Question about D_CAN implementation

Question about D_CAN implementation

[T Thayer]

Hi,

I'm experimenting with the combined C_CAN/D_CAN Linux driver for the
Altera CycloneV SoC which uses the Bosch D_CAN module..

I'm having some problems though. Specifically, I'm seeing 2 copies of
a frame get transmitted.

Strangely, the IRQ Status register = 0x800A inside the c_can_poll()
function indicates a RX message object (0 <  RX < 16) instead of a TX
message object.

If I use the standalone D_CAN driver which was in the
drivers/net/can/d_can directory, it seems to work fine (IRQ Status =
0x8011). I believe this directory was removed when the c_can and d_can
drivers were combined (SHA 69927fccd96b15bd228bb82d)

I noticed some significant differences - particularly in the
c_can_poll() function.
IRQ Status Test.
- d_can.c ANDs the irqstatus with STATUS_INTERRUPT
- c_can.c checks to see if the irqstatus == STATUS_INTERRUPT.
After IRQ Status Test.
- d_can.c has a separate if/else if loop to determine whether the
Message was RX/TX
- c_can.c has is a continuation of the IRQ status test (else if, else if)

Another symptom is that the tx_packet statistics aren't incrementing
as I send data. However, this may be related to the IRQ indicating a
RX message object.

Thank you for any insight you may have,


Thor

Re: Question about D_CAN implementation

[Marc Kleine-Budde]

[-- Attachment #1: Type: text/plain, Size: 1737 bytes --]

On 11/12/2013 11:30 PM, T Thayer wrote:
> I'm experimenting with the combined C_CAN/D_CAN Linux driver for the
> Altera CycloneV SoC which uses the Bosch D_CAN module..
> 
> I'm having some problems though. Specifically, I'm seeing 2 copies of
> a frame get transmitted.

Where do you see 2 copies of the frame?

> Strangely, the IRQ Status register = 0x800A inside the c_can_poll()
> function indicates a RX message object (0 <  RX < 16) instead of a TX
> message object.
> 
> If I use the standalone D_CAN driver which was in the
> drivers/net/can/d_can directory, it seems to work fine (IRQ Status =
> 0x8011). I believe this directory was removed when the c_can and d_can
> drivers were combined (SHA 69927fccd96b15bd228bb82d)

There never was a mainline kernel with drivers/net/can/d_can.

> I noticed some significant differences - particularly in the
> c_can_poll() function.
> IRQ Status Test.
> - d_can.c ANDs the irqstatus with STATUS_INTERRUPT
> - c_can.c checks to see if the irqstatus == STATUS_INTERRUPT.
> After IRQ Status Test.
> - d_can.c has a separate if/else if loop to determine whether the
> Message was RX/TX
> - c_can.c has is a continuation of the IRQ status test (else if, else if)
> 
> Another symptom is that the tx_packet statistics aren't incrementing
> as I send data. However, this may be related to the IRQ indicating a
> RX message object.
> 
> Thank you for any insight you may have,

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   |


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 259 bytes --]

Re: Question about D_CAN implementation

[T Thayer]

On Tue, Nov 12, 2013 at 4:50 PM, Marc Kleine-Budde <mkl@pengutronix.de> wrote:
> On 11/12/2013 11:30 PM, T Thayer wrote:
>> I'm experimenting with the combined C_CAN/D_CAN Linux driver for the
>> Altera CycloneV SoC which uses the Bosch D_CAN module..
>>
>> I'm having some problems though. Specifically, I'm seeing 2 copies of
>> a frame get transmitted.
>
> Where do you see 2 copies of the frame?

I see 2 copies on a different board running the same c_can as well as
with a CAN Bus analyzer (Microchip).

>
>> Strangely, the IRQ Status register = 0x800A inside the c_can_poll()
>> function indicates a RX message object (0 <  RX < 16) instead of a TX
>> message object.
>>
>> If I use the standalone D_CAN driver which was in the
>> drivers/net/can/d_can directory, it seems to work fine (IRQ Status =
>> 0x8011). I believe this directory was removed when the c_can and d_can
>> drivers were combined (SHA 69927fccd96b15bd228bb82d)
>
> There never was a mainline kernel with drivers/net/can/d_can.
>

OK. Thank you for this clarification.

>> I noticed some significant differences - particularly in the
>> c_can_poll() function.
>> IRQ Status Test.
>> - d_can.c ANDs the irqstatus with STATUS_INTERRUPT
>> - c_can.c checks to see if the irqstatus == STATUS_INTERRUPT.
>> After IRQ Status Test.
>> - d_can.c has a separate if/else if loop to determine whether the
>> Message was RX/TX
>> - c_can.c has is a continuation of the IRQ status test (else if, else if)
>>
>> Another symptom is that the tx_packet statistics aren't incrementing
>> as I send data. However, this may be related to the IRQ indicating a
>> RX message object.
>>
>> Thank you for any insight you may have,
>
> 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   |
>

Below is a trace of the program flow after a transmit. I was expecting
to get interrupted with a 0x8011 but I'm getting a 0x800A.
My notes have << >> around them.

root@socfpga_cyclone5:~# cansend can0 -i 0x123 x011 0x22 0x33 0x44
interface = can0, family = 29, type = 3, proto = 1
CAN Start Xmit 17.
<<c_can_start_xmit() after msg_obj_no>>
It will send can frame with 0 us interval!
CAN Write Message Object 17
<<c_can_write_msg_object() @ entry >>
CAN Object Put (17)
<<c_can_object_put() @ entry>>
CAN Poll. IRQ Status = 0x0000800A.
<<c_can_poll() before NULL irqstatus test>>
CAN Poll. IRQ Status = 0x0000000A.                                  ""
CAN Poll. Keep RXing.
<<else if irqstatus > RX_FIRST && irqstatus < RX_LAST>>
CAN Do RX Poll
 <<c_can_do_rx_poll() before for loop>>
CAN Object Get
 <<c_can_object_get() @ entry>>

Thanks,

Thor