xref: /freebsd/sys/kern/kern_proc.c (revision 3e0efd2ec4fcb4cd68fb8ccf8aea6fc6151c454b)
1 /*-
2  * Copyright (c) 1982, 1986, 1989, 1991, 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  * 4. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	@(#)kern_proc.c	8.7 (Berkeley) 2/14/95
30  */
31 
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
34 
35 #include "opt_compat.h"
36 #include "opt_ddb.h"
37 #include "opt_kdtrace.h"
38 #include "opt_ktrace.h"
39 #include "opt_kstack_pages.h"
40 #include "opt_stack.h"
41 
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/elf.h>
45 #include <sys/exec.h>
46 #include <sys/kernel.h>
47 #include <sys/limits.h>
48 #include <sys/lock.h>
49 #include <sys/loginclass.h>
50 #include <sys/malloc.h>
51 #include <sys/mman.h>
52 #include <sys/mount.h>
53 #include <sys/mutex.h>
54 #include <sys/proc.h>
55 #include <sys/ptrace.h>
56 #include <sys/refcount.h>
57 #include <sys/resourcevar.h>
58 #include <sys/sbuf.h>
59 #include <sys/sysent.h>
60 #include <sys/sched.h>
61 #include <sys/smp.h>
62 #include <sys/stack.h>
63 #include <sys/stat.h>
64 #include <sys/sysctl.h>
65 #include <sys/filedesc.h>
66 #include <sys/tty.h>
67 #include <sys/signalvar.h>
68 #include <sys/sdt.h>
69 #include <sys/sx.h>
70 #include <sys/user.h>
71 #include <sys/jail.h>
72 #include <sys/vnode.h>
73 #include <sys/eventhandler.h>
74 
75 #ifdef DDB
76 #include <ddb/ddb.h>
77 #endif
78 
79 #include <vm/vm.h>
80 #include <vm/vm_extern.h>
81 #include <vm/pmap.h>
82 #include <vm/vm_map.h>
83 #include <vm/vm_object.h>
84 #include <vm/vm_page.h>
85 #include <vm/uma.h>
86 
87 #ifdef COMPAT_FREEBSD32
88 #include <compat/freebsd32/freebsd32.h>
89 #include <compat/freebsd32/freebsd32_util.h>
90 #endif
91 
92 SDT_PROVIDER_DEFINE(proc);
93 SDT_PROBE_DEFINE(proc, kernel, ctor, entry, entry);
94 SDT_PROBE_ARGTYPE(proc, kernel, ctor, entry, 0, "struct proc *");
95 SDT_PROBE_ARGTYPE(proc, kernel, ctor, entry, 1, "int");
96 SDT_PROBE_ARGTYPE(proc, kernel, ctor, entry, 2, "void *");
97 SDT_PROBE_ARGTYPE(proc, kernel, ctor, entry, 3, "int");
98 SDT_PROBE_DEFINE(proc, kernel, ctor, return, return);
99 SDT_PROBE_ARGTYPE(proc, kernel, ctor, return, 0, "struct proc *");
100 SDT_PROBE_ARGTYPE(proc, kernel, ctor, return, 1, "int");
101 SDT_PROBE_ARGTYPE(proc, kernel, ctor, return, 2, "void *");
102 SDT_PROBE_ARGTYPE(proc, kernel, ctor, return, 3, "int");
103 SDT_PROBE_DEFINE(proc, kernel, dtor, entry, entry);
104 SDT_PROBE_ARGTYPE(proc, kernel, dtor, entry, 0, "struct proc *");
105 SDT_PROBE_ARGTYPE(proc, kernel, dtor, entry, 1, "int");
106 SDT_PROBE_ARGTYPE(proc, kernel, dtor, entry, 2, "void *");
107 SDT_PROBE_ARGTYPE(proc, kernel, dtor, entry, 3, "struct thread *");
108 SDT_PROBE_DEFINE(proc, kernel, dtor, return, return);
109 SDT_PROBE_ARGTYPE(proc, kernel, dtor, return, 0, "struct proc *");
110 SDT_PROBE_ARGTYPE(proc, kernel, dtor, return, 1, "int");
111 SDT_PROBE_ARGTYPE(proc, kernel, dtor, return, 2, "void *");
112 SDT_PROBE_DEFINE(proc, kernel, init, entry, entry);
113 SDT_PROBE_ARGTYPE(proc, kernel, init, entry, 0, "struct proc *");
114 SDT_PROBE_ARGTYPE(proc, kernel, init, entry, 1, "int");
115 SDT_PROBE_ARGTYPE(proc, kernel, init, entry, 2, "int");
116 SDT_PROBE_DEFINE(proc, kernel, init, return, return);
117 SDT_PROBE_ARGTYPE(proc, kernel, init, return, 0, "struct proc *");
118 SDT_PROBE_ARGTYPE(proc, kernel, init, return, 1, "int");
119 SDT_PROBE_ARGTYPE(proc, kernel, init, return, 2, "int");
120 
121 MALLOC_DEFINE(M_PGRP, "pgrp", "process group header");
122 MALLOC_DEFINE(M_SESSION, "session", "session header");
123 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
124 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
125 
126 static void doenterpgrp(struct proc *, struct pgrp *);
127 static void orphanpg(struct pgrp *pg);
128 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp);
129 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp);
130 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp,
131     int preferthread);
132 static void pgadjustjobc(struct pgrp *pgrp, int entering);
133 static void pgdelete(struct pgrp *);
134 static int proc_ctor(void *mem, int size, void *arg, int flags);
135 static void proc_dtor(void *mem, int size, void *arg);
136 static int proc_init(void *mem, int size, int flags);
137 static void proc_fini(void *mem, int size);
138 static void pargs_free(struct pargs *pa);
139 
140 /*
141  * Other process lists
142  */
143 struct pidhashhead *pidhashtbl;
144 u_long pidhash;
145 struct pgrphashhead *pgrphashtbl;
146 u_long pgrphash;
147 struct proclist allproc;
148 struct proclist zombproc;
149 struct sx allproc_lock;
150 struct sx proctree_lock;
151 struct mtx ppeers_lock;
152 uma_zone_t proc_zone;
153 
154 int kstack_pages = KSTACK_PAGES;
155 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0,
156     "Kernel stack size in pages");
157 
158 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
159 #ifdef COMPAT_FREEBSD32
160 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
161 #endif
162 
163 /*
164  * Initialize global process hashing structures.
165  */
166 void
167 procinit()
168 {
169 
170 	sx_init(&allproc_lock, "allproc");
171 	sx_init(&proctree_lock, "proctree");
172 	mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
173 	LIST_INIT(&allproc);
174 	LIST_INIT(&zombproc);
175 	pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
176 	pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
177 	proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
178 	    proc_ctor, proc_dtor, proc_init, proc_fini,
179 	    UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
180 	uihashinit();
181 }
182 
183 /*
184  * Prepare a proc for use.
185  */
186 static int
187 proc_ctor(void *mem, int size, void *arg, int flags)
188 {
189 	struct proc *p;
190 
191 	p = (struct proc *)mem;
192 	SDT_PROBE(proc, kernel, ctor , entry, p, size, arg, flags, 0);
193 	EVENTHANDLER_INVOKE(process_ctor, p);
194 	SDT_PROBE(proc, kernel, ctor , return, p, size, arg, flags, 0);
195 	return (0);
196 }
197 
198 /*
199  * Reclaim a proc after use.
200  */
201 static void
202 proc_dtor(void *mem, int size, void *arg)
203 {
204 	struct proc *p;
205 	struct thread *td;
206 
207 	/* INVARIANTS checks go here */
208 	p = (struct proc *)mem;
209 	td = FIRST_THREAD_IN_PROC(p);
210 	SDT_PROBE(proc, kernel, dtor, entry, p, size, arg, td, 0);
211 	if (td != NULL) {
212 #ifdef INVARIANTS
213 		KASSERT((p->p_numthreads == 1),
214 		    ("bad number of threads in exiting process"));
215 		KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
216 #endif
217 		/* Free all OSD associated to this thread. */
218 		osd_thread_exit(td);
219 	}
220 	EVENTHANDLER_INVOKE(process_dtor, p);
221 	if (p->p_ksi != NULL)
222 		KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
223 	SDT_PROBE(proc, kernel, dtor, return, p, size, arg, 0, 0);
224 }
225 
226 /*
227  * Initialize type-stable parts of a proc (when newly created).
228  */
229 static int
230 proc_init(void *mem, int size, int flags)
231 {
232 	struct proc *p;
233 
234 	p = (struct proc *)mem;
235 	SDT_PROBE(proc, kernel, init, entry, p, size, flags, 0, 0);
236 	p->p_sched = (struct p_sched *)&p[1];
237 	bzero(&p->p_mtx, sizeof(struct mtx));
238 	mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK);
239 	mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_RECURSE);
240 	cv_init(&p->p_pwait, "ppwait");
241 	cv_init(&p->p_dbgwait, "dbgwait");
242 	TAILQ_INIT(&p->p_threads);	     /* all threads in proc */
243 	EVENTHANDLER_INVOKE(process_init, p);
244 	p->p_stats = pstats_alloc();
245 	SDT_PROBE(proc, kernel, init, return, p, size, flags, 0, 0);
246 	return (0);
247 }
248 
249 /*
250  * UMA should ensure that this function is never called.
251  * Freeing a proc structure would violate type stability.
252  */
253 static void
254 proc_fini(void *mem, int size)
255 {
256 #ifdef notnow
257 	struct proc *p;
258 
259 	p = (struct proc *)mem;
260 	EVENTHANDLER_INVOKE(process_fini, p);
261 	pstats_free(p->p_stats);
262 	thread_free(FIRST_THREAD_IN_PROC(p));
263 	mtx_destroy(&p->p_mtx);
264 	if (p->p_ksi != NULL)
265 		ksiginfo_free(p->p_ksi);
266 #else
267 	panic("proc reclaimed");
268 #endif
269 }
270 
271 /*
272  * Is p an inferior of the current process?
273  */
274 int
275 inferior(p)
276 	register struct proc *p;
277 {
278 
279 	sx_assert(&proctree_lock, SX_LOCKED);
280 	for (; p != curproc; p = p->p_pptr)
281 		if (p->p_pid == 0)
282 			return (0);
283 	return (1);
284 }
285 
286 /*
287  * Locate a process by number; return only "live" processes -- i.e., neither
288  * zombies nor newly born but incompletely initialized processes.  By not
289  * returning processes in the PRS_NEW state, we allow callers to avoid
290  * testing for that condition to avoid dereferencing p_ucred, et al.
291  */
292 struct proc *
293 pfind(pid)
294 	register pid_t pid;
295 {
296 	register struct proc *p;
297 
298 	sx_slock(&allproc_lock);
299 	LIST_FOREACH(p, PIDHASH(pid), p_hash)
300 		if (p->p_pid == pid) {
301 			PROC_LOCK(p);
302 			if (p->p_state == PRS_NEW) {
303 				PROC_UNLOCK(p);
304 				p = NULL;
305 			}
306 			break;
307 		}
308 	sx_sunlock(&allproc_lock);
309 	return (p);
310 }
311 
312 static struct proc *
313 pfind_tid(pid_t tid)
314 {
315 	struct proc *p;
316 	struct thread *td;
317 
318 	sx_slock(&allproc_lock);
319 	FOREACH_PROC_IN_SYSTEM(p) {
320 		PROC_LOCK(p);
321 		if (p->p_state == PRS_NEW) {
322 			PROC_UNLOCK(p);
323 			continue;
324 		}
325 		FOREACH_THREAD_IN_PROC(p, td) {
326 			if (td->td_tid == tid)
327 				goto found;
328 		}
329 		PROC_UNLOCK(p);
330 	}
331 found:
332 	sx_sunlock(&allproc_lock);
333 	return (p);
334 }
335 
336 /*
337  * Locate a process group by number.
338  * The caller must hold proctree_lock.
339  */
340 struct pgrp *
341 pgfind(pgid)
342 	register pid_t pgid;
343 {
344 	register struct pgrp *pgrp;
345 
346 	sx_assert(&proctree_lock, SX_LOCKED);
347 
348 	LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
349 		if (pgrp->pg_id == pgid) {
350 			PGRP_LOCK(pgrp);
351 			return (pgrp);
352 		}
353 	}
354 	return (NULL);
355 }
356 
357 /*
358  * Locate process and do additional manipulations, depending on flags.
359  */
360 int
361 pget(pid_t pid, int flags, struct proc **pp)
362 {
363 	struct proc *p;
364 	int error;
365 
366 	if (pid <= PID_MAX)
367 		p = pfind(pid);
368 	else if ((flags & PGET_NOTID) == 0)
369 		p = pfind_tid(pid);
370 	else
371 		p = NULL;
372 	if (p == NULL)
373 		return (ESRCH);
374 	if ((flags & PGET_CANSEE) != 0) {
375 		error = p_cansee(curthread, p);
376 		if (error != 0)
377 			goto errout;
378 	}
379 	if ((flags & PGET_CANDEBUG) != 0) {
380 		error = p_candebug(curthread, p);
381 		if (error != 0)
382 			goto errout;
383 	}
384 	if ((flags & PGET_ISCURRENT) != 0 && curproc != p) {
385 		error = EPERM;
386 		goto errout;
387 	}
388 	if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) {
389 		error = ESRCH;
390 		goto errout;
391 	}
392 	if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) {
393 		/*
394 		 * XXXRW: Not clear ESRCH is the right error during proc
395 		 * execve().
396 		 */
397 		error = ESRCH;
398 		goto errout;
399 	}
400 	if ((flags & PGET_HOLD) != 0) {
401 		_PHOLD(p);
402 		PROC_UNLOCK(p);
403 	}
404 	*pp = p;
405 	return (0);
406 errout:
407 	PROC_UNLOCK(p);
408 	return (error);
409 }
410 
411 /*
412  * Create a new process group.
413  * pgid must be equal to the pid of p.
414  * Begin a new session if required.
415  */
416 int
417 enterpgrp(p, pgid, pgrp, sess)
418 	register struct proc *p;
419 	pid_t pgid;
420 	struct pgrp *pgrp;
421 	struct session *sess;
422 {
423 	struct pgrp *pgrp2;
424 
425 	sx_assert(&proctree_lock, SX_XLOCKED);
426 
427 	KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
428 	KASSERT(p->p_pid == pgid,
429 	    ("enterpgrp: new pgrp and pid != pgid"));
430 
431 	pgrp2 = pgfind(pgid);
432 
433 	KASSERT(pgrp2 == NULL,
434 	    ("enterpgrp: pgrp with pgid exists"));
435 	KASSERT(!SESS_LEADER(p),
436 	    ("enterpgrp: session leader attempted setpgrp"));
437 
438 	mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
439 
440 	if (sess != NULL) {
441 		/*
442 		 * new session
443 		 */
444 		mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
445 		PROC_LOCK(p);
446 		p->p_flag &= ~P_CONTROLT;
447 		PROC_UNLOCK(p);
448 		PGRP_LOCK(pgrp);
449 		sess->s_leader = p;
450 		sess->s_sid = p->p_pid;
451 		refcount_init(&sess->s_count, 1);
452 		sess->s_ttyvp = NULL;
453 		sess->s_ttydp = NULL;
454 		sess->s_ttyp = NULL;
455 		bcopy(p->p_session->s_login, sess->s_login,
456 			    sizeof(sess->s_login));
457 		pgrp->pg_session = sess;
458 		KASSERT(p == curproc,
459 		    ("enterpgrp: mksession and p != curproc"));
460 	} else {
461 		pgrp->pg_session = p->p_session;
462 		sess_hold(pgrp->pg_session);
463 		PGRP_LOCK(pgrp);
464 	}
465 	pgrp->pg_id = pgid;
466 	LIST_INIT(&pgrp->pg_members);
467 
468 	/*
469 	 * As we have an exclusive lock of proctree_lock,
470 	 * this should not deadlock.
471 	 */
472 	LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
473 	pgrp->pg_jobc = 0;
474 	SLIST_INIT(&pgrp->pg_sigiolst);
475 	PGRP_UNLOCK(pgrp);
476 
477 	doenterpgrp(p, pgrp);
478 
479 	return (0);
480 }
481 
482 /*
483  * Move p to an existing process group
484  */
485 int
486 enterthispgrp(p, pgrp)
487 	register struct proc *p;
488 	struct pgrp *pgrp;
489 {
490 
491 	sx_assert(&proctree_lock, SX_XLOCKED);
492 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
493 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
494 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
495 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
496 	KASSERT(pgrp->pg_session == p->p_session,
497 		("%s: pgrp's session %p, p->p_session %p.\n",
498 		__func__,
499 		pgrp->pg_session,
500 		p->p_session));
501 	KASSERT(pgrp != p->p_pgrp,
502 		("%s: p belongs to pgrp.", __func__));
503 
504 	doenterpgrp(p, pgrp);
505 
506 	return (0);
507 }
508 
509 /*
510  * Move p to a process group
511  */
512 static void
513 doenterpgrp(p, pgrp)
514 	struct proc *p;
515 	struct pgrp *pgrp;
516 {
517 	struct pgrp *savepgrp;
518 
519 	sx_assert(&proctree_lock, SX_XLOCKED);
520 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
521 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
522 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
523 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
524 
525 	savepgrp = p->p_pgrp;
526 
527 	/*
528 	 * Adjust eligibility of affected pgrps to participate in job control.
529 	 * Increment eligibility counts before decrementing, otherwise we
530 	 * could reach 0 spuriously during the first call.
531 	 */
532 	fixjobc(p, pgrp, 1);
533 	fixjobc(p, p->p_pgrp, 0);
534 
535 	PGRP_LOCK(pgrp);
536 	PGRP_LOCK(savepgrp);
537 	PROC_LOCK(p);
538 	LIST_REMOVE(p, p_pglist);
539 	p->p_pgrp = pgrp;
540 	PROC_UNLOCK(p);
541 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
542 	PGRP_UNLOCK(savepgrp);
543 	PGRP_UNLOCK(pgrp);
544 	if (LIST_EMPTY(&savepgrp->pg_members))
545 		pgdelete(savepgrp);
546 }
547 
548 /*
549  * remove process from process group
550  */
551 int
552 leavepgrp(p)
553 	register struct proc *p;
554 {
555 	struct pgrp *savepgrp;
556 
557 	sx_assert(&proctree_lock, SX_XLOCKED);
558 	savepgrp = p->p_pgrp;
559 	PGRP_LOCK(savepgrp);
560 	PROC_LOCK(p);
561 	LIST_REMOVE(p, p_pglist);
562 	p->p_pgrp = NULL;
563 	PROC_UNLOCK(p);
564 	PGRP_UNLOCK(savepgrp);
565 	if (LIST_EMPTY(&savepgrp->pg_members))
566 		pgdelete(savepgrp);
567 	return (0);
568 }
569 
570 /*
571  * delete a process group
572  */
573 static void
574 pgdelete(pgrp)
575 	register struct pgrp *pgrp;
576 {
577 	struct session *savesess;
578 	struct tty *tp;
579 
580 	sx_assert(&proctree_lock, SX_XLOCKED);
581 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
582 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
583 
584 	/*
585 	 * Reset any sigio structures pointing to us as a result of
586 	 * F_SETOWN with our pgid.
587 	 */
588 	funsetownlst(&pgrp->pg_sigiolst);
589 
590 	PGRP_LOCK(pgrp);
591 	tp = pgrp->pg_session->s_ttyp;
592 	LIST_REMOVE(pgrp, pg_hash);
593 	savesess = pgrp->pg_session;
594 	PGRP_UNLOCK(pgrp);
595 
596 	/* Remove the reference to the pgrp before deallocating it. */
597 	if (tp != NULL) {
598 		tty_lock(tp);
599 		tty_rel_pgrp(tp, pgrp);
600 	}
601 
602 	mtx_destroy(&pgrp->pg_mtx);
603 	free(pgrp, M_PGRP);
604 	sess_release(savesess);
605 }
606 
607 static void
608 pgadjustjobc(pgrp, entering)
609 	struct pgrp *pgrp;
610 	int entering;
611 {
612 
613 	PGRP_LOCK(pgrp);
614 	if (entering)
615 		pgrp->pg_jobc++;
616 	else {
617 		--pgrp->pg_jobc;
618 		if (pgrp->pg_jobc == 0)
619 			orphanpg(pgrp);
620 	}
621 	PGRP_UNLOCK(pgrp);
622 }
623 
624 /*
625  * Adjust pgrp jobc counters when specified process changes process group.
626  * We count the number of processes in each process group that "qualify"
627  * the group for terminal job control (those with a parent in a different
628  * process group of the same session).  If that count reaches zero, the
629  * process group becomes orphaned.  Check both the specified process'
630  * process group and that of its children.
631  * entering == 0 => p is leaving specified group.
632  * entering == 1 => p is entering specified group.
633  */
634 void
635 fixjobc(p, pgrp, entering)
636 	register struct proc *p;
637 	register struct pgrp *pgrp;
638 	int entering;
639 {
640 	register struct pgrp *hispgrp;
641 	register struct session *mysession;
642 
643 	sx_assert(&proctree_lock, SX_LOCKED);
644 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
645 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
646 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
647 
648 	/*
649 	 * Check p's parent to see whether p qualifies its own process
650 	 * group; if so, adjust count for p's process group.
651 	 */
652 	mysession = pgrp->pg_session;
653 	if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
654 	    hispgrp->pg_session == mysession)
655 		pgadjustjobc(pgrp, entering);
656 
657 	/*
658 	 * Check this process' children to see whether they qualify
659 	 * their process groups; if so, adjust counts for children's
660 	 * process groups.
661 	 */
662 	LIST_FOREACH(p, &p->p_children, p_sibling) {
663 		hispgrp = p->p_pgrp;
664 		if (hispgrp == pgrp ||
665 		    hispgrp->pg_session != mysession)
666 			continue;
667 		PROC_LOCK(p);
668 		if (p->p_state == PRS_ZOMBIE) {
669 			PROC_UNLOCK(p);
670 			continue;
671 		}
672 		PROC_UNLOCK(p);
673 		pgadjustjobc(hispgrp, entering);
674 	}
675 }
676 
677 /*
678  * A process group has become orphaned;
679  * if there are any stopped processes in the group,
680  * hang-up all process in that group.
681  */
682 static void
683 orphanpg(pg)
684 	struct pgrp *pg;
685 {
686 	register struct proc *p;
687 
688 	PGRP_LOCK_ASSERT(pg, MA_OWNED);
689 
690 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
691 		PROC_LOCK(p);
692 		if (P_SHOULDSTOP(p)) {
693 			PROC_UNLOCK(p);
694 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
695 				PROC_LOCK(p);
696 				kern_psignal(p, SIGHUP);
697 				kern_psignal(p, SIGCONT);
698 				PROC_UNLOCK(p);
699 			}
700 			return;
701 		}
702 		PROC_UNLOCK(p);
703 	}
704 }
705 
706 void
707 sess_hold(struct session *s)
708 {
709 
710 	refcount_acquire(&s->s_count);
711 }
712 
713 void
714 sess_release(struct session *s)
715 {
716 
717 	if (refcount_release(&s->s_count)) {
718 		if (s->s_ttyp != NULL) {
719 			tty_lock(s->s_ttyp);
720 			tty_rel_sess(s->s_ttyp, s);
721 		}
722 		mtx_destroy(&s->s_mtx);
723 		free(s, M_SESSION);
724 	}
725 }
726 
727 #include "opt_ddb.h"
728 #ifdef DDB
729 #include <ddb/ddb.h>
730 
731 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
732 {
733 	register struct pgrp *pgrp;
734 	register struct proc *p;
735 	register int i;
736 
737 	for (i = 0; i <= pgrphash; i++) {
738 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
739 			printf("\tindx %d\n", i);
740 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
741 				printf(
742 			"\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
743 				    (void *)pgrp, (long)pgrp->pg_id,
744 				    (void *)pgrp->pg_session,
745 				    pgrp->pg_session->s_count,
746 				    (void *)LIST_FIRST(&pgrp->pg_members));
747 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
748 					printf("\t\tpid %ld addr %p pgrp %p\n",
749 					    (long)p->p_pid, (void *)p,
750 					    (void *)p->p_pgrp);
751 				}
752 			}
753 		}
754 	}
755 }
756 #endif /* DDB */
757 
758 /*
759  * Calculate the kinfo_proc members which contain process-wide
760  * informations.
761  * Must be called with the target process locked.
762  */
763 static void
764 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
765 {
766 	struct thread *td;
767 
768 	PROC_LOCK_ASSERT(p, MA_OWNED);
769 
770 	kp->ki_estcpu = 0;
771 	kp->ki_pctcpu = 0;
772 	FOREACH_THREAD_IN_PROC(p, td) {
773 		thread_lock(td);
774 		kp->ki_pctcpu += sched_pctcpu(td);
775 		kp->ki_estcpu += td->td_estcpu;
776 		thread_unlock(td);
777 	}
778 }
779 
780 /*
781  * Clear kinfo_proc and fill in any information that is common
782  * to all threads in the process.
783  * Must be called with the target process locked.
784  */
785 static void
786 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
787 {
788 	struct thread *td0;
789 	struct tty *tp;
790 	struct session *sp;
791 	struct ucred *cred;
792 	struct sigacts *ps;
793 
794 	PROC_LOCK_ASSERT(p, MA_OWNED);
795 	bzero(kp, sizeof(*kp));
796 
797 	kp->ki_structsize = sizeof(*kp);
798 	kp->ki_paddr = p;
799 	kp->ki_addr =/* p->p_addr; */0; /* XXX */
800 	kp->ki_args = p->p_args;
801 	kp->ki_textvp = p->p_textvp;
802 #ifdef KTRACE
803 	kp->ki_tracep = p->p_tracevp;
804 	kp->ki_traceflag = p->p_traceflag;
805 #endif
806 	kp->ki_fd = p->p_fd;
807 	kp->ki_vmspace = p->p_vmspace;
808 	kp->ki_flag = p->p_flag;
809 	cred = p->p_ucred;
810 	if (cred) {
811 		kp->ki_uid = cred->cr_uid;
812 		kp->ki_ruid = cred->cr_ruid;
813 		kp->ki_svuid = cred->cr_svuid;
814 		kp->ki_cr_flags = 0;
815 		if (cred->cr_flags & CRED_FLAG_CAPMODE)
816 			kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
817 		/* XXX bde doesn't like KI_NGROUPS */
818 		if (cred->cr_ngroups > KI_NGROUPS) {
819 			kp->ki_ngroups = KI_NGROUPS;
820 			kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
821 		} else
822 			kp->ki_ngroups = cred->cr_ngroups;
823 		bcopy(cred->cr_groups, kp->ki_groups,
824 		    kp->ki_ngroups * sizeof(gid_t));
825 		kp->ki_rgid = cred->cr_rgid;
826 		kp->ki_svgid = cred->cr_svgid;
827 		/* If jailed(cred), emulate the old P_JAILED flag. */
828 		if (jailed(cred)) {
829 			kp->ki_flag |= P_JAILED;
830 			/* If inside the jail, use 0 as a jail ID. */
831 			if (cred->cr_prison != curthread->td_ucred->cr_prison)
832 				kp->ki_jid = cred->cr_prison->pr_id;
833 		}
834 		strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
835 		    sizeof(kp->ki_loginclass));
836 	}
837 	ps = p->p_sigacts;
838 	if (ps) {
839 		mtx_lock(&ps->ps_mtx);
840 		kp->ki_sigignore = ps->ps_sigignore;
841 		kp->ki_sigcatch = ps->ps_sigcatch;
842 		mtx_unlock(&ps->ps_mtx);
843 	}
844 	if (p->p_state != PRS_NEW &&
845 	    p->p_state != PRS_ZOMBIE &&
846 	    p->p_vmspace != NULL) {
847 		struct vmspace *vm = p->p_vmspace;
848 
849 		kp->ki_size = vm->vm_map.size;
850 		kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
851 		FOREACH_THREAD_IN_PROC(p, td0) {
852 			if (!TD_IS_SWAPPED(td0))
853 				kp->ki_rssize += td0->td_kstack_pages;
854 		}
855 		kp->ki_swrss = vm->vm_swrss;
856 		kp->ki_tsize = vm->vm_tsize;
857 		kp->ki_dsize = vm->vm_dsize;
858 		kp->ki_ssize = vm->vm_ssize;
859 	} else if (p->p_state == PRS_ZOMBIE)
860 		kp->ki_stat = SZOMB;
861 	if (kp->ki_flag & P_INMEM)
862 		kp->ki_sflag = PS_INMEM;
863 	else
864 		kp->ki_sflag = 0;
865 	/* Calculate legacy swtime as seconds since 'swtick'. */
866 	kp->ki_swtime = (ticks - p->p_swtick) / hz;
867 	kp->ki_pid = p->p_pid;
868 	kp->ki_nice = p->p_nice;
869 	kp->ki_start = p->p_stats->p_start;
870 	timevaladd(&kp->ki_start, &boottime);
871 	PROC_SLOCK(p);
872 	rufetch(p, &kp->ki_rusage);
873 	kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
874 	calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
875 	PROC_SUNLOCK(p);
876 	calccru(p, &kp->ki_childutime, &kp->ki_childstime);
877 	/* Some callers want child times in a single value. */
878 	kp->ki_childtime = kp->ki_childstime;
879 	timevaladd(&kp->ki_childtime, &kp->ki_childutime);
880 
881 	FOREACH_THREAD_IN_PROC(p, td0)
882 		kp->ki_cow += td0->td_cow;
883 
884 	tp = NULL;
885 	if (p->p_pgrp) {
886 		kp->ki_pgid = p->p_pgrp->pg_id;
887 		kp->ki_jobc = p->p_pgrp->pg_jobc;
888 		sp = p->p_pgrp->pg_session;
889 
890 		if (sp != NULL) {
891 			kp->ki_sid = sp->s_sid;
892 			SESS_LOCK(sp);
893 			strlcpy(kp->ki_login, sp->s_login,
894 			    sizeof(kp->ki_login));
895 			if (sp->s_ttyvp)
896 				kp->ki_kiflag |= KI_CTTY;
897 			if (SESS_LEADER(p))
898 				kp->ki_kiflag |= KI_SLEADER;
899 			/* XXX proctree_lock */
900 			tp = sp->s_ttyp;
901 			SESS_UNLOCK(sp);
902 		}
903 	}
904 	if ((p->p_flag & P_CONTROLT) && tp != NULL) {
905 		kp->ki_tdev = tty_udev(tp);
906 		kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
907 		if (tp->t_session)
908 			kp->ki_tsid = tp->t_session->s_sid;
909 	} else
910 		kp->ki_tdev = NODEV;
911 	if (p->p_comm[0] != '\0')
912 		strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
913 	if (p->p_sysent && p->p_sysent->sv_name != NULL &&
914 	    p->p_sysent->sv_name[0] != '\0')
915 		strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
916 	kp->ki_siglist = p->p_siglist;
917 	kp->ki_xstat = p->p_xstat;
918 	kp->ki_acflag = p->p_acflag;
919 	kp->ki_lock = p->p_lock;
920 	if (p->p_pptr)
921 		kp->ki_ppid = p->p_pptr->p_pid;
922 }
923 
924 /*
925  * Fill in information that is thread specific.  Must be called with
926  * target process locked.  If 'preferthread' is set, overwrite certain
927  * process-related fields that are maintained for both threads and
928  * processes.
929  */
930 static void
931 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
932 {
933 	struct proc *p;
934 
935 	p = td->td_proc;
936 	kp->ki_tdaddr = td;
937 	PROC_LOCK_ASSERT(p, MA_OWNED);
938 
939 	if (preferthread)
940 		PROC_SLOCK(p);
941 	thread_lock(td);
942 	if (td->td_wmesg != NULL)
943 		strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
944 	else
945 		bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
946 	strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname));
947 	if (TD_ON_LOCK(td)) {
948 		kp->ki_kiflag |= KI_LOCKBLOCK;
949 		strlcpy(kp->ki_lockname, td->td_lockname,
950 		    sizeof(kp->ki_lockname));
951 	} else {
952 		kp->ki_kiflag &= ~KI_LOCKBLOCK;
953 		bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
954 	}
955 
956 	if (p->p_state == PRS_NORMAL) { /* approximate. */
957 		if (TD_ON_RUNQ(td) ||
958 		    TD_CAN_RUN(td) ||
959 		    TD_IS_RUNNING(td)) {
960 			kp->ki_stat = SRUN;
961 		} else if (P_SHOULDSTOP(p)) {
962 			kp->ki_stat = SSTOP;
963 		} else if (TD_IS_SLEEPING(td)) {
964 			kp->ki_stat = SSLEEP;
965 		} else if (TD_ON_LOCK(td)) {
966 			kp->ki_stat = SLOCK;
967 		} else {
968 			kp->ki_stat = SWAIT;
969 		}
970 	} else if (p->p_state == PRS_ZOMBIE) {
971 		kp->ki_stat = SZOMB;
972 	} else {
973 		kp->ki_stat = SIDL;
974 	}
975 
976 	/* Things in the thread */
977 	kp->ki_wchan = td->td_wchan;
978 	kp->ki_pri.pri_level = td->td_priority;
979 	kp->ki_pri.pri_native = td->td_base_pri;
980 	kp->ki_lastcpu = td->td_lastcpu;
981 	kp->ki_oncpu = td->td_oncpu;
982 	kp->ki_tdflags = td->td_flags;
983 	kp->ki_tid = td->td_tid;
984 	kp->ki_numthreads = p->p_numthreads;
985 	kp->ki_pcb = td->td_pcb;
986 	kp->ki_kstack = (void *)td->td_kstack;
987 	kp->ki_slptime = (ticks - td->td_slptick) / hz;
988 	kp->ki_pri.pri_class = td->td_pri_class;
989 	kp->ki_pri.pri_user = td->td_user_pri;
990 
991 	if (preferthread) {
992 		rufetchtd(td, &kp->ki_rusage);
993 		kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
994 		kp->ki_pctcpu = sched_pctcpu(td);
995 		kp->ki_estcpu = td->td_estcpu;
996 		kp->ki_cow = td->td_cow;
997 	}
998 
999 	/* We can't get this anymore but ps etc never used it anyway. */
1000 	kp->ki_rqindex = 0;
1001 
1002 	if (preferthread)
1003 		kp->ki_siglist = td->td_siglist;
1004 	kp->ki_sigmask = td->td_sigmask;
1005 	thread_unlock(td);
1006 	if (preferthread)
1007 		PROC_SUNLOCK(p);
1008 }
1009 
1010 /*
1011  * Fill in a kinfo_proc structure for the specified process.
1012  * Must be called with the target process locked.
1013  */
1014 void
1015 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1016 {
1017 
1018 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1019 
1020 	fill_kinfo_proc_only(p, kp);
1021 	fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1022 	fill_kinfo_aggregate(p, kp);
1023 }
1024 
1025 struct pstats *
1026 pstats_alloc(void)
1027 {
1028 
1029 	return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1030 }
1031 
1032 /*
1033  * Copy parts of p_stats; zero the rest of p_stats (statistics).
1034  */
1035 void
1036 pstats_fork(struct pstats *src, struct pstats *dst)
1037 {
1038 
1039 	bzero(&dst->pstat_startzero,
1040 	    __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1041 	bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1042 	    __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1043 }
1044 
1045 void
1046 pstats_free(struct pstats *ps)
1047 {
1048 
1049 	free(ps, M_SUBPROC);
1050 }
1051 
1052 /*
1053  * Locate a zombie process by number
1054  */
1055 struct proc *
1056 zpfind(pid_t pid)
1057 {
1058 	struct proc *p;
1059 
1060 	sx_slock(&allproc_lock);
1061 	LIST_FOREACH(p, &zombproc, p_list)
1062 		if (p->p_pid == pid) {
1063 			PROC_LOCK(p);
1064 			break;
1065 		}
1066 	sx_sunlock(&allproc_lock);
1067 	return (p);
1068 }
1069 
1070 #define KERN_PROC_ZOMBMASK	0x3
1071 #define KERN_PROC_NOTHREADS	0x4
1072 
1073 #ifdef COMPAT_FREEBSD32
1074 
1075 /*
1076  * This function is typically used to copy out the kernel address, so
1077  * it can be replaced by assignment of zero.
1078  */
1079 static inline uint32_t
1080 ptr32_trim(void *ptr)
1081 {
1082 	uintptr_t uptr;
1083 
1084 	uptr = (uintptr_t)ptr;
1085 	return ((uptr > UINT_MAX) ? 0 : uptr);
1086 }
1087 
1088 #define PTRTRIM_CP(src,dst,fld) \
1089 	do { (dst).fld = ptr32_trim((src).fld); } while (0)
1090 
1091 static void
1092 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1093 {
1094 	int i;
1095 
1096 	bzero(ki32, sizeof(struct kinfo_proc32));
1097 	ki32->ki_structsize = sizeof(struct kinfo_proc32);
1098 	CP(*ki, *ki32, ki_layout);
1099 	PTRTRIM_CP(*ki, *ki32, ki_args);
1100 	PTRTRIM_CP(*ki, *ki32, ki_paddr);
1101 	PTRTRIM_CP(*ki, *ki32, ki_addr);
1102 	PTRTRIM_CP(*ki, *ki32, ki_tracep);
1103 	PTRTRIM_CP(*ki, *ki32, ki_textvp);
1104 	PTRTRIM_CP(*ki, *ki32, ki_fd);
1105 	PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1106 	PTRTRIM_CP(*ki, *ki32, ki_wchan);
1107 	CP(*ki, *ki32, ki_pid);
1108 	CP(*ki, *ki32, ki_ppid);
1109 	CP(*ki, *ki32, ki_pgid);
1110 	CP(*ki, *ki32, ki_tpgid);
1111 	CP(*ki, *ki32, ki_sid);
1112 	CP(*ki, *ki32, ki_tsid);
1113 	CP(*ki, *ki32, ki_jobc);
1114 	CP(*ki, *ki32, ki_tdev);
1115 	CP(*ki, *ki32, ki_siglist);
1116 	CP(*ki, *ki32, ki_sigmask);
1117 	CP(*ki, *ki32, ki_sigignore);
1118 	CP(*ki, *ki32, ki_sigcatch);
1119 	CP(*ki, *ki32, ki_uid);
1120 	CP(*ki, *ki32, ki_ruid);
1121 	CP(*ki, *ki32, ki_svuid);
1122 	CP(*ki, *ki32, ki_rgid);
1123 	CP(*ki, *ki32, ki_svgid);
1124 	CP(*ki, *ki32, ki_ngroups);
1125 	for (i = 0; i < KI_NGROUPS; i++)
1126 		CP(*ki, *ki32, ki_groups[i]);
1127 	CP(*ki, *ki32, ki_size);
1128 	CP(*ki, *ki32, ki_rssize);
1129 	CP(*ki, *ki32, ki_swrss);
1130 	CP(*ki, *ki32, ki_tsize);
1131 	CP(*ki, *ki32, ki_dsize);
1132 	CP(*ki, *ki32, ki_ssize);
1133 	CP(*ki, *ki32, ki_xstat);
1134 	CP(*ki, *ki32, ki_acflag);
1135 	CP(*ki, *ki32, ki_pctcpu);
1136 	CP(*ki, *ki32, ki_estcpu);
1137 	CP(*ki, *ki32, ki_slptime);
1138 	CP(*ki, *ki32, ki_swtime);
1139 	CP(*ki, *ki32, ki_cow);
1140 	CP(*ki, *ki32, ki_runtime);
1141 	TV_CP(*ki, *ki32, ki_start);
1142 	TV_CP(*ki, *ki32, ki_childtime);
1143 	CP(*ki, *ki32, ki_flag);
1144 	CP(*ki, *ki32, ki_kiflag);
1145 	CP(*ki, *ki32, ki_traceflag);
1146 	CP(*ki, *ki32, ki_stat);
1147 	CP(*ki, *ki32, ki_nice);
1148 	CP(*ki, *ki32, ki_lock);
1149 	CP(*ki, *ki32, ki_rqindex);
1150 	CP(*ki, *ki32, ki_oncpu);
1151 	CP(*ki, *ki32, ki_lastcpu);
1152 	bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1153 	bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1154 	bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1155 	bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1156 	bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1157 	bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1158 	bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1159 	CP(*ki, *ki32, ki_cr_flags);
1160 	CP(*ki, *ki32, ki_jid);
1161 	CP(*ki, *ki32, ki_numthreads);
1162 	CP(*ki, *ki32, ki_tid);
1163 	CP(*ki, *ki32, ki_pri);
1164 	freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1165 	freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1166 	PTRTRIM_CP(*ki, *ki32, ki_pcb);
1167 	PTRTRIM_CP(*ki, *ki32, ki_kstack);
1168 	PTRTRIM_CP(*ki, *ki32, ki_udata);
1169 	CP(*ki, *ki32, ki_sflag);
1170 	CP(*ki, *ki32, ki_tdflags);
1171 }
1172 
1173 static int
1174 sysctl_out_proc_copyout(struct kinfo_proc *ki, struct sysctl_req *req)
1175 {
1176 	struct kinfo_proc32 ki32;
1177 	int error;
1178 
1179 	if (req->flags & SCTL_MASK32) {
1180 		freebsd32_kinfo_proc_out(ki, &ki32);
1181 		error = SYSCTL_OUT(req, &ki32, sizeof(struct kinfo_proc32));
1182 	} else
1183 		error = SYSCTL_OUT(req, ki, sizeof(struct kinfo_proc));
1184 	return (error);
1185 }
1186 #else
1187 static int
1188 sysctl_out_proc_copyout(struct kinfo_proc *ki, struct sysctl_req *req)
1189 {
1190 
1191 	return (SYSCTL_OUT(req, ki, sizeof(struct kinfo_proc)));
1192 }
1193 #endif
1194 
1195 /*
1196  * Must be called with the process locked and will return with it unlocked.
1197  */
1198 static int
1199 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
1200 {
1201 	struct thread *td;
1202 	struct kinfo_proc kinfo_proc;
1203 	int error = 0;
1204 	struct proc *np;
1205 	pid_t pid = p->p_pid;
1206 
1207 	PROC_LOCK_ASSERT(p, MA_OWNED);
1208 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1209 
1210 	fill_kinfo_proc(p, &kinfo_proc);
1211 	if (flags & KERN_PROC_NOTHREADS)
1212 		error = sysctl_out_proc_copyout(&kinfo_proc, req);
1213 	else {
1214 		FOREACH_THREAD_IN_PROC(p, td) {
1215 			fill_kinfo_thread(td, &kinfo_proc, 1);
1216 			error = sysctl_out_proc_copyout(&kinfo_proc, req);
1217 			if (error)
1218 				break;
1219 		}
1220 	}
1221 	PROC_UNLOCK(p);
1222 	if (error)
1223 		return (error);
1224 	if (flags & KERN_PROC_ZOMBMASK)
1225 		np = zpfind(pid);
1226 	else {
1227 		if (pid == 0)
1228 			return (0);
1229 		np = pfind(pid);
1230 	}
1231 	if (np == NULL)
1232 		return (ESRCH);
1233 	if (np != p) {
1234 		PROC_UNLOCK(np);
1235 		return (ESRCH);
1236 	}
1237 	PROC_UNLOCK(np);
1238 	return (0);
1239 }
1240 
1241 static int
1242 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1243 {
1244 	int *name = (int *)arg1;
1245 	u_int namelen = arg2;
1246 	struct proc *p;
1247 	int flags, doingzomb, oid_number;
1248 	int error = 0;
1249 
1250 	oid_number = oidp->oid_number;
1251 	if (oid_number != KERN_PROC_ALL &&
1252 	    (oid_number & KERN_PROC_INC_THREAD) == 0)
1253 		flags = KERN_PROC_NOTHREADS;
1254 	else {
1255 		flags = 0;
1256 		oid_number &= ~KERN_PROC_INC_THREAD;
1257 	}
1258 	if (oid_number == KERN_PROC_PID) {
1259 		if (namelen != 1)
1260 			return (EINVAL);
1261 		error = sysctl_wire_old_buffer(req, 0);
1262 		if (error)
1263 			return (error);
1264 		error = pget((pid_t)name[0], PGET_CANSEE, &p);
1265 		if (error != 0)
1266 			return (error);
1267 		error = sysctl_out_proc(p, req, flags);
1268 		return (error);
1269 	}
1270 
1271 	switch (oid_number) {
1272 	case KERN_PROC_ALL:
1273 		if (namelen != 0)
1274 			return (EINVAL);
1275 		break;
1276 	case KERN_PROC_PROC:
1277 		if (namelen != 0 && namelen != 1)
1278 			return (EINVAL);
1279 		break;
1280 	default:
1281 		if (namelen != 1)
1282 			return (EINVAL);
1283 		break;
1284 	}
1285 
1286 	if (!req->oldptr) {
1287 		/* overestimate by 5 procs */
1288 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1289 		if (error)
1290 			return (error);
1291 	}
1292 	error = sysctl_wire_old_buffer(req, 0);
1293 	if (error != 0)
1294 		return (error);
1295 	sx_slock(&allproc_lock);
1296 	for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) {
1297 		if (!doingzomb)
1298 			p = LIST_FIRST(&allproc);
1299 		else
1300 			p = LIST_FIRST(&zombproc);
1301 		for (; p != 0; p = LIST_NEXT(p, p_list)) {
1302 			/*
1303 			 * Skip embryonic processes.
1304 			 */
1305 			PROC_LOCK(p);
1306 			if (p->p_state == PRS_NEW) {
1307 				PROC_UNLOCK(p);
1308 				continue;
1309 			}
1310 			KASSERT(p->p_ucred != NULL,
1311 			    ("process credential is NULL for non-NEW proc"));
1312 			/*
1313 			 * Show a user only appropriate processes.
1314 			 */
1315 			if (p_cansee(curthread, p)) {
1316 				PROC_UNLOCK(p);
1317 				continue;
1318 			}
1319 			/*
1320 			 * TODO - make more efficient (see notes below).
1321 			 * do by session.
1322 			 */
1323 			switch (oid_number) {
1324 
1325 			case KERN_PROC_GID:
1326 				if (p->p_ucred->cr_gid != (gid_t)name[0]) {
1327 					PROC_UNLOCK(p);
1328 					continue;
1329 				}
1330 				break;
1331 
1332 			case KERN_PROC_PGRP:
1333 				/* could do this by traversing pgrp */
1334 				if (p->p_pgrp == NULL ||
1335 				    p->p_pgrp->pg_id != (pid_t)name[0]) {
1336 					PROC_UNLOCK(p);
1337 					continue;
1338 				}
1339 				break;
1340 
1341 			case KERN_PROC_RGID:
1342 				if (p->p_ucred->cr_rgid != (gid_t)name[0]) {
1343 					PROC_UNLOCK(p);
1344 					continue;
1345 				}
1346 				break;
1347 
1348 			case KERN_PROC_SESSION:
1349 				if (p->p_session == NULL ||
1350 				    p->p_session->s_sid != (pid_t)name[0]) {
1351 					PROC_UNLOCK(p);
1352 					continue;
1353 				}
1354 				break;
1355 
1356 			case KERN_PROC_TTY:
1357 				if ((p->p_flag & P_CONTROLT) == 0 ||
1358 				    p->p_session == NULL) {
1359 					PROC_UNLOCK(p);
1360 					continue;
1361 				}
1362 				/* XXX proctree_lock */
1363 				SESS_LOCK(p->p_session);
1364 				if (p->p_session->s_ttyp == NULL ||
1365 				    tty_udev(p->p_session->s_ttyp) !=
1366 				    (dev_t)name[0]) {
1367 					SESS_UNLOCK(p->p_session);
1368 					PROC_UNLOCK(p);
1369 					continue;
1370 				}
1371 				SESS_UNLOCK(p->p_session);
1372 				break;
1373 
1374 			case KERN_PROC_UID:
1375 				if (p->p_ucred->cr_uid != (uid_t)name[0]) {
1376 					PROC_UNLOCK(p);
1377 					continue;
1378 				}
1379 				break;
1380 
1381 			case KERN_PROC_RUID:
1382 				if (p->p_ucred->cr_ruid != (uid_t)name[0]) {
1383 					PROC_UNLOCK(p);
1384 					continue;
1385 				}
1386 				break;
1387 
1388 			case KERN_PROC_PROC:
1389 				break;
1390 
1391 			default:
1392 				break;
1393 
1394 			}
1395 
1396 			error = sysctl_out_proc(p, req, flags | doingzomb);
1397 			if (error) {
1398 				sx_sunlock(&allproc_lock);
1399 				return (error);
1400 			}
1401 		}
1402 	}
1403 	sx_sunlock(&allproc_lock);
1404 	return (0);
1405 }
1406 
1407 struct pargs *
1408 pargs_alloc(int len)
1409 {
1410 	struct pargs *pa;
1411 
1412 	pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1413 		M_WAITOK);
1414 	refcount_init(&pa->ar_ref, 1);
1415 	pa->ar_length = len;
1416 	return (pa);
1417 }
1418 
1419 static void
1420 pargs_free(struct pargs *pa)
1421 {
1422 
1423 	free(pa, M_PARGS);
1424 }
1425 
1426 void
1427 pargs_hold(struct pargs *pa)
1428 {
1429 
1430 	if (pa == NULL)
1431 		return;
1432 	refcount_acquire(&pa->ar_ref);
1433 }
1434 
1435 void
1436 pargs_drop(struct pargs *pa)
1437 {
1438 
1439 	if (pa == NULL)
1440 		return;
1441 	if (refcount_release(&pa->ar_ref))
1442 		pargs_free(pa);
1443 }
1444 
1445 static int
1446 proc_read_mem(struct thread *td, struct proc *p, vm_offset_t offset, void* buf,
1447     size_t len)
1448 {
1449 	struct iovec iov;
1450 	struct uio uio;
1451 
1452 	iov.iov_base = (caddr_t)buf;
1453 	iov.iov_len = len;
1454 	uio.uio_iov = &iov;
1455 	uio.uio_iovcnt = 1;
1456 	uio.uio_offset = offset;
1457 	uio.uio_resid = (ssize_t)len;
1458 	uio.uio_segflg = UIO_SYSSPACE;
1459 	uio.uio_rw = UIO_READ;
1460 	uio.uio_td = td;
1461 
1462 	return (proc_rwmem(p, &uio));
1463 }
1464 
1465 static int
1466 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1467     size_t len)
1468 {
1469 	size_t i;
1470 	int error;
1471 
1472 	error = proc_read_mem(td, p, (vm_offset_t)sptr, buf, len);
1473 	/*
1474 	 * Reading the chunk may validly return EFAULT if the string is shorter
1475 	 * than the chunk and is aligned at the end of the page, assuming the
1476 	 * next page is not mapped.  So if EFAULT is returned do a fallback to
1477 	 * one byte read loop.
1478 	 */
1479 	if (error == EFAULT) {
1480 		for (i = 0; i < len; i++, buf++, sptr++) {
1481 			error = proc_read_mem(td, p, (vm_offset_t)sptr, buf, 1);
1482 			if (error != 0)
1483 				return (error);
1484 			if (*buf == '\0')
1485 				break;
1486 		}
1487 		error = 0;
1488 	}
1489 	return (error);
1490 }
1491 
1492 #define PROC_AUXV_MAX	256	/* Safety limit on auxv size. */
1493 
1494 enum proc_vector_type {
1495 	PROC_ARG,
1496 	PROC_ENV,
1497 	PROC_AUX,
1498 };
1499 
1500 #ifdef COMPAT_FREEBSD32
1501 static int
1502 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1503     size_t *vsizep, enum proc_vector_type type)
1504 {
1505 	struct freebsd32_ps_strings pss;
1506 	Elf32_Auxinfo aux;
1507 	vm_offset_t vptr, ptr;
1508 	uint32_t *proc_vector32;
1509 	char **proc_vector;
1510 	size_t vsize, size;
1511 	int i, error;
1512 
1513 	error = proc_read_mem(td, p, (vm_offset_t)(p->p_sysent->sv_psstrings),
1514 	    &pss, sizeof(pss));
1515 	if (error != 0)
1516 		return (error);
1517 	switch (type) {
1518 	case PROC_ARG:
1519 		vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1520 		vsize = pss.ps_nargvstr;
1521 		if (vsize > ARG_MAX)
1522 			return (ENOEXEC);
1523 		size = vsize * sizeof(int32_t);
1524 		break;
1525 	case PROC_ENV:
1526 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1527 		vsize = pss.ps_nenvstr;
1528 		if (vsize > ARG_MAX)
1529 			return (ENOEXEC);
1530 		size = vsize * sizeof(int32_t);
1531 		break;
1532 	case PROC_AUX:
1533 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1534 		    (pss.ps_nenvstr + 1) * sizeof(int32_t);
1535 		if (vptr % 4 != 0)
1536 			return (ENOEXEC);
1537 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1538 			error = proc_read_mem(td, p, ptr, &aux, sizeof(aux));
1539 			if (error != 0)
1540 				return (error);
1541 			if (aux.a_type == AT_NULL)
1542 				break;
1543 			ptr += sizeof(aux);
1544 		}
1545 		if (aux.a_type != AT_NULL)
1546 			return (ENOEXEC);
1547 		vsize = i + 1;
1548 		size = vsize * sizeof(aux);
1549 		break;
1550 	default:
1551 		KASSERT(0, ("Wrong proc vector type: %d", type));
1552 		return (EINVAL);
1553 	}
1554 	proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1555 	error = proc_read_mem(td, p, vptr, proc_vector32, size);
1556 	if (error != 0)
1557 		goto done;
1558 	if (type == PROC_AUX) {
1559 		*proc_vectorp = (char **)proc_vector32;
1560 		*vsizep = vsize;
1561 		return (0);
1562 	}
1563 	proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1564 	for (i = 0; i < (int)vsize; i++)
1565 		proc_vector[i] = PTRIN(proc_vector32[i]);
1566 	*proc_vectorp = proc_vector;
1567 	*vsizep = vsize;
1568 done:
1569 	free(proc_vector32, M_TEMP);
1570 	return (error);
1571 }
1572 #endif
1573 
1574 static int
1575 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1576     size_t *vsizep, enum proc_vector_type type)
1577 {
1578 	struct ps_strings pss;
1579 	Elf_Auxinfo aux;
1580 	vm_offset_t vptr, ptr;
1581 	char **proc_vector;
1582 	size_t vsize, size;
1583 	int error, i;
1584 
1585 #ifdef COMPAT_FREEBSD32
1586 	if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1587 		return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1588 #endif
1589 	error = proc_read_mem(td, p, (vm_offset_t)(p->p_sysent->sv_psstrings),
1590 	    &pss, sizeof(pss));
1591 	if (error != 0)
1592 		return (error);
1593 	switch (type) {
1594 	case PROC_ARG:
1595 		vptr = (vm_offset_t)pss.ps_argvstr;
1596 		vsize = pss.ps_nargvstr;
1597 		if (vsize > ARG_MAX)
1598 			return (ENOEXEC);
1599 		size = vsize * sizeof(char *);
1600 		break;
1601 	case PROC_ENV:
1602 		vptr = (vm_offset_t)pss.ps_envstr;
1603 		vsize = pss.ps_nenvstr;
1604 		if (vsize > ARG_MAX)
1605 			return (ENOEXEC);
1606 		size = vsize * sizeof(char *);
1607 		break;
1608 	case PROC_AUX:
1609 		/*
1610 		 * The aux array is just above env array on the stack. Check
1611 		 * that the address is naturally aligned.
1612 		 */
1613 		vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1614 		    * sizeof(char *);
1615 #if __ELF_WORD_SIZE == 64
1616 		if (vptr % sizeof(uint64_t) != 0)
1617 #else
1618 		if (vptr % sizeof(uint32_t) != 0)
1619 #endif
1620 			return (ENOEXEC);
1621 		/*
1622 		 * We count the array size reading the aux vectors from the
1623 		 * stack until AT_NULL vector is returned.  So (to keep the code
1624 		 * simple) we read the process stack twice: the first time here
1625 		 * to find the size and the second time when copying the vectors
1626 		 * to the allocated proc_vector.
1627 		 */
1628 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1629 			error = proc_read_mem(td, p, ptr, &aux, sizeof(aux));
1630 			if (error != 0)
1631 				return (error);
1632 			if (aux.a_type == AT_NULL)
1633 				break;
1634 			ptr += sizeof(aux);
1635 		}
1636 		/*
1637 		 * If the PROC_AUXV_MAX entries are iterated over, and we have
1638 		 * not reached AT_NULL, it is most likely we are reading wrong
1639 		 * data: either the process doesn't have auxv array or data has
1640 		 * been modified. Return the error in this case.
1641 		 */
1642 		if (aux.a_type != AT_NULL)
1643 			return (ENOEXEC);
1644 		vsize = i + 1;
1645 		size = vsize * sizeof(aux);
1646 		break;
1647 	default:
1648 		KASSERT(0, ("Wrong proc vector type: %d", type));
1649 		return (EINVAL); /* In case we are built without INVARIANTS. */
1650 	}
1651 	proc_vector = malloc(size, M_TEMP, M_WAITOK);
1652 	if (proc_vector == NULL)
1653 		return (ENOMEM);
1654 	error = proc_read_mem(td, p, vptr, proc_vector, size);
1655 	if (error != 0) {
1656 		free(proc_vector, M_TEMP);
1657 		return (error);
1658 	}
1659 	*proc_vectorp = proc_vector;
1660 	*vsizep = vsize;
1661 
1662 	return (0);
1663 }
1664 
1665 #define GET_PS_STRINGS_CHUNK_SZ	256	/* Chunk size (bytes) for ps_strings operations. */
1666 
1667 static int
1668 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
1669     enum proc_vector_type type)
1670 {
1671 	size_t done, len, nchr, vsize;
1672 	int error, i;
1673 	char **proc_vector, *sptr;
1674 	char pss_string[GET_PS_STRINGS_CHUNK_SZ];
1675 
1676 	PROC_ASSERT_HELD(p);
1677 
1678 	/*
1679 	 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
1680 	 */
1681 	nchr = 2 * (PATH_MAX + ARG_MAX);
1682 
1683 	error = get_proc_vector(td, p, &proc_vector, &vsize, type);
1684 	if (error != 0)
1685 		return (error);
1686 	for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
1687 		/*
1688 		 * The program may have scribbled into its argv array, e.g. to
1689 		 * remove some arguments.  If that has happened, break out
1690 		 * before trying to read from NULL.
1691 		 */
1692 		if (proc_vector[i] == NULL)
1693 			break;
1694 		for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
1695 			error = proc_read_string(td, p, sptr, pss_string,
1696 			    sizeof(pss_string));
1697 			if (error != 0)
1698 				goto done;
1699 			len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
1700 			if (done + len >= nchr)
1701 				len = nchr - done - 1;
1702 			sbuf_bcat(sb, pss_string, len);
1703 			if (len != GET_PS_STRINGS_CHUNK_SZ)
1704 				break;
1705 			done += GET_PS_STRINGS_CHUNK_SZ;
1706 		}
1707 		sbuf_bcat(sb, "", 1);
1708 		done += len + 1;
1709 	}
1710 done:
1711 	free(proc_vector, M_TEMP);
1712 	return (error);
1713 }
1714 
1715 int
1716 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
1717 {
1718 
1719 	return (get_ps_strings(curthread, p, sb, PROC_ARG));
1720 }
1721 
1722 int
1723 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
1724 {
1725 
1726 	return (get_ps_strings(curthread, p, sb, PROC_ENV));
1727 }
1728 
1729 /*
1730  * This sysctl allows a process to retrieve the argument list or process
1731  * title for another process without groping around in the address space
1732  * of the other process.  It also allow a process to set its own "process
1733  * title to a string of its own choice.
1734  */
1735 static int
1736 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
1737 {
1738 	int *name = (int *)arg1;
1739 	u_int namelen = arg2;
1740 	struct pargs *newpa, *pa;
1741 	struct proc *p;
1742 	struct sbuf sb;
1743 	int flags, error = 0, error2;
1744 
1745 	if (namelen != 1)
1746 		return (EINVAL);
1747 
1748 	flags = PGET_CANSEE;
1749 	if (req->newptr != NULL)
1750 		flags |= PGET_ISCURRENT;
1751 	error = pget((pid_t)name[0], flags, &p);
1752 	if (error)
1753 		return (error);
1754 
1755 	pa = p->p_args;
1756 	if (pa != NULL) {
1757 		pargs_hold(pa);
1758 		PROC_UNLOCK(p);
1759 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
1760 		pargs_drop(pa);
1761 	} else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
1762 		_PHOLD(p);
1763 		PROC_UNLOCK(p);
1764 		sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1765 		error = proc_getargv(curthread, p, &sb);
1766 		error2 = sbuf_finish(&sb);
1767 		PRELE(p);
1768 		sbuf_delete(&sb);
1769 		if (error == 0 && error2 != 0)
1770 			error = error2;
1771 	} else {
1772 		PROC_UNLOCK(p);
1773 	}
1774 	if (error != 0 || req->newptr == NULL)
1775 		return (error);
1776 
1777 	if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit)
1778 		return (ENOMEM);
1779 	newpa = pargs_alloc(req->newlen);
1780 	error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
1781 	if (error != 0) {
1782 		pargs_free(newpa);
1783 		return (error);
1784 	}
1785 	PROC_LOCK(p);
1786 	pa = p->p_args;
1787 	p->p_args = newpa;
1788 	PROC_UNLOCK(p);
1789 	pargs_drop(pa);
1790 	return (0);
1791 }
1792 
1793 /*
1794  * This sysctl allows a process to retrieve environment of another process.
1795  */
1796 static int
1797 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
1798 {
1799 	int *name = (int *)arg1;
1800 	u_int namelen = arg2;
1801 	struct proc *p;
1802 	struct sbuf sb;
1803 	int error, error2;
1804 
1805 	if (namelen != 1)
1806 		return (EINVAL);
1807 
1808 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1809 	if (error != 0)
1810 		return (error);
1811 	if ((p->p_flag & P_SYSTEM) != 0) {
1812 		PRELE(p);
1813 		return (0);
1814 	}
1815 
1816 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1817 	error = proc_getenvv(curthread, p, &sb);
1818 	error2 = sbuf_finish(&sb);
1819 	PRELE(p);
1820 	sbuf_delete(&sb);
1821 	return (error != 0 ? error : error2);
1822 }
1823 
1824 /*
1825  * This sysctl allows a process to retrieve ELF auxiliary vector of
1826  * another process.
1827  */
1828 static int
1829 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
1830 {
1831 	int *name = (int *)arg1;
1832 	u_int namelen = arg2;
1833 	struct proc *p;
1834 	size_t vsize, size;
1835 	char **auxv;
1836 	int error;
1837 
1838 	if (namelen != 1)
1839 		return (EINVAL);
1840 
1841 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1842 	if (error != 0)
1843 		return (error);
1844 	if ((p->p_flag & P_SYSTEM) != 0) {
1845 		PRELE(p);
1846 		return (0);
1847 	}
1848 	error = get_proc_vector(curthread, p, &auxv, &vsize, PROC_AUX);
1849 	if (error == 0) {
1850 #ifdef COMPAT_FREEBSD32
1851 		if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1852 			size = vsize * sizeof(Elf32_Auxinfo);
1853 		else
1854 #endif
1855 		size = vsize * sizeof(Elf_Auxinfo);
1856 		PRELE(p);
1857 		error = SYSCTL_OUT(req, auxv, size);
1858 		free(auxv, M_TEMP);
1859 	} else {
1860 		PRELE(p);
1861 	}
1862 	return (error);
1863 }
1864 
1865 /*
1866  * This sysctl allows a process to retrieve the path of the executable for
1867  * itself or another process.
1868  */
1869 static int
1870 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
1871 {
1872 	pid_t *pidp = (pid_t *)arg1;
1873 	unsigned int arglen = arg2;
1874 	struct proc *p;
1875 	struct vnode *vp;
1876 	char *retbuf, *freebuf;
1877 	int error, vfslocked;
1878 
1879 	if (arglen != 1)
1880 		return (EINVAL);
1881 	if (*pidp == -1) {	/* -1 means this process */
1882 		p = req->td->td_proc;
1883 	} else {
1884 		error = pget(*pidp, PGET_CANSEE, &p);
1885 		if (error != 0)
1886 			return (error);
1887 	}
1888 
1889 	vp = p->p_textvp;
1890 	if (vp == NULL) {
1891 		if (*pidp != -1)
1892 			PROC_UNLOCK(p);
1893 		return (0);
1894 	}
1895 	vref(vp);
1896 	if (*pidp != -1)
1897 		PROC_UNLOCK(p);
1898 	error = vn_fullpath(req->td, vp, &retbuf, &freebuf);
1899 	vfslocked = VFS_LOCK_GIANT(vp->v_mount);
1900 	vrele(vp);
1901 	VFS_UNLOCK_GIANT(vfslocked);
1902 	if (error)
1903 		return (error);
1904 	error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
1905 	free(freebuf, M_TEMP);
1906 	return (error);
1907 }
1908 
1909 static int
1910 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
1911 {
1912 	struct proc *p;
1913 	char *sv_name;
1914 	int *name;
1915 	int namelen;
1916 	int error;
1917 
1918 	namelen = arg2;
1919 	if (namelen != 1)
1920 		return (EINVAL);
1921 
1922 	name = (int *)arg1;
1923 	error = pget((pid_t)name[0], PGET_CANSEE, &p);
1924 	if (error != 0)
1925 		return (error);
1926 	sv_name = p->p_sysent->sv_name;
1927 	PROC_UNLOCK(p);
1928 	return (sysctl_handle_string(oidp, sv_name, 0, req));
1929 }
1930 
1931 #ifdef KINFO_OVMENTRY_SIZE
1932 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
1933 #endif
1934 
1935 #ifdef COMPAT_FREEBSD7
1936 static int
1937 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
1938 {
1939 	vm_map_entry_t entry, tmp_entry;
1940 	unsigned int last_timestamp;
1941 	char *fullpath, *freepath;
1942 	struct kinfo_ovmentry *kve;
1943 	struct vattr va;
1944 	struct ucred *cred;
1945 	int error, *name;
1946 	struct vnode *vp;
1947 	struct proc *p;
1948 	vm_map_t map;
1949 	struct vmspace *vm;
1950 
1951 	name = (int *)arg1;
1952 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1953 	if (error != 0)
1954 		return (error);
1955 	vm = vmspace_acquire_ref(p);
1956 	if (vm == NULL) {
1957 		PRELE(p);
1958 		return (ESRCH);
1959 	}
1960 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
1961 
1962 	map = &vm->vm_map;
1963 	vm_map_lock_read(map);
1964 	for (entry = map->header.next; entry != &map->header;
1965 	    entry = entry->next) {
1966 		vm_object_t obj, tobj, lobj;
1967 		vm_offset_t addr;
1968 		int vfslocked;
1969 
1970 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
1971 			continue;
1972 
1973 		bzero(kve, sizeof(*kve));
1974 		kve->kve_structsize = sizeof(*kve);
1975 
1976 		kve->kve_private_resident = 0;
1977 		obj = entry->object.vm_object;
1978 		if (obj != NULL) {
1979 			VM_OBJECT_LOCK(obj);
1980 			if (obj->shadow_count == 1)
1981 				kve->kve_private_resident =
1982 				    obj->resident_page_count;
1983 		}
1984 		kve->kve_resident = 0;
1985 		addr = entry->start;
1986 		while (addr < entry->end) {
1987 			if (pmap_extract(map->pmap, addr))
1988 				kve->kve_resident++;
1989 			addr += PAGE_SIZE;
1990 		}
1991 
1992 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
1993 			if (tobj != obj)
1994 				VM_OBJECT_LOCK(tobj);
1995 			if (lobj != obj)
1996 				VM_OBJECT_UNLOCK(lobj);
1997 			lobj = tobj;
1998 		}
1999 
2000 		kve->kve_start = (void*)entry->start;
2001 		kve->kve_end = (void*)entry->end;
2002 		kve->kve_offset = (off_t)entry->offset;
2003 
2004 		if (entry->protection & VM_PROT_READ)
2005 			kve->kve_protection |= KVME_PROT_READ;
2006 		if (entry->protection & VM_PROT_WRITE)
2007 			kve->kve_protection |= KVME_PROT_WRITE;
2008 		if (entry->protection & VM_PROT_EXECUTE)
2009 			kve->kve_protection |= KVME_PROT_EXEC;
2010 
2011 		if (entry->eflags & MAP_ENTRY_COW)
2012 			kve->kve_flags |= KVME_FLAG_COW;
2013 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2014 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2015 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2016 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2017 
2018 		last_timestamp = map->timestamp;
2019 		vm_map_unlock_read(map);
2020 
2021 		kve->kve_fileid = 0;
2022 		kve->kve_fsid = 0;
2023 		freepath = NULL;
2024 		fullpath = "";
2025 		if (lobj) {
2026 			vp = NULL;
2027 			switch (lobj->type) {
2028 			case OBJT_DEFAULT:
2029 				kve->kve_type = KVME_TYPE_DEFAULT;
2030 				break;
2031 			case OBJT_VNODE:
2032 				kve->kve_type = KVME_TYPE_VNODE;
2033 				vp = lobj->handle;
2034 				vref(vp);
2035 				break;
2036 			case OBJT_SWAP:
2037 				kve->kve_type = KVME_TYPE_SWAP;
2038 				break;
2039 			case OBJT_DEVICE:
2040 				kve->kve_type = KVME_TYPE_DEVICE;
2041 				break;
2042 			case OBJT_PHYS:
2043 				kve->kve_type = KVME_TYPE_PHYS;
2044 				break;
2045 			case OBJT_DEAD:
2046 				kve->kve_type = KVME_TYPE_DEAD;
2047 				break;
2048 			case OBJT_SG:
2049 				kve->kve_type = KVME_TYPE_SG;
2050 				break;
2051 			default:
2052 				kve->kve_type = KVME_TYPE_UNKNOWN;
2053 				break;
2054 			}
2055 			if (lobj != obj)
2056 				VM_OBJECT_UNLOCK(lobj);
2057 
2058 			kve->kve_ref_count = obj->ref_count;
2059 			kve->kve_shadow_count = obj->shadow_count;
2060 			VM_OBJECT_UNLOCK(obj);
2061 			if (vp != NULL) {
2062 				vn_fullpath(curthread, vp, &fullpath,
2063 				    &freepath);
2064 				cred = curthread->td_ucred;
2065 				vfslocked = VFS_LOCK_GIANT(vp->v_mount);
2066 				vn_lock(vp, LK_SHARED | LK_RETRY);
2067 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2068 					kve->kve_fileid = va.va_fileid;
2069 					kve->kve_fsid = va.va_fsid;
2070 				}
2071 				vput(vp);
2072 				VFS_UNLOCK_GIANT(vfslocked);
2073 			}
2074 		} else {
2075 			kve->kve_type = KVME_TYPE_NONE;
2076 			kve->kve_ref_count = 0;
2077 			kve->kve_shadow_count = 0;
2078 		}
2079 
2080 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2081 		if (freepath != NULL)
2082 			free(freepath, M_TEMP);
2083 
2084 		error = SYSCTL_OUT(req, kve, sizeof(*kve));
2085 		vm_map_lock_read(map);
2086 		if (error)
2087 			break;
2088 		if (last_timestamp != map->timestamp) {
2089 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2090 			entry = tmp_entry;
2091 		}
2092 	}
2093 	vm_map_unlock_read(map);
2094 	vmspace_free(vm);
2095 	PRELE(p);
2096 	free(kve, M_TEMP);
2097 	return (error);
2098 }
2099 #endif	/* COMPAT_FREEBSD7 */
2100 
2101 #ifdef KINFO_VMENTRY_SIZE
2102 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2103 #endif
2104 
2105 static int
2106 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2107 {
2108 	vm_map_entry_t entry, tmp_entry;
2109 	unsigned int last_timestamp;
2110 	char *fullpath, *freepath;
2111 	struct kinfo_vmentry *kve;
2112 	struct vattr va;
2113 	struct ucred *cred;
2114 	int error, *name;
2115 	struct vnode *vp;
2116 	struct proc *p;
2117 	struct vmspace *vm;
2118 	vm_map_t map;
2119 
2120 	name = (int *)arg1;
2121 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2122 	if (error != 0)
2123 		return (error);
2124 	vm = vmspace_acquire_ref(p);
2125 	if (vm == NULL) {
2126 		PRELE(p);
2127 		return (ESRCH);
2128 	}
2129 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2130 
2131 	map = &vm->vm_map;
2132 	vm_map_lock_read(map);
2133 	for (entry = map->header.next; entry != &map->header;
2134 	    entry = entry->next) {
2135 		vm_object_t obj, tobj, lobj;
2136 		vm_offset_t addr;
2137 		vm_paddr_t locked_pa;
2138 		int vfslocked, mincoreinfo;
2139 
2140 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2141 			continue;
2142 
2143 		bzero(kve, sizeof(*kve));
2144 
2145 		kve->kve_private_resident = 0;
2146 		obj = entry->object.vm_object;
2147 		if (obj != NULL) {
2148 			VM_OBJECT_LOCK(obj);
2149 			if (obj->shadow_count == 1)
2150 				kve->kve_private_resident =
2151 				    obj->resident_page_count;
2152 		}
2153 		kve->kve_resident = 0;
2154 		addr = entry->start;
2155 		while (addr < entry->end) {
2156 			locked_pa = 0;
2157 			mincoreinfo = pmap_mincore(map->pmap, addr, &locked_pa);
2158 			if (locked_pa != 0)
2159 				vm_page_unlock(PHYS_TO_VM_PAGE(locked_pa));
2160 			if (mincoreinfo & MINCORE_INCORE)
2161 				kve->kve_resident++;
2162 			if (mincoreinfo & MINCORE_SUPER)
2163 				kve->kve_flags |= KVME_FLAG_SUPER;
2164 			addr += PAGE_SIZE;
2165 		}
2166 
2167 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2168 			if (tobj != obj)
2169 				VM_OBJECT_LOCK(tobj);
2170 			if (lobj != obj)
2171 				VM_OBJECT_UNLOCK(lobj);
2172 			lobj = tobj;
2173 		}
2174 
2175 		kve->kve_start = entry->start;
2176 		kve->kve_end = entry->end;
2177 		kve->kve_offset = entry->offset;
2178 
2179 		if (entry->protection & VM_PROT_READ)
2180 			kve->kve_protection |= KVME_PROT_READ;
2181 		if (entry->protection & VM_PROT_WRITE)
2182 			kve->kve_protection |= KVME_PROT_WRITE;
2183 		if (entry->protection & VM_PROT_EXECUTE)
2184 			kve->kve_protection |= KVME_PROT_EXEC;
2185 
2186 		if (entry->eflags & MAP_ENTRY_COW)
2187 			kve->kve_flags |= KVME_FLAG_COW;
2188 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2189 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2190 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2191 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2192 
2193 		last_timestamp = map->timestamp;
2194 		vm_map_unlock_read(map);
2195 
2196 		freepath = NULL;
2197 		fullpath = "";
2198 		if (lobj) {
2199 			vp = NULL;
2200 			switch (lobj->type) {
2201 			case OBJT_DEFAULT:
2202 				kve->kve_type = KVME_TYPE_DEFAULT;
2203 				break;
2204 			case OBJT_VNODE:
2205 				kve->kve_type = KVME_TYPE_VNODE;
2206 				vp = lobj->handle;
2207 				vref(vp);
2208 				break;
2209 			case OBJT_SWAP:
2210 				kve->kve_type = KVME_TYPE_SWAP;
2211 				break;
2212 			case OBJT_DEVICE:
2213 				kve->kve_type = KVME_TYPE_DEVICE;
2214 				break;
2215 			case OBJT_PHYS:
2216 				kve->kve_type = KVME_TYPE_PHYS;
2217 				break;
2218 			case OBJT_DEAD:
2219 				kve->kve_type = KVME_TYPE_DEAD;
2220 				break;
2221 			case OBJT_SG:
2222 				kve->kve_type = KVME_TYPE_SG;
2223 				break;
2224 			default:
2225 				kve->kve_type = KVME_TYPE_UNKNOWN;
2226 				break;
2227 			}
2228 			if (lobj != obj)
2229 				VM_OBJECT_UNLOCK(lobj);
2230 
2231 			kve->kve_ref_count = obj->ref_count;
2232 			kve->kve_shadow_count = obj->shadow_count;
2233 			VM_OBJECT_UNLOCK(obj);
2234 			if (vp != NULL) {
2235 				vn_fullpath(curthread, vp, &fullpath,
2236 				    &freepath);
2237 				kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2238 				cred = curthread->td_ucred;
2239 				vfslocked = VFS_LOCK_GIANT(vp->v_mount);
2240 				vn_lock(vp, LK_SHARED | LK_RETRY);
2241 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2242 					kve->kve_vn_fileid = va.va_fileid;
2243 					kve->kve_vn_fsid = va.va_fsid;
2244 					kve->kve_vn_mode =
2245 					    MAKEIMODE(va.va_type, va.va_mode);
2246 					kve->kve_vn_size = va.va_size;
2247 					kve->kve_vn_rdev = va.va_rdev;
2248 					kve->kve_status = KF_ATTR_VALID;
2249 				}
2250 				vput(vp);
2251 				VFS_UNLOCK_GIANT(vfslocked);
2252 			}
2253 		} else {
2254 			kve->kve_type = KVME_TYPE_NONE;
2255 			kve->kve_ref_count = 0;
2256 			kve->kve_shadow_count = 0;
2257 		}
2258 
2259 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2260 		if (freepath != NULL)
2261 			free(freepath, M_TEMP);
2262 
2263 		/* Pack record size down */
2264 		kve->kve_structsize = offsetof(struct kinfo_vmentry, kve_path) +
2265 		    strlen(kve->kve_path) + 1;
2266 		kve->kve_structsize = roundup(kve->kve_structsize,
2267 		    sizeof(uint64_t));
2268 		error = SYSCTL_OUT(req, kve, kve->kve_structsize);
2269 		vm_map_lock_read(map);
2270 		if (error)
2271 			break;
2272 		if (last_timestamp != map->timestamp) {
2273 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2274 			entry = tmp_entry;
2275 		}
2276 	}
2277 	vm_map_unlock_read(map);
2278 	vmspace_free(vm);
2279 	PRELE(p);
2280 	free(kve, M_TEMP);
2281 	return (error);
2282 }
2283 
2284 #if defined(STACK) || defined(DDB)
2285 static int
2286 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2287 {
2288 	struct kinfo_kstack *kkstp;
2289 	int error, i, *name, numthreads;
2290 	lwpid_t *lwpidarray;
2291 	struct thread *td;
2292 	struct stack *st;
2293 	struct sbuf sb;
2294 	struct proc *p;
2295 
2296 	name = (int *)arg1;
2297 	error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2298 	if (error != 0)
2299 		return (error);
2300 
2301 	kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2302 	st = stack_create();
2303 
2304 	lwpidarray = NULL;
2305 	numthreads = 0;
2306 	PROC_LOCK(p);
2307 repeat:
2308 	if (numthreads < p->p_numthreads) {
2309 		if (lwpidarray != NULL) {
2310 			free(lwpidarray, M_TEMP);
2311 			lwpidarray = NULL;
2312 		}
2313 		numthreads = p->p_numthreads;
2314 		PROC_UNLOCK(p);
2315 		lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2316 		    M_WAITOK | M_ZERO);
2317 		PROC_LOCK(p);
2318 		goto repeat;
2319 	}
2320 	i = 0;
2321 
2322 	/*
2323 	 * XXXRW: During the below loop, execve(2) and countless other sorts
2324 	 * of changes could have taken place.  Should we check to see if the
2325 	 * vmspace has been replaced, or the like, in order to prevent
2326 	 * giving a snapshot that spans, say, execve(2), with some threads
2327 	 * before and some after?  Among other things, the credentials could
2328 	 * have changed, in which case the right to extract debug info might
2329 	 * no longer be assured.
2330 	 */
2331 	FOREACH_THREAD_IN_PROC(p, td) {
2332 		KASSERT(i < numthreads,
2333 		    ("sysctl_kern_proc_kstack: numthreads"));
2334 		lwpidarray[i] = td->td_tid;
2335 		i++;
2336 	}
2337 	numthreads = i;
2338 	for (i = 0; i < numthreads; i++) {
2339 		td = thread_find(p, lwpidarray[i]);
2340 		if (td == NULL) {
2341 			continue;
2342 		}
2343 		bzero(kkstp, sizeof(*kkstp));
2344 		(void)sbuf_new(&sb, kkstp->kkst_trace,
2345 		    sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2346 		thread_lock(td);
2347 		kkstp->kkst_tid = td->td_tid;
2348 		if (TD_IS_SWAPPED(td))
2349 			kkstp->kkst_state = KKST_STATE_SWAPPED;
2350 		else if (TD_IS_RUNNING(td))
2351 			kkstp->kkst_state = KKST_STATE_RUNNING;
2352 		else {
2353 			kkstp->kkst_state = KKST_STATE_STACKOK;
2354 			stack_save_td(st, td);
2355 		}
2356 		thread_unlock(td);
2357 		PROC_UNLOCK(p);
2358 		stack_sbuf_print(&sb, st);
2359 		sbuf_finish(&sb);
2360 		sbuf_delete(&sb);
2361 		error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2362 		PROC_LOCK(p);
2363 		if (error)
2364 			break;
2365 	}
2366 	_PRELE(p);
2367 	PROC_UNLOCK(p);
2368 	if (lwpidarray != NULL)
2369 		free(lwpidarray, M_TEMP);
2370 	stack_destroy(st);
2371 	free(kkstp, M_TEMP);
2372 	return (error);
2373 }
2374 #endif
2375 
2376 /*
2377  * This sysctl allows a process to retrieve the full list of groups from
2378  * itself or another process.
2379  */
2380 static int
2381 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2382 {
2383 	pid_t *pidp = (pid_t *)arg1;
2384 	unsigned int arglen = arg2;
2385 	struct proc *p;
2386 	struct ucred *cred;
2387 	int error;
2388 
2389 	if (arglen != 1)
2390 		return (EINVAL);
2391 	if (*pidp == -1) {	/* -1 means this process */
2392 		p = req->td->td_proc;
2393 	} else {
2394 		error = pget(*pidp, PGET_CANSEE, &p);
2395 		if (error != 0)
2396 			return (error);
2397 	}
2398 
2399 	cred = crhold(p->p_ucred);
2400 	if (*pidp != -1)
2401 		PROC_UNLOCK(p);
2402 
2403 	error = SYSCTL_OUT(req, cred->cr_groups,
2404 	    cred->cr_ngroups * sizeof(gid_t));
2405 	crfree(cred);
2406 	return (error);
2407 }
2408 
2409 /*
2410  * This sysctl allows a process to retrieve or/and set the resource limit for
2411  * another process.
2412  */
2413 static int
2414 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2415 {
2416 	int *name = (int *)arg1;
2417 	u_int namelen = arg2;
2418 	struct rlimit rlim;
2419 	struct proc *p;
2420 	u_int which;
2421 	int flags, error;
2422 
2423 	if (namelen != 2)
2424 		return (EINVAL);
2425 
2426 	which = (u_int)name[1];
2427 	if (which >= RLIM_NLIMITS)
2428 		return (EINVAL);
2429 
2430 	if (req->newptr != NULL && req->newlen != sizeof(rlim))
2431 		return (EINVAL);
2432 
2433 	flags = PGET_HOLD | PGET_NOTWEXIT;
2434 	if (req->newptr != NULL)
2435 		flags |= PGET_CANDEBUG;
2436 	else
2437 		flags |= PGET_CANSEE;
2438 	error = pget((pid_t)name[0], flags, &p);
2439 	if (error != 0)
2440 		return (error);
2441 
2442 	/*
2443 	 * Retrieve limit.
2444 	 */
2445 	if (req->oldptr != NULL) {
2446 		PROC_LOCK(p);
2447 		lim_rlimit(p, which, &rlim);
2448 		PROC_UNLOCK(p);
2449 	}
2450 	error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
2451 	if (error != 0)
2452 		goto errout;
2453 
2454 	/*
2455 	 * Set limit.
2456 	 */
2457 	if (req->newptr != NULL) {
2458 		error = SYSCTL_IN(req, &rlim, sizeof(rlim));
2459 		if (error == 0)
2460 			error = kern_proc_setrlimit(curthread, p, which, &rlim);
2461 	}
2462 
2463 errout:
2464 	PRELE(p);
2465 	return (error);
2466 }
2467 
2468 /*
2469  * This sysctl allows a process to retrieve ps_strings structure location of
2470  * another process.
2471  */
2472 static int
2473 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
2474 {
2475 	int *name = (int *)arg1;
2476 	u_int namelen = arg2;
2477 	struct proc *p;
2478 	vm_offset_t ps_strings;
2479 	int error;
2480 #ifdef COMPAT_FREEBSD32
2481 	uint32_t ps_strings32;
2482 #endif
2483 
2484 	if (namelen != 1)
2485 		return (EINVAL);
2486 
2487 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2488 	if (error != 0)
2489 		return (error);
2490 #ifdef COMPAT_FREEBSD32
2491 	if ((req->flags & SCTL_MASK32) != 0) {
2492 		/*
2493 		 * We return 0 if the 32 bit emulation request is for a 64 bit
2494 		 * process.
2495 		 */
2496 		ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
2497 		    PTROUT(p->p_sysent->sv_psstrings) : 0;
2498 		PROC_UNLOCK(p);
2499 		error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
2500 		return (error);
2501 	}
2502 #endif
2503 	ps_strings = p->p_sysent->sv_psstrings;
2504 	PROC_UNLOCK(p);
2505 	error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
2506 	return (error);
2507 }
2508 
2509 /*
2510  * This sysctl allows a process to retrieve umask of another process.
2511  */
2512 static int
2513 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
2514 {
2515 	int *name = (int *)arg1;
2516 	u_int namelen = arg2;
2517 	struct proc *p;
2518 	int error;
2519 	u_short fd_cmask;
2520 
2521 	if (namelen != 1)
2522 		return (EINVAL);
2523 
2524 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2525 	if (error != 0)
2526 		return (error);
2527 
2528 	FILEDESC_SLOCK(p->p_fd);
2529 	fd_cmask = p->p_fd->fd_cmask;
2530 	FILEDESC_SUNLOCK(p->p_fd);
2531 	PRELE(p);
2532 	error = SYSCTL_OUT(req, &fd_cmask, sizeof(fd_cmask));
2533 	return (error);
2534 }
2535 
2536 /*
2537  * This sysctl allows a process to set and retrieve binary osreldate of
2538  * another process.
2539  */
2540 static int
2541 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
2542 {
2543 	int *name = (int *)arg1;
2544 	u_int namelen = arg2;
2545 	struct proc *p;
2546 	int flags, error, osrel;
2547 
2548 	if (namelen != 1)
2549 		return (EINVAL);
2550 
2551 	if (req->newptr != NULL && req->newlen != sizeof(osrel))
2552 		return (EINVAL);
2553 
2554 	flags = PGET_HOLD | PGET_NOTWEXIT;
2555 	if (req->newptr != NULL)
2556 		flags |= PGET_CANDEBUG;
2557 	else
2558 		flags |= PGET_CANSEE;
2559 	error = pget((pid_t)name[0], flags, &p);
2560 	if (error != 0)
2561 		return (error);
2562 
2563 	error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
2564 	if (error != 0)
2565 		goto errout;
2566 
2567 	if (req->newptr != NULL) {
2568 		error = SYSCTL_IN(req, &osrel, sizeof(osrel));
2569 		if (error != 0)
2570 			goto errout;
2571 		if (osrel < 0) {
2572 			error = EINVAL;
2573 			goto errout;
2574 		}
2575 		p->p_osrel = osrel;
2576 	}
2577 errout:
2578 	PRELE(p);
2579 	return (error);
2580 }
2581 
2582 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD,  0, "Process table");
2583 
2584 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
2585 	CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
2586 	"Return entire process table");
2587 
2588 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2589 	sysctl_kern_proc, "Process table");
2590 
2591 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
2592 	sysctl_kern_proc, "Process table");
2593 
2594 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2595 	sysctl_kern_proc, "Process table");
2596 
2597 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
2598 	CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2599 
2600 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
2601 	sysctl_kern_proc, "Process table");
2602 
2603 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2604 	sysctl_kern_proc, "Process table");
2605 
2606 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2607 	sysctl_kern_proc, "Process table");
2608 
2609 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2610 	sysctl_kern_proc, "Process table");
2611 
2612 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
2613 	sysctl_kern_proc, "Return process table, no threads");
2614 
2615 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
2616 	CTLFLAG_RW | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
2617 	sysctl_kern_proc_args, "Process argument list");
2618 
2619 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
2620 	sysctl_kern_proc_env, "Process environment");
2621 
2622 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
2623 	CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
2624 
2625 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
2626 	CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
2627 
2628 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
2629 	CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
2630 	"Process syscall vector name (ABI type)");
2631 
2632 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
2633 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2634 
2635 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
2636 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2637 
2638 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
2639 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2640 
2641 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
2642 	sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2643 
2644 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
2645 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2646 
2647 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
2648 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2649 
2650 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
2651 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2652 
2653 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
2654 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2655 
2656 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
2657 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
2658 	"Return process table, no threads");
2659 
2660 #ifdef COMPAT_FREEBSD7
2661 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
2662 	CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
2663 #endif
2664 
2665 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
2666 	CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
2667 
2668 #if defined(STACK) || defined(DDB)
2669 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
2670 	CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
2671 #endif
2672 
2673 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
2674 	CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
2675 
2676 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
2677 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
2678 	"Process resource limits");
2679 
2680 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
2681 	CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
2682 	"Process ps_strings location");
2683 
2684 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
2685 	CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
2686 
2687 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
2688 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
2689 	"Process binary osreldate");
2690