xref: /freebsd/sys/kern/kern_proc.c (revision 5a91fa5a7656c99e527fe7e6f6bf6bd9e85ab589)
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/imgact.h>
47 #include <sys/ipc.h>
48 #include <sys/jail.h>
49 #include <sys/kernel.h>
50 #include <sys/limits.h>
51 #include <sys/lock.h>
52 #include <sys/loginclass.h>
53 #include <sys/malloc.h>
54 #include <sys/mman.h>
55 #include <sys/mount.h>
56 #include <sys/mutex.h>
57 #include <sys/namei.h>
58 #include <sys/proc.h>
59 #include <sys/ptrace.h>
60 #include <sys/refcount.h>
61 #include <sys/resourcevar.h>
62 #include <sys/rwlock.h>
63 #include <sys/sbuf.h>
64 #include <sys/sysent.h>
65 #include <sys/sched.h>
66 #include <sys/shm.h>
67 #include <sys/smp.h>
68 #include <sys/stack.h>
69 #include <sys/stat.h>
70 #include <sys/dtrace_bsd.h>
71 #include <sys/sysctl.h>
72 #include <sys/filedesc.h>
73 #include <sys/tty.h>
74 #include <sys/signalvar.h>
75 #include <sys/sdt.h>
76 #include <sys/sx.h>
77 #include <sys/user.h>
78 #include <sys/vnode.h>
79 #include <sys/wait.h>
80 #ifdef KTRACE
81 #include <sys/ktrace.h>
82 #endif
83 
84 #ifdef DDB
85 #include <ddb/ddb.h>
86 #endif
87 
88 #include <vm/vm.h>
89 #include <vm/vm_param.h>
90 #include <vm/vm_extern.h>
91 #include <vm/pmap.h>
92 #include <vm/vm_map.h>
93 #include <vm/vm_object.h>
94 #include <vm/vm_page.h>
95 #include <vm/vm_pager.h>
96 #include <vm/vm_radix.h>
97 #include <vm/uma.h>
98 
99 #include <fs/devfs/devfs.h>
100 
101 #ifdef COMPAT_FREEBSD32
102 #include <compat/freebsd32/freebsd32.h>
103 #include <compat/freebsd32/freebsd32_util.h>
104 #endif
105 
106 SDT_PROVIDER_DEFINE(proc);
107 
108 MALLOC_DEFINE(M_SESSION, "session", "session header");
109 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
110 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
111 
112 static void doenterpgrp(struct proc *, struct pgrp *);
113 static void orphanpg(struct pgrp *pg);
114 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp);
115 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp);
116 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp,
117     int preferthread);
118 static void pgdelete(struct pgrp *);
119 static int pgrp_init(void *mem, int size, int flags);
120 static int proc_ctor(void *mem, int size, void *arg, int flags);
121 static void proc_dtor(void *mem, int size, void *arg);
122 static int proc_init(void *mem, int size, int flags);
123 static void proc_fini(void *mem, int size);
124 static void pargs_free(struct pargs *pa);
125 
126 /*
127  * Other process lists
128  */
129 struct pidhashhead *pidhashtbl = NULL;
130 struct sx *pidhashtbl_lock;
131 u_long pidhash;
132 u_long pidhashlock;
133 struct pgrphashhead *pgrphashtbl;
134 u_long pgrphash;
135 struct proclist allproc = LIST_HEAD_INITIALIZER(allproc);
136 struct sx __exclusive_cache_line allproc_lock;
137 struct sx __exclusive_cache_line proctree_lock;
138 struct mtx __exclusive_cache_line ppeers_lock;
139 struct mtx __exclusive_cache_line procid_lock;
140 uma_zone_t proc_zone;
141 uma_zone_t pgrp_zone;
142 
143 /*
144  * The offset of various fields in struct proc and struct thread.
145  * These are used by kernel debuggers to enumerate kernel threads and
146  * processes.
147  */
148 const int proc_off_p_pid = offsetof(struct proc, p_pid);
149 const int proc_off_p_comm = offsetof(struct proc, p_comm);
150 const int proc_off_p_list = offsetof(struct proc, p_list);
151 const int proc_off_p_hash = offsetof(struct proc, p_hash);
152 const int proc_off_p_threads = offsetof(struct proc, p_threads);
153 const int thread_off_td_tid = offsetof(struct thread, td_tid);
154 const int thread_off_td_name = offsetof(struct thread, td_name);
155 const int thread_off_td_oncpu = offsetof(struct thread, td_oncpu);
156 const int thread_off_td_pcb = offsetof(struct thread, td_pcb);
157 const int thread_off_td_plist = offsetof(struct thread, td_plist);
158 
159 EVENTHANDLER_LIST_DEFINE(process_ctor);
160 EVENTHANDLER_LIST_DEFINE(process_dtor);
161 EVENTHANDLER_LIST_DEFINE(process_init);
162 EVENTHANDLER_LIST_DEFINE(process_fini);
163 EVENTHANDLER_LIST_DEFINE(process_exit);
164 EVENTHANDLER_LIST_DEFINE(process_fork);
165 EVENTHANDLER_LIST_DEFINE(process_exec);
166 
167 int kstack_pages = KSTACK_PAGES;
168 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RDTUN | CTLFLAG_NOFETCH,
169     &kstack_pages, 0,
170     "Kernel stack size in pages");
171 static int vmmap_skip_res_cnt = 0;
172 SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW,
173     &vmmap_skip_res_cnt, 0,
174     "Skip calculation of the pages resident count in kern.proc.vmmap");
175 
176 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
177 #ifdef COMPAT_FREEBSD32
178 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
179 #endif
180 
181 /*
182  * Initialize global process hashing structures.
183  */
184 void
185 procinit(void)
186 {
187 	u_long i;
188 
189 	sx_init(&allproc_lock, "allproc");
190 	sx_init(&proctree_lock, "proctree");
191 	mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
192 	mtx_init(&procid_lock, "procid", NULL, MTX_DEF);
193 	pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
194 	pidhashlock = (pidhash + 1) / 64;
195 	if (pidhashlock > 0)
196 		pidhashlock--;
197 	pidhashtbl_lock = malloc(sizeof(*pidhashtbl_lock) * (pidhashlock + 1),
198 	    M_PROC, M_WAITOK | M_ZERO);
199 	for (i = 0; i < pidhashlock + 1; i++)
200 		sx_init_flags(&pidhashtbl_lock[i], "pidhash", SX_DUPOK);
201 	pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
202 	proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
203 	    proc_ctor, proc_dtor, proc_init, proc_fini,
204 	    UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
205 	pgrp_zone = uma_zcreate("PGRP", sizeof(struct pgrp), NULL, NULL,
206 	    pgrp_init, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
207 	uihashinit();
208 }
209 
210 /*
211  * Prepare a proc for use.
212  */
213 static int
214 proc_ctor(void *mem, int size, void *arg, int flags)
215 {
216 	struct proc *p;
217 	struct thread *td;
218 
219 	p = (struct proc *)mem;
220 #ifdef KDTRACE_HOOKS
221 	kdtrace_proc_ctor(p);
222 #endif
223 	EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
224 	td = FIRST_THREAD_IN_PROC(p);
225 	if (td != NULL) {
226 		/* Make sure all thread constructors are executed */
227 		EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
228 	}
229 	return (0);
230 }
231 
232 /*
233  * Reclaim a proc after use.
234  */
235 static void
236 proc_dtor(void *mem, int size, void *arg)
237 {
238 	struct proc *p;
239 	struct thread *td;
240 
241 	p = mem;
242 	td = FIRST_THREAD_IN_PROC(p);
243 	if (td != NULL) {
244 		KASSERT(p->p_numthreads == 1,
245 		    ("too many threads in exiting process"));
246 
247 		/* Free all OSD associated to this thread. */
248 		osd_thread_exit(td);
249 		ast_kclear(td);
250 
251 		/* Make sure all thread destructors are executed */
252 		EVENTHANDLER_DIRECT_INVOKE(thread_dtor, td);
253 	}
254 	KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
255 	EVENTHANDLER_DIRECT_INVOKE(process_dtor, p);
256 #ifdef KDTRACE_HOOKS
257 	kdtrace_proc_dtor(p);
258 #endif
259 	KASSERT(p->p_ksi == NULL || !KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
260 }
261 
262 /*
263  * Initialize type-stable parts of a proc (when newly created).
264  */
265 static int
266 proc_init(void *mem, int size, int flags)
267 {
268 	struct proc *p;
269 
270 	p = (struct proc *)mem;
271 	mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK | MTX_NEW);
272 	mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_NEW);
273 	mtx_init(&p->p_statmtx, "pstatl", NULL, MTX_SPIN | MTX_NEW);
274 	mtx_init(&p->p_itimmtx, "pitiml", NULL, MTX_SPIN | MTX_NEW);
275 	mtx_init(&p->p_profmtx, "pprofl", NULL, MTX_SPIN | MTX_NEW);
276 	cv_init(&p->p_pwait, "ppwait");
277 	TAILQ_INIT(&p->p_threads);	     /* all threads in proc */
278 	EVENTHANDLER_DIRECT_INVOKE(process_init, p);
279 	p->p_stats = pstats_alloc();
280 	p->p_pgrp = NULL;
281 	TAILQ_INIT(&p->p_kqtim_stop);
282 	STAILQ_INIT(&p->p_ktr);
283 	refcount_init(&p->p_tree_refcnt, 0);
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 (1 + 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 = 1 + cred->cr_ngroups;
1120 		kp->ki_groups[0] = cred->cr_gid;
1121 		bcopy(cred->cr_groups, kp->ki_groups + 1,
1122 		    (kp->ki_ngroups - 1) * sizeof(gid_t));
1123 		kp->ki_rgid = cred->cr_rgid;
1124 		kp->ki_svgid = cred->cr_svgid;
1125 		/* If jailed(cred), emulate the old P_JAILED flag. */
1126 		if (jailed(cred)) {
1127 			kp->ki_flag |= P_JAILED;
1128 			/* If inside the jail, use 0 as a jail ID. */
1129 			if (cred->cr_prison != curthread->td_ucred->cr_prison)
1130 				kp->ki_jid = cred->cr_prison->pr_id;
1131 		}
1132 		strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
1133 		    sizeof(kp->ki_loginclass));
1134 	}
1135 	ps = p->p_sigacts;
1136 	if (ps) {
1137 		mtx_lock(&ps->ps_mtx);
1138 		kp->ki_sigignore = ps->ps_sigignore;
1139 		kp->ki_sigcatch = ps->ps_sigcatch;
1140 		mtx_unlock(&ps->ps_mtx);
1141 	}
1142 	if (p->p_state != PRS_NEW &&
1143 	    p->p_state != PRS_ZOMBIE &&
1144 	    p->p_vmspace != NULL) {
1145 		struct vmspace *vm = p->p_vmspace;
1146 
1147 		kp->ki_size = vm->vm_map.size;
1148 		kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
1149 		FOREACH_THREAD_IN_PROC(p, td0)
1150 			kp->ki_rssize += td0->td_kstack_pages;
1151 		kp->ki_swrss = vm->vm_swrss;
1152 		kp->ki_tsize = vm->vm_tsize;
1153 		kp->ki_dsize = vm->vm_dsize;
1154 		kp->ki_ssize = vm->vm_ssize;
1155 	} else if (p->p_state == PRS_ZOMBIE)
1156 		kp->ki_stat = SZOMB;
1157 	kp->ki_sflag = PS_INMEM;
1158 	/* Calculate legacy swtime as seconds since 'swtick'. */
1159 	kp->ki_swtime = (ticks - p->p_swtick) / hz;
1160 	kp->ki_pid = p->p_pid;
1161 	kp->ki_nice = p->p_nice;
1162 	kp->ki_fibnum = p->p_fibnum;
1163 	kp->ki_start = p->p_stats->p_start;
1164 	getboottime(&boottime);
1165 	timevaladd(&kp->ki_start, &boottime);
1166 	PROC_STATLOCK(p);
1167 	rufetch(p, &kp->ki_rusage);
1168 	kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
1169 	calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
1170 	PROC_STATUNLOCK(p);
1171 	calccru(p, &kp->ki_childutime, &kp->ki_childstime);
1172 	/* Some callers want child times in a single value. */
1173 	kp->ki_childtime = kp->ki_childstime;
1174 	timevaladd(&kp->ki_childtime, &kp->ki_childutime);
1175 
1176 	FOREACH_THREAD_IN_PROC(p, td0)
1177 		kp->ki_cow += td0->td_cow;
1178 
1179 	if (p->p_comm[0] != '\0')
1180 		strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
1181 	if (p->p_sysent && p->p_sysent->sv_name != NULL &&
1182 	    p->p_sysent->sv_name[0] != '\0')
1183 		strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
1184 	kp->ki_siglist = p->p_siglist;
1185 	kp->ki_xstat = KW_EXITCODE(p->p_xexit, p->p_xsig);
1186 	kp->ki_acflag = p->p_acflag;
1187 	kp->ki_lock = p->p_lock;
1188 	if (p->p_pptr) {
1189 		kp->ki_ppid = p->p_oppid;
1190 		if (p->p_flag & P_TRACED)
1191 			kp->ki_tracer = p->p_pptr->p_pid;
1192 	}
1193 }
1194 
1195 /*
1196  * Fill job-related process information.
1197  */
1198 static void
1199 fill_kinfo_proc_pgrp(struct proc *p, struct kinfo_proc *kp)
1200 {
1201 	struct tty *tp;
1202 	struct session *sp;
1203 	struct pgrp *pgrp;
1204 
1205 	sx_assert(&proctree_lock, SA_LOCKED);
1206 	PROC_LOCK_ASSERT(p, MA_OWNED);
1207 
1208 	pgrp = p->p_pgrp;
1209 	if (pgrp == NULL)
1210 		return;
1211 
1212 	kp->ki_pgid = pgrp->pg_id;
1213 	kp->ki_jobc = pgrp_calc_jobc(pgrp);
1214 
1215 	sp = pgrp->pg_session;
1216 	tp = NULL;
1217 
1218 	if (sp != NULL) {
1219 		kp->ki_sid = sp->s_sid;
1220 		SESS_LOCK(sp);
1221 		strlcpy(kp->ki_login, sp->s_login, sizeof(kp->ki_login));
1222 		if (sp->s_ttyvp)
1223 			kp->ki_kiflag |= KI_CTTY;
1224 		if (SESS_LEADER(p))
1225 			kp->ki_kiflag |= KI_SLEADER;
1226 		tp = sp->s_ttyp;
1227 		SESS_UNLOCK(sp);
1228 	}
1229 
1230 	if ((p->p_flag & P_CONTROLT) && tp != NULL) {
1231 		kp->ki_tdev = tty_udev(tp);
1232 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1233 		kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
1234 		if (tp->t_session)
1235 			kp->ki_tsid = tp->t_session->s_sid;
1236 	} else {
1237 		kp->ki_tdev = NODEV;
1238 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1239 	}
1240 	kp->ki_reaper = p->p_reaper->p_pid;
1241 	kp->ki_reapsubtree = p->p_reapsubtree;
1242 }
1243 
1244 /*
1245  * Fill in information that is thread specific.  Must be called with
1246  * target process locked.  If 'preferthread' is set, overwrite certain
1247  * process-related fields that are maintained for both threads and
1248  * processes.
1249  */
1250 static void
1251 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
1252 {
1253 	struct proc *p;
1254 
1255 	p = td->td_proc;
1256 	kp->ki_tdaddr = td;
1257 	PROC_LOCK_ASSERT(p, MA_OWNED);
1258 
1259 	if (preferthread)
1260 		PROC_STATLOCK(p);
1261 	thread_lock(td);
1262 	if (td->td_wmesg != NULL)
1263 		strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
1264 	else
1265 		bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
1266 	if (strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname)) >=
1267 	    sizeof(kp->ki_tdname)) {
1268 		strlcpy(kp->ki_moretdname,
1269 		    td->td_name + sizeof(kp->ki_tdname) - 1,
1270 		    sizeof(kp->ki_moretdname));
1271 	} else {
1272 		bzero(kp->ki_moretdname, sizeof(kp->ki_moretdname));
1273 	}
1274 	if (TD_ON_LOCK(td)) {
1275 		kp->ki_kiflag |= KI_LOCKBLOCK;
1276 		strlcpy(kp->ki_lockname, td->td_lockname,
1277 		    sizeof(kp->ki_lockname));
1278 	} else {
1279 		kp->ki_kiflag &= ~KI_LOCKBLOCK;
1280 		bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
1281 	}
1282 
1283 	if (p->p_state == PRS_NORMAL) { /* approximate. */
1284 		if (TD_ON_RUNQ(td) ||
1285 		    TD_CAN_RUN(td) ||
1286 		    TD_IS_RUNNING(td)) {
1287 			kp->ki_stat = SRUN;
1288 		} else if (P_SHOULDSTOP(p)) {
1289 			kp->ki_stat = SSTOP;
1290 		} else if (TD_IS_SLEEPING(td)) {
1291 			kp->ki_stat = SSLEEP;
1292 		} else if (TD_ON_LOCK(td)) {
1293 			kp->ki_stat = SLOCK;
1294 		} else {
1295 			kp->ki_stat = SWAIT;
1296 		}
1297 	} else if (p->p_state == PRS_ZOMBIE) {
1298 		kp->ki_stat = SZOMB;
1299 	} else {
1300 		kp->ki_stat = SIDL;
1301 	}
1302 
1303 	/* Things in the thread */
1304 	kp->ki_wchan = td->td_wchan;
1305 	kp->ki_pri.pri_level = td->td_priority;
1306 	kp->ki_pri.pri_native = td->td_base_pri;
1307 
1308 	/*
1309 	 * Note: legacy fields; clamp at the old NOCPU value and/or
1310 	 * the maximum u_char CPU value.
1311 	 */
1312 	if (td->td_lastcpu == NOCPU)
1313 		kp->ki_lastcpu_old = NOCPU_OLD;
1314 	else if (td->td_lastcpu > MAXCPU_OLD)
1315 		kp->ki_lastcpu_old = MAXCPU_OLD;
1316 	else
1317 		kp->ki_lastcpu_old = td->td_lastcpu;
1318 
1319 	if (td->td_oncpu == NOCPU)
1320 		kp->ki_oncpu_old = NOCPU_OLD;
1321 	else if (td->td_oncpu > MAXCPU_OLD)
1322 		kp->ki_oncpu_old = MAXCPU_OLD;
1323 	else
1324 		kp->ki_oncpu_old = td->td_oncpu;
1325 
1326 	kp->ki_lastcpu = td->td_lastcpu;
1327 	kp->ki_oncpu = td->td_oncpu;
1328 	kp->ki_tdflags = td->td_flags;
1329 	kp->ki_tid = td->td_tid;
1330 	kp->ki_numthreads = p->p_numthreads;
1331 	kp->ki_pcb = td->td_pcb;
1332 	kp->ki_kstack = td->td_kstack;
1333 	kp->ki_slptime = (ticks - td->td_slptick) / hz;
1334 	kp->ki_pri.pri_class = td->td_pri_class;
1335 	kp->ki_pri.pri_user = td->td_user_pri;
1336 
1337 	if (preferthread) {
1338 		rufetchtd(td, &kp->ki_rusage);
1339 		kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
1340 		kp->ki_pctcpu = sched_pctcpu(td);
1341 		kp->ki_estcpu = sched_estcpu(td);
1342 		kp->ki_cow = td->td_cow;
1343 	}
1344 
1345 	/* We can't get this anymore but ps etc never used it anyway. */
1346 	kp->ki_rqindex = 0;
1347 
1348 	if (preferthread)
1349 		kp->ki_siglist = td->td_siglist;
1350 	kp->ki_sigmask = td->td_sigmask;
1351 	thread_unlock(td);
1352 	if (preferthread)
1353 		PROC_STATUNLOCK(p);
1354 
1355 	if ((td->td_pflags & TDP2_UEXTERR) != 0)
1356 		kp->ki_uerrmsg = td->td_exterr_ptr;
1357 }
1358 
1359 /*
1360  * Fill in a kinfo_proc structure for the specified process.
1361  * Must be called with the target process locked.
1362  */
1363 void
1364 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1365 {
1366 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1367 
1368 	bzero(kp, sizeof(*kp));
1369 
1370 	fill_kinfo_proc_pgrp(p,kp);
1371 	fill_kinfo_proc_only(p, kp);
1372 	fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1373 	fill_kinfo_aggregate(p, kp);
1374 }
1375 
1376 struct pstats *
1377 pstats_alloc(void)
1378 {
1379 
1380 	return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1381 }
1382 
1383 /*
1384  * Copy parts of p_stats; zero the rest of p_stats (statistics).
1385  */
1386 void
1387 pstats_fork(struct pstats *src, struct pstats *dst)
1388 {
1389 
1390 	bzero(&dst->pstat_startzero,
1391 	    __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1392 	bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1393 	    __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1394 }
1395 
1396 void
1397 pstats_free(struct pstats *ps)
1398 {
1399 
1400 	free(ps, M_SUBPROC);
1401 }
1402 
1403 #ifdef COMPAT_FREEBSD32
1404 
1405 /*
1406  * This function is typically used to copy out the kernel address, so
1407  * it can be replaced by assignment of zero.
1408  */
1409 static inline uint32_t
1410 ptr32_trim(const void *ptr)
1411 {
1412 	uintptr_t uptr;
1413 
1414 	uptr = (uintptr_t)ptr;
1415 	return ((uptr > UINT_MAX) ? 0 : uptr);
1416 }
1417 
1418 #define PTRTRIM_CP(src,dst,fld) \
1419 	do { (dst).fld = ptr32_trim((src).fld); } while (0)
1420 
1421 static void
1422 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1423 {
1424 	int i;
1425 
1426 	bzero(ki32, sizeof(struct kinfo_proc32));
1427 	ki32->ki_structsize = sizeof(struct kinfo_proc32);
1428 	CP(*ki, *ki32, ki_layout);
1429 	PTRTRIM_CP(*ki, *ki32, ki_args);
1430 	PTRTRIM_CP(*ki, *ki32, ki_paddr);
1431 	PTRTRIM_CP(*ki, *ki32, ki_addr);
1432 	PTRTRIM_CP(*ki, *ki32, ki_tracep);
1433 	PTRTRIM_CP(*ki, *ki32, ki_textvp);
1434 	PTRTRIM_CP(*ki, *ki32, ki_fd);
1435 	PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1436 	PTRTRIM_CP(*ki, *ki32, ki_wchan);
1437 	CP(*ki, *ki32, ki_pid);
1438 	CP(*ki, *ki32, ki_ppid);
1439 	CP(*ki, *ki32, ki_pgid);
1440 	CP(*ki, *ki32, ki_tpgid);
1441 	CP(*ki, *ki32, ki_sid);
1442 	CP(*ki, *ki32, ki_tsid);
1443 	CP(*ki, *ki32, ki_jobc);
1444 	FU64_CP(*ki, *ki32, ki_tdev);
1445 	CP(*ki, *ki32, ki_tdev_freebsd11);
1446 	CP(*ki, *ki32, ki_siglist);
1447 	CP(*ki, *ki32, ki_sigmask);
1448 	CP(*ki, *ki32, ki_sigignore);
1449 	CP(*ki, *ki32, ki_sigcatch);
1450 	CP(*ki, *ki32, ki_uid);
1451 	CP(*ki, *ki32, ki_ruid);
1452 	CP(*ki, *ki32, ki_svuid);
1453 	CP(*ki, *ki32, ki_rgid);
1454 	CP(*ki, *ki32, ki_svgid);
1455 	CP(*ki, *ki32, ki_ngroups);
1456 	for (i = 0; i < KI_NGROUPS; i++)
1457 		CP(*ki, *ki32, ki_groups[i]);
1458 	CP(*ki, *ki32, ki_size);
1459 	CP(*ki, *ki32, ki_rssize);
1460 	CP(*ki, *ki32, ki_swrss);
1461 	CP(*ki, *ki32, ki_tsize);
1462 	CP(*ki, *ki32, ki_dsize);
1463 	CP(*ki, *ki32, ki_ssize);
1464 	CP(*ki, *ki32, ki_xstat);
1465 	CP(*ki, *ki32, ki_acflag);
1466 	CP(*ki, *ki32, ki_pctcpu);
1467 	CP(*ki, *ki32, ki_estcpu);
1468 	CP(*ki, *ki32, ki_slptime);
1469 	CP(*ki, *ki32, ki_swtime);
1470 	CP(*ki, *ki32, ki_cow);
1471 	FU64_CP(*ki, *ki32, ki_runtime);
1472 	TV_CP(*ki, *ki32, ki_start);
1473 	TV_CP(*ki, *ki32, ki_childtime);
1474 	CP(*ki, *ki32, ki_flag);
1475 	CP(*ki, *ki32, ki_kiflag);
1476 	CP(*ki, *ki32, ki_traceflag);
1477 	CP(*ki, *ki32, ki_stat);
1478 	CP(*ki, *ki32, ki_nice);
1479 	CP(*ki, *ki32, ki_lock);
1480 	CP(*ki, *ki32, ki_rqindex);
1481 	CP(*ki, *ki32, ki_oncpu);
1482 	CP(*ki, *ki32, ki_lastcpu);
1483 
1484 	/* XXX TODO: wrap cpu value as appropriate */
1485 	CP(*ki, *ki32, ki_oncpu_old);
1486 	CP(*ki, *ki32, ki_lastcpu_old);
1487 
1488 	bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1489 	bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1490 	bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1491 	bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1492 	bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1493 	bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1494 	bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1495 	bcopy(ki->ki_moretdname, ki32->ki_moretdname, MAXCOMLEN - TDNAMLEN + 1);
1496 	CP(*ki, *ki32, ki_tracer);
1497 	CP(*ki, *ki32, ki_flag2);
1498 	CP(*ki, *ki32, ki_fibnum);
1499 	CP(*ki, *ki32, ki_cr_flags);
1500 	CP(*ki, *ki32, ki_jid);
1501 	CP(*ki, *ki32, ki_reaper);
1502 	CP(*ki, *ki32, ki_reapsubtree);
1503 	CP(*ki, *ki32, ki_numthreads);
1504 	CP(*ki, *ki32, ki_tid);
1505 	CP(*ki, *ki32, ki_pri);
1506 	freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1507 	freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1508 	PTRTRIM_CP(*ki, *ki32, ki_pcb);
1509 	PTRTRIM_CP(*ki, *ki32, ki_kstack);
1510 	PTRTRIM_CP(*ki, *ki32, ki_udata);
1511 	PTRTRIM_CP(*ki, *ki32, ki_tdaddr);
1512 	PTRTRIM_CP(*ki, *ki32, ki_pd);
1513 	CP(*ki, *ki32, ki_sflag);
1514 	CP(*ki, *ki32, ki_tdflags);
1515 	PTRTRIM_CP(*ki, *ki32, ki_uerrmsg);
1516 }
1517 #endif
1518 
1519 static ssize_t
1520 kern_proc_out_size(struct proc *p, int flags)
1521 {
1522 	ssize_t size = 0;
1523 
1524 	PROC_LOCK_ASSERT(p, MA_OWNED);
1525 
1526 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1527 #ifdef COMPAT_FREEBSD32
1528 		if ((flags & KERN_PROC_MASK32) != 0) {
1529 			size += sizeof(struct kinfo_proc32);
1530 		} else
1531 #endif
1532 			size += sizeof(struct kinfo_proc);
1533 	} else {
1534 #ifdef COMPAT_FREEBSD32
1535 		if ((flags & KERN_PROC_MASK32) != 0)
1536 			size += sizeof(struct kinfo_proc32) * p->p_numthreads;
1537 		else
1538 #endif
1539 			size += sizeof(struct kinfo_proc) * p->p_numthreads;
1540 	}
1541 	PROC_UNLOCK(p);
1542 	return (size);
1543 }
1544 
1545 int
1546 kern_proc_out(struct proc *p, struct sbuf *sb, int flags)
1547 {
1548 	struct thread *td;
1549 	struct kinfo_proc ki;
1550 #ifdef COMPAT_FREEBSD32
1551 	struct kinfo_proc32 ki32;
1552 #endif
1553 	int error;
1554 
1555 	PROC_LOCK_ASSERT(p, MA_OWNED);
1556 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1557 
1558 	error = 0;
1559 	fill_kinfo_proc(p, &ki);
1560 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1561 #ifdef COMPAT_FREEBSD32
1562 		if ((flags & KERN_PROC_MASK32) != 0) {
1563 			freebsd32_kinfo_proc_out(&ki, &ki32);
1564 			if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1565 				error = ENOMEM;
1566 		} else
1567 #endif
1568 			if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1569 				error = ENOMEM;
1570 	} else {
1571 		FOREACH_THREAD_IN_PROC(p, td) {
1572 			fill_kinfo_thread(td, &ki, 1);
1573 #ifdef COMPAT_FREEBSD32
1574 			if ((flags & KERN_PROC_MASK32) != 0) {
1575 				freebsd32_kinfo_proc_out(&ki, &ki32);
1576 				if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1577 					error = ENOMEM;
1578 			} else
1579 #endif
1580 				if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1581 					error = ENOMEM;
1582 			if (error != 0)
1583 				break;
1584 		}
1585 	}
1586 	PROC_UNLOCK(p);
1587 	return (error);
1588 }
1589 
1590 static int
1591 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
1592 {
1593 	struct sbuf sb;
1594 	struct kinfo_proc ki;
1595 	int error, error2;
1596 
1597 	if (req->oldptr == NULL)
1598 		return (SYSCTL_OUT(req, 0, kern_proc_out_size(p, flags)));
1599 
1600 	sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req);
1601 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1602 	error = kern_proc_out(p, &sb, flags);
1603 	error2 = sbuf_finish(&sb);
1604 	sbuf_delete(&sb);
1605 	if (error != 0)
1606 		return (error);
1607 	else if (error2 != 0)
1608 		return (error2);
1609 	return (0);
1610 }
1611 
1612 int
1613 proc_iterate(int (*cb)(struct proc *, void *), void *cbarg)
1614 {
1615 	struct proc *p;
1616 	int error, i, j;
1617 
1618 	for (i = 0; i < pidhashlock + 1; i++) {
1619 		sx_slock(&proctree_lock);
1620 		sx_slock(&pidhashtbl_lock[i]);
1621 		for (j = i; j <= pidhash; j += pidhashlock + 1) {
1622 			LIST_FOREACH(p, &pidhashtbl[j], p_hash) {
1623 				if (p->p_state == PRS_NEW)
1624 					continue;
1625 				error = cb(p, cbarg);
1626 				PROC_LOCK_ASSERT(p, MA_NOTOWNED);
1627 				if (error != 0) {
1628 					sx_sunlock(&pidhashtbl_lock[i]);
1629 					sx_sunlock(&proctree_lock);
1630 					return (error);
1631 				}
1632 			}
1633 		}
1634 		sx_sunlock(&pidhashtbl_lock[i]);
1635 		sx_sunlock(&proctree_lock);
1636 	}
1637 	return (0);
1638 }
1639 
1640 struct kern_proc_out_args {
1641 	struct sysctl_req *req;
1642 	int flags;
1643 	int oid_number;
1644 	int *name;
1645 };
1646 
1647 static int
1648 sysctl_kern_proc_iterate(struct proc *p, void *origarg)
1649 {
1650 	struct kern_proc_out_args *arg = origarg;
1651 	int *name = arg->name;
1652 	int oid_number = arg->oid_number;
1653 	int flags = arg->flags;
1654 	struct sysctl_req *req = arg->req;
1655 	int error = 0;
1656 
1657 	PROC_LOCK(p);
1658 
1659 	KASSERT(p->p_ucred != NULL,
1660 	    ("process credential is NULL for non-NEW proc"));
1661 	/*
1662 	 * Show a user only appropriate processes.
1663 	 */
1664 	if (p_cansee(curthread, p))
1665 		goto skip;
1666 	/*
1667 	 * TODO - make more efficient (see notes below).
1668 	 * do by session.
1669 	 */
1670 	switch (oid_number) {
1671 	case KERN_PROC_GID:
1672 		if (p->p_ucred->cr_gid != (gid_t)name[0])
1673 			goto skip;
1674 		break;
1675 
1676 	case KERN_PROC_PGRP:
1677 		/* could do this by traversing pgrp */
1678 		if (p->p_pgrp == NULL ||
1679 		    p->p_pgrp->pg_id != (pid_t)name[0])
1680 			goto skip;
1681 		break;
1682 
1683 	case KERN_PROC_RGID:
1684 		if (p->p_ucred->cr_rgid != (gid_t)name[0])
1685 			goto skip;
1686 		break;
1687 
1688 	case KERN_PROC_SESSION:
1689 		if (p->p_session == NULL ||
1690 		    p->p_session->s_sid != (pid_t)name[0])
1691 			goto skip;
1692 		break;
1693 
1694 	case KERN_PROC_TTY:
1695 		if ((p->p_flag & P_CONTROLT) == 0 ||
1696 		    p->p_session == NULL)
1697 			goto skip;
1698 		/* XXX proctree_lock */
1699 		SESS_LOCK(p->p_session);
1700 		if (p->p_session->s_ttyp == NULL ||
1701 		    tty_udev(p->p_session->s_ttyp) !=
1702 		    (dev_t)name[0]) {
1703 			SESS_UNLOCK(p->p_session);
1704 			goto skip;
1705 		}
1706 		SESS_UNLOCK(p->p_session);
1707 		break;
1708 
1709 	case KERN_PROC_UID:
1710 		if (p->p_ucred->cr_uid != (uid_t)name[0])
1711 			goto skip;
1712 		break;
1713 
1714 	case KERN_PROC_RUID:
1715 		if (p->p_ucred->cr_ruid != (uid_t)name[0])
1716 			goto skip;
1717 		break;
1718 
1719 	case KERN_PROC_PROC:
1720 		break;
1721 
1722 	default:
1723 		break;
1724 	}
1725 	error = sysctl_out_proc(p, req, flags);
1726 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
1727 	return (error);
1728 skip:
1729 	PROC_UNLOCK(p);
1730 	return (0);
1731 }
1732 
1733 static int
1734 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1735 {
1736 	struct kern_proc_out_args iterarg;
1737 	int *name = (int *)arg1;
1738 	u_int namelen = arg2;
1739 	struct proc *p;
1740 	int flags, oid_number;
1741 	int error = 0;
1742 
1743 	oid_number = oidp->oid_number;
1744 	if (oid_number != KERN_PROC_ALL &&
1745 	    (oid_number & KERN_PROC_INC_THREAD) == 0)
1746 		flags = KERN_PROC_NOTHREADS;
1747 	else {
1748 		flags = 0;
1749 		oid_number &= ~KERN_PROC_INC_THREAD;
1750 	}
1751 #ifdef COMPAT_FREEBSD32
1752 	if (req->flags & SCTL_MASK32)
1753 		flags |= KERN_PROC_MASK32;
1754 #endif
1755 	if (oid_number == KERN_PROC_PID) {
1756 		if (namelen != 1)
1757 			return (EINVAL);
1758 		error = sysctl_wire_old_buffer(req, 0);
1759 		if (error)
1760 			return (error);
1761 		sx_slock(&proctree_lock);
1762 		error = pget((pid_t)name[0], PGET_CANSEE, &p);
1763 		if (error == 0)
1764 			error = sysctl_out_proc(p, req, flags);
1765 		sx_sunlock(&proctree_lock);
1766 		return (error);
1767 	}
1768 
1769 	switch (oid_number) {
1770 	case KERN_PROC_ALL:
1771 		if (namelen != 0)
1772 			return (EINVAL);
1773 		break;
1774 	case KERN_PROC_PROC:
1775 		if (namelen != 0 && namelen != 1)
1776 			return (EINVAL);
1777 		break;
1778 	default:
1779 		if (namelen != 1)
1780 			return (EINVAL);
1781 		break;
1782 	}
1783 
1784 	if (req->oldptr == NULL) {
1785 		/* overestimate by 5 procs */
1786 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1787 		if (error)
1788 			return (error);
1789 	} else {
1790 		error = sysctl_wire_old_buffer(req, 0);
1791 		if (error != 0)
1792 			return (error);
1793 	}
1794 	iterarg.flags = flags;
1795 	iterarg.oid_number = oid_number;
1796 	iterarg.req = req;
1797 	iterarg.name = name;
1798 	error = proc_iterate(sysctl_kern_proc_iterate, &iterarg);
1799 	return (error);
1800 }
1801 
1802 struct pargs *
1803 pargs_alloc(int len)
1804 {
1805 	struct pargs *pa;
1806 
1807 	pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1808 		M_WAITOK);
1809 	refcount_init(&pa->ar_ref, 1);
1810 	pa->ar_length = len;
1811 	return (pa);
1812 }
1813 
1814 static void
1815 pargs_free(struct pargs *pa)
1816 {
1817 
1818 	free(pa, M_PARGS);
1819 }
1820 
1821 void
1822 pargs_hold(struct pargs *pa)
1823 {
1824 
1825 	if (pa == NULL)
1826 		return;
1827 	refcount_acquire(&pa->ar_ref);
1828 }
1829 
1830 void
1831 pargs_drop(struct pargs *pa)
1832 {
1833 
1834 	if (pa == NULL)
1835 		return;
1836 	if (refcount_release(&pa->ar_ref))
1837 		pargs_free(pa);
1838 }
1839 
1840 static int
1841 proc_read_string(struct thread *td, struct vmspace *vm, const char *sptr,
1842     char *buf, size_t len)
1843 {
1844 	ssize_t n;
1845 
1846 	/*
1847 	 * This may return a short read if the string is shorter than the chunk
1848 	 * and is aligned at the end of the page, and the following page is not
1849 	 * mapped.
1850 	 */
1851 	n = vmspace_iop(td, vm, (vm_offset_t)sptr, buf, len, UIO_READ);
1852 	if (n <= 0)
1853 		return (ENOMEM);
1854 	return (0);
1855 }
1856 
1857 #define PROC_AUXV_MAX	256	/* Safety limit on auxv size. */
1858 
1859 enum proc_vector_type {
1860 	PROC_ARG,
1861 	PROC_ENV,
1862 	PROC_AUX,
1863 };
1864 
1865 #ifdef COMPAT_FREEBSD32
1866 static int
1867 get_proc_vector32(struct thread *td, struct proc *p, struct vmspace *vm,
1868     char ***proc_vectorp, size_t *vsizep, enum proc_vector_type type)
1869 {
1870 	struct freebsd32_ps_strings pss;
1871 	Elf32_Auxinfo aux;
1872 	vm_offset_t vptr, ptr;
1873 	uint32_t *proc_vector32;
1874 	char **proc_vector;
1875 	size_t vsize, size;
1876 	int i, error;
1877 
1878 	error = 0;
1879 	if (vmspace_iop(td, vm, PROC_PS_STRINGS(p), &pss, sizeof(pss),
1880 	    UIO_READ) != sizeof(pss))
1881 		return (ENOMEM);
1882 	switch (type) {
1883 	case PROC_ARG:
1884 		vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1885 		vsize = pss.ps_nargvstr;
1886 		if (vsize > ARG_MAX)
1887 			return (ENOEXEC);
1888 		size = vsize * sizeof(int32_t);
1889 		break;
1890 	case PROC_ENV:
1891 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1892 		vsize = pss.ps_nenvstr;
1893 		if (vsize > ARG_MAX)
1894 			return (ENOEXEC);
1895 		size = vsize * sizeof(int32_t);
1896 		break;
1897 	case PROC_AUX:
1898 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1899 		    (pss.ps_nenvstr + 1) * sizeof(int32_t);
1900 		if (vptr % 4 != 0)
1901 			return (ENOEXEC);
1902 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1903 			if (vmspace_iop(td, vm, ptr, &aux, sizeof(aux),
1904 			    UIO_READ) != sizeof(aux))
1905 				return (ENOMEM);
1906 			if (aux.a_type == AT_NULL)
1907 				break;
1908 			ptr += sizeof(aux);
1909 		}
1910 		if (aux.a_type != AT_NULL)
1911 			return (ENOEXEC);
1912 		vsize = i + 1;
1913 		size = vsize * sizeof(aux);
1914 		break;
1915 	default:
1916 		KASSERT(0, ("Wrong proc vector type: %d", type));
1917 		return (EINVAL);
1918 	}
1919 	proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1920 	if (vmspace_iop(td, vm, vptr, proc_vector32, size, UIO_READ) != size) {
1921 		error = ENOMEM;
1922 		goto done;
1923 	}
1924 	if (type == PROC_AUX) {
1925 		*proc_vectorp = (char **)proc_vector32;
1926 		*vsizep = vsize;
1927 		return (0);
1928 	}
1929 	proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1930 	for (i = 0; i < (int)vsize; i++)
1931 		proc_vector[i] = PTRIN(proc_vector32[i]);
1932 	*proc_vectorp = proc_vector;
1933 	*vsizep = vsize;
1934 done:
1935 	free(proc_vector32, M_TEMP);
1936 	return (error);
1937 }
1938 #endif
1939 
1940 static int
1941 get_proc_vector(struct thread *td, struct proc *p, struct vmspace *vm,
1942     char ***proc_vectorp, size_t *vsizep, enum proc_vector_type type)
1943 {
1944 	struct ps_strings pss;
1945 	Elf_Auxinfo aux;
1946 	vm_offset_t vptr, ptr;
1947 	char **proc_vector;
1948 	size_t vsize, size;
1949 	int i;
1950 
1951 #ifdef COMPAT_FREEBSD32
1952 	if (SV_PROC_FLAG(p, SV_ILP32) != 0) {
1953 		return (get_proc_vector32(td, p, vm, proc_vectorp,
1954 		    vsizep, type));
1955 	}
1956 #endif
1957 	if (vmspace_iop(td, vm, PROC_PS_STRINGS(p), &pss, sizeof(pss),
1958 	    UIO_READ) != sizeof(pss))
1959 		return (ENOMEM);
1960 	switch (type) {
1961 	case PROC_ARG:
1962 		vptr = (vm_offset_t)pss.ps_argvstr;
1963 		vsize = pss.ps_nargvstr;
1964 		if (vsize > ARG_MAX)
1965 			return (ENOEXEC);
1966 		size = vsize * sizeof(char *);
1967 		break;
1968 	case PROC_ENV:
1969 		vptr = (vm_offset_t)pss.ps_envstr;
1970 		vsize = pss.ps_nenvstr;
1971 		if (vsize > ARG_MAX)
1972 			return (ENOEXEC);
1973 		size = vsize * sizeof(char *);
1974 		break;
1975 	case PROC_AUX:
1976 		/*
1977 		 * The aux array is just above env array on the stack. Check
1978 		 * that the address is naturally aligned.
1979 		 */
1980 		vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1981 		    * sizeof(char *);
1982 #if __ELF_WORD_SIZE == 64
1983 		if (vptr % sizeof(uint64_t) != 0)
1984 #else
1985 		if (vptr % sizeof(uint32_t) != 0)
1986 #endif
1987 			return (ENOEXEC);
1988 		/*
1989 		 * We count the array size reading the aux vectors from the
1990 		 * stack until AT_NULL vector is returned.  So (to keep the code
1991 		 * simple) we read the process stack twice: the first time here
1992 		 * to find the size and the second time when copying the vectors
1993 		 * to the allocated proc_vector.
1994 		 */
1995 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1996 			if (vmspace_iop(td, vm, ptr, &aux, sizeof(aux),
1997 			    UIO_READ) != sizeof(aux))
1998 				return (ENOMEM);
1999 			if (aux.a_type == AT_NULL)
2000 				break;
2001 			ptr += sizeof(aux);
2002 		}
2003 		/*
2004 		 * If the PROC_AUXV_MAX entries are iterated over, and we have
2005 		 * not reached AT_NULL, it is most likely we are reading wrong
2006 		 * data: either the process doesn't have auxv array or data has
2007 		 * been modified. Return the error in this case.
2008 		 */
2009 		if (aux.a_type != AT_NULL)
2010 			return (ENOEXEC);
2011 		vsize = i + 1;
2012 		size = vsize * sizeof(aux);
2013 		break;
2014 	default:
2015 		KASSERT(0, ("Wrong proc vector type: %d", type));
2016 		return (EINVAL); /* In case we are built without INVARIANTS. */
2017 	}
2018 	proc_vector = malloc(size, M_TEMP, M_WAITOK);
2019 	if (vmspace_iop(td, vm, vptr, proc_vector, size, UIO_READ) != size) {
2020 		free(proc_vector, M_TEMP);
2021 		return (ENOMEM);
2022 	}
2023 	*proc_vectorp = proc_vector;
2024 	*vsizep = vsize;
2025 
2026 	return (0);
2027 }
2028 
2029 #define GET_PS_STRINGS_CHUNK_SZ	256	/* Chunk size (bytes) for ps_strings operations. */
2030 
2031 static int
2032 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
2033     enum proc_vector_type type)
2034 {
2035 	struct vmspace *vm;
2036 	size_t done, len, nchr, vsize;
2037 	int error, i;
2038 	char **proc_vector, *sptr;
2039 	char pss_string[GET_PS_STRINGS_CHUNK_SZ];
2040 
2041 	PROC_ASSERT_HELD(p);
2042 
2043 	/*
2044 	 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
2045 	 */
2046 	nchr = 2 * (PATH_MAX + ARG_MAX);
2047 
2048 	error = proc_vmspace_ref(td, p, PRVM_BLOCK_EXEC |
2049 	    PRVM_CHECK_VISIBILITY, &vm);
2050 	if (error != 0)
2051 		return (error);
2052 
2053 	error = get_proc_vector(td, p, vm, &proc_vector, &vsize, type);
2054 	if (error != 0)
2055 		goto out;
2056 	for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
2057 		/*
2058 		 * The program may have scribbled into its argv array, e.g. to
2059 		 * remove some arguments.  If that has happened, break out
2060 		 * before trying to read from NULL.
2061 		 */
2062 		if (proc_vector[i] == NULL)
2063 			break;
2064 		for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
2065 			error = proc_read_string(td, vm, sptr, pss_string,
2066 			    sizeof(pss_string));
2067 			if (error != 0)
2068 				goto done;
2069 			len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
2070 			if (done + len >= nchr)
2071 				len = nchr - done - 1;
2072 			sbuf_bcat(sb, pss_string, len);
2073 			if (len != GET_PS_STRINGS_CHUNK_SZ)
2074 				break;
2075 			done += GET_PS_STRINGS_CHUNK_SZ;
2076 		}
2077 		sbuf_bcat(sb, "", 1);
2078 		done += len + 1;
2079 	}
2080 done:
2081 	free(proc_vector, M_TEMP);
2082 out:
2083 	proc_vmspace_unref(td, p, PRVM_BLOCK_EXEC | PRVM_CHECK_VISIBILITY, vm);
2084 	return (error);
2085 }
2086 
2087 int
2088 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
2089 {
2090 
2091 	return (get_ps_strings(curthread, p, sb, PROC_ARG));
2092 }
2093 
2094 int
2095 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
2096 {
2097 
2098 	return (get_ps_strings(curthread, p, sb, PROC_ENV));
2099 }
2100 
2101 int
2102 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb)
2103 {
2104 	struct vmspace *vm;
2105 	size_t vsize, size;
2106 	char **auxv;
2107 	int error;
2108 
2109 	error = proc_vmspace_ref(td, p, PRVM_BLOCK_EXEC | PRVM_CHECK_DEBUG,
2110 	    &vm);
2111 	if (error != 0)
2112 		return (error);
2113 	error = get_proc_vector(td, p, vm, &auxv, &vsize, PROC_AUX);
2114 	proc_vmspace_unref(td, p, PRVM_BLOCK_EXEC | PRVM_CHECK_DEBUG, vm);
2115 	if (error == 0) {
2116 #ifdef COMPAT_FREEBSD32
2117 		if (SV_PROC_FLAG(p, SV_ILP32) != 0)
2118 			size = vsize * sizeof(Elf32_Auxinfo);
2119 		else
2120 #endif
2121 			size = vsize * sizeof(Elf_Auxinfo);
2122 		if (sbuf_bcat(sb, auxv, size) != 0)
2123 			error = ENOMEM;
2124 		free(auxv, M_TEMP);
2125 	}
2126 	return (error);
2127 }
2128 
2129 /*
2130  * This sysctl allows a process to retrieve the argument list or process
2131  * title for another process without groping around in the address space
2132  * of the other process.  It also allow a process to set its own "process
2133  * title to a string of its own choice.
2134  */
2135 static int
2136 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
2137 {
2138 	int *name = (int *)arg1;
2139 	u_int namelen = arg2;
2140 	struct pargs *newpa, *pa;
2141 	struct proc *p;
2142 	struct sbuf sb;
2143 	int flags, error = 0, error2;
2144 	pid_t pid;
2145 
2146 	if (namelen != 1)
2147 		return (EINVAL);
2148 
2149 	p = curproc;
2150 	pid = (pid_t)name[0];
2151 	if (pid == -1) {
2152 		pid = p->p_pid;
2153 	}
2154 
2155 	/*
2156 	 * If the query is for this process and it is single-threaded, there
2157 	 * is nobody to modify pargs, thus we can just read.
2158 	 */
2159 	if (pid == p->p_pid && p->p_numthreads == 1 && req->newptr == NULL &&
2160 	    (pa = p->p_args) != NULL)
2161 		return (SYSCTL_OUT(req, pa->ar_args, pa->ar_length));
2162 
2163 	flags = PGET_CANSEE;
2164 	if (req->newptr != NULL)
2165 		flags |= PGET_ISCURRENT;
2166 	error = pget(pid, flags, &p);
2167 	if (error)
2168 		return (error);
2169 
2170 	pa = p->p_args;
2171 	if (pa != NULL) {
2172 		pargs_hold(pa);
2173 		PROC_UNLOCK(p);
2174 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
2175 		pargs_drop(pa);
2176 	} else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
2177 		_PHOLD(p);
2178 		PROC_UNLOCK(p);
2179 		sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2180 		sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2181 		error = proc_getargv(curthread, p, &sb);
2182 		error2 = sbuf_finish(&sb);
2183 		PRELE(p);
2184 		sbuf_delete(&sb);
2185 		if (error == 0 && error2 != 0)
2186 			error = error2;
2187 	} else {
2188 		PROC_UNLOCK(p);
2189 	}
2190 	if (error != 0 || req->newptr == NULL)
2191 		return (error);
2192 
2193 	if (req->newlen > ps_arg_cache_limit - sizeof(struct pargs))
2194 		return (ENOMEM);
2195 
2196 	if (req->newlen == 0) {
2197 		/*
2198 		 * Clear the argument pointer, so that we'll fetch arguments
2199 		 * with proc_getargv() until further notice.
2200 		 */
2201 		newpa = NULL;
2202 	} else {
2203 		newpa = pargs_alloc(req->newlen);
2204 		error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
2205 		if (error != 0) {
2206 			pargs_free(newpa);
2207 			return (error);
2208 		}
2209 	}
2210 	PROC_LOCK(p);
2211 	pa = p->p_args;
2212 	p->p_args = newpa;
2213 	PROC_UNLOCK(p);
2214 	pargs_drop(pa);
2215 	return (0);
2216 }
2217 
2218 /*
2219  * This sysctl allows a process to retrieve environment of another process.
2220  */
2221 static int
2222 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
2223 {
2224 	int *name = (int *)arg1;
2225 	u_int namelen = arg2;
2226 	struct proc *p;
2227 	struct sbuf sb;
2228 	int error, error2;
2229 
2230 	if (namelen != 1)
2231 		return (EINVAL);
2232 
2233 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2234 	if (error != 0)
2235 		return (error);
2236 	if ((p->p_flag & P_SYSTEM) != 0) {
2237 		PRELE(p);
2238 		return (0);
2239 	}
2240 
2241 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2242 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2243 	error = proc_getenvv(curthread, p, &sb);
2244 	error2 = sbuf_finish(&sb);
2245 	PRELE(p);
2246 	sbuf_delete(&sb);
2247 	return (error != 0 ? error : error2);
2248 }
2249 
2250 /*
2251  * This sysctl allows a process to retrieve ELF auxiliary vector of
2252  * another process.
2253  */
2254 static int
2255 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
2256 {
2257 	int *name = (int *)arg1;
2258 	u_int namelen = arg2;
2259 	struct proc *p;
2260 	struct sbuf sb;
2261 	int error, error2;
2262 
2263 	if (namelen != 1)
2264 		return (EINVAL);
2265 
2266 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2267 	if (error != 0)
2268 		return (error);
2269 	if ((p->p_flag & P_SYSTEM) != 0) {
2270 		PRELE(p);
2271 		return (0);
2272 	}
2273 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2274 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2275 	error = proc_getauxv(curthread, p, &sb);
2276 	error2 = sbuf_finish(&sb);
2277 	PRELE(p);
2278 	sbuf_delete(&sb);
2279 	return (error != 0 ? error : error2);
2280 }
2281 
2282 /*
2283  * Look up the canonical executable path running in the specified process.
2284  * It tries to return the same hardlink name as was used for execve(2).
2285  * This allows the programs that modify their behavior based on their progname,
2286  * to operate correctly.
2287  *
2288  * Result is returned in retbuf, it must not be freed, similar to vn_fullpath()
2289  *   calling conventions.
2290  * binname is a pointer to temporary string buffer of length MAXPATHLEN,
2291  *   allocated and freed by caller.
2292  * freebuf should be freed by caller, from the M_TEMP malloc type.
2293  */
2294 int
2295 proc_get_binpath(struct proc *p, char *binname, char **retbuf,
2296     char **freebuf)
2297 {
2298 	struct nameidata nd;
2299 	struct vnode *vp, *dvp;
2300 	size_t freepath_size;
2301 	int error;
2302 	bool do_fullpath;
2303 
2304 	PROC_LOCK_ASSERT(p, MA_OWNED);
2305 
2306 	vp = p->p_textvp;
2307 	if (vp == NULL) {
2308 		PROC_UNLOCK(p);
2309 		*retbuf = "";
2310 		*freebuf = NULL;
2311 		return (0);
2312 	}
2313 	vref(vp);
2314 	dvp = p->p_textdvp;
2315 	if (dvp != NULL)
2316 		vref(dvp);
2317 	if (p->p_binname != NULL)
2318 		strlcpy(binname, p->p_binname, MAXPATHLEN);
2319 	PROC_UNLOCK(p);
2320 
2321 	do_fullpath = true;
2322 	*freebuf = NULL;
2323 	if (dvp != NULL && binname[0] != '\0') {
2324 		freepath_size = MAXPATHLEN;
2325 		if (vn_fullpath_hardlink(vp, dvp, binname, strlen(binname),
2326 		    retbuf, freebuf, &freepath_size) == 0) {
2327 			/*
2328 			 * Recheck the looked up path.  The binary
2329 			 * might have been renamed or replaced, in
2330 			 * which case we should not report old name.
2331 			 */
2332 			NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, *retbuf);
2333 			error = namei(&nd);
2334 			if (error == 0) {
2335 				if (nd.ni_vp == vp)
2336 					do_fullpath = false;
2337 				vrele(nd.ni_vp);
2338 				NDFREE_PNBUF(&nd);
2339 			}
2340 		}
2341 	}
2342 	if (do_fullpath) {
2343 		free(*freebuf, M_TEMP);
2344 		*freebuf = NULL;
2345 		error = vn_fullpath(vp, retbuf, freebuf);
2346 	}
2347 	vrele(vp);
2348 	if (dvp != NULL)
2349 		vrele(dvp);
2350 	return (error);
2351 }
2352 
2353 /*
2354  * This sysctl allows a process to retrieve the path of the executable for
2355  * itself or another process.
2356  */
2357 static int
2358 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
2359 {
2360 	pid_t *pidp = (pid_t *)arg1;
2361 	unsigned int arglen = arg2;
2362 	struct proc *p;
2363 	char *retbuf, *freebuf, *binname;
2364 	int error;
2365 
2366 	if (arglen != 1)
2367 		return (EINVAL);
2368 	binname = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
2369 	binname[0] = '\0';
2370 	if (*pidp == -1) {	/* -1 means this process */
2371 		error = 0;
2372 		p = req->td->td_proc;
2373 		PROC_LOCK(p);
2374 	} else {
2375 		error = pget(*pidp, PGET_CANSEE, &p);
2376 	}
2377 
2378 	if (error == 0)
2379 		error = proc_get_binpath(p, binname, &retbuf, &freebuf);
2380 	free(binname, M_TEMP);
2381 	if (error != 0)
2382 		return (error);
2383 	error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
2384 	free(freebuf, M_TEMP);
2385 	return (error);
2386 }
2387 
2388 static int
2389 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
2390 {
2391 	struct proc *p;
2392 	char *sv_name;
2393 	int *name;
2394 	int namelen;
2395 	int error;
2396 
2397 	namelen = arg2;
2398 	if (namelen != 1)
2399 		return (EINVAL);
2400 
2401 	name = (int *)arg1;
2402 	error = pget((pid_t)name[0], PGET_CANSEE, &p);
2403 	if (error != 0)
2404 		return (error);
2405 	sv_name = p->p_sysent->sv_name;
2406 	PROC_UNLOCK(p);
2407 	return (sysctl_handle_string(oidp, sv_name, 0, req));
2408 }
2409 
2410 #ifdef KINFO_OVMENTRY_SIZE
2411 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
2412 #endif
2413 
2414 #ifdef COMPAT_FREEBSD7
2415 static int
2416 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
2417 {
2418 	vm_map_entry_t entry, tmp_entry;
2419 	unsigned int last_timestamp, namelen;
2420 	char *fullpath, *freepath;
2421 	struct kinfo_ovmentry *kve;
2422 	struct vattr va;
2423 	struct ucred *cred;
2424 	int error, *name;
2425 	struct vnode *vp;
2426 	struct proc *p;
2427 	struct thread *td;
2428 	vm_map_t map;
2429 	struct vmspace *vm;
2430 
2431 	namelen = arg2;
2432 	if (namelen != 1)
2433 		return (EINVAL);
2434 
2435 	name = (int *)arg1;
2436 	td = curthread;
2437 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2438 	if (error != 0)
2439 		return (error);
2440 	error = proc_vmspace_ref(td, p, PRVM_CHECK_DEBUG, &vm);
2441 	if (error != 0) {
2442 		PRELE(p);
2443 		return (ESRCH);
2444 	}
2445 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2446 
2447 	map = &vm->vm_map;
2448 	vm_map_lock_read(map);
2449 	VM_MAP_ENTRY_FOREACH(entry, map) {
2450 		vm_object_t obj, tobj, lobj;
2451 		vm_offset_t addr;
2452 
2453 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2454 			continue;
2455 
2456 		bzero(kve, sizeof(*kve));
2457 		kve->kve_structsize = sizeof(*kve);
2458 
2459 		kve->kve_private_resident = 0;
2460 		obj = entry->object.vm_object;
2461 		if (obj != NULL) {
2462 			VM_OBJECT_RLOCK(obj);
2463 			if (obj->shadow_count == 1)
2464 				kve->kve_private_resident =
2465 				    obj->resident_page_count;
2466 		}
2467 		kve->kve_resident = 0;
2468 		addr = entry->start;
2469 		while (addr < entry->end) {
2470 			if (pmap_extract(map->pmap, addr))
2471 				kve->kve_resident++;
2472 			addr += PAGE_SIZE;
2473 		}
2474 
2475 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2476 			if (tobj != obj) {
2477 				VM_OBJECT_RLOCK(tobj);
2478 				kve->kve_offset += tobj->backing_object_offset;
2479 			}
2480 			if (lobj != obj)
2481 				VM_OBJECT_RUNLOCK(lobj);
2482 			lobj = tobj;
2483 		}
2484 
2485 		kve->kve_start = (void*)entry->start;
2486 		kve->kve_end = (void*)entry->end;
2487 		kve->kve_offset += (off_t)entry->offset;
2488 
2489 		if (entry->protection & VM_PROT_READ)
2490 			kve->kve_protection |= KVME_PROT_READ;
2491 		if (entry->protection & VM_PROT_WRITE)
2492 			kve->kve_protection |= KVME_PROT_WRITE;
2493 		if (entry->protection & VM_PROT_EXECUTE)
2494 			kve->kve_protection |= KVME_PROT_EXEC;
2495 
2496 		if (entry->eflags & MAP_ENTRY_COW)
2497 			kve->kve_flags |= KVME_FLAG_COW;
2498 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2499 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2500 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2501 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2502 
2503 		last_timestamp = map->timestamp;
2504 		vm_map_unlock_read(map);
2505 
2506 		kve->kve_fileid = 0;
2507 		kve->kve_fsid = 0;
2508 		freepath = NULL;
2509 		fullpath = "";
2510 		if (lobj) {
2511 			kve->kve_type = vm_object_kvme_type(lobj, &vp);
2512 			if (kve->kve_type == KVME_TYPE_MGTDEVICE)
2513 				kve->kve_type = KVME_TYPE_UNKNOWN;
2514 			if (vp != NULL)
2515 				vref(vp);
2516 			if (lobj != obj)
2517 				VM_OBJECT_RUNLOCK(lobj);
2518 
2519 			kve->kve_ref_count = obj->ref_count;
2520 			kve->kve_shadow_count = obj->shadow_count;
2521 			VM_OBJECT_RUNLOCK(obj);
2522 			if (vp != NULL) {
2523 				vn_fullpath(vp, &fullpath, &freepath);
2524 				cred = curthread->td_ucred;
2525 				vn_lock(vp, LK_SHARED | LK_RETRY);
2526 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2527 					kve->kve_fileid = va.va_fileid;
2528 					/* truncate */
2529 					kve->kve_fsid = va.va_fsid;
2530 				}
2531 				vput(vp);
2532 			}
2533 		} else {
2534 			kve->kve_type = KVME_TYPE_NONE;
2535 			kve->kve_ref_count = 0;
2536 			kve->kve_shadow_count = 0;
2537 		}
2538 
2539 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2540 		if (freepath != NULL)
2541 			free(freepath, M_TEMP);
2542 
2543 		error = SYSCTL_OUT(req, kve, sizeof(*kve));
2544 		vm_map_lock_read(map);
2545 		if (error)
2546 			break;
2547 		if (last_timestamp != map->timestamp) {
2548 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2549 			entry = tmp_entry;
2550 		}
2551 	}
2552 	vm_map_unlock_read(map);
2553 	proc_vmspace_unref(td, p, PRVM_CHECK_DEBUG, vm);
2554 	PRELE(p);
2555 	free(kve, M_TEMP);
2556 	return (error);
2557 }
2558 #endif	/* COMPAT_FREEBSD7 */
2559 
2560 #ifdef KINFO_VMENTRY_SIZE
2561 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2562 #endif
2563 
2564 void
2565 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry,
2566     int *resident_count, bool *super)
2567 {
2568 	vm_object_t obj, tobj;
2569 	vm_page_t m, m_adv;
2570 	vm_offset_t addr;
2571 	vm_paddr_t pa;
2572 	vm_pindex_t pi, pi_adv, pindex;
2573 	int incore;
2574 
2575 	*super = false;
2576 	*resident_count = 0;
2577 	if (vmmap_skip_res_cnt)
2578 		return;
2579 
2580 	pa = 0;
2581 	obj = entry->object.vm_object;
2582 	addr = entry->start;
2583 	m_adv = NULL;
2584 	pi = OFF_TO_IDX(entry->offset);
2585 	for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) {
2586 		if (m_adv != NULL) {
2587 			m = m_adv;
2588 		} else {
2589 			pi_adv = atop(entry->end - addr);
2590 			pindex = pi;
2591 			for (tobj = obj;; tobj = tobj->backing_object) {
2592 				m = vm_radix_lookup_ge(&tobj->rtree, pindex);
2593 				if (m != NULL) {
2594 					if (m->pindex == pindex)
2595 						break;
2596 					if (pi_adv > m->pindex - pindex) {
2597 						pi_adv = m->pindex - pindex;
2598 						m_adv = m;
2599 					}
2600 				}
2601 				if (tobj->backing_object == NULL)
2602 					goto next;
2603 				pindex += OFF_TO_IDX(tobj->
2604 				    backing_object_offset);
2605 			}
2606 		}
2607 		m_adv = NULL;
2608 		if (m->psind != 0 && addr + pagesizes[1] <= entry->end &&
2609 		    (addr & (pagesizes[1] - 1)) == 0 && (incore =
2610 		    pmap_mincore(map->pmap, addr, &pa) & MINCORE_SUPER) != 0) {
2611 			*super = true;
2612 			/*
2613 			 * The virtual page might be smaller than the physical
2614 			 * page, so we use the page size reported by the pmap
2615 			 * rather than m->psind.
2616 			 */
2617 			pi_adv = atop(pagesizes[incore >> MINCORE_PSIND_SHIFT]);
2618 		} else {
2619 			/*
2620 			 * We do not test the found page on validity.
2621 			 * Either the page is busy and being paged in,
2622 			 * or it was invalidated.  The first case
2623 			 * should be counted as resident, the second
2624 			 * is not so clear; we do account both.
2625 			 */
2626 			pi_adv = 1;
2627 		}
2628 		*resident_count += pi_adv;
2629 next:;
2630 	}
2631 }
2632 
2633 /*
2634  * Must be called with the process locked and will return unlocked.
2635  */
2636 int
2637 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb, ssize_t maxlen, int flags)
2638 {
2639 	vm_map_entry_t entry, tmp_entry;
2640 	struct vattr va;
2641 	vm_map_t map;
2642 	vm_object_t lobj, nobj, obj, tobj;
2643 	char *fullpath, *freepath;
2644 	struct kinfo_vmentry *kve;
2645 	struct ucred *cred;
2646 	struct vnode *vp;
2647 	struct vmspace *vm;
2648 	struct thread *td;
2649 	vm_offset_t addr;
2650 	unsigned int last_timestamp;
2651 	int error;
2652 	key_t key;
2653 	unsigned short seq;
2654 	bool guard, super;
2655 
2656 	PROC_LOCK_ASSERT(p, MA_OWNED);
2657 
2658 	_PHOLD(p);
2659 	PROC_UNLOCK(p);
2660 	td = curthread;
2661 	error = proc_vmspace_ref(td, p, PRVM_CHECK_DEBUG, &vm);
2662 	if (error != 0) {
2663 		PRELE(p);
2664 		return (error);
2665 	}
2666 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK | M_ZERO);
2667 
2668 	error = 0;
2669 	map = &vm->vm_map;
2670 	vm_map_lock_read(map);
2671 	VM_MAP_ENTRY_FOREACH(entry, map) {
2672 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2673 			continue;
2674 
2675 		addr = entry->end;
2676 		bzero(kve, sizeof(*kve));
2677 		obj = entry->object.vm_object;
2678 		if (obj != NULL) {
2679 			if ((obj->flags & OBJ_ANON) != 0)
2680 				kve->kve_obj = (uintptr_t)obj;
2681 
2682 			for (tobj = obj; tobj != NULL;
2683 			    tobj = tobj->backing_object) {
2684 				VM_OBJECT_RLOCK(tobj);
2685 				kve->kve_offset += tobj->backing_object_offset;
2686 				lobj = tobj;
2687 			}
2688 			if (obj->backing_object == NULL)
2689 				kve->kve_private_resident =
2690 				    obj->resident_page_count;
2691 			kern_proc_vmmap_resident(map, entry,
2692 			    &kve->kve_resident, &super);
2693 			if (super)
2694 				kve->kve_flags |= KVME_FLAG_SUPER;
2695 			for (tobj = obj; tobj != NULL; tobj = nobj) {
2696 				nobj = tobj->backing_object;
2697 				if (tobj != obj && tobj != lobj)
2698 					VM_OBJECT_RUNLOCK(tobj);
2699 			}
2700 		} else {
2701 			lobj = NULL;
2702 		}
2703 
2704 		kve->kve_start = entry->start;
2705 		kve->kve_end = entry->end;
2706 		kve->kve_offset += entry->offset;
2707 
2708 		if (entry->protection & VM_PROT_READ)
2709 			kve->kve_protection |= KVME_PROT_READ;
2710 		if (entry->protection & VM_PROT_WRITE)
2711 			kve->kve_protection |= KVME_PROT_WRITE;
2712 		if (entry->protection & VM_PROT_EXECUTE)
2713 			kve->kve_protection |= KVME_PROT_EXEC;
2714 		if (entry->max_protection & VM_PROT_READ)
2715 			kve->kve_protection |= KVME_MAX_PROT_READ;
2716 		if (entry->max_protection & VM_PROT_WRITE)
2717 			kve->kve_protection |= KVME_MAX_PROT_WRITE;
2718 		if (entry->max_protection & VM_PROT_EXECUTE)
2719 			kve->kve_protection |= KVME_MAX_PROT_EXEC;
2720 
2721 		if (entry->eflags & MAP_ENTRY_COW)
2722 			kve->kve_flags |= KVME_FLAG_COW;
2723 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2724 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2725 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2726 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2727 		if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2728 			kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2729 		if (entry->eflags & MAP_ENTRY_USER_WIRED)
2730 			kve->kve_flags |= KVME_FLAG_USER_WIRED;
2731 
2732 		guard = (entry->eflags & MAP_ENTRY_GUARD) != 0;
2733 
2734 		last_timestamp = map->timestamp;
2735 		vm_map_unlock_read(map);
2736 
2737 		freepath = NULL;
2738 		fullpath = "";
2739 		if (lobj != NULL) {
2740 			kve->kve_type = vm_object_kvme_type(lobj, &vp);
2741 			if (vp != NULL)
2742 				vref(vp);
2743 			if (lobj != obj)
2744 				VM_OBJECT_RUNLOCK(lobj);
2745 
2746 			kve->kve_ref_count = obj->ref_count;
2747 			kve->kve_shadow_count = obj->shadow_count;
2748 			if (obj->type == OBJT_DEVICE ||
2749 			    obj->type == OBJT_MGTDEVICE) {
2750 				cdev_pager_get_path(obj, kve->kve_path,
2751 				    sizeof(kve->kve_path));
2752 			}
2753 			VM_OBJECT_RUNLOCK(obj);
2754 			if ((lobj->flags & OBJ_SYSVSHM) != 0) {
2755 				kve->kve_flags |= KVME_FLAG_SYSVSHM;
2756 				shmobjinfo(lobj, &key, &seq);
2757 				kve->kve_vn_fileid = key;
2758 				kve->kve_vn_fsid_freebsd11 = seq;
2759 			}
2760 			if ((lobj->flags & OBJ_POSIXSHM) != 0) {
2761 				kve->kve_flags |= KVME_FLAG_POSIXSHM;
2762 				shm_get_path(lobj, kve->kve_path,
2763 				    sizeof(kve->kve_path));
2764 			}
2765 			if (vp != NULL) {
2766 				vn_fullpath(vp, &fullpath, &freepath);
2767 				kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2768 				cred = td->td_ucred;
2769 				vn_lock(vp, LK_SHARED | LK_RETRY);
2770 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2771 					kve->kve_vn_fileid = va.va_fileid;
2772 					kve->kve_vn_fsid = va.va_fsid;
2773 					kve->kve_vn_fsid_freebsd11 =
2774 					    kve->kve_vn_fsid; /* truncate */
2775 					kve->kve_vn_mode =
2776 					    MAKEIMODE(va.va_type, va.va_mode);
2777 					kve->kve_vn_size = va.va_size;
2778 					kve->kve_vn_rdev = va.va_rdev;
2779 					kve->kve_vn_rdev_freebsd11 =
2780 					    kve->kve_vn_rdev; /* truncate */
2781 					kve->kve_status = KF_ATTR_VALID;
2782 				}
2783 				vput(vp);
2784 				strlcpy(kve->kve_path, fullpath, sizeof(
2785 				    kve->kve_path));
2786 				free(freepath, M_TEMP);
2787 			}
2788 		} else {
2789 			kve->kve_type = guard ? KVME_TYPE_GUARD :
2790 			    KVME_TYPE_NONE;
2791 			kve->kve_ref_count = 0;
2792 			kve->kve_shadow_count = 0;
2793 		}
2794 
2795 		/* Pack record size down */
2796 		if ((flags & KERN_VMMAP_PACK_KINFO) != 0)
2797 			kve->kve_structsize =
2798 			    offsetof(struct kinfo_vmentry, kve_path) +
2799 			    strlen(kve->kve_path) + 1;
2800 		else
2801 			kve->kve_structsize = sizeof(*kve);
2802 		kve->kve_structsize = roundup(kve->kve_structsize,
2803 		    sizeof(uint64_t));
2804 
2805 		/* Halt filling and truncate rather than exceeding maxlen */
2806 		if (maxlen != -1 && maxlen < kve->kve_structsize) {
2807 			error = 0;
2808 			vm_map_lock_read(map);
2809 			break;
2810 		} else if (maxlen != -1)
2811 			maxlen -= kve->kve_structsize;
2812 
2813 		if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0)
2814 			error = ENOMEM;
2815 		vm_map_lock_read(map);
2816 		if (error != 0)
2817 			break;
2818 		if (last_timestamp != map->timestamp) {
2819 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2820 			entry = tmp_entry;
2821 		}
2822 	}
2823 	vm_map_unlock_read(map);
2824 	proc_vmspace_unref(td, p, PRVM_CHECK_DEBUG, vm);
2825 	PRELE(p);
2826 	free(kve, M_TEMP);
2827 	return (error);
2828 }
2829 
2830 static int
2831 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2832 {
2833 	struct proc *p;
2834 	struct sbuf sb;
2835 	u_int namelen;
2836 	int error, error2, *name;
2837 
2838 	namelen = arg2;
2839 	if (namelen != 1)
2840 		return (EINVAL);
2841 
2842 	name = (int *)arg1;
2843 	sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req);
2844 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2845 	error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
2846 	if (error != 0) {
2847 		sbuf_delete(&sb);
2848 		return (error);
2849 	}
2850 	error = kern_proc_vmmap_out(p, &sb, -1, KERN_VMMAP_PACK_KINFO);
2851 	error2 = sbuf_finish(&sb);
2852 	sbuf_delete(&sb);
2853 	return (error != 0 ? error : error2);
2854 }
2855 
2856 #if defined(STACK) || defined(DDB)
2857 static int
2858 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2859 {
2860 	struct kinfo_kstack *kkstp;
2861 	int error, i, *name, numthreads;
2862 	lwpid_t *lwpidarray;
2863 	struct thread *td, *ctd;
2864 	struct stack *st;
2865 	struct sbuf sb;
2866 	struct proc *p;
2867 	u_int namelen;
2868 
2869 	namelen = arg2;
2870 	if (namelen != 1)
2871 		return (EINVAL);
2872 
2873 	name = (int *)arg1;
2874 	ctd = curthread;
2875 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2876 	if (error != 0)
2877 		return (error);
2878 
2879 	kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2880 	st = stack_create(M_WAITOK);
2881 
2882 	lwpidarray = NULL;
2883 	PROC_LOCK(p);
2884 	execve_block_wait(ctd, p);
2885 	error = p_candebug(ctd, p);
2886 	if (error != 0) {
2887 		execve_unblock(ctd, p);
2888 		_PRELE(p);
2889 		PROC_UNLOCK(p);
2890 		return (error);
2891 	}
2892 	do {
2893 		if (lwpidarray != NULL) {
2894 			free(lwpidarray, M_TEMP);
2895 			lwpidarray = NULL;
2896 		}
2897 		numthreads = p->p_numthreads;
2898 		PROC_UNLOCK(p);
2899 		lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2900 		    M_WAITOK | M_ZERO);
2901 		PROC_LOCK(p);
2902 	} while (numthreads < p->p_numthreads);
2903 
2904 	i = 0;
2905 	FOREACH_THREAD_IN_PROC(p, td) {
2906 		KASSERT(i < numthreads,
2907 		    ("sysctl_kern_proc_kstack: numthreads"));
2908 		lwpidarray[i] = td->td_tid;
2909 		i++;
2910 	}
2911 	PROC_UNLOCK(p);
2912 	numthreads = i;
2913 	for (i = 0; i < numthreads; i++) {
2914 		td = tdfind(lwpidarray[i], p->p_pid);
2915 		if (td == NULL) {
2916 			continue;
2917 		}
2918 		bzero(kkstp, sizeof(*kkstp));
2919 		(void)sbuf_new(&sb, kkstp->kkst_trace,
2920 		    sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2921 		thread_lock(td);
2922 		kkstp->kkst_tid = td->td_tid;
2923 		if (stack_save_td(st, td) == 0)
2924 			kkstp->kkst_state = KKST_STATE_STACKOK;
2925 		else
2926 			kkstp->kkst_state = KKST_STATE_RUNNING;
2927 		thread_unlock(td);
2928 		PROC_UNLOCK(p);
2929 		stack_sbuf_print(&sb, st);
2930 		sbuf_finish(&sb);
2931 		sbuf_delete(&sb);
2932 		error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2933 		if (error)
2934 			break;
2935 	}
2936 	PROC_LOCK(p);
2937 	execve_unblock(ctd, p);
2938 	_PRELE(p);
2939 	PROC_UNLOCK(p);
2940 	if (lwpidarray != NULL)
2941 		free(lwpidarray, M_TEMP);
2942 	stack_destroy(st);
2943 	free(kkstp, M_TEMP);
2944 	return (error);
2945 }
2946 #endif
2947 
2948 /*
2949  * This sysctl allows a process to retrieve the full list of groups from
2950  * itself or another process.
2951  */
2952 static int
2953 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2954 {
2955 	pid_t *pidp = (pid_t *)arg1;
2956 	unsigned int arglen = arg2;
2957 	struct proc *p;
2958 	struct ucred *cred;
2959 	int error;
2960 
2961 	if (arglen != 1)
2962 		return (EINVAL);
2963 	if (*pidp == -1) {	/* -1 means this process */
2964 		p = req->td->td_proc;
2965 		PROC_LOCK(p);
2966 	} else {
2967 		error = pget(*pidp, PGET_CANSEE, &p);
2968 		if (error != 0)
2969 			return (error);
2970 	}
2971 
2972 	cred = crhold(p->p_ucred);
2973 	PROC_UNLOCK(p);
2974 
2975 	error = SYSCTL_OUT(req, &cred->cr_gid, sizeof(gid_t));
2976 	if (error == 0)
2977 		error = SYSCTL_OUT(req, cred->cr_groups,
2978 		    cred->cr_ngroups * sizeof(gid_t));
2979 
2980 	crfree(cred);
2981 	return (error);
2982 }
2983 
2984 /*
2985  * This sysctl allows a process to retrieve or/and set the resource limit for
2986  * another process.
2987  */
2988 static int
2989 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2990 {
2991 	int *name = (int *)arg1;
2992 	u_int namelen = arg2;
2993 	struct rlimit rlim;
2994 	struct proc *p;
2995 	struct thread *td;
2996 	u_int which;
2997 	int error;
2998 
2999 	if (namelen != 2)
3000 		return (EINVAL);
3001 
3002 	which = (u_int)name[1];
3003 	if (which >= RLIM_NLIMITS)
3004 		return (EINVAL);
3005 
3006 	if (req->newptr != NULL && req->newlen != sizeof(rlim))
3007 		return (EINVAL);
3008 
3009 	td = curthread;
3010 	error = pget((pid_t)name[0], PGET_NOTWEXIT, &p);
3011 	if (error != 0)
3012 		return (error);
3013 	_PHOLD(p);
3014 	execve_block_wait(td, p);
3015 	error = req->newptr != NULL ? p_candebug(td, p) : p_cansee(td, p);
3016 	if (error != 0)
3017 		goto errout1;
3018 
3019 	/*
3020 	 * Retrieve limit.
3021 	 */
3022 	if (req->oldptr != NULL) {
3023 		lim_rlimit_proc(p, which, &rlim);
3024 	}
3025 	PROC_UNLOCK(p);
3026 
3027 	error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
3028 	if (error != 0)
3029 		goto errout;
3030 
3031 	/*
3032 	 * Set limit.
3033 	 */
3034 	if (req->newptr != NULL) {
3035 		error = SYSCTL_IN(req, &rlim, sizeof(rlim));
3036 		if (error == 0)
3037 			error = kern_proc_setrlimit(curthread, p, which, &rlim);
3038 	}
3039 
3040 errout:
3041 	PROC_LOCK(p);
3042 errout1:
3043 	_PRELE(p);
3044 	execve_unblock(td, p);
3045 	PROC_UNLOCK(p);
3046 	return (error);
3047 }
3048 
3049 /*
3050  * This sysctl allows a process to retrieve ps_strings structure location of
3051  * another process.
3052  */
3053 static int
3054 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
3055 {
3056 	int *name = (int *)arg1;
3057 	u_int namelen = arg2;
3058 	struct proc *p;
3059 	vm_offset_t ps_strings;
3060 	int error;
3061 #ifdef COMPAT_FREEBSD32
3062 	uint32_t ps_strings32;
3063 #endif
3064 
3065 	if (namelen != 1)
3066 		return (EINVAL);
3067 
3068 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
3069 	if (error != 0)
3070 		return (error);
3071 #ifdef COMPAT_FREEBSD32
3072 	if ((req->flags & SCTL_MASK32) != 0) {
3073 		/*
3074 		 * We return 0 if the 32 bit emulation request is for a 64 bit
3075 		 * process.
3076 		 */
3077 		ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
3078 		    PTROUT(PROC_PS_STRINGS(p)) : 0;
3079 		PROC_UNLOCK(p);
3080 		error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
3081 		return (error);
3082 	}
3083 #endif
3084 	ps_strings = PROC_PS_STRINGS(p);
3085 	PROC_UNLOCK(p);
3086 	error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
3087 	return (error);
3088 }
3089 
3090 /*
3091  * This sysctl allows a process to retrieve umask of another process.
3092  */
3093 static int
3094 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
3095 {
3096 	int *name = (int *)arg1;
3097 	u_int namelen = arg2;
3098 	struct proc *p;
3099 	int error;
3100 	u_short cmask;
3101 	pid_t pid;
3102 
3103 	if (namelen != 1)
3104 		return (EINVAL);
3105 
3106 	pid = (pid_t)name[0];
3107 	p = curproc;
3108 	if (pid == p->p_pid || pid == 0) {
3109 		cmask = p->p_pd->pd_cmask;
3110 		goto out;
3111 	}
3112 
3113 	error = pget(pid, PGET_WANTREAD, &p);
3114 	if (error != 0)
3115 		return (error);
3116 
3117 	cmask = p->p_pd->pd_cmask;
3118 	PRELE(p);
3119 out:
3120 	error = SYSCTL_OUT(req, &cmask, sizeof(cmask));
3121 	return (error);
3122 }
3123 
3124 /*
3125  * This sysctl allows a process to set and retrieve binary osreldate of
3126  * another process.
3127  */
3128 static int
3129 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
3130 {
3131 	int *name = (int *)arg1;
3132 	u_int namelen = arg2;
3133 	struct proc *p;
3134 	int flags, error, old_osrel, osrel;
3135 
3136 	if (namelen != 1)
3137 		return (EINVAL);
3138 
3139 	flags = PGET_NOTWEXIT;
3140 	if (req->newptr != NULL) {
3141 		if (req->newlen != sizeof(osrel))
3142 			return (EINVAL);
3143 		error = SYSCTL_IN(req, &osrel, sizeof(osrel));
3144 		if (error != 0)
3145 			return (error);
3146 		if (osrel < 0)
3147 			return (EINVAL);
3148 		flags |= PGET_CANDEBUG;
3149 	} else {
3150 		flags |= PGET_CANSEE;
3151 	}
3152 	error = pget((pid_t)name[0], flags, &p);
3153 	if (error != 0)
3154 		return (error);
3155 	if ((p->p_flag & P_INEXEC) != 0) {
3156 		error = EBUSY;
3157 	} else {
3158 		old_osrel = p->p_osrel;
3159 		if (req->newptr != NULL)
3160 			p->p_osrel = osrel;
3161 	}
3162 	PROC_UNLOCK(p);
3163 
3164 	if (error == 0)
3165 		error = SYSCTL_OUT(req, &old_osrel, sizeof(old_osrel));
3166 	return (error);
3167 }
3168 
3169 static int
3170 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS)
3171 {
3172 	int *name = (int *)arg1;
3173 	u_int namelen = arg2;
3174 	struct proc *p;
3175 	struct kinfo_sigtramp kst;
3176 	const struct sysentvec *sv;
3177 	int error;
3178 #ifdef COMPAT_FREEBSD32
3179 	struct kinfo_sigtramp32 kst32;
3180 #endif
3181 
3182 	if (namelen != 1)
3183 		return (EINVAL);
3184 
3185 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
3186 	if (error != 0)
3187 		return (error);
3188 	sv = p->p_sysent;
3189 #ifdef COMPAT_FREEBSD32
3190 	if ((req->flags & SCTL_MASK32) != 0) {
3191 		bzero(&kst32, sizeof(kst32));
3192 		if (SV_PROC_FLAG(p, SV_ILP32)) {
3193 			if (PROC_HAS_SHP(p)) {
3194 				kst32.ksigtramp_start = PROC_SIGCODE(p);
3195 				kst32.ksigtramp_end = kst32.ksigtramp_start +
3196 				    ((sv->sv_flags & SV_DSO_SIG) == 0 ?
3197 				    *sv->sv_szsigcode :
3198 				    (uintptr_t)sv->sv_szsigcode);
3199 			} else {
3200 				kst32.ksigtramp_start = PROC_PS_STRINGS(p) -
3201 				    *sv->sv_szsigcode;
3202 				kst32.ksigtramp_end = PROC_PS_STRINGS(p);
3203 			}
3204 		}
3205 		PROC_UNLOCK(p);
3206 		error = SYSCTL_OUT(req, &kst32, sizeof(kst32));
3207 		return (error);
3208 	}
3209 #endif
3210 	bzero(&kst, sizeof(kst));
3211 	if (PROC_HAS_SHP(p)) {
3212 		kst.ksigtramp_start = (char *)PROC_SIGCODE(p);
3213 		kst.ksigtramp_end = (char *)kst.ksigtramp_start +
3214 		    ((sv->sv_flags & SV_DSO_SIG) == 0 ? *sv->sv_szsigcode :
3215 		    (uintptr_t)sv->sv_szsigcode);
3216 	} else {
3217 		kst.ksigtramp_start = (char *)PROC_PS_STRINGS(p) -
3218 		    *sv->sv_szsigcode;
3219 		kst.ksigtramp_end = (char *)PROC_PS_STRINGS(p);
3220 	}
3221 	PROC_UNLOCK(p);
3222 	error = SYSCTL_OUT(req, &kst, sizeof(kst));
3223 	return (error);
3224 }
3225 
3226 static int
3227 sysctl_kern_proc_sigfastblk(SYSCTL_HANDLER_ARGS)
3228 {
3229 	int *name = (int *)arg1;
3230 	u_int namelen = arg2;
3231 	pid_t pid;
3232 	struct proc *p;
3233 	struct thread *td1;
3234 	uintptr_t addr;
3235 #ifdef COMPAT_FREEBSD32
3236 	uint32_t addr32;
3237 #endif
3238 	int error;
3239 
3240 	if (namelen != 1 || req->newptr != NULL)
3241 		return (EINVAL);
3242 
3243 	pid = (pid_t)name[0];
3244 	error = pget(pid, PGET_HOLD | PGET_NOTWEXIT | PGET_CANDEBUG, &p);
3245 	if (error != 0)
3246 		return (error);
3247 
3248 	PROC_LOCK(p);
3249 #ifdef COMPAT_FREEBSD32
3250 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3251 		if (!SV_PROC_FLAG(p, SV_ILP32)) {
3252 			error = EINVAL;
3253 			goto errlocked;
3254 		}
3255 	}
3256 #endif
3257 	if (pid <= PID_MAX) {
3258 		td1 = FIRST_THREAD_IN_PROC(p);
3259 	} else {
3260 		FOREACH_THREAD_IN_PROC(p, td1) {
3261 			if (td1->td_tid == pid)
3262 				break;
3263 		}
3264 	}
3265 	if (td1 == NULL) {
3266 		error = ESRCH;
3267 		goto errlocked;
3268 	}
3269 	/*
3270 	 * The access to the private thread flags.  It is fine as far
3271 	 * as no out-of-thin-air values are read from td_pflags, and
3272 	 * usermode read of the td_sigblock_ptr is racy inherently,
3273 	 * since target process might have already changed it
3274 	 * meantime.
3275 	 */
3276 	if ((td1->td_pflags & TDP_SIGFASTBLOCK) != 0)
3277 		addr = (uintptr_t)td1->td_sigblock_ptr;
3278 	else
3279 		error = ENOTTY;
3280 
3281 errlocked:
3282 	_PRELE(p);
3283 	PROC_UNLOCK(p);
3284 	if (error != 0)
3285 		return (error);
3286 
3287 #ifdef COMPAT_FREEBSD32
3288 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3289 		addr32 = addr;
3290 		error = SYSCTL_OUT(req, &addr32, sizeof(addr32));
3291 	} else
3292 #endif
3293 		error = SYSCTL_OUT(req, &addr, sizeof(addr));
3294 	return (error);
3295 }
3296 
3297 static int
3298 sysctl_kern_proc_vm_layout(SYSCTL_HANDLER_ARGS)
3299 {
3300 	struct kinfo_vm_layout kvm;
3301 	struct proc *p;
3302 	struct thread *td;
3303 	struct vmspace *vmspace;
3304 	int error, *name;
3305 
3306 	name = (int *)arg1;
3307 	if ((u_int)arg2 != 1)
3308 		return (EINVAL);
3309 
3310 	td = curthread;
3311 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
3312 	if (error != 0)
3313 		return (error);
3314 #ifdef COMPAT_FREEBSD32
3315 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3316 		if (!SV_PROC_FLAG(p, SV_ILP32)) {
3317 			PROC_UNLOCK(p);
3318 			return (EINVAL);
3319 		}
3320 	}
3321 #endif
3322 	_PHOLD(p);
3323 	PROC_UNLOCK(p);
3324 	error = proc_vmspace_ref(td, p, PRVM_CHECK_DEBUG, &vmspace);
3325 	if (error != 0) {
3326 		PRELE(p);
3327 		return (error);
3328 	}
3329 
3330 	memset(&kvm, 0, sizeof(kvm));
3331 	kvm.kvm_min_user_addr = vm_map_min(&vmspace->vm_map);
3332 	kvm.kvm_max_user_addr = vm_map_max(&vmspace->vm_map);
3333 	kvm.kvm_text_addr = (uintptr_t)vmspace->vm_taddr;
3334 	kvm.kvm_text_size = vmspace->vm_tsize;
3335 	kvm.kvm_data_addr = (uintptr_t)vmspace->vm_daddr;
3336 	kvm.kvm_data_size = vmspace->vm_dsize;
3337 	kvm.kvm_stack_addr = (uintptr_t)vmspace->vm_maxsaddr;
3338 	kvm.kvm_stack_size = vmspace->vm_ssize;
3339 	kvm.kvm_shp_addr = vmspace->vm_shp_base;
3340 	kvm.kvm_shp_size = p->p_sysent->sv_shared_page_len;
3341 	if ((vmspace->vm_map.flags & MAP_WIREFUTURE) != 0)
3342 		kvm.kvm_map_flags |= KMAP_FLAG_WIREFUTURE;
3343 	if ((vmspace->vm_map.flags & MAP_ASLR) != 0)
3344 		kvm.kvm_map_flags |= KMAP_FLAG_ASLR;
3345 	if ((vmspace->vm_map.flags & MAP_ASLR_IGNSTART) != 0)
3346 		kvm.kvm_map_flags |= KMAP_FLAG_ASLR_IGNSTART;
3347 	if ((vmspace->vm_map.flags & MAP_WXORX) != 0)
3348 		kvm.kvm_map_flags |= KMAP_FLAG_WXORX;
3349 	if ((vmspace->vm_map.flags & MAP_ASLR_STACK) != 0)
3350 		kvm.kvm_map_flags |= KMAP_FLAG_ASLR_STACK;
3351 	if (vmspace->vm_shp_base != p->p_sysent->sv_shared_page_base &&
3352 	    PROC_HAS_SHP(p))
3353 		kvm.kvm_map_flags |= KMAP_FLAG_ASLR_SHARED_PAGE;
3354 
3355 #ifdef COMPAT_FREEBSD32
3356 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3357 		struct kinfo_vm_layout32 kvm32;
3358 
3359 		memset(&kvm32, 0, sizeof(kvm32));
3360 		kvm32.kvm_min_user_addr = (uint32_t)kvm.kvm_min_user_addr;
3361 		kvm32.kvm_max_user_addr = (uint32_t)kvm.kvm_max_user_addr;
3362 		kvm32.kvm_text_addr = (uint32_t)kvm.kvm_text_addr;
3363 		kvm32.kvm_text_size = (uint32_t)kvm.kvm_text_size;
3364 		kvm32.kvm_data_addr = (uint32_t)kvm.kvm_data_addr;
3365 		kvm32.kvm_data_size = (uint32_t)kvm.kvm_data_size;
3366 		kvm32.kvm_stack_addr = (uint32_t)kvm.kvm_stack_addr;
3367 		kvm32.kvm_stack_size = (uint32_t)kvm.kvm_stack_size;
3368 		kvm32.kvm_shp_addr = (uint32_t)kvm.kvm_shp_addr;
3369 		kvm32.kvm_shp_size = (uint32_t)kvm.kvm_shp_size;
3370 		kvm32.kvm_map_flags = kvm.kvm_map_flags;
3371 		error = SYSCTL_OUT(req, &kvm32, sizeof(kvm32));
3372 		goto out;
3373 	}
3374 #endif
3375 
3376 	error = SYSCTL_OUT(req, &kvm, sizeof(kvm));
3377 #ifdef COMPAT_FREEBSD32
3378 out:
3379 #endif
3380 	proc_vmspace_unref(td, p, PRVM_CHECK_DEBUG, vmspace);
3381 	PRELE(p);
3382 	return (error);
3383 }
3384 
3385 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,  0,
3386     "Process table");
3387 
3388 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
3389 	CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
3390 	"Return entire process table");
3391 
3392 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3393 	sysctl_kern_proc, "Process table");
3394 
3395 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
3396 	sysctl_kern_proc, "Process table");
3397 
3398 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3399 	sysctl_kern_proc, "Process table");
3400 
3401 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
3402 	CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3403 
3404 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
3405 	sysctl_kern_proc, "Process table");
3406 
3407 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3408 	sysctl_kern_proc, "Process table");
3409 
3410 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3411 	sysctl_kern_proc, "Process table");
3412 
3413 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3414 	sysctl_kern_proc, "Process table");
3415 
3416 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
3417 	sysctl_kern_proc, "Return process table, no threads");
3418 
3419 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
3420 	CTLFLAG_RW | CTLFLAG_CAPWR | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
3421 	sysctl_kern_proc_args, "Process argument list");
3422 
3423 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
3424 	sysctl_kern_proc_env, "Process environment");
3425 
3426 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
3427 	CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
3428 
3429 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
3430 	CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
3431 
3432 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
3433 	CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
3434 	"Process syscall vector name (ABI type)");
3435 
3436 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
3437 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3438 
3439 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
3440 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3441 
3442 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
3443 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3444 
3445 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
3446 	sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3447 
3448 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
3449 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3450 
3451 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
3452 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3453 
3454 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
3455 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3456 
3457 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
3458 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3459 
3460 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
3461 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
3462 	"Return process table, including threads");
3463 
3464 #ifdef COMPAT_FREEBSD7
3465 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
3466 	CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
3467 #endif
3468 
3469 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
3470 	CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
3471 
3472 #if defined(STACK) || defined(DDB)
3473 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
3474 	CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
3475 #endif
3476 
3477 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
3478 	CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
3479 
3480 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
3481 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
3482 	"Process resource limits");
3483 
3484 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
3485 	CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
3486 	"Process ps_strings location");
3487 
3488 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
3489 	CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
3490 
3491 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
3492 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
3493 	"Process binary osreldate");
3494 
3495 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD |
3496 	CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp,
3497 	"Process signal trampoline location");
3498 
3499 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGFASTBLK, sigfastblk, CTLFLAG_RD |
3500 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_sigfastblk,
3501 	"Thread sigfastblock address");
3502 
3503 static SYSCTL_NODE(_kern_proc, KERN_PROC_VM_LAYOUT, vm_layout, CTLFLAG_RD |
3504 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_vm_layout,
3505 	"Process virtual address space layout info");
3506 
3507 static struct sx stop_all_proc_blocker;
3508 SX_SYSINIT(stop_all_proc_blocker, &stop_all_proc_blocker, "sapblk");
3509 
3510 bool
3511 stop_all_proc_block(void)
3512 {
3513 	return (sx_xlock_sig(&stop_all_proc_blocker) == 0);
3514 }
3515 
3516 void
3517 stop_all_proc_unblock(void)
3518 {
3519 	sx_xunlock(&stop_all_proc_blocker);
3520 }
3521 
3522 int allproc_gen;
3523 
3524 /*
3525  * stop_all_proc() purpose is to stop all process which have usermode,
3526  * except current process for obvious reasons.  This makes it somewhat
3527  * unreliable when invoked from multithreaded process.  The service
3528  * must not be user-callable anyway.
3529  */
3530 void
3531 stop_all_proc(void)
3532 {
3533 	struct proc *cp, *p;
3534 	int r, gen;
3535 	bool restart, seen_stopped, seen_exiting, stopped_some;
3536 
3537 	if (!stop_all_proc_block())
3538 		return;
3539 
3540 	cp = curproc;
3541 allproc_loop:
3542 	sx_xlock(&allproc_lock);
3543 	gen = allproc_gen;
3544 	seen_exiting = seen_stopped = stopped_some = restart = false;
3545 	LIST_REMOVE(cp, p_list);
3546 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3547 	for (;;) {
3548 		p = LIST_NEXT(cp, p_list);
3549 		if (p == NULL)
3550 			break;
3551 		LIST_REMOVE(cp, p_list);
3552 		LIST_INSERT_AFTER(p, cp, p_list);
3553 		PROC_LOCK(p);
3554 		if ((p->p_flag & (P_KPROC | P_SYSTEM | P_TOTAL_STOP |
3555 		    P_STOPPED_SIG)) != 0) {
3556 			PROC_UNLOCK(p);
3557 			continue;
3558 		}
3559 		if ((p->p_flag2 & P2_WEXIT) != 0) {
3560 			seen_exiting = true;
3561 			PROC_UNLOCK(p);
3562 			continue;
3563 		}
3564 		if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) {
3565 			/*
3566 			 * Stopped processes are tolerated when there
3567 			 * are no other processes which might continue
3568 			 * them.  P_STOPPED_SINGLE but not
3569 			 * P_TOTAL_STOP process still has at least one
3570 			 * thread running.
3571 			 */
3572 			seen_stopped = true;
3573 			PROC_UNLOCK(p);
3574 			continue;
3575 		}
3576 		if ((p->p_flag & P_TRACED) != 0) {
3577 			/*
3578 			 * thread_single() below cannot stop traced p,
3579 			 * so skip it.  OTOH, we cannot require
3580 			 * restart because debugger might be either
3581 			 * already stopped or traced as well.
3582 			 */
3583 			PROC_UNLOCK(p);
3584 			continue;
3585 		}
3586 		sx_xunlock(&allproc_lock);
3587 		_PHOLD(p);
3588 		r = thread_single(p, SINGLE_ALLPROC);
3589 		if (r != 0)
3590 			restart = true;
3591 		else
3592 			stopped_some = true;
3593 		_PRELE(p);
3594 		PROC_UNLOCK(p);
3595 		sx_xlock(&allproc_lock);
3596 	}
3597 	/* Catch forked children we did not see in iteration. */
3598 	if (gen != allproc_gen)
3599 		restart = true;
3600 	sx_xunlock(&allproc_lock);
3601 	if (restart || stopped_some || seen_exiting || seen_stopped) {
3602 		kern_yield(PRI_USER);
3603 		goto allproc_loop;
3604 	}
3605 }
3606 
3607 void
3608 resume_all_proc(void)
3609 {
3610 	struct proc *cp, *p;
3611 
3612 	cp = curproc;
3613 	sx_xlock(&allproc_lock);
3614 again:
3615 	LIST_REMOVE(cp, p_list);
3616 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3617 	for (;;) {
3618 		p = LIST_NEXT(cp, p_list);
3619 		if (p == NULL)
3620 			break;
3621 		LIST_REMOVE(cp, p_list);
3622 		LIST_INSERT_AFTER(p, cp, p_list);
3623 		PROC_LOCK(p);
3624 		if ((p->p_flag & P_TOTAL_STOP) != 0) {
3625 			sx_xunlock(&allproc_lock);
3626 			_PHOLD(p);
3627 			thread_single_end(p, SINGLE_ALLPROC);
3628 			_PRELE(p);
3629 			PROC_UNLOCK(p);
3630 			sx_xlock(&allproc_lock);
3631 		} else {
3632 			PROC_UNLOCK(p);
3633 		}
3634 	}
3635 	/*  Did the loop above missed any stopped process ? */
3636 	FOREACH_PROC_IN_SYSTEM(p) {
3637 		/* No need for proc lock. */
3638 		if ((p->p_flag & P_TOTAL_STOP) != 0)
3639 			goto again;
3640 	}
3641 	sx_xunlock(&allproc_lock);
3642 
3643 	stop_all_proc_unblock();
3644 }
3645 
3646 /* #define	TOTAL_STOP_DEBUG	1 */
3647 #ifdef TOTAL_STOP_DEBUG
3648 volatile static int ap_resume;
3649 #include <sys/mount.h>
3650 
3651 static int
3652 sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS)
3653 {
3654 	int error, val;
3655 
3656 	val = 0;
3657 	ap_resume = 0;
3658 	error = sysctl_handle_int(oidp, &val, 0, req);
3659 	if (error != 0 || req->newptr == NULL)
3660 		return (error);
3661 	if (val != 0) {
3662 		stop_all_proc();
3663 		syncer_suspend();
3664 		while (ap_resume == 0)
3665 			;
3666 		syncer_resume();
3667 		resume_all_proc();
3668 	}
3669 	return (0);
3670 }
3671 
3672 SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW |
3673     CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0,
3674     sysctl_debug_stop_all_proc, "I",
3675     "");
3676 #endif
3677