From: Takashi Iwai Subject: Re: HDA codec in embedded device Date: Thu, 31 May 2012 19:54:03 +0200 Message-ID: References: <20120531061302.GA11605@avionic-0098.adnet.avionic-design.de> <4FC7960D.6000405@wwwdotorg.org> <20120531165223.GA11723@avionic-0098.mockup.avionic-design.de> Mime-Version: 1.0 (generated by SEMI 1.14.6 - "Maruoka") Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: Received: from mx2.suse.de (cantor2.suse.de [195.135.220.15]) by alsa0.perex.cz (Postfix) with ESMTP id DF12F2432F for ; Thu, 31 May 2012 19:54:06 +0200 (CEST) In-Reply-To: <20120531165223.GA11723@avionic-0098.mockup.avionic-design.de> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: alsa-devel-bounces@alsa-project.org Errors-To: alsa-devel-bounces@alsa-project.org To: Thierry Reding Cc: alsa-devel@alsa-project.org, Stephen Warren List-Id: alsa-devel@alsa-project.org At Thu, 31 May 2012 18:52:23 +0200, Thierry Reding wrote: > > * Stephen Warren wrote: > > On 05/31/2012 12:13 AM, Thierry Reding wrote: > > > Hi, > > > > > > I work on an Atom-based platform that uses a Realtek ALC892 codec connected > > > via HDA. The current code in patch_realtek.c seems to assume a standard > > > desktop setup and maps the ports accordingly. The setup that I have is > > > entirely different and involves, among other things, to reconfigure some of > > > the ports typically used as outputs as inputs and vice versa. It'll also need > > > to provide a way to pass through some inputs on the outputs without going > > > through the DAC. > > > > > > All of these features are supported by the ALC982, but I don't see how this > > > could be implemented with the current driver. My question is how this could > > > possibly be solved in a way acceptable for mainline? > > > > The way I solved this was to add a machine-specific fixup rto > > alc269_fixups[] that set up the pin default registers as required by the > > HW prior to the parser interpreting those fields. I assume the same > > would work on the alc982. > > I've been discussing this with some people internally and we came to the > conclusion that for embedded setups the best option might be to expose all > possible controls and use something like the ALSA UCM to configure the > codec to cover the different muxing options required by the setup. Well, it's not clear to me why it must be so. Really a bit more detailed description about your pin setups would be helpful to understand better. > > I think in general as Takashi mentioned, the driver assumes that the pin > > default registers are already set up to values correct for the HW. > > Typically, I believe this is done by the SBIOS or audio card BIOS prior > > to the OS starting, or perhaps by EEPROM/straps in the HW. Is there a > > way that can be made to happen? > > Unfortunately we don't have access to the BIOS sources, and I believe there > are no options to set them externally. I think the BIOS doesn't even do > anything to those registers. You can pass the pin default configs and other setups via "patch" option of snd-hda-intel driver. See Documentation/sound/alsa/HD-Audio.txt. > One other interesting option would be to encode this in the device tree. I've > been investigating to use the device tree in order to describe carrier board > differences to allow a single kernel to boot the same processor module on > different carrier boards (the processor module could be different as well, of > course). > > This is obviously common practice on ARM but x86 is lacking a bit in this > respect. I currently have a setup where I can associated device nodes with > devices behind the PCI bus to configure I2C slaves and such. The same should > be possible with HDA and codec configuration. It's because HD-audio specification is designed to parse the pin default config instead of passing the whole topology statically. We've had static mapping in the past but this was a huge pain. Of course, there might be some cases that don't match with the standard parser. But this doesn't mean that we need to move again to the static mapping. I guess it'd suffice to have some additional codes to the codec auto-parser. Takashi