Skip to contentSkip to conversation

patch kvm-svm-fix-tsc-offset-adjustment-when-running-nested.patch added to 2.6.31-stable tree

DISCUSSION Stefan Lippers-Hollmann 4 participants 5 messages Oct 10, 2009 – Oct 12, 2009 1 message not in the archive

Subsystems (2)
lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

#1 lore raw

Hi

Quoted 7 lines · On Saturday 10 October 2009, [email protected] wrote:

This is a note to let you know that we have just queued up the patch titled

Subject: KVM: SVM: Fix tsc offset adjustment when running nested

to the 2.6.31-stable tree. Its filename is

kvm-svm-fix-tsc-offset-adjustment-when-running-nested.patch

This patch breaks compilation of 2.6.31.4-rc1 and 2.6.31.4-rc2, as it seems to depend on several prior optimisations to kvm's/ svm's nested behaviour:

[...]
CC [M] arch/x86/kvm/svm.o
arch/x86/kvm/svm.c: In function 'svm_vcpu_load':
arch/x86/kvm/svm.c:713: error: 'struct vcpu_svm' has no member named 'nested' make[2]: *** [arch/x86/kvm/svm.o] Error 1
make[1]: *** [arch/x86/kvm] Error 2
make: *** [arch/x86] Error 2

Regards
Stefan Lippers-Hollmann

Quoted 39 lines

A git repo of this tree can be found at
http://www.kernel.org/git/?p=linux/kernel/git/stable/stable-queue.git;a=summary

From 77b1ab1732feb5e3dcbaf31d2f7547c5229f5f3a Mon Sep 17 00:00:00 2001
From: Joerg Roedel <[email protected]>
Date: Wed, 16 Sep 2009 15:24:17 +0200
Subject: KVM: SVM: Fix tsc offset adjustment when running nested

From: Joerg Roedel <[email protected]>

commit 77b1ab1732feb5e3dcbaf31d2f7547c5229f5f3a upstream.

When svm_vcpu_load is called while the vcpu is running in guest mode the tsc adjustment made there is lost on the next emulated #vmexit. This causes the tsc running backwards in the guest. This patch fixes the issue by also adjusting the tsc_offset in the emulated hsave area so that it will not get lost.

---
 arch/x86/kvm/svm.c |    2 ++
 1 file changed, 2 insertions(+)
arch/x86/kvm/svm.c +2 -0
@@ -709,6 +709,8 @@ static void svm_vcpu_load(struct kvm_vcp
 		rdtscll(tsc_this);
 		delta = vcpu->arch.host_tsc - tsc_this;
 		svm->vmcb->control.tsc_offset += delta;
+		if (is_nested(svm))
+			svm->nested.hsave->control.tsc_offset += delta;
 		vcpu->cpu = cpu;
 		kvm_migrate_timers(vcpu);
 		svm->asid_generation = 0;
#2 lore raw
Quoted 23 lines · Stefan Lippers-Hollmann

Hi

Quoted 7 lines · On Saturday 10 October 2009, [email protected] wrote:

This is a note to let you know that we have just queued up the patch titled

Subject: KVM: SVM: Fix tsc offset adjustment when running nested

to the 2.6.31-stable tree. Its filename is

kvm-svm-fix-tsc-offset-adjustment-when-running-nested.patch

This patch breaks compilation of 2.6.31.4-rc1 and 2.6.31.4-rc2, as it seems to depend on several prior optimisations to kvm's/ svm's nested behaviour:

[...]
CC [M] arch/x86/kvm/svm.o
arch/x86/kvm/svm.c: In function 'svm_vcpu_load':
arch/x86/kvm/svm.c:713: error: 'struct vcpu_svm' has no member named 'nested' make[2]: *** [arch/x86/kvm/svm.o] Error 1
make[1]: *** [arch/x86/kvm] Error 2
make: *** [arch/x86] Error 2

Quoted 2 lines

Regards
Stefan Lippers-Hollmann

Greg,

Please drop the patch.

Thanks

#3 lore raw
In reply to Stefan Lippers-Hollmann

All the nested stuff was moved into its own sub-struct in 2.6.32. Here is a version that works with 2.6.31.3:

Quoted 1 line

From 4505a73070f341674e0d22a8c9ee51b2beeb843e Mon Sep 17 00:00:00 2001

From: Joerg Roedel <[email protected]>
Date: Mon, 12 Oct 2009 11:19:07 +0200
Subject: [PATCH 1/2] KVM: SVM: Fix tsc offset adjustment when running nested

When svm_vcpu_load is called while the vcpu is running in guest mode the tsc adjustment made there is lost on the next emulated #vmexit. This causes the tsc running backwards in the guest. This patch fixes the issue by also adjusting the tsc_offset in the emulated hsave area so that it will not get lost.

---
 arch/x86/kvm/svm.c |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)
arch/x86/kvm/svm.c +2 -0
index b1f658a..aafd8bf 100644
@@ -709,6 +709,8 @@ static void svm_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
 		rdtscll(tsc_this);
 		delta = vcpu->arch.host_tsc - tsc_this;
 		svm->vmcb->control.tsc_offset += delta;
+		if (is_nested(svm))
+			svm->hsave->control.tsc_offset += delta;
 		vcpu->cpu = cpu;
 		kvm_migrate_timers(vcpu);
 		svm->asid_generation = 0;
Signature · 1 line
1.6.4.3
#4 lore raw
Quoted 8 lines · Joerg Roedel

All the nested stuff was moved into its own sub-struct in 2.6.32. Here is a version that works with 2.6.31.3:

Quoted 1 line

From 4505a73070f341674e0d22a8c9ee51b2beeb843e Mon Sep 17 00:00:00 2001

From: Joerg Roedel <[email protected]>
Date: Mon, 12 Oct 2009 11:19:07 +0200
Subject: [PATCH 1/2] KVM: SVM: Fix tsc offset adjustment when running nested

Quoted 8 lines

When svm_vcpu_load is called while the vcpu is running in guest mode the tsc adjustment made there is lost on the next emulated #vmexit. This causes the tsc running backwards in the guest. This patch fixes the issue by also adjusting the tsc_offset in the emulated hsave area so that it will not get lost.

Thanks, but Marcelo just said we should drop the original patch.

Marcelo, if you want this one in instead, please let me know.

thanks,

greg k-h

#5 lore raw
In reply to Joerg Roedel
Quoted 4 lines · Joerg Roedel

All the nested stuff was moved into its own sub-struct in 2.6.32. Here is a version that works with 2.6.31.3:

Quoted 1 line

From 4505a73070f341674e0d22a8c9ee51b2beeb843e Mon Sep 17 00:00:00 2001

Thanks, I've replaced the original patch with this version.

greg k-h