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)
67 #define SELECTOR_TI_MASK (1 << 2)
68 #define SELECTOR_RPL_MASK 0x03
72 #define KVM_ALIAS_SLOTS 4
74 #define KVM_PERMILLE_MMU_PAGES 20
75 #define KVM_MIN_ALLOC_MMU_PAGES 64
76 #define KVM_MMU_HASH_SHIFT 10
77 #define KVM_NUM_MMU_PAGES (1 << KVM_MMU_HASH_SHIFT)
78 #define KVM_MIN_FREE_MMU_PAGES 5
79 #define KVM_REFILL_PAGES 25
80 #define KVM_MAX_CPUID_ENTRIES 40
82 extern spinlock_t kvm_lock;
83 extern struct list_head vm_list;
121 #include <asm/kvm_x86_emulate.h>
123 #define KVM_NR_MEM_OBJS 40
125 struct kvm_guest_debug {
132 * We don't want allocation failures within the mmu code, so we preallocate
133 * enough memory for a single page fault in a cache.
135 struct kvm_mmu_memory_cache {
137 void *objects[KVM_NR_MEM_OBJS];
140 #define NR_PTE_CHAIN_ENTRIES 5
142 struct kvm_pte_chain {
143 u64 *parent_ptes[NR_PTE_CHAIN_ENTRIES];
144 struct hlist_node link;
148 * kvm_mmu_page_role, below, is defined as:
150 * bits 0:3 - total guest paging levels (2-4, or zero for real mode)
151 * bits 4:7 - page table level for this shadow (1-4)
152 * bits 8:9 - page table quadrant for 2-level guests
153 * bit 16 - "metaphysical" - gfn is not a real page (huge page/real mode)
154 * bits 17:19 - common access permissions for all ptes in this shadow page
156 union kvm_mmu_page_role {
162 unsigned pad_for_nice_hex_output:6;
163 unsigned metaphysical:1;
169 struct kvm_mmu_page {
170 struct list_head link;
171 struct hlist_node hash_link;
174 * The following two entries are used to key the shadow page in the
178 union kvm_mmu_page_role role;
181 /* hold the gfn of each spte inside spt */
183 unsigned long slot_bitmap; /* One bit set per slot which has memory
184 * in this shadow page.
186 int multimapped; /* More than one parent_pte? */
187 int root_count; /* Currently serving as active root */
189 u64 *parent_pte; /* !multimapped */
190 struct hlist_head parent_ptes; /* multimapped, kvm_pte_chain */
195 * x86 supports 3 paging modes (4-level 64-bit, 3-level 64-bit, and 2-level
196 * 32-bit). The kvm_mmu structure abstracts the details of the current mmu
200 void (*new_cr3)(struct kvm_vcpu *vcpu);
201 int (*page_fault)(struct kvm_vcpu *vcpu, gva_t gva, u32 err);
202 void (*free)(struct kvm_vcpu *vcpu);
203 gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t gva);
204 void (*prefetch_page)(struct kvm_vcpu *vcpu,
205 struct kvm_mmu_page *page);
208 int shadow_root_level;
213 struct kvm_vcpu_arch {
215 int interrupt_window_open;
216 unsigned long irq_summary; /* bit vector: 1 per word in irq_pending */
217 DECLARE_BITMAP(irq_pending, KVM_NR_INTERRUPTS);
218 unsigned long regs[NR_VCPU_REGS]; /* for rsp: vcpu_load_rsp_rip() */
219 unsigned long rip; /* needs vcpu_load_rsp_rip() */
226 u64 pdptrs[4]; /* pae */
229 struct kvm_lapic *apic; /* kernel irqchip context */
232 u64 ia32_misc_enable_msr;
233 bool tpr_access_reporting;
237 struct kvm_mmu_memory_cache mmu_pte_chain_cache;
238 struct kvm_mmu_memory_cache mmu_rmap_desc_cache;
239 struct kvm_mmu_memory_cache mmu_page_cache;
240 struct kvm_mmu_memory_cache mmu_page_header_cache;
242 gfn_t last_pt_write_gfn;
243 int last_pt_write_count;
244 u64 *last_pte_updated;
247 gfn_t gfn; /* presumed gfn during guest pte update */
248 pfn_t pfn; /* pfn corresponding to that gfn */
252 struct i387_fxsave_struct host_fx_image;
253 struct i387_fxsave_struct guest_fx_image;
255 gva_t mmio_fault_cr2;
256 struct kvm_pio_request pio;
259 struct kvm_queued_exception {
269 struct kvm_save_segment {
274 } tr, es, ds, fs, gs;
276 int halt_request; /* real mode on Intel only */
279 struct kvm_cpuid_entry2 cpuid_entries[KVM_MAX_CPUID_ENTRIES];
280 /* emulate context */
282 struct x86_emulate_ctxt emulate_ctxt;
285 struct kvm_vcpu_time_info hv_clock;
286 unsigned int time_offset;
287 struct page *time_page;
290 struct kvm_mem_alias {
292 unsigned long npages;
298 struct kvm_mem_alias aliases[KVM_ALIAS_SLOTS];
300 unsigned int n_free_mmu_pages;
301 unsigned int n_requested_mmu_pages;
302 unsigned int n_alloc_mmu_pages;
303 struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
305 * Hash table of struct kvm_mmu_page.
307 struct list_head active_mmu_pages;
308 struct kvm_pic *vpic;
309 struct kvm_ioapic *vioapic;
310 struct kvm_pit *vpit;
312 int round_robin_prev_vcpu;
313 unsigned int tss_addr;
314 struct page *apic_access_page;
318 struct page *ept_identity_pagetable;
319 bool ept_identity_pagetable_done;
323 u32 mmu_shadow_zapped;
330 u32 remote_tlb_flush;
334 struct kvm_vcpu_stat {
344 u32 irq_window_exits;
347 u32 request_irq_exits;
349 u32 host_state_reload;
353 u32 insn_emulation_fail;
357 struct descriptor_table {
360 } __attribute__((packed));
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 void (*check_processor_compatibility)(void *rtn);
368 int (*hardware_setup)(void); /* __init */
369 void (*hardware_unsetup)(void); /* __exit */
370 bool (*cpu_has_accelerated_tpr)(void);
372 /* Create, but do not attach this VCPU */
373 struct kvm_vcpu *(*vcpu_create)(struct kvm *kvm, unsigned id);
374 void (*vcpu_free)(struct kvm_vcpu *vcpu);
375 int (*vcpu_reset)(struct kvm_vcpu *vcpu);
377 void (*prepare_guest_switch)(struct kvm_vcpu *vcpu);
378 void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
379 void (*vcpu_put)(struct kvm_vcpu *vcpu);
380 void (*vcpu_decache)(struct kvm_vcpu *vcpu);
382 int (*set_guest_debug)(struct kvm_vcpu *vcpu,
383 struct kvm_debug_guest *dbg);
384 void (*guest_debug_pre)(struct kvm_vcpu *vcpu);
385 int (*get_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata);
386 int (*set_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 data);
387 u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
388 void (*get_segment)(struct kvm_vcpu *vcpu,
389 struct kvm_segment *var, int seg);
390 int (*get_cpl)(struct kvm_vcpu *vcpu);
391 void (*set_segment)(struct kvm_vcpu *vcpu,
392 struct kvm_segment *var, int seg);
393 void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
394 void (*decache_cr4_guest_bits)(struct kvm_vcpu *vcpu);
395 void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
396 void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
397 void (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
398 void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
399 void (*get_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
400 void (*set_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
401 void (*get_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
402 void (*set_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
403 unsigned long (*get_dr)(struct kvm_vcpu *vcpu, int dr);
404 void (*set_dr)(struct kvm_vcpu *vcpu, int dr, unsigned long value,
406 void (*cache_regs)(struct kvm_vcpu *vcpu);
407 void (*decache_regs)(struct kvm_vcpu *vcpu);
408 unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
409 void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
411 void (*tlb_flush)(struct kvm_vcpu *vcpu);
413 void (*run)(struct kvm_vcpu *vcpu, struct kvm_run *run);
414 int (*handle_exit)(struct kvm_run *run, struct kvm_vcpu *vcpu);
415 void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
416 void (*patch_hypercall)(struct kvm_vcpu *vcpu,
417 unsigned char *hypercall_addr);
418 int (*get_irq)(struct kvm_vcpu *vcpu);
419 void (*set_irq)(struct kvm_vcpu *vcpu, int vec);
420 void (*queue_exception)(struct kvm_vcpu *vcpu, unsigned nr,
421 bool has_error_code, u32 error_code);
422 bool (*exception_injected)(struct kvm_vcpu *vcpu);
423 void (*inject_pending_irq)(struct kvm_vcpu *vcpu);
424 void (*inject_pending_vectors)(struct kvm_vcpu *vcpu,
425 struct kvm_run *run);
427 int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
428 int (*get_tdp_level)(void);
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);
440 void kvm_mmu_set_base_ptes(u64 base_pte);
441 void kvm_mmu_set_mask_ptes(u64 user_mask, u64 accessed_mask,
442 u64 dirty_mask, u64 nx_mask, u64 x_mask);
444 int kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
445 void kvm_mmu_slot_remove_write_access(struct kvm *kvm, int slot);
446 void kvm_mmu_zap_all(struct kvm *kvm);
447 unsigned int kvm_mmu_calculate_mmu_pages(struct kvm *kvm);
448 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned int kvm_nr_mmu_pages);
450 int load_pdptrs(struct kvm_vcpu *vcpu, unsigned long cr3);
452 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
453 const void *val, int bytes);
454 int kvm_pv_mmu_op(struct kvm_vcpu *vcpu, unsigned long bytes,
455 gpa_t addr, unsigned long *ret);
457 extern bool tdp_enabled;
459 enum emulation_result {
460 EMULATE_DONE, /* no further processing */
461 EMULATE_DO_MMIO, /* kvm_run filled with mmio request */
462 EMULATE_FAIL, /* can't emulate this instruction */
465 #define EMULTYPE_NO_DECODE (1 << 0)
466 #define EMULTYPE_TRAP_UD (1 << 1)
467 int emulate_instruction(struct kvm_vcpu *vcpu, struct kvm_run *run,
468 unsigned long cr2, u16 error_code, int emulation_type);
469 void kvm_report_emulation_failure(struct kvm_vcpu *cvpu, const char *context);
470 void realmode_lgdt(struct kvm_vcpu *vcpu, u16 size, unsigned long address);
471 void realmode_lidt(struct kvm_vcpu *vcpu, u16 size, unsigned long address);
472 void realmode_lmsw(struct kvm_vcpu *vcpu, unsigned long msw,
473 unsigned long *rflags);
475 unsigned long realmode_get_cr(struct kvm_vcpu *vcpu, int cr);
476 void realmode_set_cr(struct kvm_vcpu *vcpu, int cr, unsigned long value,
477 unsigned long *rflags);
478 void kvm_enable_efer_bits(u64);
479 int kvm_get_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 *data);
480 int kvm_set_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 data);
482 struct x86_emulate_ctxt;
484 int kvm_emulate_pio(struct kvm_vcpu *vcpu, struct kvm_run *run, int in,
485 int size, unsigned port);
486 int kvm_emulate_pio_string(struct kvm_vcpu *vcpu, struct kvm_run *run, int in,
487 int size, unsigned long count, int down,
488 gva_t address, int rep, unsigned port);
489 void kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
490 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
491 int emulate_invlpg(struct kvm_vcpu *vcpu, gva_t address);
492 int emulate_clts(struct kvm_vcpu *vcpu);
493 int emulator_get_dr(struct x86_emulate_ctxt *ctxt, int dr,
494 unsigned long *dest);
495 int emulator_set_dr(struct x86_emulate_ctxt *ctxt, int dr,
496 unsigned long value);
498 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int reason);
500 void kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
501 void kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3);
502 void kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
503 void kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8);
504 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu);
505 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
506 void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l);
508 int kvm_get_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata);
509 int kvm_set_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 data);
511 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
512 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
513 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, unsigned long cr2,
516 void fx_init(struct kvm_vcpu *vcpu);
518 int emulator_read_std(unsigned long addr,
521 struct kvm_vcpu *vcpu);
522 int emulator_write_emulated(unsigned long addr,
525 struct kvm_vcpu *vcpu);
527 unsigned long segment_base(u16 selector);
529 void kvm_mmu_flush_tlb(struct kvm_vcpu *vcpu);
530 void kvm_mmu_pte_write(struct kvm_vcpu *vcpu, gpa_t gpa,
531 const u8 *new, int bytes);
532 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva);
533 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
534 int kvm_mmu_load(struct kvm_vcpu *vcpu);
535 void kvm_mmu_unload(struct kvm_vcpu *vcpu);
537 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
539 int kvm_fix_hypercall(struct kvm_vcpu *vcpu);
541 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t gva, u32 error_code);
543 void kvm_enable_tdp(void);
545 int load_pdptrs(struct kvm_vcpu *vcpu, unsigned long cr3);
546 int complete_pio(struct kvm_vcpu *vcpu);
548 static inline struct kvm_mmu_page *page_header(hpa_t shadow_page)
550 struct page *page = pfn_to_page(shadow_page >> PAGE_SHIFT);
552 return (struct kvm_mmu_page *)page_private(page);
555 static inline u16 read_fs(void)
558 asm("mov %%fs, %0" : "=g"(seg));
562 static inline u16 read_gs(void)
565 asm("mov %%gs, %0" : "=g"(seg));
569 static inline u16 read_ldt(void)
572 asm("sldt %0" : "=g"(ldt));
576 static inline void load_fs(u16 sel)
578 asm("mov %0, %%fs" : : "rm"(sel));
581 static inline void load_gs(u16 sel)
583 asm("mov %0, %%gs" : : "rm"(sel));
587 static inline void load_ldt(u16 sel)
589 asm("lldt %0" : : "rm"(sel));
593 static inline void get_idt(struct descriptor_table *table)
595 asm("sidt %0" : "=m"(*table));
598 static inline void get_gdt(struct descriptor_table *table)
600 asm("sgdt %0" : "=m"(*table));
603 static inline unsigned long read_tr_base(void)
606 asm("str %0" : "=g"(tr));
607 return segment_base(tr);
611 static inline unsigned long read_msr(unsigned long msr)
620 static inline void fx_save(struct i387_fxsave_struct *image)
622 asm("fxsave (%0)":: "r" (image));
625 static inline void fx_restore(struct i387_fxsave_struct *image)
627 asm("fxrstor (%0)":: "r" (image));
630 static inline void fx_finit(void)
635 static inline u32 get_rdx_init_val(void)
637 return 0x600; /* P6 family */
640 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
642 kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
645 #define ASM_VMX_VMCLEAR_RAX ".byte 0x66, 0x0f, 0xc7, 0x30"
646 #define ASM_VMX_VMLAUNCH ".byte 0x0f, 0x01, 0xc2"
647 #define ASM_VMX_VMRESUME ".byte 0x0f, 0x01, 0xc3"
648 #define ASM_VMX_VMPTRLD_RAX ".byte 0x0f, 0xc7, 0x30"
649 #define ASM_VMX_VMREAD_RDX_RAX ".byte 0x0f, 0x78, 0xd0"
650 #define ASM_VMX_VMWRITE_RAX_RDX ".byte 0x0f, 0x79, 0xd0"
651 #define ASM_VMX_VMWRITE_RSP_RDX ".byte 0x0f, 0x79, 0xd4"
652 #define ASM_VMX_VMXOFF ".byte 0x0f, 0x01, 0xc4"
653 #define ASM_VMX_VMXON_RAX ".byte 0xf3, 0x0f, 0xc7, 0x30"
654 #define ASM_VMX_INVEPT ".byte 0x66, 0x0f, 0x38, 0x80, 0x08"
655 #define ASM_VMX_INVVPID ".byte 0x66, 0x0f, 0x38, 0x81, 0x08"
657 #define MSR_IA32_TIME_STAMP_COUNTER 0x010
659 #define TSS_IOPB_BASE_OFFSET 0x66
660 #define TSS_BASE_SIZE 0x68
661 #define TSS_IOPB_SIZE (65536 / 8)
662 #define TSS_REDIRECTION_SIZE (256 / 8)
663 #define RMODE_TSS_SIZE \
664 (TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
667 TASK_SWITCH_CALL = 0,
668 TASK_SWITCH_IRET = 1,
670 TASK_SWITCH_GATE = 3,
673 #define KVMTRACE_5D(evt, vcpu, d1, d2, d3, d4, d5, name) \
674 trace_mark(kvm_trace_##name, "%u %p %u %u %u %u %u %u", KVM_TRC_##evt, \
675 vcpu, 5, d1, d2, d3, d4, d5)
676 #define KVMTRACE_4D(evt, vcpu, d1, d2, d3, d4, name) \
677 trace_mark(kvm_trace_##name, "%u %p %u %u %u %u %u %u", KVM_TRC_##evt, \
678 vcpu, 4, d1, d2, d3, d4, 0)
679 #define KVMTRACE_3D(evt, vcpu, d1, d2, d3, name) \
680 trace_mark(kvm_trace_##name, "%u %p %u %u %u %u %u %u", KVM_TRC_##evt, \
681 vcpu, 3, d1, d2, d3, 0, 0)
682 #define KVMTRACE_2D(evt, vcpu, d1, d2, name) \
683 trace_mark(kvm_trace_##name, "%u %p %u %u %u %u %u %u", KVM_TRC_##evt, \
684 vcpu, 2, d1, d2, 0, 0, 0)
685 #define KVMTRACE_1D(evt, vcpu, d1, name) \
686 trace_mark(kvm_trace_##name, "%u %p %u %u %u %u %u %u", KVM_TRC_##evt, \
687 vcpu, 1, d1, 0, 0, 0, 0)
688 #define KVMTRACE_0D(evt, vcpu, name) \
689 trace_mark(kvm_trace_##name, "%u %p %u %u %u %u %u %u", KVM_TRC_##evt, \
690 vcpu, 0, 0, 0, 0, 0, 0)