[alsa-devel] Headset button mapping in the kernel
Hello ALSA Devs,
I am looking to get some feedback/ideas on a possible change to headset button mapping. Locally we are carrying patches that implement the mappings in the machine driver (which we understand you do not want upstream.) We are looking to see if we can add a new API (something like a sysfs path potentially) to have userspace pass in the mapping, if it chooses to, so the mapping can still be done in the kernel. That way we can carry just the config locally and remove some of the kernel patches we are carrying locally. Thanks.
Curtis
Hi!
Is this related to Android Wired Audio Headset Specification? Just wondering (I asked about it last month[1]).
Would be neat if it were possible to get headphone buttons working on Linux in general.
- Mads
On 25.11.2019 20:25, Curtis Malainey wrote:
Hello ALSA Devs,
I am looking to get some feedback/ideas on a possible change to headset button mapping. Locally we are carrying patches that implement the mappings in the machine driver (which we understand you do not want upstream.) We are looking to see if we can add a new API (something like a sysfs path potentially) to have userspace pass in the mapping, if it chooses to, so the mapping can still be done in the kernel. That way we can carry just the config locally and remove some of the kernel patches we are carrying locally. Thanks.
Curtis _______________________________________________ Alsa-devel mailing list Alsa-devel@alsa-project.org https://mailman.alsa-project.org/mailman/listinfo/alsa-devel
On Mon, 25 Nov 2019 20:25:53 +0100, Curtis Malainey wrote:
Hello ALSA Devs,
I am looking to get some feedback/ideas on a possible change to headset button mapping. Locally we are carrying patches that implement the mappings in the machine driver (which we understand you do not want upstream.) We are looking to see if we can add a new API (something like a sysfs path potentially) to have userspace pass in the mapping, if it chooses to, so the mapping can still be done in the kernel. That way we can carry just the config locally and remove some of the kernel patches we are carrying locally. Thanks.
I guess you're referring to the mapping that is currently done via snd_jack_set_key() calls? If so, an additional sysfs or such interface would be handy, and it should be relatively easy.
But a proper sysfs entry design is another question; it's an array, and this can be set for each jack object, so not that trivial.
thanks,
Takashi
On Wed, Nov 27, 2019 at 4:01 AM Takashi Iwai tiwai@suse.de wrote:
On Mon, 25 Nov 2019 20:25:53 +0100, Curtis Malainey wrote:
Hello ALSA Devs,
I am looking to get some feedback/ideas on a possible change to headset button mapping. Locally we are carrying patches that implement the mappings in the machine driver (which we understand you do not want upstream.) We are looking to see if we can add a new API (something like a sysfs path potentially) to have userspace pass in the mapping, if it chooses to, so the mapping can still be done in the kernel. That way we can carry just the config locally and remove some of the kernel patches we are carrying locally. Thanks.
I guess you're referring to the mapping that is currently done via snd_jack_set_key() calls? If so, an additional sysfs or such interface would be handy, and it should be relatively easy.
Yes this is what I was thinking.
But a proper sysfs entry design is another question; it's an array, and this can be set for each jack object, so not that trivial.
Agreed, it will likely take some work, but it is good to know you are open to the idea. I can start looking at some possible designs and share them here. Hopefully will have them in the next week or two.
thanks,
Takashi
On 25.11.2019 20:25, Curtis Malainey wrote:
Hello ALSA Devs,
I am looking to get some feedback/ideas on a possible change to headset button mapping. Locally we are carrying patches that implement the mappings in the machine driver (which we understand you do not want upstream.) We are looking to see if we can add a new API (something like a sysfs path potentially) to have userspace pass in the mapping, if it chooses to, so the mapping can still be done in the kernel. That way we can carry just the config locally and remove some of the kernel patches we are carrying locally. Thanks.
Curtis
Sorry for the top posting in my last mail.
I just wondered, do this have anything to do with headphones that has physical buttons on the headphone wire itself? E.g the Bose QC25 is a pair of headphones that has four buttons on the wire, and as far as I can see there's no way of getting those buttons to work in vanilla Linux for now, but it works in Android and Windows 10.
I asked about this on this mailing list before[1], because I don't even know which component should be responsible for generating button events. Should it have anything to do with alsa? Is the button mapping you're asking about here about the same thing? Do anyone know how one should go about supporting these kind of button events on desktop Linux?
- Mads
[1] https://mailman.alsa-project.org/pipermail/alsa-devel/2019-October/157702.ht...
On Thu, Nov 28, 2019 at 1:13 AM Mads Lønsethagen mads@ab3.no wrote:
On 25.11.2019 20:25, Curtis Malainey wrote:
Hello ALSA Devs,
I am looking to get some feedback/ideas on a possible change to headset button mapping. Locally we are carrying patches that implement the mappings in the machine driver (which we understand you do not want upstream.) We are looking to see if we can add a new API (something like a sysfs path potentially) to have userspace pass in the mapping, if it chooses to, so the mapping can still be done in the kernel. That way we can carry just the config locally and remove some of the kernel patches we are carrying locally. Thanks.
Curtis
Hi Mads,
Apologies, apparently my spam filter grabbed your email from me (back to adjusting the rules.)
Sorry for the top posting in my last mail.
I just wondered, do this have anything to do with headphones that has physical buttons on the headphone wire itself? E.g the Bose QC25 is a pair of headphones that has four buttons on the wire, and as far as I can see there's no way of getting those buttons to work in vanilla Linux for now, but it works in Android and Windows 10.
No this is related to ChromeOS device headset button mapping, but hopefully android will pick this up as well. Yes, these buttons are the ones I am discussing. Currently in ChromeOS (and likely in Android as well) we carry patches such as https://chromium-review.googlesource.com/c/chromiumos/third_party/kernel/+/1... It appears some have started landing upstream ae09a4783b9caf9307f303ef039f8297ce0371fe ("ASoC: Intel: Headset button support in kabylake machine driver") but it would be great if we had a way for userspace to configure these buttons similar to how we handle UCMs.
I asked about this on this mailing list before[1], because I don't even
know which component should be responsible for generating button events. Should it have anything to do with alsa? Is the button mapping you're asking about here about the same thing? Do anyone know how one should go about supporting these kind of button events on desktop Linux?
This project will have to be tied to ALSA in some fashion (as you can see
it is tied to the jack which is an ALSA concept), but I still have to do the design docs. In theory, this will enable vanilla linux to be configured for any headset buttons once done.
- Mads
[1]
https://mailman.alsa-project.org/pipermail/alsa-devel/2019-October/157702.ht...
Dne 03. 12. 19 v 21:43 Curtis Malainey napsal(a):
On Thu, Nov 28, 2019 at 1:13 AM Mads Lønsethagen mads@ab3.no wrote:
On 25.11.2019 20:25, Curtis Malainey wrote:
Hello ALSA Devs,
I am looking to get some feedback/ideas on a possible change to headset button mapping. Locally we are carrying patches that implement the mappings in the machine driver (which we understand you do not want upstream.) We are looking to see if we can add a new API (something like a sysfs path potentially) to have userspace pass in the mapping, if it chooses to, so the mapping can still be done in the kernel. That way we can carry just the config locally and remove some of the kernel patches we are carrying locally. Thanks.
Curtis
Hi Mads,
Apologies, apparently my spam filter grabbed your email from me (back to adjusting the rules.)
Sorry for the top posting in my last mail.
I just wondered, do this have anything to do with headphones that has physical buttons on the headphone wire itself? E.g the Bose QC25 is a pair of headphones that has four buttons on the wire, and as far as I can see there's no way of getting those buttons to work in vanilla Linux for now, but it works in Android and Windows 10.
No this is related to ChromeOS device headset button mapping, but hopefully android will pick this up as well. Yes, these buttons are the ones I am discussing. Currently in ChromeOS (and likely in Android as well) we carry patches such as https://chromium-review.googlesource.com/c/chromiumos/third_party/kernel/+/1... It appears some have started landing upstream ae09a4783b9caf9307f303ef039f8297ce0371fe ("ASoC: Intel: Headset button support in kabylake machine driver") but it would be great if we had a way for userspace to configure these buttons similar to how we handle UCMs.
The question why you need to change this settings in the user space. I think that the device tree was designed exactly to describe this hw platform specific settings. Another possibility is to use the kernel module option to configure the settings from the user space. But it's just an idea. You are probably looking for an interface which can be used when the driver is running.
Jaroslav
It appears some have started landing upstream ae09a4783b9caf9307f303ef039f8297ce0371fe ("ASoC: Intel: Headset button support in kabylake machine driver") but it would be great if we had a way for userspace to configure these buttons similar to how we handle UCMs.
The question why you need to change this settings in the user space. I think that the device tree was designed exactly to describe this hw platform specific settings. Another possibility is to use the kernel module option to configure the settings from the user space. But it's just an idea. You are probably looking for an interface which can be used when the driver is running.
I am also unclear on the ask. We've cleaned-up all machine drivers so that the mapping is identical, except for the cases where the codec inverts the buttons. Are you saying you do an additional remapping of those buttons in userpace? If yes, why not fix the machine driver? The mapping is typically based on measured impedance, not really something userspace should really know about. Or is this a case where the ChromeOS kernel has not yet seen the upstream patches?
On Tue, Dec 3, 2019 at 4:34 PM Pierre-Louis Bossart pierre-louis.bossart@linux.intel.com wrote:
It appears some have started landing upstream ae09a4783b9caf9307f303ef039f8297ce0371fe ("ASoC: Intel: Headset button support in kabylake machine driver") but it would be great if we had a way for userspace to configure these buttons similar to how we handle UCMs.
The question why you need to change this settings in the user space. I think that the device tree was designed exactly to describe this hw platform specific settings. Another possibility is to use the kernel module option to configure the settings from the user space. But it's just an idea. You are probably looking for an interface which can be used when the driver is running.
I am also unclear on the ask. We've cleaned-up all machine drivers so that the mapping is identical, except for the cases where the codec inverts the buttons. Are you saying you do an additional remapping of those buttons in userpace? If yes, why not fix the machine driver? The mapping is typically based on measured impedance, not really something userspace should really know about.
This is something we were under the impression upstream did not want. Dylan can you clarify our stance here? If we can just add the changes to the kernel then this would be a no-op.
Or is this a case where the ChromeOS kernel has not yet seen the upstream patches?
On 03.12.2019 21:43, Curtis Malainey wrote:
Hi Mads,
...
This project will have to be tied to ALSA in some fashion (as you can see it is tied to the jack which is an ALSA concept), but I still have to do the design docs. In theory, this will enable vanilla linux to be configured for any headset buttons once done.
Thank you for replying, it's nice to know about this :)
- Mads
participants (5)
-
Curtis Malainey -
Jaroslav Kysela -
Mads Lønsethagen -
Pierre-Louis Bossart -
Takashi Iwai