xref: /netbsd-src/sys/kern/kern_proc.c (revision 181254a7b1bdde6873432bffef2d2decc4b5c22f)
1 /*	$NetBSD: kern_proc.c,v 1.259 2020/08/28 22:27:51 riastradh Exp $	*/
2 
3 /*-
4  * Copyright (c) 1999, 2006, 2007, 2008, 2020 The NetBSD Foundation, Inc.
5  * All rights reserved.
6  *
7  * This code is derived from software contributed to The NetBSD Foundation
8  * by Jason R. Thorpe of the Numerical Aerospace Simulation Facility,
9  * NASA Ames Research Center, and by Andrew Doran.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
21  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
22  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
23  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
24  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
25  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
26  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
27  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
28  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
29  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
30  * POSSIBILITY OF SUCH DAMAGE.
31  */
32 
33 /*
34  * Copyright (c) 1982, 1986, 1989, 1991, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  *
37  * Redistribution and use in source and binary forms, with or without
38  * modification, are permitted provided that the following conditions
39  * are met:
40  * 1. Redistributions of source code must retain the above copyright
41  *    notice, this list of conditions and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 3. Neither the name of the University nor the names of its contributors
46  *    may be used to endorse or promote products derived from this software
47  *    without specific prior written permission.
48  *
49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59  * SUCH DAMAGE.
60  *
61  *	@(#)kern_proc.c	8.7 (Berkeley) 2/14/95
62  */
63 
64 #include <sys/cdefs.h>
65 __KERNEL_RCSID(0, "$NetBSD: kern_proc.c,v 1.259 2020/08/28 22:27:51 riastradh Exp $");
66 
67 #ifdef _KERNEL_OPT
68 #include "opt_kstack.h"
69 #include "opt_maxuprc.h"
70 #include "opt_dtrace.h"
71 #include "opt_compat_netbsd32.h"
72 #include "opt_kaslr.h"
73 #endif
74 
75 #if defined(__HAVE_COMPAT_NETBSD32) && !defined(COMPAT_NETBSD32) \
76     && !defined(_RUMPKERNEL)
77 #define COMPAT_NETBSD32
78 #endif
79 
80 #include <sys/param.h>
81 #include <sys/systm.h>
82 #include <sys/kernel.h>
83 #include <sys/proc.h>
84 #include <sys/resourcevar.h>
85 #include <sys/buf.h>
86 #include <sys/acct.h>
87 #include <sys/wait.h>
88 #include <sys/file.h>
89 #include <ufs/ufs/quota.h>
90 #include <sys/uio.h>
91 #include <sys/pool.h>
92 #include <sys/pset.h>
93 #include <sys/ioctl.h>
94 #include <sys/tty.h>
95 #include <sys/signalvar.h>
96 #include <sys/ras.h>
97 #include <sys/filedesc.h>
98 #include <sys/syscall_stats.h>
99 #include <sys/kauth.h>
100 #include <sys/sleepq.h>
101 #include <sys/atomic.h>
102 #include <sys/kmem.h>
103 #include <sys/namei.h>
104 #include <sys/dtrace_bsd.h>
105 #include <sys/sysctl.h>
106 #include <sys/exec.h>
107 #include <sys/cpu.h>
108 #include <sys/compat_stub.h>
109 #include <sys/futex.h>
110 #include <sys/pserialize.h>
111 
112 #include <uvm/uvm_extern.h>
113 #include <uvm/uvm.h>
114 
115 /*
116  * Process lists.
117  */
118 
119 struct proclist		allproc		__cacheline_aligned;
120 struct proclist		zombproc	__cacheline_aligned;
121 
122  kmutex_t		proc_lock	__cacheline_aligned;
123 static pserialize_t	proc_psz;
124 
125 /*
126  * pid to lwp/proc lookup is done by indexing the pid_table array.
127  * Since pid numbers are only allocated when an empty slot
128  * has been found, there is no need to search any lists ever.
129  * (an orphaned pgrp will lock the slot, a session will lock
130  * the pgrp with the same number.)
131  * If the table is too small it is reallocated with twice the
132  * previous size and the entries 'unzipped' into the two halves.
133  * A linked list of free entries is passed through the pt_lwp
134  * field of 'free' items - set odd to be an invalid ptr.  Two
135  * additional bits are also used to indicate if the slot is
136  * currently occupied by a proc or lwp, and if the PID is
137  * hidden from certain kinds of lookups.  We thus require a
138  * minimum alignment for proc and lwp structures (LWPs are
139  * at least 32-byte aligned).
140  */
141 
142 struct pid_table {
143 	uintptr_t	pt_slot;
144 	struct pgrp	*pt_pgrp;
145 	pid_t		pt_pid;
146 };
147 
148 #define	PT_F_FREE		((uintptr_t)__BIT(0))
149 #define	PT_F_LWP		0	/* pseudo-flag */
150 #define	PT_F_PROC		((uintptr_t)__BIT(1))
151 
152 #define	PT_F_TYPEBITS		(PT_F_FREE|PT_F_PROC)
153 #define	PT_F_ALLBITS		(PT_F_FREE|PT_F_PROC)
154 
155 #define	PT_VALID(s)		(((s) & PT_F_FREE) == 0)
156 #define	PT_RESERVED(s)		((s) == 0)
157 #define	PT_NEXT(s)		((u_int)(s) >> 1)
158 #define	PT_SET_FREE(pid)	(((pid) << 1) | PT_F_FREE)
159 #define	PT_SET_LWP(l)		((uintptr_t)(l))
160 #define	PT_SET_PROC(p)		(((uintptr_t)(p)) | PT_F_PROC)
161 #define	PT_SET_RESERVED		0
162 #define	PT_GET_LWP(s)		((struct lwp *)((s) & ~PT_F_ALLBITS))
163 #define	PT_GET_PROC(s)		((struct proc *)((s) & ~PT_F_ALLBITS))
164 #define	PT_GET_TYPE(s)		((s) & PT_F_TYPEBITS)
165 #define	PT_IS_LWP(s)		(PT_GET_TYPE(s) == PT_F_LWP && (s) != 0)
166 #define	PT_IS_PROC(s)		(PT_GET_TYPE(s) == PT_F_PROC)
167 
168 #define	MIN_PROC_ALIGNMENT	(PT_F_ALLBITS + 1)
169 
170 /*
171  * Table of process IDs (PIDs).
172  */
173 static struct pid_table *pid_table	__read_mostly;
174 
175 #define	INITIAL_PID_TABLE_SIZE		(1 << 5)
176 
177 /* Table mask, threshold for growing and number of allocated PIDs. */
178 static u_int		pid_tbl_mask	__read_mostly;
179 static u_int		pid_alloc_lim	__read_mostly;
180 static u_int		pid_alloc_cnt	__cacheline_aligned;
181 
182 /* Next free, last free and maximum PIDs. */
183 static u_int		next_free_pt	__cacheline_aligned;
184 static u_int		last_free_pt	__cacheline_aligned;
185 static pid_t		pid_max		__read_mostly;
186 
187 /* Components of the first process -- never freed. */
188 
189 extern struct emul emul_netbsd;	/* defined in kern_exec.c */
190 
191 struct session session0 = {
192 	.s_count = 1,
193 	.s_sid = 0,
194 };
195 struct pgrp pgrp0 = {
196 	.pg_members = LIST_HEAD_INITIALIZER(&pgrp0.pg_members),
197 	.pg_session = &session0,
198 };
199 filedesc_t filedesc0;
200 struct cwdinfo cwdi0 = {
201 	.cwdi_cmask = CMASK,
202 	.cwdi_refcnt = 1,
203 };
204 struct plimit limit0;
205 struct pstats pstat0;
206 struct vmspace vmspace0;
207 struct sigacts sigacts0;
208 struct proc proc0 = {
209 	.p_lwps = LIST_HEAD_INITIALIZER(&proc0.p_lwps),
210 	.p_sigwaiters = LIST_HEAD_INITIALIZER(&proc0.p_sigwaiters),
211 	.p_nlwps = 1,
212 	.p_nrlwps = 1,
213 	.p_pgrp = &pgrp0,
214 	.p_comm = "system",
215 	/*
216 	 * Set P_NOCLDWAIT so that kernel threads are reparented to init(8)
217 	 * when they exit.  init(8) can easily wait them out for us.
218 	 */
219 	.p_flag = PK_SYSTEM | PK_NOCLDWAIT,
220 	.p_stat = SACTIVE,
221 	.p_nice = NZERO,
222 	.p_emul = &emul_netbsd,
223 	.p_cwdi = &cwdi0,
224 	.p_limit = &limit0,
225 	.p_fd = &filedesc0,
226 	.p_vmspace = &vmspace0,
227 	.p_stats = &pstat0,
228 	.p_sigacts = &sigacts0,
229 #ifdef PROC0_MD_INITIALIZERS
230 	PROC0_MD_INITIALIZERS
231 #endif
232 };
233 kauth_cred_t cred0;
234 
235 static const int	nofile	= NOFILE;
236 static const int	maxuprc	= MAXUPRC;
237 
238 static int sysctl_doeproc(SYSCTLFN_PROTO);
239 static int sysctl_kern_proc_args(SYSCTLFN_PROTO);
240 static int sysctl_security_expose_address(SYSCTLFN_PROTO);
241 
242 #ifdef KASLR
243 static int kern_expose_address = 0;
244 #else
245 static int kern_expose_address = 1;
246 #endif
247 /*
248  * The process list descriptors, used during pid allocation and
249  * by sysctl.  No locking on this data structure is needed since
250  * it is completely static.
251  */
252 const struct proclist_desc proclists[] = {
253 	{ &allproc	},
254 	{ &zombproc	},
255 	{ NULL		},
256 };
257 
258 static struct pgrp *	pg_remove(pid_t);
259 static void		pg_delete(pid_t);
260 static void		orphanpg(struct pgrp *);
261 
262 static specificdata_domain_t proc_specificdata_domain;
263 
264 static pool_cache_t proc_cache;
265 
266 static kauth_listener_t proc_listener;
267 
268 static void fill_proc(const struct proc *, struct proc *, bool);
269 static int fill_pathname(struct lwp *, pid_t, void *, size_t *);
270 static int fill_cwd(struct lwp *, pid_t, void *, size_t *);
271 
272 static int
273 proc_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
274     void *arg0, void *arg1, void *arg2, void *arg3)
275 {
276 	struct proc *p;
277 	int result;
278 
279 	result = KAUTH_RESULT_DEFER;
280 	p = arg0;
281 
282 	switch (action) {
283 	case KAUTH_PROCESS_CANSEE: {
284 		enum kauth_process_req req;
285 
286 		req = (enum kauth_process_req)(uintptr_t)arg1;
287 
288 		switch (req) {
289 		case KAUTH_REQ_PROCESS_CANSEE_ARGS:
290 		case KAUTH_REQ_PROCESS_CANSEE_ENTRY:
291 		case KAUTH_REQ_PROCESS_CANSEE_OPENFILES:
292 		case KAUTH_REQ_PROCESS_CANSEE_EPROC:
293 			result = KAUTH_RESULT_ALLOW;
294 			break;
295 
296 		case KAUTH_REQ_PROCESS_CANSEE_ENV:
297 			if (kauth_cred_getuid(cred) !=
298 			    kauth_cred_getuid(p->p_cred) ||
299 			    kauth_cred_getuid(cred) !=
300 			    kauth_cred_getsvuid(p->p_cred))
301 				break;
302 
303 			result = KAUTH_RESULT_ALLOW;
304 
305 			break;
306 
307 		case KAUTH_REQ_PROCESS_CANSEE_KPTR:
308 			if (!kern_expose_address)
309 				break;
310 
311 			if (kern_expose_address == 1 && !(p->p_flag & PK_KMEM))
312 				break;
313 
314 			result = KAUTH_RESULT_ALLOW;
315 
316 			break;
317 
318 		default:
319 			break;
320 		}
321 
322 		break;
323 		}
324 
325 	case KAUTH_PROCESS_FORK: {
326 		int lnprocs = (int)(unsigned long)arg2;
327 
328 		/*
329 		 * Don't allow a nonprivileged user to use the last few
330 		 * processes. The variable lnprocs is the current number of
331 		 * processes, maxproc is the limit.
332 		 */
333 		if (__predict_false((lnprocs >= maxproc - 5)))
334 			break;
335 
336 		result = KAUTH_RESULT_ALLOW;
337 
338 		break;
339 		}
340 
341 	case KAUTH_PROCESS_CORENAME:
342 	case KAUTH_PROCESS_STOPFLAG:
343 		if (proc_uidmatch(cred, p->p_cred) == 0)
344 			result = KAUTH_RESULT_ALLOW;
345 
346 		break;
347 
348 	default:
349 		break;
350 	}
351 
352 	return result;
353 }
354 
355 static int
356 proc_ctor(void *arg __unused, void *obj, int flags __unused)
357 {
358 	memset(obj, 0, sizeof(struct proc));
359 	return 0;
360 }
361 
362 static pid_t proc_alloc_pid_slot(struct proc *, uintptr_t);
363 
364 /*
365  * Initialize global process hashing structures.
366  */
367 void
368 procinit(void)
369 {
370 	const struct proclist_desc *pd;
371 	u_int i;
372 #define	LINK_EMPTY ((PID_MAX + INITIAL_PID_TABLE_SIZE) & ~(INITIAL_PID_TABLE_SIZE - 1))
373 
374 	for (pd = proclists; pd->pd_list != NULL; pd++)
375 		LIST_INIT(pd->pd_list);
376 
377 	mutex_init(&proc_lock, MUTEX_DEFAULT, IPL_NONE);
378 
379 	proc_psz = pserialize_create();
380 
381 	pid_table = kmem_alloc(INITIAL_PID_TABLE_SIZE
382 	    * sizeof(struct pid_table), KM_SLEEP);
383 	pid_tbl_mask = INITIAL_PID_TABLE_SIZE - 1;
384 	pid_max = PID_MAX;
385 
386 	/* Set free list running through table...
387 	   Preset 'use count' above PID_MAX so we allocate pid 1 next. */
388 	for (i = 0; i <= pid_tbl_mask; i++) {
389 		pid_table[i].pt_slot = PT_SET_FREE(LINK_EMPTY + i + 1);
390 		pid_table[i].pt_pgrp = 0;
391 		pid_table[i].pt_pid = 0;
392 	}
393 	/* slot 0 is just grabbed */
394 	next_free_pt = 1;
395 	/* Need to fix last entry. */
396 	last_free_pt = pid_tbl_mask;
397 	pid_table[last_free_pt].pt_slot = PT_SET_FREE(LINK_EMPTY);
398 	/* point at which we grow table - to avoid reusing pids too often */
399 	pid_alloc_lim = pid_tbl_mask - 1;
400 #undef LINK_EMPTY
401 
402 	/* Reserve PID 1 for init(8). */	/* XXX slightly gross */
403 	mutex_enter(&proc_lock);
404 	if (proc_alloc_pid_slot(&proc0, PT_SET_RESERVED) != 1)
405 		panic("failed to reserve PID 1 for init(8)");
406 	mutex_exit(&proc_lock);
407 
408 	proc_specificdata_domain = specificdata_domain_create();
409 	KASSERT(proc_specificdata_domain != NULL);
410 
411 	size_t proc_alignment = coherency_unit;
412 	if (proc_alignment < MIN_PROC_ALIGNMENT)
413 		proc_alignment = MIN_PROC_ALIGNMENT;
414 
415 	proc_cache = pool_cache_init(sizeof(struct proc), proc_alignment, 0, 0,
416 	    "procpl", NULL, IPL_NONE, proc_ctor, NULL, NULL);
417 
418 	proc_listener = kauth_listen_scope(KAUTH_SCOPE_PROCESS,
419 	    proc_listener_cb, NULL);
420 }
421 
422 void
423 procinit_sysctl(void)
424 {
425 	static struct sysctllog *clog;
426 
427 	sysctl_createv(&clog, 0, NULL, NULL,
428 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
429 		       CTLTYPE_INT, "expose_address",
430 		       SYSCTL_DESCR("Enable exposing kernel addresses"),
431 		       sysctl_security_expose_address, 0,
432 		       &kern_expose_address, 0, CTL_KERN, CTL_CREATE, CTL_EOL);
433 	sysctl_createv(&clog, 0, NULL, NULL,
434 		       CTLFLAG_PERMANENT,
435 		       CTLTYPE_NODE, "proc",
436 		       SYSCTL_DESCR("System-wide process information"),
437 		       sysctl_doeproc, 0, NULL, 0,
438 		       CTL_KERN, KERN_PROC, CTL_EOL);
439 	sysctl_createv(&clog, 0, NULL, NULL,
440 		       CTLFLAG_PERMANENT,
441 		       CTLTYPE_NODE, "proc2",
442 		       SYSCTL_DESCR("Machine-independent process information"),
443 		       sysctl_doeproc, 0, NULL, 0,
444 		       CTL_KERN, KERN_PROC2, CTL_EOL);
445 	sysctl_createv(&clog, 0, NULL, NULL,
446 		       CTLFLAG_PERMANENT,
447 		       CTLTYPE_NODE, "proc_args",
448 		       SYSCTL_DESCR("Process argument information"),
449 		       sysctl_kern_proc_args, 0, NULL, 0,
450 		       CTL_KERN, KERN_PROC_ARGS, CTL_EOL);
451 
452 	/*
453 	  "nodes" under these:
454 
455 	  KERN_PROC_ALL
456 	  KERN_PROC_PID pid
457 	  KERN_PROC_PGRP pgrp
458 	  KERN_PROC_SESSION sess
459 	  KERN_PROC_TTY tty
460 	  KERN_PROC_UID uid
461 	  KERN_PROC_RUID uid
462 	  KERN_PROC_GID gid
463 	  KERN_PROC_RGID gid
464 
465 	  all in all, probably not worth the effort...
466 	*/
467 }
468 
469 /*
470  * Initialize process 0.
471  */
472 void
473 proc0_init(void)
474 {
475 	struct proc *p;
476 	struct pgrp *pg;
477 	struct rlimit *rlim;
478 	rlim_t lim;
479 	int i;
480 
481 	p = &proc0;
482 	pg = &pgrp0;
483 
484 	mutex_init(&p->p_stmutex, MUTEX_DEFAULT, IPL_HIGH);
485 	mutex_init(&p->p_auxlock, MUTEX_DEFAULT, IPL_NONE);
486 	p->p_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
487 
488 	rw_init(&p->p_reflock);
489 	cv_init(&p->p_waitcv, "wait");
490 	cv_init(&p->p_lwpcv, "lwpwait");
491 
492 	LIST_INSERT_HEAD(&p->p_lwps, &lwp0, l_sibling);
493 
494 	KASSERT(lwp0.l_lid == 0);
495 	pid_table[lwp0.l_lid].pt_slot = PT_SET_LWP(&lwp0);
496 	LIST_INSERT_HEAD(&allproc, p, p_list);
497 
498 	pid_table[lwp0.l_lid].pt_pgrp = pg;
499 	LIST_INSERT_HEAD(&pg->pg_members, p, p_pglist);
500 
501 #ifdef __HAVE_SYSCALL_INTERN
502 	(*p->p_emul->e_syscall_intern)(p);
503 #endif
504 
505 	/* Create credentials. */
506 	cred0 = kauth_cred_alloc();
507 	p->p_cred = cred0;
508 
509 	/* Create the CWD info. */
510 	rw_init(&cwdi0.cwdi_lock);
511 
512 	/* Create the limits structures. */
513 	mutex_init(&limit0.pl_lock, MUTEX_DEFAULT, IPL_NONE);
514 
515 	rlim = limit0.pl_rlimit;
516 	for (i = 0; i < __arraycount(limit0.pl_rlimit); i++) {
517 		rlim[i].rlim_cur = RLIM_INFINITY;
518 		rlim[i].rlim_max = RLIM_INFINITY;
519 	}
520 
521 	rlim[RLIMIT_NOFILE].rlim_max = maxfiles;
522 	rlim[RLIMIT_NOFILE].rlim_cur = maxfiles < nofile ? maxfiles : nofile;
523 
524 	rlim[RLIMIT_NPROC].rlim_max = maxproc;
525 	rlim[RLIMIT_NPROC].rlim_cur = maxproc < maxuprc ? maxproc : maxuprc;
526 
527 	lim = MIN(VM_MAXUSER_ADDRESS, ctob((rlim_t)uvm_availmem(false)));
528 	rlim[RLIMIT_RSS].rlim_max = lim;
529 	rlim[RLIMIT_MEMLOCK].rlim_max = lim;
530 	rlim[RLIMIT_MEMLOCK].rlim_cur = lim / 3;
531 
532 	rlim[RLIMIT_NTHR].rlim_max = maxlwp;
533 	rlim[RLIMIT_NTHR].rlim_cur = maxlwp < maxuprc ? maxlwp : maxuprc;
534 
535 	/* Note that default core name has zero length. */
536 	limit0.pl_corename = defcorename;
537 	limit0.pl_cnlen = 0;
538 	limit0.pl_refcnt = 1;
539 	limit0.pl_writeable = false;
540 	limit0.pl_sv_limit = NULL;
541 
542 	/* Configure virtual memory system, set vm rlimits. */
543 	uvm_init_limits(p);
544 
545 	/* Initialize file descriptor table for proc0. */
546 	fd_init(&filedesc0);
547 
548 	/*
549 	 * Initialize proc0's vmspace, which uses the kernel pmap.
550 	 * All kernel processes (which never have user space mappings)
551 	 * share proc0's vmspace, and thus, the kernel pmap.
552 	 */
553 	uvmspace_init(&vmspace0, pmap_kernel(), round_page(VM_MIN_ADDRESS),
554 	    trunc_page(VM_MAXUSER_ADDRESS),
555 #ifdef __USE_TOPDOWN_VM
556 	    true
557 #else
558 	    false
559 #endif
560 	    );
561 
562 	/* Initialize signal state for proc0. XXX IPL_SCHED */
563 	mutex_init(&p->p_sigacts->sa_mutex, MUTEX_DEFAULT, IPL_SCHED);
564 	siginit(p);
565 
566 	proc_initspecific(p);
567 	kdtrace_proc_ctor(NULL, p);
568 }
569 
570 /*
571  * Session reference counting.
572  */
573 
574 void
575 proc_sesshold(struct session *ss)
576 {
577 
578 	KASSERT(mutex_owned(&proc_lock));
579 	ss->s_count++;
580 }
581 
582 void
583 proc_sessrele(struct session *ss)
584 {
585 	struct pgrp *pg;
586 
587 	KASSERT(mutex_owned(&proc_lock));
588 	KASSERT(ss->s_count > 0);
589 
590 	/*
591 	 * We keep the pgrp with the same id as the session in order to
592 	 * stop a process being given the same pid.  Since the pgrp holds
593 	 * a reference to the session, it must be a 'zombie' pgrp by now.
594 	 */
595 	if (--ss->s_count == 0) {
596 		pg = pg_remove(ss->s_sid);
597 	} else {
598 		pg = NULL;
599 		ss = NULL;
600 	}
601 
602 	mutex_exit(&proc_lock);
603 
604 	if (pg)
605 		kmem_free(pg, sizeof(struct pgrp));
606 	if (ss)
607 		kmem_free(ss, sizeof(struct session));
608 }
609 
610 /*
611  * Check that the specified process group is in the session of the
612  * specified process.
613  * Treats -ve ids as process ids.
614  * Used to validate TIOCSPGRP requests.
615  */
616 int
617 pgid_in_session(struct proc *p, pid_t pg_id)
618 {
619 	struct pgrp *pgrp;
620 	struct session *session;
621 	int error;
622 
623 	mutex_enter(&proc_lock);
624 	if (pg_id < 0) {
625 		struct proc *p1 = proc_find(-pg_id);
626 		if (p1 == NULL) {
627 			error = EINVAL;
628 			goto fail;
629 		}
630 		pgrp = p1->p_pgrp;
631 	} else {
632 		pgrp = pgrp_find(pg_id);
633 		if (pgrp == NULL) {
634 			error = EINVAL;
635 			goto fail;
636 		}
637 	}
638 	session = pgrp->pg_session;
639 	error = (session != p->p_pgrp->pg_session) ? EPERM : 0;
640 fail:
641 	mutex_exit(&proc_lock);
642 	return error;
643 }
644 
645 /*
646  * p_inferior: is p an inferior of q?
647  */
648 static inline bool
649 p_inferior(struct proc *p, struct proc *q)
650 {
651 
652 	KASSERT(mutex_owned(&proc_lock));
653 
654 	for (; p != q; p = p->p_pptr)
655 		if (p->p_pid == 0)
656 			return false;
657 	return true;
658 }
659 
660 /*
661  * proc_find_lwp: locate an lwp in said proc by the ID.
662  *
663  * => Must be called with p::p_lock held.
664  * => LSIDL lwps are not returned because they are only partially
665  *    constructed while occupying the slot.
666  * => Callers need to be careful about lwp::l_stat of the returned
667  *    lwp.
668  */
669 struct lwp *
670 proc_find_lwp(proc_t *p, pid_t pid)
671 {
672 	struct pid_table *pt;
673 	struct lwp *l = NULL;
674 	uintptr_t slot;
675 	int s;
676 
677 	KASSERT(mutex_owned(p->p_lock));
678 
679 	/*
680 	 * Look in the pid_table.  This is done unlocked inside a pserialize
681 	 * read section covering pid_table's memory allocation only, so take
682 	 * care to read the slot atomically and only once.  This issues a
683 	 * memory barrier for dependent loads on alpha.
684 	 */
685 	s = pserialize_read_enter();
686 	pt = &pid_table[pid & pid_tbl_mask];
687 	slot = atomic_load_consume(&pt->pt_slot);
688 	if (__predict_false(!PT_IS_LWP(slot))) {
689 		pserialize_read_exit(s);
690 		return NULL;
691 	}
692 
693 	/*
694 	 * Check to see if the LWP is from the correct process.  We won't
695 	 * see entries in pid_table from a prior process that also used "p",
696 	 * by virtue of the fact that allocating "p" means all prior updates
697 	 * to dependant data structures are visible to this thread.
698 	 */
699 	l = PT_GET_LWP(slot);
700 	if (__predict_false(atomic_load_relaxed(&l->l_proc) != p)) {
701 		pserialize_read_exit(s);
702 		return NULL;
703 	}
704 
705 	/*
706 	 * We now know that p->p_lock holds this LWP stable.
707 	 *
708 	 * If the status is not LSIDL, it means the LWP is intended to be
709 	 * findable by LID and l_lid cannot change behind us.
710 	 *
711 	 * No need to acquire the LWP's lock to check for LSIDL, as
712 	 * p->p_lock must be held to transition in and out of LSIDL.
713 	 * Any other observed state of is no particular interest.
714 	 */
715 	pserialize_read_exit(s);
716 	return l->l_stat != LSIDL && l->l_lid == pid ? l : NULL;
717 }
718 
719 /*
720  * proc_find_lwp_unlocked: locate an lwp in said proc by the ID.
721  *
722  * => Called in a pserialize read section with no locks held.
723  * => LSIDL lwps are not returned because they are only partially
724  *    constructed while occupying the slot.
725  * => Callers need to be careful about lwp::l_stat of the returned
726  *    lwp.
727  * => If an LWP is found, it's returned locked.
728  */
729 struct lwp *
730 proc_find_lwp_unlocked(proc_t *p, pid_t pid)
731 {
732 	struct pid_table *pt;
733 	struct lwp *l = NULL;
734 	uintptr_t slot;
735 
736 	KASSERT(pserialize_in_read_section());
737 
738 	/*
739 	 * Look in the pid_table.  This is done unlocked inside a pserialize
740 	 * read section covering pid_table's memory allocation only, so take
741 	 * care to read the slot atomically and only once.  This issues a
742 	 * memory barrier for dependent loads on alpha.
743 	 */
744 	pt = &pid_table[pid & pid_tbl_mask];
745 	slot = atomic_load_consume(&pt->pt_slot);
746 	if (__predict_false(!PT_IS_LWP(slot))) {
747 		return NULL;
748 	}
749 
750 	/*
751 	 * Lock the LWP we found to get it stable.  If it's embryonic or
752 	 * reaped (LSIDL) then none of the other fields can safely be
753 	 * checked.
754 	 */
755 	l = PT_GET_LWP(slot);
756 	lwp_lock(l);
757 	if (__predict_false(l->l_stat == LSIDL)) {
758 		lwp_unlock(l);
759 		return NULL;
760 	}
761 
762 	/*
763 	 * l_proc and l_lid are now known stable because the LWP is not
764 	 * LSIDL, so check those fields too to make sure we found the
765 	 * right thing.
766 	 */
767 	if (__predict_false(l->l_proc != p || l->l_lid != pid)) {
768 		lwp_unlock(l);
769 		return NULL;
770 	}
771 
772 	/* Everything checks out, return it locked. */
773 	return l;
774 }
775 
776 /*
777  * proc_find_lwp_acquire_proc: locate an lwp and acquire a lock
778  * on its containing proc.
779  *
780  * => Similar to proc_find_lwp(), but does not require you to have
781  *    the proc a priori.
782  * => Also returns proc * to caller, with p::p_lock held.
783  * => Same caveats apply.
784  */
785 struct lwp *
786 proc_find_lwp_acquire_proc(pid_t pid, struct proc **pp)
787 {
788 	struct pid_table *pt;
789 	struct proc *p = NULL;
790 	struct lwp *l = NULL;
791 	uintptr_t slot;
792 
793 	KASSERT(pp != NULL);
794 	mutex_enter(&proc_lock);
795 	pt = &pid_table[pid & pid_tbl_mask];
796 
797 	slot = pt->pt_slot;
798 	if (__predict_true(PT_IS_LWP(slot) && pt->pt_pid == pid)) {
799 		l = PT_GET_LWP(slot);
800 		p = l->l_proc;
801 		mutex_enter(p->p_lock);
802 		if (__predict_false(l->l_stat == LSIDL)) {
803 			mutex_exit(p->p_lock);
804 			l = NULL;
805 			p = NULL;
806 		}
807 	}
808 	mutex_exit(&proc_lock);
809 
810 	KASSERT(p == NULL || mutex_owned(p->p_lock));
811 	*pp = p;
812 	return l;
813 }
814 
815 /*
816  * proc_find_raw_pid_table_locked: locate a process by the ID.
817  *
818  * => Must be called with proc_lock held.
819  */
820 static proc_t *
821 proc_find_raw_pid_table_locked(pid_t pid, bool any_lwpid)
822 {
823 	struct pid_table *pt;
824 	proc_t *p = NULL;
825 	uintptr_t slot;
826 
827 	/* No - used by DDB.  KASSERT(mutex_owned(&proc_lock)); */
828 	pt = &pid_table[pid & pid_tbl_mask];
829 
830 	slot = pt->pt_slot;
831 	if (__predict_true(PT_IS_LWP(slot) && pt->pt_pid == pid)) {
832 		/*
833 		 * When looking up processes, require a direct match
834 		 * on the PID assigned to the proc, not just one of
835 		 * its LWPs.
836 		 *
837 		 * N.B. We require lwp::l_proc of LSIDL LWPs to be
838 		 * valid here.
839 		 */
840 		p = PT_GET_LWP(slot)->l_proc;
841 		if (__predict_false(p->p_pid != pid && !any_lwpid))
842 			p = NULL;
843 	} else if (PT_IS_PROC(slot) && pt->pt_pid == pid) {
844 		p = PT_GET_PROC(slot);
845 	}
846 	return p;
847 }
848 
849 proc_t *
850 proc_find_raw(pid_t pid)
851 {
852 
853 	return proc_find_raw_pid_table_locked(pid, false);
854 }
855 
856 static proc_t *
857 proc_find_internal(pid_t pid, bool any_lwpid)
858 {
859 	proc_t *p;
860 
861 	KASSERT(mutex_owned(&proc_lock));
862 
863 	p = proc_find_raw_pid_table_locked(pid, any_lwpid);
864 	if (__predict_false(p == NULL)) {
865 		return NULL;
866 	}
867 
868 	/*
869 	 * Only allow live processes to be found by PID.
870 	 * XXX: p_stat might change, since proc unlocked.
871 	 */
872 	if (__predict_true(p->p_stat == SACTIVE || p->p_stat == SSTOP)) {
873 		return p;
874 	}
875 	return NULL;
876 }
877 
878 proc_t *
879 proc_find(pid_t pid)
880 {
881 	return proc_find_internal(pid, false);
882 }
883 
884 proc_t *
885 proc_find_lwpid(pid_t pid)
886 {
887 	return proc_find_internal(pid, true);
888 }
889 
890 /*
891  * pgrp_find: locate a process group by the ID.
892  *
893  * => Must be called with proc_lock held.
894  */
895 struct pgrp *
896 pgrp_find(pid_t pgid)
897 {
898 	struct pgrp *pg;
899 
900 	KASSERT(mutex_owned(&proc_lock));
901 
902 	pg = pid_table[pgid & pid_tbl_mask].pt_pgrp;
903 
904 	/*
905 	 * Cannot look up a process group that only exists because the
906 	 * session has not died yet (traditional).
907 	 */
908 	if (pg == NULL || pg->pg_id != pgid || LIST_EMPTY(&pg->pg_members)) {
909 		return NULL;
910 	}
911 	return pg;
912 }
913 
914 static void
915 expand_pid_table(void)
916 {
917 	size_t pt_size, tsz;
918 	struct pid_table *n_pt, *new_pt;
919 	uintptr_t slot;
920 	struct pgrp *pgrp;
921 	pid_t pid, rpid;
922 	u_int i;
923 	uint new_pt_mask;
924 
925 	KASSERT(mutex_owned(&proc_lock));
926 
927 	/* Unlock the pid_table briefly to allocate memory. */
928 	pt_size = pid_tbl_mask + 1;
929 	mutex_exit(&proc_lock);
930 
931 	tsz = pt_size * 2 * sizeof(struct pid_table);
932 	new_pt = kmem_alloc(tsz, KM_SLEEP);
933 	new_pt_mask = pt_size * 2 - 1;
934 
935 	/* XXX For now.  The pratical limit is much lower anyway. */
936 	KASSERT(new_pt_mask <= FUTEX_TID_MASK);
937 
938 	mutex_enter(&proc_lock);
939 	if (pt_size != pid_tbl_mask + 1) {
940 		/* Another process beat us to it... */
941 		mutex_exit(&proc_lock);
942 		kmem_free(new_pt, tsz);
943 		goto out;
944 	}
945 
946 	/*
947 	 * Copy entries from old table into new one.
948 	 * If 'pid' is 'odd' we need to place in the upper half,
949 	 * even pid's to the lower half.
950 	 * Free items stay in the low half so we don't have to
951 	 * fixup the reference to them.
952 	 * We stuff free items on the front of the freelist
953 	 * because we can't write to unmodified entries.
954 	 * Processing the table backwards maintains a semblance
955 	 * of issuing pid numbers that increase with time.
956 	 */
957 	i = pt_size - 1;
958 	n_pt = new_pt + i;
959 	for (; ; i--, n_pt--) {
960 		slot = pid_table[i].pt_slot;
961 		pgrp = pid_table[i].pt_pgrp;
962 		if (!PT_VALID(slot)) {
963 			/* Up 'use count' so that link is valid */
964 			pid = (PT_NEXT(slot) + pt_size) & ~pt_size;
965 			rpid = 0;
966 			slot = PT_SET_FREE(pid);
967 			if (pgrp)
968 				pid = pgrp->pg_id;
969 		} else {
970 			pid = pid_table[i].pt_pid;
971 			rpid = pid;
972 		}
973 
974 		/* Save entry in appropriate half of table */
975 		n_pt[pid & pt_size].pt_slot = slot;
976 		n_pt[pid & pt_size].pt_pgrp = pgrp;
977 		n_pt[pid & pt_size].pt_pid = rpid;
978 
979 		/* Put other piece on start of free list */
980 		pid = (pid ^ pt_size) & ~pid_tbl_mask;
981 		n_pt[pid & pt_size].pt_slot =
982 			PT_SET_FREE((pid & ~pt_size) | next_free_pt);
983 		n_pt[pid & pt_size].pt_pgrp = 0;
984 		n_pt[pid & pt_size].pt_pid = 0;
985 
986 		next_free_pt = i | (pid & pt_size);
987 		if (i == 0)
988 			break;
989 	}
990 
991 	/* Save old table size and switch tables */
992 	tsz = pt_size * sizeof(struct pid_table);
993 	n_pt = pid_table;
994 	pid_table = new_pt;
995 	pid_tbl_mask = new_pt_mask;
996 
997 	/*
998 	 * pid_max starts as PID_MAX (= 30000), once we have 16384
999 	 * allocated pids we need it to be larger!
1000 	 */
1001 	if (pid_tbl_mask > PID_MAX) {
1002 		pid_max = pid_tbl_mask * 2 + 1;
1003 		pid_alloc_lim |= pid_alloc_lim << 1;
1004 	} else
1005 		pid_alloc_lim <<= 1;	/* doubles number of free slots... */
1006 
1007 	mutex_exit(&proc_lock);
1008 
1009 	/*
1010 	 * Make sure that unlocked access to the old pid_table is complete
1011 	 * and then free it.
1012 	 */
1013 	pserialize_perform(proc_psz);
1014 	kmem_free(n_pt, tsz);
1015 
1016  out:	/* Return with proc_lock held again. */
1017 	mutex_enter(&proc_lock);
1018 }
1019 
1020 struct proc *
1021 proc_alloc(void)
1022 {
1023 	struct proc *p;
1024 
1025 	p = pool_cache_get(proc_cache, PR_WAITOK);
1026 	p->p_stat = SIDL;			/* protect against others */
1027 	proc_initspecific(p);
1028 	kdtrace_proc_ctor(NULL, p);
1029 
1030 	/*
1031 	 * Allocate a placeholder in the pid_table.  When we create the
1032 	 * first LWP for this process, it will take ownership of the
1033 	 * slot.
1034 	 */
1035 	if (__predict_false(proc_alloc_pid(p) == -1)) {
1036 		/* Allocating the PID failed; unwind. */
1037 		proc_finispecific(p);
1038 		proc_free_mem(p);
1039 		p = NULL;
1040 	}
1041 	return p;
1042 }
1043 
1044 /*
1045  * proc_alloc_pid_slot: allocate PID and record the occcupant so that
1046  * proc_find_raw() can find it by the PID.
1047  */
1048 static pid_t __noinline
1049 proc_alloc_pid_slot(struct proc *p, uintptr_t slot)
1050 {
1051 	struct pid_table *pt;
1052 	pid_t pid;
1053 	int nxt;
1054 
1055 	KASSERT(mutex_owned(&proc_lock));
1056 
1057 	for (;;expand_pid_table()) {
1058 		if (__predict_false(pid_alloc_cnt >= pid_alloc_lim)) {
1059 			/* ensure pids cycle through 2000+ values */
1060 			continue;
1061 		}
1062 		/*
1063 		 * The first user process *must* be given PID 1.
1064 		 * it has already been reserved for us.  This
1065 		 * will be coming in from the proc_alloc() call
1066 		 * above, and the entry will be usurped later when
1067 		 * the first user LWP is created.
1068 		 * XXX this is slightly gross.
1069 		 */
1070 		if (__predict_false(PT_RESERVED(pid_table[1].pt_slot) &&
1071 				    p != &proc0)) {
1072 			KASSERT(PT_IS_PROC(slot));
1073 			pt = &pid_table[1];
1074 			pt->pt_slot = slot;
1075 			return 1;
1076 		}
1077 		pt = &pid_table[next_free_pt];
1078 #ifdef DIAGNOSTIC
1079 		if (__predict_false(PT_VALID(pt->pt_slot) || pt->pt_pgrp))
1080 			panic("proc_alloc: slot busy");
1081 #endif
1082 		nxt = PT_NEXT(pt->pt_slot);
1083 		if (nxt & pid_tbl_mask)
1084 			break;
1085 		/* Table full - expand (NB last entry not used....) */
1086 	}
1087 
1088 	/* pid is 'saved use count' + 'size' + entry */
1089 	pid = (nxt & ~pid_tbl_mask) + pid_tbl_mask + 1 + next_free_pt;
1090 	if ((uint)pid > (uint)pid_max)
1091 		pid &= pid_tbl_mask;
1092 	next_free_pt = nxt & pid_tbl_mask;
1093 
1094 	/* XXX For now.  The pratical limit is much lower anyway. */
1095 	KASSERT(pid <= FUTEX_TID_MASK);
1096 
1097 	/* Grab table slot */
1098 	pt->pt_slot = slot;
1099 
1100 	KASSERT(pt->pt_pid == 0);
1101 	pt->pt_pid = pid;
1102 	pid_alloc_cnt++;
1103 
1104 	return pid;
1105 }
1106 
1107 pid_t
1108 proc_alloc_pid(struct proc *p)
1109 {
1110 	pid_t pid;
1111 
1112 	KASSERT((((uintptr_t)p) & PT_F_ALLBITS) == 0);
1113 	KASSERT(p->p_stat == SIDL);
1114 
1115 	mutex_enter(&proc_lock);
1116 	pid = proc_alloc_pid_slot(p, PT_SET_PROC(p));
1117 	if (pid != -1)
1118 		p->p_pid = pid;
1119 	mutex_exit(&proc_lock);
1120 
1121 	return pid;
1122 }
1123 
1124 pid_t
1125 proc_alloc_lwpid(struct proc *p, struct lwp *l)
1126 {
1127 	struct pid_table *pt;
1128 	pid_t pid;
1129 
1130 	KASSERT((((uintptr_t)l) & PT_F_ALLBITS) == 0);
1131 	KASSERT(l->l_proc == p);
1132 	KASSERT(l->l_stat == LSIDL);
1133 
1134 	/*
1135 	 * For unlocked lookup in proc_find_lwp(), make sure l->l_proc
1136 	 * is globally visible before the LWP becomes visible via the
1137 	 * pid_table.
1138 	 */
1139 #ifndef __HAVE_ATOMIC_AS_MEMBAR
1140 	membar_producer();
1141 #endif
1142 
1143 	/*
1144 	 * If the slot for p->p_pid currently points to the proc,
1145 	 * then we should usurp this ID for the LWP.  This happens
1146 	 * at least once per process (for the first LWP), and can
1147 	 * happen again if the first LWP for a process exits and
1148 	 * before the process creates another.
1149 	 */
1150 	mutex_enter(&proc_lock);
1151 	pid = p->p_pid;
1152 	pt = &pid_table[pid & pid_tbl_mask];
1153 	KASSERT(pt->pt_pid == pid);
1154 	if (PT_IS_PROC(pt->pt_slot)) {
1155 		KASSERT(PT_GET_PROC(pt->pt_slot) == p);
1156 		l->l_lid = pid;
1157 		pt->pt_slot = PT_SET_LWP(l);
1158 	} else {
1159 		/* Need to allocate a new slot. */
1160 		pid = proc_alloc_pid_slot(p, PT_SET_LWP(l));
1161 		if (pid != -1)
1162 			l->l_lid = pid;
1163 	}
1164 	mutex_exit(&proc_lock);
1165 
1166 	return pid;
1167 }
1168 
1169 static void __noinline
1170 proc_free_pid_internal(pid_t pid, uintptr_t type __diagused)
1171 {
1172 	struct pid_table *pt;
1173 
1174 	pt = &pid_table[pid & pid_tbl_mask];
1175 
1176 	KASSERT(PT_GET_TYPE(pt->pt_slot) == type);
1177 	KASSERT(pt->pt_pid == pid);
1178 
1179 	/* save pid use count in slot */
1180 	pt->pt_slot = PT_SET_FREE(pid & ~pid_tbl_mask);
1181 	pt->pt_pid = 0;
1182 
1183 	if (pt->pt_pgrp == NULL) {
1184 		/* link last freed entry onto ours */
1185 		pid &= pid_tbl_mask;
1186 		pt = &pid_table[last_free_pt];
1187 		pt->pt_slot = PT_SET_FREE(PT_NEXT(pt->pt_slot) | pid);
1188 		pt->pt_pid = 0;
1189 		last_free_pt = pid;
1190 		pid_alloc_cnt--;
1191 	}
1192 }
1193 
1194 /*
1195  * Free a process id - called from proc_free (in kern_exit.c)
1196  *
1197  * Called with the proc_lock held.
1198  */
1199 void
1200 proc_free_pid(pid_t pid)
1201 {
1202 
1203 	KASSERT(mutex_owned(&proc_lock));
1204 	proc_free_pid_internal(pid, PT_F_PROC);
1205 }
1206 
1207 /*
1208  * Free a process id used by an LWP.  If this was the process's
1209  * first LWP, we convert the slot to point to the process; the
1210  * entry will get cleaned up later when the process finishes exiting.
1211  *
1212  * If not, then it's the same as proc_free_pid().
1213  */
1214 void
1215 proc_free_lwpid(struct proc *p, pid_t pid)
1216 {
1217 
1218 	KASSERT(mutex_owned(&proc_lock));
1219 
1220 	if (__predict_true(p->p_pid == pid)) {
1221 		struct pid_table *pt;
1222 
1223 		pt = &pid_table[pid & pid_tbl_mask];
1224 
1225 		KASSERT(pt->pt_pid == pid);
1226 		KASSERT(PT_IS_LWP(pt->pt_slot));
1227 		KASSERT(PT_GET_LWP(pt->pt_slot)->l_proc == p);
1228 
1229 		pt->pt_slot = PT_SET_PROC(p);
1230 		return;
1231 	}
1232 	proc_free_pid_internal(pid, PT_F_LWP);
1233 }
1234 
1235 void
1236 proc_free_mem(struct proc *p)
1237 {
1238 
1239 	kdtrace_proc_dtor(NULL, p);
1240 	pool_cache_put(proc_cache, p);
1241 }
1242 
1243 /*
1244  * proc_enterpgrp: move p to a new or existing process group (and session).
1245  *
1246  * If we are creating a new pgrp, the pgid should equal
1247  * the calling process' pid.
1248  * If is only valid to enter a process group that is in the session
1249  * of the process.
1250  * Also mksess should only be set if we are creating a process group
1251  *
1252  * Only called from sys_setsid, sys_setpgid and posix_spawn/spawn_return.
1253  */
1254 int
1255 proc_enterpgrp(struct proc *curp, pid_t pid, pid_t pgid, bool mksess)
1256 {
1257 	struct pgrp *new_pgrp, *pgrp;
1258 	struct session *sess;
1259 	struct proc *p;
1260 	int rval;
1261 	pid_t pg_id = NO_PGID;
1262 
1263 	/* Allocate data areas we might need before doing any validity checks */
1264 	sess = mksess ? kmem_alloc(sizeof(*sess), KM_SLEEP) : NULL;
1265 	new_pgrp = kmem_alloc(sizeof(*new_pgrp), KM_SLEEP);
1266 
1267 	mutex_enter(&proc_lock);
1268 	rval = EPERM;	/* most common error (to save typing) */
1269 
1270 	/* Check pgrp exists or can be created */
1271 	pgrp = pid_table[pgid & pid_tbl_mask].pt_pgrp;
1272 	if (pgrp != NULL && pgrp->pg_id != pgid)
1273 		goto done;
1274 
1275 	/* Can only set another process under restricted circumstances. */
1276 	if (pid != curp->p_pid) {
1277 		/* Must exist and be one of our children... */
1278 		p = proc_find_internal(pid, false);
1279 		if (p == NULL || !p_inferior(p, curp)) {
1280 			rval = ESRCH;
1281 			goto done;
1282 		}
1283 		/* ... in the same session... */
1284 		if (sess != NULL || p->p_session != curp->p_session)
1285 			goto done;
1286 		/* ... existing pgid must be in same session ... */
1287 		if (pgrp != NULL && pgrp->pg_session != p->p_session)
1288 			goto done;
1289 		/* ... and not done an exec. */
1290 		if (p->p_flag & PK_EXEC) {
1291 			rval = EACCES;
1292 			goto done;
1293 		}
1294 	} else {
1295 		/* ... setsid() cannot re-enter a pgrp */
1296 		if (mksess && (curp->p_pgid == curp->p_pid ||
1297 		    pgrp_find(curp->p_pid)))
1298 			goto done;
1299 		p = curp;
1300 	}
1301 
1302 	/* Changing the process group/session of a session
1303 	   leader is definitely off limits. */
1304 	if (SESS_LEADER(p)) {
1305 		if (sess == NULL && p->p_pgrp == pgrp)
1306 			/* unless it's a definite noop */
1307 			rval = 0;
1308 		goto done;
1309 	}
1310 
1311 	/* Can only create a process group with id of process */
1312 	if (pgrp == NULL && pgid != pid)
1313 		goto done;
1314 
1315 	/* Can only create a session if creating pgrp */
1316 	if (sess != NULL && pgrp != NULL)
1317 		goto done;
1318 
1319 	/* Check we allocated memory for a pgrp... */
1320 	if (pgrp == NULL && new_pgrp == NULL)
1321 		goto done;
1322 
1323 	/* Don't attach to 'zombie' pgrp */
1324 	if (pgrp != NULL && LIST_EMPTY(&pgrp->pg_members))
1325 		goto done;
1326 
1327 	/* Expect to succeed now */
1328 	rval = 0;
1329 
1330 	if (pgrp == p->p_pgrp)
1331 		/* nothing to do */
1332 		goto done;
1333 
1334 	/* Ok all setup, link up required structures */
1335 
1336 	if (pgrp == NULL) {
1337 		pgrp = new_pgrp;
1338 		new_pgrp = NULL;
1339 		if (sess != NULL) {
1340 			sess->s_sid = p->p_pid;
1341 			sess->s_leader = p;
1342 			sess->s_count = 1;
1343 			sess->s_ttyvp = NULL;
1344 			sess->s_ttyp = NULL;
1345 			sess->s_flags = p->p_session->s_flags & ~S_LOGIN_SET;
1346 			memcpy(sess->s_login, p->p_session->s_login,
1347 			    sizeof(sess->s_login));
1348 			p->p_lflag &= ~PL_CONTROLT;
1349 		} else {
1350 			sess = p->p_pgrp->pg_session;
1351 			proc_sesshold(sess);
1352 		}
1353 		pgrp->pg_session = sess;
1354 		sess = NULL;
1355 
1356 		pgrp->pg_id = pgid;
1357 		LIST_INIT(&pgrp->pg_members);
1358 #ifdef DIAGNOSTIC
1359 		if (__predict_false(pid_table[pgid & pid_tbl_mask].pt_pgrp))
1360 			panic("enterpgrp: pgrp table slot in use");
1361 		if (__predict_false(mksess && p != curp))
1362 			panic("enterpgrp: mksession and p != curproc");
1363 #endif
1364 		pid_table[pgid & pid_tbl_mask].pt_pgrp = pgrp;
1365 		pgrp->pg_jobc = 0;
1366 	}
1367 
1368 	/*
1369 	 * Adjust eligibility of affected pgrps to participate in job control.
1370 	 * Increment eligibility counts before decrementing, otherwise we
1371 	 * could reach 0 spuriously during the first call.
1372 	 */
1373 	fixjobc(p, pgrp, 1);
1374 	fixjobc(p, p->p_pgrp, 0);
1375 
1376 	/* Interlock with ttread(). */
1377 	mutex_spin_enter(&tty_lock);
1378 
1379 	/* Move process to requested group. */
1380 	LIST_REMOVE(p, p_pglist);
1381 	if (LIST_EMPTY(&p->p_pgrp->pg_members))
1382 		/* defer delete until we've dumped the lock */
1383 		pg_id = p->p_pgrp->pg_id;
1384 	p->p_pgrp = pgrp;
1385 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
1386 
1387 	/* Done with the swap; we can release the tty mutex. */
1388 	mutex_spin_exit(&tty_lock);
1389 
1390     done:
1391 	if (pg_id != NO_PGID) {
1392 		/* Releases proc_lock. */
1393 		pg_delete(pg_id);
1394 	} else {
1395 		mutex_exit(&proc_lock);
1396 	}
1397 	if (sess != NULL)
1398 		kmem_free(sess, sizeof(*sess));
1399 	if (new_pgrp != NULL)
1400 		kmem_free(new_pgrp, sizeof(*new_pgrp));
1401 #ifdef DEBUG_PGRP
1402 	if (__predict_false(rval))
1403 		printf("enterpgrp(%d,%d,%d), curproc %d, rval %d\n",
1404 			pid, pgid, mksess, curp->p_pid, rval);
1405 #endif
1406 	return rval;
1407 }
1408 
1409 /*
1410  * proc_leavepgrp: remove a process from its process group.
1411  *  => must be called with the proc_lock held, which will be released;
1412  */
1413 void
1414 proc_leavepgrp(struct proc *p)
1415 {
1416 	struct pgrp *pgrp;
1417 
1418 	KASSERT(mutex_owned(&proc_lock));
1419 
1420 	/* Interlock with ttread() */
1421 	mutex_spin_enter(&tty_lock);
1422 	pgrp = p->p_pgrp;
1423 	LIST_REMOVE(p, p_pglist);
1424 	p->p_pgrp = NULL;
1425 	mutex_spin_exit(&tty_lock);
1426 
1427 	if (LIST_EMPTY(&pgrp->pg_members)) {
1428 		/* Releases proc_lock. */
1429 		pg_delete(pgrp->pg_id);
1430 	} else {
1431 		mutex_exit(&proc_lock);
1432 	}
1433 }
1434 
1435 /*
1436  * pg_remove: remove a process group from the table.
1437  *  => must be called with the proc_lock held;
1438  *  => returns process group to free;
1439  */
1440 static struct pgrp *
1441 pg_remove(pid_t pg_id)
1442 {
1443 	struct pgrp *pgrp;
1444 	struct pid_table *pt;
1445 
1446 	KASSERT(mutex_owned(&proc_lock));
1447 
1448 	pt = &pid_table[pg_id & pid_tbl_mask];
1449 	pgrp = pt->pt_pgrp;
1450 
1451 	KASSERT(pgrp != NULL);
1452 	KASSERT(pgrp->pg_id == pg_id);
1453 	KASSERT(LIST_EMPTY(&pgrp->pg_members));
1454 
1455 	pt->pt_pgrp = NULL;
1456 
1457 	if (!PT_VALID(pt->pt_slot)) {
1458 		/* Orphaned pgrp, put slot onto free list. */
1459 		KASSERT((PT_NEXT(pt->pt_slot) & pid_tbl_mask) == 0);
1460 		pg_id &= pid_tbl_mask;
1461 		pt = &pid_table[last_free_pt];
1462 		pt->pt_slot = PT_SET_FREE(PT_NEXT(pt->pt_slot) | pg_id);
1463 		KASSERT(pt->pt_pid == 0);
1464 		last_free_pt = pg_id;
1465 		pid_alloc_cnt--;
1466 	}
1467 	return pgrp;
1468 }
1469 
1470 /*
1471  * pg_delete: delete and free a process group.
1472  *  => must be called with the proc_lock held, which will be released.
1473  */
1474 static void
1475 pg_delete(pid_t pg_id)
1476 {
1477 	struct pgrp *pg;
1478 	struct tty *ttyp;
1479 	struct session *ss;
1480 
1481 	KASSERT(mutex_owned(&proc_lock));
1482 
1483 	pg = pid_table[pg_id & pid_tbl_mask].pt_pgrp;
1484 	if (pg == NULL || pg->pg_id != pg_id || !LIST_EMPTY(&pg->pg_members)) {
1485 		mutex_exit(&proc_lock);
1486 		return;
1487 	}
1488 
1489 	ss = pg->pg_session;
1490 
1491 	/* Remove reference (if any) from tty to this process group */
1492 	mutex_spin_enter(&tty_lock);
1493 	ttyp = ss->s_ttyp;
1494 	if (ttyp != NULL && ttyp->t_pgrp == pg) {
1495 		ttyp->t_pgrp = NULL;
1496 		KASSERT(ttyp->t_session == ss);
1497 	}
1498 	mutex_spin_exit(&tty_lock);
1499 
1500 	/*
1501 	 * The leading process group in a session is freed by proc_sessrele(),
1502 	 * if last reference.  It will also release the locks.
1503 	 */
1504 	pg = (ss->s_sid != pg->pg_id) ? pg_remove(pg_id) : NULL;
1505 	proc_sessrele(ss);
1506 
1507 	if (pg != NULL) {
1508 		/* Free it, if was not done above. */
1509 		kmem_free(pg, sizeof(struct pgrp));
1510 	}
1511 }
1512 
1513 /*
1514  * Adjust pgrp jobc counters when specified process changes process group.
1515  * We count the number of processes in each process group that "qualify"
1516  * the group for terminal job control (those with a parent in a different
1517  * process group of the same session).  If that count reaches zero, the
1518  * process group becomes orphaned.  Check both the specified process'
1519  * process group and that of its children.
1520  * entering == 0 => p is leaving specified group.
1521  * entering == 1 => p is entering specified group.
1522  *
1523  * Call with proc_lock held.
1524  */
1525 void
1526 fixjobc(struct proc *p, struct pgrp *pgrp, int entering)
1527 {
1528 	struct pgrp *hispgrp;
1529 	struct session *mysession = pgrp->pg_session;
1530 	struct proc *child;
1531 
1532 	KASSERT(mutex_owned(&proc_lock));
1533 
1534 	/*
1535 	 * Check p's parent to see whether p qualifies its own process
1536 	 * group; if so, adjust count for p's process group.
1537 	 */
1538 	hispgrp = p->p_pptr->p_pgrp;
1539 	if (hispgrp != pgrp && hispgrp->pg_session == mysession) {
1540 		if (entering) {
1541 			pgrp->pg_jobc++;
1542 			p->p_lflag &= ~PL_ORPHANPG;
1543 		} else {
1544 			KASSERT(pgrp->pg_jobc > 0);
1545 			if (--pgrp->pg_jobc == 0)
1546 				orphanpg(pgrp);
1547 		}
1548 	}
1549 
1550 	/*
1551 	 * Check this process' children to see whether they qualify
1552 	 * their process groups; if so, adjust counts for children's
1553 	 * process groups.
1554 	 */
1555 	LIST_FOREACH(child, &p->p_children, p_sibling) {
1556 		hispgrp = child->p_pgrp;
1557 		if (hispgrp != pgrp && hispgrp->pg_session == mysession &&
1558 		    !P_ZOMBIE(child)) {
1559 			if (entering) {
1560 				child->p_lflag &= ~PL_ORPHANPG;
1561 				hispgrp->pg_jobc++;
1562 			} else {
1563 				KASSERT(hispgrp->pg_jobc > 0);
1564 				if (--hispgrp->pg_jobc == 0)
1565 					orphanpg(hispgrp);
1566 			}
1567 		}
1568 	}
1569 }
1570 
1571 /*
1572  * A process group has become orphaned;
1573  * if there are any stopped processes in the group,
1574  * hang-up all process in that group.
1575  *
1576  * Call with proc_lock held.
1577  */
1578 static void
1579 orphanpg(struct pgrp *pg)
1580 {
1581 	struct proc *p;
1582 
1583 	KASSERT(mutex_owned(&proc_lock));
1584 
1585 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
1586 		if (p->p_stat == SSTOP) {
1587 			p->p_lflag |= PL_ORPHANPG;
1588 			psignal(p, SIGHUP);
1589 			psignal(p, SIGCONT);
1590 		}
1591 	}
1592 }
1593 
1594 #ifdef DDB
1595 #include <ddb/db_output.h>
1596 void pidtbl_dump(void);
1597 void
1598 pidtbl_dump(void)
1599 {
1600 	struct pid_table *pt;
1601 	struct proc *p;
1602 	struct pgrp *pgrp;
1603 	uintptr_t slot;
1604 	int id;
1605 
1606 	db_printf("pid table %p size %x, next %x, last %x\n",
1607 		pid_table, pid_tbl_mask+1,
1608 		next_free_pt, last_free_pt);
1609 	for (pt = pid_table, id = 0; id <= pid_tbl_mask; id++, pt++) {
1610 		slot = pt->pt_slot;
1611 		if (!PT_VALID(slot) && !pt->pt_pgrp)
1612 			continue;
1613 		if (PT_IS_LWP(slot)) {
1614 			p = PT_GET_LWP(slot)->l_proc;
1615 		} else if (PT_IS_PROC(slot)) {
1616 			p = PT_GET_PROC(slot);
1617 		} else {
1618 			p = NULL;
1619 		}
1620 		db_printf("  id %x: ", id);
1621 		if (p != NULL)
1622 			db_printf("slotpid %d proc %p id %d (0x%x) %s\n",
1623 				pt->pt_pid, p, p->p_pid, p->p_pid, p->p_comm);
1624 		else
1625 			db_printf("next %x use %x\n",
1626 				PT_NEXT(slot) & pid_tbl_mask,
1627 				PT_NEXT(slot) & ~pid_tbl_mask);
1628 		if ((pgrp = pt->pt_pgrp)) {
1629 			db_printf("\tsession %p, sid %d, count %d, login %s\n",
1630 			    pgrp->pg_session, pgrp->pg_session->s_sid,
1631 			    pgrp->pg_session->s_count,
1632 			    pgrp->pg_session->s_login);
1633 			db_printf("\tpgrp %p, pg_id %d, pg_jobc %d, members %p\n",
1634 			    pgrp, pgrp->pg_id, pgrp->pg_jobc,
1635 			    LIST_FIRST(&pgrp->pg_members));
1636 			LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
1637 				db_printf("\t\tpid %d addr %p pgrp %p %s\n",
1638 				    p->p_pid, p, p->p_pgrp, p->p_comm);
1639 			}
1640 		}
1641 	}
1642 }
1643 #endif /* DDB */
1644 
1645 #ifdef KSTACK_CHECK_MAGIC
1646 
1647 #define	KSTACK_MAGIC	0xdeadbeaf
1648 
1649 /* XXX should be per process basis? */
1650 static int	kstackleftmin = KSTACK_SIZE;
1651 static int	kstackleftthres = KSTACK_SIZE / 8;
1652 
1653 void
1654 kstack_setup_magic(const struct lwp *l)
1655 {
1656 	uint32_t *ip;
1657 	uint32_t const *end;
1658 
1659 	KASSERT(l != NULL);
1660 	KASSERT(l != &lwp0);
1661 
1662 	/*
1663 	 * fill all the stack with magic number
1664 	 * so that later modification on it can be detected.
1665 	 */
1666 	ip = (uint32_t *)KSTACK_LOWEST_ADDR(l);
1667 	end = (uint32_t *)((char *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE);
1668 	for (; ip < end; ip++) {
1669 		*ip = KSTACK_MAGIC;
1670 	}
1671 }
1672 
1673 void
1674 kstack_check_magic(const struct lwp *l)
1675 {
1676 	uint32_t const *ip, *end;
1677 	int stackleft;
1678 
1679 	KASSERT(l != NULL);
1680 
1681 	/* don't check proc0 */ /*XXX*/
1682 	if (l == &lwp0)
1683 		return;
1684 
1685 #ifdef __MACHINE_STACK_GROWS_UP
1686 	/* stack grows upwards (eg. hppa) */
1687 	ip = (uint32_t *)((void *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE);
1688 	end = (uint32_t *)KSTACK_LOWEST_ADDR(l);
1689 	for (ip--; ip >= end; ip--)
1690 		if (*ip != KSTACK_MAGIC)
1691 			break;
1692 
1693 	stackleft = (void *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE - (void *)ip;
1694 #else /* __MACHINE_STACK_GROWS_UP */
1695 	/* stack grows downwards (eg. i386) */
1696 	ip = (uint32_t *)KSTACK_LOWEST_ADDR(l);
1697 	end = (uint32_t *)((char *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE);
1698 	for (; ip < end; ip++)
1699 		if (*ip != KSTACK_MAGIC)
1700 			break;
1701 
1702 	stackleft = ((const char *)ip) - (const char *)KSTACK_LOWEST_ADDR(l);
1703 #endif /* __MACHINE_STACK_GROWS_UP */
1704 
1705 	if (kstackleftmin > stackleft) {
1706 		kstackleftmin = stackleft;
1707 		if (stackleft < kstackleftthres)
1708 			printf("warning: kernel stack left %d bytes"
1709 			    "(pid %u:lid %u)\n", stackleft,
1710 			    (u_int)l->l_proc->p_pid, (u_int)l->l_lid);
1711 	}
1712 
1713 	if (stackleft <= 0) {
1714 		panic("magic on the top of kernel stack changed for "
1715 		    "pid %u, lid %u: maybe kernel stack overflow",
1716 		    (u_int)l->l_proc->p_pid, (u_int)l->l_lid);
1717 	}
1718 }
1719 #endif /* KSTACK_CHECK_MAGIC */
1720 
1721 int
1722 proclist_foreach_call(struct proclist *list,
1723     int (*callback)(struct proc *, void *arg), void *arg)
1724 {
1725 	struct proc marker;
1726 	struct proc *p;
1727 	int ret = 0;
1728 
1729 	marker.p_flag = PK_MARKER;
1730 	mutex_enter(&proc_lock);
1731 	for (p = LIST_FIRST(list); ret == 0 && p != NULL;) {
1732 		if (p->p_flag & PK_MARKER) {
1733 			p = LIST_NEXT(p, p_list);
1734 			continue;
1735 		}
1736 		LIST_INSERT_AFTER(p, &marker, p_list);
1737 		ret = (*callback)(p, arg);
1738 		KASSERT(mutex_owned(&proc_lock));
1739 		p = LIST_NEXT(&marker, p_list);
1740 		LIST_REMOVE(&marker, p_list);
1741 	}
1742 	mutex_exit(&proc_lock);
1743 
1744 	return ret;
1745 }
1746 
1747 int
1748 proc_vmspace_getref(struct proc *p, struct vmspace **vm)
1749 {
1750 
1751 	/* XXXCDC: how should locking work here? */
1752 
1753 	/* curproc exception is for coredump. */
1754 
1755 	if ((p != curproc && (p->p_sflag & PS_WEXIT) != 0) ||
1756 	    (p->p_vmspace->vm_refcnt < 1)) {
1757 		return EFAULT;
1758 	}
1759 
1760 	uvmspace_addref(p->p_vmspace);
1761 	*vm = p->p_vmspace;
1762 
1763 	return 0;
1764 }
1765 
1766 /*
1767  * Acquire a write lock on the process credential.
1768  */
1769 void
1770 proc_crmod_enter(void)
1771 {
1772 	struct lwp *l = curlwp;
1773 	struct proc *p = l->l_proc;
1774 	kauth_cred_t oc;
1775 
1776 	/* Reset what needs to be reset in plimit. */
1777 	if (p->p_limit->pl_corename != defcorename) {
1778 		lim_setcorename(p, defcorename, 0);
1779 	}
1780 
1781 	mutex_enter(p->p_lock);
1782 
1783 	/* Ensure the LWP cached credentials are up to date. */
1784 	if ((oc = l->l_cred) != p->p_cred) {
1785 		kauth_cred_hold(p->p_cred);
1786 		l->l_cred = p->p_cred;
1787 		kauth_cred_free(oc);
1788 	}
1789 }
1790 
1791 /*
1792  * Set in a new process credential, and drop the write lock.  The credential
1793  * must have a reference already.  Optionally, free a no-longer required
1794  * credential.  The scheduler also needs to inspect p_cred, so we also
1795  * briefly acquire the sched state mutex.
1796  */
1797 void
1798 proc_crmod_leave(kauth_cred_t scred, kauth_cred_t fcred, bool sugid)
1799 {
1800 	struct lwp *l = curlwp, *l2;
1801 	struct proc *p = l->l_proc;
1802 	kauth_cred_t oc;
1803 
1804 	KASSERT(mutex_owned(p->p_lock));
1805 
1806 	/* Is there a new credential to set in? */
1807 	if (scred != NULL) {
1808 		p->p_cred = scred;
1809 		LIST_FOREACH(l2, &p->p_lwps, l_sibling) {
1810 			if (l2 != l)
1811 				l2->l_prflag |= LPR_CRMOD;
1812 		}
1813 
1814 		/* Ensure the LWP cached credentials are up to date. */
1815 		if ((oc = l->l_cred) != scred) {
1816 			kauth_cred_hold(scred);
1817 			l->l_cred = scred;
1818 		}
1819 	} else
1820 		oc = NULL;	/* XXXgcc */
1821 
1822 	if (sugid) {
1823 		/*
1824 		 * Mark process as having changed credentials, stops
1825 		 * tracing etc.
1826 		 */
1827 		p->p_flag |= PK_SUGID;
1828 	}
1829 
1830 	mutex_exit(p->p_lock);
1831 
1832 	/* If there is a credential to be released, free it now. */
1833 	if (fcred != NULL) {
1834 		KASSERT(scred != NULL);
1835 		kauth_cred_free(fcred);
1836 		if (oc != scred)
1837 			kauth_cred_free(oc);
1838 	}
1839 }
1840 
1841 /*
1842  * proc_specific_key_create --
1843  *	Create a key for subsystem proc-specific data.
1844  */
1845 int
1846 proc_specific_key_create(specificdata_key_t *keyp, specificdata_dtor_t dtor)
1847 {
1848 
1849 	return (specificdata_key_create(proc_specificdata_domain, keyp, dtor));
1850 }
1851 
1852 /*
1853  * proc_specific_key_delete --
1854  *	Delete a key for subsystem proc-specific data.
1855  */
1856 void
1857 proc_specific_key_delete(specificdata_key_t key)
1858 {
1859 
1860 	specificdata_key_delete(proc_specificdata_domain, key);
1861 }
1862 
1863 /*
1864  * proc_initspecific --
1865  *	Initialize a proc's specificdata container.
1866  */
1867 void
1868 proc_initspecific(struct proc *p)
1869 {
1870 	int error __diagused;
1871 
1872 	error = specificdata_init(proc_specificdata_domain, &p->p_specdataref);
1873 	KASSERT(error == 0);
1874 }
1875 
1876 /*
1877  * proc_finispecific --
1878  *	Finalize a proc's specificdata container.
1879  */
1880 void
1881 proc_finispecific(struct proc *p)
1882 {
1883 
1884 	specificdata_fini(proc_specificdata_domain, &p->p_specdataref);
1885 }
1886 
1887 /*
1888  * proc_getspecific --
1889  *	Return proc-specific data corresponding to the specified key.
1890  */
1891 void *
1892 proc_getspecific(struct proc *p, specificdata_key_t key)
1893 {
1894 
1895 	return (specificdata_getspecific(proc_specificdata_domain,
1896 					 &p->p_specdataref, key));
1897 }
1898 
1899 /*
1900  * proc_setspecific --
1901  *	Set proc-specific data corresponding to the specified key.
1902  */
1903 void
1904 proc_setspecific(struct proc *p, specificdata_key_t key, void *data)
1905 {
1906 
1907 	specificdata_setspecific(proc_specificdata_domain,
1908 				 &p->p_specdataref, key, data);
1909 }
1910 
1911 int
1912 proc_uidmatch(kauth_cred_t cred, kauth_cred_t target)
1913 {
1914 	int r = 0;
1915 
1916 	if (kauth_cred_getuid(cred) != kauth_cred_getuid(target) ||
1917 	    kauth_cred_getuid(cred) != kauth_cred_getsvuid(target)) {
1918 		/*
1919 		 * suid proc of ours or proc not ours
1920 		 */
1921 		r = EPERM;
1922 	} else if (kauth_cred_getgid(target) != kauth_cred_getsvgid(target)) {
1923 		/*
1924 		 * sgid proc has sgid back to us temporarily
1925 		 */
1926 		r = EPERM;
1927 	} else {
1928 		/*
1929 		 * our rgid must be in target's group list (ie,
1930 		 * sub-processes started by a sgid process)
1931 		 */
1932 		int ismember = 0;
1933 
1934 		if (kauth_cred_ismember_gid(cred,
1935 		    kauth_cred_getgid(target), &ismember) != 0 ||
1936 		    !ismember)
1937 			r = EPERM;
1938 	}
1939 
1940 	return (r);
1941 }
1942 
1943 /*
1944  * sysctl stuff
1945  */
1946 
1947 #define KERN_PROCSLOP	(5 * sizeof(struct kinfo_proc))
1948 
1949 static const u_int sysctl_flagmap[] = {
1950 	PK_ADVLOCK, P_ADVLOCK,
1951 	PK_EXEC, P_EXEC,
1952 	PK_NOCLDWAIT, P_NOCLDWAIT,
1953 	PK_32, P_32,
1954 	PK_CLDSIGIGN, P_CLDSIGIGN,
1955 	PK_SUGID, P_SUGID,
1956 	0
1957 };
1958 
1959 static const u_int sysctl_sflagmap[] = {
1960 	PS_NOCLDSTOP, P_NOCLDSTOP,
1961 	PS_WEXIT, P_WEXIT,
1962 	PS_STOPFORK, P_STOPFORK,
1963 	PS_STOPEXEC, P_STOPEXEC,
1964 	PS_STOPEXIT, P_STOPEXIT,
1965 	0
1966 };
1967 
1968 static const u_int sysctl_slflagmap[] = {
1969 	PSL_TRACED, P_TRACED,
1970 	PSL_CHTRACED, P_CHTRACED,
1971 	PSL_SYSCALL, P_SYSCALL,
1972 	0
1973 };
1974 
1975 static const u_int sysctl_lflagmap[] = {
1976 	PL_CONTROLT, P_CONTROLT,
1977 	PL_PPWAIT, P_PPWAIT,
1978 	0
1979 };
1980 
1981 static const u_int sysctl_stflagmap[] = {
1982 	PST_PROFIL, P_PROFIL,
1983 	0
1984 
1985 };
1986 
1987 /* used by kern_lwp also */
1988 const u_int sysctl_lwpflagmap[] = {
1989 	LW_SINTR, L_SINTR,
1990 	LW_SYSTEM, L_SYSTEM,
1991 	0
1992 };
1993 
1994 /*
1995  * Find the most ``active'' lwp of a process and return it for ps display
1996  * purposes
1997  */
1998 static struct lwp *
1999 proc_active_lwp(struct proc *p)
2000 {
2001 	static const int ostat[] = {
2002 		0,
2003 		2,	/* LSIDL */
2004 		6,	/* LSRUN */
2005 		5,	/* LSSLEEP */
2006 		4,	/* LSSTOP */
2007 		0,	/* LSZOMB */
2008 		1,	/* LSDEAD */
2009 		7,	/* LSONPROC */
2010 		3	/* LSSUSPENDED */
2011 	};
2012 
2013 	struct lwp *l, *lp = NULL;
2014 	LIST_FOREACH(l, &p->p_lwps, l_sibling) {
2015 		KASSERT(l->l_stat >= 0 && l->l_stat < __arraycount(ostat));
2016 		if (lp == NULL ||
2017 		    ostat[l->l_stat] > ostat[lp->l_stat] ||
2018 		    (ostat[l->l_stat] == ostat[lp->l_stat] &&
2019 		    l->l_cpticks > lp->l_cpticks)) {
2020 			lp = l;
2021 			continue;
2022 		}
2023 	}
2024 	return lp;
2025 }
2026 
2027 static int
2028 sysctl_doeproc(SYSCTLFN_ARGS)
2029 {
2030 	union {
2031 		struct kinfo_proc kproc;
2032 		struct kinfo_proc2 kproc2;
2033 	} *kbuf;
2034 	struct proc *p, *next, *marker;
2035 	char *where, *dp;
2036 	int type, op, arg, error;
2037 	u_int elem_size, kelem_size, elem_count;
2038 	size_t buflen, needed;
2039 	bool match, zombie, mmmbrains;
2040 	const bool allowaddr = get_expose_address(curproc);
2041 
2042 	if (namelen == 1 && name[0] == CTL_QUERY)
2043 		return (sysctl_query(SYSCTLFN_CALL(rnode)));
2044 
2045 	dp = where = oldp;
2046 	buflen = where != NULL ? *oldlenp : 0;
2047 	error = 0;
2048 	needed = 0;
2049 	type = rnode->sysctl_num;
2050 
2051 	if (type == KERN_PROC) {
2052 		if (namelen == 0)
2053 			return EINVAL;
2054 		switch (op = name[0]) {
2055 		case KERN_PROC_ALL:
2056 			if (namelen != 1)
2057 				return EINVAL;
2058 			arg = 0;
2059 			break;
2060 		default:
2061 			if (namelen != 2)
2062 				return EINVAL;
2063 			arg = name[1];
2064 			break;
2065 		}
2066 		elem_count = 0;	/* Hush little compiler, don't you cry */
2067 		kelem_size = elem_size = sizeof(kbuf->kproc);
2068 	} else {
2069 		if (namelen != 4)
2070 			return EINVAL;
2071 		op = name[0];
2072 		arg = name[1];
2073 		elem_size = name[2];
2074 		elem_count = name[3];
2075 		kelem_size = sizeof(kbuf->kproc2);
2076 	}
2077 
2078 	sysctl_unlock();
2079 
2080 	kbuf = kmem_zalloc(sizeof(*kbuf), KM_SLEEP);
2081 	marker = kmem_alloc(sizeof(*marker), KM_SLEEP);
2082 	marker->p_flag = PK_MARKER;
2083 
2084 	mutex_enter(&proc_lock);
2085 	/*
2086 	 * Start with zombies to prevent reporting processes twice, in case they
2087 	 * are dying and being moved from the list of alive processes to zombies.
2088 	 */
2089 	mmmbrains = true;
2090 	for (p = LIST_FIRST(&zombproc);; p = next) {
2091 		if (p == NULL) {
2092 			if (mmmbrains) {
2093 				p = LIST_FIRST(&allproc);
2094 				mmmbrains = false;
2095 			}
2096 			if (p == NULL)
2097 				break;
2098 		}
2099 		next = LIST_NEXT(p, p_list);
2100 		if ((p->p_flag & PK_MARKER) != 0)
2101 			continue;
2102 
2103 		/*
2104 		 * Skip embryonic processes.
2105 		 */
2106 		if (p->p_stat == SIDL)
2107 			continue;
2108 
2109 		mutex_enter(p->p_lock);
2110 		error = kauth_authorize_process(l->l_cred,
2111 		    KAUTH_PROCESS_CANSEE, p,
2112 		    KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_EPROC), NULL, NULL);
2113 		if (error != 0) {
2114 			mutex_exit(p->p_lock);
2115 			continue;
2116 		}
2117 
2118 		/*
2119 		 * Hande all the operations in one switch on the cost of
2120 		 * algorithm complexity is on purpose. The win splitting this
2121 		 * function into several similar copies makes maintenance burden
2122 		 * burden, code grow and boost is neglible in practical systems.
2123 		 */
2124 		switch (op) {
2125 		case KERN_PROC_PID:
2126 			match = (p->p_pid == (pid_t)arg);
2127 			break;
2128 
2129 		case KERN_PROC_PGRP:
2130 			match = (p->p_pgrp->pg_id == (pid_t)arg);
2131 			break;
2132 
2133 		case KERN_PROC_SESSION:
2134 			match = (p->p_session->s_sid == (pid_t)arg);
2135 			break;
2136 
2137 		case KERN_PROC_TTY:
2138 			match = true;
2139 			if (arg == (int) KERN_PROC_TTY_REVOKE) {
2140 				if ((p->p_lflag & PL_CONTROLT) == 0 ||
2141 				    p->p_session->s_ttyp == NULL ||
2142 				    p->p_session->s_ttyvp != NULL) {
2143 				    	match = false;
2144 				}
2145 			} else if ((p->p_lflag & PL_CONTROLT) == 0 ||
2146 			    p->p_session->s_ttyp == NULL) {
2147 				if ((dev_t)arg != KERN_PROC_TTY_NODEV) {
2148 					match = false;
2149 				}
2150 			} else if (p->p_session->s_ttyp->t_dev != (dev_t)arg) {
2151 				match = false;
2152 			}
2153 			break;
2154 
2155 		case KERN_PROC_UID:
2156 			match = (kauth_cred_geteuid(p->p_cred) == (uid_t)arg);
2157 			break;
2158 
2159 		case KERN_PROC_RUID:
2160 			match = (kauth_cred_getuid(p->p_cred) == (uid_t)arg);
2161 			break;
2162 
2163 		case KERN_PROC_GID:
2164 			match = (kauth_cred_getegid(p->p_cred) == (uid_t)arg);
2165 			break;
2166 
2167 		case KERN_PROC_RGID:
2168 			match = (kauth_cred_getgid(p->p_cred) == (uid_t)arg);
2169 			break;
2170 
2171 		case KERN_PROC_ALL:
2172 			match = true;
2173 			/* allow everything */
2174 			break;
2175 
2176 		default:
2177 			error = EINVAL;
2178 			mutex_exit(p->p_lock);
2179 			goto cleanup;
2180 		}
2181 		if (!match) {
2182 			mutex_exit(p->p_lock);
2183 			continue;
2184 		}
2185 
2186 		/*
2187 		 * Grab a hold on the process.
2188 		 */
2189 		if (mmmbrains) {
2190 			zombie = true;
2191 		} else {
2192 			zombie = !rw_tryenter(&p->p_reflock, RW_READER);
2193 		}
2194 		if (zombie) {
2195 			LIST_INSERT_AFTER(p, marker, p_list);
2196 		}
2197 
2198 		if (buflen >= elem_size &&
2199 		    (type == KERN_PROC || elem_count > 0)) {
2200 			ruspace(p);	/* Update process vm resource use */
2201 
2202 			if (type == KERN_PROC) {
2203 				fill_proc(p, &kbuf->kproc.kp_proc, allowaddr);
2204 				fill_eproc(p, &kbuf->kproc.kp_eproc, zombie,
2205 				    allowaddr);
2206 			} else {
2207 				fill_kproc2(p, &kbuf->kproc2, zombie,
2208 				    allowaddr);
2209 				elem_count--;
2210 			}
2211 			mutex_exit(p->p_lock);
2212 			mutex_exit(&proc_lock);
2213 			/*
2214 			 * Copy out elem_size, but not larger than kelem_size
2215 			 */
2216 			error = sysctl_copyout(l, kbuf, dp,
2217 			    uimin(kelem_size, elem_size));
2218 			mutex_enter(&proc_lock);
2219 			if (error) {
2220 				goto bah;
2221 			}
2222 			dp += elem_size;
2223 			buflen -= elem_size;
2224 		} else {
2225 			mutex_exit(p->p_lock);
2226 		}
2227 		needed += elem_size;
2228 
2229 		/*
2230 		 * Release reference to process.
2231 		 */
2232 	 	if (zombie) {
2233 			next = LIST_NEXT(marker, p_list);
2234  			LIST_REMOVE(marker, p_list);
2235 		} else {
2236 			rw_exit(&p->p_reflock);
2237 			next = LIST_NEXT(p, p_list);
2238 		}
2239 
2240 		/*
2241 		 * Short-circuit break quickly!
2242 		 */
2243 		if (op == KERN_PROC_PID)
2244                 	break;
2245 	}
2246 	mutex_exit(&proc_lock);
2247 
2248 	if (where != NULL) {
2249 		*oldlenp = dp - where;
2250 		if (needed > *oldlenp) {
2251 			error = ENOMEM;
2252 			goto out;
2253 		}
2254 	} else {
2255 		needed += KERN_PROCSLOP;
2256 		*oldlenp = needed;
2257 	}
2258 	kmem_free(kbuf, sizeof(*kbuf));
2259 	kmem_free(marker, sizeof(*marker));
2260 	sysctl_relock();
2261 	return 0;
2262  bah:
2263  	if (zombie)
2264  		LIST_REMOVE(marker, p_list);
2265 	else
2266 		rw_exit(&p->p_reflock);
2267  cleanup:
2268 	mutex_exit(&proc_lock);
2269  out:
2270 	kmem_free(kbuf, sizeof(*kbuf));
2271 	kmem_free(marker, sizeof(*marker));
2272 	sysctl_relock();
2273 	return error;
2274 }
2275 
2276 int
2277 copyin_psstrings(struct proc *p, struct ps_strings *arginfo)
2278 {
2279 #if !defined(_RUMPKERNEL)
2280 	int retval;
2281 
2282 	if (p->p_flag & PK_32) {
2283 		MODULE_HOOK_CALL(kern_proc32_copyin_hook, (p, arginfo),
2284 		    enosys(), retval);
2285 		return retval;
2286 	}
2287 #endif /* !defined(_RUMPKERNEL) */
2288 
2289 	return copyin_proc(p, (void *)p->p_psstrp, arginfo, sizeof(*arginfo));
2290 }
2291 
2292 static int
2293 copy_procargs_sysctl_cb(void *cookie_, const void *src, size_t off, size_t len)
2294 {
2295 	void **cookie = cookie_;
2296 	struct lwp *l = cookie[0];
2297 	char *dst = cookie[1];
2298 
2299 	return sysctl_copyout(l, src, dst + off, len);
2300 }
2301 
2302 /*
2303  * sysctl helper routine for kern.proc_args pseudo-subtree.
2304  */
2305 static int
2306 sysctl_kern_proc_args(SYSCTLFN_ARGS)
2307 {
2308 	struct ps_strings pss;
2309 	struct proc *p;
2310 	pid_t pid;
2311 	int type, error;
2312 	void *cookie[2];
2313 
2314 	if (namelen == 1 && name[0] == CTL_QUERY)
2315 		return (sysctl_query(SYSCTLFN_CALL(rnode)));
2316 
2317 	if (newp != NULL || namelen != 2)
2318 		return (EINVAL);
2319 	pid = name[0];
2320 	type = name[1];
2321 
2322 	switch (type) {
2323 	case KERN_PROC_PATHNAME:
2324 		sysctl_unlock();
2325 		error = fill_pathname(l, pid, oldp, oldlenp);
2326 		sysctl_relock();
2327 		return error;
2328 
2329 	case KERN_PROC_CWD:
2330 		sysctl_unlock();
2331 		error = fill_cwd(l, pid, oldp, oldlenp);
2332 		sysctl_relock();
2333 		return error;
2334 
2335 	case KERN_PROC_ARGV:
2336 	case KERN_PROC_NARGV:
2337 	case KERN_PROC_ENV:
2338 	case KERN_PROC_NENV:
2339 		/* ok */
2340 		break;
2341 	default:
2342 		return (EINVAL);
2343 	}
2344 
2345 	sysctl_unlock();
2346 
2347 	/* check pid */
2348 	mutex_enter(&proc_lock);
2349 	if ((p = proc_find(pid)) == NULL) {
2350 		error = EINVAL;
2351 		goto out_locked;
2352 	}
2353 	mutex_enter(p->p_lock);
2354 
2355 	/* Check permission. */
2356 	if (type == KERN_PROC_ARGV || type == KERN_PROC_NARGV)
2357 		error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE,
2358 		    p, KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ARGS), NULL, NULL);
2359 	else if (type == KERN_PROC_ENV || type == KERN_PROC_NENV)
2360 		error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE,
2361 		    p, KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENV), NULL, NULL);
2362 	else
2363 		error = EINVAL; /* XXXGCC */
2364 	if (error) {
2365 		mutex_exit(p->p_lock);
2366 		goto out_locked;
2367 	}
2368 
2369 	if (oldp == NULL) {
2370 		if (type == KERN_PROC_NARGV || type == KERN_PROC_NENV)
2371 			*oldlenp = sizeof (int);
2372 		else
2373 			*oldlenp = ARG_MAX;	/* XXX XXX XXX */
2374 		error = 0;
2375 		mutex_exit(p->p_lock);
2376 		goto out_locked;
2377 	}
2378 
2379 	/*
2380 	 * Zombies don't have a stack, so we can't read their psstrings.
2381 	 * System processes also don't have a user stack.
2382 	 */
2383 	if (P_ZOMBIE(p) || (p->p_flag & PK_SYSTEM) != 0) {
2384 		error = EINVAL;
2385 		mutex_exit(p->p_lock);
2386 		goto out_locked;
2387 	}
2388 
2389 	error = rw_tryenter(&p->p_reflock, RW_READER) ? 0 : EBUSY;
2390 	mutex_exit(p->p_lock);
2391 	if (error) {
2392 		goto out_locked;
2393 	}
2394 	mutex_exit(&proc_lock);
2395 
2396 	if (type == KERN_PROC_NARGV || type == KERN_PROC_NENV) {
2397 		int value;
2398 		if ((error = copyin_psstrings(p, &pss)) == 0) {
2399 			if (type == KERN_PROC_NARGV)
2400 				value = pss.ps_nargvstr;
2401 			else
2402 				value = pss.ps_nenvstr;
2403 			error = sysctl_copyout(l, &value, oldp, sizeof(value));
2404 			*oldlenp = sizeof(value);
2405 		}
2406 	} else {
2407 		cookie[0] = l;
2408 		cookie[1] = oldp;
2409 		error = copy_procargs(p, type, oldlenp,
2410 		    copy_procargs_sysctl_cb, cookie);
2411 	}
2412 	rw_exit(&p->p_reflock);
2413 	sysctl_relock();
2414 	return error;
2415 
2416 out_locked:
2417 	mutex_exit(&proc_lock);
2418 	sysctl_relock();
2419 	return error;
2420 }
2421 
2422 int
2423 copy_procargs(struct proc *p, int oid, size_t *limit,
2424     int (*cb)(void *, const void *, size_t, size_t), void *cookie)
2425 {
2426 	struct ps_strings pss;
2427 	size_t len, i, loaded, entry_len;
2428 	struct uio auio;
2429 	struct iovec aiov;
2430 	int error, argvlen;
2431 	char *arg;
2432 	char **argv;
2433 	vaddr_t user_argv;
2434 	struct vmspace *vmspace;
2435 
2436 	/*
2437 	 * Allocate a temporary buffer to hold the argument vector and
2438 	 * the arguments themselve.
2439 	 */
2440 	arg = kmem_alloc(PAGE_SIZE, KM_SLEEP);
2441 	argv = kmem_alloc(PAGE_SIZE, KM_SLEEP);
2442 
2443 	/*
2444 	 * Lock the process down in memory.
2445 	 */
2446 	vmspace = p->p_vmspace;
2447 	uvmspace_addref(vmspace);
2448 
2449 	/*
2450 	 * Read in the ps_strings structure.
2451 	 */
2452 	if ((error = copyin_psstrings(p, &pss)) != 0)
2453 		goto done;
2454 
2455 	/*
2456 	 * Now read the address of the argument vector.
2457 	 */
2458 	switch (oid) {
2459 	case KERN_PROC_ARGV:
2460 		user_argv = (uintptr_t)pss.ps_argvstr;
2461 		argvlen = pss.ps_nargvstr;
2462 		break;
2463 	case KERN_PROC_ENV:
2464 		user_argv = (uintptr_t)pss.ps_envstr;
2465 		argvlen = pss.ps_nenvstr;
2466 		break;
2467 	default:
2468 		error = EINVAL;
2469 		goto done;
2470 	}
2471 
2472 	if (argvlen < 0) {
2473 		error = EIO;
2474 		goto done;
2475 	}
2476 
2477 
2478 	/*
2479 	 * Now copy each string.
2480 	 */
2481 	len = 0; /* bytes written to user buffer */
2482 	loaded = 0; /* bytes from argv already processed */
2483 	i = 0; /* To make compiler happy */
2484 	entry_len = PROC_PTRSZ(p);
2485 
2486 	for (; argvlen; --argvlen) {
2487 		int finished = 0;
2488 		vaddr_t base;
2489 		size_t xlen;
2490 		int j;
2491 
2492 		if (loaded == 0) {
2493 			size_t rem = entry_len * argvlen;
2494 			loaded = MIN(rem, PAGE_SIZE);
2495 			error = copyin_vmspace(vmspace,
2496 			    (const void *)user_argv, argv, loaded);
2497 			if (error)
2498 				break;
2499 			user_argv += loaded;
2500 			i = 0;
2501 		}
2502 
2503 #if !defined(_RUMPKERNEL)
2504 		if (p->p_flag & PK_32)
2505 			MODULE_HOOK_CALL(kern_proc32_base_hook,
2506 			    (argv, i++), 0, base);
2507 		else
2508 #endif /* !defined(_RUMPKERNEL) */
2509 			base = (vaddr_t)argv[i++];
2510 		loaded -= entry_len;
2511 
2512 		/*
2513 		 * The program has messed around with its arguments,
2514 		 * possibly deleting some, and replacing them with
2515 		 * NULL's. Treat this as the last argument and not
2516 		 * a failure.
2517 		 */
2518 		if (base == 0)
2519 			break;
2520 
2521 		while (!finished) {
2522 			xlen = PAGE_SIZE - (base & PAGE_MASK);
2523 
2524 			aiov.iov_base = arg;
2525 			aiov.iov_len = PAGE_SIZE;
2526 			auio.uio_iov = &aiov;
2527 			auio.uio_iovcnt = 1;
2528 			auio.uio_offset = base;
2529 			auio.uio_resid = xlen;
2530 			auio.uio_rw = UIO_READ;
2531 			UIO_SETUP_SYSSPACE(&auio);
2532 			error = uvm_io(&vmspace->vm_map, &auio, 0);
2533 			if (error)
2534 				goto done;
2535 
2536 			/* Look for the end of the string */
2537 			for (j = 0; j < xlen; j++) {
2538 				if (arg[j] == '\0') {
2539 					xlen = j + 1;
2540 					finished = 1;
2541 					break;
2542 				}
2543 			}
2544 
2545 			/* Check for user buffer overflow */
2546 			if (len + xlen > *limit) {
2547 				finished = 1;
2548 				if (len > *limit)
2549 					xlen = 0;
2550 				else
2551 					xlen = *limit - len;
2552 			}
2553 
2554 			/* Copyout the page */
2555 			error = (*cb)(cookie, arg, len, xlen);
2556 			if (error)
2557 				goto done;
2558 
2559 			len += xlen;
2560 			base += xlen;
2561 		}
2562 	}
2563 	*limit = len;
2564 
2565 done:
2566 	kmem_free(argv, PAGE_SIZE);
2567 	kmem_free(arg, PAGE_SIZE);
2568 	uvmspace_free(vmspace);
2569 	return error;
2570 }
2571 
2572 /*
2573  * Fill in a proc structure for the specified process.
2574  */
2575 static void
2576 fill_proc(const struct proc *psrc, struct proc *p, bool allowaddr)
2577 {
2578 	COND_SET_STRUCT(p->p_list, psrc->p_list, allowaddr);
2579 	memset(&p->p_auxlock, 0, sizeof(p->p_auxlock));
2580 	COND_SET_STRUCT(p->p_lock, psrc->p_lock, allowaddr);
2581 	memset(&p->p_stmutex, 0, sizeof(p->p_stmutex));
2582 	memset(&p->p_reflock, 0, sizeof(p->p_reflock));
2583 	COND_SET_STRUCT(p->p_waitcv, psrc->p_waitcv, allowaddr);
2584 	COND_SET_STRUCT(p->p_lwpcv, psrc->p_lwpcv, allowaddr);
2585 	COND_SET_PTR(p->p_cred, psrc->p_cred, allowaddr);
2586 	COND_SET_PTR(p->p_fd, psrc->p_fd, allowaddr);
2587 	COND_SET_PTR(p->p_cwdi, psrc->p_cwdi, allowaddr);
2588 	COND_SET_PTR(p->p_stats, psrc->p_stats, allowaddr);
2589 	COND_SET_PTR(p->p_limit, psrc->p_limit, allowaddr);
2590 	COND_SET_PTR(p->p_vmspace, psrc->p_vmspace, allowaddr);
2591 	COND_SET_PTR(p->p_sigacts, psrc->p_sigacts, allowaddr);
2592 	COND_SET_PTR(p->p_aio, psrc->p_aio, allowaddr);
2593 	p->p_mqueue_cnt = psrc->p_mqueue_cnt;
2594 	memset(&p->p_specdataref, 0, sizeof(p->p_specdataref));
2595 	p->p_exitsig = psrc->p_exitsig;
2596 	p->p_flag = psrc->p_flag;
2597 	p->p_sflag = psrc->p_sflag;
2598 	p->p_slflag = psrc->p_slflag;
2599 	p->p_lflag = psrc->p_lflag;
2600 	p->p_stflag = psrc->p_stflag;
2601 	p->p_stat = psrc->p_stat;
2602 	p->p_trace_enabled = psrc->p_trace_enabled;
2603 	p->p_pid = psrc->p_pid;
2604 	COND_SET_STRUCT(p->p_pglist, psrc->p_pglist, allowaddr);
2605 	COND_SET_PTR(p->p_pptr, psrc->p_pptr, allowaddr);
2606 	COND_SET_STRUCT(p->p_sibling, psrc->p_sibling, allowaddr);
2607 	COND_SET_STRUCT(p->p_children, psrc->p_children, allowaddr);
2608 	COND_SET_STRUCT(p->p_lwps, psrc->p_lwps, allowaddr);
2609 	COND_SET_PTR(p->p_raslist, psrc->p_raslist, allowaddr);
2610 	p->p_nlwps = psrc->p_nlwps;
2611 	p->p_nzlwps = psrc->p_nzlwps;
2612 	p->p_nrlwps = psrc->p_nrlwps;
2613 	p->p_nlwpwait = psrc->p_nlwpwait;
2614 	p->p_ndlwps = psrc->p_ndlwps;
2615 	p->p_nstopchild = psrc->p_nstopchild;
2616 	p->p_waited = psrc->p_waited;
2617 	COND_SET_PTR(p->p_zomblwp, psrc->p_zomblwp, allowaddr);
2618 	COND_SET_PTR(p->p_vforklwp, psrc->p_vforklwp, allowaddr);
2619 	COND_SET_PTR(p->p_sched_info, psrc->p_sched_info, allowaddr);
2620 	p->p_estcpu = psrc->p_estcpu;
2621 	p->p_estcpu_inherited = psrc->p_estcpu_inherited;
2622 	p->p_forktime = psrc->p_forktime;
2623 	p->p_pctcpu = psrc->p_pctcpu;
2624 	COND_SET_PTR(p->p_opptr, psrc->p_opptr, allowaddr);
2625 	COND_SET_PTR(p->p_timers, psrc->p_timers, allowaddr);
2626 	p->p_rtime = psrc->p_rtime;
2627 	p->p_uticks = psrc->p_uticks;
2628 	p->p_sticks = psrc->p_sticks;
2629 	p->p_iticks = psrc->p_iticks;
2630 	p->p_xutime = psrc->p_xutime;
2631 	p->p_xstime = psrc->p_xstime;
2632 	p->p_traceflag = psrc->p_traceflag;
2633 	COND_SET_PTR(p->p_tracep, psrc->p_tracep, allowaddr);
2634 	COND_SET_PTR(p->p_textvp, psrc->p_textvp, allowaddr);
2635 	COND_SET_PTR(p->p_emul, psrc->p_emul, allowaddr);
2636 	COND_SET_PTR(p->p_emuldata, psrc->p_emuldata, allowaddr);
2637 	COND_SET_CPTR(p->p_execsw, psrc->p_execsw, allowaddr);
2638 	COND_SET_STRUCT(p->p_klist, psrc->p_klist, allowaddr);
2639 	COND_SET_STRUCT(p->p_sigwaiters, psrc->p_sigwaiters, allowaddr);
2640 	COND_SET_STRUCT(p->p_sigpend.sp_info, psrc->p_sigpend.sp_info,
2641 	    allowaddr);
2642 	p->p_sigpend.sp_set = psrc->p_sigpend.sp_set;
2643 	COND_SET_PTR(p->p_lwpctl, psrc->p_lwpctl, allowaddr);
2644 	p->p_ppid = psrc->p_ppid;
2645 	p->p_oppid = psrc->p_oppid;
2646 	COND_SET_PTR(p->p_path, psrc->p_path, allowaddr);
2647 	p->p_sigctx = psrc->p_sigctx;
2648 	p->p_nice = psrc->p_nice;
2649 	memcpy(p->p_comm, psrc->p_comm, sizeof(p->p_comm));
2650 	COND_SET_PTR(p->p_pgrp, psrc->p_pgrp, allowaddr);
2651 	COND_SET_VALUE(p->p_psstrp, psrc->p_psstrp, allowaddr);
2652 	p->p_pax = psrc->p_pax;
2653 	p->p_xexit = psrc->p_xexit;
2654 	p->p_xsig = psrc->p_xsig;
2655 	p->p_acflag = psrc->p_acflag;
2656 	COND_SET_STRUCT(p->p_md, psrc->p_md, allowaddr);
2657 	p->p_stackbase = psrc->p_stackbase;
2658 	COND_SET_PTR(p->p_dtrace, psrc->p_dtrace, allowaddr);
2659 }
2660 
2661 /*
2662  * Fill in an eproc structure for the specified process.
2663  */
2664 void
2665 fill_eproc(struct proc *p, struct eproc *ep, bool zombie, bool allowaddr)
2666 {
2667 	struct tty *tp;
2668 	struct lwp *l;
2669 
2670 	KASSERT(mutex_owned(&proc_lock));
2671 	KASSERT(mutex_owned(p->p_lock));
2672 
2673 	COND_SET_PTR(ep->e_paddr, p, allowaddr);
2674 	COND_SET_PTR(ep->e_sess, p->p_session, allowaddr);
2675 	if (p->p_cred) {
2676 		kauth_cred_topcred(p->p_cred, &ep->e_pcred);
2677 		kauth_cred_toucred(p->p_cred, &ep->e_ucred);
2678 	}
2679 	if (p->p_stat != SIDL && !P_ZOMBIE(p) && !zombie) {
2680 		struct vmspace *vm = p->p_vmspace;
2681 
2682 		ep->e_vm.vm_rssize = vm_resident_count(vm);
2683 		ep->e_vm.vm_tsize = vm->vm_tsize;
2684 		ep->e_vm.vm_dsize = vm->vm_dsize;
2685 		ep->e_vm.vm_ssize = vm->vm_ssize;
2686 		ep->e_vm.vm_map.size = vm->vm_map.size;
2687 
2688 		/* Pick the primary (first) LWP */
2689 		l = proc_active_lwp(p);
2690 		KASSERT(l != NULL);
2691 		lwp_lock(l);
2692 		if (l->l_wchan)
2693 			strncpy(ep->e_wmesg, l->l_wmesg, WMESGLEN);
2694 		lwp_unlock(l);
2695 	}
2696 	ep->e_ppid = p->p_ppid;
2697 	if (p->p_pgrp && p->p_session) {
2698 		ep->e_pgid = p->p_pgrp->pg_id;
2699 		ep->e_jobc = p->p_pgrp->pg_jobc;
2700 		ep->e_sid = p->p_session->s_sid;
2701 		if ((p->p_lflag & PL_CONTROLT) &&
2702 		    (tp = p->p_session->s_ttyp)) {
2703 			ep->e_tdev = tp->t_dev;
2704 			ep->e_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PGID;
2705 			COND_SET_PTR(ep->e_tsess, tp->t_session, allowaddr);
2706 		} else
2707 			ep->e_tdev = (uint32_t)NODEV;
2708 		ep->e_flag = p->p_session->s_ttyvp ? EPROC_CTTY : 0;
2709 		if (SESS_LEADER(p))
2710 			ep->e_flag |= EPROC_SLEADER;
2711 		strncpy(ep->e_login, p->p_session->s_login, MAXLOGNAME);
2712 	}
2713 	ep->e_xsize = ep->e_xrssize = 0;
2714 	ep->e_xccount = ep->e_xswrss = 0;
2715 }
2716 
2717 /*
2718  * Fill in a kinfo_proc2 structure for the specified process.
2719  */
2720 void
2721 fill_kproc2(struct proc *p, struct kinfo_proc2 *ki, bool zombie, bool allowaddr)
2722 {
2723 	struct tty *tp;
2724 	struct lwp *l, *l2;
2725 	struct timeval ut, st, rt;
2726 	sigset_t ss1, ss2;
2727 	struct rusage ru;
2728 	struct vmspace *vm;
2729 
2730 	KASSERT(mutex_owned(&proc_lock));
2731 	KASSERT(mutex_owned(p->p_lock));
2732 
2733 	sigemptyset(&ss1);
2734 	sigemptyset(&ss2);
2735 
2736 	COND_SET_VALUE(ki->p_paddr, PTRTOUINT64(p), allowaddr);
2737 	COND_SET_VALUE(ki->p_fd, PTRTOUINT64(p->p_fd), allowaddr);
2738 	COND_SET_VALUE(ki->p_cwdi, PTRTOUINT64(p->p_cwdi), allowaddr);
2739 	COND_SET_VALUE(ki->p_stats, PTRTOUINT64(p->p_stats), allowaddr);
2740 	COND_SET_VALUE(ki->p_limit, PTRTOUINT64(p->p_limit), allowaddr);
2741 	COND_SET_VALUE(ki->p_vmspace, PTRTOUINT64(p->p_vmspace), allowaddr);
2742 	COND_SET_VALUE(ki->p_sigacts, PTRTOUINT64(p->p_sigacts), allowaddr);
2743 	COND_SET_VALUE(ki->p_sess, PTRTOUINT64(p->p_session), allowaddr);
2744 	ki->p_tsess = 0;	/* may be changed if controlling tty below */
2745 	COND_SET_VALUE(ki->p_ru, PTRTOUINT64(&p->p_stats->p_ru), allowaddr);
2746 	ki->p_eflag = 0;
2747 	ki->p_exitsig = p->p_exitsig;
2748 	ki->p_flag = L_INMEM;   /* Process never swapped out */
2749 	ki->p_flag |= sysctl_map_flags(sysctl_flagmap, p->p_flag);
2750 	ki->p_flag |= sysctl_map_flags(sysctl_sflagmap, p->p_sflag);
2751 	ki->p_flag |= sysctl_map_flags(sysctl_slflagmap, p->p_slflag);
2752 	ki->p_flag |= sysctl_map_flags(sysctl_lflagmap, p->p_lflag);
2753 	ki->p_flag |= sysctl_map_flags(sysctl_stflagmap, p->p_stflag);
2754 	ki->p_pid = p->p_pid;
2755 	ki->p_ppid = p->p_ppid;
2756 	ki->p_uid = kauth_cred_geteuid(p->p_cred);
2757 	ki->p_ruid = kauth_cred_getuid(p->p_cred);
2758 	ki->p_gid = kauth_cred_getegid(p->p_cred);
2759 	ki->p_rgid = kauth_cred_getgid(p->p_cred);
2760 	ki->p_svuid = kauth_cred_getsvuid(p->p_cred);
2761 	ki->p_svgid = kauth_cred_getsvgid(p->p_cred);
2762 	ki->p_ngroups = kauth_cred_ngroups(p->p_cred);
2763 	kauth_cred_getgroups(p->p_cred, ki->p_groups,
2764 	    uimin(ki->p_ngroups, sizeof(ki->p_groups) / sizeof(ki->p_groups[0])),
2765 	    UIO_SYSSPACE);
2766 
2767 	ki->p_uticks = p->p_uticks;
2768 	ki->p_sticks = p->p_sticks;
2769 	ki->p_iticks = p->p_iticks;
2770 	ki->p_tpgid = NO_PGID;	/* may be changed if controlling tty below */
2771 	COND_SET_VALUE(ki->p_tracep, PTRTOUINT64(p->p_tracep), allowaddr);
2772 	ki->p_traceflag = p->p_traceflag;
2773 
2774 	memcpy(&ki->p_sigignore, &p->p_sigctx.ps_sigignore,sizeof(ki_sigset_t));
2775 	memcpy(&ki->p_sigcatch, &p->p_sigctx.ps_sigcatch, sizeof(ki_sigset_t));
2776 
2777 	ki->p_cpticks = 0;
2778 	ki->p_pctcpu = p->p_pctcpu;
2779 	ki->p_estcpu = 0;
2780 	ki->p_stat = p->p_stat; /* Will likely be overridden by LWP status */
2781 	ki->p_realstat = p->p_stat;
2782 	ki->p_nice = p->p_nice;
2783 	ki->p_xstat = P_WAITSTATUS(p);
2784 	ki->p_acflag = p->p_acflag;
2785 
2786 	strncpy(ki->p_comm, p->p_comm,
2787 	    uimin(sizeof(ki->p_comm), sizeof(p->p_comm)));
2788 	strncpy(ki->p_ename, p->p_emul->e_name, sizeof(ki->p_ename));
2789 
2790 	ki->p_nlwps = p->p_nlwps;
2791 	ki->p_realflag = ki->p_flag;
2792 
2793 	if (p->p_stat != SIDL && !P_ZOMBIE(p) && !zombie) {
2794 		vm = p->p_vmspace;
2795 		ki->p_vm_rssize = vm_resident_count(vm);
2796 		ki->p_vm_tsize = vm->vm_tsize;
2797 		ki->p_vm_dsize = vm->vm_dsize;
2798 		ki->p_vm_ssize = vm->vm_ssize;
2799 		ki->p_vm_vsize = atop(vm->vm_map.size);
2800 		/*
2801 		 * Since the stack is initially mapped mostly with
2802 		 * PROT_NONE and grown as needed, adjust the "mapped size"
2803 		 * to skip the unused stack portion.
2804 		 */
2805 		ki->p_vm_msize =
2806 		    atop(vm->vm_map.size) - vm->vm_issize + vm->vm_ssize;
2807 
2808 		/* Pick the primary (first) LWP */
2809 		l = proc_active_lwp(p);
2810 		KASSERT(l != NULL);
2811 		lwp_lock(l);
2812 		ki->p_nrlwps = p->p_nrlwps;
2813 		ki->p_forw = 0;
2814 		ki->p_back = 0;
2815 		COND_SET_VALUE(ki->p_addr, PTRTOUINT64(l->l_addr), allowaddr);
2816 		ki->p_stat = l->l_stat;
2817 		ki->p_flag |= sysctl_map_flags(sysctl_lwpflagmap, l->l_flag);
2818 		ki->p_swtime = l->l_swtime;
2819 		ki->p_slptime = l->l_slptime;
2820 		if (l->l_stat == LSONPROC)
2821 			ki->p_schedflags = l->l_cpu->ci_schedstate.spc_flags;
2822 		else
2823 			ki->p_schedflags = 0;
2824 		ki->p_priority = lwp_eprio(l);
2825 		ki->p_usrpri = l->l_priority;
2826 		if (l->l_wchan)
2827 			strncpy(ki->p_wmesg, l->l_wmesg, sizeof(ki->p_wmesg));
2828 		COND_SET_VALUE(ki->p_wchan, PTRTOUINT64(l->l_wchan), allowaddr);
2829 		ki->p_cpuid = cpu_index(l->l_cpu);
2830 		lwp_unlock(l);
2831 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
2832 			/* This is hardly correct, but... */
2833 			sigplusset(&l->l_sigpend.sp_set, &ss1);
2834 			sigplusset(&l->l_sigmask, &ss2);
2835 			ki->p_cpticks += l->l_cpticks;
2836 			ki->p_pctcpu += l->l_pctcpu;
2837 			ki->p_estcpu += l->l_estcpu;
2838 		}
2839 	}
2840 	sigplusset(&p->p_sigpend.sp_set, &ss1);
2841 	memcpy(&ki->p_siglist, &ss1, sizeof(ki_sigset_t));
2842 	memcpy(&ki->p_sigmask, &ss2, sizeof(ki_sigset_t));
2843 
2844 	if (p->p_session != NULL) {
2845 		ki->p_sid = p->p_session->s_sid;
2846 		ki->p__pgid = p->p_pgrp->pg_id;
2847 		if (p->p_session->s_ttyvp)
2848 			ki->p_eflag |= EPROC_CTTY;
2849 		if (SESS_LEADER(p))
2850 			ki->p_eflag |= EPROC_SLEADER;
2851 		strncpy(ki->p_login, p->p_session->s_login,
2852 		    uimin(sizeof ki->p_login - 1, sizeof p->p_session->s_login));
2853 		ki->p_jobc = p->p_pgrp->pg_jobc;
2854 		if ((p->p_lflag & PL_CONTROLT) && (tp = p->p_session->s_ttyp)) {
2855 			ki->p_tdev = tp->t_dev;
2856 			ki->p_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PGID;
2857 			COND_SET_VALUE(ki->p_tsess, PTRTOUINT64(tp->t_session),
2858 			    allowaddr);
2859 		} else {
2860 			ki->p_tdev = (int32_t)NODEV;
2861 		}
2862 	}
2863 
2864 	if (!P_ZOMBIE(p) && !zombie) {
2865 		ki->p_uvalid = 1;
2866 		ki->p_ustart_sec = p->p_stats->p_start.tv_sec;
2867 		ki->p_ustart_usec = p->p_stats->p_start.tv_usec;
2868 
2869 		calcru(p, &ut, &st, NULL, &rt);
2870 		ki->p_rtime_sec = rt.tv_sec;
2871 		ki->p_rtime_usec = rt.tv_usec;
2872 		ki->p_uutime_sec = ut.tv_sec;
2873 		ki->p_uutime_usec = ut.tv_usec;
2874 		ki->p_ustime_sec = st.tv_sec;
2875 		ki->p_ustime_usec = st.tv_usec;
2876 
2877 		memcpy(&ru, &p->p_stats->p_ru, sizeof(ru));
2878 		ki->p_uru_nvcsw = 0;
2879 		ki->p_uru_nivcsw = 0;
2880 		LIST_FOREACH(l2, &p->p_lwps, l_sibling) {
2881 			ki->p_uru_nvcsw += (l2->l_ncsw - l2->l_nivcsw);
2882 			ki->p_uru_nivcsw += l2->l_nivcsw;
2883 			ruadd(&ru, &l2->l_ru);
2884 		}
2885 		ki->p_uru_maxrss = ru.ru_maxrss;
2886 		ki->p_uru_ixrss = ru.ru_ixrss;
2887 		ki->p_uru_idrss = ru.ru_idrss;
2888 		ki->p_uru_isrss = ru.ru_isrss;
2889 		ki->p_uru_minflt = ru.ru_minflt;
2890 		ki->p_uru_majflt = ru.ru_majflt;
2891 		ki->p_uru_nswap = ru.ru_nswap;
2892 		ki->p_uru_inblock = ru.ru_inblock;
2893 		ki->p_uru_oublock = ru.ru_oublock;
2894 		ki->p_uru_msgsnd = ru.ru_msgsnd;
2895 		ki->p_uru_msgrcv = ru.ru_msgrcv;
2896 		ki->p_uru_nsignals = ru.ru_nsignals;
2897 
2898 		timeradd(&p->p_stats->p_cru.ru_utime,
2899 			 &p->p_stats->p_cru.ru_stime, &ut);
2900 		ki->p_uctime_sec = ut.tv_sec;
2901 		ki->p_uctime_usec = ut.tv_usec;
2902 	}
2903 }
2904 
2905 
2906 int
2907 proc_find_locked(struct lwp *l, struct proc **p, pid_t pid)
2908 {
2909 	int error;
2910 
2911 	mutex_enter(&proc_lock);
2912 	if (pid == -1)
2913 		*p = l->l_proc;
2914 	else
2915 		*p = proc_find(pid);
2916 
2917 	if (*p == NULL) {
2918 		if (pid != -1)
2919 			mutex_exit(&proc_lock);
2920 		return ESRCH;
2921 	}
2922 	if (pid != -1)
2923 		mutex_enter((*p)->p_lock);
2924 	mutex_exit(&proc_lock);
2925 
2926 	error = kauth_authorize_process(l->l_cred,
2927 	    KAUTH_PROCESS_CANSEE, *p,
2928 	    KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL);
2929 	if (error) {
2930 		if (pid != -1)
2931 			mutex_exit((*p)->p_lock);
2932 	}
2933 	return error;
2934 }
2935 
2936 static int
2937 fill_pathname(struct lwp *l, pid_t pid, void *oldp, size_t *oldlenp)
2938 {
2939 	int error;
2940 	struct proc *p;
2941 
2942 	if ((error = proc_find_locked(l, &p, pid)) != 0)
2943 		return error;
2944 
2945 	if (p->p_path == NULL) {
2946 		if (pid != -1)
2947 			mutex_exit(p->p_lock);
2948 		return ENOENT;
2949 	}
2950 
2951 	size_t len = strlen(p->p_path) + 1;
2952 	if (oldp != NULL) {
2953 		size_t copylen = uimin(len, *oldlenp);
2954 		error = sysctl_copyout(l, p->p_path, oldp, copylen);
2955 		if (error == 0 && *oldlenp < len)
2956 			error = ENOSPC;
2957 	}
2958 	*oldlenp = len;
2959 	if (pid != -1)
2960 		mutex_exit(p->p_lock);
2961 	return error;
2962 }
2963 
2964 static int
2965 fill_cwd(struct lwp *l, pid_t pid, void *oldp, size_t *oldlenp)
2966 {
2967 	int error;
2968 	struct proc *p;
2969 	char *path;
2970 	char *bp, *bend;
2971 	struct cwdinfo *cwdi;
2972 	struct vnode *vp;
2973 	size_t len, lenused;
2974 
2975 	if ((error = proc_find_locked(l, &p, pid)) != 0)
2976 		return error;
2977 
2978 	len = MAXPATHLEN * 4;
2979 
2980 	path = kmem_alloc(len, KM_SLEEP);
2981 
2982 	bp = &path[len];
2983 	bend = bp;
2984 	*(--bp) = '\0';
2985 
2986 	cwdi = p->p_cwdi;
2987 	rw_enter(&cwdi->cwdi_lock, RW_READER);
2988 	vp = cwdi->cwdi_cdir;
2989 	error = getcwd_common(vp, NULL, &bp, path, len/2, 0, l);
2990 	rw_exit(&cwdi->cwdi_lock);
2991 
2992 	if (error)
2993 		goto out;
2994 
2995 	lenused = bend - bp;
2996 
2997 	if (oldp != NULL) {
2998 		size_t copylen = uimin(lenused, *oldlenp);
2999 		error = sysctl_copyout(l, bp, oldp, copylen);
3000 		if (error == 0 && *oldlenp < lenused)
3001 			error = ENOSPC;
3002 	}
3003 	*oldlenp = lenused;
3004 out:
3005 	if (pid != -1)
3006 		mutex_exit(p->p_lock);
3007 	kmem_free(path, len);
3008 	return error;
3009 }
3010 
3011 int
3012 proc_getauxv(struct proc *p, void **buf, size_t *len)
3013 {
3014 	struct ps_strings pss;
3015 	int error;
3016 	void *uauxv, *kauxv;
3017 	size_t size;
3018 
3019 	if ((error = copyin_psstrings(p, &pss)) != 0)
3020 		return error;
3021 	if (pss.ps_envstr == NULL)
3022 		return EIO;
3023 
3024 	size = p->p_execsw->es_arglen;
3025 	if (size == 0)
3026 		return EIO;
3027 
3028 	size_t ptrsz = PROC_PTRSZ(p);
3029 	uauxv = (void *)((char *)pss.ps_envstr + (pss.ps_nenvstr + 1) * ptrsz);
3030 
3031 	kauxv = kmem_alloc(size, KM_SLEEP);
3032 
3033 	error = copyin_proc(p, uauxv, kauxv, size);
3034 	if (error) {
3035 		kmem_free(kauxv, size);
3036 		return error;
3037 	}
3038 
3039 	*buf = kauxv;
3040 	*len = size;
3041 
3042 	return 0;
3043 }
3044 
3045 
3046 static int
3047 sysctl_security_expose_address(SYSCTLFN_ARGS)
3048 {
3049 	int expose_address, error;
3050 	struct sysctlnode node;
3051 
3052 	node = *rnode;
3053 	node.sysctl_data = &expose_address;
3054 	expose_address = *(int *)rnode->sysctl_data;
3055 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
3056 	if (error || newp == NULL)
3057 		return error;
3058 
3059 	if (kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_KERNADDR,
3060 	    0, NULL, NULL, NULL))
3061 		return EPERM;
3062 
3063 	switch (expose_address) {
3064 	case 0:
3065 	case 1:
3066 	case 2:
3067 		break;
3068 	default:
3069 		return EINVAL;
3070 	}
3071 
3072 	*(int *)rnode->sysctl_data = expose_address;
3073 
3074 	return 0;
3075 }
3076 
3077 bool
3078 get_expose_address(struct proc *p)
3079 {
3080 	/* allow only if sysctl variable is set or privileged */
3081 	return kauth_authorize_process(kauth_cred_get(), KAUTH_PROCESS_CANSEE,
3082 	    p, KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_KPTR), NULL, NULL) == 0;
3083 }
3084