On Sat, Nov 5, 2011 at 6:46 AM, Chris Mason <[email protected]> wrote: > On Fri, Nov 04, 2011 at 08:47:56PM -0400, Chris Mason wrote: >> On Fri, Nov 04, 2011 at 05:18:54PM -0400, Josef Bacik wrote: >> > V1->V2: I stupidly thought I could get away with some flushing if we needed >> > space but I was wrong, we could deadlock, so add a btrfs_add_bytes_noflush >> > variant that will not do any flushing and will just return ENOSPC which >> > will let >> > us fallback and do our full flushing. >> > fs/btrfs/ctree.h | 3 +++ >> > fs/btrfs/delayed-inode.c | 36 ++++++++++++++++++++++++++++-------- >> > fs/btrfs/extent-tree.c | 18 ++++++++++++++++++ >> > 3 files changed, 49 insertions(+), 8 deletions(-) >> >> This helps but it I've got a bunch of these in the log now: >> >> btrfs_dirty_inode: 205 callbacks suppressed >> btrfs: fail to dirty inode 1294 error -28 >> btrfs: fail to dirty inode 1208 error -28 >> btrfs: fail to dirty inode 714 error -28 >> btrfs: fail to dirty inode 1772 error -28 >> btrfs: fail to dirty inode 1389 error -28 >> btrfs: fail to dirty inode 770 error -28 >> btrfs: fail to dirty inode 345 error -28 >> btrfs: fail to dirty inode 1533 error -28 >> btrfs: fail to dirty inode 624 error -28 >> btrfs: fail to dirty inode 1085 error -28 > > I should add this only seems to happen during xfstest 083. >
This version of the patch clears the deadlock issue I was seeing with version 1 of this patch. -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html
