Re: [patch 10/10] can: c_can : Disable rx split as workaround

[Wolfgang Grandegger <wg@grandegger.com>]

On 04/05/2014 09:48 PM, Thomas Gleixner wrote:
> On Sat, 5 Apr 2014, Wolfgang Grandegger wrote:
>> On 04/05/2014 08:56 PM, Thomas Gleixner wrote:
>>> On Fri, 4 Apr 2014, Mark wrote:
>>>
>>>> Thomas:  a few comments on my experience with the CAN driver:
>>>>
>>>> I think I agree with your approach -- I had the same problem losing a packet
>>>> when running through the clearing of the newdat flags.
>>>>
>>>> About a month ago, I re-wrote pch_can.c  (before I discovered this mailing
>>>> list) and got it to work successfully sending / receiving CAN data at 1 MBIT
>>>
>>> Wow, that's amazing.
>>
>> We realized late that the pch_can manual [1] is an exact copy of the
>> c_can manual just with different register names (and without mentioning
>> c_can at all). Also the initial author didn't mentioning it. Maybe he
>> did not even know.
>>
>>> That driver is just a variant of c_can.c with quite some of the same
>>> bugs and some different ones.
>>
>> It's worse :(.
>>
>>> We really should switch that over to c_can.c and get rid of it.
>>
>> See http://news.gmane.org/gmane.linux.can. Should we remove pch_can
>> immediately? Or labeling it deprecated first.
> 
> Dunno, but removing it right away seems to be the better solution to
> avoid that people start fixing pch_can.c again.

I agree, especially because that driver does have serious bugs.

> One trick to keep the existing users happy, is keeping the Kconfig
> switch and select the C_CAN stuff for it. Then you can gradually phase
> it out.

I was also thinking about that. Just the module name will then be different.

Wolfgang.