"Kirill A. Korinsky" <kir...@korins.ky> writes:
> [[PGP Signed Part:Undecided]] >> On 2. Jan 2024, at 16:34, Dave Voutila <d...@sisu.io> wrote: >> >> "Kirill A. Korinsky" <kir...@korins.ky> writes: >> >>> Greetings, >>> >>> When playing with autoinstall in VM I encountered a kernel panic. It doesn't >>> happened each attempt, but often enough to be easy found. >> >> Thanks for the report. It looks like you're using 7.4 for the host and >> guest (based on the dmesg and the fact you're booting the ramdisk kernel >> from /bsd.rd). Correct? > > Yes, you're correct. I'm using both 7.4 and using the same /bsd.rd > >> Any guess how frequently it occurs? 1 in 10? 1 in 100? I haven't seen >> this myself but have a similar report from mbuhl@ when testing a diff I >> shared in November (that hasn't been committed) so I'm curious if this >> is something related to virtio(4) and vio(4) in the kernel vs. something >> in vmd(8). (The diff he was testing was a complete re-write of vmd's >> network device emulation.) > > Let say 3 out of 10. And sometime it can be reproduced on the raw, > a few times. So, quite often on this setup. Can you try the diff shared recently on bugs@? https://marc.info/?l=openbsd-bugs&m=170482622227793&q=raw It looks like the vio(4) driver doesn't wait for vmd to reset the device before it starts touching memory. This is a kernel/driver issue so should be applied to the host. I'm not sure if it will cleanly apply to 7.4, however, so you may need to be following -current on the host. -dv
