xref: /netbsd-src/sys/kern/kern_proc.c (revision daf6c4152fcddc27c445489775ed1f66ab4ea9a9)
1 /*	$NetBSD: kern_proc.c,v 1.171 2011/01/28 20:31:10 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 1.171 2011/01/28 20:31:10 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/sa.h>
93 #include <sys/savar.h>
94 #include <sys/filedesc.h>
95 #include "sys/syscall_stats.h"
96 #include <sys/kauth.h>
97 #include <sys/sleepq.h>
98 #include <sys/atomic.h>
99 #include <sys/kmem.h>
100 #include <sys/dtrace_bsd.h>
101 #include <sys/sysctl.h>
102 #include <sys/exec.h>
103 #include <sys/cpu.h>
104 
105 #include <uvm/uvm_extern.h>
106 #include <uvm/uvm_extern.h>
107 
108 #ifdef COMPAT_NETBSD32
109 #include <compat/netbsd32/netbsd32.h>
110 #endif
111 
112 /*
113  * Other process lists
114  */
115 
116 struct proclist allproc;
117 struct proclist zombproc;	/* resources have been freed */
118 
119 kmutex_t	*proc_lock;
120 
121 /*
122  * pid to proc lookup is done by indexing the pid_table array.
123  * Since pid numbers are only allocated when an empty slot
124  * has been found, there is no need to search any lists ever.
125  * (an orphaned pgrp will lock the slot, a session will lock
126  * the pgrp with the same number.)
127  * If the table is too small it is reallocated with twice the
128  * previous size and the entries 'unzipped' into the two halves.
129  * A linked list of free entries is passed through the pt_proc
130  * field of 'free' items - set odd to be an invalid ptr.
131  */
132 
133 struct pid_table {
134 	struct proc	*pt_proc;
135 	struct pgrp	*pt_pgrp;
136 	pid_t		pt_pid;
137 };
138 #if 1	/* strongly typed cast - should be a noop */
139 static inline uint p2u(struct proc *p) { return (uint)(uintptr_t)p; }
140 #else
141 #define p2u(p) ((uint)p)
142 #endif
143 #define P_VALID(p) (!(p2u(p) & 1))
144 #define P_NEXT(p) (p2u(p) >> 1)
145 #define P_FREE(pid) ((struct proc *)(uintptr_t)((pid) << 1 | 1))
146 
147 #define INITIAL_PID_TABLE_SIZE	(1 << 5)
148 static struct pid_table *pid_table;
149 static uint pid_tbl_mask = INITIAL_PID_TABLE_SIZE - 1;
150 static uint pid_alloc_lim;	/* max we allocate before growing table */
151 static uint pid_alloc_cnt;	/* number of allocated pids */
152 
153 /* links through free slots - never empty! */
154 static uint next_free_pt, last_free_pt;
155 static pid_t pid_max = PID_MAX;		/* largest value we allocate */
156 
157 /* Components of the first process -- never freed. */
158 
159 extern struct emul emul_netbsd;	/* defined in kern_exec.c */
160 
161 struct session session0 = {
162 	.s_count = 1,
163 	.s_sid = 0,
164 };
165 struct pgrp pgrp0 = {
166 	.pg_members = LIST_HEAD_INITIALIZER(&pgrp0.pg_members),
167 	.pg_session = &session0,
168 };
169 filedesc_t filedesc0;
170 struct cwdinfo cwdi0 = {
171 	.cwdi_cmask = CMASK,		/* see cmask below */
172 	.cwdi_refcnt = 1,
173 };
174 struct plimit limit0;
175 struct pstats pstat0;
176 struct vmspace vmspace0;
177 struct sigacts sigacts0;
178 struct proc proc0 = {
179 	.p_lwps = LIST_HEAD_INITIALIZER(&proc0.p_lwps),
180 	.p_sigwaiters = LIST_HEAD_INITIALIZER(&proc0.p_sigwaiters),
181 	.p_nlwps = 1,
182 	.p_nrlwps = 1,
183 	.p_nlwpid = 1,		/* must match lwp0.l_lid */
184 	.p_pgrp = &pgrp0,
185 	.p_comm = "system",
186 	/*
187 	 * Set P_NOCLDWAIT so that kernel threads are reparented to init(8)
188 	 * when they exit.  init(8) can easily wait them out for us.
189 	 */
190 	.p_flag = PK_SYSTEM | PK_NOCLDWAIT,
191 	.p_stat = SACTIVE,
192 	.p_nice = NZERO,
193 	.p_emul = &emul_netbsd,
194 	.p_cwdi = &cwdi0,
195 	.p_limit = &limit0,
196 	.p_fd = &filedesc0,
197 	.p_vmspace = &vmspace0,
198 	.p_stats = &pstat0,
199 	.p_sigacts = &sigacts0,
200 };
201 kauth_cred_t cred0;
202 
203 int nofile = NOFILE;
204 int maxuprc = MAXUPRC;
205 int cmask = CMASK;
206 
207 MALLOC_DEFINE(M_EMULDATA, "emuldata", "Per-process emulation data");
208 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
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 
324 	/* Set free list running through table...
325 	   Preset 'use count' above PID_MAX so we allocate pid 1 next. */
326 	for (i = 0; i <= pid_tbl_mask; i++) {
327 		pid_table[i].pt_proc = P_FREE(LINK_EMPTY + i + 1);
328 		pid_table[i].pt_pgrp = 0;
329 		pid_table[i].pt_pid = 0;
330 	}
331 	/* slot 0 is just grabbed */
332 	next_free_pt = 1;
333 	/* Need to fix last entry. */
334 	last_free_pt = pid_tbl_mask;
335 	pid_table[last_free_pt].pt_proc = P_FREE(LINK_EMPTY);
336 	/* point at which we grow table - to avoid reusing pids too often */
337 	pid_alloc_lim = pid_tbl_mask - 1;
338 #undef LINK_EMPTY
339 
340 	proc_specificdata_domain = specificdata_domain_create();
341 	KASSERT(proc_specificdata_domain != NULL);
342 
343 	proc_cache = pool_cache_init(sizeof(struct proc), 0, 0, 0,
344 	    "procpl", NULL, IPL_NONE, NULL, NULL, NULL);
345 
346 	proc_listener = kauth_listen_scope(KAUTH_SCOPE_PROCESS,
347 	    proc_listener_cb, NULL);
348 }
349 
350 void
351 procinit_sysctl(void)
352 {
353 	static struct sysctllog *clog;
354 
355 	sysctl_createv(&clog, 0, NULL, NULL,
356 		       CTLFLAG_PERMANENT,
357 		       CTLTYPE_NODE, "kern", NULL,
358 		       NULL, 0, NULL, 0,
359 		       CTL_KERN, CTL_EOL);
360 
361 	sysctl_createv(&clog, 0, NULL, NULL,
362 		       CTLFLAG_PERMANENT,
363 		       CTLTYPE_NODE, "proc",
364 		       SYSCTL_DESCR("System-wide process information"),
365 		       sysctl_doeproc, 0, NULL, 0,
366 		       CTL_KERN, KERN_PROC, CTL_EOL);
367 	sysctl_createv(&clog, 0, NULL, NULL,
368 		       CTLFLAG_PERMANENT,
369 		       CTLTYPE_NODE, "proc2",
370 		       SYSCTL_DESCR("Machine-independent process information"),
371 		       sysctl_doeproc, 0, NULL, 0,
372 		       CTL_KERN, KERN_PROC2, CTL_EOL);
373 	sysctl_createv(&clog, 0, NULL, NULL,
374 		       CTLFLAG_PERMANENT,
375 		       CTLTYPE_NODE, "proc_args",
376 		       SYSCTL_DESCR("Process argument information"),
377 		       sysctl_kern_proc_args, 0, NULL, 0,
378 		       CTL_KERN, KERN_PROC_ARGS, CTL_EOL);
379 
380 	/*
381 	  "nodes" under these:
382 
383 	  KERN_PROC_ALL
384 	  KERN_PROC_PID pid
385 	  KERN_PROC_PGRP pgrp
386 	  KERN_PROC_SESSION sess
387 	  KERN_PROC_TTY tty
388 	  KERN_PROC_UID uid
389 	  KERN_PROC_RUID uid
390 	  KERN_PROC_GID gid
391 	  KERN_PROC_RGID gid
392 
393 	  all in all, probably not worth the effort...
394 	*/
395 }
396 
397 /*
398  * Initialize process 0.
399  */
400 void
401 proc0_init(void)
402 {
403 	struct proc *p;
404 	struct pgrp *pg;
405 	rlim_t lim;
406 	int i;
407 
408 	p = &proc0;
409 	pg = &pgrp0;
410 
411 	mutex_init(&p->p_stmutex, MUTEX_DEFAULT, IPL_HIGH);
412 	mutex_init(&p->p_auxlock, MUTEX_DEFAULT, IPL_NONE);
413 	p->p_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
414 
415 	rw_init(&p->p_reflock);
416 	cv_init(&p->p_waitcv, "wait");
417 	cv_init(&p->p_lwpcv, "lwpwait");
418 
419 	LIST_INSERT_HEAD(&p->p_lwps, &lwp0, l_sibling);
420 
421 	pid_table[0].pt_proc = p;
422 	LIST_INSERT_HEAD(&allproc, p, p_list);
423 
424 	pid_table[0].pt_pgrp = pg;
425 	LIST_INSERT_HEAD(&pg->pg_members, p, p_pglist);
426 
427 #ifdef __HAVE_SYSCALL_INTERN
428 	(*p->p_emul->e_syscall_intern)(p);
429 #endif
430 
431 	/* Create credentials. */
432 	cred0 = kauth_cred_alloc();
433 	p->p_cred = cred0;
434 
435 	/* Create the CWD info. */
436 	rw_init(&cwdi0.cwdi_lock);
437 
438 	/* Create the limits structures. */
439 	mutex_init(&limit0.pl_lock, MUTEX_DEFAULT, IPL_NONE);
440 	for (i = 0; i < __arraycount(limit0.pl_rlimit); i++)
441 		limit0.pl_rlimit[i].rlim_cur =
442 		    limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
443 
444 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
445 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur =
446 	    maxfiles < nofile ? maxfiles : nofile;
447 
448 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
449 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur =
450 	    maxproc < maxuprc ? maxproc : maxuprc;
451 
452 	lim = MIN(VM_MAXUSER_ADDRESS, ctob((rlim_t)uvmexp.free));
453 	limit0.pl_rlimit[RLIMIT_RSS].rlim_max = lim;
454 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = lim;
455 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = lim / 3;
456 	limit0.pl_corename = defcorename;
457 	limit0.pl_refcnt = 1;
458 	limit0.pl_sv_limit = NULL;
459 
460 	/* Configure virtual memory system, set vm rlimits. */
461 	uvm_init_limits(p);
462 
463 	/* Initialize file descriptor table for proc0. */
464 	fd_init(&filedesc0);
465 
466 	/*
467 	 * Initialize proc0's vmspace, which uses the kernel pmap.
468 	 * All kernel processes (which never have user space mappings)
469 	 * share proc0's vmspace, and thus, the kernel pmap.
470 	 */
471 	uvmspace_init(&vmspace0, pmap_kernel(), round_page(VM_MIN_ADDRESS),
472 	    trunc_page(VM_MAX_ADDRESS));
473 
474 	/* Initialize signal state for proc0. XXX IPL_SCHED */
475 	mutex_init(&p->p_sigacts->sa_mutex, MUTEX_DEFAULT, IPL_SCHED);
476 	siginit(p);
477 
478 	proc_initspecific(p);
479 	kdtrace_proc_ctor(NULL, p);
480 }
481 
482 /*
483  * Session reference counting.
484  */
485 
486 void
487 proc_sesshold(struct session *ss)
488 {
489 
490 	KASSERT(mutex_owned(proc_lock));
491 	ss->s_count++;
492 }
493 
494 void
495 proc_sessrele(struct session *ss)
496 {
497 
498 	KASSERT(mutex_owned(proc_lock));
499 	/*
500 	 * We keep the pgrp with the same id as the session in order to
501 	 * stop a process being given the same pid.  Since the pgrp holds
502 	 * a reference to the session, it must be a 'zombie' pgrp by now.
503 	 */
504 	if (--ss->s_count == 0) {
505 		struct pgrp *pg;
506 
507 		pg = pg_remove(ss->s_sid);
508 		mutex_exit(proc_lock);
509 
510 		kmem_free(pg, sizeof(struct pgrp));
511 		kmem_free(ss, sizeof(struct session));
512 	} else {
513 		mutex_exit(proc_lock);
514 	}
515 }
516 
517 /*
518  * Check that the specified process group is in the session of the
519  * specified process.
520  * Treats -ve ids as process ids.
521  * Used to validate TIOCSPGRP requests.
522  */
523 int
524 pgid_in_session(struct proc *p, pid_t pg_id)
525 {
526 	struct pgrp *pgrp;
527 	struct session *session;
528 	int error;
529 
530 	mutex_enter(proc_lock);
531 	if (pg_id < 0) {
532 		struct proc *p1 = proc_find(-pg_id);
533 		if (p1 == NULL) {
534 			error = EINVAL;
535 			goto fail;
536 		}
537 		pgrp = p1->p_pgrp;
538 	} else {
539 		pgrp = pgrp_find(pg_id);
540 		if (pgrp == NULL) {
541 			error = EINVAL;
542 			goto fail;
543 		}
544 	}
545 	session = pgrp->pg_session;
546 	error = (session != p->p_pgrp->pg_session) ? EPERM : 0;
547 fail:
548 	mutex_exit(proc_lock);
549 	return error;
550 }
551 
552 /*
553  * p_inferior: is p an inferior of q?
554  */
555 static inline bool
556 p_inferior(struct proc *p, struct proc *q)
557 {
558 
559 	KASSERT(mutex_owned(proc_lock));
560 
561 	for (; p != q; p = p->p_pptr)
562 		if (p->p_pid == 0)
563 			return false;
564 	return true;
565 }
566 
567 /*
568  * proc_find: locate a process by the ID.
569  *
570  * => Must be called with proc_lock held.
571  */
572 proc_t *
573 proc_find_raw(pid_t pid)
574 {
575 	struct pid_table *pt;
576 	proc_t *p;
577 
578 	KASSERT(mutex_owned(proc_lock));
579 	pt = &pid_table[pid & pid_tbl_mask];
580 	p = pt->pt_proc;
581 	if (__predict_false(!P_VALID(p) || pt->pt_pid != pid)) {
582 		return NULL;
583 	}
584 	return p;
585 }
586 
587 proc_t *
588 proc_find(pid_t pid)
589 {
590 	proc_t *p;
591 
592 	p = proc_find_raw(pid);
593 	if (__predict_false(p == NULL)) {
594 		return NULL;
595 	}
596 
597 	/*
598 	 * Only allow live processes to be found by PID.
599 	 * XXX: p_stat might change, since unlocked.
600 	 */
601 	if (__predict_true(p->p_stat == SACTIVE || p->p_stat == SSTOP)) {
602 		return p;
603 	}
604 	return NULL;
605 }
606 
607 /*
608  * pgrp_find: locate a process group by the ID.
609  *
610  * => Must be called with proc_lock held.
611  */
612 struct pgrp *
613 pgrp_find(pid_t pgid)
614 {
615 	struct pgrp *pg;
616 
617 	KASSERT(mutex_owned(proc_lock));
618 
619 	pg = pid_table[pgid & pid_tbl_mask].pt_pgrp;
620 
621 	/*
622 	 * Cannot look up a process group that only exists because the
623 	 * session has not died yet (traditional).
624 	 */
625 	if (pg == NULL || pg->pg_id != pgid || LIST_EMPTY(&pg->pg_members)) {
626 		return NULL;
627 	}
628 	return pg;
629 }
630 
631 static void
632 expand_pid_table(void)
633 {
634 	size_t pt_size, tsz;
635 	struct pid_table *n_pt, *new_pt;
636 	struct proc *proc;
637 	struct pgrp *pgrp;
638 	pid_t pid, rpid;
639 	u_int i;
640 	uint new_pt_mask;
641 
642 	pt_size = pid_tbl_mask + 1;
643 	tsz = pt_size * 2 * sizeof(struct pid_table);
644 	new_pt = kmem_alloc(tsz, KM_SLEEP);
645 	new_pt_mask = pt_size * 2 - 1;
646 
647 	mutex_enter(proc_lock);
648 	if (pt_size != pid_tbl_mask + 1) {
649 		/* Another process beat us to it... */
650 		mutex_exit(proc_lock);
651 		kmem_free(new_pt, tsz);
652 		return;
653 	}
654 
655 	/*
656 	 * Copy entries from old table into new one.
657 	 * If 'pid' is 'odd' we need to place in the upper half,
658 	 * even pid's to the lower half.
659 	 * Free items stay in the low half so we don't have to
660 	 * fixup the reference to them.
661 	 * We stuff free items on the front of the freelist
662 	 * because we can't write to unmodified entries.
663 	 * Processing the table backwards maintains a semblance
664 	 * of issuing pid numbers that increase with time.
665 	 */
666 	i = pt_size - 1;
667 	n_pt = new_pt + i;
668 	for (; ; i--, n_pt--) {
669 		proc = pid_table[i].pt_proc;
670 		pgrp = pid_table[i].pt_pgrp;
671 		if (!P_VALID(proc)) {
672 			/* Up 'use count' so that link is valid */
673 			pid = (P_NEXT(proc) + pt_size) & ~pt_size;
674 			rpid = 0;
675 			proc = P_FREE(pid);
676 			if (pgrp)
677 				pid = pgrp->pg_id;
678 		} else {
679 			pid = pid_table[i].pt_pid;
680 			rpid = pid;
681 		}
682 
683 		/* Save entry in appropriate half of table */
684 		n_pt[pid & pt_size].pt_proc = proc;
685 		n_pt[pid & pt_size].pt_pgrp = pgrp;
686 		n_pt[pid & pt_size].pt_pid = rpid;
687 
688 		/* Put other piece on start of free list */
689 		pid = (pid ^ pt_size) & ~pid_tbl_mask;
690 		n_pt[pid & pt_size].pt_proc =
691 			P_FREE((pid & ~pt_size) | next_free_pt);
692 		n_pt[pid & pt_size].pt_pgrp = 0;
693 		n_pt[pid & pt_size].pt_pid = 0;
694 
695 		next_free_pt = i | (pid & pt_size);
696 		if (i == 0)
697 			break;
698 	}
699 
700 	/* Save old table size and switch tables */
701 	tsz = pt_size * sizeof(struct pid_table);
702 	n_pt = pid_table;
703 	pid_table = new_pt;
704 	pid_tbl_mask = new_pt_mask;
705 
706 	/*
707 	 * pid_max starts as PID_MAX (= 30000), once we have 16384
708 	 * allocated pids we need it to be larger!
709 	 */
710 	if (pid_tbl_mask > PID_MAX) {
711 		pid_max = pid_tbl_mask * 2 + 1;
712 		pid_alloc_lim |= pid_alloc_lim << 1;
713 	} else
714 		pid_alloc_lim <<= 1;	/* doubles number of free slots... */
715 
716 	mutex_exit(proc_lock);
717 	kmem_free(n_pt, tsz);
718 }
719 
720 struct proc *
721 proc_alloc(void)
722 {
723 	struct proc *p;
724 
725 	p = pool_cache_get(proc_cache, PR_WAITOK);
726 	p->p_stat = SIDL;			/* protect against others */
727 	proc_initspecific(p);
728 	kdtrace_proc_ctor(NULL, p);
729 	p->p_pid = -1;
730 	proc_alloc_pid(p);
731 	return p;
732 }
733 
734 pid_t
735 proc_alloc_pid(struct proc *p)
736 {
737 	struct pid_table *pt;
738 	pid_t pid;
739 	int nxt;
740 
741 	for (;;expand_pid_table()) {
742 		if (__predict_false(pid_alloc_cnt >= pid_alloc_lim))
743 			/* ensure pids cycle through 2000+ values */
744 			continue;
745 		mutex_enter(proc_lock);
746 		pt = &pid_table[next_free_pt];
747 #ifdef DIAGNOSTIC
748 		if (__predict_false(P_VALID(pt->pt_proc) || pt->pt_pgrp))
749 			panic("proc_alloc: slot busy");
750 #endif
751 		nxt = P_NEXT(pt->pt_proc);
752 		if (nxt & pid_tbl_mask)
753 			break;
754 		/* Table full - expand (NB last entry not used....) */
755 		mutex_exit(proc_lock);
756 	}
757 
758 	/* pid is 'saved use count' + 'size' + entry */
759 	pid = (nxt & ~pid_tbl_mask) + pid_tbl_mask + 1 + next_free_pt;
760 	if ((uint)pid > (uint)pid_max)
761 		pid &= pid_tbl_mask;
762 	next_free_pt = nxt & pid_tbl_mask;
763 
764 	/* Grab table slot */
765 	pt->pt_proc = p;
766 
767 	KASSERT(pt->pt_pid == 0);
768 	pt->pt_pid = pid;
769 	if (p->p_pid == -1) {
770 		p->p_pid = pid;
771 	}
772 	pid_alloc_cnt++;
773 	mutex_exit(proc_lock);
774 
775 	return pid;
776 }
777 
778 /*
779  * Free a process id - called from proc_free (in kern_exit.c)
780  *
781  * Called with the proc_lock held.
782  */
783 void
784 proc_free_pid(pid_t pid)
785 {
786 	struct pid_table *pt;
787 
788 	KASSERT(mutex_owned(proc_lock));
789 
790 	pt = &pid_table[pid & pid_tbl_mask];
791 
792 	/* save pid use count in slot */
793 	pt->pt_proc = P_FREE(pid & ~pid_tbl_mask);
794 	KASSERT(pt->pt_pid == pid);
795 	pt->pt_pid = 0;
796 
797 	if (pt->pt_pgrp == NULL) {
798 		/* link last freed entry onto ours */
799 		pid &= pid_tbl_mask;
800 		pt = &pid_table[last_free_pt];
801 		pt->pt_proc = P_FREE(P_NEXT(pt->pt_proc) | pid);
802 		pt->pt_pid = 0;
803 		last_free_pt = pid;
804 		pid_alloc_cnt--;
805 	}
806 
807 	atomic_dec_uint(&nprocs);
808 }
809 
810 void
811 proc_free_mem(struct proc *p)
812 {
813 
814 	kdtrace_proc_dtor(NULL, p);
815 	pool_cache_put(proc_cache, p);
816 }
817 
818 /*
819  * proc_enterpgrp: move p to a new or existing process group (and session).
820  *
821  * If we are creating a new pgrp, the pgid should equal
822  * the calling process' pid.
823  * If is only valid to enter a process group that is in the session
824  * of the process.
825  * Also mksess should only be set if we are creating a process group
826  *
827  * Only called from sys_setsid and sys_setpgid.
828  */
829 int
830 proc_enterpgrp(struct proc *curp, pid_t pid, pid_t pgid, bool mksess)
831 {
832 	struct pgrp *new_pgrp, *pgrp;
833 	struct session *sess;
834 	struct proc *p;
835 	int rval;
836 	pid_t pg_id = NO_PGID;
837 
838 	sess = mksess ? kmem_alloc(sizeof(*sess), KM_SLEEP) : NULL;
839 
840 	/* Allocate data areas we might need before doing any validity checks */
841 	mutex_enter(proc_lock);		/* Because pid_table might change */
842 	if (pid_table[pgid & pid_tbl_mask].pt_pgrp == 0) {
843 		mutex_exit(proc_lock);
844 		new_pgrp = kmem_alloc(sizeof(*new_pgrp), KM_SLEEP);
845 		mutex_enter(proc_lock);
846 	} else
847 		new_pgrp = NULL;
848 	rval = EPERM;	/* most common error (to save typing) */
849 
850 	/* Check pgrp exists or can be created */
851 	pgrp = pid_table[pgid & pid_tbl_mask].pt_pgrp;
852 	if (pgrp != NULL && pgrp->pg_id != pgid)
853 		goto done;
854 
855 	/* Can only set another process under restricted circumstances. */
856 	if (pid != curp->p_pid) {
857 		/* Must exist and be one of our children... */
858 		p = proc_find(pid);
859 		if (p == NULL || !p_inferior(p, curp)) {
860 			rval = ESRCH;
861 			goto done;
862 		}
863 		/* ... in the same session... */
864 		if (sess != NULL || p->p_session != curp->p_session)
865 			goto done;
866 		/* ... existing pgid must be in same session ... */
867 		if (pgrp != NULL && pgrp->pg_session != p->p_session)
868 			goto done;
869 		/* ... and not done an exec. */
870 		if (p->p_flag & PK_EXEC) {
871 			rval = EACCES;
872 			goto done;
873 		}
874 	} else {
875 		/* ... setsid() cannot re-enter a pgrp */
876 		if (mksess && (curp->p_pgid == curp->p_pid ||
877 		    pgrp_find(curp->p_pid)))
878 			goto done;
879 		p = curp;
880 	}
881 
882 	/* Changing the process group/session of a session
883 	   leader is definitely off limits. */
884 	if (SESS_LEADER(p)) {
885 		if (sess == NULL && p->p_pgrp == pgrp)
886 			/* unless it's a definite noop */
887 			rval = 0;
888 		goto done;
889 	}
890 
891 	/* Can only create a process group with id of process */
892 	if (pgrp == NULL && pgid != pid)
893 		goto done;
894 
895 	/* Can only create a session if creating pgrp */
896 	if (sess != NULL && pgrp != NULL)
897 		goto done;
898 
899 	/* Check we allocated memory for a pgrp... */
900 	if (pgrp == NULL && new_pgrp == NULL)
901 		goto done;
902 
903 	/* Don't attach to 'zombie' pgrp */
904 	if (pgrp != NULL && LIST_EMPTY(&pgrp->pg_members))
905 		goto done;
906 
907 	/* Expect to succeed now */
908 	rval = 0;
909 
910 	if (pgrp == p->p_pgrp)
911 		/* nothing to do */
912 		goto done;
913 
914 	/* Ok all setup, link up required structures */
915 
916 	if (pgrp == NULL) {
917 		pgrp = new_pgrp;
918 		new_pgrp = NULL;
919 		if (sess != NULL) {
920 			sess->s_sid = p->p_pid;
921 			sess->s_leader = p;
922 			sess->s_count = 1;
923 			sess->s_ttyvp = NULL;
924 			sess->s_ttyp = NULL;
925 			sess->s_flags = p->p_session->s_flags & ~S_LOGIN_SET;
926 			memcpy(sess->s_login, p->p_session->s_login,
927 			    sizeof(sess->s_login));
928 			p->p_lflag &= ~PL_CONTROLT;
929 		} else {
930 			sess = p->p_pgrp->pg_session;
931 			proc_sesshold(sess);
932 		}
933 		pgrp->pg_session = sess;
934 		sess = NULL;
935 
936 		pgrp->pg_id = pgid;
937 		LIST_INIT(&pgrp->pg_members);
938 #ifdef DIAGNOSTIC
939 		if (__predict_false(pid_table[pgid & pid_tbl_mask].pt_pgrp))
940 			panic("enterpgrp: pgrp table slot in use");
941 		if (__predict_false(mksess && p != curp))
942 			panic("enterpgrp: mksession and p != curproc");
943 #endif
944 		pid_table[pgid & pid_tbl_mask].pt_pgrp = pgrp;
945 		pgrp->pg_jobc = 0;
946 	}
947 
948 	/*
949 	 * Adjust eligibility of affected pgrps to participate in job control.
950 	 * Increment eligibility counts before decrementing, otherwise we
951 	 * could reach 0 spuriously during the first call.
952 	 */
953 	fixjobc(p, pgrp, 1);
954 	fixjobc(p, p->p_pgrp, 0);
955 
956 	/* Interlock with ttread(). */
957 	mutex_spin_enter(&tty_lock);
958 
959 	/* Move process to requested group. */
960 	LIST_REMOVE(p, p_pglist);
961 	if (LIST_EMPTY(&p->p_pgrp->pg_members))
962 		/* defer delete until we've dumped the lock */
963 		pg_id = p->p_pgrp->pg_id;
964 	p->p_pgrp = pgrp;
965 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
966 
967 	/* Done with the swap; we can release the tty mutex. */
968 	mutex_spin_exit(&tty_lock);
969 
970     done:
971 	if (pg_id != NO_PGID) {
972 		/* Releases proc_lock. */
973 		pg_delete(pg_id);
974 	} else {
975 		mutex_exit(proc_lock);
976 	}
977 	if (sess != NULL)
978 		kmem_free(sess, sizeof(*sess));
979 	if (new_pgrp != NULL)
980 		kmem_free(new_pgrp, sizeof(*new_pgrp));
981 #ifdef DEBUG_PGRP
982 	if (__predict_false(rval))
983 		printf("enterpgrp(%d,%d,%d), curproc %d, rval %d\n",
984 			pid, pgid, mksess, curp->p_pid, rval);
985 #endif
986 	return rval;
987 }
988 
989 /*
990  * proc_leavepgrp: remove a process from its process group.
991  *  => must be called with the proc_lock held, which will be released;
992  */
993 void
994 proc_leavepgrp(struct proc *p)
995 {
996 	struct pgrp *pgrp;
997 
998 	KASSERT(mutex_owned(proc_lock));
999 
1000 	/* Interlock with ttread() */
1001 	mutex_spin_enter(&tty_lock);
1002 	pgrp = p->p_pgrp;
1003 	LIST_REMOVE(p, p_pglist);
1004 	p->p_pgrp = NULL;
1005 	mutex_spin_exit(&tty_lock);
1006 
1007 	if (LIST_EMPTY(&pgrp->pg_members)) {
1008 		/* Releases proc_lock. */
1009 		pg_delete(pgrp->pg_id);
1010 	} else {
1011 		mutex_exit(proc_lock);
1012 	}
1013 }
1014 
1015 /*
1016  * pg_remove: remove a process group from the table.
1017  *  => must be called with the proc_lock held;
1018  *  => returns process group to free;
1019  */
1020 static struct pgrp *
1021 pg_remove(pid_t pg_id)
1022 {
1023 	struct pgrp *pgrp;
1024 	struct pid_table *pt;
1025 
1026 	KASSERT(mutex_owned(proc_lock));
1027 
1028 	pt = &pid_table[pg_id & pid_tbl_mask];
1029 	pgrp = pt->pt_pgrp;
1030 
1031 	KASSERT(pgrp != NULL);
1032 	KASSERT(pgrp->pg_id == pg_id);
1033 	KASSERT(LIST_EMPTY(&pgrp->pg_members));
1034 
1035 	pt->pt_pgrp = NULL;
1036 
1037 	if (!P_VALID(pt->pt_proc)) {
1038 		/* Orphaned pgrp, put slot onto free list. */
1039 		KASSERT((P_NEXT(pt->pt_proc) & pid_tbl_mask) == 0);
1040 		pg_id &= pid_tbl_mask;
1041 		pt = &pid_table[last_free_pt];
1042 		pt->pt_proc = P_FREE(P_NEXT(pt->pt_proc) | pg_id);
1043 		KASSERT(pt->pt_pid == 0);
1044 		last_free_pt = pg_id;
1045 		pid_alloc_cnt--;
1046 	}
1047 	return pgrp;
1048 }
1049 
1050 /*
1051  * pg_delete: delete and free a process group.
1052  *  => must be called with the proc_lock held, which will be released.
1053  */
1054 static void
1055 pg_delete(pid_t pg_id)
1056 {
1057 	struct pgrp *pg;
1058 	struct tty *ttyp;
1059 	struct session *ss;
1060 
1061 	KASSERT(mutex_owned(proc_lock));
1062 
1063 	pg = pid_table[pg_id & pid_tbl_mask].pt_pgrp;
1064 	if (pg == NULL || pg->pg_id != pg_id || !LIST_EMPTY(&pg->pg_members)) {
1065 		mutex_exit(proc_lock);
1066 		return;
1067 	}
1068 
1069 	ss = pg->pg_session;
1070 
1071 	/* Remove reference (if any) from tty to this process group */
1072 	mutex_spin_enter(&tty_lock);
1073 	ttyp = ss->s_ttyp;
1074 	if (ttyp != NULL && ttyp->t_pgrp == pg) {
1075 		ttyp->t_pgrp = NULL;
1076 		KASSERT(ttyp->t_session == ss);
1077 	}
1078 	mutex_spin_exit(&tty_lock);
1079 
1080 	/*
1081 	 * The leading process group in a session is freed by proc_sessrele(),
1082 	 * if last reference.  Note: proc_sessrele() releases proc_lock.
1083 	 */
1084 	pg = (ss->s_sid != pg->pg_id) ? pg_remove(pg_id) : NULL;
1085 	proc_sessrele(ss);
1086 
1087 	if (pg != NULL) {
1088 		/* Free it, if was not done by proc_sessrele(). */
1089 		kmem_free(pg, sizeof(struct pgrp));
1090 	}
1091 }
1092 
1093 /*
1094  * Adjust pgrp jobc counters when specified process changes process group.
1095  * We count the number of processes in each process group that "qualify"
1096  * the group for terminal job control (those with a parent in a different
1097  * process group of the same session).  If that count reaches zero, the
1098  * process group becomes orphaned.  Check both the specified process'
1099  * process group and that of its children.
1100  * entering == 0 => p is leaving specified group.
1101  * entering == 1 => p is entering specified group.
1102  *
1103  * Call with proc_lock held.
1104  */
1105 void
1106 fixjobc(struct proc *p, struct pgrp *pgrp, int entering)
1107 {
1108 	struct pgrp *hispgrp;
1109 	struct session *mysession = pgrp->pg_session;
1110 	struct proc *child;
1111 
1112 	KASSERT(mutex_owned(proc_lock));
1113 
1114 	/*
1115 	 * Check p's parent to see whether p qualifies its own process
1116 	 * group; if so, adjust count for p's process group.
1117 	 */
1118 	hispgrp = p->p_pptr->p_pgrp;
1119 	if (hispgrp != pgrp && hispgrp->pg_session == mysession) {
1120 		if (entering) {
1121 			pgrp->pg_jobc++;
1122 			p->p_lflag &= ~PL_ORPHANPG;
1123 		} else if (--pgrp->pg_jobc == 0)
1124 			orphanpg(pgrp);
1125 	}
1126 
1127 	/*
1128 	 * Check this process' children to see whether they qualify
1129 	 * their process groups; if so, adjust counts for children's
1130 	 * process groups.
1131 	 */
1132 	LIST_FOREACH(child, &p->p_children, p_sibling) {
1133 		hispgrp = child->p_pgrp;
1134 		if (hispgrp != pgrp && hispgrp->pg_session == mysession &&
1135 		    !P_ZOMBIE(child)) {
1136 			if (entering) {
1137 				child->p_lflag &= ~PL_ORPHANPG;
1138 				hispgrp->pg_jobc++;
1139 			} else if (--hispgrp->pg_jobc == 0)
1140 				orphanpg(hispgrp);
1141 		}
1142 	}
1143 }
1144 
1145 /*
1146  * A process group has become orphaned;
1147  * if there are any stopped processes in the group,
1148  * hang-up all process in that group.
1149  *
1150  * Call with proc_lock held.
1151  */
1152 static void
1153 orphanpg(struct pgrp *pg)
1154 {
1155 	struct proc *p;
1156 
1157 	KASSERT(mutex_owned(proc_lock));
1158 
1159 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
1160 		if (p->p_stat == SSTOP) {
1161 			p->p_lflag |= PL_ORPHANPG;
1162 			psignal(p, SIGHUP);
1163 			psignal(p, SIGCONT);
1164 		}
1165 	}
1166 }
1167 
1168 #ifdef DDB
1169 #include <ddb/db_output.h>
1170 void pidtbl_dump(void);
1171 void
1172 pidtbl_dump(void)
1173 {
1174 	struct pid_table *pt;
1175 	struct proc *p;
1176 	struct pgrp *pgrp;
1177 	int id;
1178 
1179 	db_printf("pid table %p size %x, next %x, last %x\n",
1180 		pid_table, pid_tbl_mask+1,
1181 		next_free_pt, last_free_pt);
1182 	for (pt = pid_table, id = 0; id <= pid_tbl_mask; id++, pt++) {
1183 		p = pt->pt_proc;
1184 		if (!P_VALID(p) && !pt->pt_pgrp)
1185 			continue;
1186 		db_printf("  id %x: ", id);
1187 		if (P_VALID(p))
1188 			db_printf("slotpid %d proc %p id %d (0x%x) %s\n",
1189 				pt->pt_pid, p, p->p_pid, p->p_pid, p->p_comm);
1190 		else
1191 			db_printf("next %x use %x\n",
1192 				P_NEXT(p) & pid_tbl_mask,
1193 				P_NEXT(p) & ~pid_tbl_mask);
1194 		if ((pgrp = pt->pt_pgrp)) {
1195 			db_printf("\tsession %p, sid %d, count %d, login %s\n",
1196 			    pgrp->pg_session, pgrp->pg_session->s_sid,
1197 			    pgrp->pg_session->s_count,
1198 			    pgrp->pg_session->s_login);
1199 			db_printf("\tpgrp %p, pg_id %d, pg_jobc %d, members %p\n",
1200 			    pgrp, pgrp->pg_id, pgrp->pg_jobc,
1201 			    LIST_FIRST(&pgrp->pg_members));
1202 			LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
1203 				db_printf("\t\tpid %d addr %p pgrp %p %s\n",
1204 				    p->p_pid, p, p->p_pgrp, p->p_comm);
1205 			}
1206 		}
1207 	}
1208 }
1209 #endif /* DDB */
1210 
1211 #ifdef KSTACK_CHECK_MAGIC
1212 
1213 #define	KSTACK_MAGIC	0xdeadbeaf
1214 
1215 /* XXX should be per process basis? */
1216 static int	kstackleftmin = KSTACK_SIZE;
1217 static int	kstackleftthres = KSTACK_SIZE / 8;
1218 
1219 void
1220 kstack_setup_magic(const struct lwp *l)
1221 {
1222 	uint32_t *ip;
1223 	uint32_t const *end;
1224 
1225 	KASSERT(l != NULL);
1226 	KASSERT(l != &lwp0);
1227 
1228 	/*
1229 	 * fill all the stack with magic number
1230 	 * so that later modification on it can be detected.
1231 	 */
1232 	ip = (uint32_t *)KSTACK_LOWEST_ADDR(l);
1233 	end = (uint32_t *)((char *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE);
1234 	for (; ip < end; ip++) {
1235 		*ip = KSTACK_MAGIC;
1236 	}
1237 }
1238 
1239 void
1240 kstack_check_magic(const struct lwp *l)
1241 {
1242 	uint32_t const *ip, *end;
1243 	int stackleft;
1244 
1245 	KASSERT(l != NULL);
1246 
1247 	/* don't check proc0 */ /*XXX*/
1248 	if (l == &lwp0)
1249 		return;
1250 
1251 #ifdef __MACHINE_STACK_GROWS_UP
1252 	/* stack grows upwards (eg. hppa) */
1253 	ip = (uint32_t *)((void *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE);
1254 	end = (uint32_t *)KSTACK_LOWEST_ADDR(l);
1255 	for (ip--; ip >= end; ip--)
1256 		if (*ip != KSTACK_MAGIC)
1257 			break;
1258 
1259 	stackleft = (void *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE - (void *)ip;
1260 #else /* __MACHINE_STACK_GROWS_UP */
1261 	/* stack grows downwards (eg. i386) */
1262 	ip = (uint32_t *)KSTACK_LOWEST_ADDR(l);
1263 	end = (uint32_t *)((char *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE);
1264 	for (; ip < end; ip++)
1265 		if (*ip != KSTACK_MAGIC)
1266 			break;
1267 
1268 	stackleft = ((const char *)ip) - (const char *)KSTACK_LOWEST_ADDR(l);
1269 #endif /* __MACHINE_STACK_GROWS_UP */
1270 
1271 	if (kstackleftmin > stackleft) {
1272 		kstackleftmin = stackleft;
1273 		if (stackleft < kstackleftthres)
1274 			printf("warning: kernel stack left %d bytes"
1275 			    "(pid %u:lid %u)\n", stackleft,
1276 			    (u_int)l->l_proc->p_pid, (u_int)l->l_lid);
1277 	}
1278 
1279 	if (stackleft <= 0) {
1280 		panic("magic on the top of kernel stack changed for "
1281 		    "pid %u, lid %u: maybe kernel stack overflow",
1282 		    (u_int)l->l_proc->p_pid, (u_int)l->l_lid);
1283 	}
1284 }
1285 #endif /* KSTACK_CHECK_MAGIC */
1286 
1287 int
1288 proclist_foreach_call(struct proclist *list,
1289     int (*callback)(struct proc *, void *arg), void *arg)
1290 {
1291 	struct proc marker;
1292 	struct proc *p;
1293 	int ret = 0;
1294 
1295 	marker.p_flag = PK_MARKER;
1296 	mutex_enter(proc_lock);
1297 	for (p = LIST_FIRST(list); ret == 0 && p != NULL;) {
1298 		if (p->p_flag & PK_MARKER) {
1299 			p = LIST_NEXT(p, p_list);
1300 			continue;
1301 		}
1302 		LIST_INSERT_AFTER(p, &marker, p_list);
1303 		ret = (*callback)(p, arg);
1304 		KASSERT(mutex_owned(proc_lock));
1305 		p = LIST_NEXT(&marker, p_list);
1306 		LIST_REMOVE(&marker, p_list);
1307 	}
1308 	mutex_exit(proc_lock);
1309 
1310 	return ret;
1311 }
1312 
1313 int
1314 proc_vmspace_getref(struct proc *p, struct vmspace **vm)
1315 {
1316 
1317 	/* XXXCDC: how should locking work here? */
1318 
1319 	/* curproc exception is for coredump. */
1320 
1321 	if ((p != curproc && (p->p_sflag & PS_WEXIT) != 0) ||
1322 	    (p->p_vmspace->vm_refcnt < 1)) { /* XXX */
1323 		return EFAULT;
1324 	}
1325 
1326 	uvmspace_addref(p->p_vmspace);
1327 	*vm = p->p_vmspace;
1328 
1329 	return 0;
1330 }
1331 
1332 /*
1333  * Acquire a write lock on the process credential.
1334  */
1335 void
1336 proc_crmod_enter(void)
1337 {
1338 	struct lwp *l = curlwp;
1339 	struct proc *p = l->l_proc;
1340 	struct plimit *lim;
1341 	kauth_cred_t oc;
1342 	char *cn;
1343 
1344 	/* Reset what needs to be reset in plimit. */
1345 	if (p->p_limit->pl_corename != defcorename) {
1346 		lim_privatise(p, false);
1347 		lim = p->p_limit;
1348 		mutex_enter(&lim->pl_lock);
1349 		cn = lim->pl_corename;
1350 		lim->pl_corename = defcorename;
1351 		mutex_exit(&lim->pl_lock);
1352 		if (cn != defcorename)
1353 			free(cn, M_TEMP);
1354 	}
1355 
1356 	mutex_enter(p->p_lock);
1357 
1358 	/* Ensure the LWP cached credentials are up to date. */
1359 	if ((oc = l->l_cred) != p->p_cred) {
1360 		kauth_cred_hold(p->p_cred);
1361 		l->l_cred = p->p_cred;
1362 		kauth_cred_free(oc);
1363 	}
1364 
1365 }
1366 
1367 /*
1368  * Set in a new process credential, and drop the write lock.  The credential
1369  * must have a reference already.  Optionally, free a no-longer required
1370  * credential.  The scheduler also needs to inspect p_cred, so we also
1371  * briefly acquire the sched state mutex.
1372  */
1373 void
1374 proc_crmod_leave(kauth_cred_t scred, kauth_cred_t fcred, bool sugid)
1375 {
1376 	struct lwp *l = curlwp, *l2;
1377 	struct proc *p = l->l_proc;
1378 	kauth_cred_t oc;
1379 
1380 	KASSERT(mutex_owned(p->p_lock));
1381 
1382 	/* Is there a new credential to set in? */
1383 	if (scred != NULL) {
1384 		p->p_cred = scred;
1385 		LIST_FOREACH(l2, &p->p_lwps, l_sibling) {
1386 			if (l2 != l)
1387 				l2->l_prflag |= LPR_CRMOD;
1388 		}
1389 
1390 		/* Ensure the LWP cached credentials are up to date. */
1391 		if ((oc = l->l_cred) != scred) {
1392 			kauth_cred_hold(scred);
1393 			l->l_cred = scred;
1394 		}
1395 	} else
1396 		oc = NULL;	/* XXXgcc */
1397 
1398 	if (sugid) {
1399 		/*
1400 		 * Mark process as having changed credentials, stops
1401 		 * tracing etc.
1402 		 */
1403 		p->p_flag |= PK_SUGID;
1404 	}
1405 
1406 	mutex_exit(p->p_lock);
1407 
1408 	/* If there is a credential to be released, free it now. */
1409 	if (fcred != NULL) {
1410 		KASSERT(scred != NULL);
1411 		kauth_cred_free(fcred);
1412 		if (oc != scred)
1413 			kauth_cred_free(oc);
1414 	}
1415 }
1416 
1417 /*
1418  * proc_specific_key_create --
1419  *	Create a key for subsystem proc-specific data.
1420  */
1421 int
1422 proc_specific_key_create(specificdata_key_t *keyp, specificdata_dtor_t dtor)
1423 {
1424 
1425 	return (specificdata_key_create(proc_specificdata_domain, keyp, dtor));
1426 }
1427 
1428 /*
1429  * proc_specific_key_delete --
1430  *	Delete a key for subsystem proc-specific data.
1431  */
1432 void
1433 proc_specific_key_delete(specificdata_key_t key)
1434 {
1435 
1436 	specificdata_key_delete(proc_specificdata_domain, key);
1437 }
1438 
1439 /*
1440  * proc_initspecific --
1441  *	Initialize a proc's specificdata container.
1442  */
1443 void
1444 proc_initspecific(struct proc *p)
1445 {
1446 	int error;
1447 
1448 	error = specificdata_init(proc_specificdata_domain, &p->p_specdataref);
1449 	KASSERT(error == 0);
1450 }
1451 
1452 /*
1453  * proc_finispecific --
1454  *	Finalize a proc's specificdata container.
1455  */
1456 void
1457 proc_finispecific(struct proc *p)
1458 {
1459 
1460 	specificdata_fini(proc_specificdata_domain, &p->p_specdataref);
1461 }
1462 
1463 /*
1464  * proc_getspecific --
1465  *	Return proc-specific data corresponding to the specified key.
1466  */
1467 void *
1468 proc_getspecific(struct proc *p, specificdata_key_t key)
1469 {
1470 
1471 	return (specificdata_getspecific(proc_specificdata_domain,
1472 					 &p->p_specdataref, key));
1473 }
1474 
1475 /*
1476  * proc_setspecific --
1477  *	Set proc-specific data corresponding to the specified key.
1478  */
1479 void
1480 proc_setspecific(struct proc *p, specificdata_key_t key, void *data)
1481 {
1482 
1483 	specificdata_setspecific(proc_specificdata_domain,
1484 				 &p->p_specdataref, key, data);
1485 }
1486 
1487 int
1488 proc_uidmatch(kauth_cred_t cred, kauth_cred_t target)
1489 {
1490 	int r = 0;
1491 
1492 	if (kauth_cred_getuid(cred) != kauth_cred_getuid(target) ||
1493 	    kauth_cred_getuid(cred) != kauth_cred_getsvuid(target)) {
1494 		/*
1495 		 * suid proc of ours or proc not ours
1496 		 */
1497 		r = EPERM;
1498 	} else if (kauth_cred_getgid(target) != kauth_cred_getsvgid(target)) {
1499 		/*
1500 		 * sgid proc has sgid back to us temporarily
1501 		 */
1502 		r = EPERM;
1503 	} else {
1504 		/*
1505 		 * our rgid must be in target's group list (ie,
1506 		 * sub-processes started by a sgid process)
1507 		 */
1508 		int ismember = 0;
1509 
1510 		if (kauth_cred_ismember_gid(cred,
1511 		    kauth_cred_getgid(target), &ismember) != 0 ||
1512 		    !ismember)
1513 			r = EPERM;
1514 	}
1515 
1516 	return (r);
1517 }
1518 
1519 /*
1520  * sysctl stuff
1521  */
1522 
1523 #define KERN_PROCSLOP	(5 * sizeof(struct kinfo_proc))
1524 
1525 static const u_int sysctl_flagmap[] = {
1526 	PK_ADVLOCK, P_ADVLOCK,
1527 	PK_EXEC, P_EXEC,
1528 	PK_NOCLDWAIT, P_NOCLDWAIT,
1529 	PK_32, P_32,
1530 	PK_CLDSIGIGN, P_CLDSIGIGN,
1531 	PK_SUGID, P_SUGID,
1532 	0
1533 };
1534 
1535 static const u_int sysctl_sflagmap[] = {
1536 	PS_NOCLDSTOP, P_NOCLDSTOP,
1537 	PS_WEXIT, P_WEXIT,
1538 	PS_STOPFORK, P_STOPFORK,
1539 	PS_STOPEXEC, P_STOPEXEC,
1540 	PS_STOPEXIT, P_STOPEXIT,
1541 	0
1542 };
1543 
1544 static const u_int sysctl_slflagmap[] = {
1545 	PSL_TRACED, P_TRACED,
1546 	PSL_FSTRACE, P_FSTRACE,
1547 	PSL_CHTRACED, P_CHTRACED,
1548 	PSL_SYSCALL, P_SYSCALL,
1549 	0
1550 };
1551 
1552 static const u_int sysctl_lflagmap[] = {
1553 	PL_CONTROLT, P_CONTROLT,
1554 	PL_PPWAIT, P_PPWAIT,
1555 	0
1556 };
1557 
1558 static const u_int sysctl_stflagmap[] = {
1559 	PST_PROFIL, P_PROFIL,
1560 	0
1561 
1562 };
1563 
1564 /* used by kern_lwp also */
1565 const u_int sysctl_lwpflagmap[] = {
1566 	LW_SINTR, L_SINTR,
1567 	LW_SYSTEM, L_SYSTEM,
1568 	LW_SA, L_SA,	/* WRS ??? */
1569 	0
1570 };
1571 
1572 /*
1573  * Find the most ``active'' lwp of a process and return it for ps display
1574  * purposes
1575  */
1576 static struct lwp *
1577 proc_active_lwp(struct proc *p)
1578 {
1579 	static const int ostat[] = {
1580 		0,
1581 		2,	/* LSIDL */
1582 		6,	/* LSRUN */
1583 		5,	/* LSSLEEP */
1584 		4,	/* LSSTOP */
1585 		0,	/* LSZOMB */
1586 		1,	/* LSDEAD */
1587 		7,	/* LSONPROC */
1588 		3	/* LSSUSPENDED */
1589 	};
1590 
1591 	struct lwp *l, *lp = NULL;
1592 	LIST_FOREACH(l, &p->p_lwps, l_sibling) {
1593 		KASSERT(l->l_stat >= 0 && l->l_stat < __arraycount(ostat));
1594 		if (lp == NULL ||
1595 		    ostat[l->l_stat] > ostat[lp->l_stat] ||
1596 		    (ostat[l->l_stat] == ostat[lp->l_stat] &&
1597 		    l->l_cpticks > lp->l_cpticks)) {
1598 			lp = l;
1599 			continue;
1600 		}
1601 	}
1602 	return lp;
1603 }
1604 
1605 static int
1606 sysctl_doeproc(SYSCTLFN_ARGS)
1607 {
1608 	union {
1609 		struct kinfo_proc kproc;
1610 		struct kinfo_proc2 kproc2;
1611 	} *kbuf;
1612 	struct proc *p, *next, *marker;
1613 	char *where, *dp;
1614 	int type, op, arg, error;
1615 	u_int elem_size, kelem_size, elem_count;
1616 	size_t buflen, needed;
1617 	bool match, zombie, mmmbrains;
1618 
1619 	if (namelen == 1 && name[0] == CTL_QUERY)
1620 		return (sysctl_query(SYSCTLFN_CALL(rnode)));
1621 
1622 	dp = where = oldp;
1623 	buflen = where != NULL ? *oldlenp : 0;
1624 	error = 0;
1625 	needed = 0;
1626 	type = rnode->sysctl_num;
1627 
1628 	if (type == KERN_PROC) {
1629 		if (namelen != 2 && !(namelen == 1 && name[0] == KERN_PROC_ALL))
1630 			return (EINVAL);
1631 		op = name[0];
1632 		if (op != KERN_PROC_ALL)
1633 			arg = name[1];
1634 		else
1635 			arg = 0;		/* Quell compiler warning */
1636 		elem_count = 0;	/* Ditto */
1637 		kelem_size = elem_size = sizeof(kbuf->kproc);
1638 	} else {
1639 		if (namelen != 4)
1640 			return (EINVAL);
1641 		op = name[0];
1642 		arg = name[1];
1643 		elem_size = name[2];
1644 		elem_count = name[3];
1645 		kelem_size = sizeof(kbuf->kproc2);
1646 	}
1647 
1648 	sysctl_unlock();
1649 
1650 	kbuf = kmem_alloc(sizeof(*kbuf), KM_SLEEP);
1651 	marker = kmem_alloc(sizeof(*marker), KM_SLEEP);
1652 	marker->p_flag = PK_MARKER;
1653 
1654 	mutex_enter(proc_lock);
1655 	mmmbrains = false;
1656 	for (p = LIST_FIRST(&allproc);; p = next) {
1657 		if (p == NULL) {
1658 			if (!mmmbrains) {
1659 				p = LIST_FIRST(&zombproc);
1660 				mmmbrains = true;
1661 			}
1662 			if (p == NULL)
1663 				break;
1664 		}
1665 		next = LIST_NEXT(p, p_list);
1666 		if ((p->p_flag & PK_MARKER) != 0)
1667 			continue;
1668 
1669 		/*
1670 		 * Skip embryonic processes.
1671 		 */
1672 		if (p->p_stat == SIDL)
1673 			continue;
1674 
1675 		mutex_enter(p->p_lock);
1676 		error = kauth_authorize_process(l->l_cred,
1677 		    KAUTH_PROCESS_CANSEE, p,
1678 		    KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL);
1679 		if (error != 0) {
1680 			mutex_exit(p->p_lock);
1681 			continue;
1682 		}
1683 
1684 		/*
1685 		 * TODO - make more efficient (see notes below).
1686 		 * do by session.
1687 		 */
1688 		switch (op) {
1689 		case KERN_PROC_PID:
1690 			/* could do this with just a lookup */
1691 			match = (p->p_pid == (pid_t)arg);
1692 			break;
1693 
1694 		case KERN_PROC_PGRP:
1695 			/* could do this by traversing pgrp */
1696 			match = (p->p_pgrp->pg_id == (pid_t)arg);
1697 			break;
1698 
1699 		case KERN_PROC_SESSION:
1700 			match = (p->p_session->s_sid == (pid_t)arg);
1701 			break;
1702 
1703 		case KERN_PROC_TTY:
1704 			match = true;
1705 			if (arg == (int) KERN_PROC_TTY_REVOKE) {
1706 				if ((p->p_lflag & PL_CONTROLT) == 0 ||
1707 				    p->p_session->s_ttyp == NULL ||
1708 				    p->p_session->s_ttyvp != NULL) {
1709 				    	match = false;
1710 				}
1711 			} else if ((p->p_lflag & PL_CONTROLT) == 0 ||
1712 			    p->p_session->s_ttyp == NULL) {
1713 				if ((dev_t)arg != KERN_PROC_TTY_NODEV) {
1714 					match = false;
1715 				}
1716 			} else if (p->p_session->s_ttyp->t_dev != (dev_t)arg) {
1717 				match = false;
1718 			}
1719 			break;
1720 
1721 		case KERN_PROC_UID:
1722 			match = (kauth_cred_geteuid(p->p_cred) == (uid_t)arg);
1723 			break;
1724 
1725 		case KERN_PROC_RUID:
1726 			match = (kauth_cred_getuid(p->p_cred) == (uid_t)arg);
1727 			break;
1728 
1729 		case KERN_PROC_GID:
1730 			match = (kauth_cred_getegid(p->p_cred) == (uid_t)arg);
1731 			break;
1732 
1733 		case KERN_PROC_RGID:
1734 			match = (kauth_cred_getgid(p->p_cred) == (uid_t)arg);
1735 			break;
1736 
1737 		case KERN_PROC_ALL:
1738 			match = true;
1739 			/* allow everything */
1740 			break;
1741 
1742 		default:
1743 			error = EINVAL;
1744 			mutex_exit(p->p_lock);
1745 			goto cleanup;
1746 		}
1747 		if (!match) {
1748 			mutex_exit(p->p_lock);
1749 			continue;
1750 		}
1751 
1752 		/*
1753 		 * Grab a hold on the process.
1754 		 */
1755 		if (mmmbrains) {
1756 			zombie = true;
1757 		} else {
1758 			zombie = !rw_tryenter(&p->p_reflock, RW_READER);
1759 		}
1760 		if (zombie) {
1761 			LIST_INSERT_AFTER(p, marker, p_list);
1762 		}
1763 
1764 		if (buflen >= elem_size &&
1765 		    (type == KERN_PROC || elem_count > 0)) {
1766 			if (type == KERN_PROC) {
1767 				kbuf->kproc.kp_proc = *p;
1768 				fill_eproc(p, &kbuf->kproc.kp_eproc, zombie);
1769 			} else {
1770 				fill_kproc2(p, &kbuf->kproc2, zombie);
1771 				elem_count--;
1772 			}
1773 			mutex_exit(p->p_lock);
1774 			mutex_exit(proc_lock);
1775 			/*
1776 			 * Copy out elem_size, but not larger than kelem_size
1777 			 */
1778 			error = sysctl_copyout(l, kbuf, dp,
1779 			    min(kelem_size, elem_size));
1780 			mutex_enter(proc_lock);
1781 			if (error) {
1782 				goto bah;
1783 			}
1784 			dp += elem_size;
1785 			buflen -= elem_size;
1786 		} else {
1787 			mutex_exit(p->p_lock);
1788 		}
1789 		needed += elem_size;
1790 
1791 		/*
1792 		 * Release reference to process.
1793 		 */
1794 	 	if (zombie) {
1795 			next = LIST_NEXT(marker, p_list);
1796  			LIST_REMOVE(marker, p_list);
1797 		} else {
1798 			rw_exit(&p->p_reflock);
1799 			next = LIST_NEXT(p, p_list);
1800 		}
1801 	}
1802 	mutex_exit(proc_lock);
1803 
1804 	if (where != NULL) {
1805 		*oldlenp = dp - where;
1806 		if (needed > *oldlenp) {
1807 			error = ENOMEM;
1808 			goto out;
1809 		}
1810 	} else {
1811 		needed += KERN_PROCSLOP;
1812 		*oldlenp = needed;
1813 	}
1814 	if (kbuf)
1815 		kmem_free(kbuf, sizeof(*kbuf));
1816 	if (marker)
1817 		kmem_free(marker, sizeof(*marker));
1818 	sysctl_relock();
1819 	return 0;
1820  bah:
1821  	if (zombie)
1822  		LIST_REMOVE(marker, p_list);
1823 	else
1824 		rw_exit(&p->p_reflock);
1825  cleanup:
1826 	mutex_exit(proc_lock);
1827  out:
1828 	if (kbuf)
1829 		kmem_free(kbuf, sizeof(*kbuf));
1830 	if (marker)
1831 		kmem_free(marker, sizeof(*marker));
1832 	sysctl_relock();
1833 	return error;
1834 }
1835 
1836 /*
1837  * sysctl helper routine for kern.proc_args pseudo-subtree.
1838  */
1839 static int
1840 sysctl_kern_proc_args(SYSCTLFN_ARGS)
1841 {
1842 	struct ps_strings pss;
1843 	struct proc *p;
1844 	size_t len, i;
1845 	struct uio auio;
1846 	struct iovec aiov;
1847 	pid_t pid;
1848 	int nargv, type, error, argvlen;
1849 	char *arg;
1850 	char **argv = NULL;
1851 	char *tmp;
1852 	struct vmspace *vmspace;
1853 	vaddr_t psstr_addr;
1854 	vaddr_t offsetn;
1855 	vaddr_t offsetv;
1856 
1857 	if (namelen == 1 && name[0] == CTL_QUERY)
1858 		return (sysctl_query(SYSCTLFN_CALL(rnode)));
1859 
1860 	if (newp != NULL || namelen != 2)
1861 		return (EINVAL);
1862 	pid = name[0];
1863 	type = name[1];
1864 	argv = NULL;
1865 	argvlen = 0;
1866 
1867 	switch (type) {
1868 	case KERN_PROC_ARGV:
1869 	case KERN_PROC_NARGV:
1870 	case KERN_PROC_ENV:
1871 	case KERN_PROC_NENV:
1872 		/* ok */
1873 		break;
1874 	default:
1875 		return (EINVAL);
1876 	}
1877 
1878 	sysctl_unlock();
1879 
1880 	/* check pid */
1881 	mutex_enter(proc_lock);
1882 	if ((p = proc_find(pid)) == NULL) {
1883 		error = EINVAL;
1884 		goto out_locked;
1885 	}
1886 	mutex_enter(p->p_lock);
1887 
1888 	/* Check permission. */
1889 	if (type == KERN_PROC_ARGV || type == KERN_PROC_NARGV)
1890 		error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE,
1891 		    p, KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ARGS), NULL, NULL);
1892 	else if (type == KERN_PROC_ENV || type == KERN_PROC_NENV)
1893 		error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE,
1894 		    p, KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENV), NULL, NULL);
1895 	else
1896 		error = EINVAL; /* XXXGCC */
1897 	if (error) {
1898 		mutex_exit(p->p_lock);
1899 		goto out_locked;
1900 	}
1901 
1902 	if (oldp == NULL) {
1903 		if (type == KERN_PROC_NARGV || type == KERN_PROC_NENV)
1904 			*oldlenp = sizeof (int);
1905 		else
1906 			*oldlenp = ARG_MAX;	/* XXX XXX XXX */
1907 		error = 0;
1908 		mutex_exit(p->p_lock);
1909 		goto out_locked;
1910 	}
1911 
1912 	/*
1913 	 * Zombies don't have a stack, so we can't read their psstrings.
1914 	 * System processes also don't have a user stack.
1915 	 */
1916 	if (P_ZOMBIE(p) || (p->p_flag & PK_SYSTEM) != 0) {
1917 		error = EINVAL;
1918 		mutex_exit(p->p_lock);
1919 		goto out_locked;
1920 	}
1921 
1922 	/*
1923 	 * Lock the process down in memory.
1924 	 */
1925 	psstr_addr = (vaddr_t)p->p_psstr;
1926 	if (type == KERN_PROC_ARGV || type == KERN_PROC_NARGV) {
1927 		offsetn = p->p_psnargv;
1928 		offsetv = p->p_psargv;
1929 	} else {
1930 		offsetn = p->p_psnenv;
1931 		offsetv = p->p_psenv;
1932 	}
1933 	vmspace = p->p_vmspace;
1934 	uvmspace_addref(vmspace);
1935 	mutex_exit(p->p_lock);
1936 	mutex_exit(proc_lock);
1937 
1938 	/*
1939 	 * Allocate a temporary buffer to hold the arguments.
1940 	 */
1941 	arg = kmem_alloc(PAGE_SIZE, KM_SLEEP);
1942 
1943 	/*
1944 	 * Read in the ps_strings structure.
1945 	 */
1946 	aiov.iov_base = &pss;
1947 	aiov.iov_len = sizeof(pss);
1948 	auio.uio_iov = &aiov;
1949 	auio.uio_iovcnt = 1;
1950 	auio.uio_offset = psstr_addr;
1951 	auio.uio_resid = sizeof(pss);
1952 	auio.uio_rw = UIO_READ;
1953 	UIO_SETUP_SYSSPACE(&auio);
1954 	error = uvm_io(&vmspace->vm_map, &auio);
1955 	if (error)
1956 		goto done;
1957 
1958 	memcpy(&nargv, (char *)&pss + offsetn, sizeof(nargv));
1959 	if (type == KERN_PROC_NARGV || type == KERN_PROC_NENV) {
1960 		error = sysctl_copyout(l, &nargv, oldp, sizeof(nargv));
1961 		*oldlenp = sizeof(nargv);
1962 		goto done;
1963 	}
1964 	/*
1965 	 * Now read the address of the argument vector.
1966 	 */
1967 	switch (type) {
1968 	case KERN_PROC_ARGV:
1969 		/* FALLTHROUGH */
1970 	case KERN_PROC_ENV:
1971 		memcpy(&tmp, (char *)&pss + offsetv, sizeof(tmp));
1972 		break;
1973 	default:
1974 		error = EINVAL;
1975 		goto done;
1976 	}
1977 
1978 #ifdef COMPAT_NETBSD32
1979 	if (p->p_flag & PK_32)
1980 		len = sizeof(netbsd32_charp) * nargv;
1981 	else
1982 #endif
1983 		len = sizeof(char *) * nargv;
1984 
1985 	if ((argvlen = len) != 0)
1986 		argv = kmem_alloc(len, KM_SLEEP);
1987 
1988 	aiov.iov_base = argv;
1989 	aiov.iov_len = len;
1990 	auio.uio_iov = &aiov;
1991 	auio.uio_iovcnt = 1;
1992 	auio.uio_offset = (off_t)(unsigned long)tmp;
1993 	auio.uio_resid = len;
1994 	auio.uio_rw = UIO_READ;
1995 	UIO_SETUP_SYSSPACE(&auio);
1996 	error = uvm_io(&vmspace->vm_map, &auio);
1997 	if (error)
1998 		goto done;
1999 
2000 	/*
2001 	 * Now copy each string.
2002 	 */
2003 	len = 0; /* bytes written to user buffer */
2004 	for (i = 0; i < nargv; i++) {
2005 		int finished = 0;
2006 		vaddr_t base;
2007 		size_t xlen;
2008 		int j;
2009 
2010 #ifdef COMPAT_NETBSD32
2011 		if (p->p_flag & PK_32) {
2012 			netbsd32_charp *argv32;
2013 
2014 			argv32 = (netbsd32_charp *)argv;
2015 			base = (vaddr_t)NETBSD32PTR64(argv32[i]);
2016 		} else
2017 #endif
2018 			base = (vaddr_t)argv[i];
2019 
2020 		/*
2021 		 * The program has messed around with its arguments,
2022 		 * possibly deleting some, and replacing them with
2023 		 * NULL's. Treat this as the last argument and not
2024 		 * a failure.
2025 		 */
2026 		if (base == 0)
2027 			break;
2028 
2029 		while (!finished) {
2030 			xlen = PAGE_SIZE - (base & PAGE_MASK);
2031 
2032 			aiov.iov_base = arg;
2033 			aiov.iov_len = PAGE_SIZE;
2034 			auio.uio_iov = &aiov;
2035 			auio.uio_iovcnt = 1;
2036 			auio.uio_offset = base;
2037 			auio.uio_resid = xlen;
2038 			auio.uio_rw = UIO_READ;
2039 			UIO_SETUP_SYSSPACE(&auio);
2040 			error = uvm_io(&vmspace->vm_map, &auio);
2041 			if (error)
2042 				goto done;
2043 
2044 			/* Look for the end of the string */
2045 			for (j = 0; j < xlen; j++) {
2046 				if (arg[j] == '\0') {
2047 					xlen = j + 1;
2048 					finished = 1;
2049 					break;
2050 				}
2051 			}
2052 
2053 			/* Check for user buffer overflow */
2054 			if (len + xlen > *oldlenp) {
2055 				finished = 1;
2056 				if (len > *oldlenp)
2057 					xlen = 0;
2058 				else
2059 					xlen = *oldlenp - len;
2060 			}
2061 
2062 			/* Copyout the page */
2063 			error = sysctl_copyout(l, arg, (char*)oldp + len, xlen);
2064 			if (error)
2065 				goto done;
2066 
2067 			len += xlen;
2068 			base += xlen;
2069 		}
2070 	}
2071 	*oldlenp = len;
2072 
2073 done:
2074 	if (argvlen != 0)
2075 		kmem_free(argv, argvlen);
2076 	uvmspace_free(vmspace);
2077 	kmem_free(arg, PAGE_SIZE);
2078 	sysctl_relock();
2079 	return error;
2080 
2081 out_locked:
2082 	mutex_exit(proc_lock);
2083 	sysctl_relock();
2084 	return error;
2085 }
2086 
2087 /*
2088  * Fill in an eproc structure for the specified process.
2089  */
2090 void
2091 fill_eproc(struct proc *p, struct eproc *ep, bool zombie)
2092 {
2093 	struct tty *tp;
2094 	struct lwp *l;
2095 
2096 	KASSERT(mutex_owned(proc_lock));
2097 	KASSERT(mutex_owned(p->p_lock));
2098 
2099 	memset(ep, 0, sizeof(*ep));
2100 
2101 	ep->e_paddr = p;
2102 	ep->e_sess = p->p_session;
2103 	if (p->p_cred) {
2104 		kauth_cred_topcred(p->p_cred, &ep->e_pcred);
2105 		kauth_cred_toucred(p->p_cred, &ep->e_ucred);
2106 	}
2107 	if (p->p_stat != SIDL && !P_ZOMBIE(p) && !zombie) {
2108 		struct vmspace *vm = p->p_vmspace;
2109 
2110 		ep->e_vm.vm_rssize = vm_resident_count(vm);
2111 		ep->e_vm.vm_tsize = vm->vm_tsize;
2112 		ep->e_vm.vm_dsize = vm->vm_dsize;
2113 		ep->e_vm.vm_ssize = vm->vm_ssize;
2114 		ep->e_vm.vm_map.size = vm->vm_map.size;
2115 
2116 		/* Pick the primary (first) LWP */
2117 		l = proc_active_lwp(p);
2118 		KASSERT(l != NULL);
2119 		lwp_lock(l);
2120 		if (l->l_wchan)
2121 			strncpy(ep->e_wmesg, l->l_wmesg, WMESGLEN);
2122 		lwp_unlock(l);
2123 	}
2124 	if (p->p_pptr)
2125 		ep->e_ppid = p->p_pptr->p_pid;
2126 	if (p->p_pgrp && p->p_session) {
2127 		ep->e_pgid = p->p_pgrp->pg_id;
2128 		ep->e_jobc = p->p_pgrp->pg_jobc;
2129 		ep->e_sid = p->p_session->s_sid;
2130 		if ((p->p_lflag & PL_CONTROLT) &&
2131 		    (tp = ep->e_sess->s_ttyp)) {
2132 			ep->e_tdev = tp->t_dev;
2133 			ep->e_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PGID;
2134 			ep->e_tsess = tp->t_session;
2135 		} else
2136 			ep->e_tdev = (uint32_t)NODEV;
2137 		ep->e_flag = ep->e_sess->s_ttyvp ? EPROC_CTTY : 0;
2138 		if (SESS_LEADER(p))
2139 			ep->e_flag |= EPROC_SLEADER;
2140 		strncpy(ep->e_login, ep->e_sess->s_login, MAXLOGNAME);
2141 	}
2142 	ep->e_xsize = ep->e_xrssize = 0;
2143 	ep->e_xccount = ep->e_xswrss = 0;
2144 }
2145 
2146 /*
2147  * Fill in a kinfo_proc2 structure for the specified process.
2148  */
2149 static void
2150 fill_kproc2(struct proc *p, struct kinfo_proc2 *ki, bool zombie)
2151 {
2152 	struct tty *tp;
2153 	struct lwp *l, *l2;
2154 	struct timeval ut, st, rt;
2155 	sigset_t ss1, ss2;
2156 	struct rusage ru;
2157 	struct vmspace *vm;
2158 
2159 	KASSERT(mutex_owned(proc_lock));
2160 	KASSERT(mutex_owned(p->p_lock));
2161 
2162 	sigemptyset(&ss1);
2163 	sigemptyset(&ss2);
2164 	memset(ki, 0, sizeof(*ki));
2165 
2166 	ki->p_paddr = PTRTOUINT64(p);
2167 	ki->p_fd = PTRTOUINT64(p->p_fd);
2168 	ki->p_cwdi = PTRTOUINT64(p->p_cwdi);
2169 	ki->p_stats = PTRTOUINT64(p->p_stats);
2170 	ki->p_limit = PTRTOUINT64(p->p_limit);
2171 	ki->p_vmspace = PTRTOUINT64(p->p_vmspace);
2172 	ki->p_sigacts = PTRTOUINT64(p->p_sigacts);
2173 	ki->p_sess = PTRTOUINT64(p->p_session);
2174 	ki->p_tsess = 0;	/* may be changed if controlling tty below */
2175 	ki->p_ru = PTRTOUINT64(&p->p_stats->p_ru);
2176 	ki->p_eflag = 0;
2177 	ki->p_exitsig = p->p_exitsig;
2178 	ki->p_flag = L_INMEM;   /* Process never swapped out */
2179 	ki->p_flag |= sysctl_map_flags(sysctl_flagmap, p->p_flag);
2180 	ki->p_flag |= sysctl_map_flags(sysctl_sflagmap, p->p_sflag);
2181 	ki->p_flag |= sysctl_map_flags(sysctl_slflagmap, p->p_slflag);
2182 	ki->p_flag |= sysctl_map_flags(sysctl_lflagmap, p->p_lflag);
2183 	ki->p_flag |= sysctl_map_flags(sysctl_stflagmap, p->p_stflag);
2184 	ki->p_pid = p->p_pid;
2185 	if (p->p_pptr)
2186 		ki->p_ppid = p->p_pptr->p_pid;
2187 	else
2188 		ki->p_ppid = 0;
2189 	ki->p_uid = kauth_cred_geteuid(p->p_cred);
2190 	ki->p_ruid = kauth_cred_getuid(p->p_cred);
2191 	ki->p_gid = kauth_cred_getegid(p->p_cred);
2192 	ki->p_rgid = kauth_cred_getgid(p->p_cred);
2193 	ki->p_svuid = kauth_cred_getsvuid(p->p_cred);
2194 	ki->p_svgid = kauth_cred_getsvgid(p->p_cred);
2195 	ki->p_ngroups = kauth_cred_ngroups(p->p_cred);
2196 	kauth_cred_getgroups(p->p_cred, ki->p_groups,
2197 	    min(ki->p_ngroups, sizeof(ki->p_groups) / sizeof(ki->p_groups[0])),
2198 	    UIO_SYSSPACE);
2199 
2200 	ki->p_uticks = p->p_uticks;
2201 	ki->p_sticks = p->p_sticks;
2202 	ki->p_iticks = p->p_iticks;
2203 	ki->p_tpgid = NO_PGID;	/* may be changed if controlling tty below */
2204 	ki->p_tracep = PTRTOUINT64(p->p_tracep);
2205 	ki->p_traceflag = p->p_traceflag;
2206 
2207 	memcpy(&ki->p_sigignore, &p->p_sigctx.ps_sigignore,sizeof(ki_sigset_t));
2208 	memcpy(&ki->p_sigcatch, &p->p_sigctx.ps_sigcatch, sizeof(ki_sigset_t));
2209 
2210 	ki->p_cpticks = 0;
2211 	ki->p_pctcpu = p->p_pctcpu;
2212 	ki->p_estcpu = 0;
2213 	ki->p_stat = p->p_stat; /* Will likely be overridden by LWP status */
2214 	ki->p_realstat = p->p_stat;
2215 	ki->p_nice = p->p_nice;
2216 	ki->p_xstat = p->p_xstat;
2217 	ki->p_acflag = p->p_acflag;
2218 
2219 	strncpy(ki->p_comm, p->p_comm,
2220 	    min(sizeof(ki->p_comm), sizeof(p->p_comm)));
2221 	strncpy(ki->p_ename, p->p_emul->e_name, sizeof(ki->p_ename));
2222 
2223 	ki->p_nlwps = p->p_nlwps;
2224 	ki->p_realflag = ki->p_flag;
2225 
2226 	if (p->p_stat != SIDL && !P_ZOMBIE(p) && !zombie) {
2227 		vm = p->p_vmspace;
2228 		ki->p_vm_rssize = vm_resident_count(vm);
2229 		ki->p_vm_tsize = vm->vm_tsize;
2230 		ki->p_vm_dsize = vm->vm_dsize;
2231 		ki->p_vm_ssize = vm->vm_ssize;
2232 		ki->p_vm_vsize = vm->vm_map.size;
2233 		/*
2234 		 * Since the stack is initially mapped mostly with
2235 		 * PROT_NONE and grown as needed, adjust the "mapped size"
2236 		 * to skip the unused stack portion.
2237 		 */
2238 		ki->p_vm_msize =
2239 		    atop(vm->vm_map.size) - vm->vm_issize + vm->vm_ssize;
2240 
2241 		/* Pick the primary (first) LWP */
2242 		l = proc_active_lwp(p);
2243 		KASSERT(l != NULL);
2244 		lwp_lock(l);
2245 		ki->p_nrlwps = p->p_nrlwps;
2246 		ki->p_forw = 0;
2247 		ki->p_back = 0;
2248 		ki->p_addr = PTRTOUINT64(l->l_addr);
2249 		ki->p_stat = l->l_stat;
2250 		ki->p_flag |= sysctl_map_flags(sysctl_lwpflagmap, l->l_flag);
2251 		ki->p_swtime = l->l_swtime;
2252 		ki->p_slptime = l->l_slptime;
2253 		if (l->l_stat == LSONPROC)
2254 			ki->p_schedflags = l->l_cpu->ci_schedstate.spc_flags;
2255 		else
2256 			ki->p_schedflags = 0;
2257 		ki->p_priority = lwp_eprio(l);
2258 		ki->p_usrpri = l->l_priority;
2259 		if (l->l_wchan)
2260 			strncpy(ki->p_wmesg, l->l_wmesg, sizeof(ki->p_wmesg));
2261 		ki->p_wchan = PTRTOUINT64(l->l_wchan);
2262 		ki->p_cpuid = cpu_index(l->l_cpu);
2263 		lwp_unlock(l);
2264 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
2265 			/* This is hardly correct, but... */
2266 			sigplusset(&l->l_sigpend.sp_set, &ss1);
2267 			sigplusset(&l->l_sigmask, &ss2);
2268 			ki->p_cpticks += l->l_cpticks;
2269 			ki->p_pctcpu += l->l_pctcpu;
2270 			ki->p_estcpu += l->l_estcpu;
2271 		}
2272 	}
2273 	sigplusset(&p->p_sigpend.sp_set, &ss2);
2274 	memcpy(&ki->p_siglist, &ss1, sizeof(ki_sigset_t));
2275 	memcpy(&ki->p_sigmask, &ss2, sizeof(ki_sigset_t));
2276 
2277 	if (p->p_session != NULL) {
2278 		ki->p_sid = p->p_session->s_sid;
2279 		ki->p__pgid = p->p_pgrp->pg_id;
2280 		if (p->p_session->s_ttyvp)
2281 			ki->p_eflag |= EPROC_CTTY;
2282 		if (SESS_LEADER(p))
2283 			ki->p_eflag |= EPROC_SLEADER;
2284 		strncpy(ki->p_login, p->p_session->s_login,
2285 		    min(sizeof ki->p_login - 1, sizeof p->p_session->s_login));
2286 		ki->p_jobc = p->p_pgrp->pg_jobc;
2287 		if ((p->p_lflag & PL_CONTROLT) && (tp = p->p_session->s_ttyp)) {
2288 			ki->p_tdev = tp->t_dev;
2289 			ki->p_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PGID;
2290 			ki->p_tsess = PTRTOUINT64(tp->t_session);
2291 		} else {
2292 			ki->p_tdev = (int32_t)NODEV;
2293 		}
2294 	}
2295 
2296 	if (!P_ZOMBIE(p) && !zombie) {
2297 		ki->p_uvalid = 1;
2298 		ki->p_ustart_sec = p->p_stats->p_start.tv_sec;
2299 		ki->p_ustart_usec = p->p_stats->p_start.tv_usec;
2300 
2301 		calcru(p, &ut, &st, NULL, &rt);
2302 		ki->p_rtime_sec = rt.tv_sec;
2303 		ki->p_rtime_usec = rt.tv_usec;
2304 		ki->p_uutime_sec = ut.tv_sec;
2305 		ki->p_uutime_usec = ut.tv_usec;
2306 		ki->p_ustime_sec = st.tv_sec;
2307 		ki->p_ustime_usec = st.tv_usec;
2308 
2309 		memcpy(&ru, &p->p_stats->p_ru, sizeof(ru));
2310 		ki->p_uru_nvcsw = 0;
2311 		ki->p_uru_nivcsw = 0;
2312 		LIST_FOREACH(l2, &p->p_lwps, l_sibling) {
2313 			ki->p_uru_nvcsw += (l2->l_ncsw - l2->l_nivcsw);
2314 			ki->p_uru_nivcsw += l2->l_nivcsw;
2315 			ruadd(&ru, &l2->l_ru);
2316 		}
2317 		ki->p_uru_maxrss = ru.ru_maxrss;
2318 		ki->p_uru_ixrss = ru.ru_ixrss;
2319 		ki->p_uru_idrss = ru.ru_idrss;
2320 		ki->p_uru_isrss = ru.ru_isrss;
2321 		ki->p_uru_minflt = ru.ru_minflt;
2322 		ki->p_uru_majflt = ru.ru_majflt;
2323 		ki->p_uru_nswap = ru.ru_nswap;
2324 		ki->p_uru_inblock = ru.ru_inblock;
2325 		ki->p_uru_oublock = ru.ru_oublock;
2326 		ki->p_uru_msgsnd = ru.ru_msgsnd;
2327 		ki->p_uru_msgrcv = ru.ru_msgrcv;
2328 		ki->p_uru_nsignals = ru.ru_nsignals;
2329 
2330 		timeradd(&p->p_stats->p_cru.ru_utime,
2331 			 &p->p_stats->p_cru.ru_stime, &ut);
2332 		ki->p_uctime_sec = ut.tv_sec;
2333 		ki->p_uctime_usec = ut.tv_usec;
2334 	}
2335 }
2336