Re: [PATCH v15 23/26] sched: early boot clock

From: Dominique Martinet
Date: Mon Jan 07 2019 - 18:48:59 EST

Pavel Tatashin wrote on Mon, Jan 07, 2019:
> I could not reproduce the problem. Did you suspend to memory between
> wake ups? Does this time jump happen every time, even if your laptop
> sleeps for a minute?

I'm not sure I understand "suspend to memory between the wake ups".
The full sequence is:
- start a VM (just in case, I let it boot till the end)
- suspend to memory (aka systemctl suspend) the host
- after resuming the host, soft reboot the VM (login through
serial/ssh/whatever and reboot or in the qemu console 'system_reset')

I've just slept exactly one minute and reproduced again with the fedora
stock kernel now (4.19.13-300.fc29.x86_64) in the VM.

Interestingly I'm not getting the same offset between multiple reboots
now despite not suspending again; but if I don't suspend I cannot seem
to get it to give an offset at all (only tried for a few minutes; this
might not be true) ; OTOH I pushed my luck further and even with a five
seconds sleep I'm getting a noticeable offset on first VM reboot after

[ 0.000000] Hypervisor detected: KVM
[ 0.000000] kvm-clock: Using msrs 4b564d01 and 4b564d00
[ 179.362163] kvm-clock: cpu 0, msr 13c01001, primary cpu clock
[ 179.362163] clocksource: kvm-clock: mask: 0xffffffffffffffff max_cycles: 0x1cd42e4dffb, max_idle_ns: 881590591483 ns

Honestly not sure what more information I could give, I'll try on some
other hardware than my laptop (if I can get a server to resume after
suspend through ipmi or wake on lan); but I don't have anything I could
install ubuntu on to try their qemu's version... although I really don't
want to believe that's the difference...