Re: CAN-USB adapter unplug

[Martin Kelly <mkelly@xevo.com>]

On 11/29/2017 07:21 AM, Stefan Mätje wrote:
> Am 29.11.2017 um 13:20 schrieb Marc Kleine-Budde:
>> On 11/28/2017 10:09 PM, Martin Kelly wrote:
>>>> Both applied to can.
>>>
>>> Thanks! By the way as far as I can tell from code inspection, it appears
>>> that most of the other drivers in net/can/usb should have the same
>>> disconnect bug. gs_usb appears to be clear, as it returns in its default
>>> case. Unfortunately mcba_usb is the only device I have to test with, but
>>> those with other devices may want to check for this.
>>
>> Can you create patches for the affected drivers and send them to the
>> list and the maintainers of the driver on Cc?
>>
>> I don't have access to every USB adapter neither.
>>
>> Marc
>>
> 
> Hi,
> 
> I have seen Martin's emails these days and tried to reproduce the error here
> with an esd CAN-USB/2 device (handled by esd_usb2.c). The only thing I get
> in the log are some messages like this:
> 
> kernel: [14776.152459] usb 2-1: Rx URB aborted (-71)
> 
> There is no endless loop. How could I reproduce the bad behavior? For the
> quick test I used an Ubuntu 4.4.0-101-generic kernel.

Interesting. Do you see just a few -71 RX URB aborted messages (one per 
outstanding URB) rather than an endless loop? If so, then I think 
everything is OK on that device, as the URBs are not being resubmitted.

In case it helps, my test case for the mcba_usb is on a Raspberry Pi 3. 
I don't know whether or not that could influence the USB error code we 
see, since you are seeing EPROTO instead of EPIPE when the device gets 
unplugged.

> 
> In any case I will add the patch handling EPIPE on a test system and have a
> look what it might change.
> 
> Regards,
> Stefan Mätje
>