xref: /freebsd/sys/kern/kern_proc.c (revision 00b6911b7f95bffeff6fdc1843562375d14e9cd8)
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 #ifdef COMPAT_FREEBSD32
92 #include <compat/freebsd32/freebsd32.h>
93 #include <compat/freebsd32/freebsd32_util.h>
94 #endif
95 
96 SDT_PROVIDER_DEFINE(proc);
97 
98 MALLOC_DEFINE(M_PGRP, "pgrp", "process group header");
99 MALLOC_DEFINE(M_SESSION, "session", "session header");
100 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
101 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
102 
103 static void doenterpgrp(struct proc *, struct pgrp *);
104 static void orphanpg(struct pgrp *pg);
105 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp);
106 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp);
107 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp,
108     int preferthread);
109 static void pgadjustjobc(struct pgrp *pgrp, int entering);
110 static void pgdelete(struct pgrp *);
111 static int proc_ctor(void *mem, int size, void *arg, int flags);
112 static void proc_dtor(void *mem, int size, void *arg);
113 static int proc_init(void *mem, int size, int flags);
114 static void proc_fini(void *mem, int size);
115 static void pargs_free(struct pargs *pa);
116 
117 /*
118  * Other process lists
119  */
120 struct pidhashhead *pidhashtbl;
121 struct sx *pidhashtbl_lock;
122 u_long pidhash;
123 u_long pidhashlock;
124 struct pgrphashhead *pgrphashtbl;
125 u_long pgrphash;
126 struct proclist allproc;
127 struct sx __exclusive_cache_line allproc_lock;
128 struct sx __exclusive_cache_line proctree_lock;
129 struct mtx __exclusive_cache_line ppeers_lock;
130 struct mtx __exclusive_cache_line procid_lock;
131 uma_zone_t proc_zone;
132 
133 /*
134  * The offset of various fields in struct proc and struct thread.
135  * These are used by kernel debuggers to enumerate kernel threads and
136  * processes.
137  */
138 const int proc_off_p_pid = offsetof(struct proc, p_pid);
139 const int proc_off_p_comm = offsetof(struct proc, p_comm);
140 const int proc_off_p_list = offsetof(struct proc, p_list);
141 const int proc_off_p_threads = offsetof(struct proc, p_threads);
142 const int thread_off_td_tid = offsetof(struct thread, td_tid);
143 const int thread_off_td_name = offsetof(struct thread, td_name);
144 const int thread_off_td_oncpu = offsetof(struct thread, td_oncpu);
145 const int thread_off_td_pcb = offsetof(struct thread, td_pcb);
146 const int thread_off_td_plist = offsetof(struct thread, td_plist);
147 
148 EVENTHANDLER_LIST_DEFINE(process_ctor);
149 EVENTHANDLER_LIST_DEFINE(process_dtor);
150 EVENTHANDLER_LIST_DEFINE(process_init);
151 EVENTHANDLER_LIST_DEFINE(process_fini);
152 EVENTHANDLER_LIST_DEFINE(process_exit);
153 EVENTHANDLER_LIST_DEFINE(process_fork);
154 EVENTHANDLER_LIST_DEFINE(process_exec);
155 
156 int kstack_pages = KSTACK_PAGES;
157 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0,
158     "Kernel stack size in pages");
159 static int vmmap_skip_res_cnt = 0;
160 SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW,
161     &vmmap_skip_res_cnt, 0,
162     "Skip calculation of the pages resident count in kern.proc.vmmap");
163 
164 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
165 #ifdef COMPAT_FREEBSD32
166 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
167 #endif
168 
169 /*
170  * Initialize global process hashing structures.
171  */
172 void
173 procinit(void)
174 {
175 	u_long i;
176 
177 	sx_init(&allproc_lock, "allproc");
178 	sx_init(&proctree_lock, "proctree");
179 	mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
180 	mtx_init(&procid_lock, "procid", NULL, MTX_DEF);
181 	LIST_INIT(&allproc);
182 	pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
183 	pidhashlock = (pidhash + 1) / 64;
184 	if (pidhashlock > 0)
185 		pidhashlock--;
186 	pidhashtbl_lock = malloc(sizeof(*pidhashtbl_lock) * (pidhashlock + 1),
187 	    M_PROC, M_WAITOK | M_ZERO);
188 	for (i = 0; i < pidhashlock + 1; i++)
189 		sx_init_flags(&pidhashtbl_lock[i], "pidhash", SX_DUPOK);
190 	pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
191 	proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
192 	    proc_ctor, proc_dtor, proc_init, proc_fini,
193 	    UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
194 	uihashinit();
195 }
196 
197 /*
198  * Prepare a proc for use.
199  */
200 static int
201 proc_ctor(void *mem, int size, void *arg, int flags)
202 {
203 	struct proc *p;
204 	struct thread *td;
205 
206 	p = (struct proc *)mem;
207 	EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
208 	td = FIRST_THREAD_IN_PROC(p);
209 	if (td != NULL) {
210 		/* Make sure all thread constructors are executed */
211 		EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
212 	}
213 	return (0);
214 }
215 
216 /*
217  * Reclaim a proc after use.
218  */
219 static void
220 proc_dtor(void *mem, int size, void *arg)
221 {
222 	struct proc *p;
223 	struct thread *td;
224 
225 	/* INVARIANTS checks go here */
226 	p = (struct proc *)mem;
227 	td = FIRST_THREAD_IN_PROC(p);
228 	if (td != NULL) {
229 #ifdef INVARIANTS
230 		KASSERT((p->p_numthreads == 1),
231 		    ("bad number of threads in exiting process"));
232 		KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
233 #endif
234 		/* Free all OSD associated to this thread. */
235 		osd_thread_exit(td);
236 		td_softdep_cleanup(td);
237 		MPASS(td->td_su == NULL);
238 
239 		/* Make sure all thread destructors are executed */
240 		EVENTHANDLER_DIRECT_INVOKE(thread_dtor, td);
241 	}
242 	EVENTHANDLER_DIRECT_INVOKE(process_dtor, p);
243 	if (p->p_ksi != NULL)
244 		KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
245 }
246 
247 /*
248  * Initialize type-stable parts of a proc (when newly created).
249  */
250 static int
251 proc_init(void *mem, int size, int flags)
252 {
253 	struct proc *p;
254 
255 	p = (struct proc *)mem;
256 	mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK | MTX_NEW);
257 	mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_NEW);
258 	mtx_init(&p->p_statmtx, "pstatl", NULL, MTX_SPIN | MTX_NEW);
259 	mtx_init(&p->p_itimmtx, "pitiml", NULL, MTX_SPIN | MTX_NEW);
260 	mtx_init(&p->p_profmtx, "pprofl", NULL, MTX_SPIN | MTX_NEW);
261 	cv_init(&p->p_pwait, "ppwait");
262 	TAILQ_INIT(&p->p_threads);	     /* all threads in proc */
263 	EVENTHANDLER_DIRECT_INVOKE(process_init, p);
264 	p->p_stats = pstats_alloc();
265 	p->p_pgrp = NULL;
266 	return (0);
267 }
268 
269 /*
270  * UMA should ensure that this function is never called.
271  * Freeing a proc structure would violate type stability.
272  */
273 static void
274 proc_fini(void *mem, int size)
275 {
276 #ifdef notnow
277 	struct proc *p;
278 
279 	p = (struct proc *)mem;
280 	EVENTHANDLER_DIRECT_INVOKE(process_fini, p);
281 	pstats_free(p->p_stats);
282 	thread_free(FIRST_THREAD_IN_PROC(p));
283 	mtx_destroy(&p->p_mtx);
284 	if (p->p_ksi != NULL)
285 		ksiginfo_free(p->p_ksi);
286 #else
287 	panic("proc reclaimed");
288 #endif
289 }
290 
291 /*
292  * PID space management.
293  *
294  * These bitmaps are used by fork_findpid.
295  */
296 bitstr_t bit_decl(proc_id_pidmap, PID_MAX);
297 bitstr_t bit_decl(proc_id_grpidmap, PID_MAX);
298 bitstr_t bit_decl(proc_id_sessidmap, PID_MAX);
299 bitstr_t bit_decl(proc_id_reapmap, PID_MAX);
300 
301 static bitstr_t *proc_id_array[] = {
302 	proc_id_pidmap,
303 	proc_id_grpidmap,
304 	proc_id_sessidmap,
305 	proc_id_reapmap,
306 };
307 
308 void
309 proc_id_set(int type, pid_t id)
310 {
311 
312 	KASSERT(type >= 0 && type < nitems(proc_id_array),
313 	    ("invalid type %d\n", type));
314 	mtx_lock(&procid_lock);
315 	KASSERT(bit_test(proc_id_array[type], id) == 0,
316 	    ("bit %d already set in %d\n", id, type));
317 	bit_set(proc_id_array[type], id);
318 	mtx_unlock(&procid_lock);
319 }
320 
321 void
322 proc_id_set_cond(int type, pid_t id)
323 {
324 
325 	KASSERT(type >= 0 && type < nitems(proc_id_array),
326 	    ("invalid type %d\n", type));
327 	if (bit_test(proc_id_array[type], id))
328 		return;
329 	mtx_lock(&procid_lock);
330 	bit_set(proc_id_array[type], id);
331 	mtx_unlock(&procid_lock);
332 }
333 
334 void
335 proc_id_clear(int type, pid_t id)
336 {
337 
338 	KASSERT(type >= 0 && type < nitems(proc_id_array),
339 	    ("invalid type %d\n", type));
340 	mtx_lock(&procid_lock);
341 	KASSERT(bit_test(proc_id_array[type], id) != 0,
342 	    ("bit %d not set in %d\n", id, type));
343 	bit_clear(proc_id_array[type], id);
344 	mtx_unlock(&procid_lock);
345 }
346 
347 /*
348  * Is p an inferior of the current process?
349  */
350 int
351 inferior(struct proc *p)
352 {
353 
354 	sx_assert(&proctree_lock, SX_LOCKED);
355 	PROC_LOCK_ASSERT(p, MA_OWNED);
356 	for (; p != curproc; p = proc_realparent(p)) {
357 		if (p->p_pid == 0)
358 			return (0);
359 	}
360 	return (1);
361 }
362 
363 /*
364  * Shared lock all the pid hash lists.
365  */
366 void
367 pidhash_slockall(void)
368 {
369 	u_long i;
370 
371 	for (i = 0; i < pidhashlock + 1; i++)
372 		sx_slock(&pidhashtbl_lock[i]);
373 }
374 
375 /*
376  * Shared unlock all the pid hash lists.
377  */
378 void
379 pidhash_sunlockall(void)
380 {
381 	u_long i;
382 
383 	for (i = 0; i < pidhashlock + 1; i++)
384 		sx_sunlock(&pidhashtbl_lock[i]);
385 }
386 
387 /*
388  * Similar to pfind_any(), this function finds zombies.
389  */
390 struct proc *
391 pfind_any_locked(pid_t pid)
392 {
393 	struct proc *p;
394 
395 	sx_assert(PIDHASHLOCK(pid), SX_LOCKED);
396 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
397 		if (p->p_pid == pid) {
398 			PROC_LOCK(p);
399 			if (p->p_state == PRS_NEW) {
400 				PROC_UNLOCK(p);
401 				p = NULL;
402 			}
403 			break;
404 		}
405 	}
406 	return (p);
407 }
408 
409 /*
410  * Locate a process by number.
411  *
412  * By not returning processes in the PRS_NEW state, we allow callers to avoid
413  * testing for that condition to avoid dereferencing p_ucred, et al.
414  */
415 static __always_inline struct proc *
416 _pfind(pid_t pid, bool zombie)
417 {
418 	struct proc *p;
419 
420 	p = curproc;
421 	if (p->p_pid == pid) {
422 		PROC_LOCK(p);
423 		return (p);
424 	}
425 	sx_slock(PIDHASHLOCK(pid));
426 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
427 		if (p->p_pid == pid) {
428 			PROC_LOCK(p);
429 			if (p->p_state == PRS_NEW ||
430 			    (!zombie && p->p_state == PRS_ZOMBIE)) {
431 				PROC_UNLOCK(p);
432 				p = NULL;
433 			}
434 			break;
435 		}
436 	}
437 	sx_sunlock(PIDHASHLOCK(pid));
438 	return (p);
439 }
440 
441 struct proc *
442 pfind(pid_t pid)
443 {
444 
445 	return (_pfind(pid, false));
446 }
447 
448 /*
449  * Same as pfind but allow zombies.
450  */
451 struct proc *
452 pfind_any(pid_t pid)
453 {
454 
455 	return (_pfind(pid, true));
456 }
457 
458 static struct proc *
459 pfind_tid(pid_t tid)
460 {
461 	struct proc *p;
462 	struct thread *td;
463 
464 	sx_slock(&allproc_lock);
465 	FOREACH_PROC_IN_SYSTEM(p) {
466 		PROC_LOCK(p);
467 		if (p->p_state == PRS_NEW) {
468 			PROC_UNLOCK(p);
469 			continue;
470 		}
471 		FOREACH_THREAD_IN_PROC(p, td) {
472 			if (td->td_tid == tid)
473 				goto found;
474 		}
475 		PROC_UNLOCK(p);
476 	}
477 found:
478 	sx_sunlock(&allproc_lock);
479 	return (p);
480 }
481 
482 /*
483  * Locate a process group by number.
484  * The caller must hold proctree_lock.
485  */
486 struct pgrp *
487 pgfind(pid_t pgid)
488 {
489 	struct pgrp *pgrp;
490 
491 	sx_assert(&proctree_lock, SX_LOCKED);
492 
493 	LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
494 		if (pgrp->pg_id == pgid) {
495 			PGRP_LOCK(pgrp);
496 			return (pgrp);
497 		}
498 	}
499 	return (NULL);
500 }
501 
502 /*
503  * Locate process and do additional manipulations, depending on flags.
504  */
505 int
506 pget(pid_t pid, int flags, struct proc **pp)
507 {
508 	struct proc *p;
509 	int error;
510 
511 	p = curproc;
512 	if (p->p_pid == pid) {
513 		PROC_LOCK(p);
514 	} else {
515 		p = NULL;
516 		if (pid <= PID_MAX) {
517 			if ((flags & PGET_NOTWEXIT) == 0)
518 				p = pfind_any(pid);
519 			else
520 				p = pfind(pid);
521 		} else if ((flags & PGET_NOTID) == 0) {
522 			p = pfind_tid(pid);
523 		}
524 		if (p == NULL)
525 			return (ESRCH);
526 		if ((flags & PGET_CANSEE) != 0) {
527 			error = p_cansee(curthread, p);
528 			if (error != 0)
529 				goto errout;
530 		}
531 	}
532 	if ((flags & PGET_CANDEBUG) != 0) {
533 		error = p_candebug(curthread, p);
534 		if (error != 0)
535 			goto errout;
536 	}
537 	if ((flags & PGET_ISCURRENT) != 0 && curproc != p) {
538 		error = EPERM;
539 		goto errout;
540 	}
541 	if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) {
542 		error = ESRCH;
543 		goto errout;
544 	}
545 	if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) {
546 		/*
547 		 * XXXRW: Not clear ESRCH is the right error during proc
548 		 * execve().
549 		 */
550 		error = ESRCH;
551 		goto errout;
552 	}
553 	if ((flags & PGET_HOLD) != 0) {
554 		_PHOLD(p);
555 		PROC_UNLOCK(p);
556 	}
557 	*pp = p;
558 	return (0);
559 errout:
560 	PROC_UNLOCK(p);
561 	return (error);
562 }
563 
564 /*
565  * Create a new process group.
566  * pgid must be equal to the pid of p.
567  * Begin a new session if required.
568  */
569 int
570 enterpgrp(struct proc *p, pid_t pgid, struct pgrp *pgrp, struct session *sess)
571 {
572 
573 	sx_assert(&proctree_lock, SX_XLOCKED);
574 
575 	KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
576 	KASSERT(p->p_pid == pgid,
577 	    ("enterpgrp: new pgrp and pid != pgid"));
578 	KASSERT(pgfind(pgid) == NULL,
579 	    ("enterpgrp: pgrp with pgid exists"));
580 	KASSERT(!SESS_LEADER(p),
581 	    ("enterpgrp: session leader attempted setpgrp"));
582 
583 	mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
584 
585 	if (sess != NULL) {
586 		/*
587 		 * new session
588 		 */
589 		mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
590 		PROC_LOCK(p);
591 		p->p_flag &= ~P_CONTROLT;
592 		PROC_UNLOCK(p);
593 		PGRP_LOCK(pgrp);
594 		sess->s_leader = p;
595 		sess->s_sid = p->p_pid;
596 		proc_id_set(PROC_ID_SESSION, p->p_pid);
597 		refcount_init(&sess->s_count, 1);
598 		sess->s_ttyvp = NULL;
599 		sess->s_ttydp = NULL;
600 		sess->s_ttyp = NULL;
601 		bcopy(p->p_session->s_login, sess->s_login,
602 			    sizeof(sess->s_login));
603 		pgrp->pg_session = sess;
604 		KASSERT(p == curproc,
605 		    ("enterpgrp: mksession and p != curproc"));
606 	} else {
607 		pgrp->pg_session = p->p_session;
608 		sess_hold(pgrp->pg_session);
609 		PGRP_LOCK(pgrp);
610 	}
611 	pgrp->pg_id = pgid;
612 	proc_id_set(PROC_ID_GROUP, p->p_pid);
613 	LIST_INIT(&pgrp->pg_members);
614 
615 	/*
616 	 * As we have an exclusive lock of proctree_lock,
617 	 * this should not deadlock.
618 	 */
619 	LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
620 	pgrp->pg_jobc = 0;
621 	SLIST_INIT(&pgrp->pg_sigiolst);
622 	PGRP_UNLOCK(pgrp);
623 
624 	doenterpgrp(p, pgrp);
625 
626 	return (0);
627 }
628 
629 /*
630  * Move p to an existing process group
631  */
632 int
633 enterthispgrp(struct proc *p, struct pgrp *pgrp)
634 {
635 
636 	sx_assert(&proctree_lock, SX_XLOCKED);
637 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
638 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
639 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
640 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
641 	KASSERT(pgrp->pg_session == p->p_session,
642 		("%s: pgrp's session %p, p->p_session %p.\n",
643 		__func__,
644 		pgrp->pg_session,
645 		p->p_session));
646 	KASSERT(pgrp != p->p_pgrp,
647 		("%s: p belongs to pgrp.", __func__));
648 
649 	doenterpgrp(p, pgrp);
650 
651 	return (0);
652 }
653 
654 /*
655  * Move p to a process group
656  */
657 static void
658 doenterpgrp(struct proc *p, struct pgrp *pgrp)
659 {
660 	struct pgrp *savepgrp;
661 
662 	sx_assert(&proctree_lock, SX_XLOCKED);
663 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
664 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
665 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
666 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
667 
668 	savepgrp = p->p_pgrp;
669 
670 	/*
671 	 * Adjust eligibility of affected pgrps to participate in job control.
672 	 * Increment eligibility counts before decrementing, otherwise we
673 	 * could reach 0 spuriously during the first call.
674 	 */
675 	fixjobc(p, pgrp, 1);
676 	fixjobc(p, p->p_pgrp, 0);
677 
678 	PGRP_LOCK(pgrp);
679 	PGRP_LOCK(savepgrp);
680 	PROC_LOCK(p);
681 	LIST_REMOVE(p, p_pglist);
682 	p->p_pgrp = pgrp;
683 	PROC_UNLOCK(p);
684 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
685 	PGRP_UNLOCK(savepgrp);
686 	PGRP_UNLOCK(pgrp);
687 	if (LIST_EMPTY(&savepgrp->pg_members))
688 		pgdelete(savepgrp);
689 }
690 
691 /*
692  * remove process from process group
693  */
694 int
695 leavepgrp(struct proc *p)
696 {
697 	struct pgrp *savepgrp;
698 
699 	sx_assert(&proctree_lock, SX_XLOCKED);
700 	savepgrp = p->p_pgrp;
701 	PGRP_LOCK(savepgrp);
702 	PROC_LOCK(p);
703 	LIST_REMOVE(p, p_pglist);
704 	p->p_pgrp = NULL;
705 	PROC_UNLOCK(p);
706 	PGRP_UNLOCK(savepgrp);
707 	if (LIST_EMPTY(&savepgrp->pg_members))
708 		pgdelete(savepgrp);
709 	return (0);
710 }
711 
712 /*
713  * delete a process group
714  */
715 static void
716 pgdelete(struct pgrp *pgrp)
717 {
718 	struct session *savesess;
719 	struct tty *tp;
720 
721 	sx_assert(&proctree_lock, SX_XLOCKED);
722 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
723 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
724 
725 	/*
726 	 * Reset any sigio structures pointing to us as a result of
727 	 * F_SETOWN with our pgid.
728 	 */
729 	funsetownlst(&pgrp->pg_sigiolst);
730 
731 	PGRP_LOCK(pgrp);
732 	tp = pgrp->pg_session->s_ttyp;
733 	LIST_REMOVE(pgrp, pg_hash);
734 	savesess = pgrp->pg_session;
735 	PGRP_UNLOCK(pgrp);
736 
737 	/* Remove the reference to the pgrp before deallocating it. */
738 	if (tp != NULL) {
739 		tty_lock(tp);
740 		tty_rel_pgrp(tp, pgrp);
741 	}
742 
743 	proc_id_clear(PROC_ID_GROUP, pgrp->pg_id);
744 	mtx_destroy(&pgrp->pg_mtx);
745 	free(pgrp, M_PGRP);
746 	sess_release(savesess);
747 }
748 
749 static void
750 pgadjustjobc(struct pgrp *pgrp, int entering)
751 {
752 
753 	PGRP_LOCK(pgrp);
754 	if (entering)
755 		pgrp->pg_jobc++;
756 	else {
757 		--pgrp->pg_jobc;
758 		if (pgrp->pg_jobc == 0)
759 			orphanpg(pgrp);
760 	}
761 	PGRP_UNLOCK(pgrp);
762 }
763 
764 /*
765  * Adjust pgrp jobc counters when specified process changes process group.
766  * We count the number of processes in each process group that "qualify"
767  * the group for terminal job control (those with a parent in a different
768  * process group of the same session).  If that count reaches zero, the
769  * process group becomes orphaned.  Check both the specified process'
770  * process group and that of its children.
771  * entering == 0 => p is leaving specified group.
772  * entering == 1 => p is entering specified group.
773  */
774 void
775 fixjobc(struct proc *p, struct pgrp *pgrp, int entering)
776 {
777 	struct pgrp *hispgrp;
778 	struct session *mysession;
779 	struct proc *q;
780 
781 	sx_assert(&proctree_lock, SX_LOCKED);
782 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
783 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
784 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
785 
786 	/*
787 	 * Check p's parent to see whether p qualifies its own process
788 	 * group; if so, adjust count for p's process group.
789 	 */
790 	mysession = pgrp->pg_session;
791 	if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
792 	    hispgrp->pg_session == mysession)
793 		pgadjustjobc(pgrp, entering);
794 
795 	/*
796 	 * Check this process' children to see whether they qualify
797 	 * their process groups; if so, adjust counts for children's
798 	 * process groups.
799 	 */
800 	LIST_FOREACH(q, &p->p_children, p_sibling) {
801 		hispgrp = q->p_pgrp;
802 		if (hispgrp == pgrp ||
803 		    hispgrp->pg_session != mysession)
804 			continue;
805 		if (q->p_state == PRS_ZOMBIE)
806 			continue;
807 		pgadjustjobc(hispgrp, entering);
808 	}
809 }
810 
811 void
812 killjobc(void)
813 {
814 	struct session *sp;
815 	struct tty *tp;
816 	struct proc *p;
817 	struct vnode *ttyvp;
818 
819 	p = curproc;
820 	MPASS(p->p_flag & P_WEXIT);
821 	/*
822 	 * Do a quick check to see if there is anything to do with the
823 	 * proctree_lock held. pgrp and LIST_EMPTY checks are for fixjobc().
824 	 */
825 	PROC_LOCK(p);
826 	if (!SESS_LEADER(p) &&
827 	    (p->p_pgrp == p->p_pptr->p_pgrp) &&
828 	    LIST_EMPTY(&p->p_children)) {
829 		PROC_UNLOCK(p);
830 		return;
831 	}
832 	PROC_UNLOCK(p);
833 
834 	sx_xlock(&proctree_lock);
835 	if (SESS_LEADER(p)) {
836 		sp = p->p_session;
837 
838 		/*
839 		 * s_ttyp is not zero'd; we use this to indicate that
840 		 * the session once had a controlling terminal. (for
841 		 * logging and informational purposes)
842 		 */
843 		SESS_LOCK(sp);
844 		ttyvp = sp->s_ttyvp;
845 		tp = sp->s_ttyp;
846 		sp->s_ttyvp = NULL;
847 		sp->s_ttydp = NULL;
848 		sp->s_leader = NULL;
849 		SESS_UNLOCK(sp);
850 
851 		/*
852 		 * Signal foreground pgrp and revoke access to
853 		 * controlling terminal if it has not been revoked
854 		 * already.
855 		 *
856 		 * Because the TTY may have been revoked in the mean
857 		 * time and could already have a new session associated
858 		 * with it, make sure we don't send a SIGHUP to a
859 		 * foreground process group that does not belong to this
860 		 * session.
861 		 */
862 
863 		if (tp != NULL) {
864 			tty_lock(tp);
865 			if (tp->t_session == sp)
866 				tty_signal_pgrp(tp, SIGHUP);
867 			tty_unlock(tp);
868 		}
869 
870 		if (ttyvp != NULL) {
871 			sx_xunlock(&proctree_lock);
872 			if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) {
873 				VOP_REVOKE(ttyvp, REVOKEALL);
874 				VOP_UNLOCK(ttyvp, 0);
875 			}
876 			vrele(ttyvp);
877 			sx_xlock(&proctree_lock);
878 		}
879 	}
880 	fixjobc(p, p->p_pgrp, 0);
881 	sx_xunlock(&proctree_lock);
882 }
883 
884 /*
885  * A process group has become orphaned;
886  * if there are any stopped processes in the group,
887  * hang-up all process in that group.
888  */
889 static void
890 orphanpg(struct pgrp *pg)
891 {
892 	struct proc *p;
893 
894 	PGRP_LOCK_ASSERT(pg, MA_OWNED);
895 
896 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
897 		PROC_LOCK(p);
898 		if (P_SHOULDSTOP(p) == P_STOPPED_SIG) {
899 			PROC_UNLOCK(p);
900 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
901 				PROC_LOCK(p);
902 				kern_psignal(p, SIGHUP);
903 				kern_psignal(p, SIGCONT);
904 				PROC_UNLOCK(p);
905 			}
906 			return;
907 		}
908 		PROC_UNLOCK(p);
909 	}
910 }
911 
912 void
913 sess_hold(struct session *s)
914 {
915 
916 	refcount_acquire(&s->s_count);
917 }
918 
919 void
920 sess_release(struct session *s)
921 {
922 
923 	if (refcount_release(&s->s_count)) {
924 		if (s->s_ttyp != NULL) {
925 			tty_lock(s->s_ttyp);
926 			tty_rel_sess(s->s_ttyp, s);
927 		}
928 		proc_id_clear(PROC_ID_SESSION, s->s_sid);
929 		mtx_destroy(&s->s_mtx);
930 		free(s, M_SESSION);
931 	}
932 }
933 
934 #ifdef DDB
935 
936 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
937 {
938 	struct pgrp *pgrp;
939 	struct proc *p;
940 	int i;
941 
942 	for (i = 0; i <= pgrphash; i++) {
943 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
944 			printf("\tindx %d\n", i);
945 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
946 				printf(
947 			"\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
948 				    (void *)pgrp, (long)pgrp->pg_id,
949 				    (void *)pgrp->pg_session,
950 				    pgrp->pg_session->s_count,
951 				    (void *)LIST_FIRST(&pgrp->pg_members));
952 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
953 					printf("\t\tpid %ld addr %p pgrp %p\n",
954 					    (long)p->p_pid, (void *)p,
955 					    (void *)p->p_pgrp);
956 				}
957 			}
958 		}
959 	}
960 }
961 #endif /* DDB */
962 
963 /*
964  * Calculate the kinfo_proc members which contain process-wide
965  * informations.
966  * Must be called with the target process locked.
967  */
968 static void
969 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
970 {
971 	struct thread *td;
972 
973 	PROC_LOCK_ASSERT(p, MA_OWNED);
974 
975 	kp->ki_estcpu = 0;
976 	kp->ki_pctcpu = 0;
977 	FOREACH_THREAD_IN_PROC(p, td) {
978 		thread_lock(td);
979 		kp->ki_pctcpu += sched_pctcpu(td);
980 		kp->ki_estcpu += sched_estcpu(td);
981 		thread_unlock(td);
982 	}
983 }
984 
985 /*
986  * Clear kinfo_proc and fill in any information that is common
987  * to all threads in the process.
988  * Must be called with the target process locked.
989  */
990 static void
991 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
992 {
993 	struct thread *td0;
994 	struct tty *tp;
995 	struct session *sp;
996 	struct ucred *cred;
997 	struct sigacts *ps;
998 	struct timeval boottime;
999 
1000 	PROC_LOCK_ASSERT(p, MA_OWNED);
1001 	bzero(kp, sizeof(*kp));
1002 
1003 	kp->ki_structsize = sizeof(*kp);
1004 	kp->ki_paddr = p;
1005 	kp->ki_addr =/* p->p_addr; */0; /* XXX */
1006 	kp->ki_args = p->p_args;
1007 	kp->ki_textvp = p->p_textvp;
1008 #ifdef KTRACE
1009 	kp->ki_tracep = p->p_tracevp;
1010 	kp->ki_traceflag = p->p_traceflag;
1011 #endif
1012 	kp->ki_fd = p->p_fd;
1013 	kp->ki_vmspace = p->p_vmspace;
1014 	kp->ki_flag = p->p_flag;
1015 	kp->ki_flag2 = p->p_flag2;
1016 	cred = p->p_ucred;
1017 	if (cred) {
1018 		kp->ki_uid = cred->cr_uid;
1019 		kp->ki_ruid = cred->cr_ruid;
1020 		kp->ki_svuid = cred->cr_svuid;
1021 		kp->ki_cr_flags = 0;
1022 		if (cred->cr_flags & CRED_FLAG_CAPMODE)
1023 			kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
1024 		/* XXX bde doesn't like KI_NGROUPS */
1025 		if (cred->cr_ngroups > KI_NGROUPS) {
1026 			kp->ki_ngroups = KI_NGROUPS;
1027 			kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
1028 		} else
1029 			kp->ki_ngroups = cred->cr_ngroups;
1030 		bcopy(cred->cr_groups, kp->ki_groups,
1031 		    kp->ki_ngroups * sizeof(gid_t));
1032 		kp->ki_rgid = cred->cr_rgid;
1033 		kp->ki_svgid = cred->cr_svgid;
1034 		/* If jailed(cred), emulate the old P_JAILED flag. */
1035 		if (jailed(cred)) {
1036 			kp->ki_flag |= P_JAILED;
1037 			/* If inside the jail, use 0 as a jail ID. */
1038 			if (cred->cr_prison != curthread->td_ucred->cr_prison)
1039 				kp->ki_jid = cred->cr_prison->pr_id;
1040 		}
1041 		strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
1042 		    sizeof(kp->ki_loginclass));
1043 	}
1044 	ps = p->p_sigacts;
1045 	if (ps) {
1046 		mtx_lock(&ps->ps_mtx);
1047 		kp->ki_sigignore = ps->ps_sigignore;
1048 		kp->ki_sigcatch = ps->ps_sigcatch;
1049 		mtx_unlock(&ps->ps_mtx);
1050 	}
1051 	if (p->p_state != PRS_NEW &&
1052 	    p->p_state != PRS_ZOMBIE &&
1053 	    p->p_vmspace != NULL) {
1054 		struct vmspace *vm = p->p_vmspace;
1055 
1056 		kp->ki_size = vm->vm_map.size;
1057 		kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
1058 		FOREACH_THREAD_IN_PROC(p, td0) {
1059 			if (!TD_IS_SWAPPED(td0))
1060 				kp->ki_rssize += td0->td_kstack_pages;
1061 		}
1062 		kp->ki_swrss = vm->vm_swrss;
1063 		kp->ki_tsize = vm->vm_tsize;
1064 		kp->ki_dsize = vm->vm_dsize;
1065 		kp->ki_ssize = vm->vm_ssize;
1066 	} else if (p->p_state == PRS_ZOMBIE)
1067 		kp->ki_stat = SZOMB;
1068 	if (kp->ki_flag & P_INMEM)
1069 		kp->ki_sflag = PS_INMEM;
1070 	else
1071 		kp->ki_sflag = 0;
1072 	/* Calculate legacy swtime as seconds since 'swtick'. */
1073 	kp->ki_swtime = (ticks - p->p_swtick) / hz;
1074 	kp->ki_pid = p->p_pid;
1075 	kp->ki_nice = p->p_nice;
1076 	kp->ki_fibnum = p->p_fibnum;
1077 	kp->ki_start = p->p_stats->p_start;
1078 	getboottime(&boottime);
1079 	timevaladd(&kp->ki_start, &boottime);
1080 	PROC_STATLOCK(p);
1081 	rufetch(p, &kp->ki_rusage);
1082 	kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
1083 	calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
1084 	PROC_STATUNLOCK(p);
1085 	calccru(p, &kp->ki_childutime, &kp->ki_childstime);
1086 	/* Some callers want child times in a single value. */
1087 	kp->ki_childtime = kp->ki_childstime;
1088 	timevaladd(&kp->ki_childtime, &kp->ki_childutime);
1089 
1090 	FOREACH_THREAD_IN_PROC(p, td0)
1091 		kp->ki_cow += td0->td_cow;
1092 
1093 	tp = NULL;
1094 	if (p->p_pgrp) {
1095 		kp->ki_pgid = p->p_pgrp->pg_id;
1096 		kp->ki_jobc = p->p_pgrp->pg_jobc;
1097 		sp = p->p_pgrp->pg_session;
1098 
1099 		if (sp != NULL) {
1100 			kp->ki_sid = sp->s_sid;
1101 			SESS_LOCK(sp);
1102 			strlcpy(kp->ki_login, sp->s_login,
1103 			    sizeof(kp->ki_login));
1104 			if (sp->s_ttyvp)
1105 				kp->ki_kiflag |= KI_CTTY;
1106 			if (SESS_LEADER(p))
1107 				kp->ki_kiflag |= KI_SLEADER;
1108 			/* XXX proctree_lock */
1109 			tp = sp->s_ttyp;
1110 			SESS_UNLOCK(sp);
1111 		}
1112 	}
1113 	if ((p->p_flag & P_CONTROLT) && tp != NULL) {
1114 		kp->ki_tdev = tty_udev(tp);
1115 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1116 		kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
1117 		if (tp->t_session)
1118 			kp->ki_tsid = tp->t_session->s_sid;
1119 	} else {
1120 		kp->ki_tdev = NODEV;
1121 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1122 	}
1123 	if (p->p_comm[0] != '\0')
1124 		strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
1125 	if (p->p_sysent && p->p_sysent->sv_name != NULL &&
1126 	    p->p_sysent->sv_name[0] != '\0')
1127 		strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
1128 	kp->ki_siglist = p->p_siglist;
1129 	kp->ki_xstat = KW_EXITCODE(p->p_xexit, p->p_xsig);
1130 	kp->ki_acflag = p->p_acflag;
1131 	kp->ki_lock = p->p_lock;
1132 	if (p->p_pptr) {
1133 		kp->ki_ppid = p->p_oppid;
1134 		if (p->p_flag & P_TRACED)
1135 			kp->ki_tracer = p->p_pptr->p_pid;
1136 	}
1137 }
1138 
1139 /*
1140  * Fill in information that is thread specific.  Must be called with
1141  * target process locked.  If 'preferthread' is set, overwrite certain
1142  * process-related fields that are maintained for both threads and
1143  * processes.
1144  */
1145 static void
1146 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
1147 {
1148 	struct proc *p;
1149 
1150 	p = td->td_proc;
1151 	kp->ki_tdaddr = td;
1152 	PROC_LOCK_ASSERT(p, MA_OWNED);
1153 
1154 	if (preferthread)
1155 		PROC_STATLOCK(p);
1156 	thread_lock(td);
1157 	if (td->td_wmesg != NULL)
1158 		strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
1159 	else
1160 		bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
1161 	if (strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname)) >=
1162 	    sizeof(kp->ki_tdname)) {
1163 		strlcpy(kp->ki_moretdname,
1164 		    td->td_name + sizeof(kp->ki_tdname) - 1,
1165 		    sizeof(kp->ki_moretdname));
1166 	} else {
1167 		bzero(kp->ki_moretdname, sizeof(kp->ki_moretdname));
1168 	}
1169 	if (TD_ON_LOCK(td)) {
1170 		kp->ki_kiflag |= KI_LOCKBLOCK;
1171 		strlcpy(kp->ki_lockname, td->td_lockname,
1172 		    sizeof(kp->ki_lockname));
1173 	} else {
1174 		kp->ki_kiflag &= ~KI_LOCKBLOCK;
1175 		bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
1176 	}
1177 
1178 	if (p->p_state == PRS_NORMAL) { /* approximate. */
1179 		if (TD_ON_RUNQ(td) ||
1180 		    TD_CAN_RUN(td) ||
1181 		    TD_IS_RUNNING(td)) {
1182 			kp->ki_stat = SRUN;
1183 		} else if (P_SHOULDSTOP(p)) {
1184 			kp->ki_stat = SSTOP;
1185 		} else if (TD_IS_SLEEPING(td)) {
1186 			kp->ki_stat = SSLEEP;
1187 		} else if (TD_ON_LOCK(td)) {
1188 			kp->ki_stat = SLOCK;
1189 		} else {
1190 			kp->ki_stat = SWAIT;
1191 		}
1192 	} else if (p->p_state == PRS_ZOMBIE) {
1193 		kp->ki_stat = SZOMB;
1194 	} else {
1195 		kp->ki_stat = SIDL;
1196 	}
1197 
1198 	/* Things in the thread */
1199 	kp->ki_wchan = td->td_wchan;
1200 	kp->ki_pri.pri_level = td->td_priority;
1201 	kp->ki_pri.pri_native = td->td_base_pri;
1202 
1203 	/*
1204 	 * Note: legacy fields; clamp at the old NOCPU value and/or
1205 	 * the maximum u_char CPU value.
1206 	 */
1207 	if (td->td_lastcpu == NOCPU)
1208 		kp->ki_lastcpu_old = NOCPU_OLD;
1209 	else if (td->td_lastcpu > MAXCPU_OLD)
1210 		kp->ki_lastcpu_old = MAXCPU_OLD;
1211 	else
1212 		kp->ki_lastcpu_old = td->td_lastcpu;
1213 
1214 	if (td->td_oncpu == NOCPU)
1215 		kp->ki_oncpu_old = NOCPU_OLD;
1216 	else if (td->td_oncpu > MAXCPU_OLD)
1217 		kp->ki_oncpu_old = MAXCPU_OLD;
1218 	else
1219 		kp->ki_oncpu_old = td->td_oncpu;
1220 
1221 	kp->ki_lastcpu = td->td_lastcpu;
1222 	kp->ki_oncpu = td->td_oncpu;
1223 	kp->ki_tdflags = td->td_flags;
1224 	kp->ki_tid = td->td_tid;
1225 	kp->ki_numthreads = p->p_numthreads;
1226 	kp->ki_pcb = td->td_pcb;
1227 	kp->ki_kstack = (void *)td->td_kstack;
1228 	kp->ki_slptime = (ticks - td->td_slptick) / hz;
1229 	kp->ki_pri.pri_class = td->td_pri_class;
1230 	kp->ki_pri.pri_user = td->td_user_pri;
1231 
1232 	if (preferthread) {
1233 		rufetchtd(td, &kp->ki_rusage);
1234 		kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
1235 		kp->ki_pctcpu = sched_pctcpu(td);
1236 		kp->ki_estcpu = sched_estcpu(td);
1237 		kp->ki_cow = td->td_cow;
1238 	}
1239 
1240 	/* We can't get this anymore but ps etc never used it anyway. */
1241 	kp->ki_rqindex = 0;
1242 
1243 	if (preferthread)
1244 		kp->ki_siglist = td->td_siglist;
1245 	kp->ki_sigmask = td->td_sigmask;
1246 	thread_unlock(td);
1247 	if (preferthread)
1248 		PROC_STATUNLOCK(p);
1249 }
1250 
1251 /*
1252  * Fill in a kinfo_proc structure for the specified process.
1253  * Must be called with the target process locked.
1254  */
1255 void
1256 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1257 {
1258 
1259 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1260 
1261 	fill_kinfo_proc_only(p, kp);
1262 	fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1263 	fill_kinfo_aggregate(p, kp);
1264 }
1265 
1266 struct pstats *
1267 pstats_alloc(void)
1268 {
1269 
1270 	return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1271 }
1272 
1273 /*
1274  * Copy parts of p_stats; zero the rest of p_stats (statistics).
1275  */
1276 void
1277 pstats_fork(struct pstats *src, struct pstats *dst)
1278 {
1279 
1280 	bzero(&dst->pstat_startzero,
1281 	    __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1282 	bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1283 	    __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1284 }
1285 
1286 void
1287 pstats_free(struct pstats *ps)
1288 {
1289 
1290 	free(ps, M_SUBPROC);
1291 }
1292 
1293 #ifdef COMPAT_FREEBSD32
1294 
1295 /*
1296  * This function is typically used to copy out the kernel address, so
1297  * it can be replaced by assignment of zero.
1298  */
1299 static inline uint32_t
1300 ptr32_trim(void *ptr)
1301 {
1302 	uintptr_t uptr;
1303 
1304 	uptr = (uintptr_t)ptr;
1305 	return ((uptr > UINT_MAX) ? 0 : uptr);
1306 }
1307 
1308 #define PTRTRIM_CP(src,dst,fld) \
1309 	do { (dst).fld = ptr32_trim((src).fld); } while (0)
1310 
1311 static void
1312 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1313 {
1314 	int i;
1315 
1316 	bzero(ki32, sizeof(struct kinfo_proc32));
1317 	ki32->ki_structsize = sizeof(struct kinfo_proc32);
1318 	CP(*ki, *ki32, ki_layout);
1319 	PTRTRIM_CP(*ki, *ki32, ki_args);
1320 	PTRTRIM_CP(*ki, *ki32, ki_paddr);
1321 	PTRTRIM_CP(*ki, *ki32, ki_addr);
1322 	PTRTRIM_CP(*ki, *ki32, ki_tracep);
1323 	PTRTRIM_CP(*ki, *ki32, ki_textvp);
1324 	PTRTRIM_CP(*ki, *ki32, ki_fd);
1325 	PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1326 	PTRTRIM_CP(*ki, *ki32, ki_wchan);
1327 	CP(*ki, *ki32, ki_pid);
1328 	CP(*ki, *ki32, ki_ppid);
1329 	CP(*ki, *ki32, ki_pgid);
1330 	CP(*ki, *ki32, ki_tpgid);
1331 	CP(*ki, *ki32, ki_sid);
1332 	CP(*ki, *ki32, ki_tsid);
1333 	CP(*ki, *ki32, ki_jobc);
1334 	CP(*ki, *ki32, ki_tdev);
1335 	CP(*ki, *ki32, ki_tdev_freebsd11);
1336 	CP(*ki, *ki32, ki_siglist);
1337 	CP(*ki, *ki32, ki_sigmask);
1338 	CP(*ki, *ki32, ki_sigignore);
1339 	CP(*ki, *ki32, ki_sigcatch);
1340 	CP(*ki, *ki32, ki_uid);
1341 	CP(*ki, *ki32, ki_ruid);
1342 	CP(*ki, *ki32, ki_svuid);
1343 	CP(*ki, *ki32, ki_rgid);
1344 	CP(*ki, *ki32, ki_svgid);
1345 	CP(*ki, *ki32, ki_ngroups);
1346 	for (i = 0; i < KI_NGROUPS; i++)
1347 		CP(*ki, *ki32, ki_groups[i]);
1348 	CP(*ki, *ki32, ki_size);
1349 	CP(*ki, *ki32, ki_rssize);
1350 	CP(*ki, *ki32, ki_swrss);
1351 	CP(*ki, *ki32, ki_tsize);
1352 	CP(*ki, *ki32, ki_dsize);
1353 	CP(*ki, *ki32, ki_ssize);
1354 	CP(*ki, *ki32, ki_xstat);
1355 	CP(*ki, *ki32, ki_acflag);
1356 	CP(*ki, *ki32, ki_pctcpu);
1357 	CP(*ki, *ki32, ki_estcpu);
1358 	CP(*ki, *ki32, ki_slptime);
1359 	CP(*ki, *ki32, ki_swtime);
1360 	CP(*ki, *ki32, ki_cow);
1361 	CP(*ki, *ki32, ki_runtime);
1362 	TV_CP(*ki, *ki32, ki_start);
1363 	TV_CP(*ki, *ki32, ki_childtime);
1364 	CP(*ki, *ki32, ki_flag);
1365 	CP(*ki, *ki32, ki_kiflag);
1366 	CP(*ki, *ki32, ki_traceflag);
1367 	CP(*ki, *ki32, ki_stat);
1368 	CP(*ki, *ki32, ki_nice);
1369 	CP(*ki, *ki32, ki_lock);
1370 	CP(*ki, *ki32, ki_rqindex);
1371 	CP(*ki, *ki32, ki_oncpu);
1372 	CP(*ki, *ki32, ki_lastcpu);
1373 
1374 	/* XXX TODO: wrap cpu value as appropriate */
1375 	CP(*ki, *ki32, ki_oncpu_old);
1376 	CP(*ki, *ki32, ki_lastcpu_old);
1377 
1378 	bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1379 	bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1380 	bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1381 	bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1382 	bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1383 	bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1384 	bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1385 	bcopy(ki->ki_moretdname, ki32->ki_moretdname, MAXCOMLEN - TDNAMLEN + 1);
1386 	CP(*ki, *ki32, ki_tracer);
1387 	CP(*ki, *ki32, ki_flag2);
1388 	CP(*ki, *ki32, ki_fibnum);
1389 	CP(*ki, *ki32, ki_cr_flags);
1390 	CP(*ki, *ki32, ki_jid);
1391 	CP(*ki, *ki32, ki_numthreads);
1392 	CP(*ki, *ki32, ki_tid);
1393 	CP(*ki, *ki32, ki_pri);
1394 	freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1395 	freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1396 	PTRTRIM_CP(*ki, *ki32, ki_pcb);
1397 	PTRTRIM_CP(*ki, *ki32, ki_kstack);
1398 	PTRTRIM_CP(*ki, *ki32, ki_udata);
1399 	PTRTRIM_CP(*ki, *ki32, ki_tdaddr);
1400 	CP(*ki, *ki32, ki_sflag);
1401 	CP(*ki, *ki32, ki_tdflags);
1402 }
1403 #endif
1404 
1405 static ssize_t
1406 kern_proc_out_size(struct proc *p, int flags)
1407 {
1408 	ssize_t size = 0;
1409 
1410 	PROC_LOCK_ASSERT(p, MA_OWNED);
1411 
1412 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1413 #ifdef COMPAT_FREEBSD32
1414 		if ((flags & KERN_PROC_MASK32) != 0) {
1415 			size += sizeof(struct kinfo_proc32);
1416 		} else
1417 #endif
1418 			size += sizeof(struct kinfo_proc);
1419 	} else {
1420 #ifdef COMPAT_FREEBSD32
1421 		if ((flags & KERN_PROC_MASK32) != 0)
1422 			size += sizeof(struct kinfo_proc32) * p->p_numthreads;
1423 		else
1424 #endif
1425 			size += sizeof(struct kinfo_proc) * p->p_numthreads;
1426 	}
1427 	PROC_UNLOCK(p);
1428 	return (size);
1429 }
1430 
1431 int
1432 kern_proc_out(struct proc *p, struct sbuf *sb, int flags)
1433 {
1434 	struct thread *td;
1435 	struct kinfo_proc ki;
1436 #ifdef COMPAT_FREEBSD32
1437 	struct kinfo_proc32 ki32;
1438 #endif
1439 	int error;
1440 
1441 	PROC_LOCK_ASSERT(p, MA_OWNED);
1442 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1443 
1444 	error = 0;
1445 	fill_kinfo_proc(p, &ki);
1446 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1447 #ifdef COMPAT_FREEBSD32
1448 		if ((flags & KERN_PROC_MASK32) != 0) {
1449 			freebsd32_kinfo_proc_out(&ki, &ki32);
1450 			if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1451 				error = ENOMEM;
1452 		} else
1453 #endif
1454 			if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1455 				error = ENOMEM;
1456 	} else {
1457 		FOREACH_THREAD_IN_PROC(p, td) {
1458 			fill_kinfo_thread(td, &ki, 1);
1459 #ifdef COMPAT_FREEBSD32
1460 			if ((flags & KERN_PROC_MASK32) != 0) {
1461 				freebsd32_kinfo_proc_out(&ki, &ki32);
1462 				if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1463 					error = ENOMEM;
1464 			} else
1465 #endif
1466 				if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1467 					error = ENOMEM;
1468 			if (error != 0)
1469 				break;
1470 		}
1471 	}
1472 	PROC_UNLOCK(p);
1473 	return (error);
1474 }
1475 
1476 static int
1477 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
1478 {
1479 	struct sbuf sb;
1480 	struct kinfo_proc ki;
1481 	int error, error2;
1482 
1483 	if (req->oldptr == NULL)
1484 		return (SYSCTL_OUT(req, 0, kern_proc_out_size(p, flags)));
1485 
1486 	sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req);
1487 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1488 	error = kern_proc_out(p, &sb, flags);
1489 	error2 = sbuf_finish(&sb);
1490 	sbuf_delete(&sb);
1491 	if (error != 0)
1492 		return (error);
1493 	else if (error2 != 0)
1494 		return (error2);
1495 	return (0);
1496 }
1497 
1498 int
1499 proc_iterate(int (*cb)(struct proc *, void *), void *cbarg)
1500 {
1501 	struct proc *p;
1502 	int error, i, j;
1503 
1504 	for (i = 0; i < pidhashlock + 1; i++) {
1505 		sx_slock(&pidhashtbl_lock[i]);
1506 		for (j = i; j <= pidhash; j += pidhashlock + 1) {
1507 			LIST_FOREACH(p, &pidhashtbl[j], p_hash) {
1508 				if (p->p_state == PRS_NEW)
1509 					continue;
1510 				error = cb(p, cbarg);
1511 				PROC_LOCK_ASSERT(p, MA_NOTOWNED);
1512 				if (error != 0) {
1513 					sx_sunlock(&pidhashtbl_lock[i]);
1514 					return (error);
1515 				}
1516 			}
1517 		}
1518 		sx_sunlock(&pidhashtbl_lock[i]);
1519 	}
1520 	return (0);
1521 }
1522 
1523 struct kern_proc_out_args {
1524 	struct sysctl_req *req;
1525 	int flags;
1526 	int oid_number;
1527 	int *name;
1528 };
1529 
1530 static int
1531 sysctl_kern_proc_iterate(struct proc *p, void *origarg)
1532 {
1533 	struct kern_proc_out_args *arg = origarg;
1534 	int *name = arg->name;
1535 	int oid_number = arg->oid_number;
1536 	int flags = arg->flags;
1537 	struct sysctl_req *req = arg->req;
1538 	int error = 0;
1539 
1540 	PROC_LOCK(p);
1541 
1542 	KASSERT(p->p_ucred != NULL,
1543 	    ("process credential is NULL for non-NEW proc"));
1544 	/*
1545 	 * Show a user only appropriate processes.
1546 	 */
1547 	if (p_cansee(curthread, p))
1548 		goto skip;
1549 	/*
1550 	 * TODO - make more efficient (see notes below).
1551 	 * do by session.
1552 	 */
1553 	switch (oid_number) {
1554 
1555 	case KERN_PROC_GID:
1556 		if (p->p_ucred->cr_gid != (gid_t)name[0])
1557 			goto skip;
1558 		break;
1559 
1560 	case KERN_PROC_PGRP:
1561 		/* could do this by traversing pgrp */
1562 		if (p->p_pgrp == NULL ||
1563 		    p->p_pgrp->pg_id != (pid_t)name[0])
1564 			goto skip;
1565 		break;
1566 
1567 	case KERN_PROC_RGID:
1568 		if (p->p_ucred->cr_rgid != (gid_t)name[0])
1569 			goto skip;
1570 		break;
1571 
1572 	case KERN_PROC_SESSION:
1573 		if (p->p_session == NULL ||
1574 		    p->p_session->s_sid != (pid_t)name[0])
1575 			goto skip;
1576 		break;
1577 
1578 	case KERN_PROC_TTY:
1579 		if ((p->p_flag & P_CONTROLT) == 0 ||
1580 		    p->p_session == NULL)
1581 			goto skip;
1582 		/* XXX proctree_lock */
1583 		SESS_LOCK(p->p_session);
1584 		if (p->p_session->s_ttyp == NULL ||
1585 		    tty_udev(p->p_session->s_ttyp) !=
1586 		    (dev_t)name[0]) {
1587 			SESS_UNLOCK(p->p_session);
1588 			goto skip;
1589 		}
1590 		SESS_UNLOCK(p->p_session);
1591 		break;
1592 
1593 	case KERN_PROC_UID:
1594 		if (p->p_ucred->cr_uid != (uid_t)name[0])
1595 			goto skip;
1596 		break;
1597 
1598 	case KERN_PROC_RUID:
1599 		if (p->p_ucred->cr_ruid != (uid_t)name[0])
1600 			goto skip;
1601 		break;
1602 
1603 	case KERN_PROC_PROC:
1604 		break;
1605 
1606 	default:
1607 		break;
1608 
1609 	}
1610 	error = sysctl_out_proc(p, req, flags);
1611 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
1612 	return (error);
1613 skip:
1614 	PROC_UNLOCK(p);
1615 	return (0);
1616 }
1617 
1618 static int
1619 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1620 {
1621 	struct kern_proc_out_args iterarg;
1622 	int *name = (int *)arg1;
1623 	u_int namelen = arg2;
1624 	struct proc *p;
1625 	int flags, oid_number;
1626 	int error = 0;
1627 
1628 	oid_number = oidp->oid_number;
1629 	if (oid_number != KERN_PROC_ALL &&
1630 	    (oid_number & KERN_PROC_INC_THREAD) == 0)
1631 		flags = KERN_PROC_NOTHREADS;
1632 	else {
1633 		flags = 0;
1634 		oid_number &= ~KERN_PROC_INC_THREAD;
1635 	}
1636 #ifdef COMPAT_FREEBSD32
1637 	if (req->flags & SCTL_MASK32)
1638 		flags |= KERN_PROC_MASK32;
1639 #endif
1640 	if (oid_number == KERN_PROC_PID) {
1641 		if (namelen != 1)
1642 			return (EINVAL);
1643 		error = sysctl_wire_old_buffer(req, 0);
1644 		if (error)
1645 			return (error);
1646 		error = pget((pid_t)name[0], PGET_CANSEE, &p);
1647 		if (error == 0)
1648 			error = sysctl_out_proc(p, req, flags);
1649 		return (error);
1650 	}
1651 
1652 	switch (oid_number) {
1653 	case KERN_PROC_ALL:
1654 		if (namelen != 0)
1655 			return (EINVAL);
1656 		break;
1657 	case KERN_PROC_PROC:
1658 		if (namelen != 0 && namelen != 1)
1659 			return (EINVAL);
1660 		break;
1661 	default:
1662 		if (namelen != 1)
1663 			return (EINVAL);
1664 		break;
1665 	}
1666 
1667 	if (req->oldptr == NULL) {
1668 		/* overestimate by 5 procs */
1669 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1670 		if (error)
1671 			return (error);
1672 	} else {
1673 		error = sysctl_wire_old_buffer(req, 0);
1674 		if (error != 0)
1675 			return (error);
1676 	}
1677 	iterarg.flags = flags;
1678 	iterarg.oid_number = oid_number;
1679 	iterarg.req = req;
1680 	iterarg.name = name;
1681 	error = proc_iterate(sysctl_kern_proc_iterate, &iterarg);
1682 	return (error);
1683 }
1684 
1685 struct pargs *
1686 pargs_alloc(int len)
1687 {
1688 	struct pargs *pa;
1689 
1690 	pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1691 		M_WAITOK);
1692 	refcount_init(&pa->ar_ref, 1);
1693 	pa->ar_length = len;
1694 	return (pa);
1695 }
1696 
1697 static void
1698 pargs_free(struct pargs *pa)
1699 {
1700 
1701 	free(pa, M_PARGS);
1702 }
1703 
1704 void
1705 pargs_hold(struct pargs *pa)
1706 {
1707 
1708 	if (pa == NULL)
1709 		return;
1710 	refcount_acquire(&pa->ar_ref);
1711 }
1712 
1713 void
1714 pargs_drop(struct pargs *pa)
1715 {
1716 
1717 	if (pa == NULL)
1718 		return;
1719 	if (refcount_release(&pa->ar_ref))
1720 		pargs_free(pa);
1721 }
1722 
1723 static int
1724 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1725     size_t len)
1726 {
1727 	ssize_t n;
1728 
1729 	/*
1730 	 * This may return a short read if the string is shorter than the chunk
1731 	 * and is aligned at the end of the page, and the following page is not
1732 	 * mapped.
1733 	 */
1734 	n = proc_readmem(td, p, (vm_offset_t)sptr, buf, len);
1735 	if (n <= 0)
1736 		return (ENOMEM);
1737 	return (0);
1738 }
1739 
1740 #define PROC_AUXV_MAX	256	/* Safety limit on auxv size. */
1741 
1742 enum proc_vector_type {
1743 	PROC_ARG,
1744 	PROC_ENV,
1745 	PROC_AUX,
1746 };
1747 
1748 #ifdef COMPAT_FREEBSD32
1749 static int
1750 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1751     size_t *vsizep, enum proc_vector_type type)
1752 {
1753 	struct freebsd32_ps_strings pss;
1754 	Elf32_Auxinfo aux;
1755 	vm_offset_t vptr, ptr;
1756 	uint32_t *proc_vector32;
1757 	char **proc_vector;
1758 	size_t vsize, size;
1759 	int i, error;
1760 
1761 	error = 0;
1762 	if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss,
1763 	    sizeof(pss)) != sizeof(pss))
1764 		return (ENOMEM);
1765 	switch (type) {
1766 	case PROC_ARG:
1767 		vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1768 		vsize = pss.ps_nargvstr;
1769 		if (vsize > ARG_MAX)
1770 			return (ENOEXEC);
1771 		size = vsize * sizeof(int32_t);
1772 		break;
1773 	case PROC_ENV:
1774 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1775 		vsize = pss.ps_nenvstr;
1776 		if (vsize > ARG_MAX)
1777 			return (ENOEXEC);
1778 		size = vsize * sizeof(int32_t);
1779 		break;
1780 	case PROC_AUX:
1781 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1782 		    (pss.ps_nenvstr + 1) * sizeof(int32_t);
1783 		if (vptr % 4 != 0)
1784 			return (ENOEXEC);
1785 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1786 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1787 			    sizeof(aux))
1788 				return (ENOMEM);
1789 			if (aux.a_type == AT_NULL)
1790 				break;
1791 			ptr += sizeof(aux);
1792 		}
1793 		if (aux.a_type != AT_NULL)
1794 			return (ENOEXEC);
1795 		vsize = i + 1;
1796 		size = vsize * sizeof(aux);
1797 		break;
1798 	default:
1799 		KASSERT(0, ("Wrong proc vector type: %d", type));
1800 		return (EINVAL);
1801 	}
1802 	proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1803 	if (proc_readmem(td, p, vptr, proc_vector32, size) != size) {
1804 		error = ENOMEM;
1805 		goto done;
1806 	}
1807 	if (type == PROC_AUX) {
1808 		*proc_vectorp = (char **)proc_vector32;
1809 		*vsizep = vsize;
1810 		return (0);
1811 	}
1812 	proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1813 	for (i = 0; i < (int)vsize; i++)
1814 		proc_vector[i] = PTRIN(proc_vector32[i]);
1815 	*proc_vectorp = proc_vector;
1816 	*vsizep = vsize;
1817 done:
1818 	free(proc_vector32, M_TEMP);
1819 	return (error);
1820 }
1821 #endif
1822 
1823 static int
1824 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1825     size_t *vsizep, enum proc_vector_type type)
1826 {
1827 	struct ps_strings pss;
1828 	Elf_Auxinfo aux;
1829 	vm_offset_t vptr, ptr;
1830 	char **proc_vector;
1831 	size_t vsize, size;
1832 	int i;
1833 
1834 #ifdef COMPAT_FREEBSD32
1835 	if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1836 		return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1837 #endif
1838 	if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss,
1839 	    sizeof(pss)) != sizeof(pss))
1840 		return (ENOMEM);
1841 	switch (type) {
1842 	case PROC_ARG:
1843 		vptr = (vm_offset_t)pss.ps_argvstr;
1844 		vsize = pss.ps_nargvstr;
1845 		if (vsize > ARG_MAX)
1846 			return (ENOEXEC);
1847 		size = vsize * sizeof(char *);
1848 		break;
1849 	case PROC_ENV:
1850 		vptr = (vm_offset_t)pss.ps_envstr;
1851 		vsize = pss.ps_nenvstr;
1852 		if (vsize > ARG_MAX)
1853 			return (ENOEXEC);
1854 		size = vsize * sizeof(char *);
1855 		break;
1856 	case PROC_AUX:
1857 		/*
1858 		 * The aux array is just above env array on the stack. Check
1859 		 * that the address is naturally aligned.
1860 		 */
1861 		vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1862 		    * sizeof(char *);
1863 #if __ELF_WORD_SIZE == 64
1864 		if (vptr % sizeof(uint64_t) != 0)
1865 #else
1866 		if (vptr % sizeof(uint32_t) != 0)
1867 #endif
1868 			return (ENOEXEC);
1869 		/*
1870 		 * We count the array size reading the aux vectors from the
1871 		 * stack until AT_NULL vector is returned.  So (to keep the code
1872 		 * simple) we read the process stack twice: the first time here
1873 		 * to find the size and the second time when copying the vectors
1874 		 * to the allocated proc_vector.
1875 		 */
1876 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1877 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1878 			    sizeof(aux))
1879 				return (ENOMEM);
1880 			if (aux.a_type == AT_NULL)
1881 				break;
1882 			ptr += sizeof(aux);
1883 		}
1884 		/*
1885 		 * If the PROC_AUXV_MAX entries are iterated over, and we have
1886 		 * not reached AT_NULL, it is most likely we are reading wrong
1887 		 * data: either the process doesn't have auxv array or data has
1888 		 * been modified. Return the error in this case.
1889 		 */
1890 		if (aux.a_type != AT_NULL)
1891 			return (ENOEXEC);
1892 		vsize = i + 1;
1893 		size = vsize * sizeof(aux);
1894 		break;
1895 	default:
1896 		KASSERT(0, ("Wrong proc vector type: %d", type));
1897 		return (EINVAL); /* In case we are built without INVARIANTS. */
1898 	}
1899 	proc_vector = malloc(size, M_TEMP, M_WAITOK);
1900 	if (proc_readmem(td, p, vptr, proc_vector, size) != size) {
1901 		free(proc_vector, M_TEMP);
1902 		return (ENOMEM);
1903 	}
1904 	*proc_vectorp = proc_vector;
1905 	*vsizep = vsize;
1906 
1907 	return (0);
1908 }
1909 
1910 #define GET_PS_STRINGS_CHUNK_SZ	256	/* Chunk size (bytes) for ps_strings operations. */
1911 
1912 static int
1913 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
1914     enum proc_vector_type type)
1915 {
1916 	size_t done, len, nchr, vsize;
1917 	int error, i;
1918 	char **proc_vector, *sptr;
1919 	char pss_string[GET_PS_STRINGS_CHUNK_SZ];
1920 
1921 	PROC_ASSERT_HELD(p);
1922 
1923 	/*
1924 	 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
1925 	 */
1926 	nchr = 2 * (PATH_MAX + ARG_MAX);
1927 
1928 	error = get_proc_vector(td, p, &proc_vector, &vsize, type);
1929 	if (error != 0)
1930 		return (error);
1931 	for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
1932 		/*
1933 		 * The program may have scribbled into its argv array, e.g. to
1934 		 * remove some arguments.  If that has happened, break out
1935 		 * before trying to read from NULL.
1936 		 */
1937 		if (proc_vector[i] == NULL)
1938 			break;
1939 		for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
1940 			error = proc_read_string(td, p, sptr, pss_string,
1941 			    sizeof(pss_string));
1942 			if (error != 0)
1943 				goto done;
1944 			len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
1945 			if (done + len >= nchr)
1946 				len = nchr - done - 1;
1947 			sbuf_bcat(sb, pss_string, len);
1948 			if (len != GET_PS_STRINGS_CHUNK_SZ)
1949 				break;
1950 			done += GET_PS_STRINGS_CHUNK_SZ;
1951 		}
1952 		sbuf_bcat(sb, "", 1);
1953 		done += len + 1;
1954 	}
1955 done:
1956 	free(proc_vector, M_TEMP);
1957 	return (error);
1958 }
1959 
1960 int
1961 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
1962 {
1963 
1964 	return (get_ps_strings(curthread, p, sb, PROC_ARG));
1965 }
1966 
1967 int
1968 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
1969 {
1970 
1971 	return (get_ps_strings(curthread, p, sb, PROC_ENV));
1972 }
1973 
1974 int
1975 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb)
1976 {
1977 	size_t vsize, size;
1978 	char **auxv;
1979 	int error;
1980 
1981 	error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX);
1982 	if (error == 0) {
1983 #ifdef COMPAT_FREEBSD32
1984 		if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1985 			size = vsize * sizeof(Elf32_Auxinfo);
1986 		else
1987 #endif
1988 			size = vsize * sizeof(Elf_Auxinfo);
1989 		if (sbuf_bcat(sb, auxv, size) != 0)
1990 			error = ENOMEM;
1991 		free(auxv, M_TEMP);
1992 	}
1993 	return (error);
1994 }
1995 
1996 /*
1997  * This sysctl allows a process to retrieve the argument list or process
1998  * title for another process without groping around in the address space
1999  * of the other process.  It also allow a process to set its own "process
2000  * title to a string of its own choice.
2001  */
2002 static int
2003 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
2004 {
2005 	int *name = (int *)arg1;
2006 	u_int namelen = arg2;
2007 	struct pargs *newpa, *pa;
2008 	struct proc *p;
2009 	struct sbuf sb;
2010 	int flags, error = 0, error2;
2011 	pid_t pid;
2012 
2013 	if (namelen != 1)
2014 		return (EINVAL);
2015 
2016 	pid = (pid_t)name[0];
2017 	/*
2018 	 * If the query is for this process and it is single-threaded, there
2019 	 * is nobody to modify pargs, thus we can just read.
2020 	 */
2021 	p = curproc;
2022 	if (pid == p->p_pid && p->p_numthreads == 1 && req->newptr == NULL &&
2023 	    (pa = p->p_args) != NULL)
2024 		return (SYSCTL_OUT(req, pa->ar_args, pa->ar_length));
2025 
2026 	flags = PGET_CANSEE;
2027 	if (req->newptr != NULL)
2028 		flags |= PGET_ISCURRENT;
2029 	error = pget(pid, flags, &p);
2030 	if (error)
2031 		return (error);
2032 
2033 	pa = p->p_args;
2034 	if (pa != NULL) {
2035 		pargs_hold(pa);
2036 		PROC_UNLOCK(p);
2037 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
2038 		pargs_drop(pa);
2039 	} else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
2040 		_PHOLD(p);
2041 		PROC_UNLOCK(p);
2042 		sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2043 		sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2044 		error = proc_getargv(curthread, p, &sb);
2045 		error2 = sbuf_finish(&sb);
2046 		PRELE(p);
2047 		sbuf_delete(&sb);
2048 		if (error == 0 && error2 != 0)
2049 			error = error2;
2050 	} else {
2051 		PROC_UNLOCK(p);
2052 	}
2053 	if (error != 0 || req->newptr == NULL)
2054 		return (error);
2055 
2056 	if (req->newlen > ps_arg_cache_limit - sizeof(struct pargs))
2057 		return (ENOMEM);
2058 
2059 	if (req->newlen == 0) {
2060 		/*
2061 		 * Clear the argument pointer, so that we'll fetch arguments
2062 		 * with proc_getargv() until further notice.
2063 		 */
2064 		newpa = NULL;
2065 	} else {
2066 		newpa = pargs_alloc(req->newlen);
2067 		error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
2068 		if (error != 0) {
2069 			pargs_free(newpa);
2070 			return (error);
2071 		}
2072 	}
2073 	PROC_LOCK(p);
2074 	pa = p->p_args;
2075 	p->p_args = newpa;
2076 	PROC_UNLOCK(p);
2077 	pargs_drop(pa);
2078 	return (0);
2079 }
2080 
2081 /*
2082  * This sysctl allows a process to retrieve environment of another process.
2083  */
2084 static int
2085 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
2086 {
2087 	int *name = (int *)arg1;
2088 	u_int namelen = arg2;
2089 	struct proc *p;
2090 	struct sbuf sb;
2091 	int error, error2;
2092 
2093 	if (namelen != 1)
2094 		return (EINVAL);
2095 
2096 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2097 	if (error != 0)
2098 		return (error);
2099 	if ((p->p_flag & P_SYSTEM) != 0) {
2100 		PRELE(p);
2101 		return (0);
2102 	}
2103 
2104 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2105 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2106 	error = proc_getenvv(curthread, p, &sb);
2107 	error2 = sbuf_finish(&sb);
2108 	PRELE(p);
2109 	sbuf_delete(&sb);
2110 	return (error != 0 ? error : error2);
2111 }
2112 
2113 /*
2114  * This sysctl allows a process to retrieve ELF auxiliary vector of
2115  * another process.
2116  */
2117 static int
2118 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
2119 {
2120 	int *name = (int *)arg1;
2121 	u_int namelen = arg2;
2122 	struct proc *p;
2123 	struct sbuf sb;
2124 	int error, error2;
2125 
2126 	if (namelen != 1)
2127 		return (EINVAL);
2128 
2129 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2130 	if (error != 0)
2131 		return (error);
2132 	if ((p->p_flag & P_SYSTEM) != 0) {
2133 		PRELE(p);
2134 		return (0);
2135 	}
2136 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2137 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2138 	error = proc_getauxv(curthread, p, &sb);
2139 	error2 = sbuf_finish(&sb);
2140 	PRELE(p);
2141 	sbuf_delete(&sb);
2142 	return (error != 0 ? error : error2);
2143 }
2144 
2145 /*
2146  * This sysctl allows a process to retrieve the path of the executable for
2147  * itself or another process.
2148  */
2149 static int
2150 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
2151 {
2152 	pid_t *pidp = (pid_t *)arg1;
2153 	unsigned int arglen = arg2;
2154 	struct proc *p;
2155 	struct vnode *vp;
2156 	char *retbuf, *freebuf;
2157 	int error;
2158 
2159 	if (arglen != 1)
2160 		return (EINVAL);
2161 	if (*pidp == -1) {	/* -1 means this process */
2162 		p = req->td->td_proc;
2163 	} else {
2164 		error = pget(*pidp, PGET_CANSEE, &p);
2165 		if (error != 0)
2166 			return (error);
2167 	}
2168 
2169 	vp = p->p_textvp;
2170 	if (vp == NULL) {
2171 		if (*pidp != -1)
2172 			PROC_UNLOCK(p);
2173 		return (0);
2174 	}
2175 	vref(vp);
2176 	if (*pidp != -1)
2177 		PROC_UNLOCK(p);
2178 	error = vn_fullpath(req->td, vp, &retbuf, &freebuf);
2179 	vrele(vp);
2180 	if (error)
2181 		return (error);
2182 	error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
2183 	free(freebuf, M_TEMP);
2184 	return (error);
2185 }
2186 
2187 static int
2188 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
2189 {
2190 	struct proc *p;
2191 	char *sv_name;
2192 	int *name;
2193 	int namelen;
2194 	int error;
2195 
2196 	namelen = arg2;
2197 	if (namelen != 1)
2198 		return (EINVAL);
2199 
2200 	name = (int *)arg1;
2201 	error = pget((pid_t)name[0], PGET_CANSEE, &p);
2202 	if (error != 0)
2203 		return (error);
2204 	sv_name = p->p_sysent->sv_name;
2205 	PROC_UNLOCK(p);
2206 	return (sysctl_handle_string(oidp, sv_name, 0, req));
2207 }
2208 
2209 #ifdef KINFO_OVMENTRY_SIZE
2210 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
2211 #endif
2212 
2213 #ifdef COMPAT_FREEBSD7
2214 static int
2215 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
2216 {
2217 	vm_map_entry_t entry, tmp_entry;
2218 	unsigned int last_timestamp;
2219 	char *fullpath, *freepath;
2220 	struct kinfo_ovmentry *kve;
2221 	struct vattr va;
2222 	struct ucred *cred;
2223 	int error, *name;
2224 	struct vnode *vp;
2225 	struct proc *p;
2226 	vm_map_t map;
2227 	struct vmspace *vm;
2228 
2229 	name = (int *)arg1;
2230 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2231 	if (error != 0)
2232 		return (error);
2233 	vm = vmspace_acquire_ref(p);
2234 	if (vm == NULL) {
2235 		PRELE(p);
2236 		return (ESRCH);
2237 	}
2238 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2239 
2240 	map = &vm->vm_map;
2241 	vm_map_lock_read(map);
2242 	VM_MAP_ENTRY_FOREACH(entry, map) {
2243 		vm_object_t obj, tobj, lobj;
2244 		vm_offset_t addr;
2245 
2246 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2247 			continue;
2248 
2249 		bzero(kve, sizeof(*kve));
2250 		kve->kve_structsize = sizeof(*kve);
2251 
2252 		kve->kve_private_resident = 0;
2253 		obj = entry->object.vm_object;
2254 		if (obj != NULL) {
2255 			VM_OBJECT_RLOCK(obj);
2256 			if (obj->shadow_count == 1)
2257 				kve->kve_private_resident =
2258 				    obj->resident_page_count;
2259 		}
2260 		kve->kve_resident = 0;
2261 		addr = entry->start;
2262 		while (addr < entry->end) {
2263 			if (pmap_extract(map->pmap, addr))
2264 				kve->kve_resident++;
2265 			addr += PAGE_SIZE;
2266 		}
2267 
2268 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2269 			if (tobj != obj) {
2270 				VM_OBJECT_RLOCK(tobj);
2271 				kve->kve_offset += tobj->backing_object_offset;
2272 			}
2273 			if (lobj != obj)
2274 				VM_OBJECT_RUNLOCK(lobj);
2275 			lobj = tobj;
2276 		}
2277 
2278 		kve->kve_start = (void*)entry->start;
2279 		kve->kve_end = (void*)entry->end;
2280 		kve->kve_offset += (off_t)entry->offset;
2281 
2282 		if (entry->protection & VM_PROT_READ)
2283 			kve->kve_protection |= KVME_PROT_READ;
2284 		if (entry->protection & VM_PROT_WRITE)
2285 			kve->kve_protection |= KVME_PROT_WRITE;
2286 		if (entry->protection & VM_PROT_EXECUTE)
2287 			kve->kve_protection |= KVME_PROT_EXEC;
2288 
2289 		if (entry->eflags & MAP_ENTRY_COW)
2290 			kve->kve_flags |= KVME_FLAG_COW;
2291 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2292 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2293 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2294 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2295 
2296 		last_timestamp = map->timestamp;
2297 		vm_map_unlock_read(map);
2298 
2299 		kve->kve_fileid = 0;
2300 		kve->kve_fsid = 0;
2301 		freepath = NULL;
2302 		fullpath = "";
2303 		if (lobj) {
2304 			kve->kve_type = vm_object_kvme_type(lobj, &vp);
2305 			if (kve->kve_type == KVME_TYPE_MGTDEVICE)
2306 				kve->kve_type = KVME_TYPE_UNKNOWN;
2307 			if (vp != NULL)
2308 				vref(vp);
2309 			if (lobj != obj)
2310 				VM_OBJECT_RUNLOCK(lobj);
2311 
2312 			kve->kve_ref_count = obj->ref_count;
2313 			kve->kve_shadow_count = obj->shadow_count;
2314 			VM_OBJECT_RUNLOCK(obj);
2315 			if (vp != NULL) {
2316 				vn_fullpath(curthread, vp, &fullpath,
2317 				    &freepath);
2318 				cred = curthread->td_ucred;
2319 				vn_lock(vp, LK_SHARED | LK_RETRY);
2320 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2321 					kve->kve_fileid = va.va_fileid;
2322 					/* truncate */
2323 					kve->kve_fsid = va.va_fsid;
2324 				}
2325 				vput(vp);
2326 			}
2327 		} else {
2328 			kve->kve_type = KVME_TYPE_NONE;
2329 			kve->kve_ref_count = 0;
2330 			kve->kve_shadow_count = 0;
2331 		}
2332 
2333 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2334 		if (freepath != NULL)
2335 			free(freepath, M_TEMP);
2336 
2337 		error = SYSCTL_OUT(req, kve, sizeof(*kve));
2338 		vm_map_lock_read(map);
2339 		if (error)
2340 			break;
2341 		if (last_timestamp != map->timestamp) {
2342 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2343 			entry = tmp_entry;
2344 		}
2345 	}
2346 	vm_map_unlock_read(map);
2347 	vmspace_free(vm);
2348 	PRELE(p);
2349 	free(kve, M_TEMP);
2350 	return (error);
2351 }
2352 #endif	/* COMPAT_FREEBSD7 */
2353 
2354 #ifdef KINFO_VMENTRY_SIZE
2355 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2356 #endif
2357 
2358 void
2359 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry,
2360     int *resident_count, bool *super)
2361 {
2362 	vm_object_t obj, tobj;
2363 	vm_page_t m, m_adv;
2364 	vm_offset_t addr;
2365 	vm_paddr_t pa;
2366 	vm_pindex_t pi, pi_adv, pindex;
2367 
2368 	*super = false;
2369 	*resident_count = 0;
2370 	if (vmmap_skip_res_cnt)
2371 		return;
2372 
2373 	pa = 0;
2374 	obj = entry->object.vm_object;
2375 	addr = entry->start;
2376 	m_adv = NULL;
2377 	pi = OFF_TO_IDX(entry->offset);
2378 	for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) {
2379 		if (m_adv != NULL) {
2380 			m = m_adv;
2381 		} else {
2382 			pi_adv = atop(entry->end - addr);
2383 			pindex = pi;
2384 			for (tobj = obj;; tobj = tobj->backing_object) {
2385 				m = vm_page_find_least(tobj, pindex);
2386 				if (m != NULL) {
2387 					if (m->pindex == pindex)
2388 						break;
2389 					if (pi_adv > m->pindex - pindex) {
2390 						pi_adv = m->pindex - pindex;
2391 						m_adv = m;
2392 					}
2393 				}
2394 				if (tobj->backing_object == NULL)
2395 					goto next;
2396 				pindex += OFF_TO_IDX(tobj->
2397 				    backing_object_offset);
2398 			}
2399 		}
2400 		m_adv = NULL;
2401 		if (m->psind != 0 && addr + pagesizes[1] <= entry->end &&
2402 		    (addr & (pagesizes[1] - 1)) == 0 &&
2403 		    (pmap_mincore(map->pmap, addr, &pa) & MINCORE_SUPER) != 0) {
2404 			*super = true;
2405 			pi_adv = atop(pagesizes[1]);
2406 		} else {
2407 			/*
2408 			 * We do not test the found page on validity.
2409 			 * Either the page is busy and being paged in,
2410 			 * or it was invalidated.  The first case
2411 			 * should be counted as resident, the second
2412 			 * is not so clear; we do account both.
2413 			 */
2414 			pi_adv = 1;
2415 		}
2416 		*resident_count += pi_adv;
2417 next:;
2418 	}
2419 }
2420 
2421 /*
2422  * Must be called with the process locked and will return unlocked.
2423  */
2424 int
2425 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb, ssize_t maxlen, int flags)
2426 {
2427 	vm_map_entry_t entry, tmp_entry;
2428 	struct vattr va;
2429 	vm_map_t map;
2430 	vm_object_t obj, tobj, lobj;
2431 	char *fullpath, *freepath;
2432 	struct kinfo_vmentry *kve;
2433 	struct ucred *cred;
2434 	struct vnode *vp;
2435 	struct vmspace *vm;
2436 	vm_offset_t addr;
2437 	unsigned int last_timestamp;
2438 	int error;
2439 	bool super;
2440 
2441 	PROC_LOCK_ASSERT(p, MA_OWNED);
2442 
2443 	_PHOLD(p);
2444 	PROC_UNLOCK(p);
2445 	vm = vmspace_acquire_ref(p);
2446 	if (vm == NULL) {
2447 		PRELE(p);
2448 		return (ESRCH);
2449 	}
2450 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK | M_ZERO);
2451 
2452 	error = 0;
2453 	map = &vm->vm_map;
2454 	vm_map_lock_read(map);
2455 	VM_MAP_ENTRY_FOREACH(entry, map) {
2456 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2457 			continue;
2458 
2459 		addr = entry->end;
2460 		bzero(kve, sizeof(*kve));
2461 		obj = entry->object.vm_object;
2462 		if (obj != NULL) {
2463 			for (tobj = obj; tobj != NULL;
2464 			    tobj = tobj->backing_object) {
2465 				VM_OBJECT_RLOCK(tobj);
2466 				kve->kve_offset += tobj->backing_object_offset;
2467 				lobj = tobj;
2468 			}
2469 			if (obj->backing_object == NULL)
2470 				kve->kve_private_resident =
2471 				    obj->resident_page_count;
2472 			kern_proc_vmmap_resident(map, entry,
2473 			    &kve->kve_resident, &super);
2474 			if (super)
2475 				kve->kve_flags |= KVME_FLAG_SUPER;
2476 			for (tobj = obj; tobj != NULL;
2477 			    tobj = tobj->backing_object) {
2478 				if (tobj != obj && tobj != lobj)
2479 					VM_OBJECT_RUNLOCK(tobj);
2480 			}
2481 		} else {
2482 			lobj = NULL;
2483 		}
2484 
2485 		kve->kve_start = entry->start;
2486 		kve->kve_end = entry->end;
2487 		kve->kve_offset += entry->offset;
2488 
2489 		if (entry->protection & VM_PROT_READ)
2490 			kve->kve_protection |= KVME_PROT_READ;
2491 		if (entry->protection & VM_PROT_WRITE)
2492 			kve->kve_protection |= KVME_PROT_WRITE;
2493 		if (entry->protection & VM_PROT_EXECUTE)
2494 			kve->kve_protection |= KVME_PROT_EXEC;
2495 
2496 		if (entry->eflags & MAP_ENTRY_COW)
2497 			kve->kve_flags |= KVME_FLAG_COW;
2498 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2499 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2500 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2501 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2502 		if (entry->eflags & MAP_ENTRY_GROWS_UP)
2503 			kve->kve_flags |= KVME_FLAG_GROWS_UP;
2504 		if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2505 			kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2506 		if (entry->eflags & MAP_ENTRY_USER_WIRED)
2507 			kve->kve_flags |= KVME_FLAG_USER_WIRED;
2508 
2509 		last_timestamp = map->timestamp;
2510 		vm_map_unlock_read(map);
2511 
2512 		freepath = NULL;
2513 		fullpath = "";
2514 		if (lobj != NULL) {
2515 			kve->kve_type = vm_object_kvme_type(lobj, &vp);
2516 			if (vp != NULL)
2517 				vref(vp);
2518 			if (lobj != obj)
2519 				VM_OBJECT_RUNLOCK(lobj);
2520 
2521 			kve->kve_ref_count = obj->ref_count;
2522 			kve->kve_shadow_count = obj->shadow_count;
2523 			VM_OBJECT_RUNLOCK(obj);
2524 			if (vp != NULL) {
2525 				vn_fullpath(curthread, vp, &fullpath,
2526 				    &freepath);
2527 				kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2528 				cred = curthread->td_ucred;
2529 				vn_lock(vp, LK_SHARED | LK_RETRY);
2530 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2531 					kve->kve_vn_fileid = va.va_fileid;
2532 					kve->kve_vn_fsid = va.va_fsid;
2533 					kve->kve_vn_fsid_freebsd11 =
2534 					    kve->kve_vn_fsid; /* truncate */
2535 					kve->kve_vn_mode =
2536 					    MAKEIMODE(va.va_type, va.va_mode);
2537 					kve->kve_vn_size = va.va_size;
2538 					kve->kve_vn_rdev = va.va_rdev;
2539 					kve->kve_vn_rdev_freebsd11 =
2540 					    kve->kve_vn_rdev; /* truncate */
2541 					kve->kve_status = KF_ATTR_VALID;
2542 				}
2543 				vput(vp);
2544 			}
2545 		} else {
2546 			kve->kve_type = KVME_TYPE_NONE;
2547 			kve->kve_ref_count = 0;
2548 			kve->kve_shadow_count = 0;
2549 		}
2550 
2551 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2552 		if (freepath != NULL)
2553 			free(freepath, M_TEMP);
2554 
2555 		/* Pack record size down */
2556 		if ((flags & KERN_VMMAP_PACK_KINFO) != 0)
2557 			kve->kve_structsize =
2558 			    offsetof(struct kinfo_vmentry, kve_path) +
2559 			    strlen(kve->kve_path) + 1;
2560 		else
2561 			kve->kve_structsize = sizeof(*kve);
2562 		kve->kve_structsize = roundup(kve->kve_structsize,
2563 		    sizeof(uint64_t));
2564 
2565 		/* Halt filling and truncate rather than exceeding maxlen */
2566 		if (maxlen != -1 && maxlen < kve->kve_structsize) {
2567 			error = 0;
2568 			vm_map_lock_read(map);
2569 			break;
2570 		} else if (maxlen != -1)
2571 			maxlen -= kve->kve_structsize;
2572 
2573 		if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0)
2574 			error = ENOMEM;
2575 		vm_map_lock_read(map);
2576 		if (error != 0)
2577 			break;
2578 		if (last_timestamp != map->timestamp) {
2579 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2580 			entry = tmp_entry;
2581 		}
2582 	}
2583 	vm_map_unlock_read(map);
2584 	vmspace_free(vm);
2585 	PRELE(p);
2586 	free(kve, M_TEMP);
2587 	return (error);
2588 }
2589 
2590 static int
2591 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2592 {
2593 	struct proc *p;
2594 	struct sbuf sb;
2595 	int error, error2, *name;
2596 
2597 	name = (int *)arg1;
2598 	sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req);
2599 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2600 	error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
2601 	if (error != 0) {
2602 		sbuf_delete(&sb);
2603 		return (error);
2604 	}
2605 	error = kern_proc_vmmap_out(p, &sb, -1, KERN_VMMAP_PACK_KINFO);
2606 	error2 = sbuf_finish(&sb);
2607 	sbuf_delete(&sb);
2608 	return (error != 0 ? error : error2);
2609 }
2610 
2611 #if defined(STACK) || defined(DDB)
2612 static int
2613 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2614 {
2615 	struct kinfo_kstack *kkstp;
2616 	int error, i, *name, numthreads;
2617 	lwpid_t *lwpidarray;
2618 	struct thread *td;
2619 	struct stack *st;
2620 	struct sbuf sb;
2621 	struct proc *p;
2622 
2623 	name = (int *)arg1;
2624 	error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2625 	if (error != 0)
2626 		return (error);
2627 
2628 	kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2629 	st = stack_create(M_WAITOK);
2630 
2631 	lwpidarray = NULL;
2632 	PROC_LOCK(p);
2633 	do {
2634 		if (lwpidarray != NULL) {
2635 			free(lwpidarray, M_TEMP);
2636 			lwpidarray = NULL;
2637 		}
2638 		numthreads = p->p_numthreads;
2639 		PROC_UNLOCK(p);
2640 		lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2641 		    M_WAITOK | M_ZERO);
2642 		PROC_LOCK(p);
2643 	} while (numthreads < p->p_numthreads);
2644 
2645 	/*
2646 	 * XXXRW: During the below loop, execve(2) and countless other sorts
2647 	 * of changes could have taken place.  Should we check to see if the
2648 	 * vmspace has been replaced, or the like, in order to prevent
2649 	 * giving a snapshot that spans, say, execve(2), with some threads
2650 	 * before and some after?  Among other things, the credentials could
2651 	 * have changed, in which case the right to extract debug info might
2652 	 * no longer be assured.
2653 	 */
2654 	i = 0;
2655 	FOREACH_THREAD_IN_PROC(p, td) {
2656 		KASSERT(i < numthreads,
2657 		    ("sysctl_kern_proc_kstack: numthreads"));
2658 		lwpidarray[i] = td->td_tid;
2659 		i++;
2660 	}
2661 	numthreads = i;
2662 	for (i = 0; i < numthreads; i++) {
2663 		td = thread_find(p, lwpidarray[i]);
2664 		if (td == NULL) {
2665 			continue;
2666 		}
2667 		bzero(kkstp, sizeof(*kkstp));
2668 		(void)sbuf_new(&sb, kkstp->kkst_trace,
2669 		    sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2670 		thread_lock(td);
2671 		kkstp->kkst_tid = td->td_tid;
2672 		if (TD_IS_SWAPPED(td)) {
2673 			kkstp->kkst_state = KKST_STATE_SWAPPED;
2674 		} else if (TD_IS_RUNNING(td)) {
2675 			if (stack_save_td_running(st, td) == 0)
2676 				kkstp->kkst_state = KKST_STATE_STACKOK;
2677 			else
2678 				kkstp->kkst_state = KKST_STATE_RUNNING;
2679 		} else {
2680 			kkstp->kkst_state = KKST_STATE_STACKOK;
2681 			stack_save_td(st, td);
2682 		}
2683 		thread_unlock(td);
2684 		PROC_UNLOCK(p);
2685 		stack_sbuf_print(&sb, st);
2686 		sbuf_finish(&sb);
2687 		sbuf_delete(&sb);
2688 		error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2689 		PROC_LOCK(p);
2690 		if (error)
2691 			break;
2692 	}
2693 	_PRELE(p);
2694 	PROC_UNLOCK(p);
2695 	if (lwpidarray != NULL)
2696 		free(lwpidarray, M_TEMP);
2697 	stack_destroy(st);
2698 	free(kkstp, M_TEMP);
2699 	return (error);
2700 }
2701 #endif
2702 
2703 /*
2704  * This sysctl allows a process to retrieve the full list of groups from
2705  * itself or another process.
2706  */
2707 static int
2708 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2709 {
2710 	pid_t *pidp = (pid_t *)arg1;
2711 	unsigned int arglen = arg2;
2712 	struct proc *p;
2713 	struct ucred *cred;
2714 	int error;
2715 
2716 	if (arglen != 1)
2717 		return (EINVAL);
2718 	if (*pidp == -1) {	/* -1 means this process */
2719 		p = req->td->td_proc;
2720 		PROC_LOCK(p);
2721 	} else {
2722 		error = pget(*pidp, PGET_CANSEE, &p);
2723 		if (error != 0)
2724 			return (error);
2725 	}
2726 
2727 	cred = crhold(p->p_ucred);
2728 	PROC_UNLOCK(p);
2729 
2730 	error = SYSCTL_OUT(req, cred->cr_groups,
2731 	    cred->cr_ngroups * sizeof(gid_t));
2732 	crfree(cred);
2733 	return (error);
2734 }
2735 
2736 /*
2737  * This sysctl allows a process to retrieve or/and set the resource limit for
2738  * another process.
2739  */
2740 static int
2741 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2742 {
2743 	int *name = (int *)arg1;
2744 	u_int namelen = arg2;
2745 	struct rlimit rlim;
2746 	struct proc *p;
2747 	u_int which;
2748 	int flags, error;
2749 
2750 	if (namelen != 2)
2751 		return (EINVAL);
2752 
2753 	which = (u_int)name[1];
2754 	if (which >= RLIM_NLIMITS)
2755 		return (EINVAL);
2756 
2757 	if (req->newptr != NULL && req->newlen != sizeof(rlim))
2758 		return (EINVAL);
2759 
2760 	flags = PGET_HOLD | PGET_NOTWEXIT;
2761 	if (req->newptr != NULL)
2762 		flags |= PGET_CANDEBUG;
2763 	else
2764 		flags |= PGET_CANSEE;
2765 	error = pget((pid_t)name[0], flags, &p);
2766 	if (error != 0)
2767 		return (error);
2768 
2769 	/*
2770 	 * Retrieve limit.
2771 	 */
2772 	if (req->oldptr != NULL) {
2773 		PROC_LOCK(p);
2774 		lim_rlimit_proc(p, which, &rlim);
2775 		PROC_UNLOCK(p);
2776 	}
2777 	error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
2778 	if (error != 0)
2779 		goto errout;
2780 
2781 	/*
2782 	 * Set limit.
2783 	 */
2784 	if (req->newptr != NULL) {
2785 		error = SYSCTL_IN(req, &rlim, sizeof(rlim));
2786 		if (error == 0)
2787 			error = kern_proc_setrlimit(curthread, p, which, &rlim);
2788 	}
2789 
2790 errout:
2791 	PRELE(p);
2792 	return (error);
2793 }
2794 
2795 /*
2796  * This sysctl allows a process to retrieve ps_strings structure location of
2797  * another process.
2798  */
2799 static int
2800 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
2801 {
2802 	int *name = (int *)arg1;
2803 	u_int namelen = arg2;
2804 	struct proc *p;
2805 	vm_offset_t ps_strings;
2806 	int error;
2807 #ifdef COMPAT_FREEBSD32
2808 	uint32_t ps_strings32;
2809 #endif
2810 
2811 	if (namelen != 1)
2812 		return (EINVAL);
2813 
2814 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2815 	if (error != 0)
2816 		return (error);
2817 #ifdef COMPAT_FREEBSD32
2818 	if ((req->flags & SCTL_MASK32) != 0) {
2819 		/*
2820 		 * We return 0 if the 32 bit emulation request is for a 64 bit
2821 		 * process.
2822 		 */
2823 		ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
2824 		    PTROUT(p->p_sysent->sv_psstrings) : 0;
2825 		PROC_UNLOCK(p);
2826 		error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
2827 		return (error);
2828 	}
2829 #endif
2830 	ps_strings = p->p_sysent->sv_psstrings;
2831 	PROC_UNLOCK(p);
2832 	error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
2833 	return (error);
2834 }
2835 
2836 /*
2837  * This sysctl allows a process to retrieve umask of another process.
2838  */
2839 static int
2840 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
2841 {
2842 	int *name = (int *)arg1;
2843 	u_int namelen = arg2;
2844 	struct proc *p;
2845 	int error;
2846 	u_short fd_cmask;
2847 	pid_t pid;
2848 
2849 	if (namelen != 1)
2850 		return (EINVAL);
2851 
2852 	pid = (pid_t)name[0];
2853 	p = curproc;
2854 	if (pid == p->p_pid || pid == 0) {
2855 		fd_cmask = p->p_fd->fd_cmask;
2856 		goto out;
2857 	}
2858 
2859 	error = pget(pid, PGET_WANTREAD, &p);
2860 	if (error != 0)
2861 		return (error);
2862 
2863 	fd_cmask = p->p_fd->fd_cmask;
2864 	PRELE(p);
2865 out:
2866 	error = SYSCTL_OUT(req, &fd_cmask, sizeof(fd_cmask));
2867 	return (error);
2868 }
2869 
2870 /*
2871  * This sysctl allows a process to set and retrieve binary osreldate of
2872  * another process.
2873  */
2874 static int
2875 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
2876 {
2877 	int *name = (int *)arg1;
2878 	u_int namelen = arg2;
2879 	struct proc *p;
2880 	int flags, error, osrel;
2881 
2882 	if (namelen != 1)
2883 		return (EINVAL);
2884 
2885 	if (req->newptr != NULL && req->newlen != sizeof(osrel))
2886 		return (EINVAL);
2887 
2888 	flags = PGET_HOLD | PGET_NOTWEXIT;
2889 	if (req->newptr != NULL)
2890 		flags |= PGET_CANDEBUG;
2891 	else
2892 		flags |= PGET_CANSEE;
2893 	error = pget((pid_t)name[0], flags, &p);
2894 	if (error != 0)
2895 		return (error);
2896 
2897 	error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
2898 	if (error != 0)
2899 		goto errout;
2900 
2901 	if (req->newptr != NULL) {
2902 		error = SYSCTL_IN(req, &osrel, sizeof(osrel));
2903 		if (error != 0)
2904 			goto errout;
2905 		if (osrel < 0) {
2906 			error = EINVAL;
2907 			goto errout;
2908 		}
2909 		p->p_osrel = osrel;
2910 	}
2911 errout:
2912 	PRELE(p);
2913 	return (error);
2914 }
2915 
2916 static int
2917 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS)
2918 {
2919 	int *name = (int *)arg1;
2920 	u_int namelen = arg2;
2921 	struct proc *p;
2922 	struct kinfo_sigtramp kst;
2923 	const struct sysentvec *sv;
2924 	int error;
2925 #ifdef COMPAT_FREEBSD32
2926 	struct kinfo_sigtramp32 kst32;
2927 #endif
2928 
2929 	if (namelen != 1)
2930 		return (EINVAL);
2931 
2932 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2933 	if (error != 0)
2934 		return (error);
2935 	sv = p->p_sysent;
2936 #ifdef COMPAT_FREEBSD32
2937 	if ((req->flags & SCTL_MASK32) != 0) {
2938 		bzero(&kst32, sizeof(kst32));
2939 		if (SV_PROC_FLAG(p, SV_ILP32)) {
2940 			if (sv->sv_sigcode_base != 0) {
2941 				kst32.ksigtramp_start = sv->sv_sigcode_base;
2942 				kst32.ksigtramp_end = sv->sv_sigcode_base +
2943 				    *sv->sv_szsigcode;
2944 			} else {
2945 				kst32.ksigtramp_start = sv->sv_psstrings -
2946 				    *sv->sv_szsigcode;
2947 				kst32.ksigtramp_end = sv->sv_psstrings;
2948 			}
2949 		}
2950 		PROC_UNLOCK(p);
2951 		error = SYSCTL_OUT(req, &kst32, sizeof(kst32));
2952 		return (error);
2953 	}
2954 #endif
2955 	bzero(&kst, sizeof(kst));
2956 	if (sv->sv_sigcode_base != 0) {
2957 		kst.ksigtramp_start = (char *)sv->sv_sigcode_base;
2958 		kst.ksigtramp_end = (char *)sv->sv_sigcode_base +
2959 		    *sv->sv_szsigcode;
2960 	} else {
2961 		kst.ksigtramp_start = (char *)sv->sv_psstrings -
2962 		    *sv->sv_szsigcode;
2963 		kst.ksigtramp_end = (char *)sv->sv_psstrings;
2964 	}
2965 	PROC_UNLOCK(p);
2966 	error = SYSCTL_OUT(req, &kst, sizeof(kst));
2967 	return (error);
2968 }
2969 
2970 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD,  0, "Process table");
2971 
2972 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
2973 	CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
2974 	"Return entire process table");
2975 
2976 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2977 	sysctl_kern_proc, "Process table");
2978 
2979 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
2980 	sysctl_kern_proc, "Process table");
2981 
2982 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2983 	sysctl_kern_proc, "Process table");
2984 
2985 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
2986 	CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2987 
2988 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
2989 	sysctl_kern_proc, "Process table");
2990 
2991 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2992 	sysctl_kern_proc, "Process table");
2993 
2994 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2995 	sysctl_kern_proc, "Process table");
2996 
2997 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2998 	sysctl_kern_proc, "Process table");
2999 
3000 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
3001 	sysctl_kern_proc, "Return process table, no threads");
3002 
3003 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
3004 	CTLFLAG_RW | CTLFLAG_CAPWR | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
3005 	sysctl_kern_proc_args, "Process argument list");
3006 
3007 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
3008 	sysctl_kern_proc_env, "Process environment");
3009 
3010 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
3011 	CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
3012 
3013 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
3014 	CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
3015 
3016 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
3017 	CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
3018 	"Process syscall vector name (ABI type)");
3019 
3020 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
3021 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3022 
3023 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
3024 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3025 
3026 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
3027 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3028 
3029 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
3030 	sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3031 
3032 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
3033 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3034 
3035 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
3036 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3037 
3038 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
3039 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3040 
3041 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
3042 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3043 
3044 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
3045 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
3046 	"Return process table, no threads");
3047 
3048 #ifdef COMPAT_FREEBSD7
3049 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
3050 	CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
3051 #endif
3052 
3053 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
3054 	CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
3055 
3056 #if defined(STACK) || defined(DDB)
3057 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
3058 	CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
3059 #endif
3060 
3061 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
3062 	CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
3063 
3064 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
3065 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
3066 	"Process resource limits");
3067 
3068 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
3069 	CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
3070 	"Process ps_strings location");
3071 
3072 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
3073 	CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
3074 
3075 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
3076 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
3077 	"Process binary osreldate");
3078 
3079 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD |
3080 	CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp,
3081 	"Process signal trampoline location");
3082 
3083 int allproc_gen;
3084 
3085 /*
3086  * stop_all_proc() purpose is to stop all process which have usermode,
3087  * except current process for obvious reasons.  This makes it somewhat
3088  * unreliable when invoked from multithreaded process.  The service
3089  * must not be user-callable anyway.
3090  */
3091 void
3092 stop_all_proc(void)
3093 {
3094 	struct proc *cp, *p;
3095 	int r, gen;
3096 	bool restart, seen_stopped, seen_exiting, stopped_some;
3097 
3098 	cp = curproc;
3099 allproc_loop:
3100 	sx_xlock(&allproc_lock);
3101 	gen = allproc_gen;
3102 	seen_exiting = seen_stopped = stopped_some = restart = false;
3103 	LIST_REMOVE(cp, p_list);
3104 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3105 	for (;;) {
3106 		p = LIST_NEXT(cp, p_list);
3107 		if (p == NULL)
3108 			break;
3109 		LIST_REMOVE(cp, p_list);
3110 		LIST_INSERT_AFTER(p, cp, p_list);
3111 		PROC_LOCK(p);
3112 		if ((p->p_flag & (P_KPROC | P_SYSTEM | P_TOTAL_STOP)) != 0) {
3113 			PROC_UNLOCK(p);
3114 			continue;
3115 		}
3116 		if ((p->p_flag & P_WEXIT) != 0) {
3117 			seen_exiting = true;
3118 			PROC_UNLOCK(p);
3119 			continue;
3120 		}
3121 		if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) {
3122 			/*
3123 			 * Stopped processes are tolerated when there
3124 			 * are no other processes which might continue
3125 			 * them.  P_STOPPED_SINGLE but not
3126 			 * P_TOTAL_STOP process still has at least one
3127 			 * thread running.
3128 			 */
3129 			seen_stopped = true;
3130 			PROC_UNLOCK(p);
3131 			continue;
3132 		}
3133 		sx_xunlock(&allproc_lock);
3134 		_PHOLD(p);
3135 		r = thread_single(p, SINGLE_ALLPROC);
3136 		if (r != 0)
3137 			restart = true;
3138 		else
3139 			stopped_some = true;
3140 		_PRELE(p);
3141 		PROC_UNLOCK(p);
3142 		sx_xlock(&allproc_lock);
3143 	}
3144 	/* Catch forked children we did not see in iteration. */
3145 	if (gen != allproc_gen)
3146 		restart = true;
3147 	sx_xunlock(&allproc_lock);
3148 	if (restart || stopped_some || seen_exiting || seen_stopped) {
3149 		kern_yield(PRI_USER);
3150 		goto allproc_loop;
3151 	}
3152 }
3153 
3154 void
3155 resume_all_proc(void)
3156 {
3157 	struct proc *cp, *p;
3158 
3159 	cp = curproc;
3160 	sx_xlock(&allproc_lock);
3161 again:
3162 	LIST_REMOVE(cp, p_list);
3163 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3164 	for (;;) {
3165 		p = LIST_NEXT(cp, p_list);
3166 		if (p == NULL)
3167 			break;
3168 		LIST_REMOVE(cp, p_list);
3169 		LIST_INSERT_AFTER(p, cp, p_list);
3170 		PROC_LOCK(p);
3171 		if ((p->p_flag & P_TOTAL_STOP) != 0) {
3172 			sx_xunlock(&allproc_lock);
3173 			_PHOLD(p);
3174 			thread_single_end(p, SINGLE_ALLPROC);
3175 			_PRELE(p);
3176 			PROC_UNLOCK(p);
3177 			sx_xlock(&allproc_lock);
3178 		} else {
3179 			PROC_UNLOCK(p);
3180 		}
3181 	}
3182 	/*  Did the loop above missed any stopped process ? */
3183 	FOREACH_PROC_IN_SYSTEM(p) {
3184 		/* No need for proc lock. */
3185 		if ((p->p_flag & P_TOTAL_STOP) != 0)
3186 			goto again;
3187 	}
3188 	sx_xunlock(&allproc_lock);
3189 }
3190 
3191 /* #define	TOTAL_STOP_DEBUG	1 */
3192 #ifdef TOTAL_STOP_DEBUG
3193 volatile static int ap_resume;
3194 #include <sys/mount.h>
3195 
3196 static int
3197 sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS)
3198 {
3199 	int error, val;
3200 
3201 	val = 0;
3202 	ap_resume = 0;
3203 	error = sysctl_handle_int(oidp, &val, 0, req);
3204 	if (error != 0 || req->newptr == NULL)
3205 		return (error);
3206 	if (val != 0) {
3207 		stop_all_proc();
3208 		syncer_suspend();
3209 		while (ap_resume == 0)
3210 			;
3211 		syncer_resume();
3212 		resume_all_proc();
3213 	}
3214 	return (0);
3215 }
3216 
3217 SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW |
3218     CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0,
3219     sysctl_debug_stop_all_proc, "I",
3220     "");
3221 #endif
3222