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|0xFFFFFF0000000000ULL)
27 #define KVM_GUEST_CR0_MASK \
28 (X86_CR0_PG | X86_CR0_PE | X86_CR0_WP | X86_CR0_NE \
29 | X86_CR0_NW | X86_CR0_CD)
30 #define KVM_VM_CR0_ALWAYS_ON \
31 (X86_CR0_PG | X86_CR0_PE | X86_CR0_WP | X86_CR0_NE | X86_CR0_TS \
33 #define KVM_GUEST_CR4_MASK \
34 (X86_CR4_VME | X86_CR4_PSE | X86_CR4_PAE | X86_CR4_PGE | X86_CR4_VMXE)
35 #define KVM_PMODE_VM_CR4_ALWAYS_ON (X86_CR4_PAE | X86_CR4_VMXE)
36 #define KVM_RMODE_VM_CR4_ALWAYS_ON (X86_CR4_VME | X86_CR4_PAE | X86_CR4_VMXE)
38 #define INVALID_PAGE (~(hpa_t)0)
39 #define UNMAPPED_GVA (~(gpa_t)0)
51 #define SELECTOR_TI_MASK (1 << 2)
52 #define SELECTOR_RPL_MASK 0x03
56 #define KVM_ALIAS_SLOTS 4
58 #define KVM_PERMILLE_MMU_PAGES 20
59 #define KVM_MIN_ALLOC_MMU_PAGES 64
60 #define KVM_NUM_MMU_PAGES 1024
61 #define KVM_MIN_FREE_MMU_PAGES 5
62 #define KVM_REFILL_PAGES 25
63 #define KVM_MAX_CPUID_ENTRIES 40
65 extern spinlock_t kvm_lock
;
66 extern struct list_head vm_list
;
104 #include <asm/kvm_x86_emulate.h>
106 #define KVM_NR_MEM_OBJS 40
109 * We don't want allocation failures within the mmu code, so we preallocate
110 * enough memory for a single page fault in a cache.
112 struct kvm_mmu_memory_cache
{
114 void *objects
[KVM_NR_MEM_OBJS
];
117 #define NR_PTE_CHAIN_ENTRIES 5
119 struct kvm_pte_chain
{
120 u64
*parent_ptes
[NR_PTE_CHAIN_ENTRIES
];
121 struct hlist_node link
;
125 * kvm_mmu_page_role, below, is defined as:
127 * bits 0:3 - total guest paging levels (2-4, or zero for real mode)
128 * bits 4:7 - page table level for this shadow (1-4)
129 * bits 8:9 - page table quadrant for 2-level guests
130 * bit 16 - "metaphysical" - gfn is not a real page (huge page/real mode)
131 * bits 17:19 - common access permissions for all ptes in this shadow page
133 union kvm_mmu_page_role
{
136 unsigned glevels
: 4;
138 unsigned quadrant
: 2;
139 unsigned pad_for_nice_hex_output
: 6;
140 unsigned metaphysical
: 1;
145 struct kvm_mmu_page
{
146 struct list_head link
;
147 struct hlist_node hash_link
;
150 * The following two entries are used to key the shadow page in the
154 union kvm_mmu_page_role role
;
157 /* hold the gfn of each spte inside spt */
159 unsigned long slot_bitmap
; /* One bit set per slot which has memory
160 * in this shadow page.
162 int multimapped
; /* More than one parent_pte? */
163 int root_count
; /* Currently serving as active root */
165 u64
*parent_pte
; /* !multimapped */
166 struct hlist_head parent_ptes
; /* multimapped, kvm_pte_chain */
171 * x86 supports 3 paging modes (4-level 64-bit, 3-level 64-bit, and 2-level
172 * 32-bit). The kvm_mmu structure abstracts the details of the current mmu
176 void (*new_cr3
)(struct kvm_vcpu
*vcpu
);
177 int (*page_fault
)(struct kvm_vcpu
*vcpu
, gva_t gva
, u32 err
);
178 void (*free
)(struct kvm_vcpu
*vcpu
);
179 gpa_t (*gva_to_gpa
)(struct kvm_vcpu
*vcpu
, gva_t gva
);
180 void (*prefetch_page
)(struct kvm_vcpu
*vcpu
,
181 struct kvm_mmu_page
*page
);
184 int shadow_root_level
;
189 struct kvm_vcpu_arch
{
191 int interrupt_window_open
;
192 unsigned long irq_summary
; /* bit vector: 1 per word in irq_pending */
193 DECLARE_BITMAP(irq_pending
, KVM_NR_INTERRUPTS
);
194 unsigned long regs
[NR_VCPU_REGS
]; /* for rsp: vcpu_load_rsp_rip() */
195 unsigned long rip
; /* needs vcpu_load_rsp_rip() */
202 u64 pdptrs
[4]; /* pae */
205 struct kvm_lapic
*apic
; /* kernel irqchip context */
206 #define VCPU_MP_STATE_RUNNABLE 0
207 #define VCPU_MP_STATE_UNINITIALIZED 1
208 #define VCPU_MP_STATE_INIT_RECEIVED 2
209 #define VCPU_MP_STATE_SIPI_RECEIVED 3
210 #define VCPU_MP_STATE_HALTED 4
213 u64 ia32_misc_enable_msr
;
214 bool tpr_access_reporting
;
218 struct kvm_mmu_memory_cache mmu_pte_chain_cache
;
219 struct kvm_mmu_memory_cache mmu_rmap_desc_cache
;
220 struct kvm_mmu_memory_cache mmu_page_cache
;
221 struct kvm_mmu_memory_cache mmu_page_header_cache
;
223 gfn_t last_pt_write_gfn
;
224 int last_pt_write_count
;
225 u64
*last_pte_updated
;
228 gfn_t gfn
; /* presumed gfn during guest pte update */
229 struct page
*page
; /* page corresponding to that gfn */
232 struct i387_fxsave_struct host_fx_image
;
233 struct i387_fxsave_struct guest_fx_image
;
235 gva_t mmio_fault_cr2
;
236 struct kvm_pio_request pio
;
239 struct kvm_queued_exception
{
249 struct kvm_save_segment
{
254 } tr
, es
, ds
, fs
, gs
;
256 int halt_request
; /* real mode on Intel only */
259 struct kvm_cpuid_entry2 cpuid_entries
[KVM_MAX_CPUID_ENTRIES
];
260 /* emulate context */
262 struct x86_emulate_ctxt emulate_ctxt
;
265 struct kvm_mem_alias
{
267 unsigned long npages
;
273 struct kvm_mem_alias aliases
[KVM_ALIAS_SLOTS
];
275 unsigned int n_free_mmu_pages
;
276 unsigned int n_requested_mmu_pages
;
277 unsigned int n_alloc_mmu_pages
;
278 struct hlist_head mmu_page_hash
[KVM_NUM_MMU_PAGES
];
280 * Hash table of struct kvm_mmu_page.
282 struct list_head active_mmu_pages
;
283 struct kvm_pic
*vpic
;
284 struct kvm_ioapic
*vioapic
;
286 int round_robin_prev_vcpu
;
287 unsigned int tss_addr
;
288 struct page
*apic_access_page
;
292 u32 mmu_shadow_zapped
;
299 u32 remote_tlb_flush
;
302 struct kvm_vcpu_stat
{
312 u32 irq_window_exits
;
315 u32 request_irq_exits
;
317 u32 host_state_reload
;
321 u32 insn_emulation_fail
;
324 struct descriptor_table
{
327 } __attribute__((packed
));
330 int (*cpu_has_kvm_support
)(void); /* __init */
331 int (*disabled_by_bios
)(void); /* __init */
332 void (*hardware_enable
)(void *dummy
); /* __init */
333 void (*hardware_disable
)(void *dummy
);
334 void (*check_processor_compatibility
)(void *rtn
);
335 int (*hardware_setup
)(void); /* __init */
336 void (*hardware_unsetup
)(void); /* __exit */
337 bool (*cpu_has_accelerated_tpr
)(void);
339 /* Create, but do not attach this VCPU */
340 struct kvm_vcpu
*(*vcpu_create
)(struct kvm
*kvm
, unsigned id
);
341 void (*vcpu_free
)(struct kvm_vcpu
*vcpu
);
342 int (*vcpu_reset
)(struct kvm_vcpu
*vcpu
);
344 void (*prepare_guest_switch
)(struct kvm_vcpu
*vcpu
);
345 void (*vcpu_load
)(struct kvm_vcpu
*vcpu
, int cpu
);
346 void (*vcpu_put
)(struct kvm_vcpu
*vcpu
);
347 void (*vcpu_decache
)(struct kvm_vcpu
*vcpu
);
349 int (*set_guest_debug
)(struct kvm_vcpu
*vcpu
,
350 struct kvm_debug_guest
*dbg
);
351 void (*guest_debug_pre
)(struct kvm_vcpu
*vcpu
);
352 int (*get_msr
)(struct kvm_vcpu
*vcpu
, u32 msr_index
, u64
*pdata
);
353 int (*set_msr
)(struct kvm_vcpu
*vcpu
, u32 msr_index
, u64 data
);
354 u64 (*get_segment_base
)(struct kvm_vcpu
*vcpu
, int seg
);
355 void (*get_segment
)(struct kvm_vcpu
*vcpu
,
356 struct kvm_segment
*var
, int seg
);
357 void (*set_segment
)(struct kvm_vcpu
*vcpu
,
358 struct kvm_segment
*var
, int seg
);
359 void (*get_cs_db_l_bits
)(struct kvm_vcpu
*vcpu
, int *db
, int *l
);
360 void (*decache_cr4_guest_bits
)(struct kvm_vcpu
*vcpu
);
361 void (*set_cr0
)(struct kvm_vcpu
*vcpu
, unsigned long cr0
);
362 void (*set_cr3
)(struct kvm_vcpu
*vcpu
, unsigned long cr3
);
363 void (*set_cr4
)(struct kvm_vcpu
*vcpu
, unsigned long cr4
);
364 void (*set_efer
)(struct kvm_vcpu
*vcpu
, u64 efer
);
365 void (*get_idt
)(struct kvm_vcpu
*vcpu
, struct descriptor_table
*dt
);
366 void (*set_idt
)(struct kvm_vcpu
*vcpu
, struct descriptor_table
*dt
);
367 void (*get_gdt
)(struct kvm_vcpu
*vcpu
, struct descriptor_table
*dt
);
368 void (*set_gdt
)(struct kvm_vcpu
*vcpu
, struct descriptor_table
*dt
);
369 unsigned long (*get_dr
)(struct kvm_vcpu
*vcpu
, int dr
);
370 void (*set_dr
)(struct kvm_vcpu
*vcpu
, int dr
, unsigned long value
,
372 void (*cache_regs
)(struct kvm_vcpu
*vcpu
);
373 void (*decache_regs
)(struct kvm_vcpu
*vcpu
);
374 unsigned long (*get_rflags
)(struct kvm_vcpu
*vcpu
);
375 void (*set_rflags
)(struct kvm_vcpu
*vcpu
, unsigned long rflags
);
377 void (*tlb_flush
)(struct kvm_vcpu
*vcpu
);
379 void (*run
)(struct kvm_vcpu
*vcpu
, struct kvm_run
*run
);
380 int (*handle_exit
)(struct kvm_run
*run
, struct kvm_vcpu
*vcpu
);
381 void (*skip_emulated_instruction
)(struct kvm_vcpu
*vcpu
);
382 void (*patch_hypercall
)(struct kvm_vcpu
*vcpu
,
383 unsigned char *hypercall_addr
);
384 int (*get_irq
)(struct kvm_vcpu
*vcpu
);
385 void (*set_irq
)(struct kvm_vcpu
*vcpu
, int vec
);
386 void (*queue_exception
)(struct kvm_vcpu
*vcpu
, unsigned nr
,
387 bool has_error_code
, u32 error_code
);
388 bool (*exception_injected
)(struct kvm_vcpu
*vcpu
);
389 void (*inject_pending_irq
)(struct kvm_vcpu
*vcpu
);
390 void (*inject_pending_vectors
)(struct kvm_vcpu
*vcpu
,
391 struct kvm_run
*run
);
393 int (*set_tss_addr
)(struct kvm
*kvm
, unsigned int addr
);
396 extern struct kvm_x86_ops
*kvm_x86_ops
;
398 int kvm_mmu_module_init(void);
399 void kvm_mmu_module_exit(void);
401 void kvm_mmu_destroy(struct kvm_vcpu
*vcpu
);
402 int kvm_mmu_create(struct kvm_vcpu
*vcpu
);
403 int kvm_mmu_setup(struct kvm_vcpu
*vcpu
);
404 void kvm_mmu_set_nonpresent_ptes(u64 trap_pte
, u64 notrap_pte
);
406 int kvm_mmu_reset_context(struct kvm_vcpu
*vcpu
);
407 void kvm_mmu_slot_remove_write_access(struct kvm
*kvm
, int slot
);
408 void kvm_mmu_zap_all(struct kvm
*kvm
);
409 unsigned int kvm_mmu_calculate_mmu_pages(struct kvm
*kvm
);
410 void kvm_mmu_change_mmu_pages(struct kvm
*kvm
, unsigned int kvm_nr_mmu_pages
);
412 enum emulation_result
{
413 EMULATE_DONE
, /* no further processing */
414 EMULATE_DO_MMIO
, /* kvm_run filled with mmio request */
415 EMULATE_FAIL
, /* can't emulate this instruction */
418 #define EMULTYPE_NO_DECODE (1 << 0)
419 #define EMULTYPE_TRAP_UD (1 << 1)
420 int emulate_instruction(struct kvm_vcpu
*vcpu
, struct kvm_run
*run
,
421 unsigned long cr2
, u16 error_code
, int emulation_type
);
422 void kvm_report_emulation_failure(struct kvm_vcpu
*cvpu
, const char *context
);
423 void realmode_lgdt(struct kvm_vcpu
*vcpu
, u16 size
, unsigned long address
);
424 void realmode_lidt(struct kvm_vcpu
*vcpu
, u16 size
, unsigned long address
);
425 void realmode_lmsw(struct kvm_vcpu
*vcpu
, unsigned long msw
,
426 unsigned long *rflags
);
428 unsigned long realmode_get_cr(struct kvm_vcpu
*vcpu
, int cr
);
429 void realmode_set_cr(struct kvm_vcpu
*vcpu
, int cr
, unsigned long value
,
430 unsigned long *rflags
);
431 int kvm_get_msr(struct kvm_vcpu
*vcpu
, u32 msr_index
, u64
*data
);
432 int kvm_set_msr(struct kvm_vcpu
*vcpu
, u32 msr_index
, u64 data
);
434 struct x86_emulate_ctxt
;
436 int kvm_emulate_pio(struct kvm_vcpu
*vcpu
, struct kvm_run
*run
, int in
,
437 int size
, unsigned port
);
438 int kvm_emulate_pio_string(struct kvm_vcpu
*vcpu
, struct kvm_run
*run
, int in
,
439 int size
, unsigned long count
, int down
,
440 gva_t address
, int rep
, unsigned port
);
441 void kvm_emulate_cpuid(struct kvm_vcpu
*vcpu
);
442 int kvm_emulate_halt(struct kvm_vcpu
*vcpu
);
443 int emulate_invlpg(struct kvm_vcpu
*vcpu
, gva_t address
);
444 int emulate_clts(struct kvm_vcpu
*vcpu
);
445 int emulator_get_dr(struct x86_emulate_ctxt
*ctxt
, int dr
,
446 unsigned long *dest
);
447 int emulator_set_dr(struct x86_emulate_ctxt
*ctxt
, int dr
,
448 unsigned long value
);
450 void set_cr0(struct kvm_vcpu
*vcpu
, unsigned long cr0
);
451 void set_cr3(struct kvm_vcpu
*vcpu
, unsigned long cr0
);
452 void set_cr4(struct kvm_vcpu
*vcpu
, unsigned long cr0
);
453 void set_cr8(struct kvm_vcpu
*vcpu
, unsigned long cr0
);
454 unsigned long get_cr8(struct kvm_vcpu
*vcpu
);
455 void lmsw(struct kvm_vcpu
*vcpu
, unsigned long msw
);
456 void kvm_get_cs_db_l_bits(struct kvm_vcpu
*vcpu
, int *db
, int *l
);
458 int kvm_get_msr_common(struct kvm_vcpu
*vcpu
, u32 msr
, u64
*pdata
);
459 int kvm_set_msr_common(struct kvm_vcpu
*vcpu
, u32 msr
, u64 data
);
461 void kvm_queue_exception(struct kvm_vcpu
*vcpu
, unsigned nr
);
462 void kvm_queue_exception_e(struct kvm_vcpu
*vcpu
, unsigned nr
, u32 error_code
);
463 void kvm_inject_page_fault(struct kvm_vcpu
*vcpu
, unsigned long cr2
,
466 void fx_init(struct kvm_vcpu
*vcpu
);
468 int emulator_read_std(unsigned long addr
,
471 struct kvm_vcpu
*vcpu
);
472 int emulator_write_emulated(unsigned long addr
,
475 struct kvm_vcpu
*vcpu
);
477 unsigned long segment_base(u16 selector
);
479 void kvm_mmu_flush_tlb(struct kvm_vcpu
*vcpu
);
480 void kvm_mmu_pte_write(struct kvm_vcpu
*vcpu
, gpa_t gpa
,
481 const u8
*new, int bytes
);
482 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu
*vcpu
, gva_t gva
);
483 void __kvm_mmu_free_some_pages(struct kvm_vcpu
*vcpu
);
484 int kvm_mmu_load(struct kvm_vcpu
*vcpu
);
485 void kvm_mmu_unload(struct kvm_vcpu
*vcpu
);
487 int kvm_emulate_hypercall(struct kvm_vcpu
*vcpu
);
489 int kvm_fix_hypercall(struct kvm_vcpu
*vcpu
);
491 int kvm_mmu_page_fault(struct kvm_vcpu
*vcpu
, gva_t gva
, u32 error_code
);
493 int load_pdptrs(struct kvm_vcpu
*vcpu
, unsigned long cr3
);
494 int complete_pio(struct kvm_vcpu
*vcpu
);
496 static inline struct kvm_mmu_page
*page_header(hpa_t shadow_page
)
498 struct page
*page
= pfn_to_page(shadow_page
>> PAGE_SHIFT
);
500 return (struct kvm_mmu_page
*)page_private(page
);
503 static inline u16
read_fs(void)
506 asm("mov %%fs, %0" : "=g"(seg
));
510 static inline u16
read_gs(void)
513 asm("mov %%gs, %0" : "=g"(seg
));
517 static inline u16
read_ldt(void)
520 asm("sldt %0" : "=g"(ldt
));
524 static inline void load_fs(u16 sel
)
526 asm("mov %0, %%fs" : : "rm"(sel
));
529 static inline void load_gs(u16 sel
)
531 asm("mov %0, %%gs" : : "rm"(sel
));
535 static inline void load_ldt(u16 sel
)
537 asm("lldt %0" : : "rm"(sel
));
541 static inline void get_idt(struct descriptor_table
*table
)
543 asm("sidt %0" : "=m"(*table
));
546 static inline void get_gdt(struct descriptor_table
*table
)
548 asm("sgdt %0" : "=m"(*table
));
551 static inline unsigned long read_tr_base(void)
554 asm("str %0" : "=g"(tr
));
555 return segment_base(tr
);
559 static inline unsigned long read_msr(unsigned long msr
)
568 static inline void fx_save(struct i387_fxsave_struct
*image
)
570 asm("fxsave (%0)":: "r" (image
));
573 static inline void fx_restore(struct i387_fxsave_struct
*image
)
575 asm("fxrstor (%0)":: "r" (image
));
578 static inline void fpu_init(void)
583 static inline u32
get_rdx_init_val(void)
585 return 0x600; /* P6 family */
588 static inline void kvm_inject_gp(struct kvm_vcpu
*vcpu
, u32 error_code
)
590 kvm_queue_exception_e(vcpu
, GP_VECTOR
, error_code
);
593 #define ASM_VMX_VMCLEAR_RAX ".byte 0x66, 0x0f, 0xc7, 0x30"
594 #define ASM_VMX_VMLAUNCH ".byte 0x0f, 0x01, 0xc2"
595 #define ASM_VMX_VMRESUME ".byte 0x0f, 0x01, 0xc3"
596 #define ASM_VMX_VMPTRLD_RAX ".byte 0x0f, 0xc7, 0x30"
597 #define ASM_VMX_VMREAD_RDX_RAX ".byte 0x0f, 0x78, 0xd0"
598 #define ASM_VMX_VMWRITE_RAX_RDX ".byte 0x0f, 0x79, 0xd0"
599 #define ASM_VMX_VMWRITE_RSP_RDX ".byte 0x0f, 0x79, 0xd4"
600 #define ASM_VMX_VMXOFF ".byte 0x0f, 0x01, 0xc4"
601 #define ASM_VMX_VMXON_RAX ".byte 0xf3, 0x0f, 0xc7, 0x30"
603 #define MSR_IA32_TIME_STAMP_COUNTER 0x010
605 #define TSS_IOPB_BASE_OFFSET 0x66
606 #define TSS_BASE_SIZE 0x68
607 #define TSS_IOPB_SIZE (65536 / 8)
608 #define TSS_REDIRECTION_SIZE (256 / 8)
609 #define RMODE_TSS_SIZE (TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)