xref: /freebsd/sys/kern/kern_proc.c (revision 82397d791966b09d344251bc709cd9db2b3a1902)
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  *	@(#)kern_proc.c	8.7 (Berkeley) 2/14/95
32  */
33 
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
36 
37 #include "opt_ddb.h"
38 #include "opt_ktrace.h"
39 #include "opt_kstack_pages.h"
40 #include "opt_stack.h"
41 
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/bitstring.h>
45 #include <sys/elf.h>
46 #include <sys/eventhandler.h>
47 #include <sys/exec.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/proc.h>
58 #include <sys/ptrace.h>
59 #include <sys/refcount.h>
60 #include <sys/resourcevar.h>
61 #include <sys/rwlock.h>
62 #include <sys/sbuf.h>
63 #include <sys/sysent.h>
64 #include <sys/sched.h>
65 #include <sys/smp.h>
66 #include <sys/stack.h>
67 #include <sys/stat.h>
68 #include <sys/dtrace_bsd.h>
69 #include <sys/sysctl.h>
70 #include <sys/filedesc.h>
71 #include <sys/tty.h>
72 #include <sys/signalvar.h>
73 #include <sys/sdt.h>
74 #include <sys/sx.h>
75 #include <sys/user.h>
76 #include <sys/vnode.h>
77 #include <sys/wait.h>
78 
79 #ifdef DDB
80 #include <ddb/ddb.h>
81 #endif
82 
83 #include <vm/vm.h>
84 #include <vm/vm_param.h>
85 #include <vm/vm_extern.h>
86 #include <vm/pmap.h>
87 #include <vm/vm_map.h>
88 #include <vm/vm_object.h>
89 #include <vm/vm_page.h>
90 #include <vm/uma.h>
91 
92 #include <fs/devfs/devfs.h>
93 
94 #ifdef COMPAT_FREEBSD32
95 #include <compat/freebsd32/freebsd32.h>
96 #include <compat/freebsd32/freebsd32_util.h>
97 #endif
98 
99 SDT_PROVIDER_DEFINE(proc);
100 
101 MALLOC_DEFINE(M_PGRP, "pgrp", "process group header");
102 MALLOC_DEFINE(M_SESSION, "session", "session header");
103 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
104 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
105 
106 static void fixjobc_enterpgrp(struct proc *p, struct pgrp *pgrp);
107 static void doenterpgrp(struct proc *, struct pgrp *);
108 static void orphanpg(struct pgrp *pg);
109 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp);
110 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp);
111 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp,
112     int preferthread);
113 static void pgadjustjobc(struct pgrp *pgrp, bool entering);
114 static void pgdelete(struct pgrp *);
115 static int proc_ctor(void *mem, int size, void *arg, int flags);
116 static void proc_dtor(void *mem, int size, void *arg);
117 static int proc_init(void *mem, int size, int flags);
118 static void proc_fini(void *mem, int size);
119 static void pargs_free(struct pargs *pa);
120 
121 /*
122  * Other process lists
123  */
124 struct pidhashhead *pidhashtbl;
125 struct sx *pidhashtbl_lock;
126 u_long pidhash;
127 u_long pidhashlock;
128 struct pgrphashhead *pgrphashtbl;
129 u_long pgrphash;
130 struct proclist allproc;
131 struct sx __exclusive_cache_line allproc_lock;
132 struct sx __exclusive_cache_line proctree_lock;
133 struct mtx __exclusive_cache_line ppeers_lock;
134 struct mtx __exclusive_cache_line procid_lock;
135 uma_zone_t proc_zone;
136 
137 /*
138  * The offset of various fields in struct proc and struct thread.
139  * These are used by kernel debuggers to enumerate kernel threads and
140  * processes.
141  */
142 const int proc_off_p_pid = offsetof(struct proc, p_pid);
143 const int proc_off_p_comm = offsetof(struct proc, p_comm);
144 const int proc_off_p_list = offsetof(struct proc, p_list);
145 const int proc_off_p_hash = offsetof(struct proc, p_hash);
146 const int proc_off_p_threads = offsetof(struct proc, p_threads);
147 const int thread_off_td_tid = offsetof(struct thread, td_tid);
148 const int thread_off_td_name = offsetof(struct thread, td_name);
149 const int thread_off_td_oncpu = offsetof(struct thread, td_oncpu);
150 const int thread_off_td_pcb = offsetof(struct thread, td_pcb);
151 const int thread_off_td_plist = offsetof(struct thread, td_plist);
152 
153 EVENTHANDLER_LIST_DEFINE(process_ctor);
154 EVENTHANDLER_LIST_DEFINE(process_dtor);
155 EVENTHANDLER_LIST_DEFINE(process_init);
156 EVENTHANDLER_LIST_DEFINE(process_fini);
157 EVENTHANDLER_LIST_DEFINE(process_exit);
158 EVENTHANDLER_LIST_DEFINE(process_fork);
159 EVENTHANDLER_LIST_DEFINE(process_exec);
160 
161 int kstack_pages = KSTACK_PAGES;
162 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0,
163     "Kernel stack size in pages");
164 static int vmmap_skip_res_cnt = 0;
165 SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW,
166     &vmmap_skip_res_cnt, 0,
167     "Skip calculation of the pages resident count in kern.proc.vmmap");
168 
169 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
170 #ifdef COMPAT_FREEBSD32
171 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
172 #endif
173 
174 /*
175  * Initialize global process hashing structures.
176  */
177 void
178 procinit(void)
179 {
180 	u_long i;
181 
182 	sx_init(&allproc_lock, "allproc");
183 	sx_init(&proctree_lock, "proctree");
184 	mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
185 	mtx_init(&procid_lock, "procid", NULL, MTX_DEF);
186 	LIST_INIT(&allproc);
187 	pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
188 	pidhashlock = (pidhash + 1) / 64;
189 	if (pidhashlock > 0)
190 		pidhashlock--;
191 	pidhashtbl_lock = malloc(sizeof(*pidhashtbl_lock) * (pidhashlock + 1),
192 	    M_PROC, M_WAITOK | M_ZERO);
193 	for (i = 0; i < pidhashlock + 1; i++)
194 		sx_init_flags(&pidhashtbl_lock[i], "pidhash", SX_DUPOK);
195 	pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
196 	proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
197 	    proc_ctor, proc_dtor, proc_init, proc_fini,
198 	    UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
199 	uihashinit();
200 }
201 
202 /*
203  * Prepare a proc for use.
204  */
205 static int
206 proc_ctor(void *mem, int size, void *arg, int flags)
207 {
208 	struct proc *p;
209 	struct thread *td;
210 
211 	p = (struct proc *)mem;
212 #ifdef KDTRACE_HOOKS
213 	kdtrace_proc_ctor(p);
214 #endif
215 	EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
216 	td = FIRST_THREAD_IN_PROC(p);
217 	if (td != NULL) {
218 		/* Make sure all thread constructors are executed */
219 		EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
220 	}
221 	return (0);
222 }
223 
224 /*
225  * Reclaim a proc after use.
226  */
227 static void
228 proc_dtor(void *mem, int size, void *arg)
229 {
230 	struct proc *p;
231 	struct thread *td;
232 
233 	/* INVARIANTS checks go here */
234 	p = (struct proc *)mem;
235 	td = FIRST_THREAD_IN_PROC(p);
236 	if (td != NULL) {
237 #ifdef INVARIANTS
238 		KASSERT((p->p_numthreads == 1),
239 		    ("bad number of threads in exiting process"));
240 		KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
241 #endif
242 		/* Free all OSD associated to this thread. */
243 		osd_thread_exit(td);
244 		td_softdep_cleanup(td);
245 		MPASS(td->td_su == NULL);
246 
247 		/* Make sure all thread destructors are executed */
248 		EVENTHANDLER_DIRECT_INVOKE(thread_dtor, td);
249 	}
250 	EVENTHANDLER_DIRECT_INVOKE(process_dtor, p);
251 #ifdef KDTRACE_HOOKS
252 	kdtrace_proc_dtor(p);
253 #endif
254 	if (p->p_ksi != NULL)
255 		KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
256 }
257 
258 /*
259  * Initialize type-stable parts of a proc (when newly created).
260  */
261 static int
262 proc_init(void *mem, int size, int flags)
263 {
264 	struct proc *p;
265 
266 	p = (struct proc *)mem;
267 	mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK | MTX_NEW);
268 	mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_NEW);
269 	mtx_init(&p->p_statmtx, "pstatl", NULL, MTX_SPIN | MTX_NEW);
270 	mtx_init(&p->p_itimmtx, "pitiml", NULL, MTX_SPIN | MTX_NEW);
271 	mtx_init(&p->p_profmtx, "pprofl", NULL, MTX_SPIN | MTX_NEW);
272 	cv_init(&p->p_pwait, "ppwait");
273 	TAILQ_INIT(&p->p_threads);	     /* all threads in proc */
274 	EVENTHANDLER_DIRECT_INVOKE(process_init, p);
275 	p->p_stats = pstats_alloc();
276 	p->p_pgrp = NULL;
277 	return (0);
278 }
279 
280 /*
281  * UMA should ensure that this function is never called.
282  * Freeing a proc structure would violate type stability.
283  */
284 static void
285 proc_fini(void *mem, int size)
286 {
287 #ifdef notnow
288 	struct proc *p;
289 
290 	p = (struct proc *)mem;
291 	EVENTHANDLER_DIRECT_INVOKE(process_fini, p);
292 	pstats_free(p->p_stats);
293 	thread_free(FIRST_THREAD_IN_PROC(p));
294 	mtx_destroy(&p->p_mtx);
295 	if (p->p_ksi != NULL)
296 		ksiginfo_free(p->p_ksi);
297 #else
298 	panic("proc reclaimed");
299 #endif
300 }
301 
302 /*
303  * PID space management.
304  *
305  * These bitmaps are used by fork_findpid.
306  */
307 bitstr_t bit_decl(proc_id_pidmap, PID_MAX);
308 bitstr_t bit_decl(proc_id_grpidmap, PID_MAX);
309 bitstr_t bit_decl(proc_id_sessidmap, PID_MAX);
310 bitstr_t bit_decl(proc_id_reapmap, PID_MAX);
311 
312 static bitstr_t *proc_id_array[] = {
313 	proc_id_pidmap,
314 	proc_id_grpidmap,
315 	proc_id_sessidmap,
316 	proc_id_reapmap,
317 };
318 
319 void
320 proc_id_set(int type, pid_t id)
321 {
322 
323 	KASSERT(type >= 0 && type < nitems(proc_id_array),
324 	    ("invalid type %d\n", type));
325 	mtx_lock(&procid_lock);
326 	KASSERT(bit_test(proc_id_array[type], id) == 0,
327 	    ("bit %d already set in %d\n", id, type));
328 	bit_set(proc_id_array[type], id);
329 	mtx_unlock(&procid_lock);
330 }
331 
332 void
333 proc_id_set_cond(int type, pid_t id)
334 {
335 
336 	KASSERT(type >= 0 && type < nitems(proc_id_array),
337 	    ("invalid type %d\n", type));
338 	if (bit_test(proc_id_array[type], id))
339 		return;
340 	mtx_lock(&procid_lock);
341 	bit_set(proc_id_array[type], id);
342 	mtx_unlock(&procid_lock);
343 }
344 
345 void
346 proc_id_clear(int type, pid_t id)
347 {
348 
349 	KASSERT(type >= 0 && type < nitems(proc_id_array),
350 	    ("invalid type %d\n", type));
351 	mtx_lock(&procid_lock);
352 	KASSERT(bit_test(proc_id_array[type], id) != 0,
353 	    ("bit %d not set in %d\n", id, type));
354 	bit_clear(proc_id_array[type], id);
355 	mtx_unlock(&procid_lock);
356 }
357 
358 /*
359  * Is p an inferior of the current process?
360  */
361 int
362 inferior(struct proc *p)
363 {
364 
365 	sx_assert(&proctree_lock, SX_LOCKED);
366 	PROC_LOCK_ASSERT(p, MA_OWNED);
367 	for (; p != curproc; p = proc_realparent(p)) {
368 		if (p->p_pid == 0)
369 			return (0);
370 	}
371 	return (1);
372 }
373 
374 /*
375  * Shared lock all the pid hash lists.
376  */
377 void
378 pidhash_slockall(void)
379 {
380 	u_long i;
381 
382 	for (i = 0; i < pidhashlock + 1; i++)
383 		sx_slock(&pidhashtbl_lock[i]);
384 }
385 
386 /*
387  * Shared unlock all the pid hash lists.
388  */
389 void
390 pidhash_sunlockall(void)
391 {
392 	u_long i;
393 
394 	for (i = 0; i < pidhashlock + 1; i++)
395 		sx_sunlock(&pidhashtbl_lock[i]);
396 }
397 
398 /*
399  * Similar to pfind_any(), this function finds zombies.
400  */
401 struct proc *
402 pfind_any_locked(pid_t pid)
403 {
404 	struct proc *p;
405 
406 	sx_assert(PIDHASHLOCK(pid), SX_LOCKED);
407 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
408 		if (p->p_pid == pid) {
409 			PROC_LOCK(p);
410 			if (p->p_state == PRS_NEW) {
411 				PROC_UNLOCK(p);
412 				p = NULL;
413 			}
414 			break;
415 		}
416 	}
417 	return (p);
418 }
419 
420 /*
421  * Locate a process by number.
422  *
423  * By not returning processes in the PRS_NEW state, we allow callers to avoid
424  * testing for that condition to avoid dereferencing p_ucred, et al.
425  */
426 static __always_inline struct proc *
427 _pfind(pid_t pid, bool zombie)
428 {
429 	struct proc *p;
430 
431 	p = curproc;
432 	if (p->p_pid == pid) {
433 		PROC_LOCK(p);
434 		return (p);
435 	}
436 	sx_slock(PIDHASHLOCK(pid));
437 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
438 		if (p->p_pid == pid) {
439 			PROC_LOCK(p);
440 			if (p->p_state == PRS_NEW ||
441 			    (!zombie && p->p_state == PRS_ZOMBIE)) {
442 				PROC_UNLOCK(p);
443 				p = NULL;
444 			}
445 			break;
446 		}
447 	}
448 	sx_sunlock(PIDHASHLOCK(pid));
449 	return (p);
450 }
451 
452 struct proc *
453 pfind(pid_t pid)
454 {
455 
456 	return (_pfind(pid, false));
457 }
458 
459 /*
460  * Same as pfind but allow zombies.
461  */
462 struct proc *
463 pfind_any(pid_t pid)
464 {
465 
466 	return (_pfind(pid, true));
467 }
468 
469 /*
470  * Locate a process group by number.
471  * The caller must hold proctree_lock.
472  */
473 struct pgrp *
474 pgfind(pid_t pgid)
475 {
476 	struct pgrp *pgrp;
477 
478 	sx_assert(&proctree_lock, SX_LOCKED);
479 
480 	LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
481 		if (pgrp->pg_id == pgid) {
482 			PGRP_LOCK(pgrp);
483 			return (pgrp);
484 		}
485 	}
486 	return (NULL);
487 }
488 
489 /*
490  * Locate process and do additional manipulations, depending on flags.
491  */
492 int
493 pget(pid_t pid, int flags, struct proc **pp)
494 {
495 	struct proc *p;
496 	struct thread *td1;
497 	int error;
498 
499 	p = curproc;
500 	if (p->p_pid == pid) {
501 		PROC_LOCK(p);
502 	} else {
503 		p = NULL;
504 		if (pid <= PID_MAX) {
505 			if ((flags & PGET_NOTWEXIT) == 0)
506 				p = pfind_any(pid);
507 			else
508 				p = pfind(pid);
509 		} else if ((flags & PGET_NOTID) == 0) {
510 			td1 = tdfind(pid, -1);
511 			if (td1 != NULL)
512 				p = td1->td_proc;
513 		}
514 		if (p == NULL)
515 			return (ESRCH);
516 		if ((flags & PGET_CANSEE) != 0) {
517 			error = p_cansee(curthread, p);
518 			if (error != 0)
519 				goto errout;
520 		}
521 	}
522 	if ((flags & PGET_CANDEBUG) != 0) {
523 		error = p_candebug(curthread, p);
524 		if (error != 0)
525 			goto errout;
526 	}
527 	if ((flags & PGET_ISCURRENT) != 0 && curproc != p) {
528 		error = EPERM;
529 		goto errout;
530 	}
531 	if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) {
532 		error = ESRCH;
533 		goto errout;
534 	}
535 	if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) {
536 		/*
537 		 * XXXRW: Not clear ESRCH is the right error during proc
538 		 * execve().
539 		 */
540 		error = ESRCH;
541 		goto errout;
542 	}
543 	if ((flags & PGET_HOLD) != 0) {
544 		_PHOLD(p);
545 		PROC_UNLOCK(p);
546 	}
547 	*pp = p;
548 	return (0);
549 errout:
550 	PROC_UNLOCK(p);
551 	return (error);
552 }
553 
554 /*
555  * Create a new process group.
556  * pgid must be equal to the pid of p.
557  * Begin a new session if required.
558  */
559 int
560 enterpgrp(struct proc *p, pid_t pgid, struct pgrp *pgrp, struct session *sess)
561 {
562 
563 	sx_assert(&proctree_lock, SX_XLOCKED);
564 
565 	KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
566 	KASSERT(p->p_pid == pgid,
567 	    ("enterpgrp: new pgrp and pid != pgid"));
568 	KASSERT(pgfind(pgid) == NULL,
569 	    ("enterpgrp: pgrp with pgid exists"));
570 	KASSERT(!SESS_LEADER(p),
571 	    ("enterpgrp: session leader attempted setpgrp"));
572 
573 	mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
574 
575 	if (sess != NULL) {
576 		/*
577 		 * new session
578 		 */
579 		mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
580 		PROC_LOCK(p);
581 		p->p_flag &= ~P_CONTROLT;
582 		PROC_UNLOCK(p);
583 		PGRP_LOCK(pgrp);
584 		sess->s_leader = p;
585 		sess->s_sid = p->p_pid;
586 		proc_id_set(PROC_ID_SESSION, p->p_pid);
587 		refcount_init(&sess->s_count, 1);
588 		sess->s_ttyvp = NULL;
589 		sess->s_ttydp = NULL;
590 		sess->s_ttyp = NULL;
591 		bcopy(p->p_session->s_login, sess->s_login,
592 			    sizeof(sess->s_login));
593 		pgrp->pg_session = sess;
594 		KASSERT(p == curproc,
595 		    ("enterpgrp: mksession and p != curproc"));
596 	} else {
597 		pgrp->pg_session = p->p_session;
598 		sess_hold(pgrp->pg_session);
599 		PGRP_LOCK(pgrp);
600 	}
601 	pgrp->pg_id = pgid;
602 	proc_id_set(PROC_ID_GROUP, p->p_pid);
603 	LIST_INIT(&pgrp->pg_members);
604 
605 	/*
606 	 * As we have an exclusive lock of proctree_lock,
607 	 * this should not deadlock.
608 	 */
609 	LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
610 	pgrp->pg_jobc = 0;
611 	SLIST_INIT(&pgrp->pg_sigiolst);
612 	PGRP_UNLOCK(pgrp);
613 
614 	doenterpgrp(p, pgrp);
615 
616 	return (0);
617 }
618 
619 /*
620  * Move p to an existing process group
621  */
622 int
623 enterthispgrp(struct proc *p, struct pgrp *pgrp)
624 {
625 
626 	sx_assert(&proctree_lock, SX_XLOCKED);
627 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
628 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
629 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
630 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
631 	KASSERT(pgrp->pg_session == p->p_session,
632 		("%s: pgrp's session %p, p->p_session %p.\n",
633 		__func__,
634 		pgrp->pg_session,
635 		p->p_session));
636 	KASSERT(pgrp != p->p_pgrp,
637 		("%s: p belongs to pgrp.", __func__));
638 
639 	doenterpgrp(p, pgrp);
640 
641 	return (0);
642 }
643 
644 /*
645  * If true, any child of q which belongs to group pgrp, qualifies the
646  * process group pgrp as not orphaned.
647  */
648 static bool
649 isjobproc(struct proc *q, struct pgrp *pgrp)
650 {
651 	sx_assert(&proctree_lock, SX_LOCKED);
652 	return (q->p_pgrp != pgrp &&
653 	    q->p_pgrp->pg_session == pgrp->pg_session);
654 }
655 
656 static struct proc *
657 jobc_reaper(struct proc *p)
658 {
659 	struct proc *pp;
660 
661 	sx_assert(&proctree_lock, SX_LOCKED);
662 
663 	for (pp = p;;) {
664 		pp = pp->p_reaper;
665 		if (pp->p_reaper == pp ||
666 		    (pp->p_treeflag & P_TREE_GRPEXITED) == 0)
667 			return (pp);
668 	}
669 }
670 
671 static struct proc *
672 jobc_parent(struct proc *p)
673 {
674 	struct proc *pp;
675 
676 	sx_assert(&proctree_lock, SX_LOCKED);
677 
678 	pp = proc_realparent(p);
679 	if (pp->p_pptr == NULL ||
680 	    (pp->p_treeflag & P_TREE_GRPEXITED) == 0)
681 		return (pp);
682 	return (jobc_reaper(pp));
683 }
684 
685 #ifdef INVARIANTS
686 static void
687 check_pgrp_jobc(struct pgrp *pgrp)
688 {
689 	struct proc *q;
690 	int cnt;
691 
692 	sx_assert(&proctree_lock, SX_LOCKED);
693 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
694 
695 	cnt = 0;
696 	PGRP_LOCK(pgrp);
697 	LIST_FOREACH(q, &pgrp->pg_members, p_pglist) {
698 		if ((q->p_treeflag & P_TREE_GRPEXITED) != 0 ||
699 		    q->p_pptr == NULL)
700 			continue;
701 		if (isjobproc(jobc_parent(q), pgrp))
702 			cnt++;
703 	}
704 	KASSERT(pgrp->pg_jobc == cnt, ("pgrp %d %p pg_jobc %d cnt %d",
705 	    pgrp->pg_id, pgrp, pgrp->pg_jobc, cnt));
706 	PGRP_UNLOCK(pgrp);
707 }
708 #endif
709 
710 /*
711  * Move p to a process group
712  */
713 static void
714 doenterpgrp(struct proc *p, struct pgrp *pgrp)
715 {
716 	struct pgrp *savepgrp;
717 
718 	sx_assert(&proctree_lock, SX_XLOCKED);
719 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
720 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
721 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
722 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
723 
724 	savepgrp = p->p_pgrp;
725 
726 #ifdef INVARIANTS
727 	check_pgrp_jobc(pgrp);
728 	check_pgrp_jobc(savepgrp);
729 #endif
730 
731 	/*
732 	 * Adjust eligibility of affected pgrps to participate in job control.
733 	 */
734 	fixjobc_enterpgrp(p, pgrp);
735 
736 	PGRP_LOCK(pgrp);
737 	PGRP_LOCK(savepgrp);
738 	PROC_LOCK(p);
739 	LIST_REMOVE(p, p_pglist);
740 	p->p_pgrp = pgrp;
741 	PROC_UNLOCK(p);
742 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
743 	PGRP_UNLOCK(savepgrp);
744 	PGRP_UNLOCK(pgrp);
745 	if (LIST_EMPTY(&savepgrp->pg_members))
746 		pgdelete(savepgrp);
747 }
748 
749 /*
750  * remove process from process group
751  */
752 int
753 leavepgrp(struct proc *p)
754 {
755 	struct pgrp *savepgrp;
756 
757 	sx_assert(&proctree_lock, SX_XLOCKED);
758 	savepgrp = p->p_pgrp;
759 	PGRP_LOCK(savepgrp);
760 	PROC_LOCK(p);
761 	LIST_REMOVE(p, p_pglist);
762 	p->p_pgrp = NULL;
763 	PROC_UNLOCK(p);
764 	PGRP_UNLOCK(savepgrp);
765 	if (LIST_EMPTY(&savepgrp->pg_members))
766 		pgdelete(savepgrp);
767 	return (0);
768 }
769 
770 /*
771  * delete a process group
772  */
773 static void
774 pgdelete(struct pgrp *pgrp)
775 {
776 	struct session *savesess;
777 	struct tty *tp;
778 
779 	sx_assert(&proctree_lock, SX_XLOCKED);
780 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
781 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
782 
783 	/*
784 	 * Reset any sigio structures pointing to us as a result of
785 	 * F_SETOWN with our pgid.  The proctree lock ensures that
786 	 * new sigio structures will not be added after this point.
787 	 */
788 	funsetownlst(&pgrp->pg_sigiolst);
789 
790 	PGRP_LOCK(pgrp);
791 	tp = pgrp->pg_session->s_ttyp;
792 	LIST_REMOVE(pgrp, pg_hash);
793 	savesess = pgrp->pg_session;
794 	PGRP_UNLOCK(pgrp);
795 
796 	/* Remove the reference to the pgrp before deallocating it. */
797 	if (tp != NULL) {
798 		tty_lock(tp);
799 		tty_rel_pgrp(tp, pgrp);
800 	}
801 
802 	proc_id_clear(PROC_ID_GROUP, pgrp->pg_id);
803 	mtx_destroy(&pgrp->pg_mtx);
804 	free(pgrp, M_PGRP);
805 	sess_release(savesess);
806 }
807 
808 static void
809 pgadjustjobc(struct pgrp *pgrp, bool entering)
810 {
811 
812 	PGRP_LOCK(pgrp);
813 	if (entering) {
814 		MPASS(pgrp->pg_jobc >= 0);
815 		pgrp->pg_jobc++;
816 	} else {
817 		MPASS(pgrp->pg_jobc > 0);
818 		--pgrp->pg_jobc;
819 		if (pgrp->pg_jobc == 0)
820 			orphanpg(pgrp);
821 	}
822 	PGRP_UNLOCK(pgrp);
823 }
824 
825 static void
826 fixjobc_enterpgrp_q(struct pgrp *pgrp, struct proc *p, struct proc *q, bool adj)
827 {
828 	struct pgrp *childpgrp;
829 	bool future_jobc;
830 
831 	sx_assert(&proctree_lock, SX_LOCKED);
832 
833 	if ((q->p_treeflag & P_TREE_GRPEXITED) != 0)
834 		return;
835 	childpgrp = q->p_pgrp;
836 	future_jobc = childpgrp != pgrp &&
837 	    childpgrp->pg_session == pgrp->pg_session;
838 
839 	if ((adj && !isjobproc(p, childpgrp) && future_jobc) ||
840 	    (!adj && isjobproc(p, childpgrp) && !future_jobc))
841 		pgadjustjobc(childpgrp, adj);
842 }
843 
844 /*
845  * Adjust pgrp jobc counters when specified process changes process group.
846  * We count the number of processes in each process group that "qualify"
847  * the group for terminal job control (those with a parent in a different
848  * process group of the same session).  If that count reaches zero, the
849  * process group becomes orphaned.  Check both the specified process'
850  * process group and that of its children.
851  * We increment eligibility counts before decrementing, otherwise we
852  * could reach 0 spuriously during the decrement.
853  */
854 static void
855 fixjobc_enterpgrp(struct proc *p, struct pgrp *pgrp)
856 {
857 	struct proc *q;
858 
859 	sx_assert(&proctree_lock, SX_LOCKED);
860 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
861 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
862 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
863 
864 	if (p->p_pgrp == pgrp)
865 		return;
866 
867 	if (isjobproc(jobc_parent(p), pgrp))
868 		pgadjustjobc(pgrp, true);
869 	LIST_FOREACH(q, &p->p_children, p_sibling) {
870 		if ((q->p_treeflag & P_TREE_ORPHANED) != 0)
871 			continue;
872 		fixjobc_enterpgrp_q(pgrp, p, q, true);
873 	}
874 	LIST_FOREACH(q, &p->p_orphans, p_orphan)
875 		fixjobc_enterpgrp_q(pgrp, p, q, true);
876 
877 	if (isjobproc(jobc_parent(p), p->p_pgrp))
878 		pgadjustjobc(p->p_pgrp, false);
879 	LIST_FOREACH(q, &p->p_children, p_sibling) {
880 		if ((q->p_treeflag & P_TREE_ORPHANED) != 0)
881 			continue;
882 		fixjobc_enterpgrp_q(pgrp, p, q, false);
883 	}
884 	LIST_FOREACH(q, &p->p_orphans, p_orphan)
885 		fixjobc_enterpgrp_q(pgrp, p, q, false);
886 }
887 
888 static void
889 fixjobc_kill_q(struct proc *p, struct proc *q, bool adj)
890 {
891 	struct pgrp *childpgrp;
892 
893 	sx_assert(&proctree_lock, SX_LOCKED);
894 
895 	if ((q->p_treeflag & P_TREE_GRPEXITED) != 0)
896 		return;
897 	childpgrp = q->p_pgrp;
898 
899 	if ((adj && isjobproc(jobc_reaper(q), childpgrp) &&
900 	    !isjobproc(p, childpgrp)) || (!adj && !isjobproc(jobc_reaper(q),
901 	    childpgrp) && isjobproc(p, childpgrp)))
902 		pgadjustjobc(childpgrp, adj);
903 }
904 
905 static void
906 fixjobc_kill(struct proc *p)
907 {
908 	struct proc *q;
909 	struct pgrp *pgrp;
910 
911 	sx_assert(&proctree_lock, SX_LOCKED);
912 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
913 	pgrp = p->p_pgrp;
914 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
915 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
916 #ifdef INVARIANTS
917 	check_pgrp_jobc(pgrp);
918 #endif
919 
920 	/*
921 	 * p no longer affects process group orphanage for children.
922 	 * It is marked by the flag because p is only physically
923 	 * removed from its process group on wait(2).
924 	 */
925 	MPASS((p->p_treeflag & P_TREE_GRPEXITED) == 0);
926 	p->p_treeflag |= P_TREE_GRPEXITED;
927 
928 	/*
929 	 * Check p's parent to see whether p qualifies its own process
930 	 * group; if so, adjust count for p's process group.
931 	 */
932 	if (isjobproc(jobc_parent(p), pgrp))
933 		pgadjustjobc(pgrp, false);
934 
935 	/*
936 	 * Check this process' children to see whether they qualify
937 	 * their process groups after reparenting to reaper.  If so,
938 	 * adjust counts for children's process groups.
939 	 */
940 	LIST_FOREACH(q, &p->p_children, p_sibling) {
941 		if ((q->p_treeflag & P_TREE_ORPHANED) != 0)
942 			continue;
943 		fixjobc_kill_q(p, q, true);
944 	}
945 	LIST_FOREACH(q, &p->p_orphans, p_orphan)
946 		fixjobc_kill_q(p, q, true);
947 	LIST_FOREACH(q, &p->p_children, p_sibling) {
948 		if ((q->p_treeflag & P_TREE_ORPHANED) != 0)
949 			continue;
950 		fixjobc_kill_q(p, q, false);
951 	}
952 	LIST_FOREACH(q, &p->p_orphans, p_orphan)
953 		fixjobc_kill_q(p, q, false);
954 
955 #ifdef INVARIANTS
956 	check_pgrp_jobc(pgrp);
957 #endif
958 }
959 
960 void
961 killjobc(void)
962 {
963 	struct session *sp;
964 	struct tty *tp;
965 	struct proc *p;
966 	struct vnode *ttyvp;
967 
968 	p = curproc;
969 	MPASS(p->p_flag & P_WEXIT);
970 	sx_assert(&proctree_lock, SX_LOCKED);
971 
972 	if (SESS_LEADER(p)) {
973 		sp = p->p_session;
974 
975 		/*
976 		 * s_ttyp is not zero'd; we use this to indicate that
977 		 * the session once had a controlling terminal. (for
978 		 * logging and informational purposes)
979 		 */
980 		SESS_LOCK(sp);
981 		ttyvp = sp->s_ttyvp;
982 		tp = sp->s_ttyp;
983 		sp->s_ttyvp = NULL;
984 		sp->s_ttydp = NULL;
985 		sp->s_leader = NULL;
986 		SESS_UNLOCK(sp);
987 
988 		/*
989 		 * Signal foreground pgrp and revoke access to
990 		 * controlling terminal if it has not been revoked
991 		 * already.
992 		 *
993 		 * Because the TTY may have been revoked in the mean
994 		 * time and could already have a new session associated
995 		 * with it, make sure we don't send a SIGHUP to a
996 		 * foreground process group that does not belong to this
997 		 * session.
998 		 */
999 
1000 		if (tp != NULL) {
1001 			tty_lock(tp);
1002 			if (tp->t_session == sp)
1003 				tty_signal_pgrp(tp, SIGHUP);
1004 			tty_unlock(tp);
1005 		}
1006 
1007 		if (ttyvp != NULL) {
1008 			sx_xunlock(&proctree_lock);
1009 			if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) {
1010 				VOP_REVOKE(ttyvp, REVOKEALL);
1011 				VOP_UNLOCK(ttyvp);
1012 			}
1013 			devfs_ctty_unref(ttyvp);
1014 			sx_xlock(&proctree_lock);
1015 		}
1016 	}
1017 	fixjobc_kill(p);
1018 }
1019 
1020 /*
1021  * A process group has become orphaned;
1022  * if there are any stopped processes in the group,
1023  * hang-up all process in that group.
1024  */
1025 static void
1026 orphanpg(struct pgrp *pg)
1027 {
1028 	struct proc *p;
1029 
1030 	PGRP_LOCK_ASSERT(pg, MA_OWNED);
1031 
1032 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
1033 		PROC_LOCK(p);
1034 		if (P_SHOULDSTOP(p) == P_STOPPED_SIG) {
1035 			PROC_UNLOCK(p);
1036 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
1037 				PROC_LOCK(p);
1038 				kern_psignal(p, SIGHUP);
1039 				kern_psignal(p, SIGCONT);
1040 				PROC_UNLOCK(p);
1041 			}
1042 			return;
1043 		}
1044 		PROC_UNLOCK(p);
1045 	}
1046 }
1047 
1048 void
1049 sess_hold(struct session *s)
1050 {
1051 
1052 	refcount_acquire(&s->s_count);
1053 }
1054 
1055 void
1056 sess_release(struct session *s)
1057 {
1058 
1059 	if (refcount_release(&s->s_count)) {
1060 		if (s->s_ttyp != NULL) {
1061 			tty_lock(s->s_ttyp);
1062 			tty_rel_sess(s->s_ttyp, s);
1063 		}
1064 		proc_id_clear(PROC_ID_SESSION, s->s_sid);
1065 		mtx_destroy(&s->s_mtx);
1066 		free(s, M_SESSION);
1067 	}
1068 }
1069 
1070 #ifdef DDB
1071 
1072 static void
1073 db_print_pgrp_one(struct pgrp *pgrp, struct proc *p)
1074 {
1075 	db_printf(
1076 	    "    pid %d at %p pr %d pgrp %p e %d jc %d\n",
1077 	    p->p_pid, p, p->p_pptr == NULL ? -1 : p->p_pptr->p_pid,
1078 	    p->p_pgrp, (p->p_treeflag & P_TREE_GRPEXITED) != 0,
1079 	    p->p_pptr == NULL ? 0 : isjobproc(p->p_pptr, pgrp));
1080 }
1081 
1082 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
1083 {
1084 	struct pgrp *pgrp;
1085 	struct proc *p;
1086 	int i;
1087 
1088 	for (i = 0; i <= pgrphash; i++) {
1089 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
1090 			db_printf("indx %d\n", i);
1091 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
1092 				db_printf(
1093 			"  pgrp %p, pgid %d, sess %p, sesscnt %d, mem %p\n",
1094 				    pgrp, (int)pgrp->pg_id, pgrp->pg_session,
1095 				    pgrp->pg_session->s_count,
1096 				    LIST_FIRST(&pgrp->pg_members));
1097 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist)
1098 					db_print_pgrp_one(pgrp, p);
1099 			}
1100 		}
1101 	}
1102 }
1103 #endif /* DDB */
1104 
1105 /*
1106  * Calculate the kinfo_proc members which contain process-wide
1107  * informations.
1108  * Must be called with the target process locked.
1109  */
1110 static void
1111 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
1112 {
1113 	struct thread *td;
1114 
1115 	PROC_LOCK_ASSERT(p, MA_OWNED);
1116 
1117 	kp->ki_estcpu = 0;
1118 	kp->ki_pctcpu = 0;
1119 	FOREACH_THREAD_IN_PROC(p, td) {
1120 		thread_lock(td);
1121 		kp->ki_pctcpu += sched_pctcpu(td);
1122 		kp->ki_estcpu += sched_estcpu(td);
1123 		thread_unlock(td);
1124 	}
1125 }
1126 
1127 /*
1128  * Clear kinfo_proc and fill in any information that is common
1129  * to all threads in the process.
1130  * Must be called with the target process locked.
1131  */
1132 static void
1133 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
1134 {
1135 	struct thread *td0;
1136 	struct tty *tp;
1137 	struct session *sp;
1138 	struct ucred *cred;
1139 	struct sigacts *ps;
1140 	struct timeval boottime;
1141 
1142 	PROC_LOCK_ASSERT(p, MA_OWNED);
1143 	bzero(kp, sizeof(*kp));
1144 
1145 	kp->ki_structsize = sizeof(*kp);
1146 	kp->ki_paddr = p;
1147 	kp->ki_addr =/* p->p_addr; */0; /* XXX */
1148 	kp->ki_args = p->p_args;
1149 	kp->ki_textvp = p->p_textvp;
1150 #ifdef KTRACE
1151 	kp->ki_tracep = p->p_tracevp;
1152 	kp->ki_traceflag = p->p_traceflag;
1153 #endif
1154 	kp->ki_fd = p->p_fd;
1155 	kp->ki_pd = p->p_pd;
1156 	kp->ki_vmspace = p->p_vmspace;
1157 	kp->ki_flag = p->p_flag;
1158 	kp->ki_flag2 = p->p_flag2;
1159 	cred = p->p_ucred;
1160 	if (cred) {
1161 		kp->ki_uid = cred->cr_uid;
1162 		kp->ki_ruid = cred->cr_ruid;
1163 		kp->ki_svuid = cred->cr_svuid;
1164 		kp->ki_cr_flags = 0;
1165 		if (cred->cr_flags & CRED_FLAG_CAPMODE)
1166 			kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
1167 		/* XXX bde doesn't like KI_NGROUPS */
1168 		if (cred->cr_ngroups > KI_NGROUPS) {
1169 			kp->ki_ngroups = KI_NGROUPS;
1170 			kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
1171 		} else
1172 			kp->ki_ngroups = cred->cr_ngroups;
1173 		bcopy(cred->cr_groups, kp->ki_groups,
1174 		    kp->ki_ngroups * sizeof(gid_t));
1175 		kp->ki_rgid = cred->cr_rgid;
1176 		kp->ki_svgid = cred->cr_svgid;
1177 		/* If jailed(cred), emulate the old P_JAILED flag. */
1178 		if (jailed(cred)) {
1179 			kp->ki_flag |= P_JAILED;
1180 			/* If inside the jail, use 0 as a jail ID. */
1181 			if (cred->cr_prison != curthread->td_ucred->cr_prison)
1182 				kp->ki_jid = cred->cr_prison->pr_id;
1183 		}
1184 		strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
1185 		    sizeof(kp->ki_loginclass));
1186 	}
1187 	ps = p->p_sigacts;
1188 	if (ps) {
1189 		mtx_lock(&ps->ps_mtx);
1190 		kp->ki_sigignore = ps->ps_sigignore;
1191 		kp->ki_sigcatch = ps->ps_sigcatch;
1192 		mtx_unlock(&ps->ps_mtx);
1193 	}
1194 	if (p->p_state != PRS_NEW &&
1195 	    p->p_state != PRS_ZOMBIE &&
1196 	    p->p_vmspace != NULL) {
1197 		struct vmspace *vm = p->p_vmspace;
1198 
1199 		kp->ki_size = vm->vm_map.size;
1200 		kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
1201 		FOREACH_THREAD_IN_PROC(p, td0) {
1202 			if (!TD_IS_SWAPPED(td0))
1203 				kp->ki_rssize += td0->td_kstack_pages;
1204 		}
1205 		kp->ki_swrss = vm->vm_swrss;
1206 		kp->ki_tsize = vm->vm_tsize;
1207 		kp->ki_dsize = vm->vm_dsize;
1208 		kp->ki_ssize = vm->vm_ssize;
1209 	} else if (p->p_state == PRS_ZOMBIE)
1210 		kp->ki_stat = SZOMB;
1211 	if (kp->ki_flag & P_INMEM)
1212 		kp->ki_sflag = PS_INMEM;
1213 	else
1214 		kp->ki_sflag = 0;
1215 	/* Calculate legacy swtime as seconds since 'swtick'. */
1216 	kp->ki_swtime = (ticks - p->p_swtick) / hz;
1217 	kp->ki_pid = p->p_pid;
1218 	kp->ki_nice = p->p_nice;
1219 	kp->ki_fibnum = p->p_fibnum;
1220 	kp->ki_start = p->p_stats->p_start;
1221 	getboottime(&boottime);
1222 	timevaladd(&kp->ki_start, &boottime);
1223 	PROC_STATLOCK(p);
1224 	rufetch(p, &kp->ki_rusage);
1225 	kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
1226 	calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
1227 	PROC_STATUNLOCK(p);
1228 	calccru(p, &kp->ki_childutime, &kp->ki_childstime);
1229 	/* Some callers want child times in a single value. */
1230 	kp->ki_childtime = kp->ki_childstime;
1231 	timevaladd(&kp->ki_childtime, &kp->ki_childutime);
1232 
1233 	FOREACH_THREAD_IN_PROC(p, td0)
1234 		kp->ki_cow += td0->td_cow;
1235 
1236 	tp = NULL;
1237 	if (p->p_pgrp) {
1238 		kp->ki_pgid = p->p_pgrp->pg_id;
1239 		kp->ki_jobc = p->p_pgrp->pg_jobc;
1240 		sp = p->p_pgrp->pg_session;
1241 
1242 		if (sp != NULL) {
1243 			kp->ki_sid = sp->s_sid;
1244 			SESS_LOCK(sp);
1245 			strlcpy(kp->ki_login, sp->s_login,
1246 			    sizeof(kp->ki_login));
1247 			if (sp->s_ttyvp)
1248 				kp->ki_kiflag |= KI_CTTY;
1249 			if (SESS_LEADER(p))
1250 				kp->ki_kiflag |= KI_SLEADER;
1251 			/* XXX proctree_lock */
1252 			tp = sp->s_ttyp;
1253 			SESS_UNLOCK(sp);
1254 		}
1255 	}
1256 	if ((p->p_flag & P_CONTROLT) && tp != NULL) {
1257 		kp->ki_tdev = tty_udev(tp);
1258 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1259 		kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
1260 		if (tp->t_session)
1261 			kp->ki_tsid = tp->t_session->s_sid;
1262 	} else {
1263 		kp->ki_tdev = NODEV;
1264 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1265 	}
1266 	if (p->p_comm[0] != '\0')
1267 		strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
1268 	if (p->p_sysent && p->p_sysent->sv_name != NULL &&
1269 	    p->p_sysent->sv_name[0] != '\0')
1270 		strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
1271 	kp->ki_siglist = p->p_siglist;
1272 	kp->ki_xstat = KW_EXITCODE(p->p_xexit, p->p_xsig);
1273 	kp->ki_acflag = p->p_acflag;
1274 	kp->ki_lock = p->p_lock;
1275 	if (p->p_pptr) {
1276 		kp->ki_ppid = p->p_oppid;
1277 		if (p->p_flag & P_TRACED)
1278 			kp->ki_tracer = p->p_pptr->p_pid;
1279 	}
1280 }
1281 
1282 /*
1283  * Fill in information that is thread specific.  Must be called with
1284  * target process locked.  If 'preferthread' is set, overwrite certain
1285  * process-related fields that are maintained for both threads and
1286  * processes.
1287  */
1288 static void
1289 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
1290 {
1291 	struct proc *p;
1292 
1293 	p = td->td_proc;
1294 	kp->ki_tdaddr = td;
1295 	PROC_LOCK_ASSERT(p, MA_OWNED);
1296 
1297 	if (preferthread)
1298 		PROC_STATLOCK(p);
1299 	thread_lock(td);
1300 	if (td->td_wmesg != NULL)
1301 		strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
1302 	else
1303 		bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
1304 	if (strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname)) >=
1305 	    sizeof(kp->ki_tdname)) {
1306 		strlcpy(kp->ki_moretdname,
1307 		    td->td_name + sizeof(kp->ki_tdname) - 1,
1308 		    sizeof(kp->ki_moretdname));
1309 	} else {
1310 		bzero(kp->ki_moretdname, sizeof(kp->ki_moretdname));
1311 	}
1312 	if (TD_ON_LOCK(td)) {
1313 		kp->ki_kiflag |= KI_LOCKBLOCK;
1314 		strlcpy(kp->ki_lockname, td->td_lockname,
1315 		    sizeof(kp->ki_lockname));
1316 	} else {
1317 		kp->ki_kiflag &= ~KI_LOCKBLOCK;
1318 		bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
1319 	}
1320 
1321 	if (p->p_state == PRS_NORMAL) { /* approximate. */
1322 		if (TD_ON_RUNQ(td) ||
1323 		    TD_CAN_RUN(td) ||
1324 		    TD_IS_RUNNING(td)) {
1325 			kp->ki_stat = SRUN;
1326 		} else if (P_SHOULDSTOP(p)) {
1327 			kp->ki_stat = SSTOP;
1328 		} else if (TD_IS_SLEEPING(td)) {
1329 			kp->ki_stat = SSLEEP;
1330 		} else if (TD_ON_LOCK(td)) {
1331 			kp->ki_stat = SLOCK;
1332 		} else {
1333 			kp->ki_stat = SWAIT;
1334 		}
1335 	} else if (p->p_state == PRS_ZOMBIE) {
1336 		kp->ki_stat = SZOMB;
1337 	} else {
1338 		kp->ki_stat = SIDL;
1339 	}
1340 
1341 	/* Things in the thread */
1342 	kp->ki_wchan = td->td_wchan;
1343 	kp->ki_pri.pri_level = td->td_priority;
1344 	kp->ki_pri.pri_native = td->td_base_pri;
1345 
1346 	/*
1347 	 * Note: legacy fields; clamp at the old NOCPU value and/or
1348 	 * the maximum u_char CPU value.
1349 	 */
1350 	if (td->td_lastcpu == NOCPU)
1351 		kp->ki_lastcpu_old = NOCPU_OLD;
1352 	else if (td->td_lastcpu > MAXCPU_OLD)
1353 		kp->ki_lastcpu_old = MAXCPU_OLD;
1354 	else
1355 		kp->ki_lastcpu_old = td->td_lastcpu;
1356 
1357 	if (td->td_oncpu == NOCPU)
1358 		kp->ki_oncpu_old = NOCPU_OLD;
1359 	else if (td->td_oncpu > MAXCPU_OLD)
1360 		kp->ki_oncpu_old = MAXCPU_OLD;
1361 	else
1362 		kp->ki_oncpu_old = td->td_oncpu;
1363 
1364 	kp->ki_lastcpu = td->td_lastcpu;
1365 	kp->ki_oncpu = td->td_oncpu;
1366 	kp->ki_tdflags = td->td_flags;
1367 	kp->ki_tid = td->td_tid;
1368 	kp->ki_numthreads = p->p_numthreads;
1369 	kp->ki_pcb = td->td_pcb;
1370 	kp->ki_kstack = (void *)td->td_kstack;
1371 	kp->ki_slptime = (ticks - td->td_slptick) / hz;
1372 	kp->ki_pri.pri_class = td->td_pri_class;
1373 	kp->ki_pri.pri_user = td->td_user_pri;
1374 
1375 	if (preferthread) {
1376 		rufetchtd(td, &kp->ki_rusage);
1377 		kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
1378 		kp->ki_pctcpu = sched_pctcpu(td);
1379 		kp->ki_estcpu = sched_estcpu(td);
1380 		kp->ki_cow = td->td_cow;
1381 	}
1382 
1383 	/* We can't get this anymore but ps etc never used it anyway. */
1384 	kp->ki_rqindex = 0;
1385 
1386 	if (preferthread)
1387 		kp->ki_siglist = td->td_siglist;
1388 	kp->ki_sigmask = td->td_sigmask;
1389 	thread_unlock(td);
1390 	if (preferthread)
1391 		PROC_STATUNLOCK(p);
1392 }
1393 
1394 /*
1395  * Fill in a kinfo_proc structure for the specified process.
1396  * Must be called with the target process locked.
1397  */
1398 void
1399 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1400 {
1401 
1402 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1403 
1404 	fill_kinfo_proc_only(p, kp);
1405 	fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1406 	fill_kinfo_aggregate(p, kp);
1407 }
1408 
1409 struct pstats *
1410 pstats_alloc(void)
1411 {
1412 
1413 	return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1414 }
1415 
1416 /*
1417  * Copy parts of p_stats; zero the rest of p_stats (statistics).
1418  */
1419 void
1420 pstats_fork(struct pstats *src, struct pstats *dst)
1421 {
1422 
1423 	bzero(&dst->pstat_startzero,
1424 	    __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1425 	bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1426 	    __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1427 }
1428 
1429 void
1430 pstats_free(struct pstats *ps)
1431 {
1432 
1433 	free(ps, M_SUBPROC);
1434 }
1435 
1436 #ifdef COMPAT_FREEBSD32
1437 
1438 /*
1439  * This function is typically used to copy out the kernel address, so
1440  * it can be replaced by assignment of zero.
1441  */
1442 static inline uint32_t
1443 ptr32_trim(const void *ptr)
1444 {
1445 	uintptr_t uptr;
1446 
1447 	uptr = (uintptr_t)ptr;
1448 	return ((uptr > UINT_MAX) ? 0 : uptr);
1449 }
1450 
1451 #define PTRTRIM_CP(src,dst,fld) \
1452 	do { (dst).fld = ptr32_trim((src).fld); } while (0)
1453 
1454 static void
1455 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1456 {
1457 	int i;
1458 
1459 	bzero(ki32, sizeof(struct kinfo_proc32));
1460 	ki32->ki_structsize = sizeof(struct kinfo_proc32);
1461 	CP(*ki, *ki32, ki_layout);
1462 	PTRTRIM_CP(*ki, *ki32, ki_args);
1463 	PTRTRIM_CP(*ki, *ki32, ki_paddr);
1464 	PTRTRIM_CP(*ki, *ki32, ki_addr);
1465 	PTRTRIM_CP(*ki, *ki32, ki_tracep);
1466 	PTRTRIM_CP(*ki, *ki32, ki_textvp);
1467 	PTRTRIM_CP(*ki, *ki32, ki_fd);
1468 	PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1469 	PTRTRIM_CP(*ki, *ki32, ki_wchan);
1470 	CP(*ki, *ki32, ki_pid);
1471 	CP(*ki, *ki32, ki_ppid);
1472 	CP(*ki, *ki32, ki_pgid);
1473 	CP(*ki, *ki32, ki_tpgid);
1474 	CP(*ki, *ki32, ki_sid);
1475 	CP(*ki, *ki32, ki_tsid);
1476 	CP(*ki, *ki32, ki_jobc);
1477 	CP(*ki, *ki32, ki_tdev);
1478 	CP(*ki, *ki32, ki_tdev_freebsd11);
1479 	CP(*ki, *ki32, ki_siglist);
1480 	CP(*ki, *ki32, ki_sigmask);
1481 	CP(*ki, *ki32, ki_sigignore);
1482 	CP(*ki, *ki32, ki_sigcatch);
1483 	CP(*ki, *ki32, ki_uid);
1484 	CP(*ki, *ki32, ki_ruid);
1485 	CP(*ki, *ki32, ki_svuid);
1486 	CP(*ki, *ki32, ki_rgid);
1487 	CP(*ki, *ki32, ki_svgid);
1488 	CP(*ki, *ki32, ki_ngroups);
1489 	for (i = 0; i < KI_NGROUPS; i++)
1490 		CP(*ki, *ki32, ki_groups[i]);
1491 	CP(*ki, *ki32, ki_size);
1492 	CP(*ki, *ki32, ki_rssize);
1493 	CP(*ki, *ki32, ki_swrss);
1494 	CP(*ki, *ki32, ki_tsize);
1495 	CP(*ki, *ki32, ki_dsize);
1496 	CP(*ki, *ki32, ki_ssize);
1497 	CP(*ki, *ki32, ki_xstat);
1498 	CP(*ki, *ki32, ki_acflag);
1499 	CP(*ki, *ki32, ki_pctcpu);
1500 	CP(*ki, *ki32, ki_estcpu);
1501 	CP(*ki, *ki32, ki_slptime);
1502 	CP(*ki, *ki32, ki_swtime);
1503 	CP(*ki, *ki32, ki_cow);
1504 	CP(*ki, *ki32, ki_runtime);
1505 	TV_CP(*ki, *ki32, ki_start);
1506 	TV_CP(*ki, *ki32, ki_childtime);
1507 	CP(*ki, *ki32, ki_flag);
1508 	CP(*ki, *ki32, ki_kiflag);
1509 	CP(*ki, *ki32, ki_traceflag);
1510 	CP(*ki, *ki32, ki_stat);
1511 	CP(*ki, *ki32, ki_nice);
1512 	CP(*ki, *ki32, ki_lock);
1513 	CP(*ki, *ki32, ki_rqindex);
1514 	CP(*ki, *ki32, ki_oncpu);
1515 	CP(*ki, *ki32, ki_lastcpu);
1516 
1517 	/* XXX TODO: wrap cpu value as appropriate */
1518 	CP(*ki, *ki32, ki_oncpu_old);
1519 	CP(*ki, *ki32, ki_lastcpu_old);
1520 
1521 	bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1522 	bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1523 	bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1524 	bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1525 	bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1526 	bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1527 	bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1528 	bcopy(ki->ki_moretdname, ki32->ki_moretdname, MAXCOMLEN - TDNAMLEN + 1);
1529 	CP(*ki, *ki32, ki_tracer);
1530 	CP(*ki, *ki32, ki_flag2);
1531 	CP(*ki, *ki32, ki_fibnum);
1532 	CP(*ki, *ki32, ki_cr_flags);
1533 	CP(*ki, *ki32, ki_jid);
1534 	CP(*ki, *ki32, ki_numthreads);
1535 	CP(*ki, *ki32, ki_tid);
1536 	CP(*ki, *ki32, ki_pri);
1537 	freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1538 	freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1539 	PTRTRIM_CP(*ki, *ki32, ki_pcb);
1540 	PTRTRIM_CP(*ki, *ki32, ki_kstack);
1541 	PTRTRIM_CP(*ki, *ki32, ki_udata);
1542 	PTRTRIM_CP(*ki, *ki32, ki_tdaddr);
1543 	CP(*ki, *ki32, ki_sflag);
1544 	CP(*ki, *ki32, ki_tdflags);
1545 }
1546 #endif
1547 
1548 static ssize_t
1549 kern_proc_out_size(struct proc *p, int flags)
1550 {
1551 	ssize_t size = 0;
1552 
1553 	PROC_LOCK_ASSERT(p, MA_OWNED);
1554 
1555 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1556 #ifdef COMPAT_FREEBSD32
1557 		if ((flags & KERN_PROC_MASK32) != 0) {
1558 			size += sizeof(struct kinfo_proc32);
1559 		} else
1560 #endif
1561 			size += sizeof(struct kinfo_proc);
1562 	} else {
1563 #ifdef COMPAT_FREEBSD32
1564 		if ((flags & KERN_PROC_MASK32) != 0)
1565 			size += sizeof(struct kinfo_proc32) * p->p_numthreads;
1566 		else
1567 #endif
1568 			size += sizeof(struct kinfo_proc) * p->p_numthreads;
1569 	}
1570 	PROC_UNLOCK(p);
1571 	return (size);
1572 }
1573 
1574 int
1575 kern_proc_out(struct proc *p, struct sbuf *sb, int flags)
1576 {
1577 	struct thread *td;
1578 	struct kinfo_proc ki;
1579 #ifdef COMPAT_FREEBSD32
1580 	struct kinfo_proc32 ki32;
1581 #endif
1582 	int error;
1583 
1584 	PROC_LOCK_ASSERT(p, MA_OWNED);
1585 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1586 
1587 	error = 0;
1588 	fill_kinfo_proc(p, &ki);
1589 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1590 #ifdef COMPAT_FREEBSD32
1591 		if ((flags & KERN_PROC_MASK32) != 0) {
1592 			freebsd32_kinfo_proc_out(&ki, &ki32);
1593 			if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1594 				error = ENOMEM;
1595 		} else
1596 #endif
1597 			if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1598 				error = ENOMEM;
1599 	} else {
1600 		FOREACH_THREAD_IN_PROC(p, td) {
1601 			fill_kinfo_thread(td, &ki, 1);
1602 #ifdef COMPAT_FREEBSD32
1603 			if ((flags & KERN_PROC_MASK32) != 0) {
1604 				freebsd32_kinfo_proc_out(&ki, &ki32);
1605 				if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1606 					error = ENOMEM;
1607 			} else
1608 #endif
1609 				if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1610 					error = ENOMEM;
1611 			if (error != 0)
1612 				break;
1613 		}
1614 	}
1615 	PROC_UNLOCK(p);
1616 	return (error);
1617 }
1618 
1619 static int
1620 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
1621 {
1622 	struct sbuf sb;
1623 	struct kinfo_proc ki;
1624 	int error, error2;
1625 
1626 	if (req->oldptr == NULL)
1627 		return (SYSCTL_OUT(req, 0, kern_proc_out_size(p, flags)));
1628 
1629 	sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req);
1630 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1631 	error = kern_proc_out(p, &sb, flags);
1632 	error2 = sbuf_finish(&sb);
1633 	sbuf_delete(&sb);
1634 	if (error != 0)
1635 		return (error);
1636 	else if (error2 != 0)
1637 		return (error2);
1638 	return (0);
1639 }
1640 
1641 int
1642 proc_iterate(int (*cb)(struct proc *, void *), void *cbarg)
1643 {
1644 	struct proc *p;
1645 	int error, i, j;
1646 
1647 	for (i = 0; i < pidhashlock + 1; i++) {
1648 		sx_slock(&pidhashtbl_lock[i]);
1649 		for (j = i; j <= pidhash; j += pidhashlock + 1) {
1650 			LIST_FOREACH(p, &pidhashtbl[j], p_hash) {
1651 				if (p->p_state == PRS_NEW)
1652 					continue;
1653 				error = cb(p, cbarg);
1654 				PROC_LOCK_ASSERT(p, MA_NOTOWNED);
1655 				if (error != 0) {
1656 					sx_sunlock(&pidhashtbl_lock[i]);
1657 					return (error);
1658 				}
1659 			}
1660 		}
1661 		sx_sunlock(&pidhashtbl_lock[i]);
1662 	}
1663 	return (0);
1664 }
1665 
1666 struct kern_proc_out_args {
1667 	struct sysctl_req *req;
1668 	int flags;
1669 	int oid_number;
1670 	int *name;
1671 };
1672 
1673 static int
1674 sysctl_kern_proc_iterate(struct proc *p, void *origarg)
1675 {
1676 	struct kern_proc_out_args *arg = origarg;
1677 	int *name = arg->name;
1678 	int oid_number = arg->oid_number;
1679 	int flags = arg->flags;
1680 	struct sysctl_req *req = arg->req;
1681 	int error = 0;
1682 
1683 	PROC_LOCK(p);
1684 
1685 	KASSERT(p->p_ucred != NULL,
1686 	    ("process credential is NULL for non-NEW proc"));
1687 	/*
1688 	 * Show a user only appropriate processes.
1689 	 */
1690 	if (p_cansee(curthread, p))
1691 		goto skip;
1692 	/*
1693 	 * TODO - make more efficient (see notes below).
1694 	 * do by session.
1695 	 */
1696 	switch (oid_number) {
1697 	case KERN_PROC_GID:
1698 		if (p->p_ucred->cr_gid != (gid_t)name[0])
1699 			goto skip;
1700 		break;
1701 
1702 	case KERN_PROC_PGRP:
1703 		/* could do this by traversing pgrp */
1704 		if (p->p_pgrp == NULL ||
1705 		    p->p_pgrp->pg_id != (pid_t)name[0])
1706 			goto skip;
1707 		break;
1708 
1709 	case KERN_PROC_RGID:
1710 		if (p->p_ucred->cr_rgid != (gid_t)name[0])
1711 			goto skip;
1712 		break;
1713 
1714 	case KERN_PROC_SESSION:
1715 		if (p->p_session == NULL ||
1716 		    p->p_session->s_sid != (pid_t)name[0])
1717 			goto skip;
1718 		break;
1719 
1720 	case KERN_PROC_TTY:
1721 		if ((p->p_flag & P_CONTROLT) == 0 ||
1722 		    p->p_session == NULL)
1723 			goto skip;
1724 		/* XXX proctree_lock */
1725 		SESS_LOCK(p->p_session);
1726 		if (p->p_session->s_ttyp == NULL ||
1727 		    tty_udev(p->p_session->s_ttyp) !=
1728 		    (dev_t)name[0]) {
1729 			SESS_UNLOCK(p->p_session);
1730 			goto skip;
1731 		}
1732 		SESS_UNLOCK(p->p_session);
1733 		break;
1734 
1735 	case KERN_PROC_UID:
1736 		if (p->p_ucred->cr_uid != (uid_t)name[0])
1737 			goto skip;
1738 		break;
1739 
1740 	case KERN_PROC_RUID:
1741 		if (p->p_ucred->cr_ruid != (uid_t)name[0])
1742 			goto skip;
1743 		break;
1744 
1745 	case KERN_PROC_PROC:
1746 		break;
1747 
1748 	default:
1749 		break;
1750 	}
1751 	error = sysctl_out_proc(p, req, flags);
1752 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
1753 	return (error);
1754 skip:
1755 	PROC_UNLOCK(p);
1756 	return (0);
1757 }
1758 
1759 static int
1760 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1761 {
1762 	struct kern_proc_out_args iterarg;
1763 	int *name = (int *)arg1;
1764 	u_int namelen = arg2;
1765 	struct proc *p;
1766 	int flags, oid_number;
1767 	int error = 0;
1768 
1769 	oid_number = oidp->oid_number;
1770 	if (oid_number != KERN_PROC_ALL &&
1771 	    (oid_number & KERN_PROC_INC_THREAD) == 0)
1772 		flags = KERN_PROC_NOTHREADS;
1773 	else {
1774 		flags = 0;
1775 		oid_number &= ~KERN_PROC_INC_THREAD;
1776 	}
1777 #ifdef COMPAT_FREEBSD32
1778 	if (req->flags & SCTL_MASK32)
1779 		flags |= KERN_PROC_MASK32;
1780 #endif
1781 	if (oid_number == KERN_PROC_PID) {
1782 		if (namelen != 1)
1783 			return (EINVAL);
1784 		error = sysctl_wire_old_buffer(req, 0);
1785 		if (error)
1786 			return (error);
1787 		error = pget((pid_t)name[0], PGET_CANSEE, &p);
1788 		if (error == 0)
1789 			error = sysctl_out_proc(p, req, flags);
1790 		return (error);
1791 	}
1792 
1793 	switch (oid_number) {
1794 	case KERN_PROC_ALL:
1795 		if (namelen != 0)
1796 			return (EINVAL);
1797 		break;
1798 	case KERN_PROC_PROC:
1799 		if (namelen != 0 && namelen != 1)
1800 			return (EINVAL);
1801 		break;
1802 	default:
1803 		if (namelen != 1)
1804 			return (EINVAL);
1805 		break;
1806 	}
1807 
1808 	if (req->oldptr == NULL) {
1809 		/* overestimate by 5 procs */
1810 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1811 		if (error)
1812 			return (error);
1813 	} else {
1814 		error = sysctl_wire_old_buffer(req, 0);
1815 		if (error != 0)
1816 			return (error);
1817 	}
1818 	iterarg.flags = flags;
1819 	iterarg.oid_number = oid_number;
1820 	iterarg.req = req;
1821 	iterarg.name = name;
1822 	error = proc_iterate(sysctl_kern_proc_iterate, &iterarg);
1823 	return (error);
1824 }
1825 
1826 struct pargs *
1827 pargs_alloc(int len)
1828 {
1829 	struct pargs *pa;
1830 
1831 	pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1832 		M_WAITOK);
1833 	refcount_init(&pa->ar_ref, 1);
1834 	pa->ar_length = len;
1835 	return (pa);
1836 }
1837 
1838 static void
1839 pargs_free(struct pargs *pa)
1840 {
1841 
1842 	free(pa, M_PARGS);
1843 }
1844 
1845 void
1846 pargs_hold(struct pargs *pa)
1847 {
1848 
1849 	if (pa == NULL)
1850 		return;
1851 	refcount_acquire(&pa->ar_ref);
1852 }
1853 
1854 void
1855 pargs_drop(struct pargs *pa)
1856 {
1857 
1858 	if (pa == NULL)
1859 		return;
1860 	if (refcount_release(&pa->ar_ref))
1861 		pargs_free(pa);
1862 }
1863 
1864 static int
1865 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1866     size_t len)
1867 {
1868 	ssize_t n;
1869 
1870 	/*
1871 	 * This may return a short read if the string is shorter than the chunk
1872 	 * and is aligned at the end of the page, and the following page is not
1873 	 * mapped.
1874 	 */
1875 	n = proc_readmem(td, p, (vm_offset_t)sptr, buf, len);
1876 	if (n <= 0)
1877 		return (ENOMEM);
1878 	return (0);
1879 }
1880 
1881 #define PROC_AUXV_MAX	256	/* Safety limit on auxv size. */
1882 
1883 enum proc_vector_type {
1884 	PROC_ARG,
1885 	PROC_ENV,
1886 	PROC_AUX,
1887 };
1888 
1889 #ifdef COMPAT_FREEBSD32
1890 static int
1891 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1892     size_t *vsizep, enum proc_vector_type type)
1893 {
1894 	struct freebsd32_ps_strings pss;
1895 	Elf32_Auxinfo aux;
1896 	vm_offset_t vptr, ptr;
1897 	uint32_t *proc_vector32;
1898 	char **proc_vector;
1899 	size_t vsize, size;
1900 	int i, error;
1901 
1902 	error = 0;
1903 	if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss,
1904 	    sizeof(pss)) != sizeof(pss))
1905 		return (ENOMEM);
1906 	switch (type) {
1907 	case PROC_ARG:
1908 		vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1909 		vsize = pss.ps_nargvstr;
1910 		if (vsize > ARG_MAX)
1911 			return (ENOEXEC);
1912 		size = vsize * sizeof(int32_t);
1913 		break;
1914 	case PROC_ENV:
1915 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1916 		vsize = pss.ps_nenvstr;
1917 		if (vsize > ARG_MAX)
1918 			return (ENOEXEC);
1919 		size = vsize * sizeof(int32_t);
1920 		break;
1921 	case PROC_AUX:
1922 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1923 		    (pss.ps_nenvstr + 1) * sizeof(int32_t);
1924 		if (vptr % 4 != 0)
1925 			return (ENOEXEC);
1926 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1927 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1928 			    sizeof(aux))
1929 				return (ENOMEM);
1930 			if (aux.a_type == AT_NULL)
1931 				break;
1932 			ptr += sizeof(aux);
1933 		}
1934 		if (aux.a_type != AT_NULL)
1935 			return (ENOEXEC);
1936 		vsize = i + 1;
1937 		size = vsize * sizeof(aux);
1938 		break;
1939 	default:
1940 		KASSERT(0, ("Wrong proc vector type: %d", type));
1941 		return (EINVAL);
1942 	}
1943 	proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1944 	if (proc_readmem(td, p, vptr, proc_vector32, size) != size) {
1945 		error = ENOMEM;
1946 		goto done;
1947 	}
1948 	if (type == PROC_AUX) {
1949 		*proc_vectorp = (char **)proc_vector32;
1950 		*vsizep = vsize;
1951 		return (0);
1952 	}
1953 	proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1954 	for (i = 0; i < (int)vsize; i++)
1955 		proc_vector[i] = PTRIN(proc_vector32[i]);
1956 	*proc_vectorp = proc_vector;
1957 	*vsizep = vsize;
1958 done:
1959 	free(proc_vector32, M_TEMP);
1960 	return (error);
1961 }
1962 #endif
1963 
1964 static int
1965 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1966     size_t *vsizep, enum proc_vector_type type)
1967 {
1968 	struct ps_strings pss;
1969 	Elf_Auxinfo aux;
1970 	vm_offset_t vptr, ptr;
1971 	char **proc_vector;
1972 	size_t vsize, size;
1973 	int i;
1974 
1975 #ifdef COMPAT_FREEBSD32
1976 	if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1977 		return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1978 #endif
1979 	if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss,
1980 	    sizeof(pss)) != sizeof(pss))
1981 		return (ENOMEM);
1982 	switch (type) {
1983 	case PROC_ARG:
1984 		vptr = (vm_offset_t)pss.ps_argvstr;
1985 		vsize = pss.ps_nargvstr;
1986 		if (vsize > ARG_MAX)
1987 			return (ENOEXEC);
1988 		size = vsize * sizeof(char *);
1989 		break;
1990 	case PROC_ENV:
1991 		vptr = (vm_offset_t)pss.ps_envstr;
1992 		vsize = pss.ps_nenvstr;
1993 		if (vsize > ARG_MAX)
1994 			return (ENOEXEC);
1995 		size = vsize * sizeof(char *);
1996 		break;
1997 	case PROC_AUX:
1998 		/*
1999 		 * The aux array is just above env array on the stack. Check
2000 		 * that the address is naturally aligned.
2001 		 */
2002 		vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
2003 		    * sizeof(char *);
2004 #if __ELF_WORD_SIZE == 64
2005 		if (vptr % sizeof(uint64_t) != 0)
2006 #else
2007 		if (vptr % sizeof(uint32_t) != 0)
2008 #endif
2009 			return (ENOEXEC);
2010 		/*
2011 		 * We count the array size reading the aux vectors from the
2012 		 * stack until AT_NULL vector is returned.  So (to keep the code
2013 		 * simple) we read the process stack twice: the first time here
2014 		 * to find the size and the second time when copying the vectors
2015 		 * to the allocated proc_vector.
2016 		 */
2017 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
2018 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
2019 			    sizeof(aux))
2020 				return (ENOMEM);
2021 			if (aux.a_type == AT_NULL)
2022 				break;
2023 			ptr += sizeof(aux);
2024 		}
2025 		/*
2026 		 * If the PROC_AUXV_MAX entries are iterated over, and we have
2027 		 * not reached AT_NULL, it is most likely we are reading wrong
2028 		 * data: either the process doesn't have auxv array or data has
2029 		 * been modified. Return the error in this case.
2030 		 */
2031 		if (aux.a_type != AT_NULL)
2032 			return (ENOEXEC);
2033 		vsize = i + 1;
2034 		size = vsize * sizeof(aux);
2035 		break;
2036 	default:
2037 		KASSERT(0, ("Wrong proc vector type: %d", type));
2038 		return (EINVAL); /* In case we are built without INVARIANTS. */
2039 	}
2040 	proc_vector = malloc(size, M_TEMP, M_WAITOK);
2041 	if (proc_readmem(td, p, vptr, proc_vector, size) != size) {
2042 		free(proc_vector, M_TEMP);
2043 		return (ENOMEM);
2044 	}
2045 	*proc_vectorp = proc_vector;
2046 	*vsizep = vsize;
2047 
2048 	return (0);
2049 }
2050 
2051 #define GET_PS_STRINGS_CHUNK_SZ	256	/* Chunk size (bytes) for ps_strings operations. */
2052 
2053 static int
2054 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
2055     enum proc_vector_type type)
2056 {
2057 	size_t done, len, nchr, vsize;
2058 	int error, i;
2059 	char **proc_vector, *sptr;
2060 	char pss_string[GET_PS_STRINGS_CHUNK_SZ];
2061 
2062 	PROC_ASSERT_HELD(p);
2063 
2064 	/*
2065 	 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
2066 	 */
2067 	nchr = 2 * (PATH_MAX + ARG_MAX);
2068 
2069 	error = get_proc_vector(td, p, &proc_vector, &vsize, type);
2070 	if (error != 0)
2071 		return (error);
2072 	for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
2073 		/*
2074 		 * The program may have scribbled into its argv array, e.g. to
2075 		 * remove some arguments.  If that has happened, break out
2076 		 * before trying to read from NULL.
2077 		 */
2078 		if (proc_vector[i] == NULL)
2079 			break;
2080 		for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
2081 			error = proc_read_string(td, p, sptr, pss_string,
2082 			    sizeof(pss_string));
2083 			if (error != 0)
2084 				goto done;
2085 			len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
2086 			if (done + len >= nchr)
2087 				len = nchr - done - 1;
2088 			sbuf_bcat(sb, pss_string, len);
2089 			if (len != GET_PS_STRINGS_CHUNK_SZ)
2090 				break;
2091 			done += GET_PS_STRINGS_CHUNK_SZ;
2092 		}
2093 		sbuf_bcat(sb, "", 1);
2094 		done += len + 1;
2095 	}
2096 done:
2097 	free(proc_vector, M_TEMP);
2098 	return (error);
2099 }
2100 
2101 int
2102 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
2103 {
2104 
2105 	return (get_ps_strings(curthread, p, sb, PROC_ARG));
2106 }
2107 
2108 int
2109 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
2110 {
2111 
2112 	return (get_ps_strings(curthread, p, sb, PROC_ENV));
2113 }
2114 
2115 int
2116 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb)
2117 {
2118 	size_t vsize, size;
2119 	char **auxv;
2120 	int error;
2121 
2122 	error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX);
2123 	if (error == 0) {
2124 #ifdef COMPAT_FREEBSD32
2125 		if (SV_PROC_FLAG(p, SV_ILP32) != 0)
2126 			size = vsize * sizeof(Elf32_Auxinfo);
2127 		else
2128 #endif
2129 			size = vsize * sizeof(Elf_Auxinfo);
2130 		if (sbuf_bcat(sb, auxv, size) != 0)
2131 			error = ENOMEM;
2132 		free(auxv, M_TEMP);
2133 	}
2134 	return (error);
2135 }
2136 
2137 /*
2138  * This sysctl allows a process to retrieve the argument list or process
2139  * title for another process without groping around in the address space
2140  * of the other process.  It also allow a process to set its own "process
2141  * title to a string of its own choice.
2142  */
2143 static int
2144 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
2145 {
2146 	int *name = (int *)arg1;
2147 	u_int namelen = arg2;
2148 	struct pargs *newpa, *pa;
2149 	struct proc *p;
2150 	struct sbuf sb;
2151 	int flags, error = 0, error2;
2152 	pid_t pid;
2153 
2154 	if (namelen != 1)
2155 		return (EINVAL);
2156 
2157 	pid = (pid_t)name[0];
2158 	/*
2159 	 * If the query is for this process and it is single-threaded, there
2160 	 * is nobody to modify pargs, thus we can just read.
2161 	 */
2162 	p = curproc;
2163 	if (pid == p->p_pid && p->p_numthreads == 1 && req->newptr == NULL &&
2164 	    (pa = p->p_args) != NULL)
2165 		return (SYSCTL_OUT(req, pa->ar_args, pa->ar_length));
2166 
2167 	flags = PGET_CANSEE;
2168 	if (req->newptr != NULL)
2169 		flags |= PGET_ISCURRENT;
2170 	error = pget(pid, flags, &p);
2171 	if (error)
2172 		return (error);
2173 
2174 	pa = p->p_args;
2175 	if (pa != NULL) {
2176 		pargs_hold(pa);
2177 		PROC_UNLOCK(p);
2178 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
2179 		pargs_drop(pa);
2180 	} else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
2181 		_PHOLD(p);
2182 		PROC_UNLOCK(p);
2183 		sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2184 		sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2185 		error = proc_getargv(curthread, p, &sb);
2186 		error2 = sbuf_finish(&sb);
2187 		PRELE(p);
2188 		sbuf_delete(&sb);
2189 		if (error == 0 && error2 != 0)
2190 			error = error2;
2191 	} else {
2192 		PROC_UNLOCK(p);
2193 	}
2194 	if (error != 0 || req->newptr == NULL)
2195 		return (error);
2196 
2197 	if (req->newlen > ps_arg_cache_limit - sizeof(struct pargs))
2198 		return (ENOMEM);
2199 
2200 	if (req->newlen == 0) {
2201 		/*
2202 		 * Clear the argument pointer, so that we'll fetch arguments
2203 		 * with proc_getargv() until further notice.
2204 		 */
2205 		newpa = NULL;
2206 	} else {
2207 		newpa = pargs_alloc(req->newlen);
2208 		error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
2209 		if (error != 0) {
2210 			pargs_free(newpa);
2211 			return (error);
2212 		}
2213 	}
2214 	PROC_LOCK(p);
2215 	pa = p->p_args;
2216 	p->p_args = newpa;
2217 	PROC_UNLOCK(p);
2218 	pargs_drop(pa);
2219 	return (0);
2220 }
2221 
2222 /*
2223  * This sysctl allows a process to retrieve environment of another process.
2224  */
2225 static int
2226 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
2227 {
2228 	int *name = (int *)arg1;
2229 	u_int namelen = arg2;
2230 	struct proc *p;
2231 	struct sbuf sb;
2232 	int error, error2;
2233 
2234 	if (namelen != 1)
2235 		return (EINVAL);
2236 
2237 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2238 	if (error != 0)
2239 		return (error);
2240 	if ((p->p_flag & P_SYSTEM) != 0) {
2241 		PRELE(p);
2242 		return (0);
2243 	}
2244 
2245 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2246 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2247 	error = proc_getenvv(curthread, p, &sb);
2248 	error2 = sbuf_finish(&sb);
2249 	PRELE(p);
2250 	sbuf_delete(&sb);
2251 	return (error != 0 ? error : error2);
2252 }
2253 
2254 /*
2255  * This sysctl allows a process to retrieve ELF auxiliary vector of
2256  * another process.
2257  */
2258 static int
2259 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
2260 {
2261 	int *name = (int *)arg1;
2262 	u_int namelen = arg2;
2263 	struct proc *p;
2264 	struct sbuf sb;
2265 	int error, error2;
2266 
2267 	if (namelen != 1)
2268 		return (EINVAL);
2269 
2270 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2271 	if (error != 0)
2272 		return (error);
2273 	if ((p->p_flag & P_SYSTEM) != 0) {
2274 		PRELE(p);
2275 		return (0);
2276 	}
2277 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2278 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2279 	error = proc_getauxv(curthread, p, &sb);
2280 	error2 = sbuf_finish(&sb);
2281 	PRELE(p);
2282 	sbuf_delete(&sb);
2283 	return (error != 0 ? error : error2);
2284 }
2285 
2286 /*
2287  * This sysctl allows a process to retrieve the path of the executable for
2288  * itself or another process.
2289  */
2290 static int
2291 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
2292 {
2293 	pid_t *pidp = (pid_t *)arg1;
2294 	unsigned int arglen = arg2;
2295 	struct proc *p;
2296 	struct vnode *vp;
2297 	char *retbuf, *freebuf;
2298 	int error;
2299 
2300 	if (arglen != 1)
2301 		return (EINVAL);
2302 	if (*pidp == -1) {	/* -1 means this process */
2303 		p = req->td->td_proc;
2304 	} else {
2305 		error = pget(*pidp, PGET_CANSEE, &p);
2306 		if (error != 0)
2307 			return (error);
2308 	}
2309 
2310 	vp = p->p_textvp;
2311 	if (vp == NULL) {
2312 		if (*pidp != -1)
2313 			PROC_UNLOCK(p);
2314 		return (0);
2315 	}
2316 	vref(vp);
2317 	if (*pidp != -1)
2318 		PROC_UNLOCK(p);
2319 	error = vn_fullpath(vp, &retbuf, &freebuf);
2320 	vrele(vp);
2321 	if (error)
2322 		return (error);
2323 	error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
2324 	free(freebuf, M_TEMP);
2325 	return (error);
2326 }
2327 
2328 static int
2329 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
2330 {
2331 	struct proc *p;
2332 	char *sv_name;
2333 	int *name;
2334 	int namelen;
2335 	int error;
2336 
2337 	namelen = arg2;
2338 	if (namelen != 1)
2339 		return (EINVAL);
2340 
2341 	name = (int *)arg1;
2342 	error = pget((pid_t)name[0], PGET_CANSEE, &p);
2343 	if (error != 0)
2344 		return (error);
2345 	sv_name = p->p_sysent->sv_name;
2346 	PROC_UNLOCK(p);
2347 	return (sysctl_handle_string(oidp, sv_name, 0, req));
2348 }
2349 
2350 #ifdef KINFO_OVMENTRY_SIZE
2351 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
2352 #endif
2353 
2354 #ifdef COMPAT_FREEBSD7
2355 static int
2356 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
2357 {
2358 	vm_map_entry_t entry, tmp_entry;
2359 	unsigned int last_timestamp;
2360 	char *fullpath, *freepath;
2361 	struct kinfo_ovmentry *kve;
2362 	struct vattr va;
2363 	struct ucred *cred;
2364 	int error, *name;
2365 	struct vnode *vp;
2366 	struct proc *p;
2367 	vm_map_t map;
2368 	struct vmspace *vm;
2369 
2370 	name = (int *)arg1;
2371 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2372 	if (error != 0)
2373 		return (error);
2374 	vm = vmspace_acquire_ref(p);
2375 	if (vm == NULL) {
2376 		PRELE(p);
2377 		return (ESRCH);
2378 	}
2379 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2380 
2381 	map = &vm->vm_map;
2382 	vm_map_lock_read(map);
2383 	VM_MAP_ENTRY_FOREACH(entry, map) {
2384 		vm_object_t obj, tobj, lobj;
2385 		vm_offset_t addr;
2386 
2387 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2388 			continue;
2389 
2390 		bzero(kve, sizeof(*kve));
2391 		kve->kve_structsize = sizeof(*kve);
2392 
2393 		kve->kve_private_resident = 0;
2394 		obj = entry->object.vm_object;
2395 		if (obj != NULL) {
2396 			VM_OBJECT_RLOCK(obj);
2397 			if (obj->shadow_count == 1)
2398 				kve->kve_private_resident =
2399 				    obj->resident_page_count;
2400 		}
2401 		kve->kve_resident = 0;
2402 		addr = entry->start;
2403 		while (addr < entry->end) {
2404 			if (pmap_extract(map->pmap, addr))
2405 				kve->kve_resident++;
2406 			addr += PAGE_SIZE;
2407 		}
2408 
2409 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2410 			if (tobj != obj) {
2411 				VM_OBJECT_RLOCK(tobj);
2412 				kve->kve_offset += tobj->backing_object_offset;
2413 			}
2414 			if (lobj != obj)
2415 				VM_OBJECT_RUNLOCK(lobj);
2416 			lobj = tobj;
2417 		}
2418 
2419 		kve->kve_start = (void*)entry->start;
2420 		kve->kve_end = (void*)entry->end;
2421 		kve->kve_offset += (off_t)entry->offset;
2422 
2423 		if (entry->protection & VM_PROT_READ)
2424 			kve->kve_protection |= KVME_PROT_READ;
2425 		if (entry->protection & VM_PROT_WRITE)
2426 			kve->kve_protection |= KVME_PROT_WRITE;
2427 		if (entry->protection & VM_PROT_EXECUTE)
2428 			kve->kve_protection |= KVME_PROT_EXEC;
2429 
2430 		if (entry->eflags & MAP_ENTRY_COW)
2431 			kve->kve_flags |= KVME_FLAG_COW;
2432 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2433 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2434 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2435 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2436 
2437 		last_timestamp = map->timestamp;
2438 		vm_map_unlock_read(map);
2439 
2440 		kve->kve_fileid = 0;
2441 		kve->kve_fsid = 0;
2442 		freepath = NULL;
2443 		fullpath = "";
2444 		if (lobj) {
2445 			kve->kve_type = vm_object_kvme_type(lobj, &vp);
2446 			if (kve->kve_type == KVME_TYPE_MGTDEVICE)
2447 				kve->kve_type = KVME_TYPE_UNKNOWN;
2448 			if (vp != NULL)
2449 				vref(vp);
2450 			if (lobj != obj)
2451 				VM_OBJECT_RUNLOCK(lobj);
2452 
2453 			kve->kve_ref_count = obj->ref_count;
2454 			kve->kve_shadow_count = obj->shadow_count;
2455 			VM_OBJECT_RUNLOCK(obj);
2456 			if (vp != NULL) {
2457 				vn_fullpath(vp, &fullpath, &freepath);
2458 				cred = curthread->td_ucred;
2459 				vn_lock(vp, LK_SHARED | LK_RETRY);
2460 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2461 					kve->kve_fileid = va.va_fileid;
2462 					/* truncate */
2463 					kve->kve_fsid = va.va_fsid;
2464 				}
2465 				vput(vp);
2466 			}
2467 		} else {
2468 			kve->kve_type = KVME_TYPE_NONE;
2469 			kve->kve_ref_count = 0;
2470 			kve->kve_shadow_count = 0;
2471 		}
2472 
2473 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2474 		if (freepath != NULL)
2475 			free(freepath, M_TEMP);
2476 
2477 		error = SYSCTL_OUT(req, kve, sizeof(*kve));
2478 		vm_map_lock_read(map);
2479 		if (error)
2480 			break;
2481 		if (last_timestamp != map->timestamp) {
2482 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2483 			entry = tmp_entry;
2484 		}
2485 	}
2486 	vm_map_unlock_read(map);
2487 	vmspace_free(vm);
2488 	PRELE(p);
2489 	free(kve, M_TEMP);
2490 	return (error);
2491 }
2492 #endif	/* COMPAT_FREEBSD7 */
2493 
2494 #ifdef KINFO_VMENTRY_SIZE
2495 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2496 #endif
2497 
2498 void
2499 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry,
2500     int *resident_count, bool *super)
2501 {
2502 	vm_object_t obj, tobj;
2503 	vm_page_t m, m_adv;
2504 	vm_offset_t addr;
2505 	vm_paddr_t pa;
2506 	vm_pindex_t pi, pi_adv, pindex;
2507 
2508 	*super = false;
2509 	*resident_count = 0;
2510 	if (vmmap_skip_res_cnt)
2511 		return;
2512 
2513 	pa = 0;
2514 	obj = entry->object.vm_object;
2515 	addr = entry->start;
2516 	m_adv = NULL;
2517 	pi = OFF_TO_IDX(entry->offset);
2518 	for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) {
2519 		if (m_adv != NULL) {
2520 			m = m_adv;
2521 		} else {
2522 			pi_adv = atop(entry->end - addr);
2523 			pindex = pi;
2524 			for (tobj = obj;; tobj = tobj->backing_object) {
2525 				m = vm_page_find_least(tobj, pindex);
2526 				if (m != NULL) {
2527 					if (m->pindex == pindex)
2528 						break;
2529 					if (pi_adv > m->pindex - pindex) {
2530 						pi_adv = m->pindex - pindex;
2531 						m_adv = m;
2532 					}
2533 				}
2534 				if (tobj->backing_object == NULL)
2535 					goto next;
2536 				pindex += OFF_TO_IDX(tobj->
2537 				    backing_object_offset);
2538 			}
2539 		}
2540 		m_adv = NULL;
2541 		if (m->psind != 0 && addr + pagesizes[1] <= entry->end &&
2542 		    (addr & (pagesizes[1] - 1)) == 0 &&
2543 		    (pmap_mincore(map->pmap, addr, &pa) & MINCORE_SUPER) != 0) {
2544 			*super = true;
2545 			pi_adv = atop(pagesizes[1]);
2546 		} else {
2547 			/*
2548 			 * We do not test the found page on validity.
2549 			 * Either the page is busy and being paged in,
2550 			 * or it was invalidated.  The first case
2551 			 * should be counted as resident, the second
2552 			 * is not so clear; we do account both.
2553 			 */
2554 			pi_adv = 1;
2555 		}
2556 		*resident_count += pi_adv;
2557 next:;
2558 	}
2559 }
2560 
2561 /*
2562  * Must be called with the process locked and will return unlocked.
2563  */
2564 int
2565 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb, ssize_t maxlen, int flags)
2566 {
2567 	vm_map_entry_t entry, tmp_entry;
2568 	struct vattr va;
2569 	vm_map_t map;
2570 	vm_object_t obj, tobj, lobj;
2571 	char *fullpath, *freepath;
2572 	struct kinfo_vmentry *kve;
2573 	struct ucred *cred;
2574 	struct vnode *vp;
2575 	struct vmspace *vm;
2576 	vm_offset_t addr;
2577 	unsigned int last_timestamp;
2578 	int error;
2579 	bool super;
2580 
2581 	PROC_LOCK_ASSERT(p, MA_OWNED);
2582 
2583 	_PHOLD(p);
2584 	PROC_UNLOCK(p);
2585 	vm = vmspace_acquire_ref(p);
2586 	if (vm == NULL) {
2587 		PRELE(p);
2588 		return (ESRCH);
2589 	}
2590 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK | M_ZERO);
2591 
2592 	error = 0;
2593 	map = &vm->vm_map;
2594 	vm_map_lock_read(map);
2595 	VM_MAP_ENTRY_FOREACH(entry, map) {
2596 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2597 			continue;
2598 
2599 		addr = entry->end;
2600 		bzero(kve, sizeof(*kve));
2601 		obj = entry->object.vm_object;
2602 		if (obj != NULL) {
2603 			for (tobj = obj; tobj != NULL;
2604 			    tobj = tobj->backing_object) {
2605 				VM_OBJECT_RLOCK(tobj);
2606 				kve->kve_offset += tobj->backing_object_offset;
2607 				lobj = tobj;
2608 			}
2609 			if (obj->backing_object == NULL)
2610 				kve->kve_private_resident =
2611 				    obj->resident_page_count;
2612 			kern_proc_vmmap_resident(map, entry,
2613 			    &kve->kve_resident, &super);
2614 			if (super)
2615 				kve->kve_flags |= KVME_FLAG_SUPER;
2616 			for (tobj = obj; tobj != NULL;
2617 			    tobj = tobj->backing_object) {
2618 				if (tobj != obj && tobj != lobj)
2619 					VM_OBJECT_RUNLOCK(tobj);
2620 			}
2621 		} else {
2622 			lobj = NULL;
2623 		}
2624 
2625 		kve->kve_start = entry->start;
2626 		kve->kve_end = entry->end;
2627 		kve->kve_offset += entry->offset;
2628 
2629 		if (entry->protection & VM_PROT_READ)
2630 			kve->kve_protection |= KVME_PROT_READ;
2631 		if (entry->protection & VM_PROT_WRITE)
2632 			kve->kve_protection |= KVME_PROT_WRITE;
2633 		if (entry->protection & VM_PROT_EXECUTE)
2634 			kve->kve_protection |= KVME_PROT_EXEC;
2635 
2636 		if (entry->eflags & MAP_ENTRY_COW)
2637 			kve->kve_flags |= KVME_FLAG_COW;
2638 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2639 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2640 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2641 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2642 		if (entry->eflags & MAP_ENTRY_GROWS_UP)
2643 			kve->kve_flags |= KVME_FLAG_GROWS_UP;
2644 		if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2645 			kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2646 		if (entry->eflags & MAP_ENTRY_USER_WIRED)
2647 			kve->kve_flags |= KVME_FLAG_USER_WIRED;
2648 
2649 		last_timestamp = map->timestamp;
2650 		vm_map_unlock_read(map);
2651 
2652 		freepath = NULL;
2653 		fullpath = "";
2654 		if (lobj != NULL) {
2655 			kve->kve_type = vm_object_kvme_type(lobj, &vp);
2656 			if (vp != NULL)
2657 				vref(vp);
2658 			if (lobj != obj)
2659 				VM_OBJECT_RUNLOCK(lobj);
2660 
2661 			kve->kve_ref_count = obj->ref_count;
2662 			kve->kve_shadow_count = obj->shadow_count;
2663 			VM_OBJECT_RUNLOCK(obj);
2664 			if (vp != NULL) {
2665 				vn_fullpath(vp, &fullpath, &freepath);
2666 				kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2667 				cred = curthread->td_ucred;
2668 				vn_lock(vp, LK_SHARED | LK_RETRY);
2669 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2670 					kve->kve_vn_fileid = va.va_fileid;
2671 					kve->kve_vn_fsid = va.va_fsid;
2672 					kve->kve_vn_fsid_freebsd11 =
2673 					    kve->kve_vn_fsid; /* truncate */
2674 					kve->kve_vn_mode =
2675 					    MAKEIMODE(va.va_type, va.va_mode);
2676 					kve->kve_vn_size = va.va_size;
2677 					kve->kve_vn_rdev = va.va_rdev;
2678 					kve->kve_vn_rdev_freebsd11 =
2679 					    kve->kve_vn_rdev; /* truncate */
2680 					kve->kve_status = KF_ATTR_VALID;
2681 				}
2682 				vput(vp);
2683 			}
2684 		} else {
2685 			kve->kve_type = KVME_TYPE_NONE;
2686 			kve->kve_ref_count = 0;
2687 			kve->kve_shadow_count = 0;
2688 		}
2689 
2690 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2691 		if (freepath != NULL)
2692 			free(freepath, M_TEMP);
2693 
2694 		/* Pack record size down */
2695 		if ((flags & KERN_VMMAP_PACK_KINFO) != 0)
2696 			kve->kve_structsize =
2697 			    offsetof(struct kinfo_vmentry, kve_path) +
2698 			    strlen(kve->kve_path) + 1;
2699 		else
2700 			kve->kve_structsize = sizeof(*kve);
2701 		kve->kve_structsize = roundup(kve->kve_structsize,
2702 		    sizeof(uint64_t));
2703 
2704 		/* Halt filling and truncate rather than exceeding maxlen */
2705 		if (maxlen != -1 && maxlen < kve->kve_structsize) {
2706 			error = 0;
2707 			vm_map_lock_read(map);
2708 			break;
2709 		} else if (maxlen != -1)
2710 			maxlen -= kve->kve_structsize;
2711 
2712 		if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0)
2713 			error = ENOMEM;
2714 		vm_map_lock_read(map);
2715 		if (error != 0)
2716 			break;
2717 		if (last_timestamp != map->timestamp) {
2718 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2719 			entry = tmp_entry;
2720 		}
2721 	}
2722 	vm_map_unlock_read(map);
2723 	vmspace_free(vm);
2724 	PRELE(p);
2725 	free(kve, M_TEMP);
2726 	return (error);
2727 }
2728 
2729 static int
2730 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2731 {
2732 	struct proc *p;
2733 	struct sbuf sb;
2734 	int error, error2, *name;
2735 
2736 	name = (int *)arg1;
2737 	sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req);
2738 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2739 	error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
2740 	if (error != 0) {
2741 		sbuf_delete(&sb);
2742 		return (error);
2743 	}
2744 	error = kern_proc_vmmap_out(p, &sb, -1, KERN_VMMAP_PACK_KINFO);
2745 	error2 = sbuf_finish(&sb);
2746 	sbuf_delete(&sb);
2747 	return (error != 0 ? error : error2);
2748 }
2749 
2750 #if defined(STACK) || defined(DDB)
2751 static int
2752 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2753 {
2754 	struct kinfo_kstack *kkstp;
2755 	int error, i, *name, numthreads;
2756 	lwpid_t *lwpidarray;
2757 	struct thread *td;
2758 	struct stack *st;
2759 	struct sbuf sb;
2760 	struct proc *p;
2761 
2762 	name = (int *)arg1;
2763 	error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2764 	if (error != 0)
2765 		return (error);
2766 
2767 	kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2768 	st = stack_create(M_WAITOK);
2769 
2770 	lwpidarray = NULL;
2771 	PROC_LOCK(p);
2772 	do {
2773 		if (lwpidarray != NULL) {
2774 			free(lwpidarray, M_TEMP);
2775 			lwpidarray = NULL;
2776 		}
2777 		numthreads = p->p_numthreads;
2778 		PROC_UNLOCK(p);
2779 		lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2780 		    M_WAITOK | M_ZERO);
2781 		PROC_LOCK(p);
2782 	} while (numthreads < p->p_numthreads);
2783 
2784 	/*
2785 	 * XXXRW: During the below loop, execve(2) and countless other sorts
2786 	 * of changes could have taken place.  Should we check to see if the
2787 	 * vmspace has been replaced, or the like, in order to prevent
2788 	 * giving a snapshot that spans, say, execve(2), with some threads
2789 	 * before and some after?  Among other things, the credentials could
2790 	 * have changed, in which case the right to extract debug info might
2791 	 * no longer be assured.
2792 	 */
2793 	i = 0;
2794 	FOREACH_THREAD_IN_PROC(p, td) {
2795 		KASSERT(i < numthreads,
2796 		    ("sysctl_kern_proc_kstack: numthreads"));
2797 		lwpidarray[i] = td->td_tid;
2798 		i++;
2799 	}
2800 	PROC_UNLOCK(p);
2801 	numthreads = i;
2802 	for (i = 0; i < numthreads; i++) {
2803 		td = tdfind(lwpidarray[i], p->p_pid);
2804 		if (td == NULL) {
2805 			continue;
2806 		}
2807 		bzero(kkstp, sizeof(*kkstp));
2808 		(void)sbuf_new(&sb, kkstp->kkst_trace,
2809 		    sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2810 		thread_lock(td);
2811 		kkstp->kkst_tid = td->td_tid;
2812 		if (TD_IS_SWAPPED(td))
2813 			kkstp->kkst_state = KKST_STATE_SWAPPED;
2814 		else if (stack_save_td(st, td) == 0)
2815 			kkstp->kkst_state = KKST_STATE_STACKOK;
2816 		else
2817 			kkstp->kkst_state = KKST_STATE_RUNNING;
2818 		thread_unlock(td);
2819 		PROC_UNLOCK(p);
2820 		stack_sbuf_print(&sb, st);
2821 		sbuf_finish(&sb);
2822 		sbuf_delete(&sb);
2823 		error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2824 		if (error)
2825 			break;
2826 	}
2827 	PRELE(p);
2828 	if (lwpidarray != NULL)
2829 		free(lwpidarray, M_TEMP);
2830 	stack_destroy(st);
2831 	free(kkstp, M_TEMP);
2832 	return (error);
2833 }
2834 #endif
2835 
2836 /*
2837  * This sysctl allows a process to retrieve the full list of groups from
2838  * itself or another process.
2839  */
2840 static int
2841 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2842 {
2843 	pid_t *pidp = (pid_t *)arg1;
2844 	unsigned int arglen = arg2;
2845 	struct proc *p;
2846 	struct ucred *cred;
2847 	int error;
2848 
2849 	if (arglen != 1)
2850 		return (EINVAL);
2851 	if (*pidp == -1) {	/* -1 means this process */
2852 		p = req->td->td_proc;
2853 		PROC_LOCK(p);
2854 	} else {
2855 		error = pget(*pidp, PGET_CANSEE, &p);
2856 		if (error != 0)
2857 			return (error);
2858 	}
2859 
2860 	cred = crhold(p->p_ucred);
2861 	PROC_UNLOCK(p);
2862 
2863 	error = SYSCTL_OUT(req, cred->cr_groups,
2864 	    cred->cr_ngroups * sizeof(gid_t));
2865 	crfree(cred);
2866 	return (error);
2867 }
2868 
2869 /*
2870  * This sysctl allows a process to retrieve or/and set the resource limit for
2871  * another process.
2872  */
2873 static int
2874 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2875 {
2876 	int *name = (int *)arg1;
2877 	u_int namelen = arg2;
2878 	struct rlimit rlim;
2879 	struct proc *p;
2880 	u_int which;
2881 	int flags, error;
2882 
2883 	if (namelen != 2)
2884 		return (EINVAL);
2885 
2886 	which = (u_int)name[1];
2887 	if (which >= RLIM_NLIMITS)
2888 		return (EINVAL);
2889 
2890 	if (req->newptr != NULL && req->newlen != sizeof(rlim))
2891 		return (EINVAL);
2892 
2893 	flags = PGET_HOLD | PGET_NOTWEXIT;
2894 	if (req->newptr != NULL)
2895 		flags |= PGET_CANDEBUG;
2896 	else
2897 		flags |= PGET_CANSEE;
2898 	error = pget((pid_t)name[0], flags, &p);
2899 	if (error != 0)
2900 		return (error);
2901 
2902 	/*
2903 	 * Retrieve limit.
2904 	 */
2905 	if (req->oldptr != NULL) {
2906 		PROC_LOCK(p);
2907 		lim_rlimit_proc(p, which, &rlim);
2908 		PROC_UNLOCK(p);
2909 	}
2910 	error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
2911 	if (error != 0)
2912 		goto errout;
2913 
2914 	/*
2915 	 * Set limit.
2916 	 */
2917 	if (req->newptr != NULL) {
2918 		error = SYSCTL_IN(req, &rlim, sizeof(rlim));
2919 		if (error == 0)
2920 			error = kern_proc_setrlimit(curthread, p, which, &rlim);
2921 	}
2922 
2923 errout:
2924 	PRELE(p);
2925 	return (error);
2926 }
2927 
2928 /*
2929  * This sysctl allows a process to retrieve ps_strings structure location of
2930  * another process.
2931  */
2932 static int
2933 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
2934 {
2935 	int *name = (int *)arg1;
2936 	u_int namelen = arg2;
2937 	struct proc *p;
2938 	vm_offset_t ps_strings;
2939 	int error;
2940 #ifdef COMPAT_FREEBSD32
2941 	uint32_t ps_strings32;
2942 #endif
2943 
2944 	if (namelen != 1)
2945 		return (EINVAL);
2946 
2947 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2948 	if (error != 0)
2949 		return (error);
2950 #ifdef COMPAT_FREEBSD32
2951 	if ((req->flags & SCTL_MASK32) != 0) {
2952 		/*
2953 		 * We return 0 if the 32 bit emulation request is for a 64 bit
2954 		 * process.
2955 		 */
2956 		ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
2957 		    PTROUT(p->p_sysent->sv_psstrings) : 0;
2958 		PROC_UNLOCK(p);
2959 		error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
2960 		return (error);
2961 	}
2962 #endif
2963 	ps_strings = p->p_sysent->sv_psstrings;
2964 	PROC_UNLOCK(p);
2965 	error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
2966 	return (error);
2967 }
2968 
2969 /*
2970  * This sysctl allows a process to retrieve umask of another process.
2971  */
2972 static int
2973 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
2974 {
2975 	int *name = (int *)arg1;
2976 	u_int namelen = arg2;
2977 	struct proc *p;
2978 	int error;
2979 	u_short cmask;
2980 	pid_t pid;
2981 
2982 	if (namelen != 1)
2983 		return (EINVAL);
2984 
2985 	pid = (pid_t)name[0];
2986 	p = curproc;
2987 	if (pid == p->p_pid || pid == 0) {
2988 		cmask = p->p_pd->pd_cmask;
2989 		goto out;
2990 	}
2991 
2992 	error = pget(pid, PGET_WANTREAD, &p);
2993 	if (error != 0)
2994 		return (error);
2995 
2996 	cmask = p->p_pd->pd_cmask;
2997 	PRELE(p);
2998 out:
2999 	error = SYSCTL_OUT(req, &cmask, sizeof(cmask));
3000 	return (error);
3001 }
3002 
3003 /*
3004  * This sysctl allows a process to set and retrieve binary osreldate of
3005  * another process.
3006  */
3007 static int
3008 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
3009 {
3010 	int *name = (int *)arg1;
3011 	u_int namelen = arg2;
3012 	struct proc *p;
3013 	int flags, error, osrel;
3014 
3015 	if (namelen != 1)
3016 		return (EINVAL);
3017 
3018 	if (req->newptr != NULL && req->newlen != sizeof(osrel))
3019 		return (EINVAL);
3020 
3021 	flags = PGET_HOLD | PGET_NOTWEXIT;
3022 	if (req->newptr != NULL)
3023 		flags |= PGET_CANDEBUG;
3024 	else
3025 		flags |= PGET_CANSEE;
3026 	error = pget((pid_t)name[0], flags, &p);
3027 	if (error != 0)
3028 		return (error);
3029 
3030 	error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
3031 	if (error != 0)
3032 		goto errout;
3033 
3034 	if (req->newptr != NULL) {
3035 		error = SYSCTL_IN(req, &osrel, sizeof(osrel));
3036 		if (error != 0)
3037 			goto errout;
3038 		if (osrel < 0) {
3039 			error = EINVAL;
3040 			goto errout;
3041 		}
3042 		p->p_osrel = osrel;
3043 	}
3044 errout:
3045 	PRELE(p);
3046 	return (error);
3047 }
3048 
3049 static int
3050 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS)
3051 {
3052 	int *name = (int *)arg1;
3053 	u_int namelen = arg2;
3054 	struct proc *p;
3055 	struct kinfo_sigtramp kst;
3056 	const struct sysentvec *sv;
3057 	int error;
3058 #ifdef COMPAT_FREEBSD32
3059 	struct kinfo_sigtramp32 kst32;
3060 #endif
3061 
3062 	if (namelen != 1)
3063 		return (EINVAL);
3064 
3065 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
3066 	if (error != 0)
3067 		return (error);
3068 	sv = p->p_sysent;
3069 #ifdef COMPAT_FREEBSD32
3070 	if ((req->flags & SCTL_MASK32) != 0) {
3071 		bzero(&kst32, sizeof(kst32));
3072 		if (SV_PROC_FLAG(p, SV_ILP32)) {
3073 			if (sv->sv_sigcode_base != 0) {
3074 				kst32.ksigtramp_start = sv->sv_sigcode_base;
3075 				kst32.ksigtramp_end = sv->sv_sigcode_base +
3076 				    *sv->sv_szsigcode;
3077 			} else {
3078 				kst32.ksigtramp_start = sv->sv_psstrings -
3079 				    *sv->sv_szsigcode;
3080 				kst32.ksigtramp_end = sv->sv_psstrings;
3081 			}
3082 		}
3083 		PROC_UNLOCK(p);
3084 		error = SYSCTL_OUT(req, &kst32, sizeof(kst32));
3085 		return (error);
3086 	}
3087 #endif
3088 	bzero(&kst, sizeof(kst));
3089 	if (sv->sv_sigcode_base != 0) {
3090 		kst.ksigtramp_start = (char *)sv->sv_sigcode_base;
3091 		kst.ksigtramp_end = (char *)sv->sv_sigcode_base +
3092 		    *sv->sv_szsigcode;
3093 	} else {
3094 		kst.ksigtramp_start = (char *)sv->sv_psstrings -
3095 		    *sv->sv_szsigcode;
3096 		kst.ksigtramp_end = (char *)sv->sv_psstrings;
3097 	}
3098 	PROC_UNLOCK(p);
3099 	error = SYSCTL_OUT(req, &kst, sizeof(kst));
3100 	return (error);
3101 }
3102 
3103 static int
3104 sysctl_kern_proc_sigfastblk(SYSCTL_HANDLER_ARGS)
3105 {
3106 	int *name = (int *)arg1;
3107 	u_int namelen = arg2;
3108 	pid_t pid;
3109 	struct proc *p;
3110 	struct thread *td1;
3111 	uintptr_t addr;
3112 #ifdef COMPAT_FREEBSD32
3113 	uint32_t addr32;
3114 #endif
3115 	int error;
3116 
3117 	if (namelen != 1 || req->newptr != NULL)
3118 		return (EINVAL);
3119 
3120 	pid = (pid_t)name[0];
3121 	error = pget(pid, PGET_HOLD | PGET_NOTWEXIT | PGET_CANDEBUG, &p);
3122 	if (error != 0)
3123 		return (error);
3124 
3125 	PROC_LOCK(p);
3126 #ifdef COMPAT_FREEBSD32
3127 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3128 		if (!SV_PROC_FLAG(p, SV_ILP32)) {
3129 			error = EINVAL;
3130 			goto errlocked;
3131 		}
3132 	}
3133 #endif
3134 	if (pid <= PID_MAX) {
3135 		td1 = FIRST_THREAD_IN_PROC(p);
3136 	} else {
3137 		FOREACH_THREAD_IN_PROC(p, td1) {
3138 			if (td1->td_tid == pid)
3139 				break;
3140 		}
3141 	}
3142 	if (td1 == NULL) {
3143 		error = ESRCH;
3144 		goto errlocked;
3145 	}
3146 	/*
3147 	 * The access to the private thread flags.  It is fine as far
3148 	 * as no out-of-thin-air values are read from td_pflags, and
3149 	 * usermode read of the td_sigblock_ptr is racy inherently,
3150 	 * since target process might have already changed it
3151 	 * meantime.
3152 	 */
3153 	if ((td1->td_pflags & TDP_SIGFASTBLOCK) != 0)
3154 		addr = (uintptr_t)td1->td_sigblock_ptr;
3155 	else
3156 		error = ENOTTY;
3157 
3158 errlocked:
3159 	_PRELE(p);
3160 	PROC_UNLOCK(p);
3161 	if (error != 0)
3162 		return (error);
3163 
3164 #ifdef COMPAT_FREEBSD32
3165 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3166 		addr32 = addr;
3167 		error = SYSCTL_OUT(req, &addr32, sizeof(addr32));
3168 	} else
3169 #endif
3170 		error = SYSCTL_OUT(req, &addr, sizeof(addr));
3171 	return (error);
3172 }
3173 
3174 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,  0,
3175     "Process table");
3176 
3177 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
3178 	CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
3179 	"Return entire process table");
3180 
3181 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3182 	sysctl_kern_proc, "Process table");
3183 
3184 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
3185 	sysctl_kern_proc, "Process table");
3186 
3187 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3188 	sysctl_kern_proc, "Process table");
3189 
3190 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
3191 	CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3192 
3193 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
3194 	sysctl_kern_proc, "Process table");
3195 
3196 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3197 	sysctl_kern_proc, "Process table");
3198 
3199 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3200 	sysctl_kern_proc, "Process table");
3201 
3202 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3203 	sysctl_kern_proc, "Process table");
3204 
3205 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
3206 	sysctl_kern_proc, "Return process table, no threads");
3207 
3208 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
3209 	CTLFLAG_RW | CTLFLAG_CAPWR | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
3210 	sysctl_kern_proc_args, "Process argument list");
3211 
3212 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
3213 	sysctl_kern_proc_env, "Process environment");
3214 
3215 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
3216 	CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
3217 
3218 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
3219 	CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
3220 
3221 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
3222 	CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
3223 	"Process syscall vector name (ABI type)");
3224 
3225 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
3226 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3227 
3228 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
3229 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3230 
3231 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
3232 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3233 
3234 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
3235 	sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3236 
3237 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
3238 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3239 
3240 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
3241 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3242 
3243 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
3244 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3245 
3246 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
3247 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3248 
3249 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
3250 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
3251 	"Return process table, no threads");
3252 
3253 #ifdef COMPAT_FREEBSD7
3254 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
3255 	CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
3256 #endif
3257 
3258 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
3259 	CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
3260 
3261 #if defined(STACK) || defined(DDB)
3262 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
3263 	CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
3264 #endif
3265 
3266 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
3267 	CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
3268 
3269 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
3270 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
3271 	"Process resource limits");
3272 
3273 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
3274 	CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
3275 	"Process ps_strings location");
3276 
3277 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
3278 	CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
3279 
3280 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
3281 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
3282 	"Process binary osreldate");
3283 
3284 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD |
3285 	CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp,
3286 	"Process signal trampoline location");
3287 
3288 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGFASTBLK, sigfastblk, CTLFLAG_RD |
3289 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_sigfastblk,
3290 	"Thread sigfastblock address");
3291 
3292 int allproc_gen;
3293 
3294 /*
3295  * stop_all_proc() purpose is to stop all process which have usermode,
3296  * except current process for obvious reasons.  This makes it somewhat
3297  * unreliable when invoked from multithreaded process.  The service
3298  * must not be user-callable anyway.
3299  */
3300 void
3301 stop_all_proc(void)
3302 {
3303 	struct proc *cp, *p;
3304 	int r, gen;
3305 	bool restart, seen_stopped, seen_exiting, stopped_some;
3306 
3307 	cp = curproc;
3308 allproc_loop:
3309 	sx_xlock(&allproc_lock);
3310 	gen = allproc_gen;
3311 	seen_exiting = seen_stopped = stopped_some = restart = false;
3312 	LIST_REMOVE(cp, p_list);
3313 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3314 	for (;;) {
3315 		p = LIST_NEXT(cp, p_list);
3316 		if (p == NULL)
3317 			break;
3318 		LIST_REMOVE(cp, p_list);
3319 		LIST_INSERT_AFTER(p, cp, p_list);
3320 		PROC_LOCK(p);
3321 		if ((p->p_flag & (P_KPROC | P_SYSTEM | P_TOTAL_STOP)) != 0) {
3322 			PROC_UNLOCK(p);
3323 			continue;
3324 		}
3325 		if ((p->p_flag & P_WEXIT) != 0) {
3326 			seen_exiting = true;
3327 			PROC_UNLOCK(p);
3328 			continue;
3329 		}
3330 		if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) {
3331 			/*
3332 			 * Stopped processes are tolerated when there
3333 			 * are no other processes which might continue
3334 			 * them.  P_STOPPED_SINGLE but not
3335 			 * P_TOTAL_STOP process still has at least one
3336 			 * thread running.
3337 			 */
3338 			seen_stopped = true;
3339 			PROC_UNLOCK(p);
3340 			continue;
3341 		}
3342 		sx_xunlock(&allproc_lock);
3343 		_PHOLD(p);
3344 		r = thread_single(p, SINGLE_ALLPROC);
3345 		if (r != 0)
3346 			restart = true;
3347 		else
3348 			stopped_some = true;
3349 		_PRELE(p);
3350 		PROC_UNLOCK(p);
3351 		sx_xlock(&allproc_lock);
3352 	}
3353 	/* Catch forked children we did not see in iteration. */
3354 	if (gen != allproc_gen)
3355 		restart = true;
3356 	sx_xunlock(&allproc_lock);
3357 	if (restart || stopped_some || seen_exiting || seen_stopped) {
3358 		kern_yield(PRI_USER);
3359 		goto allproc_loop;
3360 	}
3361 }
3362 
3363 void
3364 resume_all_proc(void)
3365 {
3366 	struct proc *cp, *p;
3367 
3368 	cp = curproc;
3369 	sx_xlock(&allproc_lock);
3370 again:
3371 	LIST_REMOVE(cp, p_list);
3372 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3373 	for (;;) {
3374 		p = LIST_NEXT(cp, p_list);
3375 		if (p == NULL)
3376 			break;
3377 		LIST_REMOVE(cp, p_list);
3378 		LIST_INSERT_AFTER(p, cp, p_list);
3379 		PROC_LOCK(p);
3380 		if ((p->p_flag & P_TOTAL_STOP) != 0) {
3381 			sx_xunlock(&allproc_lock);
3382 			_PHOLD(p);
3383 			thread_single_end(p, SINGLE_ALLPROC);
3384 			_PRELE(p);
3385 			PROC_UNLOCK(p);
3386 			sx_xlock(&allproc_lock);
3387 		} else {
3388 			PROC_UNLOCK(p);
3389 		}
3390 	}
3391 	/*  Did the loop above missed any stopped process ? */
3392 	FOREACH_PROC_IN_SYSTEM(p) {
3393 		/* No need for proc lock. */
3394 		if ((p->p_flag & P_TOTAL_STOP) != 0)
3395 			goto again;
3396 	}
3397 	sx_xunlock(&allproc_lock);
3398 }
3399 
3400 /* #define	TOTAL_STOP_DEBUG	1 */
3401 #ifdef TOTAL_STOP_DEBUG
3402 volatile static int ap_resume;
3403 #include <sys/mount.h>
3404 
3405 static int
3406 sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS)
3407 {
3408 	int error, val;
3409 
3410 	val = 0;
3411 	ap_resume = 0;
3412 	error = sysctl_handle_int(oidp, &val, 0, req);
3413 	if (error != 0 || req->newptr == NULL)
3414 		return (error);
3415 	if (val != 0) {
3416 		stop_all_proc();
3417 		syncer_suspend();
3418 		while (ap_resume == 0)
3419 			;
3420 		syncer_resume();
3421 		resume_all_proc();
3422 	}
3423 	return (0);
3424 }
3425 
3426 SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW |
3427     CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0,
3428     sysctl_debug_stop_all_proc, "I",
3429     "");
3430 #endif
3431