kvm: Pass CPUState to kvm_arch_*
[qemu/opensuse.git] / target-i386 / helper.c
blob00341c523363a4c0c90e35d717113b865dce60e3
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 "kvm.h"
22 #ifndef CONFIG_USER_ONLY
23 #include "sysemu.h"
24 #include "monitor.h"
25 #endif
27 //#define DEBUG_MMU
29 static void cpu_x86_version(CPUX86State *env, int *family, int *model)
31 int cpuver = env->cpuid_version;
33 if (family == NULL || model == NULL) {
34 return;
37 *family = (cpuver >> 8) & 0x0f;
38 *model = ((cpuver >> 12) & 0xf0) + ((cpuver >> 4) & 0x0f);
41 /* Broadcast MCA signal for processor version 06H_EH and above */
42 int cpu_x86_support_mca_broadcast(CPUX86State *env)
44 int family = 0;
45 int model = 0;
47 cpu_x86_version(env, &family, &model);
48 if ((family == 6 && model >= 14) || family > 6) {
49 return 1;
52 return 0;
55 /***********************************************************/
56 /* x86 debug */
58 static const char *cc_op_str[] = {
59 "DYNAMIC",
60 "EFLAGS",
62 "MULB",
63 "MULW",
64 "MULL",
65 "MULQ",
67 "ADDB",
68 "ADDW",
69 "ADDL",
70 "ADDQ",
72 "ADCB",
73 "ADCW",
74 "ADCL",
75 "ADCQ",
77 "SUBB",
78 "SUBW",
79 "SUBL",
80 "SUBQ",
82 "SBBB",
83 "SBBW",
84 "SBBL",
85 "SBBQ",
87 "LOGICB",
88 "LOGICW",
89 "LOGICL",
90 "LOGICQ",
92 "INCB",
93 "INCW",
94 "INCL",
95 "INCQ",
97 "DECB",
98 "DECW",
99 "DECL",
100 "DECQ",
102 "SHLB",
103 "SHLW",
104 "SHLL",
105 "SHLQ",
107 "SARB",
108 "SARW",
109 "SARL",
110 "SARQ",
113 static void
114 cpu_x86_dump_seg_cache(CPUX86State *env, FILE *f, fprintf_function cpu_fprintf,
115 const char *name, struct SegmentCache *sc)
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
124 cpu_fprintf(f, "%-3s=%04x %08x %08x %08x", name, sc->selector,
125 (uint32_t)sc->base, sc->limit, sc->flags & 0x00ffff00);
128 if (!(env->hflags & HF_PE_MASK) || !(sc->flags & DESC_P_MASK))
129 goto done;
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' : '-');
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"
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"
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]);
164 done:
165 cpu_fprintf(f, "\n");
168 #define DUMP_CODE_BYTES_TOTAL 50
169 #define DUMP_CODE_BYTES_BACKWARD 20
171 void cpu_dump_state(CPUX86State *env, FILE *f, fprintf_function cpu_fprintf,
172 int flags)
174 int eflags, i, nb;
175 char cc_op_name[32];
176 static const char *seg_name[6] = { "ES", "CS", "SS", "DS", "FS", "GS" };
178 cpu_synchronize_state(env);
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
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);
247 for(i = 0; i < 6; i++) {
248 cpu_x86_dump_seg_cache(env, f, cpu_fprintf, seg_name[i],
249 &env->segs[i]);
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);
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
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]);
284 cpu_fprintf(f, "\nDR6=" TARGET_FMT_lx " DR7=" TARGET_FMT_lx "\n",
285 env->dr[6], env->dr[7]);
287 if (flags & CPU_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
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);
305 cpu_fprintf(f, "EFER=%016" PRIx64 "\n", env->efer);
306 if (flags & CPU_DUMP_FPU) {
307 int fptag;
308 fptag = 0;
309 for(i = 0; i < 8; i++) {
310 fptag |= ((!env->fptags[i]) << i);
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, " ");
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",
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, " ");
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];
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), "??");
358 cpu_fprintf(f, "%s%s%s%s", i > 0 ? " " : "",
359 i == offs ? "<" : "", codestr, i == offs ? ">" : "");
361 cpu_fprintf(f, "\n");
365 /***********************************************************/
366 /* x86 mmu */
367 /* XXX: add PGE support */
369 void cpu_x86_set_a20(CPUX86State *env, int a20_state)
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);
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);
387 void cpu_x86_update_cr0(CPUX86State *env, uint32_t new_cr0)
389 int pe_state;
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);
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;
415 #endif
416 env->cr[0] = new_cr0 | CR0_ET_MASK;
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));
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)
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);
441 void cpu_x86_update_cr4(CPUX86State *env, uint32_t new_cr4)
443 #if defined(DEBUG_MMU)
444 printf("CR4 update: CR4=%08x\n", (uint32_t)env->cr[4]);
445 #endif
446 if ((new_cr4 ^ env->cr[4]) &
447 (CR4_PGE_MASK | CR4_PAE_MASK | CR4_PSE_MASK |
448 CR4_SMEP_MASK | CR4_SMAP_MASK)) {
449 tlb_flush(env, 1);
451 /* SSE handling */
452 if (!(env->cpuid_features & CPUID_SSE)) {
453 new_cr4 &= ~CR4_OSFXSR_MASK;
455 env->hflags &= ~HF_OSFXSR_MASK;
456 if (new_cr4 & CR4_OSFXSR_MASK) {
457 env->hflags |= HF_OSFXSR_MASK;
460 if (!(env->cpuid_7_0_ebx_features & CPUID_7_0_EBX_SMAP)) {
461 new_cr4 &= ~CR4_SMAP_MASK;
463 env->hflags &= ~HF_SMAP_MASK;
464 if (new_cr4 & CR4_SMAP_MASK) {
465 env->hflags |= HF_SMAP_MASK;
468 env->cr[4] = new_cr4;
471 #if defined(CONFIG_USER_ONLY)
473 int cpu_x86_handle_mmu_fault(CPUX86State *env, target_ulong addr,
474 int is_write, int mmu_idx)
476 /* user mode only emulation */
477 is_write &= 1;
478 env->cr[2] = addr;
479 env->error_code = (is_write << PG_ERROR_W_BIT);
480 env->error_code |= PG_ERROR_U_MASK;
481 env->exception_index = EXCP0E_PAGE;
482 return 1;
485 #else
487 /* XXX: This value should match the one returned by CPUID
488 * and in exec.c */
489 # if defined(TARGET_X86_64)
490 # define PHYS_ADDR_MASK 0xfffffff000LL
491 # else
492 # define PHYS_ADDR_MASK 0xffffff000LL
493 # endif
495 /* return value:
496 -1 = cannot handle fault
497 0 = nothing more to do
498 1 = generate PF fault
500 int cpu_x86_handle_mmu_fault(CPUX86State *env, target_ulong addr,
501 int is_write1, int mmu_idx)
503 uint64_t ptep, pte;
504 target_ulong pde_addr, pte_addr;
505 int error_code, is_dirty, prot, page_size, is_write, is_user;
506 hwaddr paddr;
507 uint32_t page_offset;
508 target_ulong vaddr, virt_addr;
510 is_user = mmu_idx == MMU_USER_IDX;
511 #if defined(DEBUG_MMU)
512 printf("MMU fault: addr=" TARGET_FMT_lx " w=%d u=%d eip=" TARGET_FMT_lx "\n",
513 addr, is_write1, is_user, env->eip);
514 #endif
515 is_write = is_write1 & 1;
517 if (!(env->cr[0] & CR0_PG_MASK)) {
518 pte = addr;
519 virt_addr = addr & TARGET_PAGE_MASK;
520 prot = PAGE_READ | PAGE_WRITE | PAGE_EXEC;
521 page_size = 4096;
522 goto do_mapping;
525 if (env->cr[4] & CR4_PAE_MASK) {
526 uint64_t pde, pdpe;
527 target_ulong pdpe_addr;
529 #ifdef TARGET_X86_64
530 if (env->hflags & HF_LMA_MASK) {
531 uint64_t pml4e_addr, pml4e;
532 int32_t sext;
534 /* test virtual address sign extension */
535 sext = (int64_t)addr >> 47;
536 if (sext != 0 && sext != -1) {
537 env->error_code = 0;
538 env->exception_index = EXCP0D_GPF;
539 return 1;
542 pml4e_addr = ((env->cr[3] & ~0xfff) + (((addr >> 39) & 0x1ff) << 3)) &
543 env->a20_mask;
544 pml4e = ldq_phys(pml4e_addr);
545 if (!(pml4e & PG_PRESENT_MASK)) {
546 error_code = 0;
547 goto do_fault;
549 if (!(env->efer & MSR_EFER_NXE) && (pml4e & PG_NX_MASK)) {
550 error_code = PG_ERROR_RSVD_MASK;
551 goto do_fault;
553 if (!(pml4e & PG_ACCESSED_MASK)) {
554 pml4e |= PG_ACCESSED_MASK;
555 stl_phys_notdirty(pml4e_addr, pml4e);
557 ptep = pml4e ^ PG_NX_MASK;
558 pdpe_addr = ((pml4e & PHYS_ADDR_MASK) + (((addr >> 30) & 0x1ff) << 3)) &
559 env->a20_mask;
560 pdpe = ldq_phys(pdpe_addr);
561 if (!(pdpe & PG_PRESENT_MASK)) {
562 error_code = 0;
563 goto do_fault;
565 if (!(env->efer & MSR_EFER_NXE) && (pdpe & PG_NX_MASK)) {
566 error_code = PG_ERROR_RSVD_MASK;
567 goto do_fault;
569 ptep &= pdpe ^ PG_NX_MASK;
570 if (!(pdpe & PG_ACCESSED_MASK)) {
571 pdpe |= PG_ACCESSED_MASK;
572 stl_phys_notdirty(pdpe_addr, pdpe);
574 } else
575 #endif
577 /* XXX: load them when cr3 is loaded ? */
578 pdpe_addr = ((env->cr[3] & ~0x1f) + ((addr >> 27) & 0x18)) &
579 env->a20_mask;
580 pdpe = ldq_phys(pdpe_addr);
581 if (!(pdpe & PG_PRESENT_MASK)) {
582 error_code = 0;
583 goto do_fault;
585 ptep = PG_NX_MASK | PG_USER_MASK | PG_RW_MASK;
588 pde_addr = ((pdpe & PHYS_ADDR_MASK) + (((addr >> 21) & 0x1ff) << 3)) &
589 env->a20_mask;
590 pde = ldq_phys(pde_addr);
591 if (!(pde & PG_PRESENT_MASK)) {
592 error_code = 0;
593 goto do_fault;
595 if (!(env->efer & MSR_EFER_NXE) && (pde & PG_NX_MASK)) {
596 error_code = PG_ERROR_RSVD_MASK;
597 goto do_fault;
599 ptep &= pde ^ PG_NX_MASK;
600 if (pde & PG_PSE_MASK) {
601 /* 2 MB page */
602 page_size = 2048 * 1024;
603 ptep ^= PG_NX_MASK;
604 if ((ptep & PG_NX_MASK) && is_write1 == 2) {
605 goto do_fault_protect;
607 switch (mmu_idx) {
608 case MMU_USER_IDX:
609 if (!(ptep & PG_USER_MASK)) {
610 goto do_fault_protect;
612 if (is_write && !(ptep & PG_RW_MASK)) {
613 goto do_fault_protect;
615 break;
617 case MMU_KERNEL_IDX:
618 if (is_write1 != 2 && (env->cr[4] & CR4_SMAP_MASK) &&
619 (ptep & PG_USER_MASK)) {
620 goto do_fault_protect;
622 /* fall through */
623 case MMU_KSMAP_IDX:
624 if (is_write1 == 2 && (env->cr[4] & CR4_SMEP_MASK) &&
625 (ptep & PG_USER_MASK)) {
626 goto do_fault_protect;
628 if ((env->cr[0] & CR0_WP_MASK) &&
629 is_write && !(ptep & PG_RW_MASK)) {
630 goto do_fault_protect;
632 break;
634 default: /* cannot happen */
635 break;
637 is_dirty = is_write && !(pde & PG_DIRTY_MASK);
638 if (!(pde & PG_ACCESSED_MASK) || is_dirty) {
639 pde |= PG_ACCESSED_MASK;
640 if (is_dirty)
641 pde |= PG_DIRTY_MASK;
642 stl_phys_notdirty(pde_addr, pde);
644 /* align to page_size */
645 pte = pde & ((PHYS_ADDR_MASK & ~(page_size - 1)) | 0xfff);
646 virt_addr = addr & ~(page_size - 1);
647 } else {
648 /* 4 KB page */
649 if (!(pde & PG_ACCESSED_MASK)) {
650 pde |= PG_ACCESSED_MASK;
651 stl_phys_notdirty(pde_addr, pde);
653 pte_addr = ((pde & PHYS_ADDR_MASK) + (((addr >> 12) & 0x1ff) << 3)) &
654 env->a20_mask;
655 pte = ldq_phys(pte_addr);
656 if (!(pte & PG_PRESENT_MASK)) {
657 error_code = 0;
658 goto do_fault;
660 if (!(env->efer & MSR_EFER_NXE) && (pte & PG_NX_MASK)) {
661 error_code = PG_ERROR_RSVD_MASK;
662 goto do_fault;
664 /* combine pde and pte nx, user and rw protections */
665 ptep &= pte ^ PG_NX_MASK;
666 ptep ^= PG_NX_MASK;
667 if ((ptep & PG_NX_MASK) && is_write1 == 2)
668 goto do_fault_protect;
669 switch (mmu_idx) {
670 case MMU_USER_IDX:
671 if (!(ptep & PG_USER_MASK)) {
672 goto do_fault_protect;
674 if (is_write && !(ptep & PG_RW_MASK)) {
675 goto do_fault_protect;
677 break;
679 case MMU_KERNEL_IDX:
680 if (is_write1 != 2 && (env->cr[4] & CR4_SMAP_MASK) &&
681 (ptep & PG_USER_MASK)) {
682 goto do_fault_protect;
684 /* fall through */
685 case MMU_KSMAP_IDX:
686 if (is_write1 == 2 && (env->cr[4] & CR4_SMEP_MASK) &&
687 (ptep & PG_USER_MASK)) {
688 goto do_fault_protect;
690 if ((env->cr[0] & CR0_WP_MASK) &&
691 is_write && !(ptep & PG_RW_MASK)) {
692 goto do_fault_protect;
694 break;
696 default: /* cannot happen */
697 break;
699 is_dirty = is_write && !(pte & PG_DIRTY_MASK);
700 if (!(pte & PG_ACCESSED_MASK) || is_dirty) {
701 pte |= PG_ACCESSED_MASK;
702 if (is_dirty)
703 pte |= PG_DIRTY_MASK;
704 stl_phys_notdirty(pte_addr, pte);
706 page_size = 4096;
707 virt_addr = addr & ~0xfff;
708 pte = pte & (PHYS_ADDR_MASK | 0xfff);
710 } else {
711 uint32_t pde;
713 /* page directory entry */
714 pde_addr = ((env->cr[3] & ~0xfff) + ((addr >> 20) & 0xffc)) &
715 env->a20_mask;
716 pde = ldl_phys(pde_addr);
717 if (!(pde & PG_PRESENT_MASK)) {
718 error_code = 0;
719 goto do_fault;
721 /* if PSE bit is set, then we use a 4MB page */
722 if ((pde & PG_PSE_MASK) && (env->cr[4] & CR4_PSE_MASK)) {
723 page_size = 4096 * 1024;
724 switch (mmu_idx) {
725 case MMU_USER_IDX:
726 if (!(pde & PG_USER_MASK)) {
727 goto do_fault_protect;
729 if (is_write && !(pde & PG_RW_MASK)) {
730 goto do_fault_protect;
732 break;
734 case MMU_KERNEL_IDX:
735 if (is_write1 != 2 && (env->cr[4] & CR4_SMAP_MASK) &&
736 (pde & PG_USER_MASK)) {
737 goto do_fault_protect;
739 /* fall through */
740 case MMU_KSMAP_IDX:
741 if (is_write1 == 2 && (env->cr[4] & CR4_SMEP_MASK) &&
742 (pde & PG_USER_MASK)) {
743 goto do_fault_protect;
745 if ((env->cr[0] & CR0_WP_MASK) &&
746 is_write && !(pde & PG_RW_MASK)) {
747 goto do_fault_protect;
749 break;
751 default: /* cannot happen */
752 break;
754 is_dirty = is_write && !(pde & PG_DIRTY_MASK);
755 if (!(pde & PG_ACCESSED_MASK) || is_dirty) {
756 pde |= PG_ACCESSED_MASK;
757 if (is_dirty)
758 pde |= PG_DIRTY_MASK;
759 stl_phys_notdirty(pde_addr, pde);
762 pte = pde & ~( (page_size - 1) & ~0xfff); /* align to page_size */
763 ptep = pte;
764 virt_addr = addr & ~(page_size - 1);
765 } else {
766 if (!(pde & PG_ACCESSED_MASK)) {
767 pde |= PG_ACCESSED_MASK;
768 stl_phys_notdirty(pde_addr, pde);
771 /* page directory entry */
772 pte_addr = ((pde & ~0xfff) + ((addr >> 10) & 0xffc)) &
773 env->a20_mask;
774 pte = ldl_phys(pte_addr);
775 if (!(pte & PG_PRESENT_MASK)) {
776 error_code = 0;
777 goto do_fault;
779 /* combine pde and pte user and rw protections */
780 ptep = pte & pde;
781 switch (mmu_idx) {
782 case MMU_USER_IDX:
783 if (!(ptep & PG_USER_MASK)) {
784 goto do_fault_protect;
786 if (is_write && !(ptep & PG_RW_MASK)) {
787 goto do_fault_protect;
789 break;
791 case MMU_KERNEL_IDX:
792 if (is_write1 != 2 && (env->cr[4] & CR4_SMAP_MASK) &&
793 (ptep & PG_USER_MASK)) {
794 goto do_fault_protect;
796 /* fall through */
797 case MMU_KSMAP_IDX:
798 if (is_write1 == 2 && (env->cr[4] & CR4_SMEP_MASK) &&
799 (ptep & PG_USER_MASK)) {
800 goto do_fault_protect;
802 if ((env->cr[0] & CR0_WP_MASK) &&
803 is_write && !(ptep & PG_RW_MASK)) {
804 goto do_fault_protect;
806 break;
808 default: /* cannot happen */
809 break;
811 is_dirty = is_write && !(pte & PG_DIRTY_MASK);
812 if (!(pte & PG_ACCESSED_MASK) || is_dirty) {
813 pte |= PG_ACCESSED_MASK;
814 if (is_dirty)
815 pte |= PG_DIRTY_MASK;
816 stl_phys_notdirty(pte_addr, pte);
818 page_size = 4096;
819 virt_addr = addr & ~0xfff;
822 /* the page can be put in the TLB */
823 prot = PAGE_READ;
824 if (!(ptep & PG_NX_MASK))
825 prot |= PAGE_EXEC;
826 if (pte & PG_DIRTY_MASK) {
827 /* only set write access if already dirty... otherwise wait
828 for dirty access */
829 if (is_user) {
830 if (ptep & PG_RW_MASK)
831 prot |= PAGE_WRITE;
832 } else {
833 if (!(env->cr[0] & CR0_WP_MASK) ||
834 (ptep & PG_RW_MASK))
835 prot |= PAGE_WRITE;
838 do_mapping:
839 pte = pte & env->a20_mask;
841 /* Even if 4MB pages, we map only one 4KB page in the cache to
842 avoid filling it too fast */
843 page_offset = (addr & TARGET_PAGE_MASK) & (page_size - 1);
844 paddr = (pte & TARGET_PAGE_MASK) + page_offset;
845 vaddr = virt_addr + page_offset;
847 tlb_set_page(env, vaddr, paddr, prot, mmu_idx, page_size);
848 return 0;
849 do_fault_protect:
850 error_code = PG_ERROR_P_MASK;
851 do_fault:
852 error_code |= (is_write << PG_ERROR_W_BIT);
853 if (is_user)
854 error_code |= PG_ERROR_U_MASK;
855 if (is_write1 == 2 &&
856 (((env->efer & MSR_EFER_NXE) &&
857 (env->cr[4] & CR4_PAE_MASK)) ||
858 (env->cr[4] & CR4_SMEP_MASK)))
859 error_code |= PG_ERROR_I_D_MASK;
860 if (env->intercept_exceptions & (1 << EXCP0E_PAGE)) {
861 /* cr2 is not modified in case of exceptions */
862 stq_phys(env->vm_vmcb + offsetof(struct vmcb, control.exit_info_2),
863 addr);
864 } else {
865 env->cr[2] = addr;
867 env->error_code = error_code;
868 env->exception_index = EXCP0E_PAGE;
869 return 1;
872 hwaddr cpu_get_phys_page_debug(CPUX86State *env, target_ulong addr)
874 target_ulong pde_addr, pte_addr;
875 uint64_t pte;
876 hwaddr paddr;
877 uint32_t page_offset;
878 int page_size;
880 if (env->cr[4] & CR4_PAE_MASK) {
881 target_ulong pdpe_addr;
882 uint64_t pde, pdpe;
884 #ifdef TARGET_X86_64
885 if (env->hflags & HF_LMA_MASK) {
886 uint64_t pml4e_addr, pml4e;
887 int32_t sext;
889 /* test virtual address sign extension */
890 sext = (int64_t)addr >> 47;
891 if (sext != 0 && sext != -1)
892 return -1;
894 pml4e_addr = ((env->cr[3] & ~0xfff) + (((addr >> 39) & 0x1ff) << 3)) &
895 env->a20_mask;
896 pml4e = ldq_phys(pml4e_addr);
897 if (!(pml4e & PG_PRESENT_MASK))
898 return -1;
900 pdpe_addr = ((pml4e & ~0xfff & ~(PG_NX_MASK | PG_HI_USER_MASK)) +
901 (((addr >> 30) & 0x1ff) << 3)) & env->a20_mask;
902 pdpe = ldq_phys(pdpe_addr);
903 if (!(pdpe & PG_PRESENT_MASK))
904 return -1;
905 } else
906 #endif
908 pdpe_addr = ((env->cr[3] & ~0x1f) + ((addr >> 27) & 0x18)) &
909 env->a20_mask;
910 pdpe = ldq_phys(pdpe_addr);
911 if (!(pdpe & PG_PRESENT_MASK))
912 return -1;
915 pde_addr = ((pdpe & ~0xfff & ~(PG_NX_MASK | PG_HI_USER_MASK)) +
916 (((addr >> 21) & 0x1ff) << 3)) & env->a20_mask;
917 pde = ldq_phys(pde_addr);
918 if (!(pde & PG_PRESENT_MASK)) {
919 return -1;
921 if (pde & PG_PSE_MASK) {
922 /* 2 MB page */
923 page_size = 2048 * 1024;
924 pte = pde & ~( (page_size - 1) & ~0xfff); /* align to page_size */
925 } else {
926 /* 4 KB page */
927 pte_addr = ((pde & ~0xfff & ~(PG_NX_MASK | PG_HI_USER_MASK)) +
928 (((addr >> 12) & 0x1ff) << 3)) & env->a20_mask;
929 page_size = 4096;
930 pte = ldq_phys(pte_addr);
932 pte &= ~(PG_NX_MASK | PG_HI_USER_MASK);
933 if (!(pte & PG_PRESENT_MASK))
934 return -1;
935 } else {
936 uint32_t pde;
938 if (!(env->cr[0] & CR0_PG_MASK)) {
939 pte = addr;
940 page_size = 4096;
941 } else {
942 /* page directory entry */
943 pde_addr = ((env->cr[3] & ~0xfff) + ((addr >> 20) & 0xffc)) & env->a20_mask;
944 pde = ldl_phys(pde_addr);
945 if (!(pde & PG_PRESENT_MASK))
946 return -1;
947 if ((pde & PG_PSE_MASK) && (env->cr[4] & CR4_PSE_MASK)) {
948 pte = pde & ~0x003ff000; /* align to 4MB */
949 page_size = 4096 * 1024;
950 } else {
951 /* page directory entry */
952 pte_addr = ((pde & ~0xfff) + ((addr >> 10) & 0xffc)) & env->a20_mask;
953 pte = ldl_phys(pte_addr);
954 if (!(pte & PG_PRESENT_MASK))
955 return -1;
956 page_size = 4096;
959 pte = pte & env->a20_mask;
962 page_offset = (addr & TARGET_PAGE_MASK) & (page_size - 1);
963 paddr = (pte & TARGET_PAGE_MASK) + page_offset;
964 return paddr;
967 void hw_breakpoint_insert(CPUX86State *env, int index)
969 int type, err = 0;
971 switch (hw_breakpoint_type(env->dr[7], index)) {
972 case 0:
973 if (hw_breakpoint_enabled(env->dr[7], index))
974 err = cpu_breakpoint_insert(env, env->dr[index], BP_CPU,
975 &env->cpu_breakpoint[index]);
976 break;
977 case 1:
978 type = BP_CPU | BP_MEM_WRITE;
979 goto insert_wp;
980 case 2:
981 /* No support for I/O watchpoints yet */
982 break;
983 case 3:
984 type = BP_CPU | BP_MEM_ACCESS;
985 insert_wp:
986 err = cpu_watchpoint_insert(env, env->dr[index],
987 hw_breakpoint_len(env->dr[7], index),
988 type, &env->cpu_watchpoint[index]);
989 break;
991 if (err)
992 env->cpu_breakpoint[index] = NULL;
995 void hw_breakpoint_remove(CPUX86State *env, int index)
997 if (!env->cpu_breakpoint[index])
998 return;
999 switch (hw_breakpoint_type(env->dr[7], index)) {
1000 case 0:
1001 if (hw_breakpoint_enabled(env->dr[7], index))
1002 cpu_breakpoint_remove_by_ref(env, env->cpu_breakpoint[index]);
1003 break;
1004 case 1:
1005 case 3:
1006 cpu_watchpoint_remove_by_ref(env, env->cpu_watchpoint[index]);
1007 break;
1008 case 2:
1009 /* No support for I/O watchpoints yet */
1010 break;
1014 int check_hw_breakpoints(CPUX86State *env, int force_dr6_update)
1016 target_ulong dr6;
1017 int reg, type;
1018 int hit_enabled = 0;
1020 dr6 = env->dr[6] & ~0xf;
1021 for (reg = 0; reg < 4; reg++) {
1022 type = hw_breakpoint_type(env->dr[7], reg);
1023 if ((type == 0 && env->dr[reg] == env->eip) ||
1024 ((type & 1) && env->cpu_watchpoint[reg] &&
1025 (env->cpu_watchpoint[reg]->flags & BP_WATCHPOINT_HIT))) {
1026 dr6 |= 1 << reg;
1027 if (hw_breakpoint_enabled(env->dr[7], reg))
1028 hit_enabled = 1;
1031 if (hit_enabled || force_dr6_update)
1032 env->dr[6] = dr6;
1033 return hit_enabled;
1036 void breakpoint_handler(CPUX86State *env)
1038 CPUBreakpoint *bp;
1040 if (env->watchpoint_hit) {
1041 if (env->watchpoint_hit->flags & BP_CPU) {
1042 env->watchpoint_hit = NULL;
1043 if (check_hw_breakpoints(env, 0))
1044 raise_exception(env, EXCP01_DB);
1045 else
1046 cpu_resume_from_signal(env, NULL);
1048 } else {
1049 QTAILQ_FOREACH(bp, &env->breakpoints, entry)
1050 if (bp->pc == env->eip) {
1051 if (bp->flags & BP_CPU) {
1052 check_hw_breakpoints(env, 1);
1053 raise_exception(env, EXCP01_DB);
1055 break;
1060 typedef struct MCEInjectionParams {
1061 Monitor *mon;
1062 CPUX86State *env;
1063 int bank;
1064 uint64_t status;
1065 uint64_t mcg_status;
1066 uint64_t addr;
1067 uint64_t misc;
1068 int flags;
1069 } MCEInjectionParams;
1071 static void do_inject_x86_mce(void *data)
1073 MCEInjectionParams *params = data;
1074 CPUX86State *cenv = params->env;
1075 uint64_t *banks = cenv->mce_banks + 4 * params->bank;
1077 cpu_synchronize_state(cenv);
1080 * If there is an MCE exception being processed, ignore this SRAO MCE
1081 * unless unconditional injection was requested.
1083 if (!(params->flags & MCE_INJECT_UNCOND_AO)
1084 && !(params->status & MCI_STATUS_AR)
1085 && (cenv->mcg_status & MCG_STATUS_MCIP)) {
1086 return;
1089 if (params->status & MCI_STATUS_UC) {
1091 * if MSR_MCG_CTL is not all 1s, the uncorrected error
1092 * reporting is disabled
1094 if ((cenv->mcg_cap & MCG_CTL_P) && cenv->mcg_ctl != ~(uint64_t)0) {
1095 monitor_printf(params->mon,
1096 "CPU %d: Uncorrected error reporting disabled\n",
1097 cenv->cpu_index);
1098 return;
1102 * if MSR_MCi_CTL is not all 1s, the uncorrected error
1103 * reporting is disabled for the bank
1105 if (banks[0] != ~(uint64_t)0) {
1106 monitor_printf(params->mon,
1107 "CPU %d: Uncorrected error reporting disabled for"
1108 " bank %d\n",
1109 cenv->cpu_index, params->bank);
1110 return;
1113 if ((cenv->mcg_status & MCG_STATUS_MCIP) ||
1114 !(cenv->cr[4] & CR4_MCE_MASK)) {
1115 monitor_printf(params->mon,
1116 "CPU %d: Previous MCE still in progress, raising"
1117 " triple fault\n",
1118 cenv->cpu_index);
1119 qemu_log_mask(CPU_LOG_RESET, "Triple fault\n");
1120 qemu_system_reset_request();
1121 return;
1123 if (banks[1] & MCI_STATUS_VAL) {
1124 params->status |= MCI_STATUS_OVER;
1126 banks[2] = params->addr;
1127 banks[3] = params->misc;
1128 cenv->mcg_status = params->mcg_status;
1129 banks[1] = params->status;
1130 cpu_interrupt(cenv, CPU_INTERRUPT_MCE);
1131 } else if (!(banks[1] & MCI_STATUS_VAL)
1132 || !(banks[1] & MCI_STATUS_UC)) {
1133 if (banks[1] & MCI_STATUS_VAL) {
1134 params->status |= MCI_STATUS_OVER;
1136 banks[2] = params->addr;
1137 banks[3] = params->misc;
1138 banks[1] = params->status;
1139 } else {
1140 banks[1] |= MCI_STATUS_OVER;
1144 void cpu_x86_inject_mce(Monitor *mon, X86CPU *cpu, int bank,
1145 uint64_t status, uint64_t mcg_status, uint64_t addr,
1146 uint64_t misc, int flags)
1148 CPUX86State *cenv = &cpu->env;
1149 MCEInjectionParams params = {
1150 .mon = mon,
1151 .env = cenv,
1152 .bank = bank,
1153 .status = status,
1154 .mcg_status = mcg_status,
1155 .addr = addr,
1156 .misc = misc,
1157 .flags = flags,
1159 unsigned bank_num = cenv->mcg_cap & 0xff;
1160 CPUX86State *env;
1162 if (!cenv->mcg_cap) {
1163 monitor_printf(mon, "MCE injection not supported\n");
1164 return;
1166 if (bank >= bank_num) {
1167 monitor_printf(mon, "Invalid MCE bank number\n");
1168 return;
1170 if (!(status & MCI_STATUS_VAL)) {
1171 monitor_printf(mon, "Invalid MCE status code\n");
1172 return;
1174 if ((flags & MCE_INJECT_BROADCAST)
1175 && !cpu_x86_support_mca_broadcast(cenv)) {
1176 monitor_printf(mon, "Guest CPU does not support MCA broadcast\n");
1177 return;
1180 run_on_cpu(CPU(cpu), do_inject_x86_mce, &params);
1181 if (flags & MCE_INJECT_BROADCAST) {
1182 params.bank = 1;
1183 params.status = MCI_STATUS_VAL | MCI_STATUS_UC;
1184 params.mcg_status = MCG_STATUS_MCIP | MCG_STATUS_RIPV;
1185 params.addr = 0;
1186 params.misc = 0;
1187 for (env = first_cpu; env != NULL; env = env->next_cpu) {
1188 if (cenv == env) {
1189 continue;
1191 params.env = env;
1192 run_on_cpu(CPU(cpu), do_inject_x86_mce, &params);
1197 void cpu_report_tpr_access(CPUX86State *env, TPRAccess access)
1199 if (kvm_enabled()) {
1200 env->tpr_access_type = access;
1202 cpu_interrupt(env, CPU_INTERRUPT_TPR);
1203 } else {
1204 cpu_restore_state(env, env->mem_io_pc);
1206 apic_handle_tpr_access_report(env->apic_state, env->eip, access);
1209 #endif /* !CONFIG_USER_ONLY */
1211 int cpu_x86_get_descr_debug(CPUX86State *env, unsigned int selector,
1212 target_ulong *base, unsigned int *limit,
1213 unsigned int *flags)
1215 SegmentCache *dt;
1216 target_ulong ptr;
1217 uint32_t e1, e2;
1218 int index;
1220 if (selector & 0x4)
1221 dt = &env->ldt;
1222 else
1223 dt = &env->gdt;
1224 index = selector & ~7;
1225 ptr = dt->base + index;
1226 if ((index + 7) > dt->limit
1227 || cpu_memory_rw_debug(env, ptr, (uint8_t *)&e1, sizeof(e1), 0) != 0
1228 || cpu_memory_rw_debug(env, ptr+4, (uint8_t *)&e2, sizeof(e2), 0) != 0)
1229 return 0;
1231 *base = ((e1 >> 16) | ((e2 & 0xff) << 16) | (e2 & 0xff000000));
1232 *limit = (e1 & 0xffff) | (e2 & 0x000f0000);
1233 if (e2 & DESC_G_MASK)
1234 *limit = (*limit << 12) | 0xfff;
1235 *flags = e2;
1237 return 1;
1240 X86CPU *cpu_x86_init(const char *cpu_model)
1242 X86CPU *cpu;
1243 CPUX86State *env;
1244 Error *error = NULL;
1246 cpu = X86_CPU(object_new(TYPE_X86_CPU));
1247 env = &cpu->env;
1248 env->cpu_model_str = cpu_model;
1250 if (cpu_x86_register(cpu, cpu_model) < 0) {
1251 object_delete(OBJECT(cpu));
1252 return NULL;
1255 x86_cpu_realize(OBJECT(cpu), &error);
1256 if (error) {
1257 error_free(error);
1258 object_delete(OBJECT(cpu));
1259 return NULL;
1261 return cpu;
1264 #if !defined(CONFIG_USER_ONLY)
1265 void do_cpu_init(X86CPU *cpu)
1267 CPUX86State *env = &cpu->env;
1268 int sipi = env->interrupt_request & CPU_INTERRUPT_SIPI;
1269 uint64_t pat = env->pat;
1271 cpu_reset(CPU(cpu));
1272 env->interrupt_request = sipi;
1273 env->pat = pat;
1274 apic_init_reset(env->apic_state);
1277 void do_cpu_sipi(X86CPU *cpu)
1279 CPUX86State *env = &cpu->env;
1281 apic_sipi(env->apic_state);
1283 #else
1284 void do_cpu_init(X86CPU *cpu)
1287 void do_cpu_sipi(X86CPU *cpu)
1290 #endif