xref: /netbsd-src/sys/kern/kern_proc.c (revision 7788a0781fe6ff2cce37368b4578a7ade0850cb1)
1 /*	$NetBSD: kern_proc.c,v 1.187 2013/06/10 14:53:52 pooka 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 1.187 2013/06/10 14:53:52 pooka 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 
105 #ifdef COMPAT_NETBSD32
106 #include <compat/netbsd32/netbsd32.h>
107 #endif
108 
109 /*
110  * Process lists.
111  */
112 
113 struct proclist		allproc		__cacheline_aligned;
114 struct proclist		zombproc	__cacheline_aligned;
115 
116 kmutex_t *		proc_lock	__cacheline_aligned;
117 
118 /*
119  * pid to proc lookup is done by indexing the pid_table array.
120  * Since pid numbers are only allocated when an empty slot
121  * has been found, there is no need to search any lists ever.
122  * (an orphaned pgrp will lock the slot, a session will lock
123  * the pgrp with the same number.)
124  * If the table is too small it is reallocated with twice the
125  * previous size and the entries 'unzipped' into the two halves.
126  * A linked list of free entries is passed through the pt_proc
127  * field of 'free' items - set odd to be an invalid ptr.
128  */
129 
130 struct pid_table {
131 	struct proc	*pt_proc;
132 	struct pgrp	*pt_pgrp;
133 	pid_t		pt_pid;
134 };
135 #if 1	/* strongly typed cast - should be a noop */
136 static inline uint p2u(struct proc *p) { return (uint)(uintptr_t)p; }
137 #else
138 #define p2u(p) ((uint)p)
139 #endif
140 #define P_VALID(p) (!(p2u(p) & 1))
141 #define P_NEXT(p) (p2u(p) >> 1)
142 #define P_FREE(pid) ((struct proc *)(uintptr_t)((pid) << 1 | 1))
143 
144 /*
145  * Table of process IDs (PIDs).
146  */
147 static struct pid_table *pid_table	__read_mostly;
148 
149 #define	INITIAL_PID_TABLE_SIZE		(1 << 5)
150 
151 /* Table mask, threshold for growing and number of allocated PIDs. */
152 static u_int		pid_tbl_mask	__read_mostly;
153 static u_int		pid_alloc_lim	__read_mostly;
154 static u_int		pid_alloc_cnt	__cacheline_aligned;
155 
156 /* Next free, last free and maximum PIDs. */
157 static u_int		next_free_pt	__cacheline_aligned;
158 static u_int		last_free_pt	__cacheline_aligned;
159 static pid_t		pid_max		__read_mostly;
160 
161 /* Components of the first process -- never freed. */
162 
163 extern struct emul emul_netbsd;	/* defined in kern_exec.c */
164 
165 struct session session0 = {
166 	.s_count = 1,
167 	.s_sid = 0,
168 };
169 struct pgrp pgrp0 = {
170 	.pg_members = LIST_HEAD_INITIALIZER(&pgrp0.pg_members),
171 	.pg_session = &session0,
172 };
173 filedesc_t filedesc0;
174 struct cwdinfo cwdi0 = {
175 	.cwdi_cmask = CMASK,
176 	.cwdi_refcnt = 1,
177 };
178 struct plimit limit0;
179 struct pstats pstat0;
180 struct vmspace vmspace0;
181 struct sigacts sigacts0;
182 struct proc proc0 = {
183 	.p_lwps = LIST_HEAD_INITIALIZER(&proc0.p_lwps),
184 	.p_sigwaiters = LIST_HEAD_INITIALIZER(&proc0.p_sigwaiters),
185 	.p_nlwps = 1,
186 	.p_nrlwps = 1,
187 	.p_nlwpid = 1,		/* must match lwp0.l_lid */
188 	.p_pgrp = &pgrp0,
189 	.p_comm = "system",
190 	/*
191 	 * Set P_NOCLDWAIT so that kernel threads are reparented to init(8)
192 	 * when they exit.  init(8) can easily wait them out for us.
193 	 */
194 	.p_flag = PK_SYSTEM | PK_NOCLDWAIT,
195 	.p_stat = SACTIVE,
196 	.p_nice = NZERO,
197 	.p_emul = &emul_netbsd,
198 	.p_cwdi = &cwdi0,
199 	.p_limit = &limit0,
200 	.p_fd = &filedesc0,
201 	.p_vmspace = &vmspace0,
202 	.p_stats = &pstat0,
203 	.p_sigacts = &sigacts0,
204 };
205 kauth_cred_t cred0;
206 
207 static const int	nofile	= NOFILE;
208 static const int	maxuprc	= MAXUPRC;
209 
210 static int sysctl_doeproc(SYSCTLFN_PROTO);
211 static int sysctl_kern_proc_args(SYSCTLFN_PROTO);
212 static void fill_kproc2(struct proc *, struct kinfo_proc2 *, bool);
213 
214 /*
215  * The process list descriptors, used during pid allocation and
216  * by sysctl.  No locking on this data structure is needed since
217  * it is completely static.
218  */
219 const struct proclist_desc proclists[] = {
220 	{ &allproc	},
221 	{ &zombproc	},
222 	{ NULL		},
223 };
224 
225 static struct pgrp *	pg_remove(pid_t);
226 static void		pg_delete(pid_t);
227 static void		orphanpg(struct pgrp *);
228 
229 static specificdata_domain_t proc_specificdata_domain;
230 
231 static pool_cache_t proc_cache;
232 
233 static kauth_listener_t proc_listener;
234 
235 static int
236 proc_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
237     void *arg0, void *arg1, void *arg2, void *arg3)
238 {
239 	struct proc *p;
240 	int result;
241 
242 	result = KAUTH_RESULT_DEFER;
243 	p = arg0;
244 
245 	switch (action) {
246 	case KAUTH_PROCESS_CANSEE: {
247 		enum kauth_process_req req;
248 
249 		req = (enum kauth_process_req)arg1;
250 
251 		switch (req) {
252 		case KAUTH_REQ_PROCESS_CANSEE_ARGS:
253 		case KAUTH_REQ_PROCESS_CANSEE_ENTRY:
254 		case KAUTH_REQ_PROCESS_CANSEE_OPENFILES:
255 			result = KAUTH_RESULT_ALLOW;
256 
257 			break;
258 
259 		case KAUTH_REQ_PROCESS_CANSEE_ENV:
260 			if (kauth_cred_getuid(cred) !=
261 			    kauth_cred_getuid(p->p_cred) ||
262 			    kauth_cred_getuid(cred) !=
263 			    kauth_cred_getsvuid(p->p_cred))
264 				break;
265 
266 			result = KAUTH_RESULT_ALLOW;
267 
268 			break;
269 
270 		default:
271 			break;
272 		}
273 
274 		break;
275 		}
276 
277 	case KAUTH_PROCESS_FORK: {
278 		int lnprocs = (int)(unsigned long)arg2;
279 
280 		/*
281 		 * Don't allow a nonprivileged user to use the last few
282 		 * processes. The variable lnprocs is the current number of
283 		 * processes, maxproc is the limit.
284 		 */
285 		if (__predict_false((lnprocs >= maxproc - 5)))
286 			break;
287 
288 		result = KAUTH_RESULT_ALLOW;
289 
290 		break;
291 		}
292 
293 	case KAUTH_PROCESS_CORENAME:
294 	case KAUTH_PROCESS_STOPFLAG:
295 		if (proc_uidmatch(cred, p->p_cred) == 0)
296 			result = KAUTH_RESULT_ALLOW;
297 
298 		break;
299 
300 	default:
301 		break;
302 	}
303 
304 	return result;
305 }
306 
307 /*
308  * Initialize global process hashing structures.
309  */
310 void
311 procinit(void)
312 {
313 	const struct proclist_desc *pd;
314 	u_int i;
315 #define	LINK_EMPTY ((PID_MAX + INITIAL_PID_TABLE_SIZE) & ~(INITIAL_PID_TABLE_SIZE - 1))
316 
317 	for (pd = proclists; pd->pd_list != NULL; pd++)
318 		LIST_INIT(pd->pd_list);
319 
320 	proc_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
321 	pid_table = kmem_alloc(INITIAL_PID_TABLE_SIZE
322 	    * sizeof(struct pid_table), KM_SLEEP);
323 	pid_tbl_mask = INITIAL_PID_TABLE_SIZE - 1;
324 	pid_max = PID_MAX;
325 
326 	/* Set free list running through table...
327 	   Preset 'use count' above PID_MAX so we allocate pid 1 next. */
328 	for (i = 0; i <= pid_tbl_mask; i++) {
329 		pid_table[i].pt_proc = P_FREE(LINK_EMPTY + i + 1);
330 		pid_table[i].pt_pgrp = 0;
331 		pid_table[i].pt_pid = 0;
332 	}
333 	/* slot 0 is just grabbed */
334 	next_free_pt = 1;
335 	/* Need to fix last entry. */
336 	last_free_pt = pid_tbl_mask;
337 	pid_table[last_free_pt].pt_proc = P_FREE(LINK_EMPTY);
338 	/* point at which we grow table - to avoid reusing pids too often */
339 	pid_alloc_lim = pid_tbl_mask - 1;
340 #undef LINK_EMPTY
341 
342 	proc_specificdata_domain = specificdata_domain_create();
343 	KASSERT(proc_specificdata_domain != NULL);
344 
345 	proc_cache = pool_cache_init(sizeof(struct proc), 0, 0, 0,
346 	    "procpl", NULL, IPL_NONE, NULL, NULL, NULL);
347 
348 	proc_listener = kauth_listen_scope(KAUTH_SCOPE_PROCESS,
349 	    proc_listener_cb, NULL);
350 }
351 
352 void
353 procinit_sysctl(void)
354 {
355 	static struct sysctllog *clog;
356 
357 	sysctl_createv(&clog, 0, NULL, NULL,
358 		       CTLFLAG_PERMANENT,
359 		       CTLTYPE_NODE, "kern", NULL,
360 		       NULL, 0, NULL, 0,
361 		       CTL_KERN, CTL_EOL);
362 
363 	sysctl_createv(&clog, 0, NULL, NULL,
364 		       CTLFLAG_PERMANENT,
365 		       CTLTYPE_NODE, "proc",
366 		       SYSCTL_DESCR("System-wide process information"),
367 		       sysctl_doeproc, 0, NULL, 0,
368 		       CTL_KERN, KERN_PROC, CTL_EOL);
369 	sysctl_createv(&clog, 0, NULL, NULL,
370 		       CTLFLAG_PERMANENT,
371 		       CTLTYPE_NODE, "proc2",
372 		       SYSCTL_DESCR("Machine-independent process information"),
373 		       sysctl_doeproc, 0, NULL, 0,
374 		       CTL_KERN, KERN_PROC2, CTL_EOL);
375 	sysctl_createv(&clog, 0, NULL, NULL,
376 		       CTLFLAG_PERMANENT,
377 		       CTLTYPE_NODE, "proc_args",
378 		       SYSCTL_DESCR("Process argument information"),
379 		       sysctl_kern_proc_args, 0, NULL, 0,
380 		       CTL_KERN, KERN_PROC_ARGS, CTL_EOL);
381 
382 	/*
383 	  "nodes" under these:
384 
385 	  KERN_PROC_ALL
386 	  KERN_PROC_PID pid
387 	  KERN_PROC_PGRP pgrp
388 	  KERN_PROC_SESSION sess
389 	  KERN_PROC_TTY tty
390 	  KERN_PROC_UID uid
391 	  KERN_PROC_RUID uid
392 	  KERN_PROC_GID gid
393 	  KERN_PROC_RGID gid
394 
395 	  all in all, probably not worth the effort...
396 	*/
397 }
398 
399 /*
400  * Initialize process 0.
401  */
402 void
403 proc0_init(void)
404 {
405 	struct proc *p;
406 	struct pgrp *pg;
407 	struct rlimit *rlim;
408 	rlim_t lim;
409 	int i;
410 
411 	p = &proc0;
412 	pg = &pgrp0;
413 
414 	mutex_init(&p->p_stmutex, MUTEX_DEFAULT, IPL_HIGH);
415 	mutex_init(&p->p_auxlock, MUTEX_DEFAULT, IPL_NONE);
416 	p->p_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
417 
418 	rw_init(&p->p_reflock);
419 	cv_init(&p->p_waitcv, "wait");
420 	cv_init(&p->p_lwpcv, "lwpwait");
421 
422 	LIST_INSERT_HEAD(&p->p_lwps, &lwp0, l_sibling);
423 
424 	pid_table[0].pt_proc = p;
425 	LIST_INSERT_HEAD(&allproc, p, p_list);
426 
427 	pid_table[0].pt_pgrp = pg;
428 	LIST_INSERT_HEAD(&pg->pg_members, p, p_pglist);
429 
430 #ifdef __HAVE_SYSCALL_INTERN
431 	(*p->p_emul->e_syscall_intern)(p);
432 #endif
433 
434 	/* Create credentials. */
435 	cred0 = kauth_cred_alloc();
436 	p->p_cred = cred0;
437 
438 	/* Create the CWD info. */
439 	rw_init(&cwdi0.cwdi_lock);
440 
441 	/* Create the limits structures. */
442 	mutex_init(&limit0.pl_lock, MUTEX_DEFAULT, IPL_NONE);
443 
444 	rlim = limit0.pl_rlimit;
445 	for (i = 0; i < __arraycount(limit0.pl_rlimit); i++) {
446 		rlim[i].rlim_cur = RLIM_INFINITY;
447 		rlim[i].rlim_max = RLIM_INFINITY;
448 	}
449 
450 	rlim[RLIMIT_NOFILE].rlim_max = maxfiles;
451 	rlim[RLIMIT_NOFILE].rlim_cur = maxfiles < nofile ? maxfiles : nofile;
452 
453 	rlim[RLIMIT_NPROC].rlim_max = maxproc;
454 	rlim[RLIMIT_NPROC].rlim_cur = maxproc < maxuprc ? maxproc : maxuprc;
455 
456 	lim = MIN(VM_MAXUSER_ADDRESS, ctob((rlim_t)uvmexp.free));
457 	rlim[RLIMIT_RSS].rlim_max = lim;
458 	rlim[RLIMIT_MEMLOCK].rlim_max = lim;
459 	rlim[RLIMIT_MEMLOCK].rlim_cur = lim / 3;
460 
461 	rlim[RLIMIT_NTHR].rlim_max = maxlwp;
462 	rlim[RLIMIT_NTHR].rlim_cur = maxlwp < maxuprc ? maxlwp : maxuprc;
463 
464 	/* Note that default core name has zero length. */
465 	limit0.pl_corename = defcorename;
466 	limit0.pl_cnlen = 0;
467 	limit0.pl_refcnt = 1;
468 	limit0.pl_writeable = false;
469 	limit0.pl_sv_limit = NULL;
470 
471 	/* Configure virtual memory system, set vm rlimits. */
472 	uvm_init_limits(p);
473 
474 	/* Initialize file descriptor table for proc0. */
475 	fd_init(&filedesc0);
476 
477 	/*
478 	 * Initialize proc0's vmspace, which uses the kernel pmap.
479 	 * All kernel processes (which never have user space mappings)
480 	 * share proc0's vmspace, and thus, the kernel pmap.
481 	 */
482 	uvmspace_init(&vmspace0, pmap_kernel(), round_page(VM_MIN_ADDRESS),
483 	    trunc_page(VM_MAX_ADDRESS));
484 
485 	/* Initialize signal state for proc0. XXX IPL_SCHED */
486 	mutex_init(&p->p_sigacts->sa_mutex, MUTEX_DEFAULT, IPL_SCHED);
487 	siginit(p);
488 
489 	proc_initspecific(p);
490 	kdtrace_proc_ctor(NULL, p);
491 }
492 
493 /*
494  * Session reference counting.
495  */
496 
497 void
498 proc_sesshold(struct session *ss)
499 {
500 
501 	KASSERT(mutex_owned(proc_lock));
502 	ss->s_count++;
503 }
504 
505 void
506 proc_sessrele(struct session *ss)
507 {
508 
509 	KASSERT(mutex_owned(proc_lock));
510 	/*
511 	 * We keep the pgrp with the same id as the session in order to
512 	 * stop a process being given the same pid.  Since the pgrp holds
513 	 * a reference to the session, it must be a 'zombie' pgrp by now.
514 	 */
515 	if (--ss->s_count == 0) {
516 		struct pgrp *pg;
517 
518 		pg = pg_remove(ss->s_sid);
519 		mutex_exit(proc_lock);
520 
521 		kmem_free(pg, sizeof(struct pgrp));
522 		kmem_free(ss, sizeof(struct session));
523 	} else {
524 		mutex_exit(proc_lock);
525 	}
526 }
527 
528 /*
529  * Check that the specified process group is in the session of the
530  * specified process.
531  * Treats -ve ids as process ids.
532  * Used to validate TIOCSPGRP requests.
533  */
534 int
535 pgid_in_session(struct proc *p, pid_t pg_id)
536 {
537 	struct pgrp *pgrp;
538 	struct session *session;
539 	int error;
540 
541 	mutex_enter(proc_lock);
542 	if (pg_id < 0) {
543 		struct proc *p1 = proc_find(-pg_id);
544 		if (p1 == NULL) {
545 			error = EINVAL;
546 			goto fail;
547 		}
548 		pgrp = p1->p_pgrp;
549 	} else {
550 		pgrp = pgrp_find(pg_id);
551 		if (pgrp == NULL) {
552 			error = EINVAL;
553 			goto fail;
554 		}
555 	}
556 	session = pgrp->pg_session;
557 	error = (session != p->p_pgrp->pg_session) ? EPERM : 0;
558 fail:
559 	mutex_exit(proc_lock);
560 	return error;
561 }
562 
563 /*
564  * p_inferior: is p an inferior of q?
565  */
566 static inline bool
567 p_inferior(struct proc *p, struct proc *q)
568 {
569 
570 	KASSERT(mutex_owned(proc_lock));
571 
572 	for (; p != q; p = p->p_pptr)
573 		if (p->p_pid == 0)
574 			return false;
575 	return true;
576 }
577 
578 /*
579  * proc_find: locate a process by the ID.
580  *
581  * => Must be called with proc_lock held.
582  */
583 proc_t *
584 proc_find_raw(pid_t pid)
585 {
586 	struct pid_table *pt;
587 	proc_t *p;
588 
589 	KASSERT(mutex_owned(proc_lock));
590 	pt = &pid_table[pid & pid_tbl_mask];
591 	p = pt->pt_proc;
592 	if (__predict_false(!P_VALID(p) || pt->pt_pid != pid)) {
593 		return NULL;
594 	}
595 	return p;
596 }
597 
598 proc_t *
599 proc_find(pid_t pid)
600 {
601 	proc_t *p;
602 
603 	p = proc_find_raw(pid);
604 	if (__predict_false(p == NULL)) {
605 		return NULL;
606 	}
607 
608 	/*
609 	 * Only allow live processes to be found by PID.
610 	 * XXX: p_stat might change, since unlocked.
611 	 */
612 	if (__predict_true(p->p_stat == SACTIVE || p->p_stat == SSTOP)) {
613 		return p;
614 	}
615 	return NULL;
616 }
617 
618 /*
619  * pgrp_find: locate a process group by the ID.
620  *
621  * => Must be called with proc_lock held.
622  */
623 struct pgrp *
624 pgrp_find(pid_t pgid)
625 {
626 	struct pgrp *pg;
627 
628 	KASSERT(mutex_owned(proc_lock));
629 
630 	pg = pid_table[pgid & pid_tbl_mask].pt_pgrp;
631 
632 	/*
633 	 * Cannot look up a process group that only exists because the
634 	 * session has not died yet (traditional).
635 	 */
636 	if (pg == NULL || pg->pg_id != pgid || LIST_EMPTY(&pg->pg_members)) {
637 		return NULL;
638 	}
639 	return pg;
640 }
641 
642 static void
643 expand_pid_table(void)
644 {
645 	size_t pt_size, tsz;
646 	struct pid_table *n_pt, *new_pt;
647 	struct proc *proc;
648 	struct pgrp *pgrp;
649 	pid_t pid, rpid;
650 	u_int i;
651 	uint new_pt_mask;
652 
653 	pt_size = pid_tbl_mask + 1;
654 	tsz = pt_size * 2 * sizeof(struct pid_table);
655 	new_pt = kmem_alloc(tsz, KM_SLEEP);
656 	new_pt_mask = pt_size * 2 - 1;
657 
658 	mutex_enter(proc_lock);
659 	if (pt_size != pid_tbl_mask + 1) {
660 		/* Another process beat us to it... */
661 		mutex_exit(proc_lock);
662 		kmem_free(new_pt, tsz);
663 		return;
664 	}
665 
666 	/*
667 	 * Copy entries from old table into new one.
668 	 * If 'pid' is 'odd' we need to place in the upper half,
669 	 * even pid's to the lower half.
670 	 * Free items stay in the low half so we don't have to
671 	 * fixup the reference to them.
672 	 * We stuff free items on the front of the freelist
673 	 * because we can't write to unmodified entries.
674 	 * Processing the table backwards maintains a semblance
675 	 * of issuing pid numbers that increase with time.
676 	 */
677 	i = pt_size - 1;
678 	n_pt = new_pt + i;
679 	for (; ; i--, n_pt--) {
680 		proc = pid_table[i].pt_proc;
681 		pgrp = pid_table[i].pt_pgrp;
682 		if (!P_VALID(proc)) {
683 			/* Up 'use count' so that link is valid */
684 			pid = (P_NEXT(proc) + pt_size) & ~pt_size;
685 			rpid = 0;
686 			proc = P_FREE(pid);
687 			if (pgrp)
688 				pid = pgrp->pg_id;
689 		} else {
690 			pid = pid_table[i].pt_pid;
691 			rpid = pid;
692 		}
693 
694 		/* Save entry in appropriate half of table */
695 		n_pt[pid & pt_size].pt_proc = proc;
696 		n_pt[pid & pt_size].pt_pgrp = pgrp;
697 		n_pt[pid & pt_size].pt_pid = rpid;
698 
699 		/* Put other piece on start of free list */
700 		pid = (pid ^ pt_size) & ~pid_tbl_mask;
701 		n_pt[pid & pt_size].pt_proc =
702 			P_FREE((pid & ~pt_size) | next_free_pt);
703 		n_pt[pid & pt_size].pt_pgrp = 0;
704 		n_pt[pid & pt_size].pt_pid = 0;
705 
706 		next_free_pt = i | (pid & pt_size);
707 		if (i == 0)
708 			break;
709 	}
710 
711 	/* Save old table size and switch tables */
712 	tsz = pt_size * sizeof(struct pid_table);
713 	n_pt = pid_table;
714 	pid_table = new_pt;
715 	pid_tbl_mask = new_pt_mask;
716 
717 	/*
718 	 * pid_max starts as PID_MAX (= 30000), once we have 16384
719 	 * allocated pids we need it to be larger!
720 	 */
721 	if (pid_tbl_mask > PID_MAX) {
722 		pid_max = pid_tbl_mask * 2 + 1;
723 		pid_alloc_lim |= pid_alloc_lim << 1;
724 	} else
725 		pid_alloc_lim <<= 1;	/* doubles number of free slots... */
726 
727 	mutex_exit(proc_lock);
728 	kmem_free(n_pt, tsz);
729 }
730 
731 struct proc *
732 proc_alloc(void)
733 {
734 	struct proc *p;
735 
736 	p = pool_cache_get(proc_cache, PR_WAITOK);
737 	p->p_stat = SIDL;			/* protect against others */
738 	proc_initspecific(p);
739 	kdtrace_proc_ctor(NULL, p);
740 	p->p_pid = -1;
741 	proc_alloc_pid(p);
742 	return p;
743 }
744 
745 /*
746  * proc_alloc_pid: allocate PID and record the given proc 'p' so that
747  * proc_find_raw() can find it by the PID.
748  */
749 
750 pid_t
751 proc_alloc_pid(struct proc *p)
752 {
753 	struct pid_table *pt;
754 	pid_t pid;
755 	int nxt;
756 
757 	for (;;expand_pid_table()) {
758 		if (__predict_false(pid_alloc_cnt >= pid_alloc_lim))
759 			/* ensure pids cycle through 2000+ values */
760 			continue;
761 		mutex_enter(proc_lock);
762 		pt = &pid_table[next_free_pt];
763 #ifdef DIAGNOSTIC
764 		if (__predict_false(P_VALID(pt->pt_proc) || pt->pt_pgrp))
765 			panic("proc_alloc: slot busy");
766 #endif
767 		nxt = P_NEXT(pt->pt_proc);
768 		if (nxt & pid_tbl_mask)
769 			break;
770 		/* Table full - expand (NB last entry not used....) */
771 		mutex_exit(proc_lock);
772 	}
773 
774 	/* pid is 'saved use count' + 'size' + entry */
775 	pid = (nxt & ~pid_tbl_mask) + pid_tbl_mask + 1 + next_free_pt;
776 	if ((uint)pid > (uint)pid_max)
777 		pid &= pid_tbl_mask;
778 	next_free_pt = nxt & pid_tbl_mask;
779 
780 	/* Grab table slot */
781 	pt->pt_proc = p;
782 
783 	KASSERT(pt->pt_pid == 0);
784 	pt->pt_pid = pid;
785 	if (p->p_pid == -1) {
786 		p->p_pid = pid;
787 	}
788 	pid_alloc_cnt++;
789 	mutex_exit(proc_lock);
790 
791 	return pid;
792 }
793 
794 /*
795  * Free a process id - called from proc_free (in kern_exit.c)
796  *
797  * Called with the proc_lock held.
798  */
799 void
800 proc_free_pid(pid_t pid)
801 {
802 	struct pid_table *pt;
803 
804 	KASSERT(mutex_owned(proc_lock));
805 
806 	pt = &pid_table[pid & pid_tbl_mask];
807 
808 	/* save pid use count in slot */
809 	pt->pt_proc = P_FREE(pid & ~pid_tbl_mask);
810 	KASSERT(pt->pt_pid == pid);
811 	pt->pt_pid = 0;
812 
813 	if (pt->pt_pgrp == NULL) {
814 		/* link last freed entry onto ours */
815 		pid &= pid_tbl_mask;
816 		pt = &pid_table[last_free_pt];
817 		pt->pt_proc = P_FREE(P_NEXT(pt->pt_proc) | pid);
818 		pt->pt_pid = 0;
819 		last_free_pt = pid;
820 		pid_alloc_cnt--;
821 	}
822 
823 	atomic_dec_uint(&nprocs);
824 }
825 
826 void
827 proc_free_mem(struct proc *p)
828 {
829 
830 	kdtrace_proc_dtor(NULL, p);
831 	pool_cache_put(proc_cache, p);
832 }
833 
834 /*
835  * proc_enterpgrp: move p to a new or existing process group (and session).
836  *
837  * If we are creating a new pgrp, the pgid should equal
838  * the calling process' pid.
839  * If is only valid to enter a process group that is in the session
840  * of the process.
841  * Also mksess should only be set if we are creating a process group
842  *
843  * Only called from sys_setsid, sys_setpgid and posix_spawn/spawn_return.
844  */
845 int
846 proc_enterpgrp(struct proc *curp, pid_t pid, pid_t pgid, bool mksess)
847 {
848 	struct pgrp *new_pgrp, *pgrp;
849 	struct session *sess;
850 	struct proc *p;
851 	int rval;
852 	pid_t pg_id = NO_PGID;
853 
854 	sess = mksess ? kmem_alloc(sizeof(*sess), KM_SLEEP) : NULL;
855 
856 	/* Allocate data areas we might need before doing any validity checks */
857 	mutex_enter(proc_lock);		/* Because pid_table might change */
858 	if (pid_table[pgid & pid_tbl_mask].pt_pgrp == 0) {
859 		mutex_exit(proc_lock);
860 		new_pgrp = kmem_alloc(sizeof(*new_pgrp), KM_SLEEP);
861 		mutex_enter(proc_lock);
862 	} else
863 		new_pgrp = NULL;
864 	rval = EPERM;	/* most common error (to save typing) */
865 
866 	/* Check pgrp exists or can be created */
867 	pgrp = pid_table[pgid & pid_tbl_mask].pt_pgrp;
868 	if (pgrp != NULL && pgrp->pg_id != pgid)
869 		goto done;
870 
871 	/* Can only set another process under restricted circumstances. */
872 	if (pid != curp->p_pid) {
873 		/* Must exist and be one of our children... */
874 		p = proc_find(pid);
875 		if (p == NULL || !p_inferior(p, curp)) {
876 			rval = ESRCH;
877 			goto done;
878 		}
879 		/* ... in the same session... */
880 		if (sess != NULL || p->p_session != curp->p_session)
881 			goto done;
882 		/* ... existing pgid must be in same session ... */
883 		if (pgrp != NULL && pgrp->pg_session != p->p_session)
884 			goto done;
885 		/* ... and not done an exec. */
886 		if (p->p_flag & PK_EXEC) {
887 			rval = EACCES;
888 			goto done;
889 		}
890 	} else {
891 		/* ... setsid() cannot re-enter a pgrp */
892 		if (mksess && (curp->p_pgid == curp->p_pid ||
893 		    pgrp_find(curp->p_pid)))
894 			goto done;
895 		p = curp;
896 	}
897 
898 	/* Changing the process group/session of a session
899 	   leader is definitely off limits. */
900 	if (SESS_LEADER(p)) {
901 		if (sess == NULL && p->p_pgrp == pgrp)
902 			/* unless it's a definite noop */
903 			rval = 0;
904 		goto done;
905 	}
906 
907 	/* Can only create a process group with id of process */
908 	if (pgrp == NULL && pgid != pid)
909 		goto done;
910 
911 	/* Can only create a session if creating pgrp */
912 	if (sess != NULL && pgrp != NULL)
913 		goto done;
914 
915 	/* Check we allocated memory for a pgrp... */
916 	if (pgrp == NULL && new_pgrp == NULL)
917 		goto done;
918 
919 	/* Don't attach to 'zombie' pgrp */
920 	if (pgrp != NULL && LIST_EMPTY(&pgrp->pg_members))
921 		goto done;
922 
923 	/* Expect to succeed now */
924 	rval = 0;
925 
926 	if (pgrp == p->p_pgrp)
927 		/* nothing to do */
928 		goto done;
929 
930 	/* Ok all setup, link up required structures */
931 
932 	if (pgrp == NULL) {
933 		pgrp = new_pgrp;
934 		new_pgrp = NULL;
935 		if (sess != NULL) {
936 			sess->s_sid = p->p_pid;
937 			sess->s_leader = p;
938 			sess->s_count = 1;
939 			sess->s_ttyvp = NULL;
940 			sess->s_ttyp = NULL;
941 			sess->s_flags = p->p_session->s_flags & ~S_LOGIN_SET;
942 			memcpy(sess->s_login, p->p_session->s_login,
943 			    sizeof(sess->s_login));
944 			p->p_lflag &= ~PL_CONTROLT;
945 		} else {
946 			sess = p->p_pgrp->pg_session;
947 			proc_sesshold(sess);
948 		}
949 		pgrp->pg_session = sess;
950 		sess = NULL;
951 
952 		pgrp->pg_id = pgid;
953 		LIST_INIT(&pgrp->pg_members);
954 #ifdef DIAGNOSTIC
955 		if (__predict_false(pid_table[pgid & pid_tbl_mask].pt_pgrp))
956 			panic("enterpgrp: pgrp table slot in use");
957 		if (__predict_false(mksess && p != curp))
958 			panic("enterpgrp: mksession and p != curproc");
959 #endif
960 		pid_table[pgid & pid_tbl_mask].pt_pgrp = pgrp;
961 		pgrp->pg_jobc = 0;
962 	}
963 
964 	/*
965 	 * Adjust eligibility of affected pgrps to participate in job control.
966 	 * Increment eligibility counts before decrementing, otherwise we
967 	 * could reach 0 spuriously during the first call.
968 	 */
969 	fixjobc(p, pgrp, 1);
970 	fixjobc(p, p->p_pgrp, 0);
971 
972 	/* Interlock with ttread(). */
973 	mutex_spin_enter(&tty_lock);
974 
975 	/* Move process to requested group. */
976 	LIST_REMOVE(p, p_pglist);
977 	if (LIST_EMPTY(&p->p_pgrp->pg_members))
978 		/* defer delete until we've dumped the lock */
979 		pg_id = p->p_pgrp->pg_id;
980 	p->p_pgrp = pgrp;
981 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
982 
983 	/* Done with the swap; we can release the tty mutex. */
984 	mutex_spin_exit(&tty_lock);
985 
986     done:
987 	if (pg_id != NO_PGID) {
988 		/* Releases proc_lock. */
989 		pg_delete(pg_id);
990 	} else {
991 		mutex_exit(proc_lock);
992 	}
993 	if (sess != NULL)
994 		kmem_free(sess, sizeof(*sess));
995 	if (new_pgrp != NULL)
996 		kmem_free(new_pgrp, sizeof(*new_pgrp));
997 #ifdef DEBUG_PGRP
998 	if (__predict_false(rval))
999 		printf("enterpgrp(%d,%d,%d), curproc %d, rval %d\n",
1000 			pid, pgid, mksess, curp->p_pid, rval);
1001 #endif
1002 	return rval;
1003 }
1004 
1005 /*
1006  * proc_leavepgrp: remove a process from its process group.
1007  *  => must be called with the proc_lock held, which will be released;
1008  */
1009 void
1010 proc_leavepgrp(struct proc *p)
1011 {
1012 	struct pgrp *pgrp;
1013 
1014 	KASSERT(mutex_owned(proc_lock));
1015 
1016 	/* Interlock with ttread() */
1017 	mutex_spin_enter(&tty_lock);
1018 	pgrp = p->p_pgrp;
1019 	LIST_REMOVE(p, p_pglist);
1020 	p->p_pgrp = NULL;
1021 	mutex_spin_exit(&tty_lock);
1022 
1023 	if (LIST_EMPTY(&pgrp->pg_members)) {
1024 		/* Releases proc_lock. */
1025 		pg_delete(pgrp->pg_id);
1026 	} else {
1027 		mutex_exit(proc_lock);
1028 	}
1029 }
1030 
1031 /*
1032  * pg_remove: remove a process group from the table.
1033  *  => must be called with the proc_lock held;
1034  *  => returns process group to free;
1035  */
1036 static struct pgrp *
1037 pg_remove(pid_t pg_id)
1038 {
1039 	struct pgrp *pgrp;
1040 	struct pid_table *pt;
1041 
1042 	KASSERT(mutex_owned(proc_lock));
1043 
1044 	pt = &pid_table[pg_id & pid_tbl_mask];
1045 	pgrp = pt->pt_pgrp;
1046 
1047 	KASSERT(pgrp != NULL);
1048 	KASSERT(pgrp->pg_id == pg_id);
1049 	KASSERT(LIST_EMPTY(&pgrp->pg_members));
1050 
1051 	pt->pt_pgrp = NULL;
1052 
1053 	if (!P_VALID(pt->pt_proc)) {
1054 		/* Orphaned pgrp, put slot onto free list. */
1055 		KASSERT((P_NEXT(pt->pt_proc) & pid_tbl_mask) == 0);
1056 		pg_id &= pid_tbl_mask;
1057 		pt = &pid_table[last_free_pt];
1058 		pt->pt_proc = P_FREE(P_NEXT(pt->pt_proc) | pg_id);
1059 		KASSERT(pt->pt_pid == 0);
1060 		last_free_pt = pg_id;
1061 		pid_alloc_cnt--;
1062 	}
1063 	return pgrp;
1064 }
1065 
1066 /*
1067  * pg_delete: delete and free a process group.
1068  *  => must be called with the proc_lock held, which will be released.
1069  */
1070 static void
1071 pg_delete(pid_t pg_id)
1072 {
1073 	struct pgrp *pg;
1074 	struct tty *ttyp;
1075 	struct session *ss;
1076 
1077 	KASSERT(mutex_owned(proc_lock));
1078 
1079 	pg = pid_table[pg_id & pid_tbl_mask].pt_pgrp;
1080 	if (pg == NULL || pg->pg_id != pg_id || !LIST_EMPTY(&pg->pg_members)) {
1081 		mutex_exit(proc_lock);
1082 		return;
1083 	}
1084 
1085 	ss = pg->pg_session;
1086 
1087 	/* Remove reference (if any) from tty to this process group */
1088 	mutex_spin_enter(&tty_lock);
1089 	ttyp = ss->s_ttyp;
1090 	if (ttyp != NULL && ttyp->t_pgrp == pg) {
1091 		ttyp->t_pgrp = NULL;
1092 		KASSERT(ttyp->t_session == ss);
1093 	}
1094 	mutex_spin_exit(&tty_lock);
1095 
1096 	/*
1097 	 * The leading process group in a session is freed by proc_sessrele(),
1098 	 * if last reference.  Note: proc_sessrele() releases proc_lock.
1099 	 */
1100 	pg = (ss->s_sid != pg->pg_id) ? pg_remove(pg_id) : NULL;
1101 	proc_sessrele(ss);
1102 
1103 	if (pg != NULL) {
1104 		/* Free it, if was not done by proc_sessrele(). */
1105 		kmem_free(pg, sizeof(struct pgrp));
1106 	}
1107 }
1108 
1109 /*
1110  * Adjust pgrp jobc counters when specified process changes process group.
1111  * We count the number of processes in each process group that "qualify"
1112  * the group for terminal job control (those with a parent in a different
1113  * process group of the same session).  If that count reaches zero, the
1114  * process group becomes orphaned.  Check both the specified process'
1115  * process group and that of its children.
1116  * entering == 0 => p is leaving specified group.
1117  * entering == 1 => p is entering specified group.
1118  *
1119  * Call with proc_lock held.
1120  */
1121 void
1122 fixjobc(struct proc *p, struct pgrp *pgrp, int entering)
1123 {
1124 	struct pgrp *hispgrp;
1125 	struct session *mysession = pgrp->pg_session;
1126 	struct proc *child;
1127 
1128 	KASSERT(mutex_owned(proc_lock));
1129 
1130 	/*
1131 	 * Check p's parent to see whether p qualifies its own process
1132 	 * group; if so, adjust count for p's process group.
1133 	 */
1134 	hispgrp = p->p_pptr->p_pgrp;
1135 	if (hispgrp != pgrp && hispgrp->pg_session == mysession) {
1136 		if (entering) {
1137 			pgrp->pg_jobc++;
1138 			p->p_lflag &= ~PL_ORPHANPG;
1139 		} else if (--pgrp->pg_jobc == 0)
1140 			orphanpg(pgrp);
1141 	}
1142 
1143 	/*
1144 	 * Check this process' children to see whether they qualify
1145 	 * their process groups; if so, adjust counts for children's
1146 	 * process groups.
1147 	 */
1148 	LIST_FOREACH(child, &p->p_children, p_sibling) {
1149 		hispgrp = child->p_pgrp;
1150 		if (hispgrp != pgrp && hispgrp->pg_session == mysession &&
1151 		    !P_ZOMBIE(child)) {
1152 			if (entering) {
1153 				child->p_lflag &= ~PL_ORPHANPG;
1154 				hispgrp->pg_jobc++;
1155 			} else if (--hispgrp->pg_jobc == 0)
1156 				orphanpg(hispgrp);
1157 		}
1158 	}
1159 }
1160 
1161 /*
1162  * A process group has become orphaned;
1163  * if there are any stopped processes in the group,
1164  * hang-up all process in that group.
1165  *
1166  * Call with proc_lock held.
1167  */
1168 static void
1169 orphanpg(struct pgrp *pg)
1170 {
1171 	struct proc *p;
1172 
1173 	KASSERT(mutex_owned(proc_lock));
1174 
1175 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
1176 		if (p->p_stat == SSTOP) {
1177 			p->p_lflag |= PL_ORPHANPG;
1178 			psignal(p, SIGHUP);
1179 			psignal(p, SIGCONT);
1180 		}
1181 	}
1182 }
1183 
1184 #ifdef DDB
1185 #include <ddb/db_output.h>
1186 void pidtbl_dump(void);
1187 void
1188 pidtbl_dump(void)
1189 {
1190 	struct pid_table *pt;
1191 	struct proc *p;
1192 	struct pgrp *pgrp;
1193 	int id;
1194 
1195 	db_printf("pid table %p size %x, next %x, last %x\n",
1196 		pid_table, pid_tbl_mask+1,
1197 		next_free_pt, last_free_pt);
1198 	for (pt = pid_table, id = 0; id <= pid_tbl_mask; id++, pt++) {
1199 		p = pt->pt_proc;
1200 		if (!P_VALID(p) && !pt->pt_pgrp)
1201 			continue;
1202 		db_printf("  id %x: ", id);
1203 		if (P_VALID(p))
1204 			db_printf("slotpid %d proc %p id %d (0x%x) %s\n",
1205 				pt->pt_pid, p, p->p_pid, p->p_pid, p->p_comm);
1206 		else
1207 			db_printf("next %x use %x\n",
1208 				P_NEXT(p) & pid_tbl_mask,
1209 				P_NEXT(p) & ~pid_tbl_mask);
1210 		if ((pgrp = pt->pt_pgrp)) {
1211 			db_printf("\tsession %p, sid %d, count %d, login %s\n",
1212 			    pgrp->pg_session, pgrp->pg_session->s_sid,
1213 			    pgrp->pg_session->s_count,
1214 			    pgrp->pg_session->s_login);
1215 			db_printf("\tpgrp %p, pg_id %d, pg_jobc %d, members %p\n",
1216 			    pgrp, pgrp->pg_id, pgrp->pg_jobc,
1217 			    LIST_FIRST(&pgrp->pg_members));
1218 			LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
1219 				db_printf("\t\tpid %d addr %p pgrp %p %s\n",
1220 				    p->p_pid, p, p->p_pgrp, p->p_comm);
1221 			}
1222 		}
1223 	}
1224 }
1225 #endif /* DDB */
1226 
1227 #ifdef KSTACK_CHECK_MAGIC
1228 
1229 #define	KSTACK_MAGIC	0xdeadbeaf
1230 
1231 /* XXX should be per process basis? */
1232 static int	kstackleftmin = KSTACK_SIZE;
1233 static int	kstackleftthres = KSTACK_SIZE / 8;
1234 
1235 void
1236 kstack_setup_magic(const struct lwp *l)
1237 {
1238 	uint32_t *ip;
1239 	uint32_t const *end;
1240 
1241 	KASSERT(l != NULL);
1242 	KASSERT(l != &lwp0);
1243 
1244 	/*
1245 	 * fill all the stack with magic number
1246 	 * so that later modification on it can be detected.
1247 	 */
1248 	ip = (uint32_t *)KSTACK_LOWEST_ADDR(l);
1249 	end = (uint32_t *)((char *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE);
1250 	for (; ip < end; ip++) {
1251 		*ip = KSTACK_MAGIC;
1252 	}
1253 }
1254 
1255 void
1256 kstack_check_magic(const struct lwp *l)
1257 {
1258 	uint32_t const *ip, *end;
1259 	int stackleft;
1260 
1261 	KASSERT(l != NULL);
1262 
1263 	/* don't check proc0 */ /*XXX*/
1264 	if (l == &lwp0)
1265 		return;
1266 
1267 #ifdef __MACHINE_STACK_GROWS_UP
1268 	/* stack grows upwards (eg. hppa) */
1269 	ip = (uint32_t *)((void *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE);
1270 	end = (uint32_t *)KSTACK_LOWEST_ADDR(l);
1271 	for (ip--; ip >= end; ip--)
1272 		if (*ip != KSTACK_MAGIC)
1273 			break;
1274 
1275 	stackleft = (void *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE - (void *)ip;
1276 #else /* __MACHINE_STACK_GROWS_UP */
1277 	/* stack grows downwards (eg. i386) */
1278 	ip = (uint32_t *)KSTACK_LOWEST_ADDR(l);
1279 	end = (uint32_t *)((char *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE);
1280 	for (; ip < end; ip++)
1281 		if (*ip != KSTACK_MAGIC)
1282 			break;
1283 
1284 	stackleft = ((const char *)ip) - (const char *)KSTACK_LOWEST_ADDR(l);
1285 #endif /* __MACHINE_STACK_GROWS_UP */
1286 
1287 	if (kstackleftmin > stackleft) {
1288 		kstackleftmin = stackleft;
1289 		if (stackleft < kstackleftthres)
1290 			printf("warning: kernel stack left %d bytes"
1291 			    "(pid %u:lid %u)\n", stackleft,
1292 			    (u_int)l->l_proc->p_pid, (u_int)l->l_lid);
1293 	}
1294 
1295 	if (stackleft <= 0) {
1296 		panic("magic on the top of kernel stack changed for "
1297 		    "pid %u, lid %u: maybe kernel stack overflow",
1298 		    (u_int)l->l_proc->p_pid, (u_int)l->l_lid);
1299 	}
1300 }
1301 #endif /* KSTACK_CHECK_MAGIC */
1302 
1303 int
1304 proclist_foreach_call(struct proclist *list,
1305     int (*callback)(struct proc *, void *arg), void *arg)
1306 {
1307 	struct proc marker;
1308 	struct proc *p;
1309 	int ret = 0;
1310 
1311 	marker.p_flag = PK_MARKER;
1312 	mutex_enter(proc_lock);
1313 	for (p = LIST_FIRST(list); ret == 0 && p != NULL;) {
1314 		if (p->p_flag & PK_MARKER) {
1315 			p = LIST_NEXT(p, p_list);
1316 			continue;
1317 		}
1318 		LIST_INSERT_AFTER(p, &marker, p_list);
1319 		ret = (*callback)(p, arg);
1320 		KASSERT(mutex_owned(proc_lock));
1321 		p = LIST_NEXT(&marker, p_list);
1322 		LIST_REMOVE(&marker, p_list);
1323 	}
1324 	mutex_exit(proc_lock);
1325 
1326 	return ret;
1327 }
1328 
1329 int
1330 proc_vmspace_getref(struct proc *p, struct vmspace **vm)
1331 {
1332 
1333 	/* XXXCDC: how should locking work here? */
1334 
1335 	/* curproc exception is for coredump. */
1336 
1337 	if ((p != curproc && (p->p_sflag & PS_WEXIT) != 0) ||
1338 	    (p->p_vmspace->vm_refcnt < 1)) { /* XXX */
1339 		return EFAULT;
1340 	}
1341 
1342 	uvmspace_addref(p->p_vmspace);
1343 	*vm = p->p_vmspace;
1344 
1345 	return 0;
1346 }
1347 
1348 /*
1349  * Acquire a write lock on the process credential.
1350  */
1351 void
1352 proc_crmod_enter(void)
1353 {
1354 	struct lwp *l = curlwp;
1355 	struct proc *p = l->l_proc;
1356 	kauth_cred_t oc;
1357 
1358 	/* Reset what needs to be reset in plimit. */
1359 	if (p->p_limit->pl_corename != defcorename) {
1360 		lim_setcorename(p, defcorename, 0);
1361 	}
1362 
1363 	mutex_enter(p->p_lock);
1364 
1365 	/* Ensure the LWP cached credentials are up to date. */
1366 	if ((oc = l->l_cred) != p->p_cred) {
1367 		kauth_cred_hold(p->p_cred);
1368 		l->l_cred = p->p_cred;
1369 		kauth_cred_free(oc);
1370 	}
1371 }
1372 
1373 /*
1374  * Set in a new process credential, and drop the write lock.  The credential
1375  * must have a reference already.  Optionally, free a no-longer required
1376  * credential.  The scheduler also needs to inspect p_cred, so we also
1377  * briefly acquire the sched state mutex.
1378  */
1379 void
1380 proc_crmod_leave(kauth_cred_t scred, kauth_cred_t fcred, bool sugid)
1381 {
1382 	struct lwp *l = curlwp, *l2;
1383 	struct proc *p = l->l_proc;
1384 	kauth_cred_t oc;
1385 
1386 	KASSERT(mutex_owned(p->p_lock));
1387 
1388 	/* Is there a new credential to set in? */
1389 	if (scred != NULL) {
1390 		p->p_cred = scred;
1391 		LIST_FOREACH(l2, &p->p_lwps, l_sibling) {
1392 			if (l2 != l)
1393 				l2->l_prflag |= LPR_CRMOD;
1394 		}
1395 
1396 		/* Ensure the LWP cached credentials are up to date. */
1397 		if ((oc = l->l_cred) != scred) {
1398 			kauth_cred_hold(scred);
1399 			l->l_cred = scred;
1400 		}
1401 	} else
1402 		oc = NULL;	/* XXXgcc */
1403 
1404 	if (sugid) {
1405 		/*
1406 		 * Mark process as having changed credentials, stops
1407 		 * tracing etc.
1408 		 */
1409 		p->p_flag |= PK_SUGID;
1410 	}
1411 
1412 	mutex_exit(p->p_lock);
1413 
1414 	/* If there is a credential to be released, free it now. */
1415 	if (fcred != NULL) {
1416 		KASSERT(scred != NULL);
1417 		kauth_cred_free(fcred);
1418 		if (oc != scred)
1419 			kauth_cred_free(oc);
1420 	}
1421 }
1422 
1423 /*
1424  * proc_specific_key_create --
1425  *	Create a key for subsystem proc-specific data.
1426  */
1427 int
1428 proc_specific_key_create(specificdata_key_t *keyp, specificdata_dtor_t dtor)
1429 {
1430 
1431 	return (specificdata_key_create(proc_specificdata_domain, keyp, dtor));
1432 }
1433 
1434 /*
1435  * proc_specific_key_delete --
1436  *	Delete a key for subsystem proc-specific data.
1437  */
1438 void
1439 proc_specific_key_delete(specificdata_key_t key)
1440 {
1441 
1442 	specificdata_key_delete(proc_specificdata_domain, key);
1443 }
1444 
1445 /*
1446  * proc_initspecific --
1447  *	Initialize a proc's specificdata container.
1448  */
1449 void
1450 proc_initspecific(struct proc *p)
1451 {
1452 	int error;
1453 
1454 	error = specificdata_init(proc_specificdata_domain, &p->p_specdataref);
1455 	KASSERT(error == 0);
1456 }
1457 
1458 /*
1459  * proc_finispecific --
1460  *	Finalize a proc's specificdata container.
1461  */
1462 void
1463 proc_finispecific(struct proc *p)
1464 {
1465 
1466 	specificdata_fini(proc_specificdata_domain, &p->p_specdataref);
1467 }
1468 
1469 /*
1470  * proc_getspecific --
1471  *	Return proc-specific data corresponding to the specified key.
1472  */
1473 void *
1474 proc_getspecific(struct proc *p, specificdata_key_t key)
1475 {
1476 
1477 	return (specificdata_getspecific(proc_specificdata_domain,
1478 					 &p->p_specdataref, key));
1479 }
1480 
1481 /*
1482  * proc_setspecific --
1483  *	Set proc-specific data corresponding to the specified key.
1484  */
1485 void
1486 proc_setspecific(struct proc *p, specificdata_key_t key, void *data)
1487 {
1488 
1489 	specificdata_setspecific(proc_specificdata_domain,
1490 				 &p->p_specdataref, key, data);
1491 }
1492 
1493 int
1494 proc_uidmatch(kauth_cred_t cred, kauth_cred_t target)
1495 {
1496 	int r = 0;
1497 
1498 	if (kauth_cred_getuid(cred) != kauth_cred_getuid(target) ||
1499 	    kauth_cred_getuid(cred) != kauth_cred_getsvuid(target)) {
1500 		/*
1501 		 * suid proc of ours or proc not ours
1502 		 */
1503 		r = EPERM;
1504 	} else if (kauth_cred_getgid(target) != kauth_cred_getsvgid(target)) {
1505 		/*
1506 		 * sgid proc has sgid back to us temporarily
1507 		 */
1508 		r = EPERM;
1509 	} else {
1510 		/*
1511 		 * our rgid must be in target's group list (ie,
1512 		 * sub-processes started by a sgid process)
1513 		 */
1514 		int ismember = 0;
1515 
1516 		if (kauth_cred_ismember_gid(cred,
1517 		    kauth_cred_getgid(target), &ismember) != 0 ||
1518 		    !ismember)
1519 			r = EPERM;
1520 	}
1521 
1522 	return (r);
1523 }
1524 
1525 /*
1526  * sysctl stuff
1527  */
1528 
1529 #define KERN_PROCSLOP	(5 * sizeof(struct kinfo_proc))
1530 
1531 static const u_int sysctl_flagmap[] = {
1532 	PK_ADVLOCK, P_ADVLOCK,
1533 	PK_EXEC, P_EXEC,
1534 	PK_NOCLDWAIT, P_NOCLDWAIT,
1535 	PK_32, P_32,
1536 	PK_CLDSIGIGN, P_CLDSIGIGN,
1537 	PK_SUGID, P_SUGID,
1538 	0
1539 };
1540 
1541 static const u_int sysctl_sflagmap[] = {
1542 	PS_NOCLDSTOP, P_NOCLDSTOP,
1543 	PS_WEXIT, P_WEXIT,
1544 	PS_STOPFORK, P_STOPFORK,
1545 	PS_STOPEXEC, P_STOPEXEC,
1546 	PS_STOPEXIT, P_STOPEXIT,
1547 	0
1548 };
1549 
1550 static const u_int sysctl_slflagmap[] = {
1551 	PSL_TRACED, P_TRACED,
1552 	PSL_FSTRACE, P_FSTRACE,
1553 	PSL_CHTRACED, P_CHTRACED,
1554 	PSL_SYSCALL, P_SYSCALL,
1555 	0
1556 };
1557 
1558 static const u_int sysctl_lflagmap[] = {
1559 	PL_CONTROLT, P_CONTROLT,
1560 	PL_PPWAIT, P_PPWAIT,
1561 	0
1562 };
1563 
1564 static const u_int sysctl_stflagmap[] = {
1565 	PST_PROFIL, P_PROFIL,
1566 	0
1567 
1568 };
1569 
1570 /* used by kern_lwp also */
1571 const u_int sysctl_lwpflagmap[] = {
1572 	LW_SINTR, L_SINTR,
1573 	LW_SYSTEM, L_SYSTEM,
1574 	0
1575 };
1576 
1577 /*
1578  * Find the most ``active'' lwp of a process and return it for ps display
1579  * purposes
1580  */
1581 static struct lwp *
1582 proc_active_lwp(struct proc *p)
1583 {
1584 	static const int ostat[] = {
1585 		0,
1586 		2,	/* LSIDL */
1587 		6,	/* LSRUN */
1588 		5,	/* LSSLEEP */
1589 		4,	/* LSSTOP */
1590 		0,	/* LSZOMB */
1591 		1,	/* LSDEAD */
1592 		7,	/* LSONPROC */
1593 		3	/* LSSUSPENDED */
1594 	};
1595 
1596 	struct lwp *l, *lp = NULL;
1597 	LIST_FOREACH(l, &p->p_lwps, l_sibling) {
1598 		KASSERT(l->l_stat >= 0 && l->l_stat < __arraycount(ostat));
1599 		if (lp == NULL ||
1600 		    ostat[l->l_stat] > ostat[lp->l_stat] ||
1601 		    (ostat[l->l_stat] == ostat[lp->l_stat] &&
1602 		    l->l_cpticks > lp->l_cpticks)) {
1603 			lp = l;
1604 			continue;
1605 		}
1606 	}
1607 	return lp;
1608 }
1609 
1610 static int
1611 sysctl_doeproc(SYSCTLFN_ARGS)
1612 {
1613 	union {
1614 		struct kinfo_proc kproc;
1615 		struct kinfo_proc2 kproc2;
1616 	} *kbuf;
1617 	struct proc *p, *next, *marker;
1618 	char *where, *dp;
1619 	int type, op, arg, error;
1620 	u_int elem_size, kelem_size, elem_count;
1621 	size_t buflen, needed;
1622 	bool match, zombie, mmmbrains;
1623 
1624 	if (namelen == 1 && name[0] == CTL_QUERY)
1625 		return (sysctl_query(SYSCTLFN_CALL(rnode)));
1626 
1627 	dp = where = oldp;
1628 	buflen = where != NULL ? *oldlenp : 0;
1629 	error = 0;
1630 	needed = 0;
1631 	type = rnode->sysctl_num;
1632 
1633 	if (type == KERN_PROC) {
1634 		if (namelen != 2 && !(namelen == 1 && name[0] == KERN_PROC_ALL))
1635 			return (EINVAL);
1636 		op = name[0];
1637 		if (op != KERN_PROC_ALL)
1638 			arg = name[1];
1639 		else
1640 			arg = 0;		/* Quell compiler warning */
1641 		elem_count = 0;	/* Ditto */
1642 		kelem_size = elem_size = sizeof(kbuf->kproc);
1643 	} else {
1644 		if (namelen != 4)
1645 			return (EINVAL);
1646 		op = name[0];
1647 		arg = name[1];
1648 		elem_size = name[2];
1649 		elem_count = name[3];
1650 		kelem_size = sizeof(kbuf->kproc2);
1651 	}
1652 
1653 	sysctl_unlock();
1654 
1655 	kbuf = kmem_alloc(sizeof(*kbuf), KM_SLEEP);
1656 	marker = kmem_alloc(sizeof(*marker), KM_SLEEP);
1657 	marker->p_flag = PK_MARKER;
1658 
1659 	mutex_enter(proc_lock);
1660 	mmmbrains = false;
1661 	for (p = LIST_FIRST(&allproc);; p = next) {
1662 		if (p == NULL) {
1663 			if (!mmmbrains) {
1664 				p = LIST_FIRST(&zombproc);
1665 				mmmbrains = true;
1666 			}
1667 			if (p == NULL)
1668 				break;
1669 		}
1670 		next = LIST_NEXT(p, p_list);
1671 		if ((p->p_flag & PK_MARKER) != 0)
1672 			continue;
1673 
1674 		/*
1675 		 * Skip embryonic processes.
1676 		 */
1677 		if (p->p_stat == SIDL)
1678 			continue;
1679 
1680 		mutex_enter(p->p_lock);
1681 		error = kauth_authorize_process(l->l_cred,
1682 		    KAUTH_PROCESS_CANSEE, p,
1683 		    KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL);
1684 		if (error != 0) {
1685 			mutex_exit(p->p_lock);
1686 			continue;
1687 		}
1688 
1689 		/*
1690 		 * TODO - make more efficient (see notes below).
1691 		 * do by session.
1692 		 */
1693 		switch (op) {
1694 		case KERN_PROC_PID:
1695 			/* could do this with just a lookup */
1696 			match = (p->p_pid == (pid_t)arg);
1697 			break;
1698 
1699 		case KERN_PROC_PGRP:
1700 			/* could do this by traversing pgrp */
1701 			match = (p->p_pgrp->pg_id == (pid_t)arg);
1702 			break;
1703 
1704 		case KERN_PROC_SESSION:
1705 			match = (p->p_session->s_sid == (pid_t)arg);
1706 			break;
1707 
1708 		case KERN_PROC_TTY:
1709 			match = true;
1710 			if (arg == (int) KERN_PROC_TTY_REVOKE) {
1711 				if ((p->p_lflag & PL_CONTROLT) == 0 ||
1712 				    p->p_session->s_ttyp == NULL ||
1713 				    p->p_session->s_ttyvp != NULL) {
1714 				    	match = false;
1715 				}
1716 			} else if ((p->p_lflag & PL_CONTROLT) == 0 ||
1717 			    p->p_session->s_ttyp == NULL) {
1718 				if ((dev_t)arg != KERN_PROC_TTY_NODEV) {
1719 					match = false;
1720 				}
1721 			} else if (p->p_session->s_ttyp->t_dev != (dev_t)arg) {
1722 				match = false;
1723 			}
1724 			break;
1725 
1726 		case KERN_PROC_UID:
1727 			match = (kauth_cred_geteuid(p->p_cred) == (uid_t)arg);
1728 			break;
1729 
1730 		case KERN_PROC_RUID:
1731 			match = (kauth_cred_getuid(p->p_cred) == (uid_t)arg);
1732 			break;
1733 
1734 		case KERN_PROC_GID:
1735 			match = (kauth_cred_getegid(p->p_cred) == (uid_t)arg);
1736 			break;
1737 
1738 		case KERN_PROC_RGID:
1739 			match = (kauth_cred_getgid(p->p_cred) == (uid_t)arg);
1740 			break;
1741 
1742 		case KERN_PROC_ALL:
1743 			match = true;
1744 			/* allow everything */
1745 			break;
1746 
1747 		default:
1748 			error = EINVAL;
1749 			mutex_exit(p->p_lock);
1750 			goto cleanup;
1751 		}
1752 		if (!match) {
1753 			mutex_exit(p->p_lock);
1754 			continue;
1755 		}
1756 
1757 		/*
1758 		 * Grab a hold on the process.
1759 		 */
1760 		if (mmmbrains) {
1761 			zombie = true;
1762 		} else {
1763 			zombie = !rw_tryenter(&p->p_reflock, RW_READER);
1764 		}
1765 		if (zombie) {
1766 			LIST_INSERT_AFTER(p, marker, p_list);
1767 		}
1768 
1769 		if (buflen >= elem_size &&
1770 		    (type == KERN_PROC || elem_count > 0)) {
1771 			if (type == KERN_PROC) {
1772 				kbuf->kproc.kp_proc = *p;
1773 				fill_eproc(p, &kbuf->kproc.kp_eproc, zombie);
1774 			} else {
1775 				fill_kproc2(p, &kbuf->kproc2, zombie);
1776 				elem_count--;
1777 			}
1778 			mutex_exit(p->p_lock);
1779 			mutex_exit(proc_lock);
1780 			/*
1781 			 * Copy out elem_size, but not larger than kelem_size
1782 			 */
1783 			error = sysctl_copyout(l, kbuf, dp,
1784 			    min(kelem_size, elem_size));
1785 			mutex_enter(proc_lock);
1786 			if (error) {
1787 				goto bah;
1788 			}
1789 			dp += elem_size;
1790 			buflen -= elem_size;
1791 		} else {
1792 			mutex_exit(p->p_lock);
1793 		}
1794 		needed += elem_size;
1795 
1796 		/*
1797 		 * Release reference to process.
1798 		 */
1799 	 	if (zombie) {
1800 			next = LIST_NEXT(marker, p_list);
1801  			LIST_REMOVE(marker, p_list);
1802 		} else {
1803 			rw_exit(&p->p_reflock);
1804 			next = LIST_NEXT(p, p_list);
1805 		}
1806 	}
1807 	mutex_exit(proc_lock);
1808 
1809 	if (where != NULL) {
1810 		*oldlenp = dp - where;
1811 		if (needed > *oldlenp) {
1812 			error = ENOMEM;
1813 			goto out;
1814 		}
1815 	} else {
1816 		needed += KERN_PROCSLOP;
1817 		*oldlenp = needed;
1818 	}
1819 	if (kbuf)
1820 		kmem_free(kbuf, sizeof(*kbuf));
1821 	if (marker)
1822 		kmem_free(marker, sizeof(*marker));
1823 	sysctl_relock();
1824 	return 0;
1825  bah:
1826  	if (zombie)
1827  		LIST_REMOVE(marker, p_list);
1828 	else
1829 		rw_exit(&p->p_reflock);
1830  cleanup:
1831 	mutex_exit(proc_lock);
1832  out:
1833 	if (kbuf)
1834 		kmem_free(kbuf, sizeof(*kbuf));
1835 	if (marker)
1836 		kmem_free(marker, sizeof(*marker));
1837 	sysctl_relock();
1838 	return error;
1839 }
1840 
1841 int
1842 copyin_psstrings(struct proc *p, struct ps_strings *arginfo)
1843 {
1844 
1845 #ifdef COMPAT_NETBSD32
1846 	if (p->p_flag & PK_32) {
1847 		struct ps_strings32 arginfo32;
1848 
1849 		int error = copyin_proc(p, (void *)p->p_psstrp, &arginfo32,
1850 		    sizeof(arginfo32));
1851 		if (error)
1852 			return error;
1853 		arginfo->ps_argvstr = (void *)(uintptr_t)arginfo32.ps_argvstr;
1854 		arginfo->ps_nargvstr = arginfo32.ps_nargvstr;
1855 		arginfo->ps_envstr = (void *)(uintptr_t)arginfo32.ps_envstr;
1856 		arginfo->ps_nenvstr = arginfo32.ps_nenvstr;
1857 		return 0;
1858 	}
1859 #endif
1860 	return copyin_proc(p, (void *)p->p_psstrp, arginfo, sizeof(*arginfo));
1861 }
1862 
1863 static int
1864 copy_procargs_sysctl_cb(void *cookie_, const void *src, size_t off, size_t len)
1865 {
1866 	void **cookie = cookie_;
1867 	struct lwp *l = cookie[0];
1868 	char *dst = cookie[1];
1869 
1870 	return sysctl_copyout(l, src, dst + off, len);
1871 }
1872 
1873 /*
1874  * sysctl helper routine for kern.proc_args pseudo-subtree.
1875  */
1876 static int
1877 sysctl_kern_proc_args(SYSCTLFN_ARGS)
1878 {
1879 	struct ps_strings pss;
1880 	struct proc *p;
1881 	pid_t pid;
1882 	int type, error;
1883 	void *cookie[2];
1884 
1885 	if (namelen == 1 && name[0] == CTL_QUERY)
1886 		return (sysctl_query(SYSCTLFN_CALL(rnode)));
1887 
1888 	if (newp != NULL || namelen != 2)
1889 		return (EINVAL);
1890 	pid = name[0];
1891 	type = name[1];
1892 
1893 	switch (type) {
1894 	case KERN_PROC_ARGV:
1895 	case KERN_PROC_NARGV:
1896 	case KERN_PROC_ENV:
1897 	case KERN_PROC_NENV:
1898 		/* ok */
1899 		break;
1900 	default:
1901 		return (EINVAL);
1902 	}
1903 
1904 	sysctl_unlock();
1905 
1906 	/* check pid */
1907 	mutex_enter(proc_lock);
1908 	if ((p = proc_find(pid)) == NULL) {
1909 		error = EINVAL;
1910 		goto out_locked;
1911 	}
1912 	mutex_enter(p->p_lock);
1913 
1914 	/* Check permission. */
1915 	if (type == KERN_PROC_ARGV || type == KERN_PROC_NARGV)
1916 		error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE,
1917 		    p, KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ARGS), NULL, NULL);
1918 	else if (type == KERN_PROC_ENV || type == KERN_PROC_NENV)
1919 		error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE,
1920 		    p, KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENV), NULL, NULL);
1921 	else
1922 		error = EINVAL; /* XXXGCC */
1923 	if (error) {
1924 		mutex_exit(p->p_lock);
1925 		goto out_locked;
1926 	}
1927 
1928 	if (oldp == NULL) {
1929 		if (type == KERN_PROC_NARGV || type == KERN_PROC_NENV)
1930 			*oldlenp = sizeof (int);
1931 		else
1932 			*oldlenp = ARG_MAX;	/* XXX XXX XXX */
1933 		error = 0;
1934 		mutex_exit(p->p_lock);
1935 		goto out_locked;
1936 	}
1937 
1938 	/*
1939 	 * Zombies don't have a stack, so we can't read their psstrings.
1940 	 * System processes also don't have a user stack.
1941 	 */
1942 	if (P_ZOMBIE(p) || (p->p_flag & PK_SYSTEM) != 0) {
1943 		error = EINVAL;
1944 		mutex_exit(p->p_lock);
1945 		goto out_locked;
1946 	}
1947 
1948 	error = rw_tryenter(&p->p_reflock, RW_READER) ? 0 : EBUSY;
1949 	mutex_exit(p->p_lock);
1950 	if (error) {
1951 		goto out_locked;
1952 	}
1953 	mutex_exit(proc_lock);
1954 
1955 	if (type == KERN_PROC_NARGV || type == KERN_PROC_NENV) {
1956 		int value;
1957 		if ((error = copyin_psstrings(p, &pss)) == 0) {
1958 			if (type == KERN_PROC_NARGV)
1959 				value = pss.ps_nargvstr;
1960 			else
1961 				value = pss.ps_nenvstr;
1962 			error = sysctl_copyout(l, &value, oldp, sizeof(value));
1963 			*oldlenp = sizeof(value);
1964 		}
1965 	} else {
1966 		cookie[0] = l;
1967 		cookie[1] = oldp;
1968 		error = copy_procargs(p, type, oldlenp,
1969 		    copy_procargs_sysctl_cb, cookie);
1970 	}
1971 	rw_exit(&p->p_reflock);
1972 	sysctl_relock();
1973 	return error;
1974 
1975 out_locked:
1976 	mutex_exit(proc_lock);
1977 	sysctl_relock();
1978 	return error;
1979 }
1980 
1981 int
1982 copy_procargs(struct proc *p, int oid, size_t *limit,
1983     int (*cb)(void *, const void *, size_t, size_t), void *cookie)
1984 {
1985 	struct ps_strings pss;
1986 	size_t len, i, loaded, entry_len;
1987 	struct uio auio;
1988 	struct iovec aiov;
1989 	int error, argvlen;
1990 	char *arg;
1991 	char **argv;
1992 	vaddr_t user_argv;
1993 	struct vmspace *vmspace;
1994 
1995 	/*
1996 	 * Allocate a temporary buffer to hold the argument vector and
1997 	 * the arguments themselve.
1998 	 */
1999 	arg = kmem_alloc(PAGE_SIZE, KM_SLEEP);
2000 	argv = kmem_alloc(PAGE_SIZE, KM_SLEEP);
2001 
2002 	/*
2003 	 * Lock the process down in memory.
2004 	 */
2005 	vmspace = p->p_vmspace;
2006 	uvmspace_addref(vmspace);
2007 
2008 	/*
2009 	 * Read in the ps_strings structure.
2010 	 */
2011 	if ((error = copyin_psstrings(p, &pss)) != 0)
2012 		goto done;
2013 
2014 	/*
2015 	 * Now read the address of the argument vector.
2016 	 */
2017 	switch (oid) {
2018 	case KERN_PROC_ARGV:
2019 		user_argv = (uintptr_t)pss.ps_argvstr;
2020 		argvlen = pss.ps_nargvstr;
2021 		break;
2022 	case KERN_PROC_ENV:
2023 		user_argv = (uintptr_t)pss.ps_envstr;
2024 		argvlen = pss.ps_nenvstr;
2025 		break;
2026 	default:
2027 		error = EINVAL;
2028 		goto done;
2029 	}
2030 
2031 	if (argvlen < 0) {
2032 		error = EIO;
2033 		goto done;
2034 	}
2035 
2036 #ifdef COMPAT_NETBSD32
2037 	if (p->p_flag & PK_32)
2038 		entry_len = sizeof(netbsd32_charp);
2039 	else
2040 #endif
2041 		entry_len = sizeof(char *);
2042 
2043 	/*
2044 	 * Now copy each string.
2045 	 */
2046 	len = 0; /* bytes written to user buffer */
2047 	loaded = 0; /* bytes from argv already processed */
2048 	i = 0; /* To make compiler happy */
2049 
2050 	for (; argvlen; --argvlen) {
2051 		int finished = 0;
2052 		vaddr_t base;
2053 		size_t xlen;
2054 		int j;
2055 
2056 		if (loaded == 0) {
2057 			size_t rem = entry_len * argvlen;
2058 			loaded = MIN(rem, PAGE_SIZE);
2059 			error = copyin_vmspace(vmspace,
2060 			    (const void *)user_argv, argv, loaded);
2061 			if (error)
2062 				break;
2063 			user_argv += loaded;
2064 			i = 0;
2065 		}
2066 
2067 #ifdef COMPAT_NETBSD32
2068 		if (p->p_flag & PK_32) {
2069 			netbsd32_charp *argv32;
2070 
2071 			argv32 = (netbsd32_charp *)argv;
2072 			base = (vaddr_t)NETBSD32PTR64(argv32[i++]);
2073 		} else
2074 #endif
2075 			base = (vaddr_t)argv[i++];
2076 		loaded -= entry_len;
2077 
2078 		/*
2079 		 * The program has messed around with its arguments,
2080 		 * possibly deleting some, and replacing them with
2081 		 * NULL's. Treat this as the last argument and not
2082 		 * a failure.
2083 		 */
2084 		if (base == 0)
2085 			break;
2086 
2087 		while (!finished) {
2088 			xlen = PAGE_SIZE - (base & PAGE_MASK);
2089 
2090 			aiov.iov_base = arg;
2091 			aiov.iov_len = PAGE_SIZE;
2092 			auio.uio_iov = &aiov;
2093 			auio.uio_iovcnt = 1;
2094 			auio.uio_offset = base;
2095 			auio.uio_resid = xlen;
2096 			auio.uio_rw = UIO_READ;
2097 			UIO_SETUP_SYSSPACE(&auio);
2098 			error = uvm_io(&vmspace->vm_map, &auio);
2099 			if (error)
2100 				goto done;
2101 
2102 			/* Look for the end of the string */
2103 			for (j = 0; j < xlen; j++) {
2104 				if (arg[j] == '\0') {
2105 					xlen = j + 1;
2106 					finished = 1;
2107 					break;
2108 				}
2109 			}
2110 
2111 			/* Check for user buffer overflow */
2112 			if (len + xlen > *limit) {
2113 				finished = 1;
2114 				if (len > *limit)
2115 					xlen = 0;
2116 				else
2117 					xlen = *limit - len;
2118 			}
2119 
2120 			/* Copyout the page */
2121 			error = (*cb)(cookie, arg, len, xlen);
2122 			if (error)
2123 				goto done;
2124 
2125 			len += xlen;
2126 			base += xlen;
2127 		}
2128 	}
2129 	*limit = len;
2130 
2131 done:
2132 	kmem_free(argv, PAGE_SIZE);
2133 	kmem_free(arg, PAGE_SIZE);
2134 	uvmspace_free(vmspace);
2135 	return error;
2136 }
2137 
2138 /*
2139  * Fill in an eproc structure for the specified process.
2140  */
2141 void
2142 fill_eproc(struct proc *p, struct eproc *ep, bool zombie)
2143 {
2144 	struct tty *tp;
2145 	struct lwp *l;
2146 
2147 	KASSERT(mutex_owned(proc_lock));
2148 	KASSERT(mutex_owned(p->p_lock));
2149 
2150 	memset(ep, 0, sizeof(*ep));
2151 
2152 	ep->e_paddr = p;
2153 	ep->e_sess = p->p_session;
2154 	if (p->p_cred) {
2155 		kauth_cred_topcred(p->p_cred, &ep->e_pcred);
2156 		kauth_cred_toucred(p->p_cred, &ep->e_ucred);
2157 	}
2158 	if (p->p_stat != SIDL && !P_ZOMBIE(p) && !zombie) {
2159 		struct vmspace *vm = p->p_vmspace;
2160 
2161 		ep->e_vm.vm_rssize = vm_resident_count(vm);
2162 		ep->e_vm.vm_tsize = vm->vm_tsize;
2163 		ep->e_vm.vm_dsize = vm->vm_dsize;
2164 		ep->e_vm.vm_ssize = vm->vm_ssize;
2165 		ep->e_vm.vm_map.size = vm->vm_map.size;
2166 
2167 		/* Pick the primary (first) LWP */
2168 		l = proc_active_lwp(p);
2169 		KASSERT(l != NULL);
2170 		lwp_lock(l);
2171 		if (l->l_wchan)
2172 			strncpy(ep->e_wmesg, l->l_wmesg, WMESGLEN);
2173 		lwp_unlock(l);
2174 	}
2175 	if (p->p_pptr)
2176 		ep->e_ppid = p->p_pptr->p_pid;
2177 	if (p->p_pgrp && p->p_session) {
2178 		ep->e_pgid = p->p_pgrp->pg_id;
2179 		ep->e_jobc = p->p_pgrp->pg_jobc;
2180 		ep->e_sid = p->p_session->s_sid;
2181 		if ((p->p_lflag & PL_CONTROLT) &&
2182 		    (tp = ep->e_sess->s_ttyp)) {
2183 			ep->e_tdev = tp->t_dev;
2184 			ep->e_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PGID;
2185 			ep->e_tsess = tp->t_session;
2186 		} else
2187 			ep->e_tdev = (uint32_t)NODEV;
2188 		ep->e_flag = ep->e_sess->s_ttyvp ? EPROC_CTTY : 0;
2189 		if (SESS_LEADER(p))
2190 			ep->e_flag |= EPROC_SLEADER;
2191 		strncpy(ep->e_login, ep->e_sess->s_login, MAXLOGNAME);
2192 	}
2193 	ep->e_xsize = ep->e_xrssize = 0;
2194 	ep->e_xccount = ep->e_xswrss = 0;
2195 }
2196 
2197 /*
2198  * Fill in a kinfo_proc2 structure for the specified process.
2199  */
2200 static void
2201 fill_kproc2(struct proc *p, struct kinfo_proc2 *ki, bool zombie)
2202 {
2203 	struct tty *tp;
2204 	struct lwp *l, *l2;
2205 	struct timeval ut, st, rt;
2206 	sigset_t ss1, ss2;
2207 	struct rusage ru;
2208 	struct vmspace *vm;
2209 
2210 	KASSERT(mutex_owned(proc_lock));
2211 	KASSERT(mutex_owned(p->p_lock));
2212 
2213 	sigemptyset(&ss1);
2214 	sigemptyset(&ss2);
2215 	memset(ki, 0, sizeof(*ki));
2216 
2217 	ki->p_paddr = PTRTOUINT64(p);
2218 	ki->p_fd = PTRTOUINT64(p->p_fd);
2219 	ki->p_cwdi = PTRTOUINT64(p->p_cwdi);
2220 	ki->p_stats = PTRTOUINT64(p->p_stats);
2221 	ki->p_limit = PTRTOUINT64(p->p_limit);
2222 	ki->p_vmspace = PTRTOUINT64(p->p_vmspace);
2223 	ki->p_sigacts = PTRTOUINT64(p->p_sigacts);
2224 	ki->p_sess = PTRTOUINT64(p->p_session);
2225 	ki->p_tsess = 0;	/* may be changed if controlling tty below */
2226 	ki->p_ru = PTRTOUINT64(&p->p_stats->p_ru);
2227 	ki->p_eflag = 0;
2228 	ki->p_exitsig = p->p_exitsig;
2229 	ki->p_flag = L_INMEM;   /* Process never swapped out */
2230 	ki->p_flag |= sysctl_map_flags(sysctl_flagmap, p->p_flag);
2231 	ki->p_flag |= sysctl_map_flags(sysctl_sflagmap, p->p_sflag);
2232 	ki->p_flag |= sysctl_map_flags(sysctl_slflagmap, p->p_slflag);
2233 	ki->p_flag |= sysctl_map_flags(sysctl_lflagmap, p->p_lflag);
2234 	ki->p_flag |= sysctl_map_flags(sysctl_stflagmap, p->p_stflag);
2235 	ki->p_pid = p->p_pid;
2236 	if (p->p_pptr)
2237 		ki->p_ppid = p->p_pptr->p_pid;
2238 	else
2239 		ki->p_ppid = 0;
2240 	ki->p_uid = kauth_cred_geteuid(p->p_cred);
2241 	ki->p_ruid = kauth_cred_getuid(p->p_cred);
2242 	ki->p_gid = kauth_cred_getegid(p->p_cred);
2243 	ki->p_rgid = kauth_cred_getgid(p->p_cred);
2244 	ki->p_svuid = kauth_cred_getsvuid(p->p_cred);
2245 	ki->p_svgid = kauth_cred_getsvgid(p->p_cred);
2246 	ki->p_ngroups = kauth_cred_ngroups(p->p_cred);
2247 	kauth_cred_getgroups(p->p_cred, ki->p_groups,
2248 	    min(ki->p_ngroups, sizeof(ki->p_groups) / sizeof(ki->p_groups[0])),
2249 	    UIO_SYSSPACE);
2250 
2251 	ki->p_uticks = p->p_uticks;
2252 	ki->p_sticks = p->p_sticks;
2253 	ki->p_iticks = p->p_iticks;
2254 	ki->p_tpgid = NO_PGID;	/* may be changed if controlling tty below */
2255 	ki->p_tracep = PTRTOUINT64(p->p_tracep);
2256 	ki->p_traceflag = p->p_traceflag;
2257 
2258 	memcpy(&ki->p_sigignore, &p->p_sigctx.ps_sigignore,sizeof(ki_sigset_t));
2259 	memcpy(&ki->p_sigcatch, &p->p_sigctx.ps_sigcatch, sizeof(ki_sigset_t));
2260 
2261 	ki->p_cpticks = 0;
2262 	ki->p_pctcpu = p->p_pctcpu;
2263 	ki->p_estcpu = 0;
2264 	ki->p_stat = p->p_stat; /* Will likely be overridden by LWP status */
2265 	ki->p_realstat = p->p_stat;
2266 	ki->p_nice = p->p_nice;
2267 	ki->p_xstat = p->p_xstat;
2268 	ki->p_acflag = p->p_acflag;
2269 
2270 	strncpy(ki->p_comm, p->p_comm,
2271 	    min(sizeof(ki->p_comm), sizeof(p->p_comm)));
2272 	strncpy(ki->p_ename, p->p_emul->e_name, sizeof(ki->p_ename));
2273 
2274 	ki->p_nlwps = p->p_nlwps;
2275 	ki->p_realflag = ki->p_flag;
2276 
2277 	if (p->p_stat != SIDL && !P_ZOMBIE(p) && !zombie) {
2278 		vm = p->p_vmspace;
2279 		ki->p_vm_rssize = vm_resident_count(vm);
2280 		ki->p_vm_tsize = vm->vm_tsize;
2281 		ki->p_vm_dsize = vm->vm_dsize;
2282 		ki->p_vm_ssize = vm->vm_ssize;
2283 		ki->p_vm_vsize = atop(vm->vm_map.size);
2284 		/*
2285 		 * Since the stack is initially mapped mostly with
2286 		 * PROT_NONE and grown as needed, adjust the "mapped size"
2287 		 * to skip the unused stack portion.
2288 		 */
2289 		ki->p_vm_msize =
2290 		    atop(vm->vm_map.size) - vm->vm_issize + vm->vm_ssize;
2291 
2292 		/* Pick the primary (first) LWP */
2293 		l = proc_active_lwp(p);
2294 		KASSERT(l != NULL);
2295 		lwp_lock(l);
2296 		ki->p_nrlwps = p->p_nrlwps;
2297 		ki->p_forw = 0;
2298 		ki->p_back = 0;
2299 		ki->p_addr = PTRTOUINT64(l->l_addr);
2300 		ki->p_stat = l->l_stat;
2301 		ki->p_flag |= sysctl_map_flags(sysctl_lwpflagmap, l->l_flag);
2302 		ki->p_swtime = l->l_swtime;
2303 		ki->p_slptime = l->l_slptime;
2304 		if (l->l_stat == LSONPROC)
2305 			ki->p_schedflags = l->l_cpu->ci_schedstate.spc_flags;
2306 		else
2307 			ki->p_schedflags = 0;
2308 		ki->p_priority = lwp_eprio(l);
2309 		ki->p_usrpri = l->l_priority;
2310 		if (l->l_wchan)
2311 			strncpy(ki->p_wmesg, l->l_wmesg, sizeof(ki->p_wmesg));
2312 		ki->p_wchan = PTRTOUINT64(l->l_wchan);
2313 		ki->p_cpuid = cpu_index(l->l_cpu);
2314 		lwp_unlock(l);
2315 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
2316 			/* This is hardly correct, but... */
2317 			sigplusset(&l->l_sigpend.sp_set, &ss1);
2318 			sigplusset(&l->l_sigmask, &ss2);
2319 			ki->p_cpticks += l->l_cpticks;
2320 			ki->p_pctcpu += l->l_pctcpu;
2321 			ki->p_estcpu += l->l_estcpu;
2322 		}
2323 	}
2324 	sigplusset(&p->p_sigpend.sp_set, &ss2);
2325 	memcpy(&ki->p_siglist, &ss1, sizeof(ki_sigset_t));
2326 	memcpy(&ki->p_sigmask, &ss2, sizeof(ki_sigset_t));
2327 
2328 	if (p->p_session != NULL) {
2329 		ki->p_sid = p->p_session->s_sid;
2330 		ki->p__pgid = p->p_pgrp->pg_id;
2331 		if (p->p_session->s_ttyvp)
2332 			ki->p_eflag |= EPROC_CTTY;
2333 		if (SESS_LEADER(p))
2334 			ki->p_eflag |= EPROC_SLEADER;
2335 		strncpy(ki->p_login, p->p_session->s_login,
2336 		    min(sizeof ki->p_login - 1, sizeof p->p_session->s_login));
2337 		ki->p_jobc = p->p_pgrp->pg_jobc;
2338 		if ((p->p_lflag & PL_CONTROLT) && (tp = p->p_session->s_ttyp)) {
2339 			ki->p_tdev = tp->t_dev;
2340 			ki->p_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PGID;
2341 			ki->p_tsess = PTRTOUINT64(tp->t_session);
2342 		} else {
2343 			ki->p_tdev = (int32_t)NODEV;
2344 		}
2345 	}
2346 
2347 	if (!P_ZOMBIE(p) && !zombie) {
2348 		ki->p_uvalid = 1;
2349 		ki->p_ustart_sec = p->p_stats->p_start.tv_sec;
2350 		ki->p_ustart_usec = p->p_stats->p_start.tv_usec;
2351 
2352 		calcru(p, &ut, &st, NULL, &rt);
2353 		ki->p_rtime_sec = rt.tv_sec;
2354 		ki->p_rtime_usec = rt.tv_usec;
2355 		ki->p_uutime_sec = ut.tv_sec;
2356 		ki->p_uutime_usec = ut.tv_usec;
2357 		ki->p_ustime_sec = st.tv_sec;
2358 		ki->p_ustime_usec = st.tv_usec;
2359 
2360 		memcpy(&ru, &p->p_stats->p_ru, sizeof(ru));
2361 		ki->p_uru_nvcsw = 0;
2362 		ki->p_uru_nivcsw = 0;
2363 		LIST_FOREACH(l2, &p->p_lwps, l_sibling) {
2364 			ki->p_uru_nvcsw += (l2->l_ncsw - l2->l_nivcsw);
2365 			ki->p_uru_nivcsw += l2->l_nivcsw;
2366 			ruadd(&ru, &l2->l_ru);
2367 		}
2368 		ki->p_uru_maxrss = ru.ru_maxrss;
2369 		ki->p_uru_ixrss = ru.ru_ixrss;
2370 		ki->p_uru_idrss = ru.ru_idrss;
2371 		ki->p_uru_isrss = ru.ru_isrss;
2372 		ki->p_uru_minflt = ru.ru_minflt;
2373 		ki->p_uru_majflt = ru.ru_majflt;
2374 		ki->p_uru_nswap = ru.ru_nswap;
2375 		ki->p_uru_inblock = ru.ru_inblock;
2376 		ki->p_uru_oublock = ru.ru_oublock;
2377 		ki->p_uru_msgsnd = ru.ru_msgsnd;
2378 		ki->p_uru_msgrcv = ru.ru_msgrcv;
2379 		ki->p_uru_nsignals = ru.ru_nsignals;
2380 
2381 		timeradd(&p->p_stats->p_cru.ru_utime,
2382 			 &p->p_stats->p_cru.ru_stime, &ut);
2383 		ki->p_uctime_sec = ut.tv_sec;
2384 		ki->p_uctime_usec = ut.tv_usec;
2385 	}
2386 }
2387