Re: [PATCH] usb: dwc3: gadget: Fail request submission if it was already queued

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

 



Hi,

On 1/11/2019 2:51 PM, Felipe Balbi wrote:
> Hi,
>
> Manu Gautam <mgautam@xxxxxxxxxxxxxx> writes:
>>> Manu Gautam <mgautam@xxxxxxxxxxxxxx> writes:
>>>> If a function driver tries to re-submit an already queued request,
>>>> it can results in corruption of pending/started request lists.
>>>> Catch such conditions and fail the request submission to DCD.
>>>>
>>>> Signed-off-by: Manu Gautam <mgautam@xxxxxxxxxxxxxx>
>>>> ---
>>>>  drivers/usb/dwc3/gadget.c | 6 ++++++
>>>>  1 file changed, 6 insertions(+)
>>>>
>>>> diff --git a/drivers/usb/dwc3/gadget.c b/drivers/usb/dwc3/gadget.c
>>>> index 679c12e14522..51716c6b286a 100644
>>>> --- a/drivers/usb/dwc3/gadget.c
>>>> +++ b/drivers/usb/dwc3/gadget.c
>>>> @@ -1290,6 +1290,12 @@ static int __dwc3_gadget_ep_queue(struct dwc3_ep *dep, struct dwc3_request *req)
>>>>  				&req->request, req->dep->name))
>>>>  		return -EINVAL;
>>>>  
>>>> +	if (req->request.status == -EINPROGRESS) {
>>> this test is really not enough. What if gadget driver set status to
>>> EINPROGRESS before submission? A better check would involve making sure
>>> req isn't part of dep->pending_list or dep->started_list or
>>> dep->cancelled_list. It's clear that this won't work very well as the
>>> amount of requests grow.
>> Thanks for quick review.
>> 'request.status' check can be replaced:
>> +if (!list_empty(&req->list) {
>>
>> And replace list_del with list_del_init from dwc3_gadget_giveback()
> I would rather avoid this. We could start tracking our own internal
> dwc3_request status. Something along the lines of:

Thanks for this quick patch.

[snip]

>  
> With this, we can remove some of the other request flags, such as "started".
>
>>> Anyway, which gadget driver did this? Why is it only affecting dwc3?
>> Function driver is not in upstream (f_mtp.c).
> So this could happen with any UDC, right? Why is f_mtp.c queueing the
> same request twice?


Looks like chipidea UDC already has such a check.
It is not yet clear that why f_mtp queued same request twice.
However, having a safeguard check in UDC should be helpful.


>
-- 
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project




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

  Powered by Linux