xref: /kvm-unit-tests/x86/vmx_tests.c (revision 1868c05a96ef600f5caf6fc4b5f15f6340d35261)
1 /*
2  * All test cases of nested virtualization should be in this file
3  *
4  * Author : Arthur Chunqi Li <yzt356@gmail.com>
5  */
6 
7 #include <asm/debugreg.h>
8 
9 #include "vmx.h"
10 #include "msr.h"
11 #include "processor.h"
12 #include "vm.h"
13 #include "pci.h"
14 #include "fwcfg.h"
15 #include "isr.h"
16 #include "desc.h"
17 #include "apic.h"
18 #include "types.h"
19 #include "vmalloc.h"
20 #include "alloc_page.h"
21 #include "smp.h"
22 #include "delay.h"
23 
24 #define NONCANONICAL            0xaaaaaaaaaaaaaaaaull
25 
26 #define VPID_CAP_INVVPID_TYPES_SHIFT 40
27 
28 u64 ia32_pat;
29 u64 ia32_efer;
30 void *io_bitmap_a, *io_bitmap_b;
31 u16 ioport;
32 
33 unsigned long *pml4;
34 u64 eptp;
35 void *data_page1, *data_page2;
36 
37 phys_addr_t pci_physaddr;
38 
39 void *pml_log;
40 #define PML_INDEX 512
41 
42 static inline unsigned ffs(unsigned x)
43 {
44 	int pos = -1;
45 
46 	__asm__ __volatile__("bsf %1, %%eax; cmovnz %%eax, %0"
47 			     : "+r"(pos) : "rm"(x) : "eax");
48 	return pos + 1;
49 }
50 
51 static inline void vmcall(void)
52 {
53 	asm volatile("vmcall");
54 }
55 
56 static void basic_guest_main(void)
57 {
58 	report("Basic VMX test", 1);
59 }
60 
61 static int basic_exit_handler(void)
62 {
63 	report("Basic VMX test", 0);
64 	print_vmexit_info();
65 	return VMX_TEST_EXIT;
66 }
67 
68 static void vmenter_main(void)
69 {
70 	u64 rax;
71 	u64 rsp, resume_rsp;
72 
73 	report("test vmlaunch", 1);
74 
75 	asm volatile(
76 		"mov %%rsp, %0\n\t"
77 		"mov %3, %%rax\n\t"
78 		"vmcall\n\t"
79 		"mov %%rax, %1\n\t"
80 		"mov %%rsp, %2\n\t"
81 		: "=r"(rsp), "=r"(rax), "=r"(resume_rsp)
82 		: "g"(0xABCD));
83 	report("test vmresume", (rax == 0xFFFF) && (rsp == resume_rsp));
84 }
85 
86 static int vmenter_exit_handler(void)
87 {
88 	u64 guest_rip;
89 	ulong reason;
90 
91 	guest_rip = vmcs_read(GUEST_RIP);
92 	reason = vmcs_read(EXI_REASON) & 0xff;
93 	switch (reason) {
94 	case VMX_VMCALL:
95 		if (regs.rax != 0xABCD) {
96 			report("test vmresume", 0);
97 			return VMX_TEST_VMEXIT;
98 		}
99 		regs.rax = 0xFFFF;
100 		vmcs_write(GUEST_RIP, guest_rip + 3);
101 		return VMX_TEST_RESUME;
102 	default:
103 		report("test vmresume", 0);
104 		print_vmexit_info();
105 	}
106 	return VMX_TEST_VMEXIT;
107 }
108 
109 u32 preempt_scale;
110 volatile unsigned long long tsc_val;
111 volatile u32 preempt_val;
112 u64 saved_rip;
113 
114 static int preemption_timer_init(struct vmcs *vmcs)
115 {
116 	if (!(ctrl_pin_rev.clr & PIN_PREEMPT)) {
117 		printf("\tPreemption timer is not supported\n");
118 		return VMX_TEST_EXIT;
119 	}
120 	vmcs_write(PIN_CONTROLS, vmcs_read(PIN_CONTROLS) | PIN_PREEMPT);
121 	preempt_val = 10000000;
122 	vmcs_write(PREEMPT_TIMER_VALUE, preempt_val);
123 	preempt_scale = rdmsr(MSR_IA32_VMX_MISC) & 0x1F;
124 
125 	if (!(ctrl_exit_rev.clr & EXI_SAVE_PREEMPT))
126 		printf("\tSave preemption value is not supported\n");
127 
128 	return VMX_TEST_START;
129 }
130 
131 static void preemption_timer_main(void)
132 {
133 	tsc_val = rdtsc();
134 	if (ctrl_exit_rev.clr & EXI_SAVE_PREEMPT) {
135 		vmx_set_test_stage(0);
136 		vmcall();
137 		if (vmx_get_test_stage() == 1)
138 			vmcall();
139 	}
140 	vmx_set_test_stage(1);
141 	while (vmx_get_test_stage() == 1) {
142 		if (((rdtsc() - tsc_val) >> preempt_scale)
143 				> 10 * preempt_val) {
144 			vmx_set_test_stage(2);
145 			vmcall();
146 		}
147 	}
148 	tsc_val = rdtsc();
149 	asm volatile ("hlt");
150 	vmcall();
151 	vmx_set_test_stage(5);
152 	vmcall();
153 }
154 
155 static int preemption_timer_exit_handler(void)
156 {
157 	bool guest_halted;
158 	u64 guest_rip;
159 	ulong reason;
160 	u32 insn_len;
161 	u32 ctrl_exit;
162 
163 	guest_rip = vmcs_read(GUEST_RIP);
164 	reason = vmcs_read(EXI_REASON) & 0xff;
165 	insn_len = vmcs_read(EXI_INST_LEN);
166 	switch (reason) {
167 	case VMX_PREEMPT:
168 		switch (vmx_get_test_stage()) {
169 		case 1:
170 		case 2:
171 			report("busy-wait for preemption timer",
172 			       ((rdtsc() - tsc_val) >> preempt_scale) >=
173 			       preempt_val);
174 			vmx_set_test_stage(3);
175 			vmcs_write(PREEMPT_TIMER_VALUE, preempt_val);
176 			return VMX_TEST_RESUME;
177 		case 3:
178 			guest_halted =
179 				(vmcs_read(GUEST_ACTV_STATE) == ACTV_HLT);
180 			report("preemption timer during hlt",
181 			       ((rdtsc() - tsc_val) >> preempt_scale) >=
182 			       preempt_val && guest_halted);
183 			vmx_set_test_stage(4);
184 			vmcs_write(PIN_CONTROLS,
185 				   vmcs_read(PIN_CONTROLS) & ~PIN_PREEMPT);
186 			vmcs_write(EXI_CONTROLS,
187 				   vmcs_read(EXI_CONTROLS) & ~EXI_SAVE_PREEMPT);
188 			vmcs_write(GUEST_ACTV_STATE, ACTV_ACTIVE);
189 			return VMX_TEST_RESUME;
190 		case 4:
191 			report("preemption timer with 0 value",
192 			       saved_rip == guest_rip);
193 			break;
194 		default:
195 			report("Invalid stage.", false);
196 			print_vmexit_info();
197 			break;
198 		}
199 		break;
200 	case VMX_VMCALL:
201 		vmcs_write(GUEST_RIP, guest_rip + insn_len);
202 		switch (vmx_get_test_stage()) {
203 		case 0:
204 			report("Keep preemption value",
205 			       vmcs_read(PREEMPT_TIMER_VALUE) == preempt_val);
206 			vmx_set_test_stage(1);
207 			vmcs_write(PREEMPT_TIMER_VALUE, preempt_val);
208 			ctrl_exit = (vmcs_read(EXI_CONTROLS) |
209 				EXI_SAVE_PREEMPT) & ctrl_exit_rev.clr;
210 			vmcs_write(EXI_CONTROLS, ctrl_exit);
211 			return VMX_TEST_RESUME;
212 		case 1:
213 			report("Save preemption value",
214 			       vmcs_read(PREEMPT_TIMER_VALUE) < preempt_val);
215 			return VMX_TEST_RESUME;
216 		case 2:
217 			report("busy-wait for preemption timer", 0);
218 			vmx_set_test_stage(3);
219 			vmcs_write(PREEMPT_TIMER_VALUE, preempt_val);
220 			return VMX_TEST_RESUME;
221 		case 3:
222 			report("preemption timer during hlt", 0);
223 			vmx_set_test_stage(4);
224 			/* fall through */
225 		case 4:
226 			vmcs_write(PIN_CONTROLS,
227 				   vmcs_read(PIN_CONTROLS) | PIN_PREEMPT);
228 			vmcs_write(PREEMPT_TIMER_VALUE, 0);
229 			saved_rip = guest_rip + insn_len;
230 			return VMX_TEST_RESUME;
231 		case 5:
232 			report("preemption timer with 0 value (vmcall stage 5)", 0);
233 			break;
234 		default:
235 			// Should not reach here
236 			report("unexpected stage, %d", false,
237 			       vmx_get_test_stage());
238 			print_vmexit_info();
239 			return VMX_TEST_VMEXIT;
240 		}
241 		break;
242 	default:
243 		report("Unknown exit reason, %ld", false, reason);
244 		print_vmexit_info();
245 	}
246 	vmcs_write(PIN_CONTROLS, vmcs_read(PIN_CONTROLS) & ~PIN_PREEMPT);
247 	return VMX_TEST_VMEXIT;
248 }
249 
250 static void msr_bmp_init(void)
251 {
252 	void *msr_bitmap;
253 	u32 ctrl_cpu0;
254 
255 	msr_bitmap = alloc_page();
256 	memset(msr_bitmap, 0x0, PAGE_SIZE);
257 	ctrl_cpu0 = vmcs_read(CPU_EXEC_CTRL0);
258 	ctrl_cpu0 |= CPU_MSR_BITMAP;
259 	vmcs_write(CPU_EXEC_CTRL0, ctrl_cpu0);
260 	vmcs_write(MSR_BITMAP, (u64)msr_bitmap);
261 }
262 
263 static void *get_msr_bitmap(void)
264 {
265 	void *msr_bitmap;
266 
267 	if (vmcs_read(CPU_EXEC_CTRL0) & CPU_MSR_BITMAP) {
268 		msr_bitmap = (void *)vmcs_read(MSR_BITMAP);
269 	} else {
270 		msr_bitmap = alloc_page();
271 		memset(msr_bitmap, 0xff, PAGE_SIZE);
272 		vmcs_write(MSR_BITMAP, (u64)msr_bitmap);
273 		vmcs_set_bits(CPU_EXEC_CTRL0, CPU_MSR_BITMAP);
274 	}
275 
276 	return msr_bitmap;
277 }
278 
279 static void disable_intercept_for_x2apic_msrs(void)
280 {
281 	unsigned long *msr_bitmap = (unsigned long *)get_msr_bitmap();
282 	u32 msr;
283 
284 	for (msr = APIC_BASE_MSR;
285 		 msr < (APIC_BASE_MSR+0xff);
286 		 msr += BITS_PER_LONG) {
287 		unsigned int word = msr / BITS_PER_LONG;
288 
289 		msr_bitmap[word] = 0;
290 		msr_bitmap[word + (0x800 / sizeof(long))] = 0;
291 	}
292 }
293 
294 static int test_ctrl_pat_init(struct vmcs *vmcs)
295 {
296 	u64 ctrl_ent;
297 	u64 ctrl_exi;
298 
299 	msr_bmp_init();
300 	if (!(ctrl_exit_rev.clr & EXI_SAVE_PAT) &&
301 	    !(ctrl_exit_rev.clr & EXI_LOAD_PAT) &&
302 	    !(ctrl_enter_rev.clr & ENT_LOAD_PAT)) {
303 		printf("\tSave/load PAT is not supported\n");
304 		return 1;
305 	}
306 
307 	ctrl_ent = vmcs_read(ENT_CONTROLS);
308 	ctrl_exi = vmcs_read(EXI_CONTROLS);
309 	ctrl_ent |= ctrl_enter_rev.clr & ENT_LOAD_PAT;
310 	ctrl_exi |= ctrl_exit_rev.clr & (EXI_SAVE_PAT | EXI_LOAD_PAT);
311 	vmcs_write(ENT_CONTROLS, ctrl_ent);
312 	vmcs_write(EXI_CONTROLS, ctrl_exi);
313 	ia32_pat = rdmsr(MSR_IA32_CR_PAT);
314 	vmcs_write(GUEST_PAT, 0x0);
315 	vmcs_write(HOST_PAT, ia32_pat);
316 	return VMX_TEST_START;
317 }
318 
319 static void test_ctrl_pat_main(void)
320 {
321 	u64 guest_ia32_pat;
322 
323 	guest_ia32_pat = rdmsr(MSR_IA32_CR_PAT);
324 	if (!(ctrl_enter_rev.clr & ENT_LOAD_PAT))
325 		printf("\tENT_LOAD_PAT is not supported.\n");
326 	else {
327 		if (guest_ia32_pat != 0) {
328 			report("Entry load PAT", 0);
329 			return;
330 		}
331 	}
332 	wrmsr(MSR_IA32_CR_PAT, 0x6);
333 	vmcall();
334 	guest_ia32_pat = rdmsr(MSR_IA32_CR_PAT);
335 	if (ctrl_enter_rev.clr & ENT_LOAD_PAT)
336 		report("Entry load PAT", guest_ia32_pat == ia32_pat);
337 }
338 
339 static int test_ctrl_pat_exit_handler(void)
340 {
341 	u64 guest_rip;
342 	ulong reason;
343 	u64 guest_pat;
344 
345 	guest_rip = vmcs_read(GUEST_RIP);
346 	reason = vmcs_read(EXI_REASON) & 0xff;
347 	switch (reason) {
348 	case VMX_VMCALL:
349 		guest_pat = vmcs_read(GUEST_PAT);
350 		if (!(ctrl_exit_rev.clr & EXI_SAVE_PAT)) {
351 			printf("\tEXI_SAVE_PAT is not supported\n");
352 			vmcs_write(GUEST_PAT, 0x6);
353 		} else {
354 			report("Exit save PAT", guest_pat == 0x6);
355 		}
356 		if (!(ctrl_exit_rev.clr & EXI_LOAD_PAT))
357 			printf("\tEXI_LOAD_PAT is not supported\n");
358 		else
359 			report("Exit load PAT", rdmsr(MSR_IA32_CR_PAT) == ia32_pat);
360 		vmcs_write(GUEST_PAT, ia32_pat);
361 		vmcs_write(GUEST_RIP, guest_rip + 3);
362 		return VMX_TEST_RESUME;
363 	default:
364 		printf("ERROR : Undefined exit reason, reason = %ld.\n", reason);
365 		break;
366 	}
367 	return VMX_TEST_VMEXIT;
368 }
369 
370 static int test_ctrl_efer_init(struct vmcs *vmcs)
371 {
372 	u64 ctrl_ent;
373 	u64 ctrl_exi;
374 
375 	msr_bmp_init();
376 	ctrl_ent = vmcs_read(ENT_CONTROLS) | ENT_LOAD_EFER;
377 	ctrl_exi = vmcs_read(EXI_CONTROLS) | EXI_SAVE_EFER | EXI_LOAD_EFER;
378 	vmcs_write(ENT_CONTROLS, ctrl_ent & ctrl_enter_rev.clr);
379 	vmcs_write(EXI_CONTROLS, ctrl_exi & ctrl_exit_rev.clr);
380 	ia32_efer = rdmsr(MSR_EFER);
381 	vmcs_write(GUEST_EFER, ia32_efer ^ EFER_NX);
382 	vmcs_write(HOST_EFER, ia32_efer ^ EFER_NX);
383 	return VMX_TEST_START;
384 }
385 
386 static void test_ctrl_efer_main(void)
387 {
388 	u64 guest_ia32_efer;
389 
390 	guest_ia32_efer = rdmsr(MSR_EFER);
391 	if (!(ctrl_enter_rev.clr & ENT_LOAD_EFER))
392 		printf("\tENT_LOAD_EFER is not supported.\n");
393 	else {
394 		if (guest_ia32_efer != (ia32_efer ^ EFER_NX)) {
395 			report("Entry load EFER", 0);
396 			return;
397 		}
398 	}
399 	wrmsr(MSR_EFER, ia32_efer);
400 	vmcall();
401 	guest_ia32_efer = rdmsr(MSR_EFER);
402 	if (ctrl_enter_rev.clr & ENT_LOAD_EFER)
403 		report("Entry load EFER", guest_ia32_efer == ia32_efer);
404 }
405 
406 static int test_ctrl_efer_exit_handler(void)
407 {
408 	u64 guest_rip;
409 	ulong reason;
410 	u64 guest_efer;
411 
412 	guest_rip = vmcs_read(GUEST_RIP);
413 	reason = vmcs_read(EXI_REASON) & 0xff;
414 	switch (reason) {
415 	case VMX_VMCALL:
416 		guest_efer = vmcs_read(GUEST_EFER);
417 		if (!(ctrl_exit_rev.clr & EXI_SAVE_EFER)) {
418 			printf("\tEXI_SAVE_EFER is not supported\n");
419 			vmcs_write(GUEST_EFER, ia32_efer);
420 		} else {
421 			report("Exit save EFER", guest_efer == ia32_efer);
422 		}
423 		if (!(ctrl_exit_rev.clr & EXI_LOAD_EFER)) {
424 			printf("\tEXI_LOAD_EFER is not supported\n");
425 			wrmsr(MSR_EFER, ia32_efer ^ EFER_NX);
426 		} else {
427 			report("Exit load EFER", rdmsr(MSR_EFER) == (ia32_efer ^ EFER_NX));
428 		}
429 		vmcs_write(GUEST_PAT, ia32_efer);
430 		vmcs_write(GUEST_RIP, guest_rip + 3);
431 		return VMX_TEST_RESUME;
432 	default:
433 		printf("ERROR : Undefined exit reason, reason = %ld.\n", reason);
434 		break;
435 	}
436 	return VMX_TEST_VMEXIT;
437 }
438 
439 u32 guest_cr0, guest_cr4;
440 
441 static void cr_shadowing_main(void)
442 {
443 	u32 cr0, cr4, tmp;
444 
445 	// Test read through
446 	vmx_set_test_stage(0);
447 	guest_cr0 = read_cr0();
448 	if (vmx_get_test_stage() == 1)
449 		report("Read through CR0", 0);
450 	else
451 		vmcall();
452 	vmx_set_test_stage(1);
453 	guest_cr4 = read_cr4();
454 	if (vmx_get_test_stage() == 2)
455 		report("Read through CR4", 0);
456 	else
457 		vmcall();
458 	// Test write through
459 	guest_cr0 = guest_cr0 ^ (X86_CR0_TS | X86_CR0_MP);
460 	guest_cr4 = guest_cr4 ^ (X86_CR4_TSD | X86_CR4_DE);
461 	vmx_set_test_stage(2);
462 	write_cr0(guest_cr0);
463 	if (vmx_get_test_stage() == 3)
464 		report("Write throuth CR0", 0);
465 	else
466 		vmcall();
467 	vmx_set_test_stage(3);
468 	write_cr4(guest_cr4);
469 	if (vmx_get_test_stage() == 4)
470 		report("Write through CR4", 0);
471 	else
472 		vmcall();
473 	// Test read shadow
474 	vmx_set_test_stage(4);
475 	vmcall();
476 	cr0 = read_cr0();
477 	if (vmx_get_test_stage() != 5)
478 		report("Read shadowing CR0", cr0 == guest_cr0);
479 	vmx_set_test_stage(5);
480 	cr4 = read_cr4();
481 	if (vmx_get_test_stage() != 6)
482 		report("Read shadowing CR4", cr4 == guest_cr4);
483 	// Test write shadow (same value with shadow)
484 	vmx_set_test_stage(6);
485 	write_cr0(guest_cr0);
486 	if (vmx_get_test_stage() == 7)
487 		report("Write shadowing CR0 (same value with shadow)", 0);
488 	else
489 		vmcall();
490 	vmx_set_test_stage(7);
491 	write_cr4(guest_cr4);
492 	if (vmx_get_test_stage() == 8)
493 		report("Write shadowing CR4 (same value with shadow)", 0);
494 	else
495 		vmcall();
496 	// Test write shadow (different value)
497 	vmx_set_test_stage(8);
498 	tmp = guest_cr0 ^ X86_CR0_TS;
499 	asm volatile("mov %0, %%rsi\n\t"
500 		"mov %%rsi, %%cr0\n\t"
501 		::"m"(tmp)
502 		:"rsi", "memory", "cc");
503 	report("Write shadowing different X86_CR0_TS", vmx_get_test_stage() == 9);
504 	vmx_set_test_stage(9);
505 	tmp = guest_cr0 ^ X86_CR0_MP;
506 	asm volatile("mov %0, %%rsi\n\t"
507 		"mov %%rsi, %%cr0\n\t"
508 		::"m"(tmp)
509 		:"rsi", "memory", "cc");
510 	report("Write shadowing different X86_CR0_MP", vmx_get_test_stage() == 10);
511 	vmx_set_test_stage(10);
512 	tmp = guest_cr4 ^ X86_CR4_TSD;
513 	asm volatile("mov %0, %%rsi\n\t"
514 		"mov %%rsi, %%cr4\n\t"
515 		::"m"(tmp)
516 		:"rsi", "memory", "cc");
517 	report("Write shadowing different X86_CR4_TSD", vmx_get_test_stage() == 11);
518 	vmx_set_test_stage(11);
519 	tmp = guest_cr4 ^ X86_CR4_DE;
520 	asm volatile("mov %0, %%rsi\n\t"
521 		"mov %%rsi, %%cr4\n\t"
522 		::"m"(tmp)
523 		:"rsi", "memory", "cc");
524 	report("Write shadowing different X86_CR4_DE", vmx_get_test_stage() == 12);
525 }
526 
527 static int cr_shadowing_exit_handler(void)
528 {
529 	u64 guest_rip;
530 	ulong reason;
531 	u32 insn_len;
532 	u32 exit_qual;
533 
534 	guest_rip = vmcs_read(GUEST_RIP);
535 	reason = vmcs_read(EXI_REASON) & 0xff;
536 	insn_len = vmcs_read(EXI_INST_LEN);
537 	exit_qual = vmcs_read(EXI_QUALIFICATION);
538 	switch (reason) {
539 	case VMX_VMCALL:
540 		switch (vmx_get_test_stage()) {
541 		case 0:
542 			report("Read through CR0", guest_cr0 == vmcs_read(GUEST_CR0));
543 			break;
544 		case 1:
545 			report("Read through CR4", guest_cr4 == vmcs_read(GUEST_CR4));
546 			break;
547 		case 2:
548 			report("Write through CR0", guest_cr0 == vmcs_read(GUEST_CR0));
549 			break;
550 		case 3:
551 			report("Write through CR4", guest_cr4 == vmcs_read(GUEST_CR4));
552 			break;
553 		case 4:
554 			guest_cr0 = vmcs_read(GUEST_CR0) ^ (X86_CR0_TS | X86_CR0_MP);
555 			guest_cr4 = vmcs_read(GUEST_CR4) ^ (X86_CR4_TSD | X86_CR4_DE);
556 			vmcs_write(CR0_MASK, X86_CR0_TS | X86_CR0_MP);
557 			vmcs_write(CR0_READ_SHADOW, guest_cr0 & (X86_CR0_TS | X86_CR0_MP));
558 			vmcs_write(CR4_MASK, X86_CR4_TSD | X86_CR4_DE);
559 			vmcs_write(CR4_READ_SHADOW, guest_cr4 & (X86_CR4_TSD | X86_CR4_DE));
560 			break;
561 		case 6:
562 			report("Write shadowing CR0 (same value)",
563 					guest_cr0 == (vmcs_read(GUEST_CR0) ^ (X86_CR0_TS | X86_CR0_MP)));
564 			break;
565 		case 7:
566 			report("Write shadowing CR4 (same value)",
567 					guest_cr4 == (vmcs_read(GUEST_CR4) ^ (X86_CR4_TSD | X86_CR4_DE)));
568 			break;
569 		default:
570 			// Should not reach here
571 			report("unexpected stage, %d", false,
572 			       vmx_get_test_stage());
573 			print_vmexit_info();
574 			return VMX_TEST_VMEXIT;
575 		}
576 		vmcs_write(GUEST_RIP, guest_rip + insn_len);
577 		return VMX_TEST_RESUME;
578 	case VMX_CR:
579 		switch (vmx_get_test_stage()) {
580 		case 4:
581 			report("Read shadowing CR0", 0);
582 			vmx_inc_test_stage();
583 			break;
584 		case 5:
585 			report("Read shadowing CR4", 0);
586 			vmx_inc_test_stage();
587 			break;
588 		case 6:
589 			report("Write shadowing CR0 (same value)", 0);
590 			vmx_inc_test_stage();
591 			break;
592 		case 7:
593 			report("Write shadowing CR4 (same value)", 0);
594 			vmx_inc_test_stage();
595 			break;
596 		case 8:
597 		case 9:
598 			// 0x600 encodes "mov %esi, %cr0"
599 			if (exit_qual == 0x600)
600 				vmx_inc_test_stage();
601 			break;
602 		case 10:
603 		case 11:
604 			// 0x604 encodes "mov %esi, %cr4"
605 			if (exit_qual == 0x604)
606 				vmx_inc_test_stage();
607 			break;
608 		default:
609 			// Should not reach here
610 			report("unexpected stage, %d", false,
611 			       vmx_get_test_stage());
612 			print_vmexit_info();
613 			return VMX_TEST_VMEXIT;
614 		}
615 		vmcs_write(GUEST_RIP, guest_rip + insn_len);
616 		return VMX_TEST_RESUME;
617 	default:
618 		report("Unknown exit reason, %ld", false, reason);
619 		print_vmexit_info();
620 	}
621 	return VMX_TEST_VMEXIT;
622 }
623 
624 static int iobmp_init(struct vmcs *vmcs)
625 {
626 	u32 ctrl_cpu0;
627 
628 	io_bitmap_a = alloc_page();
629 	io_bitmap_b = alloc_page();
630 	memset(io_bitmap_a, 0x0, PAGE_SIZE);
631 	memset(io_bitmap_b, 0x0, PAGE_SIZE);
632 	ctrl_cpu0 = vmcs_read(CPU_EXEC_CTRL0);
633 	ctrl_cpu0 |= CPU_IO_BITMAP;
634 	ctrl_cpu0 &= (~CPU_IO);
635 	vmcs_write(CPU_EXEC_CTRL0, ctrl_cpu0);
636 	vmcs_write(IO_BITMAP_A, (u64)io_bitmap_a);
637 	vmcs_write(IO_BITMAP_B, (u64)io_bitmap_b);
638 	return VMX_TEST_START;
639 }
640 
641 static void iobmp_main(void)
642 {
643 	// stage 0, test IO pass
644 	vmx_set_test_stage(0);
645 	inb(0x5000);
646 	outb(0x0, 0x5000);
647 	report("I/O bitmap - I/O pass", vmx_get_test_stage() == 0);
648 	// test IO width, in/out
649 	((u8 *)io_bitmap_a)[0] = 0xFF;
650 	vmx_set_test_stage(2);
651 	inb(0x0);
652 	report("I/O bitmap - trap in", vmx_get_test_stage() == 3);
653 	vmx_set_test_stage(3);
654 	outw(0x0, 0x0);
655 	report("I/O bitmap - trap out", vmx_get_test_stage() == 4);
656 	vmx_set_test_stage(4);
657 	inl(0x0);
658 	report("I/O bitmap - I/O width, long", vmx_get_test_stage() == 5);
659 	// test low/high IO port
660 	vmx_set_test_stage(5);
661 	((u8 *)io_bitmap_a)[0x5000 / 8] = (1 << (0x5000 % 8));
662 	inb(0x5000);
663 	report("I/O bitmap - I/O port, low part", vmx_get_test_stage() == 6);
664 	vmx_set_test_stage(6);
665 	((u8 *)io_bitmap_b)[0x1000 / 8] = (1 << (0x1000 % 8));
666 	inb(0x9000);
667 	report("I/O bitmap - I/O port, high part", vmx_get_test_stage() == 7);
668 	// test partial pass
669 	vmx_set_test_stage(7);
670 	inl(0x4FFF);
671 	report("I/O bitmap - partial pass", vmx_get_test_stage() == 8);
672 	// test overrun
673 	vmx_set_test_stage(8);
674 	memset(io_bitmap_a, 0x0, PAGE_SIZE);
675 	memset(io_bitmap_b, 0x0, PAGE_SIZE);
676 	inl(0xFFFF);
677 	report("I/O bitmap - overrun", vmx_get_test_stage() == 9);
678 	vmx_set_test_stage(9);
679 	vmcall();
680 	outb(0x0, 0x0);
681 	report("I/O bitmap - ignore unconditional exiting",
682 	       vmx_get_test_stage() == 9);
683 	vmx_set_test_stage(10);
684 	vmcall();
685 	outb(0x0, 0x0);
686 	report("I/O bitmap - unconditional exiting",
687 	       vmx_get_test_stage() == 11);
688 }
689 
690 static int iobmp_exit_handler(void)
691 {
692 	u64 guest_rip;
693 	ulong reason, exit_qual;
694 	u32 insn_len, ctrl_cpu0;
695 
696 	guest_rip = vmcs_read(GUEST_RIP);
697 	reason = vmcs_read(EXI_REASON) & 0xff;
698 	exit_qual = vmcs_read(EXI_QUALIFICATION);
699 	insn_len = vmcs_read(EXI_INST_LEN);
700 	switch (reason) {
701 	case VMX_IO:
702 		switch (vmx_get_test_stage()) {
703 		case 0:
704 		case 1:
705 			vmx_inc_test_stage();
706 			break;
707 		case 2:
708 			report("I/O bitmap - I/O width, byte",
709 					(exit_qual & VMX_IO_SIZE_MASK) == _VMX_IO_BYTE);
710 			report("I/O bitmap - I/O direction, in", exit_qual & VMX_IO_IN);
711 			vmx_inc_test_stage();
712 			break;
713 		case 3:
714 			report("I/O bitmap - I/O width, word",
715 					(exit_qual & VMX_IO_SIZE_MASK) == _VMX_IO_WORD);
716 			report("I/O bitmap - I/O direction, out",
717 					!(exit_qual & VMX_IO_IN));
718 			vmx_inc_test_stage();
719 			break;
720 		case 4:
721 			report("I/O bitmap - I/O width, long",
722 					(exit_qual & VMX_IO_SIZE_MASK) == _VMX_IO_LONG);
723 			vmx_inc_test_stage();
724 			break;
725 		case 5:
726 			if (((exit_qual & VMX_IO_PORT_MASK) >> VMX_IO_PORT_SHIFT) == 0x5000)
727 				vmx_inc_test_stage();
728 			break;
729 		case 6:
730 			if (((exit_qual & VMX_IO_PORT_MASK) >> VMX_IO_PORT_SHIFT) == 0x9000)
731 				vmx_inc_test_stage();
732 			break;
733 		case 7:
734 			if (((exit_qual & VMX_IO_PORT_MASK) >> VMX_IO_PORT_SHIFT) == 0x4FFF)
735 				vmx_inc_test_stage();
736 			break;
737 		case 8:
738 			if (((exit_qual & VMX_IO_PORT_MASK) >> VMX_IO_PORT_SHIFT) == 0xFFFF)
739 				vmx_inc_test_stage();
740 			break;
741 		case 9:
742 		case 10:
743 			ctrl_cpu0 = vmcs_read(CPU_EXEC_CTRL0);
744 			vmcs_write(CPU_EXEC_CTRL0, ctrl_cpu0 & ~CPU_IO);
745 			vmx_inc_test_stage();
746 			break;
747 		default:
748 			// Should not reach here
749 			report("unexpected stage, %d", false,
750 			       vmx_get_test_stage());
751 			print_vmexit_info();
752 			return VMX_TEST_VMEXIT;
753 		}
754 		vmcs_write(GUEST_RIP, guest_rip + insn_len);
755 		return VMX_TEST_RESUME;
756 	case VMX_VMCALL:
757 		switch (vmx_get_test_stage()) {
758 		case 9:
759 			ctrl_cpu0 = vmcs_read(CPU_EXEC_CTRL0);
760 			ctrl_cpu0 |= CPU_IO | CPU_IO_BITMAP;
761 			vmcs_write(CPU_EXEC_CTRL0, ctrl_cpu0);
762 			break;
763 		case 10:
764 			ctrl_cpu0 = vmcs_read(CPU_EXEC_CTRL0);
765 			ctrl_cpu0 = (ctrl_cpu0 & ~CPU_IO_BITMAP) | CPU_IO;
766 			vmcs_write(CPU_EXEC_CTRL0, ctrl_cpu0);
767 			break;
768 		default:
769 			// Should not reach here
770 			report("unexpected stage, %d", false,
771 			       vmx_get_test_stage());
772 			print_vmexit_info();
773 			return VMX_TEST_VMEXIT;
774 		}
775 		vmcs_write(GUEST_RIP, guest_rip + insn_len);
776 		return VMX_TEST_RESUME;
777 	default:
778 		printf("guest_rip = %#lx\n", guest_rip);
779 		printf("\tERROR : Undefined exit reason, reason = %ld.\n", reason);
780 		break;
781 	}
782 	return VMX_TEST_VMEXIT;
783 }
784 
785 #define INSN_CPU0		0
786 #define INSN_CPU1		1
787 #define INSN_ALWAYS_TRAP	2
788 
789 #define FIELD_EXIT_QUAL		(1 << 0)
790 #define FIELD_INSN_INFO		(1 << 1)
791 
792 asm(
793 	"insn_hlt: hlt;ret\n\t"
794 	"insn_invlpg: invlpg 0x12345678;ret\n\t"
795 	"insn_mwait: xor %eax, %eax; xor %ecx, %ecx; mwait;ret\n\t"
796 	"insn_rdpmc: xor %ecx, %ecx; rdpmc;ret\n\t"
797 	"insn_rdtsc: rdtsc;ret\n\t"
798 	"insn_cr3_load: mov cr3,%rax; mov %rax,%cr3;ret\n\t"
799 	"insn_cr3_store: mov %cr3,%rax;ret\n\t"
800 #ifdef __x86_64__
801 	"insn_cr8_load: xor %eax, %eax; mov %rax,%cr8;ret\n\t"
802 	"insn_cr8_store: mov %cr8,%rax;ret\n\t"
803 #endif
804 	"insn_monitor: xor %eax, %eax; xor %ecx, %ecx; xor %edx, %edx; monitor;ret\n\t"
805 	"insn_pause: pause;ret\n\t"
806 	"insn_wbinvd: wbinvd;ret\n\t"
807 	"insn_cpuid: mov $10, %eax; cpuid;ret\n\t"
808 	"insn_invd: invd;ret\n\t"
809 	"insn_sgdt: sgdt gdt64_desc;ret\n\t"
810 	"insn_lgdt: lgdt gdt64_desc;ret\n\t"
811 	"insn_sidt: sidt idt_descr;ret\n\t"
812 	"insn_lidt: lidt idt_descr;ret\n\t"
813 	"insn_sldt: sldt %ax;ret\n\t"
814 	"insn_lldt: xor %eax, %eax; lldt %ax;ret\n\t"
815 	"insn_str: str %ax;ret\n\t"
816 	"insn_rdrand: rdrand %rax;ret\n\t"
817 	"insn_rdseed: rdseed %rax;ret\n\t"
818 );
819 extern void insn_hlt(void);
820 extern void insn_invlpg(void);
821 extern void insn_mwait(void);
822 extern void insn_rdpmc(void);
823 extern void insn_rdtsc(void);
824 extern void insn_cr3_load(void);
825 extern void insn_cr3_store(void);
826 #ifdef __x86_64__
827 extern void insn_cr8_load(void);
828 extern void insn_cr8_store(void);
829 #endif
830 extern void insn_monitor(void);
831 extern void insn_pause(void);
832 extern void insn_wbinvd(void);
833 extern void insn_sgdt(void);
834 extern void insn_lgdt(void);
835 extern void insn_sidt(void);
836 extern void insn_lidt(void);
837 extern void insn_sldt(void);
838 extern void insn_lldt(void);
839 extern void insn_str(void);
840 extern void insn_cpuid(void);
841 extern void insn_invd(void);
842 extern void insn_rdrand(void);
843 extern void insn_rdseed(void);
844 
845 u32 cur_insn;
846 u64 cr3;
847 
848 struct insn_table {
849 	const char *name;
850 	u32 flag;
851 	void (*insn_func)(void);
852 	u32 type;
853 	u32 reason;
854 	ulong exit_qual;
855 	u32 insn_info;
856 	// Use FIELD_EXIT_QUAL and FIELD_INSN_INFO to define
857 	// which field need to be tested, reason is always tested
858 	u32 test_field;
859 };
860 
861 /*
862  * Add more test cases of instruction intercept here. Elements in this
863  * table is:
864  *	name/control flag/insn function/type/exit reason/exit qulification/
865  *	instruction info/field to test
866  * The last field defines which fields (exit_qual and insn_info) need to be
867  * tested in exit handler. If set to 0, only "reason" is checked.
868  */
869 static struct insn_table insn_table[] = {
870 	// Flags for Primary Processor-Based VM-Execution Controls
871 	{"HLT",  CPU_HLT, insn_hlt, INSN_CPU0, 12, 0, 0, 0},
872 	{"INVLPG", CPU_INVLPG, insn_invlpg, INSN_CPU0, 14,
873 		0x12345678, 0, FIELD_EXIT_QUAL},
874 	{"MWAIT", CPU_MWAIT, insn_mwait, INSN_CPU0, 36, 0, 0, 0},
875 	{"RDPMC", CPU_RDPMC, insn_rdpmc, INSN_CPU0, 15, 0, 0, 0},
876 	{"RDTSC", CPU_RDTSC, insn_rdtsc, INSN_CPU0, 16, 0, 0, 0},
877 	{"CR3 load", CPU_CR3_LOAD, insn_cr3_load, INSN_CPU0, 28, 0x3, 0,
878 		FIELD_EXIT_QUAL},
879 	{"CR3 store", CPU_CR3_STORE, insn_cr3_store, INSN_CPU0, 28, 0x13, 0,
880 		FIELD_EXIT_QUAL},
881 #ifdef __x86_64__
882 	{"CR8 load", CPU_CR8_LOAD, insn_cr8_load, INSN_CPU0, 28, 0x8, 0,
883 		FIELD_EXIT_QUAL},
884 	{"CR8 store", CPU_CR8_STORE, insn_cr8_store, INSN_CPU0, 28, 0x18, 0,
885 		FIELD_EXIT_QUAL},
886 #endif
887 	{"MONITOR", CPU_MONITOR, insn_monitor, INSN_CPU0, 39, 0, 0, 0},
888 	{"PAUSE", CPU_PAUSE, insn_pause, INSN_CPU0, 40, 0, 0, 0},
889 	// Flags for Secondary Processor-Based VM-Execution Controls
890 	{"WBINVD", CPU_WBINVD, insn_wbinvd, INSN_CPU1, 54, 0, 0, 0},
891 	{"DESC_TABLE (SGDT)", CPU_DESC_TABLE, insn_sgdt, INSN_CPU1, 46, 0, 0, 0},
892 	{"DESC_TABLE (LGDT)", CPU_DESC_TABLE, insn_lgdt, INSN_CPU1, 46, 0, 0, 0},
893 	{"DESC_TABLE (SIDT)", CPU_DESC_TABLE, insn_sidt, INSN_CPU1, 46, 0, 0, 0},
894 	{"DESC_TABLE (LIDT)", CPU_DESC_TABLE, insn_lidt, INSN_CPU1, 46, 0, 0, 0},
895 	{"DESC_TABLE (SLDT)", CPU_DESC_TABLE, insn_sldt, INSN_CPU1, 47, 0, 0, 0},
896 	{"DESC_TABLE (LLDT)", CPU_DESC_TABLE, insn_lldt, INSN_CPU1, 47, 0, 0, 0},
897 	{"DESC_TABLE (STR)", CPU_DESC_TABLE, insn_str, INSN_CPU1, 47, 0, 0, 0},
898 	/* LTR causes a #GP if done with a busy selector, so it is not tested.  */
899 	{"RDRAND", CPU_RDRAND, insn_rdrand, INSN_CPU1, VMX_RDRAND, 0, 0, 0},
900 	{"RDSEED", CPU_RDSEED, insn_rdseed, INSN_CPU1, VMX_RDSEED, 0, 0, 0},
901 	// Instructions always trap
902 	{"CPUID", 0, insn_cpuid, INSN_ALWAYS_TRAP, 10, 0, 0, 0},
903 	{"INVD", 0, insn_invd, INSN_ALWAYS_TRAP, 13, 0, 0, 0},
904 	// Instructions never trap
905 	{NULL},
906 };
907 
908 static int insn_intercept_init(struct vmcs *vmcs)
909 {
910 	u32 ctrl_cpu;
911 
912 	ctrl_cpu = ctrl_cpu_rev[0].set | CPU_SECONDARY;
913 	ctrl_cpu &= ctrl_cpu_rev[0].clr;
914 	vmcs_write(CPU_EXEC_CTRL0, ctrl_cpu);
915 	vmcs_write(CPU_EXEC_CTRL1, ctrl_cpu_rev[1].set);
916 	cr3 = read_cr3();
917 	return VMX_TEST_START;
918 }
919 
920 static void insn_intercept_main(void)
921 {
922 	for (cur_insn = 0; insn_table[cur_insn].name != NULL; cur_insn++) {
923 		vmx_set_test_stage(cur_insn * 2);
924 		if ((insn_table[cur_insn].type == INSN_CPU0 &&
925 		     !(ctrl_cpu_rev[0].clr & insn_table[cur_insn].flag)) ||
926 		    (insn_table[cur_insn].type == INSN_CPU1 &&
927 		     !(ctrl_cpu_rev[1].clr & insn_table[cur_insn].flag))) {
928 			printf("\tCPU_CTRL%d.CPU_%s is not supported.\n",
929 			       insn_table[cur_insn].type - INSN_CPU0,
930 			       insn_table[cur_insn].name);
931 			continue;
932 		}
933 
934 		if ((insn_table[cur_insn].type == INSN_CPU0 &&
935 		     !(ctrl_cpu_rev[0].set & insn_table[cur_insn].flag)) ||
936 		    (insn_table[cur_insn].type == INSN_CPU1 &&
937 		     !(ctrl_cpu_rev[1].set & insn_table[cur_insn].flag))) {
938 			/* skip hlt, it stalls the guest and is tested below */
939 			if (insn_table[cur_insn].insn_func != insn_hlt)
940 				insn_table[cur_insn].insn_func();
941 			report("execute %s", vmx_get_test_stage() == cur_insn * 2,
942 					insn_table[cur_insn].name);
943 		} else if (insn_table[cur_insn].type != INSN_ALWAYS_TRAP)
944 			printf("\tCPU_CTRL%d.CPU_%s always traps.\n",
945 			       insn_table[cur_insn].type - INSN_CPU0,
946 			       insn_table[cur_insn].name);
947 
948 		vmcall();
949 
950 		insn_table[cur_insn].insn_func();
951 		report("intercept %s", vmx_get_test_stage() == cur_insn * 2 + 1,
952 				insn_table[cur_insn].name);
953 
954 		vmx_set_test_stage(cur_insn * 2 + 1);
955 		vmcall();
956 	}
957 }
958 
959 static int insn_intercept_exit_handler(void)
960 {
961 	u64 guest_rip;
962 	u32 reason;
963 	ulong exit_qual;
964 	u32 insn_len;
965 	u32 insn_info;
966 	bool pass;
967 
968 	guest_rip = vmcs_read(GUEST_RIP);
969 	reason = vmcs_read(EXI_REASON) & 0xff;
970 	exit_qual = vmcs_read(EXI_QUALIFICATION);
971 	insn_len = vmcs_read(EXI_INST_LEN);
972 	insn_info = vmcs_read(EXI_INST_INFO);
973 
974 	if (reason == VMX_VMCALL) {
975 		u32 val = 0;
976 
977 		if (insn_table[cur_insn].type == INSN_CPU0)
978 			val = vmcs_read(CPU_EXEC_CTRL0);
979 		else if (insn_table[cur_insn].type == INSN_CPU1)
980 			val = vmcs_read(CPU_EXEC_CTRL1);
981 
982 		if (vmx_get_test_stage() & 1)
983 			val &= ~insn_table[cur_insn].flag;
984 		else
985 			val |= insn_table[cur_insn].flag;
986 
987 		if (insn_table[cur_insn].type == INSN_CPU0)
988 			vmcs_write(CPU_EXEC_CTRL0, val | ctrl_cpu_rev[0].set);
989 		else if (insn_table[cur_insn].type == INSN_CPU1)
990 			vmcs_write(CPU_EXEC_CTRL1, val | ctrl_cpu_rev[1].set);
991 	} else {
992 		pass = (cur_insn * 2 == vmx_get_test_stage()) &&
993 			insn_table[cur_insn].reason == reason;
994 		if (insn_table[cur_insn].test_field & FIELD_EXIT_QUAL &&
995 		    insn_table[cur_insn].exit_qual != exit_qual)
996 			pass = false;
997 		if (insn_table[cur_insn].test_field & FIELD_INSN_INFO &&
998 		    insn_table[cur_insn].insn_info != insn_info)
999 			pass = false;
1000 		if (pass)
1001 			vmx_inc_test_stage();
1002 	}
1003 	vmcs_write(GUEST_RIP, guest_rip + insn_len);
1004 	return VMX_TEST_RESUME;
1005 }
1006 
1007 /**
1008  * __setup_ept - Setup the VMCS fields to enable Extended Page Tables (EPT)
1009  * @hpa:	Host physical address of the top-level, a.k.a. root, EPT table
1010  * @enable_ad:	Whether or not to enable Access/Dirty bits for EPT entries
1011  *
1012  * Returns 0 on success, 1 on failure.
1013  *
1014  * Note that @hpa doesn't need to point at actual memory if VM-Launch is
1015  * expected to fail, e.g. setup_dummy_ept() arbitrarily passes '0' to satisfy
1016  * the various EPTP consistency checks, but doesn't ensure backing for HPA '0'.
1017  */
1018 static int __setup_ept(u64 hpa, bool enable_ad)
1019 {
1020 	if (!(ctrl_cpu_rev[0].clr & CPU_SECONDARY) ||
1021 	    !(ctrl_cpu_rev[1].clr & CPU_EPT)) {
1022 		printf("\tEPT is not supported");
1023 		return 1;
1024 	}
1025 	if (!(ept_vpid.val & EPT_CAP_WB)) {
1026 		printf("WB memtype for EPT walks not supported\n");
1027 		return 1;
1028 	}
1029 	if (!(ept_vpid.val & EPT_CAP_PWL4)) {
1030 		printf("\tPWL4 is not supported\n");
1031 		return 1;
1032 	}
1033 
1034 	eptp = EPT_MEM_TYPE_WB;
1035 	eptp |= (3 << EPTP_PG_WALK_LEN_SHIFT);
1036 	eptp |= hpa;
1037 	if (enable_ad)
1038 		eptp |= EPTP_AD_FLAG;
1039 
1040 	vmcs_write(EPTP, eptp);
1041 	vmcs_write(CPU_EXEC_CTRL0, vmcs_read(CPU_EXEC_CTRL0)| CPU_SECONDARY);
1042 	vmcs_write(CPU_EXEC_CTRL1, vmcs_read(CPU_EXEC_CTRL1)| CPU_EPT);
1043 
1044 	return 0;
1045 }
1046 
1047 /**
1048  * setup_ept - Enable Extended Page Tables (EPT) and setup an identity map
1049  * @enable_ad:	Whether or not to enable Access/Dirty bits for EPT entries
1050  *
1051  * Returns 0 on success, 1 on failure.
1052  *
1053  * This is the "real" function for setting up EPT tables, i.e. use this for
1054  * tests that need to run code in the guest with EPT enabled.
1055  */
1056 static int setup_ept(bool enable_ad)
1057 {
1058 	unsigned long end_of_memory;
1059 
1060 	pml4 = alloc_page();
1061 
1062 	if (__setup_ept(virt_to_phys(pml4), enable_ad))
1063 		return 1;
1064 
1065 	memset(pml4, 0, PAGE_SIZE);
1066 
1067 	end_of_memory = fwcfg_get_u64(FW_CFG_RAM_SIZE);
1068 	if (end_of_memory < (1ul << 32))
1069 		end_of_memory = (1ul << 32);
1070 	/* Cannot use large EPT pages if we need to track EPT
1071 	 * accessed/dirty bits at 4K granularity.
1072 	 */
1073 	setup_ept_range(pml4, 0, end_of_memory, 0,
1074 			!enable_ad && ept_2m_supported(),
1075 			EPT_WA | EPT_RA | EPT_EA);
1076 	return 0;
1077 }
1078 
1079 /**
1080  * setup_dummy_ept - Enable Extended Page Tables (EPT) with a dummy root HPA
1081  *
1082  * Setup EPT using a semi-arbitrary dummy root HPA.  This function is intended
1083  * for use by tests that need EPT enabled to verify dependent VMCS controls
1084  * but never expect to fully enter the guest, i.e. don't need setup the actual
1085  * EPT tables.
1086  */
1087 static void setup_dummy_ept(void)
1088 {
1089 	if (__setup_ept(0, false))
1090 		report_abort("EPT setup unexpectedly failed");
1091 }
1092 
1093 static int enable_unrestricted_guest(void)
1094 {
1095 	if (!(ctrl_cpu_rev[0].clr & CPU_SECONDARY) ||
1096 	    !(ctrl_cpu_rev[1].clr & CPU_URG) ||
1097 	    !(ctrl_cpu_rev[1].clr & CPU_EPT))
1098 		return 1;
1099 
1100 	setup_dummy_ept();
1101 
1102 	vmcs_write(CPU_EXEC_CTRL0, vmcs_read(CPU_EXEC_CTRL0) | CPU_SECONDARY);
1103 	vmcs_write(CPU_EXEC_CTRL1, vmcs_read(CPU_EXEC_CTRL1) | CPU_URG);
1104 
1105 	return 0;
1106 }
1107 
1108 static void ept_enable_ad_bits(void)
1109 {
1110 	eptp |= EPTP_AD_FLAG;
1111 	vmcs_write(EPTP, eptp);
1112 }
1113 
1114 static void ept_disable_ad_bits(void)
1115 {
1116 	eptp &= ~EPTP_AD_FLAG;
1117 	vmcs_write(EPTP, eptp);
1118 }
1119 
1120 static void ept_enable_ad_bits_or_skip_test(void)
1121 {
1122 	if (!ept_ad_bits_supported())
1123 		test_skip("EPT AD bits not supported.");
1124 	ept_enable_ad_bits();
1125 }
1126 
1127 static int apic_version;
1128 
1129 static int ept_init_common(bool have_ad)
1130 {
1131 	int ret;
1132 	struct pci_dev pcidev;
1133 
1134 	if (setup_ept(have_ad))
1135 		return VMX_TEST_EXIT;
1136 	data_page1 = alloc_page();
1137 	data_page2 = alloc_page();
1138 	memset(data_page1, 0x0, PAGE_SIZE);
1139 	memset(data_page2, 0x0, PAGE_SIZE);
1140 	*((u32 *)data_page1) = MAGIC_VAL_1;
1141 	*((u32 *)data_page2) = MAGIC_VAL_2;
1142 	install_ept(pml4, (unsigned long)data_page1, (unsigned long)data_page2,
1143 			EPT_RA | EPT_WA | EPT_EA);
1144 
1145 	apic_version = apic_read(APIC_LVR);
1146 
1147 	ret = pci_find_dev(PCI_VENDOR_ID_REDHAT, PCI_DEVICE_ID_REDHAT_TEST);
1148 	if (ret != PCIDEVADDR_INVALID) {
1149 		pci_dev_init(&pcidev, ret);
1150 		pci_physaddr = pcidev.resource[PCI_TESTDEV_BAR_MEM];
1151 	}
1152 
1153 	return VMX_TEST_START;
1154 }
1155 
1156 static int ept_init(struct vmcs *vmcs)
1157 {
1158 	return ept_init_common(false);
1159 }
1160 
1161 static void ept_common(void)
1162 {
1163 	vmx_set_test_stage(0);
1164 	if (*((u32 *)data_page2) != MAGIC_VAL_1 ||
1165 			*((u32 *)data_page1) != MAGIC_VAL_1)
1166 		report("EPT basic framework - read", 0);
1167 	else {
1168 		*((u32 *)data_page2) = MAGIC_VAL_3;
1169 		vmcall();
1170 		if (vmx_get_test_stage() == 1) {
1171 			if (*((u32 *)data_page1) == MAGIC_VAL_3 &&
1172 					*((u32 *)data_page2) == MAGIC_VAL_2)
1173 				report("EPT basic framework", 1);
1174 			else
1175 				report("EPT basic framework - remap", 1);
1176 		}
1177 	}
1178 	// Test EPT Misconfigurations
1179 	vmx_set_test_stage(1);
1180 	vmcall();
1181 	*((u32 *)data_page1) = MAGIC_VAL_1;
1182 	if (vmx_get_test_stage() != 2) {
1183 		report("EPT misconfigurations", 0);
1184 		goto t1;
1185 	}
1186 	vmx_set_test_stage(2);
1187 	vmcall();
1188 	*((u32 *)data_page1) = MAGIC_VAL_1;
1189 	report("EPT misconfigurations", vmx_get_test_stage() == 3);
1190 t1:
1191 	// Test EPT violation
1192 	vmx_set_test_stage(3);
1193 	vmcall();
1194 	*((u32 *)data_page1) = MAGIC_VAL_1;
1195 	report("EPT violation - page permission", vmx_get_test_stage() == 4);
1196 	// Violation caused by EPT paging structure
1197 	vmx_set_test_stage(4);
1198 	vmcall();
1199 	*((u32 *)data_page1) = MAGIC_VAL_2;
1200 	report("EPT violation - paging structure", vmx_get_test_stage() == 5);
1201 
1202 	// MMIO Read/Write
1203 	vmx_set_test_stage(5);
1204 	vmcall();
1205 
1206 	*(u32 volatile *)pci_physaddr;
1207 	report("MMIO EPT violation - read", vmx_get_test_stage() == 6);
1208 
1209 	*(u32 volatile *)pci_physaddr = MAGIC_VAL_1;
1210 	report("MMIO EPT violation - write", vmx_get_test_stage() == 7);
1211 }
1212 
1213 static void ept_main(void)
1214 {
1215 	ept_common();
1216 
1217 	// Test EPT access to L1 MMIO
1218 	vmx_set_test_stage(7);
1219 	report("EPT - MMIO access", *((u32 *)0xfee00030UL) == apic_version);
1220 
1221 	// Test invalid operand for INVEPT
1222 	vmcall();
1223 	report("EPT - unsupported INVEPT", vmx_get_test_stage() == 8);
1224 }
1225 
1226 static bool invept_test(int type, u64 eptp)
1227 {
1228 	bool ret, supported;
1229 
1230 	supported = ept_vpid.val & (EPT_CAP_INVEPT_SINGLE >> INVEPT_SINGLE << type);
1231 	ret = invept(type, eptp);
1232 
1233 	if (ret == !supported)
1234 		return false;
1235 
1236 	if (!supported)
1237 		printf("WARNING: unsupported invept passed!\n");
1238 	else
1239 		printf("WARNING: invept failed!\n");
1240 
1241 	return true;
1242 }
1243 
1244 static int pml_exit_handler(void)
1245 {
1246 	u16 index, count;
1247 	ulong reason = vmcs_read(EXI_REASON) & 0xff;
1248 	u64 *pmlbuf = pml_log;
1249 	u64 guest_rip = vmcs_read(GUEST_RIP);;
1250 	u64 guest_cr3 = vmcs_read(GUEST_CR3);
1251 	u32 insn_len = vmcs_read(EXI_INST_LEN);
1252 
1253 	switch (reason) {
1254 	case VMX_VMCALL:
1255 		switch (vmx_get_test_stage()) {
1256 		case 0:
1257 			index = vmcs_read(GUEST_PML_INDEX);
1258 			for (count = index + 1; count < PML_INDEX; count++) {
1259 				if (pmlbuf[count] == (u64)data_page2) {
1260 					vmx_inc_test_stage();
1261 					clear_ept_ad(pml4, guest_cr3, (unsigned long)data_page2);
1262 					break;
1263 				}
1264 			}
1265 			break;
1266 		case 1:
1267 			index = vmcs_read(GUEST_PML_INDEX);
1268 			/* Keep clearing the dirty bit till a overflow */
1269 			clear_ept_ad(pml4, guest_cr3, (unsigned long)data_page2);
1270 			break;
1271 		default:
1272 			report("unexpected stage, %d.", false,
1273 			       vmx_get_test_stage());
1274 			print_vmexit_info();
1275 			return VMX_TEST_VMEXIT;
1276 		}
1277 		vmcs_write(GUEST_RIP, guest_rip + insn_len);
1278 		return VMX_TEST_RESUME;
1279 	case VMX_PML_FULL:
1280 		vmx_inc_test_stage();
1281 		vmcs_write(GUEST_PML_INDEX, PML_INDEX - 1);
1282 		return VMX_TEST_RESUME;
1283 	default:
1284 		report("Unknown exit reason, %ld", false, reason);
1285 		print_vmexit_info();
1286 	}
1287 	return VMX_TEST_VMEXIT;
1288 }
1289 
1290 static int ept_exit_handler_common(bool have_ad)
1291 {
1292 	u64 guest_rip;
1293 	u64 guest_cr3;
1294 	ulong reason;
1295 	u32 insn_len;
1296 	u32 exit_qual;
1297 	static unsigned long data_page1_pte, data_page1_pte_pte, memaddr_pte;
1298 
1299 	guest_rip = vmcs_read(GUEST_RIP);
1300 	guest_cr3 = vmcs_read(GUEST_CR3);
1301 	reason = vmcs_read(EXI_REASON) & 0xff;
1302 	insn_len = vmcs_read(EXI_INST_LEN);
1303 	exit_qual = vmcs_read(EXI_QUALIFICATION);
1304 	switch (reason) {
1305 	case VMX_VMCALL:
1306 		switch (vmx_get_test_stage()) {
1307 		case 0:
1308 			check_ept_ad(pml4, guest_cr3,
1309 				     (unsigned long)data_page1,
1310 				     have_ad ? EPT_ACCESS_FLAG : 0,
1311 				     have_ad ? EPT_ACCESS_FLAG | EPT_DIRTY_FLAG : 0);
1312 			check_ept_ad(pml4, guest_cr3,
1313 				     (unsigned long)data_page2,
1314 				     have_ad ? EPT_ACCESS_FLAG | EPT_DIRTY_FLAG : 0,
1315 				     have_ad ? EPT_ACCESS_FLAG | EPT_DIRTY_FLAG : 0);
1316 			clear_ept_ad(pml4, guest_cr3, (unsigned long)data_page1);
1317 			clear_ept_ad(pml4, guest_cr3, (unsigned long)data_page2);
1318 			if (have_ad)
1319 				ept_sync(INVEPT_SINGLE, eptp);;
1320 			if (*((u32 *)data_page1) == MAGIC_VAL_3 &&
1321 					*((u32 *)data_page2) == MAGIC_VAL_2) {
1322 				vmx_inc_test_stage();
1323 				install_ept(pml4, (unsigned long)data_page2,
1324 						(unsigned long)data_page2,
1325 						EPT_RA | EPT_WA | EPT_EA);
1326 			} else
1327 				report("EPT basic framework - write", 0);
1328 			break;
1329 		case 1:
1330 			install_ept(pml4, (unsigned long)data_page1,
1331  				(unsigned long)data_page1, EPT_WA);
1332 			ept_sync(INVEPT_SINGLE, eptp);
1333 			break;
1334 		case 2:
1335 			install_ept(pml4, (unsigned long)data_page1,
1336  				(unsigned long)data_page1,
1337  				EPT_RA | EPT_WA | EPT_EA |
1338  				(2 << EPT_MEM_TYPE_SHIFT));
1339 			ept_sync(INVEPT_SINGLE, eptp);
1340 			break;
1341 		case 3:
1342 			clear_ept_ad(pml4, guest_cr3, (unsigned long)data_page1);
1343 			TEST_ASSERT(get_ept_pte(pml4, (unsigned long)data_page1,
1344 						1, &data_page1_pte));
1345 			set_ept_pte(pml4, (unsigned long)data_page1,
1346 				1, data_page1_pte & ~EPT_PRESENT);
1347 			ept_sync(INVEPT_SINGLE, eptp);
1348 			break;
1349 		case 4:
1350 			TEST_ASSERT(get_ept_pte(pml4, (unsigned long)data_page1,
1351 						2, &data_page1_pte));
1352 			data_page1_pte &= PAGE_MASK;
1353 			TEST_ASSERT(get_ept_pte(pml4, data_page1_pte,
1354 						2, &data_page1_pte_pte));
1355 			set_ept_pte(pml4, data_page1_pte, 2,
1356 				data_page1_pte_pte & ~EPT_PRESENT);
1357 			ept_sync(INVEPT_SINGLE, eptp);
1358 			break;
1359 		case 5:
1360 			install_ept(pml4, (unsigned long)pci_physaddr,
1361 				(unsigned long)pci_physaddr, 0);
1362 			ept_sync(INVEPT_SINGLE, eptp);
1363 			break;
1364 		case 7:
1365 			if (!invept_test(0, eptp))
1366 				vmx_inc_test_stage();
1367 			break;
1368 		// Should not reach here
1369 		default:
1370 			report("ERROR - unexpected stage, %d.", false,
1371 			       vmx_get_test_stage());
1372 			print_vmexit_info();
1373 			return VMX_TEST_VMEXIT;
1374 		}
1375 		vmcs_write(GUEST_RIP, guest_rip + insn_len);
1376 		return VMX_TEST_RESUME;
1377 	case VMX_EPT_MISCONFIG:
1378 		switch (vmx_get_test_stage()) {
1379 		case 1:
1380 		case 2:
1381 			vmx_inc_test_stage();
1382 			install_ept(pml4, (unsigned long)data_page1,
1383  				(unsigned long)data_page1,
1384  				EPT_RA | EPT_WA | EPT_EA);
1385 			ept_sync(INVEPT_SINGLE, eptp);
1386 			break;
1387 		// Should not reach here
1388 		default:
1389 			report("ERROR - unexpected stage, %d.", false,
1390 			       vmx_get_test_stage());
1391 			print_vmexit_info();
1392 			return VMX_TEST_VMEXIT;
1393 		}
1394 		return VMX_TEST_RESUME;
1395 	case VMX_EPT_VIOLATION:
1396 		switch(vmx_get_test_stage()) {
1397 		case 3:
1398 			check_ept_ad(pml4, guest_cr3, (unsigned long)data_page1, 0,
1399 				     have_ad ? EPT_ACCESS_FLAG | EPT_DIRTY_FLAG : 0);
1400 			clear_ept_ad(pml4, guest_cr3, (unsigned long)data_page1);
1401 			if (exit_qual == (EPT_VLT_WR | EPT_VLT_LADDR_VLD |
1402 					EPT_VLT_PADDR))
1403 				vmx_inc_test_stage();
1404 			set_ept_pte(pml4, (unsigned long)data_page1,
1405 				1, data_page1_pte | (EPT_PRESENT));
1406 			ept_sync(INVEPT_SINGLE, eptp);
1407 			break;
1408 		case 4:
1409 			check_ept_ad(pml4, guest_cr3, (unsigned long)data_page1, 0,
1410 				     have_ad ? EPT_ACCESS_FLAG | EPT_DIRTY_FLAG : 0);
1411 			clear_ept_ad(pml4, guest_cr3, (unsigned long)data_page1);
1412 			if (exit_qual == (EPT_VLT_RD |
1413 					  (have_ad ? EPT_VLT_WR : 0) |
1414 					  EPT_VLT_LADDR_VLD))
1415 				vmx_inc_test_stage();
1416 			set_ept_pte(pml4, data_page1_pte, 2,
1417 				data_page1_pte_pte | (EPT_PRESENT));
1418 			ept_sync(INVEPT_SINGLE, eptp);
1419 			break;
1420 		case 5:
1421 			if (exit_qual & EPT_VLT_RD)
1422 				vmx_inc_test_stage();
1423 			TEST_ASSERT(get_ept_pte(pml4, (unsigned long)pci_physaddr,
1424 						1, &memaddr_pte));
1425 			set_ept_pte(pml4, memaddr_pte, 1, memaddr_pte | EPT_RA);
1426 			ept_sync(INVEPT_SINGLE, eptp);
1427 			break;
1428 		case 6:
1429 			if (exit_qual & EPT_VLT_WR)
1430 				vmx_inc_test_stage();
1431 			TEST_ASSERT(get_ept_pte(pml4, (unsigned long)pci_physaddr,
1432 						1, &memaddr_pte));
1433 			set_ept_pte(pml4, memaddr_pte, 1, memaddr_pte | EPT_RA | EPT_WA);
1434 			ept_sync(INVEPT_SINGLE, eptp);
1435 			break;
1436 		default:
1437 			// Should not reach here
1438 			report("ERROR : unexpected stage, %d", false,
1439 			       vmx_get_test_stage());
1440 			print_vmexit_info();
1441 			return VMX_TEST_VMEXIT;
1442 		}
1443 		return VMX_TEST_RESUME;
1444 	default:
1445 		report("Unknown exit reason, %ld", false, reason);
1446 		print_vmexit_info();
1447 	}
1448 	return VMX_TEST_VMEXIT;
1449 }
1450 
1451 static int ept_exit_handler(void)
1452 {
1453 	return ept_exit_handler_common(false);
1454 }
1455 
1456 static int eptad_init(struct vmcs *vmcs)
1457 {
1458 	int r = ept_init_common(true);
1459 
1460 	if (r == VMX_TEST_EXIT)
1461 		return r;
1462 
1463 	if ((rdmsr(MSR_IA32_VMX_EPT_VPID_CAP) & EPT_CAP_AD_FLAG) == 0) {
1464 		printf("\tEPT A/D bits are not supported");
1465 		return VMX_TEST_EXIT;
1466 	}
1467 
1468 	return r;
1469 }
1470 
1471 static int pml_init(struct vmcs *vmcs)
1472 {
1473 	u32 ctrl_cpu;
1474 	int r = eptad_init(vmcs);
1475 
1476 	if (r == VMX_TEST_EXIT)
1477 		return r;
1478 
1479 	if (!(ctrl_cpu_rev[0].clr & CPU_SECONDARY) ||
1480 		!(ctrl_cpu_rev[1].clr & CPU_PML)) {
1481 		printf("\tPML is not supported");
1482 		return VMX_TEST_EXIT;
1483 	}
1484 
1485 	pml_log = alloc_page();
1486 	memset(pml_log, 0x0, PAGE_SIZE);
1487 	vmcs_write(PMLADDR, (u64)pml_log);
1488 	vmcs_write(GUEST_PML_INDEX, PML_INDEX - 1);
1489 
1490 	ctrl_cpu = vmcs_read(CPU_EXEC_CTRL1) | CPU_PML;
1491 	vmcs_write(CPU_EXEC_CTRL1, ctrl_cpu);
1492 
1493 	return VMX_TEST_START;
1494 }
1495 
1496 static void pml_main(void)
1497 {
1498 	int count = 0;
1499 
1500 	vmx_set_test_stage(0);
1501 	*((u32 *)data_page2) = 0x1;
1502 	vmcall();
1503 	report("PML - Dirty GPA Logging", vmx_get_test_stage() == 1);
1504 
1505 	while (vmx_get_test_stage() == 1) {
1506 		vmcall();
1507 		*((u32 *)data_page2) = 0x1;
1508 		if (count++ > PML_INDEX)
1509 			break;
1510 	}
1511 	report("PML Full Event", vmx_get_test_stage() == 2);
1512 }
1513 
1514 static void eptad_main(void)
1515 {
1516 	ept_common();
1517 }
1518 
1519 static int eptad_exit_handler(void)
1520 {
1521 	return ept_exit_handler_common(true);
1522 }
1523 
1524 static bool invvpid_test(int type, u16 vpid)
1525 {
1526 	bool ret, supported;
1527 
1528 	supported = ept_vpid.val &
1529 		(VPID_CAP_INVVPID_ADDR >> INVVPID_ADDR << type);
1530 	ret = invvpid(type, vpid, 0);
1531 
1532 	if (ret == !supported)
1533 		return false;
1534 
1535 	if (!supported)
1536 		printf("WARNING: unsupported invvpid passed!\n");
1537 	else
1538 		printf("WARNING: invvpid failed!\n");
1539 
1540 	return true;
1541 }
1542 
1543 static int vpid_init(struct vmcs *vmcs)
1544 {
1545 	u32 ctrl_cpu1;
1546 
1547 	if (!(ctrl_cpu_rev[0].clr & CPU_SECONDARY) ||
1548 		!(ctrl_cpu_rev[1].clr & CPU_VPID)) {
1549 		printf("\tVPID is not supported");
1550 		return VMX_TEST_EXIT;
1551 	}
1552 
1553 	ctrl_cpu1 = vmcs_read(CPU_EXEC_CTRL1);
1554 	ctrl_cpu1 |= CPU_VPID;
1555 	vmcs_write(CPU_EXEC_CTRL1, ctrl_cpu1);
1556 	return VMX_TEST_START;
1557 }
1558 
1559 static void vpid_main(void)
1560 {
1561 	vmx_set_test_stage(0);
1562 	vmcall();
1563 	report("INVVPID SINGLE ADDRESS", vmx_get_test_stage() == 1);
1564 	vmx_set_test_stage(2);
1565 	vmcall();
1566 	report("INVVPID SINGLE", vmx_get_test_stage() == 3);
1567 	vmx_set_test_stage(4);
1568 	vmcall();
1569 	report("INVVPID ALL", vmx_get_test_stage() == 5);
1570 }
1571 
1572 static int vpid_exit_handler(void)
1573 {
1574 	u64 guest_rip;
1575 	ulong reason;
1576 	u32 insn_len;
1577 
1578 	guest_rip = vmcs_read(GUEST_RIP);
1579 	reason = vmcs_read(EXI_REASON) & 0xff;
1580 	insn_len = vmcs_read(EXI_INST_LEN);
1581 
1582 	switch (reason) {
1583 	case VMX_VMCALL:
1584 		switch(vmx_get_test_stage()) {
1585 		case 0:
1586 			if (!invvpid_test(INVVPID_ADDR, 1))
1587 				vmx_inc_test_stage();
1588 			break;
1589 		case 2:
1590 			if (!invvpid_test(INVVPID_CONTEXT_GLOBAL, 1))
1591 				vmx_inc_test_stage();
1592 			break;
1593 		case 4:
1594 			if (!invvpid_test(INVVPID_ALL, 1))
1595 				vmx_inc_test_stage();
1596 			break;
1597 		default:
1598 			report("ERROR: unexpected stage, %d", false,
1599 					vmx_get_test_stage());
1600 			print_vmexit_info();
1601 			return VMX_TEST_VMEXIT;
1602 		}
1603 		vmcs_write(GUEST_RIP, guest_rip + insn_len);
1604 		return VMX_TEST_RESUME;
1605 	default:
1606 		report("Unknown exit reason, %ld", false, reason);
1607 		print_vmexit_info();
1608 	}
1609 	return VMX_TEST_VMEXIT;
1610 }
1611 
1612 #define TIMER_VECTOR	222
1613 
1614 static volatile bool timer_fired;
1615 
1616 static void timer_isr(isr_regs_t *regs)
1617 {
1618 	timer_fired = true;
1619 	apic_write(APIC_EOI, 0);
1620 }
1621 
1622 static int interrupt_init(struct vmcs *vmcs)
1623 {
1624 	msr_bmp_init();
1625 	vmcs_write(PIN_CONTROLS, vmcs_read(PIN_CONTROLS) & ~PIN_EXTINT);
1626 	handle_irq(TIMER_VECTOR, timer_isr);
1627 	return VMX_TEST_START;
1628 }
1629 
1630 static void interrupt_main(void)
1631 {
1632 	long long start, loops;
1633 
1634 	vmx_set_test_stage(0);
1635 
1636 	apic_write(APIC_LVTT, TIMER_VECTOR);
1637 	irq_enable();
1638 
1639 	apic_write(APIC_TMICT, 1);
1640 	for (loops = 0; loops < 10000000 && !timer_fired; loops++)
1641 		asm volatile ("nop");
1642 	report("direct interrupt while running guest", timer_fired);
1643 
1644 	apic_write(APIC_TMICT, 0);
1645 	irq_disable();
1646 	vmcall();
1647 	timer_fired = false;
1648 	apic_write(APIC_TMICT, 1);
1649 	for (loops = 0; loops < 10000000 && !timer_fired; loops++)
1650 		asm volatile ("nop");
1651 	report("intercepted interrupt while running guest", timer_fired);
1652 
1653 	irq_enable();
1654 	apic_write(APIC_TMICT, 0);
1655 	irq_disable();
1656 	vmcall();
1657 	timer_fired = false;
1658 	start = rdtsc();
1659 	apic_write(APIC_TMICT, 1000000);
1660 
1661 	asm volatile ("sti; hlt");
1662 
1663 	report("direct interrupt + hlt",
1664 	       rdtsc() - start > 1000000 && timer_fired);
1665 
1666 	apic_write(APIC_TMICT, 0);
1667 	irq_disable();
1668 	vmcall();
1669 	timer_fired = false;
1670 	start = rdtsc();
1671 	apic_write(APIC_TMICT, 1000000);
1672 
1673 	asm volatile ("sti; hlt");
1674 
1675 	report("intercepted interrupt + hlt",
1676 	       rdtsc() - start > 10000 && timer_fired);
1677 
1678 	apic_write(APIC_TMICT, 0);
1679 	irq_disable();
1680 	vmcall();
1681 	timer_fired = false;
1682 	start = rdtsc();
1683 	apic_write(APIC_TMICT, 1000000);
1684 
1685 	irq_enable();
1686 	asm volatile ("nop");
1687 	vmcall();
1688 
1689 	report("direct interrupt + activity state hlt",
1690 	       rdtsc() - start > 10000 && timer_fired);
1691 
1692 	apic_write(APIC_TMICT, 0);
1693 	irq_disable();
1694 	vmcall();
1695 	timer_fired = false;
1696 	start = rdtsc();
1697 	apic_write(APIC_TMICT, 1000000);
1698 
1699 	irq_enable();
1700 	asm volatile ("nop");
1701 	vmcall();
1702 
1703 	report("intercepted interrupt + activity state hlt",
1704 	       rdtsc() - start > 10000 && timer_fired);
1705 
1706 	apic_write(APIC_TMICT, 0);
1707 	irq_disable();
1708 	vmx_set_test_stage(7);
1709 	vmcall();
1710 	timer_fired = false;
1711 	apic_write(APIC_TMICT, 1);
1712 	for (loops = 0; loops < 10000000 && !timer_fired; loops++)
1713 		asm volatile ("nop");
1714 	report("running a guest with interrupt acknowledgement set", timer_fired);
1715 
1716 	apic_write(APIC_TMICT, 0);
1717 	irq_enable();
1718 	timer_fired = false;
1719 	vmcall();
1720 	report("Inject an event to a halted guest", timer_fired);
1721 }
1722 
1723 static int interrupt_exit_handler(void)
1724 {
1725 	u64 guest_rip = vmcs_read(GUEST_RIP);
1726 	ulong reason = vmcs_read(EXI_REASON) & 0xff;
1727 	u32 insn_len = vmcs_read(EXI_INST_LEN);
1728 
1729 	switch (reason) {
1730 	case VMX_VMCALL:
1731 		switch (vmx_get_test_stage()) {
1732 		case 0:
1733 		case 2:
1734 		case 5:
1735 			vmcs_write(PIN_CONTROLS,
1736 				   vmcs_read(PIN_CONTROLS) | PIN_EXTINT);
1737 			break;
1738 		case 7:
1739 			vmcs_write(EXI_CONTROLS, vmcs_read(EXI_CONTROLS) | EXI_INTA);
1740 			vmcs_write(PIN_CONTROLS,
1741 				   vmcs_read(PIN_CONTROLS) | PIN_EXTINT);
1742 			break;
1743 		case 1:
1744 		case 3:
1745 			vmcs_write(PIN_CONTROLS,
1746 				   vmcs_read(PIN_CONTROLS) & ~PIN_EXTINT);
1747 			break;
1748 		case 4:
1749 		case 6:
1750 			vmcs_write(GUEST_ACTV_STATE, ACTV_HLT);
1751 			break;
1752 
1753 		case 8:
1754 			vmcs_write(GUEST_ACTV_STATE, ACTV_HLT);
1755 			vmcs_write(ENT_INTR_INFO,
1756 				   TIMER_VECTOR |
1757 				   (VMX_INTR_TYPE_EXT_INTR << INTR_INFO_INTR_TYPE_SHIFT) |
1758 				   INTR_INFO_VALID_MASK);
1759 			break;
1760 		}
1761 		vmx_inc_test_stage();
1762 		vmcs_write(GUEST_RIP, guest_rip + insn_len);
1763 		return VMX_TEST_RESUME;
1764 	case VMX_EXTINT:
1765 		if (vmcs_read(EXI_CONTROLS) & EXI_INTA) {
1766 			int vector = vmcs_read(EXI_INTR_INFO) & 0xff;
1767 			handle_external_interrupt(vector);
1768 		} else {
1769 			irq_enable();
1770 			asm volatile ("nop");
1771 			irq_disable();
1772 		}
1773 		if (vmx_get_test_stage() >= 2)
1774 			vmcs_write(GUEST_ACTV_STATE, ACTV_ACTIVE);
1775 		return VMX_TEST_RESUME;
1776 	default:
1777 		report("Unknown exit reason, %ld", false, reason);
1778 		print_vmexit_info();
1779 	}
1780 
1781 	return VMX_TEST_VMEXIT;
1782 }
1783 
1784 static int dbgctls_init(struct vmcs *vmcs)
1785 {
1786 	u64 dr7 = 0x402;
1787 	u64 zero = 0;
1788 
1789 	msr_bmp_init();
1790 	asm volatile(
1791 		"mov %0,%%dr0\n\t"
1792 		"mov %0,%%dr1\n\t"
1793 		"mov %0,%%dr2\n\t"
1794 		"mov %1,%%dr7\n\t"
1795 		: : "r" (zero), "r" (dr7));
1796 	wrmsr(MSR_IA32_DEBUGCTLMSR, 0x1);
1797 	vmcs_write(GUEST_DR7, 0x404);
1798 	vmcs_write(GUEST_DEBUGCTL, 0x2);
1799 
1800 	vmcs_write(ENT_CONTROLS, vmcs_read(ENT_CONTROLS) | ENT_LOAD_DBGCTLS);
1801 	vmcs_write(EXI_CONTROLS, vmcs_read(EXI_CONTROLS) | EXI_SAVE_DBGCTLS);
1802 
1803 	return VMX_TEST_START;
1804 }
1805 
1806 static void dbgctls_main(void)
1807 {
1808 	u64 dr7, debugctl;
1809 
1810 	asm volatile("mov %%dr7,%0" : "=r" (dr7));
1811 	debugctl = rdmsr(MSR_IA32_DEBUGCTLMSR);
1812 	/* Commented out: KVM does not support DEBUGCTL so far */
1813 	(void)debugctl;
1814 	report("Load debug controls", dr7 == 0x404 /* && debugctl == 0x2 */);
1815 
1816 	dr7 = 0x408;
1817 	asm volatile("mov %0,%%dr7" : : "r" (dr7));
1818 	wrmsr(MSR_IA32_DEBUGCTLMSR, 0x3);
1819 
1820 	vmx_set_test_stage(0);
1821 	vmcall();
1822 	report("Save debug controls", vmx_get_test_stage() == 1);
1823 
1824 	if (ctrl_enter_rev.set & ENT_LOAD_DBGCTLS ||
1825 	    ctrl_exit_rev.set & EXI_SAVE_DBGCTLS) {
1826 		printf("\tDebug controls are always loaded/saved\n");
1827 		return;
1828 	}
1829 	vmx_set_test_stage(2);
1830 	vmcall();
1831 
1832 	asm volatile("mov %%dr7,%0" : "=r" (dr7));
1833 	debugctl = rdmsr(MSR_IA32_DEBUGCTLMSR);
1834 	/* Commented out: KVM does not support DEBUGCTL so far */
1835 	(void)debugctl;
1836 	report("Guest=host debug controls", dr7 == 0x402 /* && debugctl == 0x1 */);
1837 
1838 	dr7 = 0x408;
1839 	asm volatile("mov %0,%%dr7" : : "r" (dr7));
1840 	wrmsr(MSR_IA32_DEBUGCTLMSR, 0x3);
1841 
1842 	vmx_set_test_stage(3);
1843 	vmcall();
1844 	report("Don't save debug controls", vmx_get_test_stage() == 4);
1845 }
1846 
1847 static int dbgctls_exit_handler(void)
1848 {
1849 	unsigned int reason = vmcs_read(EXI_REASON) & 0xff;
1850 	u32 insn_len = vmcs_read(EXI_INST_LEN);
1851 	u64 guest_rip = vmcs_read(GUEST_RIP);
1852 	u64 dr7, debugctl;
1853 
1854 	asm volatile("mov %%dr7,%0" : "=r" (dr7));
1855 	debugctl = rdmsr(MSR_IA32_DEBUGCTLMSR);
1856 
1857 	switch (reason) {
1858 	case VMX_VMCALL:
1859 		switch (vmx_get_test_stage()) {
1860 		case 0:
1861 			if (dr7 == 0x400 && debugctl == 0 &&
1862 			    vmcs_read(GUEST_DR7) == 0x408 /* &&
1863 			    Commented out: KVM does not support DEBUGCTL so far
1864 			    vmcs_read(GUEST_DEBUGCTL) == 0x3 */)
1865 				vmx_inc_test_stage();
1866 			break;
1867 		case 2:
1868 			dr7 = 0x402;
1869 			asm volatile("mov %0,%%dr7" : : "r" (dr7));
1870 			wrmsr(MSR_IA32_DEBUGCTLMSR, 0x1);
1871 			vmcs_write(GUEST_DR7, 0x404);
1872 			vmcs_write(GUEST_DEBUGCTL, 0x2);
1873 
1874 			vmcs_write(ENT_CONTROLS,
1875 				vmcs_read(ENT_CONTROLS) & ~ENT_LOAD_DBGCTLS);
1876 			vmcs_write(EXI_CONTROLS,
1877 				vmcs_read(EXI_CONTROLS) & ~EXI_SAVE_DBGCTLS);
1878 			break;
1879 		case 3:
1880 			if (dr7 == 0x400 && debugctl == 0 &&
1881 			    vmcs_read(GUEST_DR7) == 0x404 /* &&
1882 			    Commented out: KVM does not support DEBUGCTL so far
1883 			    vmcs_read(GUEST_DEBUGCTL) == 0x2 */)
1884 				vmx_inc_test_stage();
1885 			break;
1886 		}
1887 		vmcs_write(GUEST_RIP, guest_rip + insn_len);
1888 		return VMX_TEST_RESUME;
1889 	default:
1890 		report("Unknown exit reason, %d", false, reason);
1891 		print_vmexit_info();
1892 	}
1893 	return VMX_TEST_VMEXIT;
1894 }
1895 
1896 struct vmx_msr_entry {
1897 	u32 index;
1898 	u32 reserved;
1899 	u64 value;
1900 } __attribute__((packed));
1901 
1902 #define MSR_MAGIC 0x31415926
1903 struct vmx_msr_entry *exit_msr_store, *entry_msr_load, *exit_msr_load;
1904 
1905 static int msr_switch_init(struct vmcs *vmcs)
1906 {
1907 	msr_bmp_init();
1908 	exit_msr_store = alloc_page();
1909 	exit_msr_load = alloc_page();
1910 	entry_msr_load = alloc_page();
1911 	memset(exit_msr_store, 0, PAGE_SIZE);
1912 	memset(exit_msr_load, 0, PAGE_SIZE);
1913 	memset(entry_msr_load, 0, PAGE_SIZE);
1914 	entry_msr_load[0].index = MSR_KERNEL_GS_BASE;
1915 	entry_msr_load[0].value = MSR_MAGIC;
1916 
1917 	vmx_set_test_stage(1);
1918 	vmcs_write(ENT_MSR_LD_CNT, 1);
1919 	vmcs_write(ENTER_MSR_LD_ADDR, (u64)entry_msr_load);
1920 	vmcs_write(EXI_MSR_ST_CNT, 1);
1921 	vmcs_write(EXIT_MSR_ST_ADDR, (u64)exit_msr_store);
1922 	vmcs_write(EXI_MSR_LD_CNT, 1);
1923 	vmcs_write(EXIT_MSR_LD_ADDR, (u64)exit_msr_load);
1924 	return VMX_TEST_START;
1925 }
1926 
1927 static void msr_switch_main(void)
1928 {
1929 	if (vmx_get_test_stage() == 1) {
1930 		report("VM entry MSR load",
1931 			rdmsr(MSR_KERNEL_GS_BASE) == MSR_MAGIC);
1932 		vmx_set_test_stage(2);
1933 		wrmsr(MSR_KERNEL_GS_BASE, MSR_MAGIC + 1);
1934 		exit_msr_store[0].index = MSR_KERNEL_GS_BASE;
1935 		exit_msr_load[0].index = MSR_KERNEL_GS_BASE;
1936 		exit_msr_load[0].value = MSR_MAGIC + 2;
1937 	}
1938 	vmcall();
1939 }
1940 
1941 static int msr_switch_exit_handler(void)
1942 {
1943 	ulong reason;
1944 
1945 	reason = vmcs_read(EXI_REASON);
1946 	if (reason == VMX_VMCALL && vmx_get_test_stage() == 2) {
1947 		report("VM exit MSR store",
1948 			exit_msr_store[0].value == MSR_MAGIC + 1);
1949 		report("VM exit MSR load",
1950 			rdmsr(MSR_KERNEL_GS_BASE) == MSR_MAGIC + 2);
1951 		vmx_set_test_stage(3);
1952 		entry_msr_load[0].index = MSR_FS_BASE;
1953 		return VMX_TEST_RESUME;
1954 	}
1955 	printf("ERROR %s: unexpected stage=%u or reason=%lu\n",
1956 		__func__, vmx_get_test_stage(), reason);
1957 	return VMX_TEST_EXIT;
1958 }
1959 
1960 static int msr_switch_entry_failure(struct vmentry_failure *failure)
1961 {
1962 	ulong reason;
1963 
1964 	if (failure->early) {
1965 		printf("ERROR %s: early exit\n", __func__);
1966 		return VMX_TEST_EXIT;
1967 	}
1968 
1969 	reason = vmcs_read(EXI_REASON);
1970 	if (reason == (VMX_ENTRY_FAILURE | VMX_FAIL_MSR) &&
1971 	    vmx_get_test_stage() == 3) {
1972 		report("VM entry MSR load: try to load FS_BASE",
1973 			vmcs_read(EXI_QUALIFICATION) == 1);
1974 		return VMX_TEST_VMEXIT;
1975 	}
1976 	printf("ERROR %s: unexpected stage=%u or reason=%lu\n",
1977 		__func__, vmx_get_test_stage(), reason);
1978 	return VMX_TEST_EXIT;
1979 }
1980 
1981 static int vmmcall_init(struct vmcs *vmcs)
1982 {
1983 	vmcs_write(EXC_BITMAP, 1 << UD_VECTOR);
1984 	return VMX_TEST_START;
1985 }
1986 
1987 static void vmmcall_main(void)
1988 {
1989 	asm volatile(
1990 		"mov $0xABCD, %%rax\n\t"
1991 		"vmmcall\n\t"
1992 		::: "rax");
1993 
1994 	report("VMMCALL", 0);
1995 }
1996 
1997 static int vmmcall_exit_handler(void)
1998 {
1999 	ulong reason;
2000 
2001 	reason = vmcs_read(EXI_REASON);
2002 	switch (reason) {
2003 	case VMX_VMCALL:
2004 		printf("here\n");
2005 		report("VMMCALL triggers #UD", 0);
2006 		break;
2007 	case VMX_EXC_NMI:
2008 		report("VMMCALL triggers #UD",
2009 		       (vmcs_read(EXI_INTR_INFO) & 0xff) == UD_VECTOR);
2010 		break;
2011 	default:
2012 		report("Unknown exit reason, %ld", false, reason);
2013 		print_vmexit_info();
2014 	}
2015 
2016 	return VMX_TEST_VMEXIT;
2017 }
2018 
2019 static int disable_rdtscp_init(struct vmcs *vmcs)
2020 {
2021 	u32 ctrl_cpu1;
2022 
2023 	if (ctrl_cpu_rev[0].clr & CPU_SECONDARY) {
2024 		ctrl_cpu1 = vmcs_read(CPU_EXEC_CTRL1);
2025 		ctrl_cpu1 &= ~CPU_RDTSCP;
2026 		vmcs_write(CPU_EXEC_CTRL1, ctrl_cpu1);
2027 	}
2028 
2029 	return VMX_TEST_START;
2030 }
2031 
2032 static void disable_rdtscp_ud_handler(struct ex_regs *regs)
2033 {
2034 	switch (vmx_get_test_stage()) {
2035 	case 0:
2036 		report("RDTSCP triggers #UD", true);
2037 		vmx_inc_test_stage();
2038 		regs->rip += 3;
2039 		break;
2040 	case 2:
2041 		report("RDPID triggers #UD", true);
2042 		vmx_inc_test_stage();
2043 		regs->rip += 4;
2044 		break;
2045 	}
2046 	return;
2047 
2048 }
2049 
2050 static void disable_rdtscp_main(void)
2051 {
2052 	/* Test that #UD is properly injected in L2.  */
2053 	handle_exception(UD_VECTOR, disable_rdtscp_ud_handler);
2054 
2055 	vmx_set_test_stage(0);
2056 	asm volatile("rdtscp" : : : "eax", "ecx", "edx");
2057 	vmcall();
2058 	asm volatile(".byte 0xf3, 0x0f, 0xc7, 0xf8" : : : "eax");
2059 
2060 	handle_exception(UD_VECTOR, 0);
2061 	vmcall();
2062 }
2063 
2064 static int disable_rdtscp_exit_handler(void)
2065 {
2066 	unsigned int reason = vmcs_read(EXI_REASON) & 0xff;
2067 
2068 	switch (reason) {
2069 	case VMX_VMCALL:
2070 		switch (vmx_get_test_stage()) {
2071 		case 0:
2072 			report("RDTSCP triggers #UD", false);
2073 			vmx_inc_test_stage();
2074 			/* fallthrough */
2075 		case 1:
2076 			vmx_inc_test_stage();
2077 			vmcs_write(GUEST_RIP, vmcs_read(GUEST_RIP) + 3);
2078 			return VMX_TEST_RESUME;
2079 		case 2:
2080 			report("RDPID triggers #UD", false);
2081 			break;
2082 		}
2083 		break;
2084 
2085 	default:
2086 		report("Unknown exit reason, %d", false, reason);
2087 		print_vmexit_info();
2088 	}
2089 	return VMX_TEST_VMEXIT;
2090 }
2091 
2092 static int int3_init(struct vmcs *vmcs)
2093 {
2094 	vmcs_write(EXC_BITMAP, ~0u);
2095 	return VMX_TEST_START;
2096 }
2097 
2098 static void int3_guest_main(void)
2099 {
2100 	asm volatile ("int3");
2101 }
2102 
2103 static int int3_exit_handler(void)
2104 {
2105 	u32 reason = vmcs_read(EXI_REASON);
2106 	u32 intr_info = vmcs_read(EXI_INTR_INFO);
2107 
2108 	report("L1 intercepts #BP", reason == VMX_EXC_NMI &&
2109 	       (intr_info & INTR_INFO_VALID_MASK) &&
2110 	       (intr_info & INTR_INFO_VECTOR_MASK) == BP_VECTOR &&
2111 	       ((intr_info & INTR_INFO_INTR_TYPE_MASK) >>
2112 		INTR_INFO_INTR_TYPE_SHIFT) == VMX_INTR_TYPE_SOFT_EXCEPTION);
2113 
2114 	return VMX_TEST_VMEXIT;
2115 }
2116 
2117 static int into_init(struct vmcs *vmcs)
2118 {
2119 	vmcs_write(EXC_BITMAP, ~0u);
2120 	return VMX_TEST_START;
2121 }
2122 
2123 static void into_guest_main(void)
2124 {
2125 	struct far_pointer32 fp = {
2126 		.offset = (uintptr_t)&&into,
2127 		.selector = KERNEL_CS32,
2128 	};
2129 	register uintptr_t rsp asm("rsp");
2130 
2131 	if (fp.offset != (uintptr_t)&&into) {
2132 		printf("Code address too high.\n");
2133 		return;
2134 	}
2135 	if ((u32)rsp != rsp) {
2136 		printf("Stack address too high.\n");
2137 		return;
2138 	}
2139 
2140 	asm goto ("lcall *%0" : : "m" (fp) : "rax" : into);
2141 	return;
2142 into:
2143 	asm volatile (".code32;"
2144 		      "movl $0x7fffffff, %eax;"
2145 		      "addl %eax, %eax;"
2146 		      "into;"
2147 		      "lret;"
2148 		      ".code64");
2149 	__builtin_unreachable();
2150 }
2151 
2152 static int into_exit_handler(void)
2153 {
2154 	u32 reason = vmcs_read(EXI_REASON);
2155 	u32 intr_info = vmcs_read(EXI_INTR_INFO);
2156 
2157 	report("L1 intercepts #OF", reason == VMX_EXC_NMI &&
2158 	       (intr_info & INTR_INFO_VALID_MASK) &&
2159 	       (intr_info & INTR_INFO_VECTOR_MASK) == OF_VECTOR &&
2160 	       ((intr_info & INTR_INFO_INTR_TYPE_MASK) >>
2161 		INTR_INFO_INTR_TYPE_SHIFT) == VMX_INTR_TYPE_SOFT_EXCEPTION);
2162 
2163 	return VMX_TEST_VMEXIT;
2164 }
2165 
2166 static void exit_monitor_from_l2_main(void)
2167 {
2168 	printf("Calling exit(0) from l2...\n");
2169 	exit(0);
2170 }
2171 
2172 static int exit_monitor_from_l2_handler(void)
2173 {
2174 	report("The guest should have killed the VMM", false);
2175 	return VMX_TEST_EXIT;
2176 }
2177 
2178 static void assert_exit_reason(u64 expected)
2179 {
2180 	u64 actual = vmcs_read(EXI_REASON);
2181 
2182 	TEST_ASSERT_EQ_MSG(expected, actual, "Expected %s, got %s.",
2183 			   exit_reason_description(expected),
2184 			   exit_reason_description(actual));
2185 }
2186 
2187 static void skip_exit_insn(void)
2188 {
2189 	u64 guest_rip = vmcs_read(GUEST_RIP);
2190 	u32 insn_len = vmcs_read(EXI_INST_LEN);
2191 	vmcs_write(GUEST_RIP, guest_rip + insn_len);
2192 }
2193 
2194 static void skip_exit_vmcall(void)
2195 {
2196 	assert_exit_reason(VMX_VMCALL);
2197 	skip_exit_insn();
2198 }
2199 
2200 static void v2_null_test_guest(void)
2201 {
2202 }
2203 
2204 static void v2_null_test(void)
2205 {
2206 	test_set_guest(v2_null_test_guest);
2207 	enter_guest();
2208 	report(__func__, 1);
2209 }
2210 
2211 static void v2_multiple_entries_test_guest(void)
2212 {
2213 	vmx_set_test_stage(1);
2214 	vmcall();
2215 	vmx_set_test_stage(2);
2216 }
2217 
2218 static void v2_multiple_entries_test(void)
2219 {
2220 	test_set_guest(v2_multiple_entries_test_guest);
2221 	enter_guest();
2222 	TEST_ASSERT_EQ(vmx_get_test_stage(), 1);
2223 	skip_exit_vmcall();
2224 	enter_guest();
2225 	TEST_ASSERT_EQ(vmx_get_test_stage(), 2);
2226 	report(__func__, 1);
2227 }
2228 
2229 static int fixture_test_data = 1;
2230 
2231 static void fixture_test_teardown(void *data)
2232 {
2233 	*((int *) data) = 1;
2234 }
2235 
2236 static void fixture_test_guest(void)
2237 {
2238 	fixture_test_data++;
2239 }
2240 
2241 
2242 static void fixture_test_setup(void)
2243 {
2244 	TEST_ASSERT_EQ_MSG(1, fixture_test_data,
2245 			   "fixture_test_teardown didn't run?!");
2246 	fixture_test_data = 2;
2247 	test_add_teardown(fixture_test_teardown, &fixture_test_data);
2248 	test_set_guest(fixture_test_guest);
2249 }
2250 
2251 static void fixture_test_case1(void)
2252 {
2253 	fixture_test_setup();
2254 	TEST_ASSERT_EQ(2, fixture_test_data);
2255 	enter_guest();
2256 	TEST_ASSERT_EQ(3, fixture_test_data);
2257 	report(__func__, 1);
2258 }
2259 
2260 static void fixture_test_case2(void)
2261 {
2262 	fixture_test_setup();
2263 	TEST_ASSERT_EQ(2, fixture_test_data);
2264 	enter_guest();
2265 	TEST_ASSERT_EQ(3, fixture_test_data);
2266 	report(__func__, 1);
2267 }
2268 
2269 enum ept_access_op {
2270 	OP_READ,
2271 	OP_WRITE,
2272 	OP_EXEC,
2273 	OP_FLUSH_TLB,
2274 	OP_EXIT,
2275 };
2276 
2277 static struct ept_access_test_data {
2278 	unsigned long gpa;
2279 	unsigned long *gva;
2280 	unsigned long hpa;
2281 	unsigned long *hva;
2282 	enum ept_access_op op;
2283 } ept_access_test_data;
2284 
2285 extern unsigned char ret42_start;
2286 extern unsigned char ret42_end;
2287 
2288 /* Returns 42. */
2289 asm(
2290 	".align 64\n"
2291 	"ret42_start:\n"
2292 	"mov $42, %eax\n"
2293 	"ret\n"
2294 	"ret42_end:\n"
2295 );
2296 
2297 static void
2298 diagnose_ept_violation_qual(u64 expected, u64 actual)
2299 {
2300 
2301 #define DIAGNOSE(flag)							\
2302 do {									\
2303 	if ((expected & flag) != (actual & flag))			\
2304 		printf(#flag " %sexpected\n",				\
2305 		       (expected & flag) ? "" : "un");			\
2306 } while (0)
2307 
2308 	DIAGNOSE(EPT_VLT_RD);
2309 	DIAGNOSE(EPT_VLT_WR);
2310 	DIAGNOSE(EPT_VLT_FETCH);
2311 	DIAGNOSE(EPT_VLT_PERM_RD);
2312 	DIAGNOSE(EPT_VLT_PERM_WR);
2313 	DIAGNOSE(EPT_VLT_PERM_EX);
2314 	DIAGNOSE(EPT_VLT_LADDR_VLD);
2315 	DIAGNOSE(EPT_VLT_PADDR);
2316 
2317 #undef DIAGNOSE
2318 }
2319 
2320 static void do_ept_access_op(enum ept_access_op op)
2321 {
2322 	ept_access_test_data.op = op;
2323 	enter_guest();
2324 }
2325 
2326 /*
2327  * Force the guest to flush its TLB (i.e., flush gva -> gpa mappings). Only
2328  * needed by tests that modify guest PTEs.
2329  */
2330 static void ept_access_test_guest_flush_tlb(void)
2331 {
2332 	do_ept_access_op(OP_FLUSH_TLB);
2333 	skip_exit_vmcall();
2334 }
2335 
2336 /*
2337  * Modifies the EPT entry at @level in the mapping of @gpa. First clears the
2338  * bits in @clear then sets the bits in @set. @mkhuge transforms the entry into
2339  * a huge page.
2340  */
2341 static unsigned long ept_twiddle(unsigned long gpa, bool mkhuge, int level,
2342 				 unsigned long clear, unsigned long set)
2343 {
2344 	struct ept_access_test_data *data = &ept_access_test_data;
2345 	unsigned long orig_pte;
2346 	unsigned long pte;
2347 
2348 	/* Screw with the mapping at the requested level. */
2349 	TEST_ASSERT(get_ept_pte(pml4, gpa, level, &orig_pte));
2350 	pte = orig_pte;
2351 	if (mkhuge)
2352 		pte = (orig_pte & ~EPT_ADDR_MASK) | data->hpa | EPT_LARGE_PAGE;
2353 	else
2354 		pte = orig_pte;
2355 	pte = (pte & ~clear) | set;
2356 	set_ept_pte(pml4, gpa, level, pte);
2357 	ept_sync(INVEPT_SINGLE, eptp);
2358 
2359 	return orig_pte;
2360 }
2361 
2362 static void ept_untwiddle(unsigned long gpa, int level, unsigned long orig_pte)
2363 {
2364 	set_ept_pte(pml4, gpa, level, orig_pte);
2365 }
2366 
2367 static void do_ept_violation(bool leaf, enum ept_access_op op,
2368 			     u64 expected_qual, u64 expected_paddr)
2369 {
2370 	u64 qual;
2371 
2372 	/* Try the access and observe the violation. */
2373 	do_ept_access_op(op);
2374 
2375 	assert_exit_reason(VMX_EPT_VIOLATION);
2376 
2377 	qual = vmcs_read(EXI_QUALIFICATION);
2378 
2379 	diagnose_ept_violation_qual(expected_qual, qual);
2380 	TEST_EXPECT_EQ(expected_qual, qual);
2381 
2382 	#if 0
2383 	/* Disable for now otherwise every test will fail */
2384 	TEST_EXPECT_EQ(vmcs_read(GUEST_LINEAR_ADDRESS),
2385 		       (unsigned long) (
2386 			       op == OP_EXEC ? data->gva + 1 : data->gva));
2387 	#endif
2388 	/*
2389 	 * TODO: tests that probe expected_paddr in pages other than the one at
2390 	 * the beginning of the 1g region.
2391 	 */
2392 	TEST_EXPECT_EQ(vmcs_read(INFO_PHYS_ADDR), expected_paddr);
2393 }
2394 
2395 static void
2396 ept_violation_at_level_mkhuge(bool mkhuge, int level, unsigned long clear,
2397 			      unsigned long set, enum ept_access_op op,
2398 			      u64 expected_qual)
2399 {
2400 	struct ept_access_test_data *data = &ept_access_test_data;
2401 	unsigned long orig_pte;
2402 
2403 	orig_pte = ept_twiddle(data->gpa, mkhuge, level, clear, set);
2404 
2405 	do_ept_violation(level == 1 || mkhuge, op, expected_qual,
2406 			 op == OP_EXEC ? data->gpa + sizeof(unsigned long) :
2407 					 data->gpa);
2408 
2409 	/* Fix the violation and resume the op loop. */
2410 	ept_untwiddle(data->gpa, level, orig_pte);
2411 	enter_guest();
2412 	skip_exit_vmcall();
2413 }
2414 
2415 static void
2416 ept_violation_at_level(int level, unsigned long clear, unsigned long set,
2417 		       enum ept_access_op op, u64 expected_qual)
2418 {
2419 	ept_violation_at_level_mkhuge(false, level, clear, set, op,
2420 				      expected_qual);
2421 	if (ept_huge_pages_supported(level))
2422 		ept_violation_at_level_mkhuge(true, level, clear, set, op,
2423 					      expected_qual);
2424 }
2425 
2426 static void ept_violation(unsigned long clear, unsigned long set,
2427 			  enum ept_access_op op, u64 expected_qual)
2428 {
2429 	ept_violation_at_level(1, clear, set, op, expected_qual);
2430 	ept_violation_at_level(2, clear, set, op, expected_qual);
2431 	ept_violation_at_level(3, clear, set, op, expected_qual);
2432 	ept_violation_at_level(4, clear, set, op, expected_qual);
2433 }
2434 
2435 static void ept_access_violation(unsigned long access, enum ept_access_op op,
2436 				       u64 expected_qual)
2437 {
2438 	ept_violation(EPT_PRESENT, access, op,
2439 		      expected_qual | EPT_VLT_LADDR_VLD | EPT_VLT_PADDR);
2440 }
2441 
2442 /*
2443  * For translations that don't involve a GVA, that is physical address (paddr)
2444  * accesses, EPT violations don't set the flag EPT_VLT_PADDR.  For a typical
2445  * guest memory access, the hardware does GVA -> GPA -> HPA.  However, certain
2446  * translations don't involve GVAs, such as when the hardware does the guest
2447  * page table walk. For example, in translating GVA_1 -> GPA_1, the guest MMU
2448  * might try to set an A bit on a guest PTE. If the GPA_2 that the PTE resides
2449  * on isn't present in the EPT, then the EPT violation will be for GPA_2 and
2450  * the EPT_VLT_PADDR bit will be clear in the exit qualification.
2451  *
2452  * Note that paddr violations can also be triggered by loading PAE page tables
2453  * with wonky addresses. We don't test that yet.
2454  *
2455  * This function modifies the EPT entry that maps the GPA that the guest page
2456  * table entry mapping ept_access_data.gva resides on.
2457  *
2458  *	@ept_access	EPT permissions to set. Other permissions are cleared.
2459  *
2460  *	@pte_ad		Set the A/D bits on the guest PTE accordingly.
2461  *
2462  *	@op		Guest operation to perform with ept_access_data.gva.
2463  *
2464  *	@expect_violation
2465  *			Is a violation expected during the paddr access?
2466  *
2467  *	@expected_qual	Expected qualification for the EPT violation.
2468  *			EPT_VLT_PADDR should be clear.
2469  */
2470 static void ept_access_paddr(unsigned long ept_access, unsigned long pte_ad,
2471 			     enum ept_access_op op, bool expect_violation,
2472 			     u64 expected_qual)
2473 {
2474 	struct ept_access_test_data *data = &ept_access_test_data;
2475 	unsigned long *ptep;
2476 	unsigned long gpa;
2477 	unsigned long orig_epte;
2478 
2479 	/* Modify the guest PTE mapping data->gva according to @pte_ad.  */
2480 	ptep = get_pte_level(current_page_table(), data->gva, /*level=*/1);
2481 	TEST_ASSERT(ptep);
2482 	TEST_ASSERT_EQ(*ptep & PT_ADDR_MASK, data->gpa);
2483 	*ptep = (*ptep & ~PT_AD_MASK) | pte_ad;
2484 	ept_access_test_guest_flush_tlb();
2485 
2486 	/*
2487 	 * Now modify the access bits on the EPT entry for the GPA that the
2488 	 * guest PTE resides on. Note that by modifying a single EPT entry,
2489 	 * we're potentially affecting 512 guest PTEs. However, we've carefully
2490 	 * constructed our test such that those other 511 PTEs aren't used by
2491 	 * the guest: data->gva is at the beginning of a 1G huge page, thus the
2492 	 * PTE we're modifying is at the beginning of a 4K page and the
2493 	 * following 511 entires are also under our control (and not touched by
2494 	 * the guest).
2495 	 */
2496 	gpa = virt_to_phys(ptep);
2497 	TEST_ASSERT_EQ(gpa & ~PAGE_MASK, 0);
2498 	/*
2499 	 * Make sure the guest page table page is mapped with a 4K EPT entry,
2500 	 * otherwise our level=1 twiddling below will fail. We use the
2501 	 * identity map (gpa = gpa) since page tables are shared with the host.
2502 	 */
2503 	install_ept(pml4, gpa, gpa, EPT_PRESENT);
2504 	orig_epte = ept_twiddle(gpa, /*mkhuge=*/0, /*level=*/1,
2505 				/*clear=*/EPT_PRESENT, /*set=*/ept_access);
2506 
2507 	if (expect_violation) {
2508 		do_ept_violation(/*leaf=*/true, op,
2509 				 expected_qual | EPT_VLT_LADDR_VLD, gpa);
2510 		ept_untwiddle(gpa, /*level=*/1, orig_epte);
2511 		do_ept_access_op(op);
2512 	} else {
2513 		do_ept_access_op(op);
2514 		ept_untwiddle(gpa, /*level=*/1, orig_epte);
2515 	}
2516 
2517 	TEST_ASSERT(*ptep & PT_ACCESSED_MASK);
2518 	if ((pte_ad & PT_DIRTY_MASK) || op == OP_WRITE)
2519 		TEST_ASSERT(*ptep & PT_DIRTY_MASK);
2520 
2521 	skip_exit_vmcall();
2522 }
2523 
2524 static void ept_access_allowed_paddr(unsigned long ept_access,
2525 				     unsigned long pte_ad,
2526 				     enum ept_access_op op)
2527 {
2528 	ept_access_paddr(ept_access, pte_ad, op, /*expect_violation=*/false,
2529 			 /*expected_qual=*/-1);
2530 }
2531 
2532 static void ept_access_violation_paddr(unsigned long ept_access,
2533 				       unsigned long pte_ad,
2534 				       enum ept_access_op op,
2535 				       u64 expected_qual)
2536 {
2537 	ept_access_paddr(ept_access, pte_ad, op, /*expect_violation=*/true,
2538 			 expected_qual);
2539 }
2540 
2541 
2542 static void ept_allowed_at_level_mkhuge(bool mkhuge, int level,
2543 					unsigned long clear,
2544 					unsigned long set,
2545 					enum ept_access_op op)
2546 {
2547 	struct ept_access_test_data *data = &ept_access_test_data;
2548 	unsigned long orig_pte;
2549 
2550 	orig_pte = ept_twiddle(data->gpa, mkhuge, level, clear, set);
2551 
2552 	/* No violation. Should proceed to vmcall. */
2553 	do_ept_access_op(op);
2554 	skip_exit_vmcall();
2555 
2556 	ept_untwiddle(data->gpa, level, orig_pte);
2557 }
2558 
2559 static void ept_allowed_at_level(int level, unsigned long clear,
2560 				 unsigned long set, enum ept_access_op op)
2561 {
2562 	ept_allowed_at_level_mkhuge(false, level, clear, set, op);
2563 	if (ept_huge_pages_supported(level))
2564 		ept_allowed_at_level_mkhuge(true, level, clear, set, op);
2565 }
2566 
2567 static void ept_allowed(unsigned long clear, unsigned long set,
2568 			enum ept_access_op op)
2569 {
2570 	ept_allowed_at_level(1, clear, set, op);
2571 	ept_allowed_at_level(2, clear, set, op);
2572 	ept_allowed_at_level(3, clear, set, op);
2573 	ept_allowed_at_level(4, clear, set, op);
2574 }
2575 
2576 static void ept_ignored_bit(int bit)
2577 {
2578 	/* Set the bit. */
2579 	ept_allowed(0, 1ul << bit, OP_READ);
2580 	ept_allowed(0, 1ul << bit, OP_WRITE);
2581 	ept_allowed(0, 1ul << bit, OP_EXEC);
2582 
2583 	/* Clear the bit. */
2584 	ept_allowed(1ul << bit, 0, OP_READ);
2585 	ept_allowed(1ul << bit, 0, OP_WRITE);
2586 	ept_allowed(1ul << bit, 0, OP_EXEC);
2587 }
2588 
2589 static void ept_access_allowed(unsigned long access, enum ept_access_op op)
2590 {
2591 	ept_allowed(EPT_PRESENT, access, op);
2592 }
2593 
2594 
2595 static void ept_misconfig_at_level_mkhuge_op(bool mkhuge, int level,
2596 					     unsigned long clear,
2597 					     unsigned long set,
2598 					     enum ept_access_op op)
2599 {
2600 	struct ept_access_test_data *data = &ept_access_test_data;
2601 	unsigned long orig_pte;
2602 
2603 	orig_pte = ept_twiddle(data->gpa, mkhuge, level, clear, set);
2604 
2605 	do_ept_access_op(op);
2606 	assert_exit_reason(VMX_EPT_MISCONFIG);
2607 
2608 	/* Intel 27.2.1, "For all other VM exits, this field is cleared." */
2609 	#if 0
2610 	/* broken: */
2611 	TEST_EXPECT_EQ_MSG(vmcs_read(EXI_QUALIFICATION), 0);
2612 	#endif
2613 	#if 0
2614 	/*
2615 	 * broken:
2616 	 * According to description of exit qual for EPT violation,
2617 	 * EPT_VLT_LADDR_VLD indicates if GUEST_LINEAR_ADDRESS is valid.
2618 	 * However, I can't find anything that says GUEST_LINEAR_ADDRESS ought
2619 	 * to be set for msiconfig.
2620 	 */
2621 	TEST_EXPECT_EQ(vmcs_read(GUEST_LINEAR_ADDRESS),
2622 		       (unsigned long) (
2623 			       op == OP_EXEC ? data->gva + 1 : data->gva));
2624 	#endif
2625 
2626 	/* Fix the violation and resume the op loop. */
2627 	ept_untwiddle(data->gpa, level, orig_pte);
2628 	enter_guest();
2629 	skip_exit_vmcall();
2630 }
2631 
2632 static void ept_misconfig_at_level_mkhuge(bool mkhuge, int level,
2633 					  unsigned long clear,
2634 					  unsigned long set)
2635 {
2636 	/* The op shouldn't matter (read, write, exec), so try them all! */
2637 	ept_misconfig_at_level_mkhuge_op(mkhuge, level, clear, set, OP_READ);
2638 	ept_misconfig_at_level_mkhuge_op(mkhuge, level, clear, set, OP_WRITE);
2639 	ept_misconfig_at_level_mkhuge_op(mkhuge, level, clear, set, OP_EXEC);
2640 }
2641 
2642 static void ept_misconfig_at_level(int level, unsigned long clear,
2643 				   unsigned long set)
2644 {
2645 	ept_misconfig_at_level_mkhuge(false, level, clear, set);
2646 	if (ept_huge_pages_supported(level))
2647 		ept_misconfig_at_level_mkhuge(true, level, clear, set);
2648 }
2649 
2650 static void ept_misconfig(unsigned long clear, unsigned long set)
2651 {
2652 	ept_misconfig_at_level(1, clear, set);
2653 	ept_misconfig_at_level(2, clear, set);
2654 	ept_misconfig_at_level(3, clear, set);
2655 	ept_misconfig_at_level(4, clear, set);
2656 }
2657 
2658 static void ept_access_misconfig(unsigned long access)
2659 {
2660 	ept_misconfig(EPT_PRESENT, access);
2661 }
2662 
2663 static void ept_reserved_bit_at_level_nohuge(int level, int bit)
2664 {
2665 	/* Setting the bit causes a misconfig. */
2666 	ept_misconfig_at_level_mkhuge(false, level, 0, 1ul << bit);
2667 
2668 	/* Making the entry non-present turns reserved bits into ignored. */
2669 	ept_violation_at_level(level, EPT_PRESENT, 1ul << bit, OP_READ,
2670 			       EPT_VLT_RD | EPT_VLT_LADDR_VLD | EPT_VLT_PADDR);
2671 }
2672 
2673 static void ept_reserved_bit_at_level_huge(int level, int bit)
2674 {
2675 	/* Setting the bit causes a misconfig. */
2676 	ept_misconfig_at_level_mkhuge(true, level, 0, 1ul << bit);
2677 
2678 	/* Making the entry non-present turns reserved bits into ignored. */
2679 	ept_violation_at_level(level, EPT_PRESENT, 1ul << bit, OP_READ,
2680 			       EPT_VLT_RD | EPT_VLT_LADDR_VLD | EPT_VLT_PADDR);
2681 }
2682 
2683 static void ept_reserved_bit_at_level(int level, int bit)
2684 {
2685 	/* Setting the bit causes a misconfig. */
2686 	ept_misconfig_at_level(level, 0, 1ul << bit);
2687 
2688 	/* Making the entry non-present turns reserved bits into ignored. */
2689 	ept_violation_at_level(level, EPT_PRESENT, 1ul << bit, OP_READ,
2690 			       EPT_VLT_RD | EPT_VLT_LADDR_VLD | EPT_VLT_PADDR);
2691 }
2692 
2693 static void ept_reserved_bit(int bit)
2694 {
2695 	ept_reserved_bit_at_level(1, bit);
2696 	ept_reserved_bit_at_level(2, bit);
2697 	ept_reserved_bit_at_level(3, bit);
2698 	ept_reserved_bit_at_level(4, bit);
2699 }
2700 
2701 #define PAGE_2M_ORDER 9
2702 #define PAGE_1G_ORDER 18
2703 
2704 static void *get_1g_page(void)
2705 {
2706 	static void *alloc;
2707 
2708 	if (!alloc)
2709 		alloc = alloc_pages(PAGE_1G_ORDER);
2710 	return alloc;
2711 }
2712 
2713 static void ept_access_test_teardown(void *unused)
2714 {
2715 	/* Exit the guest cleanly. */
2716 	do_ept_access_op(OP_EXIT);
2717 }
2718 
2719 static void ept_access_test_guest(void)
2720 {
2721 	struct ept_access_test_data *data = &ept_access_test_data;
2722 	int (*code)(void) = (int (*)(void)) &data->gva[1];
2723 
2724 	while (true) {
2725 		switch (data->op) {
2726 		case OP_READ:
2727 			TEST_ASSERT_EQ(*data->gva, MAGIC_VAL_1);
2728 			break;
2729 		case OP_WRITE:
2730 			*data->gva = MAGIC_VAL_2;
2731 			TEST_ASSERT_EQ(*data->gva, MAGIC_VAL_2);
2732 			*data->gva = MAGIC_VAL_1;
2733 			break;
2734 		case OP_EXEC:
2735 			TEST_ASSERT_EQ(42, code());
2736 			break;
2737 		case OP_FLUSH_TLB:
2738 			write_cr3(read_cr3());
2739 			break;
2740 		case OP_EXIT:
2741 			return;
2742 		default:
2743 			TEST_ASSERT_MSG(false, "Unknown op %d", data->op);
2744 		}
2745 		vmcall();
2746 	}
2747 }
2748 
2749 static void ept_access_test_setup(void)
2750 {
2751 	struct ept_access_test_data *data = &ept_access_test_data;
2752 	unsigned long npages = 1ul << PAGE_1G_ORDER;
2753 	unsigned long size = npages * PAGE_SIZE;
2754 	unsigned long *page_table = current_page_table();
2755 	unsigned long pte;
2756 
2757 	if (setup_ept(false))
2758 		test_skip("EPT not supported");
2759 
2760 	/* We use data->gpa = 1 << 39 so that test data has a separate pml4 entry */
2761 	if (cpuid_maxphyaddr() < 40)
2762 		test_skip("Test needs MAXPHYADDR >= 40");
2763 
2764 	test_set_guest(ept_access_test_guest);
2765 	test_add_teardown(ept_access_test_teardown, NULL);
2766 
2767 	data->hva = get_1g_page();
2768 	TEST_ASSERT(data->hva);
2769 	data->hpa = virt_to_phys(data->hva);
2770 
2771 	data->gpa = 1ul << 39;
2772 	data->gva = (void *) ALIGN((unsigned long) alloc_vpages(npages * 2),
2773 				   size);
2774 	TEST_ASSERT(!any_present_pages(page_table, data->gva, size));
2775 	install_pages(page_table, data->gpa, size, data->gva);
2776 
2777 	/*
2778 	 * Make sure nothing's mapped here so the tests that screw with the
2779 	 * pml4 entry don't inadvertently break something.
2780 	 */
2781 	TEST_ASSERT(get_ept_pte(pml4, data->gpa, 4, &pte) && pte == 0);
2782 	TEST_ASSERT(get_ept_pte(pml4, data->gpa + size - 1, 4, &pte) && pte == 0);
2783 	install_ept(pml4, data->hpa, data->gpa, EPT_PRESENT);
2784 
2785 	data->hva[0] = MAGIC_VAL_1;
2786 	memcpy(&data->hva[1], &ret42_start, &ret42_end - &ret42_start);
2787 }
2788 
2789 static void ept_access_test_not_present(void)
2790 {
2791 	ept_access_test_setup();
2792 	/* --- */
2793 	ept_access_violation(0, OP_READ, EPT_VLT_RD);
2794 	ept_access_violation(0, OP_WRITE, EPT_VLT_WR);
2795 	ept_access_violation(0, OP_EXEC, EPT_VLT_FETCH);
2796 }
2797 
2798 static void ept_access_test_read_only(void)
2799 {
2800 	ept_access_test_setup();
2801 
2802 	/* r-- */
2803 	ept_access_allowed(EPT_RA, OP_READ);
2804 	ept_access_violation(EPT_RA, OP_WRITE, EPT_VLT_WR | EPT_VLT_PERM_RD);
2805 	ept_access_violation(EPT_RA, OP_EXEC, EPT_VLT_FETCH | EPT_VLT_PERM_RD);
2806 }
2807 
2808 static void ept_access_test_write_only(void)
2809 {
2810 	ept_access_test_setup();
2811 	/* -w- */
2812 	ept_access_misconfig(EPT_WA);
2813 }
2814 
2815 static void ept_access_test_read_write(void)
2816 {
2817 	ept_access_test_setup();
2818 	/* rw- */
2819 	ept_access_allowed(EPT_RA | EPT_WA, OP_READ);
2820 	ept_access_allowed(EPT_RA | EPT_WA, OP_WRITE);
2821 	ept_access_violation(EPT_RA | EPT_WA, OP_EXEC,
2822 			   EPT_VLT_FETCH | EPT_VLT_PERM_RD | EPT_VLT_PERM_WR);
2823 }
2824 
2825 
2826 static void ept_access_test_execute_only(void)
2827 {
2828 	ept_access_test_setup();
2829 	/* --x */
2830 	if (ept_execute_only_supported()) {
2831 		ept_access_violation(EPT_EA, OP_READ,
2832 				     EPT_VLT_RD | EPT_VLT_PERM_EX);
2833 		ept_access_violation(EPT_EA, OP_WRITE,
2834 				     EPT_VLT_WR | EPT_VLT_PERM_EX);
2835 		ept_access_allowed(EPT_EA, OP_EXEC);
2836 	} else {
2837 		ept_access_misconfig(EPT_EA);
2838 	}
2839 }
2840 
2841 static void ept_access_test_read_execute(void)
2842 {
2843 	ept_access_test_setup();
2844 	/* r-x */
2845 	ept_access_allowed(EPT_RA | EPT_EA, OP_READ);
2846 	ept_access_violation(EPT_RA | EPT_EA, OP_WRITE,
2847 			   EPT_VLT_WR | EPT_VLT_PERM_RD | EPT_VLT_PERM_EX);
2848 	ept_access_allowed(EPT_RA | EPT_EA, OP_EXEC);
2849 }
2850 
2851 static void ept_access_test_write_execute(void)
2852 {
2853 	ept_access_test_setup();
2854 	/* -wx */
2855 	ept_access_misconfig(EPT_WA | EPT_EA);
2856 }
2857 
2858 static void ept_access_test_read_write_execute(void)
2859 {
2860 	ept_access_test_setup();
2861 	/* rwx */
2862 	ept_access_allowed(EPT_RA | EPT_WA | EPT_EA, OP_READ);
2863 	ept_access_allowed(EPT_RA | EPT_WA | EPT_EA, OP_WRITE);
2864 	ept_access_allowed(EPT_RA | EPT_WA | EPT_EA, OP_EXEC);
2865 }
2866 
2867 static void ept_access_test_reserved_bits(void)
2868 {
2869 	int i;
2870 	int maxphyaddr;
2871 
2872 	ept_access_test_setup();
2873 
2874 	/* Reserved bits above maxphyaddr. */
2875 	maxphyaddr = cpuid_maxphyaddr();
2876 	for (i = maxphyaddr; i <= 51; i++) {
2877 		report_prefix_pushf("reserved_bit=%d", i);
2878 		ept_reserved_bit(i);
2879 		report_prefix_pop();
2880 	}
2881 
2882 	/* Level-specific reserved bits. */
2883 	ept_reserved_bit_at_level_nohuge(2, 3);
2884 	ept_reserved_bit_at_level_nohuge(2, 4);
2885 	ept_reserved_bit_at_level_nohuge(2, 5);
2886 	ept_reserved_bit_at_level_nohuge(2, 6);
2887 	/* 2M alignment. */
2888 	for (i = 12; i < 20; i++) {
2889 		report_prefix_pushf("reserved_bit=%d", i);
2890 		ept_reserved_bit_at_level_huge(2, i);
2891 		report_prefix_pop();
2892 	}
2893 	ept_reserved_bit_at_level_nohuge(3, 3);
2894 	ept_reserved_bit_at_level_nohuge(3, 4);
2895 	ept_reserved_bit_at_level_nohuge(3, 5);
2896 	ept_reserved_bit_at_level_nohuge(3, 6);
2897 	/* 1G alignment. */
2898 	for (i = 12; i < 29; i++) {
2899 		report_prefix_pushf("reserved_bit=%d", i);
2900 		ept_reserved_bit_at_level_huge(3, i);
2901 		report_prefix_pop();
2902 	}
2903 	ept_reserved_bit_at_level(4, 3);
2904 	ept_reserved_bit_at_level(4, 4);
2905 	ept_reserved_bit_at_level(4, 5);
2906 	ept_reserved_bit_at_level(4, 6);
2907 	ept_reserved_bit_at_level(4, 7);
2908 }
2909 
2910 static void ept_access_test_ignored_bits(void)
2911 {
2912 	ept_access_test_setup();
2913 	/*
2914 	 * Bits ignored at every level. Bits 8 and 9 (A and D) are ignored as
2915 	 * far as translation is concerned even if AD bits are enabled in the
2916 	 * EPTP. Bit 63 is ignored because "EPT-violation #VE" VM-execution
2917 	 * control is 0.
2918 	 */
2919 	ept_ignored_bit(8);
2920 	ept_ignored_bit(9);
2921 	ept_ignored_bit(10);
2922 	ept_ignored_bit(11);
2923 	ept_ignored_bit(52);
2924 	ept_ignored_bit(53);
2925 	ept_ignored_bit(54);
2926 	ept_ignored_bit(55);
2927 	ept_ignored_bit(56);
2928 	ept_ignored_bit(57);
2929 	ept_ignored_bit(58);
2930 	ept_ignored_bit(59);
2931 	ept_ignored_bit(60);
2932 	ept_ignored_bit(61);
2933 	ept_ignored_bit(62);
2934 	ept_ignored_bit(63);
2935 }
2936 
2937 static void ept_access_test_paddr_not_present_ad_disabled(void)
2938 {
2939 	ept_access_test_setup();
2940 	ept_disable_ad_bits();
2941 
2942 	ept_access_violation_paddr(0, PT_AD_MASK, OP_READ, EPT_VLT_RD);
2943 	ept_access_violation_paddr(0, PT_AD_MASK, OP_WRITE, EPT_VLT_RD);
2944 	ept_access_violation_paddr(0, PT_AD_MASK, OP_EXEC, EPT_VLT_RD);
2945 }
2946 
2947 static void ept_access_test_paddr_not_present_ad_enabled(void)
2948 {
2949 	u64 qual = EPT_VLT_RD | EPT_VLT_WR;
2950 
2951 	ept_access_test_setup();
2952 	ept_enable_ad_bits_or_skip_test();
2953 
2954 	ept_access_violation_paddr(0, PT_AD_MASK, OP_READ, qual);
2955 	ept_access_violation_paddr(0, PT_AD_MASK, OP_WRITE, qual);
2956 	ept_access_violation_paddr(0, PT_AD_MASK, OP_EXEC, qual);
2957 }
2958 
2959 static void ept_access_test_paddr_read_only_ad_disabled(void)
2960 {
2961 	/*
2962 	 * When EPT AD bits are disabled, all accesses to guest paging
2963 	 * structures are reported separately as a read and (after
2964 	 * translation of the GPA to host physical address) a read+write
2965 	 * if the A/D bits have to be set.
2966 	 */
2967 	u64 qual = EPT_VLT_WR | EPT_VLT_RD | EPT_VLT_PERM_RD;
2968 
2969 	ept_access_test_setup();
2970 	ept_disable_ad_bits();
2971 
2972 	/* Can't update A bit, so all accesses fail. */
2973 	ept_access_violation_paddr(EPT_RA, 0, OP_READ, qual);
2974 	ept_access_violation_paddr(EPT_RA, 0, OP_WRITE, qual);
2975 	ept_access_violation_paddr(EPT_RA, 0, OP_EXEC, qual);
2976 	/* AD bits disabled, so only writes try to update the D bit. */
2977 	ept_access_allowed_paddr(EPT_RA, PT_ACCESSED_MASK, OP_READ);
2978 	ept_access_violation_paddr(EPT_RA, PT_ACCESSED_MASK, OP_WRITE, qual);
2979 	ept_access_allowed_paddr(EPT_RA, PT_ACCESSED_MASK, OP_EXEC);
2980 	/* Both A and D already set, so read-only is OK. */
2981 	ept_access_allowed_paddr(EPT_RA, PT_AD_MASK, OP_READ);
2982 	ept_access_allowed_paddr(EPT_RA, PT_AD_MASK, OP_WRITE);
2983 	ept_access_allowed_paddr(EPT_RA, PT_AD_MASK, OP_EXEC);
2984 }
2985 
2986 static void ept_access_test_paddr_read_only_ad_enabled(void)
2987 {
2988 	/*
2989 	 * When EPT AD bits are enabled, all accesses to guest paging
2990 	 * structures are considered writes as far as EPT translation
2991 	 * is concerned.
2992 	 */
2993 	u64 qual = EPT_VLT_WR | EPT_VLT_RD | EPT_VLT_PERM_RD;
2994 
2995 	ept_access_test_setup();
2996 	ept_enable_ad_bits_or_skip_test();
2997 
2998 	ept_access_violation_paddr(EPT_RA, 0, OP_READ, qual);
2999 	ept_access_violation_paddr(EPT_RA, 0, OP_WRITE, qual);
3000 	ept_access_violation_paddr(EPT_RA, 0, OP_EXEC, qual);
3001 	ept_access_violation_paddr(EPT_RA, PT_ACCESSED_MASK, OP_READ, qual);
3002 	ept_access_violation_paddr(EPT_RA, PT_ACCESSED_MASK, OP_WRITE, qual);
3003 	ept_access_violation_paddr(EPT_RA, PT_ACCESSED_MASK, OP_EXEC, qual);
3004 	ept_access_violation_paddr(EPT_RA, PT_AD_MASK, OP_READ, qual);
3005 	ept_access_violation_paddr(EPT_RA, PT_AD_MASK, OP_WRITE, qual);
3006 	ept_access_violation_paddr(EPT_RA, PT_AD_MASK, OP_EXEC, qual);
3007 }
3008 
3009 static void ept_access_test_paddr_read_write(void)
3010 {
3011 	ept_access_test_setup();
3012 	/* Read-write access to paging structure. */
3013 	ept_access_allowed_paddr(EPT_RA | EPT_WA, 0, OP_READ);
3014 	ept_access_allowed_paddr(EPT_RA | EPT_WA, 0, OP_WRITE);
3015 	ept_access_allowed_paddr(EPT_RA | EPT_WA, 0, OP_EXEC);
3016 }
3017 
3018 static void ept_access_test_paddr_read_write_execute(void)
3019 {
3020 	ept_access_test_setup();
3021 	/* RWX access to paging structure. */
3022 	ept_access_allowed_paddr(EPT_PRESENT, 0, OP_READ);
3023 	ept_access_allowed_paddr(EPT_PRESENT, 0, OP_WRITE);
3024 	ept_access_allowed_paddr(EPT_PRESENT, 0, OP_EXEC);
3025 }
3026 
3027 static void ept_access_test_paddr_read_execute_ad_disabled(void)
3028 {
3029   	/*
3030 	 * When EPT AD bits are disabled, all accesses to guest paging
3031 	 * structures are reported separately as a read and (after
3032 	 * translation of the GPA to host physical address) a read+write
3033 	 * if the A/D bits have to be set.
3034 	 */
3035 	u64 qual = EPT_VLT_WR | EPT_VLT_RD | EPT_VLT_PERM_RD | EPT_VLT_PERM_EX;
3036 
3037 	ept_access_test_setup();
3038 	ept_disable_ad_bits();
3039 
3040 	/* Can't update A bit, so all accesses fail. */
3041 	ept_access_violation_paddr(EPT_RA | EPT_EA, 0, OP_READ, qual);
3042 	ept_access_violation_paddr(EPT_RA | EPT_EA, 0, OP_WRITE, qual);
3043 	ept_access_violation_paddr(EPT_RA | EPT_EA, 0, OP_EXEC, qual);
3044 	/* AD bits disabled, so only writes try to update the D bit. */
3045 	ept_access_allowed_paddr(EPT_RA | EPT_EA, PT_ACCESSED_MASK, OP_READ);
3046 	ept_access_violation_paddr(EPT_RA | EPT_EA, PT_ACCESSED_MASK, OP_WRITE, qual);
3047 	ept_access_allowed_paddr(EPT_RA | EPT_EA, PT_ACCESSED_MASK, OP_EXEC);
3048 	/* Both A and D already set, so read-only is OK. */
3049 	ept_access_allowed_paddr(EPT_RA | EPT_EA, PT_AD_MASK, OP_READ);
3050 	ept_access_allowed_paddr(EPT_RA | EPT_EA, PT_AD_MASK, OP_WRITE);
3051 	ept_access_allowed_paddr(EPT_RA | EPT_EA, PT_AD_MASK, OP_EXEC);
3052 }
3053 
3054 static void ept_access_test_paddr_read_execute_ad_enabled(void)
3055 {
3056 	/*
3057 	 * When EPT AD bits are enabled, all accesses to guest paging
3058 	 * structures are considered writes as far as EPT translation
3059 	 * is concerned.
3060 	 */
3061 	u64 qual = EPT_VLT_WR | EPT_VLT_RD | EPT_VLT_PERM_RD | EPT_VLT_PERM_EX;
3062 
3063 	ept_access_test_setup();
3064 	ept_enable_ad_bits_or_skip_test();
3065 
3066 	ept_access_violation_paddr(EPT_RA | EPT_EA, 0, OP_READ, qual);
3067 	ept_access_violation_paddr(EPT_RA | EPT_EA, 0, OP_WRITE, qual);
3068 	ept_access_violation_paddr(EPT_RA | EPT_EA, 0, OP_EXEC, qual);
3069 	ept_access_violation_paddr(EPT_RA | EPT_EA, PT_ACCESSED_MASK, OP_READ, qual);
3070 	ept_access_violation_paddr(EPT_RA | EPT_EA, PT_ACCESSED_MASK, OP_WRITE, qual);
3071 	ept_access_violation_paddr(EPT_RA | EPT_EA, PT_ACCESSED_MASK, OP_EXEC, qual);
3072 	ept_access_violation_paddr(EPT_RA | EPT_EA, PT_AD_MASK, OP_READ, qual);
3073 	ept_access_violation_paddr(EPT_RA | EPT_EA, PT_AD_MASK, OP_WRITE, qual);
3074 	ept_access_violation_paddr(EPT_RA | EPT_EA, PT_AD_MASK, OP_EXEC, qual);
3075 }
3076 
3077 static void ept_access_test_paddr_not_present_page_fault(void)
3078 {
3079 	ept_access_test_setup();
3080 	/*
3081 	 * TODO: test no EPT violation as long as guest PF occurs. e.g., GPA is
3082 	 * page is read-only in EPT but GVA is also mapped read only in PT.
3083 	 * Thus guest page fault before host takes EPT violation for trying to
3084 	 * update A bit.
3085 	 */
3086 }
3087 
3088 static void ept_access_test_force_2m_page(void)
3089 {
3090 	ept_access_test_setup();
3091 
3092 	TEST_ASSERT_EQ(ept_2m_supported(), true);
3093 	ept_allowed_at_level_mkhuge(true, 2, 0, 0, OP_READ);
3094 	ept_violation_at_level_mkhuge(true, 2, EPT_PRESENT, EPT_RA, OP_WRITE,
3095 				      EPT_VLT_WR | EPT_VLT_PERM_RD |
3096 				      EPT_VLT_LADDR_VLD | EPT_VLT_PADDR);
3097 	ept_misconfig_at_level_mkhuge(true, 2, EPT_PRESENT, EPT_WA);
3098 }
3099 
3100 static bool invvpid_valid(u64 type, u64 vpid, u64 gla)
3101 {
3102 	u64 msr = rdmsr(MSR_IA32_VMX_EPT_VPID_CAP);
3103 
3104 	TEST_ASSERT(msr & VPID_CAP_INVVPID);
3105 
3106 	if (type < INVVPID_ADDR || type > INVVPID_CONTEXT_LOCAL)
3107 		return false;
3108 
3109 	if (!(msr & (1ull << (type + VPID_CAP_INVVPID_TYPES_SHIFT))))
3110 		return false;
3111 
3112 	if (vpid >> 16)
3113 		return false;
3114 
3115 	if (type != INVVPID_ALL && !vpid)
3116 		return false;
3117 
3118 	if (type == INVVPID_ADDR && !is_canonical(gla))
3119 		return false;
3120 
3121 	return true;
3122 }
3123 
3124 static void try_invvpid(u64 type, u64 vpid, u64 gla)
3125 {
3126 	int rc;
3127 	bool valid = invvpid_valid(type, vpid, gla);
3128 	u64 expected = valid ? VMXERR_UNSUPPORTED_VMCS_COMPONENT
3129 		: VMXERR_INVALID_OPERAND_TO_INVEPT_INVVPID;
3130 	/*
3131 	 * Set VMX_INST_ERROR to VMXERR_UNVALID_VMCS_COMPONENT, so
3132 	 * that we can tell if it is updated by INVVPID.
3133 	 */
3134 	vmcs_read(~0);
3135 	rc = invvpid(type, vpid, gla);
3136 	report("INVVPID type %ld VPID %lx GLA %lx %s",
3137 	       !rc == valid, type, vpid, gla,
3138 	       valid ? "passes" : "fails");
3139 	report("After %s INVVPID, VMX_INST_ERR is %ld (actual %ld)",
3140 	       vmcs_read(VMX_INST_ERROR) == expected,
3141 	       rc ? "failed" : "successful",
3142 	       expected, vmcs_read(VMX_INST_ERROR));
3143 }
3144 
3145 static void ds_invvpid(void *data)
3146 {
3147 	u64 msr = rdmsr(MSR_IA32_VMX_EPT_VPID_CAP);
3148 	u64 type = ffs(msr >> VPID_CAP_INVVPID_TYPES_SHIFT) - 1;
3149 
3150 	TEST_ASSERT(type >= INVVPID_ADDR && type <= INVVPID_CONTEXT_LOCAL);
3151 	asm volatile("invvpid %0, %1"
3152 		     :
3153 		     : "m"(*(struct invvpid_operand *)data),
3154 		       "r"(type));
3155 }
3156 
3157 /*
3158  * The SS override is ignored in 64-bit mode, so we use an addressing
3159  * mode with %rsp as the base register to generate an implicit SS
3160  * reference.
3161  */
3162 static void ss_invvpid(void *data)
3163 {
3164 	u64 msr = rdmsr(MSR_IA32_VMX_EPT_VPID_CAP);
3165 	u64 type = ffs(msr >> VPID_CAP_INVVPID_TYPES_SHIFT) - 1;
3166 
3167 	TEST_ASSERT(type >= INVVPID_ADDR && type <= INVVPID_CONTEXT_LOCAL);
3168 	asm volatile("sub %%rsp,%0; invvpid (%%rsp,%0,1), %1"
3169 		     : "+r"(data)
3170 		     : "r"(type));
3171 }
3172 
3173 static void invvpid_test_gp(void)
3174 {
3175 	bool fault;
3176 
3177 	fault = test_for_exception(GP_VECTOR, &ds_invvpid,
3178 				   (void *)NONCANONICAL);
3179 	report("INVVPID with non-canonical DS operand raises #GP", fault);
3180 }
3181 
3182 static void invvpid_test_ss(void)
3183 {
3184 	bool fault;
3185 
3186 	fault = test_for_exception(SS_VECTOR, &ss_invvpid,
3187 				   (void *)NONCANONICAL);
3188 	report("INVVPID with non-canonical SS operand raises #SS", fault);
3189 }
3190 
3191 static void invvpid_test_pf(void)
3192 {
3193 	void *vpage = alloc_vpage();
3194 	bool fault;
3195 
3196 	fault = test_for_exception(PF_VECTOR, &ds_invvpid, vpage);
3197 	report("INVVPID with unmapped operand raises #PF", fault);
3198 }
3199 
3200 static void try_compat_invvpid(void *unused)
3201 {
3202 	struct far_pointer32 fp = {
3203 		.offset = (uintptr_t)&&invvpid,
3204 		.selector = KERNEL_CS32,
3205 	};
3206 	register uintptr_t rsp asm("rsp");
3207 
3208 	TEST_ASSERT_MSG(fp.offset == (uintptr_t)&&invvpid,
3209 			"Code address too high.");
3210 	TEST_ASSERT_MSG(rsp == (u32)rsp, "Stack address too high.");
3211 
3212 	asm goto ("lcall *%0" : : "m" (fp) : "rax" : invvpid);
3213 	return;
3214 invvpid:
3215 	asm volatile (".code32;"
3216 		      "invvpid (%eax), %eax;"
3217 		      "lret;"
3218 		      ".code64");
3219 	__builtin_unreachable();
3220 }
3221 
3222 static void invvpid_test_compatibility_mode(void)
3223 {
3224 	bool fault;
3225 
3226 	fault = test_for_exception(UD_VECTOR, &try_compat_invvpid, NULL);
3227 	report("Compatibility mode INVVPID raises #UD", fault);
3228 }
3229 
3230 static void invvpid_test_not_in_vmx_operation(void)
3231 {
3232 	bool fault;
3233 
3234 	TEST_ASSERT(!vmx_off());
3235 	fault = test_for_exception(UD_VECTOR, &ds_invvpid, NULL);
3236 	report("INVVPID outside of VMX operation raises #UD", fault);
3237 	TEST_ASSERT(!vmx_on());
3238 }
3239 
3240 /*
3241  * This does not test real-address mode, virtual-8086 mode, protected mode,
3242  * or CPL > 0.
3243  */
3244 static void invvpid_test_v2(void)
3245 {
3246 	u64 msr;
3247 	int i;
3248 	unsigned types = 0;
3249 	unsigned type;
3250 
3251 	if (!(ctrl_cpu_rev[0].clr & CPU_SECONDARY) ||
3252 	    !(ctrl_cpu_rev[1].clr & CPU_VPID))
3253 		test_skip("VPID not supported");
3254 
3255 	msr = rdmsr(MSR_IA32_VMX_EPT_VPID_CAP);
3256 
3257 	if (!(msr & VPID_CAP_INVVPID))
3258 		test_skip("INVVPID not supported.\n");
3259 
3260 	if (msr & VPID_CAP_INVVPID_ADDR)
3261 		types |= 1u << INVVPID_ADDR;
3262 	if (msr & VPID_CAP_INVVPID_CXTGLB)
3263 		types |= 1u << INVVPID_CONTEXT_GLOBAL;
3264 	if (msr & VPID_CAP_INVVPID_ALL)
3265 		types |= 1u << INVVPID_ALL;
3266 	if (msr & VPID_CAP_INVVPID_CXTLOC)
3267 		types |= 1u << INVVPID_CONTEXT_LOCAL;
3268 
3269 	if (!types)
3270 		test_skip("No INVVPID types supported.\n");
3271 
3272 	for (i = -127; i < 128; i++)
3273 		try_invvpid(i, 0xffff, 0);
3274 
3275 	/*
3276 	 * VPID must not be more than 16 bits.
3277 	 */
3278 	for (i = 0; i < 64; i++)
3279 		for (type = 0; type < 4; type++)
3280 			if (types & (1u << type))
3281 				try_invvpid(type, 1ul << i, 0);
3282 
3283 	/*
3284 	 * VPID must not be zero, except for "all contexts."
3285 	 */
3286 	for (type = 0; type < 4; type++)
3287 		if (types & (1u << type))
3288 			try_invvpid(type, 0, 0);
3289 
3290 	/*
3291 	 * The gla operand is only validated for single-address INVVPID.
3292 	 */
3293 	if (types & (1u << INVVPID_ADDR))
3294 		try_invvpid(INVVPID_ADDR, 0xffff, NONCANONICAL);
3295 
3296 	invvpid_test_gp();
3297 	invvpid_test_ss();
3298 	invvpid_test_pf();
3299 	invvpid_test_compatibility_mode();
3300 	invvpid_test_not_in_vmx_operation();
3301 }
3302 
3303 /*
3304  * Test for early VMLAUNCH failure. Returns true if VMLAUNCH makes it
3305  * at least as far as the guest-state checks. Returns false if the
3306  * VMLAUNCH fails early and execution falls through to the next
3307  * instruction.
3308  */
3309 static bool vmlaunch_succeeds(void)
3310 {
3311 	u32 exit_reason;
3312 
3313 	/*
3314 	 * Indirectly set VMX_INST_ERR to 12 ("VMREAD/VMWRITE from/to
3315 	 * unsupported VMCS component"). The caller can then check
3316 	 * to see if a failed VM-entry sets VMX_INST_ERR as expected.
3317 	 */
3318 	vmcs_write(~0u, 0);
3319 
3320 	vmcs_write(HOST_RIP, (uintptr_t)&&success);
3321 	__asm__ __volatile__ goto ("vmwrite %%rsp, %0; vmlaunch"
3322 				   :
3323 				   : "r" ((u64)HOST_RSP)
3324 				   : "cc", "memory"
3325 				   : success);
3326 	return false;
3327 success:
3328 	exit_reason = vmcs_read(EXI_REASON);
3329 	TEST_ASSERT(exit_reason == (VMX_FAIL_STATE | VMX_ENTRY_FAILURE) ||
3330 		    exit_reason == (VMX_FAIL_MSR | VMX_ENTRY_FAILURE));
3331 	return true;
3332 }
3333 
3334 /*
3335  * Try to launch the current VMCS.
3336  */
3337 static void test_vmx_vmlaunch(u32 xerror, bool xfail)
3338 {
3339 	bool success = vmlaunch_succeeds();
3340 	u32 vmx_inst_err;
3341 
3342 	report_xfail("vmlaunch %s", xfail, success == !xerror,
3343 		     !xerror ? "succeeds" : "fails");
3344 	if (!success && xerror) {
3345 		vmx_inst_err = vmcs_read(VMX_INST_ERROR);
3346 		report("VMX inst error is %d (actual %d)",
3347 		       vmx_inst_err == xerror, xerror, vmx_inst_err);
3348 	}
3349 }
3350 
3351 static void test_vmx_invalid_controls(bool xfail)
3352 {
3353 	test_vmx_vmlaunch(VMXERR_ENTRY_INVALID_CONTROL_FIELD, xfail);
3354 }
3355 
3356 static void test_vmx_valid_controls(bool xfail)
3357 {
3358 	test_vmx_vmlaunch(0, xfail);
3359 }
3360 
3361 /*
3362  * Test a particular value of a VM-execution control bit, if the value
3363  * is required or if the value is zero.
3364  */
3365 static void test_rsvd_ctl_bit_value(const char *name, union vmx_ctrl_msr msr,
3366 				    enum Encoding encoding, unsigned bit,
3367 				    unsigned val)
3368 {
3369 	u32 mask = 1u << bit;
3370 	bool expected;
3371 	u32 controls;
3372 
3373 	if (msr.set & mask)
3374 		TEST_ASSERT(msr.clr & mask);
3375 
3376 	/*
3377 	 * We can't arbitrarily turn on a control bit, because it may
3378 	 * introduce dependencies on other VMCS fields. So, we only
3379 	 * test turning on bits that have a required setting.
3380 	 */
3381 	if (val && (msr.clr & mask) && !(msr.set & mask))
3382 		return;
3383 
3384 	report_prefix_pushf("%s %s bit %d",
3385 			    val ? "Set" : "Clear", name, bit);
3386 
3387 	controls = vmcs_read(encoding);
3388 	if (val) {
3389 		vmcs_write(encoding, msr.set | mask);
3390 		expected = (msr.clr & mask);
3391 	} else {
3392 		vmcs_write(encoding, msr.set & ~mask);
3393 		expected = !(msr.set & mask);
3394 	}
3395 	if (expected)
3396 		test_vmx_valid_controls(false);
3397 	else
3398 		test_vmx_invalid_controls(false);
3399 	vmcs_write(encoding, controls);
3400 	report_prefix_pop();
3401 }
3402 
3403 /*
3404  * Test reserved values of a VM-execution control bit, based on the
3405  * allowed bit settings from the corresponding VMX capability MSR.
3406  */
3407 static void test_rsvd_ctl_bit(const char *name, union vmx_ctrl_msr msr,
3408 			      enum Encoding encoding, unsigned bit)
3409 {
3410 	test_rsvd_ctl_bit_value(name, msr, encoding, bit, 0);
3411 	test_rsvd_ctl_bit_value(name, msr, encoding, bit, 1);
3412 }
3413 
3414 /*
3415  * Reserved bits in the pin-based VM-execution controls must be set
3416  * properly. Software may consult the VMX capability MSRs to determine
3417  * the proper settings.
3418  * [Intel SDM]
3419  */
3420 static void test_pin_based_ctls(void)
3421 {
3422 	unsigned bit;
3423 
3424 	printf("%s: %lx\n", basic.ctrl ? "MSR_IA32_VMX_TRUE_PIN" :
3425 	       "MSR_IA32_VMX_PINBASED_CTLS", ctrl_pin_rev.val);
3426 	for (bit = 0; bit < 32; bit++)
3427 		test_rsvd_ctl_bit("pin-based controls",
3428 				  ctrl_pin_rev, PIN_CONTROLS, bit);
3429 }
3430 
3431 /*
3432  * Reserved bits in the primary processor-based VM-execution controls
3433  * must be set properly. Software may consult the VMX capability MSRs
3434  * to determine the proper settings.
3435  * [Intel SDM]
3436  */
3437 static void test_primary_processor_based_ctls(void)
3438 {
3439 	unsigned bit;
3440 
3441 	printf("\n%s: %lx\n", basic.ctrl ? "MSR_IA32_VMX_TRUE_PROC" :
3442 	       "MSR_IA32_VMX_PROCBASED_CTLS", ctrl_cpu_rev[0].val);
3443 	for (bit = 0; bit < 32; bit++)
3444 		test_rsvd_ctl_bit("primary processor-based controls",
3445 				  ctrl_cpu_rev[0], CPU_EXEC_CTRL0, bit);
3446 }
3447 
3448 /*
3449  * If the "activate secondary controls" primary processor-based
3450  * VM-execution control is 1, reserved bits in the secondary
3451  * processor-based VM-execution controls must be cleared. Software may
3452  * consult the VMX capability MSRs to determine which bits are
3453  * reserved.
3454  * If the "activate secondary controls" primary processor-based
3455  * VM-execution control is 0 (or if the processor does not support the
3456  * 1-setting of that control), no checks are performed on the
3457  * secondary processor-based VM-execution controls.
3458  * [Intel SDM]
3459  */
3460 static void test_secondary_processor_based_ctls(void)
3461 {
3462 	u32 primary;
3463 	u32 secondary;
3464 	unsigned bit;
3465 
3466 	if (!(ctrl_cpu_rev[0].clr & CPU_SECONDARY))
3467 		return;
3468 
3469 	primary = vmcs_read(CPU_EXEC_CTRL0);
3470 	secondary = vmcs_read(CPU_EXEC_CTRL1);
3471 
3472 	vmcs_write(CPU_EXEC_CTRL0, primary | CPU_SECONDARY);
3473 	printf("\nMSR_IA32_VMX_PROCBASED_CTLS2: %lx\n", ctrl_cpu_rev[1].val);
3474 	for (bit = 0; bit < 32; bit++)
3475 		test_rsvd_ctl_bit("secondary processor-based controls",
3476 				  ctrl_cpu_rev[1], CPU_EXEC_CTRL1, bit);
3477 
3478 	/*
3479 	 * When the "activate secondary controls" VM-execution control
3480 	 * is clear, there are no checks on the secondary controls.
3481 	 */
3482 	vmcs_write(CPU_EXEC_CTRL0, primary & ~CPU_SECONDARY);
3483 	vmcs_write(CPU_EXEC_CTRL1, ~0);
3484 	report("Secondary processor-based controls ignored",
3485 	       vmlaunch_succeeds());
3486 	vmcs_write(CPU_EXEC_CTRL1, secondary);
3487 	vmcs_write(CPU_EXEC_CTRL0, primary);
3488 }
3489 
3490 static void try_cr3_target_count(unsigned i, unsigned max)
3491 {
3492 	report_prefix_pushf("CR3 target count 0x%x", i);
3493 	vmcs_write(CR3_TARGET_COUNT, i);
3494 	if (i <= max)
3495 		test_vmx_valid_controls(false);
3496 	else
3497 		test_vmx_invalid_controls(false);
3498 	report_prefix_pop();
3499 }
3500 
3501 /*
3502  * The CR3-target count must not be greater than 4. Future processors
3503  * may support a different number of CR3-target values. Software
3504  * should read the VMX capability MSR IA32_VMX_MISC to determine the
3505  * number of values supported.
3506  * [Intel SDM]
3507  */
3508 static void test_cr3_targets(void)
3509 {
3510 	unsigned supported_targets = (rdmsr(MSR_IA32_VMX_MISC) >> 16) & 0x1ff;
3511 	u32 cr3_targets = vmcs_read(CR3_TARGET_COUNT);
3512 	unsigned i;
3513 
3514 	printf("\nSupported CR3 targets: %d\n", supported_targets);
3515 	TEST_ASSERT(supported_targets <= 256);
3516 
3517 	try_cr3_target_count(-1u, supported_targets);
3518 	try_cr3_target_count(0x80000000, supported_targets);
3519 	try_cr3_target_count(0x7fffffff, supported_targets);
3520 	for (i = 0; i <= supported_targets + 1; i++)
3521 		try_cr3_target_count(i, supported_targets);
3522 	vmcs_write(CR3_TARGET_COUNT, cr3_targets);
3523 }
3524 
3525 /*
3526  * Test a particular address setting in the VMCS
3527  */
3528 static void test_vmcs_addr(const char *name,
3529 			   enum Encoding encoding,
3530 			   u64 align,
3531 			   bool ignored,
3532 			   bool xfail_beyond_mapped_ram,
3533 			   u64 addr)
3534 {
3535 	bool xfail =
3536 		(xfail_beyond_mapped_ram &&
3537 		 addr > fwcfg_get_u64(FW_CFG_RAM_SIZE) - align &&
3538 		 addr < (1ul << cpuid_maxphyaddr()));
3539 
3540 	report_prefix_pushf("%s = %lx", name, addr);
3541 	vmcs_write(encoding, addr);
3542 	if (ignored || (IS_ALIGNED(addr, align) &&
3543 	    addr < (1ul << cpuid_maxphyaddr())))
3544 		test_vmx_valid_controls(xfail);
3545 	else
3546 		test_vmx_invalid_controls(xfail);
3547 	report_prefix_pop();
3548 	xfail = false;
3549 }
3550 
3551 /*
3552  * Test interesting values for a VMCS address
3553  */
3554 static void test_vmcs_addr_values(const char *name,
3555 				  enum Encoding encoding,
3556 				  u64 align,
3557 				  bool ignored,
3558 				  bool xfail_beyond_mapped_ram,
3559 				  u32 bit_start, u32 bit_end)
3560 {
3561 	unsigned i;
3562 	u64 orig_val = vmcs_read(encoding);
3563 
3564 	for (i = bit_start; i <= bit_end; i++)
3565 		test_vmcs_addr(name, encoding, align, ignored,
3566 			       xfail_beyond_mapped_ram, 1ul << i);
3567 
3568 	test_vmcs_addr(name, encoding, align, ignored,
3569 		       xfail_beyond_mapped_ram, PAGE_SIZE - 1);
3570 	test_vmcs_addr(name, encoding, align, ignored,
3571 		       xfail_beyond_mapped_ram, PAGE_SIZE);
3572 	test_vmcs_addr(name, encoding, align, ignored,
3573 		       xfail_beyond_mapped_ram,
3574 		      (1ul << cpuid_maxphyaddr()) - PAGE_SIZE);
3575 	test_vmcs_addr(name, encoding, align, ignored,
3576 		       xfail_beyond_mapped_ram, -1ul);
3577 
3578 	vmcs_write(encoding, orig_val);
3579 }
3580 
3581 /*
3582  * Test a physical address reference in the VMCS, when the corresponding
3583  * feature is enabled and when the corresponding feature is disabled.
3584  */
3585 static void test_vmcs_addr_reference(u32 control_bit, enum Encoding field,
3586 				     const char *field_name,
3587 				     const char *control_name, u64 align,
3588 				     bool xfail_beyond_mapped_ram,
3589 				     bool control_primary)
3590 {
3591 	u32 primary = vmcs_read(CPU_EXEC_CTRL0);
3592 	u32 secondary = vmcs_read(CPU_EXEC_CTRL1);
3593 	u64 page_addr;
3594 
3595 	if (control_primary) {
3596 		if (!(ctrl_cpu_rev[0].clr & control_bit))
3597 			return;
3598 	} else {
3599 		if (!(ctrl_cpu_rev[1].clr & control_bit))
3600 			return;
3601 	}
3602 
3603 	page_addr = vmcs_read(field);
3604 
3605 	report_prefix_pushf("%s enabled", control_name);
3606 	if (control_primary) {
3607 		vmcs_write(CPU_EXEC_CTRL0, primary | control_bit);
3608 	} else {
3609 		vmcs_write(CPU_EXEC_CTRL0, primary | CPU_SECONDARY);
3610 		vmcs_write(CPU_EXEC_CTRL1, secondary | control_bit);
3611 	}
3612 
3613 	test_vmcs_addr_values(field_name, field, align, false,
3614 			      xfail_beyond_mapped_ram, 0, 63);
3615 	report_prefix_pop();
3616 
3617 	report_prefix_pushf("%s disabled", control_name);
3618 	if (control_primary) {
3619 		vmcs_write(CPU_EXEC_CTRL0, primary & ~control_bit);
3620 	} else {
3621 		vmcs_write(CPU_EXEC_CTRL0, primary & ~CPU_SECONDARY);
3622 		vmcs_write(CPU_EXEC_CTRL1, secondary & ~control_bit);
3623 	}
3624 
3625 	test_vmcs_addr_values(field_name, field, align, true, false, 0, 63);
3626 	report_prefix_pop();
3627 
3628 	vmcs_write(field, page_addr);
3629 	vmcs_write(CPU_EXEC_CTRL0, primary);
3630 }
3631 
3632 /*
3633  * If the "use I/O bitmaps" VM-execution control is 1, bits 11:0 of
3634  * each I/O-bitmap address must be 0. Neither address should set any
3635  * bits beyond the processor's physical-address width.
3636  * [Intel SDM]
3637  */
3638 static void test_io_bitmaps(void)
3639 {
3640 	test_vmcs_addr_reference(CPU_IO_BITMAP, IO_BITMAP_A,
3641 				 "I/O bitmap A", "Use I/O bitmaps",
3642 				 PAGE_SIZE, false, true);
3643 	test_vmcs_addr_reference(CPU_IO_BITMAP, IO_BITMAP_B,
3644 				 "I/O bitmap B", "Use I/O bitmaps",
3645 				 PAGE_SIZE, false, true);
3646 }
3647 
3648 /*
3649  * If the "use MSR bitmaps" VM-execution control is 1, bits 11:0 of
3650  * the MSR-bitmap address must be 0. The address should not set any
3651  * bits beyond the processor's physical-address width.
3652  * [Intel SDM]
3653  */
3654 static void test_msr_bitmap(void)
3655 {
3656 	test_vmcs_addr_reference(CPU_MSR_BITMAP, MSR_BITMAP,
3657 				 "MSR bitmap", "Use MSR bitmaps",
3658 				 PAGE_SIZE, false, true);
3659 }
3660 
3661 /*
3662  * If the "use TPR shadow" VM-execution control is 1, the virtual-APIC
3663  * address must satisfy the following checks:
3664  * - Bits 11:0 of the address must be 0.
3665  * - The address should not set any bits beyond the processor's
3666  *   physical-address width.
3667  * [Intel SDM]
3668  */
3669 static void test_apic_virt_addr(void)
3670 {
3671 	test_vmcs_addr_reference(CPU_TPR_SHADOW, APIC_VIRT_ADDR,
3672 				 "virtual-APIC address", "Use TPR shadow",
3673 				 PAGE_SIZE, true, true);
3674 }
3675 
3676 /*
3677  * If the "virtualize APIC-accesses" VM-execution control is 1, the
3678  * APIC-access address must satisfy the following checks:
3679  *  - Bits 11:0 of the address must be 0.
3680  *  - The address should not set any bits beyond the processor's
3681  *    physical-address width.
3682  * [Intel SDM]
3683  */
3684 static void test_apic_access_addr(void)
3685 {
3686 	void *apic_access_page = alloc_page();
3687 
3688 	vmcs_write(APIC_ACCS_ADDR, virt_to_phys(apic_access_page));
3689 
3690 	test_vmcs_addr_reference(CPU_VIRT_APIC_ACCESSES, APIC_ACCS_ADDR,
3691 				 "APIC-access address",
3692 				 "virtualize APIC-accesses", PAGE_SIZE,
3693 				 false, false);
3694 }
3695 
3696 static bool set_bit_pattern(u8 mask, u32 *secondary)
3697 {
3698 	u8 i;
3699 	bool flag = false;
3700 	u32 test_bits[3] = {
3701 		CPU_VIRT_X2APIC,
3702 		CPU_APIC_REG_VIRT,
3703 		CPU_VINTD
3704 	};
3705 
3706         for (i = 0; i < ARRAY_SIZE(test_bits); i++) {
3707 		if ((mask & (1u << i)) &&
3708 		    (ctrl_cpu_rev[1].clr & test_bits[i])) {
3709 			*secondary |= test_bits[i];
3710 			flag = true;
3711 		}
3712 	}
3713 
3714 	return (flag);
3715 }
3716 
3717 /*
3718  * If the "use TPR shadow" VM-execution control is 0, the following
3719  * VM-execution controls must also be 0:
3720  * 	- virtualize x2APIC mode
3721  *	- APIC-register virtualization
3722  *	- virtual-interrupt delivery
3723  *    [Intel SDM]
3724  *
3725  * 2. If the "virtualize x2APIC mode" VM-execution control is 1, the
3726  *    "virtualize APIC accesses" VM-execution control must be 0.
3727  *    [Intel SDM]
3728  */
3729 static void test_apic_virtual_ctls(void)
3730 {
3731 	u32 saved_primary = vmcs_read(CPU_EXEC_CTRL0);
3732 	u32 saved_secondary = vmcs_read(CPU_EXEC_CTRL1);
3733 	u32 primary = saved_primary;
3734 	u32 secondary = saved_secondary;
3735 	bool ctrl = false;
3736 	char str[10] = "disabled";
3737 	u8 i = 0, j;
3738 
3739 	/*
3740 	 * First test
3741 	 */
3742 	if (!((ctrl_cpu_rev[0].clr & (CPU_SECONDARY | CPU_TPR_SHADOW)) ==
3743 	    (CPU_SECONDARY | CPU_TPR_SHADOW)))
3744 		return;
3745 
3746 	primary |= CPU_SECONDARY;
3747 	primary &= ~CPU_TPR_SHADOW;
3748 	vmcs_write(CPU_EXEC_CTRL0, primary);
3749 
3750 	while (1) {
3751 		for (j = 1; j < 8; j++) {
3752 			secondary &= ~(CPU_VIRT_X2APIC | CPU_APIC_REG_VIRT | CPU_VINTD);
3753 			if (primary & CPU_TPR_SHADOW) {
3754 				ctrl = true;
3755 			} else {
3756 				if (! set_bit_pattern(j, &secondary))
3757 					ctrl = true;
3758 				else
3759 					ctrl = false;
3760 			}
3761 
3762 			vmcs_write(CPU_EXEC_CTRL1, secondary);
3763 			report_prefix_pushf("Use TPR shadow %s, virtualize x2APIC mode %s, APIC-register virtualization %s, virtual-interrupt delivery %s",
3764 				str, (secondary & CPU_VIRT_X2APIC) ? "enabled" : "disabled", (secondary & CPU_APIC_REG_VIRT) ? "enabled" : "disabled", (secondary & CPU_VINTD) ? "enabled" : "disabled");
3765 			if (ctrl)
3766 				test_vmx_valid_controls(false);
3767 			else
3768 				test_vmx_invalid_controls(false);
3769 			report_prefix_pop();
3770 		}
3771 
3772 		if (i == 1)
3773 			break;
3774 		i++;
3775 
3776 		primary |= CPU_TPR_SHADOW;
3777 		vmcs_write(CPU_EXEC_CTRL0, primary);
3778 		strcpy(str, "enabled");
3779 	}
3780 
3781 	/*
3782 	 * Second test
3783 	 */
3784 	u32 apic_virt_ctls = (CPU_VIRT_X2APIC | CPU_VIRT_APIC_ACCESSES);
3785 
3786 	primary = saved_primary;
3787 	secondary = saved_secondary;
3788 	if (!((ctrl_cpu_rev[1].clr & apic_virt_ctls) == apic_virt_ctls))
3789 		return;
3790 
3791 	vmcs_write(CPU_EXEC_CTRL0, primary | CPU_SECONDARY);
3792 	secondary &= ~CPU_VIRT_APIC_ACCESSES;
3793 	vmcs_write(CPU_EXEC_CTRL1, secondary & ~CPU_VIRT_X2APIC);
3794 	report_prefix_pushf("Virtualize x2APIC mode disabled; virtualize APIC access disabled");
3795 	test_vmx_valid_controls(false);
3796 	report_prefix_pop();
3797 
3798 	vmcs_write(CPU_EXEC_CTRL1, secondary | CPU_VIRT_APIC_ACCESSES);
3799 	report_prefix_pushf("Virtualize x2APIC mode disabled; virtualize APIC access enabled");
3800 	test_vmx_valid_controls(false);
3801 	report_prefix_pop();
3802 
3803 	vmcs_write(CPU_EXEC_CTRL1, secondary | CPU_VIRT_X2APIC);
3804 	report_prefix_pushf("Virtualize x2APIC mode enabled; virtualize APIC access enabled");
3805 	test_vmx_invalid_controls(false);
3806 	report_prefix_pop();
3807 
3808 	vmcs_write(CPU_EXEC_CTRL1, secondary & ~CPU_VIRT_APIC_ACCESSES);
3809 	report_prefix_pushf("Virtualize x2APIC mode enabled; virtualize APIC access disabled");
3810 	test_vmx_valid_controls(false);
3811 	report_prefix_pop();
3812 
3813 	vmcs_write(CPU_EXEC_CTRL0, saved_primary);
3814 	vmcs_write(CPU_EXEC_CTRL1, saved_secondary);
3815 }
3816 
3817 /*
3818  * If the "virtual-interrupt delivery" VM-execution control is 1, the
3819  * "external-interrupt exiting" VM-execution control must be 1.
3820  * [Intel SDM]
3821  */
3822 static void test_virtual_intr_ctls(void)
3823 {
3824 	u32 saved_primary = vmcs_read(CPU_EXEC_CTRL0);
3825 	u32 saved_secondary = vmcs_read(CPU_EXEC_CTRL1);
3826 	u32 saved_pin = vmcs_read(PIN_CONTROLS);
3827 	u32 primary = saved_primary;
3828 	u32 secondary = saved_secondary;
3829 	u32 pin = saved_pin;
3830 
3831 	if (!((ctrl_cpu_rev[1].clr & CPU_VINTD) &&
3832 	    (ctrl_pin_rev.clr & PIN_EXTINT)))
3833 		return;
3834 
3835 	vmcs_write(CPU_EXEC_CTRL0, primary | CPU_SECONDARY | CPU_TPR_SHADOW);
3836 	vmcs_write(CPU_EXEC_CTRL1, secondary & ~CPU_VINTD);
3837 	vmcs_write(PIN_CONTROLS, pin & ~PIN_EXTINT);
3838 	report_prefix_pushf("Virtualize interrupt-delivery disabled; external-interrupt exiting disabled");
3839 	test_vmx_valid_controls(false);
3840 	report_prefix_pop();
3841 
3842 	vmcs_write(CPU_EXEC_CTRL1, secondary | CPU_VINTD);
3843 	report_prefix_pushf("Virtualize interrupt-delivery enabled; external-interrupt exiting disabled");
3844 	test_vmx_invalid_controls(false);
3845 	report_prefix_pop();
3846 
3847 	vmcs_write(PIN_CONTROLS, pin | PIN_EXTINT);
3848 	report_prefix_pushf("Virtualize interrupt-delivery enabled; external-interrupt exiting enabled");
3849 	test_vmx_valid_controls(false);
3850 	report_prefix_pop();
3851 
3852 	vmcs_write(PIN_CONTROLS, pin & ~PIN_EXTINT);
3853 	report_prefix_pushf("Virtualize interrupt-delivery enabled; external-interrupt exiting disabled");
3854 	test_vmx_invalid_controls(false);
3855 	report_prefix_pop();
3856 
3857 	vmcs_write(CPU_EXEC_CTRL0, saved_primary);
3858 	vmcs_write(CPU_EXEC_CTRL1, saved_secondary);
3859 	vmcs_write(PIN_CONTROLS, saved_pin);
3860 }
3861 
3862 static void test_pi_desc_addr(u64 addr, bool ctrl)
3863 {
3864 	vmcs_write(POSTED_INTR_DESC_ADDR, addr);
3865 	report_prefix_pushf("Process-posted-interrupts enabled; posted-interrupt-descriptor-address 0x%lx", addr);
3866 	if (ctrl)
3867 		test_vmx_valid_controls(false);
3868 	else
3869 		test_vmx_invalid_controls(false);
3870 	report_prefix_pop();
3871 }
3872 
3873 /*
3874  * If the “process posted interrupts†VM-execution control is 1, the
3875  * following must be true:
3876  *
3877  *	- The “virtual-interrupt delivery†VM-execution control is 1.
3878  *	- The “acknowledge interrupt on exit†VM-exit control is 1.
3879  *	- The posted-interrupt notification vector has a value in the
3880  *	- range 0–255 (bits 15:8 are all 0).
3881  *	- Bits 5:0 of the posted-interrupt descriptor address are all 0.
3882  *	- The posted-interrupt descriptor address does not set any bits
3883  *	  beyond the processor's physical-address width.
3884  * [Intel SDM]
3885  */
3886 static void test_posted_intr(void)
3887 {
3888 	u32 saved_primary = vmcs_read(CPU_EXEC_CTRL0);
3889 	u32 saved_secondary = vmcs_read(CPU_EXEC_CTRL1);
3890 	u32 saved_pin = vmcs_read(PIN_CONTROLS);
3891 	u32 exit_ctl_saved = vmcs_read(EXI_CONTROLS);
3892 	u32 primary = saved_primary;
3893 	u32 secondary = saved_secondary;
3894 	u32 pin = saved_pin;
3895 	u32 exit_ctl = exit_ctl_saved;
3896 	u16 vec;
3897 	int i;
3898 
3899 	if (!((ctrl_pin_rev.clr & PIN_POST_INTR) &&
3900 	    (ctrl_cpu_rev[1].clr & CPU_VINTD) &&
3901 	    (ctrl_exit_rev.clr & EXI_INTA)))
3902 		return;
3903 
3904 	vmcs_write(CPU_EXEC_CTRL0, primary | CPU_SECONDARY | CPU_TPR_SHADOW);
3905 
3906 	/*
3907 	 * Test virtual-interrupt-delivery and acknowledge-interrupt-on-exit
3908 	 */
3909 	pin |= PIN_POST_INTR;
3910 	vmcs_write(PIN_CONTROLS, pin);
3911 	secondary &= ~CPU_VINTD;
3912 	vmcs_write(CPU_EXEC_CTRL1, secondary);
3913 	report_prefix_pushf("Process-posted-interrupts enabled; virtual-interrupt-delivery disabled");
3914 	test_vmx_invalid_controls(false);
3915 	report_prefix_pop();
3916 
3917 	secondary |= CPU_VINTD;
3918 	vmcs_write(CPU_EXEC_CTRL1, secondary);
3919 	report_prefix_pushf("Process-posted-interrupts enabled; virtual-interrupt-delivery enabled");
3920 	test_vmx_invalid_controls(false);
3921 	report_prefix_pop();
3922 
3923 	exit_ctl &= ~EXI_INTA;
3924 	vmcs_write(EXI_CONTROLS, exit_ctl);
3925 	report_prefix_pushf("Process-posted-interrupts enabled; virtual-interrupt-delivery enabled; acknowledge-interrupt-on-exit disabled");
3926 	test_vmx_invalid_controls(false);
3927 	report_prefix_pop();
3928 
3929 	exit_ctl |= EXI_INTA;
3930 	vmcs_write(EXI_CONTROLS, exit_ctl);
3931 	report_prefix_pushf("Process-posted-interrupts enabled; virtual-interrupt-delivery enabled; acknowledge-interrupt-on-exit enabled");
3932 	test_vmx_valid_controls(false);
3933 	report_prefix_pop();
3934 
3935 	secondary &= ~CPU_VINTD;
3936 	vmcs_write(CPU_EXEC_CTRL1, secondary);
3937 	report_prefix_pushf("Process-posted-interrupts enabled; virtual-interrupt-delivery disabled; acknowledge-interrupt-on-exit enabled");
3938 	test_vmx_invalid_controls(false);
3939 	report_prefix_pop();
3940 
3941 	secondary |= CPU_VINTD;
3942 	vmcs_write(CPU_EXEC_CTRL1, secondary);
3943 	report_prefix_pushf("Process-posted-interrupts enabled; virtual-interrupt-delivery enabled; acknowledge-interrupt-on-exit enabled");
3944 	test_vmx_valid_controls(false);
3945 	report_prefix_pop();
3946 
3947 	/*
3948 	 * Test posted-interrupt notification vector
3949 	 */
3950 	for (i = 0; i < 8; i++) {
3951 		vec = (1ul << i);
3952 		vmcs_write(PINV, vec);
3953 		report_prefix_pushf("Process-posted-interrupts enabled; posted-interrupt-notification-vector %u", vec);
3954 		test_vmx_valid_controls(false);
3955 		report_prefix_pop();
3956 	}
3957 	for (i = 8; i < 16; i++) {
3958 		vec = (1ul << i);
3959 		vmcs_write(PINV, vec);
3960 		report_prefix_pushf("Process-posted-interrupts enabled; posted-interrupt-notification-vector %u", vec);
3961 		test_vmx_invalid_controls(false);
3962 		report_prefix_pop();
3963 	}
3964 
3965 	vec &= ~(0xff << 8);
3966 	vmcs_write(PINV, vec);
3967 	report_prefix_pushf("Process-posted-interrupts enabled; posted-interrupt-notification-vector %u", vec);
3968 	test_vmx_valid_controls(false);
3969 	report_prefix_pop();
3970 
3971 	/*
3972 	 * Test posted-interrupt descriptor addresss
3973 	 */
3974 	for (i = 0; i < 6; i++) {
3975 		test_pi_desc_addr(1ul << i, false);
3976 	}
3977 
3978 	test_pi_desc_addr(0xf0, false);
3979 	test_pi_desc_addr(0xff, false);
3980 	test_pi_desc_addr(0x0f, false);
3981 	test_pi_desc_addr(0x8000, true);
3982 	test_pi_desc_addr(0x00, true);
3983 	test_pi_desc_addr(0xc000, true);
3984 
3985 	test_vmcs_addr_values("process-posted interrupts",
3986 			       POSTED_INTR_DESC_ADDR, 64,
3987 			       false, false, 0, 63);
3988 
3989 	vmcs_write(CPU_EXEC_CTRL0, saved_primary);
3990 	vmcs_write(CPU_EXEC_CTRL1, saved_secondary);
3991 	vmcs_write(PIN_CONTROLS, saved_pin);
3992 }
3993 
3994 static void test_apic_ctls(void)
3995 {
3996 	test_apic_virt_addr();
3997 	test_apic_access_addr();
3998 	test_apic_virtual_ctls();
3999 	test_virtual_intr_ctls();
4000 	test_posted_intr();
4001 }
4002 
4003 /*
4004  * If the “enable VPID†VM-execution control is 1, the value of the
4005  * of the VPID VM-execution control field must not be 0000H.
4006  * [Intel SDM]
4007  */
4008 static void test_vpid(void)
4009 {
4010 	u32 saved_primary = vmcs_read(CPU_EXEC_CTRL0);
4011 	u32 saved_secondary = vmcs_read(CPU_EXEC_CTRL1);
4012 	u16 vpid = 0x0000;
4013 	int i;
4014 
4015 	if (!((ctrl_cpu_rev[0].clr & CPU_SECONDARY) &&
4016 	    (ctrl_cpu_rev[1].clr & CPU_VPID))) {
4017 		test_skip("Secondary controls and/or VPID not supported");
4018 		return;
4019 	}
4020 
4021 	vmcs_write(CPU_EXEC_CTRL0, saved_primary | CPU_SECONDARY);
4022 	vmcs_write(CPU_EXEC_CTRL1, saved_secondary & ~CPU_VPID);
4023 	vmcs_write(VPID, vpid);
4024 	report_prefix_pushf("VPID disabled; VPID value %x", vpid);
4025 	test_vmx_valid_controls(false);
4026 	report_prefix_pop();
4027 
4028 	vmcs_write(CPU_EXEC_CTRL1, saved_secondary | CPU_VPID);
4029 	report_prefix_pushf("VPID enabled; VPID value %x", vpid);
4030 	test_vmx_invalid_controls(false);
4031 	report_prefix_pop();
4032 
4033 	for (i = 0; i < 16; i++) {
4034 		vpid = (short)1 << i;;
4035 		vmcs_write(VPID, vpid);
4036 		report_prefix_pushf("VPID enabled; VPID value %x", vpid);
4037 		test_vmx_valid_controls(false);
4038 		report_prefix_pop();
4039 	}
4040 
4041 	vmcs_write(CPU_EXEC_CTRL0, saved_primary);
4042 	vmcs_write(CPU_EXEC_CTRL1, saved_secondary);
4043 }
4044 
4045 static void set_vtpr(unsigned vtpr)
4046 {
4047 	*(u32 *)phys_to_virt(vmcs_read(APIC_VIRT_ADDR) + APIC_TASKPRI) = vtpr;
4048 }
4049 
4050 static void try_tpr_threshold_and_vtpr(unsigned threshold, unsigned vtpr)
4051 {
4052 	bool valid = true;
4053 	u32 primary = vmcs_read(CPU_EXEC_CTRL0);
4054 	u32 secondary = vmcs_read(CPU_EXEC_CTRL1);
4055 
4056 	if ((primary & CPU_TPR_SHADOW) &&
4057 	    (!(primary & CPU_SECONDARY) ||
4058 	     !(secondary & (CPU_VINTD | CPU_VIRT_APIC_ACCESSES))))
4059 		valid = (threshold & 0xf) <= ((vtpr >> 4) & 0xf);
4060 
4061 	set_vtpr(vtpr);
4062 	report_prefix_pushf("TPR threshold 0x%x, VTPR.class 0x%x",
4063 	    threshold, (vtpr >> 4) & 0xf);
4064 	if (valid)
4065 		test_vmx_valid_controls(false);
4066 	else
4067 		test_vmx_invalid_controls(false);
4068 	report_prefix_pop();
4069 }
4070 
4071 static void test_invalid_event_injection(void)
4072 {
4073 	u32 ent_intr_info_save = vmcs_read(ENT_INTR_INFO);
4074 	u32 ent_intr_error_save = vmcs_read(ENT_INTR_ERROR);
4075 	u32 ent_inst_len_save = vmcs_read(ENT_INST_LEN);
4076 	u32 primary_save = vmcs_read(CPU_EXEC_CTRL0);
4077 	u32 secondary_save = vmcs_read(CPU_EXEC_CTRL1);
4078 	u64 guest_cr0_save = vmcs_read(GUEST_CR0);
4079 	u32 ent_intr_info_base = INTR_INFO_VALID_MASK;
4080 	u32 ent_intr_info, ent_intr_err, ent_intr_len;
4081 	u32 cnt;
4082 
4083 	/* Setup */
4084 	report_prefix_push("invalid event injection");
4085 	vmcs_write(ENT_INTR_ERROR, 0x00000000);
4086 	vmcs_write(ENT_INST_LEN, 0x00000001);
4087 
4088 	/* The field’s interruption type is not set to a reserved value. */
4089 	ent_intr_info = ent_intr_info_base | INTR_TYPE_RESERVED | DE_VECTOR;
4090 	report_prefix_pushf("%s, VM-entry intr info=0x%x",
4091 			    "RESERVED interruption type invalid [-]",
4092 			    ent_intr_info);
4093 	vmcs_write(ENT_INTR_INFO, ent_intr_info);
4094 	test_vmx_invalid_controls(false);
4095 	report_prefix_pop();
4096 
4097 	ent_intr_info = ent_intr_info_base | INTR_TYPE_EXT_INTR |
4098 			DE_VECTOR;
4099 	report_prefix_pushf("%s, VM-entry intr info=0x%x",
4100 			    "RESERVED interruption type invalid [+]",
4101 			    ent_intr_info);
4102 	vmcs_write(ENT_INTR_INFO, ent_intr_info);
4103 	test_vmx_valid_controls(false);
4104 	report_prefix_pop();
4105 
4106 	/* If the interruption type is other event, the vector is 0. */
4107 	ent_intr_info = ent_intr_info_base | INTR_TYPE_OTHER_EVENT | DB_VECTOR;
4108 	report_prefix_pushf("%s, VM-entry intr info=0x%x",
4109 			    "(OTHER EVENT && vector != 0) invalid [-]",
4110 			    ent_intr_info);
4111 	vmcs_write(ENT_INTR_INFO, ent_intr_info);
4112 	test_vmx_invalid_controls(false);
4113 	report_prefix_pop();
4114 
4115 	/* If the interruption type is NMI, the vector is 2 (negative case). */
4116 	ent_intr_info = ent_intr_info_base | INTR_TYPE_NMI_INTR | DE_VECTOR;
4117 	report_prefix_pushf("%s, VM-entry intr info=0x%x",
4118 			    "(NMI && vector != 2) invalid [-]", ent_intr_info);
4119 	vmcs_write(ENT_INTR_INFO, ent_intr_info);
4120 	test_vmx_invalid_controls(false);
4121 	report_prefix_pop();
4122 
4123 	/* If the interruption type is NMI, the vector is 2 (positive case). */
4124 	ent_intr_info = ent_intr_info_base | INTR_TYPE_NMI_INTR | NMI_VECTOR;
4125 	report_prefix_pushf("%s, VM-entry intr info=0x%x",
4126 			    "(NMI && vector == 2) valid [+]", ent_intr_info);
4127 	vmcs_write(ENT_INTR_INFO, ent_intr_info);
4128 	test_vmx_valid_controls(false);
4129 	report_prefix_pop();
4130 
4131 	/*
4132 	 * If the interruption type
4133 	 * is HW exception, the vector is at most 31.
4134 	 */
4135 	ent_intr_info = ent_intr_info_base | INTR_TYPE_HARD_EXCEPTION | 0x20;
4136 	report_prefix_pushf("%s, VM-entry intr info=0x%x",
4137 			    "(HW exception && vector > 31) invalid [-]",
4138 			    ent_intr_info);
4139 	vmcs_write(ENT_INTR_INFO, ent_intr_info);
4140 	test_vmx_invalid_controls(false);
4141 	report_prefix_pop();
4142 
4143 	/*
4144 	 * deliver-error-code is 1 iff either
4145 	 * (a) the "unrestricted guest" VM-execution control is 0
4146 	 * (b) CR0.PE is set.
4147 	 */
4148 
4149 	/* Assert that unrestricted guest is disabled or unsupported */
4150 	assert(!(ctrl_cpu_rev[0].clr & CPU_SECONDARY) ||
4151 	       !(secondary_save & CPU_URG));
4152 
4153 	ent_intr_info = ent_intr_info_base | INTR_TYPE_HARD_EXCEPTION |
4154 			GP_VECTOR;
4155 	report_prefix_pushf("%s, VM-entry intr info=0x%x",
4156 			    "error code <-> (!URG || prot_mode) [-]",
4157 			    ent_intr_info);
4158 	vmcs_write(GUEST_CR0, guest_cr0_save & ~X86_CR0_PE & ~X86_CR0_PG);
4159 	vmcs_write(ENT_INTR_INFO, ent_intr_info);
4160 	test_vmx_invalid_controls(false);
4161 	report_prefix_pop();
4162 
4163 	ent_intr_info = ent_intr_info_base | INTR_INFO_DELIVER_CODE_MASK |
4164 			INTR_TYPE_HARD_EXCEPTION | GP_VECTOR;
4165 	report_prefix_pushf("%s, VM-entry intr info=0x%x",
4166 			    "error code <-> (!URG || prot_mode) [+]",
4167 			    ent_intr_info);
4168 	vmcs_write(GUEST_CR0, guest_cr0_save & ~X86_CR0_PE & ~X86_CR0_PG);
4169 	vmcs_write(ENT_INTR_INFO, ent_intr_info);
4170 	test_vmx_valid_controls(false);
4171 	report_prefix_pop();
4172 
4173 	if (enable_unrestricted_guest())
4174 		goto skip_unrestricted_guest;
4175 
4176 	ent_intr_info = ent_intr_info_base | INTR_INFO_DELIVER_CODE_MASK |
4177 			INTR_TYPE_HARD_EXCEPTION | GP_VECTOR;
4178 	report_prefix_pushf("%s, VM-entry intr info=0x%x",
4179 			    "error code <-> (!URG || prot_mode) [-]",
4180 			    ent_intr_info);
4181 	vmcs_write(GUEST_CR0, guest_cr0_save & ~X86_CR0_PE & ~X86_CR0_PG);
4182 	vmcs_write(ENT_INTR_INFO, ent_intr_info);
4183 	test_vmx_invalid_controls(false);
4184 	report_prefix_pop();
4185 
4186 	ent_intr_info = ent_intr_info_base | INTR_TYPE_HARD_EXCEPTION |
4187 			GP_VECTOR;
4188 	report_prefix_pushf("%s, VM-entry intr info=0x%x",
4189 			    "error code <-> (!URG || prot_mode) [-]",
4190 			    ent_intr_info);
4191 	vmcs_write(GUEST_CR0, guest_cr0_save | X86_CR0_PE);
4192 	vmcs_write(ENT_INTR_INFO, ent_intr_info);
4193 	test_vmx_invalid_controls(false);
4194 	report_prefix_pop();
4195 
4196 	vmcs_write(CPU_EXEC_CTRL1, secondary_save);
4197 	vmcs_write(CPU_EXEC_CTRL0, primary_save);
4198 
4199 skip_unrestricted_guest:
4200 	vmcs_write(GUEST_CR0, guest_cr0_save);
4201 
4202 	/* deliver-error-code is 1 iff the interruption type is HW exception */
4203 	report_prefix_push("error code <-> HW exception");
4204 	for (cnt = 0; cnt < 8; cnt++) {
4205 		u32 exception_type_mask = cnt << 8;
4206 		u32 deliver_error_code_mask =
4207 			exception_type_mask != INTR_TYPE_HARD_EXCEPTION ?
4208 			INTR_INFO_DELIVER_CODE_MASK : 0;
4209 
4210 		ent_intr_info = ent_intr_info_base | deliver_error_code_mask |
4211 				exception_type_mask | GP_VECTOR;
4212 		report_prefix_pushf("VM-entry intr info=0x%x [-]",
4213 				    ent_intr_info);
4214 		vmcs_write(ENT_INTR_INFO, ent_intr_info);
4215 		test_vmx_invalid_controls(false);
4216 		report_prefix_pop();
4217 	}
4218 	report_prefix_pop();
4219 
4220 	/*
4221 	 * deliver-error-code is 1 iff the the vector
4222 	 * indicates an exception that would normally deliver an error code
4223 	 */
4224 	report_prefix_push("error code <-> vector delivers error code");
4225 	for (cnt = 0; cnt < 32; cnt++) {
4226 		bool has_error_code = false;
4227 		u32 deliver_error_code_mask;
4228 
4229 		switch (cnt) {
4230 		case DF_VECTOR:
4231 		case TS_VECTOR:
4232 		case NP_VECTOR:
4233 		case SS_VECTOR:
4234 		case GP_VECTOR:
4235 		case PF_VECTOR:
4236 		case AC_VECTOR:
4237 			has_error_code = true;
4238 		}
4239 
4240 		/* Negative case */
4241 		deliver_error_code_mask = has_error_code ?
4242 						0 :
4243 						INTR_INFO_DELIVER_CODE_MASK;
4244 		ent_intr_info = ent_intr_info_base | deliver_error_code_mask |
4245 				INTR_TYPE_HARD_EXCEPTION | cnt;
4246 		report_prefix_pushf("VM-entry intr info=0x%x [-]",
4247 				    ent_intr_info);
4248 		vmcs_write(ENT_INTR_INFO, ent_intr_info);
4249 		test_vmx_invalid_controls(false);
4250 		report_prefix_pop();
4251 
4252 		/* Positive case */
4253 		deliver_error_code_mask = has_error_code ?
4254 						INTR_INFO_DELIVER_CODE_MASK :
4255 						0;
4256 		ent_intr_info = ent_intr_info_base | deliver_error_code_mask |
4257 				INTR_TYPE_HARD_EXCEPTION | cnt;
4258 		report_prefix_pushf("VM-entry intr info=0x%x [+]",
4259 				    ent_intr_info);
4260 		vmcs_write(ENT_INTR_INFO, ent_intr_info);
4261 		test_vmx_valid_controls(false);
4262 		report_prefix_pop();
4263 	}
4264 	report_prefix_pop();
4265 
4266 	/* Reserved bits in the field (30:12) are 0. */
4267 	report_prefix_push("reserved bits clear");
4268 	for (cnt = 12; cnt <= 30; cnt++) {
4269 		ent_intr_info = ent_intr_info_base |
4270 				INTR_INFO_DELIVER_CODE_MASK |
4271 				INTR_TYPE_HARD_EXCEPTION | GP_VECTOR |
4272 				(1U << cnt);
4273 		report_prefix_pushf("VM-entry intr info=0x%x [-]",
4274 				    ent_intr_info);
4275 		vmcs_write(ENT_INTR_INFO, ent_intr_info);
4276 		test_vmx_invalid_controls(false);
4277 		report_prefix_pop();
4278 	}
4279 	report_prefix_pop();
4280 
4281 	/*
4282 	 * If deliver-error-code is 1
4283 	 * bits 31:15 of the VM-entry exception error-code field are 0.
4284 	 */
4285 	ent_intr_info = ent_intr_info_base | INTR_INFO_DELIVER_CODE_MASK |
4286 			INTR_TYPE_HARD_EXCEPTION | GP_VECTOR;
4287 	report_prefix_pushf("%s, VM-entry intr info=0x%x",
4288 			    "VM-entry exception error code[31:15] clear",
4289 			    ent_intr_info);
4290 	vmcs_write(ENT_INTR_INFO, ent_intr_info);
4291 	for (cnt = 15; cnt <= 31; cnt++) {
4292 		ent_intr_err = 1U << cnt;
4293 		report_prefix_pushf("VM-entry intr error=0x%x [-]",
4294 				    ent_intr_err);
4295 		vmcs_write(ENT_INTR_ERROR, ent_intr_err);
4296 		test_vmx_invalid_controls(false);
4297 		report_prefix_pop();
4298 	}
4299 	vmcs_write(ENT_INTR_ERROR, 0x00000000);
4300 	report_prefix_pop();
4301 
4302 	/*
4303 	 * If the interruption type is software interrupt, software exception,
4304 	 * or privileged software exception, the VM-entry instruction-length
4305 	 * field is in the range 0–15.
4306 	 */
4307 
4308 	for (cnt = 0; cnt < 3; cnt++) {
4309 		switch (cnt) {
4310 		case 0:
4311 			ent_intr_info = ent_intr_info_base |
4312 					INTR_TYPE_SOFT_INTR;
4313 			break;
4314 		case 1:
4315 			ent_intr_info = ent_intr_info_base |
4316 					INTR_TYPE_SOFT_EXCEPTION;
4317 			break;
4318 		case 2:
4319 			ent_intr_info = ent_intr_info_base |
4320 					INTR_TYPE_PRIV_SW_EXCEPTION;
4321 			break;
4322 		}
4323 		report_prefix_pushf("%s, VM-entry intr info=0x%x",
4324 				    "VM-entry instruction-length check",
4325 				    ent_intr_info);
4326 		vmcs_write(ENT_INTR_INFO, ent_intr_info);
4327 
4328 		/* Instruction length set to -1 (0xFFFFFFFF) should fail */
4329 		ent_intr_len = -1;
4330 		report_prefix_pushf("VM-entry intr length = 0x%x [-]",
4331 				    ent_intr_len);
4332 		vmcs_write(ENT_INST_LEN, ent_intr_len);
4333 		test_vmx_invalid_controls(false);
4334 		report_prefix_pop();
4335 
4336 		/* Instruction length set to 16 should fail */
4337 		ent_intr_len = 0x00000010;
4338 		report_prefix_pushf("VM-entry intr length = 0x%x [-]",
4339 				    ent_intr_len);
4340 		vmcs_write(ENT_INST_LEN, 0x00000010);
4341 		test_vmx_invalid_controls(false);
4342 		report_prefix_pop();
4343 
4344 		report_prefix_pop();
4345 	}
4346 
4347 	/* Cleanup */
4348 	vmcs_write(ENT_INTR_INFO, ent_intr_info_save);
4349 	vmcs_write(ENT_INTR_ERROR, ent_intr_error_save);
4350 	vmcs_write(ENT_INST_LEN, ent_inst_len_save);
4351 	vmcs_write(CPU_EXEC_CTRL0, primary_save);
4352 	vmcs_write(CPU_EXEC_CTRL1, secondary_save);
4353 	vmcs_write(GUEST_CR0, guest_cr0_save);
4354 	report_prefix_pop();
4355 }
4356 
4357 /*
4358  * Test interesting vTPR values for a given TPR threshold.
4359  */
4360 static void test_vtpr_values(unsigned threshold)
4361 {
4362 	try_tpr_threshold_and_vtpr(threshold, (threshold - 1) << 4);
4363 	try_tpr_threshold_and_vtpr(threshold, threshold << 4);
4364 	try_tpr_threshold_and_vtpr(threshold, (threshold + 1) << 4);
4365 }
4366 
4367 static void try_tpr_threshold(unsigned threshold)
4368 {
4369 	bool valid = true;
4370 
4371 	u32 primary = vmcs_read(CPU_EXEC_CTRL0);
4372 	u32 secondary = vmcs_read(CPU_EXEC_CTRL1);
4373 
4374 	if ((primary & CPU_TPR_SHADOW) && !((primary & CPU_SECONDARY) &&
4375 	    (secondary & CPU_VINTD)))
4376 		valid = !(threshold >> 4);
4377 
4378 	set_vtpr(-1);
4379 	vmcs_write(TPR_THRESHOLD, threshold);
4380 	report_prefix_pushf("TPR threshold 0x%x, VTPR.class 0xf", threshold);
4381 	if (valid)
4382 		test_vmx_valid_controls(false);
4383 	else
4384 		test_vmx_invalid_controls(false);
4385 	report_prefix_pop();
4386 
4387 	if (valid)
4388 		test_vtpr_values(threshold);
4389 }
4390 
4391 /*
4392  * Test interesting TPR threshold values.
4393  */
4394 static void test_tpr_threshold_values(void)
4395 {
4396 	unsigned i;
4397 
4398 	for (i = 0; i < 0x10; i++)
4399 		try_tpr_threshold(i);
4400 	for (i = 4; i < 32; i++)
4401 		try_tpr_threshold(1u << i);
4402 	try_tpr_threshold(-1u);
4403 	try_tpr_threshold(0x7fffffff);
4404 }
4405 
4406 /*
4407  * This test covers the following two VM entry checks:
4408  *
4409  *      i) If the "use TPR shadow" VM-execution control is 1 and the
4410  *         "virtual-interrupt delivery" VM-execution control is 0, bits
4411  *         31:4 of the TPR threshold VM-execution control field must
4412 	   be 0.
4413  *         [Intel SDM]
4414  *
4415  *      ii) If the "use TPR shadow" VM-execution control is 1, the
4416  *          "virtual-interrupt delivery" VM-execution control is 0
4417  *          and the "virtualize APIC accesses" VM-execution control
4418  *          is 0, the value of bits 3:0 of the TPR threshold VM-execution
4419  *          control field must not be greater than the value of bits
4420  *          7:4 of VTPR.
4421  *          [Intel SDM]
4422  */
4423 static void test_tpr_threshold(void)
4424 {
4425 	u32 primary = vmcs_read(CPU_EXEC_CTRL0);
4426 	void *virtual_apic_page;
4427 
4428 	if (!(ctrl_cpu_rev[0].clr & CPU_TPR_SHADOW))
4429 		return;
4430 
4431 	virtual_apic_page = alloc_page();
4432 	memset(virtual_apic_page, 0xff, PAGE_SIZE);
4433 	vmcs_write(APIC_VIRT_ADDR, virt_to_phys(virtual_apic_page));
4434 
4435 	vmcs_write(CPU_EXEC_CTRL0, primary & ~(CPU_TPR_SHADOW | CPU_SECONDARY));
4436 	report_prefix_pushf("Use TPR shadow disabled, secondary controls disabled");
4437 	test_tpr_threshold_values();
4438 	report_prefix_pop();
4439 	vmcs_write(CPU_EXEC_CTRL0, vmcs_read(CPU_EXEC_CTRL0) | CPU_TPR_SHADOW);
4440 	report_prefix_pushf("Use TPR shadow enabled, secondary controls disabled");
4441 	test_tpr_threshold_values();
4442 	report_prefix_pop();
4443 
4444 	if (!((ctrl_cpu_rev[0].clr & CPU_SECONDARY) &&
4445 	    (ctrl_cpu_rev[1].clr & (CPU_VINTD  | CPU_VIRT_APIC_ACCESSES)))) {
4446 		vmcs_write(CPU_EXEC_CTRL0, primary);
4447 		return;
4448 	}
4449 
4450 	u32 secondary = vmcs_read(CPU_EXEC_CTRL1);
4451 
4452 	if (ctrl_cpu_rev[1].clr & CPU_VINTD) {
4453 		vmcs_write(CPU_EXEC_CTRL1, CPU_VINTD);
4454 		report_prefix_pushf("Use TPR shadow enabled; secondary controls disabled; virtual-interrupt delivery enabled; virtualize APIC accesses disabled");
4455 		test_tpr_threshold_values();
4456 		report_prefix_pop();
4457 
4458 		vmcs_write(CPU_EXEC_CTRL0,
4459 			   vmcs_read(CPU_EXEC_CTRL0) | CPU_SECONDARY);
4460 		report_prefix_pushf("Use TPR shadow enabled; secondary controls enabled; virtual-interrupt delivery enabled; virtualize APIC accesses disabled");
4461 		test_tpr_threshold_values();
4462 		report_prefix_pop();
4463 	}
4464 
4465 	if (ctrl_cpu_rev[1].clr & CPU_VIRT_APIC_ACCESSES) {
4466 		vmcs_write(CPU_EXEC_CTRL0,
4467 			   vmcs_read(CPU_EXEC_CTRL0) & ~CPU_SECONDARY);
4468 		vmcs_write(CPU_EXEC_CTRL1, CPU_VIRT_APIC_ACCESSES);
4469 		report_prefix_pushf("Use TPR shadow enabled; secondary controls disabled; virtual-interrupt delivery enabled; virtualize APIC accesses enabled");
4470 		test_tpr_threshold_values();
4471 		report_prefix_pop();
4472 
4473 		vmcs_write(CPU_EXEC_CTRL0,
4474 			   vmcs_read(CPU_EXEC_CTRL0) | CPU_SECONDARY);
4475 		report_prefix_pushf("Use TPR shadow enabled; secondary controls enabled; virtual-interrupt delivery enabled; virtualize APIC accesses enabled");
4476 		test_tpr_threshold_values();
4477 		report_prefix_pop();
4478 	}
4479 
4480 	if ((ctrl_cpu_rev[1].clr &
4481 	     (CPU_VINTD | CPU_VIRT_APIC_ACCESSES)) ==
4482 	    (CPU_VINTD | CPU_VIRT_APIC_ACCESSES)) {
4483 		vmcs_write(CPU_EXEC_CTRL0,
4484 			   vmcs_read(CPU_EXEC_CTRL0) & ~CPU_SECONDARY);
4485 		vmcs_write(CPU_EXEC_CTRL1,
4486 			   CPU_VINTD | CPU_VIRT_APIC_ACCESSES);
4487 		report_prefix_pushf("Use TPR shadow enabled; secondary controls disabled; virtual-interrupt delivery enabled; virtualize APIC accesses enabled");
4488 		test_tpr_threshold_values();
4489 		report_prefix_pop();
4490 
4491 		vmcs_write(CPU_EXEC_CTRL0,
4492 			   vmcs_read(CPU_EXEC_CTRL0) | CPU_SECONDARY);
4493 		report_prefix_pushf("Use TPR shadow enabled; secondary controls enabled; virtual-interrupt delivery enabled; virtualize APIC accesses enabled");
4494 		test_tpr_threshold_values();
4495 		report_prefix_pop();
4496 	}
4497 
4498 	vmcs_write(CPU_EXEC_CTRL1, secondary);
4499 	vmcs_write(CPU_EXEC_CTRL0, primary);
4500 }
4501 
4502 /*
4503  * This test verifies the following two vmentry checks:
4504  *
4505  *  If the "NMI exiting" VM-execution control is 0, "Virtual NMIs"
4506  *  VM-execution control must be 0.
4507  *  [Intel SDM]
4508  *
4509  *  If the “virtual NMIs” VM-execution control is 0, the “NMI-window
4510  *  exiting” VM-execution control must be 0.
4511  *  [Intel SDM]
4512  */
4513 static void test_nmi_ctrls(void)
4514 {
4515 	u32 pin_ctrls, cpu_ctrls0, test_pin_ctrls, test_cpu_ctrls0;
4516 
4517 	if ((ctrl_pin_rev.clr & (PIN_NMI | PIN_VIRT_NMI)) !=
4518 	    (PIN_NMI | PIN_VIRT_NMI)) {
4519 		test_skip("NMI exiting and Virtual NMIs are not supported !");
4520 		return;
4521 	}
4522 
4523 	/* Save the controls so that we can restore them after our tests */
4524 	pin_ctrls = vmcs_read(PIN_CONTROLS);
4525 	cpu_ctrls0 = vmcs_read(CPU_EXEC_CTRL0);
4526 
4527 	test_pin_ctrls = pin_ctrls & ~(PIN_NMI | PIN_VIRT_NMI);
4528 	test_cpu_ctrls0 = cpu_ctrls0 & ~CPU_NMI_WINDOW;
4529 
4530 	vmcs_write(PIN_CONTROLS, test_pin_ctrls);
4531 	report_prefix_pushf("NMI-exiting disabled, virtual-NMIs disabled");
4532 	test_vmx_valid_controls(false);
4533 	report_prefix_pop();
4534 
4535 	vmcs_write(PIN_CONTROLS, test_pin_ctrls | PIN_VIRT_NMI);
4536 	report_prefix_pushf("NMI-exiting disabled, virtual-NMIs enabled");
4537 	test_vmx_invalid_controls(false);
4538 	report_prefix_pop();
4539 
4540 	vmcs_write(PIN_CONTROLS, test_pin_ctrls | (PIN_NMI | PIN_VIRT_NMI));
4541 	report_prefix_pushf("NMI-exiting enabled, virtual-NMIs enabled");
4542 	test_vmx_valid_controls(false);
4543 	report_prefix_pop();
4544 
4545 	vmcs_write(PIN_CONTROLS, test_pin_ctrls | PIN_NMI);
4546 	report_prefix_pushf("NMI-exiting enabled, virtual-NMIs disabled");
4547 	test_vmx_valid_controls(false);
4548 	report_prefix_pop();
4549 
4550 	if (!(ctrl_cpu_rev[0].clr & CPU_NMI_WINDOW)) {
4551 		report_info("NMI-window exiting is not supported, skipping...");
4552 		goto done;
4553 	}
4554 
4555 	vmcs_write(PIN_CONTROLS, test_pin_ctrls);
4556 	vmcs_write(CPU_EXEC_CTRL0, test_cpu_ctrls0 | CPU_NMI_WINDOW);
4557 	report_prefix_pushf("Virtual-NMIs disabled, NMI-window-exiting enabled");
4558 	test_vmx_invalid_controls(false);
4559 	report_prefix_pop();
4560 
4561 	vmcs_write(PIN_CONTROLS, test_pin_ctrls);
4562 	vmcs_write(CPU_EXEC_CTRL0, test_cpu_ctrls0);
4563 	report_prefix_pushf("Virtual-NMIs disabled, NMI-window-exiting disabled");
4564 	test_vmx_valid_controls(false);
4565 	report_prefix_pop();
4566 
4567 	vmcs_write(PIN_CONTROLS, test_pin_ctrls | (PIN_NMI | PIN_VIRT_NMI));
4568 	vmcs_write(CPU_EXEC_CTRL0, test_cpu_ctrls0 | CPU_NMI_WINDOW);
4569 	report_prefix_pushf("Virtual-NMIs enabled, NMI-window-exiting enabled");
4570 	test_vmx_valid_controls(false);
4571 	report_prefix_pop();
4572 
4573 	vmcs_write(PIN_CONTROLS, test_pin_ctrls | (PIN_NMI | PIN_VIRT_NMI));
4574 	vmcs_write(CPU_EXEC_CTRL0, test_cpu_ctrls0);
4575 	report_prefix_pushf("Virtual-NMIs enabled, NMI-window-exiting disabled");
4576 	test_vmx_valid_controls(false);
4577 	report_prefix_pop();
4578 
4579 	/* Restore the controls to their original values */
4580 	vmcs_write(CPU_EXEC_CTRL0, cpu_ctrls0);
4581 done:
4582 	vmcs_write(PIN_CONTROLS, pin_ctrls);
4583 }
4584 
4585 static void test_eptp_ad_bit(u64 eptp, bool ctrl)
4586 {
4587 	vmcs_write(EPTP, eptp);
4588 	report_prefix_pushf("Enable-EPT enabled; EPT accessed and dirty flag %s",
4589 	    (eptp & EPTP_AD_FLAG) ? "1": "0");
4590 	if (ctrl)
4591 		test_vmx_valid_controls(false);
4592 	else
4593 		test_vmx_invalid_controls(false);
4594 	report_prefix_pop();
4595 
4596 }
4597 
4598 /*
4599  * 1. If the "enable EPT" VM-execution control is 1, the "EPTP VM-execution"
4600  *    control field must satisfy the following checks:
4601  *
4602  *     - The EPT memory type (bits 2:0) must be a value supported by the
4603  *	 processor as indicated in the IA32_VMX_EPT_VPID_CAP MSR.
4604  *     - Bits 5:3 (1 less than the EPT page-walk length) must be 3,
4605  *	 indicating an EPT page-walk length of 4.
4606  *     - Bit 6 (enable bit for accessed and dirty flags for EPT) must be
4607  *	 0 if bit 21 of the IA32_VMX_EPT_VPID_CAP MSR is read as 0,
4608  *	 indicating that the processor does not support accessed and dirty
4609  *	 dirty flags for EPT.
4610  *     - Reserved bits 11:7 and 63:N (where N is the processor's
4611  *	 physical-address width) must all be 0.
4612  *
4613  * 2. If the "unrestricted guest" VM-execution control is 1, the
4614  *    "enable EPT" VM-execution control must also be 1.
4615  */
4616 static void test_ept_eptp(void)
4617 {
4618 	u32 primary_saved = vmcs_read(CPU_EXEC_CTRL0);
4619 	u32 secondary_saved = vmcs_read(CPU_EXEC_CTRL1);
4620 	u64 eptp_saved = vmcs_read(EPTP);
4621 	u32 primary = primary_saved;
4622 	u32 secondary = secondary_saved;
4623 	u64 msr, eptp = eptp_saved;
4624 	bool un_cache = false;
4625 	bool wr_bk = false;
4626 	bool ctrl;
4627 	u32 i, maxphysaddr;
4628 	u64 j, resv_bits_mask = 0;
4629 
4630 	if (!((ctrl_cpu_rev[0].clr & CPU_SECONDARY) &&
4631 	    (ctrl_cpu_rev[1].clr & CPU_EPT))) {
4632 		test_skip("\"CPU secondary\" and/or \"enable EPT\" execution controls are not supported !");
4633 		return;
4634 	}
4635 
4636 	/*
4637 	 * Memory type (bits 2:0)
4638 	 */
4639 	msr = rdmsr(MSR_IA32_VMX_EPT_VPID_CAP);
4640 	if (msr & EPT_CAP_UC)
4641 		un_cache = true;
4642 	if (msr & EPT_CAP_WB)
4643 		wr_bk = true;
4644 
4645 	primary |= CPU_SECONDARY;
4646 	vmcs_write(CPU_EXEC_CTRL0, primary);
4647 	secondary |= CPU_EPT;
4648 	vmcs_write(CPU_EXEC_CTRL1, secondary);
4649 	eptp = (eptp & ~EPTP_PG_WALK_LEN_MASK) |
4650 	    (3ul << EPTP_PG_WALK_LEN_SHIFT);
4651 	vmcs_write(EPTP, eptp);
4652 
4653 	for (i = 0; i < 8; i++) {
4654 		if (i == 0) {
4655 			if (un_cache) {
4656 				report_info("EPT paging structure memory-type is Un-cacheable\n");
4657 				ctrl = true;
4658 			} else {
4659 				ctrl = false;
4660 			}
4661 		} else if (i == 6) {
4662 			if (wr_bk) {
4663 				report_info("EPT paging structure memory-type is Write-back\n");
4664 				ctrl = true;
4665 			} else {
4666 				ctrl = false;
4667 			}
4668 		} else {
4669 			ctrl = false;
4670 		}
4671 
4672 		eptp = (eptp & ~EPT_MEM_TYPE_MASK) | i;
4673 		vmcs_write(EPTP, eptp);
4674 		report_prefix_pushf("Enable-EPT enabled; EPT memory type %lu",
4675 		    eptp & EPT_MEM_TYPE_MASK);
4676 		if (ctrl)
4677 			test_vmx_valid_controls(false);
4678 		else
4679 			test_vmx_invalid_controls(false);
4680 		report_prefix_pop();
4681 	}
4682 
4683 	eptp = (eptp & ~EPT_MEM_TYPE_MASK) | 6ul;
4684 
4685 	/*
4686 	 * Page walk length (bits 5:3)
4687 	 */
4688 	for (i = 0; i < 8; i++) {
4689 		eptp = (eptp & ~EPTP_PG_WALK_LEN_MASK) |
4690 		    (i << EPTP_PG_WALK_LEN_SHIFT);
4691 		if (i == 3)
4692 			ctrl = true;
4693 		else
4694 			ctrl = false;
4695 
4696 		vmcs_write(EPTP, eptp);
4697 		report_prefix_pushf("Enable-EPT enabled; EPT page walk length %lu",
4698 		    eptp & EPTP_PG_WALK_LEN_MASK);
4699 		if (ctrl)
4700 			test_vmx_valid_controls(false);
4701 		else
4702 			test_vmx_invalid_controls(false);
4703 		report_prefix_pop();
4704 	}
4705 
4706 	eptp = (eptp & ~EPTP_PG_WALK_LEN_MASK) |
4707 	    3ul << EPTP_PG_WALK_LEN_SHIFT;
4708 
4709 	/*
4710 	 * Accessed and dirty flag (bit 6)
4711 	 */
4712 	if (msr & EPT_CAP_AD_FLAG) {
4713 		report_info("Processor supports accessed and dirty flag");
4714 		eptp &= ~EPTP_AD_FLAG;
4715 		test_eptp_ad_bit(eptp, true);
4716 
4717 		eptp |= EPTP_AD_FLAG;
4718 		test_eptp_ad_bit(eptp, true);
4719 	} else {
4720 		report_info("Processor does not supports accessed and dirty flag");
4721 		eptp &= ~EPTP_AD_FLAG;
4722 		test_eptp_ad_bit(eptp, true);
4723 
4724 		eptp |= EPTP_AD_FLAG;
4725 		test_eptp_ad_bit(eptp, false);
4726 	}
4727 
4728 	/*
4729 	 * Reserved bits [11:7] and [63:N]
4730 	 */
4731 	for (i = 0; i < 32; i++) {
4732 		eptp = (eptp &
4733 		    ~(EPTP_RESERV_BITS_MASK << EPTP_RESERV_BITS_SHIFT)) |
4734 		    (i << EPTP_RESERV_BITS_SHIFT);
4735 		vmcs_write(EPTP, eptp);
4736 		report_prefix_pushf("Enable-EPT enabled; reserved bits [11:7] %lu",
4737 		    (eptp >> EPTP_RESERV_BITS_SHIFT) &
4738 		    EPTP_RESERV_BITS_MASK);
4739 		if (i == 0)
4740 			test_vmx_valid_controls(false);
4741 		else
4742 			test_vmx_invalid_controls(false);
4743 		report_prefix_pop();
4744 	}
4745 
4746 	eptp = (eptp & ~(EPTP_RESERV_BITS_MASK << EPTP_RESERV_BITS_SHIFT));
4747 
4748 	maxphysaddr = cpuid_maxphyaddr();
4749 	for (i = 0; i < (63 - maxphysaddr + 1); i++) {
4750 		resv_bits_mask |= 1ul << i;
4751 	}
4752 
4753 	for (j = maxphysaddr - 1; j <= 63; j++) {
4754 		eptp = (eptp & ~(resv_bits_mask << maxphysaddr)) |
4755 		    (j < maxphysaddr ? 0 : 1ul << j);
4756 		vmcs_write(EPTP, eptp);
4757 		report_prefix_pushf("Enable-EPT enabled; reserved bits [63:N] %lu",
4758 		    (eptp >> maxphysaddr) & resv_bits_mask);
4759 		if (j < maxphysaddr)
4760 			test_vmx_valid_controls(false);
4761 		else
4762 			test_vmx_invalid_controls(false);
4763 		report_prefix_pop();
4764 	}
4765 
4766 	secondary &= ~(CPU_EPT | CPU_URG);
4767 	vmcs_write(CPU_EXEC_CTRL1, secondary);
4768 	report_prefix_pushf("Enable-EPT disabled, unrestricted-guest disabled");
4769 	test_vmx_valid_controls(false);
4770 	report_prefix_pop();
4771 
4772 	secondary |= CPU_URG;
4773 	vmcs_write(CPU_EXEC_CTRL1, secondary);
4774 	report_prefix_pushf("Enable-EPT disabled, unrestricted-guest enabled");
4775 	test_vmx_invalid_controls(false);
4776 	report_prefix_pop();
4777 
4778 	secondary |= CPU_EPT;
4779 	setup_dummy_ept();
4780 	report_prefix_pushf("Enable-EPT enabled, unrestricted-guest enabled");
4781 	test_vmx_valid_controls(false);
4782 	report_prefix_pop();
4783 
4784 	secondary &= ~CPU_URG;
4785 	vmcs_write(CPU_EXEC_CTRL1, secondary);
4786 	report_prefix_pushf("Enable-EPT enabled, unrestricted-guest disabled");
4787 	test_vmx_valid_controls(false);
4788 	report_prefix_pop();
4789 
4790 	vmcs_write(CPU_EXEC_CTRL0, primary_saved);
4791 	vmcs_write(CPU_EXEC_CTRL1, secondary_saved);
4792 	vmcs_write(EPTP, eptp_saved);
4793 }
4794 
4795 /*
4796  * If the 'enable PML' VM-execution control is 1, the 'enable EPT'
4797  * VM-execution control must also be 1. In addition, the PML address
4798  * must satisfy the following checks:
4799  *
4800  *    * Bits 11:0 of the address must be 0.
4801  *    * The address should not set any bits beyond the processor's
4802  *	physical-address width.
4803  *
4804  *  [Intel SDM]
4805  */
4806 static void test_pml(void)
4807 {
4808 	u32 primary_saved = vmcs_read(CPU_EXEC_CTRL0);
4809 	u32 secondary_saved = vmcs_read(CPU_EXEC_CTRL1);
4810 	u32 primary = primary_saved;
4811 	u32 secondary = secondary_saved;
4812 
4813 	if (!((ctrl_cpu_rev[0].clr & CPU_SECONDARY) &&
4814 	    (ctrl_cpu_rev[1].clr & CPU_EPT) && (ctrl_cpu_rev[1].clr & CPU_PML))) {
4815 		test_skip("\"Secondary execution\" control or \"enable EPT\" control or \"enable PML\" control is not supported !");
4816 		return;
4817 	}
4818 
4819 	primary |= CPU_SECONDARY;
4820 	vmcs_write(CPU_EXEC_CTRL0, primary);
4821 	secondary &= ~(CPU_PML | CPU_EPT);
4822 	vmcs_write(CPU_EXEC_CTRL1, secondary);
4823 	report_prefix_pushf("enable-PML disabled, enable-EPT disabled");
4824 	test_vmx_valid_controls(false);
4825 	report_prefix_pop();
4826 
4827 	secondary |= CPU_PML;
4828 	vmcs_write(CPU_EXEC_CTRL1, secondary);
4829 	report_prefix_pushf("enable-PML enabled, enable-EPT disabled");
4830 	test_vmx_invalid_controls(false);
4831 	report_prefix_pop();
4832 
4833 	secondary |= CPU_EPT;
4834 	setup_dummy_ept();
4835 	report_prefix_pushf("enable-PML enabled, enable-EPT enabled");
4836 	test_vmx_valid_controls(false);
4837 	report_prefix_pop();
4838 
4839 	secondary &= ~CPU_PML;
4840 	vmcs_write(CPU_EXEC_CTRL1, secondary);
4841 	report_prefix_pushf("enable-PML disabled, enable EPT enabled");
4842 	test_vmx_valid_controls(false);
4843 	report_prefix_pop();
4844 
4845 	test_vmcs_addr_reference(CPU_PML, PMLADDR, "PML address", "PML",
4846 				 PAGE_SIZE, false, false);
4847 
4848 	vmcs_write(CPU_EXEC_CTRL0, primary_saved);
4849 	vmcs_write(CPU_EXEC_CTRL1, secondary_saved);
4850 }
4851 
4852  /*
4853  * If the "activate VMX-preemption timer" VM-execution control is 0, the
4854  * the "save VMX-preemption timer value" VM-exit control must also be 0.
4855  *
4856  *  [Intel SDM]
4857  */
4858 static void test_vmx_preemption_timer(void)
4859 {
4860 	u32 saved_pin = vmcs_read(PIN_CONTROLS);
4861 	u32 saved_exit = vmcs_read(EXI_CONTROLS);
4862 	u32 pin = saved_pin;
4863 	u32 exit = saved_exit;
4864 
4865 	if (!((ctrl_exit_rev.clr & EXI_SAVE_PREEMPT) ||
4866 	    (ctrl_pin_rev.clr & PIN_PREEMPT))) {
4867 		printf("\"Save-VMX-preemption-timer\" control and/or \"Enable-VMX-preemption-timer\" control is not supported\n");
4868 		return;
4869 	}
4870 
4871 	pin |= PIN_PREEMPT;
4872 	vmcs_write(PIN_CONTROLS, pin);
4873 	exit &= ~EXI_SAVE_PREEMPT;
4874 	vmcs_write(EXI_CONTROLS, exit);
4875 	report_prefix_pushf("enable-VMX-preemption-timer enabled, save-VMX-preemption-timer disabled");
4876 	test_vmx_valid_controls(false);
4877 	report_prefix_pop();
4878 
4879 	exit |= EXI_SAVE_PREEMPT;
4880 	vmcs_write(EXI_CONTROLS, exit);
4881 	report_prefix_pushf("enable-VMX-preemption-timer enabled, save-VMX-preemption-timer enabled");
4882 	test_vmx_valid_controls(false);
4883 	report_prefix_pop();
4884 
4885 	pin &= ~PIN_PREEMPT;
4886 	vmcs_write(PIN_CONTROLS, pin);
4887 	report_prefix_pushf("enable-VMX-preemption-timer disabled, save-VMX-preemption-timer enabled");
4888 	test_vmx_invalid_controls(false);
4889 	report_prefix_pop();
4890 
4891 	exit &= ~EXI_SAVE_PREEMPT;
4892 	vmcs_write(EXI_CONTROLS, exit);
4893 	report_prefix_pushf("enable-VMX-preemption-timer disabled, save-VMX-preemption-timer disabled");
4894 	test_vmx_valid_controls(false);
4895 	report_prefix_pop();
4896 
4897 	vmcs_write(PIN_CONTROLS, saved_pin);
4898 	vmcs_write(EXI_CONTROLS, saved_exit);
4899 }
4900 
4901 /*
4902  * Tests for VM-execution control fields
4903  */
4904 static void test_vm_execution_ctls(void)
4905 {
4906 	test_pin_based_ctls();
4907 	test_primary_processor_based_ctls();
4908 	test_secondary_processor_based_ctls();
4909 	test_cr3_targets();
4910 	test_io_bitmaps();
4911 	test_msr_bitmap();
4912 	test_apic_ctls();
4913 	test_tpr_threshold();
4914 	test_nmi_ctrls();
4915 	test_pml();
4916 	test_vpid();
4917 	test_ept_eptp();
4918 	test_vmx_preemption_timer();
4919 }
4920 
4921  /*
4922   * The following checks are performed for the VM-entry MSR-load address if
4923   * the VM-entry MSR-load count field is non-zero:
4924   *
4925   *    - The lower 4 bits of the VM-entry MSR-load address must be 0.
4926   *      The address should not set any bits beyond the processor’s
4927   *      physical-address width.
4928   *
4929   *    - The address of the last byte in the VM-entry MSR-load area
4930   *      should not set any bits beyond the processor’s physical-address
4931   *      width. The address of this last byte is VM-entry MSR-load address
4932   *      + (MSR count * 16) - 1. (The arithmetic used for the computation
4933   *      uses more bits than the processor’s physical-address width.)
4934   *
4935   *
4936   *  [Intel SDM]
4937   */
4938 static void test_entry_msr_load(void)
4939 {
4940 	entry_msr_load = alloc_page();
4941 	u64 tmp;
4942 	u32 entry_msr_ld_cnt = 1;
4943 	int i;
4944 	u32 addr_len = 64;
4945 
4946 	vmcs_write(ENT_MSR_LD_CNT, entry_msr_ld_cnt);
4947 
4948 	/* Check first 4 bits of VM-entry MSR-load address */
4949 	for (i = 0; i < 4; i++) {
4950 		tmp = (u64)entry_msr_load | 1ull << i;
4951 		vmcs_write(ENTER_MSR_LD_ADDR, tmp);
4952 		report_prefix_pushf("VM-entry MSR-load addr [4:0] %lx",
4953 				    tmp & 0xf);
4954 		test_vmx_invalid_controls(false);
4955 		report_prefix_pop();
4956 	}
4957 
4958 	if (basic.val & (1ul << 48))
4959 		addr_len = 32;
4960 
4961 	test_vmcs_addr_values("VM-entry-MSR-load address",
4962 				ENTER_MSR_LD_ADDR, 16, false, false,
4963 				4, addr_len - 1);
4964 
4965 	/*
4966 	 * Check last byte of VM-entry MSR-load address
4967 	 */
4968 	entry_msr_load = (struct vmx_msr_entry *)((u64)entry_msr_load & ~0xf);
4969 
4970 	for (i = (addr_len == 64 ? cpuid_maxphyaddr(): addr_len);
4971 							i < 64; i++) {
4972 		tmp = ((u64)entry_msr_load + entry_msr_ld_cnt * 16 - 1) |
4973 			1ul << i;
4974 		vmcs_write(ENTER_MSR_LD_ADDR,
4975 			   tmp - (entry_msr_ld_cnt * 16 - 1));
4976 		test_vmx_invalid_controls(false);
4977 	}
4978 
4979 	vmcs_write(ENT_MSR_LD_CNT, 2);
4980 	vmcs_write(ENTER_MSR_LD_ADDR, (1ULL << cpuid_maxphyaddr()) - 16);
4981 	test_vmx_invalid_controls(false);
4982 	vmcs_write(ENTER_MSR_LD_ADDR, (1ULL << cpuid_maxphyaddr()) - 32);
4983 	test_vmx_valid_controls(false);
4984 	vmcs_write(ENTER_MSR_LD_ADDR, (1ULL << cpuid_maxphyaddr()) - 48);
4985 	test_vmx_valid_controls(false);
4986 }
4987 
4988 /*
4989  * Tests for VM-entry control fields
4990  */
4991 static void test_vm_entry_ctls(void)
4992 {
4993 	test_invalid_event_injection();
4994 	test_entry_msr_load();
4995 }
4996 
4997 /*
4998  * The following checks are performed for the VM-exit MSR-store address if
4999  * the VM-exit MSR-store count field is non-zero:
5000  *
5001  *    - The lower 4 bits of the VM-exit MSR-store address must be 0.
5002  *      The address should not set any bits beyond the processor’s
5003  *      physical-address width.
5004  *
5005  *    - The address of the last byte in the VM-exit MSR-store area
5006  *      should not set any bits beyond the processor’s physical-address
5007  *      width. The address of this last byte is VM-exit MSR-store address
5008  *      + (MSR count * 16) - 1. (The arithmetic used for the computation
5009  *      uses more bits than the processor’s physical-address width.)
5010  *
5011  * If IA32_VMX_BASIC[48] is read as 1, neither address should set any bits
5012  * in the range 63:32.
5013  *
5014  *  [Intel SDM]
5015  */
5016 static void test_exit_msr_store(void)
5017 {
5018 	exit_msr_store = alloc_page();
5019 	u64 tmp;
5020 	u32 exit_msr_st_cnt = 1;
5021 	int i;
5022 	u32 addr_len = 64;
5023 
5024 	vmcs_write(EXI_MSR_ST_CNT, exit_msr_st_cnt);
5025 
5026 	/* Check first 4 bits of VM-exit MSR-store address */
5027 	for (i = 0; i < 4; i++) {
5028 		tmp = (u64)exit_msr_store | 1ull << i;
5029 		vmcs_write(EXIT_MSR_ST_ADDR, tmp);
5030 		report_prefix_pushf("VM-exit MSR-store addr [4:0] %lx",
5031 				    tmp & 0xf);
5032 		test_vmx_invalid_controls(false);
5033 		report_prefix_pop();
5034 	}
5035 
5036 	if (basic.val & (1ul << 48))
5037 		addr_len = 32;
5038 
5039 	test_vmcs_addr_values("VM-exit-MSR-store address",
5040 				EXIT_MSR_ST_ADDR, 16, false, false,
5041 				4, addr_len - 1);
5042 
5043 	/*
5044 	 * Check last byte of VM-exit MSR-store address
5045 	 */
5046 	exit_msr_store = (struct vmx_msr_entry *)((u64)exit_msr_store & ~0xf);
5047 
5048 	for (i = (addr_len == 64 ? cpuid_maxphyaddr(): addr_len);
5049 							i < 64; i++) {
5050 		tmp = ((u64)exit_msr_store + exit_msr_st_cnt * 16 - 1) |
5051 			1ul << i;
5052 		vmcs_write(EXIT_MSR_ST_ADDR,
5053 			   tmp - (exit_msr_st_cnt * 16 - 1));
5054 		test_vmx_invalid_controls(false);
5055 	}
5056 
5057 	vmcs_write(EXI_MSR_ST_CNT, 2);
5058 	vmcs_write(EXIT_MSR_ST_ADDR, (1ULL << cpuid_maxphyaddr()) - 16);
5059 	test_vmx_invalid_controls(false);
5060 	vmcs_write(EXIT_MSR_ST_ADDR, (1ULL << cpuid_maxphyaddr()) - 32);
5061 	test_vmx_valid_controls(false);
5062 	vmcs_write(EXIT_MSR_ST_ADDR, (1ULL << cpuid_maxphyaddr()) - 48);
5063 	test_vmx_valid_controls(false);
5064 }
5065 
5066 /*
5067  * Tests for VM-exit controls
5068  */
5069 static void test_vm_exit_ctls(void)
5070 {
5071 	test_exit_msr_store();
5072 }
5073 
5074 /*
5075  * Check that the virtual CPU checks all of the VMX controls as
5076  * documented in the Intel SDM.
5077  */
5078 static void vmx_controls_test(void)
5079 {
5080 	/*
5081 	 * Bit 1 of the guest's RFLAGS must be 1, or VM-entry will
5082 	 * fail due to invalid guest state, should we make it that
5083 	 * far.
5084 	 */
5085 	vmcs_write(GUEST_RFLAGS, 0);
5086 
5087 	test_vm_execution_ctls();
5088 	test_vm_exit_ctls();
5089 	test_vm_entry_ctls();
5090 }
5091 
5092 /*
5093  * 1. The CR0 field must not set any bit to a value not supported in VMX
5094  *    operation.
5095  * 2. The CR4 field must not set any bit to a value not supported in VMX
5096  *    operation.
5097  * 3. On processors that support Intel 64 architecture, the CR3 field must
5098  *    be such that bits 63:52 and bits in the range 51:32 beyond the
5099  *    processor’s physical-address width must be 0.
5100  *
5101  *  [Intel SDM]
5102  */
5103 static void test_host_ctl_regs(void)
5104 {
5105 	u64 fixed0, fixed1, cr3, cr3_saved;
5106 	int i;
5107 
5108 	/* Test CR0 */
5109 	fixed0 = rdmsr(MSR_IA32_VMX_CR0_FIXED0);
5110 	fixed1 = rdmsr(MSR_IA32_VMX_CR0_FIXED1);
5111 	test_ctl_reg("HOST_CR0", HOST_CR0, fixed0, fixed1);
5112 
5113 	/* Test CR4 */
5114 	fixed0 = rdmsr(MSR_IA32_VMX_CR4_FIXED0);
5115 	fixed1 = rdmsr(MSR_IA32_VMX_CR4_FIXED1) &
5116 		 ~(X86_CR4_SMEP | X86_CR4_SMAP);
5117 	test_ctl_reg("HOST_CR4", HOST_CR4, fixed0, fixed1);
5118 
5119 	/* Test CR3 */
5120 	cr3_saved = vmcs_read(HOST_CR3);
5121 	for (i = cpuid_maxphyaddr(); i < 64; i++) {
5122 		cr3 = cr3_saved | (1ul << i);
5123 		vmcs_write(HOST_CR3, cr3);
5124 		report_prefix_pushf("HOST_CR3 %lx", cr3);
5125 		test_vmx_vmlaunch(VMXERR_ENTRY_INVALID_HOST_STATE_FIELD,
5126 				  false);
5127 		report_prefix_pop();
5128 	}
5129 
5130 	vmcs_write(HOST_CR3, cr3_saved);
5131 }
5132 
5133 /*
5134  * Check that the virtual CPU checks the VMX Host State Area as
5135  * documented in the Intel SDM.
5136  */
5137 static void vmx_host_state_area_test(void)
5138 {
5139 	/*
5140 	 * Bit 1 of the guest's RFLAGS must be 1, or VM-entry will
5141 	 * fail due to invalid guest state, should we make it that
5142 	 * far.
5143 	 */
5144 	vmcs_write(GUEST_RFLAGS, 0);
5145 
5146 	test_host_ctl_regs();
5147 }
5148 
5149 static bool valid_vmcs_for_vmentry(void)
5150 {
5151 	struct vmcs *current_vmcs = NULL;
5152 
5153 	if (vmcs_save(&current_vmcs))
5154 		return false;
5155 
5156 	return current_vmcs && !current_vmcs->hdr.shadow_vmcs;
5157 }
5158 
5159 static void try_vmentry_in_movss_shadow(void)
5160 {
5161 	u32 vm_inst_err;
5162 	u32 flags;
5163 	bool early_failure = false;
5164 	u32 expected_flags = X86_EFLAGS_FIXED;
5165 	bool valid_vmcs = valid_vmcs_for_vmentry();
5166 
5167 	expected_flags |= valid_vmcs ? X86_EFLAGS_ZF : X86_EFLAGS_CF;
5168 
5169 	/*
5170 	 * Indirectly set VM_INST_ERR to 12 ("VMREAD/VMWRITE from/to
5171 	 * unsupported VMCS component").
5172 	 */
5173 	vmcs_write(~0u, 0);
5174 
5175 	__asm__ __volatile__ ("mov %[host_rsp], %%edx;"
5176 			      "vmwrite %%rsp, %%rdx;"
5177 			      "mov 0f, %%rax;"
5178 			      "mov %[host_rip], %%edx;"
5179 			      "vmwrite %%rax, %%rdx;"
5180 			      "mov $-1, %%ah;"
5181 			      "sahf;"
5182 			      "mov %%ss, %%ax;"
5183 			      "mov %%ax, %%ss;"
5184 			      "vmlaunch;"
5185 			      "mov $1, %[early_failure];"
5186 			      "0: lahf;"
5187 			      "movzbl %%ah, %[flags]"
5188 			      : [early_failure] "+r" (early_failure),
5189 				[flags] "=&a" (flags)
5190 			      : [host_rsp] "i" (HOST_RSP),
5191 				[host_rip] "i" (HOST_RIP)
5192 			      : "rdx", "cc", "memory");
5193 	vm_inst_err = vmcs_read(VMX_INST_ERROR);
5194 
5195 	report("Early VM-entry failure", early_failure);
5196 	report("RFLAGS[8:0] is %x (actual %x)", flags == expected_flags,
5197 	       expected_flags, flags);
5198 	if (valid_vmcs)
5199 		report("VM-instruction error is %d (actual %d)",
5200 		       vm_inst_err == VMXERR_ENTRY_EVENTS_BLOCKED_BY_MOV_SS,
5201 		       VMXERR_ENTRY_EVENTS_BLOCKED_BY_MOV_SS, vm_inst_err);
5202 }
5203 
5204 static void vmentry_movss_shadow_test(void)
5205 {
5206 	struct vmcs *orig_vmcs;
5207 
5208 	TEST_ASSERT(!vmcs_save(&orig_vmcs));
5209 
5210 	/*
5211 	 * Set the launched flag on the current VMCS to verify the correct
5212 	 * error priority, below.
5213 	 */
5214 	test_set_guest(v2_null_test_guest);
5215 	enter_guest();
5216 
5217 	/*
5218 	 * With bit 1 of the guest's RFLAGS clear, VM-entry should
5219 	 * fail due to invalid guest state (if we make it that far).
5220 	 */
5221 	vmcs_write(GUEST_RFLAGS, 0);
5222 
5223 	/*
5224 	 * "VM entry with events blocked by MOV SS" takes precedence over
5225 	 * "VMLAUNCH with non-clear VMCS."
5226 	 */
5227 	report_prefix_push("valid current-VMCS");
5228 	try_vmentry_in_movss_shadow();
5229 	report_prefix_pop();
5230 
5231 	/*
5232 	 * VMfailInvalid takes precedence over "VM entry with events
5233 	 * blocked by MOV SS."
5234 	 */
5235 	TEST_ASSERT(!vmcs_clear(orig_vmcs));
5236 	report_prefix_push("no current-VMCS");
5237 	try_vmentry_in_movss_shadow();
5238 	report_prefix_pop();
5239 
5240 	TEST_ASSERT(!make_vmcs_current(orig_vmcs));
5241 	vmcs_write(GUEST_RFLAGS, X86_EFLAGS_FIXED);
5242 }
5243 
5244 #define X86_FEATURE_PCID       (1 << 17)
5245 #define X86_FEATURE_MCE        (1 << 7)
5246 
5247 static int write_cr4_checking(unsigned long val)
5248 {
5249 	asm volatile(ASM_TRY("1f")
5250 		     "mov %0, %%cr4\n\t"
5251 		     "1:": : "r" (val));
5252 	return exception_vector();
5253 }
5254 
5255 static void vmx_cr_load_test(void)
5256 {
5257 	struct cpuid _cpuid = cpuid(1);
5258 	unsigned long cr4 = read_cr4(), cr3 = read_cr3();
5259 
5260 	if (!(_cpuid.c & X86_FEATURE_PCID)) {
5261 		report_skip("PCID not detected");
5262 		return;
5263 	}
5264 	if (!(_cpuid.d & X86_FEATURE_MCE)) {
5265 		report_skip("MCE not detected");
5266 		return;
5267 	}
5268 
5269 	TEST_ASSERT(!(cr4 & (X86_CR4_PCIDE | X86_CR4_MCE)));
5270 	TEST_ASSERT(!(cr3 & X86_CR3_PCID_MASK));
5271 
5272 	/* Enable PCID for L1. */
5273 	cr4 |= X86_CR4_PCIDE;
5274 	cr3 |= 0x1;
5275 	TEST_ASSERT(!write_cr4_checking(cr4));
5276 	write_cr3(cr3);
5277 
5278 	test_set_guest(v2_null_test_guest);
5279 	vmcs_write(HOST_CR4, cr4);
5280 	vmcs_write(HOST_CR3, cr3);
5281 	enter_guest();
5282 
5283 	/*
5284 	 * No exception is expected.
5285 	 *
5286 	 * NB. KVM loads the last guest write to CR4 into CR4 read
5287 	 *     shadow. In order to trigger an exit to KVM, we can set a
5288 	 *     bit that was zero in the above CR4 write and is owned by
5289 	 *     KVM. We choose to set CR4.MCE, which shall have no side
5290 	 *     effect because normally no guest MCE (e.g., as the result
5291 	 *     of bad memory) would happen during this test.
5292 	 */
5293 	TEST_ASSERT(!write_cr4_checking(cr4 | X86_CR4_MCE));
5294 
5295 	/* Cleanup L1 state: disable PCID. */
5296 	write_cr3(cr3 & ~X86_CR3_PCID_MASK);
5297 	TEST_ASSERT(!write_cr4_checking(cr4 & ~X86_CR4_PCIDE));
5298 }
5299 
5300 static void vmx_nm_test_guest(void)
5301 {
5302 	write_cr0(read_cr0() | X86_CR0_TS);
5303 	asm volatile("fnop");
5304 }
5305 
5306 static void check_nm_exit(const char *test)
5307 {
5308 	u32 reason = vmcs_read(EXI_REASON);
5309 	u32 intr_info = vmcs_read(EXI_INTR_INFO);
5310 	const u32 expected = INTR_INFO_VALID_MASK | INTR_TYPE_HARD_EXCEPTION |
5311 		NM_VECTOR;
5312 
5313 	report("%s", reason == VMX_EXC_NMI && intr_info == expected, test);
5314 }
5315 
5316 /*
5317  * This test checks that:
5318  *
5319  * (a) If L2 launches with CR0.TS clear, but later sets CR0.TS, then
5320  *     a subsequent #NM VM-exit is reflected to L1.
5321  *
5322  * (b) If L2 launches with CR0.TS clear and CR0.EM set, then a
5323  *     subsequent #NM VM-exit is reflected to L1.
5324  */
5325 static void vmx_nm_test(void)
5326 {
5327 	unsigned long cr0 = read_cr0();
5328 
5329 	test_set_guest(vmx_nm_test_guest);
5330 
5331 	/*
5332 	 * L1 wants to intercept #NM exceptions encountered in L2.
5333 	 */
5334 	vmcs_write(EXC_BITMAP, 1 << NM_VECTOR);
5335 
5336 	/*
5337 	 * Launch L2 with CR0.TS clear, but don't claim host ownership of
5338 	 * any CR0 bits. L2 will set CR0.TS and then try to execute fnop,
5339 	 * which will raise #NM. L0 should reflect the #NM VM-exit to L1.
5340 	 */
5341 	vmcs_write(CR0_MASK, 0);
5342 	vmcs_write(GUEST_CR0, cr0 & ~X86_CR0_TS);
5343 	enter_guest();
5344 	check_nm_exit("fnop with CR0.TS set in L2 triggers #NM VM-exit to L1");
5345 
5346 	/*
5347 	 * Re-enter L2 at the fnop instruction, with CR0.TS clear but
5348 	 * CR0.EM set. The fnop will still raise #NM, and L0 should
5349 	 * reflect the #NM VM-exit to L1.
5350 	 */
5351 	vmcs_write(GUEST_CR0, (cr0 & ~X86_CR0_TS) | X86_CR0_EM);
5352 	enter_guest();
5353 	check_nm_exit("fnop with CR0.EM set in L2 triggers #NM VM-exit to L1");
5354 
5355 	/*
5356 	 * Re-enter L2 at the fnop instruction, with both CR0.TS and
5357 	 * CR0.EM clear. There will be no #NM, and the L2 guest should
5358 	 * exit normally.
5359 	 */
5360 	vmcs_write(GUEST_CR0, cr0 & ~(X86_CR0_TS | X86_CR0_EM));
5361 	enter_guest();
5362 }
5363 
5364 bool vmx_pending_event_ipi_fired;
5365 static void vmx_pending_event_ipi_isr(isr_regs_t *regs)
5366 {
5367 	vmx_pending_event_ipi_fired = true;
5368 	eoi();
5369 }
5370 
5371 bool vmx_pending_event_guest_run;
5372 static void vmx_pending_event_guest(void)
5373 {
5374 	vmcall();
5375 	vmx_pending_event_guest_run = true;
5376 }
5377 
5378 static void vmx_pending_event_test_core(bool guest_hlt)
5379 {
5380 	int ipi_vector = 0xf1;
5381 
5382 	vmx_pending_event_ipi_fired = false;
5383 	handle_irq(ipi_vector, vmx_pending_event_ipi_isr);
5384 
5385 	vmx_pending_event_guest_run = false;
5386 	test_set_guest(vmx_pending_event_guest);
5387 
5388 	vmcs_set_bits(PIN_CONTROLS, PIN_EXTINT);
5389 
5390 	enter_guest();
5391 	skip_exit_vmcall();
5392 
5393 	if (guest_hlt)
5394 		vmcs_write(GUEST_ACTV_STATE, ACTV_HLT);
5395 
5396 	irq_disable();
5397 	apic_icr_write(APIC_DEST_SELF | APIC_DEST_PHYSICAL |
5398 				   APIC_DM_FIXED | ipi_vector,
5399 				   0);
5400 
5401 	enter_guest();
5402 
5403 	assert_exit_reason(VMX_EXTINT);
5404 	report("Guest did not run before host received IPI",
5405 		   !vmx_pending_event_guest_run);
5406 
5407 	irq_enable();
5408 	asm volatile ("nop");
5409 	irq_disable();
5410 	report("Got pending interrupt after IRQ enabled",
5411 		   vmx_pending_event_ipi_fired);
5412 
5413 	if (guest_hlt)
5414 		vmcs_write(GUEST_ACTV_STATE, ACTV_ACTIVE);
5415 
5416 	enter_guest();
5417 	report("Guest finished running when no interrupt",
5418 		   vmx_pending_event_guest_run);
5419 }
5420 
5421 static void vmx_pending_event_test(void)
5422 {
5423 	vmx_pending_event_test_core(false);
5424 }
5425 
5426 static void vmx_pending_event_hlt_test(void)
5427 {
5428 	vmx_pending_event_test_core(true);
5429 }
5430 
5431 static int vmx_window_test_ud_count;
5432 
5433 static void vmx_window_test_ud_handler(struct ex_regs *regs)
5434 {
5435 	vmx_window_test_ud_count++;
5436 }
5437 
5438 static void vmx_nmi_window_test_guest(void)
5439 {
5440 	handle_exception(UD_VECTOR, vmx_window_test_ud_handler);
5441 
5442 	asm volatile("vmcall\n\t"
5443 		     "nop\n\t");
5444 
5445 	handle_exception(UD_VECTOR, NULL);
5446 }
5447 
5448 static void verify_nmi_window_exit(u64 rip)
5449 {
5450 	u32 exit_reason = vmcs_read(EXI_REASON);
5451 
5452 	report("Exit reason (%d) is 'NMI window'",
5453 	       exit_reason == VMX_NMI_WINDOW, exit_reason);
5454 	report("RIP (%#lx) is %#lx", vmcs_read(GUEST_RIP) == rip,
5455 	       vmcs_read(GUEST_RIP), rip);
5456 	report("Activity state (%ld) is 'ACTIVE'",
5457 	       vmcs_read(GUEST_ACTV_STATE) == ACTV_ACTIVE,
5458 	       vmcs_read(GUEST_ACTV_STATE));
5459 }
5460 
5461 static void vmx_nmi_window_test(void)
5462 {
5463 	u64 nop_addr;
5464 	void *ud_fault_addr = get_idt_addr(&boot_idt[UD_VECTOR]);
5465 
5466 	if (!(ctrl_pin_rev.clr & PIN_VIRT_NMI)) {
5467 		report_skip("CPU does not support the \"Virtual NMIs\" VM-execution control.");
5468 		return;
5469 	}
5470 
5471 	if (!(ctrl_cpu_rev[0].clr & CPU_NMI_WINDOW)) {
5472 		report_skip("CPU does not support the \"NMI-window exiting\" VM-execution control.");
5473 		return;
5474 	}
5475 
5476 	vmx_window_test_ud_count = 0;
5477 
5478 	report_prefix_push("NMI-window");
5479 	test_set_guest(vmx_nmi_window_test_guest);
5480 	vmcs_set_bits(PIN_CONTROLS, PIN_VIRT_NMI);
5481 	enter_guest();
5482 	skip_exit_vmcall();
5483 	nop_addr = vmcs_read(GUEST_RIP);
5484 
5485 	/*
5486 	 * Ask for "NMI-window exiting," and expect an immediate VM-exit.
5487 	 * RIP will not advance.
5488 	 */
5489 	report_prefix_push("active, no blocking");
5490 	vmcs_set_bits(CPU_EXEC_CTRL0, CPU_NMI_WINDOW);
5491 	enter_guest();
5492 	verify_nmi_window_exit(nop_addr);
5493 	report_prefix_pop();
5494 
5495 	/*
5496 	 * Ask for "NMI-window exiting" in a MOV-SS shadow, and expect
5497 	 * a VM-exit on the next instruction after the nop. (The nop
5498 	 * is one byte.)
5499 	 */
5500 	report_prefix_push("active, blocking by MOV-SS");
5501 	vmcs_write(GUEST_INTR_STATE, GUEST_INTR_STATE_MOVSS);
5502 	enter_guest();
5503 	verify_nmi_window_exit(nop_addr + 1);
5504 	report_prefix_pop();
5505 
5506 	/*
5507 	 * Ask for "NMI-window exiting" (with event injection), and
5508 	 * expect a VM-exit after the event is injected. (RIP should
5509 	 * be at the address specified in the IDT entry for #UD.)
5510 	 */
5511 	report_prefix_push("active, no blocking, injecting #UD");
5512 	vmcs_write(ENT_INTR_INFO,
5513 		   INTR_INFO_VALID_MASK | INTR_TYPE_HARD_EXCEPTION | UD_VECTOR);
5514 	enter_guest();
5515 	verify_nmi_window_exit((u64)ud_fault_addr);
5516 	report_prefix_pop();
5517 
5518 	/*
5519 	 * Ask for "NMI-window exiting" with NMI blocking, and expect
5520 	 * a VM-exit after the next IRET (i.e. after the #UD handler
5521 	 * returns). So, RIP should be back at one byte past the nop.
5522 	 */
5523 	report_prefix_push("active, blocking by NMI");
5524 	vmcs_write(GUEST_INTR_STATE, GUEST_INTR_STATE_NMI);
5525 	enter_guest();
5526 	verify_nmi_window_exit(nop_addr + 1);
5527 	report("#UD handler executed once (actual %d times)",
5528 	       vmx_window_test_ud_count == 1,
5529 	       vmx_window_test_ud_count);
5530 	report_prefix_pop();
5531 
5532 	if (!(rdmsr(MSR_IA32_VMX_MISC) & (1 << 6))) {
5533 		report_skip("CPU does not support activity state HLT.");
5534 	} else {
5535 		/*
5536 		 * Ask for "NMI-window exiting" when entering activity
5537 		 * state HLT, and expect an immediate VM-exit. RIP is
5538 		 * still one byte past the nop.
5539 		 */
5540 		report_prefix_push("halted, no blocking");
5541 		vmcs_write(GUEST_ACTV_STATE, ACTV_HLT);
5542 		enter_guest();
5543 		verify_nmi_window_exit(nop_addr + 1);
5544 		report_prefix_pop();
5545 
5546 		/*
5547 		 * Ask for "NMI-window exiting" when entering activity
5548 		 * state HLT (with event injection), and expect a
5549 		 * VM-exit after the event is injected. (RIP should be
5550 		 * at the address specified in the IDT entry for #UD.)
5551 		 */
5552 		report_prefix_push("halted, no blocking, injecting #UD");
5553 		vmcs_write(GUEST_ACTV_STATE, ACTV_HLT);
5554 		vmcs_write(ENT_INTR_INFO,
5555 			   INTR_INFO_VALID_MASK | INTR_TYPE_HARD_EXCEPTION |
5556 			   UD_VECTOR);
5557 		enter_guest();
5558 		verify_nmi_window_exit((u64)ud_fault_addr);
5559 		report_prefix_pop();
5560 	}
5561 
5562 	vmcs_clear_bits(CPU_EXEC_CTRL0, CPU_NMI_WINDOW);
5563 	enter_guest();
5564 	report_prefix_pop();
5565 }
5566 
5567 static void vmx_intr_window_test_guest(void)
5568 {
5569 	handle_exception(UD_VECTOR, vmx_window_test_ud_handler);
5570 
5571 	/*
5572 	 * The two consecutive STIs are to ensure that only the first
5573 	 * one has a shadow. Note that NOP and STI are one byte
5574 	 * instructions.
5575 	 */
5576 	asm volatile("vmcall\n\t"
5577 		     "nop\n\t"
5578 		     "sti\n\t"
5579 		     "sti\n\t");
5580 
5581 	handle_exception(UD_VECTOR, NULL);
5582 }
5583 
5584 static void verify_intr_window_exit(u64 rip)
5585 {
5586 	u32 exit_reason = vmcs_read(EXI_REASON);
5587 
5588 	report("Exit reason (%d) is 'interrupt window'",
5589 	       exit_reason == VMX_INTR_WINDOW, exit_reason);
5590 	report("RIP (%#lx) is %#lx", vmcs_read(GUEST_RIP) == rip,
5591 	       vmcs_read(GUEST_RIP), rip);
5592 	report("Activity state (%ld) is 'ACTIVE'",
5593 	       vmcs_read(GUEST_ACTV_STATE) == ACTV_ACTIVE,
5594 	       vmcs_read(GUEST_ACTV_STATE));
5595 }
5596 
5597 static void vmx_intr_window_test(void)
5598 {
5599 	u64 vmcall_addr;
5600 	u64 nop_addr;
5601 	unsigned int orig_ud_gate_type;
5602 	void *ud_fault_addr = get_idt_addr(&boot_idt[UD_VECTOR]);
5603 
5604 	if (!(ctrl_cpu_rev[0].clr & CPU_INTR_WINDOW)) {
5605 		report_skip("CPU does not support the \"interrupt-window exiting\" VM-execution control.");
5606 		return;
5607 	}
5608 
5609 	/*
5610 	 * Change the IDT entry for #UD from interrupt gate to trap gate,
5611 	 * so that it won't clear RFLAGS.IF. We don't want interrupts to
5612 	 * be disabled after vectoring a #UD.
5613 	 */
5614 	orig_ud_gate_type = boot_idt[UD_VECTOR].type;
5615 	boot_idt[UD_VECTOR].type = 15;
5616 
5617 	report_prefix_push("interrupt-window");
5618 	test_set_guest(vmx_intr_window_test_guest);
5619 	enter_guest();
5620 	assert_exit_reason(VMX_VMCALL);
5621 	vmcall_addr = vmcs_read(GUEST_RIP);
5622 
5623 	/*
5624 	 * Ask for "interrupt-window exiting" with RFLAGS.IF set and
5625 	 * no blocking; expect an immediate VM-exit. Note that we have
5626 	 * not advanced past the vmcall instruction yet, so RIP should
5627 	 * point to the vmcall instruction.
5628 	 */
5629 	report_prefix_push("active, no blocking, RFLAGS.IF=1");
5630 	vmcs_set_bits(CPU_EXEC_CTRL0, CPU_INTR_WINDOW);
5631 	vmcs_write(GUEST_RFLAGS, X86_EFLAGS_FIXED | X86_EFLAGS_IF);
5632 	enter_guest();
5633 	verify_intr_window_exit(vmcall_addr);
5634 	report_prefix_pop();
5635 
5636 	/*
5637 	 * Ask for "interrupt-window exiting" (with event injection)
5638 	 * with RFLAGS.IF set and no blocking; expect a VM-exit after
5639 	 * the event is injected. That is, RIP should should be at the
5640 	 * address specified in the IDT entry for #UD.
5641 	 */
5642 	report_prefix_push("active, no blocking, RFLAGS.IF=1, injecting #UD");
5643 	vmcs_write(ENT_INTR_INFO,
5644 		   INTR_INFO_VALID_MASK | INTR_TYPE_HARD_EXCEPTION | UD_VECTOR);
5645 	vmcall_addr = vmcs_read(GUEST_RIP);
5646 	enter_guest();
5647 	verify_intr_window_exit((u64)ud_fault_addr);
5648 	report_prefix_pop();
5649 
5650 	/*
5651 	 * Let the L2 guest run through the IRET, back to the VMCALL.
5652 	 * We have to clear the "interrupt-window exiting"
5653 	 * VM-execution control, or it would just keep causing
5654 	 * VM-exits. Then, advance past the VMCALL and set the
5655 	 * "interrupt-window exiting" VM-execution control again.
5656 	 */
5657 	vmcs_clear_bits(CPU_EXEC_CTRL0, CPU_INTR_WINDOW);
5658 	enter_guest();
5659 	skip_exit_vmcall();
5660 	nop_addr = vmcs_read(GUEST_RIP);
5661 	vmcs_set_bits(CPU_EXEC_CTRL0, CPU_INTR_WINDOW);
5662 
5663 	/*
5664 	 * Ask for "interrupt-window exiting" in a MOV-SS shadow with
5665 	 * RFLAGS.IF set, and expect a VM-exit on the next
5666 	 * instruction. (NOP is one byte.)
5667 	 */
5668 	report_prefix_push("active, blocking by MOV-SS, RFLAGS.IF=1");
5669 	vmcs_write(GUEST_INTR_STATE, GUEST_INTR_STATE_MOVSS);
5670 	enter_guest();
5671 	verify_intr_window_exit(nop_addr + 1);
5672 	report_prefix_pop();
5673 
5674 	/*
5675 	 * Back up to the NOP and ask for "interrupt-window exiting"
5676 	 * in an STI shadow with RFLAGS.IF set, and expect a VM-exit
5677 	 * on the next instruction. (NOP is one byte.)
5678 	 */
5679 	report_prefix_push("active, blocking by STI, RFLAGS.IF=1");
5680 	vmcs_write(GUEST_RIP, nop_addr);
5681 	vmcs_write(GUEST_INTR_STATE, GUEST_INTR_STATE_STI);
5682 	enter_guest();
5683 	verify_intr_window_exit(nop_addr + 1);
5684 	report_prefix_pop();
5685 
5686 	/*
5687 	 * Ask for "interrupt-window exiting" with RFLAGS.IF clear,
5688 	 * and expect a VM-exit on the instruction following the STI
5689 	 * shadow. Only the first STI (which is one byte past the NOP)
5690 	 * should have a shadow. The second STI (which is two bytes
5691 	 * past the NOP) has no shadow. Therefore, the interrupt
5692 	 * window opens at three bytes past the NOP.
5693 	 */
5694 	report_prefix_push("active, RFLAGS.IF = 0");
5695 	vmcs_write(GUEST_RFLAGS, X86_EFLAGS_FIXED);
5696 	enter_guest();
5697 	verify_intr_window_exit(nop_addr + 3);
5698 	report_prefix_pop();
5699 
5700 	if (!(rdmsr(MSR_IA32_VMX_MISC) & (1 << 6))) {
5701 		report_skip("CPU does not support activity state HLT.");
5702 	} else {
5703 		/*
5704 		 * Ask for "interrupt-window exiting" when entering
5705 		 * activity state HLT, and expect an immediate
5706 		 * VM-exit. RIP is still three bytes past the nop.
5707 		 */
5708 		report_prefix_push("halted, no blocking");
5709 		vmcs_write(GUEST_ACTV_STATE, ACTV_HLT);
5710 		enter_guest();
5711 		verify_intr_window_exit(nop_addr + 3);
5712 		report_prefix_pop();
5713 
5714 		/*
5715 		 * Ask for "interrupt-window exiting" when entering
5716 		 * activity state HLT (with event injection), and
5717 		 * expect a VM-exit after the event is injected. That
5718 		 * is, RIP should should be at the address specified
5719 		 * in the IDT entry for #UD.
5720 		 */
5721 		report_prefix_push("halted, no blocking, injecting #UD");
5722 		vmcs_write(GUEST_ACTV_STATE, ACTV_HLT);
5723 		vmcs_write(ENT_INTR_INFO,
5724 			   INTR_INFO_VALID_MASK | INTR_TYPE_HARD_EXCEPTION |
5725 			   UD_VECTOR);
5726 		enter_guest();
5727 		verify_intr_window_exit((u64)ud_fault_addr);
5728 		report_prefix_pop();
5729 	}
5730 
5731 	boot_idt[UD_VECTOR].type = orig_ud_gate_type;
5732 	vmcs_clear_bits(CPU_EXEC_CTRL0, CPU_INTR_WINDOW);
5733 	enter_guest();
5734 	report_prefix_pop();
5735 }
5736 
5737 #define GUEST_TSC_OFFSET (1u << 30)
5738 
5739 static u64 guest_tsc;
5740 
5741 static void vmx_store_tsc_test_guest(void)
5742 {
5743 	guest_tsc = rdtsc();
5744 }
5745 
5746 /*
5747  * This test ensures that when IA32_TSC is in the VM-exit MSR-store
5748  * list, the value saved is not subject to the TSC offset that is
5749  * applied to RDTSC/RDTSCP/RDMSR(IA32_TSC) in guest execution.
5750  */
5751 static void vmx_store_tsc_test(void)
5752 {
5753 	struct vmx_msr_entry msr_entry = { .index = MSR_IA32_TSC };
5754 	u64 low, high;
5755 
5756 	if (!(ctrl_cpu_rev[0].clr & CPU_USE_TSC_OFFSET)) {
5757 		report_skip("'Use TSC offsetting' not supported");
5758 		return;
5759 	}
5760 
5761 	test_set_guest(vmx_store_tsc_test_guest);
5762 
5763 	vmcs_set_bits(CPU_EXEC_CTRL0, CPU_USE_TSC_OFFSET);
5764 	vmcs_write(EXI_MSR_ST_CNT, 1);
5765 	vmcs_write(EXIT_MSR_ST_ADDR, virt_to_phys(&msr_entry));
5766 	vmcs_write(TSC_OFFSET, GUEST_TSC_OFFSET);
5767 
5768 	low = rdtsc();
5769 	enter_guest();
5770 	high = rdtsc();
5771 
5772 	report("RDTSC value in the guest (%lu) is in range [%lu, %lu]",
5773 	       low + GUEST_TSC_OFFSET <= guest_tsc &&
5774 	       guest_tsc <= high + GUEST_TSC_OFFSET,
5775 	       guest_tsc, low + GUEST_TSC_OFFSET, high + GUEST_TSC_OFFSET);
5776 	report("IA32_TSC value saved in the VM-exit MSR-store list (%lu) is in range [%lu, %lu]",
5777 	       low <= msr_entry.value && msr_entry.value <= high,
5778 	       msr_entry.value, low, high);
5779 }
5780 
5781 static void vmx_db_test_guest(void)
5782 {
5783 	/*
5784 	 * For a hardware generated single-step #DB.
5785 	 */
5786 	asm volatile("vmcall;"
5787 		     "nop;"
5788 		     ".Lpost_nop:");
5789 	/*
5790 	 * ...in a MOVSS shadow, with pending debug exceptions.
5791 	 */
5792 	asm volatile("vmcall;"
5793 		     "nop;"
5794 		     ".Lpost_movss_nop:");
5795 	/*
5796 	 * For an L0 synthesized single-step #DB. (L0 intercepts WBINVD and
5797 	 * emulates it in software.)
5798 	 */
5799 	asm volatile("vmcall;"
5800 		     "wbinvd;"
5801 		     ".Lpost_wbinvd:");
5802 	/*
5803 	 * ...in a MOVSS shadow, with pending debug exceptions.
5804 	 */
5805 	asm volatile("vmcall;"
5806 		     "wbinvd;"
5807 		     ".Lpost_movss_wbinvd:");
5808 	/*
5809 	 * For a hardware generated single-step #DB in a transactional region.
5810 	 */
5811 	asm volatile("vmcall;"
5812 		     ".Lxbegin: xbegin .Lskip_rtm;"
5813 		     "xend;"
5814 		     ".Lskip_rtm:");
5815 }
5816 
5817 /*
5818  * Clear the pending debug exceptions and RFLAGS.TF and re-enter
5819  * L2. No #DB is delivered and L2 continues to the next point of
5820  * interest.
5821  */
5822 static void dismiss_db(void)
5823 {
5824 	vmcs_write(GUEST_PENDING_DEBUG, 0);
5825 	vmcs_write(GUEST_RFLAGS, X86_EFLAGS_FIXED);
5826 	enter_guest();
5827 }
5828 
5829 /*
5830  * Check a variety of VMCS fields relevant to an intercepted #DB exception.
5831  * Then throw away the #DB exception and resume L2.
5832  */
5833 static void check_db_exit(bool xfail_qual, bool xfail_dr6, bool xfail_pdbg,
5834 			  void *expected_rip, u64 expected_exit_qual,
5835 			  u64 expected_dr6)
5836 {
5837 	u32 reason = vmcs_read(EXI_REASON);
5838 	u32 intr_info = vmcs_read(EXI_INTR_INFO);
5839 	u64 exit_qual = vmcs_read(EXI_QUALIFICATION);
5840 	u64 guest_rip = vmcs_read(GUEST_RIP);
5841 	u64 guest_pending_dbg = vmcs_read(GUEST_PENDING_DEBUG);
5842 	u64 dr6 = read_dr6();
5843 	const u32 expected_intr_info = INTR_INFO_VALID_MASK |
5844 		INTR_TYPE_HARD_EXCEPTION | DB_VECTOR;
5845 
5846 	report("Expected #DB VM-exit",
5847 	       reason == VMX_EXC_NMI && intr_info == expected_intr_info);
5848 	report("Expected RIP %p (actual %lx)", (u64)expected_rip == guest_rip,
5849 	       expected_rip, guest_rip);
5850 	report_xfail("Expected pending debug exceptions 0 (actual %lx)",
5851 		     xfail_pdbg, 0 == guest_pending_dbg, guest_pending_dbg);
5852 	report_xfail("Expected exit qualification %lx (actual %lx)", xfail_qual,
5853 		     expected_exit_qual == exit_qual,
5854 		     expected_exit_qual, exit_qual);
5855 	report_xfail("Expected DR6 %lx (actual %lx)", xfail_dr6,
5856 		     expected_dr6 == dr6, expected_dr6, dr6);
5857 	dismiss_db();
5858 }
5859 
5860 /*
5861  * Assuming the guest has just exited on a VMCALL instruction, skip
5862  * over the vmcall, and set the guest's RFLAGS.TF in the VMCS. If
5863  * pending debug exceptions are non-zero, set the VMCS up as if the
5864  * previous instruction was a MOVSS that generated the indicated
5865  * pending debug exceptions. Then enter L2.
5866  */
5867 static void single_step_guest(const char *test_name, u64 starting_dr6,
5868 			      u64 pending_debug_exceptions)
5869 {
5870 	printf("\n%s\n", test_name);
5871 	skip_exit_vmcall();
5872 	write_dr6(starting_dr6);
5873 	vmcs_write(GUEST_RFLAGS, X86_EFLAGS_FIXED | X86_EFLAGS_TF);
5874 	if (pending_debug_exceptions) {
5875 		vmcs_write(GUEST_PENDING_DEBUG, pending_debug_exceptions);
5876 		vmcs_write(GUEST_INTR_STATE, GUEST_INTR_STATE_MOVSS);
5877 	}
5878 	enter_guest();
5879 }
5880 
5881 /*
5882  * When L1 intercepts #DB, verify that a single-step trap clears
5883  * pending debug exceptions, populates the exit qualification field
5884  * properly, and that DR6 is not prematurely clobbered. In a
5885  * (simulated) MOVSS shadow, make sure that the pending debug
5886  * exception bits are properly accumulated into the exit qualification
5887  * field.
5888  */
5889 static void vmx_db_test(void)
5890 {
5891 	/*
5892 	 * We are going to set a few arbitrary bits in DR6 to verify that
5893 	 * (a) DR6 is not modified by an intercepted #DB, and
5894 	 * (b) stale bits in DR6 (DR6.BD, in particular) don't leak into
5895          *     the exit qualification field for a subsequent #DB exception.
5896 	 */
5897 	const u64 starting_dr6 = DR6_RESERVED | BIT(13) | DR_TRAP3 | DR_TRAP1;
5898 	extern char post_nop asm(".Lpost_nop");
5899 	extern char post_movss_nop asm(".Lpost_movss_nop");
5900 	extern char post_wbinvd asm(".Lpost_wbinvd");
5901 	extern char post_movss_wbinvd asm(".Lpost_movss_wbinvd");
5902 	extern char xbegin asm(".Lxbegin");
5903 	extern char skip_rtm asm(".Lskip_rtm");
5904 
5905 	/*
5906 	 * L1 wants to intercept #DB exceptions encountered in L2.
5907 	 */
5908 	vmcs_write(EXC_BITMAP, BIT(DB_VECTOR));
5909 
5910 	/*
5911 	 * Start L2 and run it up to the first point of interest.
5912 	 */
5913 	test_set_guest(vmx_db_test_guest);
5914 	enter_guest();
5915 
5916 	/*
5917 	 * Hardware-delivered #DB trap for single-step sets the
5918 	 * standard that L0 has to follow for emulated instructions.
5919 	 */
5920 	single_step_guest("Hardware delivered single-step", starting_dr6, 0);
5921 	check_db_exit(false, false, false, &post_nop, DR_STEP, starting_dr6);
5922 
5923 	/*
5924 	 * Hardware-delivered #DB trap for single-step in MOVSS shadow
5925 	 * also sets the standard that L0 has to follow for emulated
5926 	 * instructions. Here, we establish the VMCS pending debug
5927 	 * exceptions to indicate that the simulated MOVSS triggered a
5928 	 * data breakpoint as well as the single-step trap.
5929 	 */
5930 	single_step_guest("Hardware delivered single-step in MOVSS shadow",
5931 			  starting_dr6, BIT(12) | DR_STEP | DR_TRAP0 );
5932 	check_db_exit(false, false, false, &post_movss_nop, DR_STEP | DR_TRAP0,
5933 		      starting_dr6);
5934 
5935 	/*
5936 	 * L0 synthesized #DB trap for single-step is buggy, because
5937 	 * kvm (a) clobbers DR6 too early, and (b) tries its best to
5938 	 * reconstitute the exit qualification from the prematurely
5939 	 * modified DR6, but fails miserably.
5940 	 */
5941 	single_step_guest("Software synthesized single-step", starting_dr6, 0);
5942 	check_db_exit(true, true, false, &post_wbinvd, DR_STEP, starting_dr6);
5943 
5944 	/*
5945 	 * L0 synthesized #DB trap for single-step in MOVSS shadow is
5946 	 * even worse, because L0 also leaves the pending debug
5947 	 * exceptions in the VMCS instead of accumulating them into
5948 	 * the exit qualification field for the #DB exception.
5949 	 */
5950 	single_step_guest("Software synthesized single-step in MOVSS shadow",
5951 			  starting_dr6, BIT(12) | DR_STEP | DR_TRAP0);
5952 	check_db_exit(true, true, true, &post_movss_wbinvd, DR_STEP | DR_TRAP0,
5953 		      starting_dr6);
5954 
5955 	/*
5956 	 * Optional RTM test for hardware that supports RTM, to
5957 	 * demonstrate that the current volume 3 of the SDM
5958 	 * (325384-067US), table 27-1 is incorrect. Bit 16 of the exit
5959 	 * qualification for debug exceptions is not reserved. It is
5960 	 * set to 1 if a debug exception (#DB) or a breakpoint
5961 	 * exception (#BP) occurs inside an RTM region while advanced
5962 	 * debugging of RTM transactional regions is enabled.
5963 	 */
5964 	if (cpuid(7).b & BIT(11)) {
5965 		vmcs_write(ENT_CONTROLS,
5966 			   vmcs_read(ENT_CONTROLS) | ENT_LOAD_DBGCTLS);
5967 		/*
5968 		 * Set DR7.RTM[bit 11] and IA32_DEBUGCTL.RTM[bit 15]
5969 		 * in the guest to enable advanced debugging of RTM
5970 		 * transactional regions.
5971 		 */
5972 		vmcs_write(GUEST_DR7, BIT(11));
5973 		vmcs_write(GUEST_DEBUGCTL, BIT(15));
5974 		single_step_guest("Hardware delivered single-step in "
5975 				  "transactional region", starting_dr6, 0);
5976 		check_db_exit(false, false, false, &xbegin, BIT(16),
5977 			      starting_dr6);
5978 	} else {
5979 		vmcs_write(GUEST_RIP, (u64)&skip_rtm);
5980 		enter_guest();
5981 	}
5982 }
5983 
5984 static bool cpu_has_apicv(void)
5985 {
5986 	return ((ctrl_cpu_rev[1].clr & CPU_APIC_REG_VIRT) &&
5987 		(ctrl_cpu_rev[1].clr & CPU_VINTD) &&
5988 		(ctrl_pin_rev.clr & PIN_POST_INTR));
5989 }
5990 
5991 static void enable_vid(void)
5992 {
5993 	void *virtual_apic_page;
5994 
5995 	assert(cpu_has_apicv());
5996 
5997 	disable_intercept_for_x2apic_msrs();
5998 
5999 	virtual_apic_page = alloc_page();
6000 	vmcs_write(APIC_VIRT_ADDR, (u64)virtual_apic_page);
6001 
6002 	vmcs_set_bits(PIN_CONTROLS, PIN_EXTINT);
6003 
6004 	vmcs_write(EOI_EXIT_BITMAP0, 0x0);
6005 	vmcs_write(EOI_EXIT_BITMAP1, 0x0);
6006 	vmcs_write(EOI_EXIT_BITMAP2, 0x0);
6007 	vmcs_write(EOI_EXIT_BITMAP3, 0x0);
6008 
6009 	vmcs_set_bits(CPU_EXEC_CTRL0, CPU_SECONDARY | CPU_TPR_SHADOW);
6010 	vmcs_set_bits(CPU_EXEC_CTRL1, CPU_VINTD | CPU_VIRT_X2APIC);
6011 }
6012 
6013 static void trigger_ioapic_scan_thread(void *data)
6014 {
6015 	/* Wait until other CPU entered L2 */
6016 	while (vmx_get_test_stage() != 1)
6017 		;
6018 
6019 	/* Trigger ioapic scan */
6020 	ioapic_set_redir(0xf, 0x79, TRIGGER_LEVEL);
6021 	vmx_set_test_stage(2);
6022 }
6023 
6024 static void irq_79_handler_guest(isr_regs_t *regs)
6025 {
6026 	eoi();
6027 
6028 	/* L1 expects vmexit on VMX_VMCALL and not VMX_EOI_INDUCED */
6029 	vmcall();
6030 }
6031 
6032 /*
6033  * Constant for num of busy-loop iterations after which
6034  * a timer interrupt should have happened in host
6035  */
6036 #define TIMER_INTERRUPT_DELAY 100000000
6037 
6038 static void vmx_eoi_bitmap_ioapic_scan_test_guest(void)
6039 {
6040 	handle_irq(0x79, irq_79_handler_guest);
6041 	irq_enable();
6042 
6043 	/* Signal to L1 CPU to trigger ioapic scan */
6044 	vmx_set_test_stage(1);
6045 	/* Wait until L1 CPU to trigger ioapic scan */
6046 	while (vmx_get_test_stage() != 2)
6047 		;
6048 
6049 	/*
6050 	 * Wait for L0 timer interrupt to be raised while we run in L2
6051 	 * such that L0 will process the IOAPIC scan request before
6052 	 * resuming L2
6053 	 */
6054 	delay(TIMER_INTERRUPT_DELAY);
6055 
6056 	asm volatile ("int $0x79");
6057 }
6058 
6059 static void vmx_eoi_bitmap_ioapic_scan_test(void)
6060 {
6061 	if (!cpu_has_apicv() || (cpu_count() < 2)) {
6062 		report_skip(__func__);
6063 		return;
6064 	}
6065 
6066 	enable_vid();
6067 
6068 	on_cpu_async(1, trigger_ioapic_scan_thread, NULL);
6069 	test_set_guest(vmx_eoi_bitmap_ioapic_scan_test_guest);
6070 
6071 	/*
6072 	 * Launch L2.
6073 	 * We expect the exit reason to be VMX_VMCALL (and not EOI INDUCED).
6074 	 * In case the reason isn't VMX_VMCALL, the asserion inside
6075 	 * skip_exit_vmcall() will fail.
6076 	 */
6077 	enter_guest();
6078 	skip_exit_vmcall();
6079 
6080 	/* Let L2 finish */
6081 	enter_guest();
6082 	report(__func__, 1);
6083 }
6084 
6085 #define HLT_WITH_RVI_VECTOR		(0xf1)
6086 
6087 bool vmx_hlt_with_rvi_guest_isr_fired;
6088 static void vmx_hlt_with_rvi_guest_isr(isr_regs_t *regs)
6089 {
6090 	vmx_hlt_with_rvi_guest_isr_fired = true;
6091 	eoi();
6092 }
6093 
6094 static void vmx_hlt_with_rvi_guest(void)
6095 {
6096 	handle_irq(HLT_WITH_RVI_VECTOR, vmx_hlt_with_rvi_guest_isr);
6097 
6098 	irq_enable();
6099 	asm volatile ("nop");
6100 
6101 	vmcall();
6102 }
6103 
6104 static void vmx_hlt_with_rvi_test(void)
6105 {
6106 	if (!cpu_has_apicv()) {
6107 		report_skip(__func__);
6108 		return;
6109 	}
6110 
6111 	enable_vid();
6112 
6113 	vmx_hlt_with_rvi_guest_isr_fired = false;
6114 	test_set_guest(vmx_hlt_with_rvi_guest);
6115 
6116 	enter_guest();
6117 	skip_exit_vmcall();
6118 
6119 	vmcs_write(GUEST_ACTV_STATE, ACTV_HLT);
6120 	vmcs_write(GUEST_INT_STATUS, HLT_WITH_RVI_VECTOR);
6121 	enter_guest();
6122 
6123 	report("Interrupt raised in guest", vmx_hlt_with_rvi_guest_isr_fired);
6124 }
6125 
6126 static void set_irq_line_thread(void *data)
6127 {
6128 	/* Wait until other CPU entered L2 */
6129 	while (vmx_get_test_stage() != 1)
6130 		;
6131 
6132 	/* Set irq-line 0xf to raise vector 0x78 for vCPU 0 */
6133 	ioapic_set_redir(0xf, 0x78, TRIGGER_LEVEL);
6134 	vmx_set_test_stage(2);
6135 }
6136 
6137 static bool irq_78_handler_vmcall_before_eoi;
6138 static void irq_78_handler_guest(isr_regs_t *regs)
6139 {
6140 	set_irq_line(0xf, 0);
6141 	if (irq_78_handler_vmcall_before_eoi)
6142 		vmcall();
6143 	eoi();
6144 	vmcall();
6145 }
6146 
6147 static void vmx_apic_passthrough_guest(void)
6148 {
6149 	handle_irq(0x78, irq_78_handler_guest);
6150 	irq_enable();
6151 
6152 	/* If requested, wait for other CPU to trigger ioapic scan */
6153 	if (vmx_get_test_stage() < 1) {
6154 		vmx_set_test_stage(1);
6155 		while (vmx_get_test_stage() != 2)
6156 			;
6157 	}
6158 
6159 	set_irq_line(0xf, 1);
6160 }
6161 
6162 static void vmx_apic_passthrough(bool set_irq_line_from_thread)
6163 {
6164 	if (set_irq_line_from_thread && (cpu_count() < 2)) {
6165 		report_skip(__func__);
6166 		return;
6167 	}
6168 
6169 	u64 cpu_ctrl_0 = CPU_SECONDARY;
6170 	u64 cpu_ctrl_1 = 0;
6171 
6172 	disable_intercept_for_x2apic_msrs();
6173 
6174 	vmcs_write(PIN_CONTROLS, vmcs_read(PIN_CONTROLS) & ~PIN_EXTINT);
6175 
6176 	vmcs_write(CPU_EXEC_CTRL0, vmcs_read(CPU_EXEC_CTRL0) | cpu_ctrl_0);
6177 	vmcs_write(CPU_EXEC_CTRL1, vmcs_read(CPU_EXEC_CTRL1) | cpu_ctrl_1);
6178 
6179 	if (set_irq_line_from_thread) {
6180 		irq_78_handler_vmcall_before_eoi = false;
6181 		on_cpu_async(1, set_irq_line_thread, NULL);
6182 	} else {
6183 		irq_78_handler_vmcall_before_eoi = true;
6184 		ioapic_set_redir(0xf, 0x78, TRIGGER_LEVEL);
6185 		vmx_set_test_stage(2);
6186 	}
6187 	test_set_guest(vmx_apic_passthrough_guest);
6188 
6189 	if (irq_78_handler_vmcall_before_eoi) {
6190 		/* Before EOI remote_irr should still be set */
6191 		enter_guest();
6192 		skip_exit_vmcall();
6193 		TEST_ASSERT_EQ_MSG(1, (int)ioapic_read_redir(0xf).remote_irr,
6194 			"IOAPIC pass-through: remote_irr=1 before EOI");
6195 	}
6196 
6197 	/* After EOI remote_irr should be cleared */
6198 	enter_guest();
6199 	skip_exit_vmcall();
6200 	TEST_ASSERT_EQ_MSG(0, (int)ioapic_read_redir(0xf).remote_irr,
6201 		"IOAPIC pass-through: remote_irr=0 after EOI");
6202 
6203 	/* Let L2 finish */
6204 	enter_guest();
6205 	report(__func__, 1);
6206 }
6207 
6208 static void vmx_apic_passthrough_test(void)
6209 {
6210 	vmx_apic_passthrough(false);
6211 }
6212 
6213 static void vmx_apic_passthrough_thread_test(void)
6214 {
6215 	vmx_apic_passthrough(true);
6216 }
6217 
6218 enum vmcs_access {
6219 	ACCESS_VMREAD,
6220 	ACCESS_VMWRITE,
6221 	ACCESS_NONE,
6222 };
6223 
6224 struct vmcs_shadow_test_common {
6225 	enum vmcs_access op;
6226 	enum Reason reason;
6227 	u64 field;
6228 	u64 value;
6229 	u64 flags;
6230 	u64 time;
6231 } l1_l2_common;
6232 
6233 static inline u64 vmread_flags(u64 field, u64 *val)
6234 {
6235 	u64 flags;
6236 
6237 	asm volatile ("vmread %2, %1; pushf; pop %0"
6238 		      : "=r" (flags), "=rm" (*val) : "r" (field) : "cc");
6239 	return flags & X86_EFLAGS_ALU;
6240 }
6241 
6242 static inline u64 vmwrite_flags(u64 field, u64 val)
6243 {
6244 	u64 flags;
6245 
6246 	asm volatile ("vmwrite %1, %2; pushf; pop %0"
6247 		      : "=r"(flags) : "rm" (val), "r" (field) : "cc");
6248 	return flags & X86_EFLAGS_ALU;
6249 }
6250 
6251 static void vmx_vmcs_shadow_test_guest(void)
6252 {
6253 	struct vmcs_shadow_test_common *c = &l1_l2_common;
6254 	u64 start;
6255 
6256 	while (c->op != ACCESS_NONE) {
6257 		start = rdtsc();
6258 		switch (c->op) {
6259 		default:
6260 			c->flags = -1ull;
6261 			break;
6262 		case ACCESS_VMREAD:
6263 			c->flags = vmread_flags(c->field, &c->value);
6264 			break;
6265 		case ACCESS_VMWRITE:
6266 			c->flags = vmwrite_flags(c->field, 0);
6267 			break;
6268 		}
6269 		c->time = rdtsc() - start;
6270 		vmcall();
6271 	}
6272 }
6273 
6274 static u64 vmread_from_shadow(u64 field)
6275 {
6276 	struct vmcs *primary;
6277 	struct vmcs *shadow;
6278 	u64 value;
6279 
6280 	TEST_ASSERT(!vmcs_save(&primary));
6281 	shadow = (struct vmcs *)vmcs_read(VMCS_LINK_PTR);
6282 	TEST_ASSERT(!make_vmcs_current(shadow));
6283 	value = vmcs_read(field);
6284 	TEST_ASSERT(!make_vmcs_current(primary));
6285 	return value;
6286 }
6287 
6288 static u64 vmwrite_to_shadow(u64 field, u64 value)
6289 {
6290 	struct vmcs *primary;
6291 	struct vmcs *shadow;
6292 
6293 	TEST_ASSERT(!vmcs_save(&primary));
6294 	shadow = (struct vmcs *)vmcs_read(VMCS_LINK_PTR);
6295 	TEST_ASSERT(!make_vmcs_current(shadow));
6296 	vmcs_write(field, value);
6297 	value = vmcs_read(field);
6298 	TEST_ASSERT(!make_vmcs_current(primary));
6299 	return value;
6300 }
6301 
6302 static void vmcs_shadow_test_access(u8 *bitmap[2], enum vmcs_access access)
6303 {
6304 	struct vmcs_shadow_test_common *c = &l1_l2_common;
6305 
6306 	c->op = access;
6307 	vmcs_write(VMX_INST_ERROR, 0);
6308 	enter_guest();
6309 	c->reason = vmcs_read(EXI_REASON) & 0xffff;
6310 	if (c->reason != VMX_VMCALL) {
6311 		skip_exit_insn();
6312 		enter_guest();
6313 	}
6314 	skip_exit_vmcall();
6315 }
6316 
6317 static void vmcs_shadow_test_field(u8 *bitmap[2], u64 field)
6318 {
6319 	struct vmcs_shadow_test_common *c = &l1_l2_common;
6320 	struct vmcs *shadow;
6321 	u64 value;
6322 	uintptr_t flags[2];
6323 	bool good_shadow;
6324 	u32 vmx_inst_error;
6325 
6326 	report_prefix_pushf("field %lx", field);
6327 	c->field = field;
6328 
6329 	shadow = (struct vmcs *)vmcs_read(VMCS_LINK_PTR);
6330 	if (shadow != (struct vmcs *)-1ull) {
6331 		flags[ACCESS_VMREAD] = vmread_flags(field, &value);
6332 		flags[ACCESS_VMWRITE] = vmwrite_flags(field, value);
6333 		good_shadow = !flags[ACCESS_VMREAD] && !flags[ACCESS_VMWRITE];
6334 	} else {
6335 		/*
6336 		 * When VMCS link pointer is -1ull, VMWRITE/VMREAD on
6337 		 * shadowed-fields should fail with setting RFLAGS.CF.
6338 		 */
6339 		flags[ACCESS_VMREAD] = X86_EFLAGS_CF;
6340 		flags[ACCESS_VMWRITE] = X86_EFLAGS_CF;
6341 		good_shadow = false;
6342 	}
6343 
6344 	/* Intercept both VMREAD and VMWRITE. */
6345 	report_prefix_push("no VMREAD/VMWRITE permission");
6346 	/* VMWRITE/VMREAD done on reserved-bit should always intercept */
6347 	if (!(field >> VMCS_FIELD_RESERVED_SHIFT)) {
6348 		set_bit(field, bitmap[ACCESS_VMREAD]);
6349 		set_bit(field, bitmap[ACCESS_VMWRITE]);
6350 	}
6351 	vmcs_shadow_test_access(bitmap, ACCESS_VMWRITE);
6352 	report("not shadowed for VMWRITE", c->reason == VMX_VMWRITE);
6353 	vmcs_shadow_test_access(bitmap, ACCESS_VMREAD);
6354 	report("not shadowed for VMREAD", c->reason == VMX_VMREAD);
6355 	report_prefix_pop();
6356 
6357 	if (field >> VMCS_FIELD_RESERVED_SHIFT)
6358 		goto out;
6359 
6360 	/* Permit shadowed VMREAD. */
6361 	report_prefix_push("VMREAD permission only");
6362 	clear_bit(field, bitmap[ACCESS_VMREAD]);
6363 	set_bit(field, bitmap[ACCESS_VMWRITE]);
6364 	if (good_shadow)
6365 		value = vmwrite_to_shadow(field, MAGIC_VAL_1 + field);
6366 	vmcs_shadow_test_access(bitmap, ACCESS_VMWRITE);
6367 	report("not shadowed for VMWRITE", c->reason == VMX_VMWRITE);
6368 	vmcs_shadow_test_access(bitmap, ACCESS_VMREAD);
6369 	vmx_inst_error = vmcs_read(VMX_INST_ERROR);
6370 	report("shadowed for VMREAD (in %ld cycles)", c->reason == VMX_VMCALL,
6371 	       c->time);
6372 	report("ALU flags after VMREAD (%lx) are as expected (%lx)",
6373 	       c->flags == flags[ACCESS_VMREAD],
6374 	       c->flags, flags[ACCESS_VMREAD]);
6375 	if (good_shadow)
6376 		report("value read from shadow (%lx) is as expected (%lx)",
6377 		       c->value == value, c->value, value);
6378 	else if (shadow != (struct vmcs *)-1ull && flags[ACCESS_VMREAD])
6379 		report("VMX_INST_ERROR (%d) is as expected (%d)",
6380 		       vmx_inst_error == VMXERR_UNSUPPORTED_VMCS_COMPONENT,
6381 		       vmx_inst_error, VMXERR_UNSUPPORTED_VMCS_COMPONENT);
6382 	report_prefix_pop();
6383 
6384 	/* Permit shadowed VMWRITE. */
6385 	report_prefix_push("VMWRITE permission only");
6386 	set_bit(field, bitmap[ACCESS_VMREAD]);
6387 	clear_bit(field, bitmap[ACCESS_VMWRITE]);
6388 	if (good_shadow)
6389 		vmwrite_to_shadow(field, MAGIC_VAL_1 + field);
6390 	vmcs_shadow_test_access(bitmap, ACCESS_VMWRITE);
6391 	vmx_inst_error = vmcs_read(VMX_INST_ERROR);
6392 	report("shadowed for VMWRITE (in %ld cycles)", c->reason == VMX_VMCALL,
6393 		c->time);
6394 	report("ALU flags after VMWRITE (%lx) are as expected (%lx)",
6395 	       c->flags == flags[ACCESS_VMREAD],
6396 	       c->flags, flags[ACCESS_VMREAD]);
6397 	if (good_shadow) {
6398 		value = vmread_from_shadow(field);
6399 		report("shadow VMCS value (%lx) is as expected (%lx)",
6400 		       value == 0, value, 0ul);
6401 	} else if (shadow != (struct vmcs *)-1ull && flags[ACCESS_VMWRITE]) {
6402 		report("VMX_INST_ERROR (%d) is as expected (%d)",
6403 		       vmx_inst_error == VMXERR_UNSUPPORTED_VMCS_COMPONENT,
6404 		       vmx_inst_error, VMXERR_UNSUPPORTED_VMCS_COMPONENT);
6405 	}
6406 	vmcs_shadow_test_access(bitmap, ACCESS_VMREAD);
6407 	report("not shadowed for VMREAD", c->reason == VMX_VMREAD);
6408 	report_prefix_pop();
6409 
6410 	/* Permit shadowed VMREAD and VMWRITE. */
6411 	report_prefix_push("VMREAD and VMWRITE permission");
6412 	clear_bit(field, bitmap[ACCESS_VMREAD]);
6413 	clear_bit(field, bitmap[ACCESS_VMWRITE]);
6414 	if (good_shadow)
6415 		vmwrite_to_shadow(field, MAGIC_VAL_1 + field);
6416 	vmcs_shadow_test_access(bitmap, ACCESS_VMWRITE);
6417 	vmx_inst_error = vmcs_read(VMX_INST_ERROR);
6418 	report("shadowed for VMWRITE (in %ld cycles)", c->reason == VMX_VMCALL,
6419 		c->time);
6420 	report("ALU flags after VMWRITE (%lx) are as expected (%lx)",
6421 	       c->flags == flags[ACCESS_VMREAD],
6422 	       c->flags, flags[ACCESS_VMREAD]);
6423 	if (good_shadow) {
6424 		value = vmread_from_shadow(field);
6425 		report("shadow VMCS value (%lx) is as expected (%lx)",
6426 		       value == 0, value, 0ul);
6427 	} else if (shadow != (struct vmcs *)-1ull && flags[ACCESS_VMWRITE]) {
6428 		report("VMX_INST_ERROR (%d) is as expected (%d)",
6429 		       vmx_inst_error == VMXERR_UNSUPPORTED_VMCS_COMPONENT,
6430 		       vmx_inst_error, VMXERR_UNSUPPORTED_VMCS_COMPONENT);
6431 	}
6432 	vmcs_shadow_test_access(bitmap, ACCESS_VMREAD);
6433 	vmx_inst_error = vmcs_read(VMX_INST_ERROR);
6434 	report("shadowed for VMREAD (in %ld cycles)", c->reason == VMX_VMCALL,
6435 	       c->time);
6436 	report("ALU flags after VMREAD (%lx) are as expected (%lx)",
6437 	       c->flags == flags[ACCESS_VMREAD],
6438 	       c->flags, flags[ACCESS_VMREAD]);
6439 	if (good_shadow)
6440 		report("value read from shadow (%lx) is as expected (%lx)",
6441 		       c->value == 0, c->value, 0ul);
6442 	else if (shadow != (struct vmcs *)-1ull && flags[ACCESS_VMREAD])
6443 		report("VMX_INST_ERROR (%d) is as expected (%d)",
6444 		       vmx_inst_error == VMXERR_UNSUPPORTED_VMCS_COMPONENT,
6445 		       vmx_inst_error, VMXERR_UNSUPPORTED_VMCS_COMPONENT);
6446 	report_prefix_pop();
6447 
6448 out:
6449 	report_prefix_pop();
6450 }
6451 
6452 static void vmx_vmcs_shadow_test_body(u8 *bitmap[2])
6453 {
6454 	unsigned base;
6455 	unsigned index;
6456 	unsigned bit;
6457 	unsigned highest_index = rdmsr(MSR_IA32_VMX_VMCS_ENUM);
6458 
6459 	/* Run test on all possible valid VMCS fields */
6460 	for (base = 0;
6461 	     base < (1 << VMCS_FIELD_RESERVED_SHIFT);
6462 	     base += (1 << VMCS_FIELD_TYPE_SHIFT))
6463 		for (index = 0; index <= highest_index; index++)
6464 			vmcs_shadow_test_field(bitmap, base + index);
6465 
6466 	/*
6467 	 * Run tests on some invalid VMCS fields
6468 	 * (Have reserved bit set).
6469 	 */
6470 	for (bit = VMCS_FIELD_RESERVED_SHIFT; bit < VMCS_FIELD_BIT_SIZE; bit++)
6471 		vmcs_shadow_test_field(bitmap, (1ull << bit));
6472 }
6473 
6474 static void vmx_vmcs_shadow_test(void)
6475 {
6476 	u8 *bitmap[2];
6477 	struct vmcs *shadow;
6478 
6479 	if (!(ctrl_cpu_rev[0].clr & CPU_SECONDARY)) {
6480 		printf("\t'Activate secondary controls' not supported.\n");
6481 		return;
6482 	}
6483 
6484 	if (!(ctrl_cpu_rev[1].clr & CPU_SHADOW_VMCS)) {
6485 		printf("\t'VMCS shadowing' not supported.\n");
6486 		return;
6487 	}
6488 
6489 	if (!(rdmsr(MSR_IA32_VMX_MISC) &
6490 	      MSR_IA32_VMX_MISC_VMWRITE_SHADOW_RO_FIELDS)) {
6491 		printf("\tVMWRITE can't modify VM-exit information fields.\n");
6492 		return;
6493 	}
6494 
6495 	test_set_guest(vmx_vmcs_shadow_test_guest);
6496 
6497 	bitmap[ACCESS_VMREAD] = alloc_page();
6498 	bitmap[ACCESS_VMWRITE] = alloc_page();
6499 
6500 	vmcs_write(VMREAD_BITMAP, virt_to_phys(bitmap[ACCESS_VMREAD]));
6501 	vmcs_write(VMWRITE_BITMAP, virt_to_phys(bitmap[ACCESS_VMWRITE]));
6502 
6503 	shadow = alloc_page();
6504 	shadow->hdr.revision_id = basic.revision;
6505 	shadow->hdr.shadow_vmcs = 1;
6506 	TEST_ASSERT(!vmcs_clear(shadow));
6507 
6508 	vmcs_clear_bits(CPU_EXEC_CTRL0, CPU_RDTSC);
6509 	vmcs_set_bits(CPU_EXEC_CTRL0, CPU_SECONDARY);
6510 	vmcs_set_bits(CPU_EXEC_CTRL1, CPU_SHADOW_VMCS);
6511 
6512 	vmcs_write(VMCS_LINK_PTR, virt_to_phys(shadow));
6513 	report_prefix_push("valid link pointer");
6514 	vmx_vmcs_shadow_test_body(bitmap);
6515 	report_prefix_pop();
6516 
6517 	vmcs_write(VMCS_LINK_PTR, -1ull);
6518 	report_prefix_push("invalid link pointer");
6519 	vmx_vmcs_shadow_test_body(bitmap);
6520 	report_prefix_pop();
6521 
6522 	l1_l2_common.op = ACCESS_NONE;
6523 	enter_guest();
6524 }
6525 
6526 
6527 
6528 static int invalid_msr_init(struct vmcs *vmcs)
6529 {
6530 	if (!(ctrl_pin_rev.clr & PIN_PREEMPT)) {
6531 		printf("\tPreemption timer is not supported\n");
6532 		return VMX_TEST_EXIT;
6533 	}
6534 	vmcs_write(PIN_CONTROLS, vmcs_read(PIN_CONTROLS) | PIN_PREEMPT);
6535 	preempt_val = 10000000;
6536 	vmcs_write(PREEMPT_TIMER_VALUE, preempt_val);
6537 	preempt_scale = rdmsr(MSR_IA32_VMX_MISC) & 0x1F;
6538 
6539 	if (!(ctrl_exit_rev.clr & EXI_SAVE_PREEMPT))
6540 		printf("\tSave preemption value is not supported\n");
6541 
6542 	vmcs_write(ENT_MSR_LD_CNT, 1);
6543 	vmcs_write(ENTER_MSR_LD_ADDR, (u64)0x13370000);
6544 
6545 	return VMX_TEST_START;
6546 }
6547 
6548 
6549 static void invalid_msr_main(void)
6550 {
6551 	report("Invalid MSR load", 0);
6552 }
6553 
6554 static int invalid_msr_exit_handler(void)
6555 {
6556 	report("Invalid MSR load", 0);
6557 	print_vmexit_info();
6558 	return VMX_TEST_EXIT;
6559 }
6560 
6561 static int invalid_msr_entry_failure(struct vmentry_failure *failure)
6562 {
6563 	ulong reason;
6564 
6565 	reason = vmcs_read(EXI_REASON);
6566 	report("Invalid MSR load", reason == (0x80000000u | VMX_FAIL_MSR));
6567 	return VMX_TEST_VMEXIT;
6568 }
6569 
6570 
6571 #define TEST(name) { #name, .v2 = name }
6572 
6573 /* name/init/guest_main/exit_handler/syscall_handler/guest_regs */
6574 struct vmx_test vmx_tests[] = {
6575 	{ "null", NULL, basic_guest_main, basic_exit_handler, NULL, {0} },
6576 	{ "vmenter", NULL, vmenter_main, vmenter_exit_handler, NULL, {0} },
6577 	{ "preemption timer", preemption_timer_init, preemption_timer_main,
6578 		preemption_timer_exit_handler, NULL, {0} },
6579 	{ "control field PAT", test_ctrl_pat_init, test_ctrl_pat_main,
6580 		test_ctrl_pat_exit_handler, NULL, {0} },
6581 	{ "control field EFER", test_ctrl_efer_init, test_ctrl_efer_main,
6582 		test_ctrl_efer_exit_handler, NULL, {0} },
6583 	{ "CR shadowing", NULL, cr_shadowing_main,
6584 		cr_shadowing_exit_handler, NULL, {0} },
6585 	{ "I/O bitmap", iobmp_init, iobmp_main, iobmp_exit_handler,
6586 		NULL, {0} },
6587 	{ "instruction intercept", insn_intercept_init, insn_intercept_main,
6588 		insn_intercept_exit_handler, NULL, {0} },
6589 	{ "EPT A/D disabled", ept_init, ept_main, ept_exit_handler, NULL, {0} },
6590 	{ "EPT A/D enabled", eptad_init, eptad_main, eptad_exit_handler, NULL, {0} },
6591 	{ "PML", pml_init, pml_main, pml_exit_handler, NULL, {0} },
6592 	{ "VPID", vpid_init, vpid_main, vpid_exit_handler, NULL, {0} },
6593 	{ "interrupt", interrupt_init, interrupt_main,
6594 		interrupt_exit_handler, NULL, {0} },
6595 	{ "debug controls", dbgctls_init, dbgctls_main, dbgctls_exit_handler,
6596 		NULL, {0} },
6597 	{ "MSR switch", msr_switch_init, msr_switch_main,
6598 		msr_switch_exit_handler, NULL, {0}, msr_switch_entry_failure },
6599 	{ "vmmcall", vmmcall_init, vmmcall_main, vmmcall_exit_handler, NULL, {0} },
6600 	{ "disable RDTSCP", disable_rdtscp_init, disable_rdtscp_main,
6601 		disable_rdtscp_exit_handler, NULL, {0} },
6602 	{ "int3", int3_init, int3_guest_main, int3_exit_handler, NULL, {0} },
6603 	{ "into", into_init, into_guest_main, into_exit_handler, NULL, {0} },
6604 	{ "exit_monitor_from_l2_test", NULL, exit_monitor_from_l2_main,
6605 		exit_monitor_from_l2_handler, NULL, {0} },
6606 	{ "invalid_msr", invalid_msr_init, invalid_msr_main,
6607 		invalid_msr_exit_handler, NULL, {0}, invalid_msr_entry_failure},
6608 	/* Basic V2 tests. */
6609 	TEST(v2_null_test),
6610 	TEST(v2_multiple_entries_test),
6611 	TEST(fixture_test_case1),
6612 	TEST(fixture_test_case2),
6613 	/* Opcode tests. */
6614 	TEST(invvpid_test_v2),
6615 	/* VM-entry tests */
6616 	TEST(vmx_controls_test),
6617 	TEST(vmx_host_state_area_test),
6618 	TEST(vmentry_movss_shadow_test),
6619 	/* APICv tests */
6620 	TEST(vmx_eoi_bitmap_ioapic_scan_test),
6621 	TEST(vmx_hlt_with_rvi_test),
6622 	/* APIC pass-through tests */
6623 	TEST(vmx_apic_passthrough_test),
6624 	TEST(vmx_apic_passthrough_thread_test),
6625 	/* VMCS Shadowing tests */
6626 	TEST(vmx_vmcs_shadow_test),
6627 	/* Regression tests */
6628 	TEST(vmx_cr_load_test),
6629 	TEST(vmx_nm_test),
6630 	TEST(vmx_db_test),
6631 	TEST(vmx_nmi_window_test),
6632 	TEST(vmx_intr_window_test),
6633 	TEST(vmx_pending_event_test),
6634 	TEST(vmx_pending_event_hlt_test),
6635 	TEST(vmx_store_tsc_test),
6636 	/* EPT access tests. */
6637 	TEST(ept_access_test_not_present),
6638 	TEST(ept_access_test_read_only),
6639 	TEST(ept_access_test_write_only),
6640 	TEST(ept_access_test_read_write),
6641 	TEST(ept_access_test_execute_only),
6642 	TEST(ept_access_test_read_execute),
6643 	TEST(ept_access_test_write_execute),
6644 	TEST(ept_access_test_read_write_execute),
6645 	TEST(ept_access_test_reserved_bits),
6646 	TEST(ept_access_test_ignored_bits),
6647 	TEST(ept_access_test_paddr_not_present_ad_disabled),
6648 	TEST(ept_access_test_paddr_not_present_ad_enabled),
6649 	TEST(ept_access_test_paddr_read_only_ad_disabled),
6650 	TEST(ept_access_test_paddr_read_only_ad_enabled),
6651 	TEST(ept_access_test_paddr_read_write),
6652 	TEST(ept_access_test_paddr_read_write_execute),
6653 	TEST(ept_access_test_paddr_read_execute_ad_disabled),
6654 	TEST(ept_access_test_paddr_read_execute_ad_enabled),
6655 	TEST(ept_access_test_paddr_not_present_page_fault),
6656 	TEST(ept_access_test_force_2m_page),
6657 	{ NULL, NULL, NULL, NULL, NULL, {0} },
6658 };
6659