Re: [PATCH 3/3] can: m_can: fix RX path: use rx-offload to ensure skbs are sent from softirq context

[Torin Cooper-Bennun <torin@maxiluxsystems.com>]

On Fri, Mar 05, 2021 at 11:29:57PM +0100, Marc Kleine-Budde wrote:
> On 05.03.2021 17:20:15, Torin Cooper-Bennun wrote:
> > For peripheral devices, m_can sent skbs directly from a threaded irq
> > instead of from a softirq context. This patch transitions the driver to
> > use the rx-offload helper, ensuring the skbs are sent from the correct
> > context, with h/w timestamping to ensure correct ordering.
> 
> I think you beak the non-peripheral drivers here. They already have a
> NAPI function m_can_poll(). It makes no sense and doesn't work, if you
> do the RX in NAPI and then queue to rx-offload, which then needs to run
> from NAPI again. But it cannot as m_can_poll is the NAPI function.
> 
> For peripherals it works, as you do the RX in the threaded IRQ, queue to
> rx-offload, which then schedules a NAPI, to push the CAN frames into the
> networking stack.

Understood, I will make the skb handling conditional on
cdev->is_peripheral and retain netif_receive_skb and can_get_echo_skb
usage for non-peripherals as before.

--
Regards,

Torin Cooper-Bennun
Software Engineer | maxiluxsystems.com