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 "exec/exec-all.h"
23 #include "sysemu/kvm.h"
25 #ifndef CONFIG_USER_ONLY
26 #include "sysemu/sysemu.h"
27 #include "monitor/monitor.h"
28 #include "hw/i386/apic_internal.h"
31 static void cpu_x86_version(CPUX86State *env, int *family, int *model)
33 int cpuver = env->cpuid_version;
35 if (family == NULL || model == NULL) {
39 *family = (cpuver >> 8) & 0x0f;
40 *model = ((cpuver >> 12) & 0xf0) + ((cpuver >> 4) & 0x0f);
43 /* Broadcast MCA signal for processor version 06H_EH and above */
44 int cpu_x86_support_mca_broadcast(CPUX86State *env)
49 cpu_x86_version(env, &family, &model);
50 if ((family == 6 && model >= 14) || family > 6) {
57 /***********************************************************/
60 static const char *cc_op_str[CC_OP_NB] = {
127 cpu_x86_dump_seg_cache(CPUX86State *env, FILE *f, fprintf_function cpu_fprintf,
128 const char *name, struct SegmentCache *sc)
131 if (env->hflags & HF_CS64_MASK) {
132 cpu_fprintf(f, "%-3s=%04x %016" PRIx64 " %08x %08x", name,
133 sc->selector, sc->base, sc->limit, sc->flags & 0x00ffff00);
137 cpu_fprintf(f, "%-3s=%04x %08x %08x %08x", name, sc->selector,
138 (uint32_t)sc->base, sc->limit, sc->flags & 0x00ffff00);
141 if (!(env->hflags & HF_PE_MASK) || !(sc->flags & DESC_P_MASK))
144 cpu_fprintf(f, " DPL=%d ", (sc->flags & DESC_DPL_MASK) >> DESC_DPL_SHIFT);
145 if (sc->flags & DESC_S_MASK) {
146 if (sc->flags & DESC_CS_MASK) {
147 cpu_fprintf(f, (sc->flags & DESC_L_MASK) ? "CS64" :
148 ((sc->flags & DESC_B_MASK) ? "CS32" : "CS16"));
149 cpu_fprintf(f, " [%c%c", (sc->flags & DESC_C_MASK) ? 'C' : '-',
150 (sc->flags & DESC_R_MASK) ? 'R' : '-');
153 (sc->flags & DESC_B_MASK || env->hflags & HF_LMA_MASK)
155 cpu_fprintf(f, " [%c%c", (sc->flags & DESC_E_MASK) ? 'E' : '-',
156 (sc->flags & DESC_W_MASK) ? 'W' : '-');
158 cpu_fprintf(f, "%c]", (sc->flags & DESC_A_MASK) ? 'A' : '-');
160 static const char *sys_type_name[2][16] = {
162 "Reserved", "TSS16-avl", "LDT", "TSS16-busy",
163 "CallGate16", "TaskGate", "IntGate16", "TrapGate16",
164 "Reserved", "TSS32-avl", "Reserved", "TSS32-busy",
165 "CallGate32", "Reserved", "IntGate32", "TrapGate32"
168 "<hiword>", "Reserved", "LDT", "Reserved", "Reserved",
169 "Reserved", "Reserved", "Reserved", "Reserved",
170 "TSS64-avl", "Reserved", "TSS64-busy", "CallGate64",
171 "Reserved", "IntGate64", "TrapGate64"
175 sys_type_name[(env->hflags & HF_LMA_MASK) ? 1 : 0]
176 [(sc->flags & DESC_TYPE_MASK)
177 >> DESC_TYPE_SHIFT]);
180 cpu_fprintf(f, "\n");
183 #ifndef CONFIG_USER_ONLY
185 /* ARRAY_SIZE check is not required because
186 * DeliveryMode(dm) has a size of 3 bit.
188 static inline const char *dm2str(uint32_t dm)
190 static const char *str[] = {
203 static void dump_apic_lvt(FILE *f, fprintf_function cpu_fprintf,
204 const char *name, uint32_t lvt, bool is_timer)
206 uint32_t dm = (lvt & APIC_LVT_DELIV_MOD) >> APIC_LVT_DELIV_MOD_SHIFT;
208 "%s\t 0x%08x %s %-5s %-6s %-7s %-12s %-6s",
210 lvt & APIC_LVT_INT_POLARITY ? "active-lo" : "active-hi",
211 lvt & APIC_LVT_LEVEL_TRIGGER ? "level" : "edge",
212 lvt & APIC_LVT_MASKED ? "masked" : "",
213 lvt & APIC_LVT_DELIV_STS ? "pending" : "",
215 "" : lvt & APIC_LVT_TIMER_PERIODIC ?
216 "periodic" : lvt & APIC_LVT_TIMER_TSCDEADLINE ?
217 "tsc-deadline" : "one-shot",
219 if (dm != APIC_DM_NMI) {
220 cpu_fprintf(f, " (vec %u)\n", lvt & APIC_VECTOR_MASK);
222 cpu_fprintf(f, "\n");
226 /* ARRAY_SIZE check is not required because
227 * destination shorthand has a size of 2 bit.
229 static inline const char *shorthand2str(uint32_t shorthand)
231 const char *str[] = {
232 "no-shorthand", "self", "all-self", "all"
234 return str[shorthand];
237 static inline uint8_t divider_conf(uint32_t divide_conf)
239 uint8_t divide_val = ((divide_conf & 0x8) >> 1) | (divide_conf & 0x3);
241 return divide_val == 7 ? 1 : 2 << divide_val;
244 static inline void mask2str(char *str, uint32_t val, uint8_t size)
247 *str++ = (val >> size) & 1 ? '1' : '0';
252 #define MAX_LOGICAL_APIC_ID_MASK_SIZE 16
254 static void dump_apic_icr(FILE *f, fprintf_function cpu_fprintf,
255 APICCommonState *s, CPUX86State *env)
257 uint32_t icr = s->icr[0], icr2 = s->icr[1];
258 uint8_t dest_shorthand = \
259 (icr & APIC_ICR_DEST_SHORT) >> APIC_ICR_DEST_SHORT_SHIFT;
260 bool logical_mod = icr & APIC_ICR_DEST_MOD;
261 char apic_id_str[MAX_LOGICAL_APIC_ID_MASK_SIZE + 1];
265 cpu_fprintf(f, "ICR\t 0x%08x %s %s %s %s\n",
267 logical_mod ? "logical" : "physical",
268 icr & APIC_ICR_TRIGGER_MOD ? "level" : "edge",
269 icr & APIC_ICR_LEVEL ? "assert" : "de-assert",
270 shorthand2str(dest_shorthand));
272 cpu_fprintf(f, "ICR2\t 0x%08x", icr2);
273 if (dest_shorthand != 0) {
274 cpu_fprintf(f, "\n");
277 x2apic = env->features[FEAT_1_ECX] & CPUID_EXT_X2APIC;
278 dest_field = x2apic ? icr2 : icr2 >> APIC_ICR_DEST_SHIFT;
282 cpu_fprintf(f, " cpu %u (X2APIC ID)\n", dest_field);
284 cpu_fprintf(f, " cpu %u (APIC ID)\n",
285 dest_field & APIC_LOGDEST_XAPIC_ID);
290 if (s->dest_mode == 0xf) { /* flat mode */
291 mask2str(apic_id_str, icr2 >> APIC_ICR_DEST_SHIFT, 8);
292 cpu_fprintf(f, " mask %s (APIC ID)\n", apic_id_str);
293 } else if (s->dest_mode == 0) { /* cluster mode */
295 mask2str(apic_id_str, dest_field & APIC_LOGDEST_X2APIC_ID, 16);
296 cpu_fprintf(f, " cluster %u mask %s (X2APIC ID)\n",
297 dest_field >> APIC_LOGDEST_X2APIC_SHIFT, apic_id_str);
299 mask2str(apic_id_str, dest_field & APIC_LOGDEST_XAPIC_ID, 4);
300 cpu_fprintf(f, " cluster %u mask %s (APIC ID)\n",
301 dest_field >> APIC_LOGDEST_XAPIC_SHIFT, apic_id_str);
306 static void dump_apic_interrupt(FILE *f, fprintf_function cpu_fprintf,
307 const char *name, uint32_t *ireg_tab,
312 cpu_fprintf(f, "%s\t ", name);
313 for (i = 0; i < 256; i++) {
314 if (apic_get_bit(ireg_tab, i)) {
315 cpu_fprintf(f, "%u%s ", i,
316 apic_get_bit(tmr_tab, i) ? "(level)" : "");
320 cpu_fprintf(f, "%s\n", empty ? "(none)" : "");
323 void x86_cpu_dump_local_apic_state(CPUState *cs, FILE *f,
324 fprintf_function cpu_fprintf, int flags)
326 X86CPU *cpu = X86_CPU(cs);
327 APICCommonState *s = APIC_COMMON(cpu->apic_state);
328 uint32_t *lvt = s->lvt;
330 cpu_fprintf(f, "dumping local APIC state for CPU %-2u\n\n",
331 CPU(cpu)->cpu_index);
332 dump_apic_lvt(f, cpu_fprintf, "LVT0", lvt[APIC_LVT_LINT0], false);
333 dump_apic_lvt(f, cpu_fprintf, "LVT1", lvt[APIC_LVT_LINT1], false);
334 dump_apic_lvt(f, cpu_fprintf, "LVTPC", lvt[APIC_LVT_PERFORM], false);
335 dump_apic_lvt(f, cpu_fprintf, "LVTERR", lvt[APIC_LVT_ERROR], false);
336 dump_apic_lvt(f, cpu_fprintf, "LVTTHMR", lvt[APIC_LVT_THERMAL], false);
337 dump_apic_lvt(f, cpu_fprintf, "LVTT", lvt[APIC_LVT_TIMER], true);
339 cpu_fprintf(f, "Timer\t DCR=0x%x (divide by %u) initial_count = %u\n",
340 s->divide_conf & APIC_DCR_MASK,
341 divider_conf(s->divide_conf),
344 cpu_fprintf(f, "SPIV\t 0x%08x APIC %s, focus=%s, spurious vec %u\n",
346 s->spurious_vec & APIC_SPURIO_ENABLED ? "enabled" : "disabled",
347 s->spurious_vec & APIC_SPURIO_FOCUS ? "on" : "off",
348 s->spurious_vec & APIC_VECTOR_MASK);
350 dump_apic_icr(f, cpu_fprintf, s, &cpu->env);
352 cpu_fprintf(f, "ESR\t 0x%08x\n", s->esr);
354 dump_apic_interrupt(f, cpu_fprintf, "ISR", s->isr, s->tmr);
355 dump_apic_interrupt(f, cpu_fprintf, "IRR", s->irr, s->tmr);
357 cpu_fprintf(f, "\nAPR 0x%02x TPR 0x%02x DFR 0x%02x LDR 0x%02x",
358 s->arb_id, s->tpr, s->dest_mode, s->log_dest);
359 if (s->dest_mode == 0) {
360 cpu_fprintf(f, "(cluster %u: id %u)",
361 s->log_dest >> APIC_LOGDEST_XAPIC_SHIFT,
362 s->log_dest & APIC_LOGDEST_XAPIC_ID);
364 cpu_fprintf(f, " PPR 0x%02x\n", apic_get_ppr(s));
367 void x86_cpu_dump_local_apic_state(CPUState *cs, FILE *f,
368 fprintf_function cpu_fprintf, int flags)
371 #endif /* !CONFIG_USER_ONLY */
373 #define DUMP_CODE_BYTES_TOTAL 50
374 #define DUMP_CODE_BYTES_BACKWARD 20
376 void x86_cpu_dump_state(CPUState *cs, FILE *f, fprintf_function cpu_fprintf,
379 X86CPU *cpu = X86_CPU(cs);
380 CPUX86State *env = &cpu->env;
383 static const char *seg_name[6] = { "ES", "CS", "SS", "DS", "FS", "GS" };
385 eflags = cpu_compute_eflags(env);
387 if (env->hflags & HF_CS64_MASK) {
389 "RAX=%016" PRIx64 " RBX=%016" PRIx64 " RCX=%016" PRIx64 " RDX=%016" PRIx64 "\n"
390 "RSI=%016" PRIx64 " RDI=%016" PRIx64 " RBP=%016" PRIx64 " RSP=%016" PRIx64 "\n"
391 "R8 =%016" PRIx64 " R9 =%016" PRIx64 " R10=%016" PRIx64 " R11=%016" PRIx64 "\n"
392 "R12=%016" PRIx64 " R13=%016" PRIx64 " R14=%016" PRIx64 " R15=%016" PRIx64 "\n"
393 "RIP=%016" PRIx64 " RFL=%08x [%c%c%c%c%c%c%c] CPL=%d II=%d A20=%d SMM=%d HLT=%d\n",
411 eflags & DF_MASK ? 'D' : '-',
412 eflags & CC_O ? 'O' : '-',
413 eflags & CC_S ? 'S' : '-',
414 eflags & CC_Z ? 'Z' : '-',
415 eflags & CC_A ? 'A' : '-',
416 eflags & CC_P ? 'P' : '-',
417 eflags & CC_C ? 'C' : '-',
418 env->hflags & HF_CPL_MASK,
419 (env->hflags >> HF_INHIBIT_IRQ_SHIFT) & 1,
420 (env->a20_mask >> 20) & 1,
421 (env->hflags >> HF_SMM_SHIFT) & 1,
426 cpu_fprintf(f, "EAX=%08x EBX=%08x ECX=%08x EDX=%08x\n"
427 "ESI=%08x EDI=%08x EBP=%08x ESP=%08x\n"
428 "EIP=%08x EFL=%08x [%c%c%c%c%c%c%c] CPL=%d II=%d A20=%d SMM=%d HLT=%d\n",
429 (uint32_t)env->regs[R_EAX],
430 (uint32_t)env->regs[R_EBX],
431 (uint32_t)env->regs[R_ECX],
432 (uint32_t)env->regs[R_EDX],
433 (uint32_t)env->regs[R_ESI],
434 (uint32_t)env->regs[R_EDI],
435 (uint32_t)env->regs[R_EBP],
436 (uint32_t)env->regs[R_ESP],
437 (uint32_t)env->eip, eflags,
438 eflags & DF_MASK ? 'D' : '-',
439 eflags & CC_O ? 'O' : '-',
440 eflags & CC_S ? 'S' : '-',
441 eflags & CC_Z ? 'Z' : '-',
442 eflags & CC_A ? 'A' : '-',
443 eflags & CC_P ? 'P' : '-',
444 eflags & CC_C ? 'C' : '-',
445 env->hflags & HF_CPL_MASK,
446 (env->hflags >> HF_INHIBIT_IRQ_SHIFT) & 1,
447 (env->a20_mask >> 20) & 1,
448 (env->hflags >> HF_SMM_SHIFT) & 1,
452 for(i = 0; i < 6; i++) {
453 cpu_x86_dump_seg_cache(env, f, cpu_fprintf, seg_name[i],
456 cpu_x86_dump_seg_cache(env, f, cpu_fprintf, "LDT", &env->ldt);
457 cpu_x86_dump_seg_cache(env, f, cpu_fprintf, "TR", &env->tr);
460 if (env->hflags & HF_LMA_MASK) {
461 cpu_fprintf(f, "GDT= %016" PRIx64 " %08x\n",
462 env->gdt.base, env->gdt.limit);
463 cpu_fprintf(f, "IDT= %016" PRIx64 " %08x\n",
464 env->idt.base, env->idt.limit);
465 cpu_fprintf(f, "CR0=%08x CR2=%016" PRIx64 " CR3=%016" PRIx64 " CR4=%08x\n",
466 (uint32_t)env->cr[0],
469 (uint32_t)env->cr[4]);
470 for(i = 0; i < 4; i++)
471 cpu_fprintf(f, "DR%d=%016" PRIx64 " ", i, env->dr[i]);
472 cpu_fprintf(f, "\nDR6=%016" PRIx64 " DR7=%016" PRIx64 "\n",
473 env->dr[6], env->dr[7]);
477 cpu_fprintf(f, "GDT= %08x %08x\n",
478 (uint32_t)env->gdt.base, env->gdt.limit);
479 cpu_fprintf(f, "IDT= %08x %08x\n",
480 (uint32_t)env->idt.base, env->idt.limit);
481 cpu_fprintf(f, "CR0=%08x CR2=%08x CR3=%08x CR4=%08x\n",
482 (uint32_t)env->cr[0],
483 (uint32_t)env->cr[2],
484 (uint32_t)env->cr[3],
485 (uint32_t)env->cr[4]);
486 for(i = 0; i < 4; i++) {
487 cpu_fprintf(f, "DR%d=" TARGET_FMT_lx " ", i, env->dr[i]);
489 cpu_fprintf(f, "\nDR6=" TARGET_FMT_lx " DR7=" TARGET_FMT_lx "\n",
490 env->dr[6], env->dr[7]);
492 if (flags & CPU_DUMP_CCOP) {
493 if ((unsigned)env->cc_op < CC_OP_NB)
494 snprintf(cc_op_name, sizeof(cc_op_name), "%s", cc_op_str[env->cc_op]);
496 snprintf(cc_op_name, sizeof(cc_op_name), "[%d]", env->cc_op);
498 if (env->hflags & HF_CS64_MASK) {
499 cpu_fprintf(f, "CCS=%016" PRIx64 " CCD=%016" PRIx64 " CCO=%-8s\n",
500 env->cc_src, env->cc_dst,
505 cpu_fprintf(f, "CCS=%08x CCD=%08x CCO=%-8s\n",
506 (uint32_t)env->cc_src, (uint32_t)env->cc_dst,
510 cpu_fprintf(f, "EFER=%016" PRIx64 "\n", env->efer);
511 if (flags & CPU_DUMP_FPU) {
514 for(i = 0; i < 8; i++) {
515 fptag |= ((!env->fptags[i]) << i);
517 cpu_fprintf(f, "FCW=%04x FSW=%04x [ST=%d] FTW=%02x MXCSR=%08x\n",
519 (env->fpus & ~0x3800) | (env->fpstt & 0x7) << 11,
525 u.d = env->fpregs[i].d;
526 cpu_fprintf(f, "FPR%d=%016" PRIx64 " %04x",
527 i, u.l.lower, u.l.upper);
529 cpu_fprintf(f, "\n");
533 if (env->hflags & HF_CS64_MASK)
538 cpu_fprintf(f, "XMM%02d=%08x%08x%08x%08x",
540 env->xmm_regs[i].ZMM_L(3),
541 env->xmm_regs[i].ZMM_L(2),
542 env->xmm_regs[i].ZMM_L(1),
543 env->xmm_regs[i].ZMM_L(0));
545 cpu_fprintf(f, "\n");
550 if (flags & CPU_DUMP_CODE) {
551 target_ulong base = env->segs[R_CS].base + env->eip;
552 target_ulong offs = MIN(env->eip, DUMP_CODE_BYTES_BACKWARD);
556 cpu_fprintf(f, "Code=");
557 for (i = 0; i < DUMP_CODE_BYTES_TOTAL; i++) {
558 if (cpu_memory_rw_debug(cs, base - offs + i, &code, 1, 0) == 0) {
559 snprintf(codestr, sizeof(codestr), "%02x", code);
561 snprintf(codestr, sizeof(codestr), "??");
563 cpu_fprintf(f, "%s%s%s%s", i > 0 ? " " : "",
564 i == offs ? "<" : "", codestr, i == offs ? ">" : "");
566 cpu_fprintf(f, "\n");
570 /***********************************************************/
572 /* XXX: add PGE support */
574 void x86_cpu_set_a20(X86CPU *cpu, int a20_state)
576 CPUX86State *env = &cpu->env;
578 a20_state = (a20_state != 0);
579 if (a20_state != ((env->a20_mask >> 20) & 1)) {
580 CPUState *cs = CPU(cpu);
582 qemu_log_mask(CPU_LOG_MMU, "A20 update: a20=%d\n", a20_state);
583 /* if the cpu is currently executing code, we must unlink it and
584 all the potentially executing TB */
585 cpu_interrupt(cs, CPU_INTERRUPT_EXITTB);
587 /* when a20 is changed, all the MMU mappings are invalid, so
588 we must flush everything */
590 env->a20_mask = ~(1 << 20) | (a20_state << 20);
594 void cpu_x86_update_cr0(CPUX86State *env, uint32_t new_cr0)
596 X86CPU *cpu = x86_env_get_cpu(env);
599 qemu_log_mask(CPU_LOG_MMU, "CR0 update: CR0=0x%08x\n", new_cr0);
600 if ((new_cr0 & (CR0_PG_MASK | CR0_WP_MASK | CR0_PE_MASK)) !=
601 (env->cr[0] & (CR0_PG_MASK | CR0_WP_MASK | CR0_PE_MASK))) {
602 tlb_flush(CPU(cpu), 1);
606 if (!(env->cr[0] & CR0_PG_MASK) && (new_cr0 & CR0_PG_MASK) &&
607 (env->efer & MSR_EFER_LME)) {
608 /* enter in long mode */
609 /* XXX: generate an exception */
610 if (!(env->cr[4] & CR4_PAE_MASK))
612 env->efer |= MSR_EFER_LMA;
613 env->hflags |= HF_LMA_MASK;
614 } else if ((env->cr[0] & CR0_PG_MASK) && !(new_cr0 & CR0_PG_MASK) &&
615 (env->efer & MSR_EFER_LMA)) {
617 env->efer &= ~MSR_EFER_LMA;
618 env->hflags &= ~(HF_LMA_MASK | HF_CS64_MASK);
619 env->eip &= 0xffffffff;
622 env->cr[0] = new_cr0 | CR0_ET_MASK;
624 /* update PE flag in hidden flags */
625 pe_state = (env->cr[0] & CR0_PE_MASK);
626 env->hflags = (env->hflags & ~HF_PE_MASK) | (pe_state << HF_PE_SHIFT);
627 /* ensure that ADDSEG is always set in real mode */
628 env->hflags |= ((pe_state ^ 1) << HF_ADDSEG_SHIFT);
629 /* update FPU flags */
630 env->hflags = (env->hflags & ~(HF_MP_MASK | HF_EM_MASK | HF_TS_MASK)) |
631 ((new_cr0 << (HF_MP_SHIFT - 1)) & (HF_MP_MASK | HF_EM_MASK | HF_TS_MASK));
634 /* XXX: in legacy PAE mode, generate a GPF if reserved bits are set in
636 void cpu_x86_update_cr3(CPUX86State *env, target_ulong new_cr3)
638 X86CPU *cpu = x86_env_get_cpu(env);
640 env->cr[3] = new_cr3;
641 if (env->cr[0] & CR0_PG_MASK) {
642 qemu_log_mask(CPU_LOG_MMU,
643 "CR3 update: CR3=" TARGET_FMT_lx "\n", new_cr3);
644 tlb_flush(CPU(cpu), 0);
648 void cpu_x86_update_cr4(CPUX86State *env, uint32_t new_cr4)
650 X86CPU *cpu = x86_env_get_cpu(env);
653 #if defined(DEBUG_MMU)
654 printf("CR4 update: CR4=%08x\n", (uint32_t)env->cr[4]);
656 if ((new_cr4 ^ env->cr[4]) &
657 (CR4_PGE_MASK | CR4_PAE_MASK | CR4_PSE_MASK |
658 CR4_SMEP_MASK | CR4_SMAP_MASK)) {
659 tlb_flush(CPU(cpu), 1);
662 /* Clear bits we're going to recompute. */
663 hflags = env->hflags & ~(HF_OSFXSR_MASK | HF_SMAP_MASK);
666 if (!(env->features[FEAT_1_EDX] & CPUID_SSE)) {
667 new_cr4 &= ~CR4_OSFXSR_MASK;
669 if (new_cr4 & CR4_OSFXSR_MASK) {
670 hflags |= HF_OSFXSR_MASK;
673 if (!(env->features[FEAT_7_0_EBX] & CPUID_7_0_EBX_SMAP)) {
674 new_cr4 &= ~CR4_SMAP_MASK;
676 if (new_cr4 & CR4_SMAP_MASK) {
677 hflags |= HF_SMAP_MASK;
680 if (!(env->features[FEAT_7_0_ECX] & CPUID_7_0_ECX_PKU)) {
681 new_cr4 &= ~CR4_PKE_MASK;
684 env->cr[4] = new_cr4;
685 env->hflags = hflags;
687 cpu_sync_bndcs_hflags(env);
690 #if defined(CONFIG_USER_ONLY)
692 int x86_cpu_handle_mmu_fault(CPUState *cs, vaddr addr,
693 int is_write, int mmu_idx)
695 X86CPU *cpu = X86_CPU(cs);
696 CPUX86State *env = &cpu->env;
698 /* user mode only emulation */
701 env->error_code = (is_write << PG_ERROR_W_BIT);
702 env->error_code |= PG_ERROR_U_MASK;
703 cs->exception_index = EXCP0E_PAGE;
704 env->exception_is_int = 0;
705 env->exception_next_eip = -1;
712 * -1 = cannot handle fault
713 * 0 = nothing more to do
714 * 1 = generate PF fault
716 int x86_cpu_handle_mmu_fault(CPUState *cs, vaddr addr,
717 int is_write1, int mmu_idx)
719 X86CPU *cpu = X86_CPU(cs);
720 CPUX86State *env = &cpu->env;
722 target_ulong pde_addr, pte_addr;
724 int is_dirty, prot, page_size, is_write, is_user;
726 uint64_t rsvd_mask = PG_HI_RSVD_MASK;
727 uint32_t page_offset;
730 is_user = mmu_idx == MMU_USER_IDX;
731 #if defined(DEBUG_MMU)
732 printf("MMU fault: addr=%" VADDR_PRIx " w=%d u=%d eip=" TARGET_FMT_lx "\n",
733 addr, is_write1, is_user, env->eip);
735 is_write = is_write1 & 1;
737 if (!(env->cr[0] & CR0_PG_MASK)) {
740 if (!(env->hflags & HF_LMA_MASK)) {
741 /* Without long mode we can only address 32bits in real mode */
745 prot = PAGE_READ | PAGE_WRITE | PAGE_EXEC;
750 if (!(env->efer & MSR_EFER_NXE)) {
751 rsvd_mask |= PG_NX_MASK;
754 if (env->cr[4] & CR4_PAE_MASK) {
756 target_ulong pdpe_addr;
759 if (env->hflags & HF_LMA_MASK) {
760 uint64_t pml4e_addr, pml4e;
763 /* test virtual address sign extension */
764 sext = (int64_t)addr >> 47;
765 if (sext != 0 && sext != -1) {
767 cs->exception_index = EXCP0D_GPF;
771 pml4e_addr = ((env->cr[3] & ~0xfff) + (((addr >> 39) & 0x1ff) << 3)) &
773 pml4e = x86_ldq_phys(cs, pml4e_addr);
774 if (!(pml4e & PG_PRESENT_MASK)) {
777 if (pml4e & (rsvd_mask | PG_PSE_MASK)) {
780 if (!(pml4e & PG_ACCESSED_MASK)) {
781 pml4e |= PG_ACCESSED_MASK;
782 x86_stl_phys_notdirty(cs, pml4e_addr, pml4e);
784 ptep = pml4e ^ PG_NX_MASK;
785 pdpe_addr = ((pml4e & PG_ADDRESS_MASK) + (((addr >> 30) & 0x1ff) << 3)) &
787 pdpe = x86_ldq_phys(cs, pdpe_addr);
788 if (!(pdpe & PG_PRESENT_MASK)) {
791 if (pdpe & rsvd_mask) {
794 ptep &= pdpe ^ PG_NX_MASK;
795 if (!(pdpe & PG_ACCESSED_MASK)) {
796 pdpe |= PG_ACCESSED_MASK;
797 x86_stl_phys_notdirty(cs, pdpe_addr, pdpe);
799 if (pdpe & PG_PSE_MASK) {
801 page_size = 1024 * 1024 * 1024;
802 pte_addr = pdpe_addr;
804 goto do_check_protect;
809 /* XXX: load them when cr3 is loaded ? */
810 pdpe_addr = ((env->cr[3] & ~0x1f) + ((addr >> 27) & 0x18)) &
812 pdpe = x86_ldq_phys(cs, pdpe_addr);
813 if (!(pdpe & PG_PRESENT_MASK)) {
816 rsvd_mask |= PG_HI_USER_MASK;
817 if (pdpe & (rsvd_mask | PG_NX_MASK)) {
820 ptep = PG_NX_MASK | PG_USER_MASK | PG_RW_MASK;
823 pde_addr = ((pdpe & PG_ADDRESS_MASK) + (((addr >> 21) & 0x1ff) << 3)) &
825 pde = x86_ldq_phys(cs, pde_addr);
826 if (!(pde & PG_PRESENT_MASK)) {
829 if (pde & rsvd_mask) {
832 ptep &= pde ^ PG_NX_MASK;
833 if (pde & PG_PSE_MASK) {
835 page_size = 2048 * 1024;
838 goto do_check_protect;
841 if (!(pde & PG_ACCESSED_MASK)) {
842 pde |= PG_ACCESSED_MASK;
843 x86_stl_phys_notdirty(cs, pde_addr, pde);
845 pte_addr = ((pde & PG_ADDRESS_MASK) + (((addr >> 12) & 0x1ff) << 3)) &
847 pte = x86_ldq_phys(cs, pte_addr);
848 if (!(pte & PG_PRESENT_MASK)) {
851 if (pte & rsvd_mask) {
854 /* combine pde and pte nx, user and rw protections */
855 ptep &= pte ^ PG_NX_MASK;
860 /* page directory entry */
861 pde_addr = ((env->cr[3] & ~0xfff) + ((addr >> 20) & 0xffc)) &
863 pde = x86_ldl_phys(cs, pde_addr);
864 if (!(pde & PG_PRESENT_MASK)) {
867 ptep = pde | PG_NX_MASK;
869 /* if PSE bit is set, then we use a 4MB page */
870 if ((pde & PG_PSE_MASK) && (env->cr[4] & CR4_PSE_MASK)) {
871 page_size = 4096 * 1024;
874 /* Bits 20-13 provide bits 39-32 of the address, bit 21 is reserved.
875 * Leave bits 20-13 in place for setting accessed/dirty bits below.
877 pte = pde | ((pde & 0x1fe000LL) << (32 - 13));
878 rsvd_mask = 0x200000;
879 goto do_check_protect_pse36;
882 if (!(pde & PG_ACCESSED_MASK)) {
883 pde |= PG_ACCESSED_MASK;
884 x86_stl_phys_notdirty(cs, pde_addr, pde);
887 /* page directory entry */
888 pte_addr = ((pde & ~0xfff) + ((addr >> 10) & 0xffc)) &
890 pte = x86_ldl_phys(cs, pte_addr);
891 if (!(pte & PG_PRESENT_MASK)) {
894 /* combine pde and pte user and rw protections */
895 ptep &= pte | PG_NX_MASK;
901 rsvd_mask |= (page_size - 1) & PG_ADDRESS_MASK & ~PG_PSE_PAT_MASK;
902 do_check_protect_pse36:
903 if (pte & rsvd_mask) {
908 /* can the page can be put in the TLB? prot will tell us */
909 if (is_user && !(ptep & PG_USER_MASK)) {
910 goto do_fault_protect;
914 if (mmu_idx != MMU_KSMAP_IDX || !(ptep & PG_USER_MASK)) {
916 if ((ptep & PG_RW_MASK) || (!is_user && !(env->cr[0] & CR0_WP_MASK))) {
920 if (!(ptep & PG_NX_MASK) &&
921 (mmu_idx == MMU_USER_IDX ||
922 !((env->cr[4] & CR4_SMEP_MASK) && (ptep & PG_USER_MASK)))) {
925 if ((env->cr[4] & CR4_PKE_MASK) && (env->hflags & HF_LMA_MASK) &&
926 (ptep & PG_USER_MASK) && env->pkru) {
927 uint32_t pk = (pte & PG_PKRU_MASK) >> PG_PKRU_BIT;
928 uint32_t pkru_ad = (env->pkru >> pk * 2) & 1;
929 uint32_t pkru_wd = (env->pkru >> pk * 2) & 2;
930 uint32_t pkru_prot = PAGE_READ | PAGE_WRITE | PAGE_EXEC;
933 pkru_prot &= ~(PAGE_READ | PAGE_WRITE);
934 } else if (pkru_wd && (is_user || env->cr[0] & CR0_WP_MASK)) {
935 pkru_prot &= ~PAGE_WRITE;
939 if ((pkru_prot & (1 << is_write1)) == 0) {
940 assert(is_write1 != 2);
941 error_code |= PG_ERROR_PK_MASK;
942 goto do_fault_protect;
946 if ((prot & (1 << is_write1)) == 0) {
947 goto do_fault_protect;
951 is_dirty = is_write && !(pte & PG_DIRTY_MASK);
952 if (!(pte & PG_ACCESSED_MASK) || is_dirty) {
953 pte |= PG_ACCESSED_MASK;
955 pte |= PG_DIRTY_MASK;
957 x86_stl_phys_notdirty(cs, pte_addr, pte);
960 if (!(pte & PG_DIRTY_MASK)) {
961 /* only set write access if already dirty... otherwise wait
968 pte = pte & env->a20_mask;
970 /* align to page_size */
971 pte &= PG_ADDRESS_MASK & ~(page_size - 1);
973 /* Even if 4MB pages, we map only one 4KB page in the cache to
974 avoid filling it too fast */
975 vaddr = addr & TARGET_PAGE_MASK;
976 page_offset = vaddr & (page_size - 1);
977 paddr = pte + page_offset;
979 assert(prot & (1 << is_write1));
980 tlb_set_page_with_attrs(cs, vaddr, paddr, cpu_get_mem_attrs(env),
981 prot, mmu_idx, page_size);
984 error_code |= PG_ERROR_RSVD_MASK;
986 error_code |= PG_ERROR_P_MASK;
988 error_code |= (is_write << PG_ERROR_W_BIT);
990 error_code |= PG_ERROR_U_MASK;
991 if (is_write1 == 2 &&
992 (((env->efer & MSR_EFER_NXE) &&
993 (env->cr[4] & CR4_PAE_MASK)) ||
994 (env->cr[4] & CR4_SMEP_MASK)))
995 error_code |= PG_ERROR_I_D_MASK;
996 if (env->intercept_exceptions & (1 << EXCP0E_PAGE)) {
997 /* cr2 is not modified in case of exceptions */
999 env->vm_vmcb + offsetof(struct vmcb, control.exit_info_2),
1004 env->error_code = error_code;
1005 cs->exception_index = EXCP0E_PAGE;
1009 hwaddr x86_cpu_get_phys_page_debug(CPUState *cs, vaddr addr)
1011 X86CPU *cpu = X86_CPU(cs);
1012 CPUX86State *env = &cpu->env;
1013 target_ulong pde_addr, pte_addr;
1015 uint32_t page_offset;
1018 if (!(env->cr[0] & CR0_PG_MASK)) {
1019 pte = addr & env->a20_mask;
1021 } else if (env->cr[4] & CR4_PAE_MASK) {
1022 target_ulong pdpe_addr;
1025 #ifdef TARGET_X86_64
1026 if (env->hflags & HF_LMA_MASK) {
1027 uint64_t pml4e_addr, pml4e;
1030 /* test virtual address sign extension */
1031 sext = (int64_t)addr >> 47;
1032 if (sext != 0 && sext != -1) {
1035 pml4e_addr = ((env->cr[3] & ~0xfff) + (((addr >> 39) & 0x1ff) << 3)) &
1037 pml4e = x86_ldq_phys(cs, pml4e_addr);
1038 if (!(pml4e & PG_PRESENT_MASK)) {
1041 pdpe_addr = ((pml4e & PG_ADDRESS_MASK) +
1042 (((addr >> 30) & 0x1ff) << 3)) & env->a20_mask;
1043 pdpe = x86_ldq_phys(cs, pdpe_addr);
1044 if (!(pdpe & PG_PRESENT_MASK)) {
1047 if (pdpe & PG_PSE_MASK) {
1048 page_size = 1024 * 1024 * 1024;
1056 pdpe_addr = ((env->cr[3] & ~0x1f) + ((addr >> 27) & 0x18)) &
1058 pdpe = x86_ldq_phys(cs, pdpe_addr);
1059 if (!(pdpe & PG_PRESENT_MASK))
1063 pde_addr = ((pdpe & PG_ADDRESS_MASK) +
1064 (((addr >> 21) & 0x1ff) << 3)) & env->a20_mask;
1065 pde = x86_ldq_phys(cs, pde_addr);
1066 if (!(pde & PG_PRESENT_MASK)) {
1069 if (pde & PG_PSE_MASK) {
1071 page_size = 2048 * 1024;
1075 pte_addr = ((pde & PG_ADDRESS_MASK) +
1076 (((addr >> 12) & 0x1ff) << 3)) & env->a20_mask;
1078 pte = x86_ldq_phys(cs, pte_addr);
1080 if (!(pte & PG_PRESENT_MASK)) {
1086 /* page directory entry */
1087 pde_addr = ((env->cr[3] & ~0xfff) + ((addr >> 20) & 0xffc)) & env->a20_mask;
1088 pde = x86_ldl_phys(cs, pde_addr);
1089 if (!(pde & PG_PRESENT_MASK))
1091 if ((pde & PG_PSE_MASK) && (env->cr[4] & CR4_PSE_MASK)) {
1092 pte = pde | ((pde & 0x1fe000LL) << (32 - 13));
1093 page_size = 4096 * 1024;
1095 /* page directory entry */
1096 pte_addr = ((pde & ~0xfff) + ((addr >> 10) & 0xffc)) & env->a20_mask;
1097 pte = x86_ldl_phys(cs, pte_addr);
1098 if (!(pte & PG_PRESENT_MASK)) {
1103 pte = pte & env->a20_mask;
1106 #ifdef TARGET_X86_64
1109 pte &= PG_ADDRESS_MASK & ~(page_size - 1);
1110 page_offset = (addr & TARGET_PAGE_MASK) & (page_size - 1);
1111 return pte | page_offset;
1114 typedef struct MCEInjectionParams {
1118 uint64_t mcg_status;
1122 } MCEInjectionParams;
1124 static void do_inject_x86_mce(CPUState *cs, void *data)
1126 MCEInjectionParams *params = data;
1127 X86CPU *cpu = X86_CPU(cs);
1128 CPUX86State *cenv = &cpu->env;
1129 uint64_t *banks = cenv->mce_banks + 4 * params->bank;
1131 cpu_synchronize_state(cs);
1134 * If there is an MCE exception being processed, ignore this SRAO MCE
1135 * unless unconditional injection was requested.
1137 if (!(params->flags & MCE_INJECT_UNCOND_AO)
1138 && !(params->status & MCI_STATUS_AR)
1139 && (cenv->mcg_status & MCG_STATUS_MCIP)) {
1143 if (params->status & MCI_STATUS_UC) {
1145 * if MSR_MCG_CTL is not all 1s, the uncorrected error
1146 * reporting is disabled
1148 if ((cenv->mcg_cap & MCG_CTL_P) && cenv->mcg_ctl != ~(uint64_t)0) {
1149 monitor_printf(params->mon,
1150 "CPU %d: Uncorrected error reporting disabled\n",
1156 * if MSR_MCi_CTL is not all 1s, the uncorrected error
1157 * reporting is disabled for the bank
1159 if (banks[0] != ~(uint64_t)0) {
1160 monitor_printf(params->mon,
1161 "CPU %d: Uncorrected error reporting disabled for"
1163 cs->cpu_index, params->bank);
1167 if ((cenv->mcg_status & MCG_STATUS_MCIP) ||
1168 !(cenv->cr[4] & CR4_MCE_MASK)) {
1169 monitor_printf(params->mon,
1170 "CPU %d: Previous MCE still in progress, raising"
1173 qemu_log_mask(CPU_LOG_RESET, "Triple fault\n");
1174 qemu_system_reset_request();
1177 if (banks[1] & MCI_STATUS_VAL) {
1178 params->status |= MCI_STATUS_OVER;
1180 banks[2] = params->addr;
1181 banks[3] = params->misc;
1182 cenv->mcg_status = params->mcg_status;
1183 banks[1] = params->status;
1184 cpu_interrupt(cs, CPU_INTERRUPT_MCE);
1185 } else if (!(banks[1] & MCI_STATUS_VAL)
1186 || !(banks[1] & MCI_STATUS_UC)) {
1187 if (banks[1] & MCI_STATUS_VAL) {
1188 params->status |= MCI_STATUS_OVER;
1190 banks[2] = params->addr;
1191 banks[3] = params->misc;
1192 banks[1] = params->status;
1194 banks[1] |= MCI_STATUS_OVER;
1198 void cpu_x86_inject_mce(Monitor *mon, X86CPU *cpu, int bank,
1199 uint64_t status, uint64_t mcg_status, uint64_t addr,
1200 uint64_t misc, int flags)
1202 CPUState *cs = CPU(cpu);
1203 CPUX86State *cenv = &cpu->env;
1204 MCEInjectionParams params = {
1208 .mcg_status = mcg_status,
1213 unsigned bank_num = cenv->mcg_cap & 0xff;
1215 if (!cenv->mcg_cap) {
1216 monitor_printf(mon, "MCE injection not supported\n");
1219 if (bank >= bank_num) {
1220 monitor_printf(mon, "Invalid MCE bank number\n");
1223 if (!(status & MCI_STATUS_VAL)) {
1224 monitor_printf(mon, "Invalid MCE status code\n");
1227 if ((flags & MCE_INJECT_BROADCAST)
1228 && !cpu_x86_support_mca_broadcast(cenv)) {
1229 monitor_printf(mon, "Guest CPU does not support MCA broadcast\n");
1233 run_on_cpu(cs, do_inject_x86_mce, ¶ms);
1234 if (flags & MCE_INJECT_BROADCAST) {
1238 params.status = MCI_STATUS_VAL | MCI_STATUS_UC;
1239 params.mcg_status = MCG_STATUS_MCIP | MCG_STATUS_RIPV;
1242 CPU_FOREACH(other_cs) {
1243 if (other_cs == cs) {
1246 run_on_cpu(other_cs, do_inject_x86_mce, ¶ms);
1251 void cpu_report_tpr_access(CPUX86State *env, TPRAccess access)
1253 X86CPU *cpu = x86_env_get_cpu(env);
1254 CPUState *cs = CPU(cpu);
1256 if (kvm_enabled()) {
1257 env->tpr_access_type = access;
1259 cpu_interrupt(cs, CPU_INTERRUPT_TPR);
1261 cpu_restore_state(cs, cs->mem_io_pc);
1263 apic_handle_tpr_access_report(cpu->apic_state, env->eip, access);
1266 #endif /* !CONFIG_USER_ONLY */
1268 int cpu_x86_get_descr_debug(CPUX86State *env, unsigned int selector,
1269 target_ulong *base, unsigned int *limit,
1270 unsigned int *flags)
1272 X86CPU *cpu = x86_env_get_cpu(env);
1273 CPUState *cs = CPU(cpu);
1283 index = selector & ~7;
1284 ptr = dt->base + index;
1285 if ((index + 7) > dt->limit
1286 || cpu_memory_rw_debug(cs, ptr, (uint8_t *)&e1, sizeof(e1), 0) != 0
1287 || cpu_memory_rw_debug(cs, ptr+4, (uint8_t *)&e2, sizeof(e2), 0) != 0)
1290 *base = ((e1 >> 16) | ((e2 & 0xff) << 16) | (e2 & 0xff000000));
1291 *limit = (e1 & 0xffff) | (e2 & 0x000f0000);
1292 if (e2 & DESC_G_MASK)
1293 *limit = (*limit << 12) | 0xfff;
1299 #if !defined(CONFIG_USER_ONLY)
1300 void do_cpu_init(X86CPU *cpu)
1302 CPUState *cs = CPU(cpu);
1303 CPUX86State *env = &cpu->env;
1304 CPUX86State *save = g_new(CPUX86State, 1);
1305 int sipi = cs->interrupt_request & CPU_INTERRUPT_SIPI;
1310 cs->interrupt_request = sipi;
1311 memcpy(&env->start_init_save, &save->start_init_save,
1312 offsetof(CPUX86State, end_init_save) -
1313 offsetof(CPUX86State, start_init_save));
1316 if (kvm_enabled()) {
1317 kvm_arch_do_init_vcpu(cpu);
1319 apic_init_reset(cpu->apic_state);
1322 void do_cpu_sipi(X86CPU *cpu)
1324 apic_sipi(cpu->apic_state);
1327 void do_cpu_init(X86CPU *cpu)
1330 void do_cpu_sipi(X86CPU *cpu)
1335 /* Frob eflags into and out of the CPU temporary format. */
1337 void x86_cpu_exec_enter(CPUState *cs)
1339 X86CPU *cpu = X86_CPU(cs);
1340 CPUX86State *env = &cpu->env;
1342 CC_SRC = env->eflags & (CC_O | CC_S | CC_Z | CC_A | CC_P | CC_C);
1343 env->df = 1 - (2 * ((env->eflags >> 10) & 1));
1344 CC_OP = CC_OP_EFLAGS;
1345 env->eflags &= ~(DF_MASK | CC_O | CC_S | CC_Z | CC_A | CC_P | CC_C);
1348 void x86_cpu_exec_exit(CPUState *cs)
1350 X86CPU *cpu = X86_CPU(cs);
1351 CPUX86State *env = &cpu->env;
1353 env->eflags = cpu_compute_eflags(env);
1356 #ifndef CONFIG_USER_ONLY
1357 uint8_t x86_ldub_phys(CPUState *cs, hwaddr addr)
1359 X86CPU *cpu = X86_CPU(cs);
1360 CPUX86State *env = &cpu->env;
1362 return address_space_ldub(cs->as, addr,
1363 cpu_get_mem_attrs(env),
1367 uint32_t x86_lduw_phys(CPUState *cs, hwaddr addr)
1369 X86CPU *cpu = X86_CPU(cs);
1370 CPUX86State *env = &cpu->env;
1372 return address_space_lduw(cs->as, addr,
1373 cpu_get_mem_attrs(env),
1377 uint32_t x86_ldl_phys(CPUState *cs, hwaddr addr)
1379 X86CPU *cpu = X86_CPU(cs);
1380 CPUX86State *env = &cpu->env;
1382 return address_space_ldl(cs->as, addr,
1383 cpu_get_mem_attrs(env),
1387 uint64_t x86_ldq_phys(CPUState *cs, hwaddr addr)
1389 X86CPU *cpu = X86_CPU(cs);
1390 CPUX86State *env = &cpu->env;
1392 return address_space_ldq(cs->as, addr,
1393 cpu_get_mem_attrs(env),
1397 void x86_stb_phys(CPUState *cs, hwaddr addr, uint8_t val)
1399 X86CPU *cpu = X86_CPU(cs);
1400 CPUX86State *env = &cpu->env;
1402 address_space_stb(cs->as, addr, val,
1403 cpu_get_mem_attrs(env),
1407 void x86_stl_phys_notdirty(CPUState *cs, hwaddr addr, uint32_t val)
1409 X86CPU *cpu = X86_CPU(cs);
1410 CPUX86State *env = &cpu->env;
1412 address_space_stl_notdirty(cs->as, addr, val,
1413 cpu_get_mem_attrs(env),
1417 void x86_stw_phys(CPUState *cs, hwaddr addr, uint32_t val)
1419 X86CPU *cpu = X86_CPU(cs);
1420 CPUX86State *env = &cpu->env;
1422 address_space_stw(cs->as, addr, val,
1423 cpu_get_mem_attrs(env),
1427 void x86_stl_phys(CPUState *cs, hwaddr addr, uint32_t val)
1429 X86CPU *cpu = X86_CPU(cs);
1430 CPUX86State *env = &cpu->env;
1432 address_space_stl(cs->as, addr, val,
1433 cpu_get_mem_attrs(env),
1437 void x86_stq_phys(CPUState *cs, hwaddr addr, uint64_t val)
1439 X86CPU *cpu = X86_CPU(cs);
1440 CPUX86State *env = &cpu->env;
1442 address_space_stq(cs->as, addr, val,
1443 cpu_get_mem_attrs(env),