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
58 #define KVM_MAX_CPUID_ENTRIES 40
60 #define FX_IMAGE_SIZE 512
61 #define FX_IMAGE_ALIGN 16
62 #define FX_BUF_SIZE (2 * FX_IMAGE_SIZE + FX_IMAGE_ALIGN)
72 #define SELECTOR_TI_MASK (1 << 2)
73 #define SELECTOR_RPL_MASK 0x03
77 #define KVM_PIO_PAGE_OFFSET 1
82 * gva - guest virtual address
83 * gpa - guest physical address
84 * gfn - guest frame number
85 * hva - host virtual address
86 * hpa - host physical address
87 * hfn - host frame number
90 typedef unsigned long gva_t;
92 typedef unsigned long gfn_t;
94 typedef unsigned long hva_t;
96 typedef unsigned long hfn_t;
98 #define NR_PTE_CHAIN_ENTRIES 5
100 struct kvm_pte_chain {
101 u64 *parent_ptes[NR_PTE_CHAIN_ENTRIES];
102 struct hlist_node link;
106 * kvm_mmu_page_role, below, is defined as:
108 * bits 0:3 - total guest paging levels (2-4, or zero for real mode)
109 * bits 4:7 - page table level for this shadow (1-4)
110 * bits 8:9 - page table quadrant for 2-level guests
111 * bit 16 - "metaphysical" - gfn is not a real page (huge page/real mode)
113 union kvm_mmu_page_role {
116 unsigned glevels : 4;
118 unsigned quadrant : 2;
119 unsigned pad_for_nice_hex_output : 6;
120 unsigned metaphysical : 1;
124 struct kvm_mmu_page {
125 struct list_head link;
126 struct hlist_node hash_link;
129 * The following two entries are used to key the shadow page in the
133 union kvm_mmu_page_role role;
136 unsigned long slot_bitmap; /* One bit set per slot which has memory
137 * in this shadow page.
139 int global; /* Set if all ptes in this page are global */
140 int multimapped; /* More than one parent_pte? */
141 int root_count; /* Currently serving as active root */
143 u64 *parent_pte; /* !multimapped */
144 struct hlist_head parent_ptes; /* multimapped, kvm_pte_chain */
154 #define vmx_msr_entry kvm_msr_entry
159 * x86 supports 3 paging modes (4-level 64-bit, 3-level 64-bit, and 2-level
160 * 32-bit). The kvm_mmu structure abstracts the details of the current mmu
164 void (*new_cr3)(struct kvm_vcpu *vcpu);
165 int (*page_fault)(struct kvm_vcpu *vcpu, gva_t gva, u32 err);
166 void (*free)(struct kvm_vcpu *vcpu);
167 gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t gva);
170 int shadow_root_level;
175 #define KVM_NR_MEM_OBJS 20
177 struct kvm_mmu_memory_cache {
179 void *objects[KVM_NR_MEM_OBJS];
183 * We don't want allocation failures within the mmu code, so we preallocate
184 * enough memory for a single page fault in a cache.
186 struct kvm_guest_debug {
225 struct kvm_pio_request {
228 struct page *guest_pages[2];
229 unsigned guest_page_offset;
241 struct vcpu_svm *svm;
247 int interrupt_window_open;
248 unsigned long irq_summary; /* bit vector: 1 per word in irq_pending */
249 #define NR_IRQ_WORDS KVM_IRQ_BITMAP_SIZE(unsigned long)
250 unsigned long irq_pending[NR_IRQ_WORDS];
251 unsigned long regs[NR_VCPU_REGS]; /* for rsp: vcpu_load_rsp_rip() */
252 unsigned long rip; /* needs vcpu_load_rsp_rip() */
257 gpa_t para_state_gpa;
258 struct page *para_state_page;
262 u64 pdptrs[4]; /* pae */
265 u64 ia32_misc_enable_msr;
267 struct vmx_msr_entry *guest_msrs;
268 struct vmx_msr_entry *host_msrs;
270 struct list_head free_pages;
271 struct kvm_mmu_page page_header_buf[KVM_NUM_MMU_PAGES];
274 struct kvm_mmu_memory_cache mmu_pte_chain_cache;
275 struct kvm_mmu_memory_cache mmu_rmap_desc_cache;
277 gfn_t last_pt_write_gfn;
278 int last_pt_write_count;
280 struct kvm_guest_debug guest_debug;
282 char fx_buf[FX_BUF_SIZE];
284 char *guest_fx_image;
287 int mmio_read_completed;
290 unsigned char mmio_data[8];
291 gpa_t mmio_phys_addr;
292 struct kvm_pio_request pio;
301 struct kvm_save_segment {
306 } tr, es, ds, fs, gs;
310 struct kvm_cpuid_entry cpuid_entries[KVM_MAX_CPUID_ENTRIES];
313 struct kvm_memory_slot {
315 unsigned long npages;
317 struct page **phys_mem;
318 unsigned long *dirty_bitmap;
322 spinlock_t lock; /* protects everything except vcpus */
324 struct kvm_memory_slot memslots[KVM_MEMORY_SLOTS];
326 * Hash table of struct kvm_mmu_page.
328 struct list_head active_mmu_pages;
329 int n_free_mmu_pages;
330 struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
331 struct kvm_vcpu vcpus[KVM_MAX_VCPUS];
332 int memory_config_version;
334 unsigned long rmap_overflow;
335 struct list_head vm_list;
349 u32 irq_window_exits;
351 u32 request_irq_exits;
355 struct descriptor_table {
358 } __attribute__((packed));
360 struct kvm_arch_ops {
361 int (*cpu_has_kvm_support)(void); /* __init */
362 int (*disabled_by_bios)(void); /* __init */
363 void (*hardware_enable)(void *dummy); /* __init */
364 void (*hardware_disable)(void *dummy);
365 int (*hardware_setup)(void); /* __init */
366 void (*hardware_unsetup)(void); /* __exit */
368 int (*vcpu_create)(struct kvm_vcpu *vcpu);
369 void (*vcpu_free)(struct kvm_vcpu *vcpu);
371 void (*vcpu_load)(struct kvm_vcpu *vcpu);
372 void (*vcpu_put)(struct kvm_vcpu *vcpu);
373 void (*vcpu_decache)(struct kvm_vcpu *vcpu);
375 int (*set_guest_debug)(struct kvm_vcpu *vcpu,
376 struct kvm_debug_guest *dbg);
377 int (*get_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata);
378 int (*set_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 data);
379 u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
380 void (*get_segment)(struct kvm_vcpu *vcpu,
381 struct kvm_segment *var, int seg);
382 void (*set_segment)(struct kvm_vcpu *vcpu,
383 struct kvm_segment *var, int seg);
384 void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
385 void (*decache_cr0_cr4_guest_bits)(struct kvm_vcpu *vcpu);
386 void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
387 void (*set_cr0_no_modeswitch)(struct kvm_vcpu *vcpu,
389 void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
390 void (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
391 void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
392 void (*get_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
393 void (*set_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
394 void (*get_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
395 void (*set_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
396 unsigned long (*get_dr)(struct kvm_vcpu *vcpu, int dr);
397 void (*set_dr)(struct kvm_vcpu *vcpu, int dr, unsigned long value,
399 void (*cache_regs)(struct kvm_vcpu *vcpu);
400 void (*decache_regs)(struct kvm_vcpu *vcpu);
401 unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
402 void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
404 void (*invlpg)(struct kvm_vcpu *vcpu, gva_t addr);
405 void (*tlb_flush)(struct kvm_vcpu *vcpu);
406 void (*inject_page_fault)(struct kvm_vcpu *vcpu,
407 unsigned long addr, u32 err_code);
409 void (*inject_gp)(struct kvm_vcpu *vcpu, unsigned err_code);
411 int (*run)(struct kvm_vcpu *vcpu, struct kvm_run *run);
412 int (*vcpu_setup)(struct kvm_vcpu *vcpu);
413 void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
414 void (*patch_hypercall)(struct kvm_vcpu *vcpu,
415 unsigned char *hypercall_addr);
418 extern struct kvm_stat kvm_stat;
419 extern struct kvm_arch_ops *kvm_arch_ops;
421 #define kvm_printf(kvm, fmt ...) printk(KERN_DEBUG fmt)
422 #define vcpu_printf(vcpu, fmt...) kvm_printf(vcpu->kvm, fmt)
424 int kvm_init_arch(struct kvm_arch_ops *ops, struct module *module);
425 void kvm_exit_arch(void);
427 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
428 int kvm_mmu_create(struct kvm_vcpu *vcpu);
429 int kvm_mmu_setup(struct kvm_vcpu *vcpu);
431 int kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
432 void kvm_mmu_slot_remove_write_access(struct kvm_vcpu *vcpu, int slot);
434 hpa_t gpa_to_hpa(struct kvm_vcpu *vcpu, gpa_t gpa);
435 #define HPA_MSB ((sizeof(hpa_t) * 8) - 1)
436 #define HPA_ERR_MASK ((hpa_t)1 << HPA_MSB)
437 static inline int is_error_hpa(hpa_t hpa) { return hpa >> HPA_MSB; }
438 hpa_t gva_to_hpa(struct kvm_vcpu *vcpu, gva_t gva);
439 struct page *gva_to_page(struct kvm_vcpu *vcpu, gva_t gva);
441 void kvm_emulator_want_group7_invlpg(void);
443 extern hpa_t bad_page_address;
445 static inline struct page *gfn_to_page(struct kvm_memory_slot *slot, gfn_t gfn)
447 return slot->phys_mem[gfn - slot->base_gfn];
450 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
451 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
453 enum emulation_result {
454 EMULATE_DONE, /* no further processing */
455 EMULATE_DO_MMIO, /* kvm_run filled with mmio request */
456 EMULATE_FAIL, /* can't emulate this instruction */
459 int emulate_instruction(struct kvm_vcpu *vcpu, struct kvm_run *run,
460 unsigned long cr2, u16 error_code);
461 void realmode_lgdt(struct kvm_vcpu *vcpu, u16 size, unsigned long address);
462 void realmode_lidt(struct kvm_vcpu *vcpu, u16 size, unsigned long address);
463 void realmode_lmsw(struct kvm_vcpu *vcpu, unsigned long msw,
464 unsigned long *rflags);
466 unsigned long realmode_get_cr(struct kvm_vcpu *vcpu, int cr);
467 void realmode_set_cr(struct kvm_vcpu *vcpu, int cr, unsigned long value,
468 unsigned long *rflags);
470 struct x86_emulate_ctxt;
472 int kvm_setup_pio(struct kvm_vcpu *vcpu, struct kvm_run *run, int in,
473 int size, unsigned long count, int string, int down,
474 gva_t address, int rep, unsigned port);
475 void kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
476 int emulate_invlpg(struct kvm_vcpu *vcpu, gva_t address);
477 int emulate_clts(struct kvm_vcpu *vcpu);
478 int emulator_get_dr(struct x86_emulate_ctxt* ctxt, int dr,
479 unsigned long *dest);
480 int emulator_set_dr(struct x86_emulate_ctxt *ctxt, int dr,
481 unsigned long value);
483 void set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
484 void set_cr3(struct kvm_vcpu *vcpu, unsigned long cr0);
485 void set_cr4(struct kvm_vcpu *vcpu, unsigned long cr0);
486 void set_cr8(struct kvm_vcpu *vcpu, unsigned long cr0);
487 void lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
489 int kvm_get_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata);
490 int kvm_set_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 data);
492 void fx_init(struct kvm_vcpu *vcpu);
494 void load_msrs(struct vmx_msr_entry *e, int n);
495 void save_msrs(struct vmx_msr_entry *e, int n);
496 void kvm_resched(struct kvm_vcpu *vcpu);
498 int kvm_read_guest(struct kvm_vcpu *vcpu,
503 int kvm_write_guest(struct kvm_vcpu *vcpu,
508 unsigned long segment_base(u16 selector);
510 void kvm_mmu_pre_write(struct kvm_vcpu *vcpu, gpa_t gpa, int bytes);
511 void kvm_mmu_post_write(struct kvm_vcpu *vcpu, gpa_t gpa, int bytes);
512 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva);
513 void kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
515 int kvm_hypercall(struct kvm_vcpu *vcpu, struct kvm_run *run);
517 static inline int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t gva,
520 if (unlikely(vcpu->kvm->n_free_mmu_pages < KVM_MIN_FREE_MMU_PAGES))
521 kvm_mmu_free_some_pages(vcpu);
522 return vcpu->mmu.page_fault(vcpu, gva, error_code);
525 static inline struct page *_gfn_to_page(struct kvm *kvm, gfn_t gfn)
527 struct kvm_memory_slot *slot = gfn_to_memslot(kvm, gfn);
528 return (slot) ? slot->phys_mem[gfn - slot->base_gfn] : NULL;
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)