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"
28 static void cpu_x86_version(CPUX86State
*env
, int *family
, int *model
)
30 int cpuver
= env
->cpuid_version
;
32 if (family
== NULL
|| model
== NULL
) {
36 *family
= (cpuver
>> 8) & 0x0f;
37 *model
= ((cpuver
>> 12) & 0xf0) + ((cpuver
>> 4) & 0x0f);
40 /* Broadcast MCA signal for processor version 06H_EH and above */
41 int cpu_x86_support_mca_broadcast(CPUX86State
*env
)
46 cpu_x86_version(env
, &family
, &model
);
47 if ((family
== 6 && model
>= 14) || family
> 6) {
54 /***********************************************************/
57 static const char *cc_op_str
[CC_OP_NB
] = {
124 cpu_x86_dump_seg_cache(CPUX86State
*env
, FILE *f
, fprintf_function cpu_fprintf
,
125 const char *name
, struct SegmentCache
*sc
)
128 if (env
->hflags
& HF_CS64_MASK
) {
129 cpu_fprintf(f
, "%-3s=%04x %016" PRIx64
" %08x %08x", name
,
130 sc
->selector
, sc
->base
, sc
->limit
, sc
->flags
& 0x00ffff00);
134 cpu_fprintf(f
, "%-3s=%04x %08x %08x %08x", name
, sc
->selector
,
135 (uint32_t)sc
->base
, sc
->limit
, sc
->flags
& 0x00ffff00);
138 if (!(env
->hflags
& HF_PE_MASK
) || !(sc
->flags
& DESC_P_MASK
))
141 cpu_fprintf(f
, " DPL=%d ", (sc
->flags
& DESC_DPL_MASK
) >> DESC_DPL_SHIFT
);
142 if (sc
->flags
& DESC_S_MASK
) {
143 if (sc
->flags
& DESC_CS_MASK
) {
144 cpu_fprintf(f
, (sc
->flags
& DESC_L_MASK
) ? "CS64" :
145 ((sc
->flags
& DESC_B_MASK
) ? "CS32" : "CS16"));
146 cpu_fprintf(f
, " [%c%c", (sc
->flags
& DESC_C_MASK
) ? 'C' : '-',
147 (sc
->flags
& DESC_R_MASK
) ? 'R' : '-');
150 (sc
->flags
& DESC_B_MASK
|| env
->hflags
& HF_LMA_MASK
)
152 cpu_fprintf(f
, " [%c%c", (sc
->flags
& DESC_E_MASK
) ? 'E' : '-',
153 (sc
->flags
& DESC_W_MASK
) ? 'W' : '-');
155 cpu_fprintf(f
, "%c]", (sc
->flags
& DESC_A_MASK
) ? 'A' : '-');
157 static const char *sys_type_name
[2][16] = {
159 "Reserved", "TSS16-avl", "LDT", "TSS16-busy",
160 "CallGate16", "TaskGate", "IntGate16", "TrapGate16",
161 "Reserved", "TSS32-avl", "Reserved", "TSS32-busy",
162 "CallGate32", "Reserved", "IntGate32", "TrapGate32"
165 "<hiword>", "Reserved", "LDT", "Reserved", "Reserved",
166 "Reserved", "Reserved", "Reserved", "Reserved",
167 "TSS64-avl", "Reserved", "TSS64-busy", "CallGate64",
168 "Reserved", "IntGate64", "TrapGate64"
172 sys_type_name
[(env
->hflags
& HF_LMA_MASK
) ? 1 : 0]
173 [(sc
->flags
& DESC_TYPE_MASK
)
174 >> DESC_TYPE_SHIFT
]);
177 cpu_fprintf(f
, "\n");
180 #define DUMP_CODE_BYTES_TOTAL 50
181 #define DUMP_CODE_BYTES_BACKWARD 20
183 void x86_cpu_dump_state(CPUState
*cs
, FILE *f
, fprintf_function cpu_fprintf
,
186 X86CPU
*cpu
= X86_CPU(cs
);
187 CPUX86State
*env
= &cpu
->env
;
190 static const char *seg_name
[6] = { "ES", "CS", "SS", "DS", "FS", "GS" };
192 eflags
= cpu_compute_eflags(env
);
194 if (env
->hflags
& HF_CS64_MASK
) {
196 "RAX=%016" PRIx64
" RBX=%016" PRIx64
" RCX=%016" PRIx64
" RDX=%016" PRIx64
"\n"
197 "RSI=%016" PRIx64
" RDI=%016" PRIx64
" RBP=%016" PRIx64
" RSP=%016" PRIx64
"\n"
198 "R8 =%016" PRIx64
" R9 =%016" PRIx64
" R10=%016" PRIx64
" R11=%016" PRIx64
"\n"
199 "R12=%016" PRIx64
" R13=%016" PRIx64
" R14=%016" PRIx64
" R15=%016" PRIx64
"\n"
200 "RIP=%016" PRIx64
" RFL=%08x [%c%c%c%c%c%c%c] CPL=%d II=%d A20=%d SMM=%d HLT=%d\n",
218 eflags
& DF_MASK
? 'D' : '-',
219 eflags
& CC_O
? 'O' : '-',
220 eflags
& CC_S
? 'S' : '-',
221 eflags
& CC_Z
? 'Z' : '-',
222 eflags
& CC_A
? 'A' : '-',
223 eflags
& CC_P
? 'P' : '-',
224 eflags
& CC_C
? 'C' : '-',
225 env
->hflags
& HF_CPL_MASK
,
226 (env
->hflags
>> HF_INHIBIT_IRQ_SHIFT
) & 1,
227 (env
->a20_mask
>> 20) & 1,
228 (env
->hflags
>> HF_SMM_SHIFT
) & 1,
233 cpu_fprintf(f
, "EAX=%08x EBX=%08x ECX=%08x EDX=%08x\n"
234 "ESI=%08x EDI=%08x EBP=%08x ESP=%08x\n"
235 "EIP=%08x EFL=%08x [%c%c%c%c%c%c%c] CPL=%d II=%d A20=%d SMM=%d HLT=%d\n",
236 (uint32_t)env
->regs
[R_EAX
],
237 (uint32_t)env
->regs
[R_EBX
],
238 (uint32_t)env
->regs
[R_ECX
],
239 (uint32_t)env
->regs
[R_EDX
],
240 (uint32_t)env
->regs
[R_ESI
],
241 (uint32_t)env
->regs
[R_EDI
],
242 (uint32_t)env
->regs
[R_EBP
],
243 (uint32_t)env
->regs
[R_ESP
],
244 (uint32_t)env
->eip
, eflags
,
245 eflags
& DF_MASK
? 'D' : '-',
246 eflags
& CC_O
? 'O' : '-',
247 eflags
& CC_S
? 'S' : '-',
248 eflags
& CC_Z
? 'Z' : '-',
249 eflags
& CC_A
? 'A' : '-',
250 eflags
& CC_P
? 'P' : '-',
251 eflags
& CC_C
? 'C' : '-',
252 env
->hflags
& HF_CPL_MASK
,
253 (env
->hflags
>> HF_INHIBIT_IRQ_SHIFT
) & 1,
254 (env
->a20_mask
>> 20) & 1,
255 (env
->hflags
>> HF_SMM_SHIFT
) & 1,
259 for(i
= 0; i
< 6; i
++) {
260 cpu_x86_dump_seg_cache(env
, f
, cpu_fprintf
, seg_name
[i
],
263 cpu_x86_dump_seg_cache(env
, f
, cpu_fprintf
, "LDT", &env
->ldt
);
264 cpu_x86_dump_seg_cache(env
, f
, cpu_fprintf
, "TR", &env
->tr
);
267 if (env
->hflags
& HF_LMA_MASK
) {
268 cpu_fprintf(f
, "GDT= %016" PRIx64
" %08x\n",
269 env
->gdt
.base
, env
->gdt
.limit
);
270 cpu_fprintf(f
, "IDT= %016" PRIx64
" %08x\n",
271 env
->idt
.base
, env
->idt
.limit
);
272 cpu_fprintf(f
, "CR0=%08x CR2=%016" PRIx64
" CR3=%016" PRIx64
" CR4=%08x\n",
273 (uint32_t)env
->cr
[0],
276 (uint32_t)env
->cr
[4]);
277 for(i
= 0; i
< 4; i
++)
278 cpu_fprintf(f
, "DR%d=%016" PRIx64
" ", i
, env
->dr
[i
]);
279 cpu_fprintf(f
, "\nDR6=%016" PRIx64
" DR7=%016" PRIx64
"\n",
280 env
->dr
[6], env
->dr
[7]);
284 cpu_fprintf(f
, "GDT= %08x %08x\n",
285 (uint32_t)env
->gdt
.base
, env
->gdt
.limit
);
286 cpu_fprintf(f
, "IDT= %08x %08x\n",
287 (uint32_t)env
->idt
.base
, env
->idt
.limit
);
288 cpu_fprintf(f
, "CR0=%08x CR2=%08x CR3=%08x CR4=%08x\n",
289 (uint32_t)env
->cr
[0],
290 (uint32_t)env
->cr
[2],
291 (uint32_t)env
->cr
[3],
292 (uint32_t)env
->cr
[4]);
293 for(i
= 0; i
< 4; i
++) {
294 cpu_fprintf(f
, "DR%d=" TARGET_FMT_lx
" ", i
, env
->dr
[i
]);
296 cpu_fprintf(f
, "\nDR6=" TARGET_FMT_lx
" DR7=" TARGET_FMT_lx
"\n",
297 env
->dr
[6], env
->dr
[7]);
299 if (flags
& CPU_DUMP_CCOP
) {
300 if ((unsigned)env
->cc_op
< CC_OP_NB
)
301 snprintf(cc_op_name
, sizeof(cc_op_name
), "%s", cc_op_str
[env
->cc_op
]);
303 snprintf(cc_op_name
, sizeof(cc_op_name
), "[%d]", env
->cc_op
);
305 if (env
->hflags
& HF_CS64_MASK
) {
306 cpu_fprintf(f
, "CCS=%016" PRIx64
" CCD=%016" PRIx64
" CCO=%-8s\n",
307 env
->cc_src
, env
->cc_dst
,
312 cpu_fprintf(f
, "CCS=%08x CCD=%08x CCO=%-8s\n",
313 (uint32_t)env
->cc_src
, (uint32_t)env
->cc_dst
,
317 cpu_fprintf(f
, "EFER=%016" PRIx64
"\n", env
->efer
);
318 if (flags
& CPU_DUMP_FPU
) {
321 for(i
= 0; i
< 8; i
++) {
322 fptag
|= ((!env
->fptags
[i
]) << i
);
324 cpu_fprintf(f
, "FCW=%04x FSW=%04x [ST=%d] FTW=%02x MXCSR=%08x\n",
326 (env
->fpus
& ~0x3800) | (env
->fpstt
& 0x7) << 11,
332 u
.d
= env
->fpregs
[i
].d
;
333 cpu_fprintf(f
, "FPR%d=%016" PRIx64
" %04x",
334 i
, u
.l
.lower
, u
.l
.upper
);
336 cpu_fprintf(f
, "\n");
340 if (env
->hflags
& HF_CS64_MASK
)
345 cpu_fprintf(f
, "XMM%02d=%08x%08x%08x%08x",
347 env
->xmm_regs
[i
].XMM_L(3),
348 env
->xmm_regs
[i
].XMM_L(2),
349 env
->xmm_regs
[i
].XMM_L(1),
350 env
->xmm_regs
[i
].XMM_L(0));
352 cpu_fprintf(f
, "\n");
357 if (flags
& CPU_DUMP_CODE
) {
358 target_ulong base
= env
->segs
[R_CS
].base
+ env
->eip
;
359 target_ulong offs
= MIN(env
->eip
, DUMP_CODE_BYTES_BACKWARD
);
363 cpu_fprintf(f
, "Code=");
364 for (i
= 0; i
< DUMP_CODE_BYTES_TOTAL
; i
++) {
365 if (cpu_memory_rw_debug(cs
, base
- offs
+ i
, &code
, 1, 0) == 0) {
366 snprintf(codestr
, sizeof(codestr
), "%02x", code
);
368 snprintf(codestr
, sizeof(codestr
), "??");
370 cpu_fprintf(f
, "%s%s%s%s", i
> 0 ? " " : "",
371 i
== offs
? "<" : "", codestr
, i
== offs
? ">" : "");
373 cpu_fprintf(f
, "\n");
377 /***********************************************************/
379 /* XXX: add PGE support */
381 void x86_cpu_set_a20(X86CPU
*cpu
, int a20_state
)
383 CPUX86State
*env
= &cpu
->env
;
385 a20_state
= (a20_state
!= 0);
386 if (a20_state
!= ((env
->a20_mask
>> 20) & 1)) {
387 CPUState
*cs
= CPU(cpu
);
389 qemu_log_mask(CPU_LOG_MMU
, "A20 update: a20=%d\n", a20_state
);
390 /* if the cpu is currently executing code, we must unlink it and
391 all the potentially executing TB */
392 cpu_interrupt(cs
, CPU_INTERRUPT_EXITTB
);
394 /* when a20 is changed, all the MMU mappings are invalid, so
395 we must flush everything */
397 env
->a20_mask
= ~(1 << 20) | (a20_state
<< 20);
401 void cpu_x86_update_cr0(CPUX86State
*env
, uint32_t new_cr0
)
403 X86CPU
*cpu
= x86_env_get_cpu(env
);
406 qemu_log_mask(CPU_LOG_MMU
, "CR0 update: CR0=0x%08x\n", new_cr0
);
407 if ((new_cr0
& (CR0_PG_MASK
| CR0_WP_MASK
| CR0_PE_MASK
)) !=
408 (env
->cr
[0] & (CR0_PG_MASK
| CR0_WP_MASK
| CR0_PE_MASK
))) {
409 tlb_flush(CPU(cpu
), 1);
413 if (!(env
->cr
[0] & CR0_PG_MASK
) && (new_cr0
& CR0_PG_MASK
) &&
414 (env
->efer
& MSR_EFER_LME
)) {
415 /* enter in long mode */
416 /* XXX: generate an exception */
417 if (!(env
->cr
[4] & CR4_PAE_MASK
))
419 env
->efer
|= MSR_EFER_LMA
;
420 env
->hflags
|= HF_LMA_MASK
;
421 } else if ((env
->cr
[0] & CR0_PG_MASK
) && !(new_cr0
& CR0_PG_MASK
) &&
422 (env
->efer
& MSR_EFER_LMA
)) {
424 env
->efer
&= ~MSR_EFER_LMA
;
425 env
->hflags
&= ~(HF_LMA_MASK
| HF_CS64_MASK
);
426 env
->eip
&= 0xffffffff;
429 env
->cr
[0] = new_cr0
| CR0_ET_MASK
;
431 /* update PE flag in hidden flags */
432 pe_state
= (env
->cr
[0] & CR0_PE_MASK
);
433 env
->hflags
= (env
->hflags
& ~HF_PE_MASK
) | (pe_state
<< HF_PE_SHIFT
);
434 /* ensure that ADDSEG is always set in real mode */
435 env
->hflags
|= ((pe_state
^ 1) << HF_ADDSEG_SHIFT
);
436 /* update FPU flags */
437 env
->hflags
= (env
->hflags
& ~(HF_MP_MASK
| HF_EM_MASK
| HF_TS_MASK
)) |
438 ((new_cr0
<< (HF_MP_SHIFT
- 1)) & (HF_MP_MASK
| HF_EM_MASK
| HF_TS_MASK
));
441 /* XXX: in legacy PAE mode, generate a GPF if reserved bits are set in
443 void cpu_x86_update_cr3(CPUX86State
*env
, target_ulong new_cr3
)
445 X86CPU
*cpu
= x86_env_get_cpu(env
);
447 env
->cr
[3] = new_cr3
;
448 if (env
->cr
[0] & CR0_PG_MASK
) {
449 qemu_log_mask(CPU_LOG_MMU
,
450 "CR3 update: CR3=" TARGET_FMT_lx
"\n", new_cr3
);
451 tlb_flush(CPU(cpu
), 0);
455 void cpu_x86_update_cr4(CPUX86State
*env
, uint32_t new_cr4
)
457 X86CPU
*cpu
= x86_env_get_cpu(env
);
459 #if defined(DEBUG_MMU)
460 printf("CR4 update: CR4=%08x\n", (uint32_t)env
->cr
[4]);
462 if ((new_cr4
^ env
->cr
[4]) &
463 (CR4_PGE_MASK
| CR4_PAE_MASK
| CR4_PSE_MASK
|
464 CR4_SMEP_MASK
| CR4_SMAP_MASK
)) {
465 tlb_flush(CPU(cpu
), 1);
468 if (!(env
->features
[FEAT_1_EDX
] & CPUID_SSE
)) {
469 new_cr4
&= ~CR4_OSFXSR_MASK
;
471 env
->hflags
&= ~HF_OSFXSR_MASK
;
472 if (new_cr4
& CR4_OSFXSR_MASK
) {
473 env
->hflags
|= HF_OSFXSR_MASK
;
476 if (!(env
->features
[FEAT_7_0_EBX
] & CPUID_7_0_EBX_SMAP
)) {
477 new_cr4
&= ~CR4_SMAP_MASK
;
479 env
->hflags
&= ~HF_SMAP_MASK
;
480 if (new_cr4
& CR4_SMAP_MASK
) {
481 env
->hflags
|= HF_SMAP_MASK
;
484 env
->cr
[4] = new_cr4
;
487 #if defined(CONFIG_USER_ONLY)
489 int x86_cpu_handle_mmu_fault(CPUState
*cs
, vaddr addr
,
490 int is_write
, int mmu_idx
)
492 X86CPU
*cpu
= X86_CPU(cs
);
493 CPUX86State
*env
= &cpu
->env
;
495 /* user mode only emulation */
498 env
->error_code
= (is_write
<< PG_ERROR_W_BIT
);
499 env
->error_code
|= PG_ERROR_U_MASK
;
500 cs
->exception_index
= EXCP0E_PAGE
;
507 * -1 = cannot handle fault
508 * 0 = nothing more to do
509 * 1 = generate PF fault
511 int x86_cpu_handle_mmu_fault(CPUState
*cs
, vaddr addr
,
512 int is_write1
, int mmu_idx
)
514 X86CPU
*cpu
= X86_CPU(cs
);
515 CPUX86State
*env
= &cpu
->env
;
517 target_ulong pde_addr
, pte_addr
;
519 int is_dirty
, prot
, page_size
, is_write
, is_user
;
521 uint64_t rsvd_mask
= PG_HI_RSVD_MASK
;
522 uint32_t page_offset
;
525 is_user
= mmu_idx
== MMU_USER_IDX
;
526 #if defined(DEBUG_MMU)
527 printf("MMU fault: addr=%" VADDR_PRIx
" w=%d u=%d eip=" TARGET_FMT_lx
"\n",
528 addr
, is_write1
, is_user
, env
->eip
);
530 is_write
= is_write1
& 1;
532 if (!(env
->cr
[0] & CR0_PG_MASK
)) {
535 if (!(env
->hflags
& HF_LMA_MASK
)) {
536 /* Without long mode we can only address 32bits in real mode */
540 prot
= PAGE_READ
| PAGE_WRITE
| PAGE_EXEC
;
545 if (!(env
->efer
& MSR_EFER_NXE
)) {
546 rsvd_mask
|= PG_NX_MASK
;
549 if (env
->cr
[4] & CR4_PAE_MASK
) {
551 target_ulong pdpe_addr
;
554 if (env
->hflags
& HF_LMA_MASK
) {
555 uint64_t pml4e_addr
, pml4e
;
558 /* test virtual address sign extension */
559 sext
= (int64_t)addr
>> 47;
560 if (sext
!= 0 && sext
!= -1) {
562 cs
->exception_index
= EXCP0D_GPF
;
566 pml4e_addr
= ((env
->cr
[3] & ~0xfff) + (((addr
>> 39) & 0x1ff) << 3)) &
568 pml4e
= x86_ldq_phys(cs
, pml4e_addr
);
569 if (!(pml4e
& PG_PRESENT_MASK
)) {
572 if (pml4e
& (rsvd_mask
| PG_PSE_MASK
)) {
575 if (!(pml4e
& PG_ACCESSED_MASK
)) {
576 pml4e
|= PG_ACCESSED_MASK
;
577 x86_stl_phys_notdirty(cs
, pml4e_addr
, pml4e
);
579 ptep
= pml4e
^ PG_NX_MASK
;
580 pdpe_addr
= ((pml4e
& PG_ADDRESS_MASK
) + (((addr
>> 30) & 0x1ff) << 3)) &
582 pdpe
= x86_ldq_phys(cs
, pdpe_addr
);
583 if (!(pdpe
& PG_PRESENT_MASK
)) {
586 if (pdpe
& rsvd_mask
) {
589 ptep
&= pdpe
^ PG_NX_MASK
;
590 if (!(pdpe
& PG_ACCESSED_MASK
)) {
591 pdpe
|= PG_ACCESSED_MASK
;
592 x86_stl_phys_notdirty(cs
, pdpe_addr
, pdpe
);
594 if (pdpe
& PG_PSE_MASK
) {
596 page_size
= 1024 * 1024 * 1024;
597 pte_addr
= pdpe_addr
;
599 goto do_check_protect
;
604 /* XXX: load them when cr3 is loaded ? */
605 pdpe_addr
= ((env
->cr
[3] & ~0x1f) + ((addr
>> 27) & 0x18)) &
607 pdpe
= x86_ldq_phys(cs
, pdpe_addr
);
608 if (!(pdpe
& PG_PRESENT_MASK
)) {
611 rsvd_mask
|= PG_HI_USER_MASK
;
612 if (pdpe
& (rsvd_mask
| PG_NX_MASK
)) {
615 ptep
= PG_NX_MASK
| PG_USER_MASK
| PG_RW_MASK
;
618 pde_addr
= ((pdpe
& PG_ADDRESS_MASK
) + (((addr
>> 21) & 0x1ff) << 3)) &
620 pde
= x86_ldq_phys(cs
, pde_addr
);
621 if (!(pde
& PG_PRESENT_MASK
)) {
624 if (pde
& rsvd_mask
) {
627 ptep
&= pde
^ PG_NX_MASK
;
628 if (pde
& PG_PSE_MASK
) {
630 page_size
= 2048 * 1024;
633 goto do_check_protect
;
636 if (!(pde
& PG_ACCESSED_MASK
)) {
637 pde
|= PG_ACCESSED_MASK
;
638 x86_stl_phys_notdirty(cs
, pde_addr
, pde
);
640 pte_addr
= ((pde
& PG_ADDRESS_MASK
) + (((addr
>> 12) & 0x1ff) << 3)) &
642 pte
= x86_ldq_phys(cs
, pte_addr
);
643 if (!(pte
& PG_PRESENT_MASK
)) {
646 if (pte
& rsvd_mask
) {
649 /* combine pde and pte nx, user and rw protections */
650 ptep
&= pte
^ PG_NX_MASK
;
655 /* page directory entry */
656 pde_addr
= ((env
->cr
[3] & ~0xfff) + ((addr
>> 20) & 0xffc)) &
658 pde
= x86_ldl_phys(cs
, pde_addr
);
659 if (!(pde
& PG_PRESENT_MASK
)) {
662 ptep
= pde
| PG_NX_MASK
;
664 /* if PSE bit is set, then we use a 4MB page */
665 if ((pde
& PG_PSE_MASK
) && (env
->cr
[4] & CR4_PSE_MASK
)) {
666 page_size
= 4096 * 1024;
669 /* Bits 20-13 provide bits 39-32 of the address, bit 21 is reserved.
670 * Leave bits 20-13 in place for setting accessed/dirty bits below.
672 pte
= pde
| ((pde
& 0x1fe000) << (32 - 13));
673 rsvd_mask
= 0x200000;
674 goto do_check_protect_pse36
;
677 if (!(pde
& PG_ACCESSED_MASK
)) {
678 pde
|= PG_ACCESSED_MASK
;
679 x86_stl_phys_notdirty(cs
, pde_addr
, pde
);
682 /* page directory entry */
683 pte_addr
= ((pde
& ~0xfff) + ((addr
>> 10) & 0xffc)) &
685 pte
= x86_ldl_phys(cs
, pte_addr
);
686 if (!(pte
& PG_PRESENT_MASK
)) {
689 /* combine pde and pte user and rw protections */
690 ptep
&= pte
| PG_NX_MASK
;
696 rsvd_mask
|= (page_size
- 1) & PG_ADDRESS_MASK
& ~PG_PSE_PAT_MASK
;
697 do_check_protect_pse36
:
698 if (pte
& rsvd_mask
) {
702 if ((ptep
& PG_NX_MASK
) && is_write1
== 2) {
703 goto do_fault_protect
;
707 if (!(ptep
& PG_USER_MASK
)) {
708 goto do_fault_protect
;
710 if (is_write
&& !(ptep
& PG_RW_MASK
)) {
711 goto do_fault_protect
;
716 if (is_write1
!= 2 && (ptep
& PG_USER_MASK
)) {
717 goto do_fault_protect
;
720 case MMU_KNOSMAP_IDX
:
721 if (is_write1
== 2 && (env
->cr
[4] & CR4_SMEP_MASK
) &&
722 (ptep
& PG_USER_MASK
)) {
723 goto do_fault_protect
;
725 if ((env
->cr
[0] & CR0_WP_MASK
) &&
726 is_write
&& !(ptep
& PG_RW_MASK
)) {
727 goto do_fault_protect
;
731 default: /* cannot happen */
734 is_dirty
= is_write
&& !(pte
& PG_DIRTY_MASK
);
735 if (!(pte
& PG_ACCESSED_MASK
) || is_dirty
) {
736 pte
|= PG_ACCESSED_MASK
;
738 pte
|= PG_DIRTY_MASK
;
740 x86_stl_phys_notdirty(cs
, pte_addr
, pte
);
743 /* the page can be put in the TLB */
745 if (!(ptep
& PG_NX_MASK
) &&
746 (mmu_idx
== MMU_USER_IDX
||
747 !((env
->cr
[4] & CR4_SMEP_MASK
) && (ptep
& PG_USER_MASK
)))) {
750 if (pte
& PG_DIRTY_MASK
) {
751 /* only set write access if already dirty... otherwise wait
754 if (ptep
& PG_RW_MASK
)
757 if (!(env
->cr
[0] & CR0_WP_MASK
) ||
763 pte
= pte
& env
->a20_mask
;
765 /* align to page_size */
766 pte
&= PG_ADDRESS_MASK
& ~(page_size
- 1);
768 /* Even if 4MB pages, we map only one 4KB page in the cache to
769 avoid filling it too fast */
770 vaddr
= addr
& TARGET_PAGE_MASK
;
771 page_offset
= vaddr
& (page_size
- 1);
772 paddr
= pte
+ page_offset
;
774 tlb_set_page_with_attrs(cs
, vaddr
, paddr
, cpu_get_mem_attrs(env
),
775 prot
, mmu_idx
, page_size
);
778 error_code
|= PG_ERROR_RSVD_MASK
;
780 error_code
|= PG_ERROR_P_MASK
;
782 error_code
|= (is_write
<< PG_ERROR_W_BIT
);
784 error_code
|= PG_ERROR_U_MASK
;
785 if (is_write1
== 2 &&
786 (((env
->efer
& MSR_EFER_NXE
) &&
787 (env
->cr
[4] & CR4_PAE_MASK
)) ||
788 (env
->cr
[4] & CR4_SMEP_MASK
)))
789 error_code
|= PG_ERROR_I_D_MASK
;
790 if (env
->intercept_exceptions
& (1 << EXCP0E_PAGE
)) {
791 /* cr2 is not modified in case of exceptions */
793 env
->vm_vmcb
+ offsetof(struct vmcb
, control
.exit_info_2
),
798 env
->error_code
= error_code
;
799 cs
->exception_index
= EXCP0E_PAGE
;
803 hwaddr
x86_cpu_get_phys_page_debug(CPUState
*cs
, vaddr addr
)
805 X86CPU
*cpu
= X86_CPU(cs
);
806 CPUX86State
*env
= &cpu
->env
;
807 target_ulong pde_addr
, pte_addr
;
809 uint32_t page_offset
;
812 if (!(env
->cr
[0] & CR0_PG_MASK
)) {
813 pte
= addr
& env
->a20_mask
;
815 } else if (env
->cr
[4] & CR4_PAE_MASK
) {
816 target_ulong pdpe_addr
;
820 if (env
->hflags
& HF_LMA_MASK
) {
821 uint64_t pml4e_addr
, pml4e
;
824 /* test virtual address sign extension */
825 sext
= (int64_t)addr
>> 47;
826 if (sext
!= 0 && sext
!= -1) {
829 pml4e_addr
= ((env
->cr
[3] & ~0xfff) + (((addr
>> 39) & 0x1ff) << 3)) &
831 pml4e
= x86_ldq_phys(cs
, pml4e_addr
);
832 if (!(pml4e
& PG_PRESENT_MASK
)) {
835 pdpe_addr
= ((pml4e
& PG_ADDRESS_MASK
) +
836 (((addr
>> 30) & 0x1ff) << 3)) & env
->a20_mask
;
837 pdpe
= x86_ldq_phys(cs
, pdpe_addr
);
838 if (!(pdpe
& PG_PRESENT_MASK
)) {
841 if (pdpe
& PG_PSE_MASK
) {
842 page_size
= 1024 * 1024 * 1024;
850 pdpe_addr
= ((env
->cr
[3] & ~0x1f) + ((addr
>> 27) & 0x18)) &
852 pdpe
= x86_ldq_phys(cs
, pdpe_addr
);
853 if (!(pdpe
& PG_PRESENT_MASK
))
857 pde_addr
= ((pdpe
& PG_ADDRESS_MASK
) +
858 (((addr
>> 21) & 0x1ff) << 3)) & env
->a20_mask
;
859 pde
= x86_ldq_phys(cs
, pde_addr
);
860 if (!(pde
& PG_PRESENT_MASK
)) {
863 if (pde
& PG_PSE_MASK
) {
865 page_size
= 2048 * 1024;
869 pte_addr
= ((pde
& PG_ADDRESS_MASK
) +
870 (((addr
>> 12) & 0x1ff) << 3)) & env
->a20_mask
;
872 pte
= x86_ldq_phys(cs
, pte_addr
);
874 if (!(pte
& PG_PRESENT_MASK
)) {
880 /* page directory entry */
881 pde_addr
= ((env
->cr
[3] & ~0xfff) + ((addr
>> 20) & 0xffc)) & env
->a20_mask
;
882 pde
= x86_ldl_phys(cs
, pde_addr
);
883 if (!(pde
& PG_PRESENT_MASK
))
885 if ((pde
& PG_PSE_MASK
) && (env
->cr
[4] & CR4_PSE_MASK
)) {
886 pte
= pde
| ((pde
& 0x1fe000) << (32 - 13));
887 page_size
= 4096 * 1024;
889 /* page directory entry */
890 pte_addr
= ((pde
& ~0xfff) + ((addr
>> 10) & 0xffc)) & env
->a20_mask
;
891 pte
= x86_ldl_phys(cs
, pte_addr
);
892 if (!(pte
& PG_PRESENT_MASK
)) {
897 pte
= pte
& env
->a20_mask
;
903 pte
&= PG_ADDRESS_MASK
& ~(page_size
- 1);
904 page_offset
= (addr
& TARGET_PAGE_MASK
) & (page_size
- 1);
905 return pte
| page_offset
;
908 void hw_breakpoint_insert(CPUX86State
*env
, int index
)
910 CPUState
*cs
= CPU(x86_env_get_cpu(env
));
911 int type
= 0, err
= 0;
913 switch (hw_breakpoint_type(env
->dr
[7], index
)) {
914 case DR7_TYPE_BP_INST
:
915 if (hw_breakpoint_enabled(env
->dr
[7], index
)) {
916 err
= cpu_breakpoint_insert(cs
, env
->dr
[index
], BP_CPU
,
917 &env
->cpu_breakpoint
[index
]);
920 case DR7_TYPE_DATA_WR
:
921 type
= BP_CPU
| BP_MEM_WRITE
;
924 /* No support for I/O watchpoints yet */
926 case DR7_TYPE_DATA_RW
:
927 type
= BP_CPU
| BP_MEM_ACCESS
;
932 err
= cpu_watchpoint_insert(cs
, env
->dr
[index
],
933 hw_breakpoint_len(env
->dr
[7], index
),
934 type
, &env
->cpu_watchpoint
[index
]);
938 env
->cpu_breakpoint
[index
] = NULL
;
942 void hw_breakpoint_remove(CPUX86State
*env
, int index
)
946 if (!env
->cpu_breakpoint
[index
]) {
949 cs
= CPU(x86_env_get_cpu(env
));
950 switch (hw_breakpoint_type(env
->dr
[7], index
)) {
951 case DR7_TYPE_BP_INST
:
952 if (hw_breakpoint_enabled(env
->dr
[7], index
)) {
953 cpu_breakpoint_remove_by_ref(cs
, env
->cpu_breakpoint
[index
]);
956 case DR7_TYPE_DATA_WR
:
957 case DR7_TYPE_DATA_RW
:
958 cpu_watchpoint_remove_by_ref(cs
, env
->cpu_watchpoint
[index
]);
961 /* No support for I/O watchpoints yet */
966 bool check_hw_breakpoints(CPUX86State
*env
, bool force_dr6_update
)
970 bool hit_enabled
= false;
972 dr6
= env
->dr
[6] & ~0xf;
973 for (reg
= 0; reg
< DR7_MAX_BP
; reg
++) {
974 bool bp_match
= false;
975 bool wp_match
= false;
977 switch (hw_breakpoint_type(env
->dr
[7], reg
)) {
978 case DR7_TYPE_BP_INST
:
979 if (env
->dr
[reg
] == env
->eip
) {
983 case DR7_TYPE_DATA_WR
:
984 case DR7_TYPE_DATA_RW
:
985 if (env
->cpu_watchpoint
[reg
] &&
986 env
->cpu_watchpoint
[reg
]->flags
& BP_WATCHPOINT_HIT
) {
993 if (bp_match
|| wp_match
) {
995 if (hw_breakpoint_enabled(env
->dr
[7], reg
)) {
1001 if (hit_enabled
|| force_dr6_update
) {
1008 void breakpoint_handler(CPUState
*cs
)
1010 X86CPU
*cpu
= X86_CPU(cs
);
1011 CPUX86State
*env
= &cpu
->env
;
1014 if (cs
->watchpoint_hit
) {
1015 if (cs
->watchpoint_hit
->flags
& BP_CPU
) {
1016 cs
->watchpoint_hit
= NULL
;
1017 if (check_hw_breakpoints(env
, false)) {
1018 raise_exception(env
, EXCP01_DB
);
1020 cpu_resume_from_signal(cs
, NULL
);
1024 QTAILQ_FOREACH(bp
, &cs
->breakpoints
, entry
) {
1025 if (bp
->pc
== env
->eip
) {
1026 if (bp
->flags
& BP_CPU
) {
1027 check_hw_breakpoints(env
, true);
1028 raise_exception(env
, EXCP01_DB
);
1036 typedef struct MCEInjectionParams
{
1041 uint64_t mcg_status
;
1045 } MCEInjectionParams
;
1047 static void do_inject_x86_mce(void *data
)
1049 MCEInjectionParams
*params
= data
;
1050 CPUX86State
*cenv
= ¶ms
->cpu
->env
;
1051 CPUState
*cpu
= CPU(params
->cpu
);
1052 uint64_t *banks
= cenv
->mce_banks
+ 4 * params
->bank
;
1054 cpu_synchronize_state(cpu
);
1057 * If there is an MCE exception being processed, ignore this SRAO MCE
1058 * unless unconditional injection was requested.
1060 if (!(params
->flags
& MCE_INJECT_UNCOND_AO
)
1061 && !(params
->status
& MCI_STATUS_AR
)
1062 && (cenv
->mcg_status
& MCG_STATUS_MCIP
)) {
1066 if (params
->status
& MCI_STATUS_UC
) {
1068 * if MSR_MCG_CTL is not all 1s, the uncorrected error
1069 * reporting is disabled
1071 if ((cenv
->mcg_cap
& MCG_CTL_P
) && cenv
->mcg_ctl
!= ~(uint64_t)0) {
1072 monitor_printf(params
->mon
,
1073 "CPU %d: Uncorrected error reporting disabled\n",
1079 * if MSR_MCi_CTL is not all 1s, the uncorrected error
1080 * reporting is disabled for the bank
1082 if (banks
[0] != ~(uint64_t)0) {
1083 monitor_printf(params
->mon
,
1084 "CPU %d: Uncorrected error reporting disabled for"
1086 cpu
->cpu_index
, params
->bank
);
1090 if ((cenv
->mcg_status
& MCG_STATUS_MCIP
) ||
1091 !(cenv
->cr
[4] & CR4_MCE_MASK
)) {
1092 monitor_printf(params
->mon
,
1093 "CPU %d: Previous MCE still in progress, raising"
1096 qemu_log_mask(CPU_LOG_RESET
, "Triple fault\n");
1097 qemu_system_reset_request();
1100 if (banks
[1] & MCI_STATUS_VAL
) {
1101 params
->status
|= MCI_STATUS_OVER
;
1103 banks
[2] = params
->addr
;
1104 banks
[3] = params
->misc
;
1105 cenv
->mcg_status
= params
->mcg_status
;
1106 banks
[1] = params
->status
;
1107 cpu_interrupt(cpu
, CPU_INTERRUPT_MCE
);
1108 } else if (!(banks
[1] & MCI_STATUS_VAL
)
1109 || !(banks
[1] & MCI_STATUS_UC
)) {
1110 if (banks
[1] & MCI_STATUS_VAL
) {
1111 params
->status
|= MCI_STATUS_OVER
;
1113 banks
[2] = params
->addr
;
1114 banks
[3] = params
->misc
;
1115 banks
[1] = params
->status
;
1117 banks
[1] |= MCI_STATUS_OVER
;
1121 void cpu_x86_inject_mce(Monitor
*mon
, X86CPU
*cpu
, int bank
,
1122 uint64_t status
, uint64_t mcg_status
, uint64_t addr
,
1123 uint64_t misc
, int flags
)
1125 CPUState
*cs
= CPU(cpu
);
1126 CPUX86State
*cenv
= &cpu
->env
;
1127 MCEInjectionParams params
= {
1132 .mcg_status
= mcg_status
,
1137 unsigned bank_num
= cenv
->mcg_cap
& 0xff;
1139 if (!cenv
->mcg_cap
) {
1140 monitor_printf(mon
, "MCE injection not supported\n");
1143 if (bank
>= bank_num
) {
1144 monitor_printf(mon
, "Invalid MCE bank number\n");
1147 if (!(status
& MCI_STATUS_VAL
)) {
1148 monitor_printf(mon
, "Invalid MCE status code\n");
1151 if ((flags
& MCE_INJECT_BROADCAST
)
1152 && !cpu_x86_support_mca_broadcast(cenv
)) {
1153 monitor_printf(mon
, "Guest CPU does not support MCA broadcast\n");
1157 run_on_cpu(cs
, do_inject_x86_mce
, ¶ms
);
1158 if (flags
& MCE_INJECT_BROADCAST
) {
1162 params
.status
= MCI_STATUS_VAL
| MCI_STATUS_UC
;
1163 params
.mcg_status
= MCG_STATUS_MCIP
| MCG_STATUS_RIPV
;
1166 CPU_FOREACH(other_cs
) {
1167 if (other_cs
== cs
) {
1170 params
.cpu
= X86_CPU(other_cs
);
1171 run_on_cpu(other_cs
, do_inject_x86_mce
, ¶ms
);
1176 void cpu_report_tpr_access(CPUX86State
*env
, TPRAccess access
)
1178 X86CPU
*cpu
= x86_env_get_cpu(env
);
1179 CPUState
*cs
= CPU(cpu
);
1181 if (kvm_enabled()) {
1182 env
->tpr_access_type
= access
;
1184 cpu_interrupt(cs
, CPU_INTERRUPT_TPR
);
1186 cpu_restore_state(cs
, cs
->mem_io_pc
);
1188 apic_handle_tpr_access_report(cpu
->apic_state
, env
->eip
, access
);
1191 #endif /* !CONFIG_USER_ONLY */
1193 int cpu_x86_get_descr_debug(CPUX86State
*env
, unsigned int selector
,
1194 target_ulong
*base
, unsigned int *limit
,
1195 unsigned int *flags
)
1197 X86CPU
*cpu
= x86_env_get_cpu(env
);
1198 CPUState
*cs
= CPU(cpu
);
1208 index
= selector
& ~7;
1209 ptr
= dt
->base
+ index
;
1210 if ((index
+ 7) > dt
->limit
1211 || cpu_memory_rw_debug(cs
, ptr
, (uint8_t *)&e1
, sizeof(e1
), 0) != 0
1212 || cpu_memory_rw_debug(cs
, ptr
+4, (uint8_t *)&e2
, sizeof(e2
), 0) != 0)
1215 *base
= ((e1
>> 16) | ((e2
& 0xff) << 16) | (e2
& 0xff000000));
1216 *limit
= (e1
& 0xffff) | (e2
& 0x000f0000);
1217 if (e2
& DESC_G_MASK
)
1218 *limit
= (*limit
<< 12) | 0xfff;
1224 #if !defined(CONFIG_USER_ONLY)
1225 void do_cpu_init(X86CPU
*cpu
)
1227 CPUState
*cs
= CPU(cpu
);
1228 CPUX86State
*env
= &cpu
->env
;
1229 CPUX86State
*save
= g_new(CPUX86State
, 1);
1230 int sipi
= cs
->interrupt_request
& CPU_INTERRUPT_SIPI
;
1235 cs
->interrupt_request
= sipi
;
1236 memcpy(&env
->start_init_save
, &save
->start_init_save
,
1237 offsetof(CPUX86State
, end_init_save
) -
1238 offsetof(CPUX86State
, start_init_save
));
1241 if (kvm_enabled()) {
1242 kvm_arch_do_init_vcpu(cpu
);
1244 apic_init_reset(cpu
->apic_state
);
1247 void do_cpu_sipi(X86CPU
*cpu
)
1249 apic_sipi(cpu
->apic_state
);
1252 void do_cpu_init(X86CPU
*cpu
)
1255 void do_cpu_sipi(X86CPU
*cpu
)
1260 /* Frob eflags into and out of the CPU temporary format. */
1262 void x86_cpu_exec_enter(CPUState
*cs
)
1264 X86CPU
*cpu
= X86_CPU(cs
);
1265 CPUX86State
*env
= &cpu
->env
;
1267 CC_SRC
= env
->eflags
& (CC_O
| CC_S
| CC_Z
| CC_A
| CC_P
| CC_C
);
1268 env
->df
= 1 - (2 * ((env
->eflags
>> 10) & 1));
1269 CC_OP
= CC_OP_EFLAGS
;
1270 env
->eflags
&= ~(DF_MASK
| CC_O
| CC_S
| CC_Z
| CC_A
| CC_P
| CC_C
);
1273 void x86_cpu_exec_exit(CPUState
*cs
)
1275 X86CPU
*cpu
= X86_CPU(cs
);
1276 CPUX86State
*env
= &cpu
->env
;
1278 env
->eflags
= cpu_compute_eflags(env
);
1281 #ifndef CONFIG_USER_ONLY
1282 uint8_t x86_ldub_phys(CPUState
*cs
, hwaddr addr
)
1284 X86CPU
*cpu
= X86_CPU(cs
);
1285 CPUX86State
*env
= &cpu
->env
;
1287 return address_space_ldub(cs
->as
, addr
,
1288 cpu_get_mem_attrs(env
),
1292 uint32_t x86_lduw_phys(CPUState
*cs
, hwaddr addr
)
1294 X86CPU
*cpu
= X86_CPU(cs
);
1295 CPUX86State
*env
= &cpu
->env
;
1297 return address_space_lduw(cs
->as
, addr
,
1298 cpu_get_mem_attrs(env
),
1302 uint32_t x86_ldl_phys(CPUState
*cs
, hwaddr addr
)
1304 X86CPU
*cpu
= X86_CPU(cs
);
1305 CPUX86State
*env
= &cpu
->env
;
1307 return address_space_ldl(cs
->as
, addr
,
1308 cpu_get_mem_attrs(env
),
1312 uint64_t x86_ldq_phys(CPUState
*cs
, hwaddr addr
)
1314 X86CPU
*cpu
= X86_CPU(cs
);
1315 CPUX86State
*env
= &cpu
->env
;
1317 return address_space_ldq(cs
->as
, addr
,
1318 cpu_get_mem_attrs(env
),
1322 void x86_stb_phys(CPUState
*cs
, hwaddr addr
, uint8_t val
)
1324 X86CPU
*cpu
= X86_CPU(cs
);
1325 CPUX86State
*env
= &cpu
->env
;
1327 address_space_stb(cs
->as
, addr
, val
,
1328 cpu_get_mem_attrs(env
),
1332 void x86_stl_phys_notdirty(CPUState
*cs
, hwaddr addr
, uint32_t val
)
1334 X86CPU
*cpu
= X86_CPU(cs
);
1335 CPUX86State
*env
= &cpu
->env
;
1337 address_space_stl_notdirty(cs
->as
, addr
, val
,
1338 cpu_get_mem_attrs(env
),
1342 void x86_stw_phys(CPUState
*cs
, hwaddr addr
, uint32_t val
)
1344 X86CPU
*cpu
= X86_CPU(cs
);
1345 CPUX86State
*env
= &cpu
->env
;
1347 address_space_stw(cs
->as
, addr
, val
,
1348 cpu_get_mem_attrs(env
),
1352 void x86_stl_phys(CPUState
*cs
, hwaddr addr
, uint32_t val
)
1354 X86CPU
*cpu
= X86_CPU(cs
);
1355 CPUX86State
*env
= &cpu
->env
;
1357 address_space_stl(cs
->as
, addr
, val
,
1358 cpu_get_mem_attrs(env
),
1362 void x86_stq_phys(CPUState
*cs
, hwaddr addr
, uint64_t val
)
1364 X86CPU
*cpu
= X86_CPU(cs
);
1365 CPUX86State
*env
= &cpu
->env
;
1367 address_space_stq(cs
->as
, addr
, val
,
1368 cpu_get_mem_attrs(env
),