Re: sja1000 interrupt problem

[Oliver Hartkopp <socketcan@hartkopp.net>]

On 13.11.2013 00:22, Austin Schuh wrote:

> Here is what is in the syslog from the same machine sending to it's
> self with a non-realtime kernel.
> 
> # uname -a
> Linux vpc5 3.10-3-amd64 #1 SMP Debian 3.10.11-2 (2013-09-10) x86_64 GNU/Linux
> 
> <6>[  169.993870] peak_pci 0000:05:00.0 can1: Got an sja1000 interrupt.
> <6>[  169.993904] peak_pci 0000:05:00.0 can1: Received packet.
> <6>[  169.993923] peak_pci 0000:05:00.0 can1: sja1000_rx
> <6>[  169.993994] peak_pci 0000:05:00.0 can1: Returning IRQ_HANDLED
> <6>[  169.994013] peak_pci 0000:05:00.0 can1: Found can1, disabling tracing.
> <6>[  169.994029] peak_pci 0000:05:00.0 can0: Got an sja1000 interrupt.
> <6>[  169.994048] peak_pci 0000:05:00.0 can0: TX complete.
> <6>[  169.994061] peak_pci 0000:05:00.0 can0: Returning IRQ_HANDLED

This looks indeed much better :-)

> 
> When I attach the CAN device which continually sends, I get the
> following.  (The front might be snipped improperly, not sure.)
> 
> <6>[  640.823323] peak_pci 0000:05:00.0 can1: Got an sja1000 interrupt.
> <6>[  640.823331] peak_pci 0000:05:00.0 can1: Returning IRQ_NONE
> <6>[  640.823334] peak_pci 0000:05:00.0 can0: Got an sja1000 interrupt.
> <6>[  640.823344] peak_pci 0000:05:00.0 can0: Received packet.
> <6>[  640.823346] peak_pci 0000:05:00.0 can0: sja1000_rx
> <6>[  640.823391] peak_pci 0000:05:00.0 can0: Returning IRQ_HANDLED

This is a correct behaviour too:

The shared IRQ is handled by can1 (which had nothing to do) and then the chain
goes to can0 which handles the reception correctly.

@Wolfgang: Do we need an additional protection for the PITA handling in
peak_pci.c ?

@Austin:
I have another idea to test, if it just shows up in the mainline driver:

Can you please download the out-of-tree driver from PEAK (version 7.9):

http://www.peak-system.com/linux/index.htm

I wonder if this one actually compiles with a -rt kernel and if it shows up
the same issue then.

Best regards,
Oliver