]> CyberLeo.Net >> Repos - FreeBSD/releng/8.1.git/blob - sys/kern/kern_proc.c
Copy stable/8 to releng/8.1 in preparation for 8.1-RC1.
[FreeBSD/releng/8.1.git] / sys / kern / kern_proc.c
1 /*-
2  * Copyright (c) 1982, 1986, 1989, 1991, 1993
3  *      The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 4. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *      @(#)kern_proc.c 8.7 (Berkeley) 2/14/95
30  */
31
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
34
35 #include "opt_compat.h"
36 #include "opt_ddb.h"
37 #include "opt_kdtrace.h"
38 #include "opt_ktrace.h"
39 #include "opt_kstack_pages.h"
40 #include "opt_stack.h"
41
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/kernel.h>
45 #include <sys/limits.h>
46 #include <sys/lock.h>
47 #include <sys/malloc.h>
48 #include <sys/mount.h>
49 #include <sys/mutex.h>
50 #include <sys/proc.h>
51 #include <sys/refcount.h>
52 #include <sys/sbuf.h>
53 #include <sys/sysent.h>
54 #include <sys/sched.h>
55 #include <sys/smp.h>
56 #include <sys/stack.h>
57 #include <sys/sysctl.h>
58 #include <sys/filedesc.h>
59 #include <sys/tty.h>
60 #include <sys/signalvar.h>
61 #include <sys/sdt.h>
62 #include <sys/sx.h>
63 #include <sys/user.h>
64 #include <sys/jail.h>
65 #include <sys/vnode.h>
66 #include <sys/eventhandler.h>
67 #ifdef KTRACE
68 #include <sys/uio.h>
69 #include <sys/ktrace.h>
70 #endif
71
72 #ifdef DDB
73 #include <ddb/ddb.h>
74 #endif
75
76 #include <vm/vm.h>
77 #include <vm/vm_extern.h>
78 #include <vm/pmap.h>
79 #include <vm/vm_map.h>
80 #include <vm/vm_object.h>
81 #include <vm/uma.h>
82
83 #ifdef COMPAT_FREEBSD32
84 #include <compat/freebsd32/freebsd32.h>
85 #include <compat/freebsd32/freebsd32_util.h>
86 #endif
87
88 SDT_PROVIDER_DEFINE(proc);
89 SDT_PROBE_DEFINE(proc, kernel, ctor, entry);
90 SDT_PROBE_ARGTYPE(proc, kernel, ctor, entry, 0, "struct proc *");
91 SDT_PROBE_ARGTYPE(proc, kernel, ctor, entry, 1, "int");
92 SDT_PROBE_ARGTYPE(proc, kernel, ctor, entry, 2, "void *");
93 SDT_PROBE_ARGTYPE(proc, kernel, ctor, entry, 3, "int");
94 SDT_PROBE_DEFINE(proc, kernel, ctor, return);
95 SDT_PROBE_ARGTYPE(proc, kernel, ctor, return, 0, "struct proc *");
96 SDT_PROBE_ARGTYPE(proc, kernel, ctor, return, 1, "int");
97 SDT_PROBE_ARGTYPE(proc, kernel, ctor, return, 2, "void *");
98 SDT_PROBE_ARGTYPE(proc, kernel, ctor, return, 3, "int");
99 SDT_PROBE_DEFINE(proc, kernel, dtor, entry);
100 SDT_PROBE_ARGTYPE(proc, kernel, dtor, entry, 0, "struct proc *");
101 SDT_PROBE_ARGTYPE(proc, kernel, dtor, entry, 1, "int");
102 SDT_PROBE_ARGTYPE(proc, kernel, dtor, entry, 2, "void *");
103 SDT_PROBE_ARGTYPE(proc, kernel, dtor, entry, 3, "struct thread *");
104 SDT_PROBE_DEFINE(proc, kernel, dtor, return);
105 SDT_PROBE_ARGTYPE(proc, kernel, dtor, return, 0, "struct proc *");
106 SDT_PROBE_ARGTYPE(proc, kernel, dtor, return, 1, "int");
107 SDT_PROBE_ARGTYPE(proc, kernel, dtor, return, 2, "void *");
108 SDT_PROBE_DEFINE(proc, kernel, init, entry);
109 SDT_PROBE_ARGTYPE(proc, kernel, init, entry, 0, "struct proc *");
110 SDT_PROBE_ARGTYPE(proc, kernel, init, entry, 1, "int");
111 SDT_PROBE_ARGTYPE(proc, kernel, init, entry, 2, "int");
112 SDT_PROBE_DEFINE(proc, kernel, init, return);
113 SDT_PROBE_ARGTYPE(proc, kernel, init, return, 0, "struct proc *");
114 SDT_PROBE_ARGTYPE(proc, kernel, init, return, 1, "int");
115 SDT_PROBE_ARGTYPE(proc, kernel, init, return, 2, "int");
116
117 MALLOC_DEFINE(M_PGRP, "pgrp", "process group header");
118 MALLOC_DEFINE(M_SESSION, "session", "session header");
119 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
120 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
121
122 static void doenterpgrp(struct proc *, struct pgrp *);
123 static void orphanpg(struct pgrp *pg);
124 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp);
125 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp);
126 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp,
127     int preferthread);
128 static void pgadjustjobc(struct pgrp *pgrp, int entering);
129 static void pgdelete(struct pgrp *);
130 static int proc_ctor(void *mem, int size, void *arg, int flags);
131 static void proc_dtor(void *mem, int size, void *arg);
132 static int proc_init(void *mem, int size, int flags);
133 static void proc_fini(void *mem, int size);
134 static void pargs_free(struct pargs *pa);
135
136 /*
137  * Other process lists
138  */
139 struct pidhashhead *pidhashtbl;
140 u_long pidhash;
141 struct pgrphashhead *pgrphashtbl;
142 u_long pgrphash;
143 struct proclist allproc;
144 struct proclist zombproc;
145 struct sx allproc_lock;
146 struct sx proctree_lock;
147 struct mtx ppeers_lock;
148 uma_zone_t proc_zone;
149
150 int kstack_pages = KSTACK_PAGES;
151 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0, "");
152
153 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
154 #ifdef COMPAT_FREEBSD32
155 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
156 #endif
157
158 /*
159  * Initialize global process hashing structures.
160  */
161 void
162 procinit()
163 {
164
165         sx_init(&allproc_lock, "allproc");
166         sx_init(&proctree_lock, "proctree");
167         mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
168         LIST_INIT(&allproc);
169         LIST_INIT(&zombproc);
170         pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
171         pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
172         proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
173             proc_ctor, proc_dtor, proc_init, proc_fini,
174             UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
175         uihashinit();
176 }
177
178 /*
179  * Prepare a proc for use.
180  */
181 static int
182 proc_ctor(void *mem, int size, void *arg, int flags)
183 {
184         struct proc *p;
185
186         p = (struct proc *)mem;
187         SDT_PROBE(proc, kernel, ctor , entry, p, size, arg, flags, 0);
188         EVENTHANDLER_INVOKE(process_ctor, p);
189         SDT_PROBE(proc, kernel, ctor , return, p, size, arg, flags, 0);
190         return (0);
191 }
192
193 /*
194  * Reclaim a proc after use.
195  */
196 static void
197 proc_dtor(void *mem, int size, void *arg)
198 {
199         struct proc *p;
200         struct thread *td;
201
202         /* INVARIANTS checks go here */
203         p = (struct proc *)mem;
204         td = FIRST_THREAD_IN_PROC(p);
205         SDT_PROBE(proc, kernel, dtor, entry, p, size, arg, td, 0);
206         if (td != NULL) {
207 #ifdef INVARIANTS
208                 KASSERT((p->p_numthreads == 1),
209                     ("bad number of threads in exiting process"));
210                 KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
211 #endif
212                 /* Free all OSD associated to this thread. */
213                 osd_thread_exit(td);
214         }
215         EVENTHANDLER_INVOKE(process_dtor, p);
216         if (p->p_ksi != NULL)
217                 KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
218         SDT_PROBE(proc, kernel, dtor, return, p, size, arg, 0, 0);
219 }
220
221 /*
222  * Initialize type-stable parts of a proc (when newly created).
223  */
224 static int
225 proc_init(void *mem, int size, int flags)
226 {
227         struct proc *p;
228
229         p = (struct proc *)mem;
230         SDT_PROBE(proc, kernel, init, entry, p, size, flags, 0, 0);
231         p->p_sched = (struct p_sched *)&p[1];
232         bzero(&p->p_mtx, sizeof(struct mtx));
233         mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK);
234         mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_RECURSE);
235         cv_init(&p->p_pwait, "ppwait");
236         TAILQ_INIT(&p->p_threads);           /* all threads in proc */
237         EVENTHANDLER_INVOKE(process_init, p);
238         p->p_stats = pstats_alloc();
239         SDT_PROBE(proc, kernel, init, return, p, size, flags, 0, 0);
240         return (0);
241 }
242
243 /*
244  * UMA should ensure that this function is never called.
245  * Freeing a proc structure would violate type stability.
246  */
247 static void
248 proc_fini(void *mem, int size)
249 {
250 #ifdef notnow
251         struct proc *p;
252
253         p = (struct proc *)mem;
254         EVENTHANDLER_INVOKE(process_fini, p);
255         pstats_free(p->p_stats);
256         thread_free(FIRST_THREAD_IN_PROC(p));
257         mtx_destroy(&p->p_mtx);
258         if (p->p_ksi != NULL)
259                 ksiginfo_free(p->p_ksi);
260 #else
261         panic("proc reclaimed");
262 #endif
263 }
264
265 /*
266  * Is p an inferior of the current process?
267  */
268 int
269 inferior(p)
270         register struct proc *p;
271 {
272
273         sx_assert(&proctree_lock, SX_LOCKED);
274         for (; p != curproc; p = p->p_pptr)
275                 if (p->p_pid == 0)
276                         return (0);
277         return (1);
278 }
279
280 /*
281  * Locate a process by number; return only "live" processes -- i.e., neither
282  * zombies nor newly born but incompletely initialized processes.  By not
283  * returning processes in the PRS_NEW state, we allow callers to avoid
284  * testing for that condition to avoid dereferencing p_ucred, et al.
285  */
286 struct proc *
287 pfind(pid)
288         register pid_t pid;
289 {
290         register struct proc *p;
291
292         sx_slock(&allproc_lock);
293         LIST_FOREACH(p, PIDHASH(pid), p_hash)
294                 if (p->p_pid == pid) {
295                         if (p->p_state == PRS_NEW) {
296                                 p = NULL;
297                                 break;
298                         }
299                         PROC_LOCK(p);
300                         break;
301                 }
302         sx_sunlock(&allproc_lock);
303         return (p);
304 }
305
306 /*
307  * Locate a process group by number.
308  * The caller must hold proctree_lock.
309  */
310 struct pgrp *
311 pgfind(pgid)
312         register pid_t pgid;
313 {
314         register struct pgrp *pgrp;
315
316         sx_assert(&proctree_lock, SX_LOCKED);
317
318         LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
319                 if (pgrp->pg_id == pgid) {
320                         PGRP_LOCK(pgrp);
321                         return (pgrp);
322                 }
323         }
324         return (NULL);
325 }
326
327 /*
328  * Create a new process group.
329  * pgid must be equal to the pid of p.
330  * Begin a new session if required.
331  */
332 int
333 enterpgrp(p, pgid, pgrp, sess)
334         register struct proc *p;
335         pid_t pgid;
336         struct pgrp *pgrp;
337         struct session *sess;
338 {
339         struct pgrp *pgrp2;
340
341         sx_assert(&proctree_lock, SX_XLOCKED);
342
343         KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
344         KASSERT(p->p_pid == pgid,
345             ("enterpgrp: new pgrp and pid != pgid"));
346
347         pgrp2 = pgfind(pgid);
348
349         KASSERT(pgrp2 == NULL,
350             ("enterpgrp: pgrp with pgid exists"));
351         KASSERT(!SESS_LEADER(p),
352             ("enterpgrp: session leader attempted setpgrp"));
353
354         mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
355
356         if (sess != NULL) {
357                 /*
358                  * new session
359                  */
360                 mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
361                 PROC_LOCK(p);
362                 p->p_flag &= ~P_CONTROLT;
363                 PROC_UNLOCK(p);
364                 PGRP_LOCK(pgrp);
365                 sess->s_leader = p;
366                 sess->s_sid = p->p_pid;
367                 refcount_init(&sess->s_count, 1);
368                 sess->s_ttyvp = NULL;
369                 sess->s_ttyp = NULL;
370                 bcopy(p->p_session->s_login, sess->s_login,
371                             sizeof(sess->s_login));
372                 pgrp->pg_session = sess;
373                 KASSERT(p == curproc,
374                     ("enterpgrp: mksession and p != curproc"));
375         } else {
376                 pgrp->pg_session = p->p_session;
377                 sess_hold(pgrp->pg_session);
378                 PGRP_LOCK(pgrp);
379         }
380         pgrp->pg_id = pgid;
381         LIST_INIT(&pgrp->pg_members);
382
383         /*
384          * As we have an exclusive lock of proctree_lock,
385          * this should not deadlock.
386          */
387         LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
388         pgrp->pg_jobc = 0;
389         SLIST_INIT(&pgrp->pg_sigiolst);
390         PGRP_UNLOCK(pgrp);
391
392         doenterpgrp(p, pgrp);
393
394         return (0);
395 }
396
397 /*
398  * Move p to an existing process group
399  */
400 int
401 enterthispgrp(p, pgrp)
402         register struct proc *p;
403         struct pgrp *pgrp;
404 {
405
406         sx_assert(&proctree_lock, SX_XLOCKED);
407         PROC_LOCK_ASSERT(p, MA_NOTOWNED);
408         PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
409         PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
410         SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
411         KASSERT(pgrp->pg_session == p->p_session,
412                 ("%s: pgrp's session %p, p->p_session %p.\n",
413                 __func__,
414                 pgrp->pg_session,
415                 p->p_session));
416         KASSERT(pgrp != p->p_pgrp,
417                 ("%s: p belongs to pgrp.", __func__));
418
419         doenterpgrp(p, pgrp);
420
421         return (0);
422 }
423
424 /*
425  * Move p to a process group
426  */
427 static void
428 doenterpgrp(p, pgrp)
429         struct proc *p;
430         struct pgrp *pgrp;
431 {
432         struct pgrp *savepgrp;
433
434         sx_assert(&proctree_lock, SX_XLOCKED);
435         PROC_LOCK_ASSERT(p, MA_NOTOWNED);
436         PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
437         PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
438         SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
439
440         savepgrp = p->p_pgrp;
441
442         /*
443          * Adjust eligibility of affected pgrps to participate in job control.
444          * Increment eligibility counts before decrementing, otherwise we
445          * could reach 0 spuriously during the first call.
446          */
447         fixjobc(p, pgrp, 1);
448         fixjobc(p, p->p_pgrp, 0);
449
450         PGRP_LOCK(pgrp);
451         PGRP_LOCK(savepgrp);
452         PROC_LOCK(p);
453         LIST_REMOVE(p, p_pglist);
454         p->p_pgrp = pgrp;
455         PROC_UNLOCK(p);
456         LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
457         PGRP_UNLOCK(savepgrp);
458         PGRP_UNLOCK(pgrp);
459         if (LIST_EMPTY(&savepgrp->pg_members))
460                 pgdelete(savepgrp);
461 }
462
463 /*
464  * remove process from process group
465  */
466 int
467 leavepgrp(p)
468         register struct proc *p;
469 {
470         struct pgrp *savepgrp;
471
472         sx_assert(&proctree_lock, SX_XLOCKED);
473         savepgrp = p->p_pgrp;
474         PGRP_LOCK(savepgrp);
475         PROC_LOCK(p);
476         LIST_REMOVE(p, p_pglist);
477         p->p_pgrp = NULL;
478         PROC_UNLOCK(p);
479         PGRP_UNLOCK(savepgrp);
480         if (LIST_EMPTY(&savepgrp->pg_members))
481                 pgdelete(savepgrp);
482         return (0);
483 }
484
485 /*
486  * delete a process group
487  */
488 static void
489 pgdelete(pgrp)
490         register struct pgrp *pgrp;
491 {
492         struct session *savesess;
493         struct tty *tp;
494
495         sx_assert(&proctree_lock, SX_XLOCKED);
496         PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
497         SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
498
499         /*
500          * Reset any sigio structures pointing to us as a result of
501          * F_SETOWN with our pgid.
502          */
503         funsetownlst(&pgrp->pg_sigiolst);
504
505         PGRP_LOCK(pgrp);
506         tp = pgrp->pg_session->s_ttyp;
507         LIST_REMOVE(pgrp, pg_hash);
508         savesess = pgrp->pg_session;
509         PGRP_UNLOCK(pgrp);
510
511         /* Remove the reference to the pgrp before deallocating it. */
512         if (tp != NULL) {
513                 tty_lock(tp);
514                 tty_rel_pgrp(tp, pgrp);
515         }
516
517         mtx_destroy(&pgrp->pg_mtx);
518         free(pgrp, M_PGRP);
519         sess_release(savesess);
520 }
521
522 static void
523 pgadjustjobc(pgrp, entering)
524         struct pgrp *pgrp;
525         int entering;
526 {
527
528         PGRP_LOCK(pgrp);
529         if (entering)
530                 pgrp->pg_jobc++;
531         else {
532                 --pgrp->pg_jobc;
533                 if (pgrp->pg_jobc == 0)
534                         orphanpg(pgrp);
535         }
536         PGRP_UNLOCK(pgrp);
537 }
538
539 /*
540  * Adjust pgrp jobc counters when specified process changes process group.
541  * We count the number of processes in each process group that "qualify"
542  * the group for terminal job control (those with a parent in a different
543  * process group of the same session).  If that count reaches zero, the
544  * process group becomes orphaned.  Check both the specified process'
545  * process group and that of its children.
546  * entering == 0 => p is leaving specified group.
547  * entering == 1 => p is entering specified group.
548  */
549 void
550 fixjobc(p, pgrp, entering)
551         register struct proc *p;
552         register struct pgrp *pgrp;
553         int entering;
554 {
555         register struct pgrp *hispgrp;
556         register struct session *mysession;
557
558         sx_assert(&proctree_lock, SX_LOCKED);
559         PROC_LOCK_ASSERT(p, MA_NOTOWNED);
560         PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
561         SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
562
563         /*
564          * Check p's parent to see whether p qualifies its own process
565          * group; if so, adjust count for p's process group.
566          */
567         mysession = pgrp->pg_session;
568         if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
569             hispgrp->pg_session == mysession)
570                 pgadjustjobc(pgrp, entering);
571
572         /*
573          * Check this process' children to see whether they qualify
574          * their process groups; if so, adjust counts for children's
575          * process groups.
576          */
577         LIST_FOREACH(p, &p->p_children, p_sibling) {
578                 hispgrp = p->p_pgrp;
579                 if (hispgrp == pgrp ||
580                     hispgrp->pg_session != mysession)
581                         continue;
582                 PROC_LOCK(p);
583                 if (p->p_state == PRS_ZOMBIE) {
584                         PROC_UNLOCK(p);
585                         continue;
586                 }
587                 PROC_UNLOCK(p);
588                 pgadjustjobc(hispgrp, entering);
589         }
590 }
591
592 /*
593  * A process group has become orphaned;
594  * if there are any stopped processes in the group,
595  * hang-up all process in that group.
596  */
597 static void
598 orphanpg(pg)
599         struct pgrp *pg;
600 {
601         register struct proc *p;
602
603         PGRP_LOCK_ASSERT(pg, MA_OWNED);
604
605         LIST_FOREACH(p, &pg->pg_members, p_pglist) {
606                 PROC_LOCK(p);
607                 if (P_SHOULDSTOP(p)) {
608                         PROC_UNLOCK(p);
609                         LIST_FOREACH(p, &pg->pg_members, p_pglist) {
610                                 PROC_LOCK(p);
611                                 psignal(p, SIGHUP);
612                                 psignal(p, SIGCONT);
613                                 PROC_UNLOCK(p);
614                         }
615                         return;
616                 }
617                 PROC_UNLOCK(p);
618         }
619 }
620
621 void
622 sess_hold(struct session *s)
623 {
624
625         refcount_acquire(&s->s_count);
626 }
627
628 void
629 sess_release(struct session *s)
630 {
631
632         if (refcount_release(&s->s_count)) {
633                 if (s->s_ttyp != NULL) {
634                         tty_lock(s->s_ttyp);
635                         tty_rel_sess(s->s_ttyp, s);
636                 }
637                 mtx_destroy(&s->s_mtx);
638                 free(s, M_SESSION);
639         }
640 }
641
642 #include "opt_ddb.h"
643 #ifdef DDB
644 #include <ddb/ddb.h>
645
646 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
647 {
648         register struct pgrp *pgrp;
649         register struct proc *p;
650         register int i;
651
652         for (i = 0; i <= pgrphash; i++) {
653                 if (!LIST_EMPTY(&pgrphashtbl[i])) {
654                         printf("\tindx %d\n", i);
655                         LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
656                                 printf(
657                         "\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
658                                     (void *)pgrp, (long)pgrp->pg_id,
659                                     (void *)pgrp->pg_session,
660                                     pgrp->pg_session->s_count,
661                                     (void *)LIST_FIRST(&pgrp->pg_members));
662                                 LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
663                                         printf("\t\tpid %ld addr %p pgrp %p\n", 
664                                             (long)p->p_pid, (void *)p,
665                                             (void *)p->p_pgrp);
666                                 }
667                         }
668                 }
669         }
670 }
671 #endif /* DDB */
672
673 /*
674  * Calculate the kinfo_proc members which contain process-wide
675  * informations.
676  * Must be called with the target process locked.
677  */
678 static void
679 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
680 {
681         struct thread *td;
682
683         PROC_LOCK_ASSERT(p, MA_OWNED);
684
685         kp->ki_estcpu = 0;
686         kp->ki_pctcpu = 0;
687         FOREACH_THREAD_IN_PROC(p, td) {
688                 thread_lock(td);
689                 kp->ki_pctcpu += sched_pctcpu(td);
690                 kp->ki_estcpu += td->td_estcpu;
691                 thread_unlock(td);
692         }
693 }
694
695 /*
696  * Clear kinfo_proc and fill in any information that is common
697  * to all threads in the process.
698  * Must be called with the target process locked.
699  */
700 static void
701 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
702 {
703         struct thread *td0;
704         struct tty *tp;
705         struct session *sp;
706         struct ucred *cred;
707         struct sigacts *ps;
708
709         PROC_LOCK_ASSERT(p, MA_OWNED);
710         bzero(kp, sizeof(*kp));
711
712         kp->ki_structsize = sizeof(*kp);
713         kp->ki_paddr = p;
714         kp->ki_addr =/* p->p_addr; */0; /* XXX */
715         kp->ki_args = p->p_args;
716         kp->ki_textvp = p->p_textvp;
717 #ifdef KTRACE
718         kp->ki_tracep = p->p_tracevp;
719         mtx_lock(&ktrace_mtx);
720         kp->ki_traceflag = p->p_traceflag;
721         mtx_unlock(&ktrace_mtx);
722 #endif
723         kp->ki_fd = p->p_fd;
724         kp->ki_vmspace = p->p_vmspace;
725         kp->ki_flag = p->p_flag;
726         cred = p->p_ucred;
727         if (cred) {
728                 kp->ki_uid = cred->cr_uid;
729                 kp->ki_ruid = cred->cr_ruid;
730                 kp->ki_svuid = cred->cr_svuid;
731                 kp->ki_cr_flags = cred->cr_flags;
732                 /* XXX bde doesn't like KI_NGROUPS */
733                 if (cred->cr_ngroups > KI_NGROUPS) {
734                         kp->ki_ngroups = KI_NGROUPS;
735                         kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
736                 } else
737                         kp->ki_ngroups = cred->cr_ngroups;
738                 bcopy(cred->cr_groups, kp->ki_groups,
739                     kp->ki_ngroups * sizeof(gid_t));
740                 kp->ki_rgid = cred->cr_rgid;
741                 kp->ki_svgid = cred->cr_svgid;
742                 /* If jailed(cred), emulate the old P_JAILED flag. */
743                 if (jailed(cred)) {
744                         kp->ki_flag |= P_JAILED;
745                         /* If inside the jail, use 0 as a jail ID. */
746                         if (cred->cr_prison != curthread->td_ucred->cr_prison)
747                                 kp->ki_jid = cred->cr_prison->pr_id;
748                 }
749         }
750         ps = p->p_sigacts;
751         if (ps) {
752                 mtx_lock(&ps->ps_mtx);
753                 kp->ki_sigignore = ps->ps_sigignore;
754                 kp->ki_sigcatch = ps->ps_sigcatch;
755                 mtx_unlock(&ps->ps_mtx);
756         }
757         PROC_SLOCK(p);
758         if (p->p_state != PRS_NEW &&
759             p->p_state != PRS_ZOMBIE &&
760             p->p_vmspace != NULL) {
761                 struct vmspace *vm = p->p_vmspace;
762
763                 kp->ki_size = vm->vm_map.size;
764                 kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
765                 FOREACH_THREAD_IN_PROC(p, td0) {
766                         if (!TD_IS_SWAPPED(td0))
767                                 kp->ki_rssize += td0->td_kstack_pages;
768                 }
769                 kp->ki_swrss = vm->vm_swrss;
770                 kp->ki_tsize = vm->vm_tsize;
771                 kp->ki_dsize = vm->vm_dsize;
772                 kp->ki_ssize = vm->vm_ssize;
773         } else if (p->p_state == PRS_ZOMBIE)
774                 kp->ki_stat = SZOMB;
775         if (kp->ki_flag & P_INMEM)
776                 kp->ki_sflag = PS_INMEM;
777         else
778                 kp->ki_sflag = 0;
779         /* Calculate legacy swtime as seconds since 'swtick'. */
780         kp->ki_swtime = (ticks - p->p_swtick) / hz;
781         kp->ki_pid = p->p_pid;
782         kp->ki_nice = p->p_nice;
783         rufetch(p, &kp->ki_rusage);
784         kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
785         PROC_SUNLOCK(p);
786         if ((p->p_flag & P_INMEM) && p->p_stats != NULL) {
787                 kp->ki_start = p->p_stats->p_start;
788                 timevaladd(&kp->ki_start, &boottime);
789                 PROC_SLOCK(p);
790                 calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
791                 PROC_SUNLOCK(p);
792                 calccru(p, &kp->ki_childutime, &kp->ki_childstime);
793
794                 /* Some callers want child-times in a single value */
795                 kp->ki_childtime = kp->ki_childstime;
796                 timevaladd(&kp->ki_childtime, &kp->ki_childutime);
797         }
798         tp = NULL;
799         if (p->p_pgrp) {
800                 kp->ki_pgid = p->p_pgrp->pg_id;
801                 kp->ki_jobc = p->p_pgrp->pg_jobc;
802                 sp = p->p_pgrp->pg_session;
803
804                 if (sp != NULL) {
805                         kp->ki_sid = sp->s_sid;
806                         SESS_LOCK(sp);
807                         strlcpy(kp->ki_login, sp->s_login,
808                             sizeof(kp->ki_login));
809                         if (sp->s_ttyvp)
810                                 kp->ki_kiflag |= KI_CTTY;
811                         if (SESS_LEADER(p))
812                                 kp->ki_kiflag |= KI_SLEADER;
813                         /* XXX proctree_lock */
814                         tp = sp->s_ttyp;
815                         SESS_UNLOCK(sp);
816                 }
817         }
818         if ((p->p_flag & P_CONTROLT) && tp != NULL) {
819                 kp->ki_tdev = tty_udev(tp);
820                 kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
821                 if (tp->t_session)
822                         kp->ki_tsid = tp->t_session->s_sid;
823         } else
824                 kp->ki_tdev = NODEV;
825         if (p->p_comm[0] != '\0')
826                 strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
827         if (p->p_sysent && p->p_sysent->sv_name != NULL &&
828             p->p_sysent->sv_name[0] != '\0')
829                 strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
830         kp->ki_siglist = p->p_siglist;
831         kp->ki_xstat = p->p_xstat;
832         kp->ki_acflag = p->p_acflag;
833         kp->ki_lock = p->p_lock;
834         if (p->p_pptr)
835                 kp->ki_ppid = p->p_pptr->p_pid;
836 }
837
838 /*
839  * Fill in information that is thread specific.  Must be called with
840  * target process locked.  If 'preferthread' is set, overwrite certain
841  * process-related fields that are maintained for both threads and
842  * processes.
843  */
844 static void
845 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
846 {
847         struct proc *p;
848
849         p = td->td_proc;
850         PROC_LOCK_ASSERT(p, MA_OWNED);
851
852         thread_lock(td);
853         if (td->td_wmesg != NULL)
854                 strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
855         else
856                 bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
857         strlcpy(kp->ki_ocomm, td->td_name, sizeof(kp->ki_ocomm));
858         if (TD_ON_LOCK(td)) {
859                 kp->ki_kiflag |= KI_LOCKBLOCK;
860                 strlcpy(kp->ki_lockname, td->td_lockname,
861                     sizeof(kp->ki_lockname));
862         } else {
863                 kp->ki_kiflag &= ~KI_LOCKBLOCK;
864                 bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
865         }
866
867         if (p->p_state == PRS_NORMAL) { /* approximate. */
868                 if (TD_ON_RUNQ(td) ||
869                     TD_CAN_RUN(td) ||
870                     TD_IS_RUNNING(td)) {
871                         kp->ki_stat = SRUN;
872                 } else if (P_SHOULDSTOP(p)) {
873                         kp->ki_stat = SSTOP;
874                 } else if (TD_IS_SLEEPING(td)) {
875                         kp->ki_stat = SSLEEP;
876                 } else if (TD_ON_LOCK(td)) {
877                         kp->ki_stat = SLOCK;
878                 } else {
879                         kp->ki_stat = SWAIT;
880                 }
881         } else if (p->p_state == PRS_ZOMBIE) {
882                 kp->ki_stat = SZOMB;
883         } else {
884                 kp->ki_stat = SIDL;
885         }
886
887         /* Things in the thread */
888         kp->ki_wchan = td->td_wchan;
889         kp->ki_pri.pri_level = td->td_priority;
890         kp->ki_pri.pri_native = td->td_base_pri;
891         kp->ki_lastcpu = td->td_lastcpu;
892         kp->ki_oncpu = td->td_oncpu;
893         kp->ki_tdflags = td->td_flags;
894         kp->ki_tid = td->td_tid;
895         kp->ki_numthreads = p->p_numthreads;
896         kp->ki_pcb = td->td_pcb;
897         kp->ki_kstack = (void *)td->td_kstack;
898         kp->ki_slptime = (ticks - td->td_slptick) / hz;
899         kp->ki_pri.pri_class = td->td_pri_class;
900         kp->ki_pri.pri_user = td->td_user_pri;
901
902         if (preferthread) {
903                 kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
904                 kp->ki_pctcpu = sched_pctcpu(td);
905                 kp->ki_estcpu = td->td_estcpu;
906         }
907
908         /* We can't get this anymore but ps etc never used it anyway. */
909         kp->ki_rqindex = 0;
910
911         if (preferthread)
912                 kp->ki_siglist = td->td_siglist;
913         kp->ki_sigmask = td->td_sigmask;
914         thread_unlock(td);
915 }
916
917 /*
918  * Fill in a kinfo_proc structure for the specified process.
919  * Must be called with the target process locked.
920  */
921 void
922 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
923 {
924
925         MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
926
927         fill_kinfo_proc_only(p, kp);
928         fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
929         fill_kinfo_aggregate(p, kp);
930 }
931
932 struct pstats *
933 pstats_alloc(void)
934 {
935
936         return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
937 }
938
939 /*
940  * Copy parts of p_stats; zero the rest of p_stats (statistics).
941  */
942 void
943 pstats_fork(struct pstats *src, struct pstats *dst)
944 {
945
946         bzero(&dst->pstat_startzero,
947             __rangeof(struct pstats, pstat_startzero, pstat_endzero));
948         bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
949             __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
950 }
951
952 void
953 pstats_free(struct pstats *ps)
954 {
955
956         free(ps, M_SUBPROC);
957 }
958
959 /*
960  * Locate a zombie process by number
961  */
962 struct proc *
963 zpfind(pid_t pid)
964 {
965         struct proc *p;
966
967         sx_slock(&allproc_lock);
968         LIST_FOREACH(p, &zombproc, p_list)
969                 if (p->p_pid == pid) {
970                         PROC_LOCK(p);
971                         break;
972                 }
973         sx_sunlock(&allproc_lock);
974         return (p);
975 }
976
977 #define KERN_PROC_ZOMBMASK      0x3
978 #define KERN_PROC_NOTHREADS     0x4
979
980 #ifdef COMPAT_FREEBSD32
981
982 /*
983  * This function is typically used to copy out the kernel address, so
984  * it can be replaced by assignment of zero.
985  */
986 static inline uint32_t
987 ptr32_trim(void *ptr)
988 {
989         uintptr_t uptr;
990
991         uptr = (uintptr_t)ptr;
992         return ((uptr > UINT_MAX) ? 0 : uptr);
993 }
994
995 #define PTRTRIM_CP(src,dst,fld) \
996         do { (dst).fld = ptr32_trim((src).fld); } while (0)
997
998 static void
999 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1000 {
1001         int i;
1002
1003         bzero(ki32, sizeof(struct kinfo_proc32));
1004         ki32->ki_structsize = sizeof(struct kinfo_proc32);
1005         CP(*ki, *ki32, ki_layout);
1006         PTRTRIM_CP(*ki, *ki32, ki_args);
1007         PTRTRIM_CP(*ki, *ki32, ki_paddr);
1008         PTRTRIM_CP(*ki, *ki32, ki_addr);
1009         PTRTRIM_CP(*ki, *ki32, ki_tracep);
1010         PTRTRIM_CP(*ki, *ki32, ki_textvp);
1011         PTRTRIM_CP(*ki, *ki32, ki_fd);
1012         PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1013         PTRTRIM_CP(*ki, *ki32, ki_wchan);
1014         CP(*ki, *ki32, ki_pid);
1015         CP(*ki, *ki32, ki_ppid);
1016         CP(*ki, *ki32, ki_pgid);
1017         CP(*ki, *ki32, ki_tpgid);
1018         CP(*ki, *ki32, ki_sid);
1019         CP(*ki, *ki32, ki_tsid);
1020         CP(*ki, *ki32, ki_jobc);
1021         CP(*ki, *ki32, ki_tdev);
1022         CP(*ki, *ki32, ki_siglist);
1023         CP(*ki, *ki32, ki_sigmask);
1024         CP(*ki, *ki32, ki_sigignore);
1025         CP(*ki, *ki32, ki_sigcatch);
1026         CP(*ki, *ki32, ki_uid);
1027         CP(*ki, *ki32, ki_ruid);
1028         CP(*ki, *ki32, ki_svuid);
1029         CP(*ki, *ki32, ki_rgid);
1030         CP(*ki, *ki32, ki_svgid);
1031         CP(*ki, *ki32, ki_ngroups);
1032         for (i = 0; i < KI_NGROUPS; i++)
1033                 CP(*ki, *ki32, ki_groups[i]);
1034         CP(*ki, *ki32, ki_size);
1035         CP(*ki, *ki32, ki_rssize);
1036         CP(*ki, *ki32, ki_swrss);
1037         CP(*ki, *ki32, ki_tsize);
1038         CP(*ki, *ki32, ki_dsize);
1039         CP(*ki, *ki32, ki_ssize);
1040         CP(*ki, *ki32, ki_xstat);
1041         CP(*ki, *ki32, ki_acflag);
1042         CP(*ki, *ki32, ki_pctcpu);
1043         CP(*ki, *ki32, ki_estcpu);
1044         CP(*ki, *ki32, ki_slptime);
1045         CP(*ki, *ki32, ki_swtime);
1046         CP(*ki, *ki32, ki_runtime);
1047         TV_CP(*ki, *ki32, ki_start);
1048         TV_CP(*ki, *ki32, ki_childtime);
1049         CP(*ki, *ki32, ki_flag);
1050         CP(*ki, *ki32, ki_kiflag);
1051         CP(*ki, *ki32, ki_traceflag);
1052         CP(*ki, *ki32, ki_stat);
1053         CP(*ki, *ki32, ki_nice);
1054         CP(*ki, *ki32, ki_lock);
1055         CP(*ki, *ki32, ki_rqindex);
1056         CP(*ki, *ki32, ki_oncpu);
1057         CP(*ki, *ki32, ki_lastcpu);
1058         bcopy(ki->ki_ocomm, ki32->ki_ocomm, OCOMMLEN + 1);
1059         bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1060         bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1061         bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1062         bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1063         bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1064         CP(*ki, *ki32, ki_cr_flags);
1065         CP(*ki, *ki32, ki_jid);
1066         CP(*ki, *ki32, ki_numthreads);
1067         CP(*ki, *ki32, ki_tid);
1068         CP(*ki, *ki32, ki_pri);
1069         freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1070         freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1071         PTRTRIM_CP(*ki, *ki32, ki_pcb);
1072         PTRTRIM_CP(*ki, *ki32, ki_kstack);
1073         PTRTRIM_CP(*ki, *ki32, ki_udata);
1074         CP(*ki, *ki32, ki_sflag);
1075         CP(*ki, *ki32, ki_tdflags);
1076 }
1077
1078 static int
1079 sysctl_out_proc_copyout(struct kinfo_proc *ki, struct sysctl_req *req)
1080 {
1081         struct kinfo_proc32 ki32;
1082         int error;
1083
1084         if (req->flags & SCTL_MASK32) {
1085                 freebsd32_kinfo_proc_out(ki, &ki32);
1086                 error = SYSCTL_OUT(req, &ki32, sizeof(struct kinfo_proc32));
1087         } else
1088                 error = SYSCTL_OUT(req, ki, sizeof(struct kinfo_proc));
1089         return (error);
1090 }
1091 #else
1092 static int
1093 sysctl_out_proc_copyout(struct kinfo_proc *ki, struct sysctl_req *req)
1094 {
1095
1096         return (SYSCTL_OUT(req, ki, sizeof(struct kinfo_proc)));
1097 }
1098 #endif
1099
1100 /*
1101  * Must be called with the process locked and will return with it unlocked.
1102  */
1103 static int
1104 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
1105 {
1106         struct thread *td;
1107         struct kinfo_proc kinfo_proc;
1108         int error = 0;
1109         struct proc *np;
1110         pid_t pid = p->p_pid;
1111
1112         PROC_LOCK_ASSERT(p, MA_OWNED);
1113         MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1114
1115         fill_kinfo_proc(p, &kinfo_proc);
1116         if (flags & KERN_PROC_NOTHREADS)
1117                 error = sysctl_out_proc_copyout(&kinfo_proc, req);
1118         else {
1119                 FOREACH_THREAD_IN_PROC(p, td) {
1120                         fill_kinfo_thread(td, &kinfo_proc, 1);
1121                         error = sysctl_out_proc_copyout(&kinfo_proc, req);
1122                         if (error)
1123                                 break;
1124                 }
1125         }
1126         PROC_UNLOCK(p);
1127         if (error)
1128                 return (error);
1129         if (flags & KERN_PROC_ZOMBMASK)
1130                 np = zpfind(pid);
1131         else {
1132                 if (pid == 0)
1133                         return (0);
1134                 np = pfind(pid);
1135         }
1136         if (np == NULL)
1137                 return (ESRCH);
1138         if (np != p) {
1139                 PROC_UNLOCK(np);
1140                 return (ESRCH);
1141         }
1142         PROC_UNLOCK(np);
1143         return (0);
1144 }
1145
1146 static int
1147 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1148 {
1149         int *name = (int*) arg1;
1150         u_int namelen = arg2;
1151         struct proc *p;
1152         int flags, doingzomb, oid_number;
1153         int error = 0;
1154
1155         oid_number = oidp->oid_number;
1156         if (oid_number != KERN_PROC_ALL &&
1157             (oid_number & KERN_PROC_INC_THREAD) == 0)
1158                 flags = KERN_PROC_NOTHREADS;
1159         else {
1160                 flags = 0;
1161                 oid_number &= ~KERN_PROC_INC_THREAD;
1162         }
1163         if (oid_number == KERN_PROC_PID) {
1164                 if (namelen != 1) 
1165                         return (EINVAL);
1166                 error = sysctl_wire_old_buffer(req, 0);
1167                 if (error)
1168                         return (error);         
1169                 p = pfind((pid_t)name[0]);
1170                 if (!p)
1171                         return (ESRCH);
1172                 if ((error = p_cansee(curthread, p))) {
1173                         PROC_UNLOCK(p);
1174                         return (error);
1175                 }
1176                 error = sysctl_out_proc(p, req, flags);
1177                 return (error);
1178         }
1179
1180         switch (oid_number) {
1181         case KERN_PROC_ALL:
1182                 if (namelen != 0)
1183                         return (EINVAL);
1184                 break;
1185         case KERN_PROC_PROC:
1186                 if (namelen != 0 && namelen != 1)
1187                         return (EINVAL);
1188                 break;
1189         default:
1190                 if (namelen != 1)
1191                         return (EINVAL);
1192                 break;
1193         }
1194         
1195         if (!req->oldptr) {
1196                 /* overestimate by 5 procs */
1197                 error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1198                 if (error)
1199                         return (error);
1200         }
1201         error = sysctl_wire_old_buffer(req, 0);
1202         if (error != 0)
1203                 return (error);
1204         sx_slock(&allproc_lock);
1205         for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) {
1206                 if (!doingzomb)
1207                         p = LIST_FIRST(&allproc);
1208                 else
1209                         p = LIST_FIRST(&zombproc);
1210                 for (; p != 0; p = LIST_NEXT(p, p_list)) {
1211                         /*
1212                          * Skip embryonic processes.
1213                          */
1214                         PROC_SLOCK(p);
1215                         if (p->p_state == PRS_NEW) {
1216                                 PROC_SUNLOCK(p);
1217                                 continue;
1218                         }
1219                         PROC_SUNLOCK(p);
1220                         PROC_LOCK(p);
1221                         KASSERT(p->p_ucred != NULL,
1222                             ("process credential is NULL for non-NEW proc"));
1223                         /*
1224                          * Show a user only appropriate processes.
1225                          */
1226                         if (p_cansee(curthread, p)) {
1227                                 PROC_UNLOCK(p);
1228                                 continue;
1229                         }
1230                         /*
1231                          * TODO - make more efficient (see notes below).
1232                          * do by session.
1233                          */
1234                         switch (oid_number) {
1235
1236                         case KERN_PROC_GID:
1237                                 if (p->p_ucred->cr_gid != (gid_t)name[0]) {
1238                                         PROC_UNLOCK(p);
1239                                         continue;
1240                                 }
1241                                 break;
1242
1243                         case KERN_PROC_PGRP:
1244                                 /* could do this by traversing pgrp */
1245                                 if (p->p_pgrp == NULL ||
1246                                     p->p_pgrp->pg_id != (pid_t)name[0]) {
1247                                         PROC_UNLOCK(p);
1248                                         continue;
1249                                 }
1250                                 break;
1251
1252                         case KERN_PROC_RGID:
1253                                 if (p->p_ucred->cr_rgid != (gid_t)name[0]) {
1254                                         PROC_UNLOCK(p);
1255                                         continue;
1256                                 }
1257                                 break;
1258
1259                         case KERN_PROC_SESSION:
1260                                 if (p->p_session == NULL ||
1261                                     p->p_session->s_sid != (pid_t)name[0]) {
1262                                         PROC_UNLOCK(p);
1263                                         continue;
1264                                 }
1265                                 break;
1266
1267                         case KERN_PROC_TTY:
1268                                 if ((p->p_flag & P_CONTROLT) == 0 ||
1269                                     p->p_session == NULL) {
1270                                         PROC_UNLOCK(p);
1271                                         continue;
1272                                 }
1273                                 /* XXX proctree_lock */
1274                                 SESS_LOCK(p->p_session);
1275                                 if (p->p_session->s_ttyp == NULL ||
1276                                     tty_udev(p->p_session->s_ttyp) != 
1277                                     (dev_t)name[0]) {
1278                                         SESS_UNLOCK(p->p_session);
1279                                         PROC_UNLOCK(p);
1280                                         continue;
1281                                 }
1282                                 SESS_UNLOCK(p->p_session);
1283                                 break;
1284
1285                         case KERN_PROC_UID:
1286                                 if (p->p_ucred->cr_uid != (uid_t)name[0]) {
1287                                         PROC_UNLOCK(p);
1288                                         continue;
1289                                 }
1290                                 break;
1291
1292                         case KERN_PROC_RUID:
1293                                 if (p->p_ucred->cr_ruid != (uid_t)name[0]) {
1294                                         PROC_UNLOCK(p);
1295                                         continue;
1296                                 }
1297                                 break;
1298
1299                         case KERN_PROC_PROC:
1300                                 break;
1301
1302                         default:
1303                                 break;
1304
1305                         }
1306
1307                         error = sysctl_out_proc(p, req, flags | doingzomb);
1308                         if (error) {
1309                                 sx_sunlock(&allproc_lock);
1310                                 return (error);
1311                         }
1312                 }
1313         }
1314         sx_sunlock(&allproc_lock);
1315         return (0);
1316 }
1317
1318 struct pargs *
1319 pargs_alloc(int len)
1320 {
1321         struct pargs *pa;
1322
1323         pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1324                 M_WAITOK);
1325         refcount_init(&pa->ar_ref, 1);
1326         pa->ar_length = len;
1327         return (pa);
1328 }
1329
1330 static void
1331 pargs_free(struct pargs *pa)
1332 {
1333
1334         free(pa, M_PARGS);
1335 }
1336
1337 void
1338 pargs_hold(struct pargs *pa)
1339 {
1340
1341         if (pa == NULL)
1342                 return;
1343         refcount_acquire(&pa->ar_ref);
1344 }
1345
1346 void
1347 pargs_drop(struct pargs *pa)
1348 {
1349
1350         if (pa == NULL)
1351                 return;
1352         if (refcount_release(&pa->ar_ref))
1353                 pargs_free(pa);
1354 }
1355
1356 /*
1357  * This sysctl allows a process to retrieve the argument list or process
1358  * title for another process without groping around in the address space
1359  * of the other process.  It also allow a process to set its own "process 
1360  * title to a string of its own choice.
1361  */
1362 static int
1363 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
1364 {
1365         int *name = (int*) arg1;
1366         u_int namelen = arg2;
1367         struct pargs *newpa, *pa;
1368         struct proc *p;
1369         int error = 0;
1370
1371         if (namelen != 1) 
1372                 return (EINVAL);
1373
1374         p = pfind((pid_t)name[0]);
1375         if (!p)
1376                 return (ESRCH);
1377
1378         if ((error = p_cansee(curthread, p)) != 0) {
1379                 PROC_UNLOCK(p);
1380                 return (error);
1381         }
1382
1383         if (req->newptr && curproc != p) {
1384                 PROC_UNLOCK(p);
1385                 return (EPERM);
1386         }
1387
1388         pa = p->p_args;
1389         pargs_hold(pa);
1390         PROC_UNLOCK(p);
1391         if (req->oldptr != NULL && pa != NULL)
1392                 error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
1393         pargs_drop(pa);
1394         if (error != 0 || req->newptr == NULL)
1395                 return (error);
1396
1397         if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit)
1398                 return (ENOMEM);
1399         newpa = pargs_alloc(req->newlen);
1400         error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
1401         if (error != 0) {
1402                 pargs_free(newpa);
1403                 return (error);
1404         }
1405         PROC_LOCK(p);
1406         pa = p->p_args;
1407         p->p_args = newpa;
1408         PROC_UNLOCK(p);
1409         pargs_drop(pa);
1410         return (0);
1411 }
1412
1413 /*
1414  * This sysctl allows a process to retrieve the path of the executable for
1415  * itself or another process.
1416  */
1417 static int
1418 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
1419 {
1420         pid_t *pidp = (pid_t *)arg1;
1421         unsigned int arglen = arg2;
1422         struct proc *p;
1423         struct vnode *vp;
1424         char *retbuf, *freebuf;
1425         int error, vfslocked;
1426
1427         if (arglen != 1)
1428                 return (EINVAL);
1429         if (*pidp == -1) {      /* -1 means this process */
1430                 p = req->td->td_proc;
1431         } else {
1432                 p = pfind(*pidp);
1433                 if (p == NULL)
1434                         return (ESRCH);
1435                 if ((error = p_cansee(curthread, p)) != 0) {
1436                         PROC_UNLOCK(p);
1437                         return (error);
1438                 }
1439         }
1440
1441         vp = p->p_textvp;
1442         if (vp == NULL) {
1443                 if (*pidp != -1)
1444                         PROC_UNLOCK(p);
1445                 return (0);
1446         }
1447         vref(vp);
1448         if (*pidp != -1)
1449                 PROC_UNLOCK(p);
1450         error = vn_fullpath(req->td, vp, &retbuf, &freebuf);
1451         vfslocked = VFS_LOCK_GIANT(vp->v_mount);
1452         vrele(vp);
1453         VFS_UNLOCK_GIANT(vfslocked);
1454         if (error)
1455                 return (error);
1456         error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
1457         free(freebuf, M_TEMP);
1458         return (error);
1459 }
1460
1461 static int
1462 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
1463 {
1464         struct proc *p;
1465         char *sv_name;
1466         int *name;
1467         int namelen;
1468         int error;
1469
1470         namelen = arg2;
1471         if (namelen != 1) 
1472                 return (EINVAL);
1473
1474         name = (int *)arg1;
1475         if ((p = pfind((pid_t)name[0])) == NULL)
1476                 return (ESRCH);
1477         if ((error = p_cansee(curthread, p))) {
1478                 PROC_UNLOCK(p);
1479                 return (error);
1480         }
1481         sv_name = p->p_sysent->sv_name;
1482         PROC_UNLOCK(p);
1483         return (sysctl_handle_string(oidp, sv_name, 0, req));
1484 }
1485
1486 #ifdef KINFO_OVMENTRY_SIZE
1487 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
1488 #endif
1489
1490 #ifdef COMPAT_FREEBSD7
1491 static int
1492 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
1493 {
1494         vm_map_entry_t entry, tmp_entry;
1495         unsigned int last_timestamp;
1496         char *fullpath, *freepath;
1497         struct kinfo_ovmentry *kve;
1498         struct vattr va;
1499         struct ucred *cred;
1500         int error, *name;
1501         struct vnode *vp;
1502         struct proc *p;
1503         vm_map_t map;
1504         struct vmspace *vm;
1505
1506         name = (int *)arg1;
1507         if ((p = pfind((pid_t)name[0])) == NULL)
1508                 return (ESRCH);
1509         if (p->p_flag & P_WEXIT) {
1510                 PROC_UNLOCK(p);
1511                 return (ESRCH);
1512         }
1513         if ((error = p_candebug(curthread, p))) {
1514                 PROC_UNLOCK(p);
1515                 return (error);
1516         }
1517         _PHOLD(p);
1518         PROC_UNLOCK(p);
1519         vm = vmspace_acquire_ref(p);
1520         if (vm == NULL) {
1521                 PRELE(p);
1522                 return (ESRCH);
1523         }
1524         kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
1525
1526         map = &p->p_vmspace->vm_map;    /* XXXRW: More locking required? */
1527         vm_map_lock_read(map);
1528         for (entry = map->header.next; entry != &map->header;
1529             entry = entry->next) {
1530                 vm_object_t obj, tobj, lobj;
1531                 vm_offset_t addr;
1532                 int vfslocked;
1533
1534                 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
1535                         continue;
1536
1537                 bzero(kve, sizeof(*kve));
1538                 kve->kve_structsize = sizeof(*kve);
1539
1540                 kve->kve_private_resident = 0;
1541                 obj = entry->object.vm_object;
1542                 if (obj != NULL) {
1543                         VM_OBJECT_LOCK(obj);
1544                         if (obj->shadow_count == 1)
1545                                 kve->kve_private_resident =
1546                                     obj->resident_page_count;
1547                 }
1548                 kve->kve_resident = 0;
1549                 addr = entry->start;
1550                 while (addr < entry->end) {
1551                         if (pmap_extract(map->pmap, addr))
1552                                 kve->kve_resident++;
1553                         addr += PAGE_SIZE;
1554                 }
1555
1556                 for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
1557                         if (tobj != obj)
1558                                 VM_OBJECT_LOCK(tobj);
1559                         if (lobj != obj)
1560                                 VM_OBJECT_UNLOCK(lobj);
1561                         lobj = tobj;
1562                 }
1563
1564                 kve->kve_start = (void*)entry->start;
1565                 kve->kve_end = (void*)entry->end;
1566                 kve->kve_offset = (off_t)entry->offset;
1567
1568                 if (entry->protection & VM_PROT_READ)
1569                         kve->kve_protection |= KVME_PROT_READ;
1570                 if (entry->protection & VM_PROT_WRITE)
1571                         kve->kve_protection |= KVME_PROT_WRITE;
1572                 if (entry->protection & VM_PROT_EXECUTE)
1573                         kve->kve_protection |= KVME_PROT_EXEC;
1574
1575                 if (entry->eflags & MAP_ENTRY_COW)
1576                         kve->kve_flags |= KVME_FLAG_COW;
1577                 if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
1578                         kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
1579
1580                 last_timestamp = map->timestamp;
1581                 vm_map_unlock_read(map);
1582
1583                 kve->kve_fileid = 0;
1584                 kve->kve_fsid = 0;
1585                 freepath = NULL;
1586                 fullpath = "";
1587                 if (lobj) {
1588                         vp = NULL;
1589                         switch (lobj->type) {
1590                         case OBJT_DEFAULT:
1591                                 kve->kve_type = KVME_TYPE_DEFAULT;
1592                                 break;
1593                         case OBJT_VNODE:
1594                                 kve->kve_type = KVME_TYPE_VNODE;
1595                                 vp = lobj->handle;
1596                                 vref(vp);
1597                                 break;
1598                         case OBJT_SWAP:
1599                                 kve->kve_type = KVME_TYPE_SWAP;
1600                                 break;
1601                         case OBJT_DEVICE:
1602                                 kve->kve_type = KVME_TYPE_DEVICE;
1603                                 break;
1604                         case OBJT_PHYS:
1605                                 kve->kve_type = KVME_TYPE_PHYS;
1606                                 break;
1607                         case OBJT_DEAD:
1608                                 kve->kve_type = KVME_TYPE_DEAD;
1609                                 break;
1610                         case OBJT_SG:
1611                                 kve->kve_type = KVME_TYPE_SG;
1612                                 break;
1613                         default:
1614                                 kve->kve_type = KVME_TYPE_UNKNOWN;
1615                                 break;
1616                         }
1617                         if (lobj != obj)
1618                                 VM_OBJECT_UNLOCK(lobj);
1619
1620                         kve->kve_ref_count = obj->ref_count;
1621                         kve->kve_shadow_count = obj->shadow_count;
1622                         VM_OBJECT_UNLOCK(obj);
1623                         if (vp != NULL) {
1624                                 vn_fullpath(curthread, vp, &fullpath,
1625                                     &freepath);
1626                                 cred = curthread->td_ucred;
1627                                 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
1628                                 vn_lock(vp, LK_SHARED | LK_RETRY);
1629                                 if (VOP_GETATTR(vp, &va, cred) == 0) {
1630                                         kve->kve_fileid = va.va_fileid;
1631                                         kve->kve_fsid = va.va_fsid;
1632                                 }
1633                                 vput(vp);
1634                                 VFS_UNLOCK_GIANT(vfslocked);
1635                         }
1636                 } else {
1637                         kve->kve_type = KVME_TYPE_NONE;
1638                         kve->kve_ref_count = 0;
1639                         kve->kve_shadow_count = 0;
1640                 }
1641
1642                 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
1643                 if (freepath != NULL)
1644                         free(freepath, M_TEMP);
1645
1646                 error = SYSCTL_OUT(req, kve, sizeof(*kve));
1647                 vm_map_lock_read(map);
1648                 if (error)
1649                         break;
1650                 if (last_timestamp != map->timestamp) {
1651                         vm_map_lookup_entry(map, addr - 1, &tmp_entry);
1652                         entry = tmp_entry;
1653                 }
1654         }
1655         vm_map_unlock_read(map);
1656         vmspace_free(vm);
1657         PRELE(p);
1658         free(kve, M_TEMP);
1659         return (error);
1660 }
1661 #endif  /* COMPAT_FREEBSD7 */
1662
1663 #ifdef KINFO_VMENTRY_SIZE
1664 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
1665 #endif
1666
1667 static int
1668 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
1669 {
1670         vm_map_entry_t entry, tmp_entry;
1671         unsigned int last_timestamp;
1672         char *fullpath, *freepath;
1673         struct kinfo_vmentry *kve;
1674         struct vattr va;
1675         struct ucred *cred;
1676         int error, *name;
1677         struct vnode *vp;
1678         struct proc *p;
1679         struct vmspace *vm;
1680         vm_map_t map;
1681
1682         name = (int *)arg1;
1683         if ((p = pfind((pid_t)name[0])) == NULL)
1684                 return (ESRCH);
1685         if (p->p_flag & P_WEXIT) {
1686                 PROC_UNLOCK(p);
1687                 return (ESRCH);
1688         }
1689         if ((error = p_candebug(curthread, p))) {
1690                 PROC_UNLOCK(p);
1691                 return (error);
1692         }
1693         _PHOLD(p);
1694         PROC_UNLOCK(p);
1695         vm = vmspace_acquire_ref(p);
1696         if (vm == NULL) {
1697                 PRELE(p);
1698                 return (ESRCH);
1699         }
1700         kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
1701
1702         map = &vm->vm_map;      /* XXXRW: More locking required? */
1703         vm_map_lock_read(map);
1704         for (entry = map->header.next; entry != &map->header;
1705             entry = entry->next) {
1706                 vm_object_t obj, tobj, lobj;
1707                 vm_offset_t addr;
1708                 int vfslocked;
1709
1710                 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
1711                         continue;
1712
1713                 bzero(kve, sizeof(*kve));
1714
1715                 kve->kve_private_resident = 0;
1716                 obj = entry->object.vm_object;
1717                 if (obj != NULL) {
1718                         VM_OBJECT_LOCK(obj);
1719                         if (obj->shadow_count == 1)
1720                                 kve->kve_private_resident =
1721                                     obj->resident_page_count;
1722                 }
1723                 kve->kve_resident = 0;
1724                 addr = entry->start;
1725                 while (addr < entry->end) {
1726                         if (pmap_extract(map->pmap, addr))
1727                                 kve->kve_resident++;
1728                         addr += PAGE_SIZE;
1729                 }
1730
1731                 for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
1732                         if (tobj != obj)
1733                                 VM_OBJECT_LOCK(tobj);
1734                         if (lobj != obj)
1735                                 VM_OBJECT_UNLOCK(lobj);
1736                         lobj = tobj;
1737                 }
1738
1739                 kve->kve_start = entry->start;
1740                 kve->kve_end = entry->end;
1741                 kve->kve_offset = entry->offset;
1742
1743                 if (entry->protection & VM_PROT_READ)
1744                         kve->kve_protection |= KVME_PROT_READ;
1745                 if (entry->protection & VM_PROT_WRITE)
1746                         kve->kve_protection |= KVME_PROT_WRITE;
1747                 if (entry->protection & VM_PROT_EXECUTE)
1748                         kve->kve_protection |= KVME_PROT_EXEC;
1749
1750                 if (entry->eflags & MAP_ENTRY_COW)
1751                         kve->kve_flags |= KVME_FLAG_COW;
1752                 if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
1753                         kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
1754
1755                 last_timestamp = map->timestamp;
1756                 vm_map_unlock_read(map);
1757
1758                 kve->kve_fileid = 0;
1759                 kve->kve_fsid = 0;
1760                 freepath = NULL;
1761                 fullpath = "";
1762                 if (lobj) {
1763                         vp = NULL;
1764                         switch (lobj->type) {
1765                         case OBJT_DEFAULT:
1766                                 kve->kve_type = KVME_TYPE_DEFAULT;
1767                                 break;
1768                         case OBJT_VNODE:
1769                                 kve->kve_type = KVME_TYPE_VNODE;
1770                                 vp = lobj->handle;
1771                                 vref(vp);
1772                                 break;
1773                         case OBJT_SWAP:
1774                                 kve->kve_type = KVME_TYPE_SWAP;
1775                                 break;
1776                         case OBJT_DEVICE:
1777                                 kve->kve_type = KVME_TYPE_DEVICE;
1778                                 break;
1779                         case OBJT_PHYS:
1780                                 kve->kve_type = KVME_TYPE_PHYS;
1781                                 break;
1782                         case OBJT_DEAD:
1783                                 kve->kve_type = KVME_TYPE_DEAD;
1784                                 break;
1785                         case OBJT_SG:
1786                                 kve->kve_type = KVME_TYPE_SG;
1787                                 break;
1788                         default:
1789                                 kve->kve_type = KVME_TYPE_UNKNOWN;
1790                                 break;
1791                         }
1792                         if (lobj != obj)
1793                                 VM_OBJECT_UNLOCK(lobj);
1794
1795                         kve->kve_ref_count = obj->ref_count;
1796                         kve->kve_shadow_count = obj->shadow_count;
1797                         VM_OBJECT_UNLOCK(obj);
1798                         if (vp != NULL) {
1799                                 vn_fullpath(curthread, vp, &fullpath,
1800                                     &freepath);
1801                                 cred = curthread->td_ucred;
1802                                 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
1803                                 vn_lock(vp, LK_SHARED | LK_RETRY);
1804                                 if (VOP_GETATTR(vp, &va, cred) == 0) {
1805                                         kve->kve_fileid = va.va_fileid;
1806                                         kve->kve_fsid = va.va_fsid;
1807                                 }
1808                                 vput(vp);
1809                                 VFS_UNLOCK_GIANT(vfslocked);
1810                         }
1811                 } else {
1812                         kve->kve_type = KVME_TYPE_NONE;
1813                         kve->kve_ref_count = 0;
1814                         kve->kve_shadow_count = 0;
1815                 }
1816
1817                 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
1818                 if (freepath != NULL)
1819                         free(freepath, M_TEMP);
1820
1821                 /* Pack record size down */
1822                 kve->kve_structsize = offsetof(struct kinfo_vmentry, kve_path) +
1823                     strlen(kve->kve_path) + 1;
1824                 kve->kve_structsize = roundup(kve->kve_structsize,
1825                     sizeof(uint64_t));
1826                 error = SYSCTL_OUT(req, kve, kve->kve_structsize);
1827                 vm_map_lock_read(map);
1828                 if (error)
1829                         break;
1830                 if (last_timestamp != map->timestamp) {
1831                         vm_map_lookup_entry(map, addr - 1, &tmp_entry);
1832                         entry = tmp_entry;
1833                 }
1834         }
1835         vm_map_unlock_read(map);
1836         vmspace_free(vm);
1837         PRELE(p);
1838         free(kve, M_TEMP);
1839         return (error);
1840 }
1841
1842 #if defined(STACK) || defined(DDB)
1843 static int
1844 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
1845 {
1846         struct kinfo_kstack *kkstp;
1847         int error, i, *name, numthreads;
1848         lwpid_t *lwpidarray;
1849         struct thread *td;
1850         struct stack *st;
1851         struct sbuf sb;
1852         struct proc *p;
1853
1854         name = (int *)arg1;
1855         if ((p = pfind((pid_t)name[0])) == NULL)
1856                 return (ESRCH);
1857         /* XXXRW: Not clear ESRCH is the right error during proc execve(). */
1858         if (p->p_flag & P_WEXIT || p->p_flag & P_INEXEC) {
1859                 PROC_UNLOCK(p);
1860                 return (ESRCH);
1861         }
1862         if ((error = p_candebug(curthread, p))) {
1863                 PROC_UNLOCK(p);
1864                 return (error);
1865         }
1866         _PHOLD(p);
1867         PROC_UNLOCK(p);
1868
1869         kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
1870         st = stack_create();
1871
1872         lwpidarray = NULL;
1873         numthreads = 0;
1874         PROC_LOCK(p);
1875 repeat:
1876         if (numthreads < p->p_numthreads) {
1877                 if (lwpidarray != NULL) {
1878                         free(lwpidarray, M_TEMP);
1879                         lwpidarray = NULL;
1880                 }
1881                 numthreads = p->p_numthreads;
1882                 PROC_UNLOCK(p);
1883                 lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
1884                     M_WAITOK | M_ZERO);
1885                 PROC_LOCK(p);
1886                 goto repeat;
1887         }
1888         i = 0;
1889
1890         /*
1891          * XXXRW: During the below loop, execve(2) and countless other sorts
1892          * of changes could have taken place.  Should we check to see if the
1893          * vmspace has been replaced, or the like, in order to prevent
1894          * giving a snapshot that spans, say, execve(2), with some threads
1895          * before and some after?  Among other things, the credentials could
1896          * have changed, in which case the right to extract debug info might
1897          * no longer be assured.
1898          */
1899         FOREACH_THREAD_IN_PROC(p, td) {
1900                 KASSERT(i < numthreads,
1901                     ("sysctl_kern_proc_kstack: numthreads"));
1902                 lwpidarray[i] = td->td_tid;
1903                 i++;
1904         }
1905         numthreads = i;
1906         for (i = 0; i < numthreads; i++) {
1907                 td = thread_find(p, lwpidarray[i]);
1908                 if (td == NULL) {
1909                         continue;
1910                 }
1911                 bzero(kkstp, sizeof(*kkstp));
1912                 (void)sbuf_new(&sb, kkstp->kkst_trace,
1913                     sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
1914                 thread_lock(td);
1915                 kkstp->kkst_tid = td->td_tid;
1916                 if (TD_IS_SWAPPED(td))
1917                         kkstp->kkst_state = KKST_STATE_SWAPPED;
1918                 else if (TD_IS_RUNNING(td))
1919                         kkstp->kkst_state = KKST_STATE_RUNNING;
1920                 else {
1921                         kkstp->kkst_state = KKST_STATE_STACKOK;
1922                         stack_save_td(st, td);
1923                 }
1924                 thread_unlock(td);
1925                 PROC_UNLOCK(p);
1926                 stack_sbuf_print(&sb, st);
1927                 sbuf_finish(&sb);
1928                 sbuf_delete(&sb);
1929                 error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
1930                 PROC_LOCK(p);
1931                 if (error)
1932                         break;
1933         }
1934         _PRELE(p);
1935         PROC_UNLOCK(p);
1936         if (lwpidarray != NULL)
1937                 free(lwpidarray, M_TEMP);
1938         stack_destroy(st);
1939         free(kkstp, M_TEMP);
1940         return (error);
1941 }
1942 #endif
1943
1944 /*
1945  * This sysctl allows a process to retrieve the full list of groups from
1946  * itself or another process.
1947  */
1948 static int
1949 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
1950 {
1951         pid_t *pidp = (pid_t *)arg1;
1952         unsigned int arglen = arg2;
1953         struct proc *p;
1954         struct ucred *cred;
1955         int error;
1956
1957         if (arglen != 1)
1958                 return (EINVAL);
1959         if (*pidp == -1) {      /* -1 means this process */
1960                 p = req->td->td_proc;
1961         } else {
1962                 p = pfind(*pidp);
1963                 if (p == NULL)
1964                         return (ESRCH);
1965                 if ((error = p_cansee(curthread, p)) != 0) {
1966                         PROC_UNLOCK(p);
1967                         return (error);
1968                 }
1969         }
1970
1971         cred = crhold(p->p_ucred);
1972         if (*pidp != -1)
1973                 PROC_UNLOCK(p);
1974
1975         error = SYSCTL_OUT(req, cred->cr_groups,
1976             cred->cr_ngroups * sizeof(gid_t));
1977         crfree(cred);
1978         return (error);
1979 }
1980
1981 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD,  0, "Process table");
1982
1983 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
1984         CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
1985         "Return entire process table");
1986
1987 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
1988         sysctl_kern_proc, "Process table");
1989
1990 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
1991         sysctl_kern_proc, "Process table");
1992
1993 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
1994         sysctl_kern_proc, "Process table");
1995
1996 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
1997         CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
1998
1999 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE, 
2000         sysctl_kern_proc, "Process table");
2001
2002 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE, 
2003         sysctl_kern_proc, "Process table");
2004
2005 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2006         sysctl_kern_proc, "Process table");
2007
2008 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2009         sysctl_kern_proc, "Process table");
2010
2011 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
2012         sysctl_kern_proc, "Return process table, no threads");
2013
2014 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
2015         CTLFLAG_RW | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
2016         sysctl_kern_proc_args, "Process argument list");
2017
2018 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
2019         CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
2020
2021 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
2022         CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
2023         "Process syscall vector name (ABI type)");
2024
2025 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
2026         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2027
2028 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
2029         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2030
2031 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
2032         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2033
2034 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
2035         sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2036
2037 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
2038         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2039
2040 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
2041         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2042
2043 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
2044         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2045
2046 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
2047         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2048
2049 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
2050         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
2051         "Return process table, no threads");
2052
2053 #ifdef COMPAT_FREEBSD7
2054 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
2055         CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
2056 #endif
2057
2058 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
2059         CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
2060
2061 #if defined(STACK) || defined(DDB)
2062 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
2063         CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
2064 #endif
2065
2066 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
2067         CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");