[HVM] When guest explicitly modifies the TSC, update platform timers

_after_ updating guest time (since otherwise pt_reset() will get the
old hvm_get_guest_time() and subsequently pt_intr_post() will warp
the TSC back again).

This is the smallest fix for the HVM SMP TSC issues seen on recent
(viz. 2.6.18+) linux HVM kernels, but probably not the best. Cleaner
fix needs more thought.

Signed-off-by: Steven Hand <steven@xensource.com>

diff --git a/xen/arch/x86/hvm/svm/svm.c b/xen/arch/x86/hvm/svm/svm.c
index 7a656865eb53616da794b5e522ee844d47fa4b58..ff2d46896ffafc48900192018aad36113b34ad6d 100644 (file)
--- a/xen/arch/x86/hvm/svm/svm.c
+++ b/xen/arch/x86/hvm/svm/svm.c
@@ -2141,8 +2141,8 @@ static inline void svm_do_msr_access(
         switch (ecx)
         {
         case MSR_IA32_TIME_STAMP_COUNTER:
-            pt_reset(v);
             hvm_set_guest_time(v, msr_content);
+            pt_reset(v);
             break;
         case MSR_IA32_SYSENTER_CS:
             vmcb->sysenter_cs = msr_content;

diff --git a/xen/arch/x86/hvm/vmx/vmx.c b/xen/arch/x86/hvm/vmx/vmx.c
index 016fd0926a6b1b40e63b5309dfc35a14e49e7d70..f203ed9107750d7d9682c33163358d71052eb3f3 100644 (file)
--- a/xen/arch/x86/hvm/vmx/vmx.c
+++ b/xen/arch/x86/hvm/vmx/vmx.c
@@ -2346,8 +2346,8 @@ static inline int vmx_do_msr_write(struct cpu_user_regs *regs)
 
     switch (ecx) {
     case MSR_IA32_TIME_STAMP_COUNTER:
-        pt_reset(v);
         hvm_set_guest_time(v, msr_content);
+        pt_reset(v);
         break;
     case MSR_IA32_SYSENTER_CS:
         __vmwrite(GUEST_SYSENTER_CS, msr_content);