In bi-directional CAN transfer using M_CAN IP, with
the frame gap being set to '0', it leads to Protocol
error in Arbitration phase resulting in communication
stall.
Discussed with Bosch M_CAN IP team and the stall issue
can only be overcome by controlling the tx and rx
packets flow as done by the patch.
Rx packets would also be serviced when there is a tx
interrupt. The solution has been tested extensively for
more than 10 days, and no issues has been observed.
Setup that is used to reproduce the issue:
+---------------------+ +----------------------+
|Intel ElkhartLake | |Intel ElkhartLake |
| +--------+ | | +--------+ |
| |m_can 0 | |<=======>| |m_can 0 | |
| +--------+ | | +--------+ |
+---------------------+ +----------------------+
Steps to be run on the two Elkhartlake HW:
1. ip link set can0 type can bitrate 1000000
2. ip link set can0 txqueuelen 2048
3. ip link set can0 up
4. cangen -g 0 can0
5. candump can0
cangen -g 0 can0 & candump can0 commands are used for transmit and
receive on both the m_can HW simultaneously where -g is the frame gap
between two frames.
Signed-off-by: Kumari Pallavi <kumari.pallavi@xxxxxxxxx>
Signed-off-by: Srikanth Thokala <srikanth.thokala@xxxxxxxxx>
---
drivers/net/can/m_can/m_can.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/drivers/net/can/m_can/m_can.c b/drivers/net/can/m_can/m_can.c
index a5003435802b..94aa0ba89202 100644
--- a/drivers/net/can/m_can/m_can.c
+++ b/drivers/net/can/m_can/m_can.c
@@ -1118,7 +1118,7 @@ static irqreturn_t m_can_isr(int irq, void *dev_id)
/* New TX FIFO Element arrived */
if (m_can_echo_tx_event(dev) != 0)
goto out_fail;
-
+ m_can_write(cdev, M_CAN_IE, IR_ALL_INT & ~(IR_TEFN));
if (netif_queue_stopped(dev) &&
!m_can_tx_fifo_full(cdev))
netif_wake_queue(dev);
@@ -1787,6 +1787,7 @@ static netdev_tx_t m_can_start_xmit(struct sk_buff *skb,
}
} else {
cdev->tx_skb = skb;
+ m_can_write(cdev, M_CAN_IE, IR_ALL_INT & (IR_TEFN));
return m_can_tx_handler(cdev);
}
--
2.17.1
