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_MEMORY_SLOTS 4
55 #define KVM_NUM_MMU_PAGES 256
56 #define KVM_MIN_FREE_MMU_PAGES 5
57 #define KVM_REFILL_PAGES 25
59 #define FX_IMAGE_SIZE 512
60 #define FX_IMAGE_ALIGN 16
61 #define FX_BUF_SIZE (2 * FX_IMAGE_SIZE + FX_IMAGE_ALIGN)
71 #define SELECTOR_TI_MASK (1 << 2)
72 #define SELECTOR_RPL_MASK 0x03
79 * gva - guest virtual address
80 * gpa - guest physical address
81 * gfn - guest frame number
82 * hva - host virtual address
83 * hpa - host physical address
84 * hfn - host frame number
87 typedef unsigned long gva_t
;
89 typedef unsigned long gfn_t
;
91 typedef unsigned long hva_t
;
93 typedef unsigned long hfn_t
;
95 #define NR_PTE_CHAIN_ENTRIES 5
97 struct kvm_pte_chain
{
98 u64
*parent_ptes
[NR_PTE_CHAIN_ENTRIES
];
99 struct hlist_node link
;
103 * kvm_mmu_page_role, below, is defined as:
105 * bits 0:3 - total guest paging levels (2-4, or zero for real mode)
106 * bits 4:7 - page table level for this shadow (1-4)
107 * bits 8:9 - page table quadrant for 2-level guests
108 * bit 16 - "metaphysical" - gfn is not a real page (huge page/real mode)
110 union kvm_mmu_page_role
{
113 unsigned glevels
: 4;
115 unsigned quadrant
: 2;
116 unsigned pad_for_nice_hex_output
: 6;
117 unsigned metaphysical
: 1;
121 struct kvm_mmu_page
{
122 struct list_head link
;
123 struct hlist_node hash_link
;
126 * The following two entries are used to key the shadow page in the
130 union kvm_mmu_page_role role
;
133 unsigned long slot_bitmap
; /* One bit set per slot which has memory
134 * in this shadow page.
136 int global
; /* Set if all ptes in this page are global */
137 int multimapped
; /* More than one parent_pte? */
138 int root_count
; /* Currently serving as active root */
140 u64
*parent_pte
; /* !multimapped */
141 struct hlist_head parent_ptes
; /* multimapped, kvm_pte_chain */
151 #define vmx_msr_entry kvm_msr_entry
156 * x86 supports 3 paging modes (4-level 64-bit, 3-level 64-bit, and 2-level
157 * 32-bit). The kvm_mmu structure abstracts the details of the current mmu
161 void (*new_cr3
)(struct kvm_vcpu
*vcpu
);
162 int (*page_fault
)(struct kvm_vcpu
*vcpu
, gva_t gva
, u32 err
);
163 void (*free
)(struct kvm_vcpu
*vcpu
);
164 gpa_t (*gva_to_gpa
)(struct kvm_vcpu
*vcpu
, gva_t gva
);
167 int shadow_root_level
;
172 #define KVM_NR_MEM_OBJS 20
174 struct kvm_mmu_memory_cache
{
176 void *objects
[KVM_NR_MEM_OBJS
];
180 * We don't want allocation failures within the mmu code, so we preallocate
181 * enough memory for a single page fault in a cache.
183 struct kvm_guest_debug
{
226 struct vcpu_svm
*svm
;
231 int interrupt_window_open
;
232 unsigned long irq_summary
; /* bit vector: 1 per word in irq_pending */
233 #define NR_IRQ_WORDS KVM_IRQ_BITMAP_SIZE(unsigned long)
234 unsigned long irq_pending
[NR_IRQ_WORDS
];
235 unsigned long regs
[NR_VCPU_REGS
]; /* for rsp: vcpu_load_rsp_rip() */
236 unsigned long rip
; /* needs vcpu_load_rsp_rip() */
241 gpa_t para_state_gpa
;
242 struct page
*para_state_page
;
246 u64 pdptrs
[4]; /* pae */
249 u64 ia32_misc_enable_msr
;
251 struct vmx_msr_entry
*guest_msrs
;
252 struct vmx_msr_entry
*host_msrs
;
254 struct list_head free_pages
;
255 struct kvm_mmu_page page_header_buf
[KVM_NUM_MMU_PAGES
];
258 struct kvm_mmu_memory_cache mmu_pte_chain_cache
;
259 struct kvm_mmu_memory_cache mmu_rmap_desc_cache
;
261 gfn_t last_pt_write_gfn
;
262 int last_pt_write_count
;
264 struct kvm_guest_debug guest_debug
;
266 char fx_buf
[FX_BUF_SIZE
];
268 char *guest_fx_image
;
271 int mmio_read_completed
;
274 unsigned char mmio_data
[8];
275 gpa_t mmio_phys_addr
;
280 struct kvm_save_segment
{
285 } tr
, es
, ds
, fs
, gs
;
289 struct kvm_memory_slot
{
291 unsigned long npages
;
293 struct page
**phys_mem
;
294 unsigned long *dirty_bitmap
;
298 spinlock_t lock
; /* protects everything except vcpus */
300 struct kvm_memory_slot memslots
[KVM_MEMORY_SLOTS
];
302 * Hash table of struct kvm_mmu_page.
304 struct list_head active_mmu_pages
;
305 int n_free_mmu_pages
;
306 struct hlist_head mmu_page_hash
[KVM_NUM_MMU_PAGES
];
307 struct kvm_vcpu vcpus
[KVM_MAX_VCPUS
];
308 int memory_config_version
;
310 unsigned long rmap_overflow
;
311 struct list_head vm_list
;
324 u32 irq_window_exits
;
326 u32 request_irq_exits
;
330 struct descriptor_table
{
333 } __attribute__((packed
));
335 struct kvm_arch_ops
{
336 int (*cpu_has_kvm_support
)(void); /* __init */
337 int (*disabled_by_bios
)(void); /* __init */
338 void (*hardware_enable
)(void *dummy
); /* __init */
339 void (*hardware_disable
)(void *dummy
);
340 int (*hardware_setup
)(void); /* __init */
341 void (*hardware_unsetup
)(void); /* __exit */
343 int (*vcpu_create
)(struct kvm_vcpu
*vcpu
);
344 void (*vcpu_free
)(struct kvm_vcpu
*vcpu
);
346 struct kvm_vcpu
*(*vcpu_load
)(struct kvm_vcpu
*vcpu
);
347 void (*vcpu_put
)(struct kvm_vcpu
*vcpu
);
348 void (*vcpu_decache
)(struct kvm_vcpu
*vcpu
);
350 int (*set_guest_debug
)(struct kvm_vcpu
*vcpu
,
351 struct kvm_debug_guest
*dbg
);
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_cr0_cr4_guest_bits
)(struct kvm_vcpu
*vcpu
);
361 void (*set_cr0
)(struct kvm_vcpu
*vcpu
, unsigned long cr0
);
362 void (*set_cr0_no_modeswitch
)(struct kvm_vcpu
*vcpu
,
364 void (*set_cr3
)(struct kvm_vcpu
*vcpu
, unsigned long cr3
);
365 void (*set_cr4
)(struct kvm_vcpu
*vcpu
, unsigned long cr4
);
366 void (*set_efer
)(struct kvm_vcpu
*vcpu
, u64 efer
);
367 void (*get_idt
)(struct kvm_vcpu
*vcpu
, struct descriptor_table
*dt
);
368 void (*set_idt
)(struct kvm_vcpu
*vcpu
, struct descriptor_table
*dt
);
369 void (*get_gdt
)(struct kvm_vcpu
*vcpu
, struct descriptor_table
*dt
);
370 void (*set_gdt
)(struct kvm_vcpu
*vcpu
, struct descriptor_table
*dt
);
371 unsigned long (*get_dr
)(struct kvm_vcpu
*vcpu
, int dr
);
372 void (*set_dr
)(struct kvm_vcpu
*vcpu
, int dr
, unsigned long value
,
374 void (*cache_regs
)(struct kvm_vcpu
*vcpu
);
375 void (*decache_regs
)(struct kvm_vcpu
*vcpu
);
376 unsigned long (*get_rflags
)(struct kvm_vcpu
*vcpu
);
377 void (*set_rflags
)(struct kvm_vcpu
*vcpu
, unsigned long rflags
);
379 void (*invlpg
)(struct kvm_vcpu
*vcpu
, gva_t addr
);
380 void (*tlb_flush
)(struct kvm_vcpu
*vcpu
);
381 void (*inject_page_fault
)(struct kvm_vcpu
*vcpu
,
382 unsigned long addr
, u32 err_code
);
384 void (*inject_gp
)(struct kvm_vcpu
*vcpu
, unsigned err_code
);
386 int (*run
)(struct kvm_vcpu
*vcpu
, struct kvm_run
*run
);
387 int (*vcpu_setup
)(struct kvm_vcpu
*vcpu
);
388 void (*skip_emulated_instruction
)(struct kvm_vcpu
*vcpu
);
389 void (*patch_hypercall
)(struct kvm_vcpu
*vcpu
,
390 unsigned char *hypercall_addr
);
393 extern struct kvm_stat kvm_stat
;
394 extern struct kvm_arch_ops
*kvm_arch_ops
;
396 #define kvm_printf(kvm, fmt ...) printk(KERN_DEBUG fmt)
397 #define vcpu_printf(vcpu, fmt...) kvm_printf(vcpu->kvm, fmt)
399 int kvm_init_arch(struct kvm_arch_ops
*ops
, struct module
*module
);
400 void kvm_exit_arch(void);
402 void kvm_mmu_destroy(struct kvm_vcpu
*vcpu
);
403 int kvm_mmu_create(struct kvm_vcpu
*vcpu
);
404 int kvm_mmu_setup(struct kvm_vcpu
*vcpu
);
406 int kvm_mmu_reset_context(struct kvm_vcpu
*vcpu
);
407 void kvm_mmu_slot_remove_write_access(struct kvm_vcpu
*vcpu
, int slot
);
409 hpa_t
gpa_to_hpa(struct kvm_vcpu
*vcpu
, gpa_t gpa
);
410 #define HPA_MSB ((sizeof(hpa_t) * 8) - 1)
411 #define HPA_ERR_MASK ((hpa_t)1 << HPA_MSB)
412 static inline int is_error_hpa(hpa_t hpa
) { return hpa
>> HPA_MSB
; }
413 hpa_t
gva_to_hpa(struct kvm_vcpu
*vcpu
, gva_t gva
);
415 void kvm_emulator_want_group7_invlpg(void);
417 extern hpa_t bad_page_address
;
419 static inline struct page
*gfn_to_page(struct kvm_memory_slot
*slot
, gfn_t gfn
)
421 return slot
->phys_mem
[gfn
- slot
->base_gfn
];
424 struct kvm_memory_slot
*gfn_to_memslot(struct kvm
*kvm
, gfn_t gfn
);
425 void mark_page_dirty(struct kvm
*kvm
, gfn_t gfn
);
427 enum emulation_result
{
428 EMULATE_DONE
, /* no further processing */
429 EMULATE_DO_MMIO
, /* kvm_run filled with mmio request */
430 EMULATE_FAIL
, /* can't emulate this instruction */
433 int emulate_instruction(struct kvm_vcpu
*vcpu
, struct kvm_run
*run
,
434 unsigned long cr2
, u16 error_code
);
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
);
444 struct x86_emulate_ctxt
;
446 int emulate_invlpg(struct kvm_vcpu
*vcpu
, gva_t address
);
447 int emulate_clts(struct kvm_vcpu
*vcpu
);
448 int emulator_get_dr(struct x86_emulate_ctxt
* ctxt
, int dr
,
449 unsigned long *dest
);
450 int emulator_set_dr(struct x86_emulate_ctxt
*ctxt
, int dr
,
451 unsigned long value
);
453 void set_cr0(struct kvm_vcpu
*vcpu
, unsigned long cr0
);
454 void set_cr3(struct kvm_vcpu
*vcpu
, unsigned long cr0
);
455 void set_cr4(struct kvm_vcpu
*vcpu
, unsigned long cr0
);
456 void set_cr8(struct kvm_vcpu
*vcpu
, unsigned long cr0
);
457 void lmsw(struct kvm_vcpu
*vcpu
, unsigned long msw
);
459 int kvm_get_msr_common(struct kvm_vcpu
*vcpu
, u32 msr
, u64
*pdata
);
460 int kvm_set_msr_common(struct kvm_vcpu
*vcpu
, u32 msr
, u64 data
);
462 void fx_init(struct kvm_vcpu
*vcpu
);
464 void load_msrs(struct vmx_msr_entry
*e
, int n
);
465 void save_msrs(struct vmx_msr_entry
*e
, int n
);
466 void kvm_resched(struct kvm_vcpu
*vcpu
);
468 int kvm_read_guest(struct kvm_vcpu
*vcpu
,
473 int kvm_write_guest(struct kvm_vcpu
*vcpu
,
478 unsigned long segment_base(u16 selector
);
480 void kvm_mmu_pre_write(struct kvm_vcpu
*vcpu
, gpa_t gpa
, int bytes
);
481 void kvm_mmu_post_write(struct kvm_vcpu
*vcpu
, gpa_t gpa
, 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
);
485 int kvm_hypercall(struct kvm_vcpu
*vcpu
, struct kvm_run
*run
);
487 static inline int kvm_mmu_page_fault(struct kvm_vcpu
*vcpu
, gva_t gva
,
490 if (unlikely(vcpu
->kvm
->n_free_mmu_pages
< KVM_MIN_FREE_MMU_PAGES
))
491 kvm_mmu_free_some_pages(vcpu
);
492 return vcpu
->mmu
.page_fault(vcpu
, gva
, error_code
);
495 static inline struct page
*_gfn_to_page(struct kvm
*kvm
, gfn_t gfn
)
497 struct kvm_memory_slot
*slot
= gfn_to_memslot(kvm
, gfn
);
498 return (slot
) ? slot
->phys_mem
[gfn
- slot
->base_gfn
] : NULL
;
501 static inline int is_long_mode(struct kvm_vcpu
*vcpu
)
504 return vcpu
->shadow_efer
& EFER_LME
;
510 static inline int is_pae(struct kvm_vcpu
*vcpu
)
512 return vcpu
->cr4
& CR4_PAE_MASK
;
515 static inline int is_pse(struct kvm_vcpu
*vcpu
)
517 return vcpu
->cr4
& CR4_PSE_MASK
;
520 static inline int is_paging(struct kvm_vcpu
*vcpu
)
522 return vcpu
->cr0
& CR0_PG_MASK
;
525 static inline int memslot_id(struct kvm
*kvm
, struct kvm_memory_slot
*slot
)
527 return slot
- kvm
->memslots
;
530 static inline struct kvm_mmu_page
*page_header(hpa_t shadow_page
)
532 struct page
*page
= pfn_to_page(shadow_page
>> PAGE_SHIFT
);
534 return (struct kvm_mmu_page
*)page
->private;
537 static inline u16
read_fs(void)
540 asm ("mov %%fs, %0" : "=g"(seg
));
544 static inline u16
read_gs(void)
547 asm ("mov %%gs, %0" : "=g"(seg
));
551 static inline u16
read_ldt(void)
554 asm ("sldt %0" : "=g"(ldt
));
558 static inline void load_fs(u16 sel
)
560 asm ("mov %0, %%fs" : : "rm"(sel
));
563 static inline void load_gs(u16 sel
)
565 asm ("mov %0, %%gs" : : "rm"(sel
));
569 static inline void load_ldt(u16 sel
)
571 asm ("lldt %0" : : "rm"(sel
));
575 static inline void get_idt(struct descriptor_table
*table
)
577 asm ("sidt %0" : "=m"(*table
));
580 static inline void get_gdt(struct descriptor_table
*table
)
582 asm ("sgdt %0" : "=m"(*table
));
585 static inline unsigned long read_tr_base(void)
588 asm ("str %0" : "=g"(tr
));
589 return segment_base(tr
);
593 static inline unsigned long read_msr(unsigned long msr
)
602 static inline void fx_save(void *image
)
604 asm ("fxsave (%0)":: "r" (image
));
607 static inline void fx_restore(void *image
)
609 asm ("fxrstor (%0)":: "r" (image
));
612 static inline void fpu_init(void)
617 static inline u32
get_rdx_init_val(void)
619 return 0x600; /* P6 family */
622 #define ASM_VMX_VMCLEAR_RAX ".byte 0x66, 0x0f, 0xc7, 0x30"
623 #define ASM_VMX_VMLAUNCH ".byte 0x0f, 0x01, 0xc2"
624 #define ASM_VMX_VMRESUME ".byte 0x0f, 0x01, 0xc3"
625 #define ASM_VMX_VMPTRLD_RAX ".byte 0x0f, 0xc7, 0x30"
626 #define ASM_VMX_VMREAD_RDX_RAX ".byte 0x0f, 0x78, 0xd0"
627 #define ASM_VMX_VMWRITE_RAX_RDX ".byte 0x0f, 0x79, 0xd0"
628 #define ASM_VMX_VMWRITE_RSP_RDX ".byte 0x0f, 0x79, 0xd4"
629 #define ASM_VMX_VMXOFF ".byte 0x0f, 0x01, 0xc4"
630 #define ASM_VMX_VMXON_RAX ".byte 0xf3, 0x0f, 0xc7, 0x30"
632 #define MSR_IA32_TIME_STAMP_COUNTER 0x010
634 #define TSS_IOPB_BASE_OFFSET 0x66
635 #define TSS_BASE_SIZE 0x68
636 #define TSS_IOPB_SIZE (65536 / 8)
637 #define TSS_REDIRECTION_SIZE (256 / 8)
638 #define RMODE_TSS_SIZE (TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)