Hi folks,
I would like an opinion about changing code in such as way that we
don't select a delegation stateid for an IO operation when this
particular delegation is being returned.
The reason it's some what problematic is that we send out a
DELEG_RETURN and then we don't remove the stateid until we receive a
reply. In the mean while, an IO operation can be happening and in
nfs4_select_rw_stateid() it sees a delegation stateid and uses it.
Well, at the server, it finishes processing DELEG_RETURN before
getting an IO op and by that time the server is finished with the
stateid and can error an IO operation with BAD_STATEID.
diff --git a/fs/nfs/delegation.c b/fs/nfs/delegation.c
index 7f3f606..4f6f6c9 100644
--- a/fs/nfs/delegation.c
+++ b/fs/nfs/delegation.c
@@ -854,7 +854,8 @@ bool nfs4_copy_delegation_stateid(nfs4_stateid
*dst, struct inode *in
flags &= FMODE_READ|FMODE_WRITE;
rcu_read_lock();
delegation = rcu_dereference(nfsi->delegation);
- ret = (delegation != NULL && (delegation->type & flags) == flags);
+ ret = (delegation != NULL && (delegation->type & flags) == flags &&
+ !test_bit(NFS_DELEGATION_RETURNING, &delegation->flags));
if (ret) {
nfs4_stateid_copy(dst, &delegation->stateid);
nfs_mark_delegation_referenced(delegation);
--
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
