Revert "NFS: Make close(2) asynchronous when closing NFS O_DIRECT files"
authorTrond Myklebust <trond.myklebust@primarydata.com>
Fri, 4 Sep 2015 19:17:53 +0000 (15:17 -0400)
committerTrond Myklebust <trond.myklebust@primarydata.com>
Fri, 4 Sep 2015 20:54:29 +0000 (16:54 -0400)
commit4eae50143bcbfda819c650b7ed6739f3b6338ffc
tree61a335f9bcd639676317891587edda76c249fe2a
parent5cf9d70659594e1a75b34d18619d0bb6e0cbbafa
Revert "NFS: Make close(2) asynchronous when closing NFS O_DIRECT files"

This reverts commit f895c53f8ace3c3e49ebf9def90e63fc6d46d2bf.

This commit causes a NFSv4 regression in that close()+unlink() can end
up failing. The reason is that we no longer have a guarantee that the
CLOSE has completed on the server, meaning that the subsequent call to
REMOVE may fail with NFS4ERR_FILE_OPEN if the server implements Windows
unlink() semantics.

Reported-by: <Olga Kornievskaia <aglo@umich.edu>
Cc: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
fs/nfs/inode.c