xref: /netbsd-src/sys/kern/kern_proc.c (revision c9496f6b604074a9451a67df576a5b423068e71e)
1 /*	$NetBSD: kern_proc.c,v 1.209 2017/11/30 18:44:16 maxv 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 1.209 2017/11/30 18:44:16 maxv Exp $");
66 
67 #ifdef _KERNEL_OPT
68 #include "opt_kstack.h"
69 #include "opt_maxuprc.h"
70 #include "opt_dtrace.h"
71 #include "opt_compat_netbsd32.h"
72 #endif
73 
74 #if defined(__HAVE_COMPAT_NETBSD32) && !defined(COMPAT_NETBSD32) \
75     && !defined(_RUMPKERNEL)
76 #define COMPAT_NETBSD32
77 #endif
78 
79 #include <sys/param.h>
80 #include <sys/systm.h>
81 #include <sys/kernel.h>
82 #include <sys/proc.h>
83 #include <sys/resourcevar.h>
84 #include <sys/buf.h>
85 #include <sys/acct.h>
86 #include <sys/wait.h>
87 #include <sys/file.h>
88 #include <ufs/ufs/quota.h>
89 #include <sys/uio.h>
90 #include <sys/pool.h>
91 #include <sys/pset.h>
92 #include <sys/mbuf.h>
93 #include <sys/ioctl.h>
94 #include <sys/tty.h>
95 #include <sys/signalvar.h>
96 #include <sys/ras.h>
97 #include <sys/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 
109 #include <uvm/uvm_extern.h>
110 #include <uvm/uvm.h>
111 
112 #ifdef COMPAT_NETBSD32
113 #include <compat/netbsd32/netbsd32.h>
114 #endif
115 
116 /*
117  * Process lists.
118  */
119 
120 struct proclist		allproc		__cacheline_aligned;
121 struct proclist		zombproc	__cacheline_aligned;
122 
123 kmutex_t *		proc_lock	__cacheline_aligned;
124 
125 /*
126  * pid to proc lookup is done by indexing the pid_table array.
127  * Since pid numbers are only allocated when an empty slot
128  * has been found, there is no need to search any lists ever.
129  * (an orphaned pgrp will lock the slot, a session will lock
130  * the pgrp with the same number.)
131  * If the table is too small it is reallocated with twice the
132  * previous size and the entries 'unzipped' into the two halves.
133  * A linked list of free entries is passed through the pt_proc
134  * field of 'free' items - set odd to be an invalid ptr.
135  */
136 
137 struct pid_table {
138 	struct proc	*pt_proc;
139 	struct pgrp	*pt_pgrp;
140 	pid_t		pt_pid;
141 };
142 #if 1	/* strongly typed cast - should be a noop */
143 static inline uint p2u(struct proc *p) { return (uint)(uintptr_t)p; }
144 #else
145 #define p2u(p) ((uint)p)
146 #endif
147 #define P_VALID(p) (!(p2u(p) & 1))
148 #define P_NEXT(p) (p2u(p) >> 1)
149 #define P_FREE(pid) ((struct proc *)(uintptr_t)((pid) << 1 | 1))
150 
151 /*
152  * Table of process IDs (PIDs).
153  */
154 static struct pid_table *pid_table	__read_mostly;
155 
156 #define	INITIAL_PID_TABLE_SIZE		(1 << 5)
157 
158 /* Table mask, threshold for growing and number of allocated PIDs. */
159 static u_int		pid_tbl_mask	__read_mostly;
160 static u_int		pid_alloc_lim	__read_mostly;
161 static u_int		pid_alloc_cnt	__cacheline_aligned;
162 
163 /* Next free, last free and maximum PIDs. */
164 static u_int		next_free_pt	__cacheline_aligned;
165 static u_int		last_free_pt	__cacheline_aligned;
166 static pid_t		pid_max		__read_mostly;
167 
168 /* Components of the first process -- never freed. */
169 
170 extern struct emul emul_netbsd;	/* defined in kern_exec.c */
171 
172 struct session session0 = {
173 	.s_count = 1,
174 	.s_sid = 0,
175 };
176 struct pgrp pgrp0 = {
177 	.pg_members = LIST_HEAD_INITIALIZER(&pgrp0.pg_members),
178 	.pg_session = &session0,
179 };
180 filedesc_t filedesc0;
181 struct cwdinfo cwdi0 = {
182 	.cwdi_cmask = CMASK,
183 	.cwdi_refcnt = 1,
184 };
185 struct plimit limit0;
186 struct pstats pstat0;
187 struct vmspace vmspace0;
188 struct sigacts sigacts0;
189 struct proc proc0 = {
190 	.p_lwps = LIST_HEAD_INITIALIZER(&proc0.p_lwps),
191 	.p_sigwaiters = LIST_HEAD_INITIALIZER(&proc0.p_sigwaiters),
192 	.p_nlwps = 1,
193 	.p_nrlwps = 1,
194 	.p_nlwpid = 1,		/* must match lwp0.l_lid */
195 	.p_pgrp = &pgrp0,
196 	.p_comm = "system",
197 	/*
198 	 * Set P_NOCLDWAIT so that kernel threads are reparented to init(8)
199 	 * when they exit.  init(8) can easily wait them out for us.
200 	 */
201 	.p_flag = PK_SYSTEM | PK_NOCLDWAIT,
202 	.p_stat = SACTIVE,
203 	.p_nice = NZERO,
204 	.p_emul = &emul_netbsd,
205 	.p_cwdi = &cwdi0,
206 	.p_limit = &limit0,
207 	.p_fd = &filedesc0,
208 	.p_vmspace = &vmspace0,
209 	.p_stats = &pstat0,
210 	.p_sigacts = &sigacts0,
211 #ifdef PROC0_MD_INITIALIZERS
212 	PROC0_MD_INITIALIZERS
213 #endif
214 };
215 kauth_cred_t cred0;
216 
217 static const int	nofile	= NOFILE;
218 static const int	maxuprc	= MAXUPRC;
219 
220 static int sysctl_doeproc(SYSCTLFN_PROTO);
221 static int sysctl_kern_proc_args(SYSCTLFN_PROTO);
222 
223 /*
224  * The process list descriptors, used during pid allocation and
225  * by sysctl.  No locking on this data structure is needed since
226  * it is completely static.
227  */
228 const struct proclist_desc proclists[] = {
229 	{ &allproc	},
230 	{ &zombproc	},
231 	{ NULL		},
232 };
233 
234 static struct pgrp *	pg_remove(pid_t);
235 static void		pg_delete(pid_t);
236 static void		orphanpg(struct pgrp *);
237 
238 static specificdata_domain_t proc_specificdata_domain;
239 
240 static pool_cache_t proc_cache;
241 
242 static kauth_listener_t proc_listener;
243 
244 static int fill_pathname(struct lwp *, pid_t, void *, size_t *);
245 
246 static int
247 proc_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
248     void *arg0, void *arg1, void *arg2, void *arg3)
249 {
250 	struct proc *p;
251 	int result;
252 
253 	result = KAUTH_RESULT_DEFER;
254 	p = arg0;
255 
256 	switch (action) {
257 	case KAUTH_PROCESS_CANSEE: {
258 		enum kauth_process_req req;
259 
260 		req = (enum kauth_process_req)arg1;
261 
262 		switch (req) {
263 		case KAUTH_REQ_PROCESS_CANSEE_ARGS:
264 		case KAUTH_REQ_PROCESS_CANSEE_ENTRY:
265 		case KAUTH_REQ_PROCESS_CANSEE_OPENFILES:
266 			result = KAUTH_RESULT_ALLOW;
267 
268 			break;
269 
270 		case KAUTH_REQ_PROCESS_CANSEE_ENV:
271 			if (kauth_cred_getuid(cred) !=
272 			    kauth_cred_getuid(p->p_cred) ||
273 			    kauth_cred_getuid(cred) !=
274 			    kauth_cred_getsvuid(p->p_cred))
275 				break;
276 
277 			result = KAUTH_RESULT_ALLOW;
278 
279 			break;
280 
281 		default:
282 			break;
283 		}
284 
285 		break;
286 		}
287 
288 	case KAUTH_PROCESS_FORK: {
289 		int lnprocs = (int)(unsigned long)arg2;
290 
291 		/*
292 		 * Don't allow a nonprivileged user to use the last few
293 		 * processes. The variable lnprocs is the current number of
294 		 * processes, maxproc is the limit.
295 		 */
296 		if (__predict_false((lnprocs >= maxproc - 5)))
297 			break;
298 
299 		result = KAUTH_RESULT_ALLOW;
300 
301 		break;
302 		}
303 
304 	case KAUTH_PROCESS_CORENAME:
305 	case KAUTH_PROCESS_STOPFLAG:
306 		if (proc_uidmatch(cred, p->p_cred) == 0)
307 			result = KAUTH_RESULT_ALLOW;
308 
309 		break;
310 
311 	default:
312 		break;
313 	}
314 
315 	return result;
316 }
317 
318 /*
319  * Initialize global process hashing structures.
320  */
321 void
322 procinit(void)
323 {
324 	const struct proclist_desc *pd;
325 	u_int i;
326 #define	LINK_EMPTY ((PID_MAX + INITIAL_PID_TABLE_SIZE) & ~(INITIAL_PID_TABLE_SIZE - 1))
327 
328 	for (pd = proclists; pd->pd_list != NULL; pd++)
329 		LIST_INIT(pd->pd_list);
330 
331 	proc_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
332 	pid_table = kmem_alloc(INITIAL_PID_TABLE_SIZE
333 	    * sizeof(struct pid_table), KM_SLEEP);
334 	pid_tbl_mask = INITIAL_PID_TABLE_SIZE - 1;
335 	pid_max = PID_MAX;
336 
337 	/* Set free list running through table...
338 	   Preset 'use count' above PID_MAX so we allocate pid 1 next. */
339 	for (i = 0; i <= pid_tbl_mask; i++) {
340 		pid_table[i].pt_proc = P_FREE(LINK_EMPTY + i + 1);
341 		pid_table[i].pt_pgrp = 0;
342 		pid_table[i].pt_pid = 0;
343 	}
344 	/* slot 0 is just grabbed */
345 	next_free_pt = 1;
346 	/* Need to fix last entry. */
347 	last_free_pt = pid_tbl_mask;
348 	pid_table[last_free_pt].pt_proc = P_FREE(LINK_EMPTY);
349 	/* point at which we grow table - to avoid reusing pids too often */
350 	pid_alloc_lim = pid_tbl_mask - 1;
351 #undef LINK_EMPTY
352 
353 	proc_specificdata_domain = specificdata_domain_create();
354 	KASSERT(proc_specificdata_domain != NULL);
355 
356 	proc_cache = pool_cache_init(sizeof(struct proc), 0, 0, 0,
357 	    "procpl", NULL, IPL_NONE, NULL, NULL, NULL);
358 
359 	proc_listener = kauth_listen_scope(KAUTH_SCOPE_PROCESS,
360 	    proc_listener_cb, NULL);
361 }
362 
363 void
364 procinit_sysctl(void)
365 {
366 	static struct sysctllog *clog;
367 
368 	sysctl_createv(&clog, 0, NULL, NULL,
369 		       CTLFLAG_PERMANENT,
370 		       CTLTYPE_NODE, "proc",
371 		       SYSCTL_DESCR("System-wide process information"),
372 		       sysctl_doeproc, 0, NULL, 0,
373 		       CTL_KERN, KERN_PROC, CTL_EOL);
374 	sysctl_createv(&clog, 0, NULL, NULL,
375 		       CTLFLAG_PERMANENT,
376 		       CTLTYPE_NODE, "proc2",
377 		       SYSCTL_DESCR("Machine-independent process information"),
378 		       sysctl_doeproc, 0, NULL, 0,
379 		       CTL_KERN, KERN_PROC2, CTL_EOL);
380 	sysctl_createv(&clog, 0, NULL, NULL,
381 		       CTLFLAG_PERMANENT,
382 		       CTLTYPE_NODE, "proc_args",
383 		       SYSCTL_DESCR("Process argument information"),
384 		       sysctl_kern_proc_args, 0, NULL, 0,
385 		       CTL_KERN, KERN_PROC_ARGS, CTL_EOL);
386 
387 	/*
388 	  "nodes" under these:
389 
390 	  KERN_PROC_ALL
391 	  KERN_PROC_PID pid
392 	  KERN_PROC_PGRP pgrp
393 	  KERN_PROC_SESSION sess
394 	  KERN_PROC_TTY tty
395 	  KERN_PROC_UID uid
396 	  KERN_PROC_RUID uid
397 	  KERN_PROC_GID gid
398 	  KERN_PROC_RGID gid
399 
400 	  all in all, probably not worth the effort...
401 	*/
402 }
403 
404 /*
405  * Initialize process 0.
406  */
407 void
408 proc0_init(void)
409 {
410 	struct proc *p;
411 	struct pgrp *pg;
412 	struct rlimit *rlim;
413 	rlim_t lim;
414 	int i;
415 
416 	p = &proc0;
417 	pg = &pgrp0;
418 
419 	mutex_init(&p->p_stmutex, MUTEX_DEFAULT, IPL_HIGH);
420 	mutex_init(&p->p_auxlock, MUTEX_DEFAULT, IPL_NONE);
421 	p->p_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
422 
423 	rw_init(&p->p_reflock);
424 	cv_init(&p->p_waitcv, "wait");
425 	cv_init(&p->p_lwpcv, "lwpwait");
426 
427 	LIST_INSERT_HEAD(&p->p_lwps, &lwp0, l_sibling);
428 
429 	pid_table[0].pt_proc = p;
430 	LIST_INSERT_HEAD(&allproc, p, p_list);
431 
432 	pid_table[0].pt_pgrp = pg;
433 	LIST_INSERT_HEAD(&pg->pg_members, p, p_pglist);
434 
435 #ifdef __HAVE_SYSCALL_INTERN
436 	(*p->p_emul->e_syscall_intern)(p);
437 #endif
438 
439 	/* Create credentials. */
440 	cred0 = kauth_cred_alloc();
441 	p->p_cred = cred0;
442 
443 	/* Create the CWD info. */
444 	rw_init(&cwdi0.cwdi_lock);
445 
446 	/* Create the limits structures. */
447 	mutex_init(&limit0.pl_lock, MUTEX_DEFAULT, IPL_NONE);
448 
449 	rlim = limit0.pl_rlimit;
450 	for (i = 0; i < __arraycount(limit0.pl_rlimit); i++) {
451 		rlim[i].rlim_cur = RLIM_INFINITY;
452 		rlim[i].rlim_max = RLIM_INFINITY;
453 	}
454 
455 	rlim[RLIMIT_NOFILE].rlim_max = maxfiles;
456 	rlim[RLIMIT_NOFILE].rlim_cur = maxfiles < nofile ? maxfiles : nofile;
457 
458 	rlim[RLIMIT_NPROC].rlim_max = maxproc;
459 	rlim[RLIMIT_NPROC].rlim_cur = maxproc < maxuprc ? maxproc : maxuprc;
460 
461 	lim = MIN(VM_MAXUSER_ADDRESS, ctob((rlim_t)uvmexp.free));
462 	rlim[RLIMIT_RSS].rlim_max = lim;
463 	rlim[RLIMIT_MEMLOCK].rlim_max = lim;
464 	rlim[RLIMIT_MEMLOCK].rlim_cur = lim / 3;
465 
466 	rlim[RLIMIT_NTHR].rlim_max = maxlwp;
467 	rlim[RLIMIT_NTHR].rlim_cur = maxlwp < maxuprc ? maxlwp : maxuprc;
468 
469 	/* Note that default core name has zero length. */
470 	limit0.pl_corename = defcorename;
471 	limit0.pl_cnlen = 0;
472 	limit0.pl_refcnt = 1;
473 	limit0.pl_writeable = false;
474 	limit0.pl_sv_limit = NULL;
475 
476 	/* Configure virtual memory system, set vm rlimits. */
477 	uvm_init_limits(p);
478 
479 	/* Initialize file descriptor table for proc0. */
480 	fd_init(&filedesc0);
481 
482 	/*
483 	 * Initialize proc0's vmspace, which uses the kernel pmap.
484 	 * All kernel processes (which never have user space mappings)
485 	 * share proc0's vmspace, and thus, the kernel pmap.
486 	 */
487 	uvmspace_init(&vmspace0, pmap_kernel(), round_page(VM_MIN_ADDRESS),
488 	    trunc_page(VM_MAXUSER_ADDRESS),
489 #ifdef __USE_TOPDOWN_VM
490 	    true
491 #else
492 	    false
493 #endif
494 	    );
495 
496 	/* Initialize signal state for proc0. XXX IPL_SCHED */
497 	mutex_init(&p->p_sigacts->sa_mutex, MUTEX_DEFAULT, IPL_SCHED);
498 	siginit(p);
499 
500 	proc_initspecific(p);
501 	kdtrace_proc_ctor(NULL, p);
502 }
503 
504 /*
505  * Session reference counting.
506  */
507 
508 void
509 proc_sesshold(struct session *ss)
510 {
511 
512 	KASSERT(mutex_owned(proc_lock));
513 	ss->s_count++;
514 }
515 
516 void
517 proc_sessrele(struct session *ss)
518 {
519 
520 	KASSERT(mutex_owned(proc_lock));
521 	/*
522 	 * We keep the pgrp with the same id as the session in order to
523 	 * stop a process being given the same pid.  Since the pgrp holds
524 	 * a reference to the session, it must be a 'zombie' pgrp by now.
525 	 */
526 	if (--ss->s_count == 0) {
527 		struct pgrp *pg;
528 
529 		pg = pg_remove(ss->s_sid);
530 		mutex_exit(proc_lock);
531 
532 		kmem_free(pg, sizeof(struct pgrp));
533 		kmem_free(ss, sizeof(struct session));
534 	} else {
535 		mutex_exit(proc_lock);
536 	}
537 }
538 
539 /*
540  * Check that the specified process group is in the session of the
541  * specified process.
542  * Treats -ve ids as process ids.
543  * Used to validate TIOCSPGRP requests.
544  */
545 int
546 pgid_in_session(struct proc *p, pid_t pg_id)
547 {
548 	struct pgrp *pgrp;
549 	struct session *session;
550 	int error;
551 
552 	mutex_enter(proc_lock);
553 	if (pg_id < 0) {
554 		struct proc *p1 = proc_find(-pg_id);
555 		if (p1 == NULL) {
556 			error = EINVAL;
557 			goto fail;
558 		}
559 		pgrp = p1->p_pgrp;
560 	} else {
561 		pgrp = pgrp_find(pg_id);
562 		if (pgrp == NULL) {
563 			error = EINVAL;
564 			goto fail;
565 		}
566 	}
567 	session = pgrp->pg_session;
568 	error = (session != p->p_pgrp->pg_session) ? EPERM : 0;
569 fail:
570 	mutex_exit(proc_lock);
571 	return error;
572 }
573 
574 /*
575  * p_inferior: is p an inferior of q?
576  */
577 static inline bool
578 p_inferior(struct proc *p, struct proc *q)
579 {
580 
581 	KASSERT(mutex_owned(proc_lock));
582 
583 	for (; p != q; p = p->p_pptr)
584 		if (p->p_pid == 0)
585 			return false;
586 	return true;
587 }
588 
589 /*
590  * proc_find: locate a process by the ID.
591  *
592  * => Must be called with proc_lock held.
593  */
594 proc_t *
595 proc_find_raw(pid_t pid)
596 {
597 	struct pid_table *pt;
598 	proc_t *p;
599 
600 	KASSERT(mutex_owned(proc_lock));
601 	pt = &pid_table[pid & pid_tbl_mask];
602 	p = pt->pt_proc;
603 	if (__predict_false(!P_VALID(p) || pt->pt_pid != pid)) {
604 		return NULL;
605 	}
606 	return p;
607 }
608 
609 proc_t *
610 proc_find(pid_t pid)
611 {
612 	proc_t *p;
613 
614 	p = proc_find_raw(pid);
615 	if (__predict_false(p == NULL)) {
616 		return NULL;
617 	}
618 
619 	/*
620 	 * Only allow live processes to be found by PID.
621 	 * XXX: p_stat might change, since unlocked.
622 	 */
623 	if (__predict_true(p->p_stat == SACTIVE || p->p_stat == SSTOP)) {
624 		return p;
625 	}
626 	return NULL;
627 }
628 
629 /*
630  * pgrp_find: locate a process group by the ID.
631  *
632  * => Must be called with proc_lock held.
633  */
634 struct pgrp *
635 pgrp_find(pid_t pgid)
636 {
637 	struct pgrp *pg;
638 
639 	KASSERT(mutex_owned(proc_lock));
640 
641 	pg = pid_table[pgid & pid_tbl_mask].pt_pgrp;
642 
643 	/*
644 	 * Cannot look up a process group that only exists because the
645 	 * session has not died yet (traditional).
646 	 */
647 	if (pg == NULL || pg->pg_id != pgid || LIST_EMPTY(&pg->pg_members)) {
648 		return NULL;
649 	}
650 	return pg;
651 }
652 
653 static void
654 expand_pid_table(void)
655 {
656 	size_t pt_size, tsz;
657 	struct pid_table *n_pt, *new_pt;
658 	struct proc *proc;
659 	struct pgrp *pgrp;
660 	pid_t pid, rpid;
661 	u_int i;
662 	uint new_pt_mask;
663 
664 	pt_size = pid_tbl_mask + 1;
665 	tsz = pt_size * 2 * sizeof(struct pid_table);
666 	new_pt = kmem_alloc(tsz, KM_SLEEP);
667 	new_pt_mask = pt_size * 2 - 1;
668 
669 	mutex_enter(proc_lock);
670 	if (pt_size != pid_tbl_mask + 1) {
671 		/* Another process beat us to it... */
672 		mutex_exit(proc_lock);
673 		kmem_free(new_pt, tsz);
674 		return;
675 	}
676 
677 	/*
678 	 * Copy entries from old table into new one.
679 	 * If 'pid' is 'odd' we need to place in the upper half,
680 	 * even pid's to the lower half.
681 	 * Free items stay in the low half so we don't have to
682 	 * fixup the reference to them.
683 	 * We stuff free items on the front of the freelist
684 	 * because we can't write to unmodified entries.
685 	 * Processing the table backwards maintains a semblance
686 	 * of issuing pid numbers that increase with time.
687 	 */
688 	i = pt_size - 1;
689 	n_pt = new_pt + i;
690 	for (; ; i--, n_pt--) {
691 		proc = pid_table[i].pt_proc;
692 		pgrp = pid_table[i].pt_pgrp;
693 		if (!P_VALID(proc)) {
694 			/* Up 'use count' so that link is valid */
695 			pid = (P_NEXT(proc) + pt_size) & ~pt_size;
696 			rpid = 0;
697 			proc = P_FREE(pid);
698 			if (pgrp)
699 				pid = pgrp->pg_id;
700 		} else {
701 			pid = pid_table[i].pt_pid;
702 			rpid = pid;
703 		}
704 
705 		/* Save entry in appropriate half of table */
706 		n_pt[pid & pt_size].pt_proc = proc;
707 		n_pt[pid & pt_size].pt_pgrp = pgrp;
708 		n_pt[pid & pt_size].pt_pid = rpid;
709 
710 		/* Put other piece on start of free list */
711 		pid = (pid ^ pt_size) & ~pid_tbl_mask;
712 		n_pt[pid & pt_size].pt_proc =
713 			P_FREE((pid & ~pt_size) | next_free_pt);
714 		n_pt[pid & pt_size].pt_pgrp = 0;
715 		n_pt[pid & pt_size].pt_pid = 0;
716 
717 		next_free_pt = i | (pid & pt_size);
718 		if (i == 0)
719 			break;
720 	}
721 
722 	/* Save old table size and switch tables */
723 	tsz = pt_size * sizeof(struct pid_table);
724 	n_pt = pid_table;
725 	pid_table = new_pt;
726 	pid_tbl_mask = new_pt_mask;
727 
728 	/*
729 	 * pid_max starts as PID_MAX (= 30000), once we have 16384
730 	 * allocated pids we need it to be larger!
731 	 */
732 	if (pid_tbl_mask > PID_MAX) {
733 		pid_max = pid_tbl_mask * 2 + 1;
734 		pid_alloc_lim |= pid_alloc_lim << 1;
735 	} else
736 		pid_alloc_lim <<= 1;	/* doubles number of free slots... */
737 
738 	mutex_exit(proc_lock);
739 	kmem_free(n_pt, tsz);
740 }
741 
742 struct proc *
743 proc_alloc(void)
744 {
745 	struct proc *p;
746 
747 	p = pool_cache_get(proc_cache, PR_WAITOK);
748 	p->p_stat = SIDL;			/* protect against others */
749 	proc_initspecific(p);
750 	kdtrace_proc_ctor(NULL, p);
751 	p->p_pid = -1;
752 	proc_alloc_pid(p);
753 	return p;
754 }
755 
756 /*
757  * proc_alloc_pid: allocate PID and record the given proc 'p' so that
758  * proc_find_raw() can find it by the PID.
759  */
760 
761 pid_t
762 proc_alloc_pid(struct proc *p)
763 {
764 	struct pid_table *pt;
765 	pid_t pid;
766 	int nxt;
767 
768 	for (;;expand_pid_table()) {
769 		if (__predict_false(pid_alloc_cnt >= pid_alloc_lim))
770 			/* ensure pids cycle through 2000+ values */
771 			continue;
772 		mutex_enter(proc_lock);
773 		pt = &pid_table[next_free_pt];
774 #ifdef DIAGNOSTIC
775 		if (__predict_false(P_VALID(pt->pt_proc) || pt->pt_pgrp))
776 			panic("proc_alloc: slot busy");
777 #endif
778 		nxt = P_NEXT(pt->pt_proc);
779 		if (nxt & pid_tbl_mask)
780 			break;
781 		/* Table full - expand (NB last entry not used....) */
782 		mutex_exit(proc_lock);
783 	}
784 
785 	/* pid is 'saved use count' + 'size' + entry */
786 	pid = (nxt & ~pid_tbl_mask) + pid_tbl_mask + 1 + next_free_pt;
787 	if ((uint)pid > (uint)pid_max)
788 		pid &= pid_tbl_mask;
789 	next_free_pt = nxt & pid_tbl_mask;
790 
791 	/* Grab table slot */
792 	pt->pt_proc = p;
793 
794 	KASSERT(pt->pt_pid == 0);
795 	pt->pt_pid = pid;
796 	if (p->p_pid == -1) {
797 		p->p_pid = pid;
798 	}
799 	pid_alloc_cnt++;
800 	mutex_exit(proc_lock);
801 
802 	return pid;
803 }
804 
805 /*
806  * Free a process id - called from proc_free (in kern_exit.c)
807  *
808  * Called with the proc_lock held.
809  */
810 void
811 proc_free_pid(pid_t pid)
812 {
813 	struct pid_table *pt;
814 
815 	KASSERT(mutex_owned(proc_lock));
816 
817 	pt = &pid_table[pid & pid_tbl_mask];
818 
819 	/* save pid use count in slot */
820 	pt->pt_proc = P_FREE(pid & ~pid_tbl_mask);
821 	KASSERT(pt->pt_pid == pid);
822 	pt->pt_pid = 0;
823 
824 	if (pt->pt_pgrp == NULL) {
825 		/* link last freed entry onto ours */
826 		pid &= pid_tbl_mask;
827 		pt = &pid_table[last_free_pt];
828 		pt->pt_proc = P_FREE(P_NEXT(pt->pt_proc) | pid);
829 		pt->pt_pid = 0;
830 		last_free_pt = pid;
831 		pid_alloc_cnt--;
832 	}
833 
834 	atomic_dec_uint(&nprocs);
835 }
836 
837 void
838 proc_free_mem(struct proc *p)
839 {
840 
841 	kdtrace_proc_dtor(NULL, p);
842 	pool_cache_put(proc_cache, p);
843 }
844 
845 /*
846  * proc_enterpgrp: move p to a new or existing process group (and session).
847  *
848  * If we are creating a new pgrp, the pgid should equal
849  * the calling process' pid.
850  * If is only valid to enter a process group that is in the session
851  * of the process.
852  * Also mksess should only be set if we are creating a process group
853  *
854  * Only called from sys_setsid, sys_setpgid and posix_spawn/spawn_return.
855  */
856 int
857 proc_enterpgrp(struct proc *curp, pid_t pid, pid_t pgid, bool mksess)
858 {
859 	struct pgrp *new_pgrp, *pgrp;
860 	struct session *sess;
861 	struct proc *p;
862 	int rval;
863 	pid_t pg_id = NO_PGID;
864 
865 	sess = mksess ? kmem_alloc(sizeof(*sess), KM_SLEEP) : NULL;
866 
867 	/* Allocate data areas we might need before doing any validity checks */
868 	mutex_enter(proc_lock);		/* Because pid_table might change */
869 	if (pid_table[pgid & pid_tbl_mask].pt_pgrp == 0) {
870 		mutex_exit(proc_lock);
871 		new_pgrp = kmem_alloc(sizeof(*new_pgrp), KM_SLEEP);
872 		mutex_enter(proc_lock);
873 	} else
874 		new_pgrp = NULL;
875 	rval = EPERM;	/* most common error (to save typing) */
876 
877 	/* Check pgrp exists or can be created */
878 	pgrp = pid_table[pgid & pid_tbl_mask].pt_pgrp;
879 	if (pgrp != NULL && pgrp->pg_id != pgid)
880 		goto done;
881 
882 	/* Can only set another process under restricted circumstances. */
883 	if (pid != curp->p_pid) {
884 		/* Must exist and be one of our children... */
885 		p = proc_find(pid);
886 		if (p == NULL || !p_inferior(p, curp)) {
887 			rval = ESRCH;
888 			goto done;
889 		}
890 		/* ... in the same session... */
891 		if (sess != NULL || p->p_session != curp->p_session)
892 			goto done;
893 		/* ... existing pgid must be in same session ... */
894 		if (pgrp != NULL && pgrp->pg_session != p->p_session)
895 			goto done;
896 		/* ... and not done an exec. */
897 		if (p->p_flag & PK_EXEC) {
898 			rval = EACCES;
899 			goto done;
900 		}
901 	} else {
902 		/* ... setsid() cannot re-enter a pgrp */
903 		if (mksess && (curp->p_pgid == curp->p_pid ||
904 		    pgrp_find(curp->p_pid)))
905 			goto done;
906 		p = curp;
907 	}
908 
909 	/* Changing the process group/session of a session
910 	   leader is definitely off limits. */
911 	if (SESS_LEADER(p)) {
912 		if (sess == NULL && p->p_pgrp == pgrp)
913 			/* unless it's a definite noop */
914 			rval = 0;
915 		goto done;
916 	}
917 
918 	/* Can only create a process group with id of process */
919 	if (pgrp == NULL && pgid != pid)
920 		goto done;
921 
922 	/* Can only create a session if creating pgrp */
923 	if (sess != NULL && pgrp != NULL)
924 		goto done;
925 
926 	/* Check we allocated memory for a pgrp... */
927 	if (pgrp == NULL && new_pgrp == NULL)
928 		goto done;
929 
930 	/* Don't attach to 'zombie' pgrp */
931 	if (pgrp != NULL && LIST_EMPTY(&pgrp->pg_members))
932 		goto done;
933 
934 	/* Expect to succeed now */
935 	rval = 0;
936 
937 	if (pgrp == p->p_pgrp)
938 		/* nothing to do */
939 		goto done;
940 
941 	/* Ok all setup, link up required structures */
942 
943 	if (pgrp == NULL) {
944 		pgrp = new_pgrp;
945 		new_pgrp = NULL;
946 		if (sess != NULL) {
947 			sess->s_sid = p->p_pid;
948 			sess->s_leader = p;
949 			sess->s_count = 1;
950 			sess->s_ttyvp = NULL;
951 			sess->s_ttyp = NULL;
952 			sess->s_flags = p->p_session->s_flags & ~S_LOGIN_SET;
953 			memcpy(sess->s_login, p->p_session->s_login,
954 			    sizeof(sess->s_login));
955 			p->p_lflag &= ~PL_CONTROLT;
956 		} else {
957 			sess = p->p_pgrp->pg_session;
958 			proc_sesshold(sess);
959 		}
960 		pgrp->pg_session = sess;
961 		sess = NULL;
962 
963 		pgrp->pg_id = pgid;
964 		LIST_INIT(&pgrp->pg_members);
965 #ifdef DIAGNOSTIC
966 		if (__predict_false(pid_table[pgid & pid_tbl_mask].pt_pgrp))
967 			panic("enterpgrp: pgrp table slot in use");
968 		if (__predict_false(mksess && p != curp))
969 			panic("enterpgrp: mksession and p != curproc");
970 #endif
971 		pid_table[pgid & pid_tbl_mask].pt_pgrp = pgrp;
972 		pgrp->pg_jobc = 0;
973 	}
974 
975 	/*
976 	 * Adjust eligibility of affected pgrps to participate in job control.
977 	 * Increment eligibility counts before decrementing, otherwise we
978 	 * could reach 0 spuriously during the first call.
979 	 */
980 	fixjobc(p, pgrp, 1);
981 	fixjobc(p, p->p_pgrp, 0);
982 
983 	/* Interlock with ttread(). */
984 	mutex_spin_enter(&tty_lock);
985 
986 	/* Move process to requested group. */
987 	LIST_REMOVE(p, p_pglist);
988 	if (LIST_EMPTY(&p->p_pgrp->pg_members))
989 		/* defer delete until we've dumped the lock */
990 		pg_id = p->p_pgrp->pg_id;
991 	p->p_pgrp = pgrp;
992 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
993 
994 	/* Done with the swap; we can release the tty mutex. */
995 	mutex_spin_exit(&tty_lock);
996 
997     done:
998 	if (pg_id != NO_PGID) {
999 		/* Releases proc_lock. */
1000 		pg_delete(pg_id);
1001 	} else {
1002 		mutex_exit(proc_lock);
1003 	}
1004 	if (sess != NULL)
1005 		kmem_free(sess, sizeof(*sess));
1006 	if (new_pgrp != NULL)
1007 		kmem_free(new_pgrp, sizeof(*new_pgrp));
1008 #ifdef DEBUG_PGRP
1009 	if (__predict_false(rval))
1010 		printf("enterpgrp(%d,%d,%d), curproc %d, rval %d\n",
1011 			pid, pgid, mksess, curp->p_pid, rval);
1012 #endif
1013 	return rval;
1014 }
1015 
1016 /*
1017  * proc_leavepgrp: remove a process from its process group.
1018  *  => must be called with the proc_lock held, which will be released;
1019  */
1020 void
1021 proc_leavepgrp(struct proc *p)
1022 {
1023 	struct pgrp *pgrp;
1024 
1025 	KASSERT(mutex_owned(proc_lock));
1026 
1027 	/* Interlock with ttread() */
1028 	mutex_spin_enter(&tty_lock);
1029 	pgrp = p->p_pgrp;
1030 	LIST_REMOVE(p, p_pglist);
1031 	p->p_pgrp = NULL;
1032 	mutex_spin_exit(&tty_lock);
1033 
1034 	if (LIST_EMPTY(&pgrp->pg_members)) {
1035 		/* Releases proc_lock. */
1036 		pg_delete(pgrp->pg_id);
1037 	} else {
1038 		mutex_exit(proc_lock);
1039 	}
1040 }
1041 
1042 /*
1043  * pg_remove: remove a process group from the table.
1044  *  => must be called with the proc_lock held;
1045  *  => returns process group to free;
1046  */
1047 static struct pgrp *
1048 pg_remove(pid_t pg_id)
1049 {
1050 	struct pgrp *pgrp;
1051 	struct pid_table *pt;
1052 
1053 	KASSERT(mutex_owned(proc_lock));
1054 
1055 	pt = &pid_table[pg_id & pid_tbl_mask];
1056 	pgrp = pt->pt_pgrp;
1057 
1058 	KASSERT(pgrp != NULL);
1059 	KASSERT(pgrp->pg_id == pg_id);
1060 	KASSERT(LIST_EMPTY(&pgrp->pg_members));
1061 
1062 	pt->pt_pgrp = NULL;
1063 
1064 	if (!P_VALID(pt->pt_proc)) {
1065 		/* Orphaned pgrp, put slot onto free list. */
1066 		KASSERT((P_NEXT(pt->pt_proc) & pid_tbl_mask) == 0);
1067 		pg_id &= pid_tbl_mask;
1068 		pt = &pid_table[last_free_pt];
1069 		pt->pt_proc = P_FREE(P_NEXT(pt->pt_proc) | pg_id);
1070 		KASSERT(pt->pt_pid == 0);
1071 		last_free_pt = pg_id;
1072 		pid_alloc_cnt--;
1073 	}
1074 	return pgrp;
1075 }
1076 
1077 /*
1078  * pg_delete: delete and free a process group.
1079  *  => must be called with the proc_lock held, which will be released.
1080  */
1081 static void
1082 pg_delete(pid_t pg_id)
1083 {
1084 	struct pgrp *pg;
1085 	struct tty *ttyp;
1086 	struct session *ss;
1087 
1088 	KASSERT(mutex_owned(proc_lock));
1089 
1090 	pg = pid_table[pg_id & pid_tbl_mask].pt_pgrp;
1091 	if (pg == NULL || pg->pg_id != pg_id || !LIST_EMPTY(&pg->pg_members)) {
1092 		mutex_exit(proc_lock);
1093 		return;
1094 	}
1095 
1096 	ss = pg->pg_session;
1097 
1098 	/* Remove reference (if any) from tty to this process group */
1099 	mutex_spin_enter(&tty_lock);
1100 	ttyp = ss->s_ttyp;
1101 	if (ttyp != NULL && ttyp->t_pgrp == pg) {
1102 		ttyp->t_pgrp = NULL;
1103 		KASSERT(ttyp->t_session == ss);
1104 	}
1105 	mutex_spin_exit(&tty_lock);
1106 
1107 	/*
1108 	 * The leading process group in a session is freed by proc_sessrele(),
1109 	 * if last reference.  Note: proc_sessrele() releases proc_lock.
1110 	 */
1111 	pg = (ss->s_sid != pg->pg_id) ? pg_remove(pg_id) : NULL;
1112 	proc_sessrele(ss);
1113 
1114 	if (pg != NULL) {
1115 		/* Free it, if was not done by proc_sessrele(). */
1116 		kmem_free(pg, sizeof(struct pgrp));
1117 	}
1118 }
1119 
1120 /*
1121  * Adjust pgrp jobc counters when specified process changes process group.
1122  * We count the number of processes in each process group that "qualify"
1123  * the group for terminal job control (those with a parent in a different
1124  * process group of the same session).  If that count reaches zero, the
1125  * process group becomes orphaned.  Check both the specified process'
1126  * process group and that of its children.
1127  * entering == 0 => p is leaving specified group.
1128  * entering == 1 => p is entering specified group.
1129  *
1130  * Call with proc_lock held.
1131  */
1132 void
1133 fixjobc(struct proc *p, struct pgrp *pgrp, int entering)
1134 {
1135 	struct pgrp *hispgrp;
1136 	struct session *mysession = pgrp->pg_session;
1137 	struct proc *child;
1138 
1139 	KASSERT(mutex_owned(proc_lock));
1140 
1141 	/*
1142 	 * Check p's parent to see whether p qualifies its own process
1143 	 * group; if so, adjust count for p's process group.
1144 	 */
1145 	hispgrp = p->p_pptr->p_pgrp;
1146 	if (hispgrp != pgrp && hispgrp->pg_session == mysession) {
1147 		if (entering) {
1148 			pgrp->pg_jobc++;
1149 			p->p_lflag &= ~PL_ORPHANPG;
1150 		} else if (--pgrp->pg_jobc == 0)
1151 			orphanpg(pgrp);
1152 	}
1153 
1154 	/*
1155 	 * Check this process' children to see whether they qualify
1156 	 * their process groups; if so, adjust counts for children's
1157 	 * process groups.
1158 	 */
1159 	LIST_FOREACH(child, &p->p_children, p_sibling) {
1160 		hispgrp = child->p_pgrp;
1161 		if (hispgrp != pgrp && hispgrp->pg_session == mysession &&
1162 		    !P_ZOMBIE(child)) {
1163 			if (entering) {
1164 				child->p_lflag &= ~PL_ORPHANPG;
1165 				hispgrp->pg_jobc++;
1166 			} else if (--hispgrp->pg_jobc == 0)
1167 				orphanpg(hispgrp);
1168 		}
1169 	}
1170 }
1171 
1172 /*
1173  * A process group has become orphaned;
1174  * if there are any stopped processes in the group,
1175  * hang-up all process in that group.
1176  *
1177  * Call with proc_lock held.
1178  */
1179 static void
1180 orphanpg(struct pgrp *pg)
1181 {
1182 	struct proc *p;
1183 
1184 	KASSERT(mutex_owned(proc_lock));
1185 
1186 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
1187 		if (p->p_stat == SSTOP) {
1188 			p->p_lflag |= PL_ORPHANPG;
1189 			psignal(p, SIGHUP);
1190 			psignal(p, SIGCONT);
1191 		}
1192 	}
1193 }
1194 
1195 #ifdef DDB
1196 #include <ddb/db_output.h>
1197 void pidtbl_dump(void);
1198 void
1199 pidtbl_dump(void)
1200 {
1201 	struct pid_table *pt;
1202 	struct proc *p;
1203 	struct pgrp *pgrp;
1204 	int id;
1205 
1206 	db_printf("pid table %p size %x, next %x, last %x\n",
1207 		pid_table, pid_tbl_mask+1,
1208 		next_free_pt, last_free_pt);
1209 	for (pt = pid_table, id = 0; id <= pid_tbl_mask; id++, pt++) {
1210 		p = pt->pt_proc;
1211 		if (!P_VALID(p) && !pt->pt_pgrp)
1212 			continue;
1213 		db_printf("  id %x: ", id);
1214 		if (P_VALID(p))
1215 			db_printf("slotpid %d proc %p id %d (0x%x) %s\n",
1216 				pt->pt_pid, p, p->p_pid, p->p_pid, p->p_comm);
1217 		else
1218 			db_printf("next %x use %x\n",
1219 				P_NEXT(p) & pid_tbl_mask,
1220 				P_NEXT(p) & ~pid_tbl_mask);
1221 		if ((pgrp = pt->pt_pgrp)) {
1222 			db_printf("\tsession %p, sid %d, count %d, login %s\n",
1223 			    pgrp->pg_session, pgrp->pg_session->s_sid,
1224 			    pgrp->pg_session->s_count,
1225 			    pgrp->pg_session->s_login);
1226 			db_printf("\tpgrp %p, pg_id %d, pg_jobc %d, members %p\n",
1227 			    pgrp, pgrp->pg_id, pgrp->pg_jobc,
1228 			    LIST_FIRST(&pgrp->pg_members));
1229 			LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
1230 				db_printf("\t\tpid %d addr %p pgrp %p %s\n",
1231 				    p->p_pid, p, p->p_pgrp, p->p_comm);
1232 			}
1233 		}
1234 	}
1235 }
1236 #endif /* DDB */
1237 
1238 #ifdef KSTACK_CHECK_MAGIC
1239 
1240 #define	KSTACK_MAGIC	0xdeadbeaf
1241 
1242 /* XXX should be per process basis? */
1243 static int	kstackleftmin = KSTACK_SIZE;
1244 static int	kstackleftthres = KSTACK_SIZE / 8;
1245 
1246 void
1247 kstack_setup_magic(const struct lwp *l)
1248 {
1249 	uint32_t *ip;
1250 	uint32_t const *end;
1251 
1252 	KASSERT(l != NULL);
1253 	KASSERT(l != &lwp0);
1254 
1255 	/*
1256 	 * fill all the stack with magic number
1257 	 * so that later modification on it can be detected.
1258 	 */
1259 	ip = (uint32_t *)KSTACK_LOWEST_ADDR(l);
1260 	end = (uint32_t *)((char *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE);
1261 	for (; ip < end; ip++) {
1262 		*ip = KSTACK_MAGIC;
1263 	}
1264 }
1265 
1266 void
1267 kstack_check_magic(const struct lwp *l)
1268 {
1269 	uint32_t const *ip, *end;
1270 	int stackleft;
1271 
1272 	KASSERT(l != NULL);
1273 
1274 	/* don't check proc0 */ /*XXX*/
1275 	if (l == &lwp0)
1276 		return;
1277 
1278 #ifdef __MACHINE_STACK_GROWS_UP
1279 	/* stack grows upwards (eg. hppa) */
1280 	ip = (uint32_t *)((void *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE);
1281 	end = (uint32_t *)KSTACK_LOWEST_ADDR(l);
1282 	for (ip--; ip >= end; ip--)
1283 		if (*ip != KSTACK_MAGIC)
1284 			break;
1285 
1286 	stackleft = (void *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE - (void *)ip;
1287 #else /* __MACHINE_STACK_GROWS_UP */
1288 	/* stack grows downwards (eg. i386) */
1289 	ip = (uint32_t *)KSTACK_LOWEST_ADDR(l);
1290 	end = (uint32_t *)((char *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE);
1291 	for (; ip < end; ip++)
1292 		if (*ip != KSTACK_MAGIC)
1293 			break;
1294 
1295 	stackleft = ((const char *)ip) - (const char *)KSTACK_LOWEST_ADDR(l);
1296 #endif /* __MACHINE_STACK_GROWS_UP */
1297 
1298 	if (kstackleftmin > stackleft) {
1299 		kstackleftmin = stackleft;
1300 		if (stackleft < kstackleftthres)
1301 			printf("warning: kernel stack left %d bytes"
1302 			    "(pid %u:lid %u)\n", stackleft,
1303 			    (u_int)l->l_proc->p_pid, (u_int)l->l_lid);
1304 	}
1305 
1306 	if (stackleft <= 0) {
1307 		panic("magic on the top of kernel stack changed for "
1308 		    "pid %u, lid %u: maybe kernel stack overflow",
1309 		    (u_int)l->l_proc->p_pid, (u_int)l->l_lid);
1310 	}
1311 }
1312 #endif /* KSTACK_CHECK_MAGIC */
1313 
1314 int
1315 proclist_foreach_call(struct proclist *list,
1316     int (*callback)(struct proc *, void *arg), void *arg)
1317 {
1318 	struct proc marker;
1319 	struct proc *p;
1320 	int ret = 0;
1321 
1322 	marker.p_flag = PK_MARKER;
1323 	mutex_enter(proc_lock);
1324 	for (p = LIST_FIRST(list); ret == 0 && p != NULL;) {
1325 		if (p->p_flag & PK_MARKER) {
1326 			p = LIST_NEXT(p, p_list);
1327 			continue;
1328 		}
1329 		LIST_INSERT_AFTER(p, &marker, p_list);
1330 		ret = (*callback)(p, arg);
1331 		KASSERT(mutex_owned(proc_lock));
1332 		p = LIST_NEXT(&marker, p_list);
1333 		LIST_REMOVE(&marker, p_list);
1334 	}
1335 	mutex_exit(proc_lock);
1336 
1337 	return ret;
1338 }
1339 
1340 int
1341 proc_vmspace_getref(struct proc *p, struct vmspace **vm)
1342 {
1343 
1344 	/* XXXCDC: how should locking work here? */
1345 
1346 	/* curproc exception is for coredump. */
1347 
1348 	if ((p != curproc && (p->p_sflag & PS_WEXIT) != 0) ||
1349 	    (p->p_vmspace->vm_refcnt < 1)) { /* XXX */
1350 		return EFAULT;
1351 	}
1352 
1353 	uvmspace_addref(p->p_vmspace);
1354 	*vm = p->p_vmspace;
1355 
1356 	return 0;
1357 }
1358 
1359 /*
1360  * Acquire a write lock on the process credential.
1361  */
1362 void
1363 proc_crmod_enter(void)
1364 {
1365 	struct lwp *l = curlwp;
1366 	struct proc *p = l->l_proc;
1367 	kauth_cred_t oc;
1368 
1369 	/* Reset what needs to be reset in plimit. */
1370 	if (p->p_limit->pl_corename != defcorename) {
1371 		lim_setcorename(p, defcorename, 0);
1372 	}
1373 
1374 	mutex_enter(p->p_lock);
1375 
1376 	/* Ensure the LWP cached credentials are up to date. */
1377 	if ((oc = l->l_cred) != p->p_cred) {
1378 		kauth_cred_hold(p->p_cred);
1379 		l->l_cred = p->p_cred;
1380 		kauth_cred_free(oc);
1381 	}
1382 }
1383 
1384 /*
1385  * Set in a new process credential, and drop the write lock.  The credential
1386  * must have a reference already.  Optionally, free a no-longer required
1387  * credential.  The scheduler also needs to inspect p_cred, so we also
1388  * briefly acquire the sched state mutex.
1389  */
1390 void
1391 proc_crmod_leave(kauth_cred_t scred, kauth_cred_t fcred, bool sugid)
1392 {
1393 	struct lwp *l = curlwp, *l2;
1394 	struct proc *p = l->l_proc;
1395 	kauth_cred_t oc;
1396 
1397 	KASSERT(mutex_owned(p->p_lock));
1398 
1399 	/* Is there a new credential to set in? */
1400 	if (scred != NULL) {
1401 		p->p_cred = scred;
1402 		LIST_FOREACH(l2, &p->p_lwps, l_sibling) {
1403 			if (l2 != l)
1404 				l2->l_prflag |= LPR_CRMOD;
1405 		}
1406 
1407 		/* Ensure the LWP cached credentials are up to date. */
1408 		if ((oc = l->l_cred) != scred) {
1409 			kauth_cred_hold(scred);
1410 			l->l_cred = scred;
1411 		}
1412 	} else
1413 		oc = NULL;	/* XXXgcc */
1414 
1415 	if (sugid) {
1416 		/*
1417 		 * Mark process as having changed credentials, stops
1418 		 * tracing etc.
1419 		 */
1420 		p->p_flag |= PK_SUGID;
1421 	}
1422 
1423 	mutex_exit(p->p_lock);
1424 
1425 	/* If there is a credential to be released, free it now. */
1426 	if (fcred != NULL) {
1427 		KASSERT(scred != NULL);
1428 		kauth_cred_free(fcred);
1429 		if (oc != scred)
1430 			kauth_cred_free(oc);
1431 	}
1432 }
1433 
1434 /*
1435  * proc_specific_key_create --
1436  *	Create a key for subsystem proc-specific data.
1437  */
1438 int
1439 proc_specific_key_create(specificdata_key_t *keyp, specificdata_dtor_t dtor)
1440 {
1441 
1442 	return (specificdata_key_create(proc_specificdata_domain, keyp, dtor));
1443 }
1444 
1445 /*
1446  * proc_specific_key_delete --
1447  *	Delete a key for subsystem proc-specific data.
1448  */
1449 void
1450 proc_specific_key_delete(specificdata_key_t key)
1451 {
1452 
1453 	specificdata_key_delete(proc_specificdata_domain, key);
1454 }
1455 
1456 /*
1457  * proc_initspecific --
1458  *	Initialize a proc's specificdata container.
1459  */
1460 void
1461 proc_initspecific(struct proc *p)
1462 {
1463 	int error __diagused;
1464 
1465 	error = specificdata_init(proc_specificdata_domain, &p->p_specdataref);
1466 	KASSERT(error == 0);
1467 }
1468 
1469 /*
1470  * proc_finispecific --
1471  *	Finalize a proc's specificdata container.
1472  */
1473 void
1474 proc_finispecific(struct proc *p)
1475 {
1476 
1477 	specificdata_fini(proc_specificdata_domain, &p->p_specdataref);
1478 }
1479 
1480 /*
1481  * proc_getspecific --
1482  *	Return proc-specific data corresponding to the specified key.
1483  */
1484 void *
1485 proc_getspecific(struct proc *p, specificdata_key_t key)
1486 {
1487 
1488 	return (specificdata_getspecific(proc_specificdata_domain,
1489 					 &p->p_specdataref, key));
1490 }
1491 
1492 /*
1493  * proc_setspecific --
1494  *	Set proc-specific data corresponding to the specified key.
1495  */
1496 void
1497 proc_setspecific(struct proc *p, specificdata_key_t key, void *data)
1498 {
1499 
1500 	specificdata_setspecific(proc_specificdata_domain,
1501 				 &p->p_specdataref, key, data);
1502 }
1503 
1504 int
1505 proc_uidmatch(kauth_cred_t cred, kauth_cred_t target)
1506 {
1507 	int r = 0;
1508 
1509 	if (kauth_cred_getuid(cred) != kauth_cred_getuid(target) ||
1510 	    kauth_cred_getuid(cred) != kauth_cred_getsvuid(target)) {
1511 		/*
1512 		 * suid proc of ours or proc not ours
1513 		 */
1514 		r = EPERM;
1515 	} else if (kauth_cred_getgid(target) != kauth_cred_getsvgid(target)) {
1516 		/*
1517 		 * sgid proc has sgid back to us temporarily
1518 		 */
1519 		r = EPERM;
1520 	} else {
1521 		/*
1522 		 * our rgid must be in target's group list (ie,
1523 		 * sub-processes started by a sgid process)
1524 		 */
1525 		int ismember = 0;
1526 
1527 		if (kauth_cred_ismember_gid(cred,
1528 		    kauth_cred_getgid(target), &ismember) != 0 ||
1529 		    !ismember)
1530 			r = EPERM;
1531 	}
1532 
1533 	return (r);
1534 }
1535 
1536 /*
1537  * sysctl stuff
1538  */
1539 
1540 #define KERN_PROCSLOP	(5 * sizeof(struct kinfo_proc))
1541 
1542 static const u_int sysctl_flagmap[] = {
1543 	PK_ADVLOCK, P_ADVLOCK,
1544 	PK_EXEC, P_EXEC,
1545 	PK_NOCLDWAIT, P_NOCLDWAIT,
1546 	PK_32, P_32,
1547 	PK_CLDSIGIGN, P_CLDSIGIGN,
1548 	PK_SUGID, P_SUGID,
1549 	0
1550 };
1551 
1552 static const u_int sysctl_sflagmap[] = {
1553 	PS_NOCLDSTOP, P_NOCLDSTOP,
1554 	PS_WEXIT, P_WEXIT,
1555 	PS_STOPFORK, P_STOPFORK,
1556 	PS_STOPEXEC, P_STOPEXEC,
1557 	PS_STOPEXIT, P_STOPEXIT,
1558 	0
1559 };
1560 
1561 static const u_int sysctl_slflagmap[] = {
1562 	PSL_TRACED, P_TRACED,
1563 	PSL_CHTRACED, P_CHTRACED,
1564 	PSL_SYSCALL, P_SYSCALL,
1565 	0
1566 };
1567 
1568 static const u_int sysctl_lflagmap[] = {
1569 	PL_CONTROLT, P_CONTROLT,
1570 	PL_PPWAIT, P_PPWAIT,
1571 	0
1572 };
1573 
1574 static const u_int sysctl_stflagmap[] = {
1575 	PST_PROFIL, P_PROFIL,
1576 	0
1577 
1578 };
1579 
1580 /* used by kern_lwp also */
1581 const u_int sysctl_lwpflagmap[] = {
1582 	LW_SINTR, L_SINTR,
1583 	LW_SYSTEM, L_SYSTEM,
1584 	0
1585 };
1586 
1587 /*
1588  * Find the most ``active'' lwp of a process and return it for ps display
1589  * purposes
1590  */
1591 static struct lwp *
1592 proc_active_lwp(struct proc *p)
1593 {
1594 	static const int ostat[] = {
1595 		0,
1596 		2,	/* LSIDL */
1597 		6,	/* LSRUN */
1598 		5,	/* LSSLEEP */
1599 		4,	/* LSSTOP */
1600 		0,	/* LSZOMB */
1601 		1,	/* LSDEAD */
1602 		7,	/* LSONPROC */
1603 		3	/* LSSUSPENDED */
1604 	};
1605 
1606 	struct lwp *l, *lp = NULL;
1607 	LIST_FOREACH(l, &p->p_lwps, l_sibling) {
1608 		KASSERT(l->l_stat >= 0 && l->l_stat < __arraycount(ostat));
1609 		if (lp == NULL ||
1610 		    ostat[l->l_stat] > ostat[lp->l_stat] ||
1611 		    (ostat[l->l_stat] == ostat[lp->l_stat] &&
1612 		    l->l_cpticks > lp->l_cpticks)) {
1613 			lp = l;
1614 			continue;
1615 		}
1616 	}
1617 	return lp;
1618 }
1619 
1620 static int
1621 sysctl_doeproc(SYSCTLFN_ARGS)
1622 {
1623 	union {
1624 		struct kinfo_proc kproc;
1625 		struct kinfo_proc2 kproc2;
1626 	} *kbuf;
1627 	struct proc *p, *next, *marker;
1628 	char *where, *dp;
1629 	int type, op, arg, error;
1630 	u_int elem_size, kelem_size, elem_count;
1631 	size_t buflen, needed;
1632 	bool match, zombie, mmmbrains;
1633 
1634 	if (namelen == 1 && name[0] == CTL_QUERY)
1635 		return (sysctl_query(SYSCTLFN_CALL(rnode)));
1636 
1637 	dp = where = oldp;
1638 	buflen = where != NULL ? *oldlenp : 0;
1639 	error = 0;
1640 	needed = 0;
1641 	type = rnode->sysctl_num;
1642 
1643 	if (type == KERN_PROC) {
1644 		if (namelen == 0)
1645 			return EINVAL;
1646 		switch (op = name[0]) {
1647 		case KERN_PROC_ALL:
1648 			if (namelen != 1)
1649 				return EINVAL;
1650 			arg = 0;
1651 			break;
1652 		default:
1653 			if (namelen != 2)
1654 				return EINVAL;
1655 			arg = name[1];
1656 			break;
1657 		}
1658 		elem_count = 0;	/* Ditto */
1659 		kelem_size = elem_size = sizeof(kbuf->kproc);
1660 	} else {
1661 		if (namelen != 4)
1662 			return EINVAL;
1663 		op = name[0];
1664 		arg = name[1];
1665 		elem_size = name[2];
1666 		elem_count = name[3];
1667 		kelem_size = sizeof(kbuf->kproc2);
1668 	}
1669 
1670 	sysctl_unlock();
1671 
1672 	kbuf = kmem_alloc(sizeof(*kbuf), KM_SLEEP);
1673 	marker = kmem_alloc(sizeof(*marker), KM_SLEEP);
1674 	marker->p_flag = PK_MARKER;
1675 
1676 	mutex_enter(proc_lock);
1677 	mmmbrains = false;
1678 	for (p = LIST_FIRST(&allproc);; p = next) {
1679 		if (p == NULL) {
1680 			if (!mmmbrains) {
1681 				p = LIST_FIRST(&zombproc);
1682 				mmmbrains = true;
1683 			}
1684 			if (p == NULL)
1685 				break;
1686 		}
1687 		next = LIST_NEXT(p, p_list);
1688 		if ((p->p_flag & PK_MARKER) != 0)
1689 			continue;
1690 
1691 		/*
1692 		 * Skip embryonic processes.
1693 		 */
1694 		if (p->p_stat == SIDL)
1695 			continue;
1696 
1697 		mutex_enter(p->p_lock);
1698 		error = kauth_authorize_process(l->l_cred,
1699 		    KAUTH_PROCESS_CANSEE, p,
1700 		    KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL);
1701 		if (error != 0) {
1702 			mutex_exit(p->p_lock);
1703 			continue;
1704 		}
1705 
1706 		/*
1707 		 * TODO - make more efficient (see notes below).
1708 		 * do by session.
1709 		 */
1710 		switch (op) {
1711 		case KERN_PROC_PID:
1712 			/* could do this with just a lookup */
1713 			match = (p->p_pid == (pid_t)arg);
1714 			break;
1715 
1716 		case KERN_PROC_PGRP:
1717 			/* could do this by traversing pgrp */
1718 			match = (p->p_pgrp->pg_id == (pid_t)arg);
1719 			break;
1720 
1721 		case KERN_PROC_SESSION:
1722 			match = (p->p_session->s_sid == (pid_t)arg);
1723 			break;
1724 
1725 		case KERN_PROC_TTY:
1726 			match = true;
1727 			if (arg == (int) KERN_PROC_TTY_REVOKE) {
1728 				if ((p->p_lflag & PL_CONTROLT) == 0 ||
1729 				    p->p_session->s_ttyp == NULL ||
1730 				    p->p_session->s_ttyvp != NULL) {
1731 				    	match = false;
1732 				}
1733 			} else if ((p->p_lflag & PL_CONTROLT) == 0 ||
1734 			    p->p_session->s_ttyp == NULL) {
1735 				if ((dev_t)arg != KERN_PROC_TTY_NODEV) {
1736 					match = false;
1737 				}
1738 			} else if (p->p_session->s_ttyp->t_dev != (dev_t)arg) {
1739 				match = false;
1740 			}
1741 			break;
1742 
1743 		case KERN_PROC_UID:
1744 			match = (kauth_cred_geteuid(p->p_cred) == (uid_t)arg);
1745 			break;
1746 
1747 		case KERN_PROC_RUID:
1748 			match = (kauth_cred_getuid(p->p_cred) == (uid_t)arg);
1749 			break;
1750 
1751 		case KERN_PROC_GID:
1752 			match = (kauth_cred_getegid(p->p_cred) == (uid_t)arg);
1753 			break;
1754 
1755 		case KERN_PROC_RGID:
1756 			match = (kauth_cred_getgid(p->p_cred) == (uid_t)arg);
1757 			break;
1758 
1759 		case KERN_PROC_ALL:
1760 			match = true;
1761 			/* allow everything */
1762 			break;
1763 
1764 		default:
1765 			error = EINVAL;
1766 			mutex_exit(p->p_lock);
1767 			goto cleanup;
1768 		}
1769 		if (!match) {
1770 			mutex_exit(p->p_lock);
1771 			continue;
1772 		}
1773 
1774 		/*
1775 		 * Grab a hold on the process.
1776 		 */
1777 		if (mmmbrains) {
1778 			zombie = true;
1779 		} else {
1780 			zombie = !rw_tryenter(&p->p_reflock, RW_READER);
1781 		}
1782 		if (zombie) {
1783 			LIST_INSERT_AFTER(p, marker, p_list);
1784 		}
1785 
1786 		if (buflen >= elem_size &&
1787 		    (type == KERN_PROC || elem_count > 0)) {
1788 			if (type == KERN_PROC) {
1789 				kbuf->kproc.kp_proc = *p;
1790 				fill_eproc(p, &kbuf->kproc.kp_eproc, zombie);
1791 			} else {
1792 				fill_kproc2(p, &kbuf->kproc2, zombie);
1793 				elem_count--;
1794 			}
1795 			mutex_exit(p->p_lock);
1796 			mutex_exit(proc_lock);
1797 			/*
1798 			 * Copy out elem_size, but not larger than kelem_size
1799 			 */
1800 			error = sysctl_copyout(l, kbuf, dp,
1801 			    min(kelem_size, elem_size));
1802 			mutex_enter(proc_lock);
1803 			if (error) {
1804 				goto bah;
1805 			}
1806 			dp += elem_size;
1807 			buflen -= elem_size;
1808 		} else {
1809 			mutex_exit(p->p_lock);
1810 		}
1811 		needed += elem_size;
1812 
1813 		/*
1814 		 * Release reference to process.
1815 		 */
1816 	 	if (zombie) {
1817 			next = LIST_NEXT(marker, p_list);
1818  			LIST_REMOVE(marker, p_list);
1819 		} else {
1820 			rw_exit(&p->p_reflock);
1821 			next = LIST_NEXT(p, p_list);
1822 		}
1823 	}
1824 	mutex_exit(proc_lock);
1825 
1826 	if (where != NULL) {
1827 		*oldlenp = dp - where;
1828 		if (needed > *oldlenp) {
1829 			error = ENOMEM;
1830 			goto out;
1831 		}
1832 	} else {
1833 		needed += KERN_PROCSLOP;
1834 		*oldlenp = needed;
1835 	}
1836 	if (kbuf)
1837 		kmem_free(kbuf, sizeof(*kbuf));
1838 	if (marker)
1839 		kmem_free(marker, sizeof(*marker));
1840 	sysctl_relock();
1841 	return 0;
1842  bah:
1843  	if (zombie)
1844  		LIST_REMOVE(marker, p_list);
1845 	else
1846 		rw_exit(&p->p_reflock);
1847  cleanup:
1848 	mutex_exit(proc_lock);
1849  out:
1850 	if (kbuf)
1851 		kmem_free(kbuf, sizeof(*kbuf));
1852 	if (marker)
1853 		kmem_free(marker, sizeof(*marker));
1854 	sysctl_relock();
1855 	return error;
1856 }
1857 
1858 int
1859 copyin_psstrings(struct proc *p, struct ps_strings *arginfo)
1860 {
1861 
1862 #ifdef COMPAT_NETBSD32
1863 	if (p->p_flag & PK_32) {
1864 		struct ps_strings32 arginfo32;
1865 
1866 		int error = copyin_proc(p, (void *)p->p_psstrp, &arginfo32,
1867 		    sizeof(arginfo32));
1868 		if (error)
1869 			return error;
1870 		arginfo->ps_argvstr = (void *)(uintptr_t)arginfo32.ps_argvstr;
1871 		arginfo->ps_nargvstr = arginfo32.ps_nargvstr;
1872 		arginfo->ps_envstr = (void *)(uintptr_t)arginfo32.ps_envstr;
1873 		arginfo->ps_nenvstr = arginfo32.ps_nenvstr;
1874 		return 0;
1875 	}
1876 #endif
1877 	return copyin_proc(p, (void *)p->p_psstrp, arginfo, sizeof(*arginfo));
1878 }
1879 
1880 static int
1881 copy_procargs_sysctl_cb(void *cookie_, const void *src, size_t off, size_t len)
1882 {
1883 	void **cookie = cookie_;
1884 	struct lwp *l = cookie[0];
1885 	char *dst = cookie[1];
1886 
1887 	return sysctl_copyout(l, src, dst + off, len);
1888 }
1889 
1890 /*
1891  * sysctl helper routine for kern.proc_args pseudo-subtree.
1892  */
1893 static int
1894 sysctl_kern_proc_args(SYSCTLFN_ARGS)
1895 {
1896 	struct ps_strings pss;
1897 	struct proc *p;
1898 	pid_t pid;
1899 	int type, error;
1900 	void *cookie[2];
1901 
1902 	if (namelen == 1 && name[0] == CTL_QUERY)
1903 		return (sysctl_query(SYSCTLFN_CALL(rnode)));
1904 
1905 	if (newp != NULL || namelen != 2)
1906 		return (EINVAL);
1907 	pid = name[0];
1908 	type = name[1];
1909 
1910 	switch (type) {
1911 	case KERN_PROC_PATHNAME:
1912 		sysctl_unlock();
1913 		error = fill_pathname(l, pid, oldp, oldlenp);
1914 		sysctl_relock();
1915 		return error;
1916 
1917 	case KERN_PROC_ARGV:
1918 	case KERN_PROC_NARGV:
1919 	case KERN_PROC_ENV:
1920 	case KERN_PROC_NENV:
1921 		/* ok */
1922 		break;
1923 	default:
1924 		return (EINVAL);
1925 	}
1926 
1927 	sysctl_unlock();
1928 
1929 	/* check pid */
1930 	mutex_enter(proc_lock);
1931 	if ((p = proc_find(pid)) == NULL) {
1932 		error = EINVAL;
1933 		goto out_locked;
1934 	}
1935 	mutex_enter(p->p_lock);
1936 
1937 	/* Check permission. */
1938 	if (type == KERN_PROC_ARGV || type == KERN_PROC_NARGV)
1939 		error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE,
1940 		    p, KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ARGS), NULL, NULL);
1941 	else if (type == KERN_PROC_ENV || type == KERN_PROC_NENV)
1942 		error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE,
1943 		    p, KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENV), NULL, NULL);
1944 	else
1945 		error = EINVAL; /* XXXGCC */
1946 	if (error) {
1947 		mutex_exit(p->p_lock);
1948 		goto out_locked;
1949 	}
1950 
1951 	if (oldp == NULL) {
1952 		if (type == KERN_PROC_NARGV || type == KERN_PROC_NENV)
1953 			*oldlenp = sizeof (int);
1954 		else
1955 			*oldlenp = ARG_MAX;	/* XXX XXX XXX */
1956 		error = 0;
1957 		mutex_exit(p->p_lock);
1958 		goto out_locked;
1959 	}
1960 
1961 	/*
1962 	 * Zombies don't have a stack, so we can't read their psstrings.
1963 	 * System processes also don't have a user stack.
1964 	 */
1965 	if (P_ZOMBIE(p) || (p->p_flag & PK_SYSTEM) != 0) {
1966 		error = EINVAL;
1967 		mutex_exit(p->p_lock);
1968 		goto out_locked;
1969 	}
1970 
1971 	error = rw_tryenter(&p->p_reflock, RW_READER) ? 0 : EBUSY;
1972 	mutex_exit(p->p_lock);
1973 	if (error) {
1974 		goto out_locked;
1975 	}
1976 	mutex_exit(proc_lock);
1977 
1978 	if (type == KERN_PROC_NARGV || type == KERN_PROC_NENV) {
1979 		int value;
1980 		if ((error = copyin_psstrings(p, &pss)) == 0) {
1981 			if (type == KERN_PROC_NARGV)
1982 				value = pss.ps_nargvstr;
1983 			else
1984 				value = pss.ps_nenvstr;
1985 			error = sysctl_copyout(l, &value, oldp, sizeof(value));
1986 			*oldlenp = sizeof(value);
1987 		}
1988 	} else {
1989 		cookie[0] = l;
1990 		cookie[1] = oldp;
1991 		error = copy_procargs(p, type, oldlenp,
1992 		    copy_procargs_sysctl_cb, cookie);
1993 	}
1994 	rw_exit(&p->p_reflock);
1995 	sysctl_relock();
1996 	return error;
1997 
1998 out_locked:
1999 	mutex_exit(proc_lock);
2000 	sysctl_relock();
2001 	return error;
2002 }
2003 
2004 int
2005 copy_procargs(struct proc *p, int oid, size_t *limit,
2006     int (*cb)(void *, const void *, size_t, size_t), void *cookie)
2007 {
2008 	struct ps_strings pss;
2009 	size_t len, i, loaded, entry_len;
2010 	struct uio auio;
2011 	struct iovec aiov;
2012 	int error, argvlen;
2013 	char *arg;
2014 	char **argv;
2015 	vaddr_t user_argv;
2016 	struct vmspace *vmspace;
2017 
2018 	/*
2019 	 * Allocate a temporary buffer to hold the argument vector and
2020 	 * the arguments themselve.
2021 	 */
2022 	arg = kmem_alloc(PAGE_SIZE, KM_SLEEP);
2023 	argv = kmem_alloc(PAGE_SIZE, KM_SLEEP);
2024 
2025 	/*
2026 	 * Lock the process down in memory.
2027 	 */
2028 	vmspace = p->p_vmspace;
2029 	uvmspace_addref(vmspace);
2030 
2031 	/*
2032 	 * Read in the ps_strings structure.
2033 	 */
2034 	if ((error = copyin_psstrings(p, &pss)) != 0)
2035 		goto done;
2036 
2037 	/*
2038 	 * Now read the address of the argument vector.
2039 	 */
2040 	switch (oid) {
2041 	case KERN_PROC_ARGV:
2042 		user_argv = (uintptr_t)pss.ps_argvstr;
2043 		argvlen = pss.ps_nargvstr;
2044 		break;
2045 	case KERN_PROC_ENV:
2046 		user_argv = (uintptr_t)pss.ps_envstr;
2047 		argvlen = pss.ps_nenvstr;
2048 		break;
2049 	default:
2050 		error = EINVAL;
2051 		goto done;
2052 	}
2053 
2054 	if (argvlen < 0) {
2055 		error = EIO;
2056 		goto done;
2057 	}
2058 
2059 
2060 	/*
2061 	 * Now copy each string.
2062 	 */
2063 	len = 0; /* bytes written to user buffer */
2064 	loaded = 0; /* bytes from argv already processed */
2065 	i = 0; /* To make compiler happy */
2066 	entry_len = PROC_PTRSZ(p);
2067 
2068 	for (; argvlen; --argvlen) {
2069 		int finished = 0;
2070 		vaddr_t base;
2071 		size_t xlen;
2072 		int j;
2073 
2074 		if (loaded == 0) {
2075 			size_t rem = entry_len * argvlen;
2076 			loaded = MIN(rem, PAGE_SIZE);
2077 			error = copyin_vmspace(vmspace,
2078 			    (const void *)user_argv, argv, loaded);
2079 			if (error)
2080 				break;
2081 			user_argv += loaded;
2082 			i = 0;
2083 		}
2084 
2085 #ifdef COMPAT_NETBSD32
2086 		if (p->p_flag & PK_32) {
2087 			netbsd32_charp *argv32;
2088 
2089 			argv32 = (netbsd32_charp *)argv;
2090 			base = (vaddr_t)NETBSD32PTR64(argv32[i++]);
2091 		} else
2092 #endif
2093 			base = (vaddr_t)argv[i++];
2094 		loaded -= entry_len;
2095 
2096 		/*
2097 		 * The program has messed around with its arguments,
2098 		 * possibly deleting some, and replacing them with
2099 		 * NULL's. Treat this as the last argument and not
2100 		 * a failure.
2101 		 */
2102 		if (base == 0)
2103 			break;
2104 
2105 		while (!finished) {
2106 			xlen = PAGE_SIZE - (base & PAGE_MASK);
2107 
2108 			aiov.iov_base = arg;
2109 			aiov.iov_len = PAGE_SIZE;
2110 			auio.uio_iov = &aiov;
2111 			auio.uio_iovcnt = 1;
2112 			auio.uio_offset = base;
2113 			auio.uio_resid = xlen;
2114 			auio.uio_rw = UIO_READ;
2115 			UIO_SETUP_SYSSPACE(&auio);
2116 			error = uvm_io(&vmspace->vm_map, &auio, 0);
2117 			if (error)
2118 				goto done;
2119 
2120 			/* Look for the end of the string */
2121 			for (j = 0; j < xlen; j++) {
2122 				if (arg[j] == '\0') {
2123 					xlen = j + 1;
2124 					finished = 1;
2125 					break;
2126 				}
2127 			}
2128 
2129 			/* Check for user buffer overflow */
2130 			if (len + xlen > *limit) {
2131 				finished = 1;
2132 				if (len > *limit)
2133 					xlen = 0;
2134 				else
2135 					xlen = *limit - len;
2136 			}
2137 
2138 			/* Copyout the page */
2139 			error = (*cb)(cookie, arg, len, xlen);
2140 			if (error)
2141 				goto done;
2142 
2143 			len += xlen;
2144 			base += xlen;
2145 		}
2146 	}
2147 	*limit = len;
2148 
2149 done:
2150 	kmem_free(argv, PAGE_SIZE);
2151 	kmem_free(arg, PAGE_SIZE);
2152 	uvmspace_free(vmspace);
2153 	return error;
2154 }
2155 
2156 /*
2157  * Fill in an eproc structure for the specified process.
2158  */
2159 void
2160 fill_eproc(struct proc *p, struct eproc *ep, bool zombie)
2161 {
2162 	struct tty *tp;
2163 	struct lwp *l;
2164 
2165 	KASSERT(mutex_owned(proc_lock));
2166 	KASSERT(mutex_owned(p->p_lock));
2167 
2168 	memset(ep, 0, sizeof(*ep));
2169 
2170 	ep->e_paddr = p;
2171 	ep->e_sess = p->p_session;
2172 	if (p->p_cred) {
2173 		kauth_cred_topcred(p->p_cred, &ep->e_pcred);
2174 		kauth_cred_toucred(p->p_cred, &ep->e_ucred);
2175 	}
2176 	if (p->p_stat != SIDL && !P_ZOMBIE(p) && !zombie) {
2177 		struct vmspace *vm = p->p_vmspace;
2178 
2179 		ep->e_vm.vm_rssize = vm_resident_count(vm);
2180 		ep->e_vm.vm_tsize = vm->vm_tsize;
2181 		ep->e_vm.vm_dsize = vm->vm_dsize;
2182 		ep->e_vm.vm_ssize = vm->vm_ssize;
2183 		ep->e_vm.vm_map.size = vm->vm_map.size;
2184 
2185 		/* Pick the primary (first) LWP */
2186 		l = proc_active_lwp(p);
2187 		KASSERT(l != NULL);
2188 		lwp_lock(l);
2189 		if (l->l_wchan)
2190 			strncpy(ep->e_wmesg, l->l_wmesg, WMESGLEN);
2191 		lwp_unlock(l);
2192 	}
2193 	ep->e_ppid = p->p_ppid;
2194 	if (p->p_pgrp && p->p_session) {
2195 		ep->e_pgid = p->p_pgrp->pg_id;
2196 		ep->e_jobc = p->p_pgrp->pg_jobc;
2197 		ep->e_sid = p->p_session->s_sid;
2198 		if ((p->p_lflag & PL_CONTROLT) &&
2199 		    (tp = ep->e_sess->s_ttyp)) {
2200 			ep->e_tdev = tp->t_dev;
2201 			ep->e_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PGID;
2202 			ep->e_tsess = tp->t_session;
2203 		} else
2204 			ep->e_tdev = (uint32_t)NODEV;
2205 		ep->e_flag = ep->e_sess->s_ttyvp ? EPROC_CTTY : 0;
2206 		if (SESS_LEADER(p))
2207 			ep->e_flag |= EPROC_SLEADER;
2208 		strncpy(ep->e_login, ep->e_sess->s_login, MAXLOGNAME);
2209 	}
2210 	ep->e_xsize = ep->e_xrssize = 0;
2211 	ep->e_xccount = ep->e_xswrss = 0;
2212 }
2213 
2214 /*
2215  * Fill in a kinfo_proc2 structure for the specified process.
2216  */
2217 void
2218 fill_kproc2(struct proc *p, struct kinfo_proc2 *ki, bool zombie)
2219 {
2220 	struct tty *tp;
2221 	struct lwp *l, *l2;
2222 	struct timeval ut, st, rt;
2223 	sigset_t ss1, ss2;
2224 	struct rusage ru;
2225 	struct vmspace *vm;
2226 
2227 	KASSERT(mutex_owned(proc_lock));
2228 	KASSERT(mutex_owned(p->p_lock));
2229 
2230 	sigemptyset(&ss1);
2231 	sigemptyset(&ss2);
2232 	memset(ki, 0, sizeof(*ki));
2233 
2234 	ki->p_paddr = PTRTOUINT64(p);
2235 	ki->p_fd = PTRTOUINT64(p->p_fd);
2236 	ki->p_cwdi = PTRTOUINT64(p->p_cwdi);
2237 	ki->p_stats = PTRTOUINT64(p->p_stats);
2238 	ki->p_limit = PTRTOUINT64(p->p_limit);
2239 	ki->p_vmspace = PTRTOUINT64(p->p_vmspace);
2240 	ki->p_sigacts = PTRTOUINT64(p->p_sigacts);
2241 	ki->p_sess = PTRTOUINT64(p->p_session);
2242 	ki->p_tsess = 0;	/* may be changed if controlling tty below */
2243 	ki->p_ru = PTRTOUINT64(&p->p_stats->p_ru);
2244 	ki->p_eflag = 0;
2245 	ki->p_exitsig = p->p_exitsig;
2246 	ki->p_flag = L_INMEM;   /* Process never swapped out */
2247 	ki->p_flag |= sysctl_map_flags(sysctl_flagmap, p->p_flag);
2248 	ki->p_flag |= sysctl_map_flags(sysctl_sflagmap, p->p_sflag);
2249 	ki->p_flag |= sysctl_map_flags(sysctl_slflagmap, p->p_slflag);
2250 	ki->p_flag |= sysctl_map_flags(sysctl_lflagmap, p->p_lflag);
2251 	ki->p_flag |= sysctl_map_flags(sysctl_stflagmap, p->p_stflag);
2252 	ki->p_pid = p->p_pid;
2253 	ki->p_ppid = p->p_ppid;
2254 	ki->p_uid = kauth_cred_geteuid(p->p_cred);
2255 	ki->p_ruid = kauth_cred_getuid(p->p_cred);
2256 	ki->p_gid = kauth_cred_getegid(p->p_cred);
2257 	ki->p_rgid = kauth_cred_getgid(p->p_cred);
2258 	ki->p_svuid = kauth_cred_getsvuid(p->p_cred);
2259 	ki->p_svgid = kauth_cred_getsvgid(p->p_cred);
2260 	ki->p_ngroups = kauth_cred_ngroups(p->p_cred);
2261 	kauth_cred_getgroups(p->p_cred, ki->p_groups,
2262 	    min(ki->p_ngroups, sizeof(ki->p_groups) / sizeof(ki->p_groups[0])),
2263 	    UIO_SYSSPACE);
2264 
2265 	ki->p_uticks = p->p_uticks;
2266 	ki->p_sticks = p->p_sticks;
2267 	ki->p_iticks = p->p_iticks;
2268 	ki->p_tpgid = NO_PGID;	/* may be changed if controlling tty below */
2269 	ki->p_tracep = PTRTOUINT64(p->p_tracep);
2270 	ki->p_traceflag = p->p_traceflag;
2271 
2272 	memcpy(&ki->p_sigignore, &p->p_sigctx.ps_sigignore,sizeof(ki_sigset_t));
2273 	memcpy(&ki->p_sigcatch, &p->p_sigctx.ps_sigcatch, sizeof(ki_sigset_t));
2274 
2275 	ki->p_cpticks = 0;
2276 	ki->p_pctcpu = p->p_pctcpu;
2277 	ki->p_estcpu = 0;
2278 	ki->p_stat = p->p_stat; /* Will likely be overridden by LWP status */
2279 	ki->p_realstat = p->p_stat;
2280 	ki->p_nice = p->p_nice;
2281 	ki->p_xstat = P_WAITSTATUS(p);
2282 	ki->p_acflag = p->p_acflag;
2283 
2284 	strncpy(ki->p_comm, p->p_comm,
2285 	    min(sizeof(ki->p_comm), sizeof(p->p_comm)));
2286 	strncpy(ki->p_ename, p->p_emul->e_name, sizeof(ki->p_ename));
2287 
2288 	ki->p_nlwps = p->p_nlwps;
2289 	ki->p_realflag = ki->p_flag;
2290 
2291 	if (p->p_stat != SIDL && !P_ZOMBIE(p) && !zombie) {
2292 		vm = p->p_vmspace;
2293 		ki->p_vm_rssize = vm_resident_count(vm);
2294 		ki->p_vm_tsize = vm->vm_tsize;
2295 		ki->p_vm_dsize = vm->vm_dsize;
2296 		ki->p_vm_ssize = vm->vm_ssize;
2297 		ki->p_vm_vsize = atop(vm->vm_map.size);
2298 		/*
2299 		 * Since the stack is initially mapped mostly with
2300 		 * PROT_NONE and grown as needed, adjust the "mapped size"
2301 		 * to skip the unused stack portion.
2302 		 */
2303 		ki->p_vm_msize =
2304 		    atop(vm->vm_map.size) - vm->vm_issize + vm->vm_ssize;
2305 
2306 		/* Pick the primary (first) LWP */
2307 		l = proc_active_lwp(p);
2308 		KASSERT(l != NULL);
2309 		lwp_lock(l);
2310 		ki->p_nrlwps = p->p_nrlwps;
2311 		ki->p_forw = 0;
2312 		ki->p_back = 0;
2313 		ki->p_addr = PTRTOUINT64(l->l_addr);
2314 		ki->p_stat = l->l_stat;
2315 		ki->p_flag |= sysctl_map_flags(sysctl_lwpflagmap, l->l_flag);
2316 		ki->p_swtime = l->l_swtime;
2317 		ki->p_slptime = l->l_slptime;
2318 		if (l->l_stat == LSONPROC)
2319 			ki->p_schedflags = l->l_cpu->ci_schedstate.spc_flags;
2320 		else
2321 			ki->p_schedflags = 0;
2322 		ki->p_priority = lwp_eprio(l);
2323 		ki->p_usrpri = l->l_priority;
2324 		if (l->l_wchan)
2325 			strncpy(ki->p_wmesg, l->l_wmesg, sizeof(ki->p_wmesg));
2326 		ki->p_wchan = PTRTOUINT64(l->l_wchan);
2327 		ki->p_cpuid = cpu_index(l->l_cpu);
2328 		lwp_unlock(l);
2329 		LIST_FOREACH(l, &p->p_lwps, l_sibling) {
2330 			/* This is hardly correct, but... */
2331 			sigplusset(&l->l_sigpend.sp_set, &ss1);
2332 			sigplusset(&l->l_sigmask, &ss2);
2333 			ki->p_cpticks += l->l_cpticks;
2334 			ki->p_pctcpu += l->l_pctcpu;
2335 			ki->p_estcpu += l->l_estcpu;
2336 		}
2337 	}
2338 	sigplusset(&p->p_sigpend.sp_set, &ss2);
2339 	memcpy(&ki->p_siglist, &ss1, sizeof(ki_sigset_t));
2340 	memcpy(&ki->p_sigmask, &ss2, sizeof(ki_sigset_t));
2341 
2342 	if (p->p_session != NULL) {
2343 		ki->p_sid = p->p_session->s_sid;
2344 		ki->p__pgid = p->p_pgrp->pg_id;
2345 		if (p->p_session->s_ttyvp)
2346 			ki->p_eflag |= EPROC_CTTY;
2347 		if (SESS_LEADER(p))
2348 			ki->p_eflag |= EPROC_SLEADER;
2349 		strncpy(ki->p_login, p->p_session->s_login,
2350 		    min(sizeof ki->p_login - 1, sizeof p->p_session->s_login));
2351 		ki->p_jobc = p->p_pgrp->pg_jobc;
2352 		if ((p->p_lflag & PL_CONTROLT) && (tp = p->p_session->s_ttyp)) {
2353 			ki->p_tdev = tp->t_dev;
2354 			ki->p_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PGID;
2355 			ki->p_tsess = PTRTOUINT64(tp->t_session);
2356 		} else {
2357 			ki->p_tdev = (int32_t)NODEV;
2358 		}
2359 	}
2360 
2361 	if (!P_ZOMBIE(p) && !zombie) {
2362 		ki->p_uvalid = 1;
2363 		ki->p_ustart_sec = p->p_stats->p_start.tv_sec;
2364 		ki->p_ustart_usec = p->p_stats->p_start.tv_usec;
2365 
2366 		calcru(p, &ut, &st, NULL, &rt);
2367 		ki->p_rtime_sec = rt.tv_sec;
2368 		ki->p_rtime_usec = rt.tv_usec;
2369 		ki->p_uutime_sec = ut.tv_sec;
2370 		ki->p_uutime_usec = ut.tv_usec;
2371 		ki->p_ustime_sec = st.tv_sec;
2372 		ki->p_ustime_usec = st.tv_usec;
2373 
2374 		memcpy(&ru, &p->p_stats->p_ru, sizeof(ru));
2375 		ki->p_uru_nvcsw = 0;
2376 		ki->p_uru_nivcsw = 0;
2377 		LIST_FOREACH(l2, &p->p_lwps, l_sibling) {
2378 			ki->p_uru_nvcsw += (l2->l_ncsw - l2->l_nivcsw);
2379 			ki->p_uru_nivcsw += l2->l_nivcsw;
2380 			ruadd(&ru, &l2->l_ru);
2381 		}
2382 		ki->p_uru_maxrss = ru.ru_maxrss;
2383 		ki->p_uru_ixrss = ru.ru_ixrss;
2384 		ki->p_uru_idrss = ru.ru_idrss;
2385 		ki->p_uru_isrss = ru.ru_isrss;
2386 		ki->p_uru_minflt = ru.ru_minflt;
2387 		ki->p_uru_majflt = ru.ru_majflt;
2388 		ki->p_uru_nswap = ru.ru_nswap;
2389 		ki->p_uru_inblock = ru.ru_inblock;
2390 		ki->p_uru_oublock = ru.ru_oublock;
2391 		ki->p_uru_msgsnd = ru.ru_msgsnd;
2392 		ki->p_uru_msgrcv = ru.ru_msgrcv;
2393 		ki->p_uru_nsignals = ru.ru_nsignals;
2394 
2395 		timeradd(&p->p_stats->p_cru.ru_utime,
2396 			 &p->p_stats->p_cru.ru_stime, &ut);
2397 		ki->p_uctime_sec = ut.tv_sec;
2398 		ki->p_uctime_usec = ut.tv_usec;
2399 	}
2400 }
2401 
2402 
2403 int
2404 proc_find_locked(struct lwp *l, struct proc **p, pid_t pid)
2405 {
2406 	int error;
2407 
2408 	mutex_enter(proc_lock);
2409 	if (pid == -1)
2410 		*p = l->l_proc;
2411 	else
2412 		*p = proc_find(pid);
2413 
2414 	if (*p == NULL) {
2415 		if (pid != -1)
2416 			mutex_exit(proc_lock);
2417 		return ESRCH;
2418 	}
2419 	if (pid != -1)
2420 		mutex_enter((*p)->p_lock);
2421 	mutex_exit(proc_lock);
2422 
2423 	error = kauth_authorize_process(l->l_cred,
2424 	    KAUTH_PROCESS_CANSEE, *p,
2425 	    KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL);
2426 	if (error) {
2427 		if (pid != -1)
2428 			mutex_exit((*p)->p_lock);
2429 	}
2430 	return error;
2431 }
2432 
2433 static int
2434 fill_pathname(struct lwp *l, pid_t pid, void *oldp, size_t *oldlenp)
2435 {
2436 	int error;
2437 	struct proc *p;
2438 
2439 	if ((error = proc_find_locked(l, &p, pid)) != 0)
2440 		return error;
2441 
2442 	if (p->p_path == NULL) {
2443 		if (pid != -1)
2444 			mutex_exit(p->p_lock);
2445 		return ENOENT;
2446 	}
2447 
2448 	size_t len = strlen(p->p_path) + 1;
2449 	if (oldp != NULL) {
2450 		error = sysctl_copyout(l, p->p_path, oldp, *oldlenp);
2451 		if (error == 0 && *oldlenp < len)
2452 			error = ENOSPC;
2453 	}
2454 	*oldlenp = len;
2455 	if (pid != -1)
2456 		mutex_exit(p->p_lock);
2457 	return error;
2458 }
2459 
2460 int
2461 proc_getauxv(struct proc *p, void **buf, size_t *len)
2462 {
2463 	struct ps_strings pss;
2464 	int error;
2465 	void *uauxv, *kauxv;
2466 	size_t size;
2467 
2468 	if ((error = copyin_psstrings(p, &pss)) != 0)
2469 		return error;
2470 	if (pss.ps_envstr == NULL)
2471 		return EIO;
2472 
2473 	size = p->p_execsw->es_arglen;
2474 	if (size == 0)
2475 		return EIO;
2476 
2477 	size_t ptrsz = PROC_PTRSZ(p);
2478 	uauxv = (void *)((char *)pss.ps_envstr + (pss.ps_nenvstr + 1) * ptrsz);
2479 
2480 	kauxv = kmem_alloc(size, KM_SLEEP);
2481 
2482 	error = copyin_proc(p, uauxv, kauxv, size);
2483 	if (error) {
2484 		kmem_free(kauxv, size);
2485 		return error;
2486 	}
2487 
2488 	*buf = kauxv;
2489 	*len = size;
2490 
2491 	return 0;
2492 }
2493