xref: /freebsd/sys/kern/kern_proc.c (revision 52f72944b8f5abb2386eae924357dee8aea17d5b)
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_compat.h"
38 #include "opt_ddb.h"
39 #include "opt_ktrace.h"
40 #include "opt_kstack_pages.h"
41 #include "opt_stack.h"
42 
43 #include <sys/param.h>
44 #include <sys/systm.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 SDT_PROBE_DEFINE4(proc, , ctor, entry, "struct proc *", "int", "void *",
98     "int");
99 SDT_PROBE_DEFINE4(proc, , ctor, return, "struct proc *", "int", "void *",
100     "int");
101 SDT_PROBE_DEFINE4(proc, , dtor, entry, "struct proc *", "int", "void *",
102     "struct thread *");
103 SDT_PROBE_DEFINE3(proc, , dtor, return, "struct proc *", "int", "void *");
104 SDT_PROBE_DEFINE3(proc, , init, entry, "struct proc *", "int", "int");
105 SDT_PROBE_DEFINE3(proc, , init, return, "struct proc *", "int", "int");
106 
107 MALLOC_DEFINE(M_PGRP, "pgrp", "process group header");
108 MALLOC_DEFINE(M_SESSION, "session", "session header");
109 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
110 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
111 
112 static void doenterpgrp(struct proc *, struct pgrp *);
113 static void orphanpg(struct pgrp *pg);
114 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp);
115 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp);
116 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp,
117     int preferthread);
118 static void pgadjustjobc(struct pgrp *pgrp, int entering);
119 static void pgdelete(struct pgrp *);
120 static int proc_ctor(void *mem, int size, void *arg, int flags);
121 static void proc_dtor(void *mem, int size, void *arg);
122 static int proc_init(void *mem, int size, int flags);
123 static void proc_fini(void *mem, int size);
124 static void pargs_free(struct pargs *pa);
125 static struct proc *zpfind_locked(pid_t pid);
126 
127 /*
128  * Other process lists
129  */
130 struct pidhashhead *pidhashtbl;
131 u_long pidhash;
132 struct pgrphashhead *pgrphashtbl;
133 u_long pgrphash;
134 struct proclist allproc;
135 struct proclist zombproc;
136 struct sx __exclusive_cache_line allproc_lock;
137 struct sx __exclusive_cache_line proctree_lock;
138 struct mtx __exclusive_cache_line ppeers_lock;
139 uma_zone_t proc_zone;
140 
141 /*
142  * The offset of various fields in struct proc and struct thread.
143  * These are used by kernel debuggers to enumerate kernel threads and
144  * processes.
145  */
146 const int proc_off_p_pid = offsetof(struct proc, p_pid);
147 const int proc_off_p_comm = offsetof(struct proc, p_comm);
148 const int proc_off_p_list = offsetof(struct proc, p_list);
149 const int proc_off_p_threads = offsetof(struct proc, p_threads);
150 const int thread_off_td_tid = offsetof(struct thread, td_tid);
151 const int thread_off_td_name = offsetof(struct thread, td_name);
152 const int thread_off_td_oncpu = offsetof(struct thread, td_oncpu);
153 const int thread_off_td_pcb = offsetof(struct thread, td_pcb);
154 const int thread_off_td_plist = offsetof(struct thread, td_plist);
155 
156 EVENTHANDLER_LIST_DEFINE(process_ctor);
157 EVENTHANDLER_LIST_DEFINE(process_dtor);
158 EVENTHANDLER_LIST_DEFINE(process_init);
159 EVENTHANDLER_LIST_DEFINE(process_fini);
160 EVENTHANDLER_LIST_DEFINE(process_exit);
161 EVENTHANDLER_LIST_DEFINE(process_fork);
162 EVENTHANDLER_LIST_DEFINE(process_exec);
163 
164 EVENTHANDLER_LIST_DECLARE(thread_ctor);
165 EVENTHANDLER_LIST_DECLARE(thread_dtor);
166 
167 int kstack_pages = KSTACK_PAGES;
168 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0,
169     "Kernel stack size in pages");
170 static int vmmap_skip_res_cnt = 0;
171 SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW,
172     &vmmap_skip_res_cnt, 0,
173     "Skip calculation of the pages resident count in kern.proc.vmmap");
174 
175 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
176 #ifdef COMPAT_FREEBSD32
177 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
178 #endif
179 
180 /*
181  * Initialize global process hashing structures.
182  */
183 void
184 procinit(void)
185 {
186 
187 	sx_init(&allproc_lock, "allproc");
188 	sx_init(&proctree_lock, "proctree");
189 	mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
190 	LIST_INIT(&allproc);
191 	LIST_INIT(&zombproc);
192 	pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
193 	pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
194 	proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
195 	    proc_ctor, proc_dtor, proc_init, proc_fini,
196 	    UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
197 	uihashinit();
198 }
199 
200 /*
201  * Prepare a proc for use.
202  */
203 static int
204 proc_ctor(void *mem, int size, void *arg, int flags)
205 {
206 	struct proc *p;
207 	struct thread *td;
208 
209 	p = (struct proc *)mem;
210 	SDT_PROBE4(proc, , ctor , entry, p, size, arg, flags);
211 	EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
212 	SDT_PROBE4(proc, , ctor , return, p, size, arg, flags);
213 	td = FIRST_THREAD_IN_PROC(p);
214 	if (td != NULL) {
215 		/* Make sure all thread constructors are executed */
216 		EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
217 	}
218 	return (0);
219 }
220 
221 /*
222  * Reclaim a proc after use.
223  */
224 static void
225 proc_dtor(void *mem, int size, void *arg)
226 {
227 	struct proc *p;
228 	struct thread *td;
229 
230 	/* INVARIANTS checks go here */
231 	p = (struct proc *)mem;
232 	td = FIRST_THREAD_IN_PROC(p);
233 	SDT_PROBE4(proc, , dtor, entry, p, size, arg, td);
234 	if (td != NULL) {
235 #ifdef INVARIANTS
236 		KASSERT((p->p_numthreads == 1),
237 		    ("bad number of threads in exiting process"));
238 		KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
239 #endif
240 		/* Free all OSD associated to this thread. */
241 		osd_thread_exit(td);
242 		td_softdep_cleanup(td);
243 		MPASS(td->td_su == NULL);
244 
245 		/* Make sure all thread destructors are executed */
246 		EVENTHANDLER_DIRECT_INVOKE(thread_dtor, td);
247 	}
248 	EVENTHANDLER_DIRECT_INVOKE(process_dtor, p);
249 	if (p->p_ksi != NULL)
250 		KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
251 	SDT_PROBE3(proc, , dtor, return, p, size, arg);
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 	SDT_PROBE3(proc, , init, entry, p, size, flags);
264 	mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK | MTX_NEW);
265 	mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_NEW);
266 	mtx_init(&p->p_statmtx, "pstatl", NULL, MTX_SPIN | MTX_NEW);
267 	mtx_init(&p->p_itimmtx, "pitiml", NULL, MTX_SPIN | MTX_NEW);
268 	mtx_init(&p->p_profmtx, "pprofl", NULL, MTX_SPIN | MTX_NEW);
269 	cv_init(&p->p_pwait, "ppwait");
270 	cv_init(&p->p_dbgwait, "dbgwait");
271 	TAILQ_INIT(&p->p_threads);	     /* all threads in proc */
272 	EVENTHANDLER_DIRECT_INVOKE(process_init, p);
273 	p->p_stats = pstats_alloc();
274 	p->p_pgrp = NULL;
275 	SDT_PROBE3(proc, , init, return, p, size, flags);
276 	return (0);
277 }
278 
279 /*
280  * UMA should ensure that this function is never called.
281  * Freeing a proc structure would violate type stability.
282  */
283 static void
284 proc_fini(void *mem, int size)
285 {
286 #ifdef notnow
287 	struct proc *p;
288 
289 	p = (struct proc *)mem;
290 	EVENTHANDLER_DIRECT_INVOKE(process_fini, p);
291 	pstats_free(p->p_stats);
292 	thread_free(FIRST_THREAD_IN_PROC(p));
293 	mtx_destroy(&p->p_mtx);
294 	if (p->p_ksi != NULL)
295 		ksiginfo_free(p->p_ksi);
296 #else
297 	panic("proc reclaimed");
298 #endif
299 }
300 
301 /*
302  * Is p an inferior of the current process?
303  */
304 int
305 inferior(struct proc *p)
306 {
307 
308 	sx_assert(&proctree_lock, SX_LOCKED);
309 	PROC_LOCK_ASSERT(p, MA_OWNED);
310 	for (; p != curproc; p = proc_realparent(p)) {
311 		if (p->p_pid == 0)
312 			return (0);
313 	}
314 	return (1);
315 }
316 
317 struct proc *
318 pfind_locked(pid_t pid)
319 {
320 	struct proc *p;
321 
322 	sx_assert(&allproc_lock, SX_LOCKED);
323 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
324 		if (p->p_pid == pid) {
325 			PROC_LOCK(p);
326 			if (p->p_state == PRS_NEW) {
327 				PROC_UNLOCK(p);
328 				p = NULL;
329 			}
330 			break;
331 		}
332 	}
333 	return (p);
334 }
335 
336 /*
337  * Locate a process by number; return only "live" processes -- i.e., neither
338  * zombies nor newly born but incompletely initialized processes.  By not
339  * returning processes in the PRS_NEW state, we allow callers to avoid
340  * testing for that condition to avoid dereferencing p_ucred, et al.
341  */
342 struct proc *
343 pfind(pid_t pid)
344 {
345 	struct proc *p;
346 
347 	p = curproc;
348 	if (p->p_pid == pid) {
349 		PROC_LOCK(p);
350 		return (p);
351 	}
352 	sx_slock(&allproc_lock);
353 	p = pfind_locked(pid);
354 	sx_sunlock(&allproc_lock);
355 	return (p);
356 }
357 
358 /*
359  * Same as pfind but allow zombies.
360  */
361 struct proc *
362 pfind_any(pid_t pid)
363 {
364 	struct proc *p;
365 
366 	sx_slock(&allproc_lock);
367 	p = pfind_locked(pid);
368 	if (p == NULL)
369 		p = zpfind_locked(pid);
370 	sx_sunlock(&allproc_lock);
371 
372 	return (p);
373 }
374 
375 static struct proc *
376 pfind_tid_locked(pid_t tid)
377 {
378 	struct proc *p;
379 	struct thread *td;
380 
381 	sx_assert(&allproc_lock, SX_LOCKED);
382 	FOREACH_PROC_IN_SYSTEM(p) {
383 		PROC_LOCK(p);
384 		if (p->p_state == PRS_NEW) {
385 			PROC_UNLOCK(p);
386 			continue;
387 		}
388 		FOREACH_THREAD_IN_PROC(p, td) {
389 			if (td->td_tid == tid)
390 				goto found;
391 		}
392 		PROC_UNLOCK(p);
393 	}
394 found:
395 	return (p);
396 }
397 
398 /*
399  * Locate a process group by number.
400  * The caller must hold proctree_lock.
401  */
402 struct pgrp *
403 pgfind(pid_t pgid)
404 {
405 	struct pgrp *pgrp;
406 
407 	sx_assert(&proctree_lock, SX_LOCKED);
408 
409 	LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
410 		if (pgrp->pg_id == pgid) {
411 			PGRP_LOCK(pgrp);
412 			return (pgrp);
413 		}
414 	}
415 	return (NULL);
416 }
417 
418 /*
419  * Locate process and do additional manipulations, depending on flags.
420  */
421 int
422 pget(pid_t pid, int flags, struct proc **pp)
423 {
424 	struct proc *p;
425 	int error;
426 
427 	p = curproc;
428 	if (p->p_pid == pid) {
429 		PROC_LOCK(p);
430 	} else {
431 		sx_slock(&allproc_lock);
432 		if (pid <= PID_MAX) {
433 			p = pfind_locked(pid);
434 			if (p == NULL && (flags & PGET_NOTWEXIT) == 0)
435 				p = zpfind_locked(pid);
436 		} else if ((flags & PGET_NOTID) == 0) {
437 			p = pfind_tid_locked(pid);
438 		} else {
439 			p = NULL;
440 		}
441 		sx_sunlock(&allproc_lock);
442 		if (p == NULL)
443 			return (ESRCH);
444 		if ((flags & PGET_CANSEE) != 0) {
445 			error = p_cansee(curthread, p);
446 			if (error != 0)
447 				goto errout;
448 		}
449 	}
450 	if ((flags & PGET_CANDEBUG) != 0) {
451 		error = p_candebug(curthread, p);
452 		if (error != 0)
453 			goto errout;
454 	}
455 	if ((flags & PGET_ISCURRENT) != 0 && curproc != p) {
456 		error = EPERM;
457 		goto errout;
458 	}
459 	if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) {
460 		error = ESRCH;
461 		goto errout;
462 	}
463 	if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) {
464 		/*
465 		 * XXXRW: Not clear ESRCH is the right error during proc
466 		 * execve().
467 		 */
468 		error = ESRCH;
469 		goto errout;
470 	}
471 	if ((flags & PGET_HOLD) != 0) {
472 		_PHOLD(p);
473 		PROC_UNLOCK(p);
474 	}
475 	*pp = p;
476 	return (0);
477 errout:
478 	PROC_UNLOCK(p);
479 	return (error);
480 }
481 
482 /*
483  * Create a new process group.
484  * pgid must be equal to the pid of p.
485  * Begin a new session if required.
486  */
487 int
488 enterpgrp(struct proc *p, pid_t pgid, struct pgrp *pgrp, struct session *sess)
489 {
490 
491 	sx_assert(&proctree_lock, SX_XLOCKED);
492 
493 	KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
494 	KASSERT(p->p_pid == pgid,
495 	    ("enterpgrp: new pgrp and pid != pgid"));
496 	KASSERT(pgfind(pgid) == NULL,
497 	    ("enterpgrp: pgrp with pgid exists"));
498 	KASSERT(!SESS_LEADER(p),
499 	    ("enterpgrp: session leader attempted setpgrp"));
500 
501 	mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
502 
503 	if (sess != NULL) {
504 		/*
505 		 * new session
506 		 */
507 		mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
508 		PROC_LOCK(p);
509 		p->p_flag &= ~P_CONTROLT;
510 		PROC_UNLOCK(p);
511 		PGRP_LOCK(pgrp);
512 		sess->s_leader = p;
513 		sess->s_sid = p->p_pid;
514 		refcount_init(&sess->s_count, 1);
515 		sess->s_ttyvp = NULL;
516 		sess->s_ttydp = NULL;
517 		sess->s_ttyp = NULL;
518 		bcopy(p->p_session->s_login, sess->s_login,
519 			    sizeof(sess->s_login));
520 		pgrp->pg_session = sess;
521 		KASSERT(p == curproc,
522 		    ("enterpgrp: mksession and p != curproc"));
523 	} else {
524 		pgrp->pg_session = p->p_session;
525 		sess_hold(pgrp->pg_session);
526 		PGRP_LOCK(pgrp);
527 	}
528 	pgrp->pg_id = pgid;
529 	LIST_INIT(&pgrp->pg_members);
530 
531 	/*
532 	 * As we have an exclusive lock of proctree_lock,
533 	 * this should not deadlock.
534 	 */
535 	LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
536 	pgrp->pg_jobc = 0;
537 	SLIST_INIT(&pgrp->pg_sigiolst);
538 	PGRP_UNLOCK(pgrp);
539 
540 	doenterpgrp(p, pgrp);
541 
542 	return (0);
543 }
544 
545 /*
546  * Move p to an existing process group
547  */
548 int
549 enterthispgrp(struct proc *p, struct pgrp *pgrp)
550 {
551 
552 	sx_assert(&proctree_lock, SX_XLOCKED);
553 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
554 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
555 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
556 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
557 	KASSERT(pgrp->pg_session == p->p_session,
558 		("%s: pgrp's session %p, p->p_session %p.\n",
559 		__func__,
560 		pgrp->pg_session,
561 		p->p_session));
562 	KASSERT(pgrp != p->p_pgrp,
563 		("%s: p belongs to pgrp.", __func__));
564 
565 	doenterpgrp(p, pgrp);
566 
567 	return (0);
568 }
569 
570 /*
571  * Move p to a process group
572  */
573 static void
574 doenterpgrp(struct proc *p, struct pgrp *pgrp)
575 {
576 	struct pgrp *savepgrp;
577 
578 	sx_assert(&proctree_lock, SX_XLOCKED);
579 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
580 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
581 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
582 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
583 
584 	savepgrp = p->p_pgrp;
585 
586 	/*
587 	 * Adjust eligibility of affected pgrps to participate in job control.
588 	 * Increment eligibility counts before decrementing, otherwise we
589 	 * could reach 0 spuriously during the first call.
590 	 */
591 	fixjobc(p, pgrp, 1);
592 	fixjobc(p, p->p_pgrp, 0);
593 
594 	PGRP_LOCK(pgrp);
595 	PGRP_LOCK(savepgrp);
596 	PROC_LOCK(p);
597 	LIST_REMOVE(p, p_pglist);
598 	p->p_pgrp = pgrp;
599 	PROC_UNLOCK(p);
600 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
601 	PGRP_UNLOCK(savepgrp);
602 	PGRP_UNLOCK(pgrp);
603 	if (LIST_EMPTY(&savepgrp->pg_members))
604 		pgdelete(savepgrp);
605 }
606 
607 /*
608  * remove process from process group
609  */
610 int
611 leavepgrp(struct proc *p)
612 {
613 	struct pgrp *savepgrp;
614 
615 	sx_assert(&proctree_lock, SX_XLOCKED);
616 	savepgrp = p->p_pgrp;
617 	PGRP_LOCK(savepgrp);
618 	PROC_LOCK(p);
619 	LIST_REMOVE(p, p_pglist);
620 	p->p_pgrp = NULL;
621 	PROC_UNLOCK(p);
622 	PGRP_UNLOCK(savepgrp);
623 	if (LIST_EMPTY(&savepgrp->pg_members))
624 		pgdelete(savepgrp);
625 	return (0);
626 }
627 
628 /*
629  * delete a process group
630  */
631 static void
632 pgdelete(struct pgrp *pgrp)
633 {
634 	struct session *savesess;
635 	struct tty *tp;
636 
637 	sx_assert(&proctree_lock, SX_XLOCKED);
638 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
639 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
640 
641 	/*
642 	 * Reset any sigio structures pointing to us as a result of
643 	 * F_SETOWN with our pgid.
644 	 */
645 	funsetownlst(&pgrp->pg_sigiolst);
646 
647 	PGRP_LOCK(pgrp);
648 	tp = pgrp->pg_session->s_ttyp;
649 	LIST_REMOVE(pgrp, pg_hash);
650 	savesess = pgrp->pg_session;
651 	PGRP_UNLOCK(pgrp);
652 
653 	/* Remove the reference to the pgrp before deallocating it. */
654 	if (tp != NULL) {
655 		tty_lock(tp);
656 		tty_rel_pgrp(tp, pgrp);
657 	}
658 
659 	mtx_destroy(&pgrp->pg_mtx);
660 	free(pgrp, M_PGRP);
661 	sess_release(savesess);
662 }
663 
664 static void
665 pgadjustjobc(struct pgrp *pgrp, int entering)
666 {
667 
668 	PGRP_LOCK(pgrp);
669 	if (entering)
670 		pgrp->pg_jobc++;
671 	else {
672 		--pgrp->pg_jobc;
673 		if (pgrp->pg_jobc == 0)
674 			orphanpg(pgrp);
675 	}
676 	PGRP_UNLOCK(pgrp);
677 }
678 
679 /*
680  * Adjust pgrp jobc counters when specified process changes process group.
681  * We count the number of processes in each process group that "qualify"
682  * the group for terminal job control (those with a parent in a different
683  * process group of the same session).  If that count reaches zero, the
684  * process group becomes orphaned.  Check both the specified process'
685  * process group and that of its children.
686  * entering == 0 => p is leaving specified group.
687  * entering == 1 => p is entering specified group.
688  */
689 void
690 fixjobc(struct proc *p, struct pgrp *pgrp, int entering)
691 {
692 	struct pgrp *hispgrp;
693 	struct session *mysession;
694 	struct proc *q;
695 
696 	sx_assert(&proctree_lock, SX_LOCKED);
697 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
698 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
699 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
700 
701 	/*
702 	 * Check p's parent to see whether p qualifies its own process
703 	 * group; if so, adjust count for p's process group.
704 	 */
705 	mysession = pgrp->pg_session;
706 	if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
707 	    hispgrp->pg_session == mysession)
708 		pgadjustjobc(pgrp, entering);
709 
710 	/*
711 	 * Check this process' children to see whether they qualify
712 	 * their process groups; if so, adjust counts for children's
713 	 * process groups.
714 	 */
715 	LIST_FOREACH(q, &p->p_children, p_sibling) {
716 		hispgrp = q->p_pgrp;
717 		if (hispgrp == pgrp ||
718 		    hispgrp->pg_session != mysession)
719 			continue;
720 		if (q->p_state == PRS_ZOMBIE)
721 			continue;
722 		pgadjustjobc(hispgrp, entering);
723 	}
724 }
725 
726 void
727 killjobc(void)
728 {
729 	struct session *sp;
730 	struct tty *tp;
731 	struct proc *p;
732 	struct vnode *ttyvp;
733 
734 	p = curproc;
735 	MPASS(p->p_flag & P_WEXIT);
736 	/*
737 	 * Do a quick check to see if there is anything to do with the
738 	 * proctree_lock held. pgrp and LIST_EMPTY checks are for fixjobc().
739 	 */
740 	PROC_LOCK(p);
741 	if (!SESS_LEADER(p) &&
742 	    (p->p_pgrp == p->p_pptr->p_pgrp) &&
743 	    LIST_EMPTY(&p->p_children)) {
744 		PROC_UNLOCK(p);
745 		return;
746 	}
747 	PROC_UNLOCK(p);
748 
749 	sx_xlock(&proctree_lock);
750 	if (SESS_LEADER(p)) {
751 		sp = p->p_session;
752 
753 		/*
754 		 * s_ttyp is not zero'd; we use this to indicate that
755 		 * the session once had a controlling terminal. (for
756 		 * logging and informational purposes)
757 		 */
758 		SESS_LOCK(sp);
759 		ttyvp = sp->s_ttyvp;
760 		tp = sp->s_ttyp;
761 		sp->s_ttyvp = NULL;
762 		sp->s_ttydp = NULL;
763 		sp->s_leader = NULL;
764 		SESS_UNLOCK(sp);
765 
766 		/*
767 		 * Signal foreground pgrp and revoke access to
768 		 * controlling terminal if it has not been revoked
769 		 * already.
770 		 *
771 		 * Because the TTY may have been revoked in the mean
772 		 * time and could already have a new session associated
773 		 * with it, make sure we don't send a SIGHUP to a
774 		 * foreground process group that does not belong to this
775 		 * session.
776 		 */
777 
778 		if (tp != NULL) {
779 			tty_lock(tp);
780 			if (tp->t_session == sp)
781 				tty_signal_pgrp(tp, SIGHUP);
782 			tty_unlock(tp);
783 		}
784 
785 		if (ttyvp != NULL) {
786 			sx_xunlock(&proctree_lock);
787 			if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) {
788 				VOP_REVOKE(ttyvp, REVOKEALL);
789 				VOP_UNLOCK(ttyvp, 0);
790 			}
791 			vrele(ttyvp);
792 			sx_xlock(&proctree_lock);
793 		}
794 	}
795 	fixjobc(p, p->p_pgrp, 0);
796 	sx_xunlock(&proctree_lock);
797 }
798 
799 /*
800  * A process group has become orphaned;
801  * if there are any stopped processes in the group,
802  * hang-up all process in that group.
803  */
804 static void
805 orphanpg(struct pgrp *pg)
806 {
807 	struct proc *p;
808 
809 	PGRP_LOCK_ASSERT(pg, MA_OWNED);
810 
811 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
812 		PROC_LOCK(p);
813 		if (P_SHOULDSTOP(p) == P_STOPPED_SIG) {
814 			PROC_UNLOCK(p);
815 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
816 				PROC_LOCK(p);
817 				kern_psignal(p, SIGHUP);
818 				kern_psignal(p, SIGCONT);
819 				PROC_UNLOCK(p);
820 			}
821 			return;
822 		}
823 		PROC_UNLOCK(p);
824 	}
825 }
826 
827 void
828 sess_hold(struct session *s)
829 {
830 
831 	refcount_acquire(&s->s_count);
832 }
833 
834 void
835 sess_release(struct session *s)
836 {
837 
838 	if (refcount_release(&s->s_count)) {
839 		if (s->s_ttyp != NULL) {
840 			tty_lock(s->s_ttyp);
841 			tty_rel_sess(s->s_ttyp, s);
842 		}
843 		mtx_destroy(&s->s_mtx);
844 		free(s, M_SESSION);
845 	}
846 }
847 
848 #ifdef DDB
849 
850 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
851 {
852 	struct pgrp *pgrp;
853 	struct proc *p;
854 	int i;
855 
856 	for (i = 0; i <= pgrphash; i++) {
857 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
858 			printf("\tindx %d\n", i);
859 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
860 				printf(
861 			"\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
862 				    (void *)pgrp, (long)pgrp->pg_id,
863 				    (void *)pgrp->pg_session,
864 				    pgrp->pg_session->s_count,
865 				    (void *)LIST_FIRST(&pgrp->pg_members));
866 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
867 					printf("\t\tpid %ld addr %p pgrp %p\n",
868 					    (long)p->p_pid, (void *)p,
869 					    (void *)p->p_pgrp);
870 				}
871 			}
872 		}
873 	}
874 }
875 #endif /* DDB */
876 
877 /*
878  * Calculate the kinfo_proc members which contain process-wide
879  * informations.
880  * Must be called with the target process locked.
881  */
882 static void
883 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
884 {
885 	struct thread *td;
886 
887 	PROC_LOCK_ASSERT(p, MA_OWNED);
888 
889 	kp->ki_estcpu = 0;
890 	kp->ki_pctcpu = 0;
891 	FOREACH_THREAD_IN_PROC(p, td) {
892 		thread_lock(td);
893 		kp->ki_pctcpu += sched_pctcpu(td);
894 		kp->ki_estcpu += sched_estcpu(td);
895 		thread_unlock(td);
896 	}
897 }
898 
899 /*
900  * Clear kinfo_proc and fill in any information that is common
901  * to all threads in the process.
902  * Must be called with the target process locked.
903  */
904 static void
905 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
906 {
907 	struct thread *td0;
908 	struct tty *tp;
909 	struct session *sp;
910 	struct ucred *cred;
911 	struct sigacts *ps;
912 	struct timeval boottime;
913 
914 	/* For proc_realparent. */
915 	sx_assert(&proctree_lock, SX_LOCKED);
916 	PROC_LOCK_ASSERT(p, MA_OWNED);
917 	bzero(kp, sizeof(*kp));
918 
919 	kp->ki_structsize = sizeof(*kp);
920 	kp->ki_paddr = p;
921 	kp->ki_addr =/* p->p_addr; */0; /* XXX */
922 	kp->ki_args = p->p_args;
923 	kp->ki_textvp = p->p_textvp;
924 #ifdef KTRACE
925 	kp->ki_tracep = p->p_tracevp;
926 	kp->ki_traceflag = p->p_traceflag;
927 #endif
928 	kp->ki_fd = p->p_fd;
929 	kp->ki_vmspace = p->p_vmspace;
930 	kp->ki_flag = p->p_flag;
931 	kp->ki_flag2 = p->p_flag2;
932 	cred = p->p_ucred;
933 	if (cred) {
934 		kp->ki_uid = cred->cr_uid;
935 		kp->ki_ruid = cred->cr_ruid;
936 		kp->ki_svuid = cred->cr_svuid;
937 		kp->ki_cr_flags = 0;
938 		if (cred->cr_flags & CRED_FLAG_CAPMODE)
939 			kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
940 		/* XXX bde doesn't like KI_NGROUPS */
941 		if (cred->cr_ngroups > KI_NGROUPS) {
942 			kp->ki_ngroups = KI_NGROUPS;
943 			kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
944 		} else
945 			kp->ki_ngroups = cred->cr_ngroups;
946 		bcopy(cred->cr_groups, kp->ki_groups,
947 		    kp->ki_ngroups * sizeof(gid_t));
948 		kp->ki_rgid = cred->cr_rgid;
949 		kp->ki_svgid = cred->cr_svgid;
950 		/* If jailed(cred), emulate the old P_JAILED flag. */
951 		if (jailed(cred)) {
952 			kp->ki_flag |= P_JAILED;
953 			/* If inside the jail, use 0 as a jail ID. */
954 			if (cred->cr_prison != curthread->td_ucred->cr_prison)
955 				kp->ki_jid = cred->cr_prison->pr_id;
956 		}
957 		strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
958 		    sizeof(kp->ki_loginclass));
959 	}
960 	ps = p->p_sigacts;
961 	if (ps) {
962 		mtx_lock(&ps->ps_mtx);
963 		kp->ki_sigignore = ps->ps_sigignore;
964 		kp->ki_sigcatch = ps->ps_sigcatch;
965 		mtx_unlock(&ps->ps_mtx);
966 	}
967 	if (p->p_state != PRS_NEW &&
968 	    p->p_state != PRS_ZOMBIE &&
969 	    p->p_vmspace != NULL) {
970 		struct vmspace *vm = p->p_vmspace;
971 
972 		kp->ki_size = vm->vm_map.size;
973 		kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
974 		FOREACH_THREAD_IN_PROC(p, td0) {
975 			if (!TD_IS_SWAPPED(td0))
976 				kp->ki_rssize += td0->td_kstack_pages;
977 		}
978 		kp->ki_swrss = vm->vm_swrss;
979 		kp->ki_tsize = vm->vm_tsize;
980 		kp->ki_dsize = vm->vm_dsize;
981 		kp->ki_ssize = vm->vm_ssize;
982 	} else if (p->p_state == PRS_ZOMBIE)
983 		kp->ki_stat = SZOMB;
984 	if (kp->ki_flag & P_INMEM)
985 		kp->ki_sflag = PS_INMEM;
986 	else
987 		kp->ki_sflag = 0;
988 	/* Calculate legacy swtime as seconds since 'swtick'. */
989 	kp->ki_swtime = (ticks - p->p_swtick) / hz;
990 	kp->ki_pid = p->p_pid;
991 	kp->ki_nice = p->p_nice;
992 	kp->ki_fibnum = p->p_fibnum;
993 	kp->ki_start = p->p_stats->p_start;
994 	getboottime(&boottime);
995 	timevaladd(&kp->ki_start, &boottime);
996 	PROC_STATLOCK(p);
997 	rufetch(p, &kp->ki_rusage);
998 	kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
999 	calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
1000 	PROC_STATUNLOCK(p);
1001 	calccru(p, &kp->ki_childutime, &kp->ki_childstime);
1002 	/* Some callers want child times in a single value. */
1003 	kp->ki_childtime = kp->ki_childstime;
1004 	timevaladd(&kp->ki_childtime, &kp->ki_childutime);
1005 
1006 	FOREACH_THREAD_IN_PROC(p, td0)
1007 		kp->ki_cow += td0->td_cow;
1008 
1009 	tp = NULL;
1010 	if (p->p_pgrp) {
1011 		kp->ki_pgid = p->p_pgrp->pg_id;
1012 		kp->ki_jobc = p->p_pgrp->pg_jobc;
1013 		sp = p->p_pgrp->pg_session;
1014 
1015 		if (sp != NULL) {
1016 			kp->ki_sid = sp->s_sid;
1017 			SESS_LOCK(sp);
1018 			strlcpy(kp->ki_login, sp->s_login,
1019 			    sizeof(kp->ki_login));
1020 			if (sp->s_ttyvp)
1021 				kp->ki_kiflag |= KI_CTTY;
1022 			if (SESS_LEADER(p))
1023 				kp->ki_kiflag |= KI_SLEADER;
1024 			/* XXX proctree_lock */
1025 			tp = sp->s_ttyp;
1026 			SESS_UNLOCK(sp);
1027 		}
1028 	}
1029 	if ((p->p_flag & P_CONTROLT) && tp != NULL) {
1030 		kp->ki_tdev = tty_udev(tp);
1031 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1032 		kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
1033 		if (tp->t_session)
1034 			kp->ki_tsid = tp->t_session->s_sid;
1035 	} else {
1036 		kp->ki_tdev = NODEV;
1037 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1038 	}
1039 	if (p->p_comm[0] != '\0')
1040 		strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
1041 	if (p->p_sysent && p->p_sysent->sv_name != NULL &&
1042 	    p->p_sysent->sv_name[0] != '\0')
1043 		strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
1044 	kp->ki_siglist = p->p_siglist;
1045 	kp->ki_xstat = KW_EXITCODE(p->p_xexit, p->p_xsig);
1046 	kp->ki_acflag = p->p_acflag;
1047 	kp->ki_lock = p->p_lock;
1048 	if (p->p_pptr) {
1049 		kp->ki_ppid = proc_realparent(p)->p_pid;
1050 		if (p->p_flag & P_TRACED)
1051 			kp->ki_tracer = p->p_pptr->p_pid;
1052 	}
1053 }
1054 
1055 /*
1056  * Fill in information that is thread specific.  Must be called with
1057  * target process locked.  If 'preferthread' is set, overwrite certain
1058  * process-related fields that are maintained for both threads and
1059  * processes.
1060  */
1061 static void
1062 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
1063 {
1064 	struct proc *p;
1065 
1066 	p = td->td_proc;
1067 	kp->ki_tdaddr = td;
1068 	PROC_LOCK_ASSERT(p, MA_OWNED);
1069 
1070 	if (preferthread)
1071 		PROC_STATLOCK(p);
1072 	thread_lock(td);
1073 	if (td->td_wmesg != NULL)
1074 		strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
1075 	else
1076 		bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
1077 	if (strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname)) >=
1078 	    sizeof(kp->ki_tdname)) {
1079 		strlcpy(kp->ki_moretdname,
1080 		    td->td_name + sizeof(kp->ki_tdname) - 1,
1081 		    sizeof(kp->ki_moretdname));
1082 	} else {
1083 		bzero(kp->ki_moretdname, sizeof(kp->ki_moretdname));
1084 	}
1085 	if (TD_ON_LOCK(td)) {
1086 		kp->ki_kiflag |= KI_LOCKBLOCK;
1087 		strlcpy(kp->ki_lockname, td->td_lockname,
1088 		    sizeof(kp->ki_lockname));
1089 	} else {
1090 		kp->ki_kiflag &= ~KI_LOCKBLOCK;
1091 		bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
1092 	}
1093 
1094 	if (p->p_state == PRS_NORMAL) { /* approximate. */
1095 		if (TD_ON_RUNQ(td) ||
1096 		    TD_CAN_RUN(td) ||
1097 		    TD_IS_RUNNING(td)) {
1098 			kp->ki_stat = SRUN;
1099 		} else if (P_SHOULDSTOP(p)) {
1100 			kp->ki_stat = SSTOP;
1101 		} else if (TD_IS_SLEEPING(td)) {
1102 			kp->ki_stat = SSLEEP;
1103 		} else if (TD_ON_LOCK(td)) {
1104 			kp->ki_stat = SLOCK;
1105 		} else {
1106 			kp->ki_stat = SWAIT;
1107 		}
1108 	} else if (p->p_state == PRS_ZOMBIE) {
1109 		kp->ki_stat = SZOMB;
1110 	} else {
1111 		kp->ki_stat = SIDL;
1112 	}
1113 
1114 	/* Things in the thread */
1115 	kp->ki_wchan = td->td_wchan;
1116 	kp->ki_pri.pri_level = td->td_priority;
1117 	kp->ki_pri.pri_native = td->td_base_pri;
1118 
1119 	/*
1120 	 * Note: legacy fields; clamp at the old NOCPU value and/or
1121 	 * the maximum u_char CPU value.
1122 	 */
1123 	if (td->td_lastcpu == NOCPU)
1124 		kp->ki_lastcpu_old = NOCPU_OLD;
1125 	else if (td->td_lastcpu > MAXCPU_OLD)
1126 		kp->ki_lastcpu_old = MAXCPU_OLD;
1127 	else
1128 		kp->ki_lastcpu_old = td->td_lastcpu;
1129 
1130 	if (td->td_oncpu == NOCPU)
1131 		kp->ki_oncpu_old = NOCPU_OLD;
1132 	else if (td->td_oncpu > MAXCPU_OLD)
1133 		kp->ki_oncpu_old = MAXCPU_OLD;
1134 	else
1135 		kp->ki_oncpu_old = td->td_oncpu;
1136 
1137 	kp->ki_lastcpu = td->td_lastcpu;
1138 	kp->ki_oncpu = td->td_oncpu;
1139 	kp->ki_tdflags = td->td_flags;
1140 	kp->ki_tid = td->td_tid;
1141 	kp->ki_numthreads = p->p_numthreads;
1142 	kp->ki_pcb = td->td_pcb;
1143 	kp->ki_kstack = (void *)td->td_kstack;
1144 	kp->ki_slptime = (ticks - td->td_slptick) / hz;
1145 	kp->ki_pri.pri_class = td->td_pri_class;
1146 	kp->ki_pri.pri_user = td->td_user_pri;
1147 
1148 	if (preferthread) {
1149 		rufetchtd(td, &kp->ki_rusage);
1150 		kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
1151 		kp->ki_pctcpu = sched_pctcpu(td);
1152 		kp->ki_estcpu = sched_estcpu(td);
1153 		kp->ki_cow = td->td_cow;
1154 	}
1155 
1156 	/* We can't get this anymore but ps etc never used it anyway. */
1157 	kp->ki_rqindex = 0;
1158 
1159 	if (preferthread)
1160 		kp->ki_siglist = td->td_siglist;
1161 	kp->ki_sigmask = td->td_sigmask;
1162 	thread_unlock(td);
1163 	if (preferthread)
1164 		PROC_STATUNLOCK(p);
1165 }
1166 
1167 /*
1168  * Fill in a kinfo_proc structure for the specified process.
1169  * Must be called with the target process locked.
1170  */
1171 void
1172 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1173 {
1174 
1175 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1176 
1177 	fill_kinfo_proc_only(p, kp);
1178 	fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1179 	fill_kinfo_aggregate(p, kp);
1180 }
1181 
1182 struct pstats *
1183 pstats_alloc(void)
1184 {
1185 
1186 	return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1187 }
1188 
1189 /*
1190  * Copy parts of p_stats; zero the rest of p_stats (statistics).
1191  */
1192 void
1193 pstats_fork(struct pstats *src, struct pstats *dst)
1194 {
1195 
1196 	bzero(&dst->pstat_startzero,
1197 	    __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1198 	bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1199 	    __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1200 }
1201 
1202 void
1203 pstats_free(struct pstats *ps)
1204 {
1205 
1206 	free(ps, M_SUBPROC);
1207 }
1208 
1209 static struct proc *
1210 zpfind_locked(pid_t pid)
1211 {
1212 	struct proc *p;
1213 
1214 	sx_assert(&allproc_lock, SX_LOCKED);
1215 	LIST_FOREACH(p, &zombproc, p_list) {
1216 		if (p->p_pid == pid) {
1217 			PROC_LOCK(p);
1218 			break;
1219 		}
1220 	}
1221 	return (p);
1222 }
1223 
1224 /*
1225  * Locate a zombie process by number
1226  */
1227 struct proc *
1228 zpfind(pid_t pid)
1229 {
1230 	struct proc *p;
1231 
1232 	sx_slock(&allproc_lock);
1233 	p = zpfind_locked(pid);
1234 	sx_sunlock(&allproc_lock);
1235 	return (p);
1236 }
1237 
1238 #ifdef COMPAT_FREEBSD32
1239 
1240 /*
1241  * This function is typically used to copy out the kernel address, so
1242  * it can be replaced by assignment of zero.
1243  */
1244 static inline uint32_t
1245 ptr32_trim(void *ptr)
1246 {
1247 	uintptr_t uptr;
1248 
1249 	uptr = (uintptr_t)ptr;
1250 	return ((uptr > UINT_MAX) ? 0 : uptr);
1251 }
1252 
1253 #define PTRTRIM_CP(src,dst,fld) \
1254 	do { (dst).fld = ptr32_trim((src).fld); } while (0)
1255 
1256 static void
1257 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1258 {
1259 	int i;
1260 
1261 	bzero(ki32, sizeof(struct kinfo_proc32));
1262 	ki32->ki_structsize = sizeof(struct kinfo_proc32);
1263 	CP(*ki, *ki32, ki_layout);
1264 	PTRTRIM_CP(*ki, *ki32, ki_args);
1265 	PTRTRIM_CP(*ki, *ki32, ki_paddr);
1266 	PTRTRIM_CP(*ki, *ki32, ki_addr);
1267 	PTRTRIM_CP(*ki, *ki32, ki_tracep);
1268 	PTRTRIM_CP(*ki, *ki32, ki_textvp);
1269 	PTRTRIM_CP(*ki, *ki32, ki_fd);
1270 	PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1271 	PTRTRIM_CP(*ki, *ki32, ki_wchan);
1272 	CP(*ki, *ki32, ki_pid);
1273 	CP(*ki, *ki32, ki_ppid);
1274 	CP(*ki, *ki32, ki_pgid);
1275 	CP(*ki, *ki32, ki_tpgid);
1276 	CP(*ki, *ki32, ki_sid);
1277 	CP(*ki, *ki32, ki_tsid);
1278 	CP(*ki, *ki32, ki_jobc);
1279 	CP(*ki, *ki32, ki_tdev);
1280 	CP(*ki, *ki32, ki_tdev_freebsd11);
1281 	CP(*ki, *ki32, ki_siglist);
1282 	CP(*ki, *ki32, ki_sigmask);
1283 	CP(*ki, *ki32, ki_sigignore);
1284 	CP(*ki, *ki32, ki_sigcatch);
1285 	CP(*ki, *ki32, ki_uid);
1286 	CP(*ki, *ki32, ki_ruid);
1287 	CP(*ki, *ki32, ki_svuid);
1288 	CP(*ki, *ki32, ki_rgid);
1289 	CP(*ki, *ki32, ki_svgid);
1290 	CP(*ki, *ki32, ki_ngroups);
1291 	for (i = 0; i < KI_NGROUPS; i++)
1292 		CP(*ki, *ki32, ki_groups[i]);
1293 	CP(*ki, *ki32, ki_size);
1294 	CP(*ki, *ki32, ki_rssize);
1295 	CP(*ki, *ki32, ki_swrss);
1296 	CP(*ki, *ki32, ki_tsize);
1297 	CP(*ki, *ki32, ki_dsize);
1298 	CP(*ki, *ki32, ki_ssize);
1299 	CP(*ki, *ki32, ki_xstat);
1300 	CP(*ki, *ki32, ki_acflag);
1301 	CP(*ki, *ki32, ki_pctcpu);
1302 	CP(*ki, *ki32, ki_estcpu);
1303 	CP(*ki, *ki32, ki_slptime);
1304 	CP(*ki, *ki32, ki_swtime);
1305 	CP(*ki, *ki32, ki_cow);
1306 	CP(*ki, *ki32, ki_runtime);
1307 	TV_CP(*ki, *ki32, ki_start);
1308 	TV_CP(*ki, *ki32, ki_childtime);
1309 	CP(*ki, *ki32, ki_flag);
1310 	CP(*ki, *ki32, ki_kiflag);
1311 	CP(*ki, *ki32, ki_traceflag);
1312 	CP(*ki, *ki32, ki_stat);
1313 	CP(*ki, *ki32, ki_nice);
1314 	CP(*ki, *ki32, ki_lock);
1315 	CP(*ki, *ki32, ki_rqindex);
1316 	CP(*ki, *ki32, ki_oncpu);
1317 	CP(*ki, *ki32, ki_lastcpu);
1318 
1319 	/* XXX TODO: wrap cpu value as appropriate */
1320 	CP(*ki, *ki32, ki_oncpu_old);
1321 	CP(*ki, *ki32, ki_lastcpu_old);
1322 
1323 	bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1324 	bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1325 	bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1326 	bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1327 	bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1328 	bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1329 	bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1330 	bcopy(ki->ki_moretdname, ki32->ki_moretdname, MAXCOMLEN - TDNAMLEN + 1);
1331 	CP(*ki, *ki32, ki_tracer);
1332 	CP(*ki, *ki32, ki_flag2);
1333 	CP(*ki, *ki32, ki_fibnum);
1334 	CP(*ki, *ki32, ki_cr_flags);
1335 	CP(*ki, *ki32, ki_jid);
1336 	CP(*ki, *ki32, ki_numthreads);
1337 	CP(*ki, *ki32, ki_tid);
1338 	CP(*ki, *ki32, ki_pri);
1339 	freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1340 	freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1341 	PTRTRIM_CP(*ki, *ki32, ki_pcb);
1342 	PTRTRIM_CP(*ki, *ki32, ki_kstack);
1343 	PTRTRIM_CP(*ki, *ki32, ki_udata);
1344 	PTRTRIM_CP(*ki, *ki32, ki_tdaddr);
1345 	CP(*ki, *ki32, ki_sflag);
1346 	CP(*ki, *ki32, ki_tdflags);
1347 }
1348 #endif
1349 
1350 static ssize_t
1351 kern_proc_out_size(struct proc *p, int flags)
1352 {
1353 	ssize_t size = 0;
1354 
1355 	PROC_LOCK_ASSERT(p, MA_OWNED);
1356 
1357 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1358 #ifdef COMPAT_FREEBSD32
1359 		if ((flags & KERN_PROC_MASK32) != 0) {
1360 			size += sizeof(struct kinfo_proc32);
1361 		} else
1362 #endif
1363 			size += sizeof(struct kinfo_proc);
1364 	} else {
1365 #ifdef COMPAT_FREEBSD32
1366 		if ((flags & KERN_PROC_MASK32) != 0)
1367 			size += sizeof(struct kinfo_proc32) * p->p_numthreads;
1368 		else
1369 #endif
1370 			size += sizeof(struct kinfo_proc) * p->p_numthreads;
1371 	}
1372 	PROC_UNLOCK(p);
1373 	return (size);
1374 }
1375 
1376 int
1377 kern_proc_out(struct proc *p, struct sbuf *sb, int flags)
1378 {
1379 	struct thread *td;
1380 	struct kinfo_proc ki;
1381 #ifdef COMPAT_FREEBSD32
1382 	struct kinfo_proc32 ki32;
1383 #endif
1384 	int error;
1385 
1386 	PROC_LOCK_ASSERT(p, MA_OWNED);
1387 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1388 
1389 	error = 0;
1390 	fill_kinfo_proc(p, &ki);
1391 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1392 #ifdef COMPAT_FREEBSD32
1393 		if ((flags & KERN_PROC_MASK32) != 0) {
1394 			freebsd32_kinfo_proc_out(&ki, &ki32);
1395 			if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1396 				error = ENOMEM;
1397 		} else
1398 #endif
1399 			if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1400 				error = ENOMEM;
1401 	} else {
1402 		FOREACH_THREAD_IN_PROC(p, td) {
1403 			fill_kinfo_thread(td, &ki, 1);
1404 #ifdef COMPAT_FREEBSD32
1405 			if ((flags & KERN_PROC_MASK32) != 0) {
1406 				freebsd32_kinfo_proc_out(&ki, &ki32);
1407 				if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1408 					error = ENOMEM;
1409 			} else
1410 #endif
1411 				if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1412 					error = ENOMEM;
1413 			if (error != 0)
1414 				break;
1415 		}
1416 	}
1417 	PROC_UNLOCK(p);
1418 	return (error);
1419 }
1420 
1421 static int
1422 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
1423 {
1424 	struct sbuf sb;
1425 	struct kinfo_proc ki;
1426 	int error, error2;
1427 
1428 	if (req->oldptr == NULL)
1429 		return (SYSCTL_OUT(req, 0, kern_proc_out_size(p, flags)));
1430 
1431 	sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req);
1432 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1433 	error = kern_proc_out(p, &sb, flags);
1434 	error2 = sbuf_finish(&sb);
1435 	sbuf_delete(&sb);
1436 	if (error != 0)
1437 		return (error);
1438 	else if (error2 != 0)
1439 		return (error2);
1440 	return (0);
1441 }
1442 
1443 static int
1444 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1445 {
1446 	int *name = (int *)arg1;
1447 	u_int namelen = arg2;
1448 	struct proc *p;
1449 	int flags, doingzomb, oid_number;
1450 	int error = 0;
1451 
1452 	oid_number = oidp->oid_number;
1453 	if (oid_number != KERN_PROC_ALL &&
1454 	    (oid_number & KERN_PROC_INC_THREAD) == 0)
1455 		flags = KERN_PROC_NOTHREADS;
1456 	else {
1457 		flags = 0;
1458 		oid_number &= ~KERN_PROC_INC_THREAD;
1459 	}
1460 #ifdef COMPAT_FREEBSD32
1461 	if (req->flags & SCTL_MASK32)
1462 		flags |= KERN_PROC_MASK32;
1463 #endif
1464 	if (oid_number == KERN_PROC_PID) {
1465 		if (namelen != 1)
1466 			return (EINVAL);
1467 		error = sysctl_wire_old_buffer(req, 0);
1468 		if (error)
1469 			return (error);
1470 		sx_slock(&proctree_lock);
1471 		error = pget((pid_t)name[0], PGET_CANSEE, &p);
1472 		if (error == 0)
1473 			error = sysctl_out_proc(p, req, flags);
1474 		sx_sunlock(&proctree_lock);
1475 		return (error);
1476 	}
1477 
1478 	switch (oid_number) {
1479 	case KERN_PROC_ALL:
1480 		if (namelen != 0)
1481 			return (EINVAL);
1482 		break;
1483 	case KERN_PROC_PROC:
1484 		if (namelen != 0 && namelen != 1)
1485 			return (EINVAL);
1486 		break;
1487 	default:
1488 		if (namelen != 1)
1489 			return (EINVAL);
1490 		break;
1491 	}
1492 
1493 	if (req->oldptr == NULL) {
1494 		/* overestimate by 5 procs */
1495 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1496 		if (error)
1497 			return (error);
1498 	} else {
1499 		error = sysctl_wire_old_buffer(req, 0);
1500 		if (error != 0)
1501 			return (error);
1502 		/*
1503 		 * This lock is only needed to safely grab the parent of a
1504 		 * traced process. Only grab it if we are producing any
1505 		 * data to begin with.
1506 		 */
1507 		sx_slock(&proctree_lock);
1508 	}
1509 	sx_slock(&allproc_lock);
1510 	for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) {
1511 		if (!doingzomb)
1512 			p = LIST_FIRST(&allproc);
1513 		else
1514 			p = LIST_FIRST(&zombproc);
1515 		for (; p != NULL; p = LIST_NEXT(p, p_list)) {
1516 			/*
1517 			 * Skip embryonic processes.
1518 			 */
1519 			if (p->p_state == PRS_NEW)
1520 				continue;
1521 			PROC_LOCK(p);
1522 			KASSERT(p->p_ucred != NULL,
1523 			    ("process credential is NULL for non-NEW proc"));
1524 			/*
1525 			 * Show a user only appropriate processes.
1526 			 */
1527 			if (p_cansee(curthread, p)) {
1528 				PROC_UNLOCK(p);
1529 				continue;
1530 			}
1531 			/*
1532 			 * TODO - make more efficient (see notes below).
1533 			 * do by session.
1534 			 */
1535 			switch (oid_number) {
1536 
1537 			case KERN_PROC_GID:
1538 				if (p->p_ucred->cr_gid != (gid_t)name[0]) {
1539 					PROC_UNLOCK(p);
1540 					continue;
1541 				}
1542 				break;
1543 
1544 			case KERN_PROC_PGRP:
1545 				/* could do this by traversing pgrp */
1546 				if (p->p_pgrp == NULL ||
1547 				    p->p_pgrp->pg_id != (pid_t)name[0]) {
1548 					PROC_UNLOCK(p);
1549 					continue;
1550 				}
1551 				break;
1552 
1553 			case KERN_PROC_RGID:
1554 				if (p->p_ucred->cr_rgid != (gid_t)name[0]) {
1555 					PROC_UNLOCK(p);
1556 					continue;
1557 				}
1558 				break;
1559 
1560 			case KERN_PROC_SESSION:
1561 				if (p->p_session == NULL ||
1562 				    p->p_session->s_sid != (pid_t)name[0]) {
1563 					PROC_UNLOCK(p);
1564 					continue;
1565 				}
1566 				break;
1567 
1568 			case KERN_PROC_TTY:
1569 				if ((p->p_flag & P_CONTROLT) == 0 ||
1570 				    p->p_session == NULL) {
1571 					PROC_UNLOCK(p);
1572 					continue;
1573 				}
1574 				/* XXX proctree_lock */
1575 				SESS_LOCK(p->p_session);
1576 				if (p->p_session->s_ttyp == NULL ||
1577 				    tty_udev(p->p_session->s_ttyp) !=
1578 				    (dev_t)name[0]) {
1579 					SESS_UNLOCK(p->p_session);
1580 					PROC_UNLOCK(p);
1581 					continue;
1582 				}
1583 				SESS_UNLOCK(p->p_session);
1584 				break;
1585 
1586 			case KERN_PROC_UID:
1587 				if (p->p_ucred->cr_uid != (uid_t)name[0]) {
1588 					PROC_UNLOCK(p);
1589 					continue;
1590 				}
1591 				break;
1592 
1593 			case KERN_PROC_RUID:
1594 				if (p->p_ucred->cr_ruid != (uid_t)name[0]) {
1595 					PROC_UNLOCK(p);
1596 					continue;
1597 				}
1598 				break;
1599 
1600 			case KERN_PROC_PROC:
1601 				break;
1602 
1603 			default:
1604 				break;
1605 
1606 			}
1607 
1608 			error = sysctl_out_proc(p, req, flags);
1609 			if (error)
1610 				goto out;
1611 		}
1612 	}
1613 out:
1614 	sx_sunlock(&allproc_lock);
1615 	if (req->oldptr != NULL)
1616 		sx_sunlock(&proctree_lock);
1617 	return (error);
1618 }
1619 
1620 struct pargs *
1621 pargs_alloc(int len)
1622 {
1623 	struct pargs *pa;
1624 
1625 	pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1626 		M_WAITOK);
1627 	refcount_init(&pa->ar_ref, 1);
1628 	pa->ar_length = len;
1629 	return (pa);
1630 }
1631 
1632 static void
1633 pargs_free(struct pargs *pa)
1634 {
1635 
1636 	free(pa, M_PARGS);
1637 }
1638 
1639 void
1640 pargs_hold(struct pargs *pa)
1641 {
1642 
1643 	if (pa == NULL)
1644 		return;
1645 	refcount_acquire(&pa->ar_ref);
1646 }
1647 
1648 void
1649 pargs_drop(struct pargs *pa)
1650 {
1651 
1652 	if (pa == NULL)
1653 		return;
1654 	if (refcount_release(&pa->ar_ref))
1655 		pargs_free(pa);
1656 }
1657 
1658 static int
1659 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1660     size_t len)
1661 {
1662 	ssize_t n;
1663 
1664 	/*
1665 	 * This may return a short read if the string is shorter than the chunk
1666 	 * and is aligned at the end of the page, and the following page is not
1667 	 * mapped.
1668 	 */
1669 	n = proc_readmem(td, p, (vm_offset_t)sptr, buf, len);
1670 	if (n <= 0)
1671 		return (ENOMEM);
1672 	return (0);
1673 }
1674 
1675 #define PROC_AUXV_MAX	256	/* Safety limit on auxv size. */
1676 
1677 enum proc_vector_type {
1678 	PROC_ARG,
1679 	PROC_ENV,
1680 	PROC_AUX,
1681 };
1682 
1683 #ifdef COMPAT_FREEBSD32
1684 static int
1685 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1686     size_t *vsizep, enum proc_vector_type type)
1687 {
1688 	struct freebsd32_ps_strings pss;
1689 	Elf32_Auxinfo aux;
1690 	vm_offset_t vptr, ptr;
1691 	uint32_t *proc_vector32;
1692 	char **proc_vector;
1693 	size_t vsize, size;
1694 	int i, error;
1695 
1696 	error = 0;
1697 	if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss,
1698 	    sizeof(pss)) != sizeof(pss))
1699 		return (ENOMEM);
1700 	switch (type) {
1701 	case PROC_ARG:
1702 		vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1703 		vsize = pss.ps_nargvstr;
1704 		if (vsize > ARG_MAX)
1705 			return (ENOEXEC);
1706 		size = vsize * sizeof(int32_t);
1707 		break;
1708 	case PROC_ENV:
1709 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1710 		vsize = pss.ps_nenvstr;
1711 		if (vsize > ARG_MAX)
1712 			return (ENOEXEC);
1713 		size = vsize * sizeof(int32_t);
1714 		break;
1715 	case PROC_AUX:
1716 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1717 		    (pss.ps_nenvstr + 1) * sizeof(int32_t);
1718 		if (vptr % 4 != 0)
1719 			return (ENOEXEC);
1720 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1721 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1722 			    sizeof(aux))
1723 				return (ENOMEM);
1724 			if (aux.a_type == AT_NULL)
1725 				break;
1726 			ptr += sizeof(aux);
1727 		}
1728 		if (aux.a_type != AT_NULL)
1729 			return (ENOEXEC);
1730 		vsize = i + 1;
1731 		size = vsize * sizeof(aux);
1732 		break;
1733 	default:
1734 		KASSERT(0, ("Wrong proc vector type: %d", type));
1735 		return (EINVAL);
1736 	}
1737 	proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1738 	if (proc_readmem(td, p, vptr, proc_vector32, size) != size) {
1739 		error = ENOMEM;
1740 		goto done;
1741 	}
1742 	if (type == PROC_AUX) {
1743 		*proc_vectorp = (char **)proc_vector32;
1744 		*vsizep = vsize;
1745 		return (0);
1746 	}
1747 	proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1748 	for (i = 0; i < (int)vsize; i++)
1749 		proc_vector[i] = PTRIN(proc_vector32[i]);
1750 	*proc_vectorp = proc_vector;
1751 	*vsizep = vsize;
1752 done:
1753 	free(proc_vector32, M_TEMP);
1754 	return (error);
1755 }
1756 #endif
1757 
1758 static int
1759 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1760     size_t *vsizep, enum proc_vector_type type)
1761 {
1762 	struct ps_strings pss;
1763 	Elf_Auxinfo aux;
1764 	vm_offset_t vptr, ptr;
1765 	char **proc_vector;
1766 	size_t vsize, size;
1767 	int i;
1768 
1769 #ifdef COMPAT_FREEBSD32
1770 	if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1771 		return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1772 #endif
1773 	if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss,
1774 	    sizeof(pss)) != sizeof(pss))
1775 		return (ENOMEM);
1776 	switch (type) {
1777 	case PROC_ARG:
1778 		vptr = (vm_offset_t)pss.ps_argvstr;
1779 		vsize = pss.ps_nargvstr;
1780 		if (vsize > ARG_MAX)
1781 			return (ENOEXEC);
1782 		size = vsize * sizeof(char *);
1783 		break;
1784 	case PROC_ENV:
1785 		vptr = (vm_offset_t)pss.ps_envstr;
1786 		vsize = pss.ps_nenvstr;
1787 		if (vsize > ARG_MAX)
1788 			return (ENOEXEC);
1789 		size = vsize * sizeof(char *);
1790 		break;
1791 	case PROC_AUX:
1792 		/*
1793 		 * The aux array is just above env array on the stack. Check
1794 		 * that the address is naturally aligned.
1795 		 */
1796 		vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1797 		    * sizeof(char *);
1798 #if __ELF_WORD_SIZE == 64
1799 		if (vptr % sizeof(uint64_t) != 0)
1800 #else
1801 		if (vptr % sizeof(uint32_t) != 0)
1802 #endif
1803 			return (ENOEXEC);
1804 		/*
1805 		 * We count the array size reading the aux vectors from the
1806 		 * stack until AT_NULL vector is returned.  So (to keep the code
1807 		 * simple) we read the process stack twice: the first time here
1808 		 * to find the size and the second time when copying the vectors
1809 		 * to the allocated proc_vector.
1810 		 */
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 		/*
1820 		 * If the PROC_AUXV_MAX entries are iterated over, and we have
1821 		 * not reached AT_NULL, it is most likely we are reading wrong
1822 		 * data: either the process doesn't have auxv array or data has
1823 		 * been modified. Return the error in this case.
1824 		 */
1825 		if (aux.a_type != AT_NULL)
1826 			return (ENOEXEC);
1827 		vsize = i + 1;
1828 		size = vsize * sizeof(aux);
1829 		break;
1830 	default:
1831 		KASSERT(0, ("Wrong proc vector type: %d", type));
1832 		return (EINVAL); /* In case we are built without INVARIANTS. */
1833 	}
1834 	proc_vector = malloc(size, M_TEMP, M_WAITOK);
1835 	if (proc_readmem(td, p, vptr, proc_vector, size) != size) {
1836 		free(proc_vector, M_TEMP);
1837 		return (ENOMEM);
1838 	}
1839 	*proc_vectorp = proc_vector;
1840 	*vsizep = vsize;
1841 
1842 	return (0);
1843 }
1844 
1845 #define GET_PS_STRINGS_CHUNK_SZ	256	/* Chunk size (bytes) for ps_strings operations. */
1846 
1847 static int
1848 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
1849     enum proc_vector_type type)
1850 {
1851 	size_t done, len, nchr, vsize;
1852 	int error, i;
1853 	char **proc_vector, *sptr;
1854 	char pss_string[GET_PS_STRINGS_CHUNK_SZ];
1855 
1856 	PROC_ASSERT_HELD(p);
1857 
1858 	/*
1859 	 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
1860 	 */
1861 	nchr = 2 * (PATH_MAX + ARG_MAX);
1862 
1863 	error = get_proc_vector(td, p, &proc_vector, &vsize, type);
1864 	if (error != 0)
1865 		return (error);
1866 	for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
1867 		/*
1868 		 * The program may have scribbled into its argv array, e.g. to
1869 		 * remove some arguments.  If that has happened, break out
1870 		 * before trying to read from NULL.
1871 		 */
1872 		if (proc_vector[i] == NULL)
1873 			break;
1874 		for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
1875 			error = proc_read_string(td, p, sptr, pss_string,
1876 			    sizeof(pss_string));
1877 			if (error != 0)
1878 				goto done;
1879 			len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
1880 			if (done + len >= nchr)
1881 				len = nchr - done - 1;
1882 			sbuf_bcat(sb, pss_string, len);
1883 			if (len != GET_PS_STRINGS_CHUNK_SZ)
1884 				break;
1885 			done += GET_PS_STRINGS_CHUNK_SZ;
1886 		}
1887 		sbuf_bcat(sb, "", 1);
1888 		done += len + 1;
1889 	}
1890 done:
1891 	free(proc_vector, M_TEMP);
1892 	return (error);
1893 }
1894 
1895 int
1896 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
1897 {
1898 
1899 	return (get_ps_strings(curthread, p, sb, PROC_ARG));
1900 }
1901 
1902 int
1903 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
1904 {
1905 
1906 	return (get_ps_strings(curthread, p, sb, PROC_ENV));
1907 }
1908 
1909 int
1910 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb)
1911 {
1912 	size_t vsize, size;
1913 	char **auxv;
1914 	int error;
1915 
1916 	error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX);
1917 	if (error == 0) {
1918 #ifdef COMPAT_FREEBSD32
1919 		if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1920 			size = vsize * sizeof(Elf32_Auxinfo);
1921 		else
1922 #endif
1923 			size = vsize * sizeof(Elf_Auxinfo);
1924 		if (sbuf_bcat(sb, auxv, size) != 0)
1925 			error = ENOMEM;
1926 		free(auxv, M_TEMP);
1927 	}
1928 	return (error);
1929 }
1930 
1931 /*
1932  * This sysctl allows a process to retrieve the argument list or process
1933  * title for another process without groping around in the address space
1934  * of the other process.  It also allow a process to set its own "process
1935  * title to a string of its own choice.
1936  */
1937 static int
1938 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
1939 {
1940 	int *name = (int *)arg1;
1941 	u_int namelen = arg2;
1942 	struct pargs *newpa, *pa;
1943 	struct proc *p;
1944 	struct sbuf sb;
1945 	int flags, error = 0, error2;
1946 	pid_t pid;
1947 
1948 	if (namelen != 1)
1949 		return (EINVAL);
1950 
1951 	pid = (pid_t)name[0];
1952 	/*
1953 	 * If the query is for this process and it is single-threaded, there
1954 	 * is nobody to modify pargs, thus we can just read.
1955 	 */
1956 	p = curproc;
1957 	if (pid == p->p_pid && p->p_numthreads == 1 && req->newptr == NULL &&
1958 	    (pa = p->p_args) != NULL)
1959 		return (SYSCTL_OUT(req, pa->ar_args, pa->ar_length));
1960 
1961 	flags = PGET_CANSEE;
1962 	if (req->newptr != NULL)
1963 		flags |= PGET_ISCURRENT;
1964 	error = pget(pid, flags, &p);
1965 	if (error)
1966 		return (error);
1967 
1968 	pa = p->p_args;
1969 	if (pa != NULL) {
1970 		pargs_hold(pa);
1971 		PROC_UNLOCK(p);
1972 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
1973 		pargs_drop(pa);
1974 	} else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
1975 		_PHOLD(p);
1976 		PROC_UNLOCK(p);
1977 		sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1978 		sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1979 		error = proc_getargv(curthread, p, &sb);
1980 		error2 = sbuf_finish(&sb);
1981 		PRELE(p);
1982 		sbuf_delete(&sb);
1983 		if (error == 0 && error2 != 0)
1984 			error = error2;
1985 	} else {
1986 		PROC_UNLOCK(p);
1987 	}
1988 	if (error != 0 || req->newptr == NULL)
1989 		return (error);
1990 
1991 	if (req->newlen > ps_arg_cache_limit - sizeof(struct pargs))
1992 		return (ENOMEM);
1993 	newpa = pargs_alloc(req->newlen);
1994 	error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
1995 	if (error != 0) {
1996 		pargs_free(newpa);
1997 		return (error);
1998 	}
1999 	PROC_LOCK(p);
2000 	pa = p->p_args;
2001 	p->p_args = newpa;
2002 	PROC_UNLOCK(p);
2003 	pargs_drop(pa);
2004 	return (0);
2005 }
2006 
2007 /*
2008  * This sysctl allows a process to retrieve environment of another process.
2009  */
2010 static int
2011 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
2012 {
2013 	int *name = (int *)arg1;
2014 	u_int namelen = arg2;
2015 	struct proc *p;
2016 	struct sbuf sb;
2017 	int error, error2;
2018 
2019 	if (namelen != 1)
2020 		return (EINVAL);
2021 
2022 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2023 	if (error != 0)
2024 		return (error);
2025 	if ((p->p_flag & P_SYSTEM) != 0) {
2026 		PRELE(p);
2027 		return (0);
2028 	}
2029 
2030 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2031 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2032 	error = proc_getenvv(curthread, p, &sb);
2033 	error2 = sbuf_finish(&sb);
2034 	PRELE(p);
2035 	sbuf_delete(&sb);
2036 	return (error != 0 ? error : error2);
2037 }
2038 
2039 /*
2040  * This sysctl allows a process to retrieve ELF auxiliary vector of
2041  * another process.
2042  */
2043 static int
2044 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
2045 {
2046 	int *name = (int *)arg1;
2047 	u_int namelen = arg2;
2048 	struct proc *p;
2049 	struct sbuf sb;
2050 	int error, error2;
2051 
2052 	if (namelen != 1)
2053 		return (EINVAL);
2054 
2055 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2056 	if (error != 0)
2057 		return (error);
2058 	if ((p->p_flag & P_SYSTEM) != 0) {
2059 		PRELE(p);
2060 		return (0);
2061 	}
2062 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2063 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2064 	error = proc_getauxv(curthread, p, &sb);
2065 	error2 = sbuf_finish(&sb);
2066 	PRELE(p);
2067 	sbuf_delete(&sb);
2068 	return (error != 0 ? error : error2);
2069 }
2070 
2071 /*
2072  * This sysctl allows a process to retrieve the path of the executable for
2073  * itself or another process.
2074  */
2075 static int
2076 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
2077 {
2078 	pid_t *pidp = (pid_t *)arg1;
2079 	unsigned int arglen = arg2;
2080 	struct proc *p;
2081 	struct vnode *vp;
2082 	char *retbuf, *freebuf;
2083 	int error;
2084 
2085 	if (arglen != 1)
2086 		return (EINVAL);
2087 	if (*pidp == -1) {	/* -1 means this process */
2088 		p = req->td->td_proc;
2089 	} else {
2090 		error = pget(*pidp, PGET_CANSEE, &p);
2091 		if (error != 0)
2092 			return (error);
2093 	}
2094 
2095 	vp = p->p_textvp;
2096 	if (vp == NULL) {
2097 		if (*pidp != -1)
2098 			PROC_UNLOCK(p);
2099 		return (0);
2100 	}
2101 	vref(vp);
2102 	if (*pidp != -1)
2103 		PROC_UNLOCK(p);
2104 	error = vn_fullpath(req->td, vp, &retbuf, &freebuf);
2105 	vrele(vp);
2106 	if (error)
2107 		return (error);
2108 	error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
2109 	free(freebuf, M_TEMP);
2110 	return (error);
2111 }
2112 
2113 static int
2114 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
2115 {
2116 	struct proc *p;
2117 	char *sv_name;
2118 	int *name;
2119 	int namelen;
2120 	int error;
2121 
2122 	namelen = arg2;
2123 	if (namelen != 1)
2124 		return (EINVAL);
2125 
2126 	name = (int *)arg1;
2127 	error = pget((pid_t)name[0], PGET_CANSEE, &p);
2128 	if (error != 0)
2129 		return (error);
2130 	sv_name = p->p_sysent->sv_name;
2131 	PROC_UNLOCK(p);
2132 	return (sysctl_handle_string(oidp, sv_name, 0, req));
2133 }
2134 
2135 #ifdef KINFO_OVMENTRY_SIZE
2136 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
2137 #endif
2138 
2139 #ifdef COMPAT_FREEBSD7
2140 static int
2141 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
2142 {
2143 	vm_map_entry_t entry, tmp_entry;
2144 	unsigned int last_timestamp;
2145 	char *fullpath, *freepath;
2146 	struct kinfo_ovmentry *kve;
2147 	struct vattr va;
2148 	struct ucred *cred;
2149 	int error, *name;
2150 	struct vnode *vp;
2151 	struct proc *p;
2152 	vm_map_t map;
2153 	struct vmspace *vm;
2154 
2155 	name = (int *)arg1;
2156 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2157 	if (error != 0)
2158 		return (error);
2159 	vm = vmspace_acquire_ref(p);
2160 	if (vm == NULL) {
2161 		PRELE(p);
2162 		return (ESRCH);
2163 	}
2164 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2165 
2166 	map = &vm->vm_map;
2167 	vm_map_lock_read(map);
2168 	for (entry = map->header.next; entry != &map->header;
2169 	    entry = entry->next) {
2170 		vm_object_t obj, tobj, lobj;
2171 		vm_offset_t addr;
2172 
2173 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2174 			continue;
2175 
2176 		bzero(kve, sizeof(*kve));
2177 		kve->kve_structsize = sizeof(*kve);
2178 
2179 		kve->kve_private_resident = 0;
2180 		obj = entry->object.vm_object;
2181 		if (obj != NULL) {
2182 			VM_OBJECT_RLOCK(obj);
2183 			if (obj->shadow_count == 1)
2184 				kve->kve_private_resident =
2185 				    obj->resident_page_count;
2186 		}
2187 		kve->kve_resident = 0;
2188 		addr = entry->start;
2189 		while (addr < entry->end) {
2190 			if (pmap_extract(map->pmap, addr))
2191 				kve->kve_resident++;
2192 			addr += PAGE_SIZE;
2193 		}
2194 
2195 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2196 			if (tobj != obj) {
2197 				VM_OBJECT_RLOCK(tobj);
2198 				kve->kve_offset += tobj->backing_object_offset;
2199 			}
2200 			if (lobj != obj)
2201 				VM_OBJECT_RUNLOCK(lobj);
2202 			lobj = tobj;
2203 		}
2204 
2205 		kve->kve_start = (void*)entry->start;
2206 		kve->kve_end = (void*)entry->end;
2207 		kve->kve_offset += (off_t)entry->offset;
2208 
2209 		if (entry->protection & VM_PROT_READ)
2210 			kve->kve_protection |= KVME_PROT_READ;
2211 		if (entry->protection & VM_PROT_WRITE)
2212 			kve->kve_protection |= KVME_PROT_WRITE;
2213 		if (entry->protection & VM_PROT_EXECUTE)
2214 			kve->kve_protection |= KVME_PROT_EXEC;
2215 
2216 		if (entry->eflags & MAP_ENTRY_COW)
2217 			kve->kve_flags |= KVME_FLAG_COW;
2218 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2219 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2220 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2221 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2222 
2223 		last_timestamp = map->timestamp;
2224 		vm_map_unlock_read(map);
2225 
2226 		kve->kve_fileid = 0;
2227 		kve->kve_fsid = 0;
2228 		freepath = NULL;
2229 		fullpath = "";
2230 		if (lobj) {
2231 			vp = NULL;
2232 			switch (lobj->type) {
2233 			case OBJT_DEFAULT:
2234 				kve->kve_type = KVME_TYPE_DEFAULT;
2235 				break;
2236 			case OBJT_VNODE:
2237 				kve->kve_type = KVME_TYPE_VNODE;
2238 				vp = lobj->handle;
2239 				vref(vp);
2240 				break;
2241 			case OBJT_SWAP:
2242 				if ((lobj->flags & OBJ_TMPFS_NODE) != 0) {
2243 					kve->kve_type = KVME_TYPE_VNODE;
2244 					if ((lobj->flags & OBJ_TMPFS) != 0) {
2245 						vp = lobj->un_pager.swp.swp_tmpfs;
2246 						vref(vp);
2247 					}
2248 				} else {
2249 					kve->kve_type = KVME_TYPE_SWAP;
2250 				}
2251 				break;
2252 			case OBJT_DEVICE:
2253 				kve->kve_type = KVME_TYPE_DEVICE;
2254 				break;
2255 			case OBJT_PHYS:
2256 				kve->kve_type = KVME_TYPE_PHYS;
2257 				break;
2258 			case OBJT_DEAD:
2259 				kve->kve_type = KVME_TYPE_DEAD;
2260 				break;
2261 			case OBJT_SG:
2262 				kve->kve_type = KVME_TYPE_SG;
2263 				break;
2264 			default:
2265 				kve->kve_type = KVME_TYPE_UNKNOWN;
2266 				break;
2267 			}
2268 			if (lobj != obj)
2269 				VM_OBJECT_RUNLOCK(lobj);
2270 
2271 			kve->kve_ref_count = obj->ref_count;
2272 			kve->kve_shadow_count = obj->shadow_count;
2273 			VM_OBJECT_RUNLOCK(obj);
2274 			if (vp != NULL) {
2275 				vn_fullpath(curthread, vp, &fullpath,
2276 				    &freepath);
2277 				cred = curthread->td_ucred;
2278 				vn_lock(vp, LK_SHARED | LK_RETRY);
2279 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2280 					kve->kve_fileid = va.va_fileid;
2281 					/* truncate */
2282 					kve->kve_fsid = va.va_fsid;
2283 				}
2284 				vput(vp);
2285 			}
2286 		} else {
2287 			kve->kve_type = KVME_TYPE_NONE;
2288 			kve->kve_ref_count = 0;
2289 			kve->kve_shadow_count = 0;
2290 		}
2291 
2292 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2293 		if (freepath != NULL)
2294 			free(freepath, M_TEMP);
2295 
2296 		error = SYSCTL_OUT(req, kve, sizeof(*kve));
2297 		vm_map_lock_read(map);
2298 		if (error)
2299 			break;
2300 		if (last_timestamp != map->timestamp) {
2301 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2302 			entry = tmp_entry;
2303 		}
2304 	}
2305 	vm_map_unlock_read(map);
2306 	vmspace_free(vm);
2307 	PRELE(p);
2308 	free(kve, M_TEMP);
2309 	return (error);
2310 }
2311 #endif	/* COMPAT_FREEBSD7 */
2312 
2313 #ifdef KINFO_VMENTRY_SIZE
2314 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2315 #endif
2316 
2317 void
2318 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry,
2319     int *resident_count, bool *super)
2320 {
2321 	vm_object_t obj, tobj;
2322 	vm_page_t m, m_adv;
2323 	vm_offset_t addr;
2324 	vm_paddr_t locked_pa;
2325 	vm_pindex_t pi, pi_adv, pindex;
2326 
2327 	*super = false;
2328 	*resident_count = 0;
2329 	if (vmmap_skip_res_cnt)
2330 		return;
2331 
2332 	locked_pa = 0;
2333 	obj = entry->object.vm_object;
2334 	addr = entry->start;
2335 	m_adv = NULL;
2336 	pi = OFF_TO_IDX(entry->offset);
2337 	for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) {
2338 		if (m_adv != NULL) {
2339 			m = m_adv;
2340 		} else {
2341 			pi_adv = atop(entry->end - addr);
2342 			pindex = pi;
2343 			for (tobj = obj;; tobj = tobj->backing_object) {
2344 				m = vm_page_find_least(tobj, pindex);
2345 				if (m != NULL) {
2346 					if (m->pindex == pindex)
2347 						break;
2348 					if (pi_adv > m->pindex - pindex) {
2349 						pi_adv = m->pindex - pindex;
2350 						m_adv = m;
2351 					}
2352 				}
2353 				if (tobj->backing_object == NULL)
2354 					goto next;
2355 				pindex += OFF_TO_IDX(tobj->
2356 				    backing_object_offset);
2357 			}
2358 		}
2359 		m_adv = NULL;
2360 		if (m->psind != 0 && addr + pagesizes[1] <= entry->end &&
2361 		    (addr & (pagesizes[1] - 1)) == 0 &&
2362 		    (pmap_mincore(map->pmap, addr, &locked_pa) &
2363 		    MINCORE_SUPER) != 0) {
2364 			*super = true;
2365 			pi_adv = atop(pagesizes[1]);
2366 		} else {
2367 			/*
2368 			 * We do not test the found page on validity.
2369 			 * Either the page is busy and being paged in,
2370 			 * or it was invalidated.  The first case
2371 			 * should be counted as resident, the second
2372 			 * is not so clear; we do account both.
2373 			 */
2374 			pi_adv = 1;
2375 		}
2376 		*resident_count += pi_adv;
2377 next:;
2378 	}
2379 	PA_UNLOCK_COND(locked_pa);
2380 }
2381 
2382 /*
2383  * Must be called with the process locked and will return unlocked.
2384  */
2385 int
2386 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb, ssize_t maxlen, int flags)
2387 {
2388 	vm_map_entry_t entry, tmp_entry;
2389 	struct vattr va;
2390 	vm_map_t map;
2391 	vm_object_t obj, tobj, lobj;
2392 	char *fullpath, *freepath;
2393 	struct kinfo_vmentry *kve;
2394 	struct ucred *cred;
2395 	struct vnode *vp;
2396 	struct vmspace *vm;
2397 	vm_offset_t addr;
2398 	unsigned int last_timestamp;
2399 	int error;
2400 	bool super;
2401 
2402 	PROC_LOCK_ASSERT(p, MA_OWNED);
2403 
2404 	_PHOLD(p);
2405 	PROC_UNLOCK(p);
2406 	vm = vmspace_acquire_ref(p);
2407 	if (vm == NULL) {
2408 		PRELE(p);
2409 		return (ESRCH);
2410 	}
2411 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK | M_ZERO);
2412 
2413 	error = 0;
2414 	map = &vm->vm_map;
2415 	vm_map_lock_read(map);
2416 	for (entry = map->header.next; entry != &map->header;
2417 	    entry = entry->next) {
2418 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2419 			continue;
2420 
2421 		addr = entry->end;
2422 		bzero(kve, sizeof(*kve));
2423 		obj = entry->object.vm_object;
2424 		if (obj != NULL) {
2425 			for (tobj = obj; tobj != NULL;
2426 			    tobj = tobj->backing_object) {
2427 				VM_OBJECT_RLOCK(tobj);
2428 				kve->kve_offset += tobj->backing_object_offset;
2429 				lobj = tobj;
2430 			}
2431 			if (obj->backing_object == NULL)
2432 				kve->kve_private_resident =
2433 				    obj->resident_page_count;
2434 			kern_proc_vmmap_resident(map, entry,
2435 			    &kve->kve_resident, &super);
2436 			if (super)
2437 				kve->kve_flags |= KVME_FLAG_SUPER;
2438 			for (tobj = obj; tobj != NULL;
2439 			    tobj = tobj->backing_object) {
2440 				if (tobj != obj && tobj != lobj)
2441 					VM_OBJECT_RUNLOCK(tobj);
2442 			}
2443 		} else {
2444 			lobj = NULL;
2445 		}
2446 
2447 		kve->kve_start = entry->start;
2448 		kve->kve_end = entry->end;
2449 		kve->kve_offset += entry->offset;
2450 
2451 		if (entry->protection & VM_PROT_READ)
2452 			kve->kve_protection |= KVME_PROT_READ;
2453 		if (entry->protection & VM_PROT_WRITE)
2454 			kve->kve_protection |= KVME_PROT_WRITE;
2455 		if (entry->protection & VM_PROT_EXECUTE)
2456 			kve->kve_protection |= KVME_PROT_EXEC;
2457 
2458 		if (entry->eflags & MAP_ENTRY_COW)
2459 			kve->kve_flags |= KVME_FLAG_COW;
2460 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2461 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2462 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2463 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2464 		if (entry->eflags & MAP_ENTRY_GROWS_UP)
2465 			kve->kve_flags |= KVME_FLAG_GROWS_UP;
2466 		if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2467 			kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2468 
2469 		last_timestamp = map->timestamp;
2470 		vm_map_unlock_read(map);
2471 
2472 		freepath = NULL;
2473 		fullpath = "";
2474 		if (lobj != NULL) {
2475 			vp = NULL;
2476 			switch (lobj->type) {
2477 			case OBJT_DEFAULT:
2478 				kve->kve_type = KVME_TYPE_DEFAULT;
2479 				break;
2480 			case OBJT_VNODE:
2481 				kve->kve_type = KVME_TYPE_VNODE;
2482 				vp = lobj->handle;
2483 				vref(vp);
2484 				break;
2485 			case OBJT_SWAP:
2486 				if ((lobj->flags & OBJ_TMPFS_NODE) != 0) {
2487 					kve->kve_type = KVME_TYPE_VNODE;
2488 					if ((lobj->flags & OBJ_TMPFS) != 0) {
2489 						vp = lobj->un_pager.swp.swp_tmpfs;
2490 						vref(vp);
2491 					}
2492 				} else {
2493 					kve->kve_type = KVME_TYPE_SWAP;
2494 				}
2495 				break;
2496 			case OBJT_DEVICE:
2497 				kve->kve_type = KVME_TYPE_DEVICE;
2498 				break;
2499 			case OBJT_PHYS:
2500 				kve->kve_type = KVME_TYPE_PHYS;
2501 				break;
2502 			case OBJT_DEAD:
2503 				kve->kve_type = KVME_TYPE_DEAD;
2504 				break;
2505 			case OBJT_SG:
2506 				kve->kve_type = KVME_TYPE_SG;
2507 				break;
2508 			case OBJT_MGTDEVICE:
2509 				kve->kve_type = KVME_TYPE_MGTDEVICE;
2510 				break;
2511 			default:
2512 				kve->kve_type = KVME_TYPE_UNKNOWN;
2513 				break;
2514 			}
2515 			if (lobj != obj)
2516 				VM_OBJECT_RUNLOCK(lobj);
2517 
2518 			kve->kve_ref_count = obj->ref_count;
2519 			kve->kve_shadow_count = obj->shadow_count;
2520 			VM_OBJECT_RUNLOCK(obj);
2521 			if (vp != NULL) {
2522 				vn_fullpath(curthread, vp, &fullpath,
2523 				    &freepath);
2524 				kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2525 				cred = curthread->td_ucred;
2526 				vn_lock(vp, LK_SHARED | LK_RETRY);
2527 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2528 					kve->kve_vn_fileid = va.va_fileid;
2529 					kve->kve_vn_fsid = va.va_fsid;
2530 					kve->kve_vn_fsid_freebsd11 =
2531 					    kve->kve_vn_fsid; /* truncate */
2532 					kve->kve_vn_mode =
2533 					    MAKEIMODE(va.va_type, va.va_mode);
2534 					kve->kve_vn_size = va.va_size;
2535 					kve->kve_vn_rdev = va.va_rdev;
2536 					kve->kve_vn_rdev_freebsd11 =
2537 					    kve->kve_vn_rdev; /* truncate */
2538 					kve->kve_status = KF_ATTR_VALID;
2539 				}
2540 				vput(vp);
2541 			}
2542 		} else {
2543 			kve->kve_type = KVME_TYPE_NONE;
2544 			kve->kve_ref_count = 0;
2545 			kve->kve_shadow_count = 0;
2546 		}
2547 
2548 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2549 		if (freepath != NULL)
2550 			free(freepath, M_TEMP);
2551 
2552 		/* Pack record size down */
2553 		if ((flags & KERN_VMMAP_PACK_KINFO) != 0)
2554 			kve->kve_structsize =
2555 			    offsetof(struct kinfo_vmentry, kve_path) +
2556 			    strlen(kve->kve_path) + 1;
2557 		else
2558 			kve->kve_structsize = sizeof(*kve);
2559 		kve->kve_structsize = roundup(kve->kve_structsize,
2560 		    sizeof(uint64_t));
2561 
2562 		/* Halt filling and truncate rather than exceeding maxlen */
2563 		if (maxlen != -1 && maxlen < kve->kve_structsize) {
2564 			error = 0;
2565 			vm_map_lock_read(map);
2566 			break;
2567 		} else if (maxlen != -1)
2568 			maxlen -= kve->kve_structsize;
2569 
2570 		if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0)
2571 			error = ENOMEM;
2572 		vm_map_lock_read(map);
2573 		if (error != 0)
2574 			break;
2575 		if (last_timestamp != map->timestamp) {
2576 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2577 			entry = tmp_entry;
2578 		}
2579 	}
2580 	vm_map_unlock_read(map);
2581 	vmspace_free(vm);
2582 	PRELE(p);
2583 	free(kve, M_TEMP);
2584 	return (error);
2585 }
2586 
2587 static int
2588 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2589 {
2590 	struct proc *p;
2591 	struct sbuf sb;
2592 	int error, error2, *name;
2593 
2594 	name = (int *)arg1;
2595 	sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req);
2596 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2597 	error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
2598 	if (error != 0) {
2599 		sbuf_delete(&sb);
2600 		return (error);
2601 	}
2602 	error = kern_proc_vmmap_out(p, &sb, -1, KERN_VMMAP_PACK_KINFO);
2603 	error2 = sbuf_finish(&sb);
2604 	sbuf_delete(&sb);
2605 	return (error != 0 ? error : error2);
2606 }
2607 
2608 #if defined(STACK) || defined(DDB)
2609 static int
2610 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2611 {
2612 	struct kinfo_kstack *kkstp;
2613 	int error, i, *name, numthreads;
2614 	lwpid_t *lwpidarray;
2615 	struct thread *td;
2616 	struct stack *st;
2617 	struct sbuf sb;
2618 	struct proc *p;
2619 
2620 	name = (int *)arg1;
2621 	error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2622 	if (error != 0)
2623 		return (error);
2624 
2625 	kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2626 	st = stack_create(M_WAITOK);
2627 
2628 	lwpidarray = NULL;
2629 	PROC_LOCK(p);
2630 	do {
2631 		if (lwpidarray != NULL) {
2632 			free(lwpidarray, M_TEMP);
2633 			lwpidarray = NULL;
2634 		}
2635 		numthreads = p->p_numthreads;
2636 		PROC_UNLOCK(p);
2637 		lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2638 		    M_WAITOK | M_ZERO);
2639 		PROC_LOCK(p);
2640 	} while (numthreads < p->p_numthreads);
2641 
2642 	/*
2643 	 * XXXRW: During the below loop, execve(2) and countless other sorts
2644 	 * of changes could have taken place.  Should we check to see if the
2645 	 * vmspace has been replaced, or the like, in order to prevent
2646 	 * giving a snapshot that spans, say, execve(2), with some threads
2647 	 * before and some after?  Among other things, the credentials could
2648 	 * have changed, in which case the right to extract debug info might
2649 	 * no longer be assured.
2650 	 */
2651 	i = 0;
2652 	FOREACH_THREAD_IN_PROC(p, td) {
2653 		KASSERT(i < numthreads,
2654 		    ("sysctl_kern_proc_kstack: numthreads"));
2655 		lwpidarray[i] = td->td_tid;
2656 		i++;
2657 	}
2658 	numthreads = i;
2659 	for (i = 0; i < numthreads; i++) {
2660 		td = thread_find(p, lwpidarray[i]);
2661 		if (td == NULL) {
2662 			continue;
2663 		}
2664 		bzero(kkstp, sizeof(*kkstp));
2665 		(void)sbuf_new(&sb, kkstp->kkst_trace,
2666 		    sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2667 		thread_lock(td);
2668 		kkstp->kkst_tid = td->td_tid;
2669 		if (TD_IS_SWAPPED(td)) {
2670 			kkstp->kkst_state = KKST_STATE_SWAPPED;
2671 		} else if (TD_IS_RUNNING(td)) {
2672 			if (stack_save_td_running(st, td) == 0)
2673 				kkstp->kkst_state = KKST_STATE_STACKOK;
2674 			else
2675 				kkstp->kkst_state = KKST_STATE_RUNNING;
2676 		} else {
2677 			kkstp->kkst_state = KKST_STATE_STACKOK;
2678 			stack_save_td(st, td);
2679 		}
2680 		thread_unlock(td);
2681 		PROC_UNLOCK(p);
2682 		stack_sbuf_print(&sb, st);
2683 		sbuf_finish(&sb);
2684 		sbuf_delete(&sb);
2685 		error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2686 		PROC_LOCK(p);
2687 		if (error)
2688 			break;
2689 	}
2690 	_PRELE(p);
2691 	PROC_UNLOCK(p);
2692 	if (lwpidarray != NULL)
2693 		free(lwpidarray, M_TEMP);
2694 	stack_destroy(st);
2695 	free(kkstp, M_TEMP);
2696 	return (error);
2697 }
2698 #endif
2699 
2700 /*
2701  * This sysctl allows a process to retrieve the full list of groups from
2702  * itself or another process.
2703  */
2704 static int
2705 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2706 {
2707 	pid_t *pidp = (pid_t *)arg1;
2708 	unsigned int arglen = arg2;
2709 	struct proc *p;
2710 	struct ucred *cred;
2711 	int error;
2712 
2713 	if (arglen != 1)
2714 		return (EINVAL);
2715 	if (*pidp == -1) {	/* -1 means this process */
2716 		p = req->td->td_proc;
2717 		PROC_LOCK(p);
2718 	} else {
2719 		error = pget(*pidp, PGET_CANSEE, &p);
2720 		if (error != 0)
2721 			return (error);
2722 	}
2723 
2724 	cred = crhold(p->p_ucred);
2725 	PROC_UNLOCK(p);
2726 
2727 	error = SYSCTL_OUT(req, cred->cr_groups,
2728 	    cred->cr_ngroups * sizeof(gid_t));
2729 	crfree(cred);
2730 	return (error);
2731 }
2732 
2733 /*
2734  * This sysctl allows a process to retrieve or/and set the resource limit for
2735  * another process.
2736  */
2737 static int
2738 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2739 {
2740 	int *name = (int *)arg1;
2741 	u_int namelen = arg2;
2742 	struct rlimit rlim;
2743 	struct proc *p;
2744 	u_int which;
2745 	int flags, error;
2746 
2747 	if (namelen != 2)
2748 		return (EINVAL);
2749 
2750 	which = (u_int)name[1];
2751 	if (which >= RLIM_NLIMITS)
2752 		return (EINVAL);
2753 
2754 	if (req->newptr != NULL && req->newlen != sizeof(rlim))
2755 		return (EINVAL);
2756 
2757 	flags = PGET_HOLD | PGET_NOTWEXIT;
2758 	if (req->newptr != NULL)
2759 		flags |= PGET_CANDEBUG;
2760 	else
2761 		flags |= PGET_CANSEE;
2762 	error = pget((pid_t)name[0], flags, &p);
2763 	if (error != 0)
2764 		return (error);
2765 
2766 	/*
2767 	 * Retrieve limit.
2768 	 */
2769 	if (req->oldptr != NULL) {
2770 		PROC_LOCK(p);
2771 		lim_rlimit_proc(p, which, &rlim);
2772 		PROC_UNLOCK(p);
2773 	}
2774 	error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
2775 	if (error != 0)
2776 		goto errout;
2777 
2778 	/*
2779 	 * Set limit.
2780 	 */
2781 	if (req->newptr != NULL) {
2782 		error = SYSCTL_IN(req, &rlim, sizeof(rlim));
2783 		if (error == 0)
2784 			error = kern_proc_setrlimit(curthread, p, which, &rlim);
2785 	}
2786 
2787 errout:
2788 	PRELE(p);
2789 	return (error);
2790 }
2791 
2792 /*
2793  * This sysctl allows a process to retrieve ps_strings structure location of
2794  * another process.
2795  */
2796 static int
2797 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
2798 {
2799 	int *name = (int *)arg1;
2800 	u_int namelen = arg2;
2801 	struct proc *p;
2802 	vm_offset_t ps_strings;
2803 	int error;
2804 #ifdef COMPAT_FREEBSD32
2805 	uint32_t ps_strings32;
2806 #endif
2807 
2808 	if (namelen != 1)
2809 		return (EINVAL);
2810 
2811 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2812 	if (error != 0)
2813 		return (error);
2814 #ifdef COMPAT_FREEBSD32
2815 	if ((req->flags & SCTL_MASK32) != 0) {
2816 		/*
2817 		 * We return 0 if the 32 bit emulation request is for a 64 bit
2818 		 * process.
2819 		 */
2820 		ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
2821 		    PTROUT(p->p_sysent->sv_psstrings) : 0;
2822 		PROC_UNLOCK(p);
2823 		error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
2824 		return (error);
2825 	}
2826 #endif
2827 	ps_strings = p->p_sysent->sv_psstrings;
2828 	PROC_UNLOCK(p);
2829 	error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
2830 	return (error);
2831 }
2832 
2833 /*
2834  * This sysctl allows a process to retrieve umask of another process.
2835  */
2836 static int
2837 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
2838 {
2839 	int *name = (int *)arg1;
2840 	u_int namelen = arg2;
2841 	struct proc *p;
2842 	int error;
2843 	u_short fd_cmask;
2844 	pid_t pid;
2845 
2846 	if (namelen != 1)
2847 		return (EINVAL);
2848 
2849 	pid = (pid_t)name[0];
2850 	p = curproc;
2851 	if (pid == p->p_pid || pid == 0) {
2852 		fd_cmask = p->p_fd->fd_cmask;
2853 		goto out;
2854 	}
2855 
2856 	error = pget(pid, PGET_WANTREAD, &p);
2857 	if (error != 0)
2858 		return (error);
2859 
2860 	fd_cmask = p->p_fd->fd_cmask;
2861 	PRELE(p);
2862 out:
2863 	error = SYSCTL_OUT(req, &fd_cmask, sizeof(fd_cmask));
2864 	return (error);
2865 }
2866 
2867 /*
2868  * This sysctl allows a process to set and retrieve binary osreldate of
2869  * another process.
2870  */
2871 static int
2872 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
2873 {
2874 	int *name = (int *)arg1;
2875 	u_int namelen = arg2;
2876 	struct proc *p;
2877 	int flags, error, osrel;
2878 
2879 	if (namelen != 1)
2880 		return (EINVAL);
2881 
2882 	if (req->newptr != NULL && req->newlen != sizeof(osrel))
2883 		return (EINVAL);
2884 
2885 	flags = PGET_HOLD | PGET_NOTWEXIT;
2886 	if (req->newptr != NULL)
2887 		flags |= PGET_CANDEBUG;
2888 	else
2889 		flags |= PGET_CANSEE;
2890 	error = pget((pid_t)name[0], flags, &p);
2891 	if (error != 0)
2892 		return (error);
2893 
2894 	error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
2895 	if (error != 0)
2896 		goto errout;
2897 
2898 	if (req->newptr != NULL) {
2899 		error = SYSCTL_IN(req, &osrel, sizeof(osrel));
2900 		if (error != 0)
2901 			goto errout;
2902 		if (osrel < 0) {
2903 			error = EINVAL;
2904 			goto errout;
2905 		}
2906 		p->p_osrel = osrel;
2907 	}
2908 errout:
2909 	PRELE(p);
2910 	return (error);
2911 }
2912 
2913 static int
2914 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS)
2915 {
2916 	int *name = (int *)arg1;
2917 	u_int namelen = arg2;
2918 	struct proc *p;
2919 	struct kinfo_sigtramp kst;
2920 	const struct sysentvec *sv;
2921 	int error;
2922 #ifdef COMPAT_FREEBSD32
2923 	struct kinfo_sigtramp32 kst32;
2924 #endif
2925 
2926 	if (namelen != 1)
2927 		return (EINVAL);
2928 
2929 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2930 	if (error != 0)
2931 		return (error);
2932 	sv = p->p_sysent;
2933 #ifdef COMPAT_FREEBSD32
2934 	if ((req->flags & SCTL_MASK32) != 0) {
2935 		bzero(&kst32, sizeof(kst32));
2936 		if (SV_PROC_FLAG(p, SV_ILP32)) {
2937 			if (sv->sv_sigcode_base != 0) {
2938 				kst32.ksigtramp_start = sv->sv_sigcode_base;
2939 				kst32.ksigtramp_end = sv->sv_sigcode_base +
2940 				    *sv->sv_szsigcode;
2941 			} else {
2942 				kst32.ksigtramp_start = sv->sv_psstrings -
2943 				    *sv->sv_szsigcode;
2944 				kst32.ksigtramp_end = sv->sv_psstrings;
2945 			}
2946 		}
2947 		PROC_UNLOCK(p);
2948 		error = SYSCTL_OUT(req, &kst32, sizeof(kst32));
2949 		return (error);
2950 	}
2951 #endif
2952 	bzero(&kst, sizeof(kst));
2953 	if (sv->sv_sigcode_base != 0) {
2954 		kst.ksigtramp_start = (char *)sv->sv_sigcode_base;
2955 		kst.ksigtramp_end = (char *)sv->sv_sigcode_base +
2956 		    *sv->sv_szsigcode;
2957 	} else {
2958 		kst.ksigtramp_start = (char *)sv->sv_psstrings -
2959 		    *sv->sv_szsigcode;
2960 		kst.ksigtramp_end = (char *)sv->sv_psstrings;
2961 	}
2962 	PROC_UNLOCK(p);
2963 	error = SYSCTL_OUT(req, &kst, sizeof(kst));
2964 	return (error);
2965 }
2966 
2967 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD,  0, "Process table");
2968 
2969 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
2970 	CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
2971 	"Return entire process table");
2972 
2973 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2974 	sysctl_kern_proc, "Process table");
2975 
2976 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
2977 	sysctl_kern_proc, "Process table");
2978 
2979 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2980 	sysctl_kern_proc, "Process table");
2981 
2982 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
2983 	CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2984 
2985 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
2986 	sysctl_kern_proc, "Process table");
2987 
2988 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2989 	sysctl_kern_proc, "Process table");
2990 
2991 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2992 	sysctl_kern_proc, "Process table");
2993 
2994 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2995 	sysctl_kern_proc, "Process table");
2996 
2997 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
2998 	sysctl_kern_proc, "Return process table, no threads");
2999 
3000 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
3001 	CTLFLAG_RW | CTLFLAG_CAPWR | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
3002 	sysctl_kern_proc_args, "Process argument list");
3003 
3004 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
3005 	sysctl_kern_proc_env, "Process environment");
3006 
3007 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
3008 	CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
3009 
3010 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
3011 	CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
3012 
3013 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
3014 	CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
3015 	"Process syscall vector name (ABI type)");
3016 
3017 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
3018 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3019 
3020 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
3021 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3022 
3023 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
3024 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3025 
3026 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
3027 	sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3028 
3029 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
3030 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3031 
3032 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
3033 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3034 
3035 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
3036 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3037 
3038 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
3039 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3040 
3041 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
3042 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
3043 	"Return process table, no threads");
3044 
3045 #ifdef COMPAT_FREEBSD7
3046 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
3047 	CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
3048 #endif
3049 
3050 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
3051 	CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
3052 
3053 #if defined(STACK) || defined(DDB)
3054 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
3055 	CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
3056 #endif
3057 
3058 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
3059 	CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
3060 
3061 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
3062 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
3063 	"Process resource limits");
3064 
3065 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
3066 	CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
3067 	"Process ps_strings location");
3068 
3069 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
3070 	CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
3071 
3072 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
3073 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
3074 	"Process binary osreldate");
3075 
3076 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD |
3077 	CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp,
3078 	"Process signal trampoline location");
3079 
3080 int allproc_gen;
3081 
3082 /*
3083  * stop_all_proc() purpose is to stop all process which have usermode,
3084  * except current process for obvious reasons.  This makes it somewhat
3085  * unreliable when invoked from multithreaded process.  The service
3086  * must not be user-callable anyway.
3087  */
3088 void
3089 stop_all_proc(void)
3090 {
3091 	struct proc *cp, *p;
3092 	int r, gen;
3093 	bool restart, seen_stopped, seen_exiting, stopped_some;
3094 
3095 	cp = curproc;
3096 allproc_loop:
3097 	sx_xlock(&allproc_lock);
3098 	gen = allproc_gen;
3099 	seen_exiting = seen_stopped = stopped_some = restart = false;
3100 	LIST_REMOVE(cp, p_list);
3101 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3102 	for (;;) {
3103 		p = LIST_NEXT(cp, p_list);
3104 		if (p == NULL)
3105 			break;
3106 		LIST_REMOVE(cp, p_list);
3107 		LIST_INSERT_AFTER(p, cp, p_list);
3108 		PROC_LOCK(p);
3109 		if ((p->p_flag & (P_KPROC | P_SYSTEM | P_TOTAL_STOP)) != 0) {
3110 			PROC_UNLOCK(p);
3111 			continue;
3112 		}
3113 		if ((p->p_flag & P_WEXIT) != 0) {
3114 			seen_exiting = true;
3115 			PROC_UNLOCK(p);
3116 			continue;
3117 		}
3118 		if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) {
3119 			/*
3120 			 * Stopped processes are tolerated when there
3121 			 * are no other processes which might continue
3122 			 * them.  P_STOPPED_SINGLE but not
3123 			 * P_TOTAL_STOP process still has at least one
3124 			 * thread running.
3125 			 */
3126 			seen_stopped = true;
3127 			PROC_UNLOCK(p);
3128 			continue;
3129 		}
3130 		_PHOLD(p);
3131 		sx_xunlock(&allproc_lock);
3132 		r = thread_single(p, SINGLE_ALLPROC);
3133 		if (r != 0)
3134 			restart = true;
3135 		else
3136 			stopped_some = true;
3137 		_PRELE(p);
3138 		PROC_UNLOCK(p);
3139 		sx_xlock(&allproc_lock);
3140 	}
3141 	/* Catch forked children we did not see in iteration. */
3142 	if (gen != allproc_gen)
3143 		restart = true;
3144 	sx_xunlock(&allproc_lock);
3145 	if (restart || stopped_some || seen_exiting || seen_stopped) {
3146 		kern_yield(PRI_USER);
3147 		goto allproc_loop;
3148 	}
3149 }
3150 
3151 void
3152 resume_all_proc(void)
3153 {
3154 	struct proc *cp, *p;
3155 
3156 	cp = curproc;
3157 	sx_xlock(&allproc_lock);
3158 again:
3159 	LIST_REMOVE(cp, p_list);
3160 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3161 	for (;;) {
3162 		p = LIST_NEXT(cp, p_list);
3163 		if (p == NULL)
3164 			break;
3165 		LIST_REMOVE(cp, p_list);
3166 		LIST_INSERT_AFTER(p, cp, p_list);
3167 		PROC_LOCK(p);
3168 		if ((p->p_flag & P_TOTAL_STOP) != 0) {
3169 			sx_xunlock(&allproc_lock);
3170 			_PHOLD(p);
3171 			thread_single_end(p, SINGLE_ALLPROC);
3172 			_PRELE(p);
3173 			PROC_UNLOCK(p);
3174 			sx_xlock(&allproc_lock);
3175 		} else {
3176 			PROC_UNLOCK(p);
3177 		}
3178 	}
3179 	/*  Did the loop above missed any stopped process ? */
3180 	LIST_FOREACH(p, &allproc, p_list) {
3181 		/* No need for proc lock. */
3182 		if ((p->p_flag & P_TOTAL_STOP) != 0)
3183 			goto again;
3184 	}
3185 	sx_xunlock(&allproc_lock);
3186 }
3187 
3188 /* #define	TOTAL_STOP_DEBUG	1 */
3189 #ifdef TOTAL_STOP_DEBUG
3190 volatile static int ap_resume;
3191 #include <sys/mount.h>
3192 
3193 static int
3194 sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS)
3195 {
3196 	int error, val;
3197 
3198 	val = 0;
3199 	ap_resume = 0;
3200 	error = sysctl_handle_int(oidp, &val, 0, req);
3201 	if (error != 0 || req->newptr == NULL)
3202 		return (error);
3203 	if (val != 0) {
3204 		stop_all_proc();
3205 		syncer_suspend();
3206 		while (ap_resume == 0)
3207 			;
3208 		syncer_resume();
3209 		resume_all_proc();
3210 	}
3211 	return (0);
3212 }
3213 
3214 SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW |
3215     CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0,
3216     sysctl_debug_stop_all_proc, "I",
3217     "");
3218 #endif
3219