Comment 51 for bug 2012763

Hi,

Just an update before I call it a day. I was able to partially reproduce the failure with "-smp cpus=300,maxcpus=300". I say partially because the logs posted by Jeff clearly indicate that there has been some sort of out-of-memory scenario, which didn't happen here for some reason. But the most interesting part of the log (the "smpboot: native_cpu_up: bad cpu NNN" messages) happened here.

I found the following interesting discussions upstream:

https://lists.gnu.org/archive/html/qemu-devel/2022-03/msg03562.html
https://<email address hidden>/

So yeah, there's really more to the problem than we initially thought. I will start my day tomorrow backporting some of the patches mentioned in the threads above and seeing how far I can get. There may likely be other patches involved, but it's hard to say now.

Something else that I'm planning to do is building a version of qemu from Noble (currently at 8.2.0, which is pretty recent) with a new machine type changing the maxcpus thing (like I did for Jammy). @Jeff, I'll put it in a PPA tomorrow and let you know, in case you can give it a try. Hopefully it will have much better support for what we're trying to achieve here.