5 * This work is licensed under the terms of the GNU GPL, version 2. See
6 * the COPYING file in the top-level directory.
9 #include <linux/types.h>
10 #include <linux/list.h>
11 #include <linux/mutex.h>
12 #include <linux/spinlock.h>
16 #include <linux/kvm.h>
18 #define CR0_PE_MASK (1ULL << 0)
19 #define CR0_TS_MASK (1ULL << 3)
20 #define CR0_NE_MASK (1ULL << 5)
21 #define CR0_WP_MASK (1ULL << 16)
22 #define CR0_NW_MASK (1ULL << 29)
23 #define CR0_CD_MASK (1ULL << 30)
24 #define CR0_PG_MASK (1ULL << 31)
26 #define CR3_WPT_MASK (1ULL << 3)
27 #define CR3_PCD_MASK (1ULL << 4)
29 #define CR3_RESEVED_BITS 0x07ULL
30 #define CR3_L_MODE_RESEVED_BITS (~((1ULL << 40) - 1) | 0x0fe7ULL)
31 #define CR3_FLAGS_MASK ((1ULL << 5) - 1)
33 #define CR4_VME_MASK (1ULL << 0)
34 #define CR4_PSE_MASK (1ULL << 4)
35 #define CR4_PAE_MASK (1ULL << 5)
36 #define CR4_PGE_MASK (1ULL << 7)
37 #define CR4_VMXE_MASK (1ULL << 13)
39 #define KVM_GUEST_CR0_MASK \
40 (CR0_PG_MASK | CR0_PE_MASK | CR0_WP_MASK | CR0_NE_MASK \
41 | CR0_NW_MASK | CR0_CD_MASK)
42 #define KVM_VM_CR0_ALWAYS_ON \
43 (CR0_PG_MASK | CR0_PE_MASK | CR0_WP_MASK | CR0_NE_MASK)
44 #define KVM_GUEST_CR4_MASK \
45 (CR4_PSE_MASK | CR4_PAE_MASK | CR4_PGE_MASK | CR4_VMXE_MASK | CR4_VME_MASK)
46 #define KVM_PMODE_VM_CR4_ALWAYS_ON (CR4_VMXE_MASK | CR4_PAE_MASK)
47 #define KVM_RMODE_VM_CR4_ALWAYS_ON (CR4_VMXE_MASK | CR4_PAE_MASK | CR4_VME_MASK)
49 #define INVALID_PAGE (~(hpa_t)0)
50 #define UNMAPPED_GVA (~(gpa_t)0)
52 #define KVM_MAX_VCPUS 1
53 #define KVM_MEMORY_SLOTS 4
54 #define KVM_NUM_MMU_PAGES 256
55 #define KVM_MIN_FREE_MMU_PAGES 5
56 #define KVM_REFILL_PAGES 25
58 #define FX_IMAGE_SIZE 512
59 #define FX_IMAGE_ALIGN 16
60 #define FX_BUF_SIZE (2 * FX_IMAGE_SIZE + FX_IMAGE_ALIGN)
70 #define SELECTOR_TI_MASK (1 << 2)
71 #define SELECTOR_RPL_MASK 0x03
78 * gva - guest virtual address
79 * gpa - guest physical address
80 * gfn - guest frame number
81 * hva - host virtual address
82 * hpa - host physical address
83 * hfn - host frame number
86 typedef unsigned long gva_t;
88 typedef unsigned long gfn_t;
90 typedef unsigned long hva_t;
92 typedef unsigned long hfn_t;
94 #define NR_PTE_CHAIN_ENTRIES 5
96 struct kvm_pte_chain {
97 u64 *parent_ptes[NR_PTE_CHAIN_ENTRIES];
98 struct hlist_node link;
102 * kvm_mmu_page_role, below, is defined as:
104 * bits 0:3 - total guest paging levels (2-4, or zero for real mode)
105 * bits 4:7 - page table level for this shadow (1-4)
106 * bits 8:9 - page table quadrant for 2-level guests
107 * bit 16 - "metaphysical" - gfn is not a real page (huge page/real mode)
109 union kvm_mmu_page_role {
112 unsigned glevels : 4;
114 unsigned quadrant : 2;
115 unsigned pad_for_nice_hex_output : 6;
116 unsigned metaphysical : 1;
120 struct kvm_mmu_page {
121 struct list_head link;
122 struct hlist_node hash_link;
125 * The following two entries are used to key the shadow page in the
129 union kvm_mmu_page_role role;
132 unsigned long slot_bitmap; /* One bit set per slot which has memory
133 * in this shadow page.
135 int global; /* Set if all ptes in this page are global */
136 int multimapped; /* More than one parent_pte? */
138 u64 *parent_pte; /* !multimapped */
139 struct hlist_head parent_ptes; /* multimapped, kvm_pte_chain */
149 #define vmx_msr_entry kvm_msr_entry
154 * x86 supports 3 paging modes (4-level 64-bit, 3-level 64-bit, and 2-level
155 * 32-bit). The kvm_mmu structure abstracts the details of the current mmu
159 void (*new_cr3)(struct kvm_vcpu *vcpu);
160 int (*page_fault)(struct kvm_vcpu *vcpu, gva_t gva, u32 err);
161 void (*free)(struct kvm_vcpu *vcpu);
162 gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t gva);
165 int shadow_root_level;
170 struct kvm_guest_debug {
213 struct vcpu_svm *svm;
218 int interrupt_window_open;
219 unsigned long irq_summary; /* bit vector: 1 per word in irq_pending */
220 #define NR_IRQ_WORDS KVM_IRQ_BITMAP_SIZE(unsigned long)
221 unsigned long irq_pending[NR_IRQ_WORDS];
222 unsigned long regs[NR_VCPU_REGS]; /* for rsp: vcpu_load_rsp_rip() */
223 unsigned long rip; /* needs vcpu_load_rsp_rip() */
230 u64 pdptrs[4]; /* pae */
234 struct vmx_msr_entry *guest_msrs;
235 struct vmx_msr_entry *host_msrs;
237 struct list_head free_pages;
238 struct kvm_mmu_page page_header_buf[KVM_NUM_MMU_PAGES];
241 gfn_t last_pt_write_gfn;
242 int last_pt_write_count;
244 struct kvm_guest_debug guest_debug;
246 char fx_buf[FX_BUF_SIZE];
248 char *guest_fx_image;
251 int mmio_read_completed;
254 unsigned char mmio_data[8];
255 gpa_t mmio_phys_addr;
260 struct kvm_save_segment {
265 } tr, es, ds, fs, gs;
269 struct kvm_memory_slot {
271 unsigned long npages;
273 struct page **phys_mem;
274 unsigned long *dirty_bitmap;
278 spinlock_t lock; /* protects everything except vcpus */
280 struct kvm_memory_slot memslots[KVM_MEMORY_SLOTS];
282 * Hash table of struct kvm_mmu_page.
284 struct list_head active_mmu_pages;
285 int n_free_mmu_pages;
286 struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
287 struct kvm_vcpu vcpus[KVM_MAX_VCPUS];
288 int memory_config_version;
290 unsigned long rmap_overflow;
303 u32 irq_window_exits;
305 u32 request_irq_exits;
309 struct descriptor_table {
312 } __attribute__((packed));
314 struct kvm_arch_ops {
315 int (*cpu_has_kvm_support)(void); /* __init */
316 int (*disabled_by_bios)(void); /* __init */
317 void (*hardware_enable)(void *dummy); /* __init */
318 void (*hardware_disable)(void *dummy);
319 int (*hardware_setup)(void); /* __init */
320 void (*hardware_unsetup)(void); /* __exit */
322 int (*vcpu_create)(struct kvm_vcpu *vcpu);
323 void (*vcpu_free)(struct kvm_vcpu *vcpu);
325 struct kvm_vcpu *(*vcpu_load)(struct kvm_vcpu *vcpu);
326 void (*vcpu_put)(struct kvm_vcpu *vcpu);
328 int (*set_guest_debug)(struct kvm_vcpu *vcpu,
329 struct kvm_debug_guest *dbg);
330 int (*get_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata);
331 int (*set_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 data);
332 u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
333 void (*get_segment)(struct kvm_vcpu *vcpu,
334 struct kvm_segment *var, int seg);
335 void (*set_segment)(struct kvm_vcpu *vcpu,
336 struct kvm_segment *var, int seg);
337 void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
338 void (*decache_cr0_cr4_guest_bits)(struct kvm_vcpu *vcpu);
339 void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
340 void (*set_cr0_no_modeswitch)(struct kvm_vcpu *vcpu,
342 void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
343 void (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
344 void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
345 void (*get_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
346 void (*set_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
347 void (*get_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
348 void (*set_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
349 unsigned long (*get_dr)(struct kvm_vcpu *vcpu, int dr);
350 void (*set_dr)(struct kvm_vcpu *vcpu, int dr, unsigned long value,
352 void (*cache_regs)(struct kvm_vcpu *vcpu);
353 void (*decache_regs)(struct kvm_vcpu *vcpu);
354 unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
355 void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
357 void (*invlpg)(struct kvm_vcpu *vcpu, gva_t addr);
358 void (*tlb_flush)(struct kvm_vcpu *vcpu);
359 void (*inject_page_fault)(struct kvm_vcpu *vcpu,
360 unsigned long addr, u32 err_code);
362 void (*inject_gp)(struct kvm_vcpu *vcpu, unsigned err_code);
364 int (*run)(struct kvm_vcpu *vcpu, struct kvm_run *run);
365 int (*vcpu_setup)(struct kvm_vcpu *vcpu);
366 void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
369 extern struct kvm_stat kvm_stat;
370 extern struct kvm_arch_ops *kvm_arch_ops;
372 #define kvm_printf(kvm, fmt ...) printk(KERN_DEBUG fmt)
373 #define vcpu_printf(vcpu, fmt...) kvm_printf(vcpu->kvm, fmt)
375 int kvm_init_arch(struct kvm_arch_ops *ops, struct module *module);
376 void kvm_exit_arch(void);
378 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
379 int kvm_mmu_create(struct kvm_vcpu *vcpu);
380 int kvm_mmu_setup(struct kvm_vcpu *vcpu);
382 int kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
383 void kvm_mmu_slot_remove_write_access(struct kvm *kvm, int slot);
385 hpa_t gpa_to_hpa(struct kvm_vcpu *vcpu, gpa_t gpa);
386 #define HPA_MSB ((sizeof(hpa_t) * 8) - 1)
387 #define HPA_ERR_MASK ((hpa_t)1 << HPA_MSB)
388 static inline int is_error_hpa(hpa_t hpa) { return hpa >> HPA_MSB; }
389 hpa_t gva_to_hpa(struct kvm_vcpu *vcpu, gva_t gva);
391 void kvm_emulator_want_group7_invlpg(void);
393 extern hpa_t bad_page_address;
395 static inline struct page *gfn_to_page(struct kvm_memory_slot *slot, gfn_t gfn)
397 return slot->phys_mem[gfn - slot->base_gfn];
400 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
401 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
403 enum emulation_result {
404 EMULATE_DONE, /* no further processing */
405 EMULATE_DO_MMIO, /* kvm_run filled with mmio request */
406 EMULATE_FAIL, /* can't emulate this instruction */
409 int emulate_instruction(struct kvm_vcpu *vcpu, struct kvm_run *run,
410 unsigned long cr2, u16 error_code);
411 void realmode_lgdt(struct kvm_vcpu *vcpu, u16 size, unsigned long address);
412 void realmode_lidt(struct kvm_vcpu *vcpu, u16 size, unsigned long address);
413 void realmode_lmsw(struct kvm_vcpu *vcpu, unsigned long msw,
414 unsigned long *rflags);
416 unsigned long realmode_get_cr(struct kvm_vcpu *vcpu, int cr);
417 void realmode_set_cr(struct kvm_vcpu *vcpu, int cr, unsigned long value,
418 unsigned long *rflags);
420 struct x86_emulate_ctxt;
422 int emulate_invlpg(struct kvm_vcpu *vcpu, gva_t address);
423 int emulate_clts(struct kvm_vcpu *vcpu);
424 int emulator_get_dr(struct x86_emulate_ctxt* ctxt, int dr,
425 unsigned long *dest);
426 int emulator_set_dr(struct x86_emulate_ctxt *ctxt, int dr,
427 unsigned long value);
429 void set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
430 void set_cr3(struct kvm_vcpu *vcpu, unsigned long cr0);
431 void set_cr4(struct kvm_vcpu *vcpu, unsigned long cr0);
432 void set_cr8(struct kvm_vcpu *vcpu, unsigned long cr0);
433 void lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
435 int kvm_get_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata);
436 int kvm_set_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 data);
438 void fx_init(struct kvm_vcpu *vcpu);
440 void load_msrs(struct vmx_msr_entry *e, int n);
441 void save_msrs(struct vmx_msr_entry *e, int n);
442 void kvm_resched(struct kvm_vcpu *vcpu);
444 int kvm_read_guest(struct kvm_vcpu *vcpu,
449 int kvm_write_guest(struct kvm_vcpu *vcpu,
454 unsigned long segment_base(u16 selector);
456 void kvm_mmu_pre_write(struct kvm_vcpu *vcpu, gpa_t gpa, int bytes);
457 void kvm_mmu_post_write(struct kvm_vcpu *vcpu, gpa_t gpa, int bytes);
458 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva);
459 void kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
461 static inline int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t gva,
464 if (unlikely(vcpu->kvm->n_free_mmu_pages < KVM_MIN_FREE_MMU_PAGES))
465 kvm_mmu_free_some_pages(vcpu);
466 return vcpu->mmu.page_fault(vcpu, gva, error_code);
469 static inline struct page *_gfn_to_page(struct kvm *kvm, gfn_t gfn)
471 struct kvm_memory_slot *slot = gfn_to_memslot(kvm, gfn);
472 return (slot) ? slot->phys_mem[gfn - slot->base_gfn] : NULL;
475 static inline int is_long_mode(struct kvm_vcpu *vcpu)
478 return vcpu->shadow_efer & EFER_LME;
484 static inline int is_pae(struct kvm_vcpu *vcpu)
486 return vcpu->cr4 & CR4_PAE_MASK;
489 static inline int is_pse(struct kvm_vcpu *vcpu)
491 return vcpu->cr4 & CR4_PSE_MASK;
494 static inline int is_paging(struct kvm_vcpu *vcpu)
496 return vcpu->cr0 & CR0_PG_MASK;
499 static inline int memslot_id(struct kvm *kvm, struct kvm_memory_slot *slot)
501 return slot - kvm->memslots;
504 static inline struct kvm_mmu_page *page_header(hpa_t shadow_page)
506 struct page *page = pfn_to_page(shadow_page >> PAGE_SHIFT);
508 return (struct kvm_mmu_page *)page->private;
511 static inline u16 read_fs(void)
514 asm ("mov %%fs, %0" : "=g"(seg));
518 static inline u16 read_gs(void)
521 asm ("mov %%gs, %0" : "=g"(seg));
525 static inline u16 read_ldt(void)
528 asm ("sldt %0" : "=g"(ldt));
532 static inline void load_fs(u16 sel)
534 asm ("mov %0, %%fs" : : "rm"(sel));
537 static inline void load_gs(u16 sel)
539 asm ("mov %0, %%gs" : : "rm"(sel));
543 static inline void load_ldt(u16 sel)
545 asm ("lldt %0" : : "g"(sel));
549 static inline void get_idt(struct descriptor_table *table)
551 asm ("sidt %0" : "=m"(*table));
554 static inline void get_gdt(struct descriptor_table *table)
556 asm ("sgdt %0" : "=m"(*table));
559 static inline unsigned long read_tr_base(void)
562 asm ("str %0" : "=g"(tr));
563 return segment_base(tr);
567 static inline unsigned long read_msr(unsigned long msr)
576 static inline void fx_save(void *image)
578 asm ("fxsave (%0)":: "r" (image));
581 static inline void fx_restore(void *image)
583 asm ("fxrstor (%0)":: "r" (image));
586 static inline void fpu_init(void)
591 static inline u32 get_rdx_init_val(void)
593 return 0x600; /* P6 family */
596 #define ASM_VMX_VMCLEAR_RAX ".byte 0x66, 0x0f, 0xc7, 0x30"
597 #define ASM_VMX_VMLAUNCH ".byte 0x0f, 0x01, 0xc2"
598 #define ASM_VMX_VMRESUME ".byte 0x0f, 0x01, 0xc3"
599 #define ASM_VMX_VMPTRLD_RAX ".byte 0x0f, 0xc7, 0x30"
600 #define ASM_VMX_VMREAD_RDX_RAX ".byte 0x0f, 0x78, 0xd0"
601 #define ASM_VMX_VMWRITE_RAX_RDX ".byte 0x0f, 0x79, 0xd0"
602 #define ASM_VMX_VMWRITE_RSP_RDX ".byte 0x0f, 0x79, 0xd4"
603 #define ASM_VMX_VMXOFF ".byte 0x0f, 0x01, 0xc4"
604 #define ASM_VMX_VMXON_RAX ".byte 0xf3, 0x0f, 0xc7, 0x30"
606 #define MSR_IA32_TIME_STAMP_COUNTER 0x010
608 #define TSS_IOPB_BASE_OFFSET 0x66
609 #define TSS_BASE_SIZE 0x68
610 #define TSS_IOPB_SIZE (65536 / 8)
611 #define TSS_REDIRECTION_SIZE (256 / 8)
612 #define RMODE_TSS_SIZE (TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)