xref: /freebsd/sys/kern/kern_proc.c (revision bc7512cc58af2e8bbe5bbf5ca0059b1daa1da897)
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/fcntl.h>
49 #include <sys/jail.h>
50 #include <sys/kernel.h>
51 #include <sys/limits.h>
52 #include <sys/lock.h>
53 #include <sys/loginclass.h>
54 #include <sys/malloc.h>
55 #include <sys/mman.h>
56 #include <sys/mount.h>
57 #include <sys/mutex.h>
58 #include <sys/namei.h>
59 #include <sys/proc.h>
60 #include <sys/ptrace.h>
61 #include <sys/refcount.h>
62 #include <sys/resourcevar.h>
63 #include <sys/rwlock.h>
64 #include <sys/sbuf.h>
65 #include <sys/sysent.h>
66 #include <sys/sched.h>
67 #include <sys/smp.h>
68 #include <sys/stack.h>
69 #include <sys/stat.h>
70 #include <sys/dtrace_bsd.h>
71 #include <sys/sysctl.h>
72 #include <sys/filedesc.h>
73 #include <sys/tty.h>
74 #include <sys/signalvar.h>
75 #include <sys/sdt.h>
76 #include <sys/sx.h>
77 #include <sys/user.h>
78 #include <sys/vnode.h>
79 #include <sys/wait.h>
80 #ifdef KTRACE
81 #include <sys/ktrace.h>
82 #endif
83 
84 #ifdef DDB
85 #include <ddb/ddb.h>
86 #endif
87 
88 #include <vm/vm.h>
89 #include <vm/vm_param.h>
90 #include <vm/vm_extern.h>
91 #include <vm/pmap.h>
92 #include <vm/vm_map.h>
93 #include <vm/vm_object.h>
94 #include <vm/vm_page.h>
95 #include <vm/uma.h>
96 
97 #include <fs/devfs/devfs.h>
98 
99 #ifdef COMPAT_FREEBSD32
100 #include <compat/freebsd32/freebsd32.h>
101 #include <compat/freebsd32/freebsd32_util.h>
102 #endif
103 
104 SDT_PROVIDER_DEFINE(proc);
105 
106 MALLOC_DEFINE(M_SESSION, "session", "session header");
107 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
108 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
109 
110 static void doenterpgrp(struct proc *, struct pgrp *);
111 static void orphanpg(struct pgrp *pg);
112 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp);
113 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp);
114 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp,
115     int preferthread);
116 static void pgdelete(struct pgrp *);
117 static int pgrp_init(void *mem, int size, int flags);
118 static int proc_ctor(void *mem, int size, void *arg, int flags);
119 static void proc_dtor(void *mem, int size, void *arg);
120 static int proc_init(void *mem, int size, int flags);
121 static void proc_fini(void *mem, int size);
122 static void pargs_free(struct pargs *pa);
123 
124 /*
125  * Other process lists
126  */
127 struct pidhashhead *pidhashtbl = NULL;
128 struct sx *pidhashtbl_lock;
129 u_long pidhash;
130 u_long pidhashlock;
131 struct pgrphashhead *pgrphashtbl;
132 u_long pgrphash;
133 struct proclist allproc = LIST_HEAD_INITIALIZER(allproc);
134 struct sx __exclusive_cache_line allproc_lock;
135 struct sx __exclusive_cache_line proctree_lock;
136 struct mtx __exclusive_cache_line ppeers_lock;
137 struct mtx __exclusive_cache_line procid_lock;
138 uma_zone_t proc_zone;
139 uma_zone_t pgrp_zone;
140 
141 /*
142  * The offset of various fields in struct proc and struct thread.
143  * These are used by kernel debuggers to enumerate kernel threads and
144  * processes.
145  */
146 const int proc_off_p_pid = offsetof(struct proc, p_pid);
147 const int proc_off_p_comm = offsetof(struct proc, p_comm);
148 const int proc_off_p_list = offsetof(struct proc, p_list);
149 const int proc_off_p_hash = offsetof(struct proc, p_hash);
150 const int proc_off_p_threads = offsetof(struct proc, p_threads);
151 const int thread_off_td_tid = offsetof(struct thread, td_tid);
152 const int thread_off_td_name = offsetof(struct thread, td_name);
153 const int thread_off_td_oncpu = offsetof(struct thread, td_oncpu);
154 const int thread_off_td_pcb = offsetof(struct thread, td_pcb);
155 const int thread_off_td_plist = offsetof(struct thread, td_plist);
156 
157 EVENTHANDLER_LIST_DEFINE(process_ctor);
158 EVENTHANDLER_LIST_DEFINE(process_dtor);
159 EVENTHANDLER_LIST_DEFINE(process_init);
160 EVENTHANDLER_LIST_DEFINE(process_fini);
161 EVENTHANDLER_LIST_DEFINE(process_exit);
162 EVENTHANDLER_LIST_DEFINE(process_fork);
163 EVENTHANDLER_LIST_DEFINE(process_exec);
164 
165 int kstack_pages = KSTACK_PAGES;
166 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0,
167     "Kernel stack size in pages");
168 static int vmmap_skip_res_cnt = 0;
169 SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW,
170     &vmmap_skip_res_cnt, 0,
171     "Skip calculation of the pages resident count in kern.proc.vmmap");
172 
173 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
174 #ifdef COMPAT_FREEBSD32
175 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
176 #endif
177 
178 /*
179  * Initialize global process hashing structures.
180  */
181 void
182 procinit(void)
183 {
184 	u_long i;
185 
186 	sx_init(&allproc_lock, "allproc");
187 	sx_init(&proctree_lock, "proctree");
188 	mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
189 	mtx_init(&procid_lock, "procid", NULL, MTX_DEF);
190 	pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
191 	pidhashlock = (pidhash + 1) / 64;
192 	if (pidhashlock > 0)
193 		pidhashlock--;
194 	pidhashtbl_lock = malloc(sizeof(*pidhashtbl_lock) * (pidhashlock + 1),
195 	    M_PROC, M_WAITOK | M_ZERO);
196 	for (i = 0; i < pidhashlock + 1; i++)
197 		sx_init_flags(&pidhashtbl_lock[i], "pidhash", SX_DUPOK);
198 	pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
199 	proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
200 	    proc_ctor, proc_dtor, proc_init, proc_fini,
201 	    UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
202 	pgrp_zone = uma_zcreate("PGRP", sizeof(struct pgrp), NULL, NULL,
203 	    pgrp_init, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
204 	uihashinit();
205 }
206 
207 /*
208  * Prepare a proc for use.
209  */
210 static int
211 proc_ctor(void *mem, int size, void *arg, int flags)
212 {
213 	struct proc *p;
214 	struct thread *td;
215 
216 	p = (struct proc *)mem;
217 #ifdef KDTRACE_HOOKS
218 	kdtrace_proc_ctor(p);
219 #endif
220 	EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
221 	td = FIRST_THREAD_IN_PROC(p);
222 	if (td != NULL) {
223 		/* Make sure all thread constructors are executed */
224 		EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
225 	}
226 	return (0);
227 }
228 
229 /*
230  * Reclaim a proc after use.
231  */
232 static void
233 proc_dtor(void *mem, int size, void *arg)
234 {
235 	struct proc *p;
236 	struct thread *td;
237 
238 	/* INVARIANTS checks go here */
239 	p = (struct proc *)mem;
240 	td = FIRST_THREAD_IN_PROC(p);
241 	if (td != NULL) {
242 #ifdef INVARIANTS
243 		KASSERT((p->p_numthreads == 1),
244 		    ("bad number of threads in exiting process"));
245 		KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
246 #endif
247 		/* Free all OSD associated to this thread. */
248 		osd_thread_exit(td);
249 		td_softdep_cleanup(td);
250 		MPASS(td->td_su == NULL);
251 
252 		/* Make sure all thread destructors are executed */
253 		EVENTHANDLER_DIRECT_INVOKE(thread_dtor, td);
254 	}
255 	EVENTHANDLER_DIRECT_INVOKE(process_dtor, p);
256 #ifdef KDTRACE_HOOKS
257 	kdtrace_proc_dtor(p);
258 #endif
259 	if (p->p_ksi != NULL)
260 		KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
261 }
262 
263 /*
264  * Initialize type-stable parts of a proc (when newly created).
265  */
266 static int
267 proc_init(void *mem, int size, int flags)
268 {
269 	struct proc *p;
270 
271 	p = (struct proc *)mem;
272 	mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK | MTX_NEW);
273 	mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_NEW);
274 	mtx_init(&p->p_statmtx, "pstatl", NULL, MTX_SPIN | MTX_NEW);
275 	mtx_init(&p->p_itimmtx, "pitiml", NULL, MTX_SPIN | MTX_NEW);
276 	mtx_init(&p->p_profmtx, "pprofl", NULL, MTX_SPIN | MTX_NEW);
277 	cv_init(&p->p_pwait, "ppwait");
278 	TAILQ_INIT(&p->p_threads);	     /* all threads in proc */
279 	EVENTHANDLER_DIRECT_INVOKE(process_init, p);
280 	p->p_stats = pstats_alloc();
281 	p->p_pgrp = NULL;
282 	return (0);
283 }
284 
285 /*
286  * UMA should ensure that this function is never called.
287  * Freeing a proc structure would violate type stability.
288  */
289 static void
290 proc_fini(void *mem, int size)
291 {
292 #ifdef notnow
293 	struct proc *p;
294 
295 	p = (struct proc *)mem;
296 	EVENTHANDLER_DIRECT_INVOKE(process_fini, p);
297 	pstats_free(p->p_stats);
298 	thread_free(FIRST_THREAD_IN_PROC(p));
299 	mtx_destroy(&p->p_mtx);
300 	if (p->p_ksi != NULL)
301 		ksiginfo_free(p->p_ksi);
302 #else
303 	panic("proc reclaimed");
304 #endif
305 }
306 
307 static int
308 pgrp_init(void *mem, int size, int flags)
309 {
310 	struct pgrp *pg;
311 
312 	pg = mem;
313 	mtx_init(&pg->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
314 	return (0);
315 }
316 
317 /*
318  * PID space management.
319  *
320  * These bitmaps are used by fork_findpid.
321  */
322 bitstr_t bit_decl(proc_id_pidmap, PID_MAX);
323 bitstr_t bit_decl(proc_id_grpidmap, PID_MAX);
324 bitstr_t bit_decl(proc_id_sessidmap, PID_MAX);
325 bitstr_t bit_decl(proc_id_reapmap, PID_MAX);
326 
327 static bitstr_t *proc_id_array[] = {
328 	proc_id_pidmap,
329 	proc_id_grpidmap,
330 	proc_id_sessidmap,
331 	proc_id_reapmap,
332 };
333 
334 void
335 proc_id_set(int type, pid_t id)
336 {
337 
338 	KASSERT(type >= 0 && type < nitems(proc_id_array),
339 	    ("invalid type %d\n", type));
340 	mtx_lock(&procid_lock);
341 	KASSERT(bit_test(proc_id_array[type], id) == 0,
342 	    ("bit %d already set in %d\n", id, type));
343 	bit_set(proc_id_array[type], id);
344 	mtx_unlock(&procid_lock);
345 }
346 
347 void
348 proc_id_set_cond(int type, pid_t id)
349 {
350 
351 	KASSERT(type >= 0 && type < nitems(proc_id_array),
352 	    ("invalid type %d\n", type));
353 	if (bit_test(proc_id_array[type], id))
354 		return;
355 	mtx_lock(&procid_lock);
356 	bit_set(proc_id_array[type], id);
357 	mtx_unlock(&procid_lock);
358 }
359 
360 void
361 proc_id_clear(int type, pid_t id)
362 {
363 
364 	KASSERT(type >= 0 && type < nitems(proc_id_array),
365 	    ("invalid type %d\n", type));
366 	mtx_lock(&procid_lock);
367 	KASSERT(bit_test(proc_id_array[type], id) != 0,
368 	    ("bit %d not set in %d\n", id, type));
369 	bit_clear(proc_id_array[type], id);
370 	mtx_unlock(&procid_lock);
371 }
372 
373 /*
374  * Is p an inferior of the current process?
375  */
376 int
377 inferior(struct proc *p)
378 {
379 
380 	sx_assert(&proctree_lock, SX_LOCKED);
381 	PROC_LOCK_ASSERT(p, MA_OWNED);
382 	for (; p != curproc; p = proc_realparent(p)) {
383 		if (p->p_pid == 0)
384 			return (0);
385 	}
386 	return (1);
387 }
388 
389 /*
390  * Shared lock all the pid hash lists.
391  */
392 void
393 pidhash_slockall(void)
394 {
395 	u_long i;
396 
397 	for (i = 0; i < pidhashlock + 1; i++)
398 		sx_slock(&pidhashtbl_lock[i]);
399 }
400 
401 /*
402  * Shared unlock all the pid hash lists.
403  */
404 void
405 pidhash_sunlockall(void)
406 {
407 	u_long i;
408 
409 	for (i = 0; i < pidhashlock + 1; i++)
410 		sx_sunlock(&pidhashtbl_lock[i]);
411 }
412 
413 /*
414  * Similar to pfind_any(), this function finds zombies.
415  */
416 struct proc *
417 pfind_any_locked(pid_t pid)
418 {
419 	struct proc *p;
420 
421 	sx_assert(PIDHASHLOCK(pid), SX_LOCKED);
422 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
423 		if (p->p_pid == pid) {
424 			PROC_LOCK(p);
425 			if (p->p_state == PRS_NEW) {
426 				PROC_UNLOCK(p);
427 				p = NULL;
428 			}
429 			break;
430 		}
431 	}
432 	return (p);
433 }
434 
435 /*
436  * Locate a process by number.
437  *
438  * By not returning processes in the PRS_NEW state, we allow callers to avoid
439  * testing for that condition to avoid dereferencing p_ucred, et al.
440  */
441 static __always_inline struct proc *
442 _pfind(pid_t pid, bool zombie)
443 {
444 	struct proc *p;
445 
446 	p = curproc;
447 	if (p->p_pid == pid) {
448 		PROC_LOCK(p);
449 		return (p);
450 	}
451 	sx_slock(PIDHASHLOCK(pid));
452 	LIST_FOREACH(p, PIDHASH(pid), p_hash) {
453 		if (p->p_pid == pid) {
454 			PROC_LOCK(p);
455 			if (p->p_state == PRS_NEW ||
456 			    (!zombie && p->p_state == PRS_ZOMBIE)) {
457 				PROC_UNLOCK(p);
458 				p = NULL;
459 			}
460 			break;
461 		}
462 	}
463 	sx_sunlock(PIDHASHLOCK(pid));
464 	return (p);
465 }
466 
467 struct proc *
468 pfind(pid_t pid)
469 {
470 
471 	return (_pfind(pid, false));
472 }
473 
474 /*
475  * Same as pfind but allow zombies.
476  */
477 struct proc *
478 pfind_any(pid_t pid)
479 {
480 
481 	return (_pfind(pid, true));
482 }
483 
484 /*
485  * Locate a process group by number.
486  * The caller must hold proctree_lock.
487  */
488 struct pgrp *
489 pgfind(pid_t pgid)
490 {
491 	struct pgrp *pgrp;
492 
493 	sx_assert(&proctree_lock, SX_LOCKED);
494 
495 	LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
496 		if (pgrp->pg_id == pgid) {
497 			PGRP_LOCK(pgrp);
498 			return (pgrp);
499 		}
500 	}
501 	return (NULL);
502 }
503 
504 /*
505  * Locate process and do additional manipulations, depending on flags.
506  */
507 int
508 pget(pid_t pid, int flags, struct proc **pp)
509 {
510 	struct proc *p;
511 	struct thread *td1;
512 	int error;
513 
514 	p = curproc;
515 	if (p->p_pid == pid) {
516 		PROC_LOCK(p);
517 	} else {
518 		p = NULL;
519 		if (pid <= PID_MAX) {
520 			if ((flags & PGET_NOTWEXIT) == 0)
521 				p = pfind_any(pid);
522 			else
523 				p = pfind(pid);
524 		} else if ((flags & PGET_NOTID) == 0) {
525 			td1 = tdfind(pid, -1);
526 			if (td1 != NULL)
527 				p = td1->td_proc;
528 		}
529 		if (p == NULL)
530 			return (ESRCH);
531 		if ((flags & PGET_CANSEE) != 0) {
532 			error = p_cansee(curthread, p);
533 			if (error != 0)
534 				goto errout;
535 		}
536 	}
537 	if ((flags & PGET_CANDEBUG) != 0) {
538 		error = p_candebug(curthread, p);
539 		if (error != 0)
540 			goto errout;
541 	}
542 	if ((flags & PGET_ISCURRENT) != 0 && curproc != p) {
543 		error = EPERM;
544 		goto errout;
545 	}
546 	if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) {
547 		error = ESRCH;
548 		goto errout;
549 	}
550 	if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) {
551 		/*
552 		 * XXXRW: Not clear ESRCH is the right error during proc
553 		 * execve().
554 		 */
555 		error = ESRCH;
556 		goto errout;
557 	}
558 	if ((flags & PGET_HOLD) != 0) {
559 		_PHOLD(p);
560 		PROC_UNLOCK(p);
561 	}
562 	*pp = p;
563 	return (0);
564 errout:
565 	PROC_UNLOCK(p);
566 	return (error);
567 }
568 
569 /*
570  * Create a new process group.
571  * pgid must be equal to the pid of p.
572  * Begin a new session if required.
573  */
574 int
575 enterpgrp(struct proc *p, pid_t pgid, struct pgrp *pgrp, struct session *sess)
576 {
577 
578 	sx_assert(&proctree_lock, SX_XLOCKED);
579 
580 	KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
581 	KASSERT(p->p_pid == pgid,
582 	    ("enterpgrp: new pgrp and pid != pgid"));
583 	KASSERT(pgfind(pgid) == NULL,
584 	    ("enterpgrp: pgrp with pgid exists"));
585 	KASSERT(!SESS_LEADER(p),
586 	    ("enterpgrp: session leader attempted setpgrp"));
587 
588 	if (sess != NULL) {
589 		/*
590 		 * new session
591 		 */
592 		mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
593 		PROC_LOCK(p);
594 		p->p_flag &= ~P_CONTROLT;
595 		PROC_UNLOCK(p);
596 		PGRP_LOCK(pgrp);
597 		sess->s_leader = p;
598 		sess->s_sid = p->p_pid;
599 		proc_id_set(PROC_ID_SESSION, p->p_pid);
600 		refcount_init(&sess->s_count, 1);
601 		sess->s_ttyvp = NULL;
602 		sess->s_ttydp = NULL;
603 		sess->s_ttyp = NULL;
604 		bcopy(p->p_session->s_login, sess->s_login,
605 			    sizeof(sess->s_login));
606 		pgrp->pg_session = sess;
607 		KASSERT(p == curproc,
608 		    ("enterpgrp: mksession and p != curproc"));
609 	} else {
610 		pgrp->pg_session = p->p_session;
611 		sess_hold(pgrp->pg_session);
612 		PGRP_LOCK(pgrp);
613 	}
614 	pgrp->pg_id = pgid;
615 	proc_id_set(PROC_ID_GROUP, p->p_pid);
616 	LIST_INIT(&pgrp->pg_members);
617 	pgrp->pg_flags = 0;
618 
619 	/*
620 	 * As we have an exclusive lock of proctree_lock,
621 	 * this should not deadlock.
622 	 */
623 	LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
624 	SLIST_INIT(&pgrp->pg_sigiolst);
625 	PGRP_UNLOCK(pgrp);
626 
627 	doenterpgrp(p, pgrp);
628 
629 	return (0);
630 }
631 
632 /*
633  * Move p to an existing process group
634  */
635 int
636 enterthispgrp(struct proc *p, struct pgrp *pgrp)
637 {
638 
639 	sx_assert(&proctree_lock, SX_XLOCKED);
640 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
641 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
642 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
643 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
644 	KASSERT(pgrp->pg_session == p->p_session,
645 	    ("%s: pgrp's session %p, p->p_session %p proc %p\n",
646 	    __func__, pgrp->pg_session, p->p_session, p));
647 	KASSERT(pgrp != p->p_pgrp,
648 	    ("%s: p %p belongs to pgrp %p", __func__, p, pgrp));
649 
650 	doenterpgrp(p, pgrp);
651 
652 	return (0);
653 }
654 
655 /*
656  * If true, any child of q which belongs to group pgrp, qualifies the
657  * process group pgrp as not orphaned.
658  */
659 static bool
660 isjobproc(struct proc *q, struct pgrp *pgrp)
661 {
662 	sx_assert(&proctree_lock, SX_LOCKED);
663 
664 	return (q->p_pgrp != pgrp &&
665 	    q->p_pgrp->pg_session == pgrp->pg_session);
666 }
667 
668 static struct proc *
669 jobc_reaper(struct proc *p)
670 {
671 	struct proc *pp;
672 
673 	sx_assert(&proctree_lock, SA_LOCKED);
674 
675 	for (pp = p;;) {
676 		pp = pp->p_reaper;
677 		if (pp->p_reaper == pp ||
678 		    (pp->p_treeflag & P_TREE_GRPEXITED) == 0)
679 			return (pp);
680 	}
681 }
682 
683 static struct proc *
684 jobc_parent(struct proc *p, struct proc *p_exiting)
685 {
686 	struct proc *pp;
687 
688 	sx_assert(&proctree_lock, SA_LOCKED);
689 
690 	pp = proc_realparent(p);
691 	if (pp->p_pptr == NULL || pp == p_exiting ||
692 	    (pp->p_treeflag & P_TREE_GRPEXITED) == 0)
693 		return (pp);
694 	return (jobc_reaper(pp));
695 }
696 
697 static int
698 pgrp_calc_jobc(struct pgrp *pgrp)
699 {
700 	struct proc *q;
701 	int cnt;
702 
703 #ifdef INVARIANTS
704 	if (!mtx_owned(&pgrp->pg_mtx))
705 		sx_assert(&proctree_lock, SA_LOCKED);
706 #endif
707 
708 	cnt = 0;
709 	LIST_FOREACH(q, &pgrp->pg_members, p_pglist) {
710 		if ((q->p_treeflag & P_TREE_GRPEXITED) != 0 ||
711 		    q->p_pptr == NULL)
712 			continue;
713 		if (isjobproc(jobc_parent(q, NULL), pgrp))
714 			cnt++;
715 	}
716 	return (cnt);
717 }
718 
719 /*
720  * Move p to a process group
721  */
722 static void
723 doenterpgrp(struct proc *p, struct pgrp *pgrp)
724 {
725 	struct pgrp *savepgrp;
726 	struct proc *pp;
727 
728 	sx_assert(&proctree_lock, SX_XLOCKED);
729 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
730 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
731 	PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
732 	SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
733 
734 	savepgrp = p->p_pgrp;
735 	pp = jobc_parent(p, NULL);
736 
737 	PGRP_LOCK(pgrp);
738 	PGRP_LOCK(savepgrp);
739 	if (isjobproc(pp, savepgrp) && pgrp_calc_jobc(savepgrp) == 1)
740 		orphanpg(savepgrp);
741 	PROC_LOCK(p);
742 	LIST_REMOVE(p, p_pglist);
743 	p->p_pgrp = pgrp;
744 	PROC_UNLOCK(p);
745 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
746 	if (isjobproc(pp, pgrp))
747 		pgrp->pg_flags &= ~PGRP_ORPHANED;
748 	PGRP_UNLOCK(savepgrp);
749 	PGRP_UNLOCK(pgrp);
750 	if (LIST_EMPTY(&savepgrp->pg_members))
751 		pgdelete(savepgrp);
752 }
753 
754 /*
755  * remove process from process group
756  */
757 int
758 leavepgrp(struct proc *p)
759 {
760 	struct pgrp *savepgrp;
761 
762 	sx_assert(&proctree_lock, SX_XLOCKED);
763 	savepgrp = p->p_pgrp;
764 	PGRP_LOCK(savepgrp);
765 	PROC_LOCK(p);
766 	LIST_REMOVE(p, p_pglist);
767 	p->p_pgrp = NULL;
768 	PROC_UNLOCK(p);
769 	PGRP_UNLOCK(savepgrp);
770 	if (LIST_EMPTY(&savepgrp->pg_members))
771 		pgdelete(savepgrp);
772 	return (0);
773 }
774 
775 /*
776  * delete a process group
777  */
778 static void
779 pgdelete(struct pgrp *pgrp)
780 {
781 	struct session *savesess;
782 	struct tty *tp;
783 
784 	sx_assert(&proctree_lock, SX_XLOCKED);
785 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
786 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
787 
788 	/*
789 	 * Reset any sigio structures pointing to us as a result of
790 	 * F_SETOWN with our pgid.  The proctree lock ensures that
791 	 * new sigio structures will not be added after this point.
792 	 */
793 	funsetownlst(&pgrp->pg_sigiolst);
794 
795 	PGRP_LOCK(pgrp);
796 	tp = pgrp->pg_session->s_ttyp;
797 	LIST_REMOVE(pgrp, pg_hash);
798 	savesess = pgrp->pg_session;
799 	PGRP_UNLOCK(pgrp);
800 
801 	/* Remove the reference to the pgrp before deallocating it. */
802 	if (tp != NULL) {
803 		tty_lock(tp);
804 		tty_rel_pgrp(tp, pgrp);
805 	}
806 
807 	proc_id_clear(PROC_ID_GROUP, pgrp->pg_id);
808 	uma_zfree(pgrp_zone, pgrp);
809 	sess_release(savesess);
810 }
811 
812 
813 static void
814 fixjobc_kill(struct proc *p)
815 {
816 	struct proc *q;
817 	struct pgrp *pgrp;
818 
819 	sx_assert(&proctree_lock, SX_LOCKED);
820 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
821 	pgrp = p->p_pgrp;
822 	PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
823 	SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
824 
825 	/*
826 	 * p no longer affects process group orphanage for children.
827 	 * It is marked by the flag because p is only physically
828 	 * removed from its process group on wait(2).
829 	 */
830 	MPASS((p->p_treeflag & P_TREE_GRPEXITED) == 0);
831 	p->p_treeflag |= P_TREE_GRPEXITED;
832 
833 	/*
834 	 * Check if exiting p orphans its own group.
835 	 */
836 	pgrp = p->p_pgrp;
837 	if (isjobproc(jobc_parent(p, NULL), pgrp)) {
838 		PGRP_LOCK(pgrp);
839 		if (pgrp_calc_jobc(pgrp) == 0)
840 			orphanpg(pgrp);
841 		PGRP_UNLOCK(pgrp);
842 	}
843 
844 	/*
845 	 * Check this process' children to see whether they qualify
846 	 * their process groups after reparenting to reaper.
847 	 */
848 	LIST_FOREACH(q, &p->p_children, p_sibling) {
849 		pgrp = q->p_pgrp;
850 		PGRP_LOCK(pgrp);
851 		if (pgrp_calc_jobc(pgrp) == 0) {
852 			/*
853 			 * We want to handle exactly the children that
854 			 * has p as realparent.  Then, when calculating
855 			 * jobc_parent for children, we should ignore
856 			 * P_TREE_GRPEXITED flag already set on p.
857 			 */
858 			if (jobc_parent(q, p) == p && isjobproc(p, pgrp))
859 				orphanpg(pgrp);
860 		} else
861 			pgrp->pg_flags &= ~PGRP_ORPHANED;
862 		PGRP_UNLOCK(pgrp);
863 	}
864 	LIST_FOREACH(q, &p->p_orphans, p_orphan) {
865 		pgrp = q->p_pgrp;
866 		PGRP_LOCK(pgrp);
867 		if (pgrp_calc_jobc(pgrp) == 0) {
868 			if (isjobproc(p, pgrp))
869 				orphanpg(pgrp);
870 		} else
871 			pgrp->pg_flags &= ~PGRP_ORPHANED;
872 		PGRP_UNLOCK(pgrp);
873 	}
874 }
875 
876 void
877 killjobc(void)
878 {
879 	struct session *sp;
880 	struct tty *tp;
881 	struct proc *p;
882 	struct vnode *ttyvp;
883 
884 	p = curproc;
885 	MPASS(p->p_flag & P_WEXIT);
886 	sx_assert(&proctree_lock, SX_LOCKED);
887 
888 	if (SESS_LEADER(p)) {
889 		sp = p->p_session;
890 
891 		/*
892 		 * s_ttyp is not zero'd; we use this to indicate that
893 		 * the session once had a controlling terminal. (for
894 		 * logging and informational purposes)
895 		 */
896 		SESS_LOCK(sp);
897 		ttyvp = sp->s_ttyvp;
898 		tp = sp->s_ttyp;
899 		sp->s_ttyvp = NULL;
900 		sp->s_ttydp = NULL;
901 		sp->s_leader = NULL;
902 		SESS_UNLOCK(sp);
903 
904 		/*
905 		 * Signal foreground pgrp and revoke access to
906 		 * controlling terminal if it has not been revoked
907 		 * already.
908 		 *
909 		 * Because the TTY may have been revoked in the mean
910 		 * time and could already have a new session associated
911 		 * with it, make sure we don't send a SIGHUP to a
912 		 * foreground process group that does not belong to this
913 		 * session.
914 		 */
915 
916 		if (tp != NULL) {
917 			tty_lock(tp);
918 			if (tp->t_session == sp)
919 				tty_signal_pgrp(tp, SIGHUP);
920 			tty_unlock(tp);
921 		}
922 
923 		if (ttyvp != NULL) {
924 			sx_xunlock(&proctree_lock);
925 			if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) {
926 				VOP_REVOKE(ttyvp, REVOKEALL);
927 				VOP_UNLOCK(ttyvp);
928 			}
929 			devfs_ctty_unref(ttyvp);
930 			sx_xlock(&proctree_lock);
931 		}
932 	}
933 	fixjobc_kill(p);
934 }
935 
936 /*
937  * A process group has become orphaned, mark it as such for signal
938  * delivery code.  If there are any stopped processes in the group,
939  * hang-up all process in that group.
940  */
941 static void
942 orphanpg(struct pgrp *pg)
943 {
944 	struct proc *p;
945 
946 	PGRP_LOCK_ASSERT(pg, MA_OWNED);
947 
948 	pg->pg_flags |= PGRP_ORPHANED;
949 
950 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
951 		PROC_LOCK(p);
952 		if (P_SHOULDSTOP(p) == P_STOPPED_SIG) {
953 			PROC_UNLOCK(p);
954 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
955 				PROC_LOCK(p);
956 				kern_psignal(p, SIGHUP);
957 				kern_psignal(p, SIGCONT);
958 				PROC_UNLOCK(p);
959 			}
960 			return;
961 		}
962 		PROC_UNLOCK(p);
963 	}
964 }
965 
966 void
967 sess_hold(struct session *s)
968 {
969 
970 	refcount_acquire(&s->s_count);
971 }
972 
973 void
974 sess_release(struct session *s)
975 {
976 
977 	if (refcount_release(&s->s_count)) {
978 		if (s->s_ttyp != NULL) {
979 			tty_lock(s->s_ttyp);
980 			tty_rel_sess(s->s_ttyp, s);
981 		}
982 		proc_id_clear(PROC_ID_SESSION, s->s_sid);
983 		mtx_destroy(&s->s_mtx);
984 		free(s, M_SESSION);
985 	}
986 }
987 
988 #ifdef DDB
989 
990 static void
991 db_print_pgrp_one(struct pgrp *pgrp, struct proc *p)
992 {
993 	db_printf(
994 	    "    pid %d at %p pr %d pgrp %p e %d jc %d\n",
995 	    p->p_pid, p, p->p_pptr == NULL ? -1 : p->p_pptr->p_pid,
996 	    p->p_pgrp, (p->p_treeflag & P_TREE_GRPEXITED) != 0,
997 	    p->p_pptr == NULL ? 0 : isjobproc(p->p_pptr, pgrp));
998 }
999 
1000 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
1001 {
1002 	struct pgrp *pgrp;
1003 	struct proc *p;
1004 	int i;
1005 
1006 	for (i = 0; i <= pgrphash; i++) {
1007 		if (!LIST_EMPTY(&pgrphashtbl[i])) {
1008 			db_printf("indx %d\n", i);
1009 			LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
1010 				db_printf(
1011 			"  pgrp %p, pgid %d, sess %p, sesscnt %d, mem %p\n",
1012 				    pgrp, (int)pgrp->pg_id, pgrp->pg_session,
1013 				    pgrp->pg_session->s_count,
1014 				    LIST_FIRST(&pgrp->pg_members));
1015 				LIST_FOREACH(p, &pgrp->pg_members, p_pglist)
1016 					db_print_pgrp_one(pgrp, p);
1017 			}
1018 		}
1019 	}
1020 }
1021 #endif /* DDB */
1022 
1023 /*
1024  * Calculate the kinfo_proc members which contain process-wide
1025  * informations.
1026  * Must be called with the target process locked.
1027  */
1028 static void
1029 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
1030 {
1031 	struct thread *td;
1032 
1033 	PROC_LOCK_ASSERT(p, MA_OWNED);
1034 
1035 	kp->ki_estcpu = 0;
1036 	kp->ki_pctcpu = 0;
1037 	FOREACH_THREAD_IN_PROC(p, td) {
1038 		thread_lock(td);
1039 		kp->ki_pctcpu += sched_pctcpu(td);
1040 		kp->ki_estcpu += sched_estcpu(td);
1041 		thread_unlock(td);
1042 	}
1043 }
1044 
1045 /*
1046  * Fill in any information that is common to all threads in the process.
1047  * Must be called with the target process locked.
1048  */
1049 static void
1050 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
1051 {
1052 	struct thread *td0;
1053 	struct ucred *cred;
1054 	struct sigacts *ps;
1055 	struct timeval boottime;
1056 
1057 	PROC_LOCK_ASSERT(p, MA_OWNED);
1058 
1059 	kp->ki_structsize = sizeof(*kp);
1060 	kp->ki_paddr = p;
1061 	kp->ki_addr =/* p->p_addr; */0; /* XXX */
1062 	kp->ki_args = p->p_args;
1063 	kp->ki_textvp = p->p_textvp;
1064 #ifdef KTRACE
1065 	kp->ki_tracep = ktr_get_tracevp(p, false);
1066 	kp->ki_traceflag = p->p_traceflag;
1067 #endif
1068 	kp->ki_fd = p->p_fd;
1069 	kp->ki_pd = p->p_pd;
1070 	kp->ki_vmspace = p->p_vmspace;
1071 	kp->ki_flag = p->p_flag;
1072 	kp->ki_flag2 = p->p_flag2;
1073 	cred = p->p_ucred;
1074 	if (cred) {
1075 		kp->ki_uid = cred->cr_uid;
1076 		kp->ki_ruid = cred->cr_ruid;
1077 		kp->ki_svuid = cred->cr_svuid;
1078 		kp->ki_cr_flags = 0;
1079 		if (cred->cr_flags & CRED_FLAG_CAPMODE)
1080 			kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
1081 		/* XXX bde doesn't like KI_NGROUPS */
1082 		if (cred->cr_ngroups > KI_NGROUPS) {
1083 			kp->ki_ngroups = KI_NGROUPS;
1084 			kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
1085 		} else
1086 			kp->ki_ngroups = cred->cr_ngroups;
1087 		bcopy(cred->cr_groups, kp->ki_groups,
1088 		    kp->ki_ngroups * sizeof(gid_t));
1089 		kp->ki_rgid = cred->cr_rgid;
1090 		kp->ki_svgid = cred->cr_svgid;
1091 		/* If jailed(cred), emulate the old P_JAILED flag. */
1092 		if (jailed(cred)) {
1093 			kp->ki_flag |= P_JAILED;
1094 			/* If inside the jail, use 0 as a jail ID. */
1095 			if (cred->cr_prison != curthread->td_ucred->cr_prison)
1096 				kp->ki_jid = cred->cr_prison->pr_id;
1097 		}
1098 		strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
1099 		    sizeof(kp->ki_loginclass));
1100 	}
1101 	ps = p->p_sigacts;
1102 	if (ps) {
1103 		mtx_lock(&ps->ps_mtx);
1104 		kp->ki_sigignore = ps->ps_sigignore;
1105 		kp->ki_sigcatch = ps->ps_sigcatch;
1106 		mtx_unlock(&ps->ps_mtx);
1107 	}
1108 	if (p->p_state != PRS_NEW &&
1109 	    p->p_state != PRS_ZOMBIE &&
1110 	    p->p_vmspace != NULL) {
1111 		struct vmspace *vm = p->p_vmspace;
1112 
1113 		kp->ki_size = vm->vm_map.size;
1114 		kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
1115 		FOREACH_THREAD_IN_PROC(p, td0) {
1116 			if (!TD_IS_SWAPPED(td0))
1117 				kp->ki_rssize += td0->td_kstack_pages;
1118 		}
1119 		kp->ki_swrss = vm->vm_swrss;
1120 		kp->ki_tsize = vm->vm_tsize;
1121 		kp->ki_dsize = vm->vm_dsize;
1122 		kp->ki_ssize = vm->vm_ssize;
1123 	} else if (p->p_state == PRS_ZOMBIE)
1124 		kp->ki_stat = SZOMB;
1125 	if (kp->ki_flag & P_INMEM)
1126 		kp->ki_sflag = PS_INMEM;
1127 	else
1128 		kp->ki_sflag = 0;
1129 	/* Calculate legacy swtime as seconds since 'swtick'. */
1130 	kp->ki_swtime = (ticks - p->p_swtick) / hz;
1131 	kp->ki_pid = p->p_pid;
1132 	kp->ki_nice = p->p_nice;
1133 	kp->ki_fibnum = p->p_fibnum;
1134 	kp->ki_start = p->p_stats->p_start;
1135 	getboottime(&boottime);
1136 	timevaladd(&kp->ki_start, &boottime);
1137 	PROC_STATLOCK(p);
1138 	rufetch(p, &kp->ki_rusage);
1139 	kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
1140 	calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
1141 	PROC_STATUNLOCK(p);
1142 	calccru(p, &kp->ki_childutime, &kp->ki_childstime);
1143 	/* Some callers want child times in a single value. */
1144 	kp->ki_childtime = kp->ki_childstime;
1145 	timevaladd(&kp->ki_childtime, &kp->ki_childutime);
1146 
1147 	FOREACH_THREAD_IN_PROC(p, td0)
1148 		kp->ki_cow += td0->td_cow;
1149 
1150 	if (p->p_comm[0] != '\0')
1151 		strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
1152 	if (p->p_sysent && p->p_sysent->sv_name != NULL &&
1153 	    p->p_sysent->sv_name[0] != '\0')
1154 		strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
1155 	kp->ki_siglist = p->p_siglist;
1156 	kp->ki_xstat = KW_EXITCODE(p->p_xexit, p->p_xsig);
1157 	kp->ki_acflag = p->p_acflag;
1158 	kp->ki_lock = p->p_lock;
1159 	if (p->p_pptr) {
1160 		kp->ki_ppid = p->p_oppid;
1161 		if (p->p_flag & P_TRACED)
1162 			kp->ki_tracer = p->p_pptr->p_pid;
1163 	}
1164 }
1165 
1166 /*
1167  * Fill job-related process information.
1168  */
1169 static void
1170 fill_kinfo_proc_pgrp(struct proc *p, struct kinfo_proc *kp)
1171 {
1172 	struct tty *tp;
1173 	struct session *sp;
1174 	struct pgrp *pgrp;
1175 
1176 	sx_assert(&proctree_lock, SA_LOCKED);
1177 	PROC_LOCK_ASSERT(p, MA_OWNED);
1178 
1179 	pgrp = p->p_pgrp;
1180 	if (pgrp == NULL)
1181 		return;
1182 
1183 	kp->ki_pgid = pgrp->pg_id;
1184 	kp->ki_jobc = pgrp_calc_jobc(pgrp);
1185 
1186 	sp = pgrp->pg_session;
1187 	tp = NULL;
1188 
1189 	if (sp != NULL) {
1190 		kp->ki_sid = sp->s_sid;
1191 		SESS_LOCK(sp);
1192 		strlcpy(kp->ki_login, sp->s_login, sizeof(kp->ki_login));
1193 		if (sp->s_ttyvp)
1194 			kp->ki_kiflag |= KI_CTTY;
1195 		if (SESS_LEADER(p))
1196 			kp->ki_kiflag |= KI_SLEADER;
1197 		tp = sp->s_ttyp;
1198 		SESS_UNLOCK(sp);
1199 	}
1200 
1201 	if ((p->p_flag & P_CONTROLT) && tp != NULL) {
1202 		kp->ki_tdev = tty_udev(tp);
1203 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1204 		kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
1205 		if (tp->t_session)
1206 			kp->ki_tsid = tp->t_session->s_sid;
1207 	} else {
1208 		kp->ki_tdev = NODEV;
1209 		kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
1210 	}
1211 }
1212 
1213 /*
1214  * Fill in information that is thread specific.  Must be called with
1215  * target process locked.  If 'preferthread' is set, overwrite certain
1216  * process-related fields that are maintained for both threads and
1217  * processes.
1218  */
1219 static void
1220 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
1221 {
1222 	struct proc *p;
1223 
1224 	p = td->td_proc;
1225 	kp->ki_tdaddr = td;
1226 	PROC_LOCK_ASSERT(p, MA_OWNED);
1227 
1228 	if (preferthread)
1229 		PROC_STATLOCK(p);
1230 	thread_lock(td);
1231 	if (td->td_wmesg != NULL)
1232 		strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
1233 	else
1234 		bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
1235 	if (strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname)) >=
1236 	    sizeof(kp->ki_tdname)) {
1237 		strlcpy(kp->ki_moretdname,
1238 		    td->td_name + sizeof(kp->ki_tdname) - 1,
1239 		    sizeof(kp->ki_moretdname));
1240 	} else {
1241 		bzero(kp->ki_moretdname, sizeof(kp->ki_moretdname));
1242 	}
1243 	if (TD_ON_LOCK(td)) {
1244 		kp->ki_kiflag |= KI_LOCKBLOCK;
1245 		strlcpy(kp->ki_lockname, td->td_lockname,
1246 		    sizeof(kp->ki_lockname));
1247 	} else {
1248 		kp->ki_kiflag &= ~KI_LOCKBLOCK;
1249 		bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
1250 	}
1251 
1252 	if (p->p_state == PRS_NORMAL) { /* approximate. */
1253 		if (TD_ON_RUNQ(td) ||
1254 		    TD_CAN_RUN(td) ||
1255 		    TD_IS_RUNNING(td)) {
1256 			kp->ki_stat = SRUN;
1257 		} else if (P_SHOULDSTOP(p)) {
1258 			kp->ki_stat = SSTOP;
1259 		} else if (TD_IS_SLEEPING(td)) {
1260 			kp->ki_stat = SSLEEP;
1261 		} else if (TD_ON_LOCK(td)) {
1262 			kp->ki_stat = SLOCK;
1263 		} else {
1264 			kp->ki_stat = SWAIT;
1265 		}
1266 	} else if (p->p_state == PRS_ZOMBIE) {
1267 		kp->ki_stat = SZOMB;
1268 	} else {
1269 		kp->ki_stat = SIDL;
1270 	}
1271 
1272 	/* Things in the thread */
1273 	kp->ki_wchan = td->td_wchan;
1274 	kp->ki_pri.pri_level = td->td_priority;
1275 	kp->ki_pri.pri_native = td->td_base_pri;
1276 
1277 	/*
1278 	 * Note: legacy fields; clamp at the old NOCPU value and/or
1279 	 * the maximum u_char CPU value.
1280 	 */
1281 	if (td->td_lastcpu == NOCPU)
1282 		kp->ki_lastcpu_old = NOCPU_OLD;
1283 	else if (td->td_lastcpu > MAXCPU_OLD)
1284 		kp->ki_lastcpu_old = MAXCPU_OLD;
1285 	else
1286 		kp->ki_lastcpu_old = td->td_lastcpu;
1287 
1288 	if (td->td_oncpu == NOCPU)
1289 		kp->ki_oncpu_old = NOCPU_OLD;
1290 	else if (td->td_oncpu > MAXCPU_OLD)
1291 		kp->ki_oncpu_old = MAXCPU_OLD;
1292 	else
1293 		kp->ki_oncpu_old = td->td_oncpu;
1294 
1295 	kp->ki_lastcpu = td->td_lastcpu;
1296 	kp->ki_oncpu = td->td_oncpu;
1297 	kp->ki_tdflags = td->td_flags;
1298 	kp->ki_tid = td->td_tid;
1299 	kp->ki_numthreads = p->p_numthreads;
1300 	kp->ki_pcb = td->td_pcb;
1301 	kp->ki_kstack = (void *)td->td_kstack;
1302 	kp->ki_slptime = (ticks - td->td_slptick) / hz;
1303 	kp->ki_pri.pri_class = td->td_pri_class;
1304 	kp->ki_pri.pri_user = td->td_user_pri;
1305 
1306 	if (preferthread) {
1307 		rufetchtd(td, &kp->ki_rusage);
1308 		kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
1309 		kp->ki_pctcpu = sched_pctcpu(td);
1310 		kp->ki_estcpu = sched_estcpu(td);
1311 		kp->ki_cow = td->td_cow;
1312 	}
1313 
1314 	/* We can't get this anymore but ps etc never used it anyway. */
1315 	kp->ki_rqindex = 0;
1316 
1317 	if (preferthread)
1318 		kp->ki_siglist = td->td_siglist;
1319 	kp->ki_sigmask = td->td_sigmask;
1320 	thread_unlock(td);
1321 	if (preferthread)
1322 		PROC_STATUNLOCK(p);
1323 }
1324 
1325 /*
1326  * Fill in a kinfo_proc structure for the specified process.
1327  * Must be called with the target process locked.
1328  */
1329 void
1330 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1331 {
1332 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1333 
1334 	bzero(kp, sizeof(*kp));
1335 
1336 	fill_kinfo_proc_pgrp(p,kp);
1337 	fill_kinfo_proc_only(p, kp);
1338 	fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1339 	fill_kinfo_aggregate(p, kp);
1340 }
1341 
1342 struct pstats *
1343 pstats_alloc(void)
1344 {
1345 
1346 	return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1347 }
1348 
1349 /*
1350  * Copy parts of p_stats; zero the rest of p_stats (statistics).
1351  */
1352 void
1353 pstats_fork(struct pstats *src, struct pstats *dst)
1354 {
1355 
1356 	bzero(&dst->pstat_startzero,
1357 	    __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1358 	bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1359 	    __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1360 }
1361 
1362 void
1363 pstats_free(struct pstats *ps)
1364 {
1365 
1366 	free(ps, M_SUBPROC);
1367 }
1368 
1369 #ifdef COMPAT_FREEBSD32
1370 
1371 /*
1372  * This function is typically used to copy out the kernel address, so
1373  * it can be replaced by assignment of zero.
1374  */
1375 static inline uint32_t
1376 ptr32_trim(const void *ptr)
1377 {
1378 	uintptr_t uptr;
1379 
1380 	uptr = (uintptr_t)ptr;
1381 	return ((uptr > UINT_MAX) ? 0 : uptr);
1382 }
1383 
1384 #define PTRTRIM_CP(src,dst,fld) \
1385 	do { (dst).fld = ptr32_trim((src).fld); } while (0)
1386 
1387 static void
1388 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1389 {
1390 	int i;
1391 
1392 	bzero(ki32, sizeof(struct kinfo_proc32));
1393 	ki32->ki_structsize = sizeof(struct kinfo_proc32);
1394 	CP(*ki, *ki32, ki_layout);
1395 	PTRTRIM_CP(*ki, *ki32, ki_args);
1396 	PTRTRIM_CP(*ki, *ki32, ki_paddr);
1397 	PTRTRIM_CP(*ki, *ki32, ki_addr);
1398 	PTRTRIM_CP(*ki, *ki32, ki_tracep);
1399 	PTRTRIM_CP(*ki, *ki32, ki_textvp);
1400 	PTRTRIM_CP(*ki, *ki32, ki_fd);
1401 	PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1402 	PTRTRIM_CP(*ki, *ki32, ki_wchan);
1403 	CP(*ki, *ki32, ki_pid);
1404 	CP(*ki, *ki32, ki_ppid);
1405 	CP(*ki, *ki32, ki_pgid);
1406 	CP(*ki, *ki32, ki_tpgid);
1407 	CP(*ki, *ki32, ki_sid);
1408 	CP(*ki, *ki32, ki_tsid);
1409 	CP(*ki, *ki32, ki_jobc);
1410 	CP(*ki, *ki32, ki_tdev);
1411 	CP(*ki, *ki32, ki_tdev_freebsd11);
1412 	CP(*ki, *ki32, ki_siglist);
1413 	CP(*ki, *ki32, ki_sigmask);
1414 	CP(*ki, *ki32, ki_sigignore);
1415 	CP(*ki, *ki32, ki_sigcatch);
1416 	CP(*ki, *ki32, ki_uid);
1417 	CP(*ki, *ki32, ki_ruid);
1418 	CP(*ki, *ki32, ki_svuid);
1419 	CP(*ki, *ki32, ki_rgid);
1420 	CP(*ki, *ki32, ki_svgid);
1421 	CP(*ki, *ki32, ki_ngroups);
1422 	for (i = 0; i < KI_NGROUPS; i++)
1423 		CP(*ki, *ki32, ki_groups[i]);
1424 	CP(*ki, *ki32, ki_size);
1425 	CP(*ki, *ki32, ki_rssize);
1426 	CP(*ki, *ki32, ki_swrss);
1427 	CP(*ki, *ki32, ki_tsize);
1428 	CP(*ki, *ki32, ki_dsize);
1429 	CP(*ki, *ki32, ki_ssize);
1430 	CP(*ki, *ki32, ki_xstat);
1431 	CP(*ki, *ki32, ki_acflag);
1432 	CP(*ki, *ki32, ki_pctcpu);
1433 	CP(*ki, *ki32, ki_estcpu);
1434 	CP(*ki, *ki32, ki_slptime);
1435 	CP(*ki, *ki32, ki_swtime);
1436 	CP(*ki, *ki32, ki_cow);
1437 	CP(*ki, *ki32, ki_runtime);
1438 	TV_CP(*ki, *ki32, ki_start);
1439 	TV_CP(*ki, *ki32, ki_childtime);
1440 	CP(*ki, *ki32, ki_flag);
1441 	CP(*ki, *ki32, ki_kiflag);
1442 	CP(*ki, *ki32, ki_traceflag);
1443 	CP(*ki, *ki32, ki_stat);
1444 	CP(*ki, *ki32, ki_nice);
1445 	CP(*ki, *ki32, ki_lock);
1446 	CP(*ki, *ki32, ki_rqindex);
1447 	CP(*ki, *ki32, ki_oncpu);
1448 	CP(*ki, *ki32, ki_lastcpu);
1449 
1450 	/* XXX TODO: wrap cpu value as appropriate */
1451 	CP(*ki, *ki32, ki_oncpu_old);
1452 	CP(*ki, *ki32, ki_lastcpu_old);
1453 
1454 	bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1455 	bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1456 	bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1457 	bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1458 	bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1459 	bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1460 	bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1461 	bcopy(ki->ki_moretdname, ki32->ki_moretdname, MAXCOMLEN - TDNAMLEN + 1);
1462 	CP(*ki, *ki32, ki_tracer);
1463 	CP(*ki, *ki32, ki_flag2);
1464 	CP(*ki, *ki32, ki_fibnum);
1465 	CP(*ki, *ki32, ki_cr_flags);
1466 	CP(*ki, *ki32, ki_jid);
1467 	CP(*ki, *ki32, ki_numthreads);
1468 	CP(*ki, *ki32, ki_tid);
1469 	CP(*ki, *ki32, ki_pri);
1470 	freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1471 	freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1472 	PTRTRIM_CP(*ki, *ki32, ki_pcb);
1473 	PTRTRIM_CP(*ki, *ki32, ki_kstack);
1474 	PTRTRIM_CP(*ki, *ki32, ki_udata);
1475 	PTRTRIM_CP(*ki, *ki32, ki_tdaddr);
1476 	CP(*ki, *ki32, ki_sflag);
1477 	CP(*ki, *ki32, ki_tdflags);
1478 }
1479 #endif
1480 
1481 static ssize_t
1482 kern_proc_out_size(struct proc *p, int flags)
1483 {
1484 	ssize_t size = 0;
1485 
1486 	PROC_LOCK_ASSERT(p, MA_OWNED);
1487 
1488 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1489 #ifdef COMPAT_FREEBSD32
1490 		if ((flags & KERN_PROC_MASK32) != 0) {
1491 			size += sizeof(struct kinfo_proc32);
1492 		} else
1493 #endif
1494 			size += sizeof(struct kinfo_proc);
1495 	} else {
1496 #ifdef COMPAT_FREEBSD32
1497 		if ((flags & KERN_PROC_MASK32) != 0)
1498 			size += sizeof(struct kinfo_proc32) * p->p_numthreads;
1499 		else
1500 #endif
1501 			size += sizeof(struct kinfo_proc) * p->p_numthreads;
1502 	}
1503 	PROC_UNLOCK(p);
1504 	return (size);
1505 }
1506 
1507 int
1508 kern_proc_out(struct proc *p, struct sbuf *sb, int flags)
1509 {
1510 	struct thread *td;
1511 	struct kinfo_proc ki;
1512 #ifdef COMPAT_FREEBSD32
1513 	struct kinfo_proc32 ki32;
1514 #endif
1515 	int error;
1516 
1517 	PROC_LOCK_ASSERT(p, MA_OWNED);
1518 	MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1519 
1520 	error = 0;
1521 	fill_kinfo_proc(p, &ki);
1522 	if ((flags & KERN_PROC_NOTHREADS) != 0) {
1523 #ifdef COMPAT_FREEBSD32
1524 		if ((flags & KERN_PROC_MASK32) != 0) {
1525 			freebsd32_kinfo_proc_out(&ki, &ki32);
1526 			if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1527 				error = ENOMEM;
1528 		} else
1529 #endif
1530 			if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1531 				error = ENOMEM;
1532 	} else {
1533 		FOREACH_THREAD_IN_PROC(p, td) {
1534 			fill_kinfo_thread(td, &ki, 1);
1535 #ifdef COMPAT_FREEBSD32
1536 			if ((flags & KERN_PROC_MASK32) != 0) {
1537 				freebsd32_kinfo_proc_out(&ki, &ki32);
1538 				if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1539 					error = ENOMEM;
1540 			} else
1541 #endif
1542 				if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1543 					error = ENOMEM;
1544 			if (error != 0)
1545 				break;
1546 		}
1547 	}
1548 	PROC_UNLOCK(p);
1549 	return (error);
1550 }
1551 
1552 static int
1553 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
1554 {
1555 	struct sbuf sb;
1556 	struct kinfo_proc ki;
1557 	int error, error2;
1558 
1559 	if (req->oldptr == NULL)
1560 		return (SYSCTL_OUT(req, 0, kern_proc_out_size(p, flags)));
1561 
1562 	sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req);
1563 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1564 	error = kern_proc_out(p, &sb, flags);
1565 	error2 = sbuf_finish(&sb);
1566 	sbuf_delete(&sb);
1567 	if (error != 0)
1568 		return (error);
1569 	else if (error2 != 0)
1570 		return (error2);
1571 	return (0);
1572 }
1573 
1574 int
1575 proc_iterate(int (*cb)(struct proc *, void *), void *cbarg)
1576 {
1577 	struct proc *p;
1578 	int error, i, j;
1579 
1580 	for (i = 0; i < pidhashlock + 1; i++) {
1581 		sx_slock(&proctree_lock);
1582 		sx_slock(&pidhashtbl_lock[i]);
1583 		for (j = i; j <= pidhash; j += pidhashlock + 1) {
1584 			LIST_FOREACH(p, &pidhashtbl[j], p_hash) {
1585 				if (p->p_state == PRS_NEW)
1586 					continue;
1587 				error = cb(p, cbarg);
1588 				PROC_LOCK_ASSERT(p, MA_NOTOWNED);
1589 				if (error != 0) {
1590 					sx_sunlock(&pidhashtbl_lock[i]);
1591 					sx_sunlock(&proctree_lock);
1592 					return (error);
1593 				}
1594 			}
1595 		}
1596 		sx_sunlock(&pidhashtbl_lock[i]);
1597 		sx_sunlock(&proctree_lock);
1598 	}
1599 	return (0);
1600 }
1601 
1602 struct kern_proc_out_args {
1603 	struct sysctl_req *req;
1604 	int flags;
1605 	int oid_number;
1606 	int *name;
1607 };
1608 
1609 static int
1610 sysctl_kern_proc_iterate(struct proc *p, void *origarg)
1611 {
1612 	struct kern_proc_out_args *arg = origarg;
1613 	int *name = arg->name;
1614 	int oid_number = arg->oid_number;
1615 	int flags = arg->flags;
1616 	struct sysctl_req *req = arg->req;
1617 	int error = 0;
1618 
1619 	PROC_LOCK(p);
1620 
1621 	KASSERT(p->p_ucred != NULL,
1622 	    ("process credential is NULL for non-NEW proc"));
1623 	/*
1624 	 * Show a user only appropriate processes.
1625 	 */
1626 	if (p_cansee(curthread, p))
1627 		goto skip;
1628 	/*
1629 	 * TODO - make more efficient (see notes below).
1630 	 * do by session.
1631 	 */
1632 	switch (oid_number) {
1633 	case KERN_PROC_GID:
1634 		if (p->p_ucred->cr_gid != (gid_t)name[0])
1635 			goto skip;
1636 		break;
1637 
1638 	case KERN_PROC_PGRP:
1639 		/* could do this by traversing pgrp */
1640 		if (p->p_pgrp == NULL ||
1641 		    p->p_pgrp->pg_id != (pid_t)name[0])
1642 			goto skip;
1643 		break;
1644 
1645 	case KERN_PROC_RGID:
1646 		if (p->p_ucred->cr_rgid != (gid_t)name[0])
1647 			goto skip;
1648 		break;
1649 
1650 	case KERN_PROC_SESSION:
1651 		if (p->p_session == NULL ||
1652 		    p->p_session->s_sid != (pid_t)name[0])
1653 			goto skip;
1654 		break;
1655 
1656 	case KERN_PROC_TTY:
1657 		if ((p->p_flag & P_CONTROLT) == 0 ||
1658 		    p->p_session == NULL)
1659 			goto skip;
1660 		/* XXX proctree_lock */
1661 		SESS_LOCK(p->p_session);
1662 		if (p->p_session->s_ttyp == NULL ||
1663 		    tty_udev(p->p_session->s_ttyp) !=
1664 		    (dev_t)name[0]) {
1665 			SESS_UNLOCK(p->p_session);
1666 			goto skip;
1667 		}
1668 		SESS_UNLOCK(p->p_session);
1669 		break;
1670 
1671 	case KERN_PROC_UID:
1672 		if (p->p_ucred->cr_uid != (uid_t)name[0])
1673 			goto skip;
1674 		break;
1675 
1676 	case KERN_PROC_RUID:
1677 		if (p->p_ucred->cr_ruid != (uid_t)name[0])
1678 			goto skip;
1679 		break;
1680 
1681 	case KERN_PROC_PROC:
1682 		break;
1683 
1684 	default:
1685 		break;
1686 	}
1687 	error = sysctl_out_proc(p, req, flags);
1688 	PROC_LOCK_ASSERT(p, MA_NOTOWNED);
1689 	return (error);
1690 skip:
1691 	PROC_UNLOCK(p);
1692 	return (0);
1693 }
1694 
1695 static int
1696 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1697 {
1698 	struct kern_proc_out_args iterarg;
1699 	int *name = (int *)arg1;
1700 	u_int namelen = arg2;
1701 	struct proc *p;
1702 	int flags, oid_number;
1703 	int error = 0;
1704 
1705 	oid_number = oidp->oid_number;
1706 	if (oid_number != KERN_PROC_ALL &&
1707 	    (oid_number & KERN_PROC_INC_THREAD) == 0)
1708 		flags = KERN_PROC_NOTHREADS;
1709 	else {
1710 		flags = 0;
1711 		oid_number &= ~KERN_PROC_INC_THREAD;
1712 	}
1713 #ifdef COMPAT_FREEBSD32
1714 	if (req->flags & SCTL_MASK32)
1715 		flags |= KERN_PROC_MASK32;
1716 #endif
1717 	if (oid_number == KERN_PROC_PID) {
1718 		if (namelen != 1)
1719 			return (EINVAL);
1720 		error = sysctl_wire_old_buffer(req, 0);
1721 		if (error)
1722 			return (error);
1723 		sx_slock(&proctree_lock);
1724 		error = pget((pid_t)name[0], PGET_CANSEE, &p);
1725 		if (error == 0)
1726 			error = sysctl_out_proc(p, req, flags);
1727 		sx_sunlock(&proctree_lock);
1728 		return (error);
1729 	}
1730 
1731 	switch (oid_number) {
1732 	case KERN_PROC_ALL:
1733 		if (namelen != 0)
1734 			return (EINVAL);
1735 		break;
1736 	case KERN_PROC_PROC:
1737 		if (namelen != 0 && namelen != 1)
1738 			return (EINVAL);
1739 		break;
1740 	default:
1741 		if (namelen != 1)
1742 			return (EINVAL);
1743 		break;
1744 	}
1745 
1746 	if (req->oldptr == NULL) {
1747 		/* overestimate by 5 procs */
1748 		error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1749 		if (error)
1750 			return (error);
1751 	} else {
1752 		error = sysctl_wire_old_buffer(req, 0);
1753 		if (error != 0)
1754 			return (error);
1755 	}
1756 	iterarg.flags = flags;
1757 	iterarg.oid_number = oid_number;
1758 	iterarg.req = req;
1759 	iterarg.name = name;
1760 	error = proc_iterate(sysctl_kern_proc_iterate, &iterarg);
1761 	return (error);
1762 }
1763 
1764 struct pargs *
1765 pargs_alloc(int len)
1766 {
1767 	struct pargs *pa;
1768 
1769 	pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1770 		M_WAITOK);
1771 	refcount_init(&pa->ar_ref, 1);
1772 	pa->ar_length = len;
1773 	return (pa);
1774 }
1775 
1776 static void
1777 pargs_free(struct pargs *pa)
1778 {
1779 
1780 	free(pa, M_PARGS);
1781 }
1782 
1783 void
1784 pargs_hold(struct pargs *pa)
1785 {
1786 
1787 	if (pa == NULL)
1788 		return;
1789 	refcount_acquire(&pa->ar_ref);
1790 }
1791 
1792 void
1793 pargs_drop(struct pargs *pa)
1794 {
1795 
1796 	if (pa == NULL)
1797 		return;
1798 	if (refcount_release(&pa->ar_ref))
1799 		pargs_free(pa);
1800 }
1801 
1802 static int
1803 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1804     size_t len)
1805 {
1806 	ssize_t n;
1807 
1808 	/*
1809 	 * This may return a short read if the string is shorter than the chunk
1810 	 * and is aligned at the end of the page, and the following page is not
1811 	 * mapped.
1812 	 */
1813 	n = proc_readmem(td, p, (vm_offset_t)sptr, buf, len);
1814 	if (n <= 0)
1815 		return (ENOMEM);
1816 	return (0);
1817 }
1818 
1819 #define PROC_AUXV_MAX	256	/* Safety limit on auxv size. */
1820 
1821 enum proc_vector_type {
1822 	PROC_ARG,
1823 	PROC_ENV,
1824 	PROC_AUX,
1825 };
1826 
1827 #ifdef COMPAT_FREEBSD32
1828 static int
1829 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1830     size_t *vsizep, enum proc_vector_type type)
1831 {
1832 	struct freebsd32_ps_strings pss;
1833 	Elf32_Auxinfo aux;
1834 	vm_offset_t vptr, ptr;
1835 	uint32_t *proc_vector32;
1836 	char **proc_vector;
1837 	size_t vsize, size;
1838 	int i, error;
1839 
1840 	error = 0;
1841 	if (proc_readmem(td, p, PROC_PS_STRINGS(p), &pss, sizeof(pss)) !=
1842 	    sizeof(pss))
1843 		return (ENOMEM);
1844 	switch (type) {
1845 	case PROC_ARG:
1846 		vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1847 		vsize = pss.ps_nargvstr;
1848 		if (vsize > ARG_MAX)
1849 			return (ENOEXEC);
1850 		size = vsize * sizeof(int32_t);
1851 		break;
1852 	case PROC_ENV:
1853 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1854 		vsize = pss.ps_nenvstr;
1855 		if (vsize > ARG_MAX)
1856 			return (ENOEXEC);
1857 		size = vsize * sizeof(int32_t);
1858 		break;
1859 	case PROC_AUX:
1860 		vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1861 		    (pss.ps_nenvstr + 1) * sizeof(int32_t);
1862 		if (vptr % 4 != 0)
1863 			return (ENOEXEC);
1864 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1865 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1866 			    sizeof(aux))
1867 				return (ENOMEM);
1868 			if (aux.a_type == AT_NULL)
1869 				break;
1870 			ptr += sizeof(aux);
1871 		}
1872 		if (aux.a_type != AT_NULL)
1873 			return (ENOEXEC);
1874 		vsize = i + 1;
1875 		size = vsize * sizeof(aux);
1876 		break;
1877 	default:
1878 		KASSERT(0, ("Wrong proc vector type: %d", type));
1879 		return (EINVAL);
1880 	}
1881 	proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1882 	if (proc_readmem(td, p, vptr, proc_vector32, size) != size) {
1883 		error = ENOMEM;
1884 		goto done;
1885 	}
1886 	if (type == PROC_AUX) {
1887 		*proc_vectorp = (char **)proc_vector32;
1888 		*vsizep = vsize;
1889 		return (0);
1890 	}
1891 	proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1892 	for (i = 0; i < (int)vsize; i++)
1893 		proc_vector[i] = PTRIN(proc_vector32[i]);
1894 	*proc_vectorp = proc_vector;
1895 	*vsizep = vsize;
1896 done:
1897 	free(proc_vector32, M_TEMP);
1898 	return (error);
1899 }
1900 #endif
1901 
1902 static int
1903 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1904     size_t *vsizep, enum proc_vector_type type)
1905 {
1906 	struct ps_strings pss;
1907 	Elf_Auxinfo aux;
1908 	vm_offset_t vptr, ptr;
1909 	char **proc_vector;
1910 	size_t vsize, size;
1911 	int i;
1912 
1913 #ifdef COMPAT_FREEBSD32
1914 	if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1915 		return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1916 #endif
1917 	if (proc_readmem(td, p, PROC_PS_STRINGS(p), &pss, sizeof(pss)) !=
1918 	    sizeof(pss))
1919 		return (ENOMEM);
1920 	switch (type) {
1921 	case PROC_ARG:
1922 		vptr = (vm_offset_t)pss.ps_argvstr;
1923 		vsize = pss.ps_nargvstr;
1924 		if (vsize > ARG_MAX)
1925 			return (ENOEXEC);
1926 		size = vsize * sizeof(char *);
1927 		break;
1928 	case PROC_ENV:
1929 		vptr = (vm_offset_t)pss.ps_envstr;
1930 		vsize = pss.ps_nenvstr;
1931 		if (vsize > ARG_MAX)
1932 			return (ENOEXEC);
1933 		size = vsize * sizeof(char *);
1934 		break;
1935 	case PROC_AUX:
1936 		/*
1937 		 * The aux array is just above env array on the stack. Check
1938 		 * that the address is naturally aligned.
1939 		 */
1940 		vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1941 		    * sizeof(char *);
1942 #if __ELF_WORD_SIZE == 64
1943 		if (vptr % sizeof(uint64_t) != 0)
1944 #else
1945 		if (vptr % sizeof(uint32_t) != 0)
1946 #endif
1947 			return (ENOEXEC);
1948 		/*
1949 		 * We count the array size reading the aux vectors from the
1950 		 * stack until AT_NULL vector is returned.  So (to keep the code
1951 		 * simple) we read the process stack twice: the first time here
1952 		 * to find the size and the second time when copying the vectors
1953 		 * to the allocated proc_vector.
1954 		 */
1955 		for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1956 			if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1957 			    sizeof(aux))
1958 				return (ENOMEM);
1959 			if (aux.a_type == AT_NULL)
1960 				break;
1961 			ptr += sizeof(aux);
1962 		}
1963 		/*
1964 		 * If the PROC_AUXV_MAX entries are iterated over, and we have
1965 		 * not reached AT_NULL, it is most likely we are reading wrong
1966 		 * data: either the process doesn't have auxv array or data has
1967 		 * been modified. Return the error in this case.
1968 		 */
1969 		if (aux.a_type != AT_NULL)
1970 			return (ENOEXEC);
1971 		vsize = i + 1;
1972 		size = vsize * sizeof(aux);
1973 		break;
1974 	default:
1975 		KASSERT(0, ("Wrong proc vector type: %d", type));
1976 		return (EINVAL); /* In case we are built without INVARIANTS. */
1977 	}
1978 	proc_vector = malloc(size, M_TEMP, M_WAITOK);
1979 	if (proc_readmem(td, p, vptr, proc_vector, size) != size) {
1980 		free(proc_vector, M_TEMP);
1981 		return (ENOMEM);
1982 	}
1983 	*proc_vectorp = proc_vector;
1984 	*vsizep = vsize;
1985 
1986 	return (0);
1987 }
1988 
1989 #define GET_PS_STRINGS_CHUNK_SZ	256	/* Chunk size (bytes) for ps_strings operations. */
1990 
1991 static int
1992 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
1993     enum proc_vector_type type)
1994 {
1995 	size_t done, len, nchr, vsize;
1996 	int error, i;
1997 	char **proc_vector, *sptr;
1998 	char pss_string[GET_PS_STRINGS_CHUNK_SZ];
1999 
2000 	PROC_ASSERT_HELD(p);
2001 
2002 	/*
2003 	 * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
2004 	 */
2005 	nchr = 2 * (PATH_MAX + ARG_MAX);
2006 
2007 	error = get_proc_vector(td, p, &proc_vector, &vsize, type);
2008 	if (error != 0)
2009 		return (error);
2010 	for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
2011 		/*
2012 		 * The program may have scribbled into its argv array, e.g. to
2013 		 * remove some arguments.  If that has happened, break out
2014 		 * before trying to read from NULL.
2015 		 */
2016 		if (proc_vector[i] == NULL)
2017 			break;
2018 		for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
2019 			error = proc_read_string(td, p, sptr, pss_string,
2020 			    sizeof(pss_string));
2021 			if (error != 0)
2022 				goto done;
2023 			len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
2024 			if (done + len >= nchr)
2025 				len = nchr - done - 1;
2026 			sbuf_bcat(sb, pss_string, len);
2027 			if (len != GET_PS_STRINGS_CHUNK_SZ)
2028 				break;
2029 			done += GET_PS_STRINGS_CHUNK_SZ;
2030 		}
2031 		sbuf_bcat(sb, "", 1);
2032 		done += len + 1;
2033 	}
2034 done:
2035 	free(proc_vector, M_TEMP);
2036 	return (error);
2037 }
2038 
2039 int
2040 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
2041 {
2042 
2043 	return (get_ps_strings(curthread, p, sb, PROC_ARG));
2044 }
2045 
2046 int
2047 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
2048 {
2049 
2050 	return (get_ps_strings(curthread, p, sb, PROC_ENV));
2051 }
2052 
2053 int
2054 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb)
2055 {
2056 	size_t vsize, size;
2057 	char **auxv;
2058 	int error;
2059 
2060 	error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX);
2061 	if (error == 0) {
2062 #ifdef COMPAT_FREEBSD32
2063 		if (SV_PROC_FLAG(p, SV_ILP32) != 0)
2064 			size = vsize * sizeof(Elf32_Auxinfo);
2065 		else
2066 #endif
2067 			size = vsize * sizeof(Elf_Auxinfo);
2068 		if (sbuf_bcat(sb, auxv, size) != 0)
2069 			error = ENOMEM;
2070 		free(auxv, M_TEMP);
2071 	}
2072 	return (error);
2073 }
2074 
2075 /*
2076  * This sysctl allows a process to retrieve the argument list or process
2077  * title for another process without groping around in the address space
2078  * of the other process.  It also allow a process to set its own "process
2079  * title to a string of its own choice.
2080  */
2081 static int
2082 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
2083 {
2084 	int *name = (int *)arg1;
2085 	u_int namelen = arg2;
2086 	struct pargs *newpa, *pa;
2087 	struct proc *p;
2088 	struct sbuf sb;
2089 	int flags, error = 0, error2;
2090 	pid_t pid;
2091 
2092 	if (namelen != 1)
2093 		return (EINVAL);
2094 
2095 	p = curproc;
2096 	pid = (pid_t)name[0];
2097 	if (pid == -1) {
2098 		pid = p->p_pid;
2099 	}
2100 
2101 	/*
2102 	 * If the query is for this process and it is single-threaded, there
2103 	 * is nobody to modify pargs, thus we can just read.
2104 	 */
2105 	if (pid == p->p_pid && p->p_numthreads == 1 && req->newptr == NULL &&
2106 	    (pa = p->p_args) != NULL)
2107 		return (SYSCTL_OUT(req, pa->ar_args, pa->ar_length));
2108 
2109 	flags = PGET_CANSEE;
2110 	if (req->newptr != NULL)
2111 		flags |= PGET_ISCURRENT;
2112 	error = pget(pid, flags, &p);
2113 	if (error)
2114 		return (error);
2115 
2116 	pa = p->p_args;
2117 	if (pa != NULL) {
2118 		pargs_hold(pa);
2119 		PROC_UNLOCK(p);
2120 		error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
2121 		pargs_drop(pa);
2122 	} else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
2123 		_PHOLD(p);
2124 		PROC_UNLOCK(p);
2125 		sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2126 		sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2127 		error = proc_getargv(curthread, p, &sb);
2128 		error2 = sbuf_finish(&sb);
2129 		PRELE(p);
2130 		sbuf_delete(&sb);
2131 		if (error == 0 && error2 != 0)
2132 			error = error2;
2133 	} else {
2134 		PROC_UNLOCK(p);
2135 	}
2136 	if (error != 0 || req->newptr == NULL)
2137 		return (error);
2138 
2139 	if (req->newlen > ps_arg_cache_limit - sizeof(struct pargs))
2140 		return (ENOMEM);
2141 
2142 	if (req->newlen == 0) {
2143 		/*
2144 		 * Clear the argument pointer, so that we'll fetch arguments
2145 		 * with proc_getargv() until further notice.
2146 		 */
2147 		newpa = NULL;
2148 	} else {
2149 		newpa = pargs_alloc(req->newlen);
2150 		error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
2151 		if (error != 0) {
2152 			pargs_free(newpa);
2153 			return (error);
2154 		}
2155 	}
2156 	PROC_LOCK(p);
2157 	pa = p->p_args;
2158 	p->p_args = newpa;
2159 	PROC_UNLOCK(p);
2160 	pargs_drop(pa);
2161 	return (0);
2162 }
2163 
2164 /*
2165  * This sysctl allows a process to retrieve environment of another process.
2166  */
2167 static int
2168 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
2169 {
2170 	int *name = (int *)arg1;
2171 	u_int namelen = arg2;
2172 	struct proc *p;
2173 	struct sbuf sb;
2174 	int error, error2;
2175 
2176 	if (namelen != 1)
2177 		return (EINVAL);
2178 
2179 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2180 	if (error != 0)
2181 		return (error);
2182 	if ((p->p_flag & P_SYSTEM) != 0) {
2183 		PRELE(p);
2184 		return (0);
2185 	}
2186 
2187 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2188 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2189 	error = proc_getenvv(curthread, p, &sb);
2190 	error2 = sbuf_finish(&sb);
2191 	PRELE(p);
2192 	sbuf_delete(&sb);
2193 	return (error != 0 ? error : error2);
2194 }
2195 
2196 /*
2197  * This sysctl allows a process to retrieve ELF auxiliary vector of
2198  * another process.
2199  */
2200 static int
2201 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
2202 {
2203 	int *name = (int *)arg1;
2204 	u_int namelen = arg2;
2205 	struct proc *p;
2206 	struct sbuf sb;
2207 	int error, error2;
2208 
2209 	if (namelen != 1)
2210 		return (EINVAL);
2211 
2212 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2213 	if (error != 0)
2214 		return (error);
2215 	if ((p->p_flag & P_SYSTEM) != 0) {
2216 		PRELE(p);
2217 		return (0);
2218 	}
2219 	sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2220 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2221 	error = proc_getauxv(curthread, p, &sb);
2222 	error2 = sbuf_finish(&sb);
2223 	PRELE(p);
2224 	sbuf_delete(&sb);
2225 	return (error != 0 ? error : error2);
2226 }
2227 
2228 /*
2229  * Look up the canonical executable path running in the specified process.
2230  * It tries to return the same hardlink name as was used for execve(2).
2231  * This allows the programs that modify their behavior based on their progname,
2232  * to operate correctly.
2233  *
2234  * Result is returned in retbuf, it must not be freed, similar to vn_fullpath()
2235  *   calling conventions.
2236  * binname is a pointer to temporary string buffer of length MAXPATHLEN,
2237  *   allocated and freed by caller.
2238  * freebuf should be freed by caller, from the M_TEMP malloc type.
2239  */
2240 int
2241 proc_get_binpath(struct proc *p, char *binname, char **retbuf,
2242     char **freebuf)
2243 {
2244 	struct nameidata nd;
2245 	struct vnode *vp, *dvp;
2246 	size_t freepath_size;
2247 	int error;
2248 	bool do_fullpath;
2249 
2250 	PROC_LOCK_ASSERT(p, MA_OWNED);
2251 
2252 	vp = p->p_textvp;
2253 	if (vp == NULL) {
2254 		PROC_UNLOCK(p);
2255 		*retbuf = "";
2256 		*freebuf = NULL;
2257 		return (0);
2258 	}
2259 	vref(vp);
2260 	dvp = p->p_textdvp;
2261 	if (dvp != NULL)
2262 		vref(dvp);
2263 	if (p->p_binname != NULL)
2264 		strlcpy(binname, p->p_binname, MAXPATHLEN);
2265 	PROC_UNLOCK(p);
2266 
2267 	do_fullpath = true;
2268 	*freebuf = NULL;
2269 	if (dvp != NULL && binname[0] != '\0') {
2270 		freepath_size = MAXPATHLEN;
2271 		if (vn_fullpath_hardlink(vp, dvp, binname, strlen(binname),
2272 		    retbuf, freebuf, &freepath_size) == 0) {
2273 			/*
2274 			 * Recheck the looked up path.  The binary
2275 			 * might have been renamed or replaced, in
2276 			 * which case we should not report old name.
2277 			 */
2278 			NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, *retbuf);
2279 			error = namei(&nd);
2280 			if (error == 0) {
2281 				if (nd.ni_vp == vp)
2282 					do_fullpath = false;
2283 				vrele(nd.ni_vp);
2284 				NDFREE_PNBUF(&nd);
2285 			}
2286 		}
2287 	}
2288 	if (do_fullpath) {
2289 		free(*freebuf, M_TEMP);
2290 		*freebuf = NULL;
2291 		error = vn_fullpath(vp, retbuf, freebuf);
2292 	}
2293 	vrele(vp);
2294 	if (dvp != NULL)
2295 		vrele(dvp);
2296 	return (error);
2297 }
2298 
2299 /*
2300  * This sysctl allows a process to retrieve the path of the executable for
2301  * itself or another process.
2302  */
2303 static int
2304 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
2305 {
2306 	pid_t *pidp = (pid_t *)arg1;
2307 	unsigned int arglen = arg2;
2308 	struct proc *p;
2309 	char *retbuf, *freebuf, *binname;
2310 	int error;
2311 
2312 	if (arglen != 1)
2313 		return (EINVAL);
2314 	binname = malloc(MAXPATHLEN, M_TEMP, M_WAITOK);
2315 	binname[0] = '\0';
2316 	if (*pidp == -1) {	/* -1 means this process */
2317 		error = 0;
2318 		p = req->td->td_proc;
2319 		PROC_LOCK(p);
2320 	} else {
2321 		error = pget(*pidp, PGET_CANSEE, &p);
2322 	}
2323 
2324 	if (error == 0)
2325 		error = proc_get_binpath(p, binname, &retbuf, &freebuf);
2326 	free(binname, M_TEMP);
2327 	if (error != 0)
2328 		return (error);
2329 	error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
2330 	free(freebuf, M_TEMP);
2331 	return (error);
2332 }
2333 
2334 static int
2335 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
2336 {
2337 	struct proc *p;
2338 	char *sv_name;
2339 	int *name;
2340 	int namelen;
2341 	int error;
2342 
2343 	namelen = arg2;
2344 	if (namelen != 1)
2345 		return (EINVAL);
2346 
2347 	name = (int *)arg1;
2348 	error = pget((pid_t)name[0], PGET_CANSEE, &p);
2349 	if (error != 0)
2350 		return (error);
2351 	sv_name = p->p_sysent->sv_name;
2352 	PROC_UNLOCK(p);
2353 	return (sysctl_handle_string(oidp, sv_name, 0, req));
2354 }
2355 
2356 #ifdef KINFO_OVMENTRY_SIZE
2357 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
2358 #endif
2359 
2360 #ifdef COMPAT_FREEBSD7
2361 static int
2362 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
2363 {
2364 	vm_map_entry_t entry, tmp_entry;
2365 	unsigned int last_timestamp, namelen;
2366 	char *fullpath, *freepath;
2367 	struct kinfo_ovmentry *kve;
2368 	struct vattr va;
2369 	struct ucred *cred;
2370 	int error, *name;
2371 	struct vnode *vp;
2372 	struct proc *p;
2373 	vm_map_t map;
2374 	struct vmspace *vm;
2375 
2376 	namelen = arg2;
2377 	if (namelen != 1)
2378 		return (EINVAL);
2379 
2380 	name = (int *)arg1;
2381 	error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2382 	if (error != 0)
2383 		return (error);
2384 	vm = vmspace_acquire_ref(p);
2385 	if (vm == NULL) {
2386 		PRELE(p);
2387 		return (ESRCH);
2388 	}
2389 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2390 
2391 	map = &vm->vm_map;
2392 	vm_map_lock_read(map);
2393 	VM_MAP_ENTRY_FOREACH(entry, map) {
2394 		vm_object_t obj, tobj, lobj;
2395 		vm_offset_t addr;
2396 
2397 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2398 			continue;
2399 
2400 		bzero(kve, sizeof(*kve));
2401 		kve->kve_structsize = sizeof(*kve);
2402 
2403 		kve->kve_private_resident = 0;
2404 		obj = entry->object.vm_object;
2405 		if (obj != NULL) {
2406 			VM_OBJECT_RLOCK(obj);
2407 			if (obj->shadow_count == 1)
2408 				kve->kve_private_resident =
2409 				    obj->resident_page_count;
2410 		}
2411 		kve->kve_resident = 0;
2412 		addr = entry->start;
2413 		while (addr < entry->end) {
2414 			if (pmap_extract(map->pmap, addr))
2415 				kve->kve_resident++;
2416 			addr += PAGE_SIZE;
2417 		}
2418 
2419 		for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2420 			if (tobj != obj) {
2421 				VM_OBJECT_RLOCK(tobj);
2422 				kve->kve_offset += tobj->backing_object_offset;
2423 			}
2424 			if (lobj != obj)
2425 				VM_OBJECT_RUNLOCK(lobj);
2426 			lobj = tobj;
2427 		}
2428 
2429 		kve->kve_start = (void*)entry->start;
2430 		kve->kve_end = (void*)entry->end;
2431 		kve->kve_offset += (off_t)entry->offset;
2432 
2433 		if (entry->protection & VM_PROT_READ)
2434 			kve->kve_protection |= KVME_PROT_READ;
2435 		if (entry->protection & VM_PROT_WRITE)
2436 			kve->kve_protection |= KVME_PROT_WRITE;
2437 		if (entry->protection & VM_PROT_EXECUTE)
2438 			kve->kve_protection |= KVME_PROT_EXEC;
2439 
2440 		if (entry->eflags & MAP_ENTRY_COW)
2441 			kve->kve_flags |= KVME_FLAG_COW;
2442 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2443 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2444 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2445 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2446 
2447 		last_timestamp = map->timestamp;
2448 		vm_map_unlock_read(map);
2449 
2450 		kve->kve_fileid = 0;
2451 		kve->kve_fsid = 0;
2452 		freepath = NULL;
2453 		fullpath = "";
2454 		if (lobj) {
2455 			kve->kve_type = vm_object_kvme_type(lobj, &vp);
2456 			if (kve->kve_type == KVME_TYPE_MGTDEVICE)
2457 				kve->kve_type = KVME_TYPE_UNKNOWN;
2458 			if (vp != NULL)
2459 				vref(vp);
2460 			if (lobj != obj)
2461 				VM_OBJECT_RUNLOCK(lobj);
2462 
2463 			kve->kve_ref_count = obj->ref_count;
2464 			kve->kve_shadow_count = obj->shadow_count;
2465 			VM_OBJECT_RUNLOCK(obj);
2466 			if (vp != NULL) {
2467 				vn_fullpath(vp, &fullpath, &freepath);
2468 				cred = curthread->td_ucred;
2469 				vn_lock(vp, LK_SHARED | LK_RETRY);
2470 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2471 					kve->kve_fileid = va.va_fileid;
2472 					/* truncate */
2473 					kve->kve_fsid = va.va_fsid;
2474 				}
2475 				vput(vp);
2476 			}
2477 		} else {
2478 			kve->kve_type = KVME_TYPE_NONE;
2479 			kve->kve_ref_count = 0;
2480 			kve->kve_shadow_count = 0;
2481 		}
2482 
2483 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2484 		if (freepath != NULL)
2485 			free(freepath, M_TEMP);
2486 
2487 		error = SYSCTL_OUT(req, kve, sizeof(*kve));
2488 		vm_map_lock_read(map);
2489 		if (error)
2490 			break;
2491 		if (last_timestamp != map->timestamp) {
2492 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2493 			entry = tmp_entry;
2494 		}
2495 	}
2496 	vm_map_unlock_read(map);
2497 	vmspace_free(vm);
2498 	PRELE(p);
2499 	free(kve, M_TEMP);
2500 	return (error);
2501 }
2502 #endif	/* COMPAT_FREEBSD7 */
2503 
2504 #ifdef KINFO_VMENTRY_SIZE
2505 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2506 #endif
2507 
2508 void
2509 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry,
2510     int *resident_count, bool *super)
2511 {
2512 	vm_object_t obj, tobj;
2513 	vm_page_t m, m_adv;
2514 	vm_offset_t addr;
2515 	vm_paddr_t pa;
2516 	vm_pindex_t pi, pi_adv, pindex;
2517 
2518 	*super = false;
2519 	*resident_count = 0;
2520 	if (vmmap_skip_res_cnt)
2521 		return;
2522 
2523 	pa = 0;
2524 	obj = entry->object.vm_object;
2525 	addr = entry->start;
2526 	m_adv = NULL;
2527 	pi = OFF_TO_IDX(entry->offset);
2528 	for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) {
2529 		if (m_adv != NULL) {
2530 			m = m_adv;
2531 		} else {
2532 			pi_adv = atop(entry->end - addr);
2533 			pindex = pi;
2534 			for (tobj = obj;; tobj = tobj->backing_object) {
2535 				m = vm_page_find_least(tobj, pindex);
2536 				if (m != NULL) {
2537 					if (m->pindex == pindex)
2538 						break;
2539 					if (pi_adv > m->pindex - pindex) {
2540 						pi_adv = m->pindex - pindex;
2541 						m_adv = m;
2542 					}
2543 				}
2544 				if (tobj->backing_object == NULL)
2545 					goto next;
2546 				pindex += OFF_TO_IDX(tobj->
2547 				    backing_object_offset);
2548 			}
2549 		}
2550 		m_adv = NULL;
2551 		if (m->psind != 0 && addr + pagesizes[1] <= entry->end &&
2552 		    (addr & (pagesizes[1] - 1)) == 0 &&
2553 		    (pmap_mincore(map->pmap, addr, &pa) & MINCORE_SUPER) != 0) {
2554 			*super = true;
2555 			pi_adv = atop(pagesizes[1]);
2556 		} else {
2557 			/*
2558 			 * We do not test the found page on validity.
2559 			 * Either the page is busy and being paged in,
2560 			 * or it was invalidated.  The first case
2561 			 * should be counted as resident, the second
2562 			 * is not so clear; we do account both.
2563 			 */
2564 			pi_adv = 1;
2565 		}
2566 		*resident_count += pi_adv;
2567 next:;
2568 	}
2569 }
2570 
2571 /*
2572  * Must be called with the process locked and will return unlocked.
2573  */
2574 int
2575 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb, ssize_t maxlen, int flags)
2576 {
2577 	vm_map_entry_t entry, tmp_entry;
2578 	struct vattr va;
2579 	vm_map_t map;
2580 	vm_object_t lobj, nobj, obj, tobj;
2581 	char *fullpath, *freepath;
2582 	struct kinfo_vmentry *kve;
2583 	struct ucred *cred;
2584 	struct vnode *vp;
2585 	struct vmspace *vm;
2586 	vm_offset_t addr;
2587 	unsigned int last_timestamp;
2588 	int error;
2589 	bool guard, super;
2590 
2591 	PROC_LOCK_ASSERT(p, MA_OWNED);
2592 
2593 	_PHOLD(p);
2594 	PROC_UNLOCK(p);
2595 	vm = vmspace_acquire_ref(p);
2596 	if (vm == NULL) {
2597 		PRELE(p);
2598 		return (ESRCH);
2599 	}
2600 	kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK | M_ZERO);
2601 
2602 	error = 0;
2603 	map = &vm->vm_map;
2604 	vm_map_lock_read(map);
2605 	VM_MAP_ENTRY_FOREACH(entry, map) {
2606 		if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2607 			continue;
2608 
2609 		addr = entry->end;
2610 		bzero(kve, sizeof(*kve));
2611 		obj = entry->object.vm_object;
2612 		if (obj != NULL) {
2613 			if ((obj->flags & OBJ_ANON) != 0)
2614 				kve->kve_obj = (uintptr_t)obj;
2615 
2616 			for (tobj = obj; tobj != NULL;
2617 			    tobj = tobj->backing_object) {
2618 				VM_OBJECT_RLOCK(tobj);
2619 				kve->kve_offset += tobj->backing_object_offset;
2620 				lobj = tobj;
2621 			}
2622 			if (obj->backing_object == NULL)
2623 				kve->kve_private_resident =
2624 				    obj->resident_page_count;
2625 			kern_proc_vmmap_resident(map, entry,
2626 			    &kve->kve_resident, &super);
2627 			if (super)
2628 				kve->kve_flags |= KVME_FLAG_SUPER;
2629 			for (tobj = obj; tobj != NULL; tobj = nobj) {
2630 				nobj = tobj->backing_object;
2631 				if (tobj != obj && tobj != lobj)
2632 					VM_OBJECT_RUNLOCK(tobj);
2633 			}
2634 		} else {
2635 			lobj = NULL;
2636 		}
2637 
2638 		kve->kve_start = entry->start;
2639 		kve->kve_end = entry->end;
2640 		kve->kve_offset += entry->offset;
2641 
2642 		if (entry->protection & VM_PROT_READ)
2643 			kve->kve_protection |= KVME_PROT_READ;
2644 		if (entry->protection & VM_PROT_WRITE)
2645 			kve->kve_protection |= KVME_PROT_WRITE;
2646 		if (entry->protection & VM_PROT_EXECUTE)
2647 			kve->kve_protection |= KVME_PROT_EXEC;
2648 
2649 		if (entry->eflags & MAP_ENTRY_COW)
2650 			kve->kve_flags |= KVME_FLAG_COW;
2651 		if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2652 			kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2653 		if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2654 			kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2655 		if (entry->eflags & MAP_ENTRY_GROWS_UP)
2656 			kve->kve_flags |= KVME_FLAG_GROWS_UP;
2657 		if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2658 			kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2659 		if (entry->eflags & MAP_ENTRY_USER_WIRED)
2660 			kve->kve_flags |= KVME_FLAG_USER_WIRED;
2661 
2662 		guard = (entry->eflags & MAP_ENTRY_GUARD) != 0;
2663 
2664 		last_timestamp = map->timestamp;
2665 		vm_map_unlock_read(map);
2666 
2667 		freepath = NULL;
2668 		fullpath = "";
2669 		if (lobj != NULL) {
2670 			kve->kve_type = vm_object_kvme_type(lobj, &vp);
2671 			if (vp != NULL)
2672 				vref(vp);
2673 			if (lobj != obj)
2674 				VM_OBJECT_RUNLOCK(lobj);
2675 
2676 			kve->kve_ref_count = obj->ref_count;
2677 			kve->kve_shadow_count = obj->shadow_count;
2678 			VM_OBJECT_RUNLOCK(obj);
2679 			if (vp != NULL) {
2680 				vn_fullpath(vp, &fullpath, &freepath);
2681 				kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2682 				cred = curthread->td_ucred;
2683 				vn_lock(vp, LK_SHARED | LK_RETRY);
2684 				if (VOP_GETATTR(vp, &va, cred) == 0) {
2685 					kve->kve_vn_fileid = va.va_fileid;
2686 					kve->kve_vn_fsid = va.va_fsid;
2687 					kve->kve_vn_fsid_freebsd11 =
2688 					    kve->kve_vn_fsid; /* truncate */
2689 					kve->kve_vn_mode =
2690 					    MAKEIMODE(va.va_type, va.va_mode);
2691 					kve->kve_vn_size = va.va_size;
2692 					kve->kve_vn_rdev = va.va_rdev;
2693 					kve->kve_vn_rdev_freebsd11 =
2694 					    kve->kve_vn_rdev; /* truncate */
2695 					kve->kve_status = KF_ATTR_VALID;
2696 				}
2697 				vput(vp);
2698 			}
2699 		} else {
2700 			kve->kve_type = guard ? KVME_TYPE_GUARD :
2701 			    KVME_TYPE_NONE;
2702 			kve->kve_ref_count = 0;
2703 			kve->kve_shadow_count = 0;
2704 		}
2705 
2706 		strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2707 		if (freepath != NULL)
2708 			free(freepath, M_TEMP);
2709 
2710 		/* Pack record size down */
2711 		if ((flags & KERN_VMMAP_PACK_KINFO) != 0)
2712 			kve->kve_structsize =
2713 			    offsetof(struct kinfo_vmentry, kve_path) +
2714 			    strlen(kve->kve_path) + 1;
2715 		else
2716 			kve->kve_structsize = sizeof(*kve);
2717 		kve->kve_structsize = roundup(kve->kve_structsize,
2718 		    sizeof(uint64_t));
2719 
2720 		/* Halt filling and truncate rather than exceeding maxlen */
2721 		if (maxlen != -1 && maxlen < kve->kve_structsize) {
2722 			error = 0;
2723 			vm_map_lock_read(map);
2724 			break;
2725 		} else if (maxlen != -1)
2726 			maxlen -= kve->kve_structsize;
2727 
2728 		if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0)
2729 			error = ENOMEM;
2730 		vm_map_lock_read(map);
2731 		if (error != 0)
2732 			break;
2733 		if (last_timestamp != map->timestamp) {
2734 			vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2735 			entry = tmp_entry;
2736 		}
2737 	}
2738 	vm_map_unlock_read(map);
2739 	vmspace_free(vm);
2740 	PRELE(p);
2741 	free(kve, M_TEMP);
2742 	return (error);
2743 }
2744 
2745 static int
2746 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2747 {
2748 	struct proc *p;
2749 	struct sbuf sb;
2750 	u_int namelen;
2751 	int error, error2, *name;
2752 
2753 	namelen = arg2;
2754 	if (namelen != 1)
2755 		return (EINVAL);
2756 
2757 	name = (int *)arg1;
2758 	sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req);
2759 	sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2760 	error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
2761 	if (error != 0) {
2762 		sbuf_delete(&sb);
2763 		return (error);
2764 	}
2765 	error = kern_proc_vmmap_out(p, &sb, -1, KERN_VMMAP_PACK_KINFO);
2766 	error2 = sbuf_finish(&sb);
2767 	sbuf_delete(&sb);
2768 	return (error != 0 ? error : error2);
2769 }
2770 
2771 #if defined(STACK) || defined(DDB)
2772 static int
2773 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2774 {
2775 	struct kinfo_kstack *kkstp;
2776 	int error, i, *name, numthreads;
2777 	lwpid_t *lwpidarray;
2778 	struct thread *td;
2779 	struct stack *st;
2780 	struct sbuf sb;
2781 	struct proc *p;
2782 	u_int namelen;
2783 
2784 	namelen = arg2;
2785 	if (namelen != 1)
2786 		return (EINVAL);
2787 
2788 	name = (int *)arg1;
2789 	error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2790 	if (error != 0)
2791 		return (error);
2792 
2793 	kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2794 	st = stack_create(M_WAITOK);
2795 
2796 	lwpidarray = NULL;
2797 	PROC_LOCK(p);
2798 	do {
2799 		if (lwpidarray != NULL) {
2800 			free(lwpidarray, M_TEMP);
2801 			lwpidarray = NULL;
2802 		}
2803 		numthreads = p->p_numthreads;
2804 		PROC_UNLOCK(p);
2805 		lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2806 		    M_WAITOK | M_ZERO);
2807 		PROC_LOCK(p);
2808 	} while (numthreads < p->p_numthreads);
2809 
2810 	/*
2811 	 * XXXRW: During the below loop, execve(2) and countless other sorts
2812 	 * of changes could have taken place.  Should we check to see if the
2813 	 * vmspace has been replaced, or the like, in order to prevent
2814 	 * giving a snapshot that spans, say, execve(2), with some threads
2815 	 * before and some after?  Among other things, the credentials could
2816 	 * have changed, in which case the right to extract debug info might
2817 	 * no longer be assured.
2818 	 */
2819 	i = 0;
2820 	FOREACH_THREAD_IN_PROC(p, td) {
2821 		KASSERT(i < numthreads,
2822 		    ("sysctl_kern_proc_kstack: numthreads"));
2823 		lwpidarray[i] = td->td_tid;
2824 		i++;
2825 	}
2826 	PROC_UNLOCK(p);
2827 	numthreads = i;
2828 	for (i = 0; i < numthreads; i++) {
2829 		td = tdfind(lwpidarray[i], p->p_pid);
2830 		if (td == NULL) {
2831 			continue;
2832 		}
2833 		bzero(kkstp, sizeof(*kkstp));
2834 		(void)sbuf_new(&sb, kkstp->kkst_trace,
2835 		    sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2836 		thread_lock(td);
2837 		kkstp->kkst_tid = td->td_tid;
2838 		if (TD_IS_SWAPPED(td))
2839 			kkstp->kkst_state = KKST_STATE_SWAPPED;
2840 		else if (stack_save_td(st, td) == 0)
2841 			kkstp->kkst_state = KKST_STATE_STACKOK;
2842 		else
2843 			kkstp->kkst_state = KKST_STATE_RUNNING;
2844 		thread_unlock(td);
2845 		PROC_UNLOCK(p);
2846 		stack_sbuf_print(&sb, st);
2847 		sbuf_finish(&sb);
2848 		sbuf_delete(&sb);
2849 		error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2850 		if (error)
2851 			break;
2852 	}
2853 	PRELE(p);
2854 	if (lwpidarray != NULL)
2855 		free(lwpidarray, M_TEMP);
2856 	stack_destroy(st);
2857 	free(kkstp, M_TEMP);
2858 	return (error);
2859 }
2860 #endif
2861 
2862 /*
2863  * This sysctl allows a process to retrieve the full list of groups from
2864  * itself or another process.
2865  */
2866 static int
2867 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2868 {
2869 	pid_t *pidp = (pid_t *)arg1;
2870 	unsigned int arglen = arg2;
2871 	struct proc *p;
2872 	struct ucred *cred;
2873 	int error;
2874 
2875 	if (arglen != 1)
2876 		return (EINVAL);
2877 	if (*pidp == -1) {	/* -1 means this process */
2878 		p = req->td->td_proc;
2879 		PROC_LOCK(p);
2880 	} else {
2881 		error = pget(*pidp, PGET_CANSEE, &p);
2882 		if (error != 0)
2883 			return (error);
2884 	}
2885 
2886 	cred = crhold(p->p_ucred);
2887 	PROC_UNLOCK(p);
2888 
2889 	error = SYSCTL_OUT(req, cred->cr_groups,
2890 	    cred->cr_ngroups * sizeof(gid_t));
2891 	crfree(cred);
2892 	return (error);
2893 }
2894 
2895 /*
2896  * This sysctl allows a process to retrieve or/and set the resource limit for
2897  * another process.
2898  */
2899 static int
2900 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2901 {
2902 	int *name = (int *)arg1;
2903 	u_int namelen = arg2;
2904 	struct rlimit rlim;
2905 	struct proc *p;
2906 	u_int which;
2907 	int flags, error;
2908 
2909 	if (namelen != 2)
2910 		return (EINVAL);
2911 
2912 	which = (u_int)name[1];
2913 	if (which >= RLIM_NLIMITS)
2914 		return (EINVAL);
2915 
2916 	if (req->newptr != NULL && req->newlen != sizeof(rlim))
2917 		return (EINVAL);
2918 
2919 	flags = PGET_HOLD | PGET_NOTWEXIT;
2920 	if (req->newptr != NULL)
2921 		flags |= PGET_CANDEBUG;
2922 	else
2923 		flags |= PGET_CANSEE;
2924 	error = pget((pid_t)name[0], flags, &p);
2925 	if (error != 0)
2926 		return (error);
2927 
2928 	/*
2929 	 * Retrieve limit.
2930 	 */
2931 	if (req->oldptr != NULL) {
2932 		PROC_LOCK(p);
2933 		lim_rlimit_proc(p, which, &rlim);
2934 		PROC_UNLOCK(p);
2935 	}
2936 	error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
2937 	if (error != 0)
2938 		goto errout;
2939 
2940 	/*
2941 	 * Set limit.
2942 	 */
2943 	if (req->newptr != NULL) {
2944 		error = SYSCTL_IN(req, &rlim, sizeof(rlim));
2945 		if (error == 0)
2946 			error = kern_proc_setrlimit(curthread, p, which, &rlim);
2947 	}
2948 
2949 errout:
2950 	PRELE(p);
2951 	return (error);
2952 }
2953 
2954 /*
2955  * This sysctl allows a process to retrieve ps_strings structure location of
2956  * another process.
2957  */
2958 static int
2959 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
2960 {
2961 	int *name = (int *)arg1;
2962 	u_int namelen = arg2;
2963 	struct proc *p;
2964 	vm_offset_t ps_strings;
2965 	int error;
2966 #ifdef COMPAT_FREEBSD32
2967 	uint32_t ps_strings32;
2968 #endif
2969 
2970 	if (namelen != 1)
2971 		return (EINVAL);
2972 
2973 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2974 	if (error != 0)
2975 		return (error);
2976 #ifdef COMPAT_FREEBSD32
2977 	if ((req->flags & SCTL_MASK32) != 0) {
2978 		/*
2979 		 * We return 0 if the 32 bit emulation request is for a 64 bit
2980 		 * process.
2981 		 */
2982 		ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
2983 		    PTROUT(PROC_PS_STRINGS(p)) : 0;
2984 		PROC_UNLOCK(p);
2985 		error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
2986 		return (error);
2987 	}
2988 #endif
2989 	ps_strings = PROC_PS_STRINGS(p);
2990 	PROC_UNLOCK(p);
2991 	error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
2992 	return (error);
2993 }
2994 
2995 /*
2996  * This sysctl allows a process to retrieve umask of another process.
2997  */
2998 static int
2999 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
3000 {
3001 	int *name = (int *)arg1;
3002 	u_int namelen = arg2;
3003 	struct proc *p;
3004 	int error;
3005 	u_short cmask;
3006 	pid_t pid;
3007 
3008 	if (namelen != 1)
3009 		return (EINVAL);
3010 
3011 	pid = (pid_t)name[0];
3012 	p = curproc;
3013 	if (pid == p->p_pid || pid == 0) {
3014 		cmask = p->p_pd->pd_cmask;
3015 		goto out;
3016 	}
3017 
3018 	error = pget(pid, PGET_WANTREAD, &p);
3019 	if (error != 0)
3020 		return (error);
3021 
3022 	cmask = p->p_pd->pd_cmask;
3023 	PRELE(p);
3024 out:
3025 	error = SYSCTL_OUT(req, &cmask, sizeof(cmask));
3026 	return (error);
3027 }
3028 
3029 /*
3030  * This sysctl allows a process to set and retrieve binary osreldate of
3031  * another process.
3032  */
3033 static int
3034 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
3035 {
3036 	int *name = (int *)arg1;
3037 	u_int namelen = arg2;
3038 	struct proc *p;
3039 	int flags, error, osrel;
3040 
3041 	if (namelen != 1)
3042 		return (EINVAL);
3043 
3044 	if (req->newptr != NULL && req->newlen != sizeof(osrel))
3045 		return (EINVAL);
3046 
3047 	flags = PGET_HOLD | PGET_NOTWEXIT;
3048 	if (req->newptr != NULL)
3049 		flags |= PGET_CANDEBUG;
3050 	else
3051 		flags |= PGET_CANSEE;
3052 	error = pget((pid_t)name[0], flags, &p);
3053 	if (error != 0)
3054 		return (error);
3055 
3056 	error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
3057 	if (error != 0)
3058 		goto errout;
3059 
3060 	if (req->newptr != NULL) {
3061 		error = SYSCTL_IN(req, &osrel, sizeof(osrel));
3062 		if (error != 0)
3063 			goto errout;
3064 		if (osrel < 0) {
3065 			error = EINVAL;
3066 			goto errout;
3067 		}
3068 		p->p_osrel = osrel;
3069 	}
3070 errout:
3071 	PRELE(p);
3072 	return (error);
3073 }
3074 
3075 static int
3076 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS)
3077 {
3078 	int *name = (int *)arg1;
3079 	u_int namelen = arg2;
3080 	struct proc *p;
3081 	struct kinfo_sigtramp kst;
3082 	const struct sysentvec *sv;
3083 	int error;
3084 #ifdef COMPAT_FREEBSD32
3085 	struct kinfo_sigtramp32 kst32;
3086 #endif
3087 
3088 	if (namelen != 1)
3089 		return (EINVAL);
3090 
3091 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
3092 	if (error != 0)
3093 		return (error);
3094 	sv = p->p_sysent;
3095 #ifdef COMPAT_FREEBSD32
3096 	if ((req->flags & SCTL_MASK32) != 0) {
3097 		bzero(&kst32, sizeof(kst32));
3098 		if (SV_PROC_FLAG(p, SV_ILP32)) {
3099 			if (sv->sv_sigcode_base != 0) {
3100 				kst32.ksigtramp_start = sv->sv_sigcode_base;
3101 				kst32.ksigtramp_end = sv->sv_sigcode_base +
3102 				    ((sv->sv_flags & SV_DSO_SIG) == 0 ?
3103 				    *sv->sv_szsigcode :
3104 				    (uintptr_t)sv->sv_szsigcode);
3105 			} else {
3106 				kst32.ksigtramp_start = PROC_PS_STRINGS(p) -
3107 				    *sv->sv_szsigcode;
3108 				kst32.ksigtramp_end = PROC_PS_STRINGS(p);
3109 			}
3110 		}
3111 		PROC_UNLOCK(p);
3112 		error = SYSCTL_OUT(req, &kst32, sizeof(kst32));
3113 		return (error);
3114 	}
3115 #endif
3116 	bzero(&kst, sizeof(kst));
3117 	if (sv->sv_sigcode_base != 0) {
3118 		kst.ksigtramp_start = (char *)sv->sv_sigcode_base;
3119 		kst.ksigtramp_end = (char *)sv->sv_sigcode_base +
3120 		    ((sv->sv_flags & SV_DSO_SIG) == 0 ? *sv->sv_szsigcode :
3121 		    (uintptr_t)sv->sv_szsigcode);
3122 	} else {
3123 		kst.ksigtramp_start = (char *)PROC_PS_STRINGS(p) -
3124 		    *sv->sv_szsigcode;
3125 		kst.ksigtramp_end = (char *)PROC_PS_STRINGS(p);
3126 	}
3127 	PROC_UNLOCK(p);
3128 	error = SYSCTL_OUT(req, &kst, sizeof(kst));
3129 	return (error);
3130 }
3131 
3132 static int
3133 sysctl_kern_proc_sigfastblk(SYSCTL_HANDLER_ARGS)
3134 {
3135 	int *name = (int *)arg1;
3136 	u_int namelen = arg2;
3137 	pid_t pid;
3138 	struct proc *p;
3139 	struct thread *td1;
3140 	uintptr_t addr;
3141 #ifdef COMPAT_FREEBSD32
3142 	uint32_t addr32;
3143 #endif
3144 	int error;
3145 
3146 	if (namelen != 1 || req->newptr != NULL)
3147 		return (EINVAL);
3148 
3149 	pid = (pid_t)name[0];
3150 	error = pget(pid, PGET_HOLD | PGET_NOTWEXIT | PGET_CANDEBUG, &p);
3151 	if (error != 0)
3152 		return (error);
3153 
3154 	PROC_LOCK(p);
3155 #ifdef COMPAT_FREEBSD32
3156 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3157 		if (!SV_PROC_FLAG(p, SV_ILP32)) {
3158 			error = EINVAL;
3159 			goto errlocked;
3160 		}
3161 	}
3162 #endif
3163 	if (pid <= PID_MAX) {
3164 		td1 = FIRST_THREAD_IN_PROC(p);
3165 	} else {
3166 		FOREACH_THREAD_IN_PROC(p, td1) {
3167 			if (td1->td_tid == pid)
3168 				break;
3169 		}
3170 	}
3171 	if (td1 == NULL) {
3172 		error = ESRCH;
3173 		goto errlocked;
3174 	}
3175 	/*
3176 	 * The access to the private thread flags.  It is fine as far
3177 	 * as no out-of-thin-air values are read from td_pflags, and
3178 	 * usermode read of the td_sigblock_ptr is racy inherently,
3179 	 * since target process might have already changed it
3180 	 * meantime.
3181 	 */
3182 	if ((td1->td_pflags & TDP_SIGFASTBLOCK) != 0)
3183 		addr = (uintptr_t)td1->td_sigblock_ptr;
3184 	else
3185 		error = ENOTTY;
3186 
3187 errlocked:
3188 	_PRELE(p);
3189 	PROC_UNLOCK(p);
3190 	if (error != 0)
3191 		return (error);
3192 
3193 #ifdef COMPAT_FREEBSD32
3194 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3195 		addr32 = addr;
3196 		error = SYSCTL_OUT(req, &addr32, sizeof(addr32));
3197 	} else
3198 #endif
3199 		error = SYSCTL_OUT(req, &addr, sizeof(addr));
3200 	return (error);
3201 }
3202 
3203 static int
3204 sysctl_kern_proc_vm_layout(SYSCTL_HANDLER_ARGS)
3205 {
3206 	struct kinfo_vm_layout kvm;
3207 	struct proc *p;
3208 	struct vmspace *vmspace;
3209 	int error, *name;
3210 
3211 	name = (int *)arg1;
3212 	if ((u_int)arg2 != 1)
3213 		return (EINVAL);
3214 
3215 	error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
3216 	if (error != 0)
3217 		return (error);
3218 #ifdef COMPAT_FREEBSD32
3219 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3220 		if (!SV_PROC_FLAG(p, SV_ILP32)) {
3221 			PROC_UNLOCK(p);
3222 			return (EINVAL);
3223 		}
3224 	}
3225 #endif
3226 	vmspace = vmspace_acquire_ref(p);
3227 	PROC_UNLOCK(p);
3228 
3229 	memset(&kvm, 0, sizeof(kvm));
3230 	kvm.kvm_min_user_addr = vm_map_min(&vmspace->vm_map);
3231 	kvm.kvm_max_user_addr = vm_map_max(&vmspace->vm_map);
3232 	kvm.kvm_text_addr = (uintptr_t)vmspace->vm_taddr;
3233 	kvm.kvm_text_size = vmspace->vm_tsize;
3234 	kvm.kvm_data_addr = (uintptr_t)vmspace->vm_daddr;
3235 	kvm.kvm_data_size = vmspace->vm_dsize;
3236 	kvm.kvm_stack_addr = (uintptr_t)vmspace->vm_maxsaddr;
3237 	kvm.kvm_stack_size = vmspace->vm_ssize;
3238 	if ((vmspace->vm_map.flags & MAP_WIREFUTURE) != 0)
3239 		kvm.kvm_map_flags |= KMAP_FLAG_WIREFUTURE;
3240 	if ((vmspace->vm_map.flags & MAP_ASLR) != 0)
3241 		kvm.kvm_map_flags |= KMAP_FLAG_ASLR;
3242 	if ((vmspace->vm_map.flags & MAP_ASLR_IGNSTART) != 0)
3243 		kvm.kvm_map_flags |= KMAP_FLAG_ASLR_IGNSTART;
3244 	if ((vmspace->vm_map.flags & MAP_WXORX) != 0)
3245 		kvm.kvm_map_flags |= KMAP_FLAG_WXORX;
3246 	if ((vmspace->vm_map.flags & MAP_ASLR_STACK) != 0)
3247 		kvm.kvm_map_flags |= KMAP_FLAG_ASLR_STACK;
3248 
3249 #ifdef COMPAT_FREEBSD32
3250 	if (SV_CURPROC_FLAG(SV_ILP32)) {
3251 		struct kinfo_vm_layout32 kvm32;
3252 
3253 		memset(&kvm32, 0, sizeof(kvm32));
3254 		kvm32.kvm_min_user_addr = (uint32_t)kvm.kvm_min_user_addr;
3255 		kvm32.kvm_max_user_addr = (uint32_t)kvm.kvm_max_user_addr;
3256 		kvm32.kvm_text_addr = (uint32_t)kvm.kvm_text_addr;
3257 		kvm32.kvm_text_size = (uint32_t)kvm.kvm_text_size;
3258 		kvm32.kvm_data_addr = (uint32_t)kvm.kvm_data_addr;
3259 		kvm32.kvm_data_size = (uint32_t)kvm.kvm_data_size;
3260 		kvm32.kvm_stack_addr = (uint32_t)kvm.kvm_stack_addr;
3261 		kvm32.kvm_stack_size = (uint32_t)kvm.kvm_stack_size;
3262 		kvm32.kvm_map_flags = kvm.kvm_map_flags;
3263 		vmspace_free(vmspace);
3264 		error = SYSCTL_OUT(req, &kvm32, sizeof(kvm32));
3265 		goto out;
3266 	}
3267 #endif
3268 
3269 	error = SYSCTL_OUT(req, &kvm, sizeof(kvm));
3270 #ifdef COMPAT_FREEBSD32
3271 out:
3272 #endif
3273 	vmspace_free(vmspace);
3274 	return (error);
3275 }
3276 
3277 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,  0,
3278     "Process table");
3279 
3280 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
3281 	CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
3282 	"Return entire process table");
3283 
3284 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3285 	sysctl_kern_proc, "Process table");
3286 
3287 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
3288 	sysctl_kern_proc, "Process table");
3289 
3290 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3291 	sysctl_kern_proc, "Process table");
3292 
3293 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
3294 	CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3295 
3296 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
3297 	sysctl_kern_proc, "Process table");
3298 
3299 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3300 	sysctl_kern_proc, "Process table");
3301 
3302 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3303 	sysctl_kern_proc, "Process table");
3304 
3305 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
3306 	sysctl_kern_proc, "Process table");
3307 
3308 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
3309 	sysctl_kern_proc, "Return process table, no threads");
3310 
3311 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
3312 	CTLFLAG_RW | CTLFLAG_CAPWR | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
3313 	sysctl_kern_proc_args, "Process argument list");
3314 
3315 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
3316 	sysctl_kern_proc_env, "Process environment");
3317 
3318 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
3319 	CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
3320 
3321 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
3322 	CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
3323 
3324 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
3325 	CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
3326 	"Process syscall vector name (ABI type)");
3327 
3328 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
3329 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3330 
3331 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
3332 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3333 
3334 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
3335 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3336 
3337 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
3338 	sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3339 
3340 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
3341 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3342 
3343 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
3344 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3345 
3346 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
3347 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3348 
3349 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
3350 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
3351 
3352 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
3353 	CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
3354 	"Return process table, including threads");
3355 
3356 #ifdef COMPAT_FREEBSD7
3357 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
3358 	CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
3359 #endif
3360 
3361 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
3362 	CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
3363 
3364 #if defined(STACK) || defined(DDB)
3365 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
3366 	CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
3367 #endif
3368 
3369 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
3370 	CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
3371 
3372 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
3373 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
3374 	"Process resource limits");
3375 
3376 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
3377 	CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
3378 	"Process ps_strings location");
3379 
3380 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
3381 	CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
3382 
3383 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
3384 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
3385 	"Process binary osreldate");
3386 
3387 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD |
3388 	CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp,
3389 	"Process signal trampoline location");
3390 
3391 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGFASTBLK, sigfastblk, CTLFLAG_RD |
3392 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_sigfastblk,
3393 	"Thread sigfastblock address");
3394 
3395 static SYSCTL_NODE(_kern_proc, KERN_PROC_VM_LAYOUT, vm_layout, CTLFLAG_RD |
3396 	CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_vm_layout,
3397 	"Process virtual address space layout info");
3398 
3399 static struct sx stop_all_proc_blocker;
3400 SX_SYSINIT(stop_all_proc_blocker, &stop_all_proc_blocker, "sapblk");
3401 
3402 void
3403 stop_all_proc_block(void)
3404 {
3405 	sx_xlock(&stop_all_proc_blocker);
3406 }
3407 
3408 void
3409 stop_all_proc_unblock(void)
3410 {
3411 	sx_xunlock(&stop_all_proc_blocker);
3412 }
3413 
3414 int allproc_gen;
3415 
3416 /*
3417  * stop_all_proc() purpose is to stop all process which have usermode,
3418  * except current process for obvious reasons.  This makes it somewhat
3419  * unreliable when invoked from multithreaded process.  The service
3420  * must not be user-callable anyway.
3421  */
3422 void
3423 stop_all_proc(void)
3424 {
3425 	struct proc *cp, *p;
3426 	int r, gen;
3427 	bool restart, seen_stopped, seen_exiting, stopped_some;
3428 
3429 	stop_all_proc_block();
3430 
3431 	cp = curproc;
3432 allproc_loop:
3433 	sx_xlock(&allproc_lock);
3434 	gen = allproc_gen;
3435 	seen_exiting = seen_stopped = stopped_some = restart = false;
3436 	LIST_REMOVE(cp, p_list);
3437 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3438 	for (;;) {
3439 		p = LIST_NEXT(cp, p_list);
3440 		if (p == NULL)
3441 			break;
3442 		LIST_REMOVE(cp, p_list);
3443 		LIST_INSERT_AFTER(p, cp, p_list);
3444 		PROC_LOCK(p);
3445 		if ((p->p_flag & (P_KPROC | P_SYSTEM | P_TOTAL_STOP)) != 0) {
3446 			PROC_UNLOCK(p);
3447 			continue;
3448 		}
3449 		if ((p->p_flag & P_WEXIT) != 0) {
3450 			seen_exiting = true;
3451 			PROC_UNLOCK(p);
3452 			continue;
3453 		}
3454 		if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) {
3455 			/*
3456 			 * Stopped processes are tolerated when there
3457 			 * are no other processes which might continue
3458 			 * them.  P_STOPPED_SINGLE but not
3459 			 * P_TOTAL_STOP process still has at least one
3460 			 * thread running.
3461 			 */
3462 			seen_stopped = true;
3463 			PROC_UNLOCK(p);
3464 			continue;
3465 		}
3466 		sx_xunlock(&allproc_lock);
3467 		_PHOLD(p);
3468 		r = thread_single(p, SINGLE_ALLPROC);
3469 		if (r != 0)
3470 			restart = true;
3471 		else
3472 			stopped_some = true;
3473 		_PRELE(p);
3474 		PROC_UNLOCK(p);
3475 		sx_xlock(&allproc_lock);
3476 	}
3477 	/* Catch forked children we did not see in iteration. */
3478 	if (gen != allproc_gen)
3479 		restart = true;
3480 	sx_xunlock(&allproc_lock);
3481 	if (restart || stopped_some || seen_exiting || seen_stopped) {
3482 		kern_yield(PRI_USER);
3483 		goto allproc_loop;
3484 	}
3485 }
3486 
3487 void
3488 resume_all_proc(void)
3489 {
3490 	struct proc *cp, *p;
3491 
3492 	cp = curproc;
3493 	sx_xlock(&allproc_lock);
3494 again:
3495 	LIST_REMOVE(cp, p_list);
3496 	LIST_INSERT_HEAD(&allproc, cp, p_list);
3497 	for (;;) {
3498 		p = LIST_NEXT(cp, p_list);
3499 		if (p == NULL)
3500 			break;
3501 		LIST_REMOVE(cp, p_list);
3502 		LIST_INSERT_AFTER(p, cp, p_list);
3503 		PROC_LOCK(p);
3504 		if ((p->p_flag & P_TOTAL_STOP) != 0) {
3505 			sx_xunlock(&allproc_lock);
3506 			_PHOLD(p);
3507 			thread_single_end(p, SINGLE_ALLPROC);
3508 			_PRELE(p);
3509 			PROC_UNLOCK(p);
3510 			sx_xlock(&allproc_lock);
3511 		} else {
3512 			PROC_UNLOCK(p);
3513 		}
3514 	}
3515 	/*  Did the loop above missed any stopped process ? */
3516 	FOREACH_PROC_IN_SYSTEM(p) {
3517 		/* No need for proc lock. */
3518 		if ((p->p_flag & P_TOTAL_STOP) != 0)
3519 			goto again;
3520 	}
3521 	sx_xunlock(&allproc_lock);
3522 
3523 	stop_all_proc_unblock();
3524 }
3525 
3526 /* #define	TOTAL_STOP_DEBUG	1 */
3527 #ifdef TOTAL_STOP_DEBUG
3528 volatile static int ap_resume;
3529 #include <sys/mount.h>
3530 
3531 static int
3532 sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS)
3533 {
3534 	int error, val;
3535 
3536 	val = 0;
3537 	ap_resume = 0;
3538 	error = sysctl_handle_int(oidp, &val, 0, req);
3539 	if (error != 0 || req->newptr == NULL)
3540 		return (error);
3541 	if (val != 0) {
3542 		stop_all_proc();
3543 		syncer_suspend();
3544 		while (ap_resume == 0)
3545 			;
3546 		syncer_resume();
3547 		resume_all_proc();
3548 	}
3549 	return (0);
3550 }
3551 
3552 SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW |
3553     CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0,
3554     sysctl_debug_stop_all_proc, "I",
3555     "");
3556 #endif
3557