[PATCH] KVM: MMU: oom handling
authorAvi Kivity <avi@qumranet.com>
Sat, 6 Jan 2007 00:36:47 +0000 (16:36 -0800)
committerLinus Torvalds <torvalds@woody.osdl.org>
Sat, 6 Jan 2007 07:55:25 +0000 (23:55 -0800)
When beginning to process a page fault, make sure we have enough shadow pages
available to service the fault.  If not, free some pages.

Signed-off-by: Avi Kivity <avi@qumranet.com>
Acked-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
drivers/kvm/kvm.h
drivers/kvm/mmu.c
drivers/kvm/paging_tmpl.h
drivers/kvm/svm.c
drivers/kvm/vmx.c

index 34c43bb4d3482242e7544a9fd4729f63dc2e267b..1d0be85651f5426b6cf0033d5d400386af5fde62 100644 (file)
@@ -52,6 +52,8 @@
 #define KVM_MAX_VCPUS 1
 #define KVM_MEMORY_SLOTS 4
 #define KVM_NUM_MMU_PAGES 256
+#define KVM_MIN_FREE_MMU_PAGES 5
+#define KVM_REFILL_PAGES 25
 
 #define FX_IMAGE_SIZE 512
 #define FX_IMAGE_ALIGN 16
@@ -278,6 +280,7 @@ struct kvm {
         * Hash table of struct kvm_mmu_page.
         */
        struct list_head active_mmu_pages;
+       int n_free_mmu_pages;
        struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
        struct kvm_vcpu vcpus[KVM_MAX_VCPUS];
        int memory_config_version;
@@ -451,6 +454,15 @@ unsigned long segment_base(u16 selector);
 void kvm_mmu_pre_write(struct kvm_vcpu *vcpu, gpa_t gpa, int bytes);
 void kvm_mmu_post_write(struct kvm_vcpu *vcpu, gpa_t gpa, int bytes);
 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva);
+void kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
+
+static inline int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t gva,
+                                    u32 error_code)
+{
+       if (unlikely(vcpu->kvm->n_free_mmu_pages < KVM_MIN_FREE_MMU_PAGES))
+               kvm_mmu_free_some_pages(vcpu);
+       return vcpu->mmu.page_fault(vcpu, gva, error_code);
+}
 
 static inline struct page *_gfn_to_page(struct kvm *kvm, gfn_t gfn)
 {
index d788866d5a6f832fb43c3c407cd912a2d2a27fea..e4a20a45d834414cc904341819e5397b3db1a254 100644 (file)
@@ -310,6 +310,7 @@ static void kvm_mmu_free_page(struct kvm_vcpu *vcpu, hpa_t page_hpa)
        list_del(&page_head->link);
        page_head->page_hpa = page_hpa;
        list_add(&page_head->link, &vcpu->free_pages);
+       ++vcpu->kvm->n_free_mmu_pages;
 }
 
 static int is_empty_shadow_page(hpa_t page_hpa)
@@ -344,6 +345,7 @@ static struct kvm_mmu_page *kvm_mmu_alloc_page(struct kvm_vcpu *vcpu,
        page->global = 1;
        page->multimapped = 0;
        page->parent_pte = parent_pte;
+       --vcpu->kvm->n_free_mmu_pages;
        return page;
 }
 
@@ -544,8 +546,7 @@ static void kvm_mmu_zap_page(struct kvm_vcpu *vcpu,
        }
        kvm_mmu_page_unlink_children(vcpu, page);
        hlist_del(&page->hash_link);
-       list_del(&page->link);
-       list_add(&page->link, &vcpu->free_pages);
+       kvm_mmu_free_page(vcpu, page->page_hpa);
 }
 
 static int kvm_mmu_unprotect_page(struct kvm_vcpu *vcpu, gfn_t gfn)
@@ -743,18 +744,6 @@ static void mmu_alloc_roots(struct kvm_vcpu *vcpu)
        vcpu->mmu.root_hpa = __pa(vcpu->mmu.pae_root);
 }
 
-static void nonpaging_flush(struct kvm_vcpu *vcpu)
-{
-       hpa_t root = vcpu->mmu.root_hpa;
-
-       ++kvm_stat.tlb_flush;
-       pgprintk("nonpaging_flush\n");
-       mmu_free_roots(vcpu);
-       mmu_alloc_roots(vcpu);
-       kvm_arch_ops->set_cr3(vcpu, root);
-       kvm_arch_ops->tlb_flush(vcpu);
-}
-
 static gpa_t nonpaging_gva_to_gpa(struct kvm_vcpu *vcpu, gva_t vaddr)
 {
        return vaddr;
@@ -763,28 +752,19 @@ static gpa_t nonpaging_gva_to_gpa(struct kvm_vcpu *vcpu, gva_t vaddr)
 static int nonpaging_page_fault(struct kvm_vcpu *vcpu, gva_t gva,
                               u32 error_code)
 {
-       int ret;
        gpa_t addr = gva;
+       hpa_t paddr;
 
        ASSERT(vcpu);
        ASSERT(VALID_PAGE(vcpu->mmu.root_hpa));
 
-       for (;;) {
-            hpa_t paddr;
 
-            paddr = gpa_to_hpa(vcpu , addr & PT64_BASE_ADDR_MASK);
+       paddr = gpa_to_hpa(vcpu , addr & PT64_BASE_ADDR_MASK);
 
-            if (is_error_hpa(paddr))
-                    return 1;
+       if (is_error_hpa(paddr))
+               return 1;
 
-            ret = nonpaging_map(vcpu, addr & PAGE_MASK, paddr);
-            if (ret) {
-                    nonpaging_flush(vcpu);
-                    continue;
-            }
-            break;
-       }
-       return ret;
+       return nonpaging_map(vcpu, addr & PAGE_MASK, paddr);
 }
 
 static void nonpaging_inval_page(struct kvm_vcpu *vcpu, gva_t addr)
@@ -1093,6 +1073,18 @@ int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva)
        return kvm_mmu_unprotect_page(vcpu, gpa >> PAGE_SHIFT);
 }
 
+void kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu)
+{
+       while (vcpu->kvm->n_free_mmu_pages < KVM_REFILL_PAGES) {
+               struct kvm_mmu_page *page;
+
+               page = container_of(vcpu->kvm->active_mmu_pages.prev,
+                                   struct kvm_mmu_page, link);
+               kvm_mmu_zap_page(vcpu, page);
+       }
+}
+EXPORT_SYMBOL_GPL(kvm_mmu_free_some_pages);
+
 static void free_mmu_pages(struct kvm_vcpu *vcpu)
 {
        while (!list_empty(&vcpu->free_pages)) {
@@ -1124,6 +1116,7 @@ static int alloc_mmu_pages(struct kvm_vcpu *vcpu)
                page_header->page_hpa = (hpa_t)page_to_pfn(page) << PAGE_SHIFT;
                memset(__va(page_header->page_hpa), 0, PAGE_SIZE);
                list_add(&page_header->link, &vcpu->free_pages);
+               ++vcpu->kvm->n_free_mmu_pages;
        }
 
        /*
index cf4b74cc75b5874d878c66bd1ac88be7e072873f..03c474aaedded0f09959cbefa5e26b006089aee2 100644 (file)
@@ -246,8 +246,6 @@ static u64 *FNAME(fetch)(struct kvm_vcpu *vcpu, gva_t addr,
                }
                shadow_page = kvm_mmu_get_page(vcpu, table_gfn, addr, level-1,
                                               metaphysical, shadow_ent);
-               if (!shadow_page)
-                       return ERR_PTR(-ENOMEM);
                shadow_addr = shadow_page->page_hpa;
                shadow_pte = shadow_addr | PT_PRESENT_MASK | PT_ACCESSED_MASK
                        | PT_WRITABLE_MASK | PT_USER_MASK;
@@ -347,17 +345,8 @@ static int FNAME(page_fault)(struct kvm_vcpu *vcpu, gva_t addr,
        /*
         * Look up the shadow pte for the faulting address.
         */
-       for (;;) {
-               FNAME(walk_addr)(&walker, vcpu, addr);
-               shadow_pte = FNAME(fetch)(vcpu, addr, &walker);
-               if (IS_ERR(shadow_pte)) {  /* must be -ENOMEM */
-                       printk("%s: oom\n", __FUNCTION__);
-                       nonpaging_flush(vcpu);
-                       FNAME(release_walker)(&walker);
-                       continue;
-               }
-               break;
-       }
+       FNAME(walk_addr)(&walker, vcpu, addr);
+       shadow_pte = FNAME(fetch)(vcpu, addr, &walker);
 
        /*
         * The page is not mapped by the guest.  Let the guest handle it.
index 91c7f60ffd42dc39640434fc6eff7c8e2199c56f..869b524dda6bdee35503b4ecc3fd8a410378ca7c 100644 (file)
@@ -861,7 +861,7 @@ static int pf_interception(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
 
        fault_address  = vcpu->svm->vmcb->control.exit_info_2;
        error_code = vcpu->svm->vmcb->control.exit_info_1;
-       if (!vcpu->mmu.page_fault(vcpu, fault_address, error_code)) {
+       if (!kvm_mmu_page_fault(vcpu, fault_address, error_code)) {
                spin_unlock(&vcpu->kvm->lock);
                return 1;
        }
index aaa98e3e9cafb51c4719f4288108115ffd2090a0..2a1c37eed71194665406aa117537556920b59e5f 100644 (file)
@@ -1318,7 +1318,7 @@ static int handle_exception(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
                cr2 = vmcs_readl(EXIT_QUALIFICATION);
 
                spin_lock(&vcpu->kvm->lock);
-               if (!vcpu->mmu.page_fault(vcpu, cr2, error_code)) {
+               if (!kvm_mmu_page_fault(vcpu, cr2, error_code)) {
                        spin_unlock(&vcpu->kvm->lock);
                        return 1;
                }