1/* $NetBSD: kern_proc.c,v 1.199 2016/11/14 08:55:51 kre 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 1.199 2016/11/14 08:55:51 kre Exp $");
66
67#ifdef _KERNEL_OPT
68#include "opt_kstack.h"
69#include "opt_maxuprc.h"
70#include "opt_dtrace.h"
71#include "opt_compat_netbsd32.h"
72#endif
73
74#include <sys/param.h>
75#include <sys/systm.h>
76#include <sys/kernel.h>
77#include <sys/proc.h>
78#include <sys/resourcevar.h>
79#include <sys/buf.h>
80#include <sys/acct.h>
81#include <sys/wait.h>
82#include <sys/file.h>
83#include <ufs/ufs/quota.h>
84#include <sys/uio.h>
85#include <sys/pool.h>
86#include <sys/pset.h>
87#include <sys/mbuf.h>
88#include <sys/ioctl.h>
89#include <sys/tty.h>
90#include <sys/signalvar.h>
91#include <sys/ras.h>
92#include <sys/filedesc.h>
93#include <sys/syscall_stats.h>
94#include <sys/kauth.h>
95#include <sys/sleepq.h>
96#include <sys/atomic.h>
97#include <sys/kmem.h>
98#include <sys/namei.h>
99#include <sys/dtrace_bsd.h>
100#include <sys/sysctl.h>
101#include <sys/exec.h>
102#include <sys/cpu.h>
103
104#include <uvm/uvm_extern.h>
105#include <uvm/uvm.h>
106
107#ifdef COMPAT_NETBSD32
108#include <compat/netbsd32/netbsd32.h>
109#endif
110
111/*
112 * Process lists.
113 */
114
115struct proclist allproc __cacheline_aligned;
116struct proclist zombproc __cacheline_aligned;
117
118kmutex_t * proc_lock __cacheline_aligned;
119
120/*
121 * pid to proc lookup is done by indexing the pid_table array.
122 * Since pid numbers are only allocated when an empty slot
123 * has been found, there is no need to search any lists ever.
124 * (an orphaned pgrp will lock the slot, a session will lock
125 * the pgrp with the same number.)
126 * If the table is too small it is reallocated with twice the
127 * previous size and the entries 'unzipped' into the two halves.
128 * A linked list of free entries is passed through the pt_proc
129 * field of 'free' items - set odd to be an invalid ptr.
130 */
131
132struct pid_table {
133 struct proc *pt_proc;
134 struct pgrp *pt_pgrp;
135 pid_t pt_pid;
136};
137#if 1 /* strongly typed cast - should be a noop */
138static inline uint p2u(struct proc *p) { return (uint)(uintptr_t)p; }
139#else
140#define p2u(p) ((uint)p)
141#endif
142#define P_VALID(p) (!(p2u(p) & 1))
143#define P_NEXT(p) (p2u(p) >> 1)
144#define P_FREE(pid) ((struct proc *)(uintptr_t)((pid) << 1 | 1))
145
146/*
147 * Table of process IDs (PIDs).
148 */
149static struct pid_table *pid_table __read_mostly;
150
151#define INITIAL_PID_TABLE_SIZE (1 << 5)
152
153/* Table mask, threshold for growing and number of allocated PIDs. */
154static u_int pid_tbl_mask __read_mostly;
155static u_int pid_alloc_lim __read_mostly;
156static u_int pid_alloc_cnt __cacheline_aligned;
157
158/* Next free, last free and maximum PIDs. */
159static u_int next_free_pt __cacheline_aligned;
160static u_int last_free_pt __cacheline_aligned;
161static pid_t pid_max __read_mostly;
162
163/* Components of the first process -- never freed. */
164
165extern struct emul emul_netbsd; /* defined in kern_exec.c */
166
167struct session session0 = {
168 .s_count = 1,
169 .s_sid = 0,
170};
171struct pgrp pgrp0 = {
172 .pg_members = LIST_HEAD_INITIALIZER(&pgrp0.pg_members),
173 .pg_session = &session0,
174};
175filedesc_t filedesc0;
176struct cwdinfo cwdi0 = {
177 .cwdi_cmask = CMASK,
178 .cwdi_refcnt = 1,
179};
180struct plimit limit0;
181struct pstats pstat0;
182struct vmspace vmspace0;
183struct sigacts sigacts0;
184struct proc proc0 = {
185 .p_lwps = LIST_HEAD_INITIALIZER(&proc0.p_lwps),
186 .p_sigwaiters = LIST_HEAD_INITIALIZER(&proc0.p_sigwaiters),
187 .p_nlwps = 1,
188 .p_nrlwps = 1,
189 .p_nlwpid = 1, /* must match lwp0.l_lid */
190 .p_pgrp = &pgrp0,
191 .p_comm = "system",
192 /*
193 * Set P_NOCLDWAIT so that kernel threads are reparented to init(8)
194 * when they exit. init(8) can easily wait them out for us.
195 */
196 .p_flag = PK_SYSTEM | PK_NOCLDWAIT,
197 .p_stat = SACTIVE,
198 .p_nice = NZERO,
199 .p_emul = &emul_netbsd,
200 .p_cwdi = &cwdi0,
201 .p_limit = &limit0,
202 .p_fd = &filedesc0,
203 .p_vmspace = &vmspace0,
204 .p_stats = &pstat0,
205 .p_sigacts = &sigacts0,
206#ifdef PROC0_MD_INITIALIZERS
207 PROC0_MD_INITIALIZERS
208#endif
209};
210kauth_cred_t cred0;
211
212static const int nofile = NOFILE;
213static const int maxuprc = MAXUPRC;
214
215static int sysctl_doeproc(SYSCTLFN_PROTO);
216static int sysctl_kern_proc_args(SYSCTLFN_PROTO);
217
218/*
219 * The process list descriptors, used during pid allocation and
220 * by sysctl. No locking on this data structure is needed since
221 * it is completely static.
222 */
223const struct proclist_desc proclists[] = {
224 { &allproc },
225 { &zombproc },
226 { NULL },
227};
228
229static struct pgrp * pg_remove(pid_t);
230static void pg_delete(pid_t);
231static void orphanpg(struct pgrp *);
232
233static specificdata_domain_t proc_specificdata_domain;
234
235static pool_cache_t proc_cache;
236
237static kauth_listener_t proc_listener;
238
239static int fill_pathname(struct lwp *, pid_t, void *, size_t *);
240
241static int
242proc_listener_cb(kauth_cred_t cred, kauth_action_t action, void *cookie,
243 void *arg0, void *arg1, void *arg2, void *arg3)
244{
245 struct proc *p;
246 int result;
247
248 result = KAUTH_RESULT_DEFER;
249 p = arg0;
250
251 switch (action) {
252 case KAUTH_PROCESS_CANSEE: {
253 enum kauth_process_req req;
254
255 req = (enum kauth_process_req)arg1;
256
257 switch (req) {
258 case KAUTH_REQ_PROCESS_CANSEE_ARGS:
259 case KAUTH_REQ_PROCESS_CANSEE_ENTRY:
260 case KAUTH_REQ_PROCESS_CANSEE_OPENFILES:
261 result = KAUTH_RESULT_ALLOW;
262
263 break;
264
265 case KAUTH_REQ_PROCESS_CANSEE_ENV:
266 if (kauth_cred_getuid(cred) !=
267 kauth_cred_getuid(p->p_cred) ||
268 kauth_cred_getuid(cred) !=
269 kauth_cred_getsvuid(p->p_cred))
270 break;
271
272 result = KAUTH_RESULT_ALLOW;
273
274 break;
275
276 default:
277 break;
278 }
279
280 break;
281 }
282
283 case KAUTH_PROCESS_FORK: {
284 int lnprocs = (int)(unsigned long)arg2;
285
286 /*
287 * Don't allow a nonprivileged user to use the last few
288 * processes. The variable lnprocs is the current number of
289 * processes, maxproc is the limit.
290 */
291 if (__predict_false((lnprocs >= maxproc - 5)))
292 break;
293
294 result = KAUTH_RESULT_ALLOW;
295
296 break;
297 }
298
299 case KAUTH_PROCESS_CORENAME:
300 case KAUTH_PROCESS_STOPFLAG:
301 if (proc_uidmatch(cred, p->p_cred) == 0)
302 result = KAUTH_RESULT_ALLOW;
303
304 break;
305
306 default:
307 break;
308 }
309
310 return result;
311}
312
313/*
314 * Initialize global process hashing structures.
315 */
316void
317procinit(void)
318{
319 const struct proclist_desc *pd;
320 u_int i;
321#define LINK_EMPTY ((PID_MAX + INITIAL_PID_TABLE_SIZE) & ~(INITIAL_PID_TABLE_SIZE - 1))
322
323 for (pd = proclists; pd->pd_list != NULL; pd++)
324 LIST_INIT(pd->pd_list);
325
326 proc_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
327 pid_table = kmem_alloc(INITIAL_PID_TABLE_SIZE
328 * sizeof(struct pid_table), KM_SLEEP);
329 pid_tbl_mask = INITIAL_PID_TABLE_SIZE - 1;
330 pid_max = PID_MAX;
331
332 /* Set free list running through table...
333 Preset 'use count' above PID_MAX so we allocate pid 1 next. */
334 for (i = 0; i <= pid_tbl_mask; i++) {
335 pid_table[i].pt_proc = P_FREE(LINK_EMPTY + i + 1);
336 pid_table[i].pt_pgrp = 0;
337 pid_table[i].pt_pid = 0;
338 }
339 /* slot 0 is just grabbed */
340 next_free_pt = 1;
341 /* Need to fix last entry. */
342 last_free_pt = pid_tbl_mask;
343 pid_table[last_free_pt].pt_proc = P_FREE(LINK_EMPTY);
344 /* point at which we grow table - to avoid reusing pids too often */
345 pid_alloc_lim = pid_tbl_mask - 1;
346#undef LINK_EMPTY
347
348 proc_specificdata_domain = specificdata_domain_create();
349 KASSERT(proc_specificdata_domain != NULL);
350
351 proc_cache = pool_cache_init(sizeof(struct proc), 0, 0, 0,
352 "procpl", NULL, IPL_NONE, NULL, NULL, NULL);
353
354 proc_listener = kauth_listen_scope(KAUTH_SCOPE_PROCESS,
355 proc_listener_cb, NULL);
356}
357
358void
359procinit_sysctl(void)
360{
361 static struct sysctllog *clog;
362
363 sysctl_createv(&clog, 0, NULL, NULL,
364 CTLFLAG_PERMANENT,
365 CTLTYPE_NODE, "proc",
366 SYSCTL_DESCR("System-wide process information"),
367 sysctl_doeproc, 0, NULL, 0,
368 CTL_KERN, KERN_PROC, CTL_EOL);
369 sysctl_createv(&clog, 0, NULL, NULL,
370 CTLFLAG_PERMANENT,
371 CTLTYPE_NODE, "proc2",
372 SYSCTL_DESCR("Machine-independent process information"),
373 sysctl_doeproc, 0, NULL, 0,
374 CTL_KERN, KERN_PROC2, CTL_EOL);
375 sysctl_createv(&clog, 0, NULL, NULL,
376 CTLFLAG_PERMANENT,
377 CTLTYPE_NODE, "proc_args",
378 SYSCTL_DESCR("Process argument information"),
379 sysctl_kern_proc_args, 0, NULL, 0,
380 CTL_KERN, KERN_PROC_ARGS, CTL_EOL);
381
382 /*
383 "nodes" under these:
384
385 KERN_PROC_ALL
386 KERN_PROC_PID pid
387 KERN_PROC_PGRP pgrp
388 KERN_PROC_SESSION sess
389 KERN_PROC_TTY tty
390 KERN_PROC_UID uid
391 KERN_PROC_RUID uid
392 KERN_PROC_GID gid
393 KERN_PROC_RGID gid
394
395 all in all, probably not worth the effort...
396 */
397}
398
399/*
400 * Initialize process 0.
401 */
402void
403proc0_init(void)
404{
405 struct proc *p;
406 struct pgrp *pg;
407 struct rlimit *rlim;
408 rlim_t lim;
409 int i;
410
411 p = &proc0;
412 pg = &pgrp0;
413
414 mutex_init(&p->p_stmutex, MUTEX_DEFAULT, IPL_HIGH);
415 mutex_init(&p->p_auxlock, MUTEX_DEFAULT, IPL_NONE);
416 p->p_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
417
418 rw_init(&p->p_reflock);
419 cv_init(&p->p_waitcv, "wait");
420 cv_init(&p->p_lwpcv, "lwpwait");
421
422 LIST_INSERT_HEAD(&p->p_lwps, &lwp0, l_sibling);
423
424 pid_table[0].pt_proc = p;
425 LIST_INSERT_HEAD(&allproc, p, p_list);
426
427 pid_table[0].pt_pgrp = pg;
428 LIST_INSERT_HEAD(&pg->pg_members, p, p_pglist);
429
430#ifdef __HAVE_SYSCALL_INTERN
431 (*p->p_emul->e_syscall_intern)(p);
432#endif
433
434 /* Create credentials. */
435 cred0 = kauth_cred_alloc();
436 p->p_cred = cred0;
437
438 /* Create the CWD info. */
439 rw_init(&cwdi0.cwdi_lock);
440
441 /* Create the limits structures. */
442 mutex_init(&limit0.pl_lock, MUTEX_DEFAULT, IPL_NONE);
443
444 rlim = limit0.pl_rlimit;
445 for (i = 0; i < __arraycount(limit0.pl_rlimit); i++) {
446 rlim[i].rlim_cur = RLIM_INFINITY;
447 rlim[i].rlim_max = RLIM_INFINITY;
448 }
449
450 rlim[RLIMIT_NOFILE].rlim_max = maxfiles;
451 rlim[RLIMIT_NOFILE].rlim_cur = maxfiles < nofile ? maxfiles : nofile;
452
453 rlim[RLIMIT_NPROC].rlim_max = maxproc;
454 rlim[RLIMIT_NPROC].rlim_cur = maxproc < maxuprc ? maxproc : maxuprc;
455
456 lim = MIN(VM_MAXUSER_ADDRESS, ctob((rlim_t)uvmexp.free));
457 rlim[RLIMIT_RSS].rlim_max = lim;
458 rlim[RLIMIT_MEMLOCK].rlim_max = lim;
459 rlim[RLIMIT_MEMLOCK].rlim_cur = lim / 3;
460
461 rlim[RLIMIT_NTHR].rlim_max = maxlwp;
462 rlim[RLIMIT_NTHR].rlim_cur = maxlwp < maxuprc ? maxlwp : maxuprc;
463
464 /* Note that default core name has zero length. */
465 limit0.pl_corename = defcorename;
466 limit0.pl_cnlen = 0;
467 limit0.pl_refcnt = 1;
468 limit0.pl_writeable = false;
469 limit0.pl_sv_limit = NULL;
470
471 /* Configure virtual memory system, set vm rlimits. */
472 uvm_init_limits(p);
473
474 /* Initialize file descriptor table for proc0. */
475 fd_init(&filedesc0);
476
477 /*
478 * Initialize proc0's vmspace, which uses the kernel pmap.
479 * All kernel processes (which never have user space mappings)
480 * share proc0's vmspace, and thus, the kernel pmap.
481 */
482 uvmspace_init(&vmspace0, pmap_kernel(), round_page(VM_MIN_ADDRESS),
483 trunc_page(VM_MAXUSER_ADDRESS),
484#ifdef __USE_TOPDOWN_VM
485 true
486#else
487 false
488#endif
489 );
490
491 /* Initialize signal state for proc0. XXX IPL_SCHED */
492 mutex_init(&p->p_sigacts->sa_mutex, MUTEX_DEFAULT, IPL_SCHED);
493 siginit(p);
494
495 proc_initspecific(p);
496 kdtrace_proc_ctor(NULL, p);
497}
498
499/*
500 * Session reference counting.
501 */
502
503void
504proc_sesshold(struct session *ss)
505{
506
507 KASSERT(mutex_owned(proc_lock));
508 ss->s_count++;
509}
510
511void
512proc_sessrele(struct session *ss)
513{
514
515 KASSERT(mutex_owned(proc_lock));
516 /*
517 * We keep the pgrp with the same id as the session in order to
518 * stop a process being given the same pid. Since the pgrp holds
519 * a reference to the session, it must be a 'zombie' pgrp by now.
520 */
521 if (--ss->s_count == 0) {
522 struct pgrp *pg;
523
524 pg = pg_remove(ss->s_sid);
525 mutex_exit(proc_lock);
526
527 kmem_free(pg, sizeof(struct pgrp));
528 kmem_free(ss, sizeof(struct session));
529 } else {
530 mutex_exit(proc_lock);
531 }
532}
533
534/*
535 * Check that the specified process group is in the session of the
536 * specified process.
537 * Treats -ve ids as process ids.
538 * Used to validate TIOCSPGRP requests.
539 */
540int
541pgid_in_session(struct proc *p, pid_t pg_id)
542{
543 struct pgrp *pgrp;
544 struct session *session;
545 int error;
546
547 mutex_enter(proc_lock);
548 if (pg_id < 0) {
549 struct proc *p1 = proc_find(-pg_id);
550 if (p1 == NULL) {
551 error = EINVAL;
552 goto fail;
553 }
554 pgrp = p1->p_pgrp;
555 } else {
556 pgrp = pgrp_find(pg_id);
557 if (pgrp == NULL) {
558 error = EINVAL;
559 goto fail;
560 }
561 }
562 session = pgrp->pg_session;
563 error = (session != p->p_pgrp->pg_session) ? EPERM : 0;
564fail:
565 mutex_exit(proc_lock);
566 return error;
567}
568
569/*
570 * p_inferior: is p an inferior of q?
571 */
572static inline bool
573p_inferior(struct proc *p, struct proc *q)
574{
575
576 KASSERT(mutex_owned(proc_lock));
577
578 for (; p != q; p = p->p_pptr)
579 if (p->p_pid == 0)
580 return false;
581 return true;
582}
583
584/*
585 * proc_find: locate a process by the ID.
586 *
587 * => Must be called with proc_lock held.
588 */
589proc_t *
590proc_find_raw(pid_t pid)
591{
592 struct pid_table *pt;
593 proc_t *p;
594
595 KASSERT(mutex_owned(proc_lock));
596 pt = &pid_table[pid & pid_tbl_mask];
597 p = pt->pt_proc;
598 if (__predict_false(!P_VALID(p) || pt->pt_pid != pid)) {
599 return NULL;
600 }
601 return p;
602}
603
604proc_t *
605proc_find(pid_t pid)
606{
607 proc_t *p;
608
609 p = proc_find_raw(pid);
610 if (__predict_false(p == NULL)) {
611 return NULL;
612 }
613
614 /*
615 * Only allow live processes to be found by PID.
616 * XXX: p_stat might change, since unlocked.
617 */
618 if (__predict_true(p->p_stat == SACTIVE || p->p_stat == SSTOP)) {
619 return p;
620 }
621 return NULL;
622}
623
624/*
625 * pgrp_find: locate a process group by the ID.
626 *
627 * => Must be called with proc_lock held.
628 */
629struct pgrp *
630pgrp_find(pid_t pgid)
631{
632 struct pgrp *pg;
633
634 KASSERT(mutex_owned(proc_lock));
635
636 pg = pid_table[pgid & pid_tbl_mask].pt_pgrp;
637
638 /*
639 * Cannot look up a process group that only exists because the
640 * session has not died yet (traditional).
641 */
642 if (pg == NULL || pg->pg_id != pgid || LIST_EMPTY(&pg->pg_members)) {
643 return NULL;
644 }
645 return pg;
646}
647
648static void
649expand_pid_table(void)
650{
651 size_t pt_size, tsz;
652 struct pid_table *n_pt, *new_pt;
653 struct proc *proc;
654 struct pgrp *pgrp;
655 pid_t pid, rpid;
656 u_int i;
657 uint new_pt_mask;
658
659 pt_size = pid_tbl_mask + 1;
660 tsz = pt_size * 2 * sizeof(struct pid_table);
661 new_pt = kmem_alloc(tsz, KM_SLEEP);
662 new_pt_mask = pt_size * 2 - 1;
663
664 mutex_enter(proc_lock);
665 if (pt_size != pid_tbl_mask + 1) {
666 /* Another process beat us to it... */
667 mutex_exit(proc_lock);
668 kmem_free(new_pt, tsz);
669 return;
670 }
671
672 /*
673 * Copy entries from old table into new one.
674 * If 'pid' is 'odd' we need to place in the upper half,
675 * even pid's to the lower half.
676 * Free items stay in the low half so we don't have to
677 * fixup the reference to them.
678 * We stuff free items on the front of the freelist
679 * because we can't write to unmodified entries.
680 * Processing the table backwards maintains a semblance
681 * of issuing pid numbers that increase with time.
682 */
683 i = pt_size - 1;
684 n_pt = new_pt + i;
685 for (; ; i--, n_pt--) {
686 proc = pid_table[i].pt_proc;
687 pgrp = pid_table[i].pt_pgrp;
688 if (!P_VALID(proc)) {
689 /* Up 'use count' so that link is valid */
690 pid = (P_NEXT(proc) + pt_size) & ~pt_size;
691 rpid = 0;
692 proc = P_FREE(pid);
693 if (pgrp)
694 pid = pgrp->pg_id;
695 } else {
696 pid = pid_table[i].pt_pid;
697 rpid = pid;
698 }
699
700 /* Save entry in appropriate half of table */
701 n_pt[pid & pt_size].pt_proc = proc;
702 n_pt[pid & pt_size].pt_pgrp = pgrp;
703 n_pt[pid & pt_size].pt_pid = rpid;
704
705 /* Put other piece on start of free list */
706 pid = (pid ^ pt_size) & ~pid_tbl_mask;
707 n_pt[pid & pt_size].pt_proc =
708 P_FREE((pid & ~pt_size) | next_free_pt);
709 n_pt[pid & pt_size].pt_pgrp = 0;
710 n_pt[pid & pt_size].pt_pid = 0;
711
712 next_free_pt = i | (pid & pt_size);
713 if (i == 0)
714 break;
715 }
716
717 /* Save old table size and switch tables */
718 tsz = pt_size * sizeof(struct pid_table);
719 n_pt = pid_table;
720 pid_table = new_pt;
721 pid_tbl_mask = new_pt_mask;
722
723 /*
724 * pid_max starts as PID_MAX (= 30000), once we have 16384
725 * allocated pids we need it to be larger!
726 */
727 if (pid_tbl_mask > PID_MAX) {
728 pid_max = pid_tbl_mask * 2 + 1;
729 pid_alloc_lim |= pid_alloc_lim << 1;
730 } else
731 pid_alloc_lim <<= 1; /* doubles number of free slots... */
732
733 mutex_exit(proc_lock);
734 kmem_free(n_pt, tsz);
735}
736
737struct proc *
738proc_alloc(void)
739{
740 struct proc *p;
741
742 p = pool_cache_get(proc_cache, PR_WAITOK);
743 p->p_stat = SIDL; /* protect against others */
744 proc_initspecific(p);
745 kdtrace_proc_ctor(NULL, p);
746 p->p_pid = -1;
747 proc_alloc_pid(p);
748 return p;
749}
750
751/*
752 * proc_alloc_pid: allocate PID and record the given proc 'p' so that
753 * proc_find_raw() can find it by the PID.
754 */
755
756pid_t
757proc_alloc_pid(struct proc *p)
758{
759 struct pid_table *pt;
760 pid_t pid;
761 int nxt;
762
763 for (;;expand_pid_table()) {
764 if (__predict_false(pid_alloc_cnt >= pid_alloc_lim))
765 /* ensure pids cycle through 2000+ values */
766 continue;
767 mutex_enter(proc_lock);
768 pt = &pid_table[next_free_pt];
769#ifdef DIAGNOSTIC
770 if (__predict_false(P_VALID(pt->pt_proc) || pt->pt_pgrp))
771 panic("proc_alloc: slot busy");
772#endif
773 nxt = P_NEXT(pt->pt_proc);
774 if (nxt & pid_tbl_mask)
775 break;
776 /* Table full - expand (NB last entry not used....) */
777 mutex_exit(proc_lock);
778 }
779
780 /* pid is 'saved use count' + 'size' + entry */
781 pid = (nxt & ~pid_tbl_mask) + pid_tbl_mask + 1 + next_free_pt;
782 if ((uint)pid > (uint)pid_max)
783 pid &= pid_tbl_mask;
784 next_free_pt = nxt & pid_tbl_mask;
785
786 /* Grab table slot */
787 pt->pt_proc = p;
788
789 KASSERT(pt->pt_pid == 0);
790 pt->pt_pid = pid;
791 if (p->p_pid == -1) {
792 p->p_pid = pid;
793 }
794 pid_alloc_cnt++;
795 mutex_exit(proc_lock);
796
797 return pid;
798}
799
800/*
801 * Free a process id - called from proc_free (in kern_exit.c)
802 *
803 * Called with the proc_lock held.
804 */
805void
806proc_free_pid(pid_t pid)
807{
808 struct pid_table *pt;
809
810 KASSERT(mutex_owned(proc_lock));
811
812 pt = &pid_table[pid & pid_tbl_mask];
813
814 /* save pid use count in slot */
815 pt->pt_proc = P_FREE(pid & ~pid_tbl_mask);
816 KASSERT(pt->pt_pid == pid);
817 pt->pt_pid = 0;
818
819 if (pt->pt_pgrp == NULL) {
820 /* link last freed entry onto ours */
821 pid &= pid_tbl_mask;
822 pt = &pid_table[last_free_pt];
823 pt->pt_proc = P_FREE(P_NEXT(pt->pt_proc) | pid);
824 pt->pt_pid = 0;
825 last_free_pt = pid;
826 pid_alloc_cnt--;
827 }
828
829 atomic_dec_uint(&nprocs);
830}
831
832void
833proc_free_mem(struct proc *p)
834{
835
836 kdtrace_proc_dtor(NULL, p);
837 pool_cache_put(proc_cache, p);
838}
839
840/*
841 * proc_enterpgrp: move p to a new or existing process group (and session).
842 *
843 * If we are creating a new pgrp, the pgid should equal
844 * the calling process' pid.
845 * If is only valid to enter a process group that is in the session
846 * of the process.
847 * Also mksess should only be set if we are creating a process group
848 *
849 * Only called from sys_setsid, sys_setpgid and posix_spawn/spawn_return.
850 */
851int
852proc_enterpgrp(struct proc *curp, pid_t pid, pid_t pgid, bool mksess)
853{
854 struct pgrp *new_pgrp, *pgrp;
855 struct session *sess;
856 struct proc *p;
857 int rval;
858 pid_t pg_id = NO_PGID;
859
860 sess = mksess ? kmem_alloc(sizeof(*sess), KM_SLEEP) : NULL;
861
862 /* Allocate data areas we might need before doing any validity checks */
863 mutex_enter(proc_lock); /* Because pid_table might change */
864 if (pid_table[pgid & pid_tbl_mask].pt_pgrp == 0) {
865 mutex_exit(proc_lock);
866 new_pgrp = kmem_alloc(sizeof(*new_pgrp), KM_SLEEP);
867 mutex_enter(proc_lock);
868 } else
869 new_pgrp = NULL;
870 rval = EPERM; /* most common error (to save typing) */
871
872 /* Check pgrp exists or can be created */
873 pgrp = pid_table[pgid & pid_tbl_mask].pt_pgrp;
874 if (pgrp != NULL && pgrp->pg_id != pgid)
875 goto done;
876
877 /* Can only set another process under restricted circumstances. */
878 if (pid != curp->p_pid) {
879 /* Must exist and be one of our children... */
880 p = proc_find(pid);
881 if (p == NULL || !p_inferior(p, curp)) {
882 rval = ESRCH;
883 goto done;
884 }
885 /* ... in the same session... */
886 if (sess != NULL || p->p_session != curp->p_session)
887 goto done;
888 /* ... existing pgid must be in same session ... */
889 if (pgrp != NULL && pgrp->pg_session != p->p_session)
890 goto done;
891 /* ... and not done an exec. */
892 if (p->p_flag & PK_EXEC) {
893 rval = EACCES;
894 goto done;
895 }
896 } else {
897 /* ... setsid() cannot re-enter a pgrp */
898 if (mksess && (curp->p_pgid == curp->p_pid ||
899 pgrp_find(curp->p_pid)))
900 goto done;
901 p = curp;
902 }
903
904 /* Changing the process group/session of a session
905 leader is definitely off limits. */
906 if (SESS_LEADER(p)) {
907 if (sess == NULL && p->p_pgrp == pgrp)
908 /* unless it's a definite noop */
909 rval = 0;
910 goto done;
911 }
912
913 /* Can only create a process group with id of process */
914 if (pgrp == NULL && pgid != pid)
915 goto done;
916
917 /* Can only create a session if creating pgrp */
918 if (sess != NULL && pgrp != NULL)
919 goto done;
920
921 /* Check we allocated memory for a pgrp... */
922 if (pgrp == NULL && new_pgrp == NULL)
923 goto done;
924
925 /* Don't attach to 'zombie' pgrp */
926 if (pgrp != NULL && LIST_EMPTY(&pgrp->pg_members))
927 goto done;
928
929 /* Expect to succeed now */
930 rval = 0;
931
932 if (pgrp == p->p_pgrp)
933 /* nothing to do */
934 goto done;
935
936 /* Ok all setup, link up required structures */
937
938 if (pgrp == NULL) {
939 pgrp = new_pgrp;
940 new_pgrp = NULL;
941 if (sess != NULL) {
942 sess->s_sid = p->p_pid;
943 sess->s_leader = p;
944 sess->s_count = 1;
945 sess->s_ttyvp = NULL;
946 sess->s_ttyp = NULL;
947 sess->s_flags = p->p_session->s_flags & ~S_LOGIN_SET;
948 memcpy(sess->s_login, p->p_session->s_login,
949 sizeof(sess->s_login));
950 p->p_lflag &= ~PL_CONTROLT;
951 } else {
952 sess = p->p_pgrp->pg_session;
953 proc_sesshold(sess);
954 }
955 pgrp->pg_session = sess;
956 sess = NULL;
957
958 pgrp->pg_id = pgid;
959 LIST_INIT(&pgrp->pg_members);
960#ifdef DIAGNOSTIC
961 if (__predict_false(pid_table[pgid & pid_tbl_mask].pt_pgrp))
962 panic("enterpgrp: pgrp table slot in use");
963 if (__predict_false(mksess && p != curp))
964 panic("enterpgrp: mksession and p != curproc");
965#endif
966 pid_table[pgid & pid_tbl_mask].pt_pgrp = pgrp;
967 pgrp->pg_jobc = 0;
968 }
969
970 /*
971 * Adjust eligibility of affected pgrps to participate in job control.
972 * Increment eligibility counts before decrementing, otherwise we
973 * could reach 0 spuriously during the first call.
974 */
975 fixjobc(p, pgrp, 1);
976 fixjobc(p, p->p_pgrp, 0);
977
978 /* Interlock with ttread(). */
979 mutex_spin_enter(&tty_lock);
980
981 /* Move process to requested group. */
982 LIST_REMOVE(p, p_pglist);
983 if (LIST_EMPTY(&p->p_pgrp->pg_members))
984 /* defer delete until we've dumped the lock */
985 pg_id = p->p_pgrp->pg_id;
986 p->p_pgrp = pgrp;
987 LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
988
989 /* Done with the swap; we can release the tty mutex. */
990 mutex_spin_exit(&tty_lock);
991
992 done:
993 if (pg_id != NO_PGID) {
994 /* Releases proc_lock. */
995 pg_delete(pg_id);
996 } else {
997 mutex_exit(proc_lock);
998 }
999 if (sess != NULL)
1000 kmem_free(sess, sizeof(*sess));
1001 if (new_pgrp != NULL)
1002 kmem_free(new_pgrp, sizeof(*new_pgrp));
1003#ifdef DEBUG_PGRP
1004 if (__predict_false(rval))
1005 printf("enterpgrp(%d,%d,%d), curproc %d, rval %d\n",
1006 pid, pgid, mksess, curp->p_pid, rval);
1007#endif
1008 return rval;
1009}
1010
1011/*
1012 * proc_leavepgrp: remove a process from its process group.
1013 * => must be called with the proc_lock held, which will be released;
1014 */
1015void
1016proc_leavepgrp(struct proc *p)
1017{
1018 struct pgrp *pgrp;
1019
1020 KASSERT(mutex_owned(proc_lock));
1021
1022 /* Interlock with ttread() */
1023 mutex_spin_enter(&tty_lock);
1024 pgrp = p->p_pgrp;
1025 LIST_REMOVE(p, p_pglist);
1026 p->p_pgrp = NULL;
1027 mutex_spin_exit(&tty_lock);
1028
1029 if (LIST_EMPTY(&pgrp->pg_members)) {
1030 /* Releases proc_lock. */
1031 pg_delete(pgrp->pg_id);
1032 } else {
1033 mutex_exit(proc_lock);
1034 }
1035}
1036
1037/*
1038 * pg_remove: remove a process group from the table.
1039 * => must be called with the proc_lock held;
1040 * => returns process group to free;
1041 */
1042static struct pgrp *
1043pg_remove(pid_t pg_id)
1044{
1045 struct pgrp *pgrp;
1046 struct pid_table *pt;
1047
1048 KASSERT(mutex_owned(proc_lock));
1049
1050 pt = &pid_table[pg_id & pid_tbl_mask];
1051 pgrp = pt->pt_pgrp;
1052
1053 KASSERT(pgrp != NULL);
1054 KASSERT(pgrp->pg_id == pg_id);
1055 KASSERT(LIST_EMPTY(&pgrp->pg_members));
1056
1057 pt->pt_pgrp = NULL;
1058
1059 if (!P_VALID(pt->pt_proc)) {
1060 /* Orphaned pgrp, put slot onto free list. */
1061 KASSERT((P_NEXT(pt->pt_proc) & pid_tbl_mask) == 0);
1062 pg_id &= pid_tbl_mask;
1063 pt = &pid_table[last_free_pt];
1064 pt->pt_proc = P_FREE(P_NEXT(pt->pt_proc) | pg_id);
1065 KASSERT(pt->pt_pid == 0);
1066 last_free_pt = pg_id;
1067 pid_alloc_cnt--;
1068 }
1069 return pgrp;
1070}
1071
1072/*
1073 * pg_delete: delete and free a process group.
1074 * => must be called with the proc_lock held, which will be released.
1075 */
1076static void
1077pg_delete(pid_t pg_id)
1078{
1079 struct pgrp *pg;
1080 struct tty *ttyp;
1081 struct session *ss;
1082
1083 KASSERT(mutex_owned(proc_lock));
1084
1085 pg = pid_table[pg_id & pid_tbl_mask].pt_pgrp;
1086 if (pg == NULL || pg->pg_id != pg_id || !LIST_EMPTY(&pg->pg_members)) {
1087 mutex_exit(proc_lock);
1088 return;
1089 }
1090
1091 ss = pg->pg_session;
1092
1093 /* Remove reference (if any) from tty to this process group */
1094 mutex_spin_enter(&tty_lock);
1095 ttyp = ss->s_ttyp;
1096 if (ttyp != NULL && ttyp->t_pgrp == pg) {
1097 ttyp->t_pgrp = NULL;
1098 KASSERT(ttyp->t_session == ss);
1099 }
1100 mutex_spin_exit(&tty_lock);
1101
1102 /*
1103 * The leading process group in a session is freed by proc_sessrele(),
1104 * if last reference. Note: proc_sessrele() releases proc_lock.
1105 */
1106 pg = (ss->s_sid != pg->pg_id) ? pg_remove(pg_id) : NULL;
1107 proc_sessrele(ss);
1108
1109 if (pg != NULL) {
1110 /* Free it, if was not done by proc_sessrele(). */
1111 kmem_free(pg, sizeof(struct pgrp));
1112 }
1113}
1114
1115/*
1116 * Adjust pgrp jobc counters when specified process changes process group.
1117 * We count the number of processes in each process group that "qualify"
1118 * the group for terminal job control (those with a parent in a different
1119 * process group of the same session). If that count reaches zero, the
1120 * process group becomes orphaned. Check both the specified process'
1121 * process group and that of its children.
1122 * entering == 0 => p is leaving specified group.
1123 * entering == 1 => p is entering specified group.
1124 *
1125 * Call with proc_lock held.
1126 */
1127void
1128fixjobc(struct proc *p, struct pgrp *pgrp, int entering)
1129{
1130 struct pgrp *hispgrp;
1131 struct session *mysession = pgrp->pg_session;
1132 struct proc *child;
1133
1134 KASSERT(mutex_owned(proc_lock));
1135
1136 /*
1137 * Check p's parent to see whether p qualifies its own process
1138 * group; if so, adjust count for p's process group.
1139 */
1140 hispgrp = p->p_pptr->p_pgrp;
1141 if (hispgrp != pgrp && hispgrp->pg_session == mysession) {
1142 if (entering) {
1143 pgrp->pg_jobc++;
1144 p->p_lflag &= ~PL_ORPHANPG;
1145 } else if (--pgrp->pg_jobc == 0)
1146 orphanpg(pgrp);
1147 }
1148
1149 /*
1150 * Check this process' children to see whether they qualify
1151 * their process groups; if so, adjust counts for children's
1152 * process groups.
1153 */
1154 LIST_FOREACH(child, &p->p_children, p_sibling) {
1155 hispgrp = child->p_pgrp;
1156 if (hispgrp != pgrp && hispgrp->pg_session == mysession &&
1157 !P_ZOMBIE(child)) {
1158 if (entering) {
1159 child->p_lflag &= ~PL_ORPHANPG;
1160 hispgrp->pg_jobc++;
1161 } else if (--hispgrp->pg_jobc == 0)
1162 orphanpg(hispgrp);
1163 }
1164 }
1165}
1166
1167/*
1168 * A process group has become orphaned;
1169 * if there are any stopped processes in the group,
1170 * hang-up all process in that group.
1171 *
1172 * Call with proc_lock held.
1173 */
1174static void
1175orphanpg(struct pgrp *pg)
1176{
1177 struct proc *p;
1178
1179 KASSERT(mutex_owned(proc_lock));
1180
1181 LIST_FOREACH(p, &pg->pg_members, p_pglist) {
1182 if (p->p_stat == SSTOP) {
1183 p->p_lflag |= PL_ORPHANPG;
1184 psignal(p, SIGHUP);
1185 psignal(p, SIGCONT);
1186 }
1187 }
1188}
1189
1190#ifdef DDB
1191#include <ddb/db_output.h>
1192void pidtbl_dump(void);
1193void
1194pidtbl_dump(void)
1195{
1196 struct pid_table *pt;
1197 struct proc *p;
1198 struct pgrp *pgrp;
1199 int id;
1200
1201 db_printf("pid table %p size %x, next %x, last %x\n",
1202 pid_table, pid_tbl_mask+1,
1203 next_free_pt, last_free_pt);
1204 for (pt = pid_table, id = 0; id <= pid_tbl_mask; id++, pt++) {
1205 p = pt->pt_proc;
1206 if (!P_VALID(p) && !pt->pt_pgrp)
1207 continue;
1208 db_printf(" id %x: ", id);
1209 if (P_VALID(p))
1210 db_printf("slotpid %d proc %p id %d (0x%x) %s\n",
1211 pt->pt_pid, p, p->p_pid, p->p_pid, p->p_comm);
1212 else
1213 db_printf("next %x use %x\n",
1214 P_NEXT(p) & pid_tbl_mask,
1215 P_NEXT(p) & ~pid_tbl_mask);
1216 if ((pgrp = pt->pt_pgrp)) {
1217 db_printf("\tsession %p, sid %d, count %d, login %s\n",
1218 pgrp->pg_session, pgrp->pg_session->s_sid,
1219 pgrp->pg_session->s_count,
1220 pgrp->pg_session->s_login);
1221 db_printf("\tpgrp %p, pg_id %d, pg_jobc %d, members %p\n",
1222 pgrp, pgrp->pg_id, pgrp->pg_jobc,
1223 LIST_FIRST(&pgrp->pg_members));
1224 LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
1225 db_printf("\t\tpid %d addr %p pgrp %p %s\n",
1226 p->p_pid, p, p->p_pgrp, p->p_comm);
1227 }
1228 }
1229 }
1230}
1231#endif /* DDB */
1232
1233#ifdef KSTACK_CHECK_MAGIC
1234
1235#define KSTACK_MAGIC 0xdeadbeaf
1236
1237/* XXX should be per process basis? */
1238static int kstackleftmin = KSTACK_SIZE;
1239static int kstackleftthres = KSTACK_SIZE / 8;
1240
1241void
1242kstack_setup_magic(const struct lwp *l)
1243{
1244 uint32_t *ip;
1245 uint32_t const *end;
1246
1247 KASSERT(l != NULL);
1248 KASSERT(l != &lwp0);
1249
1250 /*
1251 * fill all the stack with magic number
1252 * so that later modification on it can be detected.
1253 */
1254 ip = (uint32_t *)KSTACK_LOWEST_ADDR(l);
1255 end = (uint32_t *)((char *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE);
1256 for (; ip < end; ip++) {
1257 *ip = KSTACK_MAGIC;
1258 }
1259}
1260
1261void
1262kstack_check_magic(const struct lwp *l)
1263{
1264 uint32_t const *ip, *end;
1265 int stackleft;
1266
1267 KASSERT(l != NULL);
1268
1269 /* don't check proc0 */ /*XXX*/
1270 if (l == &lwp0)
1271 return;
1272
1273#ifdef __MACHINE_STACK_GROWS_UP
1274 /* stack grows upwards (eg. hppa) */
1275 ip = (uint32_t *)((void *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE);
1276 end = (uint32_t *)KSTACK_LOWEST_ADDR(l);
1277 for (ip--; ip >= end; ip--)
1278 if (*ip != KSTACK_MAGIC)
1279 break;
1280
1281 stackleft = (void *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE - (void *)ip;
1282#else /* __MACHINE_STACK_GROWS_UP */
1283 /* stack grows downwards (eg. i386) */
1284 ip = (uint32_t *)KSTACK_LOWEST_ADDR(l);
1285 end = (uint32_t *)((char *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE);
1286 for (; ip < end; ip++)
1287 if (*ip != KSTACK_MAGIC)
1288 break;
1289
1290 stackleft = ((const char *)ip) - (const char *)KSTACK_LOWEST_ADDR(l);
1291#endif /* __MACHINE_STACK_GROWS_UP */
1292
1293 if (kstackleftmin > stackleft) {
1294 kstackleftmin = stackleft;
1295 if (stackleft < kstackleftthres)
1296 printf("warning: kernel stack left %d bytes"
1297 "(pid %u:lid %u)\n", stackleft,
1298 (u_int)l->l_proc->p_pid, (u_int)l->l_lid);
1299 }
1300
1301 if (stackleft <= 0) {
1302 panic("magic on the top of kernel stack changed for "
1303 "pid %u, lid %u: maybe kernel stack overflow",
1304 (u_int)l->l_proc->p_pid, (u_int)l->l_lid);
1305 }
1306}
1307#endif /* KSTACK_CHECK_MAGIC */
1308
1309int
1310proclist_foreach_call(struct proclist *list,
1311 int (*callback)(struct proc *, void *arg), void *arg)
1312{
1313 struct proc marker;
1314 struct proc *p;
1315 int ret = 0;
1316
1317 marker.p_flag = PK_MARKER;
1318 mutex_enter(proc_lock);
1319 for (p = LIST_FIRST(list); ret == 0 && p != NULL;) {
1320 if (p->p_flag & PK_MARKER) {
1321 p = LIST_NEXT(p, p_list);
1322 continue;
1323 }
1324 LIST_INSERT_AFTER(p, &marker, p_list);
1325 ret = (*callback)(p, arg);
1326 KASSERT(mutex_owned(proc_lock));
1327 p = LIST_NEXT(&marker, p_list);
1328 LIST_REMOVE(&marker, p_list);
1329 }
1330 mutex_exit(proc_lock);
1331
1332 return ret;
1333}
1334
1335int
1336proc_vmspace_getref(struct proc *p, struct vmspace **vm)
1337{
1338
1339 /* XXXCDC: how should locking work here? */
1340
1341 /* curproc exception is for coredump. */
1342
1343 if ((p != curproc && (p->p_sflag & PS_WEXIT) != 0) ||
1344 (p->p_vmspace->vm_refcnt < 1)) { /* XXX */
1345 return EFAULT;
1346 }
1347
1348 uvmspace_addref(p->p_vmspace);
1349 *vm = p->p_vmspace;
1350
1351 return 0;
1352}
1353
1354/*
1355 * Acquire a write lock on the process credential.
1356 */
1357void
1358proc_crmod_enter(void)
1359{
1360 struct lwp *l = curlwp;
1361 struct proc *p = l->l_proc;
1362 kauth_cred_t oc;
1363
1364 /* Reset what needs to be reset in plimit. */
1365 if (p->p_limit->pl_corename != defcorename) {
1366 lim_setcorename(p, defcorename, 0);
1367 }
1368
1369 mutex_enter(p->p_lock);
1370
1371 /* Ensure the LWP cached credentials are up to date. */
1372 if ((oc = l->l_cred) != p->p_cred) {
1373 kauth_cred_hold(p->p_cred);
1374 l->l_cred = p->p_cred;
1375 kauth_cred_free(oc);
1376 }
1377}
1378
1379/*
1380 * Set in a new process credential, and drop the write lock. The credential
1381 * must have a reference already. Optionally, free a no-longer required
1382 * credential. The scheduler also needs to inspect p_cred, so we also
1383 * briefly acquire the sched state mutex.
1384 */
1385void
1386proc_crmod_leave(kauth_cred_t scred, kauth_cred_t fcred, bool sugid)
1387{
1388 struct lwp *l = curlwp, *l2;
1389 struct proc *p = l->l_proc;
1390 kauth_cred_t oc;
1391
1392 KASSERT(mutex_owned(p->p_lock));
1393
1394 /* Is there a new credential to set in? */
1395 if (scred != NULL) {
1396 p->p_cred = scred;
1397 LIST_FOREACH(l2, &p->p_lwps, l_sibling) {
1398 if (l2 != l)
1399 l2->l_prflag |= LPR_CRMOD;
1400 }
1401
1402 /* Ensure the LWP cached credentials are up to date. */
1403 if ((oc = l->l_cred) != scred) {
1404 kauth_cred_hold(scred);
1405 l->l_cred = scred;
1406 }
1407 } else
1408 oc = NULL; /* XXXgcc */
1409
1410 if (sugid) {
1411 /*
1412 * Mark process as having changed credentials, stops
1413 * tracing etc.
1414 */
1415 p->p_flag |= PK_SUGID;
1416 }
1417
1418 mutex_exit(p->p_lock);
1419
1420 /* If there is a credential to be released, free it now. */
1421 if (fcred != NULL) {
1422 KASSERT(scred != NULL);
1423 kauth_cred_free(fcred);
1424 if (oc != scred)
1425 kauth_cred_free(oc);
1426 }
1427}
1428
1429/*
1430 * proc_specific_key_create --
1431 * Create a key for subsystem proc-specific data.
1432 */
1433int
1434proc_specific_key_create(specificdata_key_t *keyp, specificdata_dtor_t dtor)
1435{
1436
1437 return (specificdata_key_create(proc_specificdata_domain, keyp, dtor));
1438}
1439
1440/*
1441 * proc_specific_key_delete --
1442 * Delete a key for subsystem proc-specific data.
1443 */
1444void
1445proc_specific_key_delete(specificdata_key_t key)
1446{
1447
1448 specificdata_key_delete(proc_specificdata_domain, key);
1449}
1450
1451/*
1452 * proc_initspecific --
1453 * Initialize a proc's specificdata container.
1454 */
1455void
1456proc_initspecific(struct proc *p)
1457{
1458 int error __diagused;
1459
1460 error = specificdata_init(proc_specificdata_domain, &p->p_specdataref);
1461 KASSERT(error == 0);
1462}
1463
1464/*
1465 * proc_finispecific --
1466 * Finalize a proc's specificdata container.
1467 */
1468void
1469proc_finispecific(struct proc *p)
1470{
1471
1472 specificdata_fini(proc_specificdata_domain, &p->p_specdataref);
1473}
1474
1475/*
1476 * proc_getspecific --
1477 * Return proc-specific data corresponding to the specified key.
1478 */
1479void *
1480proc_getspecific(struct proc *p, specificdata_key_t key)
1481{
1482
1483 return (specificdata_getspecific(proc_specificdata_domain,
1484 &p->p_specdataref, key));
1485}
1486
1487/*
1488 * proc_setspecific --
1489 * Set proc-specific data corresponding to the specified key.
1490 */
1491void
1492proc_setspecific(struct proc *p, specificdata_key_t key, void *data)
1493{
1494
1495 specificdata_setspecific(proc_specificdata_domain,
1496 &p->p_specdataref, key, data);
1497}
1498
1499int
1500proc_uidmatch(kauth_cred_t cred, kauth_cred_t target)
1501{
1502 int r = 0;
1503
1504 if (kauth_cred_getuid(cred) != kauth_cred_getuid(target) ||
1505 kauth_cred_getuid(cred) != kauth_cred_getsvuid(target)) {
1506 /*
1507 * suid proc of ours or proc not ours
1508 */
1509 r = EPERM;
1510 } else if (kauth_cred_getgid(target) != kauth_cred_getsvgid(target)) {
1511 /*
1512 * sgid proc has sgid back to us temporarily
1513 */
1514 r = EPERM;
1515 } else {
1516 /*
1517 * our rgid must be in target's group list (ie,
1518 * sub-processes started by a sgid process)
1519 */
1520 int ismember = 0;
1521
1522 if (kauth_cred_ismember_gid(cred,
1523 kauth_cred_getgid(target), &ismember) != 0 ||
1524 !ismember)
1525 r = EPERM;
1526 }
1527
1528 return (r);
1529}
1530
1531/*
1532 * sysctl stuff
1533 */
1534
1535#define KERN_PROCSLOP (5 * sizeof(struct kinfo_proc))
1536
1537static const u_int sysctl_flagmap[] = {
1538 PK_ADVLOCK, P_ADVLOCK,
1539 PK_EXEC, P_EXEC,
1540 PK_NOCLDWAIT, P_NOCLDWAIT,
1541 PK_32, P_32,
1542 PK_CLDSIGIGN, P_CLDSIGIGN,
1543 PK_SUGID, P_SUGID,
1544 0
1545};
1546
1547static const u_int sysctl_sflagmap[] = {
1548 PS_NOCLDSTOP, P_NOCLDSTOP,
1549 PS_WEXIT, P_WEXIT,
1550 PS_STOPFORK, P_STOPFORK,
1551 PS_STOPEXEC, P_STOPEXEC,
1552 PS_STOPEXIT, P_STOPEXIT,
1553 0
1554};
1555
1556static const u_int sysctl_slflagmap[] = {
1557 PSL_TRACED, P_TRACED,
1558 PSL_FSTRACE, P_FSTRACE,
1559 PSL_CHTRACED, P_CHTRACED,
1560 PSL_SYSCALL, P_SYSCALL,
1561 0
1562};
1563
1564static const u_int sysctl_lflagmap[] = {
1565 PL_CONTROLT, P_CONTROLT,
1566 PL_PPWAIT, P_PPWAIT,
1567 0
1568};
1569
1570static const u_int sysctl_stflagmap[] = {
1571 PST_PROFIL, P_PROFIL,
1572 0
1573
1574};
1575
1576/* used by kern_lwp also */
1577const u_int sysctl_lwpflagmap[] = {
1578 LW_SINTR, L_SINTR,
1579 LW_SYSTEM, L_SYSTEM,
1580 0
1581};
1582
1583/*
1584 * Find the most ``active'' lwp of a process and return it for ps display
1585 * purposes
1586 */
1587static struct lwp *
1588proc_active_lwp(struct proc *p)
1589{
1590 static const int ostat[] = {
1591 0,
1592 2, /* LSIDL */
1593 6, /* LSRUN */
1594 5, /* LSSLEEP */
1595 4, /* LSSTOP */
1596 0, /* LSZOMB */
1597 1, /* LSDEAD */
1598 7, /* LSONPROC */
1599 3 /* LSSUSPENDED */
1600 };
1601
1602 struct lwp *l, *lp = NULL;
1603 LIST_FOREACH(l, &p->p_lwps, l_sibling) {
1604 KASSERT(l->l_stat >= 0 && l->l_stat < __arraycount(ostat));
1605 if (lp == NULL ||
1606 ostat[l->l_stat] > ostat[lp->l_stat] ||
1607 (ostat[l->l_stat] == ostat[lp->l_stat] &&
1608 l->l_cpticks > lp->l_cpticks)) {
1609 lp = l;
1610 continue;
1611 }
1612 }
1613 return lp;
1614}
1615
1616static int
1617sysctl_doeproc(SYSCTLFN_ARGS)
1618{
1619 union {
1620 struct kinfo_proc kproc;
1621 struct kinfo_proc2 kproc2;
1622 } *kbuf;
1623 struct proc *p, *next, *marker;
1624 char *where, *dp;
1625 int type, op, arg, error;
1626 u_int elem_size, kelem_size, elem_count;
1627 size_t buflen, needed;
1628 bool match, zombie, mmmbrains;
1629
1630 if (namelen == 1 && name[0] == CTL_QUERY)
1631 return (sysctl_query(SYSCTLFN_CALL(rnode)));
1632
1633 dp = where = oldp;
1634 buflen = where != NULL ? *oldlenp : 0;
1635 error = 0;
1636 needed = 0;
1637 type = rnode->sysctl_num;
1638
1639 if (type == KERN_PROC) {
1640 if (namelen == 0)
1641 return EINVAL;
1642 switch (op = name[0]) {
1643 case KERN_PROC_ALL:
1644 if (namelen != 1)
1645 return EINVAL;
1646 arg = 0;
1647 break;
1648 default:
1649 if (namelen != 2)
1650 return EINVAL;
1651 arg = name[1];
1652 break;
1653 }
1654 elem_count = 0; /* Ditto */
1655 kelem_size = elem_size = sizeof(kbuf->kproc);
1656 } else {
1657 if (namelen != 4)
1658 return EINVAL;
1659 op = name[0];
1660 arg = name[1];
1661 elem_size = name[2];
1662 elem_count = name[3];
1663 kelem_size = sizeof(kbuf->kproc2);
1664 }
1665
1666 sysctl_unlock();
1667
1668 kbuf = kmem_alloc(sizeof(*kbuf), KM_SLEEP);
1669 marker = kmem_alloc(sizeof(*marker), KM_SLEEP);
1670 marker->p_flag = PK_MARKER;
1671
1672 mutex_enter(proc_lock);
1673 mmmbrains = false;
1674 for (p = LIST_FIRST(&allproc);; p = next) {
1675 if (p == NULL) {
1676 if (!mmmbrains) {
1677 p = LIST_FIRST(&zombproc);
1678 mmmbrains = true;
1679 }
1680 if (p == NULL)
1681 break;
1682 }
1683 next = LIST_NEXT(p, p_list);
1684 if ((p->p_flag & PK_MARKER) != 0)
1685 continue;
1686
1687 /*
1688 * Skip embryonic processes.
1689 */
1690 if (p->p_stat == SIDL)
1691 continue;
1692
1693 mutex_enter(p->p_lock);
1694 error = kauth_authorize_process(l->l_cred,
1695 KAUTH_PROCESS_CANSEE, p,
1696 KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL);
1697 if (error != 0) {
1698 mutex_exit(p->p_lock);
1699 continue;
1700 }
1701
1702 /*
1703 * TODO - make more efficient (see notes below).
1704 * do by session.
1705 */
1706 switch (op) {
1707 case KERN_PROC_PID:
1708 /* could do this with just a lookup */
1709 match = (p->p_pid == (pid_t)arg);
1710 break;
1711
1712 case KERN_PROC_PGRP:
1713 /* could do this by traversing pgrp */
1714 match = (p->p_pgrp->pg_id == (pid_t)arg);
1715 break;
1716
1717 case KERN_PROC_SESSION:
1718 match = (p->p_session->s_sid == (pid_t)arg);
1719 break;
1720
1721 case KERN_PROC_TTY:
1722 match = true;
1723 if (arg == (int) KERN_PROC_TTY_REVOKE) {
1724 if ((p->p_lflag & PL_CONTROLT) == 0 ||
1725 p->p_session->s_ttyp == NULL ||
1726 p->p_session->s_ttyvp != NULL) {
1727 match = false;
1728 }
1729 } else if ((p->p_lflag & PL_CONTROLT) == 0 ||
1730 p->p_session->s_ttyp == NULL) {
1731 if ((dev_t)arg != KERN_PROC_TTY_NODEV) {
1732 match = false;
1733 }
1734 } else if (p->p_session->s_ttyp->t_dev != (dev_t)arg) {
1735 match = false;
1736 }
1737 break;
1738
1739 case KERN_PROC_UID:
1740 match = (kauth_cred_geteuid(p->p_cred) == (uid_t)arg);
1741 break;
1742
1743 case KERN_PROC_RUID:
1744 match = (kauth_cred_getuid(p->p_cred) == (uid_t)arg);
1745 break;
1746
1747 case KERN_PROC_GID:
1748 match = (kauth_cred_getegid(p->p_cred) == (uid_t)arg);
1749 break;
1750
1751 case KERN_PROC_RGID:
1752 match = (kauth_cred_getgid(p->p_cred) == (uid_t)arg);
1753 break;
1754
1755 case KERN_PROC_ALL:
1756 match = true;
1757 /* allow everything */
1758 break;
1759
1760 default:
1761 error = EINVAL;
1762 mutex_exit(p->p_lock);
1763 goto cleanup;
1764 }
1765 if (!match) {
1766 mutex_exit(p->p_lock);
1767 continue;
1768 }
1769
1770 /*
1771 * Grab a hold on the process.
1772 */
1773 if (mmmbrains) {
1774 zombie = true;
1775 } else {
1776 zombie = !rw_tryenter(&p->p_reflock, RW_READER);
1777 }
1778 if (zombie) {
1779 LIST_INSERT_AFTER(p, marker, p_list);
1780 }
1781
1782 if (buflen >= elem_size &&
1783 (type == KERN_PROC || elem_count > 0)) {
1784 if (type == KERN_PROC) {
1785 kbuf->kproc.kp_proc = *p;
1786 fill_eproc(p, &kbuf->kproc.kp_eproc, zombie);
1787 } else {
1788 fill_kproc2(p, &kbuf->kproc2, zombie);
1789 elem_count--;
1790 }
1791 mutex_exit(p->p_lock);
1792 mutex_exit(proc_lock);
1793 /*
1794 * Copy out elem_size, but not larger than kelem_size
1795 */
1796 error = sysctl_copyout(l, kbuf, dp,
1797 min(kelem_size, elem_size));
1798 mutex_enter(proc_lock);
1799 if (error) {
1800 goto bah;
1801 }
1802 dp += elem_size;
1803 buflen -= elem_size;
1804 } else {
1805 mutex_exit(p->p_lock);
1806 }
1807 needed += elem_size;
1808
1809 /*
1810 * Release reference to process.
1811 */
1812 if (zombie) {
1813 next = LIST_NEXT(marker, p_list);
1814 LIST_REMOVE(marker, p_list);
1815 } else {
1816 rw_exit(&p->p_reflock);
1817 next = LIST_NEXT(p, p_list);
1818 }
1819 }
1820 mutex_exit(proc_lock);
1821
1822 if (where != NULL) {
1823 *oldlenp = dp - where;
1824 if (needed > *oldlenp) {
1825 error = ENOMEM;
1826 goto out;
1827 }
1828 } else {
1829 needed += KERN_PROCSLOP;
1830 *oldlenp = needed;
1831 }
1832 if (kbuf)
1833 kmem_free(kbuf, sizeof(*kbuf));
1834 if (marker)
1835 kmem_free(marker, sizeof(*marker));
1836 sysctl_relock();
1837 return 0;
1838 bah:
1839 if (zombie)
1840 LIST_REMOVE(marker, p_list);
1841 else
1842 rw_exit(&p->p_reflock);
1843 cleanup:
1844 mutex_exit(proc_lock);
1845 out:
1846 if (kbuf)
1847 kmem_free(kbuf, sizeof(*kbuf));
1848 if (marker)
1849 kmem_free(marker, sizeof(*marker));
1850 sysctl_relock();
1851 return error;
1852}
1853
1854int
1855copyin_psstrings(struct proc *p, struct ps_strings *arginfo)
1856{
1857
1858#ifdef COMPAT_NETBSD32
1859 if (p->p_flag & PK_32) {
1860 struct ps_strings32 arginfo32;
1861
1862 int error = copyin_proc(p, (void *)p->p_psstrp, &arginfo32,
1863 sizeof(arginfo32));
1864 if (error)
1865 return error;
1866 arginfo->ps_argvstr = (void *)(uintptr_t)arginfo32.ps_argvstr;
1867 arginfo->ps_nargvstr = arginfo32.ps_nargvstr;
1868 arginfo->ps_envstr = (void *)(uintptr_t)arginfo32.ps_envstr;
1869 arginfo->ps_nenvstr = arginfo32.ps_nenvstr;
1870 return 0;
1871 }
1872#endif
1873 return copyin_proc(p, (void *)p->p_psstrp, arginfo, sizeof(*arginfo));
1874}
1875
1876static int
1877copy_procargs_sysctl_cb(void *cookie_, const void *src, size_t off, size_t len)
1878{
1879 void **cookie = cookie_;
1880 struct lwp *l = cookie[0];
1881 char *dst = cookie[1];
1882
1883 return sysctl_copyout(l, src, dst + off, len);
1884}
1885
1886/*
1887 * sysctl helper routine for kern.proc_args pseudo-subtree.
1888 */
1889static int
1890sysctl_kern_proc_args(SYSCTLFN_ARGS)
1891{
1892 struct ps_strings pss;
1893 struct proc *p;
1894 pid_t pid;
1895 int type, error;
1896 void *cookie[2];
1897
1898 if (namelen == 1 && name[0] == CTL_QUERY)
1899 return (sysctl_query(SYSCTLFN_CALL(rnode)));
1900
1901 if (newp != NULL || namelen != 2)
1902 return (EINVAL);
1903 pid = name[0];
1904 type = name[1];
1905
1906 switch (type) {
1907 case KERN_PROC_PATHNAME:
1908 sysctl_unlock();
1909 error = fill_pathname(l, pid, oldp, oldlenp);
1910 sysctl_relock();
1911 return error;
1912
1913 case KERN_PROC_ARGV:
1914 case KERN_PROC_NARGV:
1915 case KERN_PROC_ENV:
1916 case KERN_PROC_NENV:
1917 /* ok */
1918 break;
1919 default:
1920 return (EINVAL);
1921 }
1922
1923 sysctl_unlock();
1924
1925 /* check pid */
1926 mutex_enter(proc_lock);
1927 if ((p = proc_find(pid)) == NULL) {
1928 error = EINVAL;
1929 goto out_locked;
1930 }
1931 mutex_enter(p->p_lock);
1932
1933 /* Check permission. */
1934 if (type == KERN_PROC_ARGV || type == KERN_PROC_NARGV)
1935 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE,
1936 p, KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ARGS), NULL, NULL);
1937 else if (type == KERN_PROC_ENV || type == KERN_PROC_NENV)
1938 error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE,
1939 p, KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENV), NULL, NULL);
1940 else
1941 error = EINVAL; /* XXXGCC */
1942 if (error) {
1943 mutex_exit(p->p_lock);
1944 goto out_locked;
1945 }
1946
1947 if (oldp == NULL) {
1948 if (type == KERN_PROC_NARGV || type == KERN_PROC_NENV)
1949 *oldlenp = sizeof (int);
1950 else
1951 *oldlenp = ARG_MAX; /* XXX XXX XXX */
1952 error = 0;
1953 mutex_exit(p->p_lock);
1954 goto out_locked;
1955 }
1956
1957 /*
1958 * Zombies don't have a stack, so we can't read their psstrings.
1959 * System processes also don't have a user stack.
1960 */
1961 if (P_ZOMBIE(p) || (p->p_flag & PK_SYSTEM) != 0) {
1962 error = EINVAL;
1963 mutex_exit(p->p_lock);
1964 goto out_locked;
1965 }
1966
1967 error = rw_tryenter(&p->p_reflock, RW_READER) ? 0 : EBUSY;
1968 mutex_exit(p->p_lock);
1969 if (error) {
1970 goto out_locked;
1971 }
1972 mutex_exit(proc_lock);
1973
1974 if (type == KERN_PROC_NARGV || type == KERN_PROC_NENV) {
1975 int value;
1976 if ((error = copyin_psstrings(p, &pss)) == 0) {
1977 if (type == KERN_PROC_NARGV)
1978 value = pss.ps_nargvstr;
1979 else
1980 value = pss.ps_nenvstr;
1981 error = sysctl_copyout(l, &value, oldp, sizeof(value));
1982 *oldlenp = sizeof(value);
1983 }
1984 } else {
1985 cookie[0] = l;
1986 cookie[1] = oldp;
1987 error = copy_procargs(p, type, oldlenp,
1988 copy_procargs_sysctl_cb, cookie);
1989 }
1990 rw_exit(&p->p_reflock);
1991 sysctl_relock();
1992 return error;
1993
1994out_locked:
1995 mutex_exit(proc_lock);
1996 sysctl_relock();
1997 return error;
1998}
1999
2000int
2001copy_procargs(struct proc *p, int oid, size_t *limit,
2002 int (*cb)(void *, const void *, size_t, size_t), void *cookie)
2003{
2004 struct ps_strings pss;
2005 size_t len, i, loaded, entry_len;
2006 struct uio auio;
2007 struct iovec aiov;
2008 int error, argvlen;
2009 char *arg;
2010 char **argv;
2011 vaddr_t user_argv;
2012 struct vmspace *vmspace;
2013
2014 /*
2015 * Allocate a temporary buffer to hold the argument vector and
2016 * the arguments themselve.
2017 */
2018 arg = kmem_alloc(PAGE_SIZE, KM_SLEEP);
2019 argv = kmem_alloc(PAGE_SIZE, KM_SLEEP);
2020
2021 /*
2022 * Lock the process down in memory.
2023 */
2024 vmspace = p->p_vmspace;
2025 uvmspace_addref(vmspace);
2026
2027 /*
2028 * Read in the ps_strings structure.
2029 */
2030 if ((error = copyin_psstrings(p, &pss)) != 0)
2031 goto done;
2032
2033 /*
2034 * Now read the address of the argument vector.
2035 */
2036 switch (oid) {
2037 case KERN_PROC_ARGV:
2038 user_argv = (uintptr_t)pss.ps_argvstr;
2039 argvlen = pss.ps_nargvstr;
2040 break;
2041 case KERN_PROC_ENV:
2042 user_argv = (uintptr_t)pss.ps_envstr;
2043 argvlen = pss.ps_nenvstr;
2044 break;
2045 default:
2046 error = EINVAL;
2047 goto done;
2048 }
2049
2050 if (argvlen < 0) {
2051 error = EIO;
2052 goto done;
2053 }
2054
2055
2056 /*
2057 * Now copy each string.
2058 */
2059 len = 0; /* bytes written to user buffer */
2060 loaded = 0; /* bytes from argv already processed */
2061 i = 0; /* To make compiler happy */
2062 entry_len = PROC_PTRSZ(p);
2063
2064 for (; argvlen; --argvlen) {
2065 int finished = 0;
2066 vaddr_t base;
2067 size_t xlen;
2068 int j;
2069
2070 if (loaded == 0) {
2071 size_t rem = entry_len * argvlen;
2072 loaded = MIN(rem, PAGE_SIZE);
2073 error = copyin_vmspace(vmspace,
2074 (const void *)user_argv, argv, loaded);
2075 if (error)
2076 break;
2077 user_argv += loaded;
2078 i = 0;
2079 }
2080
2081#ifdef COMPAT_NETBSD32
2082 if (p->p_flag & PK_32) {
2083 netbsd32_charp *argv32;
2084
2085 argv32 = (netbsd32_charp *)argv;
2086 base = (vaddr_t)NETBSD32PTR64(argv32[i++]);
2087 } else
2088#endif
2089 base = (vaddr_t)argv[i++];
2090 loaded -= entry_len;
2091
2092 /*
2093 * The program has messed around with its arguments,
2094 * possibly deleting some, and replacing them with
2095 * NULL's. Treat this as the last argument and not
2096 * a failure.
2097 */
2098 if (base == 0)
2099 break;
2100
2101 while (!finished) {
2102 xlen = PAGE_SIZE - (base & PAGE_MASK);
2103
2104 aiov.iov_base = arg;
2105 aiov.iov_len = PAGE_SIZE;
2106 auio.uio_iov = &aiov;
2107 auio.uio_iovcnt = 1;
2108 auio.uio_offset = base;
2109 auio.uio_resid = xlen;
2110 auio.uio_rw = UIO_READ;
2111 UIO_SETUP_SYSSPACE(&auio);
2112 error = uvm_io(&vmspace->vm_map, &auio, 0);
2113 if (error)
2114 goto done;
2115
2116 /* Look for the end of the string */
2117 for (j = 0; j < xlen; j++) {
2118 if (arg[j] == '\0') {
2119 xlen = j + 1;
2120 finished = 1;
2121 break;
2122 }
2123 }
2124
2125 /* Check for user buffer overflow */
2126 if (len + xlen > *limit) {
2127 finished = 1;
2128 if (len > *limit)
2129 xlen = 0;
2130 else
2131 xlen = *limit - len;
2132 }
2133
2134 /* Copyout the page */
2135 error = (*cb)(cookie, arg, len, xlen);
2136 if (error)
2137 goto done;
2138
2139 len += xlen;
2140 base += xlen;
2141 }
2142 }
2143 *limit = len;
2144
2145done:
2146 kmem_free(argv, PAGE_SIZE);
2147 kmem_free(arg, PAGE_SIZE);
2148 uvmspace_free(vmspace);
2149 return error;
2150}
2151
2152/*
2153 * Fill in an eproc structure for the specified process.
2154 */
2155void
2156fill_eproc(struct proc *p, struct eproc *ep, bool zombie)
2157{
2158 struct tty *tp;
2159 struct lwp *l;
2160
2161 KASSERT(mutex_owned(proc_lock));
2162 KASSERT(mutex_owned(p->p_lock));
2163
2164 memset(ep, 0, sizeof(*ep));
2165
2166 ep->e_paddr = p;
2167 ep->e_sess = p->p_session;
2168 if (p->p_cred) {
2169 kauth_cred_topcred(p->p_cred, &ep->e_pcred);
2170 kauth_cred_toucred(p->p_cred, &ep->e_ucred);
2171 }
2172 if (p->p_stat != SIDL && !P_ZOMBIE(p) && !zombie) {
2173 struct vmspace *vm = p->p_vmspace;
2174
2175 ep->e_vm.vm_rssize = vm_resident_count(vm);
2176 ep->e_vm.vm_tsize = vm->vm_tsize;
2177 ep->e_vm.vm_dsize = vm->vm_dsize;
2178 ep->e_vm.vm_ssize = vm->vm_ssize;
2179 ep->e_vm.vm_map.size = vm->vm_map.size;
2180
2181 /* Pick the primary (first) LWP */
2182 l = proc_active_lwp(p);
2183 KASSERT(l != NULL);
2184 lwp_lock(l);
2185 if (l->l_wchan)
2186 strncpy(ep->e_wmesg, l->l_wmesg, WMESGLEN);
2187 lwp_unlock(l);
2188 }
2189 ep->e_ppid = p->p_ppid;
2190 if (p->p_pgrp && p->p_session) {
2191 ep->e_pgid = p->p_pgrp->pg_id;
2192 ep->e_jobc = p->p_pgrp->pg_jobc;
2193 ep->e_sid = p->p_session->s_sid;
2194 if ((p->p_lflag & PL_CONTROLT) &&
2195 (tp = ep->e_sess->s_ttyp)) {
2196 ep->e_tdev = tp->t_dev;
2197 ep->e_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PGID;
2198 ep->e_tsess = tp->t_session;
2199 } else
2200 ep->e_tdev = (uint32_t)NODEV;
2201 ep->e_flag = ep->e_sess->s_ttyvp ? EPROC_CTTY : 0;
2202 if (SESS_LEADER(p))
2203 ep->e_flag |= EPROC_SLEADER;
2204 strncpy(ep->e_login, ep->e_sess->s_login, MAXLOGNAME);
2205 }
2206 ep->e_xsize = ep->e_xrssize = 0;
2207 ep->e_xccount = ep->e_xswrss = 0;
2208}
2209
2210/*
2211 * Fill in a kinfo_proc2 structure for the specified process.
2212 */
2213void
2214fill_kproc2(struct proc *p, struct kinfo_proc2 *ki, bool zombie)
2215{
2216 struct tty *tp;
2217 struct lwp *l, *l2;
2218 struct timeval ut, st, rt;
2219 sigset_t ss1, ss2;
2220 struct rusage ru;
2221 struct vmspace *vm;
2222
2223 KASSERT(mutex_owned(proc_lock));
2224 KASSERT(mutex_owned(p->p_lock));
2225
2226 sigemptyset(&ss1);
2227 sigemptyset(&ss2);
2228 memset(ki, 0, sizeof(*ki));
2229
2230 ki->p_paddr = PTRTOUINT64(p);
2231 ki->p_fd = PTRTOUINT64(p->p_fd);
2232 ki->p_cwdi = PTRTOUINT64(p->p_cwdi);
2233 ki->p_stats = PTRTOUINT64(p->p_stats);
2234 ki->p_limit = PTRTOUINT64(p->p_limit);
2235 ki->p_vmspace = PTRTOUINT64(p->p_vmspace);
2236 ki->p_sigacts = PTRTOUINT64(p->p_sigacts);
2237 ki->p_sess = PTRTOUINT64(p->p_session);
2238 ki->p_tsess = 0; /* may be changed if controlling tty below */
2239 ki->p_ru = PTRTOUINT64(&p->p_stats->p_ru);
2240 ki->p_eflag = 0;
2241 ki->p_exitsig = p->p_exitsig;
2242 ki->p_flag = L_INMEM; /* Process never swapped out */
2243 ki->p_flag |= sysctl_map_flags(sysctl_flagmap, p->p_flag);
2244 ki->p_flag |= sysctl_map_flags(sysctl_sflagmap, p->p_sflag);
2245 ki->p_flag |= sysctl_map_flags(sysctl_slflagmap, p->p_slflag);
2246 ki->p_flag |= sysctl_map_flags(sysctl_lflagmap, p->p_lflag);
2247 ki->p_flag |= sysctl_map_flags(sysctl_stflagmap, p->p_stflag);
2248 ki->p_pid = p->p_pid;
2249 ki->p_ppid = p->p_ppid;
2250 ki->p_uid = kauth_cred_geteuid(p->p_cred);
2251 ki->p_ruid = kauth_cred_getuid(p->p_cred);
2252 ki->p_gid = kauth_cred_getegid(p->p_cred);
2253 ki->p_rgid = kauth_cred_getgid(p->p_cred);
2254 ki->p_svuid = kauth_cred_getsvuid(p->p_cred);
2255 ki->p_svgid = kauth_cred_getsvgid(p->p_cred);
2256 ki->p_ngroups = kauth_cred_ngroups(p->p_cred);
2257 kauth_cred_getgroups(p->p_cred, ki->p_groups,
2258 min(ki->p_ngroups, sizeof(ki->p_groups) / sizeof(ki->p_groups[0])),
2259 UIO_SYSSPACE);
2260
2261 ki->p_uticks = p->p_uticks;
2262 ki->p_sticks = p->p_sticks;
2263 ki->p_iticks = p->p_iticks;
2264 ki->p_tpgid = NO_PGID; /* may be changed if controlling tty below */
2265 ki->p_tracep = PTRTOUINT64(p->p_tracep);
2266 ki->p_traceflag = p->p_traceflag;
2267
2268 memcpy(&ki->p_sigignore, &p->p_sigctx.ps_sigignore,sizeof(ki_sigset_t));
2269 memcpy(&ki->p_sigcatch, &p->p_sigctx.ps_sigcatch, sizeof(ki_sigset_t));
2270
2271 ki->p_cpticks = 0;
2272 ki->p_pctcpu = p->p_pctcpu;
2273 ki->p_estcpu = 0;
2274 ki->p_stat = p->p_stat; /* Will likely be overridden by LWP status */
2275 ki->p_realstat = p->p_stat;
2276 ki->p_nice = p->p_nice;
2277 ki->p_xstat = P_WAITSTATUS(p);
2278 ki->p_acflag = p->p_acflag;
2279
2280 strncpy(ki->p_comm, p->p_comm,
2281 min(sizeof(ki->p_comm), sizeof(p->p_comm)));
2282 strncpy(ki->p_ename, p->p_emul->e_name, sizeof(ki->p_ename));
2283
2284 ki->p_nlwps = p->p_nlwps;
2285 ki->p_realflag = ki->p_flag;
2286
2287 if (p->p_stat != SIDL && !P_ZOMBIE(p) && !zombie) {
2288 vm = p->p_vmspace;
2289 ki->p_vm_rssize = vm_resident_count(vm);
2290 ki->p_vm_tsize = vm->vm_tsize;
2291 ki->p_vm_dsize = vm->vm_dsize;
2292 ki->p_vm_ssize = vm->vm_ssize;
2293 ki->p_vm_vsize = atop(vm->vm_map.size);
2294 /*
2295 * Since the stack is initially mapped mostly with
2296 * PROT_NONE and grown as needed, adjust the "mapped size"
2297 * to skip the unused stack portion.
2298 */
2299 ki->p_vm_msize =
2300 atop(vm->vm_map.size) - vm->vm_issize + vm->vm_ssize;
2301
2302 /* Pick the primary (first) LWP */
2303 l = proc_active_lwp(p);
2304 KASSERT(l != NULL);
2305 lwp_lock(l);
2306 ki->p_nrlwps = p->p_nrlwps;
2307 ki->p_forw = 0;
2308 ki->p_back = 0;
2309 ki->p_addr = PTRTOUINT64(l->l_addr);
2310 ki->p_stat = l->l_stat;
2311 ki->p_flag |= sysctl_map_flags(sysctl_lwpflagmap, l->l_flag);
2312 ki->p_swtime = l->l_swtime;
2313 ki->p_slptime = l->l_slptime;
2314 if (l->l_stat == LSONPROC)
2315 ki->p_schedflags = l->l_cpu->ci_schedstate.spc_flags;
2316 else
2317 ki->p_schedflags = 0;
2318 ki->p_priority = lwp_eprio(l);
2319 ki->p_usrpri = l->l_priority;
2320 if (l->l_wchan)
2321 strncpy(ki->p_wmesg, l->l_wmesg, sizeof(ki->p_wmesg));
2322 ki->p_wchan = PTRTOUINT64(l->l_wchan);
2323 ki->p_cpuid = cpu_index(l->l_cpu);
2324 lwp_unlock(l);
2325 LIST_FOREACH(l, &p->p_lwps, l_sibling) {
2326 /* This is hardly correct, but... */
2327 sigplusset(&l->l_sigpend.sp_set, &ss1);
2328 sigplusset(&l->l_sigmask, &ss2);
2329 ki->p_cpticks += l->l_cpticks;
2330 ki->p_pctcpu += l->l_pctcpu;
2331 ki->p_estcpu += l->l_estcpu;
2332 }
2333 }
2334 sigplusset(&p->p_sigpend.sp_set, &ss2);
2335 memcpy(&ki->p_siglist, &ss1, sizeof(ki_sigset_t));
2336 memcpy(&ki->p_sigmask, &ss2, sizeof(ki_sigset_t));
2337
2338 if (p->p_session != NULL) {
2339 ki->p_sid = p->p_session->s_sid;
2340 ki->p__pgid = p->p_pgrp->pg_id;
2341 if (p->p_session->s_ttyvp)
2342 ki->p_eflag |= EPROC_CTTY;
2343 if (SESS_LEADER(p))
2344 ki->p_eflag |= EPROC_SLEADER;
2345 strncpy(ki->p_login, p->p_session->s_login,
2346 min(sizeof ki->p_login - 1, sizeof p->p_session->s_login));
2347 ki->p_jobc = p->p_pgrp->pg_jobc;
2348 if ((p->p_lflag & PL_CONTROLT) && (tp = p->p_session->s_ttyp)) {
2349 ki->p_tdev = tp->t_dev;
2350 ki->p_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PGID;
2351 ki->p_tsess = PTRTOUINT64(tp->t_session);
2352 } else {
2353 ki->p_tdev = (int32_t)NODEV;
2354 }
2355 }
2356
2357 if (!P_ZOMBIE(p) && !zombie) {
2358 ki->p_uvalid = 1;
2359 ki->p_ustart_sec = p->p_stats->p_start.tv_sec;
2360 ki->p_ustart_usec = p->p_stats->p_start.tv_usec;
2361
2362 calcru(p, &ut, &st, NULL, &rt);
2363 ki->p_rtime_sec = rt.tv_sec;
2364 ki->p_rtime_usec = rt.tv_usec;
2365 ki->p_uutime_sec = ut.tv_sec;
2366 ki->p_uutime_usec = ut.tv_usec;
2367 ki->p_ustime_sec = st.tv_sec;
2368 ki->p_ustime_usec = st.tv_usec;
2369
2370 memcpy(&ru, &p->p_stats->p_ru, sizeof(ru));
2371 ki->p_uru_nvcsw = 0;
2372 ki->p_uru_nivcsw = 0;
2373 LIST_FOREACH(l2, &p->p_lwps, l_sibling) {
2374 ki->p_uru_nvcsw += (l2->l_ncsw - l2->l_nivcsw);
2375 ki->p_uru_nivcsw += l2->l_nivcsw;
2376 ruadd(&ru, &l2->l_ru);
2377 }
2378 ki->p_uru_maxrss = ru.ru_maxrss;
2379 ki->p_uru_ixrss = ru.ru_ixrss;
2380 ki->p_uru_idrss = ru.ru_idrss;
2381 ki->p_uru_isrss = ru.ru_isrss;
2382 ki->p_uru_minflt = ru.ru_minflt;
2383 ki->p_uru_majflt = ru.ru_majflt;
2384 ki->p_uru_nswap = ru.ru_nswap;
2385 ki->p_uru_inblock = ru.ru_inblock;
2386 ki->p_uru_oublock = ru.ru_oublock;
2387 ki->p_uru_msgsnd = ru.ru_msgsnd;
2388 ki->p_uru_msgrcv = ru.ru_msgrcv;
2389 ki->p_uru_nsignals = ru.ru_nsignals;
2390
2391 timeradd(&p->p_stats->p_cru.ru_utime,
2392 &p->p_stats->p_cru.ru_stime, &ut);
2393 ki->p_uctime_sec = ut.tv_sec;
2394 ki->p_uctime_usec = ut.tv_usec;
2395 }
2396}
2397
2398
2399int
2400proc_find_locked(struct lwp *l, struct proc **p, pid_t pid)
2401{
2402 int error;
2403
2404 mutex_enter(proc_lock);
2405 if (pid == -1)
2406 *p = l->l_proc;
2407 else
2408 *p = proc_find(pid);
2409
2410 if (*p == NULL) {
2411 if (pid != -1)
2412 mutex_exit(proc_lock);
2413 return ESRCH;
2414 }
2415 if (pid != -1)
2416 mutex_enter((*p)->p_lock);
2417 mutex_exit(proc_lock);
2418
2419 error = kauth_authorize_process(l->l_cred,
2420 KAUTH_PROCESS_CANSEE, *p,
2421 KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL);
2422 if (error) {
2423 if (pid != -1)
2424 mutex_exit((*p)->p_lock);
2425 }
2426 return error;
2427}
2428
2429static int
2430fill_pathname(struct lwp *l, pid_t pid, void *oldp, size_t *oldlenp)
2431{
2432#ifndef _RUMPKERNEL
2433 int error;
2434 struct proc *p;
2435 char *path;
2436 size_t len;
2437
2438 if ((error = proc_find_locked(l, &p, pid)) != 0)
2439 return error;
2440
2441 if (p->p_textvp == NULL) {
2442 if (pid != -1)
2443 mutex_exit(p->p_lock);
2444 return ENOENT;
2445 }
2446
2447 path = PNBUF_GET();
2448 error = vnode_to_path(path, MAXPATHLEN / 2, p->p_textvp, l, p);
2449 if (error)
2450 goto out;
2451
2452 len = strlen(path) + 1;
2453 if (oldp != NULL) {
2454 error = sysctl_copyout(l, path, oldp, *oldlenp);
2455 if (error == 0 && *oldlenp < len)
2456 error = ENOSPC;
2457 }
2458 *oldlenp = len;
2459out:
2460 PNBUF_PUT(path);
2461 if (pid != -1)
2462 mutex_exit(p->p_lock);
2463 return error;
2464#else
2465 return 0;
2466#endif
2467}
2468