xref: /netbsd-src/sys/kern/kern_proc.c (revision 33881f779a77dce6440bdc44610d94de75bebefe)
1 /*	$NetBSD: kern_proc.c,v 1.242 2020/02/23 22:14:03 ad Exp $	*/
2 
3 /*-
4  * Copyright (c) 1999, 2006, 2007, 2008, 2020 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 1.242 2020/02/23 22:14:03 ad 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 #include "opt_kaslr.h"
73 #endif
74 
75 #if defined(__HAVE_COMPAT_NETBSD32) && !defined(COMPAT_NETBSD32) \
76     && !defined(_RUMPKERNEL)
77 #define COMPAT_NETBSD32
78 #endif
79 
80 #include <sys/param.h>
81 #include <sys/systm.h>
82 #include <sys/kernel.h>
83 #include <sys/proc.h>
84 #include <sys/resourcevar.h>
85 #include <sys/buf.h>
86 #include <sys/acct.h>
87 #include <sys/wait.h>
88 #include <sys/file.h>
89 #include <ufs/ufs/quota.h>
90 #include <sys/uio.h>
91 #include <sys/pool.h>
92 #include <sys/pset.h>
93 #include <sys/ioctl.h>
94 #include <sys/tty.h>
95 #include <sys/signalvar.h>
96 #include <sys/ras.h>
97 #include <sys/filedesc.h>
98 #include <sys/syscall_stats.h>
99 #include <sys/kauth.h>
100 #include <sys/sleepq.h>
101 #include <sys/atomic.h>
102 #include <sys/kmem.h>
103 #include <sys/namei.h>
104 #include <sys/dtrace_bsd.h>
105 #include <sys/sysctl.h>
106 #include <sys/exec.h>
107 #include <sys/cpu.h>
108 #include <sys/compat_stub.h>
109 #include <sys/vnode.h>
110 
111 #include <uvm/uvm_extern.h>
112 #include <uvm/uvm.h>
113 
114 /*
115  * Process lists.
116  */
117 
118 struct proclist		allproc		__cacheline_aligned;
119 struct proclist		zombproc	__cacheline_aligned;
120 
121 kmutex_t *		proc_lock	__cacheline_aligned;
122 
123 /*
124  * pid to proc lookup is done by indexing the pid_table array.
125  * Since pid numbers are only allocated when an empty slot
126  * has been found, there is no need to search any lists ever.
127  * (an orphaned pgrp will lock the slot, a session will lock
128  * the pgrp with the same number.)
129  * If the table is too small it is reallocated with twice the
130  * previous size and the entries 'unzipped' into the two halves.
131  * A linked list of free entries is passed through the pt_proc
132  * field of 'free' items - set odd to be an invalid ptr.
133  */
134 
135 struct pid_table {
136 	struct proc	*pt_proc;
137 	struct pgrp	*pt_pgrp;
138 	pid_t		pt_pid;
139 };
140 #if 1	/* strongly typed cast - should be a noop */
141 static inline uint p2u(struct proc *p) { return (uint)(uintptr_t)p; }
142 #else
143 #define p2u(p) ((uint)p)
144 #endif
145 #define P_VALID(p) (!(p2u(p) & 1))
146 #define P_NEXT(p) (p2u(p) >> 1)
147 #define P_FREE(pid) ((struct proc *)(uintptr_t)((pid) << 1 | 1))
148 
149 /*
150  * Table of process IDs (PIDs).
151  */
152 static struct pid_table *pid_table	__read_mostly;
153 
154 #define	INITIAL_PID_TABLE_SIZE		(1 << 5)
155 
156 /* Table mask, threshold for growing and number of allocated PIDs. */
157 static u_int		pid_tbl_mask	__read_mostly;
158 static u_int		pid_alloc_lim	__read_mostly;
159 static u_int		pid_alloc_cnt	__cacheline_aligned;
160 
161 /* Next free, last free and maximum PIDs. */
162 static u_int		next_free_pt	__cacheline_aligned;
163 static u_int		last_free_pt	__cacheline_aligned;
164 static pid_t		pid_max		__read_mostly;
165 
166 /* Components of the first process -- never freed. */
167 
168 extern struct emul emul_netbsd;	/* defined in kern_exec.c */
169 
170 struct session session0 = {
171 	.s_count = 1,
172 	.s_sid = 0,
173 };
174 struct pgrp pgrp0 = {
175 	.pg_members = LIST_HEAD_INITIALIZER(&pgrp0.pg_members),
176 	.pg_session = &session0,
177 };
178 filedesc_t filedesc0;
179 struct cwdinfo cwdi0 = {
180 	.cwdi_cmask = CMASK,
181 	.cwdi_refcnt = 1,
182 };
183 struct plimit limit0;
184 struct pstats pstat0;
185 struct vmspace vmspace0;
186 struct sigacts sigacts0;
187 struct proc proc0 = {
188 	.p_lwps = LIST_HEAD_INITIALIZER(&proc0.p_lwps),
189 	.p_sigwaiters = LIST_HEAD_INITIALIZER(&proc0.p_sigwaiters),
190 	.p_nlwps = 1,
191 	.p_nrlwps = 1,
192 	.p_nlwpid = 1,		/* must match lwp0.l_lid */
193 	.p_pgrp = &pgrp0,
194 	.p_comm = "system",
195 	/*
196 	 * Set P_NOCLDWAIT so that kernel threads are reparented to init(8)
197 	 * when they exit.  init(8) can easily wait them out for us.
198 	 */
199 	.p_flag = PK_SYSTEM | PK_NOCLDWAIT,
200 	.p_stat = SACTIVE,
201 	.p_nice = NZERO,
202 	.p_emul = &emul_netbsd,
203 	.p_cwdi = &cwdi0,
204 	.p_limit = &limit0,
205 	.p_fd = &filedesc0,
206 	.p_vmspace = &vmspace0,
207 	.p_stats = &pstat0,
208 	.p_sigacts = &sigacts0,
209 #ifdef PROC0_MD_INITIALIZERS
210 	PROC0_MD_INITIALIZERS
211 #endif
212 };
213 kauth_cred_t cred0;
214 
215 static const int	nofile	= NOFILE;
216 static const int	maxuprc	= MAXUPRC;
217 
218 static int sysctl_doeproc(SYSCTLFN_PROTO);
219 static int sysctl_kern_proc_args(SYSCTLFN_PROTO);
220 static int sysctl_security_expose_address(SYSCTLFN_PROTO);
221 
222 #ifdef KASLR
223 static int kern_expose_address = 0;
224 #else
225 static int kern_expose_address = 1;
226 #endif
227 /*
228  * The process list descriptors, used during pid allocation and
229  * by sysctl.  No locking on this data structure is needed since
230  * it is completely static.
231  */
232 const struct proclist_desc proclists[] = {
233 	{ &allproc	},
234 	{ &zombproc	},
235 	{ NULL		},
236 };
237 
238 static struct pgrp *	pg_remove(pid_t);
239 static void		pg_delete(pid_t);
240 static void		orphanpg(struct pgrp *);
241 
242 static specificdata_domain_t proc_specificdata_domain;
243 
244 static pool_cache_t proc_cache;
245 
246 static kauth_listener_t proc_listener;
247 
248 static void fill_proc(const struct proc *, struct proc *, bool);
249 static int fill_pathname(struct lwp *, pid_t, void *, size_t *);
250 static int fill_cwd(struct lwp *, pid_t, void *, size_t *);
251 
252 static int
253 proc_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
254     void *arg0, void *arg1, void *arg2, void *arg3)
255 {
256 	struct proc *p;
257 	int result;
258 
259 	result = KAUTH_RESULT_DEFER;
260 	p = arg0;
261 
262 	switch (action) {
263 	case KAUTH_PROCESS_CANSEE: {
264 		enum kauth_process_req req;
265 
266 		req = (enum kauth_process_req)(uintptr_t)arg1;
267 
268 		switch (req) {
269 		case KAUTH_REQ_PROCESS_CANSEE_ARGS:
270 		case KAUTH_REQ_PROCESS_CANSEE_ENTRY:
271 		case KAUTH_REQ_PROCESS_CANSEE_OPENFILES:
272 		case KAUTH_REQ_PROCESS_CANSEE_EPROC:
273 			result = KAUTH_RESULT_ALLOW;
274 			break;
275 
276 		case KAUTH_REQ_PROCESS_CANSEE_ENV:
277 			if (kauth_cred_getuid(cred) !=
278 			    kauth_cred_getuid(p->p_cred) ||
279 			    kauth_cred_getuid(cred) !=
280 			    kauth_cred_getsvuid(p->p_cred))
281 				break;
282 
283 			result = KAUTH_RESULT_ALLOW;
284 
285 			break;
286 
287 		case KAUTH_REQ_PROCESS_CANSEE_KPTR:
288 			if (!kern_expose_address)
289 				break;
290 
291 			if (kern_expose_address == 1 && !(p->p_flag & PK_KMEM))
292 				break;
293 
294 			result = KAUTH_RESULT_ALLOW;
295 
296 			break;
297 
298 		default:
299 			break;
300 		}
301 
302 		break;
303 		}
304 
305 	case KAUTH_PROCESS_FORK: {
306 		int lnprocs = (int)(unsigned long)arg2;
307 
308 		/*
309 		 * Don't allow a nonprivileged user to use the last few
310 		 * processes. The variable lnprocs is the current number of
311 		 * processes, maxproc is the limit.
312 		 */
313 		if (__predict_false((lnprocs >= maxproc - 5)))
314 			break;
315 
316 		result = KAUTH_RESULT_ALLOW;
317 
318 		break;
319 		}
320 
321 	case KAUTH_PROCESS_CORENAME:
322 	case KAUTH_PROCESS_STOPFLAG:
323 		if (proc_uidmatch(cred, p->p_cred) == 0)
324 			result = KAUTH_RESULT_ALLOW;
325 
326 		break;
327 
328 	default:
329 		break;
330 	}
331 
332 	return result;
333 }
334 
335 static int
336 proc_ctor(void *arg __unused, void *obj, int flags __unused)
337 {
338 	memset(obj, 0, sizeof(struct proc));
339 	return 0;
340 }
341 
342 /*
343  * Initialize global process hashing structures.
344  */
345 void
346 procinit(void)
347 {
348 	const struct proclist_desc *pd;
349 	u_int i;
350 #define	LINK_EMPTY ((PID_MAX + INITIAL_PID_TABLE_SIZE) & ~(INITIAL_PID_TABLE_SIZE - 1))
351 
352 	for (pd = proclists; pd->pd_list != NULL; pd++)
353 		LIST_INIT(pd->pd_list);
354 
355 	proc_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
356 	pid_table = kmem_alloc(INITIAL_PID_TABLE_SIZE
357 	    * sizeof(struct pid_table), KM_SLEEP);
358 	pid_tbl_mask = INITIAL_PID_TABLE_SIZE - 1;
359 	pid_max = PID_MAX;
360 
361 	/* Set free list running through table...
362 	   Preset 'use count' above PID_MAX so we allocate pid 1 next. */
363 	for (i = 0; i <= pid_tbl_mask; i++) {
364 		pid_table[i].pt_proc = P_FREE(LINK_EMPTY + i + 1);
365 		pid_table[i].pt_pgrp = 0;
366 		pid_table[i].pt_pid = 0;
367 	}
368 	/* slot 0 is just grabbed */
369 	next_free_pt = 1;
370 	/* Need to fix last entry. */
371 	last_free_pt = pid_tbl_mask;
372 	pid_table[last_free_pt].pt_proc = P_FREE(LINK_EMPTY);
373 	/* point at which we grow table - to avoid reusing pids too often */
374 	pid_alloc_lim = pid_tbl_mask - 1;
375 #undef LINK_EMPTY
376 
377 	proc_specificdata_domain = specificdata_domain_create();
378 	KASSERT(proc_specificdata_domain != NULL);
379 
380 	proc_cache = pool_cache_init(sizeof(struct proc), coherency_unit, 0, 0,
381 	    "procpl", NULL, IPL_NONE, proc_ctor, NULL, NULL);
382 
383 	proc_listener = kauth_listen_scope(KAUTH_SCOPE_PROCESS,
384 	    proc_listener_cb, NULL);
385 }
386 
387 void
388 procinit_sysctl(void)
389 {
390 	static struct sysctllog *clog;
391 
392 	sysctl_createv(&clog, 0, NULL, NULL,
393 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
394 		       CTLTYPE_INT, "expose_address",
395 		       SYSCTL_DESCR("Enable exposing kernel addresses"),
396 		       sysctl_security_expose_address, 0,
397 		       &kern_expose_address, 0, CTL_KERN, CTL_CREATE, CTL_EOL);
398 	sysctl_createv(&clog, 0, NULL, NULL,
399 		       CTLFLAG_PERMANENT,
400 		       CTLTYPE_NODE, "proc",
401 		       SYSCTL_DESCR("System-wide process information"),
402 		       sysctl_doeproc, 0, NULL, 0,
403 		       CTL_KERN, KERN_PROC, CTL_EOL);
404 	sysctl_createv(&clog, 0, NULL, NULL,
405 		       CTLFLAG_PERMANENT,
406 		       CTLTYPE_NODE, "proc2",
407 		       SYSCTL_DESCR("Machine-independent process information"),
408 		       sysctl_doeproc, 0, NULL, 0,
409 		       CTL_KERN, KERN_PROC2, CTL_EOL);
410 	sysctl_createv(&clog, 0, NULL, NULL,
411 		       CTLFLAG_PERMANENT,
412 		       CTLTYPE_NODE, "proc_args",
413 		       SYSCTL_DESCR("Process argument information"),
414 		       sysctl_kern_proc_args, 0, NULL, 0,
415 		       CTL_KERN, KERN_PROC_ARGS, CTL_EOL);
416 
417 	/*
418 	  "nodes" under these:
419 
420 	  KERN_PROC_ALL
421 	  KERN_PROC_PID pid
422 	  KERN_PROC_PGRP pgrp
423 	  KERN_PROC_SESSION sess
424 	  KERN_PROC_TTY tty
425 	  KERN_PROC_UID uid
426 	  KERN_PROC_RUID uid
427 	  KERN_PROC_GID gid
428 	  KERN_PROC_RGID gid
429 
430 	  all in all, probably not worth the effort...
431 	*/
432 }
433 
434 /*
435  * Initialize process 0.
436  */
437 void
438 proc0_init(void)
439 {
440 	struct proc *p;
441 	struct pgrp *pg;
442 	struct rlimit *rlim;
443 	rlim_t lim;
444 	int error __diagused;
445 	int i;
446 
447 	p = &proc0;
448 	pg = &pgrp0;
449 
450 	mutex_init(&p->p_stmutex, MUTEX_DEFAULT, IPL_HIGH);
451 	mutex_init(&p->p_auxlock, MUTEX_DEFAULT, IPL_NONE);
452 	p->p_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
453 
454 	rw_init(&p->p_reflock);
455 	rw_init(&p->p_treelock);
456 	cv_init(&p->p_waitcv, "wait");
457 	cv_init(&p->p_lwpcv, "lwpwait");
458 
459 	LIST_INSERT_HEAD(&p->p_lwps, &lwp0, l_sibling);
460 	radix_tree_init_tree(&p->p_lwptree);
461 	error = radix_tree_insert_node(&p->p_lwptree,
462 	    (uint64_t)(lwp0.l_lid - 1), &lwp0);
463 	KASSERT(error == 0);
464 
465 	pid_table[0].pt_proc = p;
466 	LIST_INSERT_HEAD(&allproc, p, p_list);
467 
468 	pid_table[0].pt_pgrp = pg;
469 	LIST_INSERT_HEAD(&pg->pg_members, p, p_pglist);
470 
471 #ifdef __HAVE_SYSCALL_INTERN
472 	(*p->p_emul->e_syscall_intern)(p);
473 #endif
474 
475 	/* Create credentials. */
476 	cred0 = kauth_cred_alloc();
477 	p->p_cred = cred0;
478 
479 	/* Create the CWD info. */
480 	mutex_init(&cwdi0.cwdi_lock, MUTEX_DEFAULT, IPL_NONE);
481 
482 	/* Create the limits structures. */
483 	mutex_init(&limit0.pl_lock, MUTEX_DEFAULT, IPL_NONE);
484 
485 	rlim = limit0.pl_rlimit;
486 	for (i = 0; i < __arraycount(limit0.pl_rlimit); i++) {
487 		rlim[i].rlim_cur = RLIM_INFINITY;
488 		rlim[i].rlim_max = RLIM_INFINITY;
489 	}
490 
491 	rlim[RLIMIT_NOFILE].rlim_max = maxfiles;
492 	rlim[RLIMIT_NOFILE].rlim_cur = maxfiles < nofile ? maxfiles : nofile;
493 
494 	rlim[RLIMIT_NPROC].rlim_max = maxproc;
495 	rlim[RLIMIT_NPROC].rlim_cur = maxproc < maxuprc ? maxproc : maxuprc;
496 
497 	lim = MIN(VM_MAXUSER_ADDRESS, ctob((rlim_t)uvm_availmem()));
498 	rlim[RLIMIT_RSS].rlim_max = lim;
499 	rlim[RLIMIT_MEMLOCK].rlim_max = lim;
500 	rlim[RLIMIT_MEMLOCK].rlim_cur = lim / 3;
501 
502 	rlim[RLIMIT_NTHR].rlim_max = maxlwp;
503 	rlim[RLIMIT_NTHR].rlim_cur = maxlwp < maxuprc ? maxlwp : maxuprc;
504 
505 	/* Note that default core name has zero length. */
506 	limit0.pl_corename = defcorename;
507 	limit0.pl_cnlen = 0;
508 	limit0.pl_refcnt = 1;
509 	limit0.pl_writeable = false;
510 	limit0.pl_sv_limit = NULL;
511 
512 	/* Configure virtual memory system, set vm rlimits. */
513 	uvm_init_limits(p);
514 
515 	/* Initialize file descriptor table for proc0. */
516 	fd_init(&filedesc0);
517 
518 	/*
519 	 * Initialize proc0's vmspace, which uses the kernel pmap.
520 	 * All kernel processes (which never have user space mappings)
521 	 * share proc0's vmspace, and thus, the kernel pmap.
522 	 */
523 	uvmspace_init(&vmspace0, pmap_kernel(), round_page(VM_MIN_ADDRESS),
524 	    trunc_page(VM_MAXUSER_ADDRESS),
525 #ifdef __USE_TOPDOWN_VM
526 	    true
527 #else
528 	    false
529 #endif
530 	    );
531 
532 	/* Initialize signal state for proc0. XXX IPL_SCHED */
533 	mutex_init(&p->p_sigacts->sa_mutex, MUTEX_DEFAULT, IPL_SCHED);
534 	siginit(p);
535 
536 	proc_initspecific(p);
537 	kdtrace_proc_ctor(NULL, p);
538 }
539 
540 /*
541  * Session reference counting.
542  */
543 
544 void
545 proc_sesshold(struct session *ss)
546 {
547 
548 	KASSERT(mutex_owned(proc_lock));
549 	ss->s_count++;
550 }
551 
552 void
553 proc_sessrele(struct session *ss)
554 {
555 
556 	KASSERT(mutex_owned(proc_lock));
557 	/*
558 	 * We keep the pgrp with the same id as the session in order to
559 	 * stop a process being given the same pid.  Since the pgrp holds
560 	 * a reference to the session, it must be a 'zombie' pgrp by now.
561 	 */
562 	if (--ss->s_count == 0) {
563 		struct pgrp *pg;
564 
565 		pg = pg_remove(ss->s_sid);
566 		mutex_exit(proc_lock);
567 
568 		kmem_free(pg, sizeof(struct pgrp));
569 		kmem_free(ss, sizeof(struct session));
570 	} else {
571 		mutex_exit(proc_lock);
572 	}
573 }
574 
575 /*
576  * Check that the specified process group is in the session of the
577  * specified process.
578  * Treats -ve ids as process ids.
579  * Used to validate TIOCSPGRP requests.
580  */
581 int
582 pgid_in_session(struct proc *p, pid_t pg_id)
583 {
584 	struct pgrp *pgrp;
585 	struct session *session;
586 	int error;
587 
588 	mutex_enter(proc_lock);
589 	if (pg_id < 0) {
590 		struct proc *p1 = proc_find(-pg_id);
591 		if (p1 == NULL) {
592 			error = EINVAL;
593 			goto fail;
594 		}
595 		pgrp = p1->p_pgrp;
596 	} else {
597 		pgrp = pgrp_find(pg_id);
598 		if (pgrp == NULL) {
599 			error = EINVAL;
600 			goto fail;
601 		}
602 	}
603 	session = pgrp->pg_session;
604 	error = (session != p->p_pgrp->pg_session) ? EPERM : 0;
605 fail:
606 	mutex_exit(proc_lock);
607 	return error;
608 }
609 
610 /*
611  * p_inferior: is p an inferior of q?
612  */
613 static inline bool
614 p_inferior(struct proc *p, struct proc *q)
615 {
616 
617 	KASSERT(mutex_owned(proc_lock));
618 
619 	for (; p != q; p = p->p_pptr)
620 		if (p->p_pid == 0)
621 			return false;
622 	return true;
623 }
624 
625 /*
626  * proc_find: locate a process by the ID.
627  *
628  * => Must be called with proc_lock held.
629  */
630 proc_t *
631 proc_find_raw(pid_t pid)
632 {
633 	struct pid_table *pt;
634 	proc_t *p;
635 
636 	KASSERT(mutex_owned(proc_lock));
637 	pt = &pid_table[pid & pid_tbl_mask];
638 	p = pt->pt_proc;
639 	if (__predict_false(!P_VALID(p) || pt->pt_pid != pid)) {
640 		return NULL;
641 	}
642 	return p;
643 }
644 
645 proc_t *
646 proc_find(pid_t pid)
647 {
648 	proc_t *p;
649 
650 	p = proc_find_raw(pid);
651 	if (__predict_false(p == NULL)) {
652 		return NULL;
653 	}
654 
655 	/*
656 	 * Only allow live processes to be found by PID.
657 	 * XXX: p_stat might change, since unlocked.
658 	 */
659 	if (__predict_true(p->p_stat == SACTIVE || p->p_stat == SSTOP)) {
660 		return p;
661 	}
662 	return NULL;
663 }
664 
665 /*
666  * pgrp_find: locate a process group by the ID.
667  *
668  * => Must be called with proc_lock held.
669  */
670 struct pgrp *
671 pgrp_find(pid_t pgid)
672 {
673 	struct pgrp *pg;
674 
675 	KASSERT(mutex_owned(proc_lock));
676 
677 	pg = pid_table[pgid & pid_tbl_mask].pt_pgrp;
678 
679 	/*
680 	 * Cannot look up a process group that only exists because the
681 	 * session has not died yet (traditional).
682 	 */
683 	if (pg == NULL || pg->pg_id != pgid || LIST_EMPTY(&pg->pg_members)) {
684 		return NULL;
685 	}
686 	return pg;
687 }
688 
689 static void
690 expand_pid_table(void)
691 {
692 	size_t pt_size, tsz;
693 	struct pid_table *n_pt, *new_pt;
694 	struct proc *proc;
695 	struct pgrp *pgrp;
696 	pid_t pid, rpid;
697 	u_int i;
698 	uint new_pt_mask;
699 
700 	pt_size = pid_tbl_mask + 1;
701 	tsz = pt_size * 2 * sizeof(struct pid_table);
702 	new_pt = kmem_alloc(tsz, KM_SLEEP);
703 	new_pt_mask = pt_size * 2 - 1;
704 
705 	mutex_enter(proc_lock);
706 	if (pt_size != pid_tbl_mask + 1) {
707 		/* Another process beat us to it... */
708 		mutex_exit(proc_lock);
709 		kmem_free(new_pt, tsz);
710 		return;
711 	}
712 
713 	/*
714 	 * Copy entries from old table into new one.
715 	 * If 'pid' is 'odd' we need to place in the upper half,
716 	 * even pid's to the lower half.
717 	 * Free items stay in the low half so we don't have to
718 	 * fixup the reference to them.
719 	 * We stuff free items on the front of the freelist
720 	 * because we can't write to unmodified entries.
721 	 * Processing the table backwards maintains a semblance
722 	 * of issuing pid numbers that increase with time.
723 	 */
724 	i = pt_size - 1;
725 	n_pt = new_pt + i;
726 	for (; ; i--, n_pt--) {
727 		proc = pid_table[i].pt_proc;
728 		pgrp = pid_table[i].pt_pgrp;
729 		if (!P_VALID(proc)) {
730 			/* Up 'use count' so that link is valid */
731 			pid = (P_NEXT(proc) + pt_size) & ~pt_size;
732 			rpid = 0;
733 			proc = P_FREE(pid);
734 			if (pgrp)
735 				pid = pgrp->pg_id;
736 		} else {
737 			pid = pid_table[i].pt_pid;
738 			rpid = pid;
739 		}
740 
741 		/* Save entry in appropriate half of table */
742 		n_pt[pid & pt_size].pt_proc = proc;
743 		n_pt[pid & pt_size].pt_pgrp = pgrp;
744 		n_pt[pid & pt_size].pt_pid = rpid;
745 
746 		/* Put other piece on start of free list */
747 		pid = (pid ^ pt_size) & ~pid_tbl_mask;
748 		n_pt[pid & pt_size].pt_proc =
749 			P_FREE((pid & ~pt_size) | next_free_pt);
750 		n_pt[pid & pt_size].pt_pgrp = 0;
751 		n_pt[pid & pt_size].pt_pid = 0;
752 
753 		next_free_pt = i | (pid & pt_size);
754 		if (i == 0)
755 			break;
756 	}
757 
758 	/* Save old table size and switch tables */
759 	tsz = pt_size * sizeof(struct pid_table);
760 	n_pt = pid_table;
761 	pid_table = new_pt;
762 	pid_tbl_mask = new_pt_mask;
763 
764 	/*
765 	 * pid_max starts as PID_MAX (= 30000), once we have 16384
766 	 * allocated pids we need it to be larger!
767 	 */
768 	if (pid_tbl_mask > PID_MAX) {
769 		pid_max = pid_tbl_mask * 2 + 1;
770 		pid_alloc_lim |= pid_alloc_lim << 1;
771 	} else
772 		pid_alloc_lim <<= 1;	/* doubles number of free slots... */
773 
774 	mutex_exit(proc_lock);
775 	kmem_free(n_pt, tsz);
776 }
777 
778 struct proc *
779 proc_alloc(void)
780 {
781 	struct proc *p;
782 
783 	p = pool_cache_get(proc_cache, PR_WAITOK);
784 	p->p_stat = SIDL;			/* protect against others */
785 	proc_initspecific(p);
786 	kdtrace_proc_ctor(NULL, p);
787 	p->p_pid = -1;
788 	proc_alloc_pid(p);
789 	return p;
790 }
791 
792 /*
793  * proc_alloc_pid: allocate PID and record the given proc 'p' so that
794  * proc_find_raw() can find it by the PID.
795  */
796 
797 pid_t
798 proc_alloc_pid(struct proc *p)
799 {
800 	struct pid_table *pt;
801 	pid_t pid;
802 	int nxt;
803 
804 	for (;;expand_pid_table()) {
805 		if (__predict_false(pid_alloc_cnt >= pid_alloc_lim))
806 			/* ensure pids cycle through 2000+ values */
807 			continue;
808 		mutex_enter(proc_lock);
809 		pt = &pid_table[next_free_pt];
810 #ifdef DIAGNOSTIC
811 		if (__predict_false(P_VALID(pt->pt_proc) || pt->pt_pgrp))
812 			panic("proc_alloc: slot busy");
813 #endif
814 		nxt = P_NEXT(pt->pt_proc);
815 		if (nxt & pid_tbl_mask)
816 			break;
817 		/* Table full - expand (NB last entry not used....) */
818 		mutex_exit(proc_lock);
819 	}
820 
821 	/* pid is 'saved use count' + 'size' + entry */
822 	pid = (nxt & ~pid_tbl_mask) + pid_tbl_mask + 1 + next_free_pt;
823 	if ((uint)pid > (uint)pid_max)
824 		pid &= pid_tbl_mask;
825 	next_free_pt = nxt & pid_tbl_mask;
826 
827 	/* Grab table slot */
828 	pt->pt_proc = p;
829 
830 	KASSERT(pt->pt_pid == 0);
831 	pt->pt_pid = pid;
832 	if (p->p_pid == -1) {
833 		p->p_pid = pid;
834 	}
835 	pid_alloc_cnt++;
836 	mutex_exit(proc_lock);
837 
838 	return pid;
839 }
840 
841 /*
842  * Free a process id - called from proc_free (in kern_exit.c)
843  *
844  * Called with the proc_lock held.
845  */
846 void
847 proc_free_pid(pid_t pid)
848 {
849 	struct pid_table *pt;
850 
851 	KASSERT(mutex_owned(proc_lock));
852 
853 	pt = &pid_table[pid & pid_tbl_mask];
854 
855 	/* save pid use count in slot */
856 	pt->pt_proc = P_FREE(pid & ~pid_tbl_mask);
857 	KASSERT(pt->pt_pid == pid);
858 	pt->pt_pid = 0;
859 
860 	if (pt->pt_pgrp == NULL) {
861 		/* link last freed entry onto ours */
862 		pid &= pid_tbl_mask;
863 		pt = &pid_table[last_free_pt];
864 		pt->pt_proc = P_FREE(P_NEXT(pt->pt_proc) | pid);
865 		pt->pt_pid = 0;
866 		last_free_pt = pid;
867 		pid_alloc_cnt--;
868 	}
869 
870 	atomic_dec_uint(&nprocs);
871 }
872 
873 void
874 proc_free_mem(struct proc *p)
875 {
876 
877 	kdtrace_proc_dtor(NULL, p);
878 	pool_cache_put(proc_cache, p);
879 }
880 
881 /*
882  * proc_enterpgrp: move p to a new or existing process group (and session).
883  *
884  * If we are creating a new pgrp, the pgid should equal
885  * the calling process' pid.
886  * If is only valid to enter a process group that is in the session
887  * of the process.
888  * Also mksess should only be set if we are creating a process group
889  *
890  * Only called from sys_setsid, sys_setpgid and posix_spawn/spawn_return.
891  */
892 int
893 proc_enterpgrp(struct proc *curp, pid_t pid, pid_t pgid, bool mksess)
894 {
895 	struct pgrp *new_pgrp, *pgrp;
896 	struct session *sess;
897 	struct proc *p;
898 	int rval;
899 	pid_t pg_id = NO_PGID;
900 
901 	sess = mksess ? kmem_alloc(sizeof(*sess), KM_SLEEP) : NULL;
902 
903 	/* Allocate data areas we might need before doing any validity checks */
904 	mutex_enter(proc_lock);		/* Because pid_table might change */
905 	if (pid_table[pgid & pid_tbl_mask].pt_pgrp == 0) {
906 		mutex_exit(proc_lock);
907 		new_pgrp = kmem_alloc(sizeof(*new_pgrp), KM_SLEEP);
908 		mutex_enter(proc_lock);
909 	} else
910 		new_pgrp = NULL;
911 	rval = EPERM;	/* most common error (to save typing) */
912 
913 	/* Check pgrp exists or can be created */
914 	pgrp = pid_table[pgid & pid_tbl_mask].pt_pgrp;
915 	if (pgrp != NULL && pgrp->pg_id != pgid)
916 		goto done;
917 
918 	/* Can only set another process under restricted circumstances. */
919 	if (pid != curp->p_pid) {
920 		/* Must exist and be one of our children... */
921 		p = proc_find(pid);
922 		if (p == NULL || !p_inferior(p, curp)) {
923 			rval = ESRCH;
924 			goto done;
925 		}
926 		/* ... in the same session... */
927 		if (sess != NULL || p->p_session != curp->p_session)
928 			goto done;
929 		/* ... existing pgid must be in same session ... */
930 		if (pgrp != NULL && pgrp->pg_session != p->p_session)
931 			goto done;
932 		/* ... and not done an exec. */
933 		if (p->p_flag & PK_EXEC) {
934 			rval = EACCES;
935 			goto done;
936 		}
937 	} else {
938 		/* ... setsid() cannot re-enter a pgrp */
939 		if (mksess && (curp->p_pgid == curp->p_pid ||
940 		    pgrp_find(curp->p_pid)))
941 			goto done;
942 		p = curp;
943 	}
944 
945 	/* Changing the process group/session of a session
946 	   leader is definitely off limits. */
947 	if (SESS_LEADER(p)) {
948 		if (sess == NULL && p->p_pgrp == pgrp)
949 			/* unless it's a definite noop */
950 			rval = 0;
951 		goto done;
952 	}
953 
954 	/* Can only create a process group with id of process */
955 	if (pgrp == NULL && pgid != pid)
956 		goto done;
957 
958 	/* Can only create a session if creating pgrp */
959 	if (sess != NULL && pgrp != NULL)
960 		goto done;
961 
962 	/* Check we allocated memory for a pgrp... */
963 	if (pgrp == NULL && new_pgrp == NULL)
964 		goto done;
965 
966 	/* Don't attach to 'zombie' pgrp */
967 	if (pgrp != NULL && LIST_EMPTY(&pgrp->pg_members))
968 		goto done;
969 
970 	/* Expect to succeed now */
971 	rval = 0;
972 
973 	if (pgrp == p->p_pgrp)
974 		/* nothing to do */
975 		goto done;
976 
977 	/* Ok all setup, link up required structures */
978 
979 	if (pgrp == NULL) {
980 		pgrp = new_pgrp;
981 		new_pgrp = NULL;
982 		if (sess != NULL) {
983 			sess->s_sid = p->p_pid;
984 			sess->s_leader = p;
985 			sess->s_count = 1;
986 			sess->s_ttyvp = NULL;
987 			sess->s_ttyp = NULL;
988 			sess->s_flags = p->p_session->s_flags & ~S_LOGIN_SET;
989 			memcpy(sess->s_login, p->p_session->s_login,
990 			    sizeof(sess->s_login));
991 			p->p_lflag &= ~PL_CONTROLT;
992 		} else {
993 			sess = p->p_pgrp->pg_session;
994 			proc_sesshold(sess);
995 		}
996 		pgrp->pg_session = sess;
997 		sess = NULL;
998 
999 		pgrp->pg_id = pgid;
1000 		LIST_INIT(&pgrp->pg_members);
1001 #ifdef DIAGNOSTIC
1002 		if (__predict_false(pid_table[pgid & pid_tbl_mask].pt_pgrp))
1003 			panic("enterpgrp: pgrp table slot in use");
1004 		if (__predict_false(mksess && p != curp))
1005 			panic("enterpgrp: mksession and p != curproc");
1006 #endif
1007 		pid_table[pgid & pid_tbl_mask].pt_pgrp = pgrp;
1008 		pgrp->pg_jobc = 0;
1009 	}
1010 
1011 	/*
1012 	 * Adjust eligibility of affected pgrps to participate in job control.
1013 	 * Increment eligibility counts before decrementing, otherwise we
1014 	 * could reach 0 spuriously during the first call.
1015 	 */
1016 	fixjobc(p, pgrp, 1);
1017 	fixjobc(p, p->p_pgrp, 0);
1018 
1019 	/* Interlock with ttread(). */
1020 	mutex_spin_enter(&tty_lock);
1021 
1022 	/* Move process to requested group. */
1023 	LIST_REMOVE(p, p_pglist);
1024 	if (LIST_EMPTY(&p->p_pgrp->pg_members))
1025 		/* defer delete until we've dumped the lock */
1026 		pg_id = p->p_pgrp->pg_id;
1027 	p->p_pgrp = pgrp;
1028 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
1029 
1030 	/* Done with the swap; we can release the tty mutex. */
1031 	mutex_spin_exit(&tty_lock);
1032 
1033     done:
1034 	if (pg_id != NO_PGID) {
1035 		/* Releases proc_lock. */
1036 		pg_delete(pg_id);
1037 	} else {
1038 		mutex_exit(proc_lock);
1039 	}
1040 	if (sess != NULL)
1041 		kmem_free(sess, sizeof(*sess));
1042 	if (new_pgrp != NULL)
1043 		kmem_free(new_pgrp, sizeof(*new_pgrp));
1044 #ifdef DEBUG_PGRP
1045 	if (__predict_false(rval))
1046 		printf("enterpgrp(%d,%d,%d), curproc %d, rval %d\n",
1047 			pid, pgid, mksess, curp->p_pid, rval);
1048 #endif
1049 	return rval;
1050 }
1051 
1052 /*
1053  * proc_leavepgrp: remove a process from its process group.
1054  *  => must be called with the proc_lock held, which will be released;
1055  */
1056 void
1057 proc_leavepgrp(struct proc *p)
1058 {
1059 	struct pgrp *pgrp;
1060 
1061 	KASSERT(mutex_owned(proc_lock));
1062 
1063 	/* Interlock with ttread() */
1064 	mutex_spin_enter(&tty_lock);
1065 	pgrp = p->p_pgrp;
1066 	LIST_REMOVE(p, p_pglist);
1067 	p->p_pgrp = NULL;
1068 	mutex_spin_exit(&tty_lock);
1069 
1070 	if (LIST_EMPTY(&pgrp->pg_members)) {
1071 		/* Releases proc_lock. */
1072 		pg_delete(pgrp->pg_id);
1073 	} else {
1074 		mutex_exit(proc_lock);
1075 	}
1076 }
1077 
1078 /*
1079  * pg_remove: remove a process group from the table.
1080  *  => must be called with the proc_lock held;
1081  *  => returns process group to free;
1082  */
1083 static struct pgrp *
1084 pg_remove(pid_t pg_id)
1085 {
1086 	struct pgrp *pgrp;
1087 	struct pid_table *pt;
1088 
1089 	KASSERT(mutex_owned(proc_lock));
1090 
1091 	pt = &pid_table[pg_id & pid_tbl_mask];
1092 	pgrp = pt->pt_pgrp;
1093 
1094 	KASSERT(pgrp != NULL);
1095 	KASSERT(pgrp->pg_id == pg_id);
1096 	KASSERT(LIST_EMPTY(&pgrp->pg_members));
1097 
1098 	pt->pt_pgrp = NULL;
1099 
1100 	if (!P_VALID(pt->pt_proc)) {
1101 		/* Orphaned pgrp, put slot onto free list. */
1102 		KASSERT((P_NEXT(pt->pt_proc) & pid_tbl_mask) == 0);
1103 		pg_id &= pid_tbl_mask;
1104 		pt = &pid_table[last_free_pt];
1105 		pt->pt_proc = P_FREE(P_NEXT(pt->pt_proc) | pg_id);
1106 		KASSERT(pt->pt_pid == 0);
1107 		last_free_pt = pg_id;
1108 		pid_alloc_cnt--;
1109 	}
1110 	return pgrp;
1111 }
1112 
1113 /*
1114  * pg_delete: delete and free a process group.
1115  *  => must be called with the proc_lock held, which will be released.
1116  */
1117 static void
1118 pg_delete(pid_t pg_id)
1119 {
1120 	struct pgrp *pg;
1121 	struct tty *ttyp;
1122 	struct session *ss;
1123 
1124 	KASSERT(mutex_owned(proc_lock));
1125 
1126 	pg = pid_table[pg_id & pid_tbl_mask].pt_pgrp;
1127 	if (pg == NULL || pg->pg_id != pg_id || !LIST_EMPTY(&pg->pg_members)) {
1128 		mutex_exit(proc_lock);
1129 		return;
1130 	}
1131 
1132 	ss = pg->pg_session;
1133 
1134 	/* Remove reference (if any) from tty to this process group */
1135 	mutex_spin_enter(&tty_lock);
1136 	ttyp = ss->s_ttyp;
1137 	if (ttyp != NULL && ttyp->t_pgrp == pg) {
1138 		ttyp->t_pgrp = NULL;
1139 		KASSERT(ttyp->t_session == ss);
1140 	}
1141 	mutex_spin_exit(&tty_lock);
1142 
1143 	/*
1144 	 * The leading process group in a session is freed by proc_sessrele(),
1145 	 * if last reference.  Note: proc_sessrele() releases proc_lock.
1146 	 */
1147 	pg = (ss->s_sid != pg->pg_id) ? pg_remove(pg_id) : NULL;
1148 	proc_sessrele(ss);
1149 
1150 	if (pg != NULL) {
1151 		/* Free it, if was not done by proc_sessrele(). */
1152 		kmem_free(pg, sizeof(struct pgrp));
1153 	}
1154 }
1155 
1156 /*
1157  * Adjust pgrp jobc counters when specified process changes process group.
1158  * We count the number of processes in each process group that "qualify"
1159  * the group for terminal job control (those with a parent in a different
1160  * process group of the same session).  If that count reaches zero, the
1161  * process group becomes orphaned.  Check both the specified process'
1162  * process group and that of its children.
1163  * entering == 0 => p is leaving specified group.
1164  * entering == 1 => p is entering specified group.
1165  *
1166  * Call with proc_lock held.
1167  */
1168 void
1169 fixjobc(struct proc *p, struct pgrp *pgrp, int entering)
1170 {
1171 	struct pgrp *hispgrp;
1172 	struct session *mysession = pgrp->pg_session;
1173 	struct proc *child;
1174 
1175 	KASSERT(mutex_owned(proc_lock));
1176 
1177 	/*
1178 	 * Check p's parent to see whether p qualifies its own process
1179 	 * group; if so, adjust count for p's process group.
1180 	 */
1181 	hispgrp = p->p_pptr->p_pgrp;
1182 	if (hispgrp != pgrp && hispgrp->pg_session == mysession) {
1183 		if (entering) {
1184 			pgrp->pg_jobc++;
1185 			p->p_lflag &= ~PL_ORPHANPG;
1186 		} else if (--pgrp->pg_jobc == 0)
1187 			orphanpg(pgrp);
1188 	}
1189 
1190 	/*
1191 	 * Check this process' children to see whether they qualify
1192 	 * their process groups; if so, adjust counts for children's
1193 	 * process groups.
1194 	 */
1195 	LIST_FOREACH(child, &p->p_children, p_sibling) {
1196 		hispgrp = child->p_pgrp;
1197 		if (hispgrp != pgrp && hispgrp->pg_session == mysession &&
1198 		    !P_ZOMBIE(child)) {
1199 			if (entering) {
1200 				child->p_lflag &= ~PL_ORPHANPG;
1201 				hispgrp->pg_jobc++;
1202 			} else if (--hispgrp->pg_jobc == 0)
1203 				orphanpg(hispgrp);
1204 		}
1205 	}
1206 }
1207 
1208 /*
1209  * A process group has become orphaned;
1210  * if there are any stopped processes in the group,
1211  * hang-up all process in that group.
1212  *
1213  * Call with proc_lock held.
1214  */
1215 static void
1216 orphanpg(struct pgrp *pg)
1217 {
1218 	struct proc *p;
1219 
1220 	KASSERT(mutex_owned(proc_lock));
1221 
1222 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
1223 		if (p->p_stat == SSTOP) {
1224 			p->p_lflag |= PL_ORPHANPG;
1225 			psignal(p, SIGHUP);
1226 			psignal(p, SIGCONT);
1227 		}
1228 	}
1229 }
1230 
1231 #ifdef DDB
1232 #include <ddb/db_output.h>
1233 void pidtbl_dump(void);
1234 void
1235 pidtbl_dump(void)
1236 {
1237 	struct pid_table *pt;
1238 	struct proc *p;
1239 	struct pgrp *pgrp;
1240 	int id;
1241 
1242 	db_printf("pid table %p size %x, next %x, last %x\n",
1243 		pid_table, pid_tbl_mask+1,
1244 		next_free_pt, last_free_pt);
1245 	for (pt = pid_table, id = 0; id <= pid_tbl_mask; id++, pt++) {
1246 		p = pt->pt_proc;
1247 		if (!P_VALID(p) && !pt->pt_pgrp)
1248 			continue;
1249 		db_printf("  id %x: ", id);
1250 		if (P_VALID(p))
1251 			db_printf("slotpid %d proc %p id %d (0x%x) %s\n",
1252 				pt->pt_pid, p, p->p_pid, p->p_pid, p->p_comm);
1253 		else
1254 			db_printf("next %x use %x\n",
1255 				P_NEXT(p) & pid_tbl_mask,
1256 				P_NEXT(p) & ~pid_tbl_mask);
1257 		if ((pgrp = pt->pt_pgrp)) {
1258 			db_printf("\tsession %p, sid %d, count %d, login %s\n",
1259 			    pgrp->pg_session, pgrp->pg_session->s_sid,
1260 			    pgrp->pg_session->s_count,
1261 			    pgrp->pg_session->s_login);
1262 			db_printf("\tpgrp %p, pg_id %d, pg_jobc %d, members %p\n",
1263 			    pgrp, pgrp->pg_id, pgrp->pg_jobc,
1264 			    LIST_FIRST(&pgrp->pg_members));
1265 			LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
1266 				db_printf("\t\tpid %d addr %p pgrp %p %s\n",
1267 				    p->p_pid, p, p->p_pgrp, p->p_comm);
1268 			}
1269 		}
1270 	}
1271 }
1272 #endif /* DDB */
1273 
1274 #ifdef KSTACK_CHECK_MAGIC
1275 
1276 #define	KSTACK_MAGIC	0xdeadbeaf
1277 
1278 /* XXX should be per process basis? */
1279 static int	kstackleftmin = KSTACK_SIZE;
1280 static int	kstackleftthres = KSTACK_SIZE / 8;
1281 
1282 void
1283 kstack_setup_magic(const struct lwp *l)
1284 {
1285 	uint32_t *ip;
1286 	uint32_t const *end;
1287 
1288 	KASSERT(l != NULL);
1289 	KASSERT(l != &lwp0);
1290 
1291 	/*
1292 	 * fill all the stack with magic number
1293 	 * so that later modification on it can be detected.
1294 	 */
1295 	ip = (uint32_t *)KSTACK_LOWEST_ADDR(l);
1296 	end = (uint32_t *)((char *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE);
1297 	for (; ip < end; ip++) {
1298 		*ip = KSTACK_MAGIC;
1299 	}
1300 }
1301 
1302 void
1303 kstack_check_magic(const struct lwp *l)
1304 {
1305 	uint32_t const *ip, *end;
1306 	int stackleft;
1307 
1308 	KASSERT(l != NULL);
1309 
1310 	/* don't check proc0 */ /*XXX*/
1311 	if (l == &lwp0)
1312 		return;
1313 
1314 #ifdef __MACHINE_STACK_GROWS_UP
1315 	/* stack grows upwards (eg. hppa) */
1316 	ip = (uint32_t *)((void *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE);
1317 	end = (uint32_t *)KSTACK_LOWEST_ADDR(l);
1318 	for (ip--; ip >= end; ip--)
1319 		if (*ip != KSTACK_MAGIC)
1320 			break;
1321 
1322 	stackleft = (void *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE - (void *)ip;
1323 #else /* __MACHINE_STACK_GROWS_UP */
1324 	/* stack grows downwards (eg. i386) */
1325 	ip = (uint32_t *)KSTACK_LOWEST_ADDR(l);
1326 	end = (uint32_t *)((char *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE);
1327 	for (; ip < end; ip++)
1328 		if (*ip != KSTACK_MAGIC)
1329 			break;
1330 
1331 	stackleft = ((const char *)ip) - (const char *)KSTACK_LOWEST_ADDR(l);
1332 #endif /* __MACHINE_STACK_GROWS_UP */
1333 
1334 	if (kstackleftmin > stackleft) {
1335 		kstackleftmin = stackleft;
1336 		if (stackleft < kstackleftthres)
1337 			printf("warning: kernel stack left %d bytes"
1338 			    "(pid %u:lid %u)\n", stackleft,
1339 			    (u_int)l->l_proc->p_pid, (u_int)l->l_lid);
1340 	}
1341 
1342 	if (stackleft <= 0) {
1343 		panic("magic on the top of kernel stack changed for "
1344 		    "pid %u, lid %u: maybe kernel stack overflow",
1345 		    (u_int)l->l_proc->p_pid, (u_int)l->l_lid);
1346 	}
1347 }
1348 #endif /* KSTACK_CHECK_MAGIC */
1349 
1350 int
1351 proclist_foreach_call(struct proclist *list,
1352     int (*callback)(struct proc *, void *arg), void *arg)
1353 {
1354 	struct proc marker;
1355 	struct proc *p;
1356 	int ret = 0;
1357 
1358 	marker.p_flag = PK_MARKER;
1359 	mutex_enter(proc_lock);
1360 	for (p = LIST_FIRST(list); ret == 0 && p != NULL;) {
1361 		if (p->p_flag & PK_MARKER) {
1362 			p = LIST_NEXT(p, p_list);
1363 			continue;
1364 		}
1365 		LIST_INSERT_AFTER(p, &marker, p_list);
1366 		ret = (*callback)(p, arg);
1367 		KASSERT(mutex_owned(proc_lock));
1368 		p = LIST_NEXT(&marker, p_list);
1369 		LIST_REMOVE(&marker, p_list);
1370 	}
1371 	mutex_exit(proc_lock);
1372 
1373 	return ret;
1374 }
1375 
1376 int
1377 proc_vmspace_getref(struct proc *p, struct vmspace **vm)
1378 {
1379 
1380 	/* XXXCDC: how should locking work here? */
1381 
1382 	/* curproc exception is for coredump. */
1383 
1384 	if ((p != curproc && (p->p_sflag & PS_WEXIT) != 0) ||
1385 	    (p->p_vmspace->vm_refcnt < 1)) { /* XXX */
1386 		return EFAULT;
1387 	}
1388 
1389 	uvmspace_addref(p->p_vmspace);
1390 	*vm = p->p_vmspace;
1391 
1392 	return 0;
1393 }
1394 
1395 /*
1396  * Acquire a write lock on the process credential.
1397  */
1398 void
1399 proc_crmod_enter(void)
1400 {
1401 	struct lwp *l = curlwp;
1402 	struct proc *p = l->l_proc;
1403 	kauth_cred_t oc;
1404 
1405 	/* Reset what needs to be reset in plimit. */
1406 	if (p->p_limit->pl_corename != defcorename) {
1407 		lim_setcorename(p, defcorename, 0);
1408 	}
1409 
1410 	mutex_enter(p->p_lock);
1411 
1412 	/* Ensure the LWP cached credentials are up to date. */
1413 	if ((oc = l->l_cred) != p->p_cred) {
1414 		kauth_cred_hold(p->p_cred);
1415 		l->l_cred = p->p_cred;
1416 		kauth_cred_free(oc);
1417 	}
1418 }
1419 
1420 /*
1421  * Set in a new process credential, and drop the write lock.  The credential
1422  * must have a reference already.  Optionally, free a no-longer required
1423  * credential.  The scheduler also needs to inspect p_cred, so we also
1424  * briefly acquire the sched state mutex.
1425  */
1426 void
1427 proc_crmod_leave(kauth_cred_t scred, kauth_cred_t fcred, bool sugid)
1428 {
1429 	struct lwp *l = curlwp, *l2;
1430 	struct proc *p = l->l_proc;
1431 	kauth_cred_t oc;
1432 
1433 	KASSERT(mutex_owned(p->p_lock));
1434 
1435 	/* Is there a new credential to set in? */
1436 	if (scred != NULL) {
1437 		p->p_cred = scred;
1438 		LIST_FOREACH(l2, &p->p_lwps, l_sibling) {
1439 			if (l2 != l)
1440 				l2->l_prflag |= LPR_CRMOD;
1441 		}
1442 
1443 		/* Ensure the LWP cached credentials are up to date. */
1444 		if ((oc = l->l_cred) != scred) {
1445 			kauth_cred_hold(scred);
1446 			l->l_cred = scred;
1447 		}
1448 	} else
1449 		oc = NULL;	/* XXXgcc */
1450 
1451 	if (sugid) {
1452 		/*
1453 		 * Mark process as having changed credentials, stops
1454 		 * tracing etc.
1455 		 */
1456 		p->p_flag |= PK_SUGID;
1457 	}
1458 
1459 	mutex_exit(p->p_lock);
1460 
1461 	/* If there is a credential to be released, free it now. */
1462 	if (fcred != NULL) {
1463 		KASSERT(scred != NULL);
1464 		kauth_cred_free(fcred);
1465 		if (oc != scred)
1466 			kauth_cred_free(oc);
1467 	}
1468 }
1469 
1470 /*
1471  * proc_specific_key_create --
1472  *	Create a key for subsystem proc-specific data.
1473  */
1474 int
1475 proc_specific_key_create(specificdata_key_t *keyp, specificdata_dtor_t dtor)
1476 {
1477 
1478 	return (specificdata_key_create(proc_specificdata_domain, keyp, dtor));
1479 }
1480 
1481 /*
1482  * proc_specific_key_delete --
1483  *	Delete a key for subsystem proc-specific data.
1484  */
1485 void
1486 proc_specific_key_delete(specificdata_key_t key)
1487 {
1488 
1489 	specificdata_key_delete(proc_specificdata_domain, key);
1490 }
1491 
1492 /*
1493  * proc_initspecific --
1494  *	Initialize a proc's specificdata container.
1495  */
1496 void
1497 proc_initspecific(struct proc *p)
1498 {
1499 	int error __diagused;
1500 
1501 	error = specificdata_init(proc_specificdata_domain, &p->p_specdataref);
1502 	KASSERT(error == 0);
1503 }
1504 
1505 /*
1506  * proc_finispecific --
1507  *	Finalize a proc's specificdata container.
1508  */
1509 void
1510 proc_finispecific(struct proc *p)
1511 {
1512 
1513 	specificdata_fini(proc_specificdata_domain, &p->p_specdataref);
1514 }
1515 
1516 /*
1517  * proc_getspecific --
1518  *	Return proc-specific data corresponding to the specified key.
1519  */
1520 void *
1521 proc_getspecific(struct proc *p, specificdata_key_t key)
1522 {
1523 
1524 	return (specificdata_getspecific(proc_specificdata_domain,
1525 					 &p->p_specdataref, key));
1526 }
1527 
1528 /*
1529  * proc_setspecific --
1530  *	Set proc-specific data corresponding to the specified key.
1531  */
1532 void
1533 proc_setspecific(struct proc *p, specificdata_key_t key, void *data)
1534 {
1535 
1536 	specificdata_setspecific(proc_specificdata_domain,
1537 				 &p->p_specdataref, key, data);
1538 }
1539 
1540 int
1541 proc_uidmatch(kauth_cred_t cred, kauth_cred_t target)
1542 {
1543 	int r = 0;
1544 
1545 	if (kauth_cred_getuid(cred) != kauth_cred_getuid(target) ||
1546 	    kauth_cred_getuid(cred) != kauth_cred_getsvuid(target)) {
1547 		/*
1548 		 * suid proc of ours or proc not ours
1549 		 */
1550 		r = EPERM;
1551 	} else if (kauth_cred_getgid(target) != kauth_cred_getsvgid(target)) {
1552 		/*
1553 		 * sgid proc has sgid back to us temporarily
1554 		 */
1555 		r = EPERM;
1556 	} else {
1557 		/*
1558 		 * our rgid must be in target's group list (ie,
1559 		 * sub-processes started by a sgid process)
1560 		 */
1561 		int ismember = 0;
1562 
1563 		if (kauth_cred_ismember_gid(cred,
1564 		    kauth_cred_getgid(target), &ismember) != 0 ||
1565 		    !ismember)
1566 			r = EPERM;
1567 	}
1568 
1569 	return (r);
1570 }
1571 
1572 /*
1573  * sysctl stuff
1574  */
1575 
1576 #define KERN_PROCSLOP	(5 * sizeof(struct kinfo_proc))
1577 
1578 static const u_int sysctl_flagmap[] = {
1579 	PK_ADVLOCK, P_ADVLOCK,
1580 	PK_EXEC, P_EXEC,
1581 	PK_NOCLDWAIT, P_NOCLDWAIT,
1582 	PK_32, P_32,
1583 	PK_CLDSIGIGN, P_CLDSIGIGN,
1584 	PK_SUGID, P_SUGID,
1585 	0
1586 };
1587 
1588 static const u_int sysctl_sflagmap[] = {
1589 	PS_NOCLDSTOP, P_NOCLDSTOP,
1590 	PS_WEXIT, P_WEXIT,
1591 	PS_STOPFORK, P_STOPFORK,
1592 	PS_STOPEXEC, P_STOPEXEC,
1593 	PS_STOPEXIT, P_STOPEXIT,
1594 	0
1595 };
1596 
1597 static const u_int sysctl_slflagmap[] = {
1598 	PSL_TRACED, P_TRACED,
1599 	PSL_CHTRACED, P_CHTRACED,
1600 	PSL_SYSCALL, P_SYSCALL,
1601 	0
1602 };
1603 
1604 static const u_int sysctl_lflagmap[] = {
1605 	PL_CONTROLT, P_CONTROLT,
1606 	PL_PPWAIT, P_PPWAIT,
1607 	0
1608 };
1609 
1610 static const u_int sysctl_stflagmap[] = {
1611 	PST_PROFIL, P_PROFIL,
1612 	0
1613 
1614 };
1615 
1616 /* used by kern_lwp also */
1617 const u_int sysctl_lwpflagmap[] = {
1618 	LW_SINTR, L_SINTR,
1619 	LW_SYSTEM, L_SYSTEM,
1620 	0
1621 };
1622 
1623 /*
1624  * Find the most ``active'' lwp of a process and return it for ps display
1625  * purposes
1626  */
1627 static struct lwp *
1628 proc_active_lwp(struct proc *p)
1629 {
1630 	static const int ostat[] = {
1631 		0,
1632 		2,	/* LSIDL */
1633 		6,	/* LSRUN */
1634 		5,	/* LSSLEEP */
1635 		4,	/* LSSTOP */
1636 		0,	/* LSZOMB */
1637 		1,	/* LSDEAD */
1638 		7,	/* LSONPROC */
1639 		3	/* LSSUSPENDED */
1640 	};
1641 
1642 	struct lwp *l, *lp = NULL;
1643 	LIST_FOREACH(l, &p->p_lwps, l_sibling) {
1644 		KASSERT(l->l_stat >= 0 && l->l_stat < __arraycount(ostat));
1645 		if (lp == NULL ||
1646 		    ostat[l->l_stat] > ostat[lp->l_stat] ||
1647 		    (ostat[l->l_stat] == ostat[lp->l_stat] &&
1648 		    l->l_cpticks > lp->l_cpticks)) {
1649 			lp = l;
1650 			continue;
1651 		}
1652 	}
1653 	return lp;
1654 }
1655 
1656 static int
1657 sysctl_doeproc(SYSCTLFN_ARGS)
1658 {
1659 	union {
1660 		struct kinfo_proc kproc;
1661 		struct kinfo_proc2 kproc2;
1662 	} *kbuf;
1663 	struct proc *p, *next, *marker;
1664 	char *where, *dp;
1665 	int type, op, arg, error;
1666 	u_int elem_size, kelem_size, elem_count;
1667 	size_t buflen, needed;
1668 	bool match, zombie, mmmbrains;
1669 	const bool allowaddr = get_expose_address(curproc);
1670 
1671 	if (namelen == 1 && name[0] == CTL_QUERY)
1672 		return (sysctl_query(SYSCTLFN_CALL(rnode)));
1673 
1674 	dp = where = oldp;
1675 	buflen = where != NULL ? *oldlenp : 0;
1676 	error = 0;
1677 	needed = 0;
1678 	type = rnode->sysctl_num;
1679 
1680 	if (type == KERN_PROC) {
1681 		if (namelen == 0)
1682 			return EINVAL;
1683 		switch (op = name[0]) {
1684 		case KERN_PROC_ALL:
1685 			if (namelen != 1)
1686 				return EINVAL;
1687 			arg = 0;
1688 			break;
1689 		default:
1690 			if (namelen != 2)
1691 				return EINVAL;
1692 			arg = name[1];
1693 			break;
1694 		}
1695 		elem_count = 0;	/* Hush little compiler, don't you cry */
1696 		kelem_size = elem_size = sizeof(kbuf->kproc);
1697 	} else {
1698 		if (namelen != 4)
1699 			return EINVAL;
1700 		op = name[0];
1701 		arg = name[1];
1702 		elem_size = name[2];
1703 		elem_count = name[3];
1704 		kelem_size = sizeof(kbuf->kproc2);
1705 	}
1706 
1707 	sysctl_unlock();
1708 
1709 	kbuf = kmem_zalloc(sizeof(*kbuf), KM_SLEEP);
1710 	marker = kmem_alloc(sizeof(*marker), KM_SLEEP);
1711 	marker->p_flag = PK_MARKER;
1712 
1713 	mutex_enter(proc_lock);
1714 	/*
1715 	 * Start with zombies to prevent reporting processes twice, in case they
1716 	 * are dying and being moved from the list of alive processes to zombies.
1717 	 */
1718 	mmmbrains = true;
1719 	for (p = LIST_FIRST(&zombproc);; p = next) {
1720 		if (p == NULL) {
1721 			if (mmmbrains) {
1722 				p = LIST_FIRST(&allproc);
1723 				mmmbrains = false;
1724 			}
1725 			if (p == NULL)
1726 				break;
1727 		}
1728 		next = LIST_NEXT(p, p_list);
1729 		if ((p->p_flag & PK_MARKER) != 0)
1730 			continue;
1731 
1732 		/*
1733 		 * Skip embryonic processes.
1734 		 */
1735 		if (p->p_stat == SIDL)
1736 			continue;
1737 
1738 		mutex_enter(p->p_lock);
1739 		error = kauth_authorize_process(l->l_cred,
1740 		    KAUTH_PROCESS_CANSEE, p,
1741 		    KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_EPROC), NULL, NULL);
1742 		if (error != 0) {
1743 			mutex_exit(p->p_lock);
1744 			continue;
1745 		}
1746 
1747 		/*
1748 		 * Hande all the operations in one switch on the cost of
1749 		 * algorithm complexity is on purpose. The win splitting this
1750 		 * function into several similar copies makes maintenance burden
1751 		 * burden, code grow and boost is neglible in practical systems.
1752 		 */
1753 		switch (op) {
1754 		case KERN_PROC_PID:
1755 			match = (p->p_pid == (pid_t)arg);
1756 			break;
1757 
1758 		case KERN_PROC_PGRP:
1759 			match = (p->p_pgrp->pg_id == (pid_t)arg);
1760 			break;
1761 
1762 		case KERN_PROC_SESSION:
1763 			match = (p->p_session->s_sid == (pid_t)arg);
1764 			break;
1765 
1766 		case KERN_PROC_TTY:
1767 			match = true;
1768 			if (arg == (int) KERN_PROC_TTY_REVOKE) {
1769 				if ((p->p_lflag & PL_CONTROLT) == 0 ||
1770 				    p->p_session->s_ttyp == NULL ||
1771 				    p->p_session->s_ttyvp != NULL) {
1772 				    	match = false;
1773 				}
1774 			} else if ((p->p_lflag & PL_CONTROLT) == 0 ||
1775 			    p->p_session->s_ttyp == NULL) {
1776 				if ((dev_t)arg != KERN_PROC_TTY_NODEV) {
1777 					match = false;
1778 				}
1779 			} else if (p->p_session->s_ttyp->t_dev != (dev_t)arg) {
1780 				match = false;
1781 			}
1782 			break;
1783 
1784 		case KERN_PROC_UID:
1785 			match = (kauth_cred_geteuid(p->p_cred) == (uid_t)arg);
1786 			break;
1787 
1788 		case KERN_PROC_RUID:
1789 			match = (kauth_cred_getuid(p->p_cred) == (uid_t)arg);
1790 			break;
1791 
1792 		case KERN_PROC_GID:
1793 			match = (kauth_cred_getegid(p->p_cred) == (uid_t)arg);
1794 			break;
1795 
1796 		case KERN_PROC_RGID:
1797 			match = (kauth_cred_getgid(p->p_cred) == (uid_t)arg);
1798 			break;
1799 
1800 		case KERN_PROC_ALL:
1801 			match = true;
1802 			/* allow everything */
1803 			break;
1804 
1805 		default:
1806 			error = EINVAL;
1807 			mutex_exit(p->p_lock);
1808 			goto cleanup;
1809 		}
1810 		if (!match) {
1811 			mutex_exit(p->p_lock);
1812 			continue;
1813 		}
1814 
1815 		/*
1816 		 * Grab a hold on the process.
1817 		 */
1818 		if (mmmbrains) {
1819 			zombie = true;
1820 		} else {
1821 			zombie = !rw_tryenter(&p->p_reflock, RW_READER);
1822 		}
1823 		if (zombie) {
1824 			LIST_INSERT_AFTER(p, marker, p_list);
1825 		}
1826 
1827 		if (buflen >= elem_size &&
1828 		    (type == KERN_PROC || elem_count > 0)) {
1829 			ruspace(p);	/* Update process vm resource use */
1830 
1831 			if (type == KERN_PROC) {
1832 				fill_proc(p, &kbuf->kproc.kp_proc, allowaddr);
1833 				fill_eproc(p, &kbuf->kproc.kp_eproc, zombie,
1834 				    allowaddr);
1835 			} else {
1836 				fill_kproc2(p, &kbuf->kproc2, zombie,
1837 				    allowaddr);
1838 				elem_count--;
1839 			}
1840 			mutex_exit(p->p_lock);
1841 			mutex_exit(proc_lock);
1842 			/*
1843 			 * Copy out elem_size, but not larger than kelem_size
1844 			 */
1845 			error = sysctl_copyout(l, kbuf, dp,
1846 			    uimin(kelem_size, elem_size));
1847 			mutex_enter(proc_lock);
1848 			if (error) {
1849 				goto bah;
1850 			}
1851 			dp += elem_size;
1852 			buflen -= elem_size;
1853 		} else {
1854 			mutex_exit(p->p_lock);
1855 		}
1856 		needed += elem_size;
1857 
1858 		/*
1859 		 * Release reference to process.
1860 		 */
1861 	 	if (zombie) {
1862 			next = LIST_NEXT(marker, p_list);
1863  			LIST_REMOVE(marker, p_list);
1864 		} else {
1865 			rw_exit(&p->p_reflock);
1866 			next = LIST_NEXT(p, p_list);
1867 		}
1868 
1869 		/*
1870 		 * Short-circuit break quickly!
1871 		 */
1872 		if (op == KERN_PROC_PID)
1873                 	break;
1874 	}
1875 	mutex_exit(proc_lock);
1876 
1877 	if (where != NULL) {
1878 		*oldlenp = dp - where;
1879 		if (needed > *oldlenp) {
1880 			error = ENOMEM;
1881 			goto out;
1882 		}
1883 	} else {
1884 		needed += KERN_PROCSLOP;
1885 		*oldlenp = needed;
1886 	}
1887 	kmem_free(kbuf, sizeof(*kbuf));
1888 	kmem_free(marker, sizeof(*marker));
1889 	sysctl_relock();
1890 	return 0;
1891  bah:
1892  	if (zombie)
1893  		LIST_REMOVE(marker, p_list);
1894 	else
1895 		rw_exit(&p->p_reflock);
1896  cleanup:
1897 	mutex_exit(proc_lock);
1898  out:
1899 	kmem_free(kbuf, sizeof(*kbuf));
1900 	kmem_free(marker, sizeof(*marker));
1901 	sysctl_relock();
1902 	return error;
1903 }
1904 
1905 int
1906 copyin_psstrings(struct proc *p, struct ps_strings *arginfo)
1907 {
1908 #if !defined(_RUMPKERNEL)
1909 	int retval;
1910 
1911 	if (p->p_flag & PK_32) {
1912 		MODULE_HOOK_CALL(kern_proc32_copyin_hook, (p, arginfo),
1913 		    enosys(), retval);
1914 		return retval;
1915 	}
1916 #endif /* !defined(_RUMPKERNEL) */
1917 
1918 	return copyin_proc(p, (void *)p->p_psstrp, arginfo, sizeof(*arginfo));
1919 }
1920 
1921 static int
1922 copy_procargs_sysctl_cb(void *cookie_, const void *src, size_t off, size_t len)
1923 {
1924 	void **cookie = cookie_;
1925 	struct lwp *l = cookie[0];
1926 	char *dst = cookie[1];
1927 
1928 	return sysctl_copyout(l, src, dst + off, len);
1929 }
1930 
1931 /*
1932  * sysctl helper routine for kern.proc_args pseudo-subtree.
1933  */
1934 static int
1935 sysctl_kern_proc_args(SYSCTLFN_ARGS)
1936 {
1937 	struct ps_strings pss;
1938 	struct proc *p;
1939 	pid_t pid;
1940 	int type, error;
1941 	void *cookie[2];
1942 
1943 	if (namelen == 1 && name[0] == CTL_QUERY)
1944 		return (sysctl_query(SYSCTLFN_CALL(rnode)));
1945 
1946 	if (newp != NULL || namelen != 2)
1947 		return (EINVAL);
1948 	pid = name[0];
1949 	type = name[1];
1950 
1951 	switch (type) {
1952 	case KERN_PROC_PATHNAME:
1953 		sysctl_unlock();
1954 		error = fill_pathname(l, pid, oldp, oldlenp);
1955 		sysctl_relock();
1956 		return error;
1957 
1958 	case KERN_PROC_CWD:
1959 		sysctl_unlock();
1960 		error = fill_cwd(l, pid, oldp, oldlenp);
1961 		sysctl_relock();
1962 		return error;
1963 
1964 	case KERN_PROC_ARGV:
1965 	case KERN_PROC_NARGV:
1966 	case KERN_PROC_ENV:
1967 	case KERN_PROC_NENV:
1968 		/* ok */
1969 		break;
1970 	default:
1971 		return (EINVAL);
1972 	}
1973 
1974 	sysctl_unlock();
1975 
1976 	/* check pid */
1977 	mutex_enter(proc_lock);
1978 	if ((p = proc_find(pid)) == NULL) {
1979 		error = EINVAL;
1980 		goto out_locked;
1981 	}
1982 	mutex_enter(p->p_lock);
1983 
1984 	/* Check permission. */
1985 	if (type == KERN_PROC_ARGV || type == KERN_PROC_NARGV)
1986 		error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE,
1987 		    p, KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ARGS), NULL, NULL);
1988 	else if (type == KERN_PROC_ENV || type == KERN_PROC_NENV)
1989 		error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE,
1990 		    p, KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENV), NULL, NULL);
1991 	else
1992 		error = EINVAL; /* XXXGCC */
1993 	if (error) {
1994 		mutex_exit(p->p_lock);
1995 		goto out_locked;
1996 	}
1997 
1998 	if (oldp == NULL) {
1999 		if (type == KERN_PROC_NARGV || type == KERN_PROC_NENV)
2000 			*oldlenp = sizeof (int);
2001 		else
2002 			*oldlenp = ARG_MAX;	/* XXX XXX XXX */
2003 		error = 0;
2004 		mutex_exit(p->p_lock);
2005 		goto out_locked;
2006 	}
2007 
2008 	/*
2009 	 * Zombies don't have a stack, so we can't read their psstrings.
2010 	 * System processes also don't have a user stack.
2011 	 */
2012 	if (P_ZOMBIE(p) || (p->p_flag & PK_SYSTEM) != 0) {
2013 		error = EINVAL;
2014 		mutex_exit(p->p_lock);
2015 		goto out_locked;
2016 	}
2017 
2018 	error = rw_tryenter(&p->p_reflock, RW_READER) ? 0 : EBUSY;
2019 	mutex_exit(p->p_lock);
2020 	if (error) {
2021 		goto out_locked;
2022 	}
2023 	mutex_exit(proc_lock);
2024 
2025 	if (type == KERN_PROC_NARGV || type == KERN_PROC_NENV) {
2026 		int value;
2027 		if ((error = copyin_psstrings(p, &pss)) == 0) {
2028 			if (type == KERN_PROC_NARGV)
2029 				value = pss.ps_nargvstr;
2030 			else
2031 				value = pss.ps_nenvstr;
2032 			error = sysctl_copyout(l, &value, oldp, sizeof(value));
2033 			*oldlenp = sizeof(value);
2034 		}
2035 	} else {
2036 		cookie[0] = l;
2037 		cookie[1] = oldp;
2038 		error = copy_procargs(p, type, oldlenp,
2039 		    copy_procargs_sysctl_cb, cookie);
2040 	}
2041 	rw_exit(&p->p_reflock);
2042 	sysctl_relock();
2043 	return error;
2044 
2045 out_locked:
2046 	mutex_exit(proc_lock);
2047 	sysctl_relock();
2048 	return error;
2049 }
2050 
2051 int
2052 copy_procargs(struct proc *p, int oid, size_t *limit,
2053     int (*cb)(void *, const void *, size_t, size_t), void *cookie)
2054 {
2055 	struct ps_strings pss;
2056 	size_t len, i, loaded, entry_len;
2057 	struct uio auio;
2058 	struct iovec aiov;
2059 	int error, argvlen;
2060 	char *arg;
2061 	char **argv;
2062 	vaddr_t user_argv;
2063 	struct vmspace *vmspace;
2064 
2065 	/*
2066 	 * Allocate a temporary buffer to hold the argument vector and
2067 	 * the arguments themselve.
2068 	 */
2069 	arg = kmem_alloc(PAGE_SIZE, KM_SLEEP);
2070 	argv = kmem_alloc(PAGE_SIZE, KM_SLEEP);
2071 
2072 	/*
2073 	 * Lock the process down in memory.
2074 	 */
2075 	vmspace = p->p_vmspace;
2076 	uvmspace_addref(vmspace);
2077 
2078 	/*
2079 	 * Read in the ps_strings structure.
2080 	 */
2081 	if ((error = copyin_psstrings(p, &pss)) != 0)
2082 		goto done;
2083 
2084 	/*
2085 	 * Now read the address of the argument vector.
2086 	 */
2087 	switch (oid) {
2088 	case KERN_PROC_ARGV:
2089 		user_argv = (uintptr_t)pss.ps_argvstr;
2090 		argvlen = pss.ps_nargvstr;
2091 		break;
2092 	case KERN_PROC_ENV:
2093 		user_argv = (uintptr_t)pss.ps_envstr;
2094 		argvlen = pss.ps_nenvstr;
2095 		break;
2096 	default:
2097 		error = EINVAL;
2098 		goto done;
2099 	}
2100 
2101 	if (argvlen < 0) {
2102 		error = EIO;
2103 		goto done;
2104 	}
2105 
2106 
2107 	/*
2108 	 * Now copy each string.
2109 	 */
2110 	len = 0; /* bytes written to user buffer */
2111 	loaded = 0; /* bytes from argv already processed */
2112 	i = 0; /* To make compiler happy */
2113 	entry_len = PROC_PTRSZ(p);
2114 
2115 	for (; argvlen; --argvlen) {
2116 		int finished = 0;
2117 		vaddr_t base;
2118 		size_t xlen;
2119 		int j;
2120 
2121 		if (loaded == 0) {
2122 			size_t rem = entry_len * argvlen;
2123 			loaded = MIN(rem, PAGE_SIZE);
2124 			error = copyin_vmspace(vmspace,
2125 			    (const void *)user_argv, argv, loaded);
2126 			if (error)
2127 				break;
2128 			user_argv += loaded;
2129 			i = 0;
2130 		}
2131 
2132 #if !defined(_RUMPKERNEL)
2133 		if (p->p_flag & PK_32)
2134 			MODULE_HOOK_CALL(kern_proc32_base_hook,
2135 			    (argv, i++), 0, base);
2136 		else
2137 #endif /* !defined(_RUMPKERNEL) */
2138 			base = (vaddr_t)argv[i++];
2139 		loaded -= entry_len;
2140 
2141 		/*
2142 		 * The program has messed around with its arguments,
2143 		 * possibly deleting some, and replacing them with
2144 		 * NULL's. Treat this as the last argument and not
2145 		 * a failure.
2146 		 */
2147 		if (base == 0)
2148 			break;
2149 
2150 		while (!finished) {
2151 			xlen = PAGE_SIZE - (base & PAGE_MASK);
2152 
2153 			aiov.iov_base = arg;
2154 			aiov.iov_len = PAGE_SIZE;
2155 			auio.uio_iov = &aiov;
2156 			auio.uio_iovcnt = 1;
2157 			auio.uio_offset = base;
2158 			auio.uio_resid = xlen;
2159 			auio.uio_rw = UIO_READ;
2160 			UIO_SETUP_SYSSPACE(&auio);
2161 			error = uvm_io(&vmspace->vm_map, &auio, 0);
2162 			if (error)
2163 				goto done;
2164 
2165 			/* Look for the end of the string */
2166 			for (j = 0; j < xlen; j++) {
2167 				if (arg[j] == '\0') {
2168 					xlen = j + 1;
2169 					finished = 1;
2170 					break;
2171 				}
2172 			}
2173 
2174 			/* Check for user buffer overflow */
2175 			if (len + xlen > *limit) {
2176 				finished = 1;
2177 				if (len > *limit)
2178 					xlen = 0;
2179 				else
2180 					xlen = *limit - len;
2181 			}
2182 
2183 			/* Copyout the page */
2184 			error = (*cb)(cookie, arg, len, xlen);
2185 			if (error)
2186 				goto done;
2187 
2188 			len += xlen;
2189 			base += xlen;
2190 		}
2191 	}
2192 	*limit = len;
2193 
2194 done:
2195 	kmem_free(argv, PAGE_SIZE);
2196 	kmem_free(arg, PAGE_SIZE);
2197 	uvmspace_free(vmspace);
2198 	return error;
2199 }
2200 
2201 /*
2202  * Fill in a proc structure for the specified process.
2203  */
2204 static void
2205 fill_proc(const struct proc *psrc, struct proc *p, bool allowaddr)
2206 {
2207 	COND_SET_VALUE(p->p_list, psrc->p_list, allowaddr);
2208 	COND_SET_VALUE(p->p_auxlock, psrc->p_auxlock, allowaddr);
2209 	COND_SET_VALUE(p->p_lock, psrc->p_lock, allowaddr);
2210 	COND_SET_VALUE(p->p_stmutex, psrc->p_stmutex, allowaddr);
2211 	COND_SET_VALUE(p->p_reflock, psrc->p_reflock, allowaddr);
2212 	COND_SET_VALUE(p->p_waitcv, psrc->p_waitcv, allowaddr);
2213 	COND_SET_VALUE(p->p_lwpcv, psrc->p_lwpcv, allowaddr);
2214 	COND_SET_VALUE(p->p_cred, psrc->p_cred, allowaddr);
2215 	COND_SET_VALUE(p->p_fd, psrc->p_fd, allowaddr);
2216 	COND_SET_VALUE(p->p_cwdi, psrc->p_cwdi, allowaddr);
2217 	COND_SET_VALUE(p->p_stats, psrc->p_stats, allowaddr);
2218 	COND_SET_VALUE(p->p_limit, psrc->p_limit, allowaddr);
2219 	COND_SET_VALUE(p->p_vmspace, psrc->p_vmspace, allowaddr);
2220 	COND_SET_VALUE(p->p_sigacts, psrc->p_sigacts, allowaddr);
2221 	COND_SET_VALUE(p->p_aio, psrc->p_aio, allowaddr);
2222 	p->p_mqueue_cnt = psrc->p_mqueue_cnt;
2223 	COND_SET_VALUE(p->p_specdataref, psrc->p_specdataref, allowaddr);
2224 	p->p_exitsig = psrc->p_exitsig;
2225 	p->p_flag = psrc->p_flag;
2226 	p->p_sflag = psrc->p_sflag;
2227 	p->p_slflag = psrc->p_slflag;
2228 	p->p_lflag = psrc->p_lflag;
2229 	p->p_stflag = psrc->p_stflag;
2230 	p->p_stat = psrc->p_stat;
2231 	p->p_trace_enabled = psrc->p_trace_enabled;
2232 	p->p_pid = psrc->p_pid;
2233 	COND_SET_VALUE(p->p_pglist, psrc->p_pglist, allowaddr);
2234 	COND_SET_VALUE(p->p_pptr, psrc->p_pptr, allowaddr);
2235 	COND_SET_VALUE(p->p_sibling, psrc->p_sibling, allowaddr);
2236 	COND_SET_VALUE(p->p_children, psrc->p_children, allowaddr);
2237 	COND_SET_VALUE(p->p_lwps, psrc->p_lwps, allowaddr);
2238 	COND_SET_VALUE(p->p_raslist, psrc->p_raslist, allowaddr);
2239 	p->p_nlwps = psrc->p_nlwps;
2240 	p->p_nzlwps = psrc->p_nzlwps;
2241 	p->p_nrlwps = psrc->p_nrlwps;
2242 	p->p_nlwpwait = psrc->p_nlwpwait;
2243 	p->p_ndlwps = psrc->p_ndlwps;
2244 	p->p_nlwpid = psrc->p_nlwpid;
2245 	p->p_nstopchild = psrc->p_nstopchild;
2246 	p->p_waited = psrc->p_waited;
2247 	COND_SET_VALUE(p->p_zomblwp, psrc->p_zomblwp, allowaddr);
2248 	COND_SET_VALUE(p->p_vforklwp, psrc->p_vforklwp, allowaddr);
2249 	COND_SET_VALUE(p->p_sched_info, psrc->p_sched_info, allowaddr);
2250 	p->p_estcpu = psrc->p_estcpu;
2251 	p->p_estcpu_inherited = psrc->p_estcpu_inherited;
2252 	p->p_forktime = psrc->p_forktime;
2253 	p->p_pctcpu = psrc->p_pctcpu;
2254 	COND_SET_VALUE(p->p_opptr, psrc->p_opptr, allowaddr);
2255 	COND_SET_VALUE(p->p_timers, psrc->p_timers, allowaddr);
2256 	p->p_rtime = psrc->p_rtime;
2257 	p->p_uticks = psrc->p_uticks;
2258 	p->p_sticks = psrc->p_sticks;
2259 	p->p_iticks = psrc->p_iticks;
2260 	p->p_xutime = psrc->p_xutime;
2261 	p->p_xstime = psrc->p_xstime;
2262 	p->p_traceflag = psrc->p_traceflag;
2263 	COND_SET_VALUE(p->p_tracep, psrc->p_tracep, allowaddr);
2264 	COND_SET_VALUE(p->p_textvp, psrc->p_textvp, allowaddr);
2265 	COND_SET_VALUE(p->p_emul, psrc->p_emul, allowaddr);
2266 	COND_SET_VALUE(p->p_emuldata, psrc->p_emuldata, allowaddr);
2267 	COND_SET_VALUE(p->p_execsw, psrc->p_execsw, allowaddr);
2268 	COND_SET_VALUE(p->p_klist, psrc->p_klist, allowaddr);
2269 	COND_SET_VALUE(p->p_sigwaiters, psrc->p_sigwaiters, allowaddr);
2270 	COND_SET_VALUE(p->p_sigpend, psrc->p_sigpend, allowaddr);
2271 	COND_SET_VALUE(p->p_lwpctl, psrc->p_lwpctl, allowaddr);
2272 	p->p_ppid = psrc->p_ppid;
2273 	COND_SET_VALUE(p->p_path, psrc->p_path, allowaddr);
2274 	COND_SET_VALUE(p->p_sigctx, psrc->p_sigctx, allowaddr);
2275 	p->p_nice = psrc->p_nice;
2276 	memcpy(p->p_comm, psrc->p_comm, sizeof(p->p_comm));
2277 	COND_SET_VALUE(p->p_pgrp, psrc->p_pgrp, allowaddr);
2278 	COND_SET_VALUE(p->p_psstrp, psrc->p_psstrp, allowaddr);
2279 	p->p_pax = psrc->p_pax;
2280 	p->p_xexit = psrc->p_xexit;
2281 	p->p_xsig = psrc->p_xsig;
2282 	p->p_acflag = psrc->p_acflag;
2283 	COND_SET_VALUE(p->p_md, psrc->p_md, allowaddr);
2284 	p->p_stackbase = psrc->p_stackbase;
2285 	COND_SET_VALUE(p->p_dtrace, psrc->p_dtrace, allowaddr);
2286 }
2287 
2288 /*
2289  * Fill in an eproc structure for the specified process.
2290  */
2291 void
2292 fill_eproc(struct proc *p, struct eproc *ep, bool zombie, bool allowaddr)
2293 {
2294 	struct tty *tp;
2295 	struct lwp *l;
2296 
2297 	KASSERT(mutex_owned(proc_lock));
2298 	KASSERT(mutex_owned(p->p_lock));
2299 
2300 	COND_SET_VALUE(ep->e_paddr, p, allowaddr);
2301 	COND_SET_VALUE(ep->e_sess, p->p_session, allowaddr);
2302 	if (p->p_cred) {
2303 		kauth_cred_topcred(p->p_cred, &ep->e_pcred);
2304 		kauth_cred_toucred(p->p_cred, &ep->e_ucred);
2305 	}
2306 	if (p->p_stat != SIDL && !P_ZOMBIE(p) && !zombie) {
2307 		struct vmspace *vm = p->p_vmspace;
2308 
2309 		ep->e_vm.vm_rssize = vm_resident_count(vm);
2310 		ep->e_vm.vm_tsize = vm->vm_tsize;
2311 		ep->e_vm.vm_dsize = vm->vm_dsize;
2312 		ep->e_vm.vm_ssize = vm->vm_ssize;
2313 		ep->e_vm.vm_map.size = vm->vm_map.size;
2314 
2315 		/* Pick the primary (first) LWP */
2316 		l = proc_active_lwp(p);
2317 		KASSERT(l != NULL);
2318 		lwp_lock(l);
2319 		if (l->l_wchan)
2320 			strncpy(ep->e_wmesg, l->l_wmesg, WMESGLEN);
2321 		lwp_unlock(l);
2322 	}
2323 	ep->e_ppid = p->p_ppid;
2324 	if (p->p_pgrp && p->p_session) {
2325 		ep->e_pgid = p->p_pgrp->pg_id;
2326 		ep->e_jobc = p->p_pgrp->pg_jobc;
2327 		ep->e_sid = p->p_session->s_sid;
2328 		if ((p->p_lflag & PL_CONTROLT) &&
2329 		    (tp = p->p_session->s_ttyp)) {
2330 			ep->e_tdev = tp->t_dev;
2331 			ep->e_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PGID;
2332 			COND_SET_VALUE(ep->e_tsess, tp->t_session, allowaddr);
2333 		} else
2334 			ep->e_tdev = (uint32_t)NODEV;
2335 		ep->e_flag = p->p_session->s_ttyvp ? EPROC_CTTY : 0;
2336 		if (SESS_LEADER(p))
2337 			ep->e_flag |= EPROC_SLEADER;
2338 		strncpy(ep->e_login, p->p_session->s_login, MAXLOGNAME);
2339 	}
2340 	ep->e_xsize = ep->e_xrssize = 0;
2341 	ep->e_xccount = ep->e_xswrss = 0;
2342 }
2343 
2344 /*
2345  * Fill in a kinfo_proc2 structure for the specified process.
2346  */
2347 void
2348 fill_kproc2(struct proc *p, struct kinfo_proc2 *ki, bool zombie, bool allowaddr)
2349 {
2350 	struct tty *tp;
2351 	struct lwp *l, *l2;
2352 	struct timeval ut, st, rt;
2353 	sigset_t ss1, ss2;
2354 	struct rusage ru;
2355 	struct vmspace *vm;
2356 
2357 	KASSERT(mutex_owned(proc_lock));
2358 	KASSERT(mutex_owned(p->p_lock));
2359 
2360 	sigemptyset(&ss1);
2361 	sigemptyset(&ss2);
2362 
2363 	COND_SET_VALUE(ki->p_paddr, PTRTOUINT64(p), allowaddr);
2364 	COND_SET_VALUE(ki->p_fd, PTRTOUINT64(p->p_fd), allowaddr);
2365 	COND_SET_VALUE(ki->p_cwdi, PTRTOUINT64(p->p_cwdi), allowaddr);
2366 	COND_SET_VALUE(ki->p_stats, PTRTOUINT64(p->p_stats), allowaddr);
2367 	COND_SET_VALUE(ki->p_limit, PTRTOUINT64(p->p_limit), allowaddr);
2368 	COND_SET_VALUE(ki->p_vmspace, PTRTOUINT64(p->p_vmspace), allowaddr);
2369 	COND_SET_VALUE(ki->p_sigacts, PTRTOUINT64(p->p_sigacts), allowaddr);
2370 	COND_SET_VALUE(ki->p_sess, PTRTOUINT64(p->p_session), allowaddr);
2371 	ki->p_tsess = 0;	/* may be changed if controlling tty below */
2372 	COND_SET_VALUE(ki->p_ru, PTRTOUINT64(&p->p_stats->p_ru), allowaddr);
2373 	ki->p_eflag = 0;
2374 	ki->p_exitsig = p->p_exitsig;
2375 	ki->p_flag = L_INMEM;   /* Process never swapped out */
2376 	ki->p_flag |= sysctl_map_flags(sysctl_flagmap, p->p_flag);
2377 	ki->p_flag |= sysctl_map_flags(sysctl_sflagmap, p->p_sflag);
2378 	ki->p_flag |= sysctl_map_flags(sysctl_slflagmap, p->p_slflag);
2379 	ki->p_flag |= sysctl_map_flags(sysctl_lflagmap, p->p_lflag);
2380 	ki->p_flag |= sysctl_map_flags(sysctl_stflagmap, p->p_stflag);
2381 	ki->p_pid = p->p_pid;
2382 	ki->p_ppid = p->p_ppid;
2383 	ki->p_uid = kauth_cred_geteuid(p->p_cred);
2384 	ki->p_ruid = kauth_cred_getuid(p->p_cred);
2385 	ki->p_gid = kauth_cred_getegid(p->p_cred);
2386 	ki->p_rgid = kauth_cred_getgid(p->p_cred);
2387 	ki->p_svuid = kauth_cred_getsvuid(p->p_cred);
2388 	ki->p_svgid = kauth_cred_getsvgid(p->p_cred);
2389 	ki->p_ngroups = kauth_cred_ngroups(p->p_cred);
2390 	kauth_cred_getgroups(p->p_cred, ki->p_groups,
2391 	    uimin(ki->p_ngroups, sizeof(ki->p_groups) / sizeof(ki->p_groups[0])),
2392 	    UIO_SYSSPACE);
2393 
2394 	ki->p_uticks = p->p_uticks;
2395 	ki->p_sticks = p->p_sticks;
2396 	ki->p_iticks = p->p_iticks;
2397 	ki->p_tpgid = NO_PGID;	/* may be changed if controlling tty below */
2398 	COND_SET_VALUE(ki->p_tracep, PTRTOUINT64(p->p_tracep), allowaddr);
2399 	ki->p_traceflag = p->p_traceflag;
2400 
2401 	memcpy(&ki->p_sigignore, &p->p_sigctx.ps_sigignore,sizeof(ki_sigset_t));
2402 	memcpy(&ki->p_sigcatch, &p->p_sigctx.ps_sigcatch, sizeof(ki_sigset_t));
2403 
2404 	ki->p_cpticks = 0;
2405 	ki->p_pctcpu = p->p_pctcpu;
2406 	ki->p_estcpu = 0;
2407 	ki->p_stat = p->p_stat; /* Will likely be overridden by LWP status */
2408 	ki->p_realstat = p->p_stat;
2409 	ki->p_nice = p->p_nice;
2410 	ki->p_xstat = P_WAITSTATUS(p);
2411 	ki->p_acflag = p->p_acflag;
2412 
2413 	strncpy(ki->p_comm, p->p_comm,
2414 	    uimin(sizeof(ki->p_comm), sizeof(p->p_comm)));
2415 	strncpy(ki->p_ename, p->p_emul->e_name, sizeof(ki->p_ename));
2416 
2417 	ki->p_nlwps = p->p_nlwps;
2418 	ki->p_realflag = ki->p_flag;
2419 
2420 	if (p->p_stat != SIDL && !P_ZOMBIE(p) && !zombie) {
2421 		vm = p->p_vmspace;
2422 		ki->p_vm_rssize = vm_resident_count(vm);
2423 		ki->p_vm_tsize = vm->vm_tsize;
2424 		ki->p_vm_dsize = vm->vm_dsize;
2425 		ki->p_vm_ssize = vm->vm_ssize;
2426 		ki->p_vm_vsize = atop(vm->vm_map.size);
2427 		/*
2428 		 * Since the stack is initially mapped mostly with
2429 		 * PROT_NONE and grown as needed, adjust the "mapped size"
2430 		 * to skip the unused stack portion.
2431 		 */
2432 		ki->p_vm_msize =
2433 		    atop(vm->vm_map.size) - vm->vm_issize + vm->vm_ssize;
2434 
2435 		/* Pick the primary (first) LWP */
2436 		l = proc_active_lwp(p);
2437 		KASSERT(l != NULL);
2438 		lwp_lock(l);
2439 		ki->p_nrlwps = p->p_nrlwps;
2440 		ki->p_forw = 0;
2441 		ki->p_back = 0;
2442 		COND_SET_VALUE(ki->p_addr, PTRTOUINT64(l->l_addr), allowaddr);
2443 		ki->p_stat = l->l_stat;
2444 		ki->p_flag |= sysctl_map_flags(sysctl_lwpflagmap, l->l_flag);
2445 		ki->p_swtime = l->l_swtime;
2446 		ki->p_slptime = l->l_slptime;
2447 		if (l->l_stat == LSONPROC)
2448 			ki->p_schedflags = l->l_cpu->ci_schedstate.spc_flags;
2449 		else
2450 			ki->p_schedflags = 0;
2451 		ki->p_priority = lwp_eprio(l);
2452 		ki->p_usrpri = l->l_priority;
2453 		if (l->l_wchan)
2454 			strncpy(ki->p_wmesg, l->l_wmesg, sizeof(ki->p_wmesg));
2455 		COND_SET_VALUE(ki->p_wchan, PTRTOUINT64(l->l_wchan), allowaddr);
2456 		ki->p_cpuid = cpu_index(l->l_cpu);
2457 		lwp_unlock(l);
2458 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
2459 			/* This is hardly correct, but... */
2460 			sigplusset(&l->l_sigpend.sp_set, &ss1);
2461 			sigplusset(&l->l_sigmask, &ss2);
2462 			ki->p_cpticks += l->l_cpticks;
2463 			ki->p_pctcpu += l->l_pctcpu;
2464 			ki->p_estcpu += l->l_estcpu;
2465 		}
2466 	}
2467 	sigplusset(&p->p_sigpend.sp_set, &ss1);
2468 	memcpy(&ki->p_siglist, &ss1, sizeof(ki_sigset_t));
2469 	memcpy(&ki->p_sigmask, &ss2, sizeof(ki_sigset_t));
2470 
2471 	if (p->p_session != NULL) {
2472 		ki->p_sid = p->p_session->s_sid;
2473 		ki->p__pgid = p->p_pgrp->pg_id;
2474 		if (p->p_session->s_ttyvp)
2475 			ki->p_eflag |= EPROC_CTTY;
2476 		if (SESS_LEADER(p))
2477 			ki->p_eflag |= EPROC_SLEADER;
2478 		strncpy(ki->p_login, p->p_session->s_login,
2479 		    uimin(sizeof ki->p_login - 1, sizeof p->p_session->s_login));
2480 		ki->p_jobc = p->p_pgrp->pg_jobc;
2481 		if ((p->p_lflag & PL_CONTROLT) && (tp = p->p_session->s_ttyp)) {
2482 			ki->p_tdev = tp->t_dev;
2483 			ki->p_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PGID;
2484 			COND_SET_VALUE(ki->p_tsess, PTRTOUINT64(tp->t_session),
2485 			    allowaddr);
2486 		} else {
2487 			ki->p_tdev = (int32_t)NODEV;
2488 		}
2489 	}
2490 
2491 	if (!P_ZOMBIE(p) && !zombie) {
2492 		ki->p_uvalid = 1;
2493 		ki->p_ustart_sec = p->p_stats->p_start.tv_sec;
2494 		ki->p_ustart_usec = p->p_stats->p_start.tv_usec;
2495 
2496 		calcru(p, &ut, &st, NULL, &rt);
2497 		ki->p_rtime_sec = rt.tv_sec;
2498 		ki->p_rtime_usec = rt.tv_usec;
2499 		ki->p_uutime_sec = ut.tv_sec;
2500 		ki->p_uutime_usec = ut.tv_usec;
2501 		ki->p_ustime_sec = st.tv_sec;
2502 		ki->p_ustime_usec = st.tv_usec;
2503 
2504 		memcpy(&ru, &p->p_stats->p_ru, sizeof(ru));
2505 		ki->p_uru_nvcsw = 0;
2506 		ki->p_uru_nivcsw = 0;
2507 		LIST_FOREACH(l2, &p->p_lwps, l_sibling) {
2508 			ki->p_uru_nvcsw += (l2->l_ncsw - l2->l_nivcsw);
2509 			ki->p_uru_nivcsw += l2->l_nivcsw;
2510 			ruadd(&ru, &l2->l_ru);
2511 		}
2512 		ki->p_uru_maxrss = ru.ru_maxrss;
2513 		ki->p_uru_ixrss = ru.ru_ixrss;
2514 		ki->p_uru_idrss = ru.ru_idrss;
2515 		ki->p_uru_isrss = ru.ru_isrss;
2516 		ki->p_uru_minflt = ru.ru_minflt;
2517 		ki->p_uru_majflt = ru.ru_majflt;
2518 		ki->p_uru_nswap = ru.ru_nswap;
2519 		ki->p_uru_inblock = ru.ru_inblock;
2520 		ki->p_uru_oublock = ru.ru_oublock;
2521 		ki->p_uru_msgsnd = ru.ru_msgsnd;
2522 		ki->p_uru_msgrcv = ru.ru_msgrcv;
2523 		ki->p_uru_nsignals = ru.ru_nsignals;
2524 
2525 		timeradd(&p->p_stats->p_cru.ru_utime,
2526 			 &p->p_stats->p_cru.ru_stime, &ut);
2527 		ki->p_uctime_sec = ut.tv_sec;
2528 		ki->p_uctime_usec = ut.tv_usec;
2529 	}
2530 }
2531 
2532 
2533 int
2534 proc_find_locked(struct lwp *l, struct proc **p, pid_t pid)
2535 {
2536 	int error;
2537 
2538 	mutex_enter(proc_lock);
2539 	if (pid == -1)
2540 		*p = l->l_proc;
2541 	else
2542 		*p = proc_find(pid);
2543 
2544 	if (*p == NULL) {
2545 		if (pid != -1)
2546 			mutex_exit(proc_lock);
2547 		return ESRCH;
2548 	}
2549 	if (pid != -1)
2550 		mutex_enter((*p)->p_lock);
2551 	mutex_exit(proc_lock);
2552 
2553 	error = kauth_authorize_process(l->l_cred,
2554 	    KAUTH_PROCESS_CANSEE, *p,
2555 	    KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL);
2556 	if (error) {
2557 		if (pid != -1)
2558 			mutex_exit((*p)->p_lock);
2559 	}
2560 	return error;
2561 }
2562 
2563 static int
2564 fill_pathname(struct lwp *l, pid_t pid, void *oldp, size_t *oldlenp)
2565 {
2566 	int error;
2567 	struct proc *p;
2568 
2569 	if ((error = proc_find_locked(l, &p, pid)) != 0)
2570 		return error;
2571 
2572 	if (p->p_path == NULL) {
2573 		if (pid != -1)
2574 			mutex_exit(p->p_lock);
2575 		return ENOENT;
2576 	}
2577 
2578 	size_t len = strlen(p->p_path) + 1;
2579 	if (oldp != NULL) {
2580 		size_t copylen = uimin(len, *oldlenp);
2581 		error = sysctl_copyout(l, p->p_path, oldp, copylen);
2582 		if (error == 0 && *oldlenp < len)
2583 			error = ENOSPC;
2584 	}
2585 	*oldlenp = len;
2586 	if (pid != -1)
2587 		mutex_exit(p->p_lock);
2588 	return error;
2589 }
2590 
2591 static int
2592 fill_cwd(struct lwp *l, pid_t pid, void *oldp, size_t *oldlenp)
2593 {
2594 	int error;
2595 	struct proc *p;
2596 	char *path;
2597 	char *bp, *bend;
2598 	const struct cwdinfo *cwdi;
2599 	struct vnode *vp;
2600 	size_t len, lenused;
2601 
2602 	if ((error = proc_find_locked(l, &p, pid)) != 0)
2603 		return error;
2604 
2605 	len = MAXPATHLEN * 4;
2606 
2607 	path = kmem_alloc(len, KM_SLEEP);
2608 
2609 	bp = &path[len];
2610 	bend = bp;
2611 	*(--bp) = '\0';
2612 
2613 	cwdi = cwdlock(p);
2614 	vp = cwdi->cwdi_cdir;
2615 	vref(vp);
2616 	cwdunlock(p);
2617 	error = getcwd_common(vp, NULL, &bp, path, len/2, 0, l);
2618 	vrele(vp);
2619 
2620 	if (error)
2621 		goto out;
2622 
2623 	lenused = bend - bp;
2624 
2625 	if (oldp != NULL) {
2626 		size_t copylen = uimin(lenused, *oldlenp);
2627 		error = sysctl_copyout(l, bp, oldp, copylen);
2628 		if (error == 0 && *oldlenp < lenused)
2629 			error = ENOSPC;
2630 	}
2631 	*oldlenp = lenused;
2632 out:
2633 	if (pid != -1)
2634 		mutex_exit(p->p_lock);
2635 	kmem_free(path, len);
2636 	return error;
2637 }
2638 
2639 int
2640 proc_getauxv(struct proc *p, void **buf, size_t *len)
2641 {
2642 	struct ps_strings pss;
2643 	int error;
2644 	void *uauxv, *kauxv;
2645 	size_t size;
2646 
2647 	if ((error = copyin_psstrings(p, &pss)) != 0)
2648 		return error;
2649 	if (pss.ps_envstr == NULL)
2650 		return EIO;
2651 
2652 	size = p->p_execsw->es_arglen;
2653 	if (size == 0)
2654 		return EIO;
2655 
2656 	size_t ptrsz = PROC_PTRSZ(p);
2657 	uauxv = (void *)((char *)pss.ps_envstr + (pss.ps_nenvstr + 1) * ptrsz);
2658 
2659 	kauxv = kmem_alloc(size, KM_SLEEP);
2660 
2661 	error = copyin_proc(p, uauxv, kauxv, size);
2662 	if (error) {
2663 		kmem_free(kauxv, size);
2664 		return error;
2665 	}
2666 
2667 	*buf = kauxv;
2668 	*len = size;
2669 
2670 	return 0;
2671 }
2672 
2673 
2674 static int
2675 sysctl_security_expose_address(SYSCTLFN_ARGS)
2676 {
2677 	int expose_address, error;
2678 	struct sysctlnode node;
2679 
2680 	node = *rnode;
2681 	node.sysctl_data = &expose_address;
2682 	expose_address = *(int *)rnode->sysctl_data;
2683 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
2684 	if (error || newp == NULL)
2685 		return error;
2686 
2687 	if (kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_KERNADDR,
2688 	    0, NULL, NULL, NULL))
2689 		return EPERM;
2690 
2691 	switch (expose_address) {
2692 	case 0:
2693 	case 1:
2694 	case 2:
2695 		break;
2696 	default:
2697 		return EINVAL;
2698 	}
2699 
2700 	*(int *)rnode->sysctl_data = expose_address;
2701 
2702 	return 0;
2703 }
2704 
2705 bool
2706 get_expose_address(struct proc *p)
2707 {
2708 	/* allow only if sysctl variable is set or privileged */
2709 	return kauth_authorize_process(kauth_cred_get(), KAUTH_PROCESS_CANSEE,
2710 	    p, KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_KPTR), NULL, NULL) == 0;
2711 }
2712