[alsa-devel] [PATCH 4/5] ALSA: hda - drop def association and sequence from pinconf comparing

Hui Wang hui.wang at canonical.com
Tue May 27 08:55:38 CEST 2014


On 05/27/2014 02:53 PM, Takashi Iwai wrote:
> At Tue, 27 May 2014 08:47:33 +0200,
> David Henningsson wrote:
>>
>>
>> On 2014-05-27 08:40, Hui Wang wrote:
>>> On 05/27/2014 12:41 PM, David Henningsson wrote:
>>>>
>>>> On 2014-05-27 04:25, Alex Hung wrote:
>>>>> Hi David,
>>>>>
>>>>> BIOS today implements verbtable which is provided by codec vendor
>>>>> based on hardware design, and it is indeed not uncommon that the
>>>>> verbtable includes used pin only and leaves unused pins untouched.
>>>> Sure, but those unused pins would then have the same default value
>>>> that the codec initializes it with.
>>>>
>>>> Also, it wouldn't be uncommon for BIOS (or codec vendors) to use the
>>>> same verbtable for several machines if they share the same audio
>>>> hardware.
>>>>
>>>> But none of this explains why anyone would just change def association
>>>> and sequence value between machines? It makes no sense.
>>> So far, I met one example for this case:
>>>
>>> the Dell laptops with the same 0x10ec0255 codec,
>>>
>>> On some machines:
>>> pin 0x12, 0x90a60170
>>> pin 0x14, 0x90170120
>>> pin 0x21, 0x02211030
>>>
>>> On another machine:
>>> pin 0x12, 0x90a60140
>>> pin 0x14, 0x90170110
>>> pin 0x21, 0x02211020
>>>
>>>
>>> The def config of the rest pins are same.
>> If there is only one example (and only two different options), I think
>> we should revert this patch and use two different pin-matching quirks
>> instead.
>>
>> After all, ignoring the def assoc/sequence values also means a greater
>> risk of catching unwanted machines. Better err on the more careful side.
>>
>> This is IMO, what do others think?
> I'm for the safer side, too.
>
>
> Takashi
OK, agree to revert it.

Regards,
Hui.
>>>
>>> Regards,
>>> Hui.
>>>>> On Mon, May 26, 2014 at 6:11 PM, David Henningsson
>>>>> <david.henningsson at canonical.com> wrote:
>>>>>> (Add Alex Hung to CC)
>>>>>>
>>>>>> On 2014-05-26 10:22, Hui Wang wrote:
>>>>>>> A lot a machine have the same codec, but they have different default
>>>>>>> pinconf setting just because the def association and sequence is
>>>>>>> different, as a result they can't share a hda_pintbl[], to overcome
>>>>>>> it, we don't compare def association and sequence in the pinconf
>>>>>>> matching.
>>>>>>
>>>>>> Uhm, really? Alex, does this seem reasonable from a BIOS
>>>>>> perspective, i e,
>>>>>> that BIOS people normally would set def association and sequence
>>>>>> different
>>>>>> while leaving everything else unchanged?
>>>>>>
>>>>>>> Signed-off-by: Hui Wang <hui.wang at canonical.com>
>>>>>>> ---
>>>>>>>     sound/pci/hda/hda_auto_parser.c | 3 ++-
>>>>>>>     1 file changed, 2 insertions(+), 1 deletion(-)
>>>>>>>
>>>>>>> diff --git a/sound/pci/hda/hda_auto_parser.c
>>>>>>> b/sound/pci/hda/hda_auto_parser.c
>>>>>>> index b684c6e..3cf9137 100644
>>>>>>> --- a/sound/pci/hda/hda_auto_parser.c
>>>>>>> +++ b/sound/pci/hda/hda_auto_parser.c
>>>>>>> @@ -844,7 +844,8 @@ static bool pin_config_match(struct hda_codec
>>>>>>> *codec,
>>>>>>>     {
>>>>>>>           for (; pins->nid; pins++) {
>>>>>>>                   u32 def_conf = snd_hda_codec_get_pincfg(codec,
>>>>>>> pins->nid);
>>>>>>> -               if (pins->val != def_conf)
>>>>>>> +               u32 mask = 0xffffff00;
>>>>>>> +               if ((pins->val & mask) != (def_conf & mask))
>>>>>>>                           return false;
>>>>>>>           }
>>>>>>>           return true;
>>>>>>>
>>>>>> --
>>>>>> David Henningsson, Canonical Ltd.
>>>>>> https://launchpad.net/~diwic
>>>>>
>>>>>
>> -- 
>> David Henningsson, Canonical Ltd.
>> https://launchpad.net/~diwic
>>
>



More information about the Alsa-devel mailing list