5 * This work is licensed under the terms of the GNU GPL, version 2. See
6 * the COPYING file in the top-level directory.
9 #include <linux/types.h>
10 #include <linux/list.h>
11 #include <linux/mutex.h>
12 #include <linux/spinlock.h>
16 #include <linux/kvm.h>
17 #include <linux/kvm_para.h>
19 #define CR0_PE_MASK (1ULL << 0)
20 #define CR0_TS_MASK (1ULL << 3)
21 #define CR0_NE_MASK (1ULL << 5)
22 #define CR0_WP_MASK (1ULL << 16)
23 #define CR0_NW_MASK (1ULL << 29)
24 #define CR0_CD_MASK (1ULL << 30)
25 #define CR0_PG_MASK (1ULL << 31)
27 #define CR3_WPT_MASK (1ULL << 3)
28 #define CR3_PCD_MASK (1ULL << 4)
30 #define CR3_RESEVED_BITS 0x07ULL
31 #define CR3_L_MODE_RESEVED_BITS (~((1ULL << 40) - 1) | 0x0fe7ULL)
32 #define CR3_FLAGS_MASK ((1ULL << 5) - 1)
34 #define CR4_VME_MASK (1ULL << 0)
35 #define CR4_PSE_MASK (1ULL << 4)
36 #define CR4_PAE_MASK (1ULL << 5)
37 #define CR4_PGE_MASK (1ULL << 7)
38 #define CR4_VMXE_MASK (1ULL << 13)
40 #define KVM_GUEST_CR0_MASK \
41 (CR0_PG_MASK | CR0_PE_MASK | CR0_WP_MASK | CR0_NE_MASK \
42 | CR0_NW_MASK | CR0_CD_MASK)
43 #define KVM_VM_CR0_ALWAYS_ON \
44 (CR0_PG_MASK | CR0_PE_MASK | CR0_WP_MASK | CR0_NE_MASK)
45 #define KVM_GUEST_CR4_MASK \
46 (CR4_PSE_MASK | CR4_PAE_MASK | CR4_PGE_MASK | CR4_VMXE_MASK | CR4_VME_MASK)
47 #define KVM_PMODE_VM_CR4_ALWAYS_ON (CR4_VMXE_MASK | CR4_PAE_MASK)
48 #define KVM_RMODE_VM_CR4_ALWAYS_ON (CR4_VMXE_MASK | CR4_PAE_MASK | CR4_VME_MASK)
50 #define INVALID_PAGE (~(hpa_t)0)
51 #define UNMAPPED_GVA (~(gpa_t)0)
53 #define KVM_MAX_VCPUS 1
54 #define KVM_ALIAS_SLOTS 4
55 #define KVM_MEMORY_SLOTS 4
56 #define KVM_NUM_MMU_PAGES 256
57 #define KVM_MIN_FREE_MMU_PAGES 5
58 #define KVM_REFILL_PAGES 25
59 #define KVM_MAX_CPUID_ENTRIES 40
61 #define FX_IMAGE_SIZE 512
62 #define FX_IMAGE_ALIGN 16
63 #define FX_BUF_SIZE (2 * FX_IMAGE_SIZE + FX_IMAGE_ALIGN)
74 #define SELECTOR_TI_MASK (1 << 2)
75 #define SELECTOR_RPL_MASK 0x03
79 #define KVM_PIO_PAGE_OFFSET 1
84 * gva - guest virtual address
85 * gpa - guest physical address
86 * gfn - guest frame number
87 * hva - host virtual address
88 * hpa - host physical address
89 * hfn - host frame number
92 typedef unsigned long gva_t
;
94 typedef unsigned long gfn_t
;
96 typedef unsigned long hva_t
;
98 typedef unsigned long hfn_t
;
100 #define NR_PTE_CHAIN_ENTRIES 5
102 struct kvm_pte_chain
{
103 u64
*parent_ptes
[NR_PTE_CHAIN_ENTRIES
];
104 struct hlist_node link
;
108 * kvm_mmu_page_role, below, is defined as:
110 * bits 0:3 - total guest paging levels (2-4, or zero for real mode)
111 * bits 4:7 - page table level for this shadow (1-4)
112 * bits 8:9 - page table quadrant for 2-level guests
113 * bit 16 - "metaphysical" - gfn is not a real page (huge page/real mode)
114 * bits 17:18 - "access" - the user and writable bits of a huge page pde
116 union kvm_mmu_page_role
{
119 unsigned glevels
: 4;
121 unsigned quadrant
: 2;
122 unsigned pad_for_nice_hex_output
: 6;
123 unsigned metaphysical
: 1;
124 unsigned hugepage_access
: 2;
128 struct kvm_mmu_page
{
129 struct list_head link
;
130 struct hlist_node hash_link
;
133 * The following two entries are used to key the shadow page in the
137 union kvm_mmu_page_role role
;
140 unsigned long slot_bitmap
; /* One bit set per slot which has memory
141 * in this shadow page.
143 int multimapped
; /* More than one parent_pte? */
144 int root_count
; /* Currently serving as active root */
146 u64
*parent_pte
; /* !multimapped */
147 struct hlist_head parent_ptes
; /* multimapped, kvm_pte_chain */
157 #define vmx_msr_entry kvm_msr_entry
162 * x86 supports 3 paging modes (4-level 64-bit, 3-level 64-bit, and 2-level
163 * 32-bit). The kvm_mmu structure abstracts the details of the current mmu
167 void (*new_cr3
)(struct kvm_vcpu
*vcpu
);
168 int (*page_fault
)(struct kvm_vcpu
*vcpu
, gva_t gva
, u32 err
);
169 void (*free
)(struct kvm_vcpu
*vcpu
);
170 gpa_t (*gva_to_gpa
)(struct kvm_vcpu
*vcpu
, gva_t gva
);
173 int shadow_root_level
;
178 #define KVM_NR_MEM_OBJS 20
180 struct kvm_mmu_memory_cache
{
182 void *objects
[KVM_NR_MEM_OBJS
];
186 * We don't want allocation failures within the mmu code, so we preallocate
187 * enough memory for a single page fault in a cache.
189 struct kvm_guest_debug
{
228 struct kvm_pio_request
{
231 struct page
*guest_pages
[2];
232 unsigned guest_page_offset
;
250 u32 irq_window_exits
;
252 u32 request_irq_exits
;
260 struct vcpu_svm
*svm
;
267 int interrupt_window_open
;
268 unsigned long irq_summary
; /* bit vector: 1 per word in irq_pending */
269 #define NR_IRQ_WORDS KVM_IRQ_BITMAP_SIZE(unsigned long)
270 unsigned long irq_pending
[NR_IRQ_WORDS
];
271 unsigned long regs
[NR_VCPU_REGS
]; /* for rsp: vcpu_load_rsp_rip() */
272 unsigned long rip
; /* needs vcpu_load_rsp_rip() */
277 gpa_t para_state_gpa
;
278 struct page
*para_state_page
;
282 u64 pdptrs
[4]; /* pae */
285 u64 ia32_misc_enable_msr
;
287 struct vmx_msr_entry
*guest_msrs
;
288 struct vmx_msr_entry
*host_msrs
;
290 struct list_head free_pages
;
291 struct kvm_mmu_page page_header_buf
[KVM_NUM_MMU_PAGES
];
294 struct kvm_mmu_memory_cache mmu_pte_chain_cache
;
295 struct kvm_mmu_memory_cache mmu_rmap_desc_cache
;
297 gfn_t last_pt_write_gfn
;
298 int last_pt_write_count
;
300 struct kvm_guest_debug guest_debug
;
302 char fx_buf
[FX_BUF_SIZE
];
304 char *guest_fx_image
;
308 int mmio_read_completed
;
311 unsigned char mmio_data
[8];
312 gpa_t mmio_phys_addr
;
313 gva_t mmio_fault_cr2
;
314 struct kvm_pio_request pio
;
320 struct kvm_stat stat
;
325 struct kvm_save_segment
{
330 } tr
, es
, ds
, fs
, gs
;
334 struct kvm_cpuid_entry cpuid_entries
[KVM_MAX_CPUID_ENTRIES
];
337 struct kvm_mem_alias
{
339 unsigned long npages
;
343 struct kvm_memory_slot
{
345 unsigned long npages
;
347 struct page
**phys_mem
;
348 unsigned long *dirty_bitmap
;
352 spinlock_t lock
; /* protects everything except vcpus */
354 struct kvm_mem_alias aliases
[KVM_ALIAS_SLOTS
];
356 struct kvm_memory_slot memslots
[KVM_MEMORY_SLOTS
];
358 * Hash table of struct kvm_mmu_page.
360 struct list_head active_mmu_pages
;
361 int n_free_mmu_pages
;
362 struct hlist_head mmu_page_hash
[KVM_NUM_MMU_PAGES
];
363 struct kvm_vcpu vcpus
[KVM_MAX_VCPUS
];
364 int memory_config_version
;
366 unsigned long rmap_overflow
;
367 struct list_head vm_list
;
371 struct descriptor_table
{
374 } __attribute__((packed
));
376 struct kvm_arch_ops
{
377 int (*cpu_has_kvm_support
)(void); /* __init */
378 int (*disabled_by_bios
)(void); /* __init */
379 void (*hardware_enable
)(void *dummy
); /* __init */
380 void (*hardware_disable
)(void *dummy
);
381 int (*hardware_setup
)(void); /* __init */
382 void (*hardware_unsetup
)(void); /* __exit */
384 int (*vcpu_create
)(struct kvm_vcpu
*vcpu
);
385 void (*vcpu_free
)(struct kvm_vcpu
*vcpu
);
387 void (*vcpu_load
)(struct kvm_vcpu
*vcpu
);
388 void (*vcpu_put
)(struct kvm_vcpu
*vcpu
);
389 void (*vcpu_decache
)(struct kvm_vcpu
*vcpu
);
391 int (*set_guest_debug
)(struct kvm_vcpu
*vcpu
,
392 struct kvm_debug_guest
*dbg
);
393 int (*get_msr
)(struct kvm_vcpu
*vcpu
, u32 msr_index
, u64
*pdata
);
394 int (*set_msr
)(struct kvm_vcpu
*vcpu
, u32 msr_index
, u64 data
);
395 u64 (*get_segment_base
)(struct kvm_vcpu
*vcpu
, int seg
);
396 void (*get_segment
)(struct kvm_vcpu
*vcpu
,
397 struct kvm_segment
*var
, int seg
);
398 void (*set_segment
)(struct kvm_vcpu
*vcpu
,
399 struct kvm_segment
*var
, int seg
);
400 void (*get_cs_db_l_bits
)(struct kvm_vcpu
*vcpu
, int *db
, int *l
);
401 void (*decache_cr4_guest_bits
)(struct kvm_vcpu
*vcpu
);
402 void (*set_cr0
)(struct kvm_vcpu
*vcpu
, unsigned long cr0
);
403 void (*set_cr3
)(struct kvm_vcpu
*vcpu
, unsigned long cr3
);
404 void (*set_cr4
)(struct kvm_vcpu
*vcpu
, unsigned long cr4
);
405 void (*set_efer
)(struct kvm_vcpu
*vcpu
, u64 efer
);
406 void (*get_idt
)(struct kvm_vcpu
*vcpu
, struct descriptor_table
*dt
);
407 void (*set_idt
)(struct kvm_vcpu
*vcpu
, struct descriptor_table
*dt
);
408 void (*get_gdt
)(struct kvm_vcpu
*vcpu
, struct descriptor_table
*dt
);
409 void (*set_gdt
)(struct kvm_vcpu
*vcpu
, struct descriptor_table
*dt
);
410 unsigned long (*get_dr
)(struct kvm_vcpu
*vcpu
, int dr
);
411 void (*set_dr
)(struct kvm_vcpu
*vcpu
, int dr
, unsigned long value
,
413 void (*cache_regs
)(struct kvm_vcpu
*vcpu
);
414 void (*decache_regs
)(struct kvm_vcpu
*vcpu
);
415 unsigned long (*get_rflags
)(struct kvm_vcpu
*vcpu
);
416 void (*set_rflags
)(struct kvm_vcpu
*vcpu
, unsigned long rflags
);
418 void (*invlpg
)(struct kvm_vcpu
*vcpu
, gva_t addr
);
419 void (*tlb_flush
)(struct kvm_vcpu
*vcpu
);
420 void (*inject_page_fault
)(struct kvm_vcpu
*vcpu
,
421 unsigned long addr
, u32 err_code
);
423 void (*inject_gp
)(struct kvm_vcpu
*vcpu
, unsigned err_code
);
425 int (*run
)(struct kvm_vcpu
*vcpu
, struct kvm_run
*run
);
426 int (*vcpu_setup
)(struct kvm_vcpu
*vcpu
);
427 void (*skip_emulated_instruction
)(struct kvm_vcpu
*vcpu
);
428 void (*patch_hypercall
)(struct kvm_vcpu
*vcpu
,
429 unsigned char *hypercall_addr
);
432 extern struct kvm_arch_ops
*kvm_arch_ops
;
434 #define kvm_printf(kvm, fmt ...) printk(KERN_DEBUG fmt)
435 #define vcpu_printf(vcpu, fmt...) kvm_printf(vcpu->kvm, fmt)
437 int kvm_init_arch(struct kvm_arch_ops
*ops
, struct module
*module
);
438 void kvm_exit_arch(void);
440 int kvm_mmu_module_init(void);
441 void kvm_mmu_module_exit(void);
443 void kvm_mmu_destroy(struct kvm_vcpu
*vcpu
);
444 int kvm_mmu_create(struct kvm_vcpu
*vcpu
);
445 int kvm_mmu_setup(struct kvm_vcpu
*vcpu
);
447 int kvm_mmu_reset_context(struct kvm_vcpu
*vcpu
);
448 void kvm_mmu_slot_remove_write_access(struct kvm_vcpu
*vcpu
, int slot
);
449 void kvm_mmu_zap_all(struct kvm_vcpu
*vcpu
);
451 hpa_t
gpa_to_hpa(struct kvm_vcpu
*vcpu
, gpa_t gpa
);
452 #define HPA_MSB ((sizeof(hpa_t) * 8) - 1)
453 #define HPA_ERR_MASK ((hpa_t)1 << HPA_MSB)
454 static inline int is_error_hpa(hpa_t hpa
) { return hpa
>> HPA_MSB
; }
455 hpa_t
gva_to_hpa(struct kvm_vcpu
*vcpu
, gva_t gva
);
456 struct page
*gva_to_page(struct kvm_vcpu
*vcpu
, gva_t gva
);
458 void kvm_emulator_want_group7_invlpg(void);
460 extern hpa_t bad_page_address
;
462 struct page
*gfn_to_page(struct kvm
*kvm
, gfn_t gfn
);
463 struct kvm_memory_slot
*gfn_to_memslot(struct kvm
*kvm
, gfn_t gfn
);
464 void mark_page_dirty(struct kvm
*kvm
, gfn_t gfn
);
466 enum emulation_result
{
467 EMULATE_DONE
, /* no further processing */
468 EMULATE_DO_MMIO
, /* kvm_run filled with mmio request */
469 EMULATE_FAIL
, /* can't emulate this instruction */
472 int emulate_instruction(struct kvm_vcpu
*vcpu
, struct kvm_run
*run
,
473 unsigned long cr2
, u16 error_code
);
474 void realmode_lgdt(struct kvm_vcpu
*vcpu
, u16 size
, unsigned long address
);
475 void realmode_lidt(struct kvm_vcpu
*vcpu
, u16 size
, unsigned long address
);
476 void realmode_lmsw(struct kvm_vcpu
*vcpu
, unsigned long msw
,
477 unsigned long *rflags
);
479 unsigned long realmode_get_cr(struct kvm_vcpu
*vcpu
, int cr
);
480 void realmode_set_cr(struct kvm_vcpu
*vcpu
, int cr
, unsigned long value
,
481 unsigned long *rflags
);
483 struct x86_emulate_ctxt
;
485 int kvm_setup_pio(struct kvm_vcpu
*vcpu
, struct kvm_run
*run
, int in
,
486 int size
, unsigned long count
, int string
, int down
,
487 gva_t address
, int rep
, unsigned port
);
488 void kvm_emulate_cpuid(struct kvm_vcpu
*vcpu
);
489 int emulate_invlpg(struct kvm_vcpu
*vcpu
, gva_t address
);
490 int emulate_clts(struct kvm_vcpu
*vcpu
);
491 int emulator_get_dr(struct x86_emulate_ctxt
* ctxt
, int dr
,
492 unsigned long *dest
);
493 int emulator_set_dr(struct x86_emulate_ctxt
*ctxt
, int dr
,
494 unsigned long value
);
496 void set_cr0(struct kvm_vcpu
*vcpu
, unsigned long cr0
);
497 void set_cr3(struct kvm_vcpu
*vcpu
, unsigned long cr0
);
498 void set_cr4(struct kvm_vcpu
*vcpu
, unsigned long cr0
);
499 void set_cr8(struct kvm_vcpu
*vcpu
, unsigned long cr0
);
500 void lmsw(struct kvm_vcpu
*vcpu
, unsigned long msw
);
502 int kvm_get_msr_common(struct kvm_vcpu
*vcpu
, u32 msr
, u64
*pdata
);
503 int kvm_set_msr_common(struct kvm_vcpu
*vcpu
, u32 msr
, u64 data
);
505 void fx_init(struct kvm_vcpu
*vcpu
);
507 void load_msrs(struct vmx_msr_entry
*e
, int n
);
508 void save_msrs(struct vmx_msr_entry
*e
, int n
);
509 void kvm_resched(struct kvm_vcpu
*vcpu
);
511 int kvm_read_guest(struct kvm_vcpu
*vcpu
,
516 int kvm_write_guest(struct kvm_vcpu
*vcpu
,
521 unsigned long segment_base(u16 selector
);
523 void kvm_mmu_pre_write(struct kvm_vcpu
*vcpu
, gpa_t gpa
, int bytes
);
524 void kvm_mmu_post_write(struct kvm_vcpu
*vcpu
, gpa_t gpa
, int bytes
);
525 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu
*vcpu
, gva_t gva
);
526 void kvm_mmu_free_some_pages(struct kvm_vcpu
*vcpu
);
528 int kvm_hypercall(struct kvm_vcpu
*vcpu
, struct kvm_run
*run
);
530 static inline int kvm_mmu_page_fault(struct kvm_vcpu
*vcpu
, gva_t gva
,
533 if (unlikely(vcpu
->kvm
->n_free_mmu_pages
< KVM_MIN_FREE_MMU_PAGES
))
534 kvm_mmu_free_some_pages(vcpu
);
535 return vcpu
->mmu
.page_fault(vcpu
, gva
, error_code
);
538 static inline int is_long_mode(struct kvm_vcpu
*vcpu
)
541 return vcpu
->shadow_efer
& EFER_LME
;
547 static inline int is_pae(struct kvm_vcpu
*vcpu
)
549 return vcpu
->cr4
& CR4_PAE_MASK
;
552 static inline int is_pse(struct kvm_vcpu
*vcpu
)
554 return vcpu
->cr4
& CR4_PSE_MASK
;
557 static inline int is_paging(struct kvm_vcpu
*vcpu
)
559 return vcpu
->cr0
& CR0_PG_MASK
;
562 static inline int memslot_id(struct kvm
*kvm
, struct kvm_memory_slot
*slot
)
564 return slot
- kvm
->memslots
;
567 static inline struct kvm_mmu_page
*page_header(hpa_t shadow_page
)
569 struct page
*page
= pfn_to_page(shadow_page
>> PAGE_SHIFT
);
571 return (struct kvm_mmu_page
*)page_private(page
);
574 static inline u16
read_fs(void)
577 asm ("mov %%fs, %0" : "=g"(seg
));
581 static inline u16
read_gs(void)
584 asm ("mov %%gs, %0" : "=g"(seg
));
588 static inline u16
read_ldt(void)
591 asm ("sldt %0" : "=g"(ldt
));
595 static inline void load_fs(u16 sel
)
597 asm ("mov %0, %%fs" : : "rm"(sel
));
600 static inline void load_gs(u16 sel
)
602 asm ("mov %0, %%gs" : : "rm"(sel
));
606 static inline void load_ldt(u16 sel
)
608 asm ("lldt %0" : : "rm"(sel
));
612 static inline void get_idt(struct descriptor_table
*table
)
614 asm ("sidt %0" : "=m"(*table
));
617 static inline void get_gdt(struct descriptor_table
*table
)
619 asm ("sgdt %0" : "=m"(*table
));
622 static inline unsigned long read_tr_base(void)
625 asm ("str %0" : "=g"(tr
));
626 return segment_base(tr
);
630 static inline unsigned long read_msr(unsigned long msr
)
639 static inline void fx_save(void *image
)
641 asm ("fxsave (%0)":: "r" (image
));
644 static inline void fx_restore(void *image
)
646 asm ("fxrstor (%0)":: "r" (image
));
649 static inline void fpu_init(void)
654 static inline u32
get_rdx_init_val(void)
656 return 0x600; /* P6 family */
659 #define ASM_VMX_VMCLEAR_RAX ".byte 0x66, 0x0f, 0xc7, 0x30"
660 #define ASM_VMX_VMLAUNCH ".byte 0x0f, 0x01, 0xc2"
661 #define ASM_VMX_VMRESUME ".byte 0x0f, 0x01, 0xc3"
662 #define ASM_VMX_VMPTRLD_RAX ".byte 0x0f, 0xc7, 0x30"
663 #define ASM_VMX_VMREAD_RDX_RAX ".byte 0x0f, 0x78, 0xd0"
664 #define ASM_VMX_VMWRITE_RAX_RDX ".byte 0x0f, 0x79, 0xd0"
665 #define ASM_VMX_VMWRITE_RSP_RDX ".byte 0x0f, 0x79, 0xd4"
666 #define ASM_VMX_VMXOFF ".byte 0x0f, 0x01, 0xc4"
667 #define ASM_VMX_VMXON_RAX ".byte 0xf3, 0x0f, 0xc7, 0x30"
669 #define MSR_IA32_TIME_STAMP_COUNTER 0x010
671 #define TSS_IOPB_BASE_OFFSET 0x66
672 #define TSS_BASE_SIZE 0x68
673 #define TSS_IOPB_SIZE (65536 / 8)
674 #define TSS_REDIRECTION_SIZE (256 / 8)
675 #define RMODE_TSS_SIZE (TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)