1 #include "kvm/kvm-cpu.h" 2 3 #include "kvm/symbol.h" 4 #include "kvm/util.h" 5 #include "kvm/kvm.h" 6 7 #include <asm/msr-index.h> 8 #include <asm/apicdef.h> 9 10 #include <sys/ioctl.h> 11 #include <sys/mman.h> 12 #include <signal.h> 13 #include <stdlib.h> 14 #include <string.h> 15 #include <errno.h> 16 #include <stdio.h> 17 18 static int debug_fd; 19 20 void kvm_cpu__set_debug_fd(int fd) 21 { 22 debug_fd = fd; 23 } 24 25 int kvm_cpu__get_debug_fd(void) 26 { 27 return debug_fd; 28 } 29 30 static inline bool is_in_protected_mode(struct kvm_cpu *vcpu) 31 { 32 return vcpu->sregs.cr0 & 0x01; 33 } 34 35 static inline u64 ip_to_flat(struct kvm_cpu *vcpu, u64 ip) 36 { 37 u64 cs; 38 39 /* 40 * NOTE! We should take code segment base address into account here. 41 * Luckily it's usually zero because Linux uses flat memory model. 42 */ 43 if (is_in_protected_mode(vcpu)) 44 return ip; 45 46 cs = vcpu->sregs.cs.selector; 47 48 return ip + (cs << 4); 49 } 50 51 static inline u32 selector_to_base(u16 selector) 52 { 53 /* 54 * KVM on Intel requires 'base' to be 'selector * 16' in real mode. 55 */ 56 return (u32)selector * 16; 57 } 58 59 static struct kvm_cpu *kvm_cpu__new(struct kvm *kvm) 60 { 61 struct kvm_cpu *vcpu; 62 63 vcpu = calloc(1, sizeof *vcpu); 64 if (!vcpu) 65 return NULL; 66 67 vcpu->kvm = kvm; 68 69 return vcpu; 70 } 71 72 void kvm_cpu__delete(struct kvm_cpu *vcpu) 73 { 74 if (vcpu->msrs) 75 free(vcpu->msrs); 76 77 free(vcpu); 78 } 79 80 static int kvm_cpu__set_lint(struct kvm_cpu *vcpu) 81 { 82 struct kvm_lapic_state klapic; 83 struct local_apic *lapic = (void *)&klapic; 84 u32 lvt; 85 86 if (ioctl(vcpu->vcpu_fd, KVM_GET_LAPIC, &klapic)) 87 return -1; 88 89 lvt = *(u32 *)&lapic->lvt_lint0; 90 lvt = SET_APIC_DELIVERY_MODE(lvt, APIC_MODE_EXTINT); 91 *(u32 *)&lapic->lvt_lint0 = lvt; 92 93 lvt = *(u32 *)&lapic->lvt_lint1; 94 lvt = SET_APIC_DELIVERY_MODE(lvt, APIC_MODE_NMI); 95 *(u32 *)&lapic->lvt_lint1 = lvt; 96 97 return ioctl(vcpu->vcpu_fd, KVM_SET_LAPIC, &klapic); 98 } 99 100 struct kvm_cpu *kvm_cpu__init(struct kvm *kvm, unsigned long cpu_id) 101 { 102 struct kvm_cpu *vcpu; 103 int mmap_size; 104 int coalesced_offset; 105 106 vcpu = kvm_cpu__new(kvm); 107 if (!vcpu) 108 return NULL; 109 110 vcpu->cpu_id = cpu_id; 111 112 vcpu->vcpu_fd = ioctl(vcpu->kvm->vm_fd, KVM_CREATE_VCPU, cpu_id); 113 if (vcpu->vcpu_fd < 0) 114 die_perror("KVM_CREATE_VCPU ioctl"); 115 116 mmap_size = ioctl(vcpu->kvm->sys_fd, KVM_GET_VCPU_MMAP_SIZE, 0); 117 if (mmap_size < 0) 118 die_perror("KVM_GET_VCPU_MMAP_SIZE ioctl"); 119 120 vcpu->kvm_run = mmap(NULL, mmap_size, PROT_RW, MAP_SHARED, vcpu->vcpu_fd, 0); 121 if (vcpu->kvm_run == MAP_FAILED) 122 die("unable to mmap vcpu fd"); 123 124 coalesced_offset = ioctl(kvm->sys_fd, KVM_CHECK_EXTENSION, KVM_CAP_COALESCED_MMIO); 125 if (coalesced_offset) 126 vcpu->ring = (void *)vcpu->kvm_run + (coalesced_offset * PAGE_SIZE); 127 128 if (kvm_cpu__set_lint(vcpu)) 129 die_perror("KVM_SET_LAPIC failed"); 130 131 vcpu->is_running = true; 132 133 return vcpu; 134 } 135 136 static struct kvm_msrs *kvm_msrs__new(size_t nmsrs) 137 { 138 struct kvm_msrs *vcpu = calloc(1, sizeof(*vcpu) + (sizeof(struct kvm_msr_entry) * nmsrs)); 139 140 if (!vcpu) 141 die("out of memory"); 142 143 return vcpu; 144 } 145 146 #define KVM_MSR_ENTRY(_index, _data) \ 147 (struct kvm_msr_entry) { .index = _index, .data = _data } 148 149 static void kvm_cpu__setup_msrs(struct kvm_cpu *vcpu) 150 { 151 unsigned long ndx = 0; 152 153 vcpu->msrs = kvm_msrs__new(100); 154 155 vcpu->msrs->entries[ndx++] = KVM_MSR_ENTRY(MSR_IA32_SYSENTER_CS, 0x0); 156 vcpu->msrs->entries[ndx++] = KVM_MSR_ENTRY(MSR_IA32_SYSENTER_ESP, 0x0); 157 vcpu->msrs->entries[ndx++] = KVM_MSR_ENTRY(MSR_IA32_SYSENTER_EIP, 0x0); 158 #ifdef CONFIG_X86_64 159 vcpu->msrs->entries[ndx++] = KVM_MSR_ENTRY(MSR_STAR, 0x0); 160 vcpu->msrs->entries[ndx++] = KVM_MSR_ENTRY(MSR_CSTAR, 0x0); 161 vcpu->msrs->entries[ndx++] = KVM_MSR_ENTRY(MSR_KERNEL_GS_BASE, 0x0); 162 vcpu->msrs->entries[ndx++] = KVM_MSR_ENTRY(MSR_SYSCALL_MASK, 0x0); 163 vcpu->msrs->entries[ndx++] = KVM_MSR_ENTRY(MSR_LSTAR, 0x0); 164 #endif 165 vcpu->msrs->entries[ndx++] = KVM_MSR_ENTRY(MSR_IA32_TSC, 0x0); 166 vcpu->msrs->entries[ndx++] = KVM_MSR_ENTRY(MSR_IA32_MISC_ENABLE, 167 MSR_IA32_MISC_ENABLE_FAST_STRING); 168 169 vcpu->msrs->nmsrs = ndx; 170 171 if (ioctl(vcpu->vcpu_fd, KVM_SET_MSRS, vcpu->msrs) < 0) 172 die_perror("KVM_SET_MSRS failed"); 173 } 174 175 static void kvm_cpu__setup_fpu(struct kvm_cpu *vcpu) 176 { 177 vcpu->fpu = (struct kvm_fpu) { 178 .fcw = 0x37f, 179 .mxcsr = 0x1f80, 180 }; 181 182 if (ioctl(vcpu->vcpu_fd, KVM_SET_FPU, &vcpu->fpu) < 0) 183 die_perror("KVM_SET_FPU failed"); 184 } 185 186 static void kvm_cpu__setup_regs(struct kvm_cpu *vcpu) 187 { 188 vcpu->regs = (struct kvm_regs) { 189 /* We start the guest in 16-bit real mode */ 190 .rflags = 0x0000000000000002ULL, 191 192 .rip = vcpu->kvm->boot_ip, 193 .rsp = vcpu->kvm->boot_sp, 194 .rbp = vcpu->kvm->boot_sp, 195 }; 196 197 if (vcpu->regs.rip > USHRT_MAX) 198 die("ip 0x%llx is too high for real mode", (u64) vcpu->regs.rip); 199 200 if (ioctl(vcpu->vcpu_fd, KVM_SET_REGS, &vcpu->regs) < 0) 201 die_perror("KVM_SET_REGS failed"); 202 } 203 204 static void kvm_cpu__setup_sregs(struct kvm_cpu *vcpu) 205 { 206 207 if (ioctl(vcpu->vcpu_fd, KVM_GET_SREGS, &vcpu->sregs) < 0) 208 die_perror("KVM_GET_SREGS failed"); 209 210 vcpu->sregs.cs.selector = vcpu->kvm->boot_selector; 211 vcpu->sregs.cs.base = selector_to_base(vcpu->kvm->boot_selector); 212 vcpu->sregs.ss.selector = vcpu->kvm->boot_selector; 213 vcpu->sregs.ss.base = selector_to_base(vcpu->kvm->boot_selector); 214 vcpu->sregs.ds.selector = vcpu->kvm->boot_selector; 215 vcpu->sregs.ds.base = selector_to_base(vcpu->kvm->boot_selector); 216 vcpu->sregs.es.selector = vcpu->kvm->boot_selector; 217 vcpu->sregs.es.base = selector_to_base(vcpu->kvm->boot_selector); 218 vcpu->sregs.fs.selector = vcpu->kvm->boot_selector; 219 vcpu->sregs.fs.base = selector_to_base(vcpu->kvm->boot_selector); 220 vcpu->sregs.gs.selector = vcpu->kvm->boot_selector; 221 vcpu->sregs.gs.base = selector_to_base(vcpu->kvm->boot_selector); 222 223 if (ioctl(vcpu->vcpu_fd, KVM_SET_SREGS, &vcpu->sregs) < 0) 224 die_perror("KVM_SET_SREGS failed"); 225 } 226 227 /** 228 * kvm_cpu__reset_vcpu - reset virtual CPU to a known state 229 */ 230 void kvm_cpu__reset_vcpu(struct kvm_cpu *vcpu) 231 { 232 kvm_cpu__setup_cpuid(vcpu); 233 kvm_cpu__setup_sregs(vcpu); 234 kvm_cpu__setup_regs(vcpu); 235 kvm_cpu__setup_fpu(vcpu); 236 kvm_cpu__setup_msrs(vcpu); 237 } 238 239 bool kvm_cpu__handle_exit(struct kvm_cpu *vcpu) 240 { 241 return false; 242 } 243 244 static void print_dtable(const char *name, struct kvm_dtable *dtable) 245 { 246 dprintf(debug_fd, " %s %016llx %08hx\n", 247 name, (u64) dtable->base, (u16) dtable->limit); 248 } 249 250 static void print_segment(const char *name, struct kvm_segment *seg) 251 { 252 dprintf(debug_fd, " %s %04hx %016llx %08x %02hhx %x %x %x %x %x %x %x\n", 253 name, (u16) seg->selector, (u64) seg->base, (u32) seg->limit, 254 (u8) seg->type, seg->present, seg->dpl, seg->db, seg->s, seg->l, seg->g, seg->avl); 255 } 256 257 void kvm_cpu__show_registers(struct kvm_cpu *vcpu) 258 { 259 unsigned long cr0, cr2, cr3; 260 unsigned long cr4, cr8; 261 unsigned long rax, rbx, rcx; 262 unsigned long rdx, rsi, rdi; 263 unsigned long rbp, r8, r9; 264 unsigned long r10, r11, r12; 265 unsigned long r13, r14, r15; 266 unsigned long rip, rsp; 267 struct kvm_sregs sregs; 268 unsigned long rflags; 269 struct kvm_regs regs; 270 int i; 271 272 if (ioctl(vcpu->vcpu_fd, KVM_GET_REGS, ®s) < 0) 273 die("KVM_GET_REGS failed"); 274 275 rflags = regs.rflags; 276 277 rip = regs.rip; rsp = regs.rsp; 278 rax = regs.rax; rbx = regs.rbx; rcx = regs.rcx; 279 rdx = regs.rdx; rsi = regs.rsi; rdi = regs.rdi; 280 rbp = regs.rbp; r8 = regs.r8; r9 = regs.r9; 281 r10 = regs.r10; r11 = regs.r11; r12 = regs.r12; 282 r13 = regs.r13; r14 = regs.r14; r15 = regs.r15; 283 284 dprintf(debug_fd, "\n Registers:\n"); 285 dprintf(debug_fd, " ----------\n"); 286 dprintf(debug_fd, " rip: %016lx rsp: %016lx flags: %016lx\n", rip, rsp, rflags); 287 dprintf(debug_fd, " rax: %016lx rbx: %016lx rcx: %016lx\n", rax, rbx, rcx); 288 dprintf(debug_fd, " rdx: %016lx rsi: %016lx rdi: %016lx\n", rdx, rsi, rdi); 289 dprintf(debug_fd, " rbp: %016lx r8: %016lx r9: %016lx\n", rbp, r8, r9); 290 dprintf(debug_fd, " r10: %016lx r11: %016lx r12: %016lx\n", r10, r11, r12); 291 dprintf(debug_fd, " r13: %016lx r14: %016lx r15: %016lx\n", r13, r14, r15); 292 293 if (ioctl(vcpu->vcpu_fd, KVM_GET_SREGS, &sregs) < 0) 294 die("KVM_GET_REGS failed"); 295 296 cr0 = sregs.cr0; cr2 = sregs.cr2; cr3 = sregs.cr3; 297 cr4 = sregs.cr4; cr8 = sregs.cr8; 298 299 dprintf(debug_fd, " cr0: %016lx cr2: %016lx cr3: %016lx\n", cr0, cr2, cr3); 300 dprintf(debug_fd, " cr4: %016lx cr8: %016lx\n", cr4, cr8); 301 dprintf(debug_fd, "\n Segment registers:\n"); 302 dprintf(debug_fd, " ------------------\n"); 303 dprintf(debug_fd, " register selector base limit type p dpl db s l g avl\n"); 304 print_segment("cs ", &sregs.cs); 305 print_segment("ss ", &sregs.ss); 306 print_segment("ds ", &sregs.ds); 307 print_segment("es ", &sregs.es); 308 print_segment("fs ", &sregs.fs); 309 print_segment("gs ", &sregs.gs); 310 print_segment("tr ", &sregs.tr); 311 print_segment("ldt", &sregs.ldt); 312 print_dtable("gdt", &sregs.gdt); 313 print_dtable("idt", &sregs.idt); 314 315 dprintf(debug_fd, "\n APIC:\n"); 316 dprintf(debug_fd, " -----\n"); 317 dprintf(debug_fd, " efer: %016llx apic base: %016llx nmi: %s\n", 318 (u64) sregs.efer, (u64) sregs.apic_base, 319 (vcpu->kvm->nmi_disabled ? "disabled" : "enabled")); 320 321 dprintf(debug_fd, "\n Interrupt bitmap:\n"); 322 dprintf(debug_fd, " -----------------\n"); 323 for (i = 0; i < (KVM_NR_INTERRUPTS + 63) / 64; i++) 324 dprintf(debug_fd, " %016llx", (u64) sregs.interrupt_bitmap[i]); 325 dprintf(debug_fd, "\n"); 326 } 327 328 #define MAX_SYM_LEN 128 329 330 void kvm_cpu__show_code(struct kvm_cpu *vcpu) 331 { 332 unsigned int code_bytes = 64; 333 unsigned int code_prologue = code_bytes * 43 / 64; 334 unsigned int code_len = code_bytes; 335 char sym[MAX_SYM_LEN]; 336 unsigned char c; 337 unsigned int i; 338 u8 *ip; 339 340 if (ioctl(vcpu->vcpu_fd, KVM_GET_REGS, &vcpu->regs) < 0) 341 die("KVM_GET_REGS failed"); 342 343 if (ioctl(vcpu->vcpu_fd, KVM_GET_SREGS, &vcpu->sregs) < 0) 344 die("KVM_GET_SREGS failed"); 345 346 ip = guest_flat_to_host(vcpu->kvm, ip_to_flat(vcpu, vcpu->regs.rip) - code_prologue); 347 348 dprintf(debug_fd, "\n Code:\n"); 349 dprintf(debug_fd, " -----\n"); 350 351 symbol__lookup(vcpu->kvm, vcpu->regs.rip, sym, MAX_SYM_LEN); 352 353 dprintf(debug_fd, " rip: [<%016lx>] %s\n\n", (unsigned long) vcpu->regs.rip, sym); 354 355 for (i = 0; i < code_len; i++, ip++) { 356 if (!host_ptr_in_ram(vcpu->kvm, ip)) 357 break; 358 359 c = *ip; 360 361 if (ip == guest_flat_to_host(vcpu->kvm, ip_to_flat(vcpu, vcpu->regs.rip))) 362 dprintf(debug_fd, " <%02x>", c); 363 else 364 dprintf(debug_fd, " %02x", c); 365 } 366 367 dprintf(debug_fd, "\n"); 368 369 dprintf(debug_fd, "\n Stack:\n"); 370 dprintf(debug_fd, " ------\n"); 371 kvm__dump_mem(vcpu->kvm, vcpu->regs.rsp, 32); 372 } 373 374 void kvm_cpu__show_page_tables(struct kvm_cpu *vcpu) 375 { 376 u64 *pte1; 377 u64 *pte2; 378 u64 *pte3; 379 u64 *pte4; 380 381 if (!is_in_protected_mode(vcpu)) 382 return; 383 384 if (ioctl(vcpu->vcpu_fd, KVM_GET_SREGS, &vcpu->sregs) < 0) 385 die("KVM_GET_SREGS failed"); 386 387 pte4 = guest_flat_to_host(vcpu->kvm, vcpu->sregs.cr3); 388 if (!host_ptr_in_ram(vcpu->kvm, pte4)) 389 return; 390 391 pte3 = guest_flat_to_host(vcpu->kvm, (*pte4 & ~0xfff)); 392 if (!host_ptr_in_ram(vcpu->kvm, pte3)) 393 return; 394 395 pte2 = guest_flat_to_host(vcpu->kvm, (*pte3 & ~0xfff)); 396 if (!host_ptr_in_ram(vcpu->kvm, pte2)) 397 return; 398 399 pte1 = guest_flat_to_host(vcpu->kvm, (*pte2 & ~0xfff)); 400 if (!host_ptr_in_ram(vcpu->kvm, pte1)) 401 return; 402 403 dprintf(debug_fd, "Page Tables:\n"); 404 if (*pte2 & (1 << 7)) 405 dprintf(debug_fd, " pte4: %016llx pte3: %016llx" 406 " pte2: %016llx\n", 407 *pte4, *pte3, *pte2); 408 else 409 dprintf(debug_fd, " pte4: %016llx pte3: %016llx pte2: %016" 410 "llx pte1: %016llx\n", 411 *pte4, *pte3, *pte2, *pte1); 412 } 413 414 void kvm_cpu__arch_nmi(struct kvm_cpu *cpu) 415 { 416 struct kvm_lapic_state klapic; 417 struct local_apic *lapic = (void *)&klapic; 418 419 if (ioctl(cpu->vcpu_fd, KVM_GET_LAPIC, &klapic) != 0) 420 return; 421 422 if (lapic->lvt_lint1.mask) 423 return; 424 425 if (lapic->lvt_lint1.delivery_mode != APIC_MODE_NMI) 426 return; 427 428 ioctl(cpu->vcpu_fd, KVM_NMI); 429 } 430