On Thu, Nov 11, 2004 at 12:02:48PM -0800, Matthew Dillon wrote:
>
> :> :[ adamk@sorrow - ~ ]: gnome-terminal
> :> :Fatal error 'gc cannot wait for a signal' at line ? in file /usr/src/lib/libc_r/uthread/uthread_gc.c (errno = ?)
> :> :Abort trap (core dumped)
> :> :
> :> :Same for mozilla, thunderbird, firefox, etc..
> :> :
> :> :I left my machine, brushed my teeth, came back, and suddenly it's working again. Any ideas what happened?
> :> :
> :> :Adam
> :>
> :>...
> :
> :I'm seeing this too when my laptop resumed from suspended state, and my
> :workaround is to adjust your system clock (with ntpdate for instance)
> :a few times. You'll get "Invalid argument" error from ntpdate, but just ignore
> :it and try a few times.
>
> Well, they say a fresh look always helps, and it looks like that
> is true. In about 5 seconds I found an overflow in the
> 'cputimer_freq64_nsec * delta' calculations in kern_clock.c.
>
> It turns out that cputimer_freq_nsec is such a large number that
> it doesn't take a delta much larger then cputimer_freq to overflow
> it, and the delta calculation can certainly exceed cputimer_freq
> due to all sorts of reasons... interrupt latency, blockages, and other
> things. laptop resumption only being one of them.
>
> So I would like both of you to please try this patch and tell me what
> happens.
>
> -Matt
> Matthew Dillon
> <dillon@backplane.com>
>
> Index: kern_clock.c
> ===================================================================
> RCS file: /cvs/src/sys/kern/kern_clock.c,v
> retrieving revision 1.24
> diff -u -r1.24 kern_clock.c
> --- kern_clock.c 17 Sep 2004 00:18:09 -0000 1.24
> +++ kern_clock.c 11 Nov 2004 19:53:13 -0000
> @@ -279,10 +279,14 @@
Matt, it looks to me like this is already committed as rev 1.25
as part of VFS stage 9/99. I'm going to build and install
new kernel anyway and see whatever happens.