Date: Thu, 16 Aug 2001 13:40:03 -0700 (PDT) From: Larry Rosenman <ler@lerctr.org> To: freebsd-bugs@FreeBSD.org Subject: Re: kern/26224 Message-ID: <200108162040.f7GKe3E60401@freefall.freebsd.org>
next in thread | raw e-mail | index | archive | help
The following reply was made to PR kern/26224; it has been noted by GNATS. From: Larry Rosenman <ler@lerctr.org> To: Thomas Moestl <tmoestl@gmx.net> Cc: freebsd-gnats-submit@freebsd.org Subject: Re: kern/26224 Date: Thu, 16 Aug 2001 15:34:51 -0500 * Thomas Moestl <tmoestl@gmx.net> [010816 12:59]: > On Tue, 2001/08/14 at 01:53:09 -0500, Larry Rosenman wrote: > > * Larry Rosenman <ler@lerctr.org> [010725 19:05]: > > > * Thomas Moestl <tmoestl@gmx.net> [010725 17:05]: > > > > I have attached a patch I want to post to -stable soon for people to > > > > test. I'm sure that it fixes this panic (the lock is just removed; > > > > this section is also protected by the mp_lock, so it should be > > > > safe). However, this is largely untested (I don't have an SMP box, and > > > > on UP simplelocks are disabled). So, be warned, I could have > > > > overlooked something, so other panics might surface... > > > > On the other hand, you could probably test this patch well, because > > > > your box seems rather stressed ;) > > > Ok, it's applied, and booted. We'll see. This one seems to hit me > > > every 3-4 weeks. > > Even WITH this fix in, I just had the lockup version hit me. > > Unfortunately, ctrl/break did *NOT* bring up DDB. The basic TCP stack > > was working (ping, telnet got to the connected to xxx part), but the > > box wouldn't give a login prompt nor serve www pages. > > Hmmm. Did that lockup version happen before? A certain deadlock should > also be fixed by the patch, but I'd guess that you are hitting a > different issue. Yes, we've seen the lockup before. > > > I don't know how to get the debug info. > > Dumb question: you do have ddb in your kernel, and crtrl-break is > wired to enter ddb in the key map you use? > It is unusual for the type of bug that I'm expecting that network > interrupts are still being processed while tty interrupts are not. YES. We just found that for SYSCONS, the DDB key is CTRL/PRTSC. Will try to get the DDB next time it fails (2-3 weeks :-( ). > > For this type of bug, it is not easy to get debug information, > especially if there is no easy way to trigger it. > One method to get into ddb anyway was to trigger an NMI via the ISA > bus, but this does not work with most newer main boards. > For SMP machines, however, Tor Egge has posted a patch to one of the > mailing lists (-current or -hackers, I think) that rewires the > interrupt of a serial port to the NMI using the APIC, that might help > you (although it probably should be used with care). We'll see what happens on next fail. Thanks! > > - thomas > > P.S.: sorry for the late answer, I have been kind of busy lately ;) No Problem, I understand. LER -- Larry Rosenman http://www.lerctr.org/~ler Phone: +1 972-414-9812 E-Mail: ler@lerctr.org US Mail: 1905 Steamboat Springs Drive, Garland, TX 75044-6749 To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-bugs" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200108162040.f7GKe3E60401>