Re: [PATCH v2 3/6] can: c_can: fix control interface used by c_can_do_tx

[Dario Binacchi]

> Il 01/03/2021 12:36 Marc Kleine-Budde <mkl@xxxxxxxxxxxxxx> ha scritto:
> 
>  
> On 28.02.2021 11:35:31, Dario Binacchi wrote:
> > > On 25.02.2021 22:51:52, Dario Binacchi wrote:
> > > > According to commit 640916db2bf7 ("can: c_can: Make it SMP safe") let RX use
> > > > IF1 (i.e. IF_RX) and TX use IF2 (i.e. IF_TX).
> > > 
> > > Is this a fix?
> > > 
> > 
> > I think that If I consider what is described in the 640916db2bf7
> > commit, using the IF_RX interface in a tx routine is wrong.
> 
> Yes, IF_RX is used in c_can_do_tx(), but that's called from
> c_can_poll(), which runs ins NAPI.

Yes, you are right. I was misled by the name of the function.

> 
> As far as I understand 640916db2bf7 ("can: c_can: Make it SMP safe")
> fixes the race condition that c_can_poll() and c_can_start_xmit() both
> access the same IF. See again the patch description:
> 
> | The hardware has two message control interfaces, but the code only uses the
> | first one. So on SMP the following can be observed:
> | 
> | CPU0            CPU1
> | rx_poll()
> |   write IF1     xmit()
> |                 write IF1
> |   write IF1
> 
> It's not 100% accurate, as the race condition is not just
> c_can_do_rx_poll() against the c_can_start_xmit(), but the whole
> c_can_poll() against c_can_start_xmit().
> 
> If you think my analysis is correct, please update the patch and add a
> comment to clarify why IF_RX is used instead of changing it to IF_TX.

I agree with you, I'll do it.

Thanks and regards,
Dario

> 
> regards,
> Marc
> 
> -- 
> Pengutronix e.K.                 | Marc Kleine-Budde           |
> Embedded Linux                   | https://www.pengutronix.de  |
> Vertretung West/Dortmund         | Phone: +49-231-2826-924     |
> Amtsgericht Hildesheim, HRA 2686 | Fax:   +49-5121-206917-5555 |