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;
265 struct vcpu_svm *svm;
272 int interrupt_window_open;
273 unsigned long irq_summary; /* bit vector: 1 per word in irq_pending */
274 #define NR_IRQ_WORDS KVM_IRQ_BITMAP_SIZE(unsigned long)
275 unsigned long irq_pending[NR_IRQ_WORDS];
276 unsigned long regs[NR_VCPU_REGS]; /* for rsp: vcpu_load_rsp_rip() */
277 unsigned long rip; /* needs vcpu_load_rsp_rip() */
282 gpa_t para_state_gpa;
283 struct page *para_state_page;
287 u64 pdptrs[4]; /* pae */
290 u64 ia32_misc_enable_msr;
295 int msr_offset_kernel_gs_base;
297 struct vmx_msr_entry *guest_msrs;
298 struct vmx_msr_entry *host_msrs;
300 struct list_head free_pages;
301 struct kvm_mmu_page page_header_buf[KVM_NUM_MMU_PAGES];
304 struct kvm_mmu_memory_cache mmu_pte_chain_cache;
305 struct kvm_mmu_memory_cache mmu_rmap_desc_cache;
307 gfn_t last_pt_write_gfn;
308 int last_pt_write_count;
310 struct kvm_guest_debug guest_debug;
312 char fx_buf[FX_BUF_SIZE];
314 char *guest_fx_image;
316 int guest_fpu_loaded;
317 struct vmx_host_state {
319 u16 fs_sel, gs_sel, ldt_sel;
320 int fs_gs_ldt_reload_needed;
324 int mmio_read_completed;
327 unsigned char mmio_data[8];
328 gpa_t mmio_phys_addr;
329 gva_t mmio_fault_cr2;
330 struct kvm_pio_request pio;
336 struct kvm_stat stat;
341 struct kvm_save_segment {
346 } tr, es, ds, fs, gs;
350 struct kvm_cpuid_entry cpuid_entries[KVM_MAX_CPUID_ENTRIES];
353 struct kvm_mem_alias {
355 unsigned long npages;
359 struct kvm_memory_slot {
361 unsigned long npages;
363 struct page **phys_mem;
364 unsigned long *dirty_bitmap;
368 spinlock_t lock; /* protects everything except vcpus */
370 struct kvm_mem_alias aliases[KVM_ALIAS_SLOTS];
372 struct kvm_memory_slot memslots[KVM_MEMORY_SLOTS];
374 * Hash table of struct kvm_mmu_page.
376 struct list_head active_mmu_pages;
377 int n_free_mmu_pages;
378 struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
379 struct kvm_vcpu vcpus[KVM_MAX_VCPUS];
380 int memory_config_version;
382 unsigned long rmap_overflow;
383 struct list_head vm_list;
387 struct descriptor_table {
390 } __attribute__((packed));
392 struct kvm_arch_ops {
393 int (*cpu_has_kvm_support)(void); /* __init */
394 int (*disabled_by_bios)(void); /* __init */
395 void (*hardware_enable)(void *dummy); /* __init */
396 void (*hardware_disable)(void *dummy);
397 int (*hardware_setup)(void); /* __init */
398 void (*hardware_unsetup)(void); /* __exit */
400 int (*vcpu_create)(struct kvm_vcpu *vcpu);
401 void (*vcpu_free)(struct kvm_vcpu *vcpu);
403 void (*vcpu_load)(struct kvm_vcpu *vcpu);
404 void (*vcpu_put)(struct kvm_vcpu *vcpu);
405 void (*vcpu_decache)(struct kvm_vcpu *vcpu);
407 int (*set_guest_debug)(struct kvm_vcpu *vcpu,
408 struct kvm_debug_guest *dbg);
409 int (*get_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata);
410 int (*set_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 data);
411 u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
412 void (*get_segment)(struct kvm_vcpu *vcpu,
413 struct kvm_segment *var, int seg);
414 void (*set_segment)(struct kvm_vcpu *vcpu,
415 struct kvm_segment *var, int seg);
416 void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
417 void (*decache_cr4_guest_bits)(struct kvm_vcpu *vcpu);
418 void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
419 void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
420 void (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
421 void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
422 void (*get_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
423 void (*set_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
424 void (*get_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
425 void (*set_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
426 unsigned long (*get_dr)(struct kvm_vcpu *vcpu, int dr);
427 void (*set_dr)(struct kvm_vcpu *vcpu, int dr, unsigned long value,
429 void (*cache_regs)(struct kvm_vcpu *vcpu);
430 void (*decache_regs)(struct kvm_vcpu *vcpu);
431 unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
432 void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
434 void (*invlpg)(struct kvm_vcpu *vcpu, gva_t addr);
435 void (*tlb_flush)(struct kvm_vcpu *vcpu);
436 void (*inject_page_fault)(struct kvm_vcpu *vcpu,
437 unsigned long addr, u32 err_code);
439 void (*inject_gp)(struct kvm_vcpu *vcpu, unsigned err_code);
441 int (*run)(struct kvm_vcpu *vcpu, struct kvm_run *run);
442 int (*vcpu_setup)(struct kvm_vcpu *vcpu);
443 void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
444 void (*patch_hypercall)(struct kvm_vcpu *vcpu,
445 unsigned char *hypercall_addr);
448 extern struct kvm_arch_ops *kvm_arch_ops;
450 #define kvm_printf(kvm, fmt ...) printk(KERN_DEBUG fmt)
451 #define vcpu_printf(vcpu, fmt...) kvm_printf(vcpu->kvm, fmt)
453 int kvm_init_arch(struct kvm_arch_ops *ops, struct module *module);
454 void kvm_exit_arch(void);
456 int kvm_mmu_module_init(void);
457 void kvm_mmu_module_exit(void);
459 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
460 int kvm_mmu_create(struct kvm_vcpu *vcpu);
461 int kvm_mmu_setup(struct kvm_vcpu *vcpu);
463 int kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
464 void kvm_mmu_slot_remove_write_access(struct kvm_vcpu *vcpu, int slot);
465 void kvm_mmu_zap_all(struct kvm_vcpu *vcpu);
467 hpa_t gpa_to_hpa(struct kvm_vcpu *vcpu, gpa_t gpa);
468 #define HPA_MSB ((sizeof(hpa_t) * 8) - 1)
469 #define HPA_ERR_MASK ((hpa_t)1 << HPA_MSB)
470 static inline int is_error_hpa(hpa_t hpa) { return hpa >> HPA_MSB; }
471 hpa_t gva_to_hpa(struct kvm_vcpu *vcpu, gva_t gva);
472 struct page *gva_to_page(struct kvm_vcpu *vcpu, gva_t gva);
474 void kvm_emulator_want_group7_invlpg(void);
476 extern hpa_t bad_page_address;
478 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
479 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
480 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
482 enum emulation_result {
483 EMULATE_DONE, /* no further processing */
484 EMULATE_DO_MMIO, /* kvm_run filled with mmio request */
485 EMULATE_FAIL, /* can't emulate this instruction */
488 int emulate_instruction(struct kvm_vcpu *vcpu, struct kvm_run *run,
489 unsigned long cr2, u16 error_code);
490 void realmode_lgdt(struct kvm_vcpu *vcpu, u16 size, unsigned long address);
491 void realmode_lidt(struct kvm_vcpu *vcpu, u16 size, unsigned long address);
492 void realmode_lmsw(struct kvm_vcpu *vcpu, unsigned long msw,
493 unsigned long *rflags);
495 unsigned long realmode_get_cr(struct kvm_vcpu *vcpu, int cr);
496 void realmode_set_cr(struct kvm_vcpu *vcpu, int cr, unsigned long value,
497 unsigned long *rflags);
499 struct x86_emulate_ctxt;
501 int kvm_setup_pio(struct kvm_vcpu *vcpu, struct kvm_run *run, int in,
502 int size, unsigned long count, int string, int down,
503 gva_t address, int rep, unsigned port);
504 void kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
505 int emulate_invlpg(struct kvm_vcpu *vcpu, gva_t address);
506 int emulate_clts(struct kvm_vcpu *vcpu);
507 int emulator_get_dr(struct x86_emulate_ctxt* ctxt, int dr,
508 unsigned long *dest);
509 int emulator_set_dr(struct x86_emulate_ctxt *ctxt, int dr,
510 unsigned long value);
512 void set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
513 void set_cr3(struct kvm_vcpu *vcpu, unsigned long cr0);
514 void set_cr4(struct kvm_vcpu *vcpu, unsigned long cr0);
515 void set_cr8(struct kvm_vcpu *vcpu, unsigned long cr0);
516 void lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
518 int kvm_get_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata);
519 int kvm_set_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 data);
521 void fx_init(struct kvm_vcpu *vcpu);
523 void load_msrs(struct vmx_msr_entry *e, int n);
524 void save_msrs(struct vmx_msr_entry *e, int n);
525 void kvm_resched(struct kvm_vcpu *vcpu);
526 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
527 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
529 int kvm_read_guest(struct kvm_vcpu *vcpu,
534 int kvm_write_guest(struct kvm_vcpu *vcpu,
539 unsigned long segment_base(u16 selector);
541 void kvm_mmu_pte_write(struct kvm_vcpu *vcpu, gpa_t gpa,
542 const u8 *old, const u8 *new, int bytes);
543 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva);
544 void kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
546 int kvm_hypercall(struct kvm_vcpu *vcpu, struct kvm_run *run);
548 static inline int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t gva,
551 if (unlikely(vcpu->kvm->n_free_mmu_pages < KVM_MIN_FREE_MMU_PAGES))
552 kvm_mmu_free_some_pages(vcpu);
553 return vcpu->mmu.page_fault(vcpu, gva, error_code);
556 static inline int is_long_mode(struct kvm_vcpu *vcpu)
559 return vcpu->shadow_efer & EFER_LME;
565 static inline int is_pae(struct kvm_vcpu *vcpu)
567 return vcpu->cr4 & CR4_PAE_MASK;
570 static inline int is_pse(struct kvm_vcpu *vcpu)
572 return vcpu->cr4 & CR4_PSE_MASK;
575 static inline int is_paging(struct kvm_vcpu *vcpu)
577 return vcpu->cr0 & CR0_PG_MASK;
580 static inline int memslot_id(struct kvm *kvm, struct kvm_memory_slot *slot)
582 return slot - kvm->memslots;
585 static inline struct kvm_mmu_page *page_header(hpa_t shadow_page)
587 struct page *page = pfn_to_page(shadow_page >> PAGE_SHIFT);
589 return (struct kvm_mmu_page *)page_private(page);
592 static inline u16 read_fs(void)
595 asm ("mov %%fs, %0" : "=g"(seg));
599 static inline u16 read_gs(void)
602 asm ("mov %%gs, %0" : "=g"(seg));
606 static inline u16 read_ldt(void)
609 asm ("sldt %0" : "=g"(ldt));
613 static inline void load_fs(u16 sel)
615 asm ("mov %0, %%fs" : : "rm"(sel));
618 static inline void load_gs(u16 sel)
620 asm ("mov %0, %%gs" : : "rm"(sel));
624 static inline void load_ldt(u16 sel)
626 asm ("lldt %0" : : "rm"(sel));
630 static inline void get_idt(struct descriptor_table *table)
632 asm ("sidt %0" : "=m"(*table));
635 static inline void get_gdt(struct descriptor_table *table)
637 asm ("sgdt %0" : "=m"(*table));
640 static inline unsigned long read_tr_base(void)
643 asm ("str %0" : "=g"(tr));
644 return segment_base(tr);
648 static inline unsigned long read_msr(unsigned long msr)
657 static inline void fx_save(void *image)
659 asm ("fxsave (%0)":: "r" (image));
662 static inline void fx_restore(void *image)
664 asm ("fxrstor (%0)":: "r" (image));
667 static inline void fpu_init(void)
672 static inline u32 get_rdx_init_val(void)
674 return 0x600; /* P6 family */
677 #define ASM_VMX_VMCLEAR_RAX ".byte 0x66, 0x0f, 0xc7, 0x30"
678 #define ASM_VMX_VMLAUNCH ".byte 0x0f, 0x01, 0xc2"
679 #define ASM_VMX_VMRESUME ".byte 0x0f, 0x01, 0xc3"
680 #define ASM_VMX_VMPTRLD_RAX ".byte 0x0f, 0xc7, 0x30"
681 #define ASM_VMX_VMREAD_RDX_RAX ".byte 0x0f, 0x78, 0xd0"
682 #define ASM_VMX_VMWRITE_RAX_RDX ".byte 0x0f, 0x79, 0xd0"
683 #define ASM_VMX_VMWRITE_RSP_RDX ".byte 0x0f, 0x79, 0xd4"
684 #define ASM_VMX_VMXOFF ".byte 0x0f, 0x01, 0xc4"
685 #define ASM_VMX_VMXON_RAX ".byte 0xf3, 0x0f, 0xc7, 0x30"
687 #define MSR_IA32_TIME_STAMP_COUNTER 0x010
689 #define TSS_IOPB_BASE_OFFSET 0x66
690 #define TSS_BASE_SIZE 0x68
691 #define TSS_IOPB_SIZE (65536 / 8)
692 #define TSS_REDIRECTION_SIZE (256 / 8)
693 #define RMODE_TSS_SIZE (TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)