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>
17 #include <linux/kvm_para.h>
19 #define CR0_PE_MASK (1ULL << 0)
20 #define CR0_TS_MASK (1ULL << 3)
21 #define CR0_NE_MASK (1ULL << 5)
22 #define CR0_WP_MASK (1ULL << 16)
23 #define CR0_NW_MASK (1ULL << 29)
24 #define CR0_CD_MASK (1ULL << 30)
25 #define CR0_PG_MASK (1ULL << 31)
27 #define CR3_WPT_MASK (1ULL << 3)
28 #define CR3_PCD_MASK (1ULL << 4)
30 #define CR3_RESEVED_BITS 0x07ULL
31 #define CR3_L_MODE_RESEVED_BITS (~((1ULL << 40) - 1) | 0x0fe7ULL)
32 #define CR3_FLAGS_MASK ((1ULL << 5) - 1)
34 #define CR4_VME_MASK (1ULL << 0)
35 #define CR4_PSE_MASK (1ULL << 4)
36 #define CR4_PAE_MASK (1ULL << 5)
37 #define CR4_PGE_MASK (1ULL << 7)
38 #define CR4_VMXE_MASK (1ULL << 13)
40 #define KVM_GUEST_CR0_MASK \
41 (CR0_PG_MASK | CR0_PE_MASK | CR0_WP_MASK | CR0_NE_MASK \
42 | CR0_NW_MASK | CR0_CD_MASK)
43 #define KVM_VM_CR0_ALWAYS_ON \
44 (CR0_PG_MASK | CR0_PE_MASK | CR0_WP_MASK | CR0_NE_MASK)
45 #define KVM_GUEST_CR4_MASK \
46 (CR4_PSE_MASK | CR4_PAE_MASK | CR4_PGE_MASK | CR4_VMXE_MASK | CR4_VME_MASK)
47 #define KVM_PMODE_VM_CR4_ALWAYS_ON (CR4_VMXE_MASK | CR4_PAE_MASK)
48 #define KVM_RMODE_VM_CR4_ALWAYS_ON (CR4_VMXE_MASK | CR4_PAE_MASK | CR4_VME_MASK)
50 #define INVALID_PAGE (~(hpa_t)0)
51 #define UNMAPPED_GVA (~(gpa_t)0)
53 #define KVM_MAX_VCPUS 1
54 #define KVM_MEMORY_SLOTS 4
55 #define KVM_NUM_MMU_PAGES 256
56 #define KVM_MIN_FREE_MMU_PAGES 5
57 #define KVM_REFILL_PAGES 25
58 #define KVM_MAX_CPUID_ENTRIES 40
60 #define FX_IMAGE_SIZE 512
61 #define FX_IMAGE_ALIGN 16
62 #define FX_BUF_SIZE (2 * FX_IMAGE_SIZE + FX_IMAGE_ALIGN)
72 #define SELECTOR_TI_MASK (1 << 2)
73 #define SELECTOR_RPL_MASK 0x03
80 * gva - guest virtual address
81 * gpa - guest physical address
82 * gfn - guest frame number
83 * hva - host virtual address
84 * hpa - host physical address
85 * hfn - host frame number
88 typedef unsigned long gva_t;
90 typedef unsigned long gfn_t;
92 typedef unsigned long hva_t;
94 typedef unsigned long hfn_t;
96 #define NR_PTE_CHAIN_ENTRIES 5
98 struct kvm_pte_chain {
99 u64 *parent_ptes[NR_PTE_CHAIN_ENTRIES];
100 struct hlist_node link;
104 * kvm_mmu_page_role, below, is defined as:
106 * bits 0:3 - total guest paging levels (2-4, or zero for real mode)
107 * bits 4:7 - page table level for this shadow (1-4)
108 * bits 8:9 - page table quadrant for 2-level guests
109 * bit 16 - "metaphysical" - gfn is not a real page (huge page/real mode)
111 union kvm_mmu_page_role {
114 unsigned glevels : 4;
116 unsigned quadrant : 2;
117 unsigned pad_for_nice_hex_output : 6;
118 unsigned metaphysical : 1;
122 struct kvm_mmu_page {
123 struct list_head link;
124 struct hlist_node hash_link;
127 * The following two entries are used to key the shadow page in the
131 union kvm_mmu_page_role role;
134 unsigned long slot_bitmap; /* One bit set per slot which has memory
135 * in this shadow page.
137 int global; /* Set if all ptes in this page are global */
138 int multimapped; /* More than one parent_pte? */
139 int root_count; /* Currently serving as active root */
141 u64 *parent_pte; /* !multimapped */
142 struct hlist_head parent_ptes; /* multimapped, kvm_pte_chain */
152 #define vmx_msr_entry kvm_msr_entry
157 * x86 supports 3 paging modes (4-level 64-bit, 3-level 64-bit, and 2-level
158 * 32-bit). The kvm_mmu structure abstracts the details of the current mmu
162 void (*new_cr3)(struct kvm_vcpu *vcpu);
163 int (*page_fault)(struct kvm_vcpu *vcpu, gva_t gva, u32 err);
164 void (*free)(struct kvm_vcpu *vcpu);
165 gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t gva);
168 int shadow_root_level;
173 #define KVM_NR_MEM_OBJS 20
175 struct kvm_mmu_memory_cache {
177 void *objects[KVM_NR_MEM_OBJS];
181 * We don't want allocation failures within the mmu code, so we preallocate
182 * enough memory for a single page fault in a cache.
184 struct kvm_guest_debug {
227 struct vcpu_svm *svm;
233 int interrupt_window_open;
234 unsigned long irq_summary; /* bit vector: 1 per word in irq_pending */
235 #define NR_IRQ_WORDS KVM_IRQ_BITMAP_SIZE(unsigned long)
236 unsigned long irq_pending[NR_IRQ_WORDS];
237 unsigned long regs[NR_VCPU_REGS]; /* for rsp: vcpu_load_rsp_rip() */
238 unsigned long rip; /* needs vcpu_load_rsp_rip() */
243 gpa_t para_state_gpa;
244 struct page *para_state_page;
248 u64 pdptrs[4]; /* pae */
251 u64 ia32_misc_enable_msr;
253 struct vmx_msr_entry *guest_msrs;
254 struct vmx_msr_entry *host_msrs;
256 struct list_head free_pages;
257 struct kvm_mmu_page page_header_buf[KVM_NUM_MMU_PAGES];
260 struct kvm_mmu_memory_cache mmu_pte_chain_cache;
261 struct kvm_mmu_memory_cache mmu_rmap_desc_cache;
263 gfn_t last_pt_write_gfn;
264 int last_pt_write_count;
266 struct kvm_guest_debug guest_debug;
268 char fx_buf[FX_BUF_SIZE];
270 char *guest_fx_image;
273 int mmio_read_completed;
276 unsigned char mmio_data[8];
277 gpa_t mmio_phys_addr;
286 struct kvm_save_segment {
291 } tr, es, ds, fs, gs;
295 struct kvm_cpuid_entry cpuid_entries[KVM_MAX_CPUID_ENTRIES];
298 struct kvm_memory_slot {
300 unsigned long npages;
302 struct page **phys_mem;
303 unsigned long *dirty_bitmap;
307 spinlock_t lock; /* protects everything except vcpus */
309 struct kvm_memory_slot memslots[KVM_MEMORY_SLOTS];
311 * Hash table of struct kvm_mmu_page.
313 struct list_head active_mmu_pages;
314 int n_free_mmu_pages;
315 struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
316 struct kvm_vcpu vcpus[KVM_MAX_VCPUS];
317 int memory_config_version;
319 unsigned long rmap_overflow;
320 struct list_head vm_list;
334 u32 irq_window_exits;
336 u32 request_irq_exits;
340 struct descriptor_table {
343 } __attribute__((packed));
345 struct kvm_arch_ops {
346 int (*cpu_has_kvm_support)(void); /* __init */
347 int (*disabled_by_bios)(void); /* __init */
348 void (*hardware_enable)(void *dummy); /* __init */
349 void (*hardware_disable)(void *dummy);
350 int (*hardware_setup)(void); /* __init */
351 void (*hardware_unsetup)(void); /* __exit */
353 int (*vcpu_create)(struct kvm_vcpu *vcpu);
354 void (*vcpu_free)(struct kvm_vcpu *vcpu);
356 void (*vcpu_load)(struct kvm_vcpu *vcpu);
357 void (*vcpu_put)(struct kvm_vcpu *vcpu);
358 void (*vcpu_decache)(struct kvm_vcpu *vcpu);
360 int (*set_guest_debug)(struct kvm_vcpu *vcpu,
361 struct kvm_debug_guest *dbg);
362 int (*get_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata);
363 int (*set_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 data);
364 u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
365 void (*get_segment)(struct kvm_vcpu *vcpu,
366 struct kvm_segment *var, int seg);
367 void (*set_segment)(struct kvm_vcpu *vcpu,
368 struct kvm_segment *var, int seg);
369 void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
370 void (*decache_cr0_cr4_guest_bits)(struct kvm_vcpu *vcpu);
371 void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
372 void (*set_cr0_no_modeswitch)(struct kvm_vcpu *vcpu,
374 void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
375 void (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
376 void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
377 void (*get_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
378 void (*set_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
379 void (*get_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
380 void (*set_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
381 unsigned long (*get_dr)(struct kvm_vcpu *vcpu, int dr);
382 void (*set_dr)(struct kvm_vcpu *vcpu, int dr, unsigned long value,
384 void (*cache_regs)(struct kvm_vcpu *vcpu);
385 void (*decache_regs)(struct kvm_vcpu *vcpu);
386 unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
387 void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
389 void (*invlpg)(struct kvm_vcpu *vcpu, gva_t addr);
390 void (*tlb_flush)(struct kvm_vcpu *vcpu);
391 void (*inject_page_fault)(struct kvm_vcpu *vcpu,
392 unsigned long addr, u32 err_code);
394 void (*inject_gp)(struct kvm_vcpu *vcpu, unsigned err_code);
396 int (*run)(struct kvm_vcpu *vcpu, struct kvm_run *run);
397 int (*vcpu_setup)(struct kvm_vcpu *vcpu);
398 void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
399 void (*patch_hypercall)(struct kvm_vcpu *vcpu,
400 unsigned char *hypercall_addr);
403 extern struct kvm_stat kvm_stat;
404 extern struct kvm_arch_ops *kvm_arch_ops;
406 #define kvm_printf(kvm, fmt ...) printk(KERN_DEBUG fmt)
407 #define vcpu_printf(vcpu, fmt...) kvm_printf(vcpu->kvm, fmt)
409 int kvm_init_arch(struct kvm_arch_ops *ops, struct module *module);
410 void kvm_exit_arch(void);
412 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
413 int kvm_mmu_create(struct kvm_vcpu *vcpu);
414 int kvm_mmu_setup(struct kvm_vcpu *vcpu);
416 int kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
417 void kvm_mmu_slot_remove_write_access(struct kvm_vcpu *vcpu, int slot);
419 hpa_t gpa_to_hpa(struct kvm_vcpu *vcpu, gpa_t gpa);
420 #define HPA_MSB ((sizeof(hpa_t) * 8) - 1)
421 #define HPA_ERR_MASK ((hpa_t)1 << HPA_MSB)
422 static inline int is_error_hpa(hpa_t hpa) { return hpa >> HPA_MSB; }
423 hpa_t gva_to_hpa(struct kvm_vcpu *vcpu, gva_t gva);
425 void kvm_emulator_want_group7_invlpg(void);
427 extern hpa_t bad_page_address;
429 static inline struct page *gfn_to_page(struct kvm_memory_slot *slot, gfn_t gfn)
431 return slot->phys_mem[gfn - slot->base_gfn];
434 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
435 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
437 enum emulation_result {
438 EMULATE_DONE, /* no further processing */
439 EMULATE_DO_MMIO, /* kvm_run filled with mmio request */
440 EMULATE_FAIL, /* can't emulate this instruction */
443 int emulate_instruction(struct kvm_vcpu *vcpu, struct kvm_run *run,
444 unsigned long cr2, u16 error_code);
445 void realmode_lgdt(struct kvm_vcpu *vcpu, u16 size, unsigned long address);
446 void realmode_lidt(struct kvm_vcpu *vcpu, u16 size, unsigned long address);
447 void realmode_lmsw(struct kvm_vcpu *vcpu, unsigned long msw,
448 unsigned long *rflags);
450 unsigned long realmode_get_cr(struct kvm_vcpu *vcpu, int cr);
451 void realmode_set_cr(struct kvm_vcpu *vcpu, int cr, unsigned long value,
452 unsigned long *rflags);
454 struct x86_emulate_ctxt;
456 void kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
457 int emulate_invlpg(struct kvm_vcpu *vcpu, gva_t address);
458 int emulate_clts(struct kvm_vcpu *vcpu);
459 int emulator_get_dr(struct x86_emulate_ctxt* ctxt, int dr,
460 unsigned long *dest);
461 int emulator_set_dr(struct x86_emulate_ctxt *ctxt, int dr,
462 unsigned long value);
464 void set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
465 void set_cr3(struct kvm_vcpu *vcpu, unsigned long cr0);
466 void set_cr4(struct kvm_vcpu *vcpu, unsigned long cr0);
467 void set_cr8(struct kvm_vcpu *vcpu, unsigned long cr0);
468 void lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
470 int kvm_get_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata);
471 int kvm_set_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 data);
473 void fx_init(struct kvm_vcpu *vcpu);
475 void load_msrs(struct vmx_msr_entry *e, int n);
476 void save_msrs(struct vmx_msr_entry *e, int n);
477 void kvm_resched(struct kvm_vcpu *vcpu);
479 int kvm_read_guest(struct kvm_vcpu *vcpu,
484 int kvm_write_guest(struct kvm_vcpu *vcpu,
489 unsigned long segment_base(u16 selector);
491 void kvm_mmu_pre_write(struct kvm_vcpu *vcpu, gpa_t gpa, int bytes);
492 void kvm_mmu_post_write(struct kvm_vcpu *vcpu, gpa_t gpa, int bytes);
493 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva);
494 void kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
496 int kvm_hypercall(struct kvm_vcpu *vcpu, struct kvm_run *run);
498 static inline int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t gva,
501 if (unlikely(vcpu->kvm->n_free_mmu_pages < KVM_MIN_FREE_MMU_PAGES))
502 kvm_mmu_free_some_pages(vcpu);
503 return vcpu->mmu.page_fault(vcpu, gva, error_code);
506 static inline struct page *_gfn_to_page(struct kvm *kvm, gfn_t gfn)
508 struct kvm_memory_slot *slot = gfn_to_memslot(kvm, gfn);
509 return (slot) ? slot->phys_mem[gfn - slot->base_gfn] : NULL;
512 static inline int is_long_mode(struct kvm_vcpu *vcpu)
515 return vcpu->shadow_efer & EFER_LME;
521 static inline int is_pae(struct kvm_vcpu *vcpu)
523 return vcpu->cr4 & CR4_PAE_MASK;
526 static inline int is_pse(struct kvm_vcpu *vcpu)
528 return vcpu->cr4 & CR4_PSE_MASK;
531 static inline int is_paging(struct kvm_vcpu *vcpu)
533 return vcpu->cr0 & CR0_PG_MASK;
536 static inline int memslot_id(struct kvm *kvm, struct kvm_memory_slot *slot)
538 return slot - kvm->memslots;
541 static inline struct kvm_mmu_page *page_header(hpa_t shadow_page)
543 struct page *page = pfn_to_page(shadow_page >> PAGE_SHIFT);
545 return (struct kvm_mmu_page *)page_private(page);
548 static inline u16 read_fs(void)
551 asm ("mov %%fs, %0" : "=g"(seg));
555 static inline u16 read_gs(void)
558 asm ("mov %%gs, %0" : "=g"(seg));
562 static inline u16 read_ldt(void)
565 asm ("sldt %0" : "=g"(ldt));
569 static inline void load_fs(u16 sel)
571 asm ("mov %0, %%fs" : : "rm"(sel));
574 static inline void load_gs(u16 sel)
576 asm ("mov %0, %%gs" : : "rm"(sel));
580 static inline void load_ldt(u16 sel)
582 asm ("lldt %0" : : "rm"(sel));
586 static inline void get_idt(struct descriptor_table *table)
588 asm ("sidt %0" : "=m"(*table));
591 static inline void get_gdt(struct descriptor_table *table)
593 asm ("sgdt %0" : "=m"(*table));
596 static inline unsigned long read_tr_base(void)
599 asm ("str %0" : "=g"(tr));
600 return segment_base(tr);
604 static inline unsigned long read_msr(unsigned long msr)
613 static inline void fx_save(void *image)
615 asm ("fxsave (%0)":: "r" (image));
618 static inline void fx_restore(void *image)
620 asm ("fxrstor (%0)":: "r" (image));
623 static inline void fpu_init(void)
628 static inline u32 get_rdx_init_val(void)
630 return 0x600; /* P6 family */
633 #define ASM_VMX_VMCLEAR_RAX ".byte 0x66, 0x0f, 0xc7, 0x30"
634 #define ASM_VMX_VMLAUNCH ".byte 0x0f, 0x01, 0xc2"
635 #define ASM_VMX_VMRESUME ".byte 0x0f, 0x01, 0xc3"
636 #define ASM_VMX_VMPTRLD_RAX ".byte 0x0f, 0xc7, 0x30"
637 #define ASM_VMX_VMREAD_RDX_RAX ".byte 0x0f, 0x78, 0xd0"
638 #define ASM_VMX_VMWRITE_RAX_RDX ".byte 0x0f, 0x79, 0xd0"
639 #define ASM_VMX_VMWRITE_RSP_RDX ".byte 0x0f, 0x79, 0xd4"
640 #define ASM_VMX_VMXOFF ".byte 0x0f, 0x01, 0xc4"
641 #define ASM_VMX_VMXON_RAX ".byte 0xf3, 0x0f, 0xc7, 0x30"
643 #define MSR_IA32_TIME_STAMP_COUNTER 0x010
645 #define TSS_IOPB_BASE_OFFSET 0x66
646 #define TSS_BASE_SIZE 0x68
647 #define TSS_IOPB_SIZE (65536 / 8)
648 #define TSS_REDIRECTION_SIZE (256 / 8)
649 #define RMODE_TSS_SIZE (TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)