On Tue, Nov 04, 2014 at 02:13:30PM +0100, Oliver Hartkopp wrote:

On 04.11.2014 11:33, Marc Kleine-Budde wrote:

quoted

On 11/04/2014 10:27 AM, Dong Aisheng wrote:

quoted

quoted

+       /* We meet an IC issue that we have to write the full 8

At least on the *insert SoC name here*, an issue with the Message RAM
was discovered. Sending CAN frames with dlc less than 4 bytes will lead
to bit errors, when the first 8 bytes of the Message RAM have not been
initialized (i.e. written to). To work around this issue, the first 8
bytes are initialized here.

Yes. Also put the current IP revision (3.0.x) into the comment.
Did inform the Bosch guys from this issue - or is it already in some errata sheet?

Good idea, will add it.
I will try if we can talk Bosch guys about this issue.

Regards
Dong Aisheng

quoted

quoted

+        * bytes (whatever value for the second word) in Message RAM to
+        * avoid bit error for transmit data less than 4 bytes at the first
+        * time. By initializing the first 8 bytes of tx buffer before using
+        * it can avoid such issue.
+        */
+       m_can_fifo_write(priv, 0, M_CAN_FIFO_DATA(0), 0x0);
+       m_can_fifo_write(priv, 0, M_CAN_FIFO_DATA(1), 0x0);
+
        m_can_config_endisable(priv, false);
 }

Can you trigger the issue when sending CAN-FD frames with dlc > 8 && dlc
< 64?

Just a nitpick:

DLC can just be 0 .. 15

and the length (struct canfd_frame.len) can be from 0 .. 64

See:

http://lxr.free-electrons.com/source/include/uapi/linux/can.h#L83

That's the reason for all these helpers

http://lxr.free-electrons.com/source/drivers/net/can/dev.c#L36

that hide the evil "DLC" from userspace now and make 'len' a usable
loop variable as we were able to use the former dlc for classic CAN
:-)

Regards,
Oliver