Hi Takashi,
On 9/11/2023 10:57 AM, Wesley Cheng wrote:
Hi Takashi,
On 9/7/2023 8:36 AM, Takashi Iwai wrote:
On Tue, 29 Aug 2023 23:06:36 +0200, Wesley Cheng wrote:
Allow for different platforms to be notified on USB SND connect/disconnect seqeunces. This allows for platform USB SND modules to properly initialize and populate internal structures with references to the USB SND chip device.
Signed-off-by: Wesley Cheng quic_wcheng@quicinc.com
sound/usb/card.c | 45 +++++++++++++++++++++++++++++++++++++++++++++ sound/usb/card.h | 9 +++++++++ 2 files changed, 54 insertions(+)
diff --git a/sound/usb/card.c b/sound/usb/card.c index 1b2edc0fd2e9..067a1e82f4bf 100644 --- a/sound/usb/card.c +++ b/sound/usb/card.c @@ -118,6 +118,34 @@ MODULE_PARM_DESC(skip_validation, "Skip unit descriptor validation (default: no) static DEFINE_MUTEX(register_mutex); static struct snd_usb_audio *usb_chip[SNDRV_CARDS]; static struct usb_driver usb_audio_driver; +static struct snd_usb_platform_ops *platform_ops;
+int snd_usb_register_platform_ops(struct snd_usb_platform_ops *ops) +{ + int ret;
+ mutex_lock(®ister_mutex); + if (platform_ops) { + ret = -EEXIST; + goto out; + }
+ platform_ops = ops; +out: + mutex_unlock(®ister_mutex); + return 0; +} +EXPORT_SYMBOL_GPL(snd_usb_register_platform_ops);
For adding this kind of API, please give the proper comment. Especially this API is special and need a caution, to mention that it can be used only for a single instance.
Also, it should be mentioned that all callbacks are exclusive under the global register_mutex.
Thanks for taking the time to review. Sure, I'll add some comments in these new APIs to document what they are used for and how they are protected and limited.
@@ -910,7 +938,11 @@ static int usb_audio_probe(struct usb_interface *intf, chip->num_interfaces++; usb_set_intfdata(intf, chip); atomic_dec(&chip->active);
+ if (platform_ops && platform_ops->connect_cb) + platform_ops->connect_cb(chip); mutex_unlock(®ister_mutex);
One uncertain thing is the argument for connect_cb and disconnect_cb. Those take snd_usb_audio object, but the callback gets called per interface at each probe and disconnect. How does the callee handle multiple calls?
I guess it should depend on how the platform driver wants to handle it? I haven't run into a device with multiple UAC interfaces before, so I'll need to mimic this configuration on a device, so I can see how it exposes itself.
Will investigate this a bit more on my end and come back with my findings.
So looks like if there is a device that has multiple UAC interfaces, then it just results in one USB sound card with multiple USB streams. As of now, we do expose a ksndcontrol that allows for userspace to specify which card and PCM device to issue the audio playback on. Then based on the audio profile received by the audio DSP, it will narrow it down to see if that substream supports the requested profile.
This selection is done in the DPCM backend driver (q6usb), since the overall QMI USB stream request stems from the audio dsp, and will carry this information about which PCM device and card the USB offload driver should use. This may differ based on the implementation of the offload path, hence why I mentioned that it might be platform specific.
There is one improvement I need to make in our QC offload driver to accommodate for this scenario, and that is to avoid having to re-register for the XHCI sideband if there is already a USB SND card allocated, which I will fix and submit in the next revision.
Thanks Wesley Cheng