1 /*
2 * tools/testing/selftests/kvm/lib/x86.c
3 *
4 * Copyright (C) 2018, Google LLC.
5 *
6 * This work is licensed under the terms of the GNU GPL, version 2.
7 */
8
9 #define _GNU_SOURCE /* for program_invocation_name */
10
11 #include "test_util.h"
12 #include "kvm_util.h"
13 #include "kvm_util_internal.h"
14 #include "x86.h"
15
16 /* Minimum physical address used for virtual translation tables. */
17 #define KVM_GUEST_PAGE_TABLE_MIN_PADDR 0x180000
18
19 /* Virtual translation table structure declarations */
20 struct pageMapL4Entry {
21 uint64_t present:1;
22 uint64_t writable:1;
23 uint64_t user:1;
24 uint64_t write_through:1;
25 uint64_t cache_disable:1;
26 uint64_t accessed:1;
27 uint64_t ignored_06:1;
28 uint64_t page_size:1;
29 uint64_t ignored_11_08:4;
30 uint64_t address:40;
31 uint64_t ignored_62_52:11;
32 uint64_t execute_disable:1;
33 };
34
35 struct pageDirectoryPointerEntry {
36 uint64_t present:1;
37 uint64_t writable:1;
38 uint64_t user:1;
39 uint64_t write_through:1;
40 uint64_t cache_disable:1;
41 uint64_t accessed:1;
42 uint64_t ignored_06:1;
43 uint64_t page_size:1;
44 uint64_t ignored_11_08:4;
45 uint64_t address:40;
46 uint64_t ignored_62_52:11;
47 uint64_t execute_disable:1;
48 };
49
50 struct pageDirectoryEntry {
51 uint64_t present:1;
52 uint64_t writable:1;
53 uint64_t user:1;
54 uint64_t write_through:1;
55 uint64_t cache_disable:1;
56 uint64_t accessed:1;
57 uint64_t ignored_06:1;
58 uint64_t page_size:1;
59 uint64_t ignored_11_08:4;
60 uint64_t address:40;
61 uint64_t ignored_62_52:11;
62 uint64_t execute_disable:1;
63 };
64
65 struct pageTableEntry {
66 uint64_t present:1;
67 uint64_t writable:1;
68 uint64_t user:1;
69 uint64_t write_through:1;
70 uint64_t cache_disable:1;
71 uint64_t accessed:1;
72 uint64_t dirty:1;
73 uint64_t reserved_07:1;
74 uint64_t global:1;
75 uint64_t ignored_11_09:3;
76 uint64_t address:40;
77 uint64_t ignored_62_52:11;
78 uint64_t execute_disable:1;
79 };
80
81 /* Register Dump
82 *
83 * Input Args:
84 * indent - Left margin indent amount
85 * regs - register
86 *
87 * Output Args:
88 * stream - Output FILE stream
89 *
90 * Return: None
91 *
92 * Dumps the state of the registers given by regs, to the FILE stream
93 * given by steam.
94 */
regs_dump(FILE * stream,struct kvm_regs * regs,uint8_t indent)95 void regs_dump(FILE *stream, struct kvm_regs *regs,
96 uint8_t indent)
97 {
98 fprintf(stream, "%*srax: 0x%.16llx rbx: 0x%.16llx "
99 "rcx: 0x%.16llx rdx: 0x%.16llx\n",
100 indent, "",
101 regs->rax, regs->rbx, regs->rcx, regs->rdx);
102 fprintf(stream, "%*srsi: 0x%.16llx rdi: 0x%.16llx "
103 "rsp: 0x%.16llx rbp: 0x%.16llx\n",
104 indent, "",
105 regs->rsi, regs->rdi, regs->rsp, regs->rbp);
106 fprintf(stream, "%*sr8: 0x%.16llx r9: 0x%.16llx "
107 "r10: 0x%.16llx r11: 0x%.16llx\n",
108 indent, "",
109 regs->r8, regs->r9, regs->r10, regs->r11);
110 fprintf(stream, "%*sr12: 0x%.16llx r13: 0x%.16llx "
111 "r14: 0x%.16llx r15: 0x%.16llx\n",
112 indent, "",
113 regs->r12, regs->r13, regs->r14, regs->r15);
114 fprintf(stream, "%*srip: 0x%.16llx rfl: 0x%.16llx\n",
115 indent, "",
116 regs->rip, regs->rflags);
117 }
118
119 /* Segment Dump
120 *
121 * Input Args:
122 * indent - Left margin indent amount
123 * segment - KVM segment
124 *
125 * Output Args:
126 * stream - Output FILE stream
127 *
128 * Return: None
129 *
130 * Dumps the state of the KVM segment given by segment, to the FILE stream
131 * given by steam.
132 */
segment_dump(FILE * stream,struct kvm_segment * segment,uint8_t indent)133 static void segment_dump(FILE *stream, struct kvm_segment *segment,
134 uint8_t indent)
135 {
136 fprintf(stream, "%*sbase: 0x%.16llx limit: 0x%.8x "
137 "selector: 0x%.4x type: 0x%.2x\n",
138 indent, "", segment->base, segment->limit,
139 segment->selector, segment->type);
140 fprintf(stream, "%*spresent: 0x%.2x dpl: 0x%.2x "
141 "db: 0x%.2x s: 0x%.2x l: 0x%.2x\n",
142 indent, "", segment->present, segment->dpl,
143 segment->db, segment->s, segment->l);
144 fprintf(stream, "%*sg: 0x%.2x avl: 0x%.2x "
145 "unusable: 0x%.2x padding: 0x%.2x\n",
146 indent, "", segment->g, segment->avl,
147 segment->unusable, segment->padding);
148 }
149
150 /* dtable Dump
151 *
152 * Input Args:
153 * indent - Left margin indent amount
154 * dtable - KVM dtable
155 *
156 * Output Args:
157 * stream - Output FILE stream
158 *
159 * Return: None
160 *
161 * Dumps the state of the KVM dtable given by dtable, to the FILE stream
162 * given by steam.
163 */
dtable_dump(FILE * stream,struct kvm_dtable * dtable,uint8_t indent)164 static void dtable_dump(FILE *stream, struct kvm_dtable *dtable,
165 uint8_t indent)
166 {
167 fprintf(stream, "%*sbase: 0x%.16llx limit: 0x%.4x "
168 "padding: 0x%.4x 0x%.4x 0x%.4x\n",
169 indent, "", dtable->base, dtable->limit,
170 dtable->padding[0], dtable->padding[1], dtable->padding[2]);
171 }
172
173 /* System Register Dump
174 *
175 * Input Args:
176 * indent - Left margin indent amount
177 * sregs - System registers
178 *
179 * Output Args:
180 * stream - Output FILE stream
181 *
182 * Return: None
183 *
184 * Dumps the state of the system registers given by sregs, to the FILE stream
185 * given by steam.
186 */
sregs_dump(FILE * stream,struct kvm_sregs * sregs,uint8_t indent)187 void sregs_dump(FILE *stream, struct kvm_sregs *sregs,
188 uint8_t indent)
189 {
190 unsigned int i;
191
192 fprintf(stream, "%*scs:\n", indent, "");
193 segment_dump(stream, &sregs->cs, indent + 2);
194 fprintf(stream, "%*sds:\n", indent, "");
195 segment_dump(stream, &sregs->ds, indent + 2);
196 fprintf(stream, "%*ses:\n", indent, "");
197 segment_dump(stream, &sregs->es, indent + 2);
198 fprintf(stream, "%*sfs:\n", indent, "");
199 segment_dump(stream, &sregs->fs, indent + 2);
200 fprintf(stream, "%*sgs:\n", indent, "");
201 segment_dump(stream, &sregs->gs, indent + 2);
202 fprintf(stream, "%*sss:\n", indent, "");
203 segment_dump(stream, &sregs->ss, indent + 2);
204 fprintf(stream, "%*str:\n", indent, "");
205 segment_dump(stream, &sregs->tr, indent + 2);
206 fprintf(stream, "%*sldt:\n", indent, "");
207 segment_dump(stream, &sregs->ldt, indent + 2);
208
209 fprintf(stream, "%*sgdt:\n", indent, "");
210 dtable_dump(stream, &sregs->gdt, indent + 2);
211 fprintf(stream, "%*sidt:\n", indent, "");
212 dtable_dump(stream, &sregs->idt, indent + 2);
213
214 fprintf(stream, "%*scr0: 0x%.16llx cr2: 0x%.16llx "
215 "cr3: 0x%.16llx cr4: 0x%.16llx\n",
216 indent, "",
217 sregs->cr0, sregs->cr2, sregs->cr3, sregs->cr4);
218 fprintf(stream, "%*scr8: 0x%.16llx efer: 0x%.16llx "
219 "apic_base: 0x%.16llx\n",
220 indent, "",
221 sregs->cr8, sregs->efer, sregs->apic_base);
222
223 fprintf(stream, "%*sinterrupt_bitmap:\n", indent, "");
224 for (i = 0; i < (KVM_NR_INTERRUPTS + 63) / 64; i++) {
225 fprintf(stream, "%*s%.16llx\n", indent + 2, "",
226 sregs->interrupt_bitmap[i]);
227 }
228 }
229
virt_pgd_alloc(struct kvm_vm * vm,uint32_t pgd_memslot)230 void virt_pgd_alloc(struct kvm_vm *vm, uint32_t pgd_memslot)
231 {
232 int rc;
233
234 TEST_ASSERT(vm->mode == VM_MODE_FLAT48PG, "Attempt to use "
235 "unknown or unsupported guest mode, mode: 0x%x", vm->mode);
236
237 /* If needed, create page map l4 table. */
238 if (!vm->pgd_created) {
239 vm_paddr_t paddr = vm_phy_page_alloc(vm,
240 KVM_GUEST_PAGE_TABLE_MIN_PADDR, pgd_memslot);
241 vm->pgd = paddr;
242 vm->pgd_created = true;
243 }
244 }
245
246 /* VM Virtual Page Map
247 *
248 * Input Args:
249 * vm - Virtual Machine
250 * vaddr - VM Virtual Address
251 * paddr - VM Physical Address
252 * pgd_memslot - Memory region slot for new virtual translation tables
253 *
254 * Output Args: None
255 *
256 * Return: None
257 *
258 * Within the VM given by vm, creates a virtual translation for the page
259 * starting at vaddr to the page starting at paddr.
260 */
virt_pg_map(struct kvm_vm * vm,uint64_t vaddr,uint64_t paddr,uint32_t pgd_memslot)261 void virt_pg_map(struct kvm_vm *vm, uint64_t vaddr, uint64_t paddr,
262 uint32_t pgd_memslot)
263 {
264 uint16_t index[4];
265 struct pageMapL4Entry *pml4e;
266
267 TEST_ASSERT(vm->mode == VM_MODE_FLAT48PG, "Attempt to use "
268 "unknown or unsupported guest mode, mode: 0x%x", vm->mode);
269
270 TEST_ASSERT((vaddr % vm->page_size) == 0,
271 "Virtual address not on page boundary,\n"
272 " vaddr: 0x%lx vm->page_size: 0x%x",
273 vaddr, vm->page_size);
274 TEST_ASSERT(sparsebit_is_set(vm->vpages_valid,
275 (vaddr >> vm->page_shift)),
276 "Invalid virtual address, vaddr: 0x%lx",
277 vaddr);
278 TEST_ASSERT((paddr % vm->page_size) == 0,
279 "Physical address not on page boundary,\n"
280 " paddr: 0x%lx vm->page_size: 0x%x",
281 paddr, vm->page_size);
282 TEST_ASSERT((paddr >> vm->page_shift) <= vm->max_gfn,
283 "Physical address beyond beyond maximum supported,\n"
284 " paddr: 0x%lx vm->max_gfn: 0x%lx vm->page_size: 0x%x",
285 paddr, vm->max_gfn, vm->page_size);
286
287 index[0] = (vaddr >> 12) & 0x1ffu;
288 index[1] = (vaddr >> 21) & 0x1ffu;
289 index[2] = (vaddr >> 30) & 0x1ffu;
290 index[3] = (vaddr >> 39) & 0x1ffu;
291
292 /* Allocate page directory pointer table if not present. */
293 pml4e = addr_gpa2hva(vm, vm->pgd);
294 if (!pml4e[index[3]].present) {
295 pml4e[index[3]].address = vm_phy_page_alloc(vm,
296 KVM_GUEST_PAGE_TABLE_MIN_PADDR, pgd_memslot)
297 >> vm->page_shift;
298 pml4e[index[3]].writable = true;
299 pml4e[index[3]].present = true;
300 }
301
302 /* Allocate page directory table if not present. */
303 struct pageDirectoryPointerEntry *pdpe;
304 pdpe = addr_gpa2hva(vm, pml4e[index[3]].address * vm->page_size);
305 if (!pdpe[index[2]].present) {
306 pdpe[index[2]].address = vm_phy_page_alloc(vm,
307 KVM_GUEST_PAGE_TABLE_MIN_PADDR, pgd_memslot)
308 >> vm->page_shift;
309 pdpe[index[2]].writable = true;
310 pdpe[index[2]].present = true;
311 }
312
313 /* Allocate page table if not present. */
314 struct pageDirectoryEntry *pde;
315 pde = addr_gpa2hva(vm, pdpe[index[2]].address * vm->page_size);
316 if (!pde[index[1]].present) {
317 pde[index[1]].address = vm_phy_page_alloc(vm,
318 KVM_GUEST_PAGE_TABLE_MIN_PADDR, pgd_memslot)
319 >> vm->page_shift;
320 pde[index[1]].writable = true;
321 pde[index[1]].present = true;
322 }
323
324 /* Fill in page table entry. */
325 struct pageTableEntry *pte;
326 pte = addr_gpa2hva(vm, pde[index[1]].address * vm->page_size);
327 pte[index[0]].address = paddr >> vm->page_shift;
328 pte[index[0]].writable = true;
329 pte[index[0]].present = 1;
330 }
331
332 /* Virtual Translation Tables Dump
333 *
334 * Input Args:
335 * vm - Virtual Machine
336 * indent - Left margin indent amount
337 *
338 * Output Args:
339 * stream - Output FILE stream
340 *
341 * Return: None
342 *
343 * Dumps to the FILE stream given by stream, the contents of all the
344 * virtual translation tables for the VM given by vm.
345 */
virt_dump(FILE * stream,struct kvm_vm * vm,uint8_t indent)346 void virt_dump(FILE *stream, struct kvm_vm *vm, uint8_t indent)
347 {
348 struct pageMapL4Entry *pml4e, *pml4e_start;
349 struct pageDirectoryPointerEntry *pdpe, *pdpe_start;
350 struct pageDirectoryEntry *pde, *pde_start;
351 struct pageTableEntry *pte, *pte_start;
352
353 if (!vm->pgd_created)
354 return;
355
356 fprintf(stream, "%*s "
357 " no\n", indent, "");
358 fprintf(stream, "%*s index hvaddr gpaddr "
359 "addr w exec dirty\n",
360 indent, "");
361 pml4e_start = (struct pageMapL4Entry *) addr_gpa2hva(vm,
362 vm->pgd);
363 for (uint16_t n1 = 0; n1 <= 0x1ffu; n1++) {
364 pml4e = &pml4e_start[n1];
365 if (!pml4e->present)
366 continue;
367 fprintf(stream, "%*spml4e 0x%-3zx %p 0x%-12lx 0x%-10lx %u "
368 " %u\n",
369 indent, "",
370 pml4e - pml4e_start, pml4e,
371 addr_hva2gpa(vm, pml4e), (uint64_t) pml4e->address,
372 pml4e->writable, pml4e->execute_disable);
373
374 pdpe_start = addr_gpa2hva(vm, pml4e->address
375 * vm->page_size);
376 for (uint16_t n2 = 0; n2 <= 0x1ffu; n2++) {
377 pdpe = &pdpe_start[n2];
378 if (!pdpe->present)
379 continue;
380 fprintf(stream, "%*spdpe 0x%-3zx %p 0x%-12lx 0x%-10lx "
381 "%u %u\n",
382 indent, "",
383 pdpe - pdpe_start, pdpe,
384 addr_hva2gpa(vm, pdpe),
385 (uint64_t) pdpe->address, pdpe->writable,
386 pdpe->execute_disable);
387
388 pde_start = addr_gpa2hva(vm,
389 pdpe->address * vm->page_size);
390 for (uint16_t n3 = 0; n3 <= 0x1ffu; n3++) {
391 pde = &pde_start[n3];
392 if (!pde->present)
393 continue;
394 fprintf(stream, "%*spde 0x%-3zx %p "
395 "0x%-12lx 0x%-10lx %u %u\n",
396 indent, "", pde - pde_start, pde,
397 addr_hva2gpa(vm, pde),
398 (uint64_t) pde->address, pde->writable,
399 pde->execute_disable);
400
401 pte_start = addr_gpa2hva(vm,
402 pde->address * vm->page_size);
403 for (uint16_t n4 = 0; n4 <= 0x1ffu; n4++) {
404 pte = &pte_start[n4];
405 if (!pte->present)
406 continue;
407 fprintf(stream, "%*spte 0x%-3zx %p "
408 "0x%-12lx 0x%-10lx %u %u "
409 " %u 0x%-10lx\n",
410 indent, "",
411 pte - pte_start, pte,
412 addr_hva2gpa(vm, pte),
413 (uint64_t) pte->address,
414 pte->writable,
415 pte->execute_disable,
416 pte->dirty,
417 ((uint64_t) n1 << 27)
418 | ((uint64_t) n2 << 18)
419 | ((uint64_t) n3 << 9)
420 | ((uint64_t) n4));
421 }
422 }
423 }
424 }
425 }
426
427 /* Set Unusable Segment
428 *
429 * Input Args: None
430 *
431 * Output Args:
432 * segp - Pointer to segment register
433 *
434 * Return: None
435 *
436 * Sets the segment register pointed to by segp to an unusable state.
437 */
kvm_seg_set_unusable(struct kvm_segment * segp)438 static void kvm_seg_set_unusable(struct kvm_segment *segp)
439 {
440 memset(segp, 0, sizeof(*segp));
441 segp->unusable = true;
442 }
443
kvm_seg_fill_gdt_64bit(struct kvm_vm * vm,struct kvm_segment * segp)444 static void kvm_seg_fill_gdt_64bit(struct kvm_vm *vm, struct kvm_segment *segp)
445 {
446 void *gdt = addr_gva2hva(vm, vm->gdt);
447 struct desc64 *desc = gdt + (segp->selector >> 3) * 8;
448
449 desc->limit0 = segp->limit & 0xFFFF;
450 desc->base0 = segp->base & 0xFFFF;
451 desc->base1 = segp->base >> 16;
452 desc->s = segp->s;
453 desc->type = segp->type;
454 desc->dpl = segp->dpl;
455 desc->p = segp->present;
456 desc->limit1 = segp->limit >> 16;
457 desc->l = segp->l;
458 desc->db = segp->db;
459 desc->g = segp->g;
460 desc->base2 = segp->base >> 24;
461 if (!segp->s)
462 desc->base3 = segp->base >> 32;
463 }
464
465
466 /* Set Long Mode Flat Kernel Code Segment
467 *
468 * Input Args:
469 * vm - VM whose GDT is being filled, or NULL to only write segp
470 * selector - selector value
471 *
472 * Output Args:
473 * segp - Pointer to KVM segment
474 *
475 * Return: None
476 *
477 * Sets up the KVM segment pointed to by segp, to be a code segment
478 * with the selector value given by selector.
479 */
kvm_seg_set_kernel_code_64bit(struct kvm_vm * vm,uint16_t selector,struct kvm_segment * segp)480 static void kvm_seg_set_kernel_code_64bit(struct kvm_vm *vm, uint16_t selector,
481 struct kvm_segment *segp)
482 {
483 memset(segp, 0, sizeof(*segp));
484 segp->selector = selector;
485 segp->limit = 0xFFFFFFFFu;
486 segp->s = 0x1; /* kTypeCodeData */
487 segp->type = 0x08 | 0x01 | 0x02; /* kFlagCode | kFlagCodeAccessed
488 * | kFlagCodeReadable
489 */
490 segp->g = true;
491 segp->l = true;
492 segp->present = 1;
493 if (vm)
494 kvm_seg_fill_gdt_64bit(vm, segp);
495 }
496
497 /* Set Long Mode Flat Kernel Data Segment
498 *
499 * Input Args:
500 * vm - VM whose GDT is being filled, or NULL to only write segp
501 * selector - selector value
502 *
503 * Output Args:
504 * segp - Pointer to KVM segment
505 *
506 * Return: None
507 *
508 * Sets up the KVM segment pointed to by segp, to be a data segment
509 * with the selector value given by selector.
510 */
kvm_seg_set_kernel_data_64bit(struct kvm_vm * vm,uint16_t selector,struct kvm_segment * segp)511 static void kvm_seg_set_kernel_data_64bit(struct kvm_vm *vm, uint16_t selector,
512 struct kvm_segment *segp)
513 {
514 memset(segp, 0, sizeof(*segp));
515 segp->selector = selector;
516 segp->limit = 0xFFFFFFFFu;
517 segp->s = 0x1; /* kTypeCodeData */
518 segp->type = 0x00 | 0x01 | 0x02; /* kFlagData | kFlagDataAccessed
519 * | kFlagDataWritable
520 */
521 segp->g = true;
522 segp->present = true;
523 if (vm)
524 kvm_seg_fill_gdt_64bit(vm, segp);
525 }
526
527 /* Address Guest Virtual to Guest Physical
528 *
529 * Input Args:
530 * vm - Virtual Machine
531 * gpa - VM virtual address
532 *
533 * Output Args: None
534 *
535 * Return:
536 * Equivalent VM physical address
537 *
538 * Translates the VM virtual address given by gva to a VM physical
539 * address and then locates the memory region containing the VM
540 * physical address, within the VM given by vm. When found, the host
541 * virtual address providing the memory to the vm physical address is returned.
542 * A TEST_ASSERT failure occurs if no region containing translated
543 * VM virtual address exists.
544 */
addr_gva2gpa(struct kvm_vm * vm,vm_vaddr_t gva)545 vm_paddr_t addr_gva2gpa(struct kvm_vm *vm, vm_vaddr_t gva)
546 {
547 uint16_t index[4];
548 struct pageMapL4Entry *pml4e;
549 struct pageDirectoryPointerEntry *pdpe;
550 struct pageDirectoryEntry *pde;
551 struct pageTableEntry *pte;
552 void *hva;
553
554 TEST_ASSERT(vm->mode == VM_MODE_FLAT48PG, "Attempt to use "
555 "unknown or unsupported guest mode, mode: 0x%x", vm->mode);
556
557 index[0] = (gva >> 12) & 0x1ffu;
558 index[1] = (gva >> 21) & 0x1ffu;
559 index[2] = (gva >> 30) & 0x1ffu;
560 index[3] = (gva >> 39) & 0x1ffu;
561
562 if (!vm->pgd_created)
563 goto unmapped_gva;
564 pml4e = addr_gpa2hva(vm, vm->pgd);
565 if (!pml4e[index[3]].present)
566 goto unmapped_gva;
567
568 pdpe = addr_gpa2hva(vm, pml4e[index[3]].address * vm->page_size);
569 if (!pdpe[index[2]].present)
570 goto unmapped_gva;
571
572 pde = addr_gpa2hva(vm, pdpe[index[2]].address * vm->page_size);
573 if (!pde[index[1]].present)
574 goto unmapped_gva;
575
576 pte = addr_gpa2hva(vm, pde[index[1]].address * vm->page_size);
577 if (!pte[index[0]].present)
578 goto unmapped_gva;
579
580 return (pte[index[0]].address * vm->page_size) + (gva & 0xfffu);
581
582 unmapped_gva:
583 TEST_ASSERT(false, "No mapping for vm virtual address, "
584 "gva: 0x%lx", gva);
585 }
586
kvm_setup_gdt(struct kvm_vm * vm,struct kvm_dtable * dt,int gdt_memslot,int pgd_memslot)587 static void kvm_setup_gdt(struct kvm_vm *vm, struct kvm_dtable *dt, int gdt_memslot,
588 int pgd_memslot)
589 {
590 if (!vm->gdt)
591 vm->gdt = vm_vaddr_alloc(vm, getpagesize(),
592 KVM_UTIL_MIN_VADDR, gdt_memslot, pgd_memslot);
593
594 dt->base = vm->gdt;
595 dt->limit = getpagesize();
596 }
597
kvm_setup_tss_64bit(struct kvm_vm * vm,struct kvm_segment * segp,int selector,int gdt_memslot,int pgd_memslot)598 static void kvm_setup_tss_64bit(struct kvm_vm *vm, struct kvm_segment *segp,
599 int selector, int gdt_memslot,
600 int pgd_memslot)
601 {
602 if (!vm->tss)
603 vm->tss = vm_vaddr_alloc(vm, getpagesize(),
604 KVM_UTIL_MIN_VADDR, gdt_memslot, pgd_memslot);
605
606 memset(segp, 0, sizeof(*segp));
607 segp->base = vm->tss;
608 segp->limit = 0x67;
609 segp->selector = selector;
610 segp->type = 0xb;
611 segp->present = 1;
612 kvm_seg_fill_gdt_64bit(vm, segp);
613 }
614
vcpu_setup(struct kvm_vm * vm,int vcpuid,int pgd_memslot,int gdt_memslot)615 void vcpu_setup(struct kvm_vm *vm, int vcpuid, int pgd_memslot, int gdt_memslot)
616 {
617 struct kvm_sregs sregs;
618
619 /* Set mode specific system register values. */
620 vcpu_sregs_get(vm, vcpuid, &sregs);
621
622 sregs.idt.limit = 0;
623
624 kvm_setup_gdt(vm, &sregs.gdt, gdt_memslot, pgd_memslot);
625
626 switch (vm->mode) {
627 case VM_MODE_FLAT48PG:
628 sregs.cr0 = X86_CR0_PE | X86_CR0_NE | X86_CR0_PG;
629 sregs.cr4 |= X86_CR4_PAE;
630 sregs.efer |= (EFER_LME | EFER_LMA | EFER_NX);
631
632 kvm_seg_set_unusable(&sregs.ldt);
633 kvm_seg_set_kernel_code_64bit(vm, 0x8, &sregs.cs);
634 kvm_seg_set_kernel_data_64bit(vm, 0x10, &sregs.ds);
635 kvm_seg_set_kernel_data_64bit(vm, 0x10, &sregs.es);
636 kvm_setup_tss_64bit(vm, &sregs.tr, 0x18, gdt_memslot, pgd_memslot);
637 break;
638
639 default:
640 TEST_ASSERT(false, "Unknown guest mode, mode: 0x%x", vm->mode);
641 }
642
643 sregs.cr3 = vm->pgd;
644 vcpu_sregs_set(vm, vcpuid, &sregs);
645 }
646 /* Adds a vCPU with reasonable defaults (i.e., a stack)
647 *
648 * Input Args:
649 * vcpuid - The id of the VCPU to add to the VM.
650 * guest_code - The vCPU's entry point
651 */
vm_vcpu_add_default(struct kvm_vm * vm,uint32_t vcpuid,void * guest_code)652 void vm_vcpu_add_default(struct kvm_vm *vm, uint32_t vcpuid, void *guest_code)
653 {
654 struct kvm_mp_state mp_state;
655 struct kvm_regs regs;
656 vm_vaddr_t stack_vaddr;
657 stack_vaddr = vm_vaddr_alloc(vm, DEFAULT_STACK_PGS * getpagesize(),
658 DEFAULT_GUEST_STACK_VADDR_MIN, 0, 0);
659
660 /* Create VCPU */
661 vm_vcpu_add(vm, vcpuid, 0, 0);
662
663 /* Setup guest general purpose registers */
664 vcpu_regs_get(vm, vcpuid, ®s);
665 regs.rflags = regs.rflags | 0x2;
666 regs.rsp = stack_vaddr + (DEFAULT_STACK_PGS * getpagesize());
667 regs.rip = (unsigned long) guest_code;
668 vcpu_regs_set(vm, vcpuid, ®s);
669
670 /* Setup the MP state */
671 mp_state.mp_state = 0;
672 vcpu_set_mp_state(vm, vcpuid, &mp_state);
673 }
674
675 /* VM VCPU CPUID Set
676 *
677 * Input Args:
678 * vm - Virtual Machine
679 * vcpuid - VCPU id
680 * cpuid - The CPUID values to set.
681 *
682 * Output Args: None
683 *
684 * Return: void
685 *
686 * Set the VCPU's CPUID.
687 */
vcpu_set_cpuid(struct kvm_vm * vm,uint32_t vcpuid,struct kvm_cpuid2 * cpuid)688 void vcpu_set_cpuid(struct kvm_vm *vm,
689 uint32_t vcpuid, struct kvm_cpuid2 *cpuid)
690 {
691 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
692 int rc;
693
694 TEST_ASSERT(vcpu != NULL, "vcpu not found, vcpuid: %u", vcpuid);
695
696 rc = ioctl(vcpu->fd, KVM_SET_CPUID2, cpuid);
697 TEST_ASSERT(rc == 0, "KVM_SET_CPUID2 failed, rc: %i errno: %i",
698 rc, errno);
699
700 }
701 /* Create a VM with reasonable defaults
702 *
703 * Input Args:
704 * vcpuid - The id of the single VCPU to add to the VM.
705 * extra_mem_pages - The size of extra memories to add (this will
706 * decide how much extra space we will need to
707 * setup the page tables using mem slot 0)
708 * guest_code - The vCPU's entry point
709 *
710 * Output Args: None
711 *
712 * Return:
713 * Pointer to opaque structure that describes the created VM.
714 */
vm_create_default(uint32_t vcpuid,uint64_t extra_mem_pages,void * guest_code)715 struct kvm_vm *vm_create_default(uint32_t vcpuid, uint64_t extra_mem_pages,
716 void *guest_code)
717 {
718 struct kvm_vm *vm;
719 /*
720 * For x86 the maximum page table size for a memory region
721 * will be when only 4K pages are used. In that case the
722 * total extra size for page tables (for extra N pages) will
723 * be: N/512+N/512^2+N/512^3+... which is definitely smaller
724 * than N/512*2.
725 */
726 uint64_t extra_pg_pages = extra_mem_pages / 512 * 2;
727
728 /* Create VM */
729 vm = vm_create(VM_MODE_FLAT48PG,
730 DEFAULT_GUEST_PHY_PAGES + extra_pg_pages,
731 O_RDWR);
732
733 /* Setup guest code */
734 kvm_vm_elf_load(vm, program_invocation_name, 0, 0);
735
736 /* Setup IRQ Chip */
737 vm_create_irqchip(vm);
738
739 /* Add the first vCPU. */
740 vm_vcpu_add_default(vm, vcpuid, guest_code);
741
742 return vm;
743 }
744
745 struct kvm_x86_state {
746 struct kvm_vcpu_events events;
747 struct kvm_mp_state mp_state;
748 struct kvm_regs regs;
749 struct kvm_xsave xsave;
750 struct kvm_xcrs xcrs;
751 struct kvm_sregs sregs;
752 struct kvm_debugregs debugregs;
753 union {
754 struct kvm_nested_state nested;
755 char nested_[16384];
756 };
757 struct kvm_msrs msrs;
758 };
759
kvm_get_num_msrs(struct kvm_vm * vm)760 static int kvm_get_num_msrs(struct kvm_vm *vm)
761 {
762 struct kvm_msr_list nmsrs;
763 int r;
764
765 nmsrs.nmsrs = 0;
766 r = ioctl(vm->kvm_fd, KVM_GET_MSR_INDEX_LIST, &nmsrs);
767 TEST_ASSERT(r == -1 && errno == E2BIG, "Unexpected result from KVM_GET_MSR_INDEX_LIST probe, r: %i",
768 r);
769
770 return nmsrs.nmsrs;
771 }
772
vcpu_save_state(struct kvm_vm * vm,uint32_t vcpuid)773 struct kvm_x86_state *vcpu_save_state(struct kvm_vm *vm, uint32_t vcpuid)
774 {
775 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
776 struct kvm_msr_list *list;
777 struct kvm_x86_state *state;
778 int nmsrs, r, i;
779 static int nested_size = -1;
780
781 if (nested_size == -1) {
782 nested_size = kvm_check_cap(KVM_CAP_NESTED_STATE);
783 TEST_ASSERT(nested_size <= sizeof(state->nested_),
784 "Nested state size too big, %i > %zi",
785 nested_size, sizeof(state->nested_));
786 }
787
788 nmsrs = kvm_get_num_msrs(vm);
789 list = malloc(sizeof(*list) + nmsrs * sizeof(list->indices[0]));
790 list->nmsrs = nmsrs;
791 r = ioctl(vm->kvm_fd, KVM_GET_MSR_INDEX_LIST, list);
792 TEST_ASSERT(r == 0, "Unexpected result from KVM_GET_MSR_INDEX_LIST, r: %i",
793 r);
794
795 state = malloc(sizeof(*state) + nmsrs * sizeof(state->msrs.entries[0]));
796 r = ioctl(vcpu->fd, KVM_GET_VCPU_EVENTS, &state->events);
797 TEST_ASSERT(r == 0, "Unexpected result from KVM_GET_VCPU_EVENTS, r: %i",
798 r);
799
800 r = ioctl(vcpu->fd, KVM_GET_MP_STATE, &state->mp_state);
801 TEST_ASSERT(r == 0, "Unexpected result from KVM_GET_MP_STATE, r: %i",
802 r);
803
804 r = ioctl(vcpu->fd, KVM_GET_REGS, &state->regs);
805 TEST_ASSERT(r == 0, "Unexpected result from KVM_GET_REGS, r: %i",
806 r);
807
808 r = ioctl(vcpu->fd, KVM_GET_XSAVE, &state->xsave);
809 TEST_ASSERT(r == 0, "Unexpected result from KVM_GET_XSAVE, r: %i",
810 r);
811
812 r = ioctl(vcpu->fd, KVM_GET_XCRS, &state->xcrs);
813 TEST_ASSERT(r == 0, "Unexpected result from KVM_GET_XCRS, r: %i",
814 r);
815
816 r = ioctl(vcpu->fd, KVM_GET_SREGS, &state->sregs);
817 TEST_ASSERT(r == 0, "Unexpected result from KVM_GET_SREGS, r: %i",
818 r);
819
820 if (nested_size) {
821 state->nested.size = sizeof(state->nested_);
822 r = ioctl(vcpu->fd, KVM_GET_NESTED_STATE, &state->nested);
823 TEST_ASSERT(r == 0, "Unexpected result from KVM_GET_NESTED_STATE, r: %i",
824 r);
825 TEST_ASSERT(state->nested.size <= nested_size,
826 "Nested state size too big, %i (KVM_CHECK_CAP gave %i)",
827 state->nested.size, nested_size);
828 } else
829 state->nested.size = 0;
830
831 state->msrs.nmsrs = nmsrs;
832 for (i = 0; i < nmsrs; i++)
833 state->msrs.entries[i].index = list->indices[i];
834 r = ioctl(vcpu->fd, KVM_GET_MSRS, &state->msrs);
835 TEST_ASSERT(r == nmsrs, "Unexpected result from KVM_GET_MSRS, r: %i (failed at %x)",
836 r, r == nmsrs ? -1 : list->indices[r]);
837
838 r = ioctl(vcpu->fd, KVM_GET_DEBUGREGS, &state->debugregs);
839 TEST_ASSERT(r == 0, "Unexpected result from KVM_GET_DEBUGREGS, r: %i",
840 r);
841
842 free(list);
843 return state;
844 }
845
vcpu_load_state(struct kvm_vm * vm,uint32_t vcpuid,struct kvm_x86_state * state)846 void vcpu_load_state(struct kvm_vm *vm, uint32_t vcpuid, struct kvm_x86_state *state)
847 {
848 struct vcpu *vcpu = vcpu_find(vm, vcpuid);
849 int r;
850
851 if (state->nested.size) {
852 r = ioctl(vcpu->fd, KVM_SET_NESTED_STATE, &state->nested);
853 TEST_ASSERT(r == 0, "Unexpected result from KVM_SET_NESTED_STATE, r: %i",
854 r);
855 }
856
857 r = ioctl(vcpu->fd, KVM_SET_XSAVE, &state->xsave);
858 TEST_ASSERT(r == 0, "Unexpected result from KVM_SET_XSAVE, r: %i",
859 r);
860
861 r = ioctl(vcpu->fd, KVM_SET_XCRS, &state->xcrs);
862 TEST_ASSERT(r == 0, "Unexpected result from KVM_SET_XCRS, r: %i",
863 r);
864
865 r = ioctl(vcpu->fd, KVM_SET_SREGS, &state->sregs);
866 TEST_ASSERT(r == 0, "Unexpected result from KVM_SET_SREGS, r: %i",
867 r);
868
869 r = ioctl(vcpu->fd, KVM_SET_MSRS, &state->msrs);
870 TEST_ASSERT(r == state->msrs.nmsrs, "Unexpected result from KVM_SET_MSRS, r: %i (failed at %x)",
871 r, r == state->msrs.nmsrs ? -1 : state->msrs.entries[r].index);
872
873 r = ioctl(vcpu->fd, KVM_SET_VCPU_EVENTS, &state->events);
874 TEST_ASSERT(r == 0, "Unexpected result from KVM_SET_VCPU_EVENTS, r: %i",
875 r);
876
877 r = ioctl(vcpu->fd, KVM_SET_MP_STATE, &state->mp_state);
878 TEST_ASSERT(r == 0, "Unexpected result from KVM_SET_MP_STATE, r: %i",
879 r);
880
881 r = ioctl(vcpu->fd, KVM_SET_DEBUGREGS, &state->debugregs);
882 TEST_ASSERT(r == 0, "Unexpected result from KVM_SET_DEBUGREGS, r: %i",
883 r);
884
885 r = ioctl(vcpu->fd, KVM_SET_REGS, &state->regs);
886 TEST_ASSERT(r == 0, "Unexpected result from KVM_SET_REGS, r: %i",
887 r);
888 }
889
