Re: [linux-uvc-devel] configfs: uvc and other gadget at the same time

[Gregoire G. <gre...@ge...>]

On 6/27/22 17:31, Gregoire Gentil wrote:
> Hello,
> 
> I have UVC gadget over configfs working well alone. I have also acm + 
> mtp working at the same time. But uvc doesn't work in a uvc + acm 
> configfs configuration (while acm works in that same configuration).
> 
> uvc alone gives:
> 
> [   13.544000] dwc2 c0040000.dwc2otg: new device is high-speed
> [   13.624000] dwc2 c0040000.dwc2otg: new device is high-speed
> [   13.696000] dwc2 c0040000.dwc2otg: new address 53
> [   13.720000] configfs-gadget gadget: high-speed config #1: c
> [   13.720000] configfs-gadget gadget: uvc_function_set_alt(0, 0)
> [   13.720000] configfs-gadget gadget: reset UVC Control
> [   13.720000] configfs-gadget gadget: uvc_function_set_alt(1, 0)
> [   13.720000] configfs-gadget gadget: uvc_function_set_alt(1, 0)
> 
> 
> while uvc + acm gives:
> [   13.784000] dwc2 c0040000.dwc2otg: new device is high-speed
> [   13.864000] dwc2 c0040000.dwc2otg: new device is high-speed
> [   13.932000] dwc2 c0040000.dwc2otg: new address 55
> [   13.956000] configfs-gadget gadget: high-speed config #1: c
> [   13.956000] configfs-gadget gadget: uvc_function_set_alt(2, 0)
> [   13.956000] configfs-gadget gadget: reset UVC Control
> [   13.956000] configfs-gadget gadget: uvc_function_set_alt(3, 0)
> [   13.956000] dwc2 c0040000.dwc2otg: dwc2_hsotg_ep_sethalt(ep 
> ffffffc01e1fe318 ep0, 1)
> [   13.956000] dwc2 c0040000.dwc2otg: dwc2_hsotg_ep_sethalt(ep 
> ffffffc01e1fe318 ep0, 1)
> [   13.956000] dwc2 c0040000.dwc2otg: dwc2_hsotg_ep_sethalt(ep 
> ffffffc01e1fe318 ep0, 1)
> [   13.956000] dwc2 c0040000.dwc2otg: dwc2_hsotg_ep_sethalt(ep 
> ffffffc01e1fe318 ep0, 1)
> [   13.956000] configfs-gadget gadget: uvc_function_set_alt(3, 0)
> [   13.960000] dwc2 c0040000.dwc2otg: dwc2_hsotg_ep_sethalt(ep 
> ffffffc01e1fe318 ep0, 1)
> [   13.960000] dwc2 c0040000.dwc2otg: dwc2_hsotg_ep_sethalt(ep 
> ffffffc01e1fe318 ep0, 1)
> 
> 
> And the host PC kernel gives an error when info is retrieved from the 
> OTG device.
> 
> Because of a lot other (much more complicated) constraints, I'm on an 
> old 4.4 kernel far away from the latest.
> 
> Is this a known bug? Could it be a hardware limitation of my OTG which 
> has 1 control endpoint and 16 device mode endpoints? It seems that acm 
> is handled before uvc, would it make sense to patch to swap how those 
> two are handled initially? Where should I look into in my kernel to 
> figure out my issue?
> 
> Any pointer or idea would be appreciated! Thanks,
> 
> Grégoire


Hello again,

It turns out that VIDIOC_DQEVENT gives 0x0102 for ((struct uvc_event 
*)&v4l2_event.u.data)->req.wIndex instead of 0x0100 in an uvc standalone 
case. All the userspace code is then screwed from this change.

Does anyone have any idea why DQEVENT gives a different wIndex value 
(+2) in case another gadget has been handled before uvc?

Grégoire