]> git.proxmox.com Git - qemu.git/blob - target-i386/helper.c
projects / qemu.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge remote-tracking branch 'sstabellini/xen-fixes-20120801' into staging
[qemu.git] / target-i386 / helper.c
1 /*
2 * i386 helpers (without register variable usage)
3 *
4 * Copyright (c) 2003 Fabrice Bellard
5 *
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.
10 *
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.
15 *
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/>.
18 */
19
20 #include "cpu.h"
21 #include "kvm.h"
22 #ifndef CONFIG_USER_ONLY
23 #include "sysemu.h"
24 #include "monitor.h"
25 #endif
26
27 //#define DEBUG_MMU
28
29 static void cpu_x86_version(CPUX86State *env, int *family, int *model)
30 {
31 int cpuver = env->cpuid_version;
32
33 if (family == NULL || model == NULL) {
34 return;
35 }
36
37 *family = (cpuver >> 8) & 0x0f;
38 *model = ((cpuver >> 12) & 0xf0) + ((cpuver >> 4) & 0x0f);
39 }
40
41 /* Broadcast MCA signal for processor version 06H_EH and above */
42 int cpu_x86_support_mca_broadcast(CPUX86State *env)
43 {
44 int family = 0;
45 int model = 0;
46
47 cpu_x86_version(env, &family, &model);
48 if ((family == 6 && model >= 14) || family > 6) {
49 return 1;
50 }
51
52 return 0;
53 }
54
55 /***********************************************************/
56 /* x86 debug */
57
58 static const char *cc_op_str[] = {
59 "DYNAMIC",
60 "EFLAGS",
61
62 "MULB",
63 "MULW",
64 "MULL",
65 "MULQ",
66
67 "ADDB",
68 "ADDW",
69 "ADDL",
70 "ADDQ",
71
72 "ADCB",
73 "ADCW",
74 "ADCL",
75 "ADCQ",
76
77 "SUBB",
78 "SUBW",
79 "SUBL",
80 "SUBQ",
81
82 "SBBB",
83 "SBBW",
84 "SBBL",
85 "SBBQ",
86
87 "LOGICB",
88 "LOGICW",
89 "LOGICL",
90 "LOGICQ",
91
92 "INCB",
93 "INCW",
94 "INCL",
95 "INCQ",
96
97 "DECB",
98 "DECW",
99 "DECL",
100 "DECQ",
101
102 "SHLB",
103 "SHLW",
104 "SHLL",
105 "SHLQ",
106
107 "SARB",
108 "SARW",
109 "SARL",
110 "SARQ",
111 };
112
113 static void
114 cpu_x86_dump_seg_cache(CPUX86State *env, FILE *f, fprintf_function cpu_fprintf,
115 const char *name, struct SegmentCache *sc)
116 {
117 #ifdef TARGET_X86_64
118 if (env->hflags & HF_CS64_MASK) {
119 cpu_fprintf(f, "%-3s=%04x %016" PRIx64 " %08x %08x", name,
120 sc->selector, sc->base, sc->limit, sc->flags & 0x00ffff00);
121 } else
122 #endif
123 {
124 cpu_fprintf(f, "%-3s=%04x %08x %08x %08x", name, sc->selector,
125 (uint32_t)sc->base, sc->limit, sc->flags & 0x00ffff00);
126 }
127
128 if (!(env->hflags & HF_PE_MASK) || !(sc->flags & DESC_P_MASK))
129 goto done;
130
131 cpu_fprintf(f, " DPL=%d ", (sc->flags & DESC_DPL_MASK) >> DESC_DPL_SHIFT);
132 if (sc->flags & DESC_S_MASK) {
133 if (sc->flags & DESC_CS_MASK) {
134 cpu_fprintf(f, (sc->flags & DESC_L_MASK) ? "CS64" :
135 ((sc->flags & DESC_B_MASK) ? "CS32" : "CS16"));
136 cpu_fprintf(f, " [%c%c", (sc->flags & DESC_C_MASK) ? 'C' : '-',
137 (sc->flags & DESC_R_MASK) ? 'R' : '-');
138 } else {
139 cpu_fprintf(f, (sc->flags & DESC_B_MASK) ? "DS " : "DS16");
140 cpu_fprintf(f, " [%c%c", (sc->flags & DESC_E_MASK) ? 'E' : '-',
141 (sc->flags & DESC_W_MASK) ? 'W' : '-');
142 }
143 cpu_fprintf(f, "%c]", (sc->flags & DESC_A_MASK) ? 'A' : '-');
144 } else {
145 static const char *sys_type_name[2][16] = {
146 { /* 32 bit mode */
147 "Reserved", "TSS16-avl", "LDT", "TSS16-busy",
148 "CallGate16", "TaskGate", "IntGate16", "TrapGate16",
149 "Reserved", "TSS32-avl", "Reserved", "TSS32-busy",
150 "CallGate32", "Reserved", "IntGate32", "TrapGate32"
151 },
152 { /* 64 bit mode */
153 "<hiword>", "Reserved", "LDT", "Reserved", "Reserved",
154 "Reserved", "Reserved", "Reserved", "Reserved",
155 "TSS64-avl", "Reserved", "TSS64-busy", "CallGate64",
156 "Reserved", "IntGate64", "TrapGate64"
157 }
158 };
159 cpu_fprintf(f, "%s",
160 sys_type_name[(env->hflags & HF_LMA_MASK) ? 1 : 0]
161 [(sc->flags & DESC_TYPE_MASK)
162 >> DESC_TYPE_SHIFT]);
163 }
164 done:
165 cpu_fprintf(f, "\n");
166 }
167
168 #define DUMP_CODE_BYTES_TOTAL 50
169 #define DUMP_CODE_BYTES_BACKWARD 20
170
171 void cpu_dump_state(CPUX86State *env, FILE *f, fprintf_function cpu_fprintf,
172 int flags)
173 {
174 int eflags, i, nb;
175 char cc_op_name[32];
176 static const char *seg_name[6] = { "ES", "CS", "SS", "DS", "FS", "GS" };
177
178 cpu_synchronize_state(env);
179
180 eflags = env->eflags;
181 #ifdef TARGET_X86_64
182 if (env->hflags & HF_CS64_MASK) {
183 cpu_fprintf(f,
184 "RAX=%016" PRIx64 " RBX=%016" PRIx64 " RCX=%016" PRIx64 " RDX=%016" PRIx64 "\n"
185 "RSI=%016" PRIx64 " RDI=%016" PRIx64 " RBP=%016" PRIx64 " RSP=%016" PRIx64 "\n"
186 "R8 =%016" PRIx64 " R9 =%016" PRIx64 " R10=%016" PRIx64 " R11=%016" PRIx64 "\n"
187 "R12=%016" PRIx64 " R13=%016" PRIx64 " R14=%016" PRIx64 " R15=%016" PRIx64 "\n"
188 "RIP=%016" PRIx64 " RFL=%08x [%c%c%c%c%c%c%c] CPL=%d II=%d A20=%d SMM=%d HLT=%d\n",
189 env->regs[R_EAX],
190 env->regs[R_EBX],
191 env->regs[R_ECX],
192 env->regs[R_EDX],
193 env->regs[R_ESI],
194 env->regs[R_EDI],
195 env->regs[R_EBP],
196 env->regs[R_ESP],
197 env->regs[8],
198 env->regs[9],
199 env->regs[10],
200 env->regs[11],
201 env->regs[12],
202 env->regs[13],
203 env->regs[14],
204 env->regs[15],
205 env->eip, eflags,
206 eflags & DF_MASK ? 'D' : '-',
207 eflags & CC_O ? 'O' : '-',
208 eflags & CC_S ? 'S' : '-',
209 eflags & CC_Z ? 'Z' : '-',
210 eflags & CC_A ? 'A' : '-',
211 eflags & CC_P ? 'P' : '-',
212 eflags & CC_C ? 'C' : '-',
213 env->hflags & HF_CPL_MASK,
214 (env->hflags >> HF_INHIBIT_IRQ_SHIFT) & 1,
215 (env->a20_mask >> 20) & 1,
216 (env->hflags >> HF_SMM_SHIFT) & 1,
217 env->halted);
218 } else
219 #endif
220 {
221 cpu_fprintf(f, "EAX=%08x EBX=%08x ECX=%08x EDX=%08x\n"
222 "ESI=%08x EDI=%08x EBP=%08x ESP=%08x\n"
223 "EIP=%08x EFL=%08x [%c%c%c%c%c%c%c] CPL=%d II=%d A20=%d SMM=%d HLT=%d\n",
224 (uint32_t)env->regs[R_EAX],
225 (uint32_t)env->regs[R_EBX],
226 (uint32_t)env->regs[R_ECX],
227 (uint32_t)env->regs[R_EDX],
228 (uint32_t)env->regs[R_ESI],
229 (uint32_t)env->regs[R_EDI],
230 (uint32_t)env->regs[R_EBP],
231 (uint32_t)env->regs[R_ESP],
232 (uint32_t)env->eip, eflags,
233 eflags & DF_MASK ? 'D' : '-',
234 eflags & CC_O ? 'O' : '-',
235 eflags & CC_S ? 'S' : '-',
236 eflags & CC_Z ? 'Z' : '-',
237 eflags & CC_A ? 'A' : '-',
238 eflags & CC_P ? 'P' : '-',
239 eflags & CC_C ? 'C' : '-',
240 env->hflags & HF_CPL_MASK,
241 (env->hflags >> HF_INHIBIT_IRQ_SHIFT) & 1,
242 (env->a20_mask >> 20) & 1,
243 (env->hflags >> HF_SMM_SHIFT) & 1,
244 env->halted);
245 }
246
247 for(i = 0; i < 6; i++) {
248 cpu_x86_dump_seg_cache(env, f, cpu_fprintf, seg_name[i],
249 &env->segs[i]);
250 }
251 cpu_x86_dump_seg_cache(env, f, cpu_fprintf, "LDT", &env->ldt);
252 cpu_x86_dump_seg_cache(env, f, cpu_fprintf, "TR", &env->tr);
253
254 #ifdef TARGET_X86_64
255 if (env->hflags & HF_LMA_MASK) {
256 cpu_fprintf(f, "GDT= %016" PRIx64 " %08x\n",
257 env->gdt.base, env->gdt.limit);
258 cpu_fprintf(f, "IDT= %016" PRIx64 " %08x\n",
259 env->idt.base, env->idt.limit);
260 cpu_fprintf(f, "CR0=%08x CR2=%016" PRIx64 " CR3=%016" PRIx64 " CR4=%08x\n",
261 (uint32_t)env->cr[0],
262 env->cr[2],
263 env->cr[3],
264 (uint32_t)env->cr[4]);
265 for(i = 0; i < 4; i++)
266 cpu_fprintf(f, "DR%d=%016" PRIx64 " ", i, env->dr[i]);
267 cpu_fprintf(f, "\nDR6=%016" PRIx64 " DR7=%016" PRIx64 "\n",
268 env->dr[6], env->dr[7]);
269 } else
270 #endif
271 {
272 cpu_fprintf(f, "GDT= %08x %08x\n",
273 (uint32_t)env->gdt.base, env->gdt.limit);
274 cpu_fprintf(f, "IDT= %08x %08x\n",
275 (uint32_t)env->idt.base, env->idt.limit);
276 cpu_fprintf(f, "CR0=%08x CR2=%08x CR3=%08x CR4=%08x\n",
277 (uint32_t)env->cr[0],
278 (uint32_t)env->cr[2],
279 (uint32_t)env->cr[3],
280 (uint32_t)env->cr[4]);
281 for(i = 0; i < 4; i++) {
282 cpu_fprintf(f, "DR%d=" TARGET_FMT_lx " ", i, env->dr[i]);
283 }
284 cpu_fprintf(f, "\nDR6=" TARGET_FMT_lx " DR7=" TARGET_FMT_lx "\n",
285 env->dr[6], env->dr[7]);
286 }
287 if (flags & X86_DUMP_CCOP) {
288 if ((unsigned)env->cc_op < CC_OP_NB)
289 snprintf(cc_op_name, sizeof(cc_op_name), "%s", cc_op_str[env->cc_op]);
290 else
291 snprintf(cc_op_name, sizeof(cc_op_name), "[%d]", env->cc_op);
292 #ifdef TARGET_X86_64
293 if (env->hflags & HF_CS64_MASK) {
294 cpu_fprintf(f, "CCS=%016" PRIx64 " CCD=%016" PRIx64 " CCO=%-8s\n",
295 env->cc_src, env->cc_dst,
296 cc_op_name);
297 } else
298 #endif
299 {
300 cpu_fprintf(f, "CCS=%08x CCD=%08x CCO=%-8s\n",
301 (uint32_t)env->cc_src, (uint32_t)env->cc_dst,
302 cc_op_name);
303 }
304 }
305 cpu_fprintf(f, "EFER=%016" PRIx64 "\n", env->efer);
306 if (flags & X86_DUMP_FPU) {
307 int fptag;
308 fptag = 0;
309 for(i = 0; i < 8; i++) {
310 fptag |= ((!env->fptags[i]) << i);
311 }
312 cpu_fprintf(f, "FCW=%04x FSW=%04x [ST=%d] FTW=%02x MXCSR=%08x\n",
313 env->fpuc,
314 (env->fpus & ~0x3800) | (env->fpstt & 0x7) << 11,
315 env->fpstt,
316 fptag,
317 env->mxcsr);
318 for(i=0;i<8;i++) {
319 CPU_LDoubleU u;
320 u.d = env->fpregs[i].d;
321 cpu_fprintf(f, "FPR%d=%016" PRIx64 " %04x",
322 i, u.l.lower, u.l.upper);
323 if ((i & 1) == 1)
324 cpu_fprintf(f, "\n");
325 else
326 cpu_fprintf(f, " ");
327 }
328 if (env->hflags & HF_CS64_MASK)
329 nb = 16;
330 else
331 nb = 8;
332 for(i=0;i<nb;i++) {
333 cpu_fprintf(f, "XMM%02d=%08x%08x%08x%08x",
334 i,
335 env->xmm_regs[i].XMM_L(3),
336 env->xmm_regs[i].XMM_L(2),
337 env->xmm_regs[i].XMM_L(1),
338 env->xmm_regs[i].XMM_L(0));
339 if ((i & 1) == 1)
340 cpu_fprintf(f, "\n");
341 else
342 cpu_fprintf(f, " ");
343 }
344 }
345 if (flags & CPU_DUMP_CODE) {
346 target_ulong base = env->segs[R_CS].base + env->eip;
347 target_ulong offs = MIN(env->eip, DUMP_CODE_BYTES_BACKWARD);
348 uint8_t code;
349 char codestr[3];
350
351 cpu_fprintf(f, "Code=");
352 for (i = 0; i < DUMP_CODE_BYTES_TOTAL; i++) {
353 if (cpu_memory_rw_debug(env, base - offs + i, &code, 1, 0) == 0) {
354 snprintf(codestr, sizeof(codestr), "%02x", code);
355 } else {
356 snprintf(codestr, sizeof(codestr), "??");
357 }
358 cpu_fprintf(f, "%s%s%s%s", i > 0 ? " " : "",
359 i == offs ? "<" : "", codestr, i == offs ? ">" : "");
360 }
361 cpu_fprintf(f, "\n");
362 }
363 }
364
365 /***********************************************************/
366 /* x86 mmu */
367 /* XXX: add PGE support */
368
369 void cpu_x86_set_a20(CPUX86State *env, int a20_state)
370 {
371 a20_state = (a20_state != 0);
372 if (a20_state != ((env->a20_mask >> 20) & 1)) {
373 #if defined(DEBUG_MMU)
374 printf("A20 update: a20=%d\n", a20_state);
375 #endif
376 /* if the cpu is currently executing code, we must unlink it and
377 all the potentially executing TB */
378 cpu_interrupt(env, CPU_INTERRUPT_EXITTB);
379
380 /* when a20 is changed, all the MMU mappings are invalid, so
381 we must flush everything */
382 tlb_flush(env, 1);
383 env->a20_mask = ~(1 << 20) | (a20_state << 20);
384 }
385 }
386
387 void cpu_x86_update_cr0(CPUX86State *env, uint32_t new_cr0)
388 {
389 int pe_state;
390
391 #if defined(DEBUG_MMU)
392 printf("CR0 update: CR0=0x%08x\n", new_cr0);
393 #endif
394 if ((new_cr0 & (CR0_PG_MASK | CR0_WP_MASK | CR0_PE_MASK)) !=
395 (env->cr[0] & (CR0_PG_MASK | CR0_WP_MASK | CR0_PE_MASK))) {
396 tlb_flush(env, 1);
397 }
398
399 #ifdef TARGET_X86_64
400 if (!(env->cr[0] & CR0_PG_MASK) && (new_cr0 & CR0_PG_MASK) &&
401 (env->efer & MSR_EFER_LME)) {
402 /* enter in long mode */
403 /* XXX: generate an exception */
404 if (!(env->cr[4] & CR4_PAE_MASK))
405 return;
406 env->efer |= MSR_EFER_LMA;
407 env->hflags |= HF_LMA_MASK;
408 } else if ((env->cr[0] & CR0_PG_MASK) && !(new_cr0 & CR0_PG_MASK) &&
409 (env->efer & MSR_EFER_LMA)) {
410 /* exit long mode */
411 env->efer &= ~MSR_EFER_LMA;
412 env->hflags &= ~(HF_LMA_MASK | HF_CS64_MASK);
413 env->eip &= 0xffffffff;
414 }
415 #endif
416 env->cr[0] = new_cr0 | CR0_ET_MASK;
417
418 /* update PE flag in hidden flags */
419 pe_state = (env->cr[0] & CR0_PE_MASK);
420 env->hflags = (env->hflags & ~HF_PE_MASK) | (pe_state << HF_PE_SHIFT);
421 /* ensure that ADDSEG is always set in real mode */
422 env->hflags |= ((pe_state ^ 1) << HF_ADDSEG_SHIFT);
423 /* update FPU flags */
424 env->hflags = (env->hflags & ~(HF_MP_MASK | HF_EM_MASK | HF_TS_MASK)) |
425 ((new_cr0 << (HF_MP_SHIFT - 1)) & (HF_MP_MASK | HF_EM_MASK | HF_TS_MASK));
426 }
427
428 /* XXX: in legacy PAE mode, generate a GPF if reserved bits are set in
429 the PDPT */
430 void cpu_x86_update_cr3(CPUX86State *env, target_ulong new_cr3)
431 {
432 env->cr[3] = new_cr3;
433 if (env->cr[0] & CR0_PG_MASK) {
434 #if defined(DEBUG_MMU)
435 printf("CR3 update: CR3=" TARGET_FMT_lx "\n", new_cr3);
436 #endif
437 tlb_flush(env, 0);
438 }
439 }
440
441 void cpu_x86_update_cr4(CPUX86State *env, uint32_t new_cr4)
442 {
443 #if defined(DEBUG_MMU)
444 printf("CR4 update: CR4=%08x\n", (uint32_t)env->cr[4]);
445 #endif
446 if ((new_cr4 & (CR4_PGE_MASK | CR4_PAE_MASK | CR4_PSE_MASK)) !=
447 (env->cr[4] & (CR4_PGE_MASK | CR4_PAE_MASK | CR4_PSE_MASK))) {
448 tlb_flush(env, 1);
449 }
450 /* SSE handling */
451 if (!(env->cpuid_features & CPUID_SSE))
452 new_cr4 &= ~CR4_OSFXSR_MASK;
453 if (new_cr4 & CR4_OSFXSR_MASK)
454 env->hflags |= HF_OSFXSR_MASK;
455 else
456 env->hflags &= ~HF_OSFXSR_MASK;
457
458 env->cr[4] = new_cr4;
459 }
460
461 #if defined(CONFIG_USER_ONLY)
462
463 int cpu_x86_handle_mmu_fault(CPUX86State *env, target_ulong addr,
464 int is_write, int mmu_idx)
465 {
466 /* user mode only emulation */
467 is_write &= 1;
468 env->cr[2] = addr;
469 env->error_code = (is_write << PG_ERROR_W_BIT);
470 env->error_code |= PG_ERROR_U_MASK;
471 env->exception_index = EXCP0E_PAGE;
472 return 1;
473 }
474
475 #else
476
477 /* XXX: This value should match the one returned by CPUID
478 * and in exec.c */
479 # if defined(TARGET_X86_64)
480 # define PHYS_ADDR_MASK 0xfffffff000LL
481 # else
482 # define PHYS_ADDR_MASK 0xffffff000LL
483 # endif
484
485 /* return value:
486 -1 = cannot handle fault
487 0 = nothing more to do
488 1 = generate PF fault
489 */
490 int cpu_x86_handle_mmu_fault(CPUX86State *env, target_ulong addr,
491 int is_write1, int mmu_idx)
492 {
493 uint64_t ptep, pte;
494 target_ulong pde_addr, pte_addr;
495 int error_code, is_dirty, prot, page_size, is_write, is_user;
496 target_phys_addr_t paddr;
497 uint32_t page_offset;
498 target_ulong vaddr, virt_addr;
499
500 is_user = mmu_idx == MMU_USER_IDX;
501 #if defined(DEBUG_MMU)
502 printf("MMU fault: addr=" TARGET_FMT_lx " w=%d u=%d eip=" TARGET_FMT_lx "\n",
503 addr, is_write1, is_user, env->eip);
504 #endif
505 is_write = is_write1 & 1;
506
507 if (!(env->cr[0] & CR0_PG_MASK)) {
508 pte = addr;
509 virt_addr = addr & TARGET_PAGE_MASK;
510 prot = PAGE_READ | PAGE_WRITE | PAGE_EXEC;
511 page_size = 4096;
512 goto do_mapping;
513 }
514
515 if (env->cr[4] & CR4_PAE_MASK) {
516 uint64_t pde, pdpe;
517 target_ulong pdpe_addr;
518
519 #ifdef TARGET_X86_64
520 if (env->hflags & HF_LMA_MASK) {
521 uint64_t pml4e_addr, pml4e;
522 int32_t sext;
523
524 /* test virtual address sign extension */
525 sext = (int64_t)addr >> 47;
526 if (sext != 0 && sext != -1) {
527 env->error_code = 0;
528 env->exception_index = EXCP0D_GPF;
529 return 1;
530 }
531
532 pml4e_addr = ((env->cr[3] & ~0xfff) + (((addr >> 39) & 0x1ff) << 3)) &
533 env->a20_mask;
534 pml4e = ldq_phys(pml4e_addr);
535 if (!(pml4e & PG_PRESENT_MASK)) {
536 error_code = 0;
537 goto do_fault;
538 }
539 if (!(env->efer & MSR_EFER_NXE) && (pml4e & PG_NX_MASK)) {
540 error_code = PG_ERROR_RSVD_MASK;
541 goto do_fault;
542 }
543 if (!(pml4e & PG_ACCESSED_MASK)) {
544 pml4e |= PG_ACCESSED_MASK;
545 stl_phys_notdirty(pml4e_addr, pml4e);
546 }
547 ptep = pml4e ^ PG_NX_MASK;
548 pdpe_addr = ((pml4e & PHYS_ADDR_MASK) + (((addr >> 30) & 0x1ff) << 3)) &
549 env->a20_mask;
550 pdpe = ldq_phys(pdpe_addr);
551 if (!(pdpe & PG_PRESENT_MASK)) {
552 error_code = 0;
553 goto do_fault;
554 }
555 if (!(env->efer & MSR_EFER_NXE) && (pdpe & PG_NX_MASK)) {
556 error_code = PG_ERROR_RSVD_MASK;
557 goto do_fault;
558 }
559 ptep &= pdpe ^ PG_NX_MASK;
560 if (!(pdpe & PG_ACCESSED_MASK)) {
561 pdpe |= PG_ACCESSED_MASK;
562 stl_phys_notdirty(pdpe_addr, pdpe);
563 }
564 } else
565 #endif
566 {
567 /* XXX: load them when cr3 is loaded ? */
568 pdpe_addr = ((env->cr[3] & ~0x1f) + ((addr >> 27) & 0x18)) &
569 env->a20_mask;
570 pdpe = ldq_phys(pdpe_addr);
571 if (!(pdpe & PG_PRESENT_MASK)) {
572 error_code = 0;
573 goto do_fault;
574 }
575 ptep = PG_NX_MASK | PG_USER_MASK | PG_RW_MASK;
576 }
577
578 pde_addr = ((pdpe & PHYS_ADDR_MASK) + (((addr >> 21) & 0x1ff) << 3)) &
579 env->a20_mask;
580 pde = ldq_phys(pde_addr);
581 if (!(pde & PG_PRESENT_MASK)) {
582 error_code = 0;
583 goto do_fault;
584 }
585 if (!(env->efer & MSR_EFER_NXE) && (pde & PG_NX_MASK)) {
586 error_code = PG_ERROR_RSVD_MASK;
587 goto do_fault;
588 }
589 ptep &= pde ^ PG_NX_MASK;
590 if (pde & PG_PSE_MASK) {
591 /* 2 MB page */
592 page_size = 2048 * 1024;
593 ptep ^= PG_NX_MASK;
594 if ((ptep & PG_NX_MASK) && is_write1 == 2)
595 goto do_fault_protect;
596 if (is_user) {
597 if (!(ptep & PG_USER_MASK))
598 goto do_fault_protect;
599 if (is_write && !(ptep & PG_RW_MASK))
600 goto do_fault_protect;
601 } else {
602 if ((env->cr[0] & CR0_WP_MASK) &&
603 is_write && !(ptep & PG_RW_MASK))
604 goto do_fault_protect;
605 }
606 is_dirty = is_write && !(pde & PG_DIRTY_MASK);
607 if (!(pde & PG_ACCESSED_MASK) || is_dirty) {
608 pde |= PG_ACCESSED_MASK;
609 if (is_dirty)
610 pde |= PG_DIRTY_MASK;
611 stl_phys_notdirty(pde_addr, pde);
612 }
613 /* align to page_size */
614 pte = pde & ((PHYS_ADDR_MASK & ~(page_size - 1)) | 0xfff);
615 virt_addr = addr & ~(page_size - 1);
616 } else {
617 /* 4 KB page */
618 if (!(pde & PG_ACCESSED_MASK)) {
619 pde |= PG_ACCESSED_MASK;
620 stl_phys_notdirty(pde_addr, pde);
621 }
622 pte_addr = ((pde & PHYS_ADDR_MASK) + (((addr >> 12) & 0x1ff) << 3)) &
623 env->a20_mask;
624 pte = ldq_phys(pte_addr);
625 if (!(pte & PG_PRESENT_MASK)) {
626 error_code = 0;
627 goto do_fault;
628 }
629 if (!(env->efer & MSR_EFER_NXE) && (pte & PG_NX_MASK)) {
630 error_code = PG_ERROR_RSVD_MASK;
631 goto do_fault;
632 }
633 /* combine pde and pte nx, user and rw protections */
634 ptep &= pte ^ PG_NX_MASK;
635 ptep ^= PG_NX_MASK;
636 if ((ptep & PG_NX_MASK) && is_write1 == 2)
637 goto do_fault_protect;
638 if (is_user) {
639 if (!(ptep & PG_USER_MASK))
640 goto do_fault_protect;
641 if (is_write && !(ptep & PG_RW_MASK))
642 goto do_fault_protect;
643 } else {
644 if ((env->cr[0] & CR0_WP_MASK) &&
645 is_write && !(ptep & PG_RW_MASK))
646 goto do_fault_protect;
647 }
648 is_dirty = is_write && !(pte & PG_DIRTY_MASK);
649 if (!(pte & PG_ACCESSED_MASK) || is_dirty) {
650 pte |= PG_ACCESSED_MASK;
651 if (is_dirty)
652 pte |= PG_DIRTY_MASK;
653 stl_phys_notdirty(pte_addr, pte);
654 }
655 page_size = 4096;
656 virt_addr = addr & ~0xfff;
657 pte = pte & (PHYS_ADDR_MASK | 0xfff);
658 }
659 } else {
660 uint32_t pde;
661
662 /* page directory entry */
663 pde_addr = ((env->cr[3] & ~0xfff) + ((addr >> 20) & 0xffc)) &
664 env->a20_mask;
665 pde = ldl_phys(pde_addr);
666 if (!(pde & PG_PRESENT_MASK)) {
667 error_code = 0;
668 goto do_fault;
669 }
670 /* if PSE bit is set, then we use a 4MB page */
671 if ((pde & PG_PSE_MASK) && (env->cr[4] & CR4_PSE_MASK)) {
672 page_size = 4096 * 1024;
673 if (is_user) {
674 if (!(pde & PG_USER_MASK))
675 goto do_fault_protect;
676 if (is_write && !(pde & PG_RW_MASK))
677 goto do_fault_protect;
678 } else {
679 if ((env->cr[0] & CR0_WP_MASK) &&
680 is_write && !(pde & PG_RW_MASK))
681 goto do_fault_protect;
682 }
683 is_dirty = is_write && !(pde & PG_DIRTY_MASK);
684 if (!(pde & PG_ACCESSED_MASK) || is_dirty) {
685 pde |= PG_ACCESSED_MASK;
686 if (is_dirty)
687 pde |= PG_DIRTY_MASK;
688 stl_phys_notdirty(pde_addr, pde);
689 }
690
691 pte = pde & ~( (page_size - 1) & ~0xfff); /* align to page_size */
692 ptep = pte;
693 virt_addr = addr & ~(page_size - 1);
694 } else {
695 if (!(pde & PG_ACCESSED_MASK)) {
696 pde |= PG_ACCESSED_MASK;
697 stl_phys_notdirty(pde_addr, pde);
698 }
699
700 /* page directory entry */
701 pte_addr = ((pde & ~0xfff) + ((addr >> 10) & 0xffc)) &
702 env->a20_mask;
703 pte = ldl_phys(pte_addr);
704 if (!(pte & PG_PRESENT_MASK)) {
705 error_code = 0;
706 goto do_fault;
707 }
708 /* combine pde and pte user and rw protections */
709 ptep = pte & pde;
710 if (is_user) {
711 if (!(ptep & PG_USER_MASK))
712 goto do_fault_protect;
713 if (is_write && !(ptep & PG_RW_MASK))
714 goto do_fault_protect;
715 } else {
716 if ((env->cr[0] & CR0_WP_MASK) &&
717 is_write && !(ptep & PG_RW_MASK))
718 goto do_fault_protect;
719 }
720 is_dirty = is_write && !(pte & PG_DIRTY_MASK);
721 if (!(pte & PG_ACCESSED_MASK) || is_dirty) {
722 pte |= PG_ACCESSED_MASK;
723 if (is_dirty)
724 pte |= PG_DIRTY_MASK;
725 stl_phys_notdirty(pte_addr, pte);
726 }
727 page_size = 4096;
728 virt_addr = addr & ~0xfff;
729 }
730 }
731 /* the page can be put in the TLB */
732 prot = PAGE_READ;
733 if (!(ptep & PG_NX_MASK))
734 prot |= PAGE_EXEC;
735 if (pte & PG_DIRTY_MASK) {
736 /* only set write access if already dirty... otherwise wait
737 for dirty access */
738 if (is_user) {
739 if (ptep & PG_RW_MASK)
740 prot |= PAGE_WRITE;
741 } else {
742 if (!(env->cr[0] & CR0_WP_MASK) ||
743 (ptep & PG_RW_MASK))
744 prot |= PAGE_WRITE;
745 }
746 }
747 do_mapping:
748 pte = pte & env->a20_mask;
749
750 /* Even if 4MB pages, we map only one 4KB page in the cache to
751 avoid filling it too fast */
752 page_offset = (addr & TARGET_PAGE_MASK) & (page_size - 1);
753 paddr = (pte & TARGET_PAGE_MASK) + page_offset;
754 vaddr = virt_addr + page_offset;
755
756 tlb_set_page(env, vaddr, paddr, prot, mmu_idx, page_size);
757 return 0;
758 do_fault_protect:
759 error_code = PG_ERROR_P_MASK;
760 do_fault:
761 error_code |= (is_write << PG_ERROR_W_BIT);
762 if (is_user)
763 error_code |= PG_ERROR_U_MASK;
764 if (is_write1 == 2 &&
765 (env->efer & MSR_EFER_NXE) &&
766 (env->cr[4] & CR4_PAE_MASK))
767 error_code |= PG_ERROR_I_D_MASK;
768 if (env->intercept_exceptions & (1 << EXCP0E_PAGE)) {
769 /* cr2 is not modified in case of exceptions */
770 stq_phys(env->vm_vmcb + offsetof(struct vmcb, control.exit_info_2),
771 addr);
772 } else {
773 env->cr[2] = addr;
774 }
775 env->error_code = error_code;
776 env->exception_index = EXCP0E_PAGE;
777 return 1;
778 }
779
780 target_phys_addr_t cpu_get_phys_page_debug(CPUX86State *env, target_ulong addr)
781 {
782 target_ulong pde_addr, pte_addr;
783 uint64_t pte;
784 target_phys_addr_t paddr;
785 uint32_t page_offset;
786 int page_size;
787
788 if (env->cr[4] & CR4_PAE_MASK) {
789 target_ulong pdpe_addr;
790 uint64_t pde, pdpe;
791
792 #ifdef TARGET_X86_64
793 if (env->hflags & HF_LMA_MASK) {
794 uint64_t pml4e_addr, pml4e;
795 int32_t sext;
796
797 /* test virtual address sign extension */
798 sext = (int64_t)addr >> 47;
799 if (sext != 0 && sext != -1)
800 return -1;
801
802 pml4e_addr = ((env->cr[3] & ~0xfff) + (((addr >> 39) & 0x1ff) << 3)) &
803 env->a20_mask;
804 pml4e = ldq_phys(pml4e_addr);
805 if (!(pml4e & PG_PRESENT_MASK))
806 return -1;
807
808 pdpe_addr = ((pml4e & ~0xfff & ~(PG_NX_MASK | PG_HI_USER_MASK)) +
809 (((addr >> 30) & 0x1ff) << 3)) & env->a20_mask;
810 pdpe = ldq_phys(pdpe_addr);
811 if (!(pdpe & PG_PRESENT_MASK))
812 return -1;
813 } else
814 #endif
815 {
816 pdpe_addr = ((env->cr[3] & ~0x1f) + ((addr >> 27) & 0x18)) &
817 env->a20_mask;
818 pdpe = ldq_phys(pdpe_addr);
819 if (!(pdpe & PG_PRESENT_MASK))
820 return -1;
821 }
822
823 pde_addr = ((pdpe & ~0xfff & ~(PG_NX_MASK | PG_HI_USER_MASK)) +
824 (((addr >> 21) & 0x1ff) << 3)) & env->a20_mask;
825 pde = ldq_phys(pde_addr);
826 if (!(pde & PG_PRESENT_MASK)) {
827 return -1;
828 }
829 if (pde & PG_PSE_MASK) {
830 /* 2 MB page */
831 page_size = 2048 * 1024;
832 pte = pde & ~( (page_size - 1) & ~0xfff); /* align to page_size */
833 } else {
834 /* 4 KB page */
835 pte_addr = ((pde & ~0xfff & ~(PG_NX_MASK | PG_HI_USER_MASK)) +
836 (((addr >> 12) & 0x1ff) << 3)) & env->a20_mask;
837 page_size = 4096;
838 pte = ldq_phys(pte_addr);
839 }
840 pte &= ~(PG_NX_MASK | PG_HI_USER_MASK);
841 if (!(pte & PG_PRESENT_MASK))
842 return -1;
843 } else {
844 uint32_t pde;
845
846 if (!(env->cr[0] & CR0_PG_MASK)) {
847 pte = addr;
848 page_size = 4096;
849 } else {
850 /* page directory entry */
851 pde_addr = ((env->cr[3] & ~0xfff) + ((addr >> 20) & 0xffc)) & env->a20_mask;
852 pde = ldl_phys(pde_addr);
853 if (!(pde & PG_PRESENT_MASK))
854 return -1;
855 if ((pde & PG_PSE_MASK) && (env->cr[4] & CR4_PSE_MASK)) {
856 pte = pde & ~0x003ff000; /* align to 4MB */
857 page_size = 4096 * 1024;
858 } else {
859 /* page directory entry */
860 pte_addr = ((pde & ~0xfff) + ((addr >> 10) & 0xffc)) & env->a20_mask;
861 pte = ldl_phys(pte_addr);
862 if (!(pte & PG_PRESENT_MASK))
863 return -1;
864 page_size = 4096;
865 }
866 }
867 pte = pte & env->a20_mask;
868 }
869
870 page_offset = (addr & TARGET_PAGE_MASK) & (page_size - 1);
871 paddr = (pte & TARGET_PAGE_MASK) + page_offset;
872 return paddr;
873 }
874
875 void hw_breakpoint_insert(CPUX86State *env, int index)
876 {
877 int type, err = 0;
878
879 switch (hw_breakpoint_type(env->dr[7], index)) {
880 case 0:
881 if (hw_breakpoint_enabled(env->dr[7], index))
882 err = cpu_breakpoint_insert(env, env->dr[index], BP_CPU,
883 &env->cpu_breakpoint[index]);
884 break;
885 case 1:
886 type = BP_CPU | BP_MEM_WRITE;
887 goto insert_wp;
888 case 2:
889 /* No support for I/O watchpoints yet */
890 break;
891 case 3:
892 type = BP_CPU | BP_MEM_ACCESS;
893 insert_wp:
894 err = cpu_watchpoint_insert(env, env->dr[index],
895 hw_breakpoint_len(env->dr[7], index),
896 type, &env->cpu_watchpoint[index]);
897 break;
898 }
899 if (err)
900 env->cpu_breakpoint[index] = NULL;
901 }
902
903 void hw_breakpoint_remove(CPUX86State *env, int index)
904 {
905 if (!env->cpu_breakpoint[index])
906 return;
907 switch (hw_breakpoint_type(env->dr[7], index)) {
908 case 0:
909 if (hw_breakpoint_enabled(env->dr[7], index))
910 cpu_breakpoint_remove_by_ref(env, env->cpu_breakpoint[index]);
911 break;
912 case 1:
913 case 3:
914 cpu_watchpoint_remove_by_ref(env, env->cpu_watchpoint[index]);
915 break;
916 case 2:
917 /* No support for I/O watchpoints yet */
918 break;
919 }
920 }
921
922 int check_hw_breakpoints(CPUX86State *env, int force_dr6_update)
923 {
924 target_ulong dr6;
925 int reg, type;
926 int hit_enabled = 0;
927
928 dr6 = env->dr[6] & ~0xf;
929 for (reg = 0; reg < 4; reg++) {
930 type = hw_breakpoint_type(env->dr[7], reg);
931 if ((type == 0 && env->dr[reg] == env->eip) ||
932 ((type & 1) && env->cpu_watchpoint[reg] &&
933 (env->cpu_watchpoint[reg]->flags & BP_WATCHPOINT_HIT))) {
934 dr6 |= 1 << reg;
935 if (hw_breakpoint_enabled(env->dr[7], reg))
936 hit_enabled = 1;
937 }
938 }
939 if (hit_enabled || force_dr6_update)
940 env->dr[6] = dr6;
941 return hit_enabled;
942 }
943
944 static CPUDebugExcpHandler *prev_debug_excp_handler;
945
946 static void breakpoint_handler(CPUX86State *env)
947 {
948 CPUBreakpoint *bp;
949
950 if (env->watchpoint_hit) {
951 if (env->watchpoint_hit->flags & BP_CPU) {
952 env->watchpoint_hit = NULL;
953 if (check_hw_breakpoints(env, 0))
954 raise_exception(env, EXCP01_DB);
955 else
956 cpu_resume_from_signal(env, NULL);
957 }
958 } else {
959 QTAILQ_FOREACH(bp, &env->breakpoints, entry)
960 if (bp->pc == env->eip) {
961 if (bp->flags & BP_CPU) {
962 check_hw_breakpoints(env, 1);
963 raise_exception(env, EXCP01_DB);
964 }
965 break;
966 }
967 }
968 if (prev_debug_excp_handler)
969 prev_debug_excp_handler(env);
970 }
971
972 typedef struct MCEInjectionParams {
973 Monitor *mon;
974 CPUX86State *env;
975 int bank;
976 uint64_t status;
977 uint64_t mcg_status;
978 uint64_t addr;
979 uint64_t misc;
980 int flags;
981 } MCEInjectionParams;
982
983 static void do_inject_x86_mce(void *data)
984 {
985 MCEInjectionParams *params = data;
986 CPUX86State *cenv = params->env;
987 uint64_t *banks = cenv->mce_banks + 4 * params->bank;
988
989 cpu_synchronize_state(cenv);
990
991 /*
992 * If there is an MCE exception being processed, ignore this SRAO MCE
993 * unless unconditional injection was requested.
994 */
995 if (!(params->flags & MCE_INJECT_UNCOND_AO)
996 && !(params->status & MCI_STATUS_AR)
997 && (cenv->mcg_status & MCG_STATUS_MCIP)) {
998 return;
999 }
1000
1001 if (params->status & MCI_STATUS_UC) {
1002 /*
1003 * if MSR_MCG_CTL is not all 1s, the uncorrected error
1004 * reporting is disabled
1005 */
1006 if ((cenv->mcg_cap & MCG_CTL_P) && cenv->mcg_ctl != ~(uint64_t)0) {
1007 monitor_printf(params->mon,
1008 "CPU %d: Uncorrected error reporting disabled\n",
1009 cenv->cpu_index);
1010 return;
1011 }
1012
1013 /*
1014 * if MSR_MCi_CTL is not all 1s, the uncorrected error
1015 * reporting is disabled for the bank
1016 */
1017 if (banks[0] != ~(uint64_t)0) {
1018 monitor_printf(params->mon,
1019 "CPU %d: Uncorrected error reporting disabled for"
1020 " bank %d\n",
1021 cenv->cpu_index, params->bank);
1022 return;
1023 }
1024
1025 if ((cenv->mcg_status & MCG_STATUS_MCIP) ||
1026 !(cenv->cr[4] & CR4_MCE_MASK)) {
1027 monitor_printf(params->mon,
1028 "CPU %d: Previous MCE still in progress, raising"
1029 " triple fault\n",
1030 cenv->cpu_index);
1031 qemu_log_mask(CPU_LOG_RESET, "Triple fault\n");
1032 qemu_system_reset_request();
1033 return;
1034 }
1035 if (banks[1] & MCI_STATUS_VAL) {
1036 params->status |= MCI_STATUS_OVER;
1037 }
1038 banks[2] = params->addr;
1039 banks[3] = params->misc;
1040 cenv->mcg_status = params->mcg_status;
1041 banks[1] = params->status;
1042 cpu_interrupt(cenv, CPU_INTERRUPT_MCE);
1043 } else if (!(banks[1] & MCI_STATUS_VAL)
1044 || !(banks[1] & MCI_STATUS_UC)) {
1045 if (banks[1] & MCI_STATUS_VAL) {
1046 params->status |= MCI_STATUS_OVER;
1047 }
1048 banks[2] = params->addr;
1049 banks[3] = params->misc;
1050 banks[1] = params->status;
1051 } else {
1052 banks[1] |= MCI_STATUS_OVER;
1053 }
1054 }
1055
1056 void cpu_x86_inject_mce(Monitor *mon, CPUX86State *cenv, int bank,
1057 uint64_t status, uint64_t mcg_status, uint64_t addr,
1058 uint64_t misc, int flags)
1059 {
1060 MCEInjectionParams params = {
1061 .mon = mon,
1062 .env = cenv,
1063 .bank = bank,
1064 .status = status,
1065 .mcg_status = mcg_status,
1066 .addr = addr,
1067 .misc = misc,
1068 .flags = flags,
1069 };
1070 unsigned bank_num = cenv->mcg_cap & 0xff;
1071 CPUX86State *env;
1072
1073 if (!cenv->mcg_cap) {
1074 monitor_printf(mon, "MCE injection not supported\n");
1075 return;
1076 }
1077 if (bank >= bank_num) {
1078 monitor_printf(mon, "Invalid MCE bank number\n");
1079 return;
1080 }
1081 if (!(status & MCI_STATUS_VAL)) {
1082 monitor_printf(mon, "Invalid MCE status code\n");
1083 return;
1084 }
1085 if ((flags & MCE_INJECT_BROADCAST)
1086 && !cpu_x86_support_mca_broadcast(cenv)) {
1087 monitor_printf(mon, "Guest CPU does not support MCA broadcast\n");
1088 return;
1089 }
1090
1091 run_on_cpu(cenv, do_inject_x86_mce, &params);
1092 if (flags & MCE_INJECT_BROADCAST) {
1093 params.bank = 1;
1094 params.status = MCI_STATUS_VAL | MCI_STATUS_UC;
1095 params.mcg_status = MCG_STATUS_MCIP | MCG_STATUS_RIPV;
1096 params.addr = 0;
1097 params.misc = 0;
1098 for (env = first_cpu; env != NULL; env = env->next_cpu) {
1099 if (cenv == env) {
1100 continue;
1101 }
1102 params.env = env;
1103 run_on_cpu(cenv, do_inject_x86_mce, &params);
1104 }
1105 }
1106 }
1107
1108 void cpu_report_tpr_access(CPUX86State *env, TPRAccess access)
1109 {
1110 TranslationBlock *tb;
1111
1112 if (kvm_enabled()) {
1113 env->tpr_access_type = access;
1114
1115 cpu_interrupt(env, CPU_INTERRUPT_TPR);
1116 } else {
1117 tb = tb_find_pc(env->mem_io_pc);
1118 cpu_restore_state(tb, env, env->mem_io_pc);
1119
1120 apic_handle_tpr_access_report(env->apic_state, env->eip, access);
1121 }
1122 }
1123 #endif /* !CONFIG_USER_ONLY */
1124
1125 int cpu_x86_get_descr_debug(CPUX86State *env, unsigned int selector,
1126 target_ulong *base, unsigned int *limit,
1127 unsigned int *flags)
1128 {
1129 SegmentCache *dt;
1130 target_ulong ptr;
1131 uint32_t e1, e2;
1132 int index;
1133
1134 if (selector & 0x4)
1135 dt = &env->ldt;
1136 else
1137 dt = &env->gdt;
1138 index = selector & ~7;
1139 ptr = dt->base + index;
1140 if ((index + 7) > dt->limit
1141 || cpu_memory_rw_debug(env, ptr, (uint8_t *)&e1, sizeof(e1), 0) != 0
1142 || cpu_memory_rw_debug(env, ptr+4, (uint8_t *)&e2, sizeof(e2), 0) != 0)
1143 return 0;
1144
1145 *base = ((e1 >> 16) | ((e2 & 0xff) << 16) | (e2 & 0xff000000));
1146 *limit = (e1 & 0xffff) | (e2 & 0x000f0000);
1147 if (e2 & DESC_G_MASK)
1148 *limit = (*limit << 12) | 0xfff;
1149 *flags = e2;
1150
1151 return 1;
1152 }
1153
1154 X86CPU *cpu_x86_init(const char *cpu_model)
1155 {
1156 X86CPU *cpu;
1157 CPUX86State *env;
1158 static int inited;
1159
1160 cpu = X86_CPU(object_new(TYPE_X86_CPU));
1161 env = &cpu->env;
1162 env->cpu_model_str = cpu_model;
1163
1164 /* init various static tables used in TCG mode */
1165 if (tcg_enabled() && !inited) {
1166 inited = 1;
1167 optimize_flags_init();
1168 #ifndef CONFIG_USER_ONLY
1169 prev_debug_excp_handler =
1170 cpu_set_debug_excp_handler(breakpoint_handler);
1171 #endif
1172 }
1173 if (cpu_x86_register(cpu, cpu_model) < 0) {
1174 object_delete(OBJECT(cpu));
1175 return NULL;
1176 }
1177
1178 x86_cpu_realize(OBJECT(cpu), NULL);
1179
1180 return cpu;
1181 }
1182
1183 #if !defined(CONFIG_USER_ONLY)
1184 void do_cpu_init(X86CPU *cpu)
1185 {
1186 CPUX86State *env = &cpu->env;
1187 int sipi = env->interrupt_request & CPU_INTERRUPT_SIPI;
1188 uint64_t pat = env->pat;
1189
1190 cpu_reset(CPU(cpu));
1191 env->interrupt_request = sipi;
1192 env->pat = pat;
1193 apic_init_reset(env->apic_state);
1194 }
1195
1196 void do_cpu_sipi(X86CPU *cpu)
1197 {
1198 CPUX86State *env = &cpu->env;
1199
1200 apic_sipi(env->apic_state);
1201 }
1202 #else
1203 void do_cpu_init(X86CPU *cpu)
1204 {
1205 }
1206 void do_cpu_sipi(X86CPU *cpu)
1207 {
1208 }
1209 #endif
Qemu copy for testing