2 * Kernel-based Virtual Machine driver for Linux
4 * This header defines architecture specific interfaces, x86 version
6 * This work is licensed under the terms of the GNU GPL, version 2. See
7 * the COPYING file in the top-level directory.
11 #ifndef ASM_KVM_HOST_H
12 #define ASM_KVM_HOST_H
14 #include <linux/types.h>
17 #include <linux/kvm.h>
18 #include <linux/kvm_para.h>
19 #include <linux/kvm_types.h>
23 #define KVM_MAX_VCPUS 16
24 #define KVM_MEMORY_SLOTS 32
25 /* memory slots that does not exposed to userspace */
26 #define KVM_PRIVATE_MEM_SLOTS 4
28 #define KVM_PIO_PAGE_OFFSET 1
30 #define CR3_PAE_RESERVED_BITS ((X86_CR3_PWT | X86_CR3_PCD) - 1)
31 #define CR3_NONPAE_RESERVED_BITS ((PAGE_SIZE-1) & ~(X86_CR3_PWT | X86_CR3_PCD))
32 #define CR3_L_MODE_RESERVED_BITS (CR3_NONPAE_RESERVED_BITS | \
33 0xFFFFFF0000000000ULL)
35 #define KVM_GUEST_CR0_MASK \
36 (X86_CR0_PG | X86_CR0_PE | X86_CR0_WP | X86_CR0_NE \
37 | X86_CR0_NW | X86_CR0_CD)
38 #define KVM_VM_CR0_ALWAYS_ON \
39 (X86_CR0_PG | X86_CR0_PE | X86_CR0_WP | X86_CR0_NE | X86_CR0_TS \
41 #define KVM_GUEST_CR4_MASK \
42 (X86_CR4_VME | X86_CR4_PSE | X86_CR4_PAE | X86_CR4_PGE | X86_CR4_VMXE)
43 #define KVM_PMODE_VM_CR4_ALWAYS_ON (X86_CR4_PAE | X86_CR4_VMXE)
44 #define KVM_RMODE_VM_CR4_ALWAYS_ON (X86_CR4_VME | X86_CR4_PAE | X86_CR4_VMXE)
46 #define INVALID_PAGE (~(hpa_t)0)
47 #define UNMAPPED_GVA (~(gpa_t)0)
49 /* shadow tables are PAE even on non-PAE hosts */
50 #define KVM_HPAGE_SHIFT 21
51 #define KVM_HPAGE_SIZE (1UL << KVM_HPAGE_SHIFT)
52 #define KVM_HPAGE_MASK (~(KVM_HPAGE_SIZE - 1))
54 #define KVM_PAGES_PER_HPAGE (KVM_HPAGE_SIZE / PAGE_SIZE)
66 #define SELECTOR_TI_MASK (1 << 2)
67 #define SELECTOR_RPL_MASK 0x03
71 #define KVM_ALIAS_SLOTS 4
73 #define KVM_PERMILLE_MMU_PAGES 20
74 #define KVM_MIN_ALLOC_MMU_PAGES 64
75 #define KVM_MMU_HASH_SHIFT 10
76 #define KVM_NUM_MMU_PAGES (1 << KVM_MMU_HASH_SHIFT)
77 #define KVM_MIN_FREE_MMU_PAGES 5
78 #define KVM_REFILL_PAGES 25
79 #define KVM_MAX_CPUID_ENTRIES 40
81 extern spinlock_t kvm_lock;
82 extern struct list_head vm_list;
120 #include <asm/kvm_x86_emulate.h>
122 #define KVM_NR_MEM_OBJS 40
124 struct kvm_guest_debug {
131 * We don't want allocation failures within the mmu code, so we preallocate
132 * enough memory for a single page fault in a cache.
134 struct kvm_mmu_memory_cache {
136 void *objects[KVM_NR_MEM_OBJS];
139 #define NR_PTE_CHAIN_ENTRIES 5
141 struct kvm_pte_chain {
142 u64 *parent_ptes[NR_PTE_CHAIN_ENTRIES];
143 struct hlist_node link;
147 * kvm_mmu_page_role, below, is defined as:
149 * bits 0:3 - total guest paging levels (2-4, or zero for real mode)
150 * bits 4:7 - page table level for this shadow (1-4)
151 * bits 8:9 - page table quadrant for 2-level guests
152 * bit 16 - "metaphysical" - gfn is not a real page (huge page/real mode)
153 * bits 17:19 - common access permissions for all ptes in this shadow page
155 union kvm_mmu_page_role {
161 unsigned pad_for_nice_hex_output:6;
162 unsigned metaphysical:1;
168 struct kvm_mmu_page {
169 struct list_head link;
170 struct hlist_node hash_link;
173 * The following two entries are used to key the shadow page in the
177 union kvm_mmu_page_role role;
180 /* hold the gfn of each spte inside spt */
182 unsigned long slot_bitmap; /* One bit set per slot which has memory
183 * in this shadow page.
185 int multimapped; /* More than one parent_pte? */
186 int root_count; /* Currently serving as active root */
188 u64 *parent_pte; /* !multimapped */
189 struct hlist_head parent_ptes; /* multimapped, kvm_pte_chain */
194 * x86 supports 3 paging modes (4-level 64-bit, 3-level 64-bit, and 2-level
195 * 32-bit). The kvm_mmu structure abstracts the details of the current mmu
199 void (*new_cr3)(struct kvm_vcpu *vcpu);
200 int (*page_fault)(struct kvm_vcpu *vcpu, gva_t gva, u32 err);
201 void (*free)(struct kvm_vcpu *vcpu);
202 gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t gva);
203 void (*prefetch_page)(struct kvm_vcpu *vcpu,
204 struct kvm_mmu_page *page);
207 int shadow_root_level;
212 struct kvm_vcpu_arch {
214 int interrupt_window_open;
215 unsigned long irq_summary; /* bit vector: 1 per word in irq_pending */
216 DECLARE_BITMAP(irq_pending, KVM_NR_INTERRUPTS);
217 unsigned long regs[NR_VCPU_REGS]; /* for rsp: vcpu_load_rsp_rip() */
218 unsigned long rip; /* needs vcpu_load_rsp_rip() */
225 u64 pdptrs[4]; /* pae */
228 struct kvm_lapic *apic; /* kernel irqchip context */
229 #define VCPU_MP_STATE_RUNNABLE 0
230 #define VCPU_MP_STATE_UNINITIALIZED 1
231 #define VCPU_MP_STATE_INIT_RECEIVED 2
232 #define VCPU_MP_STATE_SIPI_RECEIVED 3
233 #define VCPU_MP_STATE_HALTED 4
236 u64 ia32_misc_enable_msr;
237 bool tpr_access_reporting;
241 struct kvm_mmu_memory_cache mmu_pte_chain_cache;
242 struct kvm_mmu_memory_cache mmu_rmap_desc_cache;
243 struct kvm_mmu_memory_cache mmu_page_cache;
244 struct kvm_mmu_memory_cache mmu_page_header_cache;
246 gfn_t last_pt_write_gfn;
247 int last_pt_write_count;
248 u64 *last_pte_updated;
251 gfn_t gfn; /* presumed gfn during guest pte update */
252 struct page *page; /* page corresponding to that gfn */
256 struct i387_fxsave_struct host_fx_image;
257 struct i387_fxsave_struct guest_fx_image;
259 gva_t mmio_fault_cr2;
260 struct kvm_pio_request pio;
263 struct kvm_queued_exception {
273 struct kvm_save_segment {
278 } tr, es, ds, fs, gs;
280 int halt_request; /* real mode on Intel only */
283 struct kvm_cpuid_entry2 cpuid_entries[KVM_MAX_CPUID_ENTRIES];
284 /* emulate context */
286 struct x86_emulate_ctxt emulate_ctxt;
289 struct kvm_vcpu_time_info hv_clock;
290 unsigned int time_offset;
291 struct page *time_page;
294 struct kvm_mem_alias {
296 unsigned long npages;
302 struct kvm_mem_alias aliases[KVM_ALIAS_SLOTS];
304 unsigned int n_free_mmu_pages;
305 unsigned int n_requested_mmu_pages;
306 unsigned int n_alloc_mmu_pages;
307 struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
309 * Hash table of struct kvm_mmu_page.
311 struct list_head active_mmu_pages;
312 struct kvm_pic *vpic;
313 struct kvm_ioapic *vioapic;
314 struct kvm_pit *vpit;
316 int round_robin_prev_vcpu;
317 unsigned int tss_addr;
318 struct page *apic_access_page;
324 u32 mmu_shadow_zapped;
331 u32 remote_tlb_flush;
335 struct kvm_vcpu_stat {
345 u32 irq_window_exits;
348 u32 request_irq_exits;
350 u32 host_state_reload;
354 u32 insn_emulation_fail;
358 struct descriptor_table {
361 } __attribute__((packed));
364 int (*cpu_has_kvm_support)(void); /* __init */
365 int (*disabled_by_bios)(void); /* __init */
366 void (*hardware_enable)(void *dummy); /* __init */
367 void (*hardware_disable)(void *dummy);
368 void (*check_processor_compatibility)(void *rtn);
369 int (*hardware_setup)(void); /* __init */
370 void (*hardware_unsetup)(void); /* __exit */
371 bool (*cpu_has_accelerated_tpr)(void);
373 /* Create, but do not attach this VCPU */
374 struct kvm_vcpu *(*vcpu_create)(struct kvm *kvm, unsigned id);
375 void (*vcpu_free)(struct kvm_vcpu *vcpu);
376 int (*vcpu_reset)(struct kvm_vcpu *vcpu);
378 void (*prepare_guest_switch)(struct kvm_vcpu *vcpu);
379 void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
380 void (*vcpu_put)(struct kvm_vcpu *vcpu);
381 void (*vcpu_decache)(struct kvm_vcpu *vcpu);
383 int (*set_guest_debug)(struct kvm_vcpu *vcpu,
384 struct kvm_debug_guest *dbg);
385 void (*guest_debug_pre)(struct kvm_vcpu *vcpu);
386 int (*get_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata);
387 int (*set_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 data);
388 u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
389 void (*get_segment)(struct kvm_vcpu *vcpu,
390 struct kvm_segment *var, int seg);
391 int (*get_cpl)(struct kvm_vcpu *vcpu);
392 void (*set_segment)(struct kvm_vcpu *vcpu,
393 struct kvm_segment *var, int seg);
394 void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
395 void (*decache_cr4_guest_bits)(struct kvm_vcpu *vcpu);
396 void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
397 void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
398 void (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
399 void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
400 void (*get_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
401 void (*set_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
402 void (*get_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
403 void (*set_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
404 unsigned long (*get_dr)(struct kvm_vcpu *vcpu, int dr);
405 void (*set_dr)(struct kvm_vcpu *vcpu, int dr, unsigned long value,
407 void (*cache_regs)(struct kvm_vcpu *vcpu);
408 void (*decache_regs)(struct kvm_vcpu *vcpu);
409 unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
410 void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
412 void (*tlb_flush)(struct kvm_vcpu *vcpu);
414 void (*run)(struct kvm_vcpu *vcpu, struct kvm_run *run);
415 int (*handle_exit)(struct kvm_run *run, struct kvm_vcpu *vcpu);
416 void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
417 void (*patch_hypercall)(struct kvm_vcpu *vcpu,
418 unsigned char *hypercall_addr);
419 int (*get_irq)(struct kvm_vcpu *vcpu);
420 void (*set_irq)(struct kvm_vcpu *vcpu, int vec);
421 void (*queue_exception)(struct kvm_vcpu *vcpu, unsigned nr,
422 bool has_error_code, u32 error_code);
423 bool (*exception_injected)(struct kvm_vcpu *vcpu);
424 void (*inject_pending_irq)(struct kvm_vcpu *vcpu);
425 void (*inject_pending_vectors)(struct kvm_vcpu *vcpu,
426 struct kvm_run *run);
428 int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
431 extern struct kvm_x86_ops *kvm_x86_ops;
433 int kvm_mmu_module_init(void);
434 void kvm_mmu_module_exit(void);
436 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
437 int kvm_mmu_create(struct kvm_vcpu *vcpu);
438 int kvm_mmu_setup(struct kvm_vcpu *vcpu);
439 void kvm_mmu_set_nonpresent_ptes(u64 trap_pte, u64 notrap_pte);
441 int kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
442 void kvm_mmu_slot_remove_write_access(struct kvm *kvm, int slot);
443 void kvm_mmu_zap_all(struct kvm *kvm);
444 unsigned int kvm_mmu_calculate_mmu_pages(struct kvm *kvm);
445 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned int kvm_nr_mmu_pages);
447 int load_pdptrs(struct kvm_vcpu *vcpu, unsigned long cr3);
449 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
450 const void *val, int bytes);
451 int kvm_pv_mmu_op(struct kvm_vcpu *vcpu, unsigned long bytes,
452 gpa_t addr, unsigned long *ret);
454 extern bool tdp_enabled;
456 enum emulation_result {
457 EMULATE_DONE, /* no further processing */
458 EMULATE_DO_MMIO, /* kvm_run filled with mmio request */
459 EMULATE_FAIL, /* can't emulate this instruction */
462 #define EMULTYPE_NO_DECODE (1 << 0)
463 #define EMULTYPE_TRAP_UD (1 << 1)
464 int emulate_instruction(struct kvm_vcpu *vcpu, struct kvm_run *run,
465 unsigned long cr2, u16 error_code, int emulation_type);
466 void kvm_report_emulation_failure(struct kvm_vcpu *cvpu, const char *context);
467 void realmode_lgdt(struct kvm_vcpu *vcpu, u16 size, unsigned long address);
468 void realmode_lidt(struct kvm_vcpu *vcpu, u16 size, unsigned long address);
469 void realmode_lmsw(struct kvm_vcpu *vcpu, unsigned long msw,
470 unsigned long *rflags);
472 unsigned long realmode_get_cr(struct kvm_vcpu *vcpu, int cr);
473 void realmode_set_cr(struct kvm_vcpu *vcpu, int cr, unsigned long value,
474 unsigned long *rflags);
475 void kvm_enable_efer_bits(u64);
476 int kvm_get_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 *data);
477 int kvm_set_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 data);
479 struct x86_emulate_ctxt;
481 int kvm_emulate_pio(struct kvm_vcpu *vcpu, struct kvm_run *run, int in,
482 int size, unsigned port);
483 int kvm_emulate_pio_string(struct kvm_vcpu *vcpu, struct kvm_run *run, int in,
484 int size, unsigned long count, int down,
485 gva_t address, int rep, unsigned port);
486 void kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
487 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
488 int emulate_invlpg(struct kvm_vcpu *vcpu, gva_t address);
489 int emulate_clts(struct kvm_vcpu *vcpu);
490 int emulator_get_dr(struct x86_emulate_ctxt *ctxt, int dr,
491 unsigned long *dest);
492 int emulator_set_dr(struct x86_emulate_ctxt *ctxt, int dr,
493 unsigned long value);
495 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int reason);
497 void kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
498 void kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr0);
499 void kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr0);
500 void kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr0);
501 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu);
502 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
503 void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l);
505 int kvm_get_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata);
506 int kvm_set_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 data);
508 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
509 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
510 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, unsigned long cr2,
513 void fx_init(struct kvm_vcpu *vcpu);
515 int emulator_read_std(unsigned long addr,
518 struct kvm_vcpu *vcpu);
519 int emulator_write_emulated(unsigned long addr,
522 struct kvm_vcpu *vcpu);
524 unsigned long segment_base(u16 selector);
526 void kvm_mmu_flush_tlb(struct kvm_vcpu *vcpu);
527 void kvm_mmu_pte_write(struct kvm_vcpu *vcpu, gpa_t gpa,
528 const u8 *new, int bytes);
529 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva);
530 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
531 int kvm_mmu_load(struct kvm_vcpu *vcpu);
532 void kvm_mmu_unload(struct kvm_vcpu *vcpu);
534 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
536 int kvm_fix_hypercall(struct kvm_vcpu *vcpu);
538 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t gva, u32 error_code);
540 void kvm_enable_tdp(void);
542 int load_pdptrs(struct kvm_vcpu *vcpu, unsigned long cr3);
543 int complete_pio(struct kvm_vcpu *vcpu);
545 static inline struct kvm_mmu_page *page_header(hpa_t shadow_page)
547 struct page *page = pfn_to_page(shadow_page >> PAGE_SHIFT);
549 return (struct kvm_mmu_page *)page_private(page);
552 static inline u16 read_fs(void)
555 asm("mov %%fs, %0" : "=g"(seg));
559 static inline u16 read_gs(void)
562 asm("mov %%gs, %0" : "=g"(seg));
566 static inline u16 read_ldt(void)
569 asm("sldt %0" : "=g"(ldt));
573 static inline void load_fs(u16 sel)
575 asm("mov %0, %%fs" : : "rm"(sel));
578 static inline void load_gs(u16 sel)
580 asm("mov %0, %%gs" : : "rm"(sel));
584 static inline void load_ldt(u16 sel)
586 asm("lldt %0" : : "rm"(sel));
590 static inline void get_idt(struct descriptor_table *table)
592 asm("sidt %0" : "=m"(*table));
595 static inline void get_gdt(struct descriptor_table *table)
597 asm("sgdt %0" : "=m"(*table));
600 static inline unsigned long read_tr_base(void)
603 asm("str %0" : "=g"(tr));
604 return segment_base(tr);
608 static inline unsigned long read_msr(unsigned long msr)
617 static inline void fx_save(struct i387_fxsave_struct *image)
619 asm("fxsave (%0)":: "r" (image));
622 static inline void fx_restore(struct i387_fxsave_struct *image)
624 asm("fxrstor (%0)":: "r" (image));
627 static inline void fpu_init(void)
632 static inline u32 get_rdx_init_val(void)
634 return 0x600; /* P6 family */
637 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
639 kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
642 #define ASM_VMX_VMCLEAR_RAX ".byte 0x66, 0x0f, 0xc7, 0x30"
643 #define ASM_VMX_VMLAUNCH ".byte 0x0f, 0x01, 0xc2"
644 #define ASM_VMX_VMRESUME ".byte 0x0f, 0x01, 0xc3"
645 #define ASM_VMX_VMPTRLD_RAX ".byte 0x0f, 0xc7, 0x30"
646 #define ASM_VMX_VMREAD_RDX_RAX ".byte 0x0f, 0x78, 0xd0"
647 #define ASM_VMX_VMWRITE_RAX_RDX ".byte 0x0f, 0x79, 0xd0"
648 #define ASM_VMX_VMWRITE_RSP_RDX ".byte 0x0f, 0x79, 0xd4"
649 #define ASM_VMX_VMXOFF ".byte 0x0f, 0x01, 0xc4"
650 #define ASM_VMX_VMXON_RAX ".byte 0xf3, 0x0f, 0xc7, 0x30"
651 #define ASM_VMX_INVVPID ".byte 0x66, 0x0f, 0x38, 0x81, 0x08"
653 #define MSR_IA32_TIME_STAMP_COUNTER 0x010
655 #define TSS_IOPB_BASE_OFFSET 0x66
656 #define TSS_BASE_SIZE 0x68
657 #define TSS_IOPB_SIZE (65536 / 8)
658 #define TSS_REDIRECTION_SIZE (256 / 8)
659 #define RMODE_TSS_SIZE \
660 (TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
663 TASK_SWITCH_CALL = 0,
664 TASK_SWITCH_IRET = 1,
666 TASK_SWITCH_GATE = 3,