xref: /freebsd/sys/kern/kern_proc.c (revision b0ee263dbd3552d5b1776be0efc1c2d105f873b1)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1982, 1986, 1989, 1991, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)kern_proc.c	8.7 (Berkeley) 2/14/95
32  */
33 
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
36 
37 #include "opt_ddb.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/bitstring.h>
45 #include <sys/elf.h>
46 #include <sys/eventhandler.h>
47 #include <sys/exec.h>
48 #include <sys/jail.h>
49 #include <sys/kernel.h>
50 #include <sys/limits.h>
51 #include <sys/lock.h>
52 #include <sys/loginclass.h>
53 #include <sys/malloc.h>
54 #include <sys/mman.h>
55 #include <sys/mount.h>
56 #include <sys/mutex.h>
57 #include <sys/proc.h>
58 #include <sys/ptrace.h>
59 #include <sys/refcount.h>
60 #include <sys/resourcevar.h>
61 #include <sys/rwlock.h>
62 #include <sys/sbuf.h>
63 #include <sys/sysent.h>
64 #include <sys/sched.h>
65 #include <sys/smp.h>
66 #include <sys/stack.h>
67 #include <sys/stat.h>
68 #include <sys/sysctl.h>
69 #include <sys/filedesc.h>
70 #include <sys/tty.h>
71 #include <sys/signalvar.h>
72 #include <sys/sdt.h>
73 #include <sys/sx.h>
74 #include <sys/user.h>
75 #include <sys/vnode.h>
76 #include <sys/wait.h>
77 
78 #ifdef DDB
79 #include <ddb/ddb.h>
80 #endif
81 
82 #include <vm/vm.h>
83 #include <vm/vm_param.h>
84 #include <vm/vm_extern.h>
85 #include <vm/pmap.h>
86 #include <vm/vm_map.h>
87 #include <vm/vm_object.h>
88 #include <vm/vm_page.h>
89 #include <vm/uma.h>
90 
91 #ifdef COMPAT_FREEBSD32
92 #include <compat/freebsd32/freebsd32.h>
93 #include <compat/freebsd32/freebsd32_util.h>
94 #endif
95 
96 SDT_PROVIDER_DEFINE(proc);
97 
98 MALLOC_DEFINE(M_PGRP, "pgrp", "process group header");
99 MALLOC_DEFINE(M_SESSION, "session", "session header");
100 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
101 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
102 
103 static void doenterpgrp(struct proc *, struct pgrp *);
104 static void orphanpg(struct pgrp *pg);
105 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp);
106 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp);
107 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp,
108     int preferthread);
109 static void pgadjustjobc(struct pgrp *pgrp, int entering);
110 static void pgdelete(struct pgrp *);
111 static int proc_ctor(void *mem, int size, void *arg, int flags);
112 static void proc_dtor(void *mem, int size, void *arg);
113 static int proc_init(void *mem, int size, int flags);
114 static void proc_fini(void *mem, int size);
115 static void pargs_free(struct pargs *pa);
116 
117 /*
118  * Other process lists
119  */
120 struct pidhashhead *pidhashtbl;
121 struct sx *pidhashtbl_lock;
122 u_long pidhash;
123 u_long pidhashlock;
124 struct pgrphashhead *pgrphashtbl;
125 u_long pgrphash;
126 struct proclist allproc;
127 struct sx __exclusive_cache_line allproc_lock;
128 struct sx __exclusive_cache_line proctree_lock;
129 struct mtx __exclusive_cache_line ppeers_lock;
130 struct mtx __exclusive_cache_line procid_lock;
131 uma_zone_t proc_zone;
132 
133 /*
134  * The offset of various fields in struct proc and struct thread.
135  * These are used by kernel debuggers to enumerate kernel threads and
136  * processes.
137  */
138 const int proc_off_p_pid = offsetof(struct proc, p_pid);
139 const int proc_off_p_comm = offsetof(struct proc, p_comm);
140 const int proc_off_p_list = offsetof(struct proc, p_list);
141 const int proc_off_p_threads = offsetof(struct proc, p_threads);
142 const int thread_off_td_tid = offsetof(struct thread, td_tid);
143 const int thread_off_td_name = offsetof(struct thread, td_name);
144 const int thread_off_td_oncpu = offsetof(struct thread, td_oncpu);
145 const int thread_off_td_pcb = offsetof(struct thread, td_pcb);
146 const int thread_off_td_plist = offsetof(struct thread, td_plist);
147 
148 EVENTHANDLER_LIST_DEFINE(process_ctor);
149 EVENTHANDLER_LIST_DEFINE(process_dtor);
150 EVENTHANDLER_LIST_DEFINE(process_init);
151 EVENTHANDLER_LIST_DEFINE(process_fini);
152 EVENTHANDLER_LIST_DEFINE(process_exit);
153 EVENTHANDLER_LIST_DEFINE(process_fork);
154 EVENTHANDLER_LIST_DEFINE(process_exec);
155 
156 int kstack_pages = KSTACK_PAGES;
157 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0,
158     "Kernel stack size in pages");
159 static int vmmap_skip_res_cnt = 0;
160 SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW,
161     &vmmap_skip_res_cnt, 0,
162     "Skip calculation of the pages resident count in kern.proc.vmmap");
163 
164 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
165 #ifdef COMPAT_FREEBSD32
166 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
167 #endif
168 
169 /*
170  * Initialize global process hashing structures.
171  */
172 void
173 procinit(void)
174 {
175 	u_long i;
176 
177 	sx_init(&allproc_lock, "allproc");
178 	sx_init(&proctree_lock, "proctree");
179 	mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
180 	mtx_init(&procid_lock, "procid", NULL, MTX_DEF);
181 	LIST_INIT(&allproc);
182 	pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
183 	pidhashlock = (pidhash + 1) / 64;
184 	if (pidhashlock > 0)
185 		pidhashlock--;
186 	pidhashtbl_lock = malloc(sizeof(*pidhashtbl_lock) * (pidhashlock + 1),
187 	    M_PROC, M_WAITOK | M_ZERO);
188 	for (i = 0; i < pidhashlock + 1; i++)
189 		sx_init_flags(&pidhashtbl_lock[i], "pidhash", SX_DUPOK);
190 	pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
191 	proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
192 	    proc_ctor, proc_dtor, proc_init, proc_fini,
193 	    UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
194 	uihashinit();
195 }
196 
197 /*
198  * Prepare a proc for use.
199  */
200 static int
201 proc_ctor(void *mem, int size, void *arg, int flags)
202 {
203 	struct proc *p;
204 	struct thread *td;
205 
206 	p = (struct proc *)mem;
207 	EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
208 	td = FIRST_THREAD_IN_PROC(p);
209 	if (td != NULL) {
210 		/* Make sure all thread constructors are executed */
211 		EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
212 	}
213 	return (0);
214 }
215 
216 /*
217  * Reclaim a proc after use.
218  */
219 static void
220 proc_dtor(void *mem, int size, void *arg)
221 {
222 	struct proc *p;
223 	struct thread *td;
224 
225 	/* INVARIANTS checks go here */
226 	p = (struct proc *)mem;
227 	td = FIRST_THREAD_IN_PROC(p);
228 	if (td != NULL) {
229 #ifdef INVARIANTS
230 		KASSERT((p->p_numthreads == 1),
231 		    ("bad number of threads in exiting process"));
232 		KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
233 #endif
234 		/* Free all OSD associated to this thread. */
235 		osd_thread_exit(td);
236 		td_softdep_cleanup(td);
237 		MPASS(td->td_su == NULL);
238 
239 		/* Make sure all thread destructors are executed */
240 		EVENTHANDLER_DIRECT_INVOKE(thread_dtor, td);
241 	}
242 	EVENTHANDLER_DIRECT_INVOKE(process_dtor, p);
243 	if (p->p_ksi != NULL)
244 		KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
245 }
246 
247 /*
248  * Initialize type-stable parts of a proc (when newly created).
249  */
250 static int
251 proc_init(void *mem, int size, int flags)
252 {
253 	struct proc *p;
254 
255 	p = (struct proc *)mem;
256 	mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK | MTX_NEW);
257 	mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_NEW);
258 	mtx_init(&p->p_statmtx, "pstatl", NULL, MTX_SPIN | MTX_NEW);
259 	mtx_init(&p->p_itimmtx, "pitiml", NULL, MTX_SPIN | MTX_NEW);
260 	mtx_init(&p->p_profmtx, "pprofl", NULL, MTX_SPIN | MTX_NEW);
261 	cv_init(&p->p_pwait, "ppwait");
262 	TAILQ_INIT(&p->p_threads);	     /* all threads in proc */
263 	EVENTHANDLER_DIRECT_INVOKE(process_init, p);
264 	p->p_stats = pstats_alloc();
265 	p->p_pgrp = NULL;
266 	return (0);
267 }
268 
269 /*
270  * UMA should ensure that this function is never called.
271  * Freeing a proc structure would violate type stability.
272  */
273 static void
274 proc_fini(void *mem, int size)
275 {
276 #ifdef notnow
277 	struct proc *p;
278 
279 	p = (struct proc *)mem;
280 	EVENTHANDLER_DIRECT_INVOKE(process_fini, p);
281 	pstats_free(p->p_stats);
282 	thread_free(FIRST_THREAD_IN_PROC(p));
283 	mtx_destroy(&p->p_mtx);
284 	if (p->p_ksi != NULL)
285 		ksiginfo_free(p->p_ksi);
286 #else
287 	panic("proc reclaimed");
288 #endif
289 }
290 
291 /*
292  * PID space management.
293  *
294  * These bitmaps are used by fork_findpid.
295  */
296 bitstr_t bit_decl(proc_id_pidmap, PID_MAX);
297 bitstr_t bit_decl(proc_id_grpidmap, PID_MAX);
298 bitstr_t bit_decl(proc_id_sessidmap, PID_MAX);
299 bitstr_t bit_decl(proc_id_reapmap, PID_MAX);
300 
301 static bitstr_t *proc_id_array[] = {
302 	proc_id_pidmap,
303 	proc_id_grpidmap,
304 	proc_id_sessidmap,
305 	proc_id_reapmap,
306 };
307 
308 void
309 proc_id_set(int type, pid_t id)
310 {
311 
312 	KASSERT(type >= 0 && type < nitems(proc_id_array),
313 	    ("invalid type %d\n", type));
314 	mtx_lock(&procid_lock);
315 	KASSERT(bit_test(proc_id_array[type], id) == 0,
316 	    ("bit %d already set in %d\n", id, type));
317 	bit_set(proc_id_array[type], id);
318 	mtx_unlock(&procid_lock);
319 }
320 
321 void
322 proc_id_set_cond(int type, pid_t id)
323 {
324 
325 	KASSERT(type >= 0 && type < nitems(proc_id_array),
326 	    ("invalid type %d\n", type));
327 	if (bit_test(proc_id_array[type], id))
328 		return;
329 	mtx_lock(&procid_lock);
330 	bit_set(proc_id_array[type], id);
331 	mtx_unlock(&procid_lock);
332 }
333 
334 void
335 proc_id_clear(int type, pid_t id)
336 {
337 
338 	KASSERT(type >= 0 && type < nitems(proc_id_array),
339 	    ("invalid type %d\n", type));
340 	mtx_lock(&procid_lock);
341 	KASSERT(bit_test(proc_id_array[type], id) != 0,
342 	    ("bit %d not set in %d\n", id, type));
343 	bit_clear(proc_id_array[type], id);
344 	mtx_unlock(&procid_lock);
345 }
346 
347 /*
348  * Is p an inferior of the current process?
349  */
350 int
351 inferior(struct proc *p)
352 {
353 
354 	sx_assert(&proctree_lock, SX_LOCKED);
355 	PROC_LOCK_ASSERT(p, MA_OWNED);
356 	for (; p != curproc; p = proc_realparent(p)) {
357 		if (p->p_pid == 0)
358 			return (0);
359 	}
360 	return (1);
361 }
362 
363 /*
364  * Shared lock all the pid hash lists.
365  */
366 void
367 pidhash_slockall(void)
368 {
369 	u_long i;
370 
371 	for (i = 0; i < pidhashlock + 1; i++)
372 		sx_slock(&pidhashtbl_lock[i]);
373 }
374 
375 /*
376  * Shared unlock all the pid hash lists.
377  */
378 void
379 pidhash_sunlockall(void)
380 {
381 	u_long i;
382 
383 	for (i = 0; i < pidhashlock + 1; i++)
384 		sx_sunlock(&pidhashtbl_lock[i]);
385 }
386 
387 /*
388  * Similar to pfind_any(), this function finds zombies.
389  */
390 struct proc *
391 pfind_any_locked(pid_t pid)
392 {
393 	struct proc *p;
394 
395 	sx_assert(PIDHASHLOCK(pid), SX_LOCKED);
396 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
397 		if (p->p_pid == pid) {
398 			PROC_LOCK(p);
399 			if (p->p_state == PRS_NEW) {
400 				PROC_UNLOCK(p);
401 				p = NULL;
402 			}
403 			break;
404 		}
405 	}
406 	return (p);
407 }
408 
409 /*
410  * Locate a process by number.
411  *
412  * By not returning processes in the PRS_NEW state, we allow callers to avoid
413  * testing for that condition to avoid dereferencing p_ucred, et al.
414  */
415 static __always_inline struct proc *
416 _pfind(pid_t pid, bool zombie)
417 {
418 	struct proc *p;
419 
420 	p = curproc;
421 	if (p->p_pid == pid) {
422 		PROC_LOCK(p);
423 		return (p);
424 	}
425 	sx_slock(PIDHASHLOCK(pid));
426 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
427 		if (p->p_pid == pid) {
428 			PROC_LOCK(p);
429 			if (p->p_state == PRS_NEW ||
430 			    (!zombie && p->p_state == PRS_ZOMBIE)) {
431 				PROC_UNLOCK(p);
432 				p = NULL;
433 			}
434 			break;
435 		}
436 	}
437 	sx_sunlock(PIDHASHLOCK(pid));
438 	return (p);
439 }
440 
441 struct proc *
442 pfind(pid_t pid)
443 {
444 
445 	return (_pfind(pid, false));
446 }
447 
448 /*
449  * Same as pfind but allow zombies.
450  */
451 struct proc *
452 pfind_any(pid_t pid)
453 {
454 
455 	return (_pfind(pid, true));
456 }
457 
458 /*
459  * Locate a process group by number.
460  * The caller must hold proctree_lock.
461  */
462 struct pgrp *
463 pgfind(pid_t pgid)
464 {
465 	struct pgrp *pgrp;
466 
467 	sx_assert(&proctree_lock, SX_LOCKED);
468 
469 	LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
470 		if (pgrp->pg_id == pgid) {
471 			PGRP_LOCK(pgrp);
472 			return (pgrp);
473 		}
474 	}
475 	return (NULL);
476 }
477 
478 /*
479  * Locate process and do additional manipulations, depending on flags.
480  */
481 int
482 pget(pid_t pid, int flags, struct proc **pp)
483 {
484 	struct proc *p;
485 	struct thread *td1;
486 	int error;
487 
488 	p = curproc;
489 	if (p->p_pid == pid) {
490 		PROC_LOCK(p);
491 	} else {
492 		p = NULL;
493 		if (pid <= PID_MAX) {
494 			if ((flags & PGET_NOTWEXIT) == 0)
495 				p = pfind_any(pid);
496 			else
497 				p = pfind(pid);
498 		} else if ((flags & PGET_NOTID) == 0) {
499 			td1 = tdfind(pid, -1);
500 			if (td1 != NULL)
501 				p = td1->td_proc;
502 		}
503 		if (p == NULL)
504 			return (ESRCH);
505 		if ((flags & PGET_CANSEE) != 0) {
506 			error = p_cansee(curthread, p);
507 			if (error != 0)
508 				goto errout;
509 		}
510 	}
511 	if ((flags & PGET_CANDEBUG) != 0) {
512 		error = p_candebug(curthread, p);
513 		if (error != 0)
514 			goto errout;
515 	}
516 	if ((flags & PGET_ISCURRENT) != 0 && curproc != p) {
517 		error = EPERM;
518 		goto errout;
519 	}
520 	if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) {
521 		error = ESRCH;
522 		goto errout;
523 	}
524 	if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) {
525 		/*
526 		 * XXXRW: Not clear ESRCH is the right error during proc
527 		 * execve().
528 		 */
529 		error = ESRCH;
530 		goto errout;
531 	}
532 	if ((flags & PGET_HOLD) != 0) {
533 		_PHOLD(p);
534 		PROC_UNLOCK(p);
535 	}
536 	*pp = p;
537 	return (0);
538 errout:
539 	PROC_UNLOCK(p);
540 	return (error);
541 }
542 
543 /*
544  * Create a new process group.
545  * pgid must be equal to the pid of p.
546  * Begin a new session if required.
547  */
548 int
549 enterpgrp(struct proc *p, pid_t pgid, struct pgrp *pgrp, struct session *sess)
550 {
551 
552 	sx_assert(&proctree_lock, SX_XLOCKED);
553 
554 	KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
555 	KASSERT(p->p_pid == pgid,
556 	    ("enterpgrp: new pgrp and pid != pgid"));
557 	KASSERT(pgfind(pgid) == NULL,
558 	    ("enterpgrp: pgrp with pgid exists"));
559 	KASSERT(!SESS_LEADER(p),
560 	    ("enterpgrp: session leader attempted setpgrp"));
561 
562 	mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
563 
564 	if (sess != NULL) {
565 		/*
566 		 * new session
567 		 */
568 		mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
569 		PROC_LOCK(p);
570 		p->p_flag &= ~P_CONTROLT;
571 		PROC_UNLOCK(p);
572 		PGRP_LOCK(pgrp);
573 		sess->s_leader = p;
574 		sess->s_sid = p->p_pid;
575 		proc_id_set(PROC_ID_SESSION, p->p_pid);
576 		refcount_init(&sess->s_count, 1);
577 		sess->s_ttyvp = NULL;
578 		sess->s_ttydp = NULL;
579 		sess->s_ttyp = NULL;
580 		bcopy(p->p_session->s_login, sess->s_login,
581 			    sizeof(sess->s_login));
582 		pgrp->pg_session = sess;
583 		KASSERT(p == curproc,
584 		    ("enterpgrp: mksession and p != curproc"));
585 	} else {
586 		pgrp->pg_session = p->p_session;
587 		sess_hold(pgrp->pg_session);
588 		PGRP_LOCK(pgrp);
589 	}
590 	pgrp->pg_id = pgid;
591 	proc_id_set(PROC_ID_GROUP, p->p_pid);
592 	LIST_INIT(&pgrp->pg_members);
593 
594 	/*
595 	 * As we have an exclusive lock of proctree_lock,
596 	 * this should not deadlock.
597 	 */
598 	LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
599 	pgrp->pg_jobc = 0;
600 	SLIST_INIT(&pgrp->pg_sigiolst);
601 	PGRP_UNLOCK(pgrp);
602 
603 	doenterpgrp(p, pgrp);
604 
605 	return (0);
606 }
607 
608 /*
609  * Move p to an existing process group
610  */
611 int
612 enterthispgrp(struct proc *p, struct pgrp *pgrp)
613 {
614 
615 	sx_assert(&proctree_lock, SX_XLOCKED);
616 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
617 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
618 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
619 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
620 	KASSERT(pgrp->pg_session == p->p_session,
621 		("%s: pgrp's session %p, p->p_session %p.\n",
622 		__func__,
623 		pgrp->pg_session,
624 		p->p_session));
625 	KASSERT(pgrp != p->p_pgrp,
626 		("%s: p belongs to pgrp.", __func__));
627 
628 	doenterpgrp(p, pgrp);
629 
630 	return (0);
631 }
632 
633 /*
634  * Move p to a process group
635  */
636 static void
637 doenterpgrp(struct proc *p, struct pgrp *pgrp)
638 {
639 	struct pgrp *savepgrp;
640 
641 	sx_assert(&proctree_lock, SX_XLOCKED);
642 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
643 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
644 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
645 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
646 
647 	savepgrp = p->p_pgrp;
648 
649 	/*
650 	 * Adjust eligibility of affected pgrps to participate in job control.
651 	 * Increment eligibility counts before decrementing, otherwise we
652 	 * could reach 0 spuriously during the first call.
653 	 */
654 	fixjobc(p, pgrp, 1);
655 	fixjobc(p, p->p_pgrp, 0);
656 
657 	PGRP_LOCK(pgrp);
658 	PGRP_LOCK(savepgrp);
659 	PROC_LOCK(p);
660 	LIST_REMOVE(p, p_pglist);
661 	p->p_pgrp = pgrp;
662 	PROC_UNLOCK(p);
663 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
664 	PGRP_UNLOCK(savepgrp);
665 	PGRP_UNLOCK(pgrp);
666 	if (LIST_EMPTY(&savepgrp->pg_members))
667 		pgdelete(savepgrp);
668 }
669 
670 /*
671  * remove process from process group
672  */
673 int
674 leavepgrp(struct proc *p)
675 {
676 	struct pgrp *savepgrp;
677 
678 	sx_assert(&proctree_lock, SX_XLOCKED);
679 	savepgrp = p->p_pgrp;
680 	PGRP_LOCK(savepgrp);
681 	PROC_LOCK(p);
682 	LIST_REMOVE(p, p_pglist);
683 	p->p_pgrp = NULL;
684 	PROC_UNLOCK(p);
685 	PGRP_UNLOCK(savepgrp);
686 	if (LIST_EMPTY(&savepgrp->pg_members))
687 		pgdelete(savepgrp);
688 	return (0);
689 }
690 
691 /*
692  * delete a process group
693  */
694 static void
695 pgdelete(struct pgrp *pgrp)
696 {
697 	struct session *savesess;
698 	struct tty *tp;
699 
700 	sx_assert(&proctree_lock, SX_XLOCKED);
701 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
702 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
703 
704 	/*
705 	 * Reset any sigio structures pointing to us as a result of
706 	 * F_SETOWN with our pgid.
707 	 */
708 	funsetownlst(&pgrp->pg_sigiolst);
709 
710 	PGRP_LOCK(pgrp);
711 	tp = pgrp->pg_session->s_ttyp;
712 	LIST_REMOVE(pgrp, pg_hash);
713 	savesess = pgrp->pg_session;
714 	PGRP_UNLOCK(pgrp);
715 
716 	/* Remove the reference to the pgrp before deallocating it. */
717 	if (tp != NULL) {
718 		tty_lock(tp);
719 		tty_rel_pgrp(tp, pgrp);
720 	}
721 
722 	proc_id_clear(PROC_ID_GROUP, pgrp->pg_id);
723 	mtx_destroy(&pgrp->pg_mtx);
724 	free(pgrp, M_PGRP);
725 	sess_release(savesess);
726 }
727 
728 static void
729 pgadjustjobc(struct pgrp *pgrp, int entering)
730 {
731 
732 	PGRP_LOCK(pgrp);
733 	if (entering) {
734 		MPASS(pgrp->pg_jobc >= 0);
735 		pgrp->pg_jobc++;
736 	} else {
737 		MPASS(pgrp->pg_jobc > 0);
738 		--pgrp->pg_jobc;
739 		if (pgrp->pg_jobc == 0)
740 			orphanpg(pgrp);
741 	}
742 	PGRP_UNLOCK(pgrp);
743 }
744 
745 /*
746  * Adjust pgrp jobc counters when specified process changes process group.
747  * We count the number of processes in each process group that "qualify"
748  * the group for terminal job control (those with a parent in a different
749  * process group of the same session).  If that count reaches zero, the
750  * process group becomes orphaned.  Check both the specified process'
751  * process group and that of its children.
752  * entering == 0 => p is leaving specified group.
753  * entering == 1 => p is entering specified group.
754  */
755 void
756 fixjobc(struct proc *p, struct pgrp *pgrp, int entering)
757 {
758 	struct pgrp *hispgrp;
759 	struct session *mysession;
760 	struct proc *q;
761 
762 	sx_assert(&proctree_lock, SX_LOCKED);
763 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
764 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
765 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
766 
767 	/*
768 	 * Check p's parent to see whether p qualifies its own process
769 	 * group; if so, adjust count for p's process group.
770 	 */
771 	mysession = pgrp->pg_session;
772 	if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
773 	    hispgrp->pg_session == mysession)
774 		pgadjustjobc(pgrp, entering);
775 
776 	/*
777 	 * Check this process' children to see whether they qualify
778 	 * their process groups; if so, adjust counts for children's
779 	 * process groups.
780 	 */
781 	LIST_FOREACH(q, &p->p_children, p_sibling) {
782 		hispgrp = q->p_pgrp;
783 		if (hispgrp == pgrp ||
784 		    hispgrp->pg_session != mysession)
785 			continue;
786 		if (q->p_state == PRS_ZOMBIE)
787 			continue;
788 		pgadjustjobc(hispgrp, entering);
789 	}
790 }
791 
792 void
793 killjobc(void)
794 {
795 	struct session *sp;
796 	struct tty *tp;
797 	struct proc *p;
798 	struct vnode *ttyvp;
799 
800 	p = curproc;
801 	MPASS(p->p_flag & P_WEXIT);
802 	/*
803 	 * Do a quick check to see if there is anything to do with the
804 	 * proctree_lock held. pgrp and LIST_EMPTY checks are for fixjobc().
805 	 */
806 	PROC_LOCK(p);
807 	if (!SESS_LEADER(p) &&
808 	    (p->p_pgrp == p->p_pptr->p_pgrp) &&
809 	    LIST_EMPTY(&p->p_children)) {
810 		PROC_UNLOCK(p);
811 		return;
812 	}
813 	PROC_UNLOCK(p);
814 
815 	sx_xlock(&proctree_lock);
816 	if (SESS_LEADER(p)) {
817 		sp = p->p_session;
818 
819 		/*
820 		 * s_ttyp is not zero'd; we use this to indicate that
821 		 * the session once had a controlling terminal. (for
822 		 * logging and informational purposes)
823 		 */
824 		SESS_LOCK(sp);
825 		ttyvp = sp->s_ttyvp;
826 		tp = sp->s_ttyp;
827 		sp->s_ttyvp = NULL;
828 		sp->s_ttydp = NULL;
829 		sp->s_leader = NULL;
830 		SESS_UNLOCK(sp);
831 
832 		/*
833 		 * Signal foreground pgrp and revoke access to
834 		 * controlling terminal if it has not been revoked
835 		 * already.
836 		 *
837 		 * Because the TTY may have been revoked in the mean
838 		 * time and could already have a new session associated
839 		 * with it, make sure we don't send a SIGHUP to a
840 		 * foreground process group that does not belong to this
841 		 * session.
842 		 */
843 
844 		if (tp != NULL) {
845 			tty_lock(tp);
846 			if (tp->t_session == sp)
847 				tty_signal_pgrp(tp, SIGHUP);
848 			tty_unlock(tp);
849 		}
850 
851 		if (ttyvp != NULL) {
852 			sx_xunlock(&proctree_lock);
853 			if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) {
854 				VOP_REVOKE(ttyvp, REVOKEALL);
855 				VOP_UNLOCK(ttyvp);
856 			}
857 			vrele(ttyvp);
858 			sx_xlock(&proctree_lock);
859 		}
860 	}
861 	fixjobc(p, p->p_pgrp, 0);
862 	sx_xunlock(&proctree_lock);
863 }
864 
865 /*
866  * A process group has become orphaned;
867  * if there are any stopped processes in the group,
868  * hang-up all process in that group.
869  */
870 static void
871 orphanpg(struct pgrp *pg)
872 {
873 	struct proc *p;
874 
875 	PGRP_LOCK_ASSERT(pg, MA_OWNED);
876 
877 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
878 		PROC_LOCK(p);
879 		if (P_SHOULDSTOP(p) == P_STOPPED_SIG) {
880 			PROC_UNLOCK(p);
881 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
882 				PROC_LOCK(p);
883 				kern_psignal(p, SIGHUP);
884 				kern_psignal(p, SIGCONT);
885 				PROC_UNLOCK(p);
886 			}
887 			return;
888 		}
889 		PROC_UNLOCK(p);
890 	}
891 }
892 
893 void
894 sess_hold(struct session *s)
895 {
896 
897 	refcount_acquire(&s->s_count);
898 }
899 
900 void
901 sess_release(struct session *s)
902 {
903 
904 	if (refcount_release(&s->s_count)) {
905 		if (s->s_ttyp != NULL) {
906 			tty_lock(s->s_ttyp);
907 			tty_rel_sess(s->s_ttyp, s);
908 		}
909 		proc_id_clear(PROC_ID_SESSION, s->s_sid);
910 		mtx_destroy(&s->s_mtx);
911 		free(s, M_SESSION);
912 	}
913 }
914 
915 #ifdef DDB
916 
917 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
918 {
919 	struct pgrp *pgrp;
920 	struct proc *p;
921 	int i;
922 
923 	for (i = 0; i <= pgrphash; i++) {
924 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
925 			printf("\tindx %d\n", i);
926 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
927 				printf(
928 			"\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
929 				    (void *)pgrp, (long)pgrp->pg_id,
930 				    (void *)pgrp->pg_session,
931 				    pgrp->pg_session->s_count,
932 				    (void *)LIST_FIRST(&pgrp->pg_members));
933 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
934 					printf("\t\tpid %ld addr %p pgrp %p\n",
935 					    (long)p->p_pid, (void *)p,
936 					    (void *)p->p_pgrp);
937 				}
938 			}
939 		}
940 	}
941 }
942 #endif /* DDB */
943 
944 /*
945  * Calculate the kinfo_proc members which contain process-wide
946  * informations.
947  * Must be called with the target process locked.
948  */
949 static void
950 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
951 {
952 	struct thread *td;
953 
954 	PROC_LOCK_ASSERT(p, MA_OWNED);
955 
956 	kp->ki_estcpu = 0;
957 	kp->ki_pctcpu = 0;
958 	FOREACH_THREAD_IN_PROC(p, td) {
959 		thread_lock(td);
960 		kp->ki_pctcpu += sched_pctcpu(td);
961 		kp->ki_estcpu += sched_estcpu(td);
962 		thread_unlock(td);
963 	}
964 }
965 
966 /*
967  * Clear kinfo_proc and fill in any information that is common
968  * to all threads in the process.
969  * Must be called with the target process locked.
970  */
971 static void
972 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
973 {
974 	struct thread *td0;
975 	struct tty *tp;
976 	struct session *sp;
977 	struct ucred *cred;
978 	struct sigacts *ps;
979 	struct timeval boottime;
980 
981 	PROC_LOCK_ASSERT(p, MA_OWNED);
982 	bzero(kp, sizeof(*kp));
983 
984 	kp->ki_structsize = sizeof(*kp);
985 	kp->ki_paddr = p;
986 	kp->ki_addr =/* p->p_addr; */0; /* XXX */
987 	kp->ki_args = p->p_args;
988 	kp->ki_textvp = p->p_textvp;
989 #ifdef KTRACE
990 	kp->ki_tracep = p->p_tracevp;
991 	kp->ki_traceflag = p->p_traceflag;
992 #endif
993 	kp->ki_fd = p->p_fd;
994 	kp->ki_vmspace = p->p_vmspace;
995 	kp->ki_flag = p->p_flag;
996 	kp->ki_flag2 = p->p_flag2;
997 	cred = p->p_ucred;
998 	if (cred) {
999 		kp->ki_uid = cred->cr_uid;
1000 		kp->ki_ruid = cred->cr_ruid;
1001 		kp->ki_svuid = cred->cr_svuid;
1002 		kp->ki_cr_flags = 0;
1003 		if (cred->cr_flags & CRED_FLAG_CAPMODE)
1004 			kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
1005 		/* XXX bde doesn't like KI_NGROUPS */
1006 		if (cred->cr_ngroups > KI_NGROUPS) {
1007 			kp->ki_ngroups = KI_NGROUPS;
1008 			kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
1009 		} else
1010 			kp->ki_ngroups = cred->cr_ngroups;
1011 		bcopy(cred->cr_groups, kp->ki_groups,
1012 		    kp->ki_ngroups * sizeof(gid_t));
1013 		kp->ki_rgid = cred->cr_rgid;
1014 		kp->ki_svgid = cred->cr_svgid;
1015 		/* If jailed(cred), emulate the old P_JAILED flag. */
1016 		if (jailed(cred)) {
1017 			kp->ki_flag |= P_JAILED;
1018 			/* If inside the jail, use 0 as a jail ID. */
1019 			if (cred->cr_prison != curthread->td_ucred->cr_prison)
1020 				kp->ki_jid = cred->cr_prison->pr_id;
1021 		}
1022 		strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
1023 		    sizeof(kp->ki_loginclass));
1024 	}
1025 	ps = p->p_sigacts;
1026 	if (ps) {
1027 		mtx_lock(&ps->ps_mtx);
1028 		kp->ki_sigignore = ps->ps_sigignore;
1029 		kp->ki_sigcatch = ps->ps_sigcatch;
1030 		mtx_unlock(&ps->ps_mtx);
1031 	}
1032 	if (p->p_state != PRS_NEW &&
1033 	    p->p_state != PRS_ZOMBIE &&
1034 	    p->p_vmspace != NULL) {
1035 		struct vmspace *vm = p->p_vmspace;
1036 
1037 		kp->ki_size = vm->vm_map.size;
1038 		kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
1039 		FOREACH_THREAD_IN_PROC(p, td0) {
1040 			if (!TD_IS_SWAPPED(td0))
1041 				kp->ki_rssize += td0->td_kstack_pages;
1042 		}
1043 		kp->ki_swrss = vm->vm_swrss;
1044 		kp->ki_tsize = vm->vm_tsize;
1045 		kp->ki_dsize = vm->vm_dsize;
1046 		kp->ki_ssize = vm->vm_ssize;
1047 	} else if (p->p_state == PRS_ZOMBIE)
1048 		kp->ki_stat = SZOMB;
1049 	if (kp->ki_flag & P_INMEM)
1050 		kp->ki_sflag = PS_INMEM;
1051 	else
1052 		kp->ki_sflag = 0;
1053 	/* Calculate legacy swtime as seconds since 'swtick'. */
1054 	kp->ki_swtime = (ticks - p->p_swtick) / hz;
1055 	kp->ki_pid = p->p_pid;
1056 	kp->ki_nice = p->p_nice;
1057 	kp->ki_fibnum = p->p_fibnum;
1058 	kp->ki_start = p->p_stats->p_start;
1059 	getboottime(&boottime);
1060 	timevaladd(&kp->ki_start, &boottime);
1061 	PROC_STATLOCK(p);
1062 	rufetch(p, &kp->ki_rusage);
1063 	kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
1064 	calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
1065 	PROC_STATUNLOCK(p);
1066 	calccru(p, &kp->ki_childutime, &kp->ki_childstime);
1067 	/* Some callers want child times in a single value. */
1068 	kp->ki_childtime = kp->ki_childstime;
1069 	timevaladd(&kp->ki_childtime, &kp->ki_childutime);
1070 
1071 	FOREACH_THREAD_IN_PROC(p, td0)
1072 		kp->ki_cow += td0->td_cow;
1073 
1074 	tp = NULL;
1075 	if (p->p_pgrp) {
1076 		kp->ki_pgid = p->p_pgrp->pg_id;
1077 		kp->ki_jobc = p->p_pgrp->pg_jobc;
1078 		sp = p->p_pgrp->pg_session;
1079 
1080 		if (sp != NULL) {
1081 			kp->ki_sid = sp->s_sid;
1082 			SESS_LOCK(sp);
1083 			strlcpy(kp->ki_login, sp->s_login,
1084 			    sizeof(kp->ki_login));
1085 			if (sp->s_ttyvp)
1086 				kp->ki_kiflag |= KI_CTTY;
1087 			if (SESS_LEADER(p))
1088 				kp->ki_kiflag |= KI_SLEADER;
1089 			/* XXX proctree_lock */
1090 			tp = sp->s_ttyp;
1091 			SESS_UNLOCK(sp);
1092 		}
1093 	}
1094 	if ((p->p_flag & P_CONTROLT) && tp != NULL) {
1095 		kp->ki_tdev = tty_udev(tp);
1096 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1097 		kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
1098 		if (tp->t_session)
1099 			kp->ki_tsid = tp->t_session->s_sid;
1100 	} else {
1101 		kp->ki_tdev = NODEV;
1102 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1103 	}
1104 	if (p->p_comm[0] != '\0')
1105 		strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
1106 	if (p->p_sysent && p->p_sysent->sv_name != NULL &&
1107 	    p->p_sysent->sv_name[0] != '\0')
1108 		strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
1109 	kp->ki_siglist = p->p_siglist;
1110 	kp->ki_xstat = KW_EXITCODE(p->p_xexit, p->p_xsig);
1111 	kp->ki_acflag = p->p_acflag;
1112 	kp->ki_lock = p->p_lock;
1113 	if (p->p_pptr) {
1114 		kp->ki_ppid = p->p_oppid;
1115 		if (p->p_flag & P_TRACED)
1116 			kp->ki_tracer = p->p_pptr->p_pid;
1117 	}
1118 }
1119 
1120 /*
1121  * Fill in information that is thread specific.  Must be called with
1122  * target process locked.  If 'preferthread' is set, overwrite certain
1123  * process-related fields that are maintained for both threads and
1124  * processes.
1125  */
1126 static void
1127 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
1128 {
1129 	struct proc *p;
1130 
1131 	p = td->td_proc;
1132 	kp->ki_tdaddr = td;
1133 	PROC_LOCK_ASSERT(p, MA_OWNED);
1134 
1135 	if (preferthread)
1136 		PROC_STATLOCK(p);
1137 	thread_lock(td);
1138 	if (td->td_wmesg != NULL)
1139 		strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
1140 	else
1141 		bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
1142 	if (strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname)) >=
1143 	    sizeof(kp->ki_tdname)) {
1144 		strlcpy(kp->ki_moretdname,
1145 		    td->td_name + sizeof(kp->ki_tdname) - 1,
1146 		    sizeof(kp->ki_moretdname));
1147 	} else {
1148 		bzero(kp->ki_moretdname, sizeof(kp->ki_moretdname));
1149 	}
1150 	if (TD_ON_LOCK(td)) {
1151 		kp->ki_kiflag |= KI_LOCKBLOCK;
1152 		strlcpy(kp->ki_lockname, td->td_lockname,
1153 		    sizeof(kp->ki_lockname));
1154 	} else {
1155 		kp->ki_kiflag &= ~KI_LOCKBLOCK;
1156 		bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
1157 	}
1158 
1159 	if (p->p_state == PRS_NORMAL) { /* approximate. */
1160 		if (TD_ON_RUNQ(td) ||
1161 		    TD_CAN_RUN(td) ||
1162 		    TD_IS_RUNNING(td)) {
1163 			kp->ki_stat = SRUN;
1164 		} else if (P_SHOULDSTOP(p)) {
1165 			kp->ki_stat = SSTOP;
1166 		} else if (TD_IS_SLEEPING(td)) {
1167 			kp->ki_stat = SSLEEP;
1168 		} else if (TD_ON_LOCK(td)) {
1169 			kp->ki_stat = SLOCK;
1170 		} else {
1171 			kp->ki_stat = SWAIT;
1172 		}
1173 	} else if (p->p_state == PRS_ZOMBIE) {
1174 		kp->ki_stat = SZOMB;
1175 	} else {
1176 		kp->ki_stat = SIDL;
1177 	}
1178 
1179 	/* Things in the thread */
1180 	kp->ki_wchan = td->td_wchan;
1181 	kp->ki_pri.pri_level = td->td_priority;
1182 	kp->ki_pri.pri_native = td->td_base_pri;
1183 
1184 	/*
1185 	 * Note: legacy fields; clamp at the old NOCPU value and/or
1186 	 * the maximum u_char CPU value.
1187 	 */
1188 	if (td->td_lastcpu == NOCPU)
1189 		kp->ki_lastcpu_old = NOCPU_OLD;
1190 	else if (td->td_lastcpu > MAXCPU_OLD)
1191 		kp->ki_lastcpu_old = MAXCPU_OLD;
1192 	else
1193 		kp->ki_lastcpu_old = td->td_lastcpu;
1194 
1195 	if (td->td_oncpu == NOCPU)
1196 		kp->ki_oncpu_old = NOCPU_OLD;
1197 	else if (td->td_oncpu > MAXCPU_OLD)
1198 		kp->ki_oncpu_old = MAXCPU_OLD;
1199 	else
1200 		kp->ki_oncpu_old = td->td_oncpu;
1201 
1202 	kp->ki_lastcpu = td->td_lastcpu;
1203 	kp->ki_oncpu = td->td_oncpu;
1204 	kp->ki_tdflags = td->td_flags;
1205 	kp->ki_tid = td->td_tid;
1206 	kp->ki_numthreads = p->p_numthreads;
1207 	kp->ki_pcb = td->td_pcb;
1208 	kp->ki_kstack = (void *)td->td_kstack;
1209 	kp->ki_slptime = (ticks - td->td_slptick) / hz;
1210 	kp->ki_pri.pri_class = td->td_pri_class;
1211 	kp->ki_pri.pri_user = td->td_user_pri;
1212 
1213 	if (preferthread) {
1214 		rufetchtd(td, &kp->ki_rusage);
1215 		kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
1216 		kp->ki_pctcpu = sched_pctcpu(td);
1217 		kp->ki_estcpu = sched_estcpu(td);
1218 		kp->ki_cow = td->td_cow;
1219 	}
1220 
1221 	/* We can't get this anymore but ps etc never used it anyway. */
1222 	kp->ki_rqindex = 0;
1223 
1224 	if (preferthread)
1225 		kp->ki_siglist = td->td_siglist;
1226 	kp->ki_sigmask = td->td_sigmask;
1227 	thread_unlock(td);
1228 	if (preferthread)
1229 		PROC_STATUNLOCK(p);
1230 }
1231 
1232 /*
1233  * Fill in a kinfo_proc structure for the specified process.
1234  * Must be called with the target process locked.
1235  */
1236 void
1237 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1238 {
1239 
1240 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1241 
1242 	fill_kinfo_proc_only(p, kp);
1243 	fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1244 	fill_kinfo_aggregate(p, kp);
1245 }
1246 
1247 struct pstats *
1248 pstats_alloc(void)
1249 {
1250 
1251 	return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1252 }
1253 
1254 /*
1255  * Copy parts of p_stats; zero the rest of p_stats (statistics).
1256  */
1257 void
1258 pstats_fork(struct pstats *src, struct pstats *dst)
1259 {
1260 
1261 	bzero(&dst->pstat_startzero,
1262 	    __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1263 	bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1264 	    __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1265 }
1266 
1267 void
1268 pstats_free(struct pstats *ps)
1269 {
1270 
1271 	free(ps, M_SUBPROC);
1272 }
1273 
1274 #ifdef COMPAT_FREEBSD32
1275 
1276 /*
1277  * This function is typically used to copy out the kernel address, so
1278  * it can be replaced by assignment of zero.
1279  */
1280 static inline uint32_t
1281 ptr32_trim(const void *ptr)
1282 {
1283 	uintptr_t uptr;
1284 
1285 	uptr = (uintptr_t)ptr;
1286 	return ((uptr > UINT_MAX) ? 0 : uptr);
1287 }
1288 
1289 #define PTRTRIM_CP(src,dst,fld) \
1290 	do { (dst).fld = ptr32_trim((src).fld); } while (0)
1291 
1292 static void
1293 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1294 {
1295 	int i;
1296 
1297 	bzero(ki32, sizeof(struct kinfo_proc32));
1298 	ki32->ki_structsize = sizeof(struct kinfo_proc32);
1299 	CP(*ki, *ki32, ki_layout);
1300 	PTRTRIM_CP(*ki, *ki32, ki_args);
1301 	PTRTRIM_CP(*ki, *ki32, ki_paddr);
1302 	PTRTRIM_CP(*ki, *ki32, ki_addr);
1303 	PTRTRIM_CP(*ki, *ki32, ki_tracep);
1304 	PTRTRIM_CP(*ki, *ki32, ki_textvp);
1305 	PTRTRIM_CP(*ki, *ki32, ki_fd);
1306 	PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1307 	PTRTRIM_CP(*ki, *ki32, ki_wchan);
1308 	CP(*ki, *ki32, ki_pid);
1309 	CP(*ki, *ki32, ki_ppid);
1310 	CP(*ki, *ki32, ki_pgid);
1311 	CP(*ki, *ki32, ki_tpgid);
1312 	CP(*ki, *ki32, ki_sid);
1313 	CP(*ki, *ki32, ki_tsid);
1314 	CP(*ki, *ki32, ki_jobc);
1315 	CP(*ki, *ki32, ki_tdev);
1316 	CP(*ki, *ki32, ki_tdev_freebsd11);
1317 	CP(*ki, *ki32, ki_siglist);
1318 	CP(*ki, *ki32, ki_sigmask);
1319 	CP(*ki, *ki32, ki_sigignore);
1320 	CP(*ki, *ki32, ki_sigcatch);
1321 	CP(*ki, *ki32, ki_uid);
1322 	CP(*ki, *ki32, ki_ruid);
1323 	CP(*ki, *ki32, ki_svuid);
1324 	CP(*ki, *ki32, ki_rgid);
1325 	CP(*ki, *ki32, ki_svgid);
1326 	CP(*ki, *ki32, ki_ngroups);
1327 	for (i = 0; i < KI_NGROUPS; i++)
1328 		CP(*ki, *ki32, ki_groups[i]);
1329 	CP(*ki, *ki32, ki_size);
1330 	CP(*ki, *ki32, ki_rssize);
1331 	CP(*ki, *ki32, ki_swrss);
1332 	CP(*ki, *ki32, ki_tsize);
1333 	CP(*ki, *ki32, ki_dsize);
1334 	CP(*ki, *ki32, ki_ssize);
1335 	CP(*ki, *ki32, ki_xstat);
1336 	CP(*ki, *ki32, ki_acflag);
1337 	CP(*ki, *ki32, ki_pctcpu);
1338 	CP(*ki, *ki32, ki_estcpu);
1339 	CP(*ki, *ki32, ki_slptime);
1340 	CP(*ki, *ki32, ki_swtime);
1341 	CP(*ki, *ki32, ki_cow);
1342 	CP(*ki, *ki32, ki_runtime);
1343 	TV_CP(*ki, *ki32, ki_start);
1344 	TV_CP(*ki, *ki32, ki_childtime);
1345 	CP(*ki, *ki32, ki_flag);
1346 	CP(*ki, *ki32, ki_kiflag);
1347 	CP(*ki, *ki32, ki_traceflag);
1348 	CP(*ki, *ki32, ki_stat);
1349 	CP(*ki, *ki32, ki_nice);
1350 	CP(*ki, *ki32, ki_lock);
1351 	CP(*ki, *ki32, ki_rqindex);
1352 	CP(*ki, *ki32, ki_oncpu);
1353 	CP(*ki, *ki32, ki_lastcpu);
1354 
1355 	/* XXX TODO: wrap cpu value as appropriate */
1356 	CP(*ki, *ki32, ki_oncpu_old);
1357 	CP(*ki, *ki32, ki_lastcpu_old);
1358 
1359 	bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1360 	bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1361 	bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1362 	bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1363 	bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1364 	bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1365 	bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1366 	bcopy(ki->ki_moretdname, ki32->ki_moretdname, MAXCOMLEN - TDNAMLEN + 1);
1367 	CP(*ki, *ki32, ki_tracer);
1368 	CP(*ki, *ki32, ki_flag2);
1369 	CP(*ki, *ki32, ki_fibnum);
1370 	CP(*ki, *ki32, ki_cr_flags);
1371 	CP(*ki, *ki32, ki_jid);
1372 	CP(*ki, *ki32, ki_numthreads);
1373 	CP(*ki, *ki32, ki_tid);
1374 	CP(*ki, *ki32, ki_pri);
1375 	freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1376 	freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1377 	PTRTRIM_CP(*ki, *ki32, ki_pcb);
1378 	PTRTRIM_CP(*ki, *ki32, ki_kstack);
1379 	PTRTRIM_CP(*ki, *ki32, ki_udata);
1380 	PTRTRIM_CP(*ki, *ki32, ki_tdaddr);
1381 	CP(*ki, *ki32, ki_sflag);
1382 	CP(*ki, *ki32, ki_tdflags);
1383 }
1384 #endif
1385 
1386 static ssize_t
1387 kern_proc_out_size(struct proc *p, int flags)
1388 {
1389 	ssize_t size = 0;
1390 
1391 	PROC_LOCK_ASSERT(p, MA_OWNED);
1392 
1393 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1394 #ifdef COMPAT_FREEBSD32
1395 		if ((flags & KERN_PROC_MASK32) != 0) {
1396 			size += sizeof(struct kinfo_proc32);
1397 		} else
1398 #endif
1399 			size += sizeof(struct kinfo_proc);
1400 	} else {
1401 #ifdef COMPAT_FREEBSD32
1402 		if ((flags & KERN_PROC_MASK32) != 0)
1403 			size += sizeof(struct kinfo_proc32) * p->p_numthreads;
1404 		else
1405 #endif
1406 			size += sizeof(struct kinfo_proc) * p->p_numthreads;
1407 	}
1408 	PROC_UNLOCK(p);
1409 	return (size);
1410 }
1411 
1412 int
1413 kern_proc_out(struct proc *p, struct sbuf *sb, int flags)
1414 {
1415 	struct thread *td;
1416 	struct kinfo_proc ki;
1417 #ifdef COMPAT_FREEBSD32
1418 	struct kinfo_proc32 ki32;
1419 #endif
1420 	int error;
1421 
1422 	PROC_LOCK_ASSERT(p, MA_OWNED);
1423 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1424 
1425 	error = 0;
1426 	fill_kinfo_proc(p, &ki);
1427 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1428 #ifdef COMPAT_FREEBSD32
1429 		if ((flags & KERN_PROC_MASK32) != 0) {
1430 			freebsd32_kinfo_proc_out(&ki, &ki32);
1431 			if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1432 				error = ENOMEM;
1433 		} else
1434 #endif
1435 			if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1436 				error = ENOMEM;
1437 	} else {
1438 		FOREACH_THREAD_IN_PROC(p, td) {
1439 			fill_kinfo_thread(td, &ki, 1);
1440 #ifdef COMPAT_FREEBSD32
1441 			if ((flags & KERN_PROC_MASK32) != 0) {
1442 				freebsd32_kinfo_proc_out(&ki, &ki32);
1443 				if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1444 					error = ENOMEM;
1445 			} else
1446 #endif
1447 				if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1448 					error = ENOMEM;
1449 			if (error != 0)
1450 				break;
1451 		}
1452 	}
1453 	PROC_UNLOCK(p);
1454 	return (error);
1455 }
1456 
1457 static int
1458 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
1459 {
1460 	struct sbuf sb;
1461 	struct kinfo_proc ki;
1462 	int error, error2;
1463 
1464 	if (req->oldptr == NULL)
1465 		return (SYSCTL_OUT(req, 0, kern_proc_out_size(p, flags)));
1466 
1467 	sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req);
1468 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1469 	error = kern_proc_out(p, &sb, flags);
1470 	error2 = sbuf_finish(&sb);
1471 	sbuf_delete(&sb);
1472 	if (error != 0)
1473 		return (error);
1474 	else if (error2 != 0)
1475 		return (error2);
1476 	return (0);
1477 }
1478 
1479 int
1480 proc_iterate(int (*cb)(struct proc *, void *), void *cbarg)
1481 {
1482 	struct proc *p;
1483 	int error, i, j;
1484 
1485 	for (i = 0; i < pidhashlock + 1; i++) {
1486 		sx_slock(&pidhashtbl_lock[i]);
1487 		for (j = i; j <= pidhash; j += pidhashlock + 1) {
1488 			LIST_FOREACH(p, &pidhashtbl[j], p_hash) {
1489 				if (p->p_state == PRS_NEW)
1490 					continue;
1491 				error = cb(p, cbarg);
1492 				PROC_LOCK_ASSERT(p, MA_NOTOWNED);
1493 				if (error != 0) {
1494 					sx_sunlock(&pidhashtbl_lock[i]);
1495 					return (error);
1496 				}
1497 			}
1498 		}
1499 		sx_sunlock(&pidhashtbl_lock[i]);
1500 	}
1501 	return (0);
1502 }
1503 
1504 struct kern_proc_out_args {
1505 	struct sysctl_req *req;
1506 	int flags;
1507 	int oid_number;
1508 	int *name;
1509 };
1510 
1511 static int
1512 sysctl_kern_proc_iterate(struct proc *p, void *origarg)
1513 {
1514 	struct kern_proc_out_args *arg = origarg;
1515 	int *name = arg->name;
1516 	int oid_number = arg->oid_number;
1517 	int flags = arg->flags;
1518 	struct sysctl_req *req = arg->req;
1519 	int error = 0;
1520 
1521 	PROC_LOCK(p);
1522 
1523 	KASSERT(p->p_ucred != NULL,
1524 	    ("process credential is NULL for non-NEW proc"));
1525 	/*
1526 	 * Show a user only appropriate processes.
1527 	 */
1528 	if (p_cansee(curthread, p))
1529 		goto skip;
1530 	/*
1531 	 * TODO - make more efficient (see notes below).
1532 	 * do by session.
1533 	 */
1534 	switch (oid_number) {
1535 
1536 	case KERN_PROC_GID:
1537 		if (p->p_ucred->cr_gid != (gid_t)name[0])
1538 			goto skip;
1539 		break;
1540 
1541 	case KERN_PROC_PGRP:
1542 		/* could do this by traversing pgrp */
1543 		if (p->p_pgrp == NULL ||
1544 		    p->p_pgrp->pg_id != (pid_t)name[0])
1545 			goto skip;
1546 		break;
1547 
1548 	case KERN_PROC_RGID:
1549 		if (p->p_ucred->cr_rgid != (gid_t)name[0])
1550 			goto skip;
1551 		break;
1552 
1553 	case KERN_PROC_SESSION:
1554 		if (p->p_session == NULL ||
1555 		    p->p_session->s_sid != (pid_t)name[0])
1556 			goto skip;
1557 		break;
1558 
1559 	case KERN_PROC_TTY:
1560 		if ((p->p_flag & P_CONTROLT) == 0 ||
1561 		    p->p_session == NULL)
1562 			goto skip;
1563 		/* XXX proctree_lock */
1564 		SESS_LOCK(p->p_session);
1565 		if (p->p_session->s_ttyp == NULL ||
1566 		    tty_udev(p->p_session->s_ttyp) !=
1567 		    (dev_t)name[0]) {
1568 			SESS_UNLOCK(p->p_session);
1569 			goto skip;
1570 		}
1571 		SESS_UNLOCK(p->p_session);
1572 		break;
1573 
1574 	case KERN_PROC_UID:
1575 		if (p->p_ucred->cr_uid != (uid_t)name[0])
1576 			goto skip;
1577 		break;
1578 
1579 	case KERN_PROC_RUID:
1580 		if (p->p_ucred->cr_ruid != (uid_t)name[0])
1581 			goto skip;
1582 		break;
1583 
1584 	case KERN_PROC_PROC:
1585 		break;
1586 
1587 	default:
1588 		break;
1589 
1590 	}
1591 	error = sysctl_out_proc(p, req, flags);
1592 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
1593 	return (error);
1594 skip:
1595 	PROC_UNLOCK(p);
1596 	return (0);
1597 }
1598 
1599 static int
1600 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1601 {
1602 	struct kern_proc_out_args iterarg;
1603 	int *name = (int *)arg1;
1604 	u_int namelen = arg2;
1605 	struct proc *p;
1606 	int flags, oid_number;
1607 	int error = 0;
1608 
1609 	oid_number = oidp->oid_number;
1610 	if (oid_number != KERN_PROC_ALL &&
1611 	    (oid_number & KERN_PROC_INC_THREAD) == 0)
1612 		flags = KERN_PROC_NOTHREADS;
1613 	else {
1614 		flags = 0;
1615 		oid_number &= ~KERN_PROC_INC_THREAD;
1616 	}
1617 #ifdef COMPAT_FREEBSD32
1618 	if (req->flags & SCTL_MASK32)
1619 		flags |= KERN_PROC_MASK32;
1620 #endif
1621 	if (oid_number == KERN_PROC_PID) {
1622 		if (namelen != 1)
1623 			return (EINVAL);
1624 		error = sysctl_wire_old_buffer(req, 0);
1625 		if (error)
1626 			return (error);
1627 		error = pget((pid_t)name[0], PGET_CANSEE, &p);
1628 		if (error == 0)
1629 			error = sysctl_out_proc(p, req, flags);
1630 		return (error);
1631 	}
1632 
1633 	switch (oid_number) {
1634 	case KERN_PROC_ALL:
1635 		if (namelen != 0)
1636 			return (EINVAL);
1637 		break;
1638 	case KERN_PROC_PROC:
1639 		if (namelen != 0 && namelen != 1)
1640 			return (EINVAL);
1641 		break;
1642 	default:
1643 		if (namelen != 1)
1644 			return (EINVAL);
1645 		break;
1646 	}
1647 
1648 	if (req->oldptr == NULL) {
1649 		/* overestimate by 5 procs */
1650 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1651 		if (error)
1652 			return (error);
1653 	} else {
1654 		error = sysctl_wire_old_buffer(req, 0);
1655 		if (error != 0)
1656 			return (error);
1657 	}
1658 	iterarg.flags = flags;
1659 	iterarg.oid_number = oid_number;
1660 	iterarg.req = req;
1661 	iterarg.name = name;
1662 	error = proc_iterate(sysctl_kern_proc_iterate, &iterarg);
1663 	return (error);
1664 }
1665 
1666 struct pargs *
1667 pargs_alloc(int len)
1668 {
1669 	struct pargs *pa;
1670 
1671 	pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1672 		M_WAITOK);
1673 	refcount_init(&pa->ar_ref, 1);
1674 	pa->ar_length = len;
1675 	return (pa);
1676 }
1677 
1678 static void
1679 pargs_free(struct pargs *pa)
1680 {
1681 
1682 	free(pa, M_PARGS);
1683 }
1684 
1685 void
1686 pargs_hold(struct pargs *pa)
1687 {
1688 
1689 	if (pa == NULL)
1690 		return;
1691 	refcount_acquire(&pa->ar_ref);
1692 }
1693 
1694 void
1695 pargs_drop(struct pargs *pa)
1696 {
1697 
1698 	if (pa == NULL)
1699 		return;
1700 	if (refcount_release(&pa->ar_ref))
1701 		pargs_free(pa);
1702 }
1703 
1704 static int
1705 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1706     size_t len)
1707 {
1708 	ssize_t n;
1709 
1710 	/*
1711 	 * This may return a short read if the string is shorter than the chunk
1712 	 * and is aligned at the end of the page, and the following page is not
1713 	 * mapped.
1714 	 */
1715 	n = proc_readmem(td, p, (vm_offset_t)sptr, buf, len);
1716 	if (n <= 0)
1717 		return (ENOMEM);
1718 	return (0);
1719 }
1720 
1721 #define PROC_AUXV_MAX	256	/* Safety limit on auxv size. */
1722 
1723 enum proc_vector_type {
1724 	PROC_ARG,
1725 	PROC_ENV,
1726 	PROC_AUX,
1727 };
1728 
1729 #ifdef COMPAT_FREEBSD32
1730 static int
1731 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1732     size_t *vsizep, enum proc_vector_type type)
1733 {
1734 	struct freebsd32_ps_strings pss;
1735 	Elf32_Auxinfo aux;
1736 	vm_offset_t vptr, ptr;
1737 	uint32_t *proc_vector32;
1738 	char **proc_vector;
1739 	size_t vsize, size;
1740 	int i, error;
1741 
1742 	error = 0;
1743 	if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss,
1744 	    sizeof(pss)) != sizeof(pss))
1745 		return (ENOMEM);
1746 	switch (type) {
1747 	case PROC_ARG:
1748 		vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1749 		vsize = pss.ps_nargvstr;
1750 		if (vsize > ARG_MAX)
1751 			return (ENOEXEC);
1752 		size = vsize * sizeof(int32_t);
1753 		break;
1754 	case PROC_ENV:
1755 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1756 		vsize = pss.ps_nenvstr;
1757 		if (vsize > ARG_MAX)
1758 			return (ENOEXEC);
1759 		size = vsize * sizeof(int32_t);
1760 		break;
1761 	case PROC_AUX:
1762 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1763 		    (pss.ps_nenvstr + 1) * sizeof(int32_t);
1764 		if (vptr % 4 != 0)
1765 			return (ENOEXEC);
1766 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1767 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1768 			    sizeof(aux))
1769 				return (ENOMEM);
1770 			if (aux.a_type == AT_NULL)
1771 				break;
1772 			ptr += sizeof(aux);
1773 		}
1774 		if (aux.a_type != AT_NULL)
1775 			return (ENOEXEC);
1776 		vsize = i + 1;
1777 		size = vsize * sizeof(aux);
1778 		break;
1779 	default:
1780 		KASSERT(0, ("Wrong proc vector type: %d", type));
1781 		return (EINVAL);
1782 	}
1783 	proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1784 	if (proc_readmem(td, p, vptr, proc_vector32, size) != size) {
1785 		error = ENOMEM;
1786 		goto done;
1787 	}
1788 	if (type == PROC_AUX) {
1789 		*proc_vectorp = (char **)proc_vector32;
1790 		*vsizep = vsize;
1791 		return (0);
1792 	}
1793 	proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1794 	for (i = 0; i < (int)vsize; i++)
1795 		proc_vector[i] = PTRIN(proc_vector32[i]);
1796 	*proc_vectorp = proc_vector;
1797 	*vsizep = vsize;
1798 done:
1799 	free(proc_vector32, M_TEMP);
1800 	return (error);
1801 }
1802 #endif
1803 
1804 static int
1805 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1806     size_t *vsizep, enum proc_vector_type type)
1807 {
1808 	struct ps_strings pss;
1809 	Elf_Auxinfo aux;
1810 	vm_offset_t vptr, ptr;
1811 	char **proc_vector;
1812 	size_t vsize, size;
1813 	int i;
1814 
1815 #ifdef COMPAT_FREEBSD32
1816 	if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1817 		return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1818 #endif
1819 	if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss,
1820 	    sizeof(pss)) != sizeof(pss))
1821 		return (ENOMEM);
1822 	switch (type) {
1823 	case PROC_ARG:
1824 		vptr = (vm_offset_t)pss.ps_argvstr;
1825 		vsize = pss.ps_nargvstr;
1826 		if (vsize > ARG_MAX)
1827 			return (ENOEXEC);
1828 		size = vsize * sizeof(char *);
1829 		break;
1830 	case PROC_ENV:
1831 		vptr = (vm_offset_t)pss.ps_envstr;
1832 		vsize = pss.ps_nenvstr;
1833 		if (vsize > ARG_MAX)
1834 			return (ENOEXEC);
1835 		size = vsize * sizeof(char *);
1836 		break;
1837 	case PROC_AUX:
1838 		/*
1839 		 * The aux array is just above env array on the stack. Check
1840 		 * that the address is naturally aligned.
1841 		 */
1842 		vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1843 		    * sizeof(char *);
1844 #if __ELF_WORD_SIZE == 64
1845 		if (vptr % sizeof(uint64_t) != 0)
1846 #else
1847 		if (vptr % sizeof(uint32_t) != 0)
1848 #endif
1849 			return (ENOEXEC);
1850 		/*
1851 		 * We count the array size reading the aux vectors from the
1852 		 * stack until AT_NULL vector is returned.  So (to keep the code
1853 		 * simple) we read the process stack twice: the first time here
1854 		 * to find the size and the second time when copying the vectors
1855 		 * to the allocated proc_vector.
1856 		 */
1857 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1858 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1859 			    sizeof(aux))
1860 				return (ENOMEM);
1861 			if (aux.a_type == AT_NULL)
1862 				break;
1863 			ptr += sizeof(aux);
1864 		}
1865 		/*
1866 		 * If the PROC_AUXV_MAX entries are iterated over, and we have
1867 		 * not reached AT_NULL, it is most likely we are reading wrong
1868 		 * data: either the process doesn't have auxv array or data has
1869 		 * been modified. Return the error in this case.
1870 		 */
1871 		if (aux.a_type != AT_NULL)
1872 			return (ENOEXEC);
1873 		vsize = i + 1;
1874 		size = vsize * sizeof(aux);
1875 		break;
1876 	default:
1877 		KASSERT(0, ("Wrong proc vector type: %d", type));
1878 		return (EINVAL); /* In case we are built without INVARIANTS. */
1879 	}
1880 	proc_vector = malloc(size, M_TEMP, M_WAITOK);
1881 	if (proc_readmem(td, p, vptr, proc_vector, size) != size) {
1882 		free(proc_vector, M_TEMP);
1883 		return (ENOMEM);
1884 	}
1885 	*proc_vectorp = proc_vector;
1886 	*vsizep = vsize;
1887 
1888 	return (0);
1889 }
1890 
1891 #define GET_PS_STRINGS_CHUNK_SZ	256	/* Chunk size (bytes) for ps_strings operations. */
1892 
1893 static int
1894 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
1895     enum proc_vector_type type)
1896 {
1897 	size_t done, len, nchr, vsize;
1898 	int error, i;
1899 	char **proc_vector, *sptr;
1900 	char pss_string[GET_PS_STRINGS_CHUNK_SZ];
1901 
1902 	PROC_ASSERT_HELD(p);
1903 
1904 	/*
1905 	 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
1906 	 */
1907 	nchr = 2 * (PATH_MAX + ARG_MAX);
1908 
1909 	error = get_proc_vector(td, p, &proc_vector, &vsize, type);
1910 	if (error != 0)
1911 		return (error);
1912 	for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
1913 		/*
1914 		 * The program may have scribbled into its argv array, e.g. to
1915 		 * remove some arguments.  If that has happened, break out
1916 		 * before trying to read from NULL.
1917 		 */
1918 		if (proc_vector[i] == NULL)
1919 			break;
1920 		for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
1921 			error = proc_read_string(td, p, sptr, pss_string,
1922 			    sizeof(pss_string));
1923 			if (error != 0)
1924 				goto done;
1925 			len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
1926 			if (done + len >= nchr)
1927 				len = nchr - done - 1;
1928 			sbuf_bcat(sb, pss_string, len);
1929 			if (len != GET_PS_STRINGS_CHUNK_SZ)
1930 				break;
1931 			done += GET_PS_STRINGS_CHUNK_SZ;
1932 		}
1933 		sbuf_bcat(sb, "", 1);
1934 		done += len + 1;
1935 	}
1936 done:
1937 	free(proc_vector, M_TEMP);
1938 	return (error);
1939 }
1940 
1941 int
1942 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
1943 {
1944 
1945 	return (get_ps_strings(curthread, p, sb, PROC_ARG));
1946 }
1947 
1948 int
1949 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
1950 {
1951 
1952 	return (get_ps_strings(curthread, p, sb, PROC_ENV));
1953 }
1954 
1955 int
1956 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb)
1957 {
1958 	size_t vsize, size;
1959 	char **auxv;
1960 	int error;
1961 
1962 	error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX);
1963 	if (error == 0) {
1964 #ifdef COMPAT_FREEBSD32
1965 		if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1966 			size = vsize * sizeof(Elf32_Auxinfo);
1967 		else
1968 #endif
1969 			size = vsize * sizeof(Elf_Auxinfo);
1970 		if (sbuf_bcat(sb, auxv, size) != 0)
1971 			error = ENOMEM;
1972 		free(auxv, M_TEMP);
1973 	}
1974 	return (error);
1975 }
1976 
1977 /*
1978  * This sysctl allows a process to retrieve the argument list or process
1979  * title for another process without groping around in the address space
1980  * of the other process.  It also allow a process to set its own "process
1981  * title to a string of its own choice.
1982  */
1983 static int
1984 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
1985 {
1986 	int *name = (int *)arg1;
1987 	u_int namelen = arg2;
1988 	struct pargs *newpa, *pa;
1989 	struct proc *p;
1990 	struct sbuf sb;
1991 	int flags, error = 0, error2;
1992 	pid_t pid;
1993 
1994 	if (namelen != 1)
1995 		return (EINVAL);
1996 
1997 	pid = (pid_t)name[0];
1998 	/*
1999 	 * If the query is for this process and it is single-threaded, there
2000 	 * is nobody to modify pargs, thus we can just read.
2001 	 */
2002 	p = curproc;
2003 	if (pid == p->p_pid && p->p_numthreads == 1 && req->newptr == NULL &&
2004 	    (pa = p->p_args) != NULL)
2005 		return (SYSCTL_OUT(req, pa->ar_args, pa->ar_length));
2006 
2007 	flags = PGET_CANSEE;
2008 	if (req->newptr != NULL)
2009 		flags |= PGET_ISCURRENT;
2010 	error = pget(pid, flags, &p);
2011 	if (error)
2012 		return (error);
2013 
2014 	pa = p->p_args;
2015 	if (pa != NULL) {
2016 		pargs_hold(pa);
2017 		PROC_UNLOCK(p);
2018 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
2019 		pargs_drop(pa);
2020 	} else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
2021 		_PHOLD(p);
2022 		PROC_UNLOCK(p);
2023 		sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2024 		sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2025 		error = proc_getargv(curthread, p, &sb);
2026 		error2 = sbuf_finish(&sb);
2027 		PRELE(p);
2028 		sbuf_delete(&sb);
2029 		if (error == 0 && error2 != 0)
2030 			error = error2;
2031 	} else {
2032 		PROC_UNLOCK(p);
2033 	}
2034 	if (error != 0 || req->newptr == NULL)
2035 		return (error);
2036 
2037 	if (req->newlen > ps_arg_cache_limit - sizeof(struct pargs))
2038 		return (ENOMEM);
2039 
2040 	if (req->newlen == 0) {
2041 		/*
2042 		 * Clear the argument pointer, so that we'll fetch arguments
2043 		 * with proc_getargv() until further notice.
2044 		 */
2045 		newpa = NULL;
2046 	} else {
2047 		newpa = pargs_alloc(req->newlen);
2048 		error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
2049 		if (error != 0) {
2050 			pargs_free(newpa);
2051 			return (error);
2052 		}
2053 	}
2054 	PROC_LOCK(p);
2055 	pa = p->p_args;
2056 	p->p_args = newpa;
2057 	PROC_UNLOCK(p);
2058 	pargs_drop(pa);
2059 	return (0);
2060 }
2061 
2062 /*
2063  * This sysctl allows a process to retrieve environment of another process.
2064  */
2065 static int
2066 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
2067 {
2068 	int *name = (int *)arg1;
2069 	u_int namelen = arg2;
2070 	struct proc *p;
2071 	struct sbuf sb;
2072 	int error, error2;
2073 
2074 	if (namelen != 1)
2075 		return (EINVAL);
2076 
2077 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2078 	if (error != 0)
2079 		return (error);
2080 	if ((p->p_flag & P_SYSTEM) != 0) {
2081 		PRELE(p);
2082 		return (0);
2083 	}
2084 
2085 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2086 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2087 	error = proc_getenvv(curthread, p, &sb);
2088 	error2 = sbuf_finish(&sb);
2089 	PRELE(p);
2090 	sbuf_delete(&sb);
2091 	return (error != 0 ? error : error2);
2092 }
2093 
2094 /*
2095  * This sysctl allows a process to retrieve ELF auxiliary vector of
2096  * another process.
2097  */
2098 static int
2099 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
2100 {
2101 	int *name = (int *)arg1;
2102 	u_int namelen = arg2;
2103 	struct proc *p;
2104 	struct sbuf sb;
2105 	int error, error2;
2106 
2107 	if (namelen != 1)
2108 		return (EINVAL);
2109 
2110 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2111 	if (error != 0)
2112 		return (error);
2113 	if ((p->p_flag & P_SYSTEM) != 0) {
2114 		PRELE(p);
2115 		return (0);
2116 	}
2117 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2118 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2119 	error = proc_getauxv(curthread, p, &sb);
2120 	error2 = sbuf_finish(&sb);
2121 	PRELE(p);
2122 	sbuf_delete(&sb);
2123 	return (error != 0 ? error : error2);
2124 }
2125 
2126 /*
2127  * This sysctl allows a process to retrieve the path of the executable for
2128  * itself or another process.
2129  */
2130 static int
2131 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
2132 {
2133 	pid_t *pidp = (pid_t *)arg1;
2134 	unsigned int arglen = arg2;
2135 	struct proc *p;
2136 	struct vnode *vp;
2137 	char *retbuf, *freebuf;
2138 	int error;
2139 
2140 	if (arglen != 1)
2141 		return (EINVAL);
2142 	if (*pidp == -1) {	/* -1 means this process */
2143 		p = req->td->td_proc;
2144 	} else {
2145 		error = pget(*pidp, PGET_CANSEE, &p);
2146 		if (error != 0)
2147 			return (error);
2148 	}
2149 
2150 	vp = p->p_textvp;
2151 	if (vp == NULL) {
2152 		if (*pidp != -1)
2153 			PROC_UNLOCK(p);
2154 		return (0);
2155 	}
2156 	vref(vp);
2157 	if (*pidp != -1)
2158 		PROC_UNLOCK(p);
2159 	error = vn_fullpath(req->td, vp, &retbuf, &freebuf);
2160 	vrele(vp);
2161 	if (error)
2162 		return (error);
2163 	error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
2164 	free(freebuf, M_TEMP);
2165 	return (error);
2166 }
2167 
2168 static int
2169 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
2170 {
2171 	struct proc *p;
2172 	char *sv_name;
2173 	int *name;
2174 	int namelen;
2175 	int error;
2176 
2177 	namelen = arg2;
2178 	if (namelen != 1)
2179 		return (EINVAL);
2180 
2181 	name = (int *)arg1;
2182 	error = pget((pid_t)name[0], PGET_CANSEE, &p);
2183 	if (error != 0)
2184 		return (error);
2185 	sv_name = p->p_sysent->sv_name;
2186 	PROC_UNLOCK(p);
2187 	return (sysctl_handle_string(oidp, sv_name, 0, req));
2188 }
2189 
2190 #ifdef KINFO_OVMENTRY_SIZE
2191 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
2192 #endif
2193 
2194 #ifdef COMPAT_FREEBSD7
2195 static int
2196 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
2197 {
2198 	vm_map_entry_t entry, tmp_entry;
2199 	unsigned int last_timestamp;
2200 	char *fullpath, *freepath;
2201 	struct kinfo_ovmentry *kve;
2202 	struct vattr va;
2203 	struct ucred *cred;
2204 	int error, *name;
2205 	struct vnode *vp;
2206 	struct proc *p;
2207 	vm_map_t map;
2208 	struct vmspace *vm;
2209 
2210 	name = (int *)arg1;
2211 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2212 	if (error != 0)
2213 		return (error);
2214 	vm = vmspace_acquire_ref(p);
2215 	if (vm == NULL) {
2216 		PRELE(p);
2217 		return (ESRCH);
2218 	}
2219 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2220 
2221 	map = &vm->vm_map;
2222 	vm_map_lock_read(map);
2223 	VM_MAP_ENTRY_FOREACH(entry, map) {
2224 		vm_object_t obj, tobj, lobj;
2225 		vm_offset_t addr;
2226 
2227 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2228 			continue;
2229 
2230 		bzero(kve, sizeof(*kve));
2231 		kve->kve_structsize = sizeof(*kve);
2232 
2233 		kve->kve_private_resident = 0;
2234 		obj = entry->object.vm_object;
2235 		if (obj != NULL) {
2236 			VM_OBJECT_RLOCK(obj);
2237 			if (obj->shadow_count == 1)
2238 				kve->kve_private_resident =
2239 				    obj->resident_page_count;
2240 		}
2241 		kve->kve_resident = 0;
2242 		addr = entry->start;
2243 		while (addr < entry->end) {
2244 			if (pmap_extract(map->pmap, addr))
2245 				kve->kve_resident++;
2246 			addr += PAGE_SIZE;
2247 		}
2248 
2249 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2250 			if (tobj != obj) {
2251 				VM_OBJECT_RLOCK(tobj);
2252 				kve->kve_offset += tobj->backing_object_offset;
2253 			}
2254 			if (lobj != obj)
2255 				VM_OBJECT_RUNLOCK(lobj);
2256 			lobj = tobj;
2257 		}
2258 
2259 		kve->kve_start = (void*)entry->start;
2260 		kve->kve_end = (void*)entry->end;
2261 		kve->kve_offset += (off_t)entry->offset;
2262 
2263 		if (entry->protection & VM_PROT_READ)
2264 			kve->kve_protection |= KVME_PROT_READ;
2265 		if (entry->protection & VM_PROT_WRITE)
2266 			kve->kve_protection |= KVME_PROT_WRITE;
2267 		if (entry->protection & VM_PROT_EXECUTE)
2268 			kve->kve_protection |= KVME_PROT_EXEC;
2269 
2270 		if (entry->eflags & MAP_ENTRY_COW)
2271 			kve->kve_flags |= KVME_FLAG_COW;
2272 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2273 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2274 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2275 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2276 
2277 		last_timestamp = map->timestamp;
2278 		vm_map_unlock_read(map);
2279 
2280 		kve->kve_fileid = 0;
2281 		kve->kve_fsid = 0;
2282 		freepath = NULL;
2283 		fullpath = "";
2284 		if (lobj) {
2285 			kve->kve_type = vm_object_kvme_type(lobj, &vp);
2286 			if (kve->kve_type == KVME_TYPE_MGTDEVICE)
2287 				kve->kve_type = KVME_TYPE_UNKNOWN;
2288 			if (vp != NULL)
2289 				vref(vp);
2290 			if (lobj != obj)
2291 				VM_OBJECT_RUNLOCK(lobj);
2292 
2293 			kve->kve_ref_count = obj->ref_count;
2294 			kve->kve_shadow_count = obj->shadow_count;
2295 			VM_OBJECT_RUNLOCK(obj);
2296 			if (vp != NULL) {
2297 				vn_fullpath(curthread, vp, &fullpath,
2298 				    &freepath);
2299 				cred = curthread->td_ucred;
2300 				vn_lock(vp, LK_SHARED | LK_RETRY);
2301 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2302 					kve->kve_fileid = va.va_fileid;
2303 					/* truncate */
2304 					kve->kve_fsid = va.va_fsid;
2305 				}
2306 				vput(vp);
2307 			}
2308 		} else {
2309 			kve->kve_type = KVME_TYPE_NONE;
2310 			kve->kve_ref_count = 0;
2311 			kve->kve_shadow_count = 0;
2312 		}
2313 
2314 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2315 		if (freepath != NULL)
2316 			free(freepath, M_TEMP);
2317 
2318 		error = SYSCTL_OUT(req, kve, sizeof(*kve));
2319 		vm_map_lock_read(map);
2320 		if (error)
2321 			break;
2322 		if (last_timestamp != map->timestamp) {
2323 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2324 			entry = tmp_entry;
2325 		}
2326 	}
2327 	vm_map_unlock_read(map);
2328 	vmspace_free(vm);
2329 	PRELE(p);
2330 	free(kve, M_TEMP);
2331 	return (error);
2332 }
2333 #endif	/* COMPAT_FREEBSD7 */
2334 
2335 #ifdef KINFO_VMENTRY_SIZE
2336 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2337 #endif
2338 
2339 void
2340 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry,
2341     int *resident_count, bool *super)
2342 {
2343 	vm_object_t obj, tobj;
2344 	vm_page_t m, m_adv;
2345 	vm_offset_t addr;
2346 	vm_paddr_t pa;
2347 	vm_pindex_t pi, pi_adv, pindex;
2348 
2349 	*super = false;
2350 	*resident_count = 0;
2351 	if (vmmap_skip_res_cnt)
2352 		return;
2353 
2354 	pa = 0;
2355 	obj = entry->object.vm_object;
2356 	addr = entry->start;
2357 	m_adv = NULL;
2358 	pi = OFF_TO_IDX(entry->offset);
2359 	for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) {
2360 		if (m_adv != NULL) {
2361 			m = m_adv;
2362 		} else {
2363 			pi_adv = atop(entry->end - addr);
2364 			pindex = pi;
2365 			for (tobj = obj;; tobj = tobj->backing_object) {
2366 				m = vm_page_find_least(tobj, pindex);
2367 				if (m != NULL) {
2368 					if (m->pindex == pindex)
2369 						break;
2370 					if (pi_adv > m->pindex - pindex) {
2371 						pi_adv = m->pindex - pindex;
2372 						m_adv = m;
2373 					}
2374 				}
2375 				if (tobj->backing_object == NULL)
2376 					goto next;
2377 				pindex += OFF_TO_IDX(tobj->
2378 				    backing_object_offset);
2379 			}
2380 		}
2381 		m_adv = NULL;
2382 		if (m->psind != 0 && addr + pagesizes[1] <= entry->end &&
2383 		    (addr & (pagesizes[1] - 1)) == 0 &&
2384 		    (pmap_mincore(map->pmap, addr, &pa) & MINCORE_SUPER) != 0) {
2385 			*super = true;
2386 			pi_adv = atop(pagesizes[1]);
2387 		} else {
2388 			/*
2389 			 * We do not test the found page on validity.
2390 			 * Either the page is busy and being paged in,
2391 			 * or it was invalidated.  The first case
2392 			 * should be counted as resident, the second
2393 			 * is not so clear; we do account both.
2394 			 */
2395 			pi_adv = 1;
2396 		}
2397 		*resident_count += pi_adv;
2398 next:;
2399 	}
2400 }
2401 
2402 /*
2403  * Must be called with the process locked and will return unlocked.
2404  */
2405 int
2406 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb, ssize_t maxlen, int flags)
2407 {
2408 	vm_map_entry_t entry, tmp_entry;
2409 	struct vattr va;
2410 	vm_map_t map;
2411 	vm_object_t obj, tobj, lobj;
2412 	char *fullpath, *freepath;
2413 	struct kinfo_vmentry *kve;
2414 	struct ucred *cred;
2415 	struct vnode *vp;
2416 	struct vmspace *vm;
2417 	vm_offset_t addr;
2418 	unsigned int last_timestamp;
2419 	int error;
2420 	bool super;
2421 
2422 	PROC_LOCK_ASSERT(p, MA_OWNED);
2423 
2424 	_PHOLD(p);
2425 	PROC_UNLOCK(p);
2426 	vm = vmspace_acquire_ref(p);
2427 	if (vm == NULL) {
2428 		PRELE(p);
2429 		return (ESRCH);
2430 	}
2431 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK | M_ZERO);
2432 
2433 	error = 0;
2434 	map = &vm->vm_map;
2435 	vm_map_lock_read(map);
2436 	VM_MAP_ENTRY_FOREACH(entry, map) {
2437 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2438 			continue;
2439 
2440 		addr = entry->end;
2441 		bzero(kve, sizeof(*kve));
2442 		obj = entry->object.vm_object;
2443 		if (obj != NULL) {
2444 			for (tobj = obj; tobj != NULL;
2445 			    tobj = tobj->backing_object) {
2446 				VM_OBJECT_RLOCK(tobj);
2447 				kve->kve_offset += tobj->backing_object_offset;
2448 				lobj = tobj;
2449 			}
2450 			if (obj->backing_object == NULL)
2451 				kve->kve_private_resident =
2452 				    obj->resident_page_count;
2453 			kern_proc_vmmap_resident(map, entry,
2454 			    &kve->kve_resident, &super);
2455 			if (super)
2456 				kve->kve_flags |= KVME_FLAG_SUPER;
2457 			for (tobj = obj; tobj != NULL;
2458 			    tobj = tobj->backing_object) {
2459 				if (tobj != obj && tobj != lobj)
2460 					VM_OBJECT_RUNLOCK(tobj);
2461 			}
2462 		} else {
2463 			lobj = NULL;
2464 		}
2465 
2466 		kve->kve_start = entry->start;
2467 		kve->kve_end = entry->end;
2468 		kve->kve_offset += entry->offset;
2469 
2470 		if (entry->protection & VM_PROT_READ)
2471 			kve->kve_protection |= KVME_PROT_READ;
2472 		if (entry->protection & VM_PROT_WRITE)
2473 			kve->kve_protection |= KVME_PROT_WRITE;
2474 		if (entry->protection & VM_PROT_EXECUTE)
2475 			kve->kve_protection |= KVME_PROT_EXEC;
2476 
2477 		if (entry->eflags & MAP_ENTRY_COW)
2478 			kve->kve_flags |= KVME_FLAG_COW;
2479 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2480 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2481 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2482 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2483 		if (entry->eflags & MAP_ENTRY_GROWS_UP)
2484 			kve->kve_flags |= KVME_FLAG_GROWS_UP;
2485 		if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2486 			kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2487 		if (entry->eflags & MAP_ENTRY_USER_WIRED)
2488 			kve->kve_flags |= KVME_FLAG_USER_WIRED;
2489 
2490 		last_timestamp = map->timestamp;
2491 		vm_map_unlock_read(map);
2492 
2493 		freepath = NULL;
2494 		fullpath = "";
2495 		if (lobj != NULL) {
2496 			kve->kve_type = vm_object_kvme_type(lobj, &vp);
2497 			if (vp != NULL)
2498 				vref(vp);
2499 			if (lobj != obj)
2500 				VM_OBJECT_RUNLOCK(lobj);
2501 
2502 			kve->kve_ref_count = obj->ref_count;
2503 			kve->kve_shadow_count = obj->shadow_count;
2504 			VM_OBJECT_RUNLOCK(obj);
2505 			if (vp != NULL) {
2506 				vn_fullpath(curthread, vp, &fullpath,
2507 				    &freepath);
2508 				kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2509 				cred = curthread->td_ucred;
2510 				vn_lock(vp, LK_SHARED | LK_RETRY);
2511 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2512 					kve->kve_vn_fileid = va.va_fileid;
2513 					kve->kve_vn_fsid = va.va_fsid;
2514 					kve->kve_vn_fsid_freebsd11 =
2515 					    kve->kve_vn_fsid; /* truncate */
2516 					kve->kve_vn_mode =
2517 					    MAKEIMODE(va.va_type, va.va_mode);
2518 					kve->kve_vn_size = va.va_size;
2519 					kve->kve_vn_rdev = va.va_rdev;
2520 					kve->kve_vn_rdev_freebsd11 =
2521 					    kve->kve_vn_rdev; /* truncate */
2522 					kve->kve_status = KF_ATTR_VALID;
2523 				}
2524 				vput(vp);
2525 			}
2526 		} else {
2527 			kve->kve_type = KVME_TYPE_NONE;
2528 			kve->kve_ref_count = 0;
2529 			kve->kve_shadow_count = 0;
2530 		}
2531 
2532 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2533 		if (freepath != NULL)
2534 			free(freepath, M_TEMP);
2535 
2536 		/* Pack record size down */
2537 		if ((flags & KERN_VMMAP_PACK_KINFO) != 0)
2538 			kve->kve_structsize =
2539 			    offsetof(struct kinfo_vmentry, kve_path) +
2540 			    strlen(kve->kve_path) + 1;
2541 		else
2542 			kve->kve_structsize = sizeof(*kve);
2543 		kve->kve_structsize = roundup(kve->kve_structsize,
2544 		    sizeof(uint64_t));
2545 
2546 		/* Halt filling and truncate rather than exceeding maxlen */
2547 		if (maxlen != -1 && maxlen < kve->kve_structsize) {
2548 			error = 0;
2549 			vm_map_lock_read(map);
2550 			break;
2551 		} else if (maxlen != -1)
2552 			maxlen -= kve->kve_structsize;
2553 
2554 		if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0)
2555 			error = ENOMEM;
2556 		vm_map_lock_read(map);
2557 		if (error != 0)
2558 			break;
2559 		if (last_timestamp != map->timestamp) {
2560 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2561 			entry = tmp_entry;
2562 		}
2563 	}
2564 	vm_map_unlock_read(map);
2565 	vmspace_free(vm);
2566 	PRELE(p);
2567 	free(kve, M_TEMP);
2568 	return (error);
2569 }
2570 
2571 static int
2572 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2573 {
2574 	struct proc *p;
2575 	struct sbuf sb;
2576 	int error, error2, *name;
2577 
2578 	name = (int *)arg1;
2579 	sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req);
2580 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2581 	error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
2582 	if (error != 0) {
2583 		sbuf_delete(&sb);
2584 		return (error);
2585 	}
2586 	error = kern_proc_vmmap_out(p, &sb, -1, KERN_VMMAP_PACK_KINFO);
2587 	error2 = sbuf_finish(&sb);
2588 	sbuf_delete(&sb);
2589 	return (error != 0 ? error : error2);
2590 }
2591 
2592 #if defined(STACK) || defined(DDB)
2593 static int
2594 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2595 {
2596 	struct kinfo_kstack *kkstp;
2597 	int error, i, *name, numthreads;
2598 	lwpid_t *lwpidarray;
2599 	struct thread *td;
2600 	struct stack *st;
2601 	struct sbuf sb;
2602 	struct proc *p;
2603 
2604 	name = (int *)arg1;
2605 	error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2606 	if (error != 0)
2607 		return (error);
2608 
2609 	kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2610 	st = stack_create(M_WAITOK);
2611 
2612 	lwpidarray = NULL;
2613 	PROC_LOCK(p);
2614 	do {
2615 		if (lwpidarray != NULL) {
2616 			free(lwpidarray, M_TEMP);
2617 			lwpidarray = NULL;
2618 		}
2619 		numthreads = p->p_numthreads;
2620 		PROC_UNLOCK(p);
2621 		lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2622 		    M_WAITOK | M_ZERO);
2623 		PROC_LOCK(p);
2624 	} while (numthreads < p->p_numthreads);
2625 
2626 	/*
2627 	 * XXXRW: During the below loop, execve(2) and countless other sorts
2628 	 * of changes could have taken place.  Should we check to see if the
2629 	 * vmspace has been replaced, or the like, in order to prevent
2630 	 * giving a snapshot that spans, say, execve(2), with some threads
2631 	 * before and some after?  Among other things, the credentials could
2632 	 * have changed, in which case the right to extract debug info might
2633 	 * no longer be assured.
2634 	 */
2635 	i = 0;
2636 	FOREACH_THREAD_IN_PROC(p, td) {
2637 		KASSERT(i < numthreads,
2638 		    ("sysctl_kern_proc_kstack: numthreads"));
2639 		lwpidarray[i] = td->td_tid;
2640 		i++;
2641 	}
2642 	numthreads = i;
2643 	for (i = 0; i < numthreads; i++) {
2644 		td = thread_find(p, lwpidarray[i]);
2645 		if (td == NULL) {
2646 			continue;
2647 		}
2648 		bzero(kkstp, sizeof(*kkstp));
2649 		(void)sbuf_new(&sb, kkstp->kkst_trace,
2650 		    sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2651 		thread_lock(td);
2652 		kkstp->kkst_tid = td->td_tid;
2653 		if (TD_IS_SWAPPED(td))
2654 			kkstp->kkst_state = KKST_STATE_SWAPPED;
2655 		else if (stack_save_td(st, td) == 0)
2656 			kkstp->kkst_state = KKST_STATE_STACKOK;
2657 		else
2658 			kkstp->kkst_state = KKST_STATE_RUNNING;
2659 		thread_unlock(td);
2660 		PROC_UNLOCK(p);
2661 		stack_sbuf_print(&sb, st);
2662 		sbuf_finish(&sb);
2663 		sbuf_delete(&sb);
2664 		error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2665 		PROC_LOCK(p);
2666 		if (error)
2667 			break;
2668 	}
2669 	_PRELE(p);
2670 	PROC_UNLOCK(p);
2671 	if (lwpidarray != NULL)
2672 		free(lwpidarray, M_TEMP);
2673 	stack_destroy(st);
2674 	free(kkstp, M_TEMP);
2675 	return (error);
2676 }
2677 #endif
2678 
2679 /*
2680  * This sysctl allows a process to retrieve the full list of groups from
2681  * itself or another process.
2682  */
2683 static int
2684 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2685 {
2686 	pid_t *pidp = (pid_t *)arg1;
2687 	unsigned int arglen = arg2;
2688 	struct proc *p;
2689 	struct ucred *cred;
2690 	int error;
2691 
2692 	if (arglen != 1)
2693 		return (EINVAL);
2694 	if (*pidp == -1) {	/* -1 means this process */
2695 		p = req->td->td_proc;
2696 		PROC_LOCK(p);
2697 	} else {
2698 		error = pget(*pidp, PGET_CANSEE, &p);
2699 		if (error != 0)
2700 			return (error);
2701 	}
2702 
2703 	cred = crhold(p->p_ucred);
2704 	PROC_UNLOCK(p);
2705 
2706 	error = SYSCTL_OUT(req, cred->cr_groups,
2707 	    cred->cr_ngroups * sizeof(gid_t));
2708 	crfree(cred);
2709 	return (error);
2710 }
2711 
2712 /*
2713  * This sysctl allows a process to retrieve or/and set the resource limit for
2714  * another process.
2715  */
2716 static int
2717 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2718 {
2719 	int *name = (int *)arg1;
2720 	u_int namelen = arg2;
2721 	struct rlimit rlim;
2722 	struct proc *p;
2723 	u_int which;
2724 	int flags, error;
2725 
2726 	if (namelen != 2)
2727 		return (EINVAL);
2728 
2729 	which = (u_int)name[1];
2730 	if (which >= RLIM_NLIMITS)
2731 		return (EINVAL);
2732 
2733 	if (req->newptr != NULL && req->newlen != sizeof(rlim))
2734 		return (EINVAL);
2735 
2736 	flags = PGET_HOLD | PGET_NOTWEXIT;
2737 	if (req->newptr != NULL)
2738 		flags |= PGET_CANDEBUG;
2739 	else
2740 		flags |= PGET_CANSEE;
2741 	error = pget((pid_t)name[0], flags, &p);
2742 	if (error != 0)
2743 		return (error);
2744 
2745 	/*
2746 	 * Retrieve limit.
2747 	 */
2748 	if (req->oldptr != NULL) {
2749 		PROC_LOCK(p);
2750 		lim_rlimit_proc(p, which, &rlim);
2751 		PROC_UNLOCK(p);
2752 	}
2753 	error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
2754 	if (error != 0)
2755 		goto errout;
2756 
2757 	/*
2758 	 * Set limit.
2759 	 */
2760 	if (req->newptr != NULL) {
2761 		error = SYSCTL_IN(req, &rlim, sizeof(rlim));
2762 		if (error == 0)
2763 			error = kern_proc_setrlimit(curthread, p, which, &rlim);
2764 	}
2765 
2766 errout:
2767 	PRELE(p);
2768 	return (error);
2769 }
2770 
2771 /*
2772  * This sysctl allows a process to retrieve ps_strings structure location of
2773  * another process.
2774  */
2775 static int
2776 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
2777 {
2778 	int *name = (int *)arg1;
2779 	u_int namelen = arg2;
2780 	struct proc *p;
2781 	vm_offset_t ps_strings;
2782 	int error;
2783 #ifdef COMPAT_FREEBSD32
2784 	uint32_t ps_strings32;
2785 #endif
2786 
2787 	if (namelen != 1)
2788 		return (EINVAL);
2789 
2790 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2791 	if (error != 0)
2792 		return (error);
2793 #ifdef COMPAT_FREEBSD32
2794 	if ((req->flags & SCTL_MASK32) != 0) {
2795 		/*
2796 		 * We return 0 if the 32 bit emulation request is for a 64 bit
2797 		 * process.
2798 		 */
2799 		ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
2800 		    PTROUT(p->p_sysent->sv_psstrings) : 0;
2801 		PROC_UNLOCK(p);
2802 		error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
2803 		return (error);
2804 	}
2805 #endif
2806 	ps_strings = p->p_sysent->sv_psstrings;
2807 	PROC_UNLOCK(p);
2808 	error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
2809 	return (error);
2810 }
2811 
2812 /*
2813  * This sysctl allows a process to retrieve umask of another process.
2814  */
2815 static int
2816 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
2817 {
2818 	int *name = (int *)arg1;
2819 	u_int namelen = arg2;
2820 	struct proc *p;
2821 	int error;
2822 	u_short fd_cmask;
2823 	pid_t pid;
2824 
2825 	if (namelen != 1)
2826 		return (EINVAL);
2827 
2828 	pid = (pid_t)name[0];
2829 	p = curproc;
2830 	if (pid == p->p_pid || pid == 0) {
2831 		fd_cmask = p->p_fd->fd_cmask;
2832 		goto out;
2833 	}
2834 
2835 	error = pget(pid, PGET_WANTREAD, &p);
2836 	if (error != 0)
2837 		return (error);
2838 
2839 	fd_cmask = p->p_fd->fd_cmask;
2840 	PRELE(p);
2841 out:
2842 	error = SYSCTL_OUT(req, &fd_cmask, sizeof(fd_cmask));
2843 	return (error);
2844 }
2845 
2846 /*
2847  * This sysctl allows a process to set and retrieve binary osreldate of
2848  * another process.
2849  */
2850 static int
2851 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
2852 {
2853 	int *name = (int *)arg1;
2854 	u_int namelen = arg2;
2855 	struct proc *p;
2856 	int flags, error, osrel;
2857 
2858 	if (namelen != 1)
2859 		return (EINVAL);
2860 
2861 	if (req->newptr != NULL && req->newlen != sizeof(osrel))
2862 		return (EINVAL);
2863 
2864 	flags = PGET_HOLD | PGET_NOTWEXIT;
2865 	if (req->newptr != NULL)
2866 		flags |= PGET_CANDEBUG;
2867 	else
2868 		flags |= PGET_CANSEE;
2869 	error = pget((pid_t)name[0], flags, &p);
2870 	if (error != 0)
2871 		return (error);
2872 
2873 	error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
2874 	if (error != 0)
2875 		goto errout;
2876 
2877 	if (req->newptr != NULL) {
2878 		error = SYSCTL_IN(req, &osrel, sizeof(osrel));
2879 		if (error != 0)
2880 			goto errout;
2881 		if (osrel < 0) {
2882 			error = EINVAL;
2883 			goto errout;
2884 		}
2885 		p->p_osrel = osrel;
2886 	}
2887 errout:
2888 	PRELE(p);
2889 	return (error);
2890 }
2891 
2892 static int
2893 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS)
2894 {
2895 	int *name = (int *)arg1;
2896 	u_int namelen = arg2;
2897 	struct proc *p;
2898 	struct kinfo_sigtramp kst;
2899 	const struct sysentvec *sv;
2900 	int error;
2901 #ifdef COMPAT_FREEBSD32
2902 	struct kinfo_sigtramp32 kst32;
2903 #endif
2904 
2905 	if (namelen != 1)
2906 		return (EINVAL);
2907 
2908 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2909 	if (error != 0)
2910 		return (error);
2911 	sv = p->p_sysent;
2912 #ifdef COMPAT_FREEBSD32
2913 	if ((req->flags & SCTL_MASK32) != 0) {
2914 		bzero(&kst32, sizeof(kst32));
2915 		if (SV_PROC_FLAG(p, SV_ILP32)) {
2916 			if (sv->sv_sigcode_base != 0) {
2917 				kst32.ksigtramp_start = sv->sv_sigcode_base;
2918 				kst32.ksigtramp_end = sv->sv_sigcode_base +
2919 				    *sv->sv_szsigcode;
2920 			} else {
2921 				kst32.ksigtramp_start = sv->sv_psstrings -
2922 				    *sv->sv_szsigcode;
2923 				kst32.ksigtramp_end = sv->sv_psstrings;
2924 			}
2925 		}
2926 		PROC_UNLOCK(p);
2927 		error = SYSCTL_OUT(req, &kst32, sizeof(kst32));
2928 		return (error);
2929 	}
2930 #endif
2931 	bzero(&kst, sizeof(kst));
2932 	if (sv->sv_sigcode_base != 0) {
2933 		kst.ksigtramp_start = (char *)sv->sv_sigcode_base;
2934 		kst.ksigtramp_end = (char *)sv->sv_sigcode_base +
2935 		    *sv->sv_szsigcode;
2936 	} else {
2937 		kst.ksigtramp_start = (char *)sv->sv_psstrings -
2938 		    *sv->sv_szsigcode;
2939 		kst.ksigtramp_end = (char *)sv->sv_psstrings;
2940 	}
2941 	PROC_UNLOCK(p);
2942 	error = SYSCTL_OUT(req, &kst, sizeof(kst));
2943 	return (error);
2944 }
2945 
2946 static int
2947 sysctl_kern_proc_sigfastblk(SYSCTL_HANDLER_ARGS)
2948 {
2949 	int *name = (int *)arg1;
2950 	u_int namelen = arg2;
2951 	pid_t pid;
2952 	struct proc *p;
2953 	struct thread *td1;
2954 	uintptr_t addr;
2955 #ifdef COMPAT_FREEBSD32
2956 	uint32_t addr32;
2957 #endif
2958 	int error;
2959 
2960 	if (namelen != 1 || req->newptr != NULL)
2961 		return (EINVAL);
2962 
2963 	pid = (pid_t)name[0];
2964 	error = pget(pid, PGET_HOLD | PGET_NOTWEXIT | PGET_CANDEBUG, &p);
2965 	if (error != 0)
2966 		return (error);
2967 
2968 	PROC_LOCK(p);
2969 #ifdef COMPAT_FREEBSD32
2970 	if (SV_CURPROC_FLAG(SV_ILP32)) {
2971 		if (!SV_PROC_FLAG(p, SV_ILP32)) {
2972 			error = EINVAL;
2973 			goto errlocked;
2974 		}
2975 	}
2976 #endif
2977 	if (pid <= PID_MAX) {
2978 		td1 = FIRST_THREAD_IN_PROC(p);
2979 	} else {
2980 		FOREACH_THREAD_IN_PROC(p, td1) {
2981 			if (td1->td_tid == pid)
2982 				break;
2983 		}
2984 	}
2985 	if (td1 == NULL) {
2986 		error = ESRCH;
2987 		goto errlocked;
2988 	}
2989 	/*
2990 	 * The access to the private thread flags.  It is fine as far
2991 	 * as no out-of-thin-air values are read from td_pflags, and
2992 	 * usermode read of the td_sigblock_ptr is racy inherently,
2993 	 * since target process might have already changed it
2994 	 * meantime.
2995 	 */
2996 	if ((td1->td_pflags & TDP_SIGFASTBLOCK) != 0)
2997 		addr = (uintptr_t)td1->td_sigblock_ptr;
2998 	else
2999 		error = ENOTTY;
3000 
3001 errlocked:
3002 	_PRELE(p);
3003 	PROC_UNLOCK(p);
3004 	if (error != 0)
3005 		return (error);
3006 
3007 #ifdef COMPAT_FREEBSD32
3008 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3009 		addr32 = addr;
3010 		error = SYSCTL_OUT(req, &addr32, sizeof(addr32));
3011 	} else
3012 #endif
3013 		error = SYSCTL_OUT(req, &addr, sizeof(addr));
3014 	return (error);
3015 }
3016 
3017 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,  0,
3018     "Process table");
3019 
3020 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
3021 	CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
3022 	"Return entire process table");
3023 
3024 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3025 	sysctl_kern_proc, "Process table");
3026 
3027 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
3028 	sysctl_kern_proc, "Process table");
3029 
3030 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3031 	sysctl_kern_proc, "Process table");
3032 
3033 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
3034 	CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3035 
3036 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
3037 	sysctl_kern_proc, "Process table");
3038 
3039 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3040 	sysctl_kern_proc, "Process table");
3041 
3042 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3043 	sysctl_kern_proc, "Process table");
3044 
3045 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3046 	sysctl_kern_proc, "Process table");
3047 
3048 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
3049 	sysctl_kern_proc, "Return process table, no threads");
3050 
3051 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
3052 	CTLFLAG_RW | CTLFLAG_CAPWR | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
3053 	sysctl_kern_proc_args, "Process argument list");
3054 
3055 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
3056 	sysctl_kern_proc_env, "Process environment");
3057 
3058 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
3059 	CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
3060 
3061 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
3062 	CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
3063 
3064 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
3065 	CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
3066 	"Process syscall vector name (ABI type)");
3067 
3068 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
3069 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3070 
3071 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
3072 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3073 
3074 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
3075 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3076 
3077 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
3078 	sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3079 
3080 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
3081 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3082 
3083 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
3084 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3085 
3086 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
3087 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3088 
3089 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
3090 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3091 
3092 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
3093 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
3094 	"Return process table, no threads");
3095 
3096 #ifdef COMPAT_FREEBSD7
3097 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
3098 	CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
3099 #endif
3100 
3101 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
3102 	CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
3103 
3104 #if defined(STACK) || defined(DDB)
3105 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
3106 	CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
3107 #endif
3108 
3109 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
3110 	CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
3111 
3112 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
3113 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
3114 	"Process resource limits");
3115 
3116 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
3117 	CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
3118 	"Process ps_strings location");
3119 
3120 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
3121 	CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
3122 
3123 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
3124 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
3125 	"Process binary osreldate");
3126 
3127 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD |
3128 	CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp,
3129 	"Process signal trampoline location");
3130 
3131 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGFASTBLK, sigfastblk, CTLFLAG_RD |
3132 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_sigfastblk,
3133 	"Thread sigfastblock address");
3134 
3135 int allproc_gen;
3136 
3137 /*
3138  * stop_all_proc() purpose is to stop all process which have usermode,
3139  * except current process for obvious reasons.  This makes it somewhat
3140  * unreliable when invoked from multithreaded process.  The service
3141  * must not be user-callable anyway.
3142  */
3143 void
3144 stop_all_proc(void)
3145 {
3146 	struct proc *cp, *p;
3147 	int r, gen;
3148 	bool restart, seen_stopped, seen_exiting, stopped_some;
3149 
3150 	cp = curproc;
3151 allproc_loop:
3152 	sx_xlock(&allproc_lock);
3153 	gen = allproc_gen;
3154 	seen_exiting = seen_stopped = stopped_some = restart = false;
3155 	LIST_REMOVE(cp, p_list);
3156 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3157 	for (;;) {
3158 		p = LIST_NEXT(cp, p_list);
3159 		if (p == NULL)
3160 			break;
3161 		LIST_REMOVE(cp, p_list);
3162 		LIST_INSERT_AFTER(p, cp, p_list);
3163 		PROC_LOCK(p);
3164 		if ((p->p_flag & (P_KPROC | P_SYSTEM | P_TOTAL_STOP)) != 0) {
3165 			PROC_UNLOCK(p);
3166 			continue;
3167 		}
3168 		if ((p->p_flag & P_WEXIT) != 0) {
3169 			seen_exiting = true;
3170 			PROC_UNLOCK(p);
3171 			continue;
3172 		}
3173 		if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) {
3174 			/*
3175 			 * Stopped processes are tolerated when there
3176 			 * are no other processes which might continue
3177 			 * them.  P_STOPPED_SINGLE but not
3178 			 * P_TOTAL_STOP process still has at least one
3179 			 * thread running.
3180 			 */
3181 			seen_stopped = true;
3182 			PROC_UNLOCK(p);
3183 			continue;
3184 		}
3185 		sx_xunlock(&allproc_lock);
3186 		_PHOLD(p);
3187 		r = thread_single(p, SINGLE_ALLPROC);
3188 		if (r != 0)
3189 			restart = true;
3190 		else
3191 			stopped_some = true;
3192 		_PRELE(p);
3193 		PROC_UNLOCK(p);
3194 		sx_xlock(&allproc_lock);
3195 	}
3196 	/* Catch forked children we did not see in iteration. */
3197 	if (gen != allproc_gen)
3198 		restart = true;
3199 	sx_xunlock(&allproc_lock);
3200 	if (restart || stopped_some || seen_exiting || seen_stopped) {
3201 		kern_yield(PRI_USER);
3202 		goto allproc_loop;
3203 	}
3204 }
3205 
3206 void
3207 resume_all_proc(void)
3208 {
3209 	struct proc *cp, *p;
3210 
3211 	cp = curproc;
3212 	sx_xlock(&allproc_lock);
3213 again:
3214 	LIST_REMOVE(cp, p_list);
3215 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3216 	for (;;) {
3217 		p = LIST_NEXT(cp, p_list);
3218 		if (p == NULL)
3219 			break;
3220 		LIST_REMOVE(cp, p_list);
3221 		LIST_INSERT_AFTER(p, cp, p_list);
3222 		PROC_LOCK(p);
3223 		if ((p->p_flag & P_TOTAL_STOP) != 0) {
3224 			sx_xunlock(&allproc_lock);
3225 			_PHOLD(p);
3226 			thread_single_end(p, SINGLE_ALLPROC);
3227 			_PRELE(p);
3228 			PROC_UNLOCK(p);
3229 			sx_xlock(&allproc_lock);
3230 		} else {
3231 			PROC_UNLOCK(p);
3232 		}
3233 	}
3234 	/*  Did the loop above missed any stopped process ? */
3235 	FOREACH_PROC_IN_SYSTEM(p) {
3236 		/* No need for proc lock. */
3237 		if ((p->p_flag & P_TOTAL_STOP) != 0)
3238 			goto again;
3239 	}
3240 	sx_xunlock(&allproc_lock);
3241 }
3242 
3243 /* #define	TOTAL_STOP_DEBUG	1 */
3244 #ifdef TOTAL_STOP_DEBUG
3245 volatile static int ap_resume;
3246 #include <sys/mount.h>
3247 
3248 static int
3249 sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS)
3250 {
3251 	int error, val;
3252 
3253 	val = 0;
3254 	ap_resume = 0;
3255 	error = sysctl_handle_int(oidp, &val, 0, req);
3256 	if (error != 0 || req->newptr == NULL)
3257 		return (error);
3258 	if (val != 0) {
3259 		stop_all_proc();
3260 		syncer_suspend();
3261 		while (ap_resume == 0)
3262 			;
3263 		syncer_resume();
3264 		resume_all_proc();
3265 	}
3266 	return (0);
3267 }
3268 
3269 SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW |
3270     CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0,
3271     sysctl_debug_stop_all_proc, "I",
3272     "");
3273 #endif
3274