xref: /netbsd-src/sys/kern/kern_ktrace.c (revision bada23909e740596d0a3785a73bd3583a9807fb8) 
1 /*	$NetBSD: kern_ktrace.c,v 1.33 1998/09/11 12:50:10 mycroft Exp $	*/
2 
3 /*
4  * Copyright (c) 1989, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. All advertising materials mentioning features or use of this software
16  *    must display the following acknowledgement:
17  *	This product includes software developed by the University of
18  *	California, Berkeley and its contributors.
19  * 4. Neither the name of the University nor the names of its contributors
20  *    may be used to endorse or promote products derived from this software
21  *    without specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33  * SUCH DAMAGE.
34  *
35  *	@(#)kern_ktrace.c	8.5 (Berkeley) 5/14/95
36  */
37 
38 #include "opt_ktrace.h"
39 
40 #ifdef KTRACE
41 
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/proc.h>
45 #include <sys/file.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 #include <sys/filedesc.h>
52 
53 #include <sys/mount.h>
54 #include <sys/syscallargs.h>
55 
56 struct ktr_header *ktrgetheader __P((int));
57 int ktrops __P((struct proc *, struct proc *, int, int, void *));
58 int ktrsetchildren __P((struct proc *, struct proc *, int, int, void *));
59 void ktrwrite __P((struct proc *, void *, struct ktr_header *));
60 int ktrcanset __P((struct proc *, struct proc *));
61 
62 void
63 ktrderef(p)
64 	struct proc *p;
65 {
66 	if (p->p_tracep == NULL)
67 		return;
68 
69 	if (p->p_traceflag & KTRFAC_FD) {
70 		struct file *fp = p->p_tracep;
71 
72 		closef(fp, NULL);
73 	} else {
74 		struct vnode *vp = p->p_tracep;
75 
76 		vrele(vp);
77 	}
78 	p->p_tracep = NULL;
79 	p->p_traceflag = 0;
80 }
81 
82 void
83 ktradref(p)
84 	struct proc *p;
85 {
86 	if (p->p_traceflag & KTRFAC_FD) {
87 		struct file *fp = p->p_tracep;
88 
89 		fp->f_count++;
90 	} else {
91 		struct vnode *vp = p->p_tracep;
92 
93 		VREF(vp);
94 	}
95 }
96 
97 struct ktr_header *
98 ktrgetheader(type)
99 	int type;
100 {
101 	struct ktr_header *kth;
102 	struct proc *p = curproc;	/* XXX */
103 
104 	MALLOC(kth, struct ktr_header *, sizeof(struct ktr_header),
105 		M_TEMP, M_WAITOK);
106 	kth->ktr_type = type;
107 	microtime(&kth->ktr_time);
108 	kth->ktr_pid = p->p_pid;
109 	memcpy(kth->ktr_comm, p->p_comm, MAXCOMLEN);
110 	return (kth);
111 }
112 
113 void
114 ktrsyscall(v, code, argsize, args)
115 	void *v;
116 	register_t code;
117 	size_t argsize;
118 	register_t args[];
119 {
120 	struct	ktr_header *kth;
121 	struct	ktr_syscall *ktp;
122 	struct proc *p = curproc;	/* XXX */
123 	register_t *argp;
124 	int	len = sizeof(struct ktr_syscall) + argsize;
125 	int i;
126 
127 	p->p_traceflag |= KTRFAC_ACTIVE;
128 	kth = ktrgetheader(KTR_SYSCALL);
129 	MALLOC(ktp, struct ktr_syscall *, len, M_TEMP, M_WAITOK);
130 	ktp->ktr_code = code;
131 	ktp->ktr_argsize = argsize;
132 	argp = (register_t *)((char *)ktp + sizeof(struct ktr_syscall));
133 	for (i = 0; i < (argsize / sizeof(*argp)); i++)
134 		*argp++ = args[i];
135 	kth->ktr_buf = (caddr_t)ktp;
136 	kth->ktr_len = len;
137 	ktrwrite(p, v, kth);
138 	FREE(ktp, M_TEMP);
139 	FREE(kth, M_TEMP);
140 	p->p_traceflag &= ~KTRFAC_ACTIVE;
141 }
142 
143 void
144 ktrsysret(v, code, error, retval)
145 	void *v;
146 	register_t code;
147 	int error;
148 	register_t retval;
149 {
150 	struct ktr_header *kth;
151 	struct ktr_sysret ktp;
152 	struct proc *p = curproc;	/* XXX */
153 
154 	p->p_traceflag |= KTRFAC_ACTIVE;
155 	kth = ktrgetheader(KTR_SYSRET);
156 	ktp.ktr_code = code;
157 	ktp.ktr_error = error;
158 	ktp.ktr_retval = retval;		/* what about val2 ? */
159 
160 	kth->ktr_buf = (caddr_t)&ktp;
161 	kth->ktr_len = sizeof(struct ktr_sysret);
162 
163 	ktrwrite(p, v, kth);
164 	FREE(kth, M_TEMP);
165 	p->p_traceflag &= ~KTRFAC_ACTIVE;
166 }
167 
168 void
169 ktrnamei(v, path)
170 	void *v;
171 	char *path;
172 {
173 	struct ktr_header *kth;
174 	struct proc *p = curproc;	/* XXX */
175 
176 	p->p_traceflag |= KTRFAC_ACTIVE;
177 	kth = ktrgetheader(KTR_NAMEI);
178 	kth->ktr_len = strlen(path);
179 	kth->ktr_buf = path;
180 
181 	ktrwrite(p, v, kth);
182 	FREE(kth, M_TEMP);
183 	p->p_traceflag &= ~KTRFAC_ACTIVE;
184 }
185 
186 void
187 ktremul(v, p, emul)
188 	void *v;
189 	struct proc *p;
190 	char *emul;
191 {
192 	struct ktr_header *kth;
193 
194 	p->p_traceflag |= KTRFAC_ACTIVE;
195 	kth = ktrgetheader(KTR_EMUL);
196 	kth->ktr_len = strlen(emul);
197 	kth->ktr_buf = emul;
198 
199 	ktrwrite(p, v, kth);
200 	FREE(kth, M_TEMP);
201 	p->p_traceflag &= ~KTRFAC_ACTIVE;
202 }
203 
204 void
205 ktrgenio(v, fd, rw, iov, len, error)
206 	void *v;
207 	int fd;
208 	enum uio_rw rw;
209 	struct iovec *iov;
210 	int len, error;
211 {
212 	struct ktr_header *kth;
213 	struct ktr_genio *ktp;
214 	caddr_t cp;
215 	int resid = len, cnt;
216 	struct proc *p = curproc;	/* XXX */
217 
218 	if (error)
219 		return;
220 	p->p_traceflag |= KTRFAC_ACTIVE;
221 	kth = ktrgetheader(KTR_GENIO);
222 	MALLOC(ktp, struct ktr_genio *, sizeof(struct ktr_genio) + len,
223 		M_TEMP, M_WAITOK);
224 	ktp->ktr_fd = fd;
225 	ktp->ktr_rw = rw;
226 	cp = (caddr_t)((char *)ktp + sizeof(struct ktr_genio));
227 	while (resid > 0) {
228 		if ((cnt = iov->iov_len) > resid)
229 			cnt = resid;
230 		if (copyin(iov->iov_base, cp, (unsigned)cnt))
231 			goto done;
232 		cp += cnt;
233 		resid -= cnt;
234 		iov++;
235 	}
236 	kth->ktr_buf = (caddr_t)ktp;
237 	kth->ktr_len = sizeof(struct ktr_genio) + len;
238 
239 	ktrwrite(p, v, kth);
240 done:
241 	FREE(kth, M_TEMP);
242 	FREE(ktp, M_TEMP);
243 	p->p_traceflag &= ~KTRFAC_ACTIVE;
244 }
245 
246 void
247 ktrpsig(v, sig, action, mask, code)
248 	void *v;
249 	int sig;
250 	sig_t action;
251 	sigset_t *mask;
252 	int code;
253 {
254 	struct ktr_header *kth;
255 	struct ktr_psig	kp;
256 	struct proc *p = curproc;	/* XXX */
257 
258 	p->p_traceflag |= KTRFAC_ACTIVE;
259 	kth = ktrgetheader(KTR_PSIG);
260 	kp.signo = (char)sig;
261 	kp.action = action;
262 	kp.mask = *mask;
263 	kp.code = code;
264 	kth->ktr_buf = (caddr_t)&kp;
265 	kth->ktr_len = sizeof(struct ktr_psig);
266 
267 	ktrwrite(p, v, kth);
268 	FREE(kth, M_TEMP);
269 	p->p_traceflag &= ~KTRFAC_ACTIVE;
270 }
271 
272 void
273 ktrcsw(v, out, user)
274 	void *v;
275 	int out, user;
276 {
277 	struct ktr_header *kth;
278 	struct	ktr_csw kc;
279 	struct proc *p = curproc;	/* XXX */
280 
281 	p->p_traceflag |= KTRFAC_ACTIVE;
282 	kth = ktrgetheader(KTR_CSW);
283 	kc.out = out;
284 	kc.user = user;
285 	kth->ktr_buf = (caddr_t)&kc;
286 	kth->ktr_len = sizeof(struct ktr_csw);
287 
288 	ktrwrite(p, v, kth);
289 	FREE(kth, M_TEMP);
290 	p->p_traceflag &= ~KTRFAC_ACTIVE;
291 }
292 
293 /* Interface and common routines */
294 
295 /*
296  * ktrace system call
297  */
298 /* ARGSUSED */
299 int
300 sys_fktrace(curp, v, retval)
301 	struct proc *curp;
302 	void *v;
303 	register_t *retval;
304 {
305 	struct sys_fktrace_args /* {
306 		syscallarg(int) fd;
307 		syscallarg(int) ops;
308 		syscallarg(int) facs;
309 		syscallarg(int) pid;
310 	} */ *uap = v;
311 	struct file *fp = NULL;
312 	struct proc *p;
313 	struct filedesc *fdp = curp->p_fd;
314 	struct pgrp *pg;
315 	int facs;
316 	int ops;
317 	int descend;
318 	int ret = 0;
319 	int error = 0;
320 
321 	if (((u_int)SCARG(uap, fd)) >= fdp->fd_nfiles ||
322 	    (fp = fdp->fd_ofiles[SCARG(uap, fd)]) == NULL ||
323 	    (fp->f_flag & FWRITE) == 0)
324 		return (EBADF);
325 
326 	ops = KTROP(SCARG(uap, ops)) | KTRFLAG_FD;
327 	descend = SCARG(uap, ops) & KTRFLAG_DESCEND;
328 	facs = SCARG(uap, facs) & ~((unsigned) KTRFAC_ROOT);
329 	curp->p_traceflag |= KTRFAC_ACTIVE;
330 
331 	/*
332 	 * Clear all uses of the tracefile
333 	 */
334 	if (KTROP(ops) == KTROP_CLEARFILE) {
335 		for (p = allproc.lh_first; p != 0; p = p->p_list.le_next) {
336 			if (p->p_tracep == fp) {
337 				if (ktrcanset(curp, p))
338 					ktrderef(p);
339 				else
340 					error = EPERM;
341 			}
342 		}
343 		goto done;
344 	}
345 	/*
346 	 * need something to (un)trace (XXX - why is this here?)
347 	 */
348 	if (!facs) {
349 		error = EINVAL;
350 		goto done;
351 	}
352 	/*
353 	 * do it
354 	 */
355 	if (SCARG(uap, pid) < 0) {
356 		/*
357 		 * by process group
358 		 */
359 		pg = pgfind(-SCARG(uap, pid));
360 		if (pg == NULL) {
361 			error = ESRCH;
362 			goto done;
363 		}
364 		for (p = pg->pg_members.lh_first; p != 0; p = p->p_pglist.le_next)
365 			if (descend)
366 				ret |= ktrsetchildren(curp, p, ops, facs, fp);
367 			else
368 				ret |= ktrops(curp, p, ops, facs, fp);
369 
370 	} else {
371 		/*
372 		 * by pid
373 		 */
374 		p = pfind(SCARG(uap, pid));
375 		if (p == NULL) {
376 			error = ESRCH;
377 			goto done;
378 		}
379 		if (descend)
380 			ret |= ktrsetchildren(curp, p, ops, facs, fp);
381 		else
382 			ret |= ktrops(curp, p, ops, facs, fp);
383 	}
384 	if (!ret)
385 		error = EPERM;
386 done:
387 	curp->p_traceflag &= ~KTRFAC_ACTIVE;
388 	return (error);
389 }
390 
391 /*
392  * ktrace system call
393  */
394 /* ARGSUSED */
395 int
396 sys_ktrace(curp, v, retval)
397 	struct proc *curp;
398 	void *v;
399 	register_t *retval;
400 {
401 	struct sys_ktrace_args /* {
402 		syscallarg(const char *) fname;
403 		syscallarg(int) ops;
404 		syscallarg(int) facs;
405 		syscallarg(int) pid;
406 	} */ *uap = v;
407 	struct vnode *vp = NULL;
408 	struct proc *p;
409 	struct pgrp *pg;
410 	int facs = SCARG(uap, facs) & ~((unsigned) KTRFAC_ROOT);
411 	int ops = KTROP(SCARG(uap, ops));
412 	int descend = SCARG(uap, ops) & KTRFLAG_DESCEND;
413 	int ret = 0;
414 	int error = 0;
415 	struct nameidata nd;
416 
417 	curp->p_traceflag |= KTRFAC_ACTIVE;
418 	if (ops != KTROP_CLEAR) {
419 		/*
420 		 * an operation which requires a file argument.
421 		 */
422 		NDINIT(&nd, LOOKUP, FOLLOW, UIO_USERSPACE, SCARG(uap, fname),
423 		    curp);
424 		if ((error = vn_open(&nd, FREAD|FWRITE, 0)) != 0) {
425 			curp->p_traceflag &= ~KTRFAC_ACTIVE;
426 			return (error);
427 		}
428 		vp = nd.ni_vp;
429 		VOP_UNLOCK(vp, 0);
430 		if (vp->v_type != VREG) {
431 			(void) vn_close(vp, FREAD|FWRITE, curp->p_ucred, curp);
432 			curp->p_traceflag &= ~KTRFAC_ACTIVE;
433 			return (EACCES);
434 		}
435 	}
436 	/*
437 	 * Clear all uses of the tracefile
438 	 */
439 	if (KTROP(ops) == KTROP_CLEARFILE) {
440 		for (p = allproc.lh_first; p != 0; p = p->p_list.le_next) {
441 			if (p->p_tracep == vp &&
442 			    !ktrops(curp, p, KTROP_CLEAR, ~0, vp))
443 				error = EPERM;
444 		}
445 		goto done;
446 	}
447 	/*
448 	 * need something to (un)trace (XXX - why is this here?)
449 	 */
450 	if (!facs) {
451 		error = EINVAL;
452 		goto done;
453 	}
454 	/*
455 	 * do it
456 	 */
457 	if (SCARG(uap, pid) < 0) {
458 		/*
459 		 * by process group
460 		 */
461 		pg = pgfind(-SCARG(uap, pid));
462 		if (pg == NULL) {
463 			error = ESRCH;
464 			goto done;
465 		}
466 		for (p = pg->pg_members.lh_first; p != 0; p = p->p_pglist.le_next)
467 			if (descend)
468 				ret |= ktrsetchildren(curp, p, ops, facs, vp);
469 			else
470 				ret |= ktrops(curp, p, ops, facs, vp);
471 
472 	} else {
473 		/*
474 		 * by pid
475 		 */
476 		p = pfind(SCARG(uap, pid));
477 		if (p == NULL) {
478 			error = ESRCH;
479 			goto done;
480 		}
481 		if (descend)
482 			ret |= ktrsetchildren(curp, p, ops, facs, vp);
483 		else
484 			ret |= ktrops(curp, p, ops, facs, vp);
485 	}
486 	if (!ret)
487 		error = EPERM;
488 done:
489 	if (vp != NULL)
490 		(void) vn_close(vp, FWRITE, curp->p_ucred, curp);
491 	curp->p_traceflag &= ~KTRFAC_ACTIVE;
492 	return (error);
493 }
494 
495 int
496 ktrops(curp, p, ops, facs, v)
497 	struct proc *p, *curp;
498 	int ops, facs;
499 	void *v;
500 {
501 
502 	if (!ktrcanset(curp, p))
503 		return (0);
504 	if (KTROP(ops) == KTROP_SET) {
505 		if (p->p_tracep != v) {
506 			/*
507 			 * if trace file already in use, relinquish
508 			 */
509 			ktrderef(p);
510 			if (ops & KTRFLAG_FD)
511 				p->p_traceflag = KTRFAC_FD;
512 			p->p_tracep = v;
513 			ktradref(p);
514 		}
515 		p->p_traceflag |= facs;
516 		if (curp->p_ucred->cr_uid == 0)
517 			p->p_traceflag |= KTRFAC_ROOT;
518 	} else {
519 		/* KTROP_CLEAR */
520 		if (((p->p_traceflag &= ~facs) & KTRFAC_MASK) == 0) {
521 			/* no more tracing */
522 			ktrderef(p);
523 		}
524 	}
525 
526 	/*
527 	 * Emit an emulation record, every time there is a ktrace
528 	 * change/attach request.
529 	 */
530 	if (KTRPOINT(p, KTR_EMUL))
531 		ktremul(p->p_tracep, p, p->p_emul->e_name);
532 
533 	return (1);
534 }
535 
536 int
537 ktrsetchildren(curp, top, ops, facs, v)
538 	struct proc *curp, *top;
539 	int ops, facs;
540 	void *v;
541 {
542 	struct proc *p;
543 	int ret = 0;
544 
545 	p = top;
546 	for (;;) {
547 		ret |= ktrops(curp, p, ops, facs, v);
548 		/*
549 		 * If this process has children, descend to them next,
550 		 * otherwise do any siblings, and if done with this level,
551 		 * follow back up the tree (but not past top).
552 		 */
553 		if (p->p_children.lh_first)
554 			p = p->p_children.lh_first;
555 		else for (;;) {
556 			if (p == top)
557 				return (ret);
558 			if (p->p_sibling.le_next) {
559 				p = p->p_sibling.le_next;
560 				break;
561 			}
562 			p = p->p_pptr;
563 		}
564 	}
565 	/*NOTREACHED*/
566 }
567 
568 void
569 ktrwrite(p, v, kth)
570 	struct proc *p;
571 	void *v;
572 	struct ktr_header *kth;
573 {
574 	struct uio auio;
575 	struct iovec aiov[2];
576 	int error;
577 
578 	if (v == NULL)
579 		return;
580 	auio.uio_iov = &aiov[0];
581 	auio.uio_offset = 0;
582 	auio.uio_segflg = UIO_SYSSPACE;
583 	auio.uio_rw = UIO_WRITE;
584 	aiov[0].iov_base = (caddr_t)kth;
585 	aiov[0].iov_len = sizeof(struct ktr_header);
586 	auio.uio_resid = sizeof(struct ktr_header);
587 	auio.uio_iovcnt = 1;
588 	auio.uio_procp = (struct proc *)0;
589 	if (kth->ktr_len > 0) {
590 		auio.uio_iovcnt++;
591 		aiov[1].iov_base = kth->ktr_buf;
592 		aiov[1].iov_len = kth->ktr_len;
593 		auio.uio_resid += kth->ktr_len;
594 	}
595 	if (p->p_traceflag & KTRFAC_FD) {
596 		struct file *fp = v;
597 
598 		error = (*fp->f_ops->fo_write)(fp, &fp->f_offset, &auio,
599 		    fp->f_cred, FOF_UPDATE_OFFSET);
600 	}
601 	else {
602 		struct vnode *vp = v;
603 
604 		vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
605 		error = VOP_WRITE(vp, &auio, IO_UNIT|IO_APPEND, p->p_ucred);
606 		VOP_UNLOCK(vp, 0);
607 	}
608 	if (!error)
609 		return;
610 	/*
611 	 * If error encountered, give up tracing on this vnode.
612 	 */
613 	log(LOG_NOTICE, "ktrace write failed, errno %d, tracing stopped\n",
614 	    error);
615 	for (p = allproc.lh_first; p != 0; p = p->p_list.le_next) {
616 		if (p->p_tracep == v)
617 			ktrderef(p);
618 	}
619 }
620 
621 /*
622  * Return true if caller has permission to set the ktracing state
623  * of target.  Essentially, the target can't possess any
624  * more permissions than the caller.  KTRFAC_ROOT signifies that
625  * root previously set the tracing status on the target process, and
626  * so, only root may further change it.
627  *
628  * TODO: check groups.  use caller effective gid.
629  */
630 int
631 ktrcanset(callp, targetp)
632 	struct proc *callp, *targetp;
633 {
634 	struct pcred *caller = callp->p_cred;
635 	struct pcred *target = targetp->p_cred;
636 
637 	if ((caller->pc_ucred->cr_uid == target->p_ruid &&
638 	     target->p_ruid == target->p_svuid &&
639 	     caller->p_rgid == target->p_rgid &&	/* XXX */
640 	     target->p_rgid == target->p_svgid &&
641 	     (targetp->p_traceflag & KTRFAC_ROOT) == 0) ||
642 	     caller->pc_ucred->cr_uid == 0)
643 		return (1);
644 
645 	return (0);
646 }
647 
648 #endif
649