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
77 #define KVM_PIO_PAGE_OFFSET 1
82 * gva - guest virtual address
83 * gpa - guest physical address
84 * gfn - guest frame number
85 * hva - host virtual address
86 * hpa - host physical address
87 * hfn - host frame number
90 typedef unsigned long gva_t;
92 typedef unsigned long gfn_t;
94 typedef unsigned long hva_t;
96 typedef unsigned long hfn_t;
98 #define NR_PTE_CHAIN_ENTRIES 5
100 struct kvm_pte_chain {
101 u64 *parent_ptes[NR_PTE_CHAIN_ENTRIES];
102 struct hlist_node link;
106 * kvm_mmu_page_role, below, is defined as:
108 * bits 0:3 - total guest paging levels (2-4, or zero for real mode)
109 * bits 4:7 - page table level for this shadow (1-4)
110 * bits 8:9 - page table quadrant for 2-level guests
111 * bit 16 - "metaphysical" - gfn is not a real page (huge page/real mode)
112 * bits 17:18 - "access" - the user and writable bits of a huge page pde
114 union kvm_mmu_page_role {
117 unsigned glevels : 4;
119 unsigned quadrant : 2;
120 unsigned pad_for_nice_hex_output : 6;
121 unsigned metaphysical : 1;
122 unsigned hugepage_access : 2;
126 struct kvm_mmu_page {
127 struct list_head link;
128 struct hlist_node hash_link;
131 * The following two entries are used to key the shadow page in the
135 union kvm_mmu_page_role role;
138 unsigned long slot_bitmap; /* One bit set per slot which has memory
139 * in this shadow page.
141 int multimapped; /* More than one parent_pte? */
142 int root_count; /* Currently serving as active root */
144 u64 *parent_pte; /* !multimapped */
145 struct hlist_head parent_ptes; /* multimapped, kvm_pte_chain */
155 #define vmx_msr_entry kvm_msr_entry
160 * x86 supports 3 paging modes (4-level 64-bit, 3-level 64-bit, and 2-level
161 * 32-bit). The kvm_mmu structure abstracts the details of the current mmu
165 void (*new_cr3)(struct kvm_vcpu *vcpu);
166 int (*page_fault)(struct kvm_vcpu *vcpu, gva_t gva, u32 err);
167 void (*free)(struct kvm_vcpu *vcpu);
168 gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t gva);
171 int shadow_root_level;
176 #define KVM_NR_MEM_OBJS 20
178 struct kvm_mmu_memory_cache {
180 void *objects[KVM_NR_MEM_OBJS];
184 * We don't want allocation failures within the mmu code, so we preallocate
185 * enough memory for a single page fault in a cache.
187 struct kvm_guest_debug {
226 struct kvm_pio_request {
229 struct page *guest_pages[2];
230 unsigned guest_page_offset;
242 struct vcpu_svm *svm;
249 int interrupt_window_open;
250 unsigned long irq_summary; /* bit vector: 1 per word in irq_pending */
251 #define NR_IRQ_WORDS KVM_IRQ_BITMAP_SIZE(unsigned long)
252 unsigned long irq_pending[NR_IRQ_WORDS];
253 unsigned long regs[NR_VCPU_REGS]; /* for rsp: vcpu_load_rsp_rip() */
254 unsigned long rip; /* needs vcpu_load_rsp_rip() */
259 gpa_t para_state_gpa;
260 struct page *para_state_page;
264 u64 pdptrs[4]; /* pae */
267 u64 ia32_misc_enable_msr;
269 struct vmx_msr_entry *guest_msrs;
270 struct vmx_msr_entry *host_msrs;
272 struct list_head free_pages;
273 struct kvm_mmu_page page_header_buf[KVM_NUM_MMU_PAGES];
276 struct kvm_mmu_memory_cache mmu_pte_chain_cache;
277 struct kvm_mmu_memory_cache mmu_rmap_desc_cache;
279 gfn_t last_pt_write_gfn;
280 int last_pt_write_count;
282 struct kvm_guest_debug guest_debug;
284 char fx_buf[FX_BUF_SIZE];
286 char *guest_fx_image;
289 int mmio_read_completed;
292 unsigned char mmio_data[8];
293 gpa_t mmio_phys_addr;
294 struct kvm_pio_request pio;
303 struct kvm_save_segment {
308 } tr, es, ds, fs, gs;
312 struct kvm_cpuid_entry cpuid_entries[KVM_MAX_CPUID_ENTRIES];
315 struct kvm_memory_slot {
317 unsigned long npages;
319 struct page **phys_mem;
320 unsigned long *dirty_bitmap;
324 spinlock_t lock; /* protects everything except vcpus */
326 struct kvm_memory_slot memslots[KVM_MEMORY_SLOTS];
328 * Hash table of struct kvm_mmu_page.
330 struct list_head active_mmu_pages;
331 int n_free_mmu_pages;
332 struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
333 struct kvm_vcpu vcpus[KVM_MAX_VCPUS];
334 int memory_config_version;
336 unsigned long rmap_overflow;
337 struct list_head vm_list;
351 u32 irq_window_exits;
353 u32 request_irq_exits;
357 struct descriptor_table {
360 } __attribute__((packed));
362 struct kvm_arch_ops {
363 int (*cpu_has_kvm_support)(void); /* __init */
364 int (*disabled_by_bios)(void); /* __init */
365 void (*hardware_enable)(void *dummy); /* __init */
366 void (*hardware_disable)(void *dummy);
367 int (*hardware_setup)(void); /* __init */
368 void (*hardware_unsetup)(void); /* __exit */
370 int (*vcpu_create)(struct kvm_vcpu *vcpu);
371 void (*vcpu_free)(struct kvm_vcpu *vcpu);
373 void (*vcpu_load)(struct kvm_vcpu *vcpu);
374 void (*vcpu_put)(struct kvm_vcpu *vcpu);
375 void (*vcpu_decache)(struct kvm_vcpu *vcpu);
377 int (*set_guest_debug)(struct kvm_vcpu *vcpu,
378 struct kvm_debug_guest *dbg);
379 int (*get_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata);
380 int (*set_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 data);
381 u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
382 void (*get_segment)(struct kvm_vcpu *vcpu,
383 struct kvm_segment *var, int seg);
384 void (*set_segment)(struct kvm_vcpu *vcpu,
385 struct kvm_segment *var, int seg);
386 void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
387 void (*decache_cr0_cr4_guest_bits)(struct kvm_vcpu *vcpu);
388 void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
389 void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
390 void (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
391 void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
392 void (*get_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
393 void (*set_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
394 void (*get_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
395 void (*set_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
396 unsigned long (*get_dr)(struct kvm_vcpu *vcpu, int dr);
397 void (*set_dr)(struct kvm_vcpu *vcpu, int dr, unsigned long value,
399 void (*cache_regs)(struct kvm_vcpu *vcpu);
400 void (*decache_regs)(struct kvm_vcpu *vcpu);
401 unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
402 void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
404 void (*invlpg)(struct kvm_vcpu *vcpu, gva_t addr);
405 void (*tlb_flush)(struct kvm_vcpu *vcpu);
406 void (*inject_page_fault)(struct kvm_vcpu *vcpu,
407 unsigned long addr, u32 err_code);
409 void (*inject_gp)(struct kvm_vcpu *vcpu, unsigned err_code);
411 int (*run)(struct kvm_vcpu *vcpu, struct kvm_run *run);
412 int (*vcpu_setup)(struct kvm_vcpu *vcpu);
413 void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
414 void (*patch_hypercall)(struct kvm_vcpu *vcpu,
415 unsigned char *hypercall_addr);
418 extern struct kvm_stat kvm_stat;
419 extern struct kvm_arch_ops *kvm_arch_ops;
421 #define kvm_printf(kvm, fmt ...) printk(KERN_DEBUG fmt)
422 #define vcpu_printf(vcpu, fmt...) kvm_printf(vcpu->kvm, fmt)
424 int kvm_init_arch(struct kvm_arch_ops *ops, struct module *module);
425 void kvm_exit_arch(void);
427 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
428 int kvm_mmu_create(struct kvm_vcpu *vcpu);
429 int kvm_mmu_setup(struct kvm_vcpu *vcpu);
431 int kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
432 void kvm_mmu_slot_remove_write_access(struct kvm_vcpu *vcpu, int slot);
433 void kvm_mmu_zap_all(struct kvm_vcpu *vcpu);
435 hpa_t gpa_to_hpa(struct kvm_vcpu *vcpu, gpa_t gpa);
436 #define HPA_MSB ((sizeof(hpa_t) * 8) - 1)
437 #define HPA_ERR_MASK ((hpa_t)1 << HPA_MSB)
438 static inline int is_error_hpa(hpa_t hpa) { return hpa >> HPA_MSB; }
439 hpa_t gva_to_hpa(struct kvm_vcpu *vcpu, gva_t gva);
440 struct page *gva_to_page(struct kvm_vcpu *vcpu, gva_t gva);
442 void kvm_emulator_want_group7_invlpg(void);
444 extern hpa_t bad_page_address;
446 static inline struct page *gfn_to_page(struct kvm_memory_slot *slot, gfn_t gfn)
448 return slot->phys_mem[gfn - slot->base_gfn];
451 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
452 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
454 enum emulation_result {
455 EMULATE_DONE, /* no further processing */
456 EMULATE_DO_MMIO, /* kvm_run filled with mmio request */
457 EMULATE_FAIL, /* can't emulate this instruction */
460 int emulate_instruction(struct kvm_vcpu *vcpu, struct kvm_run *run,
461 unsigned long cr2, u16 error_code);
462 void realmode_lgdt(struct kvm_vcpu *vcpu, u16 size, unsigned long address);
463 void realmode_lidt(struct kvm_vcpu *vcpu, u16 size, unsigned long address);
464 void realmode_lmsw(struct kvm_vcpu *vcpu, unsigned long msw,
465 unsigned long *rflags);
467 unsigned long realmode_get_cr(struct kvm_vcpu *vcpu, int cr);
468 void realmode_set_cr(struct kvm_vcpu *vcpu, int cr, unsigned long value,
469 unsigned long *rflags);
471 struct x86_emulate_ctxt;
473 int kvm_setup_pio(struct kvm_vcpu *vcpu, struct kvm_run *run, int in,
474 int size, unsigned long count, int string, int down,
475 gva_t address, int rep, unsigned port);
476 void kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
477 int emulate_invlpg(struct kvm_vcpu *vcpu, gva_t address);
478 int emulate_clts(struct kvm_vcpu *vcpu);
479 int emulator_get_dr(struct x86_emulate_ctxt* ctxt, int dr,
480 unsigned long *dest);
481 int emulator_set_dr(struct x86_emulate_ctxt *ctxt, int dr,
482 unsigned long value);
484 void set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
485 void set_cr3(struct kvm_vcpu *vcpu, unsigned long cr0);
486 void set_cr4(struct kvm_vcpu *vcpu, unsigned long cr0);
487 void set_cr8(struct kvm_vcpu *vcpu, unsigned long cr0);
488 void lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
490 int kvm_get_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata);
491 int kvm_set_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 data);
493 void fx_init(struct kvm_vcpu *vcpu);
495 void load_msrs(struct vmx_msr_entry *e, int n);
496 void save_msrs(struct vmx_msr_entry *e, int n);
497 void kvm_resched(struct kvm_vcpu *vcpu);
499 int kvm_read_guest(struct kvm_vcpu *vcpu,
504 int kvm_write_guest(struct kvm_vcpu *vcpu,
509 unsigned long segment_base(u16 selector);
511 void kvm_mmu_pre_write(struct kvm_vcpu *vcpu, gpa_t gpa, int bytes);
512 void kvm_mmu_post_write(struct kvm_vcpu *vcpu, gpa_t gpa, int bytes);
513 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva);
514 void kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
516 int kvm_hypercall(struct kvm_vcpu *vcpu, struct kvm_run *run);
518 static inline int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t gva,
521 if (unlikely(vcpu->kvm->n_free_mmu_pages < KVM_MIN_FREE_MMU_PAGES))
522 kvm_mmu_free_some_pages(vcpu);
523 return vcpu->mmu.page_fault(vcpu, gva, error_code);
526 static inline struct page *_gfn_to_page(struct kvm *kvm, gfn_t gfn)
528 struct kvm_memory_slot *slot = gfn_to_memslot(kvm, gfn);
529 return (slot) ? slot->phys_mem[gfn - slot->base_gfn] : NULL;
532 static inline int is_long_mode(struct kvm_vcpu *vcpu)
535 return vcpu->shadow_efer & EFER_LME;
541 static inline int is_pae(struct kvm_vcpu *vcpu)
543 return vcpu->cr4 & CR4_PAE_MASK;
546 static inline int is_pse(struct kvm_vcpu *vcpu)
548 return vcpu->cr4 & CR4_PSE_MASK;
551 static inline int is_paging(struct kvm_vcpu *vcpu)
553 return vcpu->cr0 & CR0_PG_MASK;
556 static inline int memslot_id(struct kvm *kvm, struct kvm_memory_slot *slot)
558 return slot - kvm->memslots;
561 static inline struct kvm_mmu_page *page_header(hpa_t shadow_page)
563 struct page *page = pfn_to_page(shadow_page >> PAGE_SHIFT);
565 return (struct kvm_mmu_page *)page_private(page);
568 static inline u16 read_fs(void)
571 asm ("mov %%fs, %0" : "=g"(seg));
575 static inline u16 read_gs(void)
578 asm ("mov %%gs, %0" : "=g"(seg));
582 static inline u16 read_ldt(void)
585 asm ("sldt %0" : "=g"(ldt));
589 static inline void load_fs(u16 sel)
591 asm ("mov %0, %%fs" : : "rm"(sel));
594 static inline void load_gs(u16 sel)
596 asm ("mov %0, %%gs" : : "rm"(sel));
600 static inline void load_ldt(u16 sel)
602 asm ("lldt %0" : : "rm"(sel));
606 static inline void get_idt(struct descriptor_table *table)
608 asm ("sidt %0" : "=m"(*table));
611 static inline void get_gdt(struct descriptor_table *table)
613 asm ("sgdt %0" : "=m"(*table));
616 static inline unsigned long read_tr_base(void)
619 asm ("str %0" : "=g"(tr));
620 return segment_base(tr);
624 static inline unsigned long read_msr(unsigned long msr)
633 static inline void fx_save(void *image)
635 asm ("fxsave (%0)":: "r" (image));
638 static inline void fx_restore(void *image)
640 asm ("fxrstor (%0)":: "r" (image));
643 static inline void fpu_init(void)
648 static inline u32 get_rdx_init_val(void)
650 return 0x600; /* P6 family */
653 #define ASM_VMX_VMCLEAR_RAX ".byte 0x66, 0x0f, 0xc7, 0x30"
654 #define ASM_VMX_VMLAUNCH ".byte 0x0f, 0x01, 0xc2"
655 #define ASM_VMX_VMRESUME ".byte 0x0f, 0x01, 0xc3"
656 #define ASM_VMX_VMPTRLD_RAX ".byte 0x0f, 0xc7, 0x30"
657 #define ASM_VMX_VMREAD_RDX_RAX ".byte 0x0f, 0x78, 0xd0"
658 #define ASM_VMX_VMWRITE_RAX_RDX ".byte 0x0f, 0x79, 0xd0"
659 #define ASM_VMX_VMWRITE_RSP_RDX ".byte 0x0f, 0x79, 0xd4"
660 #define ASM_VMX_VMXOFF ".byte 0x0f, 0x01, 0xc4"
661 #define ASM_VMX_VMXON_RAX ".byte 0xf3, 0x0f, 0xc7, 0x30"
663 #define MSR_IA32_TIME_STAMP_COUNTER 0x010
665 #define TSS_IOPB_BASE_OFFSET 0x66
666 #define TSS_BASE_SIZE 0x68
667 #define TSS_IOPB_SIZE (65536 / 8)
668 #define TSS_REDIRECTION_SIZE (256 / 8)
669 #define RMODE_TSS_SIZE (TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)