imx6 flexcan

imx6 flexcan

[Cédric Chépied]

Hi,

I'm running linux-evl on my imx6ul and it seems to run fine so far.

Now I want to send can data in an oob thread. I added the SOCK_OOB flag to my
socket and tried to run oob_write. It returns -1 so I think flexcan driver has
no oob capabilities.

How much work would it take to give OOB capabilities to the driver? What level
of support can I expect?

Thanks,
-- 
Cédric Chépied
<cedric.chepied@drotek.com>

<http://events.drotek.com/>

Re: imx6 flexcan

[Philippe Gerum]

Cédric Chépied <cedric.chepied@drotek.com> writes:

> Hi,
>
> I'm running linux-evl on my imx6ul and it seems to run fine so far.
>
> Now I want to send can data in an oob thread. I added the SOCK_OOB flag to my
> socket and tried to run oob_write. It returns -1 so I think flexcan driver has
> no oob capabilities.
>
> How much work would it take to give OOB capabilities to the driver? What level
> of support can I expect?
>
> Thanks,

Real-time driver support with EVL is documented here:
https://evlproject.org/core/oob-drivers/

There is more than just enabling an oob path for the flexcan driver,
such path should be enabled in the generic CAN stack as well for this to
work, as this was done for the generic DMA engine, the generic SPI core,
and the generic GPIO layer.

-- 
Philippe.

Re: imx6 flexcan

[Cédric Chépied]

Philippe Gerum wrote:
> Real-time driver support with EVL is documented here:
> https://evlproject.org/core/oob-drivers/

Socket Interface is under construction.

> There is more than just enabling an oob path for the flexcan driver,
> such path should be enabled in the generic CAN stack as well for this to
> work, as this was done for the generic DMA engine, the generic SPI core,
> and the generic GPIO layer.

Sounds like a lot of work. Perhaps I should consider switching to Xenomai 3.

-- 
Cédric Chépied
<cedric.chepied@drotek.com>

<http://events.drotek.com/>

Re: imx6 flexcan

[Philippe Gerum]

Cédric Chépied <cedric.chepied@drotek.com> writes:

> Philippe Gerum wrote:
>> Real-time driver support with EVL is documented here:
>> https://evlproject.org/core/oob-drivers/
>
> Socket Interface is under construction.
>

Socket is an interface, an access point. The real meat is the stack underneath.

>> There is more than just enabling an oob path for the flexcan driver,
>> such path should be enabled in the generic CAN stack as well for this to
>> work, as this was done for the generic DMA engine, the generic SPI core,
>> and the generic GPIO layer.
>
> Sounds like a lot of work. Perhaps I should consider switching to Xenomai 3.

It is a lot of work.

-- 
Philippe.