kvm: optimize out smp_mb after srcu_read_unlock

[~andy/linux] / arch / x86 / kvm / x86.c

diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index d00d88455fef11652ff37ffca1141e1217f13416..21ef1ba184ae8500a70061f566ea55fde76cbfd2 100644 (file)
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -577,6 +577,7 @@ static void kvm_put_guest_xcr0(struct kvm_vcpu *vcpu)
 int __kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr)
 {
        u64 xcr0;
+       u64 valid_bits;
 
        /* Only support XCR_XFEATURE_ENABLED_MASK(xcr0) now  */
        if (index != XCR_XFEATURE_ENABLED_MASK)
@@ -586,8 +587,16 @@ int __kvm_set_xcr(struct kvm_vcpu *vcpu, u32 index, u64 xcr)
                return 1;
        if ((xcr0 & XSTATE_YMM) && !(xcr0 & XSTATE_SSE))
                return 1;
-       if (xcr0 & ~vcpu->arch.guest_supported_xcr0)
+
+       /*
+        * Do not allow the guest to set bits that we do not support
+        * saving.  However, xcr0 bit 0 is always set, even if the
+        * emulated CPU does not support XSAVE (see fx_init).
+        */
+       valid_bits = vcpu->arch.guest_supported_xcr0 | XSTATE_FP;
+       if (xcr0 & ~valid_bits)
                return 1;
+
        kvm_put_guest_xcr0(vcpu);
        vcpu->arch.xcr0 = xcr0;
        return 0;
@@ -2718,8 +2727,7 @@ static void wbinvd_ipi(void *garbage)
 
 static bool need_emulate_wbinvd(struct kvm_vcpu *vcpu)
 {
-       return vcpu->kvm->arch.iommu_domain &&
-               !(vcpu->kvm->arch.iommu_flags & KVM_IOMMU_CACHE_COHERENCY);
+       return kvm_arch_has_noncoherent_dma(vcpu->kvm);
 }
 
 void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
@@ -3056,9 +3064,9 @@ static int kvm_vcpu_ioctl_x86_set_xcrs(struct kvm_vcpu *vcpu,
 
        for (i = 0; i < guest_xcrs->nr_xcrs; i++)
                /* Only support XCR0 currently */
-               if (guest_xcrs->xcrs[0].xcr == XCR_XFEATURE_ENABLED_MASK) {
+               if (guest_xcrs->xcrs[i].xcr == XCR_XFEATURE_ENABLED_MASK) {
                        r = __kvm_set_xcr(vcpu, XCR_XFEATURE_ENABLED_MASK,
-                               guest_xcrs->xcrs[0].value);
+                               guest_xcrs->xcrs[i].value);
                        break;
                }
        if (r)
@@ -5108,8 +5116,7 @@ int x86_emulate_instruction(struct kvm_vcpu *vcpu,
                ctxt->have_exception = false;
                ctxt->perm_ok = false;
 
-               ctxt->only_vendor_specific_insn
-                       = emulation_type & EMULTYPE_TRAP_UD;
+               ctxt->ud = emulation_type & EMULTYPE_TRAP_UD;
 
                r = x86_decode_insn(ctxt, insn, insn_len);
 
@@ -5959,10 +5966,12 @@ static int vcpu_enter_guest(struct kvm_vcpu *vcpu)
 
        vcpu->mode = IN_GUEST_MODE;
 
+       srcu_read_unlock(&vcpu->kvm->srcu, vcpu->srcu_idx);
+
        /* We should set ->mode before check ->requests,
         * see the comment in make_all_cpus_request.
         */
-       smp_mb();
+       smp_mb__after_srcu_read_unlock();
 
        local_irq_disable();
 
@@ -5972,12 +5981,11 @@ static int vcpu_enter_guest(struct kvm_vcpu *vcpu)
                smp_wmb();
                local_irq_enable();
                preempt_enable();
+               vcpu->srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
                r = 1;
                goto cancel_injection;
        }
 
-       srcu_read_unlock(&vcpu->kvm->srcu, vcpu->srcu_idx);
-
        if (req_immediate_exit)
                smp_send_reschedule(vcpu->cpu);
 
@@ -6999,6 +7007,7 @@ int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)
        INIT_LIST_HEAD(&kvm->arch.active_mmu_pages);
        INIT_LIST_HEAD(&kvm->arch.zapped_obsolete_pages);
        INIT_LIST_HEAD(&kvm->arch.assigned_dev_head);
+       atomic_set(&kvm->arch.noncoherent_dma_count, 0);
 
        /* Reserve bit 0 of irq_sources_bitmap for userspace irq source */
        set_bit(KVM_USERSPACE_IRQ_SOURCE_ID, &kvm->arch.irq_sources_bitmap);
@@ -7083,7 +7092,7 @@ void kvm_arch_destroy_vm(struct kvm *kvm)
        kfree(rcu_dereference_check(kvm->arch.apic_map, 1));
 }
 
-void kvm_arch_free_memslot(struct kvm_memory_slot *free,
+void kvm_arch_free_memslot(struct kvm *kvm, struct kvm_memory_slot *free,
                           struct kvm_memory_slot *dont)
 {
        int i;
@@ -7104,7 +7113,8 @@ void kvm_arch_free_memslot(struct kvm_memory_slot *free,
        }
 }
 
-int kvm_arch_create_memslot(struct kvm_memory_slot *slot, unsigned long npages)
+int kvm_arch_create_memslot(struct kvm *kvm, struct kvm_memory_slot *slot,
+                           unsigned long npages)
 {
        int i;
 
@@ -7438,6 +7448,24 @@ bool kvm_arch_can_inject_async_page_present(struct kvm_vcpu *vcpu)
                        kvm_x86_ops->interrupt_allowed(vcpu);
 }
 
+void kvm_arch_register_noncoherent_dma(struct kvm *kvm)
+{
+       atomic_inc(&kvm->arch.noncoherent_dma_count);
+}
+EXPORT_SYMBOL_GPL(kvm_arch_register_noncoherent_dma);
+
+void kvm_arch_unregister_noncoherent_dma(struct kvm *kvm)
+{
+       atomic_dec(&kvm->arch.noncoherent_dma_count);
+}
+EXPORT_SYMBOL_GPL(kvm_arch_unregister_noncoherent_dma);
+
+bool kvm_arch_has_noncoherent_dma(struct kvm *kvm)
+{
+       return atomic_read(&kvm->arch.noncoherent_dma_count);
+}
+EXPORT_SYMBOL_GPL(kvm_arch_has_noncoherent_dma);
+
 EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_exit);
 EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_inj_virq);
 EXPORT_TRACEPOINT_SYMBOL_GPL(kvm_page_fault);