xref: /freebsd/sys/kern/kern_proc.c (revision cfd6422a5217410fbd66f7a7a8a64d9d85e61229)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1982, 1986, 1989, 1991, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)kern_proc.c	8.7 (Berkeley) 2/14/95
32  */
33 
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
36 
37 #include "opt_ddb.h"
38 #include "opt_ktrace.h"
39 #include "opt_kstack_pages.h"
40 #include "opt_stack.h"
41 
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/bitstring.h>
45 #include <sys/elf.h>
46 #include <sys/eventhandler.h>
47 #include <sys/exec.h>
48 #include <sys/jail.h>
49 #include <sys/kernel.h>
50 #include <sys/limits.h>
51 #include <sys/lock.h>
52 #include <sys/loginclass.h>
53 #include <sys/malloc.h>
54 #include <sys/mman.h>
55 #include <sys/mount.h>
56 #include <sys/mutex.h>
57 #include <sys/proc.h>
58 #include <sys/ptrace.h>
59 #include <sys/refcount.h>
60 #include <sys/resourcevar.h>
61 #include <sys/rwlock.h>
62 #include <sys/sbuf.h>
63 #include <sys/sysent.h>
64 #include <sys/sched.h>
65 #include <sys/smp.h>
66 #include <sys/stack.h>
67 #include <sys/stat.h>
68 #include <sys/dtrace_bsd.h>
69 #include <sys/sysctl.h>
70 #include <sys/filedesc.h>
71 #include <sys/tty.h>
72 #include <sys/signalvar.h>
73 #include <sys/sdt.h>
74 #include <sys/sx.h>
75 #include <sys/user.h>
76 #include <sys/vnode.h>
77 #include <sys/wait.h>
78 
79 #ifdef DDB
80 #include <ddb/ddb.h>
81 #endif
82 
83 #include <vm/vm.h>
84 #include <vm/vm_param.h>
85 #include <vm/vm_extern.h>
86 #include <vm/pmap.h>
87 #include <vm/vm_map.h>
88 #include <vm/vm_object.h>
89 #include <vm/vm_page.h>
90 #include <vm/uma.h>
91 
92 #include <fs/devfs/devfs.h>
93 
94 #ifdef COMPAT_FREEBSD32
95 #include <compat/freebsd32/freebsd32.h>
96 #include <compat/freebsd32/freebsd32_util.h>
97 #endif
98 
99 SDT_PROVIDER_DEFINE(proc);
100 
101 MALLOC_DEFINE(M_SESSION, "session", "session header");
102 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
103 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
104 
105 static void doenterpgrp(struct proc *, struct pgrp *);
106 static void orphanpg(struct pgrp *pg);
107 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp);
108 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp);
109 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp,
110     int preferthread);
111 static void pgdelete(struct pgrp *);
112 static int pgrp_init(void *mem, int size, int flags);
113 static int proc_ctor(void *mem, int size, void *arg, int flags);
114 static void proc_dtor(void *mem, int size, void *arg);
115 static int proc_init(void *mem, int size, int flags);
116 static void proc_fini(void *mem, int size);
117 static void pargs_free(struct pargs *pa);
118 
119 /*
120  * Other process lists
121  */
122 struct pidhashhead *pidhashtbl;
123 struct sx *pidhashtbl_lock;
124 u_long pidhash;
125 u_long pidhashlock;
126 struct pgrphashhead *pgrphashtbl;
127 u_long pgrphash;
128 struct proclist allproc;
129 struct sx __exclusive_cache_line allproc_lock;
130 struct sx __exclusive_cache_line proctree_lock;
131 struct mtx __exclusive_cache_line ppeers_lock;
132 struct mtx __exclusive_cache_line procid_lock;
133 uma_zone_t proc_zone;
134 uma_zone_t pgrp_zone;
135 
136 /*
137  * The offset of various fields in struct proc and struct thread.
138  * These are used by kernel debuggers to enumerate kernel threads and
139  * processes.
140  */
141 const int proc_off_p_pid = offsetof(struct proc, p_pid);
142 const int proc_off_p_comm = offsetof(struct proc, p_comm);
143 const int proc_off_p_list = offsetof(struct proc, p_list);
144 const int proc_off_p_hash = offsetof(struct proc, p_hash);
145 const int proc_off_p_threads = offsetof(struct proc, p_threads);
146 const int thread_off_td_tid = offsetof(struct thread, td_tid);
147 const int thread_off_td_name = offsetof(struct thread, td_name);
148 const int thread_off_td_oncpu = offsetof(struct thread, td_oncpu);
149 const int thread_off_td_pcb = offsetof(struct thread, td_pcb);
150 const int thread_off_td_plist = offsetof(struct thread, td_plist);
151 
152 EVENTHANDLER_LIST_DEFINE(process_ctor);
153 EVENTHANDLER_LIST_DEFINE(process_dtor);
154 EVENTHANDLER_LIST_DEFINE(process_init);
155 EVENTHANDLER_LIST_DEFINE(process_fini);
156 EVENTHANDLER_LIST_DEFINE(process_exit);
157 EVENTHANDLER_LIST_DEFINE(process_fork);
158 EVENTHANDLER_LIST_DEFINE(process_exec);
159 
160 int kstack_pages = KSTACK_PAGES;
161 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0,
162     "Kernel stack size in pages");
163 static int vmmap_skip_res_cnt = 0;
164 SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW,
165     &vmmap_skip_res_cnt, 0,
166     "Skip calculation of the pages resident count in kern.proc.vmmap");
167 
168 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
169 #ifdef COMPAT_FREEBSD32
170 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
171 #endif
172 
173 /*
174  * Initialize global process hashing structures.
175  */
176 void
177 procinit(void)
178 {
179 	u_long i;
180 
181 	sx_init(&allproc_lock, "allproc");
182 	sx_init(&proctree_lock, "proctree");
183 	mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
184 	mtx_init(&procid_lock, "procid", NULL, MTX_DEF);
185 	LIST_INIT(&allproc);
186 	pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
187 	pidhashlock = (pidhash + 1) / 64;
188 	if (pidhashlock > 0)
189 		pidhashlock--;
190 	pidhashtbl_lock = malloc(sizeof(*pidhashtbl_lock) * (pidhashlock + 1),
191 	    M_PROC, M_WAITOK | M_ZERO);
192 	for (i = 0; i < pidhashlock + 1; i++)
193 		sx_init_flags(&pidhashtbl_lock[i], "pidhash", SX_DUPOK);
194 	pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
195 	proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
196 	    proc_ctor, proc_dtor, proc_init, proc_fini,
197 	    UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
198 	pgrp_zone = uma_zcreate("PGRP", sizeof(struct pgrp), NULL, NULL,
199 	    pgrp_init, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
200 	uihashinit();
201 }
202 
203 /*
204  * Prepare a proc for use.
205  */
206 static int
207 proc_ctor(void *mem, int size, void *arg, int flags)
208 {
209 	struct proc *p;
210 	struct thread *td;
211 
212 	p = (struct proc *)mem;
213 #ifdef KDTRACE_HOOKS
214 	kdtrace_proc_ctor(p);
215 #endif
216 	EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
217 	td = FIRST_THREAD_IN_PROC(p);
218 	if (td != NULL) {
219 		/* Make sure all thread constructors are executed */
220 		EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
221 	}
222 	return (0);
223 }
224 
225 /*
226  * Reclaim a proc after use.
227  */
228 static void
229 proc_dtor(void *mem, int size, void *arg)
230 {
231 	struct proc *p;
232 	struct thread *td;
233 
234 	/* INVARIANTS checks go here */
235 	p = (struct proc *)mem;
236 	td = FIRST_THREAD_IN_PROC(p);
237 	if (td != NULL) {
238 #ifdef INVARIANTS
239 		KASSERT((p->p_numthreads == 1),
240 		    ("bad number of threads in exiting process"));
241 		KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
242 #endif
243 		/* Free all OSD associated to this thread. */
244 		osd_thread_exit(td);
245 		td_softdep_cleanup(td);
246 		MPASS(td->td_su == NULL);
247 
248 		/* Make sure all thread destructors are executed */
249 		EVENTHANDLER_DIRECT_INVOKE(thread_dtor, td);
250 	}
251 	EVENTHANDLER_DIRECT_INVOKE(process_dtor, p);
252 #ifdef KDTRACE_HOOKS
253 	kdtrace_proc_dtor(p);
254 #endif
255 	if (p->p_ksi != NULL)
256 		KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
257 }
258 
259 /*
260  * Initialize type-stable parts of a proc (when newly created).
261  */
262 static int
263 proc_init(void *mem, int size, int flags)
264 {
265 	struct proc *p;
266 
267 	p = (struct proc *)mem;
268 	mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK | MTX_NEW);
269 	mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_NEW);
270 	mtx_init(&p->p_statmtx, "pstatl", NULL, MTX_SPIN | MTX_NEW);
271 	mtx_init(&p->p_itimmtx, "pitiml", NULL, MTX_SPIN | MTX_NEW);
272 	mtx_init(&p->p_profmtx, "pprofl", NULL, MTX_SPIN | MTX_NEW);
273 	cv_init(&p->p_pwait, "ppwait");
274 	TAILQ_INIT(&p->p_threads);	     /* all threads in proc */
275 	EVENTHANDLER_DIRECT_INVOKE(process_init, p);
276 	p->p_stats = pstats_alloc();
277 	p->p_pgrp = NULL;
278 	return (0);
279 }
280 
281 /*
282  * UMA should ensure that this function is never called.
283  * Freeing a proc structure would violate type stability.
284  */
285 static void
286 proc_fini(void *mem, int size)
287 {
288 #ifdef notnow
289 	struct proc *p;
290 
291 	p = (struct proc *)mem;
292 	EVENTHANDLER_DIRECT_INVOKE(process_fini, p);
293 	pstats_free(p->p_stats);
294 	thread_free(FIRST_THREAD_IN_PROC(p));
295 	mtx_destroy(&p->p_mtx);
296 	if (p->p_ksi != NULL)
297 		ksiginfo_free(p->p_ksi);
298 #else
299 	panic("proc reclaimed");
300 #endif
301 }
302 
303 static int
304 pgrp_init(void *mem, int size, int flags)
305 {
306 	struct pgrp *pg;
307 
308 	pg = mem;
309 	mtx_init(&pg->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
310 	return (0);
311 }
312 
313 /*
314  * PID space management.
315  *
316  * These bitmaps are used by fork_findpid.
317  */
318 bitstr_t bit_decl(proc_id_pidmap, PID_MAX);
319 bitstr_t bit_decl(proc_id_grpidmap, PID_MAX);
320 bitstr_t bit_decl(proc_id_sessidmap, PID_MAX);
321 bitstr_t bit_decl(proc_id_reapmap, PID_MAX);
322 
323 static bitstr_t *proc_id_array[] = {
324 	proc_id_pidmap,
325 	proc_id_grpidmap,
326 	proc_id_sessidmap,
327 	proc_id_reapmap,
328 };
329 
330 void
331 proc_id_set(int type, pid_t id)
332 {
333 
334 	KASSERT(type >= 0 && type < nitems(proc_id_array),
335 	    ("invalid type %d\n", type));
336 	mtx_lock(&procid_lock);
337 	KASSERT(bit_test(proc_id_array[type], id) == 0,
338 	    ("bit %d already set in %d\n", id, type));
339 	bit_set(proc_id_array[type], id);
340 	mtx_unlock(&procid_lock);
341 }
342 
343 void
344 proc_id_set_cond(int type, pid_t id)
345 {
346 
347 	KASSERT(type >= 0 && type < nitems(proc_id_array),
348 	    ("invalid type %d\n", type));
349 	if (bit_test(proc_id_array[type], id))
350 		return;
351 	mtx_lock(&procid_lock);
352 	bit_set(proc_id_array[type], id);
353 	mtx_unlock(&procid_lock);
354 }
355 
356 void
357 proc_id_clear(int type, pid_t id)
358 {
359 
360 	KASSERT(type >= 0 && type < nitems(proc_id_array),
361 	    ("invalid type %d\n", type));
362 	mtx_lock(&procid_lock);
363 	KASSERT(bit_test(proc_id_array[type], id) != 0,
364 	    ("bit %d not set in %d\n", id, type));
365 	bit_clear(proc_id_array[type], id);
366 	mtx_unlock(&procid_lock);
367 }
368 
369 /*
370  * Is p an inferior of the current process?
371  */
372 int
373 inferior(struct proc *p)
374 {
375 
376 	sx_assert(&proctree_lock, SX_LOCKED);
377 	PROC_LOCK_ASSERT(p, MA_OWNED);
378 	for (; p != curproc; p = proc_realparent(p)) {
379 		if (p->p_pid == 0)
380 			return (0);
381 	}
382 	return (1);
383 }
384 
385 /*
386  * Shared lock all the pid hash lists.
387  */
388 void
389 pidhash_slockall(void)
390 {
391 	u_long i;
392 
393 	for (i = 0; i < pidhashlock + 1; i++)
394 		sx_slock(&pidhashtbl_lock[i]);
395 }
396 
397 /*
398  * Shared unlock all the pid hash lists.
399  */
400 void
401 pidhash_sunlockall(void)
402 {
403 	u_long i;
404 
405 	for (i = 0; i < pidhashlock + 1; i++)
406 		sx_sunlock(&pidhashtbl_lock[i]);
407 }
408 
409 /*
410  * Similar to pfind_any(), this function finds zombies.
411  */
412 struct proc *
413 pfind_any_locked(pid_t pid)
414 {
415 	struct proc *p;
416 
417 	sx_assert(PIDHASHLOCK(pid), SX_LOCKED);
418 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
419 		if (p->p_pid == pid) {
420 			PROC_LOCK(p);
421 			if (p->p_state == PRS_NEW) {
422 				PROC_UNLOCK(p);
423 				p = NULL;
424 			}
425 			break;
426 		}
427 	}
428 	return (p);
429 }
430 
431 /*
432  * Locate a process by number.
433  *
434  * By not returning processes in the PRS_NEW state, we allow callers to avoid
435  * testing for that condition to avoid dereferencing p_ucred, et al.
436  */
437 static __always_inline struct proc *
438 _pfind(pid_t pid, bool zombie)
439 {
440 	struct proc *p;
441 
442 	p = curproc;
443 	if (p->p_pid == pid) {
444 		PROC_LOCK(p);
445 		return (p);
446 	}
447 	sx_slock(PIDHASHLOCK(pid));
448 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
449 		if (p->p_pid == pid) {
450 			PROC_LOCK(p);
451 			if (p->p_state == PRS_NEW ||
452 			    (!zombie && p->p_state == PRS_ZOMBIE)) {
453 				PROC_UNLOCK(p);
454 				p = NULL;
455 			}
456 			break;
457 		}
458 	}
459 	sx_sunlock(PIDHASHLOCK(pid));
460 	return (p);
461 }
462 
463 struct proc *
464 pfind(pid_t pid)
465 {
466 
467 	return (_pfind(pid, false));
468 }
469 
470 /*
471  * Same as pfind but allow zombies.
472  */
473 struct proc *
474 pfind_any(pid_t pid)
475 {
476 
477 	return (_pfind(pid, true));
478 }
479 
480 /*
481  * Locate a process group by number.
482  * The caller must hold proctree_lock.
483  */
484 struct pgrp *
485 pgfind(pid_t pgid)
486 {
487 	struct pgrp *pgrp;
488 
489 	sx_assert(&proctree_lock, SX_LOCKED);
490 
491 	LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
492 		if (pgrp->pg_id == pgid) {
493 			PGRP_LOCK(pgrp);
494 			return (pgrp);
495 		}
496 	}
497 	return (NULL);
498 }
499 
500 /*
501  * Locate process and do additional manipulations, depending on flags.
502  */
503 int
504 pget(pid_t pid, int flags, struct proc **pp)
505 {
506 	struct proc *p;
507 	struct thread *td1;
508 	int error;
509 
510 	p = curproc;
511 	if (p->p_pid == pid) {
512 		PROC_LOCK(p);
513 	} else {
514 		p = NULL;
515 		if (pid <= PID_MAX) {
516 			if ((flags & PGET_NOTWEXIT) == 0)
517 				p = pfind_any(pid);
518 			else
519 				p = pfind(pid);
520 		} else if ((flags & PGET_NOTID) == 0) {
521 			td1 = tdfind(pid, -1);
522 			if (td1 != NULL)
523 				p = td1->td_proc;
524 		}
525 		if (p == NULL)
526 			return (ESRCH);
527 		if ((flags & PGET_CANSEE) != 0) {
528 			error = p_cansee(curthread, p);
529 			if (error != 0)
530 				goto errout;
531 		}
532 	}
533 	if ((flags & PGET_CANDEBUG) != 0) {
534 		error = p_candebug(curthread, p);
535 		if (error != 0)
536 			goto errout;
537 	}
538 	if ((flags & PGET_ISCURRENT) != 0 && curproc != p) {
539 		error = EPERM;
540 		goto errout;
541 	}
542 	if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) {
543 		error = ESRCH;
544 		goto errout;
545 	}
546 	if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) {
547 		/*
548 		 * XXXRW: Not clear ESRCH is the right error during proc
549 		 * execve().
550 		 */
551 		error = ESRCH;
552 		goto errout;
553 	}
554 	if ((flags & PGET_HOLD) != 0) {
555 		_PHOLD(p);
556 		PROC_UNLOCK(p);
557 	}
558 	*pp = p;
559 	return (0);
560 errout:
561 	PROC_UNLOCK(p);
562 	return (error);
563 }
564 
565 /*
566  * Create a new process group.
567  * pgid must be equal to the pid of p.
568  * Begin a new session if required.
569  */
570 int
571 enterpgrp(struct proc *p, pid_t pgid, struct pgrp *pgrp, struct session *sess)
572 {
573 
574 	sx_assert(&proctree_lock, SX_XLOCKED);
575 
576 	KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
577 	KASSERT(p->p_pid == pgid,
578 	    ("enterpgrp: new pgrp and pid != pgid"));
579 	KASSERT(pgfind(pgid) == NULL,
580 	    ("enterpgrp: pgrp with pgid exists"));
581 	KASSERT(!SESS_LEADER(p),
582 	    ("enterpgrp: session leader attempted setpgrp"));
583 
584 	if (sess != NULL) {
585 		/*
586 		 * new session
587 		 */
588 		mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
589 		PROC_LOCK(p);
590 		p->p_flag &= ~P_CONTROLT;
591 		PROC_UNLOCK(p);
592 		PGRP_LOCK(pgrp);
593 		sess->s_leader = p;
594 		sess->s_sid = p->p_pid;
595 		proc_id_set(PROC_ID_SESSION, p->p_pid);
596 		refcount_init(&sess->s_count, 1);
597 		sess->s_ttyvp = NULL;
598 		sess->s_ttydp = NULL;
599 		sess->s_ttyp = NULL;
600 		bcopy(p->p_session->s_login, sess->s_login,
601 			    sizeof(sess->s_login));
602 		pgrp->pg_session = sess;
603 		KASSERT(p == curproc,
604 		    ("enterpgrp: mksession and p != curproc"));
605 	} else {
606 		pgrp->pg_session = p->p_session;
607 		sess_hold(pgrp->pg_session);
608 		PGRP_LOCK(pgrp);
609 	}
610 	pgrp->pg_id = pgid;
611 	proc_id_set(PROC_ID_GROUP, p->p_pid);
612 	LIST_INIT(&pgrp->pg_members);
613 	pgrp->pg_flags = 0;
614 
615 	/*
616 	 * As we have an exclusive lock of proctree_lock,
617 	 * this should not deadlock.
618 	 */
619 	LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
620 	SLIST_INIT(&pgrp->pg_sigiolst);
621 	PGRP_UNLOCK(pgrp);
622 
623 	doenterpgrp(p, pgrp);
624 
625 	return (0);
626 }
627 
628 /*
629  * Move p to an existing process group
630  */
631 int
632 enterthispgrp(struct proc *p, struct pgrp *pgrp)
633 {
634 
635 	sx_assert(&proctree_lock, SX_XLOCKED);
636 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
637 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
638 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
639 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
640 	KASSERT(pgrp->pg_session == p->p_session,
641 	    ("%s: pgrp's session %p, p->p_session %p proc %p\n",
642 	    __func__, pgrp->pg_session, p->p_session, p));
643 	KASSERT(pgrp != p->p_pgrp,
644 	    ("%s: p %p belongs to pgrp %p", __func__, p, pgrp));
645 
646 	doenterpgrp(p, pgrp);
647 
648 	return (0);
649 }
650 
651 /*
652  * If true, any child of q which belongs to group pgrp, qualifies the
653  * process group pgrp as not orphaned.
654  */
655 static bool
656 isjobproc(struct proc *q, struct pgrp *pgrp)
657 {
658 	sx_assert(&proctree_lock, SX_LOCKED);
659 
660 	return (q->p_pgrp != pgrp &&
661 	    q->p_pgrp->pg_session == pgrp->pg_session);
662 }
663 
664 static struct proc *
665 jobc_reaper(struct proc *p)
666 {
667 	struct proc *pp;
668 
669 	sx_assert(&proctree_lock, SA_LOCKED);
670 
671 	for (pp = p;;) {
672 		pp = pp->p_reaper;
673 		if (pp->p_reaper == pp ||
674 		    (pp->p_treeflag & P_TREE_GRPEXITED) == 0)
675 			return (pp);
676 	}
677 }
678 
679 static struct proc *
680 jobc_parent(struct proc *p, struct proc *p_exiting)
681 {
682 	struct proc *pp;
683 
684 	sx_assert(&proctree_lock, SA_LOCKED);
685 
686 	pp = proc_realparent(p);
687 	if (pp->p_pptr == NULL || pp == p_exiting ||
688 	    (pp->p_treeflag & P_TREE_GRPEXITED) == 0)
689 		return (pp);
690 	return (jobc_reaper(pp));
691 }
692 
693 static int
694 pgrp_calc_jobc(struct pgrp *pgrp)
695 {
696 	struct proc *q;
697 	int cnt;
698 
699 #ifdef INVARIANTS
700 	if (!mtx_owned(&pgrp->pg_mtx))
701 		sx_assert(&proctree_lock, SA_LOCKED);
702 #endif
703 
704 	cnt = 0;
705 	LIST_FOREACH(q, &pgrp->pg_members, p_pglist) {
706 		if ((q->p_treeflag & P_TREE_GRPEXITED) != 0 ||
707 		    q->p_pptr == NULL)
708 			continue;
709 		if (isjobproc(jobc_parent(q, NULL), pgrp))
710 			cnt++;
711 	}
712 	return (cnt);
713 }
714 
715 /*
716  * Move p to a process group
717  */
718 static void
719 doenterpgrp(struct proc *p, struct pgrp *pgrp)
720 {
721 	struct pgrp *savepgrp;
722 	struct proc *pp;
723 
724 	sx_assert(&proctree_lock, SX_XLOCKED);
725 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
726 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
727 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
728 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
729 
730 	savepgrp = p->p_pgrp;
731 	pp = jobc_parent(p, NULL);
732 
733 	PGRP_LOCK(pgrp);
734 	PGRP_LOCK(savepgrp);
735 	if (isjobproc(pp, savepgrp) && pgrp_calc_jobc(savepgrp) == 1)
736 		orphanpg(savepgrp);
737 	PROC_LOCK(p);
738 	LIST_REMOVE(p, p_pglist);
739 	p->p_pgrp = pgrp;
740 	PROC_UNLOCK(p);
741 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
742 	if (isjobproc(pp, pgrp))
743 		pgrp->pg_flags &= ~PGRP_ORPHANED;
744 	PGRP_UNLOCK(savepgrp);
745 	PGRP_UNLOCK(pgrp);
746 	if (LIST_EMPTY(&savepgrp->pg_members))
747 		pgdelete(savepgrp);
748 }
749 
750 /*
751  * remove process from process group
752  */
753 int
754 leavepgrp(struct proc *p)
755 {
756 	struct pgrp *savepgrp;
757 
758 	sx_assert(&proctree_lock, SX_XLOCKED);
759 	savepgrp = p->p_pgrp;
760 	PGRP_LOCK(savepgrp);
761 	PROC_LOCK(p);
762 	LIST_REMOVE(p, p_pglist);
763 	p->p_pgrp = NULL;
764 	PROC_UNLOCK(p);
765 	PGRP_UNLOCK(savepgrp);
766 	if (LIST_EMPTY(&savepgrp->pg_members))
767 		pgdelete(savepgrp);
768 	return (0);
769 }
770 
771 /*
772  * delete a process group
773  */
774 static void
775 pgdelete(struct pgrp *pgrp)
776 {
777 	struct session *savesess;
778 	struct tty *tp;
779 
780 	sx_assert(&proctree_lock, SX_XLOCKED);
781 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
782 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
783 
784 	/*
785 	 * Reset any sigio structures pointing to us as a result of
786 	 * F_SETOWN with our pgid.  The proctree lock ensures that
787 	 * new sigio structures will not be added after this point.
788 	 */
789 	funsetownlst(&pgrp->pg_sigiolst);
790 
791 	PGRP_LOCK(pgrp);
792 	tp = pgrp->pg_session->s_ttyp;
793 	LIST_REMOVE(pgrp, pg_hash);
794 	savesess = pgrp->pg_session;
795 	PGRP_UNLOCK(pgrp);
796 
797 	/* Remove the reference to the pgrp before deallocating it. */
798 	if (tp != NULL) {
799 		tty_lock(tp);
800 		tty_rel_pgrp(tp, pgrp);
801 	}
802 
803 	proc_id_clear(PROC_ID_GROUP, pgrp->pg_id);
804 	uma_zfree(pgrp_zone, pgrp);
805 	sess_release(savesess);
806 }
807 
808 
809 static void
810 fixjobc_kill(struct proc *p)
811 {
812 	struct proc *q;
813 	struct pgrp *pgrp;
814 
815 	sx_assert(&proctree_lock, SX_LOCKED);
816 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
817 	pgrp = p->p_pgrp;
818 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
819 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
820 
821 	/*
822 	 * p no longer affects process group orphanage for children.
823 	 * It is marked by the flag because p is only physically
824 	 * removed from its process group on wait(2).
825 	 */
826 	MPASS((p->p_treeflag & P_TREE_GRPEXITED) == 0);
827 	p->p_treeflag |= P_TREE_GRPEXITED;
828 
829 	/*
830 	 * Check if exiting p orphans its own group.
831 	 */
832 	pgrp = p->p_pgrp;
833 	if (isjobproc(jobc_parent(p, NULL), pgrp)) {
834 		PGRP_LOCK(pgrp);
835 		if (pgrp_calc_jobc(pgrp) == 0)
836 			orphanpg(pgrp);
837 		PGRP_UNLOCK(pgrp);
838 	}
839 
840 	/*
841 	 * Check this process' children to see whether they qualify
842 	 * their process groups after reparenting to reaper.
843 	 */
844 	LIST_FOREACH(q, &p->p_children, p_sibling) {
845 		pgrp = q->p_pgrp;
846 		PGRP_LOCK(pgrp);
847 		if (pgrp_calc_jobc(pgrp) == 0) {
848 			/*
849 			 * We want to handle exactly the children that
850 			 * has p as realparent.  Then, when calculating
851 			 * jobc_parent for children, we should ignore
852 			 * P_TREE_GRPEXITED flag already set on p.
853 			 */
854 			if (jobc_parent(q, p) == p && isjobproc(p, pgrp))
855 				orphanpg(pgrp);
856 		} else
857 			pgrp->pg_flags &= ~PGRP_ORPHANED;
858 		PGRP_UNLOCK(pgrp);
859 	}
860 	LIST_FOREACH(q, &p->p_orphans, p_orphan) {
861 		pgrp = q->p_pgrp;
862 		PGRP_LOCK(pgrp);
863 		if (pgrp_calc_jobc(pgrp) == 0) {
864 			if (isjobproc(p, pgrp))
865 				orphanpg(pgrp);
866 		} else
867 			pgrp->pg_flags &= ~PGRP_ORPHANED;
868 		PGRP_UNLOCK(pgrp);
869 	}
870 }
871 
872 void
873 killjobc(void)
874 {
875 	struct session *sp;
876 	struct tty *tp;
877 	struct proc *p;
878 	struct vnode *ttyvp;
879 
880 	p = curproc;
881 	MPASS(p->p_flag & P_WEXIT);
882 	sx_assert(&proctree_lock, SX_LOCKED);
883 
884 	if (SESS_LEADER(p)) {
885 		sp = p->p_session;
886 
887 		/*
888 		 * s_ttyp is not zero'd; we use this to indicate that
889 		 * the session once had a controlling terminal. (for
890 		 * logging and informational purposes)
891 		 */
892 		SESS_LOCK(sp);
893 		ttyvp = sp->s_ttyvp;
894 		tp = sp->s_ttyp;
895 		sp->s_ttyvp = NULL;
896 		sp->s_ttydp = NULL;
897 		sp->s_leader = NULL;
898 		SESS_UNLOCK(sp);
899 
900 		/*
901 		 * Signal foreground pgrp and revoke access to
902 		 * controlling terminal if it has not been revoked
903 		 * already.
904 		 *
905 		 * Because the TTY may have been revoked in the mean
906 		 * time and could already have a new session associated
907 		 * with it, make sure we don't send a SIGHUP to a
908 		 * foreground process group that does not belong to this
909 		 * session.
910 		 */
911 
912 		if (tp != NULL) {
913 			tty_lock(tp);
914 			if (tp->t_session == sp)
915 				tty_signal_pgrp(tp, SIGHUP);
916 			tty_unlock(tp);
917 		}
918 
919 		if (ttyvp != NULL) {
920 			sx_xunlock(&proctree_lock);
921 			if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) {
922 				VOP_REVOKE(ttyvp, REVOKEALL);
923 				VOP_UNLOCK(ttyvp);
924 			}
925 			devfs_ctty_unref(ttyvp);
926 			sx_xlock(&proctree_lock);
927 		}
928 	}
929 	fixjobc_kill(p);
930 }
931 
932 /*
933  * A process group has become orphaned, mark it as such for signal
934  * delivery code.  If there are any stopped processes in the group,
935  * hang-up all process in that group.
936  */
937 static void
938 orphanpg(struct pgrp *pg)
939 {
940 	struct proc *p;
941 
942 	PGRP_LOCK_ASSERT(pg, MA_OWNED);
943 
944 	pg->pg_flags |= PGRP_ORPHANED;
945 
946 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
947 		PROC_LOCK(p);
948 		if (P_SHOULDSTOP(p) == P_STOPPED_SIG) {
949 			PROC_UNLOCK(p);
950 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
951 				PROC_LOCK(p);
952 				kern_psignal(p, SIGHUP);
953 				kern_psignal(p, SIGCONT);
954 				PROC_UNLOCK(p);
955 			}
956 			return;
957 		}
958 		PROC_UNLOCK(p);
959 	}
960 }
961 
962 void
963 sess_hold(struct session *s)
964 {
965 
966 	refcount_acquire(&s->s_count);
967 }
968 
969 void
970 sess_release(struct session *s)
971 {
972 
973 	if (refcount_release(&s->s_count)) {
974 		if (s->s_ttyp != NULL) {
975 			tty_lock(s->s_ttyp);
976 			tty_rel_sess(s->s_ttyp, s);
977 		}
978 		proc_id_clear(PROC_ID_SESSION, s->s_sid);
979 		mtx_destroy(&s->s_mtx);
980 		free(s, M_SESSION);
981 	}
982 }
983 
984 #ifdef DDB
985 
986 static void
987 db_print_pgrp_one(struct pgrp *pgrp, struct proc *p)
988 {
989 	db_printf(
990 	    "    pid %d at %p pr %d pgrp %p e %d jc %d\n",
991 	    p->p_pid, p, p->p_pptr == NULL ? -1 : p->p_pptr->p_pid,
992 	    p->p_pgrp, (p->p_treeflag & P_TREE_GRPEXITED) != 0,
993 	    p->p_pptr == NULL ? 0 : isjobproc(p->p_pptr, pgrp));
994 }
995 
996 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
997 {
998 	struct pgrp *pgrp;
999 	struct proc *p;
1000 	int i;
1001 
1002 	for (i = 0; i <= pgrphash; i++) {
1003 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
1004 			db_printf("indx %d\n", i);
1005 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
1006 				db_printf(
1007 			"  pgrp %p, pgid %d, sess %p, sesscnt %d, mem %p\n",
1008 				    pgrp, (int)pgrp->pg_id, pgrp->pg_session,
1009 				    pgrp->pg_session->s_count,
1010 				    LIST_FIRST(&pgrp->pg_members));
1011 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist)
1012 					db_print_pgrp_one(pgrp, p);
1013 			}
1014 		}
1015 	}
1016 }
1017 #endif /* DDB */
1018 
1019 /*
1020  * Calculate the kinfo_proc members which contain process-wide
1021  * informations.
1022  * Must be called with the target process locked.
1023  */
1024 static void
1025 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
1026 {
1027 	struct thread *td;
1028 
1029 	PROC_LOCK_ASSERT(p, MA_OWNED);
1030 
1031 	kp->ki_estcpu = 0;
1032 	kp->ki_pctcpu = 0;
1033 	FOREACH_THREAD_IN_PROC(p, td) {
1034 		thread_lock(td);
1035 		kp->ki_pctcpu += sched_pctcpu(td);
1036 		kp->ki_estcpu += sched_estcpu(td);
1037 		thread_unlock(td);
1038 	}
1039 }
1040 
1041 /*
1042  * Fill in any information that is common to all threads in the process.
1043  * Must be called with the target process locked.
1044  */
1045 static void
1046 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
1047 {
1048 	struct thread *td0;
1049 	struct ucred *cred;
1050 	struct sigacts *ps;
1051 	struct timeval boottime;
1052 
1053 	PROC_LOCK_ASSERT(p, MA_OWNED);
1054 
1055 	kp->ki_structsize = sizeof(*kp);
1056 	kp->ki_paddr = p;
1057 	kp->ki_addr =/* p->p_addr; */0; /* XXX */
1058 	kp->ki_args = p->p_args;
1059 	kp->ki_textvp = p->p_textvp;
1060 #ifdef KTRACE
1061 	kp->ki_tracep = p->p_tracevp;
1062 	kp->ki_traceflag = p->p_traceflag;
1063 #endif
1064 	kp->ki_fd = p->p_fd;
1065 	kp->ki_pd = p->p_pd;
1066 	kp->ki_vmspace = p->p_vmspace;
1067 	kp->ki_flag = p->p_flag;
1068 	kp->ki_flag2 = p->p_flag2;
1069 	cred = p->p_ucred;
1070 	if (cred) {
1071 		kp->ki_uid = cred->cr_uid;
1072 		kp->ki_ruid = cred->cr_ruid;
1073 		kp->ki_svuid = cred->cr_svuid;
1074 		kp->ki_cr_flags = 0;
1075 		if (cred->cr_flags & CRED_FLAG_CAPMODE)
1076 			kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
1077 		/* XXX bde doesn't like KI_NGROUPS */
1078 		if (cred->cr_ngroups > KI_NGROUPS) {
1079 			kp->ki_ngroups = KI_NGROUPS;
1080 			kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
1081 		} else
1082 			kp->ki_ngroups = cred->cr_ngroups;
1083 		bcopy(cred->cr_groups, kp->ki_groups,
1084 		    kp->ki_ngroups * sizeof(gid_t));
1085 		kp->ki_rgid = cred->cr_rgid;
1086 		kp->ki_svgid = cred->cr_svgid;
1087 		/* If jailed(cred), emulate the old P_JAILED flag. */
1088 		if (jailed(cred)) {
1089 			kp->ki_flag |= P_JAILED;
1090 			/* If inside the jail, use 0 as a jail ID. */
1091 			if (cred->cr_prison != curthread->td_ucred->cr_prison)
1092 				kp->ki_jid = cred->cr_prison->pr_id;
1093 		}
1094 		strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
1095 		    sizeof(kp->ki_loginclass));
1096 	}
1097 	ps = p->p_sigacts;
1098 	if (ps) {
1099 		mtx_lock(&ps->ps_mtx);
1100 		kp->ki_sigignore = ps->ps_sigignore;
1101 		kp->ki_sigcatch = ps->ps_sigcatch;
1102 		mtx_unlock(&ps->ps_mtx);
1103 	}
1104 	if (p->p_state != PRS_NEW &&
1105 	    p->p_state != PRS_ZOMBIE &&
1106 	    p->p_vmspace != NULL) {
1107 		struct vmspace *vm = p->p_vmspace;
1108 
1109 		kp->ki_size = vm->vm_map.size;
1110 		kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
1111 		FOREACH_THREAD_IN_PROC(p, td0) {
1112 			if (!TD_IS_SWAPPED(td0))
1113 				kp->ki_rssize += td0->td_kstack_pages;
1114 		}
1115 		kp->ki_swrss = vm->vm_swrss;
1116 		kp->ki_tsize = vm->vm_tsize;
1117 		kp->ki_dsize = vm->vm_dsize;
1118 		kp->ki_ssize = vm->vm_ssize;
1119 	} else if (p->p_state == PRS_ZOMBIE)
1120 		kp->ki_stat = SZOMB;
1121 	if (kp->ki_flag & P_INMEM)
1122 		kp->ki_sflag = PS_INMEM;
1123 	else
1124 		kp->ki_sflag = 0;
1125 	/* Calculate legacy swtime as seconds since 'swtick'. */
1126 	kp->ki_swtime = (ticks - p->p_swtick) / hz;
1127 	kp->ki_pid = p->p_pid;
1128 	kp->ki_nice = p->p_nice;
1129 	kp->ki_fibnum = p->p_fibnum;
1130 	kp->ki_start = p->p_stats->p_start;
1131 	getboottime(&boottime);
1132 	timevaladd(&kp->ki_start, &boottime);
1133 	PROC_STATLOCK(p);
1134 	rufetch(p, &kp->ki_rusage);
1135 	kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
1136 	calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
1137 	PROC_STATUNLOCK(p);
1138 	calccru(p, &kp->ki_childutime, &kp->ki_childstime);
1139 	/* Some callers want child times in a single value. */
1140 	kp->ki_childtime = kp->ki_childstime;
1141 	timevaladd(&kp->ki_childtime, &kp->ki_childutime);
1142 
1143 	FOREACH_THREAD_IN_PROC(p, td0)
1144 		kp->ki_cow += td0->td_cow;
1145 
1146 	if (p->p_comm[0] != '\0')
1147 		strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
1148 	if (p->p_sysent && p->p_sysent->sv_name != NULL &&
1149 	    p->p_sysent->sv_name[0] != '\0')
1150 		strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
1151 	kp->ki_siglist = p->p_siglist;
1152 	kp->ki_xstat = KW_EXITCODE(p->p_xexit, p->p_xsig);
1153 	kp->ki_acflag = p->p_acflag;
1154 	kp->ki_lock = p->p_lock;
1155 	if (p->p_pptr) {
1156 		kp->ki_ppid = p->p_oppid;
1157 		if (p->p_flag & P_TRACED)
1158 			kp->ki_tracer = p->p_pptr->p_pid;
1159 	}
1160 }
1161 
1162 /*
1163  * Fill job-related process information.
1164  */
1165 static void
1166 fill_kinfo_proc_pgrp(struct proc *p, struct kinfo_proc *kp)
1167 {
1168 	struct tty *tp;
1169 	struct session *sp;
1170 	struct pgrp *pgrp;
1171 
1172 	sx_assert(&proctree_lock, SA_LOCKED);
1173 	PROC_LOCK_ASSERT(p, MA_OWNED);
1174 
1175 	pgrp = p->p_pgrp;
1176 	if (pgrp == NULL)
1177 		return;
1178 
1179 	kp->ki_pgid = pgrp->pg_id;
1180 	kp->ki_jobc = pgrp_calc_jobc(pgrp);
1181 
1182 	sp = pgrp->pg_session;
1183 	tp = NULL;
1184 
1185 	if (sp != NULL) {
1186 		kp->ki_sid = sp->s_sid;
1187 		SESS_LOCK(sp);
1188 		strlcpy(kp->ki_login, sp->s_login, sizeof(kp->ki_login));
1189 		if (sp->s_ttyvp)
1190 			kp->ki_kiflag |= KI_CTTY;
1191 		if (SESS_LEADER(p))
1192 			kp->ki_kiflag |= KI_SLEADER;
1193 		tp = sp->s_ttyp;
1194 		SESS_UNLOCK(sp);
1195 	}
1196 
1197 	if ((p->p_flag & P_CONTROLT) && tp != NULL) {
1198 		kp->ki_tdev = tty_udev(tp);
1199 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1200 		kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
1201 		if (tp->t_session)
1202 			kp->ki_tsid = tp->t_session->s_sid;
1203 	} else {
1204 		kp->ki_tdev = NODEV;
1205 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1206 	}
1207 }
1208 
1209 /*
1210  * Fill in information that is thread specific.  Must be called with
1211  * target process locked.  If 'preferthread' is set, overwrite certain
1212  * process-related fields that are maintained for both threads and
1213  * processes.
1214  */
1215 static void
1216 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
1217 {
1218 	struct proc *p;
1219 
1220 	p = td->td_proc;
1221 	kp->ki_tdaddr = td;
1222 	PROC_LOCK_ASSERT(p, MA_OWNED);
1223 
1224 	if (preferthread)
1225 		PROC_STATLOCK(p);
1226 	thread_lock(td);
1227 	if (td->td_wmesg != NULL)
1228 		strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
1229 	else
1230 		bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
1231 	if (strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname)) >=
1232 	    sizeof(kp->ki_tdname)) {
1233 		strlcpy(kp->ki_moretdname,
1234 		    td->td_name + sizeof(kp->ki_tdname) - 1,
1235 		    sizeof(kp->ki_moretdname));
1236 	} else {
1237 		bzero(kp->ki_moretdname, sizeof(kp->ki_moretdname));
1238 	}
1239 	if (TD_ON_LOCK(td)) {
1240 		kp->ki_kiflag |= KI_LOCKBLOCK;
1241 		strlcpy(kp->ki_lockname, td->td_lockname,
1242 		    sizeof(kp->ki_lockname));
1243 	} else {
1244 		kp->ki_kiflag &= ~KI_LOCKBLOCK;
1245 		bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
1246 	}
1247 
1248 	if (p->p_state == PRS_NORMAL) { /* approximate. */
1249 		if (TD_ON_RUNQ(td) ||
1250 		    TD_CAN_RUN(td) ||
1251 		    TD_IS_RUNNING(td)) {
1252 			kp->ki_stat = SRUN;
1253 		} else if (P_SHOULDSTOP(p)) {
1254 			kp->ki_stat = SSTOP;
1255 		} else if (TD_IS_SLEEPING(td)) {
1256 			kp->ki_stat = SSLEEP;
1257 		} else if (TD_ON_LOCK(td)) {
1258 			kp->ki_stat = SLOCK;
1259 		} else {
1260 			kp->ki_stat = SWAIT;
1261 		}
1262 	} else if (p->p_state == PRS_ZOMBIE) {
1263 		kp->ki_stat = SZOMB;
1264 	} else {
1265 		kp->ki_stat = SIDL;
1266 	}
1267 
1268 	/* Things in the thread */
1269 	kp->ki_wchan = td->td_wchan;
1270 	kp->ki_pri.pri_level = td->td_priority;
1271 	kp->ki_pri.pri_native = td->td_base_pri;
1272 
1273 	/*
1274 	 * Note: legacy fields; clamp at the old NOCPU value and/or
1275 	 * the maximum u_char CPU value.
1276 	 */
1277 	if (td->td_lastcpu == NOCPU)
1278 		kp->ki_lastcpu_old = NOCPU_OLD;
1279 	else if (td->td_lastcpu > MAXCPU_OLD)
1280 		kp->ki_lastcpu_old = MAXCPU_OLD;
1281 	else
1282 		kp->ki_lastcpu_old = td->td_lastcpu;
1283 
1284 	if (td->td_oncpu == NOCPU)
1285 		kp->ki_oncpu_old = NOCPU_OLD;
1286 	else if (td->td_oncpu > MAXCPU_OLD)
1287 		kp->ki_oncpu_old = MAXCPU_OLD;
1288 	else
1289 		kp->ki_oncpu_old = td->td_oncpu;
1290 
1291 	kp->ki_lastcpu = td->td_lastcpu;
1292 	kp->ki_oncpu = td->td_oncpu;
1293 	kp->ki_tdflags = td->td_flags;
1294 	kp->ki_tid = td->td_tid;
1295 	kp->ki_numthreads = p->p_numthreads;
1296 	kp->ki_pcb = td->td_pcb;
1297 	kp->ki_kstack = (void *)td->td_kstack;
1298 	kp->ki_slptime = (ticks - td->td_slptick) / hz;
1299 	kp->ki_pri.pri_class = td->td_pri_class;
1300 	kp->ki_pri.pri_user = td->td_user_pri;
1301 
1302 	if (preferthread) {
1303 		rufetchtd(td, &kp->ki_rusage);
1304 		kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
1305 		kp->ki_pctcpu = sched_pctcpu(td);
1306 		kp->ki_estcpu = sched_estcpu(td);
1307 		kp->ki_cow = td->td_cow;
1308 	}
1309 
1310 	/* We can't get this anymore but ps etc never used it anyway. */
1311 	kp->ki_rqindex = 0;
1312 
1313 	if (preferthread)
1314 		kp->ki_siglist = td->td_siglist;
1315 	kp->ki_sigmask = td->td_sigmask;
1316 	thread_unlock(td);
1317 	if (preferthread)
1318 		PROC_STATUNLOCK(p);
1319 }
1320 
1321 /*
1322  * Fill in a kinfo_proc structure for the specified process.
1323  * Must be called with the target process locked.
1324  */
1325 void
1326 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1327 {
1328 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1329 
1330 	bzero(kp, sizeof(*kp));
1331 
1332 	fill_kinfo_proc_pgrp(p,kp);
1333 	fill_kinfo_proc_only(p, kp);
1334 	fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1335 	fill_kinfo_aggregate(p, kp);
1336 }
1337 
1338 struct pstats *
1339 pstats_alloc(void)
1340 {
1341 
1342 	return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1343 }
1344 
1345 /*
1346  * Copy parts of p_stats; zero the rest of p_stats (statistics).
1347  */
1348 void
1349 pstats_fork(struct pstats *src, struct pstats *dst)
1350 {
1351 
1352 	bzero(&dst->pstat_startzero,
1353 	    __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1354 	bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1355 	    __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1356 }
1357 
1358 void
1359 pstats_free(struct pstats *ps)
1360 {
1361 
1362 	free(ps, M_SUBPROC);
1363 }
1364 
1365 #ifdef COMPAT_FREEBSD32
1366 
1367 /*
1368  * This function is typically used to copy out the kernel address, so
1369  * it can be replaced by assignment of zero.
1370  */
1371 static inline uint32_t
1372 ptr32_trim(const void *ptr)
1373 {
1374 	uintptr_t uptr;
1375 
1376 	uptr = (uintptr_t)ptr;
1377 	return ((uptr > UINT_MAX) ? 0 : uptr);
1378 }
1379 
1380 #define PTRTRIM_CP(src,dst,fld) \
1381 	do { (dst).fld = ptr32_trim((src).fld); } while (0)
1382 
1383 static void
1384 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1385 {
1386 	int i;
1387 
1388 	bzero(ki32, sizeof(struct kinfo_proc32));
1389 	ki32->ki_structsize = sizeof(struct kinfo_proc32);
1390 	CP(*ki, *ki32, ki_layout);
1391 	PTRTRIM_CP(*ki, *ki32, ki_args);
1392 	PTRTRIM_CP(*ki, *ki32, ki_paddr);
1393 	PTRTRIM_CP(*ki, *ki32, ki_addr);
1394 	PTRTRIM_CP(*ki, *ki32, ki_tracep);
1395 	PTRTRIM_CP(*ki, *ki32, ki_textvp);
1396 	PTRTRIM_CP(*ki, *ki32, ki_fd);
1397 	PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1398 	PTRTRIM_CP(*ki, *ki32, ki_wchan);
1399 	CP(*ki, *ki32, ki_pid);
1400 	CP(*ki, *ki32, ki_ppid);
1401 	CP(*ki, *ki32, ki_pgid);
1402 	CP(*ki, *ki32, ki_tpgid);
1403 	CP(*ki, *ki32, ki_sid);
1404 	CP(*ki, *ki32, ki_tsid);
1405 	CP(*ki, *ki32, ki_jobc);
1406 	CP(*ki, *ki32, ki_tdev);
1407 	CP(*ki, *ki32, ki_tdev_freebsd11);
1408 	CP(*ki, *ki32, ki_siglist);
1409 	CP(*ki, *ki32, ki_sigmask);
1410 	CP(*ki, *ki32, ki_sigignore);
1411 	CP(*ki, *ki32, ki_sigcatch);
1412 	CP(*ki, *ki32, ki_uid);
1413 	CP(*ki, *ki32, ki_ruid);
1414 	CP(*ki, *ki32, ki_svuid);
1415 	CP(*ki, *ki32, ki_rgid);
1416 	CP(*ki, *ki32, ki_svgid);
1417 	CP(*ki, *ki32, ki_ngroups);
1418 	for (i = 0; i < KI_NGROUPS; i++)
1419 		CP(*ki, *ki32, ki_groups[i]);
1420 	CP(*ki, *ki32, ki_size);
1421 	CP(*ki, *ki32, ki_rssize);
1422 	CP(*ki, *ki32, ki_swrss);
1423 	CP(*ki, *ki32, ki_tsize);
1424 	CP(*ki, *ki32, ki_dsize);
1425 	CP(*ki, *ki32, ki_ssize);
1426 	CP(*ki, *ki32, ki_xstat);
1427 	CP(*ki, *ki32, ki_acflag);
1428 	CP(*ki, *ki32, ki_pctcpu);
1429 	CP(*ki, *ki32, ki_estcpu);
1430 	CP(*ki, *ki32, ki_slptime);
1431 	CP(*ki, *ki32, ki_swtime);
1432 	CP(*ki, *ki32, ki_cow);
1433 	CP(*ki, *ki32, ki_runtime);
1434 	TV_CP(*ki, *ki32, ki_start);
1435 	TV_CP(*ki, *ki32, ki_childtime);
1436 	CP(*ki, *ki32, ki_flag);
1437 	CP(*ki, *ki32, ki_kiflag);
1438 	CP(*ki, *ki32, ki_traceflag);
1439 	CP(*ki, *ki32, ki_stat);
1440 	CP(*ki, *ki32, ki_nice);
1441 	CP(*ki, *ki32, ki_lock);
1442 	CP(*ki, *ki32, ki_rqindex);
1443 	CP(*ki, *ki32, ki_oncpu);
1444 	CP(*ki, *ki32, ki_lastcpu);
1445 
1446 	/* XXX TODO: wrap cpu value as appropriate */
1447 	CP(*ki, *ki32, ki_oncpu_old);
1448 	CP(*ki, *ki32, ki_lastcpu_old);
1449 
1450 	bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1451 	bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1452 	bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1453 	bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1454 	bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1455 	bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1456 	bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1457 	bcopy(ki->ki_moretdname, ki32->ki_moretdname, MAXCOMLEN - TDNAMLEN + 1);
1458 	CP(*ki, *ki32, ki_tracer);
1459 	CP(*ki, *ki32, ki_flag2);
1460 	CP(*ki, *ki32, ki_fibnum);
1461 	CP(*ki, *ki32, ki_cr_flags);
1462 	CP(*ki, *ki32, ki_jid);
1463 	CP(*ki, *ki32, ki_numthreads);
1464 	CP(*ki, *ki32, ki_tid);
1465 	CP(*ki, *ki32, ki_pri);
1466 	freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1467 	freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1468 	PTRTRIM_CP(*ki, *ki32, ki_pcb);
1469 	PTRTRIM_CP(*ki, *ki32, ki_kstack);
1470 	PTRTRIM_CP(*ki, *ki32, ki_udata);
1471 	PTRTRIM_CP(*ki, *ki32, ki_tdaddr);
1472 	CP(*ki, *ki32, ki_sflag);
1473 	CP(*ki, *ki32, ki_tdflags);
1474 }
1475 #endif
1476 
1477 static ssize_t
1478 kern_proc_out_size(struct proc *p, int flags)
1479 {
1480 	ssize_t size = 0;
1481 
1482 	PROC_LOCK_ASSERT(p, MA_OWNED);
1483 
1484 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1485 #ifdef COMPAT_FREEBSD32
1486 		if ((flags & KERN_PROC_MASK32) != 0) {
1487 			size += sizeof(struct kinfo_proc32);
1488 		} else
1489 #endif
1490 			size += sizeof(struct kinfo_proc);
1491 	} else {
1492 #ifdef COMPAT_FREEBSD32
1493 		if ((flags & KERN_PROC_MASK32) != 0)
1494 			size += sizeof(struct kinfo_proc32) * p->p_numthreads;
1495 		else
1496 #endif
1497 			size += sizeof(struct kinfo_proc) * p->p_numthreads;
1498 	}
1499 	PROC_UNLOCK(p);
1500 	return (size);
1501 }
1502 
1503 int
1504 kern_proc_out(struct proc *p, struct sbuf *sb, int flags)
1505 {
1506 	struct thread *td;
1507 	struct kinfo_proc ki;
1508 #ifdef COMPAT_FREEBSD32
1509 	struct kinfo_proc32 ki32;
1510 #endif
1511 	int error;
1512 
1513 	PROC_LOCK_ASSERT(p, MA_OWNED);
1514 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1515 
1516 	error = 0;
1517 	fill_kinfo_proc(p, &ki);
1518 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1519 #ifdef COMPAT_FREEBSD32
1520 		if ((flags & KERN_PROC_MASK32) != 0) {
1521 			freebsd32_kinfo_proc_out(&ki, &ki32);
1522 			if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1523 				error = ENOMEM;
1524 		} else
1525 #endif
1526 			if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1527 				error = ENOMEM;
1528 	} else {
1529 		FOREACH_THREAD_IN_PROC(p, td) {
1530 			fill_kinfo_thread(td, &ki, 1);
1531 #ifdef COMPAT_FREEBSD32
1532 			if ((flags & KERN_PROC_MASK32) != 0) {
1533 				freebsd32_kinfo_proc_out(&ki, &ki32);
1534 				if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1535 					error = ENOMEM;
1536 			} else
1537 #endif
1538 				if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1539 					error = ENOMEM;
1540 			if (error != 0)
1541 				break;
1542 		}
1543 	}
1544 	PROC_UNLOCK(p);
1545 	return (error);
1546 }
1547 
1548 static int
1549 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
1550 {
1551 	struct sbuf sb;
1552 	struct kinfo_proc ki;
1553 	int error, error2;
1554 
1555 	if (req->oldptr == NULL)
1556 		return (SYSCTL_OUT(req, 0, kern_proc_out_size(p, flags)));
1557 
1558 	sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req);
1559 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1560 	error = kern_proc_out(p, &sb, flags);
1561 	error2 = sbuf_finish(&sb);
1562 	sbuf_delete(&sb);
1563 	if (error != 0)
1564 		return (error);
1565 	else if (error2 != 0)
1566 		return (error2);
1567 	return (0);
1568 }
1569 
1570 int
1571 proc_iterate(int (*cb)(struct proc *, void *), void *cbarg)
1572 {
1573 	struct proc *p;
1574 	int error, i, j;
1575 
1576 	for (i = 0; i < pidhashlock + 1; i++) {
1577 		sx_slock(&proctree_lock);
1578 		sx_slock(&pidhashtbl_lock[i]);
1579 		for (j = i; j <= pidhash; j += pidhashlock + 1) {
1580 			LIST_FOREACH(p, &pidhashtbl[j], p_hash) {
1581 				if (p->p_state == PRS_NEW)
1582 					continue;
1583 				error = cb(p, cbarg);
1584 				PROC_LOCK_ASSERT(p, MA_NOTOWNED);
1585 				if (error != 0) {
1586 					sx_sunlock(&pidhashtbl_lock[i]);
1587 					sx_sunlock(&proctree_lock);
1588 					return (error);
1589 				}
1590 			}
1591 		}
1592 		sx_sunlock(&pidhashtbl_lock[i]);
1593 		sx_sunlock(&proctree_lock);
1594 	}
1595 	return (0);
1596 }
1597 
1598 struct kern_proc_out_args {
1599 	struct sysctl_req *req;
1600 	int flags;
1601 	int oid_number;
1602 	int *name;
1603 };
1604 
1605 static int
1606 sysctl_kern_proc_iterate(struct proc *p, void *origarg)
1607 {
1608 	struct kern_proc_out_args *arg = origarg;
1609 	int *name = arg->name;
1610 	int oid_number = arg->oid_number;
1611 	int flags = arg->flags;
1612 	struct sysctl_req *req = arg->req;
1613 	int error = 0;
1614 
1615 	PROC_LOCK(p);
1616 
1617 	KASSERT(p->p_ucred != NULL,
1618 	    ("process credential is NULL for non-NEW proc"));
1619 	/*
1620 	 * Show a user only appropriate processes.
1621 	 */
1622 	if (p_cansee(curthread, p))
1623 		goto skip;
1624 	/*
1625 	 * TODO - make more efficient (see notes below).
1626 	 * do by session.
1627 	 */
1628 	switch (oid_number) {
1629 	case KERN_PROC_GID:
1630 		if (p->p_ucred->cr_gid != (gid_t)name[0])
1631 			goto skip;
1632 		break;
1633 
1634 	case KERN_PROC_PGRP:
1635 		/* could do this by traversing pgrp */
1636 		if (p->p_pgrp == NULL ||
1637 		    p->p_pgrp->pg_id != (pid_t)name[0])
1638 			goto skip;
1639 		break;
1640 
1641 	case KERN_PROC_RGID:
1642 		if (p->p_ucred->cr_rgid != (gid_t)name[0])
1643 			goto skip;
1644 		break;
1645 
1646 	case KERN_PROC_SESSION:
1647 		if (p->p_session == NULL ||
1648 		    p->p_session->s_sid != (pid_t)name[0])
1649 			goto skip;
1650 		break;
1651 
1652 	case KERN_PROC_TTY:
1653 		if ((p->p_flag & P_CONTROLT) == 0 ||
1654 		    p->p_session == NULL)
1655 			goto skip;
1656 		/* XXX proctree_lock */
1657 		SESS_LOCK(p->p_session);
1658 		if (p->p_session->s_ttyp == NULL ||
1659 		    tty_udev(p->p_session->s_ttyp) !=
1660 		    (dev_t)name[0]) {
1661 			SESS_UNLOCK(p->p_session);
1662 			goto skip;
1663 		}
1664 		SESS_UNLOCK(p->p_session);
1665 		break;
1666 
1667 	case KERN_PROC_UID:
1668 		if (p->p_ucred->cr_uid != (uid_t)name[0])
1669 			goto skip;
1670 		break;
1671 
1672 	case KERN_PROC_RUID:
1673 		if (p->p_ucred->cr_ruid != (uid_t)name[0])
1674 			goto skip;
1675 		break;
1676 
1677 	case KERN_PROC_PROC:
1678 		break;
1679 
1680 	default:
1681 		break;
1682 	}
1683 	error = sysctl_out_proc(p, req, flags);
1684 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
1685 	return (error);
1686 skip:
1687 	PROC_UNLOCK(p);
1688 	return (0);
1689 }
1690 
1691 static int
1692 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1693 {
1694 	struct kern_proc_out_args iterarg;
1695 	int *name = (int *)arg1;
1696 	u_int namelen = arg2;
1697 	struct proc *p;
1698 	int flags, oid_number;
1699 	int error = 0;
1700 
1701 	oid_number = oidp->oid_number;
1702 	if (oid_number != KERN_PROC_ALL &&
1703 	    (oid_number & KERN_PROC_INC_THREAD) == 0)
1704 		flags = KERN_PROC_NOTHREADS;
1705 	else {
1706 		flags = 0;
1707 		oid_number &= ~KERN_PROC_INC_THREAD;
1708 	}
1709 #ifdef COMPAT_FREEBSD32
1710 	if (req->flags & SCTL_MASK32)
1711 		flags |= KERN_PROC_MASK32;
1712 #endif
1713 	if (oid_number == KERN_PROC_PID) {
1714 		if (namelen != 1)
1715 			return (EINVAL);
1716 		error = sysctl_wire_old_buffer(req, 0);
1717 		if (error)
1718 			return (error);
1719 		sx_slock(&proctree_lock);
1720 		error = pget((pid_t)name[0], PGET_CANSEE, &p);
1721 		if (error == 0)
1722 			error = sysctl_out_proc(p, req, flags);
1723 		sx_sunlock(&proctree_lock);
1724 		return (error);
1725 	}
1726 
1727 	switch (oid_number) {
1728 	case KERN_PROC_ALL:
1729 		if (namelen != 0)
1730 			return (EINVAL);
1731 		break;
1732 	case KERN_PROC_PROC:
1733 		if (namelen != 0 && namelen != 1)
1734 			return (EINVAL);
1735 		break;
1736 	default:
1737 		if (namelen != 1)
1738 			return (EINVAL);
1739 		break;
1740 	}
1741 
1742 	if (req->oldptr == NULL) {
1743 		/* overestimate by 5 procs */
1744 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1745 		if (error)
1746 			return (error);
1747 	} else {
1748 		error = sysctl_wire_old_buffer(req, 0);
1749 		if (error != 0)
1750 			return (error);
1751 	}
1752 	iterarg.flags = flags;
1753 	iterarg.oid_number = oid_number;
1754 	iterarg.req = req;
1755 	iterarg.name = name;
1756 	error = proc_iterate(sysctl_kern_proc_iterate, &iterarg);
1757 	return (error);
1758 }
1759 
1760 struct pargs *
1761 pargs_alloc(int len)
1762 {
1763 	struct pargs *pa;
1764 
1765 	pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1766 		M_WAITOK);
1767 	refcount_init(&pa->ar_ref, 1);
1768 	pa->ar_length = len;
1769 	return (pa);
1770 }
1771 
1772 static void
1773 pargs_free(struct pargs *pa)
1774 {
1775 
1776 	free(pa, M_PARGS);
1777 }
1778 
1779 void
1780 pargs_hold(struct pargs *pa)
1781 {
1782 
1783 	if (pa == NULL)
1784 		return;
1785 	refcount_acquire(&pa->ar_ref);
1786 }
1787 
1788 void
1789 pargs_drop(struct pargs *pa)
1790 {
1791 
1792 	if (pa == NULL)
1793 		return;
1794 	if (refcount_release(&pa->ar_ref))
1795 		pargs_free(pa);
1796 }
1797 
1798 static int
1799 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1800     size_t len)
1801 {
1802 	ssize_t n;
1803 
1804 	/*
1805 	 * This may return a short read if the string is shorter than the chunk
1806 	 * and is aligned at the end of the page, and the following page is not
1807 	 * mapped.
1808 	 */
1809 	n = proc_readmem(td, p, (vm_offset_t)sptr, buf, len);
1810 	if (n <= 0)
1811 		return (ENOMEM);
1812 	return (0);
1813 }
1814 
1815 #define PROC_AUXV_MAX	256	/* Safety limit on auxv size. */
1816 
1817 enum proc_vector_type {
1818 	PROC_ARG,
1819 	PROC_ENV,
1820 	PROC_AUX,
1821 };
1822 
1823 #ifdef COMPAT_FREEBSD32
1824 static int
1825 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1826     size_t *vsizep, enum proc_vector_type type)
1827 {
1828 	struct freebsd32_ps_strings pss;
1829 	Elf32_Auxinfo aux;
1830 	vm_offset_t vptr, ptr;
1831 	uint32_t *proc_vector32;
1832 	char **proc_vector;
1833 	size_t vsize, size;
1834 	int i, error;
1835 
1836 	error = 0;
1837 	if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss,
1838 	    sizeof(pss)) != sizeof(pss))
1839 		return (ENOMEM);
1840 	switch (type) {
1841 	case PROC_ARG:
1842 		vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1843 		vsize = pss.ps_nargvstr;
1844 		if (vsize > ARG_MAX)
1845 			return (ENOEXEC);
1846 		size = vsize * sizeof(int32_t);
1847 		break;
1848 	case PROC_ENV:
1849 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1850 		vsize = pss.ps_nenvstr;
1851 		if (vsize > ARG_MAX)
1852 			return (ENOEXEC);
1853 		size = vsize * sizeof(int32_t);
1854 		break;
1855 	case PROC_AUX:
1856 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1857 		    (pss.ps_nenvstr + 1) * sizeof(int32_t);
1858 		if (vptr % 4 != 0)
1859 			return (ENOEXEC);
1860 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1861 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1862 			    sizeof(aux))
1863 				return (ENOMEM);
1864 			if (aux.a_type == AT_NULL)
1865 				break;
1866 			ptr += sizeof(aux);
1867 		}
1868 		if (aux.a_type != AT_NULL)
1869 			return (ENOEXEC);
1870 		vsize = i + 1;
1871 		size = vsize * sizeof(aux);
1872 		break;
1873 	default:
1874 		KASSERT(0, ("Wrong proc vector type: %d", type));
1875 		return (EINVAL);
1876 	}
1877 	proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1878 	if (proc_readmem(td, p, vptr, proc_vector32, size) != size) {
1879 		error = ENOMEM;
1880 		goto done;
1881 	}
1882 	if (type == PROC_AUX) {
1883 		*proc_vectorp = (char **)proc_vector32;
1884 		*vsizep = vsize;
1885 		return (0);
1886 	}
1887 	proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1888 	for (i = 0; i < (int)vsize; i++)
1889 		proc_vector[i] = PTRIN(proc_vector32[i]);
1890 	*proc_vectorp = proc_vector;
1891 	*vsizep = vsize;
1892 done:
1893 	free(proc_vector32, M_TEMP);
1894 	return (error);
1895 }
1896 #endif
1897 
1898 static int
1899 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1900     size_t *vsizep, enum proc_vector_type type)
1901 {
1902 	struct ps_strings pss;
1903 	Elf_Auxinfo aux;
1904 	vm_offset_t vptr, ptr;
1905 	char **proc_vector;
1906 	size_t vsize, size;
1907 	int i;
1908 
1909 #ifdef COMPAT_FREEBSD32
1910 	if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1911 		return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1912 #endif
1913 	if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss,
1914 	    sizeof(pss)) != sizeof(pss))
1915 		return (ENOMEM);
1916 	switch (type) {
1917 	case PROC_ARG:
1918 		vptr = (vm_offset_t)pss.ps_argvstr;
1919 		vsize = pss.ps_nargvstr;
1920 		if (vsize > ARG_MAX)
1921 			return (ENOEXEC);
1922 		size = vsize * sizeof(char *);
1923 		break;
1924 	case PROC_ENV:
1925 		vptr = (vm_offset_t)pss.ps_envstr;
1926 		vsize = pss.ps_nenvstr;
1927 		if (vsize > ARG_MAX)
1928 			return (ENOEXEC);
1929 		size = vsize * sizeof(char *);
1930 		break;
1931 	case PROC_AUX:
1932 		/*
1933 		 * The aux array is just above env array on the stack. Check
1934 		 * that the address is naturally aligned.
1935 		 */
1936 		vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1937 		    * sizeof(char *);
1938 #if __ELF_WORD_SIZE == 64
1939 		if (vptr % sizeof(uint64_t) != 0)
1940 #else
1941 		if (vptr % sizeof(uint32_t) != 0)
1942 #endif
1943 			return (ENOEXEC);
1944 		/*
1945 		 * We count the array size reading the aux vectors from the
1946 		 * stack until AT_NULL vector is returned.  So (to keep the code
1947 		 * simple) we read the process stack twice: the first time here
1948 		 * to find the size and the second time when copying the vectors
1949 		 * to the allocated proc_vector.
1950 		 */
1951 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1952 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1953 			    sizeof(aux))
1954 				return (ENOMEM);
1955 			if (aux.a_type == AT_NULL)
1956 				break;
1957 			ptr += sizeof(aux);
1958 		}
1959 		/*
1960 		 * If the PROC_AUXV_MAX entries are iterated over, and we have
1961 		 * not reached AT_NULL, it is most likely we are reading wrong
1962 		 * data: either the process doesn't have auxv array or data has
1963 		 * been modified. Return the error in this case.
1964 		 */
1965 		if (aux.a_type != AT_NULL)
1966 			return (ENOEXEC);
1967 		vsize = i + 1;
1968 		size = vsize * sizeof(aux);
1969 		break;
1970 	default:
1971 		KASSERT(0, ("Wrong proc vector type: %d", type));
1972 		return (EINVAL); /* In case we are built without INVARIANTS. */
1973 	}
1974 	proc_vector = malloc(size, M_TEMP, M_WAITOK);
1975 	if (proc_readmem(td, p, vptr, proc_vector, size) != size) {
1976 		free(proc_vector, M_TEMP);
1977 		return (ENOMEM);
1978 	}
1979 	*proc_vectorp = proc_vector;
1980 	*vsizep = vsize;
1981 
1982 	return (0);
1983 }
1984 
1985 #define GET_PS_STRINGS_CHUNK_SZ	256	/* Chunk size (bytes) for ps_strings operations. */
1986 
1987 static int
1988 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
1989     enum proc_vector_type type)
1990 {
1991 	size_t done, len, nchr, vsize;
1992 	int error, i;
1993 	char **proc_vector, *sptr;
1994 	char pss_string[GET_PS_STRINGS_CHUNK_SZ];
1995 
1996 	PROC_ASSERT_HELD(p);
1997 
1998 	/*
1999 	 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
2000 	 */
2001 	nchr = 2 * (PATH_MAX + ARG_MAX);
2002 
2003 	error = get_proc_vector(td, p, &proc_vector, &vsize, type);
2004 	if (error != 0)
2005 		return (error);
2006 	for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
2007 		/*
2008 		 * The program may have scribbled into its argv array, e.g. to
2009 		 * remove some arguments.  If that has happened, break out
2010 		 * before trying to read from NULL.
2011 		 */
2012 		if (proc_vector[i] == NULL)
2013 			break;
2014 		for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
2015 			error = proc_read_string(td, p, sptr, pss_string,
2016 			    sizeof(pss_string));
2017 			if (error != 0)
2018 				goto done;
2019 			len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
2020 			if (done + len >= nchr)
2021 				len = nchr - done - 1;
2022 			sbuf_bcat(sb, pss_string, len);
2023 			if (len != GET_PS_STRINGS_CHUNK_SZ)
2024 				break;
2025 			done += GET_PS_STRINGS_CHUNK_SZ;
2026 		}
2027 		sbuf_bcat(sb, "", 1);
2028 		done += len + 1;
2029 	}
2030 done:
2031 	free(proc_vector, M_TEMP);
2032 	return (error);
2033 }
2034 
2035 int
2036 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
2037 {
2038 
2039 	return (get_ps_strings(curthread, p, sb, PROC_ARG));
2040 }
2041 
2042 int
2043 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
2044 {
2045 
2046 	return (get_ps_strings(curthread, p, sb, PROC_ENV));
2047 }
2048 
2049 int
2050 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb)
2051 {
2052 	size_t vsize, size;
2053 	char **auxv;
2054 	int error;
2055 
2056 	error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX);
2057 	if (error == 0) {
2058 #ifdef COMPAT_FREEBSD32
2059 		if (SV_PROC_FLAG(p, SV_ILP32) != 0)
2060 			size = vsize * sizeof(Elf32_Auxinfo);
2061 		else
2062 #endif
2063 			size = vsize * sizeof(Elf_Auxinfo);
2064 		if (sbuf_bcat(sb, auxv, size) != 0)
2065 			error = ENOMEM;
2066 		free(auxv, M_TEMP);
2067 	}
2068 	return (error);
2069 }
2070 
2071 /*
2072  * This sysctl allows a process to retrieve the argument list or process
2073  * title for another process without groping around in the address space
2074  * of the other process.  It also allow a process to set its own "process
2075  * title to a string of its own choice.
2076  */
2077 static int
2078 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
2079 {
2080 	int *name = (int *)arg1;
2081 	u_int namelen = arg2;
2082 	struct pargs *newpa, *pa;
2083 	struct proc *p;
2084 	struct sbuf sb;
2085 	int flags, error = 0, error2;
2086 	pid_t pid;
2087 
2088 	if (namelen != 1)
2089 		return (EINVAL);
2090 
2091 	p = curproc;
2092 	pid = (pid_t)name[0];
2093 	if (pid == -1) {
2094 		pid = p->p_pid;
2095 	}
2096 
2097 	/*
2098 	 * If the query is for this process and it is single-threaded, there
2099 	 * is nobody to modify pargs, thus we can just read.
2100 	 */
2101 	if (pid == p->p_pid && p->p_numthreads == 1 && req->newptr == NULL &&
2102 	    (pa = p->p_args) != NULL)
2103 		return (SYSCTL_OUT(req, pa->ar_args, pa->ar_length));
2104 
2105 	flags = PGET_CANSEE;
2106 	if (req->newptr != NULL)
2107 		flags |= PGET_ISCURRENT;
2108 	error = pget(pid, flags, &p);
2109 	if (error)
2110 		return (error);
2111 
2112 	pa = p->p_args;
2113 	if (pa != NULL) {
2114 		pargs_hold(pa);
2115 		PROC_UNLOCK(p);
2116 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
2117 		pargs_drop(pa);
2118 	} else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
2119 		_PHOLD(p);
2120 		PROC_UNLOCK(p);
2121 		sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2122 		sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2123 		error = proc_getargv(curthread, p, &sb);
2124 		error2 = sbuf_finish(&sb);
2125 		PRELE(p);
2126 		sbuf_delete(&sb);
2127 		if (error == 0 && error2 != 0)
2128 			error = error2;
2129 	} else {
2130 		PROC_UNLOCK(p);
2131 	}
2132 	if (error != 0 || req->newptr == NULL)
2133 		return (error);
2134 
2135 	if (req->newlen > ps_arg_cache_limit - sizeof(struct pargs))
2136 		return (ENOMEM);
2137 
2138 	if (req->newlen == 0) {
2139 		/*
2140 		 * Clear the argument pointer, so that we'll fetch arguments
2141 		 * with proc_getargv() until further notice.
2142 		 */
2143 		newpa = NULL;
2144 	} else {
2145 		newpa = pargs_alloc(req->newlen);
2146 		error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
2147 		if (error != 0) {
2148 			pargs_free(newpa);
2149 			return (error);
2150 		}
2151 	}
2152 	PROC_LOCK(p);
2153 	pa = p->p_args;
2154 	p->p_args = newpa;
2155 	PROC_UNLOCK(p);
2156 	pargs_drop(pa);
2157 	return (0);
2158 }
2159 
2160 /*
2161  * This sysctl allows a process to retrieve environment of another process.
2162  */
2163 static int
2164 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
2165 {
2166 	int *name = (int *)arg1;
2167 	u_int namelen = arg2;
2168 	struct proc *p;
2169 	struct sbuf sb;
2170 	int error, error2;
2171 
2172 	if (namelen != 1)
2173 		return (EINVAL);
2174 
2175 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2176 	if (error != 0)
2177 		return (error);
2178 	if ((p->p_flag & P_SYSTEM) != 0) {
2179 		PRELE(p);
2180 		return (0);
2181 	}
2182 
2183 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2184 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2185 	error = proc_getenvv(curthread, p, &sb);
2186 	error2 = sbuf_finish(&sb);
2187 	PRELE(p);
2188 	sbuf_delete(&sb);
2189 	return (error != 0 ? error : error2);
2190 }
2191 
2192 /*
2193  * This sysctl allows a process to retrieve ELF auxiliary vector of
2194  * another process.
2195  */
2196 static int
2197 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
2198 {
2199 	int *name = (int *)arg1;
2200 	u_int namelen = arg2;
2201 	struct proc *p;
2202 	struct sbuf sb;
2203 	int error, error2;
2204 
2205 	if (namelen != 1)
2206 		return (EINVAL);
2207 
2208 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2209 	if (error != 0)
2210 		return (error);
2211 	if ((p->p_flag & P_SYSTEM) != 0) {
2212 		PRELE(p);
2213 		return (0);
2214 	}
2215 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2216 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2217 	error = proc_getauxv(curthread, p, &sb);
2218 	error2 = sbuf_finish(&sb);
2219 	PRELE(p);
2220 	sbuf_delete(&sb);
2221 	return (error != 0 ? error : error2);
2222 }
2223 
2224 /*
2225  * This sysctl allows a process to retrieve the path of the executable for
2226  * itself or another process.
2227  */
2228 static int
2229 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
2230 {
2231 	pid_t *pidp = (pid_t *)arg1;
2232 	unsigned int arglen = arg2;
2233 	struct proc *p;
2234 	struct vnode *vp;
2235 	char *retbuf, *freebuf;
2236 	int error;
2237 
2238 	if (arglen != 1)
2239 		return (EINVAL);
2240 	if (*pidp == -1) {	/* -1 means this process */
2241 		p = req->td->td_proc;
2242 	} else {
2243 		error = pget(*pidp, PGET_CANSEE, &p);
2244 		if (error != 0)
2245 			return (error);
2246 	}
2247 
2248 	vp = p->p_textvp;
2249 	if (vp == NULL) {
2250 		if (*pidp != -1)
2251 			PROC_UNLOCK(p);
2252 		return (0);
2253 	}
2254 	vref(vp);
2255 	if (*pidp != -1)
2256 		PROC_UNLOCK(p);
2257 	error = vn_fullpath(vp, &retbuf, &freebuf);
2258 	vrele(vp);
2259 	if (error)
2260 		return (error);
2261 	error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
2262 	free(freebuf, M_TEMP);
2263 	return (error);
2264 }
2265 
2266 static int
2267 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
2268 {
2269 	struct proc *p;
2270 	char *sv_name;
2271 	int *name;
2272 	int namelen;
2273 	int error;
2274 
2275 	namelen = arg2;
2276 	if (namelen != 1)
2277 		return (EINVAL);
2278 
2279 	name = (int *)arg1;
2280 	error = pget((pid_t)name[0], PGET_CANSEE, &p);
2281 	if (error != 0)
2282 		return (error);
2283 	sv_name = p->p_sysent->sv_name;
2284 	PROC_UNLOCK(p);
2285 	return (sysctl_handle_string(oidp, sv_name, 0, req));
2286 }
2287 
2288 #ifdef KINFO_OVMENTRY_SIZE
2289 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
2290 #endif
2291 
2292 #ifdef COMPAT_FREEBSD7
2293 static int
2294 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
2295 {
2296 	vm_map_entry_t entry, tmp_entry;
2297 	unsigned int last_timestamp;
2298 	char *fullpath, *freepath;
2299 	struct kinfo_ovmentry *kve;
2300 	struct vattr va;
2301 	struct ucred *cred;
2302 	int error, *name;
2303 	struct vnode *vp;
2304 	struct proc *p;
2305 	vm_map_t map;
2306 	struct vmspace *vm;
2307 
2308 	name = (int *)arg1;
2309 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2310 	if (error != 0)
2311 		return (error);
2312 	vm = vmspace_acquire_ref(p);
2313 	if (vm == NULL) {
2314 		PRELE(p);
2315 		return (ESRCH);
2316 	}
2317 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2318 
2319 	map = &vm->vm_map;
2320 	vm_map_lock_read(map);
2321 	VM_MAP_ENTRY_FOREACH(entry, map) {
2322 		vm_object_t obj, tobj, lobj;
2323 		vm_offset_t addr;
2324 
2325 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2326 			continue;
2327 
2328 		bzero(kve, sizeof(*kve));
2329 		kve->kve_structsize = sizeof(*kve);
2330 
2331 		kve->kve_private_resident = 0;
2332 		obj = entry->object.vm_object;
2333 		if (obj != NULL) {
2334 			VM_OBJECT_RLOCK(obj);
2335 			if (obj->shadow_count == 1)
2336 				kve->kve_private_resident =
2337 				    obj->resident_page_count;
2338 		}
2339 		kve->kve_resident = 0;
2340 		addr = entry->start;
2341 		while (addr < entry->end) {
2342 			if (pmap_extract(map->pmap, addr))
2343 				kve->kve_resident++;
2344 			addr += PAGE_SIZE;
2345 		}
2346 
2347 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2348 			if (tobj != obj) {
2349 				VM_OBJECT_RLOCK(tobj);
2350 				kve->kve_offset += tobj->backing_object_offset;
2351 			}
2352 			if (lobj != obj)
2353 				VM_OBJECT_RUNLOCK(lobj);
2354 			lobj = tobj;
2355 		}
2356 
2357 		kve->kve_start = (void*)entry->start;
2358 		kve->kve_end = (void*)entry->end;
2359 		kve->kve_offset += (off_t)entry->offset;
2360 
2361 		if (entry->protection & VM_PROT_READ)
2362 			kve->kve_protection |= KVME_PROT_READ;
2363 		if (entry->protection & VM_PROT_WRITE)
2364 			kve->kve_protection |= KVME_PROT_WRITE;
2365 		if (entry->protection & VM_PROT_EXECUTE)
2366 			kve->kve_protection |= KVME_PROT_EXEC;
2367 
2368 		if (entry->eflags & MAP_ENTRY_COW)
2369 			kve->kve_flags |= KVME_FLAG_COW;
2370 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2371 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2372 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2373 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2374 
2375 		last_timestamp = map->timestamp;
2376 		vm_map_unlock_read(map);
2377 
2378 		kve->kve_fileid = 0;
2379 		kve->kve_fsid = 0;
2380 		freepath = NULL;
2381 		fullpath = "";
2382 		if (lobj) {
2383 			kve->kve_type = vm_object_kvme_type(lobj, &vp);
2384 			if (kve->kve_type == KVME_TYPE_MGTDEVICE)
2385 				kve->kve_type = KVME_TYPE_UNKNOWN;
2386 			if (vp != NULL)
2387 				vref(vp);
2388 			if (lobj != obj)
2389 				VM_OBJECT_RUNLOCK(lobj);
2390 
2391 			kve->kve_ref_count = obj->ref_count;
2392 			kve->kve_shadow_count = obj->shadow_count;
2393 			VM_OBJECT_RUNLOCK(obj);
2394 			if (vp != NULL) {
2395 				vn_fullpath(vp, &fullpath, &freepath);
2396 				cred = curthread->td_ucred;
2397 				vn_lock(vp, LK_SHARED | LK_RETRY);
2398 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2399 					kve->kve_fileid = va.va_fileid;
2400 					/* truncate */
2401 					kve->kve_fsid = va.va_fsid;
2402 				}
2403 				vput(vp);
2404 			}
2405 		} else {
2406 			kve->kve_type = KVME_TYPE_NONE;
2407 			kve->kve_ref_count = 0;
2408 			kve->kve_shadow_count = 0;
2409 		}
2410 
2411 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2412 		if (freepath != NULL)
2413 			free(freepath, M_TEMP);
2414 
2415 		error = SYSCTL_OUT(req, kve, sizeof(*kve));
2416 		vm_map_lock_read(map);
2417 		if (error)
2418 			break;
2419 		if (last_timestamp != map->timestamp) {
2420 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2421 			entry = tmp_entry;
2422 		}
2423 	}
2424 	vm_map_unlock_read(map);
2425 	vmspace_free(vm);
2426 	PRELE(p);
2427 	free(kve, M_TEMP);
2428 	return (error);
2429 }
2430 #endif	/* COMPAT_FREEBSD7 */
2431 
2432 #ifdef KINFO_VMENTRY_SIZE
2433 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2434 #endif
2435 
2436 void
2437 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry,
2438     int *resident_count, bool *super)
2439 {
2440 	vm_object_t obj, tobj;
2441 	vm_page_t m, m_adv;
2442 	vm_offset_t addr;
2443 	vm_paddr_t pa;
2444 	vm_pindex_t pi, pi_adv, pindex;
2445 
2446 	*super = false;
2447 	*resident_count = 0;
2448 	if (vmmap_skip_res_cnt)
2449 		return;
2450 
2451 	pa = 0;
2452 	obj = entry->object.vm_object;
2453 	addr = entry->start;
2454 	m_adv = NULL;
2455 	pi = OFF_TO_IDX(entry->offset);
2456 	for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) {
2457 		if (m_adv != NULL) {
2458 			m = m_adv;
2459 		} else {
2460 			pi_adv = atop(entry->end - addr);
2461 			pindex = pi;
2462 			for (tobj = obj;; tobj = tobj->backing_object) {
2463 				m = vm_page_find_least(tobj, pindex);
2464 				if (m != NULL) {
2465 					if (m->pindex == pindex)
2466 						break;
2467 					if (pi_adv > m->pindex - pindex) {
2468 						pi_adv = m->pindex - pindex;
2469 						m_adv = m;
2470 					}
2471 				}
2472 				if (tobj->backing_object == NULL)
2473 					goto next;
2474 				pindex += OFF_TO_IDX(tobj->
2475 				    backing_object_offset);
2476 			}
2477 		}
2478 		m_adv = NULL;
2479 		if (m->psind != 0 && addr + pagesizes[1] <= entry->end &&
2480 		    (addr & (pagesizes[1] - 1)) == 0 &&
2481 		    (pmap_mincore(map->pmap, addr, &pa) & MINCORE_SUPER) != 0) {
2482 			*super = true;
2483 			pi_adv = atop(pagesizes[1]);
2484 		} else {
2485 			/*
2486 			 * We do not test the found page on validity.
2487 			 * Either the page is busy and being paged in,
2488 			 * or it was invalidated.  The first case
2489 			 * should be counted as resident, the second
2490 			 * is not so clear; we do account both.
2491 			 */
2492 			pi_adv = 1;
2493 		}
2494 		*resident_count += pi_adv;
2495 next:;
2496 	}
2497 }
2498 
2499 /*
2500  * Must be called with the process locked and will return unlocked.
2501  */
2502 int
2503 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb, ssize_t maxlen, int flags)
2504 {
2505 	vm_map_entry_t entry, tmp_entry;
2506 	struct vattr va;
2507 	vm_map_t map;
2508 	vm_object_t obj, tobj, lobj;
2509 	char *fullpath, *freepath;
2510 	struct kinfo_vmentry *kve;
2511 	struct ucred *cred;
2512 	struct vnode *vp;
2513 	struct vmspace *vm;
2514 	vm_offset_t addr;
2515 	unsigned int last_timestamp;
2516 	int error;
2517 	bool super;
2518 
2519 	PROC_LOCK_ASSERT(p, MA_OWNED);
2520 
2521 	_PHOLD(p);
2522 	PROC_UNLOCK(p);
2523 	vm = vmspace_acquire_ref(p);
2524 	if (vm == NULL) {
2525 		PRELE(p);
2526 		return (ESRCH);
2527 	}
2528 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK | M_ZERO);
2529 
2530 	error = 0;
2531 	map = &vm->vm_map;
2532 	vm_map_lock_read(map);
2533 	VM_MAP_ENTRY_FOREACH(entry, map) {
2534 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2535 			continue;
2536 
2537 		addr = entry->end;
2538 		bzero(kve, sizeof(*kve));
2539 		obj = entry->object.vm_object;
2540 		if (obj != NULL) {
2541 			for (tobj = obj; tobj != NULL;
2542 			    tobj = tobj->backing_object) {
2543 				VM_OBJECT_RLOCK(tobj);
2544 				kve->kve_offset += tobj->backing_object_offset;
2545 				lobj = tobj;
2546 			}
2547 			if (obj->backing_object == NULL)
2548 				kve->kve_private_resident =
2549 				    obj->resident_page_count;
2550 			kern_proc_vmmap_resident(map, entry,
2551 			    &kve->kve_resident, &super);
2552 			if (super)
2553 				kve->kve_flags |= KVME_FLAG_SUPER;
2554 			for (tobj = obj; tobj != NULL;
2555 			    tobj = tobj->backing_object) {
2556 				if (tobj != obj && tobj != lobj)
2557 					VM_OBJECT_RUNLOCK(tobj);
2558 			}
2559 		} else {
2560 			lobj = NULL;
2561 		}
2562 
2563 		kve->kve_start = entry->start;
2564 		kve->kve_end = entry->end;
2565 		kve->kve_offset += entry->offset;
2566 
2567 		if (entry->protection & VM_PROT_READ)
2568 			kve->kve_protection |= KVME_PROT_READ;
2569 		if (entry->protection & VM_PROT_WRITE)
2570 			kve->kve_protection |= KVME_PROT_WRITE;
2571 		if (entry->protection & VM_PROT_EXECUTE)
2572 			kve->kve_protection |= KVME_PROT_EXEC;
2573 
2574 		if (entry->eflags & MAP_ENTRY_COW)
2575 			kve->kve_flags |= KVME_FLAG_COW;
2576 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2577 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2578 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2579 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2580 		if (entry->eflags & MAP_ENTRY_GROWS_UP)
2581 			kve->kve_flags |= KVME_FLAG_GROWS_UP;
2582 		if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2583 			kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2584 		if (entry->eflags & MAP_ENTRY_USER_WIRED)
2585 			kve->kve_flags |= KVME_FLAG_USER_WIRED;
2586 
2587 		last_timestamp = map->timestamp;
2588 		vm_map_unlock_read(map);
2589 
2590 		freepath = NULL;
2591 		fullpath = "";
2592 		if (lobj != NULL) {
2593 			kve->kve_type = vm_object_kvme_type(lobj, &vp);
2594 			if (vp != NULL)
2595 				vref(vp);
2596 			if (lobj != obj)
2597 				VM_OBJECT_RUNLOCK(lobj);
2598 
2599 			kve->kve_ref_count = obj->ref_count;
2600 			kve->kve_shadow_count = obj->shadow_count;
2601 			VM_OBJECT_RUNLOCK(obj);
2602 			if (vp != NULL) {
2603 				vn_fullpath(vp, &fullpath, &freepath);
2604 				kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2605 				cred = curthread->td_ucred;
2606 				vn_lock(vp, LK_SHARED | LK_RETRY);
2607 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2608 					kve->kve_vn_fileid = va.va_fileid;
2609 					kve->kve_vn_fsid = va.va_fsid;
2610 					kve->kve_vn_fsid_freebsd11 =
2611 					    kve->kve_vn_fsid; /* truncate */
2612 					kve->kve_vn_mode =
2613 					    MAKEIMODE(va.va_type, va.va_mode);
2614 					kve->kve_vn_size = va.va_size;
2615 					kve->kve_vn_rdev = va.va_rdev;
2616 					kve->kve_vn_rdev_freebsd11 =
2617 					    kve->kve_vn_rdev; /* truncate */
2618 					kve->kve_status = KF_ATTR_VALID;
2619 				}
2620 				vput(vp);
2621 			}
2622 		} else {
2623 			kve->kve_type = KVME_TYPE_NONE;
2624 			kve->kve_ref_count = 0;
2625 			kve->kve_shadow_count = 0;
2626 		}
2627 
2628 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2629 		if (freepath != NULL)
2630 			free(freepath, M_TEMP);
2631 
2632 		/* Pack record size down */
2633 		if ((flags & KERN_VMMAP_PACK_KINFO) != 0)
2634 			kve->kve_structsize =
2635 			    offsetof(struct kinfo_vmentry, kve_path) +
2636 			    strlen(kve->kve_path) + 1;
2637 		else
2638 			kve->kve_structsize = sizeof(*kve);
2639 		kve->kve_structsize = roundup(kve->kve_structsize,
2640 		    sizeof(uint64_t));
2641 
2642 		/* Halt filling and truncate rather than exceeding maxlen */
2643 		if (maxlen != -1 && maxlen < kve->kve_structsize) {
2644 			error = 0;
2645 			vm_map_lock_read(map);
2646 			break;
2647 		} else if (maxlen != -1)
2648 			maxlen -= kve->kve_structsize;
2649 
2650 		if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0)
2651 			error = ENOMEM;
2652 		vm_map_lock_read(map);
2653 		if (error != 0)
2654 			break;
2655 		if (last_timestamp != map->timestamp) {
2656 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2657 			entry = tmp_entry;
2658 		}
2659 	}
2660 	vm_map_unlock_read(map);
2661 	vmspace_free(vm);
2662 	PRELE(p);
2663 	free(kve, M_TEMP);
2664 	return (error);
2665 }
2666 
2667 static int
2668 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2669 {
2670 	struct proc *p;
2671 	struct sbuf sb;
2672 	int error, error2, *name;
2673 
2674 	name = (int *)arg1;
2675 	sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req);
2676 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2677 	error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
2678 	if (error != 0) {
2679 		sbuf_delete(&sb);
2680 		return (error);
2681 	}
2682 	error = kern_proc_vmmap_out(p, &sb, -1, KERN_VMMAP_PACK_KINFO);
2683 	error2 = sbuf_finish(&sb);
2684 	sbuf_delete(&sb);
2685 	return (error != 0 ? error : error2);
2686 }
2687 
2688 #if defined(STACK) || defined(DDB)
2689 static int
2690 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2691 {
2692 	struct kinfo_kstack *kkstp;
2693 	int error, i, *name, numthreads;
2694 	lwpid_t *lwpidarray;
2695 	struct thread *td;
2696 	struct stack *st;
2697 	struct sbuf sb;
2698 	struct proc *p;
2699 
2700 	name = (int *)arg1;
2701 	error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2702 	if (error != 0)
2703 		return (error);
2704 
2705 	kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2706 	st = stack_create(M_WAITOK);
2707 
2708 	lwpidarray = NULL;
2709 	PROC_LOCK(p);
2710 	do {
2711 		if (lwpidarray != NULL) {
2712 			free(lwpidarray, M_TEMP);
2713 			lwpidarray = NULL;
2714 		}
2715 		numthreads = p->p_numthreads;
2716 		PROC_UNLOCK(p);
2717 		lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2718 		    M_WAITOK | M_ZERO);
2719 		PROC_LOCK(p);
2720 	} while (numthreads < p->p_numthreads);
2721 
2722 	/*
2723 	 * XXXRW: During the below loop, execve(2) and countless other sorts
2724 	 * of changes could have taken place.  Should we check to see if the
2725 	 * vmspace has been replaced, or the like, in order to prevent
2726 	 * giving a snapshot that spans, say, execve(2), with some threads
2727 	 * before and some after?  Among other things, the credentials could
2728 	 * have changed, in which case the right to extract debug info might
2729 	 * no longer be assured.
2730 	 */
2731 	i = 0;
2732 	FOREACH_THREAD_IN_PROC(p, td) {
2733 		KASSERT(i < numthreads,
2734 		    ("sysctl_kern_proc_kstack: numthreads"));
2735 		lwpidarray[i] = td->td_tid;
2736 		i++;
2737 	}
2738 	PROC_UNLOCK(p);
2739 	numthreads = i;
2740 	for (i = 0; i < numthreads; i++) {
2741 		td = tdfind(lwpidarray[i], p->p_pid);
2742 		if (td == NULL) {
2743 			continue;
2744 		}
2745 		bzero(kkstp, sizeof(*kkstp));
2746 		(void)sbuf_new(&sb, kkstp->kkst_trace,
2747 		    sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2748 		thread_lock(td);
2749 		kkstp->kkst_tid = td->td_tid;
2750 		if (TD_IS_SWAPPED(td))
2751 			kkstp->kkst_state = KKST_STATE_SWAPPED;
2752 		else if (stack_save_td(st, td) == 0)
2753 			kkstp->kkst_state = KKST_STATE_STACKOK;
2754 		else
2755 			kkstp->kkst_state = KKST_STATE_RUNNING;
2756 		thread_unlock(td);
2757 		PROC_UNLOCK(p);
2758 		stack_sbuf_print(&sb, st);
2759 		sbuf_finish(&sb);
2760 		sbuf_delete(&sb);
2761 		error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2762 		if (error)
2763 			break;
2764 	}
2765 	PRELE(p);
2766 	if (lwpidarray != NULL)
2767 		free(lwpidarray, M_TEMP);
2768 	stack_destroy(st);
2769 	free(kkstp, M_TEMP);
2770 	return (error);
2771 }
2772 #endif
2773 
2774 /*
2775  * This sysctl allows a process to retrieve the full list of groups from
2776  * itself or another process.
2777  */
2778 static int
2779 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2780 {
2781 	pid_t *pidp = (pid_t *)arg1;
2782 	unsigned int arglen = arg2;
2783 	struct proc *p;
2784 	struct ucred *cred;
2785 	int error;
2786 
2787 	if (arglen != 1)
2788 		return (EINVAL);
2789 	if (*pidp == -1) {	/* -1 means this process */
2790 		p = req->td->td_proc;
2791 		PROC_LOCK(p);
2792 	} else {
2793 		error = pget(*pidp, PGET_CANSEE, &p);
2794 		if (error != 0)
2795 			return (error);
2796 	}
2797 
2798 	cred = crhold(p->p_ucred);
2799 	PROC_UNLOCK(p);
2800 
2801 	error = SYSCTL_OUT(req, cred->cr_groups,
2802 	    cred->cr_ngroups * sizeof(gid_t));
2803 	crfree(cred);
2804 	return (error);
2805 }
2806 
2807 /*
2808  * This sysctl allows a process to retrieve or/and set the resource limit for
2809  * another process.
2810  */
2811 static int
2812 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2813 {
2814 	int *name = (int *)arg1;
2815 	u_int namelen = arg2;
2816 	struct rlimit rlim;
2817 	struct proc *p;
2818 	u_int which;
2819 	int flags, error;
2820 
2821 	if (namelen != 2)
2822 		return (EINVAL);
2823 
2824 	which = (u_int)name[1];
2825 	if (which >= RLIM_NLIMITS)
2826 		return (EINVAL);
2827 
2828 	if (req->newptr != NULL && req->newlen != sizeof(rlim))
2829 		return (EINVAL);
2830 
2831 	flags = PGET_HOLD | PGET_NOTWEXIT;
2832 	if (req->newptr != NULL)
2833 		flags |= PGET_CANDEBUG;
2834 	else
2835 		flags |= PGET_CANSEE;
2836 	error = pget((pid_t)name[0], flags, &p);
2837 	if (error != 0)
2838 		return (error);
2839 
2840 	/*
2841 	 * Retrieve limit.
2842 	 */
2843 	if (req->oldptr != NULL) {
2844 		PROC_LOCK(p);
2845 		lim_rlimit_proc(p, which, &rlim);
2846 		PROC_UNLOCK(p);
2847 	}
2848 	error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
2849 	if (error != 0)
2850 		goto errout;
2851 
2852 	/*
2853 	 * Set limit.
2854 	 */
2855 	if (req->newptr != NULL) {
2856 		error = SYSCTL_IN(req, &rlim, sizeof(rlim));
2857 		if (error == 0)
2858 			error = kern_proc_setrlimit(curthread, p, which, &rlim);
2859 	}
2860 
2861 errout:
2862 	PRELE(p);
2863 	return (error);
2864 }
2865 
2866 /*
2867  * This sysctl allows a process to retrieve ps_strings structure location of
2868  * another process.
2869  */
2870 static int
2871 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
2872 {
2873 	int *name = (int *)arg1;
2874 	u_int namelen = arg2;
2875 	struct proc *p;
2876 	vm_offset_t ps_strings;
2877 	int error;
2878 #ifdef COMPAT_FREEBSD32
2879 	uint32_t ps_strings32;
2880 #endif
2881 
2882 	if (namelen != 1)
2883 		return (EINVAL);
2884 
2885 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2886 	if (error != 0)
2887 		return (error);
2888 #ifdef COMPAT_FREEBSD32
2889 	if ((req->flags & SCTL_MASK32) != 0) {
2890 		/*
2891 		 * We return 0 if the 32 bit emulation request is for a 64 bit
2892 		 * process.
2893 		 */
2894 		ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
2895 		    PTROUT(p->p_sysent->sv_psstrings) : 0;
2896 		PROC_UNLOCK(p);
2897 		error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
2898 		return (error);
2899 	}
2900 #endif
2901 	ps_strings = p->p_sysent->sv_psstrings;
2902 	PROC_UNLOCK(p);
2903 	error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
2904 	return (error);
2905 }
2906 
2907 /*
2908  * This sysctl allows a process to retrieve umask of another process.
2909  */
2910 static int
2911 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
2912 {
2913 	int *name = (int *)arg1;
2914 	u_int namelen = arg2;
2915 	struct proc *p;
2916 	int error;
2917 	u_short cmask;
2918 	pid_t pid;
2919 
2920 	if (namelen != 1)
2921 		return (EINVAL);
2922 
2923 	pid = (pid_t)name[0];
2924 	p = curproc;
2925 	if (pid == p->p_pid || pid == 0) {
2926 		cmask = p->p_pd->pd_cmask;
2927 		goto out;
2928 	}
2929 
2930 	error = pget(pid, PGET_WANTREAD, &p);
2931 	if (error != 0)
2932 		return (error);
2933 
2934 	cmask = p->p_pd->pd_cmask;
2935 	PRELE(p);
2936 out:
2937 	error = SYSCTL_OUT(req, &cmask, sizeof(cmask));
2938 	return (error);
2939 }
2940 
2941 /*
2942  * This sysctl allows a process to set and retrieve binary osreldate of
2943  * another process.
2944  */
2945 static int
2946 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
2947 {
2948 	int *name = (int *)arg1;
2949 	u_int namelen = arg2;
2950 	struct proc *p;
2951 	int flags, error, osrel;
2952 
2953 	if (namelen != 1)
2954 		return (EINVAL);
2955 
2956 	if (req->newptr != NULL && req->newlen != sizeof(osrel))
2957 		return (EINVAL);
2958 
2959 	flags = PGET_HOLD | PGET_NOTWEXIT;
2960 	if (req->newptr != NULL)
2961 		flags |= PGET_CANDEBUG;
2962 	else
2963 		flags |= PGET_CANSEE;
2964 	error = pget((pid_t)name[0], flags, &p);
2965 	if (error != 0)
2966 		return (error);
2967 
2968 	error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
2969 	if (error != 0)
2970 		goto errout;
2971 
2972 	if (req->newptr != NULL) {
2973 		error = SYSCTL_IN(req, &osrel, sizeof(osrel));
2974 		if (error != 0)
2975 			goto errout;
2976 		if (osrel < 0) {
2977 			error = EINVAL;
2978 			goto errout;
2979 		}
2980 		p->p_osrel = osrel;
2981 	}
2982 errout:
2983 	PRELE(p);
2984 	return (error);
2985 }
2986 
2987 static int
2988 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS)
2989 {
2990 	int *name = (int *)arg1;
2991 	u_int namelen = arg2;
2992 	struct proc *p;
2993 	struct kinfo_sigtramp kst;
2994 	const struct sysentvec *sv;
2995 	int error;
2996 #ifdef COMPAT_FREEBSD32
2997 	struct kinfo_sigtramp32 kst32;
2998 #endif
2999 
3000 	if (namelen != 1)
3001 		return (EINVAL);
3002 
3003 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
3004 	if (error != 0)
3005 		return (error);
3006 	sv = p->p_sysent;
3007 #ifdef COMPAT_FREEBSD32
3008 	if ((req->flags & SCTL_MASK32) != 0) {
3009 		bzero(&kst32, sizeof(kst32));
3010 		if (SV_PROC_FLAG(p, SV_ILP32)) {
3011 			if (sv->sv_sigcode_base != 0) {
3012 				kst32.ksigtramp_start = sv->sv_sigcode_base;
3013 				kst32.ksigtramp_end = sv->sv_sigcode_base +
3014 				    *sv->sv_szsigcode;
3015 			} else {
3016 				kst32.ksigtramp_start = sv->sv_psstrings -
3017 				    *sv->sv_szsigcode;
3018 				kst32.ksigtramp_end = sv->sv_psstrings;
3019 			}
3020 		}
3021 		PROC_UNLOCK(p);
3022 		error = SYSCTL_OUT(req, &kst32, sizeof(kst32));
3023 		return (error);
3024 	}
3025 #endif
3026 	bzero(&kst, sizeof(kst));
3027 	if (sv->sv_sigcode_base != 0) {
3028 		kst.ksigtramp_start = (char *)sv->sv_sigcode_base;
3029 		kst.ksigtramp_end = (char *)sv->sv_sigcode_base +
3030 		    *sv->sv_szsigcode;
3031 	} else {
3032 		kst.ksigtramp_start = (char *)sv->sv_psstrings -
3033 		    *sv->sv_szsigcode;
3034 		kst.ksigtramp_end = (char *)sv->sv_psstrings;
3035 	}
3036 	PROC_UNLOCK(p);
3037 	error = SYSCTL_OUT(req, &kst, sizeof(kst));
3038 	return (error);
3039 }
3040 
3041 static int
3042 sysctl_kern_proc_sigfastblk(SYSCTL_HANDLER_ARGS)
3043 {
3044 	int *name = (int *)arg1;
3045 	u_int namelen = arg2;
3046 	pid_t pid;
3047 	struct proc *p;
3048 	struct thread *td1;
3049 	uintptr_t addr;
3050 #ifdef COMPAT_FREEBSD32
3051 	uint32_t addr32;
3052 #endif
3053 	int error;
3054 
3055 	if (namelen != 1 || req->newptr != NULL)
3056 		return (EINVAL);
3057 
3058 	pid = (pid_t)name[0];
3059 	error = pget(pid, PGET_HOLD | PGET_NOTWEXIT | PGET_CANDEBUG, &p);
3060 	if (error != 0)
3061 		return (error);
3062 
3063 	PROC_LOCK(p);
3064 #ifdef COMPAT_FREEBSD32
3065 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3066 		if (!SV_PROC_FLAG(p, SV_ILP32)) {
3067 			error = EINVAL;
3068 			goto errlocked;
3069 		}
3070 	}
3071 #endif
3072 	if (pid <= PID_MAX) {
3073 		td1 = FIRST_THREAD_IN_PROC(p);
3074 	} else {
3075 		FOREACH_THREAD_IN_PROC(p, td1) {
3076 			if (td1->td_tid == pid)
3077 				break;
3078 		}
3079 	}
3080 	if (td1 == NULL) {
3081 		error = ESRCH;
3082 		goto errlocked;
3083 	}
3084 	/*
3085 	 * The access to the private thread flags.  It is fine as far
3086 	 * as no out-of-thin-air values are read from td_pflags, and
3087 	 * usermode read of the td_sigblock_ptr is racy inherently,
3088 	 * since target process might have already changed it
3089 	 * meantime.
3090 	 */
3091 	if ((td1->td_pflags & TDP_SIGFASTBLOCK) != 0)
3092 		addr = (uintptr_t)td1->td_sigblock_ptr;
3093 	else
3094 		error = ENOTTY;
3095 
3096 errlocked:
3097 	_PRELE(p);
3098 	PROC_UNLOCK(p);
3099 	if (error != 0)
3100 		return (error);
3101 
3102 #ifdef COMPAT_FREEBSD32
3103 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3104 		addr32 = addr;
3105 		error = SYSCTL_OUT(req, &addr32, sizeof(addr32));
3106 	} else
3107 #endif
3108 		error = SYSCTL_OUT(req, &addr, sizeof(addr));
3109 	return (error);
3110 }
3111 
3112 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,  0,
3113     "Process table");
3114 
3115 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
3116 	CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
3117 	"Return entire process table");
3118 
3119 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3120 	sysctl_kern_proc, "Process table");
3121 
3122 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
3123 	sysctl_kern_proc, "Process table");
3124 
3125 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3126 	sysctl_kern_proc, "Process table");
3127 
3128 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
3129 	CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3130 
3131 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
3132 	sysctl_kern_proc, "Process table");
3133 
3134 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3135 	sysctl_kern_proc, "Process table");
3136 
3137 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3138 	sysctl_kern_proc, "Process table");
3139 
3140 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3141 	sysctl_kern_proc, "Process table");
3142 
3143 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
3144 	sysctl_kern_proc, "Return process table, no threads");
3145 
3146 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
3147 	CTLFLAG_RW | CTLFLAG_CAPWR | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
3148 	sysctl_kern_proc_args, "Process argument list");
3149 
3150 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
3151 	sysctl_kern_proc_env, "Process environment");
3152 
3153 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
3154 	CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
3155 
3156 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
3157 	CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
3158 
3159 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
3160 	CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
3161 	"Process syscall vector name (ABI type)");
3162 
3163 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
3164 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3165 
3166 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
3167 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3168 
3169 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
3170 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3171 
3172 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
3173 	sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3174 
3175 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
3176 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3177 
3178 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
3179 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3180 
3181 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
3182 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3183 
3184 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
3185 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3186 
3187 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
3188 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
3189 	"Return process table, no threads");
3190 
3191 #ifdef COMPAT_FREEBSD7
3192 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
3193 	CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
3194 #endif
3195 
3196 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
3197 	CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
3198 
3199 #if defined(STACK) || defined(DDB)
3200 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
3201 	CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
3202 #endif
3203 
3204 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
3205 	CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
3206 
3207 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
3208 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
3209 	"Process resource limits");
3210 
3211 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
3212 	CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
3213 	"Process ps_strings location");
3214 
3215 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
3216 	CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
3217 
3218 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
3219 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
3220 	"Process binary osreldate");
3221 
3222 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD |
3223 	CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp,
3224 	"Process signal trampoline location");
3225 
3226 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGFASTBLK, sigfastblk, CTLFLAG_RD |
3227 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_sigfastblk,
3228 	"Thread sigfastblock address");
3229 
3230 int allproc_gen;
3231 
3232 /*
3233  * stop_all_proc() purpose is to stop all process which have usermode,
3234  * except current process for obvious reasons.  This makes it somewhat
3235  * unreliable when invoked from multithreaded process.  The service
3236  * must not be user-callable anyway.
3237  */
3238 void
3239 stop_all_proc(void)
3240 {
3241 	struct proc *cp, *p;
3242 	int r, gen;
3243 	bool restart, seen_stopped, seen_exiting, stopped_some;
3244 
3245 	cp = curproc;
3246 allproc_loop:
3247 	sx_xlock(&allproc_lock);
3248 	gen = allproc_gen;
3249 	seen_exiting = seen_stopped = stopped_some = restart = false;
3250 	LIST_REMOVE(cp, p_list);
3251 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3252 	for (;;) {
3253 		p = LIST_NEXT(cp, p_list);
3254 		if (p == NULL)
3255 			break;
3256 		LIST_REMOVE(cp, p_list);
3257 		LIST_INSERT_AFTER(p, cp, p_list);
3258 		PROC_LOCK(p);
3259 		if ((p->p_flag & (P_KPROC | P_SYSTEM | P_TOTAL_STOP)) != 0) {
3260 			PROC_UNLOCK(p);
3261 			continue;
3262 		}
3263 		if ((p->p_flag & P_WEXIT) != 0) {
3264 			seen_exiting = true;
3265 			PROC_UNLOCK(p);
3266 			continue;
3267 		}
3268 		if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) {
3269 			/*
3270 			 * Stopped processes are tolerated when there
3271 			 * are no other processes which might continue
3272 			 * them.  P_STOPPED_SINGLE but not
3273 			 * P_TOTAL_STOP process still has at least one
3274 			 * thread running.
3275 			 */
3276 			seen_stopped = true;
3277 			PROC_UNLOCK(p);
3278 			continue;
3279 		}
3280 		sx_xunlock(&allproc_lock);
3281 		_PHOLD(p);
3282 		r = thread_single(p, SINGLE_ALLPROC);
3283 		if (r != 0)
3284 			restart = true;
3285 		else
3286 			stopped_some = true;
3287 		_PRELE(p);
3288 		PROC_UNLOCK(p);
3289 		sx_xlock(&allproc_lock);
3290 	}
3291 	/* Catch forked children we did not see in iteration. */
3292 	if (gen != allproc_gen)
3293 		restart = true;
3294 	sx_xunlock(&allproc_lock);
3295 	if (restart || stopped_some || seen_exiting || seen_stopped) {
3296 		kern_yield(PRI_USER);
3297 		goto allproc_loop;
3298 	}
3299 }
3300 
3301 void
3302 resume_all_proc(void)
3303 {
3304 	struct proc *cp, *p;
3305 
3306 	cp = curproc;
3307 	sx_xlock(&allproc_lock);
3308 again:
3309 	LIST_REMOVE(cp, p_list);
3310 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3311 	for (;;) {
3312 		p = LIST_NEXT(cp, p_list);
3313 		if (p == NULL)
3314 			break;
3315 		LIST_REMOVE(cp, p_list);
3316 		LIST_INSERT_AFTER(p, cp, p_list);
3317 		PROC_LOCK(p);
3318 		if ((p->p_flag & P_TOTAL_STOP) != 0) {
3319 			sx_xunlock(&allproc_lock);
3320 			_PHOLD(p);
3321 			thread_single_end(p, SINGLE_ALLPROC);
3322 			_PRELE(p);
3323 			PROC_UNLOCK(p);
3324 			sx_xlock(&allproc_lock);
3325 		} else {
3326 			PROC_UNLOCK(p);
3327 		}
3328 	}
3329 	/*  Did the loop above missed any stopped process ? */
3330 	FOREACH_PROC_IN_SYSTEM(p) {
3331 		/* No need for proc lock. */
3332 		if ((p->p_flag & P_TOTAL_STOP) != 0)
3333 			goto again;
3334 	}
3335 	sx_xunlock(&allproc_lock);
3336 }
3337 
3338 /* #define	TOTAL_STOP_DEBUG	1 */
3339 #ifdef TOTAL_STOP_DEBUG
3340 volatile static int ap_resume;
3341 #include <sys/mount.h>
3342 
3343 static int
3344 sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS)
3345 {
3346 	int error, val;
3347 
3348 	val = 0;
3349 	ap_resume = 0;
3350 	error = sysctl_handle_int(oidp, &val, 0, req);
3351 	if (error != 0 || req->newptr == NULL)
3352 		return (error);
3353 	if (val != 0) {
3354 		stop_all_proc();
3355 		syncer_suspend();
3356 		while (ap_resume == 0)
3357 			;
3358 		syncer_resume();
3359 		resume_all_proc();
3360 	}
3361 	return (0);
3362 }
3363 
3364 SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW |
3365     CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0,
3366     sysctl_debug_stop_all_proc, "I",
3367     "");
3368 #endif
3369