Skip to contentSkip to conversation

X86: Suppress hundreds of Intel thermal MCE messages on high temps

PATCH Ingo Molnar 1 participant 1 message Oct 23, 2009 1 message not in the archive

lore
Share & tools
lore
#1 lore raw
In reply to a message not in the archive
Quoted 20 lines · Thomas Renninger

This is against 2.6.31, but it should still patch, tell me if not. Should this also go to [email protected]?

Thanks,

Thomas

---------
This (introduced by Ingo some time ago):

   if (!(was_throttled ^ is_throttled) &&
        time_before64(tmp_jiffs, __get_cpu_var(next_check)))

does not help much.
The interrupts happening and the read out MSR seem to tell that
the CPU gets throttled and unthrottled all the time.
Thus current implementation results in hundreds (thousands?) of "above threshold" messages per minute.

This patch is more or less a revert back to the implementation in older kernels.
The message:

Quoted 7 lines
    printk(KERN_INFO "CPU%d: Temperature/speed normal\n", cpu);

is not needed as it is suppressed anyway.

Reference of an affected Core 2 and a Celeron machine:
http://bugzilla.novell.com/show_bug.cgi?id=533556

Have you seen these upstream commits:

b417c9f: x86: mce: Fix thermal throttling message storm 3967684: x86: mce: Clean up thermal throttling state tracking code

?

If they solve the problem for you then please forward them to [email protected].

Thanks,

Ingo