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"
23 #ifndef CONFIG_USER_ONLY
24 #include "sysemu/sysemu.h"
25 #include "monitor/monitor.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 #define DUMP_CODE_BYTES_TOTAL 50
183 #define DUMP_CODE_BYTES_BACKWARD 20
185 void x86_cpu_dump_state(CPUState
*cs
, FILE *f
, fprintf_function cpu_fprintf
,
188 X86CPU
*cpu
= X86_CPU(cs
);
189 CPUX86State
*env
= &cpu
->env
;
192 static const char *seg_name
[6] = { "ES", "CS", "SS", "DS", "FS", "GS" };
194 eflags
= cpu_compute_eflags(env
);
196 if (env
->hflags
& HF_CS64_MASK
) {
198 "RAX=%016" PRIx64
" RBX=%016" PRIx64
" RCX=%016" PRIx64
" RDX=%016" PRIx64
"\n"
199 "RSI=%016" PRIx64
" RDI=%016" PRIx64
" RBP=%016" PRIx64
" RSP=%016" PRIx64
"\n"
200 "R8 =%016" PRIx64
" R9 =%016" PRIx64
" R10=%016" PRIx64
" R11=%016" PRIx64
"\n"
201 "R12=%016" PRIx64
" R13=%016" PRIx64
" R14=%016" PRIx64
" R15=%016" PRIx64
"\n"
202 "RIP=%016" PRIx64
" RFL=%08x [%c%c%c%c%c%c%c] CPL=%d II=%d A20=%d SMM=%d HLT=%d\n",
220 eflags
& DF_MASK
? 'D' : '-',
221 eflags
& CC_O
? 'O' : '-',
222 eflags
& CC_S
? 'S' : '-',
223 eflags
& CC_Z
? 'Z' : '-',
224 eflags
& CC_A
? 'A' : '-',
225 eflags
& CC_P
? 'P' : '-',
226 eflags
& CC_C
? 'C' : '-',
227 env
->hflags
& HF_CPL_MASK
,
228 (env
->hflags
>> HF_INHIBIT_IRQ_SHIFT
) & 1,
229 (env
->a20_mask
>> 20) & 1,
230 (env
->hflags
>> HF_SMM_SHIFT
) & 1,
235 cpu_fprintf(f
, "EAX=%08x EBX=%08x ECX=%08x EDX=%08x\n"
236 "ESI=%08x EDI=%08x EBP=%08x ESP=%08x\n"
237 "EIP=%08x EFL=%08x [%c%c%c%c%c%c%c] CPL=%d II=%d A20=%d SMM=%d HLT=%d\n",
238 (uint32_t)env
->regs
[R_EAX
],
239 (uint32_t)env
->regs
[R_EBX
],
240 (uint32_t)env
->regs
[R_ECX
],
241 (uint32_t)env
->regs
[R_EDX
],
242 (uint32_t)env
->regs
[R_ESI
],
243 (uint32_t)env
->regs
[R_EDI
],
244 (uint32_t)env
->regs
[R_EBP
],
245 (uint32_t)env
->regs
[R_ESP
],
246 (uint32_t)env
->eip
, eflags
,
247 eflags
& DF_MASK
? 'D' : '-',
248 eflags
& CC_O
? 'O' : '-',
249 eflags
& CC_S
? 'S' : '-',
250 eflags
& CC_Z
? 'Z' : '-',
251 eflags
& CC_A
? 'A' : '-',
252 eflags
& CC_P
? 'P' : '-',
253 eflags
& CC_C
? 'C' : '-',
254 env
->hflags
& HF_CPL_MASK
,
255 (env
->hflags
>> HF_INHIBIT_IRQ_SHIFT
) & 1,
256 (env
->a20_mask
>> 20) & 1,
257 (env
->hflags
>> HF_SMM_SHIFT
) & 1,
261 for(i
= 0; i
< 6; i
++) {
262 cpu_x86_dump_seg_cache(env
, f
, cpu_fprintf
, seg_name
[i
],
265 cpu_x86_dump_seg_cache(env
, f
, cpu_fprintf
, "LDT", &env
->ldt
);
266 cpu_x86_dump_seg_cache(env
, f
, cpu_fprintf
, "TR", &env
->tr
);
269 if (env
->hflags
& HF_LMA_MASK
) {
270 cpu_fprintf(f
, "GDT= %016" PRIx64
" %08x\n",
271 env
->gdt
.base
, env
->gdt
.limit
);
272 cpu_fprintf(f
, "IDT= %016" PRIx64
" %08x\n",
273 env
->idt
.base
, env
->idt
.limit
);
274 cpu_fprintf(f
, "CR0=%08x CR2=%016" PRIx64
" CR3=%016" PRIx64
" CR4=%08x\n",
275 (uint32_t)env
->cr
[0],
278 (uint32_t)env
->cr
[4]);
279 for(i
= 0; i
< 4; i
++)
280 cpu_fprintf(f
, "DR%d=%016" PRIx64
" ", i
, env
->dr
[i
]);
281 cpu_fprintf(f
, "\nDR6=%016" PRIx64
" DR7=%016" PRIx64
"\n",
282 env
->dr
[6], env
->dr
[7]);
286 cpu_fprintf(f
, "GDT= %08x %08x\n",
287 (uint32_t)env
->gdt
.base
, env
->gdt
.limit
);
288 cpu_fprintf(f
, "IDT= %08x %08x\n",
289 (uint32_t)env
->idt
.base
, env
->idt
.limit
);
290 cpu_fprintf(f
, "CR0=%08x CR2=%08x CR3=%08x CR4=%08x\n",
291 (uint32_t)env
->cr
[0],
292 (uint32_t)env
->cr
[2],
293 (uint32_t)env
->cr
[3],
294 (uint32_t)env
->cr
[4]);
295 for(i
= 0; i
< 4; i
++) {
296 cpu_fprintf(f
, "DR%d=" TARGET_FMT_lx
" ", i
, env
->dr
[i
]);
298 cpu_fprintf(f
, "\nDR6=" TARGET_FMT_lx
" DR7=" TARGET_FMT_lx
"\n",
299 env
->dr
[6], env
->dr
[7]);
301 if (flags
& CPU_DUMP_CCOP
) {
302 if ((unsigned)env
->cc_op
< CC_OP_NB
)
303 snprintf(cc_op_name
, sizeof(cc_op_name
), "%s", cc_op_str
[env
->cc_op
]);
305 snprintf(cc_op_name
, sizeof(cc_op_name
), "[%d]", env
->cc_op
);
307 if (env
->hflags
& HF_CS64_MASK
) {
308 cpu_fprintf(f
, "CCS=%016" PRIx64
" CCD=%016" PRIx64
" CCO=%-8s\n",
309 env
->cc_src
, env
->cc_dst
,
314 cpu_fprintf(f
, "CCS=%08x CCD=%08x CCO=%-8s\n",
315 (uint32_t)env
->cc_src
, (uint32_t)env
->cc_dst
,
319 cpu_fprintf(f
, "EFER=%016" PRIx64
"\n", env
->efer
);
320 if (flags
& CPU_DUMP_FPU
) {
323 for(i
= 0; i
< 8; i
++) {
324 fptag
|= ((!env
->fptags
[i
]) << i
);
326 cpu_fprintf(f
, "FCW=%04x FSW=%04x [ST=%d] FTW=%02x MXCSR=%08x\n",
328 (env
->fpus
& ~0x3800) | (env
->fpstt
& 0x7) << 11,
334 u
.d
= env
->fpregs
[i
].d
;
335 cpu_fprintf(f
, "FPR%d=%016" PRIx64
" %04x",
336 i
, u
.l
.lower
, u
.l
.upper
);
338 cpu_fprintf(f
, "\n");
342 if (env
->hflags
& HF_CS64_MASK
)
347 cpu_fprintf(f
, "XMM%02d=%08x%08x%08x%08x",
349 env
->xmm_regs
[i
].XMM_L(3),
350 env
->xmm_regs
[i
].XMM_L(2),
351 env
->xmm_regs
[i
].XMM_L(1),
352 env
->xmm_regs
[i
].XMM_L(0));
354 cpu_fprintf(f
, "\n");
359 if (flags
& CPU_DUMP_CODE
) {
360 target_ulong base
= env
->segs
[R_CS
].base
+ env
->eip
;
361 target_ulong offs
= MIN(env
->eip
, DUMP_CODE_BYTES_BACKWARD
);
365 cpu_fprintf(f
, "Code=");
366 for (i
= 0; i
< DUMP_CODE_BYTES_TOTAL
; i
++) {
367 if (cpu_memory_rw_debug(cs
, base
- offs
+ i
, &code
, 1, 0) == 0) {
368 snprintf(codestr
, sizeof(codestr
), "%02x", code
);
370 snprintf(codestr
, sizeof(codestr
), "??");
372 cpu_fprintf(f
, "%s%s%s%s", i
> 0 ? " " : "",
373 i
== offs
? "<" : "", codestr
, i
== offs
? ">" : "");
375 cpu_fprintf(f
, "\n");
379 /***********************************************************/
381 /* XXX: add PGE support */
383 void x86_cpu_set_a20(X86CPU
*cpu
, int a20_state
)
385 CPUX86State
*env
= &cpu
->env
;
387 a20_state
= (a20_state
!= 0);
388 if (a20_state
!= ((env
->a20_mask
>> 20) & 1)) {
389 CPUState
*cs
= CPU(cpu
);
391 #if defined(DEBUG_MMU)
392 printf("A20 update: a20=%d\n", a20_state
);
394 /* if the cpu is currently executing code, we must unlink it and
395 all the potentially executing TB */
396 cpu_interrupt(cs
, CPU_INTERRUPT_EXITTB
);
398 /* when a20 is changed, all the MMU mappings are invalid, so
399 we must flush everything */
401 env
->a20_mask
= ~(1 << 20) | (a20_state
<< 20);
405 void cpu_x86_update_cr0(CPUX86State
*env
, uint32_t new_cr0
)
407 X86CPU
*cpu
= x86_env_get_cpu(env
);
410 #if defined(DEBUG_MMU)
411 printf("CR0 update: CR0=0x%08x\n", new_cr0
);
413 if ((new_cr0
& (CR0_PG_MASK
| CR0_WP_MASK
| CR0_PE_MASK
)) !=
414 (env
->cr
[0] & (CR0_PG_MASK
| CR0_WP_MASK
| CR0_PE_MASK
))) {
415 tlb_flush(CPU(cpu
), 1);
419 if (!(env
->cr
[0] & CR0_PG_MASK
) && (new_cr0
& CR0_PG_MASK
) &&
420 (env
->efer
& MSR_EFER_LME
)) {
421 /* enter in long mode */
422 /* XXX: generate an exception */
423 if (!(env
->cr
[4] & CR4_PAE_MASK
))
425 env
->efer
|= MSR_EFER_LMA
;
426 env
->hflags
|= HF_LMA_MASK
;
427 } else if ((env
->cr
[0] & CR0_PG_MASK
) && !(new_cr0
& CR0_PG_MASK
) &&
428 (env
->efer
& MSR_EFER_LMA
)) {
430 env
->efer
&= ~MSR_EFER_LMA
;
431 env
->hflags
&= ~(HF_LMA_MASK
| HF_CS64_MASK
);
432 env
->eip
&= 0xffffffff;
435 env
->cr
[0] = new_cr0
| CR0_ET_MASK
;
437 /* update PE flag in hidden flags */
438 pe_state
= (env
->cr
[0] & CR0_PE_MASK
);
439 env
->hflags
= (env
->hflags
& ~HF_PE_MASK
) | (pe_state
<< HF_PE_SHIFT
);
440 /* ensure that ADDSEG is always set in real mode */
441 env
->hflags
|= ((pe_state
^ 1) << HF_ADDSEG_SHIFT
);
442 /* update FPU flags */
443 env
->hflags
= (env
->hflags
& ~(HF_MP_MASK
| HF_EM_MASK
| HF_TS_MASK
)) |
444 ((new_cr0
<< (HF_MP_SHIFT
- 1)) & (HF_MP_MASK
| HF_EM_MASK
| HF_TS_MASK
));
447 /* XXX: in legacy PAE mode, generate a GPF if reserved bits are set in
449 void cpu_x86_update_cr3(CPUX86State
*env
, target_ulong new_cr3
)
451 X86CPU
*cpu
= x86_env_get_cpu(env
);
453 env
->cr
[3] = new_cr3
;
454 if (env
->cr
[0] & CR0_PG_MASK
) {
455 #if defined(DEBUG_MMU)
456 printf("CR3 update: CR3=" TARGET_FMT_lx
"\n", new_cr3
);
458 tlb_flush(CPU(cpu
), 0);
462 void cpu_x86_update_cr4(CPUX86State
*env
, uint32_t new_cr4
)
464 X86CPU
*cpu
= x86_env_get_cpu(env
);
466 #if defined(DEBUG_MMU)
467 printf("CR4 update: CR4=%08x\n", (uint32_t)env
->cr
[4]);
469 if ((new_cr4
^ env
->cr
[4]) &
470 (CR4_PGE_MASK
| CR4_PAE_MASK
| CR4_PSE_MASK
|
471 CR4_SMEP_MASK
| CR4_SMAP_MASK
)) {
472 tlb_flush(CPU(cpu
), 1);
475 if (!(env
->features
[FEAT_1_EDX
] & CPUID_SSE
)) {
476 new_cr4
&= ~CR4_OSFXSR_MASK
;
478 env
->hflags
&= ~HF_OSFXSR_MASK
;
479 if (new_cr4
& CR4_OSFXSR_MASK
) {
480 env
->hflags
|= HF_OSFXSR_MASK
;
483 if (!(env
->features
[FEAT_7_0_EBX
] & CPUID_7_0_EBX_SMAP
)) {
484 new_cr4
&= ~CR4_SMAP_MASK
;
486 env
->hflags
&= ~HF_SMAP_MASK
;
487 if (new_cr4
& CR4_SMAP_MASK
) {
488 env
->hflags
|= HF_SMAP_MASK
;
491 env
->cr
[4] = new_cr4
;
494 #if defined(CONFIG_USER_ONLY)
496 int x86_cpu_handle_mmu_fault(CPUState
*cs
, vaddr addr
,
497 int is_write
, int mmu_idx
)
499 X86CPU
*cpu
= X86_CPU(cs
);
500 CPUX86State
*env
= &cpu
->env
;
502 /* user mode only emulation */
505 env
->error_code
= (is_write
<< PG_ERROR_W_BIT
);
506 env
->error_code
|= PG_ERROR_U_MASK
;
507 cs
->exception_index
= EXCP0E_PAGE
;
514 * -1 = cannot handle fault
515 * 0 = nothing more to do
516 * 1 = generate PF fault
518 int x86_cpu_handle_mmu_fault(CPUState
*cs
, vaddr addr
,
519 int is_write1
, int mmu_idx
)
521 X86CPU
*cpu
= X86_CPU(cs
);
522 CPUX86State
*env
= &cpu
->env
;
524 target_ulong pde_addr
, pte_addr
;
526 int is_dirty
, prot
, page_size
, is_write
, is_user
;
528 uint64_t rsvd_mask
= PG_HI_RSVD_MASK
;
529 uint32_t page_offset
;
532 is_user
= mmu_idx
== MMU_USER_IDX
;
533 #if defined(DEBUG_MMU)
534 printf("MMU fault: addr=%" VADDR_PRIx
" w=%d u=%d eip=" TARGET_FMT_lx
"\n",
535 addr
, is_write1
, is_user
, env
->eip
);
537 is_write
= is_write1
& 1;
539 if (!(env
->cr
[0] & CR0_PG_MASK
)) {
542 if (!(env
->hflags
& HF_LMA_MASK
)) {
543 /* Without long mode we can only address 32bits in real mode */
547 prot
= PAGE_READ
| PAGE_WRITE
| PAGE_EXEC
;
552 if (!(env
->efer
& MSR_EFER_NXE
)) {
553 rsvd_mask
|= PG_NX_MASK
;
556 if (env
->cr
[4] & CR4_PAE_MASK
) {
558 target_ulong pdpe_addr
;
561 if (env
->hflags
& HF_LMA_MASK
) {
562 uint64_t pml4e_addr
, pml4e
;
565 /* test virtual address sign extension */
566 sext
= (int64_t)addr
>> 47;
567 if (sext
!= 0 && sext
!= -1) {
569 cs
->exception_index
= EXCP0D_GPF
;
573 pml4e_addr
= ((env
->cr
[3] & ~0xfff) + (((addr
>> 39) & 0x1ff) << 3)) &
575 pml4e
= ldq_phys(cs
->as
, pml4e_addr
);
576 if (!(pml4e
& PG_PRESENT_MASK
)) {
579 if (pml4e
& (rsvd_mask
| PG_PSE_MASK
)) {
582 if (!(pml4e
& PG_ACCESSED_MASK
)) {
583 pml4e
|= PG_ACCESSED_MASK
;
584 stl_phys_notdirty(cs
->as
, pml4e_addr
, pml4e
);
586 ptep
= pml4e
^ PG_NX_MASK
;
587 pdpe_addr
= ((pml4e
& PG_ADDRESS_MASK
) + (((addr
>> 30) & 0x1ff) << 3)) &
589 pdpe
= ldq_phys(cs
->as
, pdpe_addr
);
590 if (!(pdpe
& PG_PRESENT_MASK
)) {
593 if (pdpe
& rsvd_mask
) {
596 ptep
&= pdpe
^ PG_NX_MASK
;
597 if (!(pdpe
& PG_ACCESSED_MASK
)) {
598 pdpe
|= PG_ACCESSED_MASK
;
599 stl_phys_notdirty(cs
->as
, pdpe_addr
, pdpe
);
601 if (pdpe
& PG_PSE_MASK
) {
603 page_size
= 1024 * 1024 * 1024;
604 pte_addr
= pdpe_addr
;
606 goto do_check_protect
;
611 /* XXX: load them when cr3 is loaded ? */
612 pdpe_addr
= ((env
->cr
[3] & ~0x1f) + ((addr
>> 27) & 0x18)) &
614 pdpe
= ldq_phys(cs
->as
, pdpe_addr
);
615 if (!(pdpe
& PG_PRESENT_MASK
)) {
618 rsvd_mask
|= PG_HI_USER_MASK
| PG_NX_MASK
;
619 if (pdpe
& rsvd_mask
) {
622 ptep
= PG_NX_MASK
| PG_USER_MASK
| PG_RW_MASK
;
625 pde_addr
= ((pdpe
& PG_ADDRESS_MASK
) + (((addr
>> 21) & 0x1ff) << 3)) &
627 pde
= ldq_phys(cs
->as
, pde_addr
);
628 if (!(pde
& PG_PRESENT_MASK
)) {
631 if (pde
& rsvd_mask
) {
634 ptep
&= pde
^ PG_NX_MASK
;
635 if (pde
& PG_PSE_MASK
) {
637 page_size
= 2048 * 1024;
640 goto do_check_protect
;
643 if (!(pde
& PG_ACCESSED_MASK
)) {
644 pde
|= PG_ACCESSED_MASK
;
645 stl_phys_notdirty(cs
->as
, pde_addr
, pde
);
647 pte_addr
= ((pde
& PG_ADDRESS_MASK
) + (((addr
>> 12) & 0x1ff) << 3)) &
649 pte
= ldq_phys(cs
->as
, pte_addr
);
650 if (!(pte
& PG_PRESENT_MASK
)) {
653 if (pte
& rsvd_mask
) {
656 /* combine pde and pte nx, user and rw protections */
657 ptep
&= pte
^ PG_NX_MASK
;
662 /* page directory entry */
663 pde_addr
= ((env
->cr
[3] & ~0xfff) + ((addr
>> 20) & 0xffc)) &
665 pde
= ldl_phys(cs
->as
, pde_addr
);
666 if (!(pde
& PG_PRESENT_MASK
)) {
669 ptep
= pde
| PG_NX_MASK
;
671 /* if PSE bit is set, then we use a 4MB page */
672 if ((pde
& PG_PSE_MASK
) && (env
->cr
[4] & CR4_PSE_MASK
)) {
673 page_size
= 4096 * 1024;
676 /* Bits 20-13 provide bits 39-32 of the address, bit 21 is reserved.
677 * Leave bits 20-13 in place for setting accessed/dirty bits below.
679 pte
= pde
| ((pde
& 0x1fe000) << (32 - 13));
680 rsvd_mask
= 0x200000;
681 goto do_check_protect_pse36
;
684 if (!(pde
& PG_ACCESSED_MASK
)) {
685 pde
|= PG_ACCESSED_MASK
;
686 stl_phys_notdirty(cs
->as
, pde_addr
, pde
);
689 /* page directory entry */
690 pte_addr
= ((pde
& ~0xfff) + ((addr
>> 10) & 0xffc)) &
692 pte
= ldl_phys(cs
->as
, pte_addr
);
693 if (!(pte
& PG_PRESENT_MASK
)) {
696 /* combine pde and pte user and rw protections */
697 ptep
&= pte
| PG_NX_MASK
;
703 rsvd_mask
|= (page_size
- 1) & PG_ADDRESS_MASK
& ~PG_PSE_PAT_MASK
;
704 do_check_protect_pse36
:
705 if (pte
& rsvd_mask
) {
709 if ((ptep
& PG_NX_MASK
) && is_write1
== 2) {
710 goto do_fault_protect
;
714 if (!(ptep
& PG_USER_MASK
)) {
715 goto do_fault_protect
;
717 if (is_write
&& !(ptep
& PG_RW_MASK
)) {
718 goto do_fault_protect
;
723 if (is_write1
!= 2 && (ptep
& PG_USER_MASK
)) {
724 goto do_fault_protect
;
727 case MMU_KNOSMAP_IDX
:
728 if (is_write1
== 2 && (env
->cr
[4] & CR4_SMEP_MASK
) &&
729 (ptep
& PG_USER_MASK
)) {
730 goto do_fault_protect
;
732 if ((env
->cr
[0] & CR0_WP_MASK
) &&
733 is_write
&& !(ptep
& PG_RW_MASK
)) {
734 goto do_fault_protect
;
738 default: /* cannot happen */
741 is_dirty
= is_write
&& !(pte
& PG_DIRTY_MASK
);
742 if (!(pte
& PG_ACCESSED_MASK
) || is_dirty
) {
743 pte
|= PG_ACCESSED_MASK
;
745 pte
|= PG_DIRTY_MASK
;
747 stl_phys_notdirty(cs
->as
, pte_addr
, pte
);
750 /* the page can be put in the TLB */
752 if (!(ptep
& PG_NX_MASK
) &&
753 (mmu_idx
== MMU_USER_IDX
||
754 !((env
->cr
[4] & CR4_SMEP_MASK
) && (ptep
& PG_USER_MASK
)))) {
757 if (pte
& PG_DIRTY_MASK
) {
758 /* only set write access if already dirty... otherwise wait
761 if (ptep
& PG_RW_MASK
)
764 if (!(env
->cr
[0] & CR0_WP_MASK
) ||
770 pte
= pte
& env
->a20_mask
;
772 /* align to page_size */
773 pte
&= PG_ADDRESS_MASK
& ~(page_size
- 1);
775 /* Even if 4MB pages, we map only one 4KB page in the cache to
776 avoid filling it too fast */
777 vaddr
= addr
& TARGET_PAGE_MASK
;
778 page_offset
= vaddr
& (page_size
- 1);
779 paddr
= pte
+ page_offset
;
781 tlb_set_page(cs
, vaddr
, paddr
, prot
, mmu_idx
, page_size
);
784 error_code
|= PG_ERROR_RSVD_MASK
;
786 error_code
|= PG_ERROR_P_MASK
;
788 error_code
|= (is_write
<< PG_ERROR_W_BIT
);
790 error_code
|= PG_ERROR_U_MASK
;
791 if (is_write1
== 2 &&
792 (((env
->efer
& MSR_EFER_NXE
) &&
793 (env
->cr
[4] & CR4_PAE_MASK
)) ||
794 (env
->cr
[4] & CR4_SMEP_MASK
)))
795 error_code
|= PG_ERROR_I_D_MASK
;
796 if (env
->intercept_exceptions
& (1 << EXCP0E_PAGE
)) {
797 /* cr2 is not modified in case of exceptions */
799 env
->vm_vmcb
+ offsetof(struct vmcb
, control
.exit_info_2
),
804 env
->error_code
= error_code
;
805 cs
->exception_index
= EXCP0E_PAGE
;
809 hwaddr
x86_cpu_get_phys_page_debug(CPUState
*cs
, vaddr addr
)
811 X86CPU
*cpu
= X86_CPU(cs
);
812 CPUX86State
*env
= &cpu
->env
;
813 target_ulong pde_addr
, pte_addr
;
815 uint32_t page_offset
;
818 if (!(env
->cr
[0] & CR0_PG_MASK
)) {
819 pte
= addr
& env
->a20_mask
;
821 } else if (env
->cr
[4] & CR4_PAE_MASK
) {
822 target_ulong pdpe_addr
;
826 if (env
->hflags
& HF_LMA_MASK
) {
827 uint64_t pml4e_addr
, pml4e
;
830 /* test virtual address sign extension */
831 sext
= (int64_t)addr
>> 47;
832 if (sext
!= 0 && sext
!= -1) {
835 pml4e_addr
= ((env
->cr
[3] & ~0xfff) + (((addr
>> 39) & 0x1ff) << 3)) &
837 pml4e
= ldq_phys(cs
->as
, pml4e_addr
);
838 if (!(pml4e
& PG_PRESENT_MASK
)) {
841 pdpe_addr
= ((pml4e
& PG_ADDRESS_MASK
) +
842 (((addr
>> 30) & 0x1ff) << 3)) & env
->a20_mask
;
843 pdpe
= ldq_phys(cs
->as
, pdpe_addr
);
844 if (!(pdpe
& PG_PRESENT_MASK
)) {
847 if (pdpe
& PG_PSE_MASK
) {
848 page_size
= 1024 * 1024 * 1024;
856 pdpe_addr
= ((env
->cr
[3] & ~0x1f) + ((addr
>> 27) & 0x18)) &
858 pdpe
= ldq_phys(cs
->as
, pdpe_addr
);
859 if (!(pdpe
& PG_PRESENT_MASK
))
863 pde_addr
= ((pdpe
& PG_ADDRESS_MASK
) +
864 (((addr
>> 21) & 0x1ff) << 3)) & env
->a20_mask
;
865 pde
= ldq_phys(cs
->as
, pde_addr
);
866 if (!(pde
& PG_PRESENT_MASK
)) {
869 if (pde
& PG_PSE_MASK
) {
871 page_size
= 2048 * 1024;
875 pte_addr
= ((pde
& PG_ADDRESS_MASK
) +
876 (((addr
>> 12) & 0x1ff) << 3)) & env
->a20_mask
;
878 pte
= ldq_phys(cs
->as
, pte_addr
);
880 if (!(pte
& PG_PRESENT_MASK
)) {
886 /* page directory entry */
887 pde_addr
= ((env
->cr
[3] & ~0xfff) + ((addr
>> 20) & 0xffc)) & env
->a20_mask
;
888 pde
= ldl_phys(cs
->as
, pde_addr
);
889 if (!(pde
& PG_PRESENT_MASK
))
891 if ((pde
& PG_PSE_MASK
) && (env
->cr
[4] & CR4_PSE_MASK
)) {
892 pte
= pde
| ((pde
& 0x1fe000) << (32 - 13));
893 page_size
= 4096 * 1024;
895 /* page directory entry */
896 pte_addr
= ((pde
& ~0xfff) + ((addr
>> 10) & 0xffc)) & env
->a20_mask
;
897 pte
= ldl_phys(cs
->as
, pte_addr
);
898 if (!(pte
& PG_PRESENT_MASK
)) {
903 pte
= pte
& env
->a20_mask
;
909 pte
&= PG_ADDRESS_MASK
& ~(page_size
- 1);
910 page_offset
= (addr
& TARGET_PAGE_MASK
) & (page_size
- 1);
911 return pte
| page_offset
;
914 void hw_breakpoint_insert(CPUX86State
*env
, int index
)
916 CPUState
*cs
= CPU(x86_env_get_cpu(env
));
917 int type
= 0, err
= 0;
919 switch (hw_breakpoint_type(env
->dr
[7], index
)) {
920 case DR7_TYPE_BP_INST
:
921 if (hw_breakpoint_enabled(env
->dr
[7], index
)) {
922 err
= cpu_breakpoint_insert(cs
, env
->dr
[index
], BP_CPU
,
923 &env
->cpu_breakpoint
[index
]);
926 case DR7_TYPE_DATA_WR
:
927 type
= BP_CPU
| BP_MEM_WRITE
;
930 /* No support for I/O watchpoints yet */
932 case DR7_TYPE_DATA_RW
:
933 type
= BP_CPU
| BP_MEM_ACCESS
;
938 err
= cpu_watchpoint_insert(cs
, env
->dr
[index
],
939 hw_breakpoint_len(env
->dr
[7], index
),
940 type
, &env
->cpu_watchpoint
[index
]);
944 env
->cpu_breakpoint
[index
] = NULL
;
948 void hw_breakpoint_remove(CPUX86State
*env
, int index
)
952 if (!env
->cpu_breakpoint
[index
]) {
955 cs
= CPU(x86_env_get_cpu(env
));
956 switch (hw_breakpoint_type(env
->dr
[7], index
)) {
957 case DR7_TYPE_BP_INST
:
958 if (hw_breakpoint_enabled(env
->dr
[7], index
)) {
959 cpu_breakpoint_remove_by_ref(cs
, env
->cpu_breakpoint
[index
]);
962 case DR7_TYPE_DATA_WR
:
963 case DR7_TYPE_DATA_RW
:
964 cpu_watchpoint_remove_by_ref(cs
, env
->cpu_watchpoint
[index
]);
967 /* No support for I/O watchpoints yet */
972 bool check_hw_breakpoints(CPUX86State
*env
, bool force_dr6_update
)
976 bool hit_enabled
= false;
978 dr6
= env
->dr
[6] & ~0xf;
979 for (reg
= 0; reg
< DR7_MAX_BP
; reg
++) {
980 bool bp_match
= false;
981 bool wp_match
= false;
983 switch (hw_breakpoint_type(env
->dr
[7], reg
)) {
984 case DR7_TYPE_BP_INST
:
985 if (env
->dr
[reg
] == env
->eip
) {
989 case DR7_TYPE_DATA_WR
:
990 case DR7_TYPE_DATA_RW
:
991 if (env
->cpu_watchpoint
[reg
] &&
992 env
->cpu_watchpoint
[reg
]->flags
& BP_WATCHPOINT_HIT
) {
999 if (bp_match
|| wp_match
) {
1001 if (hw_breakpoint_enabled(env
->dr
[7], reg
)) {
1007 if (hit_enabled
|| force_dr6_update
) {
1014 void breakpoint_handler(CPUX86State
*env
)
1016 CPUState
*cs
= CPU(x86_env_get_cpu(env
));
1019 if (cs
->watchpoint_hit
) {
1020 if (cs
->watchpoint_hit
->flags
& BP_CPU
) {
1021 cs
->watchpoint_hit
= NULL
;
1022 if (check_hw_breakpoints(env
, false)) {
1023 raise_exception(env
, EXCP01_DB
);
1025 cpu_resume_from_signal(cs
, NULL
);
1029 QTAILQ_FOREACH(bp
, &cs
->breakpoints
, entry
) {
1030 if (bp
->pc
== env
->eip
) {
1031 if (bp
->flags
& BP_CPU
) {
1032 check_hw_breakpoints(env
, true);
1033 raise_exception(env
, EXCP01_DB
);
1041 typedef struct MCEInjectionParams
{
1046 uint64_t mcg_status
;
1050 } MCEInjectionParams
;
1052 static void do_inject_x86_mce(void *data
)
1054 MCEInjectionParams
*params
= data
;
1055 CPUX86State
*cenv
= ¶ms
->cpu
->env
;
1056 CPUState
*cpu
= CPU(params
->cpu
);
1057 uint64_t *banks
= cenv
->mce_banks
+ 4 * params
->bank
;
1059 cpu_synchronize_state(cpu
);
1062 * If there is an MCE exception being processed, ignore this SRAO MCE
1063 * unless unconditional injection was requested.
1065 if (!(params
->flags
& MCE_INJECT_UNCOND_AO
)
1066 && !(params
->status
& MCI_STATUS_AR
)
1067 && (cenv
->mcg_status
& MCG_STATUS_MCIP
)) {
1071 if (params
->status
& MCI_STATUS_UC
) {
1073 * if MSR_MCG_CTL is not all 1s, the uncorrected error
1074 * reporting is disabled
1076 if ((cenv
->mcg_cap
& MCG_CTL_P
) && cenv
->mcg_ctl
!= ~(uint64_t)0) {
1077 monitor_printf(params
->mon
,
1078 "CPU %d: Uncorrected error reporting disabled\n",
1084 * if MSR_MCi_CTL is not all 1s, the uncorrected error
1085 * reporting is disabled for the bank
1087 if (banks
[0] != ~(uint64_t)0) {
1088 monitor_printf(params
->mon
,
1089 "CPU %d: Uncorrected error reporting disabled for"
1091 cpu
->cpu_index
, params
->bank
);
1095 if ((cenv
->mcg_status
& MCG_STATUS_MCIP
) ||
1096 !(cenv
->cr
[4] & CR4_MCE_MASK
)) {
1097 monitor_printf(params
->mon
,
1098 "CPU %d: Previous MCE still in progress, raising"
1101 qemu_log_mask(CPU_LOG_RESET
, "Triple fault\n");
1102 qemu_system_reset_request();
1105 if (banks
[1] & MCI_STATUS_VAL
) {
1106 params
->status
|= MCI_STATUS_OVER
;
1108 banks
[2] = params
->addr
;
1109 banks
[3] = params
->misc
;
1110 cenv
->mcg_status
= params
->mcg_status
;
1111 banks
[1] = params
->status
;
1112 cpu_interrupt(cpu
, CPU_INTERRUPT_MCE
);
1113 } else if (!(banks
[1] & MCI_STATUS_VAL
)
1114 || !(banks
[1] & MCI_STATUS_UC
)) {
1115 if (banks
[1] & MCI_STATUS_VAL
) {
1116 params
->status
|= MCI_STATUS_OVER
;
1118 banks
[2] = params
->addr
;
1119 banks
[3] = params
->misc
;
1120 banks
[1] = params
->status
;
1122 banks
[1] |= MCI_STATUS_OVER
;
1126 void cpu_x86_inject_mce(Monitor
*mon
, X86CPU
*cpu
, int bank
,
1127 uint64_t status
, uint64_t mcg_status
, uint64_t addr
,
1128 uint64_t misc
, int flags
)
1130 CPUState
*cs
= CPU(cpu
);
1131 CPUX86State
*cenv
= &cpu
->env
;
1132 MCEInjectionParams params
= {
1137 .mcg_status
= mcg_status
,
1142 unsigned bank_num
= cenv
->mcg_cap
& 0xff;
1144 if (!cenv
->mcg_cap
) {
1145 monitor_printf(mon
, "MCE injection not supported\n");
1148 if (bank
>= bank_num
) {
1149 monitor_printf(mon
, "Invalid MCE bank number\n");
1152 if (!(status
& MCI_STATUS_VAL
)) {
1153 monitor_printf(mon
, "Invalid MCE status code\n");
1156 if ((flags
& MCE_INJECT_BROADCAST
)
1157 && !cpu_x86_support_mca_broadcast(cenv
)) {
1158 monitor_printf(mon
, "Guest CPU does not support MCA broadcast\n");
1162 run_on_cpu(cs
, do_inject_x86_mce
, ¶ms
);
1163 if (flags
& MCE_INJECT_BROADCAST
) {
1167 params
.status
= MCI_STATUS_VAL
| MCI_STATUS_UC
;
1168 params
.mcg_status
= MCG_STATUS_MCIP
| MCG_STATUS_RIPV
;
1171 CPU_FOREACH(other_cs
) {
1172 if (other_cs
== cs
) {
1175 params
.cpu
= X86_CPU(other_cs
);
1176 run_on_cpu(other_cs
, do_inject_x86_mce
, ¶ms
);
1181 void cpu_report_tpr_access(CPUX86State
*env
, TPRAccess access
)
1183 X86CPU
*cpu
= x86_env_get_cpu(env
);
1184 CPUState
*cs
= CPU(cpu
);
1186 if (kvm_enabled()) {
1187 env
->tpr_access_type
= access
;
1189 cpu_interrupt(cs
, CPU_INTERRUPT_TPR
);
1191 cpu_restore_state(cs
, cs
->mem_io_pc
);
1193 apic_handle_tpr_access_report(cpu
->apic_state
, env
->eip
, access
);
1196 #endif /* !CONFIG_USER_ONLY */
1198 int cpu_x86_get_descr_debug(CPUX86State
*env
, unsigned int selector
,
1199 target_ulong
*base
, unsigned int *limit
,
1200 unsigned int *flags
)
1202 X86CPU
*cpu
= x86_env_get_cpu(env
);
1203 CPUState
*cs
= CPU(cpu
);
1213 index
= selector
& ~7;
1214 ptr
= dt
->base
+ index
;
1215 if ((index
+ 7) > dt
->limit
1216 || cpu_memory_rw_debug(cs
, ptr
, (uint8_t *)&e1
, sizeof(e1
), 0) != 0
1217 || cpu_memory_rw_debug(cs
, ptr
+4, (uint8_t *)&e2
, sizeof(e2
), 0) != 0)
1220 *base
= ((e1
>> 16) | ((e2
& 0xff) << 16) | (e2
& 0xff000000));
1221 *limit
= (e1
& 0xffff) | (e2
& 0x000f0000);
1222 if (e2
& DESC_G_MASK
)
1223 *limit
= (*limit
<< 12) | 0xfff;
1229 #if !defined(CONFIG_USER_ONLY)
1230 void do_cpu_init(X86CPU
*cpu
)
1232 CPUState
*cs
= CPU(cpu
);
1233 CPUX86State
*env
= &cpu
->env
;
1234 CPUX86State
*save
= g_new(CPUX86State
, 1);
1235 int sipi
= cs
->interrupt_request
& CPU_INTERRUPT_SIPI
;
1240 cs
->interrupt_request
= sipi
;
1241 memcpy(&env
->start_init_save
, &save
->start_init_save
,
1242 offsetof(CPUX86State
, end_init_save
) -
1243 offsetof(CPUX86State
, start_init_save
));
1246 if (kvm_enabled()) {
1247 kvm_arch_do_init_vcpu(cpu
);
1249 apic_init_reset(cpu
->apic_state
);
1252 void do_cpu_sipi(X86CPU
*cpu
)
1254 apic_sipi(cpu
->apic_state
);
1257 void do_cpu_init(X86CPU
*cpu
)
1260 void do_cpu_sipi(X86CPU
*cpu
)