Subject: Re: Advice on debugging pthreads problems
To: Jachym Holecek <freza@liberouter.org>
From: segv <segv@netctl.net>
List: port-sparc64
Date: 08/21/2005 02:05:48
On Sat, 20 Aug 2005 18:00:42 +0200
Jachym Holecek <freza@liberouter.org> wrote:


> I think those SIGBUSes at least come from firefox or some other packages
> it is using, not from the system. There was a number of reports about
> strict-aligned bugs in Gecko based browsers on tech-sparc64 in the past.
> 
> 	Regards,
> 		-- Jachym Holecek

Well some of the problems seem to originate from the kernel, i.e.:

Firefox or Galeon sometimes just hang, sending a plain kill signal has no
effect and when doing 'kill -9 <pid>' the CPU shoots up to 100% and the whole
system freezes. So just now I had Firefox hang on me, I exited window manager
and typed reboot, after which the kernel dropped into debugger:

# reboot
data fault: pc=1213e4c addr=0
ffb0: ffbfb_unblank
kernel trap 30: data access exception
Stopped in pid 17998.1 (reboot) at netbsd:lwp_continue+0x3c: ld %10 + 0x44], %g1
db> bt
kpsignal2(d9caea0, cdefcb0, 1, 0, 0, 0) at netbsd:kpsignal2+0x678
killpg1(c795d40, cdefcb0, d9caea0, d9caea0, cdefcb0, 0) at netbsd:killpg1+0x124
sys_kill(c795d40, cdefdd0, cdefdc0, cdefdb8, cdefde, cdefdb4) at netbsd:sys_kill
+0x164 syscall_plain(cdefed0, cdefdd0, 407368d8, 407368dc, 0, 0) at
netbsd:syscall_plain+0xc4 ?(ffffffffffffffff, 9, 101870, 5, 0, 0) at 0x1008ccc