Hi all, Here's a v2 of the patches to fix the buffer delwri queue state race problem (documented in the patch 2 commit log). This takes a completely different approach from v1, as suggested by Dave during review of v1. Patch 1 refactors the buffer submission paths and patch 2 closes the race by using synchronous buffer I/O for synchronous delwri queues. This survives a full xfstests run as well as lowmem stress testing of xfs/305 [1]. Thoughts, reviews, flames appreciated. Brian [1] If run long enough, I do eventually hit some presumably unrelated issues that are not currently reproducible simply due to the fact that the delwri queue issue is more prevalent. v2: - Implement sync buffer I/O for sync delwri queues instead of buffer wait list stealing. v1: https://marc.info/?l=linux-xfs&m=152837528705511&w=2 Brian Foster (2): xfs: refactor buffer submission into a common helper xfs: use sync buffer I/O for sync delwri queue submission fs/xfs/xfs_buf.c | 164 +++++++++++++++++++++++------------------------ 1 file changed, 79 insertions(+), 85 deletions(-) -- 2.17.1 -- To unsubscribe from this list: send the line "unsubscribe linux-xfs" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
