Hi Al, The following patches break the inode LRU operations and the first half of iput_final() out from under the inode_lock. I included the dispose_one_inode factoring patch to isolate the inode_lock from iput_final() completely. It's easy enough to drop if you don't want that right now. It passes xfstests on 1-, 2- and 8-way VMs, survives 8-way parallel create/traverse/unlink workloads with 0, 1 and 65536 byte files on XFS and ext4, and shows no problems with looping 50-client dbench runs on XFS or ext4. The patches should apply to your current merge-stem tree. -- To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
