> Can you net out what the current issues you're facing? I can't seem to
grok
> the log file posted above (not sure what I'm looking at). From reviewing
the
> thread, it seems the core dumps are gone but there may (?) be some issues
> still?

The core dumps are gone indeed. Thanks for the prompt fixes!

I'm using the Oct 11 snapshot.

The issue that's still vexing is VMs sticking in limbo. There's a vmd
process spinning at 100% cpu. vmctl status doesn't show the VM. vmctl stop
hangs waiting for a response from vmd. I don't think there's enough logging
in vmd. At least I don't see anything revealing in /var/log/daemon despite
running with -vv. Pasted the data overlapping with
https://marc.info/?l=openbsd-tech&m=153955188302856&w=2 as
https://gist.github.com/blackgnezdo/ebd728246abe418b1867df1361c95e27.

ci-openbsd$ ps ax | grep vm
34144 ??  Is      0:00.39 vmd: priv (vmd)
63860 ??  Isp     0:04.99 /usr/sbin/vmd -vv
 4223 ??  Isp     0:06.20 vmd: vmm (vmd)
71772 ??  Isp     0:09.01 vmd: control (vmd)
72373 ??  Rp/3  1136:32.53 vmd: ci-openbsd-main-1 (vmd)
90783 ??  Rp/3   50:24.59 vmd: ci-openbsd-main-2 (vmd)
47967 ??  Rp/0   49:57.49 vmd: ci-openbsd-main-0 (vmd)
55129 ??  Ip      0:00.02 vmctl stop ci-openbsd-main-1 -f -w

I could work around this in a classic sysadmin fashion by writing a script
which greps for any long running vmctl stop processes, then kill the
corresponding vmd. I think this works, but I also suspect that internal vmd
accounting which imposes the 4 VMs limit will soon get in the way.

BTW, we could use a higher VM count limit. I don't want it badly enough to
implement config passing code or run the VMs as root, hi Reyk :)

The log was from syz-manager, so it was mostly for Dmitry's benefit.

Thanks
Greg

Reply via email to