xref: /netbsd-src/sys/kern/kern_proc.c (revision 7330f729ccf0bd976a06f95fad452fe774fc7fd1)
1 /*	$NetBSD: kern_proc.c,v 1.236 2019/10/12 10:55:23 kamil 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 1.236 2019/10/12 10:55:23 kamil 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 
110 #include <uvm/uvm_extern.h>
111 #include <uvm/uvm.h>
112 
113 /*
114  * Process lists.
115  */
116 
117 struct proclist		allproc		__cacheline_aligned;
118 struct proclist		zombproc	__cacheline_aligned;
119 
120 kmutex_t *		proc_lock	__cacheline_aligned;
121 
122 /*
123  * pid to proc lookup is done by indexing the pid_table array.
124  * Since pid numbers are only allocated when an empty slot
125  * has been found, there is no need to search any lists ever.
126  * (an orphaned pgrp will lock the slot, a session will lock
127  * the pgrp with the same number.)
128  * If the table is too small it is reallocated with twice the
129  * previous size and the entries 'unzipped' into the two halves.
130  * A linked list of free entries is passed through the pt_proc
131  * field of 'free' items - set odd to be an invalid ptr.
132  */
133 
134 struct pid_table {
135 	struct proc	*pt_proc;
136 	struct pgrp	*pt_pgrp;
137 	pid_t		pt_pid;
138 };
139 #if 1	/* strongly typed cast - should be a noop */
140 static inline uint p2u(struct proc *p) { return (uint)(uintptr_t)p; }
141 #else
142 #define p2u(p) ((uint)p)
143 #endif
144 #define P_VALID(p) (!(p2u(p) & 1))
145 #define P_NEXT(p) (p2u(p) >> 1)
146 #define P_FREE(pid) ((struct proc *)(uintptr_t)((pid) << 1 | 1))
147 
148 /*
149  * Table of process IDs (PIDs).
150  */
151 static struct pid_table *pid_table	__read_mostly;
152 
153 #define	INITIAL_PID_TABLE_SIZE		(1 << 5)
154 
155 /* Table mask, threshold for growing and number of allocated PIDs. */
156 static u_int		pid_tbl_mask	__read_mostly;
157 static u_int		pid_alloc_lim	__read_mostly;
158 static u_int		pid_alloc_cnt	__cacheline_aligned;
159 
160 /* Next free, last free and maximum PIDs. */
161 static u_int		next_free_pt	__cacheline_aligned;
162 static u_int		last_free_pt	__cacheline_aligned;
163 static pid_t		pid_max		__read_mostly;
164 
165 /* Components of the first process -- never freed. */
166 
167 extern struct emul emul_netbsd;	/* defined in kern_exec.c */
168 
169 struct session session0 = {
170 	.s_count = 1,
171 	.s_sid = 0,
172 };
173 struct pgrp pgrp0 = {
174 	.pg_members = LIST_HEAD_INITIALIZER(&pgrp0.pg_members),
175 	.pg_session = &session0,
176 };
177 filedesc_t filedesc0;
178 struct cwdinfo cwdi0 = {
179 	.cwdi_cmask = CMASK,
180 	.cwdi_refcnt = 1,
181 };
182 struct plimit limit0;
183 struct pstats pstat0;
184 struct vmspace vmspace0;
185 struct sigacts sigacts0;
186 struct proc proc0 = {
187 	.p_lwps = LIST_HEAD_INITIALIZER(&proc0.p_lwps),
188 	.p_sigwaiters = LIST_HEAD_INITIALIZER(&proc0.p_sigwaiters),
189 	.p_nlwps = 1,
190 	.p_nrlwps = 1,
191 	.p_nlwpid = 1,		/* must match lwp0.l_lid */
192 	.p_pgrp = &pgrp0,
193 	.p_comm = "system",
194 	/*
195 	 * Set P_NOCLDWAIT so that kernel threads are reparented to init(8)
196 	 * when they exit.  init(8) can easily wait them out for us.
197 	 */
198 	.p_flag = PK_SYSTEM | PK_NOCLDWAIT,
199 	.p_stat = SACTIVE,
200 	.p_nice = NZERO,
201 	.p_emul = &emul_netbsd,
202 	.p_cwdi = &cwdi0,
203 	.p_limit = &limit0,
204 	.p_fd = &filedesc0,
205 	.p_vmspace = &vmspace0,
206 	.p_stats = &pstat0,
207 	.p_sigacts = &sigacts0,
208 #ifdef PROC0_MD_INITIALIZERS
209 	PROC0_MD_INITIALIZERS
210 #endif
211 };
212 kauth_cred_t cred0;
213 
214 static const int	nofile	= NOFILE;
215 static const int	maxuprc	= MAXUPRC;
216 
217 static int sysctl_doeproc(SYSCTLFN_PROTO);
218 static int sysctl_kern_proc_args(SYSCTLFN_PROTO);
219 static int sysctl_security_expose_address(SYSCTLFN_PROTO);
220 
221 #ifdef KASLR
222 static int kern_expose_address = 0;
223 #else
224 static int kern_expose_address = 1;
225 #endif
226 /*
227  * The process list descriptors, used during pid allocation and
228  * by sysctl.  No locking on this data structure is needed since
229  * it is completely static.
230  */
231 const struct proclist_desc proclists[] = {
232 	{ &allproc	},
233 	{ &zombproc	},
234 	{ NULL		},
235 };
236 
237 static struct pgrp *	pg_remove(pid_t);
238 static void		pg_delete(pid_t);
239 static void		orphanpg(struct pgrp *);
240 
241 static specificdata_domain_t proc_specificdata_domain;
242 
243 static pool_cache_t proc_cache;
244 
245 static kauth_listener_t proc_listener;
246 
247 static void fill_proc(const struct proc *, struct proc *, bool);
248 static int fill_pathname(struct lwp *, pid_t, void *, size_t *);
249 static int fill_cwd(struct lwp *, pid_t, void *, size_t *);
250 
251 static int
252 proc_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
253     void *arg0, void *arg1, void *arg2, void *arg3)
254 {
255 	struct proc *p;
256 	int result;
257 
258 	result = KAUTH_RESULT_DEFER;
259 	p = arg0;
260 
261 	switch (action) {
262 	case KAUTH_PROCESS_CANSEE: {
263 		enum kauth_process_req req;
264 
265 		req = (enum kauth_process_req)arg1;
266 
267 		switch (req) {
268 		case KAUTH_REQ_PROCESS_CANSEE_ARGS:
269 		case KAUTH_REQ_PROCESS_CANSEE_ENTRY:
270 		case KAUTH_REQ_PROCESS_CANSEE_OPENFILES:
271 		case KAUTH_REQ_PROCESS_CANSEE_EPROC:
272 			result = KAUTH_RESULT_ALLOW;
273 			break;
274 
275 		case KAUTH_REQ_PROCESS_CANSEE_ENV:
276 			if (kauth_cred_getuid(cred) !=
277 			    kauth_cred_getuid(p->p_cred) ||
278 			    kauth_cred_getuid(cred) !=
279 			    kauth_cred_getsvuid(p->p_cred))
280 				break;
281 
282 			result = KAUTH_RESULT_ALLOW;
283 
284 			break;
285 
286 		case KAUTH_REQ_PROCESS_CANSEE_KPTR:
287 			if (!kern_expose_address)
288 				break;
289 
290 			if (kern_expose_address == 1 && !(p->p_flag & PK_KMEM))
291 				break;
292 
293 			result = KAUTH_RESULT_ALLOW;
294 
295 			break;
296 
297 		default:
298 			break;
299 		}
300 
301 		break;
302 		}
303 
304 	case KAUTH_PROCESS_FORK: {
305 		int lnprocs = (int)(unsigned long)arg2;
306 
307 		/*
308 		 * Don't allow a nonprivileged user to use the last few
309 		 * processes. The variable lnprocs is the current number of
310 		 * processes, maxproc is the limit.
311 		 */
312 		if (__predict_false((lnprocs >= maxproc - 5)))
313 			break;
314 
315 		result = KAUTH_RESULT_ALLOW;
316 
317 		break;
318 		}
319 
320 	case KAUTH_PROCESS_CORENAME:
321 	case KAUTH_PROCESS_STOPFLAG:
322 		if (proc_uidmatch(cred, p->p_cred) == 0)
323 			result = KAUTH_RESULT_ALLOW;
324 
325 		break;
326 
327 	default:
328 		break;
329 	}
330 
331 	return result;
332 }
333 
334 static int
335 proc_ctor(void *arg __unused, void *obj, int flags __unused)
336 {
337 	memset(obj, 0, sizeof(struct proc));
338 	return 0;
339 }
340 
341 /*
342  * Initialize global process hashing structures.
343  */
344 void
345 procinit(void)
346 {
347 	const struct proclist_desc *pd;
348 	u_int i;
349 #define	LINK_EMPTY ((PID_MAX + INITIAL_PID_TABLE_SIZE) & ~(INITIAL_PID_TABLE_SIZE - 1))
350 
351 	for (pd = proclists; pd->pd_list != NULL; pd++)
352 		LIST_INIT(pd->pd_list);
353 
354 	proc_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
355 	pid_table = kmem_alloc(INITIAL_PID_TABLE_SIZE
356 	    * sizeof(struct pid_table), KM_SLEEP);
357 	pid_tbl_mask = INITIAL_PID_TABLE_SIZE - 1;
358 	pid_max = PID_MAX;
359 
360 	/* Set free list running through table...
361 	   Preset 'use count' above PID_MAX so we allocate pid 1 next. */
362 	for (i = 0; i <= pid_tbl_mask; i++) {
363 		pid_table[i].pt_proc = P_FREE(LINK_EMPTY + i + 1);
364 		pid_table[i].pt_pgrp = 0;
365 		pid_table[i].pt_pid = 0;
366 	}
367 	/* slot 0 is just grabbed */
368 	next_free_pt = 1;
369 	/* Need to fix last entry. */
370 	last_free_pt = pid_tbl_mask;
371 	pid_table[last_free_pt].pt_proc = P_FREE(LINK_EMPTY);
372 	/* point at which we grow table - to avoid reusing pids too often */
373 	pid_alloc_lim = pid_tbl_mask - 1;
374 #undef LINK_EMPTY
375 
376 	proc_specificdata_domain = specificdata_domain_create();
377 	KASSERT(proc_specificdata_domain != NULL);
378 
379 	proc_cache = pool_cache_init(sizeof(struct proc), 0, 0, 0,
380 	    "procpl", NULL, IPL_NONE, proc_ctor, NULL, NULL);
381 
382 	proc_listener = kauth_listen_scope(KAUTH_SCOPE_PROCESS,
383 	    proc_listener_cb, NULL);
384 }
385 
386 void
387 procinit_sysctl(void)
388 {
389 	static struct sysctllog *clog;
390 
391 	sysctl_createv(&clog, 0, NULL, NULL,
392 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
393 		       CTLTYPE_INT, "expose_address",
394 		       SYSCTL_DESCR("Enable exposing kernel addresses"),
395 		       sysctl_security_expose_address, 0,
396 		       &kern_expose_address, 0, CTL_KERN, CTL_CREATE, CTL_EOL);
397 	sysctl_createv(&clog, 0, NULL, NULL,
398 		       CTLFLAG_PERMANENT,
399 		       CTLTYPE_NODE, "proc",
400 		       SYSCTL_DESCR("System-wide process information"),
401 		       sysctl_doeproc, 0, NULL, 0,
402 		       CTL_KERN, KERN_PROC, CTL_EOL);
403 	sysctl_createv(&clog, 0, NULL, NULL,
404 		       CTLFLAG_PERMANENT,
405 		       CTLTYPE_NODE, "proc2",
406 		       SYSCTL_DESCR("Machine-independent process information"),
407 		       sysctl_doeproc, 0, NULL, 0,
408 		       CTL_KERN, KERN_PROC2, CTL_EOL);
409 	sysctl_createv(&clog, 0, NULL, NULL,
410 		       CTLFLAG_PERMANENT,
411 		       CTLTYPE_NODE, "proc_args",
412 		       SYSCTL_DESCR("Process argument information"),
413 		       sysctl_kern_proc_args, 0, NULL, 0,
414 		       CTL_KERN, KERN_PROC_ARGS, CTL_EOL);
415 
416 	/*
417 	  "nodes" under these:
418 
419 	  KERN_PROC_ALL
420 	  KERN_PROC_PID pid
421 	  KERN_PROC_PGRP pgrp
422 	  KERN_PROC_SESSION sess
423 	  KERN_PROC_TTY tty
424 	  KERN_PROC_UID uid
425 	  KERN_PROC_RUID uid
426 	  KERN_PROC_GID gid
427 	  KERN_PROC_RGID gid
428 
429 	  all in all, probably not worth the effort...
430 	*/
431 }
432 
433 /*
434  * Initialize process 0.
435  */
436 void
437 proc0_init(void)
438 {
439 	struct proc *p;
440 	struct pgrp *pg;
441 	struct rlimit *rlim;
442 	rlim_t lim;
443 	int i;
444 
445 	p = &proc0;
446 	pg = &pgrp0;
447 
448 	mutex_init(&p->p_stmutex, MUTEX_DEFAULT, IPL_HIGH);
449 	mutex_init(&p->p_auxlock, MUTEX_DEFAULT, IPL_NONE);
450 	p->p_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
451 
452 	rw_init(&p->p_reflock);
453 	cv_init(&p->p_waitcv, "wait");
454 	cv_init(&p->p_lwpcv, "lwpwait");
455 
456 	LIST_INSERT_HEAD(&p->p_lwps, &lwp0, l_sibling);
457 
458 	pid_table[0].pt_proc = p;
459 	LIST_INSERT_HEAD(&allproc, p, p_list);
460 
461 	pid_table[0].pt_pgrp = pg;
462 	LIST_INSERT_HEAD(&pg->pg_members, p, p_pglist);
463 
464 #ifdef __HAVE_SYSCALL_INTERN
465 	(*p->p_emul->e_syscall_intern)(p);
466 #endif
467 
468 	/* Create credentials. */
469 	cred0 = kauth_cred_alloc();
470 	p->p_cred = cred0;
471 
472 	/* Create the CWD info. */
473 	rw_init(&cwdi0.cwdi_lock);
474 
475 	/* Create the limits structures. */
476 	mutex_init(&limit0.pl_lock, MUTEX_DEFAULT, IPL_NONE);
477 
478 	rlim = limit0.pl_rlimit;
479 	for (i = 0; i < __arraycount(limit0.pl_rlimit); i++) {
480 		rlim[i].rlim_cur = RLIM_INFINITY;
481 		rlim[i].rlim_max = RLIM_INFINITY;
482 	}
483 
484 	rlim[RLIMIT_NOFILE].rlim_max = maxfiles;
485 	rlim[RLIMIT_NOFILE].rlim_cur = maxfiles < nofile ? maxfiles : nofile;
486 
487 	rlim[RLIMIT_NPROC].rlim_max = maxproc;
488 	rlim[RLIMIT_NPROC].rlim_cur = maxproc < maxuprc ? maxproc : maxuprc;
489 
490 	lim = MIN(VM_MAXUSER_ADDRESS, ctob((rlim_t)uvmexp.free));
491 	rlim[RLIMIT_RSS].rlim_max = lim;
492 	rlim[RLIMIT_MEMLOCK].rlim_max = lim;
493 	rlim[RLIMIT_MEMLOCK].rlim_cur = lim / 3;
494 
495 	rlim[RLIMIT_NTHR].rlim_max = maxlwp;
496 	rlim[RLIMIT_NTHR].rlim_cur = maxlwp < maxuprc ? maxlwp : maxuprc;
497 
498 	/* Note that default core name has zero length. */
499 	limit0.pl_corename = defcorename;
500 	limit0.pl_cnlen = 0;
501 	limit0.pl_refcnt = 1;
502 	limit0.pl_writeable = false;
503 	limit0.pl_sv_limit = NULL;
504 
505 	/* Configure virtual memory system, set vm rlimits. */
506 	uvm_init_limits(p);
507 
508 	/* Initialize file descriptor table for proc0. */
509 	fd_init(&filedesc0);
510 
511 	/*
512 	 * Initialize proc0's vmspace, which uses the kernel pmap.
513 	 * All kernel processes (which never have user space mappings)
514 	 * share proc0's vmspace, and thus, the kernel pmap.
515 	 */
516 	uvmspace_init(&vmspace0, pmap_kernel(), round_page(VM_MIN_ADDRESS),
517 	    trunc_page(VM_MAXUSER_ADDRESS),
518 #ifdef __USE_TOPDOWN_VM
519 	    true
520 #else
521 	    false
522 #endif
523 	    );
524 
525 	/* Initialize signal state for proc0. XXX IPL_SCHED */
526 	mutex_init(&p->p_sigacts->sa_mutex, MUTEX_DEFAULT, IPL_SCHED);
527 	siginit(p);
528 
529 	proc_initspecific(p);
530 	kdtrace_proc_ctor(NULL, p);
531 }
532 
533 /*
534  * Session reference counting.
535  */
536 
537 void
538 proc_sesshold(struct session *ss)
539 {
540 
541 	KASSERT(mutex_owned(proc_lock));
542 	ss->s_count++;
543 }
544 
545 void
546 proc_sessrele(struct session *ss)
547 {
548 
549 	KASSERT(mutex_owned(proc_lock));
550 	/*
551 	 * We keep the pgrp with the same id as the session in order to
552 	 * stop a process being given the same pid.  Since the pgrp holds
553 	 * a reference to the session, it must be a 'zombie' pgrp by now.
554 	 */
555 	if (--ss->s_count == 0) {
556 		struct pgrp *pg;
557 
558 		pg = pg_remove(ss->s_sid);
559 		mutex_exit(proc_lock);
560 
561 		kmem_free(pg, sizeof(struct pgrp));
562 		kmem_free(ss, sizeof(struct session));
563 	} else {
564 		mutex_exit(proc_lock);
565 	}
566 }
567 
568 /*
569  * Check that the specified process group is in the session of the
570  * specified process.
571  * Treats -ve ids as process ids.
572  * Used to validate TIOCSPGRP requests.
573  */
574 int
575 pgid_in_session(struct proc *p, pid_t pg_id)
576 {
577 	struct pgrp *pgrp;
578 	struct session *session;
579 	int error;
580 
581 	mutex_enter(proc_lock);
582 	if (pg_id < 0) {
583 		struct proc *p1 = proc_find(-pg_id);
584 		if (p1 == NULL) {
585 			error = EINVAL;
586 			goto fail;
587 		}
588 		pgrp = p1->p_pgrp;
589 	} else {
590 		pgrp = pgrp_find(pg_id);
591 		if (pgrp == NULL) {
592 			error = EINVAL;
593 			goto fail;
594 		}
595 	}
596 	session = pgrp->pg_session;
597 	error = (session != p->p_pgrp->pg_session) ? EPERM : 0;
598 fail:
599 	mutex_exit(proc_lock);
600 	return error;
601 }
602 
603 /*
604  * p_inferior: is p an inferior of q?
605  */
606 static inline bool
607 p_inferior(struct proc *p, struct proc *q)
608 {
609 
610 	KASSERT(mutex_owned(proc_lock));
611 
612 	for (; p != q; p = p->p_pptr)
613 		if (p->p_pid == 0)
614 			return false;
615 	return true;
616 }
617 
618 /*
619  * proc_find: locate a process by the ID.
620  *
621  * => Must be called with proc_lock held.
622  */
623 proc_t *
624 proc_find_raw(pid_t pid)
625 {
626 	struct pid_table *pt;
627 	proc_t *p;
628 
629 	KASSERT(mutex_owned(proc_lock));
630 	pt = &pid_table[pid & pid_tbl_mask];
631 	p = pt->pt_proc;
632 	if (__predict_false(!P_VALID(p) || pt->pt_pid != pid)) {
633 		return NULL;
634 	}
635 	return p;
636 }
637 
638 proc_t *
639 proc_find(pid_t pid)
640 {
641 	proc_t *p;
642 
643 	p = proc_find_raw(pid);
644 	if (__predict_false(p == NULL)) {
645 		return NULL;
646 	}
647 
648 	/*
649 	 * Only allow live processes to be found by PID.
650 	 * XXX: p_stat might change, since unlocked.
651 	 */
652 	if (__predict_true(p->p_stat == SACTIVE || p->p_stat == SSTOP)) {
653 		return p;
654 	}
655 	return NULL;
656 }
657 
658 /*
659  * pgrp_find: locate a process group by the ID.
660  *
661  * => Must be called with proc_lock held.
662  */
663 struct pgrp *
664 pgrp_find(pid_t pgid)
665 {
666 	struct pgrp *pg;
667 
668 	KASSERT(mutex_owned(proc_lock));
669 
670 	pg = pid_table[pgid & pid_tbl_mask].pt_pgrp;
671 
672 	/*
673 	 * Cannot look up a process group that only exists because the
674 	 * session has not died yet (traditional).
675 	 */
676 	if (pg == NULL || pg->pg_id != pgid || LIST_EMPTY(&pg->pg_members)) {
677 		return NULL;
678 	}
679 	return pg;
680 }
681 
682 static void
683 expand_pid_table(void)
684 {
685 	size_t pt_size, tsz;
686 	struct pid_table *n_pt, *new_pt;
687 	struct proc *proc;
688 	struct pgrp *pgrp;
689 	pid_t pid, rpid;
690 	u_int i;
691 	uint new_pt_mask;
692 
693 	pt_size = pid_tbl_mask + 1;
694 	tsz = pt_size * 2 * sizeof(struct pid_table);
695 	new_pt = kmem_alloc(tsz, KM_SLEEP);
696 	new_pt_mask = pt_size * 2 - 1;
697 
698 	mutex_enter(proc_lock);
699 	if (pt_size != pid_tbl_mask + 1) {
700 		/* Another process beat us to it... */
701 		mutex_exit(proc_lock);
702 		kmem_free(new_pt, tsz);
703 		return;
704 	}
705 
706 	/*
707 	 * Copy entries from old table into new one.
708 	 * If 'pid' is 'odd' we need to place in the upper half,
709 	 * even pid's to the lower half.
710 	 * Free items stay in the low half so we don't have to
711 	 * fixup the reference to them.
712 	 * We stuff free items on the front of the freelist
713 	 * because we can't write to unmodified entries.
714 	 * Processing the table backwards maintains a semblance
715 	 * of issuing pid numbers that increase with time.
716 	 */
717 	i = pt_size - 1;
718 	n_pt = new_pt + i;
719 	for (; ; i--, n_pt--) {
720 		proc = pid_table[i].pt_proc;
721 		pgrp = pid_table[i].pt_pgrp;
722 		if (!P_VALID(proc)) {
723 			/* Up 'use count' so that link is valid */
724 			pid = (P_NEXT(proc) + pt_size) & ~pt_size;
725 			rpid = 0;
726 			proc = P_FREE(pid);
727 			if (pgrp)
728 				pid = pgrp->pg_id;
729 		} else {
730 			pid = pid_table[i].pt_pid;
731 			rpid = pid;
732 		}
733 
734 		/* Save entry in appropriate half of table */
735 		n_pt[pid & pt_size].pt_proc = proc;
736 		n_pt[pid & pt_size].pt_pgrp = pgrp;
737 		n_pt[pid & pt_size].pt_pid = rpid;
738 
739 		/* Put other piece on start of free list */
740 		pid = (pid ^ pt_size) & ~pid_tbl_mask;
741 		n_pt[pid & pt_size].pt_proc =
742 			P_FREE((pid & ~pt_size) | next_free_pt);
743 		n_pt[pid & pt_size].pt_pgrp = 0;
744 		n_pt[pid & pt_size].pt_pid = 0;
745 
746 		next_free_pt = i | (pid & pt_size);
747 		if (i == 0)
748 			break;
749 	}
750 
751 	/* Save old table size and switch tables */
752 	tsz = pt_size * sizeof(struct pid_table);
753 	n_pt = pid_table;
754 	pid_table = new_pt;
755 	pid_tbl_mask = new_pt_mask;
756 
757 	/*
758 	 * pid_max starts as PID_MAX (= 30000), once we have 16384
759 	 * allocated pids we need it to be larger!
760 	 */
761 	if (pid_tbl_mask > PID_MAX) {
762 		pid_max = pid_tbl_mask * 2 + 1;
763 		pid_alloc_lim |= pid_alloc_lim << 1;
764 	} else
765 		pid_alloc_lim <<= 1;	/* doubles number of free slots... */
766 
767 	mutex_exit(proc_lock);
768 	kmem_free(n_pt, tsz);
769 }
770 
771 struct proc *
772 proc_alloc(void)
773 {
774 	struct proc *p;
775 
776 	p = pool_cache_get(proc_cache, PR_WAITOK);
777 	p->p_stat = SIDL;			/* protect against others */
778 	proc_initspecific(p);
779 	kdtrace_proc_ctor(NULL, p);
780 	p->p_pid = -1;
781 	proc_alloc_pid(p);
782 	return p;
783 }
784 
785 /*
786  * proc_alloc_pid: allocate PID and record the given proc 'p' so that
787  * proc_find_raw() can find it by the PID.
788  */
789 
790 pid_t
791 proc_alloc_pid(struct proc *p)
792 {
793 	struct pid_table *pt;
794 	pid_t pid;
795 	int nxt;
796 
797 	for (;;expand_pid_table()) {
798 		if (__predict_false(pid_alloc_cnt >= pid_alloc_lim))
799 			/* ensure pids cycle through 2000+ values */
800 			continue;
801 		mutex_enter(proc_lock);
802 		pt = &pid_table[next_free_pt];
803 #ifdef DIAGNOSTIC
804 		if (__predict_false(P_VALID(pt->pt_proc) || pt->pt_pgrp))
805 			panic("proc_alloc: slot busy");
806 #endif
807 		nxt = P_NEXT(pt->pt_proc);
808 		if (nxt & pid_tbl_mask)
809 			break;
810 		/* Table full - expand (NB last entry not used....) */
811 		mutex_exit(proc_lock);
812 	}
813 
814 	/* pid is 'saved use count' + 'size' + entry */
815 	pid = (nxt & ~pid_tbl_mask) + pid_tbl_mask + 1 + next_free_pt;
816 	if ((uint)pid > (uint)pid_max)
817 		pid &= pid_tbl_mask;
818 	next_free_pt = nxt & pid_tbl_mask;
819 
820 	/* Grab table slot */
821 	pt->pt_proc = p;
822 
823 	KASSERT(pt->pt_pid == 0);
824 	pt->pt_pid = pid;
825 	if (p->p_pid == -1) {
826 		p->p_pid = pid;
827 	}
828 	pid_alloc_cnt++;
829 	mutex_exit(proc_lock);
830 
831 	return pid;
832 }
833 
834 /*
835  * Free a process id - called from proc_free (in kern_exit.c)
836  *
837  * Called with the proc_lock held.
838  */
839 void
840 proc_free_pid(pid_t pid)
841 {
842 	struct pid_table *pt;
843 
844 	KASSERT(mutex_owned(proc_lock));
845 
846 	pt = &pid_table[pid & pid_tbl_mask];
847 
848 	/* save pid use count in slot */
849 	pt->pt_proc = P_FREE(pid & ~pid_tbl_mask);
850 	KASSERT(pt->pt_pid == pid);
851 	pt->pt_pid = 0;
852 
853 	if (pt->pt_pgrp == NULL) {
854 		/* link last freed entry onto ours */
855 		pid &= pid_tbl_mask;
856 		pt = &pid_table[last_free_pt];
857 		pt->pt_proc = P_FREE(P_NEXT(pt->pt_proc) | pid);
858 		pt->pt_pid = 0;
859 		last_free_pt = pid;
860 		pid_alloc_cnt--;
861 	}
862 
863 	atomic_dec_uint(&nprocs);
864 }
865 
866 void
867 proc_free_mem(struct proc *p)
868 {
869 
870 	kdtrace_proc_dtor(NULL, p);
871 	pool_cache_put(proc_cache, p);
872 }
873 
874 /*
875  * proc_enterpgrp: move p to a new or existing process group (and session).
876  *
877  * If we are creating a new pgrp, the pgid should equal
878  * the calling process' pid.
879  * If is only valid to enter a process group that is in the session
880  * of the process.
881  * Also mksess should only be set if we are creating a process group
882  *
883  * Only called from sys_setsid, sys_setpgid and posix_spawn/spawn_return.
884  */
885 int
886 proc_enterpgrp(struct proc *curp, pid_t pid, pid_t pgid, bool mksess)
887 {
888 	struct pgrp *new_pgrp, *pgrp;
889 	struct session *sess;
890 	struct proc *p;
891 	int rval;
892 	pid_t pg_id = NO_PGID;
893 
894 	sess = mksess ? kmem_alloc(sizeof(*sess), KM_SLEEP) : NULL;
895 
896 	/* Allocate data areas we might need before doing any validity checks */
897 	mutex_enter(proc_lock);		/* Because pid_table might change */
898 	if (pid_table[pgid & pid_tbl_mask].pt_pgrp == 0) {
899 		mutex_exit(proc_lock);
900 		new_pgrp = kmem_alloc(sizeof(*new_pgrp), KM_SLEEP);
901 		mutex_enter(proc_lock);
902 	} else
903 		new_pgrp = NULL;
904 	rval = EPERM;	/* most common error (to save typing) */
905 
906 	/* Check pgrp exists or can be created */
907 	pgrp = pid_table[pgid & pid_tbl_mask].pt_pgrp;
908 	if (pgrp != NULL && pgrp->pg_id != pgid)
909 		goto done;
910 
911 	/* Can only set another process under restricted circumstances. */
912 	if (pid != curp->p_pid) {
913 		/* Must exist and be one of our children... */
914 		p = proc_find(pid);
915 		if (p == NULL || !p_inferior(p, curp)) {
916 			rval = ESRCH;
917 			goto done;
918 		}
919 		/* ... in the same session... */
920 		if (sess != NULL || p->p_session != curp->p_session)
921 			goto done;
922 		/* ... existing pgid must be in same session ... */
923 		if (pgrp != NULL && pgrp->pg_session != p->p_session)
924 			goto done;
925 		/* ... and not done an exec. */
926 		if (p->p_flag & PK_EXEC) {
927 			rval = EACCES;
928 			goto done;
929 		}
930 	} else {
931 		/* ... setsid() cannot re-enter a pgrp */
932 		if (mksess && (curp->p_pgid == curp->p_pid ||
933 		    pgrp_find(curp->p_pid)))
934 			goto done;
935 		p = curp;
936 	}
937 
938 	/* Changing the process group/session of a session
939 	   leader is definitely off limits. */
940 	if (SESS_LEADER(p)) {
941 		if (sess == NULL && p->p_pgrp == pgrp)
942 			/* unless it's a definite noop */
943 			rval = 0;
944 		goto done;
945 	}
946 
947 	/* Can only create a process group with id of process */
948 	if (pgrp == NULL && pgid != pid)
949 		goto done;
950 
951 	/* Can only create a session if creating pgrp */
952 	if (sess != NULL && pgrp != NULL)
953 		goto done;
954 
955 	/* Check we allocated memory for a pgrp... */
956 	if (pgrp == NULL && new_pgrp == NULL)
957 		goto done;
958 
959 	/* Don't attach to 'zombie' pgrp */
960 	if (pgrp != NULL && LIST_EMPTY(&pgrp->pg_members))
961 		goto done;
962 
963 	/* Expect to succeed now */
964 	rval = 0;
965 
966 	if (pgrp == p->p_pgrp)
967 		/* nothing to do */
968 		goto done;
969 
970 	/* Ok all setup, link up required structures */
971 
972 	if (pgrp == NULL) {
973 		pgrp = new_pgrp;
974 		new_pgrp = NULL;
975 		if (sess != NULL) {
976 			sess->s_sid = p->p_pid;
977 			sess->s_leader = p;
978 			sess->s_count = 1;
979 			sess->s_ttyvp = NULL;
980 			sess->s_ttyp = NULL;
981 			sess->s_flags = p->p_session->s_flags & ~S_LOGIN_SET;
982 			memcpy(sess->s_login, p->p_session->s_login,
983 			    sizeof(sess->s_login));
984 			p->p_lflag &= ~PL_CONTROLT;
985 		} else {
986 			sess = p->p_pgrp->pg_session;
987 			proc_sesshold(sess);
988 		}
989 		pgrp->pg_session = sess;
990 		sess = NULL;
991 
992 		pgrp->pg_id = pgid;
993 		LIST_INIT(&pgrp->pg_members);
994 #ifdef DIAGNOSTIC
995 		if (__predict_false(pid_table[pgid & pid_tbl_mask].pt_pgrp))
996 			panic("enterpgrp: pgrp table slot in use");
997 		if (__predict_false(mksess && p != curp))
998 			panic("enterpgrp: mksession and p != curproc");
999 #endif
1000 		pid_table[pgid & pid_tbl_mask].pt_pgrp = pgrp;
1001 		pgrp->pg_jobc = 0;
1002 	}
1003 
1004 	/*
1005 	 * Adjust eligibility of affected pgrps to participate in job control.
1006 	 * Increment eligibility counts before decrementing, otherwise we
1007 	 * could reach 0 spuriously during the first call.
1008 	 */
1009 	fixjobc(p, pgrp, 1);
1010 	fixjobc(p, p->p_pgrp, 0);
1011 
1012 	/* Interlock with ttread(). */
1013 	mutex_spin_enter(&tty_lock);
1014 
1015 	/* Move process to requested group. */
1016 	LIST_REMOVE(p, p_pglist);
1017 	if (LIST_EMPTY(&p->p_pgrp->pg_members))
1018 		/* defer delete until we've dumped the lock */
1019 		pg_id = p->p_pgrp->pg_id;
1020 	p->p_pgrp = pgrp;
1021 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
1022 
1023 	/* Done with the swap; we can release the tty mutex. */
1024 	mutex_spin_exit(&tty_lock);
1025 
1026     done:
1027 	if (pg_id != NO_PGID) {
1028 		/* Releases proc_lock. */
1029 		pg_delete(pg_id);
1030 	} else {
1031 		mutex_exit(proc_lock);
1032 	}
1033 	if (sess != NULL)
1034 		kmem_free(sess, sizeof(*sess));
1035 	if (new_pgrp != NULL)
1036 		kmem_free(new_pgrp, sizeof(*new_pgrp));
1037 #ifdef DEBUG_PGRP
1038 	if (__predict_false(rval))
1039 		printf("enterpgrp(%d,%d,%d), curproc %d, rval %d\n",
1040 			pid, pgid, mksess, curp->p_pid, rval);
1041 #endif
1042 	return rval;
1043 }
1044 
1045 /*
1046  * proc_leavepgrp: remove a process from its process group.
1047  *  => must be called with the proc_lock held, which will be released;
1048  */
1049 void
1050 proc_leavepgrp(struct proc *p)
1051 {
1052 	struct pgrp *pgrp;
1053 
1054 	KASSERT(mutex_owned(proc_lock));
1055 
1056 	/* Interlock with ttread() */
1057 	mutex_spin_enter(&tty_lock);
1058 	pgrp = p->p_pgrp;
1059 	LIST_REMOVE(p, p_pglist);
1060 	p->p_pgrp = NULL;
1061 	mutex_spin_exit(&tty_lock);
1062 
1063 	if (LIST_EMPTY(&pgrp->pg_members)) {
1064 		/* Releases proc_lock. */
1065 		pg_delete(pgrp->pg_id);
1066 	} else {
1067 		mutex_exit(proc_lock);
1068 	}
1069 }
1070 
1071 /*
1072  * pg_remove: remove a process group from the table.
1073  *  => must be called with the proc_lock held;
1074  *  => returns process group to free;
1075  */
1076 static struct pgrp *
1077 pg_remove(pid_t pg_id)
1078 {
1079 	struct pgrp *pgrp;
1080 	struct pid_table *pt;
1081 
1082 	KASSERT(mutex_owned(proc_lock));
1083 
1084 	pt = &pid_table[pg_id & pid_tbl_mask];
1085 	pgrp = pt->pt_pgrp;
1086 
1087 	KASSERT(pgrp != NULL);
1088 	KASSERT(pgrp->pg_id == pg_id);
1089 	KASSERT(LIST_EMPTY(&pgrp->pg_members));
1090 
1091 	pt->pt_pgrp = NULL;
1092 
1093 	if (!P_VALID(pt->pt_proc)) {
1094 		/* Orphaned pgrp, put slot onto free list. */
1095 		KASSERT((P_NEXT(pt->pt_proc) & pid_tbl_mask) == 0);
1096 		pg_id &= pid_tbl_mask;
1097 		pt = &pid_table[last_free_pt];
1098 		pt->pt_proc = P_FREE(P_NEXT(pt->pt_proc) | pg_id);
1099 		KASSERT(pt->pt_pid == 0);
1100 		last_free_pt = pg_id;
1101 		pid_alloc_cnt--;
1102 	}
1103 	return pgrp;
1104 }
1105 
1106 /*
1107  * pg_delete: delete and free a process group.
1108  *  => must be called with the proc_lock held, which will be released.
1109  */
1110 static void
1111 pg_delete(pid_t pg_id)
1112 {
1113 	struct pgrp *pg;
1114 	struct tty *ttyp;
1115 	struct session *ss;
1116 
1117 	KASSERT(mutex_owned(proc_lock));
1118 
1119 	pg = pid_table[pg_id & pid_tbl_mask].pt_pgrp;
1120 	if (pg == NULL || pg->pg_id != pg_id || !LIST_EMPTY(&pg->pg_members)) {
1121 		mutex_exit(proc_lock);
1122 		return;
1123 	}
1124 
1125 	ss = pg->pg_session;
1126 
1127 	/* Remove reference (if any) from tty to this process group */
1128 	mutex_spin_enter(&tty_lock);
1129 	ttyp = ss->s_ttyp;
1130 	if (ttyp != NULL && ttyp->t_pgrp == pg) {
1131 		ttyp->t_pgrp = NULL;
1132 		KASSERT(ttyp->t_session == ss);
1133 	}
1134 	mutex_spin_exit(&tty_lock);
1135 
1136 	/*
1137 	 * The leading process group in a session is freed by proc_sessrele(),
1138 	 * if last reference.  Note: proc_sessrele() releases proc_lock.
1139 	 */
1140 	pg = (ss->s_sid != pg->pg_id) ? pg_remove(pg_id) : NULL;
1141 	proc_sessrele(ss);
1142 
1143 	if (pg != NULL) {
1144 		/* Free it, if was not done by proc_sessrele(). */
1145 		kmem_free(pg, sizeof(struct pgrp));
1146 	}
1147 }
1148 
1149 /*
1150  * Adjust pgrp jobc counters when specified process changes process group.
1151  * We count the number of processes in each process group that "qualify"
1152  * the group for terminal job control (those with a parent in a different
1153  * process group of the same session).  If that count reaches zero, the
1154  * process group becomes orphaned.  Check both the specified process'
1155  * process group and that of its children.
1156  * entering == 0 => p is leaving specified group.
1157  * entering == 1 => p is entering specified group.
1158  *
1159  * Call with proc_lock held.
1160  */
1161 void
1162 fixjobc(struct proc *p, struct pgrp *pgrp, int entering)
1163 {
1164 	struct pgrp *hispgrp;
1165 	struct session *mysession = pgrp->pg_session;
1166 	struct proc *child;
1167 
1168 	KASSERT(mutex_owned(proc_lock));
1169 
1170 	/*
1171 	 * Check p's parent to see whether p qualifies its own process
1172 	 * group; if so, adjust count for p's process group.
1173 	 */
1174 	hispgrp = p->p_pptr->p_pgrp;
1175 	if (hispgrp != pgrp && hispgrp->pg_session == mysession) {
1176 		if (entering) {
1177 			pgrp->pg_jobc++;
1178 			p->p_lflag &= ~PL_ORPHANPG;
1179 		} else if (--pgrp->pg_jobc == 0)
1180 			orphanpg(pgrp);
1181 	}
1182 
1183 	/*
1184 	 * Check this process' children to see whether they qualify
1185 	 * their process groups; if so, adjust counts for children's
1186 	 * process groups.
1187 	 */
1188 	LIST_FOREACH(child, &p->p_children, p_sibling) {
1189 		hispgrp = child->p_pgrp;
1190 		if (hispgrp != pgrp && hispgrp->pg_session == mysession &&
1191 		    !P_ZOMBIE(child)) {
1192 			if (entering) {
1193 				child->p_lflag &= ~PL_ORPHANPG;
1194 				hispgrp->pg_jobc++;
1195 			} else if (--hispgrp->pg_jobc == 0)
1196 				orphanpg(hispgrp);
1197 		}
1198 	}
1199 }
1200 
1201 /*
1202  * A process group has become orphaned;
1203  * if there are any stopped processes in the group,
1204  * hang-up all process in that group.
1205  *
1206  * Call with proc_lock held.
1207  */
1208 static void
1209 orphanpg(struct pgrp *pg)
1210 {
1211 	struct proc *p;
1212 
1213 	KASSERT(mutex_owned(proc_lock));
1214 
1215 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
1216 		if (p->p_stat == SSTOP) {
1217 			p->p_lflag |= PL_ORPHANPG;
1218 			psignal(p, SIGHUP);
1219 			psignal(p, SIGCONT);
1220 		}
1221 	}
1222 }
1223 
1224 #ifdef DDB
1225 #include <ddb/db_output.h>
1226 void pidtbl_dump(void);
1227 void
1228 pidtbl_dump(void)
1229 {
1230 	struct pid_table *pt;
1231 	struct proc *p;
1232 	struct pgrp *pgrp;
1233 	int id;
1234 
1235 	db_printf("pid table %p size %x, next %x, last %x\n",
1236 		pid_table, pid_tbl_mask+1,
1237 		next_free_pt, last_free_pt);
1238 	for (pt = pid_table, id = 0; id <= pid_tbl_mask; id++, pt++) {
1239 		p = pt->pt_proc;
1240 		if (!P_VALID(p) && !pt->pt_pgrp)
1241 			continue;
1242 		db_printf("  id %x: ", id);
1243 		if (P_VALID(p))
1244 			db_printf("slotpid %d proc %p id %d (0x%x) %s\n",
1245 				pt->pt_pid, p, p->p_pid, p->p_pid, p->p_comm);
1246 		else
1247 			db_printf("next %x use %x\n",
1248 				P_NEXT(p) & pid_tbl_mask,
1249 				P_NEXT(p) & ~pid_tbl_mask);
1250 		if ((pgrp = pt->pt_pgrp)) {
1251 			db_printf("\tsession %p, sid %d, count %d, login %s\n",
1252 			    pgrp->pg_session, pgrp->pg_session->s_sid,
1253 			    pgrp->pg_session->s_count,
1254 			    pgrp->pg_session->s_login);
1255 			db_printf("\tpgrp %p, pg_id %d, pg_jobc %d, members %p\n",
1256 			    pgrp, pgrp->pg_id, pgrp->pg_jobc,
1257 			    LIST_FIRST(&pgrp->pg_members));
1258 			LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
1259 				db_printf("\t\tpid %d addr %p pgrp %p %s\n",
1260 				    p->p_pid, p, p->p_pgrp, p->p_comm);
1261 			}
1262 		}
1263 	}
1264 }
1265 #endif /* DDB */
1266 
1267 #ifdef KSTACK_CHECK_MAGIC
1268 
1269 #define	KSTACK_MAGIC	0xdeadbeaf
1270 
1271 /* XXX should be per process basis? */
1272 static int	kstackleftmin = KSTACK_SIZE;
1273 static int	kstackleftthres = KSTACK_SIZE / 8;
1274 
1275 void
1276 kstack_setup_magic(const struct lwp *l)
1277 {
1278 	uint32_t *ip;
1279 	uint32_t const *end;
1280 
1281 	KASSERT(l != NULL);
1282 	KASSERT(l != &lwp0);
1283 
1284 	/*
1285 	 * fill all the stack with magic number
1286 	 * so that later modification on it can be detected.
1287 	 */
1288 	ip = (uint32_t *)KSTACK_LOWEST_ADDR(l);
1289 	end = (uint32_t *)((char *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE);
1290 	for (; ip < end; ip++) {
1291 		*ip = KSTACK_MAGIC;
1292 	}
1293 }
1294 
1295 void
1296 kstack_check_magic(const struct lwp *l)
1297 {
1298 	uint32_t const *ip, *end;
1299 	int stackleft;
1300 
1301 	KASSERT(l != NULL);
1302 
1303 	/* don't check proc0 */ /*XXX*/
1304 	if (l == &lwp0)
1305 		return;
1306 
1307 #ifdef __MACHINE_STACK_GROWS_UP
1308 	/* stack grows upwards (eg. hppa) */
1309 	ip = (uint32_t *)((void *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE);
1310 	end = (uint32_t *)KSTACK_LOWEST_ADDR(l);
1311 	for (ip--; ip >= end; ip--)
1312 		if (*ip != KSTACK_MAGIC)
1313 			break;
1314 
1315 	stackleft = (void *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE - (void *)ip;
1316 #else /* __MACHINE_STACK_GROWS_UP */
1317 	/* stack grows downwards (eg. i386) */
1318 	ip = (uint32_t *)KSTACK_LOWEST_ADDR(l);
1319 	end = (uint32_t *)((char *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE);
1320 	for (; ip < end; ip++)
1321 		if (*ip != KSTACK_MAGIC)
1322 			break;
1323 
1324 	stackleft = ((const char *)ip) - (const char *)KSTACK_LOWEST_ADDR(l);
1325 #endif /* __MACHINE_STACK_GROWS_UP */
1326 
1327 	if (kstackleftmin > stackleft) {
1328 		kstackleftmin = stackleft;
1329 		if (stackleft < kstackleftthres)
1330 			printf("warning: kernel stack left %d bytes"
1331 			    "(pid %u:lid %u)\n", stackleft,
1332 			    (u_int)l->l_proc->p_pid, (u_int)l->l_lid);
1333 	}
1334 
1335 	if (stackleft <= 0) {
1336 		panic("magic on the top of kernel stack changed for "
1337 		    "pid %u, lid %u: maybe kernel stack overflow",
1338 		    (u_int)l->l_proc->p_pid, (u_int)l->l_lid);
1339 	}
1340 }
1341 #endif /* KSTACK_CHECK_MAGIC */
1342 
1343 int
1344 proclist_foreach_call(struct proclist *list,
1345     int (*callback)(struct proc *, void *arg), void *arg)
1346 {
1347 	struct proc marker;
1348 	struct proc *p;
1349 	int ret = 0;
1350 
1351 	marker.p_flag = PK_MARKER;
1352 	mutex_enter(proc_lock);
1353 	for (p = LIST_FIRST(list); ret == 0 && p != NULL;) {
1354 		if (p->p_flag & PK_MARKER) {
1355 			p = LIST_NEXT(p, p_list);
1356 			continue;
1357 		}
1358 		LIST_INSERT_AFTER(p, &marker, p_list);
1359 		ret = (*callback)(p, arg);
1360 		KASSERT(mutex_owned(proc_lock));
1361 		p = LIST_NEXT(&marker, p_list);
1362 		LIST_REMOVE(&marker, p_list);
1363 	}
1364 	mutex_exit(proc_lock);
1365 
1366 	return ret;
1367 }
1368 
1369 int
1370 proc_vmspace_getref(struct proc *p, struct vmspace **vm)
1371 {
1372 
1373 	/* XXXCDC: how should locking work here? */
1374 
1375 	/* curproc exception is for coredump. */
1376 
1377 	if ((p != curproc && (p->p_sflag & PS_WEXIT) != 0) ||
1378 	    (p->p_vmspace->vm_refcnt < 1)) { /* XXX */
1379 		return EFAULT;
1380 	}
1381 
1382 	uvmspace_addref(p->p_vmspace);
1383 	*vm = p->p_vmspace;
1384 
1385 	return 0;
1386 }
1387 
1388 /*
1389  * Acquire a write lock on the process credential.
1390  */
1391 void
1392 proc_crmod_enter(void)
1393 {
1394 	struct lwp *l = curlwp;
1395 	struct proc *p = l->l_proc;
1396 	kauth_cred_t oc;
1397 
1398 	/* Reset what needs to be reset in plimit. */
1399 	if (p->p_limit->pl_corename != defcorename) {
1400 		lim_setcorename(p, defcorename, 0);
1401 	}
1402 
1403 	mutex_enter(p->p_lock);
1404 
1405 	/* Ensure the LWP cached credentials are up to date. */
1406 	if ((oc = l->l_cred) != p->p_cred) {
1407 		kauth_cred_hold(p->p_cred);
1408 		l->l_cred = p->p_cred;
1409 		kauth_cred_free(oc);
1410 	}
1411 }
1412 
1413 /*
1414  * Set in a new process credential, and drop the write lock.  The credential
1415  * must have a reference already.  Optionally, free a no-longer required
1416  * credential.  The scheduler also needs to inspect p_cred, so we also
1417  * briefly acquire the sched state mutex.
1418  */
1419 void
1420 proc_crmod_leave(kauth_cred_t scred, kauth_cred_t fcred, bool sugid)
1421 {
1422 	struct lwp *l = curlwp, *l2;
1423 	struct proc *p = l->l_proc;
1424 	kauth_cred_t oc;
1425 
1426 	KASSERT(mutex_owned(p->p_lock));
1427 
1428 	/* Is there a new credential to set in? */
1429 	if (scred != NULL) {
1430 		p->p_cred = scred;
1431 		LIST_FOREACH(l2, &p->p_lwps, l_sibling) {
1432 			if (l2 != l)
1433 				l2->l_prflag |= LPR_CRMOD;
1434 		}
1435 
1436 		/* Ensure the LWP cached credentials are up to date. */
1437 		if ((oc = l->l_cred) != scred) {
1438 			kauth_cred_hold(scred);
1439 			l->l_cred = scred;
1440 		}
1441 	} else
1442 		oc = NULL;	/* XXXgcc */
1443 
1444 	if (sugid) {
1445 		/*
1446 		 * Mark process as having changed credentials, stops
1447 		 * tracing etc.
1448 		 */
1449 		p->p_flag |= PK_SUGID;
1450 	}
1451 
1452 	mutex_exit(p->p_lock);
1453 
1454 	/* If there is a credential to be released, free it now. */
1455 	if (fcred != NULL) {
1456 		KASSERT(scred != NULL);
1457 		kauth_cred_free(fcred);
1458 		if (oc != scred)
1459 			kauth_cred_free(oc);
1460 	}
1461 }
1462 
1463 /*
1464  * proc_specific_key_create --
1465  *	Create a key for subsystem proc-specific data.
1466  */
1467 int
1468 proc_specific_key_create(specificdata_key_t *keyp, specificdata_dtor_t dtor)
1469 {
1470 
1471 	return (specificdata_key_create(proc_specificdata_domain, keyp, dtor));
1472 }
1473 
1474 /*
1475  * proc_specific_key_delete --
1476  *	Delete a key for subsystem proc-specific data.
1477  */
1478 void
1479 proc_specific_key_delete(specificdata_key_t key)
1480 {
1481 
1482 	specificdata_key_delete(proc_specificdata_domain, key);
1483 }
1484 
1485 /*
1486  * proc_initspecific --
1487  *	Initialize a proc's specificdata container.
1488  */
1489 void
1490 proc_initspecific(struct proc *p)
1491 {
1492 	int error __diagused;
1493 
1494 	error = specificdata_init(proc_specificdata_domain, &p->p_specdataref);
1495 	KASSERT(error == 0);
1496 }
1497 
1498 /*
1499  * proc_finispecific --
1500  *	Finalize a proc's specificdata container.
1501  */
1502 void
1503 proc_finispecific(struct proc *p)
1504 {
1505 
1506 	specificdata_fini(proc_specificdata_domain, &p->p_specdataref);
1507 }
1508 
1509 /*
1510  * proc_getspecific --
1511  *	Return proc-specific data corresponding to the specified key.
1512  */
1513 void *
1514 proc_getspecific(struct proc *p, specificdata_key_t key)
1515 {
1516 
1517 	return (specificdata_getspecific(proc_specificdata_domain,
1518 					 &p->p_specdataref, key));
1519 }
1520 
1521 /*
1522  * proc_setspecific --
1523  *	Set proc-specific data corresponding to the specified key.
1524  */
1525 void
1526 proc_setspecific(struct proc *p, specificdata_key_t key, void *data)
1527 {
1528 
1529 	specificdata_setspecific(proc_specificdata_domain,
1530 				 &p->p_specdataref, key, data);
1531 }
1532 
1533 int
1534 proc_uidmatch(kauth_cred_t cred, kauth_cred_t target)
1535 {
1536 	int r = 0;
1537 
1538 	if (kauth_cred_getuid(cred) != kauth_cred_getuid(target) ||
1539 	    kauth_cred_getuid(cred) != kauth_cred_getsvuid(target)) {
1540 		/*
1541 		 * suid proc of ours or proc not ours
1542 		 */
1543 		r = EPERM;
1544 	} else if (kauth_cred_getgid(target) != kauth_cred_getsvgid(target)) {
1545 		/*
1546 		 * sgid proc has sgid back to us temporarily
1547 		 */
1548 		r = EPERM;
1549 	} else {
1550 		/*
1551 		 * our rgid must be in target's group list (ie,
1552 		 * sub-processes started by a sgid process)
1553 		 */
1554 		int ismember = 0;
1555 
1556 		if (kauth_cred_ismember_gid(cred,
1557 		    kauth_cred_getgid(target), &ismember) != 0 ||
1558 		    !ismember)
1559 			r = EPERM;
1560 	}
1561 
1562 	return (r);
1563 }
1564 
1565 /*
1566  * sysctl stuff
1567  */
1568 
1569 #define KERN_PROCSLOP	(5 * sizeof(struct kinfo_proc))
1570 
1571 static const u_int sysctl_flagmap[] = {
1572 	PK_ADVLOCK, P_ADVLOCK,
1573 	PK_EXEC, P_EXEC,
1574 	PK_NOCLDWAIT, P_NOCLDWAIT,
1575 	PK_32, P_32,
1576 	PK_CLDSIGIGN, P_CLDSIGIGN,
1577 	PK_SUGID, P_SUGID,
1578 	0
1579 };
1580 
1581 static const u_int sysctl_sflagmap[] = {
1582 	PS_NOCLDSTOP, P_NOCLDSTOP,
1583 	PS_WEXIT, P_WEXIT,
1584 	PS_STOPFORK, P_STOPFORK,
1585 	PS_STOPEXEC, P_STOPEXEC,
1586 	PS_STOPEXIT, P_STOPEXIT,
1587 	0
1588 };
1589 
1590 static const u_int sysctl_slflagmap[] = {
1591 	PSL_TRACED, P_TRACED,
1592 	PSL_CHTRACED, P_CHTRACED,
1593 	PSL_SYSCALL, P_SYSCALL,
1594 	0
1595 };
1596 
1597 static const u_int sysctl_lflagmap[] = {
1598 	PL_CONTROLT, P_CONTROLT,
1599 	PL_PPWAIT, P_PPWAIT,
1600 	0
1601 };
1602 
1603 static const u_int sysctl_stflagmap[] = {
1604 	PST_PROFIL, P_PROFIL,
1605 	0
1606 
1607 };
1608 
1609 /* used by kern_lwp also */
1610 const u_int sysctl_lwpflagmap[] = {
1611 	LW_SINTR, L_SINTR,
1612 	LW_SYSTEM, L_SYSTEM,
1613 	0
1614 };
1615 
1616 /*
1617  * Find the most ``active'' lwp of a process and return it for ps display
1618  * purposes
1619  */
1620 static struct lwp *
1621 proc_active_lwp(struct proc *p)
1622 {
1623 	static const int ostat[] = {
1624 		0,
1625 		2,	/* LSIDL */
1626 		6,	/* LSRUN */
1627 		5,	/* LSSLEEP */
1628 		4,	/* LSSTOP */
1629 		0,	/* LSZOMB */
1630 		1,	/* LSDEAD */
1631 		7,	/* LSONPROC */
1632 		3	/* LSSUSPENDED */
1633 	};
1634 
1635 	struct lwp *l, *lp = NULL;
1636 	LIST_FOREACH(l, &p->p_lwps, l_sibling) {
1637 		KASSERT(l->l_stat >= 0 && l->l_stat < __arraycount(ostat));
1638 		if (lp == NULL ||
1639 		    ostat[l->l_stat] > ostat[lp->l_stat] ||
1640 		    (ostat[l->l_stat] == ostat[lp->l_stat] &&
1641 		    l->l_cpticks > lp->l_cpticks)) {
1642 			lp = l;
1643 			continue;
1644 		}
1645 	}
1646 	return lp;
1647 }
1648 
1649 static int
1650 sysctl_doeproc(SYSCTLFN_ARGS)
1651 {
1652 	union {
1653 		struct kinfo_proc kproc;
1654 		struct kinfo_proc2 kproc2;
1655 	} *kbuf;
1656 	struct proc *p, *next, *marker;
1657 	char *where, *dp;
1658 	int type, op, arg, error;
1659 	u_int elem_size, kelem_size, elem_count;
1660 	size_t buflen, needed;
1661 	bool match, zombie, mmmbrains;
1662 	const bool allowaddr = get_expose_address(curproc);
1663 
1664 	if (namelen == 1 && name[0] == CTL_QUERY)
1665 		return (sysctl_query(SYSCTLFN_CALL(rnode)));
1666 
1667 	dp = where = oldp;
1668 	buflen = where != NULL ? *oldlenp : 0;
1669 	error = 0;
1670 	needed = 0;
1671 	type = rnode->sysctl_num;
1672 
1673 	if (type == KERN_PROC) {
1674 		if (namelen == 0)
1675 			return EINVAL;
1676 		switch (op = name[0]) {
1677 		case KERN_PROC_ALL:
1678 			if (namelen != 1)
1679 				return EINVAL;
1680 			arg = 0;
1681 			break;
1682 		default:
1683 			if (namelen != 2)
1684 				return EINVAL;
1685 			arg = name[1];
1686 			break;
1687 		}
1688 		elem_count = 0;	/* Hush little compiler, don't you cry */
1689 		kelem_size = elem_size = sizeof(kbuf->kproc);
1690 	} else {
1691 		if (namelen != 4)
1692 			return EINVAL;
1693 		op = name[0];
1694 		arg = name[1];
1695 		elem_size = name[2];
1696 		elem_count = name[3];
1697 		kelem_size = sizeof(kbuf->kproc2);
1698 	}
1699 
1700 	sysctl_unlock();
1701 
1702 	kbuf = kmem_zalloc(sizeof(*kbuf), KM_SLEEP);
1703 	marker = kmem_alloc(sizeof(*marker), KM_SLEEP);
1704 	marker->p_flag = PK_MARKER;
1705 
1706 	mutex_enter(proc_lock);
1707 	/*
1708 	 * Start with zombies to prevent reporting processes twice, in case they
1709 	 * are dying and being moved from the list of alive processes to zombies.
1710 	 */
1711 	mmmbrains = true;
1712 	for (p = LIST_FIRST(&zombproc);; p = next) {
1713 		if (p == NULL) {
1714 			if (mmmbrains) {
1715 				p = LIST_FIRST(&allproc);
1716 				mmmbrains = false;
1717 			}
1718 			if (p == NULL)
1719 				break;
1720 		}
1721 		next = LIST_NEXT(p, p_list);
1722 		if ((p->p_flag & PK_MARKER) != 0)
1723 			continue;
1724 
1725 		/*
1726 		 * Skip embryonic processes.
1727 		 */
1728 		if (p->p_stat == SIDL)
1729 			continue;
1730 
1731 		mutex_enter(p->p_lock);
1732 		error = kauth_authorize_process(l->l_cred,
1733 		    KAUTH_PROCESS_CANSEE, p,
1734 		    KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_EPROC), NULL, NULL);
1735 		if (error != 0) {
1736 			mutex_exit(p->p_lock);
1737 			continue;
1738 		}
1739 
1740 		/*
1741 		 * Hande all the operations in one switch on the cost of
1742 		 * algorithm complexity is on purpose. The win splitting this
1743 		 * function into several similar copies makes maintenance burden
1744 		 * burden, code grow and boost is neglible in practical systems.
1745 		 */
1746 		switch (op) {
1747 		case KERN_PROC_PID:
1748 			match = (p->p_pid == (pid_t)arg);
1749 			break;
1750 
1751 		case KERN_PROC_PGRP:
1752 			match = (p->p_pgrp->pg_id == (pid_t)arg);
1753 			break;
1754 
1755 		case KERN_PROC_SESSION:
1756 			match = (p->p_session->s_sid == (pid_t)arg);
1757 			break;
1758 
1759 		case KERN_PROC_TTY:
1760 			match = true;
1761 			if (arg == (int) KERN_PROC_TTY_REVOKE) {
1762 				if ((p->p_lflag & PL_CONTROLT) == 0 ||
1763 				    p->p_session->s_ttyp == NULL ||
1764 				    p->p_session->s_ttyvp != NULL) {
1765 				    	match = false;
1766 				}
1767 			} else if ((p->p_lflag & PL_CONTROLT) == 0 ||
1768 			    p->p_session->s_ttyp == NULL) {
1769 				if ((dev_t)arg != KERN_PROC_TTY_NODEV) {
1770 					match = false;
1771 				}
1772 			} else if (p->p_session->s_ttyp->t_dev != (dev_t)arg) {
1773 				match = false;
1774 			}
1775 			break;
1776 
1777 		case KERN_PROC_UID:
1778 			match = (kauth_cred_geteuid(p->p_cred) == (uid_t)arg);
1779 			break;
1780 
1781 		case KERN_PROC_RUID:
1782 			match = (kauth_cred_getuid(p->p_cred) == (uid_t)arg);
1783 			break;
1784 
1785 		case KERN_PROC_GID:
1786 			match = (kauth_cred_getegid(p->p_cred) == (uid_t)arg);
1787 			break;
1788 
1789 		case KERN_PROC_RGID:
1790 			match = (kauth_cred_getgid(p->p_cred) == (uid_t)arg);
1791 			break;
1792 
1793 		case KERN_PROC_ALL:
1794 			match = true;
1795 			/* allow everything */
1796 			break;
1797 
1798 		default:
1799 			error = EINVAL;
1800 			mutex_exit(p->p_lock);
1801 			goto cleanup;
1802 		}
1803 		if (!match) {
1804 			mutex_exit(p->p_lock);
1805 			continue;
1806 		}
1807 
1808 		/*
1809 		 * Grab a hold on the process.
1810 		 */
1811 		if (mmmbrains) {
1812 			zombie = true;
1813 		} else {
1814 			zombie = !rw_tryenter(&p->p_reflock, RW_READER);
1815 		}
1816 		if (zombie) {
1817 			LIST_INSERT_AFTER(p, marker, p_list);
1818 		}
1819 
1820 		if (buflen >= elem_size &&
1821 		    (type == KERN_PROC || elem_count > 0)) {
1822 			ruspace(p);	/* Update process vm resource use */
1823 
1824 			if (type == KERN_PROC) {
1825 				fill_proc(p, &kbuf->kproc.kp_proc, allowaddr);
1826 				fill_eproc(p, &kbuf->kproc.kp_eproc, zombie,
1827 				    allowaddr);
1828 			} else {
1829 				fill_kproc2(p, &kbuf->kproc2, zombie,
1830 				    allowaddr);
1831 				elem_count--;
1832 			}
1833 			mutex_exit(p->p_lock);
1834 			mutex_exit(proc_lock);
1835 			/*
1836 			 * Copy out elem_size, but not larger than kelem_size
1837 			 */
1838 			error = sysctl_copyout(l, kbuf, dp,
1839 			    uimin(kelem_size, elem_size));
1840 			mutex_enter(proc_lock);
1841 			if (error) {
1842 				goto bah;
1843 			}
1844 			dp += elem_size;
1845 			buflen -= elem_size;
1846 		} else {
1847 			mutex_exit(p->p_lock);
1848 		}
1849 		needed += elem_size;
1850 
1851 		/*
1852 		 * Release reference to process.
1853 		 */
1854 	 	if (zombie) {
1855 			next = LIST_NEXT(marker, p_list);
1856  			LIST_REMOVE(marker, p_list);
1857 		} else {
1858 			rw_exit(&p->p_reflock);
1859 			next = LIST_NEXT(p, p_list);
1860 		}
1861 
1862 		/*
1863 		 * Short-circuit break quickly!
1864 		 */
1865 		if (op == KERN_PROC_PID)
1866                 	break;
1867 	}
1868 	mutex_exit(proc_lock);
1869 
1870 	if (where != NULL) {
1871 		*oldlenp = dp - where;
1872 		if (needed > *oldlenp) {
1873 			error = ENOMEM;
1874 			goto out;
1875 		}
1876 	} else {
1877 		needed += KERN_PROCSLOP;
1878 		*oldlenp = needed;
1879 	}
1880 	kmem_free(kbuf, sizeof(*kbuf));
1881 	kmem_free(marker, sizeof(*marker));
1882 	sysctl_relock();
1883 	return 0;
1884  bah:
1885  	if (zombie)
1886  		LIST_REMOVE(marker, p_list);
1887 	else
1888 		rw_exit(&p->p_reflock);
1889  cleanup:
1890 	mutex_exit(proc_lock);
1891  out:
1892 	kmem_free(kbuf, sizeof(*kbuf));
1893 	kmem_free(marker, sizeof(*marker));
1894 	sysctl_relock();
1895 	return error;
1896 }
1897 
1898 int
1899 copyin_psstrings(struct proc *p, struct ps_strings *arginfo)
1900 {
1901 #if !defined(_RUMPKERNEL)
1902 	int retval;
1903 
1904 	if (p->p_flag & PK_32) {
1905 		MODULE_HOOK_CALL(kern_proc32_copyin_hook, (p, arginfo),
1906 		    enosys(), retval);
1907 		return retval;
1908 	}
1909 #endif /* !defined(_RUMPKERNEL) */
1910 
1911 	return copyin_proc(p, (void *)p->p_psstrp, arginfo, sizeof(*arginfo));
1912 }
1913 
1914 static int
1915 copy_procargs_sysctl_cb(void *cookie_, const void *src, size_t off, size_t len)
1916 {
1917 	void **cookie = cookie_;
1918 	struct lwp *l = cookie[0];
1919 	char *dst = cookie[1];
1920 
1921 	return sysctl_copyout(l, src, dst + off, len);
1922 }
1923 
1924 /*
1925  * sysctl helper routine for kern.proc_args pseudo-subtree.
1926  */
1927 static int
1928 sysctl_kern_proc_args(SYSCTLFN_ARGS)
1929 {
1930 	struct ps_strings pss;
1931 	struct proc *p;
1932 	pid_t pid;
1933 	int type, error;
1934 	void *cookie[2];
1935 
1936 	if (namelen == 1 && name[0] == CTL_QUERY)
1937 		return (sysctl_query(SYSCTLFN_CALL(rnode)));
1938 
1939 	if (newp != NULL || namelen != 2)
1940 		return (EINVAL);
1941 	pid = name[0];
1942 	type = name[1];
1943 
1944 	switch (type) {
1945 	case KERN_PROC_PATHNAME:
1946 		sysctl_unlock();
1947 		error = fill_pathname(l, pid, oldp, oldlenp);
1948 		sysctl_relock();
1949 		return error;
1950 
1951 	case KERN_PROC_CWD:
1952 		sysctl_unlock();
1953 		error = fill_cwd(l, pid, oldp, oldlenp);
1954 		sysctl_relock();
1955 		return error;
1956 
1957 	case KERN_PROC_ARGV:
1958 	case KERN_PROC_NARGV:
1959 	case KERN_PROC_ENV:
1960 	case KERN_PROC_NENV:
1961 		/* ok */
1962 		break;
1963 	default:
1964 		return (EINVAL);
1965 	}
1966 
1967 	sysctl_unlock();
1968 
1969 	/* check pid */
1970 	mutex_enter(proc_lock);
1971 	if ((p = proc_find(pid)) == NULL) {
1972 		error = EINVAL;
1973 		goto out_locked;
1974 	}
1975 	mutex_enter(p->p_lock);
1976 
1977 	/* Check permission. */
1978 	if (type == KERN_PROC_ARGV || type == KERN_PROC_NARGV)
1979 		error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE,
1980 		    p, KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ARGS), NULL, NULL);
1981 	else if (type == KERN_PROC_ENV || type == KERN_PROC_NENV)
1982 		error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE,
1983 		    p, KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENV), NULL, NULL);
1984 	else
1985 		error = EINVAL; /* XXXGCC */
1986 	if (error) {
1987 		mutex_exit(p->p_lock);
1988 		goto out_locked;
1989 	}
1990 
1991 	if (oldp == NULL) {
1992 		if (type == KERN_PROC_NARGV || type == KERN_PROC_NENV)
1993 			*oldlenp = sizeof (int);
1994 		else
1995 			*oldlenp = ARG_MAX;	/* XXX XXX XXX */
1996 		error = 0;
1997 		mutex_exit(p->p_lock);
1998 		goto out_locked;
1999 	}
2000 
2001 	/*
2002 	 * Zombies don't have a stack, so we can't read their psstrings.
2003 	 * System processes also don't have a user stack.
2004 	 */
2005 	if (P_ZOMBIE(p) || (p->p_flag & PK_SYSTEM) != 0) {
2006 		error = EINVAL;
2007 		mutex_exit(p->p_lock);
2008 		goto out_locked;
2009 	}
2010 
2011 	error = rw_tryenter(&p->p_reflock, RW_READER) ? 0 : EBUSY;
2012 	mutex_exit(p->p_lock);
2013 	if (error) {
2014 		goto out_locked;
2015 	}
2016 	mutex_exit(proc_lock);
2017 
2018 	if (type == KERN_PROC_NARGV || type == KERN_PROC_NENV) {
2019 		int value;
2020 		if ((error = copyin_psstrings(p, &pss)) == 0) {
2021 			if (type == KERN_PROC_NARGV)
2022 				value = pss.ps_nargvstr;
2023 			else
2024 				value = pss.ps_nenvstr;
2025 			error = sysctl_copyout(l, &value, oldp, sizeof(value));
2026 			*oldlenp = sizeof(value);
2027 		}
2028 	} else {
2029 		cookie[0] = l;
2030 		cookie[1] = oldp;
2031 		error = copy_procargs(p, type, oldlenp,
2032 		    copy_procargs_sysctl_cb, cookie);
2033 	}
2034 	rw_exit(&p->p_reflock);
2035 	sysctl_relock();
2036 	return error;
2037 
2038 out_locked:
2039 	mutex_exit(proc_lock);
2040 	sysctl_relock();
2041 	return error;
2042 }
2043 
2044 int
2045 copy_procargs(struct proc *p, int oid, size_t *limit,
2046     int (*cb)(void *, const void *, size_t, size_t), void *cookie)
2047 {
2048 	struct ps_strings pss;
2049 	size_t len, i, loaded, entry_len;
2050 	struct uio auio;
2051 	struct iovec aiov;
2052 	int error, argvlen;
2053 	char *arg;
2054 	char **argv;
2055 	vaddr_t user_argv;
2056 	struct vmspace *vmspace;
2057 
2058 	/*
2059 	 * Allocate a temporary buffer to hold the argument vector and
2060 	 * the arguments themselve.
2061 	 */
2062 	arg = kmem_alloc(PAGE_SIZE, KM_SLEEP);
2063 	argv = kmem_alloc(PAGE_SIZE, KM_SLEEP);
2064 
2065 	/*
2066 	 * Lock the process down in memory.
2067 	 */
2068 	vmspace = p->p_vmspace;
2069 	uvmspace_addref(vmspace);
2070 
2071 	/*
2072 	 * Read in the ps_strings structure.
2073 	 */
2074 	if ((error = copyin_psstrings(p, &pss)) != 0)
2075 		goto done;
2076 
2077 	/*
2078 	 * Now read the address of the argument vector.
2079 	 */
2080 	switch (oid) {
2081 	case KERN_PROC_ARGV:
2082 		user_argv = (uintptr_t)pss.ps_argvstr;
2083 		argvlen = pss.ps_nargvstr;
2084 		break;
2085 	case KERN_PROC_ENV:
2086 		user_argv = (uintptr_t)pss.ps_envstr;
2087 		argvlen = pss.ps_nenvstr;
2088 		break;
2089 	default:
2090 		error = EINVAL;
2091 		goto done;
2092 	}
2093 
2094 	if (argvlen < 0) {
2095 		error = EIO;
2096 		goto done;
2097 	}
2098 
2099 
2100 	/*
2101 	 * Now copy each string.
2102 	 */
2103 	len = 0; /* bytes written to user buffer */
2104 	loaded = 0; /* bytes from argv already processed */
2105 	i = 0; /* To make compiler happy */
2106 	entry_len = PROC_PTRSZ(p);
2107 
2108 	for (; argvlen; --argvlen) {
2109 		int finished = 0;
2110 		vaddr_t base;
2111 		size_t xlen;
2112 		int j;
2113 
2114 		if (loaded == 0) {
2115 			size_t rem = entry_len * argvlen;
2116 			loaded = MIN(rem, PAGE_SIZE);
2117 			error = copyin_vmspace(vmspace,
2118 			    (const void *)user_argv, argv, loaded);
2119 			if (error)
2120 				break;
2121 			user_argv += loaded;
2122 			i = 0;
2123 		}
2124 
2125 #if !defined(_RUMPKERNEL)
2126 		if (p->p_flag & PK_32)
2127 			MODULE_HOOK_CALL(kern_proc32_base_hook,
2128 			    (argv, i++), 0, base);
2129 		else
2130 #endif /* !defined(_RUMPKERNEL) */
2131 			base = (vaddr_t)argv[i++];
2132 		loaded -= entry_len;
2133 
2134 		/*
2135 		 * The program has messed around with its arguments,
2136 		 * possibly deleting some, and replacing them with
2137 		 * NULL's. Treat this as the last argument and not
2138 		 * a failure.
2139 		 */
2140 		if (base == 0)
2141 			break;
2142 
2143 		while (!finished) {
2144 			xlen = PAGE_SIZE - (base & PAGE_MASK);
2145 
2146 			aiov.iov_base = arg;
2147 			aiov.iov_len = PAGE_SIZE;
2148 			auio.uio_iov = &aiov;
2149 			auio.uio_iovcnt = 1;
2150 			auio.uio_offset = base;
2151 			auio.uio_resid = xlen;
2152 			auio.uio_rw = UIO_READ;
2153 			UIO_SETUP_SYSSPACE(&auio);
2154 			error = uvm_io(&vmspace->vm_map, &auio, 0);
2155 			if (error)
2156 				goto done;
2157 
2158 			/* Look for the end of the string */
2159 			for (j = 0; j < xlen; j++) {
2160 				if (arg[j] == '\0') {
2161 					xlen = j + 1;
2162 					finished = 1;
2163 					break;
2164 				}
2165 			}
2166 
2167 			/* Check for user buffer overflow */
2168 			if (len + xlen > *limit) {
2169 				finished = 1;
2170 				if (len > *limit)
2171 					xlen = 0;
2172 				else
2173 					xlen = *limit - len;
2174 			}
2175 
2176 			/* Copyout the page */
2177 			error = (*cb)(cookie, arg, len, xlen);
2178 			if (error)
2179 				goto done;
2180 
2181 			len += xlen;
2182 			base += xlen;
2183 		}
2184 	}
2185 	*limit = len;
2186 
2187 done:
2188 	kmem_free(argv, PAGE_SIZE);
2189 	kmem_free(arg, PAGE_SIZE);
2190 	uvmspace_free(vmspace);
2191 	return error;
2192 }
2193 
2194 /*
2195  * Fill in a proc structure for the specified process.
2196  */
2197 static void
2198 fill_proc(const struct proc *psrc, struct proc *p, bool allowaddr)
2199 {
2200 	COND_SET_VALUE(p->p_list, psrc->p_list, allowaddr);
2201 	COND_SET_VALUE(p->p_auxlock, psrc->p_auxlock, allowaddr);
2202 	COND_SET_VALUE(p->p_lock, psrc->p_lock, allowaddr);
2203 	COND_SET_VALUE(p->p_stmutex, psrc->p_stmutex, allowaddr);
2204 	COND_SET_VALUE(p->p_reflock, psrc->p_reflock, allowaddr);
2205 	COND_SET_VALUE(p->p_waitcv, psrc->p_waitcv, allowaddr);
2206 	COND_SET_VALUE(p->p_lwpcv, psrc->p_lwpcv, allowaddr);
2207 	COND_SET_VALUE(p->p_cred, psrc->p_cred, allowaddr);
2208 	COND_SET_VALUE(p->p_fd, psrc->p_fd, allowaddr);
2209 	COND_SET_VALUE(p->p_cwdi, psrc->p_cwdi, allowaddr);
2210 	COND_SET_VALUE(p->p_stats, psrc->p_stats, allowaddr);
2211 	COND_SET_VALUE(p->p_limit, psrc->p_limit, allowaddr);
2212 	COND_SET_VALUE(p->p_vmspace, psrc->p_vmspace, allowaddr);
2213 	COND_SET_VALUE(p->p_sigacts, psrc->p_sigacts, allowaddr);
2214 	COND_SET_VALUE(p->p_aio, psrc->p_aio, allowaddr);
2215 	p->p_mqueue_cnt = psrc->p_mqueue_cnt;
2216 	COND_SET_VALUE(p->p_specdataref, psrc->p_specdataref, allowaddr);
2217 	p->p_exitsig = psrc->p_exitsig;
2218 	p->p_flag = psrc->p_flag;
2219 	p->p_sflag = psrc->p_sflag;
2220 	p->p_slflag = psrc->p_slflag;
2221 	p->p_lflag = psrc->p_lflag;
2222 	p->p_stflag = psrc->p_stflag;
2223 	p->p_stat = psrc->p_stat;
2224 	p->p_trace_enabled = psrc->p_trace_enabled;
2225 	p->p_pid = psrc->p_pid;
2226 	COND_SET_VALUE(p->p_pglist, psrc->p_pglist, allowaddr);
2227 	COND_SET_VALUE(p->p_pptr, psrc->p_pptr, allowaddr);
2228 	COND_SET_VALUE(p->p_sibling, psrc->p_sibling, allowaddr);
2229 	COND_SET_VALUE(p->p_children, psrc->p_children, allowaddr);
2230 	COND_SET_VALUE(p->p_lwps, psrc->p_lwps, allowaddr);
2231 	COND_SET_VALUE(p->p_raslist, psrc->p_raslist, allowaddr);
2232 	p->p_nlwps = psrc->p_nlwps;
2233 	p->p_nzlwps = psrc->p_nzlwps;
2234 	p->p_nrlwps = psrc->p_nrlwps;
2235 	p->p_nlwpwait = psrc->p_nlwpwait;
2236 	p->p_ndlwps = psrc->p_ndlwps;
2237 	p->p_nlwpid = psrc->p_nlwpid;
2238 	p->p_nstopchild = psrc->p_nstopchild;
2239 	p->p_waited = psrc->p_waited;
2240 	COND_SET_VALUE(p->p_zomblwp, psrc->p_zomblwp, allowaddr);
2241 	COND_SET_VALUE(p->p_vforklwp, psrc->p_vforklwp, allowaddr);
2242 	COND_SET_VALUE(p->p_sched_info, psrc->p_sched_info, allowaddr);
2243 	p->p_estcpu = psrc->p_estcpu;
2244 	p->p_estcpu_inherited = psrc->p_estcpu_inherited;
2245 	p->p_forktime = psrc->p_forktime;
2246 	p->p_pctcpu = psrc->p_pctcpu;
2247 	COND_SET_VALUE(p->p_opptr, psrc->p_opptr, allowaddr);
2248 	COND_SET_VALUE(p->p_timers, psrc->p_timers, allowaddr);
2249 	p->p_rtime = psrc->p_rtime;
2250 	p->p_uticks = psrc->p_uticks;
2251 	p->p_sticks = psrc->p_sticks;
2252 	p->p_iticks = psrc->p_iticks;
2253 	p->p_xutime = psrc->p_xutime;
2254 	p->p_xstime = psrc->p_xstime;
2255 	p->p_traceflag = psrc->p_traceflag;
2256 	COND_SET_VALUE(p->p_tracep, psrc->p_tracep, allowaddr);
2257 	COND_SET_VALUE(p->p_textvp, psrc->p_textvp, allowaddr);
2258 	COND_SET_VALUE(p->p_emul, psrc->p_emul, allowaddr);
2259 	COND_SET_VALUE(p->p_emuldata, psrc->p_emuldata, allowaddr);
2260 	COND_SET_VALUE(p->p_execsw, psrc->p_execsw, allowaddr);
2261 	COND_SET_VALUE(p->p_klist, psrc->p_klist, allowaddr);
2262 	COND_SET_VALUE(p->p_sigwaiters, psrc->p_sigwaiters, allowaddr);
2263 	COND_SET_VALUE(p->p_sigpend, psrc->p_sigpend, allowaddr);
2264 	COND_SET_VALUE(p->p_lwpctl, psrc->p_lwpctl, allowaddr);
2265 	p->p_ppid = psrc->p_ppid;
2266 	COND_SET_VALUE(p->p_path, psrc->p_path, allowaddr);
2267 	COND_SET_VALUE(p->p_sigctx, psrc->p_sigctx, allowaddr);
2268 	p->p_nice = psrc->p_nice;
2269 	memcpy(p->p_comm, psrc->p_comm, sizeof(p->p_comm));
2270 	COND_SET_VALUE(p->p_pgrp, psrc->p_pgrp, allowaddr);
2271 	COND_SET_VALUE(p->p_psstrp, psrc->p_psstrp, allowaddr);
2272 	p->p_pax = psrc->p_pax;
2273 	p->p_xexit = psrc->p_xexit;
2274 	p->p_xsig = psrc->p_xsig;
2275 	p->p_acflag = psrc->p_acflag;
2276 	COND_SET_VALUE(p->p_md, psrc->p_md, allowaddr);
2277 	p->p_stackbase = psrc->p_stackbase;
2278 	COND_SET_VALUE(p->p_dtrace, psrc->p_dtrace, allowaddr);
2279 }
2280 
2281 /*
2282  * Fill in an eproc structure for the specified process.
2283  */
2284 void
2285 fill_eproc(struct proc *p, struct eproc *ep, bool zombie, bool allowaddr)
2286 {
2287 	struct tty *tp;
2288 	struct lwp *l;
2289 
2290 	KASSERT(mutex_owned(proc_lock));
2291 	KASSERT(mutex_owned(p->p_lock));
2292 
2293 	COND_SET_VALUE(ep->e_paddr, p, allowaddr);
2294 	COND_SET_VALUE(ep->e_sess, p->p_session, allowaddr);
2295 	if (p->p_cred) {
2296 		kauth_cred_topcred(p->p_cred, &ep->e_pcred);
2297 		kauth_cred_toucred(p->p_cred, &ep->e_ucred);
2298 	}
2299 	if (p->p_stat != SIDL && !P_ZOMBIE(p) && !zombie) {
2300 		struct vmspace *vm = p->p_vmspace;
2301 
2302 		ep->e_vm.vm_rssize = vm_resident_count(vm);
2303 		ep->e_vm.vm_tsize = vm->vm_tsize;
2304 		ep->e_vm.vm_dsize = vm->vm_dsize;
2305 		ep->e_vm.vm_ssize = vm->vm_ssize;
2306 		ep->e_vm.vm_map.size = vm->vm_map.size;
2307 
2308 		/* Pick the primary (first) LWP */
2309 		l = proc_active_lwp(p);
2310 		KASSERT(l != NULL);
2311 		lwp_lock(l);
2312 		if (l->l_wchan)
2313 			strncpy(ep->e_wmesg, l->l_wmesg, WMESGLEN);
2314 		lwp_unlock(l);
2315 	}
2316 	ep->e_ppid = p->p_ppid;
2317 	if (p->p_pgrp && p->p_session) {
2318 		ep->e_pgid = p->p_pgrp->pg_id;
2319 		ep->e_jobc = p->p_pgrp->pg_jobc;
2320 		ep->e_sid = p->p_session->s_sid;
2321 		if ((p->p_lflag & PL_CONTROLT) &&
2322 		    (tp = p->p_session->s_ttyp)) {
2323 			ep->e_tdev = tp->t_dev;
2324 			ep->e_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PGID;
2325 			COND_SET_VALUE(ep->e_tsess, tp->t_session, allowaddr);
2326 		} else
2327 			ep->e_tdev = (uint32_t)NODEV;
2328 		ep->e_flag = p->p_session->s_ttyvp ? EPROC_CTTY : 0;
2329 		if (SESS_LEADER(p))
2330 			ep->e_flag |= EPROC_SLEADER;
2331 		strncpy(ep->e_login, p->p_session->s_login, MAXLOGNAME);
2332 	}
2333 	ep->e_xsize = ep->e_xrssize = 0;
2334 	ep->e_xccount = ep->e_xswrss = 0;
2335 }
2336 
2337 /*
2338  * Fill in a kinfo_proc2 structure for the specified process.
2339  */
2340 void
2341 fill_kproc2(struct proc *p, struct kinfo_proc2 *ki, bool zombie, bool allowaddr)
2342 {
2343 	struct tty *tp;
2344 	struct lwp *l, *l2;
2345 	struct timeval ut, st, rt;
2346 	sigset_t ss1, ss2;
2347 	struct rusage ru;
2348 	struct vmspace *vm;
2349 
2350 	KASSERT(mutex_owned(proc_lock));
2351 	KASSERT(mutex_owned(p->p_lock));
2352 
2353 	sigemptyset(&ss1);
2354 	sigemptyset(&ss2);
2355 
2356 	COND_SET_VALUE(ki->p_paddr, PTRTOUINT64(p), allowaddr);
2357 	COND_SET_VALUE(ki->p_fd, PTRTOUINT64(p->p_fd), allowaddr);
2358 	COND_SET_VALUE(ki->p_cwdi, PTRTOUINT64(p->p_cwdi), allowaddr);
2359 	COND_SET_VALUE(ki->p_stats, PTRTOUINT64(p->p_stats), allowaddr);
2360 	COND_SET_VALUE(ki->p_limit, PTRTOUINT64(p->p_limit), allowaddr);
2361 	COND_SET_VALUE(ki->p_vmspace, PTRTOUINT64(p->p_vmspace), allowaddr);
2362 	COND_SET_VALUE(ki->p_sigacts, PTRTOUINT64(p->p_sigacts), allowaddr);
2363 	COND_SET_VALUE(ki->p_sess, PTRTOUINT64(p->p_session), allowaddr);
2364 	ki->p_tsess = 0;	/* may be changed if controlling tty below */
2365 	COND_SET_VALUE(ki->p_ru, PTRTOUINT64(&p->p_stats->p_ru), allowaddr);
2366 	ki->p_eflag = 0;
2367 	ki->p_exitsig = p->p_exitsig;
2368 	ki->p_flag = L_INMEM;   /* Process never swapped out */
2369 	ki->p_flag |= sysctl_map_flags(sysctl_flagmap, p->p_flag);
2370 	ki->p_flag |= sysctl_map_flags(sysctl_sflagmap, p->p_sflag);
2371 	ki->p_flag |= sysctl_map_flags(sysctl_slflagmap, p->p_slflag);
2372 	ki->p_flag |= sysctl_map_flags(sysctl_lflagmap, p->p_lflag);
2373 	ki->p_flag |= sysctl_map_flags(sysctl_stflagmap, p->p_stflag);
2374 	ki->p_pid = p->p_pid;
2375 	ki->p_ppid = p->p_ppid;
2376 	ki->p_uid = kauth_cred_geteuid(p->p_cred);
2377 	ki->p_ruid = kauth_cred_getuid(p->p_cred);
2378 	ki->p_gid = kauth_cred_getegid(p->p_cred);
2379 	ki->p_rgid = kauth_cred_getgid(p->p_cred);
2380 	ki->p_svuid = kauth_cred_getsvuid(p->p_cred);
2381 	ki->p_svgid = kauth_cred_getsvgid(p->p_cred);
2382 	ki->p_ngroups = kauth_cred_ngroups(p->p_cred);
2383 	kauth_cred_getgroups(p->p_cred, ki->p_groups,
2384 	    uimin(ki->p_ngroups, sizeof(ki->p_groups) / sizeof(ki->p_groups[0])),
2385 	    UIO_SYSSPACE);
2386 
2387 	ki->p_uticks = p->p_uticks;
2388 	ki->p_sticks = p->p_sticks;
2389 	ki->p_iticks = p->p_iticks;
2390 	ki->p_tpgid = NO_PGID;	/* may be changed if controlling tty below */
2391 	COND_SET_VALUE(ki->p_tracep, PTRTOUINT64(p->p_tracep), allowaddr);
2392 	ki->p_traceflag = p->p_traceflag;
2393 
2394 	memcpy(&ki->p_sigignore, &p->p_sigctx.ps_sigignore,sizeof(ki_sigset_t));
2395 	memcpy(&ki->p_sigcatch, &p->p_sigctx.ps_sigcatch, sizeof(ki_sigset_t));
2396 
2397 	ki->p_cpticks = 0;
2398 	ki->p_pctcpu = p->p_pctcpu;
2399 	ki->p_estcpu = 0;
2400 	ki->p_stat = p->p_stat; /* Will likely be overridden by LWP status */
2401 	ki->p_realstat = p->p_stat;
2402 	ki->p_nice = p->p_nice;
2403 	ki->p_xstat = P_WAITSTATUS(p);
2404 	ki->p_acflag = p->p_acflag;
2405 
2406 	strncpy(ki->p_comm, p->p_comm,
2407 	    uimin(sizeof(ki->p_comm), sizeof(p->p_comm)));
2408 	strncpy(ki->p_ename, p->p_emul->e_name, sizeof(ki->p_ename));
2409 
2410 	ki->p_nlwps = p->p_nlwps;
2411 	ki->p_realflag = ki->p_flag;
2412 
2413 	if (p->p_stat != SIDL && !P_ZOMBIE(p) && !zombie) {
2414 		vm = p->p_vmspace;
2415 		ki->p_vm_rssize = vm_resident_count(vm);
2416 		ki->p_vm_tsize = vm->vm_tsize;
2417 		ki->p_vm_dsize = vm->vm_dsize;
2418 		ki->p_vm_ssize = vm->vm_ssize;
2419 		ki->p_vm_vsize = atop(vm->vm_map.size);
2420 		/*
2421 		 * Since the stack is initially mapped mostly with
2422 		 * PROT_NONE and grown as needed, adjust the "mapped size"
2423 		 * to skip the unused stack portion.
2424 		 */
2425 		ki->p_vm_msize =
2426 		    atop(vm->vm_map.size) - vm->vm_issize + vm->vm_ssize;
2427 
2428 		/* Pick the primary (first) LWP */
2429 		l = proc_active_lwp(p);
2430 		KASSERT(l != NULL);
2431 		lwp_lock(l);
2432 		ki->p_nrlwps = p->p_nrlwps;
2433 		ki->p_forw = 0;
2434 		ki->p_back = 0;
2435 		COND_SET_VALUE(ki->p_addr, PTRTOUINT64(l->l_addr), allowaddr);
2436 		ki->p_stat = l->l_stat;
2437 		ki->p_flag |= sysctl_map_flags(sysctl_lwpflagmap, l->l_flag);
2438 		ki->p_swtime = l->l_swtime;
2439 		ki->p_slptime = l->l_slptime;
2440 		if (l->l_stat == LSONPROC)
2441 			ki->p_schedflags = l->l_cpu->ci_schedstate.spc_flags;
2442 		else
2443 			ki->p_schedflags = 0;
2444 		ki->p_priority = lwp_eprio(l);
2445 		ki->p_usrpri = l->l_priority;
2446 		if (l->l_wchan)
2447 			strncpy(ki->p_wmesg, l->l_wmesg, sizeof(ki->p_wmesg));
2448 		COND_SET_VALUE(ki->p_wchan, PTRTOUINT64(l->l_wchan), allowaddr);
2449 		ki->p_cpuid = cpu_index(l->l_cpu);
2450 		lwp_unlock(l);
2451 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
2452 			/* This is hardly correct, but... */
2453 			sigplusset(&l->l_sigpend.sp_set, &ss1);
2454 			sigplusset(&l->l_sigmask, &ss2);
2455 			ki->p_cpticks += l->l_cpticks;
2456 			ki->p_pctcpu += l->l_pctcpu;
2457 			ki->p_estcpu += l->l_estcpu;
2458 		}
2459 	}
2460 	sigplusset(&p->p_sigpend.sp_set, &ss2);
2461 	memcpy(&ki->p_siglist, &ss1, sizeof(ki_sigset_t));
2462 	memcpy(&ki->p_sigmask, &ss2, sizeof(ki_sigset_t));
2463 
2464 	if (p->p_session != NULL) {
2465 		ki->p_sid = p->p_session->s_sid;
2466 		ki->p__pgid = p->p_pgrp->pg_id;
2467 		if (p->p_session->s_ttyvp)
2468 			ki->p_eflag |= EPROC_CTTY;
2469 		if (SESS_LEADER(p))
2470 			ki->p_eflag |= EPROC_SLEADER;
2471 		strncpy(ki->p_login, p->p_session->s_login,
2472 		    uimin(sizeof ki->p_login - 1, sizeof p->p_session->s_login));
2473 		ki->p_jobc = p->p_pgrp->pg_jobc;
2474 		if ((p->p_lflag & PL_CONTROLT) && (tp = p->p_session->s_ttyp)) {
2475 			ki->p_tdev = tp->t_dev;
2476 			ki->p_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PGID;
2477 			COND_SET_VALUE(ki->p_tsess, PTRTOUINT64(tp->t_session),
2478 			    allowaddr);
2479 		} else {
2480 			ki->p_tdev = (int32_t)NODEV;
2481 		}
2482 	}
2483 
2484 	if (!P_ZOMBIE(p) && !zombie) {
2485 		ki->p_uvalid = 1;
2486 		ki->p_ustart_sec = p->p_stats->p_start.tv_sec;
2487 		ki->p_ustart_usec = p->p_stats->p_start.tv_usec;
2488 
2489 		calcru(p, &ut, &st, NULL, &rt);
2490 		ki->p_rtime_sec = rt.tv_sec;
2491 		ki->p_rtime_usec = rt.tv_usec;
2492 		ki->p_uutime_sec = ut.tv_sec;
2493 		ki->p_uutime_usec = ut.tv_usec;
2494 		ki->p_ustime_sec = st.tv_sec;
2495 		ki->p_ustime_usec = st.tv_usec;
2496 
2497 		memcpy(&ru, &p->p_stats->p_ru, sizeof(ru));
2498 		ki->p_uru_nvcsw = 0;
2499 		ki->p_uru_nivcsw = 0;
2500 		LIST_FOREACH(l2, &p->p_lwps, l_sibling) {
2501 			ki->p_uru_nvcsw += (l2->l_ncsw - l2->l_nivcsw);
2502 			ki->p_uru_nivcsw += l2->l_nivcsw;
2503 			ruadd(&ru, &l2->l_ru);
2504 		}
2505 		ki->p_uru_maxrss = ru.ru_maxrss;
2506 		ki->p_uru_ixrss = ru.ru_ixrss;
2507 		ki->p_uru_idrss = ru.ru_idrss;
2508 		ki->p_uru_isrss = ru.ru_isrss;
2509 		ki->p_uru_minflt = ru.ru_minflt;
2510 		ki->p_uru_majflt = ru.ru_majflt;
2511 		ki->p_uru_nswap = ru.ru_nswap;
2512 		ki->p_uru_inblock = ru.ru_inblock;
2513 		ki->p_uru_oublock = ru.ru_oublock;
2514 		ki->p_uru_msgsnd = ru.ru_msgsnd;
2515 		ki->p_uru_msgrcv = ru.ru_msgrcv;
2516 		ki->p_uru_nsignals = ru.ru_nsignals;
2517 
2518 		timeradd(&p->p_stats->p_cru.ru_utime,
2519 			 &p->p_stats->p_cru.ru_stime, &ut);
2520 		ki->p_uctime_sec = ut.tv_sec;
2521 		ki->p_uctime_usec = ut.tv_usec;
2522 	}
2523 }
2524 
2525 
2526 int
2527 proc_find_locked(struct lwp *l, struct proc **p, pid_t pid)
2528 {
2529 	int error;
2530 
2531 	mutex_enter(proc_lock);
2532 	if (pid == -1)
2533 		*p = l->l_proc;
2534 	else
2535 		*p = proc_find(pid);
2536 
2537 	if (*p == NULL) {
2538 		if (pid != -1)
2539 			mutex_exit(proc_lock);
2540 		return ESRCH;
2541 	}
2542 	if (pid != -1)
2543 		mutex_enter((*p)->p_lock);
2544 	mutex_exit(proc_lock);
2545 
2546 	error = kauth_authorize_process(l->l_cred,
2547 	    KAUTH_PROCESS_CANSEE, *p,
2548 	    KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL);
2549 	if (error) {
2550 		if (pid != -1)
2551 			mutex_exit((*p)->p_lock);
2552 	}
2553 	return error;
2554 }
2555 
2556 static int
2557 fill_pathname(struct lwp *l, pid_t pid, void *oldp, size_t *oldlenp)
2558 {
2559 	int error;
2560 	struct proc *p;
2561 
2562 	if ((error = proc_find_locked(l, &p, pid)) != 0)
2563 		return error;
2564 
2565 	if (p->p_path == NULL) {
2566 		if (pid != -1)
2567 			mutex_exit(p->p_lock);
2568 		return ENOENT;
2569 	}
2570 
2571 	size_t len = strlen(p->p_path) + 1;
2572 	if (oldp != NULL) {
2573 		size_t copylen = uimin(len, *oldlenp);
2574 		error = sysctl_copyout(l, p->p_path, oldp, copylen);
2575 		if (error == 0 && *oldlenp < len)
2576 			error = ENOSPC;
2577 	}
2578 	*oldlenp = len;
2579 	if (pid != -1)
2580 		mutex_exit(p->p_lock);
2581 	return error;
2582 }
2583 
2584 static int
2585 fill_cwd(struct lwp *l, pid_t pid, void *oldp, size_t *oldlenp)
2586 {
2587 	int error;
2588 	struct proc *p;
2589 	char *path;
2590 	char *bp, *bend;
2591 	struct cwdinfo *cwdi;
2592 	struct vnode *vp;
2593 	size_t len, lenused;
2594 
2595 	if ((error = proc_find_locked(l, &p, pid)) != 0)
2596 		return error;
2597 
2598 	len = MAXPATHLEN * 4;
2599 
2600 	path = kmem_alloc(len, KM_SLEEP);
2601 
2602 	bp = &path[len];
2603 	bend = bp;
2604 	*(--bp) = '\0';
2605 
2606 	cwdi = p->p_cwdi;
2607 	rw_enter(&cwdi->cwdi_lock, RW_READER);
2608 	vp = cwdi->cwdi_cdir;
2609 	error = getcwd_common(vp, NULL, &bp, path, len/2, 0, l);
2610 	rw_exit(&cwdi->cwdi_lock);
2611 
2612 	if (error)
2613 		goto out;
2614 
2615 	lenused = bend - bp;
2616 
2617 	if (oldp != NULL) {
2618 		size_t copylen = uimin(lenused, *oldlenp);
2619 		error = sysctl_copyout(l, bp, oldp, copylen);
2620 		if (error == 0 && *oldlenp < lenused)
2621 			error = ENOSPC;
2622 	}
2623 	*oldlenp = lenused;
2624 out:
2625 	if (pid != -1)
2626 		mutex_exit(p->p_lock);
2627 	kmem_free(path, len);
2628 	return error;
2629 }
2630 
2631 int
2632 proc_getauxv(struct proc *p, void **buf, size_t *len)
2633 {
2634 	struct ps_strings pss;
2635 	int error;
2636 	void *uauxv, *kauxv;
2637 	size_t size;
2638 
2639 	if ((error = copyin_psstrings(p, &pss)) != 0)
2640 		return error;
2641 	if (pss.ps_envstr == NULL)
2642 		return EIO;
2643 
2644 	size = p->p_execsw->es_arglen;
2645 	if (size == 0)
2646 		return EIO;
2647 
2648 	size_t ptrsz = PROC_PTRSZ(p);
2649 	uauxv = (void *)((char *)pss.ps_envstr + (pss.ps_nenvstr + 1) * ptrsz);
2650 
2651 	kauxv = kmem_alloc(size, KM_SLEEP);
2652 
2653 	error = copyin_proc(p, uauxv, kauxv, size);
2654 	if (error) {
2655 		kmem_free(kauxv, size);
2656 		return error;
2657 	}
2658 
2659 	*buf = kauxv;
2660 	*len = size;
2661 
2662 	return 0;
2663 }
2664 
2665 
2666 static int
2667 sysctl_security_expose_address(SYSCTLFN_ARGS)
2668 {
2669 	int expose_address, error;
2670 	struct sysctlnode node;
2671 
2672 	node = *rnode;
2673 	node.sysctl_data = &expose_address;
2674 	expose_address = *(int *)rnode->sysctl_data;
2675 	error = sysctl_lookup(SYSCTLFN_CALL(&node));
2676 	if (error || newp == NULL)
2677 		return error;
2678 
2679 	if (kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_KERNADDR,
2680 	    0, NULL, NULL, NULL))
2681 		return EPERM;
2682 
2683 	switch (expose_address) {
2684 	case 0:
2685 	case 1:
2686 	case 2:
2687 		break;
2688 	default:
2689 		return EINVAL;
2690 	}
2691 
2692 	*(int *)rnode->sysctl_data = expose_address;
2693 
2694 	return 0;
2695 }
2696 
2697 bool
2698 get_expose_address(struct proc *p)
2699 {
2700 	/* allow only if sysctl variable is set or privileged */
2701 	return kauth_authorize_process(kauth_cred_get(), KAUTH_PROCESS_CANSEE,
2702 	    p, KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_KPTR), NULL, NULL) == 0;
2703 }
2704