svm: Add mutex_lock to protect apic_access_page_done on AMD systems

Commit Message

Suthikulpanit, Suravee Nov. 12, 2018, 12:23 p.m. UTC

From: Wei Wang <[e-mail]>

There is a race condition when accessing kvm->arch.apic_access_page_done.
Due to it, x86_set_memory_region will fail when creating the second vcpu
for a svm guest.

Add a mutex_lock to serialize the accesses to apic_access_page_done.
This lock is also used by vmx for the same purpose.

Signed-off-by: Wei Wang <[e-mail]>
Signed-off-by: Amadeusz Juskowiak <[e-mail]>
Signed-off-by: Julian Stecklina <[e-mail]>
Signed-off-by: Suravee Suthikulpanit <[e-mail]>
---
 arch/x86/kvm/svm.c | 19 +++++++++++--------
 1 file changed, 11 insertions(+), 8 deletions(-)

Comments

Joerg Roedel Nov. 12, 2018, 1:37 p.m. UTC | #1

On Mon, Nov 12, 2018 at 12:23:14PM +0000, Suthikulpanit, Suravee wrote:
> From: Wei Wang <[e-mail]>
> 
> There is a race condition when accessing kvm->arch.apic_access_page_done.
> Due to it, x86_set_memory_region will fail when creating the second vcpu
> for a svm guest.
> 
> Add a mutex_lock to serialize the accesses to apic_access_page_done.
> This lock is also used by vmx for the same purpose.
> 
> Signed-off-by: Wei Wang <[e-mail]>
> Signed-off-by: Amadeusz Juskowiak <[e-mail]>
> Signed-off-by: Julian Stecklina <[e-mail]>
> Signed-off-by: Suravee Suthikulpanit <[e-mail]>

Reviewed-by: Joerg Roedel <[e-mail]>

Paolo Bonzini Nov. 25, 2018, 5:34 p.m. UTC | #2

On 12/11/18 14:37, [e-mail] wrote:
> On Mon, Nov 12, 2018 at 12:23:14PM +0000, Suthikulpanit, Suravee wrote:
>> From: Wei Wang <[e-mail]>
>>
>> There is a race condition when accessing kvm->arch.apic_access_page_done.
>> Due to it, x86_set_memory_region will fail when creating the second vcpu
>> for a svm guest.
>>
>> Add a mutex_lock to serialize the accesses to apic_access_page_done.
>> This lock is also used by vmx for the same purpose.
>>
>> Signed-off-by: Wei Wang <[e-mail]>
>> Signed-off-by: Amadeusz Juskowiak <[e-mail]>
>> Signed-off-by: Julian Stecklina <[e-mail]>
>> Signed-off-by: Suravee Suthikulpanit <[e-mail]>
> Reviewed-by: Joerg Roedel <[e-mail]>

Queued, thanks.

Paolo

Patch

diff --git a/arch/x86/kvm/svm.c b/arch/x86/kvm/svm.c
index d96092b35936..8be1718fed1d 100644
--- a/arch/x86/kvm/svm.c
+++ b/arch/x86/kvm/svm.c
@@ -1658,20 +1658,23 @@  static u64 *avic_get_physical_id_entry(struct kvm_vcpu *vcpu,
 static int avic_init_access_page(struct kvm_vcpu *vcpu)
 {
 	struct kvm *kvm = vcpu->kvm;
-	int ret;
+	int ret = 0;
 
+	mutex_lock(&kvm->slots_lock);
 	if (kvm->arch.apic_access_page_done)
-		return 0;
+		goto out;
 
-	ret = x86_set_memory_region(kvm,
-				    APIC_ACCESS_PAGE_PRIVATE_MEMSLOT,
-				    APIC_DEFAULT_PHYS_BASE,
-				    PAGE_SIZE);
+	ret = __x86_set_memory_region(kvm,
+				      APIC_ACCESS_PAGE_PRIVATE_MEMSLOT,
+				      APIC_DEFAULT_PHYS_BASE,
+				      PAGE_SIZE);
 	if (ret)
-		return ret;
+		goto out;
 
 	kvm->arch.apic_access_page_done = true;
-	return 0;
+out:
+	mutex_unlock(&kvm->slots_lock);
+	return ret;
 }
 
 static int avic_init_backing_page(struct kvm_vcpu *vcpu)