xref: /freebsd/sys/kern/kern_proc.c (revision 5596f836e7e04a272113e57b3a80f1f67c0fec7f)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1982, 1986, 1989, 1991, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)kern_proc.c	8.7 (Berkeley) 2/14/95
32  */
33 
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
36 
37 #include "opt_ddb.h"
38 #include "opt_ktrace.h"
39 #include "opt_kstack_pages.h"
40 #include "opt_stack.h"
41 
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/bitstring.h>
45 #include <sys/elf.h>
46 #include <sys/eventhandler.h>
47 #include <sys/exec.h>
48 #include <sys/jail.h>
49 #include <sys/kernel.h>
50 #include <sys/limits.h>
51 #include <sys/lock.h>
52 #include <sys/loginclass.h>
53 #include <sys/malloc.h>
54 #include <sys/mman.h>
55 #include <sys/mount.h>
56 #include <sys/mutex.h>
57 #include <sys/proc.h>
58 #include <sys/ptrace.h>
59 #include <sys/refcount.h>
60 #include <sys/resourcevar.h>
61 #include <sys/rwlock.h>
62 #include <sys/sbuf.h>
63 #include <sys/sysent.h>
64 #include <sys/sched.h>
65 #include <sys/smp.h>
66 #include <sys/stack.h>
67 #include <sys/stat.h>
68 #include <sys/sysctl.h>
69 #include <sys/filedesc.h>
70 #include <sys/tty.h>
71 #include <sys/signalvar.h>
72 #include <sys/sdt.h>
73 #include <sys/sx.h>
74 #include <sys/user.h>
75 #include <sys/vnode.h>
76 #include <sys/wait.h>
77 
78 #ifdef DDB
79 #include <ddb/ddb.h>
80 #endif
81 
82 #include <vm/vm.h>
83 #include <vm/vm_param.h>
84 #include <vm/vm_extern.h>
85 #include <vm/pmap.h>
86 #include <vm/vm_map.h>
87 #include <vm/vm_object.h>
88 #include <vm/vm_page.h>
89 #include <vm/uma.h>
90 
91 #include <fs/devfs/devfs.h>
92 
93 #ifdef COMPAT_FREEBSD32
94 #include <compat/freebsd32/freebsd32.h>
95 #include <compat/freebsd32/freebsd32_util.h>
96 #endif
97 
98 SDT_PROVIDER_DEFINE(proc);
99 
100 MALLOC_DEFINE(M_PGRP, "pgrp", "process group header");
101 MALLOC_DEFINE(M_SESSION, "session", "session header");
102 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
103 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
104 
105 static void fixjobc_enterpgrp(struct proc *p, struct pgrp *pgrp);
106 static void doenterpgrp(struct proc *, struct pgrp *);
107 static void orphanpg(struct pgrp *pg);
108 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp);
109 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp);
110 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp,
111     int preferthread);
112 static void pgadjustjobc(struct pgrp *pgrp, bool entering);
113 static void pgdelete(struct pgrp *);
114 static int proc_ctor(void *mem, int size, void *arg, int flags);
115 static void proc_dtor(void *mem, int size, void *arg);
116 static int proc_init(void *mem, int size, int flags);
117 static void proc_fini(void *mem, int size);
118 static void pargs_free(struct pargs *pa);
119 
120 /*
121  * Other process lists
122  */
123 struct pidhashhead *pidhashtbl;
124 struct sx *pidhashtbl_lock;
125 u_long pidhash;
126 u_long pidhashlock;
127 struct pgrphashhead *pgrphashtbl;
128 u_long pgrphash;
129 struct proclist allproc;
130 struct sx __exclusive_cache_line allproc_lock;
131 struct sx __exclusive_cache_line proctree_lock;
132 struct mtx __exclusive_cache_line ppeers_lock;
133 struct mtx __exclusive_cache_line procid_lock;
134 uma_zone_t proc_zone;
135 
136 /*
137  * The offset of various fields in struct proc and struct thread.
138  * These are used by kernel debuggers to enumerate kernel threads and
139  * processes.
140  */
141 const int proc_off_p_pid = offsetof(struct proc, p_pid);
142 const int proc_off_p_comm = offsetof(struct proc, p_comm);
143 const int proc_off_p_list = offsetof(struct proc, p_list);
144 const int proc_off_p_threads = offsetof(struct proc, p_threads);
145 const int thread_off_td_tid = offsetof(struct thread, td_tid);
146 const int thread_off_td_name = offsetof(struct thread, td_name);
147 const int thread_off_td_oncpu = offsetof(struct thread, td_oncpu);
148 const int thread_off_td_pcb = offsetof(struct thread, td_pcb);
149 const int thread_off_td_plist = offsetof(struct thread, td_plist);
150 
151 EVENTHANDLER_LIST_DEFINE(process_ctor);
152 EVENTHANDLER_LIST_DEFINE(process_dtor);
153 EVENTHANDLER_LIST_DEFINE(process_init);
154 EVENTHANDLER_LIST_DEFINE(process_fini);
155 EVENTHANDLER_LIST_DEFINE(process_exit);
156 EVENTHANDLER_LIST_DEFINE(process_fork);
157 EVENTHANDLER_LIST_DEFINE(process_exec);
158 
159 int kstack_pages = KSTACK_PAGES;
160 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0,
161     "Kernel stack size in pages");
162 static int vmmap_skip_res_cnt = 0;
163 SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW,
164     &vmmap_skip_res_cnt, 0,
165     "Skip calculation of the pages resident count in kern.proc.vmmap");
166 
167 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
168 #ifdef COMPAT_FREEBSD32
169 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
170 #endif
171 
172 /*
173  * Initialize global process hashing structures.
174  */
175 void
176 procinit(void)
177 {
178 	u_long i;
179 
180 	sx_init(&allproc_lock, "allproc");
181 	sx_init(&proctree_lock, "proctree");
182 	mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
183 	mtx_init(&procid_lock, "procid", NULL, MTX_DEF);
184 	LIST_INIT(&allproc);
185 	pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
186 	pidhashlock = (pidhash + 1) / 64;
187 	if (pidhashlock > 0)
188 		pidhashlock--;
189 	pidhashtbl_lock = malloc(sizeof(*pidhashtbl_lock) * (pidhashlock + 1),
190 	    M_PROC, M_WAITOK | M_ZERO);
191 	for (i = 0; i < pidhashlock + 1; i++)
192 		sx_init_flags(&pidhashtbl_lock[i], "pidhash", SX_DUPOK);
193 	pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
194 	proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
195 	    proc_ctor, proc_dtor, proc_init, proc_fini,
196 	    UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
197 	uihashinit();
198 }
199 
200 /*
201  * Prepare a proc for use.
202  */
203 static int
204 proc_ctor(void *mem, int size, void *arg, int flags)
205 {
206 	struct proc *p;
207 	struct thread *td;
208 
209 	p = (struct proc *)mem;
210 	EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
211 	td = FIRST_THREAD_IN_PROC(p);
212 	if (td != NULL) {
213 		/* Make sure all thread constructors are executed */
214 		EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
215 	}
216 	return (0);
217 }
218 
219 /*
220  * Reclaim a proc after use.
221  */
222 static void
223 proc_dtor(void *mem, int size, void *arg)
224 {
225 	struct proc *p;
226 	struct thread *td;
227 
228 	/* INVARIANTS checks go here */
229 	p = (struct proc *)mem;
230 	td = FIRST_THREAD_IN_PROC(p);
231 	if (td != NULL) {
232 #ifdef INVARIANTS
233 		KASSERT((p->p_numthreads == 1),
234 		    ("bad number of threads in exiting process"));
235 		KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
236 #endif
237 		/* Free all OSD associated to this thread. */
238 		osd_thread_exit(td);
239 		td_softdep_cleanup(td);
240 		MPASS(td->td_su == NULL);
241 
242 		/* Make sure all thread destructors are executed */
243 		EVENTHANDLER_DIRECT_INVOKE(thread_dtor, td);
244 	}
245 	EVENTHANDLER_DIRECT_INVOKE(process_dtor, p);
246 	if (p->p_ksi != NULL)
247 		KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
248 }
249 
250 /*
251  * Initialize type-stable parts of a proc (when newly created).
252  */
253 static int
254 proc_init(void *mem, int size, int flags)
255 {
256 	struct proc *p;
257 
258 	p = (struct proc *)mem;
259 	mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK | MTX_NEW);
260 	mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_NEW);
261 	mtx_init(&p->p_statmtx, "pstatl", NULL, MTX_SPIN | MTX_NEW);
262 	mtx_init(&p->p_itimmtx, "pitiml", NULL, MTX_SPIN | MTX_NEW);
263 	mtx_init(&p->p_profmtx, "pprofl", NULL, MTX_SPIN | MTX_NEW);
264 	cv_init(&p->p_pwait, "ppwait");
265 	TAILQ_INIT(&p->p_threads);	     /* all threads in proc */
266 	EVENTHANDLER_DIRECT_INVOKE(process_init, p);
267 	p->p_stats = pstats_alloc();
268 	p->p_pgrp = NULL;
269 	return (0);
270 }
271 
272 /*
273  * UMA should ensure that this function is never called.
274  * Freeing a proc structure would violate type stability.
275  */
276 static void
277 proc_fini(void *mem, int size)
278 {
279 #ifdef notnow
280 	struct proc *p;
281 
282 	p = (struct proc *)mem;
283 	EVENTHANDLER_DIRECT_INVOKE(process_fini, p);
284 	pstats_free(p->p_stats);
285 	thread_free(FIRST_THREAD_IN_PROC(p));
286 	mtx_destroy(&p->p_mtx);
287 	if (p->p_ksi != NULL)
288 		ksiginfo_free(p->p_ksi);
289 #else
290 	panic("proc reclaimed");
291 #endif
292 }
293 
294 /*
295  * PID space management.
296  *
297  * These bitmaps are used by fork_findpid.
298  */
299 bitstr_t bit_decl(proc_id_pidmap, PID_MAX);
300 bitstr_t bit_decl(proc_id_grpidmap, PID_MAX);
301 bitstr_t bit_decl(proc_id_sessidmap, PID_MAX);
302 bitstr_t bit_decl(proc_id_reapmap, PID_MAX);
303 
304 static bitstr_t *proc_id_array[] = {
305 	proc_id_pidmap,
306 	proc_id_grpidmap,
307 	proc_id_sessidmap,
308 	proc_id_reapmap,
309 };
310 
311 void
312 proc_id_set(int type, pid_t id)
313 {
314 
315 	KASSERT(type >= 0 && type < nitems(proc_id_array),
316 	    ("invalid type %d\n", type));
317 	mtx_lock(&procid_lock);
318 	KASSERT(bit_test(proc_id_array[type], id) == 0,
319 	    ("bit %d already set in %d\n", id, type));
320 	bit_set(proc_id_array[type], id);
321 	mtx_unlock(&procid_lock);
322 }
323 
324 void
325 proc_id_set_cond(int type, pid_t id)
326 {
327 
328 	KASSERT(type >= 0 && type < nitems(proc_id_array),
329 	    ("invalid type %d\n", type));
330 	if (bit_test(proc_id_array[type], id))
331 		return;
332 	mtx_lock(&procid_lock);
333 	bit_set(proc_id_array[type], id);
334 	mtx_unlock(&procid_lock);
335 }
336 
337 void
338 proc_id_clear(int type, pid_t id)
339 {
340 
341 	KASSERT(type >= 0 && type < nitems(proc_id_array),
342 	    ("invalid type %d\n", type));
343 	mtx_lock(&procid_lock);
344 	KASSERT(bit_test(proc_id_array[type], id) != 0,
345 	    ("bit %d not set in %d\n", id, type));
346 	bit_clear(proc_id_array[type], id);
347 	mtx_unlock(&procid_lock);
348 }
349 
350 /*
351  * Is p an inferior of the current process?
352  */
353 int
354 inferior(struct proc *p)
355 {
356 
357 	sx_assert(&proctree_lock, SX_LOCKED);
358 	PROC_LOCK_ASSERT(p, MA_OWNED);
359 	for (; p != curproc; p = proc_realparent(p)) {
360 		if (p->p_pid == 0)
361 			return (0);
362 	}
363 	return (1);
364 }
365 
366 /*
367  * Shared lock all the pid hash lists.
368  */
369 void
370 pidhash_slockall(void)
371 {
372 	u_long i;
373 
374 	for (i = 0; i < pidhashlock + 1; i++)
375 		sx_slock(&pidhashtbl_lock[i]);
376 }
377 
378 /*
379  * Shared unlock all the pid hash lists.
380  */
381 void
382 pidhash_sunlockall(void)
383 {
384 	u_long i;
385 
386 	for (i = 0; i < pidhashlock + 1; i++)
387 		sx_sunlock(&pidhashtbl_lock[i]);
388 }
389 
390 /*
391  * Similar to pfind_any(), this function finds zombies.
392  */
393 struct proc *
394 pfind_any_locked(pid_t pid)
395 {
396 	struct proc *p;
397 
398 	sx_assert(PIDHASHLOCK(pid), SX_LOCKED);
399 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
400 		if (p->p_pid == pid) {
401 			PROC_LOCK(p);
402 			if (p->p_state == PRS_NEW) {
403 				PROC_UNLOCK(p);
404 				p = NULL;
405 			}
406 			break;
407 		}
408 	}
409 	return (p);
410 }
411 
412 /*
413  * Locate a process by number.
414  *
415  * By not returning processes in the PRS_NEW state, we allow callers to avoid
416  * testing for that condition to avoid dereferencing p_ucred, et al.
417  */
418 static __always_inline struct proc *
419 _pfind(pid_t pid, bool zombie)
420 {
421 	struct proc *p;
422 
423 	p = curproc;
424 	if (p->p_pid == pid) {
425 		PROC_LOCK(p);
426 		return (p);
427 	}
428 	sx_slock(PIDHASHLOCK(pid));
429 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
430 		if (p->p_pid == pid) {
431 			PROC_LOCK(p);
432 			if (p->p_state == PRS_NEW ||
433 			    (!zombie && p->p_state == PRS_ZOMBIE)) {
434 				PROC_UNLOCK(p);
435 				p = NULL;
436 			}
437 			break;
438 		}
439 	}
440 	sx_sunlock(PIDHASHLOCK(pid));
441 	return (p);
442 }
443 
444 struct proc *
445 pfind(pid_t pid)
446 {
447 
448 	return (_pfind(pid, false));
449 }
450 
451 /*
452  * Same as pfind but allow zombies.
453  */
454 struct proc *
455 pfind_any(pid_t pid)
456 {
457 
458 	return (_pfind(pid, true));
459 }
460 
461 /*
462  * Locate a process group by number.
463  * The caller must hold proctree_lock.
464  */
465 struct pgrp *
466 pgfind(pid_t pgid)
467 {
468 	struct pgrp *pgrp;
469 
470 	sx_assert(&proctree_lock, SX_LOCKED);
471 
472 	LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
473 		if (pgrp->pg_id == pgid) {
474 			PGRP_LOCK(pgrp);
475 			return (pgrp);
476 		}
477 	}
478 	return (NULL);
479 }
480 
481 /*
482  * Locate process and do additional manipulations, depending on flags.
483  */
484 int
485 pget(pid_t pid, int flags, struct proc **pp)
486 {
487 	struct proc *p;
488 	struct thread *td1;
489 	int error;
490 
491 	p = curproc;
492 	if (p->p_pid == pid) {
493 		PROC_LOCK(p);
494 	} else {
495 		p = NULL;
496 		if (pid <= PID_MAX) {
497 			if ((flags & PGET_NOTWEXIT) == 0)
498 				p = pfind_any(pid);
499 			else
500 				p = pfind(pid);
501 		} else if ((flags & PGET_NOTID) == 0) {
502 			td1 = tdfind(pid, -1);
503 			if (td1 != NULL)
504 				p = td1->td_proc;
505 		}
506 		if (p == NULL)
507 			return (ESRCH);
508 		if ((flags & PGET_CANSEE) != 0) {
509 			error = p_cansee(curthread, p);
510 			if (error != 0)
511 				goto errout;
512 		}
513 	}
514 	if ((flags & PGET_CANDEBUG) != 0) {
515 		error = p_candebug(curthread, p);
516 		if (error != 0)
517 			goto errout;
518 	}
519 	if ((flags & PGET_ISCURRENT) != 0 && curproc != p) {
520 		error = EPERM;
521 		goto errout;
522 	}
523 	if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) {
524 		error = ESRCH;
525 		goto errout;
526 	}
527 	if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) {
528 		/*
529 		 * XXXRW: Not clear ESRCH is the right error during proc
530 		 * execve().
531 		 */
532 		error = ESRCH;
533 		goto errout;
534 	}
535 	if ((flags & PGET_HOLD) != 0) {
536 		_PHOLD(p);
537 		PROC_UNLOCK(p);
538 	}
539 	*pp = p;
540 	return (0);
541 errout:
542 	PROC_UNLOCK(p);
543 	return (error);
544 }
545 
546 /*
547  * Create a new process group.
548  * pgid must be equal to the pid of p.
549  * Begin a new session if required.
550  */
551 int
552 enterpgrp(struct proc *p, pid_t pgid, struct pgrp *pgrp, struct session *sess)
553 {
554 
555 	sx_assert(&proctree_lock, SX_XLOCKED);
556 
557 	KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
558 	KASSERT(p->p_pid == pgid,
559 	    ("enterpgrp: new pgrp and pid != pgid"));
560 	KASSERT(pgfind(pgid) == NULL,
561 	    ("enterpgrp: pgrp with pgid exists"));
562 	KASSERT(!SESS_LEADER(p),
563 	    ("enterpgrp: session leader attempted setpgrp"));
564 
565 	mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
566 
567 	if (sess != NULL) {
568 		/*
569 		 * new session
570 		 */
571 		mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
572 		PROC_LOCK(p);
573 		p->p_flag &= ~P_CONTROLT;
574 		PROC_UNLOCK(p);
575 		PGRP_LOCK(pgrp);
576 		sess->s_leader = p;
577 		sess->s_sid = p->p_pid;
578 		proc_id_set(PROC_ID_SESSION, p->p_pid);
579 		refcount_init(&sess->s_count, 1);
580 		sess->s_ttyvp = NULL;
581 		sess->s_ttydp = NULL;
582 		sess->s_ttyp = NULL;
583 		bcopy(p->p_session->s_login, sess->s_login,
584 			    sizeof(sess->s_login));
585 		pgrp->pg_session = sess;
586 		KASSERT(p == curproc,
587 		    ("enterpgrp: mksession and p != curproc"));
588 	} else {
589 		pgrp->pg_session = p->p_session;
590 		sess_hold(pgrp->pg_session);
591 		PGRP_LOCK(pgrp);
592 	}
593 	pgrp->pg_id = pgid;
594 	proc_id_set(PROC_ID_GROUP, p->p_pid);
595 	LIST_INIT(&pgrp->pg_members);
596 
597 	/*
598 	 * As we have an exclusive lock of proctree_lock,
599 	 * this should not deadlock.
600 	 */
601 	LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
602 	pgrp->pg_jobc = 0;
603 	SLIST_INIT(&pgrp->pg_sigiolst);
604 	PGRP_UNLOCK(pgrp);
605 
606 	doenterpgrp(p, pgrp);
607 
608 	return (0);
609 }
610 
611 /*
612  * Move p to an existing process group
613  */
614 int
615 enterthispgrp(struct proc *p, struct pgrp *pgrp)
616 {
617 
618 	sx_assert(&proctree_lock, SX_XLOCKED);
619 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
620 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
621 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
622 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
623 	KASSERT(pgrp->pg_session == p->p_session,
624 		("%s: pgrp's session %p, p->p_session %p.\n",
625 		__func__,
626 		pgrp->pg_session,
627 		p->p_session));
628 	KASSERT(pgrp != p->p_pgrp,
629 		("%s: p belongs to pgrp.", __func__));
630 
631 	doenterpgrp(p, pgrp);
632 
633 	return (0);
634 }
635 
636 /*
637  * If true, any child of q which belongs to group pgrp, qualifies the
638  * process group pgrp as not orphaned.
639  */
640 static bool
641 isjobproc(struct proc *q, struct pgrp *pgrp)
642 {
643 	sx_assert(&proctree_lock, SX_LOCKED);
644 	return (q->p_pgrp != pgrp &&
645 	    q->p_pgrp->pg_session == pgrp->pg_session);
646 }
647 
648 static struct proc *
649 jobc_reaper(struct proc *p)
650 {
651 	struct proc *pp;
652 
653 	sx_assert(&proctree_lock, SX_LOCKED);
654 
655 	for (pp = p;;) {
656 		pp = pp->p_reaper;
657 		if (pp->p_reaper == pp ||
658 		    (pp->p_treeflag & P_TREE_GRPEXITED) == 0)
659 			return (pp);
660 	}
661 }
662 
663 static struct proc *
664 jobc_parent(struct proc *p)
665 {
666 	struct proc *pp;
667 
668 	sx_assert(&proctree_lock, SX_LOCKED);
669 
670 	pp = proc_realparent(p);
671 	if (pp->p_pptr == NULL ||
672 	    (pp->p_treeflag & P_TREE_GRPEXITED) == 0)
673 		return (pp);
674 	return (jobc_reaper(pp));
675 }
676 
677 #ifdef INVARIANTS
678 static void
679 check_pgrp_jobc(struct pgrp *pgrp)
680 {
681 	struct proc *q;
682 	int cnt;
683 
684 	sx_assert(&proctree_lock, SX_LOCKED);
685 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
686 
687 	cnt = 0;
688 	PGRP_LOCK(pgrp);
689 	LIST_FOREACH(q, &pgrp->pg_members, p_pglist) {
690 		if ((q->p_treeflag & P_TREE_GRPEXITED) != 0 ||
691 		    q->p_pptr == NULL)
692 			continue;
693 		if (isjobproc(jobc_parent(q), pgrp))
694 			cnt++;
695 	}
696 	KASSERT(pgrp->pg_jobc == cnt, ("pgrp %d %p pg_jobc %d cnt %d",
697 	    pgrp->pg_id, pgrp, pgrp->pg_jobc, cnt));
698 	PGRP_UNLOCK(pgrp);
699 }
700 #endif
701 
702 /*
703  * Move p to a process group
704  */
705 static void
706 doenterpgrp(struct proc *p, struct pgrp *pgrp)
707 {
708 	struct pgrp *savepgrp;
709 
710 	sx_assert(&proctree_lock, SX_XLOCKED);
711 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
712 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
713 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
714 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
715 
716 	savepgrp = p->p_pgrp;
717 
718 #ifdef INVARIANTS
719 	check_pgrp_jobc(pgrp);
720 	check_pgrp_jobc(savepgrp);
721 #endif
722 
723 	/*
724 	 * Adjust eligibility of affected pgrps to participate in job control.
725 	 */
726 	fixjobc_enterpgrp(p, pgrp);
727 
728 	PGRP_LOCK(pgrp);
729 	PGRP_LOCK(savepgrp);
730 	PROC_LOCK(p);
731 	LIST_REMOVE(p, p_pglist);
732 	p->p_pgrp = pgrp;
733 	PROC_UNLOCK(p);
734 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
735 	PGRP_UNLOCK(savepgrp);
736 	PGRP_UNLOCK(pgrp);
737 	if (LIST_EMPTY(&savepgrp->pg_members))
738 		pgdelete(savepgrp);
739 }
740 
741 /*
742  * remove process from process group
743  */
744 int
745 leavepgrp(struct proc *p)
746 {
747 	struct pgrp *savepgrp;
748 
749 	sx_assert(&proctree_lock, SX_XLOCKED);
750 	savepgrp = p->p_pgrp;
751 	PGRP_LOCK(savepgrp);
752 	PROC_LOCK(p);
753 	LIST_REMOVE(p, p_pglist);
754 	p->p_pgrp = NULL;
755 	PROC_UNLOCK(p);
756 	PGRP_UNLOCK(savepgrp);
757 	if (LIST_EMPTY(&savepgrp->pg_members))
758 		pgdelete(savepgrp);
759 	return (0);
760 }
761 
762 /*
763  * delete a process group
764  */
765 static void
766 pgdelete(struct pgrp *pgrp)
767 {
768 	struct session *savesess;
769 	struct tty *tp;
770 
771 	sx_assert(&proctree_lock, SX_XLOCKED);
772 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
773 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
774 
775 	/*
776 	 * Reset any sigio structures pointing to us as a result of
777 	 * F_SETOWN with our pgid.  The proctree lock ensures that
778 	 * new sigio structures will not be added after this point.
779 	 */
780 	funsetownlst(&pgrp->pg_sigiolst);
781 
782 	PGRP_LOCK(pgrp);
783 	tp = pgrp->pg_session->s_ttyp;
784 	LIST_REMOVE(pgrp, pg_hash);
785 	savesess = pgrp->pg_session;
786 	PGRP_UNLOCK(pgrp);
787 
788 	/* Remove the reference to the pgrp before deallocating it. */
789 	if (tp != NULL) {
790 		tty_lock(tp);
791 		tty_rel_pgrp(tp, pgrp);
792 	}
793 
794 	proc_id_clear(PROC_ID_GROUP, pgrp->pg_id);
795 	mtx_destroy(&pgrp->pg_mtx);
796 	free(pgrp, M_PGRP);
797 	sess_release(savesess);
798 }
799 
800 static void
801 pgadjustjobc(struct pgrp *pgrp, bool entering)
802 {
803 
804 	PGRP_LOCK(pgrp);
805 	if (entering) {
806 		MPASS(pgrp->pg_jobc >= 0);
807 		pgrp->pg_jobc++;
808 	} else {
809 		MPASS(pgrp->pg_jobc > 0);
810 		--pgrp->pg_jobc;
811 		if (pgrp->pg_jobc == 0)
812 			orphanpg(pgrp);
813 	}
814 	PGRP_UNLOCK(pgrp);
815 }
816 
817 static void
818 fixjobc_enterpgrp_q(struct pgrp *pgrp, struct proc *p, struct proc *q, bool adj)
819 {
820 	struct pgrp *childpgrp;
821 	bool future_jobc;
822 
823 	sx_assert(&proctree_lock, SX_LOCKED);
824 
825 	if ((q->p_treeflag & P_TREE_GRPEXITED) != 0)
826 		return;
827 	childpgrp = q->p_pgrp;
828 	future_jobc = childpgrp != pgrp &&
829 	    childpgrp->pg_session == pgrp->pg_session;
830 
831 	if ((adj && !isjobproc(p, childpgrp) && future_jobc) ||
832 	    (!adj && isjobproc(p, childpgrp) && !future_jobc))
833 		pgadjustjobc(childpgrp, adj);
834 }
835 
836 /*
837  * Adjust pgrp jobc counters when specified process changes process group.
838  * We count the number of processes in each process group that "qualify"
839  * the group for terminal job control (those with a parent in a different
840  * process group of the same session).  If that count reaches zero, the
841  * process group becomes orphaned.  Check both the specified process'
842  * process group and that of its children.
843  * We increment eligibility counts before decrementing, otherwise we
844  * could reach 0 spuriously during the decrement.
845  */
846 static void
847 fixjobc_enterpgrp(struct proc *p, struct pgrp *pgrp)
848 {
849 	struct proc *q;
850 
851 	sx_assert(&proctree_lock, SX_LOCKED);
852 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
853 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
854 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
855 
856 	if (p->p_pgrp == pgrp)
857 		return;
858 
859 	if (isjobproc(jobc_parent(p), pgrp))
860 		pgadjustjobc(pgrp, true);
861 	LIST_FOREACH(q, &p->p_children, p_sibling) {
862 		if ((q->p_treeflag & P_TREE_ORPHANED) != 0)
863 			continue;
864 		fixjobc_enterpgrp_q(pgrp, p, q, true);
865 	}
866 	LIST_FOREACH(q, &p->p_orphans, p_orphan)
867 		fixjobc_enterpgrp_q(pgrp, p, q, true);
868 
869 	if (isjobproc(jobc_parent(p), p->p_pgrp))
870 		pgadjustjobc(p->p_pgrp, false);
871 	LIST_FOREACH(q, &p->p_children, p_sibling) {
872 		if ((q->p_treeflag & P_TREE_ORPHANED) != 0)
873 			continue;
874 		fixjobc_enterpgrp_q(pgrp, p, q, false);
875 	}
876 	LIST_FOREACH(q, &p->p_orphans, p_orphan)
877 		fixjobc_enterpgrp_q(pgrp, p, q, false);
878 }
879 
880 static void
881 fixjobc_kill_q(struct proc *p, struct proc *q, bool adj)
882 {
883 	struct pgrp *childpgrp;
884 
885 	sx_assert(&proctree_lock, SX_LOCKED);
886 
887 	if ((q->p_treeflag & P_TREE_GRPEXITED) != 0)
888 		return;
889 	childpgrp = q->p_pgrp;
890 
891 	if ((adj && isjobproc(jobc_reaper(q), childpgrp) &&
892 	    !isjobproc(p, childpgrp)) || (!adj && !isjobproc(jobc_reaper(q),
893 	    childpgrp) && isjobproc(p, childpgrp)))
894 		pgadjustjobc(childpgrp, adj);
895 }
896 
897 static void
898 fixjobc_kill(struct proc *p)
899 {
900 	struct proc *q;
901 	struct pgrp *pgrp;
902 
903 	sx_assert(&proctree_lock, SX_LOCKED);
904 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
905 	pgrp = p->p_pgrp;
906 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
907 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
908 #ifdef INVARIANTS
909 	check_pgrp_jobc(pgrp);
910 #endif
911 
912 	/*
913 	 * p no longer affects process group orphanage for children.
914 	 * It is marked by the flag because p is only physically
915 	 * removed from its process group on wait(2).
916 	 */
917 	MPASS((p->p_treeflag & P_TREE_GRPEXITED) == 0);
918 	p->p_treeflag |= P_TREE_GRPEXITED;
919 
920 	/*
921 	 * Check p's parent to see whether p qualifies its own process
922 	 * group; if so, adjust count for p's process group.
923 	 */
924 	if (isjobproc(jobc_parent(p), pgrp))
925 		pgadjustjobc(pgrp, false);
926 
927 	/*
928 	 * Check this process' children to see whether they qualify
929 	 * their process groups after reparenting to reaper.  If so,
930 	 * adjust counts for children's process groups.
931 	 */
932 	LIST_FOREACH(q, &p->p_children, p_sibling) {
933 		if ((q->p_treeflag & P_TREE_ORPHANED) != 0)
934 			continue;
935 		fixjobc_kill_q(p, q, true);
936 	}
937 	LIST_FOREACH(q, &p->p_orphans, p_orphan)
938 		fixjobc_kill_q(p, q, true);
939 	LIST_FOREACH(q, &p->p_children, p_sibling) {
940 		if ((q->p_treeflag & P_TREE_ORPHANED) != 0)
941 			continue;
942 		fixjobc_kill_q(p, q, false);
943 	}
944 	LIST_FOREACH(q, &p->p_orphans, p_orphan)
945 		fixjobc_kill_q(p, q, false);
946 
947 #ifdef INVARIANTS
948 	check_pgrp_jobc(pgrp);
949 #endif
950 }
951 
952 void
953 killjobc(void)
954 {
955 	struct session *sp;
956 	struct tty *tp;
957 	struct proc *p;
958 	struct vnode *ttyvp;
959 
960 	p = curproc;
961 	MPASS(p->p_flag & P_WEXIT);
962 	sx_assert(&proctree_lock, SX_LOCKED);
963 
964 	if (SESS_LEADER(p)) {
965 		sp = p->p_session;
966 
967 		/*
968 		 * s_ttyp is not zero'd; we use this to indicate that
969 		 * the session once had a controlling terminal. (for
970 		 * logging and informational purposes)
971 		 */
972 		SESS_LOCK(sp);
973 		ttyvp = sp->s_ttyvp;
974 		tp = sp->s_ttyp;
975 		sp->s_ttyvp = NULL;
976 		sp->s_ttydp = NULL;
977 		sp->s_leader = NULL;
978 		SESS_UNLOCK(sp);
979 
980 		/*
981 		 * Signal foreground pgrp and revoke access to
982 		 * controlling terminal if it has not been revoked
983 		 * already.
984 		 *
985 		 * Because the TTY may have been revoked in the mean
986 		 * time and could already have a new session associated
987 		 * with it, make sure we don't send a SIGHUP to a
988 		 * foreground process group that does not belong to this
989 		 * session.
990 		 */
991 
992 		if (tp != NULL) {
993 			tty_lock(tp);
994 			if (tp->t_session == sp)
995 				tty_signal_pgrp(tp, SIGHUP);
996 			tty_unlock(tp);
997 		}
998 
999 		if (ttyvp != NULL) {
1000 			sx_xunlock(&proctree_lock);
1001 			if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) {
1002 				VOP_REVOKE(ttyvp, REVOKEALL);
1003 				VOP_UNLOCK(ttyvp);
1004 			}
1005 			devfs_ctty_unref(ttyvp);
1006 			sx_xlock(&proctree_lock);
1007 		}
1008 	}
1009 	fixjobc_kill(p);
1010 }
1011 
1012 /*
1013  * A process group has become orphaned;
1014  * if there are any stopped processes in the group,
1015  * hang-up all process in that group.
1016  */
1017 static void
1018 orphanpg(struct pgrp *pg)
1019 {
1020 	struct proc *p;
1021 
1022 	PGRP_LOCK_ASSERT(pg, MA_OWNED);
1023 
1024 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
1025 		PROC_LOCK(p);
1026 		if (P_SHOULDSTOP(p) == P_STOPPED_SIG) {
1027 			PROC_UNLOCK(p);
1028 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
1029 				PROC_LOCK(p);
1030 				kern_psignal(p, SIGHUP);
1031 				kern_psignal(p, SIGCONT);
1032 				PROC_UNLOCK(p);
1033 			}
1034 			return;
1035 		}
1036 		PROC_UNLOCK(p);
1037 	}
1038 }
1039 
1040 void
1041 sess_hold(struct session *s)
1042 {
1043 
1044 	refcount_acquire(&s->s_count);
1045 }
1046 
1047 void
1048 sess_release(struct session *s)
1049 {
1050 
1051 	if (refcount_release(&s->s_count)) {
1052 		if (s->s_ttyp != NULL) {
1053 			tty_lock(s->s_ttyp);
1054 			tty_rel_sess(s->s_ttyp, s);
1055 		}
1056 		proc_id_clear(PROC_ID_SESSION, s->s_sid);
1057 		mtx_destroy(&s->s_mtx);
1058 		free(s, M_SESSION);
1059 	}
1060 }
1061 
1062 #ifdef DDB
1063 
1064 static void
1065 db_print_pgrp_one(struct pgrp *pgrp, struct proc *p)
1066 {
1067 	db_printf(
1068 	    "    pid %d at %p pr %d pgrp %p e %d jc %d\n",
1069 	    p->p_pid, p, p->p_pptr == NULL ? -1 : p->p_pptr->p_pid,
1070 	    p->p_pgrp, (p->p_treeflag & P_TREE_GRPEXITED) != 0,
1071 	    p->p_pptr == NULL ? 0 : isjobproc(p->p_pptr, pgrp));
1072 }
1073 
1074 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
1075 {
1076 	struct pgrp *pgrp;
1077 	struct proc *p;
1078 	int i;
1079 
1080 	for (i = 0; i <= pgrphash; i++) {
1081 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
1082 			db_printf("indx %d\n", i);
1083 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
1084 				db_printf(
1085 			"  pgrp %p, pgid %d, sess %p, sesscnt %d, mem %p\n",
1086 				    pgrp, (int)pgrp->pg_id, pgrp->pg_session,
1087 				    pgrp->pg_session->s_count,
1088 				    LIST_FIRST(&pgrp->pg_members));
1089 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist)
1090 					db_print_pgrp_one(pgrp, p);
1091 			}
1092 		}
1093 	}
1094 }
1095 #endif /* DDB */
1096 
1097 /*
1098  * Calculate the kinfo_proc members which contain process-wide
1099  * informations.
1100  * Must be called with the target process locked.
1101  */
1102 static void
1103 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
1104 {
1105 	struct thread *td;
1106 
1107 	PROC_LOCK_ASSERT(p, MA_OWNED);
1108 
1109 	kp->ki_estcpu = 0;
1110 	kp->ki_pctcpu = 0;
1111 	FOREACH_THREAD_IN_PROC(p, td) {
1112 		thread_lock(td);
1113 		kp->ki_pctcpu += sched_pctcpu(td);
1114 		kp->ki_estcpu += sched_estcpu(td);
1115 		thread_unlock(td);
1116 	}
1117 }
1118 
1119 /*
1120  * Clear kinfo_proc and fill in any information that is common
1121  * to all threads in the process.
1122  * Must be called with the target process locked.
1123  */
1124 static void
1125 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
1126 {
1127 	struct thread *td0;
1128 	struct tty *tp;
1129 	struct session *sp;
1130 	struct ucred *cred;
1131 	struct sigacts *ps;
1132 	struct timeval boottime;
1133 
1134 	PROC_LOCK_ASSERT(p, MA_OWNED);
1135 	bzero(kp, sizeof(*kp));
1136 
1137 	kp->ki_structsize = sizeof(*kp);
1138 	kp->ki_paddr = p;
1139 	kp->ki_addr =/* p->p_addr; */0; /* XXX */
1140 	kp->ki_args = p->p_args;
1141 	kp->ki_textvp = p->p_textvp;
1142 #ifdef KTRACE
1143 	kp->ki_tracep = p->p_tracevp;
1144 	kp->ki_traceflag = p->p_traceflag;
1145 #endif
1146 	kp->ki_fd = p->p_fd;
1147 	kp->ki_vmspace = p->p_vmspace;
1148 	kp->ki_flag = p->p_flag;
1149 	kp->ki_flag2 = p->p_flag2;
1150 	cred = p->p_ucred;
1151 	if (cred) {
1152 		kp->ki_uid = cred->cr_uid;
1153 		kp->ki_ruid = cred->cr_ruid;
1154 		kp->ki_svuid = cred->cr_svuid;
1155 		kp->ki_cr_flags = 0;
1156 		if (cred->cr_flags & CRED_FLAG_CAPMODE)
1157 			kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
1158 		/* XXX bde doesn't like KI_NGROUPS */
1159 		if (cred->cr_ngroups > KI_NGROUPS) {
1160 			kp->ki_ngroups = KI_NGROUPS;
1161 			kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
1162 		} else
1163 			kp->ki_ngroups = cred->cr_ngroups;
1164 		bcopy(cred->cr_groups, kp->ki_groups,
1165 		    kp->ki_ngroups * sizeof(gid_t));
1166 		kp->ki_rgid = cred->cr_rgid;
1167 		kp->ki_svgid = cred->cr_svgid;
1168 		/* If jailed(cred), emulate the old P_JAILED flag. */
1169 		if (jailed(cred)) {
1170 			kp->ki_flag |= P_JAILED;
1171 			/* If inside the jail, use 0 as a jail ID. */
1172 			if (cred->cr_prison != curthread->td_ucred->cr_prison)
1173 				kp->ki_jid = cred->cr_prison->pr_id;
1174 		}
1175 		strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
1176 		    sizeof(kp->ki_loginclass));
1177 	}
1178 	ps = p->p_sigacts;
1179 	if (ps) {
1180 		mtx_lock(&ps->ps_mtx);
1181 		kp->ki_sigignore = ps->ps_sigignore;
1182 		kp->ki_sigcatch = ps->ps_sigcatch;
1183 		mtx_unlock(&ps->ps_mtx);
1184 	}
1185 	if (p->p_state != PRS_NEW &&
1186 	    p->p_state != PRS_ZOMBIE &&
1187 	    p->p_vmspace != NULL) {
1188 		struct vmspace *vm = p->p_vmspace;
1189 
1190 		kp->ki_size = vm->vm_map.size;
1191 		kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
1192 		FOREACH_THREAD_IN_PROC(p, td0) {
1193 			if (!TD_IS_SWAPPED(td0))
1194 				kp->ki_rssize += td0->td_kstack_pages;
1195 		}
1196 		kp->ki_swrss = vm->vm_swrss;
1197 		kp->ki_tsize = vm->vm_tsize;
1198 		kp->ki_dsize = vm->vm_dsize;
1199 		kp->ki_ssize = vm->vm_ssize;
1200 	} else if (p->p_state == PRS_ZOMBIE)
1201 		kp->ki_stat = SZOMB;
1202 	if (kp->ki_flag & P_INMEM)
1203 		kp->ki_sflag = PS_INMEM;
1204 	else
1205 		kp->ki_sflag = 0;
1206 	/* Calculate legacy swtime as seconds since 'swtick'. */
1207 	kp->ki_swtime = (ticks - p->p_swtick) / hz;
1208 	kp->ki_pid = p->p_pid;
1209 	kp->ki_nice = p->p_nice;
1210 	kp->ki_fibnum = p->p_fibnum;
1211 	kp->ki_start = p->p_stats->p_start;
1212 	getboottime(&boottime);
1213 	timevaladd(&kp->ki_start, &boottime);
1214 	PROC_STATLOCK(p);
1215 	rufetch(p, &kp->ki_rusage);
1216 	kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
1217 	calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
1218 	PROC_STATUNLOCK(p);
1219 	calccru(p, &kp->ki_childutime, &kp->ki_childstime);
1220 	/* Some callers want child times in a single value. */
1221 	kp->ki_childtime = kp->ki_childstime;
1222 	timevaladd(&kp->ki_childtime, &kp->ki_childutime);
1223 
1224 	FOREACH_THREAD_IN_PROC(p, td0)
1225 		kp->ki_cow += td0->td_cow;
1226 
1227 	tp = NULL;
1228 	if (p->p_pgrp) {
1229 		kp->ki_pgid = p->p_pgrp->pg_id;
1230 		kp->ki_jobc = p->p_pgrp->pg_jobc;
1231 		sp = p->p_pgrp->pg_session;
1232 
1233 		if (sp != NULL) {
1234 			kp->ki_sid = sp->s_sid;
1235 			SESS_LOCK(sp);
1236 			strlcpy(kp->ki_login, sp->s_login,
1237 			    sizeof(kp->ki_login));
1238 			if (sp->s_ttyvp)
1239 				kp->ki_kiflag |= KI_CTTY;
1240 			if (SESS_LEADER(p))
1241 				kp->ki_kiflag |= KI_SLEADER;
1242 			/* XXX proctree_lock */
1243 			tp = sp->s_ttyp;
1244 			SESS_UNLOCK(sp);
1245 		}
1246 	}
1247 	if ((p->p_flag & P_CONTROLT) && tp != NULL) {
1248 		kp->ki_tdev = tty_udev(tp);
1249 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1250 		kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
1251 		if (tp->t_session)
1252 			kp->ki_tsid = tp->t_session->s_sid;
1253 	} else {
1254 		kp->ki_tdev = NODEV;
1255 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1256 	}
1257 	if (p->p_comm[0] != '\0')
1258 		strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
1259 	if (p->p_sysent && p->p_sysent->sv_name != NULL &&
1260 	    p->p_sysent->sv_name[0] != '\0')
1261 		strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
1262 	kp->ki_siglist = p->p_siglist;
1263 	kp->ki_xstat = KW_EXITCODE(p->p_xexit, p->p_xsig);
1264 	kp->ki_acflag = p->p_acflag;
1265 	kp->ki_lock = p->p_lock;
1266 	if (p->p_pptr) {
1267 		kp->ki_ppid = p->p_oppid;
1268 		if (p->p_flag & P_TRACED)
1269 			kp->ki_tracer = p->p_pptr->p_pid;
1270 	}
1271 }
1272 
1273 /*
1274  * Fill in information that is thread specific.  Must be called with
1275  * target process locked.  If 'preferthread' is set, overwrite certain
1276  * process-related fields that are maintained for both threads and
1277  * processes.
1278  */
1279 static void
1280 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
1281 {
1282 	struct proc *p;
1283 
1284 	p = td->td_proc;
1285 	kp->ki_tdaddr = td;
1286 	PROC_LOCK_ASSERT(p, MA_OWNED);
1287 
1288 	if (preferthread)
1289 		PROC_STATLOCK(p);
1290 	thread_lock(td);
1291 	if (td->td_wmesg != NULL)
1292 		strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
1293 	else
1294 		bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
1295 	if (strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname)) >=
1296 	    sizeof(kp->ki_tdname)) {
1297 		strlcpy(kp->ki_moretdname,
1298 		    td->td_name + sizeof(kp->ki_tdname) - 1,
1299 		    sizeof(kp->ki_moretdname));
1300 	} else {
1301 		bzero(kp->ki_moretdname, sizeof(kp->ki_moretdname));
1302 	}
1303 	if (TD_ON_LOCK(td)) {
1304 		kp->ki_kiflag |= KI_LOCKBLOCK;
1305 		strlcpy(kp->ki_lockname, td->td_lockname,
1306 		    sizeof(kp->ki_lockname));
1307 	} else {
1308 		kp->ki_kiflag &= ~KI_LOCKBLOCK;
1309 		bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
1310 	}
1311 
1312 	if (p->p_state == PRS_NORMAL) { /* approximate. */
1313 		if (TD_ON_RUNQ(td) ||
1314 		    TD_CAN_RUN(td) ||
1315 		    TD_IS_RUNNING(td)) {
1316 			kp->ki_stat = SRUN;
1317 		} else if (P_SHOULDSTOP(p)) {
1318 			kp->ki_stat = SSTOP;
1319 		} else if (TD_IS_SLEEPING(td)) {
1320 			kp->ki_stat = SSLEEP;
1321 		} else if (TD_ON_LOCK(td)) {
1322 			kp->ki_stat = SLOCK;
1323 		} else {
1324 			kp->ki_stat = SWAIT;
1325 		}
1326 	} else if (p->p_state == PRS_ZOMBIE) {
1327 		kp->ki_stat = SZOMB;
1328 	} else {
1329 		kp->ki_stat = SIDL;
1330 	}
1331 
1332 	/* Things in the thread */
1333 	kp->ki_wchan = td->td_wchan;
1334 	kp->ki_pri.pri_level = td->td_priority;
1335 	kp->ki_pri.pri_native = td->td_base_pri;
1336 
1337 	/*
1338 	 * Note: legacy fields; clamp at the old NOCPU value and/or
1339 	 * the maximum u_char CPU value.
1340 	 */
1341 	if (td->td_lastcpu == NOCPU)
1342 		kp->ki_lastcpu_old = NOCPU_OLD;
1343 	else if (td->td_lastcpu > MAXCPU_OLD)
1344 		kp->ki_lastcpu_old = MAXCPU_OLD;
1345 	else
1346 		kp->ki_lastcpu_old = td->td_lastcpu;
1347 
1348 	if (td->td_oncpu == NOCPU)
1349 		kp->ki_oncpu_old = NOCPU_OLD;
1350 	else if (td->td_oncpu > MAXCPU_OLD)
1351 		kp->ki_oncpu_old = MAXCPU_OLD;
1352 	else
1353 		kp->ki_oncpu_old = td->td_oncpu;
1354 
1355 	kp->ki_lastcpu = td->td_lastcpu;
1356 	kp->ki_oncpu = td->td_oncpu;
1357 	kp->ki_tdflags = td->td_flags;
1358 	kp->ki_tid = td->td_tid;
1359 	kp->ki_numthreads = p->p_numthreads;
1360 	kp->ki_pcb = td->td_pcb;
1361 	kp->ki_kstack = (void *)td->td_kstack;
1362 	kp->ki_slptime = (ticks - td->td_slptick) / hz;
1363 	kp->ki_pri.pri_class = td->td_pri_class;
1364 	kp->ki_pri.pri_user = td->td_user_pri;
1365 
1366 	if (preferthread) {
1367 		rufetchtd(td, &kp->ki_rusage);
1368 		kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
1369 		kp->ki_pctcpu = sched_pctcpu(td);
1370 		kp->ki_estcpu = sched_estcpu(td);
1371 		kp->ki_cow = td->td_cow;
1372 	}
1373 
1374 	/* We can't get this anymore but ps etc never used it anyway. */
1375 	kp->ki_rqindex = 0;
1376 
1377 	if (preferthread)
1378 		kp->ki_siglist = td->td_siglist;
1379 	kp->ki_sigmask = td->td_sigmask;
1380 	thread_unlock(td);
1381 	if (preferthread)
1382 		PROC_STATUNLOCK(p);
1383 }
1384 
1385 /*
1386  * Fill in a kinfo_proc structure for the specified process.
1387  * Must be called with the target process locked.
1388  */
1389 void
1390 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1391 {
1392 
1393 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1394 
1395 	fill_kinfo_proc_only(p, kp);
1396 	fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1397 	fill_kinfo_aggregate(p, kp);
1398 }
1399 
1400 struct pstats *
1401 pstats_alloc(void)
1402 {
1403 
1404 	return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1405 }
1406 
1407 /*
1408  * Copy parts of p_stats; zero the rest of p_stats (statistics).
1409  */
1410 void
1411 pstats_fork(struct pstats *src, struct pstats *dst)
1412 {
1413 
1414 	bzero(&dst->pstat_startzero,
1415 	    __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1416 	bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1417 	    __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1418 }
1419 
1420 void
1421 pstats_free(struct pstats *ps)
1422 {
1423 
1424 	free(ps, M_SUBPROC);
1425 }
1426 
1427 #ifdef COMPAT_FREEBSD32
1428 
1429 /*
1430  * This function is typically used to copy out the kernel address, so
1431  * it can be replaced by assignment of zero.
1432  */
1433 static inline uint32_t
1434 ptr32_trim(const void *ptr)
1435 {
1436 	uintptr_t uptr;
1437 
1438 	uptr = (uintptr_t)ptr;
1439 	return ((uptr > UINT_MAX) ? 0 : uptr);
1440 }
1441 
1442 #define PTRTRIM_CP(src,dst,fld) \
1443 	do { (dst).fld = ptr32_trim((src).fld); } while (0)
1444 
1445 static void
1446 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1447 {
1448 	int i;
1449 
1450 	bzero(ki32, sizeof(struct kinfo_proc32));
1451 	ki32->ki_structsize = sizeof(struct kinfo_proc32);
1452 	CP(*ki, *ki32, ki_layout);
1453 	PTRTRIM_CP(*ki, *ki32, ki_args);
1454 	PTRTRIM_CP(*ki, *ki32, ki_paddr);
1455 	PTRTRIM_CP(*ki, *ki32, ki_addr);
1456 	PTRTRIM_CP(*ki, *ki32, ki_tracep);
1457 	PTRTRIM_CP(*ki, *ki32, ki_textvp);
1458 	PTRTRIM_CP(*ki, *ki32, ki_fd);
1459 	PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1460 	PTRTRIM_CP(*ki, *ki32, ki_wchan);
1461 	CP(*ki, *ki32, ki_pid);
1462 	CP(*ki, *ki32, ki_ppid);
1463 	CP(*ki, *ki32, ki_pgid);
1464 	CP(*ki, *ki32, ki_tpgid);
1465 	CP(*ki, *ki32, ki_sid);
1466 	CP(*ki, *ki32, ki_tsid);
1467 	CP(*ki, *ki32, ki_jobc);
1468 	CP(*ki, *ki32, ki_tdev);
1469 	CP(*ki, *ki32, ki_tdev_freebsd11);
1470 	CP(*ki, *ki32, ki_siglist);
1471 	CP(*ki, *ki32, ki_sigmask);
1472 	CP(*ki, *ki32, ki_sigignore);
1473 	CP(*ki, *ki32, ki_sigcatch);
1474 	CP(*ki, *ki32, ki_uid);
1475 	CP(*ki, *ki32, ki_ruid);
1476 	CP(*ki, *ki32, ki_svuid);
1477 	CP(*ki, *ki32, ki_rgid);
1478 	CP(*ki, *ki32, ki_svgid);
1479 	CP(*ki, *ki32, ki_ngroups);
1480 	for (i = 0; i < KI_NGROUPS; i++)
1481 		CP(*ki, *ki32, ki_groups[i]);
1482 	CP(*ki, *ki32, ki_size);
1483 	CP(*ki, *ki32, ki_rssize);
1484 	CP(*ki, *ki32, ki_swrss);
1485 	CP(*ki, *ki32, ki_tsize);
1486 	CP(*ki, *ki32, ki_dsize);
1487 	CP(*ki, *ki32, ki_ssize);
1488 	CP(*ki, *ki32, ki_xstat);
1489 	CP(*ki, *ki32, ki_acflag);
1490 	CP(*ki, *ki32, ki_pctcpu);
1491 	CP(*ki, *ki32, ki_estcpu);
1492 	CP(*ki, *ki32, ki_slptime);
1493 	CP(*ki, *ki32, ki_swtime);
1494 	CP(*ki, *ki32, ki_cow);
1495 	CP(*ki, *ki32, ki_runtime);
1496 	TV_CP(*ki, *ki32, ki_start);
1497 	TV_CP(*ki, *ki32, ki_childtime);
1498 	CP(*ki, *ki32, ki_flag);
1499 	CP(*ki, *ki32, ki_kiflag);
1500 	CP(*ki, *ki32, ki_traceflag);
1501 	CP(*ki, *ki32, ki_stat);
1502 	CP(*ki, *ki32, ki_nice);
1503 	CP(*ki, *ki32, ki_lock);
1504 	CP(*ki, *ki32, ki_rqindex);
1505 	CP(*ki, *ki32, ki_oncpu);
1506 	CP(*ki, *ki32, ki_lastcpu);
1507 
1508 	/* XXX TODO: wrap cpu value as appropriate */
1509 	CP(*ki, *ki32, ki_oncpu_old);
1510 	CP(*ki, *ki32, ki_lastcpu_old);
1511 
1512 	bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1513 	bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1514 	bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1515 	bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1516 	bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1517 	bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1518 	bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1519 	bcopy(ki->ki_moretdname, ki32->ki_moretdname, MAXCOMLEN - TDNAMLEN + 1);
1520 	CP(*ki, *ki32, ki_tracer);
1521 	CP(*ki, *ki32, ki_flag2);
1522 	CP(*ki, *ki32, ki_fibnum);
1523 	CP(*ki, *ki32, ki_cr_flags);
1524 	CP(*ki, *ki32, ki_jid);
1525 	CP(*ki, *ki32, ki_numthreads);
1526 	CP(*ki, *ki32, ki_tid);
1527 	CP(*ki, *ki32, ki_pri);
1528 	freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1529 	freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1530 	PTRTRIM_CP(*ki, *ki32, ki_pcb);
1531 	PTRTRIM_CP(*ki, *ki32, ki_kstack);
1532 	PTRTRIM_CP(*ki, *ki32, ki_udata);
1533 	PTRTRIM_CP(*ki, *ki32, ki_tdaddr);
1534 	CP(*ki, *ki32, ki_sflag);
1535 	CP(*ki, *ki32, ki_tdflags);
1536 }
1537 #endif
1538 
1539 static ssize_t
1540 kern_proc_out_size(struct proc *p, int flags)
1541 {
1542 	ssize_t size = 0;
1543 
1544 	PROC_LOCK_ASSERT(p, MA_OWNED);
1545 
1546 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1547 #ifdef COMPAT_FREEBSD32
1548 		if ((flags & KERN_PROC_MASK32) != 0) {
1549 			size += sizeof(struct kinfo_proc32);
1550 		} else
1551 #endif
1552 			size += sizeof(struct kinfo_proc);
1553 	} else {
1554 #ifdef COMPAT_FREEBSD32
1555 		if ((flags & KERN_PROC_MASK32) != 0)
1556 			size += sizeof(struct kinfo_proc32) * p->p_numthreads;
1557 		else
1558 #endif
1559 			size += sizeof(struct kinfo_proc) * p->p_numthreads;
1560 	}
1561 	PROC_UNLOCK(p);
1562 	return (size);
1563 }
1564 
1565 int
1566 kern_proc_out(struct proc *p, struct sbuf *sb, int flags)
1567 {
1568 	struct thread *td;
1569 	struct kinfo_proc ki;
1570 #ifdef COMPAT_FREEBSD32
1571 	struct kinfo_proc32 ki32;
1572 #endif
1573 	int error;
1574 
1575 	PROC_LOCK_ASSERT(p, MA_OWNED);
1576 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1577 
1578 	error = 0;
1579 	fill_kinfo_proc(p, &ki);
1580 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1581 #ifdef COMPAT_FREEBSD32
1582 		if ((flags & KERN_PROC_MASK32) != 0) {
1583 			freebsd32_kinfo_proc_out(&ki, &ki32);
1584 			if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1585 				error = ENOMEM;
1586 		} else
1587 #endif
1588 			if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1589 				error = ENOMEM;
1590 	} else {
1591 		FOREACH_THREAD_IN_PROC(p, td) {
1592 			fill_kinfo_thread(td, &ki, 1);
1593 #ifdef COMPAT_FREEBSD32
1594 			if ((flags & KERN_PROC_MASK32) != 0) {
1595 				freebsd32_kinfo_proc_out(&ki, &ki32);
1596 				if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1597 					error = ENOMEM;
1598 			} else
1599 #endif
1600 				if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1601 					error = ENOMEM;
1602 			if (error != 0)
1603 				break;
1604 		}
1605 	}
1606 	PROC_UNLOCK(p);
1607 	return (error);
1608 }
1609 
1610 static int
1611 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
1612 {
1613 	struct sbuf sb;
1614 	struct kinfo_proc ki;
1615 	int error, error2;
1616 
1617 	if (req->oldptr == NULL)
1618 		return (SYSCTL_OUT(req, 0, kern_proc_out_size(p, flags)));
1619 
1620 	sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req);
1621 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1622 	error = kern_proc_out(p, &sb, flags);
1623 	error2 = sbuf_finish(&sb);
1624 	sbuf_delete(&sb);
1625 	if (error != 0)
1626 		return (error);
1627 	else if (error2 != 0)
1628 		return (error2);
1629 	return (0);
1630 }
1631 
1632 int
1633 proc_iterate(int (*cb)(struct proc *, void *), void *cbarg)
1634 {
1635 	struct proc *p;
1636 	int error, i, j;
1637 
1638 	for (i = 0; i < pidhashlock + 1; i++) {
1639 		sx_slock(&pidhashtbl_lock[i]);
1640 		for (j = i; j <= pidhash; j += pidhashlock + 1) {
1641 			LIST_FOREACH(p, &pidhashtbl[j], p_hash) {
1642 				if (p->p_state == PRS_NEW)
1643 					continue;
1644 				error = cb(p, cbarg);
1645 				PROC_LOCK_ASSERT(p, MA_NOTOWNED);
1646 				if (error != 0) {
1647 					sx_sunlock(&pidhashtbl_lock[i]);
1648 					return (error);
1649 				}
1650 			}
1651 		}
1652 		sx_sunlock(&pidhashtbl_lock[i]);
1653 	}
1654 	return (0);
1655 }
1656 
1657 struct kern_proc_out_args {
1658 	struct sysctl_req *req;
1659 	int flags;
1660 	int oid_number;
1661 	int *name;
1662 };
1663 
1664 static int
1665 sysctl_kern_proc_iterate(struct proc *p, void *origarg)
1666 {
1667 	struct kern_proc_out_args *arg = origarg;
1668 	int *name = arg->name;
1669 	int oid_number = arg->oid_number;
1670 	int flags = arg->flags;
1671 	struct sysctl_req *req = arg->req;
1672 	int error = 0;
1673 
1674 	PROC_LOCK(p);
1675 
1676 	KASSERT(p->p_ucred != NULL,
1677 	    ("process credential is NULL for non-NEW proc"));
1678 	/*
1679 	 * Show a user only appropriate processes.
1680 	 */
1681 	if (p_cansee(curthread, p))
1682 		goto skip;
1683 	/*
1684 	 * TODO - make more efficient (see notes below).
1685 	 * do by session.
1686 	 */
1687 	switch (oid_number) {
1688 	case KERN_PROC_GID:
1689 		if (p->p_ucred->cr_gid != (gid_t)name[0])
1690 			goto skip;
1691 		break;
1692 
1693 	case KERN_PROC_PGRP:
1694 		/* could do this by traversing pgrp */
1695 		if (p->p_pgrp == NULL ||
1696 		    p->p_pgrp->pg_id != (pid_t)name[0])
1697 			goto skip;
1698 		break;
1699 
1700 	case KERN_PROC_RGID:
1701 		if (p->p_ucred->cr_rgid != (gid_t)name[0])
1702 			goto skip;
1703 		break;
1704 
1705 	case KERN_PROC_SESSION:
1706 		if (p->p_session == NULL ||
1707 		    p->p_session->s_sid != (pid_t)name[0])
1708 			goto skip;
1709 		break;
1710 
1711 	case KERN_PROC_TTY:
1712 		if ((p->p_flag & P_CONTROLT) == 0 ||
1713 		    p->p_session == NULL)
1714 			goto skip;
1715 		/* XXX proctree_lock */
1716 		SESS_LOCK(p->p_session);
1717 		if (p->p_session->s_ttyp == NULL ||
1718 		    tty_udev(p->p_session->s_ttyp) !=
1719 		    (dev_t)name[0]) {
1720 			SESS_UNLOCK(p->p_session);
1721 			goto skip;
1722 		}
1723 		SESS_UNLOCK(p->p_session);
1724 		break;
1725 
1726 	case KERN_PROC_UID:
1727 		if (p->p_ucred->cr_uid != (uid_t)name[0])
1728 			goto skip;
1729 		break;
1730 
1731 	case KERN_PROC_RUID:
1732 		if (p->p_ucred->cr_ruid != (uid_t)name[0])
1733 			goto skip;
1734 		break;
1735 
1736 	case KERN_PROC_PROC:
1737 		break;
1738 
1739 	default:
1740 		break;
1741 	}
1742 	error = sysctl_out_proc(p, req, flags);
1743 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
1744 	return (error);
1745 skip:
1746 	PROC_UNLOCK(p);
1747 	return (0);
1748 }
1749 
1750 static int
1751 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1752 {
1753 	struct kern_proc_out_args iterarg;
1754 	int *name = (int *)arg1;
1755 	u_int namelen = arg2;
1756 	struct proc *p;
1757 	int flags, oid_number;
1758 	int error = 0;
1759 
1760 	oid_number = oidp->oid_number;
1761 	if (oid_number != KERN_PROC_ALL &&
1762 	    (oid_number & KERN_PROC_INC_THREAD) == 0)
1763 		flags = KERN_PROC_NOTHREADS;
1764 	else {
1765 		flags = 0;
1766 		oid_number &= ~KERN_PROC_INC_THREAD;
1767 	}
1768 #ifdef COMPAT_FREEBSD32
1769 	if (req->flags & SCTL_MASK32)
1770 		flags |= KERN_PROC_MASK32;
1771 #endif
1772 	if (oid_number == KERN_PROC_PID) {
1773 		if (namelen != 1)
1774 			return (EINVAL);
1775 		error = sysctl_wire_old_buffer(req, 0);
1776 		if (error)
1777 			return (error);
1778 		error = pget((pid_t)name[0], PGET_CANSEE, &p);
1779 		if (error == 0)
1780 			error = sysctl_out_proc(p, req, flags);
1781 		return (error);
1782 	}
1783 
1784 	switch (oid_number) {
1785 	case KERN_PROC_ALL:
1786 		if (namelen != 0)
1787 			return (EINVAL);
1788 		break;
1789 	case KERN_PROC_PROC:
1790 		if (namelen != 0 && namelen != 1)
1791 			return (EINVAL);
1792 		break;
1793 	default:
1794 		if (namelen != 1)
1795 			return (EINVAL);
1796 		break;
1797 	}
1798 
1799 	if (req->oldptr == NULL) {
1800 		/* overestimate by 5 procs */
1801 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1802 		if (error)
1803 			return (error);
1804 	} else {
1805 		error = sysctl_wire_old_buffer(req, 0);
1806 		if (error != 0)
1807 			return (error);
1808 	}
1809 	iterarg.flags = flags;
1810 	iterarg.oid_number = oid_number;
1811 	iterarg.req = req;
1812 	iterarg.name = name;
1813 	error = proc_iterate(sysctl_kern_proc_iterate, &iterarg);
1814 	return (error);
1815 }
1816 
1817 struct pargs *
1818 pargs_alloc(int len)
1819 {
1820 	struct pargs *pa;
1821 
1822 	pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1823 		M_WAITOK);
1824 	refcount_init(&pa->ar_ref, 1);
1825 	pa->ar_length = len;
1826 	return (pa);
1827 }
1828 
1829 static void
1830 pargs_free(struct pargs *pa)
1831 {
1832 
1833 	free(pa, M_PARGS);
1834 }
1835 
1836 void
1837 pargs_hold(struct pargs *pa)
1838 {
1839 
1840 	if (pa == NULL)
1841 		return;
1842 	refcount_acquire(&pa->ar_ref);
1843 }
1844 
1845 void
1846 pargs_drop(struct pargs *pa)
1847 {
1848 
1849 	if (pa == NULL)
1850 		return;
1851 	if (refcount_release(&pa->ar_ref))
1852 		pargs_free(pa);
1853 }
1854 
1855 static int
1856 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1857     size_t len)
1858 {
1859 	ssize_t n;
1860 
1861 	/*
1862 	 * This may return a short read if the string is shorter than the chunk
1863 	 * and is aligned at the end of the page, and the following page is not
1864 	 * mapped.
1865 	 */
1866 	n = proc_readmem(td, p, (vm_offset_t)sptr, buf, len);
1867 	if (n <= 0)
1868 		return (ENOMEM);
1869 	return (0);
1870 }
1871 
1872 #define PROC_AUXV_MAX	256	/* Safety limit on auxv size. */
1873 
1874 enum proc_vector_type {
1875 	PROC_ARG,
1876 	PROC_ENV,
1877 	PROC_AUX,
1878 };
1879 
1880 #ifdef COMPAT_FREEBSD32
1881 static int
1882 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1883     size_t *vsizep, enum proc_vector_type type)
1884 {
1885 	struct freebsd32_ps_strings pss;
1886 	Elf32_Auxinfo aux;
1887 	vm_offset_t vptr, ptr;
1888 	uint32_t *proc_vector32;
1889 	char **proc_vector;
1890 	size_t vsize, size;
1891 	int i, error;
1892 
1893 	error = 0;
1894 	if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss,
1895 	    sizeof(pss)) != sizeof(pss))
1896 		return (ENOMEM);
1897 	switch (type) {
1898 	case PROC_ARG:
1899 		vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1900 		vsize = pss.ps_nargvstr;
1901 		if (vsize > ARG_MAX)
1902 			return (ENOEXEC);
1903 		size = vsize * sizeof(int32_t);
1904 		break;
1905 	case PROC_ENV:
1906 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1907 		vsize = pss.ps_nenvstr;
1908 		if (vsize > ARG_MAX)
1909 			return (ENOEXEC);
1910 		size = vsize * sizeof(int32_t);
1911 		break;
1912 	case PROC_AUX:
1913 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1914 		    (pss.ps_nenvstr + 1) * sizeof(int32_t);
1915 		if (vptr % 4 != 0)
1916 			return (ENOEXEC);
1917 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1918 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1919 			    sizeof(aux))
1920 				return (ENOMEM);
1921 			if (aux.a_type == AT_NULL)
1922 				break;
1923 			ptr += sizeof(aux);
1924 		}
1925 		if (aux.a_type != AT_NULL)
1926 			return (ENOEXEC);
1927 		vsize = i + 1;
1928 		size = vsize * sizeof(aux);
1929 		break;
1930 	default:
1931 		KASSERT(0, ("Wrong proc vector type: %d", type));
1932 		return (EINVAL);
1933 	}
1934 	proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1935 	if (proc_readmem(td, p, vptr, proc_vector32, size) != size) {
1936 		error = ENOMEM;
1937 		goto done;
1938 	}
1939 	if (type == PROC_AUX) {
1940 		*proc_vectorp = (char **)proc_vector32;
1941 		*vsizep = vsize;
1942 		return (0);
1943 	}
1944 	proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1945 	for (i = 0; i < (int)vsize; i++)
1946 		proc_vector[i] = PTRIN(proc_vector32[i]);
1947 	*proc_vectorp = proc_vector;
1948 	*vsizep = vsize;
1949 done:
1950 	free(proc_vector32, M_TEMP);
1951 	return (error);
1952 }
1953 #endif
1954 
1955 static int
1956 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1957     size_t *vsizep, enum proc_vector_type type)
1958 {
1959 	struct ps_strings pss;
1960 	Elf_Auxinfo aux;
1961 	vm_offset_t vptr, ptr;
1962 	char **proc_vector;
1963 	size_t vsize, size;
1964 	int i;
1965 
1966 #ifdef COMPAT_FREEBSD32
1967 	if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1968 		return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1969 #endif
1970 	if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss,
1971 	    sizeof(pss)) != sizeof(pss))
1972 		return (ENOMEM);
1973 	switch (type) {
1974 	case PROC_ARG:
1975 		vptr = (vm_offset_t)pss.ps_argvstr;
1976 		vsize = pss.ps_nargvstr;
1977 		if (vsize > ARG_MAX)
1978 			return (ENOEXEC);
1979 		size = vsize * sizeof(char *);
1980 		break;
1981 	case PROC_ENV:
1982 		vptr = (vm_offset_t)pss.ps_envstr;
1983 		vsize = pss.ps_nenvstr;
1984 		if (vsize > ARG_MAX)
1985 			return (ENOEXEC);
1986 		size = vsize * sizeof(char *);
1987 		break;
1988 	case PROC_AUX:
1989 		/*
1990 		 * The aux array is just above env array on the stack. Check
1991 		 * that the address is naturally aligned.
1992 		 */
1993 		vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1994 		    * sizeof(char *);
1995 #if __ELF_WORD_SIZE == 64
1996 		if (vptr % sizeof(uint64_t) != 0)
1997 #else
1998 		if (vptr % sizeof(uint32_t) != 0)
1999 #endif
2000 			return (ENOEXEC);
2001 		/*
2002 		 * We count the array size reading the aux vectors from the
2003 		 * stack until AT_NULL vector is returned.  So (to keep the code
2004 		 * simple) we read the process stack twice: the first time here
2005 		 * to find the size and the second time when copying the vectors
2006 		 * to the allocated proc_vector.
2007 		 */
2008 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
2009 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
2010 			    sizeof(aux))
2011 				return (ENOMEM);
2012 			if (aux.a_type == AT_NULL)
2013 				break;
2014 			ptr += sizeof(aux);
2015 		}
2016 		/*
2017 		 * If the PROC_AUXV_MAX entries are iterated over, and we have
2018 		 * not reached AT_NULL, it is most likely we are reading wrong
2019 		 * data: either the process doesn't have auxv array or data has
2020 		 * been modified. Return the error in this case.
2021 		 */
2022 		if (aux.a_type != AT_NULL)
2023 			return (ENOEXEC);
2024 		vsize = i + 1;
2025 		size = vsize * sizeof(aux);
2026 		break;
2027 	default:
2028 		KASSERT(0, ("Wrong proc vector type: %d", type));
2029 		return (EINVAL); /* In case we are built without INVARIANTS. */
2030 	}
2031 	proc_vector = malloc(size, M_TEMP, M_WAITOK);
2032 	if (proc_readmem(td, p, vptr, proc_vector, size) != size) {
2033 		free(proc_vector, M_TEMP);
2034 		return (ENOMEM);
2035 	}
2036 	*proc_vectorp = proc_vector;
2037 	*vsizep = vsize;
2038 
2039 	return (0);
2040 }
2041 
2042 #define GET_PS_STRINGS_CHUNK_SZ	256	/* Chunk size (bytes) for ps_strings operations. */
2043 
2044 static int
2045 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
2046     enum proc_vector_type type)
2047 {
2048 	size_t done, len, nchr, vsize;
2049 	int error, i;
2050 	char **proc_vector, *sptr;
2051 	char pss_string[GET_PS_STRINGS_CHUNK_SZ];
2052 
2053 	PROC_ASSERT_HELD(p);
2054 
2055 	/*
2056 	 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
2057 	 */
2058 	nchr = 2 * (PATH_MAX + ARG_MAX);
2059 
2060 	error = get_proc_vector(td, p, &proc_vector, &vsize, type);
2061 	if (error != 0)
2062 		return (error);
2063 	for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
2064 		/*
2065 		 * The program may have scribbled into its argv array, e.g. to
2066 		 * remove some arguments.  If that has happened, break out
2067 		 * before trying to read from NULL.
2068 		 */
2069 		if (proc_vector[i] == NULL)
2070 			break;
2071 		for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
2072 			error = proc_read_string(td, p, sptr, pss_string,
2073 			    sizeof(pss_string));
2074 			if (error != 0)
2075 				goto done;
2076 			len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
2077 			if (done + len >= nchr)
2078 				len = nchr - done - 1;
2079 			sbuf_bcat(sb, pss_string, len);
2080 			if (len != GET_PS_STRINGS_CHUNK_SZ)
2081 				break;
2082 			done += GET_PS_STRINGS_CHUNK_SZ;
2083 		}
2084 		sbuf_bcat(sb, "", 1);
2085 		done += len + 1;
2086 	}
2087 done:
2088 	free(proc_vector, M_TEMP);
2089 	return (error);
2090 }
2091 
2092 int
2093 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
2094 {
2095 
2096 	return (get_ps_strings(curthread, p, sb, PROC_ARG));
2097 }
2098 
2099 int
2100 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
2101 {
2102 
2103 	return (get_ps_strings(curthread, p, sb, PROC_ENV));
2104 }
2105 
2106 int
2107 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb)
2108 {
2109 	size_t vsize, size;
2110 	char **auxv;
2111 	int error;
2112 
2113 	error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX);
2114 	if (error == 0) {
2115 #ifdef COMPAT_FREEBSD32
2116 		if (SV_PROC_FLAG(p, SV_ILP32) != 0)
2117 			size = vsize * sizeof(Elf32_Auxinfo);
2118 		else
2119 #endif
2120 			size = vsize * sizeof(Elf_Auxinfo);
2121 		if (sbuf_bcat(sb, auxv, size) != 0)
2122 			error = ENOMEM;
2123 		free(auxv, M_TEMP);
2124 	}
2125 	return (error);
2126 }
2127 
2128 /*
2129  * This sysctl allows a process to retrieve the argument list or process
2130  * title for another process without groping around in the address space
2131  * of the other process.  It also allow a process to set its own "process
2132  * title to a string of its own choice.
2133  */
2134 static int
2135 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
2136 {
2137 	int *name = (int *)arg1;
2138 	u_int namelen = arg2;
2139 	struct pargs *newpa, *pa;
2140 	struct proc *p;
2141 	struct sbuf sb;
2142 	int flags, error = 0, error2;
2143 	pid_t pid;
2144 
2145 	if (namelen != 1)
2146 		return (EINVAL);
2147 
2148 	pid = (pid_t)name[0];
2149 	/*
2150 	 * If the query is for this process and it is single-threaded, there
2151 	 * is nobody to modify pargs, thus we can just read.
2152 	 */
2153 	p = curproc;
2154 	if (pid == p->p_pid && p->p_numthreads == 1 && req->newptr == NULL &&
2155 	    (pa = p->p_args) != NULL)
2156 		return (SYSCTL_OUT(req, pa->ar_args, pa->ar_length));
2157 
2158 	flags = PGET_CANSEE;
2159 	if (req->newptr != NULL)
2160 		flags |= PGET_ISCURRENT;
2161 	error = pget(pid, flags, &p);
2162 	if (error)
2163 		return (error);
2164 
2165 	pa = p->p_args;
2166 	if (pa != NULL) {
2167 		pargs_hold(pa);
2168 		PROC_UNLOCK(p);
2169 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
2170 		pargs_drop(pa);
2171 	} else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
2172 		_PHOLD(p);
2173 		PROC_UNLOCK(p);
2174 		sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2175 		sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2176 		error = proc_getargv(curthread, p, &sb);
2177 		error2 = sbuf_finish(&sb);
2178 		PRELE(p);
2179 		sbuf_delete(&sb);
2180 		if (error == 0 && error2 != 0)
2181 			error = error2;
2182 	} else {
2183 		PROC_UNLOCK(p);
2184 	}
2185 	if (error != 0 || req->newptr == NULL)
2186 		return (error);
2187 
2188 	if (req->newlen > ps_arg_cache_limit - sizeof(struct pargs))
2189 		return (ENOMEM);
2190 
2191 	if (req->newlen == 0) {
2192 		/*
2193 		 * Clear the argument pointer, so that we'll fetch arguments
2194 		 * with proc_getargv() until further notice.
2195 		 */
2196 		newpa = NULL;
2197 	} else {
2198 		newpa = pargs_alloc(req->newlen);
2199 		error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
2200 		if (error != 0) {
2201 			pargs_free(newpa);
2202 			return (error);
2203 		}
2204 	}
2205 	PROC_LOCK(p);
2206 	pa = p->p_args;
2207 	p->p_args = newpa;
2208 	PROC_UNLOCK(p);
2209 	pargs_drop(pa);
2210 	return (0);
2211 }
2212 
2213 /*
2214  * This sysctl allows a process to retrieve environment of another process.
2215  */
2216 static int
2217 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
2218 {
2219 	int *name = (int *)arg1;
2220 	u_int namelen = arg2;
2221 	struct proc *p;
2222 	struct sbuf sb;
2223 	int error, error2;
2224 
2225 	if (namelen != 1)
2226 		return (EINVAL);
2227 
2228 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2229 	if (error != 0)
2230 		return (error);
2231 	if ((p->p_flag & P_SYSTEM) != 0) {
2232 		PRELE(p);
2233 		return (0);
2234 	}
2235 
2236 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2237 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2238 	error = proc_getenvv(curthread, p, &sb);
2239 	error2 = sbuf_finish(&sb);
2240 	PRELE(p);
2241 	sbuf_delete(&sb);
2242 	return (error != 0 ? error : error2);
2243 }
2244 
2245 /*
2246  * This sysctl allows a process to retrieve ELF auxiliary vector of
2247  * another process.
2248  */
2249 static int
2250 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
2251 {
2252 	int *name = (int *)arg1;
2253 	u_int namelen = arg2;
2254 	struct proc *p;
2255 	struct sbuf sb;
2256 	int error, error2;
2257 
2258 	if (namelen != 1)
2259 		return (EINVAL);
2260 
2261 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2262 	if (error != 0)
2263 		return (error);
2264 	if ((p->p_flag & P_SYSTEM) != 0) {
2265 		PRELE(p);
2266 		return (0);
2267 	}
2268 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2269 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2270 	error = proc_getauxv(curthread, p, &sb);
2271 	error2 = sbuf_finish(&sb);
2272 	PRELE(p);
2273 	sbuf_delete(&sb);
2274 	return (error != 0 ? error : error2);
2275 }
2276 
2277 /*
2278  * This sysctl allows a process to retrieve the path of the executable for
2279  * itself or another process.
2280  */
2281 static int
2282 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
2283 {
2284 	pid_t *pidp = (pid_t *)arg1;
2285 	unsigned int arglen = arg2;
2286 	struct proc *p;
2287 	struct vnode *vp;
2288 	char *retbuf, *freebuf;
2289 	int error;
2290 
2291 	if (arglen != 1)
2292 		return (EINVAL);
2293 	if (*pidp == -1) {	/* -1 means this process */
2294 		p = req->td->td_proc;
2295 	} else {
2296 		error = pget(*pidp, PGET_CANSEE, &p);
2297 		if (error != 0)
2298 			return (error);
2299 	}
2300 
2301 	vp = p->p_textvp;
2302 	if (vp == NULL) {
2303 		if (*pidp != -1)
2304 			PROC_UNLOCK(p);
2305 		return (0);
2306 	}
2307 	vref(vp);
2308 	if (*pidp != -1)
2309 		PROC_UNLOCK(p);
2310 	error = vn_fullpath(vp, &retbuf, &freebuf);
2311 	vrele(vp);
2312 	if (error)
2313 		return (error);
2314 	error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
2315 	free(freebuf, M_TEMP);
2316 	return (error);
2317 }
2318 
2319 static int
2320 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
2321 {
2322 	struct proc *p;
2323 	char *sv_name;
2324 	int *name;
2325 	int namelen;
2326 	int error;
2327 
2328 	namelen = arg2;
2329 	if (namelen != 1)
2330 		return (EINVAL);
2331 
2332 	name = (int *)arg1;
2333 	error = pget((pid_t)name[0], PGET_CANSEE, &p);
2334 	if (error != 0)
2335 		return (error);
2336 	sv_name = p->p_sysent->sv_name;
2337 	PROC_UNLOCK(p);
2338 	return (sysctl_handle_string(oidp, sv_name, 0, req));
2339 }
2340 
2341 #ifdef KINFO_OVMENTRY_SIZE
2342 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
2343 #endif
2344 
2345 #ifdef COMPAT_FREEBSD7
2346 static int
2347 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
2348 {
2349 	vm_map_entry_t entry, tmp_entry;
2350 	unsigned int last_timestamp;
2351 	char *fullpath, *freepath;
2352 	struct kinfo_ovmentry *kve;
2353 	struct vattr va;
2354 	struct ucred *cred;
2355 	int error, *name;
2356 	struct vnode *vp;
2357 	struct proc *p;
2358 	vm_map_t map;
2359 	struct vmspace *vm;
2360 
2361 	name = (int *)arg1;
2362 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2363 	if (error != 0)
2364 		return (error);
2365 	vm = vmspace_acquire_ref(p);
2366 	if (vm == NULL) {
2367 		PRELE(p);
2368 		return (ESRCH);
2369 	}
2370 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2371 
2372 	map = &vm->vm_map;
2373 	vm_map_lock_read(map);
2374 	VM_MAP_ENTRY_FOREACH(entry, map) {
2375 		vm_object_t obj, tobj, lobj;
2376 		vm_offset_t addr;
2377 
2378 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2379 			continue;
2380 
2381 		bzero(kve, sizeof(*kve));
2382 		kve->kve_structsize = sizeof(*kve);
2383 
2384 		kve->kve_private_resident = 0;
2385 		obj = entry->object.vm_object;
2386 		if (obj != NULL) {
2387 			VM_OBJECT_RLOCK(obj);
2388 			if (obj->shadow_count == 1)
2389 				kve->kve_private_resident =
2390 				    obj->resident_page_count;
2391 		}
2392 		kve->kve_resident = 0;
2393 		addr = entry->start;
2394 		while (addr < entry->end) {
2395 			if (pmap_extract(map->pmap, addr))
2396 				kve->kve_resident++;
2397 			addr += PAGE_SIZE;
2398 		}
2399 
2400 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2401 			if (tobj != obj) {
2402 				VM_OBJECT_RLOCK(tobj);
2403 				kve->kve_offset += tobj->backing_object_offset;
2404 			}
2405 			if (lobj != obj)
2406 				VM_OBJECT_RUNLOCK(lobj);
2407 			lobj = tobj;
2408 		}
2409 
2410 		kve->kve_start = (void*)entry->start;
2411 		kve->kve_end = (void*)entry->end;
2412 		kve->kve_offset += (off_t)entry->offset;
2413 
2414 		if (entry->protection & VM_PROT_READ)
2415 			kve->kve_protection |= KVME_PROT_READ;
2416 		if (entry->protection & VM_PROT_WRITE)
2417 			kve->kve_protection |= KVME_PROT_WRITE;
2418 		if (entry->protection & VM_PROT_EXECUTE)
2419 			kve->kve_protection |= KVME_PROT_EXEC;
2420 
2421 		if (entry->eflags & MAP_ENTRY_COW)
2422 			kve->kve_flags |= KVME_FLAG_COW;
2423 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2424 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2425 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2426 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2427 
2428 		last_timestamp = map->timestamp;
2429 		vm_map_unlock_read(map);
2430 
2431 		kve->kve_fileid = 0;
2432 		kve->kve_fsid = 0;
2433 		freepath = NULL;
2434 		fullpath = "";
2435 		if (lobj) {
2436 			kve->kve_type = vm_object_kvme_type(lobj, &vp);
2437 			if (kve->kve_type == KVME_TYPE_MGTDEVICE)
2438 				kve->kve_type = KVME_TYPE_UNKNOWN;
2439 			if (vp != NULL)
2440 				vref(vp);
2441 			if (lobj != obj)
2442 				VM_OBJECT_RUNLOCK(lobj);
2443 
2444 			kve->kve_ref_count = obj->ref_count;
2445 			kve->kve_shadow_count = obj->shadow_count;
2446 			VM_OBJECT_RUNLOCK(obj);
2447 			if (vp != NULL) {
2448 				vn_fullpath(vp, &fullpath, &freepath);
2449 				cred = curthread->td_ucred;
2450 				vn_lock(vp, LK_SHARED | LK_RETRY);
2451 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2452 					kve->kve_fileid = va.va_fileid;
2453 					/* truncate */
2454 					kve->kve_fsid = va.va_fsid;
2455 				}
2456 				vput(vp);
2457 			}
2458 		} else {
2459 			kve->kve_type = KVME_TYPE_NONE;
2460 			kve->kve_ref_count = 0;
2461 			kve->kve_shadow_count = 0;
2462 		}
2463 
2464 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2465 		if (freepath != NULL)
2466 			free(freepath, M_TEMP);
2467 
2468 		error = SYSCTL_OUT(req, kve, sizeof(*kve));
2469 		vm_map_lock_read(map);
2470 		if (error)
2471 			break;
2472 		if (last_timestamp != map->timestamp) {
2473 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2474 			entry = tmp_entry;
2475 		}
2476 	}
2477 	vm_map_unlock_read(map);
2478 	vmspace_free(vm);
2479 	PRELE(p);
2480 	free(kve, M_TEMP);
2481 	return (error);
2482 }
2483 #endif	/* COMPAT_FREEBSD7 */
2484 
2485 #ifdef KINFO_VMENTRY_SIZE
2486 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2487 #endif
2488 
2489 void
2490 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry,
2491     int *resident_count, bool *super)
2492 {
2493 	vm_object_t obj, tobj;
2494 	vm_page_t m, m_adv;
2495 	vm_offset_t addr;
2496 	vm_paddr_t pa;
2497 	vm_pindex_t pi, pi_adv, pindex;
2498 
2499 	*super = false;
2500 	*resident_count = 0;
2501 	if (vmmap_skip_res_cnt)
2502 		return;
2503 
2504 	pa = 0;
2505 	obj = entry->object.vm_object;
2506 	addr = entry->start;
2507 	m_adv = NULL;
2508 	pi = OFF_TO_IDX(entry->offset);
2509 	for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) {
2510 		if (m_adv != NULL) {
2511 			m = m_adv;
2512 		} else {
2513 			pi_adv = atop(entry->end - addr);
2514 			pindex = pi;
2515 			for (tobj = obj;; tobj = tobj->backing_object) {
2516 				m = vm_page_find_least(tobj, pindex);
2517 				if (m != NULL) {
2518 					if (m->pindex == pindex)
2519 						break;
2520 					if (pi_adv > m->pindex - pindex) {
2521 						pi_adv = m->pindex - pindex;
2522 						m_adv = m;
2523 					}
2524 				}
2525 				if (tobj->backing_object == NULL)
2526 					goto next;
2527 				pindex += OFF_TO_IDX(tobj->
2528 				    backing_object_offset);
2529 			}
2530 		}
2531 		m_adv = NULL;
2532 		if (m->psind != 0 && addr + pagesizes[1] <= entry->end &&
2533 		    (addr & (pagesizes[1] - 1)) == 0 &&
2534 		    (pmap_mincore(map->pmap, addr, &pa) & MINCORE_SUPER) != 0) {
2535 			*super = true;
2536 			pi_adv = atop(pagesizes[1]);
2537 		} else {
2538 			/*
2539 			 * We do not test the found page on validity.
2540 			 * Either the page is busy and being paged in,
2541 			 * or it was invalidated.  The first case
2542 			 * should be counted as resident, the second
2543 			 * is not so clear; we do account both.
2544 			 */
2545 			pi_adv = 1;
2546 		}
2547 		*resident_count += pi_adv;
2548 next:;
2549 	}
2550 }
2551 
2552 /*
2553  * Must be called with the process locked and will return unlocked.
2554  */
2555 int
2556 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb, ssize_t maxlen, int flags)
2557 {
2558 	vm_map_entry_t entry, tmp_entry;
2559 	struct vattr va;
2560 	vm_map_t map;
2561 	vm_object_t obj, tobj, lobj;
2562 	char *fullpath, *freepath;
2563 	struct kinfo_vmentry *kve;
2564 	struct ucred *cred;
2565 	struct vnode *vp;
2566 	struct vmspace *vm;
2567 	vm_offset_t addr;
2568 	unsigned int last_timestamp;
2569 	int error;
2570 	bool super;
2571 
2572 	PROC_LOCK_ASSERT(p, MA_OWNED);
2573 
2574 	_PHOLD(p);
2575 	PROC_UNLOCK(p);
2576 	vm = vmspace_acquire_ref(p);
2577 	if (vm == NULL) {
2578 		PRELE(p);
2579 		return (ESRCH);
2580 	}
2581 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK | M_ZERO);
2582 
2583 	error = 0;
2584 	map = &vm->vm_map;
2585 	vm_map_lock_read(map);
2586 	VM_MAP_ENTRY_FOREACH(entry, map) {
2587 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2588 			continue;
2589 
2590 		addr = entry->end;
2591 		bzero(kve, sizeof(*kve));
2592 		obj = entry->object.vm_object;
2593 		if (obj != NULL) {
2594 			for (tobj = obj; tobj != NULL;
2595 			    tobj = tobj->backing_object) {
2596 				VM_OBJECT_RLOCK(tobj);
2597 				kve->kve_offset += tobj->backing_object_offset;
2598 				lobj = tobj;
2599 			}
2600 			if (obj->backing_object == NULL)
2601 				kve->kve_private_resident =
2602 				    obj->resident_page_count;
2603 			kern_proc_vmmap_resident(map, entry,
2604 			    &kve->kve_resident, &super);
2605 			if (super)
2606 				kve->kve_flags |= KVME_FLAG_SUPER;
2607 			for (tobj = obj; tobj != NULL;
2608 			    tobj = tobj->backing_object) {
2609 				if (tobj != obj && tobj != lobj)
2610 					VM_OBJECT_RUNLOCK(tobj);
2611 			}
2612 		} else {
2613 			lobj = NULL;
2614 		}
2615 
2616 		kve->kve_start = entry->start;
2617 		kve->kve_end = entry->end;
2618 		kve->kve_offset += entry->offset;
2619 
2620 		if (entry->protection & VM_PROT_READ)
2621 			kve->kve_protection |= KVME_PROT_READ;
2622 		if (entry->protection & VM_PROT_WRITE)
2623 			kve->kve_protection |= KVME_PROT_WRITE;
2624 		if (entry->protection & VM_PROT_EXECUTE)
2625 			kve->kve_protection |= KVME_PROT_EXEC;
2626 
2627 		if (entry->eflags & MAP_ENTRY_COW)
2628 			kve->kve_flags |= KVME_FLAG_COW;
2629 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2630 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2631 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2632 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2633 		if (entry->eflags & MAP_ENTRY_GROWS_UP)
2634 			kve->kve_flags |= KVME_FLAG_GROWS_UP;
2635 		if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2636 			kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2637 		if (entry->eflags & MAP_ENTRY_USER_WIRED)
2638 			kve->kve_flags |= KVME_FLAG_USER_WIRED;
2639 
2640 		last_timestamp = map->timestamp;
2641 		vm_map_unlock_read(map);
2642 
2643 		freepath = NULL;
2644 		fullpath = "";
2645 		if (lobj != NULL) {
2646 			kve->kve_type = vm_object_kvme_type(lobj, &vp);
2647 			if (vp != NULL)
2648 				vref(vp);
2649 			if (lobj != obj)
2650 				VM_OBJECT_RUNLOCK(lobj);
2651 
2652 			kve->kve_ref_count = obj->ref_count;
2653 			kve->kve_shadow_count = obj->shadow_count;
2654 			VM_OBJECT_RUNLOCK(obj);
2655 			if (vp != NULL) {
2656 				vn_fullpath(vp, &fullpath, &freepath);
2657 				kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2658 				cred = curthread->td_ucred;
2659 				vn_lock(vp, LK_SHARED | LK_RETRY);
2660 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2661 					kve->kve_vn_fileid = va.va_fileid;
2662 					kve->kve_vn_fsid = va.va_fsid;
2663 					kve->kve_vn_fsid_freebsd11 =
2664 					    kve->kve_vn_fsid; /* truncate */
2665 					kve->kve_vn_mode =
2666 					    MAKEIMODE(va.va_type, va.va_mode);
2667 					kve->kve_vn_size = va.va_size;
2668 					kve->kve_vn_rdev = va.va_rdev;
2669 					kve->kve_vn_rdev_freebsd11 =
2670 					    kve->kve_vn_rdev; /* truncate */
2671 					kve->kve_status = KF_ATTR_VALID;
2672 				}
2673 				vput(vp);
2674 			}
2675 		} else {
2676 			kve->kve_type = KVME_TYPE_NONE;
2677 			kve->kve_ref_count = 0;
2678 			kve->kve_shadow_count = 0;
2679 		}
2680 
2681 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2682 		if (freepath != NULL)
2683 			free(freepath, M_TEMP);
2684 
2685 		/* Pack record size down */
2686 		if ((flags & KERN_VMMAP_PACK_KINFO) != 0)
2687 			kve->kve_structsize =
2688 			    offsetof(struct kinfo_vmentry, kve_path) +
2689 			    strlen(kve->kve_path) + 1;
2690 		else
2691 			kve->kve_structsize = sizeof(*kve);
2692 		kve->kve_structsize = roundup(kve->kve_structsize,
2693 		    sizeof(uint64_t));
2694 
2695 		/* Halt filling and truncate rather than exceeding maxlen */
2696 		if (maxlen != -1 && maxlen < kve->kve_structsize) {
2697 			error = 0;
2698 			vm_map_lock_read(map);
2699 			break;
2700 		} else if (maxlen != -1)
2701 			maxlen -= kve->kve_structsize;
2702 
2703 		if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0)
2704 			error = ENOMEM;
2705 		vm_map_lock_read(map);
2706 		if (error != 0)
2707 			break;
2708 		if (last_timestamp != map->timestamp) {
2709 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2710 			entry = tmp_entry;
2711 		}
2712 	}
2713 	vm_map_unlock_read(map);
2714 	vmspace_free(vm);
2715 	PRELE(p);
2716 	free(kve, M_TEMP);
2717 	return (error);
2718 }
2719 
2720 static int
2721 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2722 {
2723 	struct proc *p;
2724 	struct sbuf sb;
2725 	int error, error2, *name;
2726 
2727 	name = (int *)arg1;
2728 	sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req);
2729 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2730 	error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
2731 	if (error != 0) {
2732 		sbuf_delete(&sb);
2733 		return (error);
2734 	}
2735 	error = kern_proc_vmmap_out(p, &sb, -1, KERN_VMMAP_PACK_KINFO);
2736 	error2 = sbuf_finish(&sb);
2737 	sbuf_delete(&sb);
2738 	return (error != 0 ? error : error2);
2739 }
2740 
2741 #if defined(STACK) || defined(DDB)
2742 static int
2743 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2744 {
2745 	struct kinfo_kstack *kkstp;
2746 	int error, i, *name, numthreads;
2747 	lwpid_t *lwpidarray;
2748 	struct thread *td;
2749 	struct stack *st;
2750 	struct sbuf sb;
2751 	struct proc *p;
2752 
2753 	name = (int *)arg1;
2754 	error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2755 	if (error != 0)
2756 		return (error);
2757 
2758 	kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2759 	st = stack_create(M_WAITOK);
2760 
2761 	lwpidarray = NULL;
2762 	PROC_LOCK(p);
2763 	do {
2764 		if (lwpidarray != NULL) {
2765 			free(lwpidarray, M_TEMP);
2766 			lwpidarray = NULL;
2767 		}
2768 		numthreads = p->p_numthreads;
2769 		PROC_UNLOCK(p);
2770 		lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2771 		    M_WAITOK | M_ZERO);
2772 		PROC_LOCK(p);
2773 	} while (numthreads < p->p_numthreads);
2774 
2775 	/*
2776 	 * XXXRW: During the below loop, execve(2) and countless other sorts
2777 	 * of changes could have taken place.  Should we check to see if the
2778 	 * vmspace has been replaced, or the like, in order to prevent
2779 	 * giving a snapshot that spans, say, execve(2), with some threads
2780 	 * before and some after?  Among other things, the credentials could
2781 	 * have changed, in which case the right to extract debug info might
2782 	 * no longer be assured.
2783 	 */
2784 	i = 0;
2785 	FOREACH_THREAD_IN_PROC(p, td) {
2786 		KASSERT(i < numthreads,
2787 		    ("sysctl_kern_proc_kstack: numthreads"));
2788 		lwpidarray[i] = td->td_tid;
2789 		i++;
2790 	}
2791 	PROC_UNLOCK(p);
2792 	numthreads = i;
2793 	for (i = 0; i < numthreads; i++) {
2794 		td = tdfind(lwpidarray[i], p->p_pid);
2795 		if (td == NULL) {
2796 			continue;
2797 		}
2798 		bzero(kkstp, sizeof(*kkstp));
2799 		(void)sbuf_new(&sb, kkstp->kkst_trace,
2800 		    sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2801 		thread_lock(td);
2802 		kkstp->kkst_tid = td->td_tid;
2803 		if (TD_IS_SWAPPED(td))
2804 			kkstp->kkst_state = KKST_STATE_SWAPPED;
2805 		else if (stack_save_td(st, td) == 0)
2806 			kkstp->kkst_state = KKST_STATE_STACKOK;
2807 		else
2808 			kkstp->kkst_state = KKST_STATE_RUNNING;
2809 		thread_unlock(td);
2810 		PROC_UNLOCK(p);
2811 		stack_sbuf_print(&sb, st);
2812 		sbuf_finish(&sb);
2813 		sbuf_delete(&sb);
2814 		error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2815 		if (error)
2816 			break;
2817 	}
2818 	PRELE(p);
2819 	if (lwpidarray != NULL)
2820 		free(lwpidarray, M_TEMP);
2821 	stack_destroy(st);
2822 	free(kkstp, M_TEMP);
2823 	return (error);
2824 }
2825 #endif
2826 
2827 /*
2828  * This sysctl allows a process to retrieve the full list of groups from
2829  * itself or another process.
2830  */
2831 static int
2832 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2833 {
2834 	pid_t *pidp = (pid_t *)arg1;
2835 	unsigned int arglen = arg2;
2836 	struct proc *p;
2837 	struct ucred *cred;
2838 	int error;
2839 
2840 	if (arglen != 1)
2841 		return (EINVAL);
2842 	if (*pidp == -1) {	/* -1 means this process */
2843 		p = req->td->td_proc;
2844 		PROC_LOCK(p);
2845 	} else {
2846 		error = pget(*pidp, PGET_CANSEE, &p);
2847 		if (error != 0)
2848 			return (error);
2849 	}
2850 
2851 	cred = crhold(p->p_ucred);
2852 	PROC_UNLOCK(p);
2853 
2854 	error = SYSCTL_OUT(req, cred->cr_groups,
2855 	    cred->cr_ngroups * sizeof(gid_t));
2856 	crfree(cred);
2857 	return (error);
2858 }
2859 
2860 /*
2861  * This sysctl allows a process to retrieve or/and set the resource limit for
2862  * another process.
2863  */
2864 static int
2865 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2866 {
2867 	int *name = (int *)arg1;
2868 	u_int namelen = arg2;
2869 	struct rlimit rlim;
2870 	struct proc *p;
2871 	u_int which;
2872 	int flags, error;
2873 
2874 	if (namelen != 2)
2875 		return (EINVAL);
2876 
2877 	which = (u_int)name[1];
2878 	if (which >= RLIM_NLIMITS)
2879 		return (EINVAL);
2880 
2881 	if (req->newptr != NULL && req->newlen != sizeof(rlim))
2882 		return (EINVAL);
2883 
2884 	flags = PGET_HOLD | PGET_NOTWEXIT;
2885 	if (req->newptr != NULL)
2886 		flags |= PGET_CANDEBUG;
2887 	else
2888 		flags |= PGET_CANSEE;
2889 	error = pget((pid_t)name[0], flags, &p);
2890 	if (error != 0)
2891 		return (error);
2892 
2893 	/*
2894 	 * Retrieve limit.
2895 	 */
2896 	if (req->oldptr != NULL) {
2897 		PROC_LOCK(p);
2898 		lim_rlimit_proc(p, which, &rlim);
2899 		PROC_UNLOCK(p);
2900 	}
2901 	error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
2902 	if (error != 0)
2903 		goto errout;
2904 
2905 	/*
2906 	 * Set limit.
2907 	 */
2908 	if (req->newptr != NULL) {
2909 		error = SYSCTL_IN(req, &rlim, sizeof(rlim));
2910 		if (error == 0)
2911 			error = kern_proc_setrlimit(curthread, p, which, &rlim);
2912 	}
2913 
2914 errout:
2915 	PRELE(p);
2916 	return (error);
2917 }
2918 
2919 /*
2920  * This sysctl allows a process to retrieve ps_strings structure location of
2921  * another process.
2922  */
2923 static int
2924 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
2925 {
2926 	int *name = (int *)arg1;
2927 	u_int namelen = arg2;
2928 	struct proc *p;
2929 	vm_offset_t ps_strings;
2930 	int error;
2931 #ifdef COMPAT_FREEBSD32
2932 	uint32_t ps_strings32;
2933 #endif
2934 
2935 	if (namelen != 1)
2936 		return (EINVAL);
2937 
2938 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2939 	if (error != 0)
2940 		return (error);
2941 #ifdef COMPAT_FREEBSD32
2942 	if ((req->flags & SCTL_MASK32) != 0) {
2943 		/*
2944 		 * We return 0 if the 32 bit emulation request is for a 64 bit
2945 		 * process.
2946 		 */
2947 		ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
2948 		    PTROUT(p->p_sysent->sv_psstrings) : 0;
2949 		PROC_UNLOCK(p);
2950 		error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
2951 		return (error);
2952 	}
2953 #endif
2954 	ps_strings = p->p_sysent->sv_psstrings;
2955 	PROC_UNLOCK(p);
2956 	error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
2957 	return (error);
2958 }
2959 
2960 /*
2961  * This sysctl allows a process to retrieve umask of another process.
2962  */
2963 static int
2964 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
2965 {
2966 	int *name = (int *)arg1;
2967 	u_int namelen = arg2;
2968 	struct proc *p;
2969 	int error;
2970 	u_short fd_cmask;
2971 	pid_t pid;
2972 
2973 	if (namelen != 1)
2974 		return (EINVAL);
2975 
2976 	pid = (pid_t)name[0];
2977 	p = curproc;
2978 	if (pid == p->p_pid || pid == 0) {
2979 		fd_cmask = p->p_fd->fd_cmask;
2980 		goto out;
2981 	}
2982 
2983 	error = pget(pid, PGET_WANTREAD, &p);
2984 	if (error != 0)
2985 		return (error);
2986 
2987 	fd_cmask = p->p_fd->fd_cmask;
2988 	PRELE(p);
2989 out:
2990 	error = SYSCTL_OUT(req, &fd_cmask, sizeof(fd_cmask));
2991 	return (error);
2992 }
2993 
2994 /*
2995  * This sysctl allows a process to set and retrieve binary osreldate of
2996  * another process.
2997  */
2998 static int
2999 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
3000 {
3001 	int *name = (int *)arg1;
3002 	u_int namelen = arg2;
3003 	struct proc *p;
3004 	int flags, error, osrel;
3005 
3006 	if (namelen != 1)
3007 		return (EINVAL);
3008 
3009 	if (req->newptr != NULL && req->newlen != sizeof(osrel))
3010 		return (EINVAL);
3011 
3012 	flags = PGET_HOLD | PGET_NOTWEXIT;
3013 	if (req->newptr != NULL)
3014 		flags |= PGET_CANDEBUG;
3015 	else
3016 		flags |= PGET_CANSEE;
3017 	error = pget((pid_t)name[0], flags, &p);
3018 	if (error != 0)
3019 		return (error);
3020 
3021 	error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
3022 	if (error != 0)
3023 		goto errout;
3024 
3025 	if (req->newptr != NULL) {
3026 		error = SYSCTL_IN(req, &osrel, sizeof(osrel));
3027 		if (error != 0)
3028 			goto errout;
3029 		if (osrel < 0) {
3030 			error = EINVAL;
3031 			goto errout;
3032 		}
3033 		p->p_osrel = osrel;
3034 	}
3035 errout:
3036 	PRELE(p);
3037 	return (error);
3038 }
3039 
3040 static int
3041 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS)
3042 {
3043 	int *name = (int *)arg1;
3044 	u_int namelen = arg2;
3045 	struct proc *p;
3046 	struct kinfo_sigtramp kst;
3047 	const struct sysentvec *sv;
3048 	int error;
3049 #ifdef COMPAT_FREEBSD32
3050 	struct kinfo_sigtramp32 kst32;
3051 #endif
3052 
3053 	if (namelen != 1)
3054 		return (EINVAL);
3055 
3056 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
3057 	if (error != 0)
3058 		return (error);
3059 	sv = p->p_sysent;
3060 #ifdef COMPAT_FREEBSD32
3061 	if ((req->flags & SCTL_MASK32) != 0) {
3062 		bzero(&kst32, sizeof(kst32));
3063 		if (SV_PROC_FLAG(p, SV_ILP32)) {
3064 			if (sv->sv_sigcode_base != 0) {
3065 				kst32.ksigtramp_start = sv->sv_sigcode_base;
3066 				kst32.ksigtramp_end = sv->sv_sigcode_base +
3067 				    *sv->sv_szsigcode;
3068 			} else {
3069 				kst32.ksigtramp_start = sv->sv_psstrings -
3070 				    *sv->sv_szsigcode;
3071 				kst32.ksigtramp_end = sv->sv_psstrings;
3072 			}
3073 		}
3074 		PROC_UNLOCK(p);
3075 		error = SYSCTL_OUT(req, &kst32, sizeof(kst32));
3076 		return (error);
3077 	}
3078 #endif
3079 	bzero(&kst, sizeof(kst));
3080 	if (sv->sv_sigcode_base != 0) {
3081 		kst.ksigtramp_start = (char *)sv->sv_sigcode_base;
3082 		kst.ksigtramp_end = (char *)sv->sv_sigcode_base +
3083 		    *sv->sv_szsigcode;
3084 	} else {
3085 		kst.ksigtramp_start = (char *)sv->sv_psstrings -
3086 		    *sv->sv_szsigcode;
3087 		kst.ksigtramp_end = (char *)sv->sv_psstrings;
3088 	}
3089 	PROC_UNLOCK(p);
3090 	error = SYSCTL_OUT(req, &kst, sizeof(kst));
3091 	return (error);
3092 }
3093 
3094 static int
3095 sysctl_kern_proc_sigfastblk(SYSCTL_HANDLER_ARGS)
3096 {
3097 	int *name = (int *)arg1;
3098 	u_int namelen = arg2;
3099 	pid_t pid;
3100 	struct proc *p;
3101 	struct thread *td1;
3102 	uintptr_t addr;
3103 #ifdef COMPAT_FREEBSD32
3104 	uint32_t addr32;
3105 #endif
3106 	int error;
3107 
3108 	if (namelen != 1 || req->newptr != NULL)
3109 		return (EINVAL);
3110 
3111 	pid = (pid_t)name[0];
3112 	error = pget(pid, PGET_HOLD | PGET_NOTWEXIT | PGET_CANDEBUG, &p);
3113 	if (error != 0)
3114 		return (error);
3115 
3116 	PROC_LOCK(p);
3117 #ifdef COMPAT_FREEBSD32
3118 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3119 		if (!SV_PROC_FLAG(p, SV_ILP32)) {
3120 			error = EINVAL;
3121 			goto errlocked;
3122 		}
3123 	}
3124 #endif
3125 	if (pid <= PID_MAX) {
3126 		td1 = FIRST_THREAD_IN_PROC(p);
3127 	} else {
3128 		FOREACH_THREAD_IN_PROC(p, td1) {
3129 			if (td1->td_tid == pid)
3130 				break;
3131 		}
3132 	}
3133 	if (td1 == NULL) {
3134 		error = ESRCH;
3135 		goto errlocked;
3136 	}
3137 	/*
3138 	 * The access to the private thread flags.  It is fine as far
3139 	 * as no out-of-thin-air values are read from td_pflags, and
3140 	 * usermode read of the td_sigblock_ptr is racy inherently,
3141 	 * since target process might have already changed it
3142 	 * meantime.
3143 	 */
3144 	if ((td1->td_pflags & TDP_SIGFASTBLOCK) != 0)
3145 		addr = (uintptr_t)td1->td_sigblock_ptr;
3146 	else
3147 		error = ENOTTY;
3148 
3149 errlocked:
3150 	_PRELE(p);
3151 	PROC_UNLOCK(p);
3152 	if (error != 0)
3153 		return (error);
3154 
3155 #ifdef COMPAT_FREEBSD32
3156 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3157 		addr32 = addr;
3158 		error = SYSCTL_OUT(req, &addr32, sizeof(addr32));
3159 	} else
3160 #endif
3161 		error = SYSCTL_OUT(req, &addr, sizeof(addr));
3162 	return (error);
3163 }
3164 
3165 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,  0,
3166     "Process table");
3167 
3168 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
3169 	CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
3170 	"Return entire process table");
3171 
3172 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3173 	sysctl_kern_proc, "Process table");
3174 
3175 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
3176 	sysctl_kern_proc, "Process table");
3177 
3178 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3179 	sysctl_kern_proc, "Process table");
3180 
3181 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
3182 	CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3183 
3184 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
3185 	sysctl_kern_proc, "Process table");
3186 
3187 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3188 	sysctl_kern_proc, "Process table");
3189 
3190 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3191 	sysctl_kern_proc, "Process table");
3192 
3193 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3194 	sysctl_kern_proc, "Process table");
3195 
3196 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
3197 	sysctl_kern_proc, "Return process table, no threads");
3198 
3199 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
3200 	CTLFLAG_RW | CTLFLAG_CAPWR | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
3201 	sysctl_kern_proc_args, "Process argument list");
3202 
3203 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
3204 	sysctl_kern_proc_env, "Process environment");
3205 
3206 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
3207 	CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
3208 
3209 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
3210 	CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
3211 
3212 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
3213 	CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
3214 	"Process syscall vector name (ABI type)");
3215 
3216 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
3217 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3218 
3219 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
3220 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3221 
3222 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
3223 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3224 
3225 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
3226 	sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3227 
3228 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
3229 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3230 
3231 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
3232 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3233 
3234 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
3235 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3236 
3237 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
3238 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3239 
3240 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
3241 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
3242 	"Return process table, no threads");
3243 
3244 #ifdef COMPAT_FREEBSD7
3245 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
3246 	CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
3247 #endif
3248 
3249 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
3250 	CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
3251 
3252 #if defined(STACK) || defined(DDB)
3253 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
3254 	CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
3255 #endif
3256 
3257 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
3258 	CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
3259 
3260 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
3261 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
3262 	"Process resource limits");
3263 
3264 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
3265 	CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
3266 	"Process ps_strings location");
3267 
3268 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
3269 	CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
3270 
3271 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
3272 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
3273 	"Process binary osreldate");
3274 
3275 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD |
3276 	CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp,
3277 	"Process signal trampoline location");
3278 
3279 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGFASTBLK, sigfastblk, CTLFLAG_RD |
3280 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_sigfastblk,
3281 	"Thread sigfastblock address");
3282 
3283 int allproc_gen;
3284 
3285 /*
3286  * stop_all_proc() purpose is to stop all process which have usermode,
3287  * except current process for obvious reasons.  This makes it somewhat
3288  * unreliable when invoked from multithreaded process.  The service
3289  * must not be user-callable anyway.
3290  */
3291 void
3292 stop_all_proc(void)
3293 {
3294 	struct proc *cp, *p;
3295 	int r, gen;
3296 	bool restart, seen_stopped, seen_exiting, stopped_some;
3297 
3298 	cp = curproc;
3299 allproc_loop:
3300 	sx_xlock(&allproc_lock);
3301 	gen = allproc_gen;
3302 	seen_exiting = seen_stopped = stopped_some = restart = false;
3303 	LIST_REMOVE(cp, p_list);
3304 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3305 	for (;;) {
3306 		p = LIST_NEXT(cp, p_list);
3307 		if (p == NULL)
3308 			break;
3309 		LIST_REMOVE(cp, p_list);
3310 		LIST_INSERT_AFTER(p, cp, p_list);
3311 		PROC_LOCK(p);
3312 		if ((p->p_flag & (P_KPROC | P_SYSTEM | P_TOTAL_STOP)) != 0) {
3313 			PROC_UNLOCK(p);
3314 			continue;
3315 		}
3316 		if ((p->p_flag & P_WEXIT) != 0) {
3317 			seen_exiting = true;
3318 			PROC_UNLOCK(p);
3319 			continue;
3320 		}
3321 		if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) {
3322 			/*
3323 			 * Stopped processes are tolerated when there
3324 			 * are no other processes which might continue
3325 			 * them.  P_STOPPED_SINGLE but not
3326 			 * P_TOTAL_STOP process still has at least one
3327 			 * thread running.
3328 			 */
3329 			seen_stopped = true;
3330 			PROC_UNLOCK(p);
3331 			continue;
3332 		}
3333 		sx_xunlock(&allproc_lock);
3334 		_PHOLD(p);
3335 		r = thread_single(p, SINGLE_ALLPROC);
3336 		if (r != 0)
3337 			restart = true;
3338 		else
3339 			stopped_some = true;
3340 		_PRELE(p);
3341 		PROC_UNLOCK(p);
3342 		sx_xlock(&allproc_lock);
3343 	}
3344 	/* Catch forked children we did not see in iteration. */
3345 	if (gen != allproc_gen)
3346 		restart = true;
3347 	sx_xunlock(&allproc_lock);
3348 	if (restart || stopped_some || seen_exiting || seen_stopped) {
3349 		kern_yield(PRI_USER);
3350 		goto allproc_loop;
3351 	}
3352 }
3353 
3354 void
3355 resume_all_proc(void)
3356 {
3357 	struct proc *cp, *p;
3358 
3359 	cp = curproc;
3360 	sx_xlock(&allproc_lock);
3361 again:
3362 	LIST_REMOVE(cp, p_list);
3363 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3364 	for (;;) {
3365 		p = LIST_NEXT(cp, p_list);
3366 		if (p == NULL)
3367 			break;
3368 		LIST_REMOVE(cp, p_list);
3369 		LIST_INSERT_AFTER(p, cp, p_list);
3370 		PROC_LOCK(p);
3371 		if ((p->p_flag & P_TOTAL_STOP) != 0) {
3372 			sx_xunlock(&allproc_lock);
3373 			_PHOLD(p);
3374 			thread_single_end(p, SINGLE_ALLPROC);
3375 			_PRELE(p);
3376 			PROC_UNLOCK(p);
3377 			sx_xlock(&allproc_lock);
3378 		} else {
3379 			PROC_UNLOCK(p);
3380 		}
3381 	}
3382 	/*  Did the loop above missed any stopped process ? */
3383 	FOREACH_PROC_IN_SYSTEM(p) {
3384 		/* No need for proc lock. */
3385 		if ((p->p_flag & P_TOTAL_STOP) != 0)
3386 			goto again;
3387 	}
3388 	sx_xunlock(&allproc_lock);
3389 }
3390 
3391 /* #define	TOTAL_STOP_DEBUG	1 */
3392 #ifdef TOTAL_STOP_DEBUG
3393 volatile static int ap_resume;
3394 #include <sys/mount.h>
3395 
3396 static int
3397 sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS)
3398 {
3399 	int error, val;
3400 
3401 	val = 0;
3402 	ap_resume = 0;
3403 	error = sysctl_handle_int(oidp, &val, 0, req);
3404 	if (error != 0 || req->newptr == NULL)
3405 		return (error);
3406 	if (val != 0) {
3407 		stop_all_proc();
3408 		syncer_suspend();
3409 		while (ap_resume == 0)
3410 			;
3411 		syncer_resume();
3412 		resume_all_proc();
3413 	}
3414 	return (0);
3415 }
3416 
3417 SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW |
3418     CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0,
3419     sysctl_debug_stop_all_proc, "I",
3420     "");
3421 #endif
3422