PRU /dev/rpmsg_pruXX is not closed when PRU stopped and queue was fully consumed

[nahuel]

To reproduce the problem:

cd /tmp
git clone git://git.ti.com/pru-software-support-package/pru-software-support-package.git
cd /tmp/pru-software-support-package/examples/am335x/PRU_RPMsg_Echo_Interrupt0
export PRU_CGT=/usr/share/ti/cgt-pru
make
cp gen/PRU_RPMsg_Echo_Interrupt0.out /lib/firmware/am335x-pru0-fw
echo start > /sys/class/remoteproc/remoteproc1/state

Now dmesg will show:

[  142.480770] remoteproc remoteproc1: powering up 4a334000.pru
[  142.487884] remoteproc remoteproc1: Booting fw image am335x-pru0-fw, size 74432
[  142.500918] pruss 4a300000.pruss: configured system_events[63-0] = 0x00000000.00030000
[  142.511928] pruss 4a300000.pruss: configured intr_channels = 0x00000005 host_intr = 0x00000005
[  142.528225] remoteproc remoteproc1: registered virtio0 (type 7)
[  142.540164] remoteproc remoteproc1: remote processor 4a334000.pru is now up
[  142.597314] virtio_rpmsg_bus virtio0: creating channel rpmsg-pru addr 0x1e
[  142.605230] virtio_rpmsg_bus virtio0: rpmsg host is online
[  142.641390] rpmsg_pru virtio0.rpmsg-pru.-1.30: new rpmsg_pru device: /dev/rpmsg_pru30

Then do:

echo hello > /dev/rpmsg_pru30

And open a new console and execute this without killing cat, keep it active:

# cat /dev/rpmsg_pru30
hello

Now stop the PRU:

echo stop > /sys/class/remoteproc/remoteproc1/state

dmesg will show:

[  163.736825] pruss 4a300000.pruss: unconfigured system_events[63-0] = 0x00000000.00030000
[  163.755473] pruss 4a300000.pruss: unconfigured host_intr = 0x00000005
[  163.773323] remoteproc remoteproc1: stopped remote processor 4a334000.pru

And now /dev/rpmsg_pru30 is removed but cat has the file descriptor opened. The problem is cat DOESN'T STOP. I think the correct kernel behaviour is to close the file descriptor after the PRU is stopped and the queue is fully consumed, so the user-space program has a way to know it needs to finish.

[jadonk]

Makes sense. We should bring it up on linux-omap.

[pdp7]

@nahuel @jadonk does this still happen with the current images?

@MarkAYoder this might be a good thing to check while doing the PRU cookbook

[MarkAYoder]

@pdp7 I was able to reproduce this on both a BB Black and the AI.

Sorry, no solution.