xref: /freebsd/sys/kern/kern_proc.c (revision 43d1e6ee299ad4e143d90d3ad374d1c24bd3306f)
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 proclist zombproc;
128 struct sx __exclusive_cache_line allproc_lock;
129 struct sx __exclusive_cache_line zombproc_lock;
130 struct sx __exclusive_cache_line proctree_lock;
131 struct mtx __exclusive_cache_line ppeers_lock;
132 struct mtx __exclusive_cache_line procid_lock;
133 uma_zone_t proc_zone;
134 
135 /*
136  * The offset of various fields in struct proc and struct thread.
137  * These are used by kernel debuggers to enumerate kernel threads and
138  * processes.
139  */
140 const int proc_off_p_pid = offsetof(struct proc, p_pid);
141 const int proc_off_p_comm = offsetof(struct proc, p_comm);
142 const int proc_off_p_list = offsetof(struct proc, p_list);
143 const int proc_off_p_threads = offsetof(struct proc, p_threads);
144 const int thread_off_td_tid = offsetof(struct thread, td_tid);
145 const int thread_off_td_name = offsetof(struct thread, td_name);
146 const int thread_off_td_oncpu = offsetof(struct thread, td_oncpu);
147 const int thread_off_td_pcb = offsetof(struct thread, td_pcb);
148 const int thread_off_td_plist = offsetof(struct thread, td_plist);
149 
150 EVENTHANDLER_LIST_DEFINE(process_ctor);
151 EVENTHANDLER_LIST_DEFINE(process_dtor);
152 EVENTHANDLER_LIST_DEFINE(process_init);
153 EVENTHANDLER_LIST_DEFINE(process_fini);
154 EVENTHANDLER_LIST_DEFINE(process_exit);
155 EVENTHANDLER_LIST_DEFINE(process_fork);
156 EVENTHANDLER_LIST_DEFINE(process_exec);
157 
158 EVENTHANDLER_LIST_DECLARE(thread_ctor);
159 EVENTHANDLER_LIST_DECLARE(thread_dtor);
160 
161 int kstack_pages = KSTACK_PAGES;
162 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0,
163     "Kernel stack size in pages");
164 static int vmmap_skip_res_cnt = 0;
165 SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW,
166     &vmmap_skip_res_cnt, 0,
167     "Skip calculation of the pages resident count in kern.proc.vmmap");
168 
169 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
170 #ifdef COMPAT_FREEBSD32
171 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
172 #endif
173 
174 /*
175  * Initialize global process hashing structures.
176  */
177 void
178 procinit(void)
179 {
180 	u_long i;
181 
182 	sx_init(&allproc_lock, "allproc");
183 	sx_init(&zombproc_lock, "zombproc");
184 	sx_init(&proctree_lock, "proctree");
185 	mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
186 	mtx_init(&procid_lock, "procid", NULL, MTX_DEF);
187 	LIST_INIT(&allproc);
188 	LIST_INIT(&zombproc);
189 	pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
190 	pidhashlock = (pidhash + 1) / 64;
191 	if (pidhashlock > 0)
192 		pidhashlock--;
193 	pidhashtbl_lock = malloc(sizeof(*pidhashtbl_lock) * (pidhashlock + 1),
194 	    M_PROC, M_WAITOK | M_ZERO);
195 	for (i = 0; i < pidhashlock + 1; i++)
196 		sx_init_flags(&pidhashtbl_lock[i], "pidhash", SX_DUPOK);
197 	pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
198 	proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
199 	    proc_ctor, proc_dtor, proc_init, proc_fini,
200 	    UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
201 	uihashinit();
202 }
203 
204 /*
205  * Prepare a proc for use.
206  */
207 static int
208 proc_ctor(void *mem, int size, void *arg, int flags)
209 {
210 	struct proc *p;
211 	struct thread *td;
212 
213 	p = (struct proc *)mem;
214 	EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
215 	td = FIRST_THREAD_IN_PROC(p);
216 	if (td != NULL) {
217 		/* Make sure all thread constructors are executed */
218 		EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
219 	}
220 	return (0);
221 }
222 
223 /*
224  * Reclaim a proc after use.
225  */
226 static void
227 proc_dtor(void *mem, int size, void *arg)
228 {
229 	struct proc *p;
230 	struct thread *td;
231 
232 	/* INVARIANTS checks go here */
233 	p = (struct proc *)mem;
234 	td = FIRST_THREAD_IN_PROC(p);
235 	if (td != NULL) {
236 #ifdef INVARIANTS
237 		KASSERT((p->p_numthreads == 1),
238 		    ("bad number of threads in exiting process"));
239 		KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
240 #endif
241 		/* Free all OSD associated to this thread. */
242 		osd_thread_exit(td);
243 		td_softdep_cleanup(td);
244 		MPASS(td->td_su == NULL);
245 
246 		/* Make sure all thread destructors are executed */
247 		EVENTHANDLER_DIRECT_INVOKE(thread_dtor, td);
248 	}
249 	EVENTHANDLER_DIRECT_INVOKE(process_dtor, p);
250 	if (p->p_ksi != NULL)
251 		KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
252 }
253 
254 /*
255  * Initialize type-stable parts of a proc (when newly created).
256  */
257 static int
258 proc_init(void *mem, int size, int flags)
259 {
260 	struct proc *p;
261 
262 	p = (struct proc *)mem;
263 	mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK | MTX_NEW);
264 	mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_NEW);
265 	mtx_init(&p->p_statmtx, "pstatl", NULL, MTX_SPIN | MTX_NEW);
266 	mtx_init(&p->p_itimmtx, "pitiml", NULL, MTX_SPIN | MTX_NEW);
267 	mtx_init(&p->p_profmtx, "pprofl", NULL, MTX_SPIN | MTX_NEW);
268 	cv_init(&p->p_pwait, "ppwait");
269 	TAILQ_INIT(&p->p_threads);	     /* all threads in proc */
270 	EVENTHANDLER_DIRECT_INVOKE(process_init, p);
271 	p->p_stats = pstats_alloc();
272 	p->p_pgrp = NULL;
273 	return (0);
274 }
275 
276 /*
277  * UMA should ensure that this function is never called.
278  * Freeing a proc structure would violate type stability.
279  */
280 static void
281 proc_fini(void *mem, int size)
282 {
283 #ifdef notnow
284 	struct proc *p;
285 
286 	p = (struct proc *)mem;
287 	EVENTHANDLER_DIRECT_INVOKE(process_fini, p);
288 	pstats_free(p->p_stats);
289 	thread_free(FIRST_THREAD_IN_PROC(p));
290 	mtx_destroy(&p->p_mtx);
291 	if (p->p_ksi != NULL)
292 		ksiginfo_free(p->p_ksi);
293 #else
294 	panic("proc reclaimed");
295 #endif
296 }
297 
298 /*
299  * PID space management.
300  *
301  * These bitmaps are used by fork_findpid.
302  */
303 bitstr_t bit_decl(proc_id_pidmap, PID_MAX);
304 bitstr_t bit_decl(proc_id_grpidmap, PID_MAX);
305 bitstr_t bit_decl(proc_id_sessidmap, PID_MAX);
306 bitstr_t bit_decl(proc_id_reapmap, PID_MAX);
307 
308 static bitstr_t *proc_id_array[] = {
309 	proc_id_pidmap,
310 	proc_id_grpidmap,
311 	proc_id_sessidmap,
312 	proc_id_reapmap,
313 };
314 
315 void
316 proc_id_set(int type, pid_t id)
317 {
318 
319 	KASSERT(type >= 0 && type < nitems(proc_id_array),
320 	    ("invalid type %d\n", type));
321 	mtx_lock(&procid_lock);
322 	KASSERT(bit_test(proc_id_array[type], id) == 0,
323 	    ("bit %d already set in %d\n", id, type));
324 	bit_set(proc_id_array[type], id);
325 	mtx_unlock(&procid_lock);
326 }
327 
328 void
329 proc_id_set_cond(int type, pid_t id)
330 {
331 
332 	KASSERT(type >= 0 && type < nitems(proc_id_array),
333 	    ("invalid type %d\n", type));
334 	if (bit_test(proc_id_array[type], id))
335 		return;
336 	mtx_lock(&procid_lock);
337 	bit_set(proc_id_array[type], id);
338 	mtx_unlock(&procid_lock);
339 }
340 
341 void
342 proc_id_clear(int type, pid_t id)
343 {
344 
345 	KASSERT(type >= 0 && type < nitems(proc_id_array),
346 	    ("invalid type %d\n", type));
347 	mtx_lock(&procid_lock);
348 	KASSERT(bit_test(proc_id_array[type], id) != 0,
349 	    ("bit %d not set in %d\n", id, type));
350 	bit_clear(proc_id_array[type], id);
351 	mtx_unlock(&procid_lock);
352 }
353 
354 /*
355  * Is p an inferior of the current process?
356  */
357 int
358 inferior(struct proc *p)
359 {
360 
361 	sx_assert(&proctree_lock, SX_LOCKED);
362 	PROC_LOCK_ASSERT(p, MA_OWNED);
363 	for (; p != curproc; p = proc_realparent(p)) {
364 		if (p->p_pid == 0)
365 			return (0);
366 	}
367 	return (1);
368 }
369 
370 /*
371  * Shared lock all the pid hash lists.
372  */
373 void
374 pidhash_slockall(void)
375 {
376 	u_long i;
377 
378 	for (i = 0; i < pidhashlock + 1; i++)
379 		sx_slock(&pidhashtbl_lock[i]);
380 }
381 
382 /*
383  * Shared unlock all the pid hash lists.
384  */
385 void
386 pidhash_sunlockall(void)
387 {
388 	u_long i;
389 
390 	for (i = 0; i < pidhashlock + 1; i++)
391 		sx_sunlock(&pidhashtbl_lock[i]);
392 }
393 
394 /*
395  * Similar to pfind_any(), this function finds zombies.
396  */
397 struct proc *
398 pfind_any_locked(pid_t pid)
399 {
400 	struct proc *p;
401 
402 	sx_assert(PIDHASHLOCK(pid), SX_LOCKED);
403 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
404 		if (p->p_pid == pid) {
405 			PROC_LOCK(p);
406 			if (p->p_state == PRS_NEW) {
407 				PROC_UNLOCK(p);
408 				p = NULL;
409 			}
410 			break;
411 		}
412 	}
413 	return (p);
414 }
415 
416 /*
417  * Locate a process by number.
418  *
419  * By not returning processes in the PRS_NEW state, we allow callers to avoid
420  * testing for that condition to avoid dereferencing p_ucred, et al.
421  */
422 static __always_inline struct proc *
423 _pfind(pid_t pid, bool zombie)
424 {
425 	struct proc *p;
426 
427 	p = curproc;
428 	if (p->p_pid == pid) {
429 		PROC_LOCK(p);
430 		return (p);
431 	}
432 	sx_slock(PIDHASHLOCK(pid));
433 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
434 		if (p->p_pid == pid) {
435 			PROC_LOCK(p);
436 			if (p->p_state == PRS_NEW ||
437 			    (!zombie && p->p_state == PRS_ZOMBIE)) {
438 				PROC_UNLOCK(p);
439 				p = NULL;
440 			}
441 			break;
442 		}
443 	}
444 	sx_sunlock(PIDHASHLOCK(pid));
445 	return (p);
446 }
447 
448 struct proc *
449 pfind(pid_t pid)
450 {
451 
452 	return (_pfind(pid, false));
453 }
454 
455 /*
456  * Same as pfind but allow zombies.
457  */
458 struct proc *
459 pfind_any(pid_t pid)
460 {
461 
462 	return (_pfind(pid, true));
463 }
464 
465 static struct proc *
466 pfind_tid(pid_t tid)
467 {
468 	struct proc *p;
469 	struct thread *td;
470 
471 	sx_slock(&allproc_lock);
472 	FOREACH_PROC_IN_SYSTEM(p) {
473 		PROC_LOCK(p);
474 		if (p->p_state == PRS_NEW) {
475 			PROC_UNLOCK(p);
476 			continue;
477 		}
478 		FOREACH_THREAD_IN_PROC(p, td) {
479 			if (td->td_tid == tid)
480 				goto found;
481 		}
482 		PROC_UNLOCK(p);
483 	}
484 found:
485 	sx_sunlock(&allproc_lock);
486 	return (p);
487 }
488 
489 /*
490  * Locate a process group by number.
491  * The caller must hold proctree_lock.
492  */
493 struct pgrp *
494 pgfind(pid_t pgid)
495 {
496 	struct pgrp *pgrp;
497 
498 	sx_assert(&proctree_lock, SX_LOCKED);
499 
500 	LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
501 		if (pgrp->pg_id == pgid) {
502 			PGRP_LOCK(pgrp);
503 			return (pgrp);
504 		}
505 	}
506 	return (NULL);
507 }
508 
509 /*
510  * Locate process and do additional manipulations, depending on flags.
511  */
512 int
513 pget(pid_t pid, int flags, struct proc **pp)
514 {
515 	struct proc *p;
516 	int error;
517 
518 	p = curproc;
519 	if (p->p_pid == pid) {
520 		PROC_LOCK(p);
521 	} else {
522 		p = NULL;
523 		if (pid <= PID_MAX) {
524 			if ((flags & PGET_NOTWEXIT) == 0)
525 				p = pfind_any(pid);
526 			else
527 				p = pfind(pid);
528 		} else if ((flags & PGET_NOTID) == 0) {
529 			p = pfind_tid(pid);
530 		}
531 		if (p == NULL)
532 			return (ESRCH);
533 		if ((flags & PGET_CANSEE) != 0) {
534 			error = p_cansee(curthread, p);
535 			if (error != 0)
536 				goto errout;
537 		}
538 	}
539 	if ((flags & PGET_CANDEBUG) != 0) {
540 		error = p_candebug(curthread, p);
541 		if (error != 0)
542 			goto errout;
543 	}
544 	if ((flags & PGET_ISCURRENT) != 0 && curproc != p) {
545 		error = EPERM;
546 		goto errout;
547 	}
548 	if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) {
549 		error = ESRCH;
550 		goto errout;
551 	}
552 	if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) {
553 		/*
554 		 * XXXRW: Not clear ESRCH is the right error during proc
555 		 * execve().
556 		 */
557 		error = ESRCH;
558 		goto errout;
559 	}
560 	if ((flags & PGET_HOLD) != 0) {
561 		_PHOLD(p);
562 		PROC_UNLOCK(p);
563 	}
564 	*pp = p;
565 	return (0);
566 errout:
567 	PROC_UNLOCK(p);
568 	return (error);
569 }
570 
571 /*
572  * Create a new process group.
573  * pgid must be equal to the pid of p.
574  * Begin a new session if required.
575  */
576 int
577 enterpgrp(struct proc *p, pid_t pgid, struct pgrp *pgrp, struct session *sess)
578 {
579 
580 	sx_assert(&proctree_lock, SX_XLOCKED);
581 
582 	KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
583 	KASSERT(p->p_pid == pgid,
584 	    ("enterpgrp: new pgrp and pid != pgid"));
585 	KASSERT(pgfind(pgid) == NULL,
586 	    ("enterpgrp: pgrp with pgid exists"));
587 	KASSERT(!SESS_LEADER(p),
588 	    ("enterpgrp: session leader attempted setpgrp"));
589 
590 	mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
591 
592 	if (sess != NULL) {
593 		/*
594 		 * new session
595 		 */
596 		mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
597 		PROC_LOCK(p);
598 		p->p_flag &= ~P_CONTROLT;
599 		PROC_UNLOCK(p);
600 		PGRP_LOCK(pgrp);
601 		sess->s_leader = p;
602 		sess->s_sid = p->p_pid;
603 		proc_id_set(PROC_ID_SESSION, p->p_pid);
604 		refcount_init(&sess->s_count, 1);
605 		sess->s_ttyvp = NULL;
606 		sess->s_ttydp = NULL;
607 		sess->s_ttyp = NULL;
608 		bcopy(p->p_session->s_login, sess->s_login,
609 			    sizeof(sess->s_login));
610 		pgrp->pg_session = sess;
611 		KASSERT(p == curproc,
612 		    ("enterpgrp: mksession and p != curproc"));
613 	} else {
614 		pgrp->pg_session = p->p_session;
615 		sess_hold(pgrp->pg_session);
616 		PGRP_LOCK(pgrp);
617 	}
618 	pgrp->pg_id = pgid;
619 	proc_id_set(PROC_ID_GROUP, p->p_pid);
620 	LIST_INIT(&pgrp->pg_members);
621 
622 	/*
623 	 * As we have an exclusive lock of proctree_lock,
624 	 * this should not deadlock.
625 	 */
626 	LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
627 	pgrp->pg_jobc = 0;
628 	SLIST_INIT(&pgrp->pg_sigiolst);
629 	PGRP_UNLOCK(pgrp);
630 
631 	doenterpgrp(p, pgrp);
632 
633 	return (0);
634 }
635 
636 /*
637  * Move p to an existing process group
638  */
639 int
640 enterthispgrp(struct proc *p, struct pgrp *pgrp)
641 {
642 
643 	sx_assert(&proctree_lock, SX_XLOCKED);
644 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
645 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
646 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
647 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
648 	KASSERT(pgrp->pg_session == p->p_session,
649 		("%s: pgrp's session %p, p->p_session %p.\n",
650 		__func__,
651 		pgrp->pg_session,
652 		p->p_session));
653 	KASSERT(pgrp != p->p_pgrp,
654 		("%s: p belongs to pgrp.", __func__));
655 
656 	doenterpgrp(p, pgrp);
657 
658 	return (0);
659 }
660 
661 /*
662  * Move p to a process group
663  */
664 static void
665 doenterpgrp(struct proc *p, struct pgrp *pgrp)
666 {
667 	struct pgrp *savepgrp;
668 
669 	sx_assert(&proctree_lock, SX_XLOCKED);
670 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
671 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
672 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
673 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
674 
675 	savepgrp = p->p_pgrp;
676 
677 	/*
678 	 * Adjust eligibility of affected pgrps to participate in job control.
679 	 * Increment eligibility counts before decrementing, otherwise we
680 	 * could reach 0 spuriously during the first call.
681 	 */
682 	fixjobc(p, pgrp, 1);
683 	fixjobc(p, p->p_pgrp, 0);
684 
685 	PGRP_LOCK(pgrp);
686 	PGRP_LOCK(savepgrp);
687 	PROC_LOCK(p);
688 	LIST_REMOVE(p, p_pglist);
689 	p->p_pgrp = pgrp;
690 	PROC_UNLOCK(p);
691 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
692 	PGRP_UNLOCK(savepgrp);
693 	PGRP_UNLOCK(pgrp);
694 	if (LIST_EMPTY(&savepgrp->pg_members))
695 		pgdelete(savepgrp);
696 }
697 
698 /*
699  * remove process from process group
700  */
701 int
702 leavepgrp(struct proc *p)
703 {
704 	struct pgrp *savepgrp;
705 
706 	sx_assert(&proctree_lock, SX_XLOCKED);
707 	savepgrp = p->p_pgrp;
708 	PGRP_LOCK(savepgrp);
709 	PROC_LOCK(p);
710 	LIST_REMOVE(p, p_pglist);
711 	p->p_pgrp = NULL;
712 	PROC_UNLOCK(p);
713 	PGRP_UNLOCK(savepgrp);
714 	if (LIST_EMPTY(&savepgrp->pg_members))
715 		pgdelete(savepgrp);
716 	return (0);
717 }
718 
719 /*
720  * delete a process group
721  */
722 static void
723 pgdelete(struct pgrp *pgrp)
724 {
725 	struct session *savesess;
726 	struct tty *tp;
727 
728 	sx_assert(&proctree_lock, SX_XLOCKED);
729 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
730 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
731 
732 	/*
733 	 * Reset any sigio structures pointing to us as a result of
734 	 * F_SETOWN with our pgid.
735 	 */
736 	funsetownlst(&pgrp->pg_sigiolst);
737 
738 	PGRP_LOCK(pgrp);
739 	tp = pgrp->pg_session->s_ttyp;
740 	LIST_REMOVE(pgrp, pg_hash);
741 	savesess = pgrp->pg_session;
742 	PGRP_UNLOCK(pgrp);
743 
744 	/* Remove the reference to the pgrp before deallocating it. */
745 	if (tp != NULL) {
746 		tty_lock(tp);
747 		tty_rel_pgrp(tp, pgrp);
748 	}
749 
750 	proc_id_clear(PROC_ID_GROUP, pgrp->pg_id);
751 	mtx_destroy(&pgrp->pg_mtx);
752 	free(pgrp, M_PGRP);
753 	sess_release(savesess);
754 }
755 
756 static void
757 pgadjustjobc(struct pgrp *pgrp, int entering)
758 {
759 
760 	PGRP_LOCK(pgrp);
761 	if (entering)
762 		pgrp->pg_jobc++;
763 	else {
764 		--pgrp->pg_jobc;
765 		if (pgrp->pg_jobc == 0)
766 			orphanpg(pgrp);
767 	}
768 	PGRP_UNLOCK(pgrp);
769 }
770 
771 /*
772  * Adjust pgrp jobc counters when specified process changes process group.
773  * We count the number of processes in each process group that "qualify"
774  * the group for terminal job control (those with a parent in a different
775  * process group of the same session).  If that count reaches zero, the
776  * process group becomes orphaned.  Check both the specified process'
777  * process group and that of its children.
778  * entering == 0 => p is leaving specified group.
779  * entering == 1 => p is entering specified group.
780  */
781 void
782 fixjobc(struct proc *p, struct pgrp *pgrp, int entering)
783 {
784 	struct pgrp *hispgrp;
785 	struct session *mysession;
786 	struct proc *q;
787 
788 	sx_assert(&proctree_lock, SX_LOCKED);
789 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
790 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
791 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
792 
793 	/*
794 	 * Check p's parent to see whether p qualifies its own process
795 	 * group; if so, adjust count for p's process group.
796 	 */
797 	mysession = pgrp->pg_session;
798 	if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
799 	    hispgrp->pg_session == mysession)
800 		pgadjustjobc(pgrp, entering);
801 
802 	/*
803 	 * Check this process' children to see whether they qualify
804 	 * their process groups; if so, adjust counts for children's
805 	 * process groups.
806 	 */
807 	LIST_FOREACH(q, &p->p_children, p_sibling) {
808 		hispgrp = q->p_pgrp;
809 		if (hispgrp == pgrp ||
810 		    hispgrp->pg_session != mysession)
811 			continue;
812 		if (q->p_state == PRS_ZOMBIE)
813 			continue;
814 		pgadjustjobc(hispgrp, entering);
815 	}
816 }
817 
818 void
819 killjobc(void)
820 {
821 	struct session *sp;
822 	struct tty *tp;
823 	struct proc *p;
824 	struct vnode *ttyvp;
825 
826 	p = curproc;
827 	MPASS(p->p_flag & P_WEXIT);
828 	/*
829 	 * Do a quick check to see if there is anything to do with the
830 	 * proctree_lock held. pgrp and LIST_EMPTY checks are for fixjobc().
831 	 */
832 	PROC_LOCK(p);
833 	if (!SESS_LEADER(p) &&
834 	    (p->p_pgrp == p->p_pptr->p_pgrp) &&
835 	    LIST_EMPTY(&p->p_children)) {
836 		PROC_UNLOCK(p);
837 		return;
838 	}
839 	PROC_UNLOCK(p);
840 
841 	sx_xlock(&proctree_lock);
842 	if (SESS_LEADER(p)) {
843 		sp = p->p_session;
844 
845 		/*
846 		 * s_ttyp is not zero'd; we use this to indicate that
847 		 * the session once had a controlling terminal. (for
848 		 * logging and informational purposes)
849 		 */
850 		SESS_LOCK(sp);
851 		ttyvp = sp->s_ttyvp;
852 		tp = sp->s_ttyp;
853 		sp->s_ttyvp = NULL;
854 		sp->s_ttydp = NULL;
855 		sp->s_leader = NULL;
856 		SESS_UNLOCK(sp);
857 
858 		/*
859 		 * Signal foreground pgrp and revoke access to
860 		 * controlling terminal if it has not been revoked
861 		 * already.
862 		 *
863 		 * Because the TTY may have been revoked in the mean
864 		 * time and could already have a new session associated
865 		 * with it, make sure we don't send a SIGHUP to a
866 		 * foreground process group that does not belong to this
867 		 * session.
868 		 */
869 
870 		if (tp != NULL) {
871 			tty_lock(tp);
872 			if (tp->t_session == sp)
873 				tty_signal_pgrp(tp, SIGHUP);
874 			tty_unlock(tp);
875 		}
876 
877 		if (ttyvp != NULL) {
878 			sx_xunlock(&proctree_lock);
879 			if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) {
880 				VOP_REVOKE(ttyvp, REVOKEALL);
881 				VOP_UNLOCK(ttyvp, 0);
882 			}
883 			vrele(ttyvp);
884 			sx_xlock(&proctree_lock);
885 		}
886 	}
887 	fixjobc(p, p->p_pgrp, 0);
888 	sx_xunlock(&proctree_lock);
889 }
890 
891 /*
892  * A process group has become orphaned;
893  * if there are any stopped processes in the group,
894  * hang-up all process in that group.
895  */
896 static void
897 orphanpg(struct pgrp *pg)
898 {
899 	struct proc *p;
900 
901 	PGRP_LOCK_ASSERT(pg, MA_OWNED);
902 
903 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
904 		PROC_LOCK(p);
905 		if (P_SHOULDSTOP(p) == P_STOPPED_SIG) {
906 			PROC_UNLOCK(p);
907 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
908 				PROC_LOCK(p);
909 				kern_psignal(p, SIGHUP);
910 				kern_psignal(p, SIGCONT);
911 				PROC_UNLOCK(p);
912 			}
913 			return;
914 		}
915 		PROC_UNLOCK(p);
916 	}
917 }
918 
919 void
920 sess_hold(struct session *s)
921 {
922 
923 	refcount_acquire(&s->s_count);
924 }
925 
926 void
927 sess_release(struct session *s)
928 {
929 
930 	if (refcount_release(&s->s_count)) {
931 		if (s->s_ttyp != NULL) {
932 			tty_lock(s->s_ttyp);
933 			tty_rel_sess(s->s_ttyp, s);
934 		}
935 		proc_id_clear(PROC_ID_SESSION, s->s_sid);
936 		mtx_destroy(&s->s_mtx);
937 		free(s, M_SESSION);
938 	}
939 }
940 
941 #ifdef DDB
942 
943 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
944 {
945 	struct pgrp *pgrp;
946 	struct proc *p;
947 	int i;
948 
949 	for (i = 0; i <= pgrphash; i++) {
950 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
951 			printf("\tindx %d\n", i);
952 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
953 				printf(
954 			"\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
955 				    (void *)pgrp, (long)pgrp->pg_id,
956 				    (void *)pgrp->pg_session,
957 				    pgrp->pg_session->s_count,
958 				    (void *)LIST_FIRST(&pgrp->pg_members));
959 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
960 					printf("\t\tpid %ld addr %p pgrp %p\n",
961 					    (long)p->p_pid, (void *)p,
962 					    (void *)p->p_pgrp);
963 				}
964 			}
965 		}
966 	}
967 }
968 #endif /* DDB */
969 
970 /*
971  * Calculate the kinfo_proc members which contain process-wide
972  * informations.
973  * Must be called with the target process locked.
974  */
975 static void
976 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
977 {
978 	struct thread *td;
979 
980 	PROC_LOCK_ASSERT(p, MA_OWNED);
981 
982 	kp->ki_estcpu = 0;
983 	kp->ki_pctcpu = 0;
984 	FOREACH_THREAD_IN_PROC(p, td) {
985 		thread_lock(td);
986 		kp->ki_pctcpu += sched_pctcpu(td);
987 		kp->ki_estcpu += sched_estcpu(td);
988 		thread_unlock(td);
989 	}
990 }
991 
992 /*
993  * Clear kinfo_proc and fill in any information that is common
994  * to all threads in the process.
995  * Must be called with the target process locked.
996  */
997 static void
998 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
999 {
1000 	struct thread *td0;
1001 	struct tty *tp;
1002 	struct session *sp;
1003 	struct ucred *cred;
1004 	struct sigacts *ps;
1005 	struct timeval boottime;
1006 
1007 	PROC_LOCK_ASSERT(p, MA_OWNED);
1008 	bzero(kp, sizeof(*kp));
1009 
1010 	kp->ki_structsize = sizeof(*kp);
1011 	kp->ki_paddr = p;
1012 	kp->ki_addr =/* p->p_addr; */0; /* XXX */
1013 	kp->ki_args = p->p_args;
1014 	kp->ki_textvp = p->p_textvp;
1015 #ifdef KTRACE
1016 	kp->ki_tracep = p->p_tracevp;
1017 	kp->ki_traceflag = p->p_traceflag;
1018 #endif
1019 	kp->ki_fd = p->p_fd;
1020 	kp->ki_vmspace = p->p_vmspace;
1021 	kp->ki_flag = p->p_flag;
1022 	kp->ki_flag2 = p->p_flag2;
1023 	cred = p->p_ucred;
1024 	if (cred) {
1025 		kp->ki_uid = cred->cr_uid;
1026 		kp->ki_ruid = cred->cr_ruid;
1027 		kp->ki_svuid = cred->cr_svuid;
1028 		kp->ki_cr_flags = 0;
1029 		if (cred->cr_flags & CRED_FLAG_CAPMODE)
1030 			kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
1031 		/* XXX bde doesn't like KI_NGROUPS */
1032 		if (cred->cr_ngroups > KI_NGROUPS) {
1033 			kp->ki_ngroups = KI_NGROUPS;
1034 			kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
1035 		} else
1036 			kp->ki_ngroups = cred->cr_ngroups;
1037 		bcopy(cred->cr_groups, kp->ki_groups,
1038 		    kp->ki_ngroups * sizeof(gid_t));
1039 		kp->ki_rgid = cred->cr_rgid;
1040 		kp->ki_svgid = cred->cr_svgid;
1041 		/* If jailed(cred), emulate the old P_JAILED flag. */
1042 		if (jailed(cred)) {
1043 			kp->ki_flag |= P_JAILED;
1044 			/* If inside the jail, use 0 as a jail ID. */
1045 			if (cred->cr_prison != curthread->td_ucred->cr_prison)
1046 				kp->ki_jid = cred->cr_prison->pr_id;
1047 		}
1048 		strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
1049 		    sizeof(kp->ki_loginclass));
1050 	}
1051 	ps = p->p_sigacts;
1052 	if (ps) {
1053 		mtx_lock(&ps->ps_mtx);
1054 		kp->ki_sigignore = ps->ps_sigignore;
1055 		kp->ki_sigcatch = ps->ps_sigcatch;
1056 		mtx_unlock(&ps->ps_mtx);
1057 	}
1058 	if (p->p_state != PRS_NEW &&
1059 	    p->p_state != PRS_ZOMBIE &&
1060 	    p->p_vmspace != NULL) {
1061 		struct vmspace *vm = p->p_vmspace;
1062 
1063 		kp->ki_size = vm->vm_map.size;
1064 		kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
1065 		FOREACH_THREAD_IN_PROC(p, td0) {
1066 			if (!TD_IS_SWAPPED(td0))
1067 				kp->ki_rssize += td0->td_kstack_pages;
1068 		}
1069 		kp->ki_swrss = vm->vm_swrss;
1070 		kp->ki_tsize = vm->vm_tsize;
1071 		kp->ki_dsize = vm->vm_dsize;
1072 		kp->ki_ssize = vm->vm_ssize;
1073 	} else if (p->p_state == PRS_ZOMBIE)
1074 		kp->ki_stat = SZOMB;
1075 	if (kp->ki_flag & P_INMEM)
1076 		kp->ki_sflag = PS_INMEM;
1077 	else
1078 		kp->ki_sflag = 0;
1079 	/* Calculate legacy swtime as seconds since 'swtick'. */
1080 	kp->ki_swtime = (ticks - p->p_swtick) / hz;
1081 	kp->ki_pid = p->p_pid;
1082 	kp->ki_nice = p->p_nice;
1083 	kp->ki_fibnum = p->p_fibnum;
1084 	kp->ki_start = p->p_stats->p_start;
1085 	getboottime(&boottime);
1086 	timevaladd(&kp->ki_start, &boottime);
1087 	PROC_STATLOCK(p);
1088 	rufetch(p, &kp->ki_rusage);
1089 	kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
1090 	calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
1091 	PROC_STATUNLOCK(p);
1092 	calccru(p, &kp->ki_childutime, &kp->ki_childstime);
1093 	/* Some callers want child times in a single value. */
1094 	kp->ki_childtime = kp->ki_childstime;
1095 	timevaladd(&kp->ki_childtime, &kp->ki_childutime);
1096 
1097 	FOREACH_THREAD_IN_PROC(p, td0)
1098 		kp->ki_cow += td0->td_cow;
1099 
1100 	tp = NULL;
1101 	if (p->p_pgrp) {
1102 		kp->ki_pgid = p->p_pgrp->pg_id;
1103 		kp->ki_jobc = p->p_pgrp->pg_jobc;
1104 		sp = p->p_pgrp->pg_session;
1105 
1106 		if (sp != NULL) {
1107 			kp->ki_sid = sp->s_sid;
1108 			SESS_LOCK(sp);
1109 			strlcpy(kp->ki_login, sp->s_login,
1110 			    sizeof(kp->ki_login));
1111 			if (sp->s_ttyvp)
1112 				kp->ki_kiflag |= KI_CTTY;
1113 			if (SESS_LEADER(p))
1114 				kp->ki_kiflag |= KI_SLEADER;
1115 			/* XXX proctree_lock */
1116 			tp = sp->s_ttyp;
1117 			SESS_UNLOCK(sp);
1118 		}
1119 	}
1120 	if ((p->p_flag & P_CONTROLT) && tp != NULL) {
1121 		kp->ki_tdev = tty_udev(tp);
1122 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1123 		kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
1124 		if (tp->t_session)
1125 			kp->ki_tsid = tp->t_session->s_sid;
1126 	} else {
1127 		kp->ki_tdev = NODEV;
1128 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1129 	}
1130 	if (p->p_comm[0] != '\0')
1131 		strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
1132 	if (p->p_sysent && p->p_sysent->sv_name != NULL &&
1133 	    p->p_sysent->sv_name[0] != '\0')
1134 		strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
1135 	kp->ki_siglist = p->p_siglist;
1136 	kp->ki_xstat = KW_EXITCODE(p->p_xexit, p->p_xsig);
1137 	kp->ki_acflag = p->p_acflag;
1138 	kp->ki_lock = p->p_lock;
1139 	if (p->p_pptr) {
1140 		kp->ki_ppid = p->p_oppid;
1141 		if (p->p_flag & P_TRACED)
1142 			kp->ki_tracer = p->p_pptr->p_pid;
1143 	}
1144 }
1145 
1146 /*
1147  * Fill in information that is thread specific.  Must be called with
1148  * target process locked.  If 'preferthread' is set, overwrite certain
1149  * process-related fields that are maintained for both threads and
1150  * processes.
1151  */
1152 static void
1153 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
1154 {
1155 	struct proc *p;
1156 
1157 	p = td->td_proc;
1158 	kp->ki_tdaddr = td;
1159 	PROC_LOCK_ASSERT(p, MA_OWNED);
1160 
1161 	if (preferthread)
1162 		PROC_STATLOCK(p);
1163 	thread_lock(td);
1164 	if (td->td_wmesg != NULL)
1165 		strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
1166 	else
1167 		bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
1168 	if (strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname)) >=
1169 	    sizeof(kp->ki_tdname)) {
1170 		strlcpy(kp->ki_moretdname,
1171 		    td->td_name + sizeof(kp->ki_tdname) - 1,
1172 		    sizeof(kp->ki_moretdname));
1173 	} else {
1174 		bzero(kp->ki_moretdname, sizeof(kp->ki_moretdname));
1175 	}
1176 	if (TD_ON_LOCK(td)) {
1177 		kp->ki_kiflag |= KI_LOCKBLOCK;
1178 		strlcpy(kp->ki_lockname, td->td_lockname,
1179 		    sizeof(kp->ki_lockname));
1180 	} else {
1181 		kp->ki_kiflag &= ~KI_LOCKBLOCK;
1182 		bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
1183 	}
1184 
1185 	if (p->p_state == PRS_NORMAL) { /* approximate. */
1186 		if (TD_ON_RUNQ(td) ||
1187 		    TD_CAN_RUN(td) ||
1188 		    TD_IS_RUNNING(td)) {
1189 			kp->ki_stat = SRUN;
1190 		} else if (P_SHOULDSTOP(p)) {
1191 			kp->ki_stat = SSTOP;
1192 		} else if (TD_IS_SLEEPING(td)) {
1193 			kp->ki_stat = SSLEEP;
1194 		} else if (TD_ON_LOCK(td)) {
1195 			kp->ki_stat = SLOCK;
1196 		} else {
1197 			kp->ki_stat = SWAIT;
1198 		}
1199 	} else if (p->p_state == PRS_ZOMBIE) {
1200 		kp->ki_stat = SZOMB;
1201 	} else {
1202 		kp->ki_stat = SIDL;
1203 	}
1204 
1205 	/* Things in the thread */
1206 	kp->ki_wchan = td->td_wchan;
1207 	kp->ki_pri.pri_level = td->td_priority;
1208 	kp->ki_pri.pri_native = td->td_base_pri;
1209 
1210 	/*
1211 	 * Note: legacy fields; clamp at the old NOCPU value and/or
1212 	 * the maximum u_char CPU value.
1213 	 */
1214 	if (td->td_lastcpu == NOCPU)
1215 		kp->ki_lastcpu_old = NOCPU_OLD;
1216 	else if (td->td_lastcpu > MAXCPU_OLD)
1217 		kp->ki_lastcpu_old = MAXCPU_OLD;
1218 	else
1219 		kp->ki_lastcpu_old = td->td_lastcpu;
1220 
1221 	if (td->td_oncpu == NOCPU)
1222 		kp->ki_oncpu_old = NOCPU_OLD;
1223 	else if (td->td_oncpu > MAXCPU_OLD)
1224 		kp->ki_oncpu_old = MAXCPU_OLD;
1225 	else
1226 		kp->ki_oncpu_old = td->td_oncpu;
1227 
1228 	kp->ki_lastcpu = td->td_lastcpu;
1229 	kp->ki_oncpu = td->td_oncpu;
1230 	kp->ki_tdflags = td->td_flags;
1231 	kp->ki_tid = td->td_tid;
1232 	kp->ki_numthreads = p->p_numthreads;
1233 	kp->ki_pcb = td->td_pcb;
1234 	kp->ki_kstack = (void *)td->td_kstack;
1235 	kp->ki_slptime = (ticks - td->td_slptick) / hz;
1236 	kp->ki_pri.pri_class = td->td_pri_class;
1237 	kp->ki_pri.pri_user = td->td_user_pri;
1238 
1239 	if (preferthread) {
1240 		rufetchtd(td, &kp->ki_rusage);
1241 		kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
1242 		kp->ki_pctcpu = sched_pctcpu(td);
1243 		kp->ki_estcpu = sched_estcpu(td);
1244 		kp->ki_cow = td->td_cow;
1245 	}
1246 
1247 	/* We can't get this anymore but ps etc never used it anyway. */
1248 	kp->ki_rqindex = 0;
1249 
1250 	if (preferthread)
1251 		kp->ki_siglist = td->td_siglist;
1252 	kp->ki_sigmask = td->td_sigmask;
1253 	thread_unlock(td);
1254 	if (preferthread)
1255 		PROC_STATUNLOCK(p);
1256 }
1257 
1258 /*
1259  * Fill in a kinfo_proc structure for the specified process.
1260  * Must be called with the target process locked.
1261  */
1262 void
1263 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1264 {
1265 
1266 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1267 
1268 	fill_kinfo_proc_only(p, kp);
1269 	fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1270 	fill_kinfo_aggregate(p, kp);
1271 }
1272 
1273 struct pstats *
1274 pstats_alloc(void)
1275 {
1276 
1277 	return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1278 }
1279 
1280 /*
1281  * Copy parts of p_stats; zero the rest of p_stats (statistics).
1282  */
1283 void
1284 pstats_fork(struct pstats *src, struct pstats *dst)
1285 {
1286 
1287 	bzero(&dst->pstat_startzero,
1288 	    __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1289 	bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1290 	    __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1291 }
1292 
1293 void
1294 pstats_free(struct pstats *ps)
1295 {
1296 
1297 	free(ps, M_SUBPROC);
1298 }
1299 
1300 /*
1301  * Locate a zombie process by number
1302  */
1303 struct proc *
1304 zpfind(pid_t pid)
1305 {
1306 	struct proc *p;
1307 
1308 	sx_slock(&zombproc_lock);
1309 	LIST_FOREACH(p, &zombproc, p_list) {
1310 		if (p->p_pid == pid) {
1311 			PROC_LOCK(p);
1312 			break;
1313 		}
1314 	}
1315 	sx_sunlock(&zombproc_lock);
1316 	return (p);
1317 }
1318 
1319 #ifdef COMPAT_FREEBSD32
1320 
1321 /*
1322  * This function is typically used to copy out the kernel address, so
1323  * it can be replaced by assignment of zero.
1324  */
1325 static inline uint32_t
1326 ptr32_trim(void *ptr)
1327 {
1328 	uintptr_t uptr;
1329 
1330 	uptr = (uintptr_t)ptr;
1331 	return ((uptr > UINT_MAX) ? 0 : uptr);
1332 }
1333 
1334 #define PTRTRIM_CP(src,dst,fld) \
1335 	do { (dst).fld = ptr32_trim((src).fld); } while (0)
1336 
1337 static void
1338 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1339 {
1340 	int i;
1341 
1342 	bzero(ki32, sizeof(struct kinfo_proc32));
1343 	ki32->ki_structsize = sizeof(struct kinfo_proc32);
1344 	CP(*ki, *ki32, ki_layout);
1345 	PTRTRIM_CP(*ki, *ki32, ki_args);
1346 	PTRTRIM_CP(*ki, *ki32, ki_paddr);
1347 	PTRTRIM_CP(*ki, *ki32, ki_addr);
1348 	PTRTRIM_CP(*ki, *ki32, ki_tracep);
1349 	PTRTRIM_CP(*ki, *ki32, ki_textvp);
1350 	PTRTRIM_CP(*ki, *ki32, ki_fd);
1351 	PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1352 	PTRTRIM_CP(*ki, *ki32, ki_wchan);
1353 	CP(*ki, *ki32, ki_pid);
1354 	CP(*ki, *ki32, ki_ppid);
1355 	CP(*ki, *ki32, ki_pgid);
1356 	CP(*ki, *ki32, ki_tpgid);
1357 	CP(*ki, *ki32, ki_sid);
1358 	CP(*ki, *ki32, ki_tsid);
1359 	CP(*ki, *ki32, ki_jobc);
1360 	CP(*ki, *ki32, ki_tdev);
1361 	CP(*ki, *ki32, ki_tdev_freebsd11);
1362 	CP(*ki, *ki32, ki_siglist);
1363 	CP(*ki, *ki32, ki_sigmask);
1364 	CP(*ki, *ki32, ki_sigignore);
1365 	CP(*ki, *ki32, ki_sigcatch);
1366 	CP(*ki, *ki32, ki_uid);
1367 	CP(*ki, *ki32, ki_ruid);
1368 	CP(*ki, *ki32, ki_svuid);
1369 	CP(*ki, *ki32, ki_rgid);
1370 	CP(*ki, *ki32, ki_svgid);
1371 	CP(*ki, *ki32, ki_ngroups);
1372 	for (i = 0; i < KI_NGROUPS; i++)
1373 		CP(*ki, *ki32, ki_groups[i]);
1374 	CP(*ki, *ki32, ki_size);
1375 	CP(*ki, *ki32, ki_rssize);
1376 	CP(*ki, *ki32, ki_swrss);
1377 	CP(*ki, *ki32, ki_tsize);
1378 	CP(*ki, *ki32, ki_dsize);
1379 	CP(*ki, *ki32, ki_ssize);
1380 	CP(*ki, *ki32, ki_xstat);
1381 	CP(*ki, *ki32, ki_acflag);
1382 	CP(*ki, *ki32, ki_pctcpu);
1383 	CP(*ki, *ki32, ki_estcpu);
1384 	CP(*ki, *ki32, ki_slptime);
1385 	CP(*ki, *ki32, ki_swtime);
1386 	CP(*ki, *ki32, ki_cow);
1387 	CP(*ki, *ki32, ki_runtime);
1388 	TV_CP(*ki, *ki32, ki_start);
1389 	TV_CP(*ki, *ki32, ki_childtime);
1390 	CP(*ki, *ki32, ki_flag);
1391 	CP(*ki, *ki32, ki_kiflag);
1392 	CP(*ki, *ki32, ki_traceflag);
1393 	CP(*ki, *ki32, ki_stat);
1394 	CP(*ki, *ki32, ki_nice);
1395 	CP(*ki, *ki32, ki_lock);
1396 	CP(*ki, *ki32, ki_rqindex);
1397 	CP(*ki, *ki32, ki_oncpu);
1398 	CP(*ki, *ki32, ki_lastcpu);
1399 
1400 	/* XXX TODO: wrap cpu value as appropriate */
1401 	CP(*ki, *ki32, ki_oncpu_old);
1402 	CP(*ki, *ki32, ki_lastcpu_old);
1403 
1404 	bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1405 	bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1406 	bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1407 	bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1408 	bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1409 	bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1410 	bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1411 	bcopy(ki->ki_moretdname, ki32->ki_moretdname, MAXCOMLEN - TDNAMLEN + 1);
1412 	CP(*ki, *ki32, ki_tracer);
1413 	CP(*ki, *ki32, ki_flag2);
1414 	CP(*ki, *ki32, ki_fibnum);
1415 	CP(*ki, *ki32, ki_cr_flags);
1416 	CP(*ki, *ki32, ki_jid);
1417 	CP(*ki, *ki32, ki_numthreads);
1418 	CP(*ki, *ki32, ki_tid);
1419 	CP(*ki, *ki32, ki_pri);
1420 	freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1421 	freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1422 	PTRTRIM_CP(*ki, *ki32, ki_pcb);
1423 	PTRTRIM_CP(*ki, *ki32, ki_kstack);
1424 	PTRTRIM_CP(*ki, *ki32, ki_udata);
1425 	PTRTRIM_CP(*ki, *ki32, ki_tdaddr);
1426 	CP(*ki, *ki32, ki_sflag);
1427 	CP(*ki, *ki32, ki_tdflags);
1428 }
1429 #endif
1430 
1431 static ssize_t
1432 kern_proc_out_size(struct proc *p, int flags)
1433 {
1434 	ssize_t size = 0;
1435 
1436 	PROC_LOCK_ASSERT(p, MA_OWNED);
1437 
1438 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1439 #ifdef COMPAT_FREEBSD32
1440 		if ((flags & KERN_PROC_MASK32) != 0) {
1441 			size += sizeof(struct kinfo_proc32);
1442 		} else
1443 #endif
1444 			size += sizeof(struct kinfo_proc);
1445 	} else {
1446 #ifdef COMPAT_FREEBSD32
1447 		if ((flags & KERN_PROC_MASK32) != 0)
1448 			size += sizeof(struct kinfo_proc32) * p->p_numthreads;
1449 		else
1450 #endif
1451 			size += sizeof(struct kinfo_proc) * p->p_numthreads;
1452 	}
1453 	PROC_UNLOCK(p);
1454 	return (size);
1455 }
1456 
1457 int
1458 kern_proc_out(struct proc *p, struct sbuf *sb, int flags)
1459 {
1460 	struct thread *td;
1461 	struct kinfo_proc ki;
1462 #ifdef COMPAT_FREEBSD32
1463 	struct kinfo_proc32 ki32;
1464 #endif
1465 	int error;
1466 
1467 	PROC_LOCK_ASSERT(p, MA_OWNED);
1468 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1469 
1470 	error = 0;
1471 	fill_kinfo_proc(p, &ki);
1472 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1473 #ifdef COMPAT_FREEBSD32
1474 		if ((flags & KERN_PROC_MASK32) != 0) {
1475 			freebsd32_kinfo_proc_out(&ki, &ki32);
1476 			if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1477 				error = ENOMEM;
1478 		} else
1479 #endif
1480 			if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1481 				error = ENOMEM;
1482 	} else {
1483 		FOREACH_THREAD_IN_PROC(p, td) {
1484 			fill_kinfo_thread(td, &ki, 1);
1485 #ifdef COMPAT_FREEBSD32
1486 			if ((flags & KERN_PROC_MASK32) != 0) {
1487 				freebsd32_kinfo_proc_out(&ki, &ki32);
1488 				if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1489 					error = ENOMEM;
1490 			} else
1491 #endif
1492 				if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1493 					error = ENOMEM;
1494 			if (error != 0)
1495 				break;
1496 		}
1497 	}
1498 	PROC_UNLOCK(p);
1499 	return (error);
1500 }
1501 
1502 static int
1503 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
1504 {
1505 	struct sbuf sb;
1506 	struct kinfo_proc ki;
1507 	int error, error2;
1508 
1509 	if (req->oldptr == NULL)
1510 		return (SYSCTL_OUT(req, 0, kern_proc_out_size(p, flags)));
1511 
1512 	sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req);
1513 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1514 	error = kern_proc_out(p, &sb, flags);
1515 	error2 = sbuf_finish(&sb);
1516 	sbuf_delete(&sb);
1517 	if (error != 0)
1518 		return (error);
1519 	else if (error2 != 0)
1520 		return (error2);
1521 	return (0);
1522 }
1523 
1524 int
1525 proc_iterate(int (*cb)(struct proc *, void *), void *cbarg)
1526 {
1527 	struct proc *p;
1528 	int error, i, j;
1529 
1530 	for (i = 0; i < pidhashlock + 1; i++) {
1531 		sx_slock(&pidhashtbl_lock[i]);
1532 		for (j = i; j <= pidhash; j += pidhashlock + 1) {
1533 			LIST_FOREACH(p, &pidhashtbl[j], p_hash) {
1534 				if (p->p_state == PRS_NEW)
1535 					continue;
1536 				error = cb(p, cbarg);
1537 				PROC_LOCK_ASSERT(p, MA_NOTOWNED);
1538 				if (error != 0) {
1539 					sx_sunlock(&pidhashtbl_lock[i]);
1540 					return (error);
1541 				}
1542 			}
1543 		}
1544 		sx_sunlock(&pidhashtbl_lock[i]);
1545 	}
1546 	return (0);
1547 }
1548 
1549 struct kern_proc_out_args {
1550 	struct sysctl_req *req;
1551 	int flags;
1552 	int oid_number;
1553 	int *name;
1554 };
1555 
1556 static int
1557 sysctl_kern_proc_iterate(struct proc *p, void *origarg)
1558 {
1559 	struct kern_proc_out_args *arg = origarg;
1560 	int *name = arg->name;
1561 	int oid_number = arg->oid_number;
1562 	int flags = arg->flags;
1563 	struct sysctl_req *req = arg->req;
1564 	int error = 0;
1565 
1566 	PROC_LOCK(p);
1567 
1568 	KASSERT(p->p_ucred != NULL,
1569 	    ("process credential is NULL for non-NEW proc"));
1570 	/*
1571 	 * Show a user only appropriate processes.
1572 	 */
1573 	if (p_cansee(curthread, p))
1574 		goto skip;
1575 	/*
1576 	 * TODO - make more efficient (see notes below).
1577 	 * do by session.
1578 	 */
1579 	switch (oid_number) {
1580 
1581 	case KERN_PROC_GID:
1582 		if (p->p_ucred->cr_gid != (gid_t)name[0])
1583 			goto skip;
1584 		break;
1585 
1586 	case KERN_PROC_PGRP:
1587 		/* could do this by traversing pgrp */
1588 		if (p->p_pgrp == NULL ||
1589 		    p->p_pgrp->pg_id != (pid_t)name[0])
1590 			goto skip;
1591 		break;
1592 
1593 	case KERN_PROC_RGID:
1594 		if (p->p_ucred->cr_rgid != (gid_t)name[0])
1595 			goto skip;
1596 		break;
1597 
1598 	case KERN_PROC_SESSION:
1599 		if (p->p_session == NULL ||
1600 		    p->p_session->s_sid != (pid_t)name[0])
1601 			goto skip;
1602 		break;
1603 
1604 	case KERN_PROC_TTY:
1605 		if ((p->p_flag & P_CONTROLT) == 0 ||
1606 		    p->p_session == NULL)
1607 			goto skip;
1608 		/* XXX proctree_lock */
1609 		SESS_LOCK(p->p_session);
1610 		if (p->p_session->s_ttyp == NULL ||
1611 		    tty_udev(p->p_session->s_ttyp) !=
1612 		    (dev_t)name[0]) {
1613 			SESS_UNLOCK(p->p_session);
1614 			goto skip;
1615 		}
1616 		SESS_UNLOCK(p->p_session);
1617 		break;
1618 
1619 	case KERN_PROC_UID:
1620 		if (p->p_ucred->cr_uid != (uid_t)name[0])
1621 			goto skip;
1622 		break;
1623 
1624 	case KERN_PROC_RUID:
1625 		if (p->p_ucred->cr_ruid != (uid_t)name[0])
1626 			goto skip;
1627 		break;
1628 
1629 	case KERN_PROC_PROC:
1630 		break;
1631 
1632 	default:
1633 		break;
1634 
1635 	}
1636 	error = sysctl_out_proc(p, req, flags);
1637 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
1638 	return (error);
1639 skip:
1640 	PROC_UNLOCK(p);
1641 	return (0);
1642 }
1643 
1644 static int
1645 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1646 {
1647 	struct kern_proc_out_args iterarg;
1648 	int *name = (int *)arg1;
1649 	u_int namelen = arg2;
1650 	struct proc *p;
1651 	int flags, oid_number;
1652 	int error = 0;
1653 
1654 	oid_number = oidp->oid_number;
1655 	if (oid_number != KERN_PROC_ALL &&
1656 	    (oid_number & KERN_PROC_INC_THREAD) == 0)
1657 		flags = KERN_PROC_NOTHREADS;
1658 	else {
1659 		flags = 0;
1660 		oid_number &= ~KERN_PROC_INC_THREAD;
1661 	}
1662 #ifdef COMPAT_FREEBSD32
1663 	if (req->flags & SCTL_MASK32)
1664 		flags |= KERN_PROC_MASK32;
1665 #endif
1666 	if (oid_number == KERN_PROC_PID) {
1667 		if (namelen != 1)
1668 			return (EINVAL);
1669 		error = sysctl_wire_old_buffer(req, 0);
1670 		if (error)
1671 			return (error);
1672 		error = pget((pid_t)name[0], PGET_CANSEE, &p);
1673 		if (error == 0)
1674 			error = sysctl_out_proc(p, req, flags);
1675 		return (error);
1676 	}
1677 
1678 	switch (oid_number) {
1679 	case KERN_PROC_ALL:
1680 		if (namelen != 0)
1681 			return (EINVAL);
1682 		break;
1683 	case KERN_PROC_PROC:
1684 		if (namelen != 0 && namelen != 1)
1685 			return (EINVAL);
1686 		break;
1687 	default:
1688 		if (namelen != 1)
1689 			return (EINVAL);
1690 		break;
1691 	}
1692 
1693 	if (req->oldptr == NULL) {
1694 		/* overestimate by 5 procs */
1695 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1696 		if (error)
1697 			return (error);
1698 	} else {
1699 		error = sysctl_wire_old_buffer(req, 0);
1700 		if (error != 0)
1701 			return (error);
1702 	}
1703 	iterarg.flags = flags;
1704 	iterarg.oid_number = oid_number;
1705 	iterarg.req = req;
1706 	iterarg.name = name;
1707 	error = proc_iterate(sysctl_kern_proc_iterate, &iterarg);
1708 	return (error);
1709 }
1710 
1711 struct pargs *
1712 pargs_alloc(int len)
1713 {
1714 	struct pargs *pa;
1715 
1716 	pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1717 		M_WAITOK);
1718 	refcount_init(&pa->ar_ref, 1);
1719 	pa->ar_length = len;
1720 	return (pa);
1721 }
1722 
1723 static void
1724 pargs_free(struct pargs *pa)
1725 {
1726 
1727 	free(pa, M_PARGS);
1728 }
1729 
1730 void
1731 pargs_hold(struct pargs *pa)
1732 {
1733 
1734 	if (pa == NULL)
1735 		return;
1736 	refcount_acquire(&pa->ar_ref);
1737 }
1738 
1739 void
1740 pargs_drop(struct pargs *pa)
1741 {
1742 
1743 	if (pa == NULL)
1744 		return;
1745 	if (refcount_release(&pa->ar_ref))
1746 		pargs_free(pa);
1747 }
1748 
1749 static int
1750 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1751     size_t len)
1752 {
1753 	ssize_t n;
1754 
1755 	/*
1756 	 * This may return a short read if the string is shorter than the chunk
1757 	 * and is aligned at the end of the page, and the following page is not
1758 	 * mapped.
1759 	 */
1760 	n = proc_readmem(td, p, (vm_offset_t)sptr, buf, len);
1761 	if (n <= 0)
1762 		return (ENOMEM);
1763 	return (0);
1764 }
1765 
1766 #define PROC_AUXV_MAX	256	/* Safety limit on auxv size. */
1767 
1768 enum proc_vector_type {
1769 	PROC_ARG,
1770 	PROC_ENV,
1771 	PROC_AUX,
1772 };
1773 
1774 #ifdef COMPAT_FREEBSD32
1775 static int
1776 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1777     size_t *vsizep, enum proc_vector_type type)
1778 {
1779 	struct freebsd32_ps_strings pss;
1780 	Elf32_Auxinfo aux;
1781 	vm_offset_t vptr, ptr;
1782 	uint32_t *proc_vector32;
1783 	char **proc_vector;
1784 	size_t vsize, size;
1785 	int i, error;
1786 
1787 	error = 0;
1788 	if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss,
1789 	    sizeof(pss)) != sizeof(pss))
1790 		return (ENOMEM);
1791 	switch (type) {
1792 	case PROC_ARG:
1793 		vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1794 		vsize = pss.ps_nargvstr;
1795 		if (vsize > ARG_MAX)
1796 			return (ENOEXEC);
1797 		size = vsize * sizeof(int32_t);
1798 		break;
1799 	case PROC_ENV:
1800 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1801 		vsize = pss.ps_nenvstr;
1802 		if (vsize > ARG_MAX)
1803 			return (ENOEXEC);
1804 		size = vsize * sizeof(int32_t);
1805 		break;
1806 	case PROC_AUX:
1807 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1808 		    (pss.ps_nenvstr + 1) * sizeof(int32_t);
1809 		if (vptr % 4 != 0)
1810 			return (ENOEXEC);
1811 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1812 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1813 			    sizeof(aux))
1814 				return (ENOMEM);
1815 			if (aux.a_type == AT_NULL)
1816 				break;
1817 			ptr += sizeof(aux);
1818 		}
1819 		if (aux.a_type != AT_NULL)
1820 			return (ENOEXEC);
1821 		vsize = i + 1;
1822 		size = vsize * sizeof(aux);
1823 		break;
1824 	default:
1825 		KASSERT(0, ("Wrong proc vector type: %d", type));
1826 		return (EINVAL);
1827 	}
1828 	proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1829 	if (proc_readmem(td, p, vptr, proc_vector32, size) != size) {
1830 		error = ENOMEM;
1831 		goto done;
1832 	}
1833 	if (type == PROC_AUX) {
1834 		*proc_vectorp = (char **)proc_vector32;
1835 		*vsizep = vsize;
1836 		return (0);
1837 	}
1838 	proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1839 	for (i = 0; i < (int)vsize; i++)
1840 		proc_vector[i] = PTRIN(proc_vector32[i]);
1841 	*proc_vectorp = proc_vector;
1842 	*vsizep = vsize;
1843 done:
1844 	free(proc_vector32, M_TEMP);
1845 	return (error);
1846 }
1847 #endif
1848 
1849 static int
1850 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1851     size_t *vsizep, enum proc_vector_type type)
1852 {
1853 	struct ps_strings pss;
1854 	Elf_Auxinfo aux;
1855 	vm_offset_t vptr, ptr;
1856 	char **proc_vector;
1857 	size_t vsize, size;
1858 	int i;
1859 
1860 #ifdef COMPAT_FREEBSD32
1861 	if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1862 		return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1863 #endif
1864 	if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss,
1865 	    sizeof(pss)) != sizeof(pss))
1866 		return (ENOMEM);
1867 	switch (type) {
1868 	case PROC_ARG:
1869 		vptr = (vm_offset_t)pss.ps_argvstr;
1870 		vsize = pss.ps_nargvstr;
1871 		if (vsize > ARG_MAX)
1872 			return (ENOEXEC);
1873 		size = vsize * sizeof(char *);
1874 		break;
1875 	case PROC_ENV:
1876 		vptr = (vm_offset_t)pss.ps_envstr;
1877 		vsize = pss.ps_nenvstr;
1878 		if (vsize > ARG_MAX)
1879 			return (ENOEXEC);
1880 		size = vsize * sizeof(char *);
1881 		break;
1882 	case PROC_AUX:
1883 		/*
1884 		 * The aux array is just above env array on the stack. Check
1885 		 * that the address is naturally aligned.
1886 		 */
1887 		vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1888 		    * sizeof(char *);
1889 #if __ELF_WORD_SIZE == 64
1890 		if (vptr % sizeof(uint64_t) != 0)
1891 #else
1892 		if (vptr % sizeof(uint32_t) != 0)
1893 #endif
1894 			return (ENOEXEC);
1895 		/*
1896 		 * We count the array size reading the aux vectors from the
1897 		 * stack until AT_NULL vector is returned.  So (to keep the code
1898 		 * simple) we read the process stack twice: the first time here
1899 		 * to find the size and the second time when copying the vectors
1900 		 * to the allocated proc_vector.
1901 		 */
1902 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1903 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1904 			    sizeof(aux))
1905 				return (ENOMEM);
1906 			if (aux.a_type == AT_NULL)
1907 				break;
1908 			ptr += sizeof(aux);
1909 		}
1910 		/*
1911 		 * If the PROC_AUXV_MAX entries are iterated over, and we have
1912 		 * not reached AT_NULL, it is most likely we are reading wrong
1913 		 * data: either the process doesn't have auxv array or data has
1914 		 * been modified. Return the error in this case.
1915 		 */
1916 		if (aux.a_type != AT_NULL)
1917 			return (ENOEXEC);
1918 		vsize = i + 1;
1919 		size = vsize * sizeof(aux);
1920 		break;
1921 	default:
1922 		KASSERT(0, ("Wrong proc vector type: %d", type));
1923 		return (EINVAL); /* In case we are built without INVARIANTS. */
1924 	}
1925 	proc_vector = malloc(size, M_TEMP, M_WAITOK);
1926 	if (proc_readmem(td, p, vptr, proc_vector, size) != size) {
1927 		free(proc_vector, M_TEMP);
1928 		return (ENOMEM);
1929 	}
1930 	*proc_vectorp = proc_vector;
1931 	*vsizep = vsize;
1932 
1933 	return (0);
1934 }
1935 
1936 #define GET_PS_STRINGS_CHUNK_SZ	256	/* Chunk size (bytes) for ps_strings operations. */
1937 
1938 static int
1939 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
1940     enum proc_vector_type type)
1941 {
1942 	size_t done, len, nchr, vsize;
1943 	int error, i;
1944 	char **proc_vector, *sptr;
1945 	char pss_string[GET_PS_STRINGS_CHUNK_SZ];
1946 
1947 	PROC_ASSERT_HELD(p);
1948 
1949 	/*
1950 	 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
1951 	 */
1952 	nchr = 2 * (PATH_MAX + ARG_MAX);
1953 
1954 	error = get_proc_vector(td, p, &proc_vector, &vsize, type);
1955 	if (error != 0)
1956 		return (error);
1957 	for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
1958 		/*
1959 		 * The program may have scribbled into its argv array, e.g. to
1960 		 * remove some arguments.  If that has happened, break out
1961 		 * before trying to read from NULL.
1962 		 */
1963 		if (proc_vector[i] == NULL)
1964 			break;
1965 		for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
1966 			error = proc_read_string(td, p, sptr, pss_string,
1967 			    sizeof(pss_string));
1968 			if (error != 0)
1969 				goto done;
1970 			len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
1971 			if (done + len >= nchr)
1972 				len = nchr - done - 1;
1973 			sbuf_bcat(sb, pss_string, len);
1974 			if (len != GET_PS_STRINGS_CHUNK_SZ)
1975 				break;
1976 			done += GET_PS_STRINGS_CHUNK_SZ;
1977 		}
1978 		sbuf_bcat(sb, "", 1);
1979 		done += len + 1;
1980 	}
1981 done:
1982 	free(proc_vector, M_TEMP);
1983 	return (error);
1984 }
1985 
1986 int
1987 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
1988 {
1989 
1990 	return (get_ps_strings(curthread, p, sb, PROC_ARG));
1991 }
1992 
1993 int
1994 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
1995 {
1996 
1997 	return (get_ps_strings(curthread, p, sb, PROC_ENV));
1998 }
1999 
2000 int
2001 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb)
2002 {
2003 	size_t vsize, size;
2004 	char **auxv;
2005 	int error;
2006 
2007 	error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX);
2008 	if (error == 0) {
2009 #ifdef COMPAT_FREEBSD32
2010 		if (SV_PROC_FLAG(p, SV_ILP32) != 0)
2011 			size = vsize * sizeof(Elf32_Auxinfo);
2012 		else
2013 #endif
2014 			size = vsize * sizeof(Elf_Auxinfo);
2015 		if (sbuf_bcat(sb, auxv, size) != 0)
2016 			error = ENOMEM;
2017 		free(auxv, M_TEMP);
2018 	}
2019 	return (error);
2020 }
2021 
2022 /*
2023  * This sysctl allows a process to retrieve the argument list or process
2024  * title for another process without groping around in the address space
2025  * of the other process.  It also allow a process to set its own "process
2026  * title to a string of its own choice.
2027  */
2028 static int
2029 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
2030 {
2031 	int *name = (int *)arg1;
2032 	u_int namelen = arg2;
2033 	struct pargs *newpa, *pa;
2034 	struct proc *p;
2035 	struct sbuf sb;
2036 	int flags, error = 0, error2;
2037 	pid_t pid;
2038 
2039 	if (namelen != 1)
2040 		return (EINVAL);
2041 
2042 	pid = (pid_t)name[0];
2043 	/*
2044 	 * If the query is for this process and it is single-threaded, there
2045 	 * is nobody to modify pargs, thus we can just read.
2046 	 */
2047 	p = curproc;
2048 	if (pid == p->p_pid && p->p_numthreads == 1 && req->newptr == NULL &&
2049 	    (pa = p->p_args) != NULL)
2050 		return (SYSCTL_OUT(req, pa->ar_args, pa->ar_length));
2051 
2052 	flags = PGET_CANSEE;
2053 	if (req->newptr != NULL)
2054 		flags |= PGET_ISCURRENT;
2055 	error = pget(pid, flags, &p);
2056 	if (error)
2057 		return (error);
2058 
2059 	pa = p->p_args;
2060 	if (pa != NULL) {
2061 		pargs_hold(pa);
2062 		PROC_UNLOCK(p);
2063 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
2064 		pargs_drop(pa);
2065 	} else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
2066 		_PHOLD(p);
2067 		PROC_UNLOCK(p);
2068 		sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2069 		sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2070 		error = proc_getargv(curthread, p, &sb);
2071 		error2 = sbuf_finish(&sb);
2072 		PRELE(p);
2073 		sbuf_delete(&sb);
2074 		if (error == 0 && error2 != 0)
2075 			error = error2;
2076 	} else {
2077 		PROC_UNLOCK(p);
2078 	}
2079 	if (error != 0 || req->newptr == NULL)
2080 		return (error);
2081 
2082 	if (req->newlen > ps_arg_cache_limit - sizeof(struct pargs))
2083 		return (ENOMEM);
2084 
2085 	if (req->newlen == 0) {
2086 		/*
2087 		 * Clear the argument pointer, so that we'll fetch arguments
2088 		 * with proc_getargv() until further notice.
2089 		 */
2090 		newpa = NULL;
2091 	} else {
2092 		newpa = pargs_alloc(req->newlen);
2093 		error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
2094 		if (error != 0) {
2095 			pargs_free(newpa);
2096 			return (error);
2097 		}
2098 	}
2099 	PROC_LOCK(p);
2100 	pa = p->p_args;
2101 	p->p_args = newpa;
2102 	PROC_UNLOCK(p);
2103 	pargs_drop(pa);
2104 	return (0);
2105 }
2106 
2107 /*
2108  * This sysctl allows a process to retrieve environment of another process.
2109  */
2110 static int
2111 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
2112 {
2113 	int *name = (int *)arg1;
2114 	u_int namelen = arg2;
2115 	struct proc *p;
2116 	struct sbuf sb;
2117 	int error, error2;
2118 
2119 	if (namelen != 1)
2120 		return (EINVAL);
2121 
2122 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2123 	if (error != 0)
2124 		return (error);
2125 	if ((p->p_flag & P_SYSTEM) != 0) {
2126 		PRELE(p);
2127 		return (0);
2128 	}
2129 
2130 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2131 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2132 	error = proc_getenvv(curthread, p, &sb);
2133 	error2 = sbuf_finish(&sb);
2134 	PRELE(p);
2135 	sbuf_delete(&sb);
2136 	return (error != 0 ? error : error2);
2137 }
2138 
2139 /*
2140  * This sysctl allows a process to retrieve ELF auxiliary vector of
2141  * another process.
2142  */
2143 static int
2144 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
2145 {
2146 	int *name = (int *)arg1;
2147 	u_int namelen = arg2;
2148 	struct proc *p;
2149 	struct sbuf sb;
2150 	int error, error2;
2151 
2152 	if (namelen != 1)
2153 		return (EINVAL);
2154 
2155 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2156 	if (error != 0)
2157 		return (error);
2158 	if ((p->p_flag & P_SYSTEM) != 0) {
2159 		PRELE(p);
2160 		return (0);
2161 	}
2162 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2163 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2164 	error = proc_getauxv(curthread, p, &sb);
2165 	error2 = sbuf_finish(&sb);
2166 	PRELE(p);
2167 	sbuf_delete(&sb);
2168 	return (error != 0 ? error : error2);
2169 }
2170 
2171 /*
2172  * This sysctl allows a process to retrieve the path of the executable for
2173  * itself or another process.
2174  */
2175 static int
2176 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
2177 {
2178 	pid_t *pidp = (pid_t *)arg1;
2179 	unsigned int arglen = arg2;
2180 	struct proc *p;
2181 	struct vnode *vp;
2182 	char *retbuf, *freebuf;
2183 	int error;
2184 
2185 	if (arglen != 1)
2186 		return (EINVAL);
2187 	if (*pidp == -1) {	/* -1 means this process */
2188 		p = req->td->td_proc;
2189 	} else {
2190 		error = pget(*pidp, PGET_CANSEE, &p);
2191 		if (error != 0)
2192 			return (error);
2193 	}
2194 
2195 	vp = p->p_textvp;
2196 	if (vp == NULL) {
2197 		if (*pidp != -1)
2198 			PROC_UNLOCK(p);
2199 		return (0);
2200 	}
2201 	vref(vp);
2202 	if (*pidp != -1)
2203 		PROC_UNLOCK(p);
2204 	error = vn_fullpath(req->td, vp, &retbuf, &freebuf);
2205 	vrele(vp);
2206 	if (error)
2207 		return (error);
2208 	error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
2209 	free(freebuf, M_TEMP);
2210 	return (error);
2211 }
2212 
2213 static int
2214 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
2215 {
2216 	struct proc *p;
2217 	char *sv_name;
2218 	int *name;
2219 	int namelen;
2220 	int error;
2221 
2222 	namelen = arg2;
2223 	if (namelen != 1)
2224 		return (EINVAL);
2225 
2226 	name = (int *)arg1;
2227 	error = pget((pid_t)name[0], PGET_CANSEE, &p);
2228 	if (error != 0)
2229 		return (error);
2230 	sv_name = p->p_sysent->sv_name;
2231 	PROC_UNLOCK(p);
2232 	return (sysctl_handle_string(oidp, sv_name, 0, req));
2233 }
2234 
2235 #ifdef KINFO_OVMENTRY_SIZE
2236 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
2237 #endif
2238 
2239 #ifdef COMPAT_FREEBSD7
2240 static int
2241 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
2242 {
2243 	vm_map_entry_t entry, tmp_entry;
2244 	unsigned int last_timestamp;
2245 	char *fullpath, *freepath;
2246 	struct kinfo_ovmentry *kve;
2247 	struct vattr va;
2248 	struct ucred *cred;
2249 	int error, *name;
2250 	struct vnode *vp;
2251 	struct proc *p;
2252 	vm_map_t map;
2253 	struct vmspace *vm;
2254 
2255 	name = (int *)arg1;
2256 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2257 	if (error != 0)
2258 		return (error);
2259 	vm = vmspace_acquire_ref(p);
2260 	if (vm == NULL) {
2261 		PRELE(p);
2262 		return (ESRCH);
2263 	}
2264 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2265 
2266 	map = &vm->vm_map;
2267 	vm_map_lock_read(map);
2268 	for (entry = map->header.next; entry != &map->header;
2269 	    entry = entry->next) {
2270 		vm_object_t obj, tobj, lobj;
2271 		vm_offset_t addr;
2272 
2273 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2274 			continue;
2275 
2276 		bzero(kve, sizeof(*kve));
2277 		kve->kve_structsize = sizeof(*kve);
2278 
2279 		kve->kve_private_resident = 0;
2280 		obj = entry->object.vm_object;
2281 		if (obj != NULL) {
2282 			VM_OBJECT_RLOCK(obj);
2283 			if (obj->shadow_count == 1)
2284 				kve->kve_private_resident =
2285 				    obj->resident_page_count;
2286 		}
2287 		kve->kve_resident = 0;
2288 		addr = entry->start;
2289 		while (addr < entry->end) {
2290 			if (pmap_extract(map->pmap, addr))
2291 				kve->kve_resident++;
2292 			addr += PAGE_SIZE;
2293 		}
2294 
2295 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2296 			if (tobj != obj) {
2297 				VM_OBJECT_RLOCK(tobj);
2298 				kve->kve_offset += tobj->backing_object_offset;
2299 			}
2300 			if (lobj != obj)
2301 				VM_OBJECT_RUNLOCK(lobj);
2302 			lobj = tobj;
2303 		}
2304 
2305 		kve->kve_start = (void*)entry->start;
2306 		kve->kve_end = (void*)entry->end;
2307 		kve->kve_offset += (off_t)entry->offset;
2308 
2309 		if (entry->protection & VM_PROT_READ)
2310 			kve->kve_protection |= KVME_PROT_READ;
2311 		if (entry->protection & VM_PROT_WRITE)
2312 			kve->kve_protection |= KVME_PROT_WRITE;
2313 		if (entry->protection & VM_PROT_EXECUTE)
2314 			kve->kve_protection |= KVME_PROT_EXEC;
2315 
2316 		if (entry->eflags & MAP_ENTRY_COW)
2317 			kve->kve_flags |= KVME_FLAG_COW;
2318 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2319 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2320 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2321 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2322 
2323 		last_timestamp = map->timestamp;
2324 		vm_map_unlock_read(map);
2325 
2326 		kve->kve_fileid = 0;
2327 		kve->kve_fsid = 0;
2328 		freepath = NULL;
2329 		fullpath = "";
2330 		if (lobj) {
2331 			kve->kve_type = vm_object_kvme_type(lobj, &vp);
2332 			if (kve->kve_type == KVME_TYPE_MGTDEVICE)
2333 				kve->kve_type = KVME_TYPE_UNKNOWN;
2334 			if (vp != NULL)
2335 				vref(vp);
2336 			if (lobj != obj)
2337 				VM_OBJECT_RUNLOCK(lobj);
2338 
2339 			kve->kve_ref_count = obj->ref_count;
2340 			kve->kve_shadow_count = obj->shadow_count;
2341 			VM_OBJECT_RUNLOCK(obj);
2342 			if (vp != NULL) {
2343 				vn_fullpath(curthread, vp, &fullpath,
2344 				    &freepath);
2345 				cred = curthread->td_ucred;
2346 				vn_lock(vp, LK_SHARED | LK_RETRY);
2347 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2348 					kve->kve_fileid = va.va_fileid;
2349 					/* truncate */
2350 					kve->kve_fsid = va.va_fsid;
2351 				}
2352 				vput(vp);
2353 			}
2354 		} else {
2355 			kve->kve_type = KVME_TYPE_NONE;
2356 			kve->kve_ref_count = 0;
2357 			kve->kve_shadow_count = 0;
2358 		}
2359 
2360 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2361 		if (freepath != NULL)
2362 			free(freepath, M_TEMP);
2363 
2364 		error = SYSCTL_OUT(req, kve, sizeof(*kve));
2365 		vm_map_lock_read(map);
2366 		if (error)
2367 			break;
2368 		if (last_timestamp != map->timestamp) {
2369 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2370 			entry = tmp_entry;
2371 		}
2372 	}
2373 	vm_map_unlock_read(map);
2374 	vmspace_free(vm);
2375 	PRELE(p);
2376 	free(kve, M_TEMP);
2377 	return (error);
2378 }
2379 #endif	/* COMPAT_FREEBSD7 */
2380 
2381 #ifdef KINFO_VMENTRY_SIZE
2382 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2383 #endif
2384 
2385 void
2386 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry,
2387     int *resident_count, bool *super)
2388 {
2389 	vm_object_t obj, tobj;
2390 	vm_page_t m, m_adv;
2391 	vm_offset_t addr;
2392 	vm_paddr_t locked_pa;
2393 	vm_pindex_t pi, pi_adv, pindex;
2394 
2395 	*super = false;
2396 	*resident_count = 0;
2397 	if (vmmap_skip_res_cnt)
2398 		return;
2399 
2400 	locked_pa = 0;
2401 	obj = entry->object.vm_object;
2402 	addr = entry->start;
2403 	m_adv = NULL;
2404 	pi = OFF_TO_IDX(entry->offset);
2405 	for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) {
2406 		if (m_adv != NULL) {
2407 			m = m_adv;
2408 		} else {
2409 			pi_adv = atop(entry->end - addr);
2410 			pindex = pi;
2411 			for (tobj = obj;; tobj = tobj->backing_object) {
2412 				m = vm_page_find_least(tobj, pindex);
2413 				if (m != NULL) {
2414 					if (m->pindex == pindex)
2415 						break;
2416 					if (pi_adv > m->pindex - pindex) {
2417 						pi_adv = m->pindex - pindex;
2418 						m_adv = m;
2419 					}
2420 				}
2421 				if (tobj->backing_object == NULL)
2422 					goto next;
2423 				pindex += OFF_TO_IDX(tobj->
2424 				    backing_object_offset);
2425 			}
2426 		}
2427 		m_adv = NULL;
2428 		if (m->psind != 0 && addr + pagesizes[1] <= entry->end &&
2429 		    (addr & (pagesizes[1] - 1)) == 0 &&
2430 		    (pmap_mincore(map->pmap, addr, &locked_pa) &
2431 		    MINCORE_SUPER) != 0) {
2432 			*super = true;
2433 			pi_adv = atop(pagesizes[1]);
2434 		} else {
2435 			/*
2436 			 * We do not test the found page on validity.
2437 			 * Either the page is busy and being paged in,
2438 			 * or it was invalidated.  The first case
2439 			 * should be counted as resident, the second
2440 			 * is not so clear; we do account both.
2441 			 */
2442 			pi_adv = 1;
2443 		}
2444 		*resident_count += pi_adv;
2445 next:;
2446 	}
2447 	PA_UNLOCK_COND(locked_pa);
2448 }
2449 
2450 /*
2451  * Must be called with the process locked and will return unlocked.
2452  */
2453 int
2454 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb, ssize_t maxlen, int flags)
2455 {
2456 	vm_map_entry_t entry, tmp_entry;
2457 	struct vattr va;
2458 	vm_map_t map;
2459 	vm_object_t obj, tobj, lobj;
2460 	char *fullpath, *freepath;
2461 	struct kinfo_vmentry *kve;
2462 	struct ucred *cred;
2463 	struct vnode *vp;
2464 	struct vmspace *vm;
2465 	vm_offset_t addr;
2466 	unsigned int last_timestamp;
2467 	int error;
2468 	bool super;
2469 
2470 	PROC_LOCK_ASSERT(p, MA_OWNED);
2471 
2472 	_PHOLD(p);
2473 	PROC_UNLOCK(p);
2474 	vm = vmspace_acquire_ref(p);
2475 	if (vm == NULL) {
2476 		PRELE(p);
2477 		return (ESRCH);
2478 	}
2479 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK | M_ZERO);
2480 
2481 	error = 0;
2482 	map = &vm->vm_map;
2483 	vm_map_lock_read(map);
2484 	for (entry = map->header.next; entry != &map->header;
2485 	    entry = entry->next) {
2486 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2487 			continue;
2488 
2489 		addr = entry->end;
2490 		bzero(kve, sizeof(*kve));
2491 		obj = entry->object.vm_object;
2492 		if (obj != NULL) {
2493 			for (tobj = obj; tobj != NULL;
2494 			    tobj = tobj->backing_object) {
2495 				VM_OBJECT_RLOCK(tobj);
2496 				kve->kve_offset += tobj->backing_object_offset;
2497 				lobj = tobj;
2498 			}
2499 			if (obj->backing_object == NULL)
2500 				kve->kve_private_resident =
2501 				    obj->resident_page_count;
2502 			kern_proc_vmmap_resident(map, entry,
2503 			    &kve->kve_resident, &super);
2504 			if (super)
2505 				kve->kve_flags |= KVME_FLAG_SUPER;
2506 			for (tobj = obj; tobj != NULL;
2507 			    tobj = tobj->backing_object) {
2508 				if (tobj != obj && tobj != lobj)
2509 					VM_OBJECT_RUNLOCK(tobj);
2510 			}
2511 		} else {
2512 			lobj = NULL;
2513 		}
2514 
2515 		kve->kve_start = entry->start;
2516 		kve->kve_end = entry->end;
2517 		kve->kve_offset += entry->offset;
2518 
2519 		if (entry->protection & VM_PROT_READ)
2520 			kve->kve_protection |= KVME_PROT_READ;
2521 		if (entry->protection & VM_PROT_WRITE)
2522 			kve->kve_protection |= KVME_PROT_WRITE;
2523 		if (entry->protection & VM_PROT_EXECUTE)
2524 			kve->kve_protection |= KVME_PROT_EXEC;
2525 
2526 		if (entry->eflags & MAP_ENTRY_COW)
2527 			kve->kve_flags |= KVME_FLAG_COW;
2528 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2529 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2530 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2531 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2532 		if (entry->eflags & MAP_ENTRY_GROWS_UP)
2533 			kve->kve_flags |= KVME_FLAG_GROWS_UP;
2534 		if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2535 			kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2536 		if (entry->eflags & MAP_ENTRY_USER_WIRED)
2537 			kve->kve_flags |= KVME_FLAG_USER_WIRED;
2538 
2539 		last_timestamp = map->timestamp;
2540 		vm_map_unlock_read(map);
2541 
2542 		freepath = NULL;
2543 		fullpath = "";
2544 		if (lobj != NULL) {
2545 			kve->kve_type = vm_object_kvme_type(lobj, &vp);
2546 			if (vp != NULL)
2547 				vref(vp);
2548 			if (lobj != obj)
2549 				VM_OBJECT_RUNLOCK(lobj);
2550 
2551 			kve->kve_ref_count = obj->ref_count;
2552 			kve->kve_shadow_count = obj->shadow_count;
2553 			VM_OBJECT_RUNLOCK(obj);
2554 			if (vp != NULL) {
2555 				vn_fullpath(curthread, vp, &fullpath,
2556 				    &freepath);
2557 				kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2558 				cred = curthread->td_ucred;
2559 				vn_lock(vp, LK_SHARED | LK_RETRY);
2560 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2561 					kve->kve_vn_fileid = va.va_fileid;
2562 					kve->kve_vn_fsid = va.va_fsid;
2563 					kve->kve_vn_fsid_freebsd11 =
2564 					    kve->kve_vn_fsid; /* truncate */
2565 					kve->kve_vn_mode =
2566 					    MAKEIMODE(va.va_type, va.va_mode);
2567 					kve->kve_vn_size = va.va_size;
2568 					kve->kve_vn_rdev = va.va_rdev;
2569 					kve->kve_vn_rdev_freebsd11 =
2570 					    kve->kve_vn_rdev; /* truncate */
2571 					kve->kve_status = KF_ATTR_VALID;
2572 				}
2573 				vput(vp);
2574 			}
2575 		} else {
2576 			kve->kve_type = KVME_TYPE_NONE;
2577 			kve->kve_ref_count = 0;
2578 			kve->kve_shadow_count = 0;
2579 		}
2580 
2581 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2582 		if (freepath != NULL)
2583 			free(freepath, M_TEMP);
2584 
2585 		/* Pack record size down */
2586 		if ((flags & KERN_VMMAP_PACK_KINFO) != 0)
2587 			kve->kve_structsize =
2588 			    offsetof(struct kinfo_vmentry, kve_path) +
2589 			    strlen(kve->kve_path) + 1;
2590 		else
2591 			kve->kve_structsize = sizeof(*kve);
2592 		kve->kve_structsize = roundup(kve->kve_structsize,
2593 		    sizeof(uint64_t));
2594 
2595 		/* Halt filling and truncate rather than exceeding maxlen */
2596 		if (maxlen != -1 && maxlen < kve->kve_structsize) {
2597 			error = 0;
2598 			vm_map_lock_read(map);
2599 			break;
2600 		} else if (maxlen != -1)
2601 			maxlen -= kve->kve_structsize;
2602 
2603 		if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0)
2604 			error = ENOMEM;
2605 		vm_map_lock_read(map);
2606 		if (error != 0)
2607 			break;
2608 		if (last_timestamp != map->timestamp) {
2609 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2610 			entry = tmp_entry;
2611 		}
2612 	}
2613 	vm_map_unlock_read(map);
2614 	vmspace_free(vm);
2615 	PRELE(p);
2616 	free(kve, M_TEMP);
2617 	return (error);
2618 }
2619 
2620 static int
2621 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2622 {
2623 	struct proc *p;
2624 	struct sbuf sb;
2625 	int error, error2, *name;
2626 
2627 	name = (int *)arg1;
2628 	sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req);
2629 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2630 	error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
2631 	if (error != 0) {
2632 		sbuf_delete(&sb);
2633 		return (error);
2634 	}
2635 	error = kern_proc_vmmap_out(p, &sb, -1, KERN_VMMAP_PACK_KINFO);
2636 	error2 = sbuf_finish(&sb);
2637 	sbuf_delete(&sb);
2638 	return (error != 0 ? error : error2);
2639 }
2640 
2641 #if defined(STACK) || defined(DDB)
2642 static int
2643 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2644 {
2645 	struct kinfo_kstack *kkstp;
2646 	int error, i, *name, numthreads;
2647 	lwpid_t *lwpidarray;
2648 	struct thread *td;
2649 	struct stack *st;
2650 	struct sbuf sb;
2651 	struct proc *p;
2652 
2653 	name = (int *)arg1;
2654 	error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2655 	if (error != 0)
2656 		return (error);
2657 
2658 	kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2659 	st = stack_create(M_WAITOK);
2660 
2661 	lwpidarray = NULL;
2662 	PROC_LOCK(p);
2663 	do {
2664 		if (lwpidarray != NULL) {
2665 			free(lwpidarray, M_TEMP);
2666 			lwpidarray = NULL;
2667 		}
2668 		numthreads = p->p_numthreads;
2669 		PROC_UNLOCK(p);
2670 		lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2671 		    M_WAITOK | M_ZERO);
2672 		PROC_LOCK(p);
2673 	} while (numthreads < p->p_numthreads);
2674 
2675 	/*
2676 	 * XXXRW: During the below loop, execve(2) and countless other sorts
2677 	 * of changes could have taken place.  Should we check to see if the
2678 	 * vmspace has been replaced, or the like, in order to prevent
2679 	 * giving a snapshot that spans, say, execve(2), with some threads
2680 	 * before and some after?  Among other things, the credentials could
2681 	 * have changed, in which case the right to extract debug info might
2682 	 * no longer be assured.
2683 	 */
2684 	i = 0;
2685 	FOREACH_THREAD_IN_PROC(p, td) {
2686 		KASSERT(i < numthreads,
2687 		    ("sysctl_kern_proc_kstack: numthreads"));
2688 		lwpidarray[i] = td->td_tid;
2689 		i++;
2690 	}
2691 	numthreads = i;
2692 	for (i = 0; i < numthreads; i++) {
2693 		td = thread_find(p, lwpidarray[i]);
2694 		if (td == NULL) {
2695 			continue;
2696 		}
2697 		bzero(kkstp, sizeof(*kkstp));
2698 		(void)sbuf_new(&sb, kkstp->kkst_trace,
2699 		    sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2700 		thread_lock(td);
2701 		kkstp->kkst_tid = td->td_tid;
2702 		if (TD_IS_SWAPPED(td)) {
2703 			kkstp->kkst_state = KKST_STATE_SWAPPED;
2704 		} else if (TD_IS_RUNNING(td)) {
2705 			if (stack_save_td_running(st, td) == 0)
2706 				kkstp->kkst_state = KKST_STATE_STACKOK;
2707 			else
2708 				kkstp->kkst_state = KKST_STATE_RUNNING;
2709 		} else {
2710 			kkstp->kkst_state = KKST_STATE_STACKOK;
2711 			stack_save_td(st, td);
2712 		}
2713 		thread_unlock(td);
2714 		PROC_UNLOCK(p);
2715 		stack_sbuf_print(&sb, st);
2716 		sbuf_finish(&sb);
2717 		sbuf_delete(&sb);
2718 		error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2719 		PROC_LOCK(p);
2720 		if (error)
2721 			break;
2722 	}
2723 	_PRELE(p);
2724 	PROC_UNLOCK(p);
2725 	if (lwpidarray != NULL)
2726 		free(lwpidarray, M_TEMP);
2727 	stack_destroy(st);
2728 	free(kkstp, M_TEMP);
2729 	return (error);
2730 }
2731 #endif
2732 
2733 /*
2734  * This sysctl allows a process to retrieve the full list of groups from
2735  * itself or another process.
2736  */
2737 static int
2738 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2739 {
2740 	pid_t *pidp = (pid_t *)arg1;
2741 	unsigned int arglen = arg2;
2742 	struct proc *p;
2743 	struct ucred *cred;
2744 	int error;
2745 
2746 	if (arglen != 1)
2747 		return (EINVAL);
2748 	if (*pidp == -1) {	/* -1 means this process */
2749 		p = req->td->td_proc;
2750 		PROC_LOCK(p);
2751 	} else {
2752 		error = pget(*pidp, PGET_CANSEE, &p);
2753 		if (error != 0)
2754 			return (error);
2755 	}
2756 
2757 	cred = crhold(p->p_ucred);
2758 	PROC_UNLOCK(p);
2759 
2760 	error = SYSCTL_OUT(req, cred->cr_groups,
2761 	    cred->cr_ngroups * sizeof(gid_t));
2762 	crfree(cred);
2763 	return (error);
2764 }
2765 
2766 /*
2767  * This sysctl allows a process to retrieve or/and set the resource limit for
2768  * another process.
2769  */
2770 static int
2771 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2772 {
2773 	int *name = (int *)arg1;
2774 	u_int namelen = arg2;
2775 	struct rlimit rlim;
2776 	struct proc *p;
2777 	u_int which;
2778 	int flags, error;
2779 
2780 	if (namelen != 2)
2781 		return (EINVAL);
2782 
2783 	which = (u_int)name[1];
2784 	if (which >= RLIM_NLIMITS)
2785 		return (EINVAL);
2786 
2787 	if (req->newptr != NULL && req->newlen != sizeof(rlim))
2788 		return (EINVAL);
2789 
2790 	flags = PGET_HOLD | PGET_NOTWEXIT;
2791 	if (req->newptr != NULL)
2792 		flags |= PGET_CANDEBUG;
2793 	else
2794 		flags |= PGET_CANSEE;
2795 	error = pget((pid_t)name[0], flags, &p);
2796 	if (error != 0)
2797 		return (error);
2798 
2799 	/*
2800 	 * Retrieve limit.
2801 	 */
2802 	if (req->oldptr != NULL) {
2803 		PROC_LOCK(p);
2804 		lim_rlimit_proc(p, which, &rlim);
2805 		PROC_UNLOCK(p);
2806 	}
2807 	error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
2808 	if (error != 0)
2809 		goto errout;
2810 
2811 	/*
2812 	 * Set limit.
2813 	 */
2814 	if (req->newptr != NULL) {
2815 		error = SYSCTL_IN(req, &rlim, sizeof(rlim));
2816 		if (error == 0)
2817 			error = kern_proc_setrlimit(curthread, p, which, &rlim);
2818 	}
2819 
2820 errout:
2821 	PRELE(p);
2822 	return (error);
2823 }
2824 
2825 /*
2826  * This sysctl allows a process to retrieve ps_strings structure location of
2827  * another process.
2828  */
2829 static int
2830 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
2831 {
2832 	int *name = (int *)arg1;
2833 	u_int namelen = arg2;
2834 	struct proc *p;
2835 	vm_offset_t ps_strings;
2836 	int error;
2837 #ifdef COMPAT_FREEBSD32
2838 	uint32_t ps_strings32;
2839 #endif
2840 
2841 	if (namelen != 1)
2842 		return (EINVAL);
2843 
2844 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2845 	if (error != 0)
2846 		return (error);
2847 #ifdef COMPAT_FREEBSD32
2848 	if ((req->flags & SCTL_MASK32) != 0) {
2849 		/*
2850 		 * We return 0 if the 32 bit emulation request is for a 64 bit
2851 		 * process.
2852 		 */
2853 		ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
2854 		    PTROUT(p->p_sysent->sv_psstrings) : 0;
2855 		PROC_UNLOCK(p);
2856 		error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
2857 		return (error);
2858 	}
2859 #endif
2860 	ps_strings = p->p_sysent->sv_psstrings;
2861 	PROC_UNLOCK(p);
2862 	error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
2863 	return (error);
2864 }
2865 
2866 /*
2867  * This sysctl allows a process to retrieve umask of another process.
2868  */
2869 static int
2870 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
2871 {
2872 	int *name = (int *)arg1;
2873 	u_int namelen = arg2;
2874 	struct proc *p;
2875 	int error;
2876 	u_short fd_cmask;
2877 	pid_t pid;
2878 
2879 	if (namelen != 1)
2880 		return (EINVAL);
2881 
2882 	pid = (pid_t)name[0];
2883 	p = curproc;
2884 	if (pid == p->p_pid || pid == 0) {
2885 		fd_cmask = p->p_fd->fd_cmask;
2886 		goto out;
2887 	}
2888 
2889 	error = pget(pid, PGET_WANTREAD, &p);
2890 	if (error != 0)
2891 		return (error);
2892 
2893 	fd_cmask = p->p_fd->fd_cmask;
2894 	PRELE(p);
2895 out:
2896 	error = SYSCTL_OUT(req, &fd_cmask, sizeof(fd_cmask));
2897 	return (error);
2898 }
2899 
2900 /*
2901  * This sysctl allows a process to set and retrieve binary osreldate of
2902  * another process.
2903  */
2904 static int
2905 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
2906 {
2907 	int *name = (int *)arg1;
2908 	u_int namelen = arg2;
2909 	struct proc *p;
2910 	int flags, error, osrel;
2911 
2912 	if (namelen != 1)
2913 		return (EINVAL);
2914 
2915 	if (req->newptr != NULL && req->newlen != sizeof(osrel))
2916 		return (EINVAL);
2917 
2918 	flags = PGET_HOLD | PGET_NOTWEXIT;
2919 	if (req->newptr != NULL)
2920 		flags |= PGET_CANDEBUG;
2921 	else
2922 		flags |= PGET_CANSEE;
2923 	error = pget((pid_t)name[0], flags, &p);
2924 	if (error != 0)
2925 		return (error);
2926 
2927 	error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
2928 	if (error != 0)
2929 		goto errout;
2930 
2931 	if (req->newptr != NULL) {
2932 		error = SYSCTL_IN(req, &osrel, sizeof(osrel));
2933 		if (error != 0)
2934 			goto errout;
2935 		if (osrel < 0) {
2936 			error = EINVAL;
2937 			goto errout;
2938 		}
2939 		p->p_osrel = osrel;
2940 	}
2941 errout:
2942 	PRELE(p);
2943 	return (error);
2944 }
2945 
2946 static int
2947 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS)
2948 {
2949 	int *name = (int *)arg1;
2950 	u_int namelen = arg2;
2951 	struct proc *p;
2952 	struct kinfo_sigtramp kst;
2953 	const struct sysentvec *sv;
2954 	int error;
2955 #ifdef COMPAT_FREEBSD32
2956 	struct kinfo_sigtramp32 kst32;
2957 #endif
2958 
2959 	if (namelen != 1)
2960 		return (EINVAL);
2961 
2962 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2963 	if (error != 0)
2964 		return (error);
2965 	sv = p->p_sysent;
2966 #ifdef COMPAT_FREEBSD32
2967 	if ((req->flags & SCTL_MASK32) != 0) {
2968 		bzero(&kst32, sizeof(kst32));
2969 		if (SV_PROC_FLAG(p, SV_ILP32)) {
2970 			if (sv->sv_sigcode_base != 0) {
2971 				kst32.ksigtramp_start = sv->sv_sigcode_base;
2972 				kst32.ksigtramp_end = sv->sv_sigcode_base +
2973 				    *sv->sv_szsigcode;
2974 			} else {
2975 				kst32.ksigtramp_start = sv->sv_psstrings -
2976 				    *sv->sv_szsigcode;
2977 				kst32.ksigtramp_end = sv->sv_psstrings;
2978 			}
2979 		}
2980 		PROC_UNLOCK(p);
2981 		error = SYSCTL_OUT(req, &kst32, sizeof(kst32));
2982 		return (error);
2983 	}
2984 #endif
2985 	bzero(&kst, sizeof(kst));
2986 	if (sv->sv_sigcode_base != 0) {
2987 		kst.ksigtramp_start = (char *)sv->sv_sigcode_base;
2988 		kst.ksigtramp_end = (char *)sv->sv_sigcode_base +
2989 		    *sv->sv_szsigcode;
2990 	} else {
2991 		kst.ksigtramp_start = (char *)sv->sv_psstrings -
2992 		    *sv->sv_szsigcode;
2993 		kst.ksigtramp_end = (char *)sv->sv_psstrings;
2994 	}
2995 	PROC_UNLOCK(p);
2996 	error = SYSCTL_OUT(req, &kst, sizeof(kst));
2997 	return (error);
2998 }
2999 
3000 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD,  0, "Process table");
3001 
3002 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
3003 	CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
3004 	"Return entire process table");
3005 
3006 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3007 	sysctl_kern_proc, "Process table");
3008 
3009 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
3010 	sysctl_kern_proc, "Process table");
3011 
3012 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3013 	sysctl_kern_proc, "Process table");
3014 
3015 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
3016 	CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3017 
3018 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
3019 	sysctl_kern_proc, "Process table");
3020 
3021 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3022 	sysctl_kern_proc, "Process table");
3023 
3024 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3025 	sysctl_kern_proc, "Process table");
3026 
3027 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3028 	sysctl_kern_proc, "Process table");
3029 
3030 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
3031 	sysctl_kern_proc, "Return process table, no threads");
3032 
3033 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
3034 	CTLFLAG_RW | CTLFLAG_CAPWR | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
3035 	sysctl_kern_proc_args, "Process argument list");
3036 
3037 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
3038 	sysctl_kern_proc_env, "Process environment");
3039 
3040 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
3041 	CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
3042 
3043 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
3044 	CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
3045 
3046 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
3047 	CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
3048 	"Process syscall vector name (ABI type)");
3049 
3050 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
3051 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3052 
3053 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
3054 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3055 
3056 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
3057 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3058 
3059 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
3060 	sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3061 
3062 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
3063 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3064 
3065 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
3066 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3067 
3068 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
3069 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3070 
3071 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
3072 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3073 
3074 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
3075 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
3076 	"Return process table, no threads");
3077 
3078 #ifdef COMPAT_FREEBSD7
3079 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
3080 	CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
3081 #endif
3082 
3083 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
3084 	CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
3085 
3086 #if defined(STACK) || defined(DDB)
3087 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
3088 	CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
3089 #endif
3090 
3091 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
3092 	CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
3093 
3094 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
3095 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
3096 	"Process resource limits");
3097 
3098 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
3099 	CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
3100 	"Process ps_strings location");
3101 
3102 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
3103 	CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
3104 
3105 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
3106 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
3107 	"Process binary osreldate");
3108 
3109 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD |
3110 	CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp,
3111 	"Process signal trampoline location");
3112 
3113 int allproc_gen;
3114 
3115 /*
3116  * stop_all_proc() purpose is to stop all process which have usermode,
3117  * except current process for obvious reasons.  This makes it somewhat
3118  * unreliable when invoked from multithreaded process.  The service
3119  * must not be user-callable anyway.
3120  */
3121 void
3122 stop_all_proc(void)
3123 {
3124 	struct proc *cp, *p;
3125 	int r, gen;
3126 	bool restart, seen_stopped, seen_exiting, stopped_some;
3127 
3128 	cp = curproc;
3129 allproc_loop:
3130 	sx_xlock(&allproc_lock);
3131 	gen = allproc_gen;
3132 	seen_exiting = seen_stopped = stopped_some = restart = false;
3133 	LIST_REMOVE(cp, p_list);
3134 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3135 	for (;;) {
3136 		p = LIST_NEXT(cp, p_list);
3137 		if (p == NULL)
3138 			break;
3139 		LIST_REMOVE(cp, p_list);
3140 		LIST_INSERT_AFTER(p, cp, p_list);
3141 		PROC_LOCK(p);
3142 		if ((p->p_flag & (P_KPROC | P_SYSTEM | P_TOTAL_STOP)) != 0) {
3143 			PROC_UNLOCK(p);
3144 			continue;
3145 		}
3146 		if ((p->p_flag & P_WEXIT) != 0) {
3147 			seen_exiting = true;
3148 			PROC_UNLOCK(p);
3149 			continue;
3150 		}
3151 		if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) {
3152 			/*
3153 			 * Stopped processes are tolerated when there
3154 			 * are no other processes which might continue
3155 			 * them.  P_STOPPED_SINGLE but not
3156 			 * P_TOTAL_STOP process still has at least one
3157 			 * thread running.
3158 			 */
3159 			seen_stopped = true;
3160 			PROC_UNLOCK(p);
3161 			continue;
3162 		}
3163 		sx_xunlock(&allproc_lock);
3164 		_PHOLD(p);
3165 		r = thread_single(p, SINGLE_ALLPROC);
3166 		if (r != 0)
3167 			restart = true;
3168 		else
3169 			stopped_some = true;
3170 		_PRELE(p);
3171 		PROC_UNLOCK(p);
3172 		sx_xlock(&allproc_lock);
3173 	}
3174 	/* Catch forked children we did not see in iteration. */
3175 	if (gen != allproc_gen)
3176 		restart = true;
3177 	sx_xunlock(&allproc_lock);
3178 	if (restart || stopped_some || seen_exiting || seen_stopped) {
3179 		kern_yield(PRI_USER);
3180 		goto allproc_loop;
3181 	}
3182 }
3183 
3184 void
3185 resume_all_proc(void)
3186 {
3187 	struct proc *cp, *p;
3188 
3189 	cp = curproc;
3190 	sx_xlock(&allproc_lock);
3191 again:
3192 	LIST_REMOVE(cp, p_list);
3193 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3194 	for (;;) {
3195 		p = LIST_NEXT(cp, p_list);
3196 		if (p == NULL)
3197 			break;
3198 		LIST_REMOVE(cp, p_list);
3199 		LIST_INSERT_AFTER(p, cp, p_list);
3200 		PROC_LOCK(p);
3201 		if ((p->p_flag & P_TOTAL_STOP) != 0) {
3202 			sx_xunlock(&allproc_lock);
3203 			_PHOLD(p);
3204 			thread_single_end(p, SINGLE_ALLPROC);
3205 			_PRELE(p);
3206 			PROC_UNLOCK(p);
3207 			sx_xlock(&allproc_lock);
3208 		} else {
3209 			PROC_UNLOCK(p);
3210 		}
3211 	}
3212 	/*  Did the loop above missed any stopped process ? */
3213 	FOREACH_PROC_IN_SYSTEM(p) {
3214 		/* No need for proc lock. */
3215 		if ((p->p_flag & P_TOTAL_STOP) != 0)
3216 			goto again;
3217 	}
3218 	sx_xunlock(&allproc_lock);
3219 }
3220 
3221 /* #define	TOTAL_STOP_DEBUG	1 */
3222 #ifdef TOTAL_STOP_DEBUG
3223 volatile static int ap_resume;
3224 #include <sys/mount.h>
3225 
3226 static int
3227 sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS)
3228 {
3229 	int error, val;
3230 
3231 	val = 0;
3232 	ap_resume = 0;
3233 	error = sysctl_handle_int(oidp, &val, 0, req);
3234 	if (error != 0 || req->newptr == NULL)
3235 		return (error);
3236 	if (val != 0) {
3237 		stop_all_proc();
3238 		syncer_suspend();
3239 		while (ap_resume == 0)
3240 			;
3241 		syncer_resume();
3242 		resume_all_proc();
3243 	}
3244 	return (0);
3245 }
3246 
3247 SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW |
3248     CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0,
3249     sysctl_debug_stop_all_proc, "I",
3250     "");
3251 #endif
3252