xref: /freebsd/sys/kern/kern_proc.c (revision 9bc300465e48e19d794d88d0c158a2adb92c7197)
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/jail.h>
46 #include <sys/kernel.h>
47 #include <sys/limits.h>
48 #include <sys/lock.h>
49 #include <sys/loginclass.h>
50 #include <sys/malloc.h>
51 #include <sys/mman.h>
52 #include <sys/mount.h>
53 #include <sys/mutex.h>
54 #include <sys/namei.h>
55 #include <sys/proc.h>
56 #include <sys/ptrace.h>
57 #include <sys/refcount.h>
58 #include <sys/resourcevar.h>
59 #include <sys/rwlock.h>
60 #include <sys/sbuf.h>
61 #include <sys/sysent.h>
62 #include <sys/sched.h>
63 #include <sys/smp.h>
64 #include <sys/stack.h>
65 #include <sys/stat.h>
66 #include <sys/dtrace_bsd.h>
67 #include <sys/sysctl.h>
68 #include <sys/filedesc.h>
69 #include <sys/tty.h>
70 #include <sys/signalvar.h>
71 #include <sys/sdt.h>
72 #include <sys/sx.h>
73 #include <sys/user.h>
74 #include <sys/vnode.h>
75 #include <sys/wait.h>
76 #ifdef KTRACE
77 #include <sys/ktrace.h>
78 #endif
79 
80 #ifdef DDB
81 #include <ddb/ddb.h>
82 #endif
83 
84 #include <vm/vm.h>
85 #include <vm/vm_param.h>
86 #include <vm/vm_extern.h>
87 #include <vm/pmap.h>
88 #include <vm/vm_map.h>
89 #include <vm/vm_object.h>
90 #include <vm/vm_page.h>
91 #include <vm/uma.h>
92 
93 #include <fs/devfs/devfs.h>
94 
95 #ifdef COMPAT_FREEBSD32
96 #include <compat/freebsd32/freebsd32.h>
97 #include <compat/freebsd32/freebsd32_util.h>
98 #endif
99 
100 SDT_PROVIDER_DEFINE(proc);
101 
102 MALLOC_DEFINE(M_SESSION, "session", "session header");
103 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
104 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
105 
106 static void doenterpgrp(struct proc *, struct pgrp *);
107 static void orphanpg(struct pgrp *pg);
108 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp);
109 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp);
110 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp,
111     int preferthread);
112 static void pgdelete(struct pgrp *);
113 static int pgrp_init(void *mem, int size, int flags);
114 static int proc_ctor(void *mem, int size, void *arg, int flags);
115 static void proc_dtor(void *mem, int size, void *arg);
116 static int proc_init(void *mem, int size, int flags);
117 static void proc_fini(void *mem, int size);
118 static void pargs_free(struct pargs *pa);
119 
120 /*
121  * Other process lists
122  */
123 struct pidhashhead *pidhashtbl = NULL;
124 struct sx *pidhashtbl_lock;
125 u_long pidhash;
126 u_long pidhashlock;
127 struct pgrphashhead *pgrphashtbl;
128 u_long pgrphash;
129 struct proclist allproc = LIST_HEAD_INITIALIZER(allproc);
130 struct sx __exclusive_cache_line allproc_lock;
131 struct sx __exclusive_cache_line proctree_lock;
132 struct mtx __exclusive_cache_line ppeers_lock;
133 struct mtx __exclusive_cache_line procid_lock;
134 uma_zone_t proc_zone;
135 uma_zone_t pgrp_zone;
136 
137 /*
138  * The offset of various fields in struct proc and struct thread.
139  * These are used by kernel debuggers to enumerate kernel threads and
140  * processes.
141  */
142 const int proc_off_p_pid = offsetof(struct proc, p_pid);
143 const int proc_off_p_comm = offsetof(struct proc, p_comm);
144 const int proc_off_p_list = offsetof(struct proc, p_list);
145 const int proc_off_p_hash = offsetof(struct proc, p_hash);
146 const int proc_off_p_threads = offsetof(struct proc, p_threads);
147 const int thread_off_td_tid = offsetof(struct thread, td_tid);
148 const int thread_off_td_name = offsetof(struct thread, td_name);
149 const int thread_off_td_oncpu = offsetof(struct thread, td_oncpu);
150 const int thread_off_td_pcb = offsetof(struct thread, td_pcb);
151 const int thread_off_td_plist = offsetof(struct thread, td_plist);
152 
153 EVENTHANDLER_LIST_DEFINE(process_ctor);
154 EVENTHANDLER_LIST_DEFINE(process_dtor);
155 EVENTHANDLER_LIST_DEFINE(process_init);
156 EVENTHANDLER_LIST_DEFINE(process_fini);
157 EVENTHANDLER_LIST_DEFINE(process_exit);
158 EVENTHANDLER_LIST_DEFINE(process_fork);
159 EVENTHANDLER_LIST_DEFINE(process_exec);
160 
161 int kstack_pages = KSTACK_PAGES;
162 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RDTUN | CTLFLAG_NOFETCH,
163     &kstack_pages, 0,
164     "Kernel stack size in pages");
165 static int vmmap_skip_res_cnt = 0;
166 SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW,
167     &vmmap_skip_res_cnt, 0,
168     "Skip calculation of the pages resident count in kern.proc.vmmap");
169 
170 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
171 #ifdef COMPAT_FREEBSD32
172 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
173 #endif
174 
175 /*
176  * Initialize global process hashing structures.
177  */
178 void
179 procinit(void)
180 {
181 	u_long i;
182 
183 	sx_init(&allproc_lock, "allproc");
184 	sx_init(&proctree_lock, "proctree");
185 	mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
186 	mtx_init(&procid_lock, "procid", NULL, MTX_DEF);
187 	pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
188 	pidhashlock = (pidhash + 1) / 64;
189 	if (pidhashlock > 0)
190 		pidhashlock--;
191 	pidhashtbl_lock = malloc(sizeof(*pidhashtbl_lock) * (pidhashlock + 1),
192 	    M_PROC, M_WAITOK | M_ZERO);
193 	for (i = 0; i < pidhashlock + 1; i++)
194 		sx_init_flags(&pidhashtbl_lock[i], "pidhash", SX_DUPOK);
195 	pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
196 	proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
197 	    proc_ctor, proc_dtor, proc_init, proc_fini,
198 	    UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
199 	pgrp_zone = uma_zcreate("PGRP", sizeof(struct pgrp), NULL, NULL,
200 	    pgrp_init, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
201 	uihashinit();
202 }
203 
204 /*
205  * Prepare a proc for use.
206  */
207 static int
208 proc_ctor(void *mem, int size, void *arg, int flags)
209 {
210 	struct proc *p;
211 	struct thread *td;
212 
213 	p = (struct proc *)mem;
214 #ifdef KDTRACE_HOOKS
215 	kdtrace_proc_ctor(p);
216 #endif
217 	EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
218 	td = FIRST_THREAD_IN_PROC(p);
219 	if (td != NULL) {
220 		/* Make sure all thread constructors are executed */
221 		EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
222 	}
223 	return (0);
224 }
225 
226 /*
227  * Reclaim a proc after use.
228  */
229 static void
230 proc_dtor(void *mem, int size, void *arg)
231 {
232 	struct proc *p;
233 	struct thread *td;
234 
235 	/* INVARIANTS checks go here */
236 	p = (struct proc *)mem;
237 	td = FIRST_THREAD_IN_PROC(p);
238 	if (td != NULL) {
239 #ifdef INVARIANTS
240 		KASSERT((p->p_numthreads == 1),
241 		    ("bad number of threads in exiting process"));
242 		KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
243 #endif
244 		/* Free all OSD associated to this thread. */
245 		osd_thread_exit(td);
246 		ast_kclear(td);
247 
248 		/* Make sure all thread destructors are executed */
249 		EVENTHANDLER_DIRECT_INVOKE(thread_dtor, td);
250 	}
251 	EVENTHANDLER_DIRECT_INVOKE(process_dtor, p);
252 #ifdef KDTRACE_HOOKS
253 	kdtrace_proc_dtor(p);
254 #endif
255 	if (p->p_ksi != NULL)
256 		KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
257 }
258 
259 /*
260  * Initialize type-stable parts of a proc (when newly created).
261  */
262 static int
263 proc_init(void *mem, int size, int flags)
264 {
265 	struct proc *p;
266 
267 	p = (struct proc *)mem;
268 	mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK | MTX_NEW);
269 	mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_NEW);
270 	mtx_init(&p->p_statmtx, "pstatl", NULL, MTX_SPIN | MTX_NEW);
271 	mtx_init(&p->p_itimmtx, "pitiml", NULL, MTX_SPIN | MTX_NEW);
272 	mtx_init(&p->p_profmtx, "pprofl", NULL, MTX_SPIN | MTX_NEW);
273 	cv_init(&p->p_pwait, "ppwait");
274 	TAILQ_INIT(&p->p_threads);	     /* all threads in proc */
275 	EVENTHANDLER_DIRECT_INVOKE(process_init, p);
276 	p->p_stats = pstats_alloc();
277 	p->p_pgrp = NULL;
278 	TAILQ_INIT(&p->p_kqtim_stop);
279 	return (0);
280 }
281 
282 /*
283  * UMA should ensure that this function is never called.
284  * Freeing a proc structure would violate type stability.
285  */
286 static void
287 proc_fini(void *mem, int size)
288 {
289 #ifdef notnow
290 	struct proc *p;
291 
292 	p = (struct proc *)mem;
293 	EVENTHANDLER_DIRECT_INVOKE(process_fini, p);
294 	pstats_free(p->p_stats);
295 	thread_free(FIRST_THREAD_IN_PROC(p));
296 	mtx_destroy(&p->p_mtx);
297 	if (p->p_ksi != NULL)
298 		ksiginfo_free(p->p_ksi);
299 #else
300 	panic("proc reclaimed");
301 #endif
302 }
303 
304 static int
305 pgrp_init(void *mem, int size, int flags)
306 {
307 	struct pgrp *pg;
308 
309 	pg = mem;
310 	mtx_init(&pg->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
311 	sx_init(&pg->pg_killsx, "killpg racer");
312 	return (0);
313 }
314 
315 /*
316  * PID space management.
317  *
318  * These bitmaps are used by fork_findpid.
319  */
320 bitstr_t bit_decl(proc_id_pidmap, PID_MAX);
321 bitstr_t bit_decl(proc_id_grpidmap, PID_MAX);
322 bitstr_t bit_decl(proc_id_sessidmap, PID_MAX);
323 bitstr_t bit_decl(proc_id_reapmap, PID_MAX);
324 
325 static bitstr_t *proc_id_array[] = {
326 	proc_id_pidmap,
327 	proc_id_grpidmap,
328 	proc_id_sessidmap,
329 	proc_id_reapmap,
330 };
331 
332 void
333 proc_id_set(int type, pid_t id)
334 {
335 
336 	KASSERT(type >= 0 && type < nitems(proc_id_array),
337 	    ("invalid type %d\n", type));
338 	mtx_lock(&procid_lock);
339 	KASSERT(bit_test(proc_id_array[type], id) == 0,
340 	    ("bit %d already set in %d\n", id, type));
341 	bit_set(proc_id_array[type], id);
342 	mtx_unlock(&procid_lock);
343 }
344 
345 void
346 proc_id_set_cond(int type, pid_t id)
347 {
348 
349 	KASSERT(type >= 0 && type < nitems(proc_id_array),
350 	    ("invalid type %d\n", type));
351 	if (bit_test(proc_id_array[type], id))
352 		return;
353 	mtx_lock(&procid_lock);
354 	bit_set(proc_id_array[type], id);
355 	mtx_unlock(&procid_lock);
356 }
357 
358 void
359 proc_id_clear(int type, pid_t id)
360 {
361 
362 	KASSERT(type >= 0 && type < nitems(proc_id_array),
363 	    ("invalid type %d\n", type));
364 	mtx_lock(&procid_lock);
365 	KASSERT(bit_test(proc_id_array[type], id) != 0,
366 	    ("bit %d not set in %d\n", id, type));
367 	bit_clear(proc_id_array[type], id);
368 	mtx_unlock(&procid_lock);
369 }
370 
371 /*
372  * Is p an inferior of the current process?
373  */
374 int
375 inferior(struct proc *p)
376 {
377 
378 	sx_assert(&proctree_lock, SX_LOCKED);
379 	PROC_LOCK_ASSERT(p, MA_OWNED);
380 	for (; p != curproc; p = proc_realparent(p)) {
381 		if (p->p_pid == 0)
382 			return (0);
383 	}
384 	return (1);
385 }
386 
387 /*
388  * Shared lock all the pid hash lists.
389  */
390 void
391 pidhash_slockall(void)
392 {
393 	u_long i;
394 
395 	for (i = 0; i < pidhashlock + 1; i++)
396 		sx_slock(&pidhashtbl_lock[i]);
397 }
398 
399 /*
400  * Shared unlock all the pid hash lists.
401  */
402 void
403 pidhash_sunlockall(void)
404 {
405 	u_long i;
406 
407 	for (i = 0; i < pidhashlock + 1; i++)
408 		sx_sunlock(&pidhashtbl_lock[i]);
409 }
410 
411 /*
412  * Similar to pfind_any(), this function finds zombies.
413  */
414 struct proc *
415 pfind_any_locked(pid_t pid)
416 {
417 	struct proc *p;
418 
419 	sx_assert(PIDHASHLOCK(pid), SX_LOCKED);
420 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
421 		if (p->p_pid == pid) {
422 			PROC_LOCK(p);
423 			if (p->p_state == PRS_NEW) {
424 				PROC_UNLOCK(p);
425 				p = NULL;
426 			}
427 			break;
428 		}
429 	}
430 	return (p);
431 }
432 
433 /*
434  * Locate a process by number.
435  *
436  * By not returning processes in the PRS_NEW state, we allow callers to avoid
437  * testing for that condition to avoid dereferencing p_ucred, et al.
438  */
439 static __always_inline struct proc *
440 _pfind(pid_t pid, bool zombie)
441 {
442 	struct proc *p;
443 
444 	p = curproc;
445 	if (p->p_pid == pid) {
446 		PROC_LOCK(p);
447 		return (p);
448 	}
449 	sx_slock(PIDHASHLOCK(pid));
450 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
451 		if (p->p_pid == pid) {
452 			PROC_LOCK(p);
453 			if (p->p_state == PRS_NEW ||
454 			    (!zombie && p->p_state == PRS_ZOMBIE)) {
455 				PROC_UNLOCK(p);
456 				p = NULL;
457 			}
458 			break;
459 		}
460 	}
461 	sx_sunlock(PIDHASHLOCK(pid));
462 	return (p);
463 }
464 
465 struct proc *
466 pfind(pid_t pid)
467 {
468 
469 	return (_pfind(pid, false));
470 }
471 
472 /*
473  * Same as pfind but allow zombies.
474  */
475 struct proc *
476 pfind_any(pid_t pid)
477 {
478 
479 	return (_pfind(pid, true));
480 }
481 
482 /*
483  * Locate a process group by number.
484  * The caller must hold proctree_lock.
485  */
486 struct pgrp *
487 pgfind(pid_t pgid)
488 {
489 	struct pgrp *pgrp;
490 
491 	sx_assert(&proctree_lock, SX_LOCKED);
492 
493 	LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
494 		if (pgrp->pg_id == pgid) {
495 			PGRP_LOCK(pgrp);
496 			return (pgrp);
497 		}
498 	}
499 	return (NULL);
500 }
501 
502 /*
503  * Locate process and do additional manipulations, depending on flags.
504  */
505 int
506 pget(pid_t pid, int flags, struct proc **pp)
507 {
508 	struct proc *p;
509 	struct thread *td1;
510 	int error;
511 
512 	p = curproc;
513 	if (p->p_pid == pid) {
514 		PROC_LOCK(p);
515 	} else {
516 		p = NULL;
517 		if (pid <= PID_MAX) {
518 			if ((flags & PGET_NOTWEXIT) == 0)
519 				p = pfind_any(pid);
520 			else
521 				p = pfind(pid);
522 		} else if ((flags & PGET_NOTID) == 0) {
523 			td1 = tdfind(pid, -1);
524 			if (td1 != NULL)
525 				p = td1->td_proc;
526 		}
527 		if (p == NULL)
528 			return (ESRCH);
529 		if ((flags & PGET_CANSEE) != 0) {
530 			error = p_cansee(curthread, p);
531 			if (error != 0)
532 				goto errout;
533 		}
534 	}
535 	if ((flags & PGET_CANDEBUG) != 0) {
536 		error = p_candebug(curthread, p);
537 		if (error != 0)
538 			goto errout;
539 	}
540 	if ((flags & PGET_ISCURRENT) != 0 && curproc != p) {
541 		error = EPERM;
542 		goto errout;
543 	}
544 	if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) {
545 		error = ESRCH;
546 		goto errout;
547 	}
548 	if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) {
549 		/*
550 		 * XXXRW: Not clear ESRCH is the right error during proc
551 		 * execve().
552 		 */
553 		error = ESRCH;
554 		goto errout;
555 	}
556 	if ((flags & PGET_HOLD) != 0) {
557 		_PHOLD(p);
558 		PROC_UNLOCK(p);
559 	}
560 	*pp = p;
561 	return (0);
562 errout:
563 	PROC_UNLOCK(p);
564 	return (error);
565 }
566 
567 /*
568  * Create a new process group.
569  * pgid must be equal to the pid of p.
570  * Begin a new session if required.
571  */
572 int
573 enterpgrp(struct proc *p, pid_t pgid, struct pgrp *pgrp, struct session *sess)
574 {
575 	struct pgrp *old_pgrp;
576 
577 	sx_assert(&proctree_lock, SX_XLOCKED);
578 
579 	KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
580 	KASSERT(p->p_pid == pgid,
581 	    ("enterpgrp: new pgrp and pid != pgid"));
582 	KASSERT(pgfind(pgid) == NULL,
583 	    ("enterpgrp: pgrp with pgid exists"));
584 	KASSERT(!SESS_LEADER(p),
585 	    ("enterpgrp: session leader attempted setpgrp"));
586 
587 	old_pgrp = p->p_pgrp;
588 	if (!sx_try_xlock(&old_pgrp->pg_killsx)) {
589 		sx_xunlock(&proctree_lock);
590 		sx_xlock(&old_pgrp->pg_killsx);
591 		sx_xunlock(&old_pgrp->pg_killsx);
592 		return (ERESTART);
593 	}
594 	MPASS(old_pgrp == p->p_pgrp);
595 
596 	if (sess != NULL) {
597 		/*
598 		 * new session
599 		 */
600 		mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
601 		PROC_LOCK(p);
602 		p->p_flag &= ~P_CONTROLT;
603 		PROC_UNLOCK(p);
604 		PGRP_LOCK(pgrp);
605 		sess->s_leader = p;
606 		sess->s_sid = p->p_pid;
607 		proc_id_set(PROC_ID_SESSION, p->p_pid);
608 		refcount_init(&sess->s_count, 1);
609 		sess->s_ttyvp = NULL;
610 		sess->s_ttydp = NULL;
611 		sess->s_ttyp = NULL;
612 		bcopy(p->p_session->s_login, sess->s_login,
613 			    sizeof(sess->s_login));
614 		pgrp->pg_session = sess;
615 		KASSERT(p == curproc,
616 		    ("enterpgrp: mksession and p != curproc"));
617 	} else {
618 		pgrp->pg_session = p->p_session;
619 		sess_hold(pgrp->pg_session);
620 		PGRP_LOCK(pgrp);
621 	}
622 	pgrp->pg_id = pgid;
623 	proc_id_set(PROC_ID_GROUP, p->p_pid);
624 	LIST_INIT(&pgrp->pg_members);
625 	pgrp->pg_flags = 0;
626 
627 	/*
628 	 * As we have an exclusive lock of proctree_lock,
629 	 * this should not deadlock.
630 	 */
631 	LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
632 	SLIST_INIT(&pgrp->pg_sigiolst);
633 	PGRP_UNLOCK(pgrp);
634 
635 	doenterpgrp(p, pgrp);
636 
637 	sx_xunlock(&old_pgrp->pg_killsx);
638 	return (0);
639 }
640 
641 /*
642  * Move p to an existing process group
643  */
644 int
645 enterthispgrp(struct proc *p, struct pgrp *pgrp)
646 {
647 	struct pgrp *old_pgrp;
648 
649 	sx_assert(&proctree_lock, SX_XLOCKED);
650 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
651 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
652 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
653 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
654 	KASSERT(pgrp->pg_session == p->p_session,
655 	    ("%s: pgrp's session %p, p->p_session %p proc %p\n",
656 	    __func__, pgrp->pg_session, p->p_session, p));
657 	KASSERT(pgrp != p->p_pgrp,
658 	    ("%s: p %p belongs to pgrp %p", __func__, p, pgrp));
659 
660 	old_pgrp = p->p_pgrp;
661 	if (!sx_try_xlock(&old_pgrp->pg_killsx)) {
662 		sx_xunlock(&proctree_lock);
663 		sx_xlock(&old_pgrp->pg_killsx);
664 		sx_xunlock(&old_pgrp->pg_killsx);
665 		return (ERESTART);
666 	}
667 	MPASS(old_pgrp == p->p_pgrp);
668 	if (!sx_try_xlock(&pgrp->pg_killsx)) {
669 		sx_xunlock(&old_pgrp->pg_killsx);
670 		sx_xunlock(&proctree_lock);
671 		sx_xlock(&pgrp->pg_killsx);
672 		sx_xunlock(&pgrp->pg_killsx);
673 		return (ERESTART);
674 	}
675 
676 	doenterpgrp(p, pgrp);
677 
678 	sx_xunlock(&pgrp->pg_killsx);
679 	sx_xunlock(&old_pgrp->pg_killsx);
680 	return (0);
681 }
682 
683 /*
684  * If true, any child of q which belongs to group pgrp, qualifies the
685  * process group pgrp as not orphaned.
686  */
687 static bool
688 isjobproc(struct proc *q, struct pgrp *pgrp)
689 {
690 	sx_assert(&proctree_lock, SX_LOCKED);
691 
692 	return (q->p_pgrp != pgrp &&
693 	    q->p_pgrp->pg_session == pgrp->pg_session);
694 }
695 
696 static struct proc *
697 jobc_reaper(struct proc *p)
698 {
699 	struct proc *pp;
700 
701 	sx_assert(&proctree_lock, SA_LOCKED);
702 
703 	for (pp = p;;) {
704 		pp = pp->p_reaper;
705 		if (pp->p_reaper == pp ||
706 		    (pp->p_treeflag & P_TREE_GRPEXITED) == 0)
707 			return (pp);
708 	}
709 }
710 
711 static struct proc *
712 jobc_parent(struct proc *p, struct proc *p_exiting)
713 {
714 	struct proc *pp;
715 
716 	sx_assert(&proctree_lock, SA_LOCKED);
717 
718 	pp = proc_realparent(p);
719 	if (pp->p_pptr == NULL || pp == p_exiting ||
720 	    (pp->p_treeflag & P_TREE_GRPEXITED) == 0)
721 		return (pp);
722 	return (jobc_reaper(pp));
723 }
724 
725 static int
726 pgrp_calc_jobc(struct pgrp *pgrp)
727 {
728 	struct proc *q;
729 	int cnt;
730 
731 #ifdef INVARIANTS
732 	if (!mtx_owned(&pgrp->pg_mtx))
733 		sx_assert(&proctree_lock, SA_LOCKED);
734 #endif
735 
736 	cnt = 0;
737 	LIST_FOREACH(q, &pgrp->pg_members, p_pglist) {
738 		if ((q->p_treeflag & P_TREE_GRPEXITED) != 0 ||
739 		    q->p_pptr == NULL)
740 			continue;
741 		if (isjobproc(jobc_parent(q, NULL), pgrp))
742 			cnt++;
743 	}
744 	return (cnt);
745 }
746 
747 /*
748  * Move p to a process group
749  */
750 static void
751 doenterpgrp(struct proc *p, struct pgrp *pgrp)
752 {
753 	struct pgrp *savepgrp;
754 	struct proc *pp;
755 
756 	sx_assert(&proctree_lock, SX_XLOCKED);
757 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
758 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
759 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
760 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
761 
762 	savepgrp = p->p_pgrp;
763 	pp = jobc_parent(p, NULL);
764 
765 	PGRP_LOCK(pgrp);
766 	PGRP_LOCK(savepgrp);
767 	if (isjobproc(pp, savepgrp) && pgrp_calc_jobc(savepgrp) == 1)
768 		orphanpg(savepgrp);
769 	PROC_LOCK(p);
770 	LIST_REMOVE(p, p_pglist);
771 	p->p_pgrp = pgrp;
772 	PROC_UNLOCK(p);
773 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
774 	if (isjobproc(pp, pgrp))
775 		pgrp->pg_flags &= ~PGRP_ORPHANED;
776 	PGRP_UNLOCK(savepgrp);
777 	PGRP_UNLOCK(pgrp);
778 	if (LIST_EMPTY(&savepgrp->pg_members))
779 		pgdelete(savepgrp);
780 }
781 
782 /*
783  * remove process from process group
784  */
785 int
786 leavepgrp(struct proc *p)
787 {
788 	struct pgrp *savepgrp;
789 
790 	sx_assert(&proctree_lock, SX_XLOCKED);
791 	savepgrp = p->p_pgrp;
792 	PGRP_LOCK(savepgrp);
793 	PROC_LOCK(p);
794 	LIST_REMOVE(p, p_pglist);
795 	p->p_pgrp = NULL;
796 	PROC_UNLOCK(p);
797 	PGRP_UNLOCK(savepgrp);
798 	if (LIST_EMPTY(&savepgrp->pg_members))
799 		pgdelete(savepgrp);
800 	return (0);
801 }
802 
803 /*
804  * delete a process group
805  */
806 static void
807 pgdelete(struct pgrp *pgrp)
808 {
809 	struct session *savesess;
810 	struct tty *tp;
811 
812 	sx_assert(&proctree_lock, SX_XLOCKED);
813 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
814 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
815 
816 	/*
817 	 * Reset any sigio structures pointing to us as a result of
818 	 * F_SETOWN with our pgid.  The proctree lock ensures that
819 	 * new sigio structures will not be added after this point.
820 	 */
821 	funsetownlst(&pgrp->pg_sigiolst);
822 
823 	PGRP_LOCK(pgrp);
824 	tp = pgrp->pg_session->s_ttyp;
825 	LIST_REMOVE(pgrp, pg_hash);
826 	savesess = pgrp->pg_session;
827 	PGRP_UNLOCK(pgrp);
828 
829 	/* Remove the reference to the pgrp before deallocating it. */
830 	if (tp != NULL) {
831 		tty_lock(tp);
832 		tty_rel_pgrp(tp, pgrp);
833 	}
834 
835 	proc_id_clear(PROC_ID_GROUP, pgrp->pg_id);
836 	uma_zfree(pgrp_zone, pgrp);
837 	sess_release(savesess);
838 }
839 
840 
841 static void
842 fixjobc_kill(struct proc *p)
843 {
844 	struct proc *q;
845 	struct pgrp *pgrp;
846 
847 	sx_assert(&proctree_lock, SX_LOCKED);
848 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
849 	pgrp = p->p_pgrp;
850 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
851 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
852 
853 	/*
854 	 * p no longer affects process group orphanage for children.
855 	 * It is marked by the flag because p is only physically
856 	 * removed from its process group on wait(2).
857 	 */
858 	MPASS((p->p_treeflag & P_TREE_GRPEXITED) == 0);
859 	p->p_treeflag |= P_TREE_GRPEXITED;
860 
861 	/*
862 	 * Check if exiting p orphans its own group.
863 	 */
864 	pgrp = p->p_pgrp;
865 	if (isjobproc(jobc_parent(p, NULL), pgrp)) {
866 		PGRP_LOCK(pgrp);
867 		if (pgrp_calc_jobc(pgrp) == 0)
868 			orphanpg(pgrp);
869 		PGRP_UNLOCK(pgrp);
870 	}
871 
872 	/*
873 	 * Check this process' children to see whether they qualify
874 	 * their process groups after reparenting to reaper.
875 	 */
876 	LIST_FOREACH(q, &p->p_children, p_sibling) {
877 		pgrp = q->p_pgrp;
878 		PGRP_LOCK(pgrp);
879 		if (pgrp_calc_jobc(pgrp) == 0) {
880 			/*
881 			 * We want to handle exactly the children that
882 			 * has p as realparent.  Then, when calculating
883 			 * jobc_parent for children, we should ignore
884 			 * P_TREE_GRPEXITED flag already set on p.
885 			 */
886 			if (jobc_parent(q, p) == p && isjobproc(p, pgrp))
887 				orphanpg(pgrp);
888 		} else
889 			pgrp->pg_flags &= ~PGRP_ORPHANED;
890 		PGRP_UNLOCK(pgrp);
891 	}
892 	LIST_FOREACH(q, &p->p_orphans, p_orphan) {
893 		pgrp = q->p_pgrp;
894 		PGRP_LOCK(pgrp);
895 		if (pgrp_calc_jobc(pgrp) == 0) {
896 			if (isjobproc(p, pgrp))
897 				orphanpg(pgrp);
898 		} else
899 			pgrp->pg_flags &= ~PGRP_ORPHANED;
900 		PGRP_UNLOCK(pgrp);
901 	}
902 }
903 
904 void
905 killjobc(void)
906 {
907 	struct session *sp;
908 	struct tty *tp;
909 	struct proc *p;
910 	struct vnode *ttyvp;
911 
912 	p = curproc;
913 	MPASS(p->p_flag & P_WEXIT);
914 	sx_assert(&proctree_lock, SX_LOCKED);
915 
916 	if (SESS_LEADER(p)) {
917 		sp = p->p_session;
918 
919 		/*
920 		 * s_ttyp is not zero'd; we use this to indicate that
921 		 * the session once had a controlling terminal. (for
922 		 * logging and informational purposes)
923 		 */
924 		SESS_LOCK(sp);
925 		ttyvp = sp->s_ttyvp;
926 		tp = sp->s_ttyp;
927 		sp->s_ttyvp = NULL;
928 		sp->s_ttydp = NULL;
929 		sp->s_leader = NULL;
930 		SESS_UNLOCK(sp);
931 
932 		/*
933 		 * Signal foreground pgrp and revoke access to
934 		 * controlling terminal if it has not been revoked
935 		 * already.
936 		 *
937 		 * Because the TTY may have been revoked in the mean
938 		 * time and could already have a new session associated
939 		 * with it, make sure we don't send a SIGHUP to a
940 		 * foreground process group that does not belong to this
941 		 * session.
942 		 */
943 
944 		if (tp != NULL) {
945 			tty_lock(tp);
946 			if (tp->t_session == sp)
947 				tty_signal_pgrp(tp, SIGHUP);
948 			tty_unlock(tp);
949 		}
950 
951 		if (ttyvp != NULL) {
952 			sx_xunlock(&proctree_lock);
953 			if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) {
954 				VOP_REVOKE(ttyvp, REVOKEALL);
955 				VOP_UNLOCK(ttyvp);
956 			}
957 			devfs_ctty_unref(ttyvp);
958 			sx_xlock(&proctree_lock);
959 		}
960 	}
961 	fixjobc_kill(p);
962 }
963 
964 /*
965  * A process group has become orphaned, mark it as such for signal
966  * delivery code.  If there are any stopped processes in the group,
967  * hang-up all process in that group.
968  */
969 static void
970 orphanpg(struct pgrp *pg)
971 {
972 	struct proc *p;
973 
974 	PGRP_LOCK_ASSERT(pg, MA_OWNED);
975 
976 	pg->pg_flags |= PGRP_ORPHANED;
977 
978 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
979 		PROC_LOCK(p);
980 		if (P_SHOULDSTOP(p) == P_STOPPED_SIG) {
981 			PROC_UNLOCK(p);
982 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
983 				PROC_LOCK(p);
984 				kern_psignal(p, SIGHUP);
985 				kern_psignal(p, SIGCONT);
986 				PROC_UNLOCK(p);
987 			}
988 			return;
989 		}
990 		PROC_UNLOCK(p);
991 	}
992 }
993 
994 void
995 sess_hold(struct session *s)
996 {
997 
998 	refcount_acquire(&s->s_count);
999 }
1000 
1001 void
1002 sess_release(struct session *s)
1003 {
1004 
1005 	if (refcount_release(&s->s_count)) {
1006 		if (s->s_ttyp != NULL) {
1007 			tty_lock(s->s_ttyp);
1008 			tty_rel_sess(s->s_ttyp, s);
1009 		}
1010 		proc_id_clear(PROC_ID_SESSION, s->s_sid);
1011 		mtx_destroy(&s->s_mtx);
1012 		free(s, M_SESSION);
1013 	}
1014 }
1015 
1016 #ifdef DDB
1017 
1018 static void
1019 db_print_pgrp_one(struct pgrp *pgrp, struct proc *p)
1020 {
1021 	db_printf(
1022 	    "    pid %d at %p pr %d pgrp %p e %d jc %d\n",
1023 	    p->p_pid, p, p->p_pptr == NULL ? -1 : p->p_pptr->p_pid,
1024 	    p->p_pgrp, (p->p_treeflag & P_TREE_GRPEXITED) != 0,
1025 	    p->p_pptr == NULL ? 0 : isjobproc(p->p_pptr, pgrp));
1026 }
1027 
1028 DB_SHOW_COMMAND_FLAGS(pgrpdump, pgrpdump, DB_CMD_MEMSAFE)
1029 {
1030 	struct pgrp *pgrp;
1031 	struct proc *p;
1032 	int i;
1033 
1034 	for (i = 0; i <= pgrphash; i++) {
1035 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
1036 			db_printf("indx %d\n", i);
1037 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
1038 				db_printf(
1039 			"  pgrp %p, pgid %d, sess %p, sesscnt %d, mem %p\n",
1040 				    pgrp, (int)pgrp->pg_id, pgrp->pg_session,
1041 				    pgrp->pg_session->s_count,
1042 				    LIST_FIRST(&pgrp->pg_members));
1043 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist)
1044 					db_print_pgrp_one(pgrp, p);
1045 			}
1046 		}
1047 	}
1048 }
1049 #endif /* DDB */
1050 
1051 /*
1052  * Calculate the kinfo_proc members which contain process-wide
1053  * informations.
1054  * Must be called with the target process locked.
1055  */
1056 static void
1057 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
1058 {
1059 	struct thread *td;
1060 
1061 	PROC_LOCK_ASSERT(p, MA_OWNED);
1062 
1063 	kp->ki_estcpu = 0;
1064 	kp->ki_pctcpu = 0;
1065 	FOREACH_THREAD_IN_PROC(p, td) {
1066 		thread_lock(td);
1067 		kp->ki_pctcpu += sched_pctcpu(td);
1068 		kp->ki_estcpu += sched_estcpu(td);
1069 		thread_unlock(td);
1070 	}
1071 }
1072 
1073 /*
1074  * Fill in any information that is common to all threads in the process.
1075  * Must be called with the target process locked.
1076  */
1077 static void
1078 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
1079 {
1080 	struct thread *td0;
1081 	struct ucred *cred;
1082 	struct sigacts *ps;
1083 	struct timeval boottime;
1084 
1085 	PROC_LOCK_ASSERT(p, MA_OWNED);
1086 
1087 	kp->ki_structsize = sizeof(*kp);
1088 	kp->ki_paddr = p;
1089 	kp->ki_addr =/* p->p_addr; */0; /* XXX */
1090 	kp->ki_args = p->p_args;
1091 	kp->ki_textvp = p->p_textvp;
1092 #ifdef KTRACE
1093 	kp->ki_tracep = ktr_get_tracevp(p, false);
1094 	kp->ki_traceflag = p->p_traceflag;
1095 #endif
1096 	kp->ki_fd = p->p_fd;
1097 	kp->ki_pd = p->p_pd;
1098 	kp->ki_vmspace = p->p_vmspace;
1099 	kp->ki_flag = p->p_flag;
1100 	kp->ki_flag2 = p->p_flag2;
1101 	cred = p->p_ucred;
1102 	if (cred) {
1103 		kp->ki_uid = cred->cr_uid;
1104 		kp->ki_ruid = cred->cr_ruid;
1105 		kp->ki_svuid = cred->cr_svuid;
1106 		kp->ki_cr_flags = 0;
1107 		if (cred->cr_flags & CRED_FLAG_CAPMODE)
1108 			kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
1109 		/* XXX bde doesn't like KI_NGROUPS */
1110 		if (cred->cr_ngroups > KI_NGROUPS) {
1111 			kp->ki_ngroups = KI_NGROUPS;
1112 			kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
1113 		} else
1114 			kp->ki_ngroups = cred->cr_ngroups;
1115 		bcopy(cred->cr_groups, kp->ki_groups,
1116 		    kp->ki_ngroups * sizeof(gid_t));
1117 		kp->ki_rgid = cred->cr_rgid;
1118 		kp->ki_svgid = cred->cr_svgid;
1119 		/* If jailed(cred), emulate the old P_JAILED flag. */
1120 		if (jailed(cred)) {
1121 			kp->ki_flag |= P_JAILED;
1122 			/* If inside the jail, use 0 as a jail ID. */
1123 			if (cred->cr_prison != curthread->td_ucred->cr_prison)
1124 				kp->ki_jid = cred->cr_prison->pr_id;
1125 		}
1126 		strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
1127 		    sizeof(kp->ki_loginclass));
1128 	}
1129 	ps = p->p_sigacts;
1130 	if (ps) {
1131 		mtx_lock(&ps->ps_mtx);
1132 		kp->ki_sigignore = ps->ps_sigignore;
1133 		kp->ki_sigcatch = ps->ps_sigcatch;
1134 		mtx_unlock(&ps->ps_mtx);
1135 	}
1136 	if (p->p_state != PRS_NEW &&
1137 	    p->p_state != PRS_ZOMBIE &&
1138 	    p->p_vmspace != NULL) {
1139 		struct vmspace *vm = p->p_vmspace;
1140 
1141 		kp->ki_size = vm->vm_map.size;
1142 		kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
1143 		FOREACH_THREAD_IN_PROC(p, td0)
1144 			kp->ki_rssize += td0->td_kstack_pages;
1145 		kp->ki_swrss = vm->vm_swrss;
1146 		kp->ki_tsize = vm->vm_tsize;
1147 		kp->ki_dsize = vm->vm_dsize;
1148 		kp->ki_ssize = vm->vm_ssize;
1149 	} else if (p->p_state == PRS_ZOMBIE)
1150 		kp->ki_stat = SZOMB;
1151 	kp->ki_sflag = PS_INMEM;
1152 	/* Calculate legacy swtime as seconds since 'swtick'. */
1153 	kp->ki_swtime = (ticks - p->p_swtick) / hz;
1154 	kp->ki_pid = p->p_pid;
1155 	kp->ki_nice = p->p_nice;
1156 	kp->ki_fibnum = p->p_fibnum;
1157 	kp->ki_start = p->p_stats->p_start;
1158 	getboottime(&boottime);
1159 	timevaladd(&kp->ki_start, &boottime);
1160 	PROC_STATLOCK(p);
1161 	rufetch(p, &kp->ki_rusage);
1162 	kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
1163 	calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
1164 	PROC_STATUNLOCK(p);
1165 	calccru(p, &kp->ki_childutime, &kp->ki_childstime);
1166 	/* Some callers want child times in a single value. */
1167 	kp->ki_childtime = kp->ki_childstime;
1168 	timevaladd(&kp->ki_childtime, &kp->ki_childutime);
1169 
1170 	FOREACH_THREAD_IN_PROC(p, td0)
1171 		kp->ki_cow += td0->td_cow;
1172 
1173 	if (p->p_comm[0] != '\0')
1174 		strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
1175 	if (p->p_sysent && p->p_sysent->sv_name != NULL &&
1176 	    p->p_sysent->sv_name[0] != '\0')
1177 		strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
1178 	kp->ki_siglist = p->p_siglist;
1179 	kp->ki_xstat = KW_EXITCODE(p->p_xexit, p->p_xsig);
1180 	kp->ki_acflag = p->p_acflag;
1181 	kp->ki_lock = p->p_lock;
1182 	if (p->p_pptr) {
1183 		kp->ki_ppid = p->p_oppid;
1184 		if (p->p_flag & P_TRACED)
1185 			kp->ki_tracer = p->p_pptr->p_pid;
1186 	}
1187 }
1188 
1189 /*
1190  * Fill job-related process information.
1191  */
1192 static void
1193 fill_kinfo_proc_pgrp(struct proc *p, struct kinfo_proc *kp)
1194 {
1195 	struct tty *tp;
1196 	struct session *sp;
1197 	struct pgrp *pgrp;
1198 
1199 	sx_assert(&proctree_lock, SA_LOCKED);
1200 	PROC_LOCK_ASSERT(p, MA_OWNED);
1201 
1202 	pgrp = p->p_pgrp;
1203 	if (pgrp == NULL)
1204 		return;
1205 
1206 	kp->ki_pgid = pgrp->pg_id;
1207 	kp->ki_jobc = pgrp_calc_jobc(pgrp);
1208 
1209 	sp = pgrp->pg_session;
1210 	tp = NULL;
1211 
1212 	if (sp != NULL) {
1213 		kp->ki_sid = sp->s_sid;
1214 		SESS_LOCK(sp);
1215 		strlcpy(kp->ki_login, sp->s_login, sizeof(kp->ki_login));
1216 		if (sp->s_ttyvp)
1217 			kp->ki_kiflag |= KI_CTTY;
1218 		if (SESS_LEADER(p))
1219 			kp->ki_kiflag |= KI_SLEADER;
1220 		tp = sp->s_ttyp;
1221 		SESS_UNLOCK(sp);
1222 	}
1223 
1224 	if ((p->p_flag & P_CONTROLT) && tp != NULL) {
1225 		kp->ki_tdev = tty_udev(tp);
1226 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1227 		kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
1228 		if (tp->t_session)
1229 			kp->ki_tsid = tp->t_session->s_sid;
1230 	} else {
1231 		kp->ki_tdev = NODEV;
1232 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1233 	}
1234 }
1235 
1236 /*
1237  * Fill in information that is thread specific.  Must be called with
1238  * target process locked.  If 'preferthread' is set, overwrite certain
1239  * process-related fields that are maintained for both threads and
1240  * processes.
1241  */
1242 static void
1243 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
1244 {
1245 	struct proc *p;
1246 
1247 	p = td->td_proc;
1248 	kp->ki_tdaddr = td;
1249 	PROC_LOCK_ASSERT(p, MA_OWNED);
1250 
1251 	if (preferthread)
1252 		PROC_STATLOCK(p);
1253 	thread_lock(td);
1254 	if (td->td_wmesg != NULL)
1255 		strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
1256 	else
1257 		bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
1258 	if (strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname)) >=
1259 	    sizeof(kp->ki_tdname)) {
1260 		strlcpy(kp->ki_moretdname,
1261 		    td->td_name + sizeof(kp->ki_tdname) - 1,
1262 		    sizeof(kp->ki_moretdname));
1263 	} else {
1264 		bzero(kp->ki_moretdname, sizeof(kp->ki_moretdname));
1265 	}
1266 	if (TD_ON_LOCK(td)) {
1267 		kp->ki_kiflag |= KI_LOCKBLOCK;
1268 		strlcpy(kp->ki_lockname, td->td_lockname,
1269 		    sizeof(kp->ki_lockname));
1270 	} else {
1271 		kp->ki_kiflag &= ~KI_LOCKBLOCK;
1272 		bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
1273 	}
1274 
1275 	if (p->p_state == PRS_NORMAL) { /* approximate. */
1276 		if (TD_ON_RUNQ(td) ||
1277 		    TD_CAN_RUN(td) ||
1278 		    TD_IS_RUNNING(td)) {
1279 			kp->ki_stat = SRUN;
1280 		} else if (P_SHOULDSTOP(p)) {
1281 			kp->ki_stat = SSTOP;
1282 		} else if (TD_IS_SLEEPING(td)) {
1283 			kp->ki_stat = SSLEEP;
1284 		} else if (TD_ON_LOCK(td)) {
1285 			kp->ki_stat = SLOCK;
1286 		} else {
1287 			kp->ki_stat = SWAIT;
1288 		}
1289 	} else if (p->p_state == PRS_ZOMBIE) {
1290 		kp->ki_stat = SZOMB;
1291 	} else {
1292 		kp->ki_stat = SIDL;
1293 	}
1294 
1295 	/* Things in the thread */
1296 	kp->ki_wchan = td->td_wchan;
1297 	kp->ki_pri.pri_level = td->td_priority;
1298 	kp->ki_pri.pri_native = td->td_base_pri;
1299 
1300 	/*
1301 	 * Note: legacy fields; clamp at the old NOCPU value and/or
1302 	 * the maximum u_char CPU value.
1303 	 */
1304 	if (td->td_lastcpu == NOCPU)
1305 		kp->ki_lastcpu_old = NOCPU_OLD;
1306 	else if (td->td_lastcpu > MAXCPU_OLD)
1307 		kp->ki_lastcpu_old = MAXCPU_OLD;
1308 	else
1309 		kp->ki_lastcpu_old = td->td_lastcpu;
1310 
1311 	if (td->td_oncpu == NOCPU)
1312 		kp->ki_oncpu_old = NOCPU_OLD;
1313 	else if (td->td_oncpu > MAXCPU_OLD)
1314 		kp->ki_oncpu_old = MAXCPU_OLD;
1315 	else
1316 		kp->ki_oncpu_old = td->td_oncpu;
1317 
1318 	kp->ki_lastcpu = td->td_lastcpu;
1319 	kp->ki_oncpu = td->td_oncpu;
1320 	kp->ki_tdflags = td->td_flags;
1321 	kp->ki_tid = td->td_tid;
1322 	kp->ki_numthreads = p->p_numthreads;
1323 	kp->ki_pcb = td->td_pcb;
1324 	kp->ki_kstack = (void *)td->td_kstack;
1325 	kp->ki_slptime = (ticks - td->td_slptick) / hz;
1326 	kp->ki_pri.pri_class = td->td_pri_class;
1327 	kp->ki_pri.pri_user = td->td_user_pri;
1328 
1329 	if (preferthread) {
1330 		rufetchtd(td, &kp->ki_rusage);
1331 		kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
1332 		kp->ki_pctcpu = sched_pctcpu(td);
1333 		kp->ki_estcpu = sched_estcpu(td);
1334 		kp->ki_cow = td->td_cow;
1335 	}
1336 
1337 	/* We can't get this anymore but ps etc never used it anyway. */
1338 	kp->ki_rqindex = 0;
1339 
1340 	if (preferthread)
1341 		kp->ki_siglist = td->td_siglist;
1342 	kp->ki_sigmask = td->td_sigmask;
1343 	thread_unlock(td);
1344 	if (preferthread)
1345 		PROC_STATUNLOCK(p);
1346 }
1347 
1348 /*
1349  * Fill in a kinfo_proc structure for the specified process.
1350  * Must be called with the target process locked.
1351  */
1352 void
1353 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1354 {
1355 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1356 
1357 	bzero(kp, sizeof(*kp));
1358 
1359 	fill_kinfo_proc_pgrp(p,kp);
1360 	fill_kinfo_proc_only(p, kp);
1361 	fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1362 	fill_kinfo_aggregate(p, kp);
1363 }
1364 
1365 struct pstats *
1366 pstats_alloc(void)
1367 {
1368 
1369 	return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1370 }
1371 
1372 /*
1373  * Copy parts of p_stats; zero the rest of p_stats (statistics).
1374  */
1375 void
1376 pstats_fork(struct pstats *src, struct pstats *dst)
1377 {
1378 
1379 	bzero(&dst->pstat_startzero,
1380 	    __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1381 	bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1382 	    __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1383 }
1384 
1385 void
1386 pstats_free(struct pstats *ps)
1387 {
1388 
1389 	free(ps, M_SUBPROC);
1390 }
1391 
1392 #ifdef COMPAT_FREEBSD32
1393 
1394 /*
1395  * This function is typically used to copy out the kernel address, so
1396  * it can be replaced by assignment of zero.
1397  */
1398 static inline uint32_t
1399 ptr32_trim(const void *ptr)
1400 {
1401 	uintptr_t uptr;
1402 
1403 	uptr = (uintptr_t)ptr;
1404 	return ((uptr > UINT_MAX) ? 0 : uptr);
1405 }
1406 
1407 #define PTRTRIM_CP(src,dst,fld) \
1408 	do { (dst).fld = ptr32_trim((src).fld); } while (0)
1409 
1410 static void
1411 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1412 {
1413 	int i;
1414 
1415 	bzero(ki32, sizeof(struct kinfo_proc32));
1416 	ki32->ki_structsize = sizeof(struct kinfo_proc32);
1417 	CP(*ki, *ki32, ki_layout);
1418 	PTRTRIM_CP(*ki, *ki32, ki_args);
1419 	PTRTRIM_CP(*ki, *ki32, ki_paddr);
1420 	PTRTRIM_CP(*ki, *ki32, ki_addr);
1421 	PTRTRIM_CP(*ki, *ki32, ki_tracep);
1422 	PTRTRIM_CP(*ki, *ki32, ki_textvp);
1423 	PTRTRIM_CP(*ki, *ki32, ki_fd);
1424 	PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1425 	PTRTRIM_CP(*ki, *ki32, ki_wchan);
1426 	CP(*ki, *ki32, ki_pid);
1427 	CP(*ki, *ki32, ki_ppid);
1428 	CP(*ki, *ki32, ki_pgid);
1429 	CP(*ki, *ki32, ki_tpgid);
1430 	CP(*ki, *ki32, ki_sid);
1431 	CP(*ki, *ki32, ki_tsid);
1432 	CP(*ki, *ki32, ki_jobc);
1433 	CP(*ki, *ki32, ki_tdev);
1434 	CP(*ki, *ki32, ki_tdev_freebsd11);
1435 	CP(*ki, *ki32, ki_siglist);
1436 	CP(*ki, *ki32, ki_sigmask);
1437 	CP(*ki, *ki32, ki_sigignore);
1438 	CP(*ki, *ki32, ki_sigcatch);
1439 	CP(*ki, *ki32, ki_uid);
1440 	CP(*ki, *ki32, ki_ruid);
1441 	CP(*ki, *ki32, ki_svuid);
1442 	CP(*ki, *ki32, ki_rgid);
1443 	CP(*ki, *ki32, ki_svgid);
1444 	CP(*ki, *ki32, ki_ngroups);
1445 	for (i = 0; i < KI_NGROUPS; i++)
1446 		CP(*ki, *ki32, ki_groups[i]);
1447 	CP(*ki, *ki32, ki_size);
1448 	CP(*ki, *ki32, ki_rssize);
1449 	CP(*ki, *ki32, ki_swrss);
1450 	CP(*ki, *ki32, ki_tsize);
1451 	CP(*ki, *ki32, ki_dsize);
1452 	CP(*ki, *ki32, ki_ssize);
1453 	CP(*ki, *ki32, ki_xstat);
1454 	CP(*ki, *ki32, ki_acflag);
1455 	CP(*ki, *ki32, ki_pctcpu);
1456 	CP(*ki, *ki32, ki_estcpu);
1457 	CP(*ki, *ki32, ki_slptime);
1458 	CP(*ki, *ki32, ki_swtime);
1459 	CP(*ki, *ki32, ki_cow);
1460 	CP(*ki, *ki32, ki_runtime);
1461 	TV_CP(*ki, *ki32, ki_start);
1462 	TV_CP(*ki, *ki32, ki_childtime);
1463 	CP(*ki, *ki32, ki_flag);
1464 	CP(*ki, *ki32, ki_kiflag);
1465 	CP(*ki, *ki32, ki_traceflag);
1466 	CP(*ki, *ki32, ki_stat);
1467 	CP(*ki, *ki32, ki_nice);
1468 	CP(*ki, *ki32, ki_lock);
1469 	CP(*ki, *ki32, ki_rqindex);
1470 	CP(*ki, *ki32, ki_oncpu);
1471 	CP(*ki, *ki32, ki_lastcpu);
1472 
1473 	/* XXX TODO: wrap cpu value as appropriate */
1474 	CP(*ki, *ki32, ki_oncpu_old);
1475 	CP(*ki, *ki32, ki_lastcpu_old);
1476 
1477 	bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1478 	bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1479 	bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1480 	bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1481 	bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1482 	bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1483 	bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1484 	bcopy(ki->ki_moretdname, ki32->ki_moretdname, MAXCOMLEN - TDNAMLEN + 1);
1485 	CP(*ki, *ki32, ki_tracer);
1486 	CP(*ki, *ki32, ki_flag2);
1487 	CP(*ki, *ki32, ki_fibnum);
1488 	CP(*ki, *ki32, ki_cr_flags);
1489 	CP(*ki, *ki32, ki_jid);
1490 	CP(*ki, *ki32, ki_numthreads);
1491 	CP(*ki, *ki32, ki_tid);
1492 	CP(*ki, *ki32, ki_pri);
1493 	freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1494 	freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1495 	PTRTRIM_CP(*ki, *ki32, ki_pcb);
1496 	PTRTRIM_CP(*ki, *ki32, ki_kstack);
1497 	PTRTRIM_CP(*ki, *ki32, ki_udata);
1498 	PTRTRIM_CP(*ki, *ki32, ki_tdaddr);
1499 	CP(*ki, *ki32, ki_sflag);
1500 	CP(*ki, *ki32, ki_tdflags);
1501 }
1502 #endif
1503 
1504 static ssize_t
1505 kern_proc_out_size(struct proc *p, int flags)
1506 {
1507 	ssize_t size = 0;
1508 
1509 	PROC_LOCK_ASSERT(p, MA_OWNED);
1510 
1511 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1512 #ifdef COMPAT_FREEBSD32
1513 		if ((flags & KERN_PROC_MASK32) != 0) {
1514 			size += sizeof(struct kinfo_proc32);
1515 		} else
1516 #endif
1517 			size += sizeof(struct kinfo_proc);
1518 	} else {
1519 #ifdef COMPAT_FREEBSD32
1520 		if ((flags & KERN_PROC_MASK32) != 0)
1521 			size += sizeof(struct kinfo_proc32) * p->p_numthreads;
1522 		else
1523 #endif
1524 			size += sizeof(struct kinfo_proc) * p->p_numthreads;
1525 	}
1526 	PROC_UNLOCK(p);
1527 	return (size);
1528 }
1529 
1530 int
1531 kern_proc_out(struct proc *p, struct sbuf *sb, int flags)
1532 {
1533 	struct thread *td;
1534 	struct kinfo_proc ki;
1535 #ifdef COMPAT_FREEBSD32
1536 	struct kinfo_proc32 ki32;
1537 #endif
1538 	int error;
1539 
1540 	PROC_LOCK_ASSERT(p, MA_OWNED);
1541 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1542 
1543 	error = 0;
1544 	fill_kinfo_proc(p, &ki);
1545 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1546 #ifdef COMPAT_FREEBSD32
1547 		if ((flags & KERN_PROC_MASK32) != 0) {
1548 			freebsd32_kinfo_proc_out(&ki, &ki32);
1549 			if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1550 				error = ENOMEM;
1551 		} else
1552 #endif
1553 			if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1554 				error = ENOMEM;
1555 	} else {
1556 		FOREACH_THREAD_IN_PROC(p, td) {
1557 			fill_kinfo_thread(td, &ki, 1);
1558 #ifdef COMPAT_FREEBSD32
1559 			if ((flags & KERN_PROC_MASK32) != 0) {
1560 				freebsd32_kinfo_proc_out(&ki, &ki32);
1561 				if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1562 					error = ENOMEM;
1563 			} else
1564 #endif
1565 				if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1566 					error = ENOMEM;
1567 			if (error != 0)
1568 				break;
1569 		}
1570 	}
1571 	PROC_UNLOCK(p);
1572 	return (error);
1573 }
1574 
1575 static int
1576 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
1577 {
1578 	struct sbuf sb;
1579 	struct kinfo_proc ki;
1580 	int error, error2;
1581 
1582 	if (req->oldptr == NULL)
1583 		return (SYSCTL_OUT(req, 0, kern_proc_out_size(p, flags)));
1584 
1585 	sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req);
1586 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1587 	error = kern_proc_out(p, &sb, flags);
1588 	error2 = sbuf_finish(&sb);
1589 	sbuf_delete(&sb);
1590 	if (error != 0)
1591 		return (error);
1592 	else if (error2 != 0)
1593 		return (error2);
1594 	return (0);
1595 }
1596 
1597 int
1598 proc_iterate(int (*cb)(struct proc *, void *), void *cbarg)
1599 {
1600 	struct proc *p;
1601 	int error, i, j;
1602 
1603 	for (i = 0; i < pidhashlock + 1; i++) {
1604 		sx_slock(&proctree_lock);
1605 		sx_slock(&pidhashtbl_lock[i]);
1606 		for (j = i; j <= pidhash; j += pidhashlock + 1) {
1607 			LIST_FOREACH(p, &pidhashtbl[j], p_hash) {
1608 				if (p->p_state == PRS_NEW)
1609 					continue;
1610 				error = cb(p, cbarg);
1611 				PROC_LOCK_ASSERT(p, MA_NOTOWNED);
1612 				if (error != 0) {
1613 					sx_sunlock(&pidhashtbl_lock[i]);
1614 					sx_sunlock(&proctree_lock);
1615 					return (error);
1616 				}
1617 			}
1618 		}
1619 		sx_sunlock(&pidhashtbl_lock[i]);
1620 		sx_sunlock(&proctree_lock);
1621 	}
1622 	return (0);
1623 }
1624 
1625 struct kern_proc_out_args {
1626 	struct sysctl_req *req;
1627 	int flags;
1628 	int oid_number;
1629 	int *name;
1630 };
1631 
1632 static int
1633 sysctl_kern_proc_iterate(struct proc *p, void *origarg)
1634 {
1635 	struct kern_proc_out_args *arg = origarg;
1636 	int *name = arg->name;
1637 	int oid_number = arg->oid_number;
1638 	int flags = arg->flags;
1639 	struct sysctl_req *req = arg->req;
1640 	int error = 0;
1641 
1642 	PROC_LOCK(p);
1643 
1644 	KASSERT(p->p_ucred != NULL,
1645 	    ("process credential is NULL for non-NEW proc"));
1646 	/*
1647 	 * Show a user only appropriate processes.
1648 	 */
1649 	if (p_cansee(curthread, p))
1650 		goto skip;
1651 	/*
1652 	 * TODO - make more efficient (see notes below).
1653 	 * do by session.
1654 	 */
1655 	switch (oid_number) {
1656 	case KERN_PROC_GID:
1657 		if (p->p_ucred->cr_gid != (gid_t)name[0])
1658 			goto skip;
1659 		break;
1660 
1661 	case KERN_PROC_PGRP:
1662 		/* could do this by traversing pgrp */
1663 		if (p->p_pgrp == NULL ||
1664 		    p->p_pgrp->pg_id != (pid_t)name[0])
1665 			goto skip;
1666 		break;
1667 
1668 	case KERN_PROC_RGID:
1669 		if (p->p_ucred->cr_rgid != (gid_t)name[0])
1670 			goto skip;
1671 		break;
1672 
1673 	case KERN_PROC_SESSION:
1674 		if (p->p_session == NULL ||
1675 		    p->p_session->s_sid != (pid_t)name[0])
1676 			goto skip;
1677 		break;
1678 
1679 	case KERN_PROC_TTY:
1680 		if ((p->p_flag & P_CONTROLT) == 0 ||
1681 		    p->p_session == NULL)
1682 			goto skip;
1683 		/* XXX proctree_lock */
1684 		SESS_LOCK(p->p_session);
1685 		if (p->p_session->s_ttyp == NULL ||
1686 		    tty_udev(p->p_session->s_ttyp) !=
1687 		    (dev_t)name[0]) {
1688 			SESS_UNLOCK(p->p_session);
1689 			goto skip;
1690 		}
1691 		SESS_UNLOCK(p->p_session);
1692 		break;
1693 
1694 	case KERN_PROC_UID:
1695 		if (p->p_ucred->cr_uid != (uid_t)name[0])
1696 			goto skip;
1697 		break;
1698 
1699 	case KERN_PROC_RUID:
1700 		if (p->p_ucred->cr_ruid != (uid_t)name[0])
1701 			goto skip;
1702 		break;
1703 
1704 	case KERN_PROC_PROC:
1705 		break;
1706 
1707 	default:
1708 		break;
1709 	}
1710 	error = sysctl_out_proc(p, req, flags);
1711 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
1712 	return (error);
1713 skip:
1714 	PROC_UNLOCK(p);
1715 	return (0);
1716 }
1717 
1718 static int
1719 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1720 {
1721 	struct kern_proc_out_args iterarg;
1722 	int *name = (int *)arg1;
1723 	u_int namelen = arg2;
1724 	struct proc *p;
1725 	int flags, oid_number;
1726 	int error = 0;
1727 
1728 	oid_number = oidp->oid_number;
1729 	if (oid_number != KERN_PROC_ALL &&
1730 	    (oid_number & KERN_PROC_INC_THREAD) == 0)
1731 		flags = KERN_PROC_NOTHREADS;
1732 	else {
1733 		flags = 0;
1734 		oid_number &= ~KERN_PROC_INC_THREAD;
1735 	}
1736 #ifdef COMPAT_FREEBSD32
1737 	if (req->flags & SCTL_MASK32)
1738 		flags |= KERN_PROC_MASK32;
1739 #endif
1740 	if (oid_number == KERN_PROC_PID) {
1741 		if (namelen != 1)
1742 			return (EINVAL);
1743 		error = sysctl_wire_old_buffer(req, 0);
1744 		if (error)
1745 			return (error);
1746 		sx_slock(&proctree_lock);
1747 		error = pget((pid_t)name[0], PGET_CANSEE, &p);
1748 		if (error == 0)
1749 			error = sysctl_out_proc(p, req, flags);
1750 		sx_sunlock(&proctree_lock);
1751 		return (error);
1752 	}
1753 
1754 	switch (oid_number) {
1755 	case KERN_PROC_ALL:
1756 		if (namelen != 0)
1757 			return (EINVAL);
1758 		break;
1759 	case KERN_PROC_PROC:
1760 		if (namelen != 0 && namelen != 1)
1761 			return (EINVAL);
1762 		break;
1763 	default:
1764 		if (namelen != 1)
1765 			return (EINVAL);
1766 		break;
1767 	}
1768 
1769 	if (req->oldptr == NULL) {
1770 		/* overestimate by 5 procs */
1771 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1772 		if (error)
1773 			return (error);
1774 	} else {
1775 		error = sysctl_wire_old_buffer(req, 0);
1776 		if (error != 0)
1777 			return (error);
1778 	}
1779 	iterarg.flags = flags;
1780 	iterarg.oid_number = oid_number;
1781 	iterarg.req = req;
1782 	iterarg.name = name;
1783 	error = proc_iterate(sysctl_kern_proc_iterate, &iterarg);
1784 	return (error);
1785 }
1786 
1787 struct pargs *
1788 pargs_alloc(int len)
1789 {
1790 	struct pargs *pa;
1791 
1792 	pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1793 		M_WAITOK);
1794 	refcount_init(&pa->ar_ref, 1);
1795 	pa->ar_length = len;
1796 	return (pa);
1797 }
1798 
1799 static void
1800 pargs_free(struct pargs *pa)
1801 {
1802 
1803 	free(pa, M_PARGS);
1804 }
1805 
1806 void
1807 pargs_hold(struct pargs *pa)
1808 {
1809 
1810 	if (pa == NULL)
1811 		return;
1812 	refcount_acquire(&pa->ar_ref);
1813 }
1814 
1815 void
1816 pargs_drop(struct pargs *pa)
1817 {
1818 
1819 	if (pa == NULL)
1820 		return;
1821 	if (refcount_release(&pa->ar_ref))
1822 		pargs_free(pa);
1823 }
1824 
1825 static int
1826 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1827     size_t len)
1828 {
1829 	ssize_t n;
1830 
1831 	/*
1832 	 * This may return a short read if the string is shorter than the chunk
1833 	 * and is aligned at the end of the page, and the following page is not
1834 	 * mapped.
1835 	 */
1836 	n = proc_readmem(td, p, (vm_offset_t)sptr, buf, len);
1837 	if (n <= 0)
1838 		return (ENOMEM);
1839 	return (0);
1840 }
1841 
1842 #define PROC_AUXV_MAX	256	/* Safety limit on auxv size. */
1843 
1844 enum proc_vector_type {
1845 	PROC_ARG,
1846 	PROC_ENV,
1847 	PROC_AUX,
1848 };
1849 
1850 #ifdef COMPAT_FREEBSD32
1851 static int
1852 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1853     size_t *vsizep, enum proc_vector_type type)
1854 {
1855 	struct freebsd32_ps_strings pss;
1856 	Elf32_Auxinfo aux;
1857 	vm_offset_t vptr, ptr;
1858 	uint32_t *proc_vector32;
1859 	char **proc_vector;
1860 	size_t vsize, size;
1861 	int i, error;
1862 
1863 	error = 0;
1864 	if (proc_readmem(td, p, PROC_PS_STRINGS(p), &pss, sizeof(pss)) !=
1865 	    sizeof(pss))
1866 		return (ENOMEM);
1867 	switch (type) {
1868 	case PROC_ARG:
1869 		vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1870 		vsize = pss.ps_nargvstr;
1871 		if (vsize > ARG_MAX)
1872 			return (ENOEXEC);
1873 		size = vsize * sizeof(int32_t);
1874 		break;
1875 	case PROC_ENV:
1876 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1877 		vsize = pss.ps_nenvstr;
1878 		if (vsize > ARG_MAX)
1879 			return (ENOEXEC);
1880 		size = vsize * sizeof(int32_t);
1881 		break;
1882 	case PROC_AUX:
1883 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1884 		    (pss.ps_nenvstr + 1) * sizeof(int32_t);
1885 		if (vptr % 4 != 0)
1886 			return (ENOEXEC);
1887 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1888 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1889 			    sizeof(aux))
1890 				return (ENOMEM);
1891 			if (aux.a_type == AT_NULL)
1892 				break;
1893 			ptr += sizeof(aux);
1894 		}
1895 		if (aux.a_type != AT_NULL)
1896 			return (ENOEXEC);
1897 		vsize = i + 1;
1898 		size = vsize * sizeof(aux);
1899 		break;
1900 	default:
1901 		KASSERT(0, ("Wrong proc vector type: %d", type));
1902 		return (EINVAL);
1903 	}
1904 	proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1905 	if (proc_readmem(td, p, vptr, proc_vector32, size) != size) {
1906 		error = ENOMEM;
1907 		goto done;
1908 	}
1909 	if (type == PROC_AUX) {
1910 		*proc_vectorp = (char **)proc_vector32;
1911 		*vsizep = vsize;
1912 		return (0);
1913 	}
1914 	proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1915 	for (i = 0; i < (int)vsize; i++)
1916 		proc_vector[i] = PTRIN(proc_vector32[i]);
1917 	*proc_vectorp = proc_vector;
1918 	*vsizep = vsize;
1919 done:
1920 	free(proc_vector32, M_TEMP);
1921 	return (error);
1922 }
1923 #endif
1924 
1925 static int
1926 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1927     size_t *vsizep, enum proc_vector_type type)
1928 {
1929 	struct ps_strings pss;
1930 	Elf_Auxinfo aux;
1931 	vm_offset_t vptr, ptr;
1932 	char **proc_vector;
1933 	size_t vsize, size;
1934 	int i;
1935 
1936 #ifdef COMPAT_FREEBSD32
1937 	if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1938 		return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1939 #endif
1940 	if (proc_readmem(td, p, PROC_PS_STRINGS(p), &pss, sizeof(pss)) !=
1941 	    sizeof(pss))
1942 		return (ENOMEM);
1943 	switch (type) {
1944 	case PROC_ARG:
1945 		vptr = (vm_offset_t)pss.ps_argvstr;
1946 		vsize = pss.ps_nargvstr;
1947 		if (vsize > ARG_MAX)
1948 			return (ENOEXEC);
1949 		size = vsize * sizeof(char *);
1950 		break;
1951 	case PROC_ENV:
1952 		vptr = (vm_offset_t)pss.ps_envstr;
1953 		vsize = pss.ps_nenvstr;
1954 		if (vsize > ARG_MAX)
1955 			return (ENOEXEC);
1956 		size = vsize * sizeof(char *);
1957 		break;
1958 	case PROC_AUX:
1959 		/*
1960 		 * The aux array is just above env array on the stack. Check
1961 		 * that the address is naturally aligned.
1962 		 */
1963 		vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1964 		    * sizeof(char *);
1965 #if __ELF_WORD_SIZE == 64
1966 		if (vptr % sizeof(uint64_t) != 0)
1967 #else
1968 		if (vptr % sizeof(uint32_t) != 0)
1969 #endif
1970 			return (ENOEXEC);
1971 		/*
1972 		 * We count the array size reading the aux vectors from the
1973 		 * stack until AT_NULL vector is returned.  So (to keep the code
1974 		 * simple) we read the process stack twice: the first time here
1975 		 * to find the size and the second time when copying the vectors
1976 		 * to the allocated proc_vector.
1977 		 */
1978 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1979 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1980 			    sizeof(aux))
1981 				return (ENOMEM);
1982 			if (aux.a_type == AT_NULL)
1983 				break;
1984 			ptr += sizeof(aux);
1985 		}
1986 		/*
1987 		 * If the PROC_AUXV_MAX entries are iterated over, and we have
1988 		 * not reached AT_NULL, it is most likely we are reading wrong
1989 		 * data: either the process doesn't have auxv array or data has
1990 		 * been modified. Return the error in this case.
1991 		 */
1992 		if (aux.a_type != AT_NULL)
1993 			return (ENOEXEC);
1994 		vsize = i + 1;
1995 		size = vsize * sizeof(aux);
1996 		break;
1997 	default:
1998 		KASSERT(0, ("Wrong proc vector type: %d", type));
1999 		return (EINVAL); /* In case we are built without INVARIANTS. */
2000 	}
2001 	proc_vector = malloc(size, M_TEMP, M_WAITOK);
2002 	if (proc_readmem(td, p, vptr, proc_vector, size) != size) {
2003 		free(proc_vector, M_TEMP);
2004 		return (ENOMEM);
2005 	}
2006 	*proc_vectorp = proc_vector;
2007 	*vsizep = vsize;
2008 
2009 	return (0);
2010 }
2011 
2012 #define GET_PS_STRINGS_CHUNK_SZ	256	/* Chunk size (bytes) for ps_strings operations. */
2013 
2014 static int
2015 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
2016     enum proc_vector_type type)
2017 {
2018 	size_t done, len, nchr, vsize;
2019 	int error, i;
2020 	char **proc_vector, *sptr;
2021 	char pss_string[GET_PS_STRINGS_CHUNK_SZ];
2022 
2023 	PROC_ASSERT_HELD(p);
2024 
2025 	/*
2026 	 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
2027 	 */
2028 	nchr = 2 * (PATH_MAX + ARG_MAX);
2029 
2030 	error = get_proc_vector(td, p, &proc_vector, &vsize, type);
2031 	if (error != 0)
2032 		return (error);
2033 	for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
2034 		/*
2035 		 * The program may have scribbled into its argv array, e.g. to
2036 		 * remove some arguments.  If that has happened, break out
2037 		 * before trying to read from NULL.
2038 		 */
2039 		if (proc_vector[i] == NULL)
2040 			break;
2041 		for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
2042 			error = proc_read_string(td, p, sptr, pss_string,
2043 			    sizeof(pss_string));
2044 			if (error != 0)
2045 				goto done;
2046 			len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
2047 			if (done + len >= nchr)
2048 				len = nchr - done - 1;
2049 			sbuf_bcat(sb, pss_string, len);
2050 			if (len != GET_PS_STRINGS_CHUNK_SZ)
2051 				break;
2052 			done += GET_PS_STRINGS_CHUNK_SZ;
2053 		}
2054 		sbuf_bcat(sb, "", 1);
2055 		done += len + 1;
2056 	}
2057 done:
2058 	free(proc_vector, M_TEMP);
2059 	return (error);
2060 }
2061 
2062 int
2063 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
2064 {
2065 
2066 	return (get_ps_strings(curthread, p, sb, PROC_ARG));
2067 }
2068 
2069 int
2070 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
2071 {
2072 
2073 	return (get_ps_strings(curthread, p, sb, PROC_ENV));
2074 }
2075 
2076 int
2077 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb)
2078 {
2079 	size_t vsize, size;
2080 	char **auxv;
2081 	int error;
2082 
2083 	error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX);
2084 	if (error == 0) {
2085 #ifdef COMPAT_FREEBSD32
2086 		if (SV_PROC_FLAG(p, SV_ILP32) != 0)
2087 			size = vsize * sizeof(Elf32_Auxinfo);
2088 		else
2089 #endif
2090 			size = vsize * sizeof(Elf_Auxinfo);
2091 		if (sbuf_bcat(sb, auxv, size) != 0)
2092 			error = ENOMEM;
2093 		free(auxv, M_TEMP);
2094 	}
2095 	return (error);
2096 }
2097 
2098 /*
2099  * This sysctl allows a process to retrieve the argument list or process
2100  * title for another process without groping around in the address space
2101  * of the other process.  It also allow a process to set its own "process
2102  * title to a string of its own choice.
2103  */
2104 static int
2105 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
2106 {
2107 	int *name = (int *)arg1;
2108 	u_int namelen = arg2;
2109 	struct pargs *newpa, *pa;
2110 	struct proc *p;
2111 	struct sbuf sb;
2112 	int flags, error = 0, error2;
2113 	pid_t pid;
2114 
2115 	if (namelen != 1)
2116 		return (EINVAL);
2117 
2118 	p = curproc;
2119 	pid = (pid_t)name[0];
2120 	if (pid == -1) {
2121 		pid = p->p_pid;
2122 	}
2123 
2124 	/*
2125 	 * If the query is for this process and it is single-threaded, there
2126 	 * is nobody to modify pargs, thus we can just read.
2127 	 */
2128 	if (pid == p->p_pid && p->p_numthreads == 1 && req->newptr == NULL &&
2129 	    (pa = p->p_args) != NULL)
2130 		return (SYSCTL_OUT(req, pa->ar_args, pa->ar_length));
2131 
2132 	flags = PGET_CANSEE;
2133 	if (req->newptr != NULL)
2134 		flags |= PGET_ISCURRENT;
2135 	error = pget(pid, flags, &p);
2136 	if (error)
2137 		return (error);
2138 
2139 	pa = p->p_args;
2140 	if (pa != NULL) {
2141 		pargs_hold(pa);
2142 		PROC_UNLOCK(p);
2143 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
2144 		pargs_drop(pa);
2145 	} else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
2146 		_PHOLD(p);
2147 		PROC_UNLOCK(p);
2148 		sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2149 		sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2150 		error = proc_getargv(curthread, p, &sb);
2151 		error2 = sbuf_finish(&sb);
2152 		PRELE(p);
2153 		sbuf_delete(&sb);
2154 		if (error == 0 && error2 != 0)
2155 			error = error2;
2156 	} else {
2157 		PROC_UNLOCK(p);
2158 	}
2159 	if (error != 0 || req->newptr == NULL)
2160 		return (error);
2161 
2162 	if (req->newlen > ps_arg_cache_limit - sizeof(struct pargs))
2163 		return (ENOMEM);
2164 
2165 	if (req->newlen == 0) {
2166 		/*
2167 		 * Clear the argument pointer, so that we'll fetch arguments
2168 		 * with proc_getargv() until further notice.
2169 		 */
2170 		newpa = NULL;
2171 	} else {
2172 		newpa = pargs_alloc(req->newlen);
2173 		error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
2174 		if (error != 0) {
2175 			pargs_free(newpa);
2176 			return (error);
2177 		}
2178 	}
2179 	PROC_LOCK(p);
2180 	pa = p->p_args;
2181 	p->p_args = newpa;
2182 	PROC_UNLOCK(p);
2183 	pargs_drop(pa);
2184 	return (0);
2185 }
2186 
2187 /*
2188  * This sysctl allows a process to retrieve environment of another process.
2189  */
2190 static int
2191 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
2192 {
2193 	int *name = (int *)arg1;
2194 	u_int namelen = arg2;
2195 	struct proc *p;
2196 	struct sbuf sb;
2197 	int error, error2;
2198 
2199 	if (namelen != 1)
2200 		return (EINVAL);
2201 
2202 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2203 	if (error != 0)
2204 		return (error);
2205 	if ((p->p_flag & P_SYSTEM) != 0) {
2206 		PRELE(p);
2207 		return (0);
2208 	}
2209 
2210 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2211 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2212 	error = proc_getenvv(curthread, p, &sb);
2213 	error2 = sbuf_finish(&sb);
2214 	PRELE(p);
2215 	sbuf_delete(&sb);
2216 	return (error != 0 ? error : error2);
2217 }
2218 
2219 /*
2220  * This sysctl allows a process to retrieve ELF auxiliary vector of
2221  * another process.
2222  */
2223 static int
2224 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
2225 {
2226 	int *name = (int *)arg1;
2227 	u_int namelen = arg2;
2228 	struct proc *p;
2229 	struct sbuf sb;
2230 	int error, error2;
2231 
2232 	if (namelen != 1)
2233 		return (EINVAL);
2234 
2235 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2236 	if (error != 0)
2237 		return (error);
2238 	if ((p->p_flag & P_SYSTEM) != 0) {
2239 		PRELE(p);
2240 		return (0);
2241 	}
2242 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2243 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2244 	error = proc_getauxv(curthread, p, &sb);
2245 	error2 = sbuf_finish(&sb);
2246 	PRELE(p);
2247 	sbuf_delete(&sb);
2248 	return (error != 0 ? error : error2);
2249 }
2250 
2251 /*
2252  * Look up the canonical executable path running in the specified process.
2253  * It tries to return the same hardlink name as was used for execve(2).
2254  * This allows the programs that modify their behavior based on their progname,
2255  * to operate correctly.
2256  *
2257  * Result is returned in retbuf, it must not be freed, similar to vn_fullpath()
2258  *   calling conventions.
2259  * binname is a pointer to temporary string buffer of length MAXPATHLEN,
2260  *   allocated and freed by caller.
2261  * freebuf should be freed by caller, from the M_TEMP malloc type.
2262  */
2263 int
2264 proc_get_binpath(struct proc *p, char *binname, char **retbuf,
2265     char **freebuf)
2266 {
2267 	struct nameidata nd;
2268 	struct vnode *vp, *dvp;
2269 	size_t freepath_size;
2270 	int error;
2271 	bool do_fullpath;
2272 
2273 	PROC_LOCK_ASSERT(p, MA_OWNED);
2274 
2275 	vp = p->p_textvp;
2276 	if (vp == NULL) {
2277 		PROC_UNLOCK(p);
2278 		*retbuf = "";
2279 		*freebuf = NULL;
2280 		return (0);
2281 	}
2282 	vref(vp);
2283 	dvp = p->p_textdvp;
2284 	if (dvp != NULL)
2285 		vref(dvp);
2286 	if (p->p_binname != NULL)
2287 		strlcpy(binname, p->p_binname, MAXPATHLEN);
2288 	PROC_UNLOCK(p);
2289 
2290 	do_fullpath = true;
2291 	*freebuf = NULL;
2292 	if (dvp != NULL && binname[0] != '\0') {
2293 		freepath_size = MAXPATHLEN;
2294 		if (vn_fullpath_hardlink(vp, dvp, binname, strlen(binname),
2295 		    retbuf, freebuf, &freepath_size) == 0) {
2296 			/*
2297 			 * Recheck the looked up path.  The binary
2298 			 * might have been renamed or replaced, in
2299 			 * which case we should not report old name.
2300 			 */
2301 			NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, *retbuf);
2302 			error = namei(&nd);
2303 			if (error == 0) {
2304 				if (nd.ni_vp == vp)
2305 					do_fullpath = false;
2306 				vrele(nd.ni_vp);
2307 				NDFREE_PNBUF(&nd);
2308 			}
2309 		}
2310 	}
2311 	if (do_fullpath) {
2312 		free(*freebuf, M_TEMP);
2313 		*freebuf = NULL;
2314 		error = vn_fullpath(vp, retbuf, freebuf);
2315 	}
2316 	vrele(vp);
2317 	if (dvp != NULL)
2318 		vrele(dvp);
2319 	return (error);
2320 }
2321 
2322 /*
2323  * This sysctl allows a process to retrieve the path of the executable for
2324  * itself or another process.
2325  */
2326 static int
2327 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
2328 {
2329 	pid_t *pidp = (pid_t *)arg1;
2330 	unsigned int arglen = arg2;
2331 	struct proc *p;
2332 	char *retbuf, *freebuf, *binname;
2333 	int error;
2334 
2335 	if (arglen != 1)
2336 		return (EINVAL);
2337 	binname = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
2338 	binname[0] = '\0';
2339 	if (*pidp == -1) {	/* -1 means this process */
2340 		error = 0;
2341 		p = req->td->td_proc;
2342 		PROC_LOCK(p);
2343 	} else {
2344 		error = pget(*pidp, PGET_CANSEE, &p);
2345 	}
2346 
2347 	if (error == 0)
2348 		error = proc_get_binpath(p, binname, &retbuf, &freebuf);
2349 	free(binname, M_TEMP);
2350 	if (error != 0)
2351 		return (error);
2352 	error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
2353 	free(freebuf, M_TEMP);
2354 	return (error);
2355 }
2356 
2357 static int
2358 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
2359 {
2360 	struct proc *p;
2361 	char *sv_name;
2362 	int *name;
2363 	int namelen;
2364 	int error;
2365 
2366 	namelen = arg2;
2367 	if (namelen != 1)
2368 		return (EINVAL);
2369 
2370 	name = (int *)arg1;
2371 	error = pget((pid_t)name[0], PGET_CANSEE, &p);
2372 	if (error != 0)
2373 		return (error);
2374 	sv_name = p->p_sysent->sv_name;
2375 	PROC_UNLOCK(p);
2376 	return (sysctl_handle_string(oidp, sv_name, 0, req));
2377 }
2378 
2379 #ifdef KINFO_OVMENTRY_SIZE
2380 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
2381 #endif
2382 
2383 #ifdef COMPAT_FREEBSD7
2384 static int
2385 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
2386 {
2387 	vm_map_entry_t entry, tmp_entry;
2388 	unsigned int last_timestamp, namelen;
2389 	char *fullpath, *freepath;
2390 	struct kinfo_ovmentry *kve;
2391 	struct vattr va;
2392 	struct ucred *cred;
2393 	int error, *name;
2394 	struct vnode *vp;
2395 	struct proc *p;
2396 	vm_map_t map;
2397 	struct vmspace *vm;
2398 
2399 	namelen = arg2;
2400 	if (namelen != 1)
2401 		return (EINVAL);
2402 
2403 	name = (int *)arg1;
2404 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2405 	if (error != 0)
2406 		return (error);
2407 	vm = vmspace_acquire_ref(p);
2408 	if (vm == NULL) {
2409 		PRELE(p);
2410 		return (ESRCH);
2411 	}
2412 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2413 
2414 	map = &vm->vm_map;
2415 	vm_map_lock_read(map);
2416 	VM_MAP_ENTRY_FOREACH(entry, map) {
2417 		vm_object_t obj, tobj, lobj;
2418 		vm_offset_t addr;
2419 
2420 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2421 			continue;
2422 
2423 		bzero(kve, sizeof(*kve));
2424 		kve->kve_structsize = sizeof(*kve);
2425 
2426 		kve->kve_private_resident = 0;
2427 		obj = entry->object.vm_object;
2428 		if (obj != NULL) {
2429 			VM_OBJECT_RLOCK(obj);
2430 			if (obj->shadow_count == 1)
2431 				kve->kve_private_resident =
2432 				    obj->resident_page_count;
2433 		}
2434 		kve->kve_resident = 0;
2435 		addr = entry->start;
2436 		while (addr < entry->end) {
2437 			if (pmap_extract(map->pmap, addr))
2438 				kve->kve_resident++;
2439 			addr += PAGE_SIZE;
2440 		}
2441 
2442 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2443 			if (tobj != obj) {
2444 				VM_OBJECT_RLOCK(tobj);
2445 				kve->kve_offset += tobj->backing_object_offset;
2446 			}
2447 			if (lobj != obj)
2448 				VM_OBJECT_RUNLOCK(lobj);
2449 			lobj = tobj;
2450 		}
2451 
2452 		kve->kve_start = (void*)entry->start;
2453 		kve->kve_end = (void*)entry->end;
2454 		kve->kve_offset += (off_t)entry->offset;
2455 
2456 		if (entry->protection & VM_PROT_READ)
2457 			kve->kve_protection |= KVME_PROT_READ;
2458 		if (entry->protection & VM_PROT_WRITE)
2459 			kve->kve_protection |= KVME_PROT_WRITE;
2460 		if (entry->protection & VM_PROT_EXECUTE)
2461 			kve->kve_protection |= KVME_PROT_EXEC;
2462 
2463 		if (entry->eflags & MAP_ENTRY_COW)
2464 			kve->kve_flags |= KVME_FLAG_COW;
2465 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2466 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2467 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2468 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2469 
2470 		last_timestamp = map->timestamp;
2471 		vm_map_unlock_read(map);
2472 
2473 		kve->kve_fileid = 0;
2474 		kve->kve_fsid = 0;
2475 		freepath = NULL;
2476 		fullpath = "";
2477 		if (lobj) {
2478 			kve->kve_type = vm_object_kvme_type(lobj, &vp);
2479 			if (kve->kve_type == KVME_TYPE_MGTDEVICE)
2480 				kve->kve_type = KVME_TYPE_UNKNOWN;
2481 			if (vp != NULL)
2482 				vref(vp);
2483 			if (lobj != obj)
2484 				VM_OBJECT_RUNLOCK(lobj);
2485 
2486 			kve->kve_ref_count = obj->ref_count;
2487 			kve->kve_shadow_count = obj->shadow_count;
2488 			VM_OBJECT_RUNLOCK(obj);
2489 			if (vp != NULL) {
2490 				vn_fullpath(vp, &fullpath, &freepath);
2491 				cred = curthread->td_ucred;
2492 				vn_lock(vp, LK_SHARED | LK_RETRY);
2493 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2494 					kve->kve_fileid = va.va_fileid;
2495 					/* truncate */
2496 					kve->kve_fsid = va.va_fsid;
2497 				}
2498 				vput(vp);
2499 			}
2500 		} else {
2501 			kve->kve_type = KVME_TYPE_NONE;
2502 			kve->kve_ref_count = 0;
2503 			kve->kve_shadow_count = 0;
2504 		}
2505 
2506 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2507 		if (freepath != NULL)
2508 			free(freepath, M_TEMP);
2509 
2510 		error = SYSCTL_OUT(req, kve, sizeof(*kve));
2511 		vm_map_lock_read(map);
2512 		if (error)
2513 			break;
2514 		if (last_timestamp != map->timestamp) {
2515 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2516 			entry = tmp_entry;
2517 		}
2518 	}
2519 	vm_map_unlock_read(map);
2520 	vmspace_free(vm);
2521 	PRELE(p);
2522 	free(kve, M_TEMP);
2523 	return (error);
2524 }
2525 #endif	/* COMPAT_FREEBSD7 */
2526 
2527 #ifdef KINFO_VMENTRY_SIZE
2528 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2529 #endif
2530 
2531 void
2532 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry,
2533     int *resident_count, bool *super)
2534 {
2535 	vm_object_t obj, tobj;
2536 	vm_page_t m, m_adv;
2537 	vm_offset_t addr;
2538 	vm_paddr_t pa;
2539 	vm_pindex_t pi, pi_adv, pindex;
2540 	int incore;
2541 
2542 	*super = false;
2543 	*resident_count = 0;
2544 	if (vmmap_skip_res_cnt)
2545 		return;
2546 
2547 	pa = 0;
2548 	obj = entry->object.vm_object;
2549 	addr = entry->start;
2550 	m_adv = NULL;
2551 	pi = OFF_TO_IDX(entry->offset);
2552 	for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) {
2553 		if (m_adv != NULL) {
2554 			m = m_adv;
2555 		} else {
2556 			pi_adv = atop(entry->end - addr);
2557 			pindex = pi;
2558 			for (tobj = obj;; tobj = tobj->backing_object) {
2559 				m = vm_page_find_least(tobj, pindex);
2560 				if (m != NULL) {
2561 					if (m->pindex == pindex)
2562 						break;
2563 					if (pi_adv > m->pindex - pindex) {
2564 						pi_adv = m->pindex - pindex;
2565 						m_adv = m;
2566 					}
2567 				}
2568 				if (tobj->backing_object == NULL)
2569 					goto next;
2570 				pindex += OFF_TO_IDX(tobj->
2571 				    backing_object_offset);
2572 			}
2573 		}
2574 		m_adv = NULL;
2575 		if (m->psind != 0 && addr + pagesizes[1] <= entry->end &&
2576 		    (addr & (pagesizes[1] - 1)) == 0 && (incore =
2577 		    pmap_mincore(map->pmap, addr, &pa) & MINCORE_SUPER) != 0) {
2578 			*super = true;
2579 			/*
2580 			 * The virtual page might be smaller than the physical
2581 			 * page, so we use the page size reported by the pmap
2582 			 * rather than m->psind.
2583 			 */
2584 			pi_adv = atop(pagesizes[incore >> MINCORE_PSIND_SHIFT]);
2585 		} else {
2586 			/*
2587 			 * We do not test the found page on validity.
2588 			 * Either the page is busy and being paged in,
2589 			 * or it was invalidated.  The first case
2590 			 * should be counted as resident, the second
2591 			 * is not so clear; we do account both.
2592 			 */
2593 			pi_adv = 1;
2594 		}
2595 		*resident_count += pi_adv;
2596 next:;
2597 	}
2598 }
2599 
2600 /*
2601  * Must be called with the process locked and will return unlocked.
2602  */
2603 int
2604 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb, ssize_t maxlen, int flags)
2605 {
2606 	vm_map_entry_t entry, tmp_entry;
2607 	struct vattr va;
2608 	vm_map_t map;
2609 	vm_object_t lobj, nobj, obj, tobj;
2610 	char *fullpath, *freepath;
2611 	struct kinfo_vmentry *kve;
2612 	struct ucred *cred;
2613 	struct vnode *vp;
2614 	struct vmspace *vm;
2615 	vm_offset_t addr;
2616 	unsigned int last_timestamp;
2617 	int error;
2618 	bool guard, super;
2619 
2620 	PROC_LOCK_ASSERT(p, MA_OWNED);
2621 
2622 	_PHOLD(p);
2623 	PROC_UNLOCK(p);
2624 	vm = vmspace_acquire_ref(p);
2625 	if (vm == NULL) {
2626 		PRELE(p);
2627 		return (ESRCH);
2628 	}
2629 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK | M_ZERO);
2630 
2631 	error = 0;
2632 	map = &vm->vm_map;
2633 	vm_map_lock_read(map);
2634 	VM_MAP_ENTRY_FOREACH(entry, map) {
2635 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2636 			continue;
2637 
2638 		addr = entry->end;
2639 		bzero(kve, sizeof(*kve));
2640 		obj = entry->object.vm_object;
2641 		if (obj != NULL) {
2642 			if ((obj->flags & OBJ_ANON) != 0)
2643 				kve->kve_obj = (uintptr_t)obj;
2644 
2645 			for (tobj = obj; tobj != NULL;
2646 			    tobj = tobj->backing_object) {
2647 				VM_OBJECT_RLOCK(tobj);
2648 				kve->kve_offset += tobj->backing_object_offset;
2649 				lobj = tobj;
2650 			}
2651 			if (obj->backing_object == NULL)
2652 				kve->kve_private_resident =
2653 				    obj->resident_page_count;
2654 			kern_proc_vmmap_resident(map, entry,
2655 			    &kve->kve_resident, &super);
2656 			if (super)
2657 				kve->kve_flags |= KVME_FLAG_SUPER;
2658 			for (tobj = obj; tobj != NULL; tobj = nobj) {
2659 				nobj = tobj->backing_object;
2660 				if (tobj != obj && tobj != lobj)
2661 					VM_OBJECT_RUNLOCK(tobj);
2662 			}
2663 		} else {
2664 			lobj = NULL;
2665 		}
2666 
2667 		kve->kve_start = entry->start;
2668 		kve->kve_end = entry->end;
2669 		kve->kve_offset += entry->offset;
2670 
2671 		if (entry->protection & VM_PROT_READ)
2672 			kve->kve_protection |= KVME_PROT_READ;
2673 		if (entry->protection & VM_PROT_WRITE)
2674 			kve->kve_protection |= KVME_PROT_WRITE;
2675 		if (entry->protection & VM_PROT_EXECUTE)
2676 			kve->kve_protection |= KVME_PROT_EXEC;
2677 
2678 		if (entry->eflags & MAP_ENTRY_COW)
2679 			kve->kve_flags |= KVME_FLAG_COW;
2680 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2681 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2682 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2683 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2684 		if (entry->eflags & MAP_ENTRY_GROWS_UP)
2685 			kve->kve_flags |= KVME_FLAG_GROWS_UP;
2686 		if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2687 			kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2688 		if (entry->eflags & MAP_ENTRY_USER_WIRED)
2689 			kve->kve_flags |= KVME_FLAG_USER_WIRED;
2690 
2691 		guard = (entry->eflags & MAP_ENTRY_GUARD) != 0;
2692 
2693 		last_timestamp = map->timestamp;
2694 		vm_map_unlock_read(map);
2695 
2696 		freepath = NULL;
2697 		fullpath = "";
2698 		if (lobj != NULL) {
2699 			kve->kve_type = vm_object_kvme_type(lobj, &vp);
2700 			if (vp != NULL)
2701 				vref(vp);
2702 			if (lobj != obj)
2703 				VM_OBJECT_RUNLOCK(lobj);
2704 
2705 			kve->kve_ref_count = obj->ref_count;
2706 			kve->kve_shadow_count = obj->shadow_count;
2707 			VM_OBJECT_RUNLOCK(obj);
2708 			if (vp != NULL) {
2709 				vn_fullpath(vp, &fullpath, &freepath);
2710 				kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2711 				cred = curthread->td_ucred;
2712 				vn_lock(vp, LK_SHARED | LK_RETRY);
2713 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2714 					kve->kve_vn_fileid = va.va_fileid;
2715 					kve->kve_vn_fsid = va.va_fsid;
2716 					kve->kve_vn_fsid_freebsd11 =
2717 					    kve->kve_vn_fsid; /* truncate */
2718 					kve->kve_vn_mode =
2719 					    MAKEIMODE(va.va_type, va.va_mode);
2720 					kve->kve_vn_size = va.va_size;
2721 					kve->kve_vn_rdev = va.va_rdev;
2722 					kve->kve_vn_rdev_freebsd11 =
2723 					    kve->kve_vn_rdev; /* truncate */
2724 					kve->kve_status = KF_ATTR_VALID;
2725 				}
2726 				vput(vp);
2727 			}
2728 		} else {
2729 			kve->kve_type = guard ? KVME_TYPE_GUARD :
2730 			    KVME_TYPE_NONE;
2731 			kve->kve_ref_count = 0;
2732 			kve->kve_shadow_count = 0;
2733 		}
2734 
2735 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2736 		if (freepath != NULL)
2737 			free(freepath, M_TEMP);
2738 
2739 		/* Pack record size down */
2740 		if ((flags & KERN_VMMAP_PACK_KINFO) != 0)
2741 			kve->kve_structsize =
2742 			    offsetof(struct kinfo_vmentry, kve_path) +
2743 			    strlen(kve->kve_path) + 1;
2744 		else
2745 			kve->kve_structsize = sizeof(*kve);
2746 		kve->kve_structsize = roundup(kve->kve_structsize,
2747 		    sizeof(uint64_t));
2748 
2749 		/* Halt filling and truncate rather than exceeding maxlen */
2750 		if (maxlen != -1 && maxlen < kve->kve_structsize) {
2751 			error = 0;
2752 			vm_map_lock_read(map);
2753 			break;
2754 		} else if (maxlen != -1)
2755 			maxlen -= kve->kve_structsize;
2756 
2757 		if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0)
2758 			error = ENOMEM;
2759 		vm_map_lock_read(map);
2760 		if (error != 0)
2761 			break;
2762 		if (last_timestamp != map->timestamp) {
2763 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2764 			entry = tmp_entry;
2765 		}
2766 	}
2767 	vm_map_unlock_read(map);
2768 	vmspace_free(vm);
2769 	PRELE(p);
2770 	free(kve, M_TEMP);
2771 	return (error);
2772 }
2773 
2774 static int
2775 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2776 {
2777 	struct proc *p;
2778 	struct sbuf sb;
2779 	u_int namelen;
2780 	int error, error2, *name;
2781 
2782 	namelen = arg2;
2783 	if (namelen != 1)
2784 		return (EINVAL);
2785 
2786 	name = (int *)arg1;
2787 	sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req);
2788 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2789 	error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
2790 	if (error != 0) {
2791 		sbuf_delete(&sb);
2792 		return (error);
2793 	}
2794 	error = kern_proc_vmmap_out(p, &sb, -1, KERN_VMMAP_PACK_KINFO);
2795 	error2 = sbuf_finish(&sb);
2796 	sbuf_delete(&sb);
2797 	return (error != 0 ? error : error2);
2798 }
2799 
2800 #if defined(STACK) || defined(DDB)
2801 static int
2802 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2803 {
2804 	struct kinfo_kstack *kkstp;
2805 	int error, i, *name, numthreads;
2806 	lwpid_t *lwpidarray;
2807 	struct thread *td;
2808 	struct stack *st;
2809 	struct sbuf sb;
2810 	struct proc *p;
2811 	u_int namelen;
2812 
2813 	namelen = arg2;
2814 	if (namelen != 1)
2815 		return (EINVAL);
2816 
2817 	name = (int *)arg1;
2818 	error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2819 	if (error != 0)
2820 		return (error);
2821 
2822 	kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2823 	st = stack_create(M_WAITOK);
2824 
2825 	lwpidarray = NULL;
2826 	PROC_LOCK(p);
2827 	do {
2828 		if (lwpidarray != NULL) {
2829 			free(lwpidarray, M_TEMP);
2830 			lwpidarray = NULL;
2831 		}
2832 		numthreads = p->p_numthreads;
2833 		PROC_UNLOCK(p);
2834 		lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2835 		    M_WAITOK | M_ZERO);
2836 		PROC_LOCK(p);
2837 	} while (numthreads < p->p_numthreads);
2838 
2839 	/*
2840 	 * XXXRW: During the below loop, execve(2) and countless other sorts
2841 	 * of changes could have taken place.  Should we check to see if the
2842 	 * vmspace has been replaced, or the like, in order to prevent
2843 	 * giving a snapshot that spans, say, execve(2), with some threads
2844 	 * before and some after?  Among other things, the credentials could
2845 	 * have changed, in which case the right to extract debug info might
2846 	 * no longer be assured.
2847 	 */
2848 	i = 0;
2849 	FOREACH_THREAD_IN_PROC(p, td) {
2850 		KASSERT(i < numthreads,
2851 		    ("sysctl_kern_proc_kstack: numthreads"));
2852 		lwpidarray[i] = td->td_tid;
2853 		i++;
2854 	}
2855 	PROC_UNLOCK(p);
2856 	numthreads = i;
2857 	for (i = 0; i < numthreads; i++) {
2858 		td = tdfind(lwpidarray[i], p->p_pid);
2859 		if (td == NULL) {
2860 			continue;
2861 		}
2862 		bzero(kkstp, sizeof(*kkstp));
2863 		(void)sbuf_new(&sb, kkstp->kkst_trace,
2864 		    sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2865 		thread_lock(td);
2866 		kkstp->kkst_tid = td->td_tid;
2867 		if (stack_save_td(st, td) == 0)
2868 			kkstp->kkst_state = KKST_STATE_STACKOK;
2869 		else
2870 			kkstp->kkst_state = KKST_STATE_RUNNING;
2871 		thread_unlock(td);
2872 		PROC_UNLOCK(p);
2873 		stack_sbuf_print(&sb, st);
2874 		sbuf_finish(&sb);
2875 		sbuf_delete(&sb);
2876 		error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2877 		if (error)
2878 			break;
2879 	}
2880 	PRELE(p);
2881 	if (lwpidarray != NULL)
2882 		free(lwpidarray, M_TEMP);
2883 	stack_destroy(st);
2884 	free(kkstp, M_TEMP);
2885 	return (error);
2886 }
2887 #endif
2888 
2889 /*
2890  * This sysctl allows a process to retrieve the full list of groups from
2891  * itself or another process.
2892  */
2893 static int
2894 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2895 {
2896 	pid_t *pidp = (pid_t *)arg1;
2897 	unsigned int arglen = arg2;
2898 	struct proc *p;
2899 	struct ucred *cred;
2900 	int error;
2901 
2902 	if (arglen != 1)
2903 		return (EINVAL);
2904 	if (*pidp == -1) {	/* -1 means this process */
2905 		p = req->td->td_proc;
2906 		PROC_LOCK(p);
2907 	} else {
2908 		error = pget(*pidp, PGET_CANSEE, &p);
2909 		if (error != 0)
2910 			return (error);
2911 	}
2912 
2913 	cred = crhold(p->p_ucred);
2914 	PROC_UNLOCK(p);
2915 
2916 	error = SYSCTL_OUT(req, cred->cr_groups,
2917 	    cred->cr_ngroups * sizeof(gid_t));
2918 	crfree(cred);
2919 	return (error);
2920 }
2921 
2922 /*
2923  * This sysctl allows a process to retrieve or/and set the resource limit for
2924  * another process.
2925  */
2926 static int
2927 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2928 {
2929 	int *name = (int *)arg1;
2930 	u_int namelen = arg2;
2931 	struct rlimit rlim;
2932 	struct proc *p;
2933 	u_int which;
2934 	int flags, error;
2935 
2936 	if (namelen != 2)
2937 		return (EINVAL);
2938 
2939 	which = (u_int)name[1];
2940 	if (which >= RLIM_NLIMITS)
2941 		return (EINVAL);
2942 
2943 	if (req->newptr != NULL && req->newlen != sizeof(rlim))
2944 		return (EINVAL);
2945 
2946 	flags = PGET_HOLD | PGET_NOTWEXIT;
2947 	if (req->newptr != NULL)
2948 		flags |= PGET_CANDEBUG;
2949 	else
2950 		flags |= PGET_CANSEE;
2951 	error = pget((pid_t)name[0], flags, &p);
2952 	if (error != 0)
2953 		return (error);
2954 
2955 	/*
2956 	 * Retrieve limit.
2957 	 */
2958 	if (req->oldptr != NULL) {
2959 		PROC_LOCK(p);
2960 		lim_rlimit_proc(p, which, &rlim);
2961 		PROC_UNLOCK(p);
2962 	}
2963 	error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
2964 	if (error != 0)
2965 		goto errout;
2966 
2967 	/*
2968 	 * Set limit.
2969 	 */
2970 	if (req->newptr != NULL) {
2971 		error = SYSCTL_IN(req, &rlim, sizeof(rlim));
2972 		if (error == 0)
2973 			error = kern_proc_setrlimit(curthread, p, which, &rlim);
2974 	}
2975 
2976 errout:
2977 	PRELE(p);
2978 	return (error);
2979 }
2980 
2981 /*
2982  * This sysctl allows a process to retrieve ps_strings structure location of
2983  * another process.
2984  */
2985 static int
2986 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
2987 {
2988 	int *name = (int *)arg1;
2989 	u_int namelen = arg2;
2990 	struct proc *p;
2991 	vm_offset_t ps_strings;
2992 	int error;
2993 #ifdef COMPAT_FREEBSD32
2994 	uint32_t ps_strings32;
2995 #endif
2996 
2997 	if (namelen != 1)
2998 		return (EINVAL);
2999 
3000 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
3001 	if (error != 0)
3002 		return (error);
3003 #ifdef COMPAT_FREEBSD32
3004 	if ((req->flags & SCTL_MASK32) != 0) {
3005 		/*
3006 		 * We return 0 if the 32 bit emulation request is for a 64 bit
3007 		 * process.
3008 		 */
3009 		ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
3010 		    PTROUT(PROC_PS_STRINGS(p)) : 0;
3011 		PROC_UNLOCK(p);
3012 		error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
3013 		return (error);
3014 	}
3015 #endif
3016 	ps_strings = PROC_PS_STRINGS(p);
3017 	PROC_UNLOCK(p);
3018 	error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
3019 	return (error);
3020 }
3021 
3022 /*
3023  * This sysctl allows a process to retrieve umask of another process.
3024  */
3025 static int
3026 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
3027 {
3028 	int *name = (int *)arg1;
3029 	u_int namelen = arg2;
3030 	struct proc *p;
3031 	int error;
3032 	u_short cmask;
3033 	pid_t pid;
3034 
3035 	if (namelen != 1)
3036 		return (EINVAL);
3037 
3038 	pid = (pid_t)name[0];
3039 	p = curproc;
3040 	if (pid == p->p_pid || pid == 0) {
3041 		cmask = p->p_pd->pd_cmask;
3042 		goto out;
3043 	}
3044 
3045 	error = pget(pid, PGET_WANTREAD, &p);
3046 	if (error != 0)
3047 		return (error);
3048 
3049 	cmask = p->p_pd->pd_cmask;
3050 	PRELE(p);
3051 out:
3052 	error = SYSCTL_OUT(req, &cmask, sizeof(cmask));
3053 	return (error);
3054 }
3055 
3056 /*
3057  * This sysctl allows a process to set and retrieve binary osreldate of
3058  * another process.
3059  */
3060 static int
3061 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
3062 {
3063 	int *name = (int *)arg1;
3064 	u_int namelen = arg2;
3065 	struct proc *p;
3066 	int flags, error, osrel;
3067 
3068 	if (namelen != 1)
3069 		return (EINVAL);
3070 
3071 	if (req->newptr != NULL && req->newlen != sizeof(osrel))
3072 		return (EINVAL);
3073 
3074 	flags = PGET_HOLD | PGET_NOTWEXIT;
3075 	if (req->newptr != NULL)
3076 		flags |= PGET_CANDEBUG;
3077 	else
3078 		flags |= PGET_CANSEE;
3079 	error = pget((pid_t)name[0], flags, &p);
3080 	if (error != 0)
3081 		return (error);
3082 
3083 	error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
3084 	if (error != 0)
3085 		goto errout;
3086 
3087 	if (req->newptr != NULL) {
3088 		error = SYSCTL_IN(req, &osrel, sizeof(osrel));
3089 		if (error != 0)
3090 			goto errout;
3091 		if (osrel < 0) {
3092 			error = EINVAL;
3093 			goto errout;
3094 		}
3095 		p->p_osrel = osrel;
3096 	}
3097 errout:
3098 	PRELE(p);
3099 	return (error);
3100 }
3101 
3102 static int
3103 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS)
3104 {
3105 	int *name = (int *)arg1;
3106 	u_int namelen = arg2;
3107 	struct proc *p;
3108 	struct kinfo_sigtramp kst;
3109 	const struct sysentvec *sv;
3110 	int error;
3111 #ifdef COMPAT_FREEBSD32
3112 	struct kinfo_sigtramp32 kst32;
3113 #endif
3114 
3115 	if (namelen != 1)
3116 		return (EINVAL);
3117 
3118 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
3119 	if (error != 0)
3120 		return (error);
3121 	sv = p->p_sysent;
3122 #ifdef COMPAT_FREEBSD32
3123 	if ((req->flags & SCTL_MASK32) != 0) {
3124 		bzero(&kst32, sizeof(kst32));
3125 		if (SV_PROC_FLAG(p, SV_ILP32)) {
3126 			if (PROC_HAS_SHP(p)) {
3127 				kst32.ksigtramp_start = PROC_SIGCODE(p);
3128 				kst32.ksigtramp_end = kst32.ksigtramp_start +
3129 				    ((sv->sv_flags & SV_DSO_SIG) == 0 ?
3130 				    *sv->sv_szsigcode :
3131 				    (uintptr_t)sv->sv_szsigcode);
3132 			} else {
3133 				kst32.ksigtramp_start = PROC_PS_STRINGS(p) -
3134 				    *sv->sv_szsigcode;
3135 				kst32.ksigtramp_end = PROC_PS_STRINGS(p);
3136 			}
3137 		}
3138 		PROC_UNLOCK(p);
3139 		error = SYSCTL_OUT(req, &kst32, sizeof(kst32));
3140 		return (error);
3141 	}
3142 #endif
3143 	bzero(&kst, sizeof(kst));
3144 	if (PROC_HAS_SHP(p)) {
3145 		kst.ksigtramp_start = (char *)PROC_SIGCODE(p);
3146 		kst.ksigtramp_end = (char *)kst.ksigtramp_start +
3147 		    ((sv->sv_flags & SV_DSO_SIG) == 0 ? *sv->sv_szsigcode :
3148 		    (uintptr_t)sv->sv_szsigcode);
3149 	} else {
3150 		kst.ksigtramp_start = (char *)PROC_PS_STRINGS(p) -
3151 		    *sv->sv_szsigcode;
3152 		kst.ksigtramp_end = (char *)PROC_PS_STRINGS(p);
3153 	}
3154 	PROC_UNLOCK(p);
3155 	error = SYSCTL_OUT(req, &kst, sizeof(kst));
3156 	return (error);
3157 }
3158 
3159 static int
3160 sysctl_kern_proc_sigfastblk(SYSCTL_HANDLER_ARGS)
3161 {
3162 	int *name = (int *)arg1;
3163 	u_int namelen = arg2;
3164 	pid_t pid;
3165 	struct proc *p;
3166 	struct thread *td1;
3167 	uintptr_t addr;
3168 #ifdef COMPAT_FREEBSD32
3169 	uint32_t addr32;
3170 #endif
3171 	int error;
3172 
3173 	if (namelen != 1 || req->newptr != NULL)
3174 		return (EINVAL);
3175 
3176 	pid = (pid_t)name[0];
3177 	error = pget(pid, PGET_HOLD | PGET_NOTWEXIT | PGET_CANDEBUG, &p);
3178 	if (error != 0)
3179 		return (error);
3180 
3181 	PROC_LOCK(p);
3182 #ifdef COMPAT_FREEBSD32
3183 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3184 		if (!SV_PROC_FLAG(p, SV_ILP32)) {
3185 			error = EINVAL;
3186 			goto errlocked;
3187 		}
3188 	}
3189 #endif
3190 	if (pid <= PID_MAX) {
3191 		td1 = FIRST_THREAD_IN_PROC(p);
3192 	} else {
3193 		FOREACH_THREAD_IN_PROC(p, td1) {
3194 			if (td1->td_tid == pid)
3195 				break;
3196 		}
3197 	}
3198 	if (td1 == NULL) {
3199 		error = ESRCH;
3200 		goto errlocked;
3201 	}
3202 	/*
3203 	 * The access to the private thread flags.  It is fine as far
3204 	 * as no out-of-thin-air values are read from td_pflags, and
3205 	 * usermode read of the td_sigblock_ptr is racy inherently,
3206 	 * since target process might have already changed it
3207 	 * meantime.
3208 	 */
3209 	if ((td1->td_pflags & TDP_SIGFASTBLOCK) != 0)
3210 		addr = (uintptr_t)td1->td_sigblock_ptr;
3211 	else
3212 		error = ENOTTY;
3213 
3214 errlocked:
3215 	_PRELE(p);
3216 	PROC_UNLOCK(p);
3217 	if (error != 0)
3218 		return (error);
3219 
3220 #ifdef COMPAT_FREEBSD32
3221 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3222 		addr32 = addr;
3223 		error = SYSCTL_OUT(req, &addr32, sizeof(addr32));
3224 	} else
3225 #endif
3226 		error = SYSCTL_OUT(req, &addr, sizeof(addr));
3227 	return (error);
3228 }
3229 
3230 static int
3231 sysctl_kern_proc_vm_layout(SYSCTL_HANDLER_ARGS)
3232 {
3233 	struct kinfo_vm_layout kvm;
3234 	struct proc *p;
3235 	struct vmspace *vmspace;
3236 	int error, *name;
3237 
3238 	name = (int *)arg1;
3239 	if ((u_int)arg2 != 1)
3240 		return (EINVAL);
3241 
3242 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
3243 	if (error != 0)
3244 		return (error);
3245 #ifdef COMPAT_FREEBSD32
3246 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3247 		if (!SV_PROC_FLAG(p, SV_ILP32)) {
3248 			PROC_UNLOCK(p);
3249 			return (EINVAL);
3250 		}
3251 	}
3252 #endif
3253 	vmspace = vmspace_acquire_ref(p);
3254 	PROC_UNLOCK(p);
3255 
3256 	memset(&kvm, 0, sizeof(kvm));
3257 	kvm.kvm_min_user_addr = vm_map_min(&vmspace->vm_map);
3258 	kvm.kvm_max_user_addr = vm_map_max(&vmspace->vm_map);
3259 	kvm.kvm_text_addr = (uintptr_t)vmspace->vm_taddr;
3260 	kvm.kvm_text_size = vmspace->vm_tsize;
3261 	kvm.kvm_data_addr = (uintptr_t)vmspace->vm_daddr;
3262 	kvm.kvm_data_size = vmspace->vm_dsize;
3263 	kvm.kvm_stack_addr = (uintptr_t)vmspace->vm_maxsaddr;
3264 	kvm.kvm_stack_size = vmspace->vm_ssize;
3265 	kvm.kvm_shp_addr = vmspace->vm_shp_base;
3266 	kvm.kvm_shp_size = p->p_sysent->sv_shared_page_len;
3267 	if ((vmspace->vm_map.flags & MAP_WIREFUTURE) != 0)
3268 		kvm.kvm_map_flags |= KMAP_FLAG_WIREFUTURE;
3269 	if ((vmspace->vm_map.flags & MAP_ASLR) != 0)
3270 		kvm.kvm_map_flags |= KMAP_FLAG_ASLR;
3271 	if ((vmspace->vm_map.flags & MAP_ASLR_IGNSTART) != 0)
3272 		kvm.kvm_map_flags |= KMAP_FLAG_ASLR_IGNSTART;
3273 	if ((vmspace->vm_map.flags & MAP_WXORX) != 0)
3274 		kvm.kvm_map_flags |= KMAP_FLAG_WXORX;
3275 	if ((vmspace->vm_map.flags & MAP_ASLR_STACK) != 0)
3276 		kvm.kvm_map_flags |= KMAP_FLAG_ASLR_STACK;
3277 	if (vmspace->vm_shp_base != p->p_sysent->sv_shared_page_base &&
3278 	    PROC_HAS_SHP(p))
3279 		kvm.kvm_map_flags |= KMAP_FLAG_ASLR_SHARED_PAGE;
3280 
3281 #ifdef COMPAT_FREEBSD32
3282 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3283 		struct kinfo_vm_layout32 kvm32;
3284 
3285 		memset(&kvm32, 0, sizeof(kvm32));
3286 		kvm32.kvm_min_user_addr = (uint32_t)kvm.kvm_min_user_addr;
3287 		kvm32.kvm_max_user_addr = (uint32_t)kvm.kvm_max_user_addr;
3288 		kvm32.kvm_text_addr = (uint32_t)kvm.kvm_text_addr;
3289 		kvm32.kvm_text_size = (uint32_t)kvm.kvm_text_size;
3290 		kvm32.kvm_data_addr = (uint32_t)kvm.kvm_data_addr;
3291 		kvm32.kvm_data_size = (uint32_t)kvm.kvm_data_size;
3292 		kvm32.kvm_stack_addr = (uint32_t)kvm.kvm_stack_addr;
3293 		kvm32.kvm_stack_size = (uint32_t)kvm.kvm_stack_size;
3294 		kvm32.kvm_shp_addr = (uint32_t)kvm.kvm_shp_addr;
3295 		kvm32.kvm_shp_size = (uint32_t)kvm.kvm_shp_size;
3296 		kvm32.kvm_map_flags = kvm.kvm_map_flags;
3297 		error = SYSCTL_OUT(req, &kvm32, sizeof(kvm32));
3298 		goto out;
3299 	}
3300 #endif
3301 
3302 	error = SYSCTL_OUT(req, &kvm, sizeof(kvm));
3303 #ifdef COMPAT_FREEBSD32
3304 out:
3305 #endif
3306 	vmspace_free(vmspace);
3307 	return (error);
3308 }
3309 
3310 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,  0,
3311     "Process table");
3312 
3313 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
3314 	CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
3315 	"Return entire process table");
3316 
3317 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3318 	sysctl_kern_proc, "Process table");
3319 
3320 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
3321 	sysctl_kern_proc, "Process table");
3322 
3323 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3324 	sysctl_kern_proc, "Process table");
3325 
3326 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
3327 	CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3328 
3329 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
3330 	sysctl_kern_proc, "Process table");
3331 
3332 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3333 	sysctl_kern_proc, "Process table");
3334 
3335 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3336 	sysctl_kern_proc, "Process table");
3337 
3338 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3339 	sysctl_kern_proc, "Process table");
3340 
3341 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
3342 	sysctl_kern_proc, "Return process table, no threads");
3343 
3344 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
3345 	CTLFLAG_RW | CTLFLAG_CAPWR | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
3346 	sysctl_kern_proc_args, "Process argument list");
3347 
3348 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
3349 	sysctl_kern_proc_env, "Process environment");
3350 
3351 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
3352 	CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
3353 
3354 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
3355 	CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
3356 
3357 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
3358 	CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
3359 	"Process syscall vector name (ABI type)");
3360 
3361 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
3362 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3363 
3364 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
3365 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3366 
3367 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
3368 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3369 
3370 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
3371 	sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3372 
3373 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
3374 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3375 
3376 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
3377 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3378 
3379 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
3380 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3381 
3382 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
3383 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3384 
3385 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
3386 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
3387 	"Return process table, including threads");
3388 
3389 #ifdef COMPAT_FREEBSD7
3390 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
3391 	CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
3392 #endif
3393 
3394 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
3395 	CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
3396 
3397 #if defined(STACK) || defined(DDB)
3398 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
3399 	CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
3400 #endif
3401 
3402 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
3403 	CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
3404 
3405 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
3406 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
3407 	"Process resource limits");
3408 
3409 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
3410 	CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
3411 	"Process ps_strings location");
3412 
3413 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
3414 	CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
3415 
3416 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
3417 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
3418 	"Process binary osreldate");
3419 
3420 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD |
3421 	CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp,
3422 	"Process signal trampoline location");
3423 
3424 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGFASTBLK, sigfastblk, CTLFLAG_RD |
3425 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_sigfastblk,
3426 	"Thread sigfastblock address");
3427 
3428 static SYSCTL_NODE(_kern_proc, KERN_PROC_VM_LAYOUT, vm_layout, CTLFLAG_RD |
3429 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_vm_layout,
3430 	"Process virtual address space layout info");
3431 
3432 static struct sx stop_all_proc_blocker;
3433 SX_SYSINIT(stop_all_proc_blocker, &stop_all_proc_blocker, "sapblk");
3434 
3435 bool
3436 stop_all_proc_block(void)
3437 {
3438 	return (sx_xlock_sig(&stop_all_proc_blocker) == 0);
3439 }
3440 
3441 void
3442 stop_all_proc_unblock(void)
3443 {
3444 	sx_xunlock(&stop_all_proc_blocker);
3445 }
3446 
3447 int allproc_gen;
3448 
3449 /*
3450  * stop_all_proc() purpose is to stop all process which have usermode,
3451  * except current process for obvious reasons.  This makes it somewhat
3452  * unreliable when invoked from multithreaded process.  The service
3453  * must not be user-callable anyway.
3454  */
3455 void
3456 stop_all_proc(void)
3457 {
3458 	struct proc *cp, *p;
3459 	int r, gen;
3460 	bool restart, seen_stopped, seen_exiting, stopped_some;
3461 
3462 	if (!stop_all_proc_block())
3463 		return;
3464 
3465 	cp = curproc;
3466 allproc_loop:
3467 	sx_xlock(&allproc_lock);
3468 	gen = allproc_gen;
3469 	seen_exiting = seen_stopped = stopped_some = restart = false;
3470 	LIST_REMOVE(cp, p_list);
3471 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3472 	for (;;) {
3473 		p = LIST_NEXT(cp, p_list);
3474 		if (p == NULL)
3475 			break;
3476 		LIST_REMOVE(cp, p_list);
3477 		LIST_INSERT_AFTER(p, cp, p_list);
3478 		PROC_LOCK(p);
3479 		if ((p->p_flag & (P_KPROC | P_SYSTEM | P_TOTAL_STOP |
3480 		    P_STOPPED_SIG)) != 0) {
3481 			PROC_UNLOCK(p);
3482 			continue;
3483 		}
3484 		if ((p->p_flag2 & P2_WEXIT) != 0) {
3485 			seen_exiting = true;
3486 			PROC_UNLOCK(p);
3487 			continue;
3488 		}
3489 		if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) {
3490 			/*
3491 			 * Stopped processes are tolerated when there
3492 			 * are no other processes which might continue
3493 			 * them.  P_STOPPED_SINGLE but not
3494 			 * P_TOTAL_STOP process still has at least one
3495 			 * thread running.
3496 			 */
3497 			seen_stopped = true;
3498 			PROC_UNLOCK(p);
3499 			continue;
3500 		}
3501 		if ((p->p_flag & P_TRACED) != 0) {
3502 			/*
3503 			 * thread_single() below cannot stop traced p,
3504 			 * so skip it.  OTOH, we cannot require
3505 			 * restart because debugger might be either
3506 			 * already stopped or traced as well.
3507 			 */
3508 			PROC_UNLOCK(p);
3509 			continue;
3510 		}
3511 		sx_xunlock(&allproc_lock);
3512 		_PHOLD(p);
3513 		r = thread_single(p, SINGLE_ALLPROC);
3514 		if (r != 0)
3515 			restart = true;
3516 		else
3517 			stopped_some = true;
3518 		_PRELE(p);
3519 		PROC_UNLOCK(p);
3520 		sx_xlock(&allproc_lock);
3521 	}
3522 	/* Catch forked children we did not see in iteration. */
3523 	if (gen != allproc_gen)
3524 		restart = true;
3525 	sx_xunlock(&allproc_lock);
3526 	if (restart || stopped_some || seen_exiting || seen_stopped) {
3527 		kern_yield(PRI_USER);
3528 		goto allproc_loop;
3529 	}
3530 }
3531 
3532 void
3533 resume_all_proc(void)
3534 {
3535 	struct proc *cp, *p;
3536 
3537 	cp = curproc;
3538 	sx_xlock(&allproc_lock);
3539 again:
3540 	LIST_REMOVE(cp, p_list);
3541 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3542 	for (;;) {
3543 		p = LIST_NEXT(cp, p_list);
3544 		if (p == NULL)
3545 			break;
3546 		LIST_REMOVE(cp, p_list);
3547 		LIST_INSERT_AFTER(p, cp, p_list);
3548 		PROC_LOCK(p);
3549 		if ((p->p_flag & P_TOTAL_STOP) != 0) {
3550 			sx_xunlock(&allproc_lock);
3551 			_PHOLD(p);
3552 			thread_single_end(p, SINGLE_ALLPROC);
3553 			_PRELE(p);
3554 			PROC_UNLOCK(p);
3555 			sx_xlock(&allproc_lock);
3556 		} else {
3557 			PROC_UNLOCK(p);
3558 		}
3559 	}
3560 	/*  Did the loop above missed any stopped process ? */
3561 	FOREACH_PROC_IN_SYSTEM(p) {
3562 		/* No need for proc lock. */
3563 		if ((p->p_flag & P_TOTAL_STOP) != 0)
3564 			goto again;
3565 	}
3566 	sx_xunlock(&allproc_lock);
3567 
3568 	stop_all_proc_unblock();
3569 }
3570 
3571 /* #define	TOTAL_STOP_DEBUG	1 */
3572 #ifdef TOTAL_STOP_DEBUG
3573 volatile static int ap_resume;
3574 #include <sys/mount.h>
3575 
3576 static int
3577 sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS)
3578 {
3579 	int error, val;
3580 
3581 	val = 0;
3582 	ap_resume = 0;
3583 	error = sysctl_handle_int(oidp, &val, 0, req);
3584 	if (error != 0 || req->newptr == NULL)
3585 		return (error);
3586 	if (val != 0) {
3587 		stop_all_proc();
3588 		syncer_suspend();
3589 		while (ap_resume == 0)
3590 			;
3591 		syncer_resume();
3592 		resume_all_proc();
3593 	}
3594 	return (0);
3595 }
3596 
3597 SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW |
3598     CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0,
3599     sysctl_debug_stop_all_proc, "I",
3600     "");
3601 #endif
3602