xref: /openbmc/qemu/target/i386/hvf/hvf.c (revision e615c157)
1 /* Copyright 2008 IBM Corporation
2  *           2008 Red Hat, Inc.
3  * Copyright 2011 Intel Corporation
4  * Copyright 2016 Veertu, Inc.
5  * Copyright 2017 The Android Open Source Project
6  *
7  * QEMU Hypervisor.framework support
8  *
9  * This program is free software; you can redistribute it and/or
10  * modify it under the terms of version 2 of the GNU General Public
11  * License as published by the Free Software Foundation.
12  *
13  * This program is distributed in the hope that it will be useful,
14  * but WITHOUT ANY WARRANTY; without even the implied warranty of
15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
16  * General Public License for more details.
17  *
18  * You should have received a copy of the GNU General Public License
19  * along with this program; if not, see <http://www.gnu.org/licenses/>.
20  *
21  * This file contain code under public domain from the hvdos project:
22  * https://github.com/mist64/hvdos
23  *
24  * Parts Copyright (c) 2011 NetApp, Inc.
25  * All rights reserved.
26  *
27  * Redistribution and use in source and binary forms, with or without
28  * modification, are permitted provided that the following conditions
29  * are met:
30  * 1. Redistributions of source code must retain the above copyright
31  *    notice, this list of conditions and the following disclaimer.
32  * 2. Redistributions in binary form must reproduce the above copyright
33  *    notice, this list of conditions and the following disclaimer in the
34  *    documentation and/or other materials provided with the distribution.
35  *
36  * THIS SOFTWARE IS PROVIDED BY NETAPP, INC ``AS IS'' AND
37  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
38  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
39  * ARE DISCLAIMED.  IN NO EVENT SHALL NETAPP, INC OR CONTRIBUTORS BE LIABLE
40  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
41  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
42  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
43  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
44  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
45  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
46  * SUCH DAMAGE.
47  */
48 
49 #include "qemu/osdep.h"
50 #include "qemu-common.h"
51 #include "qemu/error-report.h"
52 
53 #include "sysemu/hvf.h"
54 #include "sysemu/runstate.h"
55 #include "hvf-i386.h"
56 #include "vmcs.h"
57 #include "vmx.h"
58 #include "x86.h"
59 #include "x86_descr.h"
60 #include "x86_mmu.h"
61 #include "x86_decode.h"
62 #include "x86_emu.h"
63 #include "x86_task.h"
64 #include "x86hvf.h"
65 
66 #include <Hypervisor/hv.h>
67 #include <Hypervisor/hv_vmx.h>
68 
69 #include "exec/address-spaces.h"
70 #include "hw/i386/apic_internal.h"
71 #include "qemu/main-loop.h"
72 #include "sysemu/accel.h"
73 #include "target/i386/cpu.h"
74 
75 HVFState *hvf_state;
76 
77 static void assert_hvf_ok(hv_return_t ret)
78 {
79     if (ret == HV_SUCCESS) {
80         return;
81     }
82 
83     switch (ret) {
84     case HV_ERROR:
85         error_report("Error: HV_ERROR");
86         break;
87     case HV_BUSY:
88         error_report("Error: HV_BUSY");
89         break;
90     case HV_BAD_ARGUMENT:
91         error_report("Error: HV_BAD_ARGUMENT");
92         break;
93     case HV_NO_RESOURCES:
94         error_report("Error: HV_NO_RESOURCES");
95         break;
96     case HV_NO_DEVICE:
97         error_report("Error: HV_NO_DEVICE");
98         break;
99     case HV_UNSUPPORTED:
100         error_report("Error: HV_UNSUPPORTED");
101         break;
102     default:
103         error_report("Unknown Error");
104     }
105 
106     abort();
107 }
108 
109 /* Memory slots */
110 hvf_slot *hvf_find_overlap_slot(uint64_t start, uint64_t size)
111 {
112     hvf_slot *slot;
113     int x;
114     for (x = 0; x < hvf_state->num_slots; ++x) {
115         slot = &hvf_state->slots[x];
116         if (slot->size && start < (slot->start + slot->size) &&
117             (start + size) > slot->start) {
118             return slot;
119         }
120     }
121     return NULL;
122 }
123 
124 struct mac_slot {
125     int present;
126     uint64_t size;
127     uint64_t gpa_start;
128     uint64_t gva;
129 };
130 
131 struct mac_slot mac_slots[32];
132 
133 static int do_hvf_set_memory(hvf_slot *slot, hv_memory_flags_t flags)
134 {
135     struct mac_slot *macslot;
136     hv_return_t ret;
137 
138     macslot = &mac_slots[slot->slot_id];
139 
140     if (macslot->present) {
141         if (macslot->size != slot->size) {
142             macslot->present = 0;
143             ret = hv_vm_unmap(macslot->gpa_start, macslot->size);
144             assert_hvf_ok(ret);
145         }
146     }
147 
148     if (!slot->size) {
149         return 0;
150     }
151 
152     macslot->present = 1;
153     macslot->gpa_start = slot->start;
154     macslot->size = slot->size;
155     ret = hv_vm_map((hv_uvaddr_t)slot->mem, slot->start, slot->size, flags);
156     assert_hvf_ok(ret);
157     return 0;
158 }
159 
160 void hvf_set_phys_mem(MemoryRegionSection *section, bool add)
161 {
162     hvf_slot *mem;
163     MemoryRegion *area = section->mr;
164     bool writeable = !area->readonly && !area->rom_device;
165     hv_memory_flags_t flags;
166 
167     if (!memory_region_is_ram(area)) {
168         if (writeable) {
169             return;
170         } else if (!memory_region_is_romd(area)) {
171             /*
172              * If the memory device is not in romd_mode, then we actually want
173              * to remove the hvf memory slot so all accesses will trap.
174              */
175              add = false;
176         }
177     }
178 
179     mem = hvf_find_overlap_slot(
180             section->offset_within_address_space,
181             int128_get64(section->size));
182 
183     if (mem && add) {
184         if (mem->size == int128_get64(section->size) &&
185             mem->start == section->offset_within_address_space &&
186             mem->mem == (memory_region_get_ram_ptr(area) +
187             section->offset_within_region)) {
188             return; /* Same region was attempted to register, go away. */
189         }
190     }
191 
192     /* Region needs to be reset. set the size to 0 and remap it. */
193     if (mem) {
194         mem->size = 0;
195         if (do_hvf_set_memory(mem, 0)) {
196             error_report("Failed to reset overlapping slot");
197             abort();
198         }
199     }
200 
201     if (!add) {
202         return;
203     }
204 
205     if (area->readonly ||
206         (!memory_region_is_ram(area) && memory_region_is_romd(area))) {
207         flags = HV_MEMORY_READ | HV_MEMORY_EXEC;
208     } else {
209         flags = HV_MEMORY_READ | HV_MEMORY_WRITE | HV_MEMORY_EXEC;
210     }
211 
212     /* Now make a new slot. */
213     int x;
214 
215     for (x = 0; x < hvf_state->num_slots; ++x) {
216         mem = &hvf_state->slots[x];
217         if (!mem->size) {
218             break;
219         }
220     }
221 
222     if (x == hvf_state->num_slots) {
223         error_report("No free slots");
224         abort();
225     }
226 
227     mem->size = int128_get64(section->size);
228     mem->mem = memory_region_get_ram_ptr(area) + section->offset_within_region;
229     mem->start = section->offset_within_address_space;
230     mem->region = area;
231 
232     if (do_hvf_set_memory(mem, flags)) {
233         error_report("Error registering new memory slot");
234         abort();
235     }
236 }
237 
238 void vmx_update_tpr(CPUState *cpu)
239 {
240     /* TODO: need integrate APIC handling */
241     X86CPU *x86_cpu = X86_CPU(cpu);
242     int tpr = cpu_get_apic_tpr(x86_cpu->apic_state) << 4;
243     int irr = apic_get_highest_priority_irr(x86_cpu->apic_state);
244 
245     wreg(cpu->hvf_fd, HV_X86_TPR, tpr);
246     if (irr == -1) {
247         wvmcs(cpu->hvf_fd, VMCS_TPR_THRESHOLD, 0);
248     } else {
249         wvmcs(cpu->hvf_fd, VMCS_TPR_THRESHOLD, (irr > tpr) ? tpr >> 4 :
250               irr >> 4);
251     }
252 }
253 
254 void update_apic_tpr(CPUState *cpu)
255 {
256     X86CPU *x86_cpu = X86_CPU(cpu);
257     int tpr = rreg(cpu->hvf_fd, HV_X86_TPR) >> 4;
258     cpu_set_apic_tpr(x86_cpu->apic_state, tpr);
259 }
260 
261 #define VECTORING_INFO_VECTOR_MASK     0xff
262 
263 static void hvf_handle_interrupt(CPUState * cpu, int mask)
264 {
265     cpu->interrupt_request |= mask;
266     if (!qemu_cpu_is_self(cpu)) {
267         qemu_cpu_kick(cpu);
268     }
269 }
270 
271 void hvf_handle_io(CPUArchState *env, uint16_t port, void *buffer,
272                   int direction, int size, int count)
273 {
274     int i;
275     uint8_t *ptr = buffer;
276 
277     for (i = 0; i < count; i++) {
278         address_space_rw(&address_space_io, port, MEMTXATTRS_UNSPECIFIED,
279                          ptr, size,
280                          direction);
281         ptr += size;
282     }
283 }
284 
285 /* TODO: synchronize vcpu state */
286 static void do_hvf_cpu_synchronize_state(CPUState *cpu, run_on_cpu_data arg)
287 {
288     CPUState *cpu_state = cpu;
289     if (cpu_state->vcpu_dirty == 0) {
290         hvf_get_registers(cpu_state);
291     }
292 
293     cpu_state->vcpu_dirty = 1;
294 }
295 
296 void hvf_cpu_synchronize_state(CPUState *cpu_state)
297 {
298     if (cpu_state->vcpu_dirty == 0) {
299         run_on_cpu(cpu_state, do_hvf_cpu_synchronize_state, RUN_ON_CPU_NULL);
300     }
301 }
302 
303 static void do_hvf_cpu_synchronize_post_reset(CPUState *cpu, run_on_cpu_data arg)
304 {
305     CPUState *cpu_state = cpu;
306     hvf_put_registers(cpu_state);
307     cpu_state->vcpu_dirty = false;
308 }
309 
310 void hvf_cpu_synchronize_post_reset(CPUState *cpu_state)
311 {
312     run_on_cpu(cpu_state, do_hvf_cpu_synchronize_post_reset, RUN_ON_CPU_NULL);
313 }
314 
315 void _hvf_cpu_synchronize_post_init(CPUState *cpu, run_on_cpu_data arg)
316 {
317     CPUState *cpu_state = cpu;
318     hvf_put_registers(cpu_state);
319     cpu_state->vcpu_dirty = false;
320 }
321 
322 void hvf_cpu_synchronize_post_init(CPUState *cpu_state)
323 {
324     run_on_cpu(cpu_state, _hvf_cpu_synchronize_post_init, RUN_ON_CPU_NULL);
325 }
326 
327 static bool ept_emulation_fault(hvf_slot *slot, uint64_t gpa, uint64_t ept_qual)
328 {
329     int read, write;
330 
331     /* EPT fault on an instruction fetch doesn't make sense here */
332     if (ept_qual & EPT_VIOLATION_INST_FETCH) {
333         return false;
334     }
335 
336     /* EPT fault must be a read fault or a write fault */
337     read = ept_qual & EPT_VIOLATION_DATA_READ ? 1 : 0;
338     write = ept_qual & EPT_VIOLATION_DATA_WRITE ? 1 : 0;
339     if ((read | write) == 0) {
340         return false;
341     }
342 
343     if (write && slot) {
344         if (slot->flags & HVF_SLOT_LOG) {
345             memory_region_set_dirty(slot->region, gpa - slot->start, 1);
346             hv_vm_protect((hv_gpaddr_t)slot->start, (size_t)slot->size,
347                           HV_MEMORY_READ | HV_MEMORY_WRITE);
348         }
349     }
350 
351     /*
352      * The EPT violation must have been caused by accessing a
353      * guest-physical address that is a translation of a guest-linear
354      * address.
355      */
356     if ((ept_qual & EPT_VIOLATION_GLA_VALID) == 0 ||
357         (ept_qual & EPT_VIOLATION_XLAT_VALID) == 0) {
358         return false;
359     }
360 
361     if (!slot) {
362         return true;
363     }
364     if (!memory_region_is_ram(slot->region) &&
365         !(read && memory_region_is_romd(slot->region))) {
366         return true;
367     }
368     return false;
369 }
370 
371 static void hvf_set_dirty_tracking(MemoryRegionSection *section, bool on)
372 {
373     hvf_slot *slot;
374 
375     slot = hvf_find_overlap_slot(
376             section->offset_within_address_space,
377             int128_get64(section->size));
378 
379     /* protect region against writes; begin tracking it */
380     if (on) {
381         slot->flags |= HVF_SLOT_LOG;
382         hv_vm_protect((hv_gpaddr_t)slot->start, (size_t)slot->size,
383                       HV_MEMORY_READ);
384     /* stop tracking region*/
385     } else {
386         slot->flags &= ~HVF_SLOT_LOG;
387         hv_vm_protect((hv_gpaddr_t)slot->start, (size_t)slot->size,
388                       HV_MEMORY_READ | HV_MEMORY_WRITE);
389     }
390 }
391 
392 static void hvf_log_start(MemoryListener *listener,
393                           MemoryRegionSection *section, int old, int new)
394 {
395     if (old != 0) {
396         return;
397     }
398 
399     hvf_set_dirty_tracking(section, 1);
400 }
401 
402 static void hvf_log_stop(MemoryListener *listener,
403                          MemoryRegionSection *section, int old, int new)
404 {
405     if (new != 0) {
406         return;
407     }
408 
409     hvf_set_dirty_tracking(section, 0);
410 }
411 
412 static void hvf_log_sync(MemoryListener *listener,
413                          MemoryRegionSection *section)
414 {
415     /*
416      * sync of dirty pages is handled elsewhere; just make sure we keep
417      * tracking the region.
418      */
419     hvf_set_dirty_tracking(section, 1);
420 }
421 
422 static void hvf_region_add(MemoryListener *listener,
423                            MemoryRegionSection *section)
424 {
425     hvf_set_phys_mem(section, true);
426 }
427 
428 static void hvf_region_del(MemoryListener *listener,
429                            MemoryRegionSection *section)
430 {
431     hvf_set_phys_mem(section, false);
432 }
433 
434 static MemoryListener hvf_memory_listener = {
435     .priority = 10,
436     .region_add = hvf_region_add,
437     .region_del = hvf_region_del,
438     .log_start = hvf_log_start,
439     .log_stop = hvf_log_stop,
440     .log_sync = hvf_log_sync,
441 };
442 
443 void hvf_reset_vcpu(CPUState *cpu) {
444     uint64_t pdpte[4] = {0, 0, 0, 0};
445     int i;
446 
447     /* TODO: this shouldn't be needed; there is already a call to
448      * cpu_synchronize_all_post_reset in vl.c
449      */
450     wvmcs(cpu->hvf_fd, VMCS_ENTRY_CTLS, 0);
451     wvmcs(cpu->hvf_fd, VMCS_GUEST_IA32_EFER, 0);
452 
453     /* Initialize PDPTE */
454     for (i = 0; i < 4; i++) {
455         wvmcs(cpu->hvf_fd, VMCS_GUEST_PDPTE0 + i * 2, pdpte[i]);
456     }
457 
458     macvm_set_cr0(cpu->hvf_fd, 0x60000010);
459 
460     wvmcs(cpu->hvf_fd, VMCS_CR4_MASK, CR4_VMXE_MASK);
461     wvmcs(cpu->hvf_fd, VMCS_CR4_SHADOW, 0x0);
462     wvmcs(cpu->hvf_fd, VMCS_GUEST_CR4, CR4_VMXE_MASK);
463 
464     /* set VMCS guest state fields */
465     wvmcs(cpu->hvf_fd, VMCS_GUEST_CS_SELECTOR, 0xf000);
466     wvmcs(cpu->hvf_fd, VMCS_GUEST_CS_LIMIT, 0xffff);
467     wvmcs(cpu->hvf_fd, VMCS_GUEST_CS_ACCESS_RIGHTS, 0x9b);
468     wvmcs(cpu->hvf_fd, VMCS_GUEST_CS_BASE, 0xffff0000);
469 
470     wvmcs(cpu->hvf_fd, VMCS_GUEST_DS_SELECTOR, 0);
471     wvmcs(cpu->hvf_fd, VMCS_GUEST_DS_LIMIT, 0xffff);
472     wvmcs(cpu->hvf_fd, VMCS_GUEST_DS_ACCESS_RIGHTS, 0x93);
473     wvmcs(cpu->hvf_fd, VMCS_GUEST_DS_BASE, 0);
474 
475     wvmcs(cpu->hvf_fd, VMCS_GUEST_ES_SELECTOR, 0);
476     wvmcs(cpu->hvf_fd, VMCS_GUEST_ES_LIMIT, 0xffff);
477     wvmcs(cpu->hvf_fd, VMCS_GUEST_ES_ACCESS_RIGHTS, 0x93);
478     wvmcs(cpu->hvf_fd, VMCS_GUEST_ES_BASE, 0);
479 
480     wvmcs(cpu->hvf_fd, VMCS_GUEST_FS_SELECTOR, 0);
481     wvmcs(cpu->hvf_fd, VMCS_GUEST_FS_LIMIT, 0xffff);
482     wvmcs(cpu->hvf_fd, VMCS_GUEST_FS_ACCESS_RIGHTS, 0x93);
483     wvmcs(cpu->hvf_fd, VMCS_GUEST_FS_BASE, 0);
484 
485     wvmcs(cpu->hvf_fd, VMCS_GUEST_GS_SELECTOR, 0);
486     wvmcs(cpu->hvf_fd, VMCS_GUEST_GS_LIMIT, 0xffff);
487     wvmcs(cpu->hvf_fd, VMCS_GUEST_GS_ACCESS_RIGHTS, 0x93);
488     wvmcs(cpu->hvf_fd, VMCS_GUEST_GS_BASE, 0);
489 
490     wvmcs(cpu->hvf_fd, VMCS_GUEST_SS_SELECTOR, 0);
491     wvmcs(cpu->hvf_fd, VMCS_GUEST_SS_LIMIT, 0xffff);
492     wvmcs(cpu->hvf_fd, VMCS_GUEST_SS_ACCESS_RIGHTS, 0x93);
493     wvmcs(cpu->hvf_fd, VMCS_GUEST_SS_BASE, 0);
494 
495     wvmcs(cpu->hvf_fd, VMCS_GUEST_LDTR_SELECTOR, 0);
496     wvmcs(cpu->hvf_fd, VMCS_GUEST_LDTR_LIMIT, 0);
497     wvmcs(cpu->hvf_fd, VMCS_GUEST_LDTR_ACCESS_RIGHTS, 0x10000);
498     wvmcs(cpu->hvf_fd, VMCS_GUEST_LDTR_BASE, 0);
499 
500     wvmcs(cpu->hvf_fd, VMCS_GUEST_TR_SELECTOR, 0);
501     wvmcs(cpu->hvf_fd, VMCS_GUEST_TR_LIMIT, 0);
502     wvmcs(cpu->hvf_fd, VMCS_GUEST_TR_ACCESS_RIGHTS, 0x83);
503     wvmcs(cpu->hvf_fd, VMCS_GUEST_TR_BASE, 0);
504 
505     wvmcs(cpu->hvf_fd, VMCS_GUEST_GDTR_LIMIT, 0);
506     wvmcs(cpu->hvf_fd, VMCS_GUEST_GDTR_BASE, 0);
507 
508     wvmcs(cpu->hvf_fd, VMCS_GUEST_IDTR_LIMIT, 0);
509     wvmcs(cpu->hvf_fd, VMCS_GUEST_IDTR_BASE, 0);
510 
511     /*wvmcs(cpu->hvf_fd, VMCS_GUEST_CR2, 0x0);*/
512     wvmcs(cpu->hvf_fd, VMCS_GUEST_CR3, 0x0);
513 
514     wreg(cpu->hvf_fd, HV_X86_RIP, 0xfff0);
515     wreg(cpu->hvf_fd, HV_X86_RDX, 0x623);
516     wreg(cpu->hvf_fd, HV_X86_RFLAGS, 0x2);
517     wreg(cpu->hvf_fd, HV_X86_RSP, 0x0);
518     wreg(cpu->hvf_fd, HV_X86_RAX, 0x0);
519     wreg(cpu->hvf_fd, HV_X86_RBX, 0x0);
520     wreg(cpu->hvf_fd, HV_X86_RCX, 0x0);
521     wreg(cpu->hvf_fd, HV_X86_RSI, 0x0);
522     wreg(cpu->hvf_fd, HV_X86_RDI, 0x0);
523     wreg(cpu->hvf_fd, HV_X86_RBP, 0x0);
524 
525     for (int i = 0; i < 8; i++) {
526         wreg(cpu->hvf_fd, HV_X86_R8 + i, 0x0);
527     }
528 
529     hv_vcpu_invalidate_tlb(cpu->hvf_fd);
530     hv_vcpu_flush(cpu->hvf_fd);
531 }
532 
533 void hvf_vcpu_destroy(CPUState *cpu)
534 {
535     hv_return_t ret = hv_vcpu_destroy((hv_vcpuid_t)cpu->hvf_fd);
536     assert_hvf_ok(ret);
537 }
538 
539 static void dummy_signal(int sig)
540 {
541 }
542 
543 int hvf_init_vcpu(CPUState *cpu)
544 {
545 
546     X86CPU *x86cpu = X86_CPU(cpu);
547     CPUX86State *env = &x86cpu->env;
548     int r;
549 
550     /* init cpu signals */
551     sigset_t set;
552     struct sigaction sigact;
553 
554     memset(&sigact, 0, sizeof(sigact));
555     sigact.sa_handler = dummy_signal;
556     sigaction(SIG_IPI, &sigact, NULL);
557 
558     pthread_sigmask(SIG_BLOCK, NULL, &set);
559     sigdelset(&set, SIG_IPI);
560 
561     init_emu();
562     init_decoder();
563 
564     hvf_state->hvf_caps = g_new0(struct hvf_vcpu_caps, 1);
565     env->hvf_emul = g_new0(HVFX86EmulatorState, 1);
566 
567     r = hv_vcpu_create((hv_vcpuid_t *)&cpu->hvf_fd, HV_VCPU_DEFAULT);
568     cpu->vcpu_dirty = 1;
569     assert_hvf_ok(r);
570 
571     if (hv_vmx_read_capability(HV_VMX_CAP_PINBASED,
572         &hvf_state->hvf_caps->vmx_cap_pinbased)) {
573         abort();
574     }
575     if (hv_vmx_read_capability(HV_VMX_CAP_PROCBASED,
576         &hvf_state->hvf_caps->vmx_cap_procbased)) {
577         abort();
578     }
579     if (hv_vmx_read_capability(HV_VMX_CAP_PROCBASED2,
580         &hvf_state->hvf_caps->vmx_cap_procbased2)) {
581         abort();
582     }
583     if (hv_vmx_read_capability(HV_VMX_CAP_ENTRY,
584         &hvf_state->hvf_caps->vmx_cap_entry)) {
585         abort();
586     }
587 
588     /* set VMCS control fields */
589     wvmcs(cpu->hvf_fd, VMCS_PIN_BASED_CTLS,
590           cap2ctrl(hvf_state->hvf_caps->vmx_cap_pinbased,
591           VMCS_PIN_BASED_CTLS_EXTINT |
592           VMCS_PIN_BASED_CTLS_NMI |
593           VMCS_PIN_BASED_CTLS_VNMI));
594     wvmcs(cpu->hvf_fd, VMCS_PRI_PROC_BASED_CTLS,
595           cap2ctrl(hvf_state->hvf_caps->vmx_cap_procbased,
596           VMCS_PRI_PROC_BASED_CTLS_HLT |
597           VMCS_PRI_PROC_BASED_CTLS_MWAIT |
598           VMCS_PRI_PROC_BASED_CTLS_TSC_OFFSET |
599           VMCS_PRI_PROC_BASED_CTLS_TPR_SHADOW) |
600           VMCS_PRI_PROC_BASED_CTLS_SEC_CONTROL);
601     wvmcs(cpu->hvf_fd, VMCS_SEC_PROC_BASED_CTLS,
602           cap2ctrl(hvf_state->hvf_caps->vmx_cap_procbased2,
603                    VMCS_PRI_PROC_BASED2_CTLS_APIC_ACCESSES));
604 
605     wvmcs(cpu->hvf_fd, VMCS_ENTRY_CTLS, cap2ctrl(hvf_state->hvf_caps->vmx_cap_entry,
606           0));
607     wvmcs(cpu->hvf_fd, VMCS_EXCEPTION_BITMAP, 0); /* Double fault */
608 
609     wvmcs(cpu->hvf_fd, VMCS_TPR_THRESHOLD, 0);
610 
611     x86cpu = X86_CPU(cpu);
612     x86cpu->env.xsave_buf = qemu_memalign(4096, 4096);
613 
614     hv_vcpu_enable_native_msr(cpu->hvf_fd, MSR_STAR, 1);
615     hv_vcpu_enable_native_msr(cpu->hvf_fd, MSR_LSTAR, 1);
616     hv_vcpu_enable_native_msr(cpu->hvf_fd, MSR_CSTAR, 1);
617     hv_vcpu_enable_native_msr(cpu->hvf_fd, MSR_FMASK, 1);
618     hv_vcpu_enable_native_msr(cpu->hvf_fd, MSR_FSBASE, 1);
619     hv_vcpu_enable_native_msr(cpu->hvf_fd, MSR_GSBASE, 1);
620     hv_vcpu_enable_native_msr(cpu->hvf_fd, MSR_KERNELGSBASE, 1);
621     hv_vcpu_enable_native_msr(cpu->hvf_fd, MSR_TSC_AUX, 1);
622     hv_vcpu_enable_native_msr(cpu->hvf_fd, MSR_IA32_TSC, 1);
623     hv_vcpu_enable_native_msr(cpu->hvf_fd, MSR_IA32_SYSENTER_CS, 1);
624     hv_vcpu_enable_native_msr(cpu->hvf_fd, MSR_IA32_SYSENTER_EIP, 1);
625     hv_vcpu_enable_native_msr(cpu->hvf_fd, MSR_IA32_SYSENTER_ESP, 1);
626 
627     return 0;
628 }
629 
630 static void hvf_store_events(CPUState *cpu, uint32_t ins_len, uint64_t idtvec_info)
631 {
632     X86CPU *x86_cpu = X86_CPU(cpu);
633     CPUX86State *env = &x86_cpu->env;
634 
635     env->exception_nr = -1;
636     env->exception_pending = 0;
637     env->exception_injected = 0;
638     env->interrupt_injected = -1;
639     env->nmi_injected = false;
640     env->ins_len = 0;
641     env->has_error_code = false;
642     if (idtvec_info & VMCS_IDT_VEC_VALID) {
643         switch (idtvec_info & VMCS_IDT_VEC_TYPE) {
644         case VMCS_IDT_VEC_HWINTR:
645         case VMCS_IDT_VEC_SWINTR:
646             env->interrupt_injected = idtvec_info & VMCS_IDT_VEC_VECNUM;
647             break;
648         case VMCS_IDT_VEC_NMI:
649             env->nmi_injected = true;
650             break;
651         case VMCS_IDT_VEC_HWEXCEPTION:
652         case VMCS_IDT_VEC_SWEXCEPTION:
653             env->exception_nr = idtvec_info & VMCS_IDT_VEC_VECNUM;
654             env->exception_injected = 1;
655             break;
656         case VMCS_IDT_VEC_PRIV_SWEXCEPTION:
657         default:
658             abort();
659         }
660         if ((idtvec_info & VMCS_IDT_VEC_TYPE) == VMCS_IDT_VEC_SWEXCEPTION ||
661             (idtvec_info & VMCS_IDT_VEC_TYPE) == VMCS_IDT_VEC_SWINTR) {
662             env->ins_len = ins_len;
663         }
664         if (idtvec_info & VMCS_IDT_VEC_ERRCODE_VALID) {
665             env->has_error_code = true;
666             env->error_code = rvmcs(cpu->hvf_fd, VMCS_IDT_VECTORING_ERROR);
667         }
668     }
669     if ((rvmcs(cpu->hvf_fd, VMCS_GUEST_INTERRUPTIBILITY) &
670         VMCS_INTERRUPTIBILITY_NMI_BLOCKING)) {
671         env->hflags2 |= HF2_NMI_MASK;
672     } else {
673         env->hflags2 &= ~HF2_NMI_MASK;
674     }
675     if (rvmcs(cpu->hvf_fd, VMCS_GUEST_INTERRUPTIBILITY) &
676          (VMCS_INTERRUPTIBILITY_STI_BLOCKING |
677          VMCS_INTERRUPTIBILITY_MOVSS_BLOCKING)) {
678         env->hflags |= HF_INHIBIT_IRQ_MASK;
679     } else {
680         env->hflags &= ~HF_INHIBIT_IRQ_MASK;
681     }
682 }
683 
684 int hvf_vcpu_exec(CPUState *cpu)
685 {
686     X86CPU *x86_cpu = X86_CPU(cpu);
687     CPUX86State *env = &x86_cpu->env;
688     int ret = 0;
689     uint64_t rip = 0;
690 
691     if (hvf_process_events(cpu)) {
692         return EXCP_HLT;
693     }
694 
695     do {
696         if (cpu->vcpu_dirty) {
697             hvf_put_registers(cpu);
698             cpu->vcpu_dirty = false;
699         }
700 
701         if (hvf_inject_interrupts(cpu)) {
702             return EXCP_INTERRUPT;
703         }
704         vmx_update_tpr(cpu);
705 
706         qemu_mutex_unlock_iothread();
707         if (!cpu_is_bsp(X86_CPU(cpu)) && cpu->halted) {
708             qemu_mutex_lock_iothread();
709             return EXCP_HLT;
710         }
711 
712         hv_return_t r  = hv_vcpu_run(cpu->hvf_fd);
713         assert_hvf_ok(r);
714 
715         /* handle VMEXIT */
716         uint64_t exit_reason = rvmcs(cpu->hvf_fd, VMCS_EXIT_REASON);
717         uint64_t exit_qual = rvmcs(cpu->hvf_fd, VMCS_EXIT_QUALIFICATION);
718         uint32_t ins_len = (uint32_t)rvmcs(cpu->hvf_fd,
719                                            VMCS_EXIT_INSTRUCTION_LENGTH);
720 
721         uint64_t idtvec_info = rvmcs(cpu->hvf_fd, VMCS_IDT_VECTORING_INFO);
722 
723         hvf_store_events(cpu, ins_len, idtvec_info);
724         rip = rreg(cpu->hvf_fd, HV_X86_RIP);
725         RFLAGS(env) = rreg(cpu->hvf_fd, HV_X86_RFLAGS);
726         env->eflags = RFLAGS(env);
727 
728         qemu_mutex_lock_iothread();
729 
730         update_apic_tpr(cpu);
731         current_cpu = cpu;
732 
733         ret = 0;
734         switch (exit_reason) {
735         case EXIT_REASON_HLT: {
736             macvm_set_rip(cpu, rip + ins_len);
737             if (!((cpu->interrupt_request & CPU_INTERRUPT_HARD) &&
738                 (EFLAGS(env) & IF_MASK))
739                 && !(cpu->interrupt_request & CPU_INTERRUPT_NMI) &&
740                 !(idtvec_info & VMCS_IDT_VEC_VALID)) {
741                 cpu->halted = 1;
742                 ret = EXCP_HLT;
743                 break;
744             }
745             ret = EXCP_INTERRUPT;
746             break;
747         }
748         case EXIT_REASON_MWAIT: {
749             ret = EXCP_INTERRUPT;
750             break;
751         }
752         /* Need to check if MMIO or unmapped fault */
753         case EXIT_REASON_EPT_FAULT:
754         {
755             hvf_slot *slot;
756             uint64_t gpa = rvmcs(cpu->hvf_fd, VMCS_GUEST_PHYSICAL_ADDRESS);
757 
758             if (((idtvec_info & VMCS_IDT_VEC_VALID) == 0) &&
759                 ((exit_qual & EXIT_QUAL_NMIUDTI) != 0)) {
760                 vmx_set_nmi_blocking(cpu);
761             }
762 
763             slot = hvf_find_overlap_slot(gpa, 1);
764             /* mmio */
765             if (ept_emulation_fault(slot, gpa, exit_qual)) {
766                 struct x86_decode decode;
767 
768                 load_regs(cpu);
769                 env->hvf_emul->fetch_rip = rip;
770 
771                 decode_instruction(env, &decode);
772                 exec_instruction(env, &decode);
773                 store_regs(cpu);
774                 break;
775             }
776             break;
777         }
778         case EXIT_REASON_INOUT:
779         {
780             uint32_t in = (exit_qual & 8) != 0;
781             uint32_t size =  (exit_qual & 7) + 1;
782             uint32_t string =  (exit_qual & 16) != 0;
783             uint32_t port =  exit_qual >> 16;
784             /*uint32_t rep = (exit_qual & 0x20) != 0;*/
785 
786             if (!string && in) {
787                 uint64_t val = 0;
788                 load_regs(cpu);
789                 hvf_handle_io(env, port, &val, 0, size, 1);
790                 if (size == 1) {
791                     AL(env) = val;
792                 } else if (size == 2) {
793                     AX(env) = val;
794                 } else if (size == 4) {
795                     RAX(env) = (uint32_t)val;
796                 } else {
797                     RAX(env) = (uint64_t)val;
798                 }
799                 RIP(env) += ins_len;
800                 store_regs(cpu);
801                 break;
802             } else if (!string && !in) {
803                 RAX(env) = rreg(cpu->hvf_fd, HV_X86_RAX);
804                 hvf_handle_io(env, port, &RAX(env), 1, size, 1);
805                 macvm_set_rip(cpu, rip + ins_len);
806                 break;
807             }
808             struct x86_decode decode;
809 
810             load_regs(cpu);
811             env->hvf_emul->fetch_rip = rip;
812 
813             decode_instruction(env, &decode);
814             assert(ins_len == decode.len);
815             exec_instruction(env, &decode);
816             store_regs(cpu);
817 
818             break;
819         }
820         case EXIT_REASON_CPUID: {
821             uint32_t rax = (uint32_t)rreg(cpu->hvf_fd, HV_X86_RAX);
822             uint32_t rbx = (uint32_t)rreg(cpu->hvf_fd, HV_X86_RBX);
823             uint32_t rcx = (uint32_t)rreg(cpu->hvf_fd, HV_X86_RCX);
824             uint32_t rdx = (uint32_t)rreg(cpu->hvf_fd, HV_X86_RDX);
825 
826             cpu_x86_cpuid(env, rax, rcx, &rax, &rbx, &rcx, &rdx);
827 
828             wreg(cpu->hvf_fd, HV_X86_RAX, rax);
829             wreg(cpu->hvf_fd, HV_X86_RBX, rbx);
830             wreg(cpu->hvf_fd, HV_X86_RCX, rcx);
831             wreg(cpu->hvf_fd, HV_X86_RDX, rdx);
832 
833             macvm_set_rip(cpu, rip + ins_len);
834             break;
835         }
836         case EXIT_REASON_XSETBV: {
837             X86CPU *x86_cpu = X86_CPU(cpu);
838             CPUX86State *env = &x86_cpu->env;
839             uint32_t eax = (uint32_t)rreg(cpu->hvf_fd, HV_X86_RAX);
840             uint32_t ecx = (uint32_t)rreg(cpu->hvf_fd, HV_X86_RCX);
841             uint32_t edx = (uint32_t)rreg(cpu->hvf_fd, HV_X86_RDX);
842 
843             if (ecx) {
844                 macvm_set_rip(cpu, rip + ins_len);
845                 break;
846             }
847             env->xcr0 = ((uint64_t)edx << 32) | eax;
848             wreg(cpu->hvf_fd, HV_X86_XCR0, env->xcr0 | 1);
849             macvm_set_rip(cpu, rip + ins_len);
850             break;
851         }
852         case EXIT_REASON_INTR_WINDOW:
853             vmx_clear_int_window_exiting(cpu);
854             ret = EXCP_INTERRUPT;
855             break;
856         case EXIT_REASON_NMI_WINDOW:
857             vmx_clear_nmi_window_exiting(cpu);
858             ret = EXCP_INTERRUPT;
859             break;
860         case EXIT_REASON_EXT_INTR:
861             /* force exit and allow io handling */
862             ret = EXCP_INTERRUPT;
863             break;
864         case EXIT_REASON_RDMSR:
865         case EXIT_REASON_WRMSR:
866         {
867             load_regs(cpu);
868             if (exit_reason == EXIT_REASON_RDMSR) {
869                 simulate_rdmsr(cpu);
870             } else {
871                 simulate_wrmsr(cpu);
872             }
873             RIP(env) += rvmcs(cpu->hvf_fd, VMCS_EXIT_INSTRUCTION_LENGTH);
874             store_regs(cpu);
875             break;
876         }
877         case EXIT_REASON_CR_ACCESS: {
878             int cr;
879             int reg;
880 
881             load_regs(cpu);
882             cr = exit_qual & 15;
883             reg = (exit_qual >> 8) & 15;
884 
885             switch (cr) {
886             case 0x0: {
887                 macvm_set_cr0(cpu->hvf_fd, RRX(env, reg));
888                 break;
889             }
890             case 4: {
891                 macvm_set_cr4(cpu->hvf_fd, RRX(env, reg));
892                 break;
893             }
894             case 8: {
895                 X86CPU *x86_cpu = X86_CPU(cpu);
896                 if (exit_qual & 0x10) {
897                     RRX(env, reg) = cpu_get_apic_tpr(x86_cpu->apic_state);
898                 } else {
899                     int tpr = RRX(env, reg);
900                     cpu_set_apic_tpr(x86_cpu->apic_state, tpr);
901                     ret = EXCP_INTERRUPT;
902                 }
903                 break;
904             }
905             default:
906                 error_report("Unrecognized CR %d", cr);
907                 abort();
908             }
909             RIP(env) += ins_len;
910             store_regs(cpu);
911             break;
912         }
913         case EXIT_REASON_APIC_ACCESS: { /* TODO */
914             struct x86_decode decode;
915 
916             load_regs(cpu);
917             env->hvf_emul->fetch_rip = rip;
918 
919             decode_instruction(env, &decode);
920             exec_instruction(env, &decode);
921             store_regs(cpu);
922             break;
923         }
924         case EXIT_REASON_TPR: {
925             ret = 1;
926             break;
927         }
928         case EXIT_REASON_TASK_SWITCH: {
929             uint64_t vinfo = rvmcs(cpu->hvf_fd, VMCS_IDT_VECTORING_INFO);
930             x68_segment_selector sel = {.sel = exit_qual & 0xffff};
931             vmx_handle_task_switch(cpu, sel, (exit_qual >> 30) & 0x3,
932              vinfo & VMCS_INTR_VALID, vinfo & VECTORING_INFO_VECTOR_MASK, vinfo
933              & VMCS_INTR_T_MASK);
934             break;
935         }
936         case EXIT_REASON_TRIPLE_FAULT: {
937             qemu_system_reset_request(SHUTDOWN_CAUSE_GUEST_RESET);
938             ret = EXCP_INTERRUPT;
939             break;
940         }
941         case EXIT_REASON_RDPMC:
942             wreg(cpu->hvf_fd, HV_X86_RAX, 0);
943             wreg(cpu->hvf_fd, HV_X86_RDX, 0);
944             macvm_set_rip(cpu, rip + ins_len);
945             break;
946         case VMX_REASON_VMCALL:
947             env->exception_nr = EXCP0D_GPF;
948             env->exception_injected = 1;
949             env->has_error_code = true;
950             env->error_code = 0;
951             break;
952         default:
953             error_report("%llx: unhandled exit %llx", rip, exit_reason);
954         }
955     } while (ret == 0);
956 
957     return ret;
958 }
959 
960 bool hvf_allowed;
961 
962 static int hvf_accel_init(MachineState *ms)
963 {
964     int x;
965     hv_return_t ret;
966     HVFState *s;
967 
968     ret = hv_vm_create(HV_VM_DEFAULT);
969     assert_hvf_ok(ret);
970 
971     s = g_new0(HVFState, 1);
972 
973     s->num_slots = 32;
974     for (x = 0; x < s->num_slots; ++x) {
975         s->slots[x].size = 0;
976         s->slots[x].slot_id = x;
977     }
978 
979     hvf_state = s;
980     cpu_interrupt_handler = hvf_handle_interrupt;
981     memory_listener_register(&hvf_memory_listener, &address_space_memory);
982     return 0;
983 }
984 
985 static void hvf_accel_class_init(ObjectClass *oc, void *data)
986 {
987     AccelClass *ac = ACCEL_CLASS(oc);
988     ac->name = "HVF";
989     ac->init_machine = hvf_accel_init;
990     ac->allowed = &hvf_allowed;
991 }
992 
993 static const TypeInfo hvf_accel_type = {
994     .name = TYPE_HVF_ACCEL,
995     .parent = TYPE_ACCEL,
996     .class_init = hvf_accel_class_init,
997 };
998 
999 static void hvf_type_init(void)
1000 {
1001     type_register_static(&hvf_accel_type);
1002 }
1003 
1004 type_init(hvf_type_init);
1005