From: Chris Lalancette <clalance@redhat.com>
Date: Wed, 8 Apr 2009 10:52:42 +0200
Subject: [xen] x86: update the earlier APERF/MPERF patch
Message-id: 49DC65DA.6070709@redhat.com
O-Subject: [RHEL5.4 PATCH]: Update the earlier APERF/MPERF patch
Bugzilla: 493557
RH-Acked-by: Gerd Hoffmann <kraxel@redhat.com>
RH-Acked-by: Rik van Riel <riel@redhat.com>

All,
     The patch to fix Xen APERF/MPERF updates that I submitted (and was
subsequently committed to -138) had a slight bug in it.  Basically, if you used
the APERF/MPERF writes in conjunction with a "dom0_max_cpus" setting less than
the number of physical CPUs in the box, you would get the same crash as pointed
out in BZ 485119.
     This patch just updates the APERF/MPERF code path so it will not crash in
that situation.  To avoid confusion, I've opened up BZ 493557 to track it, which
is what this patch should be committed against.
     This has been reviewed and tested by the reporter, and solves the issue for
them.
     Please review and ACK.

--
Chris Lalancette

diff --git a/arch/x86/traps.c b/arch/x86/traps.c
index b6bf936..1f8d269 100644
--- a/arch/x86/traps.c
+++ b/arch/x86/traps.c
@@ -1810,6 +1810,8 @@ static int emulate_privileged_op(struct cpu_user_regs *regs)
             if ( wrmsr_safe(regs->ecx, eax, edx) != 0 )
                 goto fail;
             break;
+        case MSR_IA32_MPERF:
+        case MSR_IA32_APERF:
         case MSR_IA32_PERF_CTL:
             if ( boot_cpu_data.x86_vendor != X86_VENDOR_INTEL )
                 goto fail;
@@ -1818,12 +1820,6 @@ static int emulate_privileged_op(struct cpu_user_regs *regs)
             if ( wrmsr_safe(regs->ecx, eax, edx) != 0 )
                 goto fail;
             break;
-	case MSR_IA32_MPERF:
-	case MSR_IA32_APERF:
-            if ( (cpufreq_controller != FREQCTL_dom0_kernel) ||
-                 wrmsr_safe(regs->ecx, eax, edx) )
-                goto fail;
-            break;
         default:
             if ( wrmsr_hypervisor_regs(regs->ecx, eax, edx) )
                 break;