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 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 CR4_VME_MASK (1ULL << 0)
27 #define CR4_PSE_MASK (1ULL << 4)
28 #define CR4_PAE_MASK (1ULL << 5)
29 #define CR4_PGE_MASK (1ULL << 7)
30 #define CR4_VMXE_MASK (1ULL << 13)
32 #define KVM_GUEST_CR0_MASK \
33 (X86_CR0_PG | X86_CR0_PE | X86_CR0_WP | X86_CR0_NE \
34 | X86_CR0_NW | X86_CR0_CD)
35 #define KVM_VM_CR0_ALWAYS_ON \
36 (X86_CR0_PG | X86_CR0_PE | X86_CR0_WP | X86_CR0_NE | X86_CR0_TS \
38 #define KVM_GUEST_CR4_MASK \
39 (CR4_PSE_MASK | CR4_PAE_MASK | CR4_PGE_MASK | CR4_VMXE_MASK | CR4_VME_MASK)
40 #define KVM_PMODE_VM_CR4_ALWAYS_ON (CR4_VMXE_MASK | CR4_PAE_MASK)
41 #define KVM_RMODE_VM_CR4_ALWAYS_ON (CR4_VMXE_MASK | CR4_PAE_MASK | CR4_VME_MASK)
43 #define INVALID_PAGE (~(hpa_t)0)
44 #define UNMAPPED_GVA (~(gpa_t)0)
46 #define KVM_MAX_VCPUS 4
47 #define KVM_ALIAS_SLOTS 4
48 #define KVM_MEMORY_SLOTS 4
49 #define KVM_NUM_MMU_PAGES 1024
50 #define KVM_MIN_FREE_MMU_PAGES 5
51 #define KVM_REFILL_PAGES 25
52 #define KVM_MAX_CPUID_ENTRIES 40
54 #define FX_IMAGE_SIZE 512
55 #define FX_IMAGE_ALIGN 16
56 #define FX_BUF_SIZE (2 * FX_IMAGE_SIZE + FX_IMAGE_ALIGN)
67 #define SELECTOR_TI_MASK (1 << 2)
68 #define SELECTOR_RPL_MASK 0x03
72 #define KVM_PIO_PAGE_OFFSET 1
75 * vcpu->requests bit members
77 #define KVM_TLB_FLUSH 0
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)
112 * bits 17:19 - "access" - the user, writable, and nx bits of a huge page pde
114 union kvm_mmu_page_role {
117 unsigned glevels : 4;
119 unsigned quadrant : 2;
120 unsigned pad_for_nice_hex_output : 6;
121 unsigned metaphysical : 1;
122 unsigned hugepage_access : 3;
126 struct kvm_mmu_page {
127 struct list_head link;
128 struct hlist_node hash_link;
131 * The following two entries are used to key the shadow page in the
135 union kvm_mmu_page_role role;
138 unsigned long slot_bitmap; /* One bit set per slot which has memory
139 * in this shadow page.
141 int multimapped; /* More than one parent_pte? */
142 int root_count; /* Currently serving as active root */
144 u64 *parent_pte; /* !multimapped */
145 struct hlist_head parent_ptes; /* multimapped, kvm_pte_chain */
155 #define vmx_msr_entry kvm_msr_entry
160 * x86 supports 3 paging modes (4-level 64-bit, 3-level 64-bit, and 2-level
161 * 32-bit). The kvm_mmu structure abstracts the details of the current mmu
165 void (*new_cr3)(struct kvm_vcpu *vcpu);
166 int (*page_fault)(struct kvm_vcpu *vcpu, gva_t gva, u32 err);
167 void (*free)(struct kvm_vcpu *vcpu);
168 gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t gva);
171 int shadow_root_level;
176 #define KVM_NR_MEM_OBJS 20
178 struct kvm_mmu_memory_cache {
180 void *objects[KVM_NR_MEM_OBJS];
184 * We don't want allocation failures within the mmu code, so we preallocate
185 * enough memory for a single page fault in a cache.
187 struct kvm_guest_debug {
226 struct kvm_pio_request {
229 struct page *guest_pages[2];
230 unsigned guest_page_offset;
249 u32 irq_window_exits;
251 u32 request_irq_exits;
257 struct kvm_io_device {
258 void (*read)(struct kvm_io_device *this,
262 void (*write)(struct kvm_io_device *this,
266 int (*in_range)(struct kvm_io_device *this, gpa_t addr);
267 void (*destructor)(struct kvm_io_device *this);
272 static inline void kvm_iodevice_read(struct kvm_io_device *dev,
277 dev->read(dev, addr, len, val);
280 static inline void kvm_iodevice_write(struct kvm_io_device *dev,
285 dev->write(dev, addr, len, val);
288 static inline int kvm_iodevice_inrange(struct kvm_io_device *dev, gpa_t addr)
290 return dev->in_range(dev, addr);
293 static inline void kvm_iodevice_destructor(struct kvm_io_device *dev)
296 dev->destructor(dev);
300 * It would be nice to use something smarter than a linear search, TBD...
301 * Thankfully we dont expect many devices to register (famous last words :),
302 * so until then it will suffice. At least its abstracted so we can change
307 #define NR_IOBUS_DEVS 6
308 struct kvm_io_device *devs[NR_IOBUS_DEVS];
311 void kvm_io_bus_init(struct kvm_io_bus *bus);
312 void kvm_io_bus_destroy(struct kvm_io_bus *bus);
313 struct kvm_io_device *kvm_io_bus_find_dev(struct kvm_io_bus *bus, gpa_t addr);
314 void kvm_io_bus_register_dev(struct kvm_io_bus *bus,
315 struct kvm_io_device *dev);
322 struct vcpu_svm *svm;
329 int interrupt_window_open;
331 unsigned long requests;
332 unsigned long irq_summary; /* bit vector: 1 per word in irq_pending */
333 #define NR_IRQ_WORDS KVM_IRQ_BITMAP_SIZE(unsigned long)
334 unsigned long irq_pending[NR_IRQ_WORDS];
335 unsigned long regs[NR_VCPU_REGS]; /* for rsp: vcpu_load_rsp_rip() */
336 unsigned long rip; /* needs vcpu_load_rsp_rip() */
341 gpa_t para_state_gpa;
342 struct page *para_state_page;
346 u64 pdptrs[4]; /* pae */
349 u64 ia32_misc_enable_msr;
354 int msr_offset_kernel_gs_base;
356 struct vmx_msr_entry *guest_msrs;
357 struct vmx_msr_entry *host_msrs;
361 struct kvm_mmu_memory_cache mmu_pte_chain_cache;
362 struct kvm_mmu_memory_cache mmu_rmap_desc_cache;
363 struct kvm_mmu_memory_cache mmu_page_cache;
364 struct kvm_mmu_memory_cache mmu_page_header_cache;
366 gfn_t last_pt_write_gfn;
367 int last_pt_write_count;
369 struct kvm_guest_debug guest_debug;
371 char fx_buf[FX_BUF_SIZE];
373 char *guest_fx_image;
375 int guest_fpu_loaded;
376 struct vmx_host_state {
378 u16 fs_sel, gs_sel, ldt_sel;
379 int fs_gs_ldt_reload_needed;
383 int mmio_read_completed;
386 unsigned char mmio_data[8];
387 gpa_t mmio_phys_addr;
388 gva_t mmio_fault_cr2;
389 struct kvm_pio_request pio;
395 struct kvm_stat stat;
400 struct kvm_save_segment {
405 } tr, es, ds, fs, gs;
407 int halt_request; /* real mode on Intel only */
410 struct kvm_cpuid_entry cpuid_entries[KVM_MAX_CPUID_ENTRIES];
413 struct kvm_mem_alias {
415 unsigned long npages;
419 struct kvm_memory_slot {
421 unsigned long npages;
423 struct page **phys_mem;
424 unsigned long *dirty_bitmap;
428 spinlock_t lock; /* protects everything except vcpus */
430 struct kvm_mem_alias aliases[KVM_ALIAS_SLOTS];
432 struct kvm_memory_slot memslots[KVM_MEMORY_SLOTS];
434 * Hash table of struct kvm_mmu_page.
436 struct list_head active_mmu_pages;
437 int n_free_mmu_pages;
438 struct hlist_head mmu_page_hash[KVM_NUM_MMU_PAGES];
440 struct kvm_vcpu vcpus[KVM_MAX_VCPUS];
441 int memory_config_version;
443 unsigned long rmap_overflow;
444 struct list_head vm_list;
446 struct kvm_io_bus mmio_bus;
447 struct kvm_io_bus pio_bus;
450 struct descriptor_table {
453 } __attribute__((packed));
455 struct kvm_arch_ops {
456 int (*cpu_has_kvm_support)(void); /* __init */
457 int (*disabled_by_bios)(void); /* __init */
458 void (*hardware_enable)(void *dummy); /* __init */
459 void (*hardware_disable)(void *dummy);
460 int (*hardware_setup)(void); /* __init */
461 void (*hardware_unsetup)(void); /* __exit */
463 int (*vcpu_create)(struct kvm_vcpu *vcpu);
464 void (*vcpu_free)(struct kvm_vcpu *vcpu);
466 void (*vcpu_load)(struct kvm_vcpu *vcpu);
467 void (*vcpu_put)(struct kvm_vcpu *vcpu);
468 void (*vcpu_decache)(struct kvm_vcpu *vcpu);
470 int (*set_guest_debug)(struct kvm_vcpu *vcpu,
471 struct kvm_debug_guest *dbg);
472 int (*get_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata);
473 int (*set_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 data);
474 u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
475 void (*get_segment)(struct kvm_vcpu *vcpu,
476 struct kvm_segment *var, int seg);
477 void (*set_segment)(struct kvm_vcpu *vcpu,
478 struct kvm_segment *var, int seg);
479 void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
480 void (*decache_cr4_guest_bits)(struct kvm_vcpu *vcpu);
481 void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
482 void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
483 void (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
484 void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
485 void (*get_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
486 void (*set_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
487 void (*get_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
488 void (*set_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
489 unsigned long (*get_dr)(struct kvm_vcpu *vcpu, int dr);
490 void (*set_dr)(struct kvm_vcpu *vcpu, int dr, unsigned long value,
492 void (*cache_regs)(struct kvm_vcpu *vcpu);
493 void (*decache_regs)(struct kvm_vcpu *vcpu);
494 unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
495 void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
497 void (*invlpg)(struct kvm_vcpu *vcpu, gva_t addr);
498 void (*tlb_flush)(struct kvm_vcpu *vcpu);
499 void (*inject_page_fault)(struct kvm_vcpu *vcpu,
500 unsigned long addr, u32 err_code);
502 void (*inject_gp)(struct kvm_vcpu *vcpu, unsigned err_code);
504 int (*run)(struct kvm_vcpu *vcpu, struct kvm_run *run);
505 int (*vcpu_setup)(struct kvm_vcpu *vcpu);
506 void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
507 void (*patch_hypercall)(struct kvm_vcpu *vcpu,
508 unsigned char *hypercall_addr);
511 extern struct kvm_arch_ops *kvm_arch_ops;
513 #define kvm_printf(kvm, fmt ...) printk(KERN_DEBUG fmt)
514 #define vcpu_printf(vcpu, fmt...) kvm_printf(vcpu->kvm, fmt)
516 int kvm_init_arch(struct kvm_arch_ops *ops, struct module *module);
517 void kvm_exit_arch(void);
519 int kvm_mmu_module_init(void);
520 void kvm_mmu_module_exit(void);
522 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
523 int kvm_mmu_create(struct kvm_vcpu *vcpu);
524 int kvm_mmu_setup(struct kvm_vcpu *vcpu);
526 int kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
527 void kvm_mmu_slot_remove_write_access(struct kvm *kvm, int slot);
528 void kvm_mmu_zap_all(struct kvm *kvm);
530 hpa_t gpa_to_hpa(struct kvm_vcpu *vcpu, gpa_t gpa);
531 #define HPA_MSB ((sizeof(hpa_t) * 8) - 1)
532 #define HPA_ERR_MASK ((hpa_t)1 << HPA_MSB)
533 static inline int is_error_hpa(hpa_t hpa) { return hpa >> HPA_MSB; }
534 hpa_t gva_to_hpa(struct kvm_vcpu *vcpu, gva_t gva);
535 struct page *gva_to_page(struct kvm_vcpu *vcpu, gva_t gva);
537 void kvm_emulator_want_group7_invlpg(void);
539 extern hpa_t bad_page_address;
541 struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn);
542 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
543 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
545 enum emulation_result {
546 EMULATE_DONE, /* no further processing */
547 EMULATE_DO_MMIO, /* kvm_run filled with mmio request */
548 EMULATE_FAIL, /* can't emulate this instruction */
551 int emulate_instruction(struct kvm_vcpu *vcpu, struct kvm_run *run,
552 unsigned long cr2, u16 error_code);
553 void realmode_lgdt(struct kvm_vcpu *vcpu, u16 size, unsigned long address);
554 void realmode_lidt(struct kvm_vcpu *vcpu, u16 size, unsigned long address);
555 void realmode_lmsw(struct kvm_vcpu *vcpu, unsigned long msw,
556 unsigned long *rflags);
558 unsigned long realmode_get_cr(struct kvm_vcpu *vcpu, int cr);
559 void realmode_set_cr(struct kvm_vcpu *vcpu, int cr, unsigned long value,
560 unsigned long *rflags);
561 int kvm_get_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 *data);
562 int kvm_set_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 data);
564 struct x86_emulate_ctxt;
566 int kvm_setup_pio(struct kvm_vcpu *vcpu, struct kvm_run *run, int in,
567 int size, unsigned long count, int string, int down,
568 gva_t address, int rep, unsigned port);
569 void kvm_emulate_cpuid(struct kvm_vcpu *vcpu);
570 int kvm_emulate_halt(struct kvm_vcpu *vcpu);
571 int emulate_invlpg(struct kvm_vcpu *vcpu, gva_t address);
572 int emulate_clts(struct kvm_vcpu *vcpu);
573 int emulator_get_dr(struct x86_emulate_ctxt* ctxt, int dr,
574 unsigned long *dest);
575 int emulator_set_dr(struct x86_emulate_ctxt *ctxt, int dr,
576 unsigned long value);
578 void set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
579 void set_cr3(struct kvm_vcpu *vcpu, unsigned long cr0);
580 void set_cr4(struct kvm_vcpu *vcpu, unsigned long cr0);
581 void set_cr8(struct kvm_vcpu *vcpu, unsigned long cr0);
582 void lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
584 int kvm_get_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata);
585 int kvm_set_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 data);
587 void fx_init(struct kvm_vcpu *vcpu);
589 void load_msrs(struct vmx_msr_entry *e, int n);
590 void save_msrs(struct vmx_msr_entry *e, int n);
591 void kvm_resched(struct kvm_vcpu *vcpu);
592 void kvm_load_guest_fpu(struct kvm_vcpu *vcpu);
593 void kvm_put_guest_fpu(struct kvm_vcpu *vcpu);
594 void kvm_flush_remote_tlbs(struct kvm *kvm);
596 int kvm_read_guest(struct kvm_vcpu *vcpu,
601 int kvm_write_guest(struct kvm_vcpu *vcpu,
606 unsigned long segment_base(u16 selector);
608 void kvm_mmu_pte_write(struct kvm_vcpu *vcpu, gpa_t gpa,
609 const u8 *old, const u8 *new, int bytes);
610 int kvm_mmu_unprotect_page_virt(struct kvm_vcpu *vcpu, gva_t gva);
611 void __kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu);
612 int kvm_mmu_load(struct kvm_vcpu *vcpu);
613 void kvm_mmu_unload(struct kvm_vcpu *vcpu);
615 int kvm_hypercall(struct kvm_vcpu *vcpu, struct kvm_run *run);
617 static inline int kvm_mmu_page_fault(struct kvm_vcpu *vcpu, gva_t gva,
620 return vcpu->mmu.page_fault(vcpu, gva, error_code);
623 static inline void kvm_mmu_free_some_pages(struct kvm_vcpu *vcpu)
625 if (unlikely(vcpu->kvm->n_free_mmu_pages < KVM_MIN_FREE_MMU_PAGES))
626 __kvm_mmu_free_some_pages(vcpu);
629 static inline int kvm_mmu_reload(struct kvm_vcpu *vcpu)
631 if (likely(vcpu->mmu.root_hpa != INVALID_PAGE))
634 return kvm_mmu_load(vcpu);
637 static inline int is_long_mode(struct kvm_vcpu *vcpu)
640 return vcpu->shadow_efer & EFER_LME;
646 static inline int is_pae(struct kvm_vcpu *vcpu)
648 return vcpu->cr4 & CR4_PAE_MASK;
651 static inline int is_pse(struct kvm_vcpu *vcpu)
653 return vcpu->cr4 & CR4_PSE_MASK;
656 static inline int is_paging(struct kvm_vcpu *vcpu)
658 return vcpu->cr0 & X86_CR0_PG;
661 static inline int memslot_id(struct kvm *kvm, struct kvm_memory_slot *slot)
663 return slot - kvm->memslots;
666 static inline struct kvm_mmu_page *page_header(hpa_t shadow_page)
668 struct page *page = pfn_to_page(shadow_page >> PAGE_SHIFT);
670 return (struct kvm_mmu_page *)page_private(page);
673 static inline u16 read_fs(void)
676 asm ("mov %%fs, %0" : "=g"(seg));
680 static inline u16 read_gs(void)
683 asm ("mov %%gs, %0" : "=g"(seg));
687 static inline u16 read_ldt(void)
690 asm ("sldt %0" : "=g"(ldt));
694 static inline void load_fs(u16 sel)
696 asm ("mov %0, %%fs" : : "rm"(sel));
699 static inline void load_gs(u16 sel)
701 asm ("mov %0, %%gs" : : "rm"(sel));
705 static inline void load_ldt(u16 sel)
707 asm ("lldt %0" : : "rm"(sel));
711 static inline void get_idt(struct descriptor_table *table)
713 asm ("sidt %0" : "=m"(*table));
716 static inline void get_gdt(struct descriptor_table *table)
718 asm ("sgdt %0" : "=m"(*table));
721 static inline unsigned long read_tr_base(void)
724 asm ("str %0" : "=g"(tr));
725 return segment_base(tr);
729 static inline unsigned long read_msr(unsigned long msr)
738 static inline void fx_save(void *image)
740 asm ("fxsave (%0)":: "r" (image));
743 static inline void fx_restore(void *image)
745 asm ("fxrstor (%0)":: "r" (image));
748 static inline void fpu_init(void)
753 static inline u32 get_rdx_init_val(void)
755 return 0x600; /* P6 family */
758 #define ASM_VMX_VMCLEAR_RAX ".byte 0x66, 0x0f, 0xc7, 0x30"
759 #define ASM_VMX_VMLAUNCH ".byte 0x0f, 0x01, 0xc2"
760 #define ASM_VMX_VMRESUME ".byte 0x0f, 0x01, 0xc3"
761 #define ASM_VMX_VMPTRLD_RAX ".byte 0x0f, 0xc7, 0x30"
762 #define ASM_VMX_VMREAD_RDX_RAX ".byte 0x0f, 0x78, 0xd0"
763 #define ASM_VMX_VMWRITE_RAX_RDX ".byte 0x0f, 0x79, 0xd0"
764 #define ASM_VMX_VMWRITE_RSP_RDX ".byte 0x0f, 0x79, 0xd4"
765 #define ASM_VMX_VMXOFF ".byte 0x0f, 0x01, 0xc4"
766 #define ASM_VMX_VMXON_RAX ".byte 0xf3, 0x0f, 0xc7, 0x30"
768 #define MSR_IA32_TIME_STAMP_COUNTER 0x010
770 #define TSS_IOPB_BASE_OFFSET 0x66
771 #define TSS_BASE_SIZE 0x68
772 #define TSS_IOPB_SIZE (65536 / 8)
773 #define TSS_REDIRECTION_SIZE (256 / 8)
774 #define RMODE_TSS_SIZE (TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)