xref: /dflybsd-src/sys/kern/kern_ktrace.c (revision b370aff7747b2e03ce9b829fbf2877dffdadfb64)
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: src/sys/kern/kern_ktrace.c,v 1.35.2.6 2002/07/05 22:36:38 darrenr Exp $
35  * $DragonFly: src/sys/kern/kern_ktrace.c,v 1.30 2008/04/14 12:01:50 dillon Exp $
36  */
37 
38 #include "opt_ktrace.h"
39 
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/sysproto.h>
43 #include <sys/kernel.h>
44 #include <sys/proc.h>
45 #include <sys/fcntl.h>
46 #include <sys/lock.h>
47 #include <sys/nlookup.h>
48 #include <sys/vnode.h>
49 #include <sys/ktrace.h>
50 #include <sys/malloc.h>
51 #include <sys/syslog.h>
52 #include <sys/sysent.h>
53 
54 #include <vm/vm_zone.h>
55 
56 #include <sys/mplock2.h>
57 
58 static MALLOC_DEFINE(M_KTRACE, "KTRACE", "KTRACE");
59 
60 #ifdef KTRACE
61 static struct ktr_header *ktrgetheader (int type);
62 static void ktrwrite (struct lwp *, struct ktr_header *, struct uio *);
63 static int ktrcanset (struct thread *,struct proc *);
64 static int ktrsetchildren (struct thread *,struct proc *,int,int, ktrace_node_t);
65 static int ktrops (struct thread *,struct proc *,int,int, ktrace_node_t);
66 
67 /*
68  * MPSAFE
69  */
70 static struct ktr_header *
71 ktrgetheader(int type)
72 {
73 	struct ktr_header *kth;
74 	struct proc *p = curproc;	/* XXX */
75 	struct lwp *lp = curthread->td_lwp;
76 
77 	MALLOC(kth, struct ktr_header *, sizeof (struct ktr_header),
78 		M_KTRACE, M_WAITOK);
79 	kth->ktr_type = type;
80 	/* XXX threaded flag is a hack at the moment */
81 	kth->ktr_flags = (p->p_nthreads > 1) ? KTRH_THREADED : 0;
82 	microtime(&kth->ktr_time);
83 	kth->ktr_pid = p->p_pid;
84 	kth->ktr_tid = lp->lwp_tid;
85 	bcopy(p->p_comm, kth->ktr_comm, MAXCOMLEN + 1);
86 	return (kth);
87 }
88 
89 void
90 ktrsyscall(struct lwp *lp, int code, int narg, register_t args[])
91 {
92 	struct	ktr_header *kth;
93 	struct	ktr_syscall *ktp;
94 	int len;
95 	register_t *argp;
96 	int i;
97 
98 	len = offsetof(struct ktr_syscall, ktr_args) +
99 	      (narg * sizeof(register_t));
100 
101 	/*
102 	 * Setting the active bit prevents a ktrace recursion from the
103 	 * ktracing op itself.
104 	 */
105 	lp->lwp_traceflag |= KTRFAC_ACTIVE;
106 	kth = ktrgetheader(KTR_SYSCALL);
107 	MALLOC(ktp, struct ktr_syscall *, len, M_KTRACE, M_WAITOK);
108 	ktp->ktr_code = code;
109 	ktp->ktr_narg = narg;
110 	argp = &ktp->ktr_args[0];
111 	for (i = 0; i < narg; i++)
112 		*argp++ = args[i];
113 	kth->ktr_buf = (caddr_t)ktp;
114 	kth->ktr_len = len;
115 	ktrwrite(lp, kth, NULL);
116 	FREE(ktp, M_KTRACE);
117 	FREE(kth, M_KTRACE);
118 	lp->lwp_traceflag &= ~KTRFAC_ACTIVE;
119 }
120 
121 void
122 ktrsysret(struct lwp *lp, int code, int error, register_t retval)
123 {
124 	struct ktr_header *kth;
125 	struct ktr_sysret ktp;
126 
127 	lp->lwp_traceflag |= KTRFAC_ACTIVE;
128 	kth = ktrgetheader(KTR_SYSRET);
129 	ktp.ktr_code = code;
130 	ktp.ktr_error = error;
131 	if (error == 0)
132 		ktp.ktr_retval = retval;		/* what about val2 ? */
133 	else
134 		ktp.ktr_retval = 0;
135 
136 	kth->ktr_buf = (caddr_t)&ktp;
137 	kth->ktr_len = sizeof(struct ktr_sysret);
138 
139 	ktrwrite(lp, kth, NULL);
140 	FREE(kth, M_KTRACE);
141 	lp->lwp_traceflag &= ~KTRFAC_ACTIVE;
142 }
143 
144 void
145 ktrnamei(struct lwp *lp, char *path)
146 {
147 	struct ktr_header *kth;
148 
149 	lp->lwp_traceflag |= KTRFAC_ACTIVE;
150 	kth = ktrgetheader(KTR_NAMEI);
151 	kth->ktr_len = strlen(path);
152 	kth->ktr_buf = path;
153 
154 	ktrwrite(lp, kth, NULL);
155 	FREE(kth, M_KTRACE);
156 	lp->lwp_traceflag &= ~KTRFAC_ACTIVE;
157 }
158 
159 void
160 ktrgenio(struct lwp *lp, int fd, enum uio_rw rw, struct uio *uio, int error)
161 {
162 	struct ktr_header *kth;
163 	struct ktr_genio ktg;
164 
165 	if (error)
166 		return;
167 	lp->lwp_traceflag |= KTRFAC_ACTIVE;
168 	kth = ktrgetheader(KTR_GENIO);
169 	ktg.ktr_fd = fd;
170 	ktg.ktr_rw = rw;
171 	kth->ktr_buf = (caddr_t)&ktg;
172 	kth->ktr_len = sizeof(struct ktr_genio);
173 	uio->uio_offset = 0;
174 	uio->uio_rw = UIO_WRITE;
175 
176 	ktrwrite(lp, kth, uio);
177 	FREE(kth, M_KTRACE);
178 	lp->lwp_traceflag &= ~KTRFAC_ACTIVE;
179 }
180 
181 void
182 ktrpsig(struct lwp *lp, int sig, sig_t action, sigset_t *mask, int code)
183 {
184 	struct ktr_header *kth;
185 	struct ktr_psig	kp;
186 
187 	lp->lwp_traceflag |= KTRFAC_ACTIVE;
188 	kth = ktrgetheader(KTR_PSIG);
189 	kp.signo = (char)sig;
190 	kp.action = action;
191 	kp.mask = *mask;
192 	kp.code = code;
193 	kth->ktr_buf = (caddr_t)&kp;
194 	kth->ktr_len = sizeof (struct ktr_psig);
195 
196 	ktrwrite(lp, kth, NULL);
197 	FREE(kth, M_KTRACE);
198 	lp->lwp_traceflag &= ~KTRFAC_ACTIVE;
199 }
200 
201 void
202 ktrcsw(struct lwp *lp, int out, int user)
203 {
204 	struct ktr_header *kth;
205 	struct	ktr_csw kc;
206 
207 	lp->lwp_traceflag |= KTRFAC_ACTIVE;
208 	kth = ktrgetheader(KTR_CSW);
209 	kc.out = out;
210 	kc.user = user;
211 	kth->ktr_buf = (caddr_t)&kc;
212 	kth->ktr_len = sizeof (struct ktr_csw);
213 
214 	ktrwrite(lp, kth, NULL);
215 	FREE(kth, M_KTRACE);
216 	lp->lwp_traceflag &= ~KTRFAC_ACTIVE;
217 }
218 #endif
219 
220 /* Interface and common routines */
221 
222 #ifdef KTRACE
223 /*
224  * ktrace system call
225  */
226 struct ktrace_clear_info {
227 	ktrace_node_t tracenode;
228 	int rootclear;
229 	int error;
230 };
231 
232 static int ktrace_clear_callback(struct proc *p, void *data);
233 
234 #endif
235 
236 /*
237  * MPALMOSTSAFE
238  */
239 int
240 sys_ktrace(struct ktrace_args *uap)
241 {
242 #ifdef KTRACE
243 	struct ktrace_clear_info info;
244 	struct thread *td = curthread;
245 	struct proc *curp = td->td_proc;
246 	struct proc *p;
247 	struct pgrp *pg;
248 	int facs = uap->facs & ~KTRFAC_ROOT;
249 	int ops = KTROP(uap->ops);
250 	int descend = uap->ops & KTRFLAG_DESCEND;
251 	int ret = 0;
252 	int error = 0;
253 	struct nlookupdata nd;
254 	ktrace_node_t tracenode = NULL;
255 
256 	get_mplock();
257 	curp->p_traceflag |= KTRFAC_ACTIVE;
258 	if (ops != KTROP_CLEAR) {
259 		/*
260 		 * an operation which requires a file argument.
261 		 */
262 		error = nlookup_init(&nd, uap->fname,
263 					UIO_USERSPACE, NLC_LOCKVP);
264 		if (error == 0)
265 			error = vn_open(&nd, NULL, FREAD|FWRITE|O_NOFOLLOW, 0);
266 		if (error == 0 && nd.nl_open_vp->v_type != VREG)
267 			error = EACCES;
268 		if (error) {
269 			curp->p_traceflag &= ~KTRFAC_ACTIVE;
270 			nlookup_done(&nd);
271 			goto done;
272 		}
273 		MALLOC(tracenode, ktrace_node_t, sizeof (struct ktrace_node),
274 		       M_KTRACE, M_WAITOK | M_ZERO);
275 		tracenode->kn_vp = nd.nl_open_vp;
276 		tracenode->kn_refs = 1;
277 		nd.nl_open_vp = NULL;
278 		nlookup_done(&nd);
279 		vn_unlock(tracenode->kn_vp);
280 	}
281 	/*
282 	 * Clear all uses of the tracefile.  Not the most efficient operation
283 	 * in the world.
284 	 */
285 	if (ops == KTROP_CLEARFILE) {
286 		info.tracenode = tracenode;
287 		info.error = 0;
288 		info.rootclear = 0;
289 		allproc_scan(ktrace_clear_callback, &info);
290 		error = info.error;
291 		goto done;
292 	}
293 	/*
294 	 * need something to (un)trace (XXX - why is this here?)
295 	 */
296 	if (!facs) {
297 		error = EINVAL;
298 		goto done;
299 	}
300 	/*
301 	 * do it
302 	 */
303 	if (uap->pid < 0) {
304 		/*
305 		 * By process group.  Process group is referenced, preventing
306 		 * disposal.
307 		 */
308 		pg = pgfind(-uap->pid);
309 		if (pg == NULL) {
310 			error = ESRCH;
311 			goto done;
312 		}
313 		lwkt_gettoken(&pg->pg_token);
314 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
315 			PHOLD(p);
316 			if (descend)
317 				ret |= ktrsetchildren(td, p, ops, facs, tracenode);
318 			else
319 				ret |= ktrops(td, p, ops, facs, tracenode);
320 			PRELE(p);
321 		}
322 		lwkt_reltoken(&pg->pg_token);
323 		pgrel(pg);
324 	} else {
325 		/*
326 		 * by pid
327 		 */
328 		p = pfind(uap->pid);
329 		if (p == NULL) {
330 			error = ESRCH;
331 			goto done;
332 		}
333 		if (descend)
334 			ret |= ktrsetchildren(td, p, ops, facs, tracenode);
335 		else
336 			ret |= ktrops(td, p, ops, facs, tracenode);
337 		PRELE(p);
338 	}
339 	if (!ret)
340 		error = EPERM;
341 done:
342 	if (tracenode)
343 		ktrdestroy(&tracenode);
344 	curp->p_traceflag &= ~KTRFAC_ACTIVE;
345 	rel_mplock();
346 	return (error);
347 #else
348 	return ENOSYS;
349 #endif
350 }
351 
352 #ifdef KTRACE
353 
354 /*
355  * NOTE: NOT MPSAFE (yet)
356  */
357 static int
358 ktrace_clear_callback(struct proc *p, void *data)
359 {
360 	struct ktrace_clear_info *info = data;
361 
362 	if (p->p_tracenode) {
363 		if (info->rootclear) {
364 			if (p->p_tracenode == info->tracenode) {
365 				ktrdestroy(&p->p_tracenode);
366 				p->p_traceflag = 0;
367 			}
368 		} else {
369 			if (p->p_tracenode->kn_vp == info->tracenode->kn_vp) {
370 				if (ktrcanset(curthread, p)) {
371 					ktrdestroy(&p->p_tracenode);
372 					p->p_traceflag = 0;
373 				} else {
374 					info->error = EPERM;
375 				}
376 			}
377 		}
378 	}
379 	return(0);
380 }
381 
382 #endif
383 
384 /*
385  * utrace system call
386  *
387  * MPALMOSTSAFE
388  */
389 int
390 sys_utrace(struct utrace_args *uap)
391 {
392 #ifdef KTRACE
393 	struct ktr_header *kth;
394 	struct thread *td = curthread;	/* XXX */
395 	caddr_t cp;
396 
397 	if (!KTRPOINT(td, KTR_USER))
398 		return (0);
399 	if (uap->len > KTR_USER_MAXLEN)
400 		return (EINVAL);
401 	td->td_lwp->lwp_traceflag |= KTRFAC_ACTIVE;
402 	kth = ktrgetheader(KTR_USER);
403 	MALLOC(cp, caddr_t, uap->len, M_KTRACE, M_WAITOK);
404 	if (!copyin(uap->addr, cp, uap->len)) {
405 		kth->ktr_buf = cp;
406 		kth->ktr_len = uap->len;
407 		ktrwrite(td->td_lwp, kth, NULL);
408 	}
409 	FREE(kth, M_KTRACE);
410 	FREE(cp, M_KTRACE);
411 	td->td_lwp->lwp_traceflag &= ~KTRFAC_ACTIVE;
412 
413 	return (0);
414 #else
415 	return (ENOSYS);
416 #endif
417 }
418 
419 void
420 ktrdestroy(struct ktrace_node **tracenodep)
421 {
422 	ktrace_node_t tracenode;
423 
424 	if ((tracenode = *tracenodep) != NULL) {
425 		*tracenodep = NULL;
426 		KKASSERT(tracenode->kn_refs > 0);
427 		if (atomic_fetchadd_int(&tracenode->kn_refs, -1) == 1) {
428 			vn_close(tracenode->kn_vp, FREAD|FWRITE);
429 			tracenode->kn_vp = NULL;
430 			FREE(tracenode, M_KTRACE);
431 		}
432 	}
433 }
434 
435 /*
436  * This allows a process to inherit a ref on a tracenode and is also used
437  * as a temporary ref to prevent a tracenode from being destroyed out from
438  * under an active operation.
439  */
440 ktrace_node_t
441 ktrinherit(ktrace_node_t tracenode)
442 {
443 	if (tracenode) {
444 		KKASSERT(tracenode->kn_refs > 0);
445 		atomic_add_int(&tracenode->kn_refs, 1);
446 	}
447 	return(tracenode);
448 }
449 
450 #ifdef KTRACE
451 static int
452 ktrops(struct thread *td, struct proc *p, int ops, int facs,
453        ktrace_node_t tracenode)
454 {
455 	ktrace_node_t oldnode;
456 
457 	if (!ktrcanset(td, p))
458 		return (0);
459 	if (ops == KTROP_SET) {
460 		if ((oldnode = p->p_tracenode) != tracenode) {
461 			p->p_tracenode = ktrinherit(tracenode);
462 			ktrdestroy(&oldnode);
463 		}
464 		p->p_traceflag |= facs;
465 		if (td->td_ucred->cr_uid == 0)
466 			p->p_traceflag |= KTRFAC_ROOT;
467 	} else {
468 		/* KTROP_CLEAR */
469 		if (((p->p_traceflag &= ~facs) & KTRFAC_MASK) == 0) {
470 			/* no more tracing */
471 			p->p_traceflag = 0;
472 			ktrdestroy(&p->p_tracenode);
473 		}
474 	}
475 
476 	return (1);
477 }
478 
479 static int
480 ktrsetchildren(struct thread *td, struct proc *top, int ops, int facs,
481 	       ktrace_node_t tracenode)
482 {
483 	struct proc *p;
484 	struct proc *np;
485 	int ret = 0;
486 
487 	p = top;
488 	PHOLD(p);
489 	lwkt_gettoken(&p->p_token);
490 
491 	for (;;) {
492 		ret |= ktrops(td, p, ops, facs, tracenode);
493 
494 		/*
495 		 * If this process has children, descend to them next,
496 		 * otherwise do any siblings, and if done with this level,
497 		 * follow back up the tree (but not past top).
498 		 */
499 		if ((np = LIST_FIRST(&p->p_children)) != NULL) {
500 			PHOLD(np);
501 		}
502 		while (np == NULL) {
503 			if (p == top)
504 				break;
505 			if ((np = LIST_NEXT(p, p_sibling)) != NULL) {
506 				PHOLD(np);
507 				break;
508 			}
509 
510 			/*
511 			 * recurse up to parent, set p in our inner
512 			 * loop when doing this.  np can be NULL if
513 			 * we race a reparenting to init (thus 'top'
514 			 * is skipped past and never encountered).
515 			 */
516 			np = p->p_pptr;
517 			if (np == NULL)
518 				break;
519 			PHOLD(np);
520 			lwkt_reltoken(&p->p_token);
521 			PRELE(p);
522 			p = np;
523 			lwkt_gettoken(&p->p_token);
524 			np = NULL;
525 		}
526 		lwkt_reltoken(&p->p_token);
527 		PRELE(p);
528 		p = np;
529 		if (p == NULL)
530 			break;
531 		/* Already held, but we need the token too */
532 		lwkt_gettoken(&p->p_token);
533 	}
534 	return (ret);
535 }
536 
537 static void
538 ktrwrite(struct lwp *lp, struct ktr_header *kth, struct uio *uio)
539 {
540 	struct ktrace_clear_info info;
541 	struct uio auio;
542 	struct iovec aiov[2];
543 	int error;
544 	ktrace_node_t tracenode;
545 
546 	/*
547 	 * We have to ref our tracenode to prevent it from being ripped out
548 	 * from under us while we are trying to use it.   p_tracenode can
549 	 * go away at any time if another process gets a write error.
550 	 *
551 	 * XXX not MP safe
552 	 */
553 	if (lp->lwp_proc->p_tracenode == NULL)
554 		return;
555 	tracenode = ktrinherit(lp->lwp_proc->p_tracenode);
556 	auio.uio_iov = &aiov[0];
557 	auio.uio_offset = 0;
558 	auio.uio_segflg = UIO_SYSSPACE;
559 	auio.uio_rw = UIO_WRITE;
560 	aiov[0].iov_base = (caddr_t)kth;
561 	aiov[0].iov_len = sizeof(struct ktr_header);
562 	auio.uio_resid = sizeof(struct ktr_header);
563 	auio.uio_iovcnt = 1;
564 	auio.uio_td = curthread;
565 	if (kth->ktr_len > 0) {
566 		auio.uio_iovcnt++;
567 		aiov[1].iov_base = kth->ktr_buf;
568 		aiov[1].iov_len = kth->ktr_len;
569 		auio.uio_resid += kth->ktr_len;
570 		if (uio != NULL)
571 			kth->ktr_len += uio->uio_resid;
572 	}
573 	vn_lock(tracenode->kn_vp, LK_EXCLUSIVE | LK_RETRY);
574 	error = VOP_WRITE(tracenode->kn_vp, &auio,
575 			  IO_UNIT | IO_APPEND, lp->lwp_thread->td_ucred);
576 	if (error == 0 && uio != NULL) {
577 		error = VOP_WRITE(tracenode->kn_vp, uio,
578 			      IO_UNIT | IO_APPEND, lp->lwp_thread->td_ucred);
579 	}
580 	vn_unlock(tracenode->kn_vp);
581 	if (error) {
582 		/*
583 		 * If an error occured, give up tracing on all processes
584 		 * using this tracenode.  This is not MP safe but is
585 		 * blocking-safe.
586 		 */
587 		log(LOG_NOTICE,
588 		    "ktrace write failed, errno %d, tracing stopped\n", error);
589 		info.tracenode = tracenode;
590 		info.error = 0;
591 		info.rootclear = 1;
592 		allproc_scan(ktrace_clear_callback, &info);
593 	}
594 	ktrdestroy(&tracenode);
595 }
596 
597 /*
598  * Return true if caller has permission to set the ktracing state
599  * of target.  Essentially, the target can't possess any
600  * more permissions than the caller.  KTRFAC_ROOT signifies that
601  * root previously set the tracing status on the target process, and
602  * so, only root may further change it.
603  *
604  * TODO: check groups.  use caller effective gid.
605  */
606 static int
607 ktrcanset(struct thread *calltd, struct proc *targetp)
608 {
609 	struct ucred *caller = calltd->td_ucred;
610 	struct ucred *target = targetp->p_ucred;
611 
612 	if (!PRISON_CHECK(caller, target))
613 		return (0);
614 	if ((caller->cr_uid == target->cr_ruid &&
615 	     target->cr_ruid == target->cr_svuid &&
616 	     caller->cr_rgid == target->cr_rgid &&	/* XXX */
617 	     target->cr_rgid == target->cr_svgid &&
618 	     (targetp->p_traceflag & KTRFAC_ROOT) == 0 &&
619 	     (targetp->p_flags & P_SUGID) == 0) ||
620 	     caller->cr_uid == 0)
621 		return (1);
622 
623 	return (0);
624 }
625 
626 #endif /* KTRACE */
627