run resume code only for S1-S4 states

[Andriy Gapon]

on 18/04/2009 19:28 Nate Lawson said the following:
> Fabian Keil wrote:
>> Andriy Gapon <avg at freebsd.org> wrote:
>>
>>> An updated version of the patch, the only difference is: do-while(0) is gone,
>>> breaks are replaces with gotos, indentation is reduced.
>>>
>>> Per Nate's request I am calling for people with SMP systems to test if powering
>>> off via power button still works with this change. It's desirable to test power
>>> off at least two times to increase a chance of non-BSP CPU being used.
>> With an AMD Athlon(tm) 64 X2 Dual Core Processor 4600+ (2542.15-MHz K8-class CPU)
>> the first few shutdowns were successful, but on the fourth try pressing the
>> power button only lead to:
>>
>> Apr 18 12:52:42 kendra kernel: acpi: suspend request ignored (not ready yet)
>> Apr 18 12:52:42 kendra kernel: acpi: request to enter state S5 failed (err 6)
>> Apr 18 12:52:43 kendra kernel: acpi: suspend request ignored (not ready yet)
>> Apr 18 12:52:43 kendra kernel: acpi: request to enter state S5 failed (err 6)
>> Apr 18 12:52:43 kendra kernel: acpi: suspend request ignored (not ready yet)
>> Apr 18 12:52:43 kendra kernel: acpi: request to enter state S5 failed (err 6)
>> [...]
> 
> Yes, I think the case for S5 should probably come before
> acpi_sleep_disable().

Right now the patch tries to preserve the same behavior in this respect
as the current code has. I don't have a good understanding of
overlapping requests to enter different sleep states and potential bad
effects (e.g. S1 request while soft power off is already in progress).

But in this case I actually wonder what left ACPI driver is "sleep
disabled" state. Did the first soft poweroff attempt fail and caused
subsequent attempts to be disabled? Hmm, if so, then I wonder why it
could have failed.

-- 
Andriy Gapon