xref: /freebsd/sys/kern/kern_proc.c (revision d5e3895ea4fe4ef9db8823774e07b4368180a23e)
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/sysctl.h>
69 #include <sys/filedesc.h>
70 #include <sys/tty.h>
71 #include <sys/signalvar.h>
72 #include <sys/sdt.h>
73 #include <sys/sx.h>
74 #include <sys/user.h>
75 #include <sys/vnode.h>
76 #include <sys/wait.h>
77 
78 #ifdef DDB
79 #include <ddb/ddb.h>
80 #endif
81 
82 #include <vm/vm.h>
83 #include <vm/vm_param.h>
84 #include <vm/vm_extern.h>
85 #include <vm/pmap.h>
86 #include <vm/vm_map.h>
87 #include <vm/vm_object.h>
88 #include <vm/vm_page.h>
89 #include <vm/uma.h>
90 
91 #include <fs/devfs/devfs.h>
92 
93 #ifdef COMPAT_FREEBSD32
94 #include <compat/freebsd32/freebsd32.h>
95 #include <compat/freebsd32/freebsd32_util.h>
96 #endif
97 
98 SDT_PROVIDER_DEFINE(proc);
99 
100 MALLOC_DEFINE(M_PGRP, "pgrp", "process group header");
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 pgadjustjobc(struct pgrp *pgrp, int entering);
112 static void pgdelete(struct pgrp *);
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 
135 /*
136  * The offset of various fields in struct proc and struct thread.
137  * These are used by kernel debuggers to enumerate kernel threads and
138  * processes.
139  */
140 const int proc_off_p_pid = offsetof(struct proc, p_pid);
141 const int proc_off_p_comm = offsetof(struct proc, p_comm);
142 const int proc_off_p_list = offsetof(struct proc, p_list);
143 const int proc_off_p_threads = offsetof(struct proc, p_threads);
144 const int thread_off_td_tid = offsetof(struct thread, td_tid);
145 const int thread_off_td_name = offsetof(struct thread, td_name);
146 const int thread_off_td_oncpu = offsetof(struct thread, td_oncpu);
147 const int thread_off_td_pcb = offsetof(struct thread, td_pcb);
148 const int thread_off_td_plist = offsetof(struct thread, td_plist);
149 
150 EVENTHANDLER_LIST_DEFINE(process_ctor);
151 EVENTHANDLER_LIST_DEFINE(process_dtor);
152 EVENTHANDLER_LIST_DEFINE(process_init);
153 EVENTHANDLER_LIST_DEFINE(process_fini);
154 EVENTHANDLER_LIST_DEFINE(process_exit);
155 EVENTHANDLER_LIST_DEFINE(process_fork);
156 EVENTHANDLER_LIST_DEFINE(process_exec);
157 
158 int kstack_pages = KSTACK_PAGES;
159 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0,
160     "Kernel stack size in pages");
161 static int vmmap_skip_res_cnt = 0;
162 SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW,
163     &vmmap_skip_res_cnt, 0,
164     "Skip calculation of the pages resident count in kern.proc.vmmap");
165 
166 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
167 #ifdef COMPAT_FREEBSD32
168 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
169 #endif
170 
171 /*
172  * Initialize global process hashing structures.
173  */
174 void
175 procinit(void)
176 {
177 	u_long i;
178 
179 	sx_init(&allproc_lock, "allproc");
180 	sx_init(&proctree_lock, "proctree");
181 	mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
182 	mtx_init(&procid_lock, "procid", NULL, MTX_DEF);
183 	LIST_INIT(&allproc);
184 	pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
185 	pidhashlock = (pidhash + 1) / 64;
186 	if (pidhashlock > 0)
187 		pidhashlock--;
188 	pidhashtbl_lock = malloc(sizeof(*pidhashtbl_lock) * (pidhashlock + 1),
189 	    M_PROC, M_WAITOK | M_ZERO);
190 	for (i = 0; i < pidhashlock + 1; i++)
191 		sx_init_flags(&pidhashtbl_lock[i], "pidhash", SX_DUPOK);
192 	pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
193 	proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
194 	    proc_ctor, proc_dtor, proc_init, proc_fini,
195 	    UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
196 	uihashinit();
197 }
198 
199 /*
200  * Prepare a proc for use.
201  */
202 static int
203 proc_ctor(void *mem, int size, void *arg, int flags)
204 {
205 	struct proc *p;
206 	struct thread *td;
207 
208 	p = (struct proc *)mem;
209 	EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
210 	td = FIRST_THREAD_IN_PROC(p);
211 	if (td != NULL) {
212 		/* Make sure all thread constructors are executed */
213 		EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
214 	}
215 	return (0);
216 }
217 
218 /*
219  * Reclaim a proc after use.
220  */
221 static void
222 proc_dtor(void *mem, int size, void *arg)
223 {
224 	struct proc *p;
225 	struct thread *td;
226 
227 	/* INVARIANTS checks go here */
228 	p = (struct proc *)mem;
229 	td = FIRST_THREAD_IN_PROC(p);
230 	if (td != NULL) {
231 #ifdef INVARIANTS
232 		KASSERT((p->p_numthreads == 1),
233 		    ("bad number of threads in exiting process"));
234 		KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
235 #endif
236 		/* Free all OSD associated to this thread. */
237 		osd_thread_exit(td);
238 		td_softdep_cleanup(td);
239 		MPASS(td->td_su == NULL);
240 
241 		/* Make sure all thread destructors are executed */
242 		EVENTHANDLER_DIRECT_INVOKE(thread_dtor, td);
243 	}
244 	EVENTHANDLER_DIRECT_INVOKE(process_dtor, p);
245 	if (p->p_ksi != NULL)
246 		KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
247 }
248 
249 /*
250  * Initialize type-stable parts of a proc (when newly created).
251  */
252 static int
253 proc_init(void *mem, int size, int flags)
254 {
255 	struct proc *p;
256 
257 	p = (struct proc *)mem;
258 	mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK | MTX_NEW);
259 	mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_NEW);
260 	mtx_init(&p->p_statmtx, "pstatl", NULL, MTX_SPIN | MTX_NEW);
261 	mtx_init(&p->p_itimmtx, "pitiml", NULL, MTX_SPIN | MTX_NEW);
262 	mtx_init(&p->p_profmtx, "pprofl", NULL, MTX_SPIN | MTX_NEW);
263 	cv_init(&p->p_pwait, "ppwait");
264 	TAILQ_INIT(&p->p_threads);	     /* all threads in proc */
265 	EVENTHANDLER_DIRECT_INVOKE(process_init, p);
266 	p->p_stats = pstats_alloc();
267 	p->p_pgrp = NULL;
268 	return (0);
269 }
270 
271 /*
272  * UMA should ensure that this function is never called.
273  * Freeing a proc structure would violate type stability.
274  */
275 static void
276 proc_fini(void *mem, int size)
277 {
278 #ifdef notnow
279 	struct proc *p;
280 
281 	p = (struct proc *)mem;
282 	EVENTHANDLER_DIRECT_INVOKE(process_fini, p);
283 	pstats_free(p->p_stats);
284 	thread_free(FIRST_THREAD_IN_PROC(p));
285 	mtx_destroy(&p->p_mtx);
286 	if (p->p_ksi != NULL)
287 		ksiginfo_free(p->p_ksi);
288 #else
289 	panic("proc reclaimed");
290 #endif
291 }
292 
293 /*
294  * PID space management.
295  *
296  * These bitmaps are used by fork_findpid.
297  */
298 bitstr_t bit_decl(proc_id_pidmap, PID_MAX);
299 bitstr_t bit_decl(proc_id_grpidmap, PID_MAX);
300 bitstr_t bit_decl(proc_id_sessidmap, PID_MAX);
301 bitstr_t bit_decl(proc_id_reapmap, PID_MAX);
302 
303 static bitstr_t *proc_id_array[] = {
304 	proc_id_pidmap,
305 	proc_id_grpidmap,
306 	proc_id_sessidmap,
307 	proc_id_reapmap,
308 };
309 
310 void
311 proc_id_set(int type, pid_t id)
312 {
313 
314 	KASSERT(type >= 0 && type < nitems(proc_id_array),
315 	    ("invalid type %d\n", type));
316 	mtx_lock(&procid_lock);
317 	KASSERT(bit_test(proc_id_array[type], id) == 0,
318 	    ("bit %d already set in %d\n", id, type));
319 	bit_set(proc_id_array[type], id);
320 	mtx_unlock(&procid_lock);
321 }
322 
323 void
324 proc_id_set_cond(int type, pid_t id)
325 {
326 
327 	KASSERT(type >= 0 && type < nitems(proc_id_array),
328 	    ("invalid type %d\n", type));
329 	if (bit_test(proc_id_array[type], id))
330 		return;
331 	mtx_lock(&procid_lock);
332 	bit_set(proc_id_array[type], id);
333 	mtx_unlock(&procid_lock);
334 }
335 
336 void
337 proc_id_clear(int type, pid_t id)
338 {
339 
340 	KASSERT(type >= 0 && type < nitems(proc_id_array),
341 	    ("invalid type %d\n", type));
342 	mtx_lock(&procid_lock);
343 	KASSERT(bit_test(proc_id_array[type], id) != 0,
344 	    ("bit %d not set in %d\n", id, type));
345 	bit_clear(proc_id_array[type], id);
346 	mtx_unlock(&procid_lock);
347 }
348 
349 /*
350  * Is p an inferior of the current process?
351  */
352 int
353 inferior(struct proc *p)
354 {
355 
356 	sx_assert(&proctree_lock, SX_LOCKED);
357 	PROC_LOCK_ASSERT(p, MA_OWNED);
358 	for (; p != curproc; p = proc_realparent(p)) {
359 		if (p->p_pid == 0)
360 			return (0);
361 	}
362 	return (1);
363 }
364 
365 /*
366  * Shared lock all the pid hash lists.
367  */
368 void
369 pidhash_slockall(void)
370 {
371 	u_long i;
372 
373 	for (i = 0; i < pidhashlock + 1; i++)
374 		sx_slock(&pidhashtbl_lock[i]);
375 }
376 
377 /*
378  * Shared unlock all the pid hash lists.
379  */
380 void
381 pidhash_sunlockall(void)
382 {
383 	u_long i;
384 
385 	for (i = 0; i < pidhashlock + 1; i++)
386 		sx_sunlock(&pidhashtbl_lock[i]);
387 }
388 
389 /*
390  * Similar to pfind_any(), this function finds zombies.
391  */
392 struct proc *
393 pfind_any_locked(pid_t pid)
394 {
395 	struct proc *p;
396 
397 	sx_assert(PIDHASHLOCK(pid), SX_LOCKED);
398 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
399 		if (p->p_pid == pid) {
400 			PROC_LOCK(p);
401 			if (p->p_state == PRS_NEW) {
402 				PROC_UNLOCK(p);
403 				p = NULL;
404 			}
405 			break;
406 		}
407 	}
408 	return (p);
409 }
410 
411 /*
412  * Locate a process by number.
413  *
414  * By not returning processes in the PRS_NEW state, we allow callers to avoid
415  * testing for that condition to avoid dereferencing p_ucred, et al.
416  */
417 static __always_inline struct proc *
418 _pfind(pid_t pid, bool zombie)
419 {
420 	struct proc *p;
421 
422 	p = curproc;
423 	if (p->p_pid == pid) {
424 		PROC_LOCK(p);
425 		return (p);
426 	}
427 	sx_slock(PIDHASHLOCK(pid));
428 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
429 		if (p->p_pid == pid) {
430 			PROC_LOCK(p);
431 			if (p->p_state == PRS_NEW ||
432 			    (!zombie && p->p_state == PRS_ZOMBIE)) {
433 				PROC_UNLOCK(p);
434 				p = NULL;
435 			}
436 			break;
437 		}
438 	}
439 	sx_sunlock(PIDHASHLOCK(pid));
440 	return (p);
441 }
442 
443 struct proc *
444 pfind(pid_t pid)
445 {
446 
447 	return (_pfind(pid, false));
448 }
449 
450 /*
451  * Same as pfind but allow zombies.
452  */
453 struct proc *
454 pfind_any(pid_t pid)
455 {
456 
457 	return (_pfind(pid, true));
458 }
459 
460 /*
461  * Locate a process group by number.
462  * The caller must hold proctree_lock.
463  */
464 struct pgrp *
465 pgfind(pid_t pgid)
466 {
467 	struct pgrp *pgrp;
468 
469 	sx_assert(&proctree_lock, SX_LOCKED);
470 
471 	LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
472 		if (pgrp->pg_id == pgid) {
473 			PGRP_LOCK(pgrp);
474 			return (pgrp);
475 		}
476 	}
477 	return (NULL);
478 }
479 
480 /*
481  * Locate process and do additional manipulations, depending on flags.
482  */
483 int
484 pget(pid_t pid, int flags, struct proc **pp)
485 {
486 	struct proc *p;
487 	struct thread *td1;
488 	int error;
489 
490 	p = curproc;
491 	if (p->p_pid == pid) {
492 		PROC_LOCK(p);
493 	} else {
494 		p = NULL;
495 		if (pid <= PID_MAX) {
496 			if ((flags & PGET_NOTWEXIT) == 0)
497 				p = pfind_any(pid);
498 			else
499 				p = pfind(pid);
500 		} else if ((flags & PGET_NOTID) == 0) {
501 			td1 = tdfind(pid, -1);
502 			if (td1 != NULL)
503 				p = td1->td_proc;
504 		}
505 		if (p == NULL)
506 			return (ESRCH);
507 		if ((flags & PGET_CANSEE) != 0) {
508 			error = p_cansee(curthread, p);
509 			if (error != 0)
510 				goto errout;
511 		}
512 	}
513 	if ((flags & PGET_CANDEBUG) != 0) {
514 		error = p_candebug(curthread, p);
515 		if (error != 0)
516 			goto errout;
517 	}
518 	if ((flags & PGET_ISCURRENT) != 0 && curproc != p) {
519 		error = EPERM;
520 		goto errout;
521 	}
522 	if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) {
523 		error = ESRCH;
524 		goto errout;
525 	}
526 	if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) {
527 		/*
528 		 * XXXRW: Not clear ESRCH is the right error during proc
529 		 * execve().
530 		 */
531 		error = ESRCH;
532 		goto errout;
533 	}
534 	if ((flags & PGET_HOLD) != 0) {
535 		_PHOLD(p);
536 		PROC_UNLOCK(p);
537 	}
538 	*pp = p;
539 	return (0);
540 errout:
541 	PROC_UNLOCK(p);
542 	return (error);
543 }
544 
545 /*
546  * Create a new process group.
547  * pgid must be equal to the pid of p.
548  * Begin a new session if required.
549  */
550 int
551 enterpgrp(struct proc *p, pid_t pgid, struct pgrp *pgrp, struct session *sess)
552 {
553 
554 	sx_assert(&proctree_lock, SX_XLOCKED);
555 
556 	KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
557 	KASSERT(p->p_pid == pgid,
558 	    ("enterpgrp: new pgrp and pid != pgid"));
559 	KASSERT(pgfind(pgid) == NULL,
560 	    ("enterpgrp: pgrp with pgid exists"));
561 	KASSERT(!SESS_LEADER(p),
562 	    ("enterpgrp: session leader attempted setpgrp"));
563 
564 	mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
565 
566 	if (sess != NULL) {
567 		/*
568 		 * new session
569 		 */
570 		mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
571 		PROC_LOCK(p);
572 		p->p_flag &= ~P_CONTROLT;
573 		PROC_UNLOCK(p);
574 		PGRP_LOCK(pgrp);
575 		sess->s_leader = p;
576 		sess->s_sid = p->p_pid;
577 		proc_id_set(PROC_ID_SESSION, p->p_pid);
578 		refcount_init(&sess->s_count, 1);
579 		sess->s_ttyvp = NULL;
580 		sess->s_ttydp = NULL;
581 		sess->s_ttyp = NULL;
582 		bcopy(p->p_session->s_login, sess->s_login,
583 			    sizeof(sess->s_login));
584 		pgrp->pg_session = sess;
585 		KASSERT(p == curproc,
586 		    ("enterpgrp: mksession and p != curproc"));
587 	} else {
588 		pgrp->pg_session = p->p_session;
589 		sess_hold(pgrp->pg_session);
590 		PGRP_LOCK(pgrp);
591 	}
592 	pgrp->pg_id = pgid;
593 	proc_id_set(PROC_ID_GROUP, p->p_pid);
594 	LIST_INIT(&pgrp->pg_members);
595 
596 	/*
597 	 * As we have an exclusive lock of proctree_lock,
598 	 * this should not deadlock.
599 	 */
600 	LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
601 	pgrp->pg_jobc = 0;
602 	SLIST_INIT(&pgrp->pg_sigiolst);
603 	PGRP_UNLOCK(pgrp);
604 
605 	doenterpgrp(p, pgrp);
606 
607 	return (0);
608 }
609 
610 /*
611  * Move p to an existing process group
612  */
613 int
614 enterthispgrp(struct proc *p, struct pgrp *pgrp)
615 {
616 
617 	sx_assert(&proctree_lock, SX_XLOCKED);
618 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
619 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
620 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
621 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
622 	KASSERT(pgrp->pg_session == p->p_session,
623 		("%s: pgrp's session %p, p->p_session %p.\n",
624 		__func__,
625 		pgrp->pg_session,
626 		p->p_session));
627 	KASSERT(pgrp != p->p_pgrp,
628 		("%s: p belongs to pgrp.", __func__));
629 
630 	doenterpgrp(p, pgrp);
631 
632 	return (0);
633 }
634 
635 /*
636  * Move p to a process group
637  */
638 static void
639 doenterpgrp(struct proc *p, struct pgrp *pgrp)
640 {
641 	struct pgrp *savepgrp;
642 
643 	sx_assert(&proctree_lock, SX_XLOCKED);
644 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
645 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
646 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
647 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
648 
649 	savepgrp = p->p_pgrp;
650 
651 	/*
652 	 * Adjust eligibility of affected pgrps to participate in job control.
653 	 * Increment eligibility counts before decrementing, otherwise we
654 	 * could reach 0 spuriously during the first call.
655 	 */
656 	fixjobc(p, pgrp, 1);
657 	fixjobc(p, p->p_pgrp, 0);
658 
659 	PGRP_LOCK(pgrp);
660 	PGRP_LOCK(savepgrp);
661 	PROC_LOCK(p);
662 	LIST_REMOVE(p, p_pglist);
663 	p->p_pgrp = pgrp;
664 	PROC_UNLOCK(p);
665 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
666 	PGRP_UNLOCK(savepgrp);
667 	PGRP_UNLOCK(pgrp);
668 	if (LIST_EMPTY(&savepgrp->pg_members))
669 		pgdelete(savepgrp);
670 }
671 
672 /*
673  * remove process from process group
674  */
675 int
676 leavepgrp(struct proc *p)
677 {
678 	struct pgrp *savepgrp;
679 
680 	sx_assert(&proctree_lock, SX_XLOCKED);
681 	savepgrp = p->p_pgrp;
682 	PGRP_LOCK(savepgrp);
683 	PROC_LOCK(p);
684 	LIST_REMOVE(p, p_pglist);
685 	p->p_pgrp = NULL;
686 	PROC_UNLOCK(p);
687 	PGRP_UNLOCK(savepgrp);
688 	if (LIST_EMPTY(&savepgrp->pg_members))
689 		pgdelete(savepgrp);
690 	return (0);
691 }
692 
693 /*
694  * delete a process group
695  */
696 static void
697 pgdelete(struct pgrp *pgrp)
698 {
699 	struct session *savesess;
700 	struct tty *tp;
701 
702 	sx_assert(&proctree_lock, SX_XLOCKED);
703 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
704 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
705 
706 	/*
707 	 * Reset any sigio structures pointing to us as a result of
708 	 * F_SETOWN with our pgid.
709 	 */
710 	funsetownlst(&pgrp->pg_sigiolst);
711 
712 	PGRP_LOCK(pgrp);
713 	tp = pgrp->pg_session->s_ttyp;
714 	LIST_REMOVE(pgrp, pg_hash);
715 	savesess = pgrp->pg_session;
716 	PGRP_UNLOCK(pgrp);
717 
718 	/* Remove the reference to the pgrp before deallocating it. */
719 	if (tp != NULL) {
720 		tty_lock(tp);
721 		tty_rel_pgrp(tp, pgrp);
722 	}
723 
724 	proc_id_clear(PROC_ID_GROUP, pgrp->pg_id);
725 	mtx_destroy(&pgrp->pg_mtx);
726 	free(pgrp, M_PGRP);
727 	sess_release(savesess);
728 }
729 
730 static void
731 pgadjustjobc(struct pgrp *pgrp, int entering)
732 {
733 
734 	PGRP_LOCK(pgrp);
735 	if (entering) {
736 #ifdef notyet
737 		MPASS(pgrp->pg_jobc >= 0);
738 #endif
739 		pgrp->pg_jobc++;
740 	} else {
741 #ifdef notyet
742 		MPASS(pgrp->pg_jobc > 0);
743 #endif
744 		--pgrp->pg_jobc;
745 		if (pgrp->pg_jobc == 0)
746 			orphanpg(pgrp);
747 	}
748 	PGRP_UNLOCK(pgrp);
749 }
750 
751 /*
752  * Adjust pgrp jobc counters when specified process changes process group.
753  * We count the number of processes in each process group that "qualify"
754  * the group for terminal job control (those with a parent in a different
755  * process group of the same session).  If that count reaches zero, the
756  * process group becomes orphaned.  Check both the specified process'
757  * process group and that of its children.
758  * entering == 0 => p is leaving specified group.
759  * entering == 1 => p is entering specified group.
760  */
761 void
762 fixjobc(struct proc *p, struct pgrp *pgrp, int entering)
763 {
764 	struct pgrp *hispgrp;
765 	struct session *mysession;
766 	struct proc *q;
767 
768 	sx_assert(&proctree_lock, SX_LOCKED);
769 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
770 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
771 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
772 
773 	/*
774 	 * Check p's parent to see whether p qualifies its own process
775 	 * group; if so, adjust count for p's process group.
776 	 */
777 	mysession = pgrp->pg_session;
778 	if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
779 	    hispgrp->pg_session == mysession)
780 		pgadjustjobc(pgrp, entering);
781 
782 	/*
783 	 * Check this process' children to see whether they qualify
784 	 * their process groups; if so, adjust counts for children's
785 	 * process groups.
786 	 */
787 	LIST_FOREACH(q, &p->p_children, p_sibling) {
788 		hispgrp = q->p_pgrp;
789 		if (hispgrp == pgrp ||
790 		    hispgrp->pg_session != mysession)
791 			continue;
792 		if (q->p_state == PRS_ZOMBIE)
793 			continue;
794 		pgadjustjobc(hispgrp, entering);
795 	}
796 }
797 
798 void
799 killjobc(void)
800 {
801 	struct session *sp;
802 	struct tty *tp;
803 	struct proc *p;
804 	struct vnode *ttyvp;
805 
806 	p = curproc;
807 	MPASS(p->p_flag & P_WEXIT);
808 	/*
809 	 * Do a quick check to see if there is anything to do with the
810 	 * proctree_lock held. pgrp and LIST_EMPTY checks are for fixjobc().
811 	 */
812 	PROC_LOCK(p);
813 	if (!SESS_LEADER(p) &&
814 	    (p->p_pgrp == p->p_pptr->p_pgrp) &&
815 	    LIST_EMPTY(&p->p_children)) {
816 		PROC_UNLOCK(p);
817 		return;
818 	}
819 	PROC_UNLOCK(p);
820 
821 	sx_xlock(&proctree_lock);
822 	if (SESS_LEADER(p)) {
823 		sp = p->p_session;
824 
825 		/*
826 		 * s_ttyp is not zero'd; we use this to indicate that
827 		 * the session once had a controlling terminal. (for
828 		 * logging and informational purposes)
829 		 */
830 		SESS_LOCK(sp);
831 		ttyvp = sp->s_ttyvp;
832 		tp = sp->s_ttyp;
833 		sp->s_ttyvp = NULL;
834 		sp->s_ttydp = NULL;
835 		sp->s_leader = NULL;
836 		SESS_UNLOCK(sp);
837 
838 		/*
839 		 * Signal foreground pgrp and revoke access to
840 		 * controlling terminal if it has not been revoked
841 		 * already.
842 		 *
843 		 * Because the TTY may have been revoked in the mean
844 		 * time and could already have a new session associated
845 		 * with it, make sure we don't send a SIGHUP to a
846 		 * foreground process group that does not belong to this
847 		 * session.
848 		 */
849 
850 		if (tp != NULL) {
851 			tty_lock(tp);
852 			if (tp->t_session == sp)
853 				tty_signal_pgrp(tp, SIGHUP);
854 			tty_unlock(tp);
855 		}
856 
857 		if (ttyvp != NULL) {
858 			sx_xunlock(&proctree_lock);
859 			if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) {
860 				VOP_REVOKE(ttyvp, REVOKEALL);
861 				VOP_UNLOCK(ttyvp);
862 			}
863 			devfs_ctty_unref(ttyvp);
864 			sx_xlock(&proctree_lock);
865 		}
866 	}
867 	fixjobc(p, p->p_pgrp, 0);
868 	sx_xunlock(&proctree_lock);
869 }
870 
871 /*
872  * A process group has become orphaned;
873  * if there are any stopped processes in the group,
874  * hang-up all process in that group.
875  */
876 static void
877 orphanpg(struct pgrp *pg)
878 {
879 	struct proc *p;
880 
881 	PGRP_LOCK_ASSERT(pg, MA_OWNED);
882 
883 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
884 		PROC_LOCK(p);
885 		if (P_SHOULDSTOP(p) == P_STOPPED_SIG) {
886 			PROC_UNLOCK(p);
887 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
888 				PROC_LOCK(p);
889 				kern_psignal(p, SIGHUP);
890 				kern_psignal(p, SIGCONT);
891 				PROC_UNLOCK(p);
892 			}
893 			return;
894 		}
895 		PROC_UNLOCK(p);
896 	}
897 }
898 
899 void
900 sess_hold(struct session *s)
901 {
902 
903 	refcount_acquire(&s->s_count);
904 }
905 
906 void
907 sess_release(struct session *s)
908 {
909 
910 	if (refcount_release(&s->s_count)) {
911 		if (s->s_ttyp != NULL) {
912 			tty_lock(s->s_ttyp);
913 			tty_rel_sess(s->s_ttyp, s);
914 		}
915 		proc_id_clear(PROC_ID_SESSION, s->s_sid);
916 		mtx_destroy(&s->s_mtx);
917 		free(s, M_SESSION);
918 	}
919 }
920 
921 #ifdef DDB
922 
923 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
924 {
925 	struct pgrp *pgrp;
926 	struct proc *p;
927 	int i;
928 
929 	for (i = 0; i <= pgrphash; i++) {
930 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
931 			printf("\tindx %d\n", i);
932 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
933 				printf(
934 			"\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
935 				    (void *)pgrp, (long)pgrp->pg_id,
936 				    (void *)pgrp->pg_session,
937 				    pgrp->pg_session->s_count,
938 				    (void *)LIST_FIRST(&pgrp->pg_members));
939 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
940 					printf("\t\tpid %ld addr %p pgrp %p\n",
941 					    (long)p->p_pid, (void *)p,
942 					    (void *)p->p_pgrp);
943 				}
944 			}
945 		}
946 	}
947 }
948 #endif /* DDB */
949 
950 /*
951  * Calculate the kinfo_proc members which contain process-wide
952  * informations.
953  * Must be called with the target process locked.
954  */
955 static void
956 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
957 {
958 	struct thread *td;
959 
960 	PROC_LOCK_ASSERT(p, MA_OWNED);
961 
962 	kp->ki_estcpu = 0;
963 	kp->ki_pctcpu = 0;
964 	FOREACH_THREAD_IN_PROC(p, td) {
965 		thread_lock(td);
966 		kp->ki_pctcpu += sched_pctcpu(td);
967 		kp->ki_estcpu += sched_estcpu(td);
968 		thread_unlock(td);
969 	}
970 }
971 
972 /*
973  * Clear kinfo_proc and fill in any information that is common
974  * to all threads in the process.
975  * Must be called with the target process locked.
976  */
977 static void
978 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
979 {
980 	struct thread *td0;
981 	struct tty *tp;
982 	struct session *sp;
983 	struct ucred *cred;
984 	struct sigacts *ps;
985 	struct timeval boottime;
986 
987 	PROC_LOCK_ASSERT(p, MA_OWNED);
988 	bzero(kp, sizeof(*kp));
989 
990 	kp->ki_structsize = sizeof(*kp);
991 	kp->ki_paddr = p;
992 	kp->ki_addr =/* p->p_addr; */0; /* XXX */
993 	kp->ki_args = p->p_args;
994 	kp->ki_textvp = p->p_textvp;
995 #ifdef KTRACE
996 	kp->ki_tracep = p->p_tracevp;
997 	kp->ki_traceflag = p->p_traceflag;
998 #endif
999 	kp->ki_fd = p->p_fd;
1000 	kp->ki_vmspace = p->p_vmspace;
1001 	kp->ki_flag = p->p_flag;
1002 	kp->ki_flag2 = p->p_flag2;
1003 	cred = p->p_ucred;
1004 	if (cred) {
1005 		kp->ki_uid = cred->cr_uid;
1006 		kp->ki_ruid = cred->cr_ruid;
1007 		kp->ki_svuid = cred->cr_svuid;
1008 		kp->ki_cr_flags = 0;
1009 		if (cred->cr_flags & CRED_FLAG_CAPMODE)
1010 			kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
1011 		/* XXX bde doesn't like KI_NGROUPS */
1012 		if (cred->cr_ngroups > KI_NGROUPS) {
1013 			kp->ki_ngroups = KI_NGROUPS;
1014 			kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
1015 		} else
1016 			kp->ki_ngroups = cred->cr_ngroups;
1017 		bcopy(cred->cr_groups, kp->ki_groups,
1018 		    kp->ki_ngroups * sizeof(gid_t));
1019 		kp->ki_rgid = cred->cr_rgid;
1020 		kp->ki_svgid = cred->cr_svgid;
1021 		/* If jailed(cred), emulate the old P_JAILED flag. */
1022 		if (jailed(cred)) {
1023 			kp->ki_flag |= P_JAILED;
1024 			/* If inside the jail, use 0 as a jail ID. */
1025 			if (cred->cr_prison != curthread->td_ucred->cr_prison)
1026 				kp->ki_jid = cred->cr_prison->pr_id;
1027 		}
1028 		strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
1029 		    sizeof(kp->ki_loginclass));
1030 	}
1031 	ps = p->p_sigacts;
1032 	if (ps) {
1033 		mtx_lock(&ps->ps_mtx);
1034 		kp->ki_sigignore = ps->ps_sigignore;
1035 		kp->ki_sigcatch = ps->ps_sigcatch;
1036 		mtx_unlock(&ps->ps_mtx);
1037 	}
1038 	if (p->p_state != PRS_NEW &&
1039 	    p->p_state != PRS_ZOMBIE &&
1040 	    p->p_vmspace != NULL) {
1041 		struct vmspace *vm = p->p_vmspace;
1042 
1043 		kp->ki_size = vm->vm_map.size;
1044 		kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
1045 		FOREACH_THREAD_IN_PROC(p, td0) {
1046 			if (!TD_IS_SWAPPED(td0))
1047 				kp->ki_rssize += td0->td_kstack_pages;
1048 		}
1049 		kp->ki_swrss = vm->vm_swrss;
1050 		kp->ki_tsize = vm->vm_tsize;
1051 		kp->ki_dsize = vm->vm_dsize;
1052 		kp->ki_ssize = vm->vm_ssize;
1053 	} else if (p->p_state == PRS_ZOMBIE)
1054 		kp->ki_stat = SZOMB;
1055 	if (kp->ki_flag & P_INMEM)
1056 		kp->ki_sflag = PS_INMEM;
1057 	else
1058 		kp->ki_sflag = 0;
1059 	/* Calculate legacy swtime as seconds since 'swtick'. */
1060 	kp->ki_swtime = (ticks - p->p_swtick) / hz;
1061 	kp->ki_pid = p->p_pid;
1062 	kp->ki_nice = p->p_nice;
1063 	kp->ki_fibnum = p->p_fibnum;
1064 	kp->ki_start = p->p_stats->p_start;
1065 	getboottime(&boottime);
1066 	timevaladd(&kp->ki_start, &boottime);
1067 	PROC_STATLOCK(p);
1068 	rufetch(p, &kp->ki_rusage);
1069 	kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
1070 	calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
1071 	PROC_STATUNLOCK(p);
1072 	calccru(p, &kp->ki_childutime, &kp->ki_childstime);
1073 	/* Some callers want child times in a single value. */
1074 	kp->ki_childtime = kp->ki_childstime;
1075 	timevaladd(&kp->ki_childtime, &kp->ki_childutime);
1076 
1077 	FOREACH_THREAD_IN_PROC(p, td0)
1078 		kp->ki_cow += td0->td_cow;
1079 
1080 	tp = NULL;
1081 	if (p->p_pgrp) {
1082 		kp->ki_pgid = p->p_pgrp->pg_id;
1083 		kp->ki_jobc = p->p_pgrp->pg_jobc;
1084 		sp = p->p_pgrp->pg_session;
1085 
1086 		if (sp != NULL) {
1087 			kp->ki_sid = sp->s_sid;
1088 			SESS_LOCK(sp);
1089 			strlcpy(kp->ki_login, sp->s_login,
1090 			    sizeof(kp->ki_login));
1091 			if (sp->s_ttyvp)
1092 				kp->ki_kiflag |= KI_CTTY;
1093 			if (SESS_LEADER(p))
1094 				kp->ki_kiflag |= KI_SLEADER;
1095 			/* XXX proctree_lock */
1096 			tp = sp->s_ttyp;
1097 			SESS_UNLOCK(sp);
1098 		}
1099 	}
1100 	if ((p->p_flag & P_CONTROLT) && tp != NULL) {
1101 		kp->ki_tdev = tty_udev(tp);
1102 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1103 		kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
1104 		if (tp->t_session)
1105 			kp->ki_tsid = tp->t_session->s_sid;
1106 	} else {
1107 		kp->ki_tdev = NODEV;
1108 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1109 	}
1110 	if (p->p_comm[0] != '\0')
1111 		strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
1112 	if (p->p_sysent && p->p_sysent->sv_name != NULL &&
1113 	    p->p_sysent->sv_name[0] != '\0')
1114 		strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
1115 	kp->ki_siglist = p->p_siglist;
1116 	kp->ki_xstat = KW_EXITCODE(p->p_xexit, p->p_xsig);
1117 	kp->ki_acflag = p->p_acflag;
1118 	kp->ki_lock = p->p_lock;
1119 	if (p->p_pptr) {
1120 		kp->ki_ppid = p->p_oppid;
1121 		if (p->p_flag & P_TRACED)
1122 			kp->ki_tracer = p->p_pptr->p_pid;
1123 	}
1124 }
1125 
1126 /*
1127  * Fill in information that is thread specific.  Must be called with
1128  * target process locked.  If 'preferthread' is set, overwrite certain
1129  * process-related fields that are maintained for both threads and
1130  * processes.
1131  */
1132 static void
1133 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
1134 {
1135 	struct proc *p;
1136 
1137 	p = td->td_proc;
1138 	kp->ki_tdaddr = td;
1139 	PROC_LOCK_ASSERT(p, MA_OWNED);
1140 
1141 	if (preferthread)
1142 		PROC_STATLOCK(p);
1143 	thread_lock(td);
1144 	if (td->td_wmesg != NULL)
1145 		strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
1146 	else
1147 		bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
1148 	if (strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname)) >=
1149 	    sizeof(kp->ki_tdname)) {
1150 		strlcpy(kp->ki_moretdname,
1151 		    td->td_name + sizeof(kp->ki_tdname) - 1,
1152 		    sizeof(kp->ki_moretdname));
1153 	} else {
1154 		bzero(kp->ki_moretdname, sizeof(kp->ki_moretdname));
1155 	}
1156 	if (TD_ON_LOCK(td)) {
1157 		kp->ki_kiflag |= KI_LOCKBLOCK;
1158 		strlcpy(kp->ki_lockname, td->td_lockname,
1159 		    sizeof(kp->ki_lockname));
1160 	} else {
1161 		kp->ki_kiflag &= ~KI_LOCKBLOCK;
1162 		bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
1163 	}
1164 
1165 	if (p->p_state == PRS_NORMAL) { /* approximate. */
1166 		if (TD_ON_RUNQ(td) ||
1167 		    TD_CAN_RUN(td) ||
1168 		    TD_IS_RUNNING(td)) {
1169 			kp->ki_stat = SRUN;
1170 		} else if (P_SHOULDSTOP(p)) {
1171 			kp->ki_stat = SSTOP;
1172 		} else if (TD_IS_SLEEPING(td)) {
1173 			kp->ki_stat = SSLEEP;
1174 		} else if (TD_ON_LOCK(td)) {
1175 			kp->ki_stat = SLOCK;
1176 		} else {
1177 			kp->ki_stat = SWAIT;
1178 		}
1179 	} else if (p->p_state == PRS_ZOMBIE) {
1180 		kp->ki_stat = SZOMB;
1181 	} else {
1182 		kp->ki_stat = SIDL;
1183 	}
1184 
1185 	/* Things in the thread */
1186 	kp->ki_wchan = td->td_wchan;
1187 	kp->ki_pri.pri_level = td->td_priority;
1188 	kp->ki_pri.pri_native = td->td_base_pri;
1189 
1190 	/*
1191 	 * Note: legacy fields; clamp at the old NOCPU value and/or
1192 	 * the maximum u_char CPU value.
1193 	 */
1194 	if (td->td_lastcpu == NOCPU)
1195 		kp->ki_lastcpu_old = NOCPU_OLD;
1196 	else if (td->td_lastcpu > MAXCPU_OLD)
1197 		kp->ki_lastcpu_old = MAXCPU_OLD;
1198 	else
1199 		kp->ki_lastcpu_old = td->td_lastcpu;
1200 
1201 	if (td->td_oncpu == NOCPU)
1202 		kp->ki_oncpu_old = NOCPU_OLD;
1203 	else if (td->td_oncpu > MAXCPU_OLD)
1204 		kp->ki_oncpu_old = MAXCPU_OLD;
1205 	else
1206 		kp->ki_oncpu_old = td->td_oncpu;
1207 
1208 	kp->ki_lastcpu = td->td_lastcpu;
1209 	kp->ki_oncpu = td->td_oncpu;
1210 	kp->ki_tdflags = td->td_flags;
1211 	kp->ki_tid = td->td_tid;
1212 	kp->ki_numthreads = p->p_numthreads;
1213 	kp->ki_pcb = td->td_pcb;
1214 	kp->ki_kstack = (void *)td->td_kstack;
1215 	kp->ki_slptime = (ticks - td->td_slptick) / hz;
1216 	kp->ki_pri.pri_class = td->td_pri_class;
1217 	kp->ki_pri.pri_user = td->td_user_pri;
1218 
1219 	if (preferthread) {
1220 		rufetchtd(td, &kp->ki_rusage);
1221 		kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
1222 		kp->ki_pctcpu = sched_pctcpu(td);
1223 		kp->ki_estcpu = sched_estcpu(td);
1224 		kp->ki_cow = td->td_cow;
1225 	}
1226 
1227 	/* We can't get this anymore but ps etc never used it anyway. */
1228 	kp->ki_rqindex = 0;
1229 
1230 	if (preferthread)
1231 		kp->ki_siglist = td->td_siglist;
1232 	kp->ki_sigmask = td->td_sigmask;
1233 	thread_unlock(td);
1234 	if (preferthread)
1235 		PROC_STATUNLOCK(p);
1236 }
1237 
1238 /*
1239  * Fill in a kinfo_proc structure for the specified process.
1240  * Must be called with the target process locked.
1241  */
1242 void
1243 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1244 {
1245 
1246 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1247 
1248 	fill_kinfo_proc_only(p, kp);
1249 	fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1250 	fill_kinfo_aggregate(p, kp);
1251 }
1252 
1253 struct pstats *
1254 pstats_alloc(void)
1255 {
1256 
1257 	return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1258 }
1259 
1260 /*
1261  * Copy parts of p_stats; zero the rest of p_stats (statistics).
1262  */
1263 void
1264 pstats_fork(struct pstats *src, struct pstats *dst)
1265 {
1266 
1267 	bzero(&dst->pstat_startzero,
1268 	    __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1269 	bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1270 	    __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1271 }
1272 
1273 void
1274 pstats_free(struct pstats *ps)
1275 {
1276 
1277 	free(ps, M_SUBPROC);
1278 }
1279 
1280 #ifdef COMPAT_FREEBSD32
1281 
1282 /*
1283  * This function is typically used to copy out the kernel address, so
1284  * it can be replaced by assignment of zero.
1285  */
1286 static inline uint32_t
1287 ptr32_trim(const void *ptr)
1288 {
1289 	uintptr_t uptr;
1290 
1291 	uptr = (uintptr_t)ptr;
1292 	return ((uptr > UINT_MAX) ? 0 : uptr);
1293 }
1294 
1295 #define PTRTRIM_CP(src,dst,fld) \
1296 	do { (dst).fld = ptr32_trim((src).fld); } while (0)
1297 
1298 static void
1299 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1300 {
1301 	int i;
1302 
1303 	bzero(ki32, sizeof(struct kinfo_proc32));
1304 	ki32->ki_structsize = sizeof(struct kinfo_proc32);
1305 	CP(*ki, *ki32, ki_layout);
1306 	PTRTRIM_CP(*ki, *ki32, ki_args);
1307 	PTRTRIM_CP(*ki, *ki32, ki_paddr);
1308 	PTRTRIM_CP(*ki, *ki32, ki_addr);
1309 	PTRTRIM_CP(*ki, *ki32, ki_tracep);
1310 	PTRTRIM_CP(*ki, *ki32, ki_textvp);
1311 	PTRTRIM_CP(*ki, *ki32, ki_fd);
1312 	PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1313 	PTRTRIM_CP(*ki, *ki32, ki_wchan);
1314 	CP(*ki, *ki32, ki_pid);
1315 	CP(*ki, *ki32, ki_ppid);
1316 	CP(*ki, *ki32, ki_pgid);
1317 	CP(*ki, *ki32, ki_tpgid);
1318 	CP(*ki, *ki32, ki_sid);
1319 	CP(*ki, *ki32, ki_tsid);
1320 	CP(*ki, *ki32, ki_jobc);
1321 	CP(*ki, *ki32, ki_tdev);
1322 	CP(*ki, *ki32, ki_tdev_freebsd11);
1323 	CP(*ki, *ki32, ki_siglist);
1324 	CP(*ki, *ki32, ki_sigmask);
1325 	CP(*ki, *ki32, ki_sigignore);
1326 	CP(*ki, *ki32, ki_sigcatch);
1327 	CP(*ki, *ki32, ki_uid);
1328 	CP(*ki, *ki32, ki_ruid);
1329 	CP(*ki, *ki32, ki_svuid);
1330 	CP(*ki, *ki32, ki_rgid);
1331 	CP(*ki, *ki32, ki_svgid);
1332 	CP(*ki, *ki32, ki_ngroups);
1333 	for (i = 0; i < KI_NGROUPS; i++)
1334 		CP(*ki, *ki32, ki_groups[i]);
1335 	CP(*ki, *ki32, ki_size);
1336 	CP(*ki, *ki32, ki_rssize);
1337 	CP(*ki, *ki32, ki_swrss);
1338 	CP(*ki, *ki32, ki_tsize);
1339 	CP(*ki, *ki32, ki_dsize);
1340 	CP(*ki, *ki32, ki_ssize);
1341 	CP(*ki, *ki32, ki_xstat);
1342 	CP(*ki, *ki32, ki_acflag);
1343 	CP(*ki, *ki32, ki_pctcpu);
1344 	CP(*ki, *ki32, ki_estcpu);
1345 	CP(*ki, *ki32, ki_slptime);
1346 	CP(*ki, *ki32, ki_swtime);
1347 	CP(*ki, *ki32, ki_cow);
1348 	CP(*ki, *ki32, ki_runtime);
1349 	TV_CP(*ki, *ki32, ki_start);
1350 	TV_CP(*ki, *ki32, ki_childtime);
1351 	CP(*ki, *ki32, ki_flag);
1352 	CP(*ki, *ki32, ki_kiflag);
1353 	CP(*ki, *ki32, ki_traceflag);
1354 	CP(*ki, *ki32, ki_stat);
1355 	CP(*ki, *ki32, ki_nice);
1356 	CP(*ki, *ki32, ki_lock);
1357 	CP(*ki, *ki32, ki_rqindex);
1358 	CP(*ki, *ki32, ki_oncpu);
1359 	CP(*ki, *ki32, ki_lastcpu);
1360 
1361 	/* XXX TODO: wrap cpu value as appropriate */
1362 	CP(*ki, *ki32, ki_oncpu_old);
1363 	CP(*ki, *ki32, ki_lastcpu_old);
1364 
1365 	bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1366 	bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1367 	bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1368 	bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1369 	bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1370 	bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1371 	bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1372 	bcopy(ki->ki_moretdname, ki32->ki_moretdname, MAXCOMLEN - TDNAMLEN + 1);
1373 	CP(*ki, *ki32, ki_tracer);
1374 	CP(*ki, *ki32, ki_flag2);
1375 	CP(*ki, *ki32, ki_fibnum);
1376 	CP(*ki, *ki32, ki_cr_flags);
1377 	CP(*ki, *ki32, ki_jid);
1378 	CP(*ki, *ki32, ki_numthreads);
1379 	CP(*ki, *ki32, ki_tid);
1380 	CP(*ki, *ki32, ki_pri);
1381 	freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1382 	freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1383 	PTRTRIM_CP(*ki, *ki32, ki_pcb);
1384 	PTRTRIM_CP(*ki, *ki32, ki_kstack);
1385 	PTRTRIM_CP(*ki, *ki32, ki_udata);
1386 	PTRTRIM_CP(*ki, *ki32, ki_tdaddr);
1387 	CP(*ki, *ki32, ki_sflag);
1388 	CP(*ki, *ki32, ki_tdflags);
1389 }
1390 #endif
1391 
1392 static ssize_t
1393 kern_proc_out_size(struct proc *p, int flags)
1394 {
1395 	ssize_t size = 0;
1396 
1397 	PROC_LOCK_ASSERT(p, MA_OWNED);
1398 
1399 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1400 #ifdef COMPAT_FREEBSD32
1401 		if ((flags & KERN_PROC_MASK32) != 0) {
1402 			size += sizeof(struct kinfo_proc32);
1403 		} else
1404 #endif
1405 			size += sizeof(struct kinfo_proc);
1406 	} else {
1407 #ifdef COMPAT_FREEBSD32
1408 		if ((flags & KERN_PROC_MASK32) != 0)
1409 			size += sizeof(struct kinfo_proc32) * p->p_numthreads;
1410 		else
1411 #endif
1412 			size += sizeof(struct kinfo_proc) * p->p_numthreads;
1413 	}
1414 	PROC_UNLOCK(p);
1415 	return (size);
1416 }
1417 
1418 int
1419 kern_proc_out(struct proc *p, struct sbuf *sb, int flags)
1420 {
1421 	struct thread *td;
1422 	struct kinfo_proc ki;
1423 #ifdef COMPAT_FREEBSD32
1424 	struct kinfo_proc32 ki32;
1425 #endif
1426 	int error;
1427 
1428 	PROC_LOCK_ASSERT(p, MA_OWNED);
1429 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1430 
1431 	error = 0;
1432 	fill_kinfo_proc(p, &ki);
1433 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1434 #ifdef COMPAT_FREEBSD32
1435 		if ((flags & KERN_PROC_MASK32) != 0) {
1436 			freebsd32_kinfo_proc_out(&ki, &ki32);
1437 			if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1438 				error = ENOMEM;
1439 		} else
1440 #endif
1441 			if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1442 				error = ENOMEM;
1443 	} else {
1444 		FOREACH_THREAD_IN_PROC(p, td) {
1445 			fill_kinfo_thread(td, &ki, 1);
1446 #ifdef COMPAT_FREEBSD32
1447 			if ((flags & KERN_PROC_MASK32) != 0) {
1448 				freebsd32_kinfo_proc_out(&ki, &ki32);
1449 				if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1450 					error = ENOMEM;
1451 			} else
1452 #endif
1453 				if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1454 					error = ENOMEM;
1455 			if (error != 0)
1456 				break;
1457 		}
1458 	}
1459 	PROC_UNLOCK(p);
1460 	return (error);
1461 }
1462 
1463 static int
1464 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
1465 {
1466 	struct sbuf sb;
1467 	struct kinfo_proc ki;
1468 	int error, error2;
1469 
1470 	if (req->oldptr == NULL)
1471 		return (SYSCTL_OUT(req, 0, kern_proc_out_size(p, flags)));
1472 
1473 	sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req);
1474 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1475 	error = kern_proc_out(p, &sb, flags);
1476 	error2 = sbuf_finish(&sb);
1477 	sbuf_delete(&sb);
1478 	if (error != 0)
1479 		return (error);
1480 	else if (error2 != 0)
1481 		return (error2);
1482 	return (0);
1483 }
1484 
1485 int
1486 proc_iterate(int (*cb)(struct proc *, void *), void *cbarg)
1487 {
1488 	struct proc *p;
1489 	int error, i, j;
1490 
1491 	for (i = 0; i < pidhashlock + 1; i++) {
1492 		sx_slock(&pidhashtbl_lock[i]);
1493 		for (j = i; j <= pidhash; j += pidhashlock + 1) {
1494 			LIST_FOREACH(p, &pidhashtbl[j], p_hash) {
1495 				if (p->p_state == PRS_NEW)
1496 					continue;
1497 				error = cb(p, cbarg);
1498 				PROC_LOCK_ASSERT(p, MA_NOTOWNED);
1499 				if (error != 0) {
1500 					sx_sunlock(&pidhashtbl_lock[i]);
1501 					return (error);
1502 				}
1503 			}
1504 		}
1505 		sx_sunlock(&pidhashtbl_lock[i]);
1506 	}
1507 	return (0);
1508 }
1509 
1510 struct kern_proc_out_args {
1511 	struct sysctl_req *req;
1512 	int flags;
1513 	int oid_number;
1514 	int *name;
1515 };
1516 
1517 static int
1518 sysctl_kern_proc_iterate(struct proc *p, void *origarg)
1519 {
1520 	struct kern_proc_out_args *arg = origarg;
1521 	int *name = arg->name;
1522 	int oid_number = arg->oid_number;
1523 	int flags = arg->flags;
1524 	struct sysctl_req *req = arg->req;
1525 	int error = 0;
1526 
1527 	PROC_LOCK(p);
1528 
1529 	KASSERT(p->p_ucred != NULL,
1530 	    ("process credential is NULL for non-NEW proc"));
1531 	/*
1532 	 * Show a user only appropriate processes.
1533 	 */
1534 	if (p_cansee(curthread, p))
1535 		goto skip;
1536 	/*
1537 	 * TODO - make more efficient (see notes below).
1538 	 * do by session.
1539 	 */
1540 	switch (oid_number) {
1541 
1542 	case KERN_PROC_GID:
1543 		if (p->p_ucred->cr_gid != (gid_t)name[0])
1544 			goto skip;
1545 		break;
1546 
1547 	case KERN_PROC_PGRP:
1548 		/* could do this by traversing pgrp */
1549 		if (p->p_pgrp == NULL ||
1550 		    p->p_pgrp->pg_id != (pid_t)name[0])
1551 			goto skip;
1552 		break;
1553 
1554 	case KERN_PROC_RGID:
1555 		if (p->p_ucred->cr_rgid != (gid_t)name[0])
1556 			goto skip;
1557 		break;
1558 
1559 	case KERN_PROC_SESSION:
1560 		if (p->p_session == NULL ||
1561 		    p->p_session->s_sid != (pid_t)name[0])
1562 			goto skip;
1563 		break;
1564 
1565 	case KERN_PROC_TTY:
1566 		if ((p->p_flag & P_CONTROLT) == 0 ||
1567 		    p->p_session == NULL)
1568 			goto skip;
1569 		/* XXX proctree_lock */
1570 		SESS_LOCK(p->p_session);
1571 		if (p->p_session->s_ttyp == NULL ||
1572 		    tty_udev(p->p_session->s_ttyp) !=
1573 		    (dev_t)name[0]) {
1574 			SESS_UNLOCK(p->p_session);
1575 			goto skip;
1576 		}
1577 		SESS_UNLOCK(p->p_session);
1578 		break;
1579 
1580 	case KERN_PROC_UID:
1581 		if (p->p_ucred->cr_uid != (uid_t)name[0])
1582 			goto skip;
1583 		break;
1584 
1585 	case KERN_PROC_RUID:
1586 		if (p->p_ucred->cr_ruid != (uid_t)name[0])
1587 			goto skip;
1588 		break;
1589 
1590 	case KERN_PROC_PROC:
1591 		break;
1592 
1593 	default:
1594 		break;
1595 
1596 	}
1597 	error = sysctl_out_proc(p, req, flags);
1598 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
1599 	return (error);
1600 skip:
1601 	PROC_UNLOCK(p);
1602 	return (0);
1603 }
1604 
1605 static int
1606 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1607 {
1608 	struct kern_proc_out_args iterarg;
1609 	int *name = (int *)arg1;
1610 	u_int namelen = arg2;
1611 	struct proc *p;
1612 	int flags, oid_number;
1613 	int error = 0;
1614 
1615 	oid_number = oidp->oid_number;
1616 	if (oid_number != KERN_PROC_ALL &&
1617 	    (oid_number & KERN_PROC_INC_THREAD) == 0)
1618 		flags = KERN_PROC_NOTHREADS;
1619 	else {
1620 		flags = 0;
1621 		oid_number &= ~KERN_PROC_INC_THREAD;
1622 	}
1623 #ifdef COMPAT_FREEBSD32
1624 	if (req->flags & SCTL_MASK32)
1625 		flags |= KERN_PROC_MASK32;
1626 #endif
1627 	if (oid_number == KERN_PROC_PID) {
1628 		if (namelen != 1)
1629 			return (EINVAL);
1630 		error = sysctl_wire_old_buffer(req, 0);
1631 		if (error)
1632 			return (error);
1633 		error = pget((pid_t)name[0], PGET_CANSEE, &p);
1634 		if (error == 0)
1635 			error = sysctl_out_proc(p, req, flags);
1636 		return (error);
1637 	}
1638 
1639 	switch (oid_number) {
1640 	case KERN_PROC_ALL:
1641 		if (namelen != 0)
1642 			return (EINVAL);
1643 		break;
1644 	case KERN_PROC_PROC:
1645 		if (namelen != 0 && namelen != 1)
1646 			return (EINVAL);
1647 		break;
1648 	default:
1649 		if (namelen != 1)
1650 			return (EINVAL);
1651 		break;
1652 	}
1653 
1654 	if (req->oldptr == NULL) {
1655 		/* overestimate by 5 procs */
1656 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1657 		if (error)
1658 			return (error);
1659 	} else {
1660 		error = sysctl_wire_old_buffer(req, 0);
1661 		if (error != 0)
1662 			return (error);
1663 	}
1664 	iterarg.flags = flags;
1665 	iterarg.oid_number = oid_number;
1666 	iterarg.req = req;
1667 	iterarg.name = name;
1668 	error = proc_iterate(sysctl_kern_proc_iterate, &iterarg);
1669 	return (error);
1670 }
1671 
1672 struct pargs *
1673 pargs_alloc(int len)
1674 {
1675 	struct pargs *pa;
1676 
1677 	pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1678 		M_WAITOK);
1679 	refcount_init(&pa->ar_ref, 1);
1680 	pa->ar_length = len;
1681 	return (pa);
1682 }
1683 
1684 static void
1685 pargs_free(struct pargs *pa)
1686 {
1687 
1688 	free(pa, M_PARGS);
1689 }
1690 
1691 void
1692 pargs_hold(struct pargs *pa)
1693 {
1694 
1695 	if (pa == NULL)
1696 		return;
1697 	refcount_acquire(&pa->ar_ref);
1698 }
1699 
1700 void
1701 pargs_drop(struct pargs *pa)
1702 {
1703 
1704 	if (pa == NULL)
1705 		return;
1706 	if (refcount_release(&pa->ar_ref))
1707 		pargs_free(pa);
1708 }
1709 
1710 static int
1711 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1712     size_t len)
1713 {
1714 	ssize_t n;
1715 
1716 	/*
1717 	 * This may return a short read if the string is shorter than the chunk
1718 	 * and is aligned at the end of the page, and the following page is not
1719 	 * mapped.
1720 	 */
1721 	n = proc_readmem(td, p, (vm_offset_t)sptr, buf, len);
1722 	if (n <= 0)
1723 		return (ENOMEM);
1724 	return (0);
1725 }
1726 
1727 #define PROC_AUXV_MAX	256	/* Safety limit on auxv size. */
1728 
1729 enum proc_vector_type {
1730 	PROC_ARG,
1731 	PROC_ENV,
1732 	PROC_AUX,
1733 };
1734 
1735 #ifdef COMPAT_FREEBSD32
1736 static int
1737 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1738     size_t *vsizep, enum proc_vector_type type)
1739 {
1740 	struct freebsd32_ps_strings pss;
1741 	Elf32_Auxinfo aux;
1742 	vm_offset_t vptr, ptr;
1743 	uint32_t *proc_vector32;
1744 	char **proc_vector;
1745 	size_t vsize, size;
1746 	int i, error;
1747 
1748 	error = 0;
1749 	if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss,
1750 	    sizeof(pss)) != sizeof(pss))
1751 		return (ENOMEM);
1752 	switch (type) {
1753 	case PROC_ARG:
1754 		vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1755 		vsize = pss.ps_nargvstr;
1756 		if (vsize > ARG_MAX)
1757 			return (ENOEXEC);
1758 		size = vsize * sizeof(int32_t);
1759 		break;
1760 	case PROC_ENV:
1761 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1762 		vsize = pss.ps_nenvstr;
1763 		if (vsize > ARG_MAX)
1764 			return (ENOEXEC);
1765 		size = vsize * sizeof(int32_t);
1766 		break;
1767 	case PROC_AUX:
1768 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1769 		    (pss.ps_nenvstr + 1) * sizeof(int32_t);
1770 		if (vptr % 4 != 0)
1771 			return (ENOEXEC);
1772 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1773 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1774 			    sizeof(aux))
1775 				return (ENOMEM);
1776 			if (aux.a_type == AT_NULL)
1777 				break;
1778 			ptr += sizeof(aux);
1779 		}
1780 		if (aux.a_type != AT_NULL)
1781 			return (ENOEXEC);
1782 		vsize = i + 1;
1783 		size = vsize * sizeof(aux);
1784 		break;
1785 	default:
1786 		KASSERT(0, ("Wrong proc vector type: %d", type));
1787 		return (EINVAL);
1788 	}
1789 	proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1790 	if (proc_readmem(td, p, vptr, proc_vector32, size) != size) {
1791 		error = ENOMEM;
1792 		goto done;
1793 	}
1794 	if (type == PROC_AUX) {
1795 		*proc_vectorp = (char **)proc_vector32;
1796 		*vsizep = vsize;
1797 		return (0);
1798 	}
1799 	proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1800 	for (i = 0; i < (int)vsize; i++)
1801 		proc_vector[i] = PTRIN(proc_vector32[i]);
1802 	*proc_vectorp = proc_vector;
1803 	*vsizep = vsize;
1804 done:
1805 	free(proc_vector32, M_TEMP);
1806 	return (error);
1807 }
1808 #endif
1809 
1810 static int
1811 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1812     size_t *vsizep, enum proc_vector_type type)
1813 {
1814 	struct ps_strings pss;
1815 	Elf_Auxinfo aux;
1816 	vm_offset_t vptr, ptr;
1817 	char **proc_vector;
1818 	size_t vsize, size;
1819 	int i;
1820 
1821 #ifdef COMPAT_FREEBSD32
1822 	if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1823 		return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1824 #endif
1825 	if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss,
1826 	    sizeof(pss)) != sizeof(pss))
1827 		return (ENOMEM);
1828 	switch (type) {
1829 	case PROC_ARG:
1830 		vptr = (vm_offset_t)pss.ps_argvstr;
1831 		vsize = pss.ps_nargvstr;
1832 		if (vsize > ARG_MAX)
1833 			return (ENOEXEC);
1834 		size = vsize * sizeof(char *);
1835 		break;
1836 	case PROC_ENV:
1837 		vptr = (vm_offset_t)pss.ps_envstr;
1838 		vsize = pss.ps_nenvstr;
1839 		if (vsize > ARG_MAX)
1840 			return (ENOEXEC);
1841 		size = vsize * sizeof(char *);
1842 		break;
1843 	case PROC_AUX:
1844 		/*
1845 		 * The aux array is just above env array on the stack. Check
1846 		 * that the address is naturally aligned.
1847 		 */
1848 		vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1849 		    * sizeof(char *);
1850 #if __ELF_WORD_SIZE == 64
1851 		if (vptr % sizeof(uint64_t) != 0)
1852 #else
1853 		if (vptr % sizeof(uint32_t) != 0)
1854 #endif
1855 			return (ENOEXEC);
1856 		/*
1857 		 * We count the array size reading the aux vectors from the
1858 		 * stack until AT_NULL vector is returned.  So (to keep the code
1859 		 * simple) we read the process stack twice: the first time here
1860 		 * to find the size and the second time when copying the vectors
1861 		 * to the allocated proc_vector.
1862 		 */
1863 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1864 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1865 			    sizeof(aux))
1866 				return (ENOMEM);
1867 			if (aux.a_type == AT_NULL)
1868 				break;
1869 			ptr += sizeof(aux);
1870 		}
1871 		/*
1872 		 * If the PROC_AUXV_MAX entries are iterated over, and we have
1873 		 * not reached AT_NULL, it is most likely we are reading wrong
1874 		 * data: either the process doesn't have auxv array or data has
1875 		 * been modified. Return the error in this case.
1876 		 */
1877 		if (aux.a_type != AT_NULL)
1878 			return (ENOEXEC);
1879 		vsize = i + 1;
1880 		size = vsize * sizeof(aux);
1881 		break;
1882 	default:
1883 		KASSERT(0, ("Wrong proc vector type: %d", type));
1884 		return (EINVAL); /* In case we are built without INVARIANTS. */
1885 	}
1886 	proc_vector = malloc(size, M_TEMP, M_WAITOK);
1887 	if (proc_readmem(td, p, vptr, proc_vector, size) != size) {
1888 		free(proc_vector, M_TEMP);
1889 		return (ENOMEM);
1890 	}
1891 	*proc_vectorp = proc_vector;
1892 	*vsizep = vsize;
1893 
1894 	return (0);
1895 }
1896 
1897 #define GET_PS_STRINGS_CHUNK_SZ	256	/* Chunk size (bytes) for ps_strings operations. */
1898 
1899 static int
1900 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
1901     enum proc_vector_type type)
1902 {
1903 	size_t done, len, nchr, vsize;
1904 	int error, i;
1905 	char **proc_vector, *sptr;
1906 	char pss_string[GET_PS_STRINGS_CHUNK_SZ];
1907 
1908 	PROC_ASSERT_HELD(p);
1909 
1910 	/*
1911 	 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
1912 	 */
1913 	nchr = 2 * (PATH_MAX + ARG_MAX);
1914 
1915 	error = get_proc_vector(td, p, &proc_vector, &vsize, type);
1916 	if (error != 0)
1917 		return (error);
1918 	for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
1919 		/*
1920 		 * The program may have scribbled into its argv array, e.g. to
1921 		 * remove some arguments.  If that has happened, break out
1922 		 * before trying to read from NULL.
1923 		 */
1924 		if (proc_vector[i] == NULL)
1925 			break;
1926 		for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
1927 			error = proc_read_string(td, p, sptr, pss_string,
1928 			    sizeof(pss_string));
1929 			if (error != 0)
1930 				goto done;
1931 			len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
1932 			if (done + len >= nchr)
1933 				len = nchr - done - 1;
1934 			sbuf_bcat(sb, pss_string, len);
1935 			if (len != GET_PS_STRINGS_CHUNK_SZ)
1936 				break;
1937 			done += GET_PS_STRINGS_CHUNK_SZ;
1938 		}
1939 		sbuf_bcat(sb, "", 1);
1940 		done += len + 1;
1941 	}
1942 done:
1943 	free(proc_vector, M_TEMP);
1944 	return (error);
1945 }
1946 
1947 int
1948 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
1949 {
1950 
1951 	return (get_ps_strings(curthread, p, sb, PROC_ARG));
1952 }
1953 
1954 int
1955 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
1956 {
1957 
1958 	return (get_ps_strings(curthread, p, sb, PROC_ENV));
1959 }
1960 
1961 int
1962 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb)
1963 {
1964 	size_t vsize, size;
1965 	char **auxv;
1966 	int error;
1967 
1968 	error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX);
1969 	if (error == 0) {
1970 #ifdef COMPAT_FREEBSD32
1971 		if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1972 			size = vsize * sizeof(Elf32_Auxinfo);
1973 		else
1974 #endif
1975 			size = vsize * sizeof(Elf_Auxinfo);
1976 		if (sbuf_bcat(sb, auxv, size) != 0)
1977 			error = ENOMEM;
1978 		free(auxv, M_TEMP);
1979 	}
1980 	return (error);
1981 }
1982 
1983 /*
1984  * This sysctl allows a process to retrieve the argument list or process
1985  * title for another process without groping around in the address space
1986  * of the other process.  It also allow a process to set its own "process
1987  * title to a string of its own choice.
1988  */
1989 static int
1990 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
1991 {
1992 	int *name = (int *)arg1;
1993 	u_int namelen = arg2;
1994 	struct pargs *newpa, *pa;
1995 	struct proc *p;
1996 	struct sbuf sb;
1997 	int flags, error = 0, error2;
1998 	pid_t pid;
1999 
2000 	if (namelen != 1)
2001 		return (EINVAL);
2002 
2003 	pid = (pid_t)name[0];
2004 	/*
2005 	 * If the query is for this process and it is single-threaded, there
2006 	 * is nobody to modify pargs, thus we can just read.
2007 	 */
2008 	p = curproc;
2009 	if (pid == p->p_pid && p->p_numthreads == 1 && req->newptr == NULL &&
2010 	    (pa = p->p_args) != NULL)
2011 		return (SYSCTL_OUT(req, pa->ar_args, pa->ar_length));
2012 
2013 	flags = PGET_CANSEE;
2014 	if (req->newptr != NULL)
2015 		flags |= PGET_ISCURRENT;
2016 	error = pget(pid, flags, &p);
2017 	if (error)
2018 		return (error);
2019 
2020 	pa = p->p_args;
2021 	if (pa != NULL) {
2022 		pargs_hold(pa);
2023 		PROC_UNLOCK(p);
2024 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
2025 		pargs_drop(pa);
2026 	} else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
2027 		_PHOLD(p);
2028 		PROC_UNLOCK(p);
2029 		sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2030 		sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2031 		error = proc_getargv(curthread, p, &sb);
2032 		error2 = sbuf_finish(&sb);
2033 		PRELE(p);
2034 		sbuf_delete(&sb);
2035 		if (error == 0 && error2 != 0)
2036 			error = error2;
2037 	} else {
2038 		PROC_UNLOCK(p);
2039 	}
2040 	if (error != 0 || req->newptr == NULL)
2041 		return (error);
2042 
2043 	if (req->newlen > ps_arg_cache_limit - sizeof(struct pargs))
2044 		return (ENOMEM);
2045 
2046 	if (req->newlen == 0) {
2047 		/*
2048 		 * Clear the argument pointer, so that we'll fetch arguments
2049 		 * with proc_getargv() until further notice.
2050 		 */
2051 		newpa = NULL;
2052 	} else {
2053 		newpa = pargs_alloc(req->newlen);
2054 		error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
2055 		if (error != 0) {
2056 			pargs_free(newpa);
2057 			return (error);
2058 		}
2059 	}
2060 	PROC_LOCK(p);
2061 	pa = p->p_args;
2062 	p->p_args = newpa;
2063 	PROC_UNLOCK(p);
2064 	pargs_drop(pa);
2065 	return (0);
2066 }
2067 
2068 /*
2069  * This sysctl allows a process to retrieve environment of another process.
2070  */
2071 static int
2072 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
2073 {
2074 	int *name = (int *)arg1;
2075 	u_int namelen = arg2;
2076 	struct proc *p;
2077 	struct sbuf sb;
2078 	int error, error2;
2079 
2080 	if (namelen != 1)
2081 		return (EINVAL);
2082 
2083 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2084 	if (error != 0)
2085 		return (error);
2086 	if ((p->p_flag & P_SYSTEM) != 0) {
2087 		PRELE(p);
2088 		return (0);
2089 	}
2090 
2091 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2092 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2093 	error = proc_getenvv(curthread, p, &sb);
2094 	error2 = sbuf_finish(&sb);
2095 	PRELE(p);
2096 	sbuf_delete(&sb);
2097 	return (error != 0 ? error : error2);
2098 }
2099 
2100 /*
2101  * This sysctl allows a process to retrieve ELF auxiliary vector of
2102  * another process.
2103  */
2104 static int
2105 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
2106 {
2107 	int *name = (int *)arg1;
2108 	u_int namelen = arg2;
2109 	struct proc *p;
2110 	struct sbuf sb;
2111 	int error, error2;
2112 
2113 	if (namelen != 1)
2114 		return (EINVAL);
2115 
2116 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2117 	if (error != 0)
2118 		return (error);
2119 	if ((p->p_flag & P_SYSTEM) != 0) {
2120 		PRELE(p);
2121 		return (0);
2122 	}
2123 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2124 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2125 	error = proc_getauxv(curthread, p, &sb);
2126 	error2 = sbuf_finish(&sb);
2127 	PRELE(p);
2128 	sbuf_delete(&sb);
2129 	return (error != 0 ? error : error2);
2130 }
2131 
2132 /*
2133  * This sysctl allows a process to retrieve the path of the executable for
2134  * itself or another process.
2135  */
2136 static int
2137 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
2138 {
2139 	pid_t *pidp = (pid_t *)arg1;
2140 	unsigned int arglen = arg2;
2141 	struct proc *p;
2142 	struct vnode *vp;
2143 	char *retbuf, *freebuf;
2144 	int error;
2145 
2146 	if (arglen != 1)
2147 		return (EINVAL);
2148 	if (*pidp == -1) {	/* -1 means this process */
2149 		p = req->td->td_proc;
2150 	} else {
2151 		error = pget(*pidp, PGET_CANSEE, &p);
2152 		if (error != 0)
2153 			return (error);
2154 	}
2155 
2156 	vp = p->p_textvp;
2157 	if (vp == NULL) {
2158 		if (*pidp != -1)
2159 			PROC_UNLOCK(p);
2160 		return (0);
2161 	}
2162 	vref(vp);
2163 	if (*pidp != -1)
2164 		PROC_UNLOCK(p);
2165 	error = vn_fullpath(req->td, vp, &retbuf, &freebuf);
2166 	vrele(vp);
2167 	if (error)
2168 		return (error);
2169 	error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
2170 	free(freebuf, M_TEMP);
2171 	return (error);
2172 }
2173 
2174 static int
2175 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
2176 {
2177 	struct proc *p;
2178 	char *sv_name;
2179 	int *name;
2180 	int namelen;
2181 	int error;
2182 
2183 	namelen = arg2;
2184 	if (namelen != 1)
2185 		return (EINVAL);
2186 
2187 	name = (int *)arg1;
2188 	error = pget((pid_t)name[0], PGET_CANSEE, &p);
2189 	if (error != 0)
2190 		return (error);
2191 	sv_name = p->p_sysent->sv_name;
2192 	PROC_UNLOCK(p);
2193 	return (sysctl_handle_string(oidp, sv_name, 0, req));
2194 }
2195 
2196 #ifdef KINFO_OVMENTRY_SIZE
2197 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
2198 #endif
2199 
2200 #ifdef COMPAT_FREEBSD7
2201 static int
2202 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
2203 {
2204 	vm_map_entry_t entry, tmp_entry;
2205 	unsigned int last_timestamp;
2206 	char *fullpath, *freepath;
2207 	struct kinfo_ovmentry *kve;
2208 	struct vattr va;
2209 	struct ucred *cred;
2210 	int error, *name;
2211 	struct vnode *vp;
2212 	struct proc *p;
2213 	vm_map_t map;
2214 	struct vmspace *vm;
2215 
2216 	name = (int *)arg1;
2217 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2218 	if (error != 0)
2219 		return (error);
2220 	vm = vmspace_acquire_ref(p);
2221 	if (vm == NULL) {
2222 		PRELE(p);
2223 		return (ESRCH);
2224 	}
2225 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2226 
2227 	map = &vm->vm_map;
2228 	vm_map_lock_read(map);
2229 	VM_MAP_ENTRY_FOREACH(entry, map) {
2230 		vm_object_t obj, tobj, lobj;
2231 		vm_offset_t addr;
2232 
2233 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2234 			continue;
2235 
2236 		bzero(kve, sizeof(*kve));
2237 		kve->kve_structsize = sizeof(*kve);
2238 
2239 		kve->kve_private_resident = 0;
2240 		obj = entry->object.vm_object;
2241 		if (obj != NULL) {
2242 			VM_OBJECT_RLOCK(obj);
2243 			if (obj->shadow_count == 1)
2244 				kve->kve_private_resident =
2245 				    obj->resident_page_count;
2246 		}
2247 		kve->kve_resident = 0;
2248 		addr = entry->start;
2249 		while (addr < entry->end) {
2250 			if (pmap_extract(map->pmap, addr))
2251 				kve->kve_resident++;
2252 			addr += PAGE_SIZE;
2253 		}
2254 
2255 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2256 			if (tobj != obj) {
2257 				VM_OBJECT_RLOCK(tobj);
2258 				kve->kve_offset += tobj->backing_object_offset;
2259 			}
2260 			if (lobj != obj)
2261 				VM_OBJECT_RUNLOCK(lobj);
2262 			lobj = tobj;
2263 		}
2264 
2265 		kve->kve_start = (void*)entry->start;
2266 		kve->kve_end = (void*)entry->end;
2267 		kve->kve_offset += (off_t)entry->offset;
2268 
2269 		if (entry->protection & VM_PROT_READ)
2270 			kve->kve_protection |= KVME_PROT_READ;
2271 		if (entry->protection & VM_PROT_WRITE)
2272 			kve->kve_protection |= KVME_PROT_WRITE;
2273 		if (entry->protection & VM_PROT_EXECUTE)
2274 			kve->kve_protection |= KVME_PROT_EXEC;
2275 
2276 		if (entry->eflags & MAP_ENTRY_COW)
2277 			kve->kve_flags |= KVME_FLAG_COW;
2278 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2279 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2280 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2281 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2282 
2283 		last_timestamp = map->timestamp;
2284 		vm_map_unlock_read(map);
2285 
2286 		kve->kve_fileid = 0;
2287 		kve->kve_fsid = 0;
2288 		freepath = NULL;
2289 		fullpath = "";
2290 		if (lobj) {
2291 			kve->kve_type = vm_object_kvme_type(lobj, &vp);
2292 			if (kve->kve_type == KVME_TYPE_MGTDEVICE)
2293 				kve->kve_type = KVME_TYPE_UNKNOWN;
2294 			if (vp != NULL)
2295 				vref(vp);
2296 			if (lobj != obj)
2297 				VM_OBJECT_RUNLOCK(lobj);
2298 
2299 			kve->kve_ref_count = obj->ref_count;
2300 			kve->kve_shadow_count = obj->shadow_count;
2301 			VM_OBJECT_RUNLOCK(obj);
2302 			if (vp != NULL) {
2303 				vn_fullpath(curthread, vp, &fullpath,
2304 				    &freepath);
2305 				cred = curthread->td_ucred;
2306 				vn_lock(vp, LK_SHARED | LK_RETRY);
2307 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2308 					kve->kve_fileid = va.va_fileid;
2309 					/* truncate */
2310 					kve->kve_fsid = va.va_fsid;
2311 				}
2312 				vput(vp);
2313 			}
2314 		} else {
2315 			kve->kve_type = KVME_TYPE_NONE;
2316 			kve->kve_ref_count = 0;
2317 			kve->kve_shadow_count = 0;
2318 		}
2319 
2320 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2321 		if (freepath != NULL)
2322 			free(freepath, M_TEMP);
2323 
2324 		error = SYSCTL_OUT(req, kve, sizeof(*kve));
2325 		vm_map_lock_read(map);
2326 		if (error)
2327 			break;
2328 		if (last_timestamp != map->timestamp) {
2329 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2330 			entry = tmp_entry;
2331 		}
2332 	}
2333 	vm_map_unlock_read(map);
2334 	vmspace_free(vm);
2335 	PRELE(p);
2336 	free(kve, M_TEMP);
2337 	return (error);
2338 }
2339 #endif	/* COMPAT_FREEBSD7 */
2340 
2341 #ifdef KINFO_VMENTRY_SIZE
2342 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2343 #endif
2344 
2345 void
2346 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry,
2347     int *resident_count, bool *super)
2348 {
2349 	vm_object_t obj, tobj;
2350 	vm_page_t m, m_adv;
2351 	vm_offset_t addr;
2352 	vm_paddr_t pa;
2353 	vm_pindex_t pi, pi_adv, pindex;
2354 
2355 	*super = false;
2356 	*resident_count = 0;
2357 	if (vmmap_skip_res_cnt)
2358 		return;
2359 
2360 	pa = 0;
2361 	obj = entry->object.vm_object;
2362 	addr = entry->start;
2363 	m_adv = NULL;
2364 	pi = OFF_TO_IDX(entry->offset);
2365 	for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) {
2366 		if (m_adv != NULL) {
2367 			m = m_adv;
2368 		} else {
2369 			pi_adv = atop(entry->end - addr);
2370 			pindex = pi;
2371 			for (tobj = obj;; tobj = tobj->backing_object) {
2372 				m = vm_page_find_least(tobj, pindex);
2373 				if (m != NULL) {
2374 					if (m->pindex == pindex)
2375 						break;
2376 					if (pi_adv > m->pindex - pindex) {
2377 						pi_adv = m->pindex - pindex;
2378 						m_adv = m;
2379 					}
2380 				}
2381 				if (tobj->backing_object == NULL)
2382 					goto next;
2383 				pindex += OFF_TO_IDX(tobj->
2384 				    backing_object_offset);
2385 			}
2386 		}
2387 		m_adv = NULL;
2388 		if (m->psind != 0 && addr + pagesizes[1] <= entry->end &&
2389 		    (addr & (pagesizes[1] - 1)) == 0 &&
2390 		    (pmap_mincore(map->pmap, addr, &pa) & MINCORE_SUPER) != 0) {
2391 			*super = true;
2392 			pi_adv = atop(pagesizes[1]);
2393 		} else {
2394 			/*
2395 			 * We do not test the found page on validity.
2396 			 * Either the page is busy and being paged in,
2397 			 * or it was invalidated.  The first case
2398 			 * should be counted as resident, the second
2399 			 * is not so clear; we do account both.
2400 			 */
2401 			pi_adv = 1;
2402 		}
2403 		*resident_count += pi_adv;
2404 next:;
2405 	}
2406 }
2407 
2408 /*
2409  * Must be called with the process locked and will return unlocked.
2410  */
2411 int
2412 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb, ssize_t maxlen, int flags)
2413 {
2414 	vm_map_entry_t entry, tmp_entry;
2415 	struct vattr va;
2416 	vm_map_t map;
2417 	vm_object_t obj, tobj, lobj;
2418 	char *fullpath, *freepath;
2419 	struct kinfo_vmentry *kve;
2420 	struct ucred *cred;
2421 	struct vnode *vp;
2422 	struct vmspace *vm;
2423 	vm_offset_t addr;
2424 	unsigned int last_timestamp;
2425 	int error;
2426 	bool super;
2427 
2428 	PROC_LOCK_ASSERT(p, MA_OWNED);
2429 
2430 	_PHOLD(p);
2431 	PROC_UNLOCK(p);
2432 	vm = vmspace_acquire_ref(p);
2433 	if (vm == NULL) {
2434 		PRELE(p);
2435 		return (ESRCH);
2436 	}
2437 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK | M_ZERO);
2438 
2439 	error = 0;
2440 	map = &vm->vm_map;
2441 	vm_map_lock_read(map);
2442 	VM_MAP_ENTRY_FOREACH(entry, map) {
2443 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2444 			continue;
2445 
2446 		addr = entry->end;
2447 		bzero(kve, sizeof(*kve));
2448 		obj = entry->object.vm_object;
2449 		if (obj != NULL) {
2450 			for (tobj = obj; tobj != NULL;
2451 			    tobj = tobj->backing_object) {
2452 				VM_OBJECT_RLOCK(tobj);
2453 				kve->kve_offset += tobj->backing_object_offset;
2454 				lobj = tobj;
2455 			}
2456 			if (obj->backing_object == NULL)
2457 				kve->kve_private_resident =
2458 				    obj->resident_page_count;
2459 			kern_proc_vmmap_resident(map, entry,
2460 			    &kve->kve_resident, &super);
2461 			if (super)
2462 				kve->kve_flags |= KVME_FLAG_SUPER;
2463 			for (tobj = obj; tobj != NULL;
2464 			    tobj = tobj->backing_object) {
2465 				if (tobj != obj && tobj != lobj)
2466 					VM_OBJECT_RUNLOCK(tobj);
2467 			}
2468 		} else {
2469 			lobj = NULL;
2470 		}
2471 
2472 		kve->kve_start = entry->start;
2473 		kve->kve_end = entry->end;
2474 		kve->kve_offset += entry->offset;
2475 
2476 		if (entry->protection & VM_PROT_READ)
2477 			kve->kve_protection |= KVME_PROT_READ;
2478 		if (entry->protection & VM_PROT_WRITE)
2479 			kve->kve_protection |= KVME_PROT_WRITE;
2480 		if (entry->protection & VM_PROT_EXECUTE)
2481 			kve->kve_protection |= KVME_PROT_EXEC;
2482 
2483 		if (entry->eflags & MAP_ENTRY_COW)
2484 			kve->kve_flags |= KVME_FLAG_COW;
2485 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2486 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2487 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2488 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2489 		if (entry->eflags & MAP_ENTRY_GROWS_UP)
2490 			kve->kve_flags |= KVME_FLAG_GROWS_UP;
2491 		if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2492 			kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2493 		if (entry->eflags & MAP_ENTRY_USER_WIRED)
2494 			kve->kve_flags |= KVME_FLAG_USER_WIRED;
2495 
2496 		last_timestamp = map->timestamp;
2497 		vm_map_unlock_read(map);
2498 
2499 		freepath = NULL;
2500 		fullpath = "";
2501 		if (lobj != NULL) {
2502 			kve->kve_type = vm_object_kvme_type(lobj, &vp);
2503 			if (vp != NULL)
2504 				vref(vp);
2505 			if (lobj != obj)
2506 				VM_OBJECT_RUNLOCK(lobj);
2507 
2508 			kve->kve_ref_count = obj->ref_count;
2509 			kve->kve_shadow_count = obj->shadow_count;
2510 			VM_OBJECT_RUNLOCK(obj);
2511 			if (vp != NULL) {
2512 				vn_fullpath(curthread, vp, &fullpath,
2513 				    &freepath);
2514 				kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2515 				cred = curthread->td_ucred;
2516 				vn_lock(vp, LK_SHARED | LK_RETRY);
2517 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2518 					kve->kve_vn_fileid = va.va_fileid;
2519 					kve->kve_vn_fsid = va.va_fsid;
2520 					kve->kve_vn_fsid_freebsd11 =
2521 					    kve->kve_vn_fsid; /* truncate */
2522 					kve->kve_vn_mode =
2523 					    MAKEIMODE(va.va_type, va.va_mode);
2524 					kve->kve_vn_size = va.va_size;
2525 					kve->kve_vn_rdev = va.va_rdev;
2526 					kve->kve_vn_rdev_freebsd11 =
2527 					    kve->kve_vn_rdev; /* truncate */
2528 					kve->kve_status = KF_ATTR_VALID;
2529 				}
2530 				vput(vp);
2531 			}
2532 		} else {
2533 			kve->kve_type = KVME_TYPE_NONE;
2534 			kve->kve_ref_count = 0;
2535 			kve->kve_shadow_count = 0;
2536 		}
2537 
2538 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2539 		if (freepath != NULL)
2540 			free(freepath, M_TEMP);
2541 
2542 		/* Pack record size down */
2543 		if ((flags & KERN_VMMAP_PACK_KINFO) != 0)
2544 			kve->kve_structsize =
2545 			    offsetof(struct kinfo_vmentry, kve_path) +
2546 			    strlen(kve->kve_path) + 1;
2547 		else
2548 			kve->kve_structsize = sizeof(*kve);
2549 		kve->kve_structsize = roundup(kve->kve_structsize,
2550 		    sizeof(uint64_t));
2551 
2552 		/* Halt filling and truncate rather than exceeding maxlen */
2553 		if (maxlen != -1 && maxlen < kve->kve_structsize) {
2554 			error = 0;
2555 			vm_map_lock_read(map);
2556 			break;
2557 		} else if (maxlen != -1)
2558 			maxlen -= kve->kve_structsize;
2559 
2560 		if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0)
2561 			error = ENOMEM;
2562 		vm_map_lock_read(map);
2563 		if (error != 0)
2564 			break;
2565 		if (last_timestamp != map->timestamp) {
2566 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2567 			entry = tmp_entry;
2568 		}
2569 	}
2570 	vm_map_unlock_read(map);
2571 	vmspace_free(vm);
2572 	PRELE(p);
2573 	free(kve, M_TEMP);
2574 	return (error);
2575 }
2576 
2577 static int
2578 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2579 {
2580 	struct proc *p;
2581 	struct sbuf sb;
2582 	int error, error2, *name;
2583 
2584 	name = (int *)arg1;
2585 	sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req);
2586 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2587 	error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
2588 	if (error != 0) {
2589 		sbuf_delete(&sb);
2590 		return (error);
2591 	}
2592 	error = kern_proc_vmmap_out(p, &sb, -1, KERN_VMMAP_PACK_KINFO);
2593 	error2 = sbuf_finish(&sb);
2594 	sbuf_delete(&sb);
2595 	return (error != 0 ? error : error2);
2596 }
2597 
2598 #if defined(STACK) || defined(DDB)
2599 static int
2600 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2601 {
2602 	struct kinfo_kstack *kkstp;
2603 	int error, i, *name, numthreads;
2604 	lwpid_t *lwpidarray;
2605 	struct thread *td;
2606 	struct stack *st;
2607 	struct sbuf sb;
2608 	struct proc *p;
2609 
2610 	name = (int *)arg1;
2611 	error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2612 	if (error != 0)
2613 		return (error);
2614 
2615 	kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2616 	st = stack_create(M_WAITOK);
2617 
2618 	lwpidarray = NULL;
2619 	PROC_LOCK(p);
2620 	do {
2621 		if (lwpidarray != NULL) {
2622 			free(lwpidarray, M_TEMP);
2623 			lwpidarray = NULL;
2624 		}
2625 		numthreads = p->p_numthreads;
2626 		PROC_UNLOCK(p);
2627 		lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2628 		    M_WAITOK | M_ZERO);
2629 		PROC_LOCK(p);
2630 	} while (numthreads < p->p_numthreads);
2631 
2632 	/*
2633 	 * XXXRW: During the below loop, execve(2) and countless other sorts
2634 	 * of changes could have taken place.  Should we check to see if the
2635 	 * vmspace has been replaced, or the like, in order to prevent
2636 	 * giving a snapshot that spans, say, execve(2), with some threads
2637 	 * before and some after?  Among other things, the credentials could
2638 	 * have changed, in which case the right to extract debug info might
2639 	 * no longer be assured.
2640 	 */
2641 	i = 0;
2642 	FOREACH_THREAD_IN_PROC(p, td) {
2643 		KASSERT(i < numthreads,
2644 		    ("sysctl_kern_proc_kstack: numthreads"));
2645 		lwpidarray[i] = td->td_tid;
2646 		i++;
2647 	}
2648 	numthreads = i;
2649 	for (i = 0; i < numthreads; i++) {
2650 		td = thread_find(p, lwpidarray[i]);
2651 		if (td == NULL) {
2652 			continue;
2653 		}
2654 		bzero(kkstp, sizeof(*kkstp));
2655 		(void)sbuf_new(&sb, kkstp->kkst_trace,
2656 		    sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2657 		thread_lock(td);
2658 		kkstp->kkst_tid = td->td_tid;
2659 		if (TD_IS_SWAPPED(td))
2660 			kkstp->kkst_state = KKST_STATE_SWAPPED;
2661 		else if (stack_save_td(st, td) == 0)
2662 			kkstp->kkst_state = KKST_STATE_STACKOK;
2663 		else
2664 			kkstp->kkst_state = KKST_STATE_RUNNING;
2665 		thread_unlock(td);
2666 		PROC_UNLOCK(p);
2667 		stack_sbuf_print(&sb, st);
2668 		sbuf_finish(&sb);
2669 		sbuf_delete(&sb);
2670 		error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2671 		PROC_LOCK(p);
2672 		if (error)
2673 			break;
2674 	}
2675 	_PRELE(p);
2676 	PROC_UNLOCK(p);
2677 	if (lwpidarray != NULL)
2678 		free(lwpidarray, M_TEMP);
2679 	stack_destroy(st);
2680 	free(kkstp, M_TEMP);
2681 	return (error);
2682 }
2683 #endif
2684 
2685 /*
2686  * This sysctl allows a process to retrieve the full list of groups from
2687  * itself or another process.
2688  */
2689 static int
2690 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2691 {
2692 	pid_t *pidp = (pid_t *)arg1;
2693 	unsigned int arglen = arg2;
2694 	struct proc *p;
2695 	struct ucred *cred;
2696 	int error;
2697 
2698 	if (arglen != 1)
2699 		return (EINVAL);
2700 	if (*pidp == -1) {	/* -1 means this process */
2701 		p = req->td->td_proc;
2702 		PROC_LOCK(p);
2703 	} else {
2704 		error = pget(*pidp, PGET_CANSEE, &p);
2705 		if (error != 0)
2706 			return (error);
2707 	}
2708 
2709 	cred = crhold(p->p_ucred);
2710 	PROC_UNLOCK(p);
2711 
2712 	error = SYSCTL_OUT(req, cred->cr_groups,
2713 	    cred->cr_ngroups * sizeof(gid_t));
2714 	crfree(cred);
2715 	return (error);
2716 }
2717 
2718 /*
2719  * This sysctl allows a process to retrieve or/and set the resource limit for
2720  * another process.
2721  */
2722 static int
2723 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2724 {
2725 	int *name = (int *)arg1;
2726 	u_int namelen = arg2;
2727 	struct rlimit rlim;
2728 	struct proc *p;
2729 	u_int which;
2730 	int flags, error;
2731 
2732 	if (namelen != 2)
2733 		return (EINVAL);
2734 
2735 	which = (u_int)name[1];
2736 	if (which >= RLIM_NLIMITS)
2737 		return (EINVAL);
2738 
2739 	if (req->newptr != NULL && req->newlen != sizeof(rlim))
2740 		return (EINVAL);
2741 
2742 	flags = PGET_HOLD | PGET_NOTWEXIT;
2743 	if (req->newptr != NULL)
2744 		flags |= PGET_CANDEBUG;
2745 	else
2746 		flags |= PGET_CANSEE;
2747 	error = pget((pid_t)name[0], flags, &p);
2748 	if (error != 0)
2749 		return (error);
2750 
2751 	/*
2752 	 * Retrieve limit.
2753 	 */
2754 	if (req->oldptr != NULL) {
2755 		PROC_LOCK(p);
2756 		lim_rlimit_proc(p, which, &rlim);
2757 		PROC_UNLOCK(p);
2758 	}
2759 	error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
2760 	if (error != 0)
2761 		goto errout;
2762 
2763 	/*
2764 	 * Set limit.
2765 	 */
2766 	if (req->newptr != NULL) {
2767 		error = SYSCTL_IN(req, &rlim, sizeof(rlim));
2768 		if (error == 0)
2769 			error = kern_proc_setrlimit(curthread, p, which, &rlim);
2770 	}
2771 
2772 errout:
2773 	PRELE(p);
2774 	return (error);
2775 }
2776 
2777 /*
2778  * This sysctl allows a process to retrieve ps_strings structure location of
2779  * another process.
2780  */
2781 static int
2782 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
2783 {
2784 	int *name = (int *)arg1;
2785 	u_int namelen = arg2;
2786 	struct proc *p;
2787 	vm_offset_t ps_strings;
2788 	int error;
2789 #ifdef COMPAT_FREEBSD32
2790 	uint32_t ps_strings32;
2791 #endif
2792 
2793 	if (namelen != 1)
2794 		return (EINVAL);
2795 
2796 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2797 	if (error != 0)
2798 		return (error);
2799 #ifdef COMPAT_FREEBSD32
2800 	if ((req->flags & SCTL_MASK32) != 0) {
2801 		/*
2802 		 * We return 0 if the 32 bit emulation request is for a 64 bit
2803 		 * process.
2804 		 */
2805 		ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
2806 		    PTROUT(p->p_sysent->sv_psstrings) : 0;
2807 		PROC_UNLOCK(p);
2808 		error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
2809 		return (error);
2810 	}
2811 #endif
2812 	ps_strings = p->p_sysent->sv_psstrings;
2813 	PROC_UNLOCK(p);
2814 	error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
2815 	return (error);
2816 }
2817 
2818 /*
2819  * This sysctl allows a process to retrieve umask of another process.
2820  */
2821 static int
2822 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
2823 {
2824 	int *name = (int *)arg1;
2825 	u_int namelen = arg2;
2826 	struct proc *p;
2827 	int error;
2828 	u_short fd_cmask;
2829 	pid_t pid;
2830 
2831 	if (namelen != 1)
2832 		return (EINVAL);
2833 
2834 	pid = (pid_t)name[0];
2835 	p = curproc;
2836 	if (pid == p->p_pid || pid == 0) {
2837 		fd_cmask = p->p_fd->fd_cmask;
2838 		goto out;
2839 	}
2840 
2841 	error = pget(pid, PGET_WANTREAD, &p);
2842 	if (error != 0)
2843 		return (error);
2844 
2845 	fd_cmask = p->p_fd->fd_cmask;
2846 	PRELE(p);
2847 out:
2848 	error = SYSCTL_OUT(req, &fd_cmask, sizeof(fd_cmask));
2849 	return (error);
2850 }
2851 
2852 /*
2853  * This sysctl allows a process to set and retrieve binary osreldate of
2854  * another process.
2855  */
2856 static int
2857 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
2858 {
2859 	int *name = (int *)arg1;
2860 	u_int namelen = arg2;
2861 	struct proc *p;
2862 	int flags, error, osrel;
2863 
2864 	if (namelen != 1)
2865 		return (EINVAL);
2866 
2867 	if (req->newptr != NULL && req->newlen != sizeof(osrel))
2868 		return (EINVAL);
2869 
2870 	flags = PGET_HOLD | PGET_NOTWEXIT;
2871 	if (req->newptr != NULL)
2872 		flags |= PGET_CANDEBUG;
2873 	else
2874 		flags |= PGET_CANSEE;
2875 	error = pget((pid_t)name[0], flags, &p);
2876 	if (error != 0)
2877 		return (error);
2878 
2879 	error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
2880 	if (error != 0)
2881 		goto errout;
2882 
2883 	if (req->newptr != NULL) {
2884 		error = SYSCTL_IN(req, &osrel, sizeof(osrel));
2885 		if (error != 0)
2886 			goto errout;
2887 		if (osrel < 0) {
2888 			error = EINVAL;
2889 			goto errout;
2890 		}
2891 		p->p_osrel = osrel;
2892 	}
2893 errout:
2894 	PRELE(p);
2895 	return (error);
2896 }
2897 
2898 static int
2899 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS)
2900 {
2901 	int *name = (int *)arg1;
2902 	u_int namelen = arg2;
2903 	struct proc *p;
2904 	struct kinfo_sigtramp kst;
2905 	const struct sysentvec *sv;
2906 	int error;
2907 #ifdef COMPAT_FREEBSD32
2908 	struct kinfo_sigtramp32 kst32;
2909 #endif
2910 
2911 	if (namelen != 1)
2912 		return (EINVAL);
2913 
2914 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2915 	if (error != 0)
2916 		return (error);
2917 	sv = p->p_sysent;
2918 #ifdef COMPAT_FREEBSD32
2919 	if ((req->flags & SCTL_MASK32) != 0) {
2920 		bzero(&kst32, sizeof(kst32));
2921 		if (SV_PROC_FLAG(p, SV_ILP32)) {
2922 			if (sv->sv_sigcode_base != 0) {
2923 				kst32.ksigtramp_start = sv->sv_sigcode_base;
2924 				kst32.ksigtramp_end = sv->sv_sigcode_base +
2925 				    *sv->sv_szsigcode;
2926 			} else {
2927 				kst32.ksigtramp_start = sv->sv_psstrings -
2928 				    *sv->sv_szsigcode;
2929 				kst32.ksigtramp_end = sv->sv_psstrings;
2930 			}
2931 		}
2932 		PROC_UNLOCK(p);
2933 		error = SYSCTL_OUT(req, &kst32, sizeof(kst32));
2934 		return (error);
2935 	}
2936 #endif
2937 	bzero(&kst, sizeof(kst));
2938 	if (sv->sv_sigcode_base != 0) {
2939 		kst.ksigtramp_start = (char *)sv->sv_sigcode_base;
2940 		kst.ksigtramp_end = (char *)sv->sv_sigcode_base +
2941 		    *sv->sv_szsigcode;
2942 	} else {
2943 		kst.ksigtramp_start = (char *)sv->sv_psstrings -
2944 		    *sv->sv_szsigcode;
2945 		kst.ksigtramp_end = (char *)sv->sv_psstrings;
2946 	}
2947 	PROC_UNLOCK(p);
2948 	error = SYSCTL_OUT(req, &kst, sizeof(kst));
2949 	return (error);
2950 }
2951 
2952 static int
2953 sysctl_kern_proc_sigfastblk(SYSCTL_HANDLER_ARGS)
2954 {
2955 	int *name = (int *)arg1;
2956 	u_int namelen = arg2;
2957 	pid_t pid;
2958 	struct proc *p;
2959 	struct thread *td1;
2960 	uintptr_t addr;
2961 #ifdef COMPAT_FREEBSD32
2962 	uint32_t addr32;
2963 #endif
2964 	int error;
2965 
2966 	if (namelen != 1 || req->newptr != NULL)
2967 		return (EINVAL);
2968 
2969 	pid = (pid_t)name[0];
2970 	error = pget(pid, PGET_HOLD | PGET_NOTWEXIT | PGET_CANDEBUG, &p);
2971 	if (error != 0)
2972 		return (error);
2973 
2974 	PROC_LOCK(p);
2975 #ifdef COMPAT_FREEBSD32
2976 	if (SV_CURPROC_FLAG(SV_ILP32)) {
2977 		if (!SV_PROC_FLAG(p, SV_ILP32)) {
2978 			error = EINVAL;
2979 			goto errlocked;
2980 		}
2981 	}
2982 #endif
2983 	if (pid <= PID_MAX) {
2984 		td1 = FIRST_THREAD_IN_PROC(p);
2985 	} else {
2986 		FOREACH_THREAD_IN_PROC(p, td1) {
2987 			if (td1->td_tid == pid)
2988 				break;
2989 		}
2990 	}
2991 	if (td1 == NULL) {
2992 		error = ESRCH;
2993 		goto errlocked;
2994 	}
2995 	/*
2996 	 * The access to the private thread flags.  It is fine as far
2997 	 * as no out-of-thin-air values are read from td_pflags, and
2998 	 * usermode read of the td_sigblock_ptr is racy inherently,
2999 	 * since target process might have already changed it
3000 	 * meantime.
3001 	 */
3002 	if ((td1->td_pflags & TDP_SIGFASTBLOCK) != 0)
3003 		addr = (uintptr_t)td1->td_sigblock_ptr;
3004 	else
3005 		error = ENOTTY;
3006 
3007 errlocked:
3008 	_PRELE(p);
3009 	PROC_UNLOCK(p);
3010 	if (error != 0)
3011 		return (error);
3012 
3013 #ifdef COMPAT_FREEBSD32
3014 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3015 		addr32 = addr;
3016 		error = SYSCTL_OUT(req, &addr32, sizeof(addr32));
3017 	} else
3018 #endif
3019 		error = SYSCTL_OUT(req, &addr, sizeof(addr));
3020 	return (error);
3021 }
3022 
3023 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,  0,
3024     "Process table");
3025 
3026 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
3027 	CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
3028 	"Return entire process table");
3029 
3030 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3031 	sysctl_kern_proc, "Process table");
3032 
3033 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
3034 	sysctl_kern_proc, "Process table");
3035 
3036 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3037 	sysctl_kern_proc, "Process table");
3038 
3039 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
3040 	CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3041 
3042 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
3043 	sysctl_kern_proc, "Process table");
3044 
3045 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3046 	sysctl_kern_proc, "Process table");
3047 
3048 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3049 	sysctl_kern_proc, "Process table");
3050 
3051 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3052 	sysctl_kern_proc, "Process table");
3053 
3054 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
3055 	sysctl_kern_proc, "Return process table, no threads");
3056 
3057 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
3058 	CTLFLAG_RW | CTLFLAG_CAPWR | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
3059 	sysctl_kern_proc_args, "Process argument list");
3060 
3061 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
3062 	sysctl_kern_proc_env, "Process environment");
3063 
3064 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
3065 	CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
3066 
3067 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
3068 	CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
3069 
3070 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
3071 	CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
3072 	"Process syscall vector name (ABI type)");
3073 
3074 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
3075 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3076 
3077 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
3078 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3079 
3080 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
3081 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3082 
3083 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
3084 	sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3085 
3086 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
3087 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3088 
3089 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
3090 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3091 
3092 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
3093 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3094 
3095 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
3096 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3097 
3098 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
3099 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
3100 	"Return process table, no threads");
3101 
3102 #ifdef COMPAT_FREEBSD7
3103 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
3104 	CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
3105 #endif
3106 
3107 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
3108 	CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
3109 
3110 #if defined(STACK) || defined(DDB)
3111 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
3112 	CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
3113 #endif
3114 
3115 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
3116 	CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
3117 
3118 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
3119 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
3120 	"Process resource limits");
3121 
3122 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
3123 	CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
3124 	"Process ps_strings location");
3125 
3126 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
3127 	CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
3128 
3129 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
3130 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
3131 	"Process binary osreldate");
3132 
3133 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD |
3134 	CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp,
3135 	"Process signal trampoline location");
3136 
3137 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGFASTBLK, sigfastblk, CTLFLAG_RD |
3138 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_sigfastblk,
3139 	"Thread sigfastblock address");
3140 
3141 int allproc_gen;
3142 
3143 /*
3144  * stop_all_proc() purpose is to stop all process which have usermode,
3145  * except current process for obvious reasons.  This makes it somewhat
3146  * unreliable when invoked from multithreaded process.  The service
3147  * must not be user-callable anyway.
3148  */
3149 void
3150 stop_all_proc(void)
3151 {
3152 	struct proc *cp, *p;
3153 	int r, gen;
3154 	bool restart, seen_stopped, seen_exiting, stopped_some;
3155 
3156 	cp = curproc;
3157 allproc_loop:
3158 	sx_xlock(&allproc_lock);
3159 	gen = allproc_gen;
3160 	seen_exiting = seen_stopped = stopped_some = restart = false;
3161 	LIST_REMOVE(cp, p_list);
3162 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3163 	for (;;) {
3164 		p = LIST_NEXT(cp, p_list);
3165 		if (p == NULL)
3166 			break;
3167 		LIST_REMOVE(cp, p_list);
3168 		LIST_INSERT_AFTER(p, cp, p_list);
3169 		PROC_LOCK(p);
3170 		if ((p->p_flag & (P_KPROC | P_SYSTEM | P_TOTAL_STOP)) != 0) {
3171 			PROC_UNLOCK(p);
3172 			continue;
3173 		}
3174 		if ((p->p_flag & P_WEXIT) != 0) {
3175 			seen_exiting = true;
3176 			PROC_UNLOCK(p);
3177 			continue;
3178 		}
3179 		if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) {
3180 			/*
3181 			 * Stopped processes are tolerated when there
3182 			 * are no other processes which might continue
3183 			 * them.  P_STOPPED_SINGLE but not
3184 			 * P_TOTAL_STOP process still has at least one
3185 			 * thread running.
3186 			 */
3187 			seen_stopped = true;
3188 			PROC_UNLOCK(p);
3189 			continue;
3190 		}
3191 		sx_xunlock(&allproc_lock);
3192 		_PHOLD(p);
3193 		r = thread_single(p, SINGLE_ALLPROC);
3194 		if (r != 0)
3195 			restart = true;
3196 		else
3197 			stopped_some = true;
3198 		_PRELE(p);
3199 		PROC_UNLOCK(p);
3200 		sx_xlock(&allproc_lock);
3201 	}
3202 	/* Catch forked children we did not see in iteration. */
3203 	if (gen != allproc_gen)
3204 		restart = true;
3205 	sx_xunlock(&allproc_lock);
3206 	if (restart || stopped_some || seen_exiting || seen_stopped) {
3207 		kern_yield(PRI_USER);
3208 		goto allproc_loop;
3209 	}
3210 }
3211 
3212 void
3213 resume_all_proc(void)
3214 {
3215 	struct proc *cp, *p;
3216 
3217 	cp = curproc;
3218 	sx_xlock(&allproc_lock);
3219 again:
3220 	LIST_REMOVE(cp, p_list);
3221 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3222 	for (;;) {
3223 		p = LIST_NEXT(cp, p_list);
3224 		if (p == NULL)
3225 			break;
3226 		LIST_REMOVE(cp, p_list);
3227 		LIST_INSERT_AFTER(p, cp, p_list);
3228 		PROC_LOCK(p);
3229 		if ((p->p_flag & P_TOTAL_STOP) != 0) {
3230 			sx_xunlock(&allproc_lock);
3231 			_PHOLD(p);
3232 			thread_single_end(p, SINGLE_ALLPROC);
3233 			_PRELE(p);
3234 			PROC_UNLOCK(p);
3235 			sx_xlock(&allproc_lock);
3236 		} else {
3237 			PROC_UNLOCK(p);
3238 		}
3239 	}
3240 	/*  Did the loop above missed any stopped process ? */
3241 	FOREACH_PROC_IN_SYSTEM(p) {
3242 		/* No need for proc lock. */
3243 		if ((p->p_flag & P_TOTAL_STOP) != 0)
3244 			goto again;
3245 	}
3246 	sx_xunlock(&allproc_lock);
3247 }
3248 
3249 /* #define	TOTAL_STOP_DEBUG	1 */
3250 #ifdef TOTAL_STOP_DEBUG
3251 volatile static int ap_resume;
3252 #include <sys/mount.h>
3253 
3254 static int
3255 sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS)
3256 {
3257 	int error, val;
3258 
3259 	val = 0;
3260 	ap_resume = 0;
3261 	error = sysctl_handle_int(oidp, &val, 0, req);
3262 	if (error != 0 || req->newptr == NULL)
3263 		return (error);
3264 	if (val != 0) {
3265 		stop_all_proc();
3266 		syncer_suspend();
3267 		while (ap_resume == 0)
3268 			;
3269 		syncer_resume();
3270 		resume_all_proc();
3271 	}
3272 	return (0);
3273 }
3274 
3275 SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW |
3276     CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0,
3277     sysctl_debug_stop_all_proc, "I",
3278     "");
3279 #endif
3280