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 "qemu/osdep.h"
22 #include "sysemu/kvm.h"
24 #ifndef CONFIG_USER_ONLY
25 #include "sysemu/sysemu.h"
26 #include "monitor/monitor.h"
27 #include "hw/i386/apic_internal.h"
30 static void cpu_x86_version(CPUX86State *env, int *family, int *model)
32 int cpuver = env->cpuid_version;
34 if (family == NULL || model == NULL) {
38 *family = (cpuver >> 8) & 0x0f;
39 *model = ((cpuver >> 12) & 0xf0) + ((cpuver >> 4) & 0x0f);
42 /* Broadcast MCA signal for processor version 06H_EH and above */
43 int cpu_x86_support_mca_broadcast(CPUX86State *env)
48 cpu_x86_version(env, &family, &model);
49 if ((family == 6 && model >= 14) || family > 6) {
56 /***********************************************************/
59 static const char *cc_op_str[CC_OP_NB] = {
126 cpu_x86_dump_seg_cache(CPUX86State *env, FILE *f, fprintf_function cpu_fprintf,
127 const char *name, struct SegmentCache *sc)
130 if (env->hflags & HF_CS64_MASK) {
131 cpu_fprintf(f, "%-3s=%04x %016" PRIx64 " %08x %08x", name,
132 sc->selector, sc->base, sc->limit, sc->flags & 0x00ffff00);
136 cpu_fprintf(f, "%-3s=%04x %08x %08x %08x", name, sc->selector,
137 (uint32_t)sc->base, sc->limit, sc->flags & 0x00ffff00);
140 if (!(env->hflags & HF_PE_MASK) || !(sc->flags & DESC_P_MASK))
143 cpu_fprintf(f, " DPL=%d ", (sc->flags & DESC_DPL_MASK) >> DESC_DPL_SHIFT);
144 if (sc->flags & DESC_S_MASK) {
145 if (sc->flags & DESC_CS_MASK) {
146 cpu_fprintf(f, (sc->flags & DESC_L_MASK) ? "CS64" :
147 ((sc->flags & DESC_B_MASK) ? "CS32" : "CS16"));
148 cpu_fprintf(f, " [%c%c", (sc->flags & DESC_C_MASK) ? 'C' : '-',
149 (sc->flags & DESC_R_MASK) ? 'R' : '-');
152 (sc->flags & DESC_B_MASK || env->hflags & HF_LMA_MASK)
154 cpu_fprintf(f, " [%c%c", (sc->flags & DESC_E_MASK) ? 'E' : '-',
155 (sc->flags & DESC_W_MASK) ? 'W' : '-');
157 cpu_fprintf(f, "%c]", (sc->flags & DESC_A_MASK) ? 'A' : '-');
159 static const char *sys_type_name[2][16] = {
161 "Reserved", "TSS16-avl", "LDT", "TSS16-busy",
162 "CallGate16", "TaskGate", "IntGate16", "TrapGate16",
163 "Reserved", "TSS32-avl", "Reserved", "TSS32-busy",
164 "CallGate32", "Reserved", "IntGate32", "TrapGate32"
167 "<hiword>", "Reserved", "LDT", "Reserved", "Reserved",
168 "Reserved", "Reserved", "Reserved", "Reserved",
169 "TSS64-avl", "Reserved", "TSS64-busy", "CallGate64",
170 "Reserved", "IntGate64", "TrapGate64"
174 sys_type_name[(env->hflags & HF_LMA_MASK) ? 1 : 0]
175 [(sc->flags & DESC_TYPE_MASK)
176 >> DESC_TYPE_SHIFT]);
179 cpu_fprintf(f, "\n");
182 #ifndef CONFIG_USER_ONLY
184 /* ARRAY_SIZE check is not required because
185 * DeliveryMode(dm) has a size of 3 bit.
187 static inline const char *dm2str(uint32_t dm)
189 static const char *str[] = {
202 static void dump_apic_lvt(FILE *f, fprintf_function cpu_fprintf,
203 const char *name, uint32_t lvt, bool is_timer)
205 uint32_t dm = (lvt & APIC_LVT_DELIV_MOD) >> APIC_LVT_DELIV_MOD_SHIFT;
207 "%s\t 0x%08x %s %-5s %-6s %-7s %-12s %-6s",
209 lvt & APIC_LVT_INT_POLARITY ? "active-lo" : "active-hi",
210 lvt & APIC_LVT_LEVEL_TRIGGER ? "level" : "edge",
211 lvt & APIC_LVT_MASKED ? "masked" : "",
212 lvt & APIC_LVT_DELIV_STS ? "pending" : "",
214 "" : lvt & APIC_LVT_TIMER_PERIODIC ?
215 "periodic" : lvt & APIC_LVT_TIMER_TSCDEADLINE ?
216 "tsc-deadline" : "one-shot",
218 if (dm != APIC_DM_NMI) {
219 cpu_fprintf(f, " (vec %u)\n", lvt & APIC_VECTOR_MASK);
221 cpu_fprintf(f, "\n");
225 /* ARRAY_SIZE check is not required because
226 * destination shorthand has a size of 2 bit.
228 static inline const char *shorthand2str(uint32_t shorthand)
230 const char *str[] = {
231 "no-shorthand", "self", "all-self", "all"
233 return str[shorthand];
236 static inline uint8_t divider_conf(uint32_t divide_conf)
238 uint8_t divide_val = ((divide_conf & 0x8) >> 1) | (divide_conf & 0x3);
240 return divide_val == 7 ? 1 : 2 << divide_val;
243 static inline void mask2str(char *str, uint32_t val, uint8_t size)
246 *str++ = (val >> size) & 1 ? '1' : '0';
251 #define MAX_LOGICAL_APIC_ID_MASK_SIZE 16
253 static void dump_apic_icr(FILE *f, fprintf_function cpu_fprintf,
254 APICCommonState *s, CPUX86State *env)
256 uint32_t icr = s->icr[0], icr2 = s->icr[1];
257 uint8_t dest_shorthand = \
258 (icr & APIC_ICR_DEST_SHORT) >> APIC_ICR_DEST_SHORT_SHIFT;
259 bool logical_mod = icr & APIC_ICR_DEST_MOD;
260 char apic_id_str[MAX_LOGICAL_APIC_ID_MASK_SIZE + 1];
264 cpu_fprintf(f, "ICR\t 0x%08x %s %s %s %s\n",
266 logical_mod ? "logical" : "physical",
267 icr & APIC_ICR_TRIGGER_MOD ? "level" : "edge",
268 icr & APIC_ICR_LEVEL ? "assert" : "de-assert",
269 shorthand2str(dest_shorthand));
271 cpu_fprintf(f, "ICR2\t 0x%08x", icr2);
272 if (dest_shorthand != 0) {
273 cpu_fprintf(f, "\n");
276 x2apic = env->features[FEAT_1_ECX] & CPUID_EXT_X2APIC;
277 dest_field = x2apic ? icr2 : icr2 >> APIC_ICR_DEST_SHIFT;
281 cpu_fprintf(f, " cpu %u (X2APIC ID)\n", dest_field);
283 cpu_fprintf(f, " cpu %u (APIC ID)\n",
284 dest_field & APIC_LOGDEST_XAPIC_ID);
289 if (s->dest_mode == 0xf) { /* flat mode */
290 mask2str(apic_id_str, icr2 >> APIC_ICR_DEST_SHIFT, 8);
291 cpu_fprintf(f, " mask %s (APIC ID)\n", apic_id_str);
292 } else if (s->dest_mode == 0) { /* cluster mode */
294 mask2str(apic_id_str, dest_field & APIC_LOGDEST_X2APIC_ID, 16);
295 cpu_fprintf(f, " cluster %u mask %s (X2APIC ID)\n",
296 dest_field >> APIC_LOGDEST_X2APIC_SHIFT, apic_id_str);
298 mask2str(apic_id_str, dest_field & APIC_LOGDEST_XAPIC_ID, 4);
299 cpu_fprintf(f, " cluster %u mask %s (APIC ID)\n",
300 dest_field >> APIC_LOGDEST_XAPIC_SHIFT, apic_id_str);
305 static void dump_apic_interrupt(FILE *f, fprintf_function cpu_fprintf,
306 const char *name, uint32_t *ireg_tab,
311 cpu_fprintf(f, "%s\t ", name);
312 for (i = 0; i < 256; i++) {
313 if (apic_get_bit(ireg_tab, i)) {
314 cpu_fprintf(f, "%u%s ", i,
315 apic_get_bit(tmr_tab, i) ? "(level)" : "");
319 cpu_fprintf(f, "%s\n", empty ? "(none)" : "");
322 void x86_cpu_dump_local_apic_state(CPUState *cs, FILE *f,
323 fprintf_function cpu_fprintf, int flags)
325 X86CPU *cpu = X86_CPU(cs);
326 APICCommonState *s = APIC_COMMON(cpu->apic_state);
327 uint32_t *lvt = s->lvt;
329 cpu_fprintf(f, "dumping local APIC state for CPU %-2u\n\n",
330 CPU(cpu)->cpu_index);
331 dump_apic_lvt(f, cpu_fprintf, "LVT0", lvt[APIC_LVT_LINT0], false);
332 dump_apic_lvt(f, cpu_fprintf, "LVT1", lvt[APIC_LVT_LINT1], false);
333 dump_apic_lvt(f, cpu_fprintf, "LVTPC", lvt[APIC_LVT_PERFORM], false);
334 dump_apic_lvt(f, cpu_fprintf, "LVTERR", lvt[APIC_LVT_ERROR], false);
335 dump_apic_lvt(f, cpu_fprintf, "LVTTHMR", lvt[APIC_LVT_THERMAL], false);
336 dump_apic_lvt(f, cpu_fprintf, "LVTT", lvt[APIC_LVT_TIMER], true);
338 cpu_fprintf(f, "Timer\t DCR=0x%x (divide by %u) initial_count = %u\n",
339 s->divide_conf & APIC_DCR_MASK,
340 divider_conf(s->divide_conf),
343 cpu_fprintf(f, "SPIV\t 0x%08x APIC %s, focus=%s, spurious vec %u\n",
345 s->spurious_vec & APIC_SPURIO_ENABLED ? "enabled" : "disabled",
346 s->spurious_vec & APIC_SPURIO_FOCUS ? "on" : "off",
347 s->spurious_vec & APIC_VECTOR_MASK);
349 dump_apic_icr(f, cpu_fprintf, s, &cpu->env);
351 cpu_fprintf(f, "ESR\t 0x%08x\n", s->esr);
353 dump_apic_interrupt(f, cpu_fprintf, "ISR", s->isr, s->tmr);
354 dump_apic_interrupt(f, cpu_fprintf, "IRR", s->irr, s->tmr);
356 cpu_fprintf(f, "\nAPR 0x%02x TPR 0x%02x DFR 0x%02x LDR 0x%02x",
357 s->arb_id, s->tpr, s->dest_mode, s->log_dest);
358 if (s->dest_mode == 0) {
359 cpu_fprintf(f, "(cluster %u: id %u)",
360 s->log_dest >> APIC_LOGDEST_XAPIC_SHIFT,
361 s->log_dest & APIC_LOGDEST_XAPIC_ID);
363 cpu_fprintf(f, " PPR 0x%02x\n", apic_get_ppr(s));
366 void x86_cpu_dump_local_apic_state(CPUState *cs, FILE *f,
367 fprintf_function cpu_fprintf, int flags)
370 #endif /* !CONFIG_USER_ONLY */
372 #define DUMP_CODE_BYTES_TOTAL 50
373 #define DUMP_CODE_BYTES_BACKWARD 20
375 void x86_cpu_dump_state(CPUState *cs, FILE *f, fprintf_function cpu_fprintf,
378 X86CPU *cpu = X86_CPU(cs);
379 CPUX86State *env = &cpu->env;
382 static const char *seg_name[6] = { "ES", "CS", "SS", "DS", "FS", "GS" };
384 eflags = cpu_compute_eflags(env);
386 if (env->hflags & HF_CS64_MASK) {
388 "RAX=%016" PRIx64 " RBX=%016" PRIx64 " RCX=%016" PRIx64 " RDX=%016" PRIx64 "\n"
389 "RSI=%016" PRIx64 " RDI=%016" PRIx64 " RBP=%016" PRIx64 " RSP=%016" PRIx64 "\n"
390 "R8 =%016" PRIx64 " R9 =%016" PRIx64 " R10=%016" PRIx64 " R11=%016" PRIx64 "\n"
391 "R12=%016" PRIx64 " R13=%016" PRIx64 " R14=%016" PRIx64 " R15=%016" PRIx64 "\n"
392 "RIP=%016" PRIx64 " RFL=%08x [%c%c%c%c%c%c%c] CPL=%d II=%d A20=%d SMM=%d HLT=%d\n",
410 eflags & DF_MASK ? 'D' : '-',
411 eflags & CC_O ? 'O' : '-',
412 eflags & CC_S ? 'S' : '-',
413 eflags & CC_Z ? 'Z' : '-',
414 eflags & CC_A ? 'A' : '-',
415 eflags & CC_P ? 'P' : '-',
416 eflags & CC_C ? 'C' : '-',
417 env->hflags & HF_CPL_MASK,
418 (env->hflags >> HF_INHIBIT_IRQ_SHIFT) & 1,
419 (env->a20_mask >> 20) & 1,
420 (env->hflags >> HF_SMM_SHIFT) & 1,
425 cpu_fprintf(f, "EAX=%08x EBX=%08x ECX=%08x EDX=%08x\n"
426 "ESI=%08x EDI=%08x EBP=%08x ESP=%08x\n"
427 "EIP=%08x EFL=%08x [%c%c%c%c%c%c%c] CPL=%d II=%d A20=%d SMM=%d HLT=%d\n",
428 (uint32_t)env->regs[R_EAX],
429 (uint32_t)env->regs[R_EBX],
430 (uint32_t)env->regs[R_ECX],
431 (uint32_t)env->regs[R_EDX],
432 (uint32_t)env->regs[R_ESI],
433 (uint32_t)env->regs[R_EDI],
434 (uint32_t)env->regs[R_EBP],
435 (uint32_t)env->regs[R_ESP],
436 (uint32_t)env->eip, eflags,
437 eflags & DF_MASK ? 'D' : '-',
438 eflags & CC_O ? 'O' : '-',
439 eflags & CC_S ? 'S' : '-',
440 eflags & CC_Z ? 'Z' : '-',
441 eflags & CC_A ? 'A' : '-',
442 eflags & CC_P ? 'P' : '-',
443 eflags & CC_C ? 'C' : '-',
444 env->hflags & HF_CPL_MASK,
445 (env->hflags >> HF_INHIBIT_IRQ_SHIFT) & 1,
446 (env->a20_mask >> 20) & 1,
447 (env->hflags >> HF_SMM_SHIFT) & 1,
451 for(i = 0; i < 6; i++) {
452 cpu_x86_dump_seg_cache(env, f, cpu_fprintf, seg_name[i],
455 cpu_x86_dump_seg_cache(env, f, cpu_fprintf, "LDT", &env->ldt);
456 cpu_x86_dump_seg_cache(env, f, cpu_fprintf, "TR", &env->tr);
459 if (env->hflags & HF_LMA_MASK) {
460 cpu_fprintf(f, "GDT= %016" PRIx64 " %08x\n",
461 env->gdt.base, env->gdt.limit);
462 cpu_fprintf(f, "IDT= %016" PRIx64 " %08x\n",
463 env->idt.base, env->idt.limit);
464 cpu_fprintf(f, "CR0=%08x CR2=%016" PRIx64 " CR3=%016" PRIx64 " CR4=%08x\n",
465 (uint32_t)env->cr[0],
468 (uint32_t)env->cr[4]);
469 for(i = 0; i < 4; i++)
470 cpu_fprintf(f, "DR%d=%016" PRIx64 " ", i, env->dr[i]);
471 cpu_fprintf(f, "\nDR6=%016" PRIx64 " DR7=%016" PRIx64 "\n",
472 env->dr[6], env->dr[7]);
476 cpu_fprintf(f, "GDT= %08x %08x\n",
477 (uint32_t)env->gdt.base, env->gdt.limit);
478 cpu_fprintf(f, "IDT= %08x %08x\n",
479 (uint32_t)env->idt.base, env->idt.limit);
480 cpu_fprintf(f, "CR0=%08x CR2=%08x CR3=%08x CR4=%08x\n",
481 (uint32_t)env->cr[0],
482 (uint32_t)env->cr[2],
483 (uint32_t)env->cr[3],
484 (uint32_t)env->cr[4]);
485 for(i = 0; i < 4; i++) {
486 cpu_fprintf(f, "DR%d=" TARGET_FMT_lx " ", i, env->dr[i]);
488 cpu_fprintf(f, "\nDR6=" TARGET_FMT_lx " DR7=" TARGET_FMT_lx "\n",
489 env->dr[6], env->dr[7]);
491 if (flags & CPU_DUMP_CCOP) {
492 if ((unsigned)env->cc_op < CC_OP_NB)
493 snprintf(cc_op_name, sizeof(cc_op_name), "%s", cc_op_str[env->cc_op]);
495 snprintf(cc_op_name, sizeof(cc_op_name), "[%d]", env->cc_op);
497 if (env->hflags & HF_CS64_MASK) {
498 cpu_fprintf(f, "CCS=%016" PRIx64 " CCD=%016" PRIx64 " CCO=%-8s\n",
499 env->cc_src, env->cc_dst,
504 cpu_fprintf(f, "CCS=%08x CCD=%08x CCO=%-8s\n",
505 (uint32_t)env->cc_src, (uint32_t)env->cc_dst,
509 cpu_fprintf(f, "EFER=%016" PRIx64 "\n", env->efer);
510 if (flags & CPU_DUMP_FPU) {
513 for(i = 0; i < 8; i++) {
514 fptag |= ((!env->fptags[i]) << i);
516 cpu_fprintf(f, "FCW=%04x FSW=%04x [ST=%d] FTW=%02x MXCSR=%08x\n",
518 (env->fpus & ~0x3800) | (env->fpstt & 0x7) << 11,
524 u.d = env->fpregs[i].d;
525 cpu_fprintf(f, "FPR%d=%016" PRIx64 " %04x",
526 i, u.l.lower, u.l.upper);
528 cpu_fprintf(f, "\n");
532 if (env->hflags & HF_CS64_MASK)
537 cpu_fprintf(f, "XMM%02d=%08x%08x%08x%08x",
539 env->xmm_regs[i].ZMM_L(3),
540 env->xmm_regs[i].ZMM_L(2),
541 env->xmm_regs[i].ZMM_L(1),
542 env->xmm_regs[i].ZMM_L(0));
544 cpu_fprintf(f, "\n");
549 if (flags & CPU_DUMP_CODE) {
550 target_ulong base = env->segs[R_CS].base + env->eip;
551 target_ulong offs = MIN(env->eip, DUMP_CODE_BYTES_BACKWARD);
555 cpu_fprintf(f, "Code=");
556 for (i = 0; i < DUMP_CODE_BYTES_TOTAL; i++) {
557 if (cpu_memory_rw_debug(cs, base - offs + i, &code, 1, 0) == 0) {
558 snprintf(codestr, sizeof(codestr), "%02x", code);
560 snprintf(codestr, sizeof(codestr), "??");
562 cpu_fprintf(f, "%s%s%s%s", i > 0 ? " " : "",
563 i == offs ? "<" : "", codestr, i == offs ? ">" : "");
565 cpu_fprintf(f, "\n");
569 /***********************************************************/
571 /* XXX: add PGE support */
573 void x86_cpu_set_a20(X86CPU *cpu, int a20_state)
575 CPUX86State *env = &cpu->env;
577 a20_state = (a20_state != 0);
578 if (a20_state != ((env->a20_mask >> 20) & 1)) {
579 CPUState *cs = CPU(cpu);
581 qemu_log_mask(CPU_LOG_MMU, "A20 update: a20=%d\n", a20_state);
582 /* if the cpu is currently executing code, we must unlink it and
583 all the potentially executing TB */
584 cpu_interrupt(cs, CPU_INTERRUPT_EXITTB);
586 /* when a20 is changed, all the MMU mappings are invalid, so
587 we must flush everything */
589 env->a20_mask = ~(1 << 20) | (a20_state << 20);
593 void cpu_x86_update_cr0(CPUX86State *env, uint32_t new_cr0)
595 X86CPU *cpu = x86_env_get_cpu(env);
598 qemu_log_mask(CPU_LOG_MMU, "CR0 update: CR0=0x%08x\n", new_cr0);
599 if ((new_cr0 & (CR0_PG_MASK | CR0_WP_MASK | CR0_PE_MASK)) !=
600 (env->cr[0] & (CR0_PG_MASK | CR0_WP_MASK | CR0_PE_MASK))) {
601 tlb_flush(CPU(cpu), 1);
605 if (!(env->cr[0] & CR0_PG_MASK) && (new_cr0 & CR0_PG_MASK) &&
606 (env->efer & MSR_EFER_LME)) {
607 /* enter in long mode */
608 /* XXX: generate an exception */
609 if (!(env->cr[4] & CR4_PAE_MASK))
611 env->efer |= MSR_EFER_LMA;
612 env->hflags |= HF_LMA_MASK;
613 } else if ((env->cr[0] & CR0_PG_MASK) && !(new_cr0 & CR0_PG_MASK) &&
614 (env->efer & MSR_EFER_LMA)) {
616 env->efer &= ~MSR_EFER_LMA;
617 env->hflags &= ~(HF_LMA_MASK | HF_CS64_MASK);
618 env->eip &= 0xffffffff;
621 env->cr[0] = new_cr0 | CR0_ET_MASK;
623 /* update PE flag in hidden flags */
624 pe_state = (env->cr[0] & CR0_PE_MASK);
625 env->hflags = (env->hflags & ~HF_PE_MASK) | (pe_state << HF_PE_SHIFT);
626 /* ensure that ADDSEG is always set in real mode */
627 env->hflags |= ((pe_state ^ 1) << HF_ADDSEG_SHIFT);
628 /* update FPU flags */
629 env->hflags = (env->hflags & ~(HF_MP_MASK | HF_EM_MASK | HF_TS_MASK)) |
630 ((new_cr0 << (HF_MP_SHIFT - 1)) & (HF_MP_MASK | HF_EM_MASK | HF_TS_MASK));
633 /* XXX: in legacy PAE mode, generate a GPF if reserved bits are set in
635 void cpu_x86_update_cr3(CPUX86State *env, target_ulong new_cr3)
637 X86CPU *cpu = x86_env_get_cpu(env);
639 env->cr[3] = new_cr3;
640 if (env->cr[0] & CR0_PG_MASK) {
641 qemu_log_mask(CPU_LOG_MMU,
642 "CR3 update: CR3=" TARGET_FMT_lx "\n", new_cr3);
643 tlb_flush(CPU(cpu), 0);
647 void cpu_x86_update_cr4(CPUX86State *env, uint32_t new_cr4)
649 X86CPU *cpu = x86_env_get_cpu(env);
652 #if defined(DEBUG_MMU)
653 printf("CR4 update: CR4=%08x\n", (uint32_t)env->cr[4]);
655 if ((new_cr4 ^ env->cr[4]) &
656 (CR4_PGE_MASK | CR4_PAE_MASK | CR4_PSE_MASK |
657 CR4_SMEP_MASK | CR4_SMAP_MASK)) {
658 tlb_flush(CPU(cpu), 1);
661 /* Clear bits we're going to recompute. */
662 hflags = env->hflags & ~(HF_OSFXSR_MASK | HF_SMAP_MASK);
665 if (!(env->features[FEAT_1_EDX] & CPUID_SSE)) {
666 new_cr4 &= ~CR4_OSFXSR_MASK;
668 if (new_cr4 & CR4_OSFXSR_MASK) {
669 hflags |= HF_OSFXSR_MASK;
672 if (!(env->features[FEAT_7_0_EBX] & CPUID_7_0_EBX_SMAP)) {
673 new_cr4 &= ~CR4_SMAP_MASK;
675 if (new_cr4 & CR4_SMAP_MASK) {
676 hflags |= HF_SMAP_MASK;
679 env->cr[4] = new_cr4;
680 env->hflags = hflags;
683 #if defined(CONFIG_USER_ONLY)
685 int x86_cpu_handle_mmu_fault(CPUState *cs, vaddr addr,
686 int is_write, int mmu_idx)
688 X86CPU *cpu = X86_CPU(cs);
689 CPUX86State *env = &cpu->env;
691 /* user mode only emulation */
694 env->error_code = (is_write << PG_ERROR_W_BIT);
695 env->error_code |= PG_ERROR_U_MASK;
696 cs->exception_index = EXCP0E_PAGE;
703 * -1 = cannot handle fault
704 * 0 = nothing more to do
705 * 1 = generate PF fault
707 int x86_cpu_handle_mmu_fault(CPUState *cs, vaddr addr,
708 int is_write1, int mmu_idx)
710 X86CPU *cpu = X86_CPU(cs);
711 CPUX86State *env = &cpu->env;
713 target_ulong pde_addr, pte_addr;
715 int is_dirty, prot, page_size, is_write, is_user;
717 uint64_t rsvd_mask = PG_HI_RSVD_MASK;
718 uint32_t page_offset;
721 is_user = mmu_idx == MMU_USER_IDX;
722 #if defined(DEBUG_MMU)
723 printf("MMU fault: addr=%" VADDR_PRIx " w=%d u=%d eip=" TARGET_FMT_lx "\n",
724 addr, is_write1, is_user, env->eip);
726 is_write = is_write1 & 1;
728 if (!(env->cr[0] & CR0_PG_MASK)) {
731 if (!(env->hflags & HF_LMA_MASK)) {
732 /* Without long mode we can only address 32bits in real mode */
736 prot = PAGE_READ | PAGE_WRITE | PAGE_EXEC;
741 if (!(env->efer & MSR_EFER_NXE)) {
742 rsvd_mask |= PG_NX_MASK;
745 if (env->cr[4] & CR4_PAE_MASK) {
747 target_ulong pdpe_addr;
750 if (env->hflags & HF_LMA_MASK) {
751 uint64_t pml4e_addr, pml4e;
754 /* test virtual address sign extension */
755 sext = (int64_t)addr >> 47;
756 if (sext != 0 && sext != -1) {
758 cs->exception_index = EXCP0D_GPF;
762 pml4e_addr = ((env->cr[3] & ~0xfff) + (((addr >> 39) & 0x1ff) << 3)) &
764 pml4e = x86_ldq_phys(cs, pml4e_addr);
765 if (!(pml4e & PG_PRESENT_MASK)) {
768 if (pml4e & (rsvd_mask | PG_PSE_MASK)) {
771 if (!(pml4e & PG_ACCESSED_MASK)) {
772 pml4e |= PG_ACCESSED_MASK;
773 x86_stl_phys_notdirty(cs, pml4e_addr, pml4e);
775 ptep = pml4e ^ PG_NX_MASK;
776 pdpe_addr = ((pml4e & PG_ADDRESS_MASK) + (((addr >> 30) & 0x1ff) << 3)) &
778 pdpe = x86_ldq_phys(cs, pdpe_addr);
779 if (!(pdpe & PG_PRESENT_MASK)) {
782 if (pdpe & rsvd_mask) {
785 ptep &= pdpe ^ PG_NX_MASK;
786 if (!(pdpe & PG_ACCESSED_MASK)) {
787 pdpe |= PG_ACCESSED_MASK;
788 x86_stl_phys_notdirty(cs, pdpe_addr, pdpe);
790 if (pdpe & PG_PSE_MASK) {
792 page_size = 1024 * 1024 * 1024;
793 pte_addr = pdpe_addr;
795 goto do_check_protect;
800 /* XXX: load them when cr3 is loaded ? */
801 pdpe_addr = ((env->cr[3] & ~0x1f) + ((addr >> 27) & 0x18)) &
803 pdpe = x86_ldq_phys(cs, pdpe_addr);
804 if (!(pdpe & PG_PRESENT_MASK)) {
807 rsvd_mask |= PG_HI_USER_MASK;
808 if (pdpe & (rsvd_mask | PG_NX_MASK)) {
811 ptep = PG_NX_MASK | PG_USER_MASK | PG_RW_MASK;
814 pde_addr = ((pdpe & PG_ADDRESS_MASK) + (((addr >> 21) & 0x1ff) << 3)) &
816 pde = x86_ldq_phys(cs, pde_addr);
817 if (!(pde & PG_PRESENT_MASK)) {
820 if (pde & rsvd_mask) {
823 ptep &= pde ^ PG_NX_MASK;
824 if (pde & PG_PSE_MASK) {
826 page_size = 2048 * 1024;
829 goto do_check_protect;
832 if (!(pde & PG_ACCESSED_MASK)) {
833 pde |= PG_ACCESSED_MASK;
834 x86_stl_phys_notdirty(cs, pde_addr, pde);
836 pte_addr = ((pde & PG_ADDRESS_MASK) + (((addr >> 12) & 0x1ff) << 3)) &
838 pte = x86_ldq_phys(cs, pte_addr);
839 if (!(pte & PG_PRESENT_MASK)) {
842 if (pte & rsvd_mask) {
845 /* combine pde and pte nx, user and rw protections */
846 ptep &= pte ^ PG_NX_MASK;
851 /* page directory entry */
852 pde_addr = ((env->cr[3] & ~0xfff) + ((addr >> 20) & 0xffc)) &
854 pde = x86_ldl_phys(cs, pde_addr);
855 if (!(pde & PG_PRESENT_MASK)) {
858 ptep = pde | PG_NX_MASK;
860 /* if PSE bit is set, then we use a 4MB page */
861 if ((pde & PG_PSE_MASK) && (env->cr[4] & CR4_PSE_MASK)) {
862 page_size = 4096 * 1024;
865 /* Bits 20-13 provide bits 39-32 of the address, bit 21 is reserved.
866 * Leave bits 20-13 in place for setting accessed/dirty bits below.
868 pte = pde | ((pde & 0x1fe000LL) << (32 - 13));
869 rsvd_mask = 0x200000;
870 goto do_check_protect_pse36;
873 if (!(pde & PG_ACCESSED_MASK)) {
874 pde |= PG_ACCESSED_MASK;
875 x86_stl_phys_notdirty(cs, pde_addr, pde);
878 /* page directory entry */
879 pte_addr = ((pde & ~0xfff) + ((addr >> 10) & 0xffc)) &
881 pte = x86_ldl_phys(cs, pte_addr);
882 if (!(pte & PG_PRESENT_MASK)) {
885 /* combine pde and pte user and rw protections */
886 ptep &= pte | PG_NX_MASK;
892 rsvd_mask |= (page_size - 1) & PG_ADDRESS_MASK & ~PG_PSE_PAT_MASK;
893 do_check_protect_pse36:
894 if (pte & rsvd_mask) {
899 /* can the page can be put in the TLB? prot will tell us */
900 if (is_user && !(ptep & PG_USER_MASK)) {
901 goto do_fault_protect;
905 if (mmu_idx != MMU_KSMAP_IDX || !(ptep & PG_USER_MASK)) {
907 if ((ptep & PG_RW_MASK) || (!is_user && !(env->cr[0] & CR0_WP_MASK))) {
911 if (!(ptep & PG_NX_MASK) &&
912 (mmu_idx == MMU_USER_IDX ||
913 !((env->cr[4] & CR4_SMEP_MASK) && (ptep & PG_USER_MASK)))) {
917 if ((prot & (1 << is_write1)) == 0) {
918 goto do_fault_protect;
922 is_dirty = is_write && !(pte & PG_DIRTY_MASK);
923 if (!(pte & PG_ACCESSED_MASK) || is_dirty) {
924 pte |= PG_ACCESSED_MASK;
926 pte |= PG_DIRTY_MASK;
928 x86_stl_phys_notdirty(cs, pte_addr, pte);
931 if (!(pte & PG_DIRTY_MASK)) {
932 /* only set write access if already dirty... otherwise wait
939 pte = pte & env->a20_mask;
941 /* align to page_size */
942 pte &= PG_ADDRESS_MASK & ~(page_size - 1);
944 /* Even if 4MB pages, we map only one 4KB page in the cache to
945 avoid filling it too fast */
946 vaddr = addr & TARGET_PAGE_MASK;
947 page_offset = vaddr & (page_size - 1);
948 paddr = pte + page_offset;
950 assert(prot & (1 << is_write1));
951 tlb_set_page_with_attrs(cs, vaddr, paddr, cpu_get_mem_attrs(env),
952 prot, mmu_idx, page_size);
955 error_code |= PG_ERROR_RSVD_MASK;
957 error_code |= PG_ERROR_P_MASK;
959 error_code |= (is_write << PG_ERROR_W_BIT);
961 error_code |= PG_ERROR_U_MASK;
962 if (is_write1 == 2 &&
963 (((env->efer & MSR_EFER_NXE) &&
964 (env->cr[4] & CR4_PAE_MASK)) ||
965 (env->cr[4] & CR4_SMEP_MASK)))
966 error_code |= PG_ERROR_I_D_MASK;
967 if (env->intercept_exceptions & (1 << EXCP0E_PAGE)) {
968 /* cr2 is not modified in case of exceptions */
970 env->vm_vmcb + offsetof(struct vmcb, control.exit_info_2),
975 env->error_code = error_code;
976 cs->exception_index = EXCP0E_PAGE;
980 hwaddr x86_cpu_get_phys_page_debug(CPUState *cs, vaddr addr)
982 X86CPU *cpu = X86_CPU(cs);
983 CPUX86State *env = &cpu->env;
984 target_ulong pde_addr, pte_addr;
986 uint32_t page_offset;
989 if (!(env->cr[0] & CR0_PG_MASK)) {
990 pte = addr & env->a20_mask;
992 } else if (env->cr[4] & CR4_PAE_MASK) {
993 target_ulong pdpe_addr;
997 if (env->hflags & HF_LMA_MASK) {
998 uint64_t pml4e_addr, pml4e;
1001 /* test virtual address sign extension */
1002 sext = (int64_t)addr >> 47;
1003 if (sext != 0 && sext != -1) {
1006 pml4e_addr = ((env->cr[3] & ~0xfff) + (((addr >> 39) & 0x1ff) << 3)) &
1008 pml4e = x86_ldq_phys(cs, pml4e_addr);
1009 if (!(pml4e & PG_PRESENT_MASK)) {
1012 pdpe_addr = ((pml4e & PG_ADDRESS_MASK) +
1013 (((addr >> 30) & 0x1ff) << 3)) & env->a20_mask;
1014 pdpe = x86_ldq_phys(cs, pdpe_addr);
1015 if (!(pdpe & PG_PRESENT_MASK)) {
1018 if (pdpe & PG_PSE_MASK) {
1019 page_size = 1024 * 1024 * 1024;
1027 pdpe_addr = ((env->cr[3] & ~0x1f) + ((addr >> 27) & 0x18)) &
1029 pdpe = x86_ldq_phys(cs, pdpe_addr);
1030 if (!(pdpe & PG_PRESENT_MASK))
1034 pde_addr = ((pdpe & PG_ADDRESS_MASK) +
1035 (((addr >> 21) & 0x1ff) << 3)) & env->a20_mask;
1036 pde = x86_ldq_phys(cs, pde_addr);
1037 if (!(pde & PG_PRESENT_MASK)) {
1040 if (pde & PG_PSE_MASK) {
1042 page_size = 2048 * 1024;
1046 pte_addr = ((pde & PG_ADDRESS_MASK) +
1047 (((addr >> 12) & 0x1ff) << 3)) & env->a20_mask;
1049 pte = x86_ldq_phys(cs, pte_addr);
1051 if (!(pte & PG_PRESENT_MASK)) {
1057 /* page directory entry */
1058 pde_addr = ((env->cr[3] & ~0xfff) + ((addr >> 20) & 0xffc)) & env->a20_mask;
1059 pde = x86_ldl_phys(cs, pde_addr);
1060 if (!(pde & PG_PRESENT_MASK))
1062 if ((pde & PG_PSE_MASK) && (env->cr[4] & CR4_PSE_MASK)) {
1063 pte = pde | ((pde & 0x1fe000LL) << (32 - 13));
1064 page_size = 4096 * 1024;
1066 /* page directory entry */
1067 pte_addr = ((pde & ~0xfff) + ((addr >> 10) & 0xffc)) & env->a20_mask;
1068 pte = x86_ldl_phys(cs, pte_addr);
1069 if (!(pte & PG_PRESENT_MASK)) {
1074 pte = pte & env->a20_mask;
1077 #ifdef TARGET_X86_64
1080 pte &= PG_ADDRESS_MASK & ~(page_size - 1);
1081 page_offset = (addr & TARGET_PAGE_MASK) & (page_size - 1);
1082 return pte | page_offset;
1085 typedef struct MCEInjectionParams {
1090 uint64_t mcg_status;
1094 } MCEInjectionParams;
1096 static void do_inject_x86_mce(void *data)
1098 MCEInjectionParams *params = data;
1099 CPUX86State *cenv = ¶ms->cpu->env;
1100 CPUState *cpu = CPU(params->cpu);
1101 uint64_t *banks = cenv->mce_banks + 4 * params->bank;
1103 cpu_synchronize_state(cpu);
1106 * If there is an MCE exception being processed, ignore this SRAO MCE
1107 * unless unconditional injection was requested.
1109 if (!(params->flags & MCE_INJECT_UNCOND_AO)
1110 && !(params->status & MCI_STATUS_AR)
1111 && (cenv->mcg_status & MCG_STATUS_MCIP)) {
1115 if (params->status & MCI_STATUS_UC) {
1117 * if MSR_MCG_CTL is not all 1s, the uncorrected error
1118 * reporting is disabled
1120 if ((cenv->mcg_cap & MCG_CTL_P) && cenv->mcg_ctl != ~(uint64_t)0) {
1121 monitor_printf(params->mon,
1122 "CPU %d: Uncorrected error reporting disabled\n",
1128 * if MSR_MCi_CTL is not all 1s, the uncorrected error
1129 * reporting is disabled for the bank
1131 if (banks[0] != ~(uint64_t)0) {
1132 monitor_printf(params->mon,
1133 "CPU %d: Uncorrected error reporting disabled for"
1135 cpu->cpu_index, params->bank);
1139 if ((cenv->mcg_status & MCG_STATUS_MCIP) ||
1140 !(cenv->cr[4] & CR4_MCE_MASK)) {
1141 monitor_printf(params->mon,
1142 "CPU %d: Previous MCE still in progress, raising"
1145 qemu_log_mask(CPU_LOG_RESET, "Triple fault\n");
1146 qemu_system_reset_request();
1149 if (banks[1] & MCI_STATUS_VAL) {
1150 params->status |= MCI_STATUS_OVER;
1152 banks[2] = params->addr;
1153 banks[3] = params->misc;
1154 cenv->mcg_status = params->mcg_status;
1155 banks[1] = params->status;
1156 cpu_interrupt(cpu, CPU_INTERRUPT_MCE);
1157 } else if (!(banks[1] & MCI_STATUS_VAL)
1158 || !(banks[1] & MCI_STATUS_UC)) {
1159 if (banks[1] & MCI_STATUS_VAL) {
1160 params->status |= MCI_STATUS_OVER;
1162 banks[2] = params->addr;
1163 banks[3] = params->misc;
1164 banks[1] = params->status;
1166 banks[1] |= MCI_STATUS_OVER;
1170 void cpu_x86_inject_mce(Monitor *mon, X86CPU *cpu, int bank,
1171 uint64_t status, uint64_t mcg_status, uint64_t addr,
1172 uint64_t misc, int flags)
1174 CPUState *cs = CPU(cpu);
1175 CPUX86State *cenv = &cpu->env;
1176 MCEInjectionParams params = {
1181 .mcg_status = mcg_status,
1186 unsigned bank_num = cenv->mcg_cap & 0xff;
1188 if (!cenv->mcg_cap) {
1189 monitor_printf(mon, "MCE injection not supported\n");
1192 if (bank >= bank_num) {
1193 monitor_printf(mon, "Invalid MCE bank number\n");
1196 if (!(status & MCI_STATUS_VAL)) {
1197 monitor_printf(mon, "Invalid MCE status code\n");
1200 if ((flags & MCE_INJECT_BROADCAST)
1201 && !cpu_x86_support_mca_broadcast(cenv)) {
1202 monitor_printf(mon, "Guest CPU does not support MCA broadcast\n");
1206 run_on_cpu(cs, do_inject_x86_mce, ¶ms);
1207 if (flags & MCE_INJECT_BROADCAST) {
1211 params.status = MCI_STATUS_VAL | MCI_STATUS_UC;
1212 params.mcg_status = MCG_STATUS_MCIP | MCG_STATUS_RIPV;
1215 CPU_FOREACH(other_cs) {
1216 if (other_cs == cs) {
1219 params.cpu = X86_CPU(other_cs);
1220 run_on_cpu(other_cs, do_inject_x86_mce, ¶ms);
1225 void cpu_report_tpr_access(CPUX86State *env, TPRAccess access)
1227 X86CPU *cpu = x86_env_get_cpu(env);
1228 CPUState *cs = CPU(cpu);
1230 if (kvm_enabled()) {
1231 env->tpr_access_type = access;
1233 cpu_interrupt(cs, CPU_INTERRUPT_TPR);
1235 cpu_restore_state(cs, cs->mem_io_pc);
1237 apic_handle_tpr_access_report(cpu->apic_state, env->eip, access);
1240 #endif /* !CONFIG_USER_ONLY */
1242 int cpu_x86_get_descr_debug(CPUX86State *env, unsigned int selector,
1243 target_ulong *base, unsigned int *limit,
1244 unsigned int *flags)
1246 X86CPU *cpu = x86_env_get_cpu(env);
1247 CPUState *cs = CPU(cpu);
1257 index = selector & ~7;
1258 ptr = dt->base + index;
1259 if ((index + 7) > dt->limit
1260 || cpu_memory_rw_debug(cs, ptr, (uint8_t *)&e1, sizeof(e1), 0) != 0
1261 || cpu_memory_rw_debug(cs, ptr+4, (uint8_t *)&e2, sizeof(e2), 0) != 0)
1264 *base = ((e1 >> 16) | ((e2 & 0xff) << 16) | (e2 & 0xff000000));
1265 *limit = (e1 & 0xffff) | (e2 & 0x000f0000);
1266 if (e2 & DESC_G_MASK)
1267 *limit = (*limit << 12) | 0xfff;
1273 #if !defined(CONFIG_USER_ONLY)
1274 void do_cpu_init(X86CPU *cpu)
1276 CPUState *cs = CPU(cpu);
1277 CPUX86State *env = &cpu->env;
1278 CPUX86State *save = g_new(CPUX86State, 1);
1279 int sipi = cs->interrupt_request & CPU_INTERRUPT_SIPI;
1284 cs->interrupt_request = sipi;
1285 memcpy(&env->start_init_save, &save->start_init_save,
1286 offsetof(CPUX86State, end_init_save) -
1287 offsetof(CPUX86State, start_init_save));
1290 if (kvm_enabled()) {
1291 kvm_arch_do_init_vcpu(cpu);
1293 apic_init_reset(cpu->apic_state);
1296 void do_cpu_sipi(X86CPU *cpu)
1298 apic_sipi(cpu->apic_state);
1301 void do_cpu_init(X86CPU *cpu)
1304 void do_cpu_sipi(X86CPU *cpu)
1309 /* Frob eflags into and out of the CPU temporary format. */
1311 void x86_cpu_exec_enter(CPUState *cs)
1313 X86CPU *cpu = X86_CPU(cs);
1314 CPUX86State *env = &cpu->env;
1316 CC_SRC = env->eflags & (CC_O | CC_S | CC_Z | CC_A | CC_P | CC_C);
1317 env->df = 1 - (2 * ((env->eflags >> 10) & 1));
1318 CC_OP = CC_OP_EFLAGS;
1319 env->eflags &= ~(DF_MASK | CC_O | CC_S | CC_Z | CC_A | CC_P | CC_C);
1322 void x86_cpu_exec_exit(CPUState *cs)
1324 X86CPU *cpu = X86_CPU(cs);
1325 CPUX86State *env = &cpu->env;
1327 env->eflags = cpu_compute_eflags(env);
1330 #ifndef CONFIG_USER_ONLY
1331 uint8_t x86_ldub_phys(CPUState *cs, hwaddr addr)
1333 X86CPU *cpu = X86_CPU(cs);
1334 CPUX86State *env = &cpu->env;
1336 return address_space_ldub(cs->as, addr,
1337 cpu_get_mem_attrs(env),
1341 uint32_t x86_lduw_phys(CPUState *cs, hwaddr addr)
1343 X86CPU *cpu = X86_CPU(cs);
1344 CPUX86State *env = &cpu->env;
1346 return address_space_lduw(cs->as, addr,
1347 cpu_get_mem_attrs(env),
1351 uint32_t x86_ldl_phys(CPUState *cs, hwaddr addr)
1353 X86CPU *cpu = X86_CPU(cs);
1354 CPUX86State *env = &cpu->env;
1356 return address_space_ldl(cs->as, addr,
1357 cpu_get_mem_attrs(env),
1361 uint64_t x86_ldq_phys(CPUState *cs, hwaddr addr)
1363 X86CPU *cpu = X86_CPU(cs);
1364 CPUX86State *env = &cpu->env;
1366 return address_space_ldq(cs->as, addr,
1367 cpu_get_mem_attrs(env),
1371 void x86_stb_phys(CPUState *cs, hwaddr addr, uint8_t val)
1373 X86CPU *cpu = X86_CPU(cs);
1374 CPUX86State *env = &cpu->env;
1376 address_space_stb(cs->as, addr, val,
1377 cpu_get_mem_attrs(env),
1381 void x86_stl_phys_notdirty(CPUState *cs, hwaddr addr, uint32_t val)
1383 X86CPU *cpu = X86_CPU(cs);
1384 CPUX86State *env = &cpu->env;
1386 address_space_stl_notdirty(cs->as, addr, val,
1387 cpu_get_mem_attrs(env),
1391 void x86_stw_phys(CPUState *cs, hwaddr addr, uint32_t val)
1393 X86CPU *cpu = X86_CPU(cs);
1394 CPUX86State *env = &cpu->env;
1396 address_space_stw(cs->as, addr, val,
1397 cpu_get_mem_attrs(env),
1401 void x86_stl_phys(CPUState *cs, hwaddr addr, uint32_t val)
1403 X86CPU *cpu = X86_CPU(cs);
1404 CPUX86State *env = &cpu->env;
1406 address_space_stl(cs->as, addr, val,
1407 cpu_get_mem_attrs(env),
1411 void x86_stq_phys(CPUState *cs, hwaddr addr, uint64_t val)
1413 X86CPU *cpu = X86_CPU(cs);
1414 CPUX86State *env = &cpu->env;
1416 address_space_stq(cs->as, addr, val,
1417 cpu_get_mem_attrs(env),