DragonFlyBSD

On Tue, Sep 14, 2010 at 02:46:30PM +0200, Francois Tigeot wrote:

The complete panic message is a bit long:
panic: assertion: (&proc_token)->t_ref && (&proc_token)->t_ref->tr_owner ==
curthread in setrunnable

This bug is still present.

I got a crash this morning with a 2.7.3.1084.g7b033-DEVELOPMENT kernel.
The machine had a 3 day uptime and was doing a buildworld.

Core files can be downloaded from http://dl.zefyris.com/

:On Tue, Sep 14, 2010 at 02:46:30PM +0200, Francois Tigeot wrote:
:> The complete panic message is a bit long:
:> panic: assertion: (&proc_token)->t_ref && (&proc_token)->t_ref->tr_owner ==
:> curthread in setrunnable
:
:This bug is still present.
:
:I got a crash this morning with a 2.7.3.1084.g7b033-DEVELOPMENT kernel.
:The machine had a 3 day uptime and was doing a buildworld.
:
:Core files can be downloaded from http://dl.zefyris.com/
:
:--
:Francois Tigeot

That URL doesn't resolve.  Address not found.

-Matt
                    Matthew Dillon 
                    <dillon@backplane.com>

:I have tested and retested the DNS settings and didn't find anything wrong;
:besides there are already Google crawlers acting on this URL.
:
:The only DNS breakage I have seen recently was due to fubar network components
:which broke EDNS replies and/or didn't understand IPv6 addresses.
:
:--
:Francois Tigeot

Ah, well, that's the problem then.  I have no IPV6 access.

-Matt
                    Matthew Dillon 
                    <dillon@backplane.com>

On Fri, Sep 24, 2010 at 09:42:47PM -0700, Matthew Dillon wrote:

:On Tue, Sep 14, 2010 at 02:46:30PM +0200, Francois Tigeot wrote:
:> The complete panic message is a bit long:
:> panic: assertion: (&proc_token)->t_ref && (&proc_token)->t_ref->tr_owner ==
:> curthread in setrunnable
:
:This bug is still present.
:
:I got a crash this morning with a 2.7.3.1084.g7b033-DEVELOPMENT kernel.
:The machine had a 3 day uptime and was doing a buildworld.
:
:Core files can be downloaded from http://dl.zefyris.com/
:
:--
:Francois Tigeot

That URL doesn't resolve. Address not found.

I have tested and retested the DNS settings and didn't find anything wrong;
besides there are already Google crawlers acting on this URL.

The only DNS breakage I have seen recently was due to fubar network components
which broke EDNS replies and/or didn't understand IPv6 addresses.

On Sat, 25 Sep 2010 08:56:32 -0700 (PDT)
Matthew Dillon <dillon@apollo.backplane.com> wrote:

:I have tested and retested the DNS settings and didn't find anything
wrong; :besides there are already Google crawlers acting on this URL.
:
:The only DNS breakage I have seen recently was due to fubar network
components :which broke EDNS replies and/or didn't understand IPv6
addresses. :
:--
:Francois Tigeot

Ah, well, that's the problem then. I have no IPV6 access.

Neither do I - but the site works for me.

Perhaps this will help

host dl.zefyris.com
dl.zefyris.com is an alias for serenal.zefyris.com.
serenal.zefyris.com has address 213.215.11.27
serenal.zefyris.com has IPv6 address 2001:7a8:600:1:219:d1ff:fe81:27

On Sat, Sep 25, 2010 at 05:54:47PM +0100, Steve O'Hara-Smith wrote:

On Sat, 25 Sep 2010 08:56:32 -0700 (PDT)
Matthew Dillon <dillon@apollo.backplane.com> wrote:

:I have tested and retested the DNS settings and didn't find anything
wrong; :besides there are already Google crawlers acting on this URL.
:
:The only DNS breakage I have seen recently was due to fubar network
components :which broke EDNS replies and/or didn't understand IPv6
addresses. :

Ah, well, that's the problem then. I have no IPV6 access.

Neither do I - but the site works for me.

In the end, it was due to a stupid error from myself: one of the name
servers had an invalid copy of the zefyris.com zone.

Sorry for that.

:This bug is still present.
:
:I got a crash this morning with a 2.7.3.1084.g7b033-DEVELOPMENT kernel.
:The machine had a 3 day uptime and was doing a buildworld.
:
:Core files can be downloaded from http://dl.zefyris.com/
:
:--
:Francois Tigeot

Very nice crash dump.  I think I found the case that can cause the
    situation, it is a very rare but nasty bug which can cause the proc_token
    to be unexpectedly released.

I have made a commit to address the issue (b528f10fca...).  Hopefully
    this will crush the bug.

-Matt
                    Matthew Dillon 
                    <dillon@backplane.com>

Status: Chatting->Testing.. Confirm fix?

Cheers,
Edward.

On Tue, Sep 28, 2010 at 03:18:53AM +0000, Edward O'Callaghan (via DragonFly issue tracker) wrote:

Edward O'Callaghan <eocallaghan@auroraux.org> added the comment:

Status: Chatting->Testing.. Confirm fix?

All my 2.7 machines now run with the fix.

Howewer, the crash could not easily be triggered in the first place; I think
we need a few more days just to be sure.

On Tue, Sep 28, 2010 at 03:18:53AM +0000, Edward O'Callaghan (via DragonFly issue tracker) wrote:

Edward O'Callaghan <eocallaghan@auroraux.org> added the comment:

Status: Chatting->Testing.. Confirm fix?

I have not seen any related crashes in a week.

I think this issue can be closed.

Status: Testing -> Resolved.

Cheers,
Edward.