]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/commitdiff
KVM: VMX: simplify and fix vmx_vcpu_pi_load
authorPaolo Bonzini <pbonzini@redhat.com>
Tue, 6 Jun 2017 10:57:06 +0000 (12:57 +0200)
committerPaolo Bonzini <pbonzini@redhat.com>
Wed, 27 Sep 2017 11:45:42 +0000 (13:45 +0200)
The simplify part: do not touch pi_desc.nv, we can set it when the
VCPU is first created.  Likewise, pi_desc.sn is only handled by
vmx_vcpu_pi_load, do not touch it in __pi_post_block.

The fix part: do not check kvm_arch_has_assigned_device, instead
check the SN bit to figure out whether vmx_vcpu_pi_put ran before.
This matches what the previous patch did in pi_post_block.

Cc: Huangweidong <weidong.huang@huawei.com>
Cc: Gonglei <arei.gonglei@huawei.com>
Cc: wangxin <wangxinxin.wang@huawei.com>
Cc: Radim Krčmář <rkrcmar@redhat.com>
Tested-by: Longpeng (Mike) <longpeng2@huawei.com>
Cc: stable@vger.kernel.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
arch/x86/kvm/vmx.c

index 0bfe97e50a401263a0a2660e9b83642beb1f8b2b..b9d2140eb2128784c6979482cf153d701f6540d7 100644 (file)
@@ -2202,43 +2202,41 @@ static void vmx_vcpu_pi_load(struct kvm_vcpu *vcpu, int cpu)
        struct pi_desc old, new;
        unsigned int dest;
 
-       if (!kvm_arch_has_assigned_device(vcpu->kvm) ||
-               !irq_remapping_cap(IRQ_POSTING_CAP)  ||
-               !kvm_vcpu_apicv_active(vcpu))
+       /*
+        * In case of hot-plug or hot-unplug, we may have to undo
+        * vmx_vcpu_pi_put even if there is no assigned device.  And we
+        * always keep PI.NDST up to date for simplicity: it makes the
+        * code easier, and CPU migration is not a fast path.
+        */
+       if (!pi_test_sn(pi_desc) && vcpu->cpu == cpu)
+               return;
+
+       /*
+        * First handle the simple case where no cmpxchg is necessary; just
+        * allow posting non-urgent interrupts.
+        *
+        * If the 'nv' field is POSTED_INTR_WAKEUP_VECTOR, do not change
+        * PI.NDST: pi_post_block will do it for us and the wakeup_handler
+        * expects the VCPU to be on the blocked_vcpu_list that matches
+        * PI.NDST.
+        */
+       if (pi_desc->nv == POSTED_INTR_WAKEUP_VECTOR ||
+           vcpu->cpu == cpu) {
+               pi_clear_sn(pi_desc);
                return;
+       }
 
+       /* The full case.  */
        do {
                old.control = new.control = pi_desc->control;
 
-               /*
-                * If 'nv' field is POSTED_INTR_WAKEUP_VECTOR, there
-                * are two possible cases:
-                * 1. After running 'pre_block', context switch
-                *    happened. For this case, 'sn' was set in
-                *    vmx_vcpu_put(), so we need to clear it here.
-                * 2. After running 'pre_block', we were blocked,
-                *    and woken up by some other guy. For this case,
-                *    we don't need to do anything, 'pi_post_block'
-                *    will do everything for us. However, we cannot
-                *    check whether it is case #1 or case #2 here
-                *    (maybe, not needed), so we also clear sn here,
-                *    I think it is not a big deal.
-                */
-               if (pi_desc->nv != POSTED_INTR_WAKEUP_VECTOR) {
-                       if (vcpu->cpu != cpu) {
-                               dest = cpu_physical_id(cpu);
-
-                               if (x2apic_enabled())
-                                       new.ndst = dest;
-                               else
-                                       new.ndst = (dest << 8) & 0xFF00;
-                       }
+               dest = cpu_physical_id(cpu);
 
-                       /* set 'NV' to 'notification vector' */
-                       new.nv = POSTED_INTR_VECTOR;
-               }
+               if (x2apic_enabled())
+                       new.ndst = dest;
+               else
+                       new.ndst = (dest << 8) & 0xFF00;
 
-               /* Allow posting non-urgent interrupts */
                new.sn = 0;
        } while (cmpxchg(&pi_desc->control, old.control,
                        new.control) != old.control);
@@ -9592,6 +9590,13 @@ static struct kvm_vcpu *vmx_create_vcpu(struct kvm *kvm, unsigned int id)
 
        vmx->msr_ia32_feature_control_valid_bits = FEATURE_CONTROL_LOCKED;
 
+       /*
+        * Enforce invariant: pi_desc.nv is always either POSTED_INTR_VECTOR
+        * or POSTED_INTR_WAKEUP_VECTOR.
+        */
+       vmx->pi_desc.nv = POSTED_INTR_VECTOR;
+       vmx->pi_desc.sn = 1;
+
        return &vmx->vcpu;
 
 free_vmcs:
@@ -11723,9 +11728,6 @@ static void __pi_post_block(struct kvm_vcpu *vcpu)
                else
                        new.ndst = (dest << 8) & 0xFF00;
 
-               /* Allow posting non-urgent interrupts */
-               new.sn = 0;
-
                /* set 'NV' to 'notification vector' */
                new.nv = POSTED_INTR_VECTOR;
        } while (cmpxchg(&pi_desc->control, old.control,