xref: /freebsd/sys/kern/kern_proc.c (revision bcccd559e20d22c29805034b912659ffd0b97813)
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 __exclusive_cache_line allproc_lock;
135 struct sx __exclusive_cache_line proctree_lock;
136 struct mtx __exclusive_cache_line 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 EVENTHANDLER_LIST_DEFINE(process_ctor);
155 EVENTHANDLER_LIST_DEFINE(process_dtor);
156 EVENTHANDLER_LIST_DEFINE(process_init);
157 EVENTHANDLER_LIST_DEFINE(process_fini);
158 EVENTHANDLER_LIST_DEFINE(process_exit);
159 EVENTHANDLER_LIST_DEFINE(process_fork);
160 EVENTHANDLER_LIST_DEFINE(process_exec);
161 
162 EVENTHANDLER_LIST_DECLARE(thread_ctor);
163 EVENTHANDLER_LIST_DECLARE(thread_dtor);
164 
165 int kstack_pages = KSTACK_PAGES;
166 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0,
167     "Kernel stack size in pages");
168 static int vmmap_skip_res_cnt = 0;
169 SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW,
170     &vmmap_skip_res_cnt, 0,
171     "Skip calculation of the pages resident count in kern.proc.vmmap");
172 
173 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
174 #ifdef COMPAT_FREEBSD32
175 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
176 #endif
177 
178 /*
179  * Initialize global process hashing structures.
180  */
181 void
182 procinit(void)
183 {
184 
185 	sx_init(&allproc_lock, "allproc");
186 	sx_init(&proctree_lock, "proctree");
187 	mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
188 	LIST_INIT(&allproc);
189 	LIST_INIT(&zombproc);
190 	pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
191 	pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
192 	proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
193 	    proc_ctor, proc_dtor, proc_init, proc_fini,
194 	    UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
195 	uihashinit();
196 }
197 
198 /*
199  * Prepare a proc for use.
200  */
201 static int
202 proc_ctor(void *mem, int size, void *arg, int flags)
203 {
204 	struct proc *p;
205 	struct thread *td;
206 
207 	p = (struct proc *)mem;
208 	SDT_PROBE4(proc, , ctor , entry, p, size, arg, flags);
209 	EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
210 	SDT_PROBE4(proc, , ctor , return, p, size, arg, flags);
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 	SDT_PROBE4(proc, , dtor, entry, p, size, arg, td);
232 	if (td != NULL) {
233 #ifdef INVARIANTS
234 		KASSERT((p->p_numthreads == 1),
235 		    ("bad number of threads in exiting process"));
236 		KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
237 #endif
238 		/* Free all OSD associated to this thread. */
239 		osd_thread_exit(td);
240 		td_softdep_cleanup(td);
241 		MPASS(td->td_su == NULL);
242 
243 		/* Make sure all thread destructors are executed */
244 		EVENTHANDLER_DIRECT_INVOKE(thread_dtor, td);
245 	}
246 	EVENTHANDLER_DIRECT_INVOKE(process_dtor, p);
247 	if (p->p_ksi != NULL)
248 		KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
249 	SDT_PROBE3(proc, , dtor, return, p, size, arg);
250 }
251 
252 /*
253  * Initialize type-stable parts of a proc (when newly created).
254  */
255 static int
256 proc_init(void *mem, int size, int flags)
257 {
258 	struct proc *p;
259 
260 	p = (struct proc *)mem;
261 	SDT_PROBE3(proc, , init, entry, p, size, flags);
262 	mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK | MTX_NEW);
263 	mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_NEW);
264 	mtx_init(&p->p_statmtx, "pstatl", NULL, MTX_SPIN | MTX_NEW);
265 	mtx_init(&p->p_itimmtx, "pitiml", NULL, MTX_SPIN | MTX_NEW);
266 	mtx_init(&p->p_profmtx, "pprofl", NULL, MTX_SPIN | MTX_NEW);
267 	cv_init(&p->p_pwait, "ppwait");
268 	cv_init(&p->p_dbgwait, "dbgwait");
269 	TAILQ_INIT(&p->p_threads);	     /* all threads in proc */
270 	EVENTHANDLER_DIRECT_INVOKE(process_init, p);
271 	p->p_stats = pstats_alloc();
272 	p->p_pgrp = NULL;
273 	SDT_PROBE3(proc, , init, return, p, size, flags);
274 	return (0);
275 }
276 
277 /*
278  * UMA should ensure that this function is never called.
279  * Freeing a proc structure would violate type stability.
280  */
281 static void
282 proc_fini(void *mem, int size)
283 {
284 #ifdef notnow
285 	struct proc *p;
286 
287 	p = (struct proc *)mem;
288 	EVENTHANDLER_DIRECT_INVOKE(process_fini, p);
289 	pstats_free(p->p_stats);
290 	thread_free(FIRST_THREAD_IN_PROC(p));
291 	mtx_destroy(&p->p_mtx);
292 	if (p->p_ksi != NULL)
293 		ksiginfo_free(p->p_ksi);
294 #else
295 	panic("proc reclaimed");
296 #endif
297 }
298 
299 /*
300  * Is p an inferior of the current process?
301  */
302 int
303 inferior(struct proc *p)
304 {
305 
306 	sx_assert(&proctree_lock, SX_LOCKED);
307 	PROC_LOCK_ASSERT(p, MA_OWNED);
308 	for (; p != curproc; p = proc_realparent(p)) {
309 		if (p->p_pid == 0)
310 			return (0);
311 	}
312 	return (1);
313 }
314 
315 struct proc *
316 pfind_locked(pid_t pid)
317 {
318 	struct proc *p;
319 
320 	sx_assert(&allproc_lock, SX_LOCKED);
321 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
322 		if (p->p_pid == pid) {
323 			PROC_LOCK(p);
324 			if (p->p_state == PRS_NEW) {
325 				PROC_UNLOCK(p);
326 				p = NULL;
327 			}
328 			break;
329 		}
330 	}
331 	return (p);
332 }
333 
334 /*
335  * Locate a process by number; return only "live" processes -- i.e., neither
336  * zombies nor newly born but incompletely initialized processes.  By not
337  * returning processes in the PRS_NEW state, we allow callers to avoid
338  * testing for that condition to avoid dereferencing p_ucred, et al.
339  */
340 struct proc *
341 pfind(pid_t pid)
342 {
343 	struct proc *p;
344 
345 	p = curproc;
346 	if (p->p_pid == pid) {
347 		PROC_LOCK(p);
348 		return (p);
349 	}
350 	sx_slock(&allproc_lock);
351 	p = pfind_locked(pid);
352 	sx_sunlock(&allproc_lock);
353 	return (p);
354 }
355 
356 /*
357  * Same as pfind but allow zombies.
358  */
359 struct proc *
360 pfind_any(pid_t pid)
361 {
362 	struct proc *p;
363 
364 	sx_slock(&allproc_lock);
365 	p = pfind_locked(pid);
366 	if (p == NULL)
367 		p = zpfind_locked(pid);
368 	sx_sunlock(&allproc_lock);
369 
370 	return (p);
371 }
372 
373 static struct proc *
374 pfind_tid_locked(pid_t tid)
375 {
376 	struct proc *p;
377 	struct thread *td;
378 
379 	sx_assert(&allproc_lock, SX_LOCKED);
380 	FOREACH_PROC_IN_SYSTEM(p) {
381 		PROC_LOCK(p);
382 		if (p->p_state == PRS_NEW) {
383 			PROC_UNLOCK(p);
384 			continue;
385 		}
386 		FOREACH_THREAD_IN_PROC(p, td) {
387 			if (td->td_tid == tid)
388 				goto found;
389 		}
390 		PROC_UNLOCK(p);
391 	}
392 found:
393 	return (p);
394 }
395 
396 /*
397  * Locate a process group by number.
398  * The caller must hold proctree_lock.
399  */
400 struct pgrp *
401 pgfind(pid_t pgid)
402 {
403 	struct pgrp *pgrp;
404 
405 	sx_assert(&proctree_lock, SX_LOCKED);
406 
407 	LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
408 		if (pgrp->pg_id == pgid) {
409 			PGRP_LOCK(pgrp);
410 			return (pgrp);
411 		}
412 	}
413 	return (NULL);
414 }
415 
416 /*
417  * Locate process and do additional manipulations, depending on flags.
418  */
419 int
420 pget(pid_t pid, int flags, struct proc **pp)
421 {
422 	struct proc *p;
423 	int error;
424 
425 	p = curproc;
426 	if (p->p_pid == pid) {
427 		PROC_LOCK(p);
428 	} else {
429 		sx_slock(&allproc_lock);
430 		if (pid <= PID_MAX) {
431 			p = pfind_locked(pid);
432 			if (p == NULL && (flags & PGET_NOTWEXIT) == 0)
433 				p = zpfind_locked(pid);
434 		} else if ((flags & PGET_NOTID) == 0) {
435 			p = pfind_tid_locked(pid);
436 		} else {
437 			p = NULL;
438 		}
439 		sx_sunlock(&allproc_lock);
440 		if (p == NULL)
441 			return (ESRCH);
442 		if ((flags & PGET_CANSEE) != 0) {
443 			error = p_cansee(curthread, p);
444 			if (error != 0)
445 				goto errout;
446 		}
447 	}
448 	if ((flags & PGET_CANDEBUG) != 0) {
449 		error = p_candebug(curthread, p);
450 		if (error != 0)
451 			goto errout;
452 	}
453 	if ((flags & PGET_ISCURRENT) != 0 && curproc != p) {
454 		error = EPERM;
455 		goto errout;
456 	}
457 	if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) {
458 		error = ESRCH;
459 		goto errout;
460 	}
461 	if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) {
462 		/*
463 		 * XXXRW: Not clear ESRCH is the right error during proc
464 		 * execve().
465 		 */
466 		error = ESRCH;
467 		goto errout;
468 	}
469 	if ((flags & PGET_HOLD) != 0) {
470 		_PHOLD(p);
471 		PROC_UNLOCK(p);
472 	}
473 	*pp = p;
474 	return (0);
475 errout:
476 	PROC_UNLOCK(p);
477 	return (error);
478 }
479 
480 /*
481  * Create a new process group.
482  * pgid must be equal to the pid of p.
483  * Begin a new session if required.
484  */
485 int
486 enterpgrp(struct proc *p, pid_t pgid, struct pgrp *pgrp, struct session *sess)
487 {
488 
489 	sx_assert(&proctree_lock, SX_XLOCKED);
490 
491 	KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
492 	KASSERT(p->p_pid == pgid,
493 	    ("enterpgrp: new pgrp and pid != pgid"));
494 	KASSERT(pgfind(pgid) == NULL,
495 	    ("enterpgrp: pgrp with pgid exists"));
496 	KASSERT(!SESS_LEADER(p),
497 	    ("enterpgrp: session leader attempted setpgrp"));
498 
499 	mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
500 
501 	if (sess != NULL) {
502 		/*
503 		 * new session
504 		 */
505 		mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
506 		PROC_LOCK(p);
507 		p->p_flag &= ~P_CONTROLT;
508 		PROC_UNLOCK(p);
509 		PGRP_LOCK(pgrp);
510 		sess->s_leader = p;
511 		sess->s_sid = p->p_pid;
512 		refcount_init(&sess->s_count, 1);
513 		sess->s_ttyvp = NULL;
514 		sess->s_ttydp = NULL;
515 		sess->s_ttyp = NULL;
516 		bcopy(p->p_session->s_login, sess->s_login,
517 			    sizeof(sess->s_login));
518 		pgrp->pg_session = sess;
519 		KASSERT(p == curproc,
520 		    ("enterpgrp: mksession and p != curproc"));
521 	} else {
522 		pgrp->pg_session = p->p_session;
523 		sess_hold(pgrp->pg_session);
524 		PGRP_LOCK(pgrp);
525 	}
526 	pgrp->pg_id = pgid;
527 	LIST_INIT(&pgrp->pg_members);
528 
529 	/*
530 	 * As we have an exclusive lock of proctree_lock,
531 	 * this should not deadlock.
532 	 */
533 	LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
534 	pgrp->pg_jobc = 0;
535 	SLIST_INIT(&pgrp->pg_sigiolst);
536 	PGRP_UNLOCK(pgrp);
537 
538 	doenterpgrp(p, pgrp);
539 
540 	return (0);
541 }
542 
543 /*
544  * Move p to an existing process group
545  */
546 int
547 enterthispgrp(struct proc *p, struct pgrp *pgrp)
548 {
549 
550 	sx_assert(&proctree_lock, SX_XLOCKED);
551 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
552 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
553 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
554 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
555 	KASSERT(pgrp->pg_session == p->p_session,
556 		("%s: pgrp's session %p, p->p_session %p.\n",
557 		__func__,
558 		pgrp->pg_session,
559 		p->p_session));
560 	KASSERT(pgrp != p->p_pgrp,
561 		("%s: p belongs to pgrp.", __func__));
562 
563 	doenterpgrp(p, pgrp);
564 
565 	return (0);
566 }
567 
568 /*
569  * Move p to a process group
570  */
571 static void
572 doenterpgrp(struct proc *p, struct pgrp *pgrp)
573 {
574 	struct pgrp *savepgrp;
575 
576 	sx_assert(&proctree_lock, SX_XLOCKED);
577 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
578 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
579 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
580 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
581 
582 	savepgrp = p->p_pgrp;
583 
584 	/*
585 	 * Adjust eligibility of affected pgrps to participate in job control.
586 	 * Increment eligibility counts before decrementing, otherwise we
587 	 * could reach 0 spuriously during the first call.
588 	 */
589 	fixjobc(p, pgrp, 1);
590 	fixjobc(p, p->p_pgrp, 0);
591 
592 	PGRP_LOCK(pgrp);
593 	PGRP_LOCK(savepgrp);
594 	PROC_LOCK(p);
595 	LIST_REMOVE(p, p_pglist);
596 	p->p_pgrp = pgrp;
597 	PROC_UNLOCK(p);
598 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
599 	PGRP_UNLOCK(savepgrp);
600 	PGRP_UNLOCK(pgrp);
601 	if (LIST_EMPTY(&savepgrp->pg_members))
602 		pgdelete(savepgrp);
603 }
604 
605 /*
606  * remove process from process group
607  */
608 int
609 leavepgrp(struct proc *p)
610 {
611 	struct pgrp *savepgrp;
612 
613 	sx_assert(&proctree_lock, SX_XLOCKED);
614 	savepgrp = p->p_pgrp;
615 	PGRP_LOCK(savepgrp);
616 	PROC_LOCK(p);
617 	LIST_REMOVE(p, p_pglist);
618 	p->p_pgrp = NULL;
619 	PROC_UNLOCK(p);
620 	PGRP_UNLOCK(savepgrp);
621 	if (LIST_EMPTY(&savepgrp->pg_members))
622 		pgdelete(savepgrp);
623 	return (0);
624 }
625 
626 /*
627  * delete a process group
628  */
629 static void
630 pgdelete(struct pgrp *pgrp)
631 {
632 	struct session *savesess;
633 	struct tty *tp;
634 
635 	sx_assert(&proctree_lock, SX_XLOCKED);
636 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
637 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
638 
639 	/*
640 	 * Reset any sigio structures pointing to us as a result of
641 	 * F_SETOWN with our pgid.
642 	 */
643 	funsetownlst(&pgrp->pg_sigiolst);
644 
645 	PGRP_LOCK(pgrp);
646 	tp = pgrp->pg_session->s_ttyp;
647 	LIST_REMOVE(pgrp, pg_hash);
648 	savesess = pgrp->pg_session;
649 	PGRP_UNLOCK(pgrp);
650 
651 	/* Remove the reference to the pgrp before deallocating it. */
652 	if (tp != NULL) {
653 		tty_lock(tp);
654 		tty_rel_pgrp(tp, pgrp);
655 	}
656 
657 	mtx_destroy(&pgrp->pg_mtx);
658 	free(pgrp, M_PGRP);
659 	sess_release(savesess);
660 }
661 
662 static void
663 pgadjustjobc(struct pgrp *pgrp, int entering)
664 {
665 
666 	PGRP_LOCK(pgrp);
667 	if (entering)
668 		pgrp->pg_jobc++;
669 	else {
670 		--pgrp->pg_jobc;
671 		if (pgrp->pg_jobc == 0)
672 			orphanpg(pgrp);
673 	}
674 	PGRP_UNLOCK(pgrp);
675 }
676 
677 /*
678  * Adjust pgrp jobc counters when specified process changes process group.
679  * We count the number of processes in each process group that "qualify"
680  * the group for terminal job control (those with a parent in a different
681  * process group of the same session).  If that count reaches zero, the
682  * process group becomes orphaned.  Check both the specified process'
683  * process group and that of its children.
684  * entering == 0 => p is leaving specified group.
685  * entering == 1 => p is entering specified group.
686  */
687 void
688 fixjobc(struct proc *p, struct pgrp *pgrp, int entering)
689 {
690 	struct pgrp *hispgrp;
691 	struct session *mysession;
692 	struct proc *q;
693 
694 	sx_assert(&proctree_lock, SX_LOCKED);
695 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
696 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
697 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
698 
699 	/*
700 	 * Check p's parent to see whether p qualifies its own process
701 	 * group; if so, adjust count for p's process group.
702 	 */
703 	mysession = pgrp->pg_session;
704 	if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
705 	    hispgrp->pg_session == mysession)
706 		pgadjustjobc(pgrp, entering);
707 
708 	/*
709 	 * Check this process' children to see whether they qualify
710 	 * their process groups; if so, adjust counts for children's
711 	 * process groups.
712 	 */
713 	LIST_FOREACH(q, &p->p_children, p_sibling) {
714 		hispgrp = q->p_pgrp;
715 		if (hispgrp == pgrp ||
716 		    hispgrp->pg_session != mysession)
717 			continue;
718 		if (q->p_state == PRS_ZOMBIE)
719 			continue;
720 		pgadjustjobc(hispgrp, entering);
721 	}
722 }
723 
724 void
725 killjobc(void)
726 {
727 	struct session *sp;
728 	struct tty *tp;
729 	struct proc *p;
730 	struct vnode *ttyvp;
731 
732 	p = curproc;
733 	MPASS(p->p_flag & P_WEXIT);
734 	/*
735 	 * Do a quick check to see if there is anything to do with the
736 	 * proctree_lock held. pgrp and LIST_EMPTY checks are for fixjobc().
737 	 */
738 	PROC_LOCK(p);
739 	if (!SESS_LEADER(p) &&
740 	    (p->p_pgrp == p->p_pptr->p_pgrp) &&
741 	    LIST_EMPTY(&p->p_children)) {
742 		PROC_UNLOCK(p);
743 		return;
744 	}
745 	PROC_UNLOCK(p);
746 
747 	sx_xlock(&proctree_lock);
748 	if (SESS_LEADER(p)) {
749 		sp = p->p_session;
750 
751 		/*
752 		 * s_ttyp is not zero'd; we use this to indicate that
753 		 * the session once had a controlling terminal. (for
754 		 * logging and informational purposes)
755 		 */
756 		SESS_LOCK(sp);
757 		ttyvp = sp->s_ttyvp;
758 		tp = sp->s_ttyp;
759 		sp->s_ttyvp = NULL;
760 		sp->s_ttydp = NULL;
761 		sp->s_leader = NULL;
762 		SESS_UNLOCK(sp);
763 
764 		/*
765 		 * Signal foreground pgrp and revoke access to
766 		 * controlling terminal if it has not been revoked
767 		 * already.
768 		 *
769 		 * Because the TTY may have been revoked in the mean
770 		 * time and could already have a new session associated
771 		 * with it, make sure we don't send a SIGHUP to a
772 		 * foreground process group that does not belong to this
773 		 * session.
774 		 */
775 
776 		if (tp != NULL) {
777 			tty_lock(tp);
778 			if (tp->t_session == sp)
779 				tty_signal_pgrp(tp, SIGHUP);
780 			tty_unlock(tp);
781 		}
782 
783 		if (ttyvp != NULL) {
784 			sx_xunlock(&proctree_lock);
785 			if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) {
786 				VOP_REVOKE(ttyvp, REVOKEALL);
787 				VOP_UNLOCK(ttyvp, 0);
788 			}
789 			vrele(ttyvp);
790 			sx_xlock(&proctree_lock);
791 		}
792 	}
793 	fixjobc(p, p->p_pgrp, 0);
794 	sx_xunlock(&proctree_lock);
795 }
796 
797 /*
798  * A process group has become orphaned;
799  * if there are any stopped processes in the group,
800  * hang-up all process in that group.
801  */
802 static void
803 orphanpg(struct pgrp *pg)
804 {
805 	struct proc *p;
806 
807 	PGRP_LOCK_ASSERT(pg, MA_OWNED);
808 
809 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
810 		PROC_LOCK(p);
811 		if (P_SHOULDSTOP(p) == P_STOPPED_SIG) {
812 			PROC_UNLOCK(p);
813 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
814 				PROC_LOCK(p);
815 				kern_psignal(p, SIGHUP);
816 				kern_psignal(p, SIGCONT);
817 				PROC_UNLOCK(p);
818 			}
819 			return;
820 		}
821 		PROC_UNLOCK(p);
822 	}
823 }
824 
825 void
826 sess_hold(struct session *s)
827 {
828 
829 	refcount_acquire(&s->s_count);
830 }
831 
832 void
833 sess_release(struct session *s)
834 {
835 
836 	if (refcount_release(&s->s_count)) {
837 		if (s->s_ttyp != NULL) {
838 			tty_lock(s->s_ttyp);
839 			tty_rel_sess(s->s_ttyp, s);
840 		}
841 		mtx_destroy(&s->s_mtx);
842 		free(s, M_SESSION);
843 	}
844 }
845 
846 #ifdef DDB
847 
848 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
849 {
850 	struct pgrp *pgrp;
851 	struct proc *p;
852 	int i;
853 
854 	for (i = 0; i <= pgrphash; i++) {
855 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
856 			printf("\tindx %d\n", i);
857 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
858 				printf(
859 			"\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
860 				    (void *)pgrp, (long)pgrp->pg_id,
861 				    (void *)pgrp->pg_session,
862 				    pgrp->pg_session->s_count,
863 				    (void *)LIST_FIRST(&pgrp->pg_members));
864 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
865 					printf("\t\tpid %ld addr %p pgrp %p\n",
866 					    (long)p->p_pid, (void *)p,
867 					    (void *)p->p_pgrp);
868 				}
869 			}
870 		}
871 	}
872 }
873 #endif /* DDB */
874 
875 /*
876  * Calculate the kinfo_proc members which contain process-wide
877  * informations.
878  * Must be called with the target process locked.
879  */
880 static void
881 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
882 {
883 	struct thread *td;
884 
885 	PROC_LOCK_ASSERT(p, MA_OWNED);
886 
887 	kp->ki_estcpu = 0;
888 	kp->ki_pctcpu = 0;
889 	FOREACH_THREAD_IN_PROC(p, td) {
890 		thread_lock(td);
891 		kp->ki_pctcpu += sched_pctcpu(td);
892 		kp->ki_estcpu += sched_estcpu(td);
893 		thread_unlock(td);
894 	}
895 }
896 
897 /*
898  * Clear kinfo_proc and fill in any information that is common
899  * to all threads in the process.
900  * Must be called with the target process locked.
901  */
902 static void
903 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
904 {
905 	struct thread *td0;
906 	struct tty *tp;
907 	struct session *sp;
908 	struct ucred *cred;
909 	struct sigacts *ps;
910 	struct timeval boottime;
911 
912 	/* For proc_realparent. */
913 	sx_assert(&proctree_lock, SX_LOCKED);
914 	PROC_LOCK_ASSERT(p, MA_OWNED);
915 	bzero(kp, sizeof(*kp));
916 
917 	kp->ki_structsize = sizeof(*kp);
918 	kp->ki_paddr = p;
919 	kp->ki_addr =/* p->p_addr; */0; /* XXX */
920 	kp->ki_args = p->p_args;
921 	kp->ki_textvp = p->p_textvp;
922 #ifdef KTRACE
923 	kp->ki_tracep = p->p_tracevp;
924 	kp->ki_traceflag = p->p_traceflag;
925 #endif
926 	kp->ki_fd = p->p_fd;
927 	kp->ki_vmspace = p->p_vmspace;
928 	kp->ki_flag = p->p_flag;
929 	kp->ki_flag2 = p->p_flag2;
930 	cred = p->p_ucred;
931 	if (cred) {
932 		kp->ki_uid = cred->cr_uid;
933 		kp->ki_ruid = cred->cr_ruid;
934 		kp->ki_svuid = cred->cr_svuid;
935 		kp->ki_cr_flags = 0;
936 		if (cred->cr_flags & CRED_FLAG_CAPMODE)
937 			kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
938 		/* XXX bde doesn't like KI_NGROUPS */
939 		if (cred->cr_ngroups > KI_NGROUPS) {
940 			kp->ki_ngroups = KI_NGROUPS;
941 			kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
942 		} else
943 			kp->ki_ngroups = cred->cr_ngroups;
944 		bcopy(cred->cr_groups, kp->ki_groups,
945 		    kp->ki_ngroups * sizeof(gid_t));
946 		kp->ki_rgid = cred->cr_rgid;
947 		kp->ki_svgid = cred->cr_svgid;
948 		/* If jailed(cred), emulate the old P_JAILED flag. */
949 		if (jailed(cred)) {
950 			kp->ki_flag |= P_JAILED;
951 			/* If inside the jail, use 0 as a jail ID. */
952 			if (cred->cr_prison != curthread->td_ucred->cr_prison)
953 				kp->ki_jid = cred->cr_prison->pr_id;
954 		}
955 		strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
956 		    sizeof(kp->ki_loginclass));
957 	}
958 	ps = p->p_sigacts;
959 	if (ps) {
960 		mtx_lock(&ps->ps_mtx);
961 		kp->ki_sigignore = ps->ps_sigignore;
962 		kp->ki_sigcatch = ps->ps_sigcatch;
963 		mtx_unlock(&ps->ps_mtx);
964 	}
965 	if (p->p_state != PRS_NEW &&
966 	    p->p_state != PRS_ZOMBIE &&
967 	    p->p_vmspace != NULL) {
968 		struct vmspace *vm = p->p_vmspace;
969 
970 		kp->ki_size = vm->vm_map.size;
971 		kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
972 		FOREACH_THREAD_IN_PROC(p, td0) {
973 			if (!TD_IS_SWAPPED(td0))
974 				kp->ki_rssize += td0->td_kstack_pages;
975 		}
976 		kp->ki_swrss = vm->vm_swrss;
977 		kp->ki_tsize = vm->vm_tsize;
978 		kp->ki_dsize = vm->vm_dsize;
979 		kp->ki_ssize = vm->vm_ssize;
980 	} else if (p->p_state == PRS_ZOMBIE)
981 		kp->ki_stat = SZOMB;
982 	if (kp->ki_flag & P_INMEM)
983 		kp->ki_sflag = PS_INMEM;
984 	else
985 		kp->ki_sflag = 0;
986 	/* Calculate legacy swtime as seconds since 'swtick'. */
987 	kp->ki_swtime = (ticks - p->p_swtick) / hz;
988 	kp->ki_pid = p->p_pid;
989 	kp->ki_nice = p->p_nice;
990 	kp->ki_fibnum = p->p_fibnum;
991 	kp->ki_start = p->p_stats->p_start;
992 	getboottime(&boottime);
993 	timevaladd(&kp->ki_start, &boottime);
994 	PROC_STATLOCK(p);
995 	rufetch(p, &kp->ki_rusage);
996 	kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
997 	calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
998 	PROC_STATUNLOCK(p);
999 	calccru(p, &kp->ki_childutime, &kp->ki_childstime);
1000 	/* Some callers want child times in a single value. */
1001 	kp->ki_childtime = kp->ki_childstime;
1002 	timevaladd(&kp->ki_childtime, &kp->ki_childutime);
1003 
1004 	FOREACH_THREAD_IN_PROC(p, td0)
1005 		kp->ki_cow += td0->td_cow;
1006 
1007 	tp = NULL;
1008 	if (p->p_pgrp) {
1009 		kp->ki_pgid = p->p_pgrp->pg_id;
1010 		kp->ki_jobc = p->p_pgrp->pg_jobc;
1011 		sp = p->p_pgrp->pg_session;
1012 
1013 		if (sp != NULL) {
1014 			kp->ki_sid = sp->s_sid;
1015 			SESS_LOCK(sp);
1016 			strlcpy(kp->ki_login, sp->s_login,
1017 			    sizeof(kp->ki_login));
1018 			if (sp->s_ttyvp)
1019 				kp->ki_kiflag |= KI_CTTY;
1020 			if (SESS_LEADER(p))
1021 				kp->ki_kiflag |= KI_SLEADER;
1022 			/* XXX proctree_lock */
1023 			tp = sp->s_ttyp;
1024 			SESS_UNLOCK(sp);
1025 		}
1026 	}
1027 	if ((p->p_flag & P_CONTROLT) && tp != NULL) {
1028 		kp->ki_tdev = tty_udev(tp);
1029 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1030 		kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
1031 		if (tp->t_session)
1032 			kp->ki_tsid = tp->t_session->s_sid;
1033 	} else {
1034 		kp->ki_tdev = NODEV;
1035 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1036 	}
1037 	if (p->p_comm[0] != '\0')
1038 		strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
1039 	if (p->p_sysent && p->p_sysent->sv_name != NULL &&
1040 	    p->p_sysent->sv_name[0] != '\0')
1041 		strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
1042 	kp->ki_siglist = p->p_siglist;
1043 	kp->ki_xstat = KW_EXITCODE(p->p_xexit, p->p_xsig);
1044 	kp->ki_acflag = p->p_acflag;
1045 	kp->ki_lock = p->p_lock;
1046 	if (p->p_pptr) {
1047 		kp->ki_ppid = proc_realparent(p)->p_pid;
1048 		if (p->p_flag & P_TRACED)
1049 			kp->ki_tracer = p->p_pptr->p_pid;
1050 	}
1051 }
1052 
1053 /*
1054  * Fill in information that is thread specific.  Must be called with
1055  * target process locked.  If 'preferthread' is set, overwrite certain
1056  * process-related fields that are maintained for both threads and
1057  * processes.
1058  */
1059 static void
1060 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
1061 {
1062 	struct proc *p;
1063 
1064 	p = td->td_proc;
1065 	kp->ki_tdaddr = td;
1066 	PROC_LOCK_ASSERT(p, MA_OWNED);
1067 
1068 	if (preferthread)
1069 		PROC_STATLOCK(p);
1070 	thread_lock(td);
1071 	if (td->td_wmesg != NULL)
1072 		strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
1073 	else
1074 		bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
1075 	if (strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname)) >=
1076 	    sizeof(kp->ki_tdname)) {
1077 		strlcpy(kp->ki_moretdname,
1078 		    td->td_name + sizeof(kp->ki_tdname) - 1,
1079 		    sizeof(kp->ki_moretdname));
1080 	} else {
1081 		bzero(kp->ki_moretdname, sizeof(kp->ki_moretdname));
1082 	}
1083 	if (TD_ON_LOCK(td)) {
1084 		kp->ki_kiflag |= KI_LOCKBLOCK;
1085 		strlcpy(kp->ki_lockname, td->td_lockname,
1086 		    sizeof(kp->ki_lockname));
1087 	} else {
1088 		kp->ki_kiflag &= ~KI_LOCKBLOCK;
1089 		bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
1090 	}
1091 
1092 	if (p->p_state == PRS_NORMAL) { /* approximate. */
1093 		if (TD_ON_RUNQ(td) ||
1094 		    TD_CAN_RUN(td) ||
1095 		    TD_IS_RUNNING(td)) {
1096 			kp->ki_stat = SRUN;
1097 		} else if (P_SHOULDSTOP(p)) {
1098 			kp->ki_stat = SSTOP;
1099 		} else if (TD_IS_SLEEPING(td)) {
1100 			kp->ki_stat = SSLEEP;
1101 		} else if (TD_ON_LOCK(td)) {
1102 			kp->ki_stat = SLOCK;
1103 		} else {
1104 			kp->ki_stat = SWAIT;
1105 		}
1106 	} else if (p->p_state == PRS_ZOMBIE) {
1107 		kp->ki_stat = SZOMB;
1108 	} else {
1109 		kp->ki_stat = SIDL;
1110 	}
1111 
1112 	/* Things in the thread */
1113 	kp->ki_wchan = td->td_wchan;
1114 	kp->ki_pri.pri_level = td->td_priority;
1115 	kp->ki_pri.pri_native = td->td_base_pri;
1116 
1117 	/*
1118 	 * Note: legacy fields; clamp at the old NOCPU value and/or
1119 	 * the maximum u_char CPU value.
1120 	 */
1121 	if (td->td_lastcpu == NOCPU)
1122 		kp->ki_lastcpu_old = NOCPU_OLD;
1123 	else if (td->td_lastcpu > MAXCPU_OLD)
1124 		kp->ki_lastcpu_old = MAXCPU_OLD;
1125 	else
1126 		kp->ki_lastcpu_old = td->td_lastcpu;
1127 
1128 	if (td->td_oncpu == NOCPU)
1129 		kp->ki_oncpu_old = NOCPU_OLD;
1130 	else if (td->td_oncpu > MAXCPU_OLD)
1131 		kp->ki_oncpu_old = MAXCPU_OLD;
1132 	else
1133 		kp->ki_oncpu_old = td->td_oncpu;
1134 
1135 	kp->ki_lastcpu = td->td_lastcpu;
1136 	kp->ki_oncpu = td->td_oncpu;
1137 	kp->ki_tdflags = td->td_flags;
1138 	kp->ki_tid = td->td_tid;
1139 	kp->ki_numthreads = p->p_numthreads;
1140 	kp->ki_pcb = td->td_pcb;
1141 	kp->ki_kstack = (void *)td->td_kstack;
1142 	kp->ki_slptime = (ticks - td->td_slptick) / hz;
1143 	kp->ki_pri.pri_class = td->td_pri_class;
1144 	kp->ki_pri.pri_user = td->td_user_pri;
1145 
1146 	if (preferthread) {
1147 		rufetchtd(td, &kp->ki_rusage);
1148 		kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
1149 		kp->ki_pctcpu = sched_pctcpu(td);
1150 		kp->ki_estcpu = sched_estcpu(td);
1151 		kp->ki_cow = td->td_cow;
1152 	}
1153 
1154 	/* We can't get this anymore but ps etc never used it anyway. */
1155 	kp->ki_rqindex = 0;
1156 
1157 	if (preferthread)
1158 		kp->ki_siglist = td->td_siglist;
1159 	kp->ki_sigmask = td->td_sigmask;
1160 	thread_unlock(td);
1161 	if (preferthread)
1162 		PROC_STATUNLOCK(p);
1163 }
1164 
1165 /*
1166  * Fill in a kinfo_proc structure for the specified process.
1167  * Must be called with the target process locked.
1168  */
1169 void
1170 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1171 {
1172 
1173 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1174 
1175 	fill_kinfo_proc_only(p, kp);
1176 	fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1177 	fill_kinfo_aggregate(p, kp);
1178 }
1179 
1180 struct pstats *
1181 pstats_alloc(void)
1182 {
1183 
1184 	return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1185 }
1186 
1187 /*
1188  * Copy parts of p_stats; zero the rest of p_stats (statistics).
1189  */
1190 void
1191 pstats_fork(struct pstats *src, struct pstats *dst)
1192 {
1193 
1194 	bzero(&dst->pstat_startzero,
1195 	    __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1196 	bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1197 	    __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1198 }
1199 
1200 void
1201 pstats_free(struct pstats *ps)
1202 {
1203 
1204 	free(ps, M_SUBPROC);
1205 }
1206 
1207 static struct proc *
1208 zpfind_locked(pid_t pid)
1209 {
1210 	struct proc *p;
1211 
1212 	sx_assert(&allproc_lock, SX_LOCKED);
1213 	LIST_FOREACH(p, &zombproc, p_list) {
1214 		if (p->p_pid == pid) {
1215 			PROC_LOCK(p);
1216 			break;
1217 		}
1218 	}
1219 	return (p);
1220 }
1221 
1222 /*
1223  * Locate a zombie process by number
1224  */
1225 struct proc *
1226 zpfind(pid_t pid)
1227 {
1228 	struct proc *p;
1229 
1230 	sx_slock(&allproc_lock);
1231 	p = zpfind_locked(pid);
1232 	sx_sunlock(&allproc_lock);
1233 	return (p);
1234 }
1235 
1236 #ifdef COMPAT_FREEBSD32
1237 
1238 /*
1239  * This function is typically used to copy out the kernel address, so
1240  * it can be replaced by assignment of zero.
1241  */
1242 static inline uint32_t
1243 ptr32_trim(void *ptr)
1244 {
1245 	uintptr_t uptr;
1246 
1247 	uptr = (uintptr_t)ptr;
1248 	return ((uptr > UINT_MAX) ? 0 : uptr);
1249 }
1250 
1251 #define PTRTRIM_CP(src,dst,fld) \
1252 	do { (dst).fld = ptr32_trim((src).fld); } while (0)
1253 
1254 static void
1255 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1256 {
1257 	int i;
1258 
1259 	bzero(ki32, sizeof(struct kinfo_proc32));
1260 	ki32->ki_structsize = sizeof(struct kinfo_proc32);
1261 	CP(*ki, *ki32, ki_layout);
1262 	PTRTRIM_CP(*ki, *ki32, ki_args);
1263 	PTRTRIM_CP(*ki, *ki32, ki_paddr);
1264 	PTRTRIM_CP(*ki, *ki32, ki_addr);
1265 	PTRTRIM_CP(*ki, *ki32, ki_tracep);
1266 	PTRTRIM_CP(*ki, *ki32, ki_textvp);
1267 	PTRTRIM_CP(*ki, *ki32, ki_fd);
1268 	PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1269 	PTRTRIM_CP(*ki, *ki32, ki_wchan);
1270 	CP(*ki, *ki32, ki_pid);
1271 	CP(*ki, *ki32, ki_ppid);
1272 	CP(*ki, *ki32, ki_pgid);
1273 	CP(*ki, *ki32, ki_tpgid);
1274 	CP(*ki, *ki32, ki_sid);
1275 	CP(*ki, *ki32, ki_tsid);
1276 	CP(*ki, *ki32, ki_jobc);
1277 	CP(*ki, *ki32, ki_tdev);
1278 	CP(*ki, *ki32, ki_tdev_freebsd11);
1279 	CP(*ki, *ki32, ki_siglist);
1280 	CP(*ki, *ki32, ki_sigmask);
1281 	CP(*ki, *ki32, ki_sigignore);
1282 	CP(*ki, *ki32, ki_sigcatch);
1283 	CP(*ki, *ki32, ki_uid);
1284 	CP(*ki, *ki32, ki_ruid);
1285 	CP(*ki, *ki32, ki_svuid);
1286 	CP(*ki, *ki32, ki_rgid);
1287 	CP(*ki, *ki32, ki_svgid);
1288 	CP(*ki, *ki32, ki_ngroups);
1289 	for (i = 0; i < KI_NGROUPS; i++)
1290 		CP(*ki, *ki32, ki_groups[i]);
1291 	CP(*ki, *ki32, ki_size);
1292 	CP(*ki, *ki32, ki_rssize);
1293 	CP(*ki, *ki32, ki_swrss);
1294 	CP(*ki, *ki32, ki_tsize);
1295 	CP(*ki, *ki32, ki_dsize);
1296 	CP(*ki, *ki32, ki_ssize);
1297 	CP(*ki, *ki32, ki_xstat);
1298 	CP(*ki, *ki32, ki_acflag);
1299 	CP(*ki, *ki32, ki_pctcpu);
1300 	CP(*ki, *ki32, ki_estcpu);
1301 	CP(*ki, *ki32, ki_slptime);
1302 	CP(*ki, *ki32, ki_swtime);
1303 	CP(*ki, *ki32, ki_cow);
1304 	CP(*ki, *ki32, ki_runtime);
1305 	TV_CP(*ki, *ki32, ki_start);
1306 	TV_CP(*ki, *ki32, ki_childtime);
1307 	CP(*ki, *ki32, ki_flag);
1308 	CP(*ki, *ki32, ki_kiflag);
1309 	CP(*ki, *ki32, ki_traceflag);
1310 	CP(*ki, *ki32, ki_stat);
1311 	CP(*ki, *ki32, ki_nice);
1312 	CP(*ki, *ki32, ki_lock);
1313 	CP(*ki, *ki32, ki_rqindex);
1314 	CP(*ki, *ki32, ki_oncpu);
1315 	CP(*ki, *ki32, ki_lastcpu);
1316 
1317 	/* XXX TODO: wrap cpu value as appropriate */
1318 	CP(*ki, *ki32, ki_oncpu_old);
1319 	CP(*ki, *ki32, ki_lastcpu_old);
1320 
1321 	bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1322 	bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1323 	bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1324 	bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1325 	bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1326 	bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1327 	bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1328 	bcopy(ki->ki_moretdname, ki32->ki_moretdname, MAXCOMLEN - TDNAMLEN + 1);
1329 	CP(*ki, *ki32, ki_tracer);
1330 	CP(*ki, *ki32, ki_flag2);
1331 	CP(*ki, *ki32, ki_fibnum);
1332 	CP(*ki, *ki32, ki_cr_flags);
1333 	CP(*ki, *ki32, ki_jid);
1334 	CP(*ki, *ki32, ki_numthreads);
1335 	CP(*ki, *ki32, ki_tid);
1336 	CP(*ki, *ki32, ki_pri);
1337 	freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1338 	freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1339 	PTRTRIM_CP(*ki, *ki32, ki_pcb);
1340 	PTRTRIM_CP(*ki, *ki32, ki_kstack);
1341 	PTRTRIM_CP(*ki, *ki32, ki_udata);
1342 	PTRTRIM_CP(*ki, *ki32, ki_tdaddr);
1343 	CP(*ki, *ki32, ki_sflag);
1344 	CP(*ki, *ki32, ki_tdflags);
1345 }
1346 #endif
1347 
1348 int
1349 kern_proc_out(struct proc *p, struct sbuf *sb, int flags)
1350 {
1351 	struct thread *td;
1352 	struct kinfo_proc ki;
1353 #ifdef COMPAT_FREEBSD32
1354 	struct kinfo_proc32 ki32;
1355 #endif
1356 	int error;
1357 
1358 	PROC_LOCK_ASSERT(p, MA_OWNED);
1359 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1360 
1361 	error = 0;
1362 	fill_kinfo_proc(p, &ki);
1363 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1364 #ifdef COMPAT_FREEBSD32
1365 		if ((flags & KERN_PROC_MASK32) != 0) {
1366 			freebsd32_kinfo_proc_out(&ki, &ki32);
1367 			if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1368 				error = ENOMEM;
1369 		} else
1370 #endif
1371 			if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1372 				error = ENOMEM;
1373 	} else {
1374 		FOREACH_THREAD_IN_PROC(p, td) {
1375 			fill_kinfo_thread(td, &ki, 1);
1376 #ifdef COMPAT_FREEBSD32
1377 			if ((flags & KERN_PROC_MASK32) != 0) {
1378 				freebsd32_kinfo_proc_out(&ki, &ki32);
1379 				if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1380 					error = ENOMEM;
1381 			} else
1382 #endif
1383 				if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1384 					error = ENOMEM;
1385 			if (error != 0)
1386 				break;
1387 		}
1388 	}
1389 	PROC_UNLOCK(p);
1390 	return (error);
1391 }
1392 
1393 static int
1394 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
1395 {
1396 	struct sbuf sb;
1397 	struct kinfo_proc ki;
1398 	int error, error2;
1399 
1400 	sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req);
1401 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1402 	error = kern_proc_out(p, &sb, flags);
1403 	error2 = sbuf_finish(&sb);
1404 	sbuf_delete(&sb);
1405 	if (error != 0)
1406 		return (error);
1407 	else if (error2 != 0)
1408 		return (error2);
1409 	return (0);
1410 }
1411 
1412 static int
1413 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1414 {
1415 	int *name = (int *)arg1;
1416 	u_int namelen = arg2;
1417 	struct proc *p;
1418 	int flags, doingzomb, oid_number;
1419 	int error = 0;
1420 
1421 	oid_number = oidp->oid_number;
1422 	if (oid_number != KERN_PROC_ALL &&
1423 	    (oid_number & KERN_PROC_INC_THREAD) == 0)
1424 		flags = KERN_PROC_NOTHREADS;
1425 	else {
1426 		flags = 0;
1427 		oid_number &= ~KERN_PROC_INC_THREAD;
1428 	}
1429 #ifdef COMPAT_FREEBSD32
1430 	if (req->flags & SCTL_MASK32)
1431 		flags |= KERN_PROC_MASK32;
1432 #endif
1433 	if (oid_number == KERN_PROC_PID) {
1434 		if (namelen != 1)
1435 			return (EINVAL);
1436 		error = sysctl_wire_old_buffer(req, 0);
1437 		if (error)
1438 			return (error);
1439 		sx_slock(&proctree_lock);
1440 		error = pget((pid_t)name[0], PGET_CANSEE, &p);
1441 		if (error == 0)
1442 			error = sysctl_out_proc(p, req, flags);
1443 		sx_sunlock(&proctree_lock);
1444 		return (error);
1445 	}
1446 
1447 	switch (oid_number) {
1448 	case KERN_PROC_ALL:
1449 		if (namelen != 0)
1450 			return (EINVAL);
1451 		break;
1452 	case KERN_PROC_PROC:
1453 		if (namelen != 0 && namelen != 1)
1454 			return (EINVAL);
1455 		break;
1456 	default:
1457 		if (namelen != 1)
1458 			return (EINVAL);
1459 		break;
1460 	}
1461 
1462 	if (!req->oldptr) {
1463 		/* overestimate by 5 procs */
1464 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1465 		if (error)
1466 			return (error);
1467 	}
1468 	error = sysctl_wire_old_buffer(req, 0);
1469 	if (error != 0)
1470 		return (error);
1471 	sx_slock(&proctree_lock);
1472 	sx_slock(&allproc_lock);
1473 	for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) {
1474 		if (!doingzomb)
1475 			p = LIST_FIRST(&allproc);
1476 		else
1477 			p = LIST_FIRST(&zombproc);
1478 		for (; p != NULL; p = LIST_NEXT(p, p_list)) {
1479 			/*
1480 			 * Skip embryonic processes.
1481 			 */
1482 			if (p->p_state == PRS_NEW)
1483 				continue;
1484 			PROC_LOCK(p);
1485 			KASSERT(p->p_ucred != NULL,
1486 			    ("process credential is NULL for non-NEW proc"));
1487 			/*
1488 			 * Show a user only appropriate processes.
1489 			 */
1490 			if (p_cansee(curthread, p)) {
1491 				PROC_UNLOCK(p);
1492 				continue;
1493 			}
1494 			/*
1495 			 * TODO - make more efficient (see notes below).
1496 			 * do by session.
1497 			 */
1498 			switch (oid_number) {
1499 
1500 			case KERN_PROC_GID:
1501 				if (p->p_ucred->cr_gid != (gid_t)name[0]) {
1502 					PROC_UNLOCK(p);
1503 					continue;
1504 				}
1505 				break;
1506 
1507 			case KERN_PROC_PGRP:
1508 				/* could do this by traversing pgrp */
1509 				if (p->p_pgrp == NULL ||
1510 				    p->p_pgrp->pg_id != (pid_t)name[0]) {
1511 					PROC_UNLOCK(p);
1512 					continue;
1513 				}
1514 				break;
1515 
1516 			case KERN_PROC_RGID:
1517 				if (p->p_ucred->cr_rgid != (gid_t)name[0]) {
1518 					PROC_UNLOCK(p);
1519 					continue;
1520 				}
1521 				break;
1522 
1523 			case KERN_PROC_SESSION:
1524 				if (p->p_session == NULL ||
1525 				    p->p_session->s_sid != (pid_t)name[0]) {
1526 					PROC_UNLOCK(p);
1527 					continue;
1528 				}
1529 				break;
1530 
1531 			case KERN_PROC_TTY:
1532 				if ((p->p_flag & P_CONTROLT) == 0 ||
1533 				    p->p_session == NULL) {
1534 					PROC_UNLOCK(p);
1535 					continue;
1536 				}
1537 				/* XXX proctree_lock */
1538 				SESS_LOCK(p->p_session);
1539 				if (p->p_session->s_ttyp == NULL ||
1540 				    tty_udev(p->p_session->s_ttyp) !=
1541 				    (dev_t)name[0]) {
1542 					SESS_UNLOCK(p->p_session);
1543 					PROC_UNLOCK(p);
1544 					continue;
1545 				}
1546 				SESS_UNLOCK(p->p_session);
1547 				break;
1548 
1549 			case KERN_PROC_UID:
1550 				if (p->p_ucred->cr_uid != (uid_t)name[0]) {
1551 					PROC_UNLOCK(p);
1552 					continue;
1553 				}
1554 				break;
1555 
1556 			case KERN_PROC_RUID:
1557 				if (p->p_ucred->cr_ruid != (uid_t)name[0]) {
1558 					PROC_UNLOCK(p);
1559 					continue;
1560 				}
1561 				break;
1562 
1563 			case KERN_PROC_PROC:
1564 				break;
1565 
1566 			default:
1567 				break;
1568 
1569 			}
1570 
1571 			error = sysctl_out_proc(p, req, flags);
1572 			if (error) {
1573 				sx_sunlock(&allproc_lock);
1574 				sx_sunlock(&proctree_lock);
1575 				return (error);
1576 			}
1577 		}
1578 	}
1579 	sx_sunlock(&allproc_lock);
1580 	sx_sunlock(&proctree_lock);
1581 	return (0);
1582 }
1583 
1584 struct pargs *
1585 pargs_alloc(int len)
1586 {
1587 	struct pargs *pa;
1588 
1589 	pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1590 		M_WAITOK);
1591 	refcount_init(&pa->ar_ref, 1);
1592 	pa->ar_length = len;
1593 	return (pa);
1594 }
1595 
1596 static void
1597 pargs_free(struct pargs *pa)
1598 {
1599 
1600 	free(pa, M_PARGS);
1601 }
1602 
1603 void
1604 pargs_hold(struct pargs *pa)
1605 {
1606 
1607 	if (pa == NULL)
1608 		return;
1609 	refcount_acquire(&pa->ar_ref);
1610 }
1611 
1612 void
1613 pargs_drop(struct pargs *pa)
1614 {
1615 
1616 	if (pa == NULL)
1617 		return;
1618 	if (refcount_release(&pa->ar_ref))
1619 		pargs_free(pa);
1620 }
1621 
1622 static int
1623 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1624     size_t len)
1625 {
1626 	ssize_t n;
1627 
1628 	/*
1629 	 * This may return a short read if the string is shorter than the chunk
1630 	 * and is aligned at the end of the page, and the following page is not
1631 	 * mapped.
1632 	 */
1633 	n = proc_readmem(td, p, (vm_offset_t)sptr, buf, len);
1634 	if (n <= 0)
1635 		return (ENOMEM);
1636 	return (0);
1637 }
1638 
1639 #define PROC_AUXV_MAX	256	/* Safety limit on auxv size. */
1640 
1641 enum proc_vector_type {
1642 	PROC_ARG,
1643 	PROC_ENV,
1644 	PROC_AUX,
1645 };
1646 
1647 #ifdef COMPAT_FREEBSD32
1648 static int
1649 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1650     size_t *vsizep, enum proc_vector_type type)
1651 {
1652 	struct freebsd32_ps_strings pss;
1653 	Elf32_Auxinfo aux;
1654 	vm_offset_t vptr, ptr;
1655 	uint32_t *proc_vector32;
1656 	char **proc_vector;
1657 	size_t vsize, size;
1658 	int i, error;
1659 
1660 	error = 0;
1661 	if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss,
1662 	    sizeof(pss)) != sizeof(pss))
1663 		return (ENOMEM);
1664 	switch (type) {
1665 	case PROC_ARG:
1666 		vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1667 		vsize = pss.ps_nargvstr;
1668 		if (vsize > ARG_MAX)
1669 			return (ENOEXEC);
1670 		size = vsize * sizeof(int32_t);
1671 		break;
1672 	case PROC_ENV:
1673 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1674 		vsize = pss.ps_nenvstr;
1675 		if (vsize > ARG_MAX)
1676 			return (ENOEXEC);
1677 		size = vsize * sizeof(int32_t);
1678 		break;
1679 	case PROC_AUX:
1680 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1681 		    (pss.ps_nenvstr + 1) * sizeof(int32_t);
1682 		if (vptr % 4 != 0)
1683 			return (ENOEXEC);
1684 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1685 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1686 			    sizeof(aux))
1687 				return (ENOMEM);
1688 			if (aux.a_type == AT_NULL)
1689 				break;
1690 			ptr += sizeof(aux);
1691 		}
1692 		if (aux.a_type != AT_NULL)
1693 			return (ENOEXEC);
1694 		vsize = i + 1;
1695 		size = vsize * sizeof(aux);
1696 		break;
1697 	default:
1698 		KASSERT(0, ("Wrong proc vector type: %d", type));
1699 		return (EINVAL);
1700 	}
1701 	proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1702 	if (proc_readmem(td, p, vptr, proc_vector32, size) != size) {
1703 		error = ENOMEM;
1704 		goto done;
1705 	}
1706 	if (type == PROC_AUX) {
1707 		*proc_vectorp = (char **)proc_vector32;
1708 		*vsizep = vsize;
1709 		return (0);
1710 	}
1711 	proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1712 	for (i = 0; i < (int)vsize; i++)
1713 		proc_vector[i] = PTRIN(proc_vector32[i]);
1714 	*proc_vectorp = proc_vector;
1715 	*vsizep = vsize;
1716 done:
1717 	free(proc_vector32, M_TEMP);
1718 	return (error);
1719 }
1720 #endif
1721 
1722 static int
1723 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1724     size_t *vsizep, enum proc_vector_type type)
1725 {
1726 	struct ps_strings pss;
1727 	Elf_Auxinfo aux;
1728 	vm_offset_t vptr, ptr;
1729 	char **proc_vector;
1730 	size_t vsize, size;
1731 	int i;
1732 
1733 #ifdef COMPAT_FREEBSD32
1734 	if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1735 		return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1736 #endif
1737 	if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss,
1738 	    sizeof(pss)) != sizeof(pss))
1739 		return (ENOMEM);
1740 	switch (type) {
1741 	case PROC_ARG:
1742 		vptr = (vm_offset_t)pss.ps_argvstr;
1743 		vsize = pss.ps_nargvstr;
1744 		if (vsize > ARG_MAX)
1745 			return (ENOEXEC);
1746 		size = vsize * sizeof(char *);
1747 		break;
1748 	case PROC_ENV:
1749 		vptr = (vm_offset_t)pss.ps_envstr;
1750 		vsize = pss.ps_nenvstr;
1751 		if (vsize > ARG_MAX)
1752 			return (ENOEXEC);
1753 		size = vsize * sizeof(char *);
1754 		break;
1755 	case PROC_AUX:
1756 		/*
1757 		 * The aux array is just above env array on the stack. Check
1758 		 * that the address is naturally aligned.
1759 		 */
1760 		vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1761 		    * sizeof(char *);
1762 #if __ELF_WORD_SIZE == 64
1763 		if (vptr % sizeof(uint64_t) != 0)
1764 #else
1765 		if (vptr % sizeof(uint32_t) != 0)
1766 #endif
1767 			return (ENOEXEC);
1768 		/*
1769 		 * We count the array size reading the aux vectors from the
1770 		 * stack until AT_NULL vector is returned.  So (to keep the code
1771 		 * simple) we read the process stack twice: the first time here
1772 		 * to find the size and the second time when copying the vectors
1773 		 * to the allocated proc_vector.
1774 		 */
1775 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1776 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1777 			    sizeof(aux))
1778 				return (ENOMEM);
1779 			if (aux.a_type == AT_NULL)
1780 				break;
1781 			ptr += sizeof(aux);
1782 		}
1783 		/*
1784 		 * If the PROC_AUXV_MAX entries are iterated over, and we have
1785 		 * not reached AT_NULL, it is most likely we are reading wrong
1786 		 * data: either the process doesn't have auxv array or data has
1787 		 * been modified. Return the error in this case.
1788 		 */
1789 		if (aux.a_type != AT_NULL)
1790 			return (ENOEXEC);
1791 		vsize = i + 1;
1792 		size = vsize * sizeof(aux);
1793 		break;
1794 	default:
1795 		KASSERT(0, ("Wrong proc vector type: %d", type));
1796 		return (EINVAL); /* In case we are built without INVARIANTS. */
1797 	}
1798 	proc_vector = malloc(size, M_TEMP, M_WAITOK);
1799 	if (proc_readmem(td, p, vptr, proc_vector, size) != size) {
1800 		free(proc_vector, M_TEMP);
1801 		return (ENOMEM);
1802 	}
1803 	*proc_vectorp = proc_vector;
1804 	*vsizep = vsize;
1805 
1806 	return (0);
1807 }
1808 
1809 #define GET_PS_STRINGS_CHUNK_SZ	256	/* Chunk size (bytes) for ps_strings operations. */
1810 
1811 static int
1812 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
1813     enum proc_vector_type type)
1814 {
1815 	size_t done, len, nchr, vsize;
1816 	int error, i;
1817 	char **proc_vector, *sptr;
1818 	char pss_string[GET_PS_STRINGS_CHUNK_SZ];
1819 
1820 	PROC_ASSERT_HELD(p);
1821 
1822 	/*
1823 	 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
1824 	 */
1825 	nchr = 2 * (PATH_MAX + ARG_MAX);
1826 
1827 	error = get_proc_vector(td, p, &proc_vector, &vsize, type);
1828 	if (error != 0)
1829 		return (error);
1830 	for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
1831 		/*
1832 		 * The program may have scribbled into its argv array, e.g. to
1833 		 * remove some arguments.  If that has happened, break out
1834 		 * before trying to read from NULL.
1835 		 */
1836 		if (proc_vector[i] == NULL)
1837 			break;
1838 		for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
1839 			error = proc_read_string(td, p, sptr, pss_string,
1840 			    sizeof(pss_string));
1841 			if (error != 0)
1842 				goto done;
1843 			len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
1844 			if (done + len >= nchr)
1845 				len = nchr - done - 1;
1846 			sbuf_bcat(sb, pss_string, len);
1847 			if (len != GET_PS_STRINGS_CHUNK_SZ)
1848 				break;
1849 			done += GET_PS_STRINGS_CHUNK_SZ;
1850 		}
1851 		sbuf_bcat(sb, "", 1);
1852 		done += len + 1;
1853 	}
1854 done:
1855 	free(proc_vector, M_TEMP);
1856 	return (error);
1857 }
1858 
1859 int
1860 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
1861 {
1862 
1863 	return (get_ps_strings(curthread, p, sb, PROC_ARG));
1864 }
1865 
1866 int
1867 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
1868 {
1869 
1870 	return (get_ps_strings(curthread, p, sb, PROC_ENV));
1871 }
1872 
1873 int
1874 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb)
1875 {
1876 	size_t vsize, size;
1877 	char **auxv;
1878 	int error;
1879 
1880 	error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX);
1881 	if (error == 0) {
1882 #ifdef COMPAT_FREEBSD32
1883 		if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1884 			size = vsize * sizeof(Elf32_Auxinfo);
1885 		else
1886 #endif
1887 			size = vsize * sizeof(Elf_Auxinfo);
1888 		if (sbuf_bcat(sb, auxv, size) != 0)
1889 			error = ENOMEM;
1890 		free(auxv, M_TEMP);
1891 	}
1892 	return (error);
1893 }
1894 
1895 /*
1896  * This sysctl allows a process to retrieve the argument list or process
1897  * title for another process without groping around in the address space
1898  * of the other process.  It also allow a process to set its own "process
1899  * title to a string of its own choice.
1900  */
1901 static int
1902 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
1903 {
1904 	int *name = (int *)arg1;
1905 	u_int namelen = arg2;
1906 	struct pargs *newpa, *pa;
1907 	struct proc *p;
1908 	struct sbuf sb;
1909 	int flags, error = 0, error2;
1910 	pid_t pid;
1911 
1912 	if (namelen != 1)
1913 		return (EINVAL);
1914 
1915 	pid = (pid_t)name[0];
1916 	/*
1917 	 * If the query is for this process and it is single-threaded, there
1918 	 * is nobody to modify pargs, thus we can just read.
1919 	 */
1920 	p = curproc;
1921 	if (pid == p->p_pid && p->p_numthreads == 1 && req->newptr == NULL) {
1922 		if ((pa = p->p_args) != NULL)
1923 			error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
1924 		return (error);
1925 	}
1926 
1927 	flags = PGET_CANSEE;
1928 	if (req->newptr != NULL)
1929 		flags |= PGET_ISCURRENT;
1930 	error = pget(pid, flags, &p);
1931 	if (error)
1932 		return (error);
1933 
1934 	pa = p->p_args;
1935 	if (pa != NULL) {
1936 		pargs_hold(pa);
1937 		PROC_UNLOCK(p);
1938 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
1939 		pargs_drop(pa);
1940 	} else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
1941 		_PHOLD(p);
1942 		PROC_UNLOCK(p);
1943 		sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1944 		sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1945 		error = proc_getargv(curthread, p, &sb);
1946 		error2 = sbuf_finish(&sb);
1947 		PRELE(p);
1948 		sbuf_delete(&sb);
1949 		if (error == 0 && error2 != 0)
1950 			error = error2;
1951 	} else {
1952 		PROC_UNLOCK(p);
1953 	}
1954 	if (error != 0 || req->newptr == NULL)
1955 		return (error);
1956 
1957 	if (req->newlen > ps_arg_cache_limit - sizeof(struct pargs))
1958 		return (ENOMEM);
1959 	newpa = pargs_alloc(req->newlen);
1960 	error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
1961 	if (error != 0) {
1962 		pargs_free(newpa);
1963 		return (error);
1964 	}
1965 	PROC_LOCK(p);
1966 	pa = p->p_args;
1967 	p->p_args = newpa;
1968 	PROC_UNLOCK(p);
1969 	pargs_drop(pa);
1970 	return (0);
1971 }
1972 
1973 /*
1974  * This sysctl allows a process to retrieve environment of another process.
1975  */
1976 static int
1977 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
1978 {
1979 	int *name = (int *)arg1;
1980 	u_int namelen = arg2;
1981 	struct proc *p;
1982 	struct sbuf sb;
1983 	int error, error2;
1984 
1985 	if (namelen != 1)
1986 		return (EINVAL);
1987 
1988 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1989 	if (error != 0)
1990 		return (error);
1991 	if ((p->p_flag & P_SYSTEM) != 0) {
1992 		PRELE(p);
1993 		return (0);
1994 	}
1995 
1996 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1997 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1998 	error = proc_getenvv(curthread, p, &sb);
1999 	error2 = sbuf_finish(&sb);
2000 	PRELE(p);
2001 	sbuf_delete(&sb);
2002 	return (error != 0 ? error : error2);
2003 }
2004 
2005 /*
2006  * This sysctl allows a process to retrieve ELF auxiliary vector of
2007  * another process.
2008  */
2009 static int
2010 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
2011 {
2012 	int *name = (int *)arg1;
2013 	u_int namelen = arg2;
2014 	struct proc *p;
2015 	struct sbuf sb;
2016 	int error, error2;
2017 
2018 	if (namelen != 1)
2019 		return (EINVAL);
2020 
2021 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2022 	if (error != 0)
2023 		return (error);
2024 	if ((p->p_flag & P_SYSTEM) != 0) {
2025 		PRELE(p);
2026 		return (0);
2027 	}
2028 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2029 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2030 	error = proc_getauxv(curthread, p, &sb);
2031 	error2 = sbuf_finish(&sb);
2032 	PRELE(p);
2033 	sbuf_delete(&sb);
2034 	return (error != 0 ? error : error2);
2035 }
2036 
2037 /*
2038  * This sysctl allows a process to retrieve the path of the executable for
2039  * itself or another process.
2040  */
2041 static int
2042 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
2043 {
2044 	pid_t *pidp = (pid_t *)arg1;
2045 	unsigned int arglen = arg2;
2046 	struct proc *p;
2047 	struct vnode *vp;
2048 	char *retbuf, *freebuf;
2049 	int error;
2050 
2051 	if (arglen != 1)
2052 		return (EINVAL);
2053 	if (*pidp == -1) {	/* -1 means this process */
2054 		p = req->td->td_proc;
2055 	} else {
2056 		error = pget(*pidp, PGET_CANSEE, &p);
2057 		if (error != 0)
2058 			return (error);
2059 	}
2060 
2061 	vp = p->p_textvp;
2062 	if (vp == NULL) {
2063 		if (*pidp != -1)
2064 			PROC_UNLOCK(p);
2065 		return (0);
2066 	}
2067 	vref(vp);
2068 	if (*pidp != -1)
2069 		PROC_UNLOCK(p);
2070 	error = vn_fullpath(req->td, vp, &retbuf, &freebuf);
2071 	vrele(vp);
2072 	if (error)
2073 		return (error);
2074 	error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
2075 	free(freebuf, M_TEMP);
2076 	return (error);
2077 }
2078 
2079 static int
2080 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
2081 {
2082 	struct proc *p;
2083 	char *sv_name;
2084 	int *name;
2085 	int namelen;
2086 	int error;
2087 
2088 	namelen = arg2;
2089 	if (namelen != 1)
2090 		return (EINVAL);
2091 
2092 	name = (int *)arg1;
2093 	error = pget((pid_t)name[0], PGET_CANSEE, &p);
2094 	if (error != 0)
2095 		return (error);
2096 	sv_name = p->p_sysent->sv_name;
2097 	PROC_UNLOCK(p);
2098 	return (sysctl_handle_string(oidp, sv_name, 0, req));
2099 }
2100 
2101 #ifdef KINFO_OVMENTRY_SIZE
2102 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
2103 #endif
2104 
2105 #ifdef COMPAT_FREEBSD7
2106 static int
2107 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
2108 {
2109 	vm_map_entry_t entry, tmp_entry;
2110 	unsigned int last_timestamp;
2111 	char *fullpath, *freepath;
2112 	struct kinfo_ovmentry *kve;
2113 	struct vattr va;
2114 	struct ucred *cred;
2115 	int error, *name;
2116 	struct vnode *vp;
2117 	struct proc *p;
2118 	vm_map_t map;
2119 	struct vmspace *vm;
2120 
2121 	name = (int *)arg1;
2122 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2123 	if (error != 0)
2124 		return (error);
2125 	vm = vmspace_acquire_ref(p);
2126 	if (vm == NULL) {
2127 		PRELE(p);
2128 		return (ESRCH);
2129 	}
2130 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2131 
2132 	map = &vm->vm_map;
2133 	vm_map_lock_read(map);
2134 	for (entry = map->header.next; entry != &map->header;
2135 	    entry = entry->next) {
2136 		vm_object_t obj, tobj, lobj;
2137 		vm_offset_t addr;
2138 
2139 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2140 			continue;
2141 
2142 		bzero(kve, sizeof(*kve));
2143 		kve->kve_structsize = sizeof(*kve);
2144 
2145 		kve->kve_private_resident = 0;
2146 		obj = entry->object.vm_object;
2147 		if (obj != NULL) {
2148 			VM_OBJECT_RLOCK(obj);
2149 			if (obj->shadow_count == 1)
2150 				kve->kve_private_resident =
2151 				    obj->resident_page_count;
2152 		}
2153 		kve->kve_resident = 0;
2154 		addr = entry->start;
2155 		while (addr < entry->end) {
2156 			if (pmap_extract(map->pmap, addr))
2157 				kve->kve_resident++;
2158 			addr += PAGE_SIZE;
2159 		}
2160 
2161 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2162 			if (tobj != obj)
2163 				VM_OBJECT_RLOCK(tobj);
2164 			if (lobj != obj)
2165 				VM_OBJECT_RUNLOCK(lobj);
2166 			lobj = tobj;
2167 		}
2168 
2169 		kve->kve_start = (void*)entry->start;
2170 		kve->kve_end = (void*)entry->end;
2171 		kve->kve_offset = (off_t)entry->offset;
2172 
2173 		if (entry->protection & VM_PROT_READ)
2174 			kve->kve_protection |= KVME_PROT_READ;
2175 		if (entry->protection & VM_PROT_WRITE)
2176 			kve->kve_protection |= KVME_PROT_WRITE;
2177 		if (entry->protection & VM_PROT_EXECUTE)
2178 			kve->kve_protection |= KVME_PROT_EXEC;
2179 
2180 		if (entry->eflags & MAP_ENTRY_COW)
2181 			kve->kve_flags |= KVME_FLAG_COW;
2182 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2183 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2184 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2185 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2186 
2187 		last_timestamp = map->timestamp;
2188 		vm_map_unlock_read(map);
2189 
2190 		kve->kve_fileid = 0;
2191 		kve->kve_fsid = 0;
2192 		freepath = NULL;
2193 		fullpath = "";
2194 		if (lobj) {
2195 			vp = NULL;
2196 			switch (lobj->type) {
2197 			case OBJT_DEFAULT:
2198 				kve->kve_type = KVME_TYPE_DEFAULT;
2199 				break;
2200 			case OBJT_VNODE:
2201 				kve->kve_type = KVME_TYPE_VNODE;
2202 				vp = lobj->handle;
2203 				vref(vp);
2204 				break;
2205 			case OBJT_SWAP:
2206 				if ((lobj->flags & OBJ_TMPFS_NODE) != 0) {
2207 					kve->kve_type = KVME_TYPE_VNODE;
2208 					if ((lobj->flags & OBJ_TMPFS) != 0) {
2209 						vp = lobj->un_pager.swp.swp_tmpfs;
2210 						vref(vp);
2211 					}
2212 				} else {
2213 					kve->kve_type = KVME_TYPE_SWAP;
2214 				}
2215 				break;
2216 			case OBJT_DEVICE:
2217 				kve->kve_type = KVME_TYPE_DEVICE;
2218 				break;
2219 			case OBJT_PHYS:
2220 				kve->kve_type = KVME_TYPE_PHYS;
2221 				break;
2222 			case OBJT_DEAD:
2223 				kve->kve_type = KVME_TYPE_DEAD;
2224 				break;
2225 			case OBJT_SG:
2226 				kve->kve_type = KVME_TYPE_SG;
2227 				break;
2228 			default:
2229 				kve->kve_type = KVME_TYPE_UNKNOWN;
2230 				break;
2231 			}
2232 			if (lobj != obj)
2233 				VM_OBJECT_RUNLOCK(lobj);
2234 
2235 			kve->kve_ref_count = obj->ref_count;
2236 			kve->kve_shadow_count = obj->shadow_count;
2237 			VM_OBJECT_RUNLOCK(obj);
2238 			if (vp != NULL) {
2239 				vn_fullpath(curthread, vp, &fullpath,
2240 				    &freepath);
2241 				cred = curthread->td_ucred;
2242 				vn_lock(vp, LK_SHARED | LK_RETRY);
2243 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2244 					kve->kve_fileid = va.va_fileid;
2245 					/* truncate */
2246 					kve->kve_fsid = va.va_fsid;
2247 				}
2248 				vput(vp);
2249 			}
2250 		} else {
2251 			kve->kve_type = KVME_TYPE_NONE;
2252 			kve->kve_ref_count = 0;
2253 			kve->kve_shadow_count = 0;
2254 		}
2255 
2256 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2257 		if (freepath != NULL)
2258 			free(freepath, M_TEMP);
2259 
2260 		error = SYSCTL_OUT(req, kve, sizeof(*kve));
2261 		vm_map_lock_read(map);
2262 		if (error)
2263 			break;
2264 		if (last_timestamp != map->timestamp) {
2265 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2266 			entry = tmp_entry;
2267 		}
2268 	}
2269 	vm_map_unlock_read(map);
2270 	vmspace_free(vm);
2271 	PRELE(p);
2272 	free(kve, M_TEMP);
2273 	return (error);
2274 }
2275 #endif	/* COMPAT_FREEBSD7 */
2276 
2277 #ifdef KINFO_VMENTRY_SIZE
2278 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2279 #endif
2280 
2281 static void
2282 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry,
2283     struct kinfo_vmentry *kve)
2284 {
2285 	vm_object_t obj, tobj;
2286 	vm_page_t m, m_adv;
2287 	vm_offset_t addr;
2288 	vm_paddr_t locked_pa;
2289 	vm_pindex_t pi, pi_adv, pindex;
2290 
2291 	locked_pa = 0;
2292 	obj = entry->object.vm_object;
2293 	addr = entry->start;
2294 	m_adv = NULL;
2295 	pi = OFF_TO_IDX(entry->offset);
2296 	for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) {
2297 		if (m_adv != NULL) {
2298 			m = m_adv;
2299 		} else {
2300 			pi_adv = atop(entry->end - addr);
2301 			pindex = pi;
2302 			for (tobj = obj;; tobj = tobj->backing_object) {
2303 				m = vm_page_find_least(tobj, pindex);
2304 				if (m != NULL) {
2305 					if (m->pindex == pindex)
2306 						break;
2307 					if (pi_adv > m->pindex - pindex) {
2308 						pi_adv = m->pindex - pindex;
2309 						m_adv = m;
2310 					}
2311 				}
2312 				if (tobj->backing_object == NULL)
2313 					goto next;
2314 				pindex += OFF_TO_IDX(tobj->
2315 				    backing_object_offset);
2316 			}
2317 		}
2318 		m_adv = NULL;
2319 		if (m->psind != 0 && addr + pagesizes[1] <= entry->end &&
2320 		    (addr & (pagesizes[1] - 1)) == 0 &&
2321 		    (pmap_mincore(map->pmap, addr, &locked_pa) &
2322 		    MINCORE_SUPER) != 0) {
2323 			kve->kve_flags |= KVME_FLAG_SUPER;
2324 			pi_adv = atop(pagesizes[1]);
2325 		} else {
2326 			/*
2327 			 * We do not test the found page on validity.
2328 			 * Either the page is busy and being paged in,
2329 			 * or it was invalidated.  The first case
2330 			 * should be counted as resident, the second
2331 			 * is not so clear; we do account both.
2332 			 */
2333 			pi_adv = 1;
2334 		}
2335 		kve->kve_resident += pi_adv;
2336 next:;
2337 	}
2338 	PA_UNLOCK_COND(locked_pa);
2339 }
2340 
2341 /*
2342  * Must be called with the process locked and will return unlocked.
2343  */
2344 int
2345 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb, ssize_t maxlen, int flags)
2346 {
2347 	vm_map_entry_t entry, tmp_entry;
2348 	struct vattr va;
2349 	vm_map_t map;
2350 	vm_object_t obj, tobj, lobj;
2351 	char *fullpath, *freepath;
2352 	struct kinfo_vmentry *kve;
2353 	struct ucred *cred;
2354 	struct vnode *vp;
2355 	struct vmspace *vm;
2356 	vm_offset_t addr;
2357 	unsigned int last_timestamp;
2358 	int error;
2359 
2360 	PROC_LOCK_ASSERT(p, MA_OWNED);
2361 
2362 	_PHOLD(p);
2363 	PROC_UNLOCK(p);
2364 	vm = vmspace_acquire_ref(p);
2365 	if (vm == NULL) {
2366 		PRELE(p);
2367 		return (ESRCH);
2368 	}
2369 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK | M_ZERO);
2370 
2371 	error = 0;
2372 	map = &vm->vm_map;
2373 	vm_map_lock_read(map);
2374 	for (entry = map->header.next; entry != &map->header;
2375 	    entry = entry->next) {
2376 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2377 			continue;
2378 
2379 		addr = entry->end;
2380 		bzero(kve, sizeof(*kve));
2381 		obj = entry->object.vm_object;
2382 		if (obj != NULL) {
2383 			for (tobj = obj; tobj != NULL;
2384 			    tobj = tobj->backing_object) {
2385 				VM_OBJECT_RLOCK(tobj);
2386 				lobj = tobj;
2387 			}
2388 			if (obj->backing_object == NULL)
2389 				kve->kve_private_resident =
2390 				    obj->resident_page_count;
2391 			if (!vmmap_skip_res_cnt)
2392 				kern_proc_vmmap_resident(map, entry, kve);
2393 			for (tobj = obj; tobj != NULL;
2394 			    tobj = tobj->backing_object) {
2395 				if (tobj != obj && tobj != lobj)
2396 					VM_OBJECT_RUNLOCK(tobj);
2397 			}
2398 		} else {
2399 			lobj = NULL;
2400 		}
2401 
2402 		kve->kve_start = entry->start;
2403 		kve->kve_end = entry->end;
2404 		kve->kve_offset = entry->offset;
2405 
2406 		if (entry->protection & VM_PROT_READ)
2407 			kve->kve_protection |= KVME_PROT_READ;
2408 		if (entry->protection & VM_PROT_WRITE)
2409 			kve->kve_protection |= KVME_PROT_WRITE;
2410 		if (entry->protection & VM_PROT_EXECUTE)
2411 			kve->kve_protection |= KVME_PROT_EXEC;
2412 
2413 		if (entry->eflags & MAP_ENTRY_COW)
2414 			kve->kve_flags |= KVME_FLAG_COW;
2415 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2416 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2417 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2418 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2419 		if (entry->eflags & MAP_ENTRY_GROWS_UP)
2420 			kve->kve_flags |= KVME_FLAG_GROWS_UP;
2421 		if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2422 			kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2423 
2424 		last_timestamp = map->timestamp;
2425 		vm_map_unlock_read(map);
2426 
2427 		freepath = NULL;
2428 		fullpath = "";
2429 		if (lobj != NULL) {
2430 			vp = NULL;
2431 			switch (lobj->type) {
2432 			case OBJT_DEFAULT:
2433 				kve->kve_type = KVME_TYPE_DEFAULT;
2434 				break;
2435 			case OBJT_VNODE:
2436 				kve->kve_type = KVME_TYPE_VNODE;
2437 				vp = lobj->handle;
2438 				vref(vp);
2439 				break;
2440 			case OBJT_SWAP:
2441 				if ((lobj->flags & OBJ_TMPFS_NODE) != 0) {
2442 					kve->kve_type = KVME_TYPE_VNODE;
2443 					if ((lobj->flags & OBJ_TMPFS) != 0) {
2444 						vp = lobj->un_pager.swp.swp_tmpfs;
2445 						vref(vp);
2446 					}
2447 				} else {
2448 					kve->kve_type = KVME_TYPE_SWAP;
2449 				}
2450 				break;
2451 			case OBJT_DEVICE:
2452 				kve->kve_type = KVME_TYPE_DEVICE;
2453 				break;
2454 			case OBJT_PHYS:
2455 				kve->kve_type = KVME_TYPE_PHYS;
2456 				break;
2457 			case OBJT_DEAD:
2458 				kve->kve_type = KVME_TYPE_DEAD;
2459 				break;
2460 			case OBJT_SG:
2461 				kve->kve_type = KVME_TYPE_SG;
2462 				break;
2463 			case OBJT_MGTDEVICE:
2464 				kve->kve_type = KVME_TYPE_MGTDEVICE;
2465 				break;
2466 			default:
2467 				kve->kve_type = KVME_TYPE_UNKNOWN;
2468 				break;
2469 			}
2470 			if (lobj != obj)
2471 				VM_OBJECT_RUNLOCK(lobj);
2472 
2473 			kve->kve_ref_count = obj->ref_count;
2474 			kve->kve_shadow_count = obj->shadow_count;
2475 			VM_OBJECT_RUNLOCK(obj);
2476 			if (vp != NULL) {
2477 				vn_fullpath(curthread, vp, &fullpath,
2478 				    &freepath);
2479 				kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2480 				cred = curthread->td_ucred;
2481 				vn_lock(vp, LK_SHARED | LK_RETRY);
2482 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2483 					kve->kve_vn_fileid = va.va_fileid;
2484 					kve->kve_vn_fsid = va.va_fsid;
2485 					kve->kve_vn_fsid_freebsd11 =
2486 					    kve->kve_vn_fsid; /* truncate */
2487 					kve->kve_vn_mode =
2488 					    MAKEIMODE(va.va_type, va.va_mode);
2489 					kve->kve_vn_size = va.va_size;
2490 					kve->kve_vn_rdev = va.va_rdev;
2491 					kve->kve_vn_rdev_freebsd11 =
2492 					    kve->kve_vn_rdev; /* truncate */
2493 					kve->kve_status = KF_ATTR_VALID;
2494 				}
2495 				vput(vp);
2496 			}
2497 		} else {
2498 			kve->kve_type = KVME_TYPE_NONE;
2499 			kve->kve_ref_count = 0;
2500 			kve->kve_shadow_count = 0;
2501 		}
2502 
2503 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2504 		if (freepath != NULL)
2505 			free(freepath, M_TEMP);
2506 
2507 		/* Pack record size down */
2508 		if ((flags & KERN_VMMAP_PACK_KINFO) != 0)
2509 			kve->kve_structsize =
2510 			    offsetof(struct kinfo_vmentry, kve_path) +
2511 			    strlen(kve->kve_path) + 1;
2512 		else
2513 			kve->kve_structsize = sizeof(*kve);
2514 		kve->kve_structsize = roundup(kve->kve_structsize,
2515 		    sizeof(uint64_t));
2516 
2517 		/* Halt filling and truncate rather than exceeding maxlen */
2518 		if (maxlen != -1 && maxlen < kve->kve_structsize) {
2519 			error = 0;
2520 			vm_map_lock_read(map);
2521 			break;
2522 		} else if (maxlen != -1)
2523 			maxlen -= kve->kve_structsize;
2524 
2525 		if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0)
2526 			error = ENOMEM;
2527 		vm_map_lock_read(map);
2528 		if (error != 0)
2529 			break;
2530 		if (last_timestamp != map->timestamp) {
2531 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2532 			entry = tmp_entry;
2533 		}
2534 	}
2535 	vm_map_unlock_read(map);
2536 	vmspace_free(vm);
2537 	PRELE(p);
2538 	free(kve, M_TEMP);
2539 	return (error);
2540 }
2541 
2542 static int
2543 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2544 {
2545 	struct proc *p;
2546 	struct sbuf sb;
2547 	int error, error2, *name;
2548 
2549 	name = (int *)arg1;
2550 	sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req);
2551 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2552 	error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
2553 	if (error != 0) {
2554 		sbuf_delete(&sb);
2555 		return (error);
2556 	}
2557 	error = kern_proc_vmmap_out(p, &sb, -1, KERN_VMMAP_PACK_KINFO);
2558 	error2 = sbuf_finish(&sb);
2559 	sbuf_delete(&sb);
2560 	return (error != 0 ? error : error2);
2561 }
2562 
2563 #if defined(STACK) || defined(DDB)
2564 static int
2565 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2566 {
2567 	struct kinfo_kstack *kkstp;
2568 	int error, i, *name, numthreads;
2569 	lwpid_t *lwpidarray;
2570 	struct thread *td;
2571 	struct stack *st;
2572 	struct sbuf sb;
2573 	struct proc *p;
2574 
2575 	name = (int *)arg1;
2576 	error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2577 	if (error != 0)
2578 		return (error);
2579 
2580 	kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2581 	st = stack_create(M_WAITOK);
2582 
2583 	lwpidarray = NULL;
2584 	PROC_LOCK(p);
2585 	do {
2586 		if (lwpidarray != NULL) {
2587 			free(lwpidarray, M_TEMP);
2588 			lwpidarray = NULL;
2589 		}
2590 		numthreads = p->p_numthreads;
2591 		PROC_UNLOCK(p);
2592 		lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2593 		    M_WAITOK | M_ZERO);
2594 		PROC_LOCK(p);
2595 	} while (numthreads < p->p_numthreads);
2596 
2597 	/*
2598 	 * XXXRW: During the below loop, execve(2) and countless other sorts
2599 	 * of changes could have taken place.  Should we check to see if the
2600 	 * vmspace has been replaced, or the like, in order to prevent
2601 	 * giving a snapshot that spans, say, execve(2), with some threads
2602 	 * before and some after?  Among other things, the credentials could
2603 	 * have changed, in which case the right to extract debug info might
2604 	 * no longer be assured.
2605 	 */
2606 	i = 0;
2607 	FOREACH_THREAD_IN_PROC(p, td) {
2608 		KASSERT(i < numthreads,
2609 		    ("sysctl_kern_proc_kstack: numthreads"));
2610 		lwpidarray[i] = td->td_tid;
2611 		i++;
2612 	}
2613 	numthreads = i;
2614 	for (i = 0; i < numthreads; i++) {
2615 		td = thread_find(p, lwpidarray[i]);
2616 		if (td == NULL) {
2617 			continue;
2618 		}
2619 		bzero(kkstp, sizeof(*kkstp));
2620 		(void)sbuf_new(&sb, kkstp->kkst_trace,
2621 		    sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2622 		thread_lock(td);
2623 		kkstp->kkst_tid = td->td_tid;
2624 		if (TD_IS_SWAPPED(td)) {
2625 			kkstp->kkst_state = KKST_STATE_SWAPPED;
2626 		} else if (TD_IS_RUNNING(td)) {
2627 			if (stack_save_td_running(st, td) == 0)
2628 				kkstp->kkst_state = KKST_STATE_STACKOK;
2629 			else
2630 				kkstp->kkst_state = KKST_STATE_RUNNING;
2631 		} else {
2632 			kkstp->kkst_state = KKST_STATE_STACKOK;
2633 			stack_save_td(st, td);
2634 		}
2635 		thread_unlock(td);
2636 		PROC_UNLOCK(p);
2637 		stack_sbuf_print(&sb, st);
2638 		sbuf_finish(&sb);
2639 		sbuf_delete(&sb);
2640 		error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2641 		PROC_LOCK(p);
2642 		if (error)
2643 			break;
2644 	}
2645 	_PRELE(p);
2646 	PROC_UNLOCK(p);
2647 	if (lwpidarray != NULL)
2648 		free(lwpidarray, M_TEMP);
2649 	stack_destroy(st);
2650 	free(kkstp, M_TEMP);
2651 	return (error);
2652 }
2653 #endif
2654 
2655 /*
2656  * This sysctl allows a process to retrieve the full list of groups from
2657  * itself or another process.
2658  */
2659 static int
2660 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2661 {
2662 	pid_t *pidp = (pid_t *)arg1;
2663 	unsigned int arglen = arg2;
2664 	struct proc *p;
2665 	struct ucred *cred;
2666 	int error;
2667 
2668 	if (arglen != 1)
2669 		return (EINVAL);
2670 	if (*pidp == -1) {	/* -1 means this process */
2671 		p = req->td->td_proc;
2672 		PROC_LOCK(p);
2673 	} else {
2674 		error = pget(*pidp, PGET_CANSEE, &p);
2675 		if (error != 0)
2676 			return (error);
2677 	}
2678 
2679 	cred = crhold(p->p_ucred);
2680 	PROC_UNLOCK(p);
2681 
2682 	error = SYSCTL_OUT(req, cred->cr_groups,
2683 	    cred->cr_ngroups * sizeof(gid_t));
2684 	crfree(cred);
2685 	return (error);
2686 }
2687 
2688 /*
2689  * This sysctl allows a process to retrieve or/and set the resource limit for
2690  * another process.
2691  */
2692 static int
2693 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2694 {
2695 	int *name = (int *)arg1;
2696 	u_int namelen = arg2;
2697 	struct rlimit rlim;
2698 	struct proc *p;
2699 	u_int which;
2700 	int flags, error;
2701 
2702 	if (namelen != 2)
2703 		return (EINVAL);
2704 
2705 	which = (u_int)name[1];
2706 	if (which >= RLIM_NLIMITS)
2707 		return (EINVAL);
2708 
2709 	if (req->newptr != NULL && req->newlen != sizeof(rlim))
2710 		return (EINVAL);
2711 
2712 	flags = PGET_HOLD | PGET_NOTWEXIT;
2713 	if (req->newptr != NULL)
2714 		flags |= PGET_CANDEBUG;
2715 	else
2716 		flags |= PGET_CANSEE;
2717 	error = pget((pid_t)name[0], flags, &p);
2718 	if (error != 0)
2719 		return (error);
2720 
2721 	/*
2722 	 * Retrieve limit.
2723 	 */
2724 	if (req->oldptr != NULL) {
2725 		PROC_LOCK(p);
2726 		lim_rlimit_proc(p, which, &rlim);
2727 		PROC_UNLOCK(p);
2728 	}
2729 	error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
2730 	if (error != 0)
2731 		goto errout;
2732 
2733 	/*
2734 	 * Set limit.
2735 	 */
2736 	if (req->newptr != NULL) {
2737 		error = SYSCTL_IN(req, &rlim, sizeof(rlim));
2738 		if (error == 0)
2739 			error = kern_proc_setrlimit(curthread, p, which, &rlim);
2740 	}
2741 
2742 errout:
2743 	PRELE(p);
2744 	return (error);
2745 }
2746 
2747 /*
2748  * This sysctl allows a process to retrieve ps_strings structure location of
2749  * another process.
2750  */
2751 static int
2752 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
2753 {
2754 	int *name = (int *)arg1;
2755 	u_int namelen = arg2;
2756 	struct proc *p;
2757 	vm_offset_t ps_strings;
2758 	int error;
2759 #ifdef COMPAT_FREEBSD32
2760 	uint32_t ps_strings32;
2761 #endif
2762 
2763 	if (namelen != 1)
2764 		return (EINVAL);
2765 
2766 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2767 	if (error != 0)
2768 		return (error);
2769 #ifdef COMPAT_FREEBSD32
2770 	if ((req->flags & SCTL_MASK32) != 0) {
2771 		/*
2772 		 * We return 0 if the 32 bit emulation request is for a 64 bit
2773 		 * process.
2774 		 */
2775 		ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
2776 		    PTROUT(p->p_sysent->sv_psstrings) : 0;
2777 		PROC_UNLOCK(p);
2778 		error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
2779 		return (error);
2780 	}
2781 #endif
2782 	ps_strings = p->p_sysent->sv_psstrings;
2783 	PROC_UNLOCK(p);
2784 	error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
2785 	return (error);
2786 }
2787 
2788 /*
2789  * This sysctl allows a process to retrieve umask of another process.
2790  */
2791 static int
2792 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
2793 {
2794 	int *name = (int *)arg1;
2795 	u_int namelen = arg2;
2796 	struct proc *p;
2797 	int error;
2798 	u_short fd_cmask;
2799 	pid_t pid;
2800 
2801 	if (namelen != 1)
2802 		return (EINVAL);
2803 
2804 	pid = (pid_t)name[0];
2805 	p = curproc;
2806 	if (pid == p->p_pid || pid == 0) {
2807 		fd_cmask = p->p_fd->fd_cmask;
2808 		goto out;
2809 	}
2810 
2811 	error = pget(pid, PGET_WANTREAD, &p);
2812 	if (error != 0)
2813 		return (error);
2814 
2815 	fd_cmask = p->p_fd->fd_cmask;
2816 	PRELE(p);
2817 out:
2818 	error = SYSCTL_OUT(req, &fd_cmask, sizeof(fd_cmask));
2819 	return (error);
2820 }
2821 
2822 /*
2823  * This sysctl allows a process to set and retrieve binary osreldate of
2824  * another process.
2825  */
2826 static int
2827 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
2828 {
2829 	int *name = (int *)arg1;
2830 	u_int namelen = arg2;
2831 	struct proc *p;
2832 	int flags, error, osrel;
2833 
2834 	if (namelen != 1)
2835 		return (EINVAL);
2836 
2837 	if (req->newptr != NULL && req->newlen != sizeof(osrel))
2838 		return (EINVAL);
2839 
2840 	flags = PGET_HOLD | PGET_NOTWEXIT;
2841 	if (req->newptr != NULL)
2842 		flags |= PGET_CANDEBUG;
2843 	else
2844 		flags |= PGET_CANSEE;
2845 	error = pget((pid_t)name[0], flags, &p);
2846 	if (error != 0)
2847 		return (error);
2848 
2849 	error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
2850 	if (error != 0)
2851 		goto errout;
2852 
2853 	if (req->newptr != NULL) {
2854 		error = SYSCTL_IN(req, &osrel, sizeof(osrel));
2855 		if (error != 0)
2856 			goto errout;
2857 		if (osrel < 0) {
2858 			error = EINVAL;
2859 			goto errout;
2860 		}
2861 		p->p_osrel = osrel;
2862 	}
2863 errout:
2864 	PRELE(p);
2865 	return (error);
2866 }
2867 
2868 static int
2869 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS)
2870 {
2871 	int *name = (int *)arg1;
2872 	u_int namelen = arg2;
2873 	struct proc *p;
2874 	struct kinfo_sigtramp kst;
2875 	const struct sysentvec *sv;
2876 	int error;
2877 #ifdef COMPAT_FREEBSD32
2878 	struct kinfo_sigtramp32 kst32;
2879 #endif
2880 
2881 	if (namelen != 1)
2882 		return (EINVAL);
2883 
2884 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2885 	if (error != 0)
2886 		return (error);
2887 	sv = p->p_sysent;
2888 #ifdef COMPAT_FREEBSD32
2889 	if ((req->flags & SCTL_MASK32) != 0) {
2890 		bzero(&kst32, sizeof(kst32));
2891 		if (SV_PROC_FLAG(p, SV_ILP32)) {
2892 			if (sv->sv_sigcode_base != 0) {
2893 				kst32.ksigtramp_start = sv->sv_sigcode_base;
2894 				kst32.ksigtramp_end = sv->sv_sigcode_base +
2895 				    *sv->sv_szsigcode;
2896 			} else {
2897 				kst32.ksigtramp_start = sv->sv_psstrings -
2898 				    *sv->sv_szsigcode;
2899 				kst32.ksigtramp_end = sv->sv_psstrings;
2900 			}
2901 		}
2902 		PROC_UNLOCK(p);
2903 		error = SYSCTL_OUT(req, &kst32, sizeof(kst32));
2904 		return (error);
2905 	}
2906 #endif
2907 	bzero(&kst, sizeof(kst));
2908 	if (sv->sv_sigcode_base != 0) {
2909 		kst.ksigtramp_start = (char *)sv->sv_sigcode_base;
2910 		kst.ksigtramp_end = (char *)sv->sv_sigcode_base +
2911 		    *sv->sv_szsigcode;
2912 	} else {
2913 		kst.ksigtramp_start = (char *)sv->sv_psstrings -
2914 		    *sv->sv_szsigcode;
2915 		kst.ksigtramp_end = (char *)sv->sv_psstrings;
2916 	}
2917 	PROC_UNLOCK(p);
2918 	error = SYSCTL_OUT(req, &kst, sizeof(kst));
2919 	return (error);
2920 }
2921 
2922 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD,  0, "Process table");
2923 
2924 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
2925 	CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
2926 	"Return entire process table");
2927 
2928 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2929 	sysctl_kern_proc, "Process table");
2930 
2931 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
2932 	sysctl_kern_proc, "Process table");
2933 
2934 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2935 	sysctl_kern_proc, "Process table");
2936 
2937 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
2938 	CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2939 
2940 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
2941 	sysctl_kern_proc, "Process table");
2942 
2943 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2944 	sysctl_kern_proc, "Process table");
2945 
2946 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2947 	sysctl_kern_proc, "Process table");
2948 
2949 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2950 	sysctl_kern_proc, "Process table");
2951 
2952 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
2953 	sysctl_kern_proc, "Return process table, no threads");
2954 
2955 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
2956 	CTLFLAG_RW | CTLFLAG_CAPWR | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
2957 	sysctl_kern_proc_args, "Process argument list");
2958 
2959 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
2960 	sysctl_kern_proc_env, "Process environment");
2961 
2962 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
2963 	CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
2964 
2965 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
2966 	CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
2967 
2968 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
2969 	CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
2970 	"Process syscall vector name (ABI type)");
2971 
2972 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
2973 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2974 
2975 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
2976 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2977 
2978 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
2979 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2980 
2981 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
2982 	sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2983 
2984 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
2985 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2986 
2987 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
2988 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2989 
2990 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
2991 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2992 
2993 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
2994 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2995 
2996 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
2997 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
2998 	"Return process table, no threads");
2999 
3000 #ifdef COMPAT_FREEBSD7
3001 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
3002 	CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
3003 #endif
3004 
3005 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
3006 	CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
3007 
3008 #if defined(STACK) || defined(DDB)
3009 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
3010 	CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
3011 #endif
3012 
3013 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
3014 	CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
3015 
3016 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
3017 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
3018 	"Process resource limits");
3019 
3020 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
3021 	CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
3022 	"Process ps_strings location");
3023 
3024 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
3025 	CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
3026 
3027 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
3028 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
3029 	"Process binary osreldate");
3030 
3031 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD |
3032 	CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp,
3033 	"Process signal trampoline location");
3034 
3035 int allproc_gen;
3036 
3037 /*
3038  * stop_all_proc() purpose is to stop all process which have usermode,
3039  * except current process for obvious reasons.  This makes it somewhat
3040  * unreliable when invoked from multithreaded process.  The service
3041  * must not be user-callable anyway.
3042  */
3043 void
3044 stop_all_proc(void)
3045 {
3046 	struct proc *cp, *p;
3047 	int r, gen;
3048 	bool restart, seen_stopped, seen_exiting, stopped_some;
3049 
3050 	cp = curproc;
3051 allproc_loop:
3052 	sx_xlock(&allproc_lock);
3053 	gen = allproc_gen;
3054 	seen_exiting = seen_stopped = stopped_some = restart = false;
3055 	LIST_REMOVE(cp, p_list);
3056 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3057 	for (;;) {
3058 		p = LIST_NEXT(cp, p_list);
3059 		if (p == NULL)
3060 			break;
3061 		LIST_REMOVE(cp, p_list);
3062 		LIST_INSERT_AFTER(p, cp, p_list);
3063 		PROC_LOCK(p);
3064 		if ((p->p_flag & (P_KPROC | P_SYSTEM | P_TOTAL_STOP)) != 0) {
3065 			PROC_UNLOCK(p);
3066 			continue;
3067 		}
3068 		if ((p->p_flag & P_WEXIT) != 0) {
3069 			seen_exiting = true;
3070 			PROC_UNLOCK(p);
3071 			continue;
3072 		}
3073 		if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) {
3074 			/*
3075 			 * Stopped processes are tolerated when there
3076 			 * are no other processes which might continue
3077 			 * them.  P_STOPPED_SINGLE but not
3078 			 * P_TOTAL_STOP process still has at least one
3079 			 * thread running.
3080 			 */
3081 			seen_stopped = true;
3082 			PROC_UNLOCK(p);
3083 			continue;
3084 		}
3085 		_PHOLD(p);
3086 		sx_xunlock(&allproc_lock);
3087 		r = thread_single(p, SINGLE_ALLPROC);
3088 		if (r != 0)
3089 			restart = true;
3090 		else
3091 			stopped_some = true;
3092 		_PRELE(p);
3093 		PROC_UNLOCK(p);
3094 		sx_xlock(&allproc_lock);
3095 	}
3096 	/* Catch forked children we did not see in iteration. */
3097 	if (gen != allproc_gen)
3098 		restart = true;
3099 	sx_xunlock(&allproc_lock);
3100 	if (restart || stopped_some || seen_exiting || seen_stopped) {
3101 		kern_yield(PRI_USER);
3102 		goto allproc_loop;
3103 	}
3104 }
3105 
3106 void
3107 resume_all_proc(void)
3108 {
3109 	struct proc *cp, *p;
3110 
3111 	cp = curproc;
3112 	sx_xlock(&allproc_lock);
3113 	LIST_REMOVE(cp, p_list);
3114 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3115 	for (;;) {
3116 		p = LIST_NEXT(cp, p_list);
3117 		if (p == NULL)
3118 			break;
3119 		LIST_REMOVE(cp, p_list);
3120 		LIST_INSERT_AFTER(p, cp, p_list);
3121 		PROC_LOCK(p);
3122 		if ((p->p_flag & P_TOTAL_STOP) != 0) {
3123 			sx_xunlock(&allproc_lock);
3124 			_PHOLD(p);
3125 			thread_single_end(p, SINGLE_ALLPROC);
3126 			_PRELE(p);
3127 			PROC_UNLOCK(p);
3128 			sx_xlock(&allproc_lock);
3129 		} else {
3130 			PROC_UNLOCK(p);
3131 		}
3132 	}
3133 	sx_xunlock(&allproc_lock);
3134 }
3135 
3136 /* #define	TOTAL_STOP_DEBUG	1 */
3137 #ifdef TOTAL_STOP_DEBUG
3138 volatile static int ap_resume;
3139 #include <sys/mount.h>
3140 
3141 static int
3142 sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS)
3143 {
3144 	int error, val;
3145 
3146 	val = 0;
3147 	ap_resume = 0;
3148 	error = sysctl_handle_int(oidp, &val, 0, req);
3149 	if (error != 0 || req->newptr == NULL)
3150 		return (error);
3151 	if (val != 0) {
3152 		stop_all_proc();
3153 		syncer_suspend();
3154 		while (ap_resume == 0)
3155 			;
3156 		syncer_resume();
3157 		resume_all_proc();
3158 	}
3159 	return (0);
3160 }
3161 
3162 SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW |
3163     CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0,
3164     sysctl_debug_stop_all_proc, "I",
3165     "");
3166 #endif
3167