Re: [PATCH v2 3/7] usb: gadget: f_uvc: change endpoint allocation in uvc_function_bind()

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On 2023/12/21 00:02, Frank Li wrote:
>>> Some UDC driver use gadget_check_config() and match_ep() to allocate EP
>>> internal fifo memory resource, if only pass download full speed EP.
>> Could you share  the detail of problem ? do you mean find another different endpoint compared
> The problem is little bit complex. I try to use simple words.
>
> The background:
>
> Generally, UDC have some EP<0..15> and have some internal memory as FIFO.
> for example 16K.  You can simple assign EP<n> to 1K memory, which can hold
> whole package.
>
> But for UVC, some controller required internal FIFO hold whole frame data
>
> (mult+1) * (MaxBurst +1) * wPackageSize.
>
> For most case,  not every gadget use all 16 EPs. So you can assgin more
> memory into one EP, so it will reduce bus 'ping' package number and reduce
> NACK to improve transfer speed.
>
> The problem:
> pass fs_stream to udc driver, udc driver's check_config function will see
> mult and maxburst is 0. so only reserve 1K for ISO EP, but when try to 
> enable EP,  mult is 2, so there are not enough internal memory for it
> because more memory already assign to other EPs.
>
> Ideally, when gadget frame work can call check_config again when know
> usb speed, but it is not easy to fix it.
>
> Simple method use ss_stream_ep here and other function drviers. Super
> speed's package size is bigger than high/full speed. If resource can
> support super speed, it can support high/full speed.


I don't find any difference of uvc_ss_streaming_ep, uvc_hs_streaming_ep, uvc_fs_streaming_ep

descriptors. how difference happen in UDC ?


>
>
> /**
>  * gadget_is_superspeed() - return true if the hardware handles superspeed
>  * @g: controller that might support superspeed
>  */
>
> @max_speed: Highest speed the driver handles
>
> And according to gadget_is_superspeed() define, it indicate if udc
> controller support supersped, not link speed. 
>
> Orignial code is correct. If UDC support superspeed, then use ss_stream_ep.
>
> becasue superspeed is worse case compared as high and full speed.
>
> So I think original is correct.
>
> Frank.
>
>





[Index of Archives]     [Linux Media]     [Linux Input]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]     [Old Linux USB Devel Archive]

  Powered by Linux