[PATCH 7/7] can: m_can: workaround for transmit data less than 4 bytes

[Dong Aisheng]

On Mon, Nov 03, 2014 at 05:48:17PM +0100, Marc Kleine-Budde wrote:
> On 10/29/2014 11:45 AM, Dong Aisheng wrote:
> > We meet an IC issue that we have to write the full 8 bytes (whatever
> > value for the second word) in Message RAM to avoid bit error for transmit
> > data less than 4 bytes.
> 
> Is this a SoC or a m_can problem? Are all versions of the SoC/m_can
> affected? Is there a m_can version register somewhere?
> 

I'm still not sure it's SoC or m_can problem.
Our IC guys ran the simulation code and found this issue.
But due to some reasons, it may be very slow for they to investigate
and get the conclusion.

> > Without the workaround, we can easily see the following errors:
> > root at imx6qdlsolo:~# ip link set can0 up type can bitrate 1000000
> > [   66.882520] IPv6: ADDRCONF(NETDEV_CHANGE): can0: link becomes ready
> > root at imx6qdlsolo:~# cansend can0 123#112233
> > [   66.935640] m_can 20e8000.can can0: Bit Error Uncorrected
> > 
> > Signed-off-by: Dong Aisheng <b29396 at freescale.com>
> > ---
> >  drivers/net/can/m_can/m_can.c | 11 ++++++++++-
> >  1 file changed, 10 insertions(+), 1 deletion(-)
> > 
> > diff --git a/drivers/net/can/m_can/m_can.c b/drivers/net/can/m_can/m_can.c
> > index 219e0e3..f2d9ebe 100644
> > --- a/drivers/net/can/m_can/m_can.c
> > +++ b/drivers/net/can/m_can/m_can.c
> > @@ -1058,10 +1058,19 @@ static netdev_tx_t m_can_start_xmit(struct sk_buff *skb,
> >  	m_can_fifo_write(priv, 0, M_CAN_FIFO_ID, id);
> >  	m_can_fifo_write(priv, 0, M_CAN_FIFO_DLC, can_len2dlc(cf->len) << 16);
> >  
> > -	for (i = 0; i < cf->len; i += 4)
> > +	for (i = 0; i < cf->len; i += 4) {
> >  		m_can_fifo_write(priv, 0, M_CAN_FIFO_DATA(i / 4),
> >  				 *(u32 *)(cf->data + i));
> >  
> > +		/* FIXME: we meet an IC issue that we have to write the full 8
> 
> FIXME usually indicates that the driver needs some work here. Just
> describe your hardware bug, you might add a reference to an errata if
> available, though.
> 

We don't have an errata for it now.
Because i'm not sure this is the final workaround and also not sure if other
SoC vendors having the same issue, so i used FIXME here firstly.
Since the code is harmless, so i wish we could put it here first
until we find evidence no need for other SoC or only belong to specific
IP version.

> > +		 * bytes (whatever value for the second word) in Message RAM to
> > +		 * avoid bit error for transmit data less than 4 bytes
> > +		 */
> > +		if (cf->len <= 4)
> > +			m_can_fifo_write(priv, 0, M_CAN_FIFO_DATA(i / 4 + 1),
> > +					 0x0);
> 
> This workaround doesn't handle the dlc == 0 case, your error description
> isn't completely if this is a problem, too.
> 

You're right.
I just checked the dlc == 0 case also had such issue and it also needs
the extra 8 bytes write to avoid such issue.

BTW the issue only happened on the first time when you send a frame with no
data(dlc == 0) at the first time.
e.g.
root at imx6sxsabresd:~# ip link set can0 up type can bitrate 1000000
[   62.326014] IPv6: ADDRCONF(NETDEV_CHANGE): can0: link becomes ready
root at imx6sxsabresd:~# cansend can0 123#R
[   69.233645] m_can 20e8000.can can0: Bit Error Uncorrected
[   69.239167] m_can 20e8000.can can0: Bit Error Corrected

If we send a frame success first (e.g. 5 bytes data), it will not fail
again even you send no data frame (dlc == 0) later.

The former failure of sending data less than 4 bytes is similar.

Looks like the first 8 bytes of message ram has to be initialised
for the first using.

> It should be possible to change the for loop to go always to 8, or
> simply unroll the loop:
> 
> /* errata description goes here */
> m_can_fifo_write(priv, 0, M_CAN_FIFO_DATA(0), *(u32 *)(cf->data + 0));
> m_can_fifo_write(priv, 0, M_CAN_FIFO_DATA(1), *(u32 *)(cf->data + 4));
> 

Yes, i tried to fix it as follows.

/* FIXME: we meet an IC issue that we have to write the full 8
 * bytes (whatever value for the second word) in Message RAM to
 * avoid bit error for transmit data less than 4 bytes
 */
if (cf->len <= 4) {
        m_can_fifo_write(priv, 0, M_CAN_FIFO_DATA(0),
                         *(u32 *)(cf->data + 0));
        m_can_fifo_write(priv, 0, M_CAN_FIFO_DATA(1),
                         *(u32 *)(cf->data + 4));
} else {
        for (i = 0; i < cf->len; i += 4)
                m_can_fifo_write(priv, 0, M_CAN_FIFO_DATA(i / 4),
                                 *(u32 *)(cf->data + i));

Will update the patch.

Regards
Dong Aisheng

> > +	}
> > +
> >  	can_put_echo_skb(skb, dev, 0);
> >  
> >  	if (priv->can.ctrlmode & CAN_CTRLMODE_FD) {
> > 
> 
> Marc
> 
> -- 
> Pengutronix e.K.                  | Marc Kleine-Budde           |
> Industrial Linux Solutions        | Phone: +49-231-2826-924     |
> Vertretung West/Dortmund          | Fax:   +49-5121-206917-5555 |
> Amtsgericht Hildesheim, HRA 2686  | http://www.pengutronix.de   |
>