]> CyberLeo.Net >> Repos - FreeBSD/releng/9.0.git/blob - sys/kern/kern_proc.c
Copy stable/9 to releng/9.0 as part of the FreeBSD 9.0-RELEASE release
[FreeBSD/releng/9.0.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/loginclass.h>
48 #include <sys/malloc.h>
49 #include <sys/mount.h>
50 #include <sys/mutex.h>
51 #include <sys/proc.h>
52 #include <sys/refcount.h>
53 #include <sys/sbuf.h>
54 #include <sys/sysent.h>
55 #include <sys/sched.h>
56 #include <sys/smp.h>
57 #include <sys/stack.h>
58 #include <sys/sysctl.h>
59 #include <sys/filedesc.h>
60 #include <sys/tty.h>
61 #include <sys/signalvar.h>
62 #include <sys/sdt.h>
63 #include <sys/sx.h>
64 #include <sys/user.h>
65 #include <sys/jail.h>
66 #include <sys/vnode.h>
67 #include <sys/eventhandler.h>
68
69 #ifdef DDB
70 #include <ddb/ddb.h>
71 #endif
72
73 #include <vm/vm.h>
74 #include <vm/vm_extern.h>
75 #include <vm/pmap.h>
76 #include <vm/vm_map.h>
77 #include <vm/vm_object.h>
78 #include <vm/uma.h>
79
80 #ifdef COMPAT_FREEBSD32
81 #include <compat/freebsd32/freebsd32.h>
82 #include <compat/freebsd32/freebsd32_util.h>
83 #endif
84
85 SDT_PROVIDER_DEFINE(proc);
86 SDT_PROBE_DEFINE(proc, kernel, ctor, entry, entry);
87 SDT_PROBE_ARGTYPE(proc, kernel, ctor, entry, 0, "struct proc *");
88 SDT_PROBE_ARGTYPE(proc, kernel, ctor, entry, 1, "int");
89 SDT_PROBE_ARGTYPE(proc, kernel, ctor, entry, 2, "void *");
90 SDT_PROBE_ARGTYPE(proc, kernel, ctor, entry, 3, "int");
91 SDT_PROBE_DEFINE(proc, kernel, ctor, return, return);
92 SDT_PROBE_ARGTYPE(proc, kernel, ctor, return, 0, "struct proc *");
93 SDT_PROBE_ARGTYPE(proc, kernel, ctor, return, 1, "int");
94 SDT_PROBE_ARGTYPE(proc, kernel, ctor, return, 2, "void *");
95 SDT_PROBE_ARGTYPE(proc, kernel, ctor, return, 3, "int");
96 SDT_PROBE_DEFINE(proc, kernel, dtor, entry, entry);
97 SDT_PROBE_ARGTYPE(proc, kernel, dtor, entry, 0, "struct proc *");
98 SDT_PROBE_ARGTYPE(proc, kernel, dtor, entry, 1, "int");
99 SDT_PROBE_ARGTYPE(proc, kernel, dtor, entry, 2, "void *");
100 SDT_PROBE_ARGTYPE(proc, kernel, dtor, entry, 3, "struct thread *");
101 SDT_PROBE_DEFINE(proc, kernel, dtor, return, return);
102 SDT_PROBE_ARGTYPE(proc, kernel, dtor, return, 0, "struct proc *");
103 SDT_PROBE_ARGTYPE(proc, kernel, dtor, return, 1, "int");
104 SDT_PROBE_ARGTYPE(proc, kernel, dtor, return, 2, "void *");
105 SDT_PROBE_DEFINE(proc, kernel, init, entry, entry);
106 SDT_PROBE_ARGTYPE(proc, kernel, init, entry, 0, "struct proc *");
107 SDT_PROBE_ARGTYPE(proc, kernel, init, entry, 1, "int");
108 SDT_PROBE_ARGTYPE(proc, kernel, init, entry, 2, "int");
109 SDT_PROBE_DEFINE(proc, kernel, init, return, return);
110 SDT_PROBE_ARGTYPE(proc, kernel, init, return, 0, "struct proc *");
111 SDT_PROBE_ARGTYPE(proc, kernel, init, return, 1, "int");
112 SDT_PROBE_ARGTYPE(proc, kernel, init, return, 2, "int");
113
114 MALLOC_DEFINE(M_PGRP, "pgrp", "process group header");
115 MALLOC_DEFINE(M_SESSION, "session", "session header");
116 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
117 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
118
119 static void doenterpgrp(struct proc *, struct pgrp *);
120 static void orphanpg(struct pgrp *pg);
121 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp);
122 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp);
123 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp,
124     int preferthread);
125 static void pgadjustjobc(struct pgrp *pgrp, int entering);
126 static void pgdelete(struct pgrp *);
127 static int proc_ctor(void *mem, int size, void *arg, int flags);
128 static void proc_dtor(void *mem, int size, void *arg);
129 static int proc_init(void *mem, int size, int flags);
130 static void proc_fini(void *mem, int size);
131 static void pargs_free(struct pargs *pa);
132
133 /*
134  * Other process lists
135  */
136 struct pidhashhead *pidhashtbl;
137 u_long pidhash;
138 struct pgrphashhead *pgrphashtbl;
139 u_long pgrphash;
140 struct proclist allproc;
141 struct proclist zombproc;
142 struct sx allproc_lock;
143 struct sx proctree_lock;
144 struct mtx ppeers_lock;
145 uma_zone_t proc_zone;
146
147 int kstack_pages = KSTACK_PAGES;
148 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0,
149     "Kernel stack size in pages");
150
151 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
152 #ifdef COMPAT_FREEBSD32
153 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
154 #endif
155
156 /*
157  * Initialize global process hashing structures.
158  */
159 void
160 procinit()
161 {
162
163         sx_init(&allproc_lock, "allproc");
164         sx_init(&proctree_lock, "proctree");
165         mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
166         LIST_INIT(&allproc);
167         LIST_INIT(&zombproc);
168         pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
169         pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
170         proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
171             proc_ctor, proc_dtor, proc_init, proc_fini,
172             UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
173         uihashinit();
174 }
175
176 /*
177  * Prepare a proc for use.
178  */
179 static int
180 proc_ctor(void *mem, int size, void *arg, int flags)
181 {
182         struct proc *p;
183
184         p = (struct proc *)mem;
185         SDT_PROBE(proc, kernel, ctor , entry, p, size, arg, flags, 0);
186         EVENTHANDLER_INVOKE(process_ctor, p);
187         SDT_PROBE(proc, kernel, ctor , return, p, size, arg, flags, 0);
188         return (0);
189 }
190
191 /*
192  * Reclaim a proc after use.
193  */
194 static void
195 proc_dtor(void *mem, int size, void *arg)
196 {
197         struct proc *p;
198         struct thread *td;
199
200         /* INVARIANTS checks go here */
201         p = (struct proc *)mem;
202         td = FIRST_THREAD_IN_PROC(p);
203         SDT_PROBE(proc, kernel, dtor, entry, p, size, arg, td, 0);
204         if (td != NULL) {
205 #ifdef INVARIANTS
206                 KASSERT((p->p_numthreads == 1),
207                     ("bad number of threads in exiting process"));
208                 KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
209 #endif
210                 /* Free all OSD associated to this thread. */
211                 osd_thread_exit(td);
212         }
213         EVENTHANDLER_INVOKE(process_dtor, p);
214         if (p->p_ksi != NULL)
215                 KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
216         SDT_PROBE(proc, kernel, dtor, return, p, size, arg, 0, 0);
217 }
218
219 /*
220  * Initialize type-stable parts of a proc (when newly created).
221  */
222 static int
223 proc_init(void *mem, int size, int flags)
224 {
225         struct proc *p;
226
227         p = (struct proc *)mem;
228         SDT_PROBE(proc, kernel, init, entry, p, size, flags, 0, 0);
229         p->p_sched = (struct p_sched *)&p[1];
230         bzero(&p->p_mtx, sizeof(struct mtx));
231         mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK);
232         mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_RECURSE);
233         cv_init(&p->p_pwait, "ppwait");
234         cv_init(&p->p_dbgwait, "dbgwait");
235         TAILQ_INIT(&p->p_threads);           /* all threads in proc */
236         EVENTHANDLER_INVOKE(process_init, p);
237         p->p_stats = pstats_alloc();
238         SDT_PROBE(proc, kernel, init, return, p, size, flags, 0, 0);
239         return (0);
240 }
241
242 /*
243  * UMA should ensure that this function is never called.
244  * Freeing a proc structure would violate type stability.
245  */
246 static void
247 proc_fini(void *mem, int size)
248 {
249 #ifdef notnow
250         struct proc *p;
251
252         p = (struct proc *)mem;
253         EVENTHANDLER_INVOKE(process_fini, p);
254         pstats_free(p->p_stats);
255         thread_free(FIRST_THREAD_IN_PROC(p));
256         mtx_destroy(&p->p_mtx);
257         if (p->p_ksi != NULL)
258                 ksiginfo_free(p->p_ksi);
259 #else
260         panic("proc reclaimed");
261 #endif
262 }
263
264 /*
265  * Is p an inferior of the current process?
266  */
267 int
268 inferior(p)
269         register struct proc *p;
270 {
271
272         sx_assert(&proctree_lock, SX_LOCKED);
273         for (; p != curproc; p = p->p_pptr)
274                 if (p->p_pid == 0)
275                         return (0);
276         return (1);
277 }
278
279 /*
280  * Locate a process by number; return only "live" processes -- i.e., neither
281  * zombies nor newly born but incompletely initialized processes.  By not
282  * returning processes in the PRS_NEW state, we allow callers to avoid
283  * testing for that condition to avoid dereferencing p_ucred, et al.
284  */
285 struct proc *
286 pfind(pid)
287         register pid_t pid;
288 {
289         register struct proc *p;
290
291         sx_slock(&allproc_lock);
292         LIST_FOREACH(p, PIDHASH(pid), p_hash)
293                 if (p->p_pid == pid) {
294                         PROC_LOCK(p);
295                         if (p->p_state == PRS_NEW) {
296                                 PROC_UNLOCK(p);
297                                 p = NULL;
298                         }
299                         break;
300                 }
301         sx_sunlock(&allproc_lock);
302         return (p);
303 }
304
305 /*
306  * Locate a process group by number.
307  * The caller must hold proctree_lock.
308  */
309 struct pgrp *
310 pgfind(pgid)
311         register pid_t pgid;
312 {
313         register struct pgrp *pgrp;
314
315         sx_assert(&proctree_lock, SX_LOCKED);
316
317         LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
318                 if (pgrp->pg_id == pgid) {
319                         PGRP_LOCK(pgrp);
320                         return (pgrp);
321                 }
322         }
323         return (NULL);
324 }
325
326 /*
327  * Create a new process group.
328  * pgid must be equal to the pid of p.
329  * Begin a new session if required.
330  */
331 int
332 enterpgrp(p, pgid, pgrp, sess)
333         register struct proc *p;
334         pid_t pgid;
335         struct pgrp *pgrp;
336         struct session *sess;
337 {
338         struct pgrp *pgrp2;
339
340         sx_assert(&proctree_lock, SX_XLOCKED);
341
342         KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
343         KASSERT(p->p_pid == pgid,
344             ("enterpgrp: new pgrp and pid != pgid"));
345
346         pgrp2 = pgfind(pgid);
347
348         KASSERT(pgrp2 == NULL,
349             ("enterpgrp: pgrp with pgid exists"));
350         KASSERT(!SESS_LEADER(p),
351             ("enterpgrp: session leader attempted setpgrp"));
352
353         mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
354
355         if (sess != NULL) {
356                 /*
357                  * new session
358                  */
359                 mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
360                 PROC_LOCK(p);
361                 p->p_flag &= ~P_CONTROLT;
362                 PROC_UNLOCK(p);
363                 PGRP_LOCK(pgrp);
364                 sess->s_leader = p;
365                 sess->s_sid = p->p_pid;
366                 refcount_init(&sess->s_count, 1);
367                 sess->s_ttyvp = NULL;
368                 sess->s_ttydp = 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                                 kern_psignal(p, SIGHUP);
612                                 kern_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         kp->ki_traceflag = p->p_traceflag;
720 #endif
721         kp->ki_fd = p->p_fd;
722         kp->ki_vmspace = p->p_vmspace;
723         kp->ki_flag = p->p_flag;
724         cred = p->p_ucred;
725         if (cred) {
726                 kp->ki_uid = cred->cr_uid;
727                 kp->ki_ruid = cred->cr_ruid;
728                 kp->ki_svuid = cred->cr_svuid;
729                 kp->ki_cr_flags = 0;
730                 if (cred->cr_flags & CRED_FLAG_CAPMODE)
731                         kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
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                 strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
750                     sizeof(kp->ki_loginclass));
751         }
752         ps = p->p_sigacts;
753         if (ps) {
754                 mtx_lock(&ps->ps_mtx);
755                 kp->ki_sigignore = ps->ps_sigignore;
756                 kp->ki_sigcatch = ps->ps_sigcatch;
757                 mtx_unlock(&ps->ps_mtx);
758         }
759         if (p->p_state != PRS_NEW &&
760             p->p_state != PRS_ZOMBIE &&
761             p->p_vmspace != NULL) {
762                 struct vmspace *vm = p->p_vmspace;
763
764                 kp->ki_size = vm->vm_map.size;
765                 kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
766                 FOREACH_THREAD_IN_PROC(p, td0) {
767                         if (!TD_IS_SWAPPED(td0))
768                                 kp->ki_rssize += td0->td_kstack_pages;
769                 }
770                 kp->ki_swrss = vm->vm_swrss;
771                 kp->ki_tsize = vm->vm_tsize;
772                 kp->ki_dsize = vm->vm_dsize;
773                 kp->ki_ssize = vm->vm_ssize;
774         } else if (p->p_state == PRS_ZOMBIE)
775                 kp->ki_stat = SZOMB;
776         if (kp->ki_flag & P_INMEM)
777                 kp->ki_sflag = PS_INMEM;
778         else
779                 kp->ki_sflag = 0;
780         /* Calculate legacy swtime as seconds since 'swtick'. */
781         kp->ki_swtime = (ticks - p->p_swtick) / hz;
782         kp->ki_pid = p->p_pid;
783         kp->ki_nice = p->p_nice;
784         kp->ki_start = p->p_stats->p_start;
785         timevaladd(&kp->ki_start, &boottime);
786         PROC_SLOCK(p);
787         rufetch(p, &kp->ki_rusage);
788         kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
789         calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
790         PROC_SUNLOCK(p);
791         calccru(p, &kp->ki_childutime, &kp->ki_childstime);
792         /* Some callers want child times in a single value. */
793         kp->ki_childtime = kp->ki_childstime;
794         timevaladd(&kp->ki_childtime, &kp->ki_childutime);
795
796         tp = NULL;
797         if (p->p_pgrp) {
798                 kp->ki_pgid = p->p_pgrp->pg_id;
799                 kp->ki_jobc = p->p_pgrp->pg_jobc;
800                 sp = p->p_pgrp->pg_session;
801
802                 if (sp != NULL) {
803                         kp->ki_sid = sp->s_sid;
804                         SESS_LOCK(sp);
805                         strlcpy(kp->ki_login, sp->s_login,
806                             sizeof(kp->ki_login));
807                         if (sp->s_ttyvp)
808                                 kp->ki_kiflag |= KI_CTTY;
809                         if (SESS_LEADER(p))
810                                 kp->ki_kiflag |= KI_SLEADER;
811                         /* XXX proctree_lock */
812                         tp = sp->s_ttyp;
813                         SESS_UNLOCK(sp);
814                 }
815         }
816         if ((p->p_flag & P_CONTROLT) && tp != NULL) {
817                 kp->ki_tdev = tty_udev(tp);
818                 kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
819                 if (tp->t_session)
820                         kp->ki_tsid = tp->t_session->s_sid;
821         } else
822                 kp->ki_tdev = NODEV;
823         if (p->p_comm[0] != '\0')
824                 strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
825         if (p->p_sysent && p->p_sysent->sv_name != NULL &&
826             p->p_sysent->sv_name[0] != '\0')
827                 strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
828         kp->ki_siglist = p->p_siglist;
829         kp->ki_xstat = p->p_xstat;
830         kp->ki_acflag = p->p_acflag;
831         kp->ki_lock = p->p_lock;
832         if (p->p_pptr)
833                 kp->ki_ppid = p->p_pptr->p_pid;
834 }
835
836 /*
837  * Fill in information that is thread specific.  Must be called with
838  * target process locked.  If 'preferthread' is set, overwrite certain
839  * process-related fields that are maintained for both threads and
840  * processes.
841  */
842 static void
843 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
844 {
845         struct proc *p;
846
847         p = td->td_proc;
848         kp->ki_tdaddr = td;
849         PROC_LOCK_ASSERT(p, MA_OWNED);
850
851         if (preferthread)
852                 PROC_SLOCK(p);
853         thread_lock(td);
854         if (td->td_wmesg != NULL)
855                 strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
856         else
857                 bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
858         strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname));
859         if (TD_ON_LOCK(td)) {
860                 kp->ki_kiflag |= KI_LOCKBLOCK;
861                 strlcpy(kp->ki_lockname, td->td_lockname,
862                     sizeof(kp->ki_lockname));
863         } else {
864                 kp->ki_kiflag &= ~KI_LOCKBLOCK;
865                 bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
866         }
867
868         if (p->p_state == PRS_NORMAL) { /* approximate. */
869                 if (TD_ON_RUNQ(td) ||
870                     TD_CAN_RUN(td) ||
871                     TD_IS_RUNNING(td)) {
872                         kp->ki_stat = SRUN;
873                 } else if (P_SHOULDSTOP(p)) {
874                         kp->ki_stat = SSTOP;
875                 } else if (TD_IS_SLEEPING(td)) {
876                         kp->ki_stat = SSLEEP;
877                 } else if (TD_ON_LOCK(td)) {
878                         kp->ki_stat = SLOCK;
879                 } else {
880                         kp->ki_stat = SWAIT;
881                 }
882         } else if (p->p_state == PRS_ZOMBIE) {
883                 kp->ki_stat = SZOMB;
884         } else {
885                 kp->ki_stat = SIDL;
886         }
887
888         /* Things in the thread */
889         kp->ki_wchan = td->td_wchan;
890         kp->ki_pri.pri_level = td->td_priority;
891         kp->ki_pri.pri_native = td->td_base_pri;
892         kp->ki_lastcpu = td->td_lastcpu;
893         kp->ki_oncpu = td->td_oncpu;
894         kp->ki_tdflags = td->td_flags;
895         kp->ki_tid = td->td_tid;
896         kp->ki_numthreads = p->p_numthreads;
897         kp->ki_pcb = td->td_pcb;
898         kp->ki_kstack = (void *)td->td_kstack;
899         kp->ki_slptime = (ticks - td->td_slptick) / hz;
900         kp->ki_pri.pri_class = td->td_pri_class;
901         kp->ki_pri.pri_user = td->td_user_pri;
902
903         if (preferthread) {
904                 rufetchtd(td, &kp->ki_rusage);
905                 kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
906                 kp->ki_pctcpu = sched_pctcpu(td);
907                 kp->ki_estcpu = td->td_estcpu;
908         }
909
910         /* We can't get this anymore but ps etc never used it anyway. */
911         kp->ki_rqindex = 0;
912
913         if (preferthread)
914                 kp->ki_siglist = td->td_siglist;
915         kp->ki_sigmask = td->td_sigmask;
916         thread_unlock(td);
917         if (preferthread)
918                 PROC_SUNLOCK(p);
919 }
920
921 /*
922  * Fill in a kinfo_proc structure for the specified process.
923  * Must be called with the target process locked.
924  */
925 void
926 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
927 {
928
929         MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
930
931         fill_kinfo_proc_only(p, kp);
932         fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
933         fill_kinfo_aggregate(p, kp);
934 }
935
936 struct pstats *
937 pstats_alloc(void)
938 {
939
940         return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
941 }
942
943 /*
944  * Copy parts of p_stats; zero the rest of p_stats (statistics).
945  */
946 void
947 pstats_fork(struct pstats *src, struct pstats *dst)
948 {
949
950         bzero(&dst->pstat_startzero,
951             __rangeof(struct pstats, pstat_startzero, pstat_endzero));
952         bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
953             __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
954 }
955
956 void
957 pstats_free(struct pstats *ps)
958 {
959
960         free(ps, M_SUBPROC);
961 }
962
963 /*
964  * Locate a zombie process by number
965  */
966 struct proc *
967 zpfind(pid_t pid)
968 {
969         struct proc *p;
970
971         sx_slock(&allproc_lock);
972         LIST_FOREACH(p, &zombproc, p_list)
973                 if (p->p_pid == pid) {
974                         PROC_LOCK(p);
975                         break;
976                 }
977         sx_sunlock(&allproc_lock);
978         return (p);
979 }
980
981 #define KERN_PROC_ZOMBMASK      0x3
982 #define KERN_PROC_NOTHREADS     0x4
983
984 #ifdef COMPAT_FREEBSD32
985
986 /*
987  * This function is typically used to copy out the kernel address, so
988  * it can be replaced by assignment of zero.
989  */
990 static inline uint32_t
991 ptr32_trim(void *ptr)
992 {
993         uintptr_t uptr;
994
995         uptr = (uintptr_t)ptr;
996         return ((uptr > UINT_MAX) ? 0 : uptr);
997 }
998
999 #define PTRTRIM_CP(src,dst,fld) \
1000         do { (dst).fld = ptr32_trim((src).fld); } while (0)
1001
1002 static void
1003 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1004 {
1005         int i;
1006
1007         bzero(ki32, sizeof(struct kinfo_proc32));
1008         ki32->ki_structsize = sizeof(struct kinfo_proc32);
1009         CP(*ki, *ki32, ki_layout);
1010         PTRTRIM_CP(*ki, *ki32, ki_args);
1011         PTRTRIM_CP(*ki, *ki32, ki_paddr);
1012         PTRTRIM_CP(*ki, *ki32, ki_addr);
1013         PTRTRIM_CP(*ki, *ki32, ki_tracep);
1014         PTRTRIM_CP(*ki, *ki32, ki_textvp);
1015         PTRTRIM_CP(*ki, *ki32, ki_fd);
1016         PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1017         PTRTRIM_CP(*ki, *ki32, ki_wchan);
1018         CP(*ki, *ki32, ki_pid);
1019         CP(*ki, *ki32, ki_ppid);
1020         CP(*ki, *ki32, ki_pgid);
1021         CP(*ki, *ki32, ki_tpgid);
1022         CP(*ki, *ki32, ki_sid);
1023         CP(*ki, *ki32, ki_tsid);
1024         CP(*ki, *ki32, ki_jobc);
1025         CP(*ki, *ki32, ki_tdev);
1026         CP(*ki, *ki32, ki_siglist);
1027         CP(*ki, *ki32, ki_sigmask);
1028         CP(*ki, *ki32, ki_sigignore);
1029         CP(*ki, *ki32, ki_sigcatch);
1030         CP(*ki, *ki32, ki_uid);
1031         CP(*ki, *ki32, ki_ruid);
1032         CP(*ki, *ki32, ki_svuid);
1033         CP(*ki, *ki32, ki_rgid);
1034         CP(*ki, *ki32, ki_svgid);
1035         CP(*ki, *ki32, ki_ngroups);
1036         for (i = 0; i < KI_NGROUPS; i++)
1037                 CP(*ki, *ki32, ki_groups[i]);
1038         CP(*ki, *ki32, ki_size);
1039         CP(*ki, *ki32, ki_rssize);
1040         CP(*ki, *ki32, ki_swrss);
1041         CP(*ki, *ki32, ki_tsize);
1042         CP(*ki, *ki32, ki_dsize);
1043         CP(*ki, *ki32, ki_ssize);
1044         CP(*ki, *ki32, ki_xstat);
1045         CP(*ki, *ki32, ki_acflag);
1046         CP(*ki, *ki32, ki_pctcpu);
1047         CP(*ki, *ki32, ki_estcpu);
1048         CP(*ki, *ki32, ki_slptime);
1049         CP(*ki, *ki32, ki_swtime);
1050         CP(*ki, *ki32, ki_runtime);
1051         TV_CP(*ki, *ki32, ki_start);
1052         TV_CP(*ki, *ki32, ki_childtime);
1053         CP(*ki, *ki32, ki_flag);
1054         CP(*ki, *ki32, ki_kiflag);
1055         CP(*ki, *ki32, ki_traceflag);
1056         CP(*ki, *ki32, ki_stat);
1057         CP(*ki, *ki32, ki_nice);
1058         CP(*ki, *ki32, ki_lock);
1059         CP(*ki, *ki32, ki_rqindex);
1060         CP(*ki, *ki32, ki_oncpu);
1061         CP(*ki, *ki32, ki_lastcpu);
1062         bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1063         bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1064         bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1065         bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1066         bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1067         bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1068         bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1069         CP(*ki, *ki32, ki_cr_flags);
1070         CP(*ki, *ki32, ki_jid);
1071         CP(*ki, *ki32, ki_numthreads);
1072         CP(*ki, *ki32, ki_tid);
1073         CP(*ki, *ki32, ki_pri);
1074         freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1075         freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1076         PTRTRIM_CP(*ki, *ki32, ki_pcb);
1077         PTRTRIM_CP(*ki, *ki32, ki_kstack);
1078         PTRTRIM_CP(*ki, *ki32, ki_udata);
1079         CP(*ki, *ki32, ki_sflag);
1080         CP(*ki, *ki32, ki_tdflags);
1081 }
1082
1083 static int
1084 sysctl_out_proc_copyout(struct kinfo_proc *ki, struct sysctl_req *req)
1085 {
1086         struct kinfo_proc32 ki32;
1087         int error;
1088
1089         if (req->flags & SCTL_MASK32) {
1090                 freebsd32_kinfo_proc_out(ki, &ki32);
1091                 error = SYSCTL_OUT(req, &ki32, sizeof(struct kinfo_proc32));
1092         } else
1093                 error = SYSCTL_OUT(req, ki, sizeof(struct kinfo_proc));
1094         return (error);
1095 }
1096 #else
1097 static int
1098 sysctl_out_proc_copyout(struct kinfo_proc *ki, struct sysctl_req *req)
1099 {
1100
1101         return (SYSCTL_OUT(req, ki, sizeof(struct kinfo_proc)));
1102 }
1103 #endif
1104
1105 /*
1106  * Must be called with the process locked and will return with it unlocked.
1107  */
1108 static int
1109 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
1110 {
1111         struct thread *td;
1112         struct kinfo_proc kinfo_proc;
1113         int error = 0;
1114         struct proc *np;
1115         pid_t pid = p->p_pid;
1116
1117         PROC_LOCK_ASSERT(p, MA_OWNED);
1118         MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1119
1120         fill_kinfo_proc(p, &kinfo_proc);
1121         if (flags & KERN_PROC_NOTHREADS)
1122                 error = sysctl_out_proc_copyout(&kinfo_proc, req);
1123         else {
1124                 FOREACH_THREAD_IN_PROC(p, td) {
1125                         fill_kinfo_thread(td, &kinfo_proc, 1);
1126                         error = sysctl_out_proc_copyout(&kinfo_proc, req);
1127                         if (error)
1128                                 break;
1129                 }
1130         }
1131         PROC_UNLOCK(p);
1132         if (error)
1133                 return (error);
1134         if (flags & KERN_PROC_ZOMBMASK)
1135                 np = zpfind(pid);
1136         else {
1137                 if (pid == 0)
1138                         return (0);
1139                 np = pfind(pid);
1140         }
1141         if (np == NULL)
1142                 return (ESRCH);
1143         if (np != p) {
1144                 PROC_UNLOCK(np);
1145                 return (ESRCH);
1146         }
1147         PROC_UNLOCK(np);
1148         return (0);
1149 }
1150
1151 static int
1152 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1153 {
1154         int *name = (int*) arg1;
1155         u_int namelen = arg2;
1156         struct proc *p;
1157         int flags, doingzomb, oid_number;
1158         int error = 0;
1159
1160         oid_number = oidp->oid_number;
1161         if (oid_number != KERN_PROC_ALL &&
1162             (oid_number & KERN_PROC_INC_THREAD) == 0)
1163                 flags = KERN_PROC_NOTHREADS;
1164         else {
1165                 flags = 0;
1166                 oid_number &= ~KERN_PROC_INC_THREAD;
1167         }
1168         if (oid_number == KERN_PROC_PID) {
1169                 if (namelen != 1) 
1170                         return (EINVAL);
1171                 error = sysctl_wire_old_buffer(req, 0);
1172                 if (error)
1173                         return (error);         
1174                 p = pfind((pid_t)name[0]);
1175                 if (!p)
1176                         return (ESRCH);
1177                 if ((error = p_cansee(curthread, p))) {
1178                         PROC_UNLOCK(p);
1179                         return (error);
1180                 }
1181                 error = sysctl_out_proc(p, req, flags);
1182                 return (error);
1183         }
1184
1185         switch (oid_number) {
1186         case KERN_PROC_ALL:
1187                 if (namelen != 0)
1188                         return (EINVAL);
1189                 break;
1190         case KERN_PROC_PROC:
1191                 if (namelen != 0 && namelen != 1)
1192                         return (EINVAL);
1193                 break;
1194         default:
1195                 if (namelen != 1)
1196                         return (EINVAL);
1197                 break;
1198         }
1199         
1200         if (!req->oldptr) {
1201                 /* overestimate by 5 procs */
1202                 error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1203                 if (error)
1204                         return (error);
1205         }
1206         error = sysctl_wire_old_buffer(req, 0);
1207         if (error != 0)
1208                 return (error);
1209         sx_slock(&allproc_lock);
1210         for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) {
1211                 if (!doingzomb)
1212                         p = LIST_FIRST(&allproc);
1213                 else
1214                         p = LIST_FIRST(&zombproc);
1215                 for (; p != 0; p = LIST_NEXT(p, p_list)) {
1216                         /*
1217                          * Skip embryonic processes.
1218                          */
1219                         PROC_LOCK(p);
1220                         if (p->p_state == PRS_NEW) {
1221                                 PROC_UNLOCK(p);
1222                                 continue;
1223                         }
1224                         KASSERT(p->p_ucred != NULL,
1225                             ("process credential is NULL for non-NEW proc"));
1226                         /*
1227                          * Show a user only appropriate processes.
1228                          */
1229                         if (p_cansee(curthread, p)) {
1230                                 PROC_UNLOCK(p);
1231                                 continue;
1232                         }
1233                         /*
1234                          * TODO - make more efficient (see notes below).
1235                          * do by session.
1236                          */
1237                         switch (oid_number) {
1238
1239                         case KERN_PROC_GID:
1240                                 if (p->p_ucred->cr_gid != (gid_t)name[0]) {
1241                                         PROC_UNLOCK(p);
1242                                         continue;
1243                                 }
1244                                 break;
1245
1246                         case KERN_PROC_PGRP:
1247                                 /* could do this by traversing pgrp */
1248                                 if (p->p_pgrp == NULL ||
1249                                     p->p_pgrp->pg_id != (pid_t)name[0]) {
1250                                         PROC_UNLOCK(p);
1251                                         continue;
1252                                 }
1253                                 break;
1254
1255                         case KERN_PROC_RGID:
1256                                 if (p->p_ucred->cr_rgid != (gid_t)name[0]) {
1257                                         PROC_UNLOCK(p);
1258                                         continue;
1259                                 }
1260                                 break;
1261
1262                         case KERN_PROC_SESSION:
1263                                 if (p->p_session == NULL ||
1264                                     p->p_session->s_sid != (pid_t)name[0]) {
1265                                         PROC_UNLOCK(p);
1266                                         continue;
1267                                 }
1268                                 break;
1269
1270                         case KERN_PROC_TTY:
1271                                 if ((p->p_flag & P_CONTROLT) == 0 ||
1272                                     p->p_session == NULL) {
1273                                         PROC_UNLOCK(p);
1274                                         continue;
1275                                 }
1276                                 /* XXX proctree_lock */
1277                                 SESS_LOCK(p->p_session);
1278                                 if (p->p_session->s_ttyp == NULL ||
1279                                     tty_udev(p->p_session->s_ttyp) != 
1280                                     (dev_t)name[0]) {
1281                                         SESS_UNLOCK(p->p_session);
1282                                         PROC_UNLOCK(p);
1283                                         continue;
1284                                 }
1285                                 SESS_UNLOCK(p->p_session);
1286                                 break;
1287
1288                         case KERN_PROC_UID:
1289                                 if (p->p_ucred->cr_uid != (uid_t)name[0]) {
1290                                         PROC_UNLOCK(p);
1291                                         continue;
1292                                 }
1293                                 break;
1294
1295                         case KERN_PROC_RUID:
1296                                 if (p->p_ucred->cr_ruid != (uid_t)name[0]) {
1297                                         PROC_UNLOCK(p);
1298                                         continue;
1299                                 }
1300                                 break;
1301
1302                         case KERN_PROC_PROC:
1303                                 break;
1304
1305                         default:
1306                                 break;
1307
1308                         }
1309
1310                         error = sysctl_out_proc(p, req, flags | doingzomb);
1311                         if (error) {
1312                                 sx_sunlock(&allproc_lock);
1313                                 return (error);
1314                         }
1315                 }
1316         }
1317         sx_sunlock(&allproc_lock);
1318         return (0);
1319 }
1320
1321 struct pargs *
1322 pargs_alloc(int len)
1323 {
1324         struct pargs *pa;
1325
1326         pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1327                 M_WAITOK);
1328         refcount_init(&pa->ar_ref, 1);
1329         pa->ar_length = len;
1330         return (pa);
1331 }
1332
1333 static void
1334 pargs_free(struct pargs *pa)
1335 {
1336
1337         free(pa, M_PARGS);
1338 }
1339
1340 void
1341 pargs_hold(struct pargs *pa)
1342 {
1343
1344         if (pa == NULL)
1345                 return;
1346         refcount_acquire(&pa->ar_ref);
1347 }
1348
1349 void
1350 pargs_drop(struct pargs *pa)
1351 {
1352
1353         if (pa == NULL)
1354                 return;
1355         if (refcount_release(&pa->ar_ref))
1356                 pargs_free(pa);
1357 }
1358
1359 /*
1360  * This sysctl allows a process to retrieve the argument list or process
1361  * title for another process without groping around in the address space
1362  * of the other process.  It also allow a process to set its own "process 
1363  * title to a string of its own choice.
1364  */
1365 static int
1366 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
1367 {
1368         int *name = (int*) arg1;
1369         u_int namelen = arg2;
1370         struct pargs *newpa, *pa;
1371         struct proc *p;
1372         int error = 0;
1373
1374         if (namelen != 1) 
1375                 return (EINVAL);
1376
1377         p = pfind((pid_t)name[0]);
1378         if (!p)
1379                 return (ESRCH);
1380
1381         if ((error = p_cansee(curthread, p)) != 0) {
1382                 PROC_UNLOCK(p);
1383                 return (error);
1384         }
1385
1386         if (req->newptr && curproc != p) {
1387                 PROC_UNLOCK(p);
1388                 return (EPERM);
1389         }
1390
1391         pa = p->p_args;
1392         pargs_hold(pa);
1393         PROC_UNLOCK(p);
1394         if (pa != NULL)
1395                 error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
1396         pargs_drop(pa);
1397         if (error != 0 || req->newptr == NULL)
1398                 return (error);
1399
1400         if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit)
1401                 return (ENOMEM);
1402         newpa = pargs_alloc(req->newlen);
1403         error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
1404         if (error != 0) {
1405                 pargs_free(newpa);
1406                 return (error);
1407         }
1408         PROC_LOCK(p);
1409         pa = p->p_args;
1410         p->p_args = newpa;
1411         PROC_UNLOCK(p);
1412         pargs_drop(pa);
1413         return (0);
1414 }
1415
1416 /*
1417  * This sysctl allows a process to retrieve the path of the executable for
1418  * itself or another process.
1419  */
1420 static int
1421 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
1422 {
1423         pid_t *pidp = (pid_t *)arg1;
1424         unsigned int arglen = arg2;
1425         struct proc *p;
1426         struct vnode *vp;
1427         char *retbuf, *freebuf;
1428         int error, vfslocked;
1429
1430         if (arglen != 1)
1431                 return (EINVAL);
1432         if (*pidp == -1) {      /* -1 means this process */
1433                 p = req->td->td_proc;
1434         } else {
1435                 p = pfind(*pidp);
1436                 if (p == NULL)
1437                         return (ESRCH);
1438                 if ((error = p_cansee(curthread, p)) != 0) {
1439                         PROC_UNLOCK(p);
1440                         return (error);
1441                 }
1442         }
1443
1444         vp = p->p_textvp;
1445         if (vp == NULL) {
1446                 if (*pidp != -1)
1447                         PROC_UNLOCK(p);
1448                 return (0);
1449         }
1450         vref(vp);
1451         if (*pidp != -1)
1452                 PROC_UNLOCK(p);
1453         error = vn_fullpath(req->td, vp, &retbuf, &freebuf);
1454         vfslocked = VFS_LOCK_GIANT(vp->v_mount);
1455         vrele(vp);
1456         VFS_UNLOCK_GIANT(vfslocked);
1457         if (error)
1458                 return (error);
1459         error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
1460         free(freebuf, M_TEMP);
1461         return (error);
1462 }
1463
1464 static int
1465 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
1466 {
1467         struct proc *p;
1468         char *sv_name;
1469         int *name;
1470         int namelen;
1471         int error;
1472
1473         namelen = arg2;
1474         if (namelen != 1) 
1475                 return (EINVAL);
1476
1477         name = (int *)arg1;
1478         if ((p = pfind((pid_t)name[0])) == NULL)
1479                 return (ESRCH);
1480         if ((error = p_cansee(curthread, p))) {
1481                 PROC_UNLOCK(p);
1482                 return (error);
1483         }
1484         sv_name = p->p_sysent->sv_name;
1485         PROC_UNLOCK(p);
1486         return (sysctl_handle_string(oidp, sv_name, 0, req));
1487 }
1488
1489 #ifdef KINFO_OVMENTRY_SIZE
1490 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
1491 #endif
1492
1493 #ifdef COMPAT_FREEBSD7
1494 static int
1495 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
1496 {
1497         vm_map_entry_t entry, tmp_entry;
1498         unsigned int last_timestamp;
1499         char *fullpath, *freepath;
1500         struct kinfo_ovmentry *kve;
1501         struct vattr va;
1502         struct ucred *cred;
1503         int error, *name;
1504         struct vnode *vp;
1505         struct proc *p;
1506         vm_map_t map;
1507         struct vmspace *vm;
1508
1509         name = (int *)arg1;
1510         if ((p = pfind((pid_t)name[0])) == NULL)
1511                 return (ESRCH);
1512         if (p->p_flag & P_WEXIT) {
1513                 PROC_UNLOCK(p);
1514                 return (ESRCH);
1515         }
1516         if ((error = p_candebug(curthread, p))) {
1517                 PROC_UNLOCK(p);
1518                 return (error);
1519         }
1520         _PHOLD(p);
1521         PROC_UNLOCK(p);
1522         vm = vmspace_acquire_ref(p);
1523         if (vm == NULL) {
1524                 PRELE(p);
1525                 return (ESRCH);
1526         }
1527         kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
1528
1529         map = &p->p_vmspace->vm_map;    /* XXXRW: More locking required? */
1530         vm_map_lock_read(map);
1531         for (entry = map->header.next; entry != &map->header;
1532             entry = entry->next) {
1533                 vm_object_t obj, tobj, lobj;
1534                 vm_offset_t addr;
1535                 int vfslocked;
1536
1537                 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
1538                         continue;
1539
1540                 bzero(kve, sizeof(*kve));
1541                 kve->kve_structsize = sizeof(*kve);
1542
1543                 kve->kve_private_resident = 0;
1544                 obj = entry->object.vm_object;
1545                 if (obj != NULL) {
1546                         VM_OBJECT_LOCK(obj);
1547                         if (obj->shadow_count == 1)
1548                                 kve->kve_private_resident =
1549                                     obj->resident_page_count;
1550                 }
1551                 kve->kve_resident = 0;
1552                 addr = entry->start;
1553                 while (addr < entry->end) {
1554                         if (pmap_extract(map->pmap, addr))
1555                                 kve->kve_resident++;
1556                         addr += PAGE_SIZE;
1557                 }
1558
1559                 for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
1560                         if (tobj != obj)
1561                                 VM_OBJECT_LOCK(tobj);
1562                         if (lobj != obj)
1563                                 VM_OBJECT_UNLOCK(lobj);
1564                         lobj = tobj;
1565                 }
1566
1567                 kve->kve_start = (void*)entry->start;
1568                 kve->kve_end = (void*)entry->end;
1569                 kve->kve_offset = (off_t)entry->offset;
1570
1571                 if (entry->protection & VM_PROT_READ)
1572                         kve->kve_protection |= KVME_PROT_READ;
1573                 if (entry->protection & VM_PROT_WRITE)
1574                         kve->kve_protection |= KVME_PROT_WRITE;
1575                 if (entry->protection & VM_PROT_EXECUTE)
1576                         kve->kve_protection |= KVME_PROT_EXEC;
1577
1578                 if (entry->eflags & MAP_ENTRY_COW)
1579                         kve->kve_flags |= KVME_FLAG_COW;
1580                 if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
1581                         kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
1582                 if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
1583                         kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
1584
1585                 last_timestamp = map->timestamp;
1586                 vm_map_unlock_read(map);
1587
1588                 kve->kve_fileid = 0;
1589                 kve->kve_fsid = 0;
1590                 freepath = NULL;
1591                 fullpath = "";
1592                 if (lobj) {
1593                         vp = NULL;
1594                         switch (lobj->type) {
1595                         case OBJT_DEFAULT:
1596                                 kve->kve_type = KVME_TYPE_DEFAULT;
1597                                 break;
1598                         case OBJT_VNODE:
1599                                 kve->kve_type = KVME_TYPE_VNODE;
1600                                 vp = lobj->handle;
1601                                 vref(vp);
1602                                 break;
1603                         case OBJT_SWAP:
1604                                 kve->kve_type = KVME_TYPE_SWAP;
1605                                 break;
1606                         case OBJT_DEVICE:
1607                                 kve->kve_type = KVME_TYPE_DEVICE;
1608                                 break;
1609                         case OBJT_PHYS:
1610                                 kve->kve_type = KVME_TYPE_PHYS;
1611                                 break;
1612                         case OBJT_DEAD:
1613                                 kve->kve_type = KVME_TYPE_DEAD;
1614                                 break;
1615                         case OBJT_SG:
1616                                 kve->kve_type = KVME_TYPE_SG;
1617                                 break;
1618                         default:
1619                                 kve->kve_type = KVME_TYPE_UNKNOWN;
1620                                 break;
1621                         }
1622                         if (lobj != obj)
1623                                 VM_OBJECT_UNLOCK(lobj);
1624
1625                         kve->kve_ref_count = obj->ref_count;
1626                         kve->kve_shadow_count = obj->shadow_count;
1627                         VM_OBJECT_UNLOCK(obj);
1628                         if (vp != NULL) {
1629                                 vn_fullpath(curthread, vp, &fullpath,
1630                                     &freepath);
1631                                 cred = curthread->td_ucred;
1632                                 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
1633                                 vn_lock(vp, LK_SHARED | LK_RETRY);
1634                                 if (VOP_GETATTR(vp, &va, cred) == 0) {
1635                                         kve->kve_fileid = va.va_fileid;
1636                                         kve->kve_fsid = va.va_fsid;
1637                                 }
1638                                 vput(vp);
1639                                 VFS_UNLOCK_GIANT(vfslocked);
1640                         }
1641                 } else {
1642                         kve->kve_type = KVME_TYPE_NONE;
1643                         kve->kve_ref_count = 0;
1644                         kve->kve_shadow_count = 0;
1645                 }
1646
1647                 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
1648                 if (freepath != NULL)
1649                         free(freepath, M_TEMP);
1650
1651                 error = SYSCTL_OUT(req, kve, sizeof(*kve));
1652                 vm_map_lock_read(map);
1653                 if (error)
1654                         break;
1655                 if (last_timestamp != map->timestamp) {
1656                         vm_map_lookup_entry(map, addr - 1, &tmp_entry);
1657                         entry = tmp_entry;
1658                 }
1659         }
1660         vm_map_unlock_read(map);
1661         vmspace_free(vm);
1662         PRELE(p);
1663         free(kve, M_TEMP);
1664         return (error);
1665 }
1666 #endif  /* COMPAT_FREEBSD7 */
1667
1668 #ifdef KINFO_VMENTRY_SIZE
1669 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
1670 #endif
1671
1672 static int
1673 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
1674 {
1675         vm_map_entry_t entry, tmp_entry;
1676         unsigned int last_timestamp;
1677         char *fullpath, *freepath;
1678         struct kinfo_vmentry *kve;
1679         struct vattr va;
1680         struct ucred *cred;
1681         int error, *name;
1682         struct vnode *vp;
1683         struct proc *p;
1684         struct vmspace *vm;
1685         vm_map_t map;
1686
1687         name = (int *)arg1;
1688         if ((p = pfind((pid_t)name[0])) == NULL)
1689                 return (ESRCH);
1690         if (p->p_flag & P_WEXIT) {
1691                 PROC_UNLOCK(p);
1692                 return (ESRCH);
1693         }
1694         if ((error = p_candebug(curthread, p))) {
1695                 PROC_UNLOCK(p);
1696                 return (error);
1697         }
1698         _PHOLD(p);
1699         PROC_UNLOCK(p);
1700         vm = vmspace_acquire_ref(p);
1701         if (vm == NULL) {
1702                 PRELE(p);
1703                 return (ESRCH);
1704         }
1705         kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
1706
1707         map = &vm->vm_map;      /* XXXRW: More locking required? */
1708         vm_map_lock_read(map);
1709         for (entry = map->header.next; entry != &map->header;
1710             entry = entry->next) {
1711                 vm_object_t obj, tobj, lobj;
1712                 vm_offset_t addr;
1713                 int vfslocked;
1714
1715                 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
1716                         continue;
1717
1718                 bzero(kve, sizeof(*kve));
1719
1720                 kve->kve_private_resident = 0;
1721                 obj = entry->object.vm_object;
1722                 if (obj != NULL) {
1723                         VM_OBJECT_LOCK(obj);
1724                         if (obj->shadow_count == 1)
1725                                 kve->kve_private_resident =
1726                                     obj->resident_page_count;
1727                 }
1728                 kve->kve_resident = 0;
1729                 addr = entry->start;
1730                 while (addr < entry->end) {
1731                         if (pmap_extract(map->pmap, addr))
1732                                 kve->kve_resident++;
1733                         addr += PAGE_SIZE;
1734                 }
1735
1736                 for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
1737                         if (tobj != obj)
1738                                 VM_OBJECT_LOCK(tobj);
1739                         if (lobj != obj)
1740                                 VM_OBJECT_UNLOCK(lobj);
1741                         lobj = tobj;
1742                 }
1743
1744                 kve->kve_start = entry->start;
1745                 kve->kve_end = entry->end;
1746                 kve->kve_offset = entry->offset;
1747
1748                 if (entry->protection & VM_PROT_READ)
1749                         kve->kve_protection |= KVME_PROT_READ;
1750                 if (entry->protection & VM_PROT_WRITE)
1751                         kve->kve_protection |= KVME_PROT_WRITE;
1752                 if (entry->protection & VM_PROT_EXECUTE)
1753                         kve->kve_protection |= KVME_PROT_EXEC;
1754
1755                 if (entry->eflags & MAP_ENTRY_COW)
1756                         kve->kve_flags |= KVME_FLAG_COW;
1757                 if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
1758                         kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
1759                 if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
1760                         kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
1761
1762                 last_timestamp = map->timestamp;
1763                 vm_map_unlock_read(map);
1764
1765                 freepath = NULL;
1766                 fullpath = "";
1767                 if (lobj) {
1768                         vp = NULL;
1769                         switch (lobj->type) {
1770                         case OBJT_DEFAULT:
1771                                 kve->kve_type = KVME_TYPE_DEFAULT;
1772                                 break;
1773                         case OBJT_VNODE:
1774                                 kve->kve_type = KVME_TYPE_VNODE;
1775                                 vp = lobj->handle;
1776                                 vref(vp);
1777                                 break;
1778                         case OBJT_SWAP:
1779                                 kve->kve_type = KVME_TYPE_SWAP;
1780                                 break;
1781                         case OBJT_DEVICE:
1782                                 kve->kve_type = KVME_TYPE_DEVICE;
1783                                 break;
1784                         case OBJT_PHYS:
1785                                 kve->kve_type = KVME_TYPE_PHYS;
1786                                 break;
1787                         case OBJT_DEAD:
1788                                 kve->kve_type = KVME_TYPE_DEAD;
1789                                 break;
1790                         case OBJT_SG:
1791                                 kve->kve_type = KVME_TYPE_SG;
1792                                 break;
1793                         default:
1794                                 kve->kve_type = KVME_TYPE_UNKNOWN;
1795                                 break;
1796                         }
1797                         if (lobj != obj)
1798                                 VM_OBJECT_UNLOCK(lobj);
1799
1800                         kve->kve_ref_count = obj->ref_count;
1801                         kve->kve_shadow_count = obj->shadow_count;
1802                         VM_OBJECT_UNLOCK(obj);
1803                         if (vp != NULL) {
1804                                 vn_fullpath(curthread, vp, &fullpath,
1805                                     &freepath);
1806                                 kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
1807                                 cred = curthread->td_ucred;
1808                                 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
1809                                 vn_lock(vp, LK_SHARED | LK_RETRY);
1810                                 if (VOP_GETATTR(vp, &va, cred) == 0) {
1811                                         kve->kve_vn_fileid = va.va_fileid;
1812                                         kve->kve_vn_fsid = va.va_fsid;
1813                                         kve->kve_vn_mode =
1814                                             MAKEIMODE(va.va_type, va.va_mode);
1815                                         kve->kve_vn_size = va.va_size;
1816                                         kve->kve_vn_rdev = va.va_rdev;
1817                                         kve->kve_status = KF_ATTR_VALID;
1818                                 }
1819                                 vput(vp);
1820                                 VFS_UNLOCK_GIANT(vfslocked);
1821                         }
1822                 } else {
1823                         kve->kve_type = KVME_TYPE_NONE;
1824                         kve->kve_ref_count = 0;
1825                         kve->kve_shadow_count = 0;
1826                 }
1827
1828                 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
1829                 if (freepath != NULL)
1830                         free(freepath, M_TEMP);
1831
1832                 /* Pack record size down */
1833                 kve->kve_structsize = offsetof(struct kinfo_vmentry, kve_path) +
1834                     strlen(kve->kve_path) + 1;
1835                 kve->kve_structsize = roundup(kve->kve_structsize,
1836                     sizeof(uint64_t));
1837                 error = SYSCTL_OUT(req, kve, kve->kve_structsize);
1838                 vm_map_lock_read(map);
1839                 if (error)
1840                         break;
1841                 if (last_timestamp != map->timestamp) {
1842                         vm_map_lookup_entry(map, addr - 1, &tmp_entry);
1843                         entry = tmp_entry;
1844                 }
1845         }
1846         vm_map_unlock_read(map);
1847         vmspace_free(vm);
1848         PRELE(p);
1849         free(kve, M_TEMP);
1850         return (error);
1851 }
1852
1853 #if defined(STACK) || defined(DDB)
1854 static int
1855 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
1856 {
1857         struct kinfo_kstack *kkstp;
1858         int error, i, *name, numthreads;
1859         lwpid_t *lwpidarray;
1860         struct thread *td;
1861         struct stack *st;
1862         struct sbuf sb;
1863         struct proc *p;
1864
1865         name = (int *)arg1;
1866         if ((p = pfind((pid_t)name[0])) == NULL)
1867                 return (ESRCH);
1868         /* XXXRW: Not clear ESRCH is the right error during proc execve(). */
1869         if (p->p_flag & P_WEXIT || p->p_flag & P_INEXEC) {
1870                 PROC_UNLOCK(p);
1871                 return (ESRCH);
1872         }
1873         if ((error = p_candebug(curthread, p))) {
1874                 PROC_UNLOCK(p);
1875                 return (error);
1876         }
1877         _PHOLD(p);
1878         PROC_UNLOCK(p);
1879
1880         kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
1881         st = stack_create();
1882
1883         lwpidarray = NULL;
1884         numthreads = 0;
1885         PROC_LOCK(p);
1886 repeat:
1887         if (numthreads < p->p_numthreads) {
1888                 if (lwpidarray != NULL) {
1889                         free(lwpidarray, M_TEMP);
1890                         lwpidarray = NULL;
1891                 }
1892                 numthreads = p->p_numthreads;
1893                 PROC_UNLOCK(p);
1894                 lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
1895                     M_WAITOK | M_ZERO);
1896                 PROC_LOCK(p);
1897                 goto repeat;
1898         }
1899         i = 0;
1900
1901         /*
1902          * XXXRW: During the below loop, execve(2) and countless other sorts
1903          * of changes could have taken place.  Should we check to see if the
1904          * vmspace has been replaced, or the like, in order to prevent
1905          * giving a snapshot that spans, say, execve(2), with some threads
1906          * before and some after?  Among other things, the credentials could
1907          * have changed, in which case the right to extract debug info might
1908          * no longer be assured.
1909          */
1910         FOREACH_THREAD_IN_PROC(p, td) {
1911                 KASSERT(i < numthreads,
1912                     ("sysctl_kern_proc_kstack: numthreads"));
1913                 lwpidarray[i] = td->td_tid;
1914                 i++;
1915         }
1916         numthreads = i;
1917         for (i = 0; i < numthreads; i++) {
1918                 td = thread_find(p, lwpidarray[i]);
1919                 if (td == NULL) {
1920                         continue;
1921                 }
1922                 bzero(kkstp, sizeof(*kkstp));
1923                 (void)sbuf_new(&sb, kkstp->kkst_trace,
1924                     sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
1925                 thread_lock(td);
1926                 kkstp->kkst_tid = td->td_tid;
1927                 if (TD_IS_SWAPPED(td))
1928                         kkstp->kkst_state = KKST_STATE_SWAPPED;
1929                 else if (TD_IS_RUNNING(td))
1930                         kkstp->kkst_state = KKST_STATE_RUNNING;
1931                 else {
1932                         kkstp->kkst_state = KKST_STATE_STACKOK;
1933                         stack_save_td(st, td);
1934                 }
1935                 thread_unlock(td);
1936                 PROC_UNLOCK(p);
1937                 stack_sbuf_print(&sb, st);
1938                 sbuf_finish(&sb);
1939                 sbuf_delete(&sb);
1940                 error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
1941                 PROC_LOCK(p);
1942                 if (error)
1943                         break;
1944         }
1945         _PRELE(p);
1946         PROC_UNLOCK(p);
1947         if (lwpidarray != NULL)
1948                 free(lwpidarray, M_TEMP);
1949         stack_destroy(st);
1950         free(kkstp, M_TEMP);
1951         return (error);
1952 }
1953 #endif
1954
1955 /*
1956  * This sysctl allows a process to retrieve the full list of groups from
1957  * itself or another process.
1958  */
1959 static int
1960 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
1961 {
1962         pid_t *pidp = (pid_t *)arg1;
1963         unsigned int arglen = arg2;
1964         struct proc *p;
1965         struct ucred *cred;
1966         int error;
1967
1968         if (arglen != 1)
1969                 return (EINVAL);
1970         if (*pidp == -1) {      /* -1 means this process */
1971                 p = req->td->td_proc;
1972         } else {
1973                 p = pfind(*pidp);
1974                 if (p == NULL)
1975                         return (ESRCH);
1976                 if ((error = p_cansee(curthread, p)) != 0) {
1977                         PROC_UNLOCK(p);
1978                         return (error);
1979                 }
1980         }
1981
1982         cred = crhold(p->p_ucred);
1983         if (*pidp != -1)
1984                 PROC_UNLOCK(p);
1985
1986         error = SYSCTL_OUT(req, cred->cr_groups,
1987             cred->cr_ngroups * sizeof(gid_t));
1988         crfree(cred);
1989         return (error);
1990 }
1991
1992 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD,  0, "Process table");
1993
1994 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
1995         CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
1996         "Return entire process table");
1997
1998 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
1999         sysctl_kern_proc, "Process table");
2000
2001 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
2002         sysctl_kern_proc, "Process table");
2003
2004 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2005         sysctl_kern_proc, "Process table");
2006
2007 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
2008         CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2009
2010 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE, 
2011         sysctl_kern_proc, "Process table");
2012
2013 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE, 
2014         sysctl_kern_proc, "Process table");
2015
2016 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2017         sysctl_kern_proc, "Process table");
2018
2019 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2020         sysctl_kern_proc, "Process table");
2021
2022 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
2023         sysctl_kern_proc, "Return process table, no threads");
2024
2025 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
2026         CTLFLAG_RW | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
2027         sysctl_kern_proc_args, "Process argument list");
2028
2029 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
2030         CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
2031
2032 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
2033         CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
2034         "Process syscall vector name (ABI type)");
2035
2036 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
2037         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2038
2039 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
2040         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2041
2042 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
2043         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2044
2045 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
2046         sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2047
2048 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
2049         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2050
2051 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
2052         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2053
2054 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
2055         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2056
2057 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
2058         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2059
2060 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
2061         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
2062         "Return process table, no threads");
2063
2064 #ifdef COMPAT_FREEBSD7
2065 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
2066         CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
2067 #endif
2068
2069 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
2070         CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
2071
2072 #if defined(STACK) || defined(DDB)
2073 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
2074         CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
2075 #endif
2076
2077 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
2078         CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");