xref: /freebsd/sys/kern/kern_ktrace.c (revision 817420dc8eac7df799c78f5309b75092b7f7cd40)
1 /*
2  * Copyright (c) 1989, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *	This product includes software developed by the University of
16  *	California, Berkeley and its contributors.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *	@(#)kern_ktrace.c	8.2 (Berkeley) 9/23/93
34  * $FreeBSD$
35  */
36 
37 #include "opt_ktrace.h"
38 
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/sysproto.h>
42 #include <sys/kernel.h>
43 #include <sys/proc.h>
44 #include <sys/fcntl.h>
45 #include <sys/lock.h>
46 #include <sys/namei.h>
47 #include <sys/vnode.h>
48 #include <sys/ktrace.h>
49 #include <sys/malloc.h>
50 #include <sys/syslog.h>
51 
52 static MALLOC_DEFINE(M_KTRACE, "KTRACE", "KTRACE");
53 
54 #ifdef KTRACE
55 static struct ktr_header *ktrgetheader __P((int type));
56 static void ktrwrite __P((struct vnode *, struct ktr_header *, struct uio *));
57 static int ktrcanset __P((struct proc *,struct proc *));
58 static int ktrsetchildren __P((struct proc *,struct proc *,int,int,struct vnode *));
59 static int ktrops __P((struct proc *,struct proc *,int,int,struct vnode *));
60 
61 
62 static struct ktr_header *
63 ktrgetheader(type)
64 	int type;
65 {
66 	register struct ktr_header *kth;
67 	struct proc *p = curproc;	/* XXX */
68 
69 	MALLOC(kth, struct ktr_header *, sizeof (struct ktr_header),
70 		M_KTRACE, M_WAITOK);
71 	kth->ktr_type = type;
72 	microtime(&kth->ktr_time);
73 	kth->ktr_pid = p->p_pid;
74 	bcopy(p->p_comm, kth->ktr_comm, MAXCOMLEN + 1);
75 	return (kth);
76 }
77 
78 void
79 ktrsyscall(vp, code, narg, args)
80 	struct vnode *vp;
81 	int code, narg;
82 	register_t args[];
83 {
84 	struct	ktr_header *kth;
85 	struct	ktr_syscall *ktp;
86 	register int len = offsetof(struct ktr_syscall, ktr_args) +
87 	    (narg * sizeof(register_t));
88 	struct proc *p = curproc;	/* XXX */
89 	register_t *argp;
90 	int i;
91 
92 	p->p_traceflag |= KTRFAC_ACTIVE;
93 	kth = ktrgetheader(KTR_SYSCALL);
94 	MALLOC(ktp, struct ktr_syscall *, len, M_KTRACE, M_WAITOK);
95 	ktp->ktr_code = code;
96 	ktp->ktr_narg = narg;
97 	argp = &ktp->ktr_args[0];
98 	for (i = 0; i < narg; i++)
99 		*argp++ = args[i];
100 	kth->ktr_buffer = (caddr_t)ktp;
101 	kth->ktr_len = len;
102 	ktrwrite(vp, kth, NULL);
103 	FREE(ktp, M_KTRACE);
104 	FREE(kth, M_KTRACE);
105 	p->p_traceflag &= ~KTRFAC_ACTIVE;
106 }
107 
108 void
109 ktrsysret(vp, code, error, retval)
110 	struct vnode *vp;
111 	int code, error;
112 	register_t retval;
113 {
114 	struct ktr_header *kth;
115 	struct ktr_sysret ktp;
116 	struct proc *p = curproc;	/* XXX */
117 
118 	p->p_traceflag |= KTRFAC_ACTIVE;
119 	kth = ktrgetheader(KTR_SYSRET);
120 	ktp.ktr_code = code;
121 	ktp.ktr_error = error;
122 	ktp.ktr_retval = retval;		/* what about val2 ? */
123 
124 	kth->ktr_buffer = (caddr_t)&ktp;
125 	kth->ktr_len = sizeof(struct ktr_sysret);
126 
127 	ktrwrite(vp, kth, NULL);
128 	FREE(kth, M_KTRACE);
129 	p->p_traceflag &= ~KTRFAC_ACTIVE;
130 }
131 
132 void
133 ktrnamei(vp, path)
134 	struct vnode *vp;
135 	char *path;
136 {
137 	struct ktr_header *kth;
138 	struct proc *p = curproc;	/* XXX */
139 
140 	p->p_traceflag |= KTRFAC_ACTIVE;
141 	kth = ktrgetheader(KTR_NAMEI);
142 	kth->ktr_len = strlen(path);
143 	kth->ktr_buffer = path;
144 
145 	ktrwrite(vp, kth, NULL);
146 	FREE(kth, M_KTRACE);
147 	p->p_traceflag &= ~KTRFAC_ACTIVE;
148 }
149 
150 void
151 ktrgenio(vp, fd, rw, uio, error)
152 	struct vnode *vp;
153 	int fd;
154 	enum uio_rw rw;
155 	struct uio *uio;
156 	int error;
157 {
158 	struct ktr_header *kth;
159 	struct ktr_genio ktg;
160 	struct proc *p = curproc;	/* XXX */
161 
162 	if (error)
163 		return;
164 	p->p_traceflag |= KTRFAC_ACTIVE;
165 	kth = ktrgetheader(KTR_GENIO);
166 	ktg.ktr_fd = fd;
167 	ktg.ktr_rw = rw;
168 	kth->ktr_buffer = (caddr_t)&ktg;
169 	kth->ktr_len = sizeof(struct ktr_genio);
170 	uio->uio_offset = 0;
171 	uio->uio_rw = UIO_WRITE;
172 
173 	ktrwrite(vp, kth, uio);
174 	FREE(kth, M_KTRACE);
175 	p->p_traceflag &= ~KTRFAC_ACTIVE;
176 }
177 
178 void
179 ktrpsig(vp, sig, action, mask, code)
180 	struct vnode *vp;
181 	int sig;
182 	sig_t action;
183 	sigset_t *mask;
184 	int code;
185 {
186 	struct ktr_header *kth;
187 	struct ktr_psig	kp;
188 	struct proc *p = curproc;	/* XXX */
189 
190 	p->p_traceflag |= KTRFAC_ACTIVE;
191 	kth = ktrgetheader(KTR_PSIG);
192 	kp.signo = (char)sig;
193 	kp.action = action;
194 	kp.mask = *mask;
195 	kp.code = code;
196 	kth->ktr_buffer = (caddr_t)&kp;
197 	kth->ktr_len = sizeof (struct ktr_psig);
198 
199 	ktrwrite(vp, kth, NULL);
200 	FREE(kth, M_KTRACE);
201 	p->p_traceflag &= ~KTRFAC_ACTIVE;
202 }
203 
204 void
205 ktrcsw(vp, out, user)
206 	struct vnode *vp;
207 	int out, user;
208 {
209 	struct ktr_header *kth;
210 	struct	ktr_csw kc;
211 	struct proc *p = curproc;	/* XXX */
212 
213 	p->p_traceflag |= KTRFAC_ACTIVE;
214 	kth = ktrgetheader(KTR_CSW);
215 	kc.out = out;
216 	kc.user = user;
217 	kth->ktr_buffer = (caddr_t)&kc;
218 	kth->ktr_len = sizeof (struct ktr_csw);
219 
220 	ktrwrite(vp, kth, NULL);
221 	FREE(kth, M_KTRACE);
222 	p->p_traceflag &= ~KTRFAC_ACTIVE;
223 }
224 #endif
225 
226 /* Interface and common routines */
227 
228 /*
229  * ktrace system call
230  */
231 #ifndef _SYS_SYSPROTO_H_
232 struct ktrace_args {
233 	char	*fname;
234 	int	ops;
235 	int	facs;
236 	int	pid;
237 };
238 #endif
239 /* ARGSUSED */
240 int
241 ktrace(curp, uap)
242 	struct proc *curp;
243 	register struct ktrace_args *uap;
244 {
245 #ifdef KTRACE
246 	register struct vnode *vp = NULL;
247 	register struct proc *p;
248 	struct pgrp *pg;
249 	int facs = uap->facs & ~KTRFAC_ROOT;
250 	int ops = KTROP(uap->ops);
251 	int descend = uap->ops & KTRFLAG_DESCEND;
252 	int ret = 0;
253 	int flags, error = 0;
254 	struct nameidata nd;
255 
256 	curp->p_traceflag |= KTRFAC_ACTIVE;
257 	if (ops != KTROP_CLEAR) {
258 		/*
259 		 * an operation which requires a file argument.
260 		 */
261 		NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, uap->fname, curp);
262 		flags = FREAD | FWRITE | O_NOFOLLOW;
263 		error = vn_open(&nd, &flags, 0);
264 		if (error) {
265 			curp->p_traceflag &= ~KTRFAC_ACTIVE;
266 			return (error);
267 		}
268 		NDFREE(&nd, NDF_ONLY_PNBUF);
269 		vp = nd.ni_vp;
270 		VOP_UNLOCK(vp, 0, curp);
271 		if (vp->v_type != VREG) {
272 			(void) vn_close(vp, FREAD|FWRITE, curp->p_ucred, curp);
273 			curp->p_traceflag &= ~KTRFAC_ACTIVE;
274 			return (EACCES);
275 		}
276 	}
277 	/*
278 	 * Clear all uses of the tracefile
279 	 */
280 	if (ops == KTROP_CLEARFILE) {
281 		lockmgr(&allproc_lock, LK_SHARED, NULL, CURPROC);
282 		LIST_FOREACH(p, &allproc, p_list) {
283 			if (p->p_tracep == vp) {
284 				if (ktrcanset(curp, p)) {
285 					p->p_tracep = NULL;
286 					p->p_traceflag = 0;
287 					(void) vn_close(vp, FREAD|FWRITE,
288 						p->p_ucred, p);
289 				} else
290 					error = EPERM;
291 			}
292 		}
293 		lockmgr(&allproc_lock, LK_RELEASE, NULL, CURPROC);
294 		goto done;
295 	}
296 	/*
297 	 * need something to (un)trace (XXX - why is this here?)
298 	 */
299 	if (!facs) {
300 		error = EINVAL;
301 		goto done;
302 	}
303 	/*
304 	 * do it
305 	 */
306 	if (uap->pid < 0) {
307 		/*
308 		 * by process group
309 		 */
310 		pg = pgfind(-uap->pid);
311 		if (pg == NULL) {
312 			error = ESRCH;
313 			goto done;
314 		}
315 		LIST_FOREACH(p, &pg->pg_members, p_pglist)
316 			if (descend)
317 				ret |= ktrsetchildren(curp, p, ops, facs, vp);
318 			else
319 				ret |= ktrops(curp, p, ops, facs, vp);
320 
321 	} else {
322 		/*
323 		 * by pid
324 		 */
325 		p = pfind(uap->pid);
326 		if (p == NULL) {
327 			error = ESRCH;
328 			goto done;
329 		}
330 		if (descend)
331 			ret |= ktrsetchildren(curp, p, ops, facs, vp);
332 		else
333 			ret |= ktrops(curp, p, ops, facs, vp);
334 	}
335 	if (!ret)
336 		error = EPERM;
337 done:
338 	if (vp != NULL)
339 		(void) vn_close(vp, FWRITE, curp->p_ucred, curp);
340 	curp->p_traceflag &= ~KTRFAC_ACTIVE;
341 	return (error);
342 #else
343 	return ENOSYS;
344 #endif
345 }
346 
347 /*
348  * utrace system call
349  */
350 /* ARGSUSED */
351 int
352 utrace(curp, uap)
353 	struct proc *curp;
354 	register struct utrace_args *uap;
355 {
356 #ifdef KTRACE
357 	struct ktr_header *kth;
358 	struct proc *p = curproc;	/* XXX */
359 	register caddr_t cp;
360 
361 	if (!KTRPOINT(p, KTR_USER))
362 		return (0);
363 	p->p_traceflag |= KTRFAC_ACTIVE;
364 	kth = ktrgetheader(KTR_USER);
365 	MALLOC(cp, caddr_t, uap->len, M_KTRACE, M_WAITOK);
366 	if (!copyin(uap->addr, cp, uap->len)) {
367 		kth->ktr_buffer = cp;
368 		kth->ktr_len = uap->len;
369 		ktrwrite(p->p_tracep, kth, NULL);
370 	}
371 	FREE(kth, M_KTRACE);
372 	FREE(cp, M_KTRACE);
373 	p->p_traceflag &= ~KTRFAC_ACTIVE;
374 
375 	return (0);
376 #else
377 	return (ENOSYS);
378 #endif
379 }
380 
381 #ifdef KTRACE
382 static int
383 ktrops(curp, p, ops, facs, vp)
384 	struct proc *p, *curp;
385 	int ops, facs;
386 	struct vnode *vp;
387 {
388 
389 	if (!ktrcanset(curp, p))
390 		return (0);
391 	if (ops == KTROP_SET) {
392 		if (p->p_tracep != vp) {
393 			/*
394 			 * if trace file already in use, relinquish
395 			 */
396 			if (p->p_tracep != NULL)
397 				vrele(p->p_tracep);
398 			VREF(vp);
399 			p->p_tracep = vp;
400 		}
401 		p->p_traceflag |= facs;
402 		if (curp->p_ucred->cr_uid == 0)
403 			p->p_traceflag |= KTRFAC_ROOT;
404 	} else {
405 		/* KTROP_CLEAR */
406 		if (((p->p_traceflag &= ~facs) & KTRFAC_MASK) == 0) {
407 			/* no more tracing */
408 			p->p_traceflag = 0;
409 			if (p->p_tracep != NULL) {
410 				vrele(p->p_tracep);
411 				p->p_tracep = NULL;
412 			}
413 		}
414 	}
415 
416 	return (1);
417 }
418 
419 static int
420 ktrsetchildren(curp, top, ops, facs, vp)
421 	struct proc *curp, *top;
422 	int ops, facs;
423 	struct vnode *vp;
424 {
425 	register struct proc *p;
426 	register int ret = 0;
427 
428 	p = top;
429 	for (;;) {
430 		ret |= ktrops(curp, p, ops, facs, vp);
431 		/*
432 		 * If this process has children, descend to them next,
433 		 * otherwise do any siblings, and if done with this level,
434 		 * follow back up the tree (but not past top).
435 		 */
436 		if (!LIST_EMPTY(&p->p_children))
437 			p = LIST_FIRST(&p->p_children);
438 		else for (;;) {
439 			if (p == top)
440 				return (ret);
441 			if (LIST_NEXT(p, p_sibling)) {
442 				p = LIST_NEXT(p, p_sibling);
443 				break;
444 			}
445 			p = p->p_pptr;
446 		}
447 	}
448 	/*NOTREACHED*/
449 }
450 
451 static void
452 ktrwrite(vp, kth, uio)
453 	struct vnode *vp;
454 	register struct ktr_header *kth;
455 	struct uio *uio;
456 {
457 	struct uio auio;
458 	struct iovec aiov[2];
459 	struct proc *p = curproc;	/* XXX */
460 	struct mount *mp;
461 	int error;
462 
463 	if (vp == NULL)
464 		return;
465 	auio.uio_iov = &aiov[0];
466 	auio.uio_offset = 0;
467 	auio.uio_segflg = UIO_SYSSPACE;
468 	auio.uio_rw = UIO_WRITE;
469 	aiov[0].iov_base = (caddr_t)kth;
470 	aiov[0].iov_len = sizeof(struct ktr_header);
471 	auio.uio_resid = sizeof(struct ktr_header);
472 	auio.uio_iovcnt = 1;
473 	auio.uio_procp = curproc;
474 	if (kth->ktr_len > 0) {
475 		auio.uio_iovcnt++;
476 		aiov[1].iov_base = kth->ktr_buffer;
477 		aiov[1].iov_len = kth->ktr_len;
478 		auio.uio_resid += kth->ktr_len;
479 		if (uio != NULL)
480 			kth->ktr_len += uio->uio_resid;
481 	}
482 	vn_start_write(vp, &mp, V_WAIT);
483 	vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, p);
484 	(void)VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
485 	error = VOP_WRITE(vp, &auio, IO_UNIT | IO_APPEND, p->p_ucred);
486 	if (error == 0 && uio != NULL) {
487 		(void)VOP_LEASE(vp, p, p->p_ucred, LEASE_WRITE);
488 		error = VOP_WRITE(vp, uio, IO_UNIT | IO_APPEND, p->p_ucred);
489 	}
490 	VOP_UNLOCK(vp, 0, p);
491 	vn_finished_write(mp);
492 	if (!error)
493 		return;
494 	/*
495 	 * If error encountered, give up tracing on this vnode.
496 	 */
497 	log(LOG_NOTICE, "ktrace write failed, errno %d, tracing stopped\n",
498 	    error);
499 	lockmgr(&allproc_lock, LK_SHARED, NULL, CURPROC);
500 	LIST_FOREACH(p, &allproc, p_list) {
501 		if (p->p_tracep == vp) {
502 			p->p_tracep = NULL;
503 			p->p_traceflag = 0;
504 			vrele(vp);
505 		}
506 	}
507 	lockmgr(&allproc_lock, LK_RELEASE, NULL, CURPROC);
508 }
509 
510 /*
511  * Return true if caller has permission to set the ktracing state
512  * of target.  Essentially, the target can't possess any
513  * more permissions than the caller.  KTRFAC_ROOT signifies that
514  * root previously set the tracing status on the target process, and
515  * so, only root may further change it.
516  *
517  * XXX: These checks are stronger than for ptrace()
518  *
519  * TODO: check groups.  use caller effective gid.
520  */
521 static int
522 ktrcanset(callp, targetp)
523 	struct proc *callp, *targetp;
524 {
525 	register struct pcred *caller = callp->p_cred;
526 	register struct pcred *target = targetp->p_cred;
527 
528 	if (!PRISON_CHECK(callp, targetp))
529 		return (0);
530 	if ((caller->pc_ucred->cr_uid == target->p_ruid &&
531 	     target->p_ruid == target->p_svuid &&
532 	     caller->p_rgid == target->p_rgid &&	/* XXX */
533 	     target->p_rgid == target->p_svgid &&
534 	     (targetp->p_traceflag & KTRFAC_ROOT) == 0) ||
535 	     caller->pc_ucred->cr_uid == 0)
536 		return (1);
537 
538 	return (0);
539 }
540 
541 #endif /* KTRACE */
542