Hi all,
Here's a new drop of the patches that tweak collapse to be fsb based and
organize the code a bit. This has seen ~1.5b fsx ops without a failure
and some light xfstests testing so far.
The primary difference from this version and the rfc is the retention of
the eofblocks trim prior to collapse. Patch 4 is also new. It isolates
the pagecache writeback/truncate in xfs_free_file_space() to the range
affected by the free. Thoughts, reviews, flames appreciated.
Brian
v1:
- Retain the eofblocks trim and writeback/inval. the range of shifted
data only.
- Added the xfs_free_file_space() patch to no longer writeback the
entire file.
rfc: http://oss.sgi.com/archives/xfs/2014-08/msg00462.html
Brian Foster (4):
xfs: track collapse via file offset rather than extent index
xfs: refactor xfs_bmap_shift_extents() into multiple functions
xfs: writeback and inval. file range to be shifted by collapse
xfs: only writeback and truncate pages for the freed range
fs/xfs/libxfs/xfs_bmap.c | 294 +++++++++++++++++++++++++++++++----------------
fs/xfs/libxfs/xfs_bmap.h | 7 +-
fs/xfs/xfs_bmap_util.c | 54 +++++----
3 files changed, 230 insertions(+), 125 deletions(-)
--
1.8.3.1
|