Re: CURRENT + amd64 + user-ppp = panic

From: John Baldwin <jhb_at_freebsd.org>
Date: Thu, 3 Nov 2005 09:49:00 -0500
On Thursday 03 November 2005 12:40 am, Victor Snezhko wrote:
> Vladimir Kushnir <vkushnir_at_i.kiev.ua> writes:
> > On Thu, 3 Nov 2005, Max Laier wrote:
> >>> Here it is: right before panic it prints
> >>>
> >>> For 0xffffff0017531100 -1 ticks
> >>> For 0xffffff0017531100 -1 ticks
> >>
> >> This results in two consecutive callout_stop() calls, but shouldn't hurt
> >> as callout_stop is protected against that.
> >>
> >> Do you get a dump for this?  Can you compare this pointer to the global
> >> llinfo_nd6 and see if "c" from the softclock() frame is related (or
> >> maybe the previous item in the list TAILQ).
> >
> > Sorry I'm not very goot at debugging. Would you please give me some
> > instructions (lamer's level, preferably :-)). And anyway I'll be able
> > to do it tomorrow only.
>
> In the meantime, I want to debug this too. I would insert a panic()
> call somewhere in the beginning of nd6_llinfo_settimer(), but there is
> a problem. nd6_llinfo_settimer() is called not only when I start ppp,
> but at bootup too (at bootup - once, I think that this is due to
> initializing lo0 interface, which has a default ipv6 address). Could I
> somehow make the kernel not panic at boot time and panic later, when I
> call ppp?

You can either add a sysctl that starts off as 0 and only panic if the sysctl 
is 1.  You can then boot, change the sysctl to 1, and then run ppp.  
Alternatively, you could use kdb_enter() to enter the debugger instead of 
panic().  You could then just continue using 'c' when you drop into ddb 
during boot.

-- 
John Baldwin <jhb_at_FreeBSD.org>  <><  http://www.FreeBSD.org/~jhb/
"Power Users Use the Power to Serve"  =  http://www.FreeBSD.org
Received on Thu Nov 03 2005 - 13:50:01 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:38:47 UTC