Bug #3996
closed
Solarflare NIC panic with LACP
Added by Chris Buechler over 9 years ago. Updated over 7 years ago.
0%
Description
Up to and including 2.2 are affected by the bug described here.
https://bugs.freenas.org/issues/4803
There is a trivial patch attached to that ticket which reportedly fixes this problem. Should also see why this hasn't made it into FreeBSD 10.1 given it's been in FreeNAS for 7 months.
Updated by Ermal Luçi over 9 years ago
- Status changed from New to Feedback
The patch mentioned here is already part of pfSense shiped sfxge driver.
Updated by Chris Buechler over 9 years ago
it wasn't as of 2 weeks ago and I don't see any relevant changes since then.
Updated by Chris Buechler over 9 years ago
- Status changed from Feedback to Confirmed
confirmed the described scenario is an issue, and I can't find that patch's contents anywhere
Updated by Jim Thompson over 9 years ago
- Priority changed from Normal to Low
If that "Solarflare patch" is the binary blob driver for sfxge, then we should yank it back out by the roots.
Updated by Jim Thompson over 9 years ago
Ermal is correct.
Check the contents of the patch against https://svnweb.freebsd.org/base/releng/10.1/sys/dev/sfxge/sfxge_port.c?revision=272461&view=markup#l318 (which is what is in FreeBSD 10.1-RELEASE)
diff r 8dc01b10eb64 sys/dev/sfxge/sfxge_port.c a/sys/dev/sfxge/sfxge_port.c Tue Apr 15 10:32:43 2014 0100
--
++ b/sys/dev/sfxge/sfxge_port.c Sat Apr 19 14:49:46 2014 +0400@ -357,10 +357,21 @
struct sfxge_port *port = &sc->port;
int rc;
- KASSERT);
mtx_lock(&port
- rc = sfxge_mac_filter_set_locked(sc);
+ /*
+ * The function may be called without softc_lock held in the
+ * case of SIOCADDMULTI and SIOCDELMULTI ioctls. ioctl handler
+ * checks IFF_DRV_RUNNING flag which implies port started, but
+ * it is not guaranteed to remain. softc_lock shared lock can't
+ * be held in the case of these ioctls processing, since it
+ * results in failure where kernel complains that non-sleepable
+ * lock is held in sleeping thread. Both problems are repeatable
+ * on LAG with LACP proto bring up.
+ */
+ if (port->init_state == SFXGE_PORT_STARTED)
+ rc = sfxge_mac_filter_set_locked(sc);
+ else
+ rc = 0;
mtx_unlock(&port->lock);
return rc;
}
318 int
318 sfxge_mac_filter_set(struct sfxge_softc sc)
319 {
320 struct sfxge_port *port = &sc->port;
321 int rc;
322
323 mtx_lock(&port->lock);
324 /
325 * The function may be called without softc_lock held in the
326 * case of SIOCADDMULTI and SIOCDELMULTI ioctls. ioctl handler
327 * checks IFF_DRV_RUNNING flag which implies port started, but
328 * it is not guaranteed to remain. softc_lock shared lock can't
329 * be held in the case of these ioctls processing, since it
330 * results in failure where kernel complains that non-sleepable
331 * lock is held in sleeping thread. Both problems are repeatable
332 * on LAG with LACP proto bring up.
333 */
334 if (port->init_state == SFXGE_PORT_STARTED)
335 rc = sfxge_mac_filter_set_locked(sc);
336 else
337 rc = 0;
338 mtx_unlock(&port->lock);
339 return rc;
340 }
See also: https://svnweb.freebsd.org/base?view=revision&revision=265884 which reads, in-part:
264772:
Check that port is started when MAC filter is setThe MAC filter set may be called without softc_lock held in the case of
SIOCADDMULTI and SIOCDELMULTI ioctls. The ioctl handler checks IFF_DRV_RUNNING
flag which implies port started, but it is not guaranteed to remain.
softc_lock shared lock can't be held in the case of these ioctls processing,
since it results in failure where kernel complains that non-sleepable
lock is held in sleeping thread.Both problems are repeatable on LAG with LACP proto bring up.Checked in by gnn on my birthday. Sun May 11 17:18:09 2014 UTC
I'm closing this as resolved.
Updated by Jim Thompson over 9 years ago
- Status changed from Confirmed to Rejected
- Priority changed from Low to Very Low
Updated by Chris Buechler over 9 years ago
- Status changed from Rejected to Feedback
- Assignee changed from Ermal Luçi to Chris Buechler
back to me for testing after discussion with Jim. I now have a Solarflare card to test.
Updated by Chris Buechler over 9 years ago
- Target version changed from 2.2 to 2.2.1
not something we'll be able to get fixed in 2.2. needs testing and reporting upstream, not something I can make a priority pre-2.2.
Updated by Chris Buechler over 9 years ago
- Target version changed from 2.2.1 to 2.2.2
Updated by Chris Buechler about 9 years ago
- Target version changed from 2.2.2 to 2.2.3
Updated by Chris Buechler almost 9 years ago
- Status changed from Feedback to Needs Patch
- Target version deleted (
2.2.3)
not hardware we sell, so not something we'll deal with. if someone wants to pursue, report and get it fixed in FreeBSD and we can add a patch to our builds
Updated by