kvm: better advice for failed s390x startup
[qemu/ar7.git] / target-i386 / helper.c
blob4f1ddf701e5e661bbf5e112aae0e1c5e9aac9a8b
1 /*
2 * i386 helpers (without register variable usage)
4 * Copyright (c) 2003 Fabrice Bellard
6 * This library is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2 of the License, or (at your option) any later version.
11 * This library is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with this library; if not, see <http://www.gnu.org/licenses/>.
20 #include "cpu.h"
21 #include "sysemu/kvm.h"
22 #include "kvm_i386.h"
23 #ifndef CONFIG_USER_ONLY
24 #include "sysemu/sysemu.h"
25 #include "monitor/monitor.h"
26 #endif
28 static void cpu_x86_version(CPUX86State *env, int *family, int *model)
30 int cpuver = env->cpuid_version;
32 if (family == NULL || model == NULL) {
33 return;
36 *family = (cpuver >> 8) & 0x0f;
37 *model = ((cpuver >> 12) & 0xf0) + ((cpuver >> 4) & 0x0f);
40 /* Broadcast MCA signal for processor version 06H_EH and above */
41 int cpu_x86_support_mca_broadcast(CPUX86State *env)
43 int family = 0;
44 int model = 0;
46 cpu_x86_version(env, &family, &model);
47 if ((family == 6 && model >= 14) || family > 6) {
48 return 1;
51 return 0;
54 /***********************************************************/
55 /* x86 debug */
57 static const char *cc_op_str[CC_OP_NB] = {
58 "DYNAMIC",
59 "EFLAGS",
61 "MULB",
62 "MULW",
63 "MULL",
64 "MULQ",
66 "ADDB",
67 "ADDW",
68 "ADDL",
69 "ADDQ",
71 "ADCB",
72 "ADCW",
73 "ADCL",
74 "ADCQ",
76 "SUBB",
77 "SUBW",
78 "SUBL",
79 "SUBQ",
81 "SBBB",
82 "SBBW",
83 "SBBL",
84 "SBBQ",
86 "LOGICB",
87 "LOGICW",
88 "LOGICL",
89 "LOGICQ",
91 "INCB",
92 "INCW",
93 "INCL",
94 "INCQ",
96 "DECB",
97 "DECW",
98 "DECL",
99 "DECQ",
101 "SHLB",
102 "SHLW",
103 "SHLL",
104 "SHLQ",
106 "SARB",
107 "SARW",
108 "SARL",
109 "SARQ",
111 "BMILGB",
112 "BMILGW",
113 "BMILGL",
114 "BMILGQ",
116 "ADCX",
117 "ADOX",
118 "ADCOX",
120 "CLR",
123 static void
124 cpu_x86_dump_seg_cache(CPUX86State *env, FILE *f, fprintf_function cpu_fprintf,
125 const char *name, struct SegmentCache *sc)
127 #ifdef TARGET_X86_64
128 if (env->hflags & HF_CS64_MASK) {
129 cpu_fprintf(f, "%-3s=%04x %016" PRIx64 " %08x %08x", name,
130 sc->selector, sc->base, sc->limit, sc->flags & 0x00ffff00);
131 } else
132 #endif
134 cpu_fprintf(f, "%-3s=%04x %08x %08x %08x", name, sc->selector,
135 (uint32_t)sc->base, sc->limit, sc->flags & 0x00ffff00);
138 if (!(env->hflags & HF_PE_MASK) || !(sc->flags & DESC_P_MASK))
139 goto done;
141 cpu_fprintf(f, " DPL=%d ", (sc->flags & DESC_DPL_MASK) >> DESC_DPL_SHIFT);
142 if (sc->flags & DESC_S_MASK) {
143 if (sc->flags & DESC_CS_MASK) {
144 cpu_fprintf(f, (sc->flags & DESC_L_MASK) ? "CS64" :
145 ((sc->flags & DESC_B_MASK) ? "CS32" : "CS16"));
146 cpu_fprintf(f, " [%c%c", (sc->flags & DESC_C_MASK) ? 'C' : '-',
147 (sc->flags & DESC_R_MASK) ? 'R' : '-');
148 } else {
149 cpu_fprintf(f,
150 (sc->flags & DESC_B_MASK || env->hflags & HF_LMA_MASK)
151 ? "DS " : "DS16");
152 cpu_fprintf(f, " [%c%c", (sc->flags & DESC_E_MASK) ? 'E' : '-',
153 (sc->flags & DESC_W_MASK) ? 'W' : '-');
155 cpu_fprintf(f, "%c]", (sc->flags & DESC_A_MASK) ? 'A' : '-');
156 } else {
157 static const char *sys_type_name[2][16] = {
158 { /* 32 bit mode */
159 "Reserved", "TSS16-avl", "LDT", "TSS16-busy",
160 "CallGate16", "TaskGate", "IntGate16", "TrapGate16",
161 "Reserved", "TSS32-avl", "Reserved", "TSS32-busy",
162 "CallGate32", "Reserved", "IntGate32", "TrapGate32"
164 { /* 64 bit mode */
165 "<hiword>", "Reserved", "LDT", "Reserved", "Reserved",
166 "Reserved", "Reserved", "Reserved", "Reserved",
167 "TSS64-avl", "Reserved", "TSS64-busy", "CallGate64",
168 "Reserved", "IntGate64", "TrapGate64"
171 cpu_fprintf(f, "%s",
172 sys_type_name[(env->hflags & HF_LMA_MASK) ? 1 : 0]
173 [(sc->flags & DESC_TYPE_MASK)
174 >> DESC_TYPE_SHIFT]);
176 done:
177 cpu_fprintf(f, "\n");
180 #define DUMP_CODE_BYTES_TOTAL 50
181 #define DUMP_CODE_BYTES_BACKWARD 20
183 void x86_cpu_dump_state(CPUState *cs, FILE *f, fprintf_function cpu_fprintf,
184 int flags)
186 X86CPU *cpu = X86_CPU(cs);
187 CPUX86State *env = &cpu->env;
188 int eflags, i, nb;
189 char cc_op_name[32];
190 static const char *seg_name[6] = { "ES", "CS", "SS", "DS", "FS", "GS" };
192 eflags = cpu_compute_eflags(env);
193 #ifdef TARGET_X86_64
194 if (env->hflags & HF_CS64_MASK) {
195 cpu_fprintf(f,
196 "RAX=%016" PRIx64 " RBX=%016" PRIx64 " RCX=%016" PRIx64 " RDX=%016" PRIx64 "\n"
197 "RSI=%016" PRIx64 " RDI=%016" PRIx64 " RBP=%016" PRIx64 " RSP=%016" PRIx64 "\n"
198 "R8 =%016" PRIx64 " R9 =%016" PRIx64 " R10=%016" PRIx64 " R11=%016" PRIx64 "\n"
199 "R12=%016" PRIx64 " R13=%016" PRIx64 " R14=%016" PRIx64 " R15=%016" PRIx64 "\n"
200 "RIP=%016" PRIx64 " RFL=%08x [%c%c%c%c%c%c%c] CPL=%d II=%d A20=%d SMM=%d HLT=%d\n",
201 env->regs[R_EAX],
202 env->regs[R_EBX],
203 env->regs[R_ECX],
204 env->regs[R_EDX],
205 env->regs[R_ESI],
206 env->regs[R_EDI],
207 env->regs[R_EBP],
208 env->regs[R_ESP],
209 env->regs[8],
210 env->regs[9],
211 env->regs[10],
212 env->regs[11],
213 env->regs[12],
214 env->regs[13],
215 env->regs[14],
216 env->regs[15],
217 env->eip, eflags,
218 eflags & DF_MASK ? 'D' : '-',
219 eflags & CC_O ? 'O' : '-',
220 eflags & CC_S ? 'S' : '-',
221 eflags & CC_Z ? 'Z' : '-',
222 eflags & CC_A ? 'A' : '-',
223 eflags & CC_P ? 'P' : '-',
224 eflags & CC_C ? 'C' : '-',
225 env->hflags & HF_CPL_MASK,
226 (env->hflags >> HF_INHIBIT_IRQ_SHIFT) & 1,
227 (env->a20_mask >> 20) & 1,
228 (env->hflags >> HF_SMM_SHIFT) & 1,
229 cs->halted);
230 } else
231 #endif
233 cpu_fprintf(f, "EAX=%08x EBX=%08x ECX=%08x EDX=%08x\n"
234 "ESI=%08x EDI=%08x EBP=%08x ESP=%08x\n"
235 "EIP=%08x EFL=%08x [%c%c%c%c%c%c%c] CPL=%d II=%d A20=%d SMM=%d HLT=%d\n",
236 (uint32_t)env->regs[R_EAX],
237 (uint32_t)env->regs[R_EBX],
238 (uint32_t)env->regs[R_ECX],
239 (uint32_t)env->regs[R_EDX],
240 (uint32_t)env->regs[R_ESI],
241 (uint32_t)env->regs[R_EDI],
242 (uint32_t)env->regs[R_EBP],
243 (uint32_t)env->regs[R_ESP],
244 (uint32_t)env->eip, eflags,
245 eflags & DF_MASK ? 'D' : '-',
246 eflags & CC_O ? 'O' : '-',
247 eflags & CC_S ? 'S' : '-',
248 eflags & CC_Z ? 'Z' : '-',
249 eflags & CC_A ? 'A' : '-',
250 eflags & CC_P ? 'P' : '-',
251 eflags & CC_C ? 'C' : '-',
252 env->hflags & HF_CPL_MASK,
253 (env->hflags >> HF_INHIBIT_IRQ_SHIFT) & 1,
254 (env->a20_mask >> 20) & 1,
255 (env->hflags >> HF_SMM_SHIFT) & 1,
256 cs->halted);
259 for(i = 0; i < 6; i++) {
260 cpu_x86_dump_seg_cache(env, f, cpu_fprintf, seg_name[i],
261 &env->segs[i]);
263 cpu_x86_dump_seg_cache(env, f, cpu_fprintf, "LDT", &env->ldt);
264 cpu_x86_dump_seg_cache(env, f, cpu_fprintf, "TR", &env->tr);
266 #ifdef TARGET_X86_64
267 if (env->hflags & HF_LMA_MASK) {
268 cpu_fprintf(f, "GDT= %016" PRIx64 " %08x\n",
269 env->gdt.base, env->gdt.limit);
270 cpu_fprintf(f, "IDT= %016" PRIx64 " %08x\n",
271 env->idt.base, env->idt.limit);
272 cpu_fprintf(f, "CR0=%08x CR2=%016" PRIx64 " CR3=%016" PRIx64 " CR4=%08x\n",
273 (uint32_t)env->cr[0],
274 env->cr[2],
275 env->cr[3],
276 (uint32_t)env->cr[4]);
277 for(i = 0; i < 4; i++)
278 cpu_fprintf(f, "DR%d=%016" PRIx64 " ", i, env->dr[i]);
279 cpu_fprintf(f, "\nDR6=%016" PRIx64 " DR7=%016" PRIx64 "\n",
280 env->dr[6], env->dr[7]);
281 } else
282 #endif
284 cpu_fprintf(f, "GDT= %08x %08x\n",
285 (uint32_t)env->gdt.base, env->gdt.limit);
286 cpu_fprintf(f, "IDT= %08x %08x\n",
287 (uint32_t)env->idt.base, env->idt.limit);
288 cpu_fprintf(f, "CR0=%08x CR2=%08x CR3=%08x CR4=%08x\n",
289 (uint32_t)env->cr[0],
290 (uint32_t)env->cr[2],
291 (uint32_t)env->cr[3],
292 (uint32_t)env->cr[4]);
293 for(i = 0; i < 4; i++) {
294 cpu_fprintf(f, "DR%d=" TARGET_FMT_lx " ", i, env->dr[i]);
296 cpu_fprintf(f, "\nDR6=" TARGET_FMT_lx " DR7=" TARGET_FMT_lx "\n",
297 env->dr[6], env->dr[7]);
299 if (flags & CPU_DUMP_CCOP) {
300 if ((unsigned)env->cc_op < CC_OP_NB)
301 snprintf(cc_op_name, sizeof(cc_op_name), "%s", cc_op_str[env->cc_op]);
302 else
303 snprintf(cc_op_name, sizeof(cc_op_name), "[%d]", env->cc_op);
304 #ifdef TARGET_X86_64
305 if (env->hflags & HF_CS64_MASK) {
306 cpu_fprintf(f, "CCS=%016" PRIx64 " CCD=%016" PRIx64 " CCO=%-8s\n",
307 env->cc_src, env->cc_dst,
308 cc_op_name);
309 } else
310 #endif
312 cpu_fprintf(f, "CCS=%08x CCD=%08x CCO=%-8s\n",
313 (uint32_t)env->cc_src, (uint32_t)env->cc_dst,
314 cc_op_name);
317 cpu_fprintf(f, "EFER=%016" PRIx64 "\n", env->efer);
318 if (flags & CPU_DUMP_FPU) {
319 int fptag;
320 fptag = 0;
321 for(i = 0; i < 8; i++) {
322 fptag |= ((!env->fptags[i]) << i);
324 cpu_fprintf(f, "FCW=%04x FSW=%04x [ST=%d] FTW=%02x MXCSR=%08x\n",
325 env->fpuc,
326 (env->fpus & ~0x3800) | (env->fpstt & 0x7) << 11,
327 env->fpstt,
328 fptag,
329 env->mxcsr);
330 for(i=0;i<8;i++) {
331 CPU_LDoubleU u;
332 u.d = env->fpregs[i].d;
333 cpu_fprintf(f, "FPR%d=%016" PRIx64 " %04x",
334 i, u.l.lower, u.l.upper);
335 if ((i & 1) == 1)
336 cpu_fprintf(f, "\n");
337 else
338 cpu_fprintf(f, " ");
340 if (env->hflags & HF_CS64_MASK)
341 nb = 16;
342 else
343 nb = 8;
344 for(i=0;i<nb;i++) {
345 cpu_fprintf(f, "XMM%02d=%08x%08x%08x%08x",
347 env->xmm_regs[i].XMM_L(3),
348 env->xmm_regs[i].XMM_L(2),
349 env->xmm_regs[i].XMM_L(1),
350 env->xmm_regs[i].XMM_L(0));
351 if ((i & 1) == 1)
352 cpu_fprintf(f, "\n");
353 else
354 cpu_fprintf(f, " ");
357 if (flags & CPU_DUMP_CODE) {
358 target_ulong base = env->segs[R_CS].base + env->eip;
359 target_ulong offs = MIN(env->eip, DUMP_CODE_BYTES_BACKWARD);
360 uint8_t code;
361 char codestr[3];
363 cpu_fprintf(f, "Code=");
364 for (i = 0; i < DUMP_CODE_BYTES_TOTAL; i++) {
365 if (cpu_memory_rw_debug(cs, base - offs + i, &code, 1, 0) == 0) {
366 snprintf(codestr, sizeof(codestr), "%02x", code);
367 } else {
368 snprintf(codestr, sizeof(codestr), "??");
370 cpu_fprintf(f, "%s%s%s%s", i > 0 ? " " : "",
371 i == offs ? "<" : "", codestr, i == offs ? ">" : "");
373 cpu_fprintf(f, "\n");
377 /***********************************************************/
378 /* x86 mmu */
379 /* XXX: add PGE support */
381 void x86_cpu_set_a20(X86CPU *cpu, int a20_state)
383 CPUX86State *env = &cpu->env;
385 a20_state = (a20_state != 0);
386 if (a20_state != ((env->a20_mask >> 20) & 1)) {
387 CPUState *cs = CPU(cpu);
389 qemu_log_mask(CPU_LOG_MMU, "A20 update: a20=%d\n", a20_state);
390 /* if the cpu is currently executing code, we must unlink it and
391 all the potentially executing TB */
392 cpu_interrupt(cs, CPU_INTERRUPT_EXITTB);
394 /* when a20 is changed, all the MMU mappings are invalid, so
395 we must flush everything */
396 tlb_flush(cs, 1);
397 env->a20_mask = ~(1 << 20) | (a20_state << 20);
401 void cpu_x86_update_cr0(CPUX86State *env, uint32_t new_cr0)
403 X86CPU *cpu = x86_env_get_cpu(env);
404 int pe_state;
406 qemu_log_mask(CPU_LOG_MMU, "CR0 update: CR0=0x%08x\n", new_cr0);
407 if ((new_cr0 & (CR0_PG_MASK | CR0_WP_MASK | CR0_PE_MASK)) !=
408 (env->cr[0] & (CR0_PG_MASK | CR0_WP_MASK | CR0_PE_MASK))) {
409 tlb_flush(CPU(cpu), 1);
412 #ifdef TARGET_X86_64
413 if (!(env->cr[0] & CR0_PG_MASK) && (new_cr0 & CR0_PG_MASK) &&
414 (env->efer & MSR_EFER_LME)) {
415 /* enter in long mode */
416 /* XXX: generate an exception */
417 if (!(env->cr[4] & CR4_PAE_MASK))
418 return;
419 env->efer |= MSR_EFER_LMA;
420 env->hflags |= HF_LMA_MASK;
421 } else if ((env->cr[0] & CR0_PG_MASK) && !(new_cr0 & CR0_PG_MASK) &&
422 (env->efer & MSR_EFER_LMA)) {
423 /* exit long mode */
424 env->efer &= ~MSR_EFER_LMA;
425 env->hflags &= ~(HF_LMA_MASK | HF_CS64_MASK);
426 env->eip &= 0xffffffff;
428 #endif
429 env->cr[0] = new_cr0 | CR0_ET_MASK;
431 /* update PE flag in hidden flags */
432 pe_state = (env->cr[0] & CR0_PE_MASK);
433 env->hflags = (env->hflags & ~HF_PE_MASK) | (pe_state << HF_PE_SHIFT);
434 /* ensure that ADDSEG is always set in real mode */
435 env->hflags |= ((pe_state ^ 1) << HF_ADDSEG_SHIFT);
436 /* update FPU flags */
437 env->hflags = (env->hflags & ~(HF_MP_MASK | HF_EM_MASK | HF_TS_MASK)) |
438 ((new_cr0 << (HF_MP_SHIFT - 1)) & (HF_MP_MASK | HF_EM_MASK | HF_TS_MASK));
441 /* XXX: in legacy PAE mode, generate a GPF if reserved bits are set in
442 the PDPT */
443 void cpu_x86_update_cr3(CPUX86State *env, target_ulong new_cr3)
445 X86CPU *cpu = x86_env_get_cpu(env);
447 env->cr[3] = new_cr3;
448 if (env->cr[0] & CR0_PG_MASK) {
449 qemu_log_mask(CPU_LOG_MMU,
450 "CR3 update: CR3=" TARGET_FMT_lx "\n", new_cr3);
451 tlb_flush(CPU(cpu), 0);
455 void cpu_x86_update_cr4(CPUX86State *env, uint32_t new_cr4)
457 X86CPU *cpu = x86_env_get_cpu(env);
459 #if defined(DEBUG_MMU)
460 printf("CR4 update: CR4=%08x\n", (uint32_t)env->cr[4]);
461 #endif
462 if ((new_cr4 ^ env->cr[4]) &
463 (CR4_PGE_MASK | CR4_PAE_MASK | CR4_PSE_MASK |
464 CR4_SMEP_MASK | CR4_SMAP_MASK)) {
465 tlb_flush(CPU(cpu), 1);
467 /* SSE handling */
468 if (!(env->features[FEAT_1_EDX] & CPUID_SSE)) {
469 new_cr4 &= ~CR4_OSFXSR_MASK;
471 env->hflags &= ~HF_OSFXSR_MASK;
472 if (new_cr4 & CR4_OSFXSR_MASK) {
473 env->hflags |= HF_OSFXSR_MASK;
476 if (!(env->features[FEAT_7_0_EBX] & CPUID_7_0_EBX_SMAP)) {
477 new_cr4 &= ~CR4_SMAP_MASK;
479 env->hflags &= ~HF_SMAP_MASK;
480 if (new_cr4 & CR4_SMAP_MASK) {
481 env->hflags |= HF_SMAP_MASK;
484 env->cr[4] = new_cr4;
487 #if defined(CONFIG_USER_ONLY)
489 int x86_cpu_handle_mmu_fault(CPUState *cs, vaddr addr,
490 int is_write, int mmu_idx)
492 X86CPU *cpu = X86_CPU(cs);
493 CPUX86State *env = &cpu->env;
495 /* user mode only emulation */
496 is_write &= 1;
497 env->cr[2] = addr;
498 env->error_code = (is_write << PG_ERROR_W_BIT);
499 env->error_code |= PG_ERROR_U_MASK;
500 cs->exception_index = EXCP0E_PAGE;
501 return 1;
504 #else
506 /* return value:
507 * -1 = cannot handle fault
508 * 0 = nothing more to do
509 * 1 = generate PF fault
511 int x86_cpu_handle_mmu_fault(CPUState *cs, vaddr addr,
512 int is_write1, int mmu_idx)
514 X86CPU *cpu = X86_CPU(cs);
515 CPUX86State *env = &cpu->env;
516 uint64_t ptep, pte;
517 target_ulong pde_addr, pte_addr;
518 int error_code = 0;
519 int is_dirty, prot, page_size, is_write, is_user;
520 hwaddr paddr;
521 uint64_t rsvd_mask = PG_HI_RSVD_MASK;
522 uint32_t page_offset;
523 target_ulong vaddr;
525 is_user = mmu_idx == MMU_USER_IDX;
526 #if defined(DEBUG_MMU)
527 printf("MMU fault: addr=%" VADDR_PRIx " w=%d u=%d eip=" TARGET_FMT_lx "\n",
528 addr, is_write1, is_user, env->eip);
529 #endif
530 is_write = is_write1 & 1;
532 if (!(env->cr[0] & CR0_PG_MASK)) {
533 pte = addr;
534 #ifdef TARGET_X86_64
535 if (!(env->hflags & HF_LMA_MASK)) {
536 /* Without long mode we can only address 32bits in real mode */
537 pte = (uint32_t)pte;
539 #endif
540 prot = PAGE_READ | PAGE_WRITE | PAGE_EXEC;
541 page_size = 4096;
542 goto do_mapping;
545 if (!(env->efer & MSR_EFER_NXE)) {
546 rsvd_mask |= PG_NX_MASK;
549 if (env->cr[4] & CR4_PAE_MASK) {
550 uint64_t pde, pdpe;
551 target_ulong pdpe_addr;
553 #ifdef TARGET_X86_64
554 if (env->hflags & HF_LMA_MASK) {
555 uint64_t pml4e_addr, pml4e;
556 int32_t sext;
558 /* test virtual address sign extension */
559 sext = (int64_t)addr >> 47;
560 if (sext != 0 && sext != -1) {
561 env->error_code = 0;
562 cs->exception_index = EXCP0D_GPF;
563 return 1;
566 pml4e_addr = ((env->cr[3] & ~0xfff) + (((addr >> 39) & 0x1ff) << 3)) &
567 env->a20_mask;
568 pml4e = ldq_phys(cs->as, pml4e_addr);
569 if (!(pml4e & PG_PRESENT_MASK)) {
570 goto do_fault;
572 if (pml4e & (rsvd_mask | PG_PSE_MASK)) {
573 goto do_fault_rsvd;
575 if (!(pml4e & PG_ACCESSED_MASK)) {
576 pml4e |= PG_ACCESSED_MASK;
577 stl_phys_notdirty(cs->as, pml4e_addr, pml4e);
579 ptep = pml4e ^ PG_NX_MASK;
580 pdpe_addr = ((pml4e & PG_ADDRESS_MASK) + (((addr >> 30) & 0x1ff) << 3)) &
581 env->a20_mask;
582 pdpe = ldq_phys(cs->as, pdpe_addr);
583 if (!(pdpe & PG_PRESENT_MASK)) {
584 goto do_fault;
586 if (pdpe & rsvd_mask) {
587 goto do_fault_rsvd;
589 ptep &= pdpe ^ PG_NX_MASK;
590 if (!(pdpe & PG_ACCESSED_MASK)) {
591 pdpe |= PG_ACCESSED_MASK;
592 stl_phys_notdirty(cs->as, pdpe_addr, pdpe);
594 if (pdpe & PG_PSE_MASK) {
595 /* 1 GB page */
596 page_size = 1024 * 1024 * 1024;
597 pte_addr = pdpe_addr;
598 pte = pdpe;
599 goto do_check_protect;
601 } else
602 #endif
604 /* XXX: load them when cr3 is loaded ? */
605 pdpe_addr = ((env->cr[3] & ~0x1f) + ((addr >> 27) & 0x18)) &
606 env->a20_mask;
607 pdpe = ldq_phys(cs->as, pdpe_addr);
608 if (!(pdpe & PG_PRESENT_MASK)) {
609 goto do_fault;
611 rsvd_mask |= PG_HI_USER_MASK;
612 if (pdpe & (rsvd_mask | PG_NX_MASK)) {
613 goto do_fault_rsvd;
615 ptep = PG_NX_MASK | PG_USER_MASK | PG_RW_MASK;
618 pde_addr = ((pdpe & PG_ADDRESS_MASK) + (((addr >> 21) & 0x1ff) << 3)) &
619 env->a20_mask;
620 pde = ldq_phys(cs->as, pde_addr);
621 if (!(pde & PG_PRESENT_MASK)) {
622 goto do_fault;
624 if (pde & rsvd_mask) {
625 goto do_fault_rsvd;
627 ptep &= pde ^ PG_NX_MASK;
628 if (pde & PG_PSE_MASK) {
629 /* 2 MB page */
630 page_size = 2048 * 1024;
631 pte_addr = pde_addr;
632 pte = pde;
633 goto do_check_protect;
635 /* 4 KB page */
636 if (!(pde & PG_ACCESSED_MASK)) {
637 pde |= PG_ACCESSED_MASK;
638 stl_phys_notdirty(cs->as, pde_addr, pde);
640 pte_addr = ((pde & PG_ADDRESS_MASK) + (((addr >> 12) & 0x1ff) << 3)) &
641 env->a20_mask;
642 pte = ldq_phys(cs->as, pte_addr);
643 if (!(pte & PG_PRESENT_MASK)) {
644 goto do_fault;
646 if (pte & rsvd_mask) {
647 goto do_fault_rsvd;
649 /* combine pde and pte nx, user and rw protections */
650 ptep &= pte ^ PG_NX_MASK;
651 page_size = 4096;
652 } else {
653 uint32_t pde;
655 /* page directory entry */
656 pde_addr = ((env->cr[3] & ~0xfff) + ((addr >> 20) & 0xffc)) &
657 env->a20_mask;
658 pde = ldl_phys(cs->as, pde_addr);
659 if (!(pde & PG_PRESENT_MASK)) {
660 goto do_fault;
662 ptep = pde | PG_NX_MASK;
664 /* if PSE bit is set, then we use a 4MB page */
665 if ((pde & PG_PSE_MASK) && (env->cr[4] & CR4_PSE_MASK)) {
666 page_size = 4096 * 1024;
667 pte_addr = pde_addr;
669 /* Bits 20-13 provide bits 39-32 of the address, bit 21 is reserved.
670 * Leave bits 20-13 in place for setting accessed/dirty bits below.
672 pte = pde | ((pde & 0x1fe000) << (32 - 13));
673 rsvd_mask = 0x200000;
674 goto do_check_protect_pse36;
677 if (!(pde & PG_ACCESSED_MASK)) {
678 pde |= PG_ACCESSED_MASK;
679 stl_phys_notdirty(cs->as, pde_addr, pde);
682 /* page directory entry */
683 pte_addr = ((pde & ~0xfff) + ((addr >> 10) & 0xffc)) &
684 env->a20_mask;
685 pte = ldl_phys(cs->as, pte_addr);
686 if (!(pte & PG_PRESENT_MASK)) {
687 goto do_fault;
689 /* combine pde and pte user and rw protections */
690 ptep &= pte | PG_NX_MASK;
691 page_size = 4096;
692 rsvd_mask = 0;
695 do_check_protect:
696 rsvd_mask |= (page_size - 1) & PG_ADDRESS_MASK & ~PG_PSE_PAT_MASK;
697 do_check_protect_pse36:
698 if (pte & rsvd_mask) {
699 goto do_fault_rsvd;
701 ptep ^= PG_NX_MASK;
702 if ((ptep & PG_NX_MASK) && is_write1 == 2) {
703 goto do_fault_protect;
705 switch (mmu_idx) {
706 case MMU_USER_IDX:
707 if (!(ptep & PG_USER_MASK)) {
708 goto do_fault_protect;
710 if (is_write && !(ptep & PG_RW_MASK)) {
711 goto do_fault_protect;
713 break;
715 case MMU_KSMAP_IDX:
716 if (is_write1 != 2 && (ptep & PG_USER_MASK)) {
717 goto do_fault_protect;
719 /* fall through */
720 case MMU_KNOSMAP_IDX:
721 if (is_write1 == 2 && (env->cr[4] & CR4_SMEP_MASK) &&
722 (ptep & PG_USER_MASK)) {
723 goto do_fault_protect;
725 if ((env->cr[0] & CR0_WP_MASK) &&
726 is_write && !(ptep & PG_RW_MASK)) {
727 goto do_fault_protect;
729 break;
731 default: /* cannot happen */
732 break;
734 is_dirty = is_write && !(pte & PG_DIRTY_MASK);
735 if (!(pte & PG_ACCESSED_MASK) || is_dirty) {
736 pte |= PG_ACCESSED_MASK;
737 if (is_dirty) {
738 pte |= PG_DIRTY_MASK;
740 stl_phys_notdirty(cs->as, pte_addr, pte);
743 /* the page can be put in the TLB */
744 prot = PAGE_READ;
745 if (!(ptep & PG_NX_MASK) &&
746 (mmu_idx == MMU_USER_IDX ||
747 !((env->cr[4] & CR4_SMEP_MASK) && (ptep & PG_USER_MASK)))) {
748 prot |= PAGE_EXEC;
750 if (pte & PG_DIRTY_MASK) {
751 /* only set write access if already dirty... otherwise wait
752 for dirty access */
753 if (is_user) {
754 if (ptep & PG_RW_MASK)
755 prot |= PAGE_WRITE;
756 } else {
757 if (!(env->cr[0] & CR0_WP_MASK) ||
758 (ptep & PG_RW_MASK))
759 prot |= PAGE_WRITE;
762 do_mapping:
763 pte = pte & env->a20_mask;
765 /* align to page_size */
766 pte &= PG_ADDRESS_MASK & ~(page_size - 1);
768 /* Even if 4MB pages, we map only one 4KB page in the cache to
769 avoid filling it too fast */
770 vaddr = addr & TARGET_PAGE_MASK;
771 page_offset = vaddr & (page_size - 1);
772 paddr = pte + page_offset;
774 tlb_set_page(cs, vaddr, paddr, prot, mmu_idx, page_size);
775 return 0;
776 do_fault_rsvd:
777 error_code |= PG_ERROR_RSVD_MASK;
778 do_fault_protect:
779 error_code |= PG_ERROR_P_MASK;
780 do_fault:
781 error_code |= (is_write << PG_ERROR_W_BIT);
782 if (is_user)
783 error_code |= PG_ERROR_U_MASK;
784 if (is_write1 == 2 &&
785 (((env->efer & MSR_EFER_NXE) &&
786 (env->cr[4] & CR4_PAE_MASK)) ||
787 (env->cr[4] & CR4_SMEP_MASK)))
788 error_code |= PG_ERROR_I_D_MASK;
789 if (env->intercept_exceptions & (1 << EXCP0E_PAGE)) {
790 /* cr2 is not modified in case of exceptions */
791 stq_phys(cs->as,
792 env->vm_vmcb + offsetof(struct vmcb, control.exit_info_2),
793 addr);
794 } else {
795 env->cr[2] = addr;
797 env->error_code = error_code;
798 cs->exception_index = EXCP0E_PAGE;
799 return 1;
802 hwaddr x86_cpu_get_phys_page_debug(CPUState *cs, vaddr addr)
804 X86CPU *cpu = X86_CPU(cs);
805 CPUX86State *env = &cpu->env;
806 target_ulong pde_addr, pte_addr;
807 uint64_t pte;
808 uint32_t page_offset;
809 int page_size;
811 if (!(env->cr[0] & CR0_PG_MASK)) {
812 pte = addr & env->a20_mask;
813 page_size = 4096;
814 } else if (env->cr[4] & CR4_PAE_MASK) {
815 target_ulong pdpe_addr;
816 uint64_t pde, pdpe;
818 #ifdef TARGET_X86_64
819 if (env->hflags & HF_LMA_MASK) {
820 uint64_t pml4e_addr, pml4e;
821 int32_t sext;
823 /* test virtual address sign extension */
824 sext = (int64_t)addr >> 47;
825 if (sext != 0 && sext != -1) {
826 return -1;
828 pml4e_addr = ((env->cr[3] & ~0xfff) + (((addr >> 39) & 0x1ff) << 3)) &
829 env->a20_mask;
830 pml4e = ldq_phys(cs->as, pml4e_addr);
831 if (!(pml4e & PG_PRESENT_MASK)) {
832 return -1;
834 pdpe_addr = ((pml4e & PG_ADDRESS_MASK) +
835 (((addr >> 30) & 0x1ff) << 3)) & env->a20_mask;
836 pdpe = ldq_phys(cs->as, pdpe_addr);
837 if (!(pdpe & PG_PRESENT_MASK)) {
838 return -1;
840 if (pdpe & PG_PSE_MASK) {
841 page_size = 1024 * 1024 * 1024;
842 pte = pdpe;
843 goto out;
846 } else
847 #endif
849 pdpe_addr = ((env->cr[3] & ~0x1f) + ((addr >> 27) & 0x18)) &
850 env->a20_mask;
851 pdpe = ldq_phys(cs->as, pdpe_addr);
852 if (!(pdpe & PG_PRESENT_MASK))
853 return -1;
856 pde_addr = ((pdpe & PG_ADDRESS_MASK) +
857 (((addr >> 21) & 0x1ff) << 3)) & env->a20_mask;
858 pde = ldq_phys(cs->as, pde_addr);
859 if (!(pde & PG_PRESENT_MASK)) {
860 return -1;
862 if (pde & PG_PSE_MASK) {
863 /* 2 MB page */
864 page_size = 2048 * 1024;
865 pte = pde;
866 } else {
867 /* 4 KB page */
868 pte_addr = ((pde & PG_ADDRESS_MASK) +
869 (((addr >> 12) & 0x1ff) << 3)) & env->a20_mask;
870 page_size = 4096;
871 pte = ldq_phys(cs->as, pte_addr);
873 if (!(pte & PG_PRESENT_MASK)) {
874 return -1;
876 } else {
877 uint32_t pde;
879 /* page directory entry */
880 pde_addr = ((env->cr[3] & ~0xfff) + ((addr >> 20) & 0xffc)) & env->a20_mask;
881 pde = ldl_phys(cs->as, pde_addr);
882 if (!(pde & PG_PRESENT_MASK))
883 return -1;
884 if ((pde & PG_PSE_MASK) && (env->cr[4] & CR4_PSE_MASK)) {
885 pte = pde | ((pde & 0x1fe000) << (32 - 13));
886 page_size = 4096 * 1024;
887 } else {
888 /* page directory entry */
889 pte_addr = ((pde & ~0xfff) + ((addr >> 10) & 0xffc)) & env->a20_mask;
890 pte = ldl_phys(cs->as, pte_addr);
891 if (!(pte & PG_PRESENT_MASK)) {
892 return -1;
894 page_size = 4096;
896 pte = pte & env->a20_mask;
899 #ifdef TARGET_X86_64
900 out:
901 #endif
902 pte &= PG_ADDRESS_MASK & ~(page_size - 1);
903 page_offset = (addr & TARGET_PAGE_MASK) & (page_size - 1);
904 return pte | page_offset;
907 void hw_breakpoint_insert(CPUX86State *env, int index)
909 CPUState *cs = CPU(x86_env_get_cpu(env));
910 int type = 0, err = 0;
912 switch (hw_breakpoint_type(env->dr[7], index)) {
913 case DR7_TYPE_BP_INST:
914 if (hw_breakpoint_enabled(env->dr[7], index)) {
915 err = cpu_breakpoint_insert(cs, env->dr[index], BP_CPU,
916 &env->cpu_breakpoint[index]);
918 break;
919 case DR7_TYPE_DATA_WR:
920 type = BP_CPU | BP_MEM_WRITE;
921 break;
922 case DR7_TYPE_IO_RW:
923 /* No support for I/O watchpoints yet */
924 break;
925 case DR7_TYPE_DATA_RW:
926 type = BP_CPU | BP_MEM_ACCESS;
927 break;
930 if (type != 0) {
931 err = cpu_watchpoint_insert(cs, env->dr[index],
932 hw_breakpoint_len(env->dr[7], index),
933 type, &env->cpu_watchpoint[index]);
936 if (err) {
937 env->cpu_breakpoint[index] = NULL;
941 void hw_breakpoint_remove(CPUX86State *env, int index)
943 CPUState *cs;
945 if (!env->cpu_breakpoint[index]) {
946 return;
948 cs = CPU(x86_env_get_cpu(env));
949 switch (hw_breakpoint_type(env->dr[7], index)) {
950 case DR7_TYPE_BP_INST:
951 if (hw_breakpoint_enabled(env->dr[7], index)) {
952 cpu_breakpoint_remove_by_ref(cs, env->cpu_breakpoint[index]);
954 break;
955 case DR7_TYPE_DATA_WR:
956 case DR7_TYPE_DATA_RW:
957 cpu_watchpoint_remove_by_ref(cs, env->cpu_watchpoint[index]);
958 break;
959 case DR7_TYPE_IO_RW:
960 /* No support for I/O watchpoints yet */
961 break;
965 bool check_hw_breakpoints(CPUX86State *env, bool force_dr6_update)
967 target_ulong dr6;
968 int reg;
969 bool hit_enabled = false;
971 dr6 = env->dr[6] & ~0xf;
972 for (reg = 0; reg < DR7_MAX_BP; reg++) {
973 bool bp_match = false;
974 bool wp_match = false;
976 switch (hw_breakpoint_type(env->dr[7], reg)) {
977 case DR7_TYPE_BP_INST:
978 if (env->dr[reg] == env->eip) {
979 bp_match = true;
981 break;
982 case DR7_TYPE_DATA_WR:
983 case DR7_TYPE_DATA_RW:
984 if (env->cpu_watchpoint[reg] &&
985 env->cpu_watchpoint[reg]->flags & BP_WATCHPOINT_HIT) {
986 wp_match = true;
988 break;
989 case DR7_TYPE_IO_RW:
990 break;
992 if (bp_match || wp_match) {
993 dr6 |= 1 << reg;
994 if (hw_breakpoint_enabled(env->dr[7], reg)) {
995 hit_enabled = true;
1000 if (hit_enabled || force_dr6_update) {
1001 env->dr[6] = dr6;
1004 return hit_enabled;
1007 void breakpoint_handler(CPUState *cs)
1009 X86CPU *cpu = X86_CPU(cs);
1010 CPUX86State *env = &cpu->env;
1011 CPUBreakpoint *bp;
1013 if (cs->watchpoint_hit) {
1014 if (cs->watchpoint_hit->flags & BP_CPU) {
1015 cs->watchpoint_hit = NULL;
1016 if (check_hw_breakpoints(env, false)) {
1017 raise_exception(env, EXCP01_DB);
1018 } else {
1019 cpu_resume_from_signal(cs, NULL);
1022 } else {
1023 QTAILQ_FOREACH(bp, &cs->breakpoints, entry) {
1024 if (bp->pc == env->eip) {
1025 if (bp->flags & BP_CPU) {
1026 check_hw_breakpoints(env, true);
1027 raise_exception(env, EXCP01_DB);
1029 break;
1035 typedef struct MCEInjectionParams {
1036 Monitor *mon;
1037 X86CPU *cpu;
1038 int bank;
1039 uint64_t status;
1040 uint64_t mcg_status;
1041 uint64_t addr;
1042 uint64_t misc;
1043 int flags;
1044 } MCEInjectionParams;
1046 static void do_inject_x86_mce(void *data)
1048 MCEInjectionParams *params = data;
1049 CPUX86State *cenv = &params->cpu->env;
1050 CPUState *cpu = CPU(params->cpu);
1051 uint64_t *banks = cenv->mce_banks + 4 * params->bank;
1053 cpu_synchronize_state(cpu);
1056 * If there is an MCE exception being processed, ignore this SRAO MCE
1057 * unless unconditional injection was requested.
1059 if (!(params->flags & MCE_INJECT_UNCOND_AO)
1060 && !(params->status & MCI_STATUS_AR)
1061 && (cenv->mcg_status & MCG_STATUS_MCIP)) {
1062 return;
1065 if (params->status & MCI_STATUS_UC) {
1067 * if MSR_MCG_CTL is not all 1s, the uncorrected error
1068 * reporting is disabled
1070 if ((cenv->mcg_cap & MCG_CTL_P) && cenv->mcg_ctl != ~(uint64_t)0) {
1071 monitor_printf(params->mon,
1072 "CPU %d: Uncorrected error reporting disabled\n",
1073 cpu->cpu_index);
1074 return;
1078 * if MSR_MCi_CTL is not all 1s, the uncorrected error
1079 * reporting is disabled for the bank
1081 if (banks[0] != ~(uint64_t)0) {
1082 monitor_printf(params->mon,
1083 "CPU %d: Uncorrected error reporting disabled for"
1084 " bank %d\n",
1085 cpu->cpu_index, params->bank);
1086 return;
1089 if ((cenv->mcg_status & MCG_STATUS_MCIP) ||
1090 !(cenv->cr[4] & CR4_MCE_MASK)) {
1091 monitor_printf(params->mon,
1092 "CPU %d: Previous MCE still in progress, raising"
1093 " triple fault\n",
1094 cpu->cpu_index);
1095 qemu_log_mask(CPU_LOG_RESET, "Triple fault\n");
1096 qemu_system_reset_request();
1097 return;
1099 if (banks[1] & MCI_STATUS_VAL) {
1100 params->status |= MCI_STATUS_OVER;
1102 banks[2] = params->addr;
1103 banks[3] = params->misc;
1104 cenv->mcg_status = params->mcg_status;
1105 banks[1] = params->status;
1106 cpu_interrupt(cpu, CPU_INTERRUPT_MCE);
1107 } else if (!(banks[1] & MCI_STATUS_VAL)
1108 || !(banks[1] & MCI_STATUS_UC)) {
1109 if (banks[1] & MCI_STATUS_VAL) {
1110 params->status |= MCI_STATUS_OVER;
1112 banks[2] = params->addr;
1113 banks[3] = params->misc;
1114 banks[1] = params->status;
1115 } else {
1116 banks[1] |= MCI_STATUS_OVER;
1120 void cpu_x86_inject_mce(Monitor *mon, X86CPU *cpu, int bank,
1121 uint64_t status, uint64_t mcg_status, uint64_t addr,
1122 uint64_t misc, int flags)
1124 CPUState *cs = CPU(cpu);
1125 CPUX86State *cenv = &cpu->env;
1126 MCEInjectionParams params = {
1127 .mon = mon,
1128 .cpu = cpu,
1129 .bank = bank,
1130 .status = status,
1131 .mcg_status = mcg_status,
1132 .addr = addr,
1133 .misc = misc,
1134 .flags = flags,
1136 unsigned bank_num = cenv->mcg_cap & 0xff;
1138 if (!cenv->mcg_cap) {
1139 monitor_printf(mon, "MCE injection not supported\n");
1140 return;
1142 if (bank >= bank_num) {
1143 monitor_printf(mon, "Invalid MCE bank number\n");
1144 return;
1146 if (!(status & MCI_STATUS_VAL)) {
1147 monitor_printf(mon, "Invalid MCE status code\n");
1148 return;
1150 if ((flags & MCE_INJECT_BROADCAST)
1151 && !cpu_x86_support_mca_broadcast(cenv)) {
1152 monitor_printf(mon, "Guest CPU does not support MCA broadcast\n");
1153 return;
1156 run_on_cpu(cs, do_inject_x86_mce, &params);
1157 if (flags & MCE_INJECT_BROADCAST) {
1158 CPUState *other_cs;
1160 params.bank = 1;
1161 params.status = MCI_STATUS_VAL | MCI_STATUS_UC;
1162 params.mcg_status = MCG_STATUS_MCIP | MCG_STATUS_RIPV;
1163 params.addr = 0;
1164 params.misc = 0;
1165 CPU_FOREACH(other_cs) {
1166 if (other_cs == cs) {
1167 continue;
1169 params.cpu = X86_CPU(other_cs);
1170 run_on_cpu(other_cs, do_inject_x86_mce, &params);
1175 void cpu_report_tpr_access(CPUX86State *env, TPRAccess access)
1177 X86CPU *cpu = x86_env_get_cpu(env);
1178 CPUState *cs = CPU(cpu);
1180 if (kvm_enabled()) {
1181 env->tpr_access_type = access;
1183 cpu_interrupt(cs, CPU_INTERRUPT_TPR);
1184 } else {
1185 cpu_restore_state(cs, cs->mem_io_pc);
1187 apic_handle_tpr_access_report(cpu->apic_state, env->eip, access);
1190 #endif /* !CONFIG_USER_ONLY */
1192 int cpu_x86_get_descr_debug(CPUX86State *env, unsigned int selector,
1193 target_ulong *base, unsigned int *limit,
1194 unsigned int *flags)
1196 X86CPU *cpu = x86_env_get_cpu(env);
1197 CPUState *cs = CPU(cpu);
1198 SegmentCache *dt;
1199 target_ulong ptr;
1200 uint32_t e1, e2;
1201 int index;
1203 if (selector & 0x4)
1204 dt = &env->ldt;
1205 else
1206 dt = &env->gdt;
1207 index = selector & ~7;
1208 ptr = dt->base + index;
1209 if ((index + 7) > dt->limit
1210 || cpu_memory_rw_debug(cs, ptr, (uint8_t *)&e1, sizeof(e1), 0) != 0
1211 || cpu_memory_rw_debug(cs, ptr+4, (uint8_t *)&e2, sizeof(e2), 0) != 0)
1212 return 0;
1214 *base = ((e1 >> 16) | ((e2 & 0xff) << 16) | (e2 & 0xff000000));
1215 *limit = (e1 & 0xffff) | (e2 & 0x000f0000);
1216 if (e2 & DESC_G_MASK)
1217 *limit = (*limit << 12) | 0xfff;
1218 *flags = e2;
1220 return 1;
1223 #if !defined(CONFIG_USER_ONLY)
1224 void do_cpu_init(X86CPU *cpu)
1226 CPUState *cs = CPU(cpu);
1227 CPUX86State *env = &cpu->env;
1228 CPUX86State *save = g_new(CPUX86State, 1);
1229 int sipi = cs->interrupt_request & CPU_INTERRUPT_SIPI;
1231 *save = *env;
1233 cpu_reset(cs);
1234 cs->interrupt_request = sipi;
1235 memcpy(&env->start_init_save, &save->start_init_save,
1236 offsetof(CPUX86State, end_init_save) -
1237 offsetof(CPUX86State, start_init_save));
1238 g_free(save);
1240 if (kvm_enabled()) {
1241 kvm_arch_do_init_vcpu(cpu);
1243 apic_init_reset(cpu->apic_state);
1246 void do_cpu_sipi(X86CPU *cpu)
1248 apic_sipi(cpu->apic_state);
1250 #else
1251 void do_cpu_init(X86CPU *cpu)
1254 void do_cpu_sipi(X86CPU *cpu)
1257 #endif
1259 /* Frob eflags into and out of the CPU temporary format. */
1261 void x86_cpu_exec_enter(CPUState *cs)
1263 X86CPU *cpu = X86_CPU(cs);
1264 CPUX86State *env = &cpu->env;
1266 CC_SRC = env->eflags & (CC_O | CC_S | CC_Z | CC_A | CC_P | CC_C);
1267 env->df = 1 - (2 * ((env->eflags >> 10) & 1));
1268 CC_OP = CC_OP_EFLAGS;
1269 env->eflags &= ~(DF_MASK | CC_O | CC_S | CC_Z | CC_A | CC_P | CC_C);
1272 void x86_cpu_exec_exit(CPUState *cs)
1274 X86CPU *cpu = X86_CPU(cs);
1275 CPUX86State *env = &cpu->env;
1277 env->eflags = cpu_compute_eflags(env);