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 2.1 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"
21 #include "qapi/qapi-events-run-state.h"
23 #include "exec/exec-all.h"
24 #include "sysemu/runstate.h"
25 #include "kvm/kvm_i386.h"
26 #ifndef CONFIG_USER_ONLY
27 #include "sysemu/hw_accel.h"
28 #include "monitor/monitor.h"
31 void cpu_sync_bndcs_hflags(CPUX86State
*env
)
33 uint32_t hflags
= env
->hflags
;
34 uint32_t hflags2
= env
->hflags2
;
37 if ((hflags
& HF_CPL_MASK
) == 3) {
38 bndcsr
= env
->bndcs_regs
.cfgu
;
40 bndcsr
= env
->msr_bndcfgs
;
43 if ((env
->cr
[4] & CR4_OSXSAVE_MASK
)
44 && (env
->xcr0
& XSTATE_BNDCSR_MASK
)
45 && (bndcsr
& BNDCFG_ENABLE
)) {
46 hflags
|= HF_MPX_EN_MASK
;
48 hflags
&= ~HF_MPX_EN_MASK
;
51 if (bndcsr
& BNDCFG_BNDPRESERVE
) {
52 hflags2
|= HF2_MPX_PR_MASK
;
54 hflags2
&= ~HF2_MPX_PR_MASK
;
58 env
->hflags2
= hflags2
;
61 static void cpu_x86_version(CPUX86State
*env
, int *family
, int *model
)
63 int cpuver
= env
->cpuid_version
;
65 if (family
== NULL
|| model
== NULL
) {
69 *family
= (cpuver
>> 8) & 0x0f;
70 *model
= ((cpuver
>> 12) & 0xf0) + ((cpuver
>> 4) & 0x0f);
73 /* Broadcast MCA signal for processor version 06H_EH and above */
74 int cpu_x86_support_mca_broadcast(CPUX86State
*env
)
79 cpu_x86_version(env
, &family
, &model
);
80 if ((family
== 6 && model
>= 14) || family
> 6) {
87 /***********************************************************/
89 /* XXX: add PGE support */
91 void x86_cpu_set_a20(X86CPU
*cpu
, int a20_state
)
93 CPUX86State
*env
= &cpu
->env
;
95 a20_state
= (a20_state
!= 0);
96 if (a20_state
!= ((env
->a20_mask
>> 20) & 1)) {
97 CPUState
*cs
= CPU(cpu
);
99 qemu_log_mask(CPU_LOG_MMU
, "A20 update: a20=%d\n", a20_state
);
100 /* if the cpu is currently executing code, we must unlink it and
101 all the potentially executing TB */
102 cpu_interrupt(cs
, CPU_INTERRUPT_EXITTB
);
104 /* when a20 is changed, all the MMU mappings are invalid, so
105 we must flush everything */
107 env
->a20_mask
= ~(1 << 20) | (a20_state
<< 20);
111 void cpu_x86_update_cr0(CPUX86State
*env
, uint32_t new_cr0
)
113 X86CPU
*cpu
= env_archcpu(env
);
116 qemu_log_mask(CPU_LOG_MMU
, "CR0 update: CR0=0x%08x\n", new_cr0
);
117 if ((new_cr0
& (CR0_PG_MASK
| CR0_WP_MASK
| CR0_PE_MASK
)) !=
118 (env
->cr
[0] & (CR0_PG_MASK
| CR0_WP_MASK
| CR0_PE_MASK
))) {
123 if (!(env
->cr
[0] & CR0_PG_MASK
) && (new_cr0
& CR0_PG_MASK
) &&
124 (env
->efer
& MSR_EFER_LME
)) {
125 /* enter in long mode */
126 /* XXX: generate an exception */
127 if (!(env
->cr
[4] & CR4_PAE_MASK
))
129 env
->efer
|= MSR_EFER_LMA
;
130 env
->hflags
|= HF_LMA_MASK
;
131 } else if ((env
->cr
[0] & CR0_PG_MASK
) && !(new_cr0
& CR0_PG_MASK
) &&
132 (env
->efer
& MSR_EFER_LMA
)) {
134 env
->efer
&= ~MSR_EFER_LMA
;
135 env
->hflags
&= ~(HF_LMA_MASK
| HF_CS64_MASK
);
136 env
->eip
&= 0xffffffff;
139 env
->cr
[0] = new_cr0
| CR0_ET_MASK
;
141 /* update PE flag in hidden flags */
142 pe_state
= (env
->cr
[0] & CR0_PE_MASK
);
143 env
->hflags
= (env
->hflags
& ~HF_PE_MASK
) | (pe_state
<< HF_PE_SHIFT
);
144 /* ensure that ADDSEG is always set in real mode */
145 env
->hflags
|= ((pe_state
^ 1) << HF_ADDSEG_SHIFT
);
146 /* update FPU flags */
147 env
->hflags
= (env
->hflags
& ~(HF_MP_MASK
| HF_EM_MASK
| HF_TS_MASK
)) |
148 ((new_cr0
<< (HF_MP_SHIFT
- 1)) & (HF_MP_MASK
| HF_EM_MASK
| HF_TS_MASK
));
151 /* XXX: in legacy PAE mode, generate a GPF if reserved bits are set in
153 void cpu_x86_update_cr3(CPUX86State
*env
, target_ulong new_cr3
)
155 env
->cr
[3] = new_cr3
;
156 if (env
->cr
[0] & CR0_PG_MASK
) {
157 qemu_log_mask(CPU_LOG_MMU
,
158 "CR3 update: CR3=" TARGET_FMT_lx
"\n", new_cr3
);
159 tlb_flush(env_cpu(env
));
163 void cpu_x86_update_cr4(CPUX86State
*env
, uint32_t new_cr4
)
167 #if defined(DEBUG_MMU)
168 printf("CR4 update: %08x -> %08x\n", (uint32_t)env
->cr
[4], new_cr4
);
170 if ((new_cr4
^ env
->cr
[4]) &
171 (CR4_PGE_MASK
| CR4_PAE_MASK
| CR4_PSE_MASK
|
172 CR4_SMEP_MASK
| CR4_SMAP_MASK
| CR4_LA57_MASK
)) {
173 tlb_flush(env_cpu(env
));
176 /* Clear bits we're going to recompute. */
177 hflags
= env
->hflags
& ~(HF_OSFXSR_MASK
| HF_SMAP_MASK
);
180 if (!(env
->features
[FEAT_1_EDX
] & CPUID_SSE
)) {
181 new_cr4
&= ~CR4_OSFXSR_MASK
;
183 if (new_cr4
& CR4_OSFXSR_MASK
) {
184 hflags
|= HF_OSFXSR_MASK
;
187 if (!(env
->features
[FEAT_7_0_EBX
] & CPUID_7_0_EBX_SMAP
)) {
188 new_cr4
&= ~CR4_SMAP_MASK
;
190 if (new_cr4
& CR4_SMAP_MASK
) {
191 hflags
|= HF_SMAP_MASK
;
194 if (!(env
->features
[FEAT_7_0_ECX
] & CPUID_7_0_ECX_PKU
)) {
195 new_cr4
&= ~CR4_PKE_MASK
;
197 if (!(env
->features
[FEAT_7_0_ECX
] & CPUID_7_0_ECX_PKS
)) {
198 new_cr4
&= ~CR4_PKS_MASK
;
201 env
->cr
[4] = new_cr4
;
202 env
->hflags
= hflags
;
204 cpu_sync_bndcs_hflags(env
);
207 #if !defined(CONFIG_USER_ONLY)
208 hwaddr
x86_cpu_get_phys_page_attrs_debug(CPUState
*cs
, vaddr addr
,
211 X86CPU
*cpu
= X86_CPU(cs
);
212 CPUX86State
*env
= &cpu
->env
;
213 target_ulong pde_addr
, pte_addr
;
216 uint32_t page_offset
;
219 *attrs
= cpu_get_mem_attrs(env
);
221 a20_mask
= x86_get_a20_mask(env
);
222 if (!(env
->cr
[0] & CR0_PG_MASK
)) {
223 pte
= addr
& a20_mask
;
225 } else if (env
->cr
[4] & CR4_PAE_MASK
) {
226 target_ulong pdpe_addr
;
230 if (env
->hflags
& HF_LMA_MASK
) {
231 bool la57
= env
->cr
[4] & CR4_LA57_MASK
;
232 uint64_t pml5e_addr
, pml5e
;
233 uint64_t pml4e_addr
, pml4e
;
236 /* test virtual address sign extension */
237 sext
= la57
? (int64_t)addr
>> 56 : (int64_t)addr
>> 47;
238 if (sext
!= 0 && sext
!= -1) {
243 pml5e_addr
= ((env
->cr
[3] & ~0xfff) +
244 (((addr
>> 48) & 0x1ff) << 3)) & a20_mask
;
245 pml5e
= x86_ldq_phys(cs
, pml5e_addr
);
246 if (!(pml5e
& PG_PRESENT_MASK
)) {
253 pml4e_addr
= ((pml5e
& PG_ADDRESS_MASK
) +
254 (((addr
>> 39) & 0x1ff) << 3)) & a20_mask
;
255 pml4e
= x86_ldq_phys(cs
, pml4e_addr
);
256 if (!(pml4e
& PG_PRESENT_MASK
)) {
259 pdpe_addr
= ((pml4e
& PG_ADDRESS_MASK
) +
260 (((addr
>> 30) & 0x1ff) << 3)) & a20_mask
;
261 pdpe
= x86_ldq_phys(cs
, pdpe_addr
);
262 if (!(pdpe
& PG_PRESENT_MASK
)) {
265 if (pdpe
& PG_PSE_MASK
) {
266 page_size
= 1024 * 1024 * 1024;
274 pdpe_addr
= ((env
->cr
[3] & ~0x1f) + ((addr
>> 27) & 0x18)) &
276 pdpe
= x86_ldq_phys(cs
, pdpe_addr
);
277 if (!(pdpe
& PG_PRESENT_MASK
))
281 pde_addr
= ((pdpe
& PG_ADDRESS_MASK
) +
282 (((addr
>> 21) & 0x1ff) << 3)) & a20_mask
;
283 pde
= x86_ldq_phys(cs
, pde_addr
);
284 if (!(pde
& PG_PRESENT_MASK
)) {
287 if (pde
& PG_PSE_MASK
) {
289 page_size
= 2048 * 1024;
293 pte_addr
= ((pde
& PG_ADDRESS_MASK
) +
294 (((addr
>> 12) & 0x1ff) << 3)) & a20_mask
;
296 pte
= x86_ldq_phys(cs
, pte_addr
);
298 if (!(pte
& PG_PRESENT_MASK
)) {
304 /* page directory entry */
305 pde_addr
= ((env
->cr
[3] & ~0xfff) + ((addr
>> 20) & 0xffc)) & a20_mask
;
306 pde
= x86_ldl_phys(cs
, pde_addr
);
307 if (!(pde
& PG_PRESENT_MASK
))
309 if ((pde
& PG_PSE_MASK
) && (env
->cr
[4] & CR4_PSE_MASK
)) {
310 pte
= pde
| ((pde
& 0x1fe000LL
) << (32 - 13));
311 page_size
= 4096 * 1024;
313 /* page directory entry */
314 pte_addr
= ((pde
& ~0xfff) + ((addr
>> 10) & 0xffc)) & a20_mask
;
315 pte
= x86_ldl_phys(cs
, pte_addr
);
316 if (!(pte
& PG_PRESENT_MASK
)) {
321 pte
= pte
& a20_mask
;
327 pte
&= PG_ADDRESS_MASK
& ~(page_size
- 1);
328 page_offset
= (addr
& TARGET_PAGE_MASK
) & (page_size
- 1);
329 return pte
| page_offset
;
332 typedef struct MCEInjectionParams
{
340 } MCEInjectionParams
;
342 static void emit_guest_memory_failure(MemoryFailureAction action
, bool ar
,
345 MemoryFailureFlags mff
= {.action_required
= ar
, .recursive
= recursive
};
347 qapi_event_send_memory_failure(MEMORY_FAILURE_RECIPIENT_GUEST
, action
,
351 static void do_inject_x86_mce(CPUState
*cs
, run_on_cpu_data data
)
353 MCEInjectionParams
*params
= data
.host_ptr
;
354 X86CPU
*cpu
= X86_CPU(cs
);
355 CPUX86State
*cenv
= &cpu
->env
;
356 uint64_t *banks
= cenv
->mce_banks
+ 4 * params
->bank
;
357 g_autofree
char *msg
= NULL
;
358 bool need_reset
= false;
360 bool ar
= !!(params
->status
& MCI_STATUS_AR
);
362 cpu_synchronize_state(cs
);
363 recursive
= !!(cenv
->mcg_status
& MCG_STATUS_MCIP
);
366 * If there is an MCE exception being processed, ignore this SRAO MCE
367 * unless unconditional injection was requested.
369 if (!(params
->flags
& MCE_INJECT_UNCOND_AO
) && !ar
&& recursive
) {
370 emit_guest_memory_failure(MEMORY_FAILURE_ACTION_IGNORE
, ar
, recursive
);
374 if (params
->status
& MCI_STATUS_UC
) {
376 * if MSR_MCG_CTL is not all 1s, the uncorrected error
377 * reporting is disabled
379 if ((cenv
->mcg_cap
& MCG_CTL_P
) && cenv
->mcg_ctl
!= ~(uint64_t)0) {
380 monitor_printf(params
->mon
,
381 "CPU %d: Uncorrected error reporting disabled\n",
387 * if MSR_MCi_CTL is not all 1s, the uncorrected error
388 * reporting is disabled for the bank
390 if (banks
[0] != ~(uint64_t)0) {
391 monitor_printf(params
->mon
,
392 "CPU %d: Uncorrected error reporting disabled for"
394 cs
->cpu_index
, params
->bank
);
398 if (!(cenv
->cr
[4] & CR4_MCE_MASK
)) {
400 msg
= g_strdup_printf("CPU %d: MCE capability is not enabled, "
401 "raising triple fault", cs
->cpu_index
);
402 } else if (recursive
) {
404 msg
= g_strdup_printf("CPU %d: Previous MCE still in progress, "
405 "raising triple fault", cs
->cpu_index
);
409 emit_guest_memory_failure(MEMORY_FAILURE_ACTION_RESET
, ar
,
411 monitor_printf(params
->mon
, "%s", msg
);
412 qemu_log_mask(CPU_LOG_RESET
, "%s\n", msg
);
413 qemu_system_reset_request(SHUTDOWN_CAUSE_GUEST_RESET
);
417 if (banks
[1] & MCI_STATUS_VAL
) {
418 params
->status
|= MCI_STATUS_OVER
;
420 banks
[2] = params
->addr
;
421 banks
[3] = params
->misc
;
422 cenv
->mcg_status
= params
->mcg_status
;
423 banks
[1] = params
->status
;
424 cpu_interrupt(cs
, CPU_INTERRUPT_MCE
);
425 } else if (!(banks
[1] & MCI_STATUS_VAL
)
426 || !(banks
[1] & MCI_STATUS_UC
)) {
427 if (banks
[1] & MCI_STATUS_VAL
) {
428 params
->status
|= MCI_STATUS_OVER
;
430 banks
[2] = params
->addr
;
431 banks
[3] = params
->misc
;
432 banks
[1] = params
->status
;
434 banks
[1] |= MCI_STATUS_OVER
;
437 emit_guest_memory_failure(MEMORY_FAILURE_ACTION_INJECT
, ar
, recursive
);
440 void cpu_x86_inject_mce(Monitor
*mon
, X86CPU
*cpu
, int bank
,
441 uint64_t status
, uint64_t mcg_status
, uint64_t addr
,
442 uint64_t misc
, int flags
)
444 CPUState
*cs
= CPU(cpu
);
445 CPUX86State
*cenv
= &cpu
->env
;
446 MCEInjectionParams params
= {
450 .mcg_status
= mcg_status
,
455 unsigned bank_num
= cenv
->mcg_cap
& 0xff;
457 if (!cenv
->mcg_cap
) {
458 monitor_printf(mon
, "MCE injection not supported\n");
461 if (bank
>= bank_num
) {
462 monitor_printf(mon
, "Invalid MCE bank number\n");
465 if (!(status
& MCI_STATUS_VAL
)) {
466 monitor_printf(mon
, "Invalid MCE status code\n");
469 if ((flags
& MCE_INJECT_BROADCAST
)
470 && !cpu_x86_support_mca_broadcast(cenv
)) {
471 monitor_printf(mon
, "Guest CPU does not support MCA broadcast\n");
475 run_on_cpu(cs
, do_inject_x86_mce
, RUN_ON_CPU_HOST_PTR(¶ms
));
476 if (flags
& MCE_INJECT_BROADCAST
) {
480 params
.status
= MCI_STATUS_VAL
| MCI_STATUS_UC
;
481 params
.mcg_status
= MCG_STATUS_MCIP
| MCG_STATUS_RIPV
;
484 CPU_FOREACH(other_cs
) {
485 if (other_cs
== cs
) {
488 run_on_cpu(other_cs
, do_inject_x86_mce
, RUN_ON_CPU_HOST_PTR(¶ms
));
493 void cpu_report_tpr_access(CPUX86State
*env
, TPRAccess access
)
495 X86CPU
*cpu
= env_archcpu(env
);
496 CPUState
*cs
= env_cpu(env
);
498 if (kvm_enabled() || whpx_enabled() || nvmm_enabled()) {
499 env
->tpr_access_type
= access
;
501 cpu_interrupt(cs
, CPU_INTERRUPT_TPR
);
502 } else if (tcg_enabled()) {
503 cpu_restore_state(cs
, cs
->mem_io_pc
, false);
505 apic_handle_tpr_access_report(cpu
->apic_state
, env
->eip
, access
);
508 #endif /* !CONFIG_USER_ONLY */
510 int cpu_x86_get_descr_debug(CPUX86State
*env
, unsigned int selector
,
511 target_ulong
*base
, unsigned int *limit
,
514 CPUState
*cs
= env_cpu(env
);
524 index
= selector
& ~7;
525 ptr
= dt
->base
+ index
;
526 if ((index
+ 7) > dt
->limit
527 || cpu_memory_rw_debug(cs
, ptr
, (uint8_t *)&e1
, sizeof(e1
), 0) != 0
528 || cpu_memory_rw_debug(cs
, ptr
+4, (uint8_t *)&e2
, sizeof(e2
), 0) != 0)
531 *base
= ((e1
>> 16) | ((e2
& 0xff) << 16) | (e2
& 0xff000000));
532 *limit
= (e1
& 0xffff) | (e2
& 0x000f0000);
533 if (e2
& DESC_G_MASK
)
534 *limit
= (*limit
<< 12) | 0xfff;
540 #if !defined(CONFIG_USER_ONLY)
541 void do_cpu_init(X86CPU
*cpu
)
543 CPUState
*cs
= CPU(cpu
);
544 CPUX86State
*env
= &cpu
->env
;
545 CPUX86State
*save
= g_new(CPUX86State
, 1);
546 int sipi
= cs
->interrupt_request
& CPU_INTERRUPT_SIPI
;
551 cs
->interrupt_request
= sipi
;
552 memcpy(&env
->start_init_save
, &save
->start_init_save
,
553 offsetof(CPUX86State
, end_init_save
) -
554 offsetof(CPUX86State
, start_init_save
));
558 kvm_arch_do_init_vcpu(cpu
);
560 apic_init_reset(cpu
->apic_state
);
563 void do_cpu_sipi(X86CPU
*cpu
)
565 apic_sipi(cpu
->apic_state
);
568 void do_cpu_init(X86CPU
*cpu
)
571 void do_cpu_sipi(X86CPU
*cpu
)
576 #ifndef CONFIG_USER_ONLY
578 void cpu_load_efer(CPUX86State
*env
, uint64_t val
)
581 env
->hflags
&= ~(HF_LMA_MASK
| HF_SVME_MASK
);
582 if (env
->efer
& MSR_EFER_LMA
) {
583 env
->hflags
|= HF_LMA_MASK
;
585 if (env
->efer
& MSR_EFER_SVME
) {
586 env
->hflags
|= HF_SVME_MASK
;
590 uint8_t x86_ldub_phys(CPUState
*cs
, hwaddr addr
)
592 X86CPU
*cpu
= X86_CPU(cs
);
593 CPUX86State
*env
= &cpu
->env
;
594 MemTxAttrs attrs
= cpu_get_mem_attrs(env
);
595 AddressSpace
*as
= cpu_addressspace(cs
, attrs
);
597 return address_space_ldub(as
, addr
, attrs
, NULL
);
600 uint32_t x86_lduw_phys(CPUState
*cs
, hwaddr addr
)
602 X86CPU
*cpu
= X86_CPU(cs
);
603 CPUX86State
*env
= &cpu
->env
;
604 MemTxAttrs attrs
= cpu_get_mem_attrs(env
);
605 AddressSpace
*as
= cpu_addressspace(cs
, attrs
);
607 return address_space_lduw(as
, addr
, attrs
, NULL
);
610 uint32_t x86_ldl_phys(CPUState
*cs
, hwaddr addr
)
612 X86CPU
*cpu
= X86_CPU(cs
);
613 CPUX86State
*env
= &cpu
->env
;
614 MemTxAttrs attrs
= cpu_get_mem_attrs(env
);
615 AddressSpace
*as
= cpu_addressspace(cs
, attrs
);
617 return address_space_ldl(as
, addr
, attrs
, NULL
);
620 uint64_t x86_ldq_phys(CPUState
*cs
, hwaddr addr
)
622 X86CPU
*cpu
= X86_CPU(cs
);
623 CPUX86State
*env
= &cpu
->env
;
624 MemTxAttrs attrs
= cpu_get_mem_attrs(env
);
625 AddressSpace
*as
= cpu_addressspace(cs
, attrs
);
627 return address_space_ldq(as
, addr
, attrs
, NULL
);
630 void x86_stb_phys(CPUState
*cs
, hwaddr addr
, uint8_t val
)
632 X86CPU
*cpu
= X86_CPU(cs
);
633 CPUX86State
*env
= &cpu
->env
;
634 MemTxAttrs attrs
= cpu_get_mem_attrs(env
);
635 AddressSpace
*as
= cpu_addressspace(cs
, attrs
);
637 address_space_stb(as
, addr
, val
, attrs
, NULL
);
640 void x86_stl_phys_notdirty(CPUState
*cs
, hwaddr addr
, uint32_t val
)
642 X86CPU
*cpu
= X86_CPU(cs
);
643 CPUX86State
*env
= &cpu
->env
;
644 MemTxAttrs attrs
= cpu_get_mem_attrs(env
);
645 AddressSpace
*as
= cpu_addressspace(cs
, attrs
);
647 address_space_stl_notdirty(as
, addr
, val
, attrs
, NULL
);
650 void x86_stw_phys(CPUState
*cs
, hwaddr addr
, uint32_t val
)
652 X86CPU
*cpu
= X86_CPU(cs
);
653 CPUX86State
*env
= &cpu
->env
;
654 MemTxAttrs attrs
= cpu_get_mem_attrs(env
);
655 AddressSpace
*as
= cpu_addressspace(cs
, attrs
);
657 address_space_stw(as
, addr
, val
, attrs
, NULL
);
660 void x86_stl_phys(CPUState
*cs
, hwaddr addr
, uint32_t val
)
662 X86CPU
*cpu
= X86_CPU(cs
);
663 CPUX86State
*env
= &cpu
->env
;
664 MemTxAttrs attrs
= cpu_get_mem_attrs(env
);
665 AddressSpace
*as
= cpu_addressspace(cs
, attrs
);
667 address_space_stl(as
, addr
, val
, attrs
, NULL
);
670 void x86_stq_phys(CPUState
*cs
, hwaddr addr
, uint64_t val
)
672 X86CPU
*cpu
= X86_CPU(cs
);
673 CPUX86State
*env
= &cpu
->env
;
674 MemTxAttrs attrs
= cpu_get_mem_attrs(env
);
675 AddressSpace
*as
= cpu_addressspace(cs
, attrs
);
677 address_space_stq(as
, addr
, val
, attrs
, NULL
);