Re: i.MX6 SabreLite FlexCAN emulation for QEMU progress/working

[Bernhard Beschow <shentey@gmail.com>]

Am 10. Juli 2025 12:50:41 UTC schrieb Marc Kleine-Budde <mkl@pengutronix.de>:
>On 10.07.2025 12:58:05, Pavel Pisa wrote:
>> Hello Marc and others,
>> 
>> Matyas Bobek has implemented FlexCAN emulation for i.MX6
>> targets in the frame of his bachelor thesis
>> 
>>   https://dspace.cvut.cz/bitstream/handle/10467/122654/F3-BP-2025-Bobek-Matyas-BP_Bobek_FlexCAN_final_4.pdf
>> 
>> with reviews there
>> 
>>   https://dspace.cvut.cz/handle/10467/122779
>> 
>> Please, if you have problems to access documents, try that later,
>> our large archive with theses and publications is under
>> "AI" scrape-bots attacks almost permanently... or send me note
>> and I send the document directly.
>> 
>> The project continues and on base of Marc Kleine-Budde
>> review, Matyas Bobek has implemented timestamping
>> which should ensure correct messages ordering
>> even if FIFO mode is not used in the drivers,
>> Linux driver default case. There is even workaround
>> to advance timestamp if the the QEMU virtual time
>> is too coarse to provide unique values for for
>> too fast delivered messages through virtual bus
>> without real bits serialization.
>
>Nice! Good work!

Indeed!

>
>> The actual code is available on the "flexcan"
>> branch of the repository
>> 
>>   https://gitlab.fel.cvut.cz/bobekmat/qemu-flexcan
>> 
>> The clean, minimal patches sequence will be prepared
>> for review for the proposal for inclusion into QEMU
>> mainline, hopefully soon.
>> 
>> In longer term perspective, I have interrest in i.MX8
>> and other FlexCAN variants with FD support, etc.
>> I do not know if Mr. Bobek will have time and interest
>> to continue in this direction. I can offer that as the
>> another thesis or can offer that as summer job
>> etc. But latter would need some funding for students
>> motivation and their time compensation and I have had
>> never and do not have any QEMU related project for such
>> funding now and even minimal chance is even in future
>> for other than CTU CAN FD IP core where we support
>> silicon and FPGA products. So no timing, chances,
>> estimates for something beyond i.MX6 FlexCAN. 
>
>ACK, I think getting things mainline is more important than CAN-FD
>support.

Any news?

I've been playing with Matyas' flexcan branch and integrated the FlexCAN device model into the imx8mp-evk machine [1]. With some minor modifications such as allowing for 64-bit access I am able to run cangen+candump on either side and see the communication!

I think Matyas' work should be upstreamed. How shall we proceed? If desired, I could do the upstreaming process, leaving all credits to Matyas of course. IIUC we'd need Matyas' Signed-off-by tags (of the *squashed* branch) in this case for legal reasons, correct?

What do you think?

Best regards,
Bernhard

[1] https://github.com/shentok/qemu/tree/imx8mp-flexcan

>
>regards,
>Marc
>