Vexriscv SMP interrupts not working with Zephyr

[lcomino64]

Hi all,

Here is the issue described here: zephyrproject-rtos/zephyr#78053 (comment)

I'm not entirely sure this is a LiteX issue since the LiteX BIOS works perfectly on my board, it's when I run my Zephyr program that interrupts start to malfunction and freeze the program (see issue).

I managed to run a sim of a similar configuration:

litex_sim --cpu-type=vexriscv_smp --cpu-count=2 --with-sdram --sdram-init ~/zephyrproject/zephyr/build/zephyr/zephyr.bin --with-rvc --with-privileged-debug --hardware-breakpoints 4 --jtag-tap --with-jtagremote

Where the zephyr.bin is the samples/subsys/console/echo sample from Zephyr
The program in the sim doesn't get as far but I would still like to see if the issue is present there, this is the output:

I just have no idea how to 'step through' the execution or analyse the output. Can anyone help me with this? Greatly appreciate any help.

[lcomino64]

I've attached here the zephyr.bin
zephyr.bin.gz

Here is my Zephyr fork:
https://github.com/lcomino64/zephyr-lcomino64

And here is a repo containing my LiteX configuration and build output:
https://github.com/lcomino64/Thesis-public/tree/main/litex

[Dolu1990]

Hi,

Can you send the elf file of the compiled zephyr aswell ?
Thanks ^^

About debugging with openocd the simulated VexRiscv, you can use :
#1887

But be aware, this is kinda slow.

[lcomino64]

Here:
zephyr.elf.gz

Thanks, I'll look into it

[Dolu1990]

Got a "riscv64-unknown-elf-objdump: zephyr.elf: file format not recognized" when tried to make a deassembly from the elf XD
which is very weird

Overall i just wanted to check there was no PMP related code, as vexriscv has no PMP support by default. And i know that zephyr my assume there is a pmp depending how it is configured

[lcomino64]

Yes I get the same thing

$ objdump -d build/zephyr/zephyr.elf 

build/zephyr/zephyr.elf:     file format elf32-little

objdump: can't disassemble for architecture UNKNOWN!

Not sure but I think it's meant to be "elf32-littleriscv" from the generated LiteX software/include/generated/output_format.ld? not "elf32-little"

addr2line still works for me though.

As for PMP, just checked and it's definitely disabled in my Kconfig

[lcomino64]

Hey @Dolu1990,

I have the debugger on the sim up and running (turns out the sim had different csr values duh), can confirm I get the same issue with ISR1 stack maxed out.

I'm looking through the execution now, I can see the Zephyr kernel throwing stack-related errors but I need help actually identifying the issue.

Can you please lend a hand?

I've made it the hello_world sample since the problem still occurs and there's less stuff going on.

Here's the elf and binary:
elf_bin.tar.gz

And here are the zephyr files for the sim in case you want to build for yourself. Just unzip to your zephyr directory.
zephyr_changes_sim.tar.gz