xref: /freebsd/sys/kern/kern_proc.c (revision fa38579f317d5c2ff2926fab9b12ee6d429bd155)
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/conf.h>
42 #include <sys/elf.h>
43 #include <sys/eventhandler.h>
44 #include <sys/exec.h>
45 #include <sys/fcntl.h>
46 #include <sys/ipc.h>
47 #include <sys/jail.h>
48 #include <sys/kernel.h>
49 #include <sys/limits.h>
50 #include <sys/lock.h>
51 #include <sys/loginclass.h>
52 #include <sys/malloc.h>
53 #include <sys/mman.h>
54 #include <sys/mount.h>
55 #include <sys/mutex.h>
56 #include <sys/namei.h>
57 #include <sys/proc.h>
58 #include <sys/ptrace.h>
59 #include <sys/refcount.h>
60 #include <sys/resourcevar.h>
61 #include <sys/rwlock.h>
62 #include <sys/sbuf.h>
63 #include <sys/sysent.h>
64 #include <sys/sched.h>
65 #include <sys/shm.h>
66 #include <sys/smp.h>
67 #include <sys/stack.h>
68 #include <sys/stat.h>
69 #include <sys/dtrace_bsd.h>
70 #include <sys/sysctl.h>
71 #include <sys/filedesc.h>
72 #include <sys/tty.h>
73 #include <sys/signalvar.h>
74 #include <sys/sdt.h>
75 #include <sys/sx.h>
76 #include <sys/user.h>
77 #include <sys/vnode.h>
78 #include <sys/wait.h>
79 #ifdef KTRACE
80 #include <sys/ktrace.h>
81 #endif
82 
83 #ifdef DDB
84 #include <ddb/ddb.h>
85 #endif
86 
87 #include <vm/vm.h>
88 #include <vm/vm_param.h>
89 #include <vm/vm_extern.h>
90 #include <vm/pmap.h>
91 #include <vm/vm_map.h>
92 #include <vm/vm_object.h>
93 #include <vm/vm_page.h>
94 #include <vm/vm_pager.h>
95 #include <vm/vm_radix.h>
96 #include <vm/uma.h>
97 
98 #include <fs/devfs/devfs.h>
99 
100 #ifdef COMPAT_FREEBSD32
101 #include <compat/freebsd32/freebsd32.h>
102 #include <compat/freebsd32/freebsd32_util.h>
103 #endif
104 
105 SDT_PROVIDER_DEFINE(proc);
106 
107 MALLOC_DEFINE(M_SESSION, "session", "session header");
108 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
109 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
110 
111 static void doenterpgrp(struct proc *, struct pgrp *);
112 static void orphanpg(struct pgrp *pg);
113 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp);
114 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp);
115 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp,
116     int preferthread);
117 static void pgdelete(struct pgrp *);
118 static int pgrp_init(void *mem, int size, int flags);
119 static int proc_ctor(void *mem, int size, void *arg, int flags);
120 static void proc_dtor(void *mem, int size, void *arg);
121 static int proc_init(void *mem, int size, int flags);
122 static void proc_fini(void *mem, int size);
123 static void pargs_free(struct pargs *pa);
124 
125 /*
126  * Other process lists
127  */
128 struct pidhashhead *pidhashtbl = NULL;
129 struct sx *pidhashtbl_lock;
130 u_long pidhash;
131 u_long pidhashlock;
132 struct pgrphashhead *pgrphashtbl;
133 u_long pgrphash;
134 struct proclist allproc = LIST_HEAD_INITIALIZER(allproc);
135 struct sx __exclusive_cache_line allproc_lock;
136 struct sx __exclusive_cache_line proctree_lock;
137 struct mtx __exclusive_cache_line ppeers_lock;
138 struct mtx __exclusive_cache_line procid_lock;
139 uma_zone_t proc_zone;
140 uma_zone_t pgrp_zone;
141 
142 /*
143  * The offset of various fields in struct proc and struct thread.
144  * These are used by kernel debuggers to enumerate kernel threads and
145  * processes.
146  */
147 const int proc_off_p_pid = offsetof(struct proc, p_pid);
148 const int proc_off_p_comm = offsetof(struct proc, p_comm);
149 const int proc_off_p_list = offsetof(struct proc, p_list);
150 const int proc_off_p_hash = offsetof(struct proc, p_hash);
151 const int proc_off_p_threads = offsetof(struct proc, p_threads);
152 const int thread_off_td_tid = offsetof(struct thread, td_tid);
153 const int thread_off_td_name = offsetof(struct thread, td_name);
154 const int thread_off_td_oncpu = offsetof(struct thread, td_oncpu);
155 const int thread_off_td_pcb = offsetof(struct thread, td_pcb);
156 const int thread_off_td_plist = offsetof(struct thread, td_plist);
157 
158 EVENTHANDLER_LIST_DEFINE(process_ctor);
159 EVENTHANDLER_LIST_DEFINE(process_dtor);
160 EVENTHANDLER_LIST_DEFINE(process_init);
161 EVENTHANDLER_LIST_DEFINE(process_fini);
162 EVENTHANDLER_LIST_DEFINE(process_exit);
163 EVENTHANDLER_LIST_DEFINE(process_fork);
164 EVENTHANDLER_LIST_DEFINE(process_exec);
165 
166 int kstack_pages = KSTACK_PAGES;
167 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RDTUN | CTLFLAG_NOFETCH,
168     &kstack_pages, 0,
169     "Kernel stack size in pages");
170 static int vmmap_skip_res_cnt = 0;
171 SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW,
172     &vmmap_skip_res_cnt, 0,
173     "Skip calculation of the pages resident count in kern.proc.vmmap");
174 
175 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
176 #ifdef COMPAT_FREEBSD32
177 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
178 #endif
179 
180 /*
181  * Initialize global process hashing structures.
182  */
183 void
184 procinit(void)
185 {
186 	u_long i;
187 
188 	sx_init(&allproc_lock, "allproc");
189 	sx_init(&proctree_lock, "proctree");
190 	mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
191 	mtx_init(&procid_lock, "procid", NULL, MTX_DEF);
192 	pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
193 	pidhashlock = (pidhash + 1) / 64;
194 	if (pidhashlock > 0)
195 		pidhashlock--;
196 	pidhashtbl_lock = malloc(sizeof(*pidhashtbl_lock) * (pidhashlock + 1),
197 	    M_PROC, M_WAITOK | M_ZERO);
198 	for (i = 0; i < pidhashlock + 1; i++)
199 		sx_init_flags(&pidhashtbl_lock[i], "pidhash", SX_DUPOK);
200 	pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
201 	proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
202 	    proc_ctor, proc_dtor, proc_init, proc_fini,
203 	    UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
204 	pgrp_zone = uma_zcreate("PGRP", sizeof(struct pgrp), NULL, NULL,
205 	    pgrp_init, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
206 	uihashinit();
207 }
208 
209 /*
210  * Prepare a proc for use.
211  */
212 static int
213 proc_ctor(void *mem, int size, void *arg, int flags)
214 {
215 	struct proc *p;
216 	struct thread *td;
217 
218 	p = (struct proc *)mem;
219 #ifdef KDTRACE_HOOKS
220 	kdtrace_proc_ctor(p);
221 #endif
222 	EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
223 	td = FIRST_THREAD_IN_PROC(p);
224 	if (td != NULL) {
225 		/* Make sure all thread constructors are executed */
226 		EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
227 	}
228 	return (0);
229 }
230 
231 /*
232  * Reclaim a proc after use.
233  */
234 static void
235 proc_dtor(void *mem, int size, void *arg)
236 {
237 	struct proc *p;
238 	struct thread *td;
239 
240 	/* INVARIANTS checks go here */
241 	p = (struct proc *)mem;
242 	td = FIRST_THREAD_IN_PROC(p);
243 	if (td != NULL) {
244 #ifdef INVARIANTS
245 		KASSERT((p->p_numthreads == 1),
246 		    ("bad number of threads in exiting process"));
247 		KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
248 #endif
249 		/* Free all OSD associated to this thread. */
250 		osd_thread_exit(td);
251 		ast_kclear(td);
252 
253 		/* Make sure all thread destructors are executed */
254 		EVENTHANDLER_DIRECT_INVOKE(thread_dtor, td);
255 	}
256 	EVENTHANDLER_DIRECT_INVOKE(process_dtor, p);
257 #ifdef KDTRACE_HOOKS
258 	kdtrace_proc_dtor(p);
259 #endif
260 	if (p->p_ksi != NULL)
261 		KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
262 }
263 
264 /*
265  * Initialize type-stable parts of a proc (when newly created).
266  */
267 static int
268 proc_init(void *mem, int size, int flags)
269 {
270 	struct proc *p;
271 
272 	p = (struct proc *)mem;
273 	mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK | MTX_NEW);
274 	mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_NEW);
275 	mtx_init(&p->p_statmtx, "pstatl", NULL, MTX_SPIN | MTX_NEW);
276 	mtx_init(&p->p_itimmtx, "pitiml", NULL, MTX_SPIN | MTX_NEW);
277 	mtx_init(&p->p_profmtx, "pprofl", NULL, MTX_SPIN | MTX_NEW);
278 	cv_init(&p->p_pwait, "ppwait");
279 	TAILQ_INIT(&p->p_threads);	     /* all threads in proc */
280 	EVENTHANDLER_DIRECT_INVOKE(process_init, p);
281 	p->p_stats = pstats_alloc();
282 	p->p_pgrp = NULL;
283 	TAILQ_INIT(&p->p_kqtim_stop);
284 	return (0);
285 }
286 
287 /*
288  * UMA should ensure that this function is never called.
289  * Freeing a proc structure would violate type stability.
290  */
291 static void
292 proc_fini(void *mem, int size)
293 {
294 #ifdef notnow
295 	struct proc *p;
296 
297 	p = (struct proc *)mem;
298 	EVENTHANDLER_DIRECT_INVOKE(process_fini, p);
299 	pstats_free(p->p_stats);
300 	thread_free(FIRST_THREAD_IN_PROC(p));
301 	mtx_destroy(&p->p_mtx);
302 	if (p->p_ksi != NULL)
303 		ksiginfo_free(p->p_ksi);
304 #else
305 	panic("proc reclaimed");
306 #endif
307 }
308 
309 static int
310 pgrp_init(void *mem, int size, int flags)
311 {
312 	struct pgrp *pg;
313 
314 	pg = mem;
315 	mtx_init(&pg->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
316 	sx_init(&pg->pg_killsx, "killpg racer");
317 	return (0);
318 }
319 
320 /*
321  * PID space management.
322  *
323  * These bitmaps are used by fork_findpid.
324  */
325 bitstr_t bit_decl(proc_id_pidmap, PID_MAX);
326 bitstr_t bit_decl(proc_id_grpidmap, PID_MAX);
327 bitstr_t bit_decl(proc_id_sessidmap, PID_MAX);
328 bitstr_t bit_decl(proc_id_reapmap, PID_MAX);
329 
330 static bitstr_t *proc_id_array[] = {
331 	proc_id_pidmap,
332 	proc_id_grpidmap,
333 	proc_id_sessidmap,
334 	proc_id_reapmap,
335 };
336 
337 void
338 proc_id_set(int type, pid_t id)
339 {
340 
341 	KASSERT(type >= 0 && type < nitems(proc_id_array),
342 	    ("invalid type %d\n", type));
343 	mtx_lock(&procid_lock);
344 	KASSERT(bit_test(proc_id_array[type], id) == 0,
345 	    ("bit %d already set in %d\n", id, type));
346 	bit_set(proc_id_array[type], id);
347 	mtx_unlock(&procid_lock);
348 }
349 
350 void
351 proc_id_set_cond(int type, pid_t id)
352 {
353 
354 	KASSERT(type >= 0 && type < nitems(proc_id_array),
355 	    ("invalid type %d\n", type));
356 	if (bit_test(proc_id_array[type], id))
357 		return;
358 	mtx_lock(&procid_lock);
359 	bit_set(proc_id_array[type], id);
360 	mtx_unlock(&procid_lock);
361 }
362 
363 void
364 proc_id_clear(int type, pid_t id)
365 {
366 
367 	KASSERT(type >= 0 && type < nitems(proc_id_array),
368 	    ("invalid type %d\n", type));
369 	mtx_lock(&procid_lock);
370 	KASSERT(bit_test(proc_id_array[type], id) != 0,
371 	    ("bit %d not set in %d\n", id, type));
372 	bit_clear(proc_id_array[type], id);
373 	mtx_unlock(&procid_lock);
374 }
375 
376 /*
377  * Is p an inferior of the current process?
378  */
379 int
380 inferior(struct proc *p)
381 {
382 
383 	sx_assert(&proctree_lock, SX_LOCKED);
384 	PROC_LOCK_ASSERT(p, MA_OWNED);
385 	for (; p != curproc; p = proc_realparent(p)) {
386 		if (p->p_pid == 0)
387 			return (0);
388 	}
389 	return (1);
390 }
391 
392 /*
393  * Shared lock all the pid hash lists.
394  */
395 void
396 pidhash_slockall(void)
397 {
398 	u_long i;
399 
400 	for (i = 0; i < pidhashlock + 1; i++)
401 		sx_slock(&pidhashtbl_lock[i]);
402 }
403 
404 /*
405  * Shared unlock all the pid hash lists.
406  */
407 void
408 pidhash_sunlockall(void)
409 {
410 	u_long i;
411 
412 	for (i = 0; i < pidhashlock + 1; i++)
413 		sx_sunlock(&pidhashtbl_lock[i]);
414 }
415 
416 /*
417  * Similar to pfind(), this function locate a process by number.
418  */
419 struct proc *
420 pfind_any_locked(pid_t pid)
421 {
422 	struct proc *p;
423 
424 	sx_assert(PIDHASHLOCK(pid), SX_LOCKED);
425 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
426 		if (p->p_pid == pid) {
427 			PROC_LOCK(p);
428 			if (p->p_state == PRS_NEW) {
429 				PROC_UNLOCK(p);
430 				p = NULL;
431 			}
432 			break;
433 		}
434 	}
435 	return (p);
436 }
437 
438 /*
439  * Locate a process by number.
440  *
441  * By not returning processes in the PRS_NEW state, we allow callers to avoid
442  * testing for that condition to avoid dereferencing p_ucred, et al.
443  */
444 static __always_inline struct proc *
445 _pfind(pid_t pid, bool zombie)
446 {
447 	struct proc *p;
448 
449 	p = curproc;
450 	if (p->p_pid == pid) {
451 		PROC_LOCK(p);
452 		return (p);
453 	}
454 	sx_slock(PIDHASHLOCK(pid));
455 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
456 		if (p->p_pid == pid) {
457 			PROC_LOCK(p);
458 			if (p->p_state == PRS_NEW ||
459 			    (!zombie && p->p_state == PRS_ZOMBIE)) {
460 				PROC_UNLOCK(p);
461 				p = NULL;
462 			}
463 			break;
464 		}
465 	}
466 	sx_sunlock(PIDHASHLOCK(pid));
467 	return (p);
468 }
469 
470 struct proc *
471 pfind(pid_t pid)
472 {
473 
474 	return (_pfind(pid, false));
475 }
476 
477 /*
478  * Same as pfind but allow zombies.
479  */
480 struct proc *
481 pfind_any(pid_t pid)
482 {
483 
484 	return (_pfind(pid, true));
485 }
486 
487 /*
488  * Locate a process group by number.
489  * The caller must hold proctree_lock.
490  */
491 struct pgrp *
492 pgfind(pid_t pgid)
493 {
494 	struct pgrp *pgrp;
495 
496 	sx_assert(&proctree_lock, SX_LOCKED);
497 
498 	LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
499 		if (pgrp->pg_id == pgid) {
500 			PGRP_LOCK(pgrp);
501 			return (pgrp);
502 		}
503 	}
504 	return (NULL);
505 }
506 
507 /*
508  * Locate process and do additional manipulations, depending on flags.
509  */
510 int
511 pget(pid_t pid, int flags, struct proc **pp)
512 {
513 	struct proc *p;
514 	struct thread *td1;
515 	int error;
516 
517 	p = curproc;
518 	if (p->p_pid == pid) {
519 		PROC_LOCK(p);
520 	} else {
521 		p = NULL;
522 		if (pid <= PID_MAX) {
523 			if ((flags & PGET_NOTWEXIT) == 0)
524 				p = pfind_any(pid);
525 			else
526 				p = pfind(pid);
527 		} else if ((flags & PGET_NOTID) == 0) {
528 			td1 = tdfind(pid, -1);
529 			if (td1 != NULL)
530 				p = td1->td_proc;
531 		}
532 		if (p == NULL)
533 			return (ESRCH);
534 		if ((flags & PGET_CANSEE) != 0) {
535 			error = p_cansee(curthread, p);
536 			if (error != 0)
537 				goto errout;
538 		}
539 	}
540 	if ((flags & PGET_CANDEBUG) != 0) {
541 		error = p_candebug(curthread, p);
542 		if (error != 0)
543 			goto errout;
544 	}
545 	if ((flags & PGET_ISCURRENT) != 0 && curproc != p) {
546 		error = EPERM;
547 		goto errout;
548 	}
549 	if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) {
550 		error = ESRCH;
551 		goto errout;
552 	}
553 	if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) {
554 		/*
555 		 * XXXRW: Not clear ESRCH is the right error during proc
556 		 * execve().
557 		 */
558 		error = ESRCH;
559 		goto errout;
560 	}
561 	if ((flags & PGET_HOLD) != 0) {
562 		_PHOLD(p);
563 		PROC_UNLOCK(p);
564 	}
565 	*pp = p;
566 	return (0);
567 errout:
568 	PROC_UNLOCK(p);
569 	return (error);
570 }
571 
572 /*
573  * Create a new process group.
574  * pgid must be equal to the pid of p.
575  * Begin a new session if required.
576  */
577 int
578 enterpgrp(struct proc *p, pid_t pgid, struct pgrp *pgrp, struct session *sess)
579 {
580 	struct pgrp *old_pgrp;
581 
582 	sx_assert(&proctree_lock, SX_XLOCKED);
583 
584 	KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
585 	KASSERT(p->p_pid == pgid,
586 	    ("enterpgrp: new pgrp and pid != pgid"));
587 	KASSERT(pgfind(pgid) == NULL,
588 	    ("enterpgrp: pgrp with pgid exists"));
589 	KASSERT(!SESS_LEADER(p),
590 	    ("enterpgrp: session leader attempted setpgrp"));
591 
592 	old_pgrp = p->p_pgrp;
593 	if (!sx_try_xlock(&old_pgrp->pg_killsx)) {
594 		sx_xunlock(&proctree_lock);
595 		sx_xlock(&old_pgrp->pg_killsx);
596 		sx_xunlock(&old_pgrp->pg_killsx);
597 		return (ERESTART);
598 	}
599 	MPASS(old_pgrp == p->p_pgrp);
600 
601 	if (sess != NULL) {
602 		/*
603 		 * new session
604 		 */
605 		mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
606 		PROC_LOCK(p);
607 		p->p_flag &= ~P_CONTROLT;
608 		PROC_UNLOCK(p);
609 		PGRP_LOCK(pgrp);
610 		sess->s_leader = p;
611 		sess->s_sid = p->p_pid;
612 		proc_id_set(PROC_ID_SESSION, p->p_pid);
613 		refcount_init(&sess->s_count, 1);
614 		sess->s_ttyvp = NULL;
615 		sess->s_ttydp = NULL;
616 		sess->s_ttyp = NULL;
617 		bcopy(p->p_session->s_login, sess->s_login,
618 			    sizeof(sess->s_login));
619 		pgrp->pg_session = sess;
620 		KASSERT(p == curproc,
621 		    ("enterpgrp: mksession and p != curproc"));
622 	} else {
623 		pgrp->pg_session = p->p_session;
624 		sess_hold(pgrp->pg_session);
625 		PGRP_LOCK(pgrp);
626 	}
627 	pgrp->pg_id = pgid;
628 	proc_id_set(PROC_ID_GROUP, p->p_pid);
629 	LIST_INIT(&pgrp->pg_members);
630 	pgrp->pg_flags = 0;
631 
632 	/*
633 	 * As we have an exclusive lock of proctree_lock,
634 	 * this should not deadlock.
635 	 */
636 	LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
637 	SLIST_INIT(&pgrp->pg_sigiolst);
638 	PGRP_UNLOCK(pgrp);
639 
640 	doenterpgrp(p, pgrp);
641 
642 	sx_xunlock(&old_pgrp->pg_killsx);
643 	return (0);
644 }
645 
646 /*
647  * Move p to an existing process group
648  */
649 int
650 enterthispgrp(struct proc *p, struct pgrp *pgrp)
651 {
652 	struct pgrp *old_pgrp;
653 
654 	sx_assert(&proctree_lock, SX_XLOCKED);
655 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
656 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
657 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
658 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
659 	KASSERT(pgrp->pg_session == p->p_session,
660 	    ("%s: pgrp's session %p, p->p_session %p proc %p\n",
661 	    __func__, pgrp->pg_session, p->p_session, p));
662 	KASSERT(pgrp != p->p_pgrp,
663 	    ("%s: p %p belongs to pgrp %p", __func__, p, pgrp));
664 
665 	old_pgrp = p->p_pgrp;
666 	if (!sx_try_xlock(&old_pgrp->pg_killsx)) {
667 		sx_xunlock(&proctree_lock);
668 		sx_xlock(&old_pgrp->pg_killsx);
669 		sx_xunlock(&old_pgrp->pg_killsx);
670 		return (ERESTART);
671 	}
672 	MPASS(old_pgrp == p->p_pgrp);
673 	if (!sx_try_xlock(&pgrp->pg_killsx)) {
674 		sx_xunlock(&old_pgrp->pg_killsx);
675 		sx_xunlock(&proctree_lock);
676 		sx_xlock(&pgrp->pg_killsx);
677 		sx_xunlock(&pgrp->pg_killsx);
678 		return (ERESTART);
679 	}
680 
681 	doenterpgrp(p, pgrp);
682 
683 	sx_xunlock(&pgrp->pg_killsx);
684 	sx_xunlock(&old_pgrp->pg_killsx);
685 	return (0);
686 }
687 
688 /*
689  * If true, any child of q which belongs to group pgrp, qualifies the
690  * process group pgrp as not orphaned.
691  */
692 static bool
693 isjobproc(struct proc *q, struct pgrp *pgrp)
694 {
695 	sx_assert(&proctree_lock, SX_LOCKED);
696 
697 	return (q->p_pgrp != pgrp &&
698 	    q->p_pgrp->pg_session == pgrp->pg_session);
699 }
700 
701 static struct proc *
702 jobc_reaper(struct proc *p)
703 {
704 	struct proc *pp;
705 
706 	sx_assert(&proctree_lock, SA_LOCKED);
707 
708 	for (pp = p;;) {
709 		pp = pp->p_reaper;
710 		if (pp->p_reaper == pp ||
711 		    (pp->p_treeflag & P_TREE_GRPEXITED) == 0)
712 			return (pp);
713 	}
714 }
715 
716 static struct proc *
717 jobc_parent(struct proc *p, struct proc *p_exiting)
718 {
719 	struct proc *pp;
720 
721 	sx_assert(&proctree_lock, SA_LOCKED);
722 
723 	pp = proc_realparent(p);
724 	if (pp->p_pptr == NULL || pp == p_exiting ||
725 	    (pp->p_treeflag & P_TREE_GRPEXITED) == 0)
726 		return (pp);
727 	return (jobc_reaper(pp));
728 }
729 
730 int
731 pgrp_calc_jobc(struct pgrp *pgrp)
732 {
733 	struct proc *q;
734 	int cnt;
735 
736 #ifdef INVARIANTS
737 	if (!mtx_owned(&pgrp->pg_mtx))
738 		sx_assert(&proctree_lock, SA_LOCKED);
739 #endif
740 
741 	cnt = 0;
742 	LIST_FOREACH(q, &pgrp->pg_members, p_pglist) {
743 		if ((q->p_treeflag & P_TREE_GRPEXITED) != 0 ||
744 		    q->p_pptr == NULL)
745 			continue;
746 		if (isjobproc(jobc_parent(q, NULL), pgrp))
747 			cnt++;
748 	}
749 	return (cnt);
750 }
751 
752 /*
753  * Move p to a process group
754  */
755 static void
756 doenterpgrp(struct proc *p, struct pgrp *pgrp)
757 {
758 	struct pgrp *savepgrp;
759 	struct proc *pp;
760 
761 	sx_assert(&proctree_lock, SX_XLOCKED);
762 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
763 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
764 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
765 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
766 
767 	savepgrp = p->p_pgrp;
768 	pp = jobc_parent(p, NULL);
769 
770 	PGRP_LOCK(pgrp);
771 	PGRP_LOCK(savepgrp);
772 	if (isjobproc(pp, savepgrp) && pgrp_calc_jobc(savepgrp) == 1)
773 		orphanpg(savepgrp);
774 	PROC_LOCK(p);
775 	LIST_REMOVE(p, p_pglist);
776 	p->p_pgrp = pgrp;
777 	PROC_UNLOCK(p);
778 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
779 	if (isjobproc(pp, pgrp))
780 		pgrp->pg_flags &= ~PGRP_ORPHANED;
781 	PGRP_UNLOCK(savepgrp);
782 	PGRP_UNLOCK(pgrp);
783 	if (LIST_EMPTY(&savepgrp->pg_members))
784 		pgdelete(savepgrp);
785 }
786 
787 /*
788  * remove process from process group
789  */
790 int
791 leavepgrp(struct proc *p)
792 {
793 	struct pgrp *savepgrp;
794 
795 	sx_assert(&proctree_lock, SX_XLOCKED);
796 	savepgrp = p->p_pgrp;
797 	PGRP_LOCK(savepgrp);
798 	PROC_LOCK(p);
799 	LIST_REMOVE(p, p_pglist);
800 	p->p_pgrp = NULL;
801 	PROC_UNLOCK(p);
802 	PGRP_UNLOCK(savepgrp);
803 	if (LIST_EMPTY(&savepgrp->pg_members))
804 		pgdelete(savepgrp);
805 	return (0);
806 }
807 
808 /*
809  * delete a process group
810  */
811 static void
812 pgdelete(struct pgrp *pgrp)
813 {
814 	struct session *savesess;
815 	struct tty *tp;
816 
817 	sx_assert(&proctree_lock, SX_XLOCKED);
818 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
819 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
820 
821 	/*
822 	 * Reset any sigio structures pointing to us as a result of
823 	 * F_SETOWN with our pgid.  The proctree lock ensures that
824 	 * new sigio structures will not be added after this point.
825 	 */
826 	funsetownlst(&pgrp->pg_sigiolst);
827 
828 	PGRP_LOCK(pgrp);
829 	tp = pgrp->pg_session->s_ttyp;
830 	LIST_REMOVE(pgrp, pg_hash);
831 	savesess = pgrp->pg_session;
832 	PGRP_UNLOCK(pgrp);
833 
834 	/* Remove the reference to the pgrp before deallocating it. */
835 	if (tp != NULL) {
836 		tty_lock(tp);
837 		tty_rel_pgrp(tp, pgrp);
838 	}
839 
840 	proc_id_clear(PROC_ID_GROUP, pgrp->pg_id);
841 	uma_zfree(pgrp_zone, pgrp);
842 	sess_release(savesess);
843 }
844 
845 
846 static void
847 fixjobc_kill(struct proc *p)
848 {
849 	struct proc *q;
850 	struct pgrp *pgrp;
851 
852 	sx_assert(&proctree_lock, SX_LOCKED);
853 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
854 	pgrp = p->p_pgrp;
855 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
856 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
857 
858 	/*
859 	 * p no longer affects process group orphanage for children.
860 	 * It is marked by the flag because p is only physically
861 	 * removed from its process group on wait(2).
862 	 */
863 	MPASS((p->p_treeflag & P_TREE_GRPEXITED) == 0);
864 	p->p_treeflag |= P_TREE_GRPEXITED;
865 
866 	/*
867 	 * Check if exiting p orphans its own group.
868 	 */
869 	pgrp = p->p_pgrp;
870 	if (isjobproc(jobc_parent(p, NULL), pgrp)) {
871 		PGRP_LOCK(pgrp);
872 		if (pgrp_calc_jobc(pgrp) == 0)
873 			orphanpg(pgrp);
874 		PGRP_UNLOCK(pgrp);
875 	}
876 
877 	/*
878 	 * Check this process' children to see whether they qualify
879 	 * their process groups after reparenting to reaper.
880 	 */
881 	LIST_FOREACH(q, &p->p_children, p_sibling) {
882 		pgrp = q->p_pgrp;
883 		PGRP_LOCK(pgrp);
884 		if (pgrp_calc_jobc(pgrp) == 0) {
885 			/*
886 			 * We want to handle exactly the children that
887 			 * has p as realparent.  Then, when calculating
888 			 * jobc_parent for children, we should ignore
889 			 * P_TREE_GRPEXITED flag already set on p.
890 			 */
891 			if (jobc_parent(q, p) == p && isjobproc(p, pgrp))
892 				orphanpg(pgrp);
893 		} else
894 			pgrp->pg_flags &= ~PGRP_ORPHANED;
895 		PGRP_UNLOCK(pgrp);
896 	}
897 	LIST_FOREACH(q, &p->p_orphans, p_orphan) {
898 		pgrp = q->p_pgrp;
899 		PGRP_LOCK(pgrp);
900 		if (pgrp_calc_jobc(pgrp) == 0) {
901 			if (isjobproc(p, pgrp))
902 				orphanpg(pgrp);
903 		} else
904 			pgrp->pg_flags &= ~PGRP_ORPHANED;
905 		PGRP_UNLOCK(pgrp);
906 	}
907 }
908 
909 void
910 killjobc(void)
911 {
912 	struct session *sp;
913 	struct tty *tp;
914 	struct proc *p;
915 	struct vnode *ttyvp;
916 
917 	p = curproc;
918 	MPASS(p->p_flag & P_WEXIT);
919 	sx_assert(&proctree_lock, SX_LOCKED);
920 
921 	if (SESS_LEADER(p)) {
922 		sp = p->p_session;
923 
924 		/*
925 		 * s_ttyp is not zero'd; we use this to indicate that
926 		 * the session once had a controlling terminal. (for
927 		 * logging and informational purposes)
928 		 */
929 		SESS_LOCK(sp);
930 		ttyvp = sp->s_ttyvp;
931 		tp = sp->s_ttyp;
932 		sp->s_ttyvp = NULL;
933 		sp->s_ttydp = NULL;
934 		sp->s_leader = NULL;
935 		SESS_UNLOCK(sp);
936 
937 		/*
938 		 * Signal foreground pgrp and revoke access to
939 		 * controlling terminal if it has not been revoked
940 		 * already.
941 		 *
942 		 * Because the TTY may have been revoked in the mean
943 		 * time and could already have a new session associated
944 		 * with it, make sure we don't send a SIGHUP to a
945 		 * foreground process group that does not belong to this
946 		 * session.
947 		 */
948 
949 		if (tp != NULL) {
950 			tty_lock(tp);
951 			if (tp->t_session == sp)
952 				tty_signal_pgrp(tp, SIGHUP);
953 			tty_unlock(tp);
954 		}
955 
956 		if (ttyvp != NULL) {
957 			sx_xunlock(&proctree_lock);
958 			if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) {
959 				VOP_REVOKE(ttyvp, REVOKEALL);
960 				VOP_UNLOCK(ttyvp);
961 			}
962 			devfs_ctty_unref(ttyvp);
963 			sx_xlock(&proctree_lock);
964 		}
965 	}
966 	fixjobc_kill(p);
967 }
968 
969 /*
970  * A process group has become orphaned, mark it as such for signal
971  * delivery code.  If there are any stopped processes in the group,
972  * hang-up all process in that group.
973  */
974 static void
975 orphanpg(struct pgrp *pg)
976 {
977 	struct proc *p;
978 
979 	PGRP_LOCK_ASSERT(pg, MA_OWNED);
980 
981 	pg->pg_flags |= PGRP_ORPHANED;
982 
983 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
984 		PROC_LOCK(p);
985 		if (P_SHOULDSTOP(p) == P_STOPPED_SIG) {
986 			PROC_UNLOCK(p);
987 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
988 				PROC_LOCK(p);
989 				kern_psignal(p, SIGHUP);
990 				kern_psignal(p, SIGCONT);
991 				PROC_UNLOCK(p);
992 			}
993 			return;
994 		}
995 		PROC_UNLOCK(p);
996 	}
997 }
998 
999 void
1000 sess_hold(struct session *s)
1001 {
1002 
1003 	refcount_acquire(&s->s_count);
1004 }
1005 
1006 void
1007 sess_release(struct session *s)
1008 {
1009 
1010 	if (refcount_release(&s->s_count)) {
1011 		if (s->s_ttyp != NULL) {
1012 			tty_lock(s->s_ttyp);
1013 			tty_rel_sess(s->s_ttyp, s);
1014 		}
1015 		proc_id_clear(PROC_ID_SESSION, s->s_sid);
1016 		mtx_destroy(&s->s_mtx);
1017 		free(s, M_SESSION);
1018 	}
1019 }
1020 
1021 #ifdef DDB
1022 
1023 static void
1024 db_print_pgrp_one(struct pgrp *pgrp, struct proc *p)
1025 {
1026 	db_printf(
1027 	    "    pid %d at %p pr %d pgrp %p e %d jc %d\n",
1028 	    p->p_pid, p, p->p_pptr == NULL ? -1 : p->p_pptr->p_pid,
1029 	    p->p_pgrp, (p->p_treeflag & P_TREE_GRPEXITED) != 0,
1030 	    p->p_pptr == NULL ? 0 : isjobproc(p->p_pptr, pgrp));
1031 }
1032 
1033 DB_SHOW_COMMAND_FLAGS(pgrpdump, pgrpdump, DB_CMD_MEMSAFE)
1034 {
1035 	struct pgrp *pgrp;
1036 	struct proc *p;
1037 	int i;
1038 
1039 	for (i = 0; i <= pgrphash; i++) {
1040 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
1041 			db_printf("indx %d\n", i);
1042 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
1043 				db_printf(
1044 			"  pgrp %p, pgid %d, sess %p, sesscnt %d, mem %p\n",
1045 				    pgrp, (int)pgrp->pg_id, pgrp->pg_session,
1046 				    pgrp->pg_session->s_count,
1047 				    LIST_FIRST(&pgrp->pg_members));
1048 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist)
1049 					db_print_pgrp_one(pgrp, p);
1050 			}
1051 		}
1052 	}
1053 }
1054 #endif /* DDB */
1055 
1056 /*
1057  * Calculate the kinfo_proc members which contain process-wide
1058  * informations.
1059  * Must be called with the target process locked.
1060  */
1061 static void
1062 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
1063 {
1064 	struct thread *td;
1065 
1066 	PROC_LOCK_ASSERT(p, MA_OWNED);
1067 
1068 	kp->ki_estcpu = 0;
1069 	kp->ki_pctcpu = 0;
1070 	FOREACH_THREAD_IN_PROC(p, td) {
1071 		thread_lock(td);
1072 		kp->ki_pctcpu += sched_pctcpu(td);
1073 		kp->ki_estcpu += sched_estcpu(td);
1074 		thread_unlock(td);
1075 	}
1076 }
1077 
1078 /*
1079  * Fill in any information that is common to all threads in the process.
1080  * Must be called with the target process locked.
1081  */
1082 static void
1083 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
1084 {
1085 	struct thread *td0;
1086 	struct ucred *cred;
1087 	struct sigacts *ps;
1088 	struct timeval boottime;
1089 
1090 	PROC_LOCK_ASSERT(p, MA_OWNED);
1091 
1092 	kp->ki_structsize = sizeof(*kp);
1093 	kp->ki_paddr = p;
1094 	kp->ki_addr =/* p->p_addr; */0; /* XXX */
1095 	kp->ki_args = p->p_args;
1096 	kp->ki_textvp = p->p_textvp;
1097 #ifdef KTRACE
1098 	kp->ki_tracep = ktr_get_tracevp(p, false);
1099 	kp->ki_traceflag = p->p_traceflag;
1100 #endif
1101 	kp->ki_fd = p->p_fd;
1102 	kp->ki_pd = p->p_pd;
1103 	kp->ki_vmspace = p->p_vmspace;
1104 	kp->ki_flag = p->p_flag;
1105 	kp->ki_flag2 = p->p_flag2;
1106 	cred = p->p_ucred;
1107 	if (cred) {
1108 		kp->ki_uid = cred->cr_uid;
1109 		kp->ki_ruid = cred->cr_ruid;
1110 		kp->ki_svuid = cred->cr_svuid;
1111 		kp->ki_cr_flags = 0;
1112 		if (cred->cr_flags & CRED_FLAG_CAPMODE)
1113 			kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
1114 		/* XXX bde doesn't like KI_NGROUPS */
1115 		if (cred->cr_ngroups > KI_NGROUPS) {
1116 			kp->ki_ngroups = KI_NGROUPS;
1117 			kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
1118 		} else
1119 			kp->ki_ngroups = cred->cr_ngroups;
1120 		bcopy(cred->cr_groups, kp->ki_groups,
1121 		    kp->ki_ngroups * sizeof(gid_t));
1122 		kp->ki_rgid = cred->cr_rgid;
1123 		kp->ki_svgid = cred->cr_svgid;
1124 		/* If jailed(cred), emulate the old P_JAILED flag. */
1125 		if (jailed(cred)) {
1126 			kp->ki_flag |= P_JAILED;
1127 			/* If inside the jail, use 0 as a jail ID. */
1128 			if (cred->cr_prison != curthread->td_ucred->cr_prison)
1129 				kp->ki_jid = cred->cr_prison->pr_id;
1130 		}
1131 		strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
1132 		    sizeof(kp->ki_loginclass));
1133 	}
1134 	ps = p->p_sigacts;
1135 	if (ps) {
1136 		mtx_lock(&ps->ps_mtx);
1137 		kp->ki_sigignore = ps->ps_sigignore;
1138 		kp->ki_sigcatch = ps->ps_sigcatch;
1139 		mtx_unlock(&ps->ps_mtx);
1140 	}
1141 	if (p->p_state != PRS_NEW &&
1142 	    p->p_state != PRS_ZOMBIE &&
1143 	    p->p_vmspace != NULL) {
1144 		struct vmspace *vm = p->p_vmspace;
1145 
1146 		kp->ki_size = vm->vm_map.size;
1147 		kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
1148 		FOREACH_THREAD_IN_PROC(p, td0)
1149 			kp->ki_rssize += td0->td_kstack_pages;
1150 		kp->ki_swrss = vm->vm_swrss;
1151 		kp->ki_tsize = vm->vm_tsize;
1152 		kp->ki_dsize = vm->vm_dsize;
1153 		kp->ki_ssize = vm->vm_ssize;
1154 	} else if (p->p_state == PRS_ZOMBIE)
1155 		kp->ki_stat = SZOMB;
1156 	kp->ki_sflag = PS_INMEM;
1157 	/* Calculate legacy swtime as seconds since 'swtick'. */
1158 	kp->ki_swtime = (ticks - p->p_swtick) / hz;
1159 	kp->ki_pid = p->p_pid;
1160 	kp->ki_nice = p->p_nice;
1161 	kp->ki_fibnum = p->p_fibnum;
1162 	kp->ki_start = p->p_stats->p_start;
1163 	getboottime(&boottime);
1164 	timevaladd(&kp->ki_start, &boottime);
1165 	PROC_STATLOCK(p);
1166 	rufetch(p, &kp->ki_rusage);
1167 	kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
1168 	calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
1169 	PROC_STATUNLOCK(p);
1170 	calccru(p, &kp->ki_childutime, &kp->ki_childstime);
1171 	/* Some callers want child times in a single value. */
1172 	kp->ki_childtime = kp->ki_childstime;
1173 	timevaladd(&kp->ki_childtime, &kp->ki_childutime);
1174 
1175 	FOREACH_THREAD_IN_PROC(p, td0)
1176 		kp->ki_cow += td0->td_cow;
1177 
1178 	if (p->p_comm[0] != '\0')
1179 		strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
1180 	if (p->p_sysent && p->p_sysent->sv_name != NULL &&
1181 	    p->p_sysent->sv_name[0] != '\0')
1182 		strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
1183 	kp->ki_siglist = p->p_siglist;
1184 	kp->ki_xstat = KW_EXITCODE(p->p_xexit, p->p_xsig);
1185 	kp->ki_acflag = p->p_acflag;
1186 	kp->ki_lock = p->p_lock;
1187 	if (p->p_pptr) {
1188 		kp->ki_ppid = p->p_oppid;
1189 		if (p->p_flag & P_TRACED)
1190 			kp->ki_tracer = p->p_pptr->p_pid;
1191 	}
1192 }
1193 
1194 /*
1195  * Fill job-related process information.
1196  */
1197 static void
1198 fill_kinfo_proc_pgrp(struct proc *p, struct kinfo_proc *kp)
1199 {
1200 	struct tty *tp;
1201 	struct session *sp;
1202 	struct pgrp *pgrp;
1203 
1204 	sx_assert(&proctree_lock, SA_LOCKED);
1205 	PROC_LOCK_ASSERT(p, MA_OWNED);
1206 
1207 	pgrp = p->p_pgrp;
1208 	if (pgrp == NULL)
1209 		return;
1210 
1211 	kp->ki_pgid = pgrp->pg_id;
1212 	kp->ki_jobc = pgrp_calc_jobc(pgrp);
1213 
1214 	sp = pgrp->pg_session;
1215 	tp = NULL;
1216 
1217 	if (sp != NULL) {
1218 		kp->ki_sid = sp->s_sid;
1219 		SESS_LOCK(sp);
1220 		strlcpy(kp->ki_login, sp->s_login, sizeof(kp->ki_login));
1221 		if (sp->s_ttyvp)
1222 			kp->ki_kiflag |= KI_CTTY;
1223 		if (SESS_LEADER(p))
1224 			kp->ki_kiflag |= KI_SLEADER;
1225 		tp = sp->s_ttyp;
1226 		SESS_UNLOCK(sp);
1227 	}
1228 
1229 	if ((p->p_flag & P_CONTROLT) && tp != NULL) {
1230 		kp->ki_tdev = tty_udev(tp);
1231 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1232 		kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
1233 		if (tp->t_session)
1234 			kp->ki_tsid = tp->t_session->s_sid;
1235 	} else {
1236 		kp->ki_tdev = NODEV;
1237 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1238 	}
1239 }
1240 
1241 /*
1242  * Fill in information that is thread specific.  Must be called with
1243  * target process locked.  If 'preferthread' is set, overwrite certain
1244  * process-related fields that are maintained for both threads and
1245  * processes.
1246  */
1247 static void
1248 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
1249 {
1250 	struct proc *p;
1251 
1252 	p = td->td_proc;
1253 	kp->ki_tdaddr = td;
1254 	PROC_LOCK_ASSERT(p, MA_OWNED);
1255 
1256 	if (preferthread)
1257 		PROC_STATLOCK(p);
1258 	thread_lock(td);
1259 	if (td->td_wmesg != NULL)
1260 		strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
1261 	else
1262 		bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
1263 	if (strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname)) >=
1264 	    sizeof(kp->ki_tdname)) {
1265 		strlcpy(kp->ki_moretdname,
1266 		    td->td_name + sizeof(kp->ki_tdname) - 1,
1267 		    sizeof(kp->ki_moretdname));
1268 	} else {
1269 		bzero(kp->ki_moretdname, sizeof(kp->ki_moretdname));
1270 	}
1271 	if (TD_ON_LOCK(td)) {
1272 		kp->ki_kiflag |= KI_LOCKBLOCK;
1273 		strlcpy(kp->ki_lockname, td->td_lockname,
1274 		    sizeof(kp->ki_lockname));
1275 	} else {
1276 		kp->ki_kiflag &= ~KI_LOCKBLOCK;
1277 		bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
1278 	}
1279 
1280 	if (p->p_state == PRS_NORMAL) { /* approximate. */
1281 		if (TD_ON_RUNQ(td) ||
1282 		    TD_CAN_RUN(td) ||
1283 		    TD_IS_RUNNING(td)) {
1284 			kp->ki_stat = SRUN;
1285 		} else if (P_SHOULDSTOP(p)) {
1286 			kp->ki_stat = SSTOP;
1287 		} else if (TD_IS_SLEEPING(td)) {
1288 			kp->ki_stat = SSLEEP;
1289 		} else if (TD_ON_LOCK(td)) {
1290 			kp->ki_stat = SLOCK;
1291 		} else {
1292 			kp->ki_stat = SWAIT;
1293 		}
1294 	} else if (p->p_state == PRS_ZOMBIE) {
1295 		kp->ki_stat = SZOMB;
1296 	} else {
1297 		kp->ki_stat = SIDL;
1298 	}
1299 
1300 	/* Things in the thread */
1301 	kp->ki_wchan = td->td_wchan;
1302 	kp->ki_pri.pri_level = td->td_priority;
1303 	kp->ki_pri.pri_native = td->td_base_pri;
1304 
1305 	/*
1306 	 * Note: legacy fields; clamp at the old NOCPU value and/or
1307 	 * the maximum u_char CPU value.
1308 	 */
1309 	if (td->td_lastcpu == NOCPU)
1310 		kp->ki_lastcpu_old = NOCPU_OLD;
1311 	else if (td->td_lastcpu > MAXCPU_OLD)
1312 		kp->ki_lastcpu_old = MAXCPU_OLD;
1313 	else
1314 		kp->ki_lastcpu_old = td->td_lastcpu;
1315 
1316 	if (td->td_oncpu == NOCPU)
1317 		kp->ki_oncpu_old = NOCPU_OLD;
1318 	else if (td->td_oncpu > MAXCPU_OLD)
1319 		kp->ki_oncpu_old = MAXCPU_OLD;
1320 	else
1321 		kp->ki_oncpu_old = td->td_oncpu;
1322 
1323 	kp->ki_lastcpu = td->td_lastcpu;
1324 	kp->ki_oncpu = td->td_oncpu;
1325 	kp->ki_tdflags = td->td_flags;
1326 	kp->ki_tid = td->td_tid;
1327 	kp->ki_numthreads = p->p_numthreads;
1328 	kp->ki_pcb = td->td_pcb;
1329 	kp->ki_kstack = (void *)td->td_kstack;
1330 	kp->ki_slptime = (ticks - td->td_slptick) / hz;
1331 	kp->ki_pri.pri_class = td->td_pri_class;
1332 	kp->ki_pri.pri_user = td->td_user_pri;
1333 
1334 	if (preferthread) {
1335 		rufetchtd(td, &kp->ki_rusage);
1336 		kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
1337 		kp->ki_pctcpu = sched_pctcpu(td);
1338 		kp->ki_estcpu = sched_estcpu(td);
1339 		kp->ki_cow = td->td_cow;
1340 	}
1341 
1342 	/* We can't get this anymore but ps etc never used it anyway. */
1343 	kp->ki_rqindex = 0;
1344 
1345 	if (preferthread)
1346 		kp->ki_siglist = td->td_siglist;
1347 	kp->ki_sigmask = td->td_sigmask;
1348 	thread_unlock(td);
1349 	if (preferthread)
1350 		PROC_STATUNLOCK(p);
1351 }
1352 
1353 /*
1354  * Fill in a kinfo_proc structure for the specified process.
1355  * Must be called with the target process locked.
1356  */
1357 void
1358 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1359 {
1360 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1361 
1362 	bzero(kp, sizeof(*kp));
1363 
1364 	fill_kinfo_proc_pgrp(p,kp);
1365 	fill_kinfo_proc_only(p, kp);
1366 	fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1367 	fill_kinfo_aggregate(p, kp);
1368 }
1369 
1370 struct pstats *
1371 pstats_alloc(void)
1372 {
1373 
1374 	return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1375 }
1376 
1377 /*
1378  * Copy parts of p_stats; zero the rest of p_stats (statistics).
1379  */
1380 void
1381 pstats_fork(struct pstats *src, struct pstats *dst)
1382 {
1383 
1384 	bzero(&dst->pstat_startzero,
1385 	    __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1386 	bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1387 	    __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1388 }
1389 
1390 void
1391 pstats_free(struct pstats *ps)
1392 {
1393 
1394 	free(ps, M_SUBPROC);
1395 }
1396 
1397 #ifdef COMPAT_FREEBSD32
1398 
1399 /*
1400  * This function is typically used to copy out the kernel address, so
1401  * it can be replaced by assignment of zero.
1402  */
1403 static inline uint32_t
1404 ptr32_trim(const void *ptr)
1405 {
1406 	uintptr_t uptr;
1407 
1408 	uptr = (uintptr_t)ptr;
1409 	return ((uptr > UINT_MAX) ? 0 : uptr);
1410 }
1411 
1412 #define PTRTRIM_CP(src,dst,fld) \
1413 	do { (dst).fld = ptr32_trim((src).fld); } while (0)
1414 
1415 static void
1416 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1417 {
1418 	int i;
1419 
1420 	bzero(ki32, sizeof(struct kinfo_proc32));
1421 	ki32->ki_structsize = sizeof(struct kinfo_proc32);
1422 	CP(*ki, *ki32, ki_layout);
1423 	PTRTRIM_CP(*ki, *ki32, ki_args);
1424 	PTRTRIM_CP(*ki, *ki32, ki_paddr);
1425 	PTRTRIM_CP(*ki, *ki32, ki_addr);
1426 	PTRTRIM_CP(*ki, *ki32, ki_tracep);
1427 	PTRTRIM_CP(*ki, *ki32, ki_textvp);
1428 	PTRTRIM_CP(*ki, *ki32, ki_fd);
1429 	PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1430 	PTRTRIM_CP(*ki, *ki32, ki_wchan);
1431 	CP(*ki, *ki32, ki_pid);
1432 	CP(*ki, *ki32, ki_ppid);
1433 	CP(*ki, *ki32, ki_pgid);
1434 	CP(*ki, *ki32, ki_tpgid);
1435 	CP(*ki, *ki32, ki_sid);
1436 	CP(*ki, *ki32, ki_tsid);
1437 	CP(*ki, *ki32, ki_jobc);
1438 	CP(*ki, *ki32, ki_tdev);
1439 	CP(*ki, *ki32, ki_tdev_freebsd11);
1440 	CP(*ki, *ki32, ki_siglist);
1441 	CP(*ki, *ki32, ki_sigmask);
1442 	CP(*ki, *ki32, ki_sigignore);
1443 	CP(*ki, *ki32, ki_sigcatch);
1444 	CP(*ki, *ki32, ki_uid);
1445 	CP(*ki, *ki32, ki_ruid);
1446 	CP(*ki, *ki32, ki_svuid);
1447 	CP(*ki, *ki32, ki_rgid);
1448 	CP(*ki, *ki32, ki_svgid);
1449 	CP(*ki, *ki32, ki_ngroups);
1450 	for (i = 0; i < KI_NGROUPS; i++)
1451 		CP(*ki, *ki32, ki_groups[i]);
1452 	CP(*ki, *ki32, ki_size);
1453 	CP(*ki, *ki32, ki_rssize);
1454 	CP(*ki, *ki32, ki_swrss);
1455 	CP(*ki, *ki32, ki_tsize);
1456 	CP(*ki, *ki32, ki_dsize);
1457 	CP(*ki, *ki32, ki_ssize);
1458 	CP(*ki, *ki32, ki_xstat);
1459 	CP(*ki, *ki32, ki_acflag);
1460 	CP(*ki, *ki32, ki_pctcpu);
1461 	CP(*ki, *ki32, ki_estcpu);
1462 	CP(*ki, *ki32, ki_slptime);
1463 	CP(*ki, *ki32, ki_swtime);
1464 	CP(*ki, *ki32, ki_cow);
1465 	CP(*ki, *ki32, ki_runtime);
1466 	TV_CP(*ki, *ki32, ki_start);
1467 	TV_CP(*ki, *ki32, ki_childtime);
1468 	CP(*ki, *ki32, ki_flag);
1469 	CP(*ki, *ki32, ki_kiflag);
1470 	CP(*ki, *ki32, ki_traceflag);
1471 	CP(*ki, *ki32, ki_stat);
1472 	CP(*ki, *ki32, ki_nice);
1473 	CP(*ki, *ki32, ki_lock);
1474 	CP(*ki, *ki32, ki_rqindex);
1475 	CP(*ki, *ki32, ki_oncpu);
1476 	CP(*ki, *ki32, ki_lastcpu);
1477 
1478 	/* XXX TODO: wrap cpu value as appropriate */
1479 	CP(*ki, *ki32, ki_oncpu_old);
1480 	CP(*ki, *ki32, ki_lastcpu_old);
1481 
1482 	bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1483 	bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1484 	bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1485 	bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1486 	bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1487 	bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1488 	bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1489 	bcopy(ki->ki_moretdname, ki32->ki_moretdname, MAXCOMLEN - TDNAMLEN + 1);
1490 	CP(*ki, *ki32, ki_tracer);
1491 	CP(*ki, *ki32, ki_flag2);
1492 	CP(*ki, *ki32, ki_fibnum);
1493 	CP(*ki, *ki32, ki_cr_flags);
1494 	CP(*ki, *ki32, ki_jid);
1495 	CP(*ki, *ki32, ki_numthreads);
1496 	CP(*ki, *ki32, ki_tid);
1497 	CP(*ki, *ki32, ki_pri);
1498 	freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1499 	freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1500 	PTRTRIM_CP(*ki, *ki32, ki_pcb);
1501 	PTRTRIM_CP(*ki, *ki32, ki_kstack);
1502 	PTRTRIM_CP(*ki, *ki32, ki_udata);
1503 	PTRTRIM_CP(*ki, *ki32, ki_tdaddr);
1504 	CP(*ki, *ki32, ki_sflag);
1505 	CP(*ki, *ki32, ki_tdflags);
1506 }
1507 #endif
1508 
1509 static ssize_t
1510 kern_proc_out_size(struct proc *p, int flags)
1511 {
1512 	ssize_t size = 0;
1513 
1514 	PROC_LOCK_ASSERT(p, MA_OWNED);
1515 
1516 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1517 #ifdef COMPAT_FREEBSD32
1518 		if ((flags & KERN_PROC_MASK32) != 0) {
1519 			size += sizeof(struct kinfo_proc32);
1520 		} else
1521 #endif
1522 			size += sizeof(struct kinfo_proc);
1523 	} else {
1524 #ifdef COMPAT_FREEBSD32
1525 		if ((flags & KERN_PROC_MASK32) != 0)
1526 			size += sizeof(struct kinfo_proc32) * p->p_numthreads;
1527 		else
1528 #endif
1529 			size += sizeof(struct kinfo_proc) * p->p_numthreads;
1530 	}
1531 	PROC_UNLOCK(p);
1532 	return (size);
1533 }
1534 
1535 int
1536 kern_proc_out(struct proc *p, struct sbuf *sb, int flags)
1537 {
1538 	struct thread *td;
1539 	struct kinfo_proc ki;
1540 #ifdef COMPAT_FREEBSD32
1541 	struct kinfo_proc32 ki32;
1542 #endif
1543 	int error;
1544 
1545 	PROC_LOCK_ASSERT(p, MA_OWNED);
1546 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1547 
1548 	error = 0;
1549 	fill_kinfo_proc(p, &ki);
1550 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1551 #ifdef COMPAT_FREEBSD32
1552 		if ((flags & KERN_PROC_MASK32) != 0) {
1553 			freebsd32_kinfo_proc_out(&ki, &ki32);
1554 			if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1555 				error = ENOMEM;
1556 		} else
1557 #endif
1558 			if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1559 				error = ENOMEM;
1560 	} else {
1561 		FOREACH_THREAD_IN_PROC(p, td) {
1562 			fill_kinfo_thread(td, &ki, 1);
1563 #ifdef COMPAT_FREEBSD32
1564 			if ((flags & KERN_PROC_MASK32) != 0) {
1565 				freebsd32_kinfo_proc_out(&ki, &ki32);
1566 				if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1567 					error = ENOMEM;
1568 			} else
1569 #endif
1570 				if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1571 					error = ENOMEM;
1572 			if (error != 0)
1573 				break;
1574 		}
1575 	}
1576 	PROC_UNLOCK(p);
1577 	return (error);
1578 }
1579 
1580 static int
1581 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
1582 {
1583 	struct sbuf sb;
1584 	struct kinfo_proc ki;
1585 	int error, error2;
1586 
1587 	if (req->oldptr == NULL)
1588 		return (SYSCTL_OUT(req, 0, kern_proc_out_size(p, flags)));
1589 
1590 	sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req);
1591 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1592 	error = kern_proc_out(p, &sb, flags);
1593 	error2 = sbuf_finish(&sb);
1594 	sbuf_delete(&sb);
1595 	if (error != 0)
1596 		return (error);
1597 	else if (error2 != 0)
1598 		return (error2);
1599 	return (0);
1600 }
1601 
1602 int
1603 proc_iterate(int (*cb)(struct proc *, void *), void *cbarg)
1604 {
1605 	struct proc *p;
1606 	int error, i, j;
1607 
1608 	for (i = 0; i < pidhashlock + 1; i++) {
1609 		sx_slock(&proctree_lock);
1610 		sx_slock(&pidhashtbl_lock[i]);
1611 		for (j = i; j <= pidhash; j += pidhashlock + 1) {
1612 			LIST_FOREACH(p, &pidhashtbl[j], p_hash) {
1613 				if (p->p_state == PRS_NEW)
1614 					continue;
1615 				error = cb(p, cbarg);
1616 				PROC_LOCK_ASSERT(p, MA_NOTOWNED);
1617 				if (error != 0) {
1618 					sx_sunlock(&pidhashtbl_lock[i]);
1619 					sx_sunlock(&proctree_lock);
1620 					return (error);
1621 				}
1622 			}
1623 		}
1624 		sx_sunlock(&pidhashtbl_lock[i]);
1625 		sx_sunlock(&proctree_lock);
1626 	}
1627 	return (0);
1628 }
1629 
1630 struct kern_proc_out_args {
1631 	struct sysctl_req *req;
1632 	int flags;
1633 	int oid_number;
1634 	int *name;
1635 };
1636 
1637 static int
1638 sysctl_kern_proc_iterate(struct proc *p, void *origarg)
1639 {
1640 	struct kern_proc_out_args *arg = origarg;
1641 	int *name = arg->name;
1642 	int oid_number = arg->oid_number;
1643 	int flags = arg->flags;
1644 	struct sysctl_req *req = arg->req;
1645 	int error = 0;
1646 
1647 	PROC_LOCK(p);
1648 
1649 	KASSERT(p->p_ucred != NULL,
1650 	    ("process credential is NULL for non-NEW proc"));
1651 	/*
1652 	 * Show a user only appropriate processes.
1653 	 */
1654 	if (p_cansee(curthread, p))
1655 		goto skip;
1656 	/*
1657 	 * TODO - make more efficient (see notes below).
1658 	 * do by session.
1659 	 */
1660 	switch (oid_number) {
1661 	case KERN_PROC_GID:
1662 		if (p->p_ucred->cr_gid != (gid_t)name[0])
1663 			goto skip;
1664 		break;
1665 
1666 	case KERN_PROC_PGRP:
1667 		/* could do this by traversing pgrp */
1668 		if (p->p_pgrp == NULL ||
1669 		    p->p_pgrp->pg_id != (pid_t)name[0])
1670 			goto skip;
1671 		break;
1672 
1673 	case KERN_PROC_RGID:
1674 		if (p->p_ucred->cr_rgid != (gid_t)name[0])
1675 			goto skip;
1676 		break;
1677 
1678 	case KERN_PROC_SESSION:
1679 		if (p->p_session == NULL ||
1680 		    p->p_session->s_sid != (pid_t)name[0])
1681 			goto skip;
1682 		break;
1683 
1684 	case KERN_PROC_TTY:
1685 		if ((p->p_flag & P_CONTROLT) == 0 ||
1686 		    p->p_session == NULL)
1687 			goto skip;
1688 		/* XXX proctree_lock */
1689 		SESS_LOCK(p->p_session);
1690 		if (p->p_session->s_ttyp == NULL ||
1691 		    tty_udev(p->p_session->s_ttyp) !=
1692 		    (dev_t)name[0]) {
1693 			SESS_UNLOCK(p->p_session);
1694 			goto skip;
1695 		}
1696 		SESS_UNLOCK(p->p_session);
1697 		break;
1698 
1699 	case KERN_PROC_UID:
1700 		if (p->p_ucred->cr_uid != (uid_t)name[0])
1701 			goto skip;
1702 		break;
1703 
1704 	case KERN_PROC_RUID:
1705 		if (p->p_ucred->cr_ruid != (uid_t)name[0])
1706 			goto skip;
1707 		break;
1708 
1709 	case KERN_PROC_PROC:
1710 		break;
1711 
1712 	default:
1713 		break;
1714 	}
1715 	error = sysctl_out_proc(p, req, flags);
1716 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
1717 	return (error);
1718 skip:
1719 	PROC_UNLOCK(p);
1720 	return (0);
1721 }
1722 
1723 static int
1724 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1725 {
1726 	struct kern_proc_out_args iterarg;
1727 	int *name = (int *)arg1;
1728 	u_int namelen = arg2;
1729 	struct proc *p;
1730 	int flags, oid_number;
1731 	int error = 0;
1732 
1733 	oid_number = oidp->oid_number;
1734 	if (oid_number != KERN_PROC_ALL &&
1735 	    (oid_number & KERN_PROC_INC_THREAD) == 0)
1736 		flags = KERN_PROC_NOTHREADS;
1737 	else {
1738 		flags = 0;
1739 		oid_number &= ~KERN_PROC_INC_THREAD;
1740 	}
1741 #ifdef COMPAT_FREEBSD32
1742 	if (req->flags & SCTL_MASK32)
1743 		flags |= KERN_PROC_MASK32;
1744 #endif
1745 	if (oid_number == KERN_PROC_PID) {
1746 		if (namelen != 1)
1747 			return (EINVAL);
1748 		error = sysctl_wire_old_buffer(req, 0);
1749 		if (error)
1750 			return (error);
1751 		sx_slock(&proctree_lock);
1752 		error = pget((pid_t)name[0], PGET_CANSEE, &p);
1753 		if (error == 0)
1754 			error = sysctl_out_proc(p, req, flags);
1755 		sx_sunlock(&proctree_lock);
1756 		return (error);
1757 	}
1758 
1759 	switch (oid_number) {
1760 	case KERN_PROC_ALL:
1761 		if (namelen != 0)
1762 			return (EINVAL);
1763 		break;
1764 	case KERN_PROC_PROC:
1765 		if (namelen != 0 && namelen != 1)
1766 			return (EINVAL);
1767 		break;
1768 	default:
1769 		if (namelen != 1)
1770 			return (EINVAL);
1771 		break;
1772 	}
1773 
1774 	if (req->oldptr == NULL) {
1775 		/* overestimate by 5 procs */
1776 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1777 		if (error)
1778 			return (error);
1779 	} else {
1780 		error = sysctl_wire_old_buffer(req, 0);
1781 		if (error != 0)
1782 			return (error);
1783 	}
1784 	iterarg.flags = flags;
1785 	iterarg.oid_number = oid_number;
1786 	iterarg.req = req;
1787 	iterarg.name = name;
1788 	error = proc_iterate(sysctl_kern_proc_iterate, &iterarg);
1789 	return (error);
1790 }
1791 
1792 struct pargs *
1793 pargs_alloc(int len)
1794 {
1795 	struct pargs *pa;
1796 
1797 	pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1798 		M_WAITOK);
1799 	refcount_init(&pa->ar_ref, 1);
1800 	pa->ar_length = len;
1801 	return (pa);
1802 }
1803 
1804 static void
1805 pargs_free(struct pargs *pa)
1806 {
1807 
1808 	free(pa, M_PARGS);
1809 }
1810 
1811 void
1812 pargs_hold(struct pargs *pa)
1813 {
1814 
1815 	if (pa == NULL)
1816 		return;
1817 	refcount_acquire(&pa->ar_ref);
1818 }
1819 
1820 void
1821 pargs_drop(struct pargs *pa)
1822 {
1823 
1824 	if (pa == NULL)
1825 		return;
1826 	if (refcount_release(&pa->ar_ref))
1827 		pargs_free(pa);
1828 }
1829 
1830 static int
1831 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1832     size_t len)
1833 {
1834 	ssize_t n;
1835 
1836 	/*
1837 	 * This may return a short read if the string is shorter than the chunk
1838 	 * and is aligned at the end of the page, and the following page is not
1839 	 * mapped.
1840 	 */
1841 	n = proc_readmem(td, p, (vm_offset_t)sptr, buf, len);
1842 	if (n <= 0)
1843 		return (ENOMEM);
1844 	return (0);
1845 }
1846 
1847 #define PROC_AUXV_MAX	256	/* Safety limit on auxv size. */
1848 
1849 enum proc_vector_type {
1850 	PROC_ARG,
1851 	PROC_ENV,
1852 	PROC_AUX,
1853 };
1854 
1855 #ifdef COMPAT_FREEBSD32
1856 static int
1857 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1858     size_t *vsizep, enum proc_vector_type type)
1859 {
1860 	struct freebsd32_ps_strings pss;
1861 	Elf32_Auxinfo aux;
1862 	vm_offset_t vptr, ptr;
1863 	uint32_t *proc_vector32;
1864 	char **proc_vector;
1865 	size_t vsize, size;
1866 	int i, error;
1867 
1868 	error = 0;
1869 	if (proc_readmem(td, p, PROC_PS_STRINGS(p), &pss, sizeof(pss)) !=
1870 	    sizeof(pss))
1871 		return (ENOMEM);
1872 	switch (type) {
1873 	case PROC_ARG:
1874 		vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1875 		vsize = pss.ps_nargvstr;
1876 		if (vsize > ARG_MAX)
1877 			return (ENOEXEC);
1878 		size = vsize * sizeof(int32_t);
1879 		break;
1880 	case PROC_ENV:
1881 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1882 		vsize = pss.ps_nenvstr;
1883 		if (vsize > ARG_MAX)
1884 			return (ENOEXEC);
1885 		size = vsize * sizeof(int32_t);
1886 		break;
1887 	case PROC_AUX:
1888 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1889 		    (pss.ps_nenvstr + 1) * sizeof(int32_t);
1890 		if (vptr % 4 != 0)
1891 			return (ENOEXEC);
1892 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1893 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1894 			    sizeof(aux))
1895 				return (ENOMEM);
1896 			if (aux.a_type == AT_NULL)
1897 				break;
1898 			ptr += sizeof(aux);
1899 		}
1900 		if (aux.a_type != AT_NULL)
1901 			return (ENOEXEC);
1902 		vsize = i + 1;
1903 		size = vsize * sizeof(aux);
1904 		break;
1905 	default:
1906 		KASSERT(0, ("Wrong proc vector type: %d", type));
1907 		return (EINVAL);
1908 	}
1909 	proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1910 	if (proc_readmem(td, p, vptr, proc_vector32, size) != size) {
1911 		error = ENOMEM;
1912 		goto done;
1913 	}
1914 	if (type == PROC_AUX) {
1915 		*proc_vectorp = (char **)proc_vector32;
1916 		*vsizep = vsize;
1917 		return (0);
1918 	}
1919 	proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1920 	for (i = 0; i < (int)vsize; i++)
1921 		proc_vector[i] = PTRIN(proc_vector32[i]);
1922 	*proc_vectorp = proc_vector;
1923 	*vsizep = vsize;
1924 done:
1925 	free(proc_vector32, M_TEMP);
1926 	return (error);
1927 }
1928 #endif
1929 
1930 static int
1931 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1932     size_t *vsizep, enum proc_vector_type type)
1933 {
1934 	struct ps_strings pss;
1935 	Elf_Auxinfo aux;
1936 	vm_offset_t vptr, ptr;
1937 	char **proc_vector;
1938 	size_t vsize, size;
1939 	int i;
1940 
1941 #ifdef COMPAT_FREEBSD32
1942 	if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1943 		return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1944 #endif
1945 	if (proc_readmem(td, p, PROC_PS_STRINGS(p), &pss, sizeof(pss)) !=
1946 	    sizeof(pss))
1947 		return (ENOMEM);
1948 	switch (type) {
1949 	case PROC_ARG:
1950 		vptr = (vm_offset_t)pss.ps_argvstr;
1951 		vsize = pss.ps_nargvstr;
1952 		if (vsize > ARG_MAX)
1953 			return (ENOEXEC);
1954 		size = vsize * sizeof(char *);
1955 		break;
1956 	case PROC_ENV:
1957 		vptr = (vm_offset_t)pss.ps_envstr;
1958 		vsize = pss.ps_nenvstr;
1959 		if (vsize > ARG_MAX)
1960 			return (ENOEXEC);
1961 		size = vsize * sizeof(char *);
1962 		break;
1963 	case PROC_AUX:
1964 		/*
1965 		 * The aux array is just above env array on the stack. Check
1966 		 * that the address is naturally aligned.
1967 		 */
1968 		vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1969 		    * sizeof(char *);
1970 #if __ELF_WORD_SIZE == 64
1971 		if (vptr % sizeof(uint64_t) != 0)
1972 #else
1973 		if (vptr % sizeof(uint32_t) != 0)
1974 #endif
1975 			return (ENOEXEC);
1976 		/*
1977 		 * We count the array size reading the aux vectors from the
1978 		 * stack until AT_NULL vector is returned.  So (to keep the code
1979 		 * simple) we read the process stack twice: the first time here
1980 		 * to find the size and the second time when copying the vectors
1981 		 * to the allocated proc_vector.
1982 		 */
1983 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1984 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1985 			    sizeof(aux))
1986 				return (ENOMEM);
1987 			if (aux.a_type == AT_NULL)
1988 				break;
1989 			ptr += sizeof(aux);
1990 		}
1991 		/*
1992 		 * If the PROC_AUXV_MAX entries are iterated over, and we have
1993 		 * not reached AT_NULL, it is most likely we are reading wrong
1994 		 * data: either the process doesn't have auxv array or data has
1995 		 * been modified. Return the error in this case.
1996 		 */
1997 		if (aux.a_type != AT_NULL)
1998 			return (ENOEXEC);
1999 		vsize = i + 1;
2000 		size = vsize * sizeof(aux);
2001 		break;
2002 	default:
2003 		KASSERT(0, ("Wrong proc vector type: %d", type));
2004 		return (EINVAL); /* In case we are built without INVARIANTS. */
2005 	}
2006 	proc_vector = malloc(size, M_TEMP, M_WAITOK);
2007 	if (proc_readmem(td, p, vptr, proc_vector, size) != size) {
2008 		free(proc_vector, M_TEMP);
2009 		return (ENOMEM);
2010 	}
2011 	*proc_vectorp = proc_vector;
2012 	*vsizep = vsize;
2013 
2014 	return (0);
2015 }
2016 
2017 #define GET_PS_STRINGS_CHUNK_SZ	256	/* Chunk size (bytes) for ps_strings operations. */
2018 
2019 static int
2020 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
2021     enum proc_vector_type type)
2022 {
2023 	size_t done, len, nchr, vsize;
2024 	int error, i;
2025 	char **proc_vector, *sptr;
2026 	char pss_string[GET_PS_STRINGS_CHUNK_SZ];
2027 
2028 	PROC_ASSERT_HELD(p);
2029 
2030 	/*
2031 	 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
2032 	 */
2033 	nchr = 2 * (PATH_MAX + ARG_MAX);
2034 
2035 	error = get_proc_vector(td, p, &proc_vector, &vsize, type);
2036 	if (error != 0)
2037 		return (error);
2038 	for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
2039 		/*
2040 		 * The program may have scribbled into its argv array, e.g. to
2041 		 * remove some arguments.  If that has happened, break out
2042 		 * before trying to read from NULL.
2043 		 */
2044 		if (proc_vector[i] == NULL)
2045 			break;
2046 		for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
2047 			error = proc_read_string(td, p, sptr, pss_string,
2048 			    sizeof(pss_string));
2049 			if (error != 0)
2050 				goto done;
2051 			len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
2052 			if (done + len >= nchr)
2053 				len = nchr - done - 1;
2054 			sbuf_bcat(sb, pss_string, len);
2055 			if (len != GET_PS_STRINGS_CHUNK_SZ)
2056 				break;
2057 			done += GET_PS_STRINGS_CHUNK_SZ;
2058 		}
2059 		sbuf_bcat(sb, "", 1);
2060 		done += len + 1;
2061 	}
2062 done:
2063 	free(proc_vector, M_TEMP);
2064 	return (error);
2065 }
2066 
2067 int
2068 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
2069 {
2070 
2071 	return (get_ps_strings(curthread, p, sb, PROC_ARG));
2072 }
2073 
2074 int
2075 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
2076 {
2077 
2078 	return (get_ps_strings(curthread, p, sb, PROC_ENV));
2079 }
2080 
2081 int
2082 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb)
2083 {
2084 	size_t vsize, size;
2085 	char **auxv;
2086 	int error;
2087 
2088 	error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX);
2089 	if (error == 0) {
2090 #ifdef COMPAT_FREEBSD32
2091 		if (SV_PROC_FLAG(p, SV_ILP32) != 0)
2092 			size = vsize * sizeof(Elf32_Auxinfo);
2093 		else
2094 #endif
2095 			size = vsize * sizeof(Elf_Auxinfo);
2096 		if (sbuf_bcat(sb, auxv, size) != 0)
2097 			error = ENOMEM;
2098 		free(auxv, M_TEMP);
2099 	}
2100 	return (error);
2101 }
2102 
2103 /*
2104  * This sysctl allows a process to retrieve the argument list or process
2105  * title for another process without groping around in the address space
2106  * of the other process.  It also allow a process to set its own "process
2107  * title to a string of its own choice.
2108  */
2109 static int
2110 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
2111 {
2112 	int *name = (int *)arg1;
2113 	u_int namelen = arg2;
2114 	struct pargs *newpa, *pa;
2115 	struct proc *p;
2116 	struct sbuf sb;
2117 	int flags, error = 0, error2;
2118 	pid_t pid;
2119 
2120 	if (namelen != 1)
2121 		return (EINVAL);
2122 
2123 	p = curproc;
2124 	pid = (pid_t)name[0];
2125 	if (pid == -1) {
2126 		pid = p->p_pid;
2127 	}
2128 
2129 	/*
2130 	 * If the query is for this process and it is single-threaded, there
2131 	 * is nobody to modify pargs, thus we can just read.
2132 	 */
2133 	if (pid == p->p_pid && p->p_numthreads == 1 && req->newptr == NULL &&
2134 	    (pa = p->p_args) != NULL)
2135 		return (SYSCTL_OUT(req, pa->ar_args, pa->ar_length));
2136 
2137 	flags = PGET_CANSEE;
2138 	if (req->newptr != NULL)
2139 		flags |= PGET_ISCURRENT;
2140 	error = pget(pid, flags, &p);
2141 	if (error)
2142 		return (error);
2143 
2144 	pa = p->p_args;
2145 	if (pa != NULL) {
2146 		pargs_hold(pa);
2147 		PROC_UNLOCK(p);
2148 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
2149 		pargs_drop(pa);
2150 	} else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
2151 		_PHOLD(p);
2152 		PROC_UNLOCK(p);
2153 		sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2154 		sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2155 		error = proc_getargv(curthread, p, &sb);
2156 		error2 = sbuf_finish(&sb);
2157 		PRELE(p);
2158 		sbuf_delete(&sb);
2159 		if (error == 0 && error2 != 0)
2160 			error = error2;
2161 	} else {
2162 		PROC_UNLOCK(p);
2163 	}
2164 	if (error != 0 || req->newptr == NULL)
2165 		return (error);
2166 
2167 	if (req->newlen > ps_arg_cache_limit - sizeof(struct pargs))
2168 		return (ENOMEM);
2169 
2170 	if (req->newlen == 0) {
2171 		/*
2172 		 * Clear the argument pointer, so that we'll fetch arguments
2173 		 * with proc_getargv() until further notice.
2174 		 */
2175 		newpa = NULL;
2176 	} else {
2177 		newpa = pargs_alloc(req->newlen);
2178 		error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
2179 		if (error != 0) {
2180 			pargs_free(newpa);
2181 			return (error);
2182 		}
2183 	}
2184 	PROC_LOCK(p);
2185 	pa = p->p_args;
2186 	p->p_args = newpa;
2187 	PROC_UNLOCK(p);
2188 	pargs_drop(pa);
2189 	return (0);
2190 }
2191 
2192 /*
2193  * This sysctl allows a process to retrieve environment of another process.
2194  */
2195 static int
2196 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
2197 {
2198 	int *name = (int *)arg1;
2199 	u_int namelen = arg2;
2200 	struct proc *p;
2201 	struct sbuf sb;
2202 	int error, error2;
2203 
2204 	if (namelen != 1)
2205 		return (EINVAL);
2206 
2207 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2208 	if (error != 0)
2209 		return (error);
2210 	if ((p->p_flag & P_SYSTEM) != 0) {
2211 		PRELE(p);
2212 		return (0);
2213 	}
2214 
2215 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2216 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2217 	error = proc_getenvv(curthread, p, &sb);
2218 	error2 = sbuf_finish(&sb);
2219 	PRELE(p);
2220 	sbuf_delete(&sb);
2221 	return (error != 0 ? error : error2);
2222 }
2223 
2224 /*
2225  * This sysctl allows a process to retrieve ELF auxiliary vector of
2226  * another process.
2227  */
2228 static int
2229 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
2230 {
2231 	int *name = (int *)arg1;
2232 	u_int namelen = arg2;
2233 	struct proc *p;
2234 	struct sbuf sb;
2235 	int error, error2;
2236 
2237 	if (namelen != 1)
2238 		return (EINVAL);
2239 
2240 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2241 	if (error != 0)
2242 		return (error);
2243 	if ((p->p_flag & P_SYSTEM) != 0) {
2244 		PRELE(p);
2245 		return (0);
2246 	}
2247 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2248 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2249 	error = proc_getauxv(curthread, p, &sb);
2250 	error2 = sbuf_finish(&sb);
2251 	PRELE(p);
2252 	sbuf_delete(&sb);
2253 	return (error != 0 ? error : error2);
2254 }
2255 
2256 /*
2257  * Look up the canonical executable path running in the specified process.
2258  * It tries to return the same hardlink name as was used for execve(2).
2259  * This allows the programs that modify their behavior based on their progname,
2260  * to operate correctly.
2261  *
2262  * Result is returned in retbuf, it must not be freed, similar to vn_fullpath()
2263  *   calling conventions.
2264  * binname is a pointer to temporary string buffer of length MAXPATHLEN,
2265  *   allocated and freed by caller.
2266  * freebuf should be freed by caller, from the M_TEMP malloc type.
2267  */
2268 int
2269 proc_get_binpath(struct proc *p, char *binname, char **retbuf,
2270     char **freebuf)
2271 {
2272 	struct nameidata nd;
2273 	struct vnode *vp, *dvp;
2274 	size_t freepath_size;
2275 	int error;
2276 	bool do_fullpath;
2277 
2278 	PROC_LOCK_ASSERT(p, MA_OWNED);
2279 
2280 	vp = p->p_textvp;
2281 	if (vp == NULL) {
2282 		PROC_UNLOCK(p);
2283 		*retbuf = "";
2284 		*freebuf = NULL;
2285 		return (0);
2286 	}
2287 	vref(vp);
2288 	dvp = p->p_textdvp;
2289 	if (dvp != NULL)
2290 		vref(dvp);
2291 	if (p->p_binname != NULL)
2292 		strlcpy(binname, p->p_binname, MAXPATHLEN);
2293 	PROC_UNLOCK(p);
2294 
2295 	do_fullpath = true;
2296 	*freebuf = NULL;
2297 	if (dvp != NULL && binname[0] != '\0') {
2298 		freepath_size = MAXPATHLEN;
2299 		if (vn_fullpath_hardlink(vp, dvp, binname, strlen(binname),
2300 		    retbuf, freebuf, &freepath_size) == 0) {
2301 			/*
2302 			 * Recheck the looked up path.  The binary
2303 			 * might have been renamed or replaced, in
2304 			 * which case we should not report old name.
2305 			 */
2306 			NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, *retbuf);
2307 			error = namei(&nd);
2308 			if (error == 0) {
2309 				if (nd.ni_vp == vp)
2310 					do_fullpath = false;
2311 				vrele(nd.ni_vp);
2312 				NDFREE_PNBUF(&nd);
2313 			}
2314 		}
2315 	}
2316 	if (do_fullpath) {
2317 		free(*freebuf, M_TEMP);
2318 		*freebuf = NULL;
2319 		error = vn_fullpath(vp, retbuf, freebuf);
2320 	}
2321 	vrele(vp);
2322 	if (dvp != NULL)
2323 		vrele(dvp);
2324 	return (error);
2325 }
2326 
2327 /*
2328  * This sysctl allows a process to retrieve the path of the executable for
2329  * itself or another process.
2330  */
2331 static int
2332 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
2333 {
2334 	pid_t *pidp = (pid_t *)arg1;
2335 	unsigned int arglen = arg2;
2336 	struct proc *p;
2337 	char *retbuf, *freebuf, *binname;
2338 	int error;
2339 
2340 	if (arglen != 1)
2341 		return (EINVAL);
2342 	binname = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
2343 	binname[0] = '\0';
2344 	if (*pidp == -1) {	/* -1 means this process */
2345 		error = 0;
2346 		p = req->td->td_proc;
2347 		PROC_LOCK(p);
2348 	} else {
2349 		error = pget(*pidp, PGET_CANSEE, &p);
2350 	}
2351 
2352 	if (error == 0)
2353 		error = proc_get_binpath(p, binname, &retbuf, &freebuf);
2354 	free(binname, M_TEMP);
2355 	if (error != 0)
2356 		return (error);
2357 	error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
2358 	free(freebuf, M_TEMP);
2359 	return (error);
2360 }
2361 
2362 static int
2363 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
2364 {
2365 	struct proc *p;
2366 	char *sv_name;
2367 	int *name;
2368 	int namelen;
2369 	int error;
2370 
2371 	namelen = arg2;
2372 	if (namelen != 1)
2373 		return (EINVAL);
2374 
2375 	name = (int *)arg1;
2376 	error = pget((pid_t)name[0], PGET_CANSEE, &p);
2377 	if (error != 0)
2378 		return (error);
2379 	sv_name = p->p_sysent->sv_name;
2380 	PROC_UNLOCK(p);
2381 	return (sysctl_handle_string(oidp, sv_name, 0, req));
2382 }
2383 
2384 #ifdef KINFO_OVMENTRY_SIZE
2385 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
2386 #endif
2387 
2388 #ifdef COMPAT_FREEBSD7
2389 static int
2390 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
2391 {
2392 	vm_map_entry_t entry, tmp_entry;
2393 	unsigned int last_timestamp, namelen;
2394 	char *fullpath, *freepath;
2395 	struct kinfo_ovmentry *kve;
2396 	struct vattr va;
2397 	struct ucred *cred;
2398 	int error, *name;
2399 	struct vnode *vp;
2400 	struct proc *p;
2401 	vm_map_t map;
2402 	struct vmspace *vm;
2403 
2404 	namelen = arg2;
2405 	if (namelen != 1)
2406 		return (EINVAL);
2407 
2408 	name = (int *)arg1;
2409 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2410 	if (error != 0)
2411 		return (error);
2412 	vm = vmspace_acquire_ref(p);
2413 	if (vm == NULL) {
2414 		PRELE(p);
2415 		return (ESRCH);
2416 	}
2417 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2418 
2419 	map = &vm->vm_map;
2420 	vm_map_lock_read(map);
2421 	VM_MAP_ENTRY_FOREACH(entry, map) {
2422 		vm_object_t obj, tobj, lobj;
2423 		vm_offset_t addr;
2424 
2425 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2426 			continue;
2427 
2428 		bzero(kve, sizeof(*kve));
2429 		kve->kve_structsize = sizeof(*kve);
2430 
2431 		kve->kve_private_resident = 0;
2432 		obj = entry->object.vm_object;
2433 		if (obj != NULL) {
2434 			VM_OBJECT_RLOCK(obj);
2435 			if (obj->shadow_count == 1)
2436 				kve->kve_private_resident =
2437 				    obj->resident_page_count;
2438 		}
2439 		kve->kve_resident = 0;
2440 		addr = entry->start;
2441 		while (addr < entry->end) {
2442 			if (pmap_extract(map->pmap, addr))
2443 				kve->kve_resident++;
2444 			addr += PAGE_SIZE;
2445 		}
2446 
2447 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2448 			if (tobj != obj) {
2449 				VM_OBJECT_RLOCK(tobj);
2450 				kve->kve_offset += tobj->backing_object_offset;
2451 			}
2452 			if (lobj != obj)
2453 				VM_OBJECT_RUNLOCK(lobj);
2454 			lobj = tobj;
2455 		}
2456 
2457 		kve->kve_start = (void*)entry->start;
2458 		kve->kve_end = (void*)entry->end;
2459 		kve->kve_offset += (off_t)entry->offset;
2460 
2461 		if (entry->protection & VM_PROT_READ)
2462 			kve->kve_protection |= KVME_PROT_READ;
2463 		if (entry->protection & VM_PROT_WRITE)
2464 			kve->kve_protection |= KVME_PROT_WRITE;
2465 		if (entry->protection & VM_PROT_EXECUTE)
2466 			kve->kve_protection |= KVME_PROT_EXEC;
2467 
2468 		if (entry->eflags & MAP_ENTRY_COW)
2469 			kve->kve_flags |= KVME_FLAG_COW;
2470 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2471 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2472 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2473 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2474 
2475 		last_timestamp = map->timestamp;
2476 		vm_map_unlock_read(map);
2477 
2478 		kve->kve_fileid = 0;
2479 		kve->kve_fsid = 0;
2480 		freepath = NULL;
2481 		fullpath = "";
2482 		if (lobj) {
2483 			kve->kve_type = vm_object_kvme_type(lobj, &vp);
2484 			if (kve->kve_type == KVME_TYPE_MGTDEVICE)
2485 				kve->kve_type = KVME_TYPE_UNKNOWN;
2486 			if (vp != NULL)
2487 				vref(vp);
2488 			if (lobj != obj)
2489 				VM_OBJECT_RUNLOCK(lobj);
2490 
2491 			kve->kve_ref_count = obj->ref_count;
2492 			kve->kve_shadow_count = obj->shadow_count;
2493 			VM_OBJECT_RUNLOCK(obj);
2494 			if (vp != NULL) {
2495 				vn_fullpath(vp, &fullpath, &freepath);
2496 				cred = curthread->td_ucred;
2497 				vn_lock(vp, LK_SHARED | LK_RETRY);
2498 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2499 					kve->kve_fileid = va.va_fileid;
2500 					/* truncate */
2501 					kve->kve_fsid = va.va_fsid;
2502 				}
2503 				vput(vp);
2504 			}
2505 		} else {
2506 			kve->kve_type = KVME_TYPE_NONE;
2507 			kve->kve_ref_count = 0;
2508 			kve->kve_shadow_count = 0;
2509 		}
2510 
2511 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2512 		if (freepath != NULL)
2513 			free(freepath, M_TEMP);
2514 
2515 		error = SYSCTL_OUT(req, kve, sizeof(*kve));
2516 		vm_map_lock_read(map);
2517 		if (error)
2518 			break;
2519 		if (last_timestamp != map->timestamp) {
2520 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2521 			entry = tmp_entry;
2522 		}
2523 	}
2524 	vm_map_unlock_read(map);
2525 	vmspace_free(vm);
2526 	PRELE(p);
2527 	free(kve, M_TEMP);
2528 	return (error);
2529 }
2530 #endif	/* COMPAT_FREEBSD7 */
2531 
2532 #ifdef KINFO_VMENTRY_SIZE
2533 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2534 #endif
2535 
2536 void
2537 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry,
2538     int *resident_count, bool *super)
2539 {
2540 	vm_object_t obj, tobj;
2541 	vm_page_t m, m_adv;
2542 	vm_offset_t addr;
2543 	vm_paddr_t pa;
2544 	vm_pindex_t pi, pi_adv, pindex;
2545 	int incore;
2546 
2547 	*super = false;
2548 	*resident_count = 0;
2549 	if (vmmap_skip_res_cnt)
2550 		return;
2551 
2552 	pa = 0;
2553 	obj = entry->object.vm_object;
2554 	addr = entry->start;
2555 	m_adv = NULL;
2556 	pi = OFF_TO_IDX(entry->offset);
2557 	for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) {
2558 		if (m_adv != NULL) {
2559 			m = m_adv;
2560 		} else {
2561 			pi_adv = atop(entry->end - addr);
2562 			pindex = pi;
2563 			for (tobj = obj;; tobj = tobj->backing_object) {
2564 				m = vm_radix_lookup_ge(&tobj->rtree, pindex);
2565 				if (m != NULL) {
2566 					if (m->pindex == pindex)
2567 						break;
2568 					if (pi_adv > m->pindex - pindex) {
2569 						pi_adv = m->pindex - pindex;
2570 						m_adv = m;
2571 					}
2572 				}
2573 				if (tobj->backing_object == NULL)
2574 					goto next;
2575 				pindex += OFF_TO_IDX(tobj->
2576 				    backing_object_offset);
2577 			}
2578 		}
2579 		m_adv = NULL;
2580 		if (m->psind != 0 && addr + pagesizes[1] <= entry->end &&
2581 		    (addr & (pagesizes[1] - 1)) == 0 && (incore =
2582 		    pmap_mincore(map->pmap, addr, &pa) & MINCORE_SUPER) != 0) {
2583 			*super = true;
2584 			/*
2585 			 * The virtual page might be smaller than the physical
2586 			 * page, so we use the page size reported by the pmap
2587 			 * rather than m->psind.
2588 			 */
2589 			pi_adv = atop(pagesizes[incore >> MINCORE_PSIND_SHIFT]);
2590 		} else {
2591 			/*
2592 			 * We do not test the found page on validity.
2593 			 * Either the page is busy and being paged in,
2594 			 * or it was invalidated.  The first case
2595 			 * should be counted as resident, the second
2596 			 * is not so clear; we do account both.
2597 			 */
2598 			pi_adv = 1;
2599 		}
2600 		*resident_count += pi_adv;
2601 next:;
2602 	}
2603 }
2604 
2605 /*
2606  * Must be called with the process locked and will return unlocked.
2607  */
2608 int
2609 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb, ssize_t maxlen, int flags)
2610 {
2611 	vm_map_entry_t entry, tmp_entry;
2612 	struct vattr va;
2613 	vm_map_t map;
2614 	vm_object_t lobj, nobj, obj, tobj;
2615 	char *fullpath, *freepath;
2616 	struct kinfo_vmentry *kve;
2617 	struct ucred *cred;
2618 	struct vnode *vp;
2619 	struct vmspace *vm;
2620 	vm_offset_t addr;
2621 	unsigned int last_timestamp;
2622 	int error;
2623 	key_t key;
2624 	unsigned short seq;
2625 	bool guard, super;
2626 
2627 	PROC_LOCK_ASSERT(p, MA_OWNED);
2628 
2629 	_PHOLD(p);
2630 	PROC_UNLOCK(p);
2631 	vm = vmspace_acquire_ref(p);
2632 	if (vm == NULL) {
2633 		PRELE(p);
2634 		return (ESRCH);
2635 	}
2636 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK | M_ZERO);
2637 
2638 	error = 0;
2639 	map = &vm->vm_map;
2640 	vm_map_lock_read(map);
2641 	VM_MAP_ENTRY_FOREACH(entry, map) {
2642 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2643 			continue;
2644 
2645 		addr = entry->end;
2646 		bzero(kve, sizeof(*kve));
2647 		obj = entry->object.vm_object;
2648 		if (obj != NULL) {
2649 			if ((obj->flags & OBJ_ANON) != 0)
2650 				kve->kve_obj = (uintptr_t)obj;
2651 
2652 			for (tobj = obj; tobj != NULL;
2653 			    tobj = tobj->backing_object) {
2654 				VM_OBJECT_RLOCK(tobj);
2655 				kve->kve_offset += tobj->backing_object_offset;
2656 				lobj = tobj;
2657 			}
2658 			if (obj->backing_object == NULL)
2659 				kve->kve_private_resident =
2660 				    obj->resident_page_count;
2661 			kern_proc_vmmap_resident(map, entry,
2662 			    &kve->kve_resident, &super);
2663 			if (super)
2664 				kve->kve_flags |= KVME_FLAG_SUPER;
2665 			for (tobj = obj; tobj != NULL; tobj = nobj) {
2666 				nobj = tobj->backing_object;
2667 				if (tobj != obj && tobj != lobj)
2668 					VM_OBJECT_RUNLOCK(tobj);
2669 			}
2670 		} else {
2671 			lobj = NULL;
2672 		}
2673 
2674 		kve->kve_start = entry->start;
2675 		kve->kve_end = entry->end;
2676 		kve->kve_offset += entry->offset;
2677 
2678 		if (entry->protection & VM_PROT_READ)
2679 			kve->kve_protection |= KVME_PROT_READ;
2680 		if (entry->protection & VM_PROT_WRITE)
2681 			kve->kve_protection |= KVME_PROT_WRITE;
2682 		if (entry->protection & VM_PROT_EXECUTE)
2683 			kve->kve_protection |= KVME_PROT_EXEC;
2684 		if (entry->max_protection & VM_PROT_READ)
2685 			kve->kve_protection |= KVME_MAX_PROT_READ;
2686 		if (entry->max_protection & VM_PROT_WRITE)
2687 			kve->kve_protection |= KVME_MAX_PROT_WRITE;
2688 		if (entry->max_protection & VM_PROT_EXECUTE)
2689 			kve->kve_protection |= KVME_MAX_PROT_EXEC;
2690 
2691 		if (entry->eflags & MAP_ENTRY_COW)
2692 			kve->kve_flags |= KVME_FLAG_COW;
2693 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2694 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2695 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2696 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2697 		if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2698 			kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2699 		if (entry->eflags & MAP_ENTRY_USER_WIRED)
2700 			kve->kve_flags |= KVME_FLAG_USER_WIRED;
2701 
2702 		guard = (entry->eflags & MAP_ENTRY_GUARD) != 0;
2703 
2704 		last_timestamp = map->timestamp;
2705 		vm_map_unlock_read(map);
2706 
2707 		freepath = NULL;
2708 		fullpath = "";
2709 		if (lobj != NULL) {
2710 			kve->kve_type = vm_object_kvme_type(lobj, &vp);
2711 			if (vp != NULL)
2712 				vref(vp);
2713 			if (lobj != obj)
2714 				VM_OBJECT_RUNLOCK(lobj);
2715 
2716 			kve->kve_ref_count = obj->ref_count;
2717 			kve->kve_shadow_count = obj->shadow_count;
2718 			if (obj->type == OBJT_DEVICE ||
2719 			    obj->type == OBJT_MGTDEVICE) {
2720 				cdev_pager_get_path(obj, kve->kve_path,
2721 				    sizeof(kve->kve_path));
2722 			}
2723 			VM_OBJECT_RUNLOCK(obj);
2724 			if ((lobj->flags & OBJ_SYSVSHM) != 0) {
2725 				kve->kve_flags |= KVME_FLAG_SYSVSHM;
2726 				shmobjinfo(lobj, &key, &seq);
2727 				kve->kve_vn_fileid = key;
2728 				kve->kve_vn_fsid_freebsd11 = seq;
2729 			}
2730 			if ((lobj->flags & OBJ_POSIXSHM) != 0) {
2731 				kve->kve_flags |= KVME_FLAG_POSIXSHM;
2732 				shm_get_path(lobj, kve->kve_path,
2733 				    sizeof(kve->kve_path));
2734 			}
2735 			if (vp != NULL) {
2736 				vn_fullpath(vp, &fullpath, &freepath);
2737 				kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2738 				cred = curthread->td_ucred;
2739 				vn_lock(vp, LK_SHARED | LK_RETRY);
2740 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2741 					kve->kve_vn_fileid = va.va_fileid;
2742 					kve->kve_vn_fsid = va.va_fsid;
2743 					kve->kve_vn_fsid_freebsd11 =
2744 					    kve->kve_vn_fsid; /* truncate */
2745 					kve->kve_vn_mode =
2746 					    MAKEIMODE(va.va_type, va.va_mode);
2747 					kve->kve_vn_size = va.va_size;
2748 					kve->kve_vn_rdev = va.va_rdev;
2749 					kve->kve_vn_rdev_freebsd11 =
2750 					    kve->kve_vn_rdev; /* truncate */
2751 					kve->kve_status = KF_ATTR_VALID;
2752 				}
2753 				vput(vp);
2754 				strlcpy(kve->kve_path, fullpath, sizeof(
2755 				    kve->kve_path));
2756 				free(freepath, M_TEMP);
2757 			}
2758 		} else {
2759 			kve->kve_type = guard ? KVME_TYPE_GUARD :
2760 			    KVME_TYPE_NONE;
2761 			kve->kve_ref_count = 0;
2762 			kve->kve_shadow_count = 0;
2763 		}
2764 
2765 		/* Pack record size down */
2766 		if ((flags & KERN_VMMAP_PACK_KINFO) != 0)
2767 			kve->kve_structsize =
2768 			    offsetof(struct kinfo_vmentry, kve_path) +
2769 			    strlen(kve->kve_path) + 1;
2770 		else
2771 			kve->kve_structsize = sizeof(*kve);
2772 		kve->kve_structsize = roundup(kve->kve_structsize,
2773 		    sizeof(uint64_t));
2774 
2775 		/* Halt filling and truncate rather than exceeding maxlen */
2776 		if (maxlen != -1 && maxlen < kve->kve_structsize) {
2777 			error = 0;
2778 			vm_map_lock_read(map);
2779 			break;
2780 		} else if (maxlen != -1)
2781 			maxlen -= kve->kve_structsize;
2782 
2783 		if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0)
2784 			error = ENOMEM;
2785 		vm_map_lock_read(map);
2786 		if (error != 0)
2787 			break;
2788 		if (last_timestamp != map->timestamp) {
2789 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2790 			entry = tmp_entry;
2791 		}
2792 	}
2793 	vm_map_unlock_read(map);
2794 	vmspace_free(vm);
2795 	PRELE(p);
2796 	free(kve, M_TEMP);
2797 	return (error);
2798 }
2799 
2800 static int
2801 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2802 {
2803 	struct proc *p;
2804 	struct sbuf sb;
2805 	u_int namelen;
2806 	int error, error2, *name;
2807 
2808 	namelen = arg2;
2809 	if (namelen != 1)
2810 		return (EINVAL);
2811 
2812 	name = (int *)arg1;
2813 	sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req);
2814 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2815 	error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
2816 	if (error != 0) {
2817 		sbuf_delete(&sb);
2818 		return (error);
2819 	}
2820 	error = kern_proc_vmmap_out(p, &sb, -1, KERN_VMMAP_PACK_KINFO);
2821 	error2 = sbuf_finish(&sb);
2822 	sbuf_delete(&sb);
2823 	return (error != 0 ? error : error2);
2824 }
2825 
2826 #if defined(STACK) || defined(DDB)
2827 static int
2828 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2829 {
2830 	struct kinfo_kstack *kkstp;
2831 	int error, i, *name, numthreads;
2832 	lwpid_t *lwpidarray;
2833 	struct thread *td;
2834 	struct stack *st;
2835 	struct sbuf sb;
2836 	struct proc *p;
2837 	u_int namelen;
2838 
2839 	namelen = arg2;
2840 	if (namelen != 1)
2841 		return (EINVAL);
2842 
2843 	name = (int *)arg1;
2844 	error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2845 	if (error != 0)
2846 		return (error);
2847 
2848 	kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2849 	st = stack_create(M_WAITOK);
2850 
2851 	lwpidarray = NULL;
2852 	PROC_LOCK(p);
2853 	do {
2854 		if (lwpidarray != NULL) {
2855 			free(lwpidarray, M_TEMP);
2856 			lwpidarray = NULL;
2857 		}
2858 		numthreads = p->p_numthreads;
2859 		PROC_UNLOCK(p);
2860 		lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2861 		    M_WAITOK | M_ZERO);
2862 		PROC_LOCK(p);
2863 	} while (numthreads < p->p_numthreads);
2864 
2865 	/*
2866 	 * XXXRW: During the below loop, execve(2) and countless other sorts
2867 	 * of changes could have taken place.  Should we check to see if the
2868 	 * vmspace has been replaced, or the like, in order to prevent
2869 	 * giving a snapshot that spans, say, execve(2), with some threads
2870 	 * before and some after?  Among other things, the credentials could
2871 	 * have changed, in which case the right to extract debug info might
2872 	 * no longer be assured.
2873 	 */
2874 	i = 0;
2875 	FOREACH_THREAD_IN_PROC(p, td) {
2876 		KASSERT(i < numthreads,
2877 		    ("sysctl_kern_proc_kstack: numthreads"));
2878 		lwpidarray[i] = td->td_tid;
2879 		i++;
2880 	}
2881 	PROC_UNLOCK(p);
2882 	numthreads = i;
2883 	for (i = 0; i < numthreads; i++) {
2884 		td = tdfind(lwpidarray[i], p->p_pid);
2885 		if (td == NULL) {
2886 			continue;
2887 		}
2888 		bzero(kkstp, sizeof(*kkstp));
2889 		(void)sbuf_new(&sb, kkstp->kkst_trace,
2890 		    sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2891 		thread_lock(td);
2892 		kkstp->kkst_tid = td->td_tid;
2893 		if (stack_save_td(st, td) == 0)
2894 			kkstp->kkst_state = KKST_STATE_STACKOK;
2895 		else
2896 			kkstp->kkst_state = KKST_STATE_RUNNING;
2897 		thread_unlock(td);
2898 		PROC_UNLOCK(p);
2899 		stack_sbuf_print(&sb, st);
2900 		sbuf_finish(&sb);
2901 		sbuf_delete(&sb);
2902 		error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2903 		if (error)
2904 			break;
2905 	}
2906 	PRELE(p);
2907 	if (lwpidarray != NULL)
2908 		free(lwpidarray, M_TEMP);
2909 	stack_destroy(st);
2910 	free(kkstp, M_TEMP);
2911 	return (error);
2912 }
2913 #endif
2914 
2915 /*
2916  * This sysctl allows a process to retrieve the full list of groups from
2917  * itself or another process.
2918  */
2919 static int
2920 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2921 {
2922 	pid_t *pidp = (pid_t *)arg1;
2923 	unsigned int arglen = arg2;
2924 	struct proc *p;
2925 	struct ucred *cred;
2926 	int error;
2927 
2928 	if (arglen != 1)
2929 		return (EINVAL);
2930 	if (*pidp == -1) {	/* -1 means this process */
2931 		p = req->td->td_proc;
2932 		PROC_LOCK(p);
2933 	} else {
2934 		error = pget(*pidp, PGET_CANSEE, &p);
2935 		if (error != 0)
2936 			return (error);
2937 	}
2938 
2939 	cred = crhold(p->p_ucred);
2940 	PROC_UNLOCK(p);
2941 
2942 	error = SYSCTL_OUT(req, cred->cr_groups,
2943 	    cred->cr_ngroups * sizeof(gid_t));
2944 	crfree(cred);
2945 	return (error);
2946 }
2947 
2948 /*
2949  * This sysctl allows a process to retrieve or/and set the resource limit for
2950  * another process.
2951  */
2952 static int
2953 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2954 {
2955 	int *name = (int *)arg1;
2956 	u_int namelen = arg2;
2957 	struct rlimit rlim;
2958 	struct proc *p;
2959 	u_int which;
2960 	int flags, error;
2961 
2962 	if (namelen != 2)
2963 		return (EINVAL);
2964 
2965 	which = (u_int)name[1];
2966 	if (which >= RLIM_NLIMITS)
2967 		return (EINVAL);
2968 
2969 	if (req->newptr != NULL && req->newlen != sizeof(rlim))
2970 		return (EINVAL);
2971 
2972 	flags = PGET_HOLD | PGET_NOTWEXIT;
2973 	if (req->newptr != NULL)
2974 		flags |= PGET_CANDEBUG;
2975 	else
2976 		flags |= PGET_CANSEE;
2977 	error = pget((pid_t)name[0], flags, &p);
2978 	if (error != 0)
2979 		return (error);
2980 
2981 	/*
2982 	 * Retrieve limit.
2983 	 */
2984 	if (req->oldptr != NULL) {
2985 		PROC_LOCK(p);
2986 		lim_rlimit_proc(p, which, &rlim);
2987 		PROC_UNLOCK(p);
2988 	}
2989 	error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
2990 	if (error != 0)
2991 		goto errout;
2992 
2993 	/*
2994 	 * Set limit.
2995 	 */
2996 	if (req->newptr != NULL) {
2997 		error = SYSCTL_IN(req, &rlim, sizeof(rlim));
2998 		if (error == 0)
2999 			error = kern_proc_setrlimit(curthread, p, which, &rlim);
3000 	}
3001 
3002 errout:
3003 	PRELE(p);
3004 	return (error);
3005 }
3006 
3007 /*
3008  * This sysctl allows a process to retrieve ps_strings structure location of
3009  * another process.
3010  */
3011 static int
3012 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
3013 {
3014 	int *name = (int *)arg1;
3015 	u_int namelen = arg2;
3016 	struct proc *p;
3017 	vm_offset_t ps_strings;
3018 	int error;
3019 #ifdef COMPAT_FREEBSD32
3020 	uint32_t ps_strings32;
3021 #endif
3022 
3023 	if (namelen != 1)
3024 		return (EINVAL);
3025 
3026 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
3027 	if (error != 0)
3028 		return (error);
3029 #ifdef COMPAT_FREEBSD32
3030 	if ((req->flags & SCTL_MASK32) != 0) {
3031 		/*
3032 		 * We return 0 if the 32 bit emulation request is for a 64 bit
3033 		 * process.
3034 		 */
3035 		ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
3036 		    PTROUT(PROC_PS_STRINGS(p)) : 0;
3037 		PROC_UNLOCK(p);
3038 		error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
3039 		return (error);
3040 	}
3041 #endif
3042 	ps_strings = PROC_PS_STRINGS(p);
3043 	PROC_UNLOCK(p);
3044 	error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
3045 	return (error);
3046 }
3047 
3048 /*
3049  * This sysctl allows a process to retrieve umask of another process.
3050  */
3051 static int
3052 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
3053 {
3054 	int *name = (int *)arg1;
3055 	u_int namelen = arg2;
3056 	struct proc *p;
3057 	int error;
3058 	u_short cmask;
3059 	pid_t pid;
3060 
3061 	if (namelen != 1)
3062 		return (EINVAL);
3063 
3064 	pid = (pid_t)name[0];
3065 	p = curproc;
3066 	if (pid == p->p_pid || pid == 0) {
3067 		cmask = p->p_pd->pd_cmask;
3068 		goto out;
3069 	}
3070 
3071 	error = pget(pid, PGET_WANTREAD, &p);
3072 	if (error != 0)
3073 		return (error);
3074 
3075 	cmask = p->p_pd->pd_cmask;
3076 	PRELE(p);
3077 out:
3078 	error = SYSCTL_OUT(req, &cmask, sizeof(cmask));
3079 	return (error);
3080 }
3081 
3082 /*
3083  * This sysctl allows a process to set and retrieve binary osreldate of
3084  * another process.
3085  */
3086 static int
3087 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
3088 {
3089 	int *name = (int *)arg1;
3090 	u_int namelen = arg2;
3091 	struct proc *p;
3092 	int flags, error, osrel;
3093 
3094 	if (namelen != 1)
3095 		return (EINVAL);
3096 
3097 	if (req->newptr != NULL && req->newlen != sizeof(osrel))
3098 		return (EINVAL);
3099 
3100 	flags = PGET_HOLD | PGET_NOTWEXIT;
3101 	if (req->newptr != NULL)
3102 		flags |= PGET_CANDEBUG;
3103 	else
3104 		flags |= PGET_CANSEE;
3105 	error = pget((pid_t)name[0], flags, &p);
3106 	if (error != 0)
3107 		return (error);
3108 
3109 	error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
3110 	if (error != 0)
3111 		goto errout;
3112 
3113 	if (req->newptr != NULL) {
3114 		error = SYSCTL_IN(req, &osrel, sizeof(osrel));
3115 		if (error != 0)
3116 			goto errout;
3117 		if (osrel < 0) {
3118 			error = EINVAL;
3119 			goto errout;
3120 		}
3121 		p->p_osrel = osrel;
3122 	}
3123 errout:
3124 	PRELE(p);
3125 	return (error);
3126 }
3127 
3128 static int
3129 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS)
3130 {
3131 	int *name = (int *)arg1;
3132 	u_int namelen = arg2;
3133 	struct proc *p;
3134 	struct kinfo_sigtramp kst;
3135 	const struct sysentvec *sv;
3136 	int error;
3137 #ifdef COMPAT_FREEBSD32
3138 	struct kinfo_sigtramp32 kst32;
3139 #endif
3140 
3141 	if (namelen != 1)
3142 		return (EINVAL);
3143 
3144 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
3145 	if (error != 0)
3146 		return (error);
3147 	sv = p->p_sysent;
3148 #ifdef COMPAT_FREEBSD32
3149 	if ((req->flags & SCTL_MASK32) != 0) {
3150 		bzero(&kst32, sizeof(kst32));
3151 		if (SV_PROC_FLAG(p, SV_ILP32)) {
3152 			if (PROC_HAS_SHP(p)) {
3153 				kst32.ksigtramp_start = PROC_SIGCODE(p);
3154 				kst32.ksigtramp_end = kst32.ksigtramp_start +
3155 				    ((sv->sv_flags & SV_DSO_SIG) == 0 ?
3156 				    *sv->sv_szsigcode :
3157 				    (uintptr_t)sv->sv_szsigcode);
3158 			} else {
3159 				kst32.ksigtramp_start = PROC_PS_STRINGS(p) -
3160 				    *sv->sv_szsigcode;
3161 				kst32.ksigtramp_end = PROC_PS_STRINGS(p);
3162 			}
3163 		}
3164 		PROC_UNLOCK(p);
3165 		error = SYSCTL_OUT(req, &kst32, sizeof(kst32));
3166 		return (error);
3167 	}
3168 #endif
3169 	bzero(&kst, sizeof(kst));
3170 	if (PROC_HAS_SHP(p)) {
3171 		kst.ksigtramp_start = (char *)PROC_SIGCODE(p);
3172 		kst.ksigtramp_end = (char *)kst.ksigtramp_start +
3173 		    ((sv->sv_flags & SV_DSO_SIG) == 0 ? *sv->sv_szsigcode :
3174 		    (uintptr_t)sv->sv_szsigcode);
3175 	} else {
3176 		kst.ksigtramp_start = (char *)PROC_PS_STRINGS(p) -
3177 		    *sv->sv_szsigcode;
3178 		kst.ksigtramp_end = (char *)PROC_PS_STRINGS(p);
3179 	}
3180 	PROC_UNLOCK(p);
3181 	error = SYSCTL_OUT(req, &kst, sizeof(kst));
3182 	return (error);
3183 }
3184 
3185 static int
3186 sysctl_kern_proc_sigfastblk(SYSCTL_HANDLER_ARGS)
3187 {
3188 	int *name = (int *)arg1;
3189 	u_int namelen = arg2;
3190 	pid_t pid;
3191 	struct proc *p;
3192 	struct thread *td1;
3193 	uintptr_t addr;
3194 #ifdef COMPAT_FREEBSD32
3195 	uint32_t addr32;
3196 #endif
3197 	int error;
3198 
3199 	if (namelen != 1 || req->newptr != NULL)
3200 		return (EINVAL);
3201 
3202 	pid = (pid_t)name[0];
3203 	error = pget(pid, PGET_HOLD | PGET_NOTWEXIT | PGET_CANDEBUG, &p);
3204 	if (error != 0)
3205 		return (error);
3206 
3207 	PROC_LOCK(p);
3208 #ifdef COMPAT_FREEBSD32
3209 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3210 		if (!SV_PROC_FLAG(p, SV_ILP32)) {
3211 			error = EINVAL;
3212 			goto errlocked;
3213 		}
3214 	}
3215 #endif
3216 	if (pid <= PID_MAX) {
3217 		td1 = FIRST_THREAD_IN_PROC(p);
3218 	} else {
3219 		FOREACH_THREAD_IN_PROC(p, td1) {
3220 			if (td1->td_tid == pid)
3221 				break;
3222 		}
3223 	}
3224 	if (td1 == NULL) {
3225 		error = ESRCH;
3226 		goto errlocked;
3227 	}
3228 	/*
3229 	 * The access to the private thread flags.  It is fine as far
3230 	 * as no out-of-thin-air values are read from td_pflags, and
3231 	 * usermode read of the td_sigblock_ptr is racy inherently,
3232 	 * since target process might have already changed it
3233 	 * meantime.
3234 	 */
3235 	if ((td1->td_pflags & TDP_SIGFASTBLOCK) != 0)
3236 		addr = (uintptr_t)td1->td_sigblock_ptr;
3237 	else
3238 		error = ENOTTY;
3239 
3240 errlocked:
3241 	_PRELE(p);
3242 	PROC_UNLOCK(p);
3243 	if (error != 0)
3244 		return (error);
3245 
3246 #ifdef COMPAT_FREEBSD32
3247 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3248 		addr32 = addr;
3249 		error = SYSCTL_OUT(req, &addr32, sizeof(addr32));
3250 	} else
3251 #endif
3252 		error = SYSCTL_OUT(req, &addr, sizeof(addr));
3253 	return (error);
3254 }
3255 
3256 static int
3257 sysctl_kern_proc_vm_layout(SYSCTL_HANDLER_ARGS)
3258 {
3259 	struct kinfo_vm_layout kvm;
3260 	struct proc *p;
3261 	struct vmspace *vmspace;
3262 	int error, *name;
3263 
3264 	name = (int *)arg1;
3265 	if ((u_int)arg2 != 1)
3266 		return (EINVAL);
3267 
3268 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
3269 	if (error != 0)
3270 		return (error);
3271 #ifdef COMPAT_FREEBSD32
3272 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3273 		if (!SV_PROC_FLAG(p, SV_ILP32)) {
3274 			PROC_UNLOCK(p);
3275 			return (EINVAL);
3276 		}
3277 	}
3278 #endif
3279 	vmspace = vmspace_acquire_ref(p);
3280 	PROC_UNLOCK(p);
3281 
3282 	memset(&kvm, 0, sizeof(kvm));
3283 	kvm.kvm_min_user_addr = vm_map_min(&vmspace->vm_map);
3284 	kvm.kvm_max_user_addr = vm_map_max(&vmspace->vm_map);
3285 	kvm.kvm_text_addr = (uintptr_t)vmspace->vm_taddr;
3286 	kvm.kvm_text_size = vmspace->vm_tsize;
3287 	kvm.kvm_data_addr = (uintptr_t)vmspace->vm_daddr;
3288 	kvm.kvm_data_size = vmspace->vm_dsize;
3289 	kvm.kvm_stack_addr = (uintptr_t)vmspace->vm_maxsaddr;
3290 	kvm.kvm_stack_size = vmspace->vm_ssize;
3291 	kvm.kvm_shp_addr = vmspace->vm_shp_base;
3292 	kvm.kvm_shp_size = p->p_sysent->sv_shared_page_len;
3293 	if ((vmspace->vm_map.flags & MAP_WIREFUTURE) != 0)
3294 		kvm.kvm_map_flags |= KMAP_FLAG_WIREFUTURE;
3295 	if ((vmspace->vm_map.flags & MAP_ASLR) != 0)
3296 		kvm.kvm_map_flags |= KMAP_FLAG_ASLR;
3297 	if ((vmspace->vm_map.flags & MAP_ASLR_IGNSTART) != 0)
3298 		kvm.kvm_map_flags |= KMAP_FLAG_ASLR_IGNSTART;
3299 	if ((vmspace->vm_map.flags & MAP_WXORX) != 0)
3300 		kvm.kvm_map_flags |= KMAP_FLAG_WXORX;
3301 	if ((vmspace->vm_map.flags & MAP_ASLR_STACK) != 0)
3302 		kvm.kvm_map_flags |= KMAP_FLAG_ASLR_STACK;
3303 	if (vmspace->vm_shp_base != p->p_sysent->sv_shared_page_base &&
3304 	    PROC_HAS_SHP(p))
3305 		kvm.kvm_map_flags |= KMAP_FLAG_ASLR_SHARED_PAGE;
3306 
3307 #ifdef COMPAT_FREEBSD32
3308 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3309 		struct kinfo_vm_layout32 kvm32;
3310 
3311 		memset(&kvm32, 0, sizeof(kvm32));
3312 		kvm32.kvm_min_user_addr = (uint32_t)kvm.kvm_min_user_addr;
3313 		kvm32.kvm_max_user_addr = (uint32_t)kvm.kvm_max_user_addr;
3314 		kvm32.kvm_text_addr = (uint32_t)kvm.kvm_text_addr;
3315 		kvm32.kvm_text_size = (uint32_t)kvm.kvm_text_size;
3316 		kvm32.kvm_data_addr = (uint32_t)kvm.kvm_data_addr;
3317 		kvm32.kvm_data_size = (uint32_t)kvm.kvm_data_size;
3318 		kvm32.kvm_stack_addr = (uint32_t)kvm.kvm_stack_addr;
3319 		kvm32.kvm_stack_size = (uint32_t)kvm.kvm_stack_size;
3320 		kvm32.kvm_shp_addr = (uint32_t)kvm.kvm_shp_addr;
3321 		kvm32.kvm_shp_size = (uint32_t)kvm.kvm_shp_size;
3322 		kvm32.kvm_map_flags = kvm.kvm_map_flags;
3323 		error = SYSCTL_OUT(req, &kvm32, sizeof(kvm32));
3324 		goto out;
3325 	}
3326 #endif
3327 
3328 	error = SYSCTL_OUT(req, &kvm, sizeof(kvm));
3329 #ifdef COMPAT_FREEBSD32
3330 out:
3331 #endif
3332 	vmspace_free(vmspace);
3333 	return (error);
3334 }
3335 
3336 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,  0,
3337     "Process table");
3338 
3339 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
3340 	CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
3341 	"Return entire process table");
3342 
3343 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3344 	sysctl_kern_proc, "Process table");
3345 
3346 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
3347 	sysctl_kern_proc, "Process table");
3348 
3349 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3350 	sysctl_kern_proc, "Process table");
3351 
3352 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
3353 	CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3354 
3355 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
3356 	sysctl_kern_proc, "Process table");
3357 
3358 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3359 	sysctl_kern_proc, "Process table");
3360 
3361 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3362 	sysctl_kern_proc, "Process table");
3363 
3364 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3365 	sysctl_kern_proc, "Process table");
3366 
3367 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
3368 	sysctl_kern_proc, "Return process table, no threads");
3369 
3370 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
3371 	CTLFLAG_RW | CTLFLAG_CAPWR | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
3372 	sysctl_kern_proc_args, "Process argument list");
3373 
3374 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
3375 	sysctl_kern_proc_env, "Process environment");
3376 
3377 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
3378 	CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
3379 
3380 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
3381 	CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
3382 
3383 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
3384 	CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
3385 	"Process syscall vector name (ABI type)");
3386 
3387 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
3388 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3389 
3390 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
3391 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3392 
3393 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
3394 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3395 
3396 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
3397 	sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3398 
3399 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
3400 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3401 
3402 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
3403 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3404 
3405 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
3406 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3407 
3408 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
3409 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3410 
3411 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
3412 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
3413 	"Return process table, including threads");
3414 
3415 #ifdef COMPAT_FREEBSD7
3416 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
3417 	CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
3418 #endif
3419 
3420 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
3421 	CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
3422 
3423 #if defined(STACK) || defined(DDB)
3424 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
3425 	CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
3426 #endif
3427 
3428 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
3429 	CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
3430 
3431 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
3432 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
3433 	"Process resource limits");
3434 
3435 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
3436 	CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
3437 	"Process ps_strings location");
3438 
3439 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
3440 	CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
3441 
3442 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
3443 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
3444 	"Process binary osreldate");
3445 
3446 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD |
3447 	CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp,
3448 	"Process signal trampoline location");
3449 
3450 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGFASTBLK, sigfastblk, CTLFLAG_RD |
3451 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_sigfastblk,
3452 	"Thread sigfastblock address");
3453 
3454 static SYSCTL_NODE(_kern_proc, KERN_PROC_VM_LAYOUT, vm_layout, CTLFLAG_RD |
3455 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_vm_layout,
3456 	"Process virtual address space layout info");
3457 
3458 static struct sx stop_all_proc_blocker;
3459 SX_SYSINIT(stop_all_proc_blocker, &stop_all_proc_blocker, "sapblk");
3460 
3461 bool
3462 stop_all_proc_block(void)
3463 {
3464 	return (sx_xlock_sig(&stop_all_proc_blocker) == 0);
3465 }
3466 
3467 void
3468 stop_all_proc_unblock(void)
3469 {
3470 	sx_xunlock(&stop_all_proc_blocker);
3471 }
3472 
3473 int allproc_gen;
3474 
3475 /*
3476  * stop_all_proc() purpose is to stop all process which have usermode,
3477  * except current process for obvious reasons.  This makes it somewhat
3478  * unreliable when invoked from multithreaded process.  The service
3479  * must not be user-callable anyway.
3480  */
3481 void
3482 stop_all_proc(void)
3483 {
3484 	struct proc *cp, *p;
3485 	int r, gen;
3486 	bool restart, seen_stopped, seen_exiting, stopped_some;
3487 
3488 	if (!stop_all_proc_block())
3489 		return;
3490 
3491 	cp = curproc;
3492 allproc_loop:
3493 	sx_xlock(&allproc_lock);
3494 	gen = allproc_gen;
3495 	seen_exiting = seen_stopped = stopped_some = restart = false;
3496 	LIST_REMOVE(cp, p_list);
3497 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3498 	for (;;) {
3499 		p = LIST_NEXT(cp, p_list);
3500 		if (p == NULL)
3501 			break;
3502 		LIST_REMOVE(cp, p_list);
3503 		LIST_INSERT_AFTER(p, cp, p_list);
3504 		PROC_LOCK(p);
3505 		if ((p->p_flag & (P_KPROC | P_SYSTEM | P_TOTAL_STOP |
3506 		    P_STOPPED_SIG)) != 0) {
3507 			PROC_UNLOCK(p);
3508 			continue;
3509 		}
3510 		if ((p->p_flag2 & P2_WEXIT) != 0) {
3511 			seen_exiting = true;
3512 			PROC_UNLOCK(p);
3513 			continue;
3514 		}
3515 		if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) {
3516 			/*
3517 			 * Stopped processes are tolerated when there
3518 			 * are no other processes which might continue
3519 			 * them.  P_STOPPED_SINGLE but not
3520 			 * P_TOTAL_STOP process still has at least one
3521 			 * thread running.
3522 			 */
3523 			seen_stopped = true;
3524 			PROC_UNLOCK(p);
3525 			continue;
3526 		}
3527 		if ((p->p_flag & P_TRACED) != 0) {
3528 			/*
3529 			 * thread_single() below cannot stop traced p,
3530 			 * so skip it.  OTOH, we cannot require
3531 			 * restart because debugger might be either
3532 			 * already stopped or traced as well.
3533 			 */
3534 			PROC_UNLOCK(p);
3535 			continue;
3536 		}
3537 		sx_xunlock(&allproc_lock);
3538 		_PHOLD(p);
3539 		r = thread_single(p, SINGLE_ALLPROC);
3540 		if (r != 0)
3541 			restart = true;
3542 		else
3543 			stopped_some = true;
3544 		_PRELE(p);
3545 		PROC_UNLOCK(p);
3546 		sx_xlock(&allproc_lock);
3547 	}
3548 	/* Catch forked children we did not see in iteration. */
3549 	if (gen != allproc_gen)
3550 		restart = true;
3551 	sx_xunlock(&allproc_lock);
3552 	if (restart || stopped_some || seen_exiting || seen_stopped) {
3553 		kern_yield(PRI_USER);
3554 		goto allproc_loop;
3555 	}
3556 }
3557 
3558 void
3559 resume_all_proc(void)
3560 {
3561 	struct proc *cp, *p;
3562 
3563 	cp = curproc;
3564 	sx_xlock(&allproc_lock);
3565 again:
3566 	LIST_REMOVE(cp, p_list);
3567 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3568 	for (;;) {
3569 		p = LIST_NEXT(cp, p_list);
3570 		if (p == NULL)
3571 			break;
3572 		LIST_REMOVE(cp, p_list);
3573 		LIST_INSERT_AFTER(p, cp, p_list);
3574 		PROC_LOCK(p);
3575 		if ((p->p_flag & P_TOTAL_STOP) != 0) {
3576 			sx_xunlock(&allproc_lock);
3577 			_PHOLD(p);
3578 			thread_single_end(p, SINGLE_ALLPROC);
3579 			_PRELE(p);
3580 			PROC_UNLOCK(p);
3581 			sx_xlock(&allproc_lock);
3582 		} else {
3583 			PROC_UNLOCK(p);
3584 		}
3585 	}
3586 	/*  Did the loop above missed any stopped process ? */
3587 	FOREACH_PROC_IN_SYSTEM(p) {
3588 		/* No need for proc lock. */
3589 		if ((p->p_flag & P_TOTAL_STOP) != 0)
3590 			goto again;
3591 	}
3592 	sx_xunlock(&allproc_lock);
3593 
3594 	stop_all_proc_unblock();
3595 }
3596 
3597 /* #define	TOTAL_STOP_DEBUG	1 */
3598 #ifdef TOTAL_STOP_DEBUG
3599 volatile static int ap_resume;
3600 #include <sys/mount.h>
3601 
3602 static int
3603 sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS)
3604 {
3605 	int error, val;
3606 
3607 	val = 0;
3608 	ap_resume = 0;
3609 	error = sysctl_handle_int(oidp, &val, 0, req);
3610 	if (error != 0 || req->newptr == NULL)
3611 		return (error);
3612 	if (val != 0) {
3613 		stop_all_proc();
3614 		syncer_suspend();
3615 		while (ap_resume == 0)
3616 			;
3617 		syncer_resume();
3618 		resume_all_proc();
3619 	}
3620 	return (0);
3621 }
3622 
3623 SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW |
3624     CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0,
3625     sysctl_debug_stop_all_proc, "I",
3626     "");
3627 #endif
3628