xref: /freebsd/sys/kern/kern_proc.c (revision b6a05070fa77edc7ce6e60b61623fd806e807be6)
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  * 4. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	@(#)kern_proc.c	8.7 (Berkeley) 2/14/95
30  */
31 
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
34 
35 #include "opt_compat.h"
36 #include "opt_ddb.h"
37 #include "opt_ktrace.h"
38 #include "opt_kstack_pages.h"
39 #include "opt_stack.h"
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/elf.h>
44 #include <sys/eventhandler.h>
45 #include <sys/exec.h>
46 #include <sys/jail.h>
47 #include <sys/kernel.h>
48 #include <sys/limits.h>
49 #include <sys/lock.h>
50 #include <sys/loginclass.h>
51 #include <sys/malloc.h>
52 #include <sys/mman.h>
53 #include <sys/mount.h>
54 #include <sys/mutex.h>
55 #include <sys/proc.h>
56 #include <sys/ptrace.h>
57 #include <sys/refcount.h>
58 #include <sys/resourcevar.h>
59 #include <sys/rwlock.h>
60 #include <sys/sbuf.h>
61 #include <sys/sysent.h>
62 #include <sys/sched.h>
63 #include <sys/smp.h>
64 #include <sys/stack.h>
65 #include <sys/stat.h>
66 #include <sys/sysctl.h>
67 #include <sys/filedesc.h>
68 #include <sys/tty.h>
69 #include <sys/signalvar.h>
70 #include <sys/sdt.h>
71 #include <sys/sx.h>
72 #include <sys/user.h>
73 #include <sys/vnode.h>
74 #include <sys/wait.h>
75 
76 #ifdef DDB
77 #include <ddb/ddb.h>
78 #endif
79 
80 #include <vm/vm.h>
81 #include <vm/vm_param.h>
82 #include <vm/vm_extern.h>
83 #include <vm/pmap.h>
84 #include <vm/vm_map.h>
85 #include <vm/vm_object.h>
86 #include <vm/vm_page.h>
87 #include <vm/uma.h>
88 
89 #ifdef COMPAT_FREEBSD32
90 #include <compat/freebsd32/freebsd32.h>
91 #include <compat/freebsd32/freebsd32_util.h>
92 #endif
93 
94 SDT_PROVIDER_DEFINE(proc);
95 SDT_PROBE_DEFINE4(proc, , ctor, entry, "struct proc *", "int", "void *",
96     "int");
97 SDT_PROBE_DEFINE4(proc, , ctor, return, "struct proc *", "int", "void *",
98     "int");
99 SDT_PROBE_DEFINE4(proc, , dtor, entry, "struct proc *", "int", "void *",
100     "struct thread *");
101 SDT_PROBE_DEFINE3(proc, , dtor, return, "struct proc *", "int", "void *");
102 SDT_PROBE_DEFINE3(proc, , init, entry, "struct proc *", "int", "int");
103 SDT_PROBE_DEFINE3(proc, , init, return, "struct proc *", "int", "int");
104 
105 MALLOC_DEFINE(M_PGRP, "pgrp", "process group header");
106 MALLOC_DEFINE(M_SESSION, "session", "session header");
107 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
108 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
109 
110 static void doenterpgrp(struct proc *, struct pgrp *);
111 static void orphanpg(struct pgrp *pg);
112 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp);
113 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp);
114 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp,
115     int preferthread);
116 static void pgadjustjobc(struct pgrp *pgrp, int entering);
117 static void pgdelete(struct pgrp *);
118 static int proc_ctor(void *mem, int size, void *arg, int flags);
119 static void proc_dtor(void *mem, int size, void *arg);
120 static int proc_init(void *mem, int size, int flags);
121 static void proc_fini(void *mem, int size);
122 static void pargs_free(struct pargs *pa);
123 static struct proc *zpfind_locked(pid_t pid);
124 
125 /*
126  * Other process lists
127  */
128 struct pidhashhead *pidhashtbl;
129 u_long pidhash;
130 struct pgrphashhead *pgrphashtbl;
131 u_long pgrphash;
132 struct proclist allproc;
133 struct proclist zombproc;
134 struct sx allproc_lock;
135 struct sx proctree_lock;
136 struct mtx ppeers_lock;
137 uma_zone_t proc_zone;
138 
139 /*
140  * The offset of various fields in struct proc and struct thread.
141  * These are used by kernel debuggers to enumerate kernel threads and
142  * processes.
143  */
144 const int proc_off_p_pid = offsetof(struct proc, p_pid);
145 const int proc_off_p_comm = offsetof(struct proc, p_comm);
146 const int proc_off_p_list = offsetof(struct proc, p_list);
147 const int proc_off_p_threads = offsetof(struct proc, p_threads);
148 const int thread_off_td_tid = offsetof(struct thread, td_tid);
149 const int thread_off_td_name = offsetof(struct thread, td_name);
150 const int thread_off_td_oncpu = offsetof(struct thread, td_oncpu);
151 const int thread_off_td_pcb = offsetof(struct thread, td_pcb);
152 const int thread_off_td_plist = offsetof(struct thread, td_plist);
153 
154 int kstack_pages = KSTACK_PAGES;
155 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0,
156     "Kernel stack size in pages");
157 static int vmmap_skip_res_cnt = 0;
158 SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW,
159     &vmmap_skip_res_cnt, 0,
160     "Skip calculation of the pages resident count in kern.proc.vmmap");
161 
162 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
163 #ifdef COMPAT_FREEBSD32
164 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
165 #endif
166 
167 /*
168  * Initialize global process hashing structures.
169  */
170 void
171 procinit()
172 {
173 
174 	sx_init(&allproc_lock, "allproc");
175 	sx_init(&proctree_lock, "proctree");
176 	mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
177 	LIST_INIT(&allproc);
178 	LIST_INIT(&zombproc);
179 	pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
180 	pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
181 	proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
182 	    proc_ctor, proc_dtor, proc_init, proc_fini,
183 	    UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
184 	uihashinit();
185 }
186 
187 /*
188  * Prepare a proc for use.
189  */
190 static int
191 proc_ctor(void *mem, int size, void *arg, int flags)
192 {
193 	struct proc *p;
194 
195 	p = (struct proc *)mem;
196 	SDT_PROBE4(proc, , ctor , entry, p, size, arg, flags);
197 	EVENTHANDLER_INVOKE(process_ctor, p);
198 	SDT_PROBE4(proc, , ctor , return, p, size, arg, flags);
199 	return (0);
200 }
201 
202 /*
203  * Reclaim a proc after use.
204  */
205 static void
206 proc_dtor(void *mem, int size, void *arg)
207 {
208 	struct proc *p;
209 	struct thread *td;
210 
211 	/* INVARIANTS checks go here */
212 	p = (struct proc *)mem;
213 	td = FIRST_THREAD_IN_PROC(p);
214 	SDT_PROBE4(proc, , dtor, entry, p, size, arg, td);
215 	if (td != NULL) {
216 #ifdef INVARIANTS
217 		KASSERT((p->p_numthreads == 1),
218 		    ("bad number of threads in exiting process"));
219 		KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
220 #endif
221 		/* Free all OSD associated to this thread. */
222 		osd_thread_exit(td);
223 	}
224 	EVENTHANDLER_INVOKE(process_dtor, p);
225 	if (p->p_ksi != NULL)
226 		KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
227 	SDT_PROBE3(proc, , dtor, return, p, size, arg);
228 }
229 
230 /*
231  * Initialize type-stable parts of a proc (when newly created).
232  */
233 static int
234 proc_init(void *mem, int size, int flags)
235 {
236 	struct proc *p;
237 
238 	p = (struct proc *)mem;
239 	SDT_PROBE3(proc, , init, entry, p, size, flags);
240 	p->p_sched = (struct p_sched *)&p[1];
241 	mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK | MTX_NEW);
242 	mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_NEW);
243 	mtx_init(&p->p_statmtx, "pstatl", NULL, MTX_SPIN | MTX_NEW);
244 	mtx_init(&p->p_itimmtx, "pitiml", NULL, MTX_SPIN | MTX_NEW);
245 	mtx_init(&p->p_profmtx, "pprofl", NULL, MTX_SPIN | MTX_NEW);
246 	cv_init(&p->p_pwait, "ppwait");
247 	cv_init(&p->p_dbgwait, "dbgwait");
248 	TAILQ_INIT(&p->p_threads);	     /* all threads in proc */
249 	EVENTHANDLER_INVOKE(process_init, p);
250 	p->p_stats = pstats_alloc();
251 	SDT_PROBE3(proc, , init, return, p, size, flags);
252 	return (0);
253 }
254 
255 /*
256  * UMA should ensure that this function is never called.
257  * Freeing a proc structure would violate type stability.
258  */
259 static void
260 proc_fini(void *mem, int size)
261 {
262 #ifdef notnow
263 	struct proc *p;
264 
265 	p = (struct proc *)mem;
266 	EVENTHANDLER_INVOKE(process_fini, p);
267 	pstats_free(p->p_stats);
268 	thread_free(FIRST_THREAD_IN_PROC(p));
269 	mtx_destroy(&p->p_mtx);
270 	if (p->p_ksi != NULL)
271 		ksiginfo_free(p->p_ksi);
272 #else
273 	panic("proc reclaimed");
274 #endif
275 }
276 
277 /*
278  * Is p an inferior of the current process?
279  */
280 int
281 inferior(struct proc *p)
282 {
283 
284 	sx_assert(&proctree_lock, SX_LOCKED);
285 	PROC_LOCK_ASSERT(p, MA_OWNED);
286 	for (; p != curproc; p = proc_realparent(p)) {
287 		if (p->p_pid == 0)
288 			return (0);
289 	}
290 	return (1);
291 }
292 
293 struct proc *
294 pfind_locked(pid_t pid)
295 {
296 	struct proc *p;
297 
298 	sx_assert(&allproc_lock, SX_LOCKED);
299 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
300 		if (p->p_pid == pid) {
301 			PROC_LOCK(p);
302 			if (p->p_state == PRS_NEW) {
303 				PROC_UNLOCK(p);
304 				p = NULL;
305 			}
306 			break;
307 		}
308 	}
309 	return (p);
310 }
311 
312 /*
313  * Locate a process by number; return only "live" processes -- i.e., neither
314  * zombies nor newly born but incompletely initialized processes.  By not
315  * returning processes in the PRS_NEW state, we allow callers to avoid
316  * testing for that condition to avoid dereferencing p_ucred, et al.
317  */
318 struct proc *
319 pfind(pid_t pid)
320 {
321 	struct proc *p;
322 
323 	sx_slock(&allproc_lock);
324 	p = pfind_locked(pid);
325 	sx_sunlock(&allproc_lock);
326 	return (p);
327 }
328 
329 static struct proc *
330 pfind_tid_locked(pid_t tid)
331 {
332 	struct proc *p;
333 	struct thread *td;
334 
335 	sx_assert(&allproc_lock, SX_LOCKED);
336 	FOREACH_PROC_IN_SYSTEM(p) {
337 		PROC_LOCK(p);
338 		if (p->p_state == PRS_NEW) {
339 			PROC_UNLOCK(p);
340 			continue;
341 		}
342 		FOREACH_THREAD_IN_PROC(p, td) {
343 			if (td->td_tid == tid)
344 				goto found;
345 		}
346 		PROC_UNLOCK(p);
347 	}
348 found:
349 	return (p);
350 }
351 
352 /*
353  * Locate a process group by number.
354  * The caller must hold proctree_lock.
355  */
356 struct pgrp *
357 pgfind(pgid)
358 	register pid_t pgid;
359 {
360 	register struct pgrp *pgrp;
361 
362 	sx_assert(&proctree_lock, SX_LOCKED);
363 
364 	LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
365 		if (pgrp->pg_id == pgid) {
366 			PGRP_LOCK(pgrp);
367 			return (pgrp);
368 		}
369 	}
370 	return (NULL);
371 }
372 
373 /*
374  * Locate process and do additional manipulations, depending on flags.
375  */
376 int
377 pget(pid_t pid, int flags, struct proc **pp)
378 {
379 	struct proc *p;
380 	int error;
381 
382 	sx_slock(&allproc_lock);
383 	if (pid <= PID_MAX) {
384 		p = pfind_locked(pid);
385 		if (p == NULL && (flags & PGET_NOTWEXIT) == 0)
386 			p = zpfind_locked(pid);
387 	} else if ((flags & PGET_NOTID) == 0) {
388 		p = pfind_tid_locked(pid);
389 	} else {
390 		p = NULL;
391 	}
392 	sx_sunlock(&allproc_lock);
393 	if (p == NULL)
394 		return (ESRCH);
395 	if ((flags & PGET_CANSEE) != 0) {
396 		error = p_cansee(curthread, p);
397 		if (error != 0)
398 			goto errout;
399 	}
400 	if ((flags & PGET_CANDEBUG) != 0) {
401 		error = p_candebug(curthread, p);
402 		if (error != 0)
403 			goto errout;
404 	}
405 	if ((flags & PGET_ISCURRENT) != 0 && curproc != p) {
406 		error = EPERM;
407 		goto errout;
408 	}
409 	if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) {
410 		error = ESRCH;
411 		goto errout;
412 	}
413 	if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) {
414 		/*
415 		 * XXXRW: Not clear ESRCH is the right error during proc
416 		 * execve().
417 		 */
418 		error = ESRCH;
419 		goto errout;
420 	}
421 	if ((flags & PGET_HOLD) != 0) {
422 		_PHOLD(p);
423 		PROC_UNLOCK(p);
424 	}
425 	*pp = p;
426 	return (0);
427 errout:
428 	PROC_UNLOCK(p);
429 	return (error);
430 }
431 
432 /*
433  * Create a new process group.
434  * pgid must be equal to the pid of p.
435  * Begin a new session if required.
436  */
437 int
438 enterpgrp(p, pgid, pgrp, sess)
439 	register struct proc *p;
440 	pid_t pgid;
441 	struct pgrp *pgrp;
442 	struct session *sess;
443 {
444 
445 	sx_assert(&proctree_lock, SX_XLOCKED);
446 
447 	KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
448 	KASSERT(p->p_pid == pgid,
449 	    ("enterpgrp: new pgrp and pid != pgid"));
450 	KASSERT(pgfind(pgid) == NULL,
451 	    ("enterpgrp: pgrp with pgid exists"));
452 	KASSERT(!SESS_LEADER(p),
453 	    ("enterpgrp: session leader attempted setpgrp"));
454 
455 	mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
456 
457 	if (sess != NULL) {
458 		/*
459 		 * new session
460 		 */
461 		mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
462 		PROC_LOCK(p);
463 		p->p_flag &= ~P_CONTROLT;
464 		PROC_UNLOCK(p);
465 		PGRP_LOCK(pgrp);
466 		sess->s_leader = p;
467 		sess->s_sid = p->p_pid;
468 		refcount_init(&sess->s_count, 1);
469 		sess->s_ttyvp = NULL;
470 		sess->s_ttydp = NULL;
471 		sess->s_ttyp = NULL;
472 		bcopy(p->p_session->s_login, sess->s_login,
473 			    sizeof(sess->s_login));
474 		pgrp->pg_session = sess;
475 		KASSERT(p == curproc,
476 		    ("enterpgrp: mksession and p != curproc"));
477 	} else {
478 		pgrp->pg_session = p->p_session;
479 		sess_hold(pgrp->pg_session);
480 		PGRP_LOCK(pgrp);
481 	}
482 	pgrp->pg_id = pgid;
483 	LIST_INIT(&pgrp->pg_members);
484 
485 	/*
486 	 * As we have an exclusive lock of proctree_lock,
487 	 * this should not deadlock.
488 	 */
489 	LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
490 	pgrp->pg_jobc = 0;
491 	SLIST_INIT(&pgrp->pg_sigiolst);
492 	PGRP_UNLOCK(pgrp);
493 
494 	doenterpgrp(p, pgrp);
495 
496 	return (0);
497 }
498 
499 /*
500  * Move p to an existing process group
501  */
502 int
503 enterthispgrp(p, pgrp)
504 	register struct proc *p;
505 	struct pgrp *pgrp;
506 {
507 
508 	sx_assert(&proctree_lock, SX_XLOCKED);
509 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
510 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
511 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
512 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
513 	KASSERT(pgrp->pg_session == p->p_session,
514 		("%s: pgrp's session %p, p->p_session %p.\n",
515 		__func__,
516 		pgrp->pg_session,
517 		p->p_session));
518 	KASSERT(pgrp != p->p_pgrp,
519 		("%s: p belongs to pgrp.", __func__));
520 
521 	doenterpgrp(p, pgrp);
522 
523 	return (0);
524 }
525 
526 /*
527  * Move p to a process group
528  */
529 static void
530 doenterpgrp(p, pgrp)
531 	struct proc *p;
532 	struct pgrp *pgrp;
533 {
534 	struct pgrp *savepgrp;
535 
536 	sx_assert(&proctree_lock, SX_XLOCKED);
537 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
538 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
539 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
540 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
541 
542 	savepgrp = p->p_pgrp;
543 
544 	/*
545 	 * Adjust eligibility of affected pgrps to participate in job control.
546 	 * Increment eligibility counts before decrementing, otherwise we
547 	 * could reach 0 spuriously during the first call.
548 	 */
549 	fixjobc(p, pgrp, 1);
550 	fixjobc(p, p->p_pgrp, 0);
551 
552 	PGRP_LOCK(pgrp);
553 	PGRP_LOCK(savepgrp);
554 	PROC_LOCK(p);
555 	LIST_REMOVE(p, p_pglist);
556 	p->p_pgrp = pgrp;
557 	PROC_UNLOCK(p);
558 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
559 	PGRP_UNLOCK(savepgrp);
560 	PGRP_UNLOCK(pgrp);
561 	if (LIST_EMPTY(&savepgrp->pg_members))
562 		pgdelete(savepgrp);
563 }
564 
565 /*
566  * remove process from process group
567  */
568 int
569 leavepgrp(p)
570 	register struct proc *p;
571 {
572 	struct pgrp *savepgrp;
573 
574 	sx_assert(&proctree_lock, SX_XLOCKED);
575 	savepgrp = p->p_pgrp;
576 	PGRP_LOCK(savepgrp);
577 	PROC_LOCK(p);
578 	LIST_REMOVE(p, p_pglist);
579 	p->p_pgrp = NULL;
580 	PROC_UNLOCK(p);
581 	PGRP_UNLOCK(savepgrp);
582 	if (LIST_EMPTY(&savepgrp->pg_members))
583 		pgdelete(savepgrp);
584 	return (0);
585 }
586 
587 /*
588  * delete a process group
589  */
590 static void
591 pgdelete(pgrp)
592 	register struct pgrp *pgrp;
593 {
594 	struct session *savesess;
595 	struct tty *tp;
596 
597 	sx_assert(&proctree_lock, SX_XLOCKED);
598 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
599 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
600 
601 	/*
602 	 * Reset any sigio structures pointing to us as a result of
603 	 * F_SETOWN with our pgid.
604 	 */
605 	funsetownlst(&pgrp->pg_sigiolst);
606 
607 	PGRP_LOCK(pgrp);
608 	tp = pgrp->pg_session->s_ttyp;
609 	LIST_REMOVE(pgrp, pg_hash);
610 	savesess = pgrp->pg_session;
611 	PGRP_UNLOCK(pgrp);
612 
613 	/* Remove the reference to the pgrp before deallocating it. */
614 	if (tp != NULL) {
615 		tty_lock(tp);
616 		tty_rel_pgrp(tp, pgrp);
617 	}
618 
619 	mtx_destroy(&pgrp->pg_mtx);
620 	free(pgrp, M_PGRP);
621 	sess_release(savesess);
622 }
623 
624 static void
625 pgadjustjobc(pgrp, entering)
626 	struct pgrp *pgrp;
627 	int entering;
628 {
629 
630 	PGRP_LOCK(pgrp);
631 	if (entering)
632 		pgrp->pg_jobc++;
633 	else {
634 		--pgrp->pg_jobc;
635 		if (pgrp->pg_jobc == 0)
636 			orphanpg(pgrp);
637 	}
638 	PGRP_UNLOCK(pgrp);
639 }
640 
641 /*
642  * Adjust pgrp jobc counters when specified process changes process group.
643  * We count the number of processes in each process group that "qualify"
644  * the group for terminal job control (those with a parent in a different
645  * process group of the same session).  If that count reaches zero, the
646  * process group becomes orphaned.  Check both the specified process'
647  * process group and that of its children.
648  * entering == 0 => p is leaving specified group.
649  * entering == 1 => p is entering specified group.
650  */
651 void
652 fixjobc(p, pgrp, entering)
653 	register struct proc *p;
654 	register struct pgrp *pgrp;
655 	int entering;
656 {
657 	register struct pgrp *hispgrp;
658 	register struct session *mysession;
659 
660 	sx_assert(&proctree_lock, SX_LOCKED);
661 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
662 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
663 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
664 
665 	/*
666 	 * Check p's parent to see whether p qualifies its own process
667 	 * group; if so, adjust count for p's process group.
668 	 */
669 	mysession = pgrp->pg_session;
670 	if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
671 	    hispgrp->pg_session == mysession)
672 		pgadjustjobc(pgrp, entering);
673 
674 	/*
675 	 * Check this process' children to see whether they qualify
676 	 * their process groups; if so, adjust counts for children's
677 	 * process groups.
678 	 */
679 	LIST_FOREACH(p, &p->p_children, p_sibling) {
680 		hispgrp = p->p_pgrp;
681 		if (hispgrp == pgrp ||
682 		    hispgrp->pg_session != mysession)
683 			continue;
684 		PROC_LOCK(p);
685 		if (p->p_state == PRS_ZOMBIE) {
686 			PROC_UNLOCK(p);
687 			continue;
688 		}
689 		PROC_UNLOCK(p);
690 		pgadjustjobc(hispgrp, entering);
691 	}
692 }
693 
694 /*
695  * A process group has become orphaned;
696  * if there are any stopped processes in the group,
697  * hang-up all process in that group.
698  */
699 static void
700 orphanpg(pg)
701 	struct pgrp *pg;
702 {
703 	register struct proc *p;
704 
705 	PGRP_LOCK_ASSERT(pg, MA_OWNED);
706 
707 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
708 		PROC_LOCK(p);
709 		if (P_SHOULDSTOP(p) == P_STOPPED_SIG) {
710 			PROC_UNLOCK(p);
711 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
712 				PROC_LOCK(p);
713 				kern_psignal(p, SIGHUP);
714 				kern_psignal(p, SIGCONT);
715 				PROC_UNLOCK(p);
716 			}
717 			return;
718 		}
719 		PROC_UNLOCK(p);
720 	}
721 }
722 
723 void
724 sess_hold(struct session *s)
725 {
726 
727 	refcount_acquire(&s->s_count);
728 }
729 
730 void
731 sess_release(struct session *s)
732 {
733 
734 	if (refcount_release(&s->s_count)) {
735 		if (s->s_ttyp != NULL) {
736 			tty_lock(s->s_ttyp);
737 			tty_rel_sess(s->s_ttyp, s);
738 		}
739 		mtx_destroy(&s->s_mtx);
740 		free(s, M_SESSION);
741 	}
742 }
743 
744 #ifdef DDB
745 
746 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
747 {
748 	register struct pgrp *pgrp;
749 	register struct proc *p;
750 	register int i;
751 
752 	for (i = 0; i <= pgrphash; i++) {
753 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
754 			printf("\tindx %d\n", i);
755 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
756 				printf(
757 			"\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
758 				    (void *)pgrp, (long)pgrp->pg_id,
759 				    (void *)pgrp->pg_session,
760 				    pgrp->pg_session->s_count,
761 				    (void *)LIST_FIRST(&pgrp->pg_members));
762 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
763 					printf("\t\tpid %ld addr %p pgrp %p\n",
764 					    (long)p->p_pid, (void *)p,
765 					    (void *)p->p_pgrp);
766 				}
767 			}
768 		}
769 	}
770 }
771 #endif /* DDB */
772 
773 /*
774  * Calculate the kinfo_proc members which contain process-wide
775  * informations.
776  * Must be called with the target process locked.
777  */
778 static void
779 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
780 {
781 	struct thread *td;
782 
783 	PROC_LOCK_ASSERT(p, MA_OWNED);
784 
785 	kp->ki_estcpu = 0;
786 	kp->ki_pctcpu = 0;
787 	FOREACH_THREAD_IN_PROC(p, td) {
788 		thread_lock(td);
789 		kp->ki_pctcpu += sched_pctcpu(td);
790 		kp->ki_estcpu += td->td_estcpu;
791 		thread_unlock(td);
792 	}
793 }
794 
795 /*
796  * Clear kinfo_proc and fill in any information that is common
797  * to all threads in the process.
798  * Must be called with the target process locked.
799  */
800 static void
801 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
802 {
803 	struct thread *td0;
804 	struct tty *tp;
805 	struct session *sp;
806 	struct ucred *cred;
807 	struct sigacts *ps;
808 
809 	/* For proc_realparent. */
810 	sx_assert(&proctree_lock, SX_LOCKED);
811 	PROC_LOCK_ASSERT(p, MA_OWNED);
812 	bzero(kp, sizeof(*kp));
813 
814 	kp->ki_structsize = sizeof(*kp);
815 	kp->ki_paddr = p;
816 	kp->ki_addr =/* p->p_addr; */0; /* XXX */
817 	kp->ki_args = p->p_args;
818 	kp->ki_textvp = p->p_textvp;
819 #ifdef KTRACE
820 	kp->ki_tracep = p->p_tracevp;
821 	kp->ki_traceflag = p->p_traceflag;
822 #endif
823 	kp->ki_fd = p->p_fd;
824 	kp->ki_vmspace = p->p_vmspace;
825 	kp->ki_flag = p->p_flag;
826 	kp->ki_flag2 = p->p_flag2;
827 	cred = p->p_ucred;
828 	if (cred) {
829 		kp->ki_uid = cred->cr_uid;
830 		kp->ki_ruid = cred->cr_ruid;
831 		kp->ki_svuid = cred->cr_svuid;
832 		kp->ki_cr_flags = 0;
833 		if (cred->cr_flags & CRED_FLAG_CAPMODE)
834 			kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
835 		/* XXX bde doesn't like KI_NGROUPS */
836 		if (cred->cr_ngroups > KI_NGROUPS) {
837 			kp->ki_ngroups = KI_NGROUPS;
838 			kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
839 		} else
840 			kp->ki_ngroups = cred->cr_ngroups;
841 		bcopy(cred->cr_groups, kp->ki_groups,
842 		    kp->ki_ngroups * sizeof(gid_t));
843 		kp->ki_rgid = cred->cr_rgid;
844 		kp->ki_svgid = cred->cr_svgid;
845 		/* If jailed(cred), emulate the old P_JAILED flag. */
846 		if (jailed(cred)) {
847 			kp->ki_flag |= P_JAILED;
848 			/* If inside the jail, use 0 as a jail ID. */
849 			if (cred->cr_prison != curthread->td_ucred->cr_prison)
850 				kp->ki_jid = cred->cr_prison->pr_id;
851 		}
852 		strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
853 		    sizeof(kp->ki_loginclass));
854 	}
855 	ps = p->p_sigacts;
856 	if (ps) {
857 		mtx_lock(&ps->ps_mtx);
858 		kp->ki_sigignore = ps->ps_sigignore;
859 		kp->ki_sigcatch = ps->ps_sigcatch;
860 		mtx_unlock(&ps->ps_mtx);
861 	}
862 	if (p->p_state != PRS_NEW &&
863 	    p->p_state != PRS_ZOMBIE &&
864 	    p->p_vmspace != NULL) {
865 		struct vmspace *vm = p->p_vmspace;
866 
867 		kp->ki_size = vm->vm_map.size;
868 		kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
869 		FOREACH_THREAD_IN_PROC(p, td0) {
870 			if (!TD_IS_SWAPPED(td0))
871 				kp->ki_rssize += td0->td_kstack_pages;
872 		}
873 		kp->ki_swrss = vm->vm_swrss;
874 		kp->ki_tsize = vm->vm_tsize;
875 		kp->ki_dsize = vm->vm_dsize;
876 		kp->ki_ssize = vm->vm_ssize;
877 	} else if (p->p_state == PRS_ZOMBIE)
878 		kp->ki_stat = SZOMB;
879 	if (kp->ki_flag & P_INMEM)
880 		kp->ki_sflag = PS_INMEM;
881 	else
882 		kp->ki_sflag = 0;
883 	/* Calculate legacy swtime as seconds since 'swtick'. */
884 	kp->ki_swtime = (ticks - p->p_swtick) / hz;
885 	kp->ki_pid = p->p_pid;
886 	kp->ki_nice = p->p_nice;
887 	kp->ki_fibnum = p->p_fibnum;
888 	kp->ki_start = p->p_stats->p_start;
889 	timevaladd(&kp->ki_start, &boottime);
890 	PROC_STATLOCK(p);
891 	rufetch(p, &kp->ki_rusage);
892 	kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
893 	calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
894 	PROC_STATUNLOCK(p);
895 	calccru(p, &kp->ki_childutime, &kp->ki_childstime);
896 	/* Some callers want child times in a single value. */
897 	kp->ki_childtime = kp->ki_childstime;
898 	timevaladd(&kp->ki_childtime, &kp->ki_childutime);
899 
900 	FOREACH_THREAD_IN_PROC(p, td0)
901 		kp->ki_cow += td0->td_cow;
902 
903 	tp = NULL;
904 	if (p->p_pgrp) {
905 		kp->ki_pgid = p->p_pgrp->pg_id;
906 		kp->ki_jobc = p->p_pgrp->pg_jobc;
907 		sp = p->p_pgrp->pg_session;
908 
909 		if (sp != NULL) {
910 			kp->ki_sid = sp->s_sid;
911 			SESS_LOCK(sp);
912 			strlcpy(kp->ki_login, sp->s_login,
913 			    sizeof(kp->ki_login));
914 			if (sp->s_ttyvp)
915 				kp->ki_kiflag |= KI_CTTY;
916 			if (SESS_LEADER(p))
917 				kp->ki_kiflag |= KI_SLEADER;
918 			/* XXX proctree_lock */
919 			tp = sp->s_ttyp;
920 			SESS_UNLOCK(sp);
921 		}
922 	}
923 	if ((p->p_flag & P_CONTROLT) && tp != NULL) {
924 		kp->ki_tdev = tty_udev(tp);
925 		kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
926 		if (tp->t_session)
927 			kp->ki_tsid = tp->t_session->s_sid;
928 	} else
929 		kp->ki_tdev = NODEV;
930 	if (p->p_comm[0] != '\0')
931 		strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
932 	if (p->p_sysent && p->p_sysent->sv_name != NULL &&
933 	    p->p_sysent->sv_name[0] != '\0')
934 		strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
935 	kp->ki_siglist = p->p_siglist;
936 	kp->ki_xstat = KW_EXITCODE(p->p_xexit, p->p_xsig);
937 	kp->ki_acflag = p->p_acflag;
938 	kp->ki_lock = p->p_lock;
939 	if (p->p_pptr) {
940 		kp->ki_ppid = proc_realparent(p)->p_pid;
941 		if (p->p_flag & P_TRACED)
942 			kp->ki_tracer = p->p_pptr->p_pid;
943 	}
944 }
945 
946 /*
947  * Fill in information that is thread specific.  Must be called with
948  * target process locked.  If 'preferthread' is set, overwrite certain
949  * process-related fields that are maintained for both threads and
950  * processes.
951  */
952 static void
953 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
954 {
955 	struct proc *p;
956 
957 	p = td->td_proc;
958 	kp->ki_tdaddr = td;
959 	PROC_LOCK_ASSERT(p, MA_OWNED);
960 
961 	if (preferthread)
962 		PROC_STATLOCK(p);
963 	thread_lock(td);
964 	if (td->td_wmesg != NULL)
965 		strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
966 	else
967 		bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
968 	strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname));
969 	if (TD_ON_LOCK(td)) {
970 		kp->ki_kiflag |= KI_LOCKBLOCK;
971 		strlcpy(kp->ki_lockname, td->td_lockname,
972 		    sizeof(kp->ki_lockname));
973 	} else {
974 		kp->ki_kiflag &= ~KI_LOCKBLOCK;
975 		bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
976 	}
977 
978 	if (p->p_state == PRS_NORMAL) { /* approximate. */
979 		if (TD_ON_RUNQ(td) ||
980 		    TD_CAN_RUN(td) ||
981 		    TD_IS_RUNNING(td)) {
982 			kp->ki_stat = SRUN;
983 		} else if (P_SHOULDSTOP(p)) {
984 			kp->ki_stat = SSTOP;
985 		} else if (TD_IS_SLEEPING(td)) {
986 			kp->ki_stat = SSLEEP;
987 		} else if (TD_ON_LOCK(td)) {
988 			kp->ki_stat = SLOCK;
989 		} else {
990 			kp->ki_stat = SWAIT;
991 		}
992 	} else if (p->p_state == PRS_ZOMBIE) {
993 		kp->ki_stat = SZOMB;
994 	} else {
995 		kp->ki_stat = SIDL;
996 	}
997 
998 	/* Things in the thread */
999 	kp->ki_wchan = td->td_wchan;
1000 	kp->ki_pri.pri_level = td->td_priority;
1001 	kp->ki_pri.pri_native = td->td_base_pri;
1002 
1003 	/*
1004 	 * Note: legacy fields; clamp at the old NOCPU value and/or
1005 	 * the maximum u_char CPU value.
1006 	 */
1007 	if (td->td_lastcpu == NOCPU)
1008 		kp->ki_lastcpu_old = NOCPU_OLD;
1009 	else if (td->td_lastcpu > MAXCPU_OLD)
1010 		kp->ki_lastcpu_old = MAXCPU_OLD;
1011 	else
1012 		kp->ki_lastcpu_old = td->td_lastcpu;
1013 
1014 	if (td->td_oncpu == NOCPU)
1015 		kp->ki_oncpu_old = NOCPU_OLD;
1016 	else if (td->td_oncpu > MAXCPU_OLD)
1017 		kp->ki_oncpu_old = MAXCPU_OLD;
1018 	else
1019 		kp->ki_oncpu_old = td->td_oncpu;
1020 
1021 	kp->ki_lastcpu = td->td_lastcpu;
1022 	kp->ki_oncpu = td->td_oncpu;
1023 	kp->ki_tdflags = td->td_flags;
1024 	kp->ki_tid = td->td_tid;
1025 	kp->ki_numthreads = p->p_numthreads;
1026 	kp->ki_pcb = td->td_pcb;
1027 	kp->ki_kstack = (void *)td->td_kstack;
1028 	kp->ki_slptime = (ticks - td->td_slptick) / hz;
1029 	kp->ki_pri.pri_class = td->td_pri_class;
1030 	kp->ki_pri.pri_user = td->td_user_pri;
1031 
1032 	if (preferthread) {
1033 		rufetchtd(td, &kp->ki_rusage);
1034 		kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
1035 		kp->ki_pctcpu = sched_pctcpu(td);
1036 		kp->ki_estcpu = td->td_estcpu;
1037 		kp->ki_cow = td->td_cow;
1038 	}
1039 
1040 	/* We can't get this anymore but ps etc never used it anyway. */
1041 	kp->ki_rqindex = 0;
1042 
1043 	if (preferthread)
1044 		kp->ki_siglist = td->td_siglist;
1045 	kp->ki_sigmask = td->td_sigmask;
1046 	thread_unlock(td);
1047 	if (preferthread)
1048 		PROC_STATUNLOCK(p);
1049 }
1050 
1051 /*
1052  * Fill in a kinfo_proc structure for the specified process.
1053  * Must be called with the target process locked.
1054  */
1055 void
1056 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1057 {
1058 
1059 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1060 
1061 	fill_kinfo_proc_only(p, kp);
1062 	fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1063 	fill_kinfo_aggregate(p, kp);
1064 }
1065 
1066 struct pstats *
1067 pstats_alloc(void)
1068 {
1069 
1070 	return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1071 }
1072 
1073 /*
1074  * Copy parts of p_stats; zero the rest of p_stats (statistics).
1075  */
1076 void
1077 pstats_fork(struct pstats *src, struct pstats *dst)
1078 {
1079 
1080 	bzero(&dst->pstat_startzero,
1081 	    __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1082 	bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1083 	    __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1084 }
1085 
1086 void
1087 pstats_free(struct pstats *ps)
1088 {
1089 
1090 	free(ps, M_SUBPROC);
1091 }
1092 
1093 static struct proc *
1094 zpfind_locked(pid_t pid)
1095 {
1096 	struct proc *p;
1097 
1098 	sx_assert(&allproc_lock, SX_LOCKED);
1099 	LIST_FOREACH(p, &zombproc, p_list) {
1100 		if (p->p_pid == pid) {
1101 			PROC_LOCK(p);
1102 			break;
1103 		}
1104 	}
1105 	return (p);
1106 }
1107 
1108 /*
1109  * Locate a zombie process by number
1110  */
1111 struct proc *
1112 zpfind(pid_t pid)
1113 {
1114 	struct proc *p;
1115 
1116 	sx_slock(&allproc_lock);
1117 	p = zpfind_locked(pid);
1118 	sx_sunlock(&allproc_lock);
1119 	return (p);
1120 }
1121 
1122 #ifdef COMPAT_FREEBSD32
1123 
1124 /*
1125  * This function is typically used to copy out the kernel address, so
1126  * it can be replaced by assignment of zero.
1127  */
1128 static inline uint32_t
1129 ptr32_trim(void *ptr)
1130 {
1131 	uintptr_t uptr;
1132 
1133 	uptr = (uintptr_t)ptr;
1134 	return ((uptr > UINT_MAX) ? 0 : uptr);
1135 }
1136 
1137 #define PTRTRIM_CP(src,dst,fld) \
1138 	do { (dst).fld = ptr32_trim((src).fld); } while (0)
1139 
1140 static void
1141 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1142 {
1143 	int i;
1144 
1145 	bzero(ki32, sizeof(struct kinfo_proc32));
1146 	ki32->ki_structsize = sizeof(struct kinfo_proc32);
1147 	CP(*ki, *ki32, ki_layout);
1148 	PTRTRIM_CP(*ki, *ki32, ki_args);
1149 	PTRTRIM_CP(*ki, *ki32, ki_paddr);
1150 	PTRTRIM_CP(*ki, *ki32, ki_addr);
1151 	PTRTRIM_CP(*ki, *ki32, ki_tracep);
1152 	PTRTRIM_CP(*ki, *ki32, ki_textvp);
1153 	PTRTRIM_CP(*ki, *ki32, ki_fd);
1154 	PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1155 	PTRTRIM_CP(*ki, *ki32, ki_wchan);
1156 	CP(*ki, *ki32, ki_pid);
1157 	CP(*ki, *ki32, ki_ppid);
1158 	CP(*ki, *ki32, ki_pgid);
1159 	CP(*ki, *ki32, ki_tpgid);
1160 	CP(*ki, *ki32, ki_sid);
1161 	CP(*ki, *ki32, ki_tsid);
1162 	CP(*ki, *ki32, ki_jobc);
1163 	CP(*ki, *ki32, ki_tdev);
1164 	CP(*ki, *ki32, ki_siglist);
1165 	CP(*ki, *ki32, ki_sigmask);
1166 	CP(*ki, *ki32, ki_sigignore);
1167 	CP(*ki, *ki32, ki_sigcatch);
1168 	CP(*ki, *ki32, ki_uid);
1169 	CP(*ki, *ki32, ki_ruid);
1170 	CP(*ki, *ki32, ki_svuid);
1171 	CP(*ki, *ki32, ki_rgid);
1172 	CP(*ki, *ki32, ki_svgid);
1173 	CP(*ki, *ki32, ki_ngroups);
1174 	for (i = 0; i < KI_NGROUPS; i++)
1175 		CP(*ki, *ki32, ki_groups[i]);
1176 	CP(*ki, *ki32, ki_size);
1177 	CP(*ki, *ki32, ki_rssize);
1178 	CP(*ki, *ki32, ki_swrss);
1179 	CP(*ki, *ki32, ki_tsize);
1180 	CP(*ki, *ki32, ki_dsize);
1181 	CP(*ki, *ki32, ki_ssize);
1182 	CP(*ki, *ki32, ki_xstat);
1183 	CP(*ki, *ki32, ki_acflag);
1184 	CP(*ki, *ki32, ki_pctcpu);
1185 	CP(*ki, *ki32, ki_estcpu);
1186 	CP(*ki, *ki32, ki_slptime);
1187 	CP(*ki, *ki32, ki_swtime);
1188 	CP(*ki, *ki32, ki_cow);
1189 	CP(*ki, *ki32, ki_runtime);
1190 	TV_CP(*ki, *ki32, ki_start);
1191 	TV_CP(*ki, *ki32, ki_childtime);
1192 	CP(*ki, *ki32, ki_flag);
1193 	CP(*ki, *ki32, ki_kiflag);
1194 	CP(*ki, *ki32, ki_traceflag);
1195 	CP(*ki, *ki32, ki_stat);
1196 	CP(*ki, *ki32, ki_nice);
1197 	CP(*ki, *ki32, ki_lock);
1198 	CP(*ki, *ki32, ki_rqindex);
1199 	CP(*ki, *ki32, ki_oncpu);
1200 	CP(*ki, *ki32, ki_lastcpu);
1201 
1202 	/* XXX TODO: wrap cpu value as appropriate */
1203 	CP(*ki, *ki32, ki_oncpu_old);
1204 	CP(*ki, *ki32, ki_lastcpu_old);
1205 
1206 	bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1207 	bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1208 	bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1209 	bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1210 	bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1211 	bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1212 	bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1213 	CP(*ki, *ki32, ki_tracer);
1214 	CP(*ki, *ki32, ki_flag2);
1215 	CP(*ki, *ki32, ki_fibnum);
1216 	CP(*ki, *ki32, ki_cr_flags);
1217 	CP(*ki, *ki32, ki_jid);
1218 	CP(*ki, *ki32, ki_numthreads);
1219 	CP(*ki, *ki32, ki_tid);
1220 	CP(*ki, *ki32, ki_pri);
1221 	freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1222 	freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1223 	PTRTRIM_CP(*ki, *ki32, ki_pcb);
1224 	PTRTRIM_CP(*ki, *ki32, ki_kstack);
1225 	PTRTRIM_CP(*ki, *ki32, ki_udata);
1226 	CP(*ki, *ki32, ki_sflag);
1227 	CP(*ki, *ki32, ki_tdflags);
1228 }
1229 #endif
1230 
1231 int
1232 kern_proc_out(struct proc *p, struct sbuf *sb, int flags)
1233 {
1234 	struct thread *td;
1235 	struct kinfo_proc ki;
1236 #ifdef COMPAT_FREEBSD32
1237 	struct kinfo_proc32 ki32;
1238 #endif
1239 	int error;
1240 
1241 	PROC_LOCK_ASSERT(p, MA_OWNED);
1242 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1243 
1244 	error = 0;
1245 	fill_kinfo_proc(p, &ki);
1246 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1247 #ifdef COMPAT_FREEBSD32
1248 		if ((flags & KERN_PROC_MASK32) != 0) {
1249 			freebsd32_kinfo_proc_out(&ki, &ki32);
1250 			if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1251 				error = ENOMEM;
1252 		} else
1253 #endif
1254 			if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1255 				error = ENOMEM;
1256 	} else {
1257 		FOREACH_THREAD_IN_PROC(p, td) {
1258 			fill_kinfo_thread(td, &ki, 1);
1259 #ifdef COMPAT_FREEBSD32
1260 			if ((flags & KERN_PROC_MASK32) != 0) {
1261 				freebsd32_kinfo_proc_out(&ki, &ki32);
1262 				if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1263 					error = ENOMEM;
1264 			} else
1265 #endif
1266 				if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1267 					error = ENOMEM;
1268 			if (error != 0)
1269 				break;
1270 		}
1271 	}
1272 	PROC_UNLOCK(p);
1273 	return (error);
1274 }
1275 
1276 static int
1277 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags,
1278     int doingzomb)
1279 {
1280 	struct sbuf sb;
1281 	struct kinfo_proc ki;
1282 	struct proc *np;
1283 	int error, error2;
1284 	pid_t pid;
1285 
1286 	pid = p->p_pid;
1287 	sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req);
1288 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1289 	error = kern_proc_out(p, &sb, flags);
1290 	error2 = sbuf_finish(&sb);
1291 	sbuf_delete(&sb);
1292 	if (error != 0)
1293 		return (error);
1294 	else if (error2 != 0)
1295 		return (error2);
1296 	if (doingzomb)
1297 		np = zpfind(pid);
1298 	else {
1299 		if (pid == 0)
1300 			return (0);
1301 		np = pfind(pid);
1302 	}
1303 	if (np == NULL)
1304 		return (ESRCH);
1305 	if (np != p) {
1306 		PROC_UNLOCK(np);
1307 		return (ESRCH);
1308 	}
1309 	PROC_UNLOCK(np);
1310 	return (0);
1311 }
1312 
1313 static int
1314 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1315 {
1316 	int *name = (int *)arg1;
1317 	u_int namelen = arg2;
1318 	struct proc *p;
1319 	int flags, doingzomb, oid_number;
1320 	int error = 0;
1321 
1322 	oid_number = oidp->oid_number;
1323 	if (oid_number != KERN_PROC_ALL &&
1324 	    (oid_number & KERN_PROC_INC_THREAD) == 0)
1325 		flags = KERN_PROC_NOTHREADS;
1326 	else {
1327 		flags = 0;
1328 		oid_number &= ~KERN_PROC_INC_THREAD;
1329 	}
1330 #ifdef COMPAT_FREEBSD32
1331 	if (req->flags & SCTL_MASK32)
1332 		flags |= KERN_PROC_MASK32;
1333 #endif
1334 	if (oid_number == KERN_PROC_PID) {
1335 		if (namelen != 1)
1336 			return (EINVAL);
1337 		error = sysctl_wire_old_buffer(req, 0);
1338 		if (error)
1339 			return (error);
1340 		sx_slock(&proctree_lock);
1341 		error = pget((pid_t)name[0], PGET_CANSEE, &p);
1342 		if (error == 0)
1343 			error = sysctl_out_proc(p, req, flags, 0);
1344 		sx_sunlock(&proctree_lock);
1345 		return (error);
1346 	}
1347 
1348 	switch (oid_number) {
1349 	case KERN_PROC_ALL:
1350 		if (namelen != 0)
1351 			return (EINVAL);
1352 		break;
1353 	case KERN_PROC_PROC:
1354 		if (namelen != 0 && namelen != 1)
1355 			return (EINVAL);
1356 		break;
1357 	default:
1358 		if (namelen != 1)
1359 			return (EINVAL);
1360 		break;
1361 	}
1362 
1363 	if (!req->oldptr) {
1364 		/* overestimate by 5 procs */
1365 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1366 		if (error)
1367 			return (error);
1368 	}
1369 	error = sysctl_wire_old_buffer(req, 0);
1370 	if (error != 0)
1371 		return (error);
1372 	sx_slock(&proctree_lock);
1373 	sx_slock(&allproc_lock);
1374 	for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) {
1375 		if (!doingzomb)
1376 			p = LIST_FIRST(&allproc);
1377 		else
1378 			p = LIST_FIRST(&zombproc);
1379 		for (; p != 0; p = LIST_NEXT(p, p_list)) {
1380 			/*
1381 			 * Skip embryonic processes.
1382 			 */
1383 			PROC_LOCK(p);
1384 			if (p->p_state == PRS_NEW) {
1385 				PROC_UNLOCK(p);
1386 				continue;
1387 			}
1388 			KASSERT(p->p_ucred != NULL,
1389 			    ("process credential is NULL for non-NEW proc"));
1390 			/*
1391 			 * Show a user only appropriate processes.
1392 			 */
1393 			if (p_cansee(curthread, p)) {
1394 				PROC_UNLOCK(p);
1395 				continue;
1396 			}
1397 			/*
1398 			 * TODO - make more efficient (see notes below).
1399 			 * do by session.
1400 			 */
1401 			switch (oid_number) {
1402 
1403 			case KERN_PROC_GID:
1404 				if (p->p_ucred->cr_gid != (gid_t)name[0]) {
1405 					PROC_UNLOCK(p);
1406 					continue;
1407 				}
1408 				break;
1409 
1410 			case KERN_PROC_PGRP:
1411 				/* could do this by traversing pgrp */
1412 				if (p->p_pgrp == NULL ||
1413 				    p->p_pgrp->pg_id != (pid_t)name[0]) {
1414 					PROC_UNLOCK(p);
1415 					continue;
1416 				}
1417 				break;
1418 
1419 			case KERN_PROC_RGID:
1420 				if (p->p_ucred->cr_rgid != (gid_t)name[0]) {
1421 					PROC_UNLOCK(p);
1422 					continue;
1423 				}
1424 				break;
1425 
1426 			case KERN_PROC_SESSION:
1427 				if (p->p_session == NULL ||
1428 				    p->p_session->s_sid != (pid_t)name[0]) {
1429 					PROC_UNLOCK(p);
1430 					continue;
1431 				}
1432 				break;
1433 
1434 			case KERN_PROC_TTY:
1435 				if ((p->p_flag & P_CONTROLT) == 0 ||
1436 				    p->p_session == NULL) {
1437 					PROC_UNLOCK(p);
1438 					continue;
1439 				}
1440 				/* XXX proctree_lock */
1441 				SESS_LOCK(p->p_session);
1442 				if (p->p_session->s_ttyp == NULL ||
1443 				    tty_udev(p->p_session->s_ttyp) !=
1444 				    (dev_t)name[0]) {
1445 					SESS_UNLOCK(p->p_session);
1446 					PROC_UNLOCK(p);
1447 					continue;
1448 				}
1449 				SESS_UNLOCK(p->p_session);
1450 				break;
1451 
1452 			case KERN_PROC_UID:
1453 				if (p->p_ucred->cr_uid != (uid_t)name[0]) {
1454 					PROC_UNLOCK(p);
1455 					continue;
1456 				}
1457 				break;
1458 
1459 			case KERN_PROC_RUID:
1460 				if (p->p_ucred->cr_ruid != (uid_t)name[0]) {
1461 					PROC_UNLOCK(p);
1462 					continue;
1463 				}
1464 				break;
1465 
1466 			case KERN_PROC_PROC:
1467 				break;
1468 
1469 			default:
1470 				break;
1471 
1472 			}
1473 
1474 			error = sysctl_out_proc(p, req, flags, doingzomb);
1475 			if (error) {
1476 				sx_sunlock(&allproc_lock);
1477 				sx_sunlock(&proctree_lock);
1478 				return (error);
1479 			}
1480 		}
1481 	}
1482 	sx_sunlock(&allproc_lock);
1483 	sx_sunlock(&proctree_lock);
1484 	return (0);
1485 }
1486 
1487 struct pargs *
1488 pargs_alloc(int len)
1489 {
1490 	struct pargs *pa;
1491 
1492 	pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1493 		M_WAITOK);
1494 	refcount_init(&pa->ar_ref, 1);
1495 	pa->ar_length = len;
1496 	return (pa);
1497 }
1498 
1499 static void
1500 pargs_free(struct pargs *pa)
1501 {
1502 
1503 	free(pa, M_PARGS);
1504 }
1505 
1506 void
1507 pargs_hold(struct pargs *pa)
1508 {
1509 
1510 	if (pa == NULL)
1511 		return;
1512 	refcount_acquire(&pa->ar_ref);
1513 }
1514 
1515 void
1516 pargs_drop(struct pargs *pa)
1517 {
1518 
1519 	if (pa == NULL)
1520 		return;
1521 	if (refcount_release(&pa->ar_ref))
1522 		pargs_free(pa);
1523 }
1524 
1525 static int
1526 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1527     size_t len)
1528 {
1529 	ssize_t n;
1530 
1531 	/*
1532 	 * This may return a short read if the string is shorter than the chunk
1533 	 * and is aligned at the end of the page, and the following page is not
1534 	 * mapped.
1535 	 */
1536 	n = proc_readmem(td, p, (vm_offset_t)sptr, buf, len);
1537 	if (n <= 0)
1538 		return (ENOMEM);
1539 	return (0);
1540 }
1541 
1542 #define PROC_AUXV_MAX	256	/* Safety limit on auxv size. */
1543 
1544 enum proc_vector_type {
1545 	PROC_ARG,
1546 	PROC_ENV,
1547 	PROC_AUX,
1548 };
1549 
1550 #ifdef COMPAT_FREEBSD32
1551 static int
1552 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1553     size_t *vsizep, enum proc_vector_type type)
1554 {
1555 	struct freebsd32_ps_strings pss;
1556 	Elf32_Auxinfo aux;
1557 	vm_offset_t vptr, ptr;
1558 	uint32_t *proc_vector32;
1559 	char **proc_vector;
1560 	size_t vsize, size;
1561 	int i, error;
1562 
1563 	error = 0;
1564 	if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss,
1565 	    sizeof(pss)) != sizeof(pss))
1566 		return (ENOMEM);
1567 	switch (type) {
1568 	case PROC_ARG:
1569 		vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1570 		vsize = pss.ps_nargvstr;
1571 		if (vsize > ARG_MAX)
1572 			return (ENOEXEC);
1573 		size = vsize * sizeof(int32_t);
1574 		break;
1575 	case PROC_ENV:
1576 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1577 		vsize = pss.ps_nenvstr;
1578 		if (vsize > ARG_MAX)
1579 			return (ENOEXEC);
1580 		size = vsize * sizeof(int32_t);
1581 		break;
1582 	case PROC_AUX:
1583 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1584 		    (pss.ps_nenvstr + 1) * sizeof(int32_t);
1585 		if (vptr % 4 != 0)
1586 			return (ENOEXEC);
1587 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1588 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1589 			    sizeof(aux))
1590 				return (ENOMEM);
1591 			if (aux.a_type == AT_NULL)
1592 				break;
1593 			ptr += sizeof(aux);
1594 		}
1595 		if (aux.a_type != AT_NULL)
1596 			return (ENOEXEC);
1597 		vsize = i + 1;
1598 		size = vsize * sizeof(aux);
1599 		break;
1600 	default:
1601 		KASSERT(0, ("Wrong proc vector type: %d", type));
1602 		return (EINVAL);
1603 	}
1604 	proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1605 	if (proc_readmem(td, p, vptr, proc_vector32, size) != size) {
1606 		error = ENOMEM;
1607 		goto done;
1608 	}
1609 	if (type == PROC_AUX) {
1610 		*proc_vectorp = (char **)proc_vector32;
1611 		*vsizep = vsize;
1612 		return (0);
1613 	}
1614 	proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1615 	for (i = 0; i < (int)vsize; i++)
1616 		proc_vector[i] = PTRIN(proc_vector32[i]);
1617 	*proc_vectorp = proc_vector;
1618 	*vsizep = vsize;
1619 done:
1620 	free(proc_vector32, M_TEMP);
1621 	return (error);
1622 }
1623 #endif
1624 
1625 static int
1626 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1627     size_t *vsizep, enum proc_vector_type type)
1628 {
1629 	struct ps_strings pss;
1630 	Elf_Auxinfo aux;
1631 	vm_offset_t vptr, ptr;
1632 	char **proc_vector;
1633 	size_t vsize, size;
1634 	int i;
1635 
1636 #ifdef COMPAT_FREEBSD32
1637 	if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1638 		return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1639 #endif
1640 	if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss,
1641 	    sizeof(pss)) != sizeof(pss))
1642 		return (ENOMEM);
1643 	switch (type) {
1644 	case PROC_ARG:
1645 		vptr = (vm_offset_t)pss.ps_argvstr;
1646 		vsize = pss.ps_nargvstr;
1647 		if (vsize > ARG_MAX)
1648 			return (ENOEXEC);
1649 		size = vsize * sizeof(char *);
1650 		break;
1651 	case PROC_ENV:
1652 		vptr = (vm_offset_t)pss.ps_envstr;
1653 		vsize = pss.ps_nenvstr;
1654 		if (vsize > ARG_MAX)
1655 			return (ENOEXEC);
1656 		size = vsize * sizeof(char *);
1657 		break;
1658 	case PROC_AUX:
1659 		/*
1660 		 * The aux array is just above env array on the stack. Check
1661 		 * that the address is naturally aligned.
1662 		 */
1663 		vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1664 		    * sizeof(char *);
1665 #if __ELF_WORD_SIZE == 64
1666 		if (vptr % sizeof(uint64_t) != 0)
1667 #else
1668 		if (vptr % sizeof(uint32_t) != 0)
1669 #endif
1670 			return (ENOEXEC);
1671 		/*
1672 		 * We count the array size reading the aux vectors from the
1673 		 * stack until AT_NULL vector is returned.  So (to keep the code
1674 		 * simple) we read the process stack twice: the first time here
1675 		 * to find the size and the second time when copying the vectors
1676 		 * to the allocated proc_vector.
1677 		 */
1678 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1679 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1680 			    sizeof(aux))
1681 				return (ENOMEM);
1682 			if (aux.a_type == AT_NULL)
1683 				break;
1684 			ptr += sizeof(aux);
1685 		}
1686 		/*
1687 		 * If the PROC_AUXV_MAX entries are iterated over, and we have
1688 		 * not reached AT_NULL, it is most likely we are reading wrong
1689 		 * data: either the process doesn't have auxv array or data has
1690 		 * been modified. Return the error in this case.
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); /* In case we are built without INVARIANTS. */
1700 	}
1701 	proc_vector = malloc(size, M_TEMP, M_WAITOK);
1702 	if (proc_readmem(td, p, vptr, proc_vector, size) != size) {
1703 		free(proc_vector, M_TEMP);
1704 		return (ENOMEM);
1705 	}
1706 	*proc_vectorp = proc_vector;
1707 	*vsizep = vsize;
1708 
1709 	return (0);
1710 }
1711 
1712 #define GET_PS_STRINGS_CHUNK_SZ	256	/* Chunk size (bytes) for ps_strings operations. */
1713 
1714 static int
1715 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
1716     enum proc_vector_type type)
1717 {
1718 	size_t done, len, nchr, vsize;
1719 	int error, i;
1720 	char **proc_vector, *sptr;
1721 	char pss_string[GET_PS_STRINGS_CHUNK_SZ];
1722 
1723 	PROC_ASSERT_HELD(p);
1724 
1725 	/*
1726 	 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
1727 	 */
1728 	nchr = 2 * (PATH_MAX + ARG_MAX);
1729 
1730 	error = get_proc_vector(td, p, &proc_vector, &vsize, type);
1731 	if (error != 0)
1732 		return (error);
1733 	for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
1734 		/*
1735 		 * The program may have scribbled into its argv array, e.g. to
1736 		 * remove some arguments.  If that has happened, break out
1737 		 * before trying to read from NULL.
1738 		 */
1739 		if (proc_vector[i] == NULL)
1740 			break;
1741 		for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
1742 			error = proc_read_string(td, p, sptr, pss_string,
1743 			    sizeof(pss_string));
1744 			if (error != 0)
1745 				goto done;
1746 			len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
1747 			if (done + len >= nchr)
1748 				len = nchr - done - 1;
1749 			sbuf_bcat(sb, pss_string, len);
1750 			if (len != GET_PS_STRINGS_CHUNK_SZ)
1751 				break;
1752 			done += GET_PS_STRINGS_CHUNK_SZ;
1753 		}
1754 		sbuf_bcat(sb, "", 1);
1755 		done += len + 1;
1756 	}
1757 done:
1758 	free(proc_vector, M_TEMP);
1759 	return (error);
1760 }
1761 
1762 int
1763 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
1764 {
1765 
1766 	return (get_ps_strings(curthread, p, sb, PROC_ARG));
1767 }
1768 
1769 int
1770 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
1771 {
1772 
1773 	return (get_ps_strings(curthread, p, sb, PROC_ENV));
1774 }
1775 
1776 int
1777 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb)
1778 {
1779 	size_t vsize, size;
1780 	char **auxv;
1781 	int error;
1782 
1783 	error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX);
1784 	if (error == 0) {
1785 #ifdef COMPAT_FREEBSD32
1786 		if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1787 			size = vsize * sizeof(Elf32_Auxinfo);
1788 		else
1789 #endif
1790 			size = vsize * sizeof(Elf_Auxinfo);
1791 		if (sbuf_bcat(sb, auxv, size) != 0)
1792 			error = ENOMEM;
1793 		free(auxv, M_TEMP);
1794 	}
1795 	return (error);
1796 }
1797 
1798 /*
1799  * This sysctl allows a process to retrieve the argument list or process
1800  * title for another process without groping around in the address space
1801  * of the other process.  It also allow a process to set its own "process
1802  * title to a string of its own choice.
1803  */
1804 static int
1805 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
1806 {
1807 	int *name = (int *)arg1;
1808 	u_int namelen = arg2;
1809 	struct pargs *newpa, *pa;
1810 	struct proc *p;
1811 	struct sbuf sb;
1812 	int flags, error = 0, error2;
1813 
1814 	if (namelen != 1)
1815 		return (EINVAL);
1816 
1817 	flags = PGET_CANSEE;
1818 	if (req->newptr != NULL)
1819 		flags |= PGET_ISCURRENT;
1820 	error = pget((pid_t)name[0], flags, &p);
1821 	if (error)
1822 		return (error);
1823 
1824 	pa = p->p_args;
1825 	if (pa != NULL) {
1826 		pargs_hold(pa);
1827 		PROC_UNLOCK(p);
1828 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
1829 		pargs_drop(pa);
1830 	} else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
1831 		_PHOLD(p);
1832 		PROC_UNLOCK(p);
1833 		sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1834 		sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1835 		error = proc_getargv(curthread, p, &sb);
1836 		error2 = sbuf_finish(&sb);
1837 		PRELE(p);
1838 		sbuf_delete(&sb);
1839 		if (error == 0 && error2 != 0)
1840 			error = error2;
1841 	} else {
1842 		PROC_UNLOCK(p);
1843 	}
1844 	if (error != 0 || req->newptr == NULL)
1845 		return (error);
1846 
1847 	if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit)
1848 		return (ENOMEM);
1849 	newpa = pargs_alloc(req->newlen);
1850 	error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
1851 	if (error != 0) {
1852 		pargs_free(newpa);
1853 		return (error);
1854 	}
1855 	PROC_LOCK(p);
1856 	pa = p->p_args;
1857 	p->p_args = newpa;
1858 	PROC_UNLOCK(p);
1859 	pargs_drop(pa);
1860 	return (0);
1861 }
1862 
1863 /*
1864  * This sysctl allows a process to retrieve environment of another process.
1865  */
1866 static int
1867 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
1868 {
1869 	int *name = (int *)arg1;
1870 	u_int namelen = arg2;
1871 	struct proc *p;
1872 	struct sbuf sb;
1873 	int error, error2;
1874 
1875 	if (namelen != 1)
1876 		return (EINVAL);
1877 
1878 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1879 	if (error != 0)
1880 		return (error);
1881 	if ((p->p_flag & P_SYSTEM) != 0) {
1882 		PRELE(p);
1883 		return (0);
1884 	}
1885 
1886 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1887 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1888 	error = proc_getenvv(curthread, p, &sb);
1889 	error2 = sbuf_finish(&sb);
1890 	PRELE(p);
1891 	sbuf_delete(&sb);
1892 	return (error != 0 ? error : error2);
1893 }
1894 
1895 /*
1896  * This sysctl allows a process to retrieve ELF auxiliary vector of
1897  * another process.
1898  */
1899 static int
1900 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
1901 {
1902 	int *name = (int *)arg1;
1903 	u_int namelen = arg2;
1904 	struct proc *p;
1905 	struct sbuf sb;
1906 	int error, error2;
1907 
1908 	if (namelen != 1)
1909 		return (EINVAL);
1910 
1911 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1912 	if (error != 0)
1913 		return (error);
1914 	if ((p->p_flag & P_SYSTEM) != 0) {
1915 		PRELE(p);
1916 		return (0);
1917 	}
1918 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1919 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1920 	error = proc_getauxv(curthread, p, &sb);
1921 	error2 = sbuf_finish(&sb);
1922 	PRELE(p);
1923 	sbuf_delete(&sb);
1924 	return (error != 0 ? error : error2);
1925 }
1926 
1927 /*
1928  * This sysctl allows a process to retrieve the path of the executable for
1929  * itself or another process.
1930  */
1931 static int
1932 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
1933 {
1934 	pid_t *pidp = (pid_t *)arg1;
1935 	unsigned int arglen = arg2;
1936 	struct proc *p;
1937 	struct vnode *vp;
1938 	char *retbuf, *freebuf;
1939 	int error;
1940 
1941 	if (arglen != 1)
1942 		return (EINVAL);
1943 	if (*pidp == -1) {	/* -1 means this process */
1944 		p = req->td->td_proc;
1945 	} else {
1946 		error = pget(*pidp, PGET_CANSEE, &p);
1947 		if (error != 0)
1948 			return (error);
1949 	}
1950 
1951 	vp = p->p_textvp;
1952 	if (vp == NULL) {
1953 		if (*pidp != -1)
1954 			PROC_UNLOCK(p);
1955 		return (0);
1956 	}
1957 	vref(vp);
1958 	if (*pidp != -1)
1959 		PROC_UNLOCK(p);
1960 	error = vn_fullpath(req->td, vp, &retbuf, &freebuf);
1961 	vrele(vp);
1962 	if (error)
1963 		return (error);
1964 	error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
1965 	free(freebuf, M_TEMP);
1966 	return (error);
1967 }
1968 
1969 static int
1970 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
1971 {
1972 	struct proc *p;
1973 	char *sv_name;
1974 	int *name;
1975 	int namelen;
1976 	int error;
1977 
1978 	namelen = arg2;
1979 	if (namelen != 1)
1980 		return (EINVAL);
1981 
1982 	name = (int *)arg1;
1983 	error = pget((pid_t)name[0], PGET_CANSEE, &p);
1984 	if (error != 0)
1985 		return (error);
1986 	sv_name = p->p_sysent->sv_name;
1987 	PROC_UNLOCK(p);
1988 	return (sysctl_handle_string(oidp, sv_name, 0, req));
1989 }
1990 
1991 #ifdef KINFO_OVMENTRY_SIZE
1992 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
1993 #endif
1994 
1995 #ifdef COMPAT_FREEBSD7
1996 static int
1997 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
1998 {
1999 	vm_map_entry_t entry, tmp_entry;
2000 	unsigned int last_timestamp;
2001 	char *fullpath, *freepath;
2002 	struct kinfo_ovmentry *kve;
2003 	struct vattr va;
2004 	struct ucred *cred;
2005 	int error, *name;
2006 	struct vnode *vp;
2007 	struct proc *p;
2008 	vm_map_t map;
2009 	struct vmspace *vm;
2010 
2011 	name = (int *)arg1;
2012 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2013 	if (error != 0)
2014 		return (error);
2015 	vm = vmspace_acquire_ref(p);
2016 	if (vm == NULL) {
2017 		PRELE(p);
2018 		return (ESRCH);
2019 	}
2020 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2021 
2022 	map = &vm->vm_map;
2023 	vm_map_lock_read(map);
2024 	for (entry = map->header.next; entry != &map->header;
2025 	    entry = entry->next) {
2026 		vm_object_t obj, tobj, lobj;
2027 		vm_offset_t addr;
2028 
2029 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2030 			continue;
2031 
2032 		bzero(kve, sizeof(*kve));
2033 		kve->kve_structsize = sizeof(*kve);
2034 
2035 		kve->kve_private_resident = 0;
2036 		obj = entry->object.vm_object;
2037 		if (obj != NULL) {
2038 			VM_OBJECT_RLOCK(obj);
2039 			if (obj->shadow_count == 1)
2040 				kve->kve_private_resident =
2041 				    obj->resident_page_count;
2042 		}
2043 		kve->kve_resident = 0;
2044 		addr = entry->start;
2045 		while (addr < entry->end) {
2046 			if (pmap_extract(map->pmap, addr))
2047 				kve->kve_resident++;
2048 			addr += PAGE_SIZE;
2049 		}
2050 
2051 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2052 			if (tobj != obj)
2053 				VM_OBJECT_RLOCK(tobj);
2054 			if (lobj != obj)
2055 				VM_OBJECT_RUNLOCK(lobj);
2056 			lobj = tobj;
2057 		}
2058 
2059 		kve->kve_start = (void*)entry->start;
2060 		kve->kve_end = (void*)entry->end;
2061 		kve->kve_offset = (off_t)entry->offset;
2062 
2063 		if (entry->protection & VM_PROT_READ)
2064 			kve->kve_protection |= KVME_PROT_READ;
2065 		if (entry->protection & VM_PROT_WRITE)
2066 			kve->kve_protection |= KVME_PROT_WRITE;
2067 		if (entry->protection & VM_PROT_EXECUTE)
2068 			kve->kve_protection |= KVME_PROT_EXEC;
2069 
2070 		if (entry->eflags & MAP_ENTRY_COW)
2071 			kve->kve_flags |= KVME_FLAG_COW;
2072 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2073 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2074 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2075 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2076 
2077 		last_timestamp = map->timestamp;
2078 		vm_map_unlock_read(map);
2079 
2080 		kve->kve_fileid = 0;
2081 		kve->kve_fsid = 0;
2082 		freepath = NULL;
2083 		fullpath = "";
2084 		if (lobj) {
2085 			vp = NULL;
2086 			switch (lobj->type) {
2087 			case OBJT_DEFAULT:
2088 				kve->kve_type = KVME_TYPE_DEFAULT;
2089 				break;
2090 			case OBJT_VNODE:
2091 				kve->kve_type = KVME_TYPE_VNODE;
2092 				vp = lobj->handle;
2093 				vref(vp);
2094 				break;
2095 			case OBJT_SWAP:
2096 				if ((lobj->flags & OBJ_TMPFS_NODE) != 0) {
2097 					kve->kve_type = KVME_TYPE_VNODE;
2098 					if ((lobj->flags & OBJ_TMPFS) != 0) {
2099 						vp = lobj->un_pager.swp.swp_tmpfs;
2100 						vref(vp);
2101 					}
2102 				} else {
2103 					kve->kve_type = KVME_TYPE_SWAP;
2104 				}
2105 				break;
2106 			case OBJT_DEVICE:
2107 				kve->kve_type = KVME_TYPE_DEVICE;
2108 				break;
2109 			case OBJT_PHYS:
2110 				kve->kve_type = KVME_TYPE_PHYS;
2111 				break;
2112 			case OBJT_DEAD:
2113 				kve->kve_type = KVME_TYPE_DEAD;
2114 				break;
2115 			case OBJT_SG:
2116 				kve->kve_type = KVME_TYPE_SG;
2117 				break;
2118 			default:
2119 				kve->kve_type = KVME_TYPE_UNKNOWN;
2120 				break;
2121 			}
2122 			if (lobj != obj)
2123 				VM_OBJECT_RUNLOCK(lobj);
2124 
2125 			kve->kve_ref_count = obj->ref_count;
2126 			kve->kve_shadow_count = obj->shadow_count;
2127 			VM_OBJECT_RUNLOCK(obj);
2128 			if (vp != NULL) {
2129 				vn_fullpath(curthread, vp, &fullpath,
2130 				    &freepath);
2131 				cred = curthread->td_ucred;
2132 				vn_lock(vp, LK_SHARED | LK_RETRY);
2133 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2134 					kve->kve_fileid = va.va_fileid;
2135 					kve->kve_fsid = va.va_fsid;
2136 				}
2137 				vput(vp);
2138 			}
2139 		} else {
2140 			kve->kve_type = KVME_TYPE_NONE;
2141 			kve->kve_ref_count = 0;
2142 			kve->kve_shadow_count = 0;
2143 		}
2144 
2145 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2146 		if (freepath != NULL)
2147 			free(freepath, M_TEMP);
2148 
2149 		error = SYSCTL_OUT(req, kve, sizeof(*kve));
2150 		vm_map_lock_read(map);
2151 		if (error)
2152 			break;
2153 		if (last_timestamp != map->timestamp) {
2154 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2155 			entry = tmp_entry;
2156 		}
2157 	}
2158 	vm_map_unlock_read(map);
2159 	vmspace_free(vm);
2160 	PRELE(p);
2161 	free(kve, M_TEMP);
2162 	return (error);
2163 }
2164 #endif	/* COMPAT_FREEBSD7 */
2165 
2166 #ifdef KINFO_VMENTRY_SIZE
2167 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2168 #endif
2169 
2170 static void
2171 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry,
2172     struct kinfo_vmentry *kve)
2173 {
2174 	vm_object_t obj, tobj;
2175 	vm_page_t m, m_adv;
2176 	vm_offset_t addr;
2177 	vm_paddr_t locked_pa;
2178 	vm_pindex_t pi, pi_adv, pindex;
2179 
2180 	locked_pa = 0;
2181 	obj = entry->object.vm_object;
2182 	addr = entry->start;
2183 	m_adv = NULL;
2184 	pi = OFF_TO_IDX(entry->offset);
2185 	for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) {
2186 		if (m_adv != NULL) {
2187 			m = m_adv;
2188 		} else {
2189 			pi_adv = OFF_TO_IDX(entry->end - addr);
2190 			pindex = pi;
2191 			for (tobj = obj;; tobj = tobj->backing_object) {
2192 				m = vm_page_find_least(tobj, pindex);
2193 				if (m != NULL) {
2194 					if (m->pindex == pindex)
2195 						break;
2196 					if (pi_adv > m->pindex - pindex) {
2197 						pi_adv = m->pindex - pindex;
2198 						m_adv = m;
2199 					}
2200 				}
2201 				if (tobj->backing_object == NULL)
2202 					goto next;
2203 				pindex += OFF_TO_IDX(tobj->
2204 				    backing_object_offset);
2205 			}
2206 		}
2207 		m_adv = NULL;
2208 		if (m->psind != 0 && addr + pagesizes[1] <= entry->end &&
2209 		    (addr & (pagesizes[1] - 1)) == 0 &&
2210 		    (pmap_mincore(map->pmap, addr, &locked_pa) &
2211 		    MINCORE_SUPER) != 0) {
2212 			kve->kve_flags |= KVME_FLAG_SUPER;
2213 			pi_adv = OFF_TO_IDX(pagesizes[1]);
2214 		} else {
2215 			/*
2216 			 * We do not test the found page on validity.
2217 			 * Either the page is busy and being paged in,
2218 			 * or it was invalidated.  The first case
2219 			 * should be counted as resident, the second
2220 			 * is not so clear; we do account both.
2221 			 */
2222 			pi_adv = 1;
2223 		}
2224 		kve->kve_resident += pi_adv;
2225 next:;
2226 	}
2227 	PA_UNLOCK_COND(locked_pa);
2228 }
2229 
2230 /*
2231  * Must be called with the process locked and will return unlocked.
2232  */
2233 int
2234 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb, ssize_t maxlen, int flags)
2235 {
2236 	vm_map_entry_t entry, tmp_entry;
2237 	struct vattr va;
2238 	vm_map_t map;
2239 	vm_object_t obj, tobj, lobj;
2240 	char *fullpath, *freepath;
2241 	struct kinfo_vmentry *kve;
2242 	struct ucred *cred;
2243 	struct vnode *vp;
2244 	struct vmspace *vm;
2245 	vm_offset_t addr;
2246 	unsigned int last_timestamp;
2247 	int error;
2248 
2249 	PROC_LOCK_ASSERT(p, MA_OWNED);
2250 
2251 	_PHOLD(p);
2252 	PROC_UNLOCK(p);
2253 	vm = vmspace_acquire_ref(p);
2254 	if (vm == NULL) {
2255 		PRELE(p);
2256 		return (ESRCH);
2257 	}
2258 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK | M_ZERO);
2259 
2260 	error = 0;
2261 	map = &vm->vm_map;
2262 	vm_map_lock_read(map);
2263 	for (entry = map->header.next; entry != &map->header;
2264 	    entry = entry->next) {
2265 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2266 			continue;
2267 
2268 		addr = entry->end;
2269 		bzero(kve, sizeof(*kve));
2270 		obj = entry->object.vm_object;
2271 		if (obj != NULL) {
2272 			for (tobj = obj; tobj != NULL;
2273 			    tobj = tobj->backing_object) {
2274 				VM_OBJECT_RLOCK(tobj);
2275 				lobj = tobj;
2276 			}
2277 			if (obj->backing_object == NULL)
2278 				kve->kve_private_resident =
2279 				    obj->resident_page_count;
2280 			if (!vmmap_skip_res_cnt)
2281 				kern_proc_vmmap_resident(map, entry, kve);
2282 			for (tobj = obj; tobj != NULL;
2283 			    tobj = tobj->backing_object) {
2284 				if (tobj != obj && tobj != lobj)
2285 					VM_OBJECT_RUNLOCK(tobj);
2286 			}
2287 		} else {
2288 			lobj = NULL;
2289 		}
2290 
2291 		kve->kve_start = entry->start;
2292 		kve->kve_end = entry->end;
2293 		kve->kve_offset = entry->offset;
2294 
2295 		if (entry->protection & VM_PROT_READ)
2296 			kve->kve_protection |= KVME_PROT_READ;
2297 		if (entry->protection & VM_PROT_WRITE)
2298 			kve->kve_protection |= KVME_PROT_WRITE;
2299 		if (entry->protection & VM_PROT_EXECUTE)
2300 			kve->kve_protection |= KVME_PROT_EXEC;
2301 
2302 		if (entry->eflags & MAP_ENTRY_COW)
2303 			kve->kve_flags |= KVME_FLAG_COW;
2304 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2305 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2306 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2307 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2308 		if (entry->eflags & MAP_ENTRY_GROWS_UP)
2309 			kve->kve_flags |= KVME_FLAG_GROWS_UP;
2310 		if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2311 			kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2312 
2313 		last_timestamp = map->timestamp;
2314 		vm_map_unlock_read(map);
2315 
2316 		freepath = NULL;
2317 		fullpath = "";
2318 		if (lobj != NULL) {
2319 			vp = NULL;
2320 			switch (lobj->type) {
2321 			case OBJT_DEFAULT:
2322 				kve->kve_type = KVME_TYPE_DEFAULT;
2323 				break;
2324 			case OBJT_VNODE:
2325 				kve->kve_type = KVME_TYPE_VNODE;
2326 				vp = lobj->handle;
2327 				vref(vp);
2328 				break;
2329 			case OBJT_SWAP:
2330 				if ((lobj->flags & OBJ_TMPFS_NODE) != 0) {
2331 					kve->kve_type = KVME_TYPE_VNODE;
2332 					if ((lobj->flags & OBJ_TMPFS) != 0) {
2333 						vp = lobj->un_pager.swp.swp_tmpfs;
2334 						vref(vp);
2335 					}
2336 				} else {
2337 					kve->kve_type = KVME_TYPE_SWAP;
2338 				}
2339 				break;
2340 			case OBJT_DEVICE:
2341 				kve->kve_type = KVME_TYPE_DEVICE;
2342 				break;
2343 			case OBJT_PHYS:
2344 				kve->kve_type = KVME_TYPE_PHYS;
2345 				break;
2346 			case OBJT_DEAD:
2347 				kve->kve_type = KVME_TYPE_DEAD;
2348 				break;
2349 			case OBJT_SG:
2350 				kve->kve_type = KVME_TYPE_SG;
2351 				break;
2352 			case OBJT_MGTDEVICE:
2353 				kve->kve_type = KVME_TYPE_MGTDEVICE;
2354 				break;
2355 			default:
2356 				kve->kve_type = KVME_TYPE_UNKNOWN;
2357 				break;
2358 			}
2359 			if (lobj != obj)
2360 				VM_OBJECT_RUNLOCK(lobj);
2361 
2362 			kve->kve_ref_count = obj->ref_count;
2363 			kve->kve_shadow_count = obj->shadow_count;
2364 			VM_OBJECT_RUNLOCK(obj);
2365 			if (vp != NULL) {
2366 				vn_fullpath(curthread, vp, &fullpath,
2367 				    &freepath);
2368 				kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2369 				cred = curthread->td_ucred;
2370 				vn_lock(vp, LK_SHARED | LK_RETRY);
2371 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2372 					kve->kve_vn_fileid = va.va_fileid;
2373 					kve->kve_vn_fsid = va.va_fsid;
2374 					kve->kve_vn_mode =
2375 					    MAKEIMODE(va.va_type, va.va_mode);
2376 					kve->kve_vn_size = va.va_size;
2377 					kve->kve_vn_rdev = va.va_rdev;
2378 					kve->kve_status = KF_ATTR_VALID;
2379 				}
2380 				vput(vp);
2381 			}
2382 		} else {
2383 			kve->kve_type = KVME_TYPE_NONE;
2384 			kve->kve_ref_count = 0;
2385 			kve->kve_shadow_count = 0;
2386 		}
2387 
2388 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2389 		if (freepath != NULL)
2390 			free(freepath, M_TEMP);
2391 
2392 		/* Pack record size down */
2393 		if ((flags & KERN_VMMAP_PACK_KINFO) != 0)
2394 			kve->kve_structsize =
2395 			    offsetof(struct kinfo_vmentry, kve_path) +
2396 			    strlen(kve->kve_path) + 1;
2397 		else
2398 			kve->kve_structsize = sizeof(*kve);
2399 		kve->kve_structsize = roundup(kve->kve_structsize,
2400 		    sizeof(uint64_t));
2401 
2402 		/* Halt filling and truncate rather than exceeding maxlen */
2403 		if (maxlen != -1 && maxlen < kve->kve_structsize) {
2404 			error = 0;
2405 			vm_map_lock_read(map);
2406 			break;
2407 		} else if (maxlen != -1)
2408 			maxlen -= kve->kve_structsize;
2409 
2410 		if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0)
2411 			error = ENOMEM;
2412 		vm_map_lock_read(map);
2413 		if (error != 0)
2414 			break;
2415 		if (last_timestamp != map->timestamp) {
2416 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2417 			entry = tmp_entry;
2418 		}
2419 	}
2420 	vm_map_unlock_read(map);
2421 	vmspace_free(vm);
2422 	PRELE(p);
2423 	free(kve, M_TEMP);
2424 	return (error);
2425 }
2426 
2427 static int
2428 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2429 {
2430 	struct proc *p;
2431 	struct sbuf sb;
2432 	int error, error2, *name;
2433 
2434 	name = (int *)arg1;
2435 	sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req);
2436 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2437 	error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
2438 	if (error != 0) {
2439 		sbuf_delete(&sb);
2440 		return (error);
2441 	}
2442 	error = kern_proc_vmmap_out(p, &sb, -1, KERN_VMMAP_PACK_KINFO);
2443 	error2 = sbuf_finish(&sb);
2444 	sbuf_delete(&sb);
2445 	return (error != 0 ? error : error2);
2446 }
2447 
2448 #if defined(STACK) || defined(DDB)
2449 static int
2450 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2451 {
2452 	struct kinfo_kstack *kkstp;
2453 	int error, i, *name, numthreads;
2454 	lwpid_t *lwpidarray;
2455 	struct thread *td;
2456 	struct stack *st;
2457 	struct sbuf sb;
2458 	struct proc *p;
2459 
2460 	name = (int *)arg1;
2461 	error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2462 	if (error != 0)
2463 		return (error);
2464 
2465 	kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2466 	st = stack_create();
2467 
2468 	lwpidarray = NULL;
2469 	numthreads = 0;
2470 	PROC_LOCK(p);
2471 repeat:
2472 	if (numthreads < p->p_numthreads) {
2473 		if (lwpidarray != NULL) {
2474 			free(lwpidarray, M_TEMP);
2475 			lwpidarray = NULL;
2476 		}
2477 		numthreads = p->p_numthreads;
2478 		PROC_UNLOCK(p);
2479 		lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2480 		    M_WAITOK | M_ZERO);
2481 		PROC_LOCK(p);
2482 		goto repeat;
2483 	}
2484 	i = 0;
2485 
2486 	/*
2487 	 * XXXRW: During the below loop, execve(2) and countless other sorts
2488 	 * of changes could have taken place.  Should we check to see if the
2489 	 * vmspace has been replaced, or the like, in order to prevent
2490 	 * giving a snapshot that spans, say, execve(2), with some threads
2491 	 * before and some after?  Among other things, the credentials could
2492 	 * have changed, in which case the right to extract debug info might
2493 	 * no longer be assured.
2494 	 */
2495 	FOREACH_THREAD_IN_PROC(p, td) {
2496 		KASSERT(i < numthreads,
2497 		    ("sysctl_kern_proc_kstack: numthreads"));
2498 		lwpidarray[i] = td->td_tid;
2499 		i++;
2500 	}
2501 	numthreads = i;
2502 	for (i = 0; i < numthreads; i++) {
2503 		td = thread_find(p, lwpidarray[i]);
2504 		if (td == NULL) {
2505 			continue;
2506 		}
2507 		bzero(kkstp, sizeof(*kkstp));
2508 		(void)sbuf_new(&sb, kkstp->kkst_trace,
2509 		    sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2510 		thread_lock(td);
2511 		kkstp->kkst_tid = td->td_tid;
2512 		if (TD_IS_SWAPPED(td)) {
2513 			kkstp->kkst_state = KKST_STATE_SWAPPED;
2514 		} else if (TD_IS_RUNNING(td)) {
2515 			if (stack_save_td_running(st, td) == 0)
2516 				kkstp->kkst_state = KKST_STATE_STACKOK;
2517 			else
2518 				kkstp->kkst_state = KKST_STATE_RUNNING;
2519 		} else {
2520 			kkstp->kkst_state = KKST_STATE_STACKOK;
2521 			stack_save_td(st, td);
2522 		}
2523 		thread_unlock(td);
2524 		PROC_UNLOCK(p);
2525 		stack_sbuf_print(&sb, st);
2526 		sbuf_finish(&sb);
2527 		sbuf_delete(&sb);
2528 		error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2529 		PROC_LOCK(p);
2530 		if (error)
2531 			break;
2532 	}
2533 	_PRELE(p);
2534 	PROC_UNLOCK(p);
2535 	if (lwpidarray != NULL)
2536 		free(lwpidarray, M_TEMP);
2537 	stack_destroy(st);
2538 	free(kkstp, M_TEMP);
2539 	return (error);
2540 }
2541 #endif
2542 
2543 /*
2544  * This sysctl allows a process to retrieve the full list of groups from
2545  * itself or another process.
2546  */
2547 static int
2548 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2549 {
2550 	pid_t *pidp = (pid_t *)arg1;
2551 	unsigned int arglen = arg2;
2552 	struct proc *p;
2553 	struct ucred *cred;
2554 	int error;
2555 
2556 	if (arglen != 1)
2557 		return (EINVAL);
2558 	if (*pidp == -1) {	/* -1 means this process */
2559 		p = req->td->td_proc;
2560 		PROC_LOCK(p);
2561 	} else {
2562 		error = pget(*pidp, PGET_CANSEE, &p);
2563 		if (error != 0)
2564 			return (error);
2565 	}
2566 
2567 	cred = crhold(p->p_ucred);
2568 	PROC_UNLOCK(p);
2569 
2570 	error = SYSCTL_OUT(req, cred->cr_groups,
2571 	    cred->cr_ngroups * sizeof(gid_t));
2572 	crfree(cred);
2573 	return (error);
2574 }
2575 
2576 /*
2577  * This sysctl allows a process to retrieve or/and set the resource limit for
2578  * another process.
2579  */
2580 static int
2581 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2582 {
2583 	int *name = (int *)arg1;
2584 	u_int namelen = arg2;
2585 	struct rlimit rlim;
2586 	struct proc *p;
2587 	u_int which;
2588 	int flags, error;
2589 
2590 	if (namelen != 2)
2591 		return (EINVAL);
2592 
2593 	which = (u_int)name[1];
2594 	if (which >= RLIM_NLIMITS)
2595 		return (EINVAL);
2596 
2597 	if (req->newptr != NULL && req->newlen != sizeof(rlim))
2598 		return (EINVAL);
2599 
2600 	flags = PGET_HOLD | PGET_NOTWEXIT;
2601 	if (req->newptr != NULL)
2602 		flags |= PGET_CANDEBUG;
2603 	else
2604 		flags |= PGET_CANSEE;
2605 	error = pget((pid_t)name[0], flags, &p);
2606 	if (error != 0)
2607 		return (error);
2608 
2609 	/*
2610 	 * Retrieve limit.
2611 	 */
2612 	if (req->oldptr != NULL) {
2613 		PROC_LOCK(p);
2614 		lim_rlimit_proc(p, which, &rlim);
2615 		PROC_UNLOCK(p);
2616 	}
2617 	error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
2618 	if (error != 0)
2619 		goto errout;
2620 
2621 	/*
2622 	 * Set limit.
2623 	 */
2624 	if (req->newptr != NULL) {
2625 		error = SYSCTL_IN(req, &rlim, sizeof(rlim));
2626 		if (error == 0)
2627 			error = kern_proc_setrlimit(curthread, p, which, &rlim);
2628 	}
2629 
2630 errout:
2631 	PRELE(p);
2632 	return (error);
2633 }
2634 
2635 /*
2636  * This sysctl allows a process to retrieve ps_strings structure location of
2637  * another process.
2638  */
2639 static int
2640 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
2641 {
2642 	int *name = (int *)arg1;
2643 	u_int namelen = arg2;
2644 	struct proc *p;
2645 	vm_offset_t ps_strings;
2646 	int error;
2647 #ifdef COMPAT_FREEBSD32
2648 	uint32_t ps_strings32;
2649 #endif
2650 
2651 	if (namelen != 1)
2652 		return (EINVAL);
2653 
2654 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2655 	if (error != 0)
2656 		return (error);
2657 #ifdef COMPAT_FREEBSD32
2658 	if ((req->flags & SCTL_MASK32) != 0) {
2659 		/*
2660 		 * We return 0 if the 32 bit emulation request is for a 64 bit
2661 		 * process.
2662 		 */
2663 		ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
2664 		    PTROUT(p->p_sysent->sv_psstrings) : 0;
2665 		PROC_UNLOCK(p);
2666 		error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
2667 		return (error);
2668 	}
2669 #endif
2670 	ps_strings = p->p_sysent->sv_psstrings;
2671 	PROC_UNLOCK(p);
2672 	error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
2673 	return (error);
2674 }
2675 
2676 /*
2677  * This sysctl allows a process to retrieve umask of another process.
2678  */
2679 static int
2680 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
2681 {
2682 	int *name = (int *)arg1;
2683 	u_int namelen = arg2;
2684 	struct proc *p;
2685 	int error;
2686 	u_short fd_cmask;
2687 
2688 	if (namelen != 1)
2689 		return (EINVAL);
2690 
2691 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2692 	if (error != 0)
2693 		return (error);
2694 
2695 	FILEDESC_SLOCK(p->p_fd);
2696 	fd_cmask = p->p_fd->fd_cmask;
2697 	FILEDESC_SUNLOCK(p->p_fd);
2698 	PRELE(p);
2699 	error = SYSCTL_OUT(req, &fd_cmask, sizeof(fd_cmask));
2700 	return (error);
2701 }
2702 
2703 /*
2704  * This sysctl allows a process to set and retrieve binary osreldate of
2705  * another process.
2706  */
2707 static int
2708 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
2709 {
2710 	int *name = (int *)arg1;
2711 	u_int namelen = arg2;
2712 	struct proc *p;
2713 	int flags, error, osrel;
2714 
2715 	if (namelen != 1)
2716 		return (EINVAL);
2717 
2718 	if (req->newptr != NULL && req->newlen != sizeof(osrel))
2719 		return (EINVAL);
2720 
2721 	flags = PGET_HOLD | PGET_NOTWEXIT;
2722 	if (req->newptr != NULL)
2723 		flags |= PGET_CANDEBUG;
2724 	else
2725 		flags |= PGET_CANSEE;
2726 	error = pget((pid_t)name[0], flags, &p);
2727 	if (error != 0)
2728 		return (error);
2729 
2730 	error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
2731 	if (error != 0)
2732 		goto errout;
2733 
2734 	if (req->newptr != NULL) {
2735 		error = SYSCTL_IN(req, &osrel, sizeof(osrel));
2736 		if (error != 0)
2737 			goto errout;
2738 		if (osrel < 0) {
2739 			error = EINVAL;
2740 			goto errout;
2741 		}
2742 		p->p_osrel = osrel;
2743 	}
2744 errout:
2745 	PRELE(p);
2746 	return (error);
2747 }
2748 
2749 static int
2750 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS)
2751 {
2752 	int *name = (int *)arg1;
2753 	u_int namelen = arg2;
2754 	struct proc *p;
2755 	struct kinfo_sigtramp kst;
2756 	const struct sysentvec *sv;
2757 	int error;
2758 #ifdef COMPAT_FREEBSD32
2759 	struct kinfo_sigtramp32 kst32;
2760 #endif
2761 
2762 	if (namelen != 1)
2763 		return (EINVAL);
2764 
2765 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2766 	if (error != 0)
2767 		return (error);
2768 	sv = p->p_sysent;
2769 #ifdef COMPAT_FREEBSD32
2770 	if ((req->flags & SCTL_MASK32) != 0) {
2771 		bzero(&kst32, sizeof(kst32));
2772 		if (SV_PROC_FLAG(p, SV_ILP32)) {
2773 			if (sv->sv_sigcode_base != 0) {
2774 				kst32.ksigtramp_start = sv->sv_sigcode_base;
2775 				kst32.ksigtramp_end = sv->sv_sigcode_base +
2776 				    *sv->sv_szsigcode;
2777 			} else {
2778 				kst32.ksigtramp_start = sv->sv_psstrings -
2779 				    *sv->sv_szsigcode;
2780 				kst32.ksigtramp_end = sv->sv_psstrings;
2781 			}
2782 		}
2783 		PROC_UNLOCK(p);
2784 		error = SYSCTL_OUT(req, &kst32, sizeof(kst32));
2785 		return (error);
2786 	}
2787 #endif
2788 	bzero(&kst, sizeof(kst));
2789 	if (sv->sv_sigcode_base != 0) {
2790 		kst.ksigtramp_start = (char *)sv->sv_sigcode_base;
2791 		kst.ksigtramp_end = (char *)sv->sv_sigcode_base +
2792 		    *sv->sv_szsigcode;
2793 	} else {
2794 		kst.ksigtramp_start = (char *)sv->sv_psstrings -
2795 		    *sv->sv_szsigcode;
2796 		kst.ksigtramp_end = (char *)sv->sv_psstrings;
2797 	}
2798 	PROC_UNLOCK(p);
2799 	error = SYSCTL_OUT(req, &kst, sizeof(kst));
2800 	return (error);
2801 }
2802 
2803 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD,  0, "Process table");
2804 
2805 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
2806 	CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
2807 	"Return entire process table");
2808 
2809 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2810 	sysctl_kern_proc, "Process table");
2811 
2812 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
2813 	sysctl_kern_proc, "Process table");
2814 
2815 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2816 	sysctl_kern_proc, "Process table");
2817 
2818 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
2819 	CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2820 
2821 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
2822 	sysctl_kern_proc, "Process table");
2823 
2824 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2825 	sysctl_kern_proc, "Process table");
2826 
2827 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2828 	sysctl_kern_proc, "Process table");
2829 
2830 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2831 	sysctl_kern_proc, "Process table");
2832 
2833 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
2834 	sysctl_kern_proc, "Return process table, no threads");
2835 
2836 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
2837 	CTLFLAG_RW | CTLFLAG_CAPWR | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
2838 	sysctl_kern_proc_args, "Process argument list");
2839 
2840 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
2841 	sysctl_kern_proc_env, "Process environment");
2842 
2843 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
2844 	CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
2845 
2846 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
2847 	CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
2848 
2849 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
2850 	CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
2851 	"Process syscall vector name (ABI type)");
2852 
2853 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
2854 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2855 
2856 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
2857 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2858 
2859 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
2860 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2861 
2862 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
2863 	sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2864 
2865 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
2866 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2867 
2868 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
2869 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2870 
2871 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
2872 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2873 
2874 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
2875 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2876 
2877 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
2878 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
2879 	"Return process table, no threads");
2880 
2881 #ifdef COMPAT_FREEBSD7
2882 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
2883 	CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
2884 #endif
2885 
2886 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
2887 	CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
2888 
2889 #if defined(STACK) || defined(DDB)
2890 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
2891 	CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
2892 #endif
2893 
2894 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
2895 	CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
2896 
2897 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
2898 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
2899 	"Process resource limits");
2900 
2901 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
2902 	CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
2903 	"Process ps_strings location");
2904 
2905 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
2906 	CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
2907 
2908 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
2909 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
2910 	"Process binary osreldate");
2911 
2912 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD |
2913 	CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp,
2914 	"Process signal trampoline location");
2915 
2916 int allproc_gen;
2917 
2918 void
2919 stop_all_proc(void)
2920 {
2921 	struct proc *cp, *p;
2922 	int r, gen;
2923 	bool restart, seen_stopped, seen_exiting, stopped_some;
2924 
2925 	cp = curproc;
2926 	/*
2927 	 * stop_all_proc() assumes that all process which have
2928 	 * usermode must be stopped, except current process, for
2929 	 * obvious reasons.  Since other threads in the process
2930 	 * establishing global stop could unstop something, disable
2931 	 * calls from multithreaded processes as precaution.  The
2932 	 * service must not be user-callable anyway.
2933 	 */
2934 	KASSERT((cp->p_flag & P_HADTHREADS) == 0 ||
2935 	    (cp->p_flag & P_KTHREAD) != 0, ("mt stop_all_proc"));
2936 
2937 allproc_loop:
2938 	sx_xlock(&allproc_lock);
2939 	gen = allproc_gen;
2940 	seen_exiting = seen_stopped = stopped_some = restart = false;
2941 	LIST_REMOVE(cp, p_list);
2942 	LIST_INSERT_HEAD(&allproc, cp, p_list);
2943 	for (;;) {
2944 		p = LIST_NEXT(cp, p_list);
2945 		if (p == NULL)
2946 			break;
2947 		LIST_REMOVE(cp, p_list);
2948 		LIST_INSERT_AFTER(p, cp, p_list);
2949 		PROC_LOCK(p);
2950 		if ((p->p_flag & (P_KTHREAD | P_SYSTEM |
2951 		    P_TOTAL_STOP)) != 0) {
2952 			PROC_UNLOCK(p);
2953 			continue;
2954 		}
2955 		if ((p->p_flag & P_WEXIT) != 0) {
2956 			seen_exiting = true;
2957 			PROC_UNLOCK(p);
2958 			continue;
2959 		}
2960 		if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) {
2961 			/*
2962 			 * Stopped processes are tolerated when there
2963 			 * are no other processes which might continue
2964 			 * them.  P_STOPPED_SINGLE but not
2965 			 * P_TOTAL_STOP process still has at least one
2966 			 * thread running.
2967 			 */
2968 			seen_stopped = true;
2969 			PROC_UNLOCK(p);
2970 			continue;
2971 		}
2972 		_PHOLD(p);
2973 		sx_xunlock(&allproc_lock);
2974 		r = thread_single(p, SINGLE_ALLPROC);
2975 		if (r != 0)
2976 			restart = true;
2977 		else
2978 			stopped_some = true;
2979 		_PRELE(p);
2980 		PROC_UNLOCK(p);
2981 		sx_xlock(&allproc_lock);
2982 	}
2983 	/* Catch forked children we did not see in iteration. */
2984 	if (gen != allproc_gen)
2985 		restart = true;
2986 	sx_xunlock(&allproc_lock);
2987 	if (restart || stopped_some || seen_exiting || seen_stopped) {
2988 		kern_yield(PRI_USER);
2989 		goto allproc_loop;
2990 	}
2991 }
2992 
2993 void
2994 resume_all_proc(void)
2995 {
2996 	struct proc *cp, *p;
2997 
2998 	cp = curproc;
2999 	sx_xlock(&allproc_lock);
3000 	LIST_REMOVE(cp, p_list);
3001 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3002 	for (;;) {
3003 		p = LIST_NEXT(cp, p_list);
3004 		if (p == NULL)
3005 			break;
3006 		LIST_REMOVE(cp, p_list);
3007 		LIST_INSERT_AFTER(p, cp, p_list);
3008 		PROC_LOCK(p);
3009 		if ((p->p_flag & P_TOTAL_STOP) != 0) {
3010 			sx_xunlock(&allproc_lock);
3011 			_PHOLD(p);
3012 			thread_single_end(p, SINGLE_ALLPROC);
3013 			_PRELE(p);
3014 			PROC_UNLOCK(p);
3015 			sx_xlock(&allproc_lock);
3016 		} else {
3017 			PROC_UNLOCK(p);
3018 		}
3019 	}
3020 	sx_xunlock(&allproc_lock);
3021 }
3022 
3023 #define	TOTAL_STOP_DEBUG	1
3024 #ifdef TOTAL_STOP_DEBUG
3025 volatile static int ap_resume;
3026 #include <sys/mount.h>
3027 
3028 static int
3029 sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS)
3030 {
3031 	int error, val;
3032 
3033 	val = 0;
3034 	ap_resume = 0;
3035 	error = sysctl_handle_int(oidp, &val, 0, req);
3036 	if (error != 0 || req->newptr == NULL)
3037 		return (error);
3038 	if (val != 0) {
3039 		stop_all_proc();
3040 		syncer_suspend();
3041 		while (ap_resume == 0)
3042 			;
3043 		syncer_resume();
3044 		resume_all_proc();
3045 	}
3046 	return (0);
3047 }
3048 
3049 SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW |
3050     CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0,
3051     sysctl_debug_stop_all_proc, "I",
3052     "");
3053 #endif
3054