xref: /freebsd/sys/kern/kern_proc.c (revision 953111c9c536daf5304062914c0083cea5efb46b)
1 /*-
2  * Copyright (c) 1982, 1986, 1989, 1991, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	@(#)kern_proc.c	8.7 (Berkeley) 2/14/95
30  */
31 
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
34 
35 #include "opt_compat.h"
36 #include "opt_ddb.h"
37 #include "opt_ktrace.h"
38 #include "opt_kstack_pages.h"
39 #include "opt_stack.h"
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/elf.h>
44 #include <sys/eventhandler.h>
45 #include <sys/exec.h>
46 #include <sys/jail.h>
47 #include <sys/kernel.h>
48 #include <sys/limits.h>
49 #include <sys/lock.h>
50 #include <sys/loginclass.h>
51 #include <sys/malloc.h>
52 #include <sys/mman.h>
53 #include <sys/mount.h>
54 #include <sys/mutex.h>
55 #include <sys/proc.h>
56 #include <sys/ptrace.h>
57 #include <sys/refcount.h>
58 #include <sys/resourcevar.h>
59 #include <sys/rwlock.h>
60 #include <sys/sbuf.h>
61 #include <sys/sysent.h>
62 #include <sys/sched.h>
63 #include <sys/smp.h>
64 #include <sys/stack.h>
65 #include <sys/stat.h>
66 #include <sys/sysctl.h>
67 #include <sys/filedesc.h>
68 #include <sys/tty.h>
69 #include <sys/signalvar.h>
70 #include <sys/sdt.h>
71 #include <sys/sx.h>
72 #include <sys/user.h>
73 #include <sys/vnode.h>
74 #include <sys/wait.h>
75 
76 #ifdef DDB
77 #include <ddb/ddb.h>
78 #endif
79 
80 #include <vm/vm.h>
81 #include <vm/vm_param.h>
82 #include <vm/vm_extern.h>
83 #include <vm/pmap.h>
84 #include <vm/vm_map.h>
85 #include <vm/vm_object.h>
86 #include <vm/vm_page.h>
87 #include <vm/uma.h>
88 
89 #ifdef COMPAT_FREEBSD32
90 #include <compat/freebsd32/freebsd32.h>
91 #include <compat/freebsd32/freebsd32_util.h>
92 #endif
93 
94 SDT_PROVIDER_DEFINE(proc);
95 SDT_PROBE_DEFINE4(proc, , ctor, entry, "struct proc *", "int", "void *",
96     "int");
97 SDT_PROBE_DEFINE4(proc, , ctor, return, "struct proc *", "int", "void *",
98     "int");
99 SDT_PROBE_DEFINE4(proc, , dtor, entry, "struct proc *", "int", "void *",
100     "struct thread *");
101 SDT_PROBE_DEFINE3(proc, , dtor, return, "struct proc *", "int", "void *");
102 SDT_PROBE_DEFINE3(proc, , init, entry, "struct proc *", "int", "int");
103 SDT_PROBE_DEFINE3(proc, , init, return, "struct proc *", "int", "int");
104 
105 MALLOC_DEFINE(M_PGRP, "pgrp", "process group header");
106 MALLOC_DEFINE(M_SESSION, "session", "session header");
107 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
108 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
109 
110 static void doenterpgrp(struct proc *, struct pgrp *);
111 static void orphanpg(struct pgrp *pg);
112 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp);
113 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp);
114 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp,
115     int preferthread);
116 static void pgadjustjobc(struct pgrp *pgrp, int entering);
117 static void pgdelete(struct pgrp *);
118 static int proc_ctor(void *mem, int size, void *arg, int flags);
119 static void proc_dtor(void *mem, int size, void *arg);
120 static int proc_init(void *mem, int size, int flags);
121 static void proc_fini(void *mem, int size);
122 static void pargs_free(struct pargs *pa);
123 static struct proc *zpfind_locked(pid_t pid);
124 
125 /*
126  * Other process lists
127  */
128 struct pidhashhead *pidhashtbl;
129 u_long pidhash;
130 struct pgrphashhead *pgrphashtbl;
131 u_long pgrphash;
132 struct proclist allproc;
133 struct proclist zombproc;
134 struct sx allproc_lock;
135 struct sx proctree_lock;
136 struct mtx ppeers_lock;
137 uma_zone_t proc_zone;
138 
139 /*
140  * The offset of various fields in struct proc and struct thread.
141  * These are used by kernel debuggers to enumerate kernel threads and
142  * processes.
143  */
144 const int proc_off_p_pid = offsetof(struct proc, p_pid);
145 const int proc_off_p_comm = offsetof(struct proc, p_comm);
146 const int proc_off_p_list = offsetof(struct proc, p_list);
147 const int proc_off_p_threads = offsetof(struct proc, p_threads);
148 const int thread_off_td_tid = offsetof(struct thread, td_tid);
149 const int thread_off_td_name = offsetof(struct thread, td_name);
150 const int thread_off_td_oncpu = offsetof(struct thread, td_oncpu);
151 const int thread_off_td_pcb = offsetof(struct thread, td_pcb);
152 const int thread_off_td_plist = offsetof(struct thread, td_plist);
153 
154 int kstack_pages = KSTACK_PAGES;
155 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0,
156     "Kernel stack size in pages");
157 static int vmmap_skip_res_cnt = 0;
158 SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW,
159     &vmmap_skip_res_cnt, 0,
160     "Skip calculation of the pages resident count in kern.proc.vmmap");
161 
162 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
163 #ifdef COMPAT_FREEBSD32
164 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
165 #endif
166 
167 /*
168  * Initialize global process hashing structures.
169  */
170 void
171 procinit(void)
172 {
173 
174 	sx_init(&allproc_lock, "allproc");
175 	sx_init(&proctree_lock, "proctree");
176 	mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
177 	LIST_INIT(&allproc);
178 	LIST_INIT(&zombproc);
179 	pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
180 	pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
181 	proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
182 	    proc_ctor, proc_dtor, proc_init, proc_fini,
183 	    UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
184 	uihashinit();
185 }
186 
187 /*
188  * Prepare a proc for use.
189  */
190 static int
191 proc_ctor(void *mem, int size, void *arg, int flags)
192 {
193 	struct proc *p;
194 	struct thread *td;
195 
196 	p = (struct proc *)mem;
197 	SDT_PROBE4(proc, , ctor , entry, p, size, arg, flags);
198 	EVENTHANDLER_INVOKE(process_ctor, p);
199 	SDT_PROBE4(proc, , ctor , return, p, size, arg, flags);
200 	td = FIRST_THREAD_IN_PROC(p);
201 	if (td != NULL) {
202 		/* Make sure all thread constructors are executed */
203 		EVENTHANDLER_INVOKE(thread_ctor, td);
204 	}
205 	return (0);
206 }
207 
208 /*
209  * Reclaim a proc after use.
210  */
211 static void
212 proc_dtor(void *mem, int size, void *arg)
213 {
214 	struct proc *p;
215 	struct thread *td;
216 
217 	/* INVARIANTS checks go here */
218 	p = (struct proc *)mem;
219 	td = FIRST_THREAD_IN_PROC(p);
220 	SDT_PROBE4(proc, , dtor, entry, p, size, arg, td);
221 	if (td != NULL) {
222 #ifdef INVARIANTS
223 		KASSERT((p->p_numthreads == 1),
224 		    ("bad number of threads in exiting process"));
225 		KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
226 #endif
227 		/* Free all OSD associated to this thread. */
228 		osd_thread_exit(td);
229 
230 		/* Make sure all thread destructors are executed */
231 		EVENTHANDLER_INVOKE(thread_dtor, td);
232 	}
233 	EVENTHANDLER_INVOKE(process_dtor, p);
234 	if (p->p_ksi != NULL)
235 		KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
236 	SDT_PROBE3(proc, , dtor, return, p, size, arg);
237 }
238 
239 /*
240  * Initialize type-stable parts of a proc (when newly created).
241  */
242 static int
243 proc_init(void *mem, int size, int flags)
244 {
245 	struct proc *p;
246 
247 	p = (struct proc *)mem;
248 	SDT_PROBE3(proc, , init, entry, p, size, flags);
249 	mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK | MTX_NEW);
250 	mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_NEW);
251 	mtx_init(&p->p_statmtx, "pstatl", NULL, MTX_SPIN | MTX_NEW);
252 	mtx_init(&p->p_itimmtx, "pitiml", NULL, MTX_SPIN | MTX_NEW);
253 	mtx_init(&p->p_profmtx, "pprofl", NULL, MTX_SPIN | MTX_NEW);
254 	cv_init(&p->p_pwait, "ppwait");
255 	cv_init(&p->p_dbgwait, "dbgwait");
256 	TAILQ_INIT(&p->p_threads);	     /* all threads in proc */
257 	EVENTHANDLER_INVOKE(process_init, p);
258 	p->p_stats = pstats_alloc();
259 	p->p_pgrp = NULL;
260 	SDT_PROBE3(proc, , init, return, p, size, flags);
261 	return (0);
262 }
263 
264 /*
265  * UMA should ensure that this function is never called.
266  * Freeing a proc structure would violate type stability.
267  */
268 static void
269 proc_fini(void *mem, int size)
270 {
271 #ifdef notnow
272 	struct proc *p;
273 
274 	p = (struct proc *)mem;
275 	EVENTHANDLER_INVOKE(process_fini, p);
276 	pstats_free(p->p_stats);
277 	thread_free(FIRST_THREAD_IN_PROC(p));
278 	mtx_destroy(&p->p_mtx);
279 	if (p->p_ksi != NULL)
280 		ksiginfo_free(p->p_ksi);
281 #else
282 	panic("proc reclaimed");
283 #endif
284 }
285 
286 /*
287  * Is p an inferior of the current process?
288  */
289 int
290 inferior(struct proc *p)
291 {
292 
293 	sx_assert(&proctree_lock, SX_LOCKED);
294 	PROC_LOCK_ASSERT(p, MA_OWNED);
295 	for (; p != curproc; p = proc_realparent(p)) {
296 		if (p->p_pid == 0)
297 			return (0);
298 	}
299 	return (1);
300 }
301 
302 struct proc *
303 pfind_locked(pid_t pid)
304 {
305 	struct proc *p;
306 
307 	sx_assert(&allproc_lock, SX_LOCKED);
308 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
309 		if (p->p_pid == pid) {
310 			PROC_LOCK(p);
311 			if (p->p_state == PRS_NEW) {
312 				PROC_UNLOCK(p);
313 				p = NULL;
314 			}
315 			break;
316 		}
317 	}
318 	return (p);
319 }
320 
321 /*
322  * Locate a process by number; return only "live" processes -- i.e., neither
323  * zombies nor newly born but incompletely initialized processes.  By not
324  * returning processes in the PRS_NEW state, we allow callers to avoid
325  * testing for that condition to avoid dereferencing p_ucred, et al.
326  */
327 struct proc *
328 pfind(pid_t pid)
329 {
330 	struct proc *p;
331 
332 	sx_slock(&allproc_lock);
333 	p = pfind_locked(pid);
334 	sx_sunlock(&allproc_lock);
335 	return (p);
336 }
337 
338 static struct proc *
339 pfind_tid_locked(pid_t tid)
340 {
341 	struct proc *p;
342 	struct thread *td;
343 
344 	sx_assert(&allproc_lock, SX_LOCKED);
345 	FOREACH_PROC_IN_SYSTEM(p) {
346 		PROC_LOCK(p);
347 		if (p->p_state == PRS_NEW) {
348 			PROC_UNLOCK(p);
349 			continue;
350 		}
351 		FOREACH_THREAD_IN_PROC(p, td) {
352 			if (td->td_tid == tid)
353 				goto found;
354 		}
355 		PROC_UNLOCK(p);
356 	}
357 found:
358 	return (p);
359 }
360 
361 /*
362  * Locate a process group by number.
363  * The caller must hold proctree_lock.
364  */
365 struct pgrp *
366 pgfind(pid_t pgid)
367 {
368 	struct pgrp *pgrp;
369 
370 	sx_assert(&proctree_lock, SX_LOCKED);
371 
372 	LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
373 		if (pgrp->pg_id == pgid) {
374 			PGRP_LOCK(pgrp);
375 			return (pgrp);
376 		}
377 	}
378 	return (NULL);
379 }
380 
381 /*
382  * Locate process and do additional manipulations, depending on flags.
383  */
384 int
385 pget(pid_t pid, int flags, struct proc **pp)
386 {
387 	struct proc *p;
388 	int error;
389 
390 	sx_slock(&allproc_lock);
391 	if (pid <= PID_MAX) {
392 		p = pfind_locked(pid);
393 		if (p == NULL && (flags & PGET_NOTWEXIT) == 0)
394 			p = zpfind_locked(pid);
395 	} else if ((flags & PGET_NOTID) == 0) {
396 		p = pfind_tid_locked(pid);
397 	} else {
398 		p = NULL;
399 	}
400 	sx_sunlock(&allproc_lock);
401 	if (p == NULL)
402 		return (ESRCH);
403 	if ((flags & PGET_CANSEE) != 0) {
404 		error = p_cansee(curthread, p);
405 		if (error != 0)
406 			goto errout;
407 	}
408 	if ((flags & PGET_CANDEBUG) != 0) {
409 		error = p_candebug(curthread, p);
410 		if (error != 0)
411 			goto errout;
412 	}
413 	if ((flags & PGET_ISCURRENT) != 0 && curproc != p) {
414 		error = EPERM;
415 		goto errout;
416 	}
417 	if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) {
418 		error = ESRCH;
419 		goto errout;
420 	}
421 	if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) {
422 		/*
423 		 * XXXRW: Not clear ESRCH is the right error during proc
424 		 * execve().
425 		 */
426 		error = ESRCH;
427 		goto errout;
428 	}
429 	if ((flags & PGET_HOLD) != 0) {
430 		_PHOLD(p);
431 		PROC_UNLOCK(p);
432 	}
433 	*pp = p;
434 	return (0);
435 errout:
436 	PROC_UNLOCK(p);
437 	return (error);
438 }
439 
440 /*
441  * Create a new process group.
442  * pgid must be equal to the pid of p.
443  * Begin a new session if required.
444  */
445 int
446 enterpgrp(struct proc *p, pid_t pgid, struct pgrp *pgrp, struct session *sess)
447 {
448 
449 	sx_assert(&proctree_lock, SX_XLOCKED);
450 
451 	KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
452 	KASSERT(p->p_pid == pgid,
453 	    ("enterpgrp: new pgrp and pid != pgid"));
454 	KASSERT(pgfind(pgid) == NULL,
455 	    ("enterpgrp: pgrp with pgid exists"));
456 	KASSERT(!SESS_LEADER(p),
457 	    ("enterpgrp: session leader attempted setpgrp"));
458 
459 	mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
460 
461 	if (sess != NULL) {
462 		/*
463 		 * new session
464 		 */
465 		mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
466 		PROC_LOCK(p);
467 		p->p_flag &= ~P_CONTROLT;
468 		PROC_UNLOCK(p);
469 		PGRP_LOCK(pgrp);
470 		sess->s_leader = p;
471 		sess->s_sid = p->p_pid;
472 		refcount_init(&sess->s_count, 1);
473 		sess->s_ttyvp = NULL;
474 		sess->s_ttydp = NULL;
475 		sess->s_ttyp = NULL;
476 		bcopy(p->p_session->s_login, sess->s_login,
477 			    sizeof(sess->s_login));
478 		pgrp->pg_session = sess;
479 		KASSERT(p == curproc,
480 		    ("enterpgrp: mksession and p != curproc"));
481 	} else {
482 		pgrp->pg_session = p->p_session;
483 		sess_hold(pgrp->pg_session);
484 		PGRP_LOCK(pgrp);
485 	}
486 	pgrp->pg_id = pgid;
487 	LIST_INIT(&pgrp->pg_members);
488 
489 	/*
490 	 * As we have an exclusive lock of proctree_lock,
491 	 * this should not deadlock.
492 	 */
493 	LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
494 	pgrp->pg_jobc = 0;
495 	SLIST_INIT(&pgrp->pg_sigiolst);
496 	PGRP_UNLOCK(pgrp);
497 
498 	doenterpgrp(p, pgrp);
499 
500 	return (0);
501 }
502 
503 /*
504  * Move p to an existing process group
505  */
506 int
507 enterthispgrp(struct proc *p, struct pgrp *pgrp)
508 {
509 
510 	sx_assert(&proctree_lock, SX_XLOCKED);
511 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
512 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
513 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
514 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
515 	KASSERT(pgrp->pg_session == p->p_session,
516 		("%s: pgrp's session %p, p->p_session %p.\n",
517 		__func__,
518 		pgrp->pg_session,
519 		p->p_session));
520 	KASSERT(pgrp != p->p_pgrp,
521 		("%s: p belongs to pgrp.", __func__));
522 
523 	doenterpgrp(p, pgrp);
524 
525 	return (0);
526 }
527 
528 /*
529  * Move p to a process group
530  */
531 static void
532 doenterpgrp(struct proc *p, struct pgrp *pgrp)
533 {
534 	struct pgrp *savepgrp;
535 
536 	sx_assert(&proctree_lock, SX_XLOCKED);
537 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
538 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
539 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
540 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
541 
542 	savepgrp = p->p_pgrp;
543 
544 	/*
545 	 * Adjust eligibility of affected pgrps to participate in job control.
546 	 * Increment eligibility counts before decrementing, otherwise we
547 	 * could reach 0 spuriously during the first call.
548 	 */
549 	fixjobc(p, pgrp, 1);
550 	fixjobc(p, p->p_pgrp, 0);
551 
552 	PGRP_LOCK(pgrp);
553 	PGRP_LOCK(savepgrp);
554 	PROC_LOCK(p);
555 	LIST_REMOVE(p, p_pglist);
556 	p->p_pgrp = pgrp;
557 	PROC_UNLOCK(p);
558 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
559 	PGRP_UNLOCK(savepgrp);
560 	PGRP_UNLOCK(pgrp);
561 	if (LIST_EMPTY(&savepgrp->pg_members))
562 		pgdelete(savepgrp);
563 }
564 
565 /*
566  * remove process from process group
567  */
568 int
569 leavepgrp(struct proc *p)
570 {
571 	struct pgrp *savepgrp;
572 
573 	sx_assert(&proctree_lock, SX_XLOCKED);
574 	savepgrp = p->p_pgrp;
575 	PGRP_LOCK(savepgrp);
576 	PROC_LOCK(p);
577 	LIST_REMOVE(p, p_pglist);
578 	p->p_pgrp = NULL;
579 	PROC_UNLOCK(p);
580 	PGRP_UNLOCK(savepgrp);
581 	if (LIST_EMPTY(&savepgrp->pg_members))
582 		pgdelete(savepgrp);
583 	return (0);
584 }
585 
586 /*
587  * delete a process group
588  */
589 static void
590 pgdelete(struct pgrp *pgrp)
591 {
592 	struct session *savesess;
593 	struct tty *tp;
594 
595 	sx_assert(&proctree_lock, SX_XLOCKED);
596 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
597 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
598 
599 	/*
600 	 * Reset any sigio structures pointing to us as a result of
601 	 * F_SETOWN with our pgid.
602 	 */
603 	funsetownlst(&pgrp->pg_sigiolst);
604 
605 	PGRP_LOCK(pgrp);
606 	tp = pgrp->pg_session->s_ttyp;
607 	LIST_REMOVE(pgrp, pg_hash);
608 	savesess = pgrp->pg_session;
609 	PGRP_UNLOCK(pgrp);
610 
611 	/* Remove the reference to the pgrp before deallocating it. */
612 	if (tp != NULL) {
613 		tty_lock(tp);
614 		tty_rel_pgrp(tp, pgrp);
615 	}
616 
617 	mtx_destroy(&pgrp->pg_mtx);
618 	free(pgrp, M_PGRP);
619 	sess_release(savesess);
620 }
621 
622 static void
623 pgadjustjobc(struct pgrp *pgrp, int entering)
624 {
625 
626 	PGRP_LOCK(pgrp);
627 	if (entering)
628 		pgrp->pg_jobc++;
629 	else {
630 		--pgrp->pg_jobc;
631 		if (pgrp->pg_jobc == 0)
632 			orphanpg(pgrp);
633 	}
634 	PGRP_UNLOCK(pgrp);
635 }
636 
637 /*
638  * Adjust pgrp jobc counters when specified process changes process group.
639  * We count the number of processes in each process group that "qualify"
640  * the group for terminal job control (those with a parent in a different
641  * process group of the same session).  If that count reaches zero, the
642  * process group becomes orphaned.  Check both the specified process'
643  * process group and that of its children.
644  * entering == 0 => p is leaving specified group.
645  * entering == 1 => p is entering specified group.
646  */
647 void
648 fixjobc(struct proc *p, struct pgrp *pgrp, int entering)
649 {
650 	struct pgrp *hispgrp;
651 	struct session *mysession;
652 	struct proc *q;
653 
654 	sx_assert(&proctree_lock, SX_LOCKED);
655 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
656 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
657 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
658 
659 	/*
660 	 * Check p's parent to see whether p qualifies its own process
661 	 * group; if so, adjust count for p's process group.
662 	 */
663 	mysession = pgrp->pg_session;
664 	if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
665 	    hispgrp->pg_session == mysession)
666 		pgadjustjobc(pgrp, entering);
667 
668 	/*
669 	 * Check this process' children to see whether they qualify
670 	 * their process groups; if so, adjust counts for children's
671 	 * process groups.
672 	 */
673 	LIST_FOREACH(q, &p->p_children, p_sibling) {
674 		hispgrp = q->p_pgrp;
675 		if (hispgrp == pgrp ||
676 		    hispgrp->pg_session != mysession)
677 			continue;
678 		if (q->p_state == PRS_ZOMBIE)
679 			continue;
680 		pgadjustjobc(hispgrp, entering);
681 	}
682 }
683 
684 void
685 killjobc(void)
686 {
687 	struct session *sp;
688 	struct tty *tp;
689 	struct proc *p;
690 	struct vnode *ttyvp;
691 
692 	p = curproc;
693 	MPASS(p->p_flag & P_WEXIT);
694 	/*
695 	 * Do a quick check to see if there is anything to do with the
696 	 * proctree_lock held. pgrp and LIST_EMPTY checks are for fixjobc().
697 	 */
698 	PROC_LOCK(p);
699 	if (!SESS_LEADER(p) &&
700 	    (p->p_pgrp == p->p_pptr->p_pgrp) &&
701 	    LIST_EMPTY(&p->p_children)) {
702 		PROC_UNLOCK(p);
703 		return;
704 	}
705 	PROC_UNLOCK(p);
706 
707 	sx_xlock(&proctree_lock);
708 	if (SESS_LEADER(p)) {
709 		sp = p->p_session;
710 
711 		/*
712 		 * s_ttyp is not zero'd; we use this to indicate that
713 		 * the session once had a controlling terminal. (for
714 		 * logging and informational purposes)
715 		 */
716 		SESS_LOCK(sp);
717 		ttyvp = sp->s_ttyvp;
718 		tp = sp->s_ttyp;
719 		sp->s_ttyvp = NULL;
720 		sp->s_ttydp = NULL;
721 		sp->s_leader = NULL;
722 		SESS_UNLOCK(sp);
723 
724 		/*
725 		 * Signal foreground pgrp and revoke access to
726 		 * controlling terminal if it has not been revoked
727 		 * already.
728 		 *
729 		 * Because the TTY may have been revoked in the mean
730 		 * time and could already have a new session associated
731 		 * with it, make sure we don't send a SIGHUP to a
732 		 * foreground process group that does not belong to this
733 		 * session.
734 		 */
735 
736 		if (tp != NULL) {
737 			tty_lock(tp);
738 			if (tp->t_session == sp)
739 				tty_signal_pgrp(tp, SIGHUP);
740 			tty_unlock(tp);
741 		}
742 
743 		if (ttyvp != NULL) {
744 			sx_xunlock(&proctree_lock);
745 			if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) {
746 				VOP_REVOKE(ttyvp, REVOKEALL);
747 				VOP_UNLOCK(ttyvp, 0);
748 			}
749 			vrele(ttyvp);
750 			sx_xlock(&proctree_lock);
751 		}
752 	}
753 	fixjobc(p, p->p_pgrp, 0);
754 	sx_xunlock(&proctree_lock);
755 }
756 
757 /*
758  * A process group has become orphaned;
759  * if there are any stopped processes in the group,
760  * hang-up all process in that group.
761  */
762 static void
763 orphanpg(struct pgrp *pg)
764 {
765 	struct proc *p;
766 
767 	PGRP_LOCK_ASSERT(pg, MA_OWNED);
768 
769 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
770 		PROC_LOCK(p);
771 		if (P_SHOULDSTOP(p) == P_STOPPED_SIG) {
772 			PROC_UNLOCK(p);
773 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
774 				PROC_LOCK(p);
775 				kern_psignal(p, SIGHUP);
776 				kern_psignal(p, SIGCONT);
777 				PROC_UNLOCK(p);
778 			}
779 			return;
780 		}
781 		PROC_UNLOCK(p);
782 	}
783 }
784 
785 void
786 sess_hold(struct session *s)
787 {
788 
789 	refcount_acquire(&s->s_count);
790 }
791 
792 void
793 sess_release(struct session *s)
794 {
795 
796 	if (refcount_release(&s->s_count)) {
797 		if (s->s_ttyp != NULL) {
798 			tty_lock(s->s_ttyp);
799 			tty_rel_sess(s->s_ttyp, s);
800 		}
801 		mtx_destroy(&s->s_mtx);
802 		free(s, M_SESSION);
803 	}
804 }
805 
806 #ifdef DDB
807 
808 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
809 {
810 	struct pgrp *pgrp;
811 	struct proc *p;
812 	int i;
813 
814 	for (i = 0; i <= pgrphash; i++) {
815 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
816 			printf("\tindx %d\n", i);
817 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
818 				printf(
819 			"\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
820 				    (void *)pgrp, (long)pgrp->pg_id,
821 				    (void *)pgrp->pg_session,
822 				    pgrp->pg_session->s_count,
823 				    (void *)LIST_FIRST(&pgrp->pg_members));
824 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
825 					printf("\t\tpid %ld addr %p pgrp %p\n",
826 					    (long)p->p_pid, (void *)p,
827 					    (void *)p->p_pgrp);
828 				}
829 			}
830 		}
831 	}
832 }
833 #endif /* DDB */
834 
835 /*
836  * Calculate the kinfo_proc members which contain process-wide
837  * informations.
838  * Must be called with the target process locked.
839  */
840 static void
841 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
842 {
843 	struct thread *td;
844 
845 	PROC_LOCK_ASSERT(p, MA_OWNED);
846 
847 	kp->ki_estcpu = 0;
848 	kp->ki_pctcpu = 0;
849 	FOREACH_THREAD_IN_PROC(p, td) {
850 		thread_lock(td);
851 		kp->ki_pctcpu += sched_pctcpu(td);
852 		kp->ki_estcpu += sched_estcpu(td);
853 		thread_unlock(td);
854 	}
855 }
856 
857 /*
858  * Clear kinfo_proc and fill in any information that is common
859  * to all threads in the process.
860  * Must be called with the target process locked.
861  */
862 static void
863 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
864 {
865 	struct thread *td0;
866 	struct tty *tp;
867 	struct session *sp;
868 	struct ucred *cred;
869 	struct sigacts *ps;
870 	struct timeval boottime;
871 
872 	/* For proc_realparent. */
873 	sx_assert(&proctree_lock, SX_LOCKED);
874 	PROC_LOCK_ASSERT(p, MA_OWNED);
875 	bzero(kp, sizeof(*kp));
876 
877 	kp->ki_structsize = sizeof(*kp);
878 	kp->ki_paddr = p;
879 	kp->ki_addr =/* p->p_addr; */0; /* XXX */
880 	kp->ki_args = p->p_args;
881 	kp->ki_textvp = p->p_textvp;
882 #ifdef KTRACE
883 	kp->ki_tracep = p->p_tracevp;
884 	kp->ki_traceflag = p->p_traceflag;
885 #endif
886 	kp->ki_fd = p->p_fd;
887 	kp->ki_vmspace = p->p_vmspace;
888 	kp->ki_flag = p->p_flag;
889 	kp->ki_flag2 = p->p_flag2;
890 	cred = p->p_ucred;
891 	if (cred) {
892 		kp->ki_uid = cred->cr_uid;
893 		kp->ki_ruid = cred->cr_ruid;
894 		kp->ki_svuid = cred->cr_svuid;
895 		kp->ki_cr_flags = 0;
896 		if (cred->cr_flags & CRED_FLAG_CAPMODE)
897 			kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
898 		/* XXX bde doesn't like KI_NGROUPS */
899 		if (cred->cr_ngroups > KI_NGROUPS) {
900 			kp->ki_ngroups = KI_NGROUPS;
901 			kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
902 		} else
903 			kp->ki_ngroups = cred->cr_ngroups;
904 		bcopy(cred->cr_groups, kp->ki_groups,
905 		    kp->ki_ngroups * sizeof(gid_t));
906 		kp->ki_rgid = cred->cr_rgid;
907 		kp->ki_svgid = cred->cr_svgid;
908 		/* If jailed(cred), emulate the old P_JAILED flag. */
909 		if (jailed(cred)) {
910 			kp->ki_flag |= P_JAILED;
911 			/* If inside the jail, use 0 as a jail ID. */
912 			if (cred->cr_prison != curthread->td_ucred->cr_prison)
913 				kp->ki_jid = cred->cr_prison->pr_id;
914 		}
915 		strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
916 		    sizeof(kp->ki_loginclass));
917 	}
918 	ps = p->p_sigacts;
919 	if (ps) {
920 		mtx_lock(&ps->ps_mtx);
921 		kp->ki_sigignore = ps->ps_sigignore;
922 		kp->ki_sigcatch = ps->ps_sigcatch;
923 		mtx_unlock(&ps->ps_mtx);
924 	}
925 	if (p->p_state != PRS_NEW &&
926 	    p->p_state != PRS_ZOMBIE &&
927 	    p->p_vmspace != NULL) {
928 		struct vmspace *vm = p->p_vmspace;
929 
930 		kp->ki_size = vm->vm_map.size;
931 		kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
932 		FOREACH_THREAD_IN_PROC(p, td0) {
933 			if (!TD_IS_SWAPPED(td0))
934 				kp->ki_rssize += td0->td_kstack_pages;
935 		}
936 		kp->ki_swrss = vm->vm_swrss;
937 		kp->ki_tsize = vm->vm_tsize;
938 		kp->ki_dsize = vm->vm_dsize;
939 		kp->ki_ssize = vm->vm_ssize;
940 	} else if (p->p_state == PRS_ZOMBIE)
941 		kp->ki_stat = SZOMB;
942 	if (kp->ki_flag & P_INMEM)
943 		kp->ki_sflag = PS_INMEM;
944 	else
945 		kp->ki_sflag = 0;
946 	/* Calculate legacy swtime as seconds since 'swtick'. */
947 	kp->ki_swtime = (ticks - p->p_swtick) / hz;
948 	kp->ki_pid = p->p_pid;
949 	kp->ki_nice = p->p_nice;
950 	kp->ki_fibnum = p->p_fibnum;
951 	kp->ki_start = p->p_stats->p_start;
952 	getboottime(&boottime);
953 	timevaladd(&kp->ki_start, &boottime);
954 	PROC_STATLOCK(p);
955 	rufetch(p, &kp->ki_rusage);
956 	kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
957 	calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
958 	PROC_STATUNLOCK(p);
959 	calccru(p, &kp->ki_childutime, &kp->ki_childstime);
960 	/* Some callers want child times in a single value. */
961 	kp->ki_childtime = kp->ki_childstime;
962 	timevaladd(&kp->ki_childtime, &kp->ki_childutime);
963 
964 	FOREACH_THREAD_IN_PROC(p, td0)
965 		kp->ki_cow += td0->td_cow;
966 
967 	tp = NULL;
968 	if (p->p_pgrp) {
969 		kp->ki_pgid = p->p_pgrp->pg_id;
970 		kp->ki_jobc = p->p_pgrp->pg_jobc;
971 		sp = p->p_pgrp->pg_session;
972 
973 		if (sp != NULL) {
974 			kp->ki_sid = sp->s_sid;
975 			SESS_LOCK(sp);
976 			strlcpy(kp->ki_login, sp->s_login,
977 			    sizeof(kp->ki_login));
978 			if (sp->s_ttyvp)
979 				kp->ki_kiflag |= KI_CTTY;
980 			if (SESS_LEADER(p))
981 				kp->ki_kiflag |= KI_SLEADER;
982 			/* XXX proctree_lock */
983 			tp = sp->s_ttyp;
984 			SESS_UNLOCK(sp);
985 		}
986 	}
987 	if ((p->p_flag & P_CONTROLT) && tp != NULL) {
988 		kp->ki_tdev = tty_udev(tp);
989 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
990 		kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
991 		if (tp->t_session)
992 			kp->ki_tsid = tp->t_session->s_sid;
993 	} else {
994 		kp->ki_tdev = NODEV;
995 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
996 	}
997 	if (p->p_comm[0] != '\0')
998 		strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
999 	if (p->p_sysent && p->p_sysent->sv_name != NULL &&
1000 	    p->p_sysent->sv_name[0] != '\0')
1001 		strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
1002 	kp->ki_siglist = p->p_siglist;
1003 	kp->ki_xstat = KW_EXITCODE(p->p_xexit, p->p_xsig);
1004 	kp->ki_acflag = p->p_acflag;
1005 	kp->ki_lock = p->p_lock;
1006 	if (p->p_pptr) {
1007 		kp->ki_ppid = proc_realparent(p)->p_pid;
1008 		if (p->p_flag & P_TRACED)
1009 			kp->ki_tracer = p->p_pptr->p_pid;
1010 	}
1011 }
1012 
1013 /*
1014  * Fill in information that is thread specific.  Must be called with
1015  * target process locked.  If 'preferthread' is set, overwrite certain
1016  * process-related fields that are maintained for both threads and
1017  * processes.
1018  */
1019 static void
1020 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
1021 {
1022 	struct proc *p;
1023 
1024 	p = td->td_proc;
1025 	kp->ki_tdaddr = td;
1026 	PROC_LOCK_ASSERT(p, MA_OWNED);
1027 
1028 	if (preferthread)
1029 		PROC_STATLOCK(p);
1030 	thread_lock(td);
1031 	if (td->td_wmesg != NULL)
1032 		strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
1033 	else
1034 		bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
1035 	if (strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname)) >=
1036 	    sizeof(kp->ki_tdname)) {
1037 		strlcpy(kp->ki_moretdname,
1038 		    td->td_name + sizeof(kp->ki_tdname) - 1,
1039 		    sizeof(kp->ki_moretdname));
1040 	} else {
1041 		bzero(kp->ki_moretdname, sizeof(kp->ki_moretdname));
1042 	}
1043 	if (TD_ON_LOCK(td)) {
1044 		kp->ki_kiflag |= KI_LOCKBLOCK;
1045 		strlcpy(kp->ki_lockname, td->td_lockname,
1046 		    sizeof(kp->ki_lockname));
1047 	} else {
1048 		kp->ki_kiflag &= ~KI_LOCKBLOCK;
1049 		bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
1050 	}
1051 
1052 	if (p->p_state == PRS_NORMAL) { /* approximate. */
1053 		if (TD_ON_RUNQ(td) ||
1054 		    TD_CAN_RUN(td) ||
1055 		    TD_IS_RUNNING(td)) {
1056 			kp->ki_stat = SRUN;
1057 		} else if (P_SHOULDSTOP(p)) {
1058 			kp->ki_stat = SSTOP;
1059 		} else if (TD_IS_SLEEPING(td)) {
1060 			kp->ki_stat = SSLEEP;
1061 		} else if (TD_ON_LOCK(td)) {
1062 			kp->ki_stat = SLOCK;
1063 		} else {
1064 			kp->ki_stat = SWAIT;
1065 		}
1066 	} else if (p->p_state == PRS_ZOMBIE) {
1067 		kp->ki_stat = SZOMB;
1068 	} else {
1069 		kp->ki_stat = SIDL;
1070 	}
1071 
1072 	/* Things in the thread */
1073 	kp->ki_wchan = td->td_wchan;
1074 	kp->ki_pri.pri_level = td->td_priority;
1075 	kp->ki_pri.pri_native = td->td_base_pri;
1076 
1077 	/*
1078 	 * Note: legacy fields; clamp at the old NOCPU value and/or
1079 	 * the maximum u_char CPU value.
1080 	 */
1081 	if (td->td_lastcpu == NOCPU)
1082 		kp->ki_lastcpu_old = NOCPU_OLD;
1083 	else if (td->td_lastcpu > MAXCPU_OLD)
1084 		kp->ki_lastcpu_old = MAXCPU_OLD;
1085 	else
1086 		kp->ki_lastcpu_old = td->td_lastcpu;
1087 
1088 	if (td->td_oncpu == NOCPU)
1089 		kp->ki_oncpu_old = NOCPU_OLD;
1090 	else if (td->td_oncpu > MAXCPU_OLD)
1091 		kp->ki_oncpu_old = MAXCPU_OLD;
1092 	else
1093 		kp->ki_oncpu_old = td->td_oncpu;
1094 
1095 	kp->ki_lastcpu = td->td_lastcpu;
1096 	kp->ki_oncpu = td->td_oncpu;
1097 	kp->ki_tdflags = td->td_flags;
1098 	kp->ki_tid = td->td_tid;
1099 	kp->ki_numthreads = p->p_numthreads;
1100 	kp->ki_pcb = td->td_pcb;
1101 	kp->ki_kstack = (void *)td->td_kstack;
1102 	kp->ki_slptime = (ticks - td->td_slptick) / hz;
1103 	kp->ki_pri.pri_class = td->td_pri_class;
1104 	kp->ki_pri.pri_user = td->td_user_pri;
1105 
1106 	if (preferthread) {
1107 		rufetchtd(td, &kp->ki_rusage);
1108 		kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
1109 		kp->ki_pctcpu = sched_pctcpu(td);
1110 		kp->ki_estcpu = sched_estcpu(td);
1111 		kp->ki_cow = td->td_cow;
1112 	}
1113 
1114 	/* We can't get this anymore but ps etc never used it anyway. */
1115 	kp->ki_rqindex = 0;
1116 
1117 	if (preferthread)
1118 		kp->ki_siglist = td->td_siglist;
1119 	kp->ki_sigmask = td->td_sigmask;
1120 	thread_unlock(td);
1121 	if (preferthread)
1122 		PROC_STATUNLOCK(p);
1123 }
1124 
1125 /*
1126  * Fill in a kinfo_proc structure for the specified process.
1127  * Must be called with the target process locked.
1128  */
1129 void
1130 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1131 {
1132 
1133 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1134 
1135 	fill_kinfo_proc_only(p, kp);
1136 	fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1137 	fill_kinfo_aggregate(p, kp);
1138 }
1139 
1140 struct pstats *
1141 pstats_alloc(void)
1142 {
1143 
1144 	return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1145 }
1146 
1147 /*
1148  * Copy parts of p_stats; zero the rest of p_stats (statistics).
1149  */
1150 void
1151 pstats_fork(struct pstats *src, struct pstats *dst)
1152 {
1153 
1154 	bzero(&dst->pstat_startzero,
1155 	    __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1156 	bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1157 	    __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1158 }
1159 
1160 void
1161 pstats_free(struct pstats *ps)
1162 {
1163 
1164 	free(ps, M_SUBPROC);
1165 }
1166 
1167 static struct proc *
1168 zpfind_locked(pid_t pid)
1169 {
1170 	struct proc *p;
1171 
1172 	sx_assert(&allproc_lock, SX_LOCKED);
1173 	LIST_FOREACH(p, &zombproc, p_list) {
1174 		if (p->p_pid == pid) {
1175 			PROC_LOCK(p);
1176 			break;
1177 		}
1178 	}
1179 	return (p);
1180 }
1181 
1182 /*
1183  * Locate a zombie process by number
1184  */
1185 struct proc *
1186 zpfind(pid_t pid)
1187 {
1188 	struct proc *p;
1189 
1190 	sx_slock(&allproc_lock);
1191 	p = zpfind_locked(pid);
1192 	sx_sunlock(&allproc_lock);
1193 	return (p);
1194 }
1195 
1196 #ifdef COMPAT_FREEBSD32
1197 
1198 /*
1199  * This function is typically used to copy out the kernel address, so
1200  * it can be replaced by assignment of zero.
1201  */
1202 static inline uint32_t
1203 ptr32_trim(void *ptr)
1204 {
1205 	uintptr_t uptr;
1206 
1207 	uptr = (uintptr_t)ptr;
1208 	return ((uptr > UINT_MAX) ? 0 : uptr);
1209 }
1210 
1211 #define PTRTRIM_CP(src,dst,fld) \
1212 	do { (dst).fld = ptr32_trim((src).fld); } while (0)
1213 
1214 static void
1215 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1216 {
1217 	int i;
1218 
1219 	bzero(ki32, sizeof(struct kinfo_proc32));
1220 	ki32->ki_structsize = sizeof(struct kinfo_proc32);
1221 	CP(*ki, *ki32, ki_layout);
1222 	PTRTRIM_CP(*ki, *ki32, ki_args);
1223 	PTRTRIM_CP(*ki, *ki32, ki_paddr);
1224 	PTRTRIM_CP(*ki, *ki32, ki_addr);
1225 	PTRTRIM_CP(*ki, *ki32, ki_tracep);
1226 	PTRTRIM_CP(*ki, *ki32, ki_textvp);
1227 	PTRTRIM_CP(*ki, *ki32, ki_fd);
1228 	PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1229 	PTRTRIM_CP(*ki, *ki32, ki_wchan);
1230 	CP(*ki, *ki32, ki_pid);
1231 	CP(*ki, *ki32, ki_ppid);
1232 	CP(*ki, *ki32, ki_pgid);
1233 	CP(*ki, *ki32, ki_tpgid);
1234 	CP(*ki, *ki32, ki_sid);
1235 	CP(*ki, *ki32, ki_tsid);
1236 	CP(*ki, *ki32, ki_jobc);
1237 	CP(*ki, *ki32, ki_tdev);
1238 	CP(*ki, *ki32, ki_tdev_freebsd11);
1239 	CP(*ki, *ki32, ki_siglist);
1240 	CP(*ki, *ki32, ki_sigmask);
1241 	CP(*ki, *ki32, ki_sigignore);
1242 	CP(*ki, *ki32, ki_sigcatch);
1243 	CP(*ki, *ki32, ki_uid);
1244 	CP(*ki, *ki32, ki_ruid);
1245 	CP(*ki, *ki32, ki_svuid);
1246 	CP(*ki, *ki32, ki_rgid);
1247 	CP(*ki, *ki32, ki_svgid);
1248 	CP(*ki, *ki32, ki_ngroups);
1249 	for (i = 0; i < KI_NGROUPS; i++)
1250 		CP(*ki, *ki32, ki_groups[i]);
1251 	CP(*ki, *ki32, ki_size);
1252 	CP(*ki, *ki32, ki_rssize);
1253 	CP(*ki, *ki32, ki_swrss);
1254 	CP(*ki, *ki32, ki_tsize);
1255 	CP(*ki, *ki32, ki_dsize);
1256 	CP(*ki, *ki32, ki_ssize);
1257 	CP(*ki, *ki32, ki_xstat);
1258 	CP(*ki, *ki32, ki_acflag);
1259 	CP(*ki, *ki32, ki_pctcpu);
1260 	CP(*ki, *ki32, ki_estcpu);
1261 	CP(*ki, *ki32, ki_slptime);
1262 	CP(*ki, *ki32, ki_swtime);
1263 	CP(*ki, *ki32, ki_cow);
1264 	CP(*ki, *ki32, ki_runtime);
1265 	TV_CP(*ki, *ki32, ki_start);
1266 	TV_CP(*ki, *ki32, ki_childtime);
1267 	CP(*ki, *ki32, ki_flag);
1268 	CP(*ki, *ki32, ki_kiflag);
1269 	CP(*ki, *ki32, ki_traceflag);
1270 	CP(*ki, *ki32, ki_stat);
1271 	CP(*ki, *ki32, ki_nice);
1272 	CP(*ki, *ki32, ki_lock);
1273 	CP(*ki, *ki32, ki_rqindex);
1274 	CP(*ki, *ki32, ki_oncpu);
1275 	CP(*ki, *ki32, ki_lastcpu);
1276 
1277 	/* XXX TODO: wrap cpu value as appropriate */
1278 	CP(*ki, *ki32, ki_oncpu_old);
1279 	CP(*ki, *ki32, ki_lastcpu_old);
1280 
1281 	bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1282 	bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1283 	bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1284 	bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1285 	bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1286 	bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1287 	bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1288 	bcopy(ki->ki_moretdname, ki32->ki_moretdname, MAXCOMLEN - TDNAMLEN + 1);
1289 	CP(*ki, *ki32, ki_tracer);
1290 	CP(*ki, *ki32, ki_flag2);
1291 	CP(*ki, *ki32, ki_fibnum);
1292 	CP(*ki, *ki32, ki_cr_flags);
1293 	CP(*ki, *ki32, ki_jid);
1294 	CP(*ki, *ki32, ki_numthreads);
1295 	CP(*ki, *ki32, ki_tid);
1296 	CP(*ki, *ki32, ki_pri);
1297 	freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1298 	freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1299 	PTRTRIM_CP(*ki, *ki32, ki_pcb);
1300 	PTRTRIM_CP(*ki, *ki32, ki_kstack);
1301 	PTRTRIM_CP(*ki, *ki32, ki_udata);
1302 	CP(*ki, *ki32, ki_sflag);
1303 	CP(*ki, *ki32, ki_tdflags);
1304 }
1305 #endif
1306 
1307 int
1308 kern_proc_out(struct proc *p, struct sbuf *sb, int flags)
1309 {
1310 	struct thread *td;
1311 	struct kinfo_proc ki;
1312 #ifdef COMPAT_FREEBSD32
1313 	struct kinfo_proc32 ki32;
1314 #endif
1315 	int error;
1316 
1317 	PROC_LOCK_ASSERT(p, MA_OWNED);
1318 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1319 
1320 	error = 0;
1321 	fill_kinfo_proc(p, &ki);
1322 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1323 #ifdef COMPAT_FREEBSD32
1324 		if ((flags & KERN_PROC_MASK32) != 0) {
1325 			freebsd32_kinfo_proc_out(&ki, &ki32);
1326 			if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1327 				error = ENOMEM;
1328 		} else
1329 #endif
1330 			if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1331 				error = ENOMEM;
1332 	} else {
1333 		FOREACH_THREAD_IN_PROC(p, td) {
1334 			fill_kinfo_thread(td, &ki, 1);
1335 #ifdef COMPAT_FREEBSD32
1336 			if ((flags & KERN_PROC_MASK32) != 0) {
1337 				freebsd32_kinfo_proc_out(&ki, &ki32);
1338 				if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1339 					error = ENOMEM;
1340 			} else
1341 #endif
1342 				if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1343 					error = ENOMEM;
1344 			if (error != 0)
1345 				break;
1346 		}
1347 	}
1348 	PROC_UNLOCK(p);
1349 	return (error);
1350 }
1351 
1352 static int
1353 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags,
1354     int doingzomb)
1355 {
1356 	struct sbuf sb;
1357 	struct kinfo_proc ki;
1358 	struct proc *np;
1359 	int error, error2;
1360 	pid_t pid;
1361 
1362 	pid = p->p_pid;
1363 	sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req);
1364 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1365 	error = kern_proc_out(p, &sb, flags);
1366 	error2 = sbuf_finish(&sb);
1367 	sbuf_delete(&sb);
1368 	if (error != 0)
1369 		return (error);
1370 	else if (error2 != 0)
1371 		return (error2);
1372 	if (doingzomb)
1373 		np = zpfind(pid);
1374 	else {
1375 		if (pid == 0)
1376 			return (0);
1377 		np = pfind(pid);
1378 	}
1379 	if (np == NULL)
1380 		return (ESRCH);
1381 	if (np != p) {
1382 		PROC_UNLOCK(np);
1383 		return (ESRCH);
1384 	}
1385 	PROC_UNLOCK(np);
1386 	return (0);
1387 }
1388 
1389 static int
1390 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1391 {
1392 	int *name = (int *)arg1;
1393 	u_int namelen = arg2;
1394 	struct proc *p;
1395 	int flags, doingzomb, oid_number;
1396 	int error = 0;
1397 
1398 	oid_number = oidp->oid_number;
1399 	if (oid_number != KERN_PROC_ALL &&
1400 	    (oid_number & KERN_PROC_INC_THREAD) == 0)
1401 		flags = KERN_PROC_NOTHREADS;
1402 	else {
1403 		flags = 0;
1404 		oid_number &= ~KERN_PROC_INC_THREAD;
1405 	}
1406 #ifdef COMPAT_FREEBSD32
1407 	if (req->flags & SCTL_MASK32)
1408 		flags |= KERN_PROC_MASK32;
1409 #endif
1410 	if (oid_number == KERN_PROC_PID) {
1411 		if (namelen != 1)
1412 			return (EINVAL);
1413 		error = sysctl_wire_old_buffer(req, 0);
1414 		if (error)
1415 			return (error);
1416 		sx_slock(&proctree_lock);
1417 		error = pget((pid_t)name[0], PGET_CANSEE, &p);
1418 		if (error == 0)
1419 			error = sysctl_out_proc(p, req, flags, 0);
1420 		sx_sunlock(&proctree_lock);
1421 		return (error);
1422 	}
1423 
1424 	switch (oid_number) {
1425 	case KERN_PROC_ALL:
1426 		if (namelen != 0)
1427 			return (EINVAL);
1428 		break;
1429 	case KERN_PROC_PROC:
1430 		if (namelen != 0 && namelen != 1)
1431 			return (EINVAL);
1432 		break;
1433 	default:
1434 		if (namelen != 1)
1435 			return (EINVAL);
1436 		break;
1437 	}
1438 
1439 	if (!req->oldptr) {
1440 		/* overestimate by 5 procs */
1441 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1442 		if (error)
1443 			return (error);
1444 	}
1445 	error = sysctl_wire_old_buffer(req, 0);
1446 	if (error != 0)
1447 		return (error);
1448 	sx_slock(&proctree_lock);
1449 	sx_slock(&allproc_lock);
1450 	for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) {
1451 		if (!doingzomb)
1452 			p = LIST_FIRST(&allproc);
1453 		else
1454 			p = LIST_FIRST(&zombproc);
1455 		for (; p != NULL; p = LIST_NEXT(p, p_list)) {
1456 			/*
1457 			 * Skip embryonic processes.
1458 			 */
1459 			PROC_LOCK(p);
1460 			if (p->p_state == PRS_NEW) {
1461 				PROC_UNLOCK(p);
1462 				continue;
1463 			}
1464 			KASSERT(p->p_ucred != NULL,
1465 			    ("process credential is NULL for non-NEW proc"));
1466 			/*
1467 			 * Show a user only appropriate processes.
1468 			 */
1469 			if (p_cansee(curthread, p)) {
1470 				PROC_UNLOCK(p);
1471 				continue;
1472 			}
1473 			/*
1474 			 * TODO - make more efficient (see notes below).
1475 			 * do by session.
1476 			 */
1477 			switch (oid_number) {
1478 
1479 			case KERN_PROC_GID:
1480 				if (p->p_ucred->cr_gid != (gid_t)name[0]) {
1481 					PROC_UNLOCK(p);
1482 					continue;
1483 				}
1484 				break;
1485 
1486 			case KERN_PROC_PGRP:
1487 				/* could do this by traversing pgrp */
1488 				if (p->p_pgrp == NULL ||
1489 				    p->p_pgrp->pg_id != (pid_t)name[0]) {
1490 					PROC_UNLOCK(p);
1491 					continue;
1492 				}
1493 				break;
1494 
1495 			case KERN_PROC_RGID:
1496 				if (p->p_ucred->cr_rgid != (gid_t)name[0]) {
1497 					PROC_UNLOCK(p);
1498 					continue;
1499 				}
1500 				break;
1501 
1502 			case KERN_PROC_SESSION:
1503 				if (p->p_session == NULL ||
1504 				    p->p_session->s_sid != (pid_t)name[0]) {
1505 					PROC_UNLOCK(p);
1506 					continue;
1507 				}
1508 				break;
1509 
1510 			case KERN_PROC_TTY:
1511 				if ((p->p_flag & P_CONTROLT) == 0 ||
1512 				    p->p_session == NULL) {
1513 					PROC_UNLOCK(p);
1514 					continue;
1515 				}
1516 				/* XXX proctree_lock */
1517 				SESS_LOCK(p->p_session);
1518 				if (p->p_session->s_ttyp == NULL ||
1519 				    tty_udev(p->p_session->s_ttyp) !=
1520 				    (dev_t)name[0]) {
1521 					SESS_UNLOCK(p->p_session);
1522 					PROC_UNLOCK(p);
1523 					continue;
1524 				}
1525 				SESS_UNLOCK(p->p_session);
1526 				break;
1527 
1528 			case KERN_PROC_UID:
1529 				if (p->p_ucred->cr_uid != (uid_t)name[0]) {
1530 					PROC_UNLOCK(p);
1531 					continue;
1532 				}
1533 				break;
1534 
1535 			case KERN_PROC_RUID:
1536 				if (p->p_ucred->cr_ruid != (uid_t)name[0]) {
1537 					PROC_UNLOCK(p);
1538 					continue;
1539 				}
1540 				break;
1541 
1542 			case KERN_PROC_PROC:
1543 				break;
1544 
1545 			default:
1546 				break;
1547 
1548 			}
1549 
1550 			error = sysctl_out_proc(p, req, flags, doingzomb);
1551 			if (error) {
1552 				sx_sunlock(&allproc_lock);
1553 				sx_sunlock(&proctree_lock);
1554 				return (error);
1555 			}
1556 		}
1557 	}
1558 	sx_sunlock(&allproc_lock);
1559 	sx_sunlock(&proctree_lock);
1560 	return (0);
1561 }
1562 
1563 struct pargs *
1564 pargs_alloc(int len)
1565 {
1566 	struct pargs *pa;
1567 
1568 	pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1569 		M_WAITOK);
1570 	refcount_init(&pa->ar_ref, 1);
1571 	pa->ar_length = len;
1572 	return (pa);
1573 }
1574 
1575 static void
1576 pargs_free(struct pargs *pa)
1577 {
1578 
1579 	free(pa, M_PARGS);
1580 }
1581 
1582 void
1583 pargs_hold(struct pargs *pa)
1584 {
1585 
1586 	if (pa == NULL)
1587 		return;
1588 	refcount_acquire(&pa->ar_ref);
1589 }
1590 
1591 void
1592 pargs_drop(struct pargs *pa)
1593 {
1594 
1595 	if (pa == NULL)
1596 		return;
1597 	if (refcount_release(&pa->ar_ref))
1598 		pargs_free(pa);
1599 }
1600 
1601 static int
1602 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1603     size_t len)
1604 {
1605 	ssize_t n;
1606 
1607 	/*
1608 	 * This may return a short read if the string is shorter than the chunk
1609 	 * and is aligned at the end of the page, and the following page is not
1610 	 * mapped.
1611 	 */
1612 	n = proc_readmem(td, p, (vm_offset_t)sptr, buf, len);
1613 	if (n <= 0)
1614 		return (ENOMEM);
1615 	return (0);
1616 }
1617 
1618 #define PROC_AUXV_MAX	256	/* Safety limit on auxv size. */
1619 
1620 enum proc_vector_type {
1621 	PROC_ARG,
1622 	PROC_ENV,
1623 	PROC_AUX,
1624 };
1625 
1626 #ifdef COMPAT_FREEBSD32
1627 static int
1628 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1629     size_t *vsizep, enum proc_vector_type type)
1630 {
1631 	struct freebsd32_ps_strings pss;
1632 	Elf32_Auxinfo aux;
1633 	vm_offset_t vptr, ptr;
1634 	uint32_t *proc_vector32;
1635 	char **proc_vector;
1636 	size_t vsize, size;
1637 	int i, error;
1638 
1639 	error = 0;
1640 	if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss,
1641 	    sizeof(pss)) != sizeof(pss))
1642 		return (ENOMEM);
1643 	switch (type) {
1644 	case PROC_ARG:
1645 		vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1646 		vsize = pss.ps_nargvstr;
1647 		if (vsize > ARG_MAX)
1648 			return (ENOEXEC);
1649 		size = vsize * sizeof(int32_t);
1650 		break;
1651 	case PROC_ENV:
1652 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1653 		vsize = pss.ps_nenvstr;
1654 		if (vsize > ARG_MAX)
1655 			return (ENOEXEC);
1656 		size = vsize * sizeof(int32_t);
1657 		break;
1658 	case PROC_AUX:
1659 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1660 		    (pss.ps_nenvstr + 1) * sizeof(int32_t);
1661 		if (vptr % 4 != 0)
1662 			return (ENOEXEC);
1663 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1664 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1665 			    sizeof(aux))
1666 				return (ENOMEM);
1667 			if (aux.a_type == AT_NULL)
1668 				break;
1669 			ptr += sizeof(aux);
1670 		}
1671 		if (aux.a_type != AT_NULL)
1672 			return (ENOEXEC);
1673 		vsize = i + 1;
1674 		size = vsize * sizeof(aux);
1675 		break;
1676 	default:
1677 		KASSERT(0, ("Wrong proc vector type: %d", type));
1678 		return (EINVAL);
1679 	}
1680 	proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1681 	if (proc_readmem(td, p, vptr, proc_vector32, size) != size) {
1682 		error = ENOMEM;
1683 		goto done;
1684 	}
1685 	if (type == PROC_AUX) {
1686 		*proc_vectorp = (char **)proc_vector32;
1687 		*vsizep = vsize;
1688 		return (0);
1689 	}
1690 	proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1691 	for (i = 0; i < (int)vsize; i++)
1692 		proc_vector[i] = PTRIN(proc_vector32[i]);
1693 	*proc_vectorp = proc_vector;
1694 	*vsizep = vsize;
1695 done:
1696 	free(proc_vector32, M_TEMP);
1697 	return (error);
1698 }
1699 #endif
1700 
1701 static int
1702 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1703     size_t *vsizep, enum proc_vector_type type)
1704 {
1705 	struct ps_strings pss;
1706 	Elf_Auxinfo aux;
1707 	vm_offset_t vptr, ptr;
1708 	char **proc_vector;
1709 	size_t vsize, size;
1710 	int i;
1711 
1712 #ifdef COMPAT_FREEBSD32
1713 	if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1714 		return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1715 #endif
1716 	if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss,
1717 	    sizeof(pss)) != sizeof(pss))
1718 		return (ENOMEM);
1719 	switch (type) {
1720 	case PROC_ARG:
1721 		vptr = (vm_offset_t)pss.ps_argvstr;
1722 		vsize = pss.ps_nargvstr;
1723 		if (vsize > ARG_MAX)
1724 			return (ENOEXEC);
1725 		size = vsize * sizeof(char *);
1726 		break;
1727 	case PROC_ENV:
1728 		vptr = (vm_offset_t)pss.ps_envstr;
1729 		vsize = pss.ps_nenvstr;
1730 		if (vsize > ARG_MAX)
1731 			return (ENOEXEC);
1732 		size = vsize * sizeof(char *);
1733 		break;
1734 	case PROC_AUX:
1735 		/*
1736 		 * The aux array is just above env array on the stack. Check
1737 		 * that the address is naturally aligned.
1738 		 */
1739 		vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1740 		    * sizeof(char *);
1741 #if __ELF_WORD_SIZE == 64
1742 		if (vptr % sizeof(uint64_t) != 0)
1743 #else
1744 		if (vptr % sizeof(uint32_t) != 0)
1745 #endif
1746 			return (ENOEXEC);
1747 		/*
1748 		 * We count the array size reading the aux vectors from the
1749 		 * stack until AT_NULL vector is returned.  So (to keep the code
1750 		 * simple) we read the process stack twice: the first time here
1751 		 * to find the size and the second time when copying the vectors
1752 		 * to the allocated proc_vector.
1753 		 */
1754 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1755 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1756 			    sizeof(aux))
1757 				return (ENOMEM);
1758 			if (aux.a_type == AT_NULL)
1759 				break;
1760 			ptr += sizeof(aux);
1761 		}
1762 		/*
1763 		 * If the PROC_AUXV_MAX entries are iterated over, and we have
1764 		 * not reached AT_NULL, it is most likely we are reading wrong
1765 		 * data: either the process doesn't have auxv array or data has
1766 		 * been modified. Return the error in this case.
1767 		 */
1768 		if (aux.a_type != AT_NULL)
1769 			return (ENOEXEC);
1770 		vsize = i + 1;
1771 		size = vsize * sizeof(aux);
1772 		break;
1773 	default:
1774 		KASSERT(0, ("Wrong proc vector type: %d", type));
1775 		return (EINVAL); /* In case we are built without INVARIANTS. */
1776 	}
1777 	proc_vector = malloc(size, M_TEMP, M_WAITOK);
1778 	if (proc_readmem(td, p, vptr, proc_vector, size) != size) {
1779 		free(proc_vector, M_TEMP);
1780 		return (ENOMEM);
1781 	}
1782 	*proc_vectorp = proc_vector;
1783 	*vsizep = vsize;
1784 
1785 	return (0);
1786 }
1787 
1788 #define GET_PS_STRINGS_CHUNK_SZ	256	/* Chunk size (bytes) for ps_strings operations. */
1789 
1790 static int
1791 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
1792     enum proc_vector_type type)
1793 {
1794 	size_t done, len, nchr, vsize;
1795 	int error, i;
1796 	char **proc_vector, *sptr;
1797 	char pss_string[GET_PS_STRINGS_CHUNK_SZ];
1798 
1799 	PROC_ASSERT_HELD(p);
1800 
1801 	/*
1802 	 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
1803 	 */
1804 	nchr = 2 * (PATH_MAX + ARG_MAX);
1805 
1806 	error = get_proc_vector(td, p, &proc_vector, &vsize, type);
1807 	if (error != 0)
1808 		return (error);
1809 	for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
1810 		/*
1811 		 * The program may have scribbled into its argv array, e.g. to
1812 		 * remove some arguments.  If that has happened, break out
1813 		 * before trying to read from NULL.
1814 		 */
1815 		if (proc_vector[i] == NULL)
1816 			break;
1817 		for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
1818 			error = proc_read_string(td, p, sptr, pss_string,
1819 			    sizeof(pss_string));
1820 			if (error != 0)
1821 				goto done;
1822 			len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
1823 			if (done + len >= nchr)
1824 				len = nchr - done - 1;
1825 			sbuf_bcat(sb, pss_string, len);
1826 			if (len != GET_PS_STRINGS_CHUNK_SZ)
1827 				break;
1828 			done += GET_PS_STRINGS_CHUNK_SZ;
1829 		}
1830 		sbuf_bcat(sb, "", 1);
1831 		done += len + 1;
1832 	}
1833 done:
1834 	free(proc_vector, M_TEMP);
1835 	return (error);
1836 }
1837 
1838 int
1839 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
1840 {
1841 
1842 	return (get_ps_strings(curthread, p, sb, PROC_ARG));
1843 }
1844 
1845 int
1846 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
1847 {
1848 
1849 	return (get_ps_strings(curthread, p, sb, PROC_ENV));
1850 }
1851 
1852 int
1853 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb)
1854 {
1855 	size_t vsize, size;
1856 	char **auxv;
1857 	int error;
1858 
1859 	error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX);
1860 	if (error == 0) {
1861 #ifdef COMPAT_FREEBSD32
1862 		if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1863 			size = vsize * sizeof(Elf32_Auxinfo);
1864 		else
1865 #endif
1866 			size = vsize * sizeof(Elf_Auxinfo);
1867 		if (sbuf_bcat(sb, auxv, size) != 0)
1868 			error = ENOMEM;
1869 		free(auxv, M_TEMP);
1870 	}
1871 	return (error);
1872 }
1873 
1874 /*
1875  * This sysctl allows a process to retrieve the argument list or process
1876  * title for another process without groping around in the address space
1877  * of the other process.  It also allow a process to set its own "process
1878  * title to a string of its own choice.
1879  */
1880 static int
1881 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
1882 {
1883 	int *name = (int *)arg1;
1884 	u_int namelen = arg2;
1885 	struct pargs *newpa, *pa;
1886 	struct proc *p;
1887 	struct sbuf sb;
1888 	int flags, error = 0, error2;
1889 
1890 	if (namelen != 1)
1891 		return (EINVAL);
1892 
1893 	flags = PGET_CANSEE;
1894 	if (req->newptr != NULL)
1895 		flags |= PGET_ISCURRENT;
1896 	error = pget((pid_t)name[0], flags, &p);
1897 	if (error)
1898 		return (error);
1899 
1900 	pa = p->p_args;
1901 	if (pa != NULL) {
1902 		pargs_hold(pa);
1903 		PROC_UNLOCK(p);
1904 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
1905 		pargs_drop(pa);
1906 	} else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
1907 		_PHOLD(p);
1908 		PROC_UNLOCK(p);
1909 		sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1910 		sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1911 		error = proc_getargv(curthread, p, &sb);
1912 		error2 = sbuf_finish(&sb);
1913 		PRELE(p);
1914 		sbuf_delete(&sb);
1915 		if (error == 0 && error2 != 0)
1916 			error = error2;
1917 	} else {
1918 		PROC_UNLOCK(p);
1919 	}
1920 	if (error != 0 || req->newptr == NULL)
1921 		return (error);
1922 
1923 	if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit)
1924 		return (ENOMEM);
1925 	newpa = pargs_alloc(req->newlen);
1926 	error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
1927 	if (error != 0) {
1928 		pargs_free(newpa);
1929 		return (error);
1930 	}
1931 	PROC_LOCK(p);
1932 	pa = p->p_args;
1933 	p->p_args = newpa;
1934 	PROC_UNLOCK(p);
1935 	pargs_drop(pa);
1936 	return (0);
1937 }
1938 
1939 /*
1940  * This sysctl allows a process to retrieve environment of another process.
1941  */
1942 static int
1943 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
1944 {
1945 	int *name = (int *)arg1;
1946 	u_int namelen = arg2;
1947 	struct proc *p;
1948 	struct sbuf sb;
1949 	int error, error2;
1950 
1951 	if (namelen != 1)
1952 		return (EINVAL);
1953 
1954 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1955 	if (error != 0)
1956 		return (error);
1957 	if ((p->p_flag & P_SYSTEM) != 0) {
1958 		PRELE(p);
1959 		return (0);
1960 	}
1961 
1962 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1963 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1964 	error = proc_getenvv(curthread, p, &sb);
1965 	error2 = sbuf_finish(&sb);
1966 	PRELE(p);
1967 	sbuf_delete(&sb);
1968 	return (error != 0 ? error : error2);
1969 }
1970 
1971 /*
1972  * This sysctl allows a process to retrieve ELF auxiliary vector of
1973  * another process.
1974  */
1975 static int
1976 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
1977 {
1978 	int *name = (int *)arg1;
1979 	u_int namelen = arg2;
1980 	struct proc *p;
1981 	struct sbuf sb;
1982 	int error, error2;
1983 
1984 	if (namelen != 1)
1985 		return (EINVAL);
1986 
1987 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1988 	if (error != 0)
1989 		return (error);
1990 	if ((p->p_flag & P_SYSTEM) != 0) {
1991 		PRELE(p);
1992 		return (0);
1993 	}
1994 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1995 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1996 	error = proc_getauxv(curthread, p, &sb);
1997 	error2 = sbuf_finish(&sb);
1998 	PRELE(p);
1999 	sbuf_delete(&sb);
2000 	return (error != 0 ? error : error2);
2001 }
2002 
2003 /*
2004  * This sysctl allows a process to retrieve the path of the executable for
2005  * itself or another process.
2006  */
2007 static int
2008 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
2009 {
2010 	pid_t *pidp = (pid_t *)arg1;
2011 	unsigned int arglen = arg2;
2012 	struct proc *p;
2013 	struct vnode *vp;
2014 	char *retbuf, *freebuf;
2015 	int error;
2016 
2017 	if (arglen != 1)
2018 		return (EINVAL);
2019 	if (*pidp == -1) {	/* -1 means this process */
2020 		p = req->td->td_proc;
2021 	} else {
2022 		error = pget(*pidp, PGET_CANSEE, &p);
2023 		if (error != 0)
2024 			return (error);
2025 	}
2026 
2027 	vp = p->p_textvp;
2028 	if (vp == NULL) {
2029 		if (*pidp != -1)
2030 			PROC_UNLOCK(p);
2031 		return (0);
2032 	}
2033 	vref(vp);
2034 	if (*pidp != -1)
2035 		PROC_UNLOCK(p);
2036 	error = vn_fullpath(req->td, vp, &retbuf, &freebuf);
2037 	vrele(vp);
2038 	if (error)
2039 		return (error);
2040 	error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
2041 	free(freebuf, M_TEMP);
2042 	return (error);
2043 }
2044 
2045 static int
2046 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
2047 {
2048 	struct proc *p;
2049 	char *sv_name;
2050 	int *name;
2051 	int namelen;
2052 	int error;
2053 
2054 	namelen = arg2;
2055 	if (namelen != 1)
2056 		return (EINVAL);
2057 
2058 	name = (int *)arg1;
2059 	error = pget((pid_t)name[0], PGET_CANSEE, &p);
2060 	if (error != 0)
2061 		return (error);
2062 	sv_name = p->p_sysent->sv_name;
2063 	PROC_UNLOCK(p);
2064 	return (sysctl_handle_string(oidp, sv_name, 0, req));
2065 }
2066 
2067 #ifdef KINFO_OVMENTRY_SIZE
2068 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
2069 #endif
2070 
2071 #ifdef COMPAT_FREEBSD7
2072 static int
2073 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
2074 {
2075 	vm_map_entry_t entry, tmp_entry;
2076 	unsigned int last_timestamp;
2077 	char *fullpath, *freepath;
2078 	struct kinfo_ovmentry *kve;
2079 	struct vattr va;
2080 	struct ucred *cred;
2081 	int error, *name;
2082 	struct vnode *vp;
2083 	struct proc *p;
2084 	vm_map_t map;
2085 	struct vmspace *vm;
2086 
2087 	name = (int *)arg1;
2088 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2089 	if (error != 0)
2090 		return (error);
2091 	vm = vmspace_acquire_ref(p);
2092 	if (vm == NULL) {
2093 		PRELE(p);
2094 		return (ESRCH);
2095 	}
2096 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2097 
2098 	map = &vm->vm_map;
2099 	vm_map_lock_read(map);
2100 	for (entry = map->header.next; entry != &map->header;
2101 	    entry = entry->next) {
2102 		vm_object_t obj, tobj, lobj;
2103 		vm_offset_t addr;
2104 
2105 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2106 			continue;
2107 
2108 		bzero(kve, sizeof(*kve));
2109 		kve->kve_structsize = sizeof(*kve);
2110 
2111 		kve->kve_private_resident = 0;
2112 		obj = entry->object.vm_object;
2113 		if (obj != NULL) {
2114 			VM_OBJECT_RLOCK(obj);
2115 			if (obj->shadow_count == 1)
2116 				kve->kve_private_resident =
2117 				    obj->resident_page_count;
2118 		}
2119 		kve->kve_resident = 0;
2120 		addr = entry->start;
2121 		while (addr < entry->end) {
2122 			if (pmap_extract(map->pmap, addr))
2123 				kve->kve_resident++;
2124 			addr += PAGE_SIZE;
2125 		}
2126 
2127 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2128 			if (tobj != obj)
2129 				VM_OBJECT_RLOCK(tobj);
2130 			if (lobj != obj)
2131 				VM_OBJECT_RUNLOCK(lobj);
2132 			lobj = tobj;
2133 		}
2134 
2135 		kve->kve_start = (void*)entry->start;
2136 		kve->kve_end = (void*)entry->end;
2137 		kve->kve_offset = (off_t)entry->offset;
2138 
2139 		if (entry->protection & VM_PROT_READ)
2140 			kve->kve_protection |= KVME_PROT_READ;
2141 		if (entry->protection & VM_PROT_WRITE)
2142 			kve->kve_protection |= KVME_PROT_WRITE;
2143 		if (entry->protection & VM_PROT_EXECUTE)
2144 			kve->kve_protection |= KVME_PROT_EXEC;
2145 
2146 		if (entry->eflags & MAP_ENTRY_COW)
2147 			kve->kve_flags |= KVME_FLAG_COW;
2148 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2149 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2150 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2151 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2152 
2153 		last_timestamp = map->timestamp;
2154 		vm_map_unlock_read(map);
2155 
2156 		kve->kve_fileid = 0;
2157 		kve->kve_fsid = 0;
2158 		freepath = NULL;
2159 		fullpath = "";
2160 		if (lobj) {
2161 			vp = NULL;
2162 			switch (lobj->type) {
2163 			case OBJT_DEFAULT:
2164 				kve->kve_type = KVME_TYPE_DEFAULT;
2165 				break;
2166 			case OBJT_VNODE:
2167 				kve->kve_type = KVME_TYPE_VNODE;
2168 				vp = lobj->handle;
2169 				vref(vp);
2170 				break;
2171 			case OBJT_SWAP:
2172 				if ((lobj->flags & OBJ_TMPFS_NODE) != 0) {
2173 					kve->kve_type = KVME_TYPE_VNODE;
2174 					if ((lobj->flags & OBJ_TMPFS) != 0) {
2175 						vp = lobj->un_pager.swp.swp_tmpfs;
2176 						vref(vp);
2177 					}
2178 				} else {
2179 					kve->kve_type = KVME_TYPE_SWAP;
2180 				}
2181 				break;
2182 			case OBJT_DEVICE:
2183 				kve->kve_type = KVME_TYPE_DEVICE;
2184 				break;
2185 			case OBJT_PHYS:
2186 				kve->kve_type = KVME_TYPE_PHYS;
2187 				break;
2188 			case OBJT_DEAD:
2189 				kve->kve_type = KVME_TYPE_DEAD;
2190 				break;
2191 			case OBJT_SG:
2192 				kve->kve_type = KVME_TYPE_SG;
2193 				break;
2194 			default:
2195 				kve->kve_type = KVME_TYPE_UNKNOWN;
2196 				break;
2197 			}
2198 			if (lobj != obj)
2199 				VM_OBJECT_RUNLOCK(lobj);
2200 
2201 			kve->kve_ref_count = obj->ref_count;
2202 			kve->kve_shadow_count = obj->shadow_count;
2203 			VM_OBJECT_RUNLOCK(obj);
2204 			if (vp != NULL) {
2205 				vn_fullpath(curthread, vp, &fullpath,
2206 				    &freepath);
2207 				cred = curthread->td_ucred;
2208 				vn_lock(vp, LK_SHARED | LK_RETRY);
2209 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2210 					kve->kve_fileid = va.va_fileid;
2211 					/* truncate */
2212 					kve->kve_fsid = va.va_fsid;
2213 				}
2214 				vput(vp);
2215 			}
2216 		} else {
2217 			kve->kve_type = KVME_TYPE_NONE;
2218 			kve->kve_ref_count = 0;
2219 			kve->kve_shadow_count = 0;
2220 		}
2221 
2222 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2223 		if (freepath != NULL)
2224 			free(freepath, M_TEMP);
2225 
2226 		error = SYSCTL_OUT(req, kve, sizeof(*kve));
2227 		vm_map_lock_read(map);
2228 		if (error)
2229 			break;
2230 		if (last_timestamp != map->timestamp) {
2231 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2232 			entry = tmp_entry;
2233 		}
2234 	}
2235 	vm_map_unlock_read(map);
2236 	vmspace_free(vm);
2237 	PRELE(p);
2238 	free(kve, M_TEMP);
2239 	return (error);
2240 }
2241 #endif	/* COMPAT_FREEBSD7 */
2242 
2243 #ifdef KINFO_VMENTRY_SIZE
2244 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2245 #endif
2246 
2247 static void
2248 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry,
2249     struct kinfo_vmentry *kve)
2250 {
2251 	vm_object_t obj, tobj;
2252 	vm_page_t m, m_adv;
2253 	vm_offset_t addr;
2254 	vm_paddr_t locked_pa;
2255 	vm_pindex_t pi, pi_adv, pindex;
2256 
2257 	locked_pa = 0;
2258 	obj = entry->object.vm_object;
2259 	addr = entry->start;
2260 	m_adv = NULL;
2261 	pi = OFF_TO_IDX(entry->offset);
2262 	for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) {
2263 		if (m_adv != NULL) {
2264 			m = m_adv;
2265 		} else {
2266 			pi_adv = atop(entry->end - addr);
2267 			pindex = pi;
2268 			for (tobj = obj;; tobj = tobj->backing_object) {
2269 				m = vm_page_find_least(tobj, pindex);
2270 				if (m != NULL) {
2271 					if (m->pindex == pindex)
2272 						break;
2273 					if (pi_adv > m->pindex - pindex) {
2274 						pi_adv = m->pindex - pindex;
2275 						m_adv = m;
2276 					}
2277 				}
2278 				if (tobj->backing_object == NULL)
2279 					goto next;
2280 				pindex += OFF_TO_IDX(tobj->
2281 				    backing_object_offset);
2282 			}
2283 		}
2284 		m_adv = NULL;
2285 		if (m->psind != 0 && addr + pagesizes[1] <= entry->end &&
2286 		    (addr & (pagesizes[1] - 1)) == 0 &&
2287 		    (pmap_mincore(map->pmap, addr, &locked_pa) &
2288 		    MINCORE_SUPER) != 0) {
2289 			kve->kve_flags |= KVME_FLAG_SUPER;
2290 			pi_adv = atop(pagesizes[1]);
2291 		} else {
2292 			/*
2293 			 * We do not test the found page on validity.
2294 			 * Either the page is busy and being paged in,
2295 			 * or it was invalidated.  The first case
2296 			 * should be counted as resident, the second
2297 			 * is not so clear; we do account both.
2298 			 */
2299 			pi_adv = 1;
2300 		}
2301 		kve->kve_resident += pi_adv;
2302 next:;
2303 	}
2304 	PA_UNLOCK_COND(locked_pa);
2305 }
2306 
2307 /*
2308  * Must be called with the process locked and will return unlocked.
2309  */
2310 int
2311 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb, ssize_t maxlen, int flags)
2312 {
2313 	vm_map_entry_t entry, tmp_entry;
2314 	struct vattr va;
2315 	vm_map_t map;
2316 	vm_object_t obj, tobj, lobj;
2317 	char *fullpath, *freepath;
2318 	struct kinfo_vmentry *kve;
2319 	struct ucred *cred;
2320 	struct vnode *vp;
2321 	struct vmspace *vm;
2322 	vm_offset_t addr;
2323 	unsigned int last_timestamp;
2324 	int error;
2325 
2326 	PROC_LOCK_ASSERT(p, MA_OWNED);
2327 
2328 	_PHOLD(p);
2329 	PROC_UNLOCK(p);
2330 	vm = vmspace_acquire_ref(p);
2331 	if (vm == NULL) {
2332 		PRELE(p);
2333 		return (ESRCH);
2334 	}
2335 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK | M_ZERO);
2336 
2337 	error = 0;
2338 	map = &vm->vm_map;
2339 	vm_map_lock_read(map);
2340 	for (entry = map->header.next; entry != &map->header;
2341 	    entry = entry->next) {
2342 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2343 			continue;
2344 
2345 		addr = entry->end;
2346 		bzero(kve, sizeof(*kve));
2347 		obj = entry->object.vm_object;
2348 		if (obj != NULL) {
2349 			for (tobj = obj; tobj != NULL;
2350 			    tobj = tobj->backing_object) {
2351 				VM_OBJECT_RLOCK(tobj);
2352 				lobj = tobj;
2353 			}
2354 			if (obj->backing_object == NULL)
2355 				kve->kve_private_resident =
2356 				    obj->resident_page_count;
2357 			if (!vmmap_skip_res_cnt)
2358 				kern_proc_vmmap_resident(map, entry, kve);
2359 			for (tobj = obj; tobj != NULL;
2360 			    tobj = tobj->backing_object) {
2361 				if (tobj != obj && tobj != lobj)
2362 					VM_OBJECT_RUNLOCK(tobj);
2363 			}
2364 		} else {
2365 			lobj = NULL;
2366 		}
2367 
2368 		kve->kve_start = entry->start;
2369 		kve->kve_end = entry->end;
2370 		kve->kve_offset = entry->offset;
2371 
2372 		if (entry->protection & VM_PROT_READ)
2373 			kve->kve_protection |= KVME_PROT_READ;
2374 		if (entry->protection & VM_PROT_WRITE)
2375 			kve->kve_protection |= KVME_PROT_WRITE;
2376 		if (entry->protection & VM_PROT_EXECUTE)
2377 			kve->kve_protection |= KVME_PROT_EXEC;
2378 
2379 		if (entry->eflags & MAP_ENTRY_COW)
2380 			kve->kve_flags |= KVME_FLAG_COW;
2381 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2382 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2383 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2384 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2385 		if (entry->eflags & MAP_ENTRY_GROWS_UP)
2386 			kve->kve_flags |= KVME_FLAG_GROWS_UP;
2387 		if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2388 			kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2389 
2390 		last_timestamp = map->timestamp;
2391 		vm_map_unlock_read(map);
2392 
2393 		freepath = NULL;
2394 		fullpath = "";
2395 		if (lobj != NULL) {
2396 			vp = NULL;
2397 			switch (lobj->type) {
2398 			case OBJT_DEFAULT:
2399 				kve->kve_type = KVME_TYPE_DEFAULT;
2400 				break;
2401 			case OBJT_VNODE:
2402 				kve->kve_type = KVME_TYPE_VNODE;
2403 				vp = lobj->handle;
2404 				vref(vp);
2405 				break;
2406 			case OBJT_SWAP:
2407 				if ((lobj->flags & OBJ_TMPFS_NODE) != 0) {
2408 					kve->kve_type = KVME_TYPE_VNODE;
2409 					if ((lobj->flags & OBJ_TMPFS) != 0) {
2410 						vp = lobj->un_pager.swp.swp_tmpfs;
2411 						vref(vp);
2412 					}
2413 				} else {
2414 					kve->kve_type = KVME_TYPE_SWAP;
2415 				}
2416 				break;
2417 			case OBJT_DEVICE:
2418 				kve->kve_type = KVME_TYPE_DEVICE;
2419 				break;
2420 			case OBJT_PHYS:
2421 				kve->kve_type = KVME_TYPE_PHYS;
2422 				break;
2423 			case OBJT_DEAD:
2424 				kve->kve_type = KVME_TYPE_DEAD;
2425 				break;
2426 			case OBJT_SG:
2427 				kve->kve_type = KVME_TYPE_SG;
2428 				break;
2429 			case OBJT_MGTDEVICE:
2430 				kve->kve_type = KVME_TYPE_MGTDEVICE;
2431 				break;
2432 			default:
2433 				kve->kve_type = KVME_TYPE_UNKNOWN;
2434 				break;
2435 			}
2436 			if (lobj != obj)
2437 				VM_OBJECT_RUNLOCK(lobj);
2438 
2439 			kve->kve_ref_count = obj->ref_count;
2440 			kve->kve_shadow_count = obj->shadow_count;
2441 			VM_OBJECT_RUNLOCK(obj);
2442 			if (vp != NULL) {
2443 				vn_fullpath(curthread, vp, &fullpath,
2444 				    &freepath);
2445 				kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2446 				cred = curthread->td_ucred;
2447 				vn_lock(vp, LK_SHARED | LK_RETRY);
2448 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2449 					kve->kve_vn_fileid = va.va_fileid;
2450 					kve->kve_vn_fsid = va.va_fsid;
2451 					kve->kve_vn_fsid_freebsd11 =
2452 					    kve->kve_vn_fsid; /* truncate */
2453 					kve->kve_vn_mode =
2454 					    MAKEIMODE(va.va_type, va.va_mode);
2455 					kve->kve_vn_size = va.va_size;
2456 					kve->kve_vn_rdev = va.va_rdev;
2457 					kve->kve_vn_rdev_freebsd11 =
2458 					    kve->kve_vn_rdev; /* truncate */
2459 					kve->kve_status = KF_ATTR_VALID;
2460 				}
2461 				vput(vp);
2462 			}
2463 		} else {
2464 			kve->kve_type = KVME_TYPE_NONE;
2465 			kve->kve_ref_count = 0;
2466 			kve->kve_shadow_count = 0;
2467 		}
2468 
2469 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2470 		if (freepath != NULL)
2471 			free(freepath, M_TEMP);
2472 
2473 		/* Pack record size down */
2474 		if ((flags & KERN_VMMAP_PACK_KINFO) != 0)
2475 			kve->kve_structsize =
2476 			    offsetof(struct kinfo_vmentry, kve_path) +
2477 			    strlen(kve->kve_path) + 1;
2478 		else
2479 			kve->kve_structsize = sizeof(*kve);
2480 		kve->kve_structsize = roundup(kve->kve_structsize,
2481 		    sizeof(uint64_t));
2482 
2483 		/* Halt filling and truncate rather than exceeding maxlen */
2484 		if (maxlen != -1 && maxlen < kve->kve_structsize) {
2485 			error = 0;
2486 			vm_map_lock_read(map);
2487 			break;
2488 		} else if (maxlen != -1)
2489 			maxlen -= kve->kve_structsize;
2490 
2491 		if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0)
2492 			error = ENOMEM;
2493 		vm_map_lock_read(map);
2494 		if (error != 0)
2495 			break;
2496 		if (last_timestamp != map->timestamp) {
2497 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2498 			entry = tmp_entry;
2499 		}
2500 	}
2501 	vm_map_unlock_read(map);
2502 	vmspace_free(vm);
2503 	PRELE(p);
2504 	free(kve, M_TEMP);
2505 	return (error);
2506 }
2507 
2508 static int
2509 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2510 {
2511 	struct proc *p;
2512 	struct sbuf sb;
2513 	int error, error2, *name;
2514 
2515 	name = (int *)arg1;
2516 	sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req);
2517 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2518 	error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
2519 	if (error != 0) {
2520 		sbuf_delete(&sb);
2521 		return (error);
2522 	}
2523 	error = kern_proc_vmmap_out(p, &sb, -1, KERN_VMMAP_PACK_KINFO);
2524 	error2 = sbuf_finish(&sb);
2525 	sbuf_delete(&sb);
2526 	return (error != 0 ? error : error2);
2527 }
2528 
2529 #if defined(STACK) || defined(DDB)
2530 static int
2531 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2532 {
2533 	struct kinfo_kstack *kkstp;
2534 	int error, i, *name, numthreads;
2535 	lwpid_t *lwpidarray;
2536 	struct thread *td;
2537 	struct stack *st;
2538 	struct sbuf sb;
2539 	struct proc *p;
2540 
2541 	name = (int *)arg1;
2542 	error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2543 	if (error != 0)
2544 		return (error);
2545 
2546 	kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2547 	st = stack_create();
2548 
2549 	lwpidarray = NULL;
2550 	PROC_LOCK(p);
2551 	do {
2552 		if (lwpidarray != NULL) {
2553 			free(lwpidarray, M_TEMP);
2554 			lwpidarray = NULL;
2555 		}
2556 		numthreads = p->p_numthreads;
2557 		PROC_UNLOCK(p);
2558 		lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2559 		    M_WAITOK | M_ZERO);
2560 		PROC_LOCK(p);
2561 	} while (numthreads < p->p_numthreads);
2562 
2563 	/*
2564 	 * XXXRW: During the below loop, execve(2) and countless other sorts
2565 	 * of changes could have taken place.  Should we check to see if the
2566 	 * vmspace has been replaced, or the like, in order to prevent
2567 	 * giving a snapshot that spans, say, execve(2), with some threads
2568 	 * before and some after?  Among other things, the credentials could
2569 	 * have changed, in which case the right to extract debug info might
2570 	 * no longer be assured.
2571 	 */
2572 	i = 0;
2573 	FOREACH_THREAD_IN_PROC(p, td) {
2574 		KASSERT(i < numthreads,
2575 		    ("sysctl_kern_proc_kstack: numthreads"));
2576 		lwpidarray[i] = td->td_tid;
2577 		i++;
2578 	}
2579 	numthreads = i;
2580 	for (i = 0; i < numthreads; i++) {
2581 		td = thread_find(p, lwpidarray[i]);
2582 		if (td == NULL) {
2583 			continue;
2584 		}
2585 		bzero(kkstp, sizeof(*kkstp));
2586 		(void)sbuf_new(&sb, kkstp->kkst_trace,
2587 		    sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2588 		thread_lock(td);
2589 		kkstp->kkst_tid = td->td_tid;
2590 		if (TD_IS_SWAPPED(td)) {
2591 			kkstp->kkst_state = KKST_STATE_SWAPPED;
2592 		} else if (TD_IS_RUNNING(td)) {
2593 			if (stack_save_td_running(st, td) == 0)
2594 				kkstp->kkst_state = KKST_STATE_STACKOK;
2595 			else
2596 				kkstp->kkst_state = KKST_STATE_RUNNING;
2597 		} else {
2598 			kkstp->kkst_state = KKST_STATE_STACKOK;
2599 			stack_save_td(st, td);
2600 		}
2601 		thread_unlock(td);
2602 		PROC_UNLOCK(p);
2603 		stack_sbuf_print(&sb, st);
2604 		sbuf_finish(&sb);
2605 		sbuf_delete(&sb);
2606 		error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2607 		PROC_LOCK(p);
2608 		if (error)
2609 			break;
2610 	}
2611 	_PRELE(p);
2612 	PROC_UNLOCK(p);
2613 	if (lwpidarray != NULL)
2614 		free(lwpidarray, M_TEMP);
2615 	stack_destroy(st);
2616 	free(kkstp, M_TEMP);
2617 	return (error);
2618 }
2619 #endif
2620 
2621 /*
2622  * This sysctl allows a process to retrieve the full list of groups from
2623  * itself or another process.
2624  */
2625 static int
2626 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2627 {
2628 	pid_t *pidp = (pid_t *)arg1;
2629 	unsigned int arglen = arg2;
2630 	struct proc *p;
2631 	struct ucred *cred;
2632 	int error;
2633 
2634 	if (arglen != 1)
2635 		return (EINVAL);
2636 	if (*pidp == -1) {	/* -1 means this process */
2637 		p = req->td->td_proc;
2638 		PROC_LOCK(p);
2639 	} else {
2640 		error = pget(*pidp, PGET_CANSEE, &p);
2641 		if (error != 0)
2642 			return (error);
2643 	}
2644 
2645 	cred = crhold(p->p_ucred);
2646 	PROC_UNLOCK(p);
2647 
2648 	error = SYSCTL_OUT(req, cred->cr_groups,
2649 	    cred->cr_ngroups * sizeof(gid_t));
2650 	crfree(cred);
2651 	return (error);
2652 }
2653 
2654 /*
2655  * This sysctl allows a process to retrieve or/and set the resource limit for
2656  * another process.
2657  */
2658 static int
2659 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2660 {
2661 	int *name = (int *)arg1;
2662 	u_int namelen = arg2;
2663 	struct rlimit rlim;
2664 	struct proc *p;
2665 	u_int which;
2666 	int flags, error;
2667 
2668 	if (namelen != 2)
2669 		return (EINVAL);
2670 
2671 	which = (u_int)name[1];
2672 	if (which >= RLIM_NLIMITS)
2673 		return (EINVAL);
2674 
2675 	if (req->newptr != NULL && req->newlen != sizeof(rlim))
2676 		return (EINVAL);
2677 
2678 	flags = PGET_HOLD | PGET_NOTWEXIT;
2679 	if (req->newptr != NULL)
2680 		flags |= PGET_CANDEBUG;
2681 	else
2682 		flags |= PGET_CANSEE;
2683 	error = pget((pid_t)name[0], flags, &p);
2684 	if (error != 0)
2685 		return (error);
2686 
2687 	/*
2688 	 * Retrieve limit.
2689 	 */
2690 	if (req->oldptr != NULL) {
2691 		PROC_LOCK(p);
2692 		lim_rlimit_proc(p, which, &rlim);
2693 		PROC_UNLOCK(p);
2694 	}
2695 	error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
2696 	if (error != 0)
2697 		goto errout;
2698 
2699 	/*
2700 	 * Set limit.
2701 	 */
2702 	if (req->newptr != NULL) {
2703 		error = SYSCTL_IN(req, &rlim, sizeof(rlim));
2704 		if (error == 0)
2705 			error = kern_proc_setrlimit(curthread, p, which, &rlim);
2706 	}
2707 
2708 errout:
2709 	PRELE(p);
2710 	return (error);
2711 }
2712 
2713 /*
2714  * This sysctl allows a process to retrieve ps_strings structure location of
2715  * another process.
2716  */
2717 static int
2718 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
2719 {
2720 	int *name = (int *)arg1;
2721 	u_int namelen = arg2;
2722 	struct proc *p;
2723 	vm_offset_t ps_strings;
2724 	int error;
2725 #ifdef COMPAT_FREEBSD32
2726 	uint32_t ps_strings32;
2727 #endif
2728 
2729 	if (namelen != 1)
2730 		return (EINVAL);
2731 
2732 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2733 	if (error != 0)
2734 		return (error);
2735 #ifdef COMPAT_FREEBSD32
2736 	if ((req->flags & SCTL_MASK32) != 0) {
2737 		/*
2738 		 * We return 0 if the 32 bit emulation request is for a 64 bit
2739 		 * process.
2740 		 */
2741 		ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
2742 		    PTROUT(p->p_sysent->sv_psstrings) : 0;
2743 		PROC_UNLOCK(p);
2744 		error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
2745 		return (error);
2746 	}
2747 #endif
2748 	ps_strings = p->p_sysent->sv_psstrings;
2749 	PROC_UNLOCK(p);
2750 	error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
2751 	return (error);
2752 }
2753 
2754 /*
2755  * This sysctl allows a process to retrieve umask of another process.
2756  */
2757 static int
2758 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
2759 {
2760 	int *name = (int *)arg1;
2761 	u_int namelen = arg2;
2762 	struct proc *p;
2763 	int error;
2764 	u_short fd_cmask;
2765 
2766 	if (namelen != 1)
2767 		return (EINVAL);
2768 
2769 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2770 	if (error != 0)
2771 		return (error);
2772 
2773 	FILEDESC_SLOCK(p->p_fd);
2774 	fd_cmask = p->p_fd->fd_cmask;
2775 	FILEDESC_SUNLOCK(p->p_fd);
2776 	PRELE(p);
2777 	error = SYSCTL_OUT(req, &fd_cmask, sizeof(fd_cmask));
2778 	return (error);
2779 }
2780 
2781 /*
2782  * This sysctl allows a process to set and retrieve binary osreldate of
2783  * another process.
2784  */
2785 static int
2786 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
2787 {
2788 	int *name = (int *)arg1;
2789 	u_int namelen = arg2;
2790 	struct proc *p;
2791 	int flags, error, osrel;
2792 
2793 	if (namelen != 1)
2794 		return (EINVAL);
2795 
2796 	if (req->newptr != NULL && req->newlen != sizeof(osrel))
2797 		return (EINVAL);
2798 
2799 	flags = PGET_HOLD | PGET_NOTWEXIT;
2800 	if (req->newptr != NULL)
2801 		flags |= PGET_CANDEBUG;
2802 	else
2803 		flags |= PGET_CANSEE;
2804 	error = pget((pid_t)name[0], flags, &p);
2805 	if (error != 0)
2806 		return (error);
2807 
2808 	error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
2809 	if (error != 0)
2810 		goto errout;
2811 
2812 	if (req->newptr != NULL) {
2813 		error = SYSCTL_IN(req, &osrel, sizeof(osrel));
2814 		if (error != 0)
2815 			goto errout;
2816 		if (osrel < 0) {
2817 			error = EINVAL;
2818 			goto errout;
2819 		}
2820 		p->p_osrel = osrel;
2821 	}
2822 errout:
2823 	PRELE(p);
2824 	return (error);
2825 }
2826 
2827 static int
2828 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS)
2829 {
2830 	int *name = (int *)arg1;
2831 	u_int namelen = arg2;
2832 	struct proc *p;
2833 	struct kinfo_sigtramp kst;
2834 	const struct sysentvec *sv;
2835 	int error;
2836 #ifdef COMPAT_FREEBSD32
2837 	struct kinfo_sigtramp32 kst32;
2838 #endif
2839 
2840 	if (namelen != 1)
2841 		return (EINVAL);
2842 
2843 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2844 	if (error != 0)
2845 		return (error);
2846 	sv = p->p_sysent;
2847 #ifdef COMPAT_FREEBSD32
2848 	if ((req->flags & SCTL_MASK32) != 0) {
2849 		bzero(&kst32, sizeof(kst32));
2850 		if (SV_PROC_FLAG(p, SV_ILP32)) {
2851 			if (sv->sv_sigcode_base != 0) {
2852 				kst32.ksigtramp_start = sv->sv_sigcode_base;
2853 				kst32.ksigtramp_end = sv->sv_sigcode_base +
2854 				    *sv->sv_szsigcode;
2855 			} else {
2856 				kst32.ksigtramp_start = sv->sv_psstrings -
2857 				    *sv->sv_szsigcode;
2858 				kst32.ksigtramp_end = sv->sv_psstrings;
2859 			}
2860 		}
2861 		PROC_UNLOCK(p);
2862 		error = SYSCTL_OUT(req, &kst32, sizeof(kst32));
2863 		return (error);
2864 	}
2865 #endif
2866 	bzero(&kst, sizeof(kst));
2867 	if (sv->sv_sigcode_base != 0) {
2868 		kst.ksigtramp_start = (char *)sv->sv_sigcode_base;
2869 		kst.ksigtramp_end = (char *)sv->sv_sigcode_base +
2870 		    *sv->sv_szsigcode;
2871 	} else {
2872 		kst.ksigtramp_start = (char *)sv->sv_psstrings -
2873 		    *sv->sv_szsigcode;
2874 		kst.ksigtramp_end = (char *)sv->sv_psstrings;
2875 	}
2876 	PROC_UNLOCK(p);
2877 	error = SYSCTL_OUT(req, &kst, sizeof(kst));
2878 	return (error);
2879 }
2880 
2881 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD,  0, "Process table");
2882 
2883 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
2884 	CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
2885 	"Return entire process table");
2886 
2887 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2888 	sysctl_kern_proc, "Process table");
2889 
2890 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
2891 	sysctl_kern_proc, "Process table");
2892 
2893 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2894 	sysctl_kern_proc, "Process table");
2895 
2896 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
2897 	CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2898 
2899 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
2900 	sysctl_kern_proc, "Process table");
2901 
2902 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2903 	sysctl_kern_proc, "Process table");
2904 
2905 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2906 	sysctl_kern_proc, "Process table");
2907 
2908 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2909 	sysctl_kern_proc, "Process table");
2910 
2911 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
2912 	sysctl_kern_proc, "Return process table, no threads");
2913 
2914 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
2915 	CTLFLAG_RW | CTLFLAG_CAPWR | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
2916 	sysctl_kern_proc_args, "Process argument list");
2917 
2918 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
2919 	sysctl_kern_proc_env, "Process environment");
2920 
2921 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
2922 	CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
2923 
2924 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
2925 	CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
2926 
2927 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
2928 	CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
2929 	"Process syscall vector name (ABI type)");
2930 
2931 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
2932 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2933 
2934 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
2935 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2936 
2937 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
2938 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2939 
2940 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
2941 	sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2942 
2943 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
2944 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2945 
2946 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
2947 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2948 
2949 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
2950 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2951 
2952 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
2953 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2954 
2955 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
2956 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
2957 	"Return process table, no threads");
2958 
2959 #ifdef COMPAT_FREEBSD7
2960 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
2961 	CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
2962 #endif
2963 
2964 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
2965 	CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
2966 
2967 #if defined(STACK) || defined(DDB)
2968 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
2969 	CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
2970 #endif
2971 
2972 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
2973 	CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
2974 
2975 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
2976 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
2977 	"Process resource limits");
2978 
2979 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
2980 	CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
2981 	"Process ps_strings location");
2982 
2983 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
2984 	CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
2985 
2986 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
2987 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
2988 	"Process binary osreldate");
2989 
2990 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD |
2991 	CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp,
2992 	"Process signal trampoline location");
2993 
2994 int allproc_gen;
2995 
2996 /*
2997  * stop_all_proc() purpose is to stop all process which have usermode,
2998  * except current process for obvious reasons.  This makes it somewhat
2999  * unreliable when invoked from multithreaded process.  The service
3000  * must not be user-callable anyway.
3001  */
3002 void
3003 stop_all_proc(void)
3004 {
3005 	struct proc *cp, *p;
3006 	int r, gen;
3007 	bool restart, seen_stopped, seen_exiting, stopped_some;
3008 
3009 	cp = curproc;
3010 allproc_loop:
3011 	sx_xlock(&allproc_lock);
3012 	gen = allproc_gen;
3013 	seen_exiting = seen_stopped = stopped_some = restart = false;
3014 	LIST_REMOVE(cp, p_list);
3015 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3016 	for (;;) {
3017 		p = LIST_NEXT(cp, p_list);
3018 		if (p == NULL)
3019 			break;
3020 		LIST_REMOVE(cp, p_list);
3021 		LIST_INSERT_AFTER(p, cp, p_list);
3022 		PROC_LOCK(p);
3023 		if ((p->p_flag & (P_KPROC | P_SYSTEM | P_TOTAL_STOP)) != 0) {
3024 			PROC_UNLOCK(p);
3025 			continue;
3026 		}
3027 		if ((p->p_flag & P_WEXIT) != 0) {
3028 			seen_exiting = true;
3029 			PROC_UNLOCK(p);
3030 			continue;
3031 		}
3032 		if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) {
3033 			/*
3034 			 * Stopped processes are tolerated when there
3035 			 * are no other processes which might continue
3036 			 * them.  P_STOPPED_SINGLE but not
3037 			 * P_TOTAL_STOP process still has at least one
3038 			 * thread running.
3039 			 */
3040 			seen_stopped = true;
3041 			PROC_UNLOCK(p);
3042 			continue;
3043 		}
3044 		_PHOLD(p);
3045 		sx_xunlock(&allproc_lock);
3046 		r = thread_single(p, SINGLE_ALLPROC);
3047 		if (r != 0)
3048 			restart = true;
3049 		else
3050 			stopped_some = true;
3051 		_PRELE(p);
3052 		PROC_UNLOCK(p);
3053 		sx_xlock(&allproc_lock);
3054 	}
3055 	/* Catch forked children we did not see in iteration. */
3056 	if (gen != allproc_gen)
3057 		restart = true;
3058 	sx_xunlock(&allproc_lock);
3059 	if (restart || stopped_some || seen_exiting || seen_stopped) {
3060 		kern_yield(PRI_USER);
3061 		goto allproc_loop;
3062 	}
3063 }
3064 
3065 void
3066 resume_all_proc(void)
3067 {
3068 	struct proc *cp, *p;
3069 
3070 	cp = curproc;
3071 	sx_xlock(&allproc_lock);
3072 	LIST_REMOVE(cp, p_list);
3073 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3074 	for (;;) {
3075 		p = LIST_NEXT(cp, p_list);
3076 		if (p == NULL)
3077 			break;
3078 		LIST_REMOVE(cp, p_list);
3079 		LIST_INSERT_AFTER(p, cp, p_list);
3080 		PROC_LOCK(p);
3081 		if ((p->p_flag & P_TOTAL_STOP) != 0) {
3082 			sx_xunlock(&allproc_lock);
3083 			_PHOLD(p);
3084 			thread_single_end(p, SINGLE_ALLPROC);
3085 			_PRELE(p);
3086 			PROC_UNLOCK(p);
3087 			sx_xlock(&allproc_lock);
3088 		} else {
3089 			PROC_UNLOCK(p);
3090 		}
3091 	}
3092 	sx_xunlock(&allproc_lock);
3093 }
3094 
3095 /* #define	TOTAL_STOP_DEBUG	1 */
3096 #ifdef TOTAL_STOP_DEBUG
3097 volatile static int ap_resume;
3098 #include <sys/mount.h>
3099 
3100 static int
3101 sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS)
3102 {
3103 	int error, val;
3104 
3105 	val = 0;
3106 	ap_resume = 0;
3107 	error = sysctl_handle_int(oidp, &val, 0, req);
3108 	if (error != 0 || req->newptr == NULL)
3109 		return (error);
3110 	if (val != 0) {
3111 		stop_all_proc();
3112 		syncer_suspend();
3113 		while (ap_resume == 0)
3114 			;
3115 		syncer_resume();
3116 		resume_all_proc();
3117 	}
3118 	return (0);
3119 }
3120 
3121 SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW |
3122     CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0,
3123     sysctl_debug_stop_all_proc, "I",
3124     "");
3125 #endif
3126