Automatically switching snd_hda_intel.power_save value when switching from battery to ac ?

[Hans de Goede]

Hi,

On 3/6/20 10:03 AM, Takashi Iwai wrote:
> On Fri, 06 Mar 2020 09:25:49 +0100,
> Hans de Goede wrote:
>>
>> Hi,
>>
>> On 3/6/20 8:59 AM, Takashi Iwai wrote:
>>> On Thu, 05 Mar 2020 15:33:04 +0100,
>>> Hans de Goede wrote:
>>>>
>>>> Hi,
>>>>
>>>> Because of a bug-report about power-saving related plops/clocks on a
>>>> Lenovo T470s, I've asked inside Red Hat if people with a T470s and
>>>> running a recent kernel were also experiencing this.
>>>>
>>>> Most people are happy with the audio, but I did get a few bug reports
>>>> about plops on the headphones-jack.
>>>>
>>>> One of the suggestions which I got from 2 different users is to
>>>> disable power-saving for the HDA driver when on AC, esp. since most
>>>> headphones-jack use (esp. with an external amplifier which amplifies
>>>> the problem) is done while the laptop is sitting on a desk and thus
>>>> typically is connected to a charger.
>>>>
>>>> I'm personally not necessarily a fan of changing settings based
>>>> on being connected to ac or not, but I guess that in this case
>>>> it might not be such a bad idea ?
>>>
>>> Actually the power-saving-toggle-on-demand used to be the standard
>>> behavior by some power management tools including some
>>> thinkpad-specific one, IIRC.  So the behavior itself isn't too bad if
>>> the pop noise can't be fully eliminated.
>>
>> Right, atleast TLP does something like this, and maybe also some of
>> the Ubuntu pm scripts. But if we are going to do this I would like to
>> come up with some upstream, non 3th party (from a std gnu/linux
>> perspective), solution.
>>
>> Note upstream does not necessarily mean inside the kernel, it could be
>> e.g. part of systemd, e.g. a udev rule (not sure if that is possible)
>> or done by some pm suspend/resume hooks which are installed by default
>> (not sure if systemd's suspend code supports hooks though).
>>
>> If there is consensus that this is something which is a good idea /
>> a reasonable thing to do; and also consensus that this should be done
>> outside the kernel I can take a look at putting together a systemd
>> pull-req for this.
> 
> I have a mixed feeling about introducing it in standard.
> The power-saving itself is always good no matter whether it's cabled
> or not, it's merely a workaround for the problem that couldn't be
> fully fixed in kernel side.  So it was provided as an option for users
> who really need it.  But it's certainly a feasible option.

Ok, I have mixed feelings about this myself too, so I'll refrain
from trying to do a standard implementation of this then.

> OTOH, I still wonder about the cause of the noise on Thinkpad.  In
> general, if that comes from the codec power down (the Realtek one),
> it's usually fixable in a certain level.  Or if it's at the power down
> of the link (power_save_controller option), it's often hard and we
> keep the blacklist.

According to the reports which I have (I got quite a bit of feedback
when asking about T470s audio plops on an internal mailinglist) things
are fine for most users. The few users who do have issues are all hearing
the plops on the headphones. IIRC I had 2 clear reports about this and
maybe, do not remember. One of the clear reports deemed the plops as
audible but acceptable (using actual wired headphones) and one deemed
them unacceptable. The reporter who had a real issue with the plops on
the headphones was using an external amplifier hooked up to the
headphones jacks, amplifying the plops.

Since it is only on the headphones jack I think it is a codec issue,
if you can take a look at it that would be great, but otherwise I
believe that we can live with this (and the user with the external
amplifier can overrule the default value for the module option).

Regards,

Hans