HEADS UP: snd_hda MFC plan

[Alexander Leidinger]

Quoting Alexander Motin <mav at FreeBSD.org> (from Tue, 06 Jan 2009  
11:25:01 +0200):

> Alexander Leidinger wrote:
>> Quoting Alexander Motin <mav at FreeBSD.org> (from Tue, 06 Jan 2009
>> 00:13:58 +0200):
>>
>>> On most systems having several devices is not a problem, as mostly
>>> used analog connectors are usually detected first. But on some systems
>>> (for example, having separate HDMI ports on video cards, which are now
>>> also supported and often detected first), user may be required to
>>> explicitly configure his audio applications or by using
>>> hw.snd.default_unit sysctl globally specify default audio device he
>>> wants to use. This moment explicitly specified in updated man page and
>>> I am going to specify it in UPDATING.
>>
>> Is there a way to move those HDMI ports to the end (either in pcmX or in
>> the probing) somehow? If yes, it would be more POLA to do this instead
>> of requiring the users to do something.
>
> HDMI ports usually have separate PCI HDA controllers. So order of pcm
> devices there defined by PCI probe order. Two driver instances know
> nothing about each other and I don't like the idea of obtaining such
> knowledge.

I agree... so we don't have a return value for the probe which  
basically tells that we want to see the probe called for this device  
again, after all other devices where probed? Would be useful here.

If this is not possible, I think we need something in the release  
notes about this. Maybe "The snd_hda driver now supports HDMI audio  
ports of graphic cards. This may result in additional audio devices  
after an update from 7.[01] and even replacing the previous default  
sound device. To change the default device in this case do ...."

>> And related: do all detected analog ports appear in a sensible order?
>> What I mean is again POLA related. If someone updates from 7.1 to 7.2,
>> will he be required to do something to get sound out of the same
>> connector as before when connecting to the default audio device? If not,
>> is it possible for you to introduce some corresponding sorting before
>> MFCing?
>
> Order of ports withing one codec defined by hardware vendor via codec
> configuration done by BIOS. That configuration supposed to be optimal
> for the specific system. Taking that previous driver ignored most of
> this information and was less functional, there sure will be some usage
> differences, but most configurations I have seen are quite reasonable to
> work just out of the box.

Should be mentioned in the release notes too. There's the possibility  
that an user get's a different default device, so we should be able to  
say "we told you so even in the release notes".

Bye,
Alexander.

-- 
But it does move!
		-- Galileo Galilei

http://www.Leidinger.net    Alexander @ Leidinger.net: PGP ID = B0063FE7
http://www.FreeBSD.org       netchild @ FreeBSD.org  : PGP ID = 72077137