[PATCH 0/2] acpi: increase delays for s3 and s4 to make tests less aggressive

[Colin Ian King]

On 29/04/13 20:06, Alex Hung wrote:
> On 04/29/2013 11:59 PM, Colin Ian King wrote:
>> On 29/04/13 00:31, Alex Hung wrote:
>>> The default delays for s3 and s4 are too short and too aggressive, and
>>> many
>>> reported failures are fixed by simply increases the delays. If the
>>> delays
>>> are too short, the bottlenecks of the tests depend on hardware
>>> designs, ex.
>>> whether mainboards stablize fast enough when powers are removed, and OS
>>> behaviours, ex. whether drivers and services are loaded. Increasing
>>> the delays
>>> can reduce the dependency.
>>
>> The original idea was to make S3 + S4 tests aggressive to catch out
>> failures, so I think the defaults are sane because that the original
>> intention.
>
> Hi Colin,
>
> We received some feedbacks from IHV and OEM/ODM that increasing the
> delays resolve their problems of running fwts s3/s4 tests, and that's
> also exactly what they do to close the bugs. The idea "if nobody uses
> the defaults, it probably means it shouldn't be the defaults" pops up in
> my mind.

The defaults were specifically chosen to shake out bugs.  If we want to 
change them to ensure we get passes we may as well not be thoroughly 
testing.

These values were originally chosen because they exercised device 
drivers such as wireless association during suspend - and we did find a 
set of bugs with this that got fixed. That makes them legitimate in my book.

>
> I also think the defaults are too aggressive. For instance, HP's test
> suite sets the delay as 120 seconds between between wake-up and suspend,
> and currently it is 0 second in fwts. Dell/Toshiba/ASUS also does not
> set the delay to 0. If my memory does not betray me, the delay is not
> smaller than 120 seconds.
>
> I do agree that fwts should be aggressive, and that's the reason I chose
> the number to be smaller than OEM's tests.

>
> Cheers,
> Alex Hung
>
>>
>>>
>>> Alex Hung (2):
>>>    acpi: s3: increase delays to match user scanerio and to reduce false
>>>         non-firmware-related failures.
>>>    acpi: s4: increase delays to match user scanerio and to reduce false
>>>         non-firmware-related failures.
>>>
>>>   src/acpi/s3/s3.c |    6 +++---
>>>   src/acpi/s4/s4.c |    4 ++--
>>>   2 files changed, 5 insertions(+), 5 deletions(-)
>>>
>>
>>
>
>