Re: Ultrasparc III+ kernel panic

[bertrand <bertrand%systella.fr@localhost>]

Le 13.02.2015 12:09, BERTRAND Joël a écrit :
> Tobias Nygren a écrit :
> > On Sat, 07 Feb 2015 10:35:33 +0100
> > BERTRAND Joël <joel.bertrand%systella.fr@localhost> wrote:
> >
> > > cpu0: data fault: pc=14bdae8 rpc=0 addr=ffffffffffff8000
> >
> > This is in printf and addr = -32768. Obviously nothing is
> > mapped there, but ...
> >
> > Dump of assembler code for function printf:
> >     0x00000000014bdae0 <+0>:     save  %sp, -192, %sp
> >     0x00000000014bdae4 <+4>:     sethi  %hi(0x1c77800), %g1
> >     0x00000000014bdae8 <+8>:     stx  %i1, [ %fp + 0x887 ]
> >
> > The frame pointer looks wrecked, but it seems strange that
> > this memory access would resolve to -32768.
>
> New panic with :
> NetBSD legendre.systella.fr 7.99.4 NetBSD 7.99.4 (CUSTOM) #9: Thu Feb
> 5 18:00:30 CET 2015
> bertrand%legendre.systella.fr@localhost:/usr/obj/sys/arch/sparc64/compile/CUSTOM
> sparc64
>
> System panics during a cvs update -dP (on NetBSD's sources repository) 
> :
>
> Feb 13 11:58:48 legendre /netbsd: trap type 0x13e: cpu 0,
> pc=f000a8dctext_access_fault: pc=7fff754a8c21a008 va=7fff754a8c21a000
> Feb 13 11:58:48 legendre /netbsd: npc=f000a8e0 
> pstate=0x44820006<PRIV,IE>
> Feb 13 11:58:48 legendre /netbsd: Skipping crash dump on recursive 
> panic
> Feb 13 11:58:48 legendre /netbsd: panic: kernel fault
> Feb 13 11:58:48 legendre /netbsd: cpu1: Begin traceback...
> Feb 13 11:58:48 legendre /netbsd: cpu1: End traceback...
> Feb 13 11:58:48 legendre /netbsd: cpu0: shutting down
> Feb 13 11:58:48 legendre /netbsd: cpu1: rebooting
> Feb 13 11:58:48 legendre /netbsd:
>
> I'm uploading faulty kernel on 
> http://newton.systella.fr/~bertrand/netbsd.gdb
>
> Best regards,
>
> JKB

Other crash with same kernel :

Feb 14 01:28:31 legendre syslogd[463]: restart
Feb 14 01:28:31 legendre /netbsd: cpu0: data fault: pc=f001210c 
rpc=103053710 addr=24ec000
Feb 14 01:28:31 legendre /netbsd: text_access_fault: pc=5ae2dcd8 
va=5ae2c000
Feb 14 01:28:31 legendre /netbsd: Skipping crash dump on recursive panic
Feb 14 01:28:31 legendre /netbsd: panic: kernel fault
Feb 14 01:28:31 legendre /netbsd: cpu0: Begin traceback...
Feb 14 01:28:31 legendre /netbsd: cpu0: End traceback...
Feb 14 01:28:31 legendre /netbsd: cpu1: shutting down
Feb 14 01:28:31 legendre /netbsd: cpu0: rebooting
Feb 14 01:28:31 legendre /netbsd:

Please not that kernel panics just after syslogd is restarting. It's not 
the first time and I'm pretty sure that syslog and panic are correlated.

Best regards,

JKB