In-Reply-To: <20080916175629.GE7187@lenovo>

Attached is the updated dmesg from adding the nmi_watchdog=2 and acpi=debug boot params.

-J

--
[ message continues ]

[j_kernel@hoblitt.com - Tue, Sep 16, 2008 at 04:43:12PM -1000]
| In-Reply-To: <20080916175629.GE7187@lenovo>
| On Tue, Sep 16, 2008 at 09:56:29PM +0400, Cyrill Gorcunov wrote:
| > [Thomas Gleixner - Tue, Sep 16, 2008 at 07:14:40AM -0700]
| > | Can you try nmi_watchdog=2 ?
| > | 
| > | Thanks,
| > | 
| > | 	tglx
| > | 
| > 
| > And get apic=debug a try too please. I remember there
| > was a problem with SB600 on ACPI side (but they should
| > be already fixed)
| 
| Attached is the updated dmesg from adding the nmi_watchdog=2 and acpi=debug boot params.
| 
| -J
| 

Joshua, could you please attach ACPI tables dump to the bugzilla entry (~300K)?
Not sure if it help but could be usefull info. Thanks for this dmesg - trying
to analize it.

(
here is how to do it
	http://kernel.org/pub/linux/kernel/people/helgaas/debug
	http://lwn.net/Articles/237085/
)

		- Cyrill -
--

[ message continues ]

Done but I was unable to extract the DSDT table as described in the
instructions.

#  acpixtract DSDT < acpidump.asc           
Could not open DSDT

Cheers,

-J

--
--

[ message continues ]

[j_kernel@hoblitt.com - Tue, Sep 16, 2008 at 11:20:58PM -1000]
| On Wed, Sep 17, 2008 at 11:13:57AM +0400, Cyrill Gorcunov wrote:
| > Joshua, could you please attach ACPI tables dump to the bugzilla entry (~300K)?
| 
| Done but I was unable to extract the DSDT table as described in the
| instructions.
| 
| #  acpixtract DSDT < acpidump.asc           
| Could not open DSDT
| 
| Cheers,
| 
| -J
| 

 it's enough, thanks!

		- Cyrill -
--

[ message continues ]

[j_kernel@hoblitt.com - Tue, Sep 16, 2008 at 11:20:58PM -1000]
| On Wed, Sep 17, 2008 at 11:13:57AM +0400, Cyrill Gorcunov wrote:
| > Joshua, could you please attach ACPI tables dump to the bugzilla entry (~300K)?
| 
| Done but I was unable to extract the DSDT table as described in the
| instructions.
| 
| #  acpixtract DSDT < acpidump.asc           
| Could not open DSDT
| 
| Cheers,
| 
| -J
| 
| --
| 

As far as I can see there is no really issue in ACPI configuration,
at least on apic side...

Evetually nmi_watchdog=2 does work right? Though it's not
good that we've got 8259 spurious interrupt. Hmm...

		- Cyrill -
--

[ message continues ]

well, nmi_watchdog=2 changes the layout of clockevent devices and easily 
switches the system into non-highres non-dynticks mode. So it can hide 
bugs.

	Ingo
--

[ message continues ]

[Ingo Molnar - Wed, Sep 17, 2008 at 03:13:06PM +0200]
| 
| > Evetually nmi_watchdog=2 does work right? Though it's not good that 
| > we've got 8259 spurious interrupt. Hmm...
| 
| well, nmi_watchdog=2 changes the layout of clockevent devices and easily 
| switches the system into non-highres non-dynticks mode. So it can hide 
| bugs.
| 
| 	Ingo
| 

and it does since we have nmi stuck on lapics. Still investigating.

		- Cyrill -
--

[ message continues ]

[j_kernel@hoblitt.com - Tue, Sep 16, 2008 at 11:20:58PM -1000]
| On Wed, Sep 17, 2008 at 11:13:57AM +0400, Cyrill Gorcunov wrote:
| > Joshua, could you please attach ACPI tables dump to the bugzilla entry (~300K)?
| 
| Done but I was unable to extract the DSDT table as described in the
| instructions.
| 
| #  acpixtract DSDT < acpidump.asc           
| Could not open DSDT
| 
| Cheers,
| 
| -J
| 
| --
| 

Joshua, could you please do one more thing - boot the kernel
with "debug apic=debug nmi_watchdog=1" (among others) and
publish dmsg?

		- Cyrill -
--

[ message continues ]

dmesg from "debug apic=debug nmi_watchdog=2" is attached.
                                          ^

-J

--
[ message continues ]

[j_kernel@hoblitt.com - Wed, Sep 17, 2008 at 10:48:50AM -1000]
| On Wed, Sep 17, 2008 at 08:27:04PM +0400, Cyrill Gorcunov wrote:
| > Joshua, could you please do one more thing - boot the kernel
| > with "debug apic=debug nmi_watchdog=1" (among others) and
| > publish dmsg?
| 
| dmesg from "debug apic=debug nmi_watchdog=2" is attached.
|                                           ^
| 
| -J
| 
| --
...

No Joshua, exactly 'nmi_watchdog=1' since it's important to
find who is responsible that cpu is stuck while testing nmi
watchdog. 'debug and apic=debug' will show additional info
what is happening. And it would be just great if you able
to try latest -tip/master

	http://people.redhat.com/mingo/tip.git/README

since there a lot of work/fix are done. If it's a problem
to fetch this kernel thru git - me or Ingo (I suppose)
could make tar.bz2 archive so you'll not need to fetch
the whole history. I wouldn't say I'm really specialist
in this area so your report could be quite important
not for me only but for others too.

		- Cyrill -
--

[ message continues ]