ext4: move_extent improve bh vanishing success factor

[ Upstream commit 88c6b61ff1cfb4013a3523227d91ad11b2892388 ]

Xiaoguang Wang has reported sporadic EBUSY failures of ext4/302
Unfortunetly there is nothing we can do if some other task holds BH's
refenrence.  So we must return EBUSY in this case.  But we can try
kicking the journal to see if the other task releases the bh reference
after the commit is complete.  Also decrease false positives by
properly checking for ENOSPC and retrying the allocation after kicking
the journal --- which is done by ext4_should_retry_alloc().

[ Modified by tytso to properly check for ENOSPC. ]

Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>

diff --git a/fs/ext4/move_extent.c b/fs/ext4/move_extent.c
index 9f2311bc9c4f3b5fbf2d839a7491da49c174cd4c..503ea15dc5dbe07b5ffe5ea891ea2b10a287012a 100644 (file)
--- a/fs/ext4/move_extent.c
+++ b/fs/ext4/move_extent.c
@@ -273,6 +273,7 @@ move_extent_per_page(struct file *o_filp, struct inode *donor_inode,
        int replaced_count = 0;
        int from = data_offset_in_page << orig_inode->i_blkbits;
        int blocks_per_page = PAGE_CACHE_SIZE >> orig_inode->i_blkbits;
+       struct super_block *sb = orig_inode->i_sb;
 
        /*
         * It needs twice the amount of ordinary journal buffers because
@@ -405,10 +406,13 @@ unlock_pages:
        page_cache_release(pagep[1]);
 stop_journal:
        ext4_journal_stop(handle);
+       if (*err == -ENOSPC &&
+           ext4_should_retry_alloc(sb, &retries))
+               goto again;
        /* Buffer was busy because probably is pinned to journal transaction,
         * force transaction commit may help to free it. */
-       if (*err == -EBUSY && ext4_should_retry_alloc(orig_inode->i_sb,
-                                                     &retries))
+       if (*err == -EBUSY && retries++ < 4 && EXT4_SB(sb)->s_journal &&
+           jbd2_journal_force_commit_nested(EXT4_SB(sb)->s_journal))
                goto again;
        return replaced_count;