Linux/AM3358: Can't set up RPMsg

Please post which Linux version you use.

I'm using 4.14.40 linux kernel from git://git.ti.com/processor-sdk/processor-sdk-linux.git

Hello Andrey,

1) Are you running something else on the PRU core that is not using RPMsg? If so, try again without that non-RPMsg PRU core running. If that fixes your issue, jump down to 1, cont.

2) The Hands-on labs are a good resource. The labs also have a lot of non-RPMsg information, so I suggest going through the RPMsg Quick Start Guide to see if that resolves your problem.

3) If you are having trouble going through specific steps in either the Labs or the Quick Start Guide, tell me which step you are having trouble with.

1, cont) If your issue is fixed by only using one PRU core, you might have to modify the system event > interrupt channel > host interrupt mapping in the INTC. Ask me for more information on how to do this. The INTC is shared by both PRUs (see the technical reference manual for AM335x, chapter Programmable Real-Time Unit Subsystem and Industrial Communication Subsystem (PRU-ICSS) > Functional Description > Interrupt Controller (INTC)

Regards,
Nick

Thanks for the reply, Nick!) I've try to do the following to run only RPMsg things on PRU:
echo 'PRU_Halt_1.out' > /sys/class/remoteproc/remoteproc0/firmware
echo 'start' > /sys/class/remoteproc/remoteproc0/state
echo 'PRU_PRMsg_Echo_Interrupt1_1.out' > /sys/class/remoteproc/remoteproc1/firmware
echo 'start' > /sys/class/remoteproc/remoteproc1/state

And the error remained... So I will keep on digging...

Hello Andrey,

Ok, good luck with the continued debugging. I'm double-checking to see if there are known situations where the INTC will complain that it is already configured. If I don't learn anything new, I'll just wait for an update from your end.

Regards,
Nick

After some point I started to get the following error instead of mentioned above: virtio vring interrupt mechanisms are not provided (not sure what have triggered that change though, it didn't look like i'd changed anything). After some searching through linux RPMsg drivers code I realized that the reason is that pru_rproc module couldn't create mailbox for ARM<->PRU communications (the problem was in sanity check before mailbox creation that I didn't get completely - if (pru->irq_vring <= 0 && pru->irq_kick <= 0 &&!of_device_is_compatible(np, "ti,k2g-pru"))) - so I've commented it and mailbox finally was created (I also added mboxes property for pru instances in my device tree). But still when I tried to modprobe rpmsg_client_sample there is no output in dmesg and there is no pru devices in /dev/ directory. It looks like rpmsg_pru_probe function from rpmsg_pru module hadn't been called for some reason. I also found the following thread with similar problem: e2e.ti.com/.../672762. It's look like different versions of RPMsg software uses different mechanisms for ARM<->PRU communications - either mailboxes or system events. So maybe I don't need mailboxes for ARM<->PRU communications. So, I am still confused about what should I do with my kernel to get this example working... By the way here is my lsmod output:
Module Size Used by Not tainted
rpmsg_client_sample 16384 0
rpmsg_rpc 28672 0
rpmsg_proto 16384 0
rpmsg_char 16384 0
virtio_rpmsg_bus 20480 0
pruss_soc_bus 16384 0
pru_rproc 24576 2
pruss_intc 16384 1 pru_rproc
rpmsg_pru 16384 0
rpmsg_core 16384 6 rpmsg_client_sample,rpmsg_rpc,rpmsg_proto,rpmsg_char,virtio_rpmsg_bus,rpmsg_pru
pruss 16384 1 pru_rproc
omap_mailbox 20480 2

FYI: the exact kernel version from git://git.ti.com/processor-sdk/processor-sdk-linux.git is 4.14.40-4796173

Hello Andrey,

I am going to mark this thread resolved. Feel free to reply if you need to continue the discussion.

Regards,
Nick