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>
21 #include <asm/pvclock-abi.h>
24 #define KVM_MAX_VCPUS 16
25 #define KVM_MEMORY_SLOTS 32
26 /* memory slots that does not exposed to userspace */
27 #define KVM_PRIVATE_MEM_SLOTS 4
29 #define KVM_PIO_PAGE_OFFSET 1
31 #define CR3_PAE_RESERVED_BITS ((X86_CR3_PWT | X86_CR3_PCD) - 1)
32 #define CR3_NONPAE_RESERVED_BITS ((PAGE_SIZE-1) & ~(X86_CR3_PWT | X86_CR3_PCD))
33 #define CR3_L_MODE_RESERVED_BITS (CR3_NONPAE_RESERVED_BITS | \
34 0xFFFFFF0000000000ULL)
36 #define KVM_GUEST_CR0_MASK \
37 (X86_CR0_PG | X86_CR0_PE | X86_CR0_WP | X86_CR0_NE \
38 | X86_CR0_NW | X86_CR0_CD)
39 #define KVM_VM_CR0_ALWAYS_ON \
40 (X86_CR0_PG | X86_CR0_PE | X86_CR0_WP | X86_CR0_NE | X86_CR0_TS \
42 #define KVM_GUEST_CR4_MASK \
43 (X86_CR4_VME | X86_CR4_PSE | X86_CR4_PAE | X86_CR4_PGE | X86_CR4_VMXE)
44 #define KVM_PMODE_VM_CR4_ALWAYS_ON (X86_CR4_PAE | X86_CR4_VMXE)
45 #define KVM_RMODE_VM_CR4_ALWAYS_ON (X86_CR4_VME | X86_CR4_PAE | X86_CR4_VMXE)
47 #define INVALID_PAGE (~(hpa_t)0)
48 #define UNMAPPED_GVA (~(gpa_t)0)
50 /* shadow tables are PAE even on non-PAE hosts */
51 #define KVM_HPAGE_SHIFT 21
52 #define KVM_HPAGE_SIZE (1UL << KVM_HPAGE_SHIFT)
53 #define KVM_HPAGE_MASK (~(KVM_HPAGE_SIZE - 1))
55 #define KVM_PAGES_PER_HPAGE (KVM_HPAGE_SIZE / PAGE_SIZE)
68 #define SELECTOR_TI_MASK (1 << 2)
69 #define SELECTOR_RPL_MASK 0x03
73 #define KVM_ALIAS_SLOTS 4
75 #define KVM_PERMILLE_MMU_PAGES 20
76 #define KVM_MIN_ALLOC_MMU_PAGES 64
77 #define KVM_MMU_HASH_SHIFT 10
78 #define KVM_NUM_MMU_PAGES (1 << KVM_MMU_HASH_SHIFT)
79 #define KVM_MIN_FREE_MMU_PAGES 5
80 #define KVM_REFILL_PAGES 25
81 #define KVM_MAX_CPUID_ENTRIES 40
83 extern spinlock_t kvm_lock;
84 extern struct list_head vm_list;
122 #include <asm/kvm_x86_emulate.h>
124 #define KVM_NR_MEM_OBJS 40
126 struct kvm_guest_debug {
133 * We don't want allocation failures within the mmu code, so we preallocate
134 * enough memory for a single page fault in a cache.
136 struct kvm_mmu_memory_cache {
138 void *objects[KVM_NR_MEM_OBJS];
141 #define NR_PTE_CHAIN_ENTRIES 5
143 struct kvm_pte_chain {
144 u64 *parent_ptes[NR_PTE_CHAIN_ENTRIES];
145 struct hlist_node link;
149 * kvm_mmu_page_role, below, is defined as:
151 * bits 0:3 - total guest paging levels (2-4, or zero for real mode)
152 * bits 4:7 - page table level for this shadow (1-4)
153 * bits 8:9 - page table quadrant for 2-level guests
154 * bit 16 - "metaphysical" - gfn is not a real page (huge page/real mode)
155 * bits 17:19 - common access permissions for all ptes in this shadow page
157 union kvm_mmu_page_role {
163 unsigned pad_for_nice_hex_output:6;
164 unsigned metaphysical:1;
170 struct kvm_mmu_page {
171 struct list_head link;
172 struct hlist_node hash_link;
175 * The following two entries are used to key the shadow page in the
179 union kvm_mmu_page_role role;
182 /* hold the gfn of each spte inside spt */
184 unsigned long slot_bitmap; /* One bit set per slot which has memory
185 * in this shadow page.
187 int multimapped; /* More than one parent_pte? */
188 int root_count; /* Currently serving as active root */
190 u64 *parent_pte; /* !multimapped */
191 struct hlist_head parent_ptes; /* multimapped, kvm_pte_chain */
196 * x86 supports 3 paging modes (4-level 64-bit, 3-level 64-bit, and 2-level
197 * 32-bit). The kvm_mmu structure abstracts the details of the current mmu
201 void (*new_cr3)(struct kvm_vcpu *vcpu);
202 int (*page_fault)(struct kvm_vcpu *vcpu, gva_t gva, u32 err);
203 void (*free)(struct kvm_vcpu *vcpu);
204 gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t gva);
205 void (*prefetch_page)(struct kvm_vcpu *vcpu,
206 struct kvm_mmu_page *page);
209 int shadow_root_level;
214 struct kvm_vcpu_arch {
216 int interrupt_window_open;
217 unsigned long irq_summary; /* bit vector: 1 per word in irq_pending */
218 DECLARE_BITMAP(irq_pending, KVM_NR_INTERRUPTS);
219 unsigned long regs[NR_VCPU_REGS]; /* for rsp: vcpu_load_rsp_rip() */
220 unsigned long rip; /* needs vcpu_load_rsp_rip() */
227 u64 pdptrs[4]; /* pae */
230 struct kvm_lapic *apic; /* kernel irqchip context */
233 u64 ia32_misc_enable_msr;
234 bool tpr_access_reporting;
238 struct kvm_mmu_memory_cache mmu_pte_chain_cache;
239 struct kvm_mmu_memory_cache mmu_rmap_desc_cache;
240 struct kvm_mmu_memory_cache mmu_page_cache;
241 struct kvm_mmu_memory_cache mmu_page_header_cache;
243 gfn_t last_pt_write_gfn;
244 int last_pt_write_count;
245 u64 *last_pte_updated;
248 gfn_t gfn; /* presumed gfn during guest pte update */
249 pfn_t pfn; /* pfn corresponding to that gfn */
253 struct i387_fxsave_struct host_fx_image;
254 struct i387_fxsave_struct guest_fx_image;
256 gva_t mmio_fault_cr2;
257 struct kvm_pio_request pio;
260 struct kvm_queued_exception {
270 struct kvm_save_segment {
275 } tr, es, ds, fs, gs;
277 int halt_request; /* real mode on Intel only */
280 struct kvm_cpuid_entry2 cpuid_entries[KVM_MAX_CPUID_ENTRIES];
281 /* emulate context */
283 struct x86_emulate_ctxt emulate_ctxt;
286 struct pvclock_vcpu_time_info hv_clock;
287 unsigned int hv_clock_tsc_khz;
288 unsigned int time_offset;
289 struct page *time_page;
292 struct kvm_mem_alias {
294 unsigned long npages;
300 struct kvm_mem_alias aliases[KVM_ALIAS_SLOTS];
302 unsigned int n_free_mmu_pages;
303 unsigned int n_requested_mmu_pages;
304 unsigned int n_alloc_mmu_pages;
305 struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
307 * Hash table of struct kvm_mmu_page.
309 struct list_head active_mmu_pages;
310 struct kvm_pic *vpic;
311 struct kvm_ioapic *vioapic;
312 struct kvm_pit *vpit;
314 int round_robin_prev_vcpu;
315 unsigned int tss_addr;
316 struct page *apic_access_page;
320 struct page *ept_identity_pagetable;
321 bool ept_identity_pagetable_done;
325 u32 mmu_shadow_zapped;
332 u32 remote_tlb_flush;
336 struct kvm_vcpu_stat {
346 u32 irq_window_exits;
349 u32 request_irq_exits;
351 u32 host_state_reload;
355 u32 insn_emulation_fail;
359 struct descriptor_table {
362 } __attribute__((packed));
365 int (*cpu_has_kvm_support)(void); /* __init */
366 int (*disabled_by_bios)(void); /* __init */
367 void (*hardware_enable)(void *dummy); /* __init */
368 void (*hardware_disable)(void *dummy);
369 void (*check_processor_compatibility)(void *rtn);
370 int (*hardware_setup)(void); /* __init */
371 void (*hardware_unsetup)(void); /* __exit */
372 bool (*cpu_has_accelerated_tpr)(void);
374 /* Create, but do not attach this VCPU */
375 struct kvm_vcpu *(*vcpu_create)(struct kvm *kvm, unsigned id);
376 void (*vcpu_free)(struct kvm_vcpu *vcpu);
377 int (*vcpu_reset)(struct kvm_vcpu *vcpu);
379 void (*prepare_guest_switch)(struct kvm_vcpu *vcpu);
380 void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
381 void (*vcpu_put)(struct kvm_vcpu *vcpu);
382 void (*vcpu_decache)(struct kvm_vcpu *vcpu);
384 int (*set_guest_debug)(struct kvm_vcpu *vcpu,
385 struct kvm_debug_guest *dbg);
386 void (*guest_debug_pre)(struct kvm_vcpu *vcpu);
387 int (*get_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata);
388 int (*set_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 data);
389 u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
390 void (*get_segment)(struct kvm_vcpu *vcpu,
391 struct kvm_segment *var, int seg);
392 int (*get_cpl)(struct kvm_vcpu *vcpu);
393 void (*set_segment)(struct kvm_vcpu *vcpu,
394 struct kvm_segment *var, int seg);
395 void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
396 void (*decache_cr4_guest_bits)(struct kvm_vcpu *vcpu);
397 void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
398 void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
399 void (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
400 void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
401 void (*get_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
402 void (*set_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
403 void (*get_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
404 void (*set_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
405 unsigned long (*get_dr)(struct kvm_vcpu *vcpu, int dr);
406 void (*set_dr)(struct kvm_vcpu *vcpu, int dr, unsigned long value,
408 void (*cache_regs)(struct kvm_vcpu *vcpu);
409 void (*decache_regs)(struct kvm_vcpu *vcpu);
410 unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
411 void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
413 void (*tlb_flush)(struct kvm_vcpu *vcpu);
415 void (*run)(struct kvm_vcpu *vcpu, struct kvm_run *run);
416 int (*handle_exit)(struct kvm_run *run, struct kvm_vcpu *vcpu);
417 void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
418 void (*patch_hypercall)(struct kvm_vcpu *vcpu,
419 unsigned char *hypercall_addr);
420 int (*get_irq)(struct kvm_vcpu *vcpu);
421 void (*set_irq)(struct kvm_vcpu *vcpu, int vec);
422 void (*queue_exception)(struct kvm_vcpu *vcpu, unsigned nr,
423 bool has_error_code, u32 error_code);
424 bool (*exception_injected)(struct kvm_vcpu *vcpu);
425 void (*inject_pending_irq)(struct kvm_vcpu *vcpu);
426 void (*inject_pending_vectors)(struct kvm_vcpu *vcpu,
427 struct kvm_run *run);
429 int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
430 int (*get_tdp_level)(void);
433 extern struct kvm_x86_ops *kvm_x86_ops;
435 int kvm_mmu_module_init(void);
436 void kvm_mmu_module_exit(void);
438 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
439 int kvm_mmu_create(struct kvm_vcpu *vcpu);
440 int kvm_mmu_setup(struct kvm_vcpu *vcpu);
441 void kvm_mmu_set_nonpresent_ptes(u64 trap_pte, u64 notrap_pte);
442 void kvm_mmu_set_base_ptes(u64 base_pte);
443 void kvm_mmu_set_mask_ptes(u64 user_mask, u64 accessed_mask,
444 u64 dirty_mask, u64 nx_mask, u64 x_mask);
446 int kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
447 void kvm_mmu_slot_remove_write_access(struct kvm *kvm, int slot);
448 void kvm_mmu_zap_all(struct kvm *kvm);
449 unsigned int kvm_mmu_calculate_mmu_pages(struct kvm *kvm);
450 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned int kvm_nr_mmu_pages);
452 int load_pdptrs(struct kvm_vcpu *vcpu, unsigned long cr3);
454 int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
455 const void *val, int bytes);
456 int kvm_pv_mmu_op(struct kvm_vcpu *vcpu, unsigned long bytes,
457 gpa_t addr, unsigned long *ret);
459 extern bool tdp_enabled;
461 enum emulation_result {
462 EMULATE_DONE, /* no further processing */
463 EMULATE_DO_MMIO, /* kvm_run filled with mmio request */
464 EMULATE_FAIL, /* can't emulate this instruction */
467 #define EMULTYPE_NO_DECODE (1 << 0)
468 #define EMULTYPE_TRAP_UD (1 << 1)
469 int emulate_instruction(struct kvm_vcpu *vcpu, struct kvm_run *run,
470 unsigned long cr2, u16 error_code, int emulation_type);
471 void kvm_report_emulation_failure(struct kvm_vcpu *cvpu, const char *context);
472 void realmode_lgdt(struct kvm_vcpu *vcpu, u16 size, unsigned long address);
473 void realmode_lidt(struct kvm_vcpu *vcpu, u16 size, unsigned long address);
474 void realmode_lmsw(struct kvm_vcpu *vcpu, unsigned long msw,
475 unsigned long *rflags);
477 unsigned long realmode_get_cr(struct kvm_vcpu *vcpu, int cr);
478 void realmode_set_cr(struct kvm_vcpu *vcpu, int cr, unsigned long value,
479 unsigned long *rflags);
480 void kvm_enable_efer_bits(u64);
481 int kvm_get_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 *data);
482 int kvm_set_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 data);
484 struct x86_emulate_ctxt;
486 int kvm_emulate_pio(struct kvm_vcpu *vcpu, struct kvm_run *run, int in,
487 int size, unsigned port);
488 int kvm_emulate_pio_string(struct kvm_vcpu *vcpu, struct kvm_run *run, int in,
489 int size, unsigned long count, int down,
490 gva_t address, int rep, unsigned port);
491 void kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
492 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
493 int emulate_invlpg(struct kvm_vcpu *vcpu, gva_t address);
494 int emulate_clts(struct kvm_vcpu *vcpu);
495 int emulator_get_dr(struct x86_emulate_ctxt *ctxt, int dr,
496 unsigned long *dest);
497 int emulator_set_dr(struct x86_emulate_ctxt *ctxt, int dr,
498 unsigned long value);
500 int kvm_task_switch(struct kvm_vcpu *vcpu, u16 tss_selector, int reason);
502 void kvm_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
503 void kvm_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3);
504 void kvm_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4);
505 void kvm_set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8);
506 unsigned long kvm_get_cr8(struct kvm_vcpu *vcpu);
507 void kvm_lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
508 void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l);
510 int kvm_get_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata);
511 int kvm_set_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 data);
513 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
514 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
515 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, unsigned long cr2,
518 void fx_init(struct kvm_vcpu *vcpu);
520 int emulator_read_std(unsigned long addr,
523 struct kvm_vcpu *vcpu);
524 int emulator_write_emulated(unsigned long addr,
527 struct kvm_vcpu *vcpu);
529 unsigned long segment_base(u16 selector);
531 void kvm_mmu_flush_tlb(struct kvm_vcpu *vcpu);
532 void kvm_mmu_pte_write(struct kvm_vcpu *vcpu, gpa_t gpa,
533 const u8 *new, int bytes);
534 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva);
535 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
536 int kvm_mmu_load(struct kvm_vcpu *vcpu);
537 void kvm_mmu_unload(struct kvm_vcpu *vcpu);
539 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
541 int kvm_fix_hypercall(struct kvm_vcpu *vcpu);
543 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t gva, u32 error_code);
545 void kvm_enable_tdp(void);
547 int load_pdptrs(struct kvm_vcpu *vcpu, unsigned long cr3);
548 int complete_pio(struct kvm_vcpu *vcpu);
550 static inline struct kvm_mmu_page *page_header(hpa_t shadow_page)
552 struct page *page = pfn_to_page(shadow_page >> PAGE_SHIFT);
554 return (struct kvm_mmu_page *)page_private(page);
557 static inline u16 read_fs(void)
560 asm("mov %%fs, %0" : "=g"(seg));
564 static inline u16 read_gs(void)
567 asm("mov %%gs, %0" : "=g"(seg));
571 static inline u16 read_ldt(void)
574 asm("sldt %0" : "=g"(ldt));
578 static inline void load_fs(u16 sel)
580 asm("mov %0, %%fs" : : "rm"(sel));
583 static inline void load_gs(u16 sel)
585 asm("mov %0, %%gs" : : "rm"(sel));
589 static inline void load_ldt(u16 sel)
591 asm("lldt %0" : : "rm"(sel));
595 static inline void get_idt(struct descriptor_table *table)
597 asm("sidt %0" : "=m"(*table));
600 static inline void get_gdt(struct descriptor_table *table)
602 asm("sgdt %0" : "=m"(*table));
605 static inline unsigned long read_tr_base(void)
608 asm("str %0" : "=g"(tr));
609 return segment_base(tr);
613 static inline unsigned long read_msr(unsigned long msr)
622 static inline void fx_save(struct i387_fxsave_struct *image)
624 asm("fxsave (%0)":: "r" (image));
627 static inline void fx_restore(struct i387_fxsave_struct *image)
629 asm("fxrstor (%0)":: "r" (image));
632 static inline void fx_finit(void)
637 static inline u32 get_rdx_init_val(void)
639 return 0x600; /* P6 family */
642 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
644 kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
647 #define ASM_VMX_VMCLEAR_RAX ".byte 0x66, 0x0f, 0xc7, 0x30"
648 #define ASM_VMX_VMLAUNCH ".byte 0x0f, 0x01, 0xc2"
649 #define ASM_VMX_VMRESUME ".byte 0x0f, 0x01, 0xc3"
650 #define ASM_VMX_VMPTRLD_RAX ".byte 0x0f, 0xc7, 0x30"
651 #define ASM_VMX_VMREAD_RDX_RAX ".byte 0x0f, 0x78, 0xd0"
652 #define ASM_VMX_VMWRITE_RAX_RDX ".byte 0x0f, 0x79, 0xd0"
653 #define ASM_VMX_VMWRITE_RSP_RDX ".byte 0x0f, 0x79, 0xd4"
654 #define ASM_VMX_VMXOFF ".byte 0x0f, 0x01, 0xc4"
655 #define ASM_VMX_VMXON_RAX ".byte 0xf3, 0x0f, 0xc7, 0x30"
656 #define ASM_VMX_INVEPT ".byte 0x66, 0x0f, 0x38, 0x80, 0x08"
657 #define ASM_VMX_INVVPID ".byte 0x66, 0x0f, 0x38, 0x81, 0x08"
659 #define MSR_IA32_TIME_STAMP_COUNTER 0x010
661 #define TSS_IOPB_BASE_OFFSET 0x66
662 #define TSS_BASE_SIZE 0x68
663 #define TSS_IOPB_SIZE (65536 / 8)
664 #define TSS_REDIRECTION_SIZE (256 / 8)
665 #define RMODE_TSS_SIZE \
666 (TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
669 TASK_SWITCH_CALL = 0,
670 TASK_SWITCH_IRET = 1,
672 TASK_SWITCH_GATE = 3,
675 #define KVMTRACE_5D(evt, vcpu, d1, d2, d3, d4, d5, name) \
676 trace_mark(kvm_trace_##name, "%u %p %u %u %u %u %u %u", KVM_TRC_##evt, \
677 vcpu, 5, d1, d2, d3, d4, d5)
678 #define KVMTRACE_4D(evt, vcpu, d1, d2, d3, d4, name) \
679 trace_mark(kvm_trace_##name, "%u %p %u %u %u %u %u %u", KVM_TRC_##evt, \
680 vcpu, 4, d1, d2, d3, d4, 0)
681 #define KVMTRACE_3D(evt, vcpu, d1, d2, d3, name) \
682 trace_mark(kvm_trace_##name, "%u %p %u %u %u %u %u %u", KVM_TRC_##evt, \
683 vcpu, 3, d1, d2, d3, 0, 0)
684 #define KVMTRACE_2D(evt, vcpu, d1, d2, name) \
685 trace_mark(kvm_trace_##name, "%u %p %u %u %u %u %u %u", KVM_TRC_##evt, \
686 vcpu, 2, d1, d2, 0, 0, 0)
687 #define KVMTRACE_1D(evt, vcpu, d1, name) \
688 trace_mark(kvm_trace_##name, "%u %p %u %u %u %u %u %u", KVM_TRC_##evt, \
689 vcpu, 1, d1, 0, 0, 0, 0)
690 #define KVMTRACE_0D(evt, vcpu, name) \
691 trace_mark(kvm_trace_##name, "%u %p %u %u %u %u %u %u", KVM_TRC_##evt, \
692 vcpu, 0, 0, 0, 0, 0, 0)