ext4: fix ext4_da_block_invalidatepages() to handle page range properly
authorCurt Wohlgemuth <curtw@google.com>
Sat, 26 Feb 2011 17:27:52 +0000 (12:27 -0500)
committerTheodore Ts'o <tytso@mit.edu>
Sat, 26 Feb 2011 17:27:52 +0000 (12:27 -0500)
commitc7f5938adce6727b9d17785f289c1146bd88d678
tree000d17633b6b3b28dd644704afb501b366ad1bbb
parente0fd9b90765f604374c42de8ac59d6584afce264
ext4: fix ext4_da_block_invalidatepages() to handle page range properly

If ext4_da_block_invalidatepages() is called because of a
failure from ext4_map_blocks() in mpage_da_map_and_submit(),
it's supposed to clean up -- including unlock -- all the
pages in the mpd structure.  But these values may not match
up, even on a system in which block size == page size:

   mpd->b_blocknr != mpd->first_page
   mpd->b_size != (mpd->next_page - mpd->first_page)

ext4_da_block_invalidatepages() has been using b_blocknr and
b_size; this patch changes it to use first_page and
next_page.

Tested:  I injected a small number (5%) of failures in
ext4_map_blocks() in the case that the flags contain
EXT4_GET_BLOCKS_DELALLOC_RESERVE, and ran fsstress on this
kernel.  Without this patch, I got hung tasks every time.
With this patch, I see no hangs in many runs of fsstress.

Signed-off-by: Curt Wohlgemuth <curtw@google.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
fs/ext4/inode.c