diff mbox series

[2/2] ALSA: usb-audio: Avoid shoving a new stream into already registered device

Message ID 20200323170643.19181-2-tiwai@suse.de (mailing list archive)
State New, archived
Headers show
Series [1/2] ALSA: core: Add snd_device_get_state() helper | expand

Commit Message

Takashi Iwai March 23, 2020, 5:06 p.m. UTC
USB-audio driver registers the card and its devices at each probe of
USB interface, while trying to append a USB substream into the empty
PCM stream slot.  This works for most cases where the all PCM streams
are declared in the single interface description.  However, when the
device provides multiple individual interfaces, this may up with
pushing a new stream into the existing snd_pcm object that has been
already registered.  From the driver perspective, it's OK, but it
doesn't work for PulseAudio and others because they manage in the card
registration level, hence they'll miss this new device creation.

This patch tries to warn such a too-late-appended stream, and also
tries to put rather into a new snd_pcm object.

If we get a report from a user about this, we may add it to an entry
for snd_usb_registration_quirk().

Signed-off-by: Takashi Iwai <tiwai@suse.de>
---
 sound/usb/stream.c   | 9 +++++++++
 sound/usb/usbaudio.h | 1 +
 2 files changed, 10 insertions(+)

Comments

Takashi Iwai March 23, 2020, 8:25 p.m. UTC | #1
On Mon, 23 Mar 2020 18:06:43 +0100,
Takashi Iwai wrote:
> 
> USB-audio driver registers the card and its devices at each probe of
> USB interface, while trying to append a USB substream into the empty
> PCM stream slot.  This works for most cases where the all PCM streams
> are declared in the single interface description.  However, when the
> device provides multiple individual interfaces, this may up with
> pushing a new stream into the existing snd_pcm object that has been
> already registered.  From the driver perspective, it's OK, but it
> doesn't work for PulseAudio and others because they manage in the card
> registration level, hence they'll miss this new device creation.
> 
> This patch tries to warn such a too-late-appended stream, and also
> tries to put rather into a new snd_pcm object.
> 
> If we get a report from a user about this, we may add it to an entry
> for snd_usb_registration_quirk().
> 
> Signed-off-by: Takashi Iwai <tiwai@suse.de>

Let's scratch this.  It's better to be put together with a generic
module option that allows specifying the delayed registration for the
given device, and show the warning at the time of snd_card_register()
call instead.

Will respin the improved version later.


Takashi
diff mbox series

Patch

diff --git a/sound/usb/stream.c b/sound/usb/stream.c
index afd5aa574611..6c758af069e6 100644
--- a/sound/usb/stream.c
+++ b/sound/usb/stream.c
@@ -502,6 +502,15 @@  static int __snd_usb_add_audio_stream(struct snd_usb_audio *chip,
 		subs = &as->substream[stream];
 		if (subs->ep_num)
 			continue;
+		if (snd_device_get_state(chip->card, as->pcm) !=
+		    SNDRV_DEV_BUILD) {
+			if (!chip->pcm_devs_warned) {
+				usb_audio_warn(chip, "PCM stream already registered\n");
+				usb_audio_warn(chip, "Please report to upstream for assigning the delayed card registration\n");
+				chip->pcm_devs_warned = true;
+			}
+			continue;
+		}
 		err = snd_pcm_new_stream(as->pcm, stream, 1);
 		if (err < 0)
 			return err;
diff --git a/sound/usb/usbaudio.h b/sound/usb/usbaudio.h
index 6fe3ab582ec6..b2b693eeca91 100644
--- a/sound/usb/usbaudio.h
+++ b/sound/usb/usbaudio.h
@@ -34,6 +34,7 @@  struct snd_usb_audio {
 	unsigned int txfr_quirk:1; /* Subframe boundaries on transfers */
 	unsigned int tx_length_quirk:1; /* Put length specifier in transfers */
 	unsigned int setup_fmt_after_resume_quirk:1; /* setup the format to interface after resume */
+	unsigned int pcm_devs_warned:1; /* warned for delayed PCM registrations */
 	int num_interfaces;
 	int num_suspended_intf;
 	int sample_rate_read_error;