[Sound-open-firmware] SOF firmware and topology files
Hi,
I while back I was on the email list asking questions about backporting kernel fixes to Debian and getting the ucm files into Debian - and all of that has now been done. Many thanks for the pointers and the help. We almost have the Lenovo X1Carbon-7 audio working "out of the box" with Debian but there are still a few wrinkles and I'd appreciate any insight as to the best way to address these - and more importantly what I can do to help that happen :)
1) I currently add a blacklist file to stop the snd_hda_intel and snd_soc_skl drivers being loaded. I think based on some discussion and patches I saw in the alsa-devel thread you may have solved that issue - is that correct or is it still work in progress?
2) I've been using the sof-cnl-signed-intel.ri from the 1.3 release (and it's working great). Is there a plan how this will be handled downstream? For instance for Debian I didn't know whether it would become part of the intel-firmware package, or if maybe sof becomes its own package would be built by itself and the firmware grabbed from there, or does it go with some other alsa package? As an aside I tried the second approach (building it myself) but the resultant sof-cnl.ri fails to load succesfully. I haven't debugged it yet but are there any particular pointers on whether this should usually work or ideas why it wouldn't.
3) I've also got a sof-hda-generic.tplg file the origins of which I'm a bit unsure (I think I swiped it from Ubuntu as they have an image that works well on the X1C7). I tried building my own version but get ipc errors (I can share if of interest). Similarly to the sof-cnl.ri file - is there a plan/recommendations as to how this should make its way downstream?
Thanks in advance Mark Pearson
I while back I was on the email list asking questions about backporting kernel fixes to Debian and getting the ucm files into Debian - and all of that has now been done. Many thanks for the pointers and the help. We almost have the Lenovo X1Carbon-7 audio working "out of the box" with Debian but there are still a few wrinkles and I'd appreciate any insight as to the best way to address these - and more importantly what I can do to help that happen :)
- I currently add a blacklist file to stop the snd_hda_intel and snd_soc_skl drivers being loaded. I think based on some discussion and patches I saw in the alsa-devel thread you may have solved that issue - is that correct or is it still work in progress?
There is a first solution that should deal with the selection of SOF. I tested it but there are reports of additional issues on KabyLake w/ DRM-i915 stuff, so need to recheck if we did something wrong for SKL/KBL.
- I've been using the sof-cnl-signed-intel.ri from the 1.3 release (and it's working great). Is there a plan how this will be handled downstream? For instance for Debian I didn't know whether it would become part of the intel-firmware package, or if maybe sof becomes its own package would be built by itself and the firmware grabbed from there, or does it go with some other alsa package?
As an aside I tried the second approach (building it myself) but the resultant sof-cnl.ri fails to load succesfully. I haven't debugged it yet but are there any particular pointers on whether this should usually work or ideas why it wouldn't.
Yes, you cannot build your own firmware for a platform that uses the Intel production key, the firmware will not be authenticated for that platform and you will see a trace in the dmesg log.
We were planning to provide the binaries to the Linux firmware tree but it's taken a bit of a delay with travel/vacation/priorities. There is also a possible mismatch with firmware names that needs to be sorted out.
- I've also got a sof-hda-generic.tplg file the origins of which I'm a bit unsure (I think I swiped it from Ubuntu as they have an image that works well on the X1C7). I tried building my own version but get ipc errors (I can share if of interest). Similarly to the sof-cnl.ri file - is there a plan/recommendations as to how this should make its way downstream?
Our goal was to provide topology files and UCM files concurrently in a single repository that can be taken by downstream distros. I know Jaska is working on the UCM parts. Jaska please chime in, thanks!
participants (2)
-
Mark Pearson
-
Pierre-Louis Bossart