Re: Panic: HYPERVISOR_mmu_update failed on old NetBSD-5.2 domU

[Stephen Borrill <netbsd%precedence.co.uk@localhost>]

On Tue, 10 Nov 2020, Greg Troxel wrote:
> Brian Buhrow <buhrow%nfbcal.org@localhost> writes:
> > Specifically, the working installation is using an Intel Xeon 5620 processor
> > while the non-working one is using an Intel I5-2400 CPU.

In my case, I've not seen a working system, but my failed one is Xeon 
E5506 running Xen 4.7.6 from a fully-patched XenServer 7.1 LTSR system.

> > 	maybe someone knows what I am obviously missing.  Any ideas would be
> > greatly appreciated.
>
> Not all that helpful, but there have been many advisories for security
> problems in xen, and changes to various hypervisor calls as part of the
> fixes.  NetBSD 5 is long unmaintained, as is 6, and 7 is no longer
> maintained.  I know you know that, but wanted to say it for the record.
>
> It might well be that the different CPU is exposing some code path that
> has a latent bug.

Yeah, I first reported this about 2.5 years ago:

http://mail-index.netbsd.org/port-xen/2018/04/17/msg009181.html

> I would look at the "xen dmesg" or whatever the right command is to see
> the xen kernel printfs from the dom0.

Yeah, that's pretty much what Manuel suggested; I found it printed nothing 
when it panicked.

> You might try an 8 kernel on your 5 userland and see if that works.

I ended up putting on a -7 kernel and using libkver a lot:

builder# uname -a
NetBSD builder.internal 7.2 NetBSD 7.2 (XEN3PAE_DOMU.201809111420Z) i386
builder# kver -r 5.2_STABLE /bin/tcsh
builder# uname -a
NetBSD builder.internal 5.2_STABLE NetBSD 5.2_STABLE (LIBKVER) #0: Tue Jan 19 00:00:00 UTC 2038 root@localhost:/sys/arch/i386/compile/LIBKVER i386

--
Stephen