add new system file

hc

2024-02-20 102a0743326a03cd1a1202ceda21e175b7d3575c

 kernel/tools/testing/selftests/x86/test_vsyscall.c
..
..
@@ -18,6 +18,9 @@
18
18
 #include <sched.h>
19
19
 #include <stdbool.h>
20
20
 #include <setjmp.h>
21
+#include <sys/uio.h>
22
+
23
+#include "helpers.h"
21
24
 
22
25
 #ifdef __x86_64__
23
26
 # define VSYS(x) (x)
..
..
@@ -49,21 +52,21 @@
49
52
 }
50
53
 
51
54
 /* vsyscalls and vDSO */
52
-bool should_read_vsyscall = false;
55
+bool vsyscall_map_r = false, vsyscall_map_x = false;
53
56
 
54
57
 typedef long (*gtod_t)(struct timeval *tv, struct timezone *tz);
55
-gtod_t vgtod = (gtod_t)VSYS(0xffffffffff600000);
58
+const gtod_t vgtod = (gtod_t)VSYS(0xffffffffff600000);
56
59
 gtod_t vdso_gtod;
57
60
 
58
61
 typedef int (*vgettime_t)(clockid_t, struct timespec *);
59
62
 vgettime_t vdso_gettime;
60
63
 
61
64
 typedef long (*time_func_t)(time_t *t);
62
-time_func_t vtime = (time_func_t)VSYS(0xffffffffff600400);
65
+const time_func_t vtime = (time_func_t)VSYS(0xffffffffff600400);
63
66
 time_func_t vdso_time;
64
67
 
65
68
 typedef long (*getcpu_t)(unsigned *, unsigned *, void *);
66
-getcpu_t vgetcpu = (getcpu_t)VSYS(0xffffffffff600800);
69
+const getcpu_t vgetcpu = (getcpu_t)VSYS(0xffffffffff600800);
67
70
 getcpu_t vdso_getcpu;
68
71
 
69
72
 static void init_vdso(void)
..
..
@@ -107,7 +110,7 @@
107
110
 	maps = fopen("/proc/self/maps", "r");
108
111
 	if (!maps) {
109
112
 		printf("[WARN]\tCould not open /proc/self/maps -- assuming vsyscall is r-x\n");
110
-		should_read_vsyscall = true;
113
+		vsyscall_map_r = true;
111
114
 		return 0;
112
115
 	}
113
116
 
..
..
@@ -133,12 +136,8 @@
133
136
 		}
134
137
 
135
138
 		printf("\tvsyscall permissions are %c-%c\n", r, x);
136
-		should_read_vsyscall = (r == 'r');
137
-		if (x != 'x') {
138
-			vgtod = NULL;
139
-			vtime = NULL;
140
-			vgetcpu = NULL;
141
-		}
139
+		vsyscall_map_r = (r == 'r');
140
+		vsyscall_map_x = (x == 'x');
142
141
 
143
142
 		found = true;
144
143
 		break;
..
..
@@ -148,10 +147,8 @@
148
147
 
149
148
 	if (!found) {
150
149
 		printf("\tno vsyscall map in /proc/self/maps\n");
151
-		should_read_vsyscall = false;
152
-		vgtod = NULL;
153
-		vtime = NULL;
154
-		vgetcpu = NULL;
150
+		vsyscall_map_r = false;
151
+		vsyscall_map_x = false;
155
152
 	}
156
153
 
157
154
 	return nerrs;
..
..
@@ -183,9 +180,13 @@
183
180
 }
184
181
 
185
182
 static jmp_buf jmpbuf;
183
+static volatile unsigned long segv_err;
186
184
 
187
185
 static void sigsegv(int sig, siginfo_t *info, void *ctx_void)
188
186
 {
187
+	ucontext_t *ctx = (ucontext_t *)ctx_void;
188
+
189
+	segv_err =  ctx->uc_mcontext.gregs[REG_ERR];
189
190
 	siglongjmp(jmpbuf, 1);
190
191
 }
191
192
 
..
..
@@ -238,7 +239,7 @@
238
239
 		err(1, "syscall gettimeofday");
239
240
 	if (vdso_gtod)
240
241
 		ret_vdso = vdso_gtod(&tv_vdso, &tz_vdso);
241
-	if (vgtod)
242
+	if (vsyscall_map_x)
242
243
 		ret_vsys = vgtod(&tv_vsys, &tz_vsys);
243
244
 	if (sys_gtod(&tv_sys2, &tz_sys) != 0)
244
245
 		err(1, "syscall gettimeofday");
..
..
@@ -252,7 +253,7 @@
252
253
 		}
253
254
 	}
254
255
 
255
-	if (vgtod) {
256
+	if (vsyscall_map_x) {
256
257
 		if (ret_vsys == 0) {
257
258
 			nerrs += check_gtod(&tv_sys1, &tv_sys2, &tz_sys, "vsyscall", &tv_vsys, &tz_vsys);
258
259
 		} else {
..
..
@@ -273,7 +274,7 @@
273
274
 	t_sys1 = sys_time(&t2_sys1);
274
275
 	if (vdso_time)
275
276
 		t_vdso = vdso_time(&t2_vdso);
276
-	if (vtime)
277
+	if (vsyscall_map_x)
277
278
 		t_vsys = vtime(&t2_vsys);
278
279
 	t_sys2 = sys_time(&t2_sys2);
279
280
 	if (t_sys1 < 0 || t_sys1 != t2_sys1 || t_sys2 < 0 || t_sys2 != t2_sys2) {
..
..
@@ -294,7 +295,7 @@
294
295
 		}
295
296
 	}
296
297
 
297
-	if (vtime) {
298
+	if (vsyscall_map_x) {
298
299
 		if (t_vsys < 0 || t_vsys != t2_vsys) {
299
300
 			printf("[FAIL]\tvsyscall failed (ret:%ld output:%ld)\n", t_vsys, t2_vsys);
300
301
 			nerrs++;
..
..
@@ -330,7 +331,7 @@
330
331
 	ret_sys = sys_getcpu(&cpu_sys, &node_sys, 0);
331
332
 	if (vdso_getcpu)
332
333
 		ret_vdso = vdso_getcpu(&cpu_vdso, &node_vdso, 0);
333
-	if (vgetcpu)
334
+	if (vsyscall_map_x)
334
335
 		ret_vsys = vgetcpu(&cpu_vsys, &node_vsys, 0);
335
336
 
336
337
 	if (ret_sys == 0) {
..
..
@@ -369,7 +370,7 @@
369
370
 		}
370
371
 	}
371
372
 
372
-	if (vgetcpu) {
373
+	if (vsyscall_map_x) {
373
374
 		if (ret_vsys) {
374
375
 			printf("[FAIL]\tvsyscall getcpu() failed\n");
375
376
 			nerrs++;
..
..
@@ -410,36 +411,109 @@
410
411
 		can_read = false;
411
412
 	}
412
413
 
413
-	if (can_read && !should_read_vsyscall) {
414
+	if (can_read && !vsyscall_map_r) {
414
415
 		printf("[FAIL]\tWe have read access, but we shouldn't\n");
415
416
 		return 1;
416
-	} else if (!can_read && should_read_vsyscall) {
417
+	} else if (!can_read && vsyscall_map_r) {
417
418
 		printf("[FAIL]\tWe don't have read access, but we should\n");
418
419
 		return 1;
420
+	} else if (can_read) {
421
+		printf("[OK]\tWe have read access\n");
419
422
 	} else {
420
-		printf("[OK]\tgot expected result\n");
423
+		printf("[OK]\tWe do not have read access: #PF(0x%lx)\n",
424
+		       segv_err);
421
425
 	}
422
426
 #endif
423
427
 
424
428
 	return 0;
425
429
 }
426
430
 
431
+static int test_vsys_x(void)
432
+{
433
+#ifdef __x86_64__
434
+	if (vsyscall_map_x) {
435
+		/* We already tested this adequately. */
436
+		return 0;
437
+	}
438
+
439
+	printf("[RUN]\tMake sure that vsyscalls really page fault\n");
440
+
441
+	bool can_exec;
442
+	if (sigsetjmp(jmpbuf, 1) == 0) {
443
+		vgtod(NULL, NULL);
444
+		can_exec = true;
445
+	} else {
446
+		can_exec = false;
447
+	}
448
+
449
+	if (can_exec) {
450
+		printf("[FAIL]\tExecuting the vsyscall did not page fault\n");
451
+		return 1;
452
+	} else if (segv_err & (1 << 4)) { /* INSTR */
453
+		printf("[OK]\tExecuting the vsyscall page failed: #PF(0x%lx)\n",
454
+		       segv_err);
455
+	} else {
456
+		printf("[FAIL]\tExecution failed with the wrong error: #PF(0x%lx)\n",
457
+		       segv_err);
458
+		return 1;
459
+	}
460
+#endif
461
+
462
+	return 0;
463
+}
464
+
465
+/*
466
+ * Debuggers expect ptrace() to be able to peek at the vsyscall page.
467
+ * Use process_vm_readv() as a proxy for ptrace() to test this.  We
468
+ * want it to work in the vsyscall=emulate case and to fail in the
469
+ * vsyscall=xonly case.
470
+ *
471
+ * It's worth noting that this ABI is a bit nutty.  write(2) can't
472
+ * read from the vsyscall page on any kernel version or mode.  The
473
+ * fact that ptrace() ever worked was a nice courtesy of old kernels,
474
+ * but the code to support it is fairly gross.
475
+ */
476
+static int test_process_vm_readv(void)
477
+{
478
+#ifdef __x86_64__
479
+	char buf[4096];
480
+	struct iovec local, remote;
481
+	int ret;
482
+
483
+	printf("[RUN]\tprocess_vm_readv() from vsyscall page\n");
484
+
485
+	local.iov_base = buf;
486
+	local.iov_len = 4096;
487
+	remote.iov_base = (void *)0xffffffffff600000;
488
+	remote.iov_len = 4096;
489
+	ret = process_vm_readv(getpid(), &local, 1, &remote, 1, 0);
490
+	if (ret != 4096) {
491
+		/*
492
+		 * We expect process_vm_readv() to work if and only if the
493
+		 * vsyscall page is readable.
494
+		 */
495
+		printf("[%s]\tprocess_vm_readv() failed (ret = %d, errno = %d)\n", vsyscall_map_r ? "FAIL" : "OK", ret, errno);
496
+		return vsyscall_map_r ? 1 : 0;
497
+	}
498
+
499
+	if (vsyscall_map_r) {
500
+		if (!memcmp(buf, remote.iov_base, sizeof(buf))) {
501
+			printf("[OK]\tIt worked and read correct data\n");
502
+		} else {
503
+			printf("[FAIL]\tIt worked but returned incorrect data\n");
504
+			return 1;
505
+		}
506
+	} else {
507
+		printf("[FAIL]\tprocess_rm_readv() succeeded, but it should have failed in this configuration\n");
508
+		return 1;
509
+	}
510
+#endif
511
+
512
+	return 0;
513
+}
427
514
 
428
515
 #ifdef __x86_64__
429
-#define X86_EFLAGS_TF (1UL << 8)
430
516
 static volatile sig_atomic_t num_vsyscall_traps;
431
-
432
-static unsigned long get_eflags(void)
433
-{
434
-	unsigned long eflags;
435
-	asm volatile ("pushfq\n\tpopq %0" : "=rm" (eflags));
436
-	return eflags;
437
-}
438
-
439
-static void set_eflags(unsigned long eflags)
440
-{
441
-	asm volatile ("pushq %0\n\tpopfq" : : "rm" (eflags) : "flags");
442
-}
443
517
 
444
518
 static void sigtrap(int sig, siginfo_t *info, void *ctx_void)
445
519
 {
..
..
@@ -455,7 +529,7 @@
455
529
 	time_t tmp;
456
530
 	bool is_native;
457
531
 
458
-	if (!vtime)
532
+	if (!vsyscall_map_x)
459
533
 		return 0;
460
534
 
461
535
 	printf("[RUN]\tchecking that vsyscalls are emulated\n");
..
..
@@ -497,6 +571,9 @@
497
571
 
498
572
 	sethandler(SIGSEGV, sigsegv, 0);
499
573
 	nerrs += test_vsys_r();
574
+	nerrs += test_vsys_x();
575
+
576
+	nerrs += test_process_vm_readv();
500
577
 
501
578
 #ifdef __x86_64__
502
579
 	nerrs += test_emulation();