Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 24 Jan 2015 10:56:19 -0800
From:      Adrian Chadd <adrian@freebsd.org>
To:        Dmitry Sivachenko <trtrmitya@gmail.com>
Cc:        FreeBSD Net <freebsd-net@freebsd.org>
Subject:   Re: network locks up with udp traffic
Message-ID:  <CAJ-Vmon%2Bzd%2B-xURkEJbkzNVoEh%2BFNNseZPSjCJ7Xz23FjX68CA@mail.gmail.com>
In-Reply-To: <EAA00A09-DCA6-4ADF-8773-3C8A3609CCB6@gmail.com>
References:  <5B08402C-67A7-49E7-ADF8-390C94DCF1D7@gmail.com> <CAJ-Vmo=v5S=cRM3QPo0mGiNS8OS8so=915vx6R1fhDaKWS6BhQ@mail.gmail.com> <FCEBED19-25A4-4D8E-8EE5-8B2B397702C5@gmail.com> <CAJ-Vmokf9cC_26jF5B99_CRpjOaLOQUpBqMCcyqHevfi3JfFdQ@mail.gmail.com> <EAA00A09-DCA6-4ADF-8773-3C8A3609CCB6@gmail.com>

next in thread | previous in thread | raw e-mail | index | archive | help
Hi!

To be clear:

* is your kernel modified in any way; and
* did witness give you a full stacktrace as part of the lock order
reversal? All of that would be good.

Thanks,




On 24 January 2015 at 09:53, Dmitry Sivachenko <trtrmitya@gmail.com> wrote:
>
>> On 24 =D1=8F=D0=BD=D0=B2. 2015 =D0=B3., at 20:29, Adrian Chadd <adrian@f=
reebsd.org> wrote:
>>
>> Hi,
>>
>> Can you compile your kernel with WITNESS and repeat this test? It
>> looks like it's stuck in some lock ordering thing between some UDP
>> paths and an inpcb lookup.
>>
>> It should log in dmesg the first time a LOR occurs, regardless of
>> whether it actively causes a hang.
>
>
>
> Here is what I got in dmesg just before my net locked up:
>
> lock order reversal:
>  1st 0xffffffff80e79008 pcbinfohash (pcbinfohash) @ /opt/WRK/src/sys/neti=
net6/udp6_usrreq.c:1202
>  2nd 0xffffffff80e78fb0 udp (udp) @ /opt/WRK/src/sys/netinet6/in6_pcb.c:6=
14
> lock order reversal:
>  1st 0xffffffff80e79008 pcbinfohash (pcbinfohash) @ /opt/WRK/src/sys/neti=
net6/udp6_usrreq.c:1202
>  2nd 0xffffffff80e78d58 tcp (tcp) @ /opt/WRK/src/sys/netinet6/in6_pcb.c:6=
14
> lock order reversal:
>  1st 0xffffffff80e79008 pcbinfohash (pcbinfohash) @ /opt/WRK/src/sys/neti=
net6/udp6_usrreq.c:1202
>  2nd 0xffffffff80e781c0 rip (rip) @ /opt/WRK/src/sys/netinet6/in6_pcb.c:6=
14
>
> When I was saving this dmesg output to disk, I got another LOR:
>
> lock order reversal:
>  1st 0xfffffe0baf12fd78 bufwait (bufwait) @ /opt/WRK/src/sys/kern/vfs_bio=
.c:3065
>  2nd 0xfffff80011416c00 dirhash (dirhash) @ /opt/WRK/src/sys/ufs/ufs/ufs_=
dirhash.c:284
>
>
> Please tell me if I can provide more information to help tracking this do=
wn.
>
> Thanks!



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?CAJ-Vmon%2Bzd%2B-xURkEJbkzNVoEh%2BFNNseZPSjCJ7Xz23FjX68CA>