Why we put FAN to full speed in its suspend callback

Why we put FAN to full speed in its suspend callback

[Aaron Lu]

Hi,

I was trying to minimize system suspend time and it turned out there are
firmwares that would do crazy things like delaying 10ms in FAN's power on
control method. So I wonder why do we want to put the FAN into ACPI D0
state when we are going to enter a sleep state? The code seems to be
written according to Pavel's comment here:
https://bugzilla.kernel.org/show_bug.cgi?id=5000#c6
It would be great to understand why and remove that operation if
possible, thanks for reading and taking your time bringing back your
ancient memory :-)

Re: Why we put FAN to full speed in its suspend callback

[Rafael J. Wysocki]

On Wednesday, January 22, 2014 05:04:54 PM Aaron Lu wrote:
> Hi,

Hi,

> I was trying to minimize system suspend time and it turned out there are
> firmwares that would do crazy things like delaying 10ms in FAN's power on
> control method. So I wonder why do we want to put the FAN into ACPI D0
> state when we are going to enter a sleep state? The code seems to be
> written according to Pavel's comment here:
> https://bugzilla.kernel.org/show_bug.cgi?id=5000#c6
> It would be great to understand why and remove that operation if
> possible, thanks for reading and taking your time bringing back your
> ancient memory :-)

Yes, this is for image creation code during hibernation and the same set of
callbacks was used for hibernation and suspend at that time.  I think we can
remove that thing for system suspend and leave it for hibernation only.

Thanks!

-- 
I speak only for myself.
Rafael J. Wysocki, Intel Open Source Technology Center.

Re: Why we put FAN to full speed in its suspend callback

[Aaron Lu]

Removed Konstantin since his email address is undeliverable.

On 01/23/2014 05:11 AM, Rafael J. Wysocki wrote:
> On Wednesday, January 22, 2014 05:04:54 PM Aaron Lu wrote:
>> Hi,
> 
> Hi,
> 
>> I was trying to minimize system suspend time and it turned out there are
>> firmwares that would do crazy things like delaying 10ms in FAN's power on
>> control method. So I wonder why do we want to put the FAN into ACPI D0
>> state when we are going to enter a sleep state? The code seems to be
>> written according to Pavel's comment here:
>> https://bugzilla.kernel.org/show_bug.cgi?id=5000#c6
>> It would be great to understand why and remove that operation if
>> possible, thanks for reading and taking your time bringing back your
>> ancient memory :-)
> 
> Yes, this is for image creation code during hibernation and the same set of
> callbacks was used for hibernation and suspend at that time.  I think we can
> remove that thing for system suspend and leave it for hibernation only.

OK, thanks a lot for the clarification, I'll prepare a patch to do that.