[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Xen-devel] [PATCH v2 for 4.5] x86/VPMU: Clear last_vcpu when destroying VPMU

We need to make sure that last_vcpu is not pointing to VCPU whose
VPMU is being destroyed. Otherwise we may try to dereference it in
the future, when VCPU is gone.

We have to do this via IPI since otherwise there is a (somewheat
theoretical) chance that between test and subsequent clearing
of last_vcpu the remote processor (i.e. vpmu->last_pcpu) might do
both load_vpmu() and then save_vpmu() for another VCPU. The former
will clear last_vcpu and the latter will set it to something else.

Performing this operation via IPI will guarantee that nothing can
happen on the remote processor between testing and clearing of

Signed-off-by: Boris Ostrovsky <boris.ostrovsky@xxxxxxxxxx>
 xen/arch/x86/hvm/vpmu.c |   20 ++++++++++++++++++++
 1 files changed, 20 insertions(+), 0 deletions(-)

Changes in v2:
 * Test last_vcpu locally before IPI
 * Don't handle local pcpu as a special case --- on_selected_cpus
   will take care of that
 * Dont't cast variables unnecessarily

diff --git a/xen/arch/x86/hvm/vpmu.c b/xen/arch/x86/hvm/vpmu.c
index 1df74c2..37f0d9f 100644
--- a/xen/arch/x86/hvm/vpmu.c
+++ b/xen/arch/x86/hvm/vpmu.c
@@ -247,10 +247,30 @@ void vpmu_initialise(struct vcpu *v)
+static void vpmu_clear_last(void *arg)
+    if ( this_cpu(last_vcpu) == arg )
+        this_cpu(last_vcpu) = NULL;
 void vpmu_destroy(struct vcpu *v)
     struct vpmu_struct *vpmu = vcpu_vpmu(v);
+    if ( !vpmu_is_set(vpmu, VPMU_CONTEXT_ALLOCATED) )
+        return;
+    /*
+     * Need to clear last_vcpu in case it points to v.
+     * We can check here non-atomically whether it is 'v' since
+     * last_vcpu can never become 'v' again at this point.
+     * We will test it again in vpmu_clear_last() with interrupts
+     * disabled to make sure we don't clear someone else.
+     */
+    if ( per_cpu(last_vcpu, vpmu->last_pcpu) == v )
+        on_selected_cpus(cpumask_of(vpmu->last_pcpu),
+                         vpmu_clear_last, v, 1);
     if ( vpmu->arch_vpmu_ops && vpmu->arch_vpmu_ops->arch_vpmu_destroy )

Xen-devel mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.