Once I had run ftq, I realized there was no way we were running on an AC with that kind of noise, and, sure enough, we were not. ftq is a handy diag.
A bit further digging showed I had not finished the job. I had not added support code for exiting to an AC from the kernel. It's there now https://github.com/rminnich/9front/commit/acde872bca3454e40864b35d450d80c9801db6e7#diff-8d386fb2115162117b438b49040ba74afbec505d3cf3d7214c35c7cdf3410f8fR1689 I can tell it's trying to start on the AC because it panics :-) Now, the 9front branch as of today is more up to date. But if you look at 9/pc64/nix.s, you'll see lots of lines with FIXME The problem is that things in Proc have moved (particularly dbgregs), so the four functions in there need a careful look. Some of the offsets are way off. You can try to compute this by hand (easy in 2011, when dbgregs was like the 4th thing in Proc, impossible in 2025, when it has moved); just write some code, that runs in kernel main, to print offsetof() those struct members, then wrap back around and fix nix.s or write a program that runs in user mode, you'll just have to get the includes right. If you get there first, lmk. here is my current qemu command line to boot to the embedded rc qemu-system-x86_64 -kernel ~/Downloads/9pc64 -smp 4 -nographic -serial mon:stdio -cpu qemu64,+monitor -append console=0 -m 8192 you don't need a .iso because you're stopping in the embedded fs. ------------------------------------------ 9fans: 9fans Permalink: https://9fans.topicbox.com/groups/9fans/T91ed5776302b0c4c-M001e8e973b42075e1096ee11 Delivery options: https://9fans.topicbox.com/groups/9fans/subscription
