xref: /netbsd-src/sys/kern/kern_proc.c (revision 9ddb6ab554e70fb9bbd90c3d96b812bc57755a14)
1 /*	$NetBSD: kern_proc.c,v 1.182 2012/02/19 21:06:52 rmind Exp $	*/
2 
3 /*-
4  * Copyright (c) 1999, 2006, 2007, 2008 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 1.182 2012/02/19 21:06:52 rmind 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 #endif
73 
74 #include <sys/param.h>
75 #include <sys/systm.h>
76 #include <sys/kernel.h>
77 #include <sys/proc.h>
78 #include <sys/resourcevar.h>
79 #include <sys/buf.h>
80 #include <sys/acct.h>
81 #include <sys/wait.h>
82 #include <sys/file.h>
83 #include <ufs/ufs/quota.h>
84 #include <sys/uio.h>
85 #include <sys/pool.h>
86 #include <sys/pset.h>
87 #include <sys/mbuf.h>
88 #include <sys/ioctl.h>
89 #include <sys/tty.h>
90 #include <sys/signalvar.h>
91 #include <sys/ras.h>
92 #include <sys/filedesc.h>
93 #include "sys/syscall_stats.h"
94 #include <sys/kauth.h>
95 #include <sys/sleepq.h>
96 #include <sys/atomic.h>
97 #include <sys/kmem.h>
98 #include <sys/dtrace_bsd.h>
99 #include <sys/sysctl.h>
100 #include <sys/exec.h>
101 #include <sys/cpu.h>
102 
103 #include <uvm/uvm_extern.h>
104 #include <uvm/uvm_extern.h>
105 
106 #ifdef COMPAT_NETBSD32
107 #include <compat/netbsd32/netbsd32.h>
108 #endif
109 
110 /*
111  * Process lists.
112  */
113 
114 struct proclist		allproc		__cacheline_aligned;
115 struct proclist		zombproc	__cacheline_aligned;
116 
117 kmutex_t *		proc_lock	__cacheline_aligned;
118 
119 /*
120  * pid to proc lookup is done by indexing the pid_table array.
121  * Since pid numbers are only allocated when an empty slot
122  * has been found, there is no need to search any lists ever.
123  * (an orphaned pgrp will lock the slot, a session will lock
124  * the pgrp with the same number.)
125  * If the table is too small it is reallocated with twice the
126  * previous size and the entries 'unzipped' into the two halves.
127  * A linked list of free entries is passed through the pt_proc
128  * field of 'free' items - set odd to be an invalid ptr.
129  */
130 
131 struct pid_table {
132 	struct proc	*pt_proc;
133 	struct pgrp	*pt_pgrp;
134 	pid_t		pt_pid;
135 };
136 #if 1	/* strongly typed cast - should be a noop */
137 static inline uint p2u(struct proc *p) { return (uint)(uintptr_t)p; }
138 #else
139 #define p2u(p) ((uint)p)
140 #endif
141 #define P_VALID(p) (!(p2u(p) & 1))
142 #define P_NEXT(p) (p2u(p) >> 1)
143 #define P_FREE(pid) ((struct proc *)(uintptr_t)((pid) << 1 | 1))
144 
145 /*
146  * Table of process IDs (PIDs).
147  */
148 static struct pid_table *pid_table	__read_mostly;
149 
150 #define	INITIAL_PID_TABLE_SIZE		(1 << 5)
151 
152 /* Table mask, threshold for growing and number of allocated PIDs. */
153 static u_int		pid_tbl_mask	__read_mostly;
154 static u_int		pid_alloc_lim	__read_mostly;
155 static u_int		pid_alloc_cnt	__cacheline_aligned;
156 
157 /* Next free, last free and maximum PIDs. */
158 static u_int		next_free_pt	__cacheline_aligned;
159 static u_int		last_free_pt	__cacheline_aligned;
160 static pid_t		pid_max		__read_mostly;
161 
162 /* Components of the first process -- never freed. */
163 
164 extern struct emul emul_netbsd;	/* defined in kern_exec.c */
165 
166 struct session session0 = {
167 	.s_count = 1,
168 	.s_sid = 0,
169 };
170 struct pgrp pgrp0 = {
171 	.pg_members = LIST_HEAD_INITIALIZER(&pgrp0.pg_members),
172 	.pg_session = &session0,
173 };
174 filedesc_t filedesc0;
175 struct cwdinfo cwdi0 = {
176 	.cwdi_cmask = CMASK,		/* see cmask below */
177 	.cwdi_refcnt = 1,
178 };
179 struct plimit limit0;
180 struct pstats pstat0;
181 struct vmspace vmspace0;
182 struct sigacts sigacts0;
183 struct proc proc0 = {
184 	.p_lwps = LIST_HEAD_INITIALIZER(&proc0.p_lwps),
185 	.p_sigwaiters = LIST_HEAD_INITIALIZER(&proc0.p_sigwaiters),
186 	.p_nlwps = 1,
187 	.p_nrlwps = 1,
188 	.p_nlwpid = 1,		/* must match lwp0.l_lid */
189 	.p_pgrp = &pgrp0,
190 	.p_comm = "system",
191 	/*
192 	 * Set P_NOCLDWAIT so that kernel threads are reparented to init(8)
193 	 * when they exit.  init(8) can easily wait them out for us.
194 	 */
195 	.p_flag = PK_SYSTEM | PK_NOCLDWAIT,
196 	.p_stat = SACTIVE,
197 	.p_nice = NZERO,
198 	.p_emul = &emul_netbsd,
199 	.p_cwdi = &cwdi0,
200 	.p_limit = &limit0,
201 	.p_fd = &filedesc0,
202 	.p_vmspace = &vmspace0,
203 	.p_stats = &pstat0,
204 	.p_sigacts = &sigacts0,
205 };
206 kauth_cred_t cred0;
207 
208 static const int	nofile	= NOFILE;
209 static const int	maxuprc	= MAXUPRC;
210 static const int	cmask	= CMASK;
211 
212 static int sysctl_doeproc(SYSCTLFN_PROTO);
213 static int sysctl_kern_proc_args(SYSCTLFN_PROTO);
214 static void fill_kproc2(struct proc *, struct kinfo_proc2 *, bool);
215 
216 /*
217  * The process list descriptors, used during pid allocation and
218  * by sysctl.  No locking on this data structure is needed since
219  * it is completely static.
220  */
221 const struct proclist_desc proclists[] = {
222 	{ &allproc	},
223 	{ &zombproc	},
224 	{ NULL		},
225 };
226 
227 static struct pgrp *	pg_remove(pid_t);
228 static void		pg_delete(pid_t);
229 static void		orphanpg(struct pgrp *);
230 
231 static specificdata_domain_t proc_specificdata_domain;
232 
233 static pool_cache_t proc_cache;
234 
235 static kauth_listener_t proc_listener;
236 
237 static int
238 proc_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
239     void *arg0, void *arg1, void *arg2, void *arg3)
240 {
241 	struct proc *p;
242 	int result;
243 
244 	result = KAUTH_RESULT_DEFER;
245 	p = arg0;
246 
247 	switch (action) {
248 	case KAUTH_PROCESS_CANSEE: {
249 		enum kauth_process_req req;
250 
251 		req = (enum kauth_process_req)arg1;
252 
253 		switch (req) {
254 		case KAUTH_REQ_PROCESS_CANSEE_ARGS:
255 		case KAUTH_REQ_PROCESS_CANSEE_ENTRY:
256 		case KAUTH_REQ_PROCESS_CANSEE_OPENFILES:
257 			result = KAUTH_RESULT_ALLOW;
258 
259 			break;
260 
261 		case KAUTH_REQ_PROCESS_CANSEE_ENV:
262 			if (kauth_cred_getuid(cred) !=
263 			    kauth_cred_getuid(p->p_cred) ||
264 			    kauth_cred_getuid(cred) !=
265 			    kauth_cred_getsvuid(p->p_cred))
266 				break;
267 
268 			result = KAUTH_RESULT_ALLOW;
269 
270 			break;
271 
272 		default:
273 			break;
274 		}
275 
276 		break;
277 		}
278 
279 	case KAUTH_PROCESS_FORK: {
280 		int lnprocs = (int)(unsigned long)arg2;
281 
282 		/*
283 		 * Don't allow a nonprivileged user to use the last few
284 		 * processes. The variable lnprocs is the current number of
285 		 * processes, maxproc is the limit.
286 		 */
287 		if (__predict_false((lnprocs >= maxproc - 5)))
288 			break;
289 
290 		result = KAUTH_RESULT_ALLOW;
291 
292 		break;
293 		}
294 
295 	case KAUTH_PROCESS_CORENAME:
296 	case KAUTH_PROCESS_STOPFLAG:
297 		if (proc_uidmatch(cred, p->p_cred) == 0)
298 			result = KAUTH_RESULT_ALLOW;
299 
300 		break;
301 
302 	default:
303 		break;
304 	}
305 
306 	return result;
307 }
308 
309 /*
310  * Initialize global process hashing structures.
311  */
312 void
313 procinit(void)
314 {
315 	const struct proclist_desc *pd;
316 	u_int i;
317 #define	LINK_EMPTY ((PID_MAX + INITIAL_PID_TABLE_SIZE) & ~(INITIAL_PID_TABLE_SIZE - 1))
318 
319 	for (pd = proclists; pd->pd_list != NULL; pd++)
320 		LIST_INIT(pd->pd_list);
321 
322 	proc_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
323 	pid_table = kmem_alloc(INITIAL_PID_TABLE_SIZE
324 	    * sizeof(struct pid_table), KM_SLEEP);
325 	pid_tbl_mask = INITIAL_PID_TABLE_SIZE - 1;
326 	pid_max = PID_MAX;
327 
328 	/* Set free list running through table...
329 	   Preset 'use count' above PID_MAX so we allocate pid 1 next. */
330 	for (i = 0; i <= pid_tbl_mask; i++) {
331 		pid_table[i].pt_proc = P_FREE(LINK_EMPTY + i + 1);
332 		pid_table[i].pt_pgrp = 0;
333 		pid_table[i].pt_pid = 0;
334 	}
335 	/* slot 0 is just grabbed */
336 	next_free_pt = 1;
337 	/* Need to fix last entry. */
338 	last_free_pt = pid_tbl_mask;
339 	pid_table[last_free_pt].pt_proc = P_FREE(LINK_EMPTY);
340 	/* point at which we grow table - to avoid reusing pids too often */
341 	pid_alloc_lim = pid_tbl_mask - 1;
342 #undef LINK_EMPTY
343 
344 	proc_specificdata_domain = specificdata_domain_create();
345 	KASSERT(proc_specificdata_domain != NULL);
346 
347 	proc_cache = pool_cache_init(sizeof(struct proc), 0, 0, 0,
348 	    "procpl", NULL, IPL_NONE, NULL, NULL, NULL);
349 
350 	proc_listener = kauth_listen_scope(KAUTH_SCOPE_PROCESS,
351 	    proc_listener_cb, NULL);
352 }
353 
354 void
355 procinit_sysctl(void)
356 {
357 	static struct sysctllog *clog;
358 
359 	sysctl_createv(&clog, 0, NULL, NULL,
360 		       CTLFLAG_PERMANENT,
361 		       CTLTYPE_NODE, "kern", NULL,
362 		       NULL, 0, NULL, 0,
363 		       CTL_KERN, CTL_EOL);
364 
365 	sysctl_createv(&clog, 0, NULL, NULL,
366 		       CTLFLAG_PERMANENT,
367 		       CTLTYPE_NODE, "proc",
368 		       SYSCTL_DESCR("System-wide process information"),
369 		       sysctl_doeproc, 0, NULL, 0,
370 		       CTL_KERN, KERN_PROC, CTL_EOL);
371 	sysctl_createv(&clog, 0, NULL, NULL,
372 		       CTLFLAG_PERMANENT,
373 		       CTLTYPE_NODE, "proc2",
374 		       SYSCTL_DESCR("Machine-independent process information"),
375 		       sysctl_doeproc, 0, NULL, 0,
376 		       CTL_KERN, KERN_PROC2, CTL_EOL);
377 	sysctl_createv(&clog, 0, NULL, NULL,
378 		       CTLFLAG_PERMANENT,
379 		       CTLTYPE_NODE, "proc_args",
380 		       SYSCTL_DESCR("Process argument information"),
381 		       sysctl_kern_proc_args, 0, NULL, 0,
382 		       CTL_KERN, KERN_PROC_ARGS, CTL_EOL);
383 
384 	/*
385 	  "nodes" under these:
386 
387 	  KERN_PROC_ALL
388 	  KERN_PROC_PID pid
389 	  KERN_PROC_PGRP pgrp
390 	  KERN_PROC_SESSION sess
391 	  KERN_PROC_TTY tty
392 	  KERN_PROC_UID uid
393 	  KERN_PROC_RUID uid
394 	  KERN_PROC_GID gid
395 	  KERN_PROC_RGID gid
396 
397 	  all in all, probably not worth the effort...
398 	*/
399 }
400 
401 /*
402  * Initialize process 0.
403  */
404 void
405 proc0_init(void)
406 {
407 	struct proc *p;
408 	struct pgrp *pg;
409 	struct rlimit *rlim;
410 	rlim_t lim;
411 	int i;
412 
413 	p = &proc0;
414 	pg = &pgrp0;
415 
416 	mutex_init(&p->p_stmutex, MUTEX_DEFAULT, IPL_HIGH);
417 	mutex_init(&p->p_auxlock, MUTEX_DEFAULT, IPL_NONE);
418 	p->p_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
419 
420 	rw_init(&p->p_reflock);
421 	cv_init(&p->p_waitcv, "wait");
422 	cv_init(&p->p_lwpcv, "lwpwait");
423 
424 	LIST_INSERT_HEAD(&p->p_lwps, &lwp0, l_sibling);
425 
426 	pid_table[0].pt_proc = p;
427 	LIST_INSERT_HEAD(&allproc, p, p_list);
428 
429 	pid_table[0].pt_pgrp = pg;
430 	LIST_INSERT_HEAD(&pg->pg_members, p, p_pglist);
431 
432 #ifdef __HAVE_SYSCALL_INTERN
433 	(*p->p_emul->e_syscall_intern)(p);
434 #endif
435 
436 	/* Create credentials. */
437 	cred0 = kauth_cred_alloc();
438 	p->p_cred = cred0;
439 
440 	/* Create the CWD info. */
441 	rw_init(&cwdi0.cwdi_lock);
442 
443 	/* Create the limits structures. */
444 	mutex_init(&limit0.pl_lock, MUTEX_DEFAULT, IPL_NONE);
445 
446 	rlim = limit0.pl_rlimit;
447 	for (i = 0; i < __arraycount(limit0.pl_rlimit); i++) {
448 		rlim[i].rlim_cur = RLIM_INFINITY;
449 		rlim[i].rlim_max = RLIM_INFINITY;
450 	}
451 
452 	rlim[RLIMIT_NOFILE].rlim_max = maxfiles;
453 	rlim[RLIMIT_NOFILE].rlim_cur = maxfiles < nofile ? maxfiles : nofile;
454 
455 	rlim[RLIMIT_NPROC].rlim_max = maxproc;
456 	rlim[RLIMIT_NPROC].rlim_cur = maxproc < maxuprc ? maxproc : maxuprc;
457 
458 	lim = MIN(VM_MAXUSER_ADDRESS, ctob((rlim_t)uvmexp.free));
459 	rlim[RLIMIT_RSS].rlim_max = lim;
460 	rlim[RLIMIT_MEMLOCK].rlim_max = lim;
461 	rlim[RLIMIT_MEMLOCK].rlim_cur = lim / 3;
462 
463 	/* Note that default core name has zero length. */
464 	limit0.pl_corename = defcorename;
465 	limit0.pl_cnlen = 0;
466 	limit0.pl_refcnt = 1;
467 	limit0.pl_writeable = false;
468 	limit0.pl_sv_limit = NULL;
469 
470 	/* Configure virtual memory system, set vm rlimits. */
471 	uvm_init_limits(p);
472 
473 	/* Initialize file descriptor table for proc0. */
474 	fd_init(&filedesc0);
475 
476 	/*
477 	 * Initialize proc0's vmspace, which uses the kernel pmap.
478 	 * All kernel processes (which never have user space mappings)
479 	 * share proc0's vmspace, and thus, the kernel pmap.
480 	 */
481 	uvmspace_init(&vmspace0, pmap_kernel(), round_page(VM_MIN_ADDRESS),
482 	    trunc_page(VM_MAX_ADDRESS));
483 
484 	/* Initialize signal state for proc0. XXX IPL_SCHED */
485 	mutex_init(&p->p_sigacts->sa_mutex, MUTEX_DEFAULT, IPL_SCHED);
486 	siginit(p);
487 
488 	proc_initspecific(p);
489 	kdtrace_proc_ctor(NULL, p);
490 }
491 
492 /*
493  * Session reference counting.
494  */
495 
496 void
497 proc_sesshold(struct session *ss)
498 {
499 
500 	KASSERT(mutex_owned(proc_lock));
501 	ss->s_count++;
502 }
503 
504 void
505 proc_sessrele(struct session *ss)
506 {
507 
508 	KASSERT(mutex_owned(proc_lock));
509 	/*
510 	 * We keep the pgrp with the same id as the session in order to
511 	 * stop a process being given the same pid.  Since the pgrp holds
512 	 * a reference to the session, it must be a 'zombie' pgrp by now.
513 	 */
514 	if (--ss->s_count == 0) {
515 		struct pgrp *pg;
516 
517 		pg = pg_remove(ss->s_sid);
518 		mutex_exit(proc_lock);
519 
520 		kmem_free(pg, sizeof(struct pgrp));
521 		kmem_free(ss, sizeof(struct session));
522 	} else {
523 		mutex_exit(proc_lock);
524 	}
525 }
526 
527 /*
528  * Check that the specified process group is in the session of the
529  * specified process.
530  * Treats -ve ids as process ids.
531  * Used to validate TIOCSPGRP requests.
532  */
533 int
534 pgid_in_session(struct proc *p, pid_t pg_id)
535 {
536 	struct pgrp *pgrp;
537 	struct session *session;
538 	int error;
539 
540 	mutex_enter(proc_lock);
541 	if (pg_id < 0) {
542 		struct proc *p1 = proc_find(-pg_id);
543 		if (p1 == NULL) {
544 			error = EINVAL;
545 			goto fail;
546 		}
547 		pgrp = p1->p_pgrp;
548 	} else {
549 		pgrp = pgrp_find(pg_id);
550 		if (pgrp == NULL) {
551 			error = EINVAL;
552 			goto fail;
553 		}
554 	}
555 	session = pgrp->pg_session;
556 	error = (session != p->p_pgrp->pg_session) ? EPERM : 0;
557 fail:
558 	mutex_exit(proc_lock);
559 	return error;
560 }
561 
562 /*
563  * p_inferior: is p an inferior of q?
564  */
565 static inline bool
566 p_inferior(struct proc *p, struct proc *q)
567 {
568 
569 	KASSERT(mutex_owned(proc_lock));
570 
571 	for (; p != q; p = p->p_pptr)
572 		if (p->p_pid == 0)
573 			return false;
574 	return true;
575 }
576 
577 /*
578  * proc_find: locate a process by the ID.
579  *
580  * => Must be called with proc_lock held.
581  */
582 proc_t *
583 proc_find_raw(pid_t pid)
584 {
585 	struct pid_table *pt;
586 	proc_t *p;
587 
588 	KASSERT(mutex_owned(proc_lock));
589 	pt = &pid_table[pid & pid_tbl_mask];
590 	p = pt->pt_proc;
591 	if (__predict_false(!P_VALID(p) || pt->pt_pid != pid)) {
592 		return NULL;
593 	}
594 	return p;
595 }
596 
597 proc_t *
598 proc_find(pid_t pid)
599 {
600 	proc_t *p;
601 
602 	p = proc_find_raw(pid);
603 	if (__predict_false(p == NULL)) {
604 		return NULL;
605 	}
606 
607 	/*
608 	 * Only allow live processes to be found by PID.
609 	 * XXX: p_stat might change, since unlocked.
610 	 */
611 	if (__predict_true(p->p_stat == SACTIVE || p->p_stat == SSTOP)) {
612 		return p;
613 	}
614 	return NULL;
615 }
616 
617 /*
618  * pgrp_find: locate a process group by the ID.
619  *
620  * => Must be called with proc_lock held.
621  */
622 struct pgrp *
623 pgrp_find(pid_t pgid)
624 {
625 	struct pgrp *pg;
626 
627 	KASSERT(mutex_owned(proc_lock));
628 
629 	pg = pid_table[pgid & pid_tbl_mask].pt_pgrp;
630 
631 	/*
632 	 * Cannot look up a process group that only exists because the
633 	 * session has not died yet (traditional).
634 	 */
635 	if (pg == NULL || pg->pg_id != pgid || LIST_EMPTY(&pg->pg_members)) {
636 		return NULL;
637 	}
638 	return pg;
639 }
640 
641 static void
642 expand_pid_table(void)
643 {
644 	size_t pt_size, tsz;
645 	struct pid_table *n_pt, *new_pt;
646 	struct proc *proc;
647 	struct pgrp *pgrp;
648 	pid_t pid, rpid;
649 	u_int i;
650 	uint new_pt_mask;
651 
652 	pt_size = pid_tbl_mask + 1;
653 	tsz = pt_size * 2 * sizeof(struct pid_table);
654 	new_pt = kmem_alloc(tsz, KM_SLEEP);
655 	new_pt_mask = pt_size * 2 - 1;
656 
657 	mutex_enter(proc_lock);
658 	if (pt_size != pid_tbl_mask + 1) {
659 		/* Another process beat us to it... */
660 		mutex_exit(proc_lock);
661 		kmem_free(new_pt, tsz);
662 		return;
663 	}
664 
665 	/*
666 	 * Copy entries from old table into new one.
667 	 * If 'pid' is 'odd' we need to place in the upper half,
668 	 * even pid's to the lower half.
669 	 * Free items stay in the low half so we don't have to
670 	 * fixup the reference to them.
671 	 * We stuff free items on the front of the freelist
672 	 * because we can't write to unmodified entries.
673 	 * Processing the table backwards maintains a semblance
674 	 * of issuing pid numbers that increase with time.
675 	 */
676 	i = pt_size - 1;
677 	n_pt = new_pt + i;
678 	for (; ; i--, n_pt--) {
679 		proc = pid_table[i].pt_proc;
680 		pgrp = pid_table[i].pt_pgrp;
681 		if (!P_VALID(proc)) {
682 			/* Up 'use count' so that link is valid */
683 			pid = (P_NEXT(proc) + pt_size) & ~pt_size;
684 			rpid = 0;
685 			proc = P_FREE(pid);
686 			if (pgrp)
687 				pid = pgrp->pg_id;
688 		} else {
689 			pid = pid_table[i].pt_pid;
690 			rpid = pid;
691 		}
692 
693 		/* Save entry in appropriate half of table */
694 		n_pt[pid & pt_size].pt_proc = proc;
695 		n_pt[pid & pt_size].pt_pgrp = pgrp;
696 		n_pt[pid & pt_size].pt_pid = rpid;
697 
698 		/* Put other piece on start of free list */
699 		pid = (pid ^ pt_size) & ~pid_tbl_mask;
700 		n_pt[pid & pt_size].pt_proc =
701 			P_FREE((pid & ~pt_size) | next_free_pt);
702 		n_pt[pid & pt_size].pt_pgrp = 0;
703 		n_pt[pid & pt_size].pt_pid = 0;
704 
705 		next_free_pt = i | (pid & pt_size);
706 		if (i == 0)
707 			break;
708 	}
709 
710 	/* Save old table size and switch tables */
711 	tsz = pt_size * sizeof(struct pid_table);
712 	n_pt = pid_table;
713 	pid_table = new_pt;
714 	pid_tbl_mask = new_pt_mask;
715 
716 	/*
717 	 * pid_max starts as PID_MAX (= 30000), once we have 16384
718 	 * allocated pids we need it to be larger!
719 	 */
720 	if (pid_tbl_mask > PID_MAX) {
721 		pid_max = pid_tbl_mask * 2 + 1;
722 		pid_alloc_lim |= pid_alloc_lim << 1;
723 	} else
724 		pid_alloc_lim <<= 1;	/* doubles number of free slots... */
725 
726 	mutex_exit(proc_lock);
727 	kmem_free(n_pt, tsz);
728 }
729 
730 struct proc *
731 proc_alloc(void)
732 {
733 	struct proc *p;
734 
735 	p = pool_cache_get(proc_cache, PR_WAITOK);
736 	p->p_stat = SIDL;			/* protect against others */
737 	proc_initspecific(p);
738 	kdtrace_proc_ctor(NULL, p);
739 	p->p_pid = -1;
740 	proc_alloc_pid(p);
741 	return p;
742 }
743 
744 pid_t
745 proc_alloc_pid(struct proc *p)
746 {
747 	struct pid_table *pt;
748 	pid_t pid;
749 	int nxt;
750 
751 	for (;;expand_pid_table()) {
752 		if (__predict_false(pid_alloc_cnt >= pid_alloc_lim))
753 			/* ensure pids cycle through 2000+ values */
754 			continue;
755 		mutex_enter(proc_lock);
756 		pt = &pid_table[next_free_pt];
757 #ifdef DIAGNOSTIC
758 		if (__predict_false(P_VALID(pt->pt_proc) || pt->pt_pgrp))
759 			panic("proc_alloc: slot busy");
760 #endif
761 		nxt = P_NEXT(pt->pt_proc);
762 		if (nxt & pid_tbl_mask)
763 			break;
764 		/* Table full - expand (NB last entry not used....) */
765 		mutex_exit(proc_lock);
766 	}
767 
768 	/* pid is 'saved use count' + 'size' + entry */
769 	pid = (nxt & ~pid_tbl_mask) + pid_tbl_mask + 1 + next_free_pt;
770 	if ((uint)pid > (uint)pid_max)
771 		pid &= pid_tbl_mask;
772 	next_free_pt = nxt & pid_tbl_mask;
773 
774 	/* Grab table slot */
775 	pt->pt_proc = p;
776 
777 	KASSERT(pt->pt_pid == 0);
778 	pt->pt_pid = pid;
779 	if (p->p_pid == -1) {
780 		p->p_pid = pid;
781 	}
782 	pid_alloc_cnt++;
783 	mutex_exit(proc_lock);
784 
785 	return pid;
786 }
787 
788 /*
789  * Free a process id - called from proc_free (in kern_exit.c)
790  *
791  * Called with the proc_lock held.
792  */
793 void
794 proc_free_pid(pid_t pid)
795 {
796 	struct pid_table *pt;
797 
798 	KASSERT(mutex_owned(proc_lock));
799 
800 	pt = &pid_table[pid & pid_tbl_mask];
801 
802 	/* save pid use count in slot */
803 	pt->pt_proc = P_FREE(pid & ~pid_tbl_mask);
804 	KASSERT(pt->pt_pid == pid);
805 	pt->pt_pid = 0;
806 
807 	if (pt->pt_pgrp == NULL) {
808 		/* link last freed entry onto ours */
809 		pid &= pid_tbl_mask;
810 		pt = &pid_table[last_free_pt];
811 		pt->pt_proc = P_FREE(P_NEXT(pt->pt_proc) | pid);
812 		pt->pt_pid = 0;
813 		last_free_pt = pid;
814 		pid_alloc_cnt--;
815 	}
816 
817 	atomic_dec_uint(&nprocs);
818 }
819 
820 void
821 proc_free_mem(struct proc *p)
822 {
823 
824 	kdtrace_proc_dtor(NULL, p);
825 	pool_cache_put(proc_cache, p);
826 }
827 
828 /*
829  * proc_enterpgrp: move p to a new or existing process group (and session).
830  *
831  * If we are creating a new pgrp, the pgid should equal
832  * the calling process' pid.
833  * If is only valid to enter a process group that is in the session
834  * of the process.
835  * Also mksess should only be set if we are creating a process group
836  *
837  * Only called from sys_setsid, sys_setpgid and posix_spawn/spawn_return.
838  */
839 int
840 proc_enterpgrp(struct proc *curp, pid_t pid, pid_t pgid, bool mksess)
841 {
842 	struct pgrp *new_pgrp, *pgrp;
843 	struct session *sess;
844 	struct proc *p;
845 	int rval;
846 	pid_t pg_id = NO_PGID;
847 
848 	sess = mksess ? kmem_alloc(sizeof(*sess), KM_SLEEP) : NULL;
849 
850 	/* Allocate data areas we might need before doing any validity checks */
851 	mutex_enter(proc_lock);		/* Because pid_table might change */
852 	if (pid_table[pgid & pid_tbl_mask].pt_pgrp == 0) {
853 		mutex_exit(proc_lock);
854 		new_pgrp = kmem_alloc(sizeof(*new_pgrp), KM_SLEEP);
855 		mutex_enter(proc_lock);
856 	} else
857 		new_pgrp = NULL;
858 	rval = EPERM;	/* most common error (to save typing) */
859 
860 	/* Check pgrp exists or can be created */
861 	pgrp = pid_table[pgid & pid_tbl_mask].pt_pgrp;
862 	if (pgrp != NULL && pgrp->pg_id != pgid)
863 		goto done;
864 
865 	/* Can only set another process under restricted circumstances. */
866 	if (pid != curp->p_pid) {
867 		/* Must exist and be one of our children... */
868 		p = proc_find(pid);
869 		if (p == NULL || !p_inferior(p, curp)) {
870 			rval = ESRCH;
871 			goto done;
872 		}
873 		/* ... in the same session... */
874 		if (sess != NULL || p->p_session != curp->p_session)
875 			goto done;
876 		/* ... existing pgid must be in same session ... */
877 		if (pgrp != NULL && pgrp->pg_session != p->p_session)
878 			goto done;
879 		/* ... and not done an exec. */
880 		if (p->p_flag & PK_EXEC) {
881 			rval = EACCES;
882 			goto done;
883 		}
884 	} else {
885 		/* ... setsid() cannot re-enter a pgrp */
886 		if (mksess && (curp->p_pgid == curp->p_pid ||
887 		    pgrp_find(curp->p_pid)))
888 			goto done;
889 		p = curp;
890 	}
891 
892 	/* Changing the process group/session of a session
893 	   leader is definitely off limits. */
894 	if (SESS_LEADER(p)) {
895 		if (sess == NULL && p->p_pgrp == pgrp)
896 			/* unless it's a definite noop */
897 			rval = 0;
898 		goto done;
899 	}
900 
901 	/* Can only create a process group with id of process */
902 	if (pgrp == NULL && pgid != pid)
903 		goto done;
904 
905 	/* Can only create a session if creating pgrp */
906 	if (sess != NULL && pgrp != NULL)
907 		goto done;
908 
909 	/* Check we allocated memory for a pgrp... */
910 	if (pgrp == NULL && new_pgrp == NULL)
911 		goto done;
912 
913 	/* Don't attach to 'zombie' pgrp */
914 	if (pgrp != NULL && LIST_EMPTY(&pgrp->pg_members))
915 		goto done;
916 
917 	/* Expect to succeed now */
918 	rval = 0;
919 
920 	if (pgrp == p->p_pgrp)
921 		/* nothing to do */
922 		goto done;
923 
924 	/* Ok all setup, link up required structures */
925 
926 	if (pgrp == NULL) {
927 		pgrp = new_pgrp;
928 		new_pgrp = NULL;
929 		if (sess != NULL) {
930 			sess->s_sid = p->p_pid;
931 			sess->s_leader = p;
932 			sess->s_count = 1;
933 			sess->s_ttyvp = NULL;
934 			sess->s_ttyp = NULL;
935 			sess->s_flags = p->p_session->s_flags & ~S_LOGIN_SET;
936 			memcpy(sess->s_login, p->p_session->s_login,
937 			    sizeof(sess->s_login));
938 			p->p_lflag &= ~PL_CONTROLT;
939 		} else {
940 			sess = p->p_pgrp->pg_session;
941 			proc_sesshold(sess);
942 		}
943 		pgrp->pg_session = sess;
944 		sess = NULL;
945 
946 		pgrp->pg_id = pgid;
947 		LIST_INIT(&pgrp->pg_members);
948 #ifdef DIAGNOSTIC
949 		if (__predict_false(pid_table[pgid & pid_tbl_mask].pt_pgrp))
950 			panic("enterpgrp: pgrp table slot in use");
951 		if (__predict_false(mksess && p != curp))
952 			panic("enterpgrp: mksession and p != curproc");
953 #endif
954 		pid_table[pgid & pid_tbl_mask].pt_pgrp = pgrp;
955 		pgrp->pg_jobc = 0;
956 	}
957 
958 	/*
959 	 * Adjust eligibility of affected pgrps to participate in job control.
960 	 * Increment eligibility counts before decrementing, otherwise we
961 	 * could reach 0 spuriously during the first call.
962 	 */
963 	fixjobc(p, pgrp, 1);
964 	fixjobc(p, p->p_pgrp, 0);
965 
966 	/* Interlock with ttread(). */
967 	mutex_spin_enter(&tty_lock);
968 
969 	/* Move process to requested group. */
970 	LIST_REMOVE(p, p_pglist);
971 	if (LIST_EMPTY(&p->p_pgrp->pg_members))
972 		/* defer delete until we've dumped the lock */
973 		pg_id = p->p_pgrp->pg_id;
974 	p->p_pgrp = pgrp;
975 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
976 
977 	/* Done with the swap; we can release the tty mutex. */
978 	mutex_spin_exit(&tty_lock);
979 
980     done:
981 	if (pg_id != NO_PGID) {
982 		/* Releases proc_lock. */
983 		pg_delete(pg_id);
984 	} else {
985 		mutex_exit(proc_lock);
986 	}
987 	if (sess != NULL)
988 		kmem_free(sess, sizeof(*sess));
989 	if (new_pgrp != NULL)
990 		kmem_free(new_pgrp, sizeof(*new_pgrp));
991 #ifdef DEBUG_PGRP
992 	if (__predict_false(rval))
993 		printf("enterpgrp(%d,%d,%d), curproc %d, rval %d\n",
994 			pid, pgid, mksess, curp->p_pid, rval);
995 #endif
996 	return rval;
997 }
998 
999 /*
1000  * proc_leavepgrp: remove a process from its process group.
1001  *  => must be called with the proc_lock held, which will be released;
1002  */
1003 void
1004 proc_leavepgrp(struct proc *p)
1005 {
1006 	struct pgrp *pgrp;
1007 
1008 	KASSERT(mutex_owned(proc_lock));
1009 
1010 	/* Interlock with ttread() */
1011 	mutex_spin_enter(&tty_lock);
1012 	pgrp = p->p_pgrp;
1013 	LIST_REMOVE(p, p_pglist);
1014 	p->p_pgrp = NULL;
1015 	mutex_spin_exit(&tty_lock);
1016 
1017 	if (LIST_EMPTY(&pgrp->pg_members)) {
1018 		/* Releases proc_lock. */
1019 		pg_delete(pgrp->pg_id);
1020 	} else {
1021 		mutex_exit(proc_lock);
1022 	}
1023 }
1024 
1025 /*
1026  * pg_remove: remove a process group from the table.
1027  *  => must be called with the proc_lock held;
1028  *  => returns process group to free;
1029  */
1030 static struct pgrp *
1031 pg_remove(pid_t pg_id)
1032 {
1033 	struct pgrp *pgrp;
1034 	struct pid_table *pt;
1035 
1036 	KASSERT(mutex_owned(proc_lock));
1037 
1038 	pt = &pid_table[pg_id & pid_tbl_mask];
1039 	pgrp = pt->pt_pgrp;
1040 
1041 	KASSERT(pgrp != NULL);
1042 	KASSERT(pgrp->pg_id == pg_id);
1043 	KASSERT(LIST_EMPTY(&pgrp->pg_members));
1044 
1045 	pt->pt_pgrp = NULL;
1046 
1047 	if (!P_VALID(pt->pt_proc)) {
1048 		/* Orphaned pgrp, put slot onto free list. */
1049 		KASSERT((P_NEXT(pt->pt_proc) & pid_tbl_mask) == 0);
1050 		pg_id &= pid_tbl_mask;
1051 		pt = &pid_table[last_free_pt];
1052 		pt->pt_proc = P_FREE(P_NEXT(pt->pt_proc) | pg_id);
1053 		KASSERT(pt->pt_pid == 0);
1054 		last_free_pt = pg_id;
1055 		pid_alloc_cnt--;
1056 	}
1057 	return pgrp;
1058 }
1059 
1060 /*
1061  * pg_delete: delete and free a process group.
1062  *  => must be called with the proc_lock held, which will be released.
1063  */
1064 static void
1065 pg_delete(pid_t pg_id)
1066 {
1067 	struct pgrp *pg;
1068 	struct tty *ttyp;
1069 	struct session *ss;
1070 
1071 	KASSERT(mutex_owned(proc_lock));
1072 
1073 	pg = pid_table[pg_id & pid_tbl_mask].pt_pgrp;
1074 	if (pg == NULL || pg->pg_id != pg_id || !LIST_EMPTY(&pg->pg_members)) {
1075 		mutex_exit(proc_lock);
1076 		return;
1077 	}
1078 
1079 	ss = pg->pg_session;
1080 
1081 	/* Remove reference (if any) from tty to this process group */
1082 	mutex_spin_enter(&tty_lock);
1083 	ttyp = ss->s_ttyp;
1084 	if (ttyp != NULL && ttyp->t_pgrp == pg) {
1085 		ttyp->t_pgrp = NULL;
1086 		KASSERT(ttyp->t_session == ss);
1087 	}
1088 	mutex_spin_exit(&tty_lock);
1089 
1090 	/*
1091 	 * The leading process group in a session is freed by proc_sessrele(),
1092 	 * if last reference.  Note: proc_sessrele() releases proc_lock.
1093 	 */
1094 	pg = (ss->s_sid != pg->pg_id) ? pg_remove(pg_id) : NULL;
1095 	proc_sessrele(ss);
1096 
1097 	if (pg != NULL) {
1098 		/* Free it, if was not done by proc_sessrele(). */
1099 		kmem_free(pg, sizeof(struct pgrp));
1100 	}
1101 }
1102 
1103 /*
1104  * Adjust pgrp jobc counters when specified process changes process group.
1105  * We count the number of processes in each process group that "qualify"
1106  * the group for terminal job control (those with a parent in a different
1107  * process group of the same session).  If that count reaches zero, the
1108  * process group becomes orphaned.  Check both the specified process'
1109  * process group and that of its children.
1110  * entering == 0 => p is leaving specified group.
1111  * entering == 1 => p is entering specified group.
1112  *
1113  * Call with proc_lock held.
1114  */
1115 void
1116 fixjobc(struct proc *p, struct pgrp *pgrp, int entering)
1117 {
1118 	struct pgrp *hispgrp;
1119 	struct session *mysession = pgrp->pg_session;
1120 	struct proc *child;
1121 
1122 	KASSERT(mutex_owned(proc_lock));
1123 
1124 	/*
1125 	 * Check p's parent to see whether p qualifies its own process
1126 	 * group; if so, adjust count for p's process group.
1127 	 */
1128 	hispgrp = p->p_pptr->p_pgrp;
1129 	if (hispgrp != pgrp && hispgrp->pg_session == mysession) {
1130 		if (entering) {
1131 			pgrp->pg_jobc++;
1132 			p->p_lflag &= ~PL_ORPHANPG;
1133 		} else if (--pgrp->pg_jobc == 0)
1134 			orphanpg(pgrp);
1135 	}
1136 
1137 	/*
1138 	 * Check this process' children to see whether they qualify
1139 	 * their process groups; if so, adjust counts for children's
1140 	 * process groups.
1141 	 */
1142 	LIST_FOREACH(child, &p->p_children, p_sibling) {
1143 		hispgrp = child->p_pgrp;
1144 		if (hispgrp != pgrp && hispgrp->pg_session == mysession &&
1145 		    !P_ZOMBIE(child)) {
1146 			if (entering) {
1147 				child->p_lflag &= ~PL_ORPHANPG;
1148 				hispgrp->pg_jobc++;
1149 			} else if (--hispgrp->pg_jobc == 0)
1150 				orphanpg(hispgrp);
1151 		}
1152 	}
1153 }
1154 
1155 /*
1156  * A process group has become orphaned;
1157  * if there are any stopped processes in the group,
1158  * hang-up all process in that group.
1159  *
1160  * Call with proc_lock held.
1161  */
1162 static void
1163 orphanpg(struct pgrp *pg)
1164 {
1165 	struct proc *p;
1166 
1167 	KASSERT(mutex_owned(proc_lock));
1168 
1169 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
1170 		if (p->p_stat == SSTOP) {
1171 			p->p_lflag |= PL_ORPHANPG;
1172 			psignal(p, SIGHUP);
1173 			psignal(p, SIGCONT);
1174 		}
1175 	}
1176 }
1177 
1178 #ifdef DDB
1179 #include <ddb/db_output.h>
1180 void pidtbl_dump(void);
1181 void
1182 pidtbl_dump(void)
1183 {
1184 	struct pid_table *pt;
1185 	struct proc *p;
1186 	struct pgrp *pgrp;
1187 	int id;
1188 
1189 	db_printf("pid table %p size %x, next %x, last %x\n",
1190 		pid_table, pid_tbl_mask+1,
1191 		next_free_pt, last_free_pt);
1192 	for (pt = pid_table, id = 0; id <= pid_tbl_mask; id++, pt++) {
1193 		p = pt->pt_proc;
1194 		if (!P_VALID(p) && !pt->pt_pgrp)
1195 			continue;
1196 		db_printf("  id %x: ", id);
1197 		if (P_VALID(p))
1198 			db_printf("slotpid %d proc %p id %d (0x%x) %s\n",
1199 				pt->pt_pid, p, p->p_pid, p->p_pid, p->p_comm);
1200 		else
1201 			db_printf("next %x use %x\n",
1202 				P_NEXT(p) & pid_tbl_mask,
1203 				P_NEXT(p) & ~pid_tbl_mask);
1204 		if ((pgrp = pt->pt_pgrp)) {
1205 			db_printf("\tsession %p, sid %d, count %d, login %s\n",
1206 			    pgrp->pg_session, pgrp->pg_session->s_sid,
1207 			    pgrp->pg_session->s_count,
1208 			    pgrp->pg_session->s_login);
1209 			db_printf("\tpgrp %p, pg_id %d, pg_jobc %d, members %p\n",
1210 			    pgrp, pgrp->pg_id, pgrp->pg_jobc,
1211 			    LIST_FIRST(&pgrp->pg_members));
1212 			LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
1213 				db_printf("\t\tpid %d addr %p pgrp %p %s\n",
1214 				    p->p_pid, p, p->p_pgrp, p->p_comm);
1215 			}
1216 		}
1217 	}
1218 }
1219 #endif /* DDB */
1220 
1221 #ifdef KSTACK_CHECK_MAGIC
1222 
1223 #define	KSTACK_MAGIC	0xdeadbeaf
1224 
1225 /* XXX should be per process basis? */
1226 static int	kstackleftmin = KSTACK_SIZE;
1227 static int	kstackleftthres = KSTACK_SIZE / 8;
1228 
1229 void
1230 kstack_setup_magic(const struct lwp *l)
1231 {
1232 	uint32_t *ip;
1233 	uint32_t const *end;
1234 
1235 	KASSERT(l != NULL);
1236 	KASSERT(l != &lwp0);
1237 
1238 	/*
1239 	 * fill all the stack with magic number
1240 	 * so that later modification on it can be detected.
1241 	 */
1242 	ip = (uint32_t *)KSTACK_LOWEST_ADDR(l);
1243 	end = (uint32_t *)((char *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE);
1244 	for (; ip < end; ip++) {
1245 		*ip = KSTACK_MAGIC;
1246 	}
1247 }
1248 
1249 void
1250 kstack_check_magic(const struct lwp *l)
1251 {
1252 	uint32_t const *ip, *end;
1253 	int stackleft;
1254 
1255 	KASSERT(l != NULL);
1256 
1257 	/* don't check proc0 */ /*XXX*/
1258 	if (l == &lwp0)
1259 		return;
1260 
1261 #ifdef __MACHINE_STACK_GROWS_UP
1262 	/* stack grows upwards (eg. hppa) */
1263 	ip = (uint32_t *)((void *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE);
1264 	end = (uint32_t *)KSTACK_LOWEST_ADDR(l);
1265 	for (ip--; ip >= end; ip--)
1266 		if (*ip != KSTACK_MAGIC)
1267 			break;
1268 
1269 	stackleft = (void *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE - (void *)ip;
1270 #else /* __MACHINE_STACK_GROWS_UP */
1271 	/* stack grows downwards (eg. i386) */
1272 	ip = (uint32_t *)KSTACK_LOWEST_ADDR(l);
1273 	end = (uint32_t *)((char *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE);
1274 	for (; ip < end; ip++)
1275 		if (*ip != KSTACK_MAGIC)
1276 			break;
1277 
1278 	stackleft = ((const char *)ip) - (const char *)KSTACK_LOWEST_ADDR(l);
1279 #endif /* __MACHINE_STACK_GROWS_UP */
1280 
1281 	if (kstackleftmin > stackleft) {
1282 		kstackleftmin = stackleft;
1283 		if (stackleft < kstackleftthres)
1284 			printf("warning: kernel stack left %d bytes"
1285 			    "(pid %u:lid %u)\n", stackleft,
1286 			    (u_int)l->l_proc->p_pid, (u_int)l->l_lid);
1287 	}
1288 
1289 	if (stackleft <= 0) {
1290 		panic("magic on the top of kernel stack changed for "
1291 		    "pid %u, lid %u: maybe kernel stack overflow",
1292 		    (u_int)l->l_proc->p_pid, (u_int)l->l_lid);
1293 	}
1294 }
1295 #endif /* KSTACK_CHECK_MAGIC */
1296 
1297 int
1298 proclist_foreach_call(struct proclist *list,
1299     int (*callback)(struct proc *, void *arg), void *arg)
1300 {
1301 	struct proc marker;
1302 	struct proc *p;
1303 	int ret = 0;
1304 
1305 	marker.p_flag = PK_MARKER;
1306 	mutex_enter(proc_lock);
1307 	for (p = LIST_FIRST(list); ret == 0 && p != NULL;) {
1308 		if (p->p_flag & PK_MARKER) {
1309 			p = LIST_NEXT(p, p_list);
1310 			continue;
1311 		}
1312 		LIST_INSERT_AFTER(p, &marker, p_list);
1313 		ret = (*callback)(p, arg);
1314 		KASSERT(mutex_owned(proc_lock));
1315 		p = LIST_NEXT(&marker, p_list);
1316 		LIST_REMOVE(&marker, p_list);
1317 	}
1318 	mutex_exit(proc_lock);
1319 
1320 	return ret;
1321 }
1322 
1323 int
1324 proc_vmspace_getref(struct proc *p, struct vmspace **vm)
1325 {
1326 
1327 	/* XXXCDC: how should locking work here? */
1328 
1329 	/* curproc exception is for coredump. */
1330 
1331 	if ((p != curproc && (p->p_sflag & PS_WEXIT) != 0) ||
1332 	    (p->p_vmspace->vm_refcnt < 1)) { /* XXX */
1333 		return EFAULT;
1334 	}
1335 
1336 	uvmspace_addref(p->p_vmspace);
1337 	*vm = p->p_vmspace;
1338 
1339 	return 0;
1340 }
1341 
1342 /*
1343  * Acquire a write lock on the process credential.
1344  */
1345 void
1346 proc_crmod_enter(void)
1347 {
1348 	struct lwp *l = curlwp;
1349 	struct proc *p = l->l_proc;
1350 	kauth_cred_t oc;
1351 
1352 	/* Reset what needs to be reset in plimit. */
1353 	if (p->p_limit->pl_corename != defcorename) {
1354 		lim_setcorename(p, defcorename, 0);
1355 	}
1356 
1357 	mutex_enter(p->p_lock);
1358 
1359 	/* Ensure the LWP cached credentials are up to date. */
1360 	if ((oc = l->l_cred) != p->p_cred) {
1361 		kauth_cred_hold(p->p_cred);
1362 		l->l_cred = p->p_cred;
1363 		kauth_cred_free(oc);
1364 	}
1365 }
1366 
1367 /*
1368  * Set in a new process credential, and drop the write lock.  The credential
1369  * must have a reference already.  Optionally, free a no-longer required
1370  * credential.  The scheduler also needs to inspect p_cred, so we also
1371  * briefly acquire the sched state mutex.
1372  */
1373 void
1374 proc_crmod_leave(kauth_cred_t scred, kauth_cred_t fcred, bool sugid)
1375 {
1376 	struct lwp *l = curlwp, *l2;
1377 	struct proc *p = l->l_proc;
1378 	kauth_cred_t oc;
1379 
1380 	KASSERT(mutex_owned(p->p_lock));
1381 
1382 	/* Is there a new credential to set in? */
1383 	if (scred != NULL) {
1384 		p->p_cred = scred;
1385 		LIST_FOREACH(l2, &p->p_lwps, l_sibling) {
1386 			if (l2 != l)
1387 				l2->l_prflag |= LPR_CRMOD;
1388 		}
1389 
1390 		/* Ensure the LWP cached credentials are up to date. */
1391 		if ((oc = l->l_cred) != scred) {
1392 			kauth_cred_hold(scred);
1393 			l->l_cred = scred;
1394 		}
1395 	} else
1396 		oc = NULL;	/* XXXgcc */
1397 
1398 	if (sugid) {
1399 		/*
1400 		 * Mark process as having changed credentials, stops
1401 		 * tracing etc.
1402 		 */
1403 		p->p_flag |= PK_SUGID;
1404 	}
1405 
1406 	mutex_exit(p->p_lock);
1407 
1408 	/* If there is a credential to be released, free it now. */
1409 	if (fcred != NULL) {
1410 		KASSERT(scred != NULL);
1411 		kauth_cred_free(fcred);
1412 		if (oc != scred)
1413 			kauth_cred_free(oc);
1414 	}
1415 }
1416 
1417 /*
1418  * proc_specific_key_create --
1419  *	Create a key for subsystem proc-specific data.
1420  */
1421 int
1422 proc_specific_key_create(specificdata_key_t *keyp, specificdata_dtor_t dtor)
1423 {
1424 
1425 	return (specificdata_key_create(proc_specificdata_domain, keyp, dtor));
1426 }
1427 
1428 /*
1429  * proc_specific_key_delete --
1430  *	Delete a key for subsystem proc-specific data.
1431  */
1432 void
1433 proc_specific_key_delete(specificdata_key_t key)
1434 {
1435 
1436 	specificdata_key_delete(proc_specificdata_domain, key);
1437 }
1438 
1439 /*
1440  * proc_initspecific --
1441  *	Initialize a proc's specificdata container.
1442  */
1443 void
1444 proc_initspecific(struct proc *p)
1445 {
1446 	int error;
1447 
1448 	error = specificdata_init(proc_specificdata_domain, &p->p_specdataref);
1449 	KASSERT(error == 0);
1450 }
1451 
1452 /*
1453  * proc_finispecific --
1454  *	Finalize a proc's specificdata container.
1455  */
1456 void
1457 proc_finispecific(struct proc *p)
1458 {
1459 
1460 	specificdata_fini(proc_specificdata_domain, &p->p_specdataref);
1461 }
1462 
1463 /*
1464  * proc_getspecific --
1465  *	Return proc-specific data corresponding to the specified key.
1466  */
1467 void *
1468 proc_getspecific(struct proc *p, specificdata_key_t key)
1469 {
1470 
1471 	return (specificdata_getspecific(proc_specificdata_domain,
1472 					 &p->p_specdataref, key));
1473 }
1474 
1475 /*
1476  * proc_setspecific --
1477  *	Set proc-specific data corresponding to the specified key.
1478  */
1479 void
1480 proc_setspecific(struct proc *p, specificdata_key_t key, void *data)
1481 {
1482 
1483 	specificdata_setspecific(proc_specificdata_domain,
1484 				 &p->p_specdataref, key, data);
1485 }
1486 
1487 int
1488 proc_uidmatch(kauth_cred_t cred, kauth_cred_t target)
1489 {
1490 	int r = 0;
1491 
1492 	if (kauth_cred_getuid(cred) != kauth_cred_getuid(target) ||
1493 	    kauth_cred_getuid(cred) != kauth_cred_getsvuid(target)) {
1494 		/*
1495 		 * suid proc of ours or proc not ours
1496 		 */
1497 		r = EPERM;
1498 	} else if (kauth_cred_getgid(target) != kauth_cred_getsvgid(target)) {
1499 		/*
1500 		 * sgid proc has sgid back to us temporarily
1501 		 */
1502 		r = EPERM;
1503 	} else {
1504 		/*
1505 		 * our rgid must be in target's group list (ie,
1506 		 * sub-processes started by a sgid process)
1507 		 */
1508 		int ismember = 0;
1509 
1510 		if (kauth_cred_ismember_gid(cred,
1511 		    kauth_cred_getgid(target), &ismember) != 0 ||
1512 		    !ismember)
1513 			r = EPERM;
1514 	}
1515 
1516 	return (r);
1517 }
1518 
1519 /*
1520  * sysctl stuff
1521  */
1522 
1523 #define KERN_PROCSLOP	(5 * sizeof(struct kinfo_proc))
1524 
1525 static const u_int sysctl_flagmap[] = {
1526 	PK_ADVLOCK, P_ADVLOCK,
1527 	PK_EXEC, P_EXEC,
1528 	PK_NOCLDWAIT, P_NOCLDWAIT,
1529 	PK_32, P_32,
1530 	PK_CLDSIGIGN, P_CLDSIGIGN,
1531 	PK_SUGID, P_SUGID,
1532 	0
1533 };
1534 
1535 static const u_int sysctl_sflagmap[] = {
1536 	PS_NOCLDSTOP, P_NOCLDSTOP,
1537 	PS_WEXIT, P_WEXIT,
1538 	PS_STOPFORK, P_STOPFORK,
1539 	PS_STOPEXEC, P_STOPEXEC,
1540 	PS_STOPEXIT, P_STOPEXIT,
1541 	0
1542 };
1543 
1544 static const u_int sysctl_slflagmap[] = {
1545 	PSL_TRACED, P_TRACED,
1546 	PSL_FSTRACE, P_FSTRACE,
1547 	PSL_CHTRACED, P_CHTRACED,
1548 	PSL_SYSCALL, P_SYSCALL,
1549 	0
1550 };
1551 
1552 static const u_int sysctl_lflagmap[] = {
1553 	PL_CONTROLT, P_CONTROLT,
1554 	PL_PPWAIT, P_PPWAIT,
1555 	0
1556 };
1557 
1558 static const u_int sysctl_stflagmap[] = {
1559 	PST_PROFIL, P_PROFIL,
1560 	0
1561 
1562 };
1563 
1564 /* used by kern_lwp also */
1565 const u_int sysctl_lwpflagmap[] = {
1566 	LW_SINTR, L_SINTR,
1567 	LW_SYSTEM, L_SYSTEM,
1568 	0
1569 };
1570 
1571 /*
1572  * Find the most ``active'' lwp of a process and return it for ps display
1573  * purposes
1574  */
1575 static struct lwp *
1576 proc_active_lwp(struct proc *p)
1577 {
1578 	static const int ostat[] = {
1579 		0,
1580 		2,	/* LSIDL */
1581 		6,	/* LSRUN */
1582 		5,	/* LSSLEEP */
1583 		4,	/* LSSTOP */
1584 		0,	/* LSZOMB */
1585 		1,	/* LSDEAD */
1586 		7,	/* LSONPROC */
1587 		3	/* LSSUSPENDED */
1588 	};
1589 
1590 	struct lwp *l, *lp = NULL;
1591 	LIST_FOREACH(l, &p->p_lwps, l_sibling) {
1592 		KASSERT(l->l_stat >= 0 && l->l_stat < __arraycount(ostat));
1593 		if (lp == NULL ||
1594 		    ostat[l->l_stat] > ostat[lp->l_stat] ||
1595 		    (ostat[l->l_stat] == ostat[lp->l_stat] &&
1596 		    l->l_cpticks > lp->l_cpticks)) {
1597 			lp = l;
1598 			continue;
1599 		}
1600 	}
1601 	return lp;
1602 }
1603 
1604 static int
1605 sysctl_doeproc(SYSCTLFN_ARGS)
1606 {
1607 	union {
1608 		struct kinfo_proc kproc;
1609 		struct kinfo_proc2 kproc2;
1610 	} *kbuf;
1611 	struct proc *p, *next, *marker;
1612 	char *where, *dp;
1613 	int type, op, arg, error;
1614 	u_int elem_size, kelem_size, elem_count;
1615 	size_t buflen, needed;
1616 	bool match, zombie, mmmbrains;
1617 
1618 	if (namelen == 1 && name[0] == CTL_QUERY)
1619 		return (sysctl_query(SYSCTLFN_CALL(rnode)));
1620 
1621 	dp = where = oldp;
1622 	buflen = where != NULL ? *oldlenp : 0;
1623 	error = 0;
1624 	needed = 0;
1625 	type = rnode->sysctl_num;
1626 
1627 	if (type == KERN_PROC) {
1628 		if (namelen != 2 && !(namelen == 1 && name[0] == KERN_PROC_ALL))
1629 			return (EINVAL);
1630 		op = name[0];
1631 		if (op != KERN_PROC_ALL)
1632 			arg = name[1];
1633 		else
1634 			arg = 0;		/* Quell compiler warning */
1635 		elem_count = 0;	/* Ditto */
1636 		kelem_size = elem_size = sizeof(kbuf->kproc);
1637 	} else {
1638 		if (namelen != 4)
1639 			return (EINVAL);
1640 		op = name[0];
1641 		arg = name[1];
1642 		elem_size = name[2];
1643 		elem_count = name[3];
1644 		kelem_size = sizeof(kbuf->kproc2);
1645 	}
1646 
1647 	sysctl_unlock();
1648 
1649 	kbuf = kmem_alloc(sizeof(*kbuf), KM_SLEEP);
1650 	marker = kmem_alloc(sizeof(*marker), KM_SLEEP);
1651 	marker->p_flag = PK_MARKER;
1652 
1653 	mutex_enter(proc_lock);
1654 	mmmbrains = false;
1655 	for (p = LIST_FIRST(&allproc);; p = next) {
1656 		if (p == NULL) {
1657 			if (!mmmbrains) {
1658 				p = LIST_FIRST(&zombproc);
1659 				mmmbrains = true;
1660 			}
1661 			if (p == NULL)
1662 				break;
1663 		}
1664 		next = LIST_NEXT(p, p_list);
1665 		if ((p->p_flag & PK_MARKER) != 0)
1666 			continue;
1667 
1668 		/*
1669 		 * Skip embryonic processes.
1670 		 */
1671 		if (p->p_stat == SIDL)
1672 			continue;
1673 
1674 		mutex_enter(p->p_lock);
1675 		error = kauth_authorize_process(l->l_cred,
1676 		    KAUTH_PROCESS_CANSEE, p,
1677 		    KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL);
1678 		if (error != 0) {
1679 			mutex_exit(p->p_lock);
1680 			continue;
1681 		}
1682 
1683 		/*
1684 		 * TODO - make more efficient (see notes below).
1685 		 * do by session.
1686 		 */
1687 		switch (op) {
1688 		case KERN_PROC_PID:
1689 			/* could do this with just a lookup */
1690 			match = (p->p_pid == (pid_t)arg);
1691 			break;
1692 
1693 		case KERN_PROC_PGRP:
1694 			/* could do this by traversing pgrp */
1695 			match = (p->p_pgrp->pg_id == (pid_t)arg);
1696 			break;
1697 
1698 		case KERN_PROC_SESSION:
1699 			match = (p->p_session->s_sid == (pid_t)arg);
1700 			break;
1701 
1702 		case KERN_PROC_TTY:
1703 			match = true;
1704 			if (arg == (int) KERN_PROC_TTY_REVOKE) {
1705 				if ((p->p_lflag & PL_CONTROLT) == 0 ||
1706 				    p->p_session->s_ttyp == NULL ||
1707 				    p->p_session->s_ttyvp != NULL) {
1708 				    	match = false;
1709 				}
1710 			} else if ((p->p_lflag & PL_CONTROLT) == 0 ||
1711 			    p->p_session->s_ttyp == NULL) {
1712 				if ((dev_t)arg != KERN_PROC_TTY_NODEV) {
1713 					match = false;
1714 				}
1715 			} else if (p->p_session->s_ttyp->t_dev != (dev_t)arg) {
1716 				match = false;
1717 			}
1718 			break;
1719 
1720 		case KERN_PROC_UID:
1721 			match = (kauth_cred_geteuid(p->p_cred) == (uid_t)arg);
1722 			break;
1723 
1724 		case KERN_PROC_RUID:
1725 			match = (kauth_cred_getuid(p->p_cred) == (uid_t)arg);
1726 			break;
1727 
1728 		case KERN_PROC_GID:
1729 			match = (kauth_cred_getegid(p->p_cred) == (uid_t)arg);
1730 			break;
1731 
1732 		case KERN_PROC_RGID:
1733 			match = (kauth_cred_getgid(p->p_cred) == (uid_t)arg);
1734 			break;
1735 
1736 		case KERN_PROC_ALL:
1737 			match = true;
1738 			/* allow everything */
1739 			break;
1740 
1741 		default:
1742 			error = EINVAL;
1743 			mutex_exit(p->p_lock);
1744 			goto cleanup;
1745 		}
1746 		if (!match) {
1747 			mutex_exit(p->p_lock);
1748 			continue;
1749 		}
1750 
1751 		/*
1752 		 * Grab a hold on the process.
1753 		 */
1754 		if (mmmbrains) {
1755 			zombie = true;
1756 		} else {
1757 			zombie = !rw_tryenter(&p->p_reflock, RW_READER);
1758 		}
1759 		if (zombie) {
1760 			LIST_INSERT_AFTER(p, marker, p_list);
1761 		}
1762 
1763 		if (buflen >= elem_size &&
1764 		    (type == KERN_PROC || elem_count > 0)) {
1765 			if (type == KERN_PROC) {
1766 				kbuf->kproc.kp_proc = *p;
1767 				fill_eproc(p, &kbuf->kproc.kp_eproc, zombie);
1768 			} else {
1769 				fill_kproc2(p, &kbuf->kproc2, zombie);
1770 				elem_count--;
1771 			}
1772 			mutex_exit(p->p_lock);
1773 			mutex_exit(proc_lock);
1774 			/*
1775 			 * Copy out elem_size, but not larger than kelem_size
1776 			 */
1777 			error = sysctl_copyout(l, kbuf, dp,
1778 			    min(kelem_size, elem_size));
1779 			mutex_enter(proc_lock);
1780 			if (error) {
1781 				goto bah;
1782 			}
1783 			dp += elem_size;
1784 			buflen -= elem_size;
1785 		} else {
1786 			mutex_exit(p->p_lock);
1787 		}
1788 		needed += elem_size;
1789 
1790 		/*
1791 		 * Release reference to process.
1792 		 */
1793 	 	if (zombie) {
1794 			next = LIST_NEXT(marker, p_list);
1795  			LIST_REMOVE(marker, p_list);
1796 		} else {
1797 			rw_exit(&p->p_reflock);
1798 			next = LIST_NEXT(p, p_list);
1799 		}
1800 	}
1801 	mutex_exit(proc_lock);
1802 
1803 	if (where != NULL) {
1804 		*oldlenp = dp - where;
1805 		if (needed > *oldlenp) {
1806 			error = ENOMEM;
1807 			goto out;
1808 		}
1809 	} else {
1810 		needed += KERN_PROCSLOP;
1811 		*oldlenp = needed;
1812 	}
1813 	if (kbuf)
1814 		kmem_free(kbuf, sizeof(*kbuf));
1815 	if (marker)
1816 		kmem_free(marker, sizeof(*marker));
1817 	sysctl_relock();
1818 	return 0;
1819  bah:
1820  	if (zombie)
1821  		LIST_REMOVE(marker, p_list);
1822 	else
1823 		rw_exit(&p->p_reflock);
1824  cleanup:
1825 	mutex_exit(proc_lock);
1826  out:
1827 	if (kbuf)
1828 		kmem_free(kbuf, sizeof(*kbuf));
1829 	if (marker)
1830 		kmem_free(marker, sizeof(*marker));
1831 	sysctl_relock();
1832 	return error;
1833 }
1834 
1835 int
1836 copyin_psstrings(struct proc *p, struct ps_strings *arginfo)
1837 {
1838 
1839 #ifdef COMPAT_NETBSD32
1840 	if (p->p_flag & PK_32) {
1841 		struct ps_strings32 arginfo32;
1842 
1843 		int error = copyin_proc(p, (void *)p->p_psstrp, &arginfo32,
1844 		    sizeof(arginfo32));
1845 		if (error)
1846 			return error;
1847 		arginfo->ps_argvstr = (void *)(uintptr_t)arginfo32.ps_argvstr;
1848 		arginfo->ps_nargvstr = arginfo32.ps_nargvstr;
1849 		arginfo->ps_envstr = (void *)(uintptr_t)arginfo32.ps_envstr;
1850 		arginfo->ps_nenvstr = arginfo32.ps_nenvstr;
1851 		return 0;
1852 	}
1853 #endif
1854 	return copyin_proc(p, (void *)p->p_psstrp, arginfo, sizeof(*arginfo));
1855 }
1856 
1857 static int
1858 copy_procargs_sysctl_cb(void *cookie_, const void *src, size_t off, size_t len)
1859 {
1860 	void **cookie = cookie_;
1861 	struct lwp *l = cookie[0];
1862 	char *dst = cookie[1];
1863 
1864 	return sysctl_copyout(l, src, dst + off, len);
1865 }
1866 
1867 /*
1868  * sysctl helper routine for kern.proc_args pseudo-subtree.
1869  */
1870 static int
1871 sysctl_kern_proc_args(SYSCTLFN_ARGS)
1872 {
1873 	struct ps_strings pss;
1874 	struct proc *p;
1875 	pid_t pid;
1876 	int type, error;
1877 	void *cookie[2];
1878 
1879 	if (namelen == 1 && name[0] == CTL_QUERY)
1880 		return (sysctl_query(SYSCTLFN_CALL(rnode)));
1881 
1882 	if (newp != NULL || namelen != 2)
1883 		return (EINVAL);
1884 	pid = name[0];
1885 	type = name[1];
1886 
1887 	switch (type) {
1888 	case KERN_PROC_ARGV:
1889 	case KERN_PROC_NARGV:
1890 	case KERN_PROC_ENV:
1891 	case KERN_PROC_NENV:
1892 		/* ok */
1893 		break;
1894 	default:
1895 		return (EINVAL);
1896 	}
1897 
1898 	sysctl_unlock();
1899 
1900 	/* check pid */
1901 	mutex_enter(proc_lock);
1902 	if ((p = proc_find(pid)) == NULL) {
1903 		error = EINVAL;
1904 		goto out_locked;
1905 	}
1906 	mutex_enter(p->p_lock);
1907 
1908 	/* Check permission. */
1909 	if (type == KERN_PROC_ARGV || type == KERN_PROC_NARGV)
1910 		error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE,
1911 		    p, KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ARGS), NULL, NULL);
1912 	else if (type == KERN_PROC_ENV || type == KERN_PROC_NENV)
1913 		error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE,
1914 		    p, KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENV), NULL, NULL);
1915 	else
1916 		error = EINVAL; /* XXXGCC */
1917 	if (error) {
1918 		mutex_exit(p->p_lock);
1919 		goto out_locked;
1920 	}
1921 
1922 	if (oldp == NULL) {
1923 		if (type == KERN_PROC_NARGV || type == KERN_PROC_NENV)
1924 			*oldlenp = sizeof (int);
1925 		else
1926 			*oldlenp = ARG_MAX;	/* XXX XXX XXX */
1927 		error = 0;
1928 		mutex_exit(p->p_lock);
1929 		goto out_locked;
1930 	}
1931 
1932 	/*
1933 	 * Zombies don't have a stack, so we can't read their psstrings.
1934 	 * System processes also don't have a user stack.
1935 	 */
1936 	if (P_ZOMBIE(p) || (p->p_flag & PK_SYSTEM) != 0) {
1937 		error = EINVAL;
1938 		mutex_exit(p->p_lock);
1939 		goto out_locked;
1940 	}
1941 
1942 	error = rw_tryenter(&p->p_reflock, RW_READER) ? 0 : EBUSY;
1943 	mutex_exit(p->p_lock);
1944 	if (error) {
1945 		goto out_locked;
1946 	}
1947 	mutex_exit(proc_lock);
1948 
1949 	if (type == KERN_PROC_NARGV || type == KERN_PROC_NENV) {
1950 		int value;
1951 		if ((error = copyin_psstrings(p, &pss)) == 0) {
1952 			if (type == KERN_PROC_NARGV)
1953 				value = pss.ps_nargvstr;
1954 			else
1955 				value = pss.ps_nenvstr;
1956 			error = sysctl_copyout(l, &value, oldp, sizeof(value));
1957 			*oldlenp = sizeof(value);
1958 		}
1959 	} else {
1960 		cookie[0] = l;
1961 		cookie[1] = oldp;
1962 		error = copy_procargs(p, type, oldlenp,
1963 		    copy_procargs_sysctl_cb, cookie);
1964 	}
1965 	rw_exit(&p->p_reflock);
1966 	sysctl_relock();
1967 	return error;
1968 
1969 out_locked:
1970 	mutex_exit(proc_lock);
1971 	sysctl_relock();
1972 	return error;
1973 }
1974 
1975 int
1976 copy_procargs(struct proc *p, int oid, size_t *limit,
1977     int (*cb)(void *, const void *, size_t, size_t), void *cookie)
1978 {
1979 	struct ps_strings pss;
1980 	size_t len, i, loaded, entry_len;
1981 	struct uio auio;
1982 	struct iovec aiov;
1983 	int error, argvlen;
1984 	char *arg;
1985 	char **argv;
1986 	vaddr_t user_argv;
1987 	struct vmspace *vmspace;
1988 
1989 	/*
1990 	 * Allocate a temporary buffer to hold the argument vector and
1991 	 * the arguments themselve.
1992 	 */
1993 	arg = kmem_alloc(PAGE_SIZE, KM_SLEEP);
1994 	argv = kmem_alloc(PAGE_SIZE, KM_SLEEP);
1995 
1996 	/*
1997 	 * Lock the process down in memory.
1998 	 */
1999 	vmspace = p->p_vmspace;
2000 	uvmspace_addref(vmspace);
2001 
2002 	/*
2003 	 * Read in the ps_strings structure.
2004 	 */
2005 	if ((error = copyin_psstrings(p, &pss)) != 0)
2006 		goto done;
2007 
2008 	/*
2009 	 * Now read the address of the argument vector.
2010 	 */
2011 	switch (oid) {
2012 	case KERN_PROC_ARGV:
2013 		user_argv = (uintptr_t)pss.ps_argvstr;
2014 		argvlen = pss.ps_nargvstr;
2015 		break;
2016 	case KERN_PROC_ENV:
2017 		user_argv = (uintptr_t)pss.ps_envstr;
2018 		argvlen = pss.ps_nenvstr;
2019 		break;
2020 	default:
2021 		error = EINVAL;
2022 		goto done;
2023 	}
2024 
2025 	if (argvlen < 0) {
2026 		error = EIO;
2027 		goto done;
2028 	}
2029 
2030 #ifdef COMPAT_NETBSD32
2031 	if (p->p_flag & PK_32)
2032 		entry_len = sizeof(netbsd32_charp);
2033 	else
2034 #endif
2035 		entry_len = sizeof(char *);
2036 
2037 	/*
2038 	 * Now copy each string.
2039 	 */
2040 	len = 0; /* bytes written to user buffer */
2041 	loaded = 0; /* bytes from argv already processed */
2042 	i = 0; /* To make compiler happy */
2043 
2044 	for (; argvlen; --argvlen) {
2045 		int finished = 0;
2046 		vaddr_t base;
2047 		size_t xlen;
2048 		int j;
2049 
2050 		if (loaded == 0) {
2051 			size_t rem = entry_len * argvlen;
2052 			loaded = MIN(rem, PAGE_SIZE);
2053 			error = copyin_vmspace(vmspace,
2054 			    (const void *)user_argv, argv, loaded);
2055 			if (error)
2056 				break;
2057 			user_argv += loaded;
2058 			i = 0;
2059 		}
2060 
2061 #ifdef COMPAT_NETBSD32
2062 		if (p->p_flag & PK_32) {
2063 			netbsd32_charp *argv32;
2064 
2065 			argv32 = (netbsd32_charp *)argv;
2066 			base = (vaddr_t)NETBSD32PTR64(argv32[i++]);
2067 		} else
2068 #endif
2069 			base = (vaddr_t)argv[i++];
2070 		loaded -= entry_len;
2071 
2072 		/*
2073 		 * The program has messed around with its arguments,
2074 		 * possibly deleting some, and replacing them with
2075 		 * NULL's. Treat this as the last argument and not
2076 		 * a failure.
2077 		 */
2078 		if (base == 0)
2079 			break;
2080 
2081 		while (!finished) {
2082 			xlen = PAGE_SIZE - (base & PAGE_MASK);
2083 
2084 			aiov.iov_base = arg;
2085 			aiov.iov_len = PAGE_SIZE;
2086 			auio.uio_iov = &aiov;
2087 			auio.uio_iovcnt = 1;
2088 			auio.uio_offset = base;
2089 			auio.uio_resid = xlen;
2090 			auio.uio_rw = UIO_READ;
2091 			UIO_SETUP_SYSSPACE(&auio);
2092 			error = uvm_io(&vmspace->vm_map, &auio);
2093 			if (error)
2094 				goto done;
2095 
2096 			/* Look for the end of the string */
2097 			for (j = 0; j < xlen; j++) {
2098 				if (arg[j] == '\0') {
2099 					xlen = j + 1;
2100 					finished = 1;
2101 					break;
2102 				}
2103 			}
2104 
2105 			/* Check for user buffer overflow */
2106 			if (len + xlen > *limit) {
2107 				finished = 1;
2108 				if (len > *limit)
2109 					xlen = 0;
2110 				else
2111 					xlen = *limit - len;
2112 			}
2113 
2114 			/* Copyout the page */
2115 			error = (*cb)(cookie, arg, len, xlen);
2116 			if (error)
2117 				goto done;
2118 
2119 			len += xlen;
2120 			base += xlen;
2121 		}
2122 	}
2123 	*limit = len;
2124 
2125 done:
2126 	kmem_free(argv, PAGE_SIZE);
2127 	kmem_free(arg, PAGE_SIZE);
2128 	uvmspace_free(vmspace);
2129 	return error;
2130 }
2131 
2132 /*
2133  * Fill in an eproc structure for the specified process.
2134  */
2135 void
2136 fill_eproc(struct proc *p, struct eproc *ep, bool zombie)
2137 {
2138 	struct tty *tp;
2139 	struct lwp *l;
2140 
2141 	KASSERT(mutex_owned(proc_lock));
2142 	KASSERT(mutex_owned(p->p_lock));
2143 
2144 	memset(ep, 0, sizeof(*ep));
2145 
2146 	ep->e_paddr = p;
2147 	ep->e_sess = p->p_session;
2148 	if (p->p_cred) {
2149 		kauth_cred_topcred(p->p_cred, &ep->e_pcred);
2150 		kauth_cred_toucred(p->p_cred, &ep->e_ucred);
2151 	}
2152 	if (p->p_stat != SIDL && !P_ZOMBIE(p) && !zombie) {
2153 		struct vmspace *vm = p->p_vmspace;
2154 
2155 		ep->e_vm.vm_rssize = vm_resident_count(vm);
2156 		ep->e_vm.vm_tsize = vm->vm_tsize;
2157 		ep->e_vm.vm_dsize = vm->vm_dsize;
2158 		ep->e_vm.vm_ssize = vm->vm_ssize;
2159 		ep->e_vm.vm_map.size = vm->vm_map.size;
2160 
2161 		/* Pick the primary (first) LWP */
2162 		l = proc_active_lwp(p);
2163 		KASSERT(l != NULL);
2164 		lwp_lock(l);
2165 		if (l->l_wchan)
2166 			strncpy(ep->e_wmesg, l->l_wmesg, WMESGLEN);
2167 		lwp_unlock(l);
2168 	}
2169 	if (p->p_pptr)
2170 		ep->e_ppid = p->p_pptr->p_pid;
2171 	if (p->p_pgrp && p->p_session) {
2172 		ep->e_pgid = p->p_pgrp->pg_id;
2173 		ep->e_jobc = p->p_pgrp->pg_jobc;
2174 		ep->e_sid = p->p_session->s_sid;
2175 		if ((p->p_lflag & PL_CONTROLT) &&
2176 		    (tp = ep->e_sess->s_ttyp)) {
2177 			ep->e_tdev = tp->t_dev;
2178 			ep->e_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PGID;
2179 			ep->e_tsess = tp->t_session;
2180 		} else
2181 			ep->e_tdev = (uint32_t)NODEV;
2182 		ep->e_flag = ep->e_sess->s_ttyvp ? EPROC_CTTY : 0;
2183 		if (SESS_LEADER(p))
2184 			ep->e_flag |= EPROC_SLEADER;
2185 		strncpy(ep->e_login, ep->e_sess->s_login, MAXLOGNAME);
2186 	}
2187 	ep->e_xsize = ep->e_xrssize = 0;
2188 	ep->e_xccount = ep->e_xswrss = 0;
2189 }
2190 
2191 /*
2192  * Fill in a kinfo_proc2 structure for the specified process.
2193  */
2194 static void
2195 fill_kproc2(struct proc *p, struct kinfo_proc2 *ki, bool zombie)
2196 {
2197 	struct tty *tp;
2198 	struct lwp *l, *l2;
2199 	struct timeval ut, st, rt;
2200 	sigset_t ss1, ss2;
2201 	struct rusage ru;
2202 	struct vmspace *vm;
2203 
2204 	KASSERT(mutex_owned(proc_lock));
2205 	KASSERT(mutex_owned(p->p_lock));
2206 
2207 	sigemptyset(&ss1);
2208 	sigemptyset(&ss2);
2209 	memset(ki, 0, sizeof(*ki));
2210 
2211 	ki->p_paddr = PTRTOUINT64(p);
2212 	ki->p_fd = PTRTOUINT64(p->p_fd);
2213 	ki->p_cwdi = PTRTOUINT64(p->p_cwdi);
2214 	ki->p_stats = PTRTOUINT64(p->p_stats);
2215 	ki->p_limit = PTRTOUINT64(p->p_limit);
2216 	ki->p_vmspace = PTRTOUINT64(p->p_vmspace);
2217 	ki->p_sigacts = PTRTOUINT64(p->p_sigacts);
2218 	ki->p_sess = PTRTOUINT64(p->p_session);
2219 	ki->p_tsess = 0;	/* may be changed if controlling tty below */
2220 	ki->p_ru = PTRTOUINT64(&p->p_stats->p_ru);
2221 	ki->p_eflag = 0;
2222 	ki->p_exitsig = p->p_exitsig;
2223 	ki->p_flag = L_INMEM;   /* Process never swapped out */
2224 	ki->p_flag |= sysctl_map_flags(sysctl_flagmap, p->p_flag);
2225 	ki->p_flag |= sysctl_map_flags(sysctl_sflagmap, p->p_sflag);
2226 	ki->p_flag |= sysctl_map_flags(sysctl_slflagmap, p->p_slflag);
2227 	ki->p_flag |= sysctl_map_flags(sysctl_lflagmap, p->p_lflag);
2228 	ki->p_flag |= sysctl_map_flags(sysctl_stflagmap, p->p_stflag);
2229 	ki->p_pid = p->p_pid;
2230 	if (p->p_pptr)
2231 		ki->p_ppid = p->p_pptr->p_pid;
2232 	else
2233 		ki->p_ppid = 0;
2234 	ki->p_uid = kauth_cred_geteuid(p->p_cred);
2235 	ki->p_ruid = kauth_cred_getuid(p->p_cred);
2236 	ki->p_gid = kauth_cred_getegid(p->p_cred);
2237 	ki->p_rgid = kauth_cred_getgid(p->p_cred);
2238 	ki->p_svuid = kauth_cred_getsvuid(p->p_cred);
2239 	ki->p_svgid = kauth_cred_getsvgid(p->p_cred);
2240 	ki->p_ngroups = kauth_cred_ngroups(p->p_cred);
2241 	kauth_cred_getgroups(p->p_cred, ki->p_groups,
2242 	    min(ki->p_ngroups, sizeof(ki->p_groups) / sizeof(ki->p_groups[0])),
2243 	    UIO_SYSSPACE);
2244 
2245 	ki->p_uticks = p->p_uticks;
2246 	ki->p_sticks = p->p_sticks;
2247 	ki->p_iticks = p->p_iticks;
2248 	ki->p_tpgid = NO_PGID;	/* may be changed if controlling tty below */
2249 	ki->p_tracep = PTRTOUINT64(p->p_tracep);
2250 	ki->p_traceflag = p->p_traceflag;
2251 
2252 	memcpy(&ki->p_sigignore, &p->p_sigctx.ps_sigignore,sizeof(ki_sigset_t));
2253 	memcpy(&ki->p_sigcatch, &p->p_sigctx.ps_sigcatch, sizeof(ki_sigset_t));
2254 
2255 	ki->p_cpticks = 0;
2256 	ki->p_pctcpu = p->p_pctcpu;
2257 	ki->p_estcpu = 0;
2258 	ki->p_stat = p->p_stat; /* Will likely be overridden by LWP status */
2259 	ki->p_realstat = p->p_stat;
2260 	ki->p_nice = p->p_nice;
2261 	ki->p_xstat = p->p_xstat;
2262 	ki->p_acflag = p->p_acflag;
2263 
2264 	strncpy(ki->p_comm, p->p_comm,
2265 	    min(sizeof(ki->p_comm), sizeof(p->p_comm)));
2266 	strncpy(ki->p_ename, p->p_emul->e_name, sizeof(ki->p_ename));
2267 
2268 	ki->p_nlwps = p->p_nlwps;
2269 	ki->p_realflag = ki->p_flag;
2270 
2271 	if (p->p_stat != SIDL && !P_ZOMBIE(p) && !zombie) {
2272 		vm = p->p_vmspace;
2273 		ki->p_vm_rssize = vm_resident_count(vm);
2274 		ki->p_vm_tsize = vm->vm_tsize;
2275 		ki->p_vm_dsize = vm->vm_dsize;
2276 		ki->p_vm_ssize = vm->vm_ssize;
2277 		ki->p_vm_vsize = vm->vm_map.size;
2278 		/*
2279 		 * Since the stack is initially mapped mostly with
2280 		 * PROT_NONE and grown as needed, adjust the "mapped size"
2281 		 * to skip the unused stack portion.
2282 		 */
2283 		ki->p_vm_msize =
2284 		    atop(vm->vm_map.size) - vm->vm_issize + vm->vm_ssize;
2285 
2286 		/* Pick the primary (first) LWP */
2287 		l = proc_active_lwp(p);
2288 		KASSERT(l != NULL);
2289 		lwp_lock(l);
2290 		ki->p_nrlwps = p->p_nrlwps;
2291 		ki->p_forw = 0;
2292 		ki->p_back = 0;
2293 		ki->p_addr = PTRTOUINT64(l->l_addr);
2294 		ki->p_stat = l->l_stat;
2295 		ki->p_flag |= sysctl_map_flags(sysctl_lwpflagmap, l->l_flag);
2296 		ki->p_swtime = l->l_swtime;
2297 		ki->p_slptime = l->l_slptime;
2298 		if (l->l_stat == LSONPROC)
2299 			ki->p_schedflags = l->l_cpu->ci_schedstate.spc_flags;
2300 		else
2301 			ki->p_schedflags = 0;
2302 		ki->p_priority = lwp_eprio(l);
2303 		ki->p_usrpri = l->l_priority;
2304 		if (l->l_wchan)
2305 			strncpy(ki->p_wmesg, l->l_wmesg, sizeof(ki->p_wmesg));
2306 		ki->p_wchan = PTRTOUINT64(l->l_wchan);
2307 		ki->p_cpuid = cpu_index(l->l_cpu);
2308 		lwp_unlock(l);
2309 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
2310 			/* This is hardly correct, but... */
2311 			sigplusset(&l->l_sigpend.sp_set, &ss1);
2312 			sigplusset(&l->l_sigmask, &ss2);
2313 			ki->p_cpticks += l->l_cpticks;
2314 			ki->p_pctcpu += l->l_pctcpu;
2315 			ki->p_estcpu += l->l_estcpu;
2316 		}
2317 	}
2318 	sigplusset(&p->p_sigpend.sp_set, &ss2);
2319 	memcpy(&ki->p_siglist, &ss1, sizeof(ki_sigset_t));
2320 	memcpy(&ki->p_sigmask, &ss2, sizeof(ki_sigset_t));
2321 
2322 	if (p->p_session != NULL) {
2323 		ki->p_sid = p->p_session->s_sid;
2324 		ki->p__pgid = p->p_pgrp->pg_id;
2325 		if (p->p_session->s_ttyvp)
2326 			ki->p_eflag |= EPROC_CTTY;
2327 		if (SESS_LEADER(p))
2328 			ki->p_eflag |= EPROC_SLEADER;
2329 		strncpy(ki->p_login, p->p_session->s_login,
2330 		    min(sizeof ki->p_login - 1, sizeof p->p_session->s_login));
2331 		ki->p_jobc = p->p_pgrp->pg_jobc;
2332 		if ((p->p_lflag & PL_CONTROLT) && (tp = p->p_session->s_ttyp)) {
2333 			ki->p_tdev = tp->t_dev;
2334 			ki->p_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PGID;
2335 			ki->p_tsess = PTRTOUINT64(tp->t_session);
2336 		} else {
2337 			ki->p_tdev = (int32_t)NODEV;
2338 		}
2339 	}
2340 
2341 	if (!P_ZOMBIE(p) && !zombie) {
2342 		ki->p_uvalid = 1;
2343 		ki->p_ustart_sec = p->p_stats->p_start.tv_sec;
2344 		ki->p_ustart_usec = p->p_stats->p_start.tv_usec;
2345 
2346 		calcru(p, &ut, &st, NULL, &rt);
2347 		ki->p_rtime_sec = rt.tv_sec;
2348 		ki->p_rtime_usec = rt.tv_usec;
2349 		ki->p_uutime_sec = ut.tv_sec;
2350 		ki->p_uutime_usec = ut.tv_usec;
2351 		ki->p_ustime_sec = st.tv_sec;
2352 		ki->p_ustime_usec = st.tv_usec;
2353 
2354 		memcpy(&ru, &p->p_stats->p_ru, sizeof(ru));
2355 		ki->p_uru_nvcsw = 0;
2356 		ki->p_uru_nivcsw = 0;
2357 		LIST_FOREACH(l2, &p->p_lwps, l_sibling) {
2358 			ki->p_uru_nvcsw += (l2->l_ncsw - l2->l_nivcsw);
2359 			ki->p_uru_nivcsw += l2->l_nivcsw;
2360 			ruadd(&ru, &l2->l_ru);
2361 		}
2362 		ki->p_uru_maxrss = ru.ru_maxrss;
2363 		ki->p_uru_ixrss = ru.ru_ixrss;
2364 		ki->p_uru_idrss = ru.ru_idrss;
2365 		ki->p_uru_isrss = ru.ru_isrss;
2366 		ki->p_uru_minflt = ru.ru_minflt;
2367 		ki->p_uru_majflt = ru.ru_majflt;
2368 		ki->p_uru_nswap = ru.ru_nswap;
2369 		ki->p_uru_inblock = ru.ru_inblock;
2370 		ki->p_uru_oublock = ru.ru_oublock;
2371 		ki->p_uru_msgsnd = ru.ru_msgsnd;
2372 		ki->p_uru_msgrcv = ru.ru_msgrcv;
2373 		ki->p_uru_nsignals = ru.ru_nsignals;
2374 
2375 		timeradd(&p->p_stats->p_cru.ru_utime,
2376 			 &p->p_stats->p_cru.ru_stime, &ut);
2377 		ki->p_uctime_sec = ut.tv_sec;
2378 		ki->p_uctime_usec = ut.tv_usec;
2379 	}
2380 }
2381