Re: Flooding AT91_CAN peripheral with messages causes it to stop receiving any more messages

[Wolfgang Grandegger <wg@grandegger.com>]

Hello Alexander,

Am 02.05.2016 um 08:23 schrieb Alexander Stein:
> On Tuesday 05 April 2016 14:10:48, Amr Bekhit wrote:
>> I working on a board based on the AT91SAM9X25 SoC and I'm using
>> integrated CAN peripheral. I seem to have run into an issue whereby
>> sending lots of messages very rapidly in quick succession causes the
>> CAN peripheral to then stop receiving any messages at all. The only
>> way to bring it back to a functional state is to bring the network
>> interface down and then back up again.
>> [...]
>> I then start sending CAN messages to the unit using a PCAN-USB adapter
>> that is plugged into a test Linux PC. After bringing up the CAN
>> interface on the test PC, messages can be continuously sent using the
>> following bash script:
>> [...]
>> I then leave the system running for some time (1.5 hours typically,
>> may vary), periodically running ifconfig can0 to check to see if new
>> packets are being received. After a while, the can interface will stop
>> receiving new packets, even though the test PC is still transmitting
>> them. Stopping and restarting the CAN transmissions on the test PC
>> does not solve the problem. The interface does not appear to be in the
>> bus off state, as shown by running the following:
>
> That sounds a bit like my getting stuck problem in http://linux-can.vger.kernel.narkive.com/bBQqK84G/resend-patch-net-can-at91-can-c-decrease-likelyhood-of-rx-overruns#post2
>
> The patch post1 at least keeps the driver working. Although I don't know what
> has changed in at91_can meanwhile.

Thanks for pointing me to that patch. It still applies to Linux 4.1 with 
some minor fixes. Amr, could you please give it a try. Please let me 
know if you need help.

Anyway, I think the driver should not hang even in case of overflows. I 
will have a closer look later this week.

Wolfgang.