2005-12-14 23:07:24

by John Hawkes

Subject: [PATCH] ia64: eliminate softlockup warning

Fix an unnecessary softlockup watchdog warning in the ia64
uncached_build_memmap() that occurs occasionally at 256p and always at
512p. The problem occurs at boot time.

It would be good if we had a cleaner mechanism to temporarily silence
the watchdog thread, e.g.,
http://marc.theaimsgroup.com/?l=linux-kernel&m=111552476401175&w=2
but until that patch gets merged, this fix will have to suffice.

Signed-off-by: John Hawkes <[email protected]>

Index: linux/arch/ia64/kernel/uncached.c
===================================================================
--- linux.orig/arch/ia64/kernel/uncached.c 2005-12-06 15:12:14.000000000 -0800
+++ linux/arch/ia64/kernel/uncached.c 2005-12-14 14:50:55.000000000 -0800
@@ -210,6 +210,7 @@

dprintk(KERN_ERR "uncached_build_memmap(%lx %lx)\n", start, end);

+ touch_softlockup_watchdog();
memset((char *)start, 0, length);

node = paddr_to_nid(start - __IA64_UNCACHED_OFFSET);

2005-12-14 23:14:51

by Andrew Morton

Subject: Re: [PATCH] ia64: eliminate softlockup warning

[email protected] wrote:
>
> Fix an unnecessary softlockup watchdog warning in the ia64
> uncached_build_memmap() that occurs occasionally at 256p and always at
> 512p. The problem occurs at boot time.
>
> It would be good if we had a cleaner mechanism to temporarily silence
> the watchdog thread, e.g.,
> http://marc.theaimsgroup.com/?l=linux-kernel&m=111552476401175&w=2
> but until that patch gets merged, this fix will have to suffice.
>

That patch is for the hangcheck timer driver which is unrelated to the
softlockup detector.

>
> Index: linux/arch/ia64/kernel/uncached.c
> ===================================================================
> --- linux.orig/arch/ia64/kernel/uncached.c 2005-12-06 15:12:14.000000000 -0800
> +++ linux/arch/ia64/kernel/uncached.c 2005-12-14 14:50:55.000000000 -0800
> @@ -210,6 +210,7 @@
>
> dprintk(KERN_ERR "uncached_build_memmap(%lx %lx)\n", start, end);
>
> + touch_softlockup_watchdog();
> memset((char *)start, 0, length);
>
> node = paddr_to_nid(start - __IA64_UNCACHED_OFFSET);

Yes, additions like this are a bit awkward, but they're not _that_ bad. It
says "this code is doing something unusual". It's an explicit marker which
we can grep for and which conveys useful information about an exceptional
kernel state.