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/>.
21 #include "sysemu/kvm.h"
22 #ifndef CONFIG_USER_ONLY
23 #include "sysemu/sysemu.h"
24 #include "monitor/monitor.h"
29 static void cpu_x86_version(CPUX86State
*env
, int *family
, int *model
)
31 int cpuver
= env
->cpuid_version
;
33 if (family
== NULL
|| model
== NULL
) {
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
)
47 cpu_x86_version(env
, &family
, &model
);
48 if ((family
== 6 && model
>= 14) || family
> 6) {
55 /***********************************************************/
58 static const char *cc_op_str
[CC_OP_NB
] = {
125 cpu_x86_dump_seg_cache(CPUX86State
*env
, FILE *f
, fprintf_function cpu_fprintf
,
126 const char *name
, struct SegmentCache
*sc
)
129 if (env
->hflags
& HF_CS64_MASK
) {
130 cpu_fprintf(f
, "%-3s=%04x %016" PRIx64
" %08x %08x", name
,
131 sc
->selector
, sc
->base
, sc
->limit
, sc
->flags
& 0x00ffff00);
135 cpu_fprintf(f
, "%-3s=%04x %08x %08x %08x", name
, sc
->selector
,
136 (uint32_t)sc
->base
, sc
->limit
, sc
->flags
& 0x00ffff00);
139 if (!(env
->hflags
& HF_PE_MASK
) || !(sc
->flags
& DESC_P_MASK
))
142 cpu_fprintf(f
, " DPL=%d ", (sc
->flags
& DESC_DPL_MASK
) >> DESC_DPL_SHIFT
);
143 if (sc
->flags
& DESC_S_MASK
) {
144 if (sc
->flags
& DESC_CS_MASK
) {
145 cpu_fprintf(f
, (sc
->flags
& DESC_L_MASK
) ? "CS64" :
146 ((sc
->flags
& DESC_B_MASK
) ? "CS32" : "CS16"));
147 cpu_fprintf(f
, " [%c%c", (sc
->flags
& DESC_C_MASK
) ? 'C' : '-',
148 (sc
->flags
& DESC_R_MASK
) ? 'R' : '-');
151 (sc
->flags
& DESC_B_MASK
|| env
->hflags
& HF_LMA_MASK
)
153 cpu_fprintf(f
, " [%c%c", (sc
->flags
& DESC_E_MASK
) ? 'E' : '-',
154 (sc
->flags
& DESC_W_MASK
) ? 'W' : '-');
156 cpu_fprintf(f
, "%c]", (sc
->flags
& DESC_A_MASK
) ? 'A' : '-');
158 static const char *sys_type_name
[2][16] = {
160 "Reserved", "TSS16-avl", "LDT", "TSS16-busy",
161 "CallGate16", "TaskGate", "IntGate16", "TrapGate16",
162 "Reserved", "TSS32-avl", "Reserved", "TSS32-busy",
163 "CallGate32", "Reserved", "IntGate32", "TrapGate32"
166 "<hiword>", "Reserved", "LDT", "Reserved", "Reserved",
167 "Reserved", "Reserved", "Reserved", "Reserved",
168 "TSS64-avl", "Reserved", "TSS64-busy", "CallGate64",
169 "Reserved", "IntGate64", "TrapGate64"
173 sys_type_name
[(env
->hflags
& HF_LMA_MASK
) ? 1 : 0]
174 [(sc
->flags
& DESC_TYPE_MASK
)
175 >> DESC_TYPE_SHIFT
]);
178 cpu_fprintf(f
, "\n");
181 #define DUMP_CODE_BYTES_TOTAL 50
182 #define DUMP_CODE_BYTES_BACKWARD 20
184 void x86_cpu_dump_state(CPUState
*cs
, FILE *f
, fprintf_function cpu_fprintf
,
187 X86CPU
*cpu
= X86_CPU(cs
);
188 CPUX86State
*env
= &cpu
->env
;
191 static const char *seg_name
[6] = { "ES", "CS", "SS", "DS", "FS", "GS" };
193 eflags
= cpu_compute_eflags(env
);
195 if (env
->hflags
& HF_CS64_MASK
) {
197 "RAX=%016" PRIx64
" RBX=%016" PRIx64
" RCX=%016" PRIx64
" RDX=%016" PRIx64
"\n"
198 "RSI=%016" PRIx64
" RDI=%016" PRIx64
" RBP=%016" PRIx64
" RSP=%016" PRIx64
"\n"
199 "R8 =%016" PRIx64
" R9 =%016" PRIx64
" R10=%016" PRIx64
" R11=%016" PRIx64
"\n"
200 "R12=%016" PRIx64
" R13=%016" PRIx64
" R14=%016" PRIx64
" R15=%016" PRIx64
"\n"
201 "RIP=%016" PRIx64
" RFL=%08x [%c%c%c%c%c%c%c] CPL=%d II=%d A20=%d SMM=%d HLT=%d\n",
219 eflags
& DF_MASK
? 'D' : '-',
220 eflags
& CC_O
? 'O' : '-',
221 eflags
& CC_S
? 'S' : '-',
222 eflags
& CC_Z
? 'Z' : '-',
223 eflags
& CC_A
? 'A' : '-',
224 eflags
& CC_P
? 'P' : '-',
225 eflags
& CC_C
? 'C' : '-',
226 env
->hflags
& HF_CPL_MASK
,
227 (env
->hflags
>> HF_INHIBIT_IRQ_SHIFT
) & 1,
228 (env
->a20_mask
>> 20) & 1,
229 (env
->hflags
>> HF_SMM_SHIFT
) & 1,
234 cpu_fprintf(f
, "EAX=%08x EBX=%08x ECX=%08x EDX=%08x\n"
235 "ESI=%08x EDI=%08x EBP=%08x ESP=%08x\n"
236 "EIP=%08x EFL=%08x [%c%c%c%c%c%c%c] CPL=%d II=%d A20=%d SMM=%d HLT=%d\n",
237 (uint32_t)env
->regs
[R_EAX
],
238 (uint32_t)env
->regs
[R_EBX
],
239 (uint32_t)env
->regs
[R_ECX
],
240 (uint32_t)env
->regs
[R_EDX
],
241 (uint32_t)env
->regs
[R_ESI
],
242 (uint32_t)env
->regs
[R_EDI
],
243 (uint32_t)env
->regs
[R_EBP
],
244 (uint32_t)env
->regs
[R_ESP
],
245 (uint32_t)env
->eip
, eflags
,
246 eflags
& DF_MASK
? 'D' : '-',
247 eflags
& CC_O
? 'O' : '-',
248 eflags
& CC_S
? 'S' : '-',
249 eflags
& CC_Z
? 'Z' : '-',
250 eflags
& CC_A
? 'A' : '-',
251 eflags
& CC_P
? 'P' : '-',
252 eflags
& CC_C
? 'C' : '-',
253 env
->hflags
& HF_CPL_MASK
,
254 (env
->hflags
>> HF_INHIBIT_IRQ_SHIFT
) & 1,
255 (env
->a20_mask
>> 20) & 1,
256 (env
->hflags
>> HF_SMM_SHIFT
) & 1,
260 for(i
= 0; i
< 6; i
++) {
261 cpu_x86_dump_seg_cache(env
, f
, cpu_fprintf
, seg_name
[i
],
264 cpu_x86_dump_seg_cache(env
, f
, cpu_fprintf
, "LDT", &env
->ldt
);
265 cpu_x86_dump_seg_cache(env
, f
, cpu_fprintf
, "TR", &env
->tr
);
268 if (env
->hflags
& HF_LMA_MASK
) {
269 cpu_fprintf(f
, "GDT= %016" PRIx64
" %08x\n",
270 env
->gdt
.base
, env
->gdt
.limit
);
271 cpu_fprintf(f
, "IDT= %016" PRIx64
" %08x\n",
272 env
->idt
.base
, env
->idt
.limit
);
273 cpu_fprintf(f
, "CR0=%08x CR2=%016" PRIx64
" CR3=%016" PRIx64
" CR4=%08x\n",
274 (uint32_t)env
->cr
[0],
277 (uint32_t)env
->cr
[4]);
278 for(i
= 0; i
< 4; i
++)
279 cpu_fprintf(f
, "DR%d=%016" PRIx64
" ", i
, env
->dr
[i
]);
280 cpu_fprintf(f
, "\nDR6=%016" PRIx64
" DR7=%016" PRIx64
"\n",
281 env
->dr
[6], env
->dr
[7]);
285 cpu_fprintf(f
, "GDT= %08x %08x\n",
286 (uint32_t)env
->gdt
.base
, env
->gdt
.limit
);
287 cpu_fprintf(f
, "IDT= %08x %08x\n",
288 (uint32_t)env
->idt
.base
, env
->idt
.limit
);
289 cpu_fprintf(f
, "CR0=%08x CR2=%08x CR3=%08x CR4=%08x\n",
290 (uint32_t)env
->cr
[0],
291 (uint32_t)env
->cr
[2],
292 (uint32_t)env
->cr
[3],
293 (uint32_t)env
->cr
[4]);
294 for(i
= 0; i
< 4; i
++) {
295 cpu_fprintf(f
, "DR%d=" TARGET_FMT_lx
" ", i
, env
->dr
[i
]);
297 cpu_fprintf(f
, "\nDR6=" TARGET_FMT_lx
" DR7=" TARGET_FMT_lx
"\n",
298 env
->dr
[6], env
->dr
[7]);
300 if (flags
& CPU_DUMP_CCOP
) {
301 if ((unsigned)env
->cc_op
< CC_OP_NB
)
302 snprintf(cc_op_name
, sizeof(cc_op_name
), "%s", cc_op_str
[env
->cc_op
]);
304 snprintf(cc_op_name
, sizeof(cc_op_name
), "[%d]", env
->cc_op
);
306 if (env
->hflags
& HF_CS64_MASK
) {
307 cpu_fprintf(f
, "CCS=%016" PRIx64
" CCD=%016" PRIx64
" CCO=%-8s\n",
308 env
->cc_src
, env
->cc_dst
,
313 cpu_fprintf(f
, "CCS=%08x CCD=%08x CCO=%-8s\n",
314 (uint32_t)env
->cc_src
, (uint32_t)env
->cc_dst
,
318 cpu_fprintf(f
, "EFER=%016" PRIx64
"\n", env
->efer
);
319 if (flags
& CPU_DUMP_FPU
) {
322 for(i
= 0; i
< 8; i
++) {
323 fptag
|= ((!env
->fptags
[i
]) << i
);
325 cpu_fprintf(f
, "FCW=%04x FSW=%04x [ST=%d] FTW=%02x MXCSR=%08x\n",
327 (env
->fpus
& ~0x3800) | (env
->fpstt
& 0x7) << 11,
333 u
.d
= env
->fpregs
[i
].d
;
334 cpu_fprintf(f
, "FPR%d=%016" PRIx64
" %04x",
335 i
, u
.l
.lower
, u
.l
.upper
);
337 cpu_fprintf(f
, "\n");
341 if (env
->hflags
& HF_CS64_MASK
)
346 cpu_fprintf(f
, "XMM%02d=%08x%08x%08x%08x",
348 env
->xmm_regs
[i
].XMM_L(3),
349 env
->xmm_regs
[i
].XMM_L(2),
350 env
->xmm_regs
[i
].XMM_L(1),
351 env
->xmm_regs
[i
].XMM_L(0));
353 cpu_fprintf(f
, "\n");
358 if (flags
& CPU_DUMP_CODE
) {
359 target_ulong base
= env
->segs
[R_CS
].base
+ env
->eip
;
360 target_ulong offs
= MIN(env
->eip
, DUMP_CODE_BYTES_BACKWARD
);
364 cpu_fprintf(f
, "Code=");
365 for (i
= 0; i
< DUMP_CODE_BYTES_TOTAL
; i
++) {
366 if (cpu_memory_rw_debug(cs
, base
- offs
+ i
, &code
, 1, 0) == 0) {
367 snprintf(codestr
, sizeof(codestr
), "%02x", code
);
369 snprintf(codestr
, sizeof(codestr
), "??");
371 cpu_fprintf(f
, "%s%s%s%s", i
> 0 ? " " : "",
372 i
== offs
? "<" : "", codestr
, i
== offs
? ">" : "");
374 cpu_fprintf(f
, "\n");
378 /***********************************************************/
380 /* XXX: add PGE support */
382 void x86_cpu_set_a20(X86CPU
*cpu
, int a20_state
)
384 CPUX86State
*env
= &cpu
->env
;
386 a20_state
= (a20_state
!= 0);
387 if (a20_state
!= ((env
->a20_mask
>> 20) & 1)) {
388 CPUState
*cs
= CPU(cpu
);
390 #if defined(DEBUG_MMU)
391 printf("A20 update: a20=%d\n", a20_state
);
393 /* if the cpu is currently executing code, we must unlink it and
394 all the potentially executing TB */
395 cpu_interrupt(cs
, CPU_INTERRUPT_EXITTB
);
397 /* when a20 is changed, all the MMU mappings are invalid, so
398 we must flush everything */
400 env
->a20_mask
= ~(1 << 20) | (a20_state
<< 20);
404 void cpu_x86_update_cr0(CPUX86State
*env
, uint32_t new_cr0
)
406 X86CPU
*cpu
= x86_env_get_cpu(env
);
409 #if defined(DEBUG_MMU)
410 printf("CR0 update: CR0=0x%08x\n", new_cr0
);
412 if ((new_cr0
& (CR0_PG_MASK
| CR0_WP_MASK
| CR0_PE_MASK
)) !=
413 (env
->cr
[0] & (CR0_PG_MASK
| CR0_WP_MASK
| CR0_PE_MASK
))) {
414 tlb_flush(CPU(cpu
), 1);
418 if (!(env
->cr
[0] & CR0_PG_MASK
) && (new_cr0
& CR0_PG_MASK
) &&
419 (env
->efer
& MSR_EFER_LME
)) {
420 /* enter in long mode */
421 /* XXX: generate an exception */
422 if (!(env
->cr
[4] & CR4_PAE_MASK
))
424 env
->efer
|= MSR_EFER_LMA
;
425 env
->hflags
|= HF_LMA_MASK
;
426 } else if ((env
->cr
[0] & CR0_PG_MASK
) && !(new_cr0
& CR0_PG_MASK
) &&
427 (env
->efer
& MSR_EFER_LMA
)) {
429 env
->efer
&= ~MSR_EFER_LMA
;
430 env
->hflags
&= ~(HF_LMA_MASK
| HF_CS64_MASK
);
431 env
->eip
&= 0xffffffff;
434 env
->cr
[0] = new_cr0
| CR0_ET_MASK
;
436 /* update PE flag in hidden flags */
437 pe_state
= (env
->cr
[0] & CR0_PE_MASK
);
438 env
->hflags
= (env
->hflags
& ~HF_PE_MASK
) | (pe_state
<< HF_PE_SHIFT
);
439 /* ensure that ADDSEG is always set in real mode */
440 env
->hflags
|= ((pe_state
^ 1) << HF_ADDSEG_SHIFT
);
441 /* update FPU flags */
442 env
->hflags
= (env
->hflags
& ~(HF_MP_MASK
| HF_EM_MASK
| HF_TS_MASK
)) |
443 ((new_cr0
<< (HF_MP_SHIFT
- 1)) & (HF_MP_MASK
| HF_EM_MASK
| HF_TS_MASK
));
446 /* XXX: in legacy PAE mode, generate a GPF if reserved bits are set in
448 void cpu_x86_update_cr3(CPUX86State
*env
, target_ulong new_cr3
)
450 X86CPU
*cpu
= x86_env_get_cpu(env
);
452 env
->cr
[3] = new_cr3
;
453 if (env
->cr
[0] & CR0_PG_MASK
) {
454 #if defined(DEBUG_MMU)
455 printf("CR3 update: CR3=" TARGET_FMT_lx
"\n", new_cr3
);
457 tlb_flush(CPU(cpu
), 0);
461 void cpu_x86_update_cr4(CPUX86State
*env
, uint32_t new_cr4
)
463 X86CPU
*cpu
= x86_env_get_cpu(env
);
465 #if defined(DEBUG_MMU)
466 printf("CR4 update: CR4=%08x\n", (uint32_t)env
->cr
[4]);
468 if ((new_cr4
^ env
->cr
[4]) &
469 (CR4_PGE_MASK
| CR4_PAE_MASK
| CR4_PSE_MASK
|
470 CR4_SMEP_MASK
| CR4_SMAP_MASK
)) {
471 tlb_flush(CPU(cpu
), 1);
474 if (!(env
->features
[FEAT_1_EDX
] & CPUID_SSE
)) {
475 new_cr4
&= ~CR4_OSFXSR_MASK
;
477 env
->hflags
&= ~HF_OSFXSR_MASK
;
478 if (new_cr4
& CR4_OSFXSR_MASK
) {
479 env
->hflags
|= HF_OSFXSR_MASK
;
482 if (!(env
->features
[FEAT_7_0_EBX
] & CPUID_7_0_EBX_SMAP
)) {
483 new_cr4
&= ~CR4_SMAP_MASK
;
485 env
->hflags
&= ~HF_SMAP_MASK
;
486 if (new_cr4
& CR4_SMAP_MASK
) {
487 env
->hflags
|= HF_SMAP_MASK
;
490 env
->cr
[4] = new_cr4
;
493 #if defined(CONFIG_USER_ONLY)
495 int x86_cpu_handle_mmu_fault(CPUState
*cs
, vaddr addr
,
496 int is_write
, int mmu_idx
)
498 X86CPU
*cpu
= X86_CPU(cs
);
499 CPUX86State
*env
= &cpu
->env
;
501 /* user mode only emulation */
504 env
->error_code
= (is_write
<< PG_ERROR_W_BIT
);
505 env
->error_code
|= PG_ERROR_U_MASK
;
506 cs
->exception_index
= EXCP0E_PAGE
;
512 /* XXX: This value should match the one returned by CPUID
514 # if defined(TARGET_X86_64)
515 # define PHYS_ADDR_MASK 0xfffffff000LL
517 # define PHYS_ADDR_MASK 0xffffff000LL
521 * -1 = cannot handle fault
522 * 0 = nothing more to do
523 * 1 = generate PF fault
525 int x86_cpu_handle_mmu_fault(CPUState
*cs
, vaddr addr
,
526 int is_write1
, int mmu_idx
)
528 X86CPU
*cpu
= X86_CPU(cs
);
529 CPUX86State
*env
= &cpu
->env
;
531 target_ulong pde_addr
, pte_addr
;
532 int error_code
, is_dirty
, prot
, page_size
, is_write
, is_user
;
534 uint32_t page_offset
;
535 target_ulong vaddr
, virt_addr
;
537 is_user
= mmu_idx
== MMU_USER_IDX
;
538 #if defined(DEBUG_MMU)
539 printf("MMU fault: addr=%" VADDR_PRIx
" w=%d u=%d eip=" TARGET_FMT_lx
"\n",
540 addr
, is_write1
, is_user
, env
->eip
);
542 is_write
= is_write1
& 1;
544 if (!(env
->cr
[0] & CR0_PG_MASK
)) {
547 if (!(env
->hflags
& HF_LMA_MASK
)) {
548 /* Without long mode we can only address 32bits in real mode */
552 virt_addr
= addr
& TARGET_PAGE_MASK
;
553 prot
= PAGE_READ
| PAGE_WRITE
| PAGE_EXEC
;
558 if (env
->cr
[4] & CR4_PAE_MASK
) {
560 target_ulong pdpe_addr
;
563 if (env
->hflags
& HF_LMA_MASK
) {
564 uint64_t pml4e_addr
, pml4e
;
567 /* test virtual address sign extension */
568 sext
= (int64_t)addr
>> 47;
569 if (sext
!= 0 && sext
!= -1) {
571 cs
->exception_index
= EXCP0D_GPF
;
575 pml4e_addr
= ((env
->cr
[3] & ~0xfff) + (((addr
>> 39) & 0x1ff) << 3)) &
577 pml4e
= ldq_phys(cs
->as
, pml4e_addr
);
578 if (!(pml4e
& PG_PRESENT_MASK
)) {
582 if (!(env
->efer
& MSR_EFER_NXE
) && (pml4e
& PG_NX_MASK
)) {
583 error_code
= PG_ERROR_RSVD_MASK
;
586 if (!(pml4e
& PG_ACCESSED_MASK
)) {
587 pml4e
|= PG_ACCESSED_MASK
;
588 stl_phys_notdirty(cs
->as
, pml4e_addr
, pml4e
);
590 ptep
= pml4e
^ PG_NX_MASK
;
591 pdpe_addr
= ((pml4e
& PHYS_ADDR_MASK
) + (((addr
>> 30) & 0x1ff) << 3)) &
593 pdpe
= ldq_phys(cs
->as
, pdpe_addr
);
594 if (!(pdpe
& PG_PRESENT_MASK
)) {
598 if (!(env
->efer
& MSR_EFER_NXE
) && (pdpe
& PG_NX_MASK
)) {
599 error_code
= PG_ERROR_RSVD_MASK
;
602 ptep
&= pdpe
^ PG_NX_MASK
;
603 if (!(pdpe
& PG_ACCESSED_MASK
)) {
604 pdpe
|= PG_ACCESSED_MASK
;
605 stl_phys_notdirty(cs
->as
, pdpe_addr
, pdpe
);
610 /* XXX: load them when cr3 is loaded ? */
611 pdpe_addr
= ((env
->cr
[3] & ~0x1f) + ((addr
>> 27) & 0x18)) &
613 pdpe
= ldq_phys(cs
->as
, pdpe_addr
);
614 if (!(pdpe
& PG_PRESENT_MASK
)) {
618 ptep
= PG_NX_MASK
| PG_USER_MASK
| PG_RW_MASK
;
621 pde_addr
= ((pdpe
& PHYS_ADDR_MASK
) + (((addr
>> 21) & 0x1ff) << 3)) &
623 pde
= ldq_phys(cs
->as
, pde_addr
);
624 if (!(pde
& PG_PRESENT_MASK
)) {
628 if (!(env
->efer
& MSR_EFER_NXE
) && (pde
& PG_NX_MASK
)) {
629 error_code
= PG_ERROR_RSVD_MASK
;
632 ptep
&= pde
^ PG_NX_MASK
;
633 if (pde
& PG_PSE_MASK
) {
635 page_size
= 2048 * 1024;
637 if ((ptep
& PG_NX_MASK
) && is_write1
== 2) {
638 goto do_fault_protect
;
642 if (!(ptep
& PG_USER_MASK
)) {
643 goto do_fault_protect
;
645 if (is_write
&& !(ptep
& PG_RW_MASK
)) {
646 goto do_fault_protect
;
651 if (is_write1
!= 2 && (env
->cr
[4] & CR4_SMAP_MASK
) &&
652 (ptep
& PG_USER_MASK
)) {
653 goto do_fault_protect
;
657 if (is_write1
== 2 && (env
->cr
[4] & CR4_SMEP_MASK
) &&
658 (ptep
& PG_USER_MASK
)) {
659 goto do_fault_protect
;
661 if ((env
->cr
[0] & CR0_WP_MASK
) &&
662 is_write
&& !(ptep
& PG_RW_MASK
)) {
663 goto do_fault_protect
;
667 default: /* cannot happen */
670 is_dirty
= is_write
&& !(pde
& PG_DIRTY_MASK
);
671 if (!(pde
& PG_ACCESSED_MASK
) || is_dirty
) {
672 pde
|= PG_ACCESSED_MASK
;
674 pde
|= PG_DIRTY_MASK
;
675 stl_phys_notdirty(cs
->as
, pde_addr
, pde
);
677 /* align to page_size */
678 pte
= pde
& ((PHYS_ADDR_MASK
& ~(page_size
- 1)) | 0xfff);
679 virt_addr
= addr
& ~(page_size
- 1);
682 if (!(pde
& PG_ACCESSED_MASK
)) {
683 pde
|= PG_ACCESSED_MASK
;
684 stl_phys_notdirty(cs
->as
, pde_addr
, pde
);
686 pte_addr
= ((pde
& PHYS_ADDR_MASK
) + (((addr
>> 12) & 0x1ff) << 3)) &
688 pte
= ldq_phys(cs
->as
, pte_addr
);
689 if (!(pte
& PG_PRESENT_MASK
)) {
693 if (!(env
->efer
& MSR_EFER_NXE
) && (pte
& PG_NX_MASK
)) {
694 error_code
= PG_ERROR_RSVD_MASK
;
697 /* combine pde and pte nx, user and rw protections */
698 ptep
&= pte
^ PG_NX_MASK
;
700 if ((ptep
& PG_NX_MASK
) && is_write1
== 2)
701 goto do_fault_protect
;
704 if (!(ptep
& PG_USER_MASK
)) {
705 goto do_fault_protect
;
707 if (is_write
&& !(ptep
& PG_RW_MASK
)) {
708 goto do_fault_protect
;
713 if (is_write1
!= 2 && (env
->cr
[4] & CR4_SMAP_MASK
) &&
714 (ptep
& PG_USER_MASK
)) {
715 goto do_fault_protect
;
719 if (is_write1
== 2 && (env
->cr
[4] & CR4_SMEP_MASK
) &&
720 (ptep
& PG_USER_MASK
)) {
721 goto do_fault_protect
;
723 if ((env
->cr
[0] & CR0_WP_MASK
) &&
724 is_write
&& !(ptep
& PG_RW_MASK
)) {
725 goto do_fault_protect
;
729 default: /* cannot happen */
732 is_dirty
= is_write
&& !(pte
& PG_DIRTY_MASK
);
733 if (!(pte
& PG_ACCESSED_MASK
) || is_dirty
) {
734 pte
|= PG_ACCESSED_MASK
;
736 pte
|= PG_DIRTY_MASK
;
737 stl_phys_notdirty(cs
->as
, pte_addr
, pte
);
740 virt_addr
= addr
& ~0xfff;
741 pte
= pte
& (PHYS_ADDR_MASK
| 0xfff);
746 /* page directory entry */
747 pde_addr
= ((env
->cr
[3] & ~0xfff) + ((addr
>> 20) & 0xffc)) &
749 pde
= ldl_phys(cs
->as
, pde_addr
);
750 if (!(pde
& PG_PRESENT_MASK
)) {
754 /* if PSE bit is set, then we use a 4MB page */
755 if ((pde
& PG_PSE_MASK
) && (env
->cr
[4] & CR4_PSE_MASK
)) {
756 page_size
= 4096 * 1024;
759 if (!(pde
& PG_USER_MASK
)) {
760 goto do_fault_protect
;
762 if (is_write
&& !(pde
& PG_RW_MASK
)) {
763 goto do_fault_protect
;
768 if (is_write1
!= 2 && (env
->cr
[4] & CR4_SMAP_MASK
) &&
769 (pde
& PG_USER_MASK
)) {
770 goto do_fault_protect
;
774 if (is_write1
== 2 && (env
->cr
[4] & CR4_SMEP_MASK
) &&
775 (pde
& PG_USER_MASK
)) {
776 goto do_fault_protect
;
778 if ((env
->cr
[0] & CR0_WP_MASK
) &&
779 is_write
&& !(pde
& PG_RW_MASK
)) {
780 goto do_fault_protect
;
784 default: /* cannot happen */
787 is_dirty
= is_write
&& !(pde
& PG_DIRTY_MASK
);
788 if (!(pde
& PG_ACCESSED_MASK
) || is_dirty
) {
789 pde
|= PG_ACCESSED_MASK
;
791 pde
|= PG_DIRTY_MASK
;
792 stl_phys_notdirty(cs
->as
, pde_addr
, pde
);
795 pte
= pde
& ~( (page_size
- 1) & ~0xfff); /* align to page_size */
797 virt_addr
= addr
& ~(page_size
- 1);
799 if (!(pde
& PG_ACCESSED_MASK
)) {
800 pde
|= PG_ACCESSED_MASK
;
801 stl_phys_notdirty(cs
->as
, pde_addr
, pde
);
804 /* page directory entry */
805 pte_addr
= ((pde
& ~0xfff) + ((addr
>> 10) & 0xffc)) &
807 pte
= ldl_phys(cs
->as
, pte_addr
);
808 if (!(pte
& PG_PRESENT_MASK
)) {
812 /* combine pde and pte user and rw protections */
816 if (!(ptep
& PG_USER_MASK
)) {
817 goto do_fault_protect
;
819 if (is_write
&& !(ptep
& PG_RW_MASK
)) {
820 goto do_fault_protect
;
825 if (is_write1
!= 2 && (env
->cr
[4] & CR4_SMAP_MASK
) &&
826 (ptep
& PG_USER_MASK
)) {
827 goto do_fault_protect
;
831 if (is_write1
== 2 && (env
->cr
[4] & CR4_SMEP_MASK
) &&
832 (ptep
& PG_USER_MASK
)) {
833 goto do_fault_protect
;
835 if ((env
->cr
[0] & CR0_WP_MASK
) &&
836 is_write
&& !(ptep
& PG_RW_MASK
)) {
837 goto do_fault_protect
;
841 default: /* cannot happen */
844 is_dirty
= is_write
&& !(pte
& PG_DIRTY_MASK
);
845 if (!(pte
& PG_ACCESSED_MASK
) || is_dirty
) {
846 pte
|= PG_ACCESSED_MASK
;
848 pte
|= PG_DIRTY_MASK
;
849 stl_phys_notdirty(cs
->as
, pte_addr
, pte
);
852 virt_addr
= addr
& ~0xfff;
855 /* the page can be put in the TLB */
857 if (!(ptep
& PG_NX_MASK
))
859 if (pte
& PG_DIRTY_MASK
) {
860 /* only set write access if already dirty... otherwise wait
863 if (ptep
& PG_RW_MASK
)
866 if (!(env
->cr
[0] & CR0_WP_MASK
) ||
872 pte
= pte
& env
->a20_mask
;
874 /* Even if 4MB pages, we map only one 4KB page in the cache to
875 avoid filling it too fast */
876 page_offset
= (addr
& TARGET_PAGE_MASK
) & (page_size
- 1);
877 paddr
= (pte
& TARGET_PAGE_MASK
) + page_offset
;
878 vaddr
= virt_addr
+ page_offset
;
880 tlb_set_page(cs
, vaddr
, paddr
, prot
, mmu_idx
, page_size
);
883 error_code
= PG_ERROR_P_MASK
;
885 error_code
|= (is_write
<< PG_ERROR_W_BIT
);
887 error_code
|= PG_ERROR_U_MASK
;
888 if (is_write1
== 2 &&
889 (((env
->efer
& MSR_EFER_NXE
) &&
890 (env
->cr
[4] & CR4_PAE_MASK
)) ||
891 (env
->cr
[4] & CR4_SMEP_MASK
)))
892 error_code
|= PG_ERROR_I_D_MASK
;
893 if (env
->intercept_exceptions
& (1 << EXCP0E_PAGE
)) {
894 /* cr2 is not modified in case of exceptions */
896 env
->vm_vmcb
+ offsetof(struct vmcb
, control
.exit_info_2
),
901 env
->error_code
= error_code
;
902 cs
->exception_index
= EXCP0E_PAGE
;
906 hwaddr
x86_cpu_get_phys_page_debug(CPUState
*cs
, vaddr addr
)
908 X86CPU
*cpu
= X86_CPU(cs
);
909 CPUX86State
*env
= &cpu
->env
;
910 target_ulong pde_addr
, pte_addr
;
913 uint32_t page_offset
;
916 if (!(env
->cr
[0] & CR0_PG_MASK
)) {
917 pte
= addr
& env
->a20_mask
;
919 } else if (env
->cr
[4] & CR4_PAE_MASK
) {
920 target_ulong pdpe_addr
;
924 if (env
->hflags
& HF_LMA_MASK
) {
925 uint64_t pml4e_addr
, pml4e
;
928 /* test virtual address sign extension */
929 sext
= (int64_t)addr
>> 47;
930 if (sext
!= 0 && sext
!= -1)
933 pml4e_addr
= ((env
->cr
[3] & ~0xfff) + (((addr
>> 39) & 0x1ff) << 3)) &
935 pml4e
= ldq_phys(cs
->as
, pml4e_addr
);
936 if (!(pml4e
& PG_PRESENT_MASK
))
939 pdpe_addr
= ((pml4e
& ~0xfff & ~(PG_NX_MASK
| PG_HI_USER_MASK
)) +
940 (((addr
>> 30) & 0x1ff) << 3)) & env
->a20_mask
;
941 pdpe
= ldq_phys(cs
->as
, pdpe_addr
);
942 if (!(pdpe
& PG_PRESENT_MASK
))
945 if (pdpe
& PG_PSE_MASK
) {
946 page_size
= 1024 * 1024 * 1024;
947 pte
= pdpe
& ~( (page_size
- 1) & ~0xfff);
948 pte
&= ~(PG_NX_MASK
| PG_HI_USER_MASK
);
955 pdpe_addr
= ((env
->cr
[3] & ~0x1f) + ((addr
>> 27) & 0x18)) &
957 pdpe
= ldq_phys(cs
->as
, pdpe_addr
);
958 if (!(pdpe
& PG_PRESENT_MASK
))
962 pde_addr
= ((pdpe
& ~0xfff & ~(PG_NX_MASK
| PG_HI_USER_MASK
)) +
963 (((addr
>> 21) & 0x1ff) << 3)) & env
->a20_mask
;
964 pde
= ldq_phys(cs
->as
, pde_addr
);
965 if (!(pde
& PG_PRESENT_MASK
)) {
968 if (pde
& PG_PSE_MASK
) {
970 page_size
= 2048 * 1024;
971 pte
= pde
& ~( (page_size
- 1) & ~0xfff); /* align to page_size */
974 pte_addr
= ((pde
& ~0xfff & ~(PG_NX_MASK
| PG_HI_USER_MASK
)) +
975 (((addr
>> 12) & 0x1ff) << 3)) & env
->a20_mask
;
977 pte
= ldq_phys(cs
->as
, pte_addr
);
979 pte
&= ~(PG_NX_MASK
| PG_HI_USER_MASK
);
980 if (!(pte
& PG_PRESENT_MASK
))
985 /* page directory entry */
986 pde_addr
= ((env
->cr
[3] & ~0xfff) + ((addr
>> 20) & 0xffc)) & env
->a20_mask
;
987 pde
= ldl_phys(cs
->as
, pde_addr
);
988 if (!(pde
& PG_PRESENT_MASK
))
990 if ((pde
& PG_PSE_MASK
) && (env
->cr
[4] & CR4_PSE_MASK
)) {
991 pte
= pde
& ~0x003ff000; /* align to 4MB */
992 page_size
= 4096 * 1024;
994 /* page directory entry */
995 pte_addr
= ((pde
& ~0xfff) + ((addr
>> 10) & 0xffc)) & env
->a20_mask
;
996 pte
= ldl_phys(cs
->as
, pte_addr
);
997 if (!(pte
& PG_PRESENT_MASK
))
1001 pte
= pte
& env
->a20_mask
;
1004 #ifdef TARGET_X86_64
1007 page_offset
= (addr
& TARGET_PAGE_MASK
) & (page_size
- 1);
1008 paddr
= (pte
& TARGET_PAGE_MASK
) + page_offset
;
1012 void hw_breakpoint_insert(CPUX86State
*env
, int index
)
1014 CPUState
*cs
= CPU(x86_env_get_cpu(env
));
1015 int type
= 0, err
= 0;
1017 switch (hw_breakpoint_type(env
->dr
[7], index
)) {
1018 case DR7_TYPE_BP_INST
:
1019 if (hw_breakpoint_enabled(env
->dr
[7], index
)) {
1020 err
= cpu_breakpoint_insert(cs
, env
->dr
[index
], BP_CPU
,
1021 &env
->cpu_breakpoint
[index
]);
1024 case DR7_TYPE_DATA_WR
:
1025 type
= BP_CPU
| BP_MEM_WRITE
;
1027 case DR7_TYPE_IO_RW
:
1028 /* No support for I/O watchpoints yet */
1030 case DR7_TYPE_DATA_RW
:
1031 type
= BP_CPU
| BP_MEM_ACCESS
;
1036 err
= cpu_watchpoint_insert(cs
, env
->dr
[index
],
1037 hw_breakpoint_len(env
->dr
[7], index
),
1038 type
, &env
->cpu_watchpoint
[index
]);
1042 env
->cpu_breakpoint
[index
] = NULL
;
1046 void hw_breakpoint_remove(CPUX86State
*env
, int index
)
1050 if (!env
->cpu_breakpoint
[index
]) {
1053 cs
= CPU(x86_env_get_cpu(env
));
1054 switch (hw_breakpoint_type(env
->dr
[7], index
)) {
1055 case DR7_TYPE_BP_INST
:
1056 if (hw_breakpoint_enabled(env
->dr
[7], index
)) {
1057 cpu_breakpoint_remove_by_ref(cs
, env
->cpu_breakpoint
[index
]);
1060 case DR7_TYPE_DATA_WR
:
1061 case DR7_TYPE_DATA_RW
:
1062 cpu_watchpoint_remove_by_ref(cs
, env
->cpu_watchpoint
[index
]);
1064 case DR7_TYPE_IO_RW
:
1065 /* No support for I/O watchpoints yet */
1070 bool check_hw_breakpoints(CPUX86State
*env
, bool force_dr6_update
)
1074 bool hit_enabled
= false;
1076 dr6
= env
->dr
[6] & ~0xf;
1077 for (reg
= 0; reg
< DR7_MAX_BP
; reg
++) {
1078 bool bp_match
= false;
1079 bool wp_match
= false;
1081 switch (hw_breakpoint_type(env
->dr
[7], reg
)) {
1082 case DR7_TYPE_BP_INST
:
1083 if (env
->dr
[reg
] == env
->eip
) {
1087 case DR7_TYPE_DATA_WR
:
1088 case DR7_TYPE_DATA_RW
:
1089 if (env
->cpu_watchpoint
[reg
] &&
1090 env
->cpu_watchpoint
[reg
]->flags
& BP_WATCHPOINT_HIT
) {
1094 case DR7_TYPE_IO_RW
:
1097 if (bp_match
|| wp_match
) {
1099 if (hw_breakpoint_enabled(env
->dr
[7], reg
)) {
1105 if (hit_enabled
|| force_dr6_update
) {
1112 void breakpoint_handler(CPUX86State
*env
)
1114 CPUState
*cs
= CPU(x86_env_get_cpu(env
));
1117 if (cs
->watchpoint_hit
) {
1118 if (cs
->watchpoint_hit
->flags
& BP_CPU
) {
1119 cs
->watchpoint_hit
= NULL
;
1120 if (check_hw_breakpoints(env
, false)) {
1121 raise_exception(env
, EXCP01_DB
);
1123 cpu_resume_from_signal(cs
, NULL
);
1127 QTAILQ_FOREACH(bp
, &cs
->breakpoints
, entry
) {
1128 if (bp
->pc
== env
->eip
) {
1129 if (bp
->flags
& BP_CPU
) {
1130 check_hw_breakpoints(env
, true);
1131 raise_exception(env
, EXCP01_DB
);
1139 typedef struct MCEInjectionParams
{
1144 uint64_t mcg_status
;
1148 } MCEInjectionParams
;
1150 static void do_inject_x86_mce(void *data
)
1152 MCEInjectionParams
*params
= data
;
1153 CPUX86State
*cenv
= ¶ms
->cpu
->env
;
1154 CPUState
*cpu
= CPU(params
->cpu
);
1155 uint64_t *banks
= cenv
->mce_banks
+ 4 * params
->bank
;
1157 cpu_synchronize_state(cpu
);
1160 * If there is an MCE exception being processed, ignore this SRAO MCE
1161 * unless unconditional injection was requested.
1163 if (!(params
->flags
& MCE_INJECT_UNCOND_AO
)
1164 && !(params
->status
& MCI_STATUS_AR
)
1165 && (cenv
->mcg_status
& MCG_STATUS_MCIP
)) {
1169 if (params
->status
& MCI_STATUS_UC
) {
1171 * if MSR_MCG_CTL is not all 1s, the uncorrected error
1172 * reporting is disabled
1174 if ((cenv
->mcg_cap
& MCG_CTL_P
) && cenv
->mcg_ctl
!= ~(uint64_t)0) {
1175 monitor_printf(params
->mon
,
1176 "CPU %d: Uncorrected error reporting disabled\n",
1182 * if MSR_MCi_CTL is not all 1s, the uncorrected error
1183 * reporting is disabled for the bank
1185 if (banks
[0] != ~(uint64_t)0) {
1186 monitor_printf(params
->mon
,
1187 "CPU %d: Uncorrected error reporting disabled for"
1189 cpu
->cpu_index
, params
->bank
);
1193 if ((cenv
->mcg_status
& MCG_STATUS_MCIP
) ||
1194 !(cenv
->cr
[4] & CR4_MCE_MASK
)) {
1195 monitor_printf(params
->mon
,
1196 "CPU %d: Previous MCE still in progress, raising"
1199 qemu_log_mask(CPU_LOG_RESET
, "Triple fault\n");
1200 qemu_system_reset_request();
1203 if (banks
[1] & MCI_STATUS_VAL
) {
1204 params
->status
|= MCI_STATUS_OVER
;
1206 banks
[2] = params
->addr
;
1207 banks
[3] = params
->misc
;
1208 cenv
->mcg_status
= params
->mcg_status
;
1209 banks
[1] = params
->status
;
1210 cpu_interrupt(cpu
, CPU_INTERRUPT_MCE
);
1211 } else if (!(banks
[1] & MCI_STATUS_VAL
)
1212 || !(banks
[1] & MCI_STATUS_UC
)) {
1213 if (banks
[1] & MCI_STATUS_VAL
) {
1214 params
->status
|= MCI_STATUS_OVER
;
1216 banks
[2] = params
->addr
;
1217 banks
[3] = params
->misc
;
1218 banks
[1] = params
->status
;
1220 banks
[1] |= MCI_STATUS_OVER
;
1224 void cpu_x86_inject_mce(Monitor
*mon
, X86CPU
*cpu
, int bank
,
1225 uint64_t status
, uint64_t mcg_status
, uint64_t addr
,
1226 uint64_t misc
, int flags
)
1228 CPUState
*cs
= CPU(cpu
);
1229 CPUX86State
*cenv
= &cpu
->env
;
1230 MCEInjectionParams params
= {
1235 .mcg_status
= mcg_status
,
1240 unsigned bank_num
= cenv
->mcg_cap
& 0xff;
1242 if (!cenv
->mcg_cap
) {
1243 monitor_printf(mon
, "MCE injection not supported\n");
1246 if (bank
>= bank_num
) {
1247 monitor_printf(mon
, "Invalid MCE bank number\n");
1250 if (!(status
& MCI_STATUS_VAL
)) {
1251 monitor_printf(mon
, "Invalid MCE status code\n");
1254 if ((flags
& MCE_INJECT_BROADCAST
)
1255 && !cpu_x86_support_mca_broadcast(cenv
)) {
1256 monitor_printf(mon
, "Guest CPU does not support MCA broadcast\n");
1260 run_on_cpu(cs
, do_inject_x86_mce
, ¶ms
);
1261 if (flags
& MCE_INJECT_BROADCAST
) {
1265 params
.status
= MCI_STATUS_VAL
| MCI_STATUS_UC
;
1266 params
.mcg_status
= MCG_STATUS_MCIP
| MCG_STATUS_RIPV
;
1269 CPU_FOREACH(other_cs
) {
1270 if (other_cs
== cs
) {
1273 params
.cpu
= X86_CPU(other_cs
);
1274 run_on_cpu(other_cs
, do_inject_x86_mce
, ¶ms
);
1279 void cpu_report_tpr_access(CPUX86State
*env
, TPRAccess access
)
1281 X86CPU
*cpu
= x86_env_get_cpu(env
);
1282 CPUState
*cs
= CPU(cpu
);
1284 if (kvm_enabled()) {
1285 env
->tpr_access_type
= access
;
1287 cpu_interrupt(cs
, CPU_INTERRUPT_TPR
);
1289 cpu_restore_state(cs
, cs
->mem_io_pc
);
1291 apic_handle_tpr_access_report(cpu
->apic_state
, env
->eip
, access
);
1294 #endif /* !CONFIG_USER_ONLY */
1296 int cpu_x86_get_descr_debug(CPUX86State
*env
, unsigned int selector
,
1297 target_ulong
*base
, unsigned int *limit
,
1298 unsigned int *flags
)
1300 X86CPU
*cpu
= x86_env_get_cpu(env
);
1301 CPUState
*cs
= CPU(cpu
);
1311 index
= selector
& ~7;
1312 ptr
= dt
->base
+ index
;
1313 if ((index
+ 7) > dt
->limit
1314 || cpu_memory_rw_debug(cs
, ptr
, (uint8_t *)&e1
, sizeof(e1
), 0) != 0
1315 || cpu_memory_rw_debug(cs
, ptr
+4, (uint8_t *)&e2
, sizeof(e2
), 0) != 0)
1318 *base
= ((e1
>> 16) | ((e2
& 0xff) << 16) | (e2
& 0xff000000));
1319 *limit
= (e1
& 0xffff) | (e2
& 0x000f0000);
1320 if (e2
& DESC_G_MASK
)
1321 *limit
= (*limit
<< 12) | 0xfff;
1327 #if !defined(CONFIG_USER_ONLY)
1328 void do_cpu_init(X86CPU
*cpu
)
1330 CPUState
*cs
= CPU(cpu
);
1331 CPUX86State
*env
= &cpu
->env
;
1332 int sipi
= cs
->interrupt_request
& CPU_INTERRUPT_SIPI
;
1333 uint64_t pat
= env
->pat
;
1336 cs
->interrupt_request
= sipi
;
1338 apic_init_reset(cpu
->apic_state
);
1341 void do_cpu_sipi(X86CPU
*cpu
)
1343 apic_sipi(cpu
->apic_state
);
1346 void do_cpu_init(X86CPU
*cpu
)
1349 void do_cpu_sipi(X86CPU
*cpu
)