I have a lot of SMP work waiting to be merged, but still only for ARM (not aarch64 yet) so that will probably not interfere with your work. But you haven't touched the graphics drive right? (other than the vc6 gallium)

I'm very curious what you did there. When I started my own porting effort for RPi4 in May SMP made me go crazy sometimes. I was to dumb to PR my work on RPi4 port from May because I wanted to do more polishing...now we've both invested a lot of effort for the same thing. 🤷‍♂️ That's a reason I PRed the RPi5 stuff today.

Most of the RP4 work was done by @John Knipper actually, I took over some unfinished work from him before I pushed it. When I started the Raspi 3 port back in January and actually tried to make it SMP ready from the beginning. So I have had it running side by side with the non-smp build ever since. And yes, it is frustrating, debugging deadlocks is not fun. I have tried several approuches, including splitting up ready-queues to 1 per cpu and implement task stealing. But the code I have now uses 1 ready queue, and a lot of fixes in signal, irq, ifq and ipi handling. Branch is here: github.com/bsek/AROS/…/smp