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>
13 #include <linux/signal.h>
14 #include <linux/sched.h>
16 #include <asm/signal.h>
19 #include <linux/kvm.h>
20 #include <linux/kvm_para.h>
22 #define CR0_PE_MASK (1ULL << 0)
23 #define CR0_MP_MASK (1ULL << 1)
24 #define CR0_TS_MASK (1ULL << 3)
25 #define CR0_NE_MASK (1ULL << 5)
26 #define CR0_WP_MASK (1ULL << 16)
27 #define CR0_NW_MASK (1ULL << 29)
28 #define CR0_CD_MASK (1ULL << 30)
29 #define CR0_PG_MASK (1ULL << 31)
31 #define CR3_WPT_MASK (1ULL << 3)
32 #define CR3_PCD_MASK (1ULL << 4)
34 #define CR3_RESEVED_BITS 0x07ULL
35 #define CR3_L_MODE_RESEVED_BITS (~((1ULL << 40) - 1) | 0x0fe7ULL)
36 #define CR3_FLAGS_MASK ((1ULL << 5) - 1)
38 #define CR4_VME_MASK (1ULL << 0)
39 #define CR4_PSE_MASK (1ULL << 4)
40 #define CR4_PAE_MASK (1ULL << 5)
41 #define CR4_PGE_MASK (1ULL << 7)
42 #define CR4_VMXE_MASK (1ULL << 13)
44 #define KVM_GUEST_CR0_MASK \
45 (CR0_PG_MASK | CR0_PE_MASK | CR0_WP_MASK | CR0_NE_MASK \
46 | CR0_NW_MASK | CR0_CD_MASK)
47 #define KVM_VM_CR0_ALWAYS_ON \
48 (CR0_PG_MASK | CR0_PE_MASK | CR0_WP_MASK | CR0_NE_MASK | CR0_TS_MASK \
50 #define KVM_GUEST_CR4_MASK \
51 (CR4_PSE_MASK | CR4_PAE_MASK | CR4_PGE_MASK | CR4_VMXE_MASK | CR4_VME_MASK)
52 #define KVM_PMODE_VM_CR4_ALWAYS_ON (CR4_VMXE_MASK | CR4_PAE_MASK)
53 #define KVM_RMODE_VM_CR4_ALWAYS_ON (CR4_VMXE_MASK | CR4_PAE_MASK | CR4_VME_MASK)
55 #define INVALID_PAGE (~(hpa_t)0)
56 #define UNMAPPED_GVA (~(gpa_t)0)
58 #define KVM_MAX_VCPUS 4
59 #define KVM_ALIAS_SLOTS 4
60 #define KVM_MEMORY_SLOTS 4
61 #define KVM_NUM_MMU_PAGES 1024
62 #define KVM_MIN_FREE_MMU_PAGES 5
63 #define KVM_REFILL_PAGES 25
64 #define KVM_MAX_CPUID_ENTRIES 40
66 #define FX_IMAGE_SIZE 512
67 #define FX_IMAGE_ALIGN 16
68 #define FX_BUF_SIZE (2 * FX_IMAGE_SIZE + FX_IMAGE_ALIGN)
79 #define SELECTOR_TI_MASK (1 << 2)
80 #define SELECTOR_RPL_MASK 0x03
84 #define KVM_PIO_PAGE_OFFSET 1
89 * gva - guest virtual address
90 * gpa - guest physical address
91 * gfn - guest frame number
92 * hva - host virtual address
93 * hpa - host physical address
94 * hfn - host frame number
97 typedef unsigned long gva_t;
99 typedef unsigned long gfn_t;
101 typedef unsigned long hva_t;
103 typedef unsigned long hfn_t;
105 #define NR_PTE_CHAIN_ENTRIES 5
107 struct kvm_pte_chain {
108 u64 *parent_ptes[NR_PTE_CHAIN_ENTRIES];
109 struct hlist_node link;
113 * kvm_mmu_page_role, below, is defined as:
115 * bits 0:3 - total guest paging levels (2-4, or zero for real mode)
116 * bits 4:7 - page table level for this shadow (1-4)
117 * bits 8:9 - page table quadrant for 2-level guests
118 * bit 16 - "metaphysical" - gfn is not a real page (huge page/real mode)
119 * bits 17:18 - "access" - the user and writable bits of a huge page pde
121 union kvm_mmu_page_role {
124 unsigned glevels : 4;
126 unsigned quadrant : 2;
127 unsigned pad_for_nice_hex_output : 6;
128 unsigned metaphysical : 1;
129 unsigned hugepage_access : 2;
133 struct kvm_mmu_page {
134 struct list_head link;
135 struct hlist_node hash_link;
138 * The following two entries are used to key the shadow page in the
142 union kvm_mmu_page_role role;
145 unsigned long slot_bitmap; /* One bit set per slot which has memory
146 * in this shadow page.
148 int multimapped; /* More than one parent_pte? */
149 int root_count; /* Currently serving as active root */
151 u64 *parent_pte; /* !multimapped */
152 struct hlist_head parent_ptes; /* multimapped, kvm_pte_chain */
162 #define vmx_msr_entry kvm_msr_entry
167 * x86 supports 3 paging modes (4-level 64-bit, 3-level 64-bit, and 2-level
168 * 32-bit). The kvm_mmu structure abstracts the details of the current mmu
172 void (*new_cr3)(struct kvm_vcpu *vcpu);
173 int (*page_fault)(struct kvm_vcpu *vcpu, gva_t gva, u32 err);
174 void (*free)(struct kvm_vcpu *vcpu);
175 gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t gva);
178 int shadow_root_level;
183 #define KVM_NR_MEM_OBJS 20
185 struct kvm_mmu_memory_cache {
187 void *objects[KVM_NR_MEM_OBJS];
191 * We don't want allocation failures within the mmu code, so we preallocate
192 * enough memory for a single page fault in a cache.
194 struct kvm_guest_debug {
233 struct kvm_pio_request {
236 struct page *guest_pages[2];
237 unsigned guest_page_offset;
255 u32 irq_window_exits;
257 u32 request_irq_exits;
267 struct vcpu_svm *svm;
274 int interrupt_window_open;
275 unsigned long irq_summary; /* bit vector: 1 per word in irq_pending */
276 #define NR_IRQ_WORDS KVM_IRQ_BITMAP_SIZE(unsigned long)
277 unsigned long irq_pending[NR_IRQ_WORDS];
278 unsigned long regs[NR_VCPU_REGS]; /* for rsp: vcpu_load_rsp_rip() */
279 unsigned long rip; /* needs vcpu_load_rsp_rip() */
284 gpa_t para_state_gpa;
285 struct page *para_state_page;
289 u64 pdptrs[4]; /* pae */
292 u64 ia32_misc_enable_msr;
297 int msr_offset_kernel_gs_base;
299 struct vmx_msr_entry *guest_msrs;
300 struct vmx_msr_entry *host_msrs;
304 struct kvm_mmu_memory_cache mmu_pte_chain_cache;
305 struct kvm_mmu_memory_cache mmu_rmap_desc_cache;
306 struct kvm_mmu_memory_cache mmu_page_cache;
307 struct kvm_mmu_memory_cache mmu_page_header_cache;
309 gfn_t last_pt_write_gfn;
310 int last_pt_write_count;
312 struct kvm_guest_debug guest_debug;
314 char fx_buf[FX_BUF_SIZE];
316 char *guest_fx_image;
318 int guest_fpu_loaded;
319 struct vmx_host_state {
321 u16 fs_sel, gs_sel, ldt_sel;
322 int fs_gs_ldt_reload_needed;
326 int mmio_read_completed;
329 unsigned char mmio_data[8];
330 gpa_t mmio_phys_addr;
331 gva_t mmio_fault_cr2;
332 struct kvm_pio_request pio;
338 struct kvm_stat stat;
343 struct kvm_save_segment {
348 } tr, es, ds, fs, gs;
352 struct kvm_cpuid_entry cpuid_entries[KVM_MAX_CPUID_ENTRIES];
355 struct kvm_mem_alias {
357 unsigned long npages;
361 struct kvm_memory_slot {
363 unsigned long npages;
365 struct page **phys_mem;
366 unsigned long *dirty_bitmap;
370 spinlock_t lock; /* protects everything except vcpus */
372 struct kvm_mem_alias aliases[KVM_ALIAS_SLOTS];
374 struct kvm_memory_slot memslots[KVM_MEMORY_SLOTS];
376 * Hash table of struct kvm_mmu_page.
378 struct list_head active_mmu_pages;
379 int n_free_mmu_pages;
380 struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
381 struct kvm_vcpu vcpus[KVM_MAX_VCPUS];
382 int memory_config_version;
384 unsigned long rmap_overflow;
385 struct list_head vm_list;
389 struct descriptor_table {
392 } __attribute__((packed));
394 struct kvm_arch_ops {
395 int (*cpu_has_kvm_support)(void); /* __init */
396 int (*disabled_by_bios)(void); /* __init */
397 void (*hardware_enable)(void *dummy); /* __init */
398 void (*hardware_disable)(void *dummy);
399 int (*hardware_setup)(void); /* __init */
400 void (*hardware_unsetup)(void); /* __exit */
402 int (*vcpu_create)(struct kvm_vcpu *vcpu);
403 void (*vcpu_free)(struct kvm_vcpu *vcpu);
405 void (*vcpu_load)(struct kvm_vcpu *vcpu);
406 void (*vcpu_put)(struct kvm_vcpu *vcpu);
407 void (*vcpu_decache)(struct kvm_vcpu *vcpu);
409 int (*set_guest_debug)(struct kvm_vcpu *vcpu,
410 struct kvm_debug_guest *dbg);
411 int (*get_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata);
412 int (*set_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 data);
413 u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
414 void (*get_segment)(struct kvm_vcpu *vcpu,
415 struct kvm_segment *var, int seg);
416 void (*set_segment)(struct kvm_vcpu *vcpu,
417 struct kvm_segment *var, int seg);
418 void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
419 void (*decache_cr4_guest_bits)(struct kvm_vcpu *vcpu);
420 void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
421 void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
422 void (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
423 void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
424 void (*get_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
425 void (*set_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
426 void (*get_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
427 void (*set_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
428 unsigned long (*get_dr)(struct kvm_vcpu *vcpu, int dr);
429 void (*set_dr)(struct kvm_vcpu *vcpu, int dr, unsigned long value,
431 void (*cache_regs)(struct kvm_vcpu *vcpu);
432 void (*decache_regs)(struct kvm_vcpu *vcpu);
433 unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
434 void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
436 void (*invlpg)(struct kvm_vcpu *vcpu, gva_t addr);
437 void (*tlb_flush)(struct kvm_vcpu *vcpu);
438 void (*inject_page_fault)(struct kvm_vcpu *vcpu,
439 unsigned long addr, u32 err_code);
441 void (*inject_gp)(struct kvm_vcpu *vcpu, unsigned err_code);
443 int (*run)(struct kvm_vcpu *vcpu, struct kvm_run *run);
444 int (*vcpu_setup)(struct kvm_vcpu *vcpu);
445 void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
446 void (*patch_hypercall)(struct kvm_vcpu *vcpu,
447 unsigned char *hypercall_addr);
450 extern struct kvm_arch_ops *kvm_arch_ops;
452 #define kvm_printf(kvm, fmt ...) printk(KERN_DEBUG fmt)
453 #define vcpu_printf(vcpu, fmt...) kvm_printf(vcpu->kvm, fmt)
455 int kvm_init_arch(struct kvm_arch_ops *ops, struct module *module);
456 void kvm_exit_arch(void);
458 int kvm_mmu_module_init(void);
459 void kvm_mmu_module_exit(void);
461 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
462 int kvm_mmu_create(struct kvm_vcpu *vcpu);
463 int kvm_mmu_setup(struct kvm_vcpu *vcpu);
465 int kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
466 void kvm_mmu_slot_remove_write_access(struct kvm_vcpu *vcpu, int slot);
467 void kvm_mmu_zap_all(struct kvm_vcpu *vcpu);
469 hpa_t gpa_to_hpa(struct kvm_vcpu *vcpu, gpa_t gpa);
470 #define HPA_MSB ((sizeof(hpa_t) * 8) - 1)
471 #define HPA_ERR_MASK ((hpa_t)1 << HPA_MSB)
472 static inline int is_error_hpa(hpa_t hpa) { return hpa >> HPA_MSB; }
473 hpa_t gva_to_hpa(struct kvm_vcpu *vcpu, gva_t gva);
474 struct page *gva_to_page(struct kvm_vcpu *vcpu, gva_t gva);
476 void kvm_emulator_want_group7_invlpg(void);
478 extern hpa_t bad_page_address;
480 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
481 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
482 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
484 enum emulation_result {
485 EMULATE_DONE, /* no further processing */
486 EMULATE_DO_MMIO, /* kvm_run filled with mmio request */
487 EMULATE_FAIL, /* can't emulate this instruction */
490 int emulate_instruction(struct kvm_vcpu *vcpu, struct kvm_run *run,
491 unsigned long cr2, u16 error_code);
492 void realmode_lgdt(struct kvm_vcpu *vcpu, u16 size, unsigned long address);
493 void realmode_lidt(struct kvm_vcpu *vcpu, u16 size, unsigned long address);
494 void realmode_lmsw(struct kvm_vcpu *vcpu, unsigned long msw,
495 unsigned long *rflags);
497 unsigned long realmode_get_cr(struct kvm_vcpu *vcpu, int cr);
498 void realmode_set_cr(struct kvm_vcpu *vcpu, int cr, unsigned long value,
499 unsigned long *rflags);
501 struct x86_emulate_ctxt;
503 int kvm_setup_pio(struct kvm_vcpu *vcpu, struct kvm_run *run, int in,
504 int size, unsigned long count, int string, int down,
505 gva_t address, int rep, unsigned port);
506 void kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
507 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
508 int emulate_invlpg(struct kvm_vcpu *vcpu, gva_t address);
509 int emulate_clts(struct kvm_vcpu *vcpu);
510 int emulator_get_dr(struct x86_emulate_ctxt* ctxt, int dr,
511 unsigned long *dest);
512 int emulator_set_dr(struct x86_emulate_ctxt *ctxt, int dr,
513 unsigned long value);
515 void set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
516 void set_cr3(struct kvm_vcpu *vcpu, unsigned long cr0);
517 void set_cr4(struct kvm_vcpu *vcpu, unsigned long cr0);
518 void set_cr8(struct kvm_vcpu *vcpu, unsigned long cr0);
519 void lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
521 int kvm_get_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata);
522 int kvm_set_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 data);
524 void fx_init(struct kvm_vcpu *vcpu);
526 void load_msrs(struct vmx_msr_entry *e, int n);
527 void save_msrs(struct vmx_msr_entry *e, int n);
528 void kvm_resched(struct kvm_vcpu *vcpu);
529 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
530 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
532 int kvm_read_guest(struct kvm_vcpu *vcpu,
537 int kvm_write_guest(struct kvm_vcpu *vcpu,
542 unsigned long segment_base(u16 selector);
544 void kvm_mmu_pte_write(struct kvm_vcpu *vcpu, gpa_t gpa,
545 const u8 *old, const u8 *new, int bytes);
546 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva);
547 void kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
548 int kvm_mmu_load(struct kvm_vcpu *vcpu);
549 void kvm_mmu_unload(struct kvm_vcpu *vcpu);
551 int kvm_hypercall(struct kvm_vcpu *vcpu, struct kvm_run *run);
553 static inline int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t gva,
556 if (unlikely(vcpu->kvm->n_free_mmu_pages < KVM_MIN_FREE_MMU_PAGES))
557 kvm_mmu_free_some_pages(vcpu);
558 return vcpu->mmu.page_fault(vcpu, gva, error_code);
561 static inline int kvm_mmu_reload(struct kvm_vcpu *vcpu)
563 if (likely(vcpu->mmu.root_hpa != INVALID_PAGE))
566 return kvm_mmu_load(vcpu);
569 static inline int is_long_mode(struct kvm_vcpu *vcpu)
572 return vcpu->shadow_efer & EFER_LME;
578 static inline int is_pae(struct kvm_vcpu *vcpu)
580 return vcpu->cr4 & CR4_PAE_MASK;
583 static inline int is_pse(struct kvm_vcpu *vcpu)
585 return vcpu->cr4 & CR4_PSE_MASK;
588 static inline int is_paging(struct kvm_vcpu *vcpu)
590 return vcpu->cr0 & CR0_PG_MASK;
593 static inline int memslot_id(struct kvm *kvm, struct kvm_memory_slot *slot)
595 return slot - kvm->memslots;
598 static inline struct kvm_mmu_page *page_header(hpa_t shadow_page)
600 struct page *page = pfn_to_page(shadow_page >> PAGE_SHIFT);
602 return (struct kvm_mmu_page *)page_private(page);
605 static inline u16 read_fs(void)
608 asm ("mov %%fs, %0" : "=g"(seg));
612 static inline u16 read_gs(void)
615 asm ("mov %%gs, %0" : "=g"(seg));
619 static inline u16 read_ldt(void)
622 asm ("sldt %0" : "=g"(ldt));
626 static inline void load_fs(u16 sel)
628 asm ("mov %0, %%fs" : : "rm"(sel));
631 static inline void load_gs(u16 sel)
633 asm ("mov %0, %%gs" : : "rm"(sel));
637 static inline void load_ldt(u16 sel)
639 asm ("lldt %0" : : "rm"(sel));
643 static inline void get_idt(struct descriptor_table *table)
645 asm ("sidt %0" : "=m"(*table));
648 static inline void get_gdt(struct descriptor_table *table)
650 asm ("sgdt %0" : "=m"(*table));
653 static inline unsigned long read_tr_base(void)
656 asm ("str %0" : "=g"(tr));
657 return segment_base(tr);
661 static inline unsigned long read_msr(unsigned long msr)
670 static inline void fx_save(void *image)
672 asm ("fxsave (%0)":: "r" (image));
675 static inline void fx_restore(void *image)
677 asm ("fxrstor (%0)":: "r" (image));
680 static inline void fpu_init(void)
685 static inline u32 get_rdx_init_val(void)
687 return 0x600; /* P6 family */
690 #define ASM_VMX_VMCLEAR_RAX ".byte 0x66, 0x0f, 0xc7, 0x30"
691 #define ASM_VMX_VMLAUNCH ".byte 0x0f, 0x01, 0xc2"
692 #define ASM_VMX_VMRESUME ".byte 0x0f, 0x01, 0xc3"
693 #define ASM_VMX_VMPTRLD_RAX ".byte 0x0f, 0xc7, 0x30"
694 #define ASM_VMX_VMREAD_RDX_RAX ".byte 0x0f, 0x78, 0xd0"
695 #define ASM_VMX_VMWRITE_RAX_RDX ".byte 0x0f, 0x79, 0xd0"
696 #define ASM_VMX_VMWRITE_RSP_RDX ".byte 0x0f, 0x79, 0xd4"
697 #define ASM_VMX_VMXOFF ".byte 0x0f, 0x01, 0xc4"
698 #define ASM_VMX_VMXON_RAX ".byte 0xf3, 0x0f, 0xc7, 0x30"
700 #define MSR_IA32_TIME_STAMP_COUNTER 0x010
702 #define TSS_IOPB_BASE_OFFSET 0x66
703 #define TSS_BASE_SIZE 0x68
704 #define TSS_IOPB_SIZE (65536 / 8)
705 #define TSS_REDIRECTION_SIZE (256 / 8)
706 #define RMODE_TSS_SIZE (TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)