[Xen-devel] [PATCH v9 7/8] VMX: Fixup PI descriptor when cpu is offline

[Chao Gao <chao.gao@xxxxxxxxx>]

From: Feng Wu <feng.wu@xxxxxxxxx>

When cpu is offline, we need to move all the vcpus in its blocking
list to another online cpu, this patch handles it.

Signed-off-by: Feng Wu <feng.wu@xxxxxxxxx>
Signed-off-by: Chao Gao <chao.gao@xxxxxxxxx>
Reviewed-by: Jan Beulich <jbeulich@xxxxxxxx>
Acked-by: Kevin Tian <kevin.tian@xxxxxxxxx>
---
v7: 
- Pass unsigned int to vmx_pi_desc_fixup()

v6: 
- Carefully suppress 'SN' to avoid missing notification event
during moving the vcpu to the new list

v5: 
- Add some comments to explain why it doesn't cause deadlock
for the ABBA deadlock scenario. 

v4: 
- Remove the pointless check since we are in machine stop
context and no other cpus go down in parallel.

 xen/arch/x86/hvm/vmx/vmcs.c       |  1 +
 xen/arch/x86/hvm/vmx/vmx.c        | 70 +++++++++++++++++++++++++++++++++++++++
 xen/include/asm-x86/hvm/vmx/vmx.h |  1 +
 3 files changed, 72 insertions(+)

diff --git a/xen/arch/x86/hvm/vmx/vmcs.c b/xen/arch/x86/hvm/vmx/vmcs.c
index 7905d3e..7e3e093 100644
--- a/xen/arch/x86/hvm/vmx/vmcs.c
+++ b/xen/arch/x86/hvm/vmx/vmcs.c
@@ -578,6 +578,7 @@ void vmx_cpu_dead(unsigned int cpu)
     vmx_free_vmcs(per_cpu(vmxon_region, cpu));
     per_cpu(vmxon_region, cpu) = 0;
     nvmx_cpu_dead(cpu);
+    vmx_pi_desc_fixup(cpu);
 }
 
 int vmx_cpu_up(void)
diff --git a/xen/arch/x86/hvm/vmx/vmx.c b/xen/arch/x86/hvm/vmx/vmx.c
index e03786b..b8a385b 100644
--- a/xen/arch/x86/hvm/vmx/vmx.c
+++ b/xen/arch/x86/hvm/vmx/vmx.c
@@ -200,6 +200,76 @@ static void vmx_pi_do_resume(struct vcpu *v)
     vmx_pi_unblock_vcpu(v);
 }
 
+void vmx_pi_desc_fixup(unsigned int cpu)
+{
+    unsigned int new_cpu, dest;
+    unsigned long flags;
+    struct arch_vmx_struct *vmx, *tmp;
+    spinlock_t *new_lock, *old_lock = &per_cpu(vmx_pi_blocking, cpu).lock;
+    struct list_head *blocked_vcpus = &per_cpu(vmx_pi_blocking, cpu).list;
+
+    if ( !iommu_intpost )
+        return;
+
+    /*
+     * We are in the context of CPU_DEAD or CPU_UP_CANCELED notification,
+     * and it is impossible for a second CPU go down in parallel. So we
+     * can safely acquire the old cpu's lock and then acquire the new_cpu's
+     * lock after that.
+     */
+    spin_lock_irqsave(old_lock, flags);
+
+    list_for_each_entry_safe(vmx, tmp, blocked_vcpus, pi_blocking.list)
+    {
+        /*
+         * Suppress notification or we may miss an interrupt when the
+         * target cpu is dying.
+         */
+        pi_set_sn(&vmx->pi_desc);
+
+        /*
+         * Check whether a notification is pending before doing the
+         * movement, if that is the case we need to wake up it directly
+         * other than moving it to the new cpu's list.
+         */
+        if ( pi_test_on(&vmx->pi_desc) )
+        {
+            list_del(&vmx->pi_blocking.list);
+            vmx->pi_blocking.lock = NULL;
+            vcpu_unblock(container_of(vmx, struct vcpu, arch.hvm_vmx));
+        }
+        else
+        {
+            /*
+             * We need to find an online cpu as the NDST of the PI descriptor, 
it
+             * doesn't matter whether it is within the cpupool of the domain or
+             * not. As long as it is online, the vCPU will be woken up once the
+             * notification event arrives.
+             */
+            new_cpu = cpumask_any(&cpu_online_map);
+            new_lock = &per_cpu(vmx_pi_blocking, new_cpu).lock;
+
+            spin_lock(new_lock);
+
+            ASSERT(vmx->pi_blocking.lock == old_lock);
+
+            dest = cpu_physical_id(new_cpu);
+            write_atomic(&vmx->pi_desc.ndst,
+                         x2apic_enabled ? dest : MASK_INSR(dest, 
PI_xAPIC_NDST_MASK));
+
+            list_move(&vmx->pi_blocking.list,
+                      &per_cpu(vmx_pi_blocking, new_cpu).list);
+            vmx->pi_blocking.lock = new_lock;
+
+            spin_unlock(new_lock);
+        }
+
+        pi_clear_sn(&vmx->pi_desc);
+    }
+
+    spin_unlock_irqrestore(old_lock, flags);
+}
+
 /*
  * To handle posted interrupts correctly, we need to set the following
  * state:
diff --git a/xen/include/asm-x86/hvm/vmx/vmx.h 
b/xen/include/asm-x86/hvm/vmx/vmx.h
index 00e6f0d..70155fb 100644
--- a/xen/include/asm-x86/hvm/vmx/vmx.h
+++ b/xen/include/asm-x86/hvm/vmx/vmx.h
@@ -597,6 +597,7 @@ void free_p2m_hap_data(struct p2m_domain *p2m);
 void p2m_init_hap_data(struct p2m_domain *p2m);
 
 void vmx_pi_per_cpu_init(unsigned int cpu);
+void vmx_pi_desc_fixup(unsigned int cpu);
 
 void vmx_pi_hooks_assign(struct domain *d);
 void vmx_pi_hooks_deassign(struct domain *d);
-- 
1.8.3.1


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
https://lists.xen.org/xen-devel