1 /* $NetBSD: systrace.c,v 1.12 2018/06/29 11:33:46 kamil Exp $ */
2
3 /*
4 * CDDL HEADER START
5 *
6 * The contents of this file are subject to the terms of the
7 * Common Development and Distribution License (the "License").
8 * You may not use this file except in compliance with the License.
9 *
10 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
11 * or http://www.opensolaris.org/os/licensing.
12 * See the License for the specific language governing permissions
13 * and limitations under the License.
14 *
15 * When distributing Covered Code, include this CDDL HEADER in each
16 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
17 * If applicable, add the following below this CDDL HEADER, with the
18 * fields enclosed by brackets "[]" replaced with your own identifying
19 * information: Portions Copyright [yyyy] [name of copyright owner]
20 *
21 * CDDL HEADER END
22 *
23 * Portions Copyright 2006-2008 John Birrell jb@freebsd.org
24 *
25 */
26
27 /*
28 * Copyright 2006 Sun Microsystems, Inc. All rights reserved.
29 * Use is subject to license terms.
30 */
31
32 #include <sys/cdefs.h>
33 /* __FBSDID("$FreeBSD: head/sys/cddl/dev/systrace/systrace.c 306220 2016-09-22 23:22:53Z markj $"); */
34
35 #include <sys/proc.h>
36 #include <sys/param.h>
37 #include <sys/systm.h>
38 #include <sys/conf.h>
39 #include <sys/cpuvar.h>
40 #include <sys/fcntl.h>
41 #include <sys/filio.h>
42 #include <sys/kernel.h>
43 #include <sys/kmem.h>
44 #include <sys/kthread.h>
45 #include <sys/syslimits.h>
46 #include <sys/linker.h>
47 #include <sys/lock.h>
48 #include <sys/malloc.h>
49 #include <sys/module.h>
50 #include <sys/mutex.h>
51 #include <sys/poll.h>
52 #include <sys/proc.h>
53 #include <sys/selinfo.h>
54 #include <sys/syscallargs.h>
55 #include <sys/uio.h>
56 #include <sys/unistd.h>
57
58 #include <sys/dtrace.h>
59 #include "dtrace_cddl.h"
60
61 #include "emultrace.h"
62
63 #define CONCAT(x,y) __CONCAT(x,y)
64 #define STRING(s) __STRING(s)
65
66 #ifdef __FreeBSD__
67 #ifdef LINUX_SYSTRACE
68 #if defined(__amd64__)
69 #include <amd64/linux/linux.h>
70 #include <amd64/linux/linux_proto.h>
71 #include <amd64/linux/linux_syscalls.c>
72 #include <amd64/linux/linux_systrace_args.c>
73 #elif defined(__i386__)
74 #include <i386/linux/linux.h>
75 #include <i386/linux/linux_proto.h>
76 #include <i386/linux/linux_syscalls.c>
77 #include <i386/linux/linux_systrace_args.c>
78 #else
79 #error Only i386 and amd64 are supported.
80 #endif
81 #define MODNAME "linux"
82 extern struct sysent linux_sysent[];
83 #define MAXSYSCALL LINUX_SYS_MAXSYSCALL
84 #define SYSCALLNAMES linux_syscallnames
85 #define SYSENT linux_sysent
86 #elif defined(LINUX32_SYSTRACE)
87 #if defined(__amd64__)
88 #include <amd64/linux32/linux.h>
89 #include <amd64/linux32/linux32_proto.h>
90 #include <amd64/linux32/linux32_syscalls.c>
91 #include <amd64/linux32/linux32_systrace_args.c>
92 #else
93 #error Only amd64 is supported.
94 #endif
95 #define MODNAME "linux32"
96 extern struct sysent linux32_sysent[];
97 #define MAXSYSCALL LINUX32_SYS_MAXSYSCALL
98 #define SYSCALLNAMES linux32_syscallnames
99 #define SYSENT linux32_sysent
100 #elif defined(FREEBSD32_SYSTRACE)
101 /*
102 * The syscall arguments are processed into a DTrace argument array
103 * using a generated function. See sys/kern/makesyscalls.sh.
104 */
105 #include <compat/freebsd32/freebsd32_proto.h>
106 #include <compat/freebsd32/freebsd32_util.h>
107 #include <compat/freebsd32/freebsd32_syscall.h>
108 #include <compat/freebsd32/freebsd32_systrace_args.c>
109 extern const char *freebsd32_syscallnames[];
110 #define MODNAME "freebsd32"
111 #define MAXSYSCALL FREEBSD32_SYS_MAXSYSCALL
112 #define SYSCALLNAMES freebsd32_syscallnames
113 #define SYSENT freebsd32_sysent
114 #else
115 /*
116 * The syscall arguments are processed into a DTrace argument array
117 * using a generated function. See sys/kern/makesyscalls.sh.
118 */
119 #include <sys/syscall.h>
120 #include <kern/systrace_args.c>
121 #define MODNAME "freebsd"
122 #define MAXSYSCALL SYS_MAXSYSCALL
123 #define SYSCALLNAMES syscallnames
124 #define SYSENT sysent
125 #define NATIVE_ABI
126 #endif
127
128 #define PROVNAME "syscall"
129 #define DEVNAME "dtrace/systrace/" MODNAME
130 #endif /* __FreeBSD__ */
131
132 #ifdef __NetBSD__
133 #include <sys/syscallargs.h>
134
135 #ifndef NATIVE
136 extern const char * const CONCAT(emulname,_syscallnames)[];
137 extern const char * const CONCAT(alt,CONCAT(emulname,_syscallnames))[];
138 extern struct sysent CONCAT(emulname,_sysent)[];
139 #define MODNAME CONCAT(dtrace_syscall_,emulname)
140 #define MODDEP "dtrace_syscall,compat_" STRING(emulname)
141 #define MAXSYSCALL CONCAT(EMULNAME,_SYS_MAXSYSCALL)
142 #define SYSCALLNAMES CONCAT(emulname,_syscallnames)
143 #define ALTSYSCALLNAMES CONCAT(alt,CONCAT(emulname,_syscallnames))
144 #define SYSENT CONCAT(emulname,_sysent)
145 #define PROVNAME STRING(emulname) "_syscall"
146 #else
147 extern const char * const syscallnames[];
148 extern const char * const altsyscallnames[];
149 #define MODNAME dtrace_syscall
150 #define MODDEP "dtrace"
151 #define MAXSYSCALL SYS_MAXSYSCALL
152 #define SYSCALLNAMES syscallnames
153 #define ALTSYSCALLNAMES altsyscallnames
154 #define SYSENT sysent
155 #define PROVNAME "syscall"
156 #endif
157
158 #define MODCMD CONCAT(MODNAME,_modcmd)
159 #define EMUL CONCAT(emul_,emulname)
160 extern struct emul EMUL;
161 #define curthread curlwp
162 #endif /* __NetBSD__ */
163
164 #define SYSTRACE_ARTIFICIAL_FRAMES 1
165
166 #define SYSTRACE_SHIFT 16
167 #define SYSTRACE_ISENTRY(x) ((int)(x) >> SYSTRACE_SHIFT)
168 #define SYSTRACE_SYSNUM(x) ((int)(x) & ((1 << SYSTRACE_SHIFT) - 1))
169 #define SYSTRACE_ENTRY(id) ((1 << SYSTRACE_SHIFT) | (id))
170 #define SYSTRACE_RETURN(id) (id)
171
172 #if ((1 << SYSTRACE_SHIFT) <= MAXSYSCALL)
173 #error 1 << SYSTRACE_SHIFT must exceed number of system calls
174 #endif
175
176 static int systrace_unload(void);
177 static void systrace_getargdesc(void *, dtrace_id_t, void *, dtrace_argdesc_t *);
178 static uint64_t systrace_getargval(void *, dtrace_id_t, void *, int, int);
179 static void systrace_provide(void *, dtrace_probedesc_t *);
180 static void systrace_destroy(void *, dtrace_id_t, void *);
181 static int systrace_enable(void *, dtrace_id_t, void *);
182 static void systrace_disable(void *, dtrace_id_t, void *);
183 static void systrace_load(void *);
184
185 #ifdef __FreeBSD__
186 static union {
187 const char **p_constnames;
188 char **pp_syscallnames;
189 } uglyhack = { SYSCALLNAMES };
190 #endif
191
192 static dtrace_pattr_t systrace_attr = {
193 { DTRACE_STABILITY_EVOLVING, DTRACE_STABILITY_EVOLVING, DTRACE_CLASS_COMMON },
194 { DTRACE_STABILITY_PRIVATE, DTRACE_STABILITY_PRIVATE, DTRACE_CLASS_UNKNOWN },
195 { DTRACE_STABILITY_PRIVATE, DTRACE_STABILITY_PRIVATE, DTRACE_CLASS_ISA },
196 { DTRACE_STABILITY_EVOLVING, DTRACE_STABILITY_EVOLVING, DTRACE_CLASS_COMMON },
197 { DTRACE_STABILITY_PRIVATE, DTRACE_STABILITY_PRIVATE, DTRACE_CLASS_ISA },
198 };
199
200 static dtrace_pops_t systrace_pops = {
201 systrace_provide,
202 NULL,
203 systrace_enable,
204 systrace_disable,
205 NULL,
206 NULL,
207 systrace_getargdesc,
208 systrace_getargval,
209 NULL,
210 systrace_destroy
211 };
212
213 static dtrace_provider_id_t systrace_id;
214
215 /*
216 * Probe callback function.
217 *
218 * Note: This function is called for _all_ syscalls, regardless of which sysent
219 * array the syscall comes from. It could be a standard syscall or a
220 * compat syscall from something like Linux.
221 */
222 #ifdef __FreeBSD__
223 #ifdef NATIVE_ABI
224 static void
systrace_probe(struct syscall_args * sa,enum systrace_probe_t type,int retval)225 systrace_probe(struct syscall_args *sa, enum systrace_probe_t type, int retval)
226 {
227 uint64_t uargs[nitems(sa->args)];
228 dtrace_id_t id;
229 int n_args, sysnum;
230
231 sysnum = sa->code;
232 memset(uargs, 0, sizeof(uargs));
233
234 if (type == SYSTRACE_ENTRY) {
235 if ((id = sa->callp->sy_entry) == DTRACE_IDNONE)
236 return;
237
238 if (sa->callp->sy_systrace_args_func != NULL)
239 /*
240 * Convert the syscall parameters using the registered
241 * function.
242 */
243 (*sa->callp->sy_systrace_args_func)(sysnum, sa->args,
244 uargs, &n_args);
245 else
246 /*
247 * Use the built-in system call argument conversion
248 * function to translate the syscall structure fields
249 * into the array of 64-bit values that DTrace expects.
250 */
251 systrace_args(sysnum, sa->args, uargs, &n_args);
252 /*
253 * Save probe arguments now so that we can retrieve them if
254 * the getargval method is called from further down the stack.
255 */
256 curthread->t_dtrace_systrace_args = uargs;
257 } else {
258 if ((id = sa->callp->sy_return) == DTRACE_IDNONE)
259 return;
260
261 curthread->t_dtrace_systrace_args = NULL;
262 /* Set arg0 and arg1 as the return value of this syscall. */
263 uargs[0] = uargs[1] = retval;
264 }
265
266 /* Process the probe using the converted argments. */
267 dtrace_probe(id, uargs[0], uargs[1], uargs[2], uargs[3], uargs[4]);
268 }
269 #endif /* NATIVE_ABI */
270 #endif /* __FreeBSD__ */
271
272 #ifdef __NetBSD__
273 static void
systrace_probe(uint32_t id,register_t sysnum,const struct sysent * se,const void * params,const register_t * ret,int error)274 systrace_probe(uint32_t id, register_t sysnum, const struct sysent *se,
275 const void *params, const register_t *ret, int error)
276 {
277 size_t n_args = 0;
278 uintptr_t uargs[SYS_MAXSYSARGS + 3];
279
280 memset(uargs, 0, sizeof(uargs));
281 if (ret == NULL) {
282 /* entry syscall, convert params */
283 systrace_args(sysnum, params, uargs, &n_args);
284 } else {
285 /* return syscall, set values and params: */
286 uargs[0] = ret[0];
287 uargs[1] = ret[1];
288 uargs[2] = error;
289 systrace_args(sysnum, params, uargs + 3, &n_args);
290 }
291 /* Process the probe using the converted argments. */
292 /* XXX: fix for more arguments! */
293 dtrace_probe(id, uargs[0], uargs[1], uargs[2], uargs[3], uargs[4]);
294 }
295 #endif
296
297 static void
systrace_getargdesc(void * arg,dtrace_id_t id,void * parg,dtrace_argdesc_t * desc)298 systrace_getargdesc(void *arg, dtrace_id_t id, void *parg, dtrace_argdesc_t *desc)
299 {
300 int sysnum = SYSTRACE_SYSNUM((uintptr_t)parg);
301
302 if (SYSTRACE_ISENTRY((uintptr_t)parg))
303 systrace_entry_setargdesc(sysnum, desc->dtargd_ndx,
304 desc->dtargd_native, sizeof(desc->dtargd_native));
305 else
306 systrace_return_setargdesc(sysnum, desc->dtargd_ndx,
307 desc->dtargd_native, sizeof(desc->dtargd_native));
308
309 if (desc->dtargd_native[0] == '\0')
310 desc->dtargd_ndx = DTRACE_ARGNONE;
311 }
312
313 static uint64_t
systrace_getargval(void * arg,dtrace_id_t id,void * parg,int argno,int aframes)314 systrace_getargval(void *arg, dtrace_id_t id, void *parg, int argno, int aframes)
315 {
316 uint64_t *uargs;
317
318 uargs = curthread->t_dtrace_systrace_args;
319 if (uargs == NULL)
320 /* This is a return probe. */
321 return (0);
322 #ifdef __FreeBSD__
323 if (argno >= nitems(((struct syscall_args *)NULL)->args))
324 return (0);
325 #endif
326 #ifdef __NetBSD__
327 if (argno >= SYS_MAXSYSARGS)
328 return (0);
329 #endif
330
331 return (uargs[argno]);
332 }
333
334 static void
systrace_provide(void * arg,dtrace_probedesc_t * desc)335 systrace_provide(void *arg, dtrace_probedesc_t *desc)
336 {
337 int i;
338
339 if (desc != NULL)
340 return;
341
342 for (i = 0; i < MAXSYSCALL; i++) {
343 #ifdef __FreeBSD__
344 if (dtrace_probe_lookup(systrace_id, MODNAME,
345 uglyhack.pp_syscallnames[i], "entry") != 0)
346 continue;
347
348 (void)dtrace_probe_create(systrace_id, MODNAME,
349 uglyhack.pp_syscallnames[i], "entry",
350 SYSTRACE_ARTIFICIAL_FRAMES,
351 (void *)((uintptr_t)SYSTRACE_ENTRY(i)));
352 (void)dtrace_probe_create(systrace_id, MODNAME,
353 uglyhack.pp_syscallnames[i], "return",
354 SYSTRACE_ARTIFICIAL_FRAMES,
355 (void *)((uintptr_t)SYSTRACE_RETURN(i)));
356 #else
357 const char *name = ALTSYSCALLNAMES[i] ? ALTSYSCALLNAMES[i] :
358 SYSCALLNAMES[i];
359 if (dtrace_probe_lookup(systrace_id, NULL, name, "entry") != 0)
360 continue;
361
362 (void) dtrace_probe_create(systrace_id, NULL,
363 name, "entry", SYSTRACE_ARTIFICIAL_FRAMES,
364 (void *)(intptr_t)SYSTRACE_ENTRY(i));
365 (void) dtrace_probe_create(systrace_id, NULL,
366 name, "return", SYSTRACE_ARTIFICIAL_FRAMES,
367 (void *)(intptr_t)SYSTRACE_RETURN(i));
368 #endif
369 }
370 }
371
372 static void
systrace_destroy(void * arg,dtrace_id_t id,void * parg)373 systrace_destroy(void *arg, dtrace_id_t id, void *parg)
374 {
375 #ifdef DEBUG
376 int sysnum = SYSTRACE_SYSNUM((uintptr_t)parg);
377
378 /*
379 * There's nothing to do here but assert that we have actually been
380 * disabled.
381 */
382 if (SYSTRACE_ISENTRY((uintptr_t)parg)) {
383 ASSERT(sysent[sysnum].sy_entry == 0);
384 } else {
385 ASSERT(sysent[sysnum].sy_return == 0);
386 }
387 #endif
388 }
389
390 static int
systrace_enable(void * arg,dtrace_id_t id,void * parg)391 systrace_enable(void *arg, dtrace_id_t id, void *parg)
392 {
393 int sysnum = SYSTRACE_SYSNUM((uintptr_t)parg);
394
395 #ifdef __FreeBSD__
396 if (SYSENT[sysnum].sy_systrace_args_func == NULL)
397 SYSENT[sysnum].sy_systrace_args_func = systrace_args;
398 #endif
399
400 if (SYSTRACE_ISENTRY((uintptr_t)parg))
401 SYSENT[sysnum].sy_entry = id;
402 else
403 SYSENT[sysnum].sy_return = id;
404
405 return 0;
406 }
407
408 static void
systrace_disable(void * arg,dtrace_id_t id,void * parg)409 systrace_disable(void *arg, dtrace_id_t id, void *parg)
410 {
411 int sysnum = SYSTRACE_SYSNUM((uintptr_t)parg);
412
413 SYSENT[sysnum].sy_entry = 0;
414 SYSENT[sysnum].sy_return = 0;
415 }
416
417 static void
systrace_load(void * dummy)418 systrace_load(void *dummy)
419 {
420 if (dtrace_register(PROVNAME, &systrace_attr, DTRACE_PRIV_USER, NULL,
421 &systrace_pops, NULL, &systrace_id) != 0)
422 return;
423
424 #ifdef NATIVE_ABI
425 systrace_probe_func = systrace_probe;
426 #endif
427 #ifdef __NetBSD__
428 EMUL.e_dtrace_syscall = systrace_probe;
429 #endif
430 }
431
432
433 static int
systrace_unload()434 systrace_unload()
435 {
436 int error;
437
438 #ifdef NATIVE_ABI
439 systrace_probe_func = NULL;
440 #endif
441 #ifdef __NetBSD__
442 EMUL.e_dtrace_syscall = NULL;
443 #endif
444
445 if ((error = dtrace_unregister(systrace_id)) != 0)
446 return (error);
447
448 return error;
449 }
450
451 #ifdef __FreeBSD__
452 static int
systrace_modevent(module_t mod __unused,int type,void * data __unused)453 systrace_modevent(module_t mod __unused, int type, void *data __unused)
454 {
455 int error;
456
457 error = 0;
458 switch (type) {
459 case MOD_LOAD:
460 break;
461
462 case MOD_UNLOAD:
463 break;
464
465 case MOD_SHUTDOWN:
466 break;
467
468 default:
469 error = EOPNOTSUPP;
470 break;
471
472 }
473 return (error);
474 }
475
476 SYSINIT(systrace_load, SI_SUB_DTRACE_PROVIDER, SI_ORDER_ANY,
477 systrace_load, NULL);
478 SYSUNINIT(systrace_unload, SI_SUB_DTRACE_PROVIDER, SI_ORDER_ANY,
479 systrace_unload, NULL);
480
481 #ifdef LINUX_SYSTRACE
482 DEV_MODULE(systrace_linux, systrace_modevent, NULL);
483 MODULE_VERSION(systrace_linux, 1);
484 #ifdef __amd64__
485 MODULE_DEPEND(systrace_linux, linux64, 1, 1, 1);
486 #else
487 MODULE_DEPEND(systrace_linux, linux, 1, 1, 1);
488 #endif
489 MODULE_DEPEND(systrace_linux, dtrace, 1, 1, 1);
490 MODULE_DEPEND(systrace_linux, opensolaris, 1, 1, 1);
491 #elif defined(LINUX32_SYSTRACE)
492 DEV_MODULE(systrace_linux32, systrace_modevent, NULL);
493 MODULE_VERSION(systrace_linux32, 1);
494 MODULE_DEPEND(systrace_linux32, linux, 1, 1, 1);
495 MODULE_DEPEND(systrace_linux32, dtrace, 1, 1, 1);
496 MODULE_DEPEND(systrace_linux32, opensolaris, 1, 1, 1);
497 #elif defined(FREEBSD32_SYSTRACE)
498 DEV_MODULE(systrace_freebsd32, systrace_modevent, NULL);
499 MODULE_VERSION(systrace_freebsd32, 1);
500 MODULE_DEPEND(systrace_freebsd32, dtrace, 1, 1, 1);
501 MODULE_DEPEND(systrace_freebsd32, opensolaris, 1, 1, 1);
502 #else
503 DEV_MODULE(systrace, systrace_modevent, NULL);
504 MODULE_VERSION(systrace, 1);
505 MODULE_DEPEND(systrace, dtrace, 1, 1, 1);
506 MODULE_DEPEND(systrace, opensolaris, 1, 1, 1);
507 #endif
508 #endif /* __FreeBSD__ */
509
510 #ifdef __NetBSD__
511
512 static int
MODCMD(modcmd_t cmd,void * data)513 MODCMD(modcmd_t cmd, void *data)
514 {
515 switch (cmd) {
516 case MODULE_CMD_INIT:
517 systrace_load(NULL);
518 return 0;
519
520 case MODULE_CMD_FINI:
521 return systrace_unload();
522
523 case MODULE_CMD_AUTOUNLOAD:
524 return EBUSY;
525
526 default:
527 return ENOTTY;
528 }
529 }
530
531 MODULE(MODULE_CLASS_MISC, MODNAME, MODDEP)
532
533 #endif /* __NetBSD__ */
534