On 2/19/19 8:12 PM, Alexander E. Patrakov wrote:
Long Wu Yuan <[email protected]>:
(1) Mike said the same. Have you run with VMM_DEBUG ?
No. I am not even an OpenBSD user, found this thread when debugging
the lockup in KVM.
(2) That iso runs just fine on HyperV, VirtualBox & KVM.
Well, for hypervisors that allow all-virtio configurations, please try
them, as they will be (regarding the virtual hardware) the closest
thing to what vmm seems to do.
OK, just for a test, I installed OpenBSD 6.4 amd64 in a KVM-based
virtual machine (L1), and read enough manuals to set up vmm there and
run Alpine as an L2 guest (i.e. tested a nested-virtualization setup). I
did it a bit differently than you, though. Not sure if the differences
matter.
1. Used a single-processor kernel.
2. Configured bridge0 to contain vio0 (without any addresses, so that I
can use DHCP both in L1 and L2 guests), vether0 (with dhcp), and also
configured it as a switch for the virtual machines.
3. Gave only 512M of RAM to the Alpine Linux VM.
# cat /etc/vm.conf
switch "uplink" {
interface bridge0
}
# Alpine test VM
vm "alpinevmm" {
memory 512M
disk "/home/vm/alpinevmm.raw"
# Use a specific tap(4) interface with a hardcoded MAC address
interface {
lladdr 52:54:00:9c:db:0a
switch "uplink"
}
}
The VM was installed from this ISO:
SHA256 (/home/vm/alpine-virt-3.9.0-x86_64.iso) =
e87e0256d10ca2125455fb5023687b2346740a8cbae54972312fe1626f07494c
Result: it boots, and it has an uptime of 2 hours, hasn't wedged or
otherwise crashed so far. There is a problem with extremely slow clock
(~50% of the correct speed) in the VM, grossly overestimated bogomips
value in /proc/cpuinfo, and this warning in dmesg:
[ 0.060000] ----------------
[ 0.060000] | NMI testsuite:
[ 0.060000] --------------------
[ 0.060000] remote IPI: ok |
[ 0.060000] local IPI:TIMEOUT|
[ 0.200000] CPU: 0 PID: 1 Comm: swapper/0 Not tainted 4.19.18-0-virt
#1-Alpine
[ 0.200000] Hardware name: OpenBSD VMM, BIOS 1.11.0p0-OpenBSD-vmm
01/01/2011
[ 0.200000] Call Trace:
[ 0.200000] dump_stack+0x5d/0x79
[ 0.200000] dotest.constprop.0+0x3b/0x60
[ 0.200000] nmi_selftest+0x7f/0x14d
[ 0.200000] native_smp_cpus_done+0x15/0x9e
[ 0.200000] kernel_init_freeable+0x117/0x27f
[ 0.200000] ? rest_init+0x95/0x95
[ 0.200000] kernel_init+0x5/0xf5
[ 0.200000] ret_from_fork+0x35/0x40
[ 0.200000] --------------------
[ 0.200000] BUG: 1 unexpected failures (out of 2) - debugging
disabled! |
[ 0.200000]
-----------------------------------------------------------------
But well - someone here wanted a report whether an Alpine VM can survive
more than 30 minutes on vmm ;)
--
Alexander E. Patrakov
