kern/168217: [bce] Watchdog timeouts with bce(4) on BCM5716

[Xin Li]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

On 06/04/12 17:32, YongHyeon PYUN wrote:
> On Tue, May 22, 2012 at 10:34:00AM -0700, Xin Li wrote:
>> -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA256
>> 
>> On 05/22/12 03:37, yongari at FreeBSD.org wrote:
>>> Your release information(i386) and environment(amd64) does not 
>>> match. Are you using i386 or PAE?
>> 
>> Sorry if there is mismatch then it must be I didn't cleared out 
>> freefall information properly.  The system is running
>> FreeBSD/amd64 and not i386 nor PAE.
>> 
>>> Having backtrace would be nice.
>> 
>> Will do.
>> 
>>> It seems you have two differnt controllers(5716 and 5709). Does
>>> the issue happen only on 5716?
>> 
>> We have not yet tested on the 5709 part as they are not
>> connected.
>> 
>> By the way there are some updates from yesterday.  Disabling
>> hdr_split did not helped and the system stops responding again;
>> disabling both hdr_split and tso seems to make the system survive
>> after 8 hours, we will continue to watch it and report back if
>> new information available.
>> 
>> Another thing to note is that we found that on systems that does
>> not exhibit the same problem, they have oui=0x50ef for the four
>> brgphy's, and on this system the four have oui=0xaf7 (brgphy0
>> pnpinfo oui=0xaf7 model=0x3c rev=0x8 at phyno=1).  Not sure if
>> this is related though.
>> 
> 
> Xin, I'm under the impression that bce_intr() might be called when 
> IFF_DRV_RUNNING is not set. Could you check whether 
> sc->bce_ifp->if_drv_flags set IFF_DRV_RUNNING in bce_intr()? Sorry,
> one of my box that can host quad-port bce(4) controller died so I
> can't verify this at the moment until I get a new MB.

I've made the following change:

Index: sys/dev/bce/if_bce.c
===================================================================
- --- sys/dev/bce/if_bce.c        (revision 233291)
+++ sys/dev/bce/if_bce.c        (working copy)
@@ -7678,6 +7678,9 @@ bce_intr(void *xsc)

        BCE_LOCK(sc);

+       if ((ifp->if_drv_flags & IFF_DRV_RUNNING) != IFF_DRV_RUNNING)
+               BCE_PRINTF("Got interrupt without IFF_DRV_RUNNING!\n");
+
        DBRUN(sc->interrupts_generated++);

        /* Synchnorize before we read from interface's status block */

Should we run with that and see if we can catch something?

Cheers,
- -- 
Xin LI <delphij at delphij.net>	https://www.delphij.net/
FreeBSD - The Power to Serve!		Live free or die
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.19 (FreeBSD)

iQEcBAEBCAAGBQJPzHZQAAoJEG80Jeu8UPuzBBAIAL1Q/Nt+e5okB+8EB/Ah7Gkm
+4OT+Y2vB/eW+vP8HgOKsbtE6dJ3w6saTUP0RGPekpeLfdvUVxSfL1lVOCvDPyic
8IoftFoHsSOanfisrTPTiHJet2m2fYOywHPx4tfAXvYvyQON48YHRRugDaQW6WPz
azzOVHY27J4Zf2UGqDrbTY5m8yWx8MCePJ3fsBKLcY6RDLHOVpE58ebW+LfNZ9RQ
fdZPkRTXcAcvZo8Cgv2F5lo6PdRv6C5MoWgk37rR24DZD3bVcXFCM7qRrkd86TqH
OY7mNjfC0QqTDgv/EKty//TtA8C0IO5iu0mRI71ONDKEEte73RiqMz1dC43IEnE=
=inpj
-----END PGP SIGNATURE-----