Support of UART interrupts

[Dimitris Andronikou]

Hello,

I was wondering does Openpiton support uart interrupts? It seems like the uart interrupt does not work. 

I have observed a strange behavior with the UART. I can only paste 16 characters to terminal, everything else is lost. The standard 16550-type UART has a 16-byte RX FIFO. That makes me think that no interrupt is generated on character RX, so a big chunk of pasted chars instantly overflows the buffer before the periodic polling clears it.

Thanks,

Dimitris

[Jonathan Balkind]

Hi Dimitris,

Yes it should work, but there are a couple of config oddities that are worth checking which are probably why I didn't leave them turned on in the committed config. I seem to recall that the kernel command line mattered and that the device tree contents also mattered. For a while it wouldn't work if we didn't have a "chosen" node in the device tree, so there's a chance that that is absent. I also recall that once they were enabled, we got double output lines because two paths were enabled, because of the kernel command line. It may be that earlycon or something similar needs to be changed to work around that issue.

Thanks,

Jon

--
You received this message because you are subscribed to the Google Groups "OpenPiton Discussion" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openpiton+...@googlegroups.com.
To view this discussion, visit https://groups.google.com/d/msgid/openpiton/7d72f27f-a1b3-4be1-9aa0-b98c796fdc6dn%40googlegroups.com.

[Dimitris Andronikou]

Hi Jon,

Thanks for the reply!

I looked into the PLIC RTL code a bit, and I found out that the UART interrupt is mapped to source 0, which is hardwired to 0: https://github.com/pulp-platform/rv_plic/blob/5b5c5a4c1c15c3d7bb833071d344b2c2bc5f599d/rtl/plic_top.sv#L111.

I've observed that the UART interrupt doesn't work in either baremetal or Linux, so I believe we have a hardware routing issue here, in addition to the device tree config you mentioned.

Thanks,

Dimitris

[Jonathan Balkind]

Uh I think the code already offsets for that so it isn't lost? The ioDeviceNr value for the device tree also starts from 1 to compensate: https://github.com/PrincetonUniversity/openpiton/blob/openpiton-dev/piton/tools/bin/riscvlib.py#L214

Thanks,

Jon

To view this discussion, visit https://groups.google.com/d/msgid/openpiton/93d16f30-9061-4d60-94e7-9270844f3aa1n%40googlegroups.com.