Re: [PATCH 1/2] em28xx: fix usb alternate setting for analog and digital video endpoints > 0

["Frank Schäfer" <fschaefer.oss@googlemail.com>]

Am 13.01.2014 20:13, schrieb Mauro Carvalho Chehab:
> Em Mon, 13 Jan 2014 19:29:59 +0100
> Frank Schäfer <fschaefer.oss@googlemail.com> escreveu:
>
>> On 12.01.2014 18:35, Mauro Carvalho Chehab wrote:
>>> Em Sat, 11 Jan 2014 14:42:29 +0100
>>> Frank Schäfer <fschaefer.oss@googlemail.com> escreveu:
>>>
>>>> The current code assumes that the analog + digital video endpoints are always at
>>>> interface number 0 when changing the alternate setting.
>>>> This seems to work fine for most existing devices.
>>>> However, at least the SpeedLink VAD Laplace webcam has the video endpoint on
>>>> interface number 3 (which fortunately doesn't cause any trouble because ist uses
>>>> bulk transfers only).
>>>> We already consider the actual the interface number for audio endpoints, so
>>>> rename the the audio_ifnum variable and use it for all device types.
>>>> Also get get rid of a pointless (ifnum < 0) in em28xx-audio.
>>>>
>>>> Signed-off-by: Frank Schäfer <fschaefer.oss@googlemail.com>
>>>> ---
>>>>   drivers/media/usb/em28xx/em28xx-audio.c |   10 +++++-----
>>>>   drivers/media/usb/em28xx/em28xx-cards.c |    2 +-
>>>>   drivers/media/usb/em28xx/em28xx-dvb.c   |    2 +-
>>>>   drivers/media/usb/em28xx/em28xx-video.c |    2 +-
>>>>   drivers/media/usb/em28xx/em28xx.h       |    3 +--
>>>>   5 Dateien geändert, 9 Zeilen hinzugefügt(+), 10 Zeilen entfernt(-)
>>>>
>>>> diff --git a/drivers/media/usb/em28xx/em28xx-audio.c b/drivers/media/usb/em28xx/em28xx-audio.c
>>>> index 30ee389..b2ae954 100644
>>>> --- a/drivers/media/usb/em28xx/em28xx-audio.c
>>>> +++ b/drivers/media/usb/em28xx/em28xx-audio.c
>>>> @@ -243,15 +243,15 @@ static int snd_em28xx_capture_open(struct snd_pcm_substream *substream)
>>>>   	}
>>>>   
>>>>   	runtime->hw = snd_em28xx_hw_capture;
>>>> -	if ((dev->alt == 0 || dev->audio_ifnum) && dev->adev.users == 0) {
>>>> -		if (dev->audio_ifnum)
>>>> +	if ((dev->alt == 0 || dev->ifnum) && dev->adev.users == 0) {
>>> Lets keep it named as "audio_ifnum". Ok, this is equal to the video_ifnum
>>> for several devices, but on em2861 (and other em2xx1 devices), the audio
>>> interface is different.
>> Uhm, that doesn't make sense.
>> Why do you want to call it audio_ifnum although it is used for 
>> audio/video/dvb ???
> em28xx-audio only cares about the interface used by audio, and if it is
> different than the one used by video, in order to switch the alternate at
> the right interface.
>
> So, you'll likely need to add a video_ifnum and an audio_ifnum, in
> order to track both (eventually, we may need a dvb_ifnum in the future,
> if they decide to provide a separate interface for it too).
>
>>> Also, as we're trying to get rid of hardcoded values, it also makes sense
>>> to store the alternate used for the audio endpoint.
>> We already to that: dev->alt.
> See below: dev->alt = 1 or dev->alt = 7.
>
>> I think there is some general misunderstanding here.
>> There is always only a single interface number and current alternate 
>> setting per driver instance.
>> If all endpoints are on the same interface, there's only a sibgle driver 
>> instance.
>> If the audio endpoints are on a separate interface, there will be two 
>> driver instances (each of them using their own device struct etc.).
>>
>>> Btw, I'm thinking on rework on this entire code, adding a logic that would
>>> handle properly the interface used by both audio and video (when this is
>>> the case) to not just select alt = 7, but to dynamically allocate the proper
>>> value for it, by taking into account the number of allocated audio URBs.
>> alt=7 is actually bug that needs to be fixed.
> Huh? alt = 7 is a "works everytime" kind of alternate, as it has the biggest
> window size. You can say that this is not optimized, making em28xx to spend
> more bandwidth than needed, but this is not a bug.
>
>> It messes up the alt settings made by the video part.
> Yes, because it doesn't recalculate the bandwidth.
>
>> This just didn't show up so far, because the video stream is almost 
>> always started after the audio stream.
> You can't assume that. 
>
> Actually, it is/was common for people to start the video streaming
> using some non-alsa compatible application, and then use a separate
> process for audio.
>
> See, for example (at ALSA audio with other applications):
> 	http://www.linuxtv.org/wiki/index.php/Saa7134-alsa
>
> I actually use this way when I want to do some tests with alsa,
> stopping/restarting its stream without touching on the video one.
>
>> Do you know any devices with audio endpoints 0x83 on the same interface 
>> as video/dvb ?
> Yes. HVR-950 (and HVR-900).
>
>> If yes, can you check their alt settings ? I'm pretty sure the audio 
>> endpoint uses the same wMaxPacketSize and bInterval values at all alt 
>> settings.
> Yes.
>
>> Otherwise things would become very complicated...
> The problem is basically the same: when a audio or video stream starts,
> it will need to call some routine to determinate the alternate. Such
> routine should track if both streams are active, or if just one is
> active, and provide the lowest alternate that works for both (or,
> in doubt, the alternate with the highest packet size, e. g. alt = 7
> when audio and video are at the same interface, or alt = 1 otherwise).
>
>>> For now, could you please rebase this patch, keeping the interface named
>>> as "audio_ifnum"?
>>>
>>> Thanks!
>>> Mauro
>>>
>>>> +		if (dev->ifnum)
>>>>   			dev->alt = 1;
>>>>   		else
>>>>   			dev->alt = 7;
>>>>   
>>>>   		dprintk("changing alternate number on interface %d to %d\n",
>>>> -			dev->audio_ifnum, dev->alt);
>>>> -		usb_set_interface(dev->udev, dev->audio_ifnum, dev->alt);
>>>> +			dev->ifnum, dev->alt);
>>>> +		usb_set_interface(dev->udev, dev->ifnum, dev->alt);
>>>>   
>>>>   		/* Sets volume, mute, etc */
>>>>   		dev->mute = 0;
>>>> @@ -625,7 +625,7 @@ static int em28xx_audio_init(struct em28xx *dev)
>>>>   	const int sb_size = EM28XX_NUM_AUDIO_PACKETS *
>>>>   			    EM28XX_AUDIO_MAX_PACKET_SIZE;
>>>>   
>>>> -	if (!dev->has_alsa_audio || dev->audio_ifnum < 0) {
>>>> +	if (!dev->has_alsa_audio) {
>>>>   		/* This device does not support the extension (in this case
>>>>   		   the device is expecting the snd-usb-audio module or
>>>>   		   doesn't have analog audio support at all) */
>>>> diff --git a/drivers/media/usb/em28xx/em28xx-cards.c b/drivers/media/usb/em28xx/em28xx-cards.c
>>>> index 39cf49c..b2cfd5d 100644
>>>> --- a/drivers/media/usb/em28xx/em28xx-cards.c
>>>> +++ b/drivers/media/usb/em28xx/em28xx-cards.c
>>>> @@ -3224,7 +3224,7 @@ static int em28xx_usb_probe(struct usb_interface *interface,
>>>>   	dev->has_alsa_audio = has_audio;
>>>>   	dev->audio_mode.has_audio = has_audio;
>>>>   	dev->has_video = has_video;
>>>> -	dev->audio_ifnum = ifnum;
>>>> +	dev->ifnum = ifnum;
>>>>   
>>>>   	/* Checks if audio is provided by some interface */
>>>>   	for (i = 0; i < udev->config->desc.bNumInterfaces; i++) {
>>>> diff --git a/drivers/media/usb/em28xx/em28xx-dvb.c b/drivers/media/usb/em28xx/em28xx-dvb.c
>>>> index 5c6be66..9d0fcc8 100644
>>>> --- a/drivers/media/usb/em28xx/em28xx-dvb.c
>>>> +++ b/drivers/media/usb/em28xx/em28xx-dvb.c
>>>> @@ -203,7 +203,7 @@ static int em28xx_start_streaming(struct em28xx_dvb *dvb)
>>>>   		dvb_alt = dev->dvb_alt_isoc;
>>>>   	}
>>>>   
>>>> -	usb_set_interface(dev->udev, 0, dvb_alt);
>>>> +	usb_set_interface(dev->udev, dev->ifnum, dvb_alt);
>>> In this case, it should be dev->video_ifnum, as otherwise this patch
>>> will break support for devices with em2861.
>> No. Audio and video are on the same interface.
>> If they are on separate interfaces, there will be be two driver 
>> instances running, each of them handling their own interface settings 
>> (interface number and used alt setting).
> Hmm... you may be right on that. 
>
> The endpoint descriptors allow to see both interfaces. I'm not
> sure how the current binding is doing, as there were several
> changes with regards to it.
>
> I'll need to double check it with the current version, and do some
> tests.
No need to hurry. We currently don't know any broken devices.
The whole audio stuff is damn fu..ing complicated and confusing.
Thinking about it twice is usually isn't enough. :/

>
>>>>   	rc = em28xx_set_mode(dev, EM28XX_DIGITAL_MODE);
>>>>   	if (rc < 0)
>>>>   		return rc;
>>>> diff --git a/drivers/media/usb/em28xx/em28xx-video.c b/drivers/media/usb/em28xx/em28xx-video.c
>>>> index 9c44628..b65d13a 100644
>>>> --- a/drivers/media/usb/em28xx/em28xx-video.c
>>>> +++ b/drivers/media/usb/em28xx/em28xx-video.c
>>>> @@ -382,7 +382,7 @@ set_alt:
>>>>   	}
>>>>   	em28xx_videodbg("setting alternate %d with wMaxPacketSize=%u\n",
>>>>   		       dev->alt, dev->max_pkt_size);
>>>> -	errCode = usb_set_interface(dev->udev, 0, dev->alt);
>>>> +	errCode = usb_set_interface(dev->udev, dev->ifnum, dev->alt);
>>> Same here.
>> Same here, it's correct.
>>
>>>>   	if (errCode < 0) {
>>>>   		em28xx_errdev("cannot change alternate number to %d (error=%i)\n",
>>>>   			      dev->alt, errCode);
>>>> diff --git a/drivers/media/usb/em28xx/em28xx.h b/drivers/media/usb/em28xx/em28xx.h
>>>> index efdf386..8d38d00 100644
>>>> --- a/drivers/media/usb/em28xx/em28xx.h
>>>> +++ b/drivers/media/usb/em28xx/em28xx.h
>>>> @@ -549,8 +549,6 @@ struct em28xx {
>>>>   	unsigned int has_alsa_audio:1;
>>>>   	unsigned int is_audio_only:1;
>>>>   
>>>> -	int audio_ifnum;
>>>> -
>>>>   	struct v4l2_device v4l2_dev;
>>>>   	struct v4l2_ctrl_handler ctrl_handler;
>>>>   	struct v4l2_clk *clk;
>>>> @@ -664,6 +662,7 @@ struct em28xx {
>>>>   
>>>>   	/* usb transfer */
>>>>   	struct usb_device *udev;	/* the usb device */
>>>> +	u8 ifnum;		/* number of the assigned usb interface */
>>>>   	u8 analog_ep_isoc;	/* address of isoc endpoint for analog */
>>>>   	u8 analog_ep_bulk;	/* address of bulk endpoint for analog */
>>>>   	u8 dvb_ep_isoc;		/* address of isoc endpoint for DVB */
>