Date: Mon, 19 Jul 2004 21:08:11 +0200 From: "Willem Jan Withagen" <wjw@withagen.nl> To: "John Baldwin" <jhb@FreeBSD.org>, <freebsd-current@FreeBSD.org> Cc: Robert Watson <rwatson@FreeBSD.org> Subject: Re: spin lock sched lock held by 0xffffff007b712250 for > 5 seconds Message-ID: <143b01c46dc3$bc7f0db0$471b3dd4@digiware.nl> References: <Pine.NEB.3.96L.1040716121933.4361A-100000@fledge.watson.org> <200406301130.43121.jhb@FreeBSD.org>
next in thread | previous in thread | raw e-mail | index | archive | help
From: "John Baldwin" <jhb@FreeBSD.org> > On Friday 16 July 2004 12:21 pm, Robert Watson wrote: > > On Fri, 16 Jul 2004, Willem Jan Withagen wrote: > > > After todays kernelbuild the system seem to be a lot better... > > > It can take quite some buildworld abuse, but still: > > > > > > spin lock sched lock held by 0xffffff007b712250 for > 5 seconds > > > panic: spin lock held too long > > > cpuid = 1; > > > KDB: enter: panic > > > > > > But I'm not shure what I could/should do now, since the KDB > > > introduction. Normally I'd expect to see: db> > > > > We have trouble entering the debugger when in a critical section/and or > > have sched_lock held -- I think this is because we try to halt the other > > CPUs and that gets nastily stuck in some form. We need to fix this. > > > > This could well be a symptom of some of the other hangs we've been seeing, > > and I've seen similar things on my test box with preemption enabled. > > You can hack sys/i386/include/smptests.h (or smptest.h, whichever it is) and > comment out CPUSTOP_ONDDBREAK as a hack. I did that recently for some > debugging. This is i386 only ..... I'm running in 84bit mode. But in all this is a nice suggestion, to see if the same thing will cause so many crashes when running in native 32bit-i386 mode. If it does the same there, would that mean a hardware problem?? Or is i386 much more stable? --WjW
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?143b01c46dc3$bc7f0db0$471b3dd4>