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)
73 #define SELECTOR_TI_MASK (1 << 2)
74 #define SELECTOR_RPL_MASK 0x03
78 #define KVM_PIO_PAGE_OFFSET 1
83 * gva - guest virtual address
84 * gpa - guest physical address
85 * gfn - guest frame number
86 * hva - host virtual address
87 * hpa - host physical address
88 * hfn - host frame number
91 typedef unsigned long gva_t
;
93 typedef unsigned long gfn_t
;
95 typedef unsigned long hva_t
;
97 typedef unsigned long hfn_t
;
99 #define NR_PTE_CHAIN_ENTRIES 5
101 struct kvm_pte_chain
{
102 u64
*parent_ptes
[NR_PTE_CHAIN_ENTRIES
];
103 struct hlist_node link
;
107 * kvm_mmu_page_role, below, is defined as:
109 * bits 0:3 - total guest paging levels (2-4, or zero for real mode)
110 * bits 4:7 - page table level for this shadow (1-4)
111 * bits 8:9 - page table quadrant for 2-level guests
112 * bit 16 - "metaphysical" - gfn is not a real page (huge page/real mode)
113 * bits 17:18 - "access" - the user and writable bits of a huge page pde
115 union kvm_mmu_page_role
{
118 unsigned glevels
: 4;
120 unsigned quadrant
: 2;
121 unsigned pad_for_nice_hex_output
: 6;
122 unsigned metaphysical
: 1;
123 unsigned hugepage_access
: 2;
127 struct kvm_mmu_page
{
128 struct list_head link
;
129 struct hlist_node hash_link
;
132 * The following two entries are used to key the shadow page in the
136 union kvm_mmu_page_role role
;
139 unsigned long slot_bitmap
; /* One bit set per slot which has memory
140 * in this shadow page.
142 int multimapped
; /* More than one parent_pte? */
143 int root_count
; /* Currently serving as active root */
145 u64
*parent_pte
; /* !multimapped */
146 struct hlist_head parent_ptes
; /* multimapped, kvm_pte_chain */
156 #define vmx_msr_entry kvm_msr_entry
161 * x86 supports 3 paging modes (4-level 64-bit, 3-level 64-bit, and 2-level
162 * 32-bit). The kvm_mmu structure abstracts the details of the current mmu
166 void (*new_cr3
)(struct kvm_vcpu
*vcpu
);
167 int (*page_fault
)(struct kvm_vcpu
*vcpu
, gva_t gva
, u32 err
);
168 void (*free
)(struct kvm_vcpu
*vcpu
);
169 gpa_t (*gva_to_gpa
)(struct kvm_vcpu
*vcpu
, gva_t gva
);
172 int shadow_root_level
;
177 #define KVM_NR_MEM_OBJS 20
179 struct kvm_mmu_memory_cache
{
181 void *objects
[KVM_NR_MEM_OBJS
];
185 * We don't want allocation failures within the mmu code, so we preallocate
186 * enough memory for a single page fault in a cache.
188 struct kvm_guest_debug
{
227 struct kvm_pio_request
{
230 struct page
*guest_pages
[2];
231 unsigned guest_page_offset
;
243 struct vcpu_svm
*svm
;
250 int interrupt_window_open
;
251 unsigned long irq_summary
; /* bit vector: 1 per word in irq_pending */
252 #define NR_IRQ_WORDS KVM_IRQ_BITMAP_SIZE(unsigned long)
253 unsigned long irq_pending
[NR_IRQ_WORDS
];
254 unsigned long regs
[NR_VCPU_REGS
]; /* for rsp: vcpu_load_rsp_rip() */
255 unsigned long rip
; /* needs vcpu_load_rsp_rip() */
260 gpa_t para_state_gpa
;
261 struct page
*para_state_page
;
265 u64 pdptrs
[4]; /* pae */
268 u64 ia32_misc_enable_msr
;
270 struct vmx_msr_entry
*guest_msrs
;
271 struct vmx_msr_entry
*host_msrs
;
273 struct list_head free_pages
;
274 struct kvm_mmu_page page_header_buf
[KVM_NUM_MMU_PAGES
];
277 struct kvm_mmu_memory_cache mmu_pte_chain_cache
;
278 struct kvm_mmu_memory_cache mmu_rmap_desc_cache
;
280 gfn_t last_pt_write_gfn
;
281 int last_pt_write_count
;
283 struct kvm_guest_debug guest_debug
;
285 char fx_buf
[FX_BUF_SIZE
];
287 char *guest_fx_image
;
290 int mmio_read_completed
;
293 unsigned char mmio_data
[8];
294 gpa_t mmio_phys_addr
;
295 struct kvm_pio_request pio
;
304 struct kvm_save_segment
{
309 } tr
, es
, ds
, fs
, gs
;
313 struct kvm_cpuid_entry cpuid_entries
[KVM_MAX_CPUID_ENTRIES
];
316 struct kvm_mem_alias
{
318 unsigned long npages
;
322 struct kvm_memory_slot
{
324 unsigned long npages
;
326 struct page
**phys_mem
;
327 unsigned long *dirty_bitmap
;
331 spinlock_t lock
; /* protects everything except vcpus */
333 struct kvm_mem_alias aliases
[KVM_ALIAS_SLOTS
];
335 struct kvm_memory_slot memslots
[KVM_MEMORY_SLOTS
];
337 * Hash table of struct kvm_mmu_page.
339 struct list_head active_mmu_pages
;
340 int n_free_mmu_pages
;
341 struct hlist_head mmu_page_hash
[KVM_NUM_MMU_PAGES
];
342 struct kvm_vcpu vcpus
[KVM_MAX_VCPUS
];
343 int memory_config_version
;
345 unsigned long rmap_overflow
;
346 struct list_head vm_list
;
360 u32 irq_window_exits
;
362 u32 request_irq_exits
;
366 struct descriptor_table
{
369 } __attribute__((packed
));
371 struct kvm_arch_ops
{
372 int (*cpu_has_kvm_support
)(void); /* __init */
373 int (*disabled_by_bios
)(void); /* __init */
374 void (*hardware_enable
)(void *dummy
); /* __init */
375 void (*hardware_disable
)(void *dummy
);
376 int (*hardware_setup
)(void); /* __init */
377 void (*hardware_unsetup
)(void); /* __exit */
379 int (*vcpu_create
)(struct kvm_vcpu
*vcpu
);
380 void (*vcpu_free
)(struct kvm_vcpu
*vcpu
);
382 void (*vcpu_load
)(struct kvm_vcpu
*vcpu
);
383 void (*vcpu_put
)(struct kvm_vcpu
*vcpu
);
384 void (*vcpu_decache
)(struct kvm_vcpu
*vcpu
);
386 int (*set_guest_debug
)(struct kvm_vcpu
*vcpu
,
387 struct kvm_debug_guest
*dbg
);
388 int (*get_msr
)(struct kvm_vcpu
*vcpu
, u32 msr_index
, u64
*pdata
);
389 int (*set_msr
)(struct kvm_vcpu
*vcpu
, u32 msr_index
, u64 data
);
390 u64 (*get_segment_base
)(struct kvm_vcpu
*vcpu
, int seg
);
391 void (*get_segment
)(struct kvm_vcpu
*vcpu
,
392 struct kvm_segment
*var
, int seg
);
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_cr0_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 (*invlpg
)(struct kvm_vcpu
*vcpu
, gva_t addr
);
414 void (*tlb_flush
)(struct kvm_vcpu
*vcpu
);
415 void (*inject_page_fault
)(struct kvm_vcpu
*vcpu
,
416 unsigned long addr
, u32 err_code
);
418 void (*inject_gp
)(struct kvm_vcpu
*vcpu
, unsigned err_code
);
420 int (*run
)(struct kvm_vcpu
*vcpu
, struct kvm_run
*run
);
421 int (*vcpu_setup
)(struct kvm_vcpu
*vcpu
);
422 void (*skip_emulated_instruction
)(struct kvm_vcpu
*vcpu
);
423 void (*patch_hypercall
)(struct kvm_vcpu
*vcpu
,
424 unsigned char *hypercall_addr
);
427 extern struct kvm_stat kvm_stat
;
428 extern struct kvm_arch_ops
*kvm_arch_ops
;
430 #define kvm_printf(kvm, fmt ...) printk(KERN_DEBUG fmt)
431 #define vcpu_printf(vcpu, fmt...) kvm_printf(vcpu->kvm, fmt)
433 int kvm_init_arch(struct kvm_arch_ops
*ops
, struct module
*module
);
434 void kvm_exit_arch(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
);
440 int kvm_mmu_reset_context(struct kvm_vcpu
*vcpu
);
441 void kvm_mmu_slot_remove_write_access(struct kvm_vcpu
*vcpu
, int slot
);
442 void kvm_mmu_zap_all(struct kvm_vcpu
*vcpu
);
444 hpa_t
gpa_to_hpa(struct kvm_vcpu
*vcpu
, gpa_t gpa
);
445 #define HPA_MSB ((sizeof(hpa_t) * 8) - 1)
446 #define HPA_ERR_MASK ((hpa_t)1 << HPA_MSB)
447 static inline int is_error_hpa(hpa_t hpa
) { return hpa
>> HPA_MSB
; }
448 hpa_t
gva_to_hpa(struct kvm_vcpu
*vcpu
, gva_t gva
);
449 struct page
*gva_to_page(struct kvm_vcpu
*vcpu
, gva_t gva
);
451 void kvm_emulator_want_group7_invlpg(void);
453 extern hpa_t bad_page_address
;
455 struct page
*gfn_to_page(struct kvm
*kvm
, gfn_t gfn
);
456 struct kvm_memory_slot
*gfn_to_memslot(struct kvm
*kvm
, gfn_t gfn
);
457 void mark_page_dirty(struct kvm
*kvm
, gfn_t gfn
);
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 int emulate_instruction(struct kvm_vcpu
*vcpu
, struct kvm_run
*run
,
466 unsigned long cr2
, u16 error_code
);
467 void realmode_lgdt(struct kvm_vcpu
*vcpu
, u16 size
, unsigned long address
);
468 void realmode_lidt(struct kvm_vcpu
*vcpu
, u16 size
, unsigned long address
);
469 void realmode_lmsw(struct kvm_vcpu
*vcpu
, unsigned long msw
,
470 unsigned long *rflags
);
472 unsigned long realmode_get_cr(struct kvm_vcpu
*vcpu
, int cr
);
473 void realmode_set_cr(struct kvm_vcpu
*vcpu
, int cr
, unsigned long value
,
474 unsigned long *rflags
);
476 struct x86_emulate_ctxt
;
478 int kvm_setup_pio(struct kvm_vcpu
*vcpu
, struct kvm_run
*run
, int in
,
479 int size
, unsigned long count
, int string
, int down
,
480 gva_t address
, int rep
, unsigned port
);
481 void kvm_emulate_cpuid(struct kvm_vcpu
*vcpu
);
482 int emulate_invlpg(struct kvm_vcpu
*vcpu
, gva_t address
);
483 int emulate_clts(struct kvm_vcpu
*vcpu
);
484 int emulator_get_dr(struct x86_emulate_ctxt
* ctxt
, int dr
,
485 unsigned long *dest
);
486 int emulator_set_dr(struct x86_emulate_ctxt
*ctxt
, int dr
,
487 unsigned long value
);
489 void set_cr0(struct kvm_vcpu
*vcpu
, unsigned long cr0
);
490 void set_cr3(struct kvm_vcpu
*vcpu
, unsigned long cr0
);
491 void set_cr4(struct kvm_vcpu
*vcpu
, unsigned long cr0
);
492 void set_cr8(struct kvm_vcpu
*vcpu
, unsigned long cr0
);
493 void lmsw(struct kvm_vcpu
*vcpu
, unsigned long msw
);
495 int kvm_get_msr_common(struct kvm_vcpu
*vcpu
, u32 msr
, u64
*pdata
);
496 int kvm_set_msr_common(struct kvm_vcpu
*vcpu
, u32 msr
, u64 data
);
498 void fx_init(struct kvm_vcpu
*vcpu
);
500 void load_msrs(struct vmx_msr_entry
*e
, int n
);
501 void save_msrs(struct vmx_msr_entry
*e
, int n
);
502 void kvm_resched(struct kvm_vcpu
*vcpu
);
504 int kvm_read_guest(struct kvm_vcpu
*vcpu
,
509 int kvm_write_guest(struct kvm_vcpu
*vcpu
,
514 unsigned long segment_base(u16 selector
);
516 void kvm_mmu_pre_write(struct kvm_vcpu
*vcpu
, gpa_t gpa
, int bytes
);
517 void kvm_mmu_post_write(struct kvm_vcpu
*vcpu
, gpa_t gpa
, int bytes
);
518 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu
*vcpu
, gva_t gva
);
519 void kvm_mmu_free_some_pages(struct kvm_vcpu
*vcpu
);
521 int kvm_hypercall(struct kvm_vcpu
*vcpu
, struct kvm_run
*run
);
523 static inline int kvm_mmu_page_fault(struct kvm_vcpu
*vcpu
, gva_t gva
,
526 if (unlikely(vcpu
->kvm
->n_free_mmu_pages
< KVM_MIN_FREE_MMU_PAGES
))
527 kvm_mmu_free_some_pages(vcpu
);
528 return vcpu
->mmu
.page_fault(vcpu
, gva
, error_code
);
531 static inline int is_long_mode(struct kvm_vcpu
*vcpu
)
534 return vcpu
->shadow_efer
& EFER_LME
;
540 static inline int is_pae(struct kvm_vcpu
*vcpu
)
542 return vcpu
->cr4
& CR4_PAE_MASK
;
545 static inline int is_pse(struct kvm_vcpu
*vcpu
)
547 return vcpu
->cr4
& CR4_PSE_MASK
;
550 static inline int is_paging(struct kvm_vcpu
*vcpu
)
552 return vcpu
->cr0
& CR0_PG_MASK
;
555 static inline int memslot_id(struct kvm
*kvm
, struct kvm_memory_slot
*slot
)
557 return slot
- kvm
->memslots
;
560 static inline struct kvm_mmu_page
*page_header(hpa_t shadow_page
)
562 struct page
*page
= pfn_to_page(shadow_page
>> PAGE_SHIFT
);
564 return (struct kvm_mmu_page
*)page_private(page
);
567 static inline u16
read_fs(void)
570 asm ("mov %%fs, %0" : "=g"(seg
));
574 static inline u16
read_gs(void)
577 asm ("mov %%gs, %0" : "=g"(seg
));
581 static inline u16
read_ldt(void)
584 asm ("sldt %0" : "=g"(ldt
));
588 static inline void load_fs(u16 sel
)
590 asm ("mov %0, %%fs" : : "rm"(sel
));
593 static inline void load_gs(u16 sel
)
595 asm ("mov %0, %%gs" : : "rm"(sel
));
599 static inline void load_ldt(u16 sel
)
601 asm ("lldt %0" : : "rm"(sel
));
605 static inline void get_idt(struct descriptor_table
*table
)
607 asm ("sidt %0" : "=m"(*table
));
610 static inline void get_gdt(struct descriptor_table
*table
)
612 asm ("sgdt %0" : "=m"(*table
));
615 static inline unsigned long read_tr_base(void)
618 asm ("str %0" : "=g"(tr
));
619 return segment_base(tr
);
623 static inline unsigned long read_msr(unsigned long msr
)
632 static inline void fx_save(void *image
)
634 asm ("fxsave (%0)":: "r" (image
));
637 static inline void fx_restore(void *image
)
639 asm ("fxrstor (%0)":: "r" (image
));
642 static inline void fpu_init(void)
647 static inline u32
get_rdx_init_val(void)
649 return 0x600; /* P6 family */
652 #define ASM_VMX_VMCLEAR_RAX ".byte 0x66, 0x0f, 0xc7, 0x30"
653 #define ASM_VMX_VMLAUNCH ".byte 0x0f, 0x01, 0xc2"
654 #define ASM_VMX_VMRESUME ".byte 0x0f, 0x01, 0xc3"
655 #define ASM_VMX_VMPTRLD_RAX ".byte 0x0f, 0xc7, 0x30"
656 #define ASM_VMX_VMREAD_RDX_RAX ".byte 0x0f, 0x78, 0xd0"
657 #define ASM_VMX_VMWRITE_RAX_RDX ".byte 0x0f, 0x79, 0xd0"
658 #define ASM_VMX_VMWRITE_RSP_RDX ".byte 0x0f, 0x79, 0xd4"
659 #define ASM_VMX_VMXOFF ".byte 0x0f, 0x01, 0xc4"
660 #define ASM_VMX_VMXON_RAX ".byte 0xf3, 0x0f, 0xc7, 0x30"
662 #define MSR_IA32_TIME_STAMP_COUNTER 0x010
664 #define TSS_IOPB_BASE_OFFSET 0x66
665 #define TSS_BASE_SIZE 0x68
666 #define TSS_IOPB_SIZE (65536 / 8)
667 #define TSS_REDIRECTION_SIZE (256 / 8)
668 #define RMODE_TSS_SIZE (TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)