On 7 Nov 2016, at 8:09, Benjamin Coddington wrote:
On 4 Nov 2016, at 12:02, Benjamin Coddington wrote:
Hi Trond,
On 22 Sep 2016, at 13:39, Trond Myklebust wrote:
Right now, we're only running TEST/FREE_STATEID on the locks if
the open stateid recovery succeeds. The protocol requires us to
always do so.
The fix would be to move the call to TEST/FREE_STATEID and do it
before we attempt open recovery.
Signed-off-by: Trond Myklebust <trond.myklebust@xxxxxxxxxxxxxxx>
---
fs/nfs/nfs4proc.c | 92
+++++++++++++++++++++++++++++++------------------------
1 file changed, 52 insertions(+), 40 deletions(-)
diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c
index 3c1b8cb7dd95..33ca6d768bd2 100644
--- a/fs/nfs/nfs4proc.c
+++ b/fs/nfs/nfs4proc.c
@@ -2486,6 +2486,45 @@ static void
nfs41_check_delegation_stateid(struct nfs4_state *state)
}
/**
+ * nfs41_check_expired_locks - possibly free a lock stateid
+ *
+ * @state: NFSv4 state for an inode
+ *
+ * Returns NFS_OK if recovery for this stateid is now finished.
+ * Otherwise a negative NFS4ERR value is returned.
+ */
+static int nfs41_check_expired_locks(struct nfs4_state *state)
+{
+ int status, ret = NFS_OK;
+ struct nfs4_lock_state *lsp;
+ struct nfs_server *server = NFS_SERVER(state->inode);
+
+ if (!test_bit(LK_STATE_IN_USE, &state->flags))
+ goto out;
+ list_for_each_entry(lsp, &state->lock_states, ls_locks) {
+ if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
I bisected a crash to this patch (commit
c5896fc8622d57b31e1e98545d67d7089019e478).
I thought the problem was that this patch moved this path out from
under the
nfsi->rwsem in nfs4_reclaim_locks() so it ends up with a freed
nfs4_lock_state here.
I can reproduce this with generic/089. Any ideas?
Hit this on v4.9-rc4 this morning. This probably needs to take the
state_lock before traversing the lock_states list. I guess we've
never hit
this before because the old path would serialize things somehow -
maybe via
taking flc_lock in nfs4_reclaim_locks().. I'll test that fix.
Well, that's no good either as it gets stuck in a NFS4ERR_OLD_STATEID
loop
in recovery since we'd want to retry in that case, but taking the
state_lock
means we won't use the new stateid. So maybe we need both the
state_lock to
protect the list and the rwsem to stop new locks from being sent. I'll
try
that now.
Ben
--
To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
