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 CR3_PAE_RESERVED_BITS ((X86_CR3_PWT | X86_CR3_PCD) - 1)
24 #define CR3_NONPAE_RESERVED_BITS ((PAGE_SIZE-1) & ~(X86_CR3_PWT | X86_CR3_PCD))
25 #define CR3_L_MODE_RESERVED_BITS (CR3_NONPAE_RESERVED_BITS | \
26 0xFFFFFF0000000000ULL)
28 #define KVM_GUEST_CR0_MASK \
29 (X86_CR0_PG | X86_CR0_PE | X86_CR0_WP | X86_CR0_NE \
30 | X86_CR0_NW | X86_CR0_CD)
31 #define KVM_VM_CR0_ALWAYS_ON \
32 (X86_CR0_PG | X86_CR0_PE | X86_CR0_WP | X86_CR0_NE | X86_CR0_TS \
34 #define KVM_GUEST_CR4_MASK \
35 (X86_CR4_VME | X86_CR4_PSE | X86_CR4_PAE | X86_CR4_PGE | X86_CR4_VMXE)
36 #define KVM_PMODE_VM_CR4_ALWAYS_ON (X86_CR4_PAE | X86_CR4_VMXE)
37 #define KVM_RMODE_VM_CR4_ALWAYS_ON (X86_CR4_VME | X86_CR4_PAE | X86_CR4_VMXE)
39 #define INVALID_PAGE (~(hpa_t)0)
40 #define UNMAPPED_GVA (~(gpa_t)0)
52 #define SELECTOR_TI_MASK (1 << 2)
53 #define SELECTOR_RPL_MASK 0x03
57 #define KVM_ALIAS_SLOTS 4
59 #define KVM_PERMILLE_MMU_PAGES 20
60 #define KVM_MIN_ALLOC_MMU_PAGES 64
61 #define KVM_MMU_HASH_SHIFT 10
62 #define KVM_NUM_MMU_PAGES (1 << KVM_MMU_HASH_SHIFT)
63 #define KVM_MIN_FREE_MMU_PAGES 5
64 #define KVM_REFILL_PAGES 25
65 #define KVM_MAX_CPUID_ENTRIES 40
67 extern spinlock_t kvm_lock;
68 extern struct list_head vm_list;
106 #include <asm/kvm_x86_emulate.h>
108 #define KVM_NR_MEM_OBJS 40
111 * We don't want allocation failures within the mmu code, so we preallocate
112 * enough memory for a single page fault in a cache.
114 struct kvm_mmu_memory_cache {
116 void *objects[KVM_NR_MEM_OBJS];
119 #define NR_PTE_CHAIN_ENTRIES 5
121 struct kvm_pte_chain {
122 u64 *parent_ptes[NR_PTE_CHAIN_ENTRIES];
123 struct hlist_node link;
127 * kvm_mmu_page_role, below, is defined as:
129 * bits 0:3 - total guest paging levels (2-4, or zero for real mode)
130 * bits 4:7 - page table level for this shadow (1-4)
131 * bits 8:9 - page table quadrant for 2-level guests
132 * bit 16 - "metaphysical" - gfn is not a real page (huge page/real mode)
133 * bits 17:19 - common access permissions for all ptes in this shadow page
135 union kvm_mmu_page_role {
141 unsigned pad_for_nice_hex_output:6;
142 unsigned metaphysical:1;
147 struct kvm_mmu_page {
148 struct list_head link;
149 struct hlist_node hash_link;
152 * The following two entries are used to key the shadow page in the
156 union kvm_mmu_page_role role;
159 /* hold the gfn of each spte inside spt */
161 unsigned long slot_bitmap; /* One bit set per slot which has memory
162 * in this shadow page.
164 int multimapped; /* More than one parent_pte? */
165 int root_count; /* Currently serving as active root */
167 u64 *parent_pte; /* !multimapped */
168 struct hlist_head parent_ptes; /* multimapped, kvm_pte_chain */
173 * x86 supports 3 paging modes (4-level 64-bit, 3-level 64-bit, and 2-level
174 * 32-bit). The kvm_mmu structure abstracts the details of the current mmu
178 void (*new_cr3)(struct kvm_vcpu *vcpu);
179 int (*page_fault)(struct kvm_vcpu *vcpu, gva_t gva, u32 err);
180 void (*free)(struct kvm_vcpu *vcpu);
181 gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t gva);
182 void (*prefetch_page)(struct kvm_vcpu *vcpu,
183 struct kvm_mmu_page *page);
186 int shadow_root_level;
191 struct kvm_vcpu_arch {
193 int interrupt_window_open;
194 unsigned long irq_summary; /* bit vector: 1 per word in irq_pending */
195 DECLARE_BITMAP(irq_pending, KVM_NR_INTERRUPTS);
196 unsigned long regs[NR_VCPU_REGS]; /* for rsp: vcpu_load_rsp_rip() */
197 unsigned long rip; /* needs vcpu_load_rsp_rip() */
204 u64 pdptrs[4]; /* pae */
207 struct kvm_lapic *apic; /* kernel irqchip context */
208 #define VCPU_MP_STATE_RUNNABLE 0
209 #define VCPU_MP_STATE_UNINITIALIZED 1
210 #define VCPU_MP_STATE_INIT_RECEIVED 2
211 #define VCPU_MP_STATE_SIPI_RECEIVED 3
212 #define VCPU_MP_STATE_HALTED 4
215 u64 ia32_misc_enable_msr;
216 bool tpr_access_reporting;
220 struct kvm_mmu_memory_cache mmu_pte_chain_cache;
221 struct kvm_mmu_memory_cache mmu_rmap_desc_cache;
222 struct kvm_mmu_memory_cache mmu_page_cache;
223 struct kvm_mmu_memory_cache mmu_page_header_cache;
225 gfn_t last_pt_write_gfn;
226 int last_pt_write_count;
227 u64 *last_pte_updated;
230 gfn_t gfn; /* presumed gfn during guest pte update */
231 struct page *page; /* page corresponding to that gfn */
234 struct i387_fxsave_struct host_fx_image;
235 struct i387_fxsave_struct guest_fx_image;
237 gva_t mmio_fault_cr2;
238 struct kvm_pio_request pio;
241 struct kvm_queued_exception {
251 struct kvm_save_segment {
256 } tr, es, ds, fs, gs;
258 int halt_request; /* real mode on Intel only */
261 struct kvm_cpuid_entry2 cpuid_entries[KVM_MAX_CPUID_ENTRIES];
262 /* emulate context */
264 struct x86_emulate_ctxt emulate_ctxt;
267 struct kvm_vcpu_time_info hv_clock;
268 unsigned int time_offset;
269 struct page *time_page;
272 struct kvm_mem_alias {
274 unsigned long npages;
280 struct kvm_mem_alias aliases[KVM_ALIAS_SLOTS];
282 unsigned int n_free_mmu_pages;
283 unsigned int n_requested_mmu_pages;
284 unsigned int n_alloc_mmu_pages;
285 struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
287 * Hash table of struct kvm_mmu_page.
289 struct list_head active_mmu_pages;
290 struct kvm_pic *vpic;
291 struct kvm_ioapic *vioapic;
293 int round_robin_prev_vcpu;
294 unsigned int tss_addr;
295 struct page *apic_access_page;
301 u32 mmu_shadow_zapped;
308 u32 remote_tlb_flush;
311 struct kvm_vcpu_stat {
321 u32 irq_window_exits;
324 u32 request_irq_exits;
326 u32 host_state_reload;
330 u32 insn_emulation_fail;
334 struct descriptor_table {
337 } __attribute__((packed));
340 int (*cpu_has_kvm_support)(void); /* __init */
341 int (*disabled_by_bios)(void); /* __init */
342 void (*hardware_enable)(void *dummy); /* __init */
343 void (*hardware_disable)(void *dummy);
344 void (*check_processor_compatibility)(void *rtn);
345 int (*hardware_setup)(void); /* __init */
346 void (*hardware_unsetup)(void); /* __exit */
347 bool (*cpu_has_accelerated_tpr)(void);
349 /* Create, but do not attach this VCPU */
350 struct kvm_vcpu *(*vcpu_create)(struct kvm *kvm, unsigned id);
351 void (*vcpu_free)(struct kvm_vcpu *vcpu);
352 int (*vcpu_reset)(struct kvm_vcpu *vcpu);
354 void (*prepare_guest_switch)(struct kvm_vcpu *vcpu);
355 void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
356 void (*vcpu_put)(struct kvm_vcpu *vcpu);
357 void (*vcpu_decache)(struct kvm_vcpu *vcpu);
359 int (*set_guest_debug)(struct kvm_vcpu *vcpu,
360 struct kvm_debug_guest *dbg);
361 void (*guest_debug_pre)(struct kvm_vcpu *vcpu);
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_cr4_guest_bits)(struct kvm_vcpu *vcpu);
371 void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
372 void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
373 void (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
374 void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
375 void (*get_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
376 void (*set_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
377 void (*get_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
378 void (*set_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
379 unsigned long (*get_dr)(struct kvm_vcpu *vcpu, int dr);
380 void (*set_dr)(struct kvm_vcpu *vcpu, int dr, unsigned long value,
382 void (*cache_regs)(struct kvm_vcpu *vcpu);
383 void (*decache_regs)(struct kvm_vcpu *vcpu);
384 unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
385 void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
387 void (*tlb_flush)(struct kvm_vcpu *vcpu);
389 void (*run)(struct kvm_vcpu *vcpu, struct kvm_run *run);
390 int (*handle_exit)(struct kvm_run *run, struct kvm_vcpu *vcpu);
391 void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
392 void (*patch_hypercall)(struct kvm_vcpu *vcpu,
393 unsigned char *hypercall_addr);
394 int (*get_irq)(struct kvm_vcpu *vcpu);
395 void (*set_irq)(struct kvm_vcpu *vcpu, int vec);
396 void (*queue_exception)(struct kvm_vcpu *vcpu, unsigned nr,
397 bool has_error_code, u32 error_code);
398 bool (*exception_injected)(struct kvm_vcpu *vcpu);
399 void (*inject_pending_irq)(struct kvm_vcpu *vcpu);
400 void (*inject_pending_vectors)(struct kvm_vcpu *vcpu,
401 struct kvm_run *run);
403 int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
406 extern struct kvm_x86_ops *kvm_x86_ops;
408 int kvm_mmu_module_init(void);
409 void kvm_mmu_module_exit(void);
411 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
412 int kvm_mmu_create(struct kvm_vcpu *vcpu);
413 int kvm_mmu_setup(struct kvm_vcpu *vcpu);
414 void kvm_mmu_set_nonpresent_ptes(u64 trap_pte, u64 notrap_pte);
416 int kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
417 void kvm_mmu_slot_remove_write_access(struct kvm *kvm, int slot);
418 void kvm_mmu_zap_all(struct kvm *kvm);
419 unsigned int kvm_mmu_calculate_mmu_pages(struct kvm *kvm);
420 void kvm_mmu_change_mmu_pages(struct kvm *kvm, unsigned int kvm_nr_mmu_pages);
422 int load_pdptrs(struct kvm_vcpu *vcpu, unsigned long cr3);
424 enum emulation_result {
425 EMULATE_DONE, /* no further processing */
426 EMULATE_DO_MMIO, /* kvm_run filled with mmio request */
427 EMULATE_FAIL, /* can't emulate this instruction */
430 #define EMULTYPE_NO_DECODE (1 << 0)
431 #define EMULTYPE_TRAP_UD (1 << 1)
432 int emulate_instruction(struct kvm_vcpu *vcpu, struct kvm_run *run,
433 unsigned long cr2, u16 error_code, int emulation_type);
434 void kvm_report_emulation_failure(struct kvm_vcpu *cvpu, const char *context);
435 void realmode_lgdt(struct kvm_vcpu *vcpu, u16 size, unsigned long address);
436 void realmode_lidt(struct kvm_vcpu *vcpu, u16 size, unsigned long address);
437 void realmode_lmsw(struct kvm_vcpu *vcpu, unsigned long msw,
438 unsigned long *rflags);
440 unsigned long realmode_get_cr(struct kvm_vcpu *vcpu, int cr);
441 void realmode_set_cr(struct kvm_vcpu *vcpu, int cr, unsigned long value,
442 unsigned long *rflags);
443 void kvm_enable_efer_bits(u64);
444 int kvm_get_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 *data);
445 int kvm_set_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 data);
447 struct x86_emulate_ctxt;
449 int kvm_emulate_pio(struct kvm_vcpu *vcpu, struct kvm_run *run, int in,
450 int size, unsigned port);
451 int kvm_emulate_pio_string(struct kvm_vcpu *vcpu, struct kvm_run *run, int in,
452 int size, unsigned long count, int down,
453 gva_t address, int rep, unsigned port);
454 void kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
455 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
456 int emulate_invlpg(struct kvm_vcpu *vcpu, gva_t address);
457 int emulate_clts(struct kvm_vcpu *vcpu);
458 int emulator_get_dr(struct x86_emulate_ctxt *ctxt, int dr,
459 unsigned long *dest);
460 int emulator_set_dr(struct x86_emulate_ctxt *ctxt, int dr,
461 unsigned long value);
463 void set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
464 void set_cr3(struct kvm_vcpu *vcpu, unsigned long cr0);
465 void set_cr4(struct kvm_vcpu *vcpu, unsigned long cr0);
466 void set_cr8(struct kvm_vcpu *vcpu, unsigned long cr0);
467 unsigned long get_cr8(struct kvm_vcpu *vcpu);
468 void lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
469 void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l);
471 int kvm_get_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata);
472 int kvm_set_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 data);
474 void kvm_queue_exception(struct kvm_vcpu *vcpu, unsigned nr);
475 void kvm_queue_exception_e(struct kvm_vcpu *vcpu, unsigned nr, u32 error_code);
476 void kvm_inject_page_fault(struct kvm_vcpu *vcpu, unsigned long cr2,
479 void fx_init(struct kvm_vcpu *vcpu);
481 int emulator_read_std(unsigned long addr,
484 struct kvm_vcpu *vcpu);
485 int emulator_write_emulated(unsigned long addr,
488 struct kvm_vcpu *vcpu);
490 unsigned long segment_base(u16 selector);
492 void kvm_mmu_flush_tlb(struct kvm_vcpu *vcpu);
493 void kvm_mmu_pte_write(struct kvm_vcpu *vcpu, gpa_t gpa,
494 const u8 *new, int bytes);
495 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva);
496 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
497 int kvm_mmu_load(struct kvm_vcpu *vcpu);
498 void kvm_mmu_unload(struct kvm_vcpu *vcpu);
500 int kvm_emulate_hypercall(struct kvm_vcpu *vcpu);
502 int kvm_fix_hypercall(struct kvm_vcpu *vcpu);
504 int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t gva, u32 error_code);
506 void kvm_enable_tdp(void);
508 int load_pdptrs(struct kvm_vcpu *vcpu, unsigned long cr3);
509 int complete_pio(struct kvm_vcpu *vcpu);
511 static inline struct kvm_mmu_page *page_header(hpa_t shadow_page)
513 struct page *page = pfn_to_page(shadow_page >> PAGE_SHIFT);
515 return (struct kvm_mmu_page *)page_private(page);
518 static inline u16 read_fs(void)
521 asm("mov %%fs, %0" : "=g"(seg));
525 static inline u16 read_gs(void)
528 asm("mov %%gs, %0" : "=g"(seg));
532 static inline u16 read_ldt(void)
535 asm("sldt %0" : "=g"(ldt));
539 static inline void load_fs(u16 sel)
541 asm("mov %0, %%fs" : : "rm"(sel));
544 static inline void load_gs(u16 sel)
546 asm("mov %0, %%gs" : : "rm"(sel));
550 static inline void load_ldt(u16 sel)
552 asm("lldt %0" : : "rm"(sel));
556 static inline void get_idt(struct descriptor_table *table)
558 asm("sidt %0" : "=m"(*table));
561 static inline void get_gdt(struct descriptor_table *table)
563 asm("sgdt %0" : "=m"(*table));
566 static inline unsigned long read_tr_base(void)
569 asm("str %0" : "=g"(tr));
570 return segment_base(tr);
574 static inline unsigned long read_msr(unsigned long msr)
583 static inline void fx_save(struct i387_fxsave_struct *image)
585 asm("fxsave (%0)":: "r" (image));
588 static inline void fx_restore(struct i387_fxsave_struct *image)
590 asm("fxrstor (%0)":: "r" (image));
593 static inline void fpu_init(void)
598 static inline u32 get_rdx_init_val(void)
600 return 0x600; /* P6 family */
603 static inline void kvm_inject_gp(struct kvm_vcpu *vcpu, u32 error_code)
605 kvm_queue_exception_e(vcpu, GP_VECTOR, error_code);
608 #define ASM_VMX_VMCLEAR_RAX ".byte 0x66, 0x0f, 0xc7, 0x30"
609 #define ASM_VMX_VMLAUNCH ".byte 0x0f, 0x01, 0xc2"
610 #define ASM_VMX_VMRESUME ".byte 0x0f, 0x01, 0xc3"
611 #define ASM_VMX_VMPTRLD_RAX ".byte 0x0f, 0xc7, 0x30"
612 #define ASM_VMX_VMREAD_RDX_RAX ".byte 0x0f, 0x78, 0xd0"
613 #define ASM_VMX_VMWRITE_RAX_RDX ".byte 0x0f, 0x79, 0xd0"
614 #define ASM_VMX_VMWRITE_RSP_RDX ".byte 0x0f, 0x79, 0xd4"
615 #define ASM_VMX_VMXOFF ".byte 0x0f, 0x01, 0xc4"
616 #define ASM_VMX_VMXON_RAX ".byte 0xf3, 0x0f, 0xc7, 0x30"
617 #define ASM_VMX_INVVPID ".byte 0x66, 0x0f, 0x38, 0x81, 0x08"
619 #define MSR_IA32_TIME_STAMP_COUNTER 0x010
621 #define TSS_IOPB_BASE_OFFSET 0x66
622 #define TSS_BASE_SIZE 0x68
623 #define TSS_IOPB_SIZE (65536 / 8)
624 #define TSS_REDIRECTION_SIZE (256 / 8)
625 #define RMODE_TSS_SIZE \
626 (TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)