[PATCH] media: em28xx: fix a regression with HVR-950

[PATCH] media: em28xx: fix a regression with HVR-950

[Mauro Carvalho Chehab]

Changeset be7fd3c3a8c5 ("media: em28xx: Hauppauge DualHD second tuner
functionality") removed the logic with sets the alternate for the DVB
device. Without setting the right alternate, the device won't be
able to submit URBs, and userspace fails with -EMSGSIZE:

	ERROR     DMX_SET_PES_FILTER failed (PID = 0x2000): 90 Message too long

Tested with Hauppauge HVR-950 model A1C0.

Cc: Brad Love <brad@nextdimension.cc>
Fixes: be7fd3c3a8c5 ("media: em28xx: Hauppauge DualHD second tuner functionality")
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
---
 drivers/media/usb/em28xx/em28xx-dvb.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/media/usb/em28xx/em28xx-dvb.c b/drivers/media/usb/em28xx/em28xx-dvb.c
index a54cb8dc52c9..2ce7de1c7cce 100644
--- a/drivers/media/usb/em28xx/em28xx-dvb.c
+++ b/drivers/media/usb/em28xx/em28xx-dvb.c
@@ -199,6 +199,7 @@ static int em28xx_start_streaming(struct em28xx_dvb *dvb)
 	int rc;
 	struct em28xx_i2c_bus *i2c_bus = dvb->adapter.priv;
 	struct em28xx *dev = i2c_bus->dev;
+	struct usb_device *udev = interface_to_usbdev(dev->intf);
 	int dvb_max_packet_size, packet_multiplier, dvb_alt;
 
 	if (dev->dvb_xfer_bulk) {
@@ -217,6 +218,7 @@ static int em28xx_start_streaming(struct em28xx_dvb *dvb)
 		dvb_alt = dev->dvb_alt_isoc;
 	}
 
+	usb_set_interface(udev, dev->ifnum, dvb_alt);
 	rc = em28xx_set_mode(dev, EM28XX_DIGITAL_MODE);
 	if (rc < 0)
 		return rc;
-- 
2.14.3

Re: [PATCH] media: em28xx: fix a regression with HVR-950

[Brad Love]

Hi Mauro,

It turns out this patch breaks DualHD multiple tuner capability. When
alt mode is set in start_streaming it immediately kills the other tuners
stream. Essentially both tuners cannot be used together when this is
applied. I unfortunately don't have a HVR-950 to try and fix the
original regression better. Can you please take another look at this?
Until this is sorted, DualHD are effectively broken.

Cheers,

Brad




On 2018-03-09 06:21, Mauro Carvalho Chehab wrote:
> Changeset be7fd3c3a8c5 ("media: em28xx: Hauppauge DualHD second tuner
> functionality") removed the logic with sets the alternate for the DVB
> device. Without setting the right alternate, the device won't be
> able to submit URBs, and userspace fails with -EMSGSIZE:
>
> 	ERROR     DMX_SET_PES_FILTER failed (PID = 0x2000): 90 Message too long
>
> Tested with Hauppauge HVR-950 model A1C0.
>
> Cc: Brad Love <brad@nextdimension.cc>
> Fixes: be7fd3c3a8c5 ("media: em28xx: Hauppauge DualHD second tuner functionality")
> Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
> ---
>  drivers/media/usb/em28xx/em28xx-dvb.c | 2 ++
>  1 file changed, 2 insertions(+)
>
> diff --git a/drivers/media/usb/em28xx/em28xx-dvb.c b/drivers/media/usb/em28xx/em28xx-dvb.c
> index a54cb8dc52c9..2ce7de1c7cce 100644
> --- a/drivers/media/usb/em28xx/em28xx-dvb.c
> +++ b/drivers/media/usb/em28xx/em28xx-dvb.c
> @@ -199,6 +199,7 @@ static int em28xx_start_streaming(struct em28xx_dvb *dvb)
>  	int rc;
>  	struct em28xx_i2c_bus *i2c_bus = dvb->adapter.priv;
>  	struct em28xx *dev = i2c_bus->dev;
> +	struct usb_device *udev = interface_to_usbdev(dev->intf);
>  	int dvb_max_packet_size, packet_multiplier, dvb_alt;
>  
>  	if (dev->dvb_xfer_bulk) {
> @@ -217,6 +218,7 @@ static int em28xx_start_streaming(struct em28xx_dvb *dvb)
>  		dvb_alt = dev->dvb_alt_isoc;
>  	}
>  
> +	usb_set_interface(udev, dev->ifnum, dvb_alt);
>  	rc = em28xx_set_mode(dev, EM28XX_DIGITAL_MODE);
>  	if (rc < 0)
>  		return rc;

Re: [PATCH] media: em28xx: fix a regression with HVR-950

[Devin Heitmueller]

On Tue, Jun 26, 2018 at 4:09 PM, Brad Love <brad@nextdimension.cc> wrote:
> Hi Mauro,
>
> It turns out this patch breaks DualHD multiple tuner capability. When
> alt mode is set in start_streaming it immediately kills the other tuners
> stream. Essentially both tuners cannot be used together when this is
> applied. I unfortunately don't have a HVR-950 to try and fix the
> original regression better. Can you please take another look at this?
> Until this is sorted, DualHD are effectively broken.

Yeah, because the Empia device uses the same USB Interface for both
endpoints, you need to have a reference count and only change the
alternate when the first endpoint is put into use and then only reset
the alternate back to zero when neither endpoint is in use.

Devin

-- 
Devin J. Heitmueller - Kernel Labs
http://www.kernellabs.com