Ceph goes to great lengths to keep its client-side cache coherent, allowing many operations (lookup, creations, readdir) to be performed without any server interaction when the client has the proper leases. Sadly, this functionality is all currently disabled because we cannot handle races between dcache pruning and any of those activities with the current VFS interface. This patch adds a d_prune hook that allows the filesystem to be informed before a dentry is removed from the cache. Merging this for 3.2-rc1 will make Ceph users and their metadata-intensive workload very happy. If anybody has any issues at all with this, please tell me, so I can make my case or revise my approach. Thanks! sage Sage Weil (1): vfs: add d_prune dentry operation Documentation/filesystems/Locking | 1 + fs/dcache.c | 8 ++++++++ include/linux/dcache.h | 3 +++ 3 files changed, 12 insertions(+), 0 deletions(-) -- 1.7.2.5 -- 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
