xref: /freebsd/sys/kern/kern_proc.c (revision c8b31033c3971b2b7349804ffda0cea5e4835b40)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1982, 1986, 1989, 1991, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  */
31 
32 #include <sys/cdefs.h>
33 #include "opt_ddb.h"
34 #include "opt_ktrace.h"
35 #include "opt_kstack_pages.h"
36 #include "opt_stack.h"
37 
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/bitstring.h>
41 #include <sys/elf.h>
42 #include <sys/eventhandler.h>
43 #include <sys/exec.h>
44 #include <sys/fcntl.h>
45 #include <sys/ipc.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/namei.h>
56 #include <sys/proc.h>
57 #include <sys/ptrace.h>
58 #include <sys/refcount.h>
59 #include <sys/resourcevar.h>
60 #include <sys/rwlock.h>
61 #include <sys/sbuf.h>
62 #include <sys/sysent.h>
63 #include <sys/sched.h>
64 #include <sys/shm.h>
65 #include <sys/smp.h>
66 #include <sys/stack.h>
67 #include <sys/stat.h>
68 #include <sys/dtrace_bsd.h>
69 #include <sys/sysctl.h>
70 #include <sys/filedesc.h>
71 #include <sys/tty.h>
72 #include <sys/signalvar.h>
73 #include <sys/sdt.h>
74 #include <sys/sx.h>
75 #include <sys/user.h>
76 #include <sys/vnode.h>
77 #include <sys/wait.h>
78 #ifdef KTRACE
79 #include <sys/ktrace.h>
80 #endif
81 
82 #ifdef DDB
83 #include <ddb/ddb.h>
84 #endif
85 
86 #include <vm/vm.h>
87 #include <vm/vm_param.h>
88 #include <vm/vm_extern.h>
89 #include <vm/pmap.h>
90 #include <vm/vm_map.h>
91 #include <vm/vm_object.h>
92 #include <vm/vm_page.h>
93 #include <vm/uma.h>
94 
95 #include <fs/devfs/devfs.h>
96 
97 #ifdef COMPAT_FREEBSD32
98 #include <compat/freebsd32/freebsd32.h>
99 #include <compat/freebsd32/freebsd32_util.h>
100 #endif
101 
102 SDT_PROVIDER_DEFINE(proc);
103 
104 MALLOC_DEFINE(M_SESSION, "session", "session header");
105 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
106 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
107 
108 static void doenterpgrp(struct proc *, struct pgrp *);
109 static void orphanpg(struct pgrp *pg);
110 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp);
111 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp);
112 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp,
113     int preferthread);
114 static void pgdelete(struct pgrp *);
115 static int pgrp_init(void *mem, int size, int flags);
116 static int proc_ctor(void *mem, int size, void *arg, int flags);
117 static void proc_dtor(void *mem, int size, void *arg);
118 static int proc_init(void *mem, int size, int flags);
119 static void proc_fini(void *mem, int size);
120 static void pargs_free(struct pargs *pa);
121 
122 /*
123  * Other process lists
124  */
125 struct pidhashhead *pidhashtbl = NULL;
126 struct sx *pidhashtbl_lock;
127 u_long pidhash;
128 u_long pidhashlock;
129 struct pgrphashhead *pgrphashtbl;
130 u_long pgrphash;
131 struct proclist allproc = LIST_HEAD_INITIALIZER(allproc);
132 struct sx __exclusive_cache_line allproc_lock;
133 struct sx __exclusive_cache_line proctree_lock;
134 struct mtx __exclusive_cache_line ppeers_lock;
135 struct mtx __exclusive_cache_line procid_lock;
136 uma_zone_t proc_zone;
137 uma_zone_t pgrp_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_hash = offsetof(struct proc, p_hash);
148 const int proc_off_p_threads = offsetof(struct proc, p_threads);
149 const int thread_off_td_tid = offsetof(struct thread, td_tid);
150 const int thread_off_td_name = offsetof(struct thread, td_name);
151 const int thread_off_td_oncpu = offsetof(struct thread, td_oncpu);
152 const int thread_off_td_pcb = offsetof(struct thread, td_pcb);
153 const int thread_off_td_plist = offsetof(struct thread, td_plist);
154 
155 EVENTHANDLER_LIST_DEFINE(process_ctor);
156 EVENTHANDLER_LIST_DEFINE(process_dtor);
157 EVENTHANDLER_LIST_DEFINE(process_init);
158 EVENTHANDLER_LIST_DEFINE(process_fini);
159 EVENTHANDLER_LIST_DEFINE(process_exit);
160 EVENTHANDLER_LIST_DEFINE(process_fork);
161 EVENTHANDLER_LIST_DEFINE(process_exec);
162 
163 int kstack_pages = KSTACK_PAGES;
164 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RDTUN | CTLFLAG_NOFETCH,
165     &kstack_pages, 0,
166     "Kernel stack size in pages");
167 static int vmmap_skip_res_cnt = 0;
168 SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW,
169     &vmmap_skip_res_cnt, 0,
170     "Skip calculation of the pages resident count in kern.proc.vmmap");
171 
172 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
173 #ifdef COMPAT_FREEBSD32
174 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
175 #endif
176 
177 /*
178  * Initialize global process hashing structures.
179  */
180 void
181 procinit(void)
182 {
183 	u_long i;
184 
185 	sx_init(&allproc_lock, "allproc");
186 	sx_init(&proctree_lock, "proctree");
187 	mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
188 	mtx_init(&procid_lock, "procid", NULL, MTX_DEF);
189 	pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
190 	pidhashlock = (pidhash + 1) / 64;
191 	if (pidhashlock > 0)
192 		pidhashlock--;
193 	pidhashtbl_lock = malloc(sizeof(*pidhashtbl_lock) * (pidhashlock + 1),
194 	    M_PROC, M_WAITOK | M_ZERO);
195 	for (i = 0; i < pidhashlock + 1; i++)
196 		sx_init_flags(&pidhashtbl_lock[i], "pidhash", SX_DUPOK);
197 	pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
198 	proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
199 	    proc_ctor, proc_dtor, proc_init, proc_fini,
200 	    UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
201 	pgrp_zone = uma_zcreate("PGRP", sizeof(struct pgrp), NULL, NULL,
202 	    pgrp_init, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
203 	uihashinit();
204 }
205 
206 /*
207  * Prepare a proc for use.
208  */
209 static int
210 proc_ctor(void *mem, int size, void *arg, int flags)
211 {
212 	struct proc *p;
213 	struct thread *td;
214 
215 	p = (struct proc *)mem;
216 #ifdef KDTRACE_HOOKS
217 	kdtrace_proc_ctor(p);
218 #endif
219 	EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
220 	td = FIRST_THREAD_IN_PROC(p);
221 	if (td != NULL) {
222 		/* Make sure all thread constructors are executed */
223 		EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
224 	}
225 	return (0);
226 }
227 
228 /*
229  * Reclaim a proc after use.
230  */
231 static void
232 proc_dtor(void *mem, int size, void *arg)
233 {
234 	struct proc *p;
235 	struct thread *td;
236 
237 	/* INVARIANTS checks go here */
238 	p = (struct proc *)mem;
239 	td = FIRST_THREAD_IN_PROC(p);
240 	if (td != NULL) {
241 #ifdef INVARIANTS
242 		KASSERT((p->p_numthreads == 1),
243 		    ("bad number of threads in exiting process"));
244 		KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
245 #endif
246 		/* Free all OSD associated to this thread. */
247 		osd_thread_exit(td);
248 		ast_kclear(td);
249 
250 		/* Make sure all thread destructors are executed */
251 		EVENTHANDLER_DIRECT_INVOKE(thread_dtor, td);
252 	}
253 	EVENTHANDLER_DIRECT_INVOKE(process_dtor, p);
254 #ifdef KDTRACE_HOOKS
255 	kdtrace_proc_dtor(p);
256 #endif
257 	if (p->p_ksi != NULL)
258 		KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
259 }
260 
261 /*
262  * Initialize type-stable parts of a proc (when newly created).
263  */
264 static int
265 proc_init(void *mem, int size, int flags)
266 {
267 	struct proc *p;
268 
269 	p = (struct proc *)mem;
270 	mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK | MTX_NEW);
271 	mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_NEW);
272 	mtx_init(&p->p_statmtx, "pstatl", NULL, MTX_SPIN | MTX_NEW);
273 	mtx_init(&p->p_itimmtx, "pitiml", NULL, MTX_SPIN | MTX_NEW);
274 	mtx_init(&p->p_profmtx, "pprofl", NULL, MTX_SPIN | MTX_NEW);
275 	cv_init(&p->p_pwait, "ppwait");
276 	TAILQ_INIT(&p->p_threads);	     /* all threads in proc */
277 	EVENTHANDLER_DIRECT_INVOKE(process_init, p);
278 	p->p_stats = pstats_alloc();
279 	p->p_pgrp = NULL;
280 	TAILQ_INIT(&p->p_kqtim_stop);
281 	return (0);
282 }
283 
284 /*
285  * UMA should ensure that this function is never called.
286  * Freeing a proc structure would violate type stability.
287  */
288 static void
289 proc_fini(void *mem, int size)
290 {
291 #ifdef notnow
292 	struct proc *p;
293 
294 	p = (struct proc *)mem;
295 	EVENTHANDLER_DIRECT_INVOKE(process_fini, p);
296 	pstats_free(p->p_stats);
297 	thread_free(FIRST_THREAD_IN_PROC(p));
298 	mtx_destroy(&p->p_mtx);
299 	if (p->p_ksi != NULL)
300 		ksiginfo_free(p->p_ksi);
301 #else
302 	panic("proc reclaimed");
303 #endif
304 }
305 
306 static int
307 pgrp_init(void *mem, int size, int flags)
308 {
309 	struct pgrp *pg;
310 
311 	pg = mem;
312 	mtx_init(&pg->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
313 	sx_init(&pg->pg_killsx, "killpg racer");
314 	return (0);
315 }
316 
317 /*
318  * PID space management.
319  *
320  * These bitmaps are used by fork_findpid.
321  */
322 bitstr_t bit_decl(proc_id_pidmap, PID_MAX);
323 bitstr_t bit_decl(proc_id_grpidmap, PID_MAX);
324 bitstr_t bit_decl(proc_id_sessidmap, PID_MAX);
325 bitstr_t bit_decl(proc_id_reapmap, PID_MAX);
326 
327 static bitstr_t *proc_id_array[] = {
328 	proc_id_pidmap,
329 	proc_id_grpidmap,
330 	proc_id_sessidmap,
331 	proc_id_reapmap,
332 };
333 
334 void
335 proc_id_set(int type, pid_t id)
336 {
337 
338 	KASSERT(type >= 0 && type < nitems(proc_id_array),
339 	    ("invalid type %d\n", type));
340 	mtx_lock(&procid_lock);
341 	KASSERT(bit_test(proc_id_array[type], id) == 0,
342 	    ("bit %d already set in %d\n", id, type));
343 	bit_set(proc_id_array[type], id);
344 	mtx_unlock(&procid_lock);
345 }
346 
347 void
348 proc_id_set_cond(int type, pid_t id)
349 {
350 
351 	KASSERT(type >= 0 && type < nitems(proc_id_array),
352 	    ("invalid type %d\n", type));
353 	if (bit_test(proc_id_array[type], id))
354 		return;
355 	mtx_lock(&procid_lock);
356 	bit_set(proc_id_array[type], id);
357 	mtx_unlock(&procid_lock);
358 }
359 
360 void
361 proc_id_clear(int type, pid_t id)
362 {
363 
364 	KASSERT(type >= 0 && type < nitems(proc_id_array),
365 	    ("invalid type %d\n", type));
366 	mtx_lock(&procid_lock);
367 	KASSERT(bit_test(proc_id_array[type], id) != 0,
368 	    ("bit %d not set in %d\n", id, type));
369 	bit_clear(proc_id_array[type], id);
370 	mtx_unlock(&procid_lock);
371 }
372 
373 /*
374  * Is p an inferior of the current process?
375  */
376 int
377 inferior(struct proc *p)
378 {
379 
380 	sx_assert(&proctree_lock, SX_LOCKED);
381 	PROC_LOCK_ASSERT(p, MA_OWNED);
382 	for (; p != curproc; p = proc_realparent(p)) {
383 		if (p->p_pid == 0)
384 			return (0);
385 	}
386 	return (1);
387 }
388 
389 /*
390  * Shared lock all the pid hash lists.
391  */
392 void
393 pidhash_slockall(void)
394 {
395 	u_long i;
396 
397 	for (i = 0; i < pidhashlock + 1; i++)
398 		sx_slock(&pidhashtbl_lock[i]);
399 }
400 
401 /*
402  * Shared unlock all the pid hash lists.
403  */
404 void
405 pidhash_sunlockall(void)
406 {
407 	u_long i;
408 
409 	for (i = 0; i < pidhashlock + 1; i++)
410 		sx_sunlock(&pidhashtbl_lock[i]);
411 }
412 
413 /*
414  * Similar to pfind_any(), this function finds zombies.
415  */
416 struct proc *
417 pfind_any_locked(pid_t pid)
418 {
419 	struct proc *p;
420 
421 	sx_assert(PIDHASHLOCK(pid), SX_LOCKED);
422 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
423 		if (p->p_pid == pid) {
424 			PROC_LOCK(p);
425 			if (p->p_state == PRS_NEW) {
426 				PROC_UNLOCK(p);
427 				p = NULL;
428 			}
429 			break;
430 		}
431 	}
432 	return (p);
433 }
434 
435 /*
436  * Locate a process by number.
437  *
438  * By not returning processes in the PRS_NEW state, we allow callers to avoid
439  * testing for that condition to avoid dereferencing p_ucred, et al.
440  */
441 static __always_inline struct proc *
442 _pfind(pid_t pid, bool zombie)
443 {
444 	struct proc *p;
445 
446 	p = curproc;
447 	if (p->p_pid == pid) {
448 		PROC_LOCK(p);
449 		return (p);
450 	}
451 	sx_slock(PIDHASHLOCK(pid));
452 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
453 		if (p->p_pid == pid) {
454 			PROC_LOCK(p);
455 			if (p->p_state == PRS_NEW ||
456 			    (!zombie && p->p_state == PRS_ZOMBIE)) {
457 				PROC_UNLOCK(p);
458 				p = NULL;
459 			}
460 			break;
461 		}
462 	}
463 	sx_sunlock(PIDHASHLOCK(pid));
464 	return (p);
465 }
466 
467 struct proc *
468 pfind(pid_t pid)
469 {
470 
471 	return (_pfind(pid, false));
472 }
473 
474 /*
475  * Same as pfind but allow zombies.
476  */
477 struct proc *
478 pfind_any(pid_t pid)
479 {
480 
481 	return (_pfind(pid, true));
482 }
483 
484 /*
485  * Locate a process group by number.
486  * The caller must hold proctree_lock.
487  */
488 struct pgrp *
489 pgfind(pid_t pgid)
490 {
491 	struct pgrp *pgrp;
492 
493 	sx_assert(&proctree_lock, SX_LOCKED);
494 
495 	LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
496 		if (pgrp->pg_id == pgid) {
497 			PGRP_LOCK(pgrp);
498 			return (pgrp);
499 		}
500 	}
501 	return (NULL);
502 }
503 
504 /*
505  * Locate process and do additional manipulations, depending on flags.
506  */
507 int
508 pget(pid_t pid, int flags, struct proc **pp)
509 {
510 	struct proc *p;
511 	struct thread *td1;
512 	int error;
513 
514 	p = curproc;
515 	if (p->p_pid == pid) {
516 		PROC_LOCK(p);
517 	} else {
518 		p = NULL;
519 		if (pid <= PID_MAX) {
520 			if ((flags & PGET_NOTWEXIT) == 0)
521 				p = pfind_any(pid);
522 			else
523 				p = pfind(pid);
524 		} else if ((flags & PGET_NOTID) == 0) {
525 			td1 = tdfind(pid, -1);
526 			if (td1 != NULL)
527 				p = td1->td_proc;
528 		}
529 		if (p == NULL)
530 			return (ESRCH);
531 		if ((flags & PGET_CANSEE) != 0) {
532 			error = p_cansee(curthread, p);
533 			if (error != 0)
534 				goto errout;
535 		}
536 	}
537 	if ((flags & PGET_CANDEBUG) != 0) {
538 		error = p_candebug(curthread, p);
539 		if (error != 0)
540 			goto errout;
541 	}
542 	if ((flags & PGET_ISCURRENT) != 0 && curproc != p) {
543 		error = EPERM;
544 		goto errout;
545 	}
546 	if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) {
547 		error = ESRCH;
548 		goto errout;
549 	}
550 	if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) {
551 		/*
552 		 * XXXRW: Not clear ESRCH is the right error during proc
553 		 * execve().
554 		 */
555 		error = ESRCH;
556 		goto errout;
557 	}
558 	if ((flags & PGET_HOLD) != 0) {
559 		_PHOLD(p);
560 		PROC_UNLOCK(p);
561 	}
562 	*pp = p;
563 	return (0);
564 errout:
565 	PROC_UNLOCK(p);
566 	return (error);
567 }
568 
569 /*
570  * Create a new process group.
571  * pgid must be equal to the pid of p.
572  * Begin a new session if required.
573  */
574 int
575 enterpgrp(struct proc *p, pid_t pgid, struct pgrp *pgrp, struct session *sess)
576 {
577 	struct pgrp *old_pgrp;
578 
579 	sx_assert(&proctree_lock, SX_XLOCKED);
580 
581 	KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
582 	KASSERT(p->p_pid == pgid,
583 	    ("enterpgrp: new pgrp and pid != pgid"));
584 	KASSERT(pgfind(pgid) == NULL,
585 	    ("enterpgrp: pgrp with pgid exists"));
586 	KASSERT(!SESS_LEADER(p),
587 	    ("enterpgrp: session leader attempted setpgrp"));
588 
589 	old_pgrp = p->p_pgrp;
590 	if (!sx_try_xlock(&old_pgrp->pg_killsx)) {
591 		sx_xunlock(&proctree_lock);
592 		sx_xlock(&old_pgrp->pg_killsx);
593 		sx_xunlock(&old_pgrp->pg_killsx);
594 		return (ERESTART);
595 	}
596 	MPASS(old_pgrp == p->p_pgrp);
597 
598 	if (sess != NULL) {
599 		/*
600 		 * new session
601 		 */
602 		mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
603 		PROC_LOCK(p);
604 		p->p_flag &= ~P_CONTROLT;
605 		PROC_UNLOCK(p);
606 		PGRP_LOCK(pgrp);
607 		sess->s_leader = p;
608 		sess->s_sid = p->p_pid;
609 		proc_id_set(PROC_ID_SESSION, p->p_pid);
610 		refcount_init(&sess->s_count, 1);
611 		sess->s_ttyvp = NULL;
612 		sess->s_ttydp = NULL;
613 		sess->s_ttyp = NULL;
614 		bcopy(p->p_session->s_login, sess->s_login,
615 			    sizeof(sess->s_login));
616 		pgrp->pg_session = sess;
617 		KASSERT(p == curproc,
618 		    ("enterpgrp: mksession and p != curproc"));
619 	} else {
620 		pgrp->pg_session = p->p_session;
621 		sess_hold(pgrp->pg_session);
622 		PGRP_LOCK(pgrp);
623 	}
624 	pgrp->pg_id = pgid;
625 	proc_id_set(PROC_ID_GROUP, p->p_pid);
626 	LIST_INIT(&pgrp->pg_members);
627 	pgrp->pg_flags = 0;
628 
629 	/*
630 	 * As we have an exclusive lock of proctree_lock,
631 	 * this should not deadlock.
632 	 */
633 	LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
634 	SLIST_INIT(&pgrp->pg_sigiolst);
635 	PGRP_UNLOCK(pgrp);
636 
637 	doenterpgrp(p, pgrp);
638 
639 	sx_xunlock(&old_pgrp->pg_killsx);
640 	return (0);
641 }
642 
643 /*
644  * Move p to an existing process group
645  */
646 int
647 enterthispgrp(struct proc *p, struct pgrp *pgrp)
648 {
649 	struct pgrp *old_pgrp;
650 
651 	sx_assert(&proctree_lock, SX_XLOCKED);
652 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
653 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
654 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
655 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
656 	KASSERT(pgrp->pg_session == p->p_session,
657 	    ("%s: pgrp's session %p, p->p_session %p proc %p\n",
658 	    __func__, pgrp->pg_session, p->p_session, p));
659 	KASSERT(pgrp != p->p_pgrp,
660 	    ("%s: p %p belongs to pgrp %p", __func__, p, pgrp));
661 
662 	old_pgrp = p->p_pgrp;
663 	if (!sx_try_xlock(&old_pgrp->pg_killsx)) {
664 		sx_xunlock(&proctree_lock);
665 		sx_xlock(&old_pgrp->pg_killsx);
666 		sx_xunlock(&old_pgrp->pg_killsx);
667 		return (ERESTART);
668 	}
669 	MPASS(old_pgrp == p->p_pgrp);
670 	if (!sx_try_xlock(&pgrp->pg_killsx)) {
671 		sx_xunlock(&old_pgrp->pg_killsx);
672 		sx_xunlock(&proctree_lock);
673 		sx_xlock(&pgrp->pg_killsx);
674 		sx_xunlock(&pgrp->pg_killsx);
675 		return (ERESTART);
676 	}
677 
678 	doenterpgrp(p, pgrp);
679 
680 	sx_xunlock(&pgrp->pg_killsx);
681 	sx_xunlock(&old_pgrp->pg_killsx);
682 	return (0);
683 }
684 
685 /*
686  * If true, any child of q which belongs to group pgrp, qualifies the
687  * process group pgrp as not orphaned.
688  */
689 static bool
690 isjobproc(struct proc *q, struct pgrp *pgrp)
691 {
692 	sx_assert(&proctree_lock, SX_LOCKED);
693 
694 	return (q->p_pgrp != pgrp &&
695 	    q->p_pgrp->pg_session == pgrp->pg_session);
696 }
697 
698 static struct proc *
699 jobc_reaper(struct proc *p)
700 {
701 	struct proc *pp;
702 
703 	sx_assert(&proctree_lock, SA_LOCKED);
704 
705 	for (pp = p;;) {
706 		pp = pp->p_reaper;
707 		if (pp->p_reaper == pp ||
708 		    (pp->p_treeflag & P_TREE_GRPEXITED) == 0)
709 			return (pp);
710 	}
711 }
712 
713 static struct proc *
714 jobc_parent(struct proc *p, struct proc *p_exiting)
715 {
716 	struct proc *pp;
717 
718 	sx_assert(&proctree_lock, SA_LOCKED);
719 
720 	pp = proc_realparent(p);
721 	if (pp->p_pptr == NULL || pp == p_exiting ||
722 	    (pp->p_treeflag & P_TREE_GRPEXITED) == 0)
723 		return (pp);
724 	return (jobc_reaper(pp));
725 }
726 
727 static int
728 pgrp_calc_jobc(struct pgrp *pgrp)
729 {
730 	struct proc *q;
731 	int cnt;
732 
733 #ifdef INVARIANTS
734 	if (!mtx_owned(&pgrp->pg_mtx))
735 		sx_assert(&proctree_lock, SA_LOCKED);
736 #endif
737 
738 	cnt = 0;
739 	LIST_FOREACH(q, &pgrp->pg_members, p_pglist) {
740 		if ((q->p_treeflag & P_TREE_GRPEXITED) != 0 ||
741 		    q->p_pptr == NULL)
742 			continue;
743 		if (isjobproc(jobc_parent(q, NULL), pgrp))
744 			cnt++;
745 	}
746 	return (cnt);
747 }
748 
749 /*
750  * Move p to a process group
751  */
752 static void
753 doenterpgrp(struct proc *p, struct pgrp *pgrp)
754 {
755 	struct pgrp *savepgrp;
756 	struct proc *pp;
757 
758 	sx_assert(&proctree_lock, SX_XLOCKED);
759 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
760 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
761 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
762 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
763 
764 	savepgrp = p->p_pgrp;
765 	pp = jobc_parent(p, NULL);
766 
767 	PGRP_LOCK(pgrp);
768 	PGRP_LOCK(savepgrp);
769 	if (isjobproc(pp, savepgrp) && pgrp_calc_jobc(savepgrp) == 1)
770 		orphanpg(savepgrp);
771 	PROC_LOCK(p);
772 	LIST_REMOVE(p, p_pglist);
773 	p->p_pgrp = pgrp;
774 	PROC_UNLOCK(p);
775 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
776 	if (isjobproc(pp, pgrp))
777 		pgrp->pg_flags &= ~PGRP_ORPHANED;
778 	PGRP_UNLOCK(savepgrp);
779 	PGRP_UNLOCK(pgrp);
780 	if (LIST_EMPTY(&savepgrp->pg_members))
781 		pgdelete(savepgrp);
782 }
783 
784 /*
785  * remove process from process group
786  */
787 int
788 leavepgrp(struct proc *p)
789 {
790 	struct pgrp *savepgrp;
791 
792 	sx_assert(&proctree_lock, SX_XLOCKED);
793 	savepgrp = p->p_pgrp;
794 	PGRP_LOCK(savepgrp);
795 	PROC_LOCK(p);
796 	LIST_REMOVE(p, p_pglist);
797 	p->p_pgrp = NULL;
798 	PROC_UNLOCK(p);
799 	PGRP_UNLOCK(savepgrp);
800 	if (LIST_EMPTY(&savepgrp->pg_members))
801 		pgdelete(savepgrp);
802 	return (0);
803 }
804 
805 /*
806  * delete a process group
807  */
808 static void
809 pgdelete(struct pgrp *pgrp)
810 {
811 	struct session *savesess;
812 	struct tty *tp;
813 
814 	sx_assert(&proctree_lock, SX_XLOCKED);
815 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
816 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
817 
818 	/*
819 	 * Reset any sigio structures pointing to us as a result of
820 	 * F_SETOWN with our pgid.  The proctree lock ensures that
821 	 * new sigio structures will not be added after this point.
822 	 */
823 	funsetownlst(&pgrp->pg_sigiolst);
824 
825 	PGRP_LOCK(pgrp);
826 	tp = pgrp->pg_session->s_ttyp;
827 	LIST_REMOVE(pgrp, pg_hash);
828 	savesess = pgrp->pg_session;
829 	PGRP_UNLOCK(pgrp);
830 
831 	/* Remove the reference to the pgrp before deallocating it. */
832 	if (tp != NULL) {
833 		tty_lock(tp);
834 		tty_rel_pgrp(tp, pgrp);
835 	}
836 
837 	proc_id_clear(PROC_ID_GROUP, pgrp->pg_id);
838 	uma_zfree(pgrp_zone, pgrp);
839 	sess_release(savesess);
840 }
841 
842 
843 static void
844 fixjobc_kill(struct proc *p)
845 {
846 	struct proc *q;
847 	struct pgrp *pgrp;
848 
849 	sx_assert(&proctree_lock, SX_LOCKED);
850 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
851 	pgrp = p->p_pgrp;
852 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
853 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
854 
855 	/*
856 	 * p no longer affects process group orphanage for children.
857 	 * It is marked by the flag because p is only physically
858 	 * removed from its process group on wait(2).
859 	 */
860 	MPASS((p->p_treeflag & P_TREE_GRPEXITED) == 0);
861 	p->p_treeflag |= P_TREE_GRPEXITED;
862 
863 	/*
864 	 * Check if exiting p orphans its own group.
865 	 */
866 	pgrp = p->p_pgrp;
867 	if (isjobproc(jobc_parent(p, NULL), pgrp)) {
868 		PGRP_LOCK(pgrp);
869 		if (pgrp_calc_jobc(pgrp) == 0)
870 			orphanpg(pgrp);
871 		PGRP_UNLOCK(pgrp);
872 	}
873 
874 	/*
875 	 * Check this process' children to see whether they qualify
876 	 * their process groups after reparenting to reaper.
877 	 */
878 	LIST_FOREACH(q, &p->p_children, p_sibling) {
879 		pgrp = q->p_pgrp;
880 		PGRP_LOCK(pgrp);
881 		if (pgrp_calc_jobc(pgrp) == 0) {
882 			/*
883 			 * We want to handle exactly the children that
884 			 * has p as realparent.  Then, when calculating
885 			 * jobc_parent for children, we should ignore
886 			 * P_TREE_GRPEXITED flag already set on p.
887 			 */
888 			if (jobc_parent(q, p) == p && isjobproc(p, pgrp))
889 				orphanpg(pgrp);
890 		} else
891 			pgrp->pg_flags &= ~PGRP_ORPHANED;
892 		PGRP_UNLOCK(pgrp);
893 	}
894 	LIST_FOREACH(q, &p->p_orphans, p_orphan) {
895 		pgrp = q->p_pgrp;
896 		PGRP_LOCK(pgrp);
897 		if (pgrp_calc_jobc(pgrp) == 0) {
898 			if (isjobproc(p, pgrp))
899 				orphanpg(pgrp);
900 		} else
901 			pgrp->pg_flags &= ~PGRP_ORPHANED;
902 		PGRP_UNLOCK(pgrp);
903 	}
904 }
905 
906 void
907 killjobc(void)
908 {
909 	struct session *sp;
910 	struct tty *tp;
911 	struct proc *p;
912 	struct vnode *ttyvp;
913 
914 	p = curproc;
915 	MPASS(p->p_flag & P_WEXIT);
916 	sx_assert(&proctree_lock, SX_LOCKED);
917 
918 	if (SESS_LEADER(p)) {
919 		sp = p->p_session;
920 
921 		/*
922 		 * s_ttyp is not zero'd; we use this to indicate that
923 		 * the session once had a controlling terminal. (for
924 		 * logging and informational purposes)
925 		 */
926 		SESS_LOCK(sp);
927 		ttyvp = sp->s_ttyvp;
928 		tp = sp->s_ttyp;
929 		sp->s_ttyvp = NULL;
930 		sp->s_ttydp = NULL;
931 		sp->s_leader = NULL;
932 		SESS_UNLOCK(sp);
933 
934 		/*
935 		 * Signal foreground pgrp and revoke access to
936 		 * controlling terminal if it has not been revoked
937 		 * already.
938 		 *
939 		 * Because the TTY may have been revoked in the mean
940 		 * time and could already have a new session associated
941 		 * with it, make sure we don't send a SIGHUP to a
942 		 * foreground process group that does not belong to this
943 		 * session.
944 		 */
945 
946 		if (tp != NULL) {
947 			tty_lock(tp);
948 			if (tp->t_session == sp)
949 				tty_signal_pgrp(tp, SIGHUP);
950 			tty_unlock(tp);
951 		}
952 
953 		if (ttyvp != NULL) {
954 			sx_xunlock(&proctree_lock);
955 			if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) {
956 				VOP_REVOKE(ttyvp, REVOKEALL);
957 				VOP_UNLOCK(ttyvp);
958 			}
959 			devfs_ctty_unref(ttyvp);
960 			sx_xlock(&proctree_lock);
961 		}
962 	}
963 	fixjobc_kill(p);
964 }
965 
966 /*
967  * A process group has become orphaned, mark it as such for signal
968  * delivery code.  If there are any stopped processes in the group,
969  * hang-up all process in that group.
970  */
971 static void
972 orphanpg(struct pgrp *pg)
973 {
974 	struct proc *p;
975 
976 	PGRP_LOCK_ASSERT(pg, MA_OWNED);
977 
978 	pg->pg_flags |= PGRP_ORPHANED;
979 
980 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
981 		PROC_LOCK(p);
982 		if (P_SHOULDSTOP(p) == P_STOPPED_SIG) {
983 			PROC_UNLOCK(p);
984 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
985 				PROC_LOCK(p);
986 				kern_psignal(p, SIGHUP);
987 				kern_psignal(p, SIGCONT);
988 				PROC_UNLOCK(p);
989 			}
990 			return;
991 		}
992 		PROC_UNLOCK(p);
993 	}
994 }
995 
996 void
997 sess_hold(struct session *s)
998 {
999 
1000 	refcount_acquire(&s->s_count);
1001 }
1002 
1003 void
1004 sess_release(struct session *s)
1005 {
1006 
1007 	if (refcount_release(&s->s_count)) {
1008 		if (s->s_ttyp != NULL) {
1009 			tty_lock(s->s_ttyp);
1010 			tty_rel_sess(s->s_ttyp, s);
1011 		}
1012 		proc_id_clear(PROC_ID_SESSION, s->s_sid);
1013 		mtx_destroy(&s->s_mtx);
1014 		free(s, M_SESSION);
1015 	}
1016 }
1017 
1018 #ifdef DDB
1019 
1020 static void
1021 db_print_pgrp_one(struct pgrp *pgrp, struct proc *p)
1022 {
1023 	db_printf(
1024 	    "    pid %d at %p pr %d pgrp %p e %d jc %d\n",
1025 	    p->p_pid, p, p->p_pptr == NULL ? -1 : p->p_pptr->p_pid,
1026 	    p->p_pgrp, (p->p_treeflag & P_TREE_GRPEXITED) != 0,
1027 	    p->p_pptr == NULL ? 0 : isjobproc(p->p_pptr, pgrp));
1028 }
1029 
1030 DB_SHOW_COMMAND_FLAGS(pgrpdump, pgrpdump, DB_CMD_MEMSAFE)
1031 {
1032 	struct pgrp *pgrp;
1033 	struct proc *p;
1034 	int i;
1035 
1036 	for (i = 0; i <= pgrphash; i++) {
1037 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
1038 			db_printf("indx %d\n", i);
1039 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
1040 				db_printf(
1041 			"  pgrp %p, pgid %d, sess %p, sesscnt %d, mem %p\n",
1042 				    pgrp, (int)pgrp->pg_id, pgrp->pg_session,
1043 				    pgrp->pg_session->s_count,
1044 				    LIST_FIRST(&pgrp->pg_members));
1045 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist)
1046 					db_print_pgrp_one(pgrp, p);
1047 			}
1048 		}
1049 	}
1050 }
1051 #endif /* DDB */
1052 
1053 /*
1054  * Calculate the kinfo_proc members which contain process-wide
1055  * informations.
1056  * Must be called with the target process locked.
1057  */
1058 static void
1059 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
1060 {
1061 	struct thread *td;
1062 
1063 	PROC_LOCK_ASSERT(p, MA_OWNED);
1064 
1065 	kp->ki_estcpu = 0;
1066 	kp->ki_pctcpu = 0;
1067 	FOREACH_THREAD_IN_PROC(p, td) {
1068 		thread_lock(td);
1069 		kp->ki_pctcpu += sched_pctcpu(td);
1070 		kp->ki_estcpu += sched_estcpu(td);
1071 		thread_unlock(td);
1072 	}
1073 }
1074 
1075 /*
1076  * Fill in any information that is common to all threads in the process.
1077  * Must be called with the target process locked.
1078  */
1079 static void
1080 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
1081 {
1082 	struct thread *td0;
1083 	struct ucred *cred;
1084 	struct sigacts *ps;
1085 	struct timeval boottime;
1086 
1087 	PROC_LOCK_ASSERT(p, MA_OWNED);
1088 
1089 	kp->ki_structsize = sizeof(*kp);
1090 	kp->ki_paddr = p;
1091 	kp->ki_addr =/* p->p_addr; */0; /* XXX */
1092 	kp->ki_args = p->p_args;
1093 	kp->ki_textvp = p->p_textvp;
1094 #ifdef KTRACE
1095 	kp->ki_tracep = ktr_get_tracevp(p, false);
1096 	kp->ki_traceflag = p->p_traceflag;
1097 #endif
1098 	kp->ki_fd = p->p_fd;
1099 	kp->ki_pd = p->p_pd;
1100 	kp->ki_vmspace = p->p_vmspace;
1101 	kp->ki_flag = p->p_flag;
1102 	kp->ki_flag2 = p->p_flag2;
1103 	cred = p->p_ucred;
1104 	if (cred) {
1105 		kp->ki_uid = cred->cr_uid;
1106 		kp->ki_ruid = cred->cr_ruid;
1107 		kp->ki_svuid = cred->cr_svuid;
1108 		kp->ki_cr_flags = 0;
1109 		if (cred->cr_flags & CRED_FLAG_CAPMODE)
1110 			kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
1111 		/* XXX bde doesn't like KI_NGROUPS */
1112 		if (cred->cr_ngroups > KI_NGROUPS) {
1113 			kp->ki_ngroups = KI_NGROUPS;
1114 			kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
1115 		} else
1116 			kp->ki_ngroups = cred->cr_ngroups;
1117 		bcopy(cred->cr_groups, kp->ki_groups,
1118 		    kp->ki_ngroups * sizeof(gid_t));
1119 		kp->ki_rgid = cred->cr_rgid;
1120 		kp->ki_svgid = cred->cr_svgid;
1121 		/* If jailed(cred), emulate the old P_JAILED flag. */
1122 		if (jailed(cred)) {
1123 			kp->ki_flag |= P_JAILED;
1124 			/* If inside the jail, use 0 as a jail ID. */
1125 			if (cred->cr_prison != curthread->td_ucred->cr_prison)
1126 				kp->ki_jid = cred->cr_prison->pr_id;
1127 		}
1128 		strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
1129 		    sizeof(kp->ki_loginclass));
1130 	}
1131 	ps = p->p_sigacts;
1132 	if (ps) {
1133 		mtx_lock(&ps->ps_mtx);
1134 		kp->ki_sigignore = ps->ps_sigignore;
1135 		kp->ki_sigcatch = ps->ps_sigcatch;
1136 		mtx_unlock(&ps->ps_mtx);
1137 	}
1138 	if (p->p_state != PRS_NEW &&
1139 	    p->p_state != PRS_ZOMBIE &&
1140 	    p->p_vmspace != NULL) {
1141 		struct vmspace *vm = p->p_vmspace;
1142 
1143 		kp->ki_size = vm->vm_map.size;
1144 		kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
1145 		FOREACH_THREAD_IN_PROC(p, td0)
1146 			kp->ki_rssize += td0->td_kstack_pages;
1147 		kp->ki_swrss = vm->vm_swrss;
1148 		kp->ki_tsize = vm->vm_tsize;
1149 		kp->ki_dsize = vm->vm_dsize;
1150 		kp->ki_ssize = vm->vm_ssize;
1151 	} else if (p->p_state == PRS_ZOMBIE)
1152 		kp->ki_stat = SZOMB;
1153 	kp->ki_sflag = PS_INMEM;
1154 	/* Calculate legacy swtime as seconds since 'swtick'. */
1155 	kp->ki_swtime = (ticks - p->p_swtick) / hz;
1156 	kp->ki_pid = p->p_pid;
1157 	kp->ki_nice = p->p_nice;
1158 	kp->ki_fibnum = p->p_fibnum;
1159 	kp->ki_start = p->p_stats->p_start;
1160 	getboottime(&boottime);
1161 	timevaladd(&kp->ki_start, &boottime);
1162 	PROC_STATLOCK(p);
1163 	rufetch(p, &kp->ki_rusage);
1164 	kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
1165 	calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
1166 	PROC_STATUNLOCK(p);
1167 	calccru(p, &kp->ki_childutime, &kp->ki_childstime);
1168 	/* Some callers want child times in a single value. */
1169 	kp->ki_childtime = kp->ki_childstime;
1170 	timevaladd(&kp->ki_childtime, &kp->ki_childutime);
1171 
1172 	FOREACH_THREAD_IN_PROC(p, td0)
1173 		kp->ki_cow += td0->td_cow;
1174 
1175 	if (p->p_comm[0] != '\0')
1176 		strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
1177 	if (p->p_sysent && p->p_sysent->sv_name != NULL &&
1178 	    p->p_sysent->sv_name[0] != '\0')
1179 		strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
1180 	kp->ki_siglist = p->p_siglist;
1181 	kp->ki_xstat = KW_EXITCODE(p->p_xexit, p->p_xsig);
1182 	kp->ki_acflag = p->p_acflag;
1183 	kp->ki_lock = p->p_lock;
1184 	if (p->p_pptr) {
1185 		kp->ki_ppid = p->p_oppid;
1186 		if (p->p_flag & P_TRACED)
1187 			kp->ki_tracer = p->p_pptr->p_pid;
1188 	}
1189 }
1190 
1191 /*
1192  * Fill job-related process information.
1193  */
1194 static void
1195 fill_kinfo_proc_pgrp(struct proc *p, struct kinfo_proc *kp)
1196 {
1197 	struct tty *tp;
1198 	struct session *sp;
1199 	struct pgrp *pgrp;
1200 
1201 	sx_assert(&proctree_lock, SA_LOCKED);
1202 	PROC_LOCK_ASSERT(p, MA_OWNED);
1203 
1204 	pgrp = p->p_pgrp;
1205 	if (pgrp == NULL)
1206 		return;
1207 
1208 	kp->ki_pgid = pgrp->pg_id;
1209 	kp->ki_jobc = pgrp_calc_jobc(pgrp);
1210 
1211 	sp = pgrp->pg_session;
1212 	tp = NULL;
1213 
1214 	if (sp != NULL) {
1215 		kp->ki_sid = sp->s_sid;
1216 		SESS_LOCK(sp);
1217 		strlcpy(kp->ki_login, sp->s_login, sizeof(kp->ki_login));
1218 		if (sp->s_ttyvp)
1219 			kp->ki_kiflag |= KI_CTTY;
1220 		if (SESS_LEADER(p))
1221 			kp->ki_kiflag |= KI_SLEADER;
1222 		tp = sp->s_ttyp;
1223 		SESS_UNLOCK(sp);
1224 	}
1225 
1226 	if ((p->p_flag & P_CONTROLT) && tp != NULL) {
1227 		kp->ki_tdev = tty_udev(tp);
1228 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1229 		kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
1230 		if (tp->t_session)
1231 			kp->ki_tsid = tp->t_session->s_sid;
1232 	} else {
1233 		kp->ki_tdev = NODEV;
1234 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1235 	}
1236 }
1237 
1238 /*
1239  * Fill in information that is thread specific.  Must be called with
1240  * target process locked.  If 'preferthread' is set, overwrite certain
1241  * process-related fields that are maintained for both threads and
1242  * processes.
1243  */
1244 static void
1245 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
1246 {
1247 	struct proc *p;
1248 
1249 	p = td->td_proc;
1250 	kp->ki_tdaddr = td;
1251 	PROC_LOCK_ASSERT(p, MA_OWNED);
1252 
1253 	if (preferthread)
1254 		PROC_STATLOCK(p);
1255 	thread_lock(td);
1256 	if (td->td_wmesg != NULL)
1257 		strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
1258 	else
1259 		bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
1260 	if (strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname)) >=
1261 	    sizeof(kp->ki_tdname)) {
1262 		strlcpy(kp->ki_moretdname,
1263 		    td->td_name + sizeof(kp->ki_tdname) - 1,
1264 		    sizeof(kp->ki_moretdname));
1265 	} else {
1266 		bzero(kp->ki_moretdname, sizeof(kp->ki_moretdname));
1267 	}
1268 	if (TD_ON_LOCK(td)) {
1269 		kp->ki_kiflag |= KI_LOCKBLOCK;
1270 		strlcpy(kp->ki_lockname, td->td_lockname,
1271 		    sizeof(kp->ki_lockname));
1272 	} else {
1273 		kp->ki_kiflag &= ~KI_LOCKBLOCK;
1274 		bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
1275 	}
1276 
1277 	if (p->p_state == PRS_NORMAL) { /* approximate. */
1278 		if (TD_ON_RUNQ(td) ||
1279 		    TD_CAN_RUN(td) ||
1280 		    TD_IS_RUNNING(td)) {
1281 			kp->ki_stat = SRUN;
1282 		} else if (P_SHOULDSTOP(p)) {
1283 			kp->ki_stat = SSTOP;
1284 		} else if (TD_IS_SLEEPING(td)) {
1285 			kp->ki_stat = SSLEEP;
1286 		} else if (TD_ON_LOCK(td)) {
1287 			kp->ki_stat = SLOCK;
1288 		} else {
1289 			kp->ki_stat = SWAIT;
1290 		}
1291 	} else if (p->p_state == PRS_ZOMBIE) {
1292 		kp->ki_stat = SZOMB;
1293 	} else {
1294 		kp->ki_stat = SIDL;
1295 	}
1296 
1297 	/* Things in the thread */
1298 	kp->ki_wchan = td->td_wchan;
1299 	kp->ki_pri.pri_level = td->td_priority;
1300 	kp->ki_pri.pri_native = td->td_base_pri;
1301 
1302 	/*
1303 	 * Note: legacy fields; clamp at the old NOCPU value and/or
1304 	 * the maximum u_char CPU value.
1305 	 */
1306 	if (td->td_lastcpu == NOCPU)
1307 		kp->ki_lastcpu_old = NOCPU_OLD;
1308 	else if (td->td_lastcpu > MAXCPU_OLD)
1309 		kp->ki_lastcpu_old = MAXCPU_OLD;
1310 	else
1311 		kp->ki_lastcpu_old = td->td_lastcpu;
1312 
1313 	if (td->td_oncpu == NOCPU)
1314 		kp->ki_oncpu_old = NOCPU_OLD;
1315 	else if (td->td_oncpu > MAXCPU_OLD)
1316 		kp->ki_oncpu_old = MAXCPU_OLD;
1317 	else
1318 		kp->ki_oncpu_old = td->td_oncpu;
1319 
1320 	kp->ki_lastcpu = td->td_lastcpu;
1321 	kp->ki_oncpu = td->td_oncpu;
1322 	kp->ki_tdflags = td->td_flags;
1323 	kp->ki_tid = td->td_tid;
1324 	kp->ki_numthreads = p->p_numthreads;
1325 	kp->ki_pcb = td->td_pcb;
1326 	kp->ki_kstack = (void *)td->td_kstack;
1327 	kp->ki_slptime = (ticks - td->td_slptick) / hz;
1328 	kp->ki_pri.pri_class = td->td_pri_class;
1329 	kp->ki_pri.pri_user = td->td_user_pri;
1330 
1331 	if (preferthread) {
1332 		rufetchtd(td, &kp->ki_rusage);
1333 		kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
1334 		kp->ki_pctcpu = sched_pctcpu(td);
1335 		kp->ki_estcpu = sched_estcpu(td);
1336 		kp->ki_cow = td->td_cow;
1337 	}
1338 
1339 	/* We can't get this anymore but ps etc never used it anyway. */
1340 	kp->ki_rqindex = 0;
1341 
1342 	if (preferthread)
1343 		kp->ki_siglist = td->td_siglist;
1344 	kp->ki_sigmask = td->td_sigmask;
1345 	thread_unlock(td);
1346 	if (preferthread)
1347 		PROC_STATUNLOCK(p);
1348 }
1349 
1350 /*
1351  * Fill in a kinfo_proc structure for the specified process.
1352  * Must be called with the target process locked.
1353  */
1354 void
1355 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1356 {
1357 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1358 
1359 	bzero(kp, sizeof(*kp));
1360 
1361 	fill_kinfo_proc_pgrp(p,kp);
1362 	fill_kinfo_proc_only(p, kp);
1363 	fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1364 	fill_kinfo_aggregate(p, kp);
1365 }
1366 
1367 struct pstats *
1368 pstats_alloc(void)
1369 {
1370 
1371 	return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1372 }
1373 
1374 /*
1375  * Copy parts of p_stats; zero the rest of p_stats (statistics).
1376  */
1377 void
1378 pstats_fork(struct pstats *src, struct pstats *dst)
1379 {
1380 
1381 	bzero(&dst->pstat_startzero,
1382 	    __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1383 	bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1384 	    __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1385 }
1386 
1387 void
1388 pstats_free(struct pstats *ps)
1389 {
1390 
1391 	free(ps, M_SUBPROC);
1392 }
1393 
1394 #ifdef COMPAT_FREEBSD32
1395 
1396 /*
1397  * This function is typically used to copy out the kernel address, so
1398  * it can be replaced by assignment of zero.
1399  */
1400 static inline uint32_t
1401 ptr32_trim(const void *ptr)
1402 {
1403 	uintptr_t uptr;
1404 
1405 	uptr = (uintptr_t)ptr;
1406 	return ((uptr > UINT_MAX) ? 0 : uptr);
1407 }
1408 
1409 #define PTRTRIM_CP(src,dst,fld) \
1410 	do { (dst).fld = ptr32_trim((src).fld); } while (0)
1411 
1412 static void
1413 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1414 {
1415 	int i;
1416 
1417 	bzero(ki32, sizeof(struct kinfo_proc32));
1418 	ki32->ki_structsize = sizeof(struct kinfo_proc32);
1419 	CP(*ki, *ki32, ki_layout);
1420 	PTRTRIM_CP(*ki, *ki32, ki_args);
1421 	PTRTRIM_CP(*ki, *ki32, ki_paddr);
1422 	PTRTRIM_CP(*ki, *ki32, ki_addr);
1423 	PTRTRIM_CP(*ki, *ki32, ki_tracep);
1424 	PTRTRIM_CP(*ki, *ki32, ki_textvp);
1425 	PTRTRIM_CP(*ki, *ki32, ki_fd);
1426 	PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1427 	PTRTRIM_CP(*ki, *ki32, ki_wchan);
1428 	CP(*ki, *ki32, ki_pid);
1429 	CP(*ki, *ki32, ki_ppid);
1430 	CP(*ki, *ki32, ki_pgid);
1431 	CP(*ki, *ki32, ki_tpgid);
1432 	CP(*ki, *ki32, ki_sid);
1433 	CP(*ki, *ki32, ki_tsid);
1434 	CP(*ki, *ki32, ki_jobc);
1435 	CP(*ki, *ki32, ki_tdev);
1436 	CP(*ki, *ki32, ki_tdev_freebsd11);
1437 	CP(*ki, *ki32, ki_siglist);
1438 	CP(*ki, *ki32, ki_sigmask);
1439 	CP(*ki, *ki32, ki_sigignore);
1440 	CP(*ki, *ki32, ki_sigcatch);
1441 	CP(*ki, *ki32, ki_uid);
1442 	CP(*ki, *ki32, ki_ruid);
1443 	CP(*ki, *ki32, ki_svuid);
1444 	CP(*ki, *ki32, ki_rgid);
1445 	CP(*ki, *ki32, ki_svgid);
1446 	CP(*ki, *ki32, ki_ngroups);
1447 	for (i = 0; i < KI_NGROUPS; i++)
1448 		CP(*ki, *ki32, ki_groups[i]);
1449 	CP(*ki, *ki32, ki_size);
1450 	CP(*ki, *ki32, ki_rssize);
1451 	CP(*ki, *ki32, ki_swrss);
1452 	CP(*ki, *ki32, ki_tsize);
1453 	CP(*ki, *ki32, ki_dsize);
1454 	CP(*ki, *ki32, ki_ssize);
1455 	CP(*ki, *ki32, ki_xstat);
1456 	CP(*ki, *ki32, ki_acflag);
1457 	CP(*ki, *ki32, ki_pctcpu);
1458 	CP(*ki, *ki32, ki_estcpu);
1459 	CP(*ki, *ki32, ki_slptime);
1460 	CP(*ki, *ki32, ki_swtime);
1461 	CP(*ki, *ki32, ki_cow);
1462 	CP(*ki, *ki32, ki_runtime);
1463 	TV_CP(*ki, *ki32, ki_start);
1464 	TV_CP(*ki, *ki32, ki_childtime);
1465 	CP(*ki, *ki32, ki_flag);
1466 	CP(*ki, *ki32, ki_kiflag);
1467 	CP(*ki, *ki32, ki_traceflag);
1468 	CP(*ki, *ki32, ki_stat);
1469 	CP(*ki, *ki32, ki_nice);
1470 	CP(*ki, *ki32, ki_lock);
1471 	CP(*ki, *ki32, ki_rqindex);
1472 	CP(*ki, *ki32, ki_oncpu);
1473 	CP(*ki, *ki32, ki_lastcpu);
1474 
1475 	/* XXX TODO: wrap cpu value as appropriate */
1476 	CP(*ki, *ki32, ki_oncpu_old);
1477 	CP(*ki, *ki32, ki_lastcpu_old);
1478 
1479 	bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1480 	bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1481 	bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1482 	bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1483 	bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1484 	bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1485 	bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1486 	bcopy(ki->ki_moretdname, ki32->ki_moretdname, MAXCOMLEN - TDNAMLEN + 1);
1487 	CP(*ki, *ki32, ki_tracer);
1488 	CP(*ki, *ki32, ki_flag2);
1489 	CP(*ki, *ki32, ki_fibnum);
1490 	CP(*ki, *ki32, ki_cr_flags);
1491 	CP(*ki, *ki32, ki_jid);
1492 	CP(*ki, *ki32, ki_numthreads);
1493 	CP(*ki, *ki32, ki_tid);
1494 	CP(*ki, *ki32, ki_pri);
1495 	freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1496 	freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1497 	PTRTRIM_CP(*ki, *ki32, ki_pcb);
1498 	PTRTRIM_CP(*ki, *ki32, ki_kstack);
1499 	PTRTRIM_CP(*ki, *ki32, ki_udata);
1500 	PTRTRIM_CP(*ki, *ki32, ki_tdaddr);
1501 	CP(*ki, *ki32, ki_sflag);
1502 	CP(*ki, *ki32, ki_tdflags);
1503 }
1504 #endif
1505 
1506 static ssize_t
1507 kern_proc_out_size(struct proc *p, int flags)
1508 {
1509 	ssize_t size = 0;
1510 
1511 	PROC_LOCK_ASSERT(p, MA_OWNED);
1512 
1513 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1514 #ifdef COMPAT_FREEBSD32
1515 		if ((flags & KERN_PROC_MASK32) != 0) {
1516 			size += sizeof(struct kinfo_proc32);
1517 		} else
1518 #endif
1519 			size += sizeof(struct kinfo_proc);
1520 	} else {
1521 #ifdef COMPAT_FREEBSD32
1522 		if ((flags & KERN_PROC_MASK32) != 0)
1523 			size += sizeof(struct kinfo_proc32) * p->p_numthreads;
1524 		else
1525 #endif
1526 			size += sizeof(struct kinfo_proc) * p->p_numthreads;
1527 	}
1528 	PROC_UNLOCK(p);
1529 	return (size);
1530 }
1531 
1532 int
1533 kern_proc_out(struct proc *p, struct sbuf *sb, int flags)
1534 {
1535 	struct thread *td;
1536 	struct kinfo_proc ki;
1537 #ifdef COMPAT_FREEBSD32
1538 	struct kinfo_proc32 ki32;
1539 #endif
1540 	int error;
1541 
1542 	PROC_LOCK_ASSERT(p, MA_OWNED);
1543 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1544 
1545 	error = 0;
1546 	fill_kinfo_proc(p, &ki);
1547 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1548 #ifdef COMPAT_FREEBSD32
1549 		if ((flags & KERN_PROC_MASK32) != 0) {
1550 			freebsd32_kinfo_proc_out(&ki, &ki32);
1551 			if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1552 				error = ENOMEM;
1553 		} else
1554 #endif
1555 			if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1556 				error = ENOMEM;
1557 	} else {
1558 		FOREACH_THREAD_IN_PROC(p, td) {
1559 			fill_kinfo_thread(td, &ki, 1);
1560 #ifdef COMPAT_FREEBSD32
1561 			if ((flags & KERN_PROC_MASK32) != 0) {
1562 				freebsd32_kinfo_proc_out(&ki, &ki32);
1563 				if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1564 					error = ENOMEM;
1565 			} else
1566 #endif
1567 				if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1568 					error = ENOMEM;
1569 			if (error != 0)
1570 				break;
1571 		}
1572 	}
1573 	PROC_UNLOCK(p);
1574 	return (error);
1575 }
1576 
1577 static int
1578 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
1579 {
1580 	struct sbuf sb;
1581 	struct kinfo_proc ki;
1582 	int error, error2;
1583 
1584 	if (req->oldptr == NULL)
1585 		return (SYSCTL_OUT(req, 0, kern_proc_out_size(p, flags)));
1586 
1587 	sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req);
1588 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1589 	error = kern_proc_out(p, &sb, flags);
1590 	error2 = sbuf_finish(&sb);
1591 	sbuf_delete(&sb);
1592 	if (error != 0)
1593 		return (error);
1594 	else if (error2 != 0)
1595 		return (error2);
1596 	return (0);
1597 }
1598 
1599 int
1600 proc_iterate(int (*cb)(struct proc *, void *), void *cbarg)
1601 {
1602 	struct proc *p;
1603 	int error, i, j;
1604 
1605 	for (i = 0; i < pidhashlock + 1; i++) {
1606 		sx_slock(&proctree_lock);
1607 		sx_slock(&pidhashtbl_lock[i]);
1608 		for (j = i; j <= pidhash; j += pidhashlock + 1) {
1609 			LIST_FOREACH(p, &pidhashtbl[j], p_hash) {
1610 				if (p->p_state == PRS_NEW)
1611 					continue;
1612 				error = cb(p, cbarg);
1613 				PROC_LOCK_ASSERT(p, MA_NOTOWNED);
1614 				if (error != 0) {
1615 					sx_sunlock(&pidhashtbl_lock[i]);
1616 					sx_sunlock(&proctree_lock);
1617 					return (error);
1618 				}
1619 			}
1620 		}
1621 		sx_sunlock(&pidhashtbl_lock[i]);
1622 		sx_sunlock(&proctree_lock);
1623 	}
1624 	return (0);
1625 }
1626 
1627 struct kern_proc_out_args {
1628 	struct sysctl_req *req;
1629 	int flags;
1630 	int oid_number;
1631 	int *name;
1632 };
1633 
1634 static int
1635 sysctl_kern_proc_iterate(struct proc *p, void *origarg)
1636 {
1637 	struct kern_proc_out_args *arg = origarg;
1638 	int *name = arg->name;
1639 	int oid_number = arg->oid_number;
1640 	int flags = arg->flags;
1641 	struct sysctl_req *req = arg->req;
1642 	int error = 0;
1643 
1644 	PROC_LOCK(p);
1645 
1646 	KASSERT(p->p_ucred != NULL,
1647 	    ("process credential is NULL for non-NEW proc"));
1648 	/*
1649 	 * Show a user only appropriate processes.
1650 	 */
1651 	if (p_cansee(curthread, p))
1652 		goto skip;
1653 	/*
1654 	 * TODO - make more efficient (see notes below).
1655 	 * do by session.
1656 	 */
1657 	switch (oid_number) {
1658 	case KERN_PROC_GID:
1659 		if (p->p_ucred->cr_gid != (gid_t)name[0])
1660 			goto skip;
1661 		break;
1662 
1663 	case KERN_PROC_PGRP:
1664 		/* could do this by traversing pgrp */
1665 		if (p->p_pgrp == NULL ||
1666 		    p->p_pgrp->pg_id != (pid_t)name[0])
1667 			goto skip;
1668 		break;
1669 
1670 	case KERN_PROC_RGID:
1671 		if (p->p_ucred->cr_rgid != (gid_t)name[0])
1672 			goto skip;
1673 		break;
1674 
1675 	case KERN_PROC_SESSION:
1676 		if (p->p_session == NULL ||
1677 		    p->p_session->s_sid != (pid_t)name[0])
1678 			goto skip;
1679 		break;
1680 
1681 	case KERN_PROC_TTY:
1682 		if ((p->p_flag & P_CONTROLT) == 0 ||
1683 		    p->p_session == NULL)
1684 			goto skip;
1685 		/* XXX proctree_lock */
1686 		SESS_LOCK(p->p_session);
1687 		if (p->p_session->s_ttyp == NULL ||
1688 		    tty_udev(p->p_session->s_ttyp) !=
1689 		    (dev_t)name[0]) {
1690 			SESS_UNLOCK(p->p_session);
1691 			goto skip;
1692 		}
1693 		SESS_UNLOCK(p->p_session);
1694 		break;
1695 
1696 	case KERN_PROC_UID:
1697 		if (p->p_ucred->cr_uid != (uid_t)name[0])
1698 			goto skip;
1699 		break;
1700 
1701 	case KERN_PROC_RUID:
1702 		if (p->p_ucred->cr_ruid != (uid_t)name[0])
1703 			goto skip;
1704 		break;
1705 
1706 	case KERN_PROC_PROC:
1707 		break;
1708 
1709 	default:
1710 		break;
1711 	}
1712 	error = sysctl_out_proc(p, req, flags);
1713 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
1714 	return (error);
1715 skip:
1716 	PROC_UNLOCK(p);
1717 	return (0);
1718 }
1719 
1720 static int
1721 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1722 {
1723 	struct kern_proc_out_args iterarg;
1724 	int *name = (int *)arg1;
1725 	u_int namelen = arg2;
1726 	struct proc *p;
1727 	int flags, oid_number;
1728 	int error = 0;
1729 
1730 	oid_number = oidp->oid_number;
1731 	if (oid_number != KERN_PROC_ALL &&
1732 	    (oid_number & KERN_PROC_INC_THREAD) == 0)
1733 		flags = KERN_PROC_NOTHREADS;
1734 	else {
1735 		flags = 0;
1736 		oid_number &= ~KERN_PROC_INC_THREAD;
1737 	}
1738 #ifdef COMPAT_FREEBSD32
1739 	if (req->flags & SCTL_MASK32)
1740 		flags |= KERN_PROC_MASK32;
1741 #endif
1742 	if (oid_number == KERN_PROC_PID) {
1743 		if (namelen != 1)
1744 			return (EINVAL);
1745 		error = sysctl_wire_old_buffer(req, 0);
1746 		if (error)
1747 			return (error);
1748 		sx_slock(&proctree_lock);
1749 		error = pget((pid_t)name[0], PGET_CANSEE, &p);
1750 		if (error == 0)
1751 			error = sysctl_out_proc(p, req, flags);
1752 		sx_sunlock(&proctree_lock);
1753 		return (error);
1754 	}
1755 
1756 	switch (oid_number) {
1757 	case KERN_PROC_ALL:
1758 		if (namelen != 0)
1759 			return (EINVAL);
1760 		break;
1761 	case KERN_PROC_PROC:
1762 		if (namelen != 0 && namelen != 1)
1763 			return (EINVAL);
1764 		break;
1765 	default:
1766 		if (namelen != 1)
1767 			return (EINVAL);
1768 		break;
1769 	}
1770 
1771 	if (req->oldptr == NULL) {
1772 		/* overestimate by 5 procs */
1773 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1774 		if (error)
1775 			return (error);
1776 	} else {
1777 		error = sysctl_wire_old_buffer(req, 0);
1778 		if (error != 0)
1779 			return (error);
1780 	}
1781 	iterarg.flags = flags;
1782 	iterarg.oid_number = oid_number;
1783 	iterarg.req = req;
1784 	iterarg.name = name;
1785 	error = proc_iterate(sysctl_kern_proc_iterate, &iterarg);
1786 	return (error);
1787 }
1788 
1789 struct pargs *
1790 pargs_alloc(int len)
1791 {
1792 	struct pargs *pa;
1793 
1794 	pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1795 		M_WAITOK);
1796 	refcount_init(&pa->ar_ref, 1);
1797 	pa->ar_length = len;
1798 	return (pa);
1799 }
1800 
1801 static void
1802 pargs_free(struct pargs *pa)
1803 {
1804 
1805 	free(pa, M_PARGS);
1806 }
1807 
1808 void
1809 pargs_hold(struct pargs *pa)
1810 {
1811 
1812 	if (pa == NULL)
1813 		return;
1814 	refcount_acquire(&pa->ar_ref);
1815 }
1816 
1817 void
1818 pargs_drop(struct pargs *pa)
1819 {
1820 
1821 	if (pa == NULL)
1822 		return;
1823 	if (refcount_release(&pa->ar_ref))
1824 		pargs_free(pa);
1825 }
1826 
1827 static int
1828 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1829     size_t len)
1830 {
1831 	ssize_t n;
1832 
1833 	/*
1834 	 * This may return a short read if the string is shorter than the chunk
1835 	 * and is aligned at the end of the page, and the following page is not
1836 	 * mapped.
1837 	 */
1838 	n = proc_readmem(td, p, (vm_offset_t)sptr, buf, len);
1839 	if (n <= 0)
1840 		return (ENOMEM);
1841 	return (0);
1842 }
1843 
1844 #define PROC_AUXV_MAX	256	/* Safety limit on auxv size. */
1845 
1846 enum proc_vector_type {
1847 	PROC_ARG,
1848 	PROC_ENV,
1849 	PROC_AUX,
1850 };
1851 
1852 #ifdef COMPAT_FREEBSD32
1853 static int
1854 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1855     size_t *vsizep, enum proc_vector_type type)
1856 {
1857 	struct freebsd32_ps_strings pss;
1858 	Elf32_Auxinfo aux;
1859 	vm_offset_t vptr, ptr;
1860 	uint32_t *proc_vector32;
1861 	char **proc_vector;
1862 	size_t vsize, size;
1863 	int i, error;
1864 
1865 	error = 0;
1866 	if (proc_readmem(td, p, PROC_PS_STRINGS(p), &pss, sizeof(pss)) !=
1867 	    sizeof(pss))
1868 		return (ENOMEM);
1869 	switch (type) {
1870 	case PROC_ARG:
1871 		vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1872 		vsize = pss.ps_nargvstr;
1873 		if (vsize > ARG_MAX)
1874 			return (ENOEXEC);
1875 		size = vsize * sizeof(int32_t);
1876 		break;
1877 	case PROC_ENV:
1878 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1879 		vsize = pss.ps_nenvstr;
1880 		if (vsize > ARG_MAX)
1881 			return (ENOEXEC);
1882 		size = vsize * sizeof(int32_t);
1883 		break;
1884 	case PROC_AUX:
1885 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1886 		    (pss.ps_nenvstr + 1) * sizeof(int32_t);
1887 		if (vptr % 4 != 0)
1888 			return (ENOEXEC);
1889 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1890 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1891 			    sizeof(aux))
1892 				return (ENOMEM);
1893 			if (aux.a_type == AT_NULL)
1894 				break;
1895 			ptr += sizeof(aux);
1896 		}
1897 		if (aux.a_type != AT_NULL)
1898 			return (ENOEXEC);
1899 		vsize = i + 1;
1900 		size = vsize * sizeof(aux);
1901 		break;
1902 	default:
1903 		KASSERT(0, ("Wrong proc vector type: %d", type));
1904 		return (EINVAL);
1905 	}
1906 	proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1907 	if (proc_readmem(td, p, vptr, proc_vector32, size) != size) {
1908 		error = ENOMEM;
1909 		goto done;
1910 	}
1911 	if (type == PROC_AUX) {
1912 		*proc_vectorp = (char **)proc_vector32;
1913 		*vsizep = vsize;
1914 		return (0);
1915 	}
1916 	proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1917 	for (i = 0; i < (int)vsize; i++)
1918 		proc_vector[i] = PTRIN(proc_vector32[i]);
1919 	*proc_vectorp = proc_vector;
1920 	*vsizep = vsize;
1921 done:
1922 	free(proc_vector32, M_TEMP);
1923 	return (error);
1924 }
1925 #endif
1926 
1927 static int
1928 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1929     size_t *vsizep, enum proc_vector_type type)
1930 {
1931 	struct ps_strings pss;
1932 	Elf_Auxinfo aux;
1933 	vm_offset_t vptr, ptr;
1934 	char **proc_vector;
1935 	size_t vsize, size;
1936 	int i;
1937 
1938 #ifdef COMPAT_FREEBSD32
1939 	if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1940 		return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1941 #endif
1942 	if (proc_readmem(td, p, PROC_PS_STRINGS(p), &pss, sizeof(pss)) !=
1943 	    sizeof(pss))
1944 		return (ENOMEM);
1945 	switch (type) {
1946 	case PROC_ARG:
1947 		vptr = (vm_offset_t)pss.ps_argvstr;
1948 		vsize = pss.ps_nargvstr;
1949 		if (vsize > ARG_MAX)
1950 			return (ENOEXEC);
1951 		size = vsize * sizeof(char *);
1952 		break;
1953 	case PROC_ENV:
1954 		vptr = (vm_offset_t)pss.ps_envstr;
1955 		vsize = pss.ps_nenvstr;
1956 		if (vsize > ARG_MAX)
1957 			return (ENOEXEC);
1958 		size = vsize * sizeof(char *);
1959 		break;
1960 	case PROC_AUX:
1961 		/*
1962 		 * The aux array is just above env array on the stack. Check
1963 		 * that the address is naturally aligned.
1964 		 */
1965 		vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1966 		    * sizeof(char *);
1967 #if __ELF_WORD_SIZE == 64
1968 		if (vptr % sizeof(uint64_t) != 0)
1969 #else
1970 		if (vptr % sizeof(uint32_t) != 0)
1971 #endif
1972 			return (ENOEXEC);
1973 		/*
1974 		 * We count the array size reading the aux vectors from the
1975 		 * stack until AT_NULL vector is returned.  So (to keep the code
1976 		 * simple) we read the process stack twice: the first time here
1977 		 * to find the size and the second time when copying the vectors
1978 		 * to the allocated proc_vector.
1979 		 */
1980 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1981 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1982 			    sizeof(aux))
1983 				return (ENOMEM);
1984 			if (aux.a_type == AT_NULL)
1985 				break;
1986 			ptr += sizeof(aux);
1987 		}
1988 		/*
1989 		 * If the PROC_AUXV_MAX entries are iterated over, and we have
1990 		 * not reached AT_NULL, it is most likely we are reading wrong
1991 		 * data: either the process doesn't have auxv array or data has
1992 		 * been modified. Return the error in this case.
1993 		 */
1994 		if (aux.a_type != AT_NULL)
1995 			return (ENOEXEC);
1996 		vsize = i + 1;
1997 		size = vsize * sizeof(aux);
1998 		break;
1999 	default:
2000 		KASSERT(0, ("Wrong proc vector type: %d", type));
2001 		return (EINVAL); /* In case we are built without INVARIANTS. */
2002 	}
2003 	proc_vector = malloc(size, M_TEMP, M_WAITOK);
2004 	if (proc_readmem(td, p, vptr, proc_vector, size) != size) {
2005 		free(proc_vector, M_TEMP);
2006 		return (ENOMEM);
2007 	}
2008 	*proc_vectorp = proc_vector;
2009 	*vsizep = vsize;
2010 
2011 	return (0);
2012 }
2013 
2014 #define GET_PS_STRINGS_CHUNK_SZ	256	/* Chunk size (bytes) for ps_strings operations. */
2015 
2016 static int
2017 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
2018     enum proc_vector_type type)
2019 {
2020 	size_t done, len, nchr, vsize;
2021 	int error, i;
2022 	char **proc_vector, *sptr;
2023 	char pss_string[GET_PS_STRINGS_CHUNK_SZ];
2024 
2025 	PROC_ASSERT_HELD(p);
2026 
2027 	/*
2028 	 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
2029 	 */
2030 	nchr = 2 * (PATH_MAX + ARG_MAX);
2031 
2032 	error = get_proc_vector(td, p, &proc_vector, &vsize, type);
2033 	if (error != 0)
2034 		return (error);
2035 	for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
2036 		/*
2037 		 * The program may have scribbled into its argv array, e.g. to
2038 		 * remove some arguments.  If that has happened, break out
2039 		 * before trying to read from NULL.
2040 		 */
2041 		if (proc_vector[i] == NULL)
2042 			break;
2043 		for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
2044 			error = proc_read_string(td, p, sptr, pss_string,
2045 			    sizeof(pss_string));
2046 			if (error != 0)
2047 				goto done;
2048 			len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
2049 			if (done + len >= nchr)
2050 				len = nchr - done - 1;
2051 			sbuf_bcat(sb, pss_string, len);
2052 			if (len != GET_PS_STRINGS_CHUNK_SZ)
2053 				break;
2054 			done += GET_PS_STRINGS_CHUNK_SZ;
2055 		}
2056 		sbuf_bcat(sb, "", 1);
2057 		done += len + 1;
2058 	}
2059 done:
2060 	free(proc_vector, M_TEMP);
2061 	return (error);
2062 }
2063 
2064 int
2065 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
2066 {
2067 
2068 	return (get_ps_strings(curthread, p, sb, PROC_ARG));
2069 }
2070 
2071 int
2072 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
2073 {
2074 
2075 	return (get_ps_strings(curthread, p, sb, PROC_ENV));
2076 }
2077 
2078 int
2079 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb)
2080 {
2081 	size_t vsize, size;
2082 	char **auxv;
2083 	int error;
2084 
2085 	error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX);
2086 	if (error == 0) {
2087 #ifdef COMPAT_FREEBSD32
2088 		if (SV_PROC_FLAG(p, SV_ILP32) != 0)
2089 			size = vsize * sizeof(Elf32_Auxinfo);
2090 		else
2091 #endif
2092 			size = vsize * sizeof(Elf_Auxinfo);
2093 		if (sbuf_bcat(sb, auxv, size) != 0)
2094 			error = ENOMEM;
2095 		free(auxv, M_TEMP);
2096 	}
2097 	return (error);
2098 }
2099 
2100 /*
2101  * This sysctl allows a process to retrieve the argument list or process
2102  * title for another process without groping around in the address space
2103  * of the other process.  It also allow a process to set its own "process
2104  * title to a string of its own choice.
2105  */
2106 static int
2107 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
2108 {
2109 	int *name = (int *)arg1;
2110 	u_int namelen = arg2;
2111 	struct pargs *newpa, *pa;
2112 	struct proc *p;
2113 	struct sbuf sb;
2114 	int flags, error = 0, error2;
2115 	pid_t pid;
2116 
2117 	if (namelen != 1)
2118 		return (EINVAL);
2119 
2120 	p = curproc;
2121 	pid = (pid_t)name[0];
2122 	if (pid == -1) {
2123 		pid = p->p_pid;
2124 	}
2125 
2126 	/*
2127 	 * If the query is for this process and it is single-threaded, there
2128 	 * is nobody to modify pargs, thus we can just read.
2129 	 */
2130 	if (pid == p->p_pid && p->p_numthreads == 1 && req->newptr == NULL &&
2131 	    (pa = p->p_args) != NULL)
2132 		return (SYSCTL_OUT(req, pa->ar_args, pa->ar_length));
2133 
2134 	flags = PGET_CANSEE;
2135 	if (req->newptr != NULL)
2136 		flags |= PGET_ISCURRENT;
2137 	error = pget(pid, flags, &p);
2138 	if (error)
2139 		return (error);
2140 
2141 	pa = p->p_args;
2142 	if (pa != NULL) {
2143 		pargs_hold(pa);
2144 		PROC_UNLOCK(p);
2145 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
2146 		pargs_drop(pa);
2147 	} else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
2148 		_PHOLD(p);
2149 		PROC_UNLOCK(p);
2150 		sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2151 		sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2152 		error = proc_getargv(curthread, p, &sb);
2153 		error2 = sbuf_finish(&sb);
2154 		PRELE(p);
2155 		sbuf_delete(&sb);
2156 		if (error == 0 && error2 != 0)
2157 			error = error2;
2158 	} else {
2159 		PROC_UNLOCK(p);
2160 	}
2161 	if (error != 0 || req->newptr == NULL)
2162 		return (error);
2163 
2164 	if (req->newlen > ps_arg_cache_limit - sizeof(struct pargs))
2165 		return (ENOMEM);
2166 
2167 	if (req->newlen == 0) {
2168 		/*
2169 		 * Clear the argument pointer, so that we'll fetch arguments
2170 		 * with proc_getargv() until further notice.
2171 		 */
2172 		newpa = NULL;
2173 	} else {
2174 		newpa = pargs_alloc(req->newlen);
2175 		error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
2176 		if (error != 0) {
2177 			pargs_free(newpa);
2178 			return (error);
2179 		}
2180 	}
2181 	PROC_LOCK(p);
2182 	pa = p->p_args;
2183 	p->p_args = newpa;
2184 	PROC_UNLOCK(p);
2185 	pargs_drop(pa);
2186 	return (0);
2187 }
2188 
2189 /*
2190  * This sysctl allows a process to retrieve environment of another process.
2191  */
2192 static int
2193 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
2194 {
2195 	int *name = (int *)arg1;
2196 	u_int namelen = arg2;
2197 	struct proc *p;
2198 	struct sbuf sb;
2199 	int error, error2;
2200 
2201 	if (namelen != 1)
2202 		return (EINVAL);
2203 
2204 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2205 	if (error != 0)
2206 		return (error);
2207 	if ((p->p_flag & P_SYSTEM) != 0) {
2208 		PRELE(p);
2209 		return (0);
2210 	}
2211 
2212 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2213 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2214 	error = proc_getenvv(curthread, p, &sb);
2215 	error2 = sbuf_finish(&sb);
2216 	PRELE(p);
2217 	sbuf_delete(&sb);
2218 	return (error != 0 ? error : error2);
2219 }
2220 
2221 /*
2222  * This sysctl allows a process to retrieve ELF auxiliary vector of
2223  * another process.
2224  */
2225 static int
2226 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
2227 {
2228 	int *name = (int *)arg1;
2229 	u_int namelen = arg2;
2230 	struct proc *p;
2231 	struct sbuf sb;
2232 	int error, error2;
2233 
2234 	if (namelen != 1)
2235 		return (EINVAL);
2236 
2237 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2238 	if (error != 0)
2239 		return (error);
2240 	if ((p->p_flag & P_SYSTEM) != 0) {
2241 		PRELE(p);
2242 		return (0);
2243 	}
2244 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2245 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2246 	error = proc_getauxv(curthread, p, &sb);
2247 	error2 = sbuf_finish(&sb);
2248 	PRELE(p);
2249 	sbuf_delete(&sb);
2250 	return (error != 0 ? error : error2);
2251 }
2252 
2253 /*
2254  * Look up the canonical executable path running in the specified process.
2255  * It tries to return the same hardlink name as was used for execve(2).
2256  * This allows the programs that modify their behavior based on their progname,
2257  * to operate correctly.
2258  *
2259  * Result is returned in retbuf, it must not be freed, similar to vn_fullpath()
2260  *   calling conventions.
2261  * binname is a pointer to temporary string buffer of length MAXPATHLEN,
2262  *   allocated and freed by caller.
2263  * freebuf should be freed by caller, from the M_TEMP malloc type.
2264  */
2265 int
2266 proc_get_binpath(struct proc *p, char *binname, char **retbuf,
2267     char **freebuf)
2268 {
2269 	struct nameidata nd;
2270 	struct vnode *vp, *dvp;
2271 	size_t freepath_size;
2272 	int error;
2273 	bool do_fullpath;
2274 
2275 	PROC_LOCK_ASSERT(p, MA_OWNED);
2276 
2277 	vp = p->p_textvp;
2278 	if (vp == NULL) {
2279 		PROC_UNLOCK(p);
2280 		*retbuf = "";
2281 		*freebuf = NULL;
2282 		return (0);
2283 	}
2284 	vref(vp);
2285 	dvp = p->p_textdvp;
2286 	if (dvp != NULL)
2287 		vref(dvp);
2288 	if (p->p_binname != NULL)
2289 		strlcpy(binname, p->p_binname, MAXPATHLEN);
2290 	PROC_UNLOCK(p);
2291 
2292 	do_fullpath = true;
2293 	*freebuf = NULL;
2294 	if (dvp != NULL && binname[0] != '\0') {
2295 		freepath_size = MAXPATHLEN;
2296 		if (vn_fullpath_hardlink(vp, dvp, binname, strlen(binname),
2297 		    retbuf, freebuf, &freepath_size) == 0) {
2298 			/*
2299 			 * Recheck the looked up path.  The binary
2300 			 * might have been renamed or replaced, in
2301 			 * which case we should not report old name.
2302 			 */
2303 			NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, *retbuf);
2304 			error = namei(&nd);
2305 			if (error == 0) {
2306 				if (nd.ni_vp == vp)
2307 					do_fullpath = false;
2308 				vrele(nd.ni_vp);
2309 				NDFREE_PNBUF(&nd);
2310 			}
2311 		}
2312 	}
2313 	if (do_fullpath) {
2314 		free(*freebuf, M_TEMP);
2315 		*freebuf = NULL;
2316 		error = vn_fullpath(vp, retbuf, freebuf);
2317 	}
2318 	vrele(vp);
2319 	if (dvp != NULL)
2320 		vrele(dvp);
2321 	return (error);
2322 }
2323 
2324 /*
2325  * This sysctl allows a process to retrieve the path of the executable for
2326  * itself or another process.
2327  */
2328 static int
2329 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
2330 {
2331 	pid_t *pidp = (pid_t *)arg1;
2332 	unsigned int arglen = arg2;
2333 	struct proc *p;
2334 	char *retbuf, *freebuf, *binname;
2335 	int error;
2336 
2337 	if (arglen != 1)
2338 		return (EINVAL);
2339 	binname = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
2340 	binname[0] = '\0';
2341 	if (*pidp == -1) {	/* -1 means this process */
2342 		error = 0;
2343 		p = req->td->td_proc;
2344 		PROC_LOCK(p);
2345 	} else {
2346 		error = pget(*pidp, PGET_CANSEE, &p);
2347 	}
2348 
2349 	if (error == 0)
2350 		error = proc_get_binpath(p, binname, &retbuf, &freebuf);
2351 	free(binname, M_TEMP);
2352 	if (error != 0)
2353 		return (error);
2354 	error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
2355 	free(freebuf, M_TEMP);
2356 	return (error);
2357 }
2358 
2359 static int
2360 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
2361 {
2362 	struct proc *p;
2363 	char *sv_name;
2364 	int *name;
2365 	int namelen;
2366 	int error;
2367 
2368 	namelen = arg2;
2369 	if (namelen != 1)
2370 		return (EINVAL);
2371 
2372 	name = (int *)arg1;
2373 	error = pget((pid_t)name[0], PGET_CANSEE, &p);
2374 	if (error != 0)
2375 		return (error);
2376 	sv_name = p->p_sysent->sv_name;
2377 	PROC_UNLOCK(p);
2378 	return (sysctl_handle_string(oidp, sv_name, 0, req));
2379 }
2380 
2381 #ifdef KINFO_OVMENTRY_SIZE
2382 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
2383 #endif
2384 
2385 #ifdef COMPAT_FREEBSD7
2386 static int
2387 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
2388 {
2389 	vm_map_entry_t entry, tmp_entry;
2390 	unsigned int last_timestamp, namelen;
2391 	char *fullpath, *freepath;
2392 	struct kinfo_ovmentry *kve;
2393 	struct vattr va;
2394 	struct ucred *cred;
2395 	int error, *name;
2396 	struct vnode *vp;
2397 	struct proc *p;
2398 	vm_map_t map;
2399 	struct vmspace *vm;
2400 
2401 	namelen = arg2;
2402 	if (namelen != 1)
2403 		return (EINVAL);
2404 
2405 	name = (int *)arg1;
2406 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2407 	if (error != 0)
2408 		return (error);
2409 	vm = vmspace_acquire_ref(p);
2410 	if (vm == NULL) {
2411 		PRELE(p);
2412 		return (ESRCH);
2413 	}
2414 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2415 
2416 	map = &vm->vm_map;
2417 	vm_map_lock_read(map);
2418 	VM_MAP_ENTRY_FOREACH(entry, map) {
2419 		vm_object_t obj, tobj, lobj;
2420 		vm_offset_t addr;
2421 
2422 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2423 			continue;
2424 
2425 		bzero(kve, sizeof(*kve));
2426 		kve->kve_structsize = sizeof(*kve);
2427 
2428 		kve->kve_private_resident = 0;
2429 		obj = entry->object.vm_object;
2430 		if (obj != NULL) {
2431 			VM_OBJECT_RLOCK(obj);
2432 			if (obj->shadow_count == 1)
2433 				kve->kve_private_resident =
2434 				    obj->resident_page_count;
2435 		}
2436 		kve->kve_resident = 0;
2437 		addr = entry->start;
2438 		while (addr < entry->end) {
2439 			if (pmap_extract(map->pmap, addr))
2440 				kve->kve_resident++;
2441 			addr += PAGE_SIZE;
2442 		}
2443 
2444 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2445 			if (tobj != obj) {
2446 				VM_OBJECT_RLOCK(tobj);
2447 				kve->kve_offset += tobj->backing_object_offset;
2448 			}
2449 			if (lobj != obj)
2450 				VM_OBJECT_RUNLOCK(lobj);
2451 			lobj = tobj;
2452 		}
2453 
2454 		kve->kve_start = (void*)entry->start;
2455 		kve->kve_end = (void*)entry->end;
2456 		kve->kve_offset += (off_t)entry->offset;
2457 
2458 		if (entry->protection & VM_PROT_READ)
2459 			kve->kve_protection |= KVME_PROT_READ;
2460 		if (entry->protection & VM_PROT_WRITE)
2461 			kve->kve_protection |= KVME_PROT_WRITE;
2462 		if (entry->protection & VM_PROT_EXECUTE)
2463 			kve->kve_protection |= KVME_PROT_EXEC;
2464 
2465 		if (entry->eflags & MAP_ENTRY_COW)
2466 			kve->kve_flags |= KVME_FLAG_COW;
2467 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2468 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2469 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2470 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2471 
2472 		last_timestamp = map->timestamp;
2473 		vm_map_unlock_read(map);
2474 
2475 		kve->kve_fileid = 0;
2476 		kve->kve_fsid = 0;
2477 		freepath = NULL;
2478 		fullpath = "";
2479 		if (lobj) {
2480 			kve->kve_type = vm_object_kvme_type(lobj, &vp);
2481 			if (kve->kve_type == KVME_TYPE_MGTDEVICE)
2482 				kve->kve_type = KVME_TYPE_UNKNOWN;
2483 			if (vp != NULL)
2484 				vref(vp);
2485 			if (lobj != obj)
2486 				VM_OBJECT_RUNLOCK(lobj);
2487 
2488 			kve->kve_ref_count = obj->ref_count;
2489 			kve->kve_shadow_count = obj->shadow_count;
2490 			VM_OBJECT_RUNLOCK(obj);
2491 			if (vp != NULL) {
2492 				vn_fullpath(vp, &fullpath, &freepath);
2493 				cred = curthread->td_ucred;
2494 				vn_lock(vp, LK_SHARED | LK_RETRY);
2495 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2496 					kve->kve_fileid = va.va_fileid;
2497 					/* truncate */
2498 					kve->kve_fsid = va.va_fsid;
2499 				}
2500 				vput(vp);
2501 			}
2502 		} else {
2503 			kve->kve_type = KVME_TYPE_NONE;
2504 			kve->kve_ref_count = 0;
2505 			kve->kve_shadow_count = 0;
2506 		}
2507 
2508 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2509 		if (freepath != NULL)
2510 			free(freepath, M_TEMP);
2511 
2512 		error = SYSCTL_OUT(req, kve, sizeof(*kve));
2513 		vm_map_lock_read(map);
2514 		if (error)
2515 			break;
2516 		if (last_timestamp != map->timestamp) {
2517 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2518 			entry = tmp_entry;
2519 		}
2520 	}
2521 	vm_map_unlock_read(map);
2522 	vmspace_free(vm);
2523 	PRELE(p);
2524 	free(kve, M_TEMP);
2525 	return (error);
2526 }
2527 #endif	/* COMPAT_FREEBSD7 */
2528 
2529 #ifdef KINFO_VMENTRY_SIZE
2530 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2531 #endif
2532 
2533 void
2534 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry,
2535     int *resident_count, bool *super)
2536 {
2537 	vm_object_t obj, tobj;
2538 	vm_page_t m, m_adv;
2539 	vm_offset_t addr;
2540 	vm_paddr_t pa;
2541 	vm_pindex_t pi, pi_adv, pindex;
2542 	int incore;
2543 
2544 	*super = false;
2545 	*resident_count = 0;
2546 	if (vmmap_skip_res_cnt)
2547 		return;
2548 
2549 	pa = 0;
2550 	obj = entry->object.vm_object;
2551 	addr = entry->start;
2552 	m_adv = NULL;
2553 	pi = OFF_TO_IDX(entry->offset);
2554 	for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) {
2555 		if (m_adv != NULL) {
2556 			m = m_adv;
2557 		} else {
2558 			pi_adv = atop(entry->end - addr);
2559 			pindex = pi;
2560 			for (tobj = obj;; tobj = tobj->backing_object) {
2561 				m = vm_page_find_least(tobj, pindex);
2562 				if (m != NULL) {
2563 					if (m->pindex == pindex)
2564 						break;
2565 					if (pi_adv > m->pindex - pindex) {
2566 						pi_adv = m->pindex - pindex;
2567 						m_adv = m;
2568 					}
2569 				}
2570 				if (tobj->backing_object == NULL)
2571 					goto next;
2572 				pindex += OFF_TO_IDX(tobj->
2573 				    backing_object_offset);
2574 			}
2575 		}
2576 		m_adv = NULL;
2577 		if (m->psind != 0 && addr + pagesizes[1] <= entry->end &&
2578 		    (addr & (pagesizes[1] - 1)) == 0 && (incore =
2579 		    pmap_mincore(map->pmap, addr, &pa) & MINCORE_SUPER) != 0) {
2580 			*super = true;
2581 			/*
2582 			 * The virtual page might be smaller than the physical
2583 			 * page, so we use the page size reported by the pmap
2584 			 * rather than m->psind.
2585 			 */
2586 			pi_adv = atop(pagesizes[incore >> MINCORE_PSIND_SHIFT]);
2587 		} else {
2588 			/*
2589 			 * We do not test the found page on validity.
2590 			 * Either the page is busy and being paged in,
2591 			 * or it was invalidated.  The first case
2592 			 * should be counted as resident, the second
2593 			 * is not so clear; we do account both.
2594 			 */
2595 			pi_adv = 1;
2596 		}
2597 		*resident_count += pi_adv;
2598 next:;
2599 	}
2600 }
2601 
2602 /*
2603  * Must be called with the process locked and will return unlocked.
2604  */
2605 int
2606 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb, ssize_t maxlen, int flags)
2607 {
2608 	vm_map_entry_t entry, tmp_entry;
2609 	struct vattr va;
2610 	vm_map_t map;
2611 	vm_object_t lobj, nobj, obj, tobj;
2612 	char *fullpath, *freepath;
2613 	struct kinfo_vmentry *kve;
2614 	struct ucred *cred;
2615 	struct vnode *vp;
2616 	struct vmspace *vm;
2617 	vm_offset_t addr;
2618 	unsigned int last_timestamp;
2619 	int error;
2620 	key_t key;
2621 	unsigned short seq;
2622 	bool guard, super;
2623 
2624 	PROC_LOCK_ASSERT(p, MA_OWNED);
2625 
2626 	_PHOLD(p);
2627 	PROC_UNLOCK(p);
2628 	vm = vmspace_acquire_ref(p);
2629 	if (vm == NULL) {
2630 		PRELE(p);
2631 		return (ESRCH);
2632 	}
2633 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK | M_ZERO);
2634 
2635 	error = 0;
2636 	map = &vm->vm_map;
2637 	vm_map_lock_read(map);
2638 	VM_MAP_ENTRY_FOREACH(entry, map) {
2639 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2640 			continue;
2641 
2642 		addr = entry->end;
2643 		bzero(kve, sizeof(*kve));
2644 		obj = entry->object.vm_object;
2645 		if (obj != NULL) {
2646 			if ((obj->flags & OBJ_ANON) != 0)
2647 				kve->kve_obj = (uintptr_t)obj;
2648 
2649 			for (tobj = obj; tobj != NULL;
2650 			    tobj = tobj->backing_object) {
2651 				VM_OBJECT_RLOCK(tobj);
2652 				kve->kve_offset += tobj->backing_object_offset;
2653 				lobj = tobj;
2654 			}
2655 			if (obj->backing_object == NULL)
2656 				kve->kve_private_resident =
2657 				    obj->resident_page_count;
2658 			kern_proc_vmmap_resident(map, entry,
2659 			    &kve->kve_resident, &super);
2660 			if (super)
2661 				kve->kve_flags |= KVME_FLAG_SUPER;
2662 			for (tobj = obj; tobj != NULL; tobj = nobj) {
2663 				nobj = tobj->backing_object;
2664 				if (tobj != obj && tobj != lobj)
2665 					VM_OBJECT_RUNLOCK(tobj);
2666 			}
2667 		} else {
2668 			lobj = NULL;
2669 		}
2670 
2671 		kve->kve_start = entry->start;
2672 		kve->kve_end = entry->end;
2673 		kve->kve_offset += entry->offset;
2674 
2675 		if (entry->protection & VM_PROT_READ)
2676 			kve->kve_protection |= KVME_PROT_READ;
2677 		if (entry->protection & VM_PROT_WRITE)
2678 			kve->kve_protection |= KVME_PROT_WRITE;
2679 		if (entry->protection & VM_PROT_EXECUTE)
2680 			kve->kve_protection |= KVME_PROT_EXEC;
2681 
2682 		if (entry->eflags & MAP_ENTRY_COW)
2683 			kve->kve_flags |= KVME_FLAG_COW;
2684 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2685 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2686 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2687 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2688 		if (entry->eflags & MAP_ENTRY_GROWS_UP)
2689 			kve->kve_flags |= KVME_FLAG_GROWS_UP;
2690 		if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2691 			kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2692 		if (entry->eflags & MAP_ENTRY_USER_WIRED)
2693 			kve->kve_flags |= KVME_FLAG_USER_WIRED;
2694 
2695 		guard = (entry->eflags & MAP_ENTRY_GUARD) != 0;
2696 
2697 		last_timestamp = map->timestamp;
2698 		vm_map_unlock_read(map);
2699 
2700 		freepath = NULL;
2701 		fullpath = "";
2702 		if (lobj != NULL) {
2703 			kve->kve_type = vm_object_kvme_type(lobj, &vp);
2704 			if (vp != NULL)
2705 				vref(vp);
2706 			if (lobj != obj)
2707 				VM_OBJECT_RUNLOCK(lobj);
2708 
2709 			kve->kve_ref_count = obj->ref_count;
2710 			kve->kve_shadow_count = obj->shadow_count;
2711 			VM_OBJECT_RUNLOCK(obj);
2712 			if ((lobj->flags & OBJ_SYSVSHM) != 0) {
2713 				kve->kve_flags |= KVME_FLAG_SYSVSHM;
2714 				shmobjinfo(lobj, &key, &seq);
2715 				kve->kve_vn_fileid = key;
2716 				kve->kve_vn_fsid_freebsd11 = seq;
2717 			}
2718 			if (vp != NULL) {
2719 				vn_fullpath(vp, &fullpath, &freepath);
2720 				kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2721 				cred = curthread->td_ucred;
2722 				vn_lock(vp, LK_SHARED | LK_RETRY);
2723 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2724 					kve->kve_vn_fileid = va.va_fileid;
2725 					kve->kve_vn_fsid = va.va_fsid;
2726 					kve->kve_vn_fsid_freebsd11 =
2727 					    kve->kve_vn_fsid; /* truncate */
2728 					kve->kve_vn_mode =
2729 					    MAKEIMODE(va.va_type, va.va_mode);
2730 					kve->kve_vn_size = va.va_size;
2731 					kve->kve_vn_rdev = va.va_rdev;
2732 					kve->kve_vn_rdev_freebsd11 =
2733 					    kve->kve_vn_rdev; /* truncate */
2734 					kve->kve_status = KF_ATTR_VALID;
2735 				}
2736 				vput(vp);
2737 			}
2738 		} else {
2739 			kve->kve_type = guard ? KVME_TYPE_GUARD :
2740 			    KVME_TYPE_NONE;
2741 			kve->kve_ref_count = 0;
2742 			kve->kve_shadow_count = 0;
2743 		}
2744 
2745 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2746 		if (freepath != NULL)
2747 			free(freepath, M_TEMP);
2748 
2749 		/* Pack record size down */
2750 		if ((flags & KERN_VMMAP_PACK_KINFO) != 0)
2751 			kve->kve_structsize =
2752 			    offsetof(struct kinfo_vmentry, kve_path) +
2753 			    strlen(kve->kve_path) + 1;
2754 		else
2755 			kve->kve_structsize = sizeof(*kve);
2756 		kve->kve_structsize = roundup(kve->kve_structsize,
2757 		    sizeof(uint64_t));
2758 
2759 		/* Halt filling and truncate rather than exceeding maxlen */
2760 		if (maxlen != -1 && maxlen < kve->kve_structsize) {
2761 			error = 0;
2762 			vm_map_lock_read(map);
2763 			break;
2764 		} else if (maxlen != -1)
2765 			maxlen -= kve->kve_structsize;
2766 
2767 		if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0)
2768 			error = ENOMEM;
2769 		vm_map_lock_read(map);
2770 		if (error != 0)
2771 			break;
2772 		if (last_timestamp != map->timestamp) {
2773 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2774 			entry = tmp_entry;
2775 		}
2776 	}
2777 	vm_map_unlock_read(map);
2778 	vmspace_free(vm);
2779 	PRELE(p);
2780 	free(kve, M_TEMP);
2781 	return (error);
2782 }
2783 
2784 static int
2785 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2786 {
2787 	struct proc *p;
2788 	struct sbuf sb;
2789 	u_int namelen;
2790 	int error, error2, *name;
2791 
2792 	namelen = arg2;
2793 	if (namelen != 1)
2794 		return (EINVAL);
2795 
2796 	name = (int *)arg1;
2797 	sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req);
2798 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2799 	error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
2800 	if (error != 0) {
2801 		sbuf_delete(&sb);
2802 		return (error);
2803 	}
2804 	error = kern_proc_vmmap_out(p, &sb, -1, KERN_VMMAP_PACK_KINFO);
2805 	error2 = sbuf_finish(&sb);
2806 	sbuf_delete(&sb);
2807 	return (error != 0 ? error : error2);
2808 }
2809 
2810 #if defined(STACK) || defined(DDB)
2811 static int
2812 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2813 {
2814 	struct kinfo_kstack *kkstp;
2815 	int error, i, *name, numthreads;
2816 	lwpid_t *lwpidarray;
2817 	struct thread *td;
2818 	struct stack *st;
2819 	struct sbuf sb;
2820 	struct proc *p;
2821 	u_int namelen;
2822 
2823 	namelen = arg2;
2824 	if (namelen != 1)
2825 		return (EINVAL);
2826 
2827 	name = (int *)arg1;
2828 	error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2829 	if (error != 0)
2830 		return (error);
2831 
2832 	kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2833 	st = stack_create(M_WAITOK);
2834 
2835 	lwpidarray = NULL;
2836 	PROC_LOCK(p);
2837 	do {
2838 		if (lwpidarray != NULL) {
2839 			free(lwpidarray, M_TEMP);
2840 			lwpidarray = NULL;
2841 		}
2842 		numthreads = p->p_numthreads;
2843 		PROC_UNLOCK(p);
2844 		lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2845 		    M_WAITOK | M_ZERO);
2846 		PROC_LOCK(p);
2847 	} while (numthreads < p->p_numthreads);
2848 
2849 	/*
2850 	 * XXXRW: During the below loop, execve(2) and countless other sorts
2851 	 * of changes could have taken place.  Should we check to see if the
2852 	 * vmspace has been replaced, or the like, in order to prevent
2853 	 * giving a snapshot that spans, say, execve(2), with some threads
2854 	 * before and some after?  Among other things, the credentials could
2855 	 * have changed, in which case the right to extract debug info might
2856 	 * no longer be assured.
2857 	 */
2858 	i = 0;
2859 	FOREACH_THREAD_IN_PROC(p, td) {
2860 		KASSERT(i < numthreads,
2861 		    ("sysctl_kern_proc_kstack: numthreads"));
2862 		lwpidarray[i] = td->td_tid;
2863 		i++;
2864 	}
2865 	PROC_UNLOCK(p);
2866 	numthreads = i;
2867 	for (i = 0; i < numthreads; i++) {
2868 		td = tdfind(lwpidarray[i], p->p_pid);
2869 		if (td == NULL) {
2870 			continue;
2871 		}
2872 		bzero(kkstp, sizeof(*kkstp));
2873 		(void)sbuf_new(&sb, kkstp->kkst_trace,
2874 		    sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2875 		thread_lock(td);
2876 		kkstp->kkst_tid = td->td_tid;
2877 		if (stack_save_td(st, td) == 0)
2878 			kkstp->kkst_state = KKST_STATE_STACKOK;
2879 		else
2880 			kkstp->kkst_state = KKST_STATE_RUNNING;
2881 		thread_unlock(td);
2882 		PROC_UNLOCK(p);
2883 		stack_sbuf_print(&sb, st);
2884 		sbuf_finish(&sb);
2885 		sbuf_delete(&sb);
2886 		error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2887 		if (error)
2888 			break;
2889 	}
2890 	PRELE(p);
2891 	if (lwpidarray != NULL)
2892 		free(lwpidarray, M_TEMP);
2893 	stack_destroy(st);
2894 	free(kkstp, M_TEMP);
2895 	return (error);
2896 }
2897 #endif
2898 
2899 /*
2900  * This sysctl allows a process to retrieve the full list of groups from
2901  * itself or another process.
2902  */
2903 static int
2904 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2905 {
2906 	pid_t *pidp = (pid_t *)arg1;
2907 	unsigned int arglen = arg2;
2908 	struct proc *p;
2909 	struct ucred *cred;
2910 	int error;
2911 
2912 	if (arglen != 1)
2913 		return (EINVAL);
2914 	if (*pidp == -1) {	/* -1 means this process */
2915 		p = req->td->td_proc;
2916 		PROC_LOCK(p);
2917 	} else {
2918 		error = pget(*pidp, PGET_CANSEE, &p);
2919 		if (error != 0)
2920 			return (error);
2921 	}
2922 
2923 	cred = crhold(p->p_ucred);
2924 	PROC_UNLOCK(p);
2925 
2926 	error = SYSCTL_OUT(req, cred->cr_groups,
2927 	    cred->cr_ngroups * sizeof(gid_t));
2928 	crfree(cred);
2929 	return (error);
2930 }
2931 
2932 /*
2933  * This sysctl allows a process to retrieve or/and set the resource limit for
2934  * another process.
2935  */
2936 static int
2937 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2938 {
2939 	int *name = (int *)arg1;
2940 	u_int namelen = arg2;
2941 	struct rlimit rlim;
2942 	struct proc *p;
2943 	u_int which;
2944 	int flags, error;
2945 
2946 	if (namelen != 2)
2947 		return (EINVAL);
2948 
2949 	which = (u_int)name[1];
2950 	if (which >= RLIM_NLIMITS)
2951 		return (EINVAL);
2952 
2953 	if (req->newptr != NULL && req->newlen != sizeof(rlim))
2954 		return (EINVAL);
2955 
2956 	flags = PGET_HOLD | PGET_NOTWEXIT;
2957 	if (req->newptr != NULL)
2958 		flags |= PGET_CANDEBUG;
2959 	else
2960 		flags |= PGET_CANSEE;
2961 	error = pget((pid_t)name[0], flags, &p);
2962 	if (error != 0)
2963 		return (error);
2964 
2965 	/*
2966 	 * Retrieve limit.
2967 	 */
2968 	if (req->oldptr != NULL) {
2969 		PROC_LOCK(p);
2970 		lim_rlimit_proc(p, which, &rlim);
2971 		PROC_UNLOCK(p);
2972 	}
2973 	error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
2974 	if (error != 0)
2975 		goto errout;
2976 
2977 	/*
2978 	 * Set limit.
2979 	 */
2980 	if (req->newptr != NULL) {
2981 		error = SYSCTL_IN(req, &rlim, sizeof(rlim));
2982 		if (error == 0)
2983 			error = kern_proc_setrlimit(curthread, p, which, &rlim);
2984 	}
2985 
2986 errout:
2987 	PRELE(p);
2988 	return (error);
2989 }
2990 
2991 /*
2992  * This sysctl allows a process to retrieve ps_strings structure location of
2993  * another process.
2994  */
2995 static int
2996 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
2997 {
2998 	int *name = (int *)arg1;
2999 	u_int namelen = arg2;
3000 	struct proc *p;
3001 	vm_offset_t ps_strings;
3002 	int error;
3003 #ifdef COMPAT_FREEBSD32
3004 	uint32_t ps_strings32;
3005 #endif
3006 
3007 	if (namelen != 1)
3008 		return (EINVAL);
3009 
3010 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
3011 	if (error != 0)
3012 		return (error);
3013 #ifdef COMPAT_FREEBSD32
3014 	if ((req->flags & SCTL_MASK32) != 0) {
3015 		/*
3016 		 * We return 0 if the 32 bit emulation request is for a 64 bit
3017 		 * process.
3018 		 */
3019 		ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
3020 		    PTROUT(PROC_PS_STRINGS(p)) : 0;
3021 		PROC_UNLOCK(p);
3022 		error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
3023 		return (error);
3024 	}
3025 #endif
3026 	ps_strings = PROC_PS_STRINGS(p);
3027 	PROC_UNLOCK(p);
3028 	error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
3029 	return (error);
3030 }
3031 
3032 /*
3033  * This sysctl allows a process to retrieve umask of another process.
3034  */
3035 static int
3036 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
3037 {
3038 	int *name = (int *)arg1;
3039 	u_int namelen = arg2;
3040 	struct proc *p;
3041 	int error;
3042 	u_short cmask;
3043 	pid_t pid;
3044 
3045 	if (namelen != 1)
3046 		return (EINVAL);
3047 
3048 	pid = (pid_t)name[0];
3049 	p = curproc;
3050 	if (pid == p->p_pid || pid == 0) {
3051 		cmask = p->p_pd->pd_cmask;
3052 		goto out;
3053 	}
3054 
3055 	error = pget(pid, PGET_WANTREAD, &p);
3056 	if (error != 0)
3057 		return (error);
3058 
3059 	cmask = p->p_pd->pd_cmask;
3060 	PRELE(p);
3061 out:
3062 	error = SYSCTL_OUT(req, &cmask, sizeof(cmask));
3063 	return (error);
3064 }
3065 
3066 /*
3067  * This sysctl allows a process to set and retrieve binary osreldate of
3068  * another process.
3069  */
3070 static int
3071 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
3072 {
3073 	int *name = (int *)arg1;
3074 	u_int namelen = arg2;
3075 	struct proc *p;
3076 	int flags, error, osrel;
3077 
3078 	if (namelen != 1)
3079 		return (EINVAL);
3080 
3081 	if (req->newptr != NULL && req->newlen != sizeof(osrel))
3082 		return (EINVAL);
3083 
3084 	flags = PGET_HOLD | PGET_NOTWEXIT;
3085 	if (req->newptr != NULL)
3086 		flags |= PGET_CANDEBUG;
3087 	else
3088 		flags |= PGET_CANSEE;
3089 	error = pget((pid_t)name[0], flags, &p);
3090 	if (error != 0)
3091 		return (error);
3092 
3093 	error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
3094 	if (error != 0)
3095 		goto errout;
3096 
3097 	if (req->newptr != NULL) {
3098 		error = SYSCTL_IN(req, &osrel, sizeof(osrel));
3099 		if (error != 0)
3100 			goto errout;
3101 		if (osrel < 0) {
3102 			error = EINVAL;
3103 			goto errout;
3104 		}
3105 		p->p_osrel = osrel;
3106 	}
3107 errout:
3108 	PRELE(p);
3109 	return (error);
3110 }
3111 
3112 static int
3113 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS)
3114 {
3115 	int *name = (int *)arg1;
3116 	u_int namelen = arg2;
3117 	struct proc *p;
3118 	struct kinfo_sigtramp kst;
3119 	const struct sysentvec *sv;
3120 	int error;
3121 #ifdef COMPAT_FREEBSD32
3122 	struct kinfo_sigtramp32 kst32;
3123 #endif
3124 
3125 	if (namelen != 1)
3126 		return (EINVAL);
3127 
3128 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
3129 	if (error != 0)
3130 		return (error);
3131 	sv = p->p_sysent;
3132 #ifdef COMPAT_FREEBSD32
3133 	if ((req->flags & SCTL_MASK32) != 0) {
3134 		bzero(&kst32, sizeof(kst32));
3135 		if (SV_PROC_FLAG(p, SV_ILP32)) {
3136 			if (PROC_HAS_SHP(p)) {
3137 				kst32.ksigtramp_start = PROC_SIGCODE(p);
3138 				kst32.ksigtramp_end = kst32.ksigtramp_start +
3139 				    ((sv->sv_flags & SV_DSO_SIG) == 0 ?
3140 				    *sv->sv_szsigcode :
3141 				    (uintptr_t)sv->sv_szsigcode);
3142 			} else {
3143 				kst32.ksigtramp_start = PROC_PS_STRINGS(p) -
3144 				    *sv->sv_szsigcode;
3145 				kst32.ksigtramp_end = PROC_PS_STRINGS(p);
3146 			}
3147 		}
3148 		PROC_UNLOCK(p);
3149 		error = SYSCTL_OUT(req, &kst32, sizeof(kst32));
3150 		return (error);
3151 	}
3152 #endif
3153 	bzero(&kst, sizeof(kst));
3154 	if (PROC_HAS_SHP(p)) {
3155 		kst.ksigtramp_start = (char *)PROC_SIGCODE(p);
3156 		kst.ksigtramp_end = (char *)kst.ksigtramp_start +
3157 		    ((sv->sv_flags & SV_DSO_SIG) == 0 ? *sv->sv_szsigcode :
3158 		    (uintptr_t)sv->sv_szsigcode);
3159 	} else {
3160 		kst.ksigtramp_start = (char *)PROC_PS_STRINGS(p) -
3161 		    *sv->sv_szsigcode;
3162 		kst.ksigtramp_end = (char *)PROC_PS_STRINGS(p);
3163 	}
3164 	PROC_UNLOCK(p);
3165 	error = SYSCTL_OUT(req, &kst, sizeof(kst));
3166 	return (error);
3167 }
3168 
3169 static int
3170 sysctl_kern_proc_sigfastblk(SYSCTL_HANDLER_ARGS)
3171 {
3172 	int *name = (int *)arg1;
3173 	u_int namelen = arg2;
3174 	pid_t pid;
3175 	struct proc *p;
3176 	struct thread *td1;
3177 	uintptr_t addr;
3178 #ifdef COMPAT_FREEBSD32
3179 	uint32_t addr32;
3180 #endif
3181 	int error;
3182 
3183 	if (namelen != 1 || req->newptr != NULL)
3184 		return (EINVAL);
3185 
3186 	pid = (pid_t)name[0];
3187 	error = pget(pid, PGET_HOLD | PGET_NOTWEXIT | PGET_CANDEBUG, &p);
3188 	if (error != 0)
3189 		return (error);
3190 
3191 	PROC_LOCK(p);
3192 #ifdef COMPAT_FREEBSD32
3193 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3194 		if (!SV_PROC_FLAG(p, SV_ILP32)) {
3195 			error = EINVAL;
3196 			goto errlocked;
3197 		}
3198 	}
3199 #endif
3200 	if (pid <= PID_MAX) {
3201 		td1 = FIRST_THREAD_IN_PROC(p);
3202 	} else {
3203 		FOREACH_THREAD_IN_PROC(p, td1) {
3204 			if (td1->td_tid == pid)
3205 				break;
3206 		}
3207 	}
3208 	if (td1 == NULL) {
3209 		error = ESRCH;
3210 		goto errlocked;
3211 	}
3212 	/*
3213 	 * The access to the private thread flags.  It is fine as far
3214 	 * as no out-of-thin-air values are read from td_pflags, and
3215 	 * usermode read of the td_sigblock_ptr is racy inherently,
3216 	 * since target process might have already changed it
3217 	 * meantime.
3218 	 */
3219 	if ((td1->td_pflags & TDP_SIGFASTBLOCK) != 0)
3220 		addr = (uintptr_t)td1->td_sigblock_ptr;
3221 	else
3222 		error = ENOTTY;
3223 
3224 errlocked:
3225 	_PRELE(p);
3226 	PROC_UNLOCK(p);
3227 	if (error != 0)
3228 		return (error);
3229 
3230 #ifdef COMPAT_FREEBSD32
3231 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3232 		addr32 = addr;
3233 		error = SYSCTL_OUT(req, &addr32, sizeof(addr32));
3234 	} else
3235 #endif
3236 		error = SYSCTL_OUT(req, &addr, sizeof(addr));
3237 	return (error);
3238 }
3239 
3240 static int
3241 sysctl_kern_proc_vm_layout(SYSCTL_HANDLER_ARGS)
3242 {
3243 	struct kinfo_vm_layout kvm;
3244 	struct proc *p;
3245 	struct vmspace *vmspace;
3246 	int error, *name;
3247 
3248 	name = (int *)arg1;
3249 	if ((u_int)arg2 != 1)
3250 		return (EINVAL);
3251 
3252 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
3253 	if (error != 0)
3254 		return (error);
3255 #ifdef COMPAT_FREEBSD32
3256 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3257 		if (!SV_PROC_FLAG(p, SV_ILP32)) {
3258 			PROC_UNLOCK(p);
3259 			return (EINVAL);
3260 		}
3261 	}
3262 #endif
3263 	vmspace = vmspace_acquire_ref(p);
3264 	PROC_UNLOCK(p);
3265 
3266 	memset(&kvm, 0, sizeof(kvm));
3267 	kvm.kvm_min_user_addr = vm_map_min(&vmspace->vm_map);
3268 	kvm.kvm_max_user_addr = vm_map_max(&vmspace->vm_map);
3269 	kvm.kvm_text_addr = (uintptr_t)vmspace->vm_taddr;
3270 	kvm.kvm_text_size = vmspace->vm_tsize;
3271 	kvm.kvm_data_addr = (uintptr_t)vmspace->vm_daddr;
3272 	kvm.kvm_data_size = vmspace->vm_dsize;
3273 	kvm.kvm_stack_addr = (uintptr_t)vmspace->vm_maxsaddr;
3274 	kvm.kvm_stack_size = vmspace->vm_ssize;
3275 	kvm.kvm_shp_addr = vmspace->vm_shp_base;
3276 	kvm.kvm_shp_size = p->p_sysent->sv_shared_page_len;
3277 	if ((vmspace->vm_map.flags & MAP_WIREFUTURE) != 0)
3278 		kvm.kvm_map_flags |= KMAP_FLAG_WIREFUTURE;
3279 	if ((vmspace->vm_map.flags & MAP_ASLR) != 0)
3280 		kvm.kvm_map_flags |= KMAP_FLAG_ASLR;
3281 	if ((vmspace->vm_map.flags & MAP_ASLR_IGNSTART) != 0)
3282 		kvm.kvm_map_flags |= KMAP_FLAG_ASLR_IGNSTART;
3283 	if ((vmspace->vm_map.flags & MAP_WXORX) != 0)
3284 		kvm.kvm_map_flags |= KMAP_FLAG_WXORX;
3285 	if ((vmspace->vm_map.flags & MAP_ASLR_STACK) != 0)
3286 		kvm.kvm_map_flags |= KMAP_FLAG_ASLR_STACK;
3287 	if (vmspace->vm_shp_base != p->p_sysent->sv_shared_page_base &&
3288 	    PROC_HAS_SHP(p))
3289 		kvm.kvm_map_flags |= KMAP_FLAG_ASLR_SHARED_PAGE;
3290 
3291 #ifdef COMPAT_FREEBSD32
3292 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3293 		struct kinfo_vm_layout32 kvm32;
3294 
3295 		memset(&kvm32, 0, sizeof(kvm32));
3296 		kvm32.kvm_min_user_addr = (uint32_t)kvm.kvm_min_user_addr;
3297 		kvm32.kvm_max_user_addr = (uint32_t)kvm.kvm_max_user_addr;
3298 		kvm32.kvm_text_addr = (uint32_t)kvm.kvm_text_addr;
3299 		kvm32.kvm_text_size = (uint32_t)kvm.kvm_text_size;
3300 		kvm32.kvm_data_addr = (uint32_t)kvm.kvm_data_addr;
3301 		kvm32.kvm_data_size = (uint32_t)kvm.kvm_data_size;
3302 		kvm32.kvm_stack_addr = (uint32_t)kvm.kvm_stack_addr;
3303 		kvm32.kvm_stack_size = (uint32_t)kvm.kvm_stack_size;
3304 		kvm32.kvm_shp_addr = (uint32_t)kvm.kvm_shp_addr;
3305 		kvm32.kvm_shp_size = (uint32_t)kvm.kvm_shp_size;
3306 		kvm32.kvm_map_flags = kvm.kvm_map_flags;
3307 		error = SYSCTL_OUT(req, &kvm32, sizeof(kvm32));
3308 		goto out;
3309 	}
3310 #endif
3311 
3312 	error = SYSCTL_OUT(req, &kvm, sizeof(kvm));
3313 #ifdef COMPAT_FREEBSD32
3314 out:
3315 #endif
3316 	vmspace_free(vmspace);
3317 	return (error);
3318 }
3319 
3320 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,  0,
3321     "Process table");
3322 
3323 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
3324 	CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
3325 	"Return entire process table");
3326 
3327 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3328 	sysctl_kern_proc, "Process table");
3329 
3330 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
3331 	sysctl_kern_proc, "Process table");
3332 
3333 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3334 	sysctl_kern_proc, "Process table");
3335 
3336 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
3337 	CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3338 
3339 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
3340 	sysctl_kern_proc, "Process table");
3341 
3342 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3343 	sysctl_kern_proc, "Process table");
3344 
3345 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3346 	sysctl_kern_proc, "Process table");
3347 
3348 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3349 	sysctl_kern_proc, "Process table");
3350 
3351 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
3352 	sysctl_kern_proc, "Return process table, no threads");
3353 
3354 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
3355 	CTLFLAG_RW | CTLFLAG_CAPWR | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
3356 	sysctl_kern_proc_args, "Process argument list");
3357 
3358 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
3359 	sysctl_kern_proc_env, "Process environment");
3360 
3361 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
3362 	CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
3363 
3364 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
3365 	CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
3366 
3367 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
3368 	CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
3369 	"Process syscall vector name (ABI type)");
3370 
3371 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
3372 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3373 
3374 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
3375 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3376 
3377 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
3378 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3379 
3380 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
3381 	sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3382 
3383 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
3384 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3385 
3386 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
3387 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3388 
3389 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
3390 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3391 
3392 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
3393 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3394 
3395 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
3396 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
3397 	"Return process table, including threads");
3398 
3399 #ifdef COMPAT_FREEBSD7
3400 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
3401 	CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
3402 #endif
3403 
3404 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
3405 	CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
3406 
3407 #if defined(STACK) || defined(DDB)
3408 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
3409 	CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
3410 #endif
3411 
3412 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
3413 	CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
3414 
3415 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
3416 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
3417 	"Process resource limits");
3418 
3419 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
3420 	CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
3421 	"Process ps_strings location");
3422 
3423 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
3424 	CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
3425 
3426 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
3427 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
3428 	"Process binary osreldate");
3429 
3430 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD |
3431 	CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp,
3432 	"Process signal trampoline location");
3433 
3434 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGFASTBLK, sigfastblk, CTLFLAG_RD |
3435 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_sigfastblk,
3436 	"Thread sigfastblock address");
3437 
3438 static SYSCTL_NODE(_kern_proc, KERN_PROC_VM_LAYOUT, vm_layout, CTLFLAG_RD |
3439 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_vm_layout,
3440 	"Process virtual address space layout info");
3441 
3442 static struct sx stop_all_proc_blocker;
3443 SX_SYSINIT(stop_all_proc_blocker, &stop_all_proc_blocker, "sapblk");
3444 
3445 bool
3446 stop_all_proc_block(void)
3447 {
3448 	return (sx_xlock_sig(&stop_all_proc_blocker) == 0);
3449 }
3450 
3451 void
3452 stop_all_proc_unblock(void)
3453 {
3454 	sx_xunlock(&stop_all_proc_blocker);
3455 }
3456 
3457 int allproc_gen;
3458 
3459 /*
3460  * stop_all_proc() purpose is to stop all process which have usermode,
3461  * except current process for obvious reasons.  This makes it somewhat
3462  * unreliable when invoked from multithreaded process.  The service
3463  * must not be user-callable anyway.
3464  */
3465 void
3466 stop_all_proc(void)
3467 {
3468 	struct proc *cp, *p;
3469 	int r, gen;
3470 	bool restart, seen_stopped, seen_exiting, stopped_some;
3471 
3472 	if (!stop_all_proc_block())
3473 		return;
3474 
3475 	cp = curproc;
3476 allproc_loop:
3477 	sx_xlock(&allproc_lock);
3478 	gen = allproc_gen;
3479 	seen_exiting = seen_stopped = stopped_some = restart = false;
3480 	LIST_REMOVE(cp, p_list);
3481 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3482 	for (;;) {
3483 		p = LIST_NEXT(cp, p_list);
3484 		if (p == NULL)
3485 			break;
3486 		LIST_REMOVE(cp, p_list);
3487 		LIST_INSERT_AFTER(p, cp, p_list);
3488 		PROC_LOCK(p);
3489 		if ((p->p_flag & (P_KPROC | P_SYSTEM | P_TOTAL_STOP |
3490 		    P_STOPPED_SIG)) != 0) {
3491 			PROC_UNLOCK(p);
3492 			continue;
3493 		}
3494 		if ((p->p_flag2 & P2_WEXIT) != 0) {
3495 			seen_exiting = true;
3496 			PROC_UNLOCK(p);
3497 			continue;
3498 		}
3499 		if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) {
3500 			/*
3501 			 * Stopped processes are tolerated when there
3502 			 * are no other processes which might continue
3503 			 * them.  P_STOPPED_SINGLE but not
3504 			 * P_TOTAL_STOP process still has at least one
3505 			 * thread running.
3506 			 */
3507 			seen_stopped = true;
3508 			PROC_UNLOCK(p);
3509 			continue;
3510 		}
3511 		if ((p->p_flag & P_TRACED) != 0) {
3512 			/*
3513 			 * thread_single() below cannot stop traced p,
3514 			 * so skip it.  OTOH, we cannot require
3515 			 * restart because debugger might be either
3516 			 * already stopped or traced as well.
3517 			 */
3518 			PROC_UNLOCK(p);
3519 			continue;
3520 		}
3521 		sx_xunlock(&allproc_lock);
3522 		_PHOLD(p);
3523 		r = thread_single(p, SINGLE_ALLPROC);
3524 		if (r != 0)
3525 			restart = true;
3526 		else
3527 			stopped_some = true;
3528 		_PRELE(p);
3529 		PROC_UNLOCK(p);
3530 		sx_xlock(&allproc_lock);
3531 	}
3532 	/* Catch forked children we did not see in iteration. */
3533 	if (gen != allproc_gen)
3534 		restart = true;
3535 	sx_xunlock(&allproc_lock);
3536 	if (restart || stopped_some || seen_exiting || seen_stopped) {
3537 		kern_yield(PRI_USER);
3538 		goto allproc_loop;
3539 	}
3540 }
3541 
3542 void
3543 resume_all_proc(void)
3544 {
3545 	struct proc *cp, *p;
3546 
3547 	cp = curproc;
3548 	sx_xlock(&allproc_lock);
3549 again:
3550 	LIST_REMOVE(cp, p_list);
3551 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3552 	for (;;) {
3553 		p = LIST_NEXT(cp, p_list);
3554 		if (p == NULL)
3555 			break;
3556 		LIST_REMOVE(cp, p_list);
3557 		LIST_INSERT_AFTER(p, cp, p_list);
3558 		PROC_LOCK(p);
3559 		if ((p->p_flag & P_TOTAL_STOP) != 0) {
3560 			sx_xunlock(&allproc_lock);
3561 			_PHOLD(p);
3562 			thread_single_end(p, SINGLE_ALLPROC);
3563 			_PRELE(p);
3564 			PROC_UNLOCK(p);
3565 			sx_xlock(&allproc_lock);
3566 		} else {
3567 			PROC_UNLOCK(p);
3568 		}
3569 	}
3570 	/*  Did the loop above missed any stopped process ? */
3571 	FOREACH_PROC_IN_SYSTEM(p) {
3572 		/* No need for proc lock. */
3573 		if ((p->p_flag & P_TOTAL_STOP) != 0)
3574 			goto again;
3575 	}
3576 	sx_xunlock(&allproc_lock);
3577 
3578 	stop_all_proc_unblock();
3579 }
3580 
3581 /* #define	TOTAL_STOP_DEBUG	1 */
3582 #ifdef TOTAL_STOP_DEBUG
3583 volatile static int ap_resume;
3584 #include <sys/mount.h>
3585 
3586 static int
3587 sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS)
3588 {
3589 	int error, val;
3590 
3591 	val = 0;
3592 	ap_resume = 0;
3593 	error = sysctl_handle_int(oidp, &val, 0, req);
3594 	if (error != 0 || req->newptr == NULL)
3595 		return (error);
3596 	if (val != 0) {
3597 		stop_all_proc();
3598 		syncer_suspend();
3599 		while (ap_resume == 0)
3600 			;
3601 		syncer_resume();
3602 		resume_all_proc();
3603 	}
3604 	return (0);
3605 }
3606 
3607 SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW |
3608     CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0,
3609     sysctl_debug_stop_all_proc, "I",
3610     "");
3611 #endif
3612