Re: [XEN PATCH 1/2] hvm: Support more than 32 VCPUS when migrating.

From: "Jan Beulich" <JBeulich@suse.com>
To: "Konrad Rzeszutek Wilk" <konrad.wilk@oracle.com>
Cc: <david.vrabel@Citrix.com>, <konrad@kernel.org>,
	<xen-devel@lists.xenproject.org>, <boris.ostrovsky@oracle.com>,
	<linux-kernel@vger.kernel.org>, <keir@xen.org>
Subject: Re: [XEN PATCH 1/2] hvm: Support more than 32 VCPUS when migrating.
Date: Wed, 23 Apr 2014 09:57:40 +0100	[thread overview]
Message-ID: <53579CA4020000780000B342@nat28.tlf.novell.com> (raw)
In-Reply-To: <20140422183443.GA6817@phenom.dumpdata.com>

>>> On 22.04.14 at 20:34, <konrad.wilk@oracle.com> wrote:
> With this patch:
> [...]
> And with an HVM guest poking at the rest of VCPUOPs: VCPUOP_get_physid, 
> VCPUOP_initialise, and VCPUOP_send_nmi - either before the CPU is up or
> when it is up - work.
> 
> That is: the VCPUOP_get_physid would return -EINVAL; VCPUOP_initialise
> would return either -EEXIST or 0, and in either case
> the content of the ctxt was full of zeros.

Good.

> The VCPUOP_send_nmi did cause the HVM to get an NMI and it spitted out
> 'Dazed and confused'. It also noticed corruption:
> 
> [    3.611742] Corrupted low memory at c000fffc (fffc phys) = 00029b00
> [    2.386785] Corrupted low memory at ffff88000000fff8 (fff8 phys) = 
> 2990000000000
> 
> Which is odd because there does not seem to be anything in the path
> of hypervisor that would cause this.

Indeed. This looks a little like a segment descriptor got modified here
with a descriptor table base of zero and a selector of 0xfff8. That
corruption needs to be hunted down in any case before enabling
VCPUOP_send_nmi for HVM.

Jan