On 07/10/2013 01:52 PM, Christian Sünkenberg wrote:
> Hello,
> 
> On 05/01/2013 07:33 PM, H. Peter Anvin wrote:
>> On 05/01/2013 10:01 AM, Jonas Heinrich wrote:
>>> Hello, I tried the newest kernel, 3.9 today but the bug is still
>>> present. Applying the attached patch solves the bug for me.
>>>
>>> Best regards, Jonas Heinrich
>>
>> Okay... WTF is going on here?  Does pmode_behavior just not get set up
>> correctly?  Since it seems you can get it to wake up with your patch,
>> perhaps we can get read out the value of pmode_behavior and print it...
> 
> indeed, arch/x86/kernel/acpi/sleep.c tries an rdmsr_safe(MSR_EFER, ...)
> and sets WAKEUP_BEHAVIOR_RESTORE_EFER bit on success, however,
> on 90 nm Pentium M (Family 6, Model 13), reading an invalid MSR
> is not guaranteed to trap, see Erratum X4 in "Intel® Pentium® M
> Processor on 90 nm Process with 2-MB L2 Cache and Intel® Processor A100
> and A110 on 90 nm process with 512-KB L2 Cache Specification Update".
> On Jonas' T43, which has an affected Pentium M without EFER,
> rdmsr_safe(MSR_EFER, ...) succeeds and WAKEUP_BEHAVIOR_RESTORE_EFER
> gets set, while on resume the corresponding wrmsr traps and thus resume
> fails.
> 
> The pre-3.7 code snippet incidentally catched this by not restoring
> EFER when it would be restored to all 0s.
> 

That does seem like a reasonable explanation.

Does this patch fix the problem?  (Comment blatantly ripped off from
your email message.)

	-hpa