KVM: x86: switch kvm_set_memory_alias to SRCU update

Using a similar two-step procedure as for memslots.

Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>

4 files changed, 63 insertions, 11 deletions

diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
index 7cdcb3d0f770..6c8c7c578c46 100644
--- a/arch/x86/include/asm/kvm_host.h
+++ b/arch/x86/include/asm/kvm_host.h
@@ -368,8 +368,12 @@ struct kvm_mem_alias {
         gfn_t base_gfn;
         unsigned long npages;
         gfn_t target_gfn;
+#define KVM_ALIAS_INVALID     1UL
+        unsigned long flags;
 };
 
+#define KVM_ARCH_HAS_UNALIAS_INSTANTIATION
+
 struct kvm_mem_aliases {
         struct kvm_mem_alias aliases[KVM_ALIAS_SLOTS];
         int naliases;
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index e7488350ca16..28127c936c3b 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -38,6 +38,7 @@
 #include <linux/intel-iommu.h>
 #include <linux/cpufreq.h>
 #include <linux/user-return-notifier.h>
+#include <linux/srcu.h>
 #include <trace/events/kvm.h>
 #undef TRACE_INCLUDE_FILE
 #define CREATE_TRACE_POINTS
@@ -2223,11 +2224,32 @@ static int kvm_vm_ioctl_get_nr_mmu_pages(struct kvm *kvm)
         return kvm->arch.n_alloc_mmu_pages;
 }
 
+gfn_t unalias_gfn_instantiation(struct kvm *kvm, gfn_t gfn)
+{
+        int i;
+        struct kvm_mem_alias *alias;
+        struct kvm_mem_aliases *aliases;
+
+        aliases = rcu_dereference(kvm->arch.aliases);
+
+        for (i = 0; i < aliases->naliases; ++i) {
+                alias = &aliases->aliases[i];
+                if (alias->flags & KVM_ALIAS_INVALID)
+                        continue;
+                if (gfn >= alias->base_gfn
+                    && gfn < alias->base_gfn + alias->npages)
+                        return alias->target_gfn + gfn - alias->base_gfn;
+        }
+        return gfn;
+}
+
 gfn_t unalias_gfn(struct kvm *kvm, gfn_t gfn)
 {
         int i;
         struct kvm_mem_alias *alias;
-        struct kvm_mem_aliases *aliases = kvm->arch.aliases;
+        struct kvm_mem_aliases *aliases;
+
+        aliases = rcu_dereference(kvm->arch.aliases);
 
         for (i = 0; i < aliases->naliases; ++i) {
                 alias = &aliases->aliases[i];
@@ -2248,7 +2270,7 @@ static int kvm_vm_ioctl_set_memory_alias(struct kvm *kvm,
 {
         int r, n;
         struct kvm_mem_alias *p;
-        struct kvm_mem_aliases *aliases;
+        struct kvm_mem_aliases *aliases, *old_aliases;
 
         r = -EINVAL;
         /* General sanity checks */
@@ -2265,28 +2287,48 @@ static int kvm_vm_ioctl_set_memory_alias(struct kvm *kvm,
             < alias->target_phys_addr)
                 goto out;
 
+        r = -ENOMEM;
+        aliases = kzalloc(sizeof(struct kvm_mem_aliases), GFP_KERNEL);
+        if (!aliases)
+                goto out;
+
         down_write(&kvm->slots_lock);
-        spin_lock(&kvm->mmu_lock);
 
-        aliases = kvm->arch.aliases;
+        /* invalidate any gfn reference in case of deletion/shrinking */
+        memcpy(aliases, kvm->arch.aliases, sizeof(struct kvm_mem_aliases));
+        aliases->aliases[alias->slot].flags |= KVM_ALIAS_INVALID;
+        old_aliases = kvm->arch.aliases;
+        rcu_assign_pointer(kvm->arch.aliases, aliases);
+        synchronize_srcu_expedited(&kvm->srcu);
+        kvm_mmu_zap_all(kvm);
+        kfree(old_aliases);
+
+        r = -ENOMEM;
+        aliases = kzalloc(sizeof(struct kvm_mem_aliases), GFP_KERNEL);
+        if (!aliases)
+                goto out_unlock;
+
+        memcpy(aliases, kvm->arch.aliases, sizeof(struct kvm_mem_aliases));
 
         p = &aliases->aliases[alias->slot];
         p->base_gfn = alias->guest_phys_addr >> PAGE_SHIFT;
         p->npages = alias->memory_size >> PAGE_SHIFT;
         p->target_gfn = alias->target_phys_addr >> PAGE_SHIFT;
+        p->flags &= ~(KVM_ALIAS_INVALID);
 
         for (n = KVM_ALIAS_SLOTS; n > 0; --n)
                 if (aliases->aliases[n - 1].npages)
                         break;
         aliases->naliases = n;
 
-        spin_unlock(&kvm->mmu_lock);
-        kvm_mmu_zap_all(kvm);
+        old_aliases = kvm->arch.aliases;
+        rcu_assign_pointer(kvm->arch.aliases, aliases);
+        synchronize_srcu_expedited(&kvm->srcu);
+        kfree(old_aliases);
+        r = 0;
 
+out_unlock:
         up_write(&kvm->slots_lock);
-
-        return 0;
-
 out:
         return r;
 }
diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
index 93bd30701ca7..20941c0f4045 100644
--- a/include/linux/kvm_host.h
+++ b/include/linux/kvm_host.h
@@ -266,6 +266,8 @@ void kvm_arch_commit_memory_region(struct kvm *kvm,
 void kvm_disable_largepages(void);
 void kvm_arch_flush_shadow(struct kvm *kvm);
 gfn_t unalias_gfn(struct kvm *kvm, gfn_t gfn);
+gfn_t unalias_gfn_instantiation(struct kvm *kvm, gfn_t gfn);
+
 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn);
 void kvm_release_page_clean(struct page *page);
@@ -539,6 +541,10 @@ static inline int mmu_notifier_retry(struct kvm_vcpu *vcpu, unsigned long mmu_se
 }
 #endif
 
+#ifndef KVM_ARCH_HAS_UNALIAS_INSTANTIATION
+#define unalias_gfn_instantiation unalias_gfn
+#endif
+
 #ifdef CONFIG_HAVE_KVM_IRQCHIP
 
 #define KVM_MAX_IRQ_ROUTES 1024
diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
index 2bb24a814fdf..c680f7b64c6f 100644
--- a/virt/kvm/kvm_main.c
+++ b/virt/kvm/kvm_main.c
@@ -859,7 +859,7 @@ int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn)
         int i;
         struct kvm_memslots *slots = rcu_dereference(kvm->memslots);
 
-        gfn = unalias_gfn(kvm, gfn);
+        gfn = unalias_gfn_instantiation(kvm, gfn);
         for (i = 0; i < KVM_MEMORY_SLOTS; ++i) {
                 struct kvm_memory_slot *memslot = &slots->memslots[i];
 
@@ -896,7 +896,7 @@ unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn)
 {
         struct kvm_memory_slot *slot;
 
-        gfn = unalias_gfn(kvm, gfn);
+        gfn = unalias_gfn_instantiation(kvm, gfn);
         slot = gfn_to_memslot_unaliased(kvm, gfn);
         if (!slot || slot->flags & KVM_MEMSLOT_INVALID)
                 return bad_hva();