,6 @@ do_blockdev_direct_IO(int rw, struct kiocb *iocb, struct inode *inode, if (drop_refcount(dio) == 0) { retval = dio_complete(dio, offset, retval, false); - kmem_cache_free(dio_cache, dio); } else BUG_ON(retval != -EIOCBQUEUED); diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h index 0ab26fbf338..b247fbbed99 100644 --- a/fs/ext4/ext4.h +++ b/fs/ext4/ext4.h @@ -180,7 +180,6 @@ struct ext4_map_blocks { * Flags for ext4_io_end->flags */ #define EXT4_IO_END_UNWRITTEN 0x0001 -#define EXT4_IO_END_DIRECT 0x0002 /* * For converting uninitialized extents on a work queue. 'handle' is used for @@ -196,8 +195,6 @@ typedef struct ext4_io_end { unsigned int flag; /* unwritten or not */ loff_t offset; /* offset in the file */ ssize_t size; /* size of the extent */ - struct kiocb *iocb; /* iocb struct for AIO */ - int result; /* error value for AIO */ atomic_t count; /* reference counter */ } ext4_io_end_t; @@ -900,11 +897,9 @@ struct ext4_inode_info { * Completed IOs that need unwritten extents handling and don't have * transaction reserved */ - struct list_head i_unrsv_conversion_list; atomic_t i_ioend_count; /* Number of outstanding io_end structs */ atomic_t i_unwritten; /* Nr. of inflight conversions pending */ struct work_struct i_rsv_conversion_work; - struct work_struct i_unrsv_conversion_work; spinlock_t i_block_reservation_lock; @@ -1276,8 +1271,6 @@ struct ext4_sb_info { struct flex_groups *s_flex_groups; ext4_group_t s_flex_groups_allocated; - /* workqueue for unreserved extent convertions (dio) */ - struct workqueue_struct *unrsv_conversion_wq; /* workqueue for reserved extent conversions (buffered io) */ struct workqueue_struct *rsv_conversion_wq; @@ -1340,9 +1333,6 @@ static inline void ext4_set_io_unwritten_flag(struct inode *inode, struct ext4_io_end *io_end) { if (!(io_end->flag & EXT4_IO_END_UNWRITTEN)) { - /* Writeback has to have coversion transaction reserved */ - WARN_ON(EXT4_SB(inode->i_sb)->s_journal && !io_end->handle && - !(io_end->flag & EXT4_IO_END_DIRECT)); io_end->flag |= EXT4_IO_END_UNWRITTEN; atomic_inc(&EXT4_I(inode)->i_unwritten); } @@ -2716,7 +2706,6 @@ extern void ext4_put_io_end_defer(ext4_io_end_t *io_end); extern void ext4_io_submit_init(struct ext4_io_submit *io, struct writeback_control *wbc); extern void ext4_end_io_rsv_work(struct work_struct *work); -extern void ext4_end_io_unrsv_work(struct work_struct *work); extern void ext4_io_submit(struct ext4_io_submit *io); extern int ext4_bio_write_page(struct ext4_io_submit *io, struct page *page, diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c index c2ca04e67a4..123bd81692d 100644 --- a/fs/ext4/inode.c +++ b/fs/ext4/inode.c @@ -727,8 +727,12 @@ static int _ext4_get_block(struct inode *inode, sector_t iblock, ret = ext4_map_blocks(handle, inode, &map, flags); if (ret > 0) { + ext4_io_end_t *io_end = ext4_inode_aio(inode); + map_bh(bh, inode->i_sb, map.m_pblk); bh->b_state = (bh->b_state & ~EXT4_MAP_FLAGS) | map.m_flags; + if (io_end && io_end->flag & EXT4_IO_END_UNWRITTEN) + set_buffer_defer_completion(bh); bh->b_size = inode->i_sb->s_blocksize * map.m_len; ret = 0; } @@ -2991,19 +2995,13 @@ static int ext4_get_block_write_nolock(struct inode *inode, sector_t iblock, } static void ext4_end_io_dio(struct kiocb *iocb, loff_t offset, - ssize_t size, void *private, int ret, - bool is_async) + ssize_t size, void *private) { - struct inode *inode = file_inode(iocb->ki_filp); ext4_io_end_t *io_end = iocb->private; /* if not async direct IO just return */ - if (!io_end) { - inode_dio_done(inode); - if (is_async) - aio_complete(iocb, ret, 0); + if (!io_end) return; - } ext_debug("ext4_end_io_dio(): io_end 0x%p " "for inode %lu, iocb 0x%p, offset %llu, size %zd\n", @@ -3013,11 +3011,7 @@ static void ext4_end_io_dio(struct kiocb *iocb, loff_t offset, iocb->private = NULL; io_end->offset = offset; io_end->size = size; - if (is_async) { - io_end->iocb = iocb; - io_end->result = ret; - } - ext4_put_io_end_defer(io_end); + ext4_put_io_end(io_end); } /* @@ -3102,7 +3096,6 @@ static ssize_t ext4_ext_direct_IO(int rw, struct kiocb *iocb, ret = -ENOMEM; goto retake_lock; } - io_end->flag |= EXT4_IO_END_DIRECT; /* * Grab reference for DIO. Will be dropped in ext4_end_io_dio() */ @@ -3147,13 +3140,6 @@ static ssize_t ext4_ext_direct_IO(int rw, struct kiocb *iocb, if (ret <= 0 && ret != -EIOCBQUEUED && iocb->private) { WARN_ON(iocb->private != io_end); WARN_ON(io_end->flag & EXT4_IO_END_UNWRITTEN); - WARN_ON(io_end->iocb); - /* - * Generic code already did inode_dio_done() so we - * have to clear EXT4_IO_END_DIRECT to not do it for - * the second time. - */ - io_end->flag = 0; ext4_put_io_end(io_end); iocb->private = NULL; } diff --git a/fs/ext4/page-io.c b/fs/ext4/page-io.c index 6625d210fb4..d7d0c7b46ed 100644 --- a/fs/ext4/page-io.c +++ b/fs/ext4/page-io.c @@ -123,10 +123,6 @@ static void ext4_release_io_end(ext4_io_end_t *io_end) ext4_finish_bio(bio); bio_put(bio); } - if (io_end->flag & EXT4_IO_END_DIRECT) - inode_dio_done(io_end->inode); - if (io_end->iocb) - aio_complete(io_end->iocb, io_end->result, 0); kmem_cache_free(io_end_cachep, io_end); } @@ -204,19 +200,14 @@ static void ext4_add_complete_io(ext4_io_end_t *io_end) struct workqueue_struct *wq; unsigned long flags; - BUG_ON(!(io_end->flag & EXT4_IO_END_UNWRITTEN)); + /* Only reserved conversions from writeback should enter here */ + WARN_ON(!(io_end->flag & EXT4_IO_END_UNWRITTEN)); + WARN_ON(!io_end->handle); spin_lock_irqsave(&ei->i_completed_io_lock, flags); - if (io_end->handle) { - wq = EXT4_SB(io_end->inode->i_sb)->rsv_conversion_wq; - if (list_empty(&ei->i_rsv_conversion_list)) - queue_work(wq, &ei->i_rsv_conversion_work); - list_add_tail(&io_end->list, &ei->i_rsv_conversion_list); - } else { - wq = EXT4_SB(io_end->inode->i_sb)->unrsv_conversion_wq; - if (list_empty(&ei->i_unrsv_conversion_list)) - queue_work(wq, &ei->i_unrsv_conversion_work); - list_add_tail(&io_end->list, &ei->i_unrsv_conversion_list); - } + wq = EXT4_SB(io_end->inode->i_sb)->rsv_conversion_wq; + if (list_empty(&ei->i_rsv_conversion_list)) + queue_work(wq, &ei->i_rsv_conversion_work); + list_add_tail(&io_end->list, &ei->i_rsv_conversion_list); spin_unlock_irqrestore(&ei->i_completed_io_lock, flags); } @@ -256,13 +247,6 @@ void ext4_end_io_rsv_work(struct work_struct *work) ext4_do_flush_completed_IO(&ei->vfs_inode, &ei->i_rsv_conversion_list); } -void ext4_end_io_unrsv_work(struct work_struct *work) -{ - struct ext4_inode_info *ei = container_of(work, struct ext4_inode_info, - i_unrsv_conversion_work); - ext4_do_flush_completed_IO(&ei->vfs_inode, &ei->i_unrsv_conversion_list); -} - ext4_io_end_t *ext4_init_io_end(struct inode *inode, gfp_t flags) { ext4_io_end_t *io = kmem_cache_zalloc(io_end_cachep, flags); diff --git a/fs/ext4/super.c b/fs/ext4/super.c index b59373b625e..5db4f0df817 100644 --- a/fs/ext4/super.c +++ b/fs/ext4/super.c @@ -762,9 +762,7 @@ static void ext4_put_super(struct super_block *sb) ext4_unregister_li_request(sb); dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED); - flush_workqueue(sbi->unrsv_conversion_wq); flush_workqueue(sbi->rsv_conversion_wq); - destroy_workqueue(sbi->unrsv_conversion_wq); destroy_workqueue(sbi->rsv_conversion_wq); if (sbi->s_journal) { @@ -875,14 +873,12 @@ static struct inode *ext4_alloc_inode(struct super_block *sb) #endif ei->jinode = NULL; INIT_LIST_HEAD(&ei->i_rsv_conversion_list); - INIT_LIST_HEAD(&ei->i_unrsv_conversion_list); spin_lock_init(&ei->i_completed_io_lock); ei->i_sync_tid = 0; ei->i_datasync_tid = 0; atomic_set(&ei->i_ioend_count, 0); atomic_set(&ei->i_unwritten, 0); INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work); - INIT_WORK(&ei->i_unrsv_conversion_work, ext4_end_io_unrsv_work); return &ei->vfs_inode; } @@ -3954,14 +3950,6 @@ no_journal: goto failed_mount4; } - EXT4_SB(sb)->unrsv_conversion_wq = - alloc_workqueue("ext4-unrsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1); - if (!EXT4_SB(sb)->unrsv_conversion_wq) { - printk(KERN_ERR "EXT4-fs: failed to create workqueue\n"); - ret = -ENOMEM; - goto failed_mount4; - } - /* * The jbd2_journal_load will have done any necessary log recovery, * so we can safely mount the rest of the filesystem now. @@ -4115,8 +4103,6 @@ failed_mount4: ext4_msg(sb, KERN_ERR, "mount failed"); if (EXT4_SB(sb)->rsv_conversion_wq) destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq); - if (EXT4_SB(sb)->unrsv_conversion_wq) - destroy_workqueue(EXT4_SB(sb)->unrsv_conversion_wq); failed_mount_wq: if (sbi->s_journal) { jbd2_journal_destroy(sbi->s_journal); @@ -4564,7 +4550,6 @@ static int ext4_sync_fs(struct super_block *sb, int wait) trace_ext4_sync_fs(sb, wait); flush_workqueue(sbi->rsv_conversion_wq); - flush_workqueue(sbi->unrsv_conversion_wq); /* * Writeback quota in non-journalled quota case - journalled quota has * no dirty dquots @@ -4600,7 +4585,6 @@ static int ext4_sync_fs_nojournal(struct super_block *sb, int wait) trace_ext4_sync_fs(sb, wait); flush_workqueue(EXT4_SB(sb)->rsv_conversion_wq); - flush_workqueue(EXT4_SB(sb)->unrsv_conversion_wq); dquot_writeback_dquots(sb, -1); if (wait && test_opt(sb, BARRIER)) ret = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL); diff --git a/fs/ocfs2/aops.c b/fs/ocfs2/aops.c index 2abf97b2a59..94417a85ce6 100644 --- a/fs/ocfs2/aops.c +++ b/fs/ocfs2/aops.c @@ -565,9 +565,7 @@ bail: static void ocfs2_dio_end_io(struct kiocb *iocb, loff_t offset, ssize_t bytes, - void *private, - int ret, - bool is_async) + void *private) { struct inode *inode = file_inode(iocb->ki_filp); int level; @@ -592,10 +590,6 @@ static void ocfs2_dio_end_io(struct kiocb *iocb, level = ocfs2_iocb_rw_locked_level(iocb); ocfs2_rw_unlock(inode, level); - - inode_dio_done(inode); - if (is_async) - aio_complete(iocb, ret, 0); } /* diff --git a/fs/super.c b/fs/super.c index 68307c02922..5536a95186e 100644 --- a/fs/super.c +++ b/fs/super.c @@ -152,15 +152,9 @@ static struct super_block *alloc_super(struct file_system_type *type, int flags) static const struct super_operations default_op; if (s) { - if (security_sb_alloc(s)) { - /* - * We cannot call security_sb_free() without - * security_sb_alloc() succeeding. So bail out manually - */ - kfree(s); - s = NULL; - goto out; - } + if (security_sb_alloc(s)) + goto out_free_sb; + #ifdef CONFIG_SMP s->s_files = alloc_percpu(struct list_head); if (!s->s_files) @@ -228,6 +222,7 @@ err_out: free_percpu(s->s_files); #endif destroy_sb_writers(s); +out_free_sb: kfree(s); s = NULL; goto out; @@ -414,6 +409,11 @@ void generic_shutdown_super(struct super_block *sb) evict_inodes(sb); + if (sb->s_dio_done_wq) { + destroy_workqueue(sb->s_dio_done_wq); + sb->s_dio_done_wq = NULL; + } + if (sop->put_super) sop->put_super(sb); diff --git a/fs/xfs/xfs_aops.c b/fs/xfs/xfs_aops.c index 596ec71da00..e11d654af78 100644 --- a/fs/xfs/xfs_aops.c +++ b/fs/xfs/xfs_aops.c @@ -86,14 +86,6 @@ xfs_destroy_ioend( bh->b_end_io(bh, !ioend->io_error); } - if (ioend->io_iocb) { - inode_dio_done(ioend->io_inode); - if (ioend->io_isasync) { - aio_complete(ioend->io_iocb, ioend->io_error ? - ioend->io_error : ioend->io_result, 0); - } - } - mempool_free(ioend, xfs_ioend_pool); } @@ -281,7 +273,6 @@ xfs_alloc_ioend( * all the I/O from calling the completion routine too early. */ atomic_set(&ioend->io_remaining, 1); - ioend->io_isasync = 0; ioend->io_isdirect = 0; ioend->io_error = 0; ioend->io_list = NULL; @@ -291,8 +282,6 @@ xfs_alloc_ioend( ioend->io_buffer_tail = NULL; ioend->io_offset = 0; ioend->io_size = 0; - ioend->io_iocb = NULL; - ioend->io_result = 0; ioend->io_append_trans = NULL; INIT_WORK(&ioend->io_work, xfs_end_io); @@ -1292,8 +1281,10 @@ __xfs_get_blocks( if (create || !ISUNWRITTEN(&imap)) xfs_map_buffer(inode, bh_result, &imap, offset); if (create && ISUNWRITTEN(&imap)) { - if (direct) + if (direct) { bh_result->b_private = inode; + set_buffer_defer_completion(bh_result); + } set_buffer_unwritten(bh_result); } } @@ -1390,9 +1381,7 @@ xfs_end_io_direct_write( struct kiocb *iocb, loff_t offset, ssize_t size, - void *private, - int ret, - bool is_async) + void *private) { struct xfs_ioend *ioend = iocb->private; @@ -1414,17 +1403,10 @@ xfs_end_io_direct_write( ioend->io_offset = offset; ioend->io_size = size; - ioend->io_iocb = iocb; - ioend->io_result = ret; if (private && size > 0) ioend->io_type = XFS_IO_UNWRITTEN; - if (is_async) { - ioend->io_isasync = 1; - xfs_finish_ioend(ioend); - } else { - xfs_finish_ioend_sync(ioend); - } + xfs_finish_ioend_sync(ioend); } STATIC ssize_t diff --git a/fs/xfs/xfs_aops.h b/fs/xfs/xfs_aops.h index c325abb8d61..f94dd459dff 100644 --- a/fs/xfs/xfs_aops.h +++ b/fs/xfs/xfs_aops.h @@ -45,7 +45,6 @@ typedef struct xfs_ioend { unsigned int io_type; /* delalloc / unwritten */ int io_error; /* I/O error code */ atomic_t io_remaining; /* hold count */ - unsigned int io_isasync : 1; /* needs aio_complete */ unsigned int io_isdirect : 1;/* direct I/O */ struct inode *io_inode; /* file being written to */ struct buffer_head *io_buffer_head;/* buffer linked list head */ @@ -54,8 +53,6 @@ typedef struct xfs_ioend { xfs_off_t io_offset; /* offset in the file */ struct work_struct io_work; /* xfsdatad work queue */ struct xfs_trans *io_append_trans;/* xact. for size update */ - struct kiocb *io_iocb; - int io_result; } xfs_ioend_t; extern const struct address_space_operations xfs_address_space_operations; diff --git a/include/linux/buffer_head.h b/include/linux/buffer_head.h index 91fa9a94ae9..d77797a52b7 100644 --- a/include/linux/buffer_head.h +++ b/include/linux/buffer_head.h @@ -36,6 +36,7 @@ enum bh_state_bits { BH_Quiet, /* Buffer Error Prinks to be quiet */ BH_Meta, /* Buffer contains metadata */ BH_Prio, /* Buffer should be submitted with REQ_PRIO */ + BH_Defer_Completion, /* Defer AIO completion to workqueue */ BH_PrivateStart,/* not a state bit, but the first bit available * for private allocation by other entities @@ -128,6 +129,7 @@ BUFFER_FNS(Write_EIO, write_io_error) BUFFER_FNS(Unwritten, unwritten) BUFFER_FNS(Meta, meta) BUFFER_FNS(Prio, prio) +BUFFER_FNS(Defer_Completion, defer_completion) #define bh_offset(bh) ((unsigned long)(bh)->b_data & ~PAGE_MASK) diff --git a/include/linux/fs.h b/include/linux/fs.h index 65c9929cd34..c9013876eb2 100644 --- a/include/linux/fs.h +++ b/include/linux/fs.h @@ -46,6 +46,7 @@ struct vfsmount; struct cred; struct swap_info_struct; struct seq_file; +struct workqueue_struct; extern void __init inode_init(void); extern void __init inode_init_early(void); @@ -63,8 +64,7 @@ struct buffer_head; typedef int (get_block_t)(struct inode *inode, sector_t iblock, struct buffer_head *bh_result, int create); typedef void (dio_iodone_t)(struct kiocb *iocb, loff_t offset, - ssize_t bytes, void *private, int ret, - bool is_async); + ssize_t bytes, void *private); #define MAY_EXEC 0x00000001 #define MAY_WRITE 0x00000002 @@ -1328,6 +1328,9 @@ struct super_block { /* Being remounted read-only */ int s_readonly_remount; + + /* AIO completions deferred from interrupt context */ + struct workqueue_struct *s_dio_done_wq; }; /* superblock cache pruning functions */ -- cgit v1.2.3-70-g09d2