xref: /netbsd-src/sys/kern/kern_proc.c (revision 5bbd2a12505d72a8177929a37b5cee489d0a1cfd)
1 /*	$NetBSD: kern_proc.c,v 1.186 2012/06/09 02:31:14 christos 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 1.186 2012/06/09 02:31:14 christos Exp $");
66 
67 #ifdef _KERNEL_OPT
68 #include "opt_kstack.h"
69 #include "opt_maxuprc.h"
70 #include "opt_dtrace.h"
71 #include "opt_compat_netbsd32.h"
72 #endif
73 
74 #include <sys/param.h>
75 #include <sys/systm.h>
76 #include <sys/kernel.h>
77 #include <sys/proc.h>
78 #include <sys/resourcevar.h>
79 #include <sys/buf.h>
80 #include <sys/acct.h>
81 #include <sys/wait.h>
82 #include <sys/file.h>
83 #include <ufs/ufs/quota.h>
84 #include <sys/uio.h>
85 #include <sys/pool.h>
86 #include <sys/pset.h>
87 #include <sys/mbuf.h>
88 #include <sys/ioctl.h>
89 #include <sys/tty.h>
90 #include <sys/signalvar.h>
91 #include <sys/ras.h>
92 #include <sys/filedesc.h>
93 #include <sys/syscall_stats.h>
94 #include <sys/kauth.h>
95 #include <sys/sleepq.h>
96 #include <sys/atomic.h>
97 #include <sys/kmem.h>
98 #include <sys/dtrace_bsd.h>
99 #include <sys/sysctl.h>
100 #include <sys/exec.h>
101 #include <sys/cpu.h>
102 
103 #include <uvm/uvm_extern.h>
104 
105 #ifdef COMPAT_NETBSD32
106 #include <compat/netbsd32/netbsd32.h>
107 #endif
108 
109 /*
110  * Process lists.
111  */
112 
113 struct proclist		allproc		__cacheline_aligned;
114 struct proclist		zombproc	__cacheline_aligned;
115 
116 kmutex_t *		proc_lock	__cacheline_aligned;
117 
118 /*
119  * pid to proc lookup is done by indexing the pid_table array.
120  * Since pid numbers are only allocated when an empty slot
121  * has been found, there is no need to search any lists ever.
122  * (an orphaned pgrp will lock the slot, a session will lock
123  * the pgrp with the same number.)
124  * If the table is too small it is reallocated with twice the
125  * previous size and the entries 'unzipped' into the two halves.
126  * A linked list of free entries is passed through the pt_proc
127  * field of 'free' items - set odd to be an invalid ptr.
128  */
129 
130 struct pid_table {
131 	struct proc	*pt_proc;
132 	struct pgrp	*pt_pgrp;
133 	pid_t		pt_pid;
134 };
135 #if 1	/* strongly typed cast - should be a noop */
136 static inline uint p2u(struct proc *p) { return (uint)(uintptr_t)p; }
137 #else
138 #define p2u(p) ((uint)p)
139 #endif
140 #define P_VALID(p) (!(p2u(p) & 1))
141 #define P_NEXT(p) (p2u(p) >> 1)
142 #define P_FREE(pid) ((struct proc *)(uintptr_t)((pid) << 1 | 1))
143 
144 /*
145  * Table of process IDs (PIDs).
146  */
147 static struct pid_table *pid_table	__read_mostly;
148 
149 #define	INITIAL_PID_TABLE_SIZE		(1 << 5)
150 
151 /* Table mask, threshold for growing and number of allocated PIDs. */
152 static u_int		pid_tbl_mask	__read_mostly;
153 static u_int		pid_alloc_lim	__read_mostly;
154 static u_int		pid_alloc_cnt	__cacheline_aligned;
155 
156 /* Next free, last free and maximum PIDs. */
157 static u_int		next_free_pt	__cacheline_aligned;
158 static u_int		last_free_pt	__cacheline_aligned;
159 static pid_t		pid_max		__read_mostly;
160 
161 /* Components of the first process -- never freed. */
162 
163 extern struct emul emul_netbsd;	/* defined in kern_exec.c */
164 
165 struct session session0 = {
166 	.s_count = 1,
167 	.s_sid = 0,
168 };
169 struct pgrp pgrp0 = {
170 	.pg_members = LIST_HEAD_INITIALIZER(&pgrp0.pg_members),
171 	.pg_session = &session0,
172 };
173 filedesc_t filedesc0;
174 struct cwdinfo cwdi0 = {
175 	.cwdi_cmask = CMASK,		/* see cmask below */
176 	.cwdi_refcnt = 1,
177 };
178 struct plimit limit0;
179 struct pstats pstat0;
180 struct vmspace vmspace0;
181 struct sigacts sigacts0;
182 struct proc proc0 = {
183 	.p_lwps = LIST_HEAD_INITIALIZER(&proc0.p_lwps),
184 	.p_sigwaiters = LIST_HEAD_INITIALIZER(&proc0.p_sigwaiters),
185 	.p_nlwps = 1,
186 	.p_nrlwps = 1,
187 	.p_nlwpid = 1,		/* must match lwp0.l_lid */
188 	.p_pgrp = &pgrp0,
189 	.p_comm = "system",
190 	/*
191 	 * Set P_NOCLDWAIT so that kernel threads are reparented to init(8)
192 	 * when they exit.  init(8) can easily wait them out for us.
193 	 */
194 	.p_flag = PK_SYSTEM | PK_NOCLDWAIT,
195 	.p_stat = SACTIVE,
196 	.p_nice = NZERO,
197 	.p_emul = &emul_netbsd,
198 	.p_cwdi = &cwdi0,
199 	.p_limit = &limit0,
200 	.p_fd = &filedesc0,
201 	.p_vmspace = &vmspace0,
202 	.p_stats = &pstat0,
203 	.p_sigacts = &sigacts0,
204 };
205 kauth_cred_t cred0;
206 
207 static const int	nofile	= NOFILE;
208 static const int	maxuprc	= MAXUPRC;
209 static const int	cmask	= CMASK;
210 
211 static int sysctl_doeproc(SYSCTLFN_PROTO);
212 static int sysctl_kern_proc_args(SYSCTLFN_PROTO);
213 static void fill_kproc2(struct proc *, struct kinfo_proc2 *, bool);
214 
215 /*
216  * The process list descriptors, used during pid allocation and
217  * by sysctl.  No locking on this data structure is needed since
218  * it is completely static.
219  */
220 const struct proclist_desc proclists[] = {
221 	{ &allproc	},
222 	{ &zombproc	},
223 	{ NULL		},
224 };
225 
226 static struct pgrp *	pg_remove(pid_t);
227 static void		pg_delete(pid_t);
228 static void		orphanpg(struct pgrp *);
229 
230 static specificdata_domain_t proc_specificdata_domain;
231 
232 static pool_cache_t proc_cache;
233 
234 static kauth_listener_t proc_listener;
235 
236 static int
237 proc_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
238     void *arg0, void *arg1, void *arg2, void *arg3)
239 {
240 	struct proc *p;
241 	int result;
242 
243 	result = KAUTH_RESULT_DEFER;
244 	p = arg0;
245 
246 	switch (action) {
247 	case KAUTH_PROCESS_CANSEE: {
248 		enum kauth_process_req req;
249 
250 		req = (enum kauth_process_req)arg1;
251 
252 		switch (req) {
253 		case KAUTH_REQ_PROCESS_CANSEE_ARGS:
254 		case KAUTH_REQ_PROCESS_CANSEE_ENTRY:
255 		case KAUTH_REQ_PROCESS_CANSEE_OPENFILES:
256 			result = KAUTH_RESULT_ALLOW;
257 
258 			break;
259 
260 		case KAUTH_REQ_PROCESS_CANSEE_ENV:
261 			if (kauth_cred_getuid(cred) !=
262 			    kauth_cred_getuid(p->p_cred) ||
263 			    kauth_cred_getuid(cred) !=
264 			    kauth_cred_getsvuid(p->p_cred))
265 				break;
266 
267 			result = KAUTH_RESULT_ALLOW;
268 
269 			break;
270 
271 		default:
272 			break;
273 		}
274 
275 		break;
276 		}
277 
278 	case KAUTH_PROCESS_FORK: {
279 		int lnprocs = (int)(unsigned long)arg2;
280 
281 		/*
282 		 * Don't allow a nonprivileged user to use the last few
283 		 * processes. The variable lnprocs is the current number of
284 		 * processes, maxproc is the limit.
285 		 */
286 		if (__predict_false((lnprocs >= maxproc - 5)))
287 			break;
288 
289 		result = KAUTH_RESULT_ALLOW;
290 
291 		break;
292 		}
293 
294 	case KAUTH_PROCESS_CORENAME:
295 	case KAUTH_PROCESS_STOPFLAG:
296 		if (proc_uidmatch(cred, p->p_cred) == 0)
297 			result = KAUTH_RESULT_ALLOW;
298 
299 		break;
300 
301 	default:
302 		break;
303 	}
304 
305 	return result;
306 }
307 
308 /*
309  * Initialize global process hashing structures.
310  */
311 void
312 procinit(void)
313 {
314 	const struct proclist_desc *pd;
315 	u_int i;
316 #define	LINK_EMPTY ((PID_MAX + INITIAL_PID_TABLE_SIZE) & ~(INITIAL_PID_TABLE_SIZE - 1))
317 
318 	for (pd = proclists; pd->pd_list != NULL; pd++)
319 		LIST_INIT(pd->pd_list);
320 
321 	proc_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
322 	pid_table = kmem_alloc(INITIAL_PID_TABLE_SIZE
323 	    * sizeof(struct pid_table), KM_SLEEP);
324 	pid_tbl_mask = INITIAL_PID_TABLE_SIZE - 1;
325 	pid_max = PID_MAX;
326 
327 	/* Set free list running through table...
328 	   Preset 'use count' above PID_MAX so we allocate pid 1 next. */
329 	for (i = 0; i <= pid_tbl_mask; i++) {
330 		pid_table[i].pt_proc = P_FREE(LINK_EMPTY + i + 1);
331 		pid_table[i].pt_pgrp = 0;
332 		pid_table[i].pt_pid = 0;
333 	}
334 	/* slot 0 is just grabbed */
335 	next_free_pt = 1;
336 	/* Need to fix last entry. */
337 	last_free_pt = pid_tbl_mask;
338 	pid_table[last_free_pt].pt_proc = P_FREE(LINK_EMPTY);
339 	/* point at which we grow table - to avoid reusing pids too often */
340 	pid_alloc_lim = pid_tbl_mask - 1;
341 #undef LINK_EMPTY
342 
343 	proc_specificdata_domain = specificdata_domain_create();
344 	KASSERT(proc_specificdata_domain != NULL);
345 
346 	proc_cache = pool_cache_init(sizeof(struct proc), 0, 0, 0,
347 	    "procpl", NULL, IPL_NONE, NULL, NULL, NULL);
348 
349 	proc_listener = kauth_listen_scope(KAUTH_SCOPE_PROCESS,
350 	    proc_listener_cb, NULL);
351 }
352 
353 void
354 procinit_sysctl(void)
355 {
356 	static struct sysctllog *clog;
357 
358 	sysctl_createv(&clog, 0, NULL, NULL,
359 		       CTLFLAG_PERMANENT,
360 		       CTLTYPE_NODE, "kern", NULL,
361 		       NULL, 0, NULL, 0,
362 		       CTL_KERN, CTL_EOL);
363 
364 	sysctl_createv(&clog, 0, NULL, NULL,
365 		       CTLFLAG_PERMANENT,
366 		       CTLTYPE_NODE, "proc",
367 		       SYSCTL_DESCR("System-wide process information"),
368 		       sysctl_doeproc, 0, NULL, 0,
369 		       CTL_KERN, KERN_PROC, CTL_EOL);
370 	sysctl_createv(&clog, 0, NULL, NULL,
371 		       CTLFLAG_PERMANENT,
372 		       CTLTYPE_NODE, "proc2",
373 		       SYSCTL_DESCR("Machine-independent process information"),
374 		       sysctl_doeproc, 0, NULL, 0,
375 		       CTL_KERN, KERN_PROC2, CTL_EOL);
376 	sysctl_createv(&clog, 0, NULL, NULL,
377 		       CTLFLAG_PERMANENT,
378 		       CTLTYPE_NODE, "proc_args",
379 		       SYSCTL_DESCR("Process argument information"),
380 		       sysctl_kern_proc_args, 0, NULL, 0,
381 		       CTL_KERN, KERN_PROC_ARGS, CTL_EOL);
382 
383 	/*
384 	  "nodes" under these:
385 
386 	  KERN_PROC_ALL
387 	  KERN_PROC_PID pid
388 	  KERN_PROC_PGRP pgrp
389 	  KERN_PROC_SESSION sess
390 	  KERN_PROC_TTY tty
391 	  KERN_PROC_UID uid
392 	  KERN_PROC_RUID uid
393 	  KERN_PROC_GID gid
394 	  KERN_PROC_RGID gid
395 
396 	  all in all, probably not worth the effort...
397 	*/
398 }
399 
400 /*
401  * Initialize process 0.
402  */
403 void
404 proc0_init(void)
405 {
406 	struct proc *p;
407 	struct pgrp *pg;
408 	struct rlimit *rlim;
409 	rlim_t lim;
410 	int i;
411 
412 	p = &proc0;
413 	pg = &pgrp0;
414 
415 	mutex_init(&p->p_stmutex, MUTEX_DEFAULT, IPL_HIGH);
416 	mutex_init(&p->p_auxlock, MUTEX_DEFAULT, IPL_NONE);
417 	p->p_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
418 
419 	rw_init(&p->p_reflock);
420 	cv_init(&p->p_waitcv, "wait");
421 	cv_init(&p->p_lwpcv, "lwpwait");
422 
423 	LIST_INSERT_HEAD(&p->p_lwps, &lwp0, l_sibling);
424 
425 	pid_table[0].pt_proc = p;
426 	LIST_INSERT_HEAD(&allproc, p, p_list);
427 
428 	pid_table[0].pt_pgrp = pg;
429 	LIST_INSERT_HEAD(&pg->pg_members, p, p_pglist);
430 
431 #ifdef __HAVE_SYSCALL_INTERN
432 	(*p->p_emul->e_syscall_intern)(p);
433 #endif
434 
435 	/* Create credentials. */
436 	cred0 = kauth_cred_alloc();
437 	p->p_cred = cred0;
438 
439 	/* Create the CWD info. */
440 	rw_init(&cwdi0.cwdi_lock);
441 
442 	/* Create the limits structures. */
443 	mutex_init(&limit0.pl_lock, MUTEX_DEFAULT, IPL_NONE);
444 
445 	rlim = limit0.pl_rlimit;
446 	for (i = 0; i < __arraycount(limit0.pl_rlimit); i++) {
447 		rlim[i].rlim_cur = RLIM_INFINITY;
448 		rlim[i].rlim_max = RLIM_INFINITY;
449 	}
450 
451 	rlim[RLIMIT_NOFILE].rlim_max = maxfiles;
452 	rlim[RLIMIT_NOFILE].rlim_cur = maxfiles < nofile ? maxfiles : nofile;
453 
454 	rlim[RLIMIT_NPROC].rlim_max = maxproc;
455 	rlim[RLIMIT_NPROC].rlim_cur = maxproc < maxuprc ? maxproc : maxuprc;
456 
457 	lim = MIN(VM_MAXUSER_ADDRESS, ctob((rlim_t)uvmexp.free));
458 	rlim[RLIMIT_RSS].rlim_max = lim;
459 	rlim[RLIMIT_MEMLOCK].rlim_max = lim;
460 	rlim[RLIMIT_MEMLOCK].rlim_cur = lim / 3;
461 
462 	rlim[RLIMIT_NTHR].rlim_max = maxlwp;
463 	rlim[RLIMIT_NTHR].rlim_cur = maxlwp < maxuprc ? maxlwp : maxuprc;
464 
465 	/* Note that default core name has zero length. */
466 	limit0.pl_corename = defcorename;
467 	limit0.pl_cnlen = 0;
468 	limit0.pl_refcnt = 1;
469 	limit0.pl_writeable = false;
470 	limit0.pl_sv_limit = NULL;
471 
472 	/* Configure virtual memory system, set vm rlimits. */
473 	uvm_init_limits(p);
474 
475 	/* Initialize file descriptor table for proc0. */
476 	fd_init(&filedesc0);
477 
478 	/*
479 	 * Initialize proc0's vmspace, which uses the kernel pmap.
480 	 * All kernel processes (which never have user space mappings)
481 	 * share proc0's vmspace, and thus, the kernel pmap.
482 	 */
483 	uvmspace_init(&vmspace0, pmap_kernel(), round_page(VM_MIN_ADDRESS),
484 	    trunc_page(VM_MAX_ADDRESS));
485 
486 	/* Initialize signal state for proc0. XXX IPL_SCHED */
487 	mutex_init(&p->p_sigacts->sa_mutex, MUTEX_DEFAULT, IPL_SCHED);
488 	siginit(p);
489 
490 	proc_initspecific(p);
491 	kdtrace_proc_ctor(NULL, p);
492 }
493 
494 /*
495  * Session reference counting.
496  */
497 
498 void
499 proc_sesshold(struct session *ss)
500 {
501 
502 	KASSERT(mutex_owned(proc_lock));
503 	ss->s_count++;
504 }
505 
506 void
507 proc_sessrele(struct session *ss)
508 {
509 
510 	KASSERT(mutex_owned(proc_lock));
511 	/*
512 	 * We keep the pgrp with the same id as the session in order to
513 	 * stop a process being given the same pid.  Since the pgrp holds
514 	 * a reference to the session, it must be a 'zombie' pgrp by now.
515 	 */
516 	if (--ss->s_count == 0) {
517 		struct pgrp *pg;
518 
519 		pg = pg_remove(ss->s_sid);
520 		mutex_exit(proc_lock);
521 
522 		kmem_free(pg, sizeof(struct pgrp));
523 		kmem_free(ss, sizeof(struct session));
524 	} else {
525 		mutex_exit(proc_lock);
526 	}
527 }
528 
529 /*
530  * Check that the specified process group is in the session of the
531  * specified process.
532  * Treats -ve ids as process ids.
533  * Used to validate TIOCSPGRP requests.
534  */
535 int
536 pgid_in_session(struct proc *p, pid_t pg_id)
537 {
538 	struct pgrp *pgrp;
539 	struct session *session;
540 	int error;
541 
542 	mutex_enter(proc_lock);
543 	if (pg_id < 0) {
544 		struct proc *p1 = proc_find(-pg_id);
545 		if (p1 == NULL) {
546 			error = EINVAL;
547 			goto fail;
548 		}
549 		pgrp = p1->p_pgrp;
550 	} else {
551 		pgrp = pgrp_find(pg_id);
552 		if (pgrp == NULL) {
553 			error = EINVAL;
554 			goto fail;
555 		}
556 	}
557 	session = pgrp->pg_session;
558 	error = (session != p->p_pgrp->pg_session) ? EPERM : 0;
559 fail:
560 	mutex_exit(proc_lock);
561 	return error;
562 }
563 
564 /*
565  * p_inferior: is p an inferior of q?
566  */
567 static inline bool
568 p_inferior(struct proc *p, struct proc *q)
569 {
570 
571 	KASSERT(mutex_owned(proc_lock));
572 
573 	for (; p != q; p = p->p_pptr)
574 		if (p->p_pid == 0)
575 			return false;
576 	return true;
577 }
578 
579 /*
580  * proc_find: locate a process by the ID.
581  *
582  * => Must be called with proc_lock held.
583  */
584 proc_t *
585 proc_find_raw(pid_t pid)
586 {
587 	struct pid_table *pt;
588 	proc_t *p;
589 
590 	KASSERT(mutex_owned(proc_lock));
591 	pt = &pid_table[pid & pid_tbl_mask];
592 	p = pt->pt_proc;
593 	if (__predict_false(!P_VALID(p) || pt->pt_pid != pid)) {
594 		return NULL;
595 	}
596 	return p;
597 }
598 
599 proc_t *
600 proc_find(pid_t pid)
601 {
602 	proc_t *p;
603 
604 	p = proc_find_raw(pid);
605 	if (__predict_false(p == NULL)) {
606 		return NULL;
607 	}
608 
609 	/*
610 	 * Only allow live processes to be found by PID.
611 	 * XXX: p_stat might change, since unlocked.
612 	 */
613 	if (__predict_true(p->p_stat == SACTIVE || p->p_stat == SSTOP)) {
614 		return p;
615 	}
616 	return NULL;
617 }
618 
619 /*
620  * pgrp_find: locate a process group by the ID.
621  *
622  * => Must be called with proc_lock held.
623  */
624 struct pgrp *
625 pgrp_find(pid_t pgid)
626 {
627 	struct pgrp *pg;
628 
629 	KASSERT(mutex_owned(proc_lock));
630 
631 	pg = pid_table[pgid & pid_tbl_mask].pt_pgrp;
632 
633 	/*
634 	 * Cannot look up a process group that only exists because the
635 	 * session has not died yet (traditional).
636 	 */
637 	if (pg == NULL || pg->pg_id != pgid || LIST_EMPTY(&pg->pg_members)) {
638 		return NULL;
639 	}
640 	return pg;
641 }
642 
643 static void
644 expand_pid_table(void)
645 {
646 	size_t pt_size, tsz;
647 	struct pid_table *n_pt, *new_pt;
648 	struct proc *proc;
649 	struct pgrp *pgrp;
650 	pid_t pid, rpid;
651 	u_int i;
652 	uint new_pt_mask;
653 
654 	pt_size = pid_tbl_mask + 1;
655 	tsz = pt_size * 2 * sizeof(struct pid_table);
656 	new_pt = kmem_alloc(tsz, KM_SLEEP);
657 	new_pt_mask = pt_size * 2 - 1;
658 
659 	mutex_enter(proc_lock);
660 	if (pt_size != pid_tbl_mask + 1) {
661 		/* Another process beat us to it... */
662 		mutex_exit(proc_lock);
663 		kmem_free(new_pt, tsz);
664 		return;
665 	}
666 
667 	/*
668 	 * Copy entries from old table into new one.
669 	 * If 'pid' is 'odd' we need to place in the upper half,
670 	 * even pid's to the lower half.
671 	 * Free items stay in the low half so we don't have to
672 	 * fixup the reference to them.
673 	 * We stuff free items on the front of the freelist
674 	 * because we can't write to unmodified entries.
675 	 * Processing the table backwards maintains a semblance
676 	 * of issuing pid numbers that increase with time.
677 	 */
678 	i = pt_size - 1;
679 	n_pt = new_pt + i;
680 	for (; ; i--, n_pt--) {
681 		proc = pid_table[i].pt_proc;
682 		pgrp = pid_table[i].pt_pgrp;
683 		if (!P_VALID(proc)) {
684 			/* Up 'use count' so that link is valid */
685 			pid = (P_NEXT(proc) + pt_size) & ~pt_size;
686 			rpid = 0;
687 			proc = P_FREE(pid);
688 			if (pgrp)
689 				pid = pgrp->pg_id;
690 		} else {
691 			pid = pid_table[i].pt_pid;
692 			rpid = pid;
693 		}
694 
695 		/* Save entry in appropriate half of table */
696 		n_pt[pid & pt_size].pt_proc = proc;
697 		n_pt[pid & pt_size].pt_pgrp = pgrp;
698 		n_pt[pid & pt_size].pt_pid = rpid;
699 
700 		/* Put other piece on start of free list */
701 		pid = (pid ^ pt_size) & ~pid_tbl_mask;
702 		n_pt[pid & pt_size].pt_proc =
703 			P_FREE((pid & ~pt_size) | next_free_pt);
704 		n_pt[pid & pt_size].pt_pgrp = 0;
705 		n_pt[pid & pt_size].pt_pid = 0;
706 
707 		next_free_pt = i | (pid & pt_size);
708 		if (i == 0)
709 			break;
710 	}
711 
712 	/* Save old table size and switch tables */
713 	tsz = pt_size * sizeof(struct pid_table);
714 	n_pt = pid_table;
715 	pid_table = new_pt;
716 	pid_tbl_mask = new_pt_mask;
717 
718 	/*
719 	 * pid_max starts as PID_MAX (= 30000), once we have 16384
720 	 * allocated pids we need it to be larger!
721 	 */
722 	if (pid_tbl_mask > PID_MAX) {
723 		pid_max = pid_tbl_mask * 2 + 1;
724 		pid_alloc_lim |= pid_alloc_lim << 1;
725 	} else
726 		pid_alloc_lim <<= 1;	/* doubles number of free slots... */
727 
728 	mutex_exit(proc_lock);
729 	kmem_free(n_pt, tsz);
730 }
731 
732 struct proc *
733 proc_alloc(void)
734 {
735 	struct proc *p;
736 
737 	p = pool_cache_get(proc_cache, PR_WAITOK);
738 	p->p_stat = SIDL;			/* protect against others */
739 	proc_initspecific(p);
740 	kdtrace_proc_ctor(NULL, p);
741 	p->p_pid = -1;
742 	proc_alloc_pid(p);
743 	return p;
744 }
745 
746 /*
747  * proc_alloc_pid: allocate PID and record the given proc 'p' so that
748  * proc_find_raw() can find it by the PID.
749  */
750 
751 pid_t
752 proc_alloc_pid(struct proc *p)
753 {
754 	struct pid_table *pt;
755 	pid_t pid;
756 	int nxt;
757 
758 	for (;;expand_pid_table()) {
759 		if (__predict_false(pid_alloc_cnt >= pid_alloc_lim))
760 			/* ensure pids cycle through 2000+ values */
761 			continue;
762 		mutex_enter(proc_lock);
763 		pt = &pid_table[next_free_pt];
764 #ifdef DIAGNOSTIC
765 		if (__predict_false(P_VALID(pt->pt_proc) || pt->pt_pgrp))
766 			panic("proc_alloc: slot busy");
767 #endif
768 		nxt = P_NEXT(pt->pt_proc);
769 		if (nxt & pid_tbl_mask)
770 			break;
771 		/* Table full - expand (NB last entry not used....) */
772 		mutex_exit(proc_lock);
773 	}
774 
775 	/* pid is 'saved use count' + 'size' + entry */
776 	pid = (nxt & ~pid_tbl_mask) + pid_tbl_mask + 1 + next_free_pt;
777 	if ((uint)pid > (uint)pid_max)
778 		pid &= pid_tbl_mask;
779 	next_free_pt = nxt & pid_tbl_mask;
780 
781 	/* Grab table slot */
782 	pt->pt_proc = p;
783 
784 	KASSERT(pt->pt_pid == 0);
785 	pt->pt_pid = pid;
786 	if (p->p_pid == -1) {
787 		p->p_pid = pid;
788 	}
789 	pid_alloc_cnt++;
790 	mutex_exit(proc_lock);
791 
792 	return pid;
793 }
794 
795 /*
796  * Free a process id - called from proc_free (in kern_exit.c)
797  *
798  * Called with the proc_lock held.
799  */
800 void
801 proc_free_pid(pid_t pid)
802 {
803 	struct pid_table *pt;
804 
805 	KASSERT(mutex_owned(proc_lock));
806 
807 	pt = &pid_table[pid & pid_tbl_mask];
808 
809 	/* save pid use count in slot */
810 	pt->pt_proc = P_FREE(pid & ~pid_tbl_mask);
811 	KASSERT(pt->pt_pid == pid);
812 	pt->pt_pid = 0;
813 
814 	if (pt->pt_pgrp == NULL) {
815 		/* link last freed entry onto ours */
816 		pid &= pid_tbl_mask;
817 		pt = &pid_table[last_free_pt];
818 		pt->pt_proc = P_FREE(P_NEXT(pt->pt_proc) | pid);
819 		pt->pt_pid = 0;
820 		last_free_pt = pid;
821 		pid_alloc_cnt--;
822 	}
823 
824 	atomic_dec_uint(&nprocs);
825 }
826 
827 void
828 proc_free_mem(struct proc *p)
829 {
830 
831 	kdtrace_proc_dtor(NULL, p);
832 	pool_cache_put(proc_cache, p);
833 }
834 
835 /*
836  * proc_enterpgrp: move p to a new or existing process group (and session).
837  *
838  * If we are creating a new pgrp, the pgid should equal
839  * the calling process' pid.
840  * If is only valid to enter a process group that is in the session
841  * of the process.
842  * Also mksess should only be set if we are creating a process group
843  *
844  * Only called from sys_setsid, sys_setpgid and posix_spawn/spawn_return.
845  */
846 int
847 proc_enterpgrp(struct proc *curp, pid_t pid, pid_t pgid, bool mksess)
848 {
849 	struct pgrp *new_pgrp, *pgrp;
850 	struct session *sess;
851 	struct proc *p;
852 	int rval;
853 	pid_t pg_id = NO_PGID;
854 
855 	sess = mksess ? kmem_alloc(sizeof(*sess), KM_SLEEP) : NULL;
856 
857 	/* Allocate data areas we might need before doing any validity checks */
858 	mutex_enter(proc_lock);		/* Because pid_table might change */
859 	if (pid_table[pgid & pid_tbl_mask].pt_pgrp == 0) {
860 		mutex_exit(proc_lock);
861 		new_pgrp = kmem_alloc(sizeof(*new_pgrp), KM_SLEEP);
862 		mutex_enter(proc_lock);
863 	} else
864 		new_pgrp = NULL;
865 	rval = EPERM;	/* most common error (to save typing) */
866 
867 	/* Check pgrp exists or can be created */
868 	pgrp = pid_table[pgid & pid_tbl_mask].pt_pgrp;
869 	if (pgrp != NULL && pgrp->pg_id != pgid)
870 		goto done;
871 
872 	/* Can only set another process under restricted circumstances. */
873 	if (pid != curp->p_pid) {
874 		/* Must exist and be one of our children... */
875 		p = proc_find(pid);
876 		if (p == NULL || !p_inferior(p, curp)) {
877 			rval = ESRCH;
878 			goto done;
879 		}
880 		/* ... in the same session... */
881 		if (sess != NULL || p->p_session != curp->p_session)
882 			goto done;
883 		/* ... existing pgid must be in same session ... */
884 		if (pgrp != NULL && pgrp->pg_session != p->p_session)
885 			goto done;
886 		/* ... and not done an exec. */
887 		if (p->p_flag & PK_EXEC) {
888 			rval = EACCES;
889 			goto done;
890 		}
891 	} else {
892 		/* ... setsid() cannot re-enter a pgrp */
893 		if (mksess && (curp->p_pgid == curp->p_pid ||
894 		    pgrp_find(curp->p_pid)))
895 			goto done;
896 		p = curp;
897 	}
898 
899 	/* Changing the process group/session of a session
900 	   leader is definitely off limits. */
901 	if (SESS_LEADER(p)) {
902 		if (sess == NULL && p->p_pgrp == pgrp)
903 			/* unless it's a definite noop */
904 			rval = 0;
905 		goto done;
906 	}
907 
908 	/* Can only create a process group with id of process */
909 	if (pgrp == NULL && pgid != pid)
910 		goto done;
911 
912 	/* Can only create a session if creating pgrp */
913 	if (sess != NULL && pgrp != NULL)
914 		goto done;
915 
916 	/* Check we allocated memory for a pgrp... */
917 	if (pgrp == NULL && new_pgrp == NULL)
918 		goto done;
919 
920 	/* Don't attach to 'zombie' pgrp */
921 	if (pgrp != NULL && LIST_EMPTY(&pgrp->pg_members))
922 		goto done;
923 
924 	/* Expect to succeed now */
925 	rval = 0;
926 
927 	if (pgrp == p->p_pgrp)
928 		/* nothing to do */
929 		goto done;
930 
931 	/* Ok all setup, link up required structures */
932 
933 	if (pgrp == NULL) {
934 		pgrp = new_pgrp;
935 		new_pgrp = NULL;
936 		if (sess != NULL) {
937 			sess->s_sid = p->p_pid;
938 			sess->s_leader = p;
939 			sess->s_count = 1;
940 			sess->s_ttyvp = NULL;
941 			sess->s_ttyp = NULL;
942 			sess->s_flags = p->p_session->s_flags & ~S_LOGIN_SET;
943 			memcpy(sess->s_login, p->p_session->s_login,
944 			    sizeof(sess->s_login));
945 			p->p_lflag &= ~PL_CONTROLT;
946 		} else {
947 			sess = p->p_pgrp->pg_session;
948 			proc_sesshold(sess);
949 		}
950 		pgrp->pg_session = sess;
951 		sess = NULL;
952 
953 		pgrp->pg_id = pgid;
954 		LIST_INIT(&pgrp->pg_members);
955 #ifdef DIAGNOSTIC
956 		if (__predict_false(pid_table[pgid & pid_tbl_mask].pt_pgrp))
957 			panic("enterpgrp: pgrp table slot in use");
958 		if (__predict_false(mksess && p != curp))
959 			panic("enterpgrp: mksession and p != curproc");
960 #endif
961 		pid_table[pgid & pid_tbl_mask].pt_pgrp = pgrp;
962 		pgrp->pg_jobc = 0;
963 	}
964 
965 	/*
966 	 * Adjust eligibility of affected pgrps to participate in job control.
967 	 * Increment eligibility counts before decrementing, otherwise we
968 	 * could reach 0 spuriously during the first call.
969 	 */
970 	fixjobc(p, pgrp, 1);
971 	fixjobc(p, p->p_pgrp, 0);
972 
973 	/* Interlock with ttread(). */
974 	mutex_spin_enter(&tty_lock);
975 
976 	/* Move process to requested group. */
977 	LIST_REMOVE(p, p_pglist);
978 	if (LIST_EMPTY(&p->p_pgrp->pg_members))
979 		/* defer delete until we've dumped the lock */
980 		pg_id = p->p_pgrp->pg_id;
981 	p->p_pgrp = pgrp;
982 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
983 
984 	/* Done with the swap; we can release the tty mutex. */
985 	mutex_spin_exit(&tty_lock);
986 
987     done:
988 	if (pg_id != NO_PGID) {
989 		/* Releases proc_lock. */
990 		pg_delete(pg_id);
991 	} else {
992 		mutex_exit(proc_lock);
993 	}
994 	if (sess != NULL)
995 		kmem_free(sess, sizeof(*sess));
996 	if (new_pgrp != NULL)
997 		kmem_free(new_pgrp, sizeof(*new_pgrp));
998 #ifdef DEBUG_PGRP
999 	if (__predict_false(rval))
1000 		printf("enterpgrp(%d,%d,%d), curproc %d, rval %d\n",
1001 			pid, pgid, mksess, curp->p_pid, rval);
1002 #endif
1003 	return rval;
1004 }
1005 
1006 /*
1007  * proc_leavepgrp: remove a process from its process group.
1008  *  => must be called with the proc_lock held, which will be released;
1009  */
1010 void
1011 proc_leavepgrp(struct proc *p)
1012 {
1013 	struct pgrp *pgrp;
1014 
1015 	KASSERT(mutex_owned(proc_lock));
1016 
1017 	/* Interlock with ttread() */
1018 	mutex_spin_enter(&tty_lock);
1019 	pgrp = p->p_pgrp;
1020 	LIST_REMOVE(p, p_pglist);
1021 	p->p_pgrp = NULL;
1022 	mutex_spin_exit(&tty_lock);
1023 
1024 	if (LIST_EMPTY(&pgrp->pg_members)) {
1025 		/* Releases proc_lock. */
1026 		pg_delete(pgrp->pg_id);
1027 	} else {
1028 		mutex_exit(proc_lock);
1029 	}
1030 }
1031 
1032 /*
1033  * pg_remove: remove a process group from the table.
1034  *  => must be called with the proc_lock held;
1035  *  => returns process group to free;
1036  */
1037 static struct pgrp *
1038 pg_remove(pid_t pg_id)
1039 {
1040 	struct pgrp *pgrp;
1041 	struct pid_table *pt;
1042 
1043 	KASSERT(mutex_owned(proc_lock));
1044 
1045 	pt = &pid_table[pg_id & pid_tbl_mask];
1046 	pgrp = pt->pt_pgrp;
1047 
1048 	KASSERT(pgrp != NULL);
1049 	KASSERT(pgrp->pg_id == pg_id);
1050 	KASSERT(LIST_EMPTY(&pgrp->pg_members));
1051 
1052 	pt->pt_pgrp = NULL;
1053 
1054 	if (!P_VALID(pt->pt_proc)) {
1055 		/* Orphaned pgrp, put slot onto free list. */
1056 		KASSERT((P_NEXT(pt->pt_proc) & pid_tbl_mask) == 0);
1057 		pg_id &= pid_tbl_mask;
1058 		pt = &pid_table[last_free_pt];
1059 		pt->pt_proc = P_FREE(P_NEXT(pt->pt_proc) | pg_id);
1060 		KASSERT(pt->pt_pid == 0);
1061 		last_free_pt = pg_id;
1062 		pid_alloc_cnt--;
1063 	}
1064 	return pgrp;
1065 }
1066 
1067 /*
1068  * pg_delete: delete and free a process group.
1069  *  => must be called with the proc_lock held, which will be released.
1070  */
1071 static void
1072 pg_delete(pid_t pg_id)
1073 {
1074 	struct pgrp *pg;
1075 	struct tty *ttyp;
1076 	struct session *ss;
1077 
1078 	KASSERT(mutex_owned(proc_lock));
1079 
1080 	pg = pid_table[pg_id & pid_tbl_mask].pt_pgrp;
1081 	if (pg == NULL || pg->pg_id != pg_id || !LIST_EMPTY(&pg->pg_members)) {
1082 		mutex_exit(proc_lock);
1083 		return;
1084 	}
1085 
1086 	ss = pg->pg_session;
1087 
1088 	/* Remove reference (if any) from tty to this process group */
1089 	mutex_spin_enter(&tty_lock);
1090 	ttyp = ss->s_ttyp;
1091 	if (ttyp != NULL && ttyp->t_pgrp == pg) {
1092 		ttyp->t_pgrp = NULL;
1093 		KASSERT(ttyp->t_session == ss);
1094 	}
1095 	mutex_spin_exit(&tty_lock);
1096 
1097 	/*
1098 	 * The leading process group in a session is freed by proc_sessrele(),
1099 	 * if last reference.  Note: proc_sessrele() releases proc_lock.
1100 	 */
1101 	pg = (ss->s_sid != pg->pg_id) ? pg_remove(pg_id) : NULL;
1102 	proc_sessrele(ss);
1103 
1104 	if (pg != NULL) {
1105 		/* Free it, if was not done by proc_sessrele(). */
1106 		kmem_free(pg, sizeof(struct pgrp));
1107 	}
1108 }
1109 
1110 /*
1111  * Adjust pgrp jobc counters when specified process changes process group.
1112  * We count the number of processes in each process group that "qualify"
1113  * the group for terminal job control (those with a parent in a different
1114  * process group of the same session).  If that count reaches zero, the
1115  * process group becomes orphaned.  Check both the specified process'
1116  * process group and that of its children.
1117  * entering == 0 => p is leaving specified group.
1118  * entering == 1 => p is entering specified group.
1119  *
1120  * Call with proc_lock held.
1121  */
1122 void
1123 fixjobc(struct proc *p, struct pgrp *pgrp, int entering)
1124 {
1125 	struct pgrp *hispgrp;
1126 	struct session *mysession = pgrp->pg_session;
1127 	struct proc *child;
1128 
1129 	KASSERT(mutex_owned(proc_lock));
1130 
1131 	/*
1132 	 * Check p's parent to see whether p qualifies its own process
1133 	 * group; if so, adjust count for p's process group.
1134 	 */
1135 	hispgrp = p->p_pptr->p_pgrp;
1136 	if (hispgrp != pgrp && hispgrp->pg_session == mysession) {
1137 		if (entering) {
1138 			pgrp->pg_jobc++;
1139 			p->p_lflag &= ~PL_ORPHANPG;
1140 		} else if (--pgrp->pg_jobc == 0)
1141 			orphanpg(pgrp);
1142 	}
1143 
1144 	/*
1145 	 * Check this process' children to see whether they qualify
1146 	 * their process groups; if so, adjust counts for children's
1147 	 * process groups.
1148 	 */
1149 	LIST_FOREACH(child, &p->p_children, p_sibling) {
1150 		hispgrp = child->p_pgrp;
1151 		if (hispgrp != pgrp && hispgrp->pg_session == mysession &&
1152 		    !P_ZOMBIE(child)) {
1153 			if (entering) {
1154 				child->p_lflag &= ~PL_ORPHANPG;
1155 				hispgrp->pg_jobc++;
1156 			} else if (--hispgrp->pg_jobc == 0)
1157 				orphanpg(hispgrp);
1158 		}
1159 	}
1160 }
1161 
1162 /*
1163  * A process group has become orphaned;
1164  * if there are any stopped processes in the group,
1165  * hang-up all process in that group.
1166  *
1167  * Call with proc_lock held.
1168  */
1169 static void
1170 orphanpg(struct pgrp *pg)
1171 {
1172 	struct proc *p;
1173 
1174 	KASSERT(mutex_owned(proc_lock));
1175 
1176 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
1177 		if (p->p_stat == SSTOP) {
1178 			p->p_lflag |= PL_ORPHANPG;
1179 			psignal(p, SIGHUP);
1180 			psignal(p, SIGCONT);
1181 		}
1182 	}
1183 }
1184 
1185 #ifdef DDB
1186 #include <ddb/db_output.h>
1187 void pidtbl_dump(void);
1188 void
1189 pidtbl_dump(void)
1190 {
1191 	struct pid_table *pt;
1192 	struct proc *p;
1193 	struct pgrp *pgrp;
1194 	int id;
1195 
1196 	db_printf("pid table %p size %x, next %x, last %x\n",
1197 		pid_table, pid_tbl_mask+1,
1198 		next_free_pt, last_free_pt);
1199 	for (pt = pid_table, id = 0; id <= pid_tbl_mask; id++, pt++) {
1200 		p = pt->pt_proc;
1201 		if (!P_VALID(p) && !pt->pt_pgrp)
1202 			continue;
1203 		db_printf("  id %x: ", id);
1204 		if (P_VALID(p))
1205 			db_printf("slotpid %d proc %p id %d (0x%x) %s\n",
1206 				pt->pt_pid, p, p->p_pid, p->p_pid, p->p_comm);
1207 		else
1208 			db_printf("next %x use %x\n",
1209 				P_NEXT(p) & pid_tbl_mask,
1210 				P_NEXT(p) & ~pid_tbl_mask);
1211 		if ((pgrp = pt->pt_pgrp)) {
1212 			db_printf("\tsession %p, sid %d, count %d, login %s\n",
1213 			    pgrp->pg_session, pgrp->pg_session->s_sid,
1214 			    pgrp->pg_session->s_count,
1215 			    pgrp->pg_session->s_login);
1216 			db_printf("\tpgrp %p, pg_id %d, pg_jobc %d, members %p\n",
1217 			    pgrp, pgrp->pg_id, pgrp->pg_jobc,
1218 			    LIST_FIRST(&pgrp->pg_members));
1219 			LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
1220 				db_printf("\t\tpid %d addr %p pgrp %p %s\n",
1221 				    p->p_pid, p, p->p_pgrp, p->p_comm);
1222 			}
1223 		}
1224 	}
1225 }
1226 #endif /* DDB */
1227 
1228 #ifdef KSTACK_CHECK_MAGIC
1229 
1230 #define	KSTACK_MAGIC	0xdeadbeaf
1231 
1232 /* XXX should be per process basis? */
1233 static int	kstackleftmin = KSTACK_SIZE;
1234 static int	kstackleftthres = KSTACK_SIZE / 8;
1235 
1236 void
1237 kstack_setup_magic(const struct lwp *l)
1238 {
1239 	uint32_t *ip;
1240 	uint32_t const *end;
1241 
1242 	KASSERT(l != NULL);
1243 	KASSERT(l != &lwp0);
1244 
1245 	/*
1246 	 * fill all the stack with magic number
1247 	 * so that later modification on it can be detected.
1248 	 */
1249 	ip = (uint32_t *)KSTACK_LOWEST_ADDR(l);
1250 	end = (uint32_t *)((char *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE);
1251 	for (; ip < end; ip++) {
1252 		*ip = KSTACK_MAGIC;
1253 	}
1254 }
1255 
1256 void
1257 kstack_check_magic(const struct lwp *l)
1258 {
1259 	uint32_t const *ip, *end;
1260 	int stackleft;
1261 
1262 	KASSERT(l != NULL);
1263 
1264 	/* don't check proc0 */ /*XXX*/
1265 	if (l == &lwp0)
1266 		return;
1267 
1268 #ifdef __MACHINE_STACK_GROWS_UP
1269 	/* stack grows upwards (eg. hppa) */
1270 	ip = (uint32_t *)((void *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE);
1271 	end = (uint32_t *)KSTACK_LOWEST_ADDR(l);
1272 	for (ip--; ip >= end; ip--)
1273 		if (*ip != KSTACK_MAGIC)
1274 			break;
1275 
1276 	stackleft = (void *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE - (void *)ip;
1277 #else /* __MACHINE_STACK_GROWS_UP */
1278 	/* stack grows downwards (eg. i386) */
1279 	ip = (uint32_t *)KSTACK_LOWEST_ADDR(l);
1280 	end = (uint32_t *)((char *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE);
1281 	for (; ip < end; ip++)
1282 		if (*ip != KSTACK_MAGIC)
1283 			break;
1284 
1285 	stackleft = ((const char *)ip) - (const char *)KSTACK_LOWEST_ADDR(l);
1286 #endif /* __MACHINE_STACK_GROWS_UP */
1287 
1288 	if (kstackleftmin > stackleft) {
1289 		kstackleftmin = stackleft;
1290 		if (stackleft < kstackleftthres)
1291 			printf("warning: kernel stack left %d bytes"
1292 			    "(pid %u:lid %u)\n", stackleft,
1293 			    (u_int)l->l_proc->p_pid, (u_int)l->l_lid);
1294 	}
1295 
1296 	if (stackleft <= 0) {
1297 		panic("magic on the top of kernel stack changed for "
1298 		    "pid %u, lid %u: maybe kernel stack overflow",
1299 		    (u_int)l->l_proc->p_pid, (u_int)l->l_lid);
1300 	}
1301 }
1302 #endif /* KSTACK_CHECK_MAGIC */
1303 
1304 int
1305 proclist_foreach_call(struct proclist *list,
1306     int (*callback)(struct proc *, void *arg), void *arg)
1307 {
1308 	struct proc marker;
1309 	struct proc *p;
1310 	int ret = 0;
1311 
1312 	marker.p_flag = PK_MARKER;
1313 	mutex_enter(proc_lock);
1314 	for (p = LIST_FIRST(list); ret == 0 && p != NULL;) {
1315 		if (p->p_flag & PK_MARKER) {
1316 			p = LIST_NEXT(p, p_list);
1317 			continue;
1318 		}
1319 		LIST_INSERT_AFTER(p, &marker, p_list);
1320 		ret = (*callback)(p, arg);
1321 		KASSERT(mutex_owned(proc_lock));
1322 		p = LIST_NEXT(&marker, p_list);
1323 		LIST_REMOVE(&marker, p_list);
1324 	}
1325 	mutex_exit(proc_lock);
1326 
1327 	return ret;
1328 }
1329 
1330 int
1331 proc_vmspace_getref(struct proc *p, struct vmspace **vm)
1332 {
1333 
1334 	/* XXXCDC: how should locking work here? */
1335 
1336 	/* curproc exception is for coredump. */
1337 
1338 	if ((p != curproc && (p->p_sflag & PS_WEXIT) != 0) ||
1339 	    (p->p_vmspace->vm_refcnt < 1)) { /* XXX */
1340 		return EFAULT;
1341 	}
1342 
1343 	uvmspace_addref(p->p_vmspace);
1344 	*vm = p->p_vmspace;
1345 
1346 	return 0;
1347 }
1348 
1349 /*
1350  * Acquire a write lock on the process credential.
1351  */
1352 void
1353 proc_crmod_enter(void)
1354 {
1355 	struct lwp *l = curlwp;
1356 	struct proc *p = l->l_proc;
1357 	kauth_cred_t oc;
1358 
1359 	/* Reset what needs to be reset in plimit. */
1360 	if (p->p_limit->pl_corename != defcorename) {
1361 		lim_setcorename(p, defcorename, 0);
1362 	}
1363 
1364 	mutex_enter(p->p_lock);
1365 
1366 	/* Ensure the LWP cached credentials are up to date. */
1367 	if ((oc = l->l_cred) != p->p_cred) {
1368 		kauth_cred_hold(p->p_cred);
1369 		l->l_cred = p->p_cred;
1370 		kauth_cred_free(oc);
1371 	}
1372 }
1373 
1374 /*
1375  * Set in a new process credential, and drop the write lock.  The credential
1376  * must have a reference already.  Optionally, free a no-longer required
1377  * credential.  The scheduler also needs to inspect p_cred, so we also
1378  * briefly acquire the sched state mutex.
1379  */
1380 void
1381 proc_crmod_leave(kauth_cred_t scred, kauth_cred_t fcred, bool sugid)
1382 {
1383 	struct lwp *l = curlwp, *l2;
1384 	struct proc *p = l->l_proc;
1385 	kauth_cred_t oc;
1386 
1387 	KASSERT(mutex_owned(p->p_lock));
1388 
1389 	/* Is there a new credential to set in? */
1390 	if (scred != NULL) {
1391 		p->p_cred = scred;
1392 		LIST_FOREACH(l2, &p->p_lwps, l_sibling) {
1393 			if (l2 != l)
1394 				l2->l_prflag |= LPR_CRMOD;
1395 		}
1396 
1397 		/* Ensure the LWP cached credentials are up to date. */
1398 		if ((oc = l->l_cred) != scred) {
1399 			kauth_cred_hold(scred);
1400 			l->l_cred = scred;
1401 		}
1402 	} else
1403 		oc = NULL;	/* XXXgcc */
1404 
1405 	if (sugid) {
1406 		/*
1407 		 * Mark process as having changed credentials, stops
1408 		 * tracing etc.
1409 		 */
1410 		p->p_flag |= PK_SUGID;
1411 	}
1412 
1413 	mutex_exit(p->p_lock);
1414 
1415 	/* If there is a credential to be released, free it now. */
1416 	if (fcred != NULL) {
1417 		KASSERT(scred != NULL);
1418 		kauth_cred_free(fcred);
1419 		if (oc != scred)
1420 			kauth_cred_free(oc);
1421 	}
1422 }
1423 
1424 /*
1425  * proc_specific_key_create --
1426  *	Create a key for subsystem proc-specific data.
1427  */
1428 int
1429 proc_specific_key_create(specificdata_key_t *keyp, specificdata_dtor_t dtor)
1430 {
1431 
1432 	return (specificdata_key_create(proc_specificdata_domain, keyp, dtor));
1433 }
1434 
1435 /*
1436  * proc_specific_key_delete --
1437  *	Delete a key for subsystem proc-specific data.
1438  */
1439 void
1440 proc_specific_key_delete(specificdata_key_t key)
1441 {
1442 
1443 	specificdata_key_delete(proc_specificdata_domain, key);
1444 }
1445 
1446 /*
1447  * proc_initspecific --
1448  *	Initialize a proc's specificdata container.
1449  */
1450 void
1451 proc_initspecific(struct proc *p)
1452 {
1453 	int error;
1454 
1455 	error = specificdata_init(proc_specificdata_domain, &p->p_specdataref);
1456 	KASSERT(error == 0);
1457 }
1458 
1459 /*
1460  * proc_finispecific --
1461  *	Finalize a proc's specificdata container.
1462  */
1463 void
1464 proc_finispecific(struct proc *p)
1465 {
1466 
1467 	specificdata_fini(proc_specificdata_domain, &p->p_specdataref);
1468 }
1469 
1470 /*
1471  * proc_getspecific --
1472  *	Return proc-specific data corresponding to the specified key.
1473  */
1474 void *
1475 proc_getspecific(struct proc *p, specificdata_key_t key)
1476 {
1477 
1478 	return (specificdata_getspecific(proc_specificdata_domain,
1479 					 &p->p_specdataref, key));
1480 }
1481 
1482 /*
1483  * proc_setspecific --
1484  *	Set proc-specific data corresponding to the specified key.
1485  */
1486 void
1487 proc_setspecific(struct proc *p, specificdata_key_t key, void *data)
1488 {
1489 
1490 	specificdata_setspecific(proc_specificdata_domain,
1491 				 &p->p_specdataref, key, data);
1492 }
1493 
1494 int
1495 proc_uidmatch(kauth_cred_t cred, kauth_cred_t target)
1496 {
1497 	int r = 0;
1498 
1499 	if (kauth_cred_getuid(cred) != kauth_cred_getuid(target) ||
1500 	    kauth_cred_getuid(cred) != kauth_cred_getsvuid(target)) {
1501 		/*
1502 		 * suid proc of ours or proc not ours
1503 		 */
1504 		r = EPERM;
1505 	} else if (kauth_cred_getgid(target) != kauth_cred_getsvgid(target)) {
1506 		/*
1507 		 * sgid proc has sgid back to us temporarily
1508 		 */
1509 		r = EPERM;
1510 	} else {
1511 		/*
1512 		 * our rgid must be in target's group list (ie,
1513 		 * sub-processes started by a sgid process)
1514 		 */
1515 		int ismember = 0;
1516 
1517 		if (kauth_cred_ismember_gid(cred,
1518 		    kauth_cred_getgid(target), &ismember) != 0 ||
1519 		    !ismember)
1520 			r = EPERM;
1521 	}
1522 
1523 	return (r);
1524 }
1525 
1526 /*
1527  * sysctl stuff
1528  */
1529 
1530 #define KERN_PROCSLOP	(5 * sizeof(struct kinfo_proc))
1531 
1532 static const u_int sysctl_flagmap[] = {
1533 	PK_ADVLOCK, P_ADVLOCK,
1534 	PK_EXEC, P_EXEC,
1535 	PK_NOCLDWAIT, P_NOCLDWAIT,
1536 	PK_32, P_32,
1537 	PK_CLDSIGIGN, P_CLDSIGIGN,
1538 	PK_SUGID, P_SUGID,
1539 	0
1540 };
1541 
1542 static const u_int sysctl_sflagmap[] = {
1543 	PS_NOCLDSTOP, P_NOCLDSTOP,
1544 	PS_WEXIT, P_WEXIT,
1545 	PS_STOPFORK, P_STOPFORK,
1546 	PS_STOPEXEC, P_STOPEXEC,
1547 	PS_STOPEXIT, P_STOPEXIT,
1548 	0
1549 };
1550 
1551 static const u_int sysctl_slflagmap[] = {
1552 	PSL_TRACED, P_TRACED,
1553 	PSL_FSTRACE, P_FSTRACE,
1554 	PSL_CHTRACED, P_CHTRACED,
1555 	PSL_SYSCALL, P_SYSCALL,
1556 	0
1557 };
1558 
1559 static const u_int sysctl_lflagmap[] = {
1560 	PL_CONTROLT, P_CONTROLT,
1561 	PL_PPWAIT, P_PPWAIT,
1562 	0
1563 };
1564 
1565 static const u_int sysctl_stflagmap[] = {
1566 	PST_PROFIL, P_PROFIL,
1567 	0
1568 
1569 };
1570 
1571 /* used by kern_lwp also */
1572 const u_int sysctl_lwpflagmap[] = {
1573 	LW_SINTR, L_SINTR,
1574 	LW_SYSTEM, L_SYSTEM,
1575 	0
1576 };
1577 
1578 /*
1579  * Find the most ``active'' lwp of a process and return it for ps display
1580  * purposes
1581  */
1582 static struct lwp *
1583 proc_active_lwp(struct proc *p)
1584 {
1585 	static const int ostat[] = {
1586 		0,
1587 		2,	/* LSIDL */
1588 		6,	/* LSRUN */
1589 		5,	/* LSSLEEP */
1590 		4,	/* LSSTOP */
1591 		0,	/* LSZOMB */
1592 		1,	/* LSDEAD */
1593 		7,	/* LSONPROC */
1594 		3	/* LSSUSPENDED */
1595 	};
1596 
1597 	struct lwp *l, *lp = NULL;
1598 	LIST_FOREACH(l, &p->p_lwps, l_sibling) {
1599 		KASSERT(l->l_stat >= 0 && l->l_stat < __arraycount(ostat));
1600 		if (lp == NULL ||
1601 		    ostat[l->l_stat] > ostat[lp->l_stat] ||
1602 		    (ostat[l->l_stat] == ostat[lp->l_stat] &&
1603 		    l->l_cpticks > lp->l_cpticks)) {
1604 			lp = l;
1605 			continue;
1606 		}
1607 	}
1608 	return lp;
1609 }
1610 
1611 static int
1612 sysctl_doeproc(SYSCTLFN_ARGS)
1613 {
1614 	union {
1615 		struct kinfo_proc kproc;
1616 		struct kinfo_proc2 kproc2;
1617 	} *kbuf;
1618 	struct proc *p, *next, *marker;
1619 	char *where, *dp;
1620 	int type, op, arg, error;
1621 	u_int elem_size, kelem_size, elem_count;
1622 	size_t buflen, needed;
1623 	bool match, zombie, mmmbrains;
1624 
1625 	if (namelen == 1 && name[0] == CTL_QUERY)
1626 		return (sysctl_query(SYSCTLFN_CALL(rnode)));
1627 
1628 	dp = where = oldp;
1629 	buflen = where != NULL ? *oldlenp : 0;
1630 	error = 0;
1631 	needed = 0;
1632 	type = rnode->sysctl_num;
1633 
1634 	if (type == KERN_PROC) {
1635 		if (namelen != 2 && !(namelen == 1 && name[0] == KERN_PROC_ALL))
1636 			return (EINVAL);
1637 		op = name[0];
1638 		if (op != KERN_PROC_ALL)
1639 			arg = name[1];
1640 		else
1641 			arg = 0;		/* Quell compiler warning */
1642 		elem_count = 0;	/* Ditto */
1643 		kelem_size = elem_size = sizeof(kbuf->kproc);
1644 	} else {
1645 		if (namelen != 4)
1646 			return (EINVAL);
1647 		op = name[0];
1648 		arg = name[1];
1649 		elem_size = name[2];
1650 		elem_count = name[3];
1651 		kelem_size = sizeof(kbuf->kproc2);
1652 	}
1653 
1654 	sysctl_unlock();
1655 
1656 	kbuf = kmem_alloc(sizeof(*kbuf), KM_SLEEP);
1657 	marker = kmem_alloc(sizeof(*marker), KM_SLEEP);
1658 	marker->p_flag = PK_MARKER;
1659 
1660 	mutex_enter(proc_lock);
1661 	mmmbrains = false;
1662 	for (p = LIST_FIRST(&allproc);; p = next) {
1663 		if (p == NULL) {
1664 			if (!mmmbrains) {
1665 				p = LIST_FIRST(&zombproc);
1666 				mmmbrains = true;
1667 			}
1668 			if (p == NULL)
1669 				break;
1670 		}
1671 		next = LIST_NEXT(p, p_list);
1672 		if ((p->p_flag & PK_MARKER) != 0)
1673 			continue;
1674 
1675 		/*
1676 		 * Skip embryonic processes.
1677 		 */
1678 		if (p->p_stat == SIDL)
1679 			continue;
1680 
1681 		mutex_enter(p->p_lock);
1682 		error = kauth_authorize_process(l->l_cred,
1683 		    KAUTH_PROCESS_CANSEE, p,
1684 		    KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL);
1685 		if (error != 0) {
1686 			mutex_exit(p->p_lock);
1687 			continue;
1688 		}
1689 
1690 		/*
1691 		 * TODO - make more efficient (see notes below).
1692 		 * do by session.
1693 		 */
1694 		switch (op) {
1695 		case KERN_PROC_PID:
1696 			/* could do this with just a lookup */
1697 			match = (p->p_pid == (pid_t)arg);
1698 			break;
1699 
1700 		case KERN_PROC_PGRP:
1701 			/* could do this by traversing pgrp */
1702 			match = (p->p_pgrp->pg_id == (pid_t)arg);
1703 			break;
1704 
1705 		case KERN_PROC_SESSION:
1706 			match = (p->p_session->s_sid == (pid_t)arg);
1707 			break;
1708 
1709 		case KERN_PROC_TTY:
1710 			match = true;
1711 			if (arg == (int) KERN_PROC_TTY_REVOKE) {
1712 				if ((p->p_lflag & PL_CONTROLT) == 0 ||
1713 				    p->p_session->s_ttyp == NULL ||
1714 				    p->p_session->s_ttyvp != NULL) {
1715 				    	match = false;
1716 				}
1717 			} else if ((p->p_lflag & PL_CONTROLT) == 0 ||
1718 			    p->p_session->s_ttyp == NULL) {
1719 				if ((dev_t)arg != KERN_PROC_TTY_NODEV) {
1720 					match = false;
1721 				}
1722 			} else if (p->p_session->s_ttyp->t_dev != (dev_t)arg) {
1723 				match = false;
1724 			}
1725 			break;
1726 
1727 		case KERN_PROC_UID:
1728 			match = (kauth_cred_geteuid(p->p_cred) == (uid_t)arg);
1729 			break;
1730 
1731 		case KERN_PROC_RUID:
1732 			match = (kauth_cred_getuid(p->p_cred) == (uid_t)arg);
1733 			break;
1734 
1735 		case KERN_PROC_GID:
1736 			match = (kauth_cred_getegid(p->p_cred) == (uid_t)arg);
1737 			break;
1738 
1739 		case KERN_PROC_RGID:
1740 			match = (kauth_cred_getgid(p->p_cred) == (uid_t)arg);
1741 			break;
1742 
1743 		case KERN_PROC_ALL:
1744 			match = true;
1745 			/* allow everything */
1746 			break;
1747 
1748 		default:
1749 			error = EINVAL;
1750 			mutex_exit(p->p_lock);
1751 			goto cleanup;
1752 		}
1753 		if (!match) {
1754 			mutex_exit(p->p_lock);
1755 			continue;
1756 		}
1757 
1758 		/*
1759 		 * Grab a hold on the process.
1760 		 */
1761 		if (mmmbrains) {
1762 			zombie = true;
1763 		} else {
1764 			zombie = !rw_tryenter(&p->p_reflock, RW_READER);
1765 		}
1766 		if (zombie) {
1767 			LIST_INSERT_AFTER(p, marker, p_list);
1768 		}
1769 
1770 		if (buflen >= elem_size &&
1771 		    (type == KERN_PROC || elem_count > 0)) {
1772 			if (type == KERN_PROC) {
1773 				kbuf->kproc.kp_proc = *p;
1774 				fill_eproc(p, &kbuf->kproc.kp_eproc, zombie);
1775 			} else {
1776 				fill_kproc2(p, &kbuf->kproc2, zombie);
1777 				elem_count--;
1778 			}
1779 			mutex_exit(p->p_lock);
1780 			mutex_exit(proc_lock);
1781 			/*
1782 			 * Copy out elem_size, but not larger than kelem_size
1783 			 */
1784 			error = sysctl_copyout(l, kbuf, dp,
1785 			    min(kelem_size, elem_size));
1786 			mutex_enter(proc_lock);
1787 			if (error) {
1788 				goto bah;
1789 			}
1790 			dp += elem_size;
1791 			buflen -= elem_size;
1792 		} else {
1793 			mutex_exit(p->p_lock);
1794 		}
1795 		needed += elem_size;
1796 
1797 		/*
1798 		 * Release reference to process.
1799 		 */
1800 	 	if (zombie) {
1801 			next = LIST_NEXT(marker, p_list);
1802  			LIST_REMOVE(marker, p_list);
1803 		} else {
1804 			rw_exit(&p->p_reflock);
1805 			next = LIST_NEXT(p, p_list);
1806 		}
1807 	}
1808 	mutex_exit(proc_lock);
1809 
1810 	if (where != NULL) {
1811 		*oldlenp = dp - where;
1812 		if (needed > *oldlenp) {
1813 			error = ENOMEM;
1814 			goto out;
1815 		}
1816 	} else {
1817 		needed += KERN_PROCSLOP;
1818 		*oldlenp = needed;
1819 	}
1820 	if (kbuf)
1821 		kmem_free(kbuf, sizeof(*kbuf));
1822 	if (marker)
1823 		kmem_free(marker, sizeof(*marker));
1824 	sysctl_relock();
1825 	return 0;
1826  bah:
1827  	if (zombie)
1828  		LIST_REMOVE(marker, p_list);
1829 	else
1830 		rw_exit(&p->p_reflock);
1831  cleanup:
1832 	mutex_exit(proc_lock);
1833  out:
1834 	if (kbuf)
1835 		kmem_free(kbuf, sizeof(*kbuf));
1836 	if (marker)
1837 		kmem_free(marker, sizeof(*marker));
1838 	sysctl_relock();
1839 	return error;
1840 }
1841 
1842 int
1843 copyin_psstrings(struct proc *p, struct ps_strings *arginfo)
1844 {
1845 
1846 #ifdef COMPAT_NETBSD32
1847 	if (p->p_flag & PK_32) {
1848 		struct ps_strings32 arginfo32;
1849 
1850 		int error = copyin_proc(p, (void *)p->p_psstrp, &arginfo32,
1851 		    sizeof(arginfo32));
1852 		if (error)
1853 			return error;
1854 		arginfo->ps_argvstr = (void *)(uintptr_t)arginfo32.ps_argvstr;
1855 		arginfo->ps_nargvstr = arginfo32.ps_nargvstr;
1856 		arginfo->ps_envstr = (void *)(uintptr_t)arginfo32.ps_envstr;
1857 		arginfo->ps_nenvstr = arginfo32.ps_nenvstr;
1858 		return 0;
1859 	}
1860 #endif
1861 	return copyin_proc(p, (void *)p->p_psstrp, arginfo, sizeof(*arginfo));
1862 }
1863 
1864 static int
1865 copy_procargs_sysctl_cb(void *cookie_, const void *src, size_t off, size_t len)
1866 {
1867 	void **cookie = cookie_;
1868 	struct lwp *l = cookie[0];
1869 	char *dst = cookie[1];
1870 
1871 	return sysctl_copyout(l, src, dst + off, len);
1872 }
1873 
1874 /*
1875  * sysctl helper routine for kern.proc_args pseudo-subtree.
1876  */
1877 static int
1878 sysctl_kern_proc_args(SYSCTLFN_ARGS)
1879 {
1880 	struct ps_strings pss;
1881 	struct proc *p;
1882 	pid_t pid;
1883 	int type, error;
1884 	void *cookie[2];
1885 
1886 	if (namelen == 1 && name[0] == CTL_QUERY)
1887 		return (sysctl_query(SYSCTLFN_CALL(rnode)));
1888 
1889 	if (newp != NULL || namelen != 2)
1890 		return (EINVAL);
1891 	pid = name[0];
1892 	type = name[1];
1893 
1894 	switch (type) {
1895 	case KERN_PROC_ARGV:
1896 	case KERN_PROC_NARGV:
1897 	case KERN_PROC_ENV:
1898 	case KERN_PROC_NENV:
1899 		/* ok */
1900 		break;
1901 	default:
1902 		return (EINVAL);
1903 	}
1904 
1905 	sysctl_unlock();
1906 
1907 	/* check pid */
1908 	mutex_enter(proc_lock);
1909 	if ((p = proc_find(pid)) == NULL) {
1910 		error = EINVAL;
1911 		goto out_locked;
1912 	}
1913 	mutex_enter(p->p_lock);
1914 
1915 	/* Check permission. */
1916 	if (type == KERN_PROC_ARGV || type == KERN_PROC_NARGV)
1917 		error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE,
1918 		    p, KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ARGS), NULL, NULL);
1919 	else if (type == KERN_PROC_ENV || type == KERN_PROC_NENV)
1920 		error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE,
1921 		    p, KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENV), NULL, NULL);
1922 	else
1923 		error = EINVAL; /* XXXGCC */
1924 	if (error) {
1925 		mutex_exit(p->p_lock);
1926 		goto out_locked;
1927 	}
1928 
1929 	if (oldp == NULL) {
1930 		if (type == KERN_PROC_NARGV || type == KERN_PROC_NENV)
1931 			*oldlenp = sizeof (int);
1932 		else
1933 			*oldlenp = ARG_MAX;	/* XXX XXX XXX */
1934 		error = 0;
1935 		mutex_exit(p->p_lock);
1936 		goto out_locked;
1937 	}
1938 
1939 	/*
1940 	 * Zombies don't have a stack, so we can't read their psstrings.
1941 	 * System processes also don't have a user stack.
1942 	 */
1943 	if (P_ZOMBIE(p) || (p->p_flag & PK_SYSTEM) != 0) {
1944 		error = EINVAL;
1945 		mutex_exit(p->p_lock);
1946 		goto out_locked;
1947 	}
1948 
1949 	error = rw_tryenter(&p->p_reflock, RW_READER) ? 0 : EBUSY;
1950 	mutex_exit(p->p_lock);
1951 	if (error) {
1952 		goto out_locked;
1953 	}
1954 	mutex_exit(proc_lock);
1955 
1956 	if (type == KERN_PROC_NARGV || type == KERN_PROC_NENV) {
1957 		int value;
1958 		if ((error = copyin_psstrings(p, &pss)) == 0) {
1959 			if (type == KERN_PROC_NARGV)
1960 				value = pss.ps_nargvstr;
1961 			else
1962 				value = pss.ps_nenvstr;
1963 			error = sysctl_copyout(l, &value, oldp, sizeof(value));
1964 			*oldlenp = sizeof(value);
1965 		}
1966 	} else {
1967 		cookie[0] = l;
1968 		cookie[1] = oldp;
1969 		error = copy_procargs(p, type, oldlenp,
1970 		    copy_procargs_sysctl_cb, cookie);
1971 	}
1972 	rw_exit(&p->p_reflock);
1973 	sysctl_relock();
1974 	return error;
1975 
1976 out_locked:
1977 	mutex_exit(proc_lock);
1978 	sysctl_relock();
1979 	return error;
1980 }
1981 
1982 int
1983 copy_procargs(struct proc *p, int oid, size_t *limit,
1984     int (*cb)(void *, const void *, size_t, size_t), void *cookie)
1985 {
1986 	struct ps_strings pss;
1987 	size_t len, i, loaded, entry_len;
1988 	struct uio auio;
1989 	struct iovec aiov;
1990 	int error, argvlen;
1991 	char *arg;
1992 	char **argv;
1993 	vaddr_t user_argv;
1994 	struct vmspace *vmspace;
1995 
1996 	/*
1997 	 * Allocate a temporary buffer to hold the argument vector and
1998 	 * the arguments themselve.
1999 	 */
2000 	arg = kmem_alloc(PAGE_SIZE, KM_SLEEP);
2001 	argv = kmem_alloc(PAGE_SIZE, KM_SLEEP);
2002 
2003 	/*
2004 	 * Lock the process down in memory.
2005 	 */
2006 	vmspace = p->p_vmspace;
2007 	uvmspace_addref(vmspace);
2008 
2009 	/*
2010 	 * Read in the ps_strings structure.
2011 	 */
2012 	if ((error = copyin_psstrings(p, &pss)) != 0)
2013 		goto done;
2014 
2015 	/*
2016 	 * Now read the address of the argument vector.
2017 	 */
2018 	switch (oid) {
2019 	case KERN_PROC_ARGV:
2020 		user_argv = (uintptr_t)pss.ps_argvstr;
2021 		argvlen = pss.ps_nargvstr;
2022 		break;
2023 	case KERN_PROC_ENV:
2024 		user_argv = (uintptr_t)pss.ps_envstr;
2025 		argvlen = pss.ps_nenvstr;
2026 		break;
2027 	default:
2028 		error = EINVAL;
2029 		goto done;
2030 	}
2031 
2032 	if (argvlen < 0) {
2033 		error = EIO;
2034 		goto done;
2035 	}
2036 
2037 #ifdef COMPAT_NETBSD32
2038 	if (p->p_flag & PK_32)
2039 		entry_len = sizeof(netbsd32_charp);
2040 	else
2041 #endif
2042 		entry_len = sizeof(char *);
2043 
2044 	/*
2045 	 * Now copy each string.
2046 	 */
2047 	len = 0; /* bytes written to user buffer */
2048 	loaded = 0; /* bytes from argv already processed */
2049 	i = 0; /* To make compiler happy */
2050 
2051 	for (; argvlen; --argvlen) {
2052 		int finished = 0;
2053 		vaddr_t base;
2054 		size_t xlen;
2055 		int j;
2056 
2057 		if (loaded == 0) {
2058 			size_t rem = entry_len * argvlen;
2059 			loaded = MIN(rem, PAGE_SIZE);
2060 			error = copyin_vmspace(vmspace,
2061 			    (const void *)user_argv, argv, loaded);
2062 			if (error)
2063 				break;
2064 			user_argv += loaded;
2065 			i = 0;
2066 		}
2067 
2068 #ifdef COMPAT_NETBSD32
2069 		if (p->p_flag & PK_32) {
2070 			netbsd32_charp *argv32;
2071 
2072 			argv32 = (netbsd32_charp *)argv;
2073 			base = (vaddr_t)NETBSD32PTR64(argv32[i++]);
2074 		} else
2075 #endif
2076 			base = (vaddr_t)argv[i++];
2077 		loaded -= entry_len;
2078 
2079 		/*
2080 		 * The program has messed around with its arguments,
2081 		 * possibly deleting some, and replacing them with
2082 		 * NULL's. Treat this as the last argument and not
2083 		 * a failure.
2084 		 */
2085 		if (base == 0)
2086 			break;
2087 
2088 		while (!finished) {
2089 			xlen = PAGE_SIZE - (base & PAGE_MASK);
2090 
2091 			aiov.iov_base = arg;
2092 			aiov.iov_len = PAGE_SIZE;
2093 			auio.uio_iov = &aiov;
2094 			auio.uio_iovcnt = 1;
2095 			auio.uio_offset = base;
2096 			auio.uio_resid = xlen;
2097 			auio.uio_rw = UIO_READ;
2098 			UIO_SETUP_SYSSPACE(&auio);
2099 			error = uvm_io(&vmspace->vm_map, &auio);
2100 			if (error)
2101 				goto done;
2102 
2103 			/* Look for the end of the string */
2104 			for (j = 0; j < xlen; j++) {
2105 				if (arg[j] == '\0') {
2106 					xlen = j + 1;
2107 					finished = 1;
2108 					break;
2109 				}
2110 			}
2111 
2112 			/* Check for user buffer overflow */
2113 			if (len + xlen > *limit) {
2114 				finished = 1;
2115 				if (len > *limit)
2116 					xlen = 0;
2117 				else
2118 					xlen = *limit - len;
2119 			}
2120 
2121 			/* Copyout the page */
2122 			error = (*cb)(cookie, arg, len, xlen);
2123 			if (error)
2124 				goto done;
2125 
2126 			len += xlen;
2127 			base += xlen;
2128 		}
2129 	}
2130 	*limit = len;
2131 
2132 done:
2133 	kmem_free(argv, PAGE_SIZE);
2134 	kmem_free(arg, PAGE_SIZE);
2135 	uvmspace_free(vmspace);
2136 	return error;
2137 }
2138 
2139 /*
2140  * Fill in an eproc structure for the specified process.
2141  */
2142 void
2143 fill_eproc(struct proc *p, struct eproc *ep, bool zombie)
2144 {
2145 	struct tty *tp;
2146 	struct lwp *l;
2147 
2148 	KASSERT(mutex_owned(proc_lock));
2149 	KASSERT(mutex_owned(p->p_lock));
2150 
2151 	memset(ep, 0, sizeof(*ep));
2152 
2153 	ep->e_paddr = p;
2154 	ep->e_sess = p->p_session;
2155 	if (p->p_cred) {
2156 		kauth_cred_topcred(p->p_cred, &ep->e_pcred);
2157 		kauth_cred_toucred(p->p_cred, &ep->e_ucred);
2158 	}
2159 	if (p->p_stat != SIDL && !P_ZOMBIE(p) && !zombie) {
2160 		struct vmspace *vm = p->p_vmspace;
2161 
2162 		ep->e_vm.vm_rssize = vm_resident_count(vm);
2163 		ep->e_vm.vm_tsize = vm->vm_tsize;
2164 		ep->e_vm.vm_dsize = vm->vm_dsize;
2165 		ep->e_vm.vm_ssize = vm->vm_ssize;
2166 		ep->e_vm.vm_map.size = vm->vm_map.size;
2167 
2168 		/* Pick the primary (first) LWP */
2169 		l = proc_active_lwp(p);
2170 		KASSERT(l != NULL);
2171 		lwp_lock(l);
2172 		if (l->l_wchan)
2173 			strncpy(ep->e_wmesg, l->l_wmesg, WMESGLEN);
2174 		lwp_unlock(l);
2175 	}
2176 	if (p->p_pptr)
2177 		ep->e_ppid = p->p_pptr->p_pid;
2178 	if (p->p_pgrp && p->p_session) {
2179 		ep->e_pgid = p->p_pgrp->pg_id;
2180 		ep->e_jobc = p->p_pgrp->pg_jobc;
2181 		ep->e_sid = p->p_session->s_sid;
2182 		if ((p->p_lflag & PL_CONTROLT) &&
2183 		    (tp = ep->e_sess->s_ttyp)) {
2184 			ep->e_tdev = tp->t_dev;
2185 			ep->e_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PGID;
2186 			ep->e_tsess = tp->t_session;
2187 		} else
2188 			ep->e_tdev = (uint32_t)NODEV;
2189 		ep->e_flag = ep->e_sess->s_ttyvp ? EPROC_CTTY : 0;
2190 		if (SESS_LEADER(p))
2191 			ep->e_flag |= EPROC_SLEADER;
2192 		strncpy(ep->e_login, ep->e_sess->s_login, MAXLOGNAME);
2193 	}
2194 	ep->e_xsize = ep->e_xrssize = 0;
2195 	ep->e_xccount = ep->e_xswrss = 0;
2196 }
2197 
2198 /*
2199  * Fill in a kinfo_proc2 structure for the specified process.
2200  */
2201 static void
2202 fill_kproc2(struct proc *p, struct kinfo_proc2 *ki, bool zombie)
2203 {
2204 	struct tty *tp;
2205 	struct lwp *l, *l2;
2206 	struct timeval ut, st, rt;
2207 	sigset_t ss1, ss2;
2208 	struct rusage ru;
2209 	struct vmspace *vm;
2210 
2211 	KASSERT(mutex_owned(proc_lock));
2212 	KASSERT(mutex_owned(p->p_lock));
2213 
2214 	sigemptyset(&ss1);
2215 	sigemptyset(&ss2);
2216 	memset(ki, 0, sizeof(*ki));
2217 
2218 	ki->p_paddr = PTRTOUINT64(p);
2219 	ki->p_fd = PTRTOUINT64(p->p_fd);
2220 	ki->p_cwdi = PTRTOUINT64(p->p_cwdi);
2221 	ki->p_stats = PTRTOUINT64(p->p_stats);
2222 	ki->p_limit = PTRTOUINT64(p->p_limit);
2223 	ki->p_vmspace = PTRTOUINT64(p->p_vmspace);
2224 	ki->p_sigacts = PTRTOUINT64(p->p_sigacts);
2225 	ki->p_sess = PTRTOUINT64(p->p_session);
2226 	ki->p_tsess = 0;	/* may be changed if controlling tty below */
2227 	ki->p_ru = PTRTOUINT64(&p->p_stats->p_ru);
2228 	ki->p_eflag = 0;
2229 	ki->p_exitsig = p->p_exitsig;
2230 	ki->p_flag = L_INMEM;   /* Process never swapped out */
2231 	ki->p_flag |= sysctl_map_flags(sysctl_flagmap, p->p_flag);
2232 	ki->p_flag |= sysctl_map_flags(sysctl_sflagmap, p->p_sflag);
2233 	ki->p_flag |= sysctl_map_flags(sysctl_slflagmap, p->p_slflag);
2234 	ki->p_flag |= sysctl_map_flags(sysctl_lflagmap, p->p_lflag);
2235 	ki->p_flag |= sysctl_map_flags(sysctl_stflagmap, p->p_stflag);
2236 	ki->p_pid = p->p_pid;
2237 	if (p->p_pptr)
2238 		ki->p_ppid = p->p_pptr->p_pid;
2239 	else
2240 		ki->p_ppid = 0;
2241 	ki->p_uid = kauth_cred_geteuid(p->p_cred);
2242 	ki->p_ruid = kauth_cred_getuid(p->p_cred);
2243 	ki->p_gid = kauth_cred_getegid(p->p_cred);
2244 	ki->p_rgid = kauth_cred_getgid(p->p_cred);
2245 	ki->p_svuid = kauth_cred_getsvuid(p->p_cred);
2246 	ki->p_svgid = kauth_cred_getsvgid(p->p_cred);
2247 	ki->p_ngroups = kauth_cred_ngroups(p->p_cred);
2248 	kauth_cred_getgroups(p->p_cred, ki->p_groups,
2249 	    min(ki->p_ngroups, sizeof(ki->p_groups) / sizeof(ki->p_groups[0])),
2250 	    UIO_SYSSPACE);
2251 
2252 	ki->p_uticks = p->p_uticks;
2253 	ki->p_sticks = p->p_sticks;
2254 	ki->p_iticks = p->p_iticks;
2255 	ki->p_tpgid = NO_PGID;	/* may be changed if controlling tty below */
2256 	ki->p_tracep = PTRTOUINT64(p->p_tracep);
2257 	ki->p_traceflag = p->p_traceflag;
2258 
2259 	memcpy(&ki->p_sigignore, &p->p_sigctx.ps_sigignore,sizeof(ki_sigset_t));
2260 	memcpy(&ki->p_sigcatch, &p->p_sigctx.ps_sigcatch, sizeof(ki_sigset_t));
2261 
2262 	ki->p_cpticks = 0;
2263 	ki->p_pctcpu = p->p_pctcpu;
2264 	ki->p_estcpu = 0;
2265 	ki->p_stat = p->p_stat; /* Will likely be overridden by LWP status */
2266 	ki->p_realstat = p->p_stat;
2267 	ki->p_nice = p->p_nice;
2268 	ki->p_xstat = p->p_xstat;
2269 	ki->p_acflag = p->p_acflag;
2270 
2271 	strncpy(ki->p_comm, p->p_comm,
2272 	    min(sizeof(ki->p_comm), sizeof(p->p_comm)));
2273 	strncpy(ki->p_ename, p->p_emul->e_name, sizeof(ki->p_ename));
2274 
2275 	ki->p_nlwps = p->p_nlwps;
2276 	ki->p_realflag = ki->p_flag;
2277 
2278 	if (p->p_stat != SIDL && !P_ZOMBIE(p) && !zombie) {
2279 		vm = p->p_vmspace;
2280 		ki->p_vm_rssize = vm_resident_count(vm);
2281 		ki->p_vm_tsize = vm->vm_tsize;
2282 		ki->p_vm_dsize = vm->vm_dsize;
2283 		ki->p_vm_ssize = vm->vm_ssize;
2284 		ki->p_vm_vsize = atop(vm->vm_map.size);
2285 		/*
2286 		 * Since the stack is initially mapped mostly with
2287 		 * PROT_NONE and grown as needed, adjust the "mapped size"
2288 		 * to skip the unused stack portion.
2289 		 */
2290 		ki->p_vm_msize =
2291 		    atop(vm->vm_map.size) - vm->vm_issize + vm->vm_ssize;
2292 
2293 		/* Pick the primary (first) LWP */
2294 		l = proc_active_lwp(p);
2295 		KASSERT(l != NULL);
2296 		lwp_lock(l);
2297 		ki->p_nrlwps = p->p_nrlwps;
2298 		ki->p_forw = 0;
2299 		ki->p_back = 0;
2300 		ki->p_addr = PTRTOUINT64(l->l_addr);
2301 		ki->p_stat = l->l_stat;
2302 		ki->p_flag |= sysctl_map_flags(sysctl_lwpflagmap, l->l_flag);
2303 		ki->p_swtime = l->l_swtime;
2304 		ki->p_slptime = l->l_slptime;
2305 		if (l->l_stat == LSONPROC)
2306 			ki->p_schedflags = l->l_cpu->ci_schedstate.spc_flags;
2307 		else
2308 			ki->p_schedflags = 0;
2309 		ki->p_priority = lwp_eprio(l);
2310 		ki->p_usrpri = l->l_priority;
2311 		if (l->l_wchan)
2312 			strncpy(ki->p_wmesg, l->l_wmesg, sizeof(ki->p_wmesg));
2313 		ki->p_wchan = PTRTOUINT64(l->l_wchan);
2314 		ki->p_cpuid = cpu_index(l->l_cpu);
2315 		lwp_unlock(l);
2316 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
2317 			/* This is hardly correct, but... */
2318 			sigplusset(&l->l_sigpend.sp_set, &ss1);
2319 			sigplusset(&l->l_sigmask, &ss2);
2320 			ki->p_cpticks += l->l_cpticks;
2321 			ki->p_pctcpu += l->l_pctcpu;
2322 			ki->p_estcpu += l->l_estcpu;
2323 		}
2324 	}
2325 	sigplusset(&p->p_sigpend.sp_set, &ss2);
2326 	memcpy(&ki->p_siglist, &ss1, sizeof(ki_sigset_t));
2327 	memcpy(&ki->p_sigmask, &ss2, sizeof(ki_sigset_t));
2328 
2329 	if (p->p_session != NULL) {
2330 		ki->p_sid = p->p_session->s_sid;
2331 		ki->p__pgid = p->p_pgrp->pg_id;
2332 		if (p->p_session->s_ttyvp)
2333 			ki->p_eflag |= EPROC_CTTY;
2334 		if (SESS_LEADER(p))
2335 			ki->p_eflag |= EPROC_SLEADER;
2336 		strncpy(ki->p_login, p->p_session->s_login,
2337 		    min(sizeof ki->p_login - 1, sizeof p->p_session->s_login));
2338 		ki->p_jobc = p->p_pgrp->pg_jobc;
2339 		if ((p->p_lflag & PL_CONTROLT) && (tp = p->p_session->s_ttyp)) {
2340 			ki->p_tdev = tp->t_dev;
2341 			ki->p_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PGID;
2342 			ki->p_tsess = PTRTOUINT64(tp->t_session);
2343 		} else {
2344 			ki->p_tdev = (int32_t)NODEV;
2345 		}
2346 	}
2347 
2348 	if (!P_ZOMBIE(p) && !zombie) {
2349 		ki->p_uvalid = 1;
2350 		ki->p_ustart_sec = p->p_stats->p_start.tv_sec;
2351 		ki->p_ustart_usec = p->p_stats->p_start.tv_usec;
2352 
2353 		calcru(p, &ut, &st, NULL, &rt);
2354 		ki->p_rtime_sec = rt.tv_sec;
2355 		ki->p_rtime_usec = rt.tv_usec;
2356 		ki->p_uutime_sec = ut.tv_sec;
2357 		ki->p_uutime_usec = ut.tv_usec;
2358 		ki->p_ustime_sec = st.tv_sec;
2359 		ki->p_ustime_usec = st.tv_usec;
2360 
2361 		memcpy(&ru, &p->p_stats->p_ru, sizeof(ru));
2362 		ki->p_uru_nvcsw = 0;
2363 		ki->p_uru_nivcsw = 0;
2364 		LIST_FOREACH(l2, &p->p_lwps, l_sibling) {
2365 			ki->p_uru_nvcsw += (l2->l_ncsw - l2->l_nivcsw);
2366 			ki->p_uru_nivcsw += l2->l_nivcsw;
2367 			ruadd(&ru, &l2->l_ru);
2368 		}
2369 		ki->p_uru_maxrss = ru.ru_maxrss;
2370 		ki->p_uru_ixrss = ru.ru_ixrss;
2371 		ki->p_uru_idrss = ru.ru_idrss;
2372 		ki->p_uru_isrss = ru.ru_isrss;
2373 		ki->p_uru_minflt = ru.ru_minflt;
2374 		ki->p_uru_majflt = ru.ru_majflt;
2375 		ki->p_uru_nswap = ru.ru_nswap;
2376 		ki->p_uru_inblock = ru.ru_inblock;
2377 		ki->p_uru_oublock = ru.ru_oublock;
2378 		ki->p_uru_msgsnd = ru.ru_msgsnd;
2379 		ki->p_uru_msgrcv = ru.ru_msgrcv;
2380 		ki->p_uru_nsignals = ru.ru_nsignals;
2381 
2382 		timeradd(&p->p_stats->p_cru.ru_utime,
2383 			 &p->p_stats->p_cru.ru_stime, &ut);
2384 		ki->p_uctime_sec = ut.tv_sec;
2385 		ki->p_uctime_usec = ut.tv_usec;
2386 	}
2387 }
2388