Re: [PATCH 0/3 v5] can/usb: Add PEAK-System PCAN USB adapters driver

[Oliver Hartkopp <socketcan@hartkopp.net>]

On 22.02.2012 18:03, Stephane Grosjean wrote:

> Le 22/02/2012 17:40, Oliver Hartkopp a écrit :
>> On 22.02.2012 17:33, Stephane Grosjean wrote:
>>
>>> Le 22/02/2012 17:01, Oliver Hartkopp a écrit :
>>>
>>>> Pls try the following:
>>>>
>>>> PC1:
>>>> can0 ->   usb-pro ch1 ----+
>>>> can1 ->   usb-pro ch2 -+  |
>>>> can2 ->   usb ---------+  |
>>>>                        |  | 1Mpbs
>>>> PC2:                 |  |
>>>> can0 ->   pcmcia ch1 --+  |
>>>> can1 ->   pcmcia ch2 -----+
>>>>
>>>> PC1:
>>>> cangen -g0 -i can0
>>>> cangen -g0 -i can1
>>>> candump any
>>>>
>>>> PC2:
>>>> cangen -g0 -i can0
>>>> cangen -g0 -i can1
>>>> candump any
>>>>
>>>> and then pull the usb-pro from PC1 ...
>>>>
>>>> That's really unfriendly 8-)
>>>
>>> Ok, I found another cable and set up your tesbed: now the usb-pro is pulled
>>> and both "candump" always are displaying data sent by "cangen can0" from PC2
>>> and read from pcan-usb on PC1... Both drivers can be removed without any
>>> problem.
>>
>> cangen can0 only on PC2?
>>
>> pls put the cangen on PC2 _and_ PC1 as shown above.
>>
> 
> Sorry but your scheme showed cangen only on can0 and can1... which are mapped
> to usb-pro can interfaces. on PC1, which is pulled from PC1! So, no more can
> data is sent from PC1! Remains only "cangen can0" from PC2!


It's not relevant if there's any traffic left, when you pulled the usb-pro.

The trick is to get maximum load (rx AND tx) to the USB adapter when
disconnecting the USB.

Therefore putting max tx load to _both_ PCs CAN interfaces was my plan - and
this is presented in the testbed description IMHO. Or did i miss anything?

> I just did another test, adding "cangen can2" on PC1 too... Everything is
> still ok, on both sides...


Fine.


>>> And what is (to) "follow your bug report on my setup" ???
>>
>> I just wanted to make sure that you can see the same effect with your setup
>> that i had with version 5.
> 
> Ok, confusion came from the fact that, once I did modifications into v5, this
> automatically became a "v6"  for me ...
> So you want me to reproduce your crash using the "old" v5, right? I already
> did that, I confirm that the pc hanged with v5, removing the driver and/or
> pulling (one of) the usb adapters.


Ah, ok. That's important. Because if i would had any problems that you do not
see in your setup, my testbed might have been broken or my kernel might be
misconfigured for that testing purpose.

> Should I now wait for any linux-usb ml review before posting this v6?


IMO it makes no sense to review an outdated patch.

Please post your v6 as you did for the v5 - adding some changelog.

Regards,
Oliver