[patch 00/11] convert slotslock to SRCU

[patch 00/11] convert slotslock to SRCU

[Marcelo Tosatti]

Now that synchronize_srcu_expedited is in the tree, we can continue the 
convertion of slots_lock to SRCU.

Results:
up:

srcu-memslot:
cpuid 2157
vmcall 1992
mov_from_cr8 11
mov_to_cr8 17
inl_from_pmtimer 11226
ipi (skipped)
ipi+halt (skipped)

stock:
cpuid 2272
vmcall 2114
mov_from_cr8 11
mov_to_cr8 17
inl_from_pmtimer 11521
ipi (skipped)
ipi+halt (skipped)

vmcall 6.1% improvement

smp 2:

srcu-memslot:
cpuid 2333
vmcall 2165
mov_from_cr8 11
mov_to_cr8 18
inl_from_pmtimer 12233
ipi 40734
ipi+halt 41675

stock:
cpuid 2512
vmcall 2367
mov_from_cr8 11
mov_to_cr8 18
inl_from_pmtimer 12821
ipi 42941
ipi+halt 43146

vmcall 9.3% improvement

[patch 01/11] KVM: modify memslots layout in struct kvm

[Marcelo Tosatti]

[-- Attachment #1: modify-memslot-layout --]
[-- Type: text/plain, Size: 8417 bytes --]

Have a pointer to an allocated region inside struct kvm.

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

Index: kvm/arch/x86/kvm/mmu.c
===================================================================
--- kvm.orig/arch/x86/kvm/mmu.c
+++ kvm/arch/x86/kvm/mmu.c
@@ -809,13 +809,14 @@ static int kvm_handle_hva(struct kvm *kv
 {
 	int i, j;
 	int retval = 0;
+	struct kvm_memslots *slots = kvm->memslots;
 
 	/*
 	 * If mmap_sem isn't taken, we can look the memslots with only
 	 * the mmu_lock by skipping over the slots with userspace_addr == 0.
 	 */
-	for (i = 0; i < kvm->nmemslots; i++) {
-		struct kvm_memory_slot *memslot = &kvm->memslots[i];
+	for (i = 0; i < slots->nmemslots; i++) {
+		struct kvm_memory_slot *memslot = &slots->memslots[i];
 		unsigned long start = memslot->userspace_addr;
 		unsigned long end;
 
@@ -3023,8 +3024,8 @@ unsigned int kvm_mmu_calculate_mmu_pages
 	unsigned int nr_mmu_pages;
 	unsigned int  nr_pages = 0;
 
-	for (i = 0; i < kvm->nmemslots; i++)
-		nr_pages += kvm->memslots[i].npages;
+	for (i = 0; i < kvm->memslots->nmemslots; i++)
+		nr_pages += kvm->memslots->memslots[i].npages;
 
 	nr_mmu_pages = nr_pages * KVM_PERMILLE_MMU_PAGES / 1000;
 	nr_mmu_pages = max(nr_mmu_pages,
@@ -3297,7 +3298,7 @@ static int count_rmaps(struct kvm_vcpu *
 	int i, j, k;
 
 	for (i = 0; i < KVM_MEMORY_SLOTS; ++i) {
-		struct kvm_memory_slot *m = &vcpu->kvm->memslots[i];
+		struct kvm_memory_slot *m = &vcpu->kvm->memslots->memslots[i];
 		struct kvm_rmap_desc *d;
 
 		for (j = 0; j < m->npages; ++j) {
Index: kvm/arch/x86/kvm/vmx.c
===================================================================
--- kvm.orig/arch/x86/kvm/vmx.c
+++ kvm/arch/x86/kvm/vmx.c
@@ -1503,8 +1503,8 @@ static void enter_pmode(struct kvm_vcpu 
 static gva_t rmode_tss_base(struct kvm *kvm)
 {
 	if (!kvm->arch.tss_addr) {
-		gfn_t base_gfn = kvm->memslots[0].base_gfn +
-				 kvm->memslots[0].npages - 3;
+		gfn_t base_gfn = kvm->memslots->memslots[0].base_gfn +
+				 kvm->memslots->memslots[0].npages - 3;
 		return base_gfn << PAGE_SHIFT;
 	}
 	return kvm->arch.tss_addr;
Index: kvm/arch/x86/kvm/x86.c
===================================================================
--- kvm.orig/arch/x86/kvm/x86.c
+++ kvm/arch/x86/kvm/x86.c
@@ -2428,7 +2428,7 @@ int kvm_vm_ioctl_get_dirty_log(struct kv
 		spin_lock(&kvm->mmu_lock);
 		kvm_mmu_slot_remove_write_access(kvm, log->slot);
 		spin_unlock(&kvm->mmu_lock);
-		memslot = &kvm->memslots[log->slot];
+		memslot = &kvm->memslots->memslots[log->slot];
 		n = ALIGN(memslot->npages, BITS_PER_LONG) / 8;
 		memset(memslot->dirty_bitmap, 0, n);
 	}
@@ -5222,7 +5222,7 @@ int kvm_arch_set_memory_region(struct kv
 				int user_alloc)
 {
 	int npages = mem->memory_size >> PAGE_SHIFT;
-	struct kvm_memory_slot *memslot = &kvm->memslots[mem->slot];
+	struct kvm_memory_slot *memslot = &kvm->memslots->memslots[mem->slot];
 
 	/*To keep backward compatibility with older userspace,
 	 *x86 needs to hanlde !user_alloc case.
Index: kvm/include/linux/kvm_host.h
===================================================================
--- kvm.orig/include/linux/kvm_host.h
+++ kvm/include/linux/kvm_host.h
@@ -150,14 +150,18 @@ struct kvm_irq_routing_table {};
 
 #endif
 
+struct kvm_memslots {
+	int nmemslots;
+	struct kvm_memory_slot memslots[KVM_MEMORY_SLOTS +
+					KVM_PRIVATE_MEM_SLOTS];
+};
+
 struct kvm {
 	spinlock_t mmu_lock;
 	spinlock_t requests_lock;
 	struct rw_semaphore slots_lock;
 	struct mm_struct *mm; /* userspace tied to this vm */
-	int nmemslots;
-	struct kvm_memory_slot memslots[KVM_MEMORY_SLOTS +
-					KVM_PRIVATE_MEM_SLOTS];
+	struct kvm_memslots *memslots;
 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
 	u32 bsp_vcpu_id;
 	struct kvm_vcpu *bsp_vcpu;
@@ -482,7 +486,7 @@ static inline void kvm_guest_exit(void)
 
 static inline int memslot_id(struct kvm *kvm, struct kvm_memory_slot *slot)
 {
-	return slot - kvm->memslots;
+	return slot - kvm->memslots->memslots;
 }
 
 static inline gpa_t gfn_to_gpa(gfn_t gfn)
Index: kvm/virt/kvm/iommu.c
===================================================================
--- kvm.orig/virt/kvm/iommu.c
+++ kvm/virt/kvm/iommu.c
@@ -76,10 +76,13 @@ unmap_pages:
 static int kvm_iommu_map_memslots(struct kvm *kvm)
 {
 	int i, r = 0;
+	struct kvm_memslots *slots;
 
-	for (i = 0; i < kvm->nmemslots; i++) {
-		r = kvm_iommu_map_pages(kvm, kvm->memslots[i].base_gfn,
-					kvm->memslots[i].npages);
+	slots = kvm->memslots;
+
+	for (i = 0; i < slots->nmemslots; i++) {
+		r = kvm_iommu_map_pages(kvm, slots->memslots[i].base_gfn,
+					slots->memslots[i].npages);
 		if (r)
 			break;
 	}
@@ -210,10 +213,13 @@ static void kvm_iommu_put_pages(struct k
 static int kvm_iommu_unmap_memslots(struct kvm *kvm)
 {
 	int i;
+	struct kvm_memslots *slots;
+
+	slots = kvm->memslots;
 
-	for (i = 0; i < kvm->nmemslots; i++) {
-		kvm_iommu_put_pages(kvm, kvm->memslots[i].base_gfn,
-				    kvm->memslots[i].npages);
+	for (i = 0; i < slots->nmemslots; i++) {
+		kvm_iommu_put_pages(kvm, slots->memslots[i].base_gfn,
+				    slots->memslots[i].npages);
 	}
 
 	return 0;
Index: kvm/virt/kvm/kvm_main.c
===================================================================
--- kvm.orig/virt/kvm/kvm_main.c
+++ kvm/virt/kvm/kvm_main.c
@@ -375,12 +375,16 @@ static struct kvm *kvm_create_vm(void)
 	INIT_HLIST_HEAD(&kvm->irq_ack_notifier_list);
 #endif
 
+	r = -ENOMEM;
+	kvm->memslots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL);
+	if (!kvm->memslots)
+		goto out_err;
+
 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
 	page = alloc_page(GFP_KERNEL | __GFP_ZERO);
-	if (!page) {
-		r = -ENOMEM;
+	if (!page)
 		goto out_err;
-	}
+
 	kvm->coalesced_mmio_ring =
 			(struct kvm_coalesced_mmio_ring *)page_address(page);
 #endif
@@ -416,6 +420,7 @@ out:
 out_err:
 	hardware_disable_all();
 out_err_nodisable:
+	kfree(kvm->memslots);
 	kfree(kvm);
 	return ERR_PTR(r);
 }
@@ -450,9 +455,12 @@ static void kvm_free_physmem_slot(struct
 void kvm_free_physmem(struct kvm *kvm)
 {
 	int i;
+	struct kvm_memslots *slots = kvm->memslots;
+
+	for (i = 0; i < slots->nmemslots; ++i)
+		kvm_free_physmem_slot(&slots->memslots[i], NULL);
 
-	for (i = 0; i < kvm->nmemslots; ++i)
-		kvm_free_physmem_slot(&kvm->memslots[i], NULL);
+	kfree(kvm->memslots);
 }
 
 static void kvm_destroy_vm(struct kvm *kvm)
@@ -533,7 +541,7 @@ int __kvm_set_memory_region(struct kvm *
 	if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr)
 		goto out;
 
-	memslot = &kvm->memslots[mem->slot];
+	memslot = &kvm->memslots->memslots[mem->slot];
 	base_gfn = mem->guest_phys_addr >> PAGE_SHIFT;
 	npages = mem->memory_size >> PAGE_SHIFT;
 
@@ -554,7 +562,7 @@ int __kvm_set_memory_region(struct kvm *
 	/* Check for overlaps */
 	r = -EEXIST;
 	for (i = 0; i < KVM_MEMORY_SLOTS; ++i) {
-		struct kvm_memory_slot *s = &kvm->memslots[i];
+		struct kvm_memory_slot *s = &kvm->memslots->memslots[i];
 
 		if (s == memslot || !s->npages)
 			continue;
@@ -656,8 +664,8 @@ skip_lpage:
 		kvm_arch_flush_shadow(kvm);
 
 	spin_lock(&kvm->mmu_lock);
-	if (mem->slot >= kvm->nmemslots)
-		kvm->nmemslots = mem->slot + 1;
+	if (mem->slot >= kvm->memslots->nmemslots)
+		kvm->memslots->nmemslots = mem->slot + 1;
 
 	*memslot = new;
 	spin_unlock(&kvm->mmu_lock);
@@ -727,7 +735,7 @@ int kvm_get_dirty_log(struct kvm *kvm,
 	if (log->slot >= KVM_MEMORY_SLOTS)
 		goto out;
 
-	memslot = &kvm->memslots[log->slot];
+	memslot = &kvm->memslots->memslots[log->slot];
 	r = -ENOENT;
 	if (!memslot->dirty_bitmap)
 		goto out;
@@ -781,9 +789,10 @@ EXPORT_SYMBOL_GPL(kvm_is_error_hva);
 struct kvm_memory_slot *gfn_to_memslot_unaliased(struct kvm *kvm, gfn_t gfn)
 {
 	int i;
+	struct kvm_memslots *slots = kvm->memslots;
 
-	for (i = 0; i < kvm->nmemslots; ++i) {
-		struct kvm_memory_slot *memslot = &kvm->memslots[i];
+	for (i = 0; i < slots->nmemslots; ++i) {
+		struct kvm_memory_slot *memslot = &slots->memslots[i];
 
 		if (gfn >= memslot->base_gfn
 		    && gfn < memslot->base_gfn + memslot->npages)
@@ -802,10 +811,11 @@ struct kvm_memory_slot *gfn_to_memslot(s
 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn)
 {
 	int i;
+	struct kvm_memslots *slots = kvm->memslots;
 
 	gfn = unalias_gfn(kvm, gfn);
 	for (i = 0; i < KVM_MEMORY_SLOTS; ++i) {
-		struct kvm_memory_slot *memslot = &kvm->memslots[i];
+		struct kvm_memory_slot *memslot = &slots->memslots[i];
 
 		if (gfn >= memslot->base_gfn
 		    && gfn < memslot->base_gfn + memslot->npages)

[patch 02/11] KVM: modify alias layout in x86s struct kvm_arch

[Marcelo Tosatti]

[-- Attachment #1: modify-alias-layout --]
[-- Type: text/plain, Size: 2666 bytes --]

Have a pointer to an allocated region inside x86's kvm_arch.

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

Index: kvm/arch/x86/include/asm/kvm_host.h
===================================================================
--- kvm.orig/arch/x86/include/asm/kvm_host.h
+++ kvm/arch/x86/include/asm/kvm_host.h
@@ -370,9 +370,13 @@ struct kvm_mem_alias {
 	gfn_t target_gfn;
 };
 
-struct kvm_arch{
-	int naliases;
+struct kvm_mem_aliases {
 	struct kvm_mem_alias aliases[KVM_ALIAS_SLOTS];
+	int naliases;
+};
+
+struct kvm_arch {
+	struct kvm_mem_aliases *aliases;
 
 	unsigned int n_free_mmu_pages;
 	unsigned int n_requested_mmu_pages;
Index: kvm/arch/x86/kvm/x86.c
===================================================================
--- kvm.orig/arch/x86/kvm/x86.c
+++ kvm/arch/x86/kvm/x86.c
@@ -2228,9 +2228,10 @@ gfn_t unalias_gfn(struct kvm *kvm, gfn_t
 {
 	int i;
 	struct kvm_mem_alias *alias;
+	struct kvm_mem_aliases *aliases = kvm->arch.aliases;
 
-	for (i = 0; i < kvm->arch.naliases; ++i) {
-		alias = &kvm->arch.aliases[i];
+	for (i = 0; i < aliases->naliases; ++i) {
+		alias = &aliases->aliases[i];
 		if (gfn >= alias->base_gfn
 		    && gfn < alias->base_gfn + alias->npages)
 			return alias->target_gfn + gfn - alias->base_gfn;
@@ -2248,6 +2249,7 @@ static int kvm_vm_ioctl_set_memory_alias
 {
 	int r, n;
 	struct kvm_mem_alias *p;
+	struct kvm_mem_aliases *aliases;
 
 	r = -EINVAL;
 	/* General sanity checks */
@@ -2267,15 +2269,17 @@ static int kvm_vm_ioctl_set_memory_alias
 	down_write(&kvm->slots_lock);
 	spin_lock(&kvm->mmu_lock);
 
-	p = &kvm->arch.aliases[alias->slot];
+	aliases = kvm->arch.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;
 
 	for (n = KVM_ALIAS_SLOTS; n > 0; --n)
-		if (kvm->arch.aliases[n - 1].npages)
+		if (aliases->aliases[n - 1].npages)
 			break;
-	kvm->arch.naliases = n;
+	aliases->naliases = n;
 
 	spin_unlock(&kvm->mmu_lock);
 	kvm_mmu_zap_all(kvm);
@@ -5157,6 +5161,12 @@ struct  kvm *kvm_arch_create_vm(void)
 	if (!kvm)
 		return ERR_PTR(-ENOMEM);
 
+	kvm->arch.aliases = kzalloc(sizeof(struct kvm_mem_aliases), GFP_KERNEL);
+	if (!kvm->arch.aliases) {
+		kfree(kvm);
+		return ERR_PTR(-ENOMEM);
+	}
+
 	INIT_LIST_HEAD(&kvm->arch.active_mmu_pages);
 	INIT_LIST_HEAD(&kvm->arch.assigned_dev_head);
 
@@ -5213,6 +5223,7 @@ void kvm_arch_destroy_vm(struct kvm *kvm
 		put_page(kvm->arch.apic_access_page);
 	if (kvm->arch.ept_identity_pagetable)
 		put_page(kvm->arch.ept_identity_pagetable);
+	kfree(kvm->arch.aliases);
 	kfree(kvm);
 }
 

[patch 03/11] KVM: split kvm_arch_set_memory_region into prepare and commit

[Marcelo Tosatti]

[-- Attachment #1: prepare-commit-memory-region --]
[-- Type: text/plain, Size: 6354 bytes --]

Required for SRCU convertion later.

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

Index: kvm/arch/ia64/kvm/kvm-ia64.c
===================================================================
--- kvm.orig/arch/ia64/kvm/kvm-ia64.c
+++ kvm/arch/ia64/kvm/kvm-ia64.c
@@ -1576,15 +1576,14 @@ out:
 	return r;
 }
 
-int kvm_arch_set_memory_region(struct kvm *kvm,
-		struct kvm_userspace_memory_region *mem,
+int kvm_arch_prepare_memory_region(struct kvm *kvm,
+		struct kvm_memory_slot *memslot,
 		struct kvm_memory_slot old,
 		int user_alloc)
 {
 	unsigned long i;
 	unsigned long pfn;
-	int npages = mem->memory_size >> PAGE_SHIFT;
-	struct kvm_memory_slot *memslot = &kvm->memslots[mem->slot];
+	int npages = memslot->npages;
 	unsigned long base_gfn = memslot->base_gfn;
 
 	if (base_gfn + npages > (KVM_MAX_MEM_SIZE >> PAGE_SHIFT))
@@ -1608,6 +1607,14 @@ int kvm_arch_set_memory_region(struct kv
 	return 0;
 }
 
+void kvm_arch_commit_memory_region(struct kvm *kvm,
+		struct kvm_userspace_memory_region *mem,
+		struct kvm_memory_slot old,
+		int user_alloc)
+{
+	return;
+}
+
 void kvm_arch_flush_shadow(struct kvm *kvm)
 {
 	kvm_flush_remote_tlbs(kvm);
Index: kvm/arch/s390/kvm/kvm-s390.c
===================================================================
--- kvm.orig/arch/s390/kvm/kvm-s390.c
+++ kvm/arch/s390/kvm/kvm-s390.c
@@ -689,10 +689,10 @@ long kvm_arch_vcpu_ioctl(struct file *fi
 }
 
 /* Section: memory related */
-int kvm_arch_set_memory_region(struct kvm *kvm,
-				struct kvm_userspace_memory_region *mem,
-				struct kvm_memory_slot old,
-				int user_alloc)
+int kvm_arch_prepare_memory_region(struct kvm *kvm,
+				   struct kvm_userspace_memory_region *mem,
+				   struct kvm_memory_slot old,
+				   int user_alloc)
 {
 	int i;
 	struct kvm_vcpu *vcpu;
@@ -719,14 +719,20 @@ int kvm_arch_set_memory_region(struct kv
 	if (!user_alloc)
 		return -EINVAL;
 
+	return 0;
+}
+
+int kvm_arch_commit_memory_region(struct kvm *kvm,
+				struct kvm_userspace_memory_region *mem,
+				struct kvm_memory_slot old,
+				int user_alloc)
+{
 	/* request update of sie control block for all available vcpus */
 	kvm_for_each_vcpu(i, vcpu, kvm) {
 		if (test_and_set_bit(KVM_REQ_MMU_RELOAD, &vcpu->requests))
 			continue;
 		kvm_s390_inject_sigp_stop(vcpu, ACTION_RELOADVCPU_ON_STOP);
 	}
-
-	return 0;
 }
 
 void kvm_arch_flush_shadow(struct kvm *kvm)
Index: kvm/arch/x86/kvm/x86.c
===================================================================
--- kvm.orig/arch/x86/kvm/x86.c
+++ kvm/arch/x86/kvm/x86.c
@@ -5227,13 +5227,12 @@ void kvm_arch_destroy_vm(struct kvm *kvm
 	kfree(kvm);
 }
 
-int kvm_arch_set_memory_region(struct kvm *kvm,
-				struct kvm_userspace_memory_region *mem,
+int kvm_arch_prepare_memory_region(struct kvm *kvm,
+				struct kvm_memory_slot *memslot,
 				struct kvm_memory_slot old,
 				int user_alloc)
 {
-	int npages = mem->memory_size >> PAGE_SHIFT;
-	struct kvm_memory_slot *memslot = &kvm->memslots->memslots[mem->slot];
+	int npages = memslot->npages;
 
 	/*To keep backward compatibility with older userspace,
 	 *x86 needs to hanlde !user_alloc case.
@@ -5253,26 +5252,35 @@ int kvm_arch_set_memory_region(struct kv
 			if (IS_ERR((void *)userspace_addr))
 				return PTR_ERR((void *)userspace_addr);
 
-			/* set userspace_addr atomically for kvm_hva_to_rmapp */
-			spin_lock(&kvm->mmu_lock);
 			memslot->userspace_addr = userspace_addr;
-			spin_unlock(&kvm->mmu_lock);
-		} else {
-			if (!old.user_alloc && old.rmap) {
-				int ret;
-
-				down_write(&current->mm->mmap_sem);
-				ret = do_munmap(current->mm, old.userspace_addr,
-						old.npages * PAGE_SIZE);
-				up_write(&current->mm->mmap_sem);
-				if (ret < 0)
-					printk(KERN_WARNING
-				       "kvm_vm_ioctl_set_memory_region: "
-				       "failed to munmap memory\n");
-			}
 		}
 	}
 
+
+	return 0;
+}
+
+void kvm_arch_commit_memory_region(struct kvm *kvm,
+				struct kvm_userspace_memory_region *mem,
+				struct kvm_memory_slot old,
+				int user_alloc)
+{
+
+	int npages = mem->memory_size >> PAGE_SHIFT;
+
+	if (!user_alloc && !old.user_alloc && old.rmap && !npages) {
+		int ret;
+
+		down_write(&current->mm->mmap_sem);
+		ret = do_munmap(current->mm, old.userspace_addr,
+				old.npages * PAGE_SIZE);
+		up_write(&current->mm->mmap_sem);
+		if (ret < 0)
+			printk(KERN_WARNING
+			       "kvm_vm_ioctl_set_memory_region: "
+			       "failed to munmap memory\n");
+	}
+
 	spin_lock(&kvm->mmu_lock);
 	if (!kvm->arch.n_requested_mmu_pages) {
 		unsigned int nr_mmu_pages = kvm_mmu_calculate_mmu_pages(kvm);
@@ -5281,8 +5289,6 @@ int kvm_arch_set_memory_region(struct kv
 
 	kvm_mmu_slot_remove_write_access(kvm, mem->slot);
 	spin_unlock(&kvm->mmu_lock);
-
-	return 0;
 }
 
 void kvm_arch_flush_shadow(struct kvm *kvm)
Index: kvm/include/linux/kvm_host.h
===================================================================
--- kvm.orig/include/linux/kvm_host.h
+++ kvm/include/linux/kvm_host.h
@@ -253,7 +253,11 @@ int kvm_set_memory_region(struct kvm *kv
 int __kvm_set_memory_region(struct kvm *kvm,
 			    struct kvm_userspace_memory_region *mem,
 			    int user_alloc);
-int kvm_arch_set_memory_region(struct kvm *kvm,
+int kvm_arch_prepare_memory_region(struct kvm *kvm,
+				struct kvm_memory_slot *memslot,
+				struct kvm_memory_slot old,
+				int user_alloc);
+void kvm_arch_commit_memory_region(struct kvm *kvm,
 				struct kvm_userspace_memory_region *mem,
 				struct kvm_memory_slot old,
 				int user_alloc);
Index: kvm/virt/kvm/kvm_main.c
===================================================================
--- kvm.orig/virt/kvm/kvm_main.c
+++ kvm/virt/kvm/kvm_main.c
@@ -663,6 +663,10 @@ skip_lpage:
 	if (!npages)
 		kvm_arch_flush_shadow(kvm);
 
+	r = kvm_arch_prepare_memory_region(kvm, &new, old, user_alloc);
+	if (r)
+		goto out_free;
+
 	spin_lock(&kvm->mmu_lock);
 	if (mem->slot >= kvm->memslots->nmemslots)
 		kvm->memslots->nmemslots = mem->slot + 1;
@@ -670,13 +674,7 @@ skip_lpage:
 	*memslot = new;
 	spin_unlock(&kvm->mmu_lock);
 
-	r = kvm_arch_set_memory_region(kvm, mem, old, user_alloc);
-	if (r) {
-		spin_lock(&kvm->mmu_lock);
-		*memslot = old;
-		spin_unlock(&kvm->mmu_lock);
-		goto out_free;
-	}
+	kvm_arch_commit_memory_region(kvm, mem, old, user_alloc);
 
 	kvm_free_physmem_slot(&old, npages ? &new : NULL);
 	/* Slot deletion case: we have to update the current slot */

[patch 04/11] KVM: introduce gfn_to_pfn_memslot

[Marcelo Tosatti]

[-- Attachment #1: introduce-gfn_to_pfn_memslot --]
[-- Type: text/plain, Size: 2132 bytes --]

Which takes a memslot pointer instead of using kvm->memslots.

To be used by SRCU convertion later.

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

Index: kvm/include/linux/kvm_host.h
===================================================================
--- kvm.orig/include/linux/kvm_host.h
+++ kvm/include/linux/kvm_host.h
@@ -272,6 +272,8 @@ void kvm_set_page_dirty(struct page *pag
 void kvm_set_page_accessed(struct page *page);
 
 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
+pfn_t gfn_to_pfn_memslot(struct kvm *kvm,
+			 struct kvm_memory_slot *slot, gfn_t gfn);
 void kvm_release_pfn_dirty(pfn_t);
 void kvm_release_pfn_clean(pfn_t pfn);
 void kvm_set_pfn_dirty(pfn_t pfn);
Index: kvm/virt/kvm/kvm_main.c
===================================================================
--- kvm.orig/virt/kvm/kvm_main.c
+++ kvm/virt/kvm/kvm_main.c
@@ -835,21 +835,14 @@ unsigned long gfn_to_hva(struct kvm *kvm
 }
 EXPORT_SYMBOL_GPL(gfn_to_hva);
 
-pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn)
+static pfn_t __gfn_to_pfn(struct kvm *kvm, unsigned long addr)
 {
 	struct page *page[1];
-	unsigned long addr;
 	int npages;
 	pfn_t pfn;
 
 	might_sleep();
 
-	addr = gfn_to_hva(kvm, gfn);
-	if (kvm_is_error_hva(addr)) {
-		get_page(bad_page);
-		return page_to_pfn(bad_page);
-	}
-
 	npages = get_user_pages_fast(addr, 1, 1, page);
 
 	if (unlikely(npages != 1)) {
@@ -874,8 +867,32 @@ pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t 
 	return pfn;
 }
 
+pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn)
+{
+	unsigned long addr;
+
+	addr = gfn_to_hva(kvm, gfn);
+	if (kvm_is_error_hva(addr)) {
+		get_page(bad_page);
+		return page_to_pfn(bad_page);
+	}
+
+	return __gfn_to_pfn(kvm, addr);
+}
 EXPORT_SYMBOL_GPL(gfn_to_pfn);
 
+static unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
+{
+	return (slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE);
+}
+
+pfn_t gfn_to_pfn_memslot(struct kvm *kvm,
+			 struct kvm_memory_slot *slot, gfn_t gfn)
+{
+	unsigned long addr = gfn_to_hva_memslot(slot, gfn);
+	return __gfn_to_pfn(kvm, addr);
+}
+
 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn)
 {
 	pfn_t pfn;

Re: [patch 04/11] KVM: introduce gfn_to_pfn_memslot

[Avi Kivity]

On 12/23/2009 01:38 PM, Marcelo Tosatti wrote:
> Which takes a memslot pointer instead of using kvm->memslots.
>
> To be used by SRCU convertion later.
>
> Signed-off-by: Marcelo Tosatti<mtosatti@redhat.com>
>
>
> @@ -835,21 +835,14 @@ unsigned long gfn_to_hva(struct kvm *kvm
>   }
>   EXPORT_SYMBOL_GPL(gfn_to_hva);
>
> -pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn)
> +static pfn_t __gfn_to_pfn(struct kvm *kvm, unsigned long addr)
>   {
>    

That's an hva now, not a gfn; the function name should reflect this.

-- 
error compiling committee.c: too many arguments to function

[patch 05/11] KVM: use gfn_to_pfn_memslot in kvm_iommu_map_pages

[Marcelo Tosatti]

[-- Attachment #1: pass-memslot-to-kvm_iommu_map_pages --]
[-- Type: text/plain, Size: 2550 bytes --]

So its possible to iommu map a memslot before making it visible to 
kvm.

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

Index: kvm/include/linux/kvm_host.h
===================================================================
--- kvm.orig/include/linux/kvm_host.h
+++ kvm/include/linux/kvm_host.h
@@ -439,8 +439,7 @@ void kvm_free_irq_source_id(struct kvm *
 #define KVM_IOMMU_CACHE_COHERENCY	0x1
 
 #ifdef CONFIG_IOMMU_API
-int kvm_iommu_map_pages(struct kvm *kvm, gfn_t base_gfn,
-			unsigned long npages);
+int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
 int kvm_iommu_map_guest(struct kvm *kvm);
 int kvm_iommu_unmap_guest(struct kvm *kvm);
 int kvm_assign_device(struct kvm *kvm,
Index: kvm/virt/kvm/iommu.c
===================================================================
--- kvm.orig/virt/kvm/iommu.c
+++ kvm/virt/kvm/iommu.c
@@ -32,10 +32,10 @@ static int kvm_iommu_unmap_memslots(stru
 static void kvm_iommu_put_pages(struct kvm *kvm,
 				gfn_t base_gfn, unsigned long npages);
 
-int kvm_iommu_map_pages(struct kvm *kvm,
-			gfn_t base_gfn, unsigned long npages)
+int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot)
 {
-	gfn_t gfn = base_gfn;
+	gfn_t gfn = slot->base_gfn;
+	unsigned long npages = slot->npages;
 	pfn_t pfn;
 	int i, r = 0;
 	struct iommu_domain *domain = kvm->arch.iommu_domain;
@@ -54,7 +54,7 @@ int kvm_iommu_map_pages(struct kvm *kvm,
 		if (iommu_iova_to_phys(domain, gfn_to_gpa(gfn)))
 			continue;
 
-		pfn = gfn_to_pfn(kvm, gfn);
+		pfn = gfn_to_pfn_memslot(kvm, slot, gfn);
 		r = iommu_map_range(domain,
 				    gfn_to_gpa(gfn),
 				    pfn_to_hpa(pfn),
@@ -69,7 +69,7 @@ int kvm_iommu_map_pages(struct kvm *kvm,
 	return 0;
 
 unmap_pages:
-	kvm_iommu_put_pages(kvm, base_gfn, i);
+	kvm_iommu_put_pages(kvm, slot->base_gfn, i);
 	return r;
 }
 
@@ -81,8 +81,7 @@ static int kvm_iommu_map_memslots(struct
 	slots = kvm->memslots;
 
 	for (i = 0; i < slots->nmemslots; i++) {
-		r = kvm_iommu_map_pages(kvm, slots->memslots[i].base_gfn,
-					slots->memslots[i].npages);
+		r = kvm_iommu_map_pages(kvm, &slots->memslots[i]);
 		if (r)
 			break;
 	}
Index: kvm/virt/kvm/kvm_main.c
===================================================================
--- kvm.orig/virt/kvm/kvm_main.c
+++ kvm/virt/kvm/kvm_main.c
@@ -684,7 +684,7 @@ skip_lpage:
 	spin_unlock(&kvm->mmu_lock);
 #ifdef CONFIG_DMAR
 	/* map the pages in iommu page table */
-	r = kvm_iommu_map_pages(kvm, base_gfn, npages);
+	r = kvm_iommu_map_pages(kvm, memslot);
 	if (r)
 		goto out;
 #endif

[patch 06/11] KVM: introduce kvm->srcu and convert kvm_set_memory_region to SRCU update

[Marcelo Tosatti]

[-- Attachment #1: introduce-srcu-and-use-for-slots --]
[-- Type: text/plain, Size: 14621 bytes --]

Use two steps for memslot deletion: mark the slot invalid (which stops 
instantiation of new shadow pages for that slot, but allows destruction),
then instantiate the new empty slot.

Also simplifies kvm_handle_hva locking.

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

Index: kvm/include/linux/kvm.h
===================================================================
--- kvm.orig/include/linux/kvm.h
+++ kvm/include/linux/kvm.h
@@ -103,7 +103,7 @@ struct kvm_userspace_memory_region {
 
 /* for kvm_memory_region::flags */
 #define KVM_MEM_LOG_DIRTY_PAGES  1UL
-
+#define KVM_MEMSLOT_INVALID      (1UL << 1)
 
 /* for KVM_IRQ_LINE */
 struct kvm_irq_level {
Index: kvm/virt/kvm/kvm_main.c
===================================================================
--- kvm.orig/virt/kvm/kvm_main.c
+++ kvm/virt/kvm/kvm_main.c
@@ -44,6 +44,7 @@
 #include <linux/bitops.h>
 #include <linux/spinlock.h>
 #include <linux/compat.h>
+#include <linux/srcu.h>
 
 #include <asm/processor.h>
 #include <asm/io.h>
@@ -379,11 +380,15 @@ static struct kvm *kvm_create_vm(void)
 	kvm->memslots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL);
 	if (!kvm->memslots)
 		goto out_err;
+	if (init_srcu_struct(&kvm->srcu))
+		goto out_err;
 
 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
 	page = alloc_page(GFP_KERNEL | __GFP_ZERO);
-	if (!page)
+	if (!page) {
+		cleanup_srcu_struct(&kvm->srcu);
 		goto out_err;
+	}
 
 	kvm->coalesced_mmio_ring =
 			(struct kvm_coalesced_mmio_ring *)page_address(page);
@@ -391,6 +396,7 @@ static struct kvm *kvm_create_vm(void)
 
 	r = kvm_init_mmu_notifier(kvm);
 	if (r) {
+		cleanup_srcu_struct(&kvm->srcu);
 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
 		put_page(page);
 #endif
@@ -481,6 +487,7 @@ static void kvm_destroy_vm(struct kvm *k
 	kvm_arch_flush_shadow(kvm);
 #endif
 	kvm_arch_destroy_vm(kvm);
+	cleanup_srcu_struct(&kvm->srcu);
 	hardware_disable_all();
 	mmdrop(mm);
 }
@@ -521,12 +528,13 @@ int __kvm_set_memory_region(struct kvm *
 			    struct kvm_userspace_memory_region *mem,
 			    int user_alloc)
 {
-	int r;
+	int r, flush_shadow = 0;
 	gfn_t base_gfn;
 	unsigned long npages;
 	unsigned long i;
 	struct kvm_memory_slot *memslot;
 	struct kvm_memory_slot old, new;
+	struct kvm_memslots *slots, *old_memslots;
 
 	r = -EINVAL;
 	/* General sanity checks */
@@ -588,15 +596,7 @@ int __kvm_set_memory_region(struct kvm *
 		memset(new.rmap, 0, npages * sizeof(*new.rmap));
 
 		new.user_alloc = user_alloc;
-		/*
-		 * hva_to_rmmap() serialzies with the mmu_lock and to be
-		 * safe it has to ignore memslots with !user_alloc &&
-		 * !userspace_addr.
-		 */
-		if (user_alloc)
-			new.userspace_addr = mem->userspace_addr;
-		else
-			new.userspace_addr = 0;
+		new.userspace_addr = mem->userspace_addr;
 	}
 	if (!npages)
 		goto skip_lpage;
@@ -651,8 +651,9 @@ skip_lpage:
 		if (!new.dirty_bitmap)
 			goto out_free;
 		memset(new.dirty_bitmap, 0, dirty_bytes);
+		/* destroy any largepage mappings for dirty tracking */
 		if (old.npages)
-			kvm_arch_flush_shadow(kvm);
+			flush_shadow = 1;
 	}
 #else  /* not defined CONFIG_S390 */
 	new.user_alloc = user_alloc;
@@ -660,34 +661,69 @@ skip_lpage:
 		new.userspace_addr = mem->userspace_addr;
 #endif /* not defined CONFIG_S390 */
 
-	if (!npages)
+	if (!npages) {
+		slots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL);
+		if (!slots)
+			goto out_free;
+		memcpy(slots, kvm->memslots, sizeof(struct kvm_memslots));
+		if (mem->slot >= slots->nmemslots)
+			slots->nmemslots = mem->slot + 1;
+		slots->memslots[mem->slot].flags |= KVM_MEMSLOT_INVALID;
+
+		old_memslots = kvm->memslots;
+		rcu_assign_pointer(kvm->memslots, slots);
+		synchronize_srcu_expedited(&kvm->srcu);
+		/* From this point no new shadow pages pointing to a deleted
+		 * memslot will be created.
+         	 *
+         	 * validation of sp->gfn happens in:
+         	 * 	- gfn_to_hva (kvm_read_guest, gfn_to_pfn)
+         	 * 	- kvm_is_visible_gfn (mmu_check_roots)
+         	 */
 		kvm_arch_flush_shadow(kvm);
+		kfree(old_memslots);
+	}
 
 	r = kvm_arch_prepare_memory_region(kvm, &new, old, user_alloc);
 	if (r)
 		goto out_free;
 
-	spin_lock(&kvm->mmu_lock);
-	if (mem->slot >= kvm->memslots->nmemslots)
-		kvm->memslots->nmemslots = mem->slot + 1;
+#ifdef CONFIG_DMAR
+	/* map the pages in iommu page table */
+	if (npages)
+		r = kvm_iommu_map_pages(kvm, &new);
+		if (r)
+			goto out_free;
+#endif
 
-	*memslot = new;
-	spin_unlock(&kvm->mmu_lock);
+	slots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL);
+	if (!slots)
+		goto out_free;
+	memcpy(slots, kvm->memslots, sizeof(struct kvm_memslots));
+	if (mem->slot >= slots->nmemslots)
+		slots->nmemslots = mem->slot + 1;
+
+	/* actual memory is freed via old in kvm_free_physmem_slot below */
+	if (!npages) {
+		new.rmap = NULL;
+		new.dirty_bitmap = NULL;
+		for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i)
+			new.lpage_info[i] = NULL;
+	}
+
+	slots->memslots[mem->slot] = new;
+	old_memslots = kvm->memslots;
+	rcu_assign_pointer(kvm->memslots, slots);
+	synchronize_srcu_expedited(&kvm->srcu);
 
 	kvm_arch_commit_memory_region(kvm, mem, old, user_alloc);
 
-	kvm_free_physmem_slot(&old, npages ? &new : NULL);
-	/* Slot deletion case: we have to update the current slot */
-	spin_lock(&kvm->mmu_lock);
-	if (!npages)
-		*memslot = old;
-	spin_unlock(&kvm->mmu_lock);
-#ifdef CONFIG_DMAR
-	/* map the pages in iommu page table */
-	r = kvm_iommu_map_pages(kvm, memslot);
-	if (r)
-		goto out;
-#endif
+	kvm_free_physmem_slot(&old, &new);
+	kfree(old_memslots);
+
+	if (flush_shadow)
+		kvm_arch_flush_shadow(kvm);
+
 	return 0;
 
 out_free:
@@ -787,7 +823,7 @@ EXPORT_SYMBOL_GPL(kvm_is_error_hva);
 struct kvm_memory_slot *gfn_to_memslot_unaliased(struct kvm *kvm, gfn_t gfn)
 {
 	int i;
-	struct kvm_memslots *slots = kvm->memslots;
+	struct kvm_memslots *slots = rcu_dereference(kvm->memslots);
 
 	for (i = 0; i < slots->nmemslots; ++i) {
 		struct kvm_memory_slot *memslot = &slots->memslots[i];
@@ -809,12 +845,15 @@ struct kvm_memory_slot *gfn_to_memslot(s
 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn)
 {
 	int i;
-	struct kvm_memslots *slots = kvm->memslots;
+	struct kvm_memslots *slots = rcu_dereference(kvm->memslots);
 
 	gfn = unalias_gfn(kvm, gfn);
 	for (i = 0; i < KVM_MEMORY_SLOTS; ++i) {
 		struct kvm_memory_slot *memslot = &slots->memslots[i];
 
+		if (memslot->flags & KVM_MEMSLOT_INVALID)
+			continue;
+
 		if (gfn >= memslot->base_gfn
 		    && gfn < memslot->base_gfn + memslot->npages)
 			return 1;
@@ -823,13 +862,31 @@ int kvm_is_visible_gfn(struct kvm *kvm, 
 }
 EXPORT_SYMBOL_GPL(kvm_is_visible_gfn);
 
+int memslot_id(struct kvm *kvm, gfn_t gfn)
+{
+	int i;
+	struct kvm_memslots *slots = rcu_dereference(kvm->memslots);
+	struct kvm_memory_slot *memslot = NULL;
+
+	gfn = unalias_gfn(kvm, gfn);
+	for (i = 0; i < slots->nmemslots; ++i) {
+		memslot = &slots->memslots[i];
+
+		if (gfn >= memslot->base_gfn
+		    && gfn < memslot->base_gfn + memslot->npages)
+			break;
+	}
+
+	return memslot - slots->memslots;
+}
+
 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn)
 {
 	struct kvm_memory_slot *slot;
 
 	gfn = unalias_gfn(kvm, gfn);
 	slot = gfn_to_memslot_unaliased(kvm, gfn);
-	if (!slot)
+	if (!slot || slot->flags & KVM_MEMSLOT_INVALID)
 		return bad_hva();
 	return (slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE);
 }
Index: kvm/arch/ia64/kvm/kvm-ia64.c
===================================================================
--- kvm.orig/arch/ia64/kvm/kvm-ia64.c
+++ kvm/arch/ia64/kvm/kvm-ia64.c
@@ -1834,6 +1834,7 @@ int kvm_vm_ioctl_get_dirty_log(struct kv
 	struct kvm_memory_slot *memslot;
 	int is_dirty = 0;
 
+	down_write(&kvm->slots_lock);
 	spin_lock(&kvm->arch.dirty_log_lock);
 
 	r = kvm_ia64_sync_dirty_log(kvm, log);
@@ -1853,6 +1854,7 @@ int kvm_vm_ioctl_get_dirty_log(struct kv
 	}
 	r = 0;
 out:
+	up_write(&kvm->slots_lock);
 	spin_unlock(&kvm->arch.dirty_log_lock);
 	return r;
 }
Index: kvm/arch/x86/kvm/mmu.c
===================================================================
--- kvm.orig/arch/x86/kvm/mmu.c
+++ kvm/arch/x86/kvm/mmu.c
@@ -29,6 +29,7 @@
 #include <linux/swap.h>
 #include <linux/hugetlb.h>
 #include <linux/compiler.h>
+#include <linux/srcu.h>
 
 #include <asm/page.h>
 #include <asm/cmpxchg.h>
@@ -807,23 +808,18 @@ static int kvm_handle_hva(struct kvm *kv
 			  int (*handler)(struct kvm *kvm, unsigned long *rmapp,
 					 unsigned long data))
 {
-	int i, j;
+	int i, j, idx;
 	int retval = 0;
-	struct kvm_memslots *slots = kvm->memslots;
+	struct kvm_memslots *slots;
+
+	idx = srcu_read_lock(&kvm->srcu);
+	slots = rcu_dereference(kvm->memslots);
 
-	/*
-	 * If mmap_sem isn't taken, we can look the memslots with only
-	 * the mmu_lock by skipping over the slots with userspace_addr == 0.
-	 */
 	for (i = 0; i < slots->nmemslots; i++) {
 		struct kvm_memory_slot *memslot = &slots->memslots[i];
 		unsigned long start = memslot->userspace_addr;
 		unsigned long end;
 
-		/* mmu_lock protects userspace_addr */
-		if (!start)
-			continue;
-
 		end = start + (memslot->npages << PAGE_SHIFT);
 		if (hva >= start && hva < end) {
 			gfn_t gfn_offset = (hva - start) >> PAGE_SHIFT;
@@ -840,6 +836,7 @@ static int kvm_handle_hva(struct kvm *kv
 			}
 		}
 	}
+	srcu_read_unlock(&kvm->srcu, idx);
 
 	return retval;
 }
@@ -1619,7 +1616,7 @@ static void mmu_unshadow(struct kvm *kvm
 
 static void page_header_update_slot(struct kvm *kvm, void *pte, gfn_t gfn)
 {
-	int slot = memslot_id(kvm, gfn_to_memslot(kvm, gfn));
+	int slot = memslot_id(kvm, gfn);
 	struct kvm_mmu_page *sp = page_header(__pa(pte));
 
 	__set_bit(slot, sp->slot_bitmap);
@@ -3020,16 +3017,20 @@ nomem:
  */
 unsigned int kvm_mmu_calculate_mmu_pages(struct kvm *kvm)
 {
-	int i;
+	int i, idx;
 	unsigned int nr_mmu_pages;
 	unsigned int  nr_pages = 0;
+	struct kvm_memslots *slots;
 
-	for (i = 0; i < kvm->memslots->nmemslots; i++)
-		nr_pages += kvm->memslots->memslots[i].npages;
+	idx = srcu_read_lock(&kvm->srcu);
+	slots = rcu_dereference(kvm->memslots);
+	for (i = 0; i < slots->nmemslots; i++)
+		nr_pages += slots->memslots[i].npages;
 
 	nr_mmu_pages = nr_pages * KVM_PERMILLE_MMU_PAGES / 1000;
 	nr_mmu_pages = max(nr_mmu_pages,
 			(unsigned int) KVM_MIN_ALLOC_MMU_PAGES);
+	srcu_read_unlock(&kvm->srcu, idx);
 
 	return nr_mmu_pages;
 }
@@ -3295,10 +3296,12 @@ static void audit_mappings(struct kvm_vc
 static int count_rmaps(struct kvm_vcpu *vcpu)
 {
 	int nmaps = 0;
-	int i, j, k;
+	int i, j, k, idx;
 
+	idx = srcu_read_lock(&kvm->srcu);
+	slots = rcu_dereference(kvm->memslots);
 	for (i = 0; i < KVM_MEMORY_SLOTS; ++i) {
-		struct kvm_memory_slot *m = &vcpu->kvm->memslots->memslots[i];
+		struct kvm_memory_slot *m = &slots->memslots[i];
 		struct kvm_rmap_desc *d;
 
 		for (j = 0; j < m->npages; ++j) {
@@ -3321,6 +3324,7 @@ static int count_rmaps(struct kvm_vcpu *
 			}
 		}
 	}
+	srcu_read_unlock(&kvm->srcu, idx);
 	return nmaps;
 }
 
Index: kvm/arch/x86/kvm/vmx.c
===================================================================
--- kvm.orig/arch/x86/kvm/vmx.c
+++ kvm/arch/x86/kvm/vmx.c
@@ -24,6 +24,7 @@
 #include <linux/mm.h>
 #include <linux/highmem.h>
 #include <linux/sched.h>
+#include <linux/srcu.h>
 #include <linux/moduleparam.h>
 #include <linux/ftrace_event.h>
 #include "kvm_cache_regs.h"
@@ -1503,10 +1504,18 @@ static void enter_pmode(struct kvm_vcpu 
 static gva_t rmode_tss_base(struct kvm *kvm)
 {
 	if (!kvm->arch.tss_addr) {
-		gfn_t base_gfn = kvm->memslots->memslots[0].base_gfn +
-				 kvm->memslots->memslots[0].npages - 3;
+		struct kvm_memslots *slots;
+		gfn_t base_gfn;
+		int idx;
+
+		idx = srcu_read_lock(&kvm->srcu);
+		slots = rcu_dereference(kvm->memslots);
+ 		base_gfn = slots->memslots[0].base_gfn +
+				 slots->memslots[0].npages - 3;
+		srcu_read_unlock(&kvm->srcu, idx);
 		return base_gfn << PAGE_SHIFT;
 	}
+
 	return kvm->arch.tss_addr;
 }
 
Index: kvm/include/linux/kvm_host.h
===================================================================
--- kvm.orig/include/linux/kvm_host.h
+++ kvm/include/linux/kvm_host.h
@@ -162,6 +162,7 @@ struct kvm {
 	struct rw_semaphore slots_lock;
 	struct mm_struct *mm; /* userspace tied to this vm */
 	struct kvm_memslots *memslots;
+	struct srcu_struct srcu;
 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
 	u32 bsp_vcpu_id;
 	struct kvm_vcpu *bsp_vcpu;
@@ -274,6 +275,7 @@ void kvm_set_page_accessed(struct page *
 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
 pfn_t gfn_to_pfn_memslot(struct kvm *kvm,
 			 struct kvm_memory_slot *slot, gfn_t gfn);
+int memslot_id(struct kvm *kvm, gfn_t gfn);
 void kvm_release_pfn_dirty(pfn_t);
 void kvm_release_pfn_clean(pfn_t pfn);
 void kvm_set_pfn_dirty(pfn_t pfn);
@@ -489,11 +491,6 @@ static inline void kvm_guest_exit(void)
 	current->flags &= ~PF_VCPU;
 }
 
-static inline int memslot_id(struct kvm *kvm, struct kvm_memory_slot *slot)
-{
-	return slot - kvm->memslots->memslots;
-}
-
 static inline gpa_t gfn_to_gpa(gfn_t gfn)
 {
 	return (gpa_t)gfn << PAGE_SHIFT;
Index: kvm/virt/kvm/assigned-dev.c
===================================================================
--- kvm.orig/virt/kvm/assigned-dev.c
+++ kvm/virt/kvm/assigned-dev.c
@@ -504,12 +504,12 @@ out:
 static int kvm_vm_ioctl_assign_device(struct kvm *kvm,
 				      struct kvm_assigned_pci_dev *assigned_dev)
 {
-	int r = 0;
+	int r = 0, idx;
 	struct kvm_assigned_dev_kernel *match;
 	struct pci_dev *dev;
 
 	mutex_lock(&kvm->lock);
-	down_read(&kvm->slots_lock);
+	idx = srcu_read_lock(&kvm->srcu);
 
 	match = kvm_find_assigned_dev(&kvm->arch.assigned_dev_head,
 				      assigned_dev->assigned_dev_id);
@@ -573,7 +573,7 @@ static int kvm_vm_ioctl_assign_device(st
 	}
 
 out:
-	up_read(&kvm->slots_lock);
+	srcu_read_unlock(&kvm->srcu, idx);
 	mutex_unlock(&kvm->lock);
 	return r;
 out_list_del:
@@ -585,7 +585,7 @@ out_put:
 	pci_dev_put(dev);
 out_free:
 	kfree(match);
-	up_read(&kvm->slots_lock);
+	srcu_read_unlock(&kvm->srcu, idx);
 	mutex_unlock(&kvm->lock);
 	return r;
 }
Index: kvm/virt/kvm/iommu.c
===================================================================
--- kvm.orig/virt/kvm/iommu.c
+++ kvm/virt/kvm/iommu.c
@@ -78,7 +78,7 @@ static int kvm_iommu_map_memslots(struct
 	int i, r = 0;
 	struct kvm_memslots *slots;
 
-	slots = kvm->memslots;
+	slots = rcu_dereference(kvm->memslots);
 
 	for (i = 0; i < slots->nmemslots; i++) {
 		r = kvm_iommu_map_pages(kvm, &slots->memslots[i]);
@@ -214,7 +214,7 @@ static int kvm_iommu_unmap_memslots(stru
 	int i;
 	struct kvm_memslots *slots;
 
-	slots = kvm->memslots;
+	slots = rcu_dereference(kvm->memslots);
 
 	for (i = 0; i < slots->nmemslots; i++) {
 		kvm_iommu_put_pages(kvm, slots->memslots[i].base_gfn,

Re: [patch 06/11] KVM: introduce kvm->srcu and convert kvm_set_memory_region to SRCU update

[Avi Kivity]

On 12/23/2009 01:38 PM, Marcelo Tosatti wrote:
> Use two steps for memslot deletion: mark the slot invalid (which stops
> instantiation of new shadow pages for that slot, but allows destruction),
> then instantiate the new empty slot.
>
> Also simplifies kvm_handle_hva locking.
>
>
>   	r = kvm_arch_prepare_memory_region(kvm,&new, old, user_alloc);
>   	if (r)
>   		goto out_free;
>    

r == 0

>
> -	spin_lock(&kvm->mmu_lock);
> -	if (mem->slot>= kvm->memslots->nmemslots)
> -		kvm->memslots->nmemslots = mem->slot + 1;
> +#ifdef CONFIG_DMAR
> +	/* map the pages in iommu page table */
> +	if (npages)
> +		r = kvm_iommu_map_pages(kvm,&new);
> +		if (r)
> +			goto out_free;
>    

Bad indentation.

(still r == 0)

> +#endif
>
> -	*memslot = new;
> -	spin_unlock(&kvm->mmu_lock);
> +	slots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL);
> +	if (!slots)
> +		goto out_free;
>    

goto out_free with r == 0.

Best to assign r = -ENOMEM each time, to avoid these headaches.

>
> +int memslot_id(struct kvm *kvm, gfn_t gfn)
>    

Should be either static or kvm_memslot_id().  But the source is already 
like that, so leave it.

> @@ -807,23 +808,18 @@ static int kvm_handle_hva(struct kvm *kv
>   			  int (*handler)(struct kvm *kvm, unsigned long *rmapp,
>   					 unsigned long data))
>   {
> -	int i, j;
> +	int i, j, idx;
>   	int retval = 0;
> -	struct kvm_memslots *slots = kvm->memslots;
> +	struct kvm_memslots *slots;
> +
> +	idx = srcu_read_lock(&kvm->srcu);
>    

Maybe a better place is in the mmu notifiers, so the code is shared 
(once other archs start to use mmu notifiers).


> @@ -3020,16 +3017,20 @@ nomem:
>    */
>   unsigned int kvm_mmu_calculate_mmu_pages(struct kvm *kvm)
>   {
> -	int i;
> +	int i, idx;
>   	unsigned int nr_mmu_pages;
>   	unsigned int  nr_pages = 0;
> +	struct kvm_memslots *slots;
>
> -	for (i = 0; i<  kvm->memslots->nmemslots; i++)
> -		nr_pages += kvm->memslots->memslots[i].npages;
> +	idx = srcu_read_lock(&kvm->srcu);
> +	slots = rcu_dereference(kvm->memslots);
> +	for (i = 0; i<  slots->nmemslots; i++)
> +		nr_pages += slots->memslots[i].npages;
>
>   	nr_mmu_pages = nr_pages * KVM_PERMILLE_MMU_PAGES / 1000;
>   	nr_mmu_pages = max(nr_mmu_pages,
>   			(unsigned int) KVM_MIN_ALLOC_MMU_PAGES);
> +	srcu_read_unlock(&kvm->srcu, idx);
>
>    

Again, would like to move the srcu_locking to an outer scope.

> @@ -1503,10 +1504,18 @@ static void enter_pmode(struct kvm_vcpu
>   static gva_t rmode_tss_base(struct kvm *kvm)
>   {
>   	if (!kvm->arch.tss_addr) {
> -		gfn_t base_gfn = kvm->memslots->memslots[0].base_gfn +
> -				 kvm->memslots->memslots[0].npages - 3;
> +		struct kvm_memslots *slots;
> +		gfn_t base_gfn;
> +		int idx;
> +
> +		idx = srcu_read_lock(&kvm->srcu);
> +		slots = rcu_dereference(kvm->memslots);
> + 		base_gfn = slots->memslots[0].base_gfn +
> +				 slots->memslots[0].npages - 3;
> +		srcu_read_unlock(&kvm->srcu, idx);
>   		return base_gfn<<  PAGE_SHIFT;
>   	}
> +
>   	return kvm->arch.tss_addr;
>   }
>    

Shouldn't we already hold the srcu_lock as part of normal guest exit 
(similar to down_read() we do today)?

-- 
error compiling committee.c: too many arguments to function

Re: [patch 06/11] KVM: introduce kvm->srcu and convert kvm_set_memory_region to SRCU update

[Marcelo Tosatti]

On Wed, Dec 23, 2009 at 02:32:55PM +0200, Avi Kivity wrote:
> On 12/23/2009 01:38 PM, Marcelo Tosatti wrote:
>> Use two steps for memslot deletion: mark the slot invalid (which stops
>> instantiation of new shadow pages for that slot, but allows destruction),
>> then instantiate the new empty slot.
>>
>> Also simplifies kvm_handle_hva locking.
>>
>>
>>   	r = kvm_arch_prepare_memory_region(kvm,&new, old, user_alloc);
>>   	if (r)
>>   		goto out_free;
>>    
>
> r == 0

Huh? kvm_arch_prepare_memory_region returns a suitable error code on
failure, 0 on success.

>>
>> -	spin_lock(&kvm->mmu_lock);
>> -	if (mem->slot>= kvm->memslots->nmemslots)
>> -		kvm->memslots->nmemslots = mem->slot + 1;
>> +#ifdef CONFIG_DMAR
>> +	/* map the pages in iommu page table */
>> +	if (npages)
>> +		r = kvm_iommu_map_pages(kvm,&new);
>> +		if (r)
>> +			goto out_free;
>>    
>
> Bad indentation.
>
> (still r == 0)

kvm_iommu_map_pages returns 0 on success, error code otherwise? The
error code of kvm_iommu_map_pages was returned before the patchset, BTW.

>> +#endif
>>
>> -	*memslot = new;
>> -	spin_unlock(&kvm->mmu_lock);
>> +	slots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL);
>> +	if (!slots)
>> +		goto out_free;
>>    
>
> goto out_free with r == 0.
>
> Best to assign r = -ENOMEM each time, to avoid these headaches.

Fixed.

>> +int memslot_id(struct kvm *kvm, gfn_t gfn)
>>    
>
> Should be either static or kvm_memslot_id().  But the source is already  
> like that, so leave it.

OK.

>> @@ -807,23 +808,18 @@ static int kvm_handle_hva(struct kvm *kv
>>   			  int (*handler)(struct kvm *kvm, unsigned long *rmapp,
>>   					 unsigned long data))
>>   {
>> -	int i, j;
>> +	int i, j, idx;
>>   	int retval = 0;
>> -	struct kvm_memslots *slots = kvm->memslots;
>> +	struct kvm_memslots *slots;
>> +
>> +	idx = srcu_read_lock(&kvm->srcu);
>>    
>
> Maybe a better place is in the mmu notifiers, so the code is shared  
> (once other archs start to use mmu notifiers).

Done.

>> @@ -3020,16 +3017,20 @@ nomem:
>>    */
>>   unsigned int kvm_mmu_calculate_mmu_pages(struct kvm *kvm)
>>   {
>> -	int i;
>> +	int i, idx;
>>   	unsigned int nr_mmu_pages;
>>   	unsigned int  nr_pages = 0;
>> +	struct kvm_memslots *slots;
>>
>> -	for (i = 0; i<  kvm->memslots->nmemslots; i++)
>> -		nr_pages += kvm->memslots->memslots[i].npages;
>> +	idx = srcu_read_lock(&kvm->srcu);
>> +	slots = rcu_dereference(kvm->memslots);
>> +	for (i = 0; i<  slots->nmemslots; i++)
>> +		nr_pages += slots->memslots[i].npages;
>>
>>   	nr_mmu_pages = nr_pages * KVM_PERMILLE_MMU_PAGES / 1000;
>>   	nr_mmu_pages = max(nr_mmu_pages,
>>   			(unsigned int) KVM_MIN_ALLOC_MMU_PAGES);
>> +	srcu_read_unlock(&kvm->srcu, idx);
>>
>>    
>
> Again, would like to move the srcu_locking to an outer scope.

This one was for documentation purposes only, since
kvm_mmu_calculate_mmu_pages is only called from commit_memory_region
(which mutually excludes itself).

I can remove it if you'd prefer.

>> @@ -1503,10 +1504,18 @@ static void enter_pmode(struct kvm_vcpu
>>   static gva_t rmode_tss_base(struct kvm *kvm)
>>   {
>>   	if (!kvm->arch.tss_addr) {
>> -		gfn_t base_gfn = kvm->memslots->memslots[0].base_gfn +
>> -				 kvm->memslots->memslots[0].npages - 3;
>> +		struct kvm_memslots *slots;
>> +		gfn_t base_gfn;
>> +		int idx;
>> +
>> +		idx = srcu_read_lock(&kvm->srcu);
>> +		slots = rcu_dereference(kvm->memslots);
>> + 		base_gfn = slots->memslots[0].base_gfn +
>> +				 slots->memslots[0].npages - 3;
>> +		srcu_read_unlock(&kvm->srcu, idx);
>>   		return base_gfn<<  PAGE_SHIFT;
>>   	}
>> +
>>   	return kvm->arch.tss_addr;
>>   }
>>    
>
> Shouldn't we already hold the srcu_lock as part of normal guest exit  
> (similar to down_read() we do today)?

Yes, but:

kvm_arch_vcpu_setup -> vmx_vcpu_reset -> vmx_set_cr0 -> enter_pmode

So its called outside guest context path (memslot info is used outside
guest context in this case).

Re: [patch 06/11] KVM: introduce kvm->srcu and convert kvm_set_memory_region to SRCU update

[Avi Kivity]

On 12/23/2009 04:19 PM, Marcelo Tosatti wrote:
> On Wed, Dec 23, 2009 at 02:32:55PM +0200, Avi Kivity wrote:
>    
>> On 12/23/2009 01:38 PM, Marcelo Tosatti wrote:
>>      
>>> Use two steps for memslot deletion: mark the slot invalid (which stops
>>> instantiation of new shadow pages for that slot, but allows destruction),
>>> then instantiate the new empty slot.
>>>
>>> Also simplifies kvm_handle_hva locking.
>>>
>>>
>>>    	r = kvm_arch_prepare_memory_region(kvm,&new, old, user_alloc);
>>>    	if (r)
>>>    		goto out_free;
>>>
>>>        
>> r == 0
>>      
> Huh? kvm_arch_prepare_memory_region returns a suitable error code on
> failure, 0 on success.
>    

That's fine here, just pointing out you're inheriting this r == 0 later on.

> kvm_iommu_map_pages returns 0 on success, error code otherwise? The
> error code of kvm_iommu_map_pages was returned before the patchset, BTW.
>    

Ok, then please fix it separately for easier backporting.


>>>    	nr_mmu_pages = nr_pages * KVM_PERMILLE_MMU_PAGES / 1000;
>>>    	nr_mmu_pages = max(nr_mmu_pages,
>>>    			(unsigned int) KVM_MIN_ALLOC_MMU_PAGES);
>>> +	srcu_read_unlock(&kvm->srcu, idx);
>>>
>>>
>>>        
>> Again, would like to move the srcu_locking to an outer scope.
>>      
> This one was for documentation purposes only, since
> kvm_mmu_calculate_mmu_pages is only called from commit_memory_region
> (which mutually excludes itself).
>
> I can remove it if you'd prefer.
>    

A comment should be sufficient.  Otherwise someone might read the code 
and point out it's unnecessary.



>> Shouldn't we already hold the srcu_lock as part of normal guest exit
>> (similar to down_read() we do today)?
>>      
> Yes, but:
>
> kvm_arch_vcpu_setup ->  vmx_vcpu_reset ->  vmx_set_cr0 ->  enter_pmode
>
> So its called outside guest context path (memslot info is used outside
> guest context in this case).
>
>    

Didn't we take slots_lock previously?  If so, that's a bug.  We should 
probably make all vcpu ioctls take slots_lock at top level, since many 
vcpu ioctls can be caused to access memory.

-- 
error compiling committee.c: too many arguments to function

Re: [patch 06/11] KVM: introduce kvm->srcu and convert kvm_set_memory_region to SRCU update

[Marcelo Tosatti]

On Wed, Dec 23, 2009 at 04:33:06PM +0200, Avi Kivity wrote:
> On 12/23/2009 04:19 PM, Marcelo Tosatti wrote:
>> On Wed, Dec 23, 2009 at 02:32:55PM +0200, Avi Kivity wrote:
>>    
>>> On 12/23/2009 01:38 PM, Marcelo Tosatti wrote:
>>>      
>>>> Use two steps for memslot deletion: mark the slot invalid (which stops
>>>> instantiation of new shadow pages for that slot, but allows destruction),
>>>> then instantiate the new empty slot.
>>>>
>>>> Also simplifies kvm_handle_hva locking.
>>>>
>>>>
>>>>    	r = kvm_arch_prepare_memory_region(kvm,&new, old, user_alloc);
>>>>    	if (r)
>>>>    		goto out_free;
>>>>
>>>>        
>>> r == 0
>>>      
>> Huh? kvm_arch_prepare_memory_region returns a suitable error code on
>> failure, 0 on success.
>>    
>
> That's fine here, just pointing out you're inheriting this r == 0 later on.
>
>> kvm_iommu_map_pages returns 0 on success, error code otherwise? The
>> error code of kvm_iommu_map_pages was returned before the patchset, BTW.
>>    
>
> Ok, then please fix it separately for easier backporting.

Its not a bug.

>>>>    	nr_mmu_pages = nr_pages * KVM_PERMILLE_MMU_PAGES / 1000;
>>>>    	nr_mmu_pages = max(nr_mmu_pages,
>>>>    			(unsigned int) KVM_MIN_ALLOC_MMU_PAGES);
>>>> +	srcu_read_unlock(&kvm->srcu, idx);
>>>>
>>>>
>>>>        
>>> Again, would like to move the srcu_locking to an outer scope.
>>>      
>> This one was for documentation purposes only, since
>> kvm_mmu_calculate_mmu_pages is only called from commit_memory_region
>> (which mutually excludes itself).
>>
>> I can remove it if you'd prefer.
>>    
>
> A comment should be sufficient.  Otherwise someone might read the code  
> and point out it's unnecessary.
>
>
>
>>> Shouldn't we already hold the srcu_lock as part of normal guest exit
>>> (similar to down_read() we do today)?
>>>      
>> Yes, but:
>>
>> kvm_arch_vcpu_setup ->  vmx_vcpu_reset ->  vmx_set_cr0 ->  enter_pmode
>>
>> So its called outside guest context path (memslot info is used outside
>> guest context in this case).
>>
>>    
>
> Didn't we take slots_lock previously?  If so, that's a bug.  We should  
> probably make all vcpu ioctls take slots_lock at top level, since many  
> vcpu ioctls can be caused to access memory.

vmx_vcpu_reset takes it, so again, this is for documentation purposes
(its easier to make sense of srcu_read_lock + rcu_dereference together).

I'll remove it, since rcu_dereference should be enough to document the
locking.

[patch 07/11] KVM: use SRCU for dirty log

[Marcelo Tosatti]

[-- Attachment #1: use-srcu-for-dirty-log --]
[-- Type: text/plain, Size: 2025 bytes --]

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

Index: kvm/arch/x86/kvm/x86.c
===================================================================
--- kvm.orig/arch/x86/kvm/x86.c
+++ kvm/arch/x86/kvm/x86.c
@@ -2416,27 +2416,60 @@ static int kvm_vm_ioctl_reinject(struct 
 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
 				      struct kvm_dirty_log *log)
 {
-	int r;
-	int n;
+	int r, n, i;
 	struct kvm_memory_slot *memslot;
-	int is_dirty = 0;
+	unsigned long is_dirty = 0;
+	unsigned long *dirty_bitmap = NULL;
 
 	down_write(&kvm->slots_lock);
 
-	r = kvm_get_dirty_log(kvm, log, &is_dirty);
-	if (r)
+	r = -EINVAL;
+	if (log->slot >= KVM_MEMORY_SLOTS)
+		goto out;
+
+	memslot = &kvm->memslots->memslots[log->slot];
+	r = -ENOENT;
+	if (!memslot->dirty_bitmap)
+		goto out;
+
+	n = ALIGN(memslot->npages, BITS_PER_LONG) / 8;
+
+	r = -ENOMEM;
+	dirty_bitmap = vmalloc(n);
+	if (!dirty_bitmap)
 		goto out;
+	memset(dirty_bitmap, 0, n);
+
+	for (i = 0; !is_dirty && i < n/sizeof(long); i++)
+		is_dirty = memslot->dirty_bitmap[i];
 
 	/* If nothing is dirty, don't bother messing with page tables. */
 	if (is_dirty) {
+		struct kvm_memslots *slots, *old_slots;
+
 		spin_lock(&kvm->mmu_lock);
 		kvm_mmu_slot_remove_write_access(kvm, log->slot);
 		spin_unlock(&kvm->mmu_lock);
-		memslot = &kvm->memslots->memslots[log->slot];
-		n = ALIGN(memslot->npages, BITS_PER_LONG) / 8;
-		memset(memslot->dirty_bitmap, 0, n);
+
+		slots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL);
+		if (!slots)
+			goto out_free;
+
+		memcpy(slots, kvm->memslots, sizeof(struct kvm_memslots));
+		slots->memslots[log->slot].dirty_bitmap = dirty_bitmap;
+
+		old_slots = kvm->memslots;
+		rcu_assign_pointer(kvm->memslots, slots);
+		synchronize_srcu_expedited(&kvm->srcu);
+		dirty_bitmap = old_slots->memslots[log->slot].dirty_bitmap;
+		kfree(old_slots);
 	}
+
 	r = 0;
+	if (copy_to_user(log->dirty_bitmap, dirty_bitmap, n))
+		r = -EFAULT;
+out_free:
+	vfree(dirty_bitmap);
 out:
 	up_write(&kvm->slots_lock);
 	return r;

[patch 08/11] KVM: x86: switch kvm_set_memory_alias to SRCU update

[Marcelo Tosatti]

[-- Attachment #1: srcu-alias-update --]
[-- Type: text/plain, Size: 5255 bytes --]

Using a similar two-step procedure as for memslots.

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

Index: kvm/arch/x86/include/asm/kvm_host.h
===================================================================
--- kvm.orig/arch/x86/include/asm/kvm_host.h
+++ kvm/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;
Index: kvm/arch/x86/kvm/x86.c
===================================================================
--- kvm.orig/arch/x86/kvm/x86.c
+++ kvm/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
@@ -2224,11 +2225,32 @@ static int kvm_vm_ioctl_get_nr_mmu_pages
 	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];
@@ -2249,7 +2271,7 @@ static int kvm_vm_ioctl_set_memory_alias
 {
 	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 */
@@ -2266,28 +2288,48 @@ static int kvm_vm_ioctl_set_memory_alias
 	    < 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;
 }
Index: kvm/include/linux/kvm_host.h
===================================================================
--- kvm.orig/include/linux/kvm_host.h
+++ kvm/include/linux/kvm_host.h
@@ -265,6 +265,8 @@ void kvm_arch_commit_memory_region(struc
 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);
@@ -538,6 +540,10 @@ static inline int mmu_notifier_retry(str
 }
 #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
Index: kvm/virt/kvm/kvm_main.c
===================================================================
--- kvm.orig/virt/kvm/kvm_main.c
+++ kvm/virt/kvm/kvm_main.c
@@ -847,7 +847,7 @@ int kvm_is_visible_gfn(struct kvm *kvm, 
 	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];
 
@@ -884,7 +884,7 @@ unsigned long gfn_to_hva(struct kvm *kvm
 {
 	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();

[patch 09/11] KVM: convert io_bus to SRCU

[Marcelo Tosatti]

[-- Attachment #1: srcu-iobus-update --]
[-- Type: text/plain, Size: 14236 bytes --]

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

Index: kvm/arch/x86/kvm/i8254.c
===================================================================
--- kvm.orig/arch/x86/kvm/i8254.c
+++ kvm/arch/x86/kvm/i8254.c
@@ -642,13 +642,13 @@ struct kvm_pit *kvm_create_pit(struct kv
 	kvm_register_irq_mask_notifier(kvm, 0, &pit->mask_notifier);
 
 	kvm_iodevice_init(&pit->dev, &pit_dev_ops);
-	ret = __kvm_io_bus_register_dev(&kvm->pio_bus, &pit->dev);
+	ret = kvm_io_bus_register_dev(kvm, KVM_PIO_BUS, &pit->dev);
 	if (ret < 0)
 		goto fail;
 
 	if (flags & KVM_PIT_SPEAKER_DUMMY) {
 		kvm_iodevice_init(&pit->speaker_dev, &speaker_dev_ops);
-		ret = __kvm_io_bus_register_dev(&kvm->pio_bus,
+		ret = kvm_io_bus_register_dev(kvm, KVM_PIO_BUS,
 						&pit->speaker_dev);
 		if (ret < 0)
 			goto fail_unregister;
@@ -657,7 +657,7 @@ struct kvm_pit *kvm_create_pit(struct kv
 	return pit;
 
 fail_unregister:
-	__kvm_io_bus_unregister_dev(&kvm->pio_bus, &pit->dev);
+	kvm_io_bus_unregister_dev(kvm, KVM_PIO_BUS, &pit->dev);
 
 fail:
 	if (pit->irq_source_id >= 0)
Index: kvm/arch/x86/kvm/i8259.c
===================================================================
--- kvm.orig/arch/x86/kvm/i8259.c
+++ kvm/arch/x86/kvm/i8259.c
@@ -533,7 +533,9 @@ struct kvm_pic *kvm_create_pic(struct kv
 	 * Initialize PIO device
 	 */
 	kvm_iodevice_init(&s->dev, &picdev_ops);
-	ret = kvm_io_bus_register_dev(kvm, &kvm->pio_bus, &s->dev);
+	down_write(&kvm->slots_lock);
+	ret = kvm_io_bus_register_dev(kvm, KVM_PIO_BUS, &s->dev);
+	up_write(&kvm->slots_lock);
 	if (ret < 0) {
 		kfree(s);
 		return NULL;
Index: kvm/include/linux/kvm_host.h
===================================================================
--- kvm.orig/include/linux/kvm_host.h
+++ kvm/include/linux/kvm_host.h
@@ -57,20 +57,20 @@ struct kvm_io_bus {
 	struct kvm_io_device *devs[NR_IOBUS_DEVS];
 };
 
-void kvm_io_bus_init(struct kvm_io_bus *bus);
-void kvm_io_bus_destroy(struct kvm_io_bus *bus);
-int kvm_io_bus_write(struct kvm_io_bus *bus, gpa_t addr, int len,
-		     const void *val);
-int kvm_io_bus_read(struct kvm_io_bus *bus, gpa_t addr, int len,
+enum kvm_bus {
+	KVM_MMIO_BUS,
+	KVM_PIO_BUS,
+	KVM_NR_BUSES
+};
+
+int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
+		     int len, const void *val);
+int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, int len,
 		    void *val);
-int __kvm_io_bus_register_dev(struct kvm_io_bus *bus,
-			       struct kvm_io_device *dev);
-int kvm_io_bus_register_dev(struct kvm *kvm, struct kvm_io_bus *bus,
+int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx,
 			    struct kvm_io_device *dev);
-void __kvm_io_bus_unregister_dev(struct kvm_io_bus *bus,
-				 struct kvm_io_device *dev);
-void kvm_io_bus_unregister_dev(struct kvm *kvm, struct kvm_io_bus *bus,
-			       struct kvm_io_device *dev);
+int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
+			      struct kvm_io_device *dev);
 
 struct kvm_vcpu {
 	struct kvm *kvm;
@@ -171,8 +171,7 @@ struct kvm {
 	atomic_t online_vcpus;
 	struct list_head vm_list;
 	struct mutex lock;
-	struct kvm_io_bus mmio_bus;
-	struct kvm_io_bus pio_bus;
+	struct kvm_io_bus *buses[KVM_NR_BUSES];
 #ifdef CONFIG_HAVE_KVM_EVENTFD
 	struct {
 		spinlock_t        lock;
Index: kvm/virt/kvm/eventfd.c
===================================================================
--- kvm.orig/virt/kvm/eventfd.c
+++ kvm/virt/kvm/eventfd.c
@@ -451,7 +451,7 @@ static int
 kvm_assign_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
 {
 	int                       pio = args->flags & KVM_IOEVENTFD_FLAG_PIO;
-	struct kvm_io_bus        *bus = pio ? &kvm->pio_bus : &kvm->mmio_bus;
+	enum kvm_bus              bus_idx = pio ? KVM_PIO_BUS : KVM_MMIO_BUS;
 	struct _ioeventfd        *p;
 	struct eventfd_ctx       *eventfd;
 	int                       ret;
@@ -506,7 +506,7 @@ kvm_assign_ioeventfd(struct kvm *kvm, st
 
 	kvm_iodevice_init(&p->dev, &ioeventfd_ops);
 
-	ret = __kvm_io_bus_register_dev(bus, &p->dev);
+	ret = kvm_io_bus_register_dev(kvm, bus_idx, &p->dev);
 	if (ret < 0)
 		goto unlock_fail;
 
@@ -530,7 +530,7 @@ static int
 kvm_deassign_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
 {
 	int                       pio = args->flags & KVM_IOEVENTFD_FLAG_PIO;
-	struct kvm_io_bus        *bus = pio ? &kvm->pio_bus : &kvm->mmio_bus;
+	enum kvm_bus              bus_idx = pio ? KVM_PIO_BUS : KVM_MMIO_BUS;
 	struct _ioeventfd        *p, *tmp;
 	struct eventfd_ctx       *eventfd;
 	int                       ret = -ENOENT;
@@ -553,7 +553,7 @@ kvm_deassign_ioeventfd(struct kvm *kvm, 
 		if (!p->wildcard && p->datamatch != args->datamatch)
 			continue;
 
-		__kvm_io_bus_unregister_dev(bus, &p->dev);
+		kvm_io_bus_unregister_dev(kvm, bus_idx, &p->dev);
 		ioeventfd_release(p);
 		ret = 0;
 		break;
Index: kvm/virt/kvm/ioapic.c
===================================================================
--- kvm.orig/virt/kvm/ioapic.c
+++ kvm/virt/kvm/ioapic.c
@@ -372,7 +372,9 @@ int kvm_ioapic_init(struct kvm *kvm)
 	kvm_ioapic_reset(ioapic);
 	kvm_iodevice_init(&ioapic->dev, &ioapic_mmio_ops);
 	ioapic->kvm = kvm;
-	ret = kvm_io_bus_register_dev(kvm, &kvm->mmio_bus, &ioapic->dev);
+	down_write(&kvm->slots_lock);
+	ret = kvm_io_bus_register_dev(kvm, KVM_MMIO_BUS, &ioapic->dev);
+	up_write(&kvm->slots_lock);
 	if (ret < 0)
 		kfree(ioapic);
 
Index: kvm/virt/kvm/kvm_main.c
===================================================================
--- kvm.orig/virt/kvm/kvm_main.c
+++ kvm/virt/kvm/kvm_main.c
@@ -85,6 +85,8 @@ static long kvm_vcpu_ioctl(struct file *
 static int hardware_enable_all(void);
 static void hardware_disable_all(void);
 
+static void kvm_io_bus_destroy(struct kvm_io_bus *bus);
+
 static bool kvm_rebooting;
 
 static bool largepages_enabled = true;
@@ -358,7 +360,7 @@ static int kvm_init_mmu_notifier(struct 
 
 static struct kvm *kvm_create_vm(void)
 {
-	int r = 0;
+	int r = 0, i;
 	struct kvm *kvm = kvm_arch_create_vm();
 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
 	struct page *page;
@@ -382,6 +384,14 @@ static struct kvm *kvm_create_vm(void)
 		goto out_err;
 	if (init_srcu_struct(&kvm->srcu))
 		goto out_err;
+ 	for (i = 0; i < KVM_NR_BUSES; i++) {
+ 		kvm->buses[i] = kzalloc(sizeof(struct kvm_io_bus),
+ 					GFP_KERNEL);
+ 		if (!kvm->buses[i]) {
+ 			cleanup_srcu_struct(&kvm->srcu);
+ 			goto out_err;
+ 		}
+ 	}
 
 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
 	page = alloc_page(GFP_KERNEL | __GFP_ZERO);
@@ -407,11 +417,9 @@ static struct kvm *kvm_create_vm(void)
 	atomic_inc(&kvm->mm->mm_count);
 	spin_lock_init(&kvm->mmu_lock);
 	spin_lock_init(&kvm->requests_lock);
-	kvm_io_bus_init(&kvm->pio_bus);
 	kvm_eventfd_init(kvm);
 	mutex_init(&kvm->lock);
 	mutex_init(&kvm->irq_lock);
-	kvm_io_bus_init(&kvm->mmio_bus);
 	init_rwsem(&kvm->slots_lock);
 	atomic_set(&kvm->users_count, 1);
 	spin_lock(&kvm_lock);
@@ -426,6 +434,8 @@ out:
 out_err:
 	hardware_disable_all();
 out_err_nodisable:
+ 	for (i = 0; i < KVM_NR_BUSES; i++)
+ 		kfree(kvm->buses[i]);
 	kfree(kvm->memslots);
 	kfree(kvm);
 	return ERR_PTR(r);
@@ -471,6 +481,7 @@ void kvm_free_physmem(struct kvm *kvm)
 
 static void kvm_destroy_vm(struct kvm *kvm)
 {
+	int i;
 	struct mm_struct *mm = kvm->mm;
 
 	kvm_arch_sync_events(kvm);
@@ -478,8 +489,8 @@ static void kvm_destroy_vm(struct kvm *k
 	list_del(&kvm->vm_list);
 	spin_unlock(&kvm_lock);
 	kvm_free_irq_routing(kvm);
-	kvm_io_bus_destroy(&kvm->pio_bus);
-	kvm_io_bus_destroy(&kvm->mmio_bus);
+	for (i = 0; i < KVM_NR_BUSES; i++)
+		kvm_io_bus_destroy(kvm->buses[i]);
 	kvm_coalesced_mmio_free(kvm);
 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
 	mmu_notifier_unregister(&kvm->mmu_notifier, kvm->mm);
@@ -1937,12 +1948,7 @@ static struct notifier_block kvm_reboot_
 	.priority = 0,
 };
 
-void kvm_io_bus_init(struct kvm_io_bus *bus)
-{
-	memset(bus, 0, sizeof(*bus));
-}
-
-void kvm_io_bus_destroy(struct kvm_io_bus *bus)
+static void kvm_io_bus_destroy(struct kvm_io_bus *bus)
 {
 	int i;
 
@@ -1951,13 +1957,15 @@ void kvm_io_bus_destroy(struct kvm_io_bu
 
 		kvm_iodevice_destructor(pos);
 	}
+	kfree(bus);
 }
 
 /* kvm_io_bus_write - called under kvm->slots_lock */
-int kvm_io_bus_write(struct kvm_io_bus *bus, gpa_t addr,
+int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
 		     int len, const void *val)
 {
 	int i;
+	struct kvm_io_bus *bus = rcu_dereference(kvm->buses[bus_idx]);
 	for (i = 0; i < bus->dev_count; i++)
 		if (!kvm_iodevice_write(bus->devs[i], addr, len, val))
 			return 0;
@@ -1965,59 +1973,71 @@ int kvm_io_bus_write(struct kvm_io_bus *
 }
 
 /* kvm_io_bus_read - called under kvm->slots_lock */
-int kvm_io_bus_read(struct kvm_io_bus *bus, gpa_t addr, int len, void *val)
+int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
+		    int len, void *val)
 {
 	int i;
+	struct kvm_io_bus *bus = rcu_dereference(kvm->buses[bus_idx]);
+
 	for (i = 0; i < bus->dev_count; i++)
 		if (!kvm_iodevice_read(bus->devs[i], addr, len, val))
 			return 0;
 	return -EOPNOTSUPP;
 }
 
-int kvm_io_bus_register_dev(struct kvm *kvm, struct kvm_io_bus *bus,
-			     struct kvm_io_device *dev)
+/* Caller must have write lock on slots_lock. */
+int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx,
+			    struct kvm_io_device *dev)
 {
-	int ret;
+	struct kvm_io_bus *new_bus, *bus;
 
-	down_write(&kvm->slots_lock);
-	ret = __kvm_io_bus_register_dev(bus, dev);
-	up_write(&kvm->slots_lock);
-
-	return ret;
-}
-
-/* An unlocked version. Caller must have write lock on slots_lock. */
-int __kvm_io_bus_register_dev(struct kvm_io_bus *bus,
-			      struct kvm_io_device *dev)
-{
+	bus = kvm->buses[bus_idx];
 	if (bus->dev_count > NR_IOBUS_DEVS-1)
 		return -ENOSPC;
 
-	bus->devs[bus->dev_count++] = dev;
+	new_bus = kzalloc(sizeof(struct kvm_io_bus), GFP_KERNEL);
+	if (!new_bus)
+		return -ENOMEM;
+	memcpy(new_bus, bus, sizeof(struct kvm_io_bus));
+	new_bus->devs[new_bus->dev_count++] = dev;
+	rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
+	synchronize_srcu_expedited(&kvm->srcu);
+	kfree(bus);
 
 	return 0;
 }
 
-void kvm_io_bus_unregister_dev(struct kvm *kvm,
-			       struct kvm_io_bus *bus,
-			       struct kvm_io_device *dev)
+/* Caller must have write lock on slots_lock. */
+int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
+			      struct kvm_io_device *dev)
 {
-	down_write(&kvm->slots_lock);
-	__kvm_io_bus_unregister_dev(bus, dev);
-	up_write(&kvm->slots_lock);
-}
+	int i, r;
+	struct kvm_io_bus *new_bus, *bus;
 
-/* An unlocked version. Caller must have write lock on slots_lock. */
-void __kvm_io_bus_unregister_dev(struct kvm_io_bus *bus,
-				 struct kvm_io_device *dev)
-{
-	int i;
+	new_bus = kzalloc(sizeof(struct kvm_io_bus), GFP_KERNEL);
+	if (!new_bus)
+		return -ENOMEM;
 
-	for (i = 0; i < bus->dev_count; i++)
-		if (bus->devs[i] == dev) {
-			bus->devs[i] = bus->devs[--bus->dev_count];
+	bus = kvm->buses[bus_idx];
+	memcpy(new_bus, bus, sizeof(struct kvm_io_bus));
+
+	r = -ENOENT;
+	for (i = 0; i < new_bus->dev_count; i++)
+		if (new_bus->devs[i] == dev) {
+			r = 0;
+			new_bus->devs[i] = new_bus->devs[--new_bus->dev_count];
 			break;
 		}
+
+	if (r) {
+		kfree(new_bus);
+		return r;
+	}
+
+	rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
+	synchronize_srcu_expedited(&kvm->srcu);
+	kfree(bus);
+	return r;
 }
 
 static struct notifier_block kvm_cpu_notifier = {
Index: kvm/virt/kvm/coalesced_mmio.c
===================================================================
--- kvm.orig/virt/kvm/coalesced_mmio.c
+++ kvm/virt/kvm/coalesced_mmio.c
@@ -110,7 +110,9 @@ int kvm_coalesced_mmio_init(struct kvm *
 	dev->kvm = kvm;
 	kvm->coalesced_mmio_dev = dev;
 
-	ret = kvm_io_bus_register_dev(kvm, &kvm->mmio_bus, &dev->dev);
+	down_write(&kvm->slots_lock);
+	ret = kvm_io_bus_register_dev(kvm, KVM_MMIO_BUS, &dev->dev);
+	up_write(&kvm->slots_lock);
 	if (ret < 0)
 		goto out_free_dev;
 
Index: kvm/arch/x86/kvm/x86.c
===================================================================
--- kvm.orig/arch/x86/kvm/x86.c
+++ kvm/arch/x86/kvm/x86.c
@@ -2852,7 +2852,7 @@ static int vcpu_mmio_write(struct kvm_vc
 	    !kvm_iodevice_write(&vcpu->arch.apic->dev, addr, len, v))
 		return 0;
 
-	return kvm_io_bus_write(&vcpu->kvm->mmio_bus, addr, len, v);
+	return kvm_io_bus_write(vcpu->kvm, KVM_MMIO_BUS, addr, len, v);
 }
 
 static int vcpu_mmio_read(struct kvm_vcpu *vcpu, gpa_t addr, int len, void *v)
@@ -2861,7 +2861,7 @@ static int vcpu_mmio_read(struct kvm_vcp
 	    !kvm_iodevice_read(&vcpu->arch.apic->dev, addr, len, v))
 		return 0;
 
-	return kvm_io_bus_read(&vcpu->kvm->mmio_bus, addr, len, v);
+	return kvm_io_bus_read(vcpu->kvm, KVM_MMIO_BUS, addr, len, v);
 }
 
 static int kvm_read_guest_virt(gva_t addr, void *val, unsigned int bytes,
@@ -3346,11 +3346,12 @@ static int kernel_pio(struct kvm_vcpu *v
 	int r;
 
 	if (vcpu->arch.pio.in)
-		r = kvm_io_bus_read(&vcpu->kvm->pio_bus, vcpu->arch.pio.port,
+		r = kvm_io_bus_read(vcpu->kvm, KVM_PIO_BUS, vcpu->arch.pio.port,
 				    vcpu->arch.pio.size, pd);
 	else
-		r = kvm_io_bus_write(&vcpu->kvm->pio_bus, vcpu->arch.pio.port,
-				     vcpu->arch.pio.size, pd);
+		r = kvm_io_bus_write(vcpu->kvm, KVM_PIO_BUS,
+				     vcpu->arch.pio.port, vcpu->arch.pio.size,
+				     pd);
 	return r;
 }
 
@@ -3361,7 +3362,7 @@ static int pio_string_write(struct kvm_v
 	int i, r = 0;
 
 	for (i = 0; i < io->cur_count; i++) {
-		if (kvm_io_bus_write(&vcpu->kvm->pio_bus,
+		if (kvm_io_bus_write(vcpu->kvm, KVM_PIO_BUS,
 				     io->port, io->size, pd)) {
 			r = -EOPNOTSUPP;
 			break;
Index: kvm/arch/ia64/kvm/kvm-ia64.c
===================================================================
--- kvm.orig/arch/ia64/kvm/kvm-ia64.c
+++ kvm/arch/ia64/kvm/kvm-ia64.c
@@ -241,10 +241,10 @@ static int handle_mmio(struct kvm_vcpu *
 	return 0;
 mmio:
 	if (p->dir)
-		r = kvm_io_bus_read(&vcpu->kvm->mmio_bus, p->addr,
+		r = kvm_io_bus_read(vcpu->kvm, KVM_MMIO_BUS, p->addr,
 				    p->size, &p->data);
 	else
-		r = kvm_io_bus_write(&vcpu->kvm->mmio_bus, p->addr,
+		r = kvm_io_bus_write(vcpu->kvm, KVM_MMIO_BUS, p->addr,
 				     p->size, &p->data);
 	if (r)
 		printk(KERN_ERR"kvm: No iodevice found! addr:%lx\n", p->addr);

[patch 10/11] KVM: switch vcpu context to use SRCU

[Marcelo Tosatti]

[-- Attachment #1: switch-to-memslot-srcu --]
[-- Type: text/plain, Size: 8330 bytes --]

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

Index: kvm/arch/x86/kvm/x86.c
===================================================================
--- kvm.orig/arch/x86/kvm/x86.c
+++ kvm/arch/x86/kvm/x86.c
@@ -1307,15 +1307,15 @@ static int __msr_io(struct kvm_vcpu *vcp
 		    int (*do_msr)(struct kvm_vcpu *vcpu,
 				  unsigned index, u64 *data))
 {
-	int i;
+	int i, idx;
 
 	vcpu_load(vcpu);
 
-	down_read(&vcpu->kvm->slots_lock);
+	idx = srcu_read_lock(&vcpu->kvm->srcu);
 	for (i = 0; i < msrs->nmsrs; ++i)
 		if (do_msr(vcpu, entries[i].index, &entries[i].data))
 			break;
-	up_read(&vcpu->kvm->slots_lock);
+	srcu_read_unlock(&vcpu->kvm->srcu, idx);
 
 	vcpu_put(vcpu);
 
@@ -3901,14 +3901,15 @@ static void vapic_enter(struct kvm_vcpu 
 static void vapic_exit(struct kvm_vcpu *vcpu)
 {
 	struct kvm_lapic *apic = vcpu->arch.apic;
+	int idx;
 
 	if (!apic || !apic->vapic_addr)
 		return;
 
-	down_read(&vcpu->kvm->slots_lock);
+	idx = srcu_read_lock(&vcpu->kvm->srcu);
 	kvm_release_page_dirty(apic->vapic_page);
 	mark_page_dirty(vcpu->kvm, apic->vapic_addr >> PAGE_SHIFT);
-	up_read(&vcpu->kvm->slots_lock);
+	srcu_read_unlock(&vcpu->kvm->srcu, idx);
 }
 
 static void update_cr8_intercept(struct kvm_vcpu *vcpu)
@@ -4037,7 +4038,7 @@ static int vcpu_enter_guest(struct kvm_v
 		kvm_lapic_sync_to_vapic(vcpu);
 	}
 
-	up_read(&vcpu->kvm->slots_lock);
+	srcu_read_unlock(&vcpu->kvm->srcu, vcpu->srcu_idx);
 
 	kvm_guest_enter();
 
@@ -4079,7 +4080,7 @@ static int vcpu_enter_guest(struct kvm_v
 
 	preempt_enable();
 
-	down_read(&vcpu->kvm->slots_lock);
+	vcpu->srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
 
 	/*
 	 * Profile KVM exit RIPs:
@@ -4101,6 +4102,7 @@ out:
 static int __vcpu_run(struct kvm_vcpu *vcpu)
 {
 	int r;
+	struct kvm *kvm = vcpu->kvm;
 
 	if (unlikely(vcpu->arch.mp_state == KVM_MP_STATE_SIPI_RECEIVED)) {
 		pr_debug("vcpu %d received sipi with vector # %x\n",
@@ -4112,7 +4114,7 @@ static int __vcpu_run(struct kvm_vcpu *v
 		vcpu->arch.mp_state = KVM_MP_STATE_RUNNABLE;
 	}
 
-	down_read(&vcpu->kvm->slots_lock);
+	vcpu->srcu_idx = srcu_read_lock(&kvm->srcu);
 	vapic_enter(vcpu);
 
 	r = 1;
@@ -4120,9 +4122,9 @@ static int __vcpu_run(struct kvm_vcpu *v
 		if (vcpu->arch.mp_state == KVM_MP_STATE_RUNNABLE)
 			r = vcpu_enter_guest(vcpu);
 		else {
-			up_read(&vcpu->kvm->slots_lock);
+			srcu_read_unlock(&kvm->srcu, vcpu->srcu_idx);
 			kvm_vcpu_block(vcpu);
-			down_read(&vcpu->kvm->slots_lock);
+			vcpu->srcu_idx = srcu_read_lock(&kvm->srcu);
 			if (test_and_clear_bit(KVM_REQ_UNHALT, &vcpu->requests))
 			{
 				switch(vcpu->arch.mp_state) {
@@ -4157,13 +4159,13 @@ static int __vcpu_run(struct kvm_vcpu *v
 			++vcpu->stat.signal_exits;
 		}
 		if (need_resched()) {
-			up_read(&vcpu->kvm->slots_lock);
+			srcu_read_unlock(&kvm->srcu, vcpu->srcu_idx);
 			kvm_resched(vcpu);
-			down_read(&vcpu->kvm->slots_lock);
+			vcpu->srcu_idx = srcu_read_lock(&kvm->srcu);
 		}
 	}
 
-	up_read(&vcpu->kvm->slots_lock);
+	srcu_read_unlock(&kvm->srcu, vcpu->srcu_idx);
 	post_kvm_run_save(vcpu);
 
 	vapic_exit(vcpu);
@@ -4202,10 +4204,10 @@ int kvm_arch_vcpu_ioctl_run(struct kvm_v
 		vcpu->mmio_read_completed = 1;
 		vcpu->mmio_needed = 0;
 
-		down_read(&vcpu->kvm->slots_lock);
+		vcpu->srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
 		r = emulate_instruction(vcpu, vcpu->arch.mmio_fault_cr2, 0,
 					EMULTYPE_NO_DECODE);
-		up_read(&vcpu->kvm->slots_lock);
+		srcu_read_unlock(&vcpu->kvm->srcu, vcpu->srcu_idx);
 		if (r == EMULATE_DO_MMIO) {
 			/*
 			 * Read-modify-write.  Back to userspace.
@@ -4968,11 +4970,12 @@ int kvm_arch_vcpu_ioctl_translate(struct
 {
 	unsigned long vaddr = tr->linear_address;
 	gpa_t gpa;
+	int idx;
 
 	vcpu_load(vcpu);
-	down_read(&vcpu->kvm->slots_lock);
+	idx = srcu_read_lock(&vcpu->kvm->srcu);
 	gpa = vcpu->arch.mmu.gva_to_gpa(vcpu, vaddr);
-	up_read(&vcpu->kvm->slots_lock);
+	srcu_read_unlock(&vcpu->kvm->srcu, idx);
 	tr->physical_address = gpa;
 	tr->valid = gpa != UNMAPPED_GVA;
 	tr->writeable = 1;
@@ -5223,10 +5226,12 @@ fail:
 
 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu)
 {
+	int idx;
+
 	kvm_free_lapic(vcpu);
-	down_read(&vcpu->kvm->slots_lock);
+	idx = srcu_read_lock(&vcpu->kvm->srcu);
 	kvm_mmu_destroy(vcpu);
-	up_read(&vcpu->kvm->slots_lock);
+	srcu_read_unlock(&vcpu->kvm->srcu, idx);
 	free_page((unsigned long)vcpu->arch.pio_data);
 }
 
Index: kvm/include/linux/kvm_host.h
===================================================================
--- kvm.orig/include/linux/kvm_host.h
+++ kvm/include/linux/kvm_host.h
@@ -83,6 +83,8 @@ struct kvm_vcpu {
 	struct kvm_run *run;
 	unsigned long requests;
 	unsigned long guest_debug;
+	int srcu_idx;
+
 	int fpu_active;
 	int guest_fpu_loaded;
 	wait_queue_head_t wq;
Index: kvm/arch/x86/kvm/mmu.c
===================================================================
--- kvm.orig/arch/x86/kvm/mmu.c
+++ kvm/arch/x86/kvm/mmu.c
@@ -2937,10 +2937,9 @@ static int mmu_shrink(int nr_to_scan, gf
 	spin_lock(&kvm_lock);
 
 	list_for_each_entry(kvm, &vm_list, vm_list) {
-		int npages;
+		int npages, idx;
 
-		if (!down_read_trylock(&kvm->slots_lock))
-			continue;
+		idx = srcu_read_lock(&kvm->srcu);
 		spin_lock(&kvm->mmu_lock);
 		npages = kvm->arch.n_alloc_mmu_pages -
 			 kvm->arch.n_free_mmu_pages;
@@ -2953,7 +2952,7 @@ static int mmu_shrink(int nr_to_scan, gf
 		nr_to_scan--;
 
 		spin_unlock(&kvm->mmu_lock);
-		up_read(&kvm->slots_lock);
+		srcu_read_unlock(&kvm->srcu, idx);
 	}
 	if (kvm_freed)
 		list_move_tail(&kvm_freed->vm_list, &vm_list);
Index: kvm/arch/x86/kvm/vmx.c
===================================================================
--- kvm.orig/arch/x86/kvm/vmx.c
+++ kvm/arch/x86/kvm/vmx.c
@@ -2482,10 +2482,10 @@ static int vmx_vcpu_reset(struct kvm_vcp
 {
 	struct vcpu_vmx *vmx = to_vmx(vcpu);
 	u64 msr;
-	int ret;
+	int ret, idx;
 
 	vcpu->arch.regs_avail = ~((1 << VCPU_REGS_RIP) | (1 << VCPU_REGS_RSP));
-	down_read(&vcpu->kvm->slots_lock);
+	idx = srcu_read_lock(&vcpu->kvm->srcu);
 	if (!init_rmode(vmx->vcpu.kvm)) {
 		ret = -ENOMEM;
 		goto out;
@@ -2593,7 +2593,7 @@ static int vmx_vcpu_reset(struct kvm_vcp
 	vmx->emulation_required = 0;
 
 out:
-	up_read(&vcpu->kvm->slots_lock);
+	srcu_read_unlock(&vcpu->kvm->srcu, idx);
 	return ret;
 }
 
Index: kvm/arch/ia64/kvm/kvm-ia64.c
===================================================================
--- kvm.orig/arch/ia64/kvm/kvm-ia64.c
+++ kvm/arch/ia64/kvm/kvm-ia64.c
@@ -636,12 +636,9 @@ static void kvm_vcpu_post_transition(str
 static int __vcpu_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
 {
 	union context *host_ctx, *guest_ctx;
-	int r;
+	int r, idx;
 
-	/*
-	 * down_read() may sleep and return with interrupts enabled
-	 */
-	down_read(&vcpu->kvm->slots_lock);
+	idx = srcu_read_lock(&vcpu->kvm->srcu);
 
 again:
 	if (signal_pending(current)) {
@@ -663,7 +660,7 @@ again:
 	if (r < 0)
 		goto vcpu_run_fail;
 
-	up_read(&vcpu->kvm->slots_lock);
+	srcu_read_unlock(&vcpu->kvm->srcu, idx);
 	kvm_guest_enter();
 
 	/*
@@ -687,7 +684,7 @@ again:
 	kvm_guest_exit();
 	preempt_enable();
 
-	down_read(&vcpu->kvm->slots_lock);
+	idx = srcu_read_lock(&vcpu->kvm->srcu);
 
 	r = kvm_handle_exit(kvm_run, vcpu);
 
@@ -697,10 +694,10 @@ again:
 	}
 
 out:
-	up_read(&vcpu->kvm->slots_lock);
+	srcu_read_unlock(&vcpu->kvm->srcu, idx);
 	if (r > 0) {
 		kvm_resched(vcpu);
-		down_read(&vcpu->kvm->slots_lock);
+		idx = srcu_read_lock(&vcpu->kvm->srcu);
 		goto again;
 	}
 
Index: kvm/arch/s390/kvm/kvm-s390.h
===================================================================
--- kvm.orig/arch/s390/kvm/kvm-s390.h
+++ kvm/arch/s390/kvm/kvm-s390.h
@@ -67,10 +67,14 @@ static inline long kvm_s390_vcpu_get_mem
 
 static inline void kvm_s390_vcpu_set_mem(struct kvm_vcpu *vcpu)
 {
+	int idx;
 	struct kvm_memory_slot *mem;
+	struct kvm_memory_slots *memslots;
 
-	down_read(&vcpu->kvm->slots_lock);
-	mem = &vcpu->kvm->memslots[0];
+	idx = srcu_read_lock(&vcpu->kvm->srcu);
+	memslots = rcu_dereference(vcpu->kvm->memslots);
+
+	mem = &memslots->memslots[0];
 
 	vcpu->arch.sie_block->gmsor = mem->userspace_addr;
 	vcpu->arch.sie_block->gmslm =
@@ -78,7 +82,7 @@ static inline void kvm_s390_vcpu_set_mem
 		(mem->npages << PAGE_SHIFT) +
 		VIRTIODESCSPACE - 1ul;
 
-	up_read(&vcpu->kvm->slots_lock);
+	srcu_read_unlock(&vcpu->kvm->srcu, idx);
 }
 
 /* implemented in priv.c */

[patch 11/11] KVM: convert slots_lock to a mutex

[Marcelo Tosatti]

[-- Attachment #1: convert-slotslock-to-mutex --]
[-- Type: text/plain, Size: 9777 bytes --]

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

Index: kvm/arch/x86/kvm/vmx.c
===================================================================
--- kvm.orig/arch/x86/kvm/vmx.c
+++ kvm/arch/x86/kvm/vmx.c
@@ -2227,7 +2227,7 @@ static int alloc_apic_access_page(struct
 	struct kvm_userspace_memory_region kvm_userspace_mem;
 	int r = 0;
 
-	down_write(&kvm->slots_lock);
+	mutex_lock(&kvm->slots_lock);
 	if (kvm->arch.apic_access_page)
 		goto out;
 	kvm_userspace_mem.slot = APIC_ACCESS_PAGE_PRIVATE_MEMSLOT;
@@ -2240,7 +2240,7 @@ static int alloc_apic_access_page(struct
 
 	kvm->arch.apic_access_page = gfn_to_page(kvm, 0xfee00);
 out:
-	up_write(&kvm->slots_lock);
+	mutex_unlock(&kvm->slots_lock);
 	return r;
 }
 
@@ -2249,7 +2249,7 @@ static int alloc_identity_pagetable(stru
 	struct kvm_userspace_memory_region kvm_userspace_mem;
 	int r = 0;
 
-	down_write(&kvm->slots_lock);
+	mutex_lock(&kvm->slots_lock);
 	if (kvm->arch.ept_identity_pagetable)
 		goto out;
 	kvm_userspace_mem.slot = IDENTITY_PAGETABLE_PRIVATE_MEMSLOT;
@@ -2264,7 +2264,7 @@ static int alloc_identity_pagetable(stru
 	kvm->arch.ept_identity_pagetable = gfn_to_page(kvm,
 			kvm->arch.ept_identity_map_addr >> PAGE_SHIFT);
 out:
-	up_write(&kvm->slots_lock);
+	mutex_unlock(&kvm->slots_lock);
 	return r;
 }
 
Index: kvm/arch/x86/kvm/x86.c
===================================================================
--- kvm.orig/arch/x86/kvm/x86.c
+++ kvm/arch/x86/kvm/x86.c
@@ -2209,14 +2209,14 @@ static int kvm_vm_ioctl_set_nr_mmu_pages
 	if (kvm_nr_mmu_pages < KVM_MIN_ALLOC_MMU_PAGES)
 		return -EINVAL;
 
-	down_write(&kvm->slots_lock);
+	mutex_lock(&kvm->slots_lock);
 	spin_lock(&kvm->mmu_lock);
 
 	kvm_mmu_change_mmu_pages(kvm, kvm_nr_mmu_pages);
 	kvm->arch.n_requested_mmu_pages = kvm_nr_mmu_pages;
 
 	spin_unlock(&kvm->mmu_lock);
-	up_write(&kvm->slots_lock);
+	mutex_unlock(&kvm->slots_lock);
 	return 0;
 }
 
@@ -2293,7 +2293,7 @@ static int kvm_vm_ioctl_set_memory_alias
 	if (!aliases)
 		goto out;
 
-	down_write(&kvm->slots_lock);
+	mutex_lock(&kvm->slots_lock);
 
 	/* invalidate any gfn reference in case of deletion/shrinking */
 	memcpy(aliases, kvm->arch.aliases, sizeof(struct kvm_mem_aliases));
@@ -2329,7 +2329,7 @@ static int kvm_vm_ioctl_set_memory_alias
 	r = 0;
 
 out_unlock:
-	up_write(&kvm->slots_lock);
+	mutex_unlock(&kvm->slots_lock);
 out:
 	return r;
 }
@@ -2463,7 +2463,7 @@ int kvm_vm_ioctl_get_dirty_log(struct kv
 	unsigned long is_dirty = 0;
 	unsigned long *dirty_bitmap = NULL;
 
-	down_write(&kvm->slots_lock);
+	mutex_lock(&kvm->slots_lock);
 
 	r = -EINVAL;
 	if (log->slot >= KVM_MEMORY_SLOTS)
@@ -2513,7 +2513,7 @@ int kvm_vm_ioctl_get_dirty_log(struct kv
 out_free:
 	vfree(dirty_bitmap);
 out:
-	up_write(&kvm->slots_lock);
+	mutex_unlock(&kvm->slots_lock);
 	return r;
 }
 
@@ -2626,7 +2626,7 @@ long kvm_arch_vm_ioctl(struct file *filp
 				   sizeof(struct kvm_pit_config)))
 			goto out;
 	create_pit:
-		down_write(&kvm->slots_lock);
+		mutex_lock(&kvm->slots_lock);
 		r = -EEXIST;
 		if (kvm->arch.vpit)
 			goto create_pit_unlock;
@@ -2635,7 +2635,7 @@ long kvm_arch_vm_ioctl(struct file *filp
 		if (kvm->arch.vpit)
 			r = 0;
 	create_pit_unlock:
-		up_write(&kvm->slots_lock);
+		mutex_unlock(&kvm->slots_lock);
 		break;
 	case KVM_IRQ_LINE_STATUS:
 	case KVM_IRQ_LINE: {
Index: kvm/include/linux/kvm_host.h
===================================================================
--- kvm.orig/include/linux/kvm_host.h
+++ kvm/include/linux/kvm_host.h
@@ -161,7 +161,7 @@ struct kvm_memslots {
 struct kvm {
 	spinlock_t mmu_lock;
 	spinlock_t requests_lock;
-	struct rw_semaphore slots_lock;
+	struct mutex slots_lock;
 	struct mm_struct *mm; /* userspace tied to this vm */
 	struct kvm_memslots *memslots;
 	struct srcu_struct srcu;
Index: kvm/virt/kvm/coalesced_mmio.c
===================================================================
--- kvm.orig/virt/kvm/coalesced_mmio.c
+++ kvm/virt/kvm/coalesced_mmio.c
@@ -110,9 +110,9 @@ int kvm_coalesced_mmio_init(struct kvm *
 	dev->kvm = kvm;
 	kvm->coalesced_mmio_dev = dev;
 
-	down_write(&kvm->slots_lock);
+	mutex_lock(&kvm->slots_lock);
 	ret = kvm_io_bus_register_dev(kvm, KVM_MMIO_BUS, &dev->dev);
-	up_write(&kvm->slots_lock);
+	mutex_unlock(&kvm->slots_lock);
 	if (ret < 0)
 		goto out_free_dev;
 
@@ -140,16 +140,16 @@ int kvm_vm_ioctl_register_coalesced_mmio
 	if (dev == NULL)
 		return -EINVAL;
 
-	down_write(&kvm->slots_lock);
+	mutex_lock(&kvm->slots_lock);
 	if (dev->nb_zones >= KVM_COALESCED_MMIO_ZONE_MAX) {
-		up_write(&kvm->slots_lock);
+		mutex_unlock(&kvm->slots_lock);
 		return -ENOBUFS;
 	}
 
 	dev->zone[dev->nb_zones] = *zone;
 	dev->nb_zones++;
 
-	up_write(&kvm->slots_lock);
+	mutex_unlock(&kvm->slots_lock);
 	return 0;
 }
 
@@ -163,7 +163,7 @@ int kvm_vm_ioctl_unregister_coalesced_mm
 	if (dev == NULL)
 		return -EINVAL;
 
-	down_write(&kvm->slots_lock);
+	mutex_lock(&kvm->slots_lock);
 
 	i = dev->nb_zones;
 	while(i) {
@@ -181,7 +181,7 @@ int kvm_vm_ioctl_unregister_coalesced_mm
 		i--;
 	}
 
-	up_write(&kvm->slots_lock);
+	mutex_unlock(&kvm->slots_lock);
 
 	return 0;
 }
Index: kvm/virt/kvm/eventfd.c
===================================================================
--- kvm.orig/virt/kvm/eventfd.c
+++ kvm/virt/kvm/eventfd.c
@@ -496,7 +496,7 @@ kvm_assign_ioeventfd(struct kvm *kvm, st
 	else
 		p->wildcard = true;
 
-	down_write(&kvm->slots_lock);
+	mutex_lock(&kvm->slots_lock);
 
 	/* Verify that there isnt a match already */
 	if (ioeventfd_check_collision(kvm, p)) {
@@ -512,12 +512,12 @@ kvm_assign_ioeventfd(struct kvm *kvm, st
 
 	list_add_tail(&p->list, &kvm->ioeventfds);
 
-	up_write(&kvm->slots_lock);
+	mutex_unlock(&kvm->slots_lock);
 
 	return 0;
 
 unlock_fail:
-	up_write(&kvm->slots_lock);
+	mutex_unlock(&kvm->slots_lock);
 
 fail:
 	kfree(p);
@@ -539,7 +539,7 @@ kvm_deassign_ioeventfd(struct kvm *kvm, 
 	if (IS_ERR(eventfd))
 		return PTR_ERR(eventfd);
 
-	down_write(&kvm->slots_lock);
+	mutex_lock(&kvm->slots_lock);
 
 	list_for_each_entry_safe(p, tmp, &kvm->ioeventfds, list) {
 		bool wildcard = !(args->flags & KVM_IOEVENTFD_FLAG_DATAMATCH);
@@ -559,7 +559,7 @@ kvm_deassign_ioeventfd(struct kvm *kvm, 
 		break;
 	}
 
-	up_write(&kvm->slots_lock);
+	mutex_unlock(&kvm->slots_lock);
 
 	eventfd_ctx_put(eventfd);
 
Index: kvm/virt/kvm/ioapic.c
===================================================================
--- kvm.orig/virt/kvm/ioapic.c
+++ kvm/virt/kvm/ioapic.c
@@ -372,9 +372,9 @@ int kvm_ioapic_init(struct kvm *kvm)
 	kvm_ioapic_reset(ioapic);
 	kvm_iodevice_init(&ioapic->dev, &ioapic_mmio_ops);
 	ioapic->kvm = kvm;
-	down_write(&kvm->slots_lock);
+	mutex_lock(&kvm->slots_lock);
 	ret = kvm_io_bus_register_dev(kvm, KVM_MMIO_BUS, &ioapic->dev);
-	up_write(&kvm->slots_lock);
+	mutex_unlock(&kvm->slots_lock);
 	if (ret < 0)
 		kfree(ioapic);
 
Index: kvm/virt/kvm/kvm_main.c
===================================================================
--- kvm.orig/virt/kvm/kvm_main.c
+++ kvm/virt/kvm/kvm_main.c
@@ -420,7 +420,7 @@ static struct kvm *kvm_create_vm(void)
 	kvm_eventfd_init(kvm);
 	mutex_init(&kvm->lock);
 	mutex_init(&kvm->irq_lock);
-	init_rwsem(&kvm->slots_lock);
+	mutex_init(&kvm->slots_lock);
 	atomic_set(&kvm->users_count, 1);
 	spin_lock(&kvm_lock);
 	list_add(&kvm->vm_list, &vm_list);
@@ -751,9 +751,9 @@ int kvm_set_memory_region(struct kvm *kv
 {
 	int r;
 
-	down_write(&kvm->slots_lock);
+	mutex_lock(&kvm->slots_lock);
 	r = __kvm_set_memory_region(kvm, mem, user_alloc);
-	up_write(&kvm->slots_lock);
+	mutex_unlock(&kvm->slots_lock);
 	return r;
 }
 EXPORT_SYMBOL_GPL(kvm_set_memory_region);
@@ -1985,7 +1985,7 @@ int kvm_io_bus_read(struct kvm *kvm, enu
 	return -EOPNOTSUPP;
 }
 
-/* Caller must have write lock on slots_lock. */
+/* Caller must hold slots_lock. */
 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx,
 			    struct kvm_io_device *dev)
 {
@@ -2007,7 +2007,7 @@ int kvm_io_bus_register_dev(struct kvm *
 	return 0;
 }
 
-/* Caller must have write lock on slots_lock. */
+/* Caller must hold slots_lock. */
 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
 			      struct kvm_io_device *dev)
 {
Index: kvm/arch/ia64/kvm/kvm-ia64.c
===================================================================
--- kvm.orig/arch/ia64/kvm/kvm-ia64.c
+++ kvm/arch/ia64/kvm/kvm-ia64.c
@@ -1831,7 +1831,7 @@ int kvm_vm_ioctl_get_dirty_log(struct kv
 	struct kvm_memory_slot *memslot;
 	int is_dirty = 0;
 
-	down_write(&kvm->slots_lock);
+	mutex_lock(&kvm->slots_lock);
 	spin_lock(&kvm->arch.dirty_log_lock);
 
 	r = kvm_ia64_sync_dirty_log(kvm, log);
@@ -1851,7 +1851,7 @@ int kvm_vm_ioctl_get_dirty_log(struct kv
 	}
 	r = 0;
 out:
-	up_write(&kvm->slots_lock);
+	mutex_unlock(&kvm->slots_lock);
 	spin_unlock(&kvm->arch.dirty_log_lock);
 	return r;
 }
Index: kvm/arch/x86/kvm/i8254.c
===================================================================
--- kvm.orig/arch/x86/kvm/i8254.c
+++ kvm/arch/x86/kvm/i8254.c
@@ -602,7 +602,7 @@ static const struct kvm_io_device_ops sp
 	.write    = speaker_ioport_write,
 };
 
-/* Caller must have writers lock on slots_lock */
+/* Caller must hold slots_lock */
 struct kvm_pit *kvm_create_pit(struct kvm *kvm, u32 flags)
 {
 	struct kvm_pit *pit;
Index: kvm/arch/x86/kvm/i8259.c
===================================================================
--- kvm.orig/arch/x86/kvm/i8259.c
+++ kvm/arch/x86/kvm/i8259.c
@@ -533,9 +533,9 @@ struct kvm_pic *kvm_create_pic(struct kv
 	 * Initialize PIO device
 	 */
 	kvm_iodevice_init(&s->dev, &picdev_ops);
-	down_write(&kvm->slots_lock);
+	mutex_lock(&kvm->slots_lock);
 	ret = kvm_io_bus_register_dev(kvm, KVM_PIO_BUS, &s->dev);
-	up_write(&kvm->slots_lock);
+	mutex_unlock(&kvm->slots_lock);
 	if (ret < 0) {
 		kfree(s);
 		return NULL;

Re: [patch 00/11] convert slotslock to SRCU

[Avi Kivity]

On 12/23/2009 01:38 PM, Marcelo Tosatti wrote:
> Now that synchronize_srcu_expedited is in the tree, we can continue the
> convertion of slots_lock to SRCU.
>
>    

Okay, looks pretty good.  The only issue I see is where to put 
srcu_read_lock()s:

- outer scope (more reuse for archs, less places to forget)
- inner scope (tighter code)

There is now a mix of the two.

> Results:
> up:
>
> vmcall 6.1% improvement
>
> smp 2:
>
>
> vmcall 9.3% improvement
>    

Pretty impressive, all this from removing an uncontended semaphore.

-- 
error compiling committee.c: too many arguments to function

[patch 00/11] convert slotslock to SRCU v2

[Marcelo Tosatti]

Addressing comments.

[patch 01/11] KVM: modify memslots layout in struct kvm

[Marcelo Tosatti]

[-- Attachment #1: modify-memslot-layout --]
[-- Type: text/plain, Size: 8417 bytes --]

Have a pointer to an allocated region inside struct kvm.

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

Index: kvm/arch/x86/kvm/mmu.c
===================================================================
--- kvm.orig/arch/x86/kvm/mmu.c
+++ kvm/arch/x86/kvm/mmu.c
@@ -809,13 +809,14 @@ static int kvm_handle_hva(struct kvm *kv
 {
 	int i, j;
 	int retval = 0;
+	struct kvm_memslots *slots = kvm->memslots;
 
 	/*
 	 * If mmap_sem isn't taken, we can look the memslots with only
 	 * the mmu_lock by skipping over the slots with userspace_addr == 0.
 	 */
-	for (i = 0; i < kvm->nmemslots; i++) {
-		struct kvm_memory_slot *memslot = &kvm->memslots[i];
+	for (i = 0; i < slots->nmemslots; i++) {
+		struct kvm_memory_slot *memslot = &slots->memslots[i];
 		unsigned long start = memslot->userspace_addr;
 		unsigned long end;
 
@@ -3023,8 +3024,8 @@ unsigned int kvm_mmu_calculate_mmu_pages
 	unsigned int nr_mmu_pages;
 	unsigned int  nr_pages = 0;
 
-	for (i = 0; i < kvm->nmemslots; i++)
-		nr_pages += kvm->memslots[i].npages;
+	for (i = 0; i < kvm->memslots->nmemslots; i++)
+		nr_pages += kvm->memslots->memslots[i].npages;
 
 	nr_mmu_pages = nr_pages * KVM_PERMILLE_MMU_PAGES / 1000;
 	nr_mmu_pages = max(nr_mmu_pages,
@@ -3297,7 +3298,7 @@ static int count_rmaps(struct kvm_vcpu *
 	int i, j, k;
 
 	for (i = 0; i < KVM_MEMORY_SLOTS; ++i) {
-		struct kvm_memory_slot *m = &vcpu->kvm->memslots[i];
+		struct kvm_memory_slot *m = &vcpu->kvm->memslots->memslots[i];
 		struct kvm_rmap_desc *d;
 
 		for (j = 0; j < m->npages; ++j) {
Index: kvm/arch/x86/kvm/vmx.c
===================================================================
--- kvm.orig/arch/x86/kvm/vmx.c
+++ kvm/arch/x86/kvm/vmx.c
@@ -1503,8 +1503,8 @@ static void enter_pmode(struct kvm_vcpu 
 static gva_t rmode_tss_base(struct kvm *kvm)
 {
 	if (!kvm->arch.tss_addr) {
-		gfn_t base_gfn = kvm->memslots[0].base_gfn +
-				 kvm->memslots[0].npages - 3;
+		gfn_t base_gfn = kvm->memslots->memslots[0].base_gfn +
+				 kvm->memslots->memslots[0].npages - 3;
 		return base_gfn << PAGE_SHIFT;
 	}
 	return kvm->arch.tss_addr;
Index: kvm/arch/x86/kvm/x86.c
===================================================================
--- kvm.orig/arch/x86/kvm/x86.c
+++ kvm/arch/x86/kvm/x86.c
@@ -2428,7 +2428,7 @@ int kvm_vm_ioctl_get_dirty_log(struct kv
 		spin_lock(&kvm->mmu_lock);
 		kvm_mmu_slot_remove_write_access(kvm, log->slot);
 		spin_unlock(&kvm->mmu_lock);
-		memslot = &kvm->memslots[log->slot];
+		memslot = &kvm->memslots->memslots[log->slot];
 		n = ALIGN(memslot->npages, BITS_PER_LONG) / 8;
 		memset(memslot->dirty_bitmap, 0, n);
 	}
@@ -5222,7 +5222,7 @@ int kvm_arch_set_memory_region(struct kv
 				int user_alloc)
 {
 	int npages = mem->memory_size >> PAGE_SHIFT;
-	struct kvm_memory_slot *memslot = &kvm->memslots[mem->slot];
+	struct kvm_memory_slot *memslot = &kvm->memslots->memslots[mem->slot];
 
 	/*To keep backward compatibility with older userspace,
 	 *x86 needs to hanlde !user_alloc case.
Index: kvm/include/linux/kvm_host.h
===================================================================
--- kvm.orig/include/linux/kvm_host.h
+++ kvm/include/linux/kvm_host.h
@@ -150,14 +150,18 @@ struct kvm_irq_routing_table {};
 
 #endif
 
+struct kvm_memslots {
+	int nmemslots;
+	struct kvm_memory_slot memslots[KVM_MEMORY_SLOTS +
+					KVM_PRIVATE_MEM_SLOTS];
+};
+
 struct kvm {
 	spinlock_t mmu_lock;
 	spinlock_t requests_lock;
 	struct rw_semaphore slots_lock;
 	struct mm_struct *mm; /* userspace tied to this vm */
-	int nmemslots;
-	struct kvm_memory_slot memslots[KVM_MEMORY_SLOTS +
-					KVM_PRIVATE_MEM_SLOTS];
+	struct kvm_memslots *memslots;
 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
 	u32 bsp_vcpu_id;
 	struct kvm_vcpu *bsp_vcpu;
@@ -482,7 +486,7 @@ static inline void kvm_guest_exit(void)
 
 static inline int memslot_id(struct kvm *kvm, struct kvm_memory_slot *slot)
 {
-	return slot - kvm->memslots;
+	return slot - kvm->memslots->memslots;
 }
 
 static inline gpa_t gfn_to_gpa(gfn_t gfn)
Index: kvm/virt/kvm/iommu.c
===================================================================
--- kvm.orig/virt/kvm/iommu.c
+++ kvm/virt/kvm/iommu.c
@@ -76,10 +76,13 @@ unmap_pages:
 static int kvm_iommu_map_memslots(struct kvm *kvm)
 {
 	int i, r = 0;
+	struct kvm_memslots *slots;
 
-	for (i = 0; i < kvm->nmemslots; i++) {
-		r = kvm_iommu_map_pages(kvm, kvm->memslots[i].base_gfn,
-					kvm->memslots[i].npages);
+	slots = kvm->memslots;
+
+	for (i = 0; i < slots->nmemslots; i++) {
+		r = kvm_iommu_map_pages(kvm, slots->memslots[i].base_gfn,
+					slots->memslots[i].npages);
 		if (r)
 			break;
 	}
@@ -210,10 +213,13 @@ static void kvm_iommu_put_pages(struct k
 static int kvm_iommu_unmap_memslots(struct kvm *kvm)
 {
 	int i;
+	struct kvm_memslots *slots;
+
+	slots = kvm->memslots;
 
-	for (i = 0; i < kvm->nmemslots; i++) {
-		kvm_iommu_put_pages(kvm, kvm->memslots[i].base_gfn,
-				    kvm->memslots[i].npages);
+	for (i = 0; i < slots->nmemslots; i++) {
+		kvm_iommu_put_pages(kvm, slots->memslots[i].base_gfn,
+				    slots->memslots[i].npages);
 	}
 
 	return 0;
Index: kvm/virt/kvm/kvm_main.c
===================================================================
--- kvm.orig/virt/kvm/kvm_main.c
+++ kvm/virt/kvm/kvm_main.c
@@ -375,12 +375,16 @@ static struct kvm *kvm_create_vm(void)
 	INIT_HLIST_HEAD(&kvm->irq_ack_notifier_list);
 #endif
 
+	r = -ENOMEM;
+	kvm->memslots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL);
+	if (!kvm->memslots)
+		goto out_err;
+
 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
 	page = alloc_page(GFP_KERNEL | __GFP_ZERO);
-	if (!page) {
-		r = -ENOMEM;
+	if (!page)
 		goto out_err;
-	}
+
 	kvm->coalesced_mmio_ring =
 			(struct kvm_coalesced_mmio_ring *)page_address(page);
 #endif
@@ -416,6 +420,7 @@ out:
 out_err:
 	hardware_disable_all();
 out_err_nodisable:
+	kfree(kvm->memslots);
 	kfree(kvm);
 	return ERR_PTR(r);
 }
@@ -450,9 +455,12 @@ static void kvm_free_physmem_slot(struct
 void kvm_free_physmem(struct kvm *kvm)
 {
 	int i;
+	struct kvm_memslots *slots = kvm->memslots;
+
+	for (i = 0; i < slots->nmemslots; ++i)
+		kvm_free_physmem_slot(&slots->memslots[i], NULL);
 
-	for (i = 0; i < kvm->nmemslots; ++i)
-		kvm_free_physmem_slot(&kvm->memslots[i], NULL);
+	kfree(kvm->memslots);
 }
 
 static void kvm_destroy_vm(struct kvm *kvm)
@@ -533,7 +541,7 @@ int __kvm_set_memory_region(struct kvm *
 	if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr)
 		goto out;
 
-	memslot = &kvm->memslots[mem->slot];
+	memslot = &kvm->memslots->memslots[mem->slot];
 	base_gfn = mem->guest_phys_addr >> PAGE_SHIFT;
 	npages = mem->memory_size >> PAGE_SHIFT;
 
@@ -554,7 +562,7 @@ int __kvm_set_memory_region(struct kvm *
 	/* Check for overlaps */
 	r = -EEXIST;
 	for (i = 0; i < KVM_MEMORY_SLOTS; ++i) {
-		struct kvm_memory_slot *s = &kvm->memslots[i];
+		struct kvm_memory_slot *s = &kvm->memslots->memslots[i];
 
 		if (s == memslot || !s->npages)
 			continue;
@@ -656,8 +664,8 @@ skip_lpage:
 		kvm_arch_flush_shadow(kvm);
 
 	spin_lock(&kvm->mmu_lock);
-	if (mem->slot >= kvm->nmemslots)
-		kvm->nmemslots = mem->slot + 1;
+	if (mem->slot >= kvm->memslots->nmemslots)
+		kvm->memslots->nmemslots = mem->slot + 1;
 
 	*memslot = new;
 	spin_unlock(&kvm->mmu_lock);
@@ -727,7 +735,7 @@ int kvm_get_dirty_log(struct kvm *kvm,
 	if (log->slot >= KVM_MEMORY_SLOTS)
 		goto out;
 
-	memslot = &kvm->memslots[log->slot];
+	memslot = &kvm->memslots->memslots[log->slot];
 	r = -ENOENT;
 	if (!memslot->dirty_bitmap)
 		goto out;
@@ -781,9 +789,10 @@ EXPORT_SYMBOL_GPL(kvm_is_error_hva);
 struct kvm_memory_slot *gfn_to_memslot_unaliased(struct kvm *kvm, gfn_t gfn)
 {
 	int i;
+	struct kvm_memslots *slots = kvm->memslots;
 
-	for (i = 0; i < kvm->nmemslots; ++i) {
-		struct kvm_memory_slot *memslot = &kvm->memslots[i];
+	for (i = 0; i < slots->nmemslots; ++i) {
+		struct kvm_memory_slot *memslot = &slots->memslots[i];
 
 		if (gfn >= memslot->base_gfn
 		    && gfn < memslot->base_gfn + memslot->npages)
@@ -802,10 +811,11 @@ struct kvm_memory_slot *gfn_to_memslot(s
 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn)
 {
 	int i;
+	struct kvm_memslots *slots = kvm->memslots;
 
 	gfn = unalias_gfn(kvm, gfn);
 	for (i = 0; i < KVM_MEMORY_SLOTS; ++i) {
-		struct kvm_memory_slot *memslot = &kvm->memslots[i];
+		struct kvm_memory_slot *memslot = &slots->memslots[i];
 
 		if (gfn >= memslot->base_gfn
 		    && gfn < memslot->base_gfn + memslot->npages)

[patch 02/11] KVM: modify alias layout in x86s struct kvm_arch

[Marcelo Tosatti]

[-- Attachment #1: modify-alias-layout --]
[-- Type: text/plain, Size: 2666 bytes --]

Have a pointer to an allocated region inside x86's kvm_arch.

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

Index: kvm/arch/x86/include/asm/kvm_host.h
===================================================================
--- kvm.orig/arch/x86/include/asm/kvm_host.h
+++ kvm/arch/x86/include/asm/kvm_host.h
@@ -370,9 +370,13 @@ struct kvm_mem_alias {
 	gfn_t target_gfn;
 };
 
-struct kvm_arch{
-	int naliases;
+struct kvm_mem_aliases {
 	struct kvm_mem_alias aliases[KVM_ALIAS_SLOTS];
+	int naliases;
+};
+
+struct kvm_arch {
+	struct kvm_mem_aliases *aliases;
 
 	unsigned int n_free_mmu_pages;
 	unsigned int n_requested_mmu_pages;
Index: kvm/arch/x86/kvm/x86.c
===================================================================
--- kvm.orig/arch/x86/kvm/x86.c
+++ kvm/arch/x86/kvm/x86.c
@@ -2228,9 +2228,10 @@ gfn_t unalias_gfn(struct kvm *kvm, gfn_t
 {
 	int i;
 	struct kvm_mem_alias *alias;
+	struct kvm_mem_aliases *aliases = kvm->arch.aliases;
 
-	for (i = 0; i < kvm->arch.naliases; ++i) {
-		alias = &kvm->arch.aliases[i];
+	for (i = 0; i < aliases->naliases; ++i) {
+		alias = &aliases->aliases[i];
 		if (gfn >= alias->base_gfn
 		    && gfn < alias->base_gfn + alias->npages)
 			return alias->target_gfn + gfn - alias->base_gfn;
@@ -2248,6 +2249,7 @@ static int kvm_vm_ioctl_set_memory_alias
 {
 	int r, n;
 	struct kvm_mem_alias *p;
+	struct kvm_mem_aliases *aliases;
 
 	r = -EINVAL;
 	/* General sanity checks */
@@ -2267,15 +2269,17 @@ static int kvm_vm_ioctl_set_memory_alias
 	down_write(&kvm->slots_lock);
 	spin_lock(&kvm->mmu_lock);
 
-	p = &kvm->arch.aliases[alias->slot];
+	aliases = kvm->arch.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;
 
 	for (n = KVM_ALIAS_SLOTS; n > 0; --n)
-		if (kvm->arch.aliases[n - 1].npages)
+		if (aliases->aliases[n - 1].npages)
 			break;
-	kvm->arch.naliases = n;
+	aliases->naliases = n;
 
 	spin_unlock(&kvm->mmu_lock);
 	kvm_mmu_zap_all(kvm);
@@ -5157,6 +5161,12 @@ struct  kvm *kvm_arch_create_vm(void)
 	if (!kvm)
 		return ERR_PTR(-ENOMEM);
 
+	kvm->arch.aliases = kzalloc(sizeof(struct kvm_mem_aliases), GFP_KERNEL);
+	if (!kvm->arch.aliases) {
+		kfree(kvm);
+		return ERR_PTR(-ENOMEM);
+	}
+
 	INIT_LIST_HEAD(&kvm->arch.active_mmu_pages);
 	INIT_LIST_HEAD(&kvm->arch.assigned_dev_head);
 
@@ -5213,6 +5223,7 @@ void kvm_arch_destroy_vm(struct kvm *kvm
 		put_page(kvm->arch.apic_access_page);
 	if (kvm->arch.ept_identity_pagetable)
 		put_page(kvm->arch.ept_identity_pagetable);
+	kfree(kvm->arch.aliases);
 	kfree(kvm);
 }
 

[patch 03/11] KVM: split kvm_arch_set_memory_region into prepare and commit

[Marcelo Tosatti]

[-- Attachment #1: prepare-commit-memory-region --]
[-- Type: text/plain, Size: 6354 bytes --]

Required for SRCU convertion later.

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

Index: kvm/arch/ia64/kvm/kvm-ia64.c
===================================================================
--- kvm.orig/arch/ia64/kvm/kvm-ia64.c
+++ kvm/arch/ia64/kvm/kvm-ia64.c
@@ -1576,15 +1576,14 @@ out:
 	return r;
 }
 
-int kvm_arch_set_memory_region(struct kvm *kvm,
-		struct kvm_userspace_memory_region *mem,
+int kvm_arch_prepare_memory_region(struct kvm *kvm,
+		struct kvm_memory_slot *memslot,
 		struct kvm_memory_slot old,
 		int user_alloc)
 {
 	unsigned long i;
 	unsigned long pfn;
-	int npages = mem->memory_size >> PAGE_SHIFT;
-	struct kvm_memory_slot *memslot = &kvm->memslots[mem->slot];
+	int npages = memslot->npages;
 	unsigned long base_gfn = memslot->base_gfn;
 
 	if (base_gfn + npages > (KVM_MAX_MEM_SIZE >> PAGE_SHIFT))
@@ -1608,6 +1607,14 @@ int kvm_arch_set_memory_region(struct kv
 	return 0;
 }
 
+void kvm_arch_commit_memory_region(struct kvm *kvm,
+		struct kvm_userspace_memory_region *mem,
+		struct kvm_memory_slot old,
+		int user_alloc)
+{
+	return;
+}
+
 void kvm_arch_flush_shadow(struct kvm *kvm)
 {
 	kvm_flush_remote_tlbs(kvm);
Index: kvm/arch/s390/kvm/kvm-s390.c
===================================================================
--- kvm.orig/arch/s390/kvm/kvm-s390.c
+++ kvm/arch/s390/kvm/kvm-s390.c
@@ -689,10 +689,10 @@ long kvm_arch_vcpu_ioctl(struct file *fi
 }
 
 /* Section: memory related */
-int kvm_arch_set_memory_region(struct kvm *kvm,
-				struct kvm_userspace_memory_region *mem,
-				struct kvm_memory_slot old,
-				int user_alloc)
+int kvm_arch_prepare_memory_region(struct kvm *kvm,
+				   struct kvm_userspace_memory_region *mem,
+				   struct kvm_memory_slot old,
+				   int user_alloc)
 {
 	int i;
 	struct kvm_vcpu *vcpu;
@@ -719,14 +719,20 @@ int kvm_arch_set_memory_region(struct kv
 	if (!user_alloc)
 		return -EINVAL;
 
+	return 0;
+}
+
+int kvm_arch_commit_memory_region(struct kvm *kvm,
+				struct kvm_userspace_memory_region *mem,
+				struct kvm_memory_slot old,
+				int user_alloc)
+{
 	/* request update of sie control block for all available vcpus */
 	kvm_for_each_vcpu(i, vcpu, kvm) {
 		if (test_and_set_bit(KVM_REQ_MMU_RELOAD, &vcpu->requests))
 			continue;
 		kvm_s390_inject_sigp_stop(vcpu, ACTION_RELOADVCPU_ON_STOP);
 	}
-
-	return 0;
 }
 
 void kvm_arch_flush_shadow(struct kvm *kvm)
Index: kvm/arch/x86/kvm/x86.c
===================================================================
--- kvm.orig/arch/x86/kvm/x86.c
+++ kvm/arch/x86/kvm/x86.c
@@ -5227,13 +5227,12 @@ void kvm_arch_destroy_vm(struct kvm *kvm
 	kfree(kvm);
 }
 
-int kvm_arch_set_memory_region(struct kvm *kvm,
-				struct kvm_userspace_memory_region *mem,
+int kvm_arch_prepare_memory_region(struct kvm *kvm,
+				struct kvm_memory_slot *memslot,
 				struct kvm_memory_slot old,
 				int user_alloc)
 {
-	int npages = mem->memory_size >> PAGE_SHIFT;
-	struct kvm_memory_slot *memslot = &kvm->memslots->memslots[mem->slot];
+	int npages = memslot->npages;
 
 	/*To keep backward compatibility with older userspace,
 	 *x86 needs to hanlde !user_alloc case.
@@ -5253,26 +5252,35 @@ int kvm_arch_set_memory_region(struct kv
 			if (IS_ERR((void *)userspace_addr))
 				return PTR_ERR((void *)userspace_addr);
 
-			/* set userspace_addr atomically for kvm_hva_to_rmapp */
-			spin_lock(&kvm->mmu_lock);
 			memslot->userspace_addr = userspace_addr;
-			spin_unlock(&kvm->mmu_lock);
-		} else {
-			if (!old.user_alloc && old.rmap) {
-				int ret;
-
-				down_write(&current->mm->mmap_sem);
-				ret = do_munmap(current->mm, old.userspace_addr,
-						old.npages * PAGE_SIZE);
-				up_write(&current->mm->mmap_sem);
-				if (ret < 0)
-					printk(KERN_WARNING
-				       "kvm_vm_ioctl_set_memory_region: "
-				       "failed to munmap memory\n");
-			}
 		}
 	}
 
+
+	return 0;
+}
+
+void kvm_arch_commit_memory_region(struct kvm *kvm,
+				struct kvm_userspace_memory_region *mem,
+				struct kvm_memory_slot old,
+				int user_alloc)
+{
+
+	int npages = mem->memory_size >> PAGE_SHIFT;
+
+	if (!user_alloc && !old.user_alloc && old.rmap && !npages) {
+		int ret;
+
+		down_write(&current->mm->mmap_sem);
+		ret = do_munmap(current->mm, old.userspace_addr,
+				old.npages * PAGE_SIZE);
+		up_write(&current->mm->mmap_sem);
+		if (ret < 0)
+			printk(KERN_WARNING
+			       "kvm_vm_ioctl_set_memory_region: "
+			       "failed to munmap memory\n");
+	}
+
 	spin_lock(&kvm->mmu_lock);
 	if (!kvm->arch.n_requested_mmu_pages) {
 		unsigned int nr_mmu_pages = kvm_mmu_calculate_mmu_pages(kvm);
@@ -5281,8 +5289,6 @@ int kvm_arch_set_memory_region(struct kv
 
 	kvm_mmu_slot_remove_write_access(kvm, mem->slot);
 	spin_unlock(&kvm->mmu_lock);
-
-	return 0;
 }
 
 void kvm_arch_flush_shadow(struct kvm *kvm)
Index: kvm/include/linux/kvm_host.h
===================================================================
--- kvm.orig/include/linux/kvm_host.h
+++ kvm/include/linux/kvm_host.h
@@ -253,7 +253,11 @@ int kvm_set_memory_region(struct kvm *kv
 int __kvm_set_memory_region(struct kvm *kvm,
 			    struct kvm_userspace_memory_region *mem,
 			    int user_alloc);
-int kvm_arch_set_memory_region(struct kvm *kvm,
+int kvm_arch_prepare_memory_region(struct kvm *kvm,
+				struct kvm_memory_slot *memslot,
+				struct kvm_memory_slot old,
+				int user_alloc);
+void kvm_arch_commit_memory_region(struct kvm *kvm,
 				struct kvm_userspace_memory_region *mem,
 				struct kvm_memory_slot old,
 				int user_alloc);
Index: kvm/virt/kvm/kvm_main.c
===================================================================
--- kvm.orig/virt/kvm/kvm_main.c
+++ kvm/virt/kvm/kvm_main.c
@@ -663,6 +663,10 @@ skip_lpage:
 	if (!npages)
 		kvm_arch_flush_shadow(kvm);
 
+	r = kvm_arch_prepare_memory_region(kvm, &new, old, user_alloc);
+	if (r)
+		goto out_free;
+
 	spin_lock(&kvm->mmu_lock);
 	if (mem->slot >= kvm->memslots->nmemslots)
 		kvm->memslots->nmemslots = mem->slot + 1;
@@ -670,13 +674,7 @@ skip_lpage:
 	*memslot = new;
 	spin_unlock(&kvm->mmu_lock);
 
-	r = kvm_arch_set_memory_region(kvm, mem, old, user_alloc);
-	if (r) {
-		spin_lock(&kvm->mmu_lock);
-		*memslot = old;
-		spin_unlock(&kvm->mmu_lock);
-		goto out_free;
-	}
+	kvm_arch_commit_memory_region(kvm, mem, old, user_alloc);
 
 	kvm_free_physmem_slot(&old, npages ? &new : NULL);
 	/* Slot deletion case: we have to update the current slot */

[patch 04/11] KVM: introduce gfn_to_pfn_memslot

[Marcelo Tosatti]

[-- Attachment #1: introduce-gfn_to_pfn_memslot --]
[-- Type: text/plain, Size: 2126 bytes --]

Which takes a memslot pointer instead of using kvm->memslots.

To be used by SRCU convertion later.

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

Index: kvm/include/linux/kvm_host.h
===================================================================
--- kvm.orig/include/linux/kvm_host.h
+++ kvm/include/linux/kvm_host.h
@@ -272,6 +272,8 @@ void kvm_set_page_dirty(struct page *pag
 void kvm_set_page_accessed(struct page *page);
 
 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
+pfn_t gfn_to_pfn_memslot(struct kvm *kvm,
+			 struct kvm_memory_slot *slot, gfn_t gfn);
 void kvm_release_pfn_dirty(pfn_t);
 void kvm_release_pfn_clean(pfn_t pfn);
 void kvm_set_pfn_dirty(pfn_t pfn);
Index: kvm/virt/kvm/kvm_main.c
===================================================================
--- kvm.orig/virt/kvm/kvm_main.c
+++ kvm/virt/kvm/kvm_main.c
@@ -835,21 +835,14 @@ unsigned long gfn_to_hva(struct kvm *kvm
 }
 EXPORT_SYMBOL_GPL(gfn_to_hva);
 
-pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn)
+static pfn_t hva_to_pfn(struct kvm *kvm, unsigned long addr)
 {
 	struct page *page[1];
-	unsigned long addr;
 	int npages;
 	pfn_t pfn;
 
 	might_sleep();
 
-	addr = gfn_to_hva(kvm, gfn);
-	if (kvm_is_error_hva(addr)) {
-		get_page(bad_page);
-		return page_to_pfn(bad_page);
-	}
-
 	npages = get_user_pages_fast(addr, 1, 1, page);
 
 	if (unlikely(npages != 1)) {
@@ -874,8 +867,32 @@ pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t 
 	return pfn;
 }
 
+pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn)
+{
+	unsigned long addr;
+
+	addr = gfn_to_hva(kvm, gfn);
+	if (kvm_is_error_hva(addr)) {
+		get_page(bad_page);
+		return page_to_pfn(bad_page);
+	}
+
+	return hva_to_pfn(kvm, addr);
+}
 EXPORT_SYMBOL_GPL(gfn_to_pfn);
 
+static unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn)
+{
+	return (slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE);
+}
+
+pfn_t gfn_to_pfn_memslot(struct kvm *kvm,
+			 struct kvm_memory_slot *slot, gfn_t gfn)
+{
+	unsigned long addr = gfn_to_hva_memslot(slot, gfn);
+	return hva_to_pfn(kvm, addr);
+}
+
 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn)
 {
 	pfn_t pfn;

[patch 05/11] KVM: use gfn_to_pfn_memslot in kvm_iommu_map_pages

[Marcelo Tosatti]

[-- Attachment #1: pass-memslot-to-kvm_iommu_map_pages --]
[-- Type: text/plain, Size: 2550 bytes --]

So its possible to iommu map a memslot before making it visible to 
kvm.

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

Index: kvm/include/linux/kvm_host.h
===================================================================
--- kvm.orig/include/linux/kvm_host.h
+++ kvm/include/linux/kvm_host.h
@@ -439,8 +439,7 @@ void kvm_free_irq_source_id(struct kvm *
 #define KVM_IOMMU_CACHE_COHERENCY	0x1
 
 #ifdef CONFIG_IOMMU_API
-int kvm_iommu_map_pages(struct kvm *kvm, gfn_t base_gfn,
-			unsigned long npages);
+int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot);
 int kvm_iommu_map_guest(struct kvm *kvm);
 int kvm_iommu_unmap_guest(struct kvm *kvm);
 int kvm_assign_device(struct kvm *kvm,
Index: kvm/virt/kvm/iommu.c
===================================================================
--- kvm.orig/virt/kvm/iommu.c
+++ kvm/virt/kvm/iommu.c
@@ -32,10 +32,10 @@ static int kvm_iommu_unmap_memslots(stru
 static void kvm_iommu_put_pages(struct kvm *kvm,
 				gfn_t base_gfn, unsigned long npages);
 
-int kvm_iommu_map_pages(struct kvm *kvm,
-			gfn_t base_gfn, unsigned long npages)
+int kvm_iommu_map_pages(struct kvm *kvm, struct kvm_memory_slot *slot)
 {
-	gfn_t gfn = base_gfn;
+	gfn_t gfn = slot->base_gfn;
+	unsigned long npages = slot->npages;
 	pfn_t pfn;
 	int i, r = 0;
 	struct iommu_domain *domain = kvm->arch.iommu_domain;
@@ -54,7 +54,7 @@ int kvm_iommu_map_pages(struct kvm *kvm,
 		if (iommu_iova_to_phys(domain, gfn_to_gpa(gfn)))
 			continue;
 
-		pfn = gfn_to_pfn(kvm, gfn);
+		pfn = gfn_to_pfn_memslot(kvm, slot, gfn);
 		r = iommu_map_range(domain,
 				    gfn_to_gpa(gfn),
 				    pfn_to_hpa(pfn),
@@ -69,7 +69,7 @@ int kvm_iommu_map_pages(struct kvm *kvm,
 	return 0;
 
 unmap_pages:
-	kvm_iommu_put_pages(kvm, base_gfn, i);
+	kvm_iommu_put_pages(kvm, slot->base_gfn, i);
 	return r;
 }
 
@@ -81,8 +81,7 @@ static int kvm_iommu_map_memslots(struct
 	slots = kvm->memslots;
 
 	for (i = 0; i < slots->nmemslots; i++) {
-		r = kvm_iommu_map_pages(kvm, slots->memslots[i].base_gfn,
-					slots->memslots[i].npages);
+		r = kvm_iommu_map_pages(kvm, &slots->memslots[i]);
 		if (r)
 			break;
 	}
Index: kvm/virt/kvm/kvm_main.c
===================================================================
--- kvm.orig/virt/kvm/kvm_main.c
+++ kvm/virt/kvm/kvm_main.c
@@ -684,7 +684,7 @@ skip_lpage:
 	spin_unlock(&kvm->mmu_lock);
 #ifdef CONFIG_DMAR
 	/* map the pages in iommu page table */
-	r = kvm_iommu_map_pages(kvm, base_gfn, npages);
+	r = kvm_iommu_map_pages(kvm, memslot);
 	if (r)
 		goto out;
 #endif

[patch 06/11] KVM: introduce kvm->srcu and convert kvm_set_memory_region to SRCU update

[Marcelo Tosatti]

[-- Attachment #1: introduce-srcu-and-use-for-slots --]
[-- Type: text/plain, Size: 15992 bytes --]

Use two steps for memslot deletion: mark the slot invalid (which stops 
instantiation of new shadow pages for that slot, but allows destruction),
then instantiate the new empty slot.

Also simplifies kvm_handle_hva locking.

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

Index: kvm/include/linux/kvm.h
===================================================================
--- kvm.orig/include/linux/kvm.h
+++ kvm/include/linux/kvm.h
@@ -103,7 +103,7 @@ struct kvm_userspace_memory_region {
 
 /* for kvm_memory_region::flags */
 #define KVM_MEM_LOG_DIRTY_PAGES  1UL
-
+#define KVM_MEMSLOT_INVALID      (1UL << 1)
 
 /* for KVM_IRQ_LINE */
 struct kvm_irq_level {
Index: kvm/virt/kvm/kvm_main.c
===================================================================
--- kvm.orig/virt/kvm/kvm_main.c
+++ kvm/virt/kvm/kvm_main.c
@@ -44,6 +44,7 @@
 #include <linux/bitops.h>
 #include <linux/spinlock.h>
 #include <linux/compat.h>
+#include <linux/srcu.h>
 
 #include <asm/processor.h>
 #include <asm/io.h>
@@ -213,7 +214,7 @@ static void kvm_mmu_notifier_invalidate_
 					     unsigned long address)
 {
 	struct kvm *kvm = mmu_notifier_to_kvm(mn);
-	int need_tlb_flush;
+	int need_tlb_flush, idx;
 
 	/*
 	 * When ->invalidate_page runs, the linux pte has been zapped
@@ -233,10 +234,12 @@ static void kvm_mmu_notifier_invalidate_
 	 * pte after kvm_unmap_hva returned, without noticing the page
 	 * is going to be freed.
 	 */
+	idx = srcu_read_lock(&kvm->srcu);
 	spin_lock(&kvm->mmu_lock);
 	kvm->mmu_notifier_seq++;
 	need_tlb_flush = kvm_unmap_hva(kvm, address);
 	spin_unlock(&kvm->mmu_lock);
+	srcu_read_unlock(&kvm->srcu, idx);
 
 	/* we've to flush the tlb before the pages can be freed */
 	if (need_tlb_flush)
@@ -250,11 +253,14 @@ static void kvm_mmu_notifier_change_pte(
 					pte_t pte)
 {
 	struct kvm *kvm = mmu_notifier_to_kvm(mn);
+	int idx;
 
+	idx = srcu_read_lock(&kvm->srcu);
 	spin_lock(&kvm->mmu_lock);
 	kvm->mmu_notifier_seq++;
 	kvm_set_spte_hva(kvm, address, pte);
 	spin_unlock(&kvm->mmu_lock);
+	srcu_read_unlock(&kvm->srcu, idx);
 }
 
 static void kvm_mmu_notifier_invalidate_range_start(struct mmu_notifier *mn,
@@ -263,8 +269,9 @@ static void kvm_mmu_notifier_invalidate_
 						    unsigned long end)
 {
 	struct kvm *kvm = mmu_notifier_to_kvm(mn);
-	int need_tlb_flush = 0;
+	int need_tlb_flush = 0, idx;
 
+	idx = srcu_read_lock(&kvm->srcu);
 	spin_lock(&kvm->mmu_lock);
 	/*
 	 * The count increase must become visible at unlock time as no
@@ -275,6 +282,7 @@ static void kvm_mmu_notifier_invalidate_
 	for (; start < end; start += PAGE_SIZE)
 		need_tlb_flush |= kvm_unmap_hva(kvm, start);
 	spin_unlock(&kvm->mmu_lock);
+	srcu_read_unlock(&kvm->srcu, idx);
 
 	/* we've to flush the tlb before the pages can be freed */
 	if (need_tlb_flush)
@@ -312,11 +320,13 @@ static int kvm_mmu_notifier_clear_flush_
 					      unsigned long address)
 {
 	struct kvm *kvm = mmu_notifier_to_kvm(mn);
-	int young;
+	int young, idx;
 
+	idx = srcu_read_lock(&kvm->srcu);
 	spin_lock(&kvm->mmu_lock);
 	young = kvm_age_hva(kvm, address);
 	spin_unlock(&kvm->mmu_lock);
+	srcu_read_unlock(&kvm->srcu, idx);
 
 	if (young)
 		kvm_flush_remote_tlbs(kvm);
@@ -379,11 +389,15 @@ static struct kvm *kvm_create_vm(void)
 	kvm->memslots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL);
 	if (!kvm->memslots)
 		goto out_err;
+	if (init_srcu_struct(&kvm->srcu))
+		goto out_err;
 
 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
 	page = alloc_page(GFP_KERNEL | __GFP_ZERO);
-	if (!page)
+	if (!page) {
+		cleanup_srcu_struct(&kvm->srcu);
 		goto out_err;
+	}
 
 	kvm->coalesced_mmio_ring =
 			(struct kvm_coalesced_mmio_ring *)page_address(page);
@@ -391,6 +405,7 @@ static struct kvm *kvm_create_vm(void)
 
 	r = kvm_init_mmu_notifier(kvm);
 	if (r) {
+		cleanup_srcu_struct(&kvm->srcu);
 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
 		put_page(page);
 #endif
@@ -481,6 +496,7 @@ static void kvm_destroy_vm(struct kvm *k
 	kvm_arch_flush_shadow(kvm);
 #endif
 	kvm_arch_destroy_vm(kvm);
+	cleanup_srcu_struct(&kvm->srcu);
 	hardware_disable_all();
 	mmdrop(mm);
 }
@@ -521,12 +537,13 @@ int __kvm_set_memory_region(struct kvm *
 			    struct kvm_userspace_memory_region *mem,
 			    int user_alloc)
 {
-	int r;
+	int r, flush_shadow = 0;
 	gfn_t base_gfn;
 	unsigned long npages;
 	unsigned long i;
 	struct kvm_memory_slot *memslot;
 	struct kvm_memory_slot old, new;
+	struct kvm_memslots *slots, *old_memslots;
 
 	r = -EINVAL;
 	/* General sanity checks */
@@ -588,15 +605,7 @@ int __kvm_set_memory_region(struct kvm *
 		memset(new.rmap, 0, npages * sizeof(*new.rmap));
 
 		new.user_alloc = user_alloc;
-		/*
-		 * hva_to_rmmap() serialzies with the mmu_lock and to be
-		 * safe it has to ignore memslots with !user_alloc &&
-		 * !userspace_addr.
-		 */
-		if (user_alloc)
-			new.userspace_addr = mem->userspace_addr;
-		else
-			new.userspace_addr = 0;
+		new.userspace_addr = mem->userspace_addr;
 	}
 	if (!npages)
 		goto skip_lpage;
@@ -651,8 +660,9 @@ skip_lpage:
 		if (!new.dirty_bitmap)
 			goto out_free;
 		memset(new.dirty_bitmap, 0, dirty_bytes);
+		/* destroy any largepage mappings for dirty tracking */
 		if (old.npages)
-			kvm_arch_flush_shadow(kvm);
+			flush_shadow = 1;
 	}
 #else  /* not defined CONFIG_S390 */
 	new.user_alloc = user_alloc;
@@ -660,34 +670,72 @@ skip_lpage:
 		new.userspace_addr = mem->userspace_addr;
 #endif /* not defined CONFIG_S390 */
 
-	if (!npages)
+	if (!npages) {
+		r = -ENOMEM;
+		slots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL);
+		if (!slots)
+			goto out_free;
+		memcpy(slots, kvm->memslots, sizeof(struct kvm_memslots));
+		if (mem->slot >= slots->nmemslots)
+			slots->nmemslots = mem->slot + 1;
+		slots->memslots[mem->slot].flags |= KVM_MEMSLOT_INVALID;
+
+		old_memslots = kvm->memslots;
+		rcu_assign_pointer(kvm->memslots, slots);
+		synchronize_srcu_expedited(&kvm->srcu);
+		/* From this point no new shadow pages pointing to a deleted
+		 * memslot will be created.
+         	 *
+         	 * validation of sp->gfn happens in:
+         	 * 	- gfn_to_hva (kvm_read_guest, gfn_to_pfn)
+         	 * 	- kvm_is_visible_gfn (mmu_check_roots)
+         	 */
 		kvm_arch_flush_shadow(kvm);
+		kfree(old_memslots);
+	}
 
 	r = kvm_arch_prepare_memory_region(kvm, &new, old, user_alloc);
 	if (r)
 		goto out_free;
 
-	spin_lock(&kvm->mmu_lock);
-	if (mem->slot >= kvm->memslots->nmemslots)
-		kvm->memslots->nmemslots = mem->slot + 1;
+#ifdef CONFIG_DMAR
+	/* map the pages in iommu page table */
+	if (npages) {
+		r = kvm_iommu_map_pages(kvm, &new);
+		if (r)
+			goto out_free;
+	}
+#endif
 
-	*memslot = new;
-	spin_unlock(&kvm->mmu_lock);
+	r = -ENOMEM;
+	slots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL);
+	if (!slots)
+		goto out_free;
+	memcpy(slots, kvm->memslots, sizeof(struct kvm_memslots));
+	if (mem->slot >= slots->nmemslots)
+		slots->nmemslots = mem->slot + 1;
+
+	/* actual memory is freed via old in kvm_free_physmem_slot below */
+	if (!npages) {
+		new.rmap = NULL;
+		new.dirty_bitmap = NULL;
+		for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i)
+			new.lpage_info[i] = NULL;
+	}
+
+	slots->memslots[mem->slot] = new;
+	old_memslots = kvm->memslots;
+	rcu_assign_pointer(kvm->memslots, slots);
+	synchronize_srcu_expedited(&kvm->srcu);
 
 	kvm_arch_commit_memory_region(kvm, mem, old, user_alloc);
 
-	kvm_free_physmem_slot(&old, npages ? &new : NULL);
-	/* Slot deletion case: we have to update the current slot */
-	spin_lock(&kvm->mmu_lock);
-	if (!npages)
-		*memslot = old;
-	spin_unlock(&kvm->mmu_lock);
-#ifdef CONFIG_DMAR
-	/* map the pages in iommu page table */
-	r = kvm_iommu_map_pages(kvm, memslot);
-	if (r)
-		goto out;
-#endif
+	kvm_free_physmem_slot(&old, &new);
+	kfree(old_memslots);
+
+	if (flush_shadow)
+		kvm_arch_flush_shadow(kvm);
+
 	return 0;
 
 out_free:
@@ -787,7 +835,7 @@ EXPORT_SYMBOL_GPL(kvm_is_error_hva);
 struct kvm_memory_slot *gfn_to_memslot_unaliased(struct kvm *kvm, gfn_t gfn)
 {
 	int i;
-	struct kvm_memslots *slots = kvm->memslots;
+	struct kvm_memslots *slots = rcu_dereference(kvm->memslots);
 
 	for (i = 0; i < slots->nmemslots; ++i) {
 		struct kvm_memory_slot *memslot = &slots->memslots[i];
@@ -809,12 +857,15 @@ struct kvm_memory_slot *gfn_to_memslot(s
 int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn)
 {
 	int i;
-	struct kvm_memslots *slots = kvm->memslots;
+	struct kvm_memslots *slots = rcu_dereference(kvm->memslots);
 
 	gfn = unalias_gfn(kvm, gfn);
 	for (i = 0; i < KVM_MEMORY_SLOTS; ++i) {
 		struct kvm_memory_slot *memslot = &slots->memslots[i];
 
+		if (memslot->flags & KVM_MEMSLOT_INVALID)
+			continue;
+
 		if (gfn >= memslot->base_gfn
 		    && gfn < memslot->base_gfn + memslot->npages)
 			return 1;
@@ -823,13 +874,31 @@ int kvm_is_visible_gfn(struct kvm *kvm, 
 }
 EXPORT_SYMBOL_GPL(kvm_is_visible_gfn);
 
+int memslot_id(struct kvm *kvm, gfn_t gfn)
+{
+	int i;
+	struct kvm_memslots *slots = rcu_dereference(kvm->memslots);
+	struct kvm_memory_slot *memslot = NULL;
+
+	gfn = unalias_gfn(kvm, gfn);
+	for (i = 0; i < slots->nmemslots; ++i) {
+		memslot = &slots->memslots[i];
+
+		if (gfn >= memslot->base_gfn
+		    && gfn < memslot->base_gfn + memslot->npages)
+			break;
+	}
+
+	return memslot - slots->memslots;
+}
+
 unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn)
 {
 	struct kvm_memory_slot *slot;
 
 	gfn = unalias_gfn(kvm, gfn);
 	slot = gfn_to_memslot_unaliased(kvm, gfn);
-	if (!slot)
+	if (!slot || slot->flags & KVM_MEMSLOT_INVALID)
 		return bad_hva();
 	return (slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE);
 }
Index: kvm/arch/ia64/kvm/kvm-ia64.c
===================================================================
--- kvm.orig/arch/ia64/kvm/kvm-ia64.c
+++ kvm/arch/ia64/kvm/kvm-ia64.c
@@ -1834,6 +1834,7 @@ int kvm_vm_ioctl_get_dirty_log(struct kv
 	struct kvm_memory_slot *memslot;
 	int is_dirty = 0;
 
+	down_write(&kvm->slots_lock);
 	spin_lock(&kvm->arch.dirty_log_lock);
 
 	r = kvm_ia64_sync_dirty_log(kvm, log);
@@ -1853,6 +1854,7 @@ int kvm_vm_ioctl_get_dirty_log(struct kv
 	}
 	r = 0;
 out:
+	up_write(&kvm->slots_lock);
 	spin_unlock(&kvm->arch.dirty_log_lock);
 	return r;
 }
Index: kvm/arch/x86/kvm/mmu.c
===================================================================
--- kvm.orig/arch/x86/kvm/mmu.c
+++ kvm/arch/x86/kvm/mmu.c
@@ -29,6 +29,7 @@
 #include <linux/swap.h>
 #include <linux/hugetlb.h>
 #include <linux/compiler.h>
+#include <linux/srcu.h>
 
 #include <asm/page.h>
 #include <asm/cmpxchg.h>
@@ -809,21 +810,15 @@ static int kvm_handle_hva(struct kvm *kv
 {
 	int i, j;
 	int retval = 0;
-	struct kvm_memslots *slots = kvm->memslots;
+	struct kvm_memslots *slots;
+
+	slots = rcu_dereference(kvm->memslots);
 
-	/*
-	 * If mmap_sem isn't taken, we can look the memslots with only
-	 * the mmu_lock by skipping over the slots with userspace_addr == 0.
-	 */
 	for (i = 0; i < slots->nmemslots; i++) {
 		struct kvm_memory_slot *memslot = &slots->memslots[i];
 		unsigned long start = memslot->userspace_addr;
 		unsigned long end;
 
-		/* mmu_lock protects userspace_addr */
-		if (!start)
-			continue;
-
 		end = start + (memslot->npages << PAGE_SHIFT);
 		if (hva >= start && hva < end) {
 			gfn_t gfn_offset = (hva - start) >> PAGE_SHIFT;
@@ -1619,7 +1614,7 @@ static void mmu_unshadow(struct kvm *kvm
 
 static void page_header_update_slot(struct kvm *kvm, void *pte, gfn_t gfn)
 {
-	int slot = memslot_id(kvm, gfn_to_memslot(kvm, gfn));
+	int slot = memslot_id(kvm, gfn);
 	struct kvm_mmu_page *sp = page_header(__pa(pte));
 
 	__set_bit(slot, sp->slot_bitmap);
@@ -3023,9 +3018,11 @@ unsigned int kvm_mmu_calculate_mmu_pages
 	int i;
 	unsigned int nr_mmu_pages;
 	unsigned int  nr_pages = 0;
+	struct kvm_memslots *slots;
 
-	for (i = 0; i < kvm->memslots->nmemslots; i++)
-		nr_pages += kvm->memslots->memslots[i].npages;
+	slots = rcu_dereference(kvm->memslots);
+	for (i = 0; i < slots->nmemslots; i++)
+		nr_pages += slots->memslots[i].npages;
 
 	nr_mmu_pages = nr_pages * KVM_PERMILLE_MMU_PAGES / 1000;
 	nr_mmu_pages = max(nr_mmu_pages,
@@ -3295,10 +3292,12 @@ static void audit_mappings(struct kvm_vc
 static int count_rmaps(struct kvm_vcpu *vcpu)
 {
 	int nmaps = 0;
-	int i, j, k;
+	int i, j, k, idx;
 
+	idx = srcu_read_lock(&kvm->srcu);
+	slots = rcu_dereference(kvm->memslots);
 	for (i = 0; i < KVM_MEMORY_SLOTS; ++i) {
-		struct kvm_memory_slot *m = &vcpu->kvm->memslots->memslots[i];
+		struct kvm_memory_slot *m = &slots->memslots[i];
 		struct kvm_rmap_desc *d;
 
 		for (j = 0; j < m->npages; ++j) {
@@ -3321,6 +3320,7 @@ static int count_rmaps(struct kvm_vcpu *
 			}
 		}
 	}
+	srcu_read_unlock(&kvm->srcu, idx);
 	return nmaps;
 }
 
Index: kvm/arch/x86/kvm/vmx.c
===================================================================
--- kvm.orig/arch/x86/kvm/vmx.c
+++ kvm/arch/x86/kvm/vmx.c
@@ -1503,7 +1503,11 @@ static void enter_pmode(struct kvm_vcpu 
 static gva_t rmode_tss_base(struct kvm *kvm)
 {
 	if (!kvm->arch.tss_addr) {
-		gfn_t base_gfn = kvm->memslots->memslots[0].base_gfn +
+		struct kvm_memslots *slots;
+		gfn_t base_gfn;
+
+		slots = rcu_dereference(kvm->memslots);
+		base_gfn = kvm->memslots->memslots[0].base_gfn +
 				 kvm->memslots->memslots[0].npages - 3;
 		return base_gfn << PAGE_SHIFT;
 	}
Index: kvm/include/linux/kvm_host.h
===================================================================
--- kvm.orig/include/linux/kvm_host.h
+++ kvm/include/linux/kvm_host.h
@@ -162,6 +162,7 @@ struct kvm {
 	struct rw_semaphore slots_lock;
 	struct mm_struct *mm; /* userspace tied to this vm */
 	struct kvm_memslots *memslots;
+	struct srcu_struct srcu;
 #ifdef CONFIG_KVM_APIC_ARCHITECTURE
 	u32 bsp_vcpu_id;
 	struct kvm_vcpu *bsp_vcpu;
@@ -274,6 +275,7 @@ void kvm_set_page_accessed(struct page *
 pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn);
 pfn_t gfn_to_pfn_memslot(struct kvm *kvm,
 			 struct kvm_memory_slot *slot, gfn_t gfn);
+int memslot_id(struct kvm *kvm, gfn_t gfn);
 void kvm_release_pfn_dirty(pfn_t);
 void kvm_release_pfn_clean(pfn_t pfn);
 void kvm_set_pfn_dirty(pfn_t pfn);
@@ -489,11 +491,6 @@ static inline void kvm_guest_exit(void)
 	current->flags &= ~PF_VCPU;
 }
 
-static inline int memslot_id(struct kvm *kvm, struct kvm_memory_slot *slot)
-{
-	return slot - kvm->memslots->memslots;
-}
-
 static inline gpa_t gfn_to_gpa(gfn_t gfn)
 {
 	return (gpa_t)gfn << PAGE_SHIFT;
Index: kvm/virt/kvm/assigned-dev.c
===================================================================
--- kvm.orig/virt/kvm/assigned-dev.c
+++ kvm/virt/kvm/assigned-dev.c
@@ -504,12 +504,12 @@ out:
 static int kvm_vm_ioctl_assign_device(struct kvm *kvm,
 				      struct kvm_assigned_pci_dev *assigned_dev)
 {
-	int r = 0;
+	int r = 0, idx;
 	struct kvm_assigned_dev_kernel *match;
 	struct pci_dev *dev;
 
 	mutex_lock(&kvm->lock);
-	down_read(&kvm->slots_lock);
+	idx = srcu_read_lock(&kvm->srcu);
 
 	match = kvm_find_assigned_dev(&kvm->arch.assigned_dev_head,
 				      assigned_dev->assigned_dev_id);
@@ -573,7 +573,7 @@ static int kvm_vm_ioctl_assign_device(st
 	}
 
 out:
-	up_read(&kvm->slots_lock);
+	srcu_read_unlock(&kvm->srcu, idx);
 	mutex_unlock(&kvm->lock);
 	return r;
 out_list_del:
@@ -585,7 +585,7 @@ out_put:
 	pci_dev_put(dev);
 out_free:
 	kfree(match);
-	up_read(&kvm->slots_lock);
+	srcu_read_unlock(&kvm->srcu, idx);
 	mutex_unlock(&kvm->lock);
 	return r;
 }
Index: kvm/virt/kvm/iommu.c
===================================================================
--- kvm.orig/virt/kvm/iommu.c
+++ kvm/virt/kvm/iommu.c
@@ -78,7 +78,7 @@ static int kvm_iommu_map_memslots(struct
 	int i, r = 0;
 	struct kvm_memslots *slots;
 
-	slots = kvm->memslots;
+	slots = rcu_dereference(kvm->memslots);
 
 	for (i = 0; i < slots->nmemslots; i++) {
 		r = kvm_iommu_map_pages(kvm, &slots->memslots[i]);
@@ -214,7 +214,7 @@ static int kvm_iommu_unmap_memslots(stru
 	int i;
 	struct kvm_memslots *slots;
 
-	slots = kvm->memslots;
+	slots = rcu_dereference(kvm->memslots);
 
 	for (i = 0; i < slots->nmemslots; i++) {
 		kvm_iommu_put_pages(kvm, slots->memslots[i].base_gfn,

[patch 07/11] KVM: use SRCU for dirty log

[Marcelo Tosatti]

[-- Attachment #1: use-srcu-for-dirty-log --]
[-- Type: text/plain, Size: 2025 bytes --]

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

Index: kvm/arch/x86/kvm/x86.c
===================================================================
--- kvm.orig/arch/x86/kvm/x86.c
+++ kvm/arch/x86/kvm/x86.c
@@ -2416,27 +2416,60 @@ static int kvm_vm_ioctl_reinject(struct 
 int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
 				      struct kvm_dirty_log *log)
 {
-	int r;
-	int n;
+	int r, n, i;
 	struct kvm_memory_slot *memslot;
-	int is_dirty = 0;
+	unsigned long is_dirty = 0;
+	unsigned long *dirty_bitmap = NULL;
 
 	down_write(&kvm->slots_lock);
 
-	r = kvm_get_dirty_log(kvm, log, &is_dirty);
-	if (r)
+	r = -EINVAL;
+	if (log->slot >= KVM_MEMORY_SLOTS)
+		goto out;
+
+	memslot = &kvm->memslots->memslots[log->slot];
+	r = -ENOENT;
+	if (!memslot->dirty_bitmap)
+		goto out;
+
+	n = ALIGN(memslot->npages, BITS_PER_LONG) / 8;
+
+	r = -ENOMEM;
+	dirty_bitmap = vmalloc(n);
+	if (!dirty_bitmap)
 		goto out;
+	memset(dirty_bitmap, 0, n);
+
+	for (i = 0; !is_dirty && i < n/sizeof(long); i++)
+		is_dirty = memslot->dirty_bitmap[i];
 
 	/* If nothing is dirty, don't bother messing with page tables. */
 	if (is_dirty) {
+		struct kvm_memslots *slots, *old_slots;
+
 		spin_lock(&kvm->mmu_lock);
 		kvm_mmu_slot_remove_write_access(kvm, log->slot);
 		spin_unlock(&kvm->mmu_lock);
-		memslot = &kvm->memslots->memslots[log->slot];
-		n = ALIGN(memslot->npages, BITS_PER_LONG) / 8;
-		memset(memslot->dirty_bitmap, 0, n);
+
+		slots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL);
+		if (!slots)
+			goto out_free;
+
+		memcpy(slots, kvm->memslots, sizeof(struct kvm_memslots));
+		slots->memslots[log->slot].dirty_bitmap = dirty_bitmap;
+
+		old_slots = kvm->memslots;
+		rcu_assign_pointer(kvm->memslots, slots);
+		synchronize_srcu_expedited(&kvm->srcu);
+		dirty_bitmap = old_slots->memslots[log->slot].dirty_bitmap;
+		kfree(old_slots);
 	}
+
 	r = 0;
+	if (copy_to_user(log->dirty_bitmap, dirty_bitmap, n))
+		r = -EFAULT;
+out_free:
+	vfree(dirty_bitmap);
 out:
 	up_write(&kvm->slots_lock);
 	return r;

[patch 08/11] KVM: x86: switch kvm_set_memory_alias to SRCU update

[Marcelo Tosatti]

[-- Attachment #1: srcu-alias-update --]
[-- Type: text/plain, Size: 5255 bytes --]

Using a similar two-step procedure as for memslots.

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

Index: kvm/arch/x86/include/asm/kvm_host.h
===================================================================
--- kvm.orig/arch/x86/include/asm/kvm_host.h
+++ kvm/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;
Index: kvm/arch/x86/kvm/x86.c
===================================================================
--- kvm.orig/arch/x86/kvm/x86.c
+++ kvm/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
@@ -2224,11 +2225,32 @@ static int kvm_vm_ioctl_get_nr_mmu_pages
 	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];
@@ -2249,7 +2271,7 @@ static int kvm_vm_ioctl_set_memory_alias
 {
 	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 */
@@ -2266,28 +2288,48 @@ static int kvm_vm_ioctl_set_memory_alias
 	    < 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;
 }
Index: kvm/include/linux/kvm_host.h
===================================================================
--- kvm.orig/include/linux/kvm_host.h
+++ kvm/include/linux/kvm_host.h
@@ -265,6 +265,8 @@ void kvm_arch_commit_memory_region(struc
 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);
@@ -538,6 +540,10 @@ static inline int mmu_notifier_retry(str
 }
 #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
Index: kvm/virt/kvm/kvm_main.c
===================================================================
--- kvm.orig/virt/kvm/kvm_main.c
+++ kvm/virt/kvm/kvm_main.c
@@ -859,7 +859,7 @@ int kvm_is_visible_gfn(struct kvm *kvm, 
 	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
 {
 	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();

[patch 09/11] KVM: convert io_bus to SRCU

[Marcelo Tosatti]

[-- Attachment #1: srcu-iobus-update --]
[-- Type: text/plain, Size: 14236 bytes --]

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

Index: kvm/arch/x86/kvm/i8254.c
===================================================================
--- kvm.orig/arch/x86/kvm/i8254.c
+++ kvm/arch/x86/kvm/i8254.c
@@ -642,13 +642,13 @@ struct kvm_pit *kvm_create_pit(struct kv
 	kvm_register_irq_mask_notifier(kvm, 0, &pit->mask_notifier);
 
 	kvm_iodevice_init(&pit->dev, &pit_dev_ops);
-	ret = __kvm_io_bus_register_dev(&kvm->pio_bus, &pit->dev);
+	ret = kvm_io_bus_register_dev(kvm, KVM_PIO_BUS, &pit->dev);
 	if (ret < 0)
 		goto fail;
 
 	if (flags & KVM_PIT_SPEAKER_DUMMY) {
 		kvm_iodevice_init(&pit->speaker_dev, &speaker_dev_ops);
-		ret = __kvm_io_bus_register_dev(&kvm->pio_bus,
+		ret = kvm_io_bus_register_dev(kvm, KVM_PIO_BUS,
 						&pit->speaker_dev);
 		if (ret < 0)
 			goto fail_unregister;
@@ -657,7 +657,7 @@ struct kvm_pit *kvm_create_pit(struct kv
 	return pit;
 
 fail_unregister:
-	__kvm_io_bus_unregister_dev(&kvm->pio_bus, &pit->dev);
+	kvm_io_bus_unregister_dev(kvm, KVM_PIO_BUS, &pit->dev);
 
 fail:
 	if (pit->irq_source_id >= 0)
Index: kvm/arch/x86/kvm/i8259.c
===================================================================
--- kvm.orig/arch/x86/kvm/i8259.c
+++ kvm/arch/x86/kvm/i8259.c
@@ -533,7 +533,9 @@ struct kvm_pic *kvm_create_pic(struct kv
 	 * Initialize PIO device
 	 */
 	kvm_iodevice_init(&s->dev, &picdev_ops);
-	ret = kvm_io_bus_register_dev(kvm, &kvm->pio_bus, &s->dev);
+	down_write(&kvm->slots_lock);
+	ret = kvm_io_bus_register_dev(kvm, KVM_PIO_BUS, &s->dev);
+	up_write(&kvm->slots_lock);
 	if (ret < 0) {
 		kfree(s);
 		return NULL;
Index: kvm/include/linux/kvm_host.h
===================================================================
--- kvm.orig/include/linux/kvm_host.h
+++ kvm/include/linux/kvm_host.h
@@ -57,20 +57,20 @@ struct kvm_io_bus {
 	struct kvm_io_device *devs[NR_IOBUS_DEVS];
 };
 
-void kvm_io_bus_init(struct kvm_io_bus *bus);
-void kvm_io_bus_destroy(struct kvm_io_bus *bus);
-int kvm_io_bus_write(struct kvm_io_bus *bus, gpa_t addr, int len,
-		     const void *val);
-int kvm_io_bus_read(struct kvm_io_bus *bus, gpa_t addr, int len,
+enum kvm_bus {
+	KVM_MMIO_BUS,
+	KVM_PIO_BUS,
+	KVM_NR_BUSES
+};
+
+int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
+		     int len, const void *val);
+int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, int len,
 		    void *val);
-int __kvm_io_bus_register_dev(struct kvm_io_bus *bus,
-			       struct kvm_io_device *dev);
-int kvm_io_bus_register_dev(struct kvm *kvm, struct kvm_io_bus *bus,
+int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx,
 			    struct kvm_io_device *dev);
-void __kvm_io_bus_unregister_dev(struct kvm_io_bus *bus,
-				 struct kvm_io_device *dev);
-void kvm_io_bus_unregister_dev(struct kvm *kvm, struct kvm_io_bus *bus,
-			       struct kvm_io_device *dev);
+int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
+			      struct kvm_io_device *dev);
 
 struct kvm_vcpu {
 	struct kvm *kvm;
@@ -171,8 +171,7 @@ struct kvm {
 	atomic_t online_vcpus;
 	struct list_head vm_list;
 	struct mutex lock;
-	struct kvm_io_bus mmio_bus;
-	struct kvm_io_bus pio_bus;
+	struct kvm_io_bus *buses[KVM_NR_BUSES];
 #ifdef CONFIG_HAVE_KVM_EVENTFD
 	struct {
 		spinlock_t        lock;
Index: kvm/virt/kvm/eventfd.c
===================================================================
--- kvm.orig/virt/kvm/eventfd.c
+++ kvm/virt/kvm/eventfd.c
@@ -451,7 +451,7 @@ static int
 kvm_assign_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
 {
 	int                       pio = args->flags & KVM_IOEVENTFD_FLAG_PIO;
-	struct kvm_io_bus        *bus = pio ? &kvm->pio_bus : &kvm->mmio_bus;
+	enum kvm_bus              bus_idx = pio ? KVM_PIO_BUS : KVM_MMIO_BUS;
 	struct _ioeventfd        *p;
 	struct eventfd_ctx       *eventfd;
 	int                       ret;
@@ -506,7 +506,7 @@ kvm_assign_ioeventfd(struct kvm *kvm, st
 
 	kvm_iodevice_init(&p->dev, &ioeventfd_ops);
 
-	ret = __kvm_io_bus_register_dev(bus, &p->dev);
+	ret = kvm_io_bus_register_dev(kvm, bus_idx, &p->dev);
 	if (ret < 0)
 		goto unlock_fail;
 
@@ -530,7 +530,7 @@ static int
 kvm_deassign_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
 {
 	int                       pio = args->flags & KVM_IOEVENTFD_FLAG_PIO;
-	struct kvm_io_bus        *bus = pio ? &kvm->pio_bus : &kvm->mmio_bus;
+	enum kvm_bus              bus_idx = pio ? KVM_PIO_BUS : KVM_MMIO_BUS;
 	struct _ioeventfd        *p, *tmp;
 	struct eventfd_ctx       *eventfd;
 	int                       ret = -ENOENT;
@@ -553,7 +553,7 @@ kvm_deassign_ioeventfd(struct kvm *kvm, 
 		if (!p->wildcard && p->datamatch != args->datamatch)
 			continue;
 
-		__kvm_io_bus_unregister_dev(bus, &p->dev);
+		kvm_io_bus_unregister_dev(kvm, bus_idx, &p->dev);
 		ioeventfd_release(p);
 		ret = 0;
 		break;
Index: kvm/virt/kvm/ioapic.c
===================================================================
--- kvm.orig/virt/kvm/ioapic.c
+++ kvm/virt/kvm/ioapic.c
@@ -372,7 +372,9 @@ int kvm_ioapic_init(struct kvm *kvm)
 	kvm_ioapic_reset(ioapic);
 	kvm_iodevice_init(&ioapic->dev, &ioapic_mmio_ops);
 	ioapic->kvm = kvm;
-	ret = kvm_io_bus_register_dev(kvm, &kvm->mmio_bus, &ioapic->dev);
+	down_write(&kvm->slots_lock);
+	ret = kvm_io_bus_register_dev(kvm, KVM_MMIO_BUS, &ioapic->dev);
+	up_write(&kvm->slots_lock);
 	if (ret < 0)
 		kfree(ioapic);
 
Index: kvm/virt/kvm/kvm_main.c
===================================================================
--- kvm.orig/virt/kvm/kvm_main.c
+++ kvm/virt/kvm/kvm_main.c
@@ -85,6 +85,8 @@ static long kvm_vcpu_ioctl(struct file *
 static int hardware_enable_all(void);
 static void hardware_disable_all(void);
 
+static void kvm_io_bus_destroy(struct kvm_io_bus *bus);
+
 static bool kvm_rebooting;
 
 static bool largepages_enabled = true;
@@ -367,7 +369,7 @@ static int kvm_init_mmu_notifier(struct 
 
 static struct kvm *kvm_create_vm(void)
 {
-	int r = 0;
+	int r = 0, i;
 	struct kvm *kvm = kvm_arch_create_vm();
 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
 	struct page *page;
@@ -391,6 +393,14 @@ static struct kvm *kvm_create_vm(void)
 		goto out_err;
 	if (init_srcu_struct(&kvm->srcu))
 		goto out_err;
+ 	for (i = 0; i < KVM_NR_BUSES; i++) {
+ 		kvm->buses[i] = kzalloc(sizeof(struct kvm_io_bus),
+ 					GFP_KERNEL);
+ 		if (!kvm->buses[i]) {
+ 			cleanup_srcu_struct(&kvm->srcu);
+ 			goto out_err;
+ 		}
+ 	}
 
 #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
 	page = alloc_page(GFP_KERNEL | __GFP_ZERO);
@@ -416,11 +426,9 @@ static struct kvm *kvm_create_vm(void)
 	atomic_inc(&kvm->mm->mm_count);
 	spin_lock_init(&kvm->mmu_lock);
 	spin_lock_init(&kvm->requests_lock);
-	kvm_io_bus_init(&kvm->pio_bus);
 	kvm_eventfd_init(kvm);
 	mutex_init(&kvm->lock);
 	mutex_init(&kvm->irq_lock);
-	kvm_io_bus_init(&kvm->mmio_bus);
 	init_rwsem(&kvm->slots_lock);
 	atomic_set(&kvm->users_count, 1);
 	spin_lock(&kvm_lock);
@@ -435,6 +443,8 @@ out:
 out_err:
 	hardware_disable_all();
 out_err_nodisable:
+ 	for (i = 0; i < KVM_NR_BUSES; i++)
+ 		kfree(kvm->buses[i]);
 	kfree(kvm->memslots);
 	kfree(kvm);
 	return ERR_PTR(r);
@@ -480,6 +490,7 @@ void kvm_free_physmem(struct kvm *kvm)
 
 static void kvm_destroy_vm(struct kvm *kvm)
 {
+	int i;
 	struct mm_struct *mm = kvm->mm;
 
 	kvm_arch_sync_events(kvm);
@@ -487,8 +498,8 @@ static void kvm_destroy_vm(struct kvm *k
 	list_del(&kvm->vm_list);
 	spin_unlock(&kvm_lock);
 	kvm_free_irq_routing(kvm);
-	kvm_io_bus_destroy(&kvm->pio_bus);
-	kvm_io_bus_destroy(&kvm->mmio_bus);
+	for (i = 0; i < KVM_NR_BUSES; i++)
+		kvm_io_bus_destroy(kvm->buses[i]);
 	kvm_coalesced_mmio_free(kvm);
 #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
 	mmu_notifier_unregister(&kvm->mmu_notifier, kvm->mm);
@@ -1949,12 +1960,7 @@ static struct notifier_block kvm_reboot_
 	.priority = 0,
 };
 
-void kvm_io_bus_init(struct kvm_io_bus *bus)
-{
-	memset(bus, 0, sizeof(*bus));
-}
-
-void kvm_io_bus_destroy(struct kvm_io_bus *bus)
+static void kvm_io_bus_destroy(struct kvm_io_bus *bus)
 {
 	int i;
 
@@ -1963,13 +1969,15 @@ void kvm_io_bus_destroy(struct kvm_io_bu
 
 		kvm_iodevice_destructor(pos);
 	}
+	kfree(bus);
 }
 
 /* kvm_io_bus_write - called under kvm->slots_lock */
-int kvm_io_bus_write(struct kvm_io_bus *bus, gpa_t addr,
+int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
 		     int len, const void *val)
 {
 	int i;
+	struct kvm_io_bus *bus = rcu_dereference(kvm->buses[bus_idx]);
 	for (i = 0; i < bus->dev_count; i++)
 		if (!kvm_iodevice_write(bus->devs[i], addr, len, val))
 			return 0;
@@ -1977,59 +1985,71 @@ int kvm_io_bus_write(struct kvm_io_bus *
 }
 
 /* kvm_io_bus_read - called under kvm->slots_lock */
-int kvm_io_bus_read(struct kvm_io_bus *bus, gpa_t addr, int len, void *val)
+int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
+		    int len, void *val)
 {
 	int i;
+	struct kvm_io_bus *bus = rcu_dereference(kvm->buses[bus_idx]);
+
 	for (i = 0; i < bus->dev_count; i++)
 		if (!kvm_iodevice_read(bus->devs[i], addr, len, val))
 			return 0;
 	return -EOPNOTSUPP;
 }
 
-int kvm_io_bus_register_dev(struct kvm *kvm, struct kvm_io_bus *bus,
-			     struct kvm_io_device *dev)
+/* Caller must have write lock on slots_lock. */
+int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx,
+			    struct kvm_io_device *dev)
 {
-	int ret;
+	struct kvm_io_bus *new_bus, *bus;
 
-	down_write(&kvm->slots_lock);
-	ret = __kvm_io_bus_register_dev(bus, dev);
-	up_write(&kvm->slots_lock);
-
-	return ret;
-}
-
-/* An unlocked version. Caller must have write lock on slots_lock. */
-int __kvm_io_bus_register_dev(struct kvm_io_bus *bus,
-			      struct kvm_io_device *dev)
-{
+	bus = kvm->buses[bus_idx];
 	if (bus->dev_count > NR_IOBUS_DEVS-1)
 		return -ENOSPC;
 
-	bus->devs[bus->dev_count++] = dev;
+	new_bus = kzalloc(sizeof(struct kvm_io_bus), GFP_KERNEL);
+	if (!new_bus)
+		return -ENOMEM;
+	memcpy(new_bus, bus, sizeof(struct kvm_io_bus));
+	new_bus->devs[new_bus->dev_count++] = dev;
+	rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
+	synchronize_srcu_expedited(&kvm->srcu);
+	kfree(bus);
 
 	return 0;
 }
 
-void kvm_io_bus_unregister_dev(struct kvm *kvm,
-			       struct kvm_io_bus *bus,
-			       struct kvm_io_device *dev)
+/* Caller must have write lock on slots_lock. */
+int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
+			      struct kvm_io_device *dev)
 {
-	down_write(&kvm->slots_lock);
-	__kvm_io_bus_unregister_dev(bus, dev);
-	up_write(&kvm->slots_lock);
-}
+	int i, r;
+	struct kvm_io_bus *new_bus, *bus;
 
-/* An unlocked version. Caller must have write lock on slots_lock. */
-void __kvm_io_bus_unregister_dev(struct kvm_io_bus *bus,
-				 struct kvm_io_device *dev)
-{
-	int i;
+	new_bus = kzalloc(sizeof(struct kvm_io_bus), GFP_KERNEL);
+	if (!new_bus)
+		return -ENOMEM;
 
-	for (i = 0; i < bus->dev_count; i++)
-		if (bus->devs[i] == dev) {
-			bus->devs[i] = bus->devs[--bus->dev_count];
+	bus = kvm->buses[bus_idx];
+	memcpy(new_bus, bus, sizeof(struct kvm_io_bus));
+
+	r = -ENOENT;
+	for (i = 0; i < new_bus->dev_count; i++)
+		if (new_bus->devs[i] == dev) {
+			r = 0;
+			new_bus->devs[i] = new_bus->devs[--new_bus->dev_count];
 			break;
 		}
+
+	if (r) {
+		kfree(new_bus);
+		return r;
+	}
+
+	rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
+	synchronize_srcu_expedited(&kvm->srcu);
+	kfree(bus);
+	return r;
 }
 
 static struct notifier_block kvm_cpu_notifier = {
Index: kvm/virt/kvm/coalesced_mmio.c
===================================================================
--- kvm.orig/virt/kvm/coalesced_mmio.c
+++ kvm/virt/kvm/coalesced_mmio.c
@@ -110,7 +110,9 @@ int kvm_coalesced_mmio_init(struct kvm *
 	dev->kvm = kvm;
 	kvm->coalesced_mmio_dev = dev;
 
-	ret = kvm_io_bus_register_dev(kvm, &kvm->mmio_bus, &dev->dev);
+	down_write(&kvm->slots_lock);
+	ret = kvm_io_bus_register_dev(kvm, KVM_MMIO_BUS, &dev->dev);
+	up_write(&kvm->slots_lock);
 	if (ret < 0)
 		goto out_free_dev;
 
Index: kvm/arch/x86/kvm/x86.c
===================================================================
--- kvm.orig/arch/x86/kvm/x86.c
+++ kvm/arch/x86/kvm/x86.c
@@ -2852,7 +2852,7 @@ static int vcpu_mmio_write(struct kvm_vc
 	    !kvm_iodevice_write(&vcpu->arch.apic->dev, addr, len, v))
 		return 0;
 
-	return kvm_io_bus_write(&vcpu->kvm->mmio_bus, addr, len, v);
+	return kvm_io_bus_write(vcpu->kvm, KVM_MMIO_BUS, addr, len, v);
 }
 
 static int vcpu_mmio_read(struct kvm_vcpu *vcpu, gpa_t addr, int len, void *v)
@@ -2861,7 +2861,7 @@ static int vcpu_mmio_read(struct kvm_vcp
 	    !kvm_iodevice_read(&vcpu->arch.apic->dev, addr, len, v))
 		return 0;
 
-	return kvm_io_bus_read(&vcpu->kvm->mmio_bus, addr, len, v);
+	return kvm_io_bus_read(vcpu->kvm, KVM_MMIO_BUS, addr, len, v);
 }
 
 static int kvm_read_guest_virt(gva_t addr, void *val, unsigned int bytes,
@@ -3346,11 +3346,12 @@ static int kernel_pio(struct kvm_vcpu *v
 	int r;
 
 	if (vcpu->arch.pio.in)
-		r = kvm_io_bus_read(&vcpu->kvm->pio_bus, vcpu->arch.pio.port,
+		r = kvm_io_bus_read(vcpu->kvm, KVM_PIO_BUS, vcpu->arch.pio.port,
 				    vcpu->arch.pio.size, pd);
 	else
-		r = kvm_io_bus_write(&vcpu->kvm->pio_bus, vcpu->arch.pio.port,
-				     vcpu->arch.pio.size, pd);
+		r = kvm_io_bus_write(vcpu->kvm, KVM_PIO_BUS,
+				     vcpu->arch.pio.port, vcpu->arch.pio.size,
+				     pd);
 	return r;
 }
 
@@ -3361,7 +3362,7 @@ static int pio_string_write(struct kvm_v
 	int i, r = 0;
 
 	for (i = 0; i < io->cur_count; i++) {
-		if (kvm_io_bus_write(&vcpu->kvm->pio_bus,
+		if (kvm_io_bus_write(vcpu->kvm, KVM_PIO_BUS,
 				     io->port, io->size, pd)) {
 			r = -EOPNOTSUPP;
 			break;
Index: kvm/arch/ia64/kvm/kvm-ia64.c
===================================================================
--- kvm.orig/arch/ia64/kvm/kvm-ia64.c
+++ kvm/arch/ia64/kvm/kvm-ia64.c
@@ -241,10 +241,10 @@ static int handle_mmio(struct kvm_vcpu *
 	return 0;
 mmio:
 	if (p->dir)
-		r = kvm_io_bus_read(&vcpu->kvm->mmio_bus, p->addr,
+		r = kvm_io_bus_read(vcpu->kvm, KVM_MMIO_BUS, p->addr,
 				    p->size, &p->data);
 	else
-		r = kvm_io_bus_write(&vcpu->kvm->mmio_bus, p->addr,
+		r = kvm_io_bus_write(vcpu->kvm, KVM_MMIO_BUS, p->addr,
 				     p->size, &p->data);
 	if (r)
 		printk(KERN_ERR"kvm: No iodevice found! addr:%lx\n", p->addr);

[patch 10/11] KVM: switch vcpu context to use SRCU

[Marcelo Tosatti]

[-- Attachment #1: switch-to-memslot-srcu --]
[-- Type: text/plain, Size: 8330 bytes --]

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

Index: kvm/arch/x86/kvm/x86.c
===================================================================
--- kvm.orig/arch/x86/kvm/x86.c
+++ kvm/arch/x86/kvm/x86.c
@@ -1307,15 +1307,15 @@ static int __msr_io(struct kvm_vcpu *vcp
 		    int (*do_msr)(struct kvm_vcpu *vcpu,
 				  unsigned index, u64 *data))
 {
-	int i;
+	int i, idx;
 
 	vcpu_load(vcpu);
 
-	down_read(&vcpu->kvm->slots_lock);
+	idx = srcu_read_lock(&vcpu->kvm->srcu);
 	for (i = 0; i < msrs->nmsrs; ++i)
 		if (do_msr(vcpu, entries[i].index, &entries[i].data))
 			break;
-	up_read(&vcpu->kvm->slots_lock);
+	srcu_read_unlock(&vcpu->kvm->srcu, idx);
 
 	vcpu_put(vcpu);
 
@@ -3901,14 +3901,15 @@ static void vapic_enter(struct kvm_vcpu 
 static void vapic_exit(struct kvm_vcpu *vcpu)
 {
 	struct kvm_lapic *apic = vcpu->arch.apic;
+	int idx;
 
 	if (!apic || !apic->vapic_addr)
 		return;
 
-	down_read(&vcpu->kvm->slots_lock);
+	idx = srcu_read_lock(&vcpu->kvm->srcu);
 	kvm_release_page_dirty(apic->vapic_page);
 	mark_page_dirty(vcpu->kvm, apic->vapic_addr >> PAGE_SHIFT);
-	up_read(&vcpu->kvm->slots_lock);
+	srcu_read_unlock(&vcpu->kvm->srcu, idx);
 }
 
 static void update_cr8_intercept(struct kvm_vcpu *vcpu)
@@ -4037,7 +4038,7 @@ static int vcpu_enter_guest(struct kvm_v
 		kvm_lapic_sync_to_vapic(vcpu);
 	}
 
-	up_read(&vcpu->kvm->slots_lock);
+	srcu_read_unlock(&vcpu->kvm->srcu, vcpu->srcu_idx);
 
 	kvm_guest_enter();
 
@@ -4079,7 +4080,7 @@ static int vcpu_enter_guest(struct kvm_v
 
 	preempt_enable();
 
-	down_read(&vcpu->kvm->slots_lock);
+	vcpu->srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
 
 	/*
 	 * Profile KVM exit RIPs:
@@ -4101,6 +4102,7 @@ out:
 static int __vcpu_run(struct kvm_vcpu *vcpu)
 {
 	int r;
+	struct kvm *kvm = vcpu->kvm;
 
 	if (unlikely(vcpu->arch.mp_state == KVM_MP_STATE_SIPI_RECEIVED)) {
 		pr_debug("vcpu %d received sipi with vector # %x\n",
@@ -4112,7 +4114,7 @@ static int __vcpu_run(struct kvm_vcpu *v
 		vcpu->arch.mp_state = KVM_MP_STATE_RUNNABLE;
 	}
 
-	down_read(&vcpu->kvm->slots_lock);
+	vcpu->srcu_idx = srcu_read_lock(&kvm->srcu);
 	vapic_enter(vcpu);
 
 	r = 1;
@@ -4120,9 +4122,9 @@ static int __vcpu_run(struct kvm_vcpu *v
 		if (vcpu->arch.mp_state == KVM_MP_STATE_RUNNABLE)
 			r = vcpu_enter_guest(vcpu);
 		else {
-			up_read(&vcpu->kvm->slots_lock);
+			srcu_read_unlock(&kvm->srcu, vcpu->srcu_idx);
 			kvm_vcpu_block(vcpu);
-			down_read(&vcpu->kvm->slots_lock);
+			vcpu->srcu_idx = srcu_read_lock(&kvm->srcu);
 			if (test_and_clear_bit(KVM_REQ_UNHALT, &vcpu->requests))
 			{
 				switch(vcpu->arch.mp_state) {
@@ -4157,13 +4159,13 @@ static int __vcpu_run(struct kvm_vcpu *v
 			++vcpu->stat.signal_exits;
 		}
 		if (need_resched()) {
-			up_read(&vcpu->kvm->slots_lock);
+			srcu_read_unlock(&kvm->srcu, vcpu->srcu_idx);
 			kvm_resched(vcpu);
-			down_read(&vcpu->kvm->slots_lock);
+			vcpu->srcu_idx = srcu_read_lock(&kvm->srcu);
 		}
 	}
 
-	up_read(&vcpu->kvm->slots_lock);
+	srcu_read_unlock(&kvm->srcu, vcpu->srcu_idx);
 	post_kvm_run_save(vcpu);
 
 	vapic_exit(vcpu);
@@ -4202,10 +4204,10 @@ int kvm_arch_vcpu_ioctl_run(struct kvm_v
 		vcpu->mmio_read_completed = 1;
 		vcpu->mmio_needed = 0;
 
-		down_read(&vcpu->kvm->slots_lock);
+		vcpu->srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
 		r = emulate_instruction(vcpu, vcpu->arch.mmio_fault_cr2, 0,
 					EMULTYPE_NO_DECODE);
-		up_read(&vcpu->kvm->slots_lock);
+		srcu_read_unlock(&vcpu->kvm->srcu, vcpu->srcu_idx);
 		if (r == EMULATE_DO_MMIO) {
 			/*
 			 * Read-modify-write.  Back to userspace.
@@ -4968,11 +4970,12 @@ int kvm_arch_vcpu_ioctl_translate(struct
 {
 	unsigned long vaddr = tr->linear_address;
 	gpa_t gpa;
+	int idx;
 
 	vcpu_load(vcpu);
-	down_read(&vcpu->kvm->slots_lock);
+	idx = srcu_read_lock(&vcpu->kvm->srcu);
 	gpa = vcpu->arch.mmu.gva_to_gpa(vcpu, vaddr);
-	up_read(&vcpu->kvm->slots_lock);
+	srcu_read_unlock(&vcpu->kvm->srcu, idx);
 	tr->physical_address = gpa;
 	tr->valid = gpa != UNMAPPED_GVA;
 	tr->writeable = 1;
@@ -5223,10 +5226,12 @@ fail:
 
 void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu)
 {
+	int idx;
+
 	kvm_free_lapic(vcpu);
-	down_read(&vcpu->kvm->slots_lock);
+	idx = srcu_read_lock(&vcpu->kvm->srcu);
 	kvm_mmu_destroy(vcpu);
-	up_read(&vcpu->kvm->slots_lock);
+	srcu_read_unlock(&vcpu->kvm->srcu, idx);
 	free_page((unsigned long)vcpu->arch.pio_data);
 }
 
Index: kvm/include/linux/kvm_host.h
===================================================================
--- kvm.orig/include/linux/kvm_host.h
+++ kvm/include/linux/kvm_host.h
@@ -83,6 +83,8 @@ struct kvm_vcpu {
 	struct kvm_run *run;
 	unsigned long requests;
 	unsigned long guest_debug;
+	int srcu_idx;
+
 	int fpu_active;
 	int guest_fpu_loaded;
 	wait_queue_head_t wq;
Index: kvm/arch/x86/kvm/mmu.c
===================================================================
--- kvm.orig/arch/x86/kvm/mmu.c
+++ kvm/arch/x86/kvm/mmu.c
@@ -2935,10 +2935,9 @@ static int mmu_shrink(int nr_to_scan, gf
 	spin_lock(&kvm_lock);
 
 	list_for_each_entry(kvm, &vm_list, vm_list) {
-		int npages;
+		int npages, idx;
 
-		if (!down_read_trylock(&kvm->slots_lock))
-			continue;
+		idx = srcu_read_lock(&kvm->srcu);
 		spin_lock(&kvm->mmu_lock);
 		npages = kvm->arch.n_alloc_mmu_pages -
 			 kvm->arch.n_free_mmu_pages;
@@ -2951,7 +2950,7 @@ static int mmu_shrink(int nr_to_scan, gf
 		nr_to_scan--;
 
 		spin_unlock(&kvm->mmu_lock);
-		up_read(&kvm->slots_lock);
+		srcu_read_unlock(&kvm->srcu, idx);
 	}
 	if (kvm_freed)
 		list_move_tail(&kvm_freed->vm_list, &vm_list);
Index: kvm/arch/x86/kvm/vmx.c
===================================================================
--- kvm.orig/arch/x86/kvm/vmx.c
+++ kvm/arch/x86/kvm/vmx.c
@@ -2477,10 +2477,10 @@ static int vmx_vcpu_reset(struct kvm_vcp
 {
 	struct vcpu_vmx *vmx = to_vmx(vcpu);
 	u64 msr;
-	int ret;
+	int ret, idx;
 
 	vcpu->arch.regs_avail = ~((1 << VCPU_REGS_RIP) | (1 << VCPU_REGS_RSP));
-	down_read(&vcpu->kvm->slots_lock);
+	idx = srcu_read_lock(&vcpu->kvm->srcu);
 	if (!init_rmode(vmx->vcpu.kvm)) {
 		ret = -ENOMEM;
 		goto out;
@@ -2588,7 +2588,7 @@ static int vmx_vcpu_reset(struct kvm_vcp
 	vmx->emulation_required = 0;
 
 out:
-	up_read(&vcpu->kvm->slots_lock);
+	srcu_read_unlock(&vcpu->kvm->srcu, idx);
 	return ret;
 }
 
Index: kvm/arch/ia64/kvm/kvm-ia64.c
===================================================================
--- kvm.orig/arch/ia64/kvm/kvm-ia64.c
+++ kvm/arch/ia64/kvm/kvm-ia64.c
@@ -636,12 +636,9 @@ static void kvm_vcpu_post_transition(str
 static int __vcpu_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
 {
 	union context *host_ctx, *guest_ctx;
-	int r;
+	int r, idx;
 
-	/*
-	 * down_read() may sleep and return with interrupts enabled
-	 */
-	down_read(&vcpu->kvm->slots_lock);
+	idx = srcu_read_lock(&vcpu->kvm->srcu);
 
 again:
 	if (signal_pending(current)) {
@@ -663,7 +660,7 @@ again:
 	if (r < 0)
 		goto vcpu_run_fail;
 
-	up_read(&vcpu->kvm->slots_lock);
+	srcu_read_unlock(&vcpu->kvm->srcu, idx);
 	kvm_guest_enter();
 
 	/*
@@ -687,7 +684,7 @@ again:
 	kvm_guest_exit();
 	preempt_enable();
 
-	down_read(&vcpu->kvm->slots_lock);
+	idx = srcu_read_lock(&vcpu->kvm->srcu);
 
 	r = kvm_handle_exit(kvm_run, vcpu);
 
@@ -697,10 +694,10 @@ again:
 	}
 
 out:
-	up_read(&vcpu->kvm->slots_lock);
+	srcu_read_unlock(&vcpu->kvm->srcu, idx);
 	if (r > 0) {
 		kvm_resched(vcpu);
-		down_read(&vcpu->kvm->slots_lock);
+		idx = srcu_read_lock(&vcpu->kvm->srcu);
 		goto again;
 	}
 
Index: kvm/arch/s390/kvm/kvm-s390.h
===================================================================
--- kvm.orig/arch/s390/kvm/kvm-s390.h
+++ kvm/arch/s390/kvm/kvm-s390.h
@@ -67,10 +67,14 @@ static inline long kvm_s390_vcpu_get_mem
 
 static inline void kvm_s390_vcpu_set_mem(struct kvm_vcpu *vcpu)
 {
+	int idx;
 	struct kvm_memory_slot *mem;
+	struct kvm_memory_slots *memslots;
 
-	down_read(&vcpu->kvm->slots_lock);
-	mem = &vcpu->kvm->memslots[0];
+	idx = srcu_read_lock(&vcpu->kvm->srcu);
+	memslots = rcu_dereference(vcpu->kvm->memslots);
+
+	mem = &memslots->memslots[0];
 
 	vcpu->arch.sie_block->gmsor = mem->userspace_addr;
 	vcpu->arch.sie_block->gmslm =
@@ -78,7 +82,7 @@ static inline void kvm_s390_vcpu_set_mem
 		(mem->npages << PAGE_SHIFT) +
 		VIRTIODESCSPACE - 1ul;
 
-	up_read(&vcpu->kvm->slots_lock);
+	srcu_read_unlock(&vcpu->kvm->srcu, idx);
 }
 
 /* implemented in priv.c */

[patch 11/11] KVM: convert slots_lock to a mutex

[Marcelo Tosatti]

[-- Attachment #1: convert-slotslock-to-mutex --]
[-- Type: text/plain, Size: 9777 bytes --]

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

Index: kvm/arch/x86/kvm/vmx.c
===================================================================
--- kvm.orig/arch/x86/kvm/vmx.c
+++ kvm/arch/x86/kvm/vmx.c
@@ -2222,7 +2222,7 @@ static int alloc_apic_access_page(struct
 	struct kvm_userspace_memory_region kvm_userspace_mem;
 	int r = 0;
 
-	down_write(&kvm->slots_lock);
+	mutex_lock(&kvm->slots_lock);
 	if (kvm->arch.apic_access_page)
 		goto out;
 	kvm_userspace_mem.slot = APIC_ACCESS_PAGE_PRIVATE_MEMSLOT;
@@ -2235,7 +2235,7 @@ static int alloc_apic_access_page(struct
 
 	kvm->arch.apic_access_page = gfn_to_page(kvm, 0xfee00);
 out:
-	up_write(&kvm->slots_lock);
+	mutex_unlock(&kvm->slots_lock);
 	return r;
 }
 
@@ -2244,7 +2244,7 @@ static int alloc_identity_pagetable(stru
 	struct kvm_userspace_memory_region kvm_userspace_mem;
 	int r = 0;
 
-	down_write(&kvm->slots_lock);
+	mutex_lock(&kvm->slots_lock);
 	if (kvm->arch.ept_identity_pagetable)
 		goto out;
 	kvm_userspace_mem.slot = IDENTITY_PAGETABLE_PRIVATE_MEMSLOT;
@@ -2259,7 +2259,7 @@ static int alloc_identity_pagetable(stru
 	kvm->arch.ept_identity_pagetable = gfn_to_page(kvm,
 			kvm->arch.ept_identity_map_addr >> PAGE_SHIFT);
 out:
-	up_write(&kvm->slots_lock);
+	mutex_unlock(&kvm->slots_lock);
 	return r;
 }
 
Index: kvm/arch/x86/kvm/x86.c
===================================================================
--- kvm.orig/arch/x86/kvm/x86.c
+++ kvm/arch/x86/kvm/x86.c
@@ -2209,14 +2209,14 @@ static int kvm_vm_ioctl_set_nr_mmu_pages
 	if (kvm_nr_mmu_pages < KVM_MIN_ALLOC_MMU_PAGES)
 		return -EINVAL;
 
-	down_write(&kvm->slots_lock);
+	mutex_lock(&kvm->slots_lock);
 	spin_lock(&kvm->mmu_lock);
 
 	kvm_mmu_change_mmu_pages(kvm, kvm_nr_mmu_pages);
 	kvm->arch.n_requested_mmu_pages = kvm_nr_mmu_pages;
 
 	spin_unlock(&kvm->mmu_lock);
-	up_write(&kvm->slots_lock);
+	mutex_unlock(&kvm->slots_lock);
 	return 0;
 }
 
@@ -2293,7 +2293,7 @@ static int kvm_vm_ioctl_set_memory_alias
 	if (!aliases)
 		goto out;
 
-	down_write(&kvm->slots_lock);
+	mutex_lock(&kvm->slots_lock);
 
 	/* invalidate any gfn reference in case of deletion/shrinking */
 	memcpy(aliases, kvm->arch.aliases, sizeof(struct kvm_mem_aliases));
@@ -2329,7 +2329,7 @@ static int kvm_vm_ioctl_set_memory_alias
 	r = 0;
 
 out_unlock:
-	up_write(&kvm->slots_lock);
+	mutex_unlock(&kvm->slots_lock);
 out:
 	return r;
 }
@@ -2463,7 +2463,7 @@ int kvm_vm_ioctl_get_dirty_log(struct kv
 	unsigned long is_dirty = 0;
 	unsigned long *dirty_bitmap = NULL;
 
-	down_write(&kvm->slots_lock);
+	mutex_lock(&kvm->slots_lock);
 
 	r = -EINVAL;
 	if (log->slot >= KVM_MEMORY_SLOTS)
@@ -2513,7 +2513,7 @@ int kvm_vm_ioctl_get_dirty_log(struct kv
 out_free:
 	vfree(dirty_bitmap);
 out:
-	up_write(&kvm->slots_lock);
+	mutex_unlock(&kvm->slots_lock);
 	return r;
 }
 
@@ -2626,7 +2626,7 @@ long kvm_arch_vm_ioctl(struct file *filp
 				   sizeof(struct kvm_pit_config)))
 			goto out;
 	create_pit:
-		down_write(&kvm->slots_lock);
+		mutex_lock(&kvm->slots_lock);
 		r = -EEXIST;
 		if (kvm->arch.vpit)
 			goto create_pit_unlock;
@@ -2635,7 +2635,7 @@ long kvm_arch_vm_ioctl(struct file *filp
 		if (kvm->arch.vpit)
 			r = 0;
 	create_pit_unlock:
-		up_write(&kvm->slots_lock);
+		mutex_unlock(&kvm->slots_lock);
 		break;
 	case KVM_IRQ_LINE_STATUS:
 	case KVM_IRQ_LINE: {
Index: kvm/include/linux/kvm_host.h
===================================================================
--- kvm.orig/include/linux/kvm_host.h
+++ kvm/include/linux/kvm_host.h
@@ -161,7 +161,7 @@ struct kvm_memslots {
 struct kvm {
 	spinlock_t mmu_lock;
 	spinlock_t requests_lock;
-	struct rw_semaphore slots_lock;
+	struct mutex slots_lock;
 	struct mm_struct *mm; /* userspace tied to this vm */
 	struct kvm_memslots *memslots;
 	struct srcu_struct srcu;
Index: kvm/virt/kvm/coalesced_mmio.c
===================================================================
--- kvm.orig/virt/kvm/coalesced_mmio.c
+++ kvm/virt/kvm/coalesced_mmio.c
@@ -110,9 +110,9 @@ int kvm_coalesced_mmio_init(struct kvm *
 	dev->kvm = kvm;
 	kvm->coalesced_mmio_dev = dev;
 
-	down_write(&kvm->slots_lock);
+	mutex_lock(&kvm->slots_lock);
 	ret = kvm_io_bus_register_dev(kvm, KVM_MMIO_BUS, &dev->dev);
-	up_write(&kvm->slots_lock);
+	mutex_unlock(&kvm->slots_lock);
 	if (ret < 0)
 		goto out_free_dev;
 
@@ -140,16 +140,16 @@ int kvm_vm_ioctl_register_coalesced_mmio
 	if (dev == NULL)
 		return -EINVAL;
 
-	down_write(&kvm->slots_lock);
+	mutex_lock(&kvm->slots_lock);
 	if (dev->nb_zones >= KVM_COALESCED_MMIO_ZONE_MAX) {
-		up_write(&kvm->slots_lock);
+		mutex_unlock(&kvm->slots_lock);
 		return -ENOBUFS;
 	}
 
 	dev->zone[dev->nb_zones] = *zone;
 	dev->nb_zones++;
 
-	up_write(&kvm->slots_lock);
+	mutex_unlock(&kvm->slots_lock);
 	return 0;
 }
 
@@ -163,7 +163,7 @@ int kvm_vm_ioctl_unregister_coalesced_mm
 	if (dev == NULL)
 		return -EINVAL;
 
-	down_write(&kvm->slots_lock);
+	mutex_lock(&kvm->slots_lock);
 
 	i = dev->nb_zones;
 	while(i) {
@@ -181,7 +181,7 @@ int kvm_vm_ioctl_unregister_coalesced_mm
 		i--;
 	}
 
-	up_write(&kvm->slots_lock);
+	mutex_unlock(&kvm->slots_lock);
 
 	return 0;
 }
Index: kvm/virt/kvm/eventfd.c
===================================================================
--- kvm.orig/virt/kvm/eventfd.c
+++ kvm/virt/kvm/eventfd.c
@@ -496,7 +496,7 @@ kvm_assign_ioeventfd(struct kvm *kvm, st
 	else
 		p->wildcard = true;
 
-	down_write(&kvm->slots_lock);
+	mutex_lock(&kvm->slots_lock);
 
 	/* Verify that there isnt a match already */
 	if (ioeventfd_check_collision(kvm, p)) {
@@ -512,12 +512,12 @@ kvm_assign_ioeventfd(struct kvm *kvm, st
 
 	list_add_tail(&p->list, &kvm->ioeventfds);
 
-	up_write(&kvm->slots_lock);
+	mutex_unlock(&kvm->slots_lock);
 
 	return 0;
 
 unlock_fail:
-	up_write(&kvm->slots_lock);
+	mutex_unlock(&kvm->slots_lock);
 
 fail:
 	kfree(p);
@@ -539,7 +539,7 @@ kvm_deassign_ioeventfd(struct kvm *kvm, 
 	if (IS_ERR(eventfd))
 		return PTR_ERR(eventfd);
 
-	down_write(&kvm->slots_lock);
+	mutex_lock(&kvm->slots_lock);
 
 	list_for_each_entry_safe(p, tmp, &kvm->ioeventfds, list) {
 		bool wildcard = !(args->flags & KVM_IOEVENTFD_FLAG_DATAMATCH);
@@ -559,7 +559,7 @@ kvm_deassign_ioeventfd(struct kvm *kvm, 
 		break;
 	}
 
-	up_write(&kvm->slots_lock);
+	mutex_unlock(&kvm->slots_lock);
 
 	eventfd_ctx_put(eventfd);
 
Index: kvm/virt/kvm/ioapic.c
===================================================================
--- kvm.orig/virt/kvm/ioapic.c
+++ kvm/virt/kvm/ioapic.c
@@ -372,9 +372,9 @@ int kvm_ioapic_init(struct kvm *kvm)
 	kvm_ioapic_reset(ioapic);
 	kvm_iodevice_init(&ioapic->dev, &ioapic_mmio_ops);
 	ioapic->kvm = kvm;
-	down_write(&kvm->slots_lock);
+	mutex_lock(&kvm->slots_lock);
 	ret = kvm_io_bus_register_dev(kvm, KVM_MMIO_BUS, &ioapic->dev);
-	up_write(&kvm->slots_lock);
+	mutex_unlock(&kvm->slots_lock);
 	if (ret < 0)
 		kfree(ioapic);
 
Index: kvm/virt/kvm/kvm_main.c
===================================================================
--- kvm.orig/virt/kvm/kvm_main.c
+++ kvm/virt/kvm/kvm_main.c
@@ -429,7 +429,7 @@ static struct kvm *kvm_create_vm(void)
 	kvm_eventfd_init(kvm);
 	mutex_init(&kvm->lock);
 	mutex_init(&kvm->irq_lock);
-	init_rwsem(&kvm->slots_lock);
+	mutex_init(&kvm->slots_lock);
 	atomic_set(&kvm->users_count, 1);
 	spin_lock(&kvm_lock);
 	list_add(&kvm->vm_list, &vm_list);
@@ -763,9 +763,9 @@ int kvm_set_memory_region(struct kvm *kv
 {
 	int r;
 
-	down_write(&kvm->slots_lock);
+	mutex_lock(&kvm->slots_lock);
 	r = __kvm_set_memory_region(kvm, mem, user_alloc);
-	up_write(&kvm->slots_lock);
+	mutex_unlock(&kvm->slots_lock);
 	return r;
 }
 EXPORT_SYMBOL_GPL(kvm_set_memory_region);
@@ -1997,7 +1997,7 @@ int kvm_io_bus_read(struct kvm *kvm, enu
 	return -EOPNOTSUPP;
 }
 
-/* Caller must have write lock on slots_lock. */
+/* Caller must hold slots_lock. */
 int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx,
 			    struct kvm_io_device *dev)
 {
@@ -2019,7 +2019,7 @@ int kvm_io_bus_register_dev(struct kvm *
 	return 0;
 }
 
-/* Caller must have write lock on slots_lock. */
+/* Caller must hold slots_lock. */
 int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
 			      struct kvm_io_device *dev)
 {
Index: kvm/arch/ia64/kvm/kvm-ia64.c
===================================================================
--- kvm.orig/arch/ia64/kvm/kvm-ia64.c
+++ kvm/arch/ia64/kvm/kvm-ia64.c
@@ -1831,7 +1831,7 @@ int kvm_vm_ioctl_get_dirty_log(struct kv
 	struct kvm_memory_slot *memslot;
 	int is_dirty = 0;
 
-	down_write(&kvm->slots_lock);
+	mutex_lock(&kvm->slots_lock);
 	spin_lock(&kvm->arch.dirty_log_lock);
 
 	r = kvm_ia64_sync_dirty_log(kvm, log);
@@ -1851,7 +1851,7 @@ int kvm_vm_ioctl_get_dirty_log(struct kv
 	}
 	r = 0;
 out:
-	up_write(&kvm->slots_lock);
+	mutex_unlock(&kvm->slots_lock);
 	spin_unlock(&kvm->arch.dirty_log_lock);
 	return r;
 }
Index: kvm/arch/x86/kvm/i8254.c
===================================================================
--- kvm.orig/arch/x86/kvm/i8254.c
+++ kvm/arch/x86/kvm/i8254.c
@@ -602,7 +602,7 @@ static const struct kvm_io_device_ops sp
 	.write    = speaker_ioport_write,
 };
 
-/* Caller must have writers lock on slots_lock */
+/* Caller must hold slots_lock */
 struct kvm_pit *kvm_create_pit(struct kvm *kvm, u32 flags)
 {
 	struct kvm_pit *pit;
Index: kvm/arch/x86/kvm/i8259.c
===================================================================
--- kvm.orig/arch/x86/kvm/i8259.c
+++ kvm/arch/x86/kvm/i8259.c
@@ -533,9 +533,9 @@ struct kvm_pic *kvm_create_pic(struct kv
 	 * Initialize PIO device
 	 */
 	kvm_iodevice_init(&s->dev, &picdev_ops);
-	down_write(&kvm->slots_lock);
+	mutex_lock(&kvm->slots_lock);
 	ret = kvm_io_bus_register_dev(kvm, KVM_PIO_BUS, &s->dev);
-	up_write(&kvm->slots_lock);
+	mutex_unlock(&kvm->slots_lock);
 	if (ret < 0) {
 		kfree(s);
 		return NULL;

Re: [patch 00/11] convert slotslock to SRCU v2

[Avi Kivity]

On 12/23/2009 06:35 PM, Marcelo Tosatti wrote:
> Addressing comments.
>    

Looks good.


-- 
error compiling committee.c: too many arguments to function

Re: [patch 00/11] convert slotslock to SRCU v2

[Gleb Natapov]

On Thu, Dec 24, 2009 at 10:30:24AM +0200, Avi Kivity wrote:
> On 12/23/2009 06:35 PM, Marcelo Tosatti wrote:
> >Addressing comments.
> 
> Looks good.
> 
> 
Looks good to me too.

--
			Gleb.

Re: [patch 00/11] convert slotslock to SRCU v2

[Sheng Yang]

On Thursday 24 December 2009 22:56:38 Gleb Natapov wrote:
> On Thu, Dec 24, 2009 at 10:30:24AM +0200, Avi Kivity wrote:
> > On 12/23/2009 06:35 PM, Marcelo Tosatti wrote:
> > >Addressing comments.
> >
> > Looks good.
> 
> Looks good to me too.

One small comment on the naming:

kvm->memslots->memslots[i]
kvm->aliases->aliases[i]

Doesn't look good...

-- 
regards
Yang, Sheng

Re: [patch 00/11] convert slotslock to SRCU v2

[Marcelo Tosatti]

On Fri, Dec 25, 2009 at 03:24:49PM +0800, Sheng Yang wrote:
> On Thursday 24 December 2009 22:56:38 Gleb Natapov wrote:
> > On Thu, Dec 24, 2009 at 10:30:24AM +0200, Avi Kivity wrote:
> > > On 12/23/2009 06:35 PM, Marcelo Tosatti wrote:
> > > >Addressing comments.
> > >
> > > Looks good.
> > 
> > Looks good to me too.
> 
> One small comment on the naming:
> 
> kvm->memslots->memslots[i]
> kvm->aliases->aliases[i]
> 
> Doesn't look good...

Hi Sheng,

Yes its somewhat confusing but i could not come up with 
anything better. Suggestions?

Re: [patch 00/11] convert slotslock to SRCU v2

[Avi Kivity]

On 12/25/2009 02:25 PM, Marcelo Tosatti wrote:
>
>> One small comment on the naming:
>>
>> kvm->memslots->memslots[i]
>> kvm->aliases->aliases[i]
>>
>> Doesn't look good...
>>      
> Hi Sheng,
>
> Yes its somewhat confusing but i could not come up with
> anything better. Suggestions?
>    

Could all it 'mem', and put both ->memslots and ->aliases as members.  
This might simplify things a bit.

However, I don't recommend doing so now, it's just churn.  We can always 
to it later.

-- 
I have a truly marvellous patch that fixes the bug which this
signature is too narrow to contain.

Re: [patch 00/11] convert slotslock to SRCU v2

[Sheng Yang]

On Saturday 26 December 2009 00:27:11 Avi Kivity wrote:
> On 12/25/2009 02:25 PM, Marcelo Tosatti wrote:
> >> One small comment on the naming:
> >>
> >> kvm->memslots->memslots[i]
> >> kvm->aliases->aliases[i]
> >>
> >> Doesn't look good...
> >
> > Hi Sheng,
> >
> > Yes its somewhat confusing but i could not come up with
> > anything better. Suggestions?
> 
> Could all it 'mem', and put both ->memslots and ->aliases as members.
> This might simplify things a bit.
> 
> However, I don't recommend doing so now, it's just churn.  We can always
> to it later.
 
Yes. Just happen to see it, and feel a little uneasy... :)

-- 
regards
Yang, Sheng

Re: [patch 00/11] convert slotslock to SRCU

[Avi Kivity]

On 12/23/2009 01:38 PM, Marcelo Tosatti wrote:
> Now that synchronize_srcu_expedited is in the tree, we can continue the
> convertion of slots_lock to SRCU.
>
> Results:
> up:
>
>    

Without this patchset, -smp 64 is simply unusable.  An _idle_ Linux 
guest generates enough interrupts that the system is unable to make any 
progress:


[root@monster01 ~]# perf report -g
# Samples: 14214748264314
#
# Overhead          Command                      Shared Object  Symbol
# ........  ...............  .................................  ......
#
     40.81%  qemu-system-x86  [kernel]                           [k] 
_raw_spin_lock_irq
             |
             --- _raw_spin_lock_irq
                |
                |--99.94%-- __down_read
                |          down_read
                |          |
                |          |--99.82%-- 0xffffffffa00479c4
                |          |          0xffffffffa003a6c9
                |          |          vfs_ioctl
                |          |          do_vfs_ioctl
                |          |          sys_ioctl
                |          |          system_call
                |          |          __GI_ioctl
                |           --0.18%-- [...]
                 --0.06%-- [...]

     40.57%  qemu-system-x86  [kernel]                           [k] 
_raw_spin_lock_irqsave
             |
             --- _raw_spin_lock_irqsave
                |
                |--99.88%-- __up_read
                |          up_read
                |          |
                |          |--99.82%-- 0xffffffffa0047897
                |          |          0xffffffffa003a6c9
                |          |          vfs_ioctl
                |          |          do_vfs_ioctl
                |          |          sys_ioctl
                |          |          system_call
                |          |          __GI_ioctl
                |           --0.18%-- [...]
                 --0.12%-- [...]

With this patchset, -smp 64 flies.  Patch to follow.  Amazing work!

(and surprising how easy it is to saturate a large system)

-- 
error compiling committee.c: too many arguments to function