Date: Tue, 11 Sep 2001 08:47:33 +0200 (CEST) From: Michael Class <michaelc@tmbbobmc.bbn.hp.com> To: <Tor.Egge@fast.no> Cc: <current@freebsd.org> Subject: Re: boot() called on cpu #1 - hang Message-ID: <20010911084119.G809-100000@tmbbobmc.bbn.hp.com> In-Reply-To: <200109100456.GAA57477@midten.fast.no>
next in thread | previous in thread | raw e-mail | index | archive | help
On Mon, 10 Sep 2001 Tor.Egge@fast.no wrote:
> > Hello Tor,
> >
> > thank you for your quick response, unfortunately your patch did
> > not fix the problem.
>
>
> Your machine seems to hang too early for the patch to have any effect.
> (the patch affects a hang that occurs after the kernel has printed
> cpu reset called on cpu#1
> cpu_reset: Stopping other CPUs
> )
>
> > I have now tested a little bit more with the following sequence:
> >
> > boot machine to single-user
> > reboot
> >
> > I did this more then 10 times. It now got stuck every time
> > Approx. 8 time with
> >
> > boot() called on cpu #1
> > W
> >
> > And 3 times with
> >
> > boot() called on cpu #0
> > Wa
> >
> > or
> >
> > boot() called on cpu #0
> > Waiting (max
> >
> > It looks to me that the kernel-printf gets somehow stuck.
>
>
> Did you use -O2 when compiling the kernel ? That sometimes causes
> strange problems.
>
> The kernel doesn't appear do do much before printing the
>
> Waiting (max %d seconds) for system process `%s' to stop
>
> message in kproc_shutdown.
>
>
> boot() in /usr/src/sys/kern/kern_shutdown.c contains
>
> #ifdef SMP
> if (smp_active)
> printf("boot() called on cpu#%d\n", PCPU_GET(cpuid));
> #endif
> /*
> * Do any callouts that should be done BEFORE syncing the filesystems.
> */
> EVENTHANDLER_INVOKE(shutdown_pre_sync, howto);
>
>
> where the EVENTHANDLER_INVOKE macro expands to a lockmgr() call and
> invocation of the two events associated with shutdown_pre_sync:
>
> kproc_shutdown(bufdaemonproc, howto)
> kproc_shutdown(updateproc, howto)
>
> The normal output is
>
> Waiting (max 60 seconds) for system process `bufdaemon' to stop...stopped
> Waiting (max 60 seconds) for system process `syncer' to stop...stopped
>
> If the lockmgr lock for the event list is damaged, further damage
> elsewhere might occur due to the lockmgr call. If a debug printf
> before the lockmgr call in EVENTHANDLER_INVOKE() works while a debug
> printf after the lockmgr call isn't properly printed, then the
> probability for the problem being related to the lockmgr call is
> increased (cf. /usr/src/sys/sys/eventhandler.h)
>
> - Tor Egge
>
>
Hello Tor,
I have added a printf right before and after the lockmgr call in the
EVENTHANDLER_INVOKE() Macro in /usr/src/sys/sys/eventhandler.h.
But both of these printf do work! The output I am getting then is:
Boot() called on cpu #1
before lockmgr
after lockmgr
W
What else could I test?
Michael
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20010911084119.G809-100000>