xref: /freebsd/sys/kern/kern_proc.c (revision 389e4940069316fe667ffa263fa7d6390d0a960f)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1982, 1986, 1989, 1991, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)kern_proc.c	8.7 (Berkeley) 2/14/95
32  */
33 
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
36 
37 #include "opt_ddb.h"
38 #include "opt_ktrace.h"
39 #include "opt_kstack_pages.h"
40 #include "opt_stack.h"
41 
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/elf.h>
45 #include <sys/eventhandler.h>
46 #include <sys/exec.h>
47 #include <sys/jail.h>
48 #include <sys/kernel.h>
49 #include <sys/limits.h>
50 #include <sys/lock.h>
51 #include <sys/loginclass.h>
52 #include <sys/malloc.h>
53 #include <sys/mman.h>
54 #include <sys/mount.h>
55 #include <sys/mutex.h>
56 #include <sys/proc.h>
57 #include <sys/ptrace.h>
58 #include <sys/refcount.h>
59 #include <sys/resourcevar.h>
60 #include <sys/rwlock.h>
61 #include <sys/sbuf.h>
62 #include <sys/sysent.h>
63 #include <sys/sched.h>
64 #include <sys/smp.h>
65 #include <sys/stack.h>
66 #include <sys/stat.h>
67 #include <sys/sysctl.h>
68 #include <sys/filedesc.h>
69 #include <sys/tty.h>
70 #include <sys/signalvar.h>
71 #include <sys/sdt.h>
72 #include <sys/sx.h>
73 #include <sys/user.h>
74 #include <sys/vnode.h>
75 #include <sys/wait.h>
76 
77 #ifdef DDB
78 #include <ddb/ddb.h>
79 #endif
80 
81 #include <vm/vm.h>
82 #include <vm/vm_param.h>
83 #include <vm/vm_extern.h>
84 #include <vm/pmap.h>
85 #include <vm/vm_map.h>
86 #include <vm/vm_object.h>
87 #include <vm/vm_page.h>
88 #include <vm/uma.h>
89 
90 #ifdef COMPAT_FREEBSD32
91 #include <compat/freebsd32/freebsd32.h>
92 #include <compat/freebsd32/freebsd32_util.h>
93 #endif
94 
95 SDT_PROVIDER_DEFINE(proc);
96 SDT_PROBE_DEFINE4(proc, , ctor, entry, "struct proc *", "int", "void *",
97     "int");
98 SDT_PROBE_DEFINE4(proc, , ctor, return, "struct proc *", "int", "void *",
99     "int");
100 SDT_PROBE_DEFINE4(proc, , dtor, entry, "struct proc *", "int", "void *",
101     "struct thread *");
102 SDT_PROBE_DEFINE3(proc, , dtor, return, "struct proc *", "int", "void *");
103 SDT_PROBE_DEFINE3(proc, , init, entry, "struct proc *", "int", "int");
104 SDT_PROBE_DEFINE3(proc, , init, return, "struct proc *", "int", "int");
105 
106 MALLOC_DEFINE(M_PGRP, "pgrp", "process group header");
107 MALLOC_DEFINE(M_SESSION, "session", "session header");
108 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
109 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
110 
111 static void doenterpgrp(struct proc *, struct pgrp *);
112 static void orphanpg(struct pgrp *pg);
113 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp);
114 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp);
115 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp,
116     int preferthread);
117 static void pgadjustjobc(struct pgrp *pgrp, int entering);
118 static void pgdelete(struct pgrp *);
119 static int proc_ctor(void *mem, int size, void *arg, int flags);
120 static void proc_dtor(void *mem, int size, void *arg);
121 static int proc_init(void *mem, int size, int flags);
122 static void proc_fini(void *mem, int size);
123 static void pargs_free(struct pargs *pa);
124 static struct proc *zpfind_locked(pid_t pid);
125 
126 /*
127  * Other process lists
128  */
129 struct pidhashhead *pidhashtbl;
130 u_long pidhash;
131 struct pgrphashhead *pgrphashtbl;
132 u_long pgrphash;
133 struct proclist allproc;
134 struct proclist zombproc;
135 struct sx __exclusive_cache_line allproc_lock;
136 struct sx __exclusive_cache_line proctree_lock;
137 struct mtx __exclusive_cache_line ppeers_lock;
138 uma_zone_t proc_zone;
139 
140 /*
141  * The offset of various fields in struct proc and struct thread.
142  * These are used by kernel debuggers to enumerate kernel threads and
143  * processes.
144  */
145 const int proc_off_p_pid = offsetof(struct proc, p_pid);
146 const int proc_off_p_comm = offsetof(struct proc, p_comm);
147 const int proc_off_p_list = offsetof(struct proc, p_list);
148 const int proc_off_p_threads = offsetof(struct proc, p_threads);
149 const int thread_off_td_tid = offsetof(struct thread, td_tid);
150 const int thread_off_td_name = offsetof(struct thread, td_name);
151 const int thread_off_td_oncpu = offsetof(struct thread, td_oncpu);
152 const int thread_off_td_pcb = offsetof(struct thread, td_pcb);
153 const int thread_off_td_plist = offsetof(struct thread, td_plist);
154 
155 EVENTHANDLER_LIST_DEFINE(process_ctor);
156 EVENTHANDLER_LIST_DEFINE(process_dtor);
157 EVENTHANDLER_LIST_DEFINE(process_init);
158 EVENTHANDLER_LIST_DEFINE(process_fini);
159 EVENTHANDLER_LIST_DEFINE(process_exit);
160 EVENTHANDLER_LIST_DEFINE(process_fork);
161 EVENTHANDLER_LIST_DEFINE(process_exec);
162 
163 EVENTHANDLER_LIST_DECLARE(thread_ctor);
164 EVENTHANDLER_LIST_DECLARE(thread_dtor);
165 
166 int kstack_pages = KSTACK_PAGES;
167 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0,
168     "Kernel stack size in pages");
169 static int vmmap_skip_res_cnt = 0;
170 SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW,
171     &vmmap_skip_res_cnt, 0,
172     "Skip calculation of the pages resident count in kern.proc.vmmap");
173 
174 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
175 #ifdef COMPAT_FREEBSD32
176 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
177 #endif
178 
179 /*
180  * Initialize global process hashing structures.
181  */
182 void
183 procinit(void)
184 {
185 
186 	sx_init(&allproc_lock, "allproc");
187 	sx_init(&proctree_lock, "proctree");
188 	mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
189 	LIST_INIT(&allproc);
190 	LIST_INIT(&zombproc);
191 	pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
192 	pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
193 	proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
194 	    proc_ctor, proc_dtor, proc_init, proc_fini,
195 	    UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
196 	uihashinit();
197 }
198 
199 /*
200  * Prepare a proc for use.
201  */
202 static int
203 proc_ctor(void *mem, int size, void *arg, int flags)
204 {
205 	struct proc *p;
206 	struct thread *td;
207 
208 	p = (struct proc *)mem;
209 	SDT_PROBE4(proc, , ctor , entry, p, size, arg, flags);
210 	EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
211 	SDT_PROBE4(proc, , ctor , return, p, size, arg, flags);
212 	td = FIRST_THREAD_IN_PROC(p);
213 	if (td != NULL) {
214 		/* Make sure all thread constructors are executed */
215 		EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
216 	}
217 	return (0);
218 }
219 
220 /*
221  * Reclaim a proc after use.
222  */
223 static void
224 proc_dtor(void *mem, int size, void *arg)
225 {
226 	struct proc *p;
227 	struct thread *td;
228 
229 	/* INVARIANTS checks go here */
230 	p = (struct proc *)mem;
231 	td = FIRST_THREAD_IN_PROC(p);
232 	SDT_PROBE4(proc, , dtor, entry, p, size, arg, td);
233 	if (td != NULL) {
234 #ifdef INVARIANTS
235 		KASSERT((p->p_numthreads == 1),
236 		    ("bad number of threads in exiting process"));
237 		KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
238 #endif
239 		/* Free all OSD associated to this thread. */
240 		osd_thread_exit(td);
241 		td_softdep_cleanup(td);
242 		MPASS(td->td_su == NULL);
243 
244 		/* Make sure all thread destructors are executed */
245 		EVENTHANDLER_DIRECT_INVOKE(thread_dtor, td);
246 	}
247 	EVENTHANDLER_DIRECT_INVOKE(process_dtor, p);
248 	if (p->p_ksi != NULL)
249 		KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
250 	SDT_PROBE3(proc, , dtor, return, p, size, arg);
251 }
252 
253 /*
254  * Initialize type-stable parts of a proc (when newly created).
255  */
256 static int
257 proc_init(void *mem, int size, int flags)
258 {
259 	struct proc *p;
260 
261 	p = (struct proc *)mem;
262 	SDT_PROBE3(proc, , init, entry, p, size, flags);
263 	mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK | MTX_NEW);
264 	mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_NEW);
265 	mtx_init(&p->p_statmtx, "pstatl", NULL, MTX_SPIN | MTX_NEW);
266 	mtx_init(&p->p_itimmtx, "pitiml", NULL, MTX_SPIN | MTX_NEW);
267 	mtx_init(&p->p_profmtx, "pprofl", NULL, MTX_SPIN | MTX_NEW);
268 	cv_init(&p->p_pwait, "ppwait");
269 	cv_init(&p->p_dbgwait, "dbgwait");
270 	TAILQ_INIT(&p->p_threads);	     /* all threads in proc */
271 	EVENTHANDLER_DIRECT_INVOKE(process_init, p);
272 	p->p_stats = pstats_alloc();
273 	p->p_pgrp = NULL;
274 	SDT_PROBE3(proc, , init, return, p, size, flags);
275 	return (0);
276 }
277 
278 /*
279  * UMA should ensure that this function is never called.
280  * Freeing a proc structure would violate type stability.
281  */
282 static void
283 proc_fini(void *mem, int size)
284 {
285 #ifdef notnow
286 	struct proc *p;
287 
288 	p = (struct proc *)mem;
289 	EVENTHANDLER_DIRECT_INVOKE(process_fini, p);
290 	pstats_free(p->p_stats);
291 	thread_free(FIRST_THREAD_IN_PROC(p));
292 	mtx_destroy(&p->p_mtx);
293 	if (p->p_ksi != NULL)
294 		ksiginfo_free(p->p_ksi);
295 #else
296 	panic("proc reclaimed");
297 #endif
298 }
299 
300 /*
301  * Is p an inferior of the current process?
302  */
303 int
304 inferior(struct proc *p)
305 {
306 
307 	sx_assert(&proctree_lock, SX_LOCKED);
308 	PROC_LOCK_ASSERT(p, MA_OWNED);
309 	for (; p != curproc; p = proc_realparent(p)) {
310 		if (p->p_pid == 0)
311 			return (0);
312 	}
313 	return (1);
314 }
315 
316 struct proc *
317 pfind_locked(pid_t pid)
318 {
319 	struct proc *p;
320 
321 	sx_assert(&allproc_lock, SX_LOCKED);
322 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
323 		if (p->p_pid == pid) {
324 			PROC_LOCK(p);
325 			if (p->p_state == PRS_NEW) {
326 				PROC_UNLOCK(p);
327 				p = NULL;
328 			}
329 			break;
330 		}
331 	}
332 	return (p);
333 }
334 
335 /*
336  * Locate a process by number; return only "live" processes -- i.e., neither
337  * zombies nor newly born but incompletely initialized processes.  By not
338  * returning processes in the PRS_NEW state, we allow callers to avoid
339  * testing for that condition to avoid dereferencing p_ucred, et al.
340  */
341 struct proc *
342 pfind(pid_t pid)
343 {
344 	struct proc *p;
345 
346 	p = curproc;
347 	if (p->p_pid == pid) {
348 		PROC_LOCK(p);
349 		return (p);
350 	}
351 	sx_slock(&allproc_lock);
352 	p = pfind_locked(pid);
353 	sx_sunlock(&allproc_lock);
354 	return (p);
355 }
356 
357 /*
358  * Same as pfind but allow zombies.
359  */
360 struct proc *
361 pfind_any(pid_t pid)
362 {
363 	struct proc *p;
364 
365 	sx_slock(&allproc_lock);
366 	p = pfind_locked(pid);
367 	if (p == NULL)
368 		p = zpfind_locked(pid);
369 	sx_sunlock(&allproc_lock);
370 
371 	return (p);
372 }
373 
374 static struct proc *
375 pfind_tid_locked(pid_t tid)
376 {
377 	struct proc *p;
378 	struct thread *td;
379 
380 	sx_assert(&allproc_lock, SX_LOCKED);
381 	FOREACH_PROC_IN_SYSTEM(p) {
382 		PROC_LOCK(p);
383 		if (p->p_state == PRS_NEW) {
384 			PROC_UNLOCK(p);
385 			continue;
386 		}
387 		FOREACH_THREAD_IN_PROC(p, td) {
388 			if (td->td_tid == tid)
389 				goto found;
390 		}
391 		PROC_UNLOCK(p);
392 	}
393 found:
394 	return (p);
395 }
396 
397 /*
398  * Locate a process group by number.
399  * The caller must hold proctree_lock.
400  */
401 struct pgrp *
402 pgfind(pid_t pgid)
403 {
404 	struct pgrp *pgrp;
405 
406 	sx_assert(&proctree_lock, SX_LOCKED);
407 
408 	LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
409 		if (pgrp->pg_id == pgid) {
410 			PGRP_LOCK(pgrp);
411 			return (pgrp);
412 		}
413 	}
414 	return (NULL);
415 }
416 
417 /*
418  * Locate process and do additional manipulations, depending on flags.
419  */
420 int
421 pget(pid_t pid, int flags, struct proc **pp)
422 {
423 	struct proc *p;
424 	int error;
425 
426 	p = curproc;
427 	if (p->p_pid == pid) {
428 		PROC_LOCK(p);
429 	} else {
430 		sx_slock(&allproc_lock);
431 		if (pid <= PID_MAX) {
432 			p = pfind_locked(pid);
433 			if (p == NULL && (flags & PGET_NOTWEXIT) == 0)
434 				p = zpfind_locked(pid);
435 		} else if ((flags & PGET_NOTID) == 0) {
436 			p = pfind_tid_locked(pid);
437 		} else {
438 			p = NULL;
439 		}
440 		sx_sunlock(&allproc_lock);
441 		if (p == NULL)
442 			return (ESRCH);
443 		if ((flags & PGET_CANSEE) != 0) {
444 			error = p_cansee(curthread, p);
445 			if (error != 0)
446 				goto errout;
447 		}
448 	}
449 	if ((flags & PGET_CANDEBUG) != 0) {
450 		error = p_candebug(curthread, p);
451 		if (error != 0)
452 			goto errout;
453 	}
454 	if ((flags & PGET_ISCURRENT) != 0 && curproc != p) {
455 		error = EPERM;
456 		goto errout;
457 	}
458 	if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) {
459 		error = ESRCH;
460 		goto errout;
461 	}
462 	if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) {
463 		/*
464 		 * XXXRW: Not clear ESRCH is the right error during proc
465 		 * execve().
466 		 */
467 		error = ESRCH;
468 		goto errout;
469 	}
470 	if ((flags & PGET_HOLD) != 0) {
471 		_PHOLD(p);
472 		PROC_UNLOCK(p);
473 	}
474 	*pp = p;
475 	return (0);
476 errout:
477 	PROC_UNLOCK(p);
478 	return (error);
479 }
480 
481 /*
482  * Create a new process group.
483  * pgid must be equal to the pid of p.
484  * Begin a new session if required.
485  */
486 int
487 enterpgrp(struct proc *p, pid_t pgid, struct pgrp *pgrp, struct session *sess)
488 {
489 
490 	sx_assert(&proctree_lock, SX_XLOCKED);
491 
492 	KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
493 	KASSERT(p->p_pid == pgid,
494 	    ("enterpgrp: new pgrp and pid != pgid"));
495 	KASSERT(pgfind(pgid) == NULL,
496 	    ("enterpgrp: pgrp with pgid exists"));
497 	KASSERT(!SESS_LEADER(p),
498 	    ("enterpgrp: session leader attempted setpgrp"));
499 
500 	mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
501 
502 	if (sess != NULL) {
503 		/*
504 		 * new session
505 		 */
506 		mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
507 		PROC_LOCK(p);
508 		p->p_flag &= ~P_CONTROLT;
509 		PROC_UNLOCK(p);
510 		PGRP_LOCK(pgrp);
511 		sess->s_leader = p;
512 		sess->s_sid = p->p_pid;
513 		refcount_init(&sess->s_count, 1);
514 		sess->s_ttyvp = NULL;
515 		sess->s_ttydp = NULL;
516 		sess->s_ttyp = NULL;
517 		bcopy(p->p_session->s_login, sess->s_login,
518 			    sizeof(sess->s_login));
519 		pgrp->pg_session = sess;
520 		KASSERT(p == curproc,
521 		    ("enterpgrp: mksession and p != curproc"));
522 	} else {
523 		pgrp->pg_session = p->p_session;
524 		sess_hold(pgrp->pg_session);
525 		PGRP_LOCK(pgrp);
526 	}
527 	pgrp->pg_id = pgid;
528 	LIST_INIT(&pgrp->pg_members);
529 
530 	/*
531 	 * As we have an exclusive lock of proctree_lock,
532 	 * this should not deadlock.
533 	 */
534 	LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
535 	pgrp->pg_jobc = 0;
536 	SLIST_INIT(&pgrp->pg_sigiolst);
537 	PGRP_UNLOCK(pgrp);
538 
539 	doenterpgrp(p, pgrp);
540 
541 	return (0);
542 }
543 
544 /*
545  * Move p to an existing process group
546  */
547 int
548 enterthispgrp(struct proc *p, struct pgrp *pgrp)
549 {
550 
551 	sx_assert(&proctree_lock, SX_XLOCKED);
552 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
553 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
554 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
555 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
556 	KASSERT(pgrp->pg_session == p->p_session,
557 		("%s: pgrp's session %p, p->p_session %p.\n",
558 		__func__,
559 		pgrp->pg_session,
560 		p->p_session));
561 	KASSERT(pgrp != p->p_pgrp,
562 		("%s: p belongs to pgrp.", __func__));
563 
564 	doenterpgrp(p, pgrp);
565 
566 	return (0);
567 }
568 
569 /*
570  * Move p to a process group
571  */
572 static void
573 doenterpgrp(struct proc *p, struct pgrp *pgrp)
574 {
575 	struct pgrp *savepgrp;
576 
577 	sx_assert(&proctree_lock, SX_XLOCKED);
578 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
579 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
580 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
581 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
582 
583 	savepgrp = p->p_pgrp;
584 
585 	/*
586 	 * Adjust eligibility of affected pgrps to participate in job control.
587 	 * Increment eligibility counts before decrementing, otherwise we
588 	 * could reach 0 spuriously during the first call.
589 	 */
590 	fixjobc(p, pgrp, 1);
591 	fixjobc(p, p->p_pgrp, 0);
592 
593 	PGRP_LOCK(pgrp);
594 	PGRP_LOCK(savepgrp);
595 	PROC_LOCK(p);
596 	LIST_REMOVE(p, p_pglist);
597 	p->p_pgrp = pgrp;
598 	PROC_UNLOCK(p);
599 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
600 	PGRP_UNLOCK(savepgrp);
601 	PGRP_UNLOCK(pgrp);
602 	if (LIST_EMPTY(&savepgrp->pg_members))
603 		pgdelete(savepgrp);
604 }
605 
606 /*
607  * remove process from process group
608  */
609 int
610 leavepgrp(struct proc *p)
611 {
612 	struct pgrp *savepgrp;
613 
614 	sx_assert(&proctree_lock, SX_XLOCKED);
615 	savepgrp = p->p_pgrp;
616 	PGRP_LOCK(savepgrp);
617 	PROC_LOCK(p);
618 	LIST_REMOVE(p, p_pglist);
619 	p->p_pgrp = NULL;
620 	PROC_UNLOCK(p);
621 	PGRP_UNLOCK(savepgrp);
622 	if (LIST_EMPTY(&savepgrp->pg_members))
623 		pgdelete(savepgrp);
624 	return (0);
625 }
626 
627 /*
628  * delete a process group
629  */
630 static void
631 pgdelete(struct pgrp *pgrp)
632 {
633 	struct session *savesess;
634 	struct tty *tp;
635 
636 	sx_assert(&proctree_lock, SX_XLOCKED);
637 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
638 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
639 
640 	/*
641 	 * Reset any sigio structures pointing to us as a result of
642 	 * F_SETOWN with our pgid.
643 	 */
644 	funsetownlst(&pgrp->pg_sigiolst);
645 
646 	PGRP_LOCK(pgrp);
647 	tp = pgrp->pg_session->s_ttyp;
648 	LIST_REMOVE(pgrp, pg_hash);
649 	savesess = pgrp->pg_session;
650 	PGRP_UNLOCK(pgrp);
651 
652 	/* Remove the reference to the pgrp before deallocating it. */
653 	if (tp != NULL) {
654 		tty_lock(tp);
655 		tty_rel_pgrp(tp, pgrp);
656 	}
657 
658 	mtx_destroy(&pgrp->pg_mtx);
659 	free(pgrp, M_PGRP);
660 	sess_release(savesess);
661 }
662 
663 static void
664 pgadjustjobc(struct pgrp *pgrp, int entering)
665 {
666 
667 	PGRP_LOCK(pgrp);
668 	if (entering)
669 		pgrp->pg_jobc++;
670 	else {
671 		--pgrp->pg_jobc;
672 		if (pgrp->pg_jobc == 0)
673 			orphanpg(pgrp);
674 	}
675 	PGRP_UNLOCK(pgrp);
676 }
677 
678 /*
679  * Adjust pgrp jobc counters when specified process changes process group.
680  * We count the number of processes in each process group that "qualify"
681  * the group for terminal job control (those with a parent in a different
682  * process group of the same session).  If that count reaches zero, the
683  * process group becomes orphaned.  Check both the specified process'
684  * process group and that of its children.
685  * entering == 0 => p is leaving specified group.
686  * entering == 1 => p is entering specified group.
687  */
688 void
689 fixjobc(struct proc *p, struct pgrp *pgrp, int entering)
690 {
691 	struct pgrp *hispgrp;
692 	struct session *mysession;
693 	struct proc *q;
694 
695 	sx_assert(&proctree_lock, SX_LOCKED);
696 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
697 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
698 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
699 
700 	/*
701 	 * Check p's parent to see whether p qualifies its own process
702 	 * group; if so, adjust count for p's process group.
703 	 */
704 	mysession = pgrp->pg_session;
705 	if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
706 	    hispgrp->pg_session == mysession)
707 		pgadjustjobc(pgrp, entering);
708 
709 	/*
710 	 * Check this process' children to see whether they qualify
711 	 * their process groups; if so, adjust counts for children's
712 	 * process groups.
713 	 */
714 	LIST_FOREACH(q, &p->p_children, p_sibling) {
715 		hispgrp = q->p_pgrp;
716 		if (hispgrp == pgrp ||
717 		    hispgrp->pg_session != mysession)
718 			continue;
719 		if (q->p_state == PRS_ZOMBIE)
720 			continue;
721 		pgadjustjobc(hispgrp, entering);
722 	}
723 }
724 
725 void
726 killjobc(void)
727 {
728 	struct session *sp;
729 	struct tty *tp;
730 	struct proc *p;
731 	struct vnode *ttyvp;
732 
733 	p = curproc;
734 	MPASS(p->p_flag & P_WEXIT);
735 	/*
736 	 * Do a quick check to see if there is anything to do with the
737 	 * proctree_lock held. pgrp and LIST_EMPTY checks are for fixjobc().
738 	 */
739 	PROC_LOCK(p);
740 	if (!SESS_LEADER(p) &&
741 	    (p->p_pgrp == p->p_pptr->p_pgrp) &&
742 	    LIST_EMPTY(&p->p_children)) {
743 		PROC_UNLOCK(p);
744 		return;
745 	}
746 	PROC_UNLOCK(p);
747 
748 	sx_xlock(&proctree_lock);
749 	if (SESS_LEADER(p)) {
750 		sp = p->p_session;
751 
752 		/*
753 		 * s_ttyp is not zero'd; we use this to indicate that
754 		 * the session once had a controlling terminal. (for
755 		 * logging and informational purposes)
756 		 */
757 		SESS_LOCK(sp);
758 		ttyvp = sp->s_ttyvp;
759 		tp = sp->s_ttyp;
760 		sp->s_ttyvp = NULL;
761 		sp->s_ttydp = NULL;
762 		sp->s_leader = NULL;
763 		SESS_UNLOCK(sp);
764 
765 		/*
766 		 * Signal foreground pgrp and revoke access to
767 		 * controlling terminal if it has not been revoked
768 		 * already.
769 		 *
770 		 * Because the TTY may have been revoked in the mean
771 		 * time and could already have a new session associated
772 		 * with it, make sure we don't send a SIGHUP to a
773 		 * foreground process group that does not belong to this
774 		 * session.
775 		 */
776 
777 		if (tp != NULL) {
778 			tty_lock(tp);
779 			if (tp->t_session == sp)
780 				tty_signal_pgrp(tp, SIGHUP);
781 			tty_unlock(tp);
782 		}
783 
784 		if (ttyvp != NULL) {
785 			sx_xunlock(&proctree_lock);
786 			if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) {
787 				VOP_REVOKE(ttyvp, REVOKEALL);
788 				VOP_UNLOCK(ttyvp, 0);
789 			}
790 			vrele(ttyvp);
791 			sx_xlock(&proctree_lock);
792 		}
793 	}
794 	fixjobc(p, p->p_pgrp, 0);
795 	sx_xunlock(&proctree_lock);
796 }
797 
798 /*
799  * A process group has become orphaned;
800  * if there are any stopped processes in the group,
801  * hang-up all process in that group.
802  */
803 static void
804 orphanpg(struct pgrp *pg)
805 {
806 	struct proc *p;
807 
808 	PGRP_LOCK_ASSERT(pg, MA_OWNED);
809 
810 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
811 		PROC_LOCK(p);
812 		if (P_SHOULDSTOP(p) == P_STOPPED_SIG) {
813 			PROC_UNLOCK(p);
814 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
815 				PROC_LOCK(p);
816 				kern_psignal(p, SIGHUP);
817 				kern_psignal(p, SIGCONT);
818 				PROC_UNLOCK(p);
819 			}
820 			return;
821 		}
822 		PROC_UNLOCK(p);
823 	}
824 }
825 
826 void
827 sess_hold(struct session *s)
828 {
829 
830 	refcount_acquire(&s->s_count);
831 }
832 
833 void
834 sess_release(struct session *s)
835 {
836 
837 	if (refcount_release(&s->s_count)) {
838 		if (s->s_ttyp != NULL) {
839 			tty_lock(s->s_ttyp);
840 			tty_rel_sess(s->s_ttyp, s);
841 		}
842 		mtx_destroy(&s->s_mtx);
843 		free(s, M_SESSION);
844 	}
845 }
846 
847 #ifdef DDB
848 
849 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
850 {
851 	struct pgrp *pgrp;
852 	struct proc *p;
853 	int i;
854 
855 	for (i = 0; i <= pgrphash; i++) {
856 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
857 			printf("\tindx %d\n", i);
858 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
859 				printf(
860 			"\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
861 				    (void *)pgrp, (long)pgrp->pg_id,
862 				    (void *)pgrp->pg_session,
863 				    pgrp->pg_session->s_count,
864 				    (void *)LIST_FIRST(&pgrp->pg_members));
865 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
866 					printf("\t\tpid %ld addr %p pgrp %p\n",
867 					    (long)p->p_pid, (void *)p,
868 					    (void *)p->p_pgrp);
869 				}
870 			}
871 		}
872 	}
873 }
874 #endif /* DDB */
875 
876 /*
877  * Calculate the kinfo_proc members which contain process-wide
878  * informations.
879  * Must be called with the target process locked.
880  */
881 static void
882 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
883 {
884 	struct thread *td;
885 
886 	PROC_LOCK_ASSERT(p, MA_OWNED);
887 
888 	kp->ki_estcpu = 0;
889 	kp->ki_pctcpu = 0;
890 	FOREACH_THREAD_IN_PROC(p, td) {
891 		thread_lock(td);
892 		kp->ki_pctcpu += sched_pctcpu(td);
893 		kp->ki_estcpu += sched_estcpu(td);
894 		thread_unlock(td);
895 	}
896 }
897 
898 /*
899  * Clear kinfo_proc and fill in any information that is common
900  * to all threads in the process.
901  * Must be called with the target process locked.
902  */
903 static void
904 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
905 {
906 	struct thread *td0;
907 	struct tty *tp;
908 	struct session *sp;
909 	struct ucred *cred;
910 	struct sigacts *ps;
911 	struct timeval boottime;
912 
913 	/* For proc_realparent. */
914 	sx_assert(&proctree_lock, SX_LOCKED);
915 	PROC_LOCK_ASSERT(p, MA_OWNED);
916 	bzero(kp, sizeof(*kp));
917 
918 	kp->ki_structsize = sizeof(*kp);
919 	kp->ki_paddr = p;
920 	kp->ki_addr =/* p->p_addr; */0; /* XXX */
921 	kp->ki_args = p->p_args;
922 	kp->ki_textvp = p->p_textvp;
923 #ifdef KTRACE
924 	kp->ki_tracep = p->p_tracevp;
925 	kp->ki_traceflag = p->p_traceflag;
926 #endif
927 	kp->ki_fd = p->p_fd;
928 	kp->ki_vmspace = p->p_vmspace;
929 	kp->ki_flag = p->p_flag;
930 	kp->ki_flag2 = p->p_flag2;
931 	cred = p->p_ucred;
932 	if (cred) {
933 		kp->ki_uid = cred->cr_uid;
934 		kp->ki_ruid = cred->cr_ruid;
935 		kp->ki_svuid = cred->cr_svuid;
936 		kp->ki_cr_flags = 0;
937 		if (cred->cr_flags & CRED_FLAG_CAPMODE)
938 			kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
939 		/* XXX bde doesn't like KI_NGROUPS */
940 		if (cred->cr_ngroups > KI_NGROUPS) {
941 			kp->ki_ngroups = KI_NGROUPS;
942 			kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
943 		} else
944 			kp->ki_ngroups = cred->cr_ngroups;
945 		bcopy(cred->cr_groups, kp->ki_groups,
946 		    kp->ki_ngroups * sizeof(gid_t));
947 		kp->ki_rgid = cred->cr_rgid;
948 		kp->ki_svgid = cred->cr_svgid;
949 		/* If jailed(cred), emulate the old P_JAILED flag. */
950 		if (jailed(cred)) {
951 			kp->ki_flag |= P_JAILED;
952 			/* If inside the jail, use 0 as a jail ID. */
953 			if (cred->cr_prison != curthread->td_ucred->cr_prison)
954 				kp->ki_jid = cred->cr_prison->pr_id;
955 		}
956 		strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
957 		    sizeof(kp->ki_loginclass));
958 	}
959 	ps = p->p_sigacts;
960 	if (ps) {
961 		mtx_lock(&ps->ps_mtx);
962 		kp->ki_sigignore = ps->ps_sigignore;
963 		kp->ki_sigcatch = ps->ps_sigcatch;
964 		mtx_unlock(&ps->ps_mtx);
965 	}
966 	if (p->p_state != PRS_NEW &&
967 	    p->p_state != PRS_ZOMBIE &&
968 	    p->p_vmspace != NULL) {
969 		struct vmspace *vm = p->p_vmspace;
970 
971 		kp->ki_size = vm->vm_map.size;
972 		kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
973 		FOREACH_THREAD_IN_PROC(p, td0) {
974 			if (!TD_IS_SWAPPED(td0))
975 				kp->ki_rssize += td0->td_kstack_pages;
976 		}
977 		kp->ki_swrss = vm->vm_swrss;
978 		kp->ki_tsize = vm->vm_tsize;
979 		kp->ki_dsize = vm->vm_dsize;
980 		kp->ki_ssize = vm->vm_ssize;
981 	} else if (p->p_state == PRS_ZOMBIE)
982 		kp->ki_stat = SZOMB;
983 	if (kp->ki_flag & P_INMEM)
984 		kp->ki_sflag = PS_INMEM;
985 	else
986 		kp->ki_sflag = 0;
987 	/* Calculate legacy swtime as seconds since 'swtick'. */
988 	kp->ki_swtime = (ticks - p->p_swtick) / hz;
989 	kp->ki_pid = p->p_pid;
990 	kp->ki_nice = p->p_nice;
991 	kp->ki_fibnum = p->p_fibnum;
992 	kp->ki_start = p->p_stats->p_start;
993 	getboottime(&boottime);
994 	timevaladd(&kp->ki_start, &boottime);
995 	PROC_STATLOCK(p);
996 	rufetch(p, &kp->ki_rusage);
997 	kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
998 	calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
999 	PROC_STATUNLOCK(p);
1000 	calccru(p, &kp->ki_childutime, &kp->ki_childstime);
1001 	/* Some callers want child times in a single value. */
1002 	kp->ki_childtime = kp->ki_childstime;
1003 	timevaladd(&kp->ki_childtime, &kp->ki_childutime);
1004 
1005 	FOREACH_THREAD_IN_PROC(p, td0)
1006 		kp->ki_cow += td0->td_cow;
1007 
1008 	tp = NULL;
1009 	if (p->p_pgrp) {
1010 		kp->ki_pgid = p->p_pgrp->pg_id;
1011 		kp->ki_jobc = p->p_pgrp->pg_jobc;
1012 		sp = p->p_pgrp->pg_session;
1013 
1014 		if (sp != NULL) {
1015 			kp->ki_sid = sp->s_sid;
1016 			SESS_LOCK(sp);
1017 			strlcpy(kp->ki_login, sp->s_login,
1018 			    sizeof(kp->ki_login));
1019 			if (sp->s_ttyvp)
1020 				kp->ki_kiflag |= KI_CTTY;
1021 			if (SESS_LEADER(p))
1022 				kp->ki_kiflag |= KI_SLEADER;
1023 			/* XXX proctree_lock */
1024 			tp = sp->s_ttyp;
1025 			SESS_UNLOCK(sp);
1026 		}
1027 	}
1028 	if ((p->p_flag & P_CONTROLT) && tp != NULL) {
1029 		kp->ki_tdev = tty_udev(tp);
1030 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1031 		kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
1032 		if (tp->t_session)
1033 			kp->ki_tsid = tp->t_session->s_sid;
1034 	} else {
1035 		kp->ki_tdev = NODEV;
1036 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1037 	}
1038 	if (p->p_comm[0] != '\0')
1039 		strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
1040 	if (p->p_sysent && p->p_sysent->sv_name != NULL &&
1041 	    p->p_sysent->sv_name[0] != '\0')
1042 		strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
1043 	kp->ki_siglist = p->p_siglist;
1044 	kp->ki_xstat = KW_EXITCODE(p->p_xexit, p->p_xsig);
1045 	kp->ki_acflag = p->p_acflag;
1046 	kp->ki_lock = p->p_lock;
1047 	if (p->p_pptr) {
1048 		kp->ki_ppid = proc_realparent(p)->p_pid;
1049 		if (p->p_flag & P_TRACED)
1050 			kp->ki_tracer = p->p_pptr->p_pid;
1051 	}
1052 }
1053 
1054 /*
1055  * Fill in information that is thread specific.  Must be called with
1056  * target process locked.  If 'preferthread' is set, overwrite certain
1057  * process-related fields that are maintained for both threads and
1058  * processes.
1059  */
1060 static void
1061 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
1062 {
1063 	struct proc *p;
1064 
1065 	p = td->td_proc;
1066 	kp->ki_tdaddr = td;
1067 	PROC_LOCK_ASSERT(p, MA_OWNED);
1068 
1069 	if (preferthread)
1070 		PROC_STATLOCK(p);
1071 	thread_lock(td);
1072 	if (td->td_wmesg != NULL)
1073 		strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
1074 	else
1075 		bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
1076 	if (strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname)) >=
1077 	    sizeof(kp->ki_tdname)) {
1078 		strlcpy(kp->ki_moretdname,
1079 		    td->td_name + sizeof(kp->ki_tdname) - 1,
1080 		    sizeof(kp->ki_moretdname));
1081 	} else {
1082 		bzero(kp->ki_moretdname, sizeof(kp->ki_moretdname));
1083 	}
1084 	if (TD_ON_LOCK(td)) {
1085 		kp->ki_kiflag |= KI_LOCKBLOCK;
1086 		strlcpy(kp->ki_lockname, td->td_lockname,
1087 		    sizeof(kp->ki_lockname));
1088 	} else {
1089 		kp->ki_kiflag &= ~KI_LOCKBLOCK;
1090 		bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
1091 	}
1092 
1093 	if (p->p_state == PRS_NORMAL) { /* approximate. */
1094 		if (TD_ON_RUNQ(td) ||
1095 		    TD_CAN_RUN(td) ||
1096 		    TD_IS_RUNNING(td)) {
1097 			kp->ki_stat = SRUN;
1098 		} else if (P_SHOULDSTOP(p)) {
1099 			kp->ki_stat = SSTOP;
1100 		} else if (TD_IS_SLEEPING(td)) {
1101 			kp->ki_stat = SSLEEP;
1102 		} else if (TD_ON_LOCK(td)) {
1103 			kp->ki_stat = SLOCK;
1104 		} else {
1105 			kp->ki_stat = SWAIT;
1106 		}
1107 	} else if (p->p_state == PRS_ZOMBIE) {
1108 		kp->ki_stat = SZOMB;
1109 	} else {
1110 		kp->ki_stat = SIDL;
1111 	}
1112 
1113 	/* Things in the thread */
1114 	kp->ki_wchan = td->td_wchan;
1115 	kp->ki_pri.pri_level = td->td_priority;
1116 	kp->ki_pri.pri_native = td->td_base_pri;
1117 
1118 	/*
1119 	 * Note: legacy fields; clamp at the old NOCPU value and/or
1120 	 * the maximum u_char CPU value.
1121 	 */
1122 	if (td->td_lastcpu == NOCPU)
1123 		kp->ki_lastcpu_old = NOCPU_OLD;
1124 	else if (td->td_lastcpu > MAXCPU_OLD)
1125 		kp->ki_lastcpu_old = MAXCPU_OLD;
1126 	else
1127 		kp->ki_lastcpu_old = td->td_lastcpu;
1128 
1129 	if (td->td_oncpu == NOCPU)
1130 		kp->ki_oncpu_old = NOCPU_OLD;
1131 	else if (td->td_oncpu > MAXCPU_OLD)
1132 		kp->ki_oncpu_old = MAXCPU_OLD;
1133 	else
1134 		kp->ki_oncpu_old = td->td_oncpu;
1135 
1136 	kp->ki_lastcpu = td->td_lastcpu;
1137 	kp->ki_oncpu = td->td_oncpu;
1138 	kp->ki_tdflags = td->td_flags;
1139 	kp->ki_tid = td->td_tid;
1140 	kp->ki_numthreads = p->p_numthreads;
1141 	kp->ki_pcb = td->td_pcb;
1142 	kp->ki_kstack = (void *)td->td_kstack;
1143 	kp->ki_slptime = (ticks - td->td_slptick) / hz;
1144 	kp->ki_pri.pri_class = td->td_pri_class;
1145 	kp->ki_pri.pri_user = td->td_user_pri;
1146 
1147 	if (preferthread) {
1148 		rufetchtd(td, &kp->ki_rusage);
1149 		kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
1150 		kp->ki_pctcpu = sched_pctcpu(td);
1151 		kp->ki_estcpu = sched_estcpu(td);
1152 		kp->ki_cow = td->td_cow;
1153 	}
1154 
1155 	/* We can't get this anymore but ps etc never used it anyway. */
1156 	kp->ki_rqindex = 0;
1157 
1158 	if (preferthread)
1159 		kp->ki_siglist = td->td_siglist;
1160 	kp->ki_sigmask = td->td_sigmask;
1161 	thread_unlock(td);
1162 	if (preferthread)
1163 		PROC_STATUNLOCK(p);
1164 }
1165 
1166 /*
1167  * Fill in a kinfo_proc structure for the specified process.
1168  * Must be called with the target process locked.
1169  */
1170 void
1171 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1172 {
1173 
1174 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1175 
1176 	fill_kinfo_proc_only(p, kp);
1177 	fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1178 	fill_kinfo_aggregate(p, kp);
1179 }
1180 
1181 struct pstats *
1182 pstats_alloc(void)
1183 {
1184 
1185 	return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1186 }
1187 
1188 /*
1189  * Copy parts of p_stats; zero the rest of p_stats (statistics).
1190  */
1191 void
1192 pstats_fork(struct pstats *src, struct pstats *dst)
1193 {
1194 
1195 	bzero(&dst->pstat_startzero,
1196 	    __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1197 	bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1198 	    __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1199 }
1200 
1201 void
1202 pstats_free(struct pstats *ps)
1203 {
1204 
1205 	free(ps, M_SUBPROC);
1206 }
1207 
1208 static struct proc *
1209 zpfind_locked(pid_t pid)
1210 {
1211 	struct proc *p;
1212 
1213 	sx_assert(&allproc_lock, SX_LOCKED);
1214 	LIST_FOREACH(p, &zombproc, p_list) {
1215 		if (p->p_pid == pid) {
1216 			PROC_LOCK(p);
1217 			break;
1218 		}
1219 	}
1220 	return (p);
1221 }
1222 
1223 /*
1224  * Locate a zombie process by number
1225  */
1226 struct proc *
1227 zpfind(pid_t pid)
1228 {
1229 	struct proc *p;
1230 
1231 	sx_slock(&allproc_lock);
1232 	p = zpfind_locked(pid);
1233 	sx_sunlock(&allproc_lock);
1234 	return (p);
1235 }
1236 
1237 #ifdef COMPAT_FREEBSD32
1238 
1239 /*
1240  * This function is typically used to copy out the kernel address, so
1241  * it can be replaced by assignment of zero.
1242  */
1243 static inline uint32_t
1244 ptr32_trim(void *ptr)
1245 {
1246 	uintptr_t uptr;
1247 
1248 	uptr = (uintptr_t)ptr;
1249 	return ((uptr > UINT_MAX) ? 0 : uptr);
1250 }
1251 
1252 #define PTRTRIM_CP(src,dst,fld) \
1253 	do { (dst).fld = ptr32_trim((src).fld); } while (0)
1254 
1255 static void
1256 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1257 {
1258 	int i;
1259 
1260 	bzero(ki32, sizeof(struct kinfo_proc32));
1261 	ki32->ki_structsize = sizeof(struct kinfo_proc32);
1262 	CP(*ki, *ki32, ki_layout);
1263 	PTRTRIM_CP(*ki, *ki32, ki_args);
1264 	PTRTRIM_CP(*ki, *ki32, ki_paddr);
1265 	PTRTRIM_CP(*ki, *ki32, ki_addr);
1266 	PTRTRIM_CP(*ki, *ki32, ki_tracep);
1267 	PTRTRIM_CP(*ki, *ki32, ki_textvp);
1268 	PTRTRIM_CP(*ki, *ki32, ki_fd);
1269 	PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1270 	PTRTRIM_CP(*ki, *ki32, ki_wchan);
1271 	CP(*ki, *ki32, ki_pid);
1272 	CP(*ki, *ki32, ki_ppid);
1273 	CP(*ki, *ki32, ki_pgid);
1274 	CP(*ki, *ki32, ki_tpgid);
1275 	CP(*ki, *ki32, ki_sid);
1276 	CP(*ki, *ki32, ki_tsid);
1277 	CP(*ki, *ki32, ki_jobc);
1278 	CP(*ki, *ki32, ki_tdev);
1279 	CP(*ki, *ki32, ki_tdev_freebsd11);
1280 	CP(*ki, *ki32, ki_siglist);
1281 	CP(*ki, *ki32, ki_sigmask);
1282 	CP(*ki, *ki32, ki_sigignore);
1283 	CP(*ki, *ki32, ki_sigcatch);
1284 	CP(*ki, *ki32, ki_uid);
1285 	CP(*ki, *ki32, ki_ruid);
1286 	CP(*ki, *ki32, ki_svuid);
1287 	CP(*ki, *ki32, ki_rgid);
1288 	CP(*ki, *ki32, ki_svgid);
1289 	CP(*ki, *ki32, ki_ngroups);
1290 	for (i = 0; i < KI_NGROUPS; i++)
1291 		CP(*ki, *ki32, ki_groups[i]);
1292 	CP(*ki, *ki32, ki_size);
1293 	CP(*ki, *ki32, ki_rssize);
1294 	CP(*ki, *ki32, ki_swrss);
1295 	CP(*ki, *ki32, ki_tsize);
1296 	CP(*ki, *ki32, ki_dsize);
1297 	CP(*ki, *ki32, ki_ssize);
1298 	CP(*ki, *ki32, ki_xstat);
1299 	CP(*ki, *ki32, ki_acflag);
1300 	CP(*ki, *ki32, ki_pctcpu);
1301 	CP(*ki, *ki32, ki_estcpu);
1302 	CP(*ki, *ki32, ki_slptime);
1303 	CP(*ki, *ki32, ki_swtime);
1304 	CP(*ki, *ki32, ki_cow);
1305 	CP(*ki, *ki32, ki_runtime);
1306 	TV_CP(*ki, *ki32, ki_start);
1307 	TV_CP(*ki, *ki32, ki_childtime);
1308 	CP(*ki, *ki32, ki_flag);
1309 	CP(*ki, *ki32, ki_kiflag);
1310 	CP(*ki, *ki32, ki_traceflag);
1311 	CP(*ki, *ki32, ki_stat);
1312 	CP(*ki, *ki32, ki_nice);
1313 	CP(*ki, *ki32, ki_lock);
1314 	CP(*ki, *ki32, ki_rqindex);
1315 	CP(*ki, *ki32, ki_oncpu);
1316 	CP(*ki, *ki32, ki_lastcpu);
1317 
1318 	/* XXX TODO: wrap cpu value as appropriate */
1319 	CP(*ki, *ki32, ki_oncpu_old);
1320 	CP(*ki, *ki32, ki_lastcpu_old);
1321 
1322 	bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1323 	bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1324 	bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1325 	bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1326 	bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1327 	bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1328 	bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1329 	bcopy(ki->ki_moretdname, ki32->ki_moretdname, MAXCOMLEN - TDNAMLEN + 1);
1330 	CP(*ki, *ki32, ki_tracer);
1331 	CP(*ki, *ki32, ki_flag2);
1332 	CP(*ki, *ki32, ki_fibnum);
1333 	CP(*ki, *ki32, ki_cr_flags);
1334 	CP(*ki, *ki32, ki_jid);
1335 	CP(*ki, *ki32, ki_numthreads);
1336 	CP(*ki, *ki32, ki_tid);
1337 	CP(*ki, *ki32, ki_pri);
1338 	freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1339 	freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1340 	PTRTRIM_CP(*ki, *ki32, ki_pcb);
1341 	PTRTRIM_CP(*ki, *ki32, ki_kstack);
1342 	PTRTRIM_CP(*ki, *ki32, ki_udata);
1343 	PTRTRIM_CP(*ki, *ki32, ki_tdaddr);
1344 	CP(*ki, *ki32, ki_sflag);
1345 	CP(*ki, *ki32, ki_tdflags);
1346 }
1347 #endif
1348 
1349 static ssize_t
1350 kern_proc_out_size(struct proc *p, int flags)
1351 {
1352 	ssize_t size = 0;
1353 
1354 	PROC_LOCK_ASSERT(p, MA_OWNED);
1355 
1356 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1357 #ifdef COMPAT_FREEBSD32
1358 		if ((flags & KERN_PROC_MASK32) != 0) {
1359 			size += sizeof(struct kinfo_proc32);
1360 		} else
1361 #endif
1362 			size += sizeof(struct kinfo_proc);
1363 	} else {
1364 #ifdef COMPAT_FREEBSD32
1365 		if ((flags & KERN_PROC_MASK32) != 0)
1366 			size += sizeof(struct kinfo_proc32) * p->p_numthreads;
1367 		else
1368 #endif
1369 			size += sizeof(struct kinfo_proc) * p->p_numthreads;
1370 	}
1371 	PROC_UNLOCK(p);
1372 	return (size);
1373 }
1374 
1375 int
1376 kern_proc_out(struct proc *p, struct sbuf *sb, int flags)
1377 {
1378 	struct thread *td;
1379 	struct kinfo_proc ki;
1380 #ifdef COMPAT_FREEBSD32
1381 	struct kinfo_proc32 ki32;
1382 #endif
1383 	int error;
1384 
1385 	PROC_LOCK_ASSERT(p, MA_OWNED);
1386 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1387 
1388 	error = 0;
1389 	fill_kinfo_proc(p, &ki);
1390 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1391 #ifdef COMPAT_FREEBSD32
1392 		if ((flags & KERN_PROC_MASK32) != 0) {
1393 			freebsd32_kinfo_proc_out(&ki, &ki32);
1394 			if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1395 				error = ENOMEM;
1396 		} else
1397 #endif
1398 			if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1399 				error = ENOMEM;
1400 	} else {
1401 		FOREACH_THREAD_IN_PROC(p, td) {
1402 			fill_kinfo_thread(td, &ki, 1);
1403 #ifdef COMPAT_FREEBSD32
1404 			if ((flags & KERN_PROC_MASK32) != 0) {
1405 				freebsd32_kinfo_proc_out(&ki, &ki32);
1406 				if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1407 					error = ENOMEM;
1408 			} else
1409 #endif
1410 				if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1411 					error = ENOMEM;
1412 			if (error != 0)
1413 				break;
1414 		}
1415 	}
1416 	PROC_UNLOCK(p);
1417 	return (error);
1418 }
1419 
1420 static int
1421 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
1422 {
1423 	struct sbuf sb;
1424 	struct kinfo_proc ki;
1425 	int error, error2;
1426 
1427 	if (req->oldptr == NULL)
1428 		return (SYSCTL_OUT(req, 0, kern_proc_out_size(p, flags)));
1429 
1430 	sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req);
1431 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1432 	error = kern_proc_out(p, &sb, flags);
1433 	error2 = sbuf_finish(&sb);
1434 	sbuf_delete(&sb);
1435 	if (error != 0)
1436 		return (error);
1437 	else if (error2 != 0)
1438 		return (error2);
1439 	return (0);
1440 }
1441 
1442 static int
1443 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1444 {
1445 	int *name = (int *)arg1;
1446 	u_int namelen = arg2;
1447 	struct proc *p;
1448 	int flags, doingzomb, oid_number;
1449 	int error = 0;
1450 
1451 	oid_number = oidp->oid_number;
1452 	if (oid_number != KERN_PROC_ALL &&
1453 	    (oid_number & KERN_PROC_INC_THREAD) == 0)
1454 		flags = KERN_PROC_NOTHREADS;
1455 	else {
1456 		flags = 0;
1457 		oid_number &= ~KERN_PROC_INC_THREAD;
1458 	}
1459 #ifdef COMPAT_FREEBSD32
1460 	if (req->flags & SCTL_MASK32)
1461 		flags |= KERN_PROC_MASK32;
1462 #endif
1463 	if (oid_number == KERN_PROC_PID) {
1464 		if (namelen != 1)
1465 			return (EINVAL);
1466 		error = sysctl_wire_old_buffer(req, 0);
1467 		if (error)
1468 			return (error);
1469 		sx_slock(&proctree_lock);
1470 		error = pget((pid_t)name[0], PGET_CANSEE, &p);
1471 		if (error == 0)
1472 			error = sysctl_out_proc(p, req, flags);
1473 		sx_sunlock(&proctree_lock);
1474 		return (error);
1475 	}
1476 
1477 	switch (oid_number) {
1478 	case KERN_PROC_ALL:
1479 		if (namelen != 0)
1480 			return (EINVAL);
1481 		break;
1482 	case KERN_PROC_PROC:
1483 		if (namelen != 0 && namelen != 1)
1484 			return (EINVAL);
1485 		break;
1486 	default:
1487 		if (namelen != 1)
1488 			return (EINVAL);
1489 		break;
1490 	}
1491 
1492 	if (req->oldptr == NULL) {
1493 		/* overestimate by 5 procs */
1494 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1495 		if (error)
1496 			return (error);
1497 	} else {
1498 		error = sysctl_wire_old_buffer(req, 0);
1499 		if (error != 0)
1500 			return (error);
1501 		/*
1502 		 * This lock is only needed to safely grab the parent of a
1503 		 * traced process. Only grab it if we are producing any
1504 		 * data to begin with.
1505 		 */
1506 		sx_slock(&proctree_lock);
1507 	}
1508 	sx_slock(&allproc_lock);
1509 	for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) {
1510 		if (!doingzomb)
1511 			p = LIST_FIRST(&allproc);
1512 		else
1513 			p = LIST_FIRST(&zombproc);
1514 		for (; p != NULL; p = LIST_NEXT(p, p_list)) {
1515 			/*
1516 			 * Skip embryonic processes.
1517 			 */
1518 			if (p->p_state == PRS_NEW)
1519 				continue;
1520 			PROC_LOCK(p);
1521 			KASSERT(p->p_ucred != NULL,
1522 			    ("process credential is NULL for non-NEW proc"));
1523 			/*
1524 			 * Show a user only appropriate processes.
1525 			 */
1526 			if (p_cansee(curthread, p)) {
1527 				PROC_UNLOCK(p);
1528 				continue;
1529 			}
1530 			/*
1531 			 * TODO - make more efficient (see notes below).
1532 			 * do by session.
1533 			 */
1534 			switch (oid_number) {
1535 
1536 			case KERN_PROC_GID:
1537 				if (p->p_ucred->cr_gid != (gid_t)name[0]) {
1538 					PROC_UNLOCK(p);
1539 					continue;
1540 				}
1541 				break;
1542 
1543 			case KERN_PROC_PGRP:
1544 				/* could do this by traversing pgrp */
1545 				if (p->p_pgrp == NULL ||
1546 				    p->p_pgrp->pg_id != (pid_t)name[0]) {
1547 					PROC_UNLOCK(p);
1548 					continue;
1549 				}
1550 				break;
1551 
1552 			case KERN_PROC_RGID:
1553 				if (p->p_ucred->cr_rgid != (gid_t)name[0]) {
1554 					PROC_UNLOCK(p);
1555 					continue;
1556 				}
1557 				break;
1558 
1559 			case KERN_PROC_SESSION:
1560 				if (p->p_session == NULL ||
1561 				    p->p_session->s_sid != (pid_t)name[0]) {
1562 					PROC_UNLOCK(p);
1563 					continue;
1564 				}
1565 				break;
1566 
1567 			case KERN_PROC_TTY:
1568 				if ((p->p_flag & P_CONTROLT) == 0 ||
1569 				    p->p_session == NULL) {
1570 					PROC_UNLOCK(p);
1571 					continue;
1572 				}
1573 				/* XXX proctree_lock */
1574 				SESS_LOCK(p->p_session);
1575 				if (p->p_session->s_ttyp == NULL ||
1576 				    tty_udev(p->p_session->s_ttyp) !=
1577 				    (dev_t)name[0]) {
1578 					SESS_UNLOCK(p->p_session);
1579 					PROC_UNLOCK(p);
1580 					continue;
1581 				}
1582 				SESS_UNLOCK(p->p_session);
1583 				break;
1584 
1585 			case KERN_PROC_UID:
1586 				if (p->p_ucred->cr_uid != (uid_t)name[0]) {
1587 					PROC_UNLOCK(p);
1588 					continue;
1589 				}
1590 				break;
1591 
1592 			case KERN_PROC_RUID:
1593 				if (p->p_ucred->cr_ruid != (uid_t)name[0]) {
1594 					PROC_UNLOCK(p);
1595 					continue;
1596 				}
1597 				break;
1598 
1599 			case KERN_PROC_PROC:
1600 				break;
1601 
1602 			default:
1603 				break;
1604 
1605 			}
1606 
1607 			error = sysctl_out_proc(p, req, flags);
1608 			if (error)
1609 				goto out;
1610 		}
1611 	}
1612 out:
1613 	sx_sunlock(&allproc_lock);
1614 	if (req->oldptr != NULL)
1615 		sx_sunlock(&proctree_lock);
1616 	return (error);
1617 }
1618 
1619 struct pargs *
1620 pargs_alloc(int len)
1621 {
1622 	struct pargs *pa;
1623 
1624 	pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1625 		M_WAITOK);
1626 	refcount_init(&pa->ar_ref, 1);
1627 	pa->ar_length = len;
1628 	return (pa);
1629 }
1630 
1631 static void
1632 pargs_free(struct pargs *pa)
1633 {
1634 
1635 	free(pa, M_PARGS);
1636 }
1637 
1638 void
1639 pargs_hold(struct pargs *pa)
1640 {
1641 
1642 	if (pa == NULL)
1643 		return;
1644 	refcount_acquire(&pa->ar_ref);
1645 }
1646 
1647 void
1648 pargs_drop(struct pargs *pa)
1649 {
1650 
1651 	if (pa == NULL)
1652 		return;
1653 	if (refcount_release(&pa->ar_ref))
1654 		pargs_free(pa);
1655 }
1656 
1657 static int
1658 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1659     size_t len)
1660 {
1661 	ssize_t n;
1662 
1663 	/*
1664 	 * This may return a short read if the string is shorter than the chunk
1665 	 * and is aligned at the end of the page, and the following page is not
1666 	 * mapped.
1667 	 */
1668 	n = proc_readmem(td, p, (vm_offset_t)sptr, buf, len);
1669 	if (n <= 0)
1670 		return (ENOMEM);
1671 	return (0);
1672 }
1673 
1674 #define PROC_AUXV_MAX	256	/* Safety limit on auxv size. */
1675 
1676 enum proc_vector_type {
1677 	PROC_ARG,
1678 	PROC_ENV,
1679 	PROC_AUX,
1680 };
1681 
1682 #ifdef COMPAT_FREEBSD32
1683 static int
1684 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1685     size_t *vsizep, enum proc_vector_type type)
1686 {
1687 	struct freebsd32_ps_strings pss;
1688 	Elf32_Auxinfo aux;
1689 	vm_offset_t vptr, ptr;
1690 	uint32_t *proc_vector32;
1691 	char **proc_vector;
1692 	size_t vsize, size;
1693 	int i, error;
1694 
1695 	error = 0;
1696 	if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss,
1697 	    sizeof(pss)) != sizeof(pss))
1698 		return (ENOMEM);
1699 	switch (type) {
1700 	case PROC_ARG:
1701 		vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1702 		vsize = pss.ps_nargvstr;
1703 		if (vsize > ARG_MAX)
1704 			return (ENOEXEC);
1705 		size = vsize * sizeof(int32_t);
1706 		break;
1707 	case PROC_ENV:
1708 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1709 		vsize = pss.ps_nenvstr;
1710 		if (vsize > ARG_MAX)
1711 			return (ENOEXEC);
1712 		size = vsize * sizeof(int32_t);
1713 		break;
1714 	case PROC_AUX:
1715 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1716 		    (pss.ps_nenvstr + 1) * sizeof(int32_t);
1717 		if (vptr % 4 != 0)
1718 			return (ENOEXEC);
1719 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1720 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1721 			    sizeof(aux))
1722 				return (ENOMEM);
1723 			if (aux.a_type == AT_NULL)
1724 				break;
1725 			ptr += sizeof(aux);
1726 		}
1727 		if (aux.a_type != AT_NULL)
1728 			return (ENOEXEC);
1729 		vsize = i + 1;
1730 		size = vsize * sizeof(aux);
1731 		break;
1732 	default:
1733 		KASSERT(0, ("Wrong proc vector type: %d", type));
1734 		return (EINVAL);
1735 	}
1736 	proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1737 	if (proc_readmem(td, p, vptr, proc_vector32, size) != size) {
1738 		error = ENOMEM;
1739 		goto done;
1740 	}
1741 	if (type == PROC_AUX) {
1742 		*proc_vectorp = (char **)proc_vector32;
1743 		*vsizep = vsize;
1744 		return (0);
1745 	}
1746 	proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1747 	for (i = 0; i < (int)vsize; i++)
1748 		proc_vector[i] = PTRIN(proc_vector32[i]);
1749 	*proc_vectorp = proc_vector;
1750 	*vsizep = vsize;
1751 done:
1752 	free(proc_vector32, M_TEMP);
1753 	return (error);
1754 }
1755 #endif
1756 
1757 static int
1758 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1759     size_t *vsizep, enum proc_vector_type type)
1760 {
1761 	struct ps_strings pss;
1762 	Elf_Auxinfo aux;
1763 	vm_offset_t vptr, ptr;
1764 	char **proc_vector;
1765 	size_t vsize, size;
1766 	int i;
1767 
1768 #ifdef COMPAT_FREEBSD32
1769 	if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1770 		return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1771 #endif
1772 	if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss,
1773 	    sizeof(pss)) != sizeof(pss))
1774 		return (ENOMEM);
1775 	switch (type) {
1776 	case PROC_ARG:
1777 		vptr = (vm_offset_t)pss.ps_argvstr;
1778 		vsize = pss.ps_nargvstr;
1779 		if (vsize > ARG_MAX)
1780 			return (ENOEXEC);
1781 		size = vsize * sizeof(char *);
1782 		break;
1783 	case PROC_ENV:
1784 		vptr = (vm_offset_t)pss.ps_envstr;
1785 		vsize = pss.ps_nenvstr;
1786 		if (vsize > ARG_MAX)
1787 			return (ENOEXEC);
1788 		size = vsize * sizeof(char *);
1789 		break;
1790 	case PROC_AUX:
1791 		/*
1792 		 * The aux array is just above env array on the stack. Check
1793 		 * that the address is naturally aligned.
1794 		 */
1795 		vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1796 		    * sizeof(char *);
1797 #if __ELF_WORD_SIZE == 64
1798 		if (vptr % sizeof(uint64_t) != 0)
1799 #else
1800 		if (vptr % sizeof(uint32_t) != 0)
1801 #endif
1802 			return (ENOEXEC);
1803 		/*
1804 		 * We count the array size reading the aux vectors from the
1805 		 * stack until AT_NULL vector is returned.  So (to keep the code
1806 		 * simple) we read the process stack twice: the first time here
1807 		 * to find the size and the second time when copying the vectors
1808 		 * to the allocated proc_vector.
1809 		 */
1810 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1811 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1812 			    sizeof(aux))
1813 				return (ENOMEM);
1814 			if (aux.a_type == AT_NULL)
1815 				break;
1816 			ptr += sizeof(aux);
1817 		}
1818 		/*
1819 		 * If the PROC_AUXV_MAX entries are iterated over, and we have
1820 		 * not reached AT_NULL, it is most likely we are reading wrong
1821 		 * data: either the process doesn't have auxv array or data has
1822 		 * been modified. Return the error in this case.
1823 		 */
1824 		if (aux.a_type != AT_NULL)
1825 			return (ENOEXEC);
1826 		vsize = i + 1;
1827 		size = vsize * sizeof(aux);
1828 		break;
1829 	default:
1830 		KASSERT(0, ("Wrong proc vector type: %d", type));
1831 		return (EINVAL); /* In case we are built without INVARIANTS. */
1832 	}
1833 	proc_vector = malloc(size, M_TEMP, M_WAITOK);
1834 	if (proc_readmem(td, p, vptr, proc_vector, size) != size) {
1835 		free(proc_vector, M_TEMP);
1836 		return (ENOMEM);
1837 	}
1838 	*proc_vectorp = proc_vector;
1839 	*vsizep = vsize;
1840 
1841 	return (0);
1842 }
1843 
1844 #define GET_PS_STRINGS_CHUNK_SZ	256	/* Chunk size (bytes) for ps_strings operations. */
1845 
1846 static int
1847 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
1848     enum proc_vector_type type)
1849 {
1850 	size_t done, len, nchr, vsize;
1851 	int error, i;
1852 	char **proc_vector, *sptr;
1853 	char pss_string[GET_PS_STRINGS_CHUNK_SZ];
1854 
1855 	PROC_ASSERT_HELD(p);
1856 
1857 	/*
1858 	 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
1859 	 */
1860 	nchr = 2 * (PATH_MAX + ARG_MAX);
1861 
1862 	error = get_proc_vector(td, p, &proc_vector, &vsize, type);
1863 	if (error != 0)
1864 		return (error);
1865 	for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
1866 		/*
1867 		 * The program may have scribbled into its argv array, e.g. to
1868 		 * remove some arguments.  If that has happened, break out
1869 		 * before trying to read from NULL.
1870 		 */
1871 		if (proc_vector[i] == NULL)
1872 			break;
1873 		for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
1874 			error = proc_read_string(td, p, sptr, pss_string,
1875 			    sizeof(pss_string));
1876 			if (error != 0)
1877 				goto done;
1878 			len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
1879 			if (done + len >= nchr)
1880 				len = nchr - done - 1;
1881 			sbuf_bcat(sb, pss_string, len);
1882 			if (len != GET_PS_STRINGS_CHUNK_SZ)
1883 				break;
1884 			done += GET_PS_STRINGS_CHUNK_SZ;
1885 		}
1886 		sbuf_bcat(sb, "", 1);
1887 		done += len + 1;
1888 	}
1889 done:
1890 	free(proc_vector, M_TEMP);
1891 	return (error);
1892 }
1893 
1894 int
1895 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
1896 {
1897 
1898 	return (get_ps_strings(curthread, p, sb, PROC_ARG));
1899 }
1900 
1901 int
1902 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
1903 {
1904 
1905 	return (get_ps_strings(curthread, p, sb, PROC_ENV));
1906 }
1907 
1908 int
1909 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb)
1910 {
1911 	size_t vsize, size;
1912 	char **auxv;
1913 	int error;
1914 
1915 	error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX);
1916 	if (error == 0) {
1917 #ifdef COMPAT_FREEBSD32
1918 		if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1919 			size = vsize * sizeof(Elf32_Auxinfo);
1920 		else
1921 #endif
1922 			size = vsize * sizeof(Elf_Auxinfo);
1923 		if (sbuf_bcat(sb, auxv, size) != 0)
1924 			error = ENOMEM;
1925 		free(auxv, M_TEMP);
1926 	}
1927 	return (error);
1928 }
1929 
1930 /*
1931  * This sysctl allows a process to retrieve the argument list or process
1932  * title for another process without groping around in the address space
1933  * of the other process.  It also allow a process to set its own "process
1934  * title to a string of its own choice.
1935  */
1936 static int
1937 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
1938 {
1939 	int *name = (int *)arg1;
1940 	u_int namelen = arg2;
1941 	struct pargs *newpa, *pa;
1942 	struct proc *p;
1943 	struct sbuf sb;
1944 	int flags, error = 0, error2;
1945 	pid_t pid;
1946 
1947 	if (namelen != 1)
1948 		return (EINVAL);
1949 
1950 	pid = (pid_t)name[0];
1951 	/*
1952 	 * If the query is for this process and it is single-threaded, there
1953 	 * is nobody to modify pargs, thus we can just read.
1954 	 */
1955 	p = curproc;
1956 	if (pid == p->p_pid && p->p_numthreads == 1 && req->newptr == NULL &&
1957 	    (pa = p->p_args) != NULL)
1958 		return (SYSCTL_OUT(req, pa->ar_args, pa->ar_length));
1959 
1960 	flags = PGET_CANSEE;
1961 	if (req->newptr != NULL)
1962 		flags |= PGET_ISCURRENT;
1963 	error = pget(pid, flags, &p);
1964 	if (error)
1965 		return (error);
1966 
1967 	pa = p->p_args;
1968 	if (pa != NULL) {
1969 		pargs_hold(pa);
1970 		PROC_UNLOCK(p);
1971 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
1972 		pargs_drop(pa);
1973 	} else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
1974 		_PHOLD(p);
1975 		PROC_UNLOCK(p);
1976 		sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1977 		sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1978 		error = proc_getargv(curthread, p, &sb);
1979 		error2 = sbuf_finish(&sb);
1980 		PRELE(p);
1981 		sbuf_delete(&sb);
1982 		if (error == 0 && error2 != 0)
1983 			error = error2;
1984 	} else {
1985 		PROC_UNLOCK(p);
1986 	}
1987 	if (error != 0 || req->newptr == NULL)
1988 		return (error);
1989 
1990 	if (req->newlen > ps_arg_cache_limit - sizeof(struct pargs))
1991 		return (ENOMEM);
1992 	newpa = pargs_alloc(req->newlen);
1993 	error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
1994 	if (error != 0) {
1995 		pargs_free(newpa);
1996 		return (error);
1997 	}
1998 	PROC_LOCK(p);
1999 	pa = p->p_args;
2000 	p->p_args = newpa;
2001 	PROC_UNLOCK(p);
2002 	pargs_drop(pa);
2003 	return (0);
2004 }
2005 
2006 /*
2007  * This sysctl allows a process to retrieve environment of another process.
2008  */
2009 static int
2010 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
2011 {
2012 	int *name = (int *)arg1;
2013 	u_int namelen = arg2;
2014 	struct proc *p;
2015 	struct sbuf sb;
2016 	int error, error2;
2017 
2018 	if (namelen != 1)
2019 		return (EINVAL);
2020 
2021 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2022 	if (error != 0)
2023 		return (error);
2024 	if ((p->p_flag & P_SYSTEM) != 0) {
2025 		PRELE(p);
2026 		return (0);
2027 	}
2028 
2029 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2030 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2031 	error = proc_getenvv(curthread, p, &sb);
2032 	error2 = sbuf_finish(&sb);
2033 	PRELE(p);
2034 	sbuf_delete(&sb);
2035 	return (error != 0 ? error : error2);
2036 }
2037 
2038 /*
2039  * This sysctl allows a process to retrieve ELF auxiliary vector of
2040  * another process.
2041  */
2042 static int
2043 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
2044 {
2045 	int *name = (int *)arg1;
2046 	u_int namelen = arg2;
2047 	struct proc *p;
2048 	struct sbuf sb;
2049 	int error, error2;
2050 
2051 	if (namelen != 1)
2052 		return (EINVAL);
2053 
2054 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2055 	if (error != 0)
2056 		return (error);
2057 	if ((p->p_flag & P_SYSTEM) != 0) {
2058 		PRELE(p);
2059 		return (0);
2060 	}
2061 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2062 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2063 	error = proc_getauxv(curthread, p, &sb);
2064 	error2 = sbuf_finish(&sb);
2065 	PRELE(p);
2066 	sbuf_delete(&sb);
2067 	return (error != 0 ? error : error2);
2068 }
2069 
2070 /*
2071  * This sysctl allows a process to retrieve the path of the executable for
2072  * itself or another process.
2073  */
2074 static int
2075 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
2076 {
2077 	pid_t *pidp = (pid_t *)arg1;
2078 	unsigned int arglen = arg2;
2079 	struct proc *p;
2080 	struct vnode *vp;
2081 	char *retbuf, *freebuf;
2082 	int error;
2083 
2084 	if (arglen != 1)
2085 		return (EINVAL);
2086 	if (*pidp == -1) {	/* -1 means this process */
2087 		p = req->td->td_proc;
2088 	} else {
2089 		error = pget(*pidp, PGET_CANSEE, &p);
2090 		if (error != 0)
2091 			return (error);
2092 	}
2093 
2094 	vp = p->p_textvp;
2095 	if (vp == NULL) {
2096 		if (*pidp != -1)
2097 			PROC_UNLOCK(p);
2098 		return (0);
2099 	}
2100 	vref(vp);
2101 	if (*pidp != -1)
2102 		PROC_UNLOCK(p);
2103 	error = vn_fullpath(req->td, vp, &retbuf, &freebuf);
2104 	vrele(vp);
2105 	if (error)
2106 		return (error);
2107 	error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
2108 	free(freebuf, M_TEMP);
2109 	return (error);
2110 }
2111 
2112 static int
2113 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
2114 {
2115 	struct proc *p;
2116 	char *sv_name;
2117 	int *name;
2118 	int namelen;
2119 	int error;
2120 
2121 	namelen = arg2;
2122 	if (namelen != 1)
2123 		return (EINVAL);
2124 
2125 	name = (int *)arg1;
2126 	error = pget((pid_t)name[0], PGET_CANSEE, &p);
2127 	if (error != 0)
2128 		return (error);
2129 	sv_name = p->p_sysent->sv_name;
2130 	PROC_UNLOCK(p);
2131 	return (sysctl_handle_string(oidp, sv_name, 0, req));
2132 }
2133 
2134 #ifdef KINFO_OVMENTRY_SIZE
2135 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
2136 #endif
2137 
2138 #ifdef COMPAT_FREEBSD7
2139 static int
2140 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
2141 {
2142 	vm_map_entry_t entry, tmp_entry;
2143 	unsigned int last_timestamp;
2144 	char *fullpath, *freepath;
2145 	struct kinfo_ovmentry *kve;
2146 	struct vattr va;
2147 	struct ucred *cred;
2148 	int error, *name;
2149 	struct vnode *vp;
2150 	struct proc *p;
2151 	vm_map_t map;
2152 	struct vmspace *vm;
2153 
2154 	name = (int *)arg1;
2155 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2156 	if (error != 0)
2157 		return (error);
2158 	vm = vmspace_acquire_ref(p);
2159 	if (vm == NULL) {
2160 		PRELE(p);
2161 		return (ESRCH);
2162 	}
2163 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2164 
2165 	map = &vm->vm_map;
2166 	vm_map_lock_read(map);
2167 	for (entry = map->header.next; entry != &map->header;
2168 	    entry = entry->next) {
2169 		vm_object_t obj, tobj, lobj;
2170 		vm_offset_t addr;
2171 
2172 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2173 			continue;
2174 
2175 		bzero(kve, sizeof(*kve));
2176 		kve->kve_structsize = sizeof(*kve);
2177 
2178 		kve->kve_private_resident = 0;
2179 		obj = entry->object.vm_object;
2180 		if (obj != NULL) {
2181 			VM_OBJECT_RLOCK(obj);
2182 			if (obj->shadow_count == 1)
2183 				kve->kve_private_resident =
2184 				    obj->resident_page_count;
2185 		}
2186 		kve->kve_resident = 0;
2187 		addr = entry->start;
2188 		while (addr < entry->end) {
2189 			if (pmap_extract(map->pmap, addr))
2190 				kve->kve_resident++;
2191 			addr += PAGE_SIZE;
2192 		}
2193 
2194 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2195 			if (tobj != obj) {
2196 				VM_OBJECT_RLOCK(tobj);
2197 				kve->kve_offset += tobj->backing_object_offset;
2198 			}
2199 			if (lobj != obj)
2200 				VM_OBJECT_RUNLOCK(lobj);
2201 			lobj = tobj;
2202 		}
2203 
2204 		kve->kve_start = (void*)entry->start;
2205 		kve->kve_end = (void*)entry->end;
2206 		kve->kve_offset += (off_t)entry->offset;
2207 
2208 		if (entry->protection & VM_PROT_READ)
2209 			kve->kve_protection |= KVME_PROT_READ;
2210 		if (entry->protection & VM_PROT_WRITE)
2211 			kve->kve_protection |= KVME_PROT_WRITE;
2212 		if (entry->protection & VM_PROT_EXECUTE)
2213 			kve->kve_protection |= KVME_PROT_EXEC;
2214 
2215 		if (entry->eflags & MAP_ENTRY_COW)
2216 			kve->kve_flags |= KVME_FLAG_COW;
2217 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2218 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2219 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2220 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2221 
2222 		last_timestamp = map->timestamp;
2223 		vm_map_unlock_read(map);
2224 
2225 		kve->kve_fileid = 0;
2226 		kve->kve_fsid = 0;
2227 		freepath = NULL;
2228 		fullpath = "";
2229 		if (lobj) {
2230 			vp = NULL;
2231 			switch (lobj->type) {
2232 			case OBJT_DEFAULT:
2233 				kve->kve_type = KVME_TYPE_DEFAULT;
2234 				break;
2235 			case OBJT_VNODE:
2236 				kve->kve_type = KVME_TYPE_VNODE;
2237 				vp = lobj->handle;
2238 				vref(vp);
2239 				break;
2240 			case OBJT_SWAP:
2241 				if ((lobj->flags & OBJ_TMPFS_NODE) != 0) {
2242 					kve->kve_type = KVME_TYPE_VNODE;
2243 					if ((lobj->flags & OBJ_TMPFS) != 0) {
2244 						vp = lobj->un_pager.swp.swp_tmpfs;
2245 						vref(vp);
2246 					}
2247 				} else {
2248 					kve->kve_type = KVME_TYPE_SWAP;
2249 				}
2250 				break;
2251 			case OBJT_DEVICE:
2252 				kve->kve_type = KVME_TYPE_DEVICE;
2253 				break;
2254 			case OBJT_PHYS:
2255 				kve->kve_type = KVME_TYPE_PHYS;
2256 				break;
2257 			case OBJT_DEAD:
2258 				kve->kve_type = KVME_TYPE_DEAD;
2259 				break;
2260 			case OBJT_SG:
2261 				kve->kve_type = KVME_TYPE_SG;
2262 				break;
2263 			default:
2264 				kve->kve_type = KVME_TYPE_UNKNOWN;
2265 				break;
2266 			}
2267 			if (lobj != obj)
2268 				VM_OBJECT_RUNLOCK(lobj);
2269 
2270 			kve->kve_ref_count = obj->ref_count;
2271 			kve->kve_shadow_count = obj->shadow_count;
2272 			VM_OBJECT_RUNLOCK(obj);
2273 			if (vp != NULL) {
2274 				vn_fullpath(curthread, vp, &fullpath,
2275 				    &freepath);
2276 				cred = curthread->td_ucred;
2277 				vn_lock(vp, LK_SHARED | LK_RETRY);
2278 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2279 					kve->kve_fileid = va.va_fileid;
2280 					/* truncate */
2281 					kve->kve_fsid = va.va_fsid;
2282 				}
2283 				vput(vp);
2284 			}
2285 		} else {
2286 			kve->kve_type = KVME_TYPE_NONE;
2287 			kve->kve_ref_count = 0;
2288 			kve->kve_shadow_count = 0;
2289 		}
2290 
2291 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2292 		if (freepath != NULL)
2293 			free(freepath, M_TEMP);
2294 
2295 		error = SYSCTL_OUT(req, kve, sizeof(*kve));
2296 		vm_map_lock_read(map);
2297 		if (error)
2298 			break;
2299 		if (last_timestamp != map->timestamp) {
2300 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2301 			entry = tmp_entry;
2302 		}
2303 	}
2304 	vm_map_unlock_read(map);
2305 	vmspace_free(vm);
2306 	PRELE(p);
2307 	free(kve, M_TEMP);
2308 	return (error);
2309 }
2310 #endif	/* COMPAT_FREEBSD7 */
2311 
2312 #ifdef KINFO_VMENTRY_SIZE
2313 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2314 #endif
2315 
2316 void
2317 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry,
2318     int *resident_count, bool *super)
2319 {
2320 	vm_object_t obj, tobj;
2321 	vm_page_t m, m_adv;
2322 	vm_offset_t addr;
2323 	vm_paddr_t locked_pa;
2324 	vm_pindex_t pi, pi_adv, pindex;
2325 
2326 	*super = false;
2327 	*resident_count = 0;
2328 	if (vmmap_skip_res_cnt)
2329 		return;
2330 
2331 	locked_pa = 0;
2332 	obj = entry->object.vm_object;
2333 	addr = entry->start;
2334 	m_adv = NULL;
2335 	pi = OFF_TO_IDX(entry->offset);
2336 	for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) {
2337 		if (m_adv != NULL) {
2338 			m = m_adv;
2339 		} else {
2340 			pi_adv = atop(entry->end - addr);
2341 			pindex = pi;
2342 			for (tobj = obj;; tobj = tobj->backing_object) {
2343 				m = vm_page_find_least(tobj, pindex);
2344 				if (m != NULL) {
2345 					if (m->pindex == pindex)
2346 						break;
2347 					if (pi_adv > m->pindex - pindex) {
2348 						pi_adv = m->pindex - pindex;
2349 						m_adv = m;
2350 					}
2351 				}
2352 				if (tobj->backing_object == NULL)
2353 					goto next;
2354 				pindex += OFF_TO_IDX(tobj->
2355 				    backing_object_offset);
2356 			}
2357 		}
2358 		m_adv = NULL;
2359 		if (m->psind != 0 && addr + pagesizes[1] <= entry->end &&
2360 		    (addr & (pagesizes[1] - 1)) == 0 &&
2361 		    (pmap_mincore(map->pmap, addr, &locked_pa) &
2362 		    MINCORE_SUPER) != 0) {
2363 			*super = true;
2364 			pi_adv = atop(pagesizes[1]);
2365 		} else {
2366 			/*
2367 			 * We do not test the found page on validity.
2368 			 * Either the page is busy and being paged in,
2369 			 * or it was invalidated.  The first case
2370 			 * should be counted as resident, the second
2371 			 * is not so clear; we do account both.
2372 			 */
2373 			pi_adv = 1;
2374 		}
2375 		*resident_count += pi_adv;
2376 next:;
2377 	}
2378 	PA_UNLOCK_COND(locked_pa);
2379 }
2380 
2381 /*
2382  * Must be called with the process locked and will return unlocked.
2383  */
2384 int
2385 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb, ssize_t maxlen, int flags)
2386 {
2387 	vm_map_entry_t entry, tmp_entry;
2388 	struct vattr va;
2389 	vm_map_t map;
2390 	vm_object_t obj, tobj, lobj;
2391 	char *fullpath, *freepath;
2392 	struct kinfo_vmentry *kve;
2393 	struct ucred *cred;
2394 	struct vnode *vp;
2395 	struct vmspace *vm;
2396 	vm_offset_t addr;
2397 	unsigned int last_timestamp;
2398 	int error;
2399 	bool super;
2400 
2401 	PROC_LOCK_ASSERT(p, MA_OWNED);
2402 
2403 	_PHOLD(p);
2404 	PROC_UNLOCK(p);
2405 	vm = vmspace_acquire_ref(p);
2406 	if (vm == NULL) {
2407 		PRELE(p);
2408 		return (ESRCH);
2409 	}
2410 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK | M_ZERO);
2411 
2412 	error = 0;
2413 	map = &vm->vm_map;
2414 	vm_map_lock_read(map);
2415 	for (entry = map->header.next; entry != &map->header;
2416 	    entry = entry->next) {
2417 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2418 			continue;
2419 
2420 		addr = entry->end;
2421 		bzero(kve, sizeof(*kve));
2422 		obj = entry->object.vm_object;
2423 		if (obj != NULL) {
2424 			for (tobj = obj; tobj != NULL;
2425 			    tobj = tobj->backing_object) {
2426 				VM_OBJECT_RLOCK(tobj);
2427 				kve->kve_offset += tobj->backing_object_offset;
2428 				lobj = tobj;
2429 			}
2430 			if (obj->backing_object == NULL)
2431 				kve->kve_private_resident =
2432 				    obj->resident_page_count;
2433 			kern_proc_vmmap_resident(map, entry,
2434 			    &kve->kve_resident, &super);
2435 			if (super)
2436 				kve->kve_flags |= KVME_FLAG_SUPER;
2437 			for (tobj = obj; tobj != NULL;
2438 			    tobj = tobj->backing_object) {
2439 				if (tobj != obj && tobj != lobj)
2440 					VM_OBJECT_RUNLOCK(tobj);
2441 			}
2442 		} else {
2443 			lobj = NULL;
2444 		}
2445 
2446 		kve->kve_start = entry->start;
2447 		kve->kve_end = entry->end;
2448 		kve->kve_offset += entry->offset;
2449 
2450 		if (entry->protection & VM_PROT_READ)
2451 			kve->kve_protection |= KVME_PROT_READ;
2452 		if (entry->protection & VM_PROT_WRITE)
2453 			kve->kve_protection |= KVME_PROT_WRITE;
2454 		if (entry->protection & VM_PROT_EXECUTE)
2455 			kve->kve_protection |= KVME_PROT_EXEC;
2456 
2457 		if (entry->eflags & MAP_ENTRY_COW)
2458 			kve->kve_flags |= KVME_FLAG_COW;
2459 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2460 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2461 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2462 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2463 		if (entry->eflags & MAP_ENTRY_GROWS_UP)
2464 			kve->kve_flags |= KVME_FLAG_GROWS_UP;
2465 		if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2466 			kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2467 
2468 		last_timestamp = map->timestamp;
2469 		vm_map_unlock_read(map);
2470 
2471 		freepath = NULL;
2472 		fullpath = "";
2473 		if (lobj != NULL) {
2474 			vp = NULL;
2475 			switch (lobj->type) {
2476 			case OBJT_DEFAULT:
2477 				kve->kve_type = KVME_TYPE_DEFAULT;
2478 				break;
2479 			case OBJT_VNODE:
2480 				kve->kve_type = KVME_TYPE_VNODE;
2481 				vp = lobj->handle;
2482 				vref(vp);
2483 				break;
2484 			case OBJT_SWAP:
2485 				if ((lobj->flags & OBJ_TMPFS_NODE) != 0) {
2486 					kve->kve_type = KVME_TYPE_VNODE;
2487 					if ((lobj->flags & OBJ_TMPFS) != 0) {
2488 						vp = lobj->un_pager.swp.swp_tmpfs;
2489 						vref(vp);
2490 					}
2491 				} else {
2492 					kve->kve_type = KVME_TYPE_SWAP;
2493 				}
2494 				break;
2495 			case OBJT_DEVICE:
2496 				kve->kve_type = KVME_TYPE_DEVICE;
2497 				break;
2498 			case OBJT_PHYS:
2499 				kve->kve_type = KVME_TYPE_PHYS;
2500 				break;
2501 			case OBJT_DEAD:
2502 				kve->kve_type = KVME_TYPE_DEAD;
2503 				break;
2504 			case OBJT_SG:
2505 				kve->kve_type = KVME_TYPE_SG;
2506 				break;
2507 			case OBJT_MGTDEVICE:
2508 				kve->kve_type = KVME_TYPE_MGTDEVICE;
2509 				break;
2510 			default:
2511 				kve->kve_type = KVME_TYPE_UNKNOWN;
2512 				break;
2513 			}
2514 			if (lobj != obj)
2515 				VM_OBJECT_RUNLOCK(lobj);
2516 
2517 			kve->kve_ref_count = obj->ref_count;
2518 			kve->kve_shadow_count = obj->shadow_count;
2519 			VM_OBJECT_RUNLOCK(obj);
2520 			if (vp != NULL) {
2521 				vn_fullpath(curthread, vp, &fullpath,
2522 				    &freepath);
2523 				kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2524 				cred = curthread->td_ucred;
2525 				vn_lock(vp, LK_SHARED | LK_RETRY);
2526 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2527 					kve->kve_vn_fileid = va.va_fileid;
2528 					kve->kve_vn_fsid = va.va_fsid;
2529 					kve->kve_vn_fsid_freebsd11 =
2530 					    kve->kve_vn_fsid; /* truncate */
2531 					kve->kve_vn_mode =
2532 					    MAKEIMODE(va.va_type, va.va_mode);
2533 					kve->kve_vn_size = va.va_size;
2534 					kve->kve_vn_rdev = va.va_rdev;
2535 					kve->kve_vn_rdev_freebsd11 =
2536 					    kve->kve_vn_rdev; /* truncate */
2537 					kve->kve_status = KF_ATTR_VALID;
2538 				}
2539 				vput(vp);
2540 			}
2541 		} else {
2542 			kve->kve_type = KVME_TYPE_NONE;
2543 			kve->kve_ref_count = 0;
2544 			kve->kve_shadow_count = 0;
2545 		}
2546 
2547 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2548 		if (freepath != NULL)
2549 			free(freepath, M_TEMP);
2550 
2551 		/* Pack record size down */
2552 		if ((flags & KERN_VMMAP_PACK_KINFO) != 0)
2553 			kve->kve_structsize =
2554 			    offsetof(struct kinfo_vmentry, kve_path) +
2555 			    strlen(kve->kve_path) + 1;
2556 		else
2557 			kve->kve_structsize = sizeof(*kve);
2558 		kve->kve_structsize = roundup(kve->kve_structsize,
2559 		    sizeof(uint64_t));
2560 
2561 		/* Halt filling and truncate rather than exceeding maxlen */
2562 		if (maxlen != -1 && maxlen < kve->kve_structsize) {
2563 			error = 0;
2564 			vm_map_lock_read(map);
2565 			break;
2566 		} else if (maxlen != -1)
2567 			maxlen -= kve->kve_structsize;
2568 
2569 		if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0)
2570 			error = ENOMEM;
2571 		vm_map_lock_read(map);
2572 		if (error != 0)
2573 			break;
2574 		if (last_timestamp != map->timestamp) {
2575 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2576 			entry = tmp_entry;
2577 		}
2578 	}
2579 	vm_map_unlock_read(map);
2580 	vmspace_free(vm);
2581 	PRELE(p);
2582 	free(kve, M_TEMP);
2583 	return (error);
2584 }
2585 
2586 static int
2587 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2588 {
2589 	struct proc *p;
2590 	struct sbuf sb;
2591 	int error, error2, *name;
2592 
2593 	name = (int *)arg1;
2594 	sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req);
2595 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2596 	error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
2597 	if (error != 0) {
2598 		sbuf_delete(&sb);
2599 		return (error);
2600 	}
2601 	error = kern_proc_vmmap_out(p, &sb, -1, KERN_VMMAP_PACK_KINFO);
2602 	error2 = sbuf_finish(&sb);
2603 	sbuf_delete(&sb);
2604 	return (error != 0 ? error : error2);
2605 }
2606 
2607 #if defined(STACK) || defined(DDB)
2608 static int
2609 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2610 {
2611 	struct kinfo_kstack *kkstp;
2612 	int error, i, *name, numthreads;
2613 	lwpid_t *lwpidarray;
2614 	struct thread *td;
2615 	struct stack *st;
2616 	struct sbuf sb;
2617 	struct proc *p;
2618 
2619 	name = (int *)arg1;
2620 	error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2621 	if (error != 0)
2622 		return (error);
2623 
2624 	kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2625 	st = stack_create(M_WAITOK);
2626 
2627 	lwpidarray = NULL;
2628 	PROC_LOCK(p);
2629 	do {
2630 		if (lwpidarray != NULL) {
2631 			free(lwpidarray, M_TEMP);
2632 			lwpidarray = NULL;
2633 		}
2634 		numthreads = p->p_numthreads;
2635 		PROC_UNLOCK(p);
2636 		lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2637 		    M_WAITOK | M_ZERO);
2638 		PROC_LOCK(p);
2639 	} while (numthreads < p->p_numthreads);
2640 
2641 	/*
2642 	 * XXXRW: During the below loop, execve(2) and countless other sorts
2643 	 * of changes could have taken place.  Should we check to see if the
2644 	 * vmspace has been replaced, or the like, in order to prevent
2645 	 * giving a snapshot that spans, say, execve(2), with some threads
2646 	 * before and some after?  Among other things, the credentials could
2647 	 * have changed, in which case the right to extract debug info might
2648 	 * no longer be assured.
2649 	 */
2650 	i = 0;
2651 	FOREACH_THREAD_IN_PROC(p, td) {
2652 		KASSERT(i < numthreads,
2653 		    ("sysctl_kern_proc_kstack: numthreads"));
2654 		lwpidarray[i] = td->td_tid;
2655 		i++;
2656 	}
2657 	numthreads = i;
2658 	for (i = 0; i < numthreads; i++) {
2659 		td = thread_find(p, lwpidarray[i]);
2660 		if (td == NULL) {
2661 			continue;
2662 		}
2663 		bzero(kkstp, sizeof(*kkstp));
2664 		(void)sbuf_new(&sb, kkstp->kkst_trace,
2665 		    sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2666 		thread_lock(td);
2667 		kkstp->kkst_tid = td->td_tid;
2668 		if (TD_IS_SWAPPED(td)) {
2669 			kkstp->kkst_state = KKST_STATE_SWAPPED;
2670 		} else if (TD_IS_RUNNING(td)) {
2671 			if (stack_save_td_running(st, td) == 0)
2672 				kkstp->kkst_state = KKST_STATE_STACKOK;
2673 			else
2674 				kkstp->kkst_state = KKST_STATE_RUNNING;
2675 		} else {
2676 			kkstp->kkst_state = KKST_STATE_STACKOK;
2677 			stack_save_td(st, td);
2678 		}
2679 		thread_unlock(td);
2680 		PROC_UNLOCK(p);
2681 		stack_sbuf_print(&sb, st);
2682 		sbuf_finish(&sb);
2683 		sbuf_delete(&sb);
2684 		error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2685 		PROC_LOCK(p);
2686 		if (error)
2687 			break;
2688 	}
2689 	_PRELE(p);
2690 	PROC_UNLOCK(p);
2691 	if (lwpidarray != NULL)
2692 		free(lwpidarray, M_TEMP);
2693 	stack_destroy(st);
2694 	free(kkstp, M_TEMP);
2695 	return (error);
2696 }
2697 #endif
2698 
2699 /*
2700  * This sysctl allows a process to retrieve the full list of groups from
2701  * itself or another process.
2702  */
2703 static int
2704 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2705 {
2706 	pid_t *pidp = (pid_t *)arg1;
2707 	unsigned int arglen = arg2;
2708 	struct proc *p;
2709 	struct ucred *cred;
2710 	int error;
2711 
2712 	if (arglen != 1)
2713 		return (EINVAL);
2714 	if (*pidp == -1) {	/* -1 means this process */
2715 		p = req->td->td_proc;
2716 		PROC_LOCK(p);
2717 	} else {
2718 		error = pget(*pidp, PGET_CANSEE, &p);
2719 		if (error != 0)
2720 			return (error);
2721 	}
2722 
2723 	cred = crhold(p->p_ucred);
2724 	PROC_UNLOCK(p);
2725 
2726 	error = SYSCTL_OUT(req, cred->cr_groups,
2727 	    cred->cr_ngroups * sizeof(gid_t));
2728 	crfree(cred);
2729 	return (error);
2730 }
2731 
2732 /*
2733  * This sysctl allows a process to retrieve or/and set the resource limit for
2734  * another process.
2735  */
2736 static int
2737 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2738 {
2739 	int *name = (int *)arg1;
2740 	u_int namelen = arg2;
2741 	struct rlimit rlim;
2742 	struct proc *p;
2743 	u_int which;
2744 	int flags, error;
2745 
2746 	if (namelen != 2)
2747 		return (EINVAL);
2748 
2749 	which = (u_int)name[1];
2750 	if (which >= RLIM_NLIMITS)
2751 		return (EINVAL);
2752 
2753 	if (req->newptr != NULL && req->newlen != sizeof(rlim))
2754 		return (EINVAL);
2755 
2756 	flags = PGET_HOLD | PGET_NOTWEXIT;
2757 	if (req->newptr != NULL)
2758 		flags |= PGET_CANDEBUG;
2759 	else
2760 		flags |= PGET_CANSEE;
2761 	error = pget((pid_t)name[0], flags, &p);
2762 	if (error != 0)
2763 		return (error);
2764 
2765 	/*
2766 	 * Retrieve limit.
2767 	 */
2768 	if (req->oldptr != NULL) {
2769 		PROC_LOCK(p);
2770 		lim_rlimit_proc(p, which, &rlim);
2771 		PROC_UNLOCK(p);
2772 	}
2773 	error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
2774 	if (error != 0)
2775 		goto errout;
2776 
2777 	/*
2778 	 * Set limit.
2779 	 */
2780 	if (req->newptr != NULL) {
2781 		error = SYSCTL_IN(req, &rlim, sizeof(rlim));
2782 		if (error == 0)
2783 			error = kern_proc_setrlimit(curthread, p, which, &rlim);
2784 	}
2785 
2786 errout:
2787 	PRELE(p);
2788 	return (error);
2789 }
2790 
2791 /*
2792  * This sysctl allows a process to retrieve ps_strings structure location of
2793  * another process.
2794  */
2795 static int
2796 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
2797 {
2798 	int *name = (int *)arg1;
2799 	u_int namelen = arg2;
2800 	struct proc *p;
2801 	vm_offset_t ps_strings;
2802 	int error;
2803 #ifdef COMPAT_FREEBSD32
2804 	uint32_t ps_strings32;
2805 #endif
2806 
2807 	if (namelen != 1)
2808 		return (EINVAL);
2809 
2810 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2811 	if (error != 0)
2812 		return (error);
2813 #ifdef COMPAT_FREEBSD32
2814 	if ((req->flags & SCTL_MASK32) != 0) {
2815 		/*
2816 		 * We return 0 if the 32 bit emulation request is for a 64 bit
2817 		 * process.
2818 		 */
2819 		ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
2820 		    PTROUT(p->p_sysent->sv_psstrings) : 0;
2821 		PROC_UNLOCK(p);
2822 		error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
2823 		return (error);
2824 	}
2825 #endif
2826 	ps_strings = p->p_sysent->sv_psstrings;
2827 	PROC_UNLOCK(p);
2828 	error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
2829 	return (error);
2830 }
2831 
2832 /*
2833  * This sysctl allows a process to retrieve umask of another process.
2834  */
2835 static int
2836 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
2837 {
2838 	int *name = (int *)arg1;
2839 	u_int namelen = arg2;
2840 	struct proc *p;
2841 	int error;
2842 	u_short fd_cmask;
2843 	pid_t pid;
2844 
2845 	if (namelen != 1)
2846 		return (EINVAL);
2847 
2848 	pid = (pid_t)name[0];
2849 	p = curproc;
2850 	if (pid == p->p_pid || pid == 0) {
2851 		fd_cmask = p->p_fd->fd_cmask;
2852 		goto out;
2853 	}
2854 
2855 	error = pget(pid, PGET_WANTREAD, &p);
2856 	if (error != 0)
2857 		return (error);
2858 
2859 	fd_cmask = p->p_fd->fd_cmask;
2860 	PRELE(p);
2861 out:
2862 	error = SYSCTL_OUT(req, &fd_cmask, sizeof(fd_cmask));
2863 	return (error);
2864 }
2865 
2866 /*
2867  * This sysctl allows a process to set and retrieve binary osreldate of
2868  * another process.
2869  */
2870 static int
2871 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
2872 {
2873 	int *name = (int *)arg1;
2874 	u_int namelen = arg2;
2875 	struct proc *p;
2876 	int flags, error, osrel;
2877 
2878 	if (namelen != 1)
2879 		return (EINVAL);
2880 
2881 	if (req->newptr != NULL && req->newlen != sizeof(osrel))
2882 		return (EINVAL);
2883 
2884 	flags = PGET_HOLD | PGET_NOTWEXIT;
2885 	if (req->newptr != NULL)
2886 		flags |= PGET_CANDEBUG;
2887 	else
2888 		flags |= PGET_CANSEE;
2889 	error = pget((pid_t)name[0], flags, &p);
2890 	if (error != 0)
2891 		return (error);
2892 
2893 	error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
2894 	if (error != 0)
2895 		goto errout;
2896 
2897 	if (req->newptr != NULL) {
2898 		error = SYSCTL_IN(req, &osrel, sizeof(osrel));
2899 		if (error != 0)
2900 			goto errout;
2901 		if (osrel < 0) {
2902 			error = EINVAL;
2903 			goto errout;
2904 		}
2905 		p->p_osrel = osrel;
2906 	}
2907 errout:
2908 	PRELE(p);
2909 	return (error);
2910 }
2911 
2912 static int
2913 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS)
2914 {
2915 	int *name = (int *)arg1;
2916 	u_int namelen = arg2;
2917 	struct proc *p;
2918 	struct kinfo_sigtramp kst;
2919 	const struct sysentvec *sv;
2920 	int error;
2921 #ifdef COMPAT_FREEBSD32
2922 	struct kinfo_sigtramp32 kst32;
2923 #endif
2924 
2925 	if (namelen != 1)
2926 		return (EINVAL);
2927 
2928 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2929 	if (error != 0)
2930 		return (error);
2931 	sv = p->p_sysent;
2932 #ifdef COMPAT_FREEBSD32
2933 	if ((req->flags & SCTL_MASK32) != 0) {
2934 		bzero(&kst32, sizeof(kst32));
2935 		if (SV_PROC_FLAG(p, SV_ILP32)) {
2936 			if (sv->sv_sigcode_base != 0) {
2937 				kst32.ksigtramp_start = sv->sv_sigcode_base;
2938 				kst32.ksigtramp_end = sv->sv_sigcode_base +
2939 				    *sv->sv_szsigcode;
2940 			} else {
2941 				kst32.ksigtramp_start = sv->sv_psstrings -
2942 				    *sv->sv_szsigcode;
2943 				kst32.ksigtramp_end = sv->sv_psstrings;
2944 			}
2945 		}
2946 		PROC_UNLOCK(p);
2947 		error = SYSCTL_OUT(req, &kst32, sizeof(kst32));
2948 		return (error);
2949 	}
2950 #endif
2951 	bzero(&kst, sizeof(kst));
2952 	if (sv->sv_sigcode_base != 0) {
2953 		kst.ksigtramp_start = (char *)sv->sv_sigcode_base;
2954 		kst.ksigtramp_end = (char *)sv->sv_sigcode_base +
2955 		    *sv->sv_szsigcode;
2956 	} else {
2957 		kst.ksigtramp_start = (char *)sv->sv_psstrings -
2958 		    *sv->sv_szsigcode;
2959 		kst.ksigtramp_end = (char *)sv->sv_psstrings;
2960 	}
2961 	PROC_UNLOCK(p);
2962 	error = SYSCTL_OUT(req, &kst, sizeof(kst));
2963 	return (error);
2964 }
2965 
2966 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD,  0, "Process table");
2967 
2968 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
2969 	CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
2970 	"Return entire process table");
2971 
2972 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2973 	sysctl_kern_proc, "Process table");
2974 
2975 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
2976 	sysctl_kern_proc, "Process table");
2977 
2978 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2979 	sysctl_kern_proc, "Process table");
2980 
2981 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
2982 	CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2983 
2984 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
2985 	sysctl_kern_proc, "Process table");
2986 
2987 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2988 	sysctl_kern_proc, "Process table");
2989 
2990 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2991 	sysctl_kern_proc, "Process table");
2992 
2993 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2994 	sysctl_kern_proc, "Process table");
2995 
2996 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
2997 	sysctl_kern_proc, "Return process table, no threads");
2998 
2999 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
3000 	CTLFLAG_RW | CTLFLAG_CAPWR | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
3001 	sysctl_kern_proc_args, "Process argument list");
3002 
3003 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
3004 	sysctl_kern_proc_env, "Process environment");
3005 
3006 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
3007 	CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
3008 
3009 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
3010 	CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
3011 
3012 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
3013 	CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
3014 	"Process syscall vector name (ABI type)");
3015 
3016 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
3017 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3018 
3019 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
3020 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3021 
3022 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
3023 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3024 
3025 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
3026 	sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3027 
3028 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
3029 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3030 
3031 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
3032 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3033 
3034 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
3035 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3036 
3037 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
3038 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3039 
3040 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
3041 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
3042 	"Return process table, no threads");
3043 
3044 #ifdef COMPAT_FREEBSD7
3045 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
3046 	CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
3047 #endif
3048 
3049 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
3050 	CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
3051 
3052 #if defined(STACK) || defined(DDB)
3053 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
3054 	CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
3055 #endif
3056 
3057 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
3058 	CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
3059 
3060 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
3061 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
3062 	"Process resource limits");
3063 
3064 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
3065 	CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
3066 	"Process ps_strings location");
3067 
3068 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
3069 	CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
3070 
3071 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
3072 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
3073 	"Process binary osreldate");
3074 
3075 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD |
3076 	CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp,
3077 	"Process signal trampoline location");
3078 
3079 int allproc_gen;
3080 
3081 /*
3082  * stop_all_proc() purpose is to stop all process which have usermode,
3083  * except current process for obvious reasons.  This makes it somewhat
3084  * unreliable when invoked from multithreaded process.  The service
3085  * must not be user-callable anyway.
3086  */
3087 void
3088 stop_all_proc(void)
3089 {
3090 	struct proc *cp, *p;
3091 	int r, gen;
3092 	bool restart, seen_stopped, seen_exiting, stopped_some;
3093 
3094 	cp = curproc;
3095 allproc_loop:
3096 	sx_xlock(&allproc_lock);
3097 	gen = allproc_gen;
3098 	seen_exiting = seen_stopped = stopped_some = restart = false;
3099 	LIST_REMOVE(cp, p_list);
3100 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3101 	for (;;) {
3102 		p = LIST_NEXT(cp, p_list);
3103 		if (p == NULL)
3104 			break;
3105 		LIST_REMOVE(cp, p_list);
3106 		LIST_INSERT_AFTER(p, cp, p_list);
3107 		PROC_LOCK(p);
3108 		if ((p->p_flag & (P_KPROC | P_SYSTEM | P_TOTAL_STOP)) != 0) {
3109 			PROC_UNLOCK(p);
3110 			continue;
3111 		}
3112 		if ((p->p_flag & P_WEXIT) != 0) {
3113 			seen_exiting = true;
3114 			PROC_UNLOCK(p);
3115 			continue;
3116 		}
3117 		if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) {
3118 			/*
3119 			 * Stopped processes are tolerated when there
3120 			 * are no other processes which might continue
3121 			 * them.  P_STOPPED_SINGLE but not
3122 			 * P_TOTAL_STOP process still has at least one
3123 			 * thread running.
3124 			 */
3125 			seen_stopped = true;
3126 			PROC_UNLOCK(p);
3127 			continue;
3128 		}
3129 		_PHOLD(p);
3130 		sx_xunlock(&allproc_lock);
3131 		r = thread_single(p, SINGLE_ALLPROC);
3132 		if (r != 0)
3133 			restart = true;
3134 		else
3135 			stopped_some = true;
3136 		_PRELE(p);
3137 		PROC_UNLOCK(p);
3138 		sx_xlock(&allproc_lock);
3139 	}
3140 	/* Catch forked children we did not see in iteration. */
3141 	if (gen != allproc_gen)
3142 		restart = true;
3143 	sx_xunlock(&allproc_lock);
3144 	if (restart || stopped_some || seen_exiting || seen_stopped) {
3145 		kern_yield(PRI_USER);
3146 		goto allproc_loop;
3147 	}
3148 }
3149 
3150 void
3151 resume_all_proc(void)
3152 {
3153 	struct proc *cp, *p;
3154 
3155 	cp = curproc;
3156 	sx_xlock(&allproc_lock);
3157 again:
3158 	LIST_REMOVE(cp, p_list);
3159 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3160 	for (;;) {
3161 		p = LIST_NEXT(cp, p_list);
3162 		if (p == NULL)
3163 			break;
3164 		LIST_REMOVE(cp, p_list);
3165 		LIST_INSERT_AFTER(p, cp, p_list);
3166 		PROC_LOCK(p);
3167 		if ((p->p_flag & P_TOTAL_STOP) != 0) {
3168 			sx_xunlock(&allproc_lock);
3169 			_PHOLD(p);
3170 			thread_single_end(p, SINGLE_ALLPROC);
3171 			_PRELE(p);
3172 			PROC_UNLOCK(p);
3173 			sx_xlock(&allproc_lock);
3174 		} else {
3175 			PROC_UNLOCK(p);
3176 		}
3177 	}
3178 	/*  Did the loop above missed any stopped process ? */
3179 	LIST_FOREACH(p, &allproc, p_list) {
3180 		/* No need for proc lock. */
3181 		if ((p->p_flag & P_TOTAL_STOP) != 0)
3182 			goto again;
3183 	}
3184 	sx_xunlock(&allproc_lock);
3185 }
3186 
3187 /* #define	TOTAL_STOP_DEBUG	1 */
3188 #ifdef TOTAL_STOP_DEBUG
3189 volatile static int ap_resume;
3190 #include <sys/mount.h>
3191 
3192 static int
3193 sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS)
3194 {
3195 	int error, val;
3196 
3197 	val = 0;
3198 	ap_resume = 0;
3199 	error = sysctl_handle_int(oidp, &val, 0, req);
3200 	if (error != 0 || req->newptr == NULL)
3201 		return (error);
3202 	if (val != 0) {
3203 		stop_all_proc();
3204 		syncer_suspend();
3205 		while (ap_resume == 0)
3206 			;
3207 		syncer_resume();
3208 		resume_all_proc();
3209 	}
3210 	return (0);
3211 }
3212 
3213 SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW |
3214     CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0,
3215     sysctl_debug_stop_all_proc, "I",
3216     "");
3217 #endif
3218