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>
14 #include <asm/signal.h>
17 #include <linux/kvm.h>
18 #include <linux/kvm_para.h>
20 #define CR0_PE_MASK (1ULL << 0)
21 #define CR0_MP_MASK (1ULL << 1)
22 #define CR0_TS_MASK (1ULL << 3)
23 #define CR0_NE_MASK (1ULL << 5)
24 #define CR0_WP_MASK (1ULL << 16)
25 #define CR0_NW_MASK (1ULL << 29)
26 #define CR0_CD_MASK (1ULL << 30)
27 #define CR0_PG_MASK (1ULL << 31)
29 #define CR3_WPT_MASK (1ULL << 3)
30 #define CR3_PCD_MASK (1ULL << 4)
32 #define CR3_RESEVED_BITS 0x07ULL
33 #define CR3_L_MODE_RESEVED_BITS (~((1ULL << 40) - 1) | 0x0fe7ULL)
34 #define CR3_FLAGS_MASK ((1ULL << 5) - 1)
36 #define CR4_VME_MASK (1ULL << 0)
37 #define CR4_PSE_MASK (1ULL << 4)
38 #define CR4_PAE_MASK (1ULL << 5)
39 #define CR4_PGE_MASK (1ULL << 7)
40 #define CR4_VMXE_MASK (1ULL << 13)
42 #define KVM_GUEST_CR0_MASK \
43 (CR0_PG_MASK | CR0_PE_MASK | CR0_WP_MASK | CR0_NE_MASK \
44 | CR0_NW_MASK | CR0_CD_MASK)
45 #define KVM_VM_CR0_ALWAYS_ON \
46 (CR0_PG_MASK | CR0_PE_MASK | CR0_WP_MASK | CR0_NE_MASK | CR0_TS_MASK \
48 #define KVM_GUEST_CR4_MASK \
49 (CR4_PSE_MASK | CR4_PAE_MASK | CR4_PGE_MASK | CR4_VMXE_MASK | CR4_VME_MASK)
50 #define KVM_PMODE_VM_CR4_ALWAYS_ON (CR4_VMXE_MASK | CR4_PAE_MASK)
51 #define KVM_RMODE_VM_CR4_ALWAYS_ON (CR4_VMXE_MASK | CR4_PAE_MASK | CR4_VME_MASK)
53 #define INVALID_PAGE (~(hpa_t)0)
54 #define UNMAPPED_GVA (~(gpa_t)0)
56 #define KVM_MAX_VCPUS 1
57 #define KVM_ALIAS_SLOTS 4
58 #define KVM_MEMORY_SLOTS 4
59 #define KVM_NUM_MMU_PAGES 1024
60 #define KVM_MIN_FREE_MMU_PAGES 5
61 #define KVM_REFILL_PAGES 25
62 #define KVM_MAX_CPUID_ENTRIES 40
64 #define FX_IMAGE_SIZE 512
65 #define FX_IMAGE_ALIGN 16
66 #define FX_BUF_SIZE (2 * FX_IMAGE_SIZE + FX_IMAGE_ALIGN)
77 #define SELECTOR_TI_MASK (1 << 2)
78 #define SELECTOR_RPL_MASK 0x03
82 #define KVM_PIO_PAGE_OFFSET 1
87 * gva - guest virtual address
88 * gpa - guest physical address
89 * gfn - guest frame number
90 * hva - host virtual address
91 * hpa - host physical address
92 * hfn - host frame number
95 typedef unsigned long gva_t
;
97 typedef unsigned long gfn_t
;
99 typedef unsigned long hva_t
;
101 typedef unsigned long hfn_t
;
103 #define NR_PTE_CHAIN_ENTRIES 5
105 struct kvm_pte_chain
{
106 u64
*parent_ptes
[NR_PTE_CHAIN_ENTRIES
];
107 struct hlist_node link
;
111 * kvm_mmu_page_role, below, is defined as:
113 * bits 0:3 - total guest paging levels (2-4, or zero for real mode)
114 * bits 4:7 - page table level for this shadow (1-4)
115 * bits 8:9 - page table quadrant for 2-level guests
116 * bit 16 - "metaphysical" - gfn is not a real page (huge page/real mode)
117 * bits 17:18 - "access" - the user and writable bits of a huge page pde
119 union kvm_mmu_page_role
{
122 unsigned glevels
: 4;
124 unsigned quadrant
: 2;
125 unsigned pad_for_nice_hex_output
: 6;
126 unsigned metaphysical
: 1;
127 unsigned hugepage_access
: 2;
131 struct kvm_mmu_page
{
132 struct list_head link
;
133 struct hlist_node hash_link
;
136 * The following two entries are used to key the shadow page in the
140 union kvm_mmu_page_role role
;
143 unsigned long slot_bitmap
; /* One bit set per slot which has memory
144 * in this shadow page.
146 int multimapped
; /* More than one parent_pte? */
147 int root_count
; /* Currently serving as active root */
149 u64
*parent_pte
; /* !multimapped */
150 struct hlist_head parent_ptes
; /* multimapped, kvm_pte_chain */
160 #define vmx_msr_entry kvm_msr_entry
165 * x86 supports 3 paging modes (4-level 64-bit, 3-level 64-bit, and 2-level
166 * 32-bit). The kvm_mmu structure abstracts the details of the current mmu
170 void (*new_cr3
)(struct kvm_vcpu
*vcpu
);
171 int (*page_fault
)(struct kvm_vcpu
*vcpu
, gva_t gva
, u32 err
);
172 void (*free
)(struct kvm_vcpu
*vcpu
);
173 gpa_t (*gva_to_gpa
)(struct kvm_vcpu
*vcpu
, gva_t gva
);
176 int shadow_root_level
;
181 #define KVM_NR_MEM_OBJS 20
183 struct kvm_mmu_memory_cache
{
185 void *objects
[KVM_NR_MEM_OBJS
];
189 * We don't want allocation failures within the mmu code, so we preallocate
190 * enough memory for a single page fault in a cache.
192 struct kvm_guest_debug
{
231 struct kvm_pio_request
{
234 struct page
*guest_pages
[2];
235 unsigned guest_page_offset
;
253 u32 irq_window_exits
;
255 u32 request_irq_exits
;
264 struct vcpu_svm
*svm
;
271 int interrupt_window_open
;
272 unsigned long irq_summary
; /* bit vector: 1 per word in irq_pending */
273 #define NR_IRQ_WORDS KVM_IRQ_BITMAP_SIZE(unsigned long)
274 unsigned long irq_pending
[NR_IRQ_WORDS
];
275 unsigned long regs
[NR_VCPU_REGS
]; /* for rsp: vcpu_load_rsp_rip() */
276 unsigned long rip
; /* needs vcpu_load_rsp_rip() */
281 gpa_t para_state_gpa
;
282 struct page
*para_state_page
;
286 u64 pdptrs
[4]; /* pae */
289 u64 ia32_misc_enable_msr
;
293 int msr_offset_kernel_gs_base
;
295 struct vmx_msr_entry
*guest_msrs
;
296 struct vmx_msr_entry
*host_msrs
;
298 struct list_head free_pages
;
299 struct kvm_mmu_page page_header_buf
[KVM_NUM_MMU_PAGES
];
302 struct kvm_mmu_memory_cache mmu_pte_chain_cache
;
303 struct kvm_mmu_memory_cache mmu_rmap_desc_cache
;
305 gfn_t last_pt_write_gfn
;
306 int last_pt_write_count
;
308 struct kvm_guest_debug guest_debug
;
310 char fx_buf
[FX_BUF_SIZE
];
312 char *guest_fx_image
;
314 int guest_fpu_loaded
;
315 struct vmx_host_state
{
317 u16 fs_sel
, gs_sel
, ldt_sel
;
318 int fs_gs_ldt_reload_needed
;
322 int mmio_read_completed
;
325 unsigned char mmio_data
[8];
326 gpa_t mmio_phys_addr
;
327 gva_t mmio_fault_cr2
;
328 struct kvm_pio_request pio
;
334 struct kvm_stat stat
;
339 struct kvm_save_segment
{
344 } tr
, es
, ds
, fs
, gs
;
348 struct kvm_cpuid_entry cpuid_entries
[KVM_MAX_CPUID_ENTRIES
];
351 struct kvm_mem_alias
{
353 unsigned long npages
;
357 struct kvm_memory_slot
{
359 unsigned long npages
;
361 struct page
**phys_mem
;
362 unsigned long *dirty_bitmap
;
366 spinlock_t lock
; /* protects everything except vcpus */
368 struct kvm_mem_alias aliases
[KVM_ALIAS_SLOTS
];
370 struct kvm_memory_slot memslots
[KVM_MEMORY_SLOTS
];
372 * Hash table of struct kvm_mmu_page.
374 struct list_head active_mmu_pages
;
375 int n_free_mmu_pages
;
376 struct hlist_head mmu_page_hash
[KVM_NUM_MMU_PAGES
];
377 struct kvm_vcpu vcpus
[KVM_MAX_VCPUS
];
378 int memory_config_version
;
380 unsigned long rmap_overflow
;
381 struct list_head vm_list
;
385 struct descriptor_table
{
388 } __attribute__((packed
));
390 struct kvm_arch_ops
{
391 int (*cpu_has_kvm_support
)(void); /* __init */
392 int (*disabled_by_bios
)(void); /* __init */
393 void (*hardware_enable
)(void *dummy
); /* __init */
394 void (*hardware_disable
)(void *dummy
);
395 int (*hardware_setup
)(void); /* __init */
396 void (*hardware_unsetup
)(void); /* __exit */
398 int (*vcpu_create
)(struct kvm_vcpu
*vcpu
);
399 void (*vcpu_free
)(struct kvm_vcpu
*vcpu
);
401 void (*vcpu_load
)(struct kvm_vcpu
*vcpu
);
402 void (*vcpu_put
)(struct kvm_vcpu
*vcpu
);
403 void (*vcpu_decache
)(struct kvm_vcpu
*vcpu
);
405 int (*set_guest_debug
)(struct kvm_vcpu
*vcpu
,
406 struct kvm_debug_guest
*dbg
);
407 int (*get_msr
)(struct kvm_vcpu
*vcpu
, u32 msr_index
, u64
*pdata
);
408 int (*set_msr
)(struct kvm_vcpu
*vcpu
, u32 msr_index
, u64 data
);
409 u64 (*get_segment_base
)(struct kvm_vcpu
*vcpu
, int seg
);
410 void (*get_segment
)(struct kvm_vcpu
*vcpu
,
411 struct kvm_segment
*var
, int seg
);
412 void (*set_segment
)(struct kvm_vcpu
*vcpu
,
413 struct kvm_segment
*var
, int seg
);
414 void (*get_cs_db_l_bits
)(struct kvm_vcpu
*vcpu
, int *db
, int *l
);
415 void (*decache_cr4_guest_bits
)(struct kvm_vcpu
*vcpu
);
416 void (*set_cr0
)(struct kvm_vcpu
*vcpu
, unsigned long cr0
);
417 void (*set_cr3
)(struct kvm_vcpu
*vcpu
, unsigned long cr3
);
418 void (*set_cr4
)(struct kvm_vcpu
*vcpu
, unsigned long cr4
);
419 void (*set_efer
)(struct kvm_vcpu
*vcpu
, u64 efer
);
420 void (*get_idt
)(struct kvm_vcpu
*vcpu
, struct descriptor_table
*dt
);
421 void (*set_idt
)(struct kvm_vcpu
*vcpu
, struct descriptor_table
*dt
);
422 void (*get_gdt
)(struct kvm_vcpu
*vcpu
, struct descriptor_table
*dt
);
423 void (*set_gdt
)(struct kvm_vcpu
*vcpu
, struct descriptor_table
*dt
);
424 unsigned long (*get_dr
)(struct kvm_vcpu
*vcpu
, int dr
);
425 void (*set_dr
)(struct kvm_vcpu
*vcpu
, int dr
, unsigned long value
,
427 void (*cache_regs
)(struct kvm_vcpu
*vcpu
);
428 void (*decache_regs
)(struct kvm_vcpu
*vcpu
);
429 unsigned long (*get_rflags
)(struct kvm_vcpu
*vcpu
);
430 void (*set_rflags
)(struct kvm_vcpu
*vcpu
, unsigned long rflags
);
432 void (*invlpg
)(struct kvm_vcpu
*vcpu
, gva_t addr
);
433 void (*tlb_flush
)(struct kvm_vcpu
*vcpu
);
434 void (*inject_page_fault
)(struct kvm_vcpu
*vcpu
,
435 unsigned long addr
, u32 err_code
);
437 void (*inject_gp
)(struct kvm_vcpu
*vcpu
, unsigned err_code
);
439 int (*run
)(struct kvm_vcpu
*vcpu
, struct kvm_run
*run
);
440 int (*vcpu_setup
)(struct kvm_vcpu
*vcpu
);
441 void (*skip_emulated_instruction
)(struct kvm_vcpu
*vcpu
);
442 void (*patch_hypercall
)(struct kvm_vcpu
*vcpu
,
443 unsigned char *hypercall_addr
);
446 extern struct kvm_arch_ops
*kvm_arch_ops
;
448 #define kvm_printf(kvm, fmt ...) printk(KERN_DEBUG fmt)
449 #define vcpu_printf(vcpu, fmt...) kvm_printf(vcpu->kvm, fmt)
451 int kvm_init_arch(struct kvm_arch_ops
*ops
, struct module
*module
);
452 void kvm_exit_arch(void);
454 int kvm_mmu_module_init(void);
455 void kvm_mmu_module_exit(void);
457 void kvm_mmu_destroy(struct kvm_vcpu
*vcpu
);
458 int kvm_mmu_create(struct kvm_vcpu
*vcpu
);
459 int kvm_mmu_setup(struct kvm_vcpu
*vcpu
);
461 int kvm_mmu_reset_context(struct kvm_vcpu
*vcpu
);
462 void kvm_mmu_slot_remove_write_access(struct kvm_vcpu
*vcpu
, int slot
);
463 void kvm_mmu_zap_all(struct kvm_vcpu
*vcpu
);
465 hpa_t
gpa_to_hpa(struct kvm_vcpu
*vcpu
, gpa_t gpa
);
466 #define HPA_MSB ((sizeof(hpa_t) * 8) - 1)
467 #define HPA_ERR_MASK ((hpa_t)1 << HPA_MSB)
468 static inline int is_error_hpa(hpa_t hpa
) { return hpa
>> HPA_MSB
; }
469 hpa_t
gva_to_hpa(struct kvm_vcpu
*vcpu
, gva_t gva
);
470 struct page
*gva_to_page(struct kvm_vcpu
*vcpu
, gva_t gva
);
472 void kvm_emulator_want_group7_invlpg(void);
474 extern hpa_t bad_page_address
;
476 struct page
*gfn_to_page(struct kvm
*kvm
, gfn_t gfn
);
477 struct kvm_memory_slot
*gfn_to_memslot(struct kvm
*kvm
, gfn_t gfn
);
478 void mark_page_dirty(struct kvm
*kvm
, gfn_t gfn
);
480 enum emulation_result
{
481 EMULATE_DONE
, /* no further processing */
482 EMULATE_DO_MMIO
, /* kvm_run filled with mmio request */
483 EMULATE_FAIL
, /* can't emulate this instruction */
486 int emulate_instruction(struct kvm_vcpu
*vcpu
, struct kvm_run
*run
,
487 unsigned long cr2
, u16 error_code
);
488 void realmode_lgdt(struct kvm_vcpu
*vcpu
, u16 size
, unsigned long address
);
489 void realmode_lidt(struct kvm_vcpu
*vcpu
, u16 size
, unsigned long address
);
490 void realmode_lmsw(struct kvm_vcpu
*vcpu
, unsigned long msw
,
491 unsigned long *rflags
);
493 unsigned long realmode_get_cr(struct kvm_vcpu
*vcpu
, int cr
);
494 void realmode_set_cr(struct kvm_vcpu
*vcpu
, int cr
, unsigned long value
,
495 unsigned long *rflags
);
497 struct x86_emulate_ctxt
;
499 int kvm_setup_pio(struct kvm_vcpu
*vcpu
, struct kvm_run
*run
, int in
,
500 int size
, unsigned long count
, int string
, int down
,
501 gva_t address
, int rep
, unsigned port
);
502 void kvm_emulate_cpuid(struct kvm_vcpu
*vcpu
);
503 int emulate_invlpg(struct kvm_vcpu
*vcpu
, gva_t address
);
504 int emulate_clts(struct kvm_vcpu
*vcpu
);
505 int emulator_get_dr(struct x86_emulate_ctxt
* ctxt
, int dr
,
506 unsigned long *dest
);
507 int emulator_set_dr(struct x86_emulate_ctxt
*ctxt
, int dr
,
508 unsigned long value
);
510 void set_cr0(struct kvm_vcpu
*vcpu
, unsigned long cr0
);
511 void set_cr3(struct kvm_vcpu
*vcpu
, unsigned long cr0
);
512 void set_cr4(struct kvm_vcpu
*vcpu
, unsigned long cr0
);
513 void set_cr8(struct kvm_vcpu
*vcpu
, unsigned long cr0
);
514 void lmsw(struct kvm_vcpu
*vcpu
, unsigned long msw
);
516 int kvm_get_msr_common(struct kvm_vcpu
*vcpu
, u32 msr
, u64
*pdata
);
517 int kvm_set_msr_common(struct kvm_vcpu
*vcpu
, u32 msr
, u64 data
);
519 void fx_init(struct kvm_vcpu
*vcpu
);
521 void load_msrs(struct vmx_msr_entry
*e
, int n
);
522 void save_msrs(struct vmx_msr_entry
*e
, int n
);
523 void kvm_resched(struct kvm_vcpu
*vcpu
);
524 void kvm_load_guest_fpu(struct kvm_vcpu
*vcpu
);
525 void kvm_put_guest_fpu(struct kvm_vcpu
*vcpu
);
527 int kvm_read_guest(struct kvm_vcpu
*vcpu
,
532 int kvm_write_guest(struct kvm_vcpu
*vcpu
,
537 unsigned long segment_base(u16 selector
);
539 void kvm_mmu_pte_write(struct kvm_vcpu
*vcpu
, gpa_t gpa
,
540 const u8
*old
, const u8
*new, int bytes
);
541 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu
*vcpu
, gva_t gva
);
542 void kvm_mmu_free_some_pages(struct kvm_vcpu
*vcpu
);
544 int kvm_hypercall(struct kvm_vcpu
*vcpu
, struct kvm_run
*run
);
546 static inline int kvm_mmu_page_fault(struct kvm_vcpu
*vcpu
, gva_t gva
,
549 if (unlikely(vcpu
->kvm
->n_free_mmu_pages
< KVM_MIN_FREE_MMU_PAGES
))
550 kvm_mmu_free_some_pages(vcpu
);
551 return vcpu
->mmu
.page_fault(vcpu
, gva
, error_code
);
554 static inline int is_long_mode(struct kvm_vcpu
*vcpu
)
557 return vcpu
->shadow_efer
& EFER_LME
;
563 static inline int is_pae(struct kvm_vcpu
*vcpu
)
565 return vcpu
->cr4
& CR4_PAE_MASK
;
568 static inline int is_pse(struct kvm_vcpu
*vcpu
)
570 return vcpu
->cr4
& CR4_PSE_MASK
;
573 static inline int is_paging(struct kvm_vcpu
*vcpu
)
575 return vcpu
->cr0
& CR0_PG_MASK
;
578 static inline int memslot_id(struct kvm
*kvm
, struct kvm_memory_slot
*slot
)
580 return slot
- kvm
->memslots
;
583 static inline struct kvm_mmu_page
*page_header(hpa_t shadow_page
)
585 struct page
*page
= pfn_to_page(shadow_page
>> PAGE_SHIFT
);
587 return (struct kvm_mmu_page
*)page_private(page
);
590 static inline u16
read_fs(void)
593 asm ("mov %%fs, %0" : "=g"(seg
));
597 static inline u16
read_gs(void)
600 asm ("mov %%gs, %0" : "=g"(seg
));
604 static inline u16
read_ldt(void)
607 asm ("sldt %0" : "=g"(ldt
));
611 static inline void load_fs(u16 sel
)
613 asm ("mov %0, %%fs" : : "rm"(sel
));
616 static inline void load_gs(u16 sel
)
618 asm ("mov %0, %%gs" : : "rm"(sel
));
622 static inline void load_ldt(u16 sel
)
624 asm ("lldt %0" : : "rm"(sel
));
628 static inline void get_idt(struct descriptor_table
*table
)
630 asm ("sidt %0" : "=m"(*table
));
633 static inline void get_gdt(struct descriptor_table
*table
)
635 asm ("sgdt %0" : "=m"(*table
));
638 static inline unsigned long read_tr_base(void)
641 asm ("str %0" : "=g"(tr
));
642 return segment_base(tr
);
646 static inline unsigned long read_msr(unsigned long msr
)
655 static inline void fx_save(void *image
)
657 asm ("fxsave (%0)":: "r" (image
));
660 static inline void fx_restore(void *image
)
662 asm ("fxrstor (%0)":: "r" (image
));
665 static inline void fpu_init(void)
670 static inline u32
get_rdx_init_val(void)
672 return 0x600; /* P6 family */
675 #define ASM_VMX_VMCLEAR_RAX ".byte 0x66, 0x0f, 0xc7, 0x30"
676 #define ASM_VMX_VMLAUNCH ".byte 0x0f, 0x01, 0xc2"
677 #define ASM_VMX_VMRESUME ".byte 0x0f, 0x01, 0xc3"
678 #define ASM_VMX_VMPTRLD_RAX ".byte 0x0f, 0xc7, 0x30"
679 #define ASM_VMX_VMREAD_RDX_RAX ".byte 0x0f, 0x78, 0xd0"
680 #define ASM_VMX_VMWRITE_RAX_RDX ".byte 0x0f, 0x79, 0xd0"
681 #define ASM_VMX_VMWRITE_RSP_RDX ".byte 0x0f, 0x79, 0xd4"
682 #define ASM_VMX_VMXOFF ".byte 0x0f, 0x01, 0xc4"
683 #define ASM_VMX_VMXON_RAX ".byte 0xf3, 0x0f, 0xc7, 0x30"
685 #define MSR_IA32_TIME_STAMP_COUNTER 0x010
687 #define TSS_IOPB_BASE_OFFSET 0x66
688 #define TSS_BASE_SIZE 0x68
689 #define TSS_IOPB_SIZE (65536 / 8)
690 #define TSS_REDIRECTION_SIZE (256 / 8)
691 #define RMODE_TSS_SIZE (TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)