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 <linux/preempt.h>
17 #include <asm/signal.h>
19 #include <linux/kvm.h>
20 #include <linux/kvm_para.h>
22 #define CR3_PAE_RESERVED_BITS ((X86_CR3_PWT | X86_CR3_PCD) - 1)
23 #define CR3_NONPAE_RESERVED_BITS ((PAGE_SIZE-1) & ~(X86_CR3_PWT | X86_CR3_PCD))
24 #define CR3_L_MODE_RESERVED_BITS (CR3_NONPAE_RESERVED_BITS|0xFFFFFF0000000000ULL)
26 #define KVM_GUEST_CR0_MASK \
27 (X86_CR0_PG | X86_CR0_PE | X86_CR0_WP | X86_CR0_NE \
28 | X86_CR0_NW | X86_CR0_CD)
29 #define KVM_VM_CR0_ALWAYS_ON \
30 (X86_CR0_PG | X86_CR0_PE | X86_CR0_WP | X86_CR0_NE | X86_CR0_TS \
32 #define KVM_GUEST_CR4_MASK \
33 (X86_CR4_VME | X86_CR4_PSE | X86_CR4_PAE | X86_CR4_PGE | X86_CR4_VMXE)
34 #define KVM_PMODE_VM_CR4_ALWAYS_ON (X86_CR4_PAE | X86_CR4_VMXE)
35 #define KVM_RMODE_VM_CR4_ALWAYS_ON (X86_CR4_VME | X86_CR4_PAE | X86_CR4_VMXE)
37 #define INVALID_PAGE (~(hpa_t)0)
38 #define UNMAPPED_GVA (~(gpa_t)0)
40 #define KVM_MAX_VCPUS 4
41 #define KVM_ALIAS_SLOTS 4
42 #define KVM_MEMORY_SLOTS 8
43 #define KVM_NUM_MMU_PAGES 1024
44 #define KVM_MIN_FREE_MMU_PAGES 5
45 #define KVM_REFILL_PAGES 25
46 #define KVM_MAX_CPUID_ENTRIES 40
57 #define SELECTOR_TI_MASK (1 << 2)
58 #define SELECTOR_RPL_MASK 0x03
62 #define KVM_PIO_PAGE_OFFSET 1
65 * vcpu->requests bit members
67 #define KVM_TLB_FLUSH 0
72 * gva - guest virtual address
73 * gpa - guest physical address
74 * gfn - guest frame number
75 * hva - host virtual address
76 * hpa - host physical address
77 * hfn - host frame number
80 typedef unsigned long gva_t;
82 typedef unsigned long gfn_t;
84 typedef unsigned long hva_t;
86 typedef unsigned long hfn_t;
88 #define NR_PTE_CHAIN_ENTRIES 5
90 struct kvm_pte_chain {
91 u64 *parent_ptes[NR_PTE_CHAIN_ENTRIES];
92 struct hlist_node link;
96 * kvm_mmu_page_role, below, is defined as:
98 * bits 0:3 - total guest paging levels (2-4, or zero for real mode)
99 * bits 4:7 - page table level for this shadow (1-4)
100 * bits 8:9 - page table quadrant for 2-level guests
101 * bit 16 - "metaphysical" - gfn is not a real page (huge page/real mode)
102 * bits 17:19 - "access" - the user, writable, and nx bits of a huge page pde
104 union kvm_mmu_page_role {
107 unsigned glevels : 4;
109 unsigned quadrant : 2;
110 unsigned pad_for_nice_hex_output : 6;
111 unsigned metaphysical : 1;
112 unsigned hugepage_access : 3;
116 struct kvm_mmu_page {
117 struct list_head link;
118 struct hlist_node hash_link;
121 * The following two entries are used to key the shadow page in the
125 union kvm_mmu_page_role role;
128 unsigned long slot_bitmap; /* One bit set per slot which has memory
129 * in this shadow page.
131 int multimapped; /* More than one parent_pte? */
132 int root_count; /* Currently serving as active root */
134 u64 *parent_pte; /* !multimapped */
135 struct hlist_head parent_ptes; /* multimapped, kvm_pte_chain */
140 extern struct kmem_cache *kvm_vcpu_cache;
143 * x86 supports 3 paging modes (4-level 64-bit, 3-level 64-bit, and 2-level
144 * 32-bit). The kvm_mmu structure abstracts the details of the current mmu
148 void (*new_cr3)(struct kvm_vcpu *vcpu);
149 int (*page_fault)(struct kvm_vcpu *vcpu, gva_t gva, u32 err);
150 void (*free)(struct kvm_vcpu *vcpu);
151 gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t gva);
154 int shadow_root_level;
159 #define KVM_NR_MEM_OBJS 20
161 struct kvm_mmu_memory_cache {
163 void *objects[KVM_NR_MEM_OBJS];
167 * We don't want allocation failures within the mmu code, so we preallocate
168 * enough memory for a single page fault in a cache.
170 struct kvm_guest_debug {
209 struct kvm_pio_request {
212 struct page *guest_pages[2];
213 unsigned guest_page_offset;
232 u32 irq_window_exits;
234 u32 request_irq_exits;
240 struct kvm_io_device {
241 void (*read)(struct kvm_io_device *this,
245 void (*write)(struct kvm_io_device *this,
249 int (*in_range)(struct kvm_io_device *this, gpa_t addr);
250 void (*destructor)(struct kvm_io_device *this);
255 static inline void kvm_iodevice_read(struct kvm_io_device *dev,
260 dev->read(dev, addr, len, val);
263 static inline void kvm_iodevice_write(struct kvm_io_device *dev,
268 dev->write(dev, addr, len, val);
271 static inline int kvm_iodevice_inrange(struct kvm_io_device *dev, gpa_t addr)
273 return dev->in_range(dev, addr);
276 static inline void kvm_iodevice_destructor(struct kvm_io_device *dev)
279 dev->destructor(dev);
283 * It would be nice to use something smarter than a linear search, TBD...
284 * Thankfully we dont expect many devices to register (famous last words :),
285 * so until then it will suffice. At least its abstracted so we can change
290 #define NR_IOBUS_DEVS 6
291 struct kvm_io_device *devs[NR_IOBUS_DEVS];
294 void kvm_io_bus_init(struct kvm_io_bus *bus);
295 void kvm_io_bus_destroy(struct kvm_io_bus *bus);
296 struct kvm_io_device *kvm_io_bus_find_dev(struct kvm_io_bus *bus, gpa_t addr);
297 void kvm_io_bus_register_dev(struct kvm_io_bus *bus,
298 struct kvm_io_device *dev);
302 struct preempt_notifier preempt_notifier;
308 int interrupt_window_open;
310 unsigned long requests;
311 unsigned long irq_summary; /* bit vector: 1 per word in irq_pending */
312 DECLARE_BITMAP(irq_pending, KVM_NR_INTERRUPTS);
313 unsigned long regs[NR_VCPU_REGS]; /* for rsp: vcpu_load_rsp_rip() */
314 unsigned long rip; /* needs vcpu_load_rsp_rip() */
319 gpa_t para_state_gpa;
320 struct page *para_state_page;
324 u64 pdptrs[4]; /* pae */
327 struct kvm_lapic *apic; /* kernel irqchip context */
328 u64 ia32_misc_enable_msr;
332 struct kvm_mmu_memory_cache mmu_pte_chain_cache;
333 struct kvm_mmu_memory_cache mmu_rmap_desc_cache;
334 struct kvm_mmu_memory_cache mmu_page_cache;
335 struct kvm_mmu_memory_cache mmu_page_header_cache;
337 gfn_t last_pt_write_gfn;
338 int last_pt_write_count;
340 struct kvm_guest_debug guest_debug;
342 struct i387_fxsave_struct host_fx_image;
343 struct i387_fxsave_struct guest_fx_image;
345 int guest_fpu_loaded;
348 int mmio_read_completed;
351 unsigned char mmio_data[8];
352 gpa_t mmio_phys_addr;
353 gva_t mmio_fault_cr2;
354 struct kvm_pio_request pio;
360 struct kvm_stat stat;
365 struct kvm_save_segment {
370 } tr, es, ds, fs, gs;
372 int halt_request; /* real mode on Intel only */
375 struct kvm_cpuid_entry cpuid_entries[KVM_MAX_CPUID_ENTRIES];
378 struct kvm_mem_alias {
380 unsigned long npages;
384 struct kvm_memory_slot {
386 unsigned long npages;
388 struct page **phys_mem;
389 unsigned long *dirty_bitmap;
393 struct mutex lock; /* protects everything except vcpus */
395 struct kvm_mem_alias aliases[KVM_ALIAS_SLOTS];
397 struct kvm_memory_slot memslots[KVM_MEMORY_SLOTS];
399 * Hash table of struct kvm_mmu_page.
401 struct list_head active_mmu_pages;
402 int n_free_mmu_pages;
403 struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
404 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
405 int memory_config_version;
407 unsigned long rmap_overflow;
408 struct list_head vm_list;
410 struct kvm_io_bus mmio_bus;
411 struct kvm_io_bus pio_bus;
412 struct kvm_pic *vpic;
415 static inline struct kvm_pic *pic_irqchip(struct kvm *kvm)
420 static inline int irqchip_in_kernel(struct kvm *kvm)
422 return pic_irqchip(kvm) != 0;
425 struct descriptor_table {
428 } __attribute__((packed));
430 struct kvm_arch_ops {
431 int (*cpu_has_kvm_support)(void); /* __init */
432 int (*disabled_by_bios)(void); /* __init */
433 void (*hardware_enable)(void *dummy); /* __init */
434 void (*hardware_disable)(void *dummy);
435 void (*check_processor_compatibility)(void *rtn);
436 int (*hardware_setup)(void); /* __init */
437 void (*hardware_unsetup)(void); /* __exit */
439 /* Create, but do not attach this VCPU */
440 struct kvm_vcpu *(*vcpu_create)(struct kvm *kvm, unsigned id);
441 void (*vcpu_free)(struct kvm_vcpu *vcpu);
443 void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
444 void (*vcpu_put)(struct kvm_vcpu *vcpu);
445 void (*vcpu_decache)(struct kvm_vcpu *vcpu);
447 int (*set_guest_debug)(struct kvm_vcpu *vcpu,
448 struct kvm_debug_guest *dbg);
449 int (*get_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata);
450 int (*set_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 data);
451 u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
452 void (*get_segment)(struct kvm_vcpu *vcpu,
453 struct kvm_segment *var, int seg);
454 void (*set_segment)(struct kvm_vcpu *vcpu,
455 struct kvm_segment *var, int seg);
456 void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
457 void (*decache_cr4_guest_bits)(struct kvm_vcpu *vcpu);
458 void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
459 void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
460 void (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
461 void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
462 void (*get_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
463 void (*set_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
464 void (*get_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
465 void (*set_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
466 unsigned long (*get_dr)(struct kvm_vcpu *vcpu, int dr);
467 void (*set_dr)(struct kvm_vcpu *vcpu, int dr, unsigned long value,
469 void (*cache_regs)(struct kvm_vcpu *vcpu);
470 void (*decache_regs)(struct kvm_vcpu *vcpu);
471 unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
472 void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
474 void (*invlpg)(struct kvm_vcpu *vcpu, gva_t addr);
475 void (*tlb_flush)(struct kvm_vcpu *vcpu);
476 void (*inject_page_fault)(struct kvm_vcpu *vcpu,
477 unsigned long addr, u32 err_code);
479 void (*inject_gp)(struct kvm_vcpu *vcpu, unsigned err_code);
481 int (*run)(struct kvm_vcpu *vcpu, struct kvm_run *run);
482 void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
483 void (*patch_hypercall)(struct kvm_vcpu *vcpu,
484 unsigned char *hypercall_addr);
487 extern struct kvm_arch_ops *kvm_arch_ops;
489 /* The guest did something we don't support. */
490 #define pr_unimpl(vcpu, fmt, ...) \
492 if (printk_ratelimit()) \
493 printk(KERN_ERR "kvm: %i: cpu%i " fmt, \
494 current->tgid, (vcpu)->vcpu_id , ## __VA_ARGS__); \
497 #define kvm_printf(kvm, fmt ...) printk(KERN_DEBUG fmt)
498 #define vcpu_printf(vcpu, fmt...) kvm_printf(vcpu->kvm, fmt)
500 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
501 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
503 int kvm_init_arch(struct kvm_arch_ops *ops, unsigned int vcpu_size,
504 struct module *module);
505 void kvm_exit_arch(void);
507 int kvm_mmu_module_init(void);
508 void kvm_mmu_module_exit(void);
510 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
511 int kvm_mmu_create(struct kvm_vcpu *vcpu);
512 int kvm_mmu_setup(struct kvm_vcpu *vcpu);
514 int kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
515 void kvm_mmu_slot_remove_write_access(struct kvm *kvm, int slot);
516 void kvm_mmu_zap_all(struct kvm *kvm);
518 hpa_t gpa_to_hpa(struct kvm_vcpu *vcpu, gpa_t gpa);
519 #define HPA_MSB ((sizeof(hpa_t) * 8) - 1)
520 #define HPA_ERR_MASK ((hpa_t)1 << HPA_MSB)
521 static inline int is_error_hpa(hpa_t hpa) { return hpa >> HPA_MSB; }
522 hpa_t gva_to_hpa(struct kvm_vcpu *vcpu, gva_t gva);
523 struct page *gva_to_page(struct kvm_vcpu *vcpu, gva_t gva);
525 void kvm_emulator_want_group7_invlpg(void);
527 extern hpa_t bad_page_address;
529 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
530 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
531 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
533 enum emulation_result {
534 EMULATE_DONE, /* no further processing */
535 EMULATE_DO_MMIO, /* kvm_run filled with mmio request */
536 EMULATE_FAIL, /* can't emulate this instruction */
539 int emulate_instruction(struct kvm_vcpu *vcpu, struct kvm_run *run,
540 unsigned long cr2, u16 error_code);
541 void realmode_lgdt(struct kvm_vcpu *vcpu, u16 size, unsigned long address);
542 void realmode_lidt(struct kvm_vcpu *vcpu, u16 size, unsigned long address);
543 void realmode_lmsw(struct kvm_vcpu *vcpu, unsigned long msw,
544 unsigned long *rflags);
546 unsigned long realmode_get_cr(struct kvm_vcpu *vcpu, int cr);
547 void realmode_set_cr(struct kvm_vcpu *vcpu, int cr, unsigned long value,
548 unsigned long *rflags);
549 int kvm_get_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 *data);
550 int kvm_set_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 data);
552 struct x86_emulate_ctxt;
554 int kvm_emulate_pio (struct kvm_vcpu *vcpu, struct kvm_run *run, int in,
555 int size, unsigned port);
556 int kvm_emulate_pio_string(struct kvm_vcpu *vcpu, struct kvm_run *run, int in,
557 int size, unsigned long count, int down,
558 gva_t address, int rep, unsigned port);
559 void kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
560 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
561 int emulate_invlpg(struct kvm_vcpu *vcpu, gva_t address);
562 int emulate_clts(struct kvm_vcpu *vcpu);
563 int emulator_get_dr(struct x86_emulate_ctxt* ctxt, int dr,
564 unsigned long *dest);
565 int emulator_set_dr(struct x86_emulate_ctxt *ctxt, int dr,
566 unsigned long value);
568 void set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
569 void set_cr3(struct kvm_vcpu *vcpu, unsigned long cr0);
570 void set_cr4(struct kvm_vcpu *vcpu, unsigned long cr0);
571 void set_cr8(struct kvm_vcpu *vcpu, unsigned long cr0);
572 unsigned long get_cr8(struct kvm_vcpu *vcpu);
573 void lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
575 int kvm_get_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata);
576 int kvm_set_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 data);
578 void fx_init(struct kvm_vcpu *vcpu);
580 void kvm_resched(struct kvm_vcpu *vcpu);
581 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
582 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
583 void kvm_flush_remote_tlbs(struct kvm *kvm);
585 int emulator_read_std(unsigned long addr,
588 struct kvm_vcpu *vcpu);
589 int emulator_write_emulated(unsigned long addr,
592 struct kvm_vcpu *vcpu);
594 unsigned long segment_base(u16 selector);
596 void kvm_mmu_pte_write(struct kvm_vcpu *vcpu, gpa_t gpa,
597 const u8 *new, int bytes);
598 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva);
599 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
600 int kvm_mmu_load(struct kvm_vcpu *vcpu);
601 void kvm_mmu_unload(struct kvm_vcpu *vcpu);
603 int kvm_hypercall(struct kvm_vcpu *vcpu, struct kvm_run *run);
605 static inline int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t gva,
608 return vcpu->mmu.page_fault(vcpu, gva, error_code);
611 static inline void kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu)
613 if (unlikely(vcpu->kvm->n_free_mmu_pages < KVM_MIN_FREE_MMU_PAGES))
614 __kvm_mmu_free_some_pages(vcpu);
617 static inline int kvm_mmu_reload(struct kvm_vcpu *vcpu)
619 if (likely(vcpu->mmu.root_hpa != INVALID_PAGE))
622 return kvm_mmu_load(vcpu);
625 static inline int is_long_mode(struct kvm_vcpu *vcpu)
628 return vcpu->shadow_efer & EFER_LME;
634 static inline int is_pae(struct kvm_vcpu *vcpu)
636 return vcpu->cr4 & X86_CR4_PAE;
639 static inline int is_pse(struct kvm_vcpu *vcpu)
641 return vcpu->cr4 & X86_CR4_PSE;
644 static inline int is_paging(struct kvm_vcpu *vcpu)
646 return vcpu->cr0 & X86_CR0_PG;
649 static inline int memslot_id(struct kvm *kvm, struct kvm_memory_slot *slot)
651 return slot - kvm->memslots;
654 static inline struct kvm_mmu_page *page_header(hpa_t shadow_page)
656 struct page *page = pfn_to_page(shadow_page >> PAGE_SHIFT);
658 return (struct kvm_mmu_page *)page_private(page);
661 static inline u16 read_fs(void)
664 asm ("mov %%fs, %0" : "=g"(seg));
668 static inline u16 read_gs(void)
671 asm ("mov %%gs, %0" : "=g"(seg));
675 static inline u16 read_ldt(void)
678 asm ("sldt %0" : "=g"(ldt));
682 static inline void load_fs(u16 sel)
684 asm ("mov %0, %%fs" : : "rm"(sel));
687 static inline void load_gs(u16 sel)
689 asm ("mov %0, %%gs" : : "rm"(sel));
693 static inline void load_ldt(u16 sel)
695 asm ("lldt %0" : : "rm"(sel));
699 static inline void get_idt(struct descriptor_table *table)
701 asm ("sidt %0" : "=m"(*table));
704 static inline void get_gdt(struct descriptor_table *table)
706 asm ("sgdt %0" : "=m"(*table));
709 static inline unsigned long read_tr_base(void)
712 asm ("str %0" : "=g"(tr));
713 return segment_base(tr);
717 static inline unsigned long read_msr(unsigned long msr)
726 static inline void fx_save(struct i387_fxsave_struct *image)
728 asm ("fxsave (%0)":: "r" (image));
731 static inline void fx_restore(struct i387_fxsave_struct *image)
733 asm ("fxrstor (%0)":: "r" (image));
736 static inline void fpu_init(void)
741 static inline u32 get_rdx_init_val(void)
743 return 0x600; /* P6 family */
746 #define ASM_VMX_VMCLEAR_RAX ".byte 0x66, 0x0f, 0xc7, 0x30"
747 #define ASM_VMX_VMLAUNCH ".byte 0x0f, 0x01, 0xc2"
748 #define ASM_VMX_VMRESUME ".byte 0x0f, 0x01, 0xc3"
749 #define ASM_VMX_VMPTRLD_RAX ".byte 0x0f, 0xc7, 0x30"
750 #define ASM_VMX_VMREAD_RDX_RAX ".byte 0x0f, 0x78, 0xd0"
751 #define ASM_VMX_VMWRITE_RAX_RDX ".byte 0x0f, 0x79, 0xd0"
752 #define ASM_VMX_VMWRITE_RSP_RDX ".byte 0x0f, 0x79, 0xd4"
753 #define ASM_VMX_VMXOFF ".byte 0x0f, 0x01, 0xc4"
754 #define ASM_VMX_VMXON_RAX ".byte 0xf3, 0x0f, 0xc7, 0x30"
756 #define MSR_IA32_TIME_STAMP_COUNTER 0x010
758 #define TSS_IOPB_BASE_OFFSET 0x66
759 #define TSS_BASE_SIZE 0x68
760 #define TSS_IOPB_SIZE (65536 / 8)
761 #define TSS_REDIRECTION_SIZE (256 / 8)
762 #define RMODE_TSS_SIZE (TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)