[Linux-ia64] [PATCH] fix timer interrupts getting lost

[Linux-ia64] [PATCH] fix timer interrupts getting lost

[Alex Williamson]

[-- Attachment #1: Type: text/plain, Size: 402 bytes --]


    This patch fixes the issue of some CPUs not showing timer interrupts
going off.  Seems during the process of sync'ing the itc, we jumped over
the next timer value.  This patch is against 2.5.67 + ia64.  I haven't
seen the problem on 2.4, but a quick looks seems like it's potentially
an issue there too.  Thanks,

	Alex

--
Alex Williamson                             HP Linux and Open Source Lab

[-- Attachment #2: smp_timer_fix.diff --]
[-- Type: text/plain, Size: 523 bytes --]

--- linux-2.5.67.clean/arch/ia64/kernel/smpboot.c	2003-04-22 13:58:09.000000000 -0600
+++ linux-2.5.67/arch/ia64/kernel/smpboot.c	2003-04-23 20:25:00.000000000 -0600
@@ -311,6 +311,15 @@
 		 */
 		Dprintk("Going to syncup ITC with BP.\n");
 		ia64_sync_itc(0);
+
+		/*
+		 * Make sure we didn't sync the itc ahead of the next
+		 * timer interrupt, if so, just reset it.
+		 */
+		if (time_after(ia64_get_itc(),local_cpu_data->itm_next)) {
+			Dprintk("oops, jumped a timer.\n");
+			ia64_cpu_local_tick();
+		}
 	}
 
 	/*

Re: [Linux-ia64] [PATCH] fix timer interrupts getting lost

[Takayoshi Kochi]

Thanks Alex!

This patch fixed the problem I reported before.
And I could observe some "oops jumped a timer." messages
(with debug flag enabled).

>     This patch fixes the issue of some CPUs not showing timer interrupts
> going off.  Seems during the process of sync'ing the itc, we jumped over
> the next timer value.  This patch is against 2.5.67 + ia64.  I haven't
> seen the problem on 2.4, but a quick looks seems like it's potentially
> an issue there too.  Thanks,

>> Hi,
>> 
>> Thanks to Alex's fixes for PCI segment support, we could
>> boot 2.5.67 kernel on Tiger.  But a strange thing is happening.
>> Only cpu0 & 1 are getting timer interrupts and cpu 2 & 3 aren't.
>> This problem is reproducible and we see this phenomenon after every
>> boot.  It seems that these cpus stop receiving timer
>> interrupts after a very short while.
>> Has anyone seen this?

Thanks,
---
Takayoshi Kochi <kochi@hpc.bs1.fc.nec.co.jp>

Re: [Linux-ia64] [PATCH] fix timer interrupts getting lost

[David Mosberger]

>>>>> On Wed, 23 Apr 2003 22:16:43 -0600, Alex Williamson <alex_williamson@attbi.com> said:

  Alex> This patch fixes the issue of some CPUs not showing timer
  Alex> interrupts going off.  Seems during the process of sync'ing
  Alex> the itc, we jumped over the next timer value.  This patch is
  Alex> against 2.5.67 + ia64.  I haven't seen the problem on 2.4, but
  Alex> a quick looks seems like it's potentially an issue there too.
  Alex> Thanks,

Good catch.  I applied this patch now.

The reason it started to show up on 2.5 now is mostly likely because I
had to move the ITC syncing to a point where interrupts are enabled.

	--david

Re: [Linux-ia64] [PATCH] fix timer interrupts getting lost

[Bjorn Helgaas]

On Wednesday 23 April 2003 10:16 pm, Alex Williamson wrote:
> 
>     This patch fixes the issue of some CPUs not showing timer interrupts
> going off.  Seems during the process of sync'ing the itc, we jumped over
> the next timer value.  This patch is against 2.5.67 + ia64.  I haven't
> seen the problem on 2.4, but a quick looks seems like it's potentially
> an issue there too.  Thanks,

I applied this for 2.4 also.  Thanks!