xref: /freebsd/sys/kern/kern_proc.c (revision 31fab773613b21183fbd5a313fa8c69d37fcb196)
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 
32 #include <sys/cdefs.h>
33 #include "opt_ddb.h"
34 #include "opt_ktrace.h"
35 #include "opt_kstack_pages.h"
36 #include "opt_stack.h"
37 
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/bitstring.h>
41 #include <sys/conf.h>
42 #include <sys/elf.h>
43 #include <sys/eventhandler.h>
44 #include <sys/exec.h>
45 #include <sys/fcntl.h>
46 #include <sys/ipc.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/namei.h>
57 #include <sys/proc.h>
58 #include <sys/ptrace.h>
59 #include <sys/refcount.h>
60 #include <sys/resourcevar.h>
61 #include <sys/rwlock.h>
62 #include <sys/sbuf.h>
63 #include <sys/sysent.h>
64 #include <sys/sched.h>
65 #include <sys/shm.h>
66 #include <sys/smp.h>
67 #include <sys/stack.h>
68 #include <sys/stat.h>
69 #include <sys/dtrace_bsd.h>
70 #include <sys/sysctl.h>
71 #include <sys/filedesc.h>
72 #include <sys/tty.h>
73 #include <sys/signalvar.h>
74 #include <sys/sdt.h>
75 #include <sys/sx.h>
76 #include <sys/user.h>
77 #include <sys/vnode.h>
78 #include <sys/wait.h>
79 #ifdef KTRACE
80 #include <sys/ktrace.h>
81 #endif
82 
83 #ifdef DDB
84 #include <ddb/ddb.h>
85 #endif
86 
87 #include <vm/vm.h>
88 #include <vm/vm_param.h>
89 #include <vm/vm_extern.h>
90 #include <vm/pmap.h>
91 #include <vm/vm_map.h>
92 #include <vm/vm_object.h>
93 #include <vm/vm_page.h>
94 #include <vm/vm_pager.h>
95 #include <vm/vm_radix.h>
96 #include <vm/uma.h>
97 
98 #include <fs/devfs/devfs.h>
99 
100 #ifdef COMPAT_FREEBSD32
101 #include <compat/freebsd32/freebsd32.h>
102 #include <compat/freebsd32/freebsd32_util.h>
103 #endif
104 
105 SDT_PROVIDER_DEFINE(proc);
106 
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 pgdelete(struct pgrp *);
118 static int pgrp_init(void *mem, int size, int flags);
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 
125 /*
126  * Other process lists
127  */
128 struct pidhashhead *pidhashtbl = NULL;
129 struct sx *pidhashtbl_lock;
130 u_long pidhash;
131 u_long pidhashlock;
132 struct pgrphashhead *pgrphashtbl;
133 u_long pgrphash;
134 struct proclist allproc = LIST_HEAD_INITIALIZER(allproc);
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 struct mtx __exclusive_cache_line procid_lock;
139 uma_zone_t proc_zone;
140 uma_zone_t pgrp_zone;
141 
142 /*
143  * The offset of various fields in struct proc and struct thread.
144  * These are used by kernel debuggers to enumerate kernel threads and
145  * processes.
146  */
147 const int proc_off_p_pid = offsetof(struct proc, p_pid);
148 const int proc_off_p_comm = offsetof(struct proc, p_comm);
149 const int proc_off_p_list = offsetof(struct proc, p_list);
150 const int proc_off_p_hash = offsetof(struct proc, p_hash);
151 const int proc_off_p_threads = offsetof(struct proc, p_threads);
152 const int thread_off_td_tid = offsetof(struct thread, td_tid);
153 const int thread_off_td_name = offsetof(struct thread, td_name);
154 const int thread_off_td_oncpu = offsetof(struct thread, td_oncpu);
155 const int thread_off_td_pcb = offsetof(struct thread, td_pcb);
156 const int thread_off_td_plist = offsetof(struct thread, td_plist);
157 
158 EVENTHANDLER_LIST_DEFINE(process_ctor);
159 EVENTHANDLER_LIST_DEFINE(process_dtor);
160 EVENTHANDLER_LIST_DEFINE(process_init);
161 EVENTHANDLER_LIST_DEFINE(process_fini);
162 EVENTHANDLER_LIST_DEFINE(process_exit);
163 EVENTHANDLER_LIST_DEFINE(process_fork);
164 EVENTHANDLER_LIST_DEFINE(process_exec);
165 
166 int kstack_pages = KSTACK_PAGES;
167 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RDTUN | CTLFLAG_NOFETCH,
168     &kstack_pages, 0,
169     "Kernel stack size in pages");
170 static int vmmap_skip_res_cnt = 0;
171 SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW,
172     &vmmap_skip_res_cnt, 0,
173     "Skip calculation of the pages resident count in kern.proc.vmmap");
174 
175 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
176 #ifdef COMPAT_FREEBSD32
177 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
178 #endif
179 
180 /*
181  * Initialize global process hashing structures.
182  */
183 void
184 procinit(void)
185 {
186 	u_long i;
187 
188 	sx_init(&allproc_lock, "allproc");
189 	sx_init(&proctree_lock, "proctree");
190 	mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
191 	mtx_init(&procid_lock, "procid", NULL, MTX_DEF);
192 	pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
193 	pidhashlock = (pidhash + 1) / 64;
194 	if (pidhashlock > 0)
195 		pidhashlock--;
196 	pidhashtbl_lock = malloc(sizeof(*pidhashtbl_lock) * (pidhashlock + 1),
197 	    M_PROC, M_WAITOK | M_ZERO);
198 	for (i = 0; i < pidhashlock + 1; i++)
199 		sx_init_flags(&pidhashtbl_lock[i], "pidhash", SX_DUPOK);
200 	pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
201 	proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
202 	    proc_ctor, proc_dtor, proc_init, proc_fini,
203 	    UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
204 	pgrp_zone = uma_zcreate("PGRP", sizeof(struct pgrp), NULL, NULL,
205 	    pgrp_init, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
206 	uihashinit();
207 }
208 
209 /*
210  * Prepare a proc for use.
211  */
212 static int
213 proc_ctor(void *mem, int size, void *arg, int flags)
214 {
215 	struct proc *p;
216 	struct thread *td;
217 
218 	p = (struct proc *)mem;
219 #ifdef KDTRACE_HOOKS
220 	kdtrace_proc_ctor(p);
221 #endif
222 	EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
223 	td = FIRST_THREAD_IN_PROC(p);
224 	if (td != NULL) {
225 		/* Make sure all thread constructors are executed */
226 		EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
227 	}
228 	return (0);
229 }
230 
231 /*
232  * Reclaim a proc after use.
233  */
234 static void
235 proc_dtor(void *mem, int size, void *arg)
236 {
237 	struct proc *p;
238 	struct thread *td;
239 
240 	p = mem;
241 	td = FIRST_THREAD_IN_PROC(p);
242 	if (td != NULL) {
243 		KASSERT(p->p_numthreads == 1,
244 		    ("too many threads in exiting process"));
245 
246 		/* Free all OSD associated to this thread. */
247 		osd_thread_exit(td);
248 		ast_kclear(td);
249 
250 		/* Make sure all thread destructors are executed */
251 		EVENTHANDLER_DIRECT_INVOKE(thread_dtor, td);
252 	}
253 	KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
254 	EVENTHANDLER_DIRECT_INVOKE(process_dtor, p);
255 #ifdef KDTRACE_HOOKS
256 	kdtrace_proc_dtor(p);
257 #endif
258 	KASSERT(p->p_ksi == NULL || !KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
259 }
260 
261 /*
262  * Initialize type-stable parts of a proc (when newly created).
263  */
264 static int
265 proc_init(void *mem, int size, int flags)
266 {
267 	struct proc *p;
268 
269 	p = (struct proc *)mem;
270 	mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK | MTX_NEW);
271 	mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_NEW);
272 	mtx_init(&p->p_statmtx, "pstatl", NULL, MTX_SPIN | MTX_NEW);
273 	mtx_init(&p->p_itimmtx, "pitiml", NULL, MTX_SPIN | MTX_NEW);
274 	mtx_init(&p->p_profmtx, "pprofl", NULL, MTX_SPIN | MTX_NEW);
275 	cv_init(&p->p_pwait, "ppwait");
276 	TAILQ_INIT(&p->p_threads);	     /* all threads in proc */
277 	EVENTHANDLER_DIRECT_INVOKE(process_init, p);
278 	p->p_stats = pstats_alloc();
279 	p->p_pgrp = NULL;
280 	TAILQ_INIT(&p->p_kqtim_stop);
281 	STAILQ_INIT(&p->p_ktr);
282 	return (0);
283 }
284 
285 /*
286  * UMA should ensure that this function is never called.
287  * Freeing a proc structure would violate type stability.
288  */
289 static void
290 proc_fini(void *mem, int size)
291 {
292 #ifdef notnow
293 	struct proc *p;
294 
295 	p = (struct proc *)mem;
296 	EVENTHANDLER_DIRECT_INVOKE(process_fini, p);
297 	pstats_free(p->p_stats);
298 	thread_free(FIRST_THREAD_IN_PROC(p));
299 	mtx_destroy(&p->p_mtx);
300 	if (p->p_ksi != NULL)
301 		ksiginfo_free(p->p_ksi);
302 #else
303 	panic("proc reclaimed");
304 #endif
305 }
306 
307 static int
308 pgrp_init(void *mem, int size, int flags)
309 {
310 	struct pgrp *pg;
311 
312 	pg = mem;
313 	mtx_init(&pg->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
314 	sx_init(&pg->pg_killsx, "killpg racer");
315 	return (0);
316 }
317 
318 /*
319  * PID space management.
320  *
321  * These bitmaps are used by fork_findpid.
322  */
323 bitstr_t bit_decl(proc_id_pidmap, PID_MAX);
324 bitstr_t bit_decl(proc_id_grpidmap, PID_MAX);
325 bitstr_t bit_decl(proc_id_sessidmap, PID_MAX);
326 bitstr_t bit_decl(proc_id_reapmap, PID_MAX);
327 
328 static bitstr_t *proc_id_array[] = {
329 	proc_id_pidmap,
330 	proc_id_grpidmap,
331 	proc_id_sessidmap,
332 	proc_id_reapmap,
333 };
334 
335 void
336 proc_id_set(int type, pid_t id)
337 {
338 
339 	KASSERT(type >= 0 && type < nitems(proc_id_array),
340 	    ("invalid type %d\n", type));
341 	mtx_lock(&procid_lock);
342 	KASSERT(bit_test(proc_id_array[type], id) == 0,
343 	    ("bit %d already set in %d\n", id, type));
344 	bit_set(proc_id_array[type], id);
345 	mtx_unlock(&procid_lock);
346 }
347 
348 void
349 proc_id_set_cond(int type, pid_t id)
350 {
351 
352 	KASSERT(type >= 0 && type < nitems(proc_id_array),
353 	    ("invalid type %d\n", type));
354 	if (bit_test(proc_id_array[type], id))
355 		return;
356 	mtx_lock(&procid_lock);
357 	bit_set(proc_id_array[type], id);
358 	mtx_unlock(&procid_lock);
359 }
360 
361 void
362 proc_id_clear(int type, pid_t id)
363 {
364 
365 	KASSERT(type >= 0 && type < nitems(proc_id_array),
366 	    ("invalid type %d\n", type));
367 	mtx_lock(&procid_lock);
368 	KASSERT(bit_test(proc_id_array[type], id) != 0,
369 	    ("bit %d not set in %d\n", id, type));
370 	bit_clear(proc_id_array[type], id);
371 	mtx_unlock(&procid_lock);
372 }
373 
374 /*
375  * Is p an inferior of the current process?
376  */
377 int
378 inferior(struct proc *p)
379 {
380 
381 	sx_assert(&proctree_lock, SX_LOCKED);
382 	PROC_LOCK_ASSERT(p, MA_OWNED);
383 	for (; p != curproc; p = proc_realparent(p)) {
384 		if (p->p_pid == 0)
385 			return (0);
386 	}
387 	return (1);
388 }
389 
390 /*
391  * Shared lock all the pid hash lists.
392  */
393 void
394 pidhash_slockall(void)
395 {
396 	u_long i;
397 
398 	for (i = 0; i < pidhashlock + 1; i++)
399 		sx_slock(&pidhashtbl_lock[i]);
400 }
401 
402 /*
403  * Shared unlock all the pid hash lists.
404  */
405 void
406 pidhash_sunlockall(void)
407 {
408 	u_long i;
409 
410 	for (i = 0; i < pidhashlock + 1; i++)
411 		sx_sunlock(&pidhashtbl_lock[i]);
412 }
413 
414 /*
415  * Similar to pfind(), this function locate a process by number.
416  */
417 struct proc *
418 pfind_any_locked(pid_t pid)
419 {
420 	struct proc *p;
421 
422 	sx_assert(PIDHASHLOCK(pid), SX_LOCKED);
423 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
424 		if (p->p_pid == pid) {
425 			PROC_LOCK(p);
426 			if (p->p_state == PRS_NEW) {
427 				PROC_UNLOCK(p);
428 				p = NULL;
429 			}
430 			break;
431 		}
432 	}
433 	return (p);
434 }
435 
436 /*
437  * Locate a process by number.
438  *
439  * By not returning processes in the PRS_NEW state, we allow callers to avoid
440  * testing for that condition to avoid dereferencing p_ucred, et al.
441  */
442 static __always_inline struct proc *
443 _pfind(pid_t pid, bool zombie)
444 {
445 	struct proc *p;
446 
447 	p = curproc;
448 	if (p->p_pid == pid) {
449 		PROC_LOCK(p);
450 		return (p);
451 	}
452 	sx_slock(PIDHASHLOCK(pid));
453 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
454 		if (p->p_pid == pid) {
455 			PROC_LOCK(p);
456 			if (p->p_state == PRS_NEW ||
457 			    (!zombie && p->p_state == PRS_ZOMBIE)) {
458 				PROC_UNLOCK(p);
459 				p = NULL;
460 			}
461 			break;
462 		}
463 	}
464 	sx_sunlock(PIDHASHLOCK(pid));
465 	return (p);
466 }
467 
468 struct proc *
469 pfind(pid_t pid)
470 {
471 
472 	return (_pfind(pid, false));
473 }
474 
475 /*
476  * Same as pfind but allow zombies.
477  */
478 struct proc *
479 pfind_any(pid_t pid)
480 {
481 
482 	return (_pfind(pid, true));
483 }
484 
485 /*
486  * Locate a process group by number.
487  * The caller must hold proctree_lock.
488  */
489 struct pgrp *
490 pgfind(pid_t pgid)
491 {
492 	struct pgrp *pgrp;
493 
494 	sx_assert(&proctree_lock, SX_LOCKED);
495 
496 	LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
497 		if (pgrp->pg_id == pgid) {
498 			PGRP_LOCK(pgrp);
499 			return (pgrp);
500 		}
501 	}
502 	return (NULL);
503 }
504 
505 /*
506  * Locate process and do additional manipulations, depending on flags.
507  */
508 int
509 pget(pid_t pid, int flags, struct proc **pp)
510 {
511 	struct proc *p;
512 	struct thread *td1;
513 	int error;
514 
515 	p = curproc;
516 	if (p->p_pid == pid) {
517 		PROC_LOCK(p);
518 	} else {
519 		p = NULL;
520 		if (pid <= PID_MAX) {
521 			if ((flags & PGET_NOTWEXIT) == 0)
522 				p = pfind_any(pid);
523 			else
524 				p = pfind(pid);
525 		} else if ((flags & PGET_NOTID) == 0) {
526 			td1 = tdfind(pid, -1);
527 			if (td1 != NULL)
528 				p = td1->td_proc;
529 		}
530 		if (p == NULL)
531 			return (ESRCH);
532 		if ((flags & PGET_CANSEE) != 0) {
533 			error = p_cansee(curthread, p);
534 			if (error != 0)
535 				goto errout;
536 		}
537 	}
538 	if ((flags & PGET_CANDEBUG) != 0) {
539 		error = p_candebug(curthread, p);
540 		if (error != 0)
541 			goto errout;
542 	}
543 	if ((flags & PGET_ISCURRENT) != 0 && curproc != p) {
544 		error = EPERM;
545 		goto errout;
546 	}
547 	if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) {
548 		error = ESRCH;
549 		goto errout;
550 	}
551 	if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) {
552 		/*
553 		 * XXXRW: Not clear ESRCH is the right error during proc
554 		 * execve().
555 		 */
556 		error = ESRCH;
557 		goto errout;
558 	}
559 	if ((flags & PGET_HOLD) != 0) {
560 		_PHOLD(p);
561 		PROC_UNLOCK(p);
562 	}
563 	*pp = p;
564 	return (0);
565 errout:
566 	PROC_UNLOCK(p);
567 	return (error);
568 }
569 
570 /*
571  * Create a new process group.
572  * pgid must be equal to the pid of p.
573  * Begin a new session if required.
574  */
575 int
576 enterpgrp(struct proc *p, pid_t pgid, struct pgrp *pgrp, struct session *sess)
577 {
578 	struct pgrp *old_pgrp;
579 
580 	sx_assert(&proctree_lock, SX_XLOCKED);
581 
582 	KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
583 	KASSERT(p->p_pid == pgid,
584 	    ("enterpgrp: new pgrp and pid != pgid"));
585 	KASSERT(pgfind(pgid) == NULL,
586 	    ("enterpgrp: pgrp with pgid exists"));
587 	KASSERT(!SESS_LEADER(p),
588 	    ("enterpgrp: session leader attempted setpgrp"));
589 
590 	old_pgrp = p->p_pgrp;
591 	if (!sx_try_xlock(&old_pgrp->pg_killsx)) {
592 		sx_xunlock(&proctree_lock);
593 		sx_xlock(&old_pgrp->pg_killsx);
594 		sx_xunlock(&old_pgrp->pg_killsx);
595 		return (ERESTART);
596 	}
597 	MPASS(old_pgrp == p->p_pgrp);
598 
599 	if (sess != NULL) {
600 		/*
601 		 * new session
602 		 */
603 		mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
604 		PROC_LOCK(p);
605 		p->p_flag &= ~P_CONTROLT;
606 		PROC_UNLOCK(p);
607 		PGRP_LOCK(pgrp);
608 		sess->s_leader = p;
609 		sess->s_sid = p->p_pid;
610 		proc_id_set(PROC_ID_SESSION, p->p_pid);
611 		refcount_init(&sess->s_count, 1);
612 		sess->s_ttyvp = NULL;
613 		sess->s_ttydp = NULL;
614 		sess->s_ttyp = NULL;
615 		bcopy(p->p_session->s_login, sess->s_login,
616 			    sizeof(sess->s_login));
617 		pgrp->pg_session = sess;
618 		KASSERT(p == curproc,
619 		    ("enterpgrp: mksession and p != curproc"));
620 	} else {
621 		pgrp->pg_session = p->p_session;
622 		sess_hold(pgrp->pg_session);
623 		PGRP_LOCK(pgrp);
624 	}
625 	pgrp->pg_id = pgid;
626 	proc_id_set(PROC_ID_GROUP, p->p_pid);
627 	LIST_INIT(&pgrp->pg_members);
628 	pgrp->pg_flags = 0;
629 
630 	/*
631 	 * As we have an exclusive lock of proctree_lock,
632 	 * this should not deadlock.
633 	 */
634 	LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
635 	SLIST_INIT(&pgrp->pg_sigiolst);
636 	PGRP_UNLOCK(pgrp);
637 
638 	doenterpgrp(p, pgrp);
639 
640 	sx_xunlock(&old_pgrp->pg_killsx);
641 	return (0);
642 }
643 
644 /*
645  * Move p to an existing process group
646  */
647 int
648 enterthispgrp(struct proc *p, struct pgrp *pgrp)
649 {
650 	struct pgrp *old_pgrp;
651 
652 	sx_assert(&proctree_lock, SX_XLOCKED);
653 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
654 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
655 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
656 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
657 	KASSERT(pgrp->pg_session == p->p_session,
658 	    ("%s: pgrp's session %p, p->p_session %p proc %p\n",
659 	    __func__, pgrp->pg_session, p->p_session, p));
660 	KASSERT(pgrp != p->p_pgrp,
661 	    ("%s: p %p belongs to pgrp %p", __func__, p, pgrp));
662 
663 	old_pgrp = p->p_pgrp;
664 	if (!sx_try_xlock(&old_pgrp->pg_killsx)) {
665 		sx_xunlock(&proctree_lock);
666 		sx_xlock(&old_pgrp->pg_killsx);
667 		sx_xunlock(&old_pgrp->pg_killsx);
668 		return (ERESTART);
669 	}
670 	MPASS(old_pgrp == p->p_pgrp);
671 	if (!sx_try_xlock(&pgrp->pg_killsx)) {
672 		sx_xunlock(&old_pgrp->pg_killsx);
673 		sx_xunlock(&proctree_lock);
674 		sx_xlock(&pgrp->pg_killsx);
675 		sx_xunlock(&pgrp->pg_killsx);
676 		return (ERESTART);
677 	}
678 
679 	doenterpgrp(p, pgrp);
680 
681 	sx_xunlock(&pgrp->pg_killsx);
682 	sx_xunlock(&old_pgrp->pg_killsx);
683 	return (0);
684 }
685 
686 /*
687  * If true, any child of q which belongs to group pgrp, qualifies the
688  * process group pgrp as not orphaned.
689  */
690 static bool
691 isjobproc(struct proc *q, struct pgrp *pgrp)
692 {
693 	sx_assert(&proctree_lock, SX_LOCKED);
694 
695 	return (q->p_pgrp != pgrp &&
696 	    q->p_pgrp->pg_session == pgrp->pg_session);
697 }
698 
699 static struct proc *
700 jobc_reaper(struct proc *p)
701 {
702 	struct proc *pp;
703 
704 	sx_assert(&proctree_lock, SA_LOCKED);
705 
706 	for (pp = p;;) {
707 		pp = pp->p_reaper;
708 		if (pp->p_reaper == pp ||
709 		    (pp->p_treeflag & P_TREE_GRPEXITED) == 0)
710 			return (pp);
711 	}
712 }
713 
714 static struct proc *
715 jobc_parent(struct proc *p, struct proc *p_exiting)
716 {
717 	struct proc *pp;
718 
719 	sx_assert(&proctree_lock, SA_LOCKED);
720 
721 	pp = proc_realparent(p);
722 	if (pp->p_pptr == NULL || pp == p_exiting ||
723 	    (pp->p_treeflag & P_TREE_GRPEXITED) == 0)
724 		return (pp);
725 	return (jobc_reaper(pp));
726 }
727 
728 int
729 pgrp_calc_jobc(struct pgrp *pgrp)
730 {
731 	struct proc *q;
732 	int cnt;
733 
734 #ifdef INVARIANTS
735 	if (!mtx_owned(&pgrp->pg_mtx))
736 		sx_assert(&proctree_lock, SA_LOCKED);
737 #endif
738 
739 	cnt = 0;
740 	LIST_FOREACH(q, &pgrp->pg_members, p_pglist) {
741 		if ((q->p_treeflag & P_TREE_GRPEXITED) != 0 ||
742 		    q->p_pptr == NULL)
743 			continue;
744 		if (isjobproc(jobc_parent(q, NULL), pgrp))
745 			cnt++;
746 	}
747 	return (cnt);
748 }
749 
750 /*
751  * Move p to a process group
752  */
753 static void
754 doenterpgrp(struct proc *p, struct pgrp *pgrp)
755 {
756 	struct pgrp *savepgrp;
757 	struct proc *pp;
758 
759 	sx_assert(&proctree_lock, SX_XLOCKED);
760 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
761 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
762 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
763 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
764 
765 	savepgrp = p->p_pgrp;
766 	pp = jobc_parent(p, NULL);
767 
768 	PGRP_LOCK(pgrp);
769 	PGRP_LOCK(savepgrp);
770 	if (isjobproc(pp, savepgrp) && pgrp_calc_jobc(savepgrp) == 1)
771 		orphanpg(savepgrp);
772 	PROC_LOCK(p);
773 	LIST_REMOVE(p, p_pglist);
774 	p->p_pgrp = pgrp;
775 	PROC_UNLOCK(p);
776 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
777 	if (isjobproc(pp, pgrp))
778 		pgrp->pg_flags &= ~PGRP_ORPHANED;
779 	PGRP_UNLOCK(savepgrp);
780 	PGRP_UNLOCK(pgrp);
781 	if (LIST_EMPTY(&savepgrp->pg_members))
782 		pgdelete(savepgrp);
783 }
784 
785 /*
786  * remove process from process group
787  */
788 int
789 leavepgrp(struct proc *p)
790 {
791 	struct pgrp *savepgrp;
792 
793 	sx_assert(&proctree_lock, SX_XLOCKED);
794 	savepgrp = p->p_pgrp;
795 	PGRP_LOCK(savepgrp);
796 	PROC_LOCK(p);
797 	LIST_REMOVE(p, p_pglist);
798 	p->p_pgrp = NULL;
799 	PROC_UNLOCK(p);
800 	PGRP_UNLOCK(savepgrp);
801 	if (LIST_EMPTY(&savepgrp->pg_members))
802 		pgdelete(savepgrp);
803 	return (0);
804 }
805 
806 /*
807  * delete a process group
808  */
809 static void
810 pgdelete(struct pgrp *pgrp)
811 {
812 	struct session *savesess;
813 	struct tty *tp;
814 
815 	sx_assert(&proctree_lock, SX_XLOCKED);
816 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
817 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
818 
819 	/*
820 	 * Reset any sigio structures pointing to us as a result of
821 	 * F_SETOWN with our pgid.  The proctree lock ensures that
822 	 * new sigio structures will not be added after this point.
823 	 */
824 	funsetownlst(&pgrp->pg_sigiolst);
825 
826 	PGRP_LOCK(pgrp);
827 	tp = pgrp->pg_session->s_ttyp;
828 	LIST_REMOVE(pgrp, pg_hash);
829 	savesess = pgrp->pg_session;
830 	PGRP_UNLOCK(pgrp);
831 
832 	/* Remove the reference to the pgrp before deallocating it. */
833 	if (tp != NULL) {
834 		tty_lock(tp);
835 		tty_rel_pgrp(tp, pgrp);
836 	}
837 
838 	proc_id_clear(PROC_ID_GROUP, pgrp->pg_id);
839 	uma_zfree(pgrp_zone, pgrp);
840 	sess_release(savesess);
841 }
842 
843 
844 static void
845 fixjobc_kill(struct proc *p)
846 {
847 	struct proc *q;
848 	struct pgrp *pgrp;
849 
850 	sx_assert(&proctree_lock, SX_LOCKED);
851 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
852 	pgrp = p->p_pgrp;
853 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
854 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
855 
856 	/*
857 	 * p no longer affects process group orphanage for children.
858 	 * It is marked by the flag because p is only physically
859 	 * removed from its process group on wait(2).
860 	 */
861 	MPASS((p->p_treeflag & P_TREE_GRPEXITED) == 0);
862 	p->p_treeflag |= P_TREE_GRPEXITED;
863 
864 	/*
865 	 * Check if exiting p orphans its own group.
866 	 */
867 	pgrp = p->p_pgrp;
868 	if (isjobproc(jobc_parent(p, NULL), pgrp)) {
869 		PGRP_LOCK(pgrp);
870 		if (pgrp_calc_jobc(pgrp) == 0)
871 			orphanpg(pgrp);
872 		PGRP_UNLOCK(pgrp);
873 	}
874 
875 	/*
876 	 * Check this process' children to see whether they qualify
877 	 * their process groups after reparenting to reaper.
878 	 */
879 	LIST_FOREACH(q, &p->p_children, p_sibling) {
880 		pgrp = q->p_pgrp;
881 		PGRP_LOCK(pgrp);
882 		if (pgrp_calc_jobc(pgrp) == 0) {
883 			/*
884 			 * We want to handle exactly the children that
885 			 * has p as realparent.  Then, when calculating
886 			 * jobc_parent for children, we should ignore
887 			 * P_TREE_GRPEXITED flag already set on p.
888 			 */
889 			if (jobc_parent(q, p) == p && isjobproc(p, pgrp))
890 				orphanpg(pgrp);
891 		} else
892 			pgrp->pg_flags &= ~PGRP_ORPHANED;
893 		PGRP_UNLOCK(pgrp);
894 	}
895 	LIST_FOREACH(q, &p->p_orphans, p_orphan) {
896 		pgrp = q->p_pgrp;
897 		PGRP_LOCK(pgrp);
898 		if (pgrp_calc_jobc(pgrp) == 0) {
899 			if (isjobproc(p, pgrp))
900 				orphanpg(pgrp);
901 		} else
902 			pgrp->pg_flags &= ~PGRP_ORPHANED;
903 		PGRP_UNLOCK(pgrp);
904 	}
905 }
906 
907 void
908 killjobc(void)
909 {
910 	struct session *sp;
911 	struct tty *tp;
912 	struct proc *p;
913 	struct vnode *ttyvp;
914 
915 	p = curproc;
916 	MPASS(p->p_flag & P_WEXIT);
917 	sx_assert(&proctree_lock, SX_LOCKED);
918 
919 	if (SESS_LEADER(p)) {
920 		sp = p->p_session;
921 
922 		/*
923 		 * s_ttyp is not zero'd; we use this to indicate that
924 		 * the session once had a controlling terminal. (for
925 		 * logging and informational purposes)
926 		 */
927 		SESS_LOCK(sp);
928 		ttyvp = sp->s_ttyvp;
929 		tp = sp->s_ttyp;
930 		sp->s_ttyvp = NULL;
931 		sp->s_ttydp = NULL;
932 		sp->s_leader = NULL;
933 		SESS_UNLOCK(sp);
934 
935 		/*
936 		 * Signal foreground pgrp and revoke access to
937 		 * controlling terminal if it has not been revoked
938 		 * already.
939 		 *
940 		 * Because the TTY may have been revoked in the mean
941 		 * time and could already have a new session associated
942 		 * with it, make sure we don't send a SIGHUP to a
943 		 * foreground process group that does not belong to this
944 		 * session.
945 		 */
946 
947 		if (tp != NULL) {
948 			tty_lock(tp);
949 			if (tp->t_session == sp)
950 				tty_signal_pgrp(tp, SIGHUP);
951 			tty_unlock(tp);
952 		}
953 
954 		if (ttyvp != NULL) {
955 			sx_xunlock(&proctree_lock);
956 			if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) {
957 				VOP_REVOKE(ttyvp, REVOKEALL);
958 				VOP_UNLOCK(ttyvp);
959 			}
960 			devfs_ctty_unref(ttyvp);
961 			sx_xlock(&proctree_lock);
962 		}
963 	}
964 	fixjobc_kill(p);
965 }
966 
967 /*
968  * A process group has become orphaned, mark it as such for signal
969  * delivery code.  If there are any stopped processes in the group,
970  * hang-up all process in that group.
971  */
972 static void
973 orphanpg(struct pgrp *pg)
974 {
975 	struct proc *p;
976 
977 	PGRP_LOCK_ASSERT(pg, MA_OWNED);
978 
979 	pg->pg_flags |= PGRP_ORPHANED;
980 
981 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
982 		PROC_LOCK(p);
983 		if (P_SHOULDSTOP(p) == P_STOPPED_SIG) {
984 			PROC_UNLOCK(p);
985 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
986 				PROC_LOCK(p);
987 				kern_psignal(p, SIGHUP);
988 				kern_psignal(p, SIGCONT);
989 				PROC_UNLOCK(p);
990 			}
991 			return;
992 		}
993 		PROC_UNLOCK(p);
994 	}
995 }
996 
997 void
998 sess_hold(struct session *s)
999 {
1000 
1001 	refcount_acquire(&s->s_count);
1002 }
1003 
1004 void
1005 sess_release(struct session *s)
1006 {
1007 
1008 	if (refcount_release(&s->s_count)) {
1009 		if (s->s_ttyp != NULL) {
1010 			tty_lock(s->s_ttyp);
1011 			tty_rel_sess(s->s_ttyp, s);
1012 		}
1013 		proc_id_clear(PROC_ID_SESSION, s->s_sid);
1014 		mtx_destroy(&s->s_mtx);
1015 		free(s, M_SESSION);
1016 	}
1017 }
1018 
1019 #ifdef DDB
1020 
1021 static void
1022 db_print_pgrp_one(struct pgrp *pgrp, struct proc *p)
1023 {
1024 	db_printf(
1025 	    "    pid %d at %p pr %d pgrp %p e %d jc %d\n",
1026 	    p->p_pid, p, p->p_pptr == NULL ? -1 : p->p_pptr->p_pid,
1027 	    p->p_pgrp, (p->p_treeflag & P_TREE_GRPEXITED) != 0,
1028 	    p->p_pptr == NULL ? 0 : isjobproc(p->p_pptr, pgrp));
1029 }
1030 
1031 DB_SHOW_COMMAND_FLAGS(pgrpdump, pgrpdump, DB_CMD_MEMSAFE)
1032 {
1033 	struct pgrp *pgrp;
1034 	struct proc *p;
1035 	int i;
1036 
1037 	for (i = 0; i <= pgrphash; i++) {
1038 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
1039 			db_printf("indx %d\n", i);
1040 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
1041 				db_printf(
1042 			"  pgrp %p, pgid %d, sess %p, sesscnt %d, mem %p\n",
1043 				    pgrp, (int)pgrp->pg_id, pgrp->pg_session,
1044 				    pgrp->pg_session->s_count,
1045 				    LIST_FIRST(&pgrp->pg_members));
1046 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist)
1047 					db_print_pgrp_one(pgrp, p);
1048 			}
1049 		}
1050 	}
1051 }
1052 #endif /* DDB */
1053 
1054 /*
1055  * Calculate the kinfo_proc members which contain process-wide
1056  * informations.
1057  * Must be called with the target process locked.
1058  */
1059 static void
1060 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
1061 {
1062 	struct thread *td;
1063 
1064 	PROC_LOCK_ASSERT(p, MA_OWNED);
1065 
1066 	kp->ki_estcpu = 0;
1067 	kp->ki_pctcpu = 0;
1068 	FOREACH_THREAD_IN_PROC(p, td) {
1069 		thread_lock(td);
1070 		kp->ki_pctcpu += sched_pctcpu(td);
1071 		kp->ki_estcpu += sched_estcpu(td);
1072 		thread_unlock(td);
1073 	}
1074 }
1075 
1076 /*
1077  * Fill in any information that is common to all threads in the process.
1078  * Must be called with the target process locked.
1079  */
1080 static void
1081 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
1082 {
1083 	struct thread *td0;
1084 	struct ucred *cred;
1085 	struct sigacts *ps;
1086 	struct timeval boottime;
1087 
1088 	PROC_LOCK_ASSERT(p, MA_OWNED);
1089 
1090 	kp->ki_structsize = sizeof(*kp);
1091 	kp->ki_paddr = p;
1092 	kp->ki_addr =/* p->p_addr; */0; /* XXX */
1093 	kp->ki_args = p->p_args;
1094 	kp->ki_textvp = p->p_textvp;
1095 #ifdef KTRACE
1096 	kp->ki_tracep = ktr_get_tracevp(p, false);
1097 	kp->ki_traceflag = p->p_traceflag;
1098 #endif
1099 	kp->ki_fd = p->p_fd;
1100 	kp->ki_pd = p->p_pd;
1101 	kp->ki_vmspace = p->p_vmspace;
1102 	kp->ki_flag = p->p_flag;
1103 	kp->ki_flag2 = p->p_flag2;
1104 	cred = p->p_ucred;
1105 	if (cred) {
1106 		kp->ki_uid = cred->cr_uid;
1107 		kp->ki_ruid = cred->cr_ruid;
1108 		kp->ki_svuid = cred->cr_svuid;
1109 		kp->ki_cr_flags = 0;
1110 		if (cred->cr_flags & CRED_FLAG_CAPMODE)
1111 			kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
1112 		/* XXX bde doesn't like KI_NGROUPS */
1113 		if (1 + cred->cr_ngroups > KI_NGROUPS) {
1114 			kp->ki_ngroups = KI_NGROUPS;
1115 			kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
1116 		} else
1117 			kp->ki_ngroups = 1 + cred->cr_ngroups;
1118 		kp->ki_groups[0] = cred->cr_gid;
1119 		bcopy(cred->cr_groups, kp->ki_groups + 1,
1120 		    (kp->ki_ngroups - 1) * sizeof(gid_t));
1121 		kp->ki_rgid = cred->cr_rgid;
1122 		kp->ki_svgid = cred->cr_svgid;
1123 		/* If jailed(cred), emulate the old P_JAILED flag. */
1124 		if (jailed(cred)) {
1125 			kp->ki_flag |= P_JAILED;
1126 			/* If inside the jail, use 0 as a jail ID. */
1127 			if (cred->cr_prison != curthread->td_ucred->cr_prison)
1128 				kp->ki_jid = cred->cr_prison->pr_id;
1129 		}
1130 		strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
1131 		    sizeof(kp->ki_loginclass));
1132 	}
1133 	ps = p->p_sigacts;
1134 	if (ps) {
1135 		mtx_lock(&ps->ps_mtx);
1136 		kp->ki_sigignore = ps->ps_sigignore;
1137 		kp->ki_sigcatch = ps->ps_sigcatch;
1138 		mtx_unlock(&ps->ps_mtx);
1139 	}
1140 	if (p->p_state != PRS_NEW &&
1141 	    p->p_state != PRS_ZOMBIE &&
1142 	    p->p_vmspace != NULL) {
1143 		struct vmspace *vm = p->p_vmspace;
1144 
1145 		kp->ki_size = vm->vm_map.size;
1146 		kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
1147 		FOREACH_THREAD_IN_PROC(p, td0)
1148 			kp->ki_rssize += td0->td_kstack_pages;
1149 		kp->ki_swrss = vm->vm_swrss;
1150 		kp->ki_tsize = vm->vm_tsize;
1151 		kp->ki_dsize = vm->vm_dsize;
1152 		kp->ki_ssize = vm->vm_ssize;
1153 	} else if (p->p_state == PRS_ZOMBIE)
1154 		kp->ki_stat = SZOMB;
1155 	kp->ki_sflag = PS_INMEM;
1156 	/* Calculate legacy swtime as seconds since 'swtick'. */
1157 	kp->ki_swtime = (ticks - p->p_swtick) / hz;
1158 	kp->ki_pid = p->p_pid;
1159 	kp->ki_nice = p->p_nice;
1160 	kp->ki_fibnum = p->p_fibnum;
1161 	kp->ki_start = p->p_stats->p_start;
1162 	getboottime(&boottime);
1163 	timevaladd(&kp->ki_start, &boottime);
1164 	PROC_STATLOCK(p);
1165 	rufetch(p, &kp->ki_rusage);
1166 	kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
1167 	calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
1168 	PROC_STATUNLOCK(p);
1169 	calccru(p, &kp->ki_childutime, &kp->ki_childstime);
1170 	/* Some callers want child times in a single value. */
1171 	kp->ki_childtime = kp->ki_childstime;
1172 	timevaladd(&kp->ki_childtime, &kp->ki_childutime);
1173 
1174 	FOREACH_THREAD_IN_PROC(p, td0)
1175 		kp->ki_cow += td0->td_cow;
1176 
1177 	if (p->p_comm[0] != '\0')
1178 		strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
1179 	if (p->p_sysent && p->p_sysent->sv_name != NULL &&
1180 	    p->p_sysent->sv_name[0] != '\0')
1181 		strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
1182 	kp->ki_siglist = p->p_siglist;
1183 	kp->ki_xstat = KW_EXITCODE(p->p_xexit, p->p_xsig);
1184 	kp->ki_acflag = p->p_acflag;
1185 	kp->ki_lock = p->p_lock;
1186 	if (p->p_pptr) {
1187 		kp->ki_ppid = p->p_oppid;
1188 		if (p->p_flag & P_TRACED)
1189 			kp->ki_tracer = p->p_pptr->p_pid;
1190 	}
1191 }
1192 
1193 /*
1194  * Fill job-related process information.
1195  */
1196 static void
1197 fill_kinfo_proc_pgrp(struct proc *p, struct kinfo_proc *kp)
1198 {
1199 	struct tty *tp;
1200 	struct session *sp;
1201 	struct pgrp *pgrp;
1202 
1203 	sx_assert(&proctree_lock, SA_LOCKED);
1204 	PROC_LOCK_ASSERT(p, MA_OWNED);
1205 
1206 	pgrp = p->p_pgrp;
1207 	if (pgrp == NULL)
1208 		return;
1209 
1210 	kp->ki_pgid = pgrp->pg_id;
1211 	kp->ki_jobc = pgrp_calc_jobc(pgrp);
1212 
1213 	sp = pgrp->pg_session;
1214 	tp = NULL;
1215 
1216 	if (sp != NULL) {
1217 		kp->ki_sid = sp->s_sid;
1218 		SESS_LOCK(sp);
1219 		strlcpy(kp->ki_login, sp->s_login, sizeof(kp->ki_login));
1220 		if (sp->s_ttyvp)
1221 			kp->ki_kiflag |= KI_CTTY;
1222 		if (SESS_LEADER(p))
1223 			kp->ki_kiflag |= KI_SLEADER;
1224 		tp = sp->s_ttyp;
1225 		SESS_UNLOCK(sp);
1226 	}
1227 
1228 	if ((p->p_flag & P_CONTROLT) && tp != NULL) {
1229 		kp->ki_tdev = tty_udev(tp);
1230 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1231 		kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
1232 		if (tp->t_session)
1233 			kp->ki_tsid = tp->t_session->s_sid;
1234 	} else {
1235 		kp->ki_tdev = NODEV;
1236 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1237 	}
1238 }
1239 
1240 /*
1241  * Fill in information that is thread specific.  Must be called with
1242  * target process locked.  If 'preferthread' is set, overwrite certain
1243  * process-related fields that are maintained for both threads and
1244  * processes.
1245  */
1246 static void
1247 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
1248 {
1249 	struct proc *p;
1250 
1251 	p = td->td_proc;
1252 	kp->ki_tdaddr = td;
1253 	PROC_LOCK_ASSERT(p, MA_OWNED);
1254 
1255 	if (preferthread)
1256 		PROC_STATLOCK(p);
1257 	thread_lock(td);
1258 	if (td->td_wmesg != NULL)
1259 		strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
1260 	else
1261 		bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
1262 	if (strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname)) >=
1263 	    sizeof(kp->ki_tdname)) {
1264 		strlcpy(kp->ki_moretdname,
1265 		    td->td_name + sizeof(kp->ki_tdname) - 1,
1266 		    sizeof(kp->ki_moretdname));
1267 	} else {
1268 		bzero(kp->ki_moretdname, sizeof(kp->ki_moretdname));
1269 	}
1270 	if (TD_ON_LOCK(td)) {
1271 		kp->ki_kiflag |= KI_LOCKBLOCK;
1272 		strlcpy(kp->ki_lockname, td->td_lockname,
1273 		    sizeof(kp->ki_lockname));
1274 	} else {
1275 		kp->ki_kiflag &= ~KI_LOCKBLOCK;
1276 		bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
1277 	}
1278 
1279 	if (p->p_state == PRS_NORMAL) { /* approximate. */
1280 		if (TD_ON_RUNQ(td) ||
1281 		    TD_CAN_RUN(td) ||
1282 		    TD_IS_RUNNING(td)) {
1283 			kp->ki_stat = SRUN;
1284 		} else if (P_SHOULDSTOP(p)) {
1285 			kp->ki_stat = SSTOP;
1286 		} else if (TD_IS_SLEEPING(td)) {
1287 			kp->ki_stat = SSLEEP;
1288 		} else if (TD_ON_LOCK(td)) {
1289 			kp->ki_stat = SLOCK;
1290 		} else {
1291 			kp->ki_stat = SWAIT;
1292 		}
1293 	} else if (p->p_state == PRS_ZOMBIE) {
1294 		kp->ki_stat = SZOMB;
1295 	} else {
1296 		kp->ki_stat = SIDL;
1297 	}
1298 
1299 	/* Things in the thread */
1300 	kp->ki_wchan = td->td_wchan;
1301 	kp->ki_pri.pri_level = td->td_priority;
1302 	kp->ki_pri.pri_native = td->td_base_pri;
1303 
1304 	/*
1305 	 * Note: legacy fields; clamp at the old NOCPU value and/or
1306 	 * the maximum u_char CPU value.
1307 	 */
1308 	if (td->td_lastcpu == NOCPU)
1309 		kp->ki_lastcpu_old = NOCPU_OLD;
1310 	else if (td->td_lastcpu > MAXCPU_OLD)
1311 		kp->ki_lastcpu_old = MAXCPU_OLD;
1312 	else
1313 		kp->ki_lastcpu_old = td->td_lastcpu;
1314 
1315 	if (td->td_oncpu == NOCPU)
1316 		kp->ki_oncpu_old = NOCPU_OLD;
1317 	else if (td->td_oncpu > MAXCPU_OLD)
1318 		kp->ki_oncpu_old = MAXCPU_OLD;
1319 	else
1320 		kp->ki_oncpu_old = td->td_oncpu;
1321 
1322 	kp->ki_lastcpu = td->td_lastcpu;
1323 	kp->ki_oncpu = td->td_oncpu;
1324 	kp->ki_tdflags = td->td_flags;
1325 	kp->ki_tid = td->td_tid;
1326 	kp->ki_numthreads = p->p_numthreads;
1327 	kp->ki_pcb = td->td_pcb;
1328 	kp->ki_kstack = (void *)td->td_kstack;
1329 	kp->ki_slptime = (ticks - td->td_slptick) / hz;
1330 	kp->ki_pri.pri_class = td->td_pri_class;
1331 	kp->ki_pri.pri_user = td->td_user_pri;
1332 
1333 	if (preferthread) {
1334 		rufetchtd(td, &kp->ki_rusage);
1335 		kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
1336 		kp->ki_pctcpu = sched_pctcpu(td);
1337 		kp->ki_estcpu = sched_estcpu(td);
1338 		kp->ki_cow = td->td_cow;
1339 	}
1340 
1341 	/* We can't get this anymore but ps etc never used it anyway. */
1342 	kp->ki_rqindex = 0;
1343 
1344 	if (preferthread)
1345 		kp->ki_siglist = td->td_siglist;
1346 	kp->ki_sigmask = td->td_sigmask;
1347 	thread_unlock(td);
1348 	if (preferthread)
1349 		PROC_STATUNLOCK(p);
1350 
1351 	if ((td->td_pflags & TDP2_UEXTERR) != 0)
1352 		kp->ki_uerrmsg = td->td_exterr_ptr;
1353 }
1354 
1355 /*
1356  * Fill in a kinfo_proc structure for the specified process.
1357  * Must be called with the target process locked.
1358  */
1359 void
1360 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1361 {
1362 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1363 
1364 	bzero(kp, sizeof(*kp));
1365 
1366 	fill_kinfo_proc_pgrp(p,kp);
1367 	fill_kinfo_proc_only(p, kp);
1368 	fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1369 	fill_kinfo_aggregate(p, kp);
1370 }
1371 
1372 struct pstats *
1373 pstats_alloc(void)
1374 {
1375 
1376 	return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1377 }
1378 
1379 /*
1380  * Copy parts of p_stats; zero the rest of p_stats (statistics).
1381  */
1382 void
1383 pstats_fork(struct pstats *src, struct pstats *dst)
1384 {
1385 
1386 	bzero(&dst->pstat_startzero,
1387 	    __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1388 	bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1389 	    __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1390 }
1391 
1392 void
1393 pstats_free(struct pstats *ps)
1394 {
1395 
1396 	free(ps, M_SUBPROC);
1397 }
1398 
1399 #ifdef COMPAT_FREEBSD32
1400 
1401 /*
1402  * This function is typically used to copy out the kernel address, so
1403  * it can be replaced by assignment of zero.
1404  */
1405 static inline uint32_t
1406 ptr32_trim(const void *ptr)
1407 {
1408 	uintptr_t uptr;
1409 
1410 	uptr = (uintptr_t)ptr;
1411 	return ((uptr > UINT_MAX) ? 0 : uptr);
1412 }
1413 
1414 #define PTRTRIM_CP(src,dst,fld) \
1415 	do { (dst).fld = ptr32_trim((src).fld); } while (0)
1416 
1417 static void
1418 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1419 {
1420 	int i;
1421 
1422 	bzero(ki32, sizeof(struct kinfo_proc32));
1423 	ki32->ki_structsize = sizeof(struct kinfo_proc32);
1424 	CP(*ki, *ki32, ki_layout);
1425 	PTRTRIM_CP(*ki, *ki32, ki_args);
1426 	PTRTRIM_CP(*ki, *ki32, ki_paddr);
1427 	PTRTRIM_CP(*ki, *ki32, ki_addr);
1428 	PTRTRIM_CP(*ki, *ki32, ki_tracep);
1429 	PTRTRIM_CP(*ki, *ki32, ki_textvp);
1430 	PTRTRIM_CP(*ki, *ki32, ki_fd);
1431 	PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1432 	PTRTRIM_CP(*ki, *ki32, ki_wchan);
1433 	CP(*ki, *ki32, ki_pid);
1434 	CP(*ki, *ki32, ki_ppid);
1435 	CP(*ki, *ki32, ki_pgid);
1436 	CP(*ki, *ki32, ki_tpgid);
1437 	CP(*ki, *ki32, ki_sid);
1438 	CP(*ki, *ki32, ki_tsid);
1439 	CP(*ki, *ki32, ki_jobc);
1440 	FU64_CP(*ki, *ki32, ki_tdev);
1441 	CP(*ki, *ki32, ki_tdev_freebsd11);
1442 	CP(*ki, *ki32, ki_siglist);
1443 	CP(*ki, *ki32, ki_sigmask);
1444 	CP(*ki, *ki32, ki_sigignore);
1445 	CP(*ki, *ki32, ki_sigcatch);
1446 	CP(*ki, *ki32, ki_uid);
1447 	CP(*ki, *ki32, ki_ruid);
1448 	CP(*ki, *ki32, ki_svuid);
1449 	CP(*ki, *ki32, ki_rgid);
1450 	CP(*ki, *ki32, ki_svgid);
1451 	CP(*ki, *ki32, ki_ngroups);
1452 	for (i = 0; i < KI_NGROUPS; i++)
1453 		CP(*ki, *ki32, ki_groups[i]);
1454 	CP(*ki, *ki32, ki_size);
1455 	CP(*ki, *ki32, ki_rssize);
1456 	CP(*ki, *ki32, ki_swrss);
1457 	CP(*ki, *ki32, ki_tsize);
1458 	CP(*ki, *ki32, ki_dsize);
1459 	CP(*ki, *ki32, ki_ssize);
1460 	CP(*ki, *ki32, ki_xstat);
1461 	CP(*ki, *ki32, ki_acflag);
1462 	CP(*ki, *ki32, ki_pctcpu);
1463 	CP(*ki, *ki32, ki_estcpu);
1464 	CP(*ki, *ki32, ki_slptime);
1465 	CP(*ki, *ki32, ki_swtime);
1466 	CP(*ki, *ki32, ki_cow);
1467 	FU64_CP(*ki, *ki32, ki_runtime);
1468 	TV_CP(*ki, *ki32, ki_start);
1469 	TV_CP(*ki, *ki32, ki_childtime);
1470 	CP(*ki, *ki32, ki_flag);
1471 	CP(*ki, *ki32, ki_kiflag);
1472 	CP(*ki, *ki32, ki_traceflag);
1473 	CP(*ki, *ki32, ki_stat);
1474 	CP(*ki, *ki32, ki_nice);
1475 	CP(*ki, *ki32, ki_lock);
1476 	CP(*ki, *ki32, ki_rqindex);
1477 	CP(*ki, *ki32, ki_oncpu);
1478 	CP(*ki, *ki32, ki_lastcpu);
1479 
1480 	/* XXX TODO: wrap cpu value as appropriate */
1481 	CP(*ki, *ki32, ki_oncpu_old);
1482 	CP(*ki, *ki32, ki_lastcpu_old);
1483 
1484 	bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1485 	bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1486 	bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1487 	bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1488 	bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1489 	bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1490 	bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1491 	bcopy(ki->ki_moretdname, ki32->ki_moretdname, MAXCOMLEN - TDNAMLEN + 1);
1492 	CP(*ki, *ki32, ki_tracer);
1493 	CP(*ki, *ki32, ki_flag2);
1494 	CP(*ki, *ki32, ki_fibnum);
1495 	CP(*ki, *ki32, ki_cr_flags);
1496 	CP(*ki, *ki32, ki_jid);
1497 	CP(*ki, *ki32, ki_numthreads);
1498 	CP(*ki, *ki32, ki_tid);
1499 	CP(*ki, *ki32, ki_pri);
1500 	freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1501 	freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1502 	PTRTRIM_CP(*ki, *ki32, ki_pcb);
1503 	PTRTRIM_CP(*ki, *ki32, ki_kstack);
1504 	PTRTRIM_CP(*ki, *ki32, ki_udata);
1505 	PTRTRIM_CP(*ki, *ki32, ki_tdaddr);
1506 	PTRTRIM_CP(*ki, *ki32, ki_pd);
1507 	CP(*ki, *ki32, ki_sflag);
1508 	CP(*ki, *ki32, ki_tdflags);
1509 	PTRTRIM_CP(*ki, *ki32, ki_uerrmsg);
1510 }
1511 #endif
1512 
1513 static ssize_t
1514 kern_proc_out_size(struct proc *p, int flags)
1515 {
1516 	ssize_t size = 0;
1517 
1518 	PROC_LOCK_ASSERT(p, MA_OWNED);
1519 
1520 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1521 #ifdef COMPAT_FREEBSD32
1522 		if ((flags & KERN_PROC_MASK32) != 0) {
1523 			size += sizeof(struct kinfo_proc32);
1524 		} else
1525 #endif
1526 			size += sizeof(struct kinfo_proc);
1527 	} else {
1528 #ifdef COMPAT_FREEBSD32
1529 		if ((flags & KERN_PROC_MASK32) != 0)
1530 			size += sizeof(struct kinfo_proc32) * p->p_numthreads;
1531 		else
1532 #endif
1533 			size += sizeof(struct kinfo_proc) * p->p_numthreads;
1534 	}
1535 	PROC_UNLOCK(p);
1536 	return (size);
1537 }
1538 
1539 int
1540 kern_proc_out(struct proc *p, struct sbuf *sb, int flags)
1541 {
1542 	struct thread *td;
1543 	struct kinfo_proc ki;
1544 #ifdef COMPAT_FREEBSD32
1545 	struct kinfo_proc32 ki32;
1546 #endif
1547 	int error;
1548 
1549 	PROC_LOCK_ASSERT(p, MA_OWNED);
1550 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1551 
1552 	error = 0;
1553 	fill_kinfo_proc(p, &ki);
1554 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1555 #ifdef COMPAT_FREEBSD32
1556 		if ((flags & KERN_PROC_MASK32) != 0) {
1557 			freebsd32_kinfo_proc_out(&ki, &ki32);
1558 			if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1559 				error = ENOMEM;
1560 		} else
1561 #endif
1562 			if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1563 				error = ENOMEM;
1564 	} else {
1565 		FOREACH_THREAD_IN_PROC(p, td) {
1566 			fill_kinfo_thread(td, &ki, 1);
1567 #ifdef COMPAT_FREEBSD32
1568 			if ((flags & KERN_PROC_MASK32) != 0) {
1569 				freebsd32_kinfo_proc_out(&ki, &ki32);
1570 				if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1571 					error = ENOMEM;
1572 			} else
1573 #endif
1574 				if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1575 					error = ENOMEM;
1576 			if (error != 0)
1577 				break;
1578 		}
1579 	}
1580 	PROC_UNLOCK(p);
1581 	return (error);
1582 }
1583 
1584 static int
1585 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
1586 {
1587 	struct sbuf sb;
1588 	struct kinfo_proc ki;
1589 	int error, error2;
1590 
1591 	if (req->oldptr == NULL)
1592 		return (SYSCTL_OUT(req, 0, kern_proc_out_size(p, flags)));
1593 
1594 	sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req);
1595 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1596 	error = kern_proc_out(p, &sb, flags);
1597 	error2 = sbuf_finish(&sb);
1598 	sbuf_delete(&sb);
1599 	if (error != 0)
1600 		return (error);
1601 	else if (error2 != 0)
1602 		return (error2);
1603 	return (0);
1604 }
1605 
1606 int
1607 proc_iterate(int (*cb)(struct proc *, void *), void *cbarg)
1608 {
1609 	struct proc *p;
1610 	int error, i, j;
1611 
1612 	for (i = 0; i < pidhashlock + 1; i++) {
1613 		sx_slock(&proctree_lock);
1614 		sx_slock(&pidhashtbl_lock[i]);
1615 		for (j = i; j <= pidhash; j += pidhashlock + 1) {
1616 			LIST_FOREACH(p, &pidhashtbl[j], p_hash) {
1617 				if (p->p_state == PRS_NEW)
1618 					continue;
1619 				error = cb(p, cbarg);
1620 				PROC_LOCK_ASSERT(p, MA_NOTOWNED);
1621 				if (error != 0) {
1622 					sx_sunlock(&pidhashtbl_lock[i]);
1623 					sx_sunlock(&proctree_lock);
1624 					return (error);
1625 				}
1626 			}
1627 		}
1628 		sx_sunlock(&pidhashtbl_lock[i]);
1629 		sx_sunlock(&proctree_lock);
1630 	}
1631 	return (0);
1632 }
1633 
1634 struct kern_proc_out_args {
1635 	struct sysctl_req *req;
1636 	int flags;
1637 	int oid_number;
1638 	int *name;
1639 };
1640 
1641 static int
1642 sysctl_kern_proc_iterate(struct proc *p, void *origarg)
1643 {
1644 	struct kern_proc_out_args *arg = origarg;
1645 	int *name = arg->name;
1646 	int oid_number = arg->oid_number;
1647 	int flags = arg->flags;
1648 	struct sysctl_req *req = arg->req;
1649 	int error = 0;
1650 
1651 	PROC_LOCK(p);
1652 
1653 	KASSERT(p->p_ucred != NULL,
1654 	    ("process credential is NULL for non-NEW proc"));
1655 	/*
1656 	 * Show a user only appropriate processes.
1657 	 */
1658 	if (p_cansee(curthread, p))
1659 		goto skip;
1660 	/*
1661 	 * TODO - make more efficient (see notes below).
1662 	 * do by session.
1663 	 */
1664 	switch (oid_number) {
1665 	case KERN_PROC_GID:
1666 		if (p->p_ucred->cr_gid != (gid_t)name[0])
1667 			goto skip;
1668 		break;
1669 
1670 	case KERN_PROC_PGRP:
1671 		/* could do this by traversing pgrp */
1672 		if (p->p_pgrp == NULL ||
1673 		    p->p_pgrp->pg_id != (pid_t)name[0])
1674 			goto skip;
1675 		break;
1676 
1677 	case KERN_PROC_RGID:
1678 		if (p->p_ucred->cr_rgid != (gid_t)name[0])
1679 			goto skip;
1680 		break;
1681 
1682 	case KERN_PROC_SESSION:
1683 		if (p->p_session == NULL ||
1684 		    p->p_session->s_sid != (pid_t)name[0])
1685 			goto skip;
1686 		break;
1687 
1688 	case KERN_PROC_TTY:
1689 		if ((p->p_flag & P_CONTROLT) == 0 ||
1690 		    p->p_session == NULL)
1691 			goto skip;
1692 		/* XXX proctree_lock */
1693 		SESS_LOCK(p->p_session);
1694 		if (p->p_session->s_ttyp == NULL ||
1695 		    tty_udev(p->p_session->s_ttyp) !=
1696 		    (dev_t)name[0]) {
1697 			SESS_UNLOCK(p->p_session);
1698 			goto skip;
1699 		}
1700 		SESS_UNLOCK(p->p_session);
1701 		break;
1702 
1703 	case KERN_PROC_UID:
1704 		if (p->p_ucred->cr_uid != (uid_t)name[0])
1705 			goto skip;
1706 		break;
1707 
1708 	case KERN_PROC_RUID:
1709 		if (p->p_ucred->cr_ruid != (uid_t)name[0])
1710 			goto skip;
1711 		break;
1712 
1713 	case KERN_PROC_PROC:
1714 		break;
1715 
1716 	default:
1717 		break;
1718 	}
1719 	error = sysctl_out_proc(p, req, flags);
1720 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
1721 	return (error);
1722 skip:
1723 	PROC_UNLOCK(p);
1724 	return (0);
1725 }
1726 
1727 static int
1728 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1729 {
1730 	struct kern_proc_out_args iterarg;
1731 	int *name = (int *)arg1;
1732 	u_int namelen = arg2;
1733 	struct proc *p;
1734 	int flags, oid_number;
1735 	int error = 0;
1736 
1737 	oid_number = oidp->oid_number;
1738 	if (oid_number != KERN_PROC_ALL &&
1739 	    (oid_number & KERN_PROC_INC_THREAD) == 0)
1740 		flags = KERN_PROC_NOTHREADS;
1741 	else {
1742 		flags = 0;
1743 		oid_number &= ~KERN_PROC_INC_THREAD;
1744 	}
1745 #ifdef COMPAT_FREEBSD32
1746 	if (req->flags & SCTL_MASK32)
1747 		flags |= KERN_PROC_MASK32;
1748 #endif
1749 	if (oid_number == KERN_PROC_PID) {
1750 		if (namelen != 1)
1751 			return (EINVAL);
1752 		error = sysctl_wire_old_buffer(req, 0);
1753 		if (error)
1754 			return (error);
1755 		sx_slock(&proctree_lock);
1756 		error = pget((pid_t)name[0], PGET_CANSEE, &p);
1757 		if (error == 0)
1758 			error = sysctl_out_proc(p, req, flags);
1759 		sx_sunlock(&proctree_lock);
1760 		return (error);
1761 	}
1762 
1763 	switch (oid_number) {
1764 	case KERN_PROC_ALL:
1765 		if (namelen != 0)
1766 			return (EINVAL);
1767 		break;
1768 	case KERN_PROC_PROC:
1769 		if (namelen != 0 && namelen != 1)
1770 			return (EINVAL);
1771 		break;
1772 	default:
1773 		if (namelen != 1)
1774 			return (EINVAL);
1775 		break;
1776 	}
1777 
1778 	if (req->oldptr == NULL) {
1779 		/* overestimate by 5 procs */
1780 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1781 		if (error)
1782 			return (error);
1783 	} else {
1784 		error = sysctl_wire_old_buffer(req, 0);
1785 		if (error != 0)
1786 			return (error);
1787 	}
1788 	iterarg.flags = flags;
1789 	iterarg.oid_number = oid_number;
1790 	iterarg.req = req;
1791 	iterarg.name = name;
1792 	error = proc_iterate(sysctl_kern_proc_iterate, &iterarg);
1793 	return (error);
1794 }
1795 
1796 struct pargs *
1797 pargs_alloc(int len)
1798 {
1799 	struct pargs *pa;
1800 
1801 	pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1802 		M_WAITOK);
1803 	refcount_init(&pa->ar_ref, 1);
1804 	pa->ar_length = len;
1805 	return (pa);
1806 }
1807 
1808 static void
1809 pargs_free(struct pargs *pa)
1810 {
1811 
1812 	free(pa, M_PARGS);
1813 }
1814 
1815 void
1816 pargs_hold(struct pargs *pa)
1817 {
1818 
1819 	if (pa == NULL)
1820 		return;
1821 	refcount_acquire(&pa->ar_ref);
1822 }
1823 
1824 void
1825 pargs_drop(struct pargs *pa)
1826 {
1827 
1828 	if (pa == NULL)
1829 		return;
1830 	if (refcount_release(&pa->ar_ref))
1831 		pargs_free(pa);
1832 }
1833 
1834 static int
1835 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1836     size_t len)
1837 {
1838 	ssize_t n;
1839 
1840 	/*
1841 	 * This may return a short read if the string is shorter than the chunk
1842 	 * and is aligned at the end of the page, and the following page is not
1843 	 * mapped.
1844 	 */
1845 	n = proc_readmem(td, p, (vm_offset_t)sptr, buf, len);
1846 	if (n <= 0)
1847 		return (ENOMEM);
1848 	return (0);
1849 }
1850 
1851 #define PROC_AUXV_MAX	256	/* Safety limit on auxv size. */
1852 
1853 enum proc_vector_type {
1854 	PROC_ARG,
1855 	PROC_ENV,
1856 	PROC_AUX,
1857 };
1858 
1859 #ifdef COMPAT_FREEBSD32
1860 static int
1861 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1862     size_t *vsizep, enum proc_vector_type type)
1863 {
1864 	struct freebsd32_ps_strings pss;
1865 	Elf32_Auxinfo aux;
1866 	vm_offset_t vptr, ptr;
1867 	uint32_t *proc_vector32;
1868 	char **proc_vector;
1869 	size_t vsize, size;
1870 	int i, error;
1871 
1872 	error = 0;
1873 	if (proc_readmem(td, p, PROC_PS_STRINGS(p), &pss, sizeof(pss)) !=
1874 	    sizeof(pss))
1875 		return (ENOMEM);
1876 	switch (type) {
1877 	case PROC_ARG:
1878 		vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1879 		vsize = pss.ps_nargvstr;
1880 		if (vsize > ARG_MAX)
1881 			return (ENOEXEC);
1882 		size = vsize * sizeof(int32_t);
1883 		break;
1884 	case PROC_ENV:
1885 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1886 		vsize = pss.ps_nenvstr;
1887 		if (vsize > ARG_MAX)
1888 			return (ENOEXEC);
1889 		size = vsize * sizeof(int32_t);
1890 		break;
1891 	case PROC_AUX:
1892 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1893 		    (pss.ps_nenvstr + 1) * sizeof(int32_t);
1894 		if (vptr % 4 != 0)
1895 			return (ENOEXEC);
1896 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1897 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1898 			    sizeof(aux))
1899 				return (ENOMEM);
1900 			if (aux.a_type == AT_NULL)
1901 				break;
1902 			ptr += sizeof(aux);
1903 		}
1904 		if (aux.a_type != AT_NULL)
1905 			return (ENOEXEC);
1906 		vsize = i + 1;
1907 		size = vsize * sizeof(aux);
1908 		break;
1909 	default:
1910 		KASSERT(0, ("Wrong proc vector type: %d", type));
1911 		return (EINVAL);
1912 	}
1913 	proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1914 	if (proc_readmem(td, p, vptr, proc_vector32, size) != size) {
1915 		error = ENOMEM;
1916 		goto done;
1917 	}
1918 	if (type == PROC_AUX) {
1919 		*proc_vectorp = (char **)proc_vector32;
1920 		*vsizep = vsize;
1921 		return (0);
1922 	}
1923 	proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1924 	for (i = 0; i < (int)vsize; i++)
1925 		proc_vector[i] = PTRIN(proc_vector32[i]);
1926 	*proc_vectorp = proc_vector;
1927 	*vsizep = vsize;
1928 done:
1929 	free(proc_vector32, M_TEMP);
1930 	return (error);
1931 }
1932 #endif
1933 
1934 static int
1935 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1936     size_t *vsizep, enum proc_vector_type type)
1937 {
1938 	struct ps_strings pss;
1939 	Elf_Auxinfo aux;
1940 	vm_offset_t vptr, ptr;
1941 	char **proc_vector;
1942 	size_t vsize, size;
1943 	int i;
1944 
1945 #ifdef COMPAT_FREEBSD32
1946 	if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1947 		return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1948 #endif
1949 	if (proc_readmem(td, p, PROC_PS_STRINGS(p), &pss, sizeof(pss)) !=
1950 	    sizeof(pss))
1951 		return (ENOMEM);
1952 	switch (type) {
1953 	case PROC_ARG:
1954 		vptr = (vm_offset_t)pss.ps_argvstr;
1955 		vsize = pss.ps_nargvstr;
1956 		if (vsize > ARG_MAX)
1957 			return (ENOEXEC);
1958 		size = vsize * sizeof(char *);
1959 		break;
1960 	case PROC_ENV:
1961 		vptr = (vm_offset_t)pss.ps_envstr;
1962 		vsize = pss.ps_nenvstr;
1963 		if (vsize > ARG_MAX)
1964 			return (ENOEXEC);
1965 		size = vsize * sizeof(char *);
1966 		break;
1967 	case PROC_AUX:
1968 		/*
1969 		 * The aux array is just above env array on the stack. Check
1970 		 * that the address is naturally aligned.
1971 		 */
1972 		vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1973 		    * sizeof(char *);
1974 #if __ELF_WORD_SIZE == 64
1975 		if (vptr % sizeof(uint64_t) != 0)
1976 #else
1977 		if (vptr % sizeof(uint32_t) != 0)
1978 #endif
1979 			return (ENOEXEC);
1980 		/*
1981 		 * We count the array size reading the aux vectors from the
1982 		 * stack until AT_NULL vector is returned.  So (to keep the code
1983 		 * simple) we read the process stack twice: the first time here
1984 		 * to find the size and the second time when copying the vectors
1985 		 * to the allocated proc_vector.
1986 		 */
1987 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1988 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1989 			    sizeof(aux))
1990 				return (ENOMEM);
1991 			if (aux.a_type == AT_NULL)
1992 				break;
1993 			ptr += sizeof(aux);
1994 		}
1995 		/*
1996 		 * If the PROC_AUXV_MAX entries are iterated over, and we have
1997 		 * not reached AT_NULL, it is most likely we are reading wrong
1998 		 * data: either the process doesn't have auxv array or data has
1999 		 * been modified. Return the error in this case.
2000 		 */
2001 		if (aux.a_type != AT_NULL)
2002 			return (ENOEXEC);
2003 		vsize = i + 1;
2004 		size = vsize * sizeof(aux);
2005 		break;
2006 	default:
2007 		KASSERT(0, ("Wrong proc vector type: %d", type));
2008 		return (EINVAL); /* In case we are built without INVARIANTS. */
2009 	}
2010 	proc_vector = malloc(size, M_TEMP, M_WAITOK);
2011 	if (proc_readmem(td, p, vptr, proc_vector, size) != size) {
2012 		free(proc_vector, M_TEMP);
2013 		return (ENOMEM);
2014 	}
2015 	*proc_vectorp = proc_vector;
2016 	*vsizep = vsize;
2017 
2018 	return (0);
2019 }
2020 
2021 #define GET_PS_STRINGS_CHUNK_SZ	256	/* Chunk size (bytes) for ps_strings operations. */
2022 
2023 static int
2024 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
2025     enum proc_vector_type type)
2026 {
2027 	size_t done, len, nchr, vsize;
2028 	int error, i;
2029 	char **proc_vector, *sptr;
2030 	char pss_string[GET_PS_STRINGS_CHUNK_SZ];
2031 
2032 	PROC_ASSERT_HELD(p);
2033 
2034 	/*
2035 	 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
2036 	 */
2037 	nchr = 2 * (PATH_MAX + ARG_MAX);
2038 
2039 	error = get_proc_vector(td, p, &proc_vector, &vsize, type);
2040 	if (error != 0)
2041 		return (error);
2042 	for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
2043 		/*
2044 		 * The program may have scribbled into its argv array, e.g. to
2045 		 * remove some arguments.  If that has happened, break out
2046 		 * before trying to read from NULL.
2047 		 */
2048 		if (proc_vector[i] == NULL)
2049 			break;
2050 		for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
2051 			error = proc_read_string(td, p, sptr, pss_string,
2052 			    sizeof(pss_string));
2053 			if (error != 0)
2054 				goto done;
2055 			len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
2056 			if (done + len >= nchr)
2057 				len = nchr - done - 1;
2058 			sbuf_bcat(sb, pss_string, len);
2059 			if (len != GET_PS_STRINGS_CHUNK_SZ)
2060 				break;
2061 			done += GET_PS_STRINGS_CHUNK_SZ;
2062 		}
2063 		sbuf_bcat(sb, "", 1);
2064 		done += len + 1;
2065 	}
2066 done:
2067 	free(proc_vector, M_TEMP);
2068 	return (error);
2069 }
2070 
2071 int
2072 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
2073 {
2074 
2075 	return (get_ps_strings(curthread, p, sb, PROC_ARG));
2076 }
2077 
2078 int
2079 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
2080 {
2081 
2082 	return (get_ps_strings(curthread, p, sb, PROC_ENV));
2083 }
2084 
2085 int
2086 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb)
2087 {
2088 	size_t vsize, size;
2089 	char **auxv;
2090 	int error;
2091 
2092 	error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX);
2093 	if (error == 0) {
2094 #ifdef COMPAT_FREEBSD32
2095 		if (SV_PROC_FLAG(p, SV_ILP32) != 0)
2096 			size = vsize * sizeof(Elf32_Auxinfo);
2097 		else
2098 #endif
2099 			size = vsize * sizeof(Elf_Auxinfo);
2100 		if (sbuf_bcat(sb, auxv, size) != 0)
2101 			error = ENOMEM;
2102 		free(auxv, M_TEMP);
2103 	}
2104 	return (error);
2105 }
2106 
2107 /*
2108  * This sysctl allows a process to retrieve the argument list or process
2109  * title for another process without groping around in the address space
2110  * of the other process.  It also allow a process to set its own "process
2111  * title to a string of its own choice.
2112  */
2113 static int
2114 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
2115 {
2116 	int *name = (int *)arg1;
2117 	u_int namelen = arg2;
2118 	struct pargs *newpa, *pa;
2119 	struct proc *p;
2120 	struct sbuf sb;
2121 	int flags, error = 0, error2;
2122 	pid_t pid;
2123 
2124 	if (namelen != 1)
2125 		return (EINVAL);
2126 
2127 	p = curproc;
2128 	pid = (pid_t)name[0];
2129 	if (pid == -1) {
2130 		pid = p->p_pid;
2131 	}
2132 
2133 	/*
2134 	 * If the query is for this process and it is single-threaded, there
2135 	 * is nobody to modify pargs, thus we can just read.
2136 	 */
2137 	if (pid == p->p_pid && p->p_numthreads == 1 && req->newptr == NULL &&
2138 	    (pa = p->p_args) != NULL)
2139 		return (SYSCTL_OUT(req, pa->ar_args, pa->ar_length));
2140 
2141 	flags = PGET_CANSEE;
2142 	if (req->newptr != NULL)
2143 		flags |= PGET_ISCURRENT;
2144 	error = pget(pid, flags, &p);
2145 	if (error)
2146 		return (error);
2147 
2148 	pa = p->p_args;
2149 	if (pa != NULL) {
2150 		pargs_hold(pa);
2151 		PROC_UNLOCK(p);
2152 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
2153 		pargs_drop(pa);
2154 	} else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
2155 		_PHOLD(p);
2156 		PROC_UNLOCK(p);
2157 		sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2158 		sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2159 		error = proc_getargv(curthread, p, &sb);
2160 		error2 = sbuf_finish(&sb);
2161 		PRELE(p);
2162 		sbuf_delete(&sb);
2163 		if (error == 0 && error2 != 0)
2164 			error = error2;
2165 	} else {
2166 		PROC_UNLOCK(p);
2167 	}
2168 	if (error != 0 || req->newptr == NULL)
2169 		return (error);
2170 
2171 	if (req->newlen > ps_arg_cache_limit - sizeof(struct pargs))
2172 		return (ENOMEM);
2173 
2174 	if (req->newlen == 0) {
2175 		/*
2176 		 * Clear the argument pointer, so that we'll fetch arguments
2177 		 * with proc_getargv() until further notice.
2178 		 */
2179 		newpa = NULL;
2180 	} else {
2181 		newpa = pargs_alloc(req->newlen);
2182 		error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
2183 		if (error != 0) {
2184 			pargs_free(newpa);
2185 			return (error);
2186 		}
2187 	}
2188 	PROC_LOCK(p);
2189 	pa = p->p_args;
2190 	p->p_args = newpa;
2191 	PROC_UNLOCK(p);
2192 	pargs_drop(pa);
2193 	return (0);
2194 }
2195 
2196 /*
2197  * This sysctl allows a process to retrieve environment of another process.
2198  */
2199 static int
2200 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
2201 {
2202 	int *name = (int *)arg1;
2203 	u_int namelen = arg2;
2204 	struct proc *p;
2205 	struct sbuf sb;
2206 	int error, error2;
2207 
2208 	if (namelen != 1)
2209 		return (EINVAL);
2210 
2211 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2212 	if (error != 0)
2213 		return (error);
2214 	if ((p->p_flag & P_SYSTEM) != 0) {
2215 		PRELE(p);
2216 		return (0);
2217 	}
2218 
2219 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2220 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2221 	error = proc_getenvv(curthread, p, &sb);
2222 	error2 = sbuf_finish(&sb);
2223 	PRELE(p);
2224 	sbuf_delete(&sb);
2225 	return (error != 0 ? error : error2);
2226 }
2227 
2228 /*
2229  * This sysctl allows a process to retrieve ELF auxiliary vector of
2230  * another process.
2231  */
2232 static int
2233 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
2234 {
2235 	int *name = (int *)arg1;
2236 	u_int namelen = arg2;
2237 	struct proc *p;
2238 	struct sbuf sb;
2239 	int error, error2;
2240 
2241 	if (namelen != 1)
2242 		return (EINVAL);
2243 
2244 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2245 	if (error != 0)
2246 		return (error);
2247 	if ((p->p_flag & P_SYSTEM) != 0) {
2248 		PRELE(p);
2249 		return (0);
2250 	}
2251 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2252 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2253 	error = proc_getauxv(curthread, p, &sb);
2254 	error2 = sbuf_finish(&sb);
2255 	PRELE(p);
2256 	sbuf_delete(&sb);
2257 	return (error != 0 ? error : error2);
2258 }
2259 
2260 /*
2261  * Look up the canonical executable path running in the specified process.
2262  * It tries to return the same hardlink name as was used for execve(2).
2263  * This allows the programs that modify their behavior based on their progname,
2264  * to operate correctly.
2265  *
2266  * Result is returned in retbuf, it must not be freed, similar to vn_fullpath()
2267  *   calling conventions.
2268  * binname is a pointer to temporary string buffer of length MAXPATHLEN,
2269  *   allocated and freed by caller.
2270  * freebuf should be freed by caller, from the M_TEMP malloc type.
2271  */
2272 int
2273 proc_get_binpath(struct proc *p, char *binname, char **retbuf,
2274     char **freebuf)
2275 {
2276 	struct nameidata nd;
2277 	struct vnode *vp, *dvp;
2278 	size_t freepath_size;
2279 	int error;
2280 	bool do_fullpath;
2281 
2282 	PROC_LOCK_ASSERT(p, MA_OWNED);
2283 
2284 	vp = p->p_textvp;
2285 	if (vp == NULL) {
2286 		PROC_UNLOCK(p);
2287 		*retbuf = "";
2288 		*freebuf = NULL;
2289 		return (0);
2290 	}
2291 	vref(vp);
2292 	dvp = p->p_textdvp;
2293 	if (dvp != NULL)
2294 		vref(dvp);
2295 	if (p->p_binname != NULL)
2296 		strlcpy(binname, p->p_binname, MAXPATHLEN);
2297 	PROC_UNLOCK(p);
2298 
2299 	do_fullpath = true;
2300 	*freebuf = NULL;
2301 	if (dvp != NULL && binname[0] != '\0') {
2302 		freepath_size = MAXPATHLEN;
2303 		if (vn_fullpath_hardlink(vp, dvp, binname, strlen(binname),
2304 		    retbuf, freebuf, &freepath_size) == 0) {
2305 			/*
2306 			 * Recheck the looked up path.  The binary
2307 			 * might have been renamed or replaced, in
2308 			 * which case we should not report old name.
2309 			 */
2310 			NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, *retbuf);
2311 			error = namei(&nd);
2312 			if (error == 0) {
2313 				if (nd.ni_vp == vp)
2314 					do_fullpath = false;
2315 				vrele(nd.ni_vp);
2316 				NDFREE_PNBUF(&nd);
2317 			}
2318 		}
2319 	}
2320 	if (do_fullpath) {
2321 		free(*freebuf, M_TEMP);
2322 		*freebuf = NULL;
2323 		error = vn_fullpath(vp, retbuf, freebuf);
2324 	}
2325 	vrele(vp);
2326 	if (dvp != NULL)
2327 		vrele(dvp);
2328 	return (error);
2329 }
2330 
2331 /*
2332  * This sysctl allows a process to retrieve the path of the executable for
2333  * itself or another process.
2334  */
2335 static int
2336 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
2337 {
2338 	pid_t *pidp = (pid_t *)arg1;
2339 	unsigned int arglen = arg2;
2340 	struct proc *p;
2341 	char *retbuf, *freebuf, *binname;
2342 	int error;
2343 
2344 	if (arglen != 1)
2345 		return (EINVAL);
2346 	binname = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
2347 	binname[0] = '\0';
2348 	if (*pidp == -1) {	/* -1 means this process */
2349 		error = 0;
2350 		p = req->td->td_proc;
2351 		PROC_LOCK(p);
2352 	} else {
2353 		error = pget(*pidp, PGET_CANSEE, &p);
2354 	}
2355 
2356 	if (error == 0)
2357 		error = proc_get_binpath(p, binname, &retbuf, &freebuf);
2358 	free(binname, M_TEMP);
2359 	if (error != 0)
2360 		return (error);
2361 	error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
2362 	free(freebuf, M_TEMP);
2363 	return (error);
2364 }
2365 
2366 static int
2367 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
2368 {
2369 	struct proc *p;
2370 	char *sv_name;
2371 	int *name;
2372 	int namelen;
2373 	int error;
2374 
2375 	namelen = arg2;
2376 	if (namelen != 1)
2377 		return (EINVAL);
2378 
2379 	name = (int *)arg1;
2380 	error = pget((pid_t)name[0], PGET_CANSEE, &p);
2381 	if (error != 0)
2382 		return (error);
2383 	sv_name = p->p_sysent->sv_name;
2384 	PROC_UNLOCK(p);
2385 	return (sysctl_handle_string(oidp, sv_name, 0, req));
2386 }
2387 
2388 #ifdef KINFO_OVMENTRY_SIZE
2389 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
2390 #endif
2391 
2392 #ifdef COMPAT_FREEBSD7
2393 static int
2394 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
2395 {
2396 	vm_map_entry_t entry, tmp_entry;
2397 	unsigned int last_timestamp, namelen;
2398 	char *fullpath, *freepath;
2399 	struct kinfo_ovmentry *kve;
2400 	struct vattr va;
2401 	struct ucred *cred;
2402 	int error, *name;
2403 	struct vnode *vp;
2404 	struct proc *p;
2405 	vm_map_t map;
2406 	struct vmspace *vm;
2407 
2408 	namelen = arg2;
2409 	if (namelen != 1)
2410 		return (EINVAL);
2411 
2412 	name = (int *)arg1;
2413 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2414 	if (error != 0)
2415 		return (error);
2416 	vm = vmspace_acquire_ref(p);
2417 	if (vm == NULL) {
2418 		PRELE(p);
2419 		return (ESRCH);
2420 	}
2421 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2422 
2423 	map = &vm->vm_map;
2424 	vm_map_lock_read(map);
2425 	VM_MAP_ENTRY_FOREACH(entry, map) {
2426 		vm_object_t obj, tobj, lobj;
2427 		vm_offset_t addr;
2428 
2429 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2430 			continue;
2431 
2432 		bzero(kve, sizeof(*kve));
2433 		kve->kve_structsize = sizeof(*kve);
2434 
2435 		kve->kve_private_resident = 0;
2436 		obj = entry->object.vm_object;
2437 		if (obj != NULL) {
2438 			VM_OBJECT_RLOCK(obj);
2439 			if (obj->shadow_count == 1)
2440 				kve->kve_private_resident =
2441 				    obj->resident_page_count;
2442 		}
2443 		kve->kve_resident = 0;
2444 		addr = entry->start;
2445 		while (addr < entry->end) {
2446 			if (pmap_extract(map->pmap, addr))
2447 				kve->kve_resident++;
2448 			addr += PAGE_SIZE;
2449 		}
2450 
2451 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2452 			if (tobj != obj) {
2453 				VM_OBJECT_RLOCK(tobj);
2454 				kve->kve_offset += tobj->backing_object_offset;
2455 			}
2456 			if (lobj != obj)
2457 				VM_OBJECT_RUNLOCK(lobj);
2458 			lobj = tobj;
2459 		}
2460 
2461 		kve->kve_start = (void*)entry->start;
2462 		kve->kve_end = (void*)entry->end;
2463 		kve->kve_offset += (off_t)entry->offset;
2464 
2465 		if (entry->protection & VM_PROT_READ)
2466 			kve->kve_protection |= KVME_PROT_READ;
2467 		if (entry->protection & VM_PROT_WRITE)
2468 			kve->kve_protection |= KVME_PROT_WRITE;
2469 		if (entry->protection & VM_PROT_EXECUTE)
2470 			kve->kve_protection |= KVME_PROT_EXEC;
2471 
2472 		if (entry->eflags & MAP_ENTRY_COW)
2473 			kve->kve_flags |= KVME_FLAG_COW;
2474 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2475 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2476 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2477 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2478 
2479 		last_timestamp = map->timestamp;
2480 		vm_map_unlock_read(map);
2481 
2482 		kve->kve_fileid = 0;
2483 		kve->kve_fsid = 0;
2484 		freepath = NULL;
2485 		fullpath = "";
2486 		if (lobj) {
2487 			kve->kve_type = vm_object_kvme_type(lobj, &vp);
2488 			if (kve->kve_type == KVME_TYPE_MGTDEVICE)
2489 				kve->kve_type = KVME_TYPE_UNKNOWN;
2490 			if (vp != NULL)
2491 				vref(vp);
2492 			if (lobj != obj)
2493 				VM_OBJECT_RUNLOCK(lobj);
2494 
2495 			kve->kve_ref_count = obj->ref_count;
2496 			kve->kve_shadow_count = obj->shadow_count;
2497 			VM_OBJECT_RUNLOCK(obj);
2498 			if (vp != NULL) {
2499 				vn_fullpath(vp, &fullpath, &freepath);
2500 				cred = curthread->td_ucred;
2501 				vn_lock(vp, LK_SHARED | LK_RETRY);
2502 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2503 					kve->kve_fileid = va.va_fileid;
2504 					/* truncate */
2505 					kve->kve_fsid = va.va_fsid;
2506 				}
2507 				vput(vp);
2508 			}
2509 		} else {
2510 			kve->kve_type = KVME_TYPE_NONE;
2511 			kve->kve_ref_count = 0;
2512 			kve->kve_shadow_count = 0;
2513 		}
2514 
2515 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2516 		if (freepath != NULL)
2517 			free(freepath, M_TEMP);
2518 
2519 		error = SYSCTL_OUT(req, kve, sizeof(*kve));
2520 		vm_map_lock_read(map);
2521 		if (error)
2522 			break;
2523 		if (last_timestamp != map->timestamp) {
2524 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2525 			entry = tmp_entry;
2526 		}
2527 	}
2528 	vm_map_unlock_read(map);
2529 	vmspace_free(vm);
2530 	PRELE(p);
2531 	free(kve, M_TEMP);
2532 	return (error);
2533 }
2534 #endif	/* COMPAT_FREEBSD7 */
2535 
2536 #ifdef KINFO_VMENTRY_SIZE
2537 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2538 #endif
2539 
2540 void
2541 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry,
2542     int *resident_count, bool *super)
2543 {
2544 	vm_object_t obj, tobj;
2545 	vm_page_t m, m_adv;
2546 	vm_offset_t addr;
2547 	vm_paddr_t pa;
2548 	vm_pindex_t pi, pi_adv, pindex;
2549 	int incore;
2550 
2551 	*super = false;
2552 	*resident_count = 0;
2553 	if (vmmap_skip_res_cnt)
2554 		return;
2555 
2556 	pa = 0;
2557 	obj = entry->object.vm_object;
2558 	addr = entry->start;
2559 	m_adv = NULL;
2560 	pi = OFF_TO_IDX(entry->offset);
2561 	for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) {
2562 		if (m_adv != NULL) {
2563 			m = m_adv;
2564 		} else {
2565 			pi_adv = atop(entry->end - addr);
2566 			pindex = pi;
2567 			for (tobj = obj;; tobj = tobj->backing_object) {
2568 				m = vm_radix_lookup_ge(&tobj->rtree, pindex);
2569 				if (m != NULL) {
2570 					if (m->pindex == pindex)
2571 						break;
2572 					if (pi_adv > m->pindex - pindex) {
2573 						pi_adv = m->pindex - pindex;
2574 						m_adv = m;
2575 					}
2576 				}
2577 				if (tobj->backing_object == NULL)
2578 					goto next;
2579 				pindex += OFF_TO_IDX(tobj->
2580 				    backing_object_offset);
2581 			}
2582 		}
2583 		m_adv = NULL;
2584 		if (m->psind != 0 && addr + pagesizes[1] <= entry->end &&
2585 		    (addr & (pagesizes[1] - 1)) == 0 && (incore =
2586 		    pmap_mincore(map->pmap, addr, &pa) & MINCORE_SUPER) != 0) {
2587 			*super = true;
2588 			/*
2589 			 * The virtual page might be smaller than the physical
2590 			 * page, so we use the page size reported by the pmap
2591 			 * rather than m->psind.
2592 			 */
2593 			pi_adv = atop(pagesizes[incore >> MINCORE_PSIND_SHIFT]);
2594 		} else {
2595 			/*
2596 			 * We do not test the found page on validity.
2597 			 * Either the page is busy and being paged in,
2598 			 * or it was invalidated.  The first case
2599 			 * should be counted as resident, the second
2600 			 * is not so clear; we do account both.
2601 			 */
2602 			pi_adv = 1;
2603 		}
2604 		*resident_count += pi_adv;
2605 next:;
2606 	}
2607 }
2608 
2609 /*
2610  * Must be called with the process locked and will return unlocked.
2611  */
2612 int
2613 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb, ssize_t maxlen, int flags)
2614 {
2615 	vm_map_entry_t entry, tmp_entry;
2616 	struct vattr va;
2617 	vm_map_t map;
2618 	vm_object_t lobj, nobj, obj, tobj;
2619 	char *fullpath, *freepath;
2620 	struct kinfo_vmentry *kve;
2621 	struct ucred *cred;
2622 	struct vnode *vp;
2623 	struct vmspace *vm;
2624 	vm_offset_t addr;
2625 	unsigned int last_timestamp;
2626 	int error;
2627 	key_t key;
2628 	unsigned short seq;
2629 	bool guard, super;
2630 
2631 	PROC_LOCK_ASSERT(p, MA_OWNED);
2632 
2633 	_PHOLD(p);
2634 	PROC_UNLOCK(p);
2635 	vm = vmspace_acquire_ref(p);
2636 	if (vm == NULL) {
2637 		PRELE(p);
2638 		return (ESRCH);
2639 	}
2640 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK | M_ZERO);
2641 
2642 	error = 0;
2643 	map = &vm->vm_map;
2644 	vm_map_lock_read(map);
2645 	VM_MAP_ENTRY_FOREACH(entry, map) {
2646 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2647 			continue;
2648 
2649 		addr = entry->end;
2650 		bzero(kve, sizeof(*kve));
2651 		obj = entry->object.vm_object;
2652 		if (obj != NULL) {
2653 			if ((obj->flags & OBJ_ANON) != 0)
2654 				kve->kve_obj = (uintptr_t)obj;
2655 
2656 			for (tobj = obj; tobj != NULL;
2657 			    tobj = tobj->backing_object) {
2658 				VM_OBJECT_RLOCK(tobj);
2659 				kve->kve_offset += tobj->backing_object_offset;
2660 				lobj = tobj;
2661 			}
2662 			if (obj->backing_object == NULL)
2663 				kve->kve_private_resident =
2664 				    obj->resident_page_count;
2665 			kern_proc_vmmap_resident(map, entry,
2666 			    &kve->kve_resident, &super);
2667 			if (super)
2668 				kve->kve_flags |= KVME_FLAG_SUPER;
2669 			for (tobj = obj; tobj != NULL; tobj = nobj) {
2670 				nobj = tobj->backing_object;
2671 				if (tobj != obj && tobj != lobj)
2672 					VM_OBJECT_RUNLOCK(tobj);
2673 			}
2674 		} else {
2675 			lobj = NULL;
2676 		}
2677 
2678 		kve->kve_start = entry->start;
2679 		kve->kve_end = entry->end;
2680 		kve->kve_offset += entry->offset;
2681 
2682 		if (entry->protection & VM_PROT_READ)
2683 			kve->kve_protection |= KVME_PROT_READ;
2684 		if (entry->protection & VM_PROT_WRITE)
2685 			kve->kve_protection |= KVME_PROT_WRITE;
2686 		if (entry->protection & VM_PROT_EXECUTE)
2687 			kve->kve_protection |= KVME_PROT_EXEC;
2688 		if (entry->max_protection & VM_PROT_READ)
2689 			kve->kve_protection |= KVME_MAX_PROT_READ;
2690 		if (entry->max_protection & VM_PROT_WRITE)
2691 			kve->kve_protection |= KVME_MAX_PROT_WRITE;
2692 		if (entry->max_protection & VM_PROT_EXECUTE)
2693 			kve->kve_protection |= KVME_MAX_PROT_EXEC;
2694 
2695 		if (entry->eflags & MAP_ENTRY_COW)
2696 			kve->kve_flags |= KVME_FLAG_COW;
2697 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2698 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2699 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2700 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2701 		if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2702 			kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2703 		if (entry->eflags & MAP_ENTRY_USER_WIRED)
2704 			kve->kve_flags |= KVME_FLAG_USER_WIRED;
2705 
2706 		guard = (entry->eflags & MAP_ENTRY_GUARD) != 0;
2707 
2708 		last_timestamp = map->timestamp;
2709 		vm_map_unlock_read(map);
2710 
2711 		freepath = NULL;
2712 		fullpath = "";
2713 		if (lobj != NULL) {
2714 			kve->kve_type = vm_object_kvme_type(lobj, &vp);
2715 			if (vp != NULL)
2716 				vref(vp);
2717 			if (lobj != obj)
2718 				VM_OBJECT_RUNLOCK(lobj);
2719 
2720 			kve->kve_ref_count = obj->ref_count;
2721 			kve->kve_shadow_count = obj->shadow_count;
2722 			if (obj->type == OBJT_DEVICE ||
2723 			    obj->type == OBJT_MGTDEVICE) {
2724 				cdev_pager_get_path(obj, kve->kve_path,
2725 				    sizeof(kve->kve_path));
2726 			}
2727 			VM_OBJECT_RUNLOCK(obj);
2728 			if ((lobj->flags & OBJ_SYSVSHM) != 0) {
2729 				kve->kve_flags |= KVME_FLAG_SYSVSHM;
2730 				shmobjinfo(lobj, &key, &seq);
2731 				kve->kve_vn_fileid = key;
2732 				kve->kve_vn_fsid_freebsd11 = seq;
2733 			}
2734 			if ((lobj->flags & OBJ_POSIXSHM) != 0) {
2735 				kve->kve_flags |= KVME_FLAG_POSIXSHM;
2736 				shm_get_path(lobj, kve->kve_path,
2737 				    sizeof(kve->kve_path));
2738 			}
2739 			if (vp != NULL) {
2740 				vn_fullpath(vp, &fullpath, &freepath);
2741 				kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2742 				cred = curthread->td_ucred;
2743 				vn_lock(vp, LK_SHARED | LK_RETRY);
2744 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2745 					kve->kve_vn_fileid = va.va_fileid;
2746 					kve->kve_vn_fsid = va.va_fsid;
2747 					kve->kve_vn_fsid_freebsd11 =
2748 					    kve->kve_vn_fsid; /* truncate */
2749 					kve->kve_vn_mode =
2750 					    MAKEIMODE(va.va_type, va.va_mode);
2751 					kve->kve_vn_size = va.va_size;
2752 					kve->kve_vn_rdev = va.va_rdev;
2753 					kve->kve_vn_rdev_freebsd11 =
2754 					    kve->kve_vn_rdev; /* truncate */
2755 					kve->kve_status = KF_ATTR_VALID;
2756 				}
2757 				vput(vp);
2758 				strlcpy(kve->kve_path, fullpath, sizeof(
2759 				    kve->kve_path));
2760 				free(freepath, M_TEMP);
2761 			}
2762 		} else {
2763 			kve->kve_type = guard ? KVME_TYPE_GUARD :
2764 			    KVME_TYPE_NONE;
2765 			kve->kve_ref_count = 0;
2766 			kve->kve_shadow_count = 0;
2767 		}
2768 
2769 		/* Pack record size down */
2770 		if ((flags & KERN_VMMAP_PACK_KINFO) != 0)
2771 			kve->kve_structsize =
2772 			    offsetof(struct kinfo_vmentry, kve_path) +
2773 			    strlen(kve->kve_path) + 1;
2774 		else
2775 			kve->kve_structsize = sizeof(*kve);
2776 		kve->kve_structsize = roundup(kve->kve_structsize,
2777 		    sizeof(uint64_t));
2778 
2779 		/* Halt filling and truncate rather than exceeding maxlen */
2780 		if (maxlen != -1 && maxlen < kve->kve_structsize) {
2781 			error = 0;
2782 			vm_map_lock_read(map);
2783 			break;
2784 		} else if (maxlen != -1)
2785 			maxlen -= kve->kve_structsize;
2786 
2787 		if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0)
2788 			error = ENOMEM;
2789 		vm_map_lock_read(map);
2790 		if (error != 0)
2791 			break;
2792 		if (last_timestamp != map->timestamp) {
2793 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2794 			entry = tmp_entry;
2795 		}
2796 	}
2797 	vm_map_unlock_read(map);
2798 	vmspace_free(vm);
2799 	PRELE(p);
2800 	free(kve, M_TEMP);
2801 	return (error);
2802 }
2803 
2804 static int
2805 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2806 {
2807 	struct proc *p;
2808 	struct sbuf sb;
2809 	u_int namelen;
2810 	int error, error2, *name;
2811 
2812 	namelen = arg2;
2813 	if (namelen != 1)
2814 		return (EINVAL);
2815 
2816 	name = (int *)arg1;
2817 	sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req);
2818 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2819 	error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
2820 	if (error != 0) {
2821 		sbuf_delete(&sb);
2822 		return (error);
2823 	}
2824 	error = kern_proc_vmmap_out(p, &sb, -1, KERN_VMMAP_PACK_KINFO);
2825 	error2 = sbuf_finish(&sb);
2826 	sbuf_delete(&sb);
2827 	return (error != 0 ? error : error2);
2828 }
2829 
2830 #if defined(STACK) || defined(DDB)
2831 static int
2832 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2833 {
2834 	struct kinfo_kstack *kkstp;
2835 	int error, i, *name, numthreads;
2836 	lwpid_t *lwpidarray;
2837 	struct thread *td;
2838 	struct stack *st;
2839 	struct sbuf sb;
2840 	struct proc *p;
2841 	u_int namelen;
2842 
2843 	namelen = arg2;
2844 	if (namelen != 1)
2845 		return (EINVAL);
2846 
2847 	name = (int *)arg1;
2848 	error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2849 	if (error != 0)
2850 		return (error);
2851 
2852 	kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2853 	st = stack_create(M_WAITOK);
2854 
2855 	lwpidarray = NULL;
2856 	PROC_LOCK(p);
2857 	do {
2858 		if (lwpidarray != NULL) {
2859 			free(lwpidarray, M_TEMP);
2860 			lwpidarray = NULL;
2861 		}
2862 		numthreads = p->p_numthreads;
2863 		PROC_UNLOCK(p);
2864 		lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2865 		    M_WAITOK | M_ZERO);
2866 		PROC_LOCK(p);
2867 	} while (numthreads < p->p_numthreads);
2868 
2869 	/*
2870 	 * XXXRW: During the below loop, execve(2) and countless other sorts
2871 	 * of changes could have taken place.  Should we check to see if the
2872 	 * vmspace has been replaced, or the like, in order to prevent
2873 	 * giving a snapshot that spans, say, execve(2), with some threads
2874 	 * before and some after?  Among other things, the credentials could
2875 	 * have changed, in which case the right to extract debug info might
2876 	 * no longer be assured.
2877 	 */
2878 	i = 0;
2879 	FOREACH_THREAD_IN_PROC(p, td) {
2880 		KASSERT(i < numthreads,
2881 		    ("sysctl_kern_proc_kstack: numthreads"));
2882 		lwpidarray[i] = td->td_tid;
2883 		i++;
2884 	}
2885 	PROC_UNLOCK(p);
2886 	numthreads = i;
2887 	for (i = 0; i < numthreads; i++) {
2888 		td = tdfind(lwpidarray[i], p->p_pid);
2889 		if (td == NULL) {
2890 			continue;
2891 		}
2892 		bzero(kkstp, sizeof(*kkstp));
2893 		(void)sbuf_new(&sb, kkstp->kkst_trace,
2894 		    sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2895 		thread_lock(td);
2896 		kkstp->kkst_tid = td->td_tid;
2897 		if (stack_save_td(st, td) == 0)
2898 			kkstp->kkst_state = KKST_STATE_STACKOK;
2899 		else
2900 			kkstp->kkst_state = KKST_STATE_RUNNING;
2901 		thread_unlock(td);
2902 		PROC_UNLOCK(p);
2903 		stack_sbuf_print(&sb, st);
2904 		sbuf_finish(&sb);
2905 		sbuf_delete(&sb);
2906 		error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2907 		if (error)
2908 			break;
2909 	}
2910 	PRELE(p);
2911 	if (lwpidarray != NULL)
2912 		free(lwpidarray, M_TEMP);
2913 	stack_destroy(st);
2914 	free(kkstp, M_TEMP);
2915 	return (error);
2916 }
2917 #endif
2918 
2919 /*
2920  * This sysctl allows a process to retrieve the full list of groups from
2921  * itself or another process.
2922  */
2923 static int
2924 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2925 {
2926 	pid_t *pidp = (pid_t *)arg1;
2927 	unsigned int arglen = arg2;
2928 	struct proc *p;
2929 	struct ucred *cred;
2930 	int error;
2931 
2932 	if (arglen != 1)
2933 		return (EINVAL);
2934 	if (*pidp == -1) {	/* -1 means this process */
2935 		p = req->td->td_proc;
2936 		PROC_LOCK(p);
2937 	} else {
2938 		error = pget(*pidp, PGET_CANSEE, &p);
2939 		if (error != 0)
2940 			return (error);
2941 	}
2942 
2943 	cred = crhold(p->p_ucred);
2944 	PROC_UNLOCK(p);
2945 
2946 	error = SYSCTL_OUT(req, &cred->cr_gid, sizeof(gid_t));
2947 	if (error == 0)
2948 		error = SYSCTL_OUT(req, cred->cr_groups,
2949 		    cred->cr_ngroups * sizeof(gid_t));
2950 
2951 	crfree(cred);
2952 	return (error);
2953 }
2954 
2955 /*
2956  * This sysctl allows a process to retrieve or/and set the resource limit for
2957  * another process.
2958  */
2959 static int
2960 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2961 {
2962 	int *name = (int *)arg1;
2963 	u_int namelen = arg2;
2964 	struct rlimit rlim;
2965 	struct proc *p;
2966 	u_int which;
2967 	int flags, error;
2968 
2969 	if (namelen != 2)
2970 		return (EINVAL);
2971 
2972 	which = (u_int)name[1];
2973 	if (which >= RLIM_NLIMITS)
2974 		return (EINVAL);
2975 
2976 	if (req->newptr != NULL && req->newlen != sizeof(rlim))
2977 		return (EINVAL);
2978 
2979 	flags = PGET_HOLD | PGET_NOTWEXIT;
2980 	if (req->newptr != NULL)
2981 		flags |= PGET_CANDEBUG;
2982 	else
2983 		flags |= PGET_CANSEE;
2984 	error = pget((pid_t)name[0], flags, &p);
2985 	if (error != 0)
2986 		return (error);
2987 
2988 	/*
2989 	 * Retrieve limit.
2990 	 */
2991 	if (req->oldptr != NULL) {
2992 		PROC_LOCK(p);
2993 		lim_rlimit_proc(p, which, &rlim);
2994 		PROC_UNLOCK(p);
2995 	}
2996 	error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
2997 	if (error != 0)
2998 		goto errout;
2999 
3000 	/*
3001 	 * Set limit.
3002 	 */
3003 	if (req->newptr != NULL) {
3004 		error = SYSCTL_IN(req, &rlim, sizeof(rlim));
3005 		if (error == 0)
3006 			error = kern_proc_setrlimit(curthread, p, which, &rlim);
3007 	}
3008 
3009 errout:
3010 	PRELE(p);
3011 	return (error);
3012 }
3013 
3014 /*
3015  * This sysctl allows a process to retrieve ps_strings structure location of
3016  * another process.
3017  */
3018 static int
3019 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
3020 {
3021 	int *name = (int *)arg1;
3022 	u_int namelen = arg2;
3023 	struct proc *p;
3024 	vm_offset_t ps_strings;
3025 	int error;
3026 #ifdef COMPAT_FREEBSD32
3027 	uint32_t ps_strings32;
3028 #endif
3029 
3030 	if (namelen != 1)
3031 		return (EINVAL);
3032 
3033 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
3034 	if (error != 0)
3035 		return (error);
3036 #ifdef COMPAT_FREEBSD32
3037 	if ((req->flags & SCTL_MASK32) != 0) {
3038 		/*
3039 		 * We return 0 if the 32 bit emulation request is for a 64 bit
3040 		 * process.
3041 		 */
3042 		ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
3043 		    PTROUT(PROC_PS_STRINGS(p)) : 0;
3044 		PROC_UNLOCK(p);
3045 		error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
3046 		return (error);
3047 	}
3048 #endif
3049 	ps_strings = PROC_PS_STRINGS(p);
3050 	PROC_UNLOCK(p);
3051 	error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
3052 	return (error);
3053 }
3054 
3055 /*
3056  * This sysctl allows a process to retrieve umask of another process.
3057  */
3058 static int
3059 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
3060 {
3061 	int *name = (int *)arg1;
3062 	u_int namelen = arg2;
3063 	struct proc *p;
3064 	int error;
3065 	u_short cmask;
3066 	pid_t pid;
3067 
3068 	if (namelen != 1)
3069 		return (EINVAL);
3070 
3071 	pid = (pid_t)name[0];
3072 	p = curproc;
3073 	if (pid == p->p_pid || pid == 0) {
3074 		cmask = p->p_pd->pd_cmask;
3075 		goto out;
3076 	}
3077 
3078 	error = pget(pid, PGET_WANTREAD, &p);
3079 	if (error != 0)
3080 		return (error);
3081 
3082 	cmask = p->p_pd->pd_cmask;
3083 	PRELE(p);
3084 out:
3085 	error = SYSCTL_OUT(req, &cmask, sizeof(cmask));
3086 	return (error);
3087 }
3088 
3089 /*
3090  * This sysctl allows a process to set and retrieve binary osreldate of
3091  * another process.
3092  */
3093 static int
3094 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
3095 {
3096 	int *name = (int *)arg1;
3097 	u_int namelen = arg2;
3098 	struct proc *p;
3099 	int flags, error, osrel;
3100 
3101 	if (namelen != 1)
3102 		return (EINVAL);
3103 
3104 	if (req->newptr != NULL && req->newlen != sizeof(osrel))
3105 		return (EINVAL);
3106 
3107 	flags = PGET_HOLD | PGET_NOTWEXIT;
3108 	if (req->newptr != NULL)
3109 		flags |= PGET_CANDEBUG;
3110 	else
3111 		flags |= PGET_CANSEE;
3112 	error = pget((pid_t)name[0], flags, &p);
3113 	if (error != 0)
3114 		return (error);
3115 
3116 	error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
3117 	if (error != 0)
3118 		goto errout;
3119 
3120 	if (req->newptr != NULL) {
3121 		error = SYSCTL_IN(req, &osrel, sizeof(osrel));
3122 		if (error != 0)
3123 			goto errout;
3124 		if (osrel < 0) {
3125 			error = EINVAL;
3126 			goto errout;
3127 		}
3128 		p->p_osrel = osrel;
3129 	}
3130 errout:
3131 	PRELE(p);
3132 	return (error);
3133 }
3134 
3135 static int
3136 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS)
3137 {
3138 	int *name = (int *)arg1;
3139 	u_int namelen = arg2;
3140 	struct proc *p;
3141 	struct kinfo_sigtramp kst;
3142 	const struct sysentvec *sv;
3143 	int error;
3144 #ifdef COMPAT_FREEBSD32
3145 	struct kinfo_sigtramp32 kst32;
3146 #endif
3147 
3148 	if (namelen != 1)
3149 		return (EINVAL);
3150 
3151 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
3152 	if (error != 0)
3153 		return (error);
3154 	sv = p->p_sysent;
3155 #ifdef COMPAT_FREEBSD32
3156 	if ((req->flags & SCTL_MASK32) != 0) {
3157 		bzero(&kst32, sizeof(kst32));
3158 		if (SV_PROC_FLAG(p, SV_ILP32)) {
3159 			if (PROC_HAS_SHP(p)) {
3160 				kst32.ksigtramp_start = PROC_SIGCODE(p);
3161 				kst32.ksigtramp_end = kst32.ksigtramp_start +
3162 				    ((sv->sv_flags & SV_DSO_SIG) == 0 ?
3163 				    *sv->sv_szsigcode :
3164 				    (uintptr_t)sv->sv_szsigcode);
3165 			} else {
3166 				kst32.ksigtramp_start = PROC_PS_STRINGS(p) -
3167 				    *sv->sv_szsigcode;
3168 				kst32.ksigtramp_end = PROC_PS_STRINGS(p);
3169 			}
3170 		}
3171 		PROC_UNLOCK(p);
3172 		error = SYSCTL_OUT(req, &kst32, sizeof(kst32));
3173 		return (error);
3174 	}
3175 #endif
3176 	bzero(&kst, sizeof(kst));
3177 	if (PROC_HAS_SHP(p)) {
3178 		kst.ksigtramp_start = (char *)PROC_SIGCODE(p);
3179 		kst.ksigtramp_end = (char *)kst.ksigtramp_start +
3180 		    ((sv->sv_flags & SV_DSO_SIG) == 0 ? *sv->sv_szsigcode :
3181 		    (uintptr_t)sv->sv_szsigcode);
3182 	} else {
3183 		kst.ksigtramp_start = (char *)PROC_PS_STRINGS(p) -
3184 		    *sv->sv_szsigcode;
3185 		kst.ksigtramp_end = (char *)PROC_PS_STRINGS(p);
3186 	}
3187 	PROC_UNLOCK(p);
3188 	error = SYSCTL_OUT(req, &kst, sizeof(kst));
3189 	return (error);
3190 }
3191 
3192 static int
3193 sysctl_kern_proc_sigfastblk(SYSCTL_HANDLER_ARGS)
3194 {
3195 	int *name = (int *)arg1;
3196 	u_int namelen = arg2;
3197 	pid_t pid;
3198 	struct proc *p;
3199 	struct thread *td1;
3200 	uintptr_t addr;
3201 #ifdef COMPAT_FREEBSD32
3202 	uint32_t addr32;
3203 #endif
3204 	int error;
3205 
3206 	if (namelen != 1 || req->newptr != NULL)
3207 		return (EINVAL);
3208 
3209 	pid = (pid_t)name[0];
3210 	error = pget(pid, PGET_HOLD | PGET_NOTWEXIT | PGET_CANDEBUG, &p);
3211 	if (error != 0)
3212 		return (error);
3213 
3214 	PROC_LOCK(p);
3215 #ifdef COMPAT_FREEBSD32
3216 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3217 		if (!SV_PROC_FLAG(p, SV_ILP32)) {
3218 			error = EINVAL;
3219 			goto errlocked;
3220 		}
3221 	}
3222 #endif
3223 	if (pid <= PID_MAX) {
3224 		td1 = FIRST_THREAD_IN_PROC(p);
3225 	} else {
3226 		FOREACH_THREAD_IN_PROC(p, td1) {
3227 			if (td1->td_tid == pid)
3228 				break;
3229 		}
3230 	}
3231 	if (td1 == NULL) {
3232 		error = ESRCH;
3233 		goto errlocked;
3234 	}
3235 	/*
3236 	 * The access to the private thread flags.  It is fine as far
3237 	 * as no out-of-thin-air values are read from td_pflags, and
3238 	 * usermode read of the td_sigblock_ptr is racy inherently,
3239 	 * since target process might have already changed it
3240 	 * meantime.
3241 	 */
3242 	if ((td1->td_pflags & TDP_SIGFASTBLOCK) != 0)
3243 		addr = (uintptr_t)td1->td_sigblock_ptr;
3244 	else
3245 		error = ENOTTY;
3246 
3247 errlocked:
3248 	_PRELE(p);
3249 	PROC_UNLOCK(p);
3250 	if (error != 0)
3251 		return (error);
3252 
3253 #ifdef COMPAT_FREEBSD32
3254 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3255 		addr32 = addr;
3256 		error = SYSCTL_OUT(req, &addr32, sizeof(addr32));
3257 	} else
3258 #endif
3259 		error = SYSCTL_OUT(req, &addr, sizeof(addr));
3260 	return (error);
3261 }
3262 
3263 static int
3264 sysctl_kern_proc_vm_layout(SYSCTL_HANDLER_ARGS)
3265 {
3266 	struct kinfo_vm_layout kvm;
3267 	struct proc *p;
3268 	struct vmspace *vmspace;
3269 	int error, *name;
3270 
3271 	name = (int *)arg1;
3272 	if ((u_int)arg2 != 1)
3273 		return (EINVAL);
3274 
3275 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
3276 	if (error != 0)
3277 		return (error);
3278 #ifdef COMPAT_FREEBSD32
3279 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3280 		if (!SV_PROC_FLAG(p, SV_ILP32)) {
3281 			PROC_UNLOCK(p);
3282 			return (EINVAL);
3283 		}
3284 	}
3285 #endif
3286 	vmspace = vmspace_acquire_ref(p);
3287 	PROC_UNLOCK(p);
3288 
3289 	memset(&kvm, 0, sizeof(kvm));
3290 	kvm.kvm_min_user_addr = vm_map_min(&vmspace->vm_map);
3291 	kvm.kvm_max_user_addr = vm_map_max(&vmspace->vm_map);
3292 	kvm.kvm_text_addr = (uintptr_t)vmspace->vm_taddr;
3293 	kvm.kvm_text_size = vmspace->vm_tsize;
3294 	kvm.kvm_data_addr = (uintptr_t)vmspace->vm_daddr;
3295 	kvm.kvm_data_size = vmspace->vm_dsize;
3296 	kvm.kvm_stack_addr = (uintptr_t)vmspace->vm_maxsaddr;
3297 	kvm.kvm_stack_size = vmspace->vm_ssize;
3298 	kvm.kvm_shp_addr = vmspace->vm_shp_base;
3299 	kvm.kvm_shp_size = p->p_sysent->sv_shared_page_len;
3300 	if ((vmspace->vm_map.flags & MAP_WIREFUTURE) != 0)
3301 		kvm.kvm_map_flags |= KMAP_FLAG_WIREFUTURE;
3302 	if ((vmspace->vm_map.flags & MAP_ASLR) != 0)
3303 		kvm.kvm_map_flags |= KMAP_FLAG_ASLR;
3304 	if ((vmspace->vm_map.flags & MAP_ASLR_IGNSTART) != 0)
3305 		kvm.kvm_map_flags |= KMAP_FLAG_ASLR_IGNSTART;
3306 	if ((vmspace->vm_map.flags & MAP_WXORX) != 0)
3307 		kvm.kvm_map_flags |= KMAP_FLAG_WXORX;
3308 	if ((vmspace->vm_map.flags & MAP_ASLR_STACK) != 0)
3309 		kvm.kvm_map_flags |= KMAP_FLAG_ASLR_STACK;
3310 	if (vmspace->vm_shp_base != p->p_sysent->sv_shared_page_base &&
3311 	    PROC_HAS_SHP(p))
3312 		kvm.kvm_map_flags |= KMAP_FLAG_ASLR_SHARED_PAGE;
3313 
3314 #ifdef COMPAT_FREEBSD32
3315 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3316 		struct kinfo_vm_layout32 kvm32;
3317 
3318 		memset(&kvm32, 0, sizeof(kvm32));
3319 		kvm32.kvm_min_user_addr = (uint32_t)kvm.kvm_min_user_addr;
3320 		kvm32.kvm_max_user_addr = (uint32_t)kvm.kvm_max_user_addr;
3321 		kvm32.kvm_text_addr = (uint32_t)kvm.kvm_text_addr;
3322 		kvm32.kvm_text_size = (uint32_t)kvm.kvm_text_size;
3323 		kvm32.kvm_data_addr = (uint32_t)kvm.kvm_data_addr;
3324 		kvm32.kvm_data_size = (uint32_t)kvm.kvm_data_size;
3325 		kvm32.kvm_stack_addr = (uint32_t)kvm.kvm_stack_addr;
3326 		kvm32.kvm_stack_size = (uint32_t)kvm.kvm_stack_size;
3327 		kvm32.kvm_shp_addr = (uint32_t)kvm.kvm_shp_addr;
3328 		kvm32.kvm_shp_size = (uint32_t)kvm.kvm_shp_size;
3329 		kvm32.kvm_map_flags = kvm.kvm_map_flags;
3330 		error = SYSCTL_OUT(req, &kvm32, sizeof(kvm32));
3331 		goto out;
3332 	}
3333 #endif
3334 
3335 	error = SYSCTL_OUT(req, &kvm, sizeof(kvm));
3336 #ifdef COMPAT_FREEBSD32
3337 out:
3338 #endif
3339 	vmspace_free(vmspace);
3340 	return (error);
3341 }
3342 
3343 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,  0,
3344     "Process table");
3345 
3346 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
3347 	CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
3348 	"Return entire process table");
3349 
3350 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3351 	sysctl_kern_proc, "Process table");
3352 
3353 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
3354 	sysctl_kern_proc, "Process table");
3355 
3356 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3357 	sysctl_kern_proc, "Process table");
3358 
3359 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
3360 	CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3361 
3362 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
3363 	sysctl_kern_proc, "Process table");
3364 
3365 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3366 	sysctl_kern_proc, "Process table");
3367 
3368 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3369 	sysctl_kern_proc, "Process table");
3370 
3371 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3372 	sysctl_kern_proc, "Process table");
3373 
3374 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
3375 	sysctl_kern_proc, "Return process table, no threads");
3376 
3377 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
3378 	CTLFLAG_RW | CTLFLAG_CAPWR | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
3379 	sysctl_kern_proc_args, "Process argument list");
3380 
3381 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
3382 	sysctl_kern_proc_env, "Process environment");
3383 
3384 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
3385 	CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
3386 
3387 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
3388 	CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
3389 
3390 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
3391 	CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
3392 	"Process syscall vector name (ABI type)");
3393 
3394 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
3395 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3396 
3397 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
3398 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3399 
3400 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
3401 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3402 
3403 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
3404 	sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3405 
3406 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
3407 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3408 
3409 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
3410 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3411 
3412 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
3413 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3414 
3415 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
3416 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3417 
3418 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
3419 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
3420 	"Return process table, including threads");
3421 
3422 #ifdef COMPAT_FREEBSD7
3423 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
3424 	CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
3425 #endif
3426 
3427 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
3428 	CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
3429 
3430 #if defined(STACK) || defined(DDB)
3431 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
3432 	CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
3433 #endif
3434 
3435 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
3436 	CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
3437 
3438 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
3439 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
3440 	"Process resource limits");
3441 
3442 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
3443 	CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
3444 	"Process ps_strings location");
3445 
3446 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
3447 	CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
3448 
3449 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
3450 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
3451 	"Process binary osreldate");
3452 
3453 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD |
3454 	CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp,
3455 	"Process signal trampoline location");
3456 
3457 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGFASTBLK, sigfastblk, CTLFLAG_RD |
3458 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_sigfastblk,
3459 	"Thread sigfastblock address");
3460 
3461 static SYSCTL_NODE(_kern_proc, KERN_PROC_VM_LAYOUT, vm_layout, CTLFLAG_RD |
3462 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_vm_layout,
3463 	"Process virtual address space layout info");
3464 
3465 static struct sx stop_all_proc_blocker;
3466 SX_SYSINIT(stop_all_proc_blocker, &stop_all_proc_blocker, "sapblk");
3467 
3468 bool
3469 stop_all_proc_block(void)
3470 {
3471 	return (sx_xlock_sig(&stop_all_proc_blocker) == 0);
3472 }
3473 
3474 void
3475 stop_all_proc_unblock(void)
3476 {
3477 	sx_xunlock(&stop_all_proc_blocker);
3478 }
3479 
3480 int allproc_gen;
3481 
3482 /*
3483  * stop_all_proc() purpose is to stop all process which have usermode,
3484  * except current process for obvious reasons.  This makes it somewhat
3485  * unreliable when invoked from multithreaded process.  The service
3486  * must not be user-callable anyway.
3487  */
3488 void
3489 stop_all_proc(void)
3490 {
3491 	struct proc *cp, *p;
3492 	int r, gen;
3493 	bool restart, seen_stopped, seen_exiting, stopped_some;
3494 
3495 	if (!stop_all_proc_block())
3496 		return;
3497 
3498 	cp = curproc;
3499 allproc_loop:
3500 	sx_xlock(&allproc_lock);
3501 	gen = allproc_gen;
3502 	seen_exiting = seen_stopped = stopped_some = restart = false;
3503 	LIST_REMOVE(cp, p_list);
3504 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3505 	for (;;) {
3506 		p = LIST_NEXT(cp, p_list);
3507 		if (p == NULL)
3508 			break;
3509 		LIST_REMOVE(cp, p_list);
3510 		LIST_INSERT_AFTER(p, cp, p_list);
3511 		PROC_LOCK(p);
3512 		if ((p->p_flag & (P_KPROC | P_SYSTEM | P_TOTAL_STOP |
3513 		    P_STOPPED_SIG)) != 0) {
3514 			PROC_UNLOCK(p);
3515 			continue;
3516 		}
3517 		if ((p->p_flag2 & P2_WEXIT) != 0) {
3518 			seen_exiting = true;
3519 			PROC_UNLOCK(p);
3520 			continue;
3521 		}
3522 		if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) {
3523 			/*
3524 			 * Stopped processes are tolerated when there
3525 			 * are no other processes which might continue
3526 			 * them.  P_STOPPED_SINGLE but not
3527 			 * P_TOTAL_STOP process still has at least one
3528 			 * thread running.
3529 			 */
3530 			seen_stopped = true;
3531 			PROC_UNLOCK(p);
3532 			continue;
3533 		}
3534 		if ((p->p_flag & P_TRACED) != 0) {
3535 			/*
3536 			 * thread_single() below cannot stop traced p,
3537 			 * so skip it.  OTOH, we cannot require
3538 			 * restart because debugger might be either
3539 			 * already stopped or traced as well.
3540 			 */
3541 			PROC_UNLOCK(p);
3542 			continue;
3543 		}
3544 		sx_xunlock(&allproc_lock);
3545 		_PHOLD(p);
3546 		r = thread_single(p, SINGLE_ALLPROC);
3547 		if (r != 0)
3548 			restart = true;
3549 		else
3550 			stopped_some = true;
3551 		_PRELE(p);
3552 		PROC_UNLOCK(p);
3553 		sx_xlock(&allproc_lock);
3554 	}
3555 	/* Catch forked children we did not see in iteration. */
3556 	if (gen != allproc_gen)
3557 		restart = true;
3558 	sx_xunlock(&allproc_lock);
3559 	if (restart || stopped_some || seen_exiting || seen_stopped) {
3560 		kern_yield(PRI_USER);
3561 		goto allproc_loop;
3562 	}
3563 }
3564 
3565 void
3566 resume_all_proc(void)
3567 {
3568 	struct proc *cp, *p;
3569 
3570 	cp = curproc;
3571 	sx_xlock(&allproc_lock);
3572 again:
3573 	LIST_REMOVE(cp, p_list);
3574 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3575 	for (;;) {
3576 		p = LIST_NEXT(cp, p_list);
3577 		if (p == NULL)
3578 			break;
3579 		LIST_REMOVE(cp, p_list);
3580 		LIST_INSERT_AFTER(p, cp, p_list);
3581 		PROC_LOCK(p);
3582 		if ((p->p_flag & P_TOTAL_STOP) != 0) {
3583 			sx_xunlock(&allproc_lock);
3584 			_PHOLD(p);
3585 			thread_single_end(p, SINGLE_ALLPROC);
3586 			_PRELE(p);
3587 			PROC_UNLOCK(p);
3588 			sx_xlock(&allproc_lock);
3589 		} else {
3590 			PROC_UNLOCK(p);
3591 		}
3592 	}
3593 	/*  Did the loop above missed any stopped process ? */
3594 	FOREACH_PROC_IN_SYSTEM(p) {
3595 		/* No need for proc lock. */
3596 		if ((p->p_flag & P_TOTAL_STOP) != 0)
3597 			goto again;
3598 	}
3599 	sx_xunlock(&allproc_lock);
3600 
3601 	stop_all_proc_unblock();
3602 }
3603 
3604 /* #define	TOTAL_STOP_DEBUG	1 */
3605 #ifdef TOTAL_STOP_DEBUG
3606 volatile static int ap_resume;
3607 #include <sys/mount.h>
3608 
3609 static int
3610 sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS)
3611 {
3612 	int error, val;
3613 
3614 	val = 0;
3615 	ap_resume = 0;
3616 	error = sysctl_handle_int(oidp, &val, 0, req);
3617 	if (error != 0 || req->newptr == NULL)
3618 		return (error);
3619 	if (val != 0) {
3620 		stop_all_proc();
3621 		syncer_suspend();
3622 		while (ap_resume == 0)
3623 			;
3624 		syncer_resume();
3625 		resume_all_proc();
3626 	}
3627 	return (0);
3628 }
3629 
3630 SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW |
3631     CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0,
3632     sysctl_debug_stop_all_proc, "I",
3633     "");
3634 #endif
3635