diff options
author | Avi Kivity <avi@qumranet.com> | 2007-01-05 19:36:47 -0500 |
---|---|---|
committer | Linus Torvalds <torvalds@woody.osdl.org> | 2007-01-06 02:55:25 -0500 |
commit | ebeace8609205bf5e1b96fe325b7dea148042232 (patch) | |
tree | 863991639d7df2a4aa5fd8b20b6b3a28c2cc2cda /drivers/kvm | |
parent | cc4529efc7b730b596d9c7d5a917c00a357e92aa (diff) |
[PATCH] KVM: MMU: oom handling
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>
Diffstat (limited to 'drivers/kvm')
-rw-r--r-- | drivers/kvm/kvm.h | 12 | ||||
-rw-r--r-- | drivers/kvm/mmu.c | 49 | ||||
-rw-r--r-- | drivers/kvm/paging_tmpl.h | 15 | ||||
-rw-r--r-- | drivers/kvm/svm.c | 2 | ||||
-rw-r--r-- | drivers/kvm/vmx.c | 2 |
5 files changed, 37 insertions, 43 deletions
diff --git a/drivers/kvm/kvm.h b/drivers/kvm/kvm.h index 34c43bb4d348..1d0be85651f5 100644 --- a/drivers/kvm/kvm.h +++ b/drivers/kvm/kvm.h | |||
@@ -52,6 +52,8 @@ | |||
52 | #define KVM_MAX_VCPUS 1 | 52 | #define KVM_MAX_VCPUS 1 |
53 | #define KVM_MEMORY_SLOTS 4 | 53 | #define KVM_MEMORY_SLOTS 4 |
54 | #define KVM_NUM_MMU_PAGES 256 | 54 | #define KVM_NUM_MMU_PAGES 256 |
55 | #define KVM_MIN_FREE_MMU_PAGES 5 | ||
56 | #define KVM_REFILL_PAGES 25 | ||
55 | 57 | ||
56 | #define FX_IMAGE_SIZE 512 | 58 | #define FX_IMAGE_SIZE 512 |
57 | #define FX_IMAGE_ALIGN 16 | 59 | #define FX_IMAGE_ALIGN 16 |
@@ -278,6 +280,7 @@ struct kvm { | |||
278 | * Hash table of struct kvm_mmu_page. | 280 | * Hash table of struct kvm_mmu_page. |
279 | */ | 281 | */ |
280 | struct list_head active_mmu_pages; | 282 | struct list_head active_mmu_pages; |
283 | int n_free_mmu_pages; | ||
281 | struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES]; | 284 | struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES]; |
282 | struct kvm_vcpu vcpus[KVM_MAX_VCPUS]; | 285 | struct kvm_vcpu vcpus[KVM_MAX_VCPUS]; |
283 | int memory_config_version; | 286 | int memory_config_version; |
@@ -451,6 +454,15 @@ unsigned long segment_base(u16 selector); | |||
451 | void kvm_mmu_pre_write(struct kvm_vcpu *vcpu, gpa_t gpa, int bytes); | 454 | void kvm_mmu_pre_write(struct kvm_vcpu *vcpu, gpa_t gpa, int bytes); |
452 | void kvm_mmu_post_write(struct kvm_vcpu *vcpu, gpa_t gpa, int bytes); | 455 | void kvm_mmu_post_write(struct kvm_vcpu *vcpu, gpa_t gpa, int bytes); |
453 | int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva); | 456 | int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva); |
457 | void kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu); | ||
458 | |||
459 | static inline int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t gva, | ||
460 | u32 error_code) | ||
461 | { | ||
462 | if (unlikely(vcpu->kvm->n_free_mmu_pages < KVM_MIN_FREE_MMU_PAGES)) | ||
463 | kvm_mmu_free_some_pages(vcpu); | ||
464 | return vcpu->mmu.page_fault(vcpu, gva, error_code); | ||
465 | } | ||
454 | 466 | ||
455 | static inline struct page *_gfn_to_page(struct kvm *kvm, gfn_t gfn) | 467 | static inline struct page *_gfn_to_page(struct kvm *kvm, gfn_t gfn) |
456 | { | 468 | { |
diff --git a/drivers/kvm/mmu.c b/drivers/kvm/mmu.c index d788866d5a6f..e4a20a45d834 100644 --- a/drivers/kvm/mmu.c +++ b/drivers/kvm/mmu.c | |||
@@ -310,6 +310,7 @@ static void kvm_mmu_free_page(struct kvm_vcpu *vcpu, hpa_t page_hpa) | |||
310 | list_del(&page_head->link); | 310 | list_del(&page_head->link); |
311 | page_head->page_hpa = page_hpa; | 311 | page_head->page_hpa = page_hpa; |
312 | list_add(&page_head->link, &vcpu->free_pages); | 312 | list_add(&page_head->link, &vcpu->free_pages); |
313 | ++vcpu->kvm->n_free_mmu_pages; | ||
313 | } | 314 | } |
314 | 315 | ||
315 | static int is_empty_shadow_page(hpa_t page_hpa) | 316 | 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, | |||
344 | page->global = 1; | 345 | page->global = 1; |
345 | page->multimapped = 0; | 346 | page->multimapped = 0; |
346 | page->parent_pte = parent_pte; | 347 | page->parent_pte = parent_pte; |
348 | --vcpu->kvm->n_free_mmu_pages; | ||
347 | return page; | 349 | return page; |
348 | } | 350 | } |
349 | 351 | ||
@@ -544,8 +546,7 @@ static void kvm_mmu_zap_page(struct kvm_vcpu *vcpu, | |||
544 | } | 546 | } |
545 | kvm_mmu_page_unlink_children(vcpu, page); | 547 | kvm_mmu_page_unlink_children(vcpu, page); |
546 | hlist_del(&page->hash_link); | 548 | hlist_del(&page->hash_link); |
547 | list_del(&page->link); | 549 | kvm_mmu_free_page(vcpu, page->page_hpa); |
548 | list_add(&page->link, &vcpu->free_pages); | ||
549 | } | 550 | } |
550 | 551 | ||
551 | static int kvm_mmu_unprotect_page(struct kvm_vcpu *vcpu, gfn_t gfn) | 552 | 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) | |||
743 | vcpu->mmu.root_hpa = __pa(vcpu->mmu.pae_root); | 744 | vcpu->mmu.root_hpa = __pa(vcpu->mmu.pae_root); |
744 | } | 745 | } |
745 | 746 | ||
746 | static void nonpaging_flush(struct kvm_vcpu *vcpu) | ||
747 | { | ||
748 | hpa_t root = vcpu->mmu.root_hpa; | ||
749 | |||
750 | ++kvm_stat.tlb_flush; | ||
751 | pgprintk("nonpaging_flush\n"); | ||
752 | mmu_free_roots(vcpu); | ||
753 | mmu_alloc_roots(vcpu); | ||
754 | kvm_arch_ops->set_cr3(vcpu, root); | ||
755 | kvm_arch_ops->tlb_flush(vcpu); | ||
756 | } | ||
757 | |||
758 | static gpa_t nonpaging_gva_to_gpa(struct kvm_vcpu *vcpu, gva_t vaddr) | 747 | static gpa_t nonpaging_gva_to_gpa(struct kvm_vcpu *vcpu, gva_t vaddr) |
759 | { | 748 | { |
760 | return vaddr; | 749 | return vaddr; |
@@ -763,28 +752,19 @@ static gpa_t nonpaging_gva_to_gpa(struct kvm_vcpu *vcpu, gva_t vaddr) | |||
763 | static int nonpaging_page_fault(struct kvm_vcpu *vcpu, gva_t gva, | 752 | static int nonpaging_page_fault(struct kvm_vcpu *vcpu, gva_t gva, |
764 | u32 error_code) | 753 | u32 error_code) |
765 | { | 754 | { |
766 | int ret; | ||
767 | gpa_t addr = gva; | 755 | gpa_t addr = gva; |
756 | hpa_t paddr; | ||
768 | 757 | ||
769 | ASSERT(vcpu); | 758 | ASSERT(vcpu); |
770 | ASSERT(VALID_PAGE(vcpu->mmu.root_hpa)); | 759 | ASSERT(VALID_PAGE(vcpu->mmu.root_hpa)); |
771 | 760 | ||
772 | for (;;) { | ||
773 | hpa_t paddr; | ||
774 | 761 | ||
775 | paddr = gpa_to_hpa(vcpu , addr & PT64_BASE_ADDR_MASK); | 762 | paddr = gpa_to_hpa(vcpu , addr & PT64_BASE_ADDR_MASK); |
776 | 763 | ||
777 | if (is_error_hpa(paddr)) | 764 | if (is_error_hpa(paddr)) |
778 | return 1; | 765 | return 1; |
779 | 766 | ||
780 | ret = nonpaging_map(vcpu, addr & PAGE_MASK, paddr); | 767 | return nonpaging_map(vcpu, addr & PAGE_MASK, paddr); |
781 | if (ret) { | ||
782 | nonpaging_flush(vcpu); | ||
783 | continue; | ||
784 | } | ||
785 | break; | ||
786 | } | ||
787 | return ret; | ||
788 | } | 768 | } |
789 | 769 | ||
790 | static void nonpaging_inval_page(struct kvm_vcpu *vcpu, gva_t addr) | 770 | 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) | |||
1093 | return kvm_mmu_unprotect_page(vcpu, gpa >> PAGE_SHIFT); | 1073 | return kvm_mmu_unprotect_page(vcpu, gpa >> PAGE_SHIFT); |
1094 | } | 1074 | } |
1095 | 1075 | ||
1076 | void kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu) | ||
1077 | { | ||
1078 | while (vcpu->kvm->n_free_mmu_pages < KVM_REFILL_PAGES) { | ||
1079 | struct kvm_mmu_page *page; | ||
1080 | |||
1081 | page = container_of(vcpu->kvm->active_mmu_pages.prev, | ||
1082 | struct kvm_mmu_page, link); | ||
1083 | kvm_mmu_zap_page(vcpu, page); | ||
1084 | } | ||
1085 | } | ||
1086 | EXPORT_SYMBOL_GPL(kvm_mmu_free_some_pages); | ||
1087 | |||
1096 | static void free_mmu_pages(struct kvm_vcpu *vcpu) | 1088 | static void free_mmu_pages(struct kvm_vcpu *vcpu) |
1097 | { | 1089 | { |
1098 | while (!list_empty(&vcpu->free_pages)) { | 1090 | while (!list_empty(&vcpu->free_pages)) { |
@@ -1124,6 +1116,7 @@ static int alloc_mmu_pages(struct kvm_vcpu *vcpu) | |||
1124 | page_header->page_hpa = (hpa_t)page_to_pfn(page) << PAGE_SHIFT; | 1116 | page_header->page_hpa = (hpa_t)page_to_pfn(page) << PAGE_SHIFT; |
1125 | memset(__va(page_header->page_hpa), 0, PAGE_SIZE); | 1117 | memset(__va(page_header->page_hpa), 0, PAGE_SIZE); |
1126 | list_add(&page_header->link, &vcpu->free_pages); | 1118 | list_add(&page_header->link, &vcpu->free_pages); |
1119 | ++vcpu->kvm->n_free_mmu_pages; | ||
1127 | } | 1120 | } |
1128 | 1121 | ||
1129 | /* | 1122 | /* |
diff --git a/drivers/kvm/paging_tmpl.h b/drivers/kvm/paging_tmpl.h index cf4b74cc75b5..03c474aaedde 100644 --- a/drivers/kvm/paging_tmpl.h +++ b/drivers/kvm/paging_tmpl.h | |||
@@ -246,8 +246,6 @@ static u64 *FNAME(fetch)(struct kvm_vcpu *vcpu, gva_t addr, | |||
246 | } | 246 | } |
247 | shadow_page = kvm_mmu_get_page(vcpu, table_gfn, addr, level-1, | 247 | shadow_page = kvm_mmu_get_page(vcpu, table_gfn, addr, level-1, |
248 | metaphysical, shadow_ent); | 248 | metaphysical, shadow_ent); |
249 | if (!shadow_page) | ||
250 | return ERR_PTR(-ENOMEM); | ||
251 | shadow_addr = shadow_page->page_hpa; | 249 | shadow_addr = shadow_page->page_hpa; |
252 | shadow_pte = shadow_addr | PT_PRESENT_MASK | PT_ACCESSED_MASK | 250 | shadow_pte = shadow_addr | PT_PRESENT_MASK | PT_ACCESSED_MASK |
253 | | PT_WRITABLE_MASK | PT_USER_MASK; | 251 | | PT_WRITABLE_MASK | PT_USER_MASK; |
@@ -347,17 +345,8 @@ static int FNAME(page_fault)(struct kvm_vcpu *vcpu, gva_t addr, | |||
347 | /* | 345 | /* |
348 | * Look up the shadow pte for the faulting address. | 346 | * Look up the shadow pte for the faulting address. |
349 | */ | 347 | */ |
350 | for (;;) { | 348 | FNAME(walk_addr)(&walker, vcpu, addr); |
351 | FNAME(walk_addr)(&walker, vcpu, addr); | 349 | shadow_pte = FNAME(fetch)(vcpu, addr, &walker); |
352 | shadow_pte = FNAME(fetch)(vcpu, addr, &walker); | ||
353 | if (IS_ERR(shadow_pte)) { /* must be -ENOMEM */ | ||
354 | printk("%s: oom\n", __FUNCTION__); | ||
355 | nonpaging_flush(vcpu); | ||
356 | FNAME(release_walker)(&walker); | ||
357 | continue; | ||
358 | } | ||
359 | break; | ||
360 | } | ||
361 | 350 | ||
362 | /* | 351 | /* |
363 | * The page is not mapped by the guest. Let the guest handle it. | 352 | * The page is not mapped by the guest. Let the guest handle it. |
diff --git a/drivers/kvm/svm.c b/drivers/kvm/svm.c index 91c7f60ffd42..869b524dda6b 100644 --- a/drivers/kvm/svm.c +++ b/drivers/kvm/svm.c | |||
@@ -861,7 +861,7 @@ static int pf_interception(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run) | |||
861 | 861 | ||
862 | fault_address = vcpu->svm->vmcb->control.exit_info_2; | 862 | fault_address = vcpu->svm->vmcb->control.exit_info_2; |
863 | error_code = vcpu->svm->vmcb->control.exit_info_1; | 863 | error_code = vcpu->svm->vmcb->control.exit_info_1; |
864 | if (!vcpu->mmu.page_fault(vcpu, fault_address, error_code)) { | 864 | if (!kvm_mmu_page_fault(vcpu, fault_address, error_code)) { |
865 | spin_unlock(&vcpu->kvm->lock); | 865 | spin_unlock(&vcpu->kvm->lock); |
866 | return 1; | 866 | return 1; |
867 | } | 867 | } |
diff --git a/drivers/kvm/vmx.c b/drivers/kvm/vmx.c index aaa98e3e9caf..2a1c37eed711 100644 --- a/drivers/kvm/vmx.c +++ b/drivers/kvm/vmx.c | |||
@@ -1318,7 +1318,7 @@ static int handle_exception(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run) | |||
1318 | cr2 = vmcs_readl(EXIT_QUALIFICATION); | 1318 | cr2 = vmcs_readl(EXIT_QUALIFICATION); |
1319 | 1319 | ||
1320 | spin_lock(&vcpu->kvm->lock); | 1320 | spin_lock(&vcpu->kvm->lock); |
1321 | if (!vcpu->mmu.page_fault(vcpu, cr2, error_code)) { | 1321 | if (!kvm_mmu_page_fault(vcpu, cr2, error_code)) { |
1322 | spin_unlock(&vcpu->kvm->lock); | 1322 | spin_unlock(&vcpu->kvm->lock); |
1323 | return 1; | 1323 | return 1; |
1324 | } | 1324 | } |