On Wed, Oct 17, 2018 at 9:40 PM, Greg Steuck <g...@nest.cx> wrote: > On Wed, Oct 17, 2018 at 12:00 PM Dmitry Vyukov <dvyu...@google.com> wrote: >> >> On Wed, Oct 17, 2018 at 4:50 AM, Greg Steuck <g...@nest.cx> wrote: >> > I think I see some evidence of occasional VM just never finishing >> > booting to >> > the point of running sshd. At least that's what I surmised from the >> > posted >> > manager.log file. >> > https://gist.github.com/blackgnezdo/a69e83c42c0c4cbbd53c7f3b35e91632 >> >> >> This should be fine in the sense that it should not lead to vmd losing >> VMs. > > > I agree that that vmd shouldn't be losing the VMs in such a case. But some > evidence shows they become unkillable as described in my previous message: > https://marc.info/?l=openbsd-tech&m=153955188302856&w=2
But VMM does not have an idea about guest OS, OS booting, etc. For VMM it's just a CPU that runs instruction some stream. There may be no OS at all. So I don't see how this can affect vmd behavior. If killing a VM leads to hangs, then most likely that can equally happen if the kill happens during boot or not during boot. >> If test machine does not come up live within some time frame, we >> kill it and create a new one. > > > How tight a timeout does it have? 10 mins to print IP, and then +20 mins for sshd to come up: https://github.com/google/syzkaller/blob/master/vm/vmm/vmm.go#L215
