[pulseaudio-discuss] [RFC] Dynamically created PCM devices for HDMI/DisplayPort

[Alexander E. Patrakov]

03.07.2015 22:26, Tanu Kaskinen wrote:
> On Thu, 2015-07-02 at 22:07 +0500, Alexander E. Patrakov wrote:
>> 02.07.2015 20:13, Tanu Kaskinen wrote:
>>> On Tue, 2015-06-30 at 22:29 +0500, Alexander E. Patrakov wrote:
>>>> 30.06.2015 20:08, Tanu Kaskinen wrote:
>>>>> Event: monitor gets unplugged
>>>>> -----------------------------
>>>>>
>>>>> When a monitor gets unplugged, module-udev-detect gets a notified, and
>>>>> it should figure out that a PCM device has disappeared. It should then
>>>>> call pa_alsa_card_pcm_removed(), which is a new function that needs to
>>>>> be implemented.
>>>>
>>>> Again, the case of manually loaded module-alsa-sink is not considered.
>>>
>>> For simplicity, I propose that no special handling is added to
>>> module-alsa-sink. If the module is loaded for a dynamic PCM that goes
>>> away, that presumably will cause the IO thread to fail. The module then
>>> unloads itself.
>>
>> Yes, that would work. OTOH, I think that the IO thread would notice the
>> error before module-udev-detect gets the notification. So I am not sure
>> whether any special handling of unplug events is needed in
>> module-udev-detect. It would be needed if we wanted to distinguish
>> between "card failed" and "card no longer exists" (like the kernel keeps
>> dead /dev/sd* devices for USB flash drives that were yanked while mounted).
>
> If a single sink fails due to unplugged monitor, the card as whole
> should not be considered "failed". Maybe you meant the mapping object
> corresponding to the PCM device, not the card object?

Yes.

> Anyway, I think we can't just rely on the IO thread failures, because
> the set of available profiles of a card has to be updated also if the
> unplugged monitor isn't currently in active use. If it's not in use,
> there won't be any IO thread that could report the removal event.

Of course, you are right!

-- 
Alexander E. Patrakov