[PATCH] can: at91_can: fix passive-state AERR flooding

When the at91_can is a single node on the bus and a user attempts to
transmit, the can state machine will report ack errors and increment the
transmit error count until it reaches the passive-state. Per the
specification, it will then transmit with a passive error, but will stop
incrementing the transmit error count. This results in the host machine
being flooded with the AERR interrupt forever, or until another node
rejoins the bus.

To prevent the AERR flooding, disable the AERR interrupt when we are in
passive mode.

Signed-off-by: Brandon Maier <redacted>
---
 drivers/net/can/at91_can.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/drivers/net/can/at91_can.c b/drivers/net/can/at91_can.c
index b06af90a9964..2a8831127bd0 100644
--- a/drivers/net/can/at91_can.c
+++ b/drivers/net/can/at91_can.c

@@ -804,8 +804,13 @@ static int at91_poll(struct napi_struct *napi, int quota)
 		work_done += at91_poll_err(dev, quota - work_done, reg_sr);
 
 	if (work_done < quota) {
-		/* enable IRQs for frame errors and all mailboxes >= rx_next */
+		/* enable IRQs for frame errors and all mailboxes >= rx_next,
+		 * disable the ack error in passive mode to avoid flooding
+		 * ourselves with interrupts
+		 */
 		u32 reg_ier = AT91_IRQ_ERR_FRAME;
+		if (priv->can.state == CAN_STATE_ERROR_PASSIVE)
+			reg_ier &= ~AT91_IRQ_AERR;
 
 		reg_ier |= get_irq_mb_rx(priv) & ~AT91_MB_MASK(priv->rx_next);
 

-- 
2.30.2

On 05/10/2021 at 20:30, Brandon Maier wrote:

When the at91_can is a single node on the bus and a user attempts to
transmit, the can state machine will report ack errors and increment the
transmit error count until it reaches the passive-state. Per the
specification, it will then transmit with a passive error, but will stop
incrementing the transmit error count. This results in the host machine
being flooded with the AERR interrupt forever, or until another node
rejoins the bus.

To prevent the AERR flooding, disable the AERR interrupt when we are in
passive mode.

Signed-off-by: Brandon Maier <redacted>

Even if I'm not familiar with the matter, the explanation above makes sense:
Acked-by: Nicolas Ferre <nicolas.ferre@microchip.com>

Thanks Brandon, best regards,
   Nicolas

-- 
Nicolas Ferre

On 05.10.2021 13:30:23, Brandon Maier wrote:

When the at91_can is a single node on the bus and a user attempts to
transmit, the can state machine will report ack errors and increment the
transmit error count until it reaches the passive-state. Per the
specification, it will then transmit with a passive error, but will stop
incrementing the transmit error count. This results in the host machine
being flooded with the AERR interrupt forever, or until another node
rejoins the bus.

To prevent the AERR flooding, disable the AERR interrupt when we are in
passive mode.

Can you implement Bus Error Reporting?

| https://elixir.bootlin.com/linux/v5.14/source/include/uapi/linux/can/netlink.h#L99

This way the user can control if bus errors, and the ACK error is one of
them, should be reported. Bus error reporting is disabled by default. I
think enabling AT91_IRQ_ERR_FRAME only if CAN_CTRLMODE_BERR_REPORTING is
active should do the trick.

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 |