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>
16 #include <linux/kvm.h>
18 #define CR0_PE_MASK (1ULL << 0)
19 #define CR0_TS_MASK (1ULL << 3)
20 #define CR0_NE_MASK (1ULL << 5)
21 #define CR0_WP_MASK (1ULL << 16)
22 #define CR0_NW_MASK (1ULL << 29)
23 #define CR0_CD_MASK (1ULL << 30)
24 #define CR0_PG_MASK (1ULL << 31)
26 #define CR3_WPT_MASK (1ULL << 3)
27 #define CR3_PCD_MASK (1ULL << 4)
29 #define CR3_RESEVED_BITS 0x07ULL
30 #define CR3_L_MODE_RESEVED_BITS (~((1ULL << 40) - 1) | 0x0fe7ULL)
31 #define CR3_FLAGS_MASK ((1ULL << 5) - 1)
33 #define CR4_VME_MASK (1ULL << 0)
34 #define CR4_PSE_MASK (1ULL << 4)
35 #define CR4_PAE_MASK (1ULL << 5)
36 #define CR4_PGE_MASK (1ULL << 7)
37 #define CR4_VMXE_MASK (1ULL << 13)
39 #define KVM_GUEST_CR0_MASK \
40 (CR0_PG_MASK | CR0_PE_MASK | CR0_WP_MASK | CR0_NE_MASK \
41 | CR0_NW_MASK | CR0_CD_MASK)
42 #define KVM_VM_CR0_ALWAYS_ON \
43 (CR0_PG_MASK | CR0_PE_MASK | CR0_WP_MASK | CR0_NE_MASK)
44 #define KVM_GUEST_CR4_MASK \
45 (CR4_PSE_MASK | CR4_PAE_MASK | CR4_PGE_MASK | CR4_VMXE_MASK | CR4_VME_MASK)
46 #define KVM_PMODE_VM_CR4_ALWAYS_ON (CR4_VMXE_MASK | CR4_PAE_MASK)
47 #define KVM_RMODE_VM_CR4_ALWAYS_ON (CR4_VMXE_MASK | CR4_PAE_MASK | CR4_VME_MASK)
49 #define INVALID_PAGE (~(hpa_t)0)
50 #define UNMAPPED_GVA (~(gpa_t)0)
52 #define KVM_MAX_VCPUS 1
53 #define KVM_MEMORY_SLOTS 4
54 #define KVM_NUM_MMU_PAGES 256
56 #define FX_IMAGE_SIZE 512
57 #define FX_IMAGE_ALIGN 16
58 #define FX_BUF_SIZE (2 * FX_IMAGE_SIZE + FX_IMAGE_ALIGN)
68 #define SELECTOR_TI_MASK (1 << 2)
69 #define SELECTOR_RPL_MASK 0x03
76 * gva - guest virtual address
77 * gpa - guest physical address
78 * gfn - guest frame number
79 * hva - host virtual address
80 * hpa - host physical address
81 * hfn - host frame number
84 typedef unsigned long gva_t;
86 typedef unsigned long gfn_t;
88 typedef unsigned long hva_t;
90 typedef unsigned long hfn_t;
93 struct list_head link;
95 unsigned long slot_bitmap; /* One bit set per slot which has memory
96 * in this shadow page.
98 int global; /* Set if all ptes in this page are global */
108 #define vmx_msr_entry kvm_msr_entry
113 * x86 supports 3 paging modes (4-level 64-bit, 3-level 64-bit, and 2-level
114 * 32-bit). The kvm_mmu structure abstracts the details of the current mmu
118 void (*new_cr3)(struct kvm_vcpu *vcpu);
119 int (*page_fault)(struct kvm_vcpu *vcpu, gva_t gva, u32 err);
120 void (*inval_page)(struct kvm_vcpu *vcpu, gva_t gva);
121 void (*free)(struct kvm_vcpu *vcpu);
122 gpa_t (*gva_to_gpa)(struct kvm_vcpu *vcpu, gva_t gva);
125 int shadow_root_level;
128 struct kvm_guest_debug {
171 struct vcpu_svm *svm;
176 int interrupt_window_open;
177 unsigned long irq_summary; /* bit vector: 1 per word in irq_pending */
178 #define NR_IRQ_WORDS KVM_IRQ_BITMAP_SIZE(unsigned long)
179 unsigned long irq_pending[NR_IRQ_WORDS];
180 unsigned long regs[NR_VCPU_REGS]; /* for rsp: vcpu_load_rsp_rip() */
181 unsigned long rip; /* needs vcpu_load_rsp_rip() */
191 struct vmx_msr_entry *guest_msrs;
192 struct vmx_msr_entry *host_msrs;
194 struct list_head free_pages;
195 struct kvm_mmu_page page_header_buf[KVM_NUM_MMU_PAGES];
198 struct kvm_guest_debug guest_debug;
200 char fx_buf[FX_BUF_SIZE];
202 char *guest_fx_image;
205 int mmio_read_completed;
208 unsigned char mmio_data[8];
209 gpa_t mmio_phys_addr;
214 struct kvm_save_segment {
219 } tr, es, ds, fs, gs;
223 struct kvm_memory_slot {
225 unsigned long npages;
227 struct page **phys_mem;
228 unsigned long *dirty_bitmap;
232 spinlock_t lock; /* protects everything except vcpus */
234 struct kvm_memory_slot memslots[KVM_MEMORY_SLOTS];
235 struct list_head active_mmu_pages;
236 struct kvm_vcpu vcpus[KVM_MAX_VCPUS];
237 int memory_config_version;
251 u32 irq_window_exits;
253 u32 request_irq_exits;
257 struct descriptor_table {
260 } __attribute__((packed));
262 struct kvm_arch_ops {
263 int (*cpu_has_kvm_support)(void); /* __init */
264 int (*disabled_by_bios)(void); /* __init */
265 void (*hardware_enable)(void *dummy); /* __init */
266 void (*hardware_disable)(void *dummy);
267 int (*hardware_setup)(void); /* __init */
268 void (*hardware_unsetup)(void); /* __exit */
270 int (*vcpu_create)(struct kvm_vcpu *vcpu);
271 void (*vcpu_free)(struct kvm_vcpu *vcpu);
273 struct kvm_vcpu *(*vcpu_load)(struct kvm_vcpu *vcpu);
274 void (*vcpu_put)(struct kvm_vcpu *vcpu);
276 int (*set_guest_debug)(struct kvm_vcpu *vcpu,
277 struct kvm_debug_guest *dbg);
278 int (*get_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata);
279 int (*set_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 data);
280 u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
281 void (*get_segment)(struct kvm_vcpu *vcpu,
282 struct kvm_segment *var, int seg);
283 void (*set_segment)(struct kvm_vcpu *vcpu,
284 struct kvm_segment *var, int seg);
285 void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
286 void (*decache_cr0_cr4_guest_bits)(struct kvm_vcpu *vcpu);
287 void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
288 void (*set_cr0_no_modeswitch)(struct kvm_vcpu *vcpu,
290 void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
291 void (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
292 void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
293 void (*get_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
294 void (*set_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
295 void (*get_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
296 void (*set_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
297 unsigned long (*get_dr)(struct kvm_vcpu *vcpu, int dr);
298 void (*set_dr)(struct kvm_vcpu *vcpu, int dr, unsigned long value,
300 void (*cache_regs)(struct kvm_vcpu *vcpu);
301 void (*decache_regs)(struct kvm_vcpu *vcpu);
302 unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
303 void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
305 void (*invlpg)(struct kvm_vcpu *vcpu, gva_t addr);
306 void (*tlb_flush)(struct kvm_vcpu *vcpu);
307 void (*inject_page_fault)(struct kvm_vcpu *vcpu,
308 unsigned long addr, u32 err_code);
310 void (*inject_gp)(struct kvm_vcpu *vcpu, unsigned err_code);
312 int (*run)(struct kvm_vcpu *vcpu, struct kvm_run *run);
313 int (*vcpu_setup)(struct kvm_vcpu *vcpu);
314 void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
317 extern struct kvm_stat kvm_stat;
318 extern struct kvm_arch_ops *kvm_arch_ops;
320 #define kvm_printf(kvm, fmt ...) printk(KERN_DEBUG fmt)
321 #define vcpu_printf(vcpu, fmt...) kvm_printf(vcpu->kvm, fmt)
323 int kvm_init_arch(struct kvm_arch_ops *ops, struct module *module);
324 void kvm_exit_arch(void);
326 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
327 int kvm_mmu_create(struct kvm_vcpu *vcpu);
328 int kvm_mmu_setup(struct kvm_vcpu *vcpu);
330 int kvm_mmu_reset_context(struct kvm_vcpu *vcpu);
331 void kvm_mmu_slot_remove_write_access(struct kvm *kvm, int slot);
333 hpa_t gpa_to_hpa(struct kvm_vcpu *vcpu, gpa_t gpa);
334 #define HPA_MSB ((sizeof(hpa_t) * 8) - 1)
335 #define HPA_ERR_MASK ((hpa_t)1 << HPA_MSB)
336 static inline int is_error_hpa(hpa_t hpa) { return hpa >> HPA_MSB; }
337 hpa_t gva_to_hpa(struct kvm_vcpu *vcpu, gva_t gva);
339 void kvm_emulator_want_group7_invlpg(void);
341 extern hpa_t bad_page_address;
343 static inline struct page *gfn_to_page(struct kvm_memory_slot *slot, gfn_t gfn)
345 return slot->phys_mem[gfn - slot->base_gfn];
348 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
349 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
351 enum emulation_result {
352 EMULATE_DONE, /* no further processing */
353 EMULATE_DO_MMIO, /* kvm_run filled with mmio request */
354 EMULATE_FAIL, /* can't emulate this instruction */
357 int emulate_instruction(struct kvm_vcpu *vcpu, struct kvm_run *run,
358 unsigned long cr2, u16 error_code);
359 void realmode_lgdt(struct kvm_vcpu *vcpu, u16 size, unsigned long address);
360 void realmode_lidt(struct kvm_vcpu *vcpu, u16 size, unsigned long address);
361 void realmode_lmsw(struct kvm_vcpu *vcpu, unsigned long msw,
362 unsigned long *rflags);
364 unsigned long realmode_get_cr(struct kvm_vcpu *vcpu, int cr);
365 void realmode_set_cr(struct kvm_vcpu *vcpu, int cr, unsigned long value,
366 unsigned long *rflags);
368 struct x86_emulate_ctxt;
370 int emulate_invlpg(struct kvm_vcpu *vcpu, gva_t address);
371 int emulate_clts(struct kvm_vcpu *vcpu);
372 int emulator_get_dr(struct x86_emulate_ctxt* ctxt, int dr,
373 unsigned long *dest);
374 int emulator_set_dr(struct x86_emulate_ctxt *ctxt, int dr,
375 unsigned long value);
377 void set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
378 void set_cr3(struct kvm_vcpu *vcpu, unsigned long cr0);
379 void set_cr4(struct kvm_vcpu *vcpu, unsigned long cr0);
380 void set_cr8(struct kvm_vcpu *vcpu, unsigned long cr0);
381 void lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
383 int kvm_get_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata);
384 int kvm_set_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 data);
386 void fx_init(struct kvm_vcpu *vcpu);
388 void load_msrs(struct vmx_msr_entry *e, int n);
389 void save_msrs(struct vmx_msr_entry *e, int n);
390 void kvm_resched(struct kvm_vcpu *vcpu);
392 int kvm_read_guest(struct kvm_vcpu *vcpu,
397 int kvm_write_guest(struct kvm_vcpu *vcpu,
402 unsigned long segment_base(u16 selector);
404 static inline struct page *_gfn_to_page(struct kvm *kvm, gfn_t gfn)
406 struct kvm_memory_slot *slot = gfn_to_memslot(kvm, gfn);
407 return (slot) ? slot->phys_mem[gfn - slot->base_gfn] : NULL;
410 static inline int is_long_mode(struct kvm_vcpu *vcpu)
413 return vcpu->shadow_efer & EFER_LME;
419 static inline int is_pae(struct kvm_vcpu *vcpu)
421 return vcpu->cr4 & CR4_PAE_MASK;
424 static inline int is_pse(struct kvm_vcpu *vcpu)
426 return vcpu->cr4 & CR4_PSE_MASK;
429 static inline int is_paging(struct kvm_vcpu *vcpu)
431 return vcpu->cr0 & CR0_PG_MASK;
434 static inline int memslot_id(struct kvm *kvm, struct kvm_memory_slot *slot)
436 return slot - kvm->memslots;
439 static inline struct kvm_mmu_page *page_header(hpa_t shadow_page)
441 struct page *page = pfn_to_page(shadow_page >> PAGE_SHIFT);
443 return (struct kvm_mmu_page *)page->private;
446 static inline u16 read_fs(void)
449 asm ("mov %%fs, %0" : "=g"(seg));
453 static inline u16 read_gs(void)
456 asm ("mov %%gs, %0" : "=g"(seg));
460 static inline u16 read_ldt(void)
463 asm ("sldt %0" : "=g"(ldt));
467 static inline void load_fs(u16 sel)
469 asm ("mov %0, %%fs" : : "rm"(sel));
472 static inline void load_gs(u16 sel)
474 asm ("mov %0, %%gs" : : "rm"(sel));
478 static inline void load_ldt(u16 sel)
480 asm ("lldt %0" : : "g"(sel));
484 static inline void get_idt(struct descriptor_table *table)
486 asm ("sidt %0" : "=m"(*table));
489 static inline void get_gdt(struct descriptor_table *table)
491 asm ("sgdt %0" : "=m"(*table));
494 static inline unsigned long read_tr_base(void)
497 asm ("str %0" : "=g"(tr));
498 return segment_base(tr);
502 static inline unsigned long read_msr(unsigned long msr)
511 static inline void fx_save(void *image)
513 asm ("fxsave (%0)":: "r" (image));
516 static inline void fx_restore(void *image)
518 asm ("fxrstor (%0)":: "r" (image));
521 static inline void fpu_init(void)
526 static inline u32 get_rdx_init_val(void)
528 return 0x600; /* P6 family */
531 #define ASM_VMX_VMCLEAR_RAX ".byte 0x66, 0x0f, 0xc7, 0x30"
532 #define ASM_VMX_VMLAUNCH ".byte 0x0f, 0x01, 0xc2"
533 #define ASM_VMX_VMRESUME ".byte 0x0f, 0x01, 0xc3"
534 #define ASM_VMX_VMPTRLD_RAX ".byte 0x0f, 0xc7, 0x30"
535 #define ASM_VMX_VMREAD_RDX_RAX ".byte 0x0f, 0x78, 0xd0"
536 #define ASM_VMX_VMWRITE_RAX_RDX ".byte 0x0f, 0x79, 0xd0"
537 #define ASM_VMX_VMWRITE_RSP_RDX ".byte 0x0f, 0x79, 0xd4"
538 #define ASM_VMX_VMXOFF ".byte 0x0f, 0x01, 0xc4"
539 #define ASM_VMX_VMXON_RAX ".byte 0xf3, 0x0f, 0xc7, 0x30"
541 #define MSR_IA32_TIME_STAMP_COUNTER 0x010
543 #define TSS_IOPB_BASE_OFFSET 0x66
544 #define TSS_BASE_SIZE 0x68
545 #define TSS_IOPB_SIZE (65536 / 8)
546 #define TSS_REDIRECTION_SIZE (256 / 8)
547 #define RMODE_TSS_SIZE (TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
552 * When emulating 32-bit mode, cr3 is only 32 bits even on x86_64. Therefore
553 * we need to allocate shadow page tables in the first 4GB of memory, which
554 * happens to fit the DMA32 zone.
556 #define GFP_KVM_MMU (GFP_KERNEL | __GFP_DMA32)
560 #define GFP_KVM_MMU GFP_KERNEL