There are number of races exist caused by lack of synchronization between DIO workers in flight and truncate/fsync/punch_hole routines This patch series try to optimize and fix existing DIO/AIO code I observe significant performance improvements for big SMP hosts performind DIO for single file. Testcases which helps me to catch this type of bugs was posted here http://www.spinics.net/lists/linux-fsdevel/msg58312.html or available on my github repo: https://github.com/dmonakhov/xfstests/tree/ce8e3adab629b2a9be8ba2e73db7dad49eb46614 plese run 286,287,288 TOC: # first two are cleanups ext4: ext4_inode_info diet ext4: give i_aiodio_unwritten more appropriate name # Cleanup and bug fixes ext4: fix unwritten counter leakage ext4: completed_io locking cleanup V3 ext4: serialize dio nonlocked reads with defrag workers V3 ext4: punch_hole should wait for DIO writers ext4: serialize unlocked dio reads with truncate ext4: endless truncate due to nonlocked dio readers V2 ext4: serialize truncate with owerwrite DIO workers V2 ext4: fix ext_remove_space for punch_hole case Changes from V2: Fix use-after-free for queued end_io_work. fs/ext4/ext4.h | 33 ++++++++++--- fs/ext4/extents.c | 88 +++++++++++++++++++++++------------- fs/ext4/file.c | 9 --- fs/ext4/fsync.c | 85 ----------------------------------- fs/ext4/indirect.c | 19 ++++++- fs/ext4/inode.c | 49 ++++++++------------ fs/ext4/move_extent.c | 8 +++ fs/ext4/page-io.c | 121 +++++++++++++++++++++++++++++--------------------- fs/ext4/super.c | 3 - 10 files changed, 201 insertions(+), 216 deletions(-) -- To unsubscribe from this list: send the line "unsubscribe linux-ext4" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
