xref: /netbsd-src/sys/kern/kern_proc.c (revision ce2c90c7c172d95d2402a5b3d96d8f8e6d138a21)
1 /*	$NetBSD: kern_proc.c,v 1.98 2006/10/11 04:51:06 thorpej Exp $	*/
2 
3 /*-
4  * Copyright (c) 1999 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.
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  * 3. All advertising materials mentioning features or use of this software
20  *    must display the following acknowledgement:
21  *	This product includes software developed by the NetBSD
22  *	Foundation, Inc. and its contributors.
23  * 4. Neither the name of The NetBSD Foundation nor the names of its
24  *    contributors may be used to endorse or promote products derived
25  *    from this software without specific prior written permission.
26  *
27  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
28  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
29  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
30  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
31  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
32  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
33  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
34  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
35  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
36  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
37  * POSSIBILITY OF SUCH DAMAGE.
38  */
39 
40 /*
41  * Copyright (c) 1982, 1986, 1989, 1991, 1993
42  *	The Regents of the University of California.  All rights reserved.
43  *
44  * Redistribution and use in source and binary forms, with or without
45  * modification, are permitted provided that the following conditions
46  * are met:
47  * 1. Redistributions of source code must retain the above copyright
48  *    notice, this list of conditions and the following disclaimer.
49  * 2. Redistributions in binary form must reproduce the above copyright
50  *    notice, this list of conditions and the following disclaimer in the
51  *    documentation and/or other materials provided with the distribution.
52  * 3. Neither the name of the University nor the names of its contributors
53  *    may be used to endorse or promote products derived from this software
54  *    without specific prior written permission.
55  *
56  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
57  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
58  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
59  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
60  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
61  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
62  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
63  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
64  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
65  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
66  * SUCH DAMAGE.
67  *
68  *	@(#)kern_proc.c	8.7 (Berkeley) 2/14/95
69  */
70 
71 #include <sys/cdefs.h>
72 __KERNEL_RCSID(0, "$NetBSD: kern_proc.c,v 1.98 2006/10/11 04:51:06 thorpej Exp $");
73 
74 #include "opt_kstack.h"
75 #include "opt_maxuprc.h"
76 #include "opt_multiprocessor.h"
77 #include "opt_lockdebug.h"
78 
79 #include <sys/param.h>
80 #include <sys/systm.h>
81 #include <sys/kernel.h>
82 #include <sys/proc.h>
83 #include <sys/resourcevar.h>
84 #include <sys/buf.h>
85 #include <sys/acct.h>
86 #include <sys/wait.h>
87 #include <sys/file.h>
88 #include <ufs/ufs/quota.h>
89 #include <sys/uio.h>
90 #include <sys/malloc.h>
91 #include <sys/pool.h>
92 #include <sys/mbuf.h>
93 #include <sys/ioctl.h>
94 #include <sys/tty.h>
95 #include <sys/signalvar.h>
96 #include <sys/ras.h>
97 #include <sys/sa.h>
98 #include <sys/savar.h>
99 #include <sys/filedesc.h>
100 #include <sys/kauth.h>
101 
102 #include <uvm/uvm.h>
103 #include <uvm/uvm_extern.h>
104 
105 /*
106  * Other process lists
107  */
108 
109 struct proclist allproc;
110 struct proclist zombproc;	/* resources have been freed */
111 
112 
113 /*
114  * Process list locking:
115  *
116  * We have two types of locks on the proclists: read locks and write
117  * locks.  Read locks can be used in interrupt context, so while we
118  * hold the write lock, we must also block clock interrupts to
119  * lock out any scheduling changes that may happen in interrupt
120  * context.
121  *
122  * The proclist lock locks the following structures:
123  *
124  *	allproc
125  *	zombproc
126  *	pid_table
127  */
128 struct lock proclist_lock;
129 
130 /*
131  * pid to proc lookup is done by indexing the pid_table array.
132  * Since pid numbers are only allocated when an empty slot
133  * has been found, there is no need to search any lists ever.
134  * (an orphaned pgrp will lock the slot, a session will lock
135  * the pgrp with the same number.)
136  * If the table is too small it is reallocated with twice the
137  * previous size and the entries 'unzipped' into the two halves.
138  * A linked list of free entries is passed through the pt_proc
139  * field of 'free' items - set odd to be an invalid ptr.
140  */
141 
142 struct pid_table {
143 	struct proc	*pt_proc;
144 	struct pgrp	*pt_pgrp;
145 };
146 #if 1	/* strongly typed cast - should be a noop */
147 static inline uint p2u(struct proc *p) { return (uint)(uintptr_t)p; }
148 #else
149 #define p2u(p) ((uint)p)
150 #endif
151 #define P_VALID(p) (!(p2u(p) & 1))
152 #define P_NEXT(p) (p2u(p) >> 1)
153 #define P_FREE(pid) ((struct proc *)(uintptr_t)((pid) << 1 | 1))
154 
155 #define INITIAL_PID_TABLE_SIZE	(1 << 5)
156 static struct pid_table *pid_table;
157 static uint pid_tbl_mask = INITIAL_PID_TABLE_SIZE - 1;
158 static uint pid_alloc_lim;	/* max we allocate before growing table */
159 static uint pid_alloc_cnt;	/* number of allocated pids */
160 
161 /* links through free slots - never empty! */
162 static uint next_free_pt, last_free_pt;
163 static pid_t pid_max = PID_MAX;		/* largest value we allocate */
164 
165 /* Components of the first process -- never freed. */
166 struct session session0;
167 struct pgrp pgrp0;
168 struct proc proc0;
169 struct lwp lwp0;
170 kauth_cred_t cred0;
171 struct filedesc0 filedesc0;
172 struct cwdinfo cwdi0;
173 struct plimit limit0;
174 struct pstats pstat0;
175 struct vmspace vmspace0;
176 struct sigacts sigacts0;
177 
178 extern struct user *proc0paddr;
179 
180 extern const struct emul emul_netbsd;	/* defined in kern_exec.c */
181 
182 int nofile = NOFILE;
183 int maxuprc = MAXUPRC;
184 int cmask = CMASK;
185 
186 POOL_INIT(proc_pool, sizeof(struct proc), 0, 0, 0, "procpl",
187     &pool_allocator_nointr);
188 POOL_INIT(pgrp_pool, sizeof(struct pgrp), 0, 0, 0, "pgrppl",
189     &pool_allocator_nointr);
190 POOL_INIT(plimit_pool, sizeof(struct plimit), 0, 0, 0, "plimitpl",
191     &pool_allocator_nointr);
192 POOL_INIT(pstats_pool, sizeof(struct pstats), 0, 0, 0, "pstatspl",
193     &pool_allocator_nointr);
194 POOL_INIT(rusage_pool, sizeof(struct rusage), 0, 0, 0, "rusgepl",
195     &pool_allocator_nointr);
196 POOL_INIT(session_pool, sizeof(struct session), 0, 0, 0, "sessionpl",
197     &pool_allocator_nointr);
198 
199 MALLOC_DEFINE(M_EMULDATA, "emuldata", "Per-process emulation data");
200 MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
201 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
202 
203 /*
204  * The process list descriptors, used during pid allocation and
205  * by sysctl.  No locking on this data structure is needed since
206  * it is completely static.
207  */
208 const struct proclist_desc proclists[] = {
209 	{ &allproc	},
210 	{ &zombproc	},
211 	{ NULL		},
212 };
213 
214 static void orphanpg(struct pgrp *);
215 static void pg_delete(pid_t);
216 
217 static specificdata_domain_t proc_specificdata_domain;
218 
219 /*
220  * Initialize global process hashing structures.
221  */
222 void
223 procinit(void)
224 {
225 	const struct proclist_desc *pd;
226 	int i;
227 #define	LINK_EMPTY ((PID_MAX + INITIAL_PID_TABLE_SIZE) & ~(INITIAL_PID_TABLE_SIZE - 1))
228 
229 	for (pd = proclists; pd->pd_list != NULL; pd++)
230 		LIST_INIT(pd->pd_list);
231 
232 	spinlockinit(&proclist_lock, "proclk", 0);
233 
234 	pid_table = malloc(INITIAL_PID_TABLE_SIZE * sizeof *pid_table,
235 			    M_PROC, M_WAITOK);
236 	/* Set free list running through table...
237 	   Preset 'use count' above PID_MAX so we allocate pid 1 next. */
238 	for (i = 0; i <= pid_tbl_mask; i++) {
239 		pid_table[i].pt_proc = P_FREE(LINK_EMPTY + i + 1);
240 		pid_table[i].pt_pgrp = 0;
241 	}
242 	/* slot 0 is just grabbed */
243 	next_free_pt = 1;
244 	/* Need to fix last entry. */
245 	last_free_pt = pid_tbl_mask;
246 	pid_table[last_free_pt].pt_proc = P_FREE(LINK_EMPTY);
247 	/* point at which we grow table - to avoid reusing pids too often */
248 	pid_alloc_lim = pid_tbl_mask - 1;
249 #undef LINK_EMPTY
250 
251 	LIST_INIT(&alllwp);
252 
253 	uihashtbl =
254 	    hashinit(maxproc / 16, HASH_LIST, M_PROC, M_WAITOK, &uihash);
255 
256 	proc_specificdata_domain = specificdata_domain_create();
257 	KASSERT(proc_specificdata_domain != NULL);
258 }
259 
260 /*
261  * Initialize process 0.
262  */
263 void
264 proc0_init(void)
265 {
266 	struct proc *p;
267 	struct pgrp *pg;
268 	struct session *sess;
269 	struct lwp *l;
270 	int s;
271 	u_int i;
272 	rlim_t lim;
273 
274 	p = &proc0;
275 	pg = &pgrp0;
276 	sess = &session0;
277 	l = &lwp0;
278 
279 	simple_lock_init(&p->p_lock);
280 	LIST_INIT(&p->p_lwps);
281 	LIST_INSERT_HEAD(&p->p_lwps, l, l_sibling);
282 	p->p_nlwps = 1;
283 	simple_lock_init(&p->p_sigctx.ps_silock);
284 	CIRCLEQ_INIT(&p->p_sigctx.ps_siginfo);
285 
286 	s = proclist_lock_write();
287 
288 	pid_table[0].pt_proc = p;
289 	LIST_INSERT_HEAD(&allproc, p, p_list);
290 	LIST_INSERT_HEAD(&alllwp, l, l_list);
291 
292 	p->p_pgrp = pg;
293 	pid_table[0].pt_pgrp = pg;
294 	LIST_INIT(&pg->pg_members);
295 	LIST_INSERT_HEAD(&pg->pg_members, p, p_pglist);
296 
297 	pg->pg_session = sess;
298 	sess->s_count = 1;
299 	sess->s_sid = 0;
300 	sess->s_leader = p;
301 
302 	proclist_unlock_write(s);
303 
304 	/*
305 	 * Set P_NOCLDWAIT so that kernel threads are reparented to
306 	 * init(8) when they exit.  init(8) can easily wait them out
307 	 * for us.
308 	 */
309 	p->p_flag = P_SYSTEM | P_NOCLDWAIT;
310 	p->p_stat = SACTIVE;
311 	p->p_nice = NZERO;
312 	p->p_emul = &emul_netbsd;
313 #ifdef __HAVE_SYSCALL_INTERN
314 	(*p->p_emul->e_syscall_intern)(p);
315 #endif
316 	strncpy(p->p_comm, "swapper", MAXCOMLEN);
317 
318 	l->l_flag = L_INMEM;
319 	l->l_stat = LSONPROC;
320 	p->p_nrlwps = 1;
321 
322 	callout_init(&l->l_tsleep_ch);
323 
324 	/* Create credentials. */
325 	cred0 = kauth_cred_alloc();
326 	p->p_cred = cred0;
327 	lwp_update_creds(l);
328 
329 	/* Create the CWD info. */
330 	p->p_cwdi = &cwdi0;
331 	cwdi0.cwdi_cmask = cmask;
332 	cwdi0.cwdi_refcnt = 1;
333 	simple_lock_init(&cwdi0.cwdi_slock);
334 
335 	/* Create the limits structures. */
336 	p->p_limit = &limit0;
337 	simple_lock_init(&limit0.p_slock);
338 	for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++)
339 		limit0.pl_rlimit[i].rlim_cur =
340 		    limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
341 
342 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
343 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur =
344 	    maxfiles < nofile ? maxfiles : nofile;
345 
346 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
347 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur =
348 	    maxproc < maxuprc ? maxproc : maxuprc;
349 
350 	lim = ptoa(uvmexp.free);
351 	limit0.pl_rlimit[RLIMIT_RSS].rlim_max = lim;
352 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = lim;
353 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = lim / 3;
354 	limit0.pl_corename = defcorename;
355 	limit0.p_refcnt = 1;
356 
357 	/* Configure virtual memory system, set vm rlimits. */
358 	uvm_init_limits(p);
359 
360 	/* Initialize file descriptor table for proc0. */
361 	p->p_fd = &filedesc0.fd_fd;
362 	fdinit1(&filedesc0);
363 
364 	/*
365 	 * Initialize proc0's vmspace, which uses the kernel pmap.
366 	 * All kernel processes (which never have user space mappings)
367 	 * share proc0's vmspace, and thus, the kernel pmap.
368 	 */
369 	uvmspace_init(&vmspace0, pmap_kernel(), round_page(VM_MIN_ADDRESS),
370 	    trunc_page(VM_MAX_ADDRESS));
371 	p->p_vmspace = &vmspace0;
372 
373 	l->l_addr = proc0paddr;				/* XXX */
374 
375 	p->p_stats = &pstat0;
376 
377 	/* Initialize signal state for proc0. */
378 	p->p_sigacts = &sigacts0;
379 	siginit(p);
380 
381 	proc_initspecific(p);
382 	lwp_initspecific(l);
383 }
384 
385 /*
386  * Acquire a read lock on the proclist.
387  */
388 void
389 proclist_lock_read(void)
390 {
391 	int error;
392 
393 	error = spinlockmgr(&proclist_lock, LK_SHARED, NULL);
394 #ifdef DIAGNOSTIC
395 	if (__predict_false(error != 0))
396 		panic("proclist_lock_read: failed to acquire lock");
397 #endif
398 }
399 
400 /*
401  * Release a read lock on the proclist.
402  */
403 void
404 proclist_unlock_read(void)
405 {
406 
407 	(void) spinlockmgr(&proclist_lock, LK_RELEASE, NULL);
408 }
409 
410 /*
411  * Acquire a write lock on the proclist.
412  */
413 int
414 proclist_lock_write(void)
415 {
416 	int s, error;
417 
418 	s = splclock();
419 	error = spinlockmgr(&proclist_lock, LK_EXCLUSIVE, NULL);
420 #ifdef DIAGNOSTIC
421 	if (__predict_false(error != 0))
422 		panic("proclist_lock: failed to acquire lock");
423 #endif
424 	return s;
425 }
426 
427 /*
428  * Release a write lock on the proclist.
429  */
430 void
431 proclist_unlock_write(int s)
432 {
433 
434 	(void) spinlockmgr(&proclist_lock, LK_RELEASE, NULL);
435 	splx(s);
436 }
437 
438 /*
439  * Check that the specified process group is in the session of the
440  * specified process.
441  * Treats -ve ids as process ids.
442  * Used to validate TIOCSPGRP requests.
443  */
444 int
445 pgid_in_session(struct proc *p, pid_t pg_id)
446 {
447 	struct pgrp *pgrp;
448 
449 	if (pg_id < 0) {
450 		struct proc *p1 = pfind(-pg_id);
451 		if (p1 == NULL)
452 			return EINVAL;
453 		pgrp = p1->p_pgrp;
454 	} else {
455 		pgrp = pgfind(pg_id);
456 		if (pgrp == NULL)
457 			return EINVAL;
458 	}
459 	if (pgrp->pg_session != p->p_pgrp->pg_session)
460 		return EPERM;
461 	return 0;
462 }
463 
464 /*
465  * Is p an inferior of q?
466  *
467  * Call with the proclist_lock held.
468  */
469 int
470 inferior(struct proc *p, struct proc *q)
471 {
472 
473 	for (; p != q; p = p->p_pptr)
474 		if (p->p_pid == 0)
475 			return 0;
476 	return 1;
477 }
478 
479 /*
480  * Locate a process by number
481  */
482 struct proc *
483 p_find(pid_t pid, uint flags)
484 {
485 	struct proc *p;
486 	char stat;
487 
488 	if (!(flags & PFIND_LOCKED))
489 		proclist_lock_read();
490 	p = pid_table[pid & pid_tbl_mask].pt_proc;
491 	/* Only allow live processes to be found by pid. */
492 	if (P_VALID(p) && p->p_pid == pid &&
493 	    ((stat = p->p_stat) == SACTIVE || stat == SSTOP
494 		    || (stat == SZOMB && (flags & PFIND_ZOMBIE)))) {
495 		if (flags & PFIND_UNLOCK_OK)
496 			 proclist_unlock_read();
497 		return p;
498 	}
499 	if (flags & PFIND_UNLOCK_FAIL)
500 		 proclist_unlock_read();
501 	return NULL;
502 }
503 
504 
505 /*
506  * Locate a process group by number
507  */
508 struct pgrp *
509 pg_find(pid_t pgid, uint flags)
510 {
511 	struct pgrp *pg;
512 
513 	if (!(flags & PFIND_LOCKED))
514 		proclist_lock_read();
515 	pg = pid_table[pgid & pid_tbl_mask].pt_pgrp;
516 	/*
517 	 * Can't look up a pgrp that only exists because the session
518 	 * hasn't died yet (traditional)
519 	 */
520 	if (pg == NULL || pg->pg_id != pgid || LIST_EMPTY(&pg->pg_members)) {
521 		if (flags & PFIND_UNLOCK_FAIL)
522 			 proclist_unlock_read();
523 		return NULL;
524 	}
525 
526 	if (flags & PFIND_UNLOCK_OK)
527 		proclist_unlock_read();
528 	return pg;
529 }
530 
531 static void
532 expand_pid_table(void)
533 {
534 	uint pt_size = pid_tbl_mask + 1;
535 	struct pid_table *n_pt, *new_pt;
536 	struct proc *proc;
537 	struct pgrp *pgrp;
538 	int i;
539 	int s;
540 	pid_t pid;
541 
542 	new_pt = malloc(pt_size * 2 * sizeof *new_pt, M_PROC, M_WAITOK);
543 
544 	s = proclist_lock_write();
545 	if (pt_size != pid_tbl_mask + 1) {
546 		/* Another process beat us to it... */
547 		proclist_unlock_write(s);
548 		FREE(new_pt, M_PROC);
549 		return;
550 	}
551 
552 	/*
553 	 * Copy entries from old table into new one.
554 	 * If 'pid' is 'odd' we need to place in the upper half,
555 	 * even pid's to the lower half.
556 	 * Free items stay in the low half so we don't have to
557 	 * fixup the reference to them.
558 	 * We stuff free items on the front of the freelist
559 	 * because we can't write to unmodified entries.
560 	 * Processing the table backwards maintains a semblance
561 	 * of issueing pid numbers that increase with time.
562 	 */
563 	i = pt_size - 1;
564 	n_pt = new_pt + i;
565 	for (; ; i--, n_pt--) {
566 		proc = pid_table[i].pt_proc;
567 		pgrp = pid_table[i].pt_pgrp;
568 		if (!P_VALID(proc)) {
569 			/* Up 'use count' so that link is valid */
570 			pid = (P_NEXT(proc) + pt_size) & ~pt_size;
571 			proc = P_FREE(pid);
572 			if (pgrp)
573 				pid = pgrp->pg_id;
574 		} else
575 			pid = proc->p_pid;
576 
577 		/* Save entry in appropriate half of table */
578 		n_pt[pid & pt_size].pt_proc = proc;
579 		n_pt[pid & pt_size].pt_pgrp = pgrp;
580 
581 		/* Put other piece on start of free list */
582 		pid = (pid ^ pt_size) & ~pid_tbl_mask;
583 		n_pt[pid & pt_size].pt_proc =
584 				    P_FREE((pid & ~pt_size) | next_free_pt);
585 		n_pt[pid & pt_size].pt_pgrp = 0;
586 		next_free_pt = i | (pid & pt_size);
587 		if (i == 0)
588 			break;
589 	}
590 
591 	/* Switch tables */
592 	n_pt = pid_table;
593 	pid_table = new_pt;
594 	pid_tbl_mask = pt_size * 2 - 1;
595 
596 	/*
597 	 * pid_max starts as PID_MAX (= 30000), once we have 16384
598 	 * allocated pids we need it to be larger!
599 	 */
600 	if (pid_tbl_mask > PID_MAX) {
601 		pid_max = pid_tbl_mask * 2 + 1;
602 		pid_alloc_lim |= pid_alloc_lim << 1;
603 	} else
604 		pid_alloc_lim <<= 1;	/* doubles number of free slots... */
605 
606 	proclist_unlock_write(s);
607 	FREE(n_pt, M_PROC);
608 }
609 
610 struct proc *
611 proc_alloc(void)
612 {
613 	struct proc *p;
614 	int s, nxt;
615 	pid_t pid;
616 	struct pid_table *pt;
617 
618 	p = pool_get(&proc_pool, PR_WAITOK);
619 	p->p_stat = SIDL;			/* protect against others */
620 
621 	proc_initspecific(p);
622 	/* allocate next free pid */
623 
624 	for (;;expand_pid_table()) {
625 		if (__predict_false(pid_alloc_cnt >= pid_alloc_lim))
626 			/* ensure pids cycle through 2000+ values */
627 			continue;
628 		s = proclist_lock_write();
629 		pt = &pid_table[next_free_pt];
630 #ifdef DIAGNOSTIC
631 		if (__predict_false(P_VALID(pt->pt_proc) || pt->pt_pgrp))
632 			panic("proc_alloc: slot busy");
633 #endif
634 		nxt = P_NEXT(pt->pt_proc);
635 		if (nxt & pid_tbl_mask)
636 			break;
637 		/* Table full - expand (NB last entry not used....) */
638 		proclist_unlock_write(s);
639 	}
640 
641 	/* pid is 'saved use count' + 'size' + entry */
642 	pid = (nxt & ~pid_tbl_mask) + pid_tbl_mask + 1 + next_free_pt;
643 	if ((uint)pid > (uint)pid_max)
644 		pid &= pid_tbl_mask;
645 	p->p_pid = pid;
646 	next_free_pt = nxt & pid_tbl_mask;
647 
648 	/* Grab table slot */
649 	pt->pt_proc = p;
650 	pid_alloc_cnt++;
651 
652 	proclist_unlock_write(s);
653 
654 	return p;
655 }
656 
657 /*
658  * Free last resources of a process - called from proc_free (in kern_exit.c)
659  */
660 void
661 proc_free_mem(struct proc *p)
662 {
663 	int s;
664 	pid_t pid = p->p_pid;
665 	struct pid_table *pt;
666 
667 	s = proclist_lock_write();
668 
669 	pt = &pid_table[pid & pid_tbl_mask];
670 #ifdef DIAGNOSTIC
671 	if (__predict_false(pt->pt_proc != p))
672 		panic("proc_free: pid_table mismatch, pid %x, proc %p",
673 			pid, p);
674 #endif
675 	/* save pid use count in slot */
676 	pt->pt_proc = P_FREE(pid & ~pid_tbl_mask);
677 
678 	if (pt->pt_pgrp == NULL) {
679 		/* link last freed entry onto ours */
680 		pid &= pid_tbl_mask;
681 		pt = &pid_table[last_free_pt];
682 		pt->pt_proc = P_FREE(P_NEXT(pt->pt_proc) | pid);
683 		last_free_pt = pid;
684 		pid_alloc_cnt--;
685 	}
686 
687 	nprocs--;
688 	proclist_unlock_write(s);
689 
690 	pool_put(&proc_pool, p);
691 }
692 
693 /*
694  * Move p to a new or existing process group (and session)
695  *
696  * If we are creating a new pgrp, the pgid should equal
697  * the calling process' pid.
698  * If is only valid to enter a process group that is in the session
699  * of the process.
700  * Also mksess should only be set if we are creating a process group
701  *
702  * Only called from sys_setsid, sys_setpgid/sys_setpgrp and the
703  * SYSV setpgrp support for hpux == enterpgrp(curproc, curproc->p_pid)
704  */
705 int
706 enterpgrp(struct proc *p, pid_t pgid, int mksess)
707 {
708 	struct pgrp *new_pgrp, *pgrp;
709 	struct session *sess;
710 	struct proc *curp = curproc;
711 	pid_t pid = p->p_pid;
712 	int rval;
713 	int s;
714 	pid_t pg_id = NO_PGID;
715 
716 	/* Allocate data areas we might need before doing any validity checks */
717 	proclist_lock_read();		/* Because pid_table might change */
718 	if (pid_table[pgid & pid_tbl_mask].pt_pgrp == 0) {
719 		proclist_unlock_read();
720 		new_pgrp = pool_get(&pgrp_pool, PR_WAITOK);
721 	} else {
722 		proclist_unlock_read();
723 		new_pgrp = NULL;
724 	}
725 	if (mksess)
726 		sess = pool_get(&session_pool, M_WAITOK);
727 	else
728 		sess = NULL;
729 
730 	s = proclist_lock_write();
731 	rval = EPERM;	/* most common error (to save typing) */
732 
733 	/* Check pgrp exists or can be created */
734 	pgrp = pid_table[pgid & pid_tbl_mask].pt_pgrp;
735 	if (pgrp != NULL && pgrp->pg_id != pgid)
736 		goto done;
737 
738 	/* Can only set another process under restricted circumstances. */
739 	if (p != curp) {
740 		/* must exist and be one of our children... */
741 		if (p != pid_table[pid & pid_tbl_mask].pt_proc
742 		    || !inferior(p, curp)) {
743 			rval = ESRCH;
744 			goto done;
745 		}
746 		/* ... in the same session... */
747 		if (sess != NULL || p->p_session != curp->p_session)
748 			goto done;
749 		/* ... existing pgid must be in same session ... */
750 		if (pgrp != NULL && pgrp->pg_session != p->p_session)
751 			goto done;
752 		/* ... and not done an exec. */
753 		if (p->p_flag & P_EXEC) {
754 			rval = EACCES;
755 			goto done;
756 		}
757 	}
758 
759 	/* Changing the process group/session of a session
760 	   leader is definitely off limits. */
761 	if (SESS_LEADER(p)) {
762 		if (sess == NULL && p->p_pgrp == pgrp)
763 			/* unless it's a definite noop */
764 			rval = 0;
765 		goto done;
766 	}
767 
768 	/* Can only create a process group with id of process */
769 	if (pgrp == NULL && pgid != pid)
770 		goto done;
771 
772 	/* Can only create a session if creating pgrp */
773 	if (sess != NULL && pgrp != NULL)
774 		goto done;
775 
776 	/* Check we allocated memory for a pgrp... */
777 	if (pgrp == NULL && new_pgrp == NULL)
778 		goto done;
779 
780 	/* Don't attach to 'zombie' pgrp */
781 	if (pgrp != NULL && LIST_EMPTY(&pgrp->pg_members))
782 		goto done;
783 
784 	/* Expect to succeed now */
785 	rval = 0;
786 
787 	if (pgrp == p->p_pgrp)
788 		/* nothing to do */
789 		goto done;
790 
791 	/* Ok all setup, link up required structures */
792 	if (pgrp == NULL) {
793 		pgrp = new_pgrp;
794 		new_pgrp = 0;
795 		if (sess != NULL) {
796 			sess->s_sid = p->p_pid;
797 			sess->s_leader = p;
798 			sess->s_count = 1;
799 			sess->s_ttyvp = NULL;
800 			sess->s_ttyp = NULL;
801 			sess->s_flags = p->p_session->s_flags & ~S_LOGIN_SET;
802 			memcpy(sess->s_login, p->p_session->s_login,
803 			    sizeof(sess->s_login));
804 			p->p_flag &= ~P_CONTROLT;
805 		} else {
806 			sess = p->p_pgrp->pg_session;
807 			SESSHOLD(sess);
808 		}
809 		pgrp->pg_session = sess;
810 		sess = 0;
811 
812 		pgrp->pg_id = pgid;
813 		LIST_INIT(&pgrp->pg_members);
814 #ifdef DIAGNOSTIC
815 		if (__predict_false(pid_table[pgid & pid_tbl_mask].pt_pgrp))
816 			panic("enterpgrp: pgrp table slot in use");
817 		if (__predict_false(mksess && p != curp))
818 			panic("enterpgrp: mksession and p != curproc");
819 #endif
820 		pid_table[pgid & pid_tbl_mask].pt_pgrp = pgrp;
821 		pgrp->pg_jobc = 0;
822 	}
823 
824 	/*
825 	 * Adjust eligibility of affected pgrps to participate in job control.
826 	 * Increment eligibility counts before decrementing, otherwise we
827 	 * could reach 0 spuriously during the first call.
828 	 */
829 	fixjobc(p, pgrp, 1);
830 	fixjobc(p, p->p_pgrp, 0);
831 
832 	/* Move process to requested group */
833 	LIST_REMOVE(p, p_pglist);
834 	if (LIST_EMPTY(&p->p_pgrp->pg_members))
835 		/* defer delete until we've dumped the lock */
836 		pg_id = p->p_pgrp->pg_id;
837 	p->p_pgrp = pgrp;
838 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
839 
840     done:
841 	proclist_unlock_write(s);
842 	if (sess != NULL)
843 		pool_put(&session_pool, sess);
844 	if (new_pgrp != NULL)
845 		pool_put(&pgrp_pool, new_pgrp);
846 	if (pg_id != NO_PGID)
847 		pg_delete(pg_id);
848 #ifdef DEBUG_PGRP
849 	if (__predict_false(rval))
850 		printf("enterpgrp(%d,%d,%d), curproc %d, rval %d\n",
851 			pid, pgid, mksess, curp->p_pid, rval);
852 #endif
853 	return rval;
854 }
855 
856 /*
857  * Remove a process from its process group.
858  */
859 int
860 leavepgrp(struct proc *p)
861 {
862 	int s;
863 	struct pgrp *pgrp;
864 	pid_t pg_id;
865 
866 	s = proclist_lock_write();
867 	pgrp = p->p_pgrp;
868 	LIST_REMOVE(p, p_pglist);
869 	p->p_pgrp = NULL;
870 	pg_id = LIST_EMPTY(&pgrp->pg_members) ? pgrp->pg_id : NO_PGID;
871 	proclist_unlock_write(s);
872 
873 	if (pg_id != NO_PGID)
874 		pg_delete(pg_id);
875 	return 0;
876 }
877 
878 static void
879 pg_free(pid_t pg_id)
880 {
881 	struct pgrp *pgrp;
882 	struct pid_table *pt;
883 	int s;
884 
885 	s = proclist_lock_write();
886 	pt = &pid_table[pg_id & pid_tbl_mask];
887 	pgrp = pt->pt_pgrp;
888 #ifdef DIAGNOSTIC
889 	if (__predict_false(!pgrp || pgrp->pg_id != pg_id
890 	    || !LIST_EMPTY(&pgrp->pg_members)))
891 		panic("pg_free: process group absent or has members");
892 #endif
893 	pt->pt_pgrp = 0;
894 
895 	if (!P_VALID(pt->pt_proc)) {
896 		/* orphaned pgrp, put slot onto free list */
897 #ifdef DIAGNOSTIC
898 		if (__predict_false(P_NEXT(pt->pt_proc) & pid_tbl_mask))
899 			panic("pg_free: process slot on free list");
900 #endif
901 
902 		pg_id &= pid_tbl_mask;
903 		pt = &pid_table[last_free_pt];
904 		pt->pt_proc = P_FREE(P_NEXT(pt->pt_proc) | pg_id);
905 		last_free_pt = pg_id;
906 		pid_alloc_cnt--;
907 	}
908 	proclist_unlock_write(s);
909 
910 	pool_put(&pgrp_pool, pgrp);
911 }
912 
913 /*
914  * delete a process group
915  */
916 static void
917 pg_delete(pid_t pg_id)
918 {
919 	struct pgrp *pgrp;
920 	struct tty *ttyp;
921 	struct session *ss;
922 	int s, is_pgrp_leader;
923 
924 	s = proclist_lock_write();
925 	pgrp = pid_table[pg_id & pid_tbl_mask].pt_pgrp;
926 	if (pgrp == NULL || pgrp->pg_id != pg_id ||
927 	    !LIST_EMPTY(&pgrp->pg_members)) {
928 		proclist_unlock_write(s);
929 		return;
930 	}
931 
932 	ss = pgrp->pg_session;
933 
934 	/* Remove reference (if any) from tty to this process group */
935 	ttyp = ss->s_ttyp;
936 	if (ttyp != NULL && ttyp->t_pgrp == pgrp) {
937 		ttyp->t_pgrp = NULL;
938 #ifdef DIAGNOSTIC
939 		if (ttyp->t_session != ss)
940 			panic("pg_delete: wrong session on terminal");
941 #endif
942 	}
943 
944 	/*
945 	 * The leading process group in a session is freed
946 	 * by sessdelete() if last reference.
947 	 */
948 	is_pgrp_leader = (ss->s_sid == pgrp->pg_id);
949 	proclist_unlock_write(s);
950 	SESSRELE(ss);
951 
952 	if (is_pgrp_leader)
953 		return;
954 
955 	pg_free(pg_id);
956 }
957 
958 /*
959  * Delete session - called from SESSRELE when s_count becomes zero.
960  */
961 void
962 sessdelete(struct session *ss)
963 {
964 	/*
965 	 * We keep the pgrp with the same id as the session in
966 	 * order to stop a process being given the same pid.
967 	 * Since the pgrp holds a reference to the session, it
968 	 * must be a 'zombie' pgrp by now.
969 	 */
970 
971 	pg_free(ss->s_sid);
972 
973 	pool_put(&session_pool, ss);
974 }
975 
976 /*
977  * Adjust pgrp jobc counters when specified process changes process group.
978  * We count the number of processes in each process group that "qualify"
979  * the group for terminal job control (those with a parent in a different
980  * process group of the same session).  If that count reaches zero, the
981  * process group becomes orphaned.  Check both the specified process'
982  * process group and that of its children.
983  * entering == 0 => p is leaving specified group.
984  * entering == 1 => p is entering specified group.
985  *
986  * Call with proclist_lock held.
987  */
988 void
989 fixjobc(struct proc *p, struct pgrp *pgrp, int entering)
990 {
991 	struct pgrp *hispgrp;
992 	struct session *mysession = pgrp->pg_session;
993 	struct proc *child;
994 
995 	/*
996 	 * Check p's parent to see whether p qualifies its own process
997 	 * group; if so, adjust count for p's process group.
998 	 */
999 	hispgrp = p->p_pptr->p_pgrp;
1000 	if (hispgrp != pgrp && hispgrp->pg_session == mysession) {
1001 		if (entering)
1002 			pgrp->pg_jobc++;
1003 		else if (--pgrp->pg_jobc == 0)
1004 			orphanpg(pgrp);
1005 	}
1006 
1007 	/*
1008 	 * Check this process' children to see whether they qualify
1009 	 * their process groups; if so, adjust counts for children's
1010 	 * process groups.
1011 	 */
1012 	LIST_FOREACH(child, &p->p_children, p_sibling) {
1013 		hispgrp = child->p_pgrp;
1014 		if (hispgrp != pgrp && hispgrp->pg_session == mysession &&
1015 		    !P_ZOMBIE(child)) {
1016 			if (entering)
1017 				hispgrp->pg_jobc++;
1018 			else if (--hispgrp->pg_jobc == 0)
1019 				orphanpg(hispgrp);
1020 		}
1021 	}
1022 }
1023 
1024 /*
1025  * A process group has become orphaned;
1026  * if there are any stopped processes in the group,
1027  * hang-up all process in that group.
1028  *
1029  * Call with proclist_lock held.
1030  */
1031 static void
1032 orphanpg(struct pgrp *pg)
1033 {
1034 	struct proc *p;
1035 
1036 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
1037 		if (p->p_stat == SSTOP) {
1038 			LIST_FOREACH(p, &pg->pg_members, p_pglist) {
1039 				psignal(p, SIGHUP);
1040 				psignal(p, SIGCONT);
1041 			}
1042 			return;
1043 		}
1044 	}
1045 }
1046 
1047 /* mark process as suid/sgid, reset some values to defaults */
1048 void
1049 p_sugid(struct proc *p)
1050 {
1051 	struct plimit *lim;
1052 	char *cn;
1053 
1054 	p->p_flag |= P_SUGID;
1055 	/* reset what needs to be reset in plimit */
1056 	lim = p->p_limit;
1057 	if (lim->pl_corename != defcorename) {
1058 		if (lim->p_refcnt > 1 &&
1059 		    (lim->p_lflags & PL_SHAREMOD) == 0) {
1060 			p->p_limit = limcopy(lim);
1061 			limfree(lim);
1062 			lim = p->p_limit;
1063 		}
1064 		simple_lock(&lim->p_slock);
1065 		cn = lim->pl_corename;
1066 		lim->pl_corename = defcorename;
1067 		simple_unlock(&lim->p_slock);
1068 		if (cn != defcorename)
1069 			free(cn, M_TEMP);
1070 	}
1071 }
1072 
1073 #ifdef DDB
1074 #include <ddb/db_output.h>
1075 void pidtbl_dump(void);
1076 void
1077 pidtbl_dump(void)
1078 {
1079 	struct pid_table *pt;
1080 	struct proc *p;
1081 	struct pgrp *pgrp;
1082 	int id;
1083 
1084 	db_printf("pid table %p size %x, next %x, last %x\n",
1085 		pid_table, pid_tbl_mask+1,
1086 		next_free_pt, last_free_pt);
1087 	for (pt = pid_table, id = 0; id <= pid_tbl_mask; id++, pt++) {
1088 		p = pt->pt_proc;
1089 		if (!P_VALID(p) && !pt->pt_pgrp)
1090 			continue;
1091 		db_printf("  id %x: ", id);
1092 		if (P_VALID(p))
1093 			db_printf("proc %p id %d (0x%x) %s\n",
1094 				p, p->p_pid, p->p_pid, p->p_comm);
1095 		else
1096 			db_printf("next %x use %x\n",
1097 				P_NEXT(p) & pid_tbl_mask,
1098 				P_NEXT(p) & ~pid_tbl_mask);
1099 		if ((pgrp = pt->pt_pgrp)) {
1100 			db_printf("\tsession %p, sid %d, count %d, login %s\n",
1101 			    pgrp->pg_session, pgrp->pg_session->s_sid,
1102 			    pgrp->pg_session->s_count,
1103 			    pgrp->pg_session->s_login);
1104 			db_printf("\tpgrp %p, pg_id %d, pg_jobc %d, members %p\n",
1105 			    pgrp, pgrp->pg_id, pgrp->pg_jobc,
1106 			    pgrp->pg_members.lh_first);
1107 			for (p = pgrp->pg_members.lh_first; p != 0;
1108 			    p = p->p_pglist.le_next) {
1109 				db_printf("\t\tpid %d addr %p pgrp %p %s\n",
1110 				    p->p_pid, p, p->p_pgrp, p->p_comm);
1111 			}
1112 		}
1113 	}
1114 }
1115 #endif /* DDB */
1116 
1117 #ifdef KSTACK_CHECK_MAGIC
1118 #include <sys/user.h>
1119 
1120 #define	KSTACK_MAGIC	0xdeadbeaf
1121 
1122 /* XXX should be per process basis? */
1123 int kstackleftmin = KSTACK_SIZE;
1124 int kstackleftthres = KSTACK_SIZE / 8; /* warn if remaining stack is
1125 					  less than this */
1126 
1127 void
1128 kstack_setup_magic(const struct lwp *l)
1129 {
1130 	uint32_t *ip;
1131 	uint32_t const *end;
1132 
1133 	KASSERT(l != NULL);
1134 	KASSERT(l != &lwp0);
1135 
1136 	/*
1137 	 * fill all the stack with magic number
1138 	 * so that later modification on it can be detected.
1139 	 */
1140 	ip = (uint32_t *)KSTACK_LOWEST_ADDR(l);
1141 	end = (uint32_t *)((caddr_t)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE);
1142 	for (; ip < end; ip++) {
1143 		*ip = KSTACK_MAGIC;
1144 	}
1145 }
1146 
1147 void
1148 kstack_check_magic(const struct lwp *l)
1149 {
1150 	uint32_t const *ip, *end;
1151 	int stackleft;
1152 
1153 	KASSERT(l != NULL);
1154 
1155 	/* don't check proc0 */ /*XXX*/
1156 	if (l == &lwp0)
1157 		return;
1158 
1159 #ifdef __MACHINE_STACK_GROWS_UP
1160 	/* stack grows upwards (eg. hppa) */
1161 	ip = (uint32_t *)((caddr_t)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE);
1162 	end = (uint32_t *)KSTACK_LOWEST_ADDR(l);
1163 	for (ip--; ip >= end; ip--)
1164 		if (*ip != KSTACK_MAGIC)
1165 			break;
1166 
1167 	stackleft = (caddr_t)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE - (caddr_t)ip;
1168 #else /* __MACHINE_STACK_GROWS_UP */
1169 	/* stack grows downwards (eg. i386) */
1170 	ip = (uint32_t *)KSTACK_LOWEST_ADDR(l);
1171 	end = (uint32_t *)((caddr_t)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE);
1172 	for (; ip < end; ip++)
1173 		if (*ip != KSTACK_MAGIC)
1174 			break;
1175 
1176 	stackleft = ((const char *)ip) - (const char *)KSTACK_LOWEST_ADDR(l);
1177 #endif /* __MACHINE_STACK_GROWS_UP */
1178 
1179 	if (kstackleftmin > stackleft) {
1180 		kstackleftmin = stackleft;
1181 		if (stackleft < kstackleftthres)
1182 			printf("warning: kernel stack left %d bytes"
1183 			    "(pid %u:lid %u)\n", stackleft,
1184 			    (u_int)l->l_proc->p_pid, (u_int)l->l_lid);
1185 	}
1186 
1187 	if (stackleft <= 0) {
1188 		panic("magic on the top of kernel stack changed for "
1189 		    "pid %u, lid %u: maybe kernel stack overflow",
1190 		    (u_int)l->l_proc->p_pid, (u_int)l->l_lid);
1191 	}
1192 }
1193 #endif /* KSTACK_CHECK_MAGIC */
1194 
1195 /* XXX shouldn't be here */
1196 #if defined(MULTIPROCESSOR) || defined(LOCKDEBUG)
1197 #define	PROCLIST_ASSERT_LOCKED_READ()	\
1198 	KASSERT(lockstatus(&proclist_lock) == LK_SHARED)
1199 #else
1200 #define	PROCLIST_ASSERT_LOCKED_READ()	/* nothing */
1201 #endif
1202 
1203 int
1204 proclist_foreach_call(struct proclist *list,
1205     int (*callback)(struct proc *, void *arg), void *arg)
1206 {
1207 	struct proc marker;
1208 	struct proc *p;
1209 	struct lwp * const l = curlwp;
1210 	int ret = 0;
1211 
1212 	marker.p_flag = P_MARKER;
1213 	PHOLD(l);
1214 	proclist_lock_read();
1215 	for (p = LIST_FIRST(list); ret == 0 && p != NULL;) {
1216 		if (p->p_flag & P_MARKER) {
1217 			p = LIST_NEXT(p, p_list);
1218 			continue;
1219 		}
1220 		LIST_INSERT_AFTER(p, &marker, p_list);
1221 		ret = (*callback)(p, arg);
1222 		PROCLIST_ASSERT_LOCKED_READ();
1223 		p = LIST_NEXT(&marker, p_list);
1224 		LIST_REMOVE(&marker, p_list);
1225 	}
1226 	proclist_unlock_read();
1227 	PRELE(l);
1228 
1229 	return ret;
1230 }
1231 
1232 int
1233 proc_vmspace_getref(struct proc *p, struct vmspace **vm)
1234 {
1235 
1236 	/* XXXCDC: how should locking work here? */
1237 
1238 	/* curproc exception is for coredump. */
1239 
1240 	if ((p != curproc && (p->p_flag & P_WEXIT) != 0) ||
1241 	    (p->p_vmspace->vm_refcnt < 1)) { /* XXX */
1242 		return EFAULT;
1243 	}
1244 
1245 	uvmspace_addref(p->p_vmspace);
1246 	*vm = p->p_vmspace;
1247 
1248 	return 0;
1249 }
1250 
1251 /*
1252  * Acquire a write lock on the process credential.
1253  */
1254 void
1255 proc_crmod_enter(struct proc *p)
1256 {
1257 
1258 	/*
1259 	 * XXXSMP This should be a lightweight sleep lock.  'struct lock' is
1260 	 * too large.
1261 	 */
1262 	simple_lock(&p->p_lock);
1263 	while ((p->p_flag & P_CRLOCK) != 0)
1264 		ltsleep(&p->p_cred, PLOCK, "crlock", 0, &p->p_lock);
1265 	p->p_flag |= P_CRLOCK;
1266 	simple_unlock(&p->p_lock);
1267 }
1268 
1269 /*
1270  * Block out readers, set in a new process credential, and drop the write
1271  * lock.  The credential must have a reference already.  Optionally, free a
1272  * no-longer required credential.
1273  */
1274 void
1275 proc_crmod_leave(struct proc *p, kauth_cred_t scred, kauth_cred_t fcred)
1276 {
1277 
1278 	KDASSERT((p->p_flag & P_CRLOCK) != 0);
1279 	simple_lock(&p->p_lock);
1280 	p->p_cred = scred;
1281 	p->p_flag &= ~P_CRLOCK;
1282 	simple_unlock(&p->p_lock);
1283 	wakeup(&p->p_cred);
1284 	if (fcred != NULL)
1285 		kauth_cred_free(fcred);
1286 }
1287 
1288 /*
1289  * proc_specific_key_create --
1290  *	Create a key for subsystem proc-specific data.
1291  */
1292 int
1293 proc_specific_key_create(specificdata_key_t *keyp, specificdata_dtor_t dtor)
1294 {
1295 
1296 	return (specificdata_key_create(proc_specificdata_domain, keyp, dtor));
1297 }
1298 
1299 /*
1300  * proc_specific_key_delete --
1301  *	Delete a key for subsystem proc-specific data.
1302  */
1303 void
1304 proc_specific_key_delete(specificdata_key_t key)
1305 {
1306 
1307 	specificdata_key_delete(proc_specificdata_domain, key);
1308 }
1309 
1310 /*
1311  * proc_initspecific --
1312  *	Initialize a proc's specificdata container.
1313  */
1314 void
1315 proc_initspecific(struct proc *p)
1316 {
1317 	int error;
1318 
1319 	error = specificdata_init(proc_specificdata_domain, &p->p_specdataref);
1320 	KASSERT(error == 0);
1321 }
1322 
1323 /*
1324  * proc_finispecific --
1325  *	Finalize a proc's specificdata container.
1326  */
1327 void
1328 proc_finispecific(struct proc *p)
1329 {
1330 
1331 	specificdata_fini(proc_specificdata_domain, &p->p_specdataref);
1332 }
1333 
1334 /*
1335  * proc_getspecific --
1336  *	Return proc-specific data corresponding to the specified key.
1337  */
1338 void *
1339 proc_getspecific(struct proc *p, specificdata_key_t key)
1340 {
1341 
1342 	return (specificdata_getspecific(proc_specificdata_domain,
1343 					 &p->p_specdataref, key));
1344 }
1345 
1346 /*
1347  * proc_setspecific --
1348  *	Set proc-specific data corresponding to the specified key.
1349  */
1350 void
1351 proc_setspecific(struct proc *p, specificdata_key_t key, void *data)
1352 {
1353 
1354 	specificdata_setspecific(proc_specificdata_domain,
1355 				 &p->p_specdataref, key, data);
1356 }
1357