BCMK> I found the message in /fs/lockd/clntproc.c:
BCMK> switch (req->a_res.status) {
BCMK> case NLM_LCK_GRANTED:
BCMK> case NLM_LCK_DENIED_GRACE_PERIOD:
BCMK> /* Everything's good */
BCMK> break;
BCMK> case NLM_LCK_DENIED_NOLOCKS:
Interesting: The X/Open spec. says that the three return codes for
NLM_CANCEL are _GRANTED (success), _DENIED (failure), and
_DENIED_GRACE_PERIOD (call back later).
_DENIED_NOLOCKS is not listed as one of the possible (i.e. valid) return
codes, which makes sense when you think about it: cancels shouldn't fail
due to locking resources having been exhausted.
Does anyone know if this was an accidental deviation, or was it done for
some deliberate reason? Olaf? H.J.?
BCMK> So NLM_LCK_BLOCKED and NLM_LCK_DENIED are not being handled. (And
BCMK> I'm getting NLM_LCK_DENIED errors.) I checked the RFCs and it looks like
BCMK> we should just retry in this case. (It's supposed to be for temporary
BCMK> failures.) If so, I'm going to change the code to read:
BCMK> switch (req->a_res.status) {
BCMK> case NLM_LCK_GRANTED:
BCMK> case NLM_LCK_DENIED_GRACE_PERIOD:
BCMK> /* Everything's good */
BCMK> break;
BCMK> case NLM_LCK_DENIED_NOLOCKS:
BCMK> dprintk("lockd: CANCEL failed (server has no locks)\n");
BCMK> goto retry_cancel;
Arguably, that last case should be removed and picked up by the default:
printk notice (as an anomaly).
BCMK> + case NLM_LCK_DENIED:
BCMK> + dprintk("lockd: CANCEL failed (general error)\n");
BCMK> + goto retry_cancel;
And that should replace it.
--Up, inviting correction if he's misunderstanding something.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/