[PATCH] Fixes for more than 32 VCPUs migration for HVM guests (v1).

From: konrad@kernel.org
To: xen-devel@lists.xenproject.org, david.vrabel@Citrix.com,
	boris.ostrovsky@oracle.com, linux-kernel@vger.kernel.org,
	keir@xen.org, jbeulich@suse.com
Subject: [PATCH] Fixes for more than 32 VCPUs migration for HVM guests (v1).
Date: Tue,  8 Apr 2014 13:25:48 -0400	[thread overview]
Message-ID: <1396977950-8789-1-git-send-email-konrad__5415.71787694028$1396977992$gmane$org@kernel.org> (raw)
In-Reply-To: <1396859560.22845.4.camel@kazak.uk.xensource.com>

These two patches (one for Linux, one for Xen) allow PVHVM guests to use
the per-cpu VCPU mechanism after migration. Currently when an PVHVM guest
migrates all the per-cpu information is lost and we fallback on the
shared_info structure. This is regardless if the HVM guest has 2 or 128 CPUs.
Since the structure has an array for only 32 CPUs that means if we are
to migrate a PVHVM guest - we can only do it up to 32 CPUs.

These patches fix it and allow more than 32 VCPUs to be migrated with
PVHVM Linux guests.

The Linux diff is:

 arch/x86/xen/enlighten.c |   21 ++++++++++++++++-----
 arch/x86/xen/suspend.c   |    6 +-----
 arch/x86/xen/time.c      |    3 +++
 3 files changed, 20 insertions(+), 10 deletions(-)

while the Xen one is:

 xen/arch/x86/hvm/hvm.c |    3 +++
 1 files changed, 3 insertions(+), 0 deletions(-)