xref: /netbsd-src/sys/kern/sys_process.c (revision d909946ca08dceb44d7d0f22ec9488679695d976)
1 /*	$NetBSD: sys_process.c,v 1.169 2016/05/25 17:43:58 christos Exp $	*/
2 
3 /*-
4  * Copyright (c) 2008, 2009 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Andrew Doran.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
20  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
21  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
22  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
23  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
24  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
25  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
26  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
27  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
28  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
29  * POSSIBILITY OF SUCH DAMAGE.
30  */
31 
32 /*-
33  * Copyright (c) 1982, 1986, 1989, 1993
34  *	The Regents of the University of California.  All rights reserved.
35  * (c) UNIX System Laboratories, Inc.
36  * All or some portions of this file are derived from material licensed
37  * to the University of California by American Telephone and Telegraph
38  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
39  * the permission of UNIX System Laboratories, Inc.
40  *
41  * This code is derived from software contributed to Berkeley by
42  * Jan-Simon Pendry.
43  *
44  * Redistribution and use in source and binary forms, with or without
45  * modification, are permitted provided that the following conditions
46  * are met:
47  * 1. Redistributions of source code must retain the above copyright
48  *    notice, this list of conditions and the following disclaimer.
49  * 2. Redistributions in binary form must reproduce the above copyright
50  *    notice, this list of conditions and the following disclaimer in the
51  *    documentation and/or other materials provided with the distribution.
52  * 3. Neither the name of the University nor the names of its contributors
53  *    may be used to endorse or promote products derived from this software
54  *    without specific prior written permission.
55  *
56  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
57  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
58  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
59  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
60  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
61  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
62  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
63  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
64  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
65  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
66  * SUCH DAMAGE.
67  *
68  *	from: @(#)sys_process.c	8.1 (Berkeley) 6/10/93
69  */
70 
71 /*-
72  * Copyright (c) 1993 Jan-Simon Pendry.
73  * Copyright (c) 1994 Christopher G. Demetriou.  All rights reserved.
74  *
75  * This code is derived from software contributed to Berkeley by
76  * Jan-Simon Pendry.
77  *
78  * Redistribution and use in source and binary forms, with or without
79  * modification, are permitted provided that the following conditions
80  * are met:
81  * 1. Redistributions of source code must retain the above copyright
82  *    notice, this list of conditions and the following disclaimer.
83  * 2. Redistributions in binary form must reproduce the above copyright
84  *    notice, this list of conditions and the following disclaimer in the
85  *    documentation and/or other materials provided with the distribution.
86  * 3. All advertising materials mentioning features or use of this software
87  *    must display the following acknowledgement:
88  *	This product includes software developed by the University of
89  *	California, Berkeley and its contributors.
90  * 4. Neither the name of the University nor the names of its contributors
91  *    may be used to endorse or promote products derived from this software
92  *    without specific prior written permission.
93  *
94  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
95  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
96  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
97  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
98  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
99  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
100  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
101  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
102  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
103  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
104  * SUCH DAMAGE.
105  *
106  *	from: @(#)sys_process.c	8.1 (Berkeley) 6/10/93
107  */
108 
109 /*
110  * References:
111  *	(1) Bach's "The Design of the UNIX Operating System",
112  *	(2) sys/miscfs/procfs from UCB's 4.4BSD-Lite distribution,
113  *	(3) the "4.4BSD Programmer's Reference Manual" published
114  *		by USENIX and O'Reilly & Associates.
115  * The 4.4BSD PRM does a reasonably good job of documenting what the various
116  * ptrace() requests should actually do, and its text is quoted several times
117  * in this file.
118  */
119 
120 #include <sys/cdefs.h>
121 __KERNEL_RCSID(0, "$NetBSD: sys_process.c,v 1.169 2016/05/25 17:43:58 christos Exp $");
122 
123 #include "opt_ptrace.h"
124 #include "opt_ktrace.h"
125 #include "opt_pax.h"
126 
127 #include <sys/param.h>
128 #include <sys/systm.h>
129 #include <sys/proc.h>
130 #include <sys/errno.h>
131 #include <sys/exec.h>
132 #include <sys/pax.h>
133 #include <sys/ptrace.h>
134 #include <sys/uio.h>
135 #include <sys/ras.h>
136 #include <sys/kmem.h>
137 #include <sys/kauth.h>
138 #include <sys/mount.h>
139 #include <sys/syscallargs.h>
140 
141 #include <uvm/uvm_extern.h>
142 
143 #include <machine/reg.h>
144 
145 #ifdef PTRACE
146 
147 # ifdef DEBUG
148 #  define DPRINTF(a) uprintf a
149 # else
150 #  define DPRINTF(a)
151 # endif
152 
153 static kauth_listener_t ptrace_listener;
154 #ifdef PTRACE
155 static int process_auxv_offset(struct proc *, struct uio *);
156 #endif
157 
158 static int
159 ptrace_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
160     void *arg0, void *arg1, void *arg2, void *arg3)
161 {
162 	struct proc *p;
163 	int result;
164 
165 	result = KAUTH_RESULT_DEFER;
166 	p = arg0;
167 
168 	if (action != KAUTH_PROCESS_PTRACE)
169 		return result;
170 
171 	switch ((u_long)arg1) {
172 	case PT_TRACE_ME:
173 	case PT_ATTACH:
174 	case PT_WRITE_I:
175 	case PT_WRITE_D:
176 	case PT_READ_I:
177 	case PT_READ_D:
178 	case PT_IO:
179 #ifdef PT_GETREGS
180 	case PT_GETREGS:
181 #endif
182 #ifdef PT_SETREGS
183 	case PT_SETREGS:
184 #endif
185 #ifdef PT_GETFPREGS
186 	case PT_GETFPREGS:
187 #endif
188 #ifdef PT_SETFPREGS
189 	case PT_SETFPREGS:
190 #endif
191 	case PT_SET_EVENT_MASK:
192 	case PT_GET_EVENT_MASK:
193 	case PT_GET_PROCESS_STATE:
194 #ifdef __HAVE_PTRACE_MACHDEP
195 	PTRACE_MACHDEP_REQUEST_CASES
196 #endif
197 		if (kauth_cred_getuid(cred) != kauth_cred_getuid(p->p_cred) ||
198 		    ISSET(p->p_flag, PK_SUGID)) {
199 			break;
200 		}
201 
202 		result = KAUTH_RESULT_ALLOW;
203 
204 	break;
205 
206 #ifdef PT_STEP
207 	case PT_STEP:
208 #endif
209 	case PT_CONTINUE:
210 	case PT_KILL:
211 	case PT_DETACH:
212 	case PT_LWPINFO:
213 	case PT_SYSCALL:
214 	case PT_SYSCALLEMU:
215 	case PT_DUMPCORE:
216 		result = KAUTH_RESULT_ALLOW;
217 		break;
218 
219 	default:
220 		break;
221 	}
222 
223 	return result;
224 }
225 
226 void
227 ptrace_init(void)
228 {
229 
230 	ptrace_listener = kauth_listen_scope(KAUTH_SCOPE_PROCESS,
231 	    ptrace_listener_cb, NULL);
232 }
233 
234 /*
235  * Process debugging system call.
236  */
237 int
238 sys_ptrace(struct lwp *l, const struct sys_ptrace_args *uap, register_t *retval)
239 {
240 	/* {
241 		syscallarg(int) req;
242 		syscallarg(pid_t) pid;
243 		syscallarg(void *) addr;
244 		syscallarg(int) data;
245 	} */
246 	struct proc *p = l->l_proc;
247 	struct lwp *lt;
248 #ifdef PT_STEP
249 	struct lwp *lt2;
250 #endif
251 	struct proc *t;				/* target process */
252 	struct uio uio;
253 	struct iovec iov;
254 	struct ptrace_io_desc piod;
255 	struct ptrace_event pe;
256 	struct ptrace_state ps;
257 	struct ptrace_lwpinfo pl;
258 	struct vmspace *vm;
259 	int error, write, tmp, req, pheld;
260 	int signo = 0;
261 	int resume_all;
262 	ksiginfo_t ksi;
263 	char *path;
264 	int len = 0;
265 
266 	error = 0;
267 	req = SCARG(uap, req);
268 
269 	/*
270 	 * If attaching or detaching, we need to get a write hold on the
271 	 * proclist lock so that we can re-parent the target process.
272 	 */
273 	mutex_enter(proc_lock);
274 
275 	/* "A foolish consistency..." XXX */
276 	if (req == PT_TRACE_ME) {
277 		t = p;
278 		mutex_enter(t->p_lock);
279 	} else {
280 		/* Find the process we're supposed to be operating on. */
281 		t = proc_find(SCARG(uap, pid));
282 		if (t == NULL) {
283 			mutex_exit(proc_lock);
284 			return (ESRCH);
285 		}
286 
287 		/* XXX-elad */
288 		mutex_enter(t->p_lock);
289 		error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE,
290 		    t, KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL);
291 		if (error) {
292 			mutex_exit(proc_lock);
293 			mutex_exit(t->p_lock);
294 			return (ESRCH);
295 		}
296 	}
297 
298 	/*
299 	 * Grab a reference on the process to prevent it from execing or
300 	 * exiting.
301 	 */
302 	if (!rw_tryenter(&t->p_reflock, RW_READER)) {
303 		mutex_exit(proc_lock);
304 		mutex_exit(t->p_lock);
305 		return EBUSY;
306 	}
307 
308 	/* Make sure we can operate on it. */
309 	switch (req) {
310 	case  PT_TRACE_ME:
311 		/* Saying that you're being traced is always legal. */
312 		break;
313 
314 	case  PT_ATTACH:
315 		/*
316 		 * You can't attach to a process if:
317 		 *	(1) it's the process that's doing the attaching,
318 		 */
319 		if (t->p_pid == p->p_pid) {
320 			error = EINVAL;
321 			break;
322 		}
323 
324 		/*
325 		 *  (2) it's a system process
326 		 */
327 		if (t->p_flag & PK_SYSTEM) {
328 			error = EPERM;
329 			break;
330 		}
331 
332 		/*
333 		 *	(3) it's already being traced, or
334 		 */
335 		if (ISSET(t->p_slflag, PSL_TRACED)) {
336 			error = EBUSY;
337 			break;
338 		}
339 
340 		/*
341 		 * 	(4) the tracer is chrooted, and its root directory is
342 		 * 	    not at or above the root directory of the tracee
343 		 */
344 		mutex_exit(t->p_lock);	/* XXXSMP */
345 		tmp = proc_isunder(t, l);
346 		mutex_enter(t->p_lock);	/* XXXSMP */
347 		if (!tmp) {
348 			error = EPERM;
349 			break;
350 		}
351 		break;
352 
353 	case  PT_READ_I:
354 	case  PT_READ_D:
355 	case  PT_WRITE_I:
356 	case  PT_WRITE_D:
357 	case  PT_IO:
358 #ifdef PT_GETREGS
359 	case  PT_GETREGS:
360 #endif
361 #ifdef PT_SETREGS
362 	case  PT_SETREGS:
363 #endif
364 #ifdef PT_GETFPREGS
365 	case  PT_GETFPREGS:
366 #endif
367 #ifdef PT_SETFPREGS
368 	case  PT_SETFPREGS:
369 #endif
370 #ifdef __HAVE_PTRACE_MACHDEP
371 	PTRACE_MACHDEP_REQUEST_CASES
372 #endif
373 		/*
374 		 * You can't read/write the memory or registers of a process
375 		 * if the tracer is chrooted, and its root directory is not at
376 		 * or above the root directory of the tracee.
377 		 */
378 		mutex_exit(t->p_lock);	/* XXXSMP */
379 		tmp = proc_isunder(t, l);
380 		mutex_enter(t->p_lock);	/* XXXSMP */
381 		if (!tmp) {
382 			error = EPERM;
383 			break;
384 		}
385 		/*FALLTHROUGH*/
386 
387 	case  PT_CONTINUE:
388 	case  PT_KILL:
389 	case  PT_DETACH:
390 	case  PT_LWPINFO:
391 	case  PT_SYSCALL:
392 	case  PT_SYSCALLEMU:
393 	case  PT_DUMPCORE:
394 #ifdef PT_STEP
395 	case  PT_STEP:
396 #endif
397 	case  PT_SET_EVENT_MASK:
398 	case  PT_GET_EVENT_MASK:
399 	case  PT_GET_PROCESS_STATE:
400 		/*
401 		 * You can't do what you want to the process if:
402 		 *	(1) It's not being traced at all,
403 		 */
404 		if (!ISSET(t->p_slflag, PSL_TRACED)) {
405 			error = EPERM;
406 			break;
407 		}
408 
409 		/*
410 		 *	(2) it's being traced by procfs (which has
411 		 *	    different signal delivery semantics),
412 		 */
413 		if (ISSET(t->p_slflag, PSL_FSTRACE)) {
414 			DPRINTF(("file system traced\n"));
415 			error = EBUSY;
416 			break;
417 		}
418 
419 		/*
420 		 *	(3) it's not being traced by _you_, or
421 		 */
422 		if (t->p_pptr != p) {
423 			DPRINTF(("parent %d != %d\n", t->p_pptr->p_pid,
424 			    p->p_pid));
425 			error = EBUSY;
426 			break;
427 		}
428 
429 		/*
430 		 *	(4) it's not currently stopped.
431 		 */
432 		if (t->p_stat != SSTOP || !t->p_waited /* XXXSMP */) {
433 			DPRINTF(("stat %d flag %d\n", t->p_stat,
434 			    !t->p_waited));
435 			error = EBUSY;
436 			break;
437 		}
438 		break;
439 
440 	default:			/* It was not a legal request. */
441 		error = EINVAL;
442 		break;
443 	}
444 
445 	if (error == 0) {
446 		error = kauth_authorize_process(l->l_cred,
447 		    KAUTH_PROCESS_PTRACE, t, KAUTH_ARG(req),
448 		    NULL, NULL);
449 	}
450 	if (error == 0) {
451 		lt = lwp_find_first(t);
452 		if (lt == NULL)
453 			error = ESRCH;
454 	}
455 
456 	if (error != 0) {
457 		mutex_exit(proc_lock);
458 		mutex_exit(t->p_lock);
459 		rw_exit(&t->p_reflock);
460 		return error;
461 	}
462 
463 	/* Do single-step fixup if needed. */
464 	FIX_SSTEP(t);
465 	KASSERT(lt != NULL);
466 	lwp_addref(lt);
467 
468 	/*
469 	 * Which locks do we need held? XXX Ugly.
470 	 */
471 	switch (req) {
472 #ifdef PT_STEP
473 	case PT_STEP:
474 #endif
475 	case PT_CONTINUE:
476 	case PT_DETACH:
477 	case PT_KILL:
478 	case PT_SYSCALL:
479 	case PT_SYSCALLEMU:
480 	case PT_ATTACH:
481 	case PT_TRACE_ME:
482 		pheld = 1;
483 		break;
484 	default:
485 		mutex_exit(proc_lock);
486 		mutex_exit(t->p_lock);
487 		pheld = 0;
488 		break;
489 	}
490 
491 	/* Now do the operation. */
492 	write = 0;
493 	*retval = 0;
494 	tmp = 0;
495 	resume_all = 1;
496 
497 	switch (req) {
498 	case  PT_TRACE_ME:
499 		/* Just set the trace flag. */
500 		SET(t->p_slflag, PSL_TRACED);
501 		t->p_opptr = t->p_pptr;
502 		break;
503 
504 	case  PT_WRITE_I:		/* XXX no separate I and D spaces */
505 	case  PT_WRITE_D:
506 #if defined(__HAVE_RAS)
507 		/*
508 		 * Can't write to a RAS
509 		 */
510 		if (ras_lookup(t, SCARG(uap, addr)) != (void *)-1) {
511 			error = EACCES;
512 			break;
513 		}
514 #endif
515 		write = 1;
516 		tmp = SCARG(uap, data);
517 		/* FALLTHROUGH */
518 
519 	case  PT_READ_I:		/* XXX no separate I and D spaces */
520 	case  PT_READ_D:
521 		/* write = 0 done above. */
522 		iov.iov_base = (void *)&tmp;
523 		iov.iov_len = sizeof(tmp);
524 		uio.uio_iov = &iov;
525 		uio.uio_iovcnt = 1;
526 		uio.uio_offset = (off_t)(unsigned long)SCARG(uap, addr);
527 		uio.uio_resid = sizeof(tmp);
528 		uio.uio_rw = write ? UIO_WRITE : UIO_READ;
529 		UIO_SETUP_SYSSPACE(&uio);
530 
531 		error = process_domem(l, lt, &uio);
532 		if (!write)
533 			*retval = tmp;
534 		break;
535 
536 	case  PT_IO:
537 		error = copyin(SCARG(uap, addr), &piod, sizeof(piod));
538 		if (error)
539 			break;
540 
541 		iov.iov_base = piod.piod_addr;
542 		iov.iov_len = piod.piod_len;
543 		uio.uio_iov = &iov;
544 		uio.uio_iovcnt = 1;
545 		uio.uio_offset = (off_t)(unsigned long)piod.piod_offs;
546 		uio.uio_resid = piod.piod_len;
547 
548 		switch (piod.piod_op) {
549 		case PIOD_READ_D:
550 		case PIOD_READ_I:
551 			uio.uio_rw = UIO_READ;
552 			break;
553 		case PIOD_WRITE_D:
554 		case PIOD_WRITE_I:
555 			/*
556 			 * Can't write to a RAS
557 			 */
558 			if (ras_lookup(t, SCARG(uap, addr)) != (void *)-1) {
559 				return (EACCES);
560 			}
561 			uio.uio_rw = UIO_WRITE;
562 			break;
563 		case PIOD_READ_AUXV:
564 			req = PT_READ_D;
565 			uio.uio_rw = UIO_READ;
566 			tmp = t->p_execsw->es_arglen * sizeof(char *);
567 			if (uio.uio_offset > tmp)
568 				return EIO;
569 			if (uio.uio_resid > tmp - uio.uio_offset)
570 				uio.uio_resid = tmp - uio.uio_offset;
571 			piod.piod_len = iov.iov_len = uio.uio_resid;
572 			error = process_auxv_offset(t, &uio);
573 			if (error)
574 				return error;
575 			break;
576 		default:
577 			error = EINVAL;
578 			break;
579 		}
580 		if (error)
581 			break;
582 		error = proc_vmspace_getref(l->l_proc, &vm);
583 		if (error)
584 			break;
585 		uio.uio_vmspace = vm;
586 
587 		error = process_domem(l, lt, &uio);
588 		piod.piod_len -= uio.uio_resid;
589 		(void) copyout(&piod, SCARG(uap, addr), sizeof(piod));
590 		uvmspace_free(vm);
591 		break;
592 
593 	case  PT_DUMPCORE:
594 		if ((path = SCARG(uap, addr)) != NULL) {
595 			char *dst;
596 			len = SCARG(uap, data);
597 
598 			if (len < 0 || len >= MAXPATHLEN) {
599 				error = EINVAL;
600 				break;
601 			}
602 			dst = kmem_alloc(len + 1, KM_SLEEP);
603 			if ((error = copyin(path, dst, len)) != 0) {
604 				kmem_free(dst, len + 1);
605 				break;
606 			}
607 			path = dst;
608 			path[len] = '\0';
609 		}
610 		error = (*coredump_vec)(lt, path);
611 		if (path)
612 			kmem_free(path, len + 1);
613 		break;
614 
615 #ifdef PT_STEP
616 	case  PT_STEP:
617 		/*
618 		 * From the 4.4BSD PRM:
619 		 * "Execution continues as in request PT_CONTINUE; however
620 		 * as soon as possible after execution of at least one
621 		 * instruction, execution stops again. [ ... ]"
622 		 */
623 #endif
624 	case  PT_CONTINUE:
625 	case  PT_SYSCALL:
626 	case  PT_DETACH:
627 		if (req == PT_SYSCALL) {
628 			if (!ISSET(t->p_slflag, PSL_SYSCALL)) {
629 				SET(t->p_slflag, PSL_SYSCALL);
630 #ifdef __HAVE_SYSCALL_INTERN
631 				(*t->p_emul->e_syscall_intern)(t);
632 #endif
633 			}
634 		} else {
635 			if (ISSET(t->p_slflag, PSL_SYSCALL)) {
636 				CLR(t->p_slflag, PSL_SYSCALL);
637 #ifdef __HAVE_SYSCALL_INTERN
638 				(*t->p_emul->e_syscall_intern)(t);
639 #endif
640 			}
641 		}
642 		t->p_trace_enabled = trace_is_enabled(t);
643 
644 		/*
645 		 * Pick up the LWPID, if supplied.  There are two cases:
646 		 * data < 0 : step or continue single thread, lwp = -data
647 		 * data > 0 in PT_STEP : step this thread, continue others
648 		 * For operations other than PT_STEP, data > 0 means
649 		 * data is the signo to deliver to the process.
650 		 */
651 		tmp = SCARG(uap, data);
652 		if (tmp >= 0) {
653 #ifdef PT_STEP
654 			if (req == PT_STEP)
655 				signo = 0;
656 			else
657 #endif
658 			{
659 				signo = tmp;
660 				tmp = 0;	/* don't search for LWP */
661 			}
662 		}
663 		else
664 			tmp = -tmp;
665 
666 		if (tmp > 0) {
667 			if (req == PT_DETACH) {
668 				error = EINVAL;
669 				break;
670 			}
671 			lwp_delref2 (lt);
672 			lt = lwp_find(t, tmp);
673 			if (lt == NULL) {
674 				error = ESRCH;
675 				break;
676 			}
677 			lwp_addref(lt);
678 			resume_all = 0;
679 			signo = 0;
680 		}
681 
682 		/*
683 		 * From the 4.4BSD PRM:
684 		 * "The data argument is taken as a signal number and the
685 		 * child's execution continues at location addr as if it
686 		 * incurred that signal.  Normally the signal number will
687 		 * be either 0 to indicate that the signal that caused the
688 		 * stop should be ignored, or that value fetched out of
689 		 * the process's image indicating which signal caused
690 		 * the stop.  If addr is (int *)1 then execution continues
691 		 * from where it stopped."
692 		 */
693 
694 		/* Check that the data is a valid signal number or zero. */
695 		if (signo < 0 || signo >= NSIG) {
696 			error = EINVAL;
697 			break;
698 		}
699 
700 		/* If the address parameter is not (int *)1, set the pc. */
701 		if ((int *)SCARG(uap, addr) != (int *)1) {
702 			error = process_set_pc(lt, SCARG(uap, addr));
703 			if (error != 0)
704 				break;
705 		}
706 #ifdef PT_STEP
707 		/*
708 		 * Arrange for a single-step, if that's requested and possible.
709 		 * More precisely, set the single step status as requested for
710 		 * the requested thread, and clear it for other threads.
711 		 */
712 		LIST_FOREACH(lt2, &t->p_lwps, l_sibling) {
713 			if (lt != lt2)
714 			{
715 				lwp_lock(lt2);
716 				process_sstep(lt2, 0);
717 				lwp_unlock(lt2);
718 			}
719 		}
720 		error = process_sstep(lt, req == PT_STEP);
721 		if (error)
722 			break;
723 #endif
724 		if (req == PT_DETACH) {
725 			CLR(t->p_slflag, PSL_TRACED|PSL_FSTRACE|PSL_SYSCALL);
726 
727 			/* give process back to original parent or init */
728 			if (t->p_opptr != t->p_pptr) {
729 				struct proc *pp = t->p_opptr;
730 				proc_reparent(t, pp ? pp : initproc);
731 			}
732 
733 			/* not being traced any more */
734 			t->p_opptr = NULL;
735 		}
736 	sendsig:
737 		t->p_fpid = 0;
738 		/* Finally, deliver the requested signal (or none). */
739 		if (t->p_stat == SSTOP) {
740 			/*
741 			 * Unstop the process.  If it needs to take a
742 			 * signal, make all efforts to ensure that at
743 			 * an LWP runs to see it.
744 			 */
745 			t->p_xsig = signo;
746 			if (resume_all)
747 				proc_unstop(t);
748 			else
749 				lwp_unstop(lt);
750 		} else if (signo != 0) {
751 			KSI_INIT_EMPTY(&ksi);
752 			ksi.ksi_signo = signo;
753 			kpsignal2(t, &ksi);
754 		}
755 		break;
756 
757 	case  PT_SYSCALLEMU:
758 		if (!ISSET(t->p_slflag, PSL_SYSCALL) || t->p_stat != SSTOP) {
759 			error = EINVAL;
760 			break;
761 		}
762 		SET(t->p_slflag, PSL_SYSCALLEMU);
763 		break;
764 
765 	case  PT_KILL:
766 		/* just send the process a KILL signal. */
767 		signo = SIGKILL;
768 		goto sendsig;	/* in PT_CONTINUE, above. */
769 
770 	case  PT_ATTACH:
771 		/*
772 		 * Go ahead and set the trace flag.
773 		 * Save the old parent (it's reset in
774 		 *   _DETACH, and also in kern_exit.c:wait4()
775 		 * Reparent the process so that the tracing
776 		 *   proc gets to see all the action.
777 		 * Stop the target.
778 		 */
779 		t->p_opptr = t->p_pptr;
780 		if (t->p_pptr != p) {
781 			struct proc *parent = t->p_pptr;
782 
783 			if (parent->p_lock < t->p_lock) {
784 				if (!mutex_tryenter(parent->p_lock)) {
785 					mutex_exit(t->p_lock);
786 					mutex_enter(parent->p_lock);
787 					mutex_enter(t->p_lock);
788 				}
789 			} else if (parent->p_lock > t->p_lock) {
790 				mutex_enter(parent->p_lock);
791 			}
792 			parent->p_slflag |= PSL_CHTRACED;
793 			proc_reparent(t, p);
794 			if (parent->p_lock != t->p_lock)
795 				mutex_exit(parent->p_lock);
796 		}
797 		SET(t->p_slflag, PSL_TRACED);
798 		signo = SIGSTOP;
799 		goto sendsig;
800 
801 	case  PT_GET_EVENT_MASK:
802 		if (SCARG(uap, data) != sizeof(pe)) {
803 			DPRINTF(("ptrace(%d): %d != %zu\n", req,
804 			    SCARG(uap, data), sizeof(pe)));
805 			error = EINVAL;
806 			break;
807 		}
808 		memset(&pe, 0, sizeof(pe));
809 		pe.pe_set_event = ISSET(t->p_slflag, PSL_TRACEFORK) ?
810 		    PTRACE_FORK : 0;
811 		error = copyout(&pe, SCARG(uap, addr), sizeof(pe));
812 		break;
813 
814 	case  PT_SET_EVENT_MASK:
815 		if (SCARG(uap, data) != sizeof(pe)) {
816 			DPRINTF(("ptrace(%d): %d != %zu\n", req,
817 			    SCARG(uap, data), sizeof(pe)));
818 			error = EINVAL;
819 			break;
820 		}
821 		if ((error = copyin(SCARG(uap, addr), &pe, sizeof(pe))) != 0)
822 			return error;
823 		if (pe.pe_set_event & PTRACE_FORK)
824 			SET(t->p_slflag, PSL_TRACEFORK);
825 		else
826 			CLR(t->p_slflag, PSL_TRACEFORK);
827 		break;
828 
829 	case  PT_GET_PROCESS_STATE:
830 		if (SCARG(uap, data) != sizeof(ps)) {
831 			DPRINTF(("ptrace(%d): %d != %zu\n", req,
832 			    SCARG(uap, data), sizeof(ps)));
833 			error = EINVAL;
834 			break;
835 		}
836 		memset(&ps, 0, sizeof(ps));
837 		if (t->p_fpid) {
838 			ps.pe_report_event = PTRACE_FORK;
839 			ps.pe_other_pid = t->p_fpid;
840 		}
841 		error = copyout(&ps, SCARG(uap, addr), sizeof(ps));
842 		break;
843 
844 	case PT_LWPINFO:
845 		if (SCARG(uap, data) != sizeof(pl)) {
846 			DPRINTF(("ptrace(%d): %d != %zu\n", req,
847 			    SCARG(uap, data), sizeof(pl)));
848 			error = EINVAL;
849 			break;
850 		}
851 		error = copyin(SCARG(uap, addr), &pl, sizeof(pl));
852 		if (error)
853 			break;
854 		tmp = pl.pl_lwpid;
855 		lwp_delref(lt);
856 		mutex_enter(t->p_lock);
857 		if (tmp == 0)
858 			lt = lwp_find_first(t);
859 		else {
860 			lt = lwp_find(t, tmp);
861 			if (lt == NULL) {
862 				mutex_exit(t->p_lock);
863 				error = ESRCH;
864 				break;
865 			}
866 			lt = LIST_NEXT(lt, l_sibling);
867 		}
868 		while (lt != NULL && !lwp_alive(lt))
869 			lt = LIST_NEXT(lt, l_sibling);
870 		pl.pl_lwpid = 0;
871 		pl.pl_event = 0;
872 		if (lt) {
873 			lwp_addref(lt);
874 			pl.pl_lwpid = lt->l_lid;
875 			if (lt->l_lid == t->p_sigctx.ps_lwp)
876 				pl.pl_event = PL_EVENT_SIGNAL;
877 		}
878 		mutex_exit(t->p_lock);
879 
880 		error = copyout(&pl, SCARG(uap, addr), sizeof(pl));
881 		break;
882 
883 #ifdef PT_SETREGS
884 	case  PT_SETREGS:
885 		write = 1;
886 #endif
887 #ifdef PT_GETREGS
888 	case  PT_GETREGS:
889 		/* write = 0 done above. */
890 #endif
891 #if defined(PT_SETREGS) || defined(PT_GETREGS)
892 		tmp = SCARG(uap, data);
893 		if (tmp != 0 && t->p_nlwps > 1) {
894 			lwp_delref(lt);
895 			mutex_enter(t->p_lock);
896 			lt = lwp_find(t, tmp);
897 			if (lt == NULL) {
898 				mutex_exit(t->p_lock);
899 				error = ESRCH;
900 				break;
901 			}
902 			lwp_addref(lt);
903 			mutex_exit(t->p_lock);
904 		}
905 		if (!process_validregs(lt))
906 			error = EINVAL;
907 		else {
908 			error = proc_vmspace_getref(l->l_proc, &vm);
909 			if (error)
910 				break;
911 			iov.iov_base = SCARG(uap, addr);
912 			iov.iov_len = sizeof(struct reg);
913 			uio.uio_iov = &iov;
914 			uio.uio_iovcnt = 1;
915 			uio.uio_offset = 0;
916 			uio.uio_resid = sizeof(struct reg);
917 			uio.uio_rw = write ? UIO_WRITE : UIO_READ;
918 			uio.uio_vmspace = vm;
919 
920 			error = process_doregs(l, lt, &uio);
921 			uvmspace_free(vm);
922 		}
923 		break;
924 #endif
925 
926 #ifdef PT_SETFPREGS
927 	case  PT_SETFPREGS:
928 		write = 1;
929 #endif
930 #ifdef PT_GETFPREGS
931 	case  PT_GETFPREGS:
932 		/* write = 0 done above. */
933 #endif
934 #if defined(PT_SETFPREGS) || defined(PT_GETFPREGS)
935 		tmp = SCARG(uap, data);
936 		if (tmp != 0 && t->p_nlwps > 1) {
937 			lwp_delref(lt);
938 			mutex_enter(t->p_lock);
939 			lt = lwp_find(t, tmp);
940 			if (lt == NULL) {
941 				mutex_exit(t->p_lock);
942 				error = ESRCH;
943 				break;
944 			}
945 			lwp_addref(lt);
946 			mutex_exit(t->p_lock);
947 		}
948 		if (!process_validfpregs(lt))
949 			error = EINVAL;
950 		else {
951 			error = proc_vmspace_getref(l->l_proc, &vm);
952 			if (error)
953 				break;
954 			iov.iov_base = SCARG(uap, addr);
955 			iov.iov_len = sizeof(struct fpreg);
956 			uio.uio_iov = &iov;
957 			uio.uio_iovcnt = 1;
958 			uio.uio_offset = 0;
959 			uio.uio_resid = sizeof(struct fpreg);
960 			uio.uio_rw = write ? UIO_WRITE : UIO_READ;
961 			uio.uio_vmspace = vm;
962 
963 			error = process_dofpregs(l, lt, &uio);
964 			uvmspace_free(vm);
965 		}
966 		break;
967 #endif
968 
969 #ifdef __HAVE_PTRACE_MACHDEP
970 	PTRACE_MACHDEP_REQUEST_CASES
971 		error = ptrace_machdep_dorequest(l, lt,
972 		    req, SCARG(uap, addr), SCARG(uap, data));
973 		break;
974 #endif
975 	}
976 
977 	if (pheld) {
978 		mutex_exit(t->p_lock);
979 		mutex_exit(proc_lock);
980 	}
981 	if (lt != NULL)
982 		lwp_delref(lt);
983 	rw_exit(&t->p_reflock);
984 
985 	return error;
986 }
987 
988 int
989 process_doregs(struct lwp *curl /*tracer*/,
990     struct lwp *l /*traced*/,
991     struct uio *uio)
992 {
993 #if defined(PT_GETREGS) || defined(PT_SETREGS)
994 	int error;
995 	struct reg r;
996 	char *kv;
997 	int kl;
998 
999 	if (uio->uio_offset < 0 || uio->uio_offset > (off_t)sizeof(r))
1000 		return EINVAL;
1001 
1002 	kl = sizeof(r);
1003 	kv = (char *)&r;
1004 
1005 	kv += uio->uio_offset;
1006 	kl -= uio->uio_offset;
1007 	if ((size_t)kl > uio->uio_resid)
1008 		kl = uio->uio_resid;
1009 
1010 	error = process_read_regs(l, &r);
1011 	if (error == 0)
1012 		error = uiomove(kv, kl, uio);
1013 	if (error == 0 && uio->uio_rw == UIO_WRITE) {
1014 		if (l->l_stat != LSSTOP)
1015 			error = EBUSY;
1016 		else
1017 			error = process_write_regs(l, &r);
1018 	}
1019 
1020 	uio->uio_offset = 0;
1021 	return (error);
1022 #else
1023 	return (EINVAL);
1024 #endif
1025 }
1026 
1027 int
1028 process_validregs(struct lwp *l)
1029 {
1030 
1031 #if defined(PT_SETREGS) || defined(PT_GETREGS)
1032 	return ((l->l_flag & LW_SYSTEM) == 0);
1033 #else
1034 	return (0);
1035 #endif
1036 }
1037 
1038 int
1039 process_dofpregs(struct lwp *curl /*tracer*/,
1040     struct lwp *l /*traced*/,
1041     struct uio *uio)
1042 {
1043 #if defined(PT_GETFPREGS) || defined(PT_SETFPREGS)
1044 	int error;
1045 	struct fpreg r;
1046 	char *kv;
1047 	size_t kl;
1048 
1049 	if (uio->uio_offset < 0 || uio->uio_offset > (off_t)sizeof(r))
1050 		return EINVAL;
1051 
1052 	kl = sizeof(r);
1053 	kv = (char *)&r;
1054 
1055 	kv += uio->uio_offset;
1056 	kl -= uio->uio_offset;
1057 	if (kl > uio->uio_resid)
1058 		kl = uio->uio_resid;
1059 
1060 	error = process_read_fpregs(l, &r, &kl);
1061 	if (error == 0)
1062 		error = uiomove(kv, kl, uio);
1063 	if (error == 0 && uio->uio_rw == UIO_WRITE) {
1064 		if (l->l_stat != LSSTOP)
1065 			error = EBUSY;
1066 		else
1067 			error = process_write_fpregs(l, &r, kl);
1068 	}
1069 	uio->uio_offset = 0;
1070 	return (error);
1071 #else
1072 	return (EINVAL);
1073 #endif
1074 }
1075 
1076 int
1077 process_validfpregs(struct lwp *l)
1078 {
1079 
1080 #if defined(PT_SETFPREGS) || defined(PT_GETFPREGS)
1081 	return ((l->l_flag & LW_SYSTEM) == 0);
1082 #else
1083 	return (0);
1084 #endif
1085 }
1086 #endif /* PTRACE */
1087 
1088 #if defined(KTRACE) || defined(PTRACE)
1089 int
1090 process_domem(struct lwp *curl /*tracer*/,
1091     struct lwp *l /*traced*/,
1092     struct uio *uio)
1093 {
1094 	struct proc *p = l->l_proc;	/* traced */
1095 	struct vmspace *vm;
1096 	int error;
1097 
1098 	size_t len;
1099 #ifdef PMAP_NEED_PROCWR
1100 	vaddr_t	addr;
1101 #endif
1102 
1103 	error = 0;
1104 	len = uio->uio_resid;
1105 
1106 	if (len == 0)
1107 		return (0);
1108 
1109 #ifdef PMAP_NEED_PROCWR
1110 	addr = uio->uio_offset;
1111 #endif
1112 
1113 	vm = p->p_vmspace;
1114 
1115 	mutex_enter(&vm->vm_map.misc_lock);
1116 	if ((l->l_flag & LW_WEXIT) || vm->vm_refcnt < 1)
1117 		error = EFAULT;
1118 	if (error == 0)
1119 		p->p_vmspace->vm_refcnt++;  /* XXX */
1120 	mutex_exit(&vm->vm_map.misc_lock);
1121 	if (error != 0)
1122 		return (error);
1123 	error = uvm_io(&vm->vm_map, uio, pax_mprotect_prot(l));
1124 	uvmspace_free(vm);
1125 
1126 #ifdef PMAP_NEED_PROCWR
1127 	if (error == 0 && uio->uio_rw == UIO_WRITE)
1128 		pmap_procwr(p, addr, len);
1129 #endif
1130 	return (error);
1131 }
1132 #endif /* KTRACE || PTRACE */
1133 
1134 #if defined(KTRACE) || defined(PTRACE)
1135 void
1136 process_stoptrace(void)
1137 {
1138 	struct lwp *l = curlwp;
1139 	struct proc *p = l->l_proc, *pp;
1140 
1141 	mutex_enter(proc_lock);
1142 	mutex_enter(p->p_lock);
1143 	pp = p->p_pptr;
1144 	if (pp->p_pid == 1) {
1145 		CLR(p->p_slflag, PSL_SYSCALL);	/* XXXSMP */
1146 		mutex_exit(p->p_lock);
1147 		mutex_exit(proc_lock);
1148 		return;
1149 	}
1150 
1151 	p->p_xsig = SIGTRAP;
1152 	proc_stop(p, 1, SIGSTOP);
1153 	mutex_exit(proc_lock);
1154 
1155 	if (sigispending(l, 0)) {
1156 		lwp_lock(l);
1157 		l->l_flag |= LW_PENDSIG;
1158 		lwp_unlock(l);
1159 	}
1160 	mutex_exit(p->p_lock);
1161 }
1162 #endif	/* KTRACE || PTRACE */
1163 
1164 #ifdef PTRACE
1165 static int
1166 process_auxv_offset(struct proc *p, struct uio *uio)
1167 {
1168 	struct ps_strings pss;
1169 	int error;
1170 	off_t off = (off_t)p->p_psstrp;
1171 
1172 	if ((error = copyin_psstrings(p, &pss)) != 0)
1173 		return error;
1174 
1175 	if (pss.ps_envstr == NULL)
1176 		return EIO;
1177 
1178 	uio->uio_offset += (off_t)(vaddr_t)(pss.ps_envstr + pss.ps_nenvstr + 1);
1179 #ifdef __MACHINE_STACK_GROWS_UP
1180 	if (uio->uio_offset < off)
1181 		return EIO;
1182 #else
1183 	if (uio->uio_offset > off)
1184 		return EIO;
1185 	if ((uio->uio_offset + uio->uio_resid) > off)
1186 		uio->uio_resid = off - uio->uio_offset;
1187 #endif
1188 	return 0;
1189 }
1190 #endif
1191