xref: /freebsd/sys/kern/kern_proc.c (revision 85e449cfcfdf3a3ea99bae7fed9f3cd436eb21ec)
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 	refcount_init(&p->p_tree_refcnt, 0);
283 	return (0);
284 }
285 
286 /*
287  * UMA should ensure that this function is never called.
288  * Freeing a proc structure would violate type stability.
289  */
290 static void
291 proc_fini(void *mem, int size)
292 {
293 #ifdef notnow
294 	struct proc *p;
295 
296 	p = (struct proc *)mem;
297 	EVENTHANDLER_DIRECT_INVOKE(process_fini, p);
298 	pstats_free(p->p_stats);
299 	thread_free(FIRST_THREAD_IN_PROC(p));
300 	mtx_destroy(&p->p_mtx);
301 	if (p->p_ksi != NULL)
302 		ksiginfo_free(p->p_ksi);
303 #else
304 	panic("proc reclaimed");
305 #endif
306 }
307 
308 static int
309 pgrp_init(void *mem, int size, int flags)
310 {
311 	struct pgrp *pg;
312 
313 	pg = mem;
314 	mtx_init(&pg->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
315 	sx_init(&pg->pg_killsx, "killpg racer");
316 	return (0);
317 }
318 
319 /*
320  * PID space management.
321  *
322  * These bitmaps are used by fork_findpid.
323  */
324 bitstr_t bit_decl(proc_id_pidmap, PID_MAX);
325 bitstr_t bit_decl(proc_id_grpidmap, PID_MAX);
326 bitstr_t bit_decl(proc_id_sessidmap, PID_MAX);
327 bitstr_t bit_decl(proc_id_reapmap, PID_MAX);
328 
329 static bitstr_t *proc_id_array[] = {
330 	proc_id_pidmap,
331 	proc_id_grpidmap,
332 	proc_id_sessidmap,
333 	proc_id_reapmap,
334 };
335 
336 void
337 proc_id_set(int type, pid_t id)
338 {
339 
340 	KASSERT(type >= 0 && type < nitems(proc_id_array),
341 	    ("invalid type %d\n", type));
342 	mtx_lock(&procid_lock);
343 	KASSERT(bit_test(proc_id_array[type], id) == 0,
344 	    ("bit %d already set in %d\n", id, type));
345 	bit_set(proc_id_array[type], id);
346 	mtx_unlock(&procid_lock);
347 }
348 
349 void
350 proc_id_set_cond(int type, pid_t id)
351 {
352 
353 	KASSERT(type >= 0 && type < nitems(proc_id_array),
354 	    ("invalid type %d\n", type));
355 	if (bit_test(proc_id_array[type], id))
356 		return;
357 	mtx_lock(&procid_lock);
358 	bit_set(proc_id_array[type], id);
359 	mtx_unlock(&procid_lock);
360 }
361 
362 void
363 proc_id_clear(int type, pid_t id)
364 {
365 
366 	KASSERT(type >= 0 && type < nitems(proc_id_array),
367 	    ("invalid type %d\n", type));
368 	mtx_lock(&procid_lock);
369 	KASSERT(bit_test(proc_id_array[type], id) != 0,
370 	    ("bit %d not set in %d\n", id, type));
371 	bit_clear(proc_id_array[type], id);
372 	mtx_unlock(&procid_lock);
373 }
374 
375 /*
376  * Is p an inferior of the current process?
377  */
378 int
379 inferior(struct proc *p)
380 {
381 
382 	sx_assert(&proctree_lock, SX_LOCKED);
383 	PROC_LOCK_ASSERT(p, MA_OWNED);
384 	for (; p != curproc; p = proc_realparent(p)) {
385 		if (p->p_pid == 0)
386 			return (0);
387 	}
388 	return (1);
389 }
390 
391 /*
392  * Shared lock all the pid hash lists.
393  */
394 void
395 pidhash_slockall(void)
396 {
397 	u_long i;
398 
399 	for (i = 0; i < pidhashlock + 1; i++)
400 		sx_slock(&pidhashtbl_lock[i]);
401 }
402 
403 /*
404  * Shared unlock all the pid hash lists.
405  */
406 void
407 pidhash_sunlockall(void)
408 {
409 	u_long i;
410 
411 	for (i = 0; i < pidhashlock + 1; i++)
412 		sx_sunlock(&pidhashtbl_lock[i]);
413 }
414 
415 /*
416  * Similar to pfind(), this function locate a process by number.
417  */
418 struct proc *
419 pfind_any_locked(pid_t pid)
420 {
421 	struct proc *p;
422 
423 	sx_assert(PIDHASHLOCK(pid), SX_LOCKED);
424 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
425 		if (p->p_pid == pid) {
426 			PROC_LOCK(p);
427 			if (p->p_state == PRS_NEW) {
428 				PROC_UNLOCK(p);
429 				p = NULL;
430 			}
431 			break;
432 		}
433 	}
434 	return (p);
435 }
436 
437 /*
438  * Locate a process by number.
439  *
440  * By not returning processes in the PRS_NEW state, we allow callers to avoid
441  * testing for that condition to avoid dereferencing p_ucred, et al.
442  */
443 static __always_inline struct proc *
444 _pfind(pid_t pid, bool zombie)
445 {
446 	struct proc *p;
447 
448 	p = curproc;
449 	if (p->p_pid == pid) {
450 		PROC_LOCK(p);
451 		return (p);
452 	}
453 	sx_slock(PIDHASHLOCK(pid));
454 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
455 		if (p->p_pid == pid) {
456 			PROC_LOCK(p);
457 			if (p->p_state == PRS_NEW ||
458 			    (!zombie && p->p_state == PRS_ZOMBIE)) {
459 				PROC_UNLOCK(p);
460 				p = NULL;
461 			}
462 			break;
463 		}
464 	}
465 	sx_sunlock(PIDHASHLOCK(pid));
466 	return (p);
467 }
468 
469 struct proc *
470 pfind(pid_t pid)
471 {
472 
473 	return (_pfind(pid, false));
474 }
475 
476 /*
477  * Same as pfind but allow zombies.
478  */
479 struct proc *
480 pfind_any(pid_t pid)
481 {
482 
483 	return (_pfind(pid, true));
484 }
485 
486 /*
487  * Locate a process group by number.
488  * The caller must hold proctree_lock.
489  */
490 struct pgrp *
491 pgfind(pid_t pgid)
492 {
493 	struct pgrp *pgrp;
494 
495 	sx_assert(&proctree_lock, SX_LOCKED);
496 
497 	LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
498 		if (pgrp->pg_id == pgid) {
499 			PGRP_LOCK(pgrp);
500 			return (pgrp);
501 		}
502 	}
503 	return (NULL);
504 }
505 
506 /*
507  * Locate process and do additional manipulations, depending on flags.
508  */
509 int
510 pget(pid_t pid, int flags, struct proc **pp)
511 {
512 	struct proc *p;
513 	struct thread *td1;
514 	int error;
515 
516 	p = curproc;
517 	if (p->p_pid == pid) {
518 		PROC_LOCK(p);
519 	} else {
520 		p = NULL;
521 		if (pid <= PID_MAX) {
522 			if ((flags & PGET_NOTWEXIT) == 0)
523 				p = pfind_any(pid);
524 			else
525 				p = pfind(pid);
526 		} else if ((flags & PGET_NOTID) == 0) {
527 			td1 = tdfind(pid, -1);
528 			if (td1 != NULL)
529 				p = td1->td_proc;
530 		}
531 		if (p == NULL)
532 			return (ESRCH);
533 		if ((flags & PGET_CANSEE) != 0) {
534 			error = p_cansee(curthread, p);
535 			if (error != 0)
536 				goto errout;
537 		}
538 	}
539 	if ((flags & PGET_CANDEBUG) != 0) {
540 		error = p_candebug(curthread, p);
541 		if (error != 0)
542 			goto errout;
543 	}
544 	if ((flags & PGET_ISCURRENT) != 0 && curproc != p) {
545 		error = EPERM;
546 		goto errout;
547 	}
548 	if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) {
549 		error = ESRCH;
550 		goto errout;
551 	}
552 	if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) {
553 		/*
554 		 * XXXRW: Not clear ESRCH is the right error during proc
555 		 * execve().
556 		 */
557 		error = ESRCH;
558 		goto errout;
559 	}
560 	if ((flags & PGET_HOLD) != 0) {
561 		_PHOLD(p);
562 		PROC_UNLOCK(p);
563 	}
564 	*pp = p;
565 	return (0);
566 errout:
567 	PROC_UNLOCK(p);
568 	return (error);
569 }
570 
571 /*
572  * Create a new process group.
573  * pgid must be equal to the pid of p.
574  * Begin a new session if required.
575  */
576 int
577 enterpgrp(struct proc *p, pid_t pgid, struct pgrp *pgrp, struct session *sess)
578 {
579 	struct pgrp *old_pgrp;
580 
581 	sx_assert(&proctree_lock, SX_XLOCKED);
582 
583 	KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
584 	KASSERT(p->p_pid == pgid,
585 	    ("enterpgrp: new pgrp and pid != pgid"));
586 	KASSERT(pgfind(pgid) == NULL,
587 	    ("enterpgrp: pgrp with pgid exists"));
588 	KASSERT(!SESS_LEADER(p),
589 	    ("enterpgrp: session leader attempted setpgrp"));
590 
591 	old_pgrp = p->p_pgrp;
592 	if (!sx_try_xlock(&old_pgrp->pg_killsx)) {
593 		sx_xunlock(&proctree_lock);
594 		sx_xlock(&old_pgrp->pg_killsx);
595 		sx_xunlock(&old_pgrp->pg_killsx);
596 		return (ERESTART);
597 	}
598 	MPASS(old_pgrp == p->p_pgrp);
599 
600 	if (sess != NULL) {
601 		/*
602 		 * new session
603 		 */
604 		mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
605 		PROC_LOCK(p);
606 		p->p_flag &= ~P_CONTROLT;
607 		PROC_UNLOCK(p);
608 		PGRP_LOCK(pgrp);
609 		sess->s_leader = p;
610 		sess->s_sid = p->p_pid;
611 		proc_id_set(PROC_ID_SESSION, p->p_pid);
612 		refcount_init(&sess->s_count, 1);
613 		sess->s_ttyvp = NULL;
614 		sess->s_ttydp = NULL;
615 		sess->s_ttyp = NULL;
616 		bcopy(p->p_session->s_login, sess->s_login,
617 			    sizeof(sess->s_login));
618 		pgrp->pg_session = sess;
619 		KASSERT(p == curproc,
620 		    ("enterpgrp: mksession and p != curproc"));
621 	} else {
622 		pgrp->pg_session = p->p_session;
623 		sess_hold(pgrp->pg_session);
624 		PGRP_LOCK(pgrp);
625 	}
626 	pgrp->pg_id = pgid;
627 	proc_id_set(PROC_ID_GROUP, p->p_pid);
628 	LIST_INIT(&pgrp->pg_members);
629 	pgrp->pg_flags = 0;
630 
631 	/*
632 	 * As we have an exclusive lock of proctree_lock,
633 	 * this should not deadlock.
634 	 */
635 	LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
636 	SLIST_INIT(&pgrp->pg_sigiolst);
637 	PGRP_UNLOCK(pgrp);
638 
639 	doenterpgrp(p, pgrp);
640 
641 	sx_xunlock(&old_pgrp->pg_killsx);
642 	return (0);
643 }
644 
645 /*
646  * Move p to an existing process group
647  */
648 int
649 enterthispgrp(struct proc *p, struct pgrp *pgrp)
650 {
651 	struct pgrp *old_pgrp;
652 
653 	sx_assert(&proctree_lock, SX_XLOCKED);
654 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
655 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
656 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
657 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
658 	KASSERT(pgrp->pg_session == p->p_session,
659 	    ("%s: pgrp's session %p, p->p_session %p proc %p\n",
660 	    __func__, pgrp->pg_session, p->p_session, p));
661 	KASSERT(pgrp != p->p_pgrp,
662 	    ("%s: p %p belongs to pgrp %p", __func__, p, pgrp));
663 
664 	old_pgrp = p->p_pgrp;
665 	if (!sx_try_xlock(&old_pgrp->pg_killsx)) {
666 		sx_xunlock(&proctree_lock);
667 		sx_xlock(&old_pgrp->pg_killsx);
668 		sx_xunlock(&old_pgrp->pg_killsx);
669 		return (ERESTART);
670 	}
671 	MPASS(old_pgrp == p->p_pgrp);
672 	if (!sx_try_xlock(&pgrp->pg_killsx)) {
673 		sx_xunlock(&old_pgrp->pg_killsx);
674 		sx_xunlock(&proctree_lock);
675 		sx_xlock(&pgrp->pg_killsx);
676 		sx_xunlock(&pgrp->pg_killsx);
677 		return (ERESTART);
678 	}
679 
680 	doenterpgrp(p, pgrp);
681 
682 	sx_xunlock(&pgrp->pg_killsx);
683 	sx_xunlock(&old_pgrp->pg_killsx);
684 	return (0);
685 }
686 
687 /*
688  * If true, any child of q which belongs to group pgrp, qualifies the
689  * process group pgrp as not orphaned.
690  */
691 static bool
692 isjobproc(struct proc *q, struct pgrp *pgrp)
693 {
694 	sx_assert(&proctree_lock, SX_LOCKED);
695 
696 	return (q->p_pgrp != pgrp &&
697 	    q->p_pgrp->pg_session == pgrp->pg_session);
698 }
699 
700 static struct proc *
701 jobc_reaper(struct proc *p)
702 {
703 	struct proc *pp;
704 
705 	sx_assert(&proctree_lock, SA_LOCKED);
706 
707 	for (pp = p;;) {
708 		pp = pp->p_reaper;
709 		if (pp->p_reaper == pp ||
710 		    (pp->p_treeflag & P_TREE_GRPEXITED) == 0)
711 			return (pp);
712 	}
713 }
714 
715 static struct proc *
716 jobc_parent(struct proc *p, struct proc *p_exiting)
717 {
718 	struct proc *pp;
719 
720 	sx_assert(&proctree_lock, SA_LOCKED);
721 
722 	pp = proc_realparent(p);
723 	if (pp->p_pptr == NULL || pp == p_exiting ||
724 	    (pp->p_treeflag & P_TREE_GRPEXITED) == 0)
725 		return (pp);
726 	return (jobc_reaper(pp));
727 }
728 
729 int
730 pgrp_calc_jobc(struct pgrp *pgrp)
731 {
732 	struct proc *q;
733 	int cnt;
734 
735 #ifdef INVARIANTS
736 	if (!mtx_owned(&pgrp->pg_mtx))
737 		sx_assert(&proctree_lock, SA_LOCKED);
738 #endif
739 
740 	cnt = 0;
741 	LIST_FOREACH(q, &pgrp->pg_members, p_pglist) {
742 		if ((q->p_treeflag & P_TREE_GRPEXITED) != 0 ||
743 		    q->p_pptr == NULL)
744 			continue;
745 		if (isjobproc(jobc_parent(q, NULL), pgrp))
746 			cnt++;
747 	}
748 	return (cnt);
749 }
750 
751 /*
752  * Move p to a process group
753  */
754 static void
755 doenterpgrp(struct proc *p, struct pgrp *pgrp)
756 {
757 	struct pgrp *savepgrp;
758 	struct proc *pp;
759 
760 	sx_assert(&proctree_lock, SX_XLOCKED);
761 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
762 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
763 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
764 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
765 
766 	savepgrp = p->p_pgrp;
767 	pp = jobc_parent(p, NULL);
768 
769 	PGRP_LOCK(pgrp);
770 	PGRP_LOCK(savepgrp);
771 	if (isjobproc(pp, savepgrp) && pgrp_calc_jobc(savepgrp) == 1)
772 		orphanpg(savepgrp);
773 	PROC_LOCK(p);
774 	LIST_REMOVE(p, p_pglist);
775 	p->p_pgrp = pgrp;
776 	PROC_UNLOCK(p);
777 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
778 	if (isjobproc(pp, pgrp))
779 		pgrp->pg_flags &= ~PGRP_ORPHANED;
780 	PGRP_UNLOCK(savepgrp);
781 	PGRP_UNLOCK(pgrp);
782 	if (LIST_EMPTY(&savepgrp->pg_members))
783 		pgdelete(savepgrp);
784 }
785 
786 /*
787  * remove process from process group
788  */
789 int
790 leavepgrp(struct proc *p)
791 {
792 	struct pgrp *savepgrp;
793 
794 	sx_assert(&proctree_lock, SX_XLOCKED);
795 	savepgrp = p->p_pgrp;
796 	PGRP_LOCK(savepgrp);
797 	PROC_LOCK(p);
798 	LIST_REMOVE(p, p_pglist);
799 	p->p_pgrp = NULL;
800 	PROC_UNLOCK(p);
801 	PGRP_UNLOCK(savepgrp);
802 	if (LIST_EMPTY(&savepgrp->pg_members))
803 		pgdelete(savepgrp);
804 	return (0);
805 }
806 
807 /*
808  * delete a process group
809  */
810 static void
811 pgdelete(struct pgrp *pgrp)
812 {
813 	struct session *savesess;
814 	struct tty *tp;
815 
816 	sx_assert(&proctree_lock, SX_XLOCKED);
817 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
818 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
819 
820 	/*
821 	 * Reset any sigio structures pointing to us as a result of
822 	 * F_SETOWN with our pgid.  The proctree lock ensures that
823 	 * new sigio structures will not be added after this point.
824 	 */
825 	funsetownlst(&pgrp->pg_sigiolst);
826 
827 	PGRP_LOCK(pgrp);
828 	tp = pgrp->pg_session->s_ttyp;
829 	LIST_REMOVE(pgrp, pg_hash);
830 	savesess = pgrp->pg_session;
831 	PGRP_UNLOCK(pgrp);
832 
833 	/* Remove the reference to the pgrp before deallocating it. */
834 	if (tp != NULL) {
835 		tty_lock(tp);
836 		tty_rel_pgrp(tp, pgrp);
837 	}
838 
839 	proc_id_clear(PROC_ID_GROUP, pgrp->pg_id);
840 	uma_zfree(pgrp_zone, pgrp);
841 	sess_release(savesess);
842 }
843 
844 
845 static void
846 fixjobc_kill(struct proc *p)
847 {
848 	struct proc *q;
849 	struct pgrp *pgrp;
850 
851 	sx_assert(&proctree_lock, SX_LOCKED);
852 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
853 	pgrp = p->p_pgrp;
854 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
855 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
856 
857 	/*
858 	 * p no longer affects process group orphanage for children.
859 	 * It is marked by the flag because p is only physically
860 	 * removed from its process group on wait(2).
861 	 */
862 	MPASS((p->p_treeflag & P_TREE_GRPEXITED) == 0);
863 	p->p_treeflag |= P_TREE_GRPEXITED;
864 
865 	/*
866 	 * Check if exiting p orphans its own group.
867 	 */
868 	pgrp = p->p_pgrp;
869 	if (isjobproc(jobc_parent(p, NULL), pgrp)) {
870 		PGRP_LOCK(pgrp);
871 		if (pgrp_calc_jobc(pgrp) == 0)
872 			orphanpg(pgrp);
873 		PGRP_UNLOCK(pgrp);
874 	}
875 
876 	/*
877 	 * Check this process' children to see whether they qualify
878 	 * their process groups after reparenting to reaper.
879 	 */
880 	LIST_FOREACH(q, &p->p_children, p_sibling) {
881 		pgrp = q->p_pgrp;
882 		PGRP_LOCK(pgrp);
883 		if (pgrp_calc_jobc(pgrp) == 0) {
884 			/*
885 			 * We want to handle exactly the children that
886 			 * has p as realparent.  Then, when calculating
887 			 * jobc_parent for children, we should ignore
888 			 * P_TREE_GRPEXITED flag already set on p.
889 			 */
890 			if (jobc_parent(q, p) == p && isjobproc(p, pgrp))
891 				orphanpg(pgrp);
892 		} else
893 			pgrp->pg_flags &= ~PGRP_ORPHANED;
894 		PGRP_UNLOCK(pgrp);
895 	}
896 	LIST_FOREACH(q, &p->p_orphans, p_orphan) {
897 		pgrp = q->p_pgrp;
898 		PGRP_LOCK(pgrp);
899 		if (pgrp_calc_jobc(pgrp) == 0) {
900 			if (isjobproc(p, pgrp))
901 				orphanpg(pgrp);
902 		} else
903 			pgrp->pg_flags &= ~PGRP_ORPHANED;
904 		PGRP_UNLOCK(pgrp);
905 	}
906 }
907 
908 void
909 killjobc(void)
910 {
911 	struct session *sp;
912 	struct tty *tp;
913 	struct proc *p;
914 	struct vnode *ttyvp;
915 
916 	p = curproc;
917 	MPASS(p->p_flag & P_WEXIT);
918 	sx_assert(&proctree_lock, SX_LOCKED);
919 
920 	if (SESS_LEADER(p)) {
921 		sp = p->p_session;
922 
923 		/*
924 		 * s_ttyp is not zero'd; we use this to indicate that
925 		 * the session once had a controlling terminal. (for
926 		 * logging and informational purposes)
927 		 */
928 		SESS_LOCK(sp);
929 		ttyvp = sp->s_ttyvp;
930 		tp = sp->s_ttyp;
931 		sp->s_ttyvp = NULL;
932 		sp->s_ttydp = NULL;
933 		sp->s_leader = NULL;
934 		SESS_UNLOCK(sp);
935 
936 		/*
937 		 * Signal foreground pgrp and revoke access to
938 		 * controlling terminal if it has not been revoked
939 		 * already.
940 		 *
941 		 * Because the TTY may have been revoked in the mean
942 		 * time and could already have a new session associated
943 		 * with it, make sure we don't send a SIGHUP to a
944 		 * foreground process group that does not belong to this
945 		 * session.
946 		 */
947 
948 		if (tp != NULL) {
949 			tty_lock(tp);
950 			if (tp->t_session == sp)
951 				tty_signal_pgrp(tp, SIGHUP);
952 			tty_unlock(tp);
953 		}
954 
955 		if (ttyvp != NULL) {
956 			sx_xunlock(&proctree_lock);
957 			if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) {
958 				VOP_REVOKE(ttyvp, REVOKEALL);
959 				VOP_UNLOCK(ttyvp);
960 			}
961 			devfs_ctty_unref(ttyvp);
962 			sx_xlock(&proctree_lock);
963 		}
964 	}
965 	fixjobc_kill(p);
966 }
967 
968 /*
969  * A process group has become orphaned, mark it as such for signal
970  * delivery code.  If there are any stopped processes in the group,
971  * hang-up all process in that group.
972  */
973 static void
974 orphanpg(struct pgrp *pg)
975 {
976 	struct proc *p;
977 
978 	PGRP_LOCK_ASSERT(pg, MA_OWNED);
979 
980 	pg->pg_flags |= PGRP_ORPHANED;
981 
982 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
983 		PROC_LOCK(p);
984 		if (P_SHOULDSTOP(p) == P_STOPPED_SIG) {
985 			PROC_UNLOCK(p);
986 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
987 				PROC_LOCK(p);
988 				kern_psignal(p, SIGHUP);
989 				kern_psignal(p, SIGCONT);
990 				PROC_UNLOCK(p);
991 			}
992 			return;
993 		}
994 		PROC_UNLOCK(p);
995 	}
996 }
997 
998 void
999 sess_hold(struct session *s)
1000 {
1001 
1002 	refcount_acquire(&s->s_count);
1003 }
1004 
1005 void
1006 sess_release(struct session *s)
1007 {
1008 
1009 	if (refcount_release(&s->s_count)) {
1010 		if (s->s_ttyp != NULL) {
1011 			tty_lock(s->s_ttyp);
1012 			tty_rel_sess(s->s_ttyp, s);
1013 		}
1014 		proc_id_clear(PROC_ID_SESSION, s->s_sid);
1015 		mtx_destroy(&s->s_mtx);
1016 		free(s, M_SESSION);
1017 	}
1018 }
1019 
1020 #ifdef DDB
1021 
1022 static void
1023 db_print_pgrp_one(struct pgrp *pgrp, struct proc *p)
1024 {
1025 	db_printf(
1026 	    "    pid %d at %p pr %d pgrp %p e %d jc %d\n",
1027 	    p->p_pid, p, p->p_pptr == NULL ? -1 : p->p_pptr->p_pid,
1028 	    p->p_pgrp, (p->p_treeflag & P_TREE_GRPEXITED) != 0,
1029 	    p->p_pptr == NULL ? 0 : isjobproc(p->p_pptr, pgrp));
1030 }
1031 
1032 DB_SHOW_COMMAND_FLAGS(pgrpdump, pgrpdump, DB_CMD_MEMSAFE)
1033 {
1034 	struct pgrp *pgrp;
1035 	struct proc *p;
1036 	int i;
1037 
1038 	for (i = 0; i <= pgrphash; i++) {
1039 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
1040 			db_printf("indx %d\n", i);
1041 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
1042 				db_printf(
1043 			"  pgrp %p, pgid %d, sess %p, sesscnt %d, mem %p\n",
1044 				    pgrp, (int)pgrp->pg_id, pgrp->pg_session,
1045 				    pgrp->pg_session->s_count,
1046 				    LIST_FIRST(&pgrp->pg_members));
1047 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist)
1048 					db_print_pgrp_one(pgrp, p);
1049 			}
1050 		}
1051 	}
1052 }
1053 #endif /* DDB */
1054 
1055 /*
1056  * Calculate the kinfo_proc members which contain process-wide
1057  * informations.
1058  * Must be called with the target process locked.
1059  */
1060 static void
1061 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
1062 {
1063 	struct thread *td;
1064 
1065 	PROC_LOCK_ASSERT(p, MA_OWNED);
1066 
1067 	kp->ki_estcpu = 0;
1068 	kp->ki_pctcpu = 0;
1069 	FOREACH_THREAD_IN_PROC(p, td) {
1070 		thread_lock(td);
1071 		kp->ki_pctcpu += sched_pctcpu(td);
1072 		kp->ki_estcpu += sched_estcpu(td);
1073 		thread_unlock(td);
1074 	}
1075 }
1076 
1077 /*
1078  * Fill in any information that is common to all threads in the process.
1079  * Must be called with the target process locked.
1080  */
1081 static void
1082 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
1083 {
1084 	struct thread *td0;
1085 	struct ucred *cred;
1086 	struct sigacts *ps;
1087 	struct timeval boottime;
1088 
1089 	PROC_LOCK_ASSERT(p, MA_OWNED);
1090 
1091 	kp->ki_structsize = sizeof(*kp);
1092 	kp->ki_paddr = p;
1093 	kp->ki_addr =/* p->p_addr; */0; /* XXX */
1094 	kp->ki_args = p->p_args;
1095 	kp->ki_textvp = p->p_textvp;
1096 #ifdef KTRACE
1097 	kp->ki_tracep = ktr_get_tracevp(p, false);
1098 	kp->ki_traceflag = p->p_traceflag;
1099 #endif
1100 	kp->ki_fd = p->p_fd;
1101 	kp->ki_pd = p->p_pd;
1102 	kp->ki_vmspace = p->p_vmspace;
1103 	kp->ki_flag = p->p_flag;
1104 	kp->ki_flag2 = p->p_flag2;
1105 	cred = p->p_ucred;
1106 	if (cred) {
1107 		kp->ki_uid = cred->cr_uid;
1108 		kp->ki_ruid = cred->cr_ruid;
1109 		kp->ki_svuid = cred->cr_svuid;
1110 		kp->ki_cr_flags = 0;
1111 		if (cred->cr_flags & CRED_FLAG_CAPMODE)
1112 			kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
1113 		/* XXX bde doesn't like KI_NGROUPS */
1114 		if (1 + cred->cr_ngroups > KI_NGROUPS) {
1115 			kp->ki_ngroups = KI_NGROUPS;
1116 			kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
1117 		} else
1118 			kp->ki_ngroups = 1 + cred->cr_ngroups;
1119 		kp->ki_groups[0] = cred->cr_gid;
1120 		bcopy(cred->cr_groups, kp->ki_groups + 1,
1121 		    (kp->ki_ngroups - 1) * sizeof(gid_t));
1122 		kp->ki_rgid = cred->cr_rgid;
1123 		kp->ki_svgid = cred->cr_svgid;
1124 		/* If jailed(cred), emulate the old P_JAILED flag. */
1125 		if (jailed(cred)) {
1126 			kp->ki_flag |= P_JAILED;
1127 			/* If inside the jail, use 0 as a jail ID. */
1128 			if (cred->cr_prison != curthread->td_ucred->cr_prison)
1129 				kp->ki_jid = cred->cr_prison->pr_id;
1130 		}
1131 		strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
1132 		    sizeof(kp->ki_loginclass));
1133 	}
1134 	ps = p->p_sigacts;
1135 	if (ps) {
1136 		mtx_lock(&ps->ps_mtx);
1137 		kp->ki_sigignore = ps->ps_sigignore;
1138 		kp->ki_sigcatch = ps->ps_sigcatch;
1139 		mtx_unlock(&ps->ps_mtx);
1140 	}
1141 	if (p->p_state != PRS_NEW &&
1142 	    p->p_state != PRS_ZOMBIE &&
1143 	    p->p_vmspace != NULL) {
1144 		struct vmspace *vm = p->p_vmspace;
1145 
1146 		kp->ki_size = vm->vm_map.size;
1147 		kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
1148 		FOREACH_THREAD_IN_PROC(p, td0)
1149 			kp->ki_rssize += td0->td_kstack_pages;
1150 		kp->ki_swrss = vm->vm_swrss;
1151 		kp->ki_tsize = vm->vm_tsize;
1152 		kp->ki_dsize = vm->vm_dsize;
1153 		kp->ki_ssize = vm->vm_ssize;
1154 	} else if (p->p_state == PRS_ZOMBIE)
1155 		kp->ki_stat = SZOMB;
1156 	kp->ki_sflag = PS_INMEM;
1157 	/* Calculate legacy swtime as seconds since 'swtick'. */
1158 	kp->ki_swtime = (ticks - p->p_swtick) / hz;
1159 	kp->ki_pid = p->p_pid;
1160 	kp->ki_nice = p->p_nice;
1161 	kp->ki_fibnum = p->p_fibnum;
1162 	kp->ki_start = p->p_stats->p_start;
1163 	getboottime(&boottime);
1164 	timevaladd(&kp->ki_start, &boottime);
1165 	PROC_STATLOCK(p);
1166 	rufetch(p, &kp->ki_rusage);
1167 	kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
1168 	calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
1169 	PROC_STATUNLOCK(p);
1170 	calccru(p, &kp->ki_childutime, &kp->ki_childstime);
1171 	/* Some callers want child times in a single value. */
1172 	kp->ki_childtime = kp->ki_childstime;
1173 	timevaladd(&kp->ki_childtime, &kp->ki_childutime);
1174 
1175 	FOREACH_THREAD_IN_PROC(p, td0)
1176 		kp->ki_cow += td0->td_cow;
1177 
1178 	if (p->p_comm[0] != '\0')
1179 		strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
1180 	if (p->p_sysent && p->p_sysent->sv_name != NULL &&
1181 	    p->p_sysent->sv_name[0] != '\0')
1182 		strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
1183 	kp->ki_siglist = p->p_siglist;
1184 	kp->ki_xstat = KW_EXITCODE(p->p_xexit, p->p_xsig);
1185 	kp->ki_acflag = p->p_acflag;
1186 	kp->ki_lock = p->p_lock;
1187 	if (p->p_pptr) {
1188 		kp->ki_ppid = p->p_oppid;
1189 		if (p->p_flag & P_TRACED)
1190 			kp->ki_tracer = p->p_pptr->p_pid;
1191 	}
1192 }
1193 
1194 /*
1195  * Fill job-related process information.
1196  */
1197 static void
1198 fill_kinfo_proc_pgrp(struct proc *p, struct kinfo_proc *kp)
1199 {
1200 	struct tty *tp;
1201 	struct session *sp;
1202 	struct pgrp *pgrp;
1203 
1204 	sx_assert(&proctree_lock, SA_LOCKED);
1205 	PROC_LOCK_ASSERT(p, MA_OWNED);
1206 
1207 	pgrp = p->p_pgrp;
1208 	if (pgrp == NULL)
1209 		return;
1210 
1211 	kp->ki_pgid = pgrp->pg_id;
1212 	kp->ki_jobc = pgrp_calc_jobc(pgrp);
1213 
1214 	sp = pgrp->pg_session;
1215 	tp = NULL;
1216 
1217 	if (sp != NULL) {
1218 		kp->ki_sid = sp->s_sid;
1219 		SESS_LOCK(sp);
1220 		strlcpy(kp->ki_login, sp->s_login, sizeof(kp->ki_login));
1221 		if (sp->s_ttyvp)
1222 			kp->ki_kiflag |= KI_CTTY;
1223 		if (SESS_LEADER(p))
1224 			kp->ki_kiflag |= KI_SLEADER;
1225 		tp = sp->s_ttyp;
1226 		SESS_UNLOCK(sp);
1227 	}
1228 
1229 	if ((p->p_flag & P_CONTROLT) && tp != NULL) {
1230 		kp->ki_tdev = tty_udev(tp);
1231 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1232 		kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
1233 		if (tp->t_session)
1234 			kp->ki_tsid = tp->t_session->s_sid;
1235 	} else {
1236 		kp->ki_tdev = NODEV;
1237 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1238 	}
1239 	kp->ki_reaper = p->p_reaper->p_pid;
1240 	kp->ki_reapsubtree = p->p_reapsubtree;
1241 }
1242 
1243 /*
1244  * Fill in information that is thread specific.  Must be called with
1245  * target process locked.  If 'preferthread' is set, overwrite certain
1246  * process-related fields that are maintained for both threads and
1247  * processes.
1248  */
1249 static void
1250 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
1251 {
1252 	struct proc *p;
1253 
1254 	p = td->td_proc;
1255 	kp->ki_tdaddr = td;
1256 	PROC_LOCK_ASSERT(p, MA_OWNED);
1257 
1258 	if (preferthread)
1259 		PROC_STATLOCK(p);
1260 	thread_lock(td);
1261 	if (td->td_wmesg != NULL)
1262 		strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
1263 	else
1264 		bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
1265 	if (strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname)) >=
1266 	    sizeof(kp->ki_tdname)) {
1267 		strlcpy(kp->ki_moretdname,
1268 		    td->td_name + sizeof(kp->ki_tdname) - 1,
1269 		    sizeof(kp->ki_moretdname));
1270 	} else {
1271 		bzero(kp->ki_moretdname, sizeof(kp->ki_moretdname));
1272 	}
1273 	if (TD_ON_LOCK(td)) {
1274 		kp->ki_kiflag |= KI_LOCKBLOCK;
1275 		strlcpy(kp->ki_lockname, td->td_lockname,
1276 		    sizeof(kp->ki_lockname));
1277 	} else {
1278 		kp->ki_kiflag &= ~KI_LOCKBLOCK;
1279 		bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
1280 	}
1281 
1282 	if (p->p_state == PRS_NORMAL) { /* approximate. */
1283 		if (TD_ON_RUNQ(td) ||
1284 		    TD_CAN_RUN(td) ||
1285 		    TD_IS_RUNNING(td)) {
1286 			kp->ki_stat = SRUN;
1287 		} else if (P_SHOULDSTOP(p)) {
1288 			kp->ki_stat = SSTOP;
1289 		} else if (TD_IS_SLEEPING(td)) {
1290 			kp->ki_stat = SSLEEP;
1291 		} else if (TD_ON_LOCK(td)) {
1292 			kp->ki_stat = SLOCK;
1293 		} else {
1294 			kp->ki_stat = SWAIT;
1295 		}
1296 	} else if (p->p_state == PRS_ZOMBIE) {
1297 		kp->ki_stat = SZOMB;
1298 	} else {
1299 		kp->ki_stat = SIDL;
1300 	}
1301 
1302 	/* Things in the thread */
1303 	kp->ki_wchan = td->td_wchan;
1304 	kp->ki_pri.pri_level = td->td_priority;
1305 	kp->ki_pri.pri_native = td->td_base_pri;
1306 
1307 	/*
1308 	 * Note: legacy fields; clamp at the old NOCPU value and/or
1309 	 * the maximum u_char CPU value.
1310 	 */
1311 	if (td->td_lastcpu == NOCPU)
1312 		kp->ki_lastcpu_old = NOCPU_OLD;
1313 	else if (td->td_lastcpu > MAXCPU_OLD)
1314 		kp->ki_lastcpu_old = MAXCPU_OLD;
1315 	else
1316 		kp->ki_lastcpu_old = td->td_lastcpu;
1317 
1318 	if (td->td_oncpu == NOCPU)
1319 		kp->ki_oncpu_old = NOCPU_OLD;
1320 	else if (td->td_oncpu > MAXCPU_OLD)
1321 		kp->ki_oncpu_old = MAXCPU_OLD;
1322 	else
1323 		kp->ki_oncpu_old = td->td_oncpu;
1324 
1325 	kp->ki_lastcpu = td->td_lastcpu;
1326 	kp->ki_oncpu = td->td_oncpu;
1327 	kp->ki_tdflags = td->td_flags;
1328 	kp->ki_tid = td->td_tid;
1329 	kp->ki_numthreads = p->p_numthreads;
1330 	kp->ki_pcb = td->td_pcb;
1331 	kp->ki_kstack = td->td_kstack;
1332 	kp->ki_slptime = (ticks - td->td_slptick) / hz;
1333 	kp->ki_pri.pri_class = td->td_pri_class;
1334 	kp->ki_pri.pri_user = td->td_user_pri;
1335 
1336 	if (preferthread) {
1337 		rufetchtd(td, &kp->ki_rusage);
1338 		kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
1339 		kp->ki_pctcpu = sched_pctcpu(td);
1340 		kp->ki_estcpu = sched_estcpu(td);
1341 		kp->ki_cow = td->td_cow;
1342 	}
1343 
1344 	/* We can't get this anymore but ps etc never used it anyway. */
1345 	kp->ki_rqindex = 0;
1346 
1347 	if (preferthread)
1348 		kp->ki_siglist = td->td_siglist;
1349 	kp->ki_sigmask = td->td_sigmask;
1350 	thread_unlock(td);
1351 	if (preferthread)
1352 		PROC_STATUNLOCK(p);
1353 
1354 	if ((td->td_pflags & TDP2_UEXTERR) != 0)
1355 		kp->ki_uerrmsg = td->td_exterr_ptr;
1356 }
1357 
1358 /*
1359  * Fill in a kinfo_proc structure for the specified process.
1360  * Must be called with the target process locked.
1361  */
1362 void
1363 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1364 {
1365 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1366 
1367 	bzero(kp, sizeof(*kp));
1368 
1369 	fill_kinfo_proc_pgrp(p,kp);
1370 	fill_kinfo_proc_only(p, kp);
1371 	fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1372 	fill_kinfo_aggregate(p, kp);
1373 }
1374 
1375 struct pstats *
1376 pstats_alloc(void)
1377 {
1378 
1379 	return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1380 }
1381 
1382 /*
1383  * Copy parts of p_stats; zero the rest of p_stats (statistics).
1384  */
1385 void
1386 pstats_fork(struct pstats *src, struct pstats *dst)
1387 {
1388 
1389 	bzero(&dst->pstat_startzero,
1390 	    __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1391 	bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1392 	    __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1393 }
1394 
1395 void
1396 pstats_free(struct pstats *ps)
1397 {
1398 
1399 	free(ps, M_SUBPROC);
1400 }
1401 
1402 #ifdef COMPAT_FREEBSD32
1403 
1404 /*
1405  * This function is typically used to copy out the kernel address, so
1406  * it can be replaced by assignment of zero.
1407  */
1408 static inline uint32_t
1409 ptr32_trim(const void *ptr)
1410 {
1411 	uintptr_t uptr;
1412 
1413 	uptr = (uintptr_t)ptr;
1414 	return ((uptr > UINT_MAX) ? 0 : uptr);
1415 }
1416 
1417 #define PTRTRIM_CP(src,dst,fld) \
1418 	do { (dst).fld = ptr32_trim((src).fld); } while (0)
1419 
1420 static void
1421 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1422 {
1423 	int i;
1424 
1425 	bzero(ki32, sizeof(struct kinfo_proc32));
1426 	ki32->ki_structsize = sizeof(struct kinfo_proc32);
1427 	CP(*ki, *ki32, ki_layout);
1428 	PTRTRIM_CP(*ki, *ki32, ki_args);
1429 	PTRTRIM_CP(*ki, *ki32, ki_paddr);
1430 	PTRTRIM_CP(*ki, *ki32, ki_addr);
1431 	PTRTRIM_CP(*ki, *ki32, ki_tracep);
1432 	PTRTRIM_CP(*ki, *ki32, ki_textvp);
1433 	PTRTRIM_CP(*ki, *ki32, ki_fd);
1434 	PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1435 	PTRTRIM_CP(*ki, *ki32, ki_wchan);
1436 	CP(*ki, *ki32, ki_pid);
1437 	CP(*ki, *ki32, ki_ppid);
1438 	CP(*ki, *ki32, ki_pgid);
1439 	CP(*ki, *ki32, ki_tpgid);
1440 	CP(*ki, *ki32, ki_sid);
1441 	CP(*ki, *ki32, ki_tsid);
1442 	CP(*ki, *ki32, ki_jobc);
1443 	FU64_CP(*ki, *ki32, ki_tdev);
1444 	CP(*ki, *ki32, ki_tdev_freebsd11);
1445 	CP(*ki, *ki32, ki_siglist);
1446 	CP(*ki, *ki32, ki_sigmask);
1447 	CP(*ki, *ki32, ki_sigignore);
1448 	CP(*ki, *ki32, ki_sigcatch);
1449 	CP(*ki, *ki32, ki_uid);
1450 	CP(*ki, *ki32, ki_ruid);
1451 	CP(*ki, *ki32, ki_svuid);
1452 	CP(*ki, *ki32, ki_rgid);
1453 	CP(*ki, *ki32, ki_svgid);
1454 	CP(*ki, *ki32, ki_ngroups);
1455 	for (i = 0; i < KI_NGROUPS; i++)
1456 		CP(*ki, *ki32, ki_groups[i]);
1457 	CP(*ki, *ki32, ki_size);
1458 	CP(*ki, *ki32, ki_rssize);
1459 	CP(*ki, *ki32, ki_swrss);
1460 	CP(*ki, *ki32, ki_tsize);
1461 	CP(*ki, *ki32, ki_dsize);
1462 	CP(*ki, *ki32, ki_ssize);
1463 	CP(*ki, *ki32, ki_xstat);
1464 	CP(*ki, *ki32, ki_acflag);
1465 	CP(*ki, *ki32, ki_pctcpu);
1466 	CP(*ki, *ki32, ki_estcpu);
1467 	CP(*ki, *ki32, ki_slptime);
1468 	CP(*ki, *ki32, ki_swtime);
1469 	CP(*ki, *ki32, ki_cow);
1470 	FU64_CP(*ki, *ki32, ki_runtime);
1471 	TV_CP(*ki, *ki32, ki_start);
1472 	TV_CP(*ki, *ki32, ki_childtime);
1473 	CP(*ki, *ki32, ki_flag);
1474 	CP(*ki, *ki32, ki_kiflag);
1475 	CP(*ki, *ki32, ki_traceflag);
1476 	CP(*ki, *ki32, ki_stat);
1477 	CP(*ki, *ki32, ki_nice);
1478 	CP(*ki, *ki32, ki_lock);
1479 	CP(*ki, *ki32, ki_rqindex);
1480 	CP(*ki, *ki32, ki_oncpu);
1481 	CP(*ki, *ki32, ki_lastcpu);
1482 
1483 	/* XXX TODO: wrap cpu value as appropriate */
1484 	CP(*ki, *ki32, ki_oncpu_old);
1485 	CP(*ki, *ki32, ki_lastcpu_old);
1486 
1487 	bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1488 	bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1489 	bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1490 	bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1491 	bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1492 	bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1493 	bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1494 	bcopy(ki->ki_moretdname, ki32->ki_moretdname, MAXCOMLEN - TDNAMLEN + 1);
1495 	CP(*ki, *ki32, ki_tracer);
1496 	CP(*ki, *ki32, ki_flag2);
1497 	CP(*ki, *ki32, ki_fibnum);
1498 	CP(*ki, *ki32, ki_cr_flags);
1499 	CP(*ki, *ki32, ki_jid);
1500 	CP(*ki, *ki32, ki_reaper);
1501 	CP(*ki, *ki32, ki_reapsubtree);
1502 	CP(*ki, *ki32, ki_numthreads);
1503 	CP(*ki, *ki32, ki_tid);
1504 	CP(*ki, *ki32, ki_pri);
1505 	freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1506 	freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1507 	PTRTRIM_CP(*ki, *ki32, ki_pcb);
1508 	PTRTRIM_CP(*ki, *ki32, ki_kstack);
1509 	PTRTRIM_CP(*ki, *ki32, ki_udata);
1510 	PTRTRIM_CP(*ki, *ki32, ki_tdaddr);
1511 	PTRTRIM_CP(*ki, *ki32, ki_pd);
1512 	CP(*ki, *ki32, ki_sflag);
1513 	CP(*ki, *ki32, ki_tdflags);
1514 	PTRTRIM_CP(*ki, *ki32, ki_uerrmsg);
1515 }
1516 #endif
1517 
1518 static ssize_t
1519 kern_proc_out_size(struct proc *p, int flags)
1520 {
1521 	ssize_t size = 0;
1522 
1523 	PROC_LOCK_ASSERT(p, MA_OWNED);
1524 
1525 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1526 #ifdef COMPAT_FREEBSD32
1527 		if ((flags & KERN_PROC_MASK32) != 0) {
1528 			size += sizeof(struct kinfo_proc32);
1529 		} else
1530 #endif
1531 			size += sizeof(struct kinfo_proc);
1532 	} else {
1533 #ifdef COMPAT_FREEBSD32
1534 		if ((flags & KERN_PROC_MASK32) != 0)
1535 			size += sizeof(struct kinfo_proc32) * p->p_numthreads;
1536 		else
1537 #endif
1538 			size += sizeof(struct kinfo_proc) * p->p_numthreads;
1539 	}
1540 	PROC_UNLOCK(p);
1541 	return (size);
1542 }
1543 
1544 int
1545 kern_proc_out(struct proc *p, struct sbuf *sb, int flags)
1546 {
1547 	struct thread *td;
1548 	struct kinfo_proc ki;
1549 #ifdef COMPAT_FREEBSD32
1550 	struct kinfo_proc32 ki32;
1551 #endif
1552 	int error;
1553 
1554 	PROC_LOCK_ASSERT(p, MA_OWNED);
1555 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1556 
1557 	error = 0;
1558 	fill_kinfo_proc(p, &ki);
1559 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1560 #ifdef COMPAT_FREEBSD32
1561 		if ((flags & KERN_PROC_MASK32) != 0) {
1562 			freebsd32_kinfo_proc_out(&ki, &ki32);
1563 			if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1564 				error = ENOMEM;
1565 		} else
1566 #endif
1567 			if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1568 				error = ENOMEM;
1569 	} else {
1570 		FOREACH_THREAD_IN_PROC(p, td) {
1571 			fill_kinfo_thread(td, &ki, 1);
1572 #ifdef COMPAT_FREEBSD32
1573 			if ((flags & KERN_PROC_MASK32) != 0) {
1574 				freebsd32_kinfo_proc_out(&ki, &ki32);
1575 				if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1576 					error = ENOMEM;
1577 			} else
1578 #endif
1579 				if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1580 					error = ENOMEM;
1581 			if (error != 0)
1582 				break;
1583 		}
1584 	}
1585 	PROC_UNLOCK(p);
1586 	return (error);
1587 }
1588 
1589 static int
1590 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
1591 {
1592 	struct sbuf sb;
1593 	struct kinfo_proc ki;
1594 	int error, error2;
1595 
1596 	if (req->oldptr == NULL)
1597 		return (SYSCTL_OUT(req, 0, kern_proc_out_size(p, flags)));
1598 
1599 	sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req);
1600 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1601 	error = kern_proc_out(p, &sb, flags);
1602 	error2 = sbuf_finish(&sb);
1603 	sbuf_delete(&sb);
1604 	if (error != 0)
1605 		return (error);
1606 	else if (error2 != 0)
1607 		return (error2);
1608 	return (0);
1609 }
1610 
1611 int
1612 proc_iterate(int (*cb)(struct proc *, void *), void *cbarg)
1613 {
1614 	struct proc *p;
1615 	int error, i, j;
1616 
1617 	for (i = 0; i < pidhashlock + 1; i++) {
1618 		sx_slock(&proctree_lock);
1619 		sx_slock(&pidhashtbl_lock[i]);
1620 		for (j = i; j <= pidhash; j += pidhashlock + 1) {
1621 			LIST_FOREACH(p, &pidhashtbl[j], p_hash) {
1622 				if (p->p_state == PRS_NEW)
1623 					continue;
1624 				error = cb(p, cbarg);
1625 				PROC_LOCK_ASSERT(p, MA_NOTOWNED);
1626 				if (error != 0) {
1627 					sx_sunlock(&pidhashtbl_lock[i]);
1628 					sx_sunlock(&proctree_lock);
1629 					return (error);
1630 				}
1631 			}
1632 		}
1633 		sx_sunlock(&pidhashtbl_lock[i]);
1634 		sx_sunlock(&proctree_lock);
1635 	}
1636 	return (0);
1637 }
1638 
1639 struct kern_proc_out_args {
1640 	struct sysctl_req *req;
1641 	int flags;
1642 	int oid_number;
1643 	int *name;
1644 };
1645 
1646 static int
1647 sysctl_kern_proc_iterate(struct proc *p, void *origarg)
1648 {
1649 	struct kern_proc_out_args *arg = origarg;
1650 	int *name = arg->name;
1651 	int oid_number = arg->oid_number;
1652 	int flags = arg->flags;
1653 	struct sysctl_req *req = arg->req;
1654 	int error = 0;
1655 
1656 	PROC_LOCK(p);
1657 
1658 	KASSERT(p->p_ucred != NULL,
1659 	    ("process credential is NULL for non-NEW proc"));
1660 	/*
1661 	 * Show a user only appropriate processes.
1662 	 */
1663 	if (p_cansee(curthread, p))
1664 		goto skip;
1665 	/*
1666 	 * TODO - make more efficient (see notes below).
1667 	 * do by session.
1668 	 */
1669 	switch (oid_number) {
1670 	case KERN_PROC_GID:
1671 		if (p->p_ucred->cr_gid != (gid_t)name[0])
1672 			goto skip;
1673 		break;
1674 
1675 	case KERN_PROC_PGRP:
1676 		/* could do this by traversing pgrp */
1677 		if (p->p_pgrp == NULL ||
1678 		    p->p_pgrp->pg_id != (pid_t)name[0])
1679 			goto skip;
1680 		break;
1681 
1682 	case KERN_PROC_RGID:
1683 		if (p->p_ucred->cr_rgid != (gid_t)name[0])
1684 			goto skip;
1685 		break;
1686 
1687 	case KERN_PROC_SESSION:
1688 		if (p->p_session == NULL ||
1689 		    p->p_session->s_sid != (pid_t)name[0])
1690 			goto skip;
1691 		break;
1692 
1693 	case KERN_PROC_TTY:
1694 		if ((p->p_flag & P_CONTROLT) == 0 ||
1695 		    p->p_session == NULL)
1696 			goto skip;
1697 		/* XXX proctree_lock */
1698 		SESS_LOCK(p->p_session);
1699 		if (p->p_session->s_ttyp == NULL ||
1700 		    tty_udev(p->p_session->s_ttyp) !=
1701 		    (dev_t)name[0]) {
1702 			SESS_UNLOCK(p->p_session);
1703 			goto skip;
1704 		}
1705 		SESS_UNLOCK(p->p_session);
1706 		break;
1707 
1708 	case KERN_PROC_UID:
1709 		if (p->p_ucred->cr_uid != (uid_t)name[0])
1710 			goto skip;
1711 		break;
1712 
1713 	case KERN_PROC_RUID:
1714 		if (p->p_ucred->cr_ruid != (uid_t)name[0])
1715 			goto skip;
1716 		break;
1717 
1718 	case KERN_PROC_PROC:
1719 		break;
1720 
1721 	default:
1722 		break;
1723 	}
1724 	error = sysctl_out_proc(p, req, flags);
1725 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
1726 	return (error);
1727 skip:
1728 	PROC_UNLOCK(p);
1729 	return (0);
1730 }
1731 
1732 static int
1733 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1734 {
1735 	struct kern_proc_out_args iterarg;
1736 	int *name = (int *)arg1;
1737 	u_int namelen = arg2;
1738 	struct proc *p;
1739 	int flags, oid_number;
1740 	int error = 0;
1741 
1742 	oid_number = oidp->oid_number;
1743 	if (oid_number != KERN_PROC_ALL &&
1744 	    (oid_number & KERN_PROC_INC_THREAD) == 0)
1745 		flags = KERN_PROC_NOTHREADS;
1746 	else {
1747 		flags = 0;
1748 		oid_number &= ~KERN_PROC_INC_THREAD;
1749 	}
1750 #ifdef COMPAT_FREEBSD32
1751 	if (req->flags & SCTL_MASK32)
1752 		flags |= KERN_PROC_MASK32;
1753 #endif
1754 	if (oid_number == KERN_PROC_PID) {
1755 		if (namelen != 1)
1756 			return (EINVAL);
1757 		error = sysctl_wire_old_buffer(req, 0);
1758 		if (error)
1759 			return (error);
1760 		sx_slock(&proctree_lock);
1761 		error = pget((pid_t)name[0], PGET_CANSEE, &p);
1762 		if (error == 0)
1763 			error = sysctl_out_proc(p, req, flags);
1764 		sx_sunlock(&proctree_lock);
1765 		return (error);
1766 	}
1767 
1768 	switch (oid_number) {
1769 	case KERN_PROC_ALL:
1770 		if (namelen != 0)
1771 			return (EINVAL);
1772 		break;
1773 	case KERN_PROC_PROC:
1774 		if (namelen != 0 && namelen != 1)
1775 			return (EINVAL);
1776 		break;
1777 	default:
1778 		if (namelen != 1)
1779 			return (EINVAL);
1780 		break;
1781 	}
1782 
1783 	if (req->oldptr == NULL) {
1784 		/* overestimate by 5 procs */
1785 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1786 		if (error)
1787 			return (error);
1788 	} else {
1789 		error = sysctl_wire_old_buffer(req, 0);
1790 		if (error != 0)
1791 			return (error);
1792 	}
1793 	iterarg.flags = flags;
1794 	iterarg.oid_number = oid_number;
1795 	iterarg.req = req;
1796 	iterarg.name = name;
1797 	error = proc_iterate(sysctl_kern_proc_iterate, &iterarg);
1798 	return (error);
1799 }
1800 
1801 struct pargs *
1802 pargs_alloc(int len)
1803 {
1804 	struct pargs *pa;
1805 
1806 	pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1807 		M_WAITOK);
1808 	refcount_init(&pa->ar_ref, 1);
1809 	pa->ar_length = len;
1810 	return (pa);
1811 }
1812 
1813 static void
1814 pargs_free(struct pargs *pa)
1815 {
1816 
1817 	free(pa, M_PARGS);
1818 }
1819 
1820 void
1821 pargs_hold(struct pargs *pa)
1822 {
1823 
1824 	if (pa == NULL)
1825 		return;
1826 	refcount_acquire(&pa->ar_ref);
1827 }
1828 
1829 void
1830 pargs_drop(struct pargs *pa)
1831 {
1832 
1833 	if (pa == NULL)
1834 		return;
1835 	if (refcount_release(&pa->ar_ref))
1836 		pargs_free(pa);
1837 }
1838 
1839 static int
1840 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1841     size_t len)
1842 {
1843 	ssize_t n;
1844 
1845 	/*
1846 	 * This may return a short read if the string is shorter than the chunk
1847 	 * and is aligned at the end of the page, and the following page is not
1848 	 * mapped.
1849 	 */
1850 	n = proc_readmem(td, p, (vm_offset_t)sptr, buf, len);
1851 	if (n <= 0)
1852 		return (ENOMEM);
1853 	return (0);
1854 }
1855 
1856 #define PROC_AUXV_MAX	256	/* Safety limit on auxv size. */
1857 
1858 enum proc_vector_type {
1859 	PROC_ARG,
1860 	PROC_ENV,
1861 	PROC_AUX,
1862 };
1863 
1864 #ifdef COMPAT_FREEBSD32
1865 static int
1866 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1867     size_t *vsizep, enum proc_vector_type type)
1868 {
1869 	struct freebsd32_ps_strings pss;
1870 	Elf32_Auxinfo aux;
1871 	vm_offset_t vptr, ptr;
1872 	uint32_t *proc_vector32;
1873 	char **proc_vector;
1874 	size_t vsize, size;
1875 	int i, error;
1876 
1877 	error = 0;
1878 	if (proc_readmem(td, p, PROC_PS_STRINGS(p), &pss, sizeof(pss)) !=
1879 	    sizeof(pss))
1880 		return (ENOMEM);
1881 	switch (type) {
1882 	case PROC_ARG:
1883 		vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1884 		vsize = pss.ps_nargvstr;
1885 		if (vsize > ARG_MAX)
1886 			return (ENOEXEC);
1887 		size = vsize * sizeof(int32_t);
1888 		break;
1889 	case PROC_ENV:
1890 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1891 		vsize = pss.ps_nenvstr;
1892 		if (vsize > ARG_MAX)
1893 			return (ENOEXEC);
1894 		size = vsize * sizeof(int32_t);
1895 		break;
1896 	case PROC_AUX:
1897 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1898 		    (pss.ps_nenvstr + 1) * sizeof(int32_t);
1899 		if (vptr % 4 != 0)
1900 			return (ENOEXEC);
1901 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1902 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1903 			    sizeof(aux))
1904 				return (ENOMEM);
1905 			if (aux.a_type == AT_NULL)
1906 				break;
1907 			ptr += sizeof(aux);
1908 		}
1909 		if (aux.a_type != AT_NULL)
1910 			return (ENOEXEC);
1911 		vsize = i + 1;
1912 		size = vsize * sizeof(aux);
1913 		break;
1914 	default:
1915 		KASSERT(0, ("Wrong proc vector type: %d", type));
1916 		return (EINVAL);
1917 	}
1918 	proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1919 	if (proc_readmem(td, p, vptr, proc_vector32, size) != size) {
1920 		error = ENOMEM;
1921 		goto done;
1922 	}
1923 	if (type == PROC_AUX) {
1924 		*proc_vectorp = (char **)proc_vector32;
1925 		*vsizep = vsize;
1926 		return (0);
1927 	}
1928 	proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1929 	for (i = 0; i < (int)vsize; i++)
1930 		proc_vector[i] = PTRIN(proc_vector32[i]);
1931 	*proc_vectorp = proc_vector;
1932 	*vsizep = vsize;
1933 done:
1934 	free(proc_vector32, M_TEMP);
1935 	return (error);
1936 }
1937 #endif
1938 
1939 static int
1940 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1941     size_t *vsizep, enum proc_vector_type type)
1942 {
1943 	struct ps_strings pss;
1944 	Elf_Auxinfo aux;
1945 	vm_offset_t vptr, ptr;
1946 	char **proc_vector;
1947 	size_t vsize, size;
1948 	int i;
1949 
1950 #ifdef COMPAT_FREEBSD32
1951 	if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1952 		return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1953 #endif
1954 	if (proc_readmem(td, p, PROC_PS_STRINGS(p), &pss, sizeof(pss)) !=
1955 	    sizeof(pss))
1956 		return (ENOMEM);
1957 	switch (type) {
1958 	case PROC_ARG:
1959 		vptr = (vm_offset_t)pss.ps_argvstr;
1960 		vsize = pss.ps_nargvstr;
1961 		if (vsize > ARG_MAX)
1962 			return (ENOEXEC);
1963 		size = vsize * sizeof(char *);
1964 		break;
1965 	case PROC_ENV:
1966 		vptr = (vm_offset_t)pss.ps_envstr;
1967 		vsize = pss.ps_nenvstr;
1968 		if (vsize > ARG_MAX)
1969 			return (ENOEXEC);
1970 		size = vsize * sizeof(char *);
1971 		break;
1972 	case PROC_AUX:
1973 		/*
1974 		 * The aux array is just above env array on the stack. Check
1975 		 * that the address is naturally aligned.
1976 		 */
1977 		vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1978 		    * sizeof(char *);
1979 #if __ELF_WORD_SIZE == 64
1980 		if (vptr % sizeof(uint64_t) != 0)
1981 #else
1982 		if (vptr % sizeof(uint32_t) != 0)
1983 #endif
1984 			return (ENOEXEC);
1985 		/*
1986 		 * We count the array size reading the aux vectors from the
1987 		 * stack until AT_NULL vector is returned.  So (to keep the code
1988 		 * simple) we read the process stack twice: the first time here
1989 		 * to find the size and the second time when copying the vectors
1990 		 * to the allocated proc_vector.
1991 		 */
1992 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1993 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1994 			    sizeof(aux))
1995 				return (ENOMEM);
1996 			if (aux.a_type == AT_NULL)
1997 				break;
1998 			ptr += sizeof(aux);
1999 		}
2000 		/*
2001 		 * If the PROC_AUXV_MAX entries are iterated over, and we have
2002 		 * not reached AT_NULL, it is most likely we are reading wrong
2003 		 * data: either the process doesn't have auxv array or data has
2004 		 * been modified. Return the error in this case.
2005 		 */
2006 		if (aux.a_type != AT_NULL)
2007 			return (ENOEXEC);
2008 		vsize = i + 1;
2009 		size = vsize * sizeof(aux);
2010 		break;
2011 	default:
2012 		KASSERT(0, ("Wrong proc vector type: %d", type));
2013 		return (EINVAL); /* In case we are built without INVARIANTS. */
2014 	}
2015 	proc_vector = malloc(size, M_TEMP, M_WAITOK);
2016 	if (proc_readmem(td, p, vptr, proc_vector, size) != size) {
2017 		free(proc_vector, M_TEMP);
2018 		return (ENOMEM);
2019 	}
2020 	*proc_vectorp = proc_vector;
2021 	*vsizep = vsize;
2022 
2023 	return (0);
2024 }
2025 
2026 #define GET_PS_STRINGS_CHUNK_SZ	256	/* Chunk size (bytes) for ps_strings operations. */
2027 
2028 static int
2029 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
2030     enum proc_vector_type type)
2031 {
2032 	size_t done, len, nchr, vsize;
2033 	int error, i;
2034 	char **proc_vector, *sptr;
2035 	char pss_string[GET_PS_STRINGS_CHUNK_SZ];
2036 
2037 	PROC_ASSERT_HELD(p);
2038 
2039 	/*
2040 	 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
2041 	 */
2042 	nchr = 2 * (PATH_MAX + ARG_MAX);
2043 
2044 	error = get_proc_vector(td, p, &proc_vector, &vsize, type);
2045 	if (error != 0)
2046 		return (error);
2047 	for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
2048 		/*
2049 		 * The program may have scribbled into its argv array, e.g. to
2050 		 * remove some arguments.  If that has happened, break out
2051 		 * before trying to read from NULL.
2052 		 */
2053 		if (proc_vector[i] == NULL)
2054 			break;
2055 		for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
2056 			error = proc_read_string(td, p, sptr, pss_string,
2057 			    sizeof(pss_string));
2058 			if (error != 0)
2059 				goto done;
2060 			len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
2061 			if (done + len >= nchr)
2062 				len = nchr - done - 1;
2063 			sbuf_bcat(sb, pss_string, len);
2064 			if (len != GET_PS_STRINGS_CHUNK_SZ)
2065 				break;
2066 			done += GET_PS_STRINGS_CHUNK_SZ;
2067 		}
2068 		sbuf_bcat(sb, "", 1);
2069 		done += len + 1;
2070 	}
2071 done:
2072 	free(proc_vector, M_TEMP);
2073 	return (error);
2074 }
2075 
2076 int
2077 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
2078 {
2079 
2080 	return (get_ps_strings(curthread, p, sb, PROC_ARG));
2081 }
2082 
2083 int
2084 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
2085 {
2086 
2087 	return (get_ps_strings(curthread, p, sb, PROC_ENV));
2088 }
2089 
2090 int
2091 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb)
2092 {
2093 	size_t vsize, size;
2094 	char **auxv;
2095 	int error;
2096 
2097 	error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX);
2098 	if (error == 0) {
2099 #ifdef COMPAT_FREEBSD32
2100 		if (SV_PROC_FLAG(p, SV_ILP32) != 0)
2101 			size = vsize * sizeof(Elf32_Auxinfo);
2102 		else
2103 #endif
2104 			size = vsize * sizeof(Elf_Auxinfo);
2105 		if (sbuf_bcat(sb, auxv, size) != 0)
2106 			error = ENOMEM;
2107 		free(auxv, M_TEMP);
2108 	}
2109 	return (error);
2110 }
2111 
2112 /*
2113  * This sysctl allows a process to retrieve the argument list or process
2114  * title for another process without groping around in the address space
2115  * of the other process.  It also allow a process to set its own "process
2116  * title to a string of its own choice.
2117  */
2118 static int
2119 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
2120 {
2121 	int *name = (int *)arg1;
2122 	u_int namelen = arg2;
2123 	struct pargs *newpa, *pa;
2124 	struct proc *p;
2125 	struct sbuf sb;
2126 	int flags, error = 0, error2;
2127 	pid_t pid;
2128 
2129 	if (namelen != 1)
2130 		return (EINVAL);
2131 
2132 	p = curproc;
2133 	pid = (pid_t)name[0];
2134 	if (pid == -1) {
2135 		pid = p->p_pid;
2136 	}
2137 
2138 	/*
2139 	 * If the query is for this process and it is single-threaded, there
2140 	 * is nobody to modify pargs, thus we can just read.
2141 	 */
2142 	if (pid == p->p_pid && p->p_numthreads == 1 && req->newptr == NULL &&
2143 	    (pa = p->p_args) != NULL)
2144 		return (SYSCTL_OUT(req, pa->ar_args, pa->ar_length));
2145 
2146 	flags = PGET_CANSEE;
2147 	if (req->newptr != NULL)
2148 		flags |= PGET_ISCURRENT;
2149 	error = pget(pid, flags, &p);
2150 	if (error)
2151 		return (error);
2152 
2153 	pa = p->p_args;
2154 	if (pa != NULL) {
2155 		pargs_hold(pa);
2156 		PROC_UNLOCK(p);
2157 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
2158 		pargs_drop(pa);
2159 	} else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
2160 		_PHOLD(p);
2161 		PROC_UNLOCK(p);
2162 		sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2163 		sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2164 		error = proc_getargv(curthread, p, &sb);
2165 		error2 = sbuf_finish(&sb);
2166 		PRELE(p);
2167 		sbuf_delete(&sb);
2168 		if (error == 0 && error2 != 0)
2169 			error = error2;
2170 	} else {
2171 		PROC_UNLOCK(p);
2172 	}
2173 	if (error != 0 || req->newptr == NULL)
2174 		return (error);
2175 
2176 	if (req->newlen > ps_arg_cache_limit - sizeof(struct pargs))
2177 		return (ENOMEM);
2178 
2179 	if (req->newlen == 0) {
2180 		/*
2181 		 * Clear the argument pointer, so that we'll fetch arguments
2182 		 * with proc_getargv() until further notice.
2183 		 */
2184 		newpa = NULL;
2185 	} else {
2186 		newpa = pargs_alloc(req->newlen);
2187 		error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
2188 		if (error != 0) {
2189 			pargs_free(newpa);
2190 			return (error);
2191 		}
2192 	}
2193 	PROC_LOCK(p);
2194 	pa = p->p_args;
2195 	p->p_args = newpa;
2196 	PROC_UNLOCK(p);
2197 	pargs_drop(pa);
2198 	return (0);
2199 }
2200 
2201 /*
2202  * This sysctl allows a process to retrieve environment of another process.
2203  */
2204 static int
2205 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
2206 {
2207 	int *name = (int *)arg1;
2208 	u_int namelen = arg2;
2209 	struct proc *p;
2210 	struct sbuf sb;
2211 	int error, error2;
2212 
2213 	if (namelen != 1)
2214 		return (EINVAL);
2215 
2216 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2217 	if (error != 0)
2218 		return (error);
2219 	if ((p->p_flag & P_SYSTEM) != 0) {
2220 		PRELE(p);
2221 		return (0);
2222 	}
2223 
2224 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2225 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2226 	error = proc_getenvv(curthread, p, &sb);
2227 	error2 = sbuf_finish(&sb);
2228 	PRELE(p);
2229 	sbuf_delete(&sb);
2230 	return (error != 0 ? error : error2);
2231 }
2232 
2233 /*
2234  * This sysctl allows a process to retrieve ELF auxiliary vector of
2235  * another process.
2236  */
2237 static int
2238 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
2239 {
2240 	int *name = (int *)arg1;
2241 	u_int namelen = arg2;
2242 	struct proc *p;
2243 	struct sbuf sb;
2244 	int error, error2;
2245 
2246 	if (namelen != 1)
2247 		return (EINVAL);
2248 
2249 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2250 	if (error != 0)
2251 		return (error);
2252 	if ((p->p_flag & P_SYSTEM) != 0) {
2253 		PRELE(p);
2254 		return (0);
2255 	}
2256 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2257 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2258 	error = proc_getauxv(curthread, p, &sb);
2259 	error2 = sbuf_finish(&sb);
2260 	PRELE(p);
2261 	sbuf_delete(&sb);
2262 	return (error != 0 ? error : error2);
2263 }
2264 
2265 /*
2266  * Look up the canonical executable path running in the specified process.
2267  * It tries to return the same hardlink name as was used for execve(2).
2268  * This allows the programs that modify their behavior based on their progname,
2269  * to operate correctly.
2270  *
2271  * Result is returned in retbuf, it must not be freed, similar to vn_fullpath()
2272  *   calling conventions.
2273  * binname is a pointer to temporary string buffer of length MAXPATHLEN,
2274  *   allocated and freed by caller.
2275  * freebuf should be freed by caller, from the M_TEMP malloc type.
2276  */
2277 int
2278 proc_get_binpath(struct proc *p, char *binname, char **retbuf,
2279     char **freebuf)
2280 {
2281 	struct nameidata nd;
2282 	struct vnode *vp, *dvp;
2283 	size_t freepath_size;
2284 	int error;
2285 	bool do_fullpath;
2286 
2287 	PROC_LOCK_ASSERT(p, MA_OWNED);
2288 
2289 	vp = p->p_textvp;
2290 	if (vp == NULL) {
2291 		PROC_UNLOCK(p);
2292 		*retbuf = "";
2293 		*freebuf = NULL;
2294 		return (0);
2295 	}
2296 	vref(vp);
2297 	dvp = p->p_textdvp;
2298 	if (dvp != NULL)
2299 		vref(dvp);
2300 	if (p->p_binname != NULL)
2301 		strlcpy(binname, p->p_binname, MAXPATHLEN);
2302 	PROC_UNLOCK(p);
2303 
2304 	do_fullpath = true;
2305 	*freebuf = NULL;
2306 	if (dvp != NULL && binname[0] != '\0') {
2307 		freepath_size = MAXPATHLEN;
2308 		if (vn_fullpath_hardlink(vp, dvp, binname, strlen(binname),
2309 		    retbuf, freebuf, &freepath_size) == 0) {
2310 			/*
2311 			 * Recheck the looked up path.  The binary
2312 			 * might have been renamed or replaced, in
2313 			 * which case we should not report old name.
2314 			 */
2315 			NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, *retbuf);
2316 			error = namei(&nd);
2317 			if (error == 0) {
2318 				if (nd.ni_vp == vp)
2319 					do_fullpath = false;
2320 				vrele(nd.ni_vp);
2321 				NDFREE_PNBUF(&nd);
2322 			}
2323 		}
2324 	}
2325 	if (do_fullpath) {
2326 		free(*freebuf, M_TEMP);
2327 		*freebuf = NULL;
2328 		error = vn_fullpath(vp, retbuf, freebuf);
2329 	}
2330 	vrele(vp);
2331 	if (dvp != NULL)
2332 		vrele(dvp);
2333 	return (error);
2334 }
2335 
2336 /*
2337  * This sysctl allows a process to retrieve the path of the executable for
2338  * itself or another process.
2339  */
2340 static int
2341 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
2342 {
2343 	pid_t *pidp = (pid_t *)arg1;
2344 	unsigned int arglen = arg2;
2345 	struct proc *p;
2346 	char *retbuf, *freebuf, *binname;
2347 	int error;
2348 
2349 	if (arglen != 1)
2350 		return (EINVAL);
2351 	binname = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
2352 	binname[0] = '\0';
2353 	if (*pidp == -1) {	/* -1 means this process */
2354 		error = 0;
2355 		p = req->td->td_proc;
2356 		PROC_LOCK(p);
2357 	} else {
2358 		error = pget(*pidp, PGET_CANSEE, &p);
2359 	}
2360 
2361 	if (error == 0)
2362 		error = proc_get_binpath(p, binname, &retbuf, &freebuf);
2363 	free(binname, M_TEMP);
2364 	if (error != 0)
2365 		return (error);
2366 	error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
2367 	free(freebuf, M_TEMP);
2368 	return (error);
2369 }
2370 
2371 static int
2372 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
2373 {
2374 	struct proc *p;
2375 	char *sv_name;
2376 	int *name;
2377 	int namelen;
2378 	int error;
2379 
2380 	namelen = arg2;
2381 	if (namelen != 1)
2382 		return (EINVAL);
2383 
2384 	name = (int *)arg1;
2385 	error = pget((pid_t)name[0], PGET_CANSEE, &p);
2386 	if (error != 0)
2387 		return (error);
2388 	sv_name = p->p_sysent->sv_name;
2389 	PROC_UNLOCK(p);
2390 	return (sysctl_handle_string(oidp, sv_name, 0, req));
2391 }
2392 
2393 #ifdef KINFO_OVMENTRY_SIZE
2394 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
2395 #endif
2396 
2397 #ifdef COMPAT_FREEBSD7
2398 static int
2399 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
2400 {
2401 	vm_map_entry_t entry, tmp_entry;
2402 	unsigned int last_timestamp, namelen;
2403 	char *fullpath, *freepath;
2404 	struct kinfo_ovmentry *kve;
2405 	struct vattr va;
2406 	struct ucred *cred;
2407 	int error, *name;
2408 	struct vnode *vp;
2409 	struct proc *p;
2410 	vm_map_t map;
2411 	struct vmspace *vm;
2412 
2413 	namelen = arg2;
2414 	if (namelen != 1)
2415 		return (EINVAL);
2416 
2417 	name = (int *)arg1;
2418 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2419 	if (error != 0)
2420 		return (error);
2421 	vm = vmspace_acquire_ref(p);
2422 	if (vm == NULL) {
2423 		PRELE(p);
2424 		return (ESRCH);
2425 	}
2426 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2427 
2428 	map = &vm->vm_map;
2429 	vm_map_lock_read(map);
2430 	VM_MAP_ENTRY_FOREACH(entry, map) {
2431 		vm_object_t obj, tobj, lobj;
2432 		vm_offset_t addr;
2433 
2434 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2435 			continue;
2436 
2437 		bzero(kve, sizeof(*kve));
2438 		kve->kve_structsize = sizeof(*kve);
2439 
2440 		kve->kve_private_resident = 0;
2441 		obj = entry->object.vm_object;
2442 		if (obj != NULL) {
2443 			VM_OBJECT_RLOCK(obj);
2444 			if (obj->shadow_count == 1)
2445 				kve->kve_private_resident =
2446 				    obj->resident_page_count;
2447 		}
2448 		kve->kve_resident = 0;
2449 		addr = entry->start;
2450 		while (addr < entry->end) {
2451 			if (pmap_extract(map->pmap, addr))
2452 				kve->kve_resident++;
2453 			addr += PAGE_SIZE;
2454 		}
2455 
2456 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2457 			if (tobj != obj) {
2458 				VM_OBJECT_RLOCK(tobj);
2459 				kve->kve_offset += tobj->backing_object_offset;
2460 			}
2461 			if (lobj != obj)
2462 				VM_OBJECT_RUNLOCK(lobj);
2463 			lobj = tobj;
2464 		}
2465 
2466 		kve->kve_start = (void*)entry->start;
2467 		kve->kve_end = (void*)entry->end;
2468 		kve->kve_offset += (off_t)entry->offset;
2469 
2470 		if (entry->protection & VM_PROT_READ)
2471 			kve->kve_protection |= KVME_PROT_READ;
2472 		if (entry->protection & VM_PROT_WRITE)
2473 			kve->kve_protection |= KVME_PROT_WRITE;
2474 		if (entry->protection & VM_PROT_EXECUTE)
2475 			kve->kve_protection |= KVME_PROT_EXEC;
2476 
2477 		if (entry->eflags & MAP_ENTRY_COW)
2478 			kve->kve_flags |= KVME_FLAG_COW;
2479 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2480 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2481 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2482 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2483 
2484 		last_timestamp = map->timestamp;
2485 		vm_map_unlock_read(map);
2486 
2487 		kve->kve_fileid = 0;
2488 		kve->kve_fsid = 0;
2489 		freepath = NULL;
2490 		fullpath = "";
2491 		if (lobj) {
2492 			kve->kve_type = vm_object_kvme_type(lobj, &vp);
2493 			if (kve->kve_type == KVME_TYPE_MGTDEVICE)
2494 				kve->kve_type = KVME_TYPE_UNKNOWN;
2495 			if (vp != NULL)
2496 				vref(vp);
2497 			if (lobj != obj)
2498 				VM_OBJECT_RUNLOCK(lobj);
2499 
2500 			kve->kve_ref_count = obj->ref_count;
2501 			kve->kve_shadow_count = obj->shadow_count;
2502 			VM_OBJECT_RUNLOCK(obj);
2503 			if (vp != NULL) {
2504 				vn_fullpath(vp, &fullpath, &freepath);
2505 				cred = curthread->td_ucred;
2506 				vn_lock(vp, LK_SHARED | LK_RETRY);
2507 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2508 					kve->kve_fileid = va.va_fileid;
2509 					/* truncate */
2510 					kve->kve_fsid = va.va_fsid;
2511 				}
2512 				vput(vp);
2513 			}
2514 		} else {
2515 			kve->kve_type = KVME_TYPE_NONE;
2516 			kve->kve_ref_count = 0;
2517 			kve->kve_shadow_count = 0;
2518 		}
2519 
2520 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2521 		if (freepath != NULL)
2522 			free(freepath, M_TEMP);
2523 
2524 		error = SYSCTL_OUT(req, kve, sizeof(*kve));
2525 		vm_map_lock_read(map);
2526 		if (error)
2527 			break;
2528 		if (last_timestamp != map->timestamp) {
2529 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2530 			entry = tmp_entry;
2531 		}
2532 	}
2533 	vm_map_unlock_read(map);
2534 	vmspace_free(vm);
2535 	PRELE(p);
2536 	free(kve, M_TEMP);
2537 	return (error);
2538 }
2539 #endif	/* COMPAT_FREEBSD7 */
2540 
2541 #ifdef KINFO_VMENTRY_SIZE
2542 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2543 #endif
2544 
2545 void
2546 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry,
2547     int *resident_count, bool *super)
2548 {
2549 	vm_object_t obj, tobj;
2550 	vm_page_t m, m_adv;
2551 	vm_offset_t addr;
2552 	vm_paddr_t pa;
2553 	vm_pindex_t pi, pi_adv, pindex;
2554 	int incore;
2555 
2556 	*super = false;
2557 	*resident_count = 0;
2558 	if (vmmap_skip_res_cnt)
2559 		return;
2560 
2561 	pa = 0;
2562 	obj = entry->object.vm_object;
2563 	addr = entry->start;
2564 	m_adv = NULL;
2565 	pi = OFF_TO_IDX(entry->offset);
2566 	for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) {
2567 		if (m_adv != NULL) {
2568 			m = m_adv;
2569 		} else {
2570 			pi_adv = atop(entry->end - addr);
2571 			pindex = pi;
2572 			for (tobj = obj;; tobj = tobj->backing_object) {
2573 				m = vm_radix_lookup_ge(&tobj->rtree, pindex);
2574 				if (m != NULL) {
2575 					if (m->pindex == pindex)
2576 						break;
2577 					if (pi_adv > m->pindex - pindex) {
2578 						pi_adv = m->pindex - pindex;
2579 						m_adv = m;
2580 					}
2581 				}
2582 				if (tobj->backing_object == NULL)
2583 					goto next;
2584 				pindex += OFF_TO_IDX(tobj->
2585 				    backing_object_offset);
2586 			}
2587 		}
2588 		m_adv = NULL;
2589 		if (m->psind != 0 && addr + pagesizes[1] <= entry->end &&
2590 		    (addr & (pagesizes[1] - 1)) == 0 && (incore =
2591 		    pmap_mincore(map->pmap, addr, &pa) & MINCORE_SUPER) != 0) {
2592 			*super = true;
2593 			/*
2594 			 * The virtual page might be smaller than the physical
2595 			 * page, so we use the page size reported by the pmap
2596 			 * rather than m->psind.
2597 			 */
2598 			pi_adv = atop(pagesizes[incore >> MINCORE_PSIND_SHIFT]);
2599 		} else {
2600 			/*
2601 			 * We do not test the found page on validity.
2602 			 * Either the page is busy and being paged in,
2603 			 * or it was invalidated.  The first case
2604 			 * should be counted as resident, the second
2605 			 * is not so clear; we do account both.
2606 			 */
2607 			pi_adv = 1;
2608 		}
2609 		*resident_count += pi_adv;
2610 next:;
2611 	}
2612 }
2613 
2614 /*
2615  * Must be called with the process locked and will return unlocked.
2616  */
2617 int
2618 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb, ssize_t maxlen, int flags)
2619 {
2620 	vm_map_entry_t entry, tmp_entry;
2621 	struct vattr va;
2622 	vm_map_t map;
2623 	vm_object_t lobj, nobj, obj, tobj;
2624 	char *fullpath, *freepath;
2625 	struct kinfo_vmentry *kve;
2626 	struct ucred *cred;
2627 	struct vnode *vp;
2628 	struct vmspace *vm;
2629 	vm_offset_t addr;
2630 	unsigned int last_timestamp;
2631 	int error;
2632 	key_t key;
2633 	unsigned short seq;
2634 	bool guard, super;
2635 
2636 	PROC_LOCK_ASSERT(p, MA_OWNED);
2637 
2638 	_PHOLD(p);
2639 	PROC_UNLOCK(p);
2640 	vm = vmspace_acquire_ref(p);
2641 	if (vm == NULL) {
2642 		PRELE(p);
2643 		return (ESRCH);
2644 	}
2645 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK | M_ZERO);
2646 
2647 	error = 0;
2648 	map = &vm->vm_map;
2649 	vm_map_lock_read(map);
2650 	VM_MAP_ENTRY_FOREACH(entry, map) {
2651 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2652 			continue;
2653 
2654 		addr = entry->end;
2655 		bzero(kve, sizeof(*kve));
2656 		obj = entry->object.vm_object;
2657 		if (obj != NULL) {
2658 			if ((obj->flags & OBJ_ANON) != 0)
2659 				kve->kve_obj = (uintptr_t)obj;
2660 
2661 			for (tobj = obj; tobj != NULL;
2662 			    tobj = tobj->backing_object) {
2663 				VM_OBJECT_RLOCK(tobj);
2664 				kve->kve_offset += tobj->backing_object_offset;
2665 				lobj = tobj;
2666 			}
2667 			if (obj->backing_object == NULL)
2668 				kve->kve_private_resident =
2669 				    obj->resident_page_count;
2670 			kern_proc_vmmap_resident(map, entry,
2671 			    &kve->kve_resident, &super);
2672 			if (super)
2673 				kve->kve_flags |= KVME_FLAG_SUPER;
2674 			for (tobj = obj; tobj != NULL; tobj = nobj) {
2675 				nobj = tobj->backing_object;
2676 				if (tobj != obj && tobj != lobj)
2677 					VM_OBJECT_RUNLOCK(tobj);
2678 			}
2679 		} else {
2680 			lobj = NULL;
2681 		}
2682 
2683 		kve->kve_start = entry->start;
2684 		kve->kve_end = entry->end;
2685 		kve->kve_offset += entry->offset;
2686 
2687 		if (entry->protection & VM_PROT_READ)
2688 			kve->kve_protection |= KVME_PROT_READ;
2689 		if (entry->protection & VM_PROT_WRITE)
2690 			kve->kve_protection |= KVME_PROT_WRITE;
2691 		if (entry->protection & VM_PROT_EXECUTE)
2692 			kve->kve_protection |= KVME_PROT_EXEC;
2693 		if (entry->max_protection & VM_PROT_READ)
2694 			kve->kve_protection |= KVME_MAX_PROT_READ;
2695 		if (entry->max_protection & VM_PROT_WRITE)
2696 			kve->kve_protection |= KVME_MAX_PROT_WRITE;
2697 		if (entry->max_protection & VM_PROT_EXECUTE)
2698 			kve->kve_protection |= KVME_MAX_PROT_EXEC;
2699 
2700 		if (entry->eflags & MAP_ENTRY_COW)
2701 			kve->kve_flags |= KVME_FLAG_COW;
2702 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2703 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2704 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2705 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2706 		if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2707 			kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2708 		if (entry->eflags & MAP_ENTRY_USER_WIRED)
2709 			kve->kve_flags |= KVME_FLAG_USER_WIRED;
2710 
2711 		guard = (entry->eflags & MAP_ENTRY_GUARD) != 0;
2712 
2713 		last_timestamp = map->timestamp;
2714 		vm_map_unlock_read(map);
2715 
2716 		freepath = NULL;
2717 		fullpath = "";
2718 		if (lobj != NULL) {
2719 			kve->kve_type = vm_object_kvme_type(lobj, &vp);
2720 			if (vp != NULL)
2721 				vref(vp);
2722 			if (lobj != obj)
2723 				VM_OBJECT_RUNLOCK(lobj);
2724 
2725 			kve->kve_ref_count = obj->ref_count;
2726 			kve->kve_shadow_count = obj->shadow_count;
2727 			if (obj->type == OBJT_DEVICE ||
2728 			    obj->type == OBJT_MGTDEVICE) {
2729 				cdev_pager_get_path(obj, kve->kve_path,
2730 				    sizeof(kve->kve_path));
2731 			}
2732 			VM_OBJECT_RUNLOCK(obj);
2733 			if ((lobj->flags & OBJ_SYSVSHM) != 0) {
2734 				kve->kve_flags |= KVME_FLAG_SYSVSHM;
2735 				shmobjinfo(lobj, &key, &seq);
2736 				kve->kve_vn_fileid = key;
2737 				kve->kve_vn_fsid_freebsd11 = seq;
2738 			}
2739 			if ((lobj->flags & OBJ_POSIXSHM) != 0) {
2740 				kve->kve_flags |= KVME_FLAG_POSIXSHM;
2741 				shm_get_path(lobj, kve->kve_path,
2742 				    sizeof(kve->kve_path));
2743 			}
2744 			if (vp != NULL) {
2745 				vn_fullpath(vp, &fullpath, &freepath);
2746 				kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2747 				cred = curthread->td_ucred;
2748 				vn_lock(vp, LK_SHARED | LK_RETRY);
2749 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2750 					kve->kve_vn_fileid = va.va_fileid;
2751 					kve->kve_vn_fsid = va.va_fsid;
2752 					kve->kve_vn_fsid_freebsd11 =
2753 					    kve->kve_vn_fsid; /* truncate */
2754 					kve->kve_vn_mode =
2755 					    MAKEIMODE(va.va_type, va.va_mode);
2756 					kve->kve_vn_size = va.va_size;
2757 					kve->kve_vn_rdev = va.va_rdev;
2758 					kve->kve_vn_rdev_freebsd11 =
2759 					    kve->kve_vn_rdev; /* truncate */
2760 					kve->kve_status = KF_ATTR_VALID;
2761 				}
2762 				vput(vp);
2763 				strlcpy(kve->kve_path, fullpath, sizeof(
2764 				    kve->kve_path));
2765 				free(freepath, M_TEMP);
2766 			}
2767 		} else {
2768 			kve->kve_type = guard ? KVME_TYPE_GUARD :
2769 			    KVME_TYPE_NONE;
2770 			kve->kve_ref_count = 0;
2771 			kve->kve_shadow_count = 0;
2772 		}
2773 
2774 		/* Pack record size down */
2775 		if ((flags & KERN_VMMAP_PACK_KINFO) != 0)
2776 			kve->kve_structsize =
2777 			    offsetof(struct kinfo_vmentry, kve_path) +
2778 			    strlen(kve->kve_path) + 1;
2779 		else
2780 			kve->kve_structsize = sizeof(*kve);
2781 		kve->kve_structsize = roundup(kve->kve_structsize,
2782 		    sizeof(uint64_t));
2783 
2784 		/* Halt filling and truncate rather than exceeding maxlen */
2785 		if (maxlen != -1 && maxlen < kve->kve_structsize) {
2786 			error = 0;
2787 			vm_map_lock_read(map);
2788 			break;
2789 		} else if (maxlen != -1)
2790 			maxlen -= kve->kve_structsize;
2791 
2792 		if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0)
2793 			error = ENOMEM;
2794 		vm_map_lock_read(map);
2795 		if (error != 0)
2796 			break;
2797 		if (last_timestamp != map->timestamp) {
2798 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2799 			entry = tmp_entry;
2800 		}
2801 	}
2802 	vm_map_unlock_read(map);
2803 	vmspace_free(vm);
2804 	PRELE(p);
2805 	free(kve, M_TEMP);
2806 	return (error);
2807 }
2808 
2809 static int
2810 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2811 {
2812 	struct proc *p;
2813 	struct sbuf sb;
2814 	u_int namelen;
2815 	int error, error2, *name;
2816 
2817 	namelen = arg2;
2818 	if (namelen != 1)
2819 		return (EINVAL);
2820 
2821 	name = (int *)arg1;
2822 	sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req);
2823 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2824 	error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
2825 	if (error != 0) {
2826 		sbuf_delete(&sb);
2827 		return (error);
2828 	}
2829 	error = kern_proc_vmmap_out(p, &sb, -1, KERN_VMMAP_PACK_KINFO);
2830 	error2 = sbuf_finish(&sb);
2831 	sbuf_delete(&sb);
2832 	return (error != 0 ? error : error2);
2833 }
2834 
2835 #if defined(STACK) || defined(DDB)
2836 static int
2837 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2838 {
2839 	struct kinfo_kstack *kkstp;
2840 	int error, i, *name, numthreads;
2841 	lwpid_t *lwpidarray;
2842 	struct thread *td;
2843 	struct stack *st;
2844 	struct sbuf sb;
2845 	struct proc *p;
2846 	u_int namelen;
2847 
2848 	namelen = arg2;
2849 	if (namelen != 1)
2850 		return (EINVAL);
2851 
2852 	name = (int *)arg1;
2853 	error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2854 	if (error != 0)
2855 		return (error);
2856 
2857 	kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2858 	st = stack_create(M_WAITOK);
2859 
2860 	lwpidarray = NULL;
2861 	PROC_LOCK(p);
2862 	do {
2863 		if (lwpidarray != NULL) {
2864 			free(lwpidarray, M_TEMP);
2865 			lwpidarray = NULL;
2866 		}
2867 		numthreads = p->p_numthreads;
2868 		PROC_UNLOCK(p);
2869 		lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2870 		    M_WAITOK | M_ZERO);
2871 		PROC_LOCK(p);
2872 	} while (numthreads < p->p_numthreads);
2873 
2874 	/*
2875 	 * XXXRW: During the below loop, execve(2) and countless other sorts
2876 	 * of changes could have taken place.  Should we check to see if the
2877 	 * vmspace has been replaced, or the like, in order to prevent
2878 	 * giving a snapshot that spans, say, execve(2), with some threads
2879 	 * before and some after?  Among other things, the credentials could
2880 	 * have changed, in which case the right to extract debug info might
2881 	 * no longer be assured.
2882 	 */
2883 	i = 0;
2884 	FOREACH_THREAD_IN_PROC(p, td) {
2885 		KASSERT(i < numthreads,
2886 		    ("sysctl_kern_proc_kstack: numthreads"));
2887 		lwpidarray[i] = td->td_tid;
2888 		i++;
2889 	}
2890 	PROC_UNLOCK(p);
2891 	numthreads = i;
2892 	for (i = 0; i < numthreads; i++) {
2893 		td = tdfind(lwpidarray[i], p->p_pid);
2894 		if (td == NULL) {
2895 			continue;
2896 		}
2897 		bzero(kkstp, sizeof(*kkstp));
2898 		(void)sbuf_new(&sb, kkstp->kkst_trace,
2899 		    sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2900 		thread_lock(td);
2901 		kkstp->kkst_tid = td->td_tid;
2902 		if (stack_save_td(st, td) == 0)
2903 			kkstp->kkst_state = KKST_STATE_STACKOK;
2904 		else
2905 			kkstp->kkst_state = KKST_STATE_RUNNING;
2906 		thread_unlock(td);
2907 		PROC_UNLOCK(p);
2908 		stack_sbuf_print(&sb, st);
2909 		sbuf_finish(&sb);
2910 		sbuf_delete(&sb);
2911 		error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2912 		if (error)
2913 			break;
2914 	}
2915 	PRELE(p);
2916 	if (lwpidarray != NULL)
2917 		free(lwpidarray, M_TEMP);
2918 	stack_destroy(st);
2919 	free(kkstp, M_TEMP);
2920 	return (error);
2921 }
2922 #endif
2923 
2924 /*
2925  * This sysctl allows a process to retrieve the full list of groups from
2926  * itself or another process.
2927  */
2928 static int
2929 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2930 {
2931 	pid_t *pidp = (pid_t *)arg1;
2932 	unsigned int arglen = arg2;
2933 	struct proc *p;
2934 	struct ucred *cred;
2935 	int error;
2936 
2937 	if (arglen != 1)
2938 		return (EINVAL);
2939 	if (*pidp == -1) {	/* -1 means this process */
2940 		p = req->td->td_proc;
2941 		PROC_LOCK(p);
2942 	} else {
2943 		error = pget(*pidp, PGET_CANSEE, &p);
2944 		if (error != 0)
2945 			return (error);
2946 	}
2947 
2948 	cred = crhold(p->p_ucred);
2949 	PROC_UNLOCK(p);
2950 
2951 	error = SYSCTL_OUT(req, &cred->cr_gid, sizeof(gid_t));
2952 	if (error == 0)
2953 		error = SYSCTL_OUT(req, cred->cr_groups,
2954 		    cred->cr_ngroups * sizeof(gid_t));
2955 
2956 	crfree(cred);
2957 	return (error);
2958 }
2959 
2960 /*
2961  * This sysctl allows a process to retrieve or/and set the resource limit for
2962  * another process.
2963  */
2964 static int
2965 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2966 {
2967 	int *name = (int *)arg1;
2968 	u_int namelen = arg2;
2969 	struct rlimit rlim;
2970 	struct proc *p;
2971 	u_int which;
2972 	int flags, error;
2973 
2974 	if (namelen != 2)
2975 		return (EINVAL);
2976 
2977 	which = (u_int)name[1];
2978 	if (which >= RLIM_NLIMITS)
2979 		return (EINVAL);
2980 
2981 	if (req->newptr != NULL && req->newlen != sizeof(rlim))
2982 		return (EINVAL);
2983 
2984 	flags = PGET_HOLD | PGET_NOTWEXIT;
2985 	if (req->newptr != NULL)
2986 		flags |= PGET_CANDEBUG;
2987 	else
2988 		flags |= PGET_CANSEE;
2989 	error = pget((pid_t)name[0], flags, &p);
2990 	if (error != 0)
2991 		return (error);
2992 
2993 	/*
2994 	 * Retrieve limit.
2995 	 */
2996 	if (req->oldptr != NULL) {
2997 		PROC_LOCK(p);
2998 		lim_rlimit_proc(p, which, &rlim);
2999 		PROC_UNLOCK(p);
3000 	}
3001 	error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
3002 	if (error != 0)
3003 		goto errout;
3004 
3005 	/*
3006 	 * Set limit.
3007 	 */
3008 	if (req->newptr != NULL) {
3009 		error = SYSCTL_IN(req, &rlim, sizeof(rlim));
3010 		if (error == 0)
3011 			error = kern_proc_setrlimit(curthread, p, which, &rlim);
3012 	}
3013 
3014 errout:
3015 	PRELE(p);
3016 	return (error);
3017 }
3018 
3019 /*
3020  * This sysctl allows a process to retrieve ps_strings structure location of
3021  * another process.
3022  */
3023 static int
3024 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
3025 {
3026 	int *name = (int *)arg1;
3027 	u_int namelen = arg2;
3028 	struct proc *p;
3029 	vm_offset_t ps_strings;
3030 	int error;
3031 #ifdef COMPAT_FREEBSD32
3032 	uint32_t ps_strings32;
3033 #endif
3034 
3035 	if (namelen != 1)
3036 		return (EINVAL);
3037 
3038 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
3039 	if (error != 0)
3040 		return (error);
3041 #ifdef COMPAT_FREEBSD32
3042 	if ((req->flags & SCTL_MASK32) != 0) {
3043 		/*
3044 		 * We return 0 if the 32 bit emulation request is for a 64 bit
3045 		 * process.
3046 		 */
3047 		ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
3048 		    PTROUT(PROC_PS_STRINGS(p)) : 0;
3049 		PROC_UNLOCK(p);
3050 		error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
3051 		return (error);
3052 	}
3053 #endif
3054 	ps_strings = PROC_PS_STRINGS(p);
3055 	PROC_UNLOCK(p);
3056 	error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
3057 	return (error);
3058 }
3059 
3060 /*
3061  * This sysctl allows a process to retrieve umask of another process.
3062  */
3063 static int
3064 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
3065 {
3066 	int *name = (int *)arg1;
3067 	u_int namelen = arg2;
3068 	struct proc *p;
3069 	int error;
3070 	u_short cmask;
3071 	pid_t pid;
3072 
3073 	if (namelen != 1)
3074 		return (EINVAL);
3075 
3076 	pid = (pid_t)name[0];
3077 	p = curproc;
3078 	if (pid == p->p_pid || pid == 0) {
3079 		cmask = p->p_pd->pd_cmask;
3080 		goto out;
3081 	}
3082 
3083 	error = pget(pid, PGET_WANTREAD, &p);
3084 	if (error != 0)
3085 		return (error);
3086 
3087 	cmask = p->p_pd->pd_cmask;
3088 	PRELE(p);
3089 out:
3090 	error = SYSCTL_OUT(req, &cmask, sizeof(cmask));
3091 	return (error);
3092 }
3093 
3094 /*
3095  * This sysctl allows a process to set and retrieve binary osreldate of
3096  * another process.
3097  */
3098 static int
3099 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
3100 {
3101 	int *name = (int *)arg1;
3102 	u_int namelen = arg2;
3103 	struct proc *p;
3104 	int flags, error, osrel;
3105 
3106 	if (namelen != 1)
3107 		return (EINVAL);
3108 
3109 	if (req->newptr != NULL && req->newlen != sizeof(osrel))
3110 		return (EINVAL);
3111 
3112 	flags = PGET_HOLD | PGET_NOTWEXIT;
3113 	if (req->newptr != NULL)
3114 		flags |= PGET_CANDEBUG;
3115 	else
3116 		flags |= PGET_CANSEE;
3117 	error = pget((pid_t)name[0], flags, &p);
3118 	if (error != 0)
3119 		return (error);
3120 
3121 	error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
3122 	if (error != 0)
3123 		goto errout;
3124 
3125 	if (req->newptr != NULL) {
3126 		error = SYSCTL_IN(req, &osrel, sizeof(osrel));
3127 		if (error != 0)
3128 			goto errout;
3129 		if (osrel < 0) {
3130 			error = EINVAL;
3131 			goto errout;
3132 		}
3133 		p->p_osrel = osrel;
3134 	}
3135 errout:
3136 	PRELE(p);
3137 	return (error);
3138 }
3139 
3140 static int
3141 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS)
3142 {
3143 	int *name = (int *)arg1;
3144 	u_int namelen = arg2;
3145 	struct proc *p;
3146 	struct kinfo_sigtramp kst;
3147 	const struct sysentvec *sv;
3148 	int error;
3149 #ifdef COMPAT_FREEBSD32
3150 	struct kinfo_sigtramp32 kst32;
3151 #endif
3152 
3153 	if (namelen != 1)
3154 		return (EINVAL);
3155 
3156 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
3157 	if (error != 0)
3158 		return (error);
3159 	sv = p->p_sysent;
3160 #ifdef COMPAT_FREEBSD32
3161 	if ((req->flags & SCTL_MASK32) != 0) {
3162 		bzero(&kst32, sizeof(kst32));
3163 		if (SV_PROC_FLAG(p, SV_ILP32)) {
3164 			if (PROC_HAS_SHP(p)) {
3165 				kst32.ksigtramp_start = PROC_SIGCODE(p);
3166 				kst32.ksigtramp_end = kst32.ksigtramp_start +
3167 				    ((sv->sv_flags & SV_DSO_SIG) == 0 ?
3168 				    *sv->sv_szsigcode :
3169 				    (uintptr_t)sv->sv_szsigcode);
3170 			} else {
3171 				kst32.ksigtramp_start = PROC_PS_STRINGS(p) -
3172 				    *sv->sv_szsigcode;
3173 				kst32.ksigtramp_end = PROC_PS_STRINGS(p);
3174 			}
3175 		}
3176 		PROC_UNLOCK(p);
3177 		error = SYSCTL_OUT(req, &kst32, sizeof(kst32));
3178 		return (error);
3179 	}
3180 #endif
3181 	bzero(&kst, sizeof(kst));
3182 	if (PROC_HAS_SHP(p)) {
3183 		kst.ksigtramp_start = (char *)PROC_SIGCODE(p);
3184 		kst.ksigtramp_end = (char *)kst.ksigtramp_start +
3185 		    ((sv->sv_flags & SV_DSO_SIG) == 0 ? *sv->sv_szsigcode :
3186 		    (uintptr_t)sv->sv_szsigcode);
3187 	} else {
3188 		kst.ksigtramp_start = (char *)PROC_PS_STRINGS(p) -
3189 		    *sv->sv_szsigcode;
3190 		kst.ksigtramp_end = (char *)PROC_PS_STRINGS(p);
3191 	}
3192 	PROC_UNLOCK(p);
3193 	error = SYSCTL_OUT(req, &kst, sizeof(kst));
3194 	return (error);
3195 }
3196 
3197 static int
3198 sysctl_kern_proc_sigfastblk(SYSCTL_HANDLER_ARGS)
3199 {
3200 	int *name = (int *)arg1;
3201 	u_int namelen = arg2;
3202 	pid_t pid;
3203 	struct proc *p;
3204 	struct thread *td1;
3205 	uintptr_t addr;
3206 #ifdef COMPAT_FREEBSD32
3207 	uint32_t addr32;
3208 #endif
3209 	int error;
3210 
3211 	if (namelen != 1 || req->newptr != NULL)
3212 		return (EINVAL);
3213 
3214 	pid = (pid_t)name[0];
3215 	error = pget(pid, PGET_HOLD | PGET_NOTWEXIT | PGET_CANDEBUG, &p);
3216 	if (error != 0)
3217 		return (error);
3218 
3219 	PROC_LOCK(p);
3220 #ifdef COMPAT_FREEBSD32
3221 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3222 		if (!SV_PROC_FLAG(p, SV_ILP32)) {
3223 			error = EINVAL;
3224 			goto errlocked;
3225 		}
3226 	}
3227 #endif
3228 	if (pid <= PID_MAX) {
3229 		td1 = FIRST_THREAD_IN_PROC(p);
3230 	} else {
3231 		FOREACH_THREAD_IN_PROC(p, td1) {
3232 			if (td1->td_tid == pid)
3233 				break;
3234 		}
3235 	}
3236 	if (td1 == NULL) {
3237 		error = ESRCH;
3238 		goto errlocked;
3239 	}
3240 	/*
3241 	 * The access to the private thread flags.  It is fine as far
3242 	 * as no out-of-thin-air values are read from td_pflags, and
3243 	 * usermode read of the td_sigblock_ptr is racy inherently,
3244 	 * since target process might have already changed it
3245 	 * meantime.
3246 	 */
3247 	if ((td1->td_pflags & TDP_SIGFASTBLOCK) != 0)
3248 		addr = (uintptr_t)td1->td_sigblock_ptr;
3249 	else
3250 		error = ENOTTY;
3251 
3252 errlocked:
3253 	_PRELE(p);
3254 	PROC_UNLOCK(p);
3255 	if (error != 0)
3256 		return (error);
3257 
3258 #ifdef COMPAT_FREEBSD32
3259 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3260 		addr32 = addr;
3261 		error = SYSCTL_OUT(req, &addr32, sizeof(addr32));
3262 	} else
3263 #endif
3264 		error = SYSCTL_OUT(req, &addr, sizeof(addr));
3265 	return (error);
3266 }
3267 
3268 static int
3269 sysctl_kern_proc_vm_layout(SYSCTL_HANDLER_ARGS)
3270 {
3271 	struct kinfo_vm_layout kvm;
3272 	struct proc *p;
3273 	struct vmspace *vmspace;
3274 	int error, *name;
3275 
3276 	name = (int *)arg1;
3277 	if ((u_int)arg2 != 1)
3278 		return (EINVAL);
3279 
3280 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
3281 	if (error != 0)
3282 		return (error);
3283 #ifdef COMPAT_FREEBSD32
3284 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3285 		if (!SV_PROC_FLAG(p, SV_ILP32)) {
3286 			PROC_UNLOCK(p);
3287 			return (EINVAL);
3288 		}
3289 	}
3290 #endif
3291 	vmspace = vmspace_acquire_ref(p);
3292 	PROC_UNLOCK(p);
3293 
3294 	memset(&kvm, 0, sizeof(kvm));
3295 	kvm.kvm_min_user_addr = vm_map_min(&vmspace->vm_map);
3296 	kvm.kvm_max_user_addr = vm_map_max(&vmspace->vm_map);
3297 	kvm.kvm_text_addr = (uintptr_t)vmspace->vm_taddr;
3298 	kvm.kvm_text_size = vmspace->vm_tsize;
3299 	kvm.kvm_data_addr = (uintptr_t)vmspace->vm_daddr;
3300 	kvm.kvm_data_size = vmspace->vm_dsize;
3301 	kvm.kvm_stack_addr = (uintptr_t)vmspace->vm_maxsaddr;
3302 	kvm.kvm_stack_size = vmspace->vm_ssize;
3303 	kvm.kvm_shp_addr = vmspace->vm_shp_base;
3304 	kvm.kvm_shp_size = p->p_sysent->sv_shared_page_len;
3305 	if ((vmspace->vm_map.flags & MAP_WIREFUTURE) != 0)
3306 		kvm.kvm_map_flags |= KMAP_FLAG_WIREFUTURE;
3307 	if ((vmspace->vm_map.flags & MAP_ASLR) != 0)
3308 		kvm.kvm_map_flags |= KMAP_FLAG_ASLR;
3309 	if ((vmspace->vm_map.flags & MAP_ASLR_IGNSTART) != 0)
3310 		kvm.kvm_map_flags |= KMAP_FLAG_ASLR_IGNSTART;
3311 	if ((vmspace->vm_map.flags & MAP_WXORX) != 0)
3312 		kvm.kvm_map_flags |= KMAP_FLAG_WXORX;
3313 	if ((vmspace->vm_map.flags & MAP_ASLR_STACK) != 0)
3314 		kvm.kvm_map_flags |= KMAP_FLAG_ASLR_STACK;
3315 	if (vmspace->vm_shp_base != p->p_sysent->sv_shared_page_base &&
3316 	    PROC_HAS_SHP(p))
3317 		kvm.kvm_map_flags |= KMAP_FLAG_ASLR_SHARED_PAGE;
3318 
3319 #ifdef COMPAT_FREEBSD32
3320 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3321 		struct kinfo_vm_layout32 kvm32;
3322 
3323 		memset(&kvm32, 0, sizeof(kvm32));
3324 		kvm32.kvm_min_user_addr = (uint32_t)kvm.kvm_min_user_addr;
3325 		kvm32.kvm_max_user_addr = (uint32_t)kvm.kvm_max_user_addr;
3326 		kvm32.kvm_text_addr = (uint32_t)kvm.kvm_text_addr;
3327 		kvm32.kvm_text_size = (uint32_t)kvm.kvm_text_size;
3328 		kvm32.kvm_data_addr = (uint32_t)kvm.kvm_data_addr;
3329 		kvm32.kvm_data_size = (uint32_t)kvm.kvm_data_size;
3330 		kvm32.kvm_stack_addr = (uint32_t)kvm.kvm_stack_addr;
3331 		kvm32.kvm_stack_size = (uint32_t)kvm.kvm_stack_size;
3332 		kvm32.kvm_shp_addr = (uint32_t)kvm.kvm_shp_addr;
3333 		kvm32.kvm_shp_size = (uint32_t)kvm.kvm_shp_size;
3334 		kvm32.kvm_map_flags = kvm.kvm_map_flags;
3335 		error = SYSCTL_OUT(req, &kvm32, sizeof(kvm32));
3336 		goto out;
3337 	}
3338 #endif
3339 
3340 	error = SYSCTL_OUT(req, &kvm, sizeof(kvm));
3341 #ifdef COMPAT_FREEBSD32
3342 out:
3343 #endif
3344 	vmspace_free(vmspace);
3345 	return (error);
3346 }
3347 
3348 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,  0,
3349     "Process table");
3350 
3351 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
3352 	CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
3353 	"Return entire process table");
3354 
3355 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3356 	sysctl_kern_proc, "Process table");
3357 
3358 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
3359 	sysctl_kern_proc, "Process table");
3360 
3361 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3362 	sysctl_kern_proc, "Process table");
3363 
3364 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
3365 	CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3366 
3367 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
3368 	sysctl_kern_proc, "Process table");
3369 
3370 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3371 	sysctl_kern_proc, "Process table");
3372 
3373 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3374 	sysctl_kern_proc, "Process table");
3375 
3376 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3377 	sysctl_kern_proc, "Process table");
3378 
3379 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
3380 	sysctl_kern_proc, "Return process table, no threads");
3381 
3382 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
3383 	CTLFLAG_RW | CTLFLAG_CAPWR | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
3384 	sysctl_kern_proc_args, "Process argument list");
3385 
3386 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
3387 	sysctl_kern_proc_env, "Process environment");
3388 
3389 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
3390 	CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
3391 
3392 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
3393 	CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
3394 
3395 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
3396 	CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
3397 	"Process syscall vector name (ABI type)");
3398 
3399 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
3400 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3401 
3402 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
3403 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3404 
3405 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
3406 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3407 
3408 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
3409 	sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3410 
3411 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
3412 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3413 
3414 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
3415 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3416 
3417 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
3418 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3419 
3420 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
3421 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3422 
3423 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
3424 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
3425 	"Return process table, including threads");
3426 
3427 #ifdef COMPAT_FREEBSD7
3428 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
3429 	CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
3430 #endif
3431 
3432 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
3433 	CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
3434 
3435 #if defined(STACK) || defined(DDB)
3436 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
3437 	CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
3438 #endif
3439 
3440 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
3441 	CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
3442 
3443 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
3444 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
3445 	"Process resource limits");
3446 
3447 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
3448 	CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
3449 	"Process ps_strings location");
3450 
3451 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
3452 	CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
3453 
3454 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
3455 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
3456 	"Process binary osreldate");
3457 
3458 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD |
3459 	CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp,
3460 	"Process signal trampoline location");
3461 
3462 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGFASTBLK, sigfastblk, CTLFLAG_RD |
3463 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_sigfastblk,
3464 	"Thread sigfastblock address");
3465 
3466 static SYSCTL_NODE(_kern_proc, KERN_PROC_VM_LAYOUT, vm_layout, CTLFLAG_RD |
3467 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_vm_layout,
3468 	"Process virtual address space layout info");
3469 
3470 static struct sx stop_all_proc_blocker;
3471 SX_SYSINIT(stop_all_proc_blocker, &stop_all_proc_blocker, "sapblk");
3472 
3473 bool
3474 stop_all_proc_block(void)
3475 {
3476 	return (sx_xlock_sig(&stop_all_proc_blocker) == 0);
3477 }
3478 
3479 void
3480 stop_all_proc_unblock(void)
3481 {
3482 	sx_xunlock(&stop_all_proc_blocker);
3483 }
3484 
3485 int allproc_gen;
3486 
3487 /*
3488  * stop_all_proc() purpose is to stop all process which have usermode,
3489  * except current process for obvious reasons.  This makes it somewhat
3490  * unreliable when invoked from multithreaded process.  The service
3491  * must not be user-callable anyway.
3492  */
3493 void
3494 stop_all_proc(void)
3495 {
3496 	struct proc *cp, *p;
3497 	int r, gen;
3498 	bool restart, seen_stopped, seen_exiting, stopped_some;
3499 
3500 	if (!stop_all_proc_block())
3501 		return;
3502 
3503 	cp = curproc;
3504 allproc_loop:
3505 	sx_xlock(&allproc_lock);
3506 	gen = allproc_gen;
3507 	seen_exiting = seen_stopped = stopped_some = restart = false;
3508 	LIST_REMOVE(cp, p_list);
3509 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3510 	for (;;) {
3511 		p = LIST_NEXT(cp, p_list);
3512 		if (p == NULL)
3513 			break;
3514 		LIST_REMOVE(cp, p_list);
3515 		LIST_INSERT_AFTER(p, cp, p_list);
3516 		PROC_LOCK(p);
3517 		if ((p->p_flag & (P_KPROC | P_SYSTEM | P_TOTAL_STOP |
3518 		    P_STOPPED_SIG)) != 0) {
3519 			PROC_UNLOCK(p);
3520 			continue;
3521 		}
3522 		if ((p->p_flag2 & P2_WEXIT) != 0) {
3523 			seen_exiting = true;
3524 			PROC_UNLOCK(p);
3525 			continue;
3526 		}
3527 		if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) {
3528 			/*
3529 			 * Stopped processes are tolerated when there
3530 			 * are no other processes which might continue
3531 			 * them.  P_STOPPED_SINGLE but not
3532 			 * P_TOTAL_STOP process still has at least one
3533 			 * thread running.
3534 			 */
3535 			seen_stopped = true;
3536 			PROC_UNLOCK(p);
3537 			continue;
3538 		}
3539 		if ((p->p_flag & P_TRACED) != 0) {
3540 			/*
3541 			 * thread_single() below cannot stop traced p,
3542 			 * so skip it.  OTOH, we cannot require
3543 			 * restart because debugger might be either
3544 			 * already stopped or traced as well.
3545 			 */
3546 			PROC_UNLOCK(p);
3547 			continue;
3548 		}
3549 		sx_xunlock(&allproc_lock);
3550 		_PHOLD(p);
3551 		r = thread_single(p, SINGLE_ALLPROC);
3552 		if (r != 0)
3553 			restart = true;
3554 		else
3555 			stopped_some = true;
3556 		_PRELE(p);
3557 		PROC_UNLOCK(p);
3558 		sx_xlock(&allproc_lock);
3559 	}
3560 	/* Catch forked children we did not see in iteration. */
3561 	if (gen != allproc_gen)
3562 		restart = true;
3563 	sx_xunlock(&allproc_lock);
3564 	if (restart || stopped_some || seen_exiting || seen_stopped) {
3565 		kern_yield(PRI_USER);
3566 		goto allproc_loop;
3567 	}
3568 }
3569 
3570 void
3571 resume_all_proc(void)
3572 {
3573 	struct proc *cp, *p;
3574 
3575 	cp = curproc;
3576 	sx_xlock(&allproc_lock);
3577 again:
3578 	LIST_REMOVE(cp, p_list);
3579 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3580 	for (;;) {
3581 		p = LIST_NEXT(cp, p_list);
3582 		if (p == NULL)
3583 			break;
3584 		LIST_REMOVE(cp, p_list);
3585 		LIST_INSERT_AFTER(p, cp, p_list);
3586 		PROC_LOCK(p);
3587 		if ((p->p_flag & P_TOTAL_STOP) != 0) {
3588 			sx_xunlock(&allproc_lock);
3589 			_PHOLD(p);
3590 			thread_single_end(p, SINGLE_ALLPROC);
3591 			_PRELE(p);
3592 			PROC_UNLOCK(p);
3593 			sx_xlock(&allproc_lock);
3594 		} else {
3595 			PROC_UNLOCK(p);
3596 		}
3597 	}
3598 	/*  Did the loop above missed any stopped process ? */
3599 	FOREACH_PROC_IN_SYSTEM(p) {
3600 		/* No need for proc lock. */
3601 		if ((p->p_flag & P_TOTAL_STOP) != 0)
3602 			goto again;
3603 	}
3604 	sx_xunlock(&allproc_lock);
3605 
3606 	stop_all_proc_unblock();
3607 }
3608 
3609 /* #define	TOTAL_STOP_DEBUG	1 */
3610 #ifdef TOTAL_STOP_DEBUG
3611 volatile static int ap_resume;
3612 #include <sys/mount.h>
3613 
3614 static int
3615 sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS)
3616 {
3617 	int error, val;
3618 
3619 	val = 0;
3620 	ap_resume = 0;
3621 	error = sysctl_handle_int(oidp, &val, 0, req);
3622 	if (error != 0 || req->newptr == NULL)
3623 		return (error);
3624 	if (val != 0) {
3625 		stop_all_proc();
3626 		syncer_suspend();
3627 		while (ap_resume == 0)
3628 			;
3629 		syncer_resume();
3630 		resume_all_proc();
3631 	}
3632 	return (0);
3633 }
3634 
3635 SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW |
3636     CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0,
3637     sysctl_debug_stop_all_proc, "I",
3638     "");
3639 #endif
3640