From 53e872681fed6a43047e71bf927f77d06f467988 Mon Sep 17 00:00:00 2001 From: Jan Kara Date: Tue, 25 Dec 2012 13:29:52 -0500 Subject: ext4: fix deadlock in journal_unmap_buffer() We cannot wait for transaction commit in journal_unmap_buffer() because we hold page lock which ranks below transaction start. We solve the issue by bailing out of journal_unmap_buffer() and jbd2_journal_invalidatepage() with -EBUSY. Caller is then responsible for waiting for transaction commit to finish and try invalidation again. Since the issue can happen only for page stradding i_size, it is simple enough to manually call jbd2_journal_invalidatepage() for such page from ext4_setattr(), check the return value and wait if necessary. Signed-off-by: Jan Kara Signed-off-by: "Theodore Ts'o" --- fs/ext4/inode.c | 82 +++++++++++++++++++++++++++++++++++++++++++++++++-------- 1 file changed, 71 insertions(+), 11 deletions(-) (limited to 'fs/ext4') diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c index 12d3fbcff59f..cbfe13bf5b2a 100644 --- a/fs/ext4/inode.c +++ b/fs/ext4/inode.c @@ -2894,8 +2894,8 @@ static void ext4_invalidatepage(struct page *page, unsigned long offset) block_invalidatepage(page, offset); } -static void ext4_journalled_invalidatepage(struct page *page, - unsigned long offset) +static int __ext4_journalled_invalidatepage(struct page *page, + unsigned long offset) { journal_t *journal = EXT4_JOURNAL(page->mapping->host); @@ -2907,7 +2907,14 @@ static void ext4_journalled_invalidatepage(struct page *page, if (offset == 0) ClearPageChecked(page); - jbd2_journal_invalidatepage(journal, page, offset); + return jbd2_journal_invalidatepage(journal, page, offset); +} + +/* Wrapper for aops... */ +static void ext4_journalled_invalidatepage(struct page *page, + unsigned long offset) +{ + WARN_ON(__ext4_journalled_invalidatepage(page, offset) < 0); } static int ext4_releasepage(struct page *page, gfp_t wait) @@ -4313,6 +4320,47 @@ int ext4_write_inode(struct inode *inode, struct writeback_control *wbc) return err; } +/* + * In data=journal mode ext4_journalled_invalidatepage() may fail to invalidate + * buffers that are attached to a page stradding i_size and are undergoing + * commit. In that case we have to wait for commit to finish and try again. + */ +static void ext4_wait_for_tail_page_commit(struct inode *inode) +{ + struct page *page; + unsigned offset; + journal_t *journal = EXT4_SB(inode->i_sb)->s_journal; + tid_t commit_tid = 0; + int ret; + + offset = inode->i_size & (PAGE_CACHE_SIZE - 1); + /* + * All buffers in the last page remain valid? Then there's nothing to + * do. We do the check mainly to optimize the common PAGE_CACHE_SIZE == + * blocksize case + */ + if (offset > PAGE_CACHE_SIZE - (1 << inode->i_blkbits)) + return; + while (1) { + page = find_lock_page(inode->i_mapping, + inode->i_size >> PAGE_CACHE_SHIFT); + if (!page) + return; + ret = __ext4_journalled_invalidatepage(page, offset); + unlock_page(page); + page_cache_release(page); + if (ret != -EBUSY) + return; + commit_tid = 0; + read_lock(&journal->j_state_lock); + if (journal->j_committing_transaction) + commit_tid = journal->j_committing_transaction->t_tid; + read_unlock(&journal->j_state_lock); + if (commit_tid) + jbd2_log_wait_commit(journal, commit_tid); + } +} + /* * ext4_setattr() * @@ -4426,16 +4474,28 @@ int ext4_setattr(struct dentry *dentry, struct iattr *attr) } if (attr->ia_valid & ATTR_SIZE) { - if (attr->ia_size != i_size_read(inode)) { - truncate_setsize(inode, attr->ia_size); - /* Inode size will be reduced, wait for dio in flight. - * Temporarily disable dioread_nolock to prevent - * livelock. */ + 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) { - ext4_inode_block_unlocked_dio(inode); - inode_dio_wait(inode); - ext4_inode_resume_unlocked_dio(inode); + 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); } ext4_truncate(inode); } -- cgit v1.2.3