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_TS_MASK (1ULL << 3)
22 #define CR0_NE_MASK (1ULL << 5)
23 #define CR0_WP_MASK (1ULL << 16)
24 #define CR0_NW_MASK (1ULL << 29)
25 #define CR0_CD_MASK (1ULL << 30)
26 #define CR0_PG_MASK (1ULL << 31)
28 #define CR3_WPT_MASK (1ULL << 3)
29 #define CR3_PCD_MASK (1ULL << 4)
31 #define CR3_RESEVED_BITS 0x07ULL
32 #define CR3_L_MODE_RESEVED_BITS (~((1ULL << 40) - 1) | 0x0fe7ULL)
33 #define CR3_FLAGS_MASK ((1ULL << 5) - 1)
35 #define CR4_VME_MASK (1ULL << 0)
36 #define CR4_PSE_MASK (1ULL << 4)
37 #define CR4_PAE_MASK (1ULL << 5)
38 #define CR4_PGE_MASK (1ULL << 7)
39 #define CR4_VMXE_MASK (1ULL << 13)
41 #define KVM_GUEST_CR0_MASK \
42 (CR0_PG_MASK | CR0_PE_MASK | CR0_WP_MASK | CR0_NE_MASK \
43 | CR0_NW_MASK | CR0_CD_MASK)
44 #define KVM_VM_CR0_ALWAYS_ON \
45 (CR0_PG_MASK | CR0_PE_MASK | CR0_WP_MASK | CR0_NE_MASK | CR0_TS_MASK)
46 #define KVM_GUEST_CR4_MASK \
47 (CR4_PSE_MASK | CR4_PAE_MASK | CR4_PGE_MASK | CR4_VMXE_MASK | CR4_VME_MASK)
48 #define KVM_PMODE_VM_CR4_ALWAYS_ON (CR4_VMXE_MASK | CR4_PAE_MASK)
49 #define KVM_RMODE_VM_CR4_ALWAYS_ON (CR4_VMXE_MASK | CR4_PAE_MASK | CR4_VME_MASK)
51 #define INVALID_PAGE (~(hpa_t)0)
52 #define UNMAPPED_GVA (~(gpa_t)0)
54 #define KVM_MAX_VCPUS 1
55 #define KVM_ALIAS_SLOTS 4
56 #define KVM_MEMORY_SLOTS 4
57 #define KVM_NUM_MMU_PAGES 1024
58 #define KVM_MIN_FREE_MMU_PAGES 5
59 #define KVM_REFILL_PAGES 25
60 #define KVM_MAX_CPUID_ENTRIES 40
62 #define FX_IMAGE_SIZE 512
63 #define FX_IMAGE_ALIGN 16
64 #define FX_BUF_SIZE (2 * FX_IMAGE_SIZE + FX_IMAGE_ALIGN)
75 #define SELECTOR_TI_MASK (1 << 2)
76 #define SELECTOR_RPL_MASK 0x03
80 #define KVM_PIO_PAGE_OFFSET 1
85 * gva - guest virtual address
86 * gpa - guest physical address
87 * gfn - guest frame number
88 * hva - host virtual address
89 * hpa - host physical address
90 * hfn - host frame number
93 typedef unsigned long gva_t;
95 typedef unsigned long gfn_t;
97 typedef unsigned long hva_t;
99 typedef unsigned long hfn_t;
101 #define NR_PTE_CHAIN_ENTRIES 5
103 struct kvm_pte_chain {
104 u64 *parent_ptes[NR_PTE_CHAIN_ENTRIES];
105 struct hlist_node link;
109 * kvm_mmu_page_role, below, is defined as:
111 * bits 0:3 - total guest paging levels (2-4, or zero for real mode)
112 * bits 4:7 - page table level for this shadow (1-4)
113 * bits 8:9 - page table quadrant for 2-level guests
114 * bit 16 - "metaphysical" - gfn is not a real page (huge page/real mode)
115 * bits 17:18 - "access" - the user and writable bits of a huge page pde
117 union kvm_mmu_page_role {
120 unsigned glevels : 4;
122 unsigned quadrant : 2;
123 unsigned pad_for_nice_hex_output : 6;
124 unsigned metaphysical : 1;
125 unsigned hugepage_access : 2;
129 struct kvm_mmu_page {
130 struct list_head link;
131 struct hlist_node hash_link;
134 * The following two entries are used to key the shadow page in the
138 union kvm_mmu_page_role role;
141 unsigned long slot_bitmap; /* One bit set per slot which has memory
142 * in this shadow page.
144 int multimapped; /* More than one parent_pte? */
145 int root_count; /* Currently serving as active root */
147 u64 *parent_pte; /* !multimapped */
148 struct hlist_head parent_ptes; /* multimapped, kvm_pte_chain */
158 #define vmx_msr_entry kvm_msr_entry
163 * x86 supports 3 paging modes (4-level 64-bit, 3-level 64-bit, and 2-level
164 * 32-bit). The kvm_mmu structure abstracts the details of the current mmu
168 void (*new_cr3)(struct kvm_vcpu *vcpu);
169 int (*page_fault)(struct kvm_vcpu *vcpu, gva_t gva, u32 err);
170 void (*free)(struct kvm_vcpu *vcpu);
171 gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t gva);
174 int shadow_root_level;
179 #define KVM_NR_MEM_OBJS 20
181 struct kvm_mmu_memory_cache {
183 void *objects[KVM_NR_MEM_OBJS];
187 * We don't want allocation failures within the mmu code, so we preallocate
188 * enough memory for a single page fault in a cache.
190 struct kvm_guest_debug {
229 struct kvm_pio_request {
232 struct page *guest_pages[2];
233 unsigned guest_page_offset;
251 u32 irq_window_exits;
253 u32 request_irq_exits;
262 struct vcpu_svm *svm;
269 int interrupt_window_open;
270 unsigned long irq_summary; /* bit vector: 1 per word in irq_pending */
271 #define NR_IRQ_WORDS KVM_IRQ_BITMAP_SIZE(unsigned long)
272 unsigned long irq_pending[NR_IRQ_WORDS];
273 unsigned long regs[NR_VCPU_REGS]; /* for rsp: vcpu_load_rsp_rip() */
274 unsigned long rip; /* needs vcpu_load_rsp_rip() */
279 gpa_t para_state_gpa;
280 struct page *para_state_page;
284 u64 pdptrs[4]; /* pae */
287 u64 ia32_misc_enable_msr;
289 struct vmx_msr_entry *guest_msrs;
290 struct vmx_msr_entry *host_msrs;
292 struct list_head free_pages;
293 struct kvm_mmu_page page_header_buf[KVM_NUM_MMU_PAGES];
296 struct kvm_mmu_memory_cache mmu_pte_chain_cache;
297 struct kvm_mmu_memory_cache mmu_rmap_desc_cache;
299 gfn_t last_pt_write_gfn;
300 int last_pt_write_count;
302 struct kvm_guest_debug guest_debug;
304 char fx_buf[FX_BUF_SIZE];
306 char *guest_fx_image;
308 int guest_fpu_loaded;
309 struct vmx_host_state {
311 u16 fs_sel, gs_sel, ldt_sel;
312 int fs_gs_ldt_reload_needed;
316 int mmio_read_completed;
319 unsigned char mmio_data[8];
320 gpa_t mmio_phys_addr;
321 gva_t mmio_fault_cr2;
322 struct kvm_pio_request pio;
328 struct kvm_stat stat;
333 struct kvm_save_segment {
338 } tr, es, ds, fs, gs;
342 struct kvm_cpuid_entry cpuid_entries[KVM_MAX_CPUID_ENTRIES];
345 struct kvm_mem_alias {
347 unsigned long npages;
351 struct kvm_memory_slot {
353 unsigned long npages;
355 struct page **phys_mem;
356 unsigned long *dirty_bitmap;
360 spinlock_t lock; /* protects everything except vcpus */
362 struct kvm_mem_alias aliases[KVM_ALIAS_SLOTS];
364 struct kvm_memory_slot memslots[KVM_MEMORY_SLOTS];
366 * Hash table of struct kvm_mmu_page.
368 struct list_head active_mmu_pages;
369 int n_free_mmu_pages;
370 struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
371 struct kvm_vcpu vcpus[KVM_MAX_VCPUS];
372 int memory_config_version;
374 unsigned long rmap_overflow;
375 struct list_head vm_list;
379 struct descriptor_table {
382 } __attribute__((packed));
384 struct kvm_arch_ops {
385 int (*cpu_has_kvm_support)(void); /* __init */
386 int (*disabled_by_bios)(void); /* __init */
387 void (*hardware_enable)(void *dummy); /* __init */
388 void (*hardware_disable)(void *dummy);
389 int (*hardware_setup)(void); /* __init */
390 void (*hardware_unsetup)(void); /* __exit */
392 int (*vcpu_create)(struct kvm_vcpu *vcpu);
393 void (*vcpu_free)(struct kvm_vcpu *vcpu);
395 void (*vcpu_load)(struct kvm_vcpu *vcpu);
396 void (*vcpu_put)(struct kvm_vcpu *vcpu);
397 void (*vcpu_decache)(struct kvm_vcpu *vcpu);
399 int (*set_guest_debug)(struct kvm_vcpu *vcpu,
400 struct kvm_debug_guest *dbg);
401 int (*get_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata);
402 int (*set_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 data);
403 u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
404 void (*get_segment)(struct kvm_vcpu *vcpu,
405 struct kvm_segment *var, int seg);
406 void (*set_segment)(struct kvm_vcpu *vcpu,
407 struct kvm_segment *var, int seg);
408 void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
409 void (*decache_cr4_guest_bits)(struct kvm_vcpu *vcpu);
410 void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
411 void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
412 void (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
413 void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
414 void (*get_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
415 void (*set_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
416 void (*get_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
417 void (*set_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
418 unsigned long (*get_dr)(struct kvm_vcpu *vcpu, int dr);
419 void (*set_dr)(struct kvm_vcpu *vcpu, int dr, unsigned long value,
421 void (*cache_regs)(struct kvm_vcpu *vcpu);
422 void (*decache_regs)(struct kvm_vcpu *vcpu);
423 unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
424 void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
426 void (*invlpg)(struct kvm_vcpu *vcpu, gva_t addr);
427 void (*tlb_flush)(struct kvm_vcpu *vcpu);
428 void (*inject_page_fault)(struct kvm_vcpu *vcpu,
429 unsigned long addr, u32 err_code);
431 void (*inject_gp)(struct kvm_vcpu *vcpu, unsigned err_code);
433 int (*run)(struct kvm_vcpu *vcpu, struct kvm_run *run);
434 int (*vcpu_setup)(struct kvm_vcpu *vcpu);
435 void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
436 void (*patch_hypercall)(struct kvm_vcpu *vcpu,
437 unsigned char *hypercall_addr);
440 extern struct kvm_arch_ops *kvm_arch_ops;
442 #define kvm_printf(kvm, fmt ...) printk(KERN_DEBUG fmt)
443 #define vcpu_printf(vcpu, fmt...) kvm_printf(vcpu->kvm, fmt)
445 int kvm_init_arch(struct kvm_arch_ops *ops, struct module *module);
446 void kvm_exit_arch(void);
448 int kvm_mmu_module_init(void);
449 void kvm_mmu_module_exit(void);
451 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
452 int kvm_mmu_create(struct kvm_vcpu *vcpu);
453 int kvm_mmu_setup(struct kvm_vcpu *vcpu);
455 int kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
456 void kvm_mmu_slot_remove_write_access(struct kvm_vcpu *vcpu, int slot);
457 void kvm_mmu_zap_all(struct kvm_vcpu *vcpu);
459 hpa_t gpa_to_hpa(struct kvm_vcpu *vcpu, gpa_t gpa);
460 #define HPA_MSB ((sizeof(hpa_t) * 8) - 1)
461 #define HPA_ERR_MASK ((hpa_t)1 << HPA_MSB)
462 static inline int is_error_hpa(hpa_t hpa) { return hpa >> HPA_MSB; }
463 hpa_t gva_to_hpa(struct kvm_vcpu *vcpu, gva_t gva);
464 struct page *gva_to_page(struct kvm_vcpu *vcpu, gva_t gva);
466 void kvm_emulator_want_group7_invlpg(void);
468 extern hpa_t bad_page_address;
470 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
471 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
472 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
474 enum emulation_result {
475 EMULATE_DONE, /* no further processing */
476 EMULATE_DO_MMIO, /* kvm_run filled with mmio request */
477 EMULATE_FAIL, /* can't emulate this instruction */
480 int emulate_instruction(struct kvm_vcpu *vcpu, struct kvm_run *run,
481 unsigned long cr2, u16 error_code);
482 void realmode_lgdt(struct kvm_vcpu *vcpu, u16 size, unsigned long address);
483 void realmode_lidt(struct kvm_vcpu *vcpu, u16 size, unsigned long address);
484 void realmode_lmsw(struct kvm_vcpu *vcpu, unsigned long msw,
485 unsigned long *rflags);
487 unsigned long realmode_get_cr(struct kvm_vcpu *vcpu, int cr);
488 void realmode_set_cr(struct kvm_vcpu *vcpu, int cr, unsigned long value,
489 unsigned long *rflags);
491 struct x86_emulate_ctxt;
493 int kvm_setup_pio(struct kvm_vcpu *vcpu, struct kvm_run *run, int in,
494 int size, unsigned long count, int string, int down,
495 gva_t address, int rep, unsigned port);
496 void kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
497 int emulate_invlpg(struct kvm_vcpu *vcpu, gva_t address);
498 int emulate_clts(struct kvm_vcpu *vcpu);
499 int emulator_get_dr(struct x86_emulate_ctxt* ctxt, int dr,
500 unsigned long *dest);
501 int emulator_set_dr(struct x86_emulate_ctxt *ctxt, int dr,
502 unsigned long value);
504 void set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
505 void set_cr3(struct kvm_vcpu *vcpu, unsigned long cr0);
506 void set_cr4(struct kvm_vcpu *vcpu, unsigned long cr0);
507 void set_cr8(struct kvm_vcpu *vcpu, unsigned long cr0);
508 void lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
510 int kvm_get_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata);
511 int kvm_set_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 data);
513 void fx_init(struct kvm_vcpu *vcpu);
515 void load_msrs(struct vmx_msr_entry *e, int n);
516 void save_msrs(struct vmx_msr_entry *e, int n);
517 void kvm_resched(struct kvm_vcpu *vcpu);
518 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
519 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
521 int kvm_read_guest(struct kvm_vcpu *vcpu,
526 int kvm_write_guest(struct kvm_vcpu *vcpu,
531 unsigned long segment_base(u16 selector);
533 void kvm_mmu_pte_write(struct kvm_vcpu *vcpu, gpa_t gpa,
534 const u8 *old, const u8 *new, int bytes);
535 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva);
536 void kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
538 int kvm_hypercall(struct kvm_vcpu *vcpu, struct kvm_run *run);
540 static inline int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t gva,
543 if (unlikely(vcpu->kvm->n_free_mmu_pages < KVM_MIN_FREE_MMU_PAGES))
544 kvm_mmu_free_some_pages(vcpu);
545 return vcpu->mmu.page_fault(vcpu, gva, error_code);
548 static inline int is_long_mode(struct kvm_vcpu *vcpu)
551 return vcpu->shadow_efer & EFER_LME;
557 static inline int is_pae(struct kvm_vcpu *vcpu)
559 return vcpu->cr4 & CR4_PAE_MASK;
562 static inline int is_pse(struct kvm_vcpu *vcpu)
564 return vcpu->cr4 & CR4_PSE_MASK;
567 static inline int is_paging(struct kvm_vcpu *vcpu)
569 return vcpu->cr0 & CR0_PG_MASK;
572 static inline int memslot_id(struct kvm *kvm, struct kvm_memory_slot *slot)
574 return slot - kvm->memslots;
577 static inline struct kvm_mmu_page *page_header(hpa_t shadow_page)
579 struct page *page = pfn_to_page(shadow_page >> PAGE_SHIFT);
581 return (struct kvm_mmu_page *)page_private(page);
584 static inline u16 read_fs(void)
587 asm ("mov %%fs, %0" : "=g"(seg));
591 static inline u16 read_gs(void)
594 asm ("mov %%gs, %0" : "=g"(seg));
598 static inline u16 read_ldt(void)
601 asm ("sldt %0" : "=g"(ldt));
605 static inline void load_fs(u16 sel)
607 asm ("mov %0, %%fs" : : "rm"(sel));
610 static inline void load_gs(u16 sel)
612 asm ("mov %0, %%gs" : : "rm"(sel));
616 static inline void load_ldt(u16 sel)
618 asm ("lldt %0" : : "rm"(sel));
622 static inline void get_idt(struct descriptor_table *table)
624 asm ("sidt %0" : "=m"(*table));
627 static inline void get_gdt(struct descriptor_table *table)
629 asm ("sgdt %0" : "=m"(*table));
632 static inline unsigned long read_tr_base(void)
635 asm ("str %0" : "=g"(tr));
636 return segment_base(tr);
640 static inline unsigned long read_msr(unsigned long msr)
649 static inline void fx_save(void *image)
651 asm ("fxsave (%0)":: "r" (image));
654 static inline void fx_restore(void *image)
656 asm ("fxrstor (%0)":: "r" (image));
659 static inline void fpu_init(void)
664 static inline u32 get_rdx_init_val(void)
666 return 0x600; /* P6 family */
669 #define ASM_VMX_VMCLEAR_RAX ".byte 0x66, 0x0f, 0xc7, 0x30"
670 #define ASM_VMX_VMLAUNCH ".byte 0x0f, 0x01, 0xc2"
671 #define ASM_VMX_VMRESUME ".byte 0x0f, 0x01, 0xc3"
672 #define ASM_VMX_VMPTRLD_RAX ".byte 0x0f, 0xc7, 0x30"
673 #define ASM_VMX_VMREAD_RDX_RAX ".byte 0x0f, 0x78, 0xd0"
674 #define ASM_VMX_VMWRITE_RAX_RDX ".byte 0x0f, 0x79, 0xd0"
675 #define ASM_VMX_VMWRITE_RSP_RDX ".byte 0x0f, 0x79, 0xd4"
676 #define ASM_VMX_VMXOFF ".byte 0x0f, 0x01, 0xc4"
677 #define ASM_VMX_VMXON_RAX ".byte 0xf3, 0x0f, 0xc7, 0x30"
679 #define MSR_IA32_TIME_STAMP_COUNTER 0x010
681 #define TSS_IOPB_BASE_OFFSET 0x66
682 #define TSS_BASE_SIZE 0x68
683 #define TSS_IOPB_SIZE (65536 / 8)
684 #define TSS_REDIRECTION_SIZE (256 / 8)
685 #define RMODE_TSS_SIZE (TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)