Re: NetBSD 5.1-amd64 fails to boot in QEMU

[Chavdar Ivanov <ci4ic4%gmail.com@localhost>]

On 24 January 2012 19:22, Greg Fefelov <apiocera%gmail.com@localhost> wrote:
> Hello.

Hi,

> Recently I've been trying to install NetBSD to my VPS: they use QEMU
> (integrated with Xen architecture, I guess) and allow to upload
> oneself's hard disk images.

I found myself with some spare cycles on a reasonably decent Xeon
server running XenServer 6.

NetBSD fully virtualized mode is probably a non-starter. There is a
long-standing problem with the interrupt timer on the emulated RealTek
network hardware, one gets timeouts all the time and the performance
is abysmal. There have been some patches along the way which to the
best of my knowledge did not get integrated. I am running -current:

$ uname -a
NetBSD support7 5.99.60 NetBSD 5.99.60 (GENERIC) #0: Tue Jan 24
01:41:22 GMT 2012
root@support6:/usr/obj/auto-current/amd64/sys/arch/amd64/compile/GENERIC
amd64

As a data point, I am getting the following:

root@loan8 ~tscam2/snaps # iperf -c support6 -w 64k
------------------------------------------------------------
Client connecting to support6, TCP port 5001
TCP window size: 65.0 KByte (WARNING: requested 64.0 KByte)
------------------------------------------------------------
[  3] local 10.1.32.165 port 55742 connected with 10.1.32.6 port 5001
[ ID] Interval       Transfer     Bandwidth
[  3]  0.0-10.0 sec   549 MBytes   460 Mbits/sec
iperf -c support6 -w 64k  0.09s user 8.18s system 81% cpu 10.112 total
root@loan8 ~tscam2/snaps # iperf -c support7 -w 64k
------------------------------------------------------------
Client connecting to support7, TCP port 5001
TCP window size: 65.0 KByte (WARNING: requested 64.0 KByte)
------------------------------------------------------------
[  3] local 10.1.32.165 port 55739 connected with 10.1.32.7 port 5001
[ ID] Interval       Transfer     Bandwidth
[  3]  0.0-14.7 sec   512 KBytes   285 Kbits/sec
root@loan8 ~tscam2/snaps # iperf -c support8 -w 64k
------------------------------------------------------------
Client connecting to support8, TCP port 5001
TCP window size: 65.0 KByte (WARNING: requested 64.0 KByte)
------------------------------------------------------------
[  3] local 10.1.32.165 port 55736 connected with 10.1.32.8 port 5001
[ ID] Interval       Transfer     Bandwidth
[  3]  0.0-10.0 sec   659 MBytes   552 Mbits/sec
iperf -c support8 -w 64k  0.06s user 8.00s system 79% cpu 10.110 total


support6 is running yesterday's -current amd64 in PV mode with
XEN-DOMU kernel, debugging turned off and MULTIPROCESSOR on, the
machine has four CPUs allocated (this is still marked as unstable but
seems to generally work for me, if I am careful; I ran full build on
that system today).
support7 is running GENERIC amd64 -current, the benchmark shows the
networking to be some 2000 times slower than on the PV mode machine,
therefore unusable in this mode (I still intend to drag the available
patches to the re driver to try it, at least Xorg is running OK on
this machine).
support8 is running FreeBSD9 amd64 with the XENHVM kernel, showing the
not insignificant 20-25% faster networking compared to NetBSD-current
(it also works fully virtalized, but the networking is some 2.5-3
times slower than on this one).

> I have had used VirtualBox to prepare the image and it worked well,
> but when they have tried to run it, NetBSD kept reloading after
> loading the kernel into memory, even before booting process itself.

Interesting. I'll prepare an .vhd file under VirtualBox and will try
to run it under Xen tomorrow.
>
> After that, I ran QEMU on my Windows box with the same result: with
> GENERIC kernel it hangs and reboots after
> booting hd0a:netbsd
> 10122216+572416+782784 [727680+477736]=0xd19a68

Which version of QEMU for Windows?

>
> AMD64 installation .iso anso hangs and reboots after
> booting cd0a:netbsd
> 10122912+5691904+783264=0x10d52a0
>
> ..and nothing else. What should I do?

I'd prepare the VM under Xen, I guess.

Chavdar Ivanov



-- 
----