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 4
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 u64 ia32_misc_enable_msr;
331 struct kvm_mmu_memory_cache mmu_pte_chain_cache;
332 struct kvm_mmu_memory_cache mmu_rmap_desc_cache;
333 struct kvm_mmu_memory_cache mmu_page_cache;
334 struct kvm_mmu_memory_cache mmu_page_header_cache;
336 gfn_t last_pt_write_gfn;
337 int last_pt_write_count;
339 struct kvm_guest_debug guest_debug;
341 struct i387_fxsave_struct host_fx_image;
342 struct i387_fxsave_struct guest_fx_image;
344 int guest_fpu_loaded;
347 int mmio_read_completed;
350 unsigned char mmio_data[8];
351 gpa_t mmio_phys_addr;
352 gva_t mmio_fault_cr2;
353 struct kvm_pio_request pio;
359 struct kvm_stat stat;
364 struct kvm_save_segment {
369 } tr, es, ds, fs, gs;
371 int halt_request; /* real mode on Intel only */
374 struct kvm_cpuid_entry cpuid_entries[KVM_MAX_CPUID_ENTRIES];
377 struct kvm_mem_alias {
379 unsigned long npages;
383 struct kvm_memory_slot {
385 unsigned long npages;
387 struct page **phys_mem;
388 unsigned long *dirty_bitmap;
392 struct mutex lock; /* protects everything except vcpus */
394 struct kvm_mem_alias aliases[KVM_ALIAS_SLOTS];
396 struct kvm_memory_slot memslots[KVM_MEMORY_SLOTS];
398 * Hash table of struct kvm_mmu_page.
400 struct list_head active_mmu_pages;
401 int n_free_mmu_pages;
402 struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
403 struct kvm_vcpu *vcpus[KVM_MAX_VCPUS];
404 int memory_config_version;
406 unsigned long rmap_overflow;
407 struct list_head vm_list;
409 struct kvm_io_bus mmio_bus;
410 struct kvm_io_bus pio_bus;
413 struct descriptor_table {
416 } __attribute__((packed));
418 struct kvm_arch_ops {
419 int (*cpu_has_kvm_support)(void); /* __init */
420 int (*disabled_by_bios)(void); /* __init */
421 void (*hardware_enable)(void *dummy); /* __init */
422 void (*hardware_disable)(void *dummy);
423 void (*check_processor_compatibility)(void *rtn);
424 int (*hardware_setup)(void); /* __init */
425 void (*hardware_unsetup)(void); /* __exit */
427 /* Create, but do not attach this VCPU */
428 struct kvm_vcpu *(*vcpu_create)(struct kvm *kvm, unsigned id);
429 void (*vcpu_free)(struct kvm_vcpu *vcpu);
431 void (*vcpu_load)(struct kvm_vcpu *vcpu, int cpu);
432 void (*vcpu_put)(struct kvm_vcpu *vcpu);
433 void (*vcpu_decache)(struct kvm_vcpu *vcpu);
435 int (*set_guest_debug)(struct kvm_vcpu *vcpu,
436 struct kvm_debug_guest *dbg);
437 int (*get_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata);
438 int (*set_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 data);
439 u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
440 void (*get_segment)(struct kvm_vcpu *vcpu,
441 struct kvm_segment *var, int seg);
442 void (*set_segment)(struct kvm_vcpu *vcpu,
443 struct kvm_segment *var, int seg);
444 void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
445 void (*decache_cr4_guest_bits)(struct kvm_vcpu *vcpu);
446 void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
447 void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
448 void (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
449 void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
450 void (*get_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
451 void (*set_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
452 void (*get_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
453 void (*set_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
454 unsigned long (*get_dr)(struct kvm_vcpu *vcpu, int dr);
455 void (*set_dr)(struct kvm_vcpu *vcpu, int dr, unsigned long value,
457 void (*cache_regs)(struct kvm_vcpu *vcpu);
458 void (*decache_regs)(struct kvm_vcpu *vcpu);
459 unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
460 void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
462 void (*invlpg)(struct kvm_vcpu *vcpu, gva_t addr);
463 void (*tlb_flush)(struct kvm_vcpu *vcpu);
464 void (*inject_page_fault)(struct kvm_vcpu *vcpu,
465 unsigned long addr, u32 err_code);
467 void (*inject_gp)(struct kvm_vcpu *vcpu, unsigned err_code);
469 int (*run)(struct kvm_vcpu *vcpu, struct kvm_run *run);
470 void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
471 void (*patch_hypercall)(struct kvm_vcpu *vcpu,
472 unsigned char *hypercall_addr);
475 extern struct kvm_arch_ops *kvm_arch_ops;
477 /* The guest did something we don't support. */
478 #define pr_unimpl(vcpu, fmt, ...) \
480 if (printk_ratelimit()) \
481 printk(KERN_ERR "kvm: %i: cpu%i " fmt, \
482 current->tgid, (vcpu)->vcpu_id , ## __VA_ARGS__); \
485 #define kvm_printf(kvm, fmt ...) printk(KERN_DEBUG fmt)
486 #define vcpu_printf(vcpu, fmt...) kvm_printf(vcpu->kvm, fmt)
488 int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id);
489 void kvm_vcpu_uninit(struct kvm_vcpu *vcpu);
491 int kvm_init_arch(struct kvm_arch_ops *ops, unsigned int vcpu_size,
492 struct module *module);
493 void kvm_exit_arch(void);
495 int kvm_mmu_module_init(void);
496 void kvm_mmu_module_exit(void);
498 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
499 int kvm_mmu_create(struct kvm_vcpu *vcpu);
500 int kvm_mmu_setup(struct kvm_vcpu *vcpu);
502 int kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
503 void kvm_mmu_slot_remove_write_access(struct kvm *kvm, int slot);
504 void kvm_mmu_zap_all(struct kvm *kvm);
506 hpa_t gpa_to_hpa(struct kvm_vcpu *vcpu, gpa_t gpa);
507 #define HPA_MSB ((sizeof(hpa_t) * 8) - 1)
508 #define HPA_ERR_MASK ((hpa_t)1 << HPA_MSB)
509 static inline int is_error_hpa(hpa_t hpa) { return hpa >> HPA_MSB; }
510 hpa_t gva_to_hpa(struct kvm_vcpu *vcpu, gva_t gva);
511 struct page *gva_to_page(struct kvm_vcpu *vcpu, gva_t gva);
513 void kvm_emulator_want_group7_invlpg(void);
515 extern hpa_t bad_page_address;
517 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
518 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
519 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
521 enum emulation_result {
522 EMULATE_DONE, /* no further processing */
523 EMULATE_DO_MMIO, /* kvm_run filled with mmio request */
524 EMULATE_FAIL, /* can't emulate this instruction */
527 int emulate_instruction(struct kvm_vcpu *vcpu, struct kvm_run *run,
528 unsigned long cr2, u16 error_code);
529 void realmode_lgdt(struct kvm_vcpu *vcpu, u16 size, unsigned long address);
530 void realmode_lidt(struct kvm_vcpu *vcpu, u16 size, unsigned long address);
531 void realmode_lmsw(struct kvm_vcpu *vcpu, unsigned long msw,
532 unsigned long *rflags);
534 unsigned long realmode_get_cr(struct kvm_vcpu *vcpu, int cr);
535 void realmode_set_cr(struct kvm_vcpu *vcpu, int cr, unsigned long value,
536 unsigned long *rflags);
537 int kvm_get_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 *data);
538 int kvm_set_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 data);
540 struct x86_emulate_ctxt;
542 int kvm_emulate_pio (struct kvm_vcpu *vcpu, struct kvm_run *run, int in,
543 int size, unsigned port);
544 int kvm_emulate_pio_string(struct kvm_vcpu *vcpu, struct kvm_run *run, int in,
545 int size, unsigned long count, int down,
546 gva_t address, int rep, unsigned port);
547 void kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
548 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
549 int emulate_invlpg(struct kvm_vcpu *vcpu, gva_t address);
550 int emulate_clts(struct kvm_vcpu *vcpu);
551 int emulator_get_dr(struct x86_emulate_ctxt* ctxt, int dr,
552 unsigned long *dest);
553 int emulator_set_dr(struct x86_emulate_ctxt *ctxt, int dr,
554 unsigned long value);
556 void set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
557 void set_cr3(struct kvm_vcpu *vcpu, unsigned long cr0);
558 void set_cr4(struct kvm_vcpu *vcpu, unsigned long cr0);
559 void set_cr8(struct kvm_vcpu *vcpu, unsigned long cr0);
560 void lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
562 int kvm_get_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata);
563 int kvm_set_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 data);
565 void fx_init(struct kvm_vcpu *vcpu);
567 void kvm_resched(struct kvm_vcpu *vcpu);
568 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
569 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
570 void kvm_flush_remote_tlbs(struct kvm *kvm);
572 int emulator_read_std(unsigned long addr,
575 struct kvm_vcpu *vcpu);
576 int emulator_write_emulated(unsigned long addr,
579 struct kvm_vcpu *vcpu);
581 unsigned long segment_base(u16 selector);
583 void kvm_mmu_pte_write(struct kvm_vcpu *vcpu, gpa_t gpa,
584 const u8 *new, int bytes);
585 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva);
586 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
587 int kvm_mmu_load(struct kvm_vcpu *vcpu);
588 void kvm_mmu_unload(struct kvm_vcpu *vcpu);
590 int kvm_hypercall(struct kvm_vcpu *vcpu, struct kvm_run *run);
592 static inline int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t gva,
595 return vcpu->mmu.page_fault(vcpu, gva, error_code);
598 static inline void kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu)
600 if (unlikely(vcpu->kvm->n_free_mmu_pages < KVM_MIN_FREE_MMU_PAGES))
601 __kvm_mmu_free_some_pages(vcpu);
604 static inline int kvm_mmu_reload(struct kvm_vcpu *vcpu)
606 if (likely(vcpu->mmu.root_hpa != INVALID_PAGE))
609 return kvm_mmu_load(vcpu);
612 static inline int is_long_mode(struct kvm_vcpu *vcpu)
615 return vcpu->shadow_efer & EFER_LME;
621 static inline int is_pae(struct kvm_vcpu *vcpu)
623 return vcpu->cr4 & X86_CR4_PAE;
626 static inline int is_pse(struct kvm_vcpu *vcpu)
628 return vcpu->cr4 & X86_CR4_PSE;
631 static inline int is_paging(struct kvm_vcpu *vcpu)
633 return vcpu->cr0 & X86_CR0_PG;
636 static inline int memslot_id(struct kvm *kvm, struct kvm_memory_slot *slot)
638 return slot - kvm->memslots;
641 static inline struct kvm_mmu_page *page_header(hpa_t shadow_page)
643 struct page *page = pfn_to_page(shadow_page >> PAGE_SHIFT);
645 return (struct kvm_mmu_page *)page_private(page);
648 static inline u16 read_fs(void)
651 asm ("mov %%fs, %0" : "=g"(seg));
655 static inline u16 read_gs(void)
658 asm ("mov %%gs, %0" : "=g"(seg));
662 static inline u16 read_ldt(void)
665 asm ("sldt %0" : "=g"(ldt));
669 static inline void load_fs(u16 sel)
671 asm ("mov %0, %%fs" : : "rm"(sel));
674 static inline void load_gs(u16 sel)
676 asm ("mov %0, %%gs" : : "rm"(sel));
680 static inline void load_ldt(u16 sel)
682 asm ("lldt %0" : : "rm"(sel));
686 static inline void get_idt(struct descriptor_table *table)
688 asm ("sidt %0" : "=m"(*table));
691 static inline void get_gdt(struct descriptor_table *table)
693 asm ("sgdt %0" : "=m"(*table));
696 static inline unsigned long read_tr_base(void)
699 asm ("str %0" : "=g"(tr));
700 return segment_base(tr);
704 static inline unsigned long read_msr(unsigned long msr)
713 static inline void fx_save(struct i387_fxsave_struct *image)
715 asm ("fxsave (%0)":: "r" (image));
718 static inline void fx_restore(struct i387_fxsave_struct *image)
720 asm ("fxrstor (%0)":: "r" (image));
723 static inline void fpu_init(void)
728 static inline u32 get_rdx_init_val(void)
730 return 0x600; /* P6 family */
733 #define ASM_VMX_VMCLEAR_RAX ".byte 0x66, 0x0f, 0xc7, 0x30"
734 #define ASM_VMX_VMLAUNCH ".byte 0x0f, 0x01, 0xc2"
735 #define ASM_VMX_VMRESUME ".byte 0x0f, 0x01, 0xc3"
736 #define ASM_VMX_VMPTRLD_RAX ".byte 0x0f, 0xc7, 0x30"
737 #define ASM_VMX_VMREAD_RDX_RAX ".byte 0x0f, 0x78, 0xd0"
738 #define ASM_VMX_VMWRITE_RAX_RDX ".byte 0x0f, 0x79, 0xd0"
739 #define ASM_VMX_VMWRITE_RSP_RDX ".byte 0x0f, 0x79, 0xd4"
740 #define ASM_VMX_VMXOFF ".byte 0x0f, 0x01, 0xc4"
741 #define ASM_VMX_VMXON_RAX ".byte 0xf3, 0x0f, 0xc7, 0x30"
743 #define MSR_IA32_TIME_STAMP_COUNTER 0x010
745 #define TSS_IOPB_BASE_OFFSET 0x66
746 #define TSS_BASE_SIZE 0x68
747 #define TSS_IOPB_SIZE (65536 / 8)
748 #define TSS_REDIRECTION_SIZE (256 / 8)
749 #define RMODE_TSS_SIZE (TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)