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     "Kernel stack size in pages");
 153
 154 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
 155 #ifdef COMPAT_FREEBSD32
 156 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
 157 #endif
 158
 159 /*
 160  * Initialize global process hashing structures.
 161  */
 162 void
 163 procinit()
 164 {
 165
 166         sx_init(&allproc_lock, "allproc");
 167         sx_init(&proctree_lock, "proctree");
 168         mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
 169         LIST_INIT(&allproc);
 170         LIST_INIT(&zombproc);
 171         pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
 172         pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
 173         proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
 174             proc_ctor, proc_dtor, proc_init, proc_fini,
 175             UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
 176         uihashinit();
 177 }
 178
 179 /*
 180  * Prepare a proc for use.
 181  */
 182 static int
 183 proc_ctor(void *mem, int size, void *arg, int flags)
 184 {
 185         struct proc *p;
 186
 187         p = (struct proc *)mem;
 188         SDT_PROBE(proc, kernel, ctor , entry, p, size, arg, flags, 0);
 189         EVENTHANDLER_INVOKE(process_ctor, p);
 190         SDT_PROBE(proc, kernel, ctor , return, p, size, arg, flags, 0);
 191         return (0);
 192 }
 193
 194 /*
 195  * Reclaim a proc after use.
 196  */
 197 static void
 198 proc_dtor(void *mem, int size, void *arg)
 199 {
 200         struct proc *p;
 201         struct thread *td;
 202
 203         /* INVARIANTS checks go here */
 204         p = (struct proc *)mem;
 205         td = FIRST_THREAD_IN_PROC(p);
 206         SDT_PROBE(proc, kernel, dtor, entry, p, size, arg, td, 0);
 207         if (td != NULL) {
 208 #ifdef INVARIANTS
 209                 KASSERT((p->p_numthreads == 1),
 210                     ("bad number of threads in exiting process"));
 211                 KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
 212 #endif
 213                 /* Free all OSD associated to this thread. */
 214                 osd_thread_exit(td);
 215         }
 216         EVENTHANDLER_INVOKE(process_dtor, p);
 217         if (p->p_ksi != NULL)
 218                 KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
 219         SDT_PROBE(proc, kernel, dtor, return, p, size, arg, 0, 0);
 220 }
 221
 222 /*
 223  * Initialize type-stable parts of a proc (when newly created).
 224  */
 225 static int
 226 proc_init(void *mem, int size, int flags)
 227 {
 228         struct proc *p;
 229
 230         p = (struct proc *)mem;
 231         SDT_PROBE(proc, kernel, init, entry, p, size, flags, 0, 0);
 232         p->p_sched = (struct p_sched *)&p[1];
 233         bzero(&p->p_mtx, sizeof(struct mtx));
 234         mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK);
 235         mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_RECURSE);
 236         cv_init(&p->p_pwait, "ppwait");
 237         TAILQ_INIT(&p->p_threads);           /* all threads in proc */
 238         EVENTHANDLER_INVOKE(process_init, p);
 239         p->p_stats = pstats_alloc();
 240         SDT_PROBE(proc, kernel, init, return, p, size, flags, 0, 0);
 241         return (0);
 242 }
 243
 244 /*
 245  * UMA should ensure that this function is never called.
 246  * Freeing a proc structure would violate type stability.
 247  */
 248 static void
 249 proc_fini(void *mem, int size)
 250 {
 251 #ifdef notnow
 252         struct proc *p;
 253
 254         p = (struct proc *)mem;
 255         EVENTHANDLER_INVOKE(process_fini, p);
 256         pstats_free(p->p_stats);
 257         thread_free(FIRST_THREAD_IN_PROC(p));
 258         mtx_destroy(&p->p_mtx);
 259         if (p->p_ksi != NULL)
 260                 ksiginfo_free(p->p_ksi);
 261 #else
 262         panic("proc reclaimed");
 263 #endif
 264 }
 265
 266 /*
 267  * Is p an inferior of the current process?
 268  */
 269 int
 270 inferior(p)
 271         register struct proc *p;
 272 {
 273
 274         sx_assert(&proctree_lock, SX_LOCKED);
 275         for (; p != curproc; p = p->p_pptr)
 276                 if (p->p_pid == 0)
 277                         return (0);
 278         return (1);
 279 }
 280
 281 /*
 282  * Locate a process by number; return only "live" processes -- i.e., neither
 283  * zombies nor newly born but incompletely initialized processes.  By not
 284  * returning processes in the PRS_NEW state, we allow callers to avoid
 285  * testing for that condition to avoid dereferencing p_ucred, et al.
 286  */
 287 struct proc *
 288 pfind(pid)
 289         register pid_t pid;
 290 {
 291         register struct proc *p;
 292
 293         sx_slock(&allproc_lock);
 294         LIST_FOREACH(p, PIDHASH(pid), p_hash)
 295                 if (p->p_pid == pid) {
 296                         if (p->p_state == PRS_NEW) {
 297                                 p = NULL;
 298                                 break;
 299                         }
 300                         PROC_LOCK(p);
 301                         break;
 302                 }
 303         sx_sunlock(&allproc_lock);
 304         return (p);
 305 }
 306
 307 /*
 308  * Locate a process group by number.
 309  * The caller must hold proctree_lock.
 310  */
 311 struct pgrp *
 312 pgfind(pgid)
 313         register pid_t pgid;
 314 {
 315         register struct pgrp *pgrp;
 316
 317         sx_assert(&proctree_lock, SX_LOCKED);
 318
 319         LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
 320                 if (pgrp->pg_id == pgid) {
 321                         PGRP_LOCK(pgrp);
 322                         return (pgrp);
 323                 }
 324         }
 325         return (NULL);
 326 }
 327
 328 /*
 329  * Create a new process group.
 330  * pgid must be equal to the pid of p.
 331  * Begin a new session if required.
 332  */
 333 int
 334 enterpgrp(p, pgid, pgrp, sess)
 335         register struct proc *p;
 336         pid_t pgid;
 337         struct pgrp *pgrp;
 338         struct session *sess;
 339 {
 340         struct pgrp *pgrp2;
 341
 342         sx_assert(&proctree_lock, SX_XLOCKED);
 343
 344         KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
 345         KASSERT(p->p_pid == pgid,
 346             ("enterpgrp: new pgrp and pid != pgid"));
 347
 348         pgrp2 = pgfind(pgid);
 349
 350         KASSERT(pgrp2 == NULL,
 351             ("enterpgrp: pgrp with pgid exists"));
 352         KASSERT(!SESS_LEADER(p),
 353             ("enterpgrp: session leader attempted setpgrp"));
 354
 355         mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
 356
 357         if (sess != NULL) {
 358                 /*
 359                  * new session
 360                  */
 361                 mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
 362                 PROC_LOCK(p);
 363                 p->p_flag &= ~P_CONTROLT;
 364                 PROC_UNLOCK(p);
 365                 PGRP_LOCK(pgrp);
 366                 sess->s_leader = p;
 367                 sess->s_sid = p->p_pid;
 368                 refcount_init(&sess->s_count, 1);
 369                 sess->s_ttyvp = NULL;
 370                 sess->s_ttyp = NULL;
 371                 bcopy(p->p_session->s_login, sess->s_login,
 372                             sizeof(sess->s_login));
 373                 pgrp->pg_session = sess;
 374                 KASSERT(p == curproc,
 375                     ("enterpgrp: mksession and p != curproc"));
 376         } else {
 377                 pgrp->pg_session = p->p_session;
 378                 sess_hold(pgrp->pg_session);
 379                 PGRP_LOCK(pgrp);
 380         }
 381         pgrp->pg_id = pgid;
 382         LIST_INIT(&pgrp->pg_members);
 383
 384         /*
 385          * As we have an exclusive lock of proctree_lock,
 386          * this should not deadlock.
 387          */
 388         LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
 389         pgrp->pg_jobc = 0;
 390         SLIST_INIT(&pgrp->pg_sigiolst);
 391         PGRP_UNLOCK(pgrp);
 392
 393         doenterpgrp(p, pgrp);
 394
 395         return (0);
 396 }
 397
 398 /*
 399  * Move p to an existing process group
 400  */
 401 int
 402 enterthispgrp(p, pgrp)
 403         register struct proc *p;
 404         struct pgrp *pgrp;
 405 {
 406
 407         sx_assert(&proctree_lock, SX_XLOCKED);
 408         PROC_LOCK_ASSERT(p, MA_NOTOWNED);
 409         PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
 410         PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
 411         SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
 412         KASSERT(pgrp->pg_session == p->p_session,
 413                 ("%s: pgrp's session %p, p->p_session %p.\n",
 414                 __func__,
 415                 pgrp->pg_session,
 416                 p->p_session));
 417         KASSERT(pgrp != p->p_pgrp,
 418                 ("%s: p belongs to pgrp.", __func__));
 419
 420         doenterpgrp(p, pgrp);
 421
 422         return (0);
 423 }
 424
 425 /*
 426  * Move p to a process group
 427  */
 428 static void
 429 doenterpgrp(p, pgrp)
 430         struct proc *p;
 431         struct pgrp *pgrp;
 432 {
 433         struct pgrp *savepgrp;
 434
 435         sx_assert(&proctree_lock, SX_XLOCKED);
 436         PROC_LOCK_ASSERT(p, MA_NOTOWNED);
 437         PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
 438         PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
 439         SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
 440
 441         savepgrp = p->p_pgrp;
 442
 443         /*
 444          * Adjust eligibility of affected pgrps to participate in job control.
 445          * Increment eligibility counts before decrementing, otherwise we
 446          * could reach 0 spuriously during the first call.
 447          */
 448         fixjobc(p, pgrp, 1);
 449         fixjobc(p, p->p_pgrp, 0);
 450
 451         PGRP_LOCK(pgrp);
 452         PGRP_LOCK(savepgrp);
 453         PROC_LOCK(p);
 454         LIST_REMOVE(p, p_pglist);
 455         p->p_pgrp = pgrp;
 456         PROC_UNLOCK(p);
 457         LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
 458         PGRP_UNLOCK(savepgrp);
 459         PGRP_UNLOCK(pgrp);
 460         if (LIST_EMPTY(&savepgrp->pg_members))
 461                 pgdelete(savepgrp);
 462 }
 463
 464 /*
 465  * remove process from process group
 466  */
 467 int
 468 leavepgrp(p)
 469         register struct proc *p;
 470 {
 471         struct pgrp *savepgrp;
 472
 473         sx_assert(&proctree_lock, SX_XLOCKED);
 474         savepgrp = p->p_pgrp;
 475         PGRP_LOCK(savepgrp);
 476         PROC_LOCK(p);
 477         LIST_REMOVE(p, p_pglist);
 478         p->p_pgrp = NULL;
 479         PROC_UNLOCK(p);
 480         PGRP_UNLOCK(savepgrp);
 481         if (LIST_EMPTY(&savepgrp->pg_members))
 482                 pgdelete(savepgrp);
 483         return (0);
 484 }
 485
 486 /*
 487  * delete a process group
 488  */
 489 static void
 490 pgdelete(pgrp)
 491         register struct pgrp *pgrp;
 492 {
 493         struct session *savesess;
 494         struct tty *tp;
 495
 496         sx_assert(&proctree_lock, SX_XLOCKED);
 497         PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
 498         SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
 499
 500         /*
 501          * Reset any sigio structures pointing to us as a result of
 502          * F_SETOWN with our pgid.
 503          */
 504         funsetownlst(&pgrp->pg_sigiolst);
 505
 506         PGRP_LOCK(pgrp);
 507         tp = pgrp->pg_session->s_ttyp;
 508         LIST_REMOVE(pgrp, pg_hash);
 509         savesess = pgrp->pg_session;
 510         PGRP_UNLOCK(pgrp);
 511
 512         /* Remove the reference to the pgrp before deallocating it. */
 513         if (tp != NULL) {
 514                 tty_lock(tp);
 515                 tty_rel_pgrp(tp, pgrp);
 516         }
 517
 518         mtx_destroy(&pgrp->pg_mtx);
 519         free(pgrp, M_PGRP);
 520         sess_release(savesess);
 521 }
 522
 523 static void
 524 pgadjustjobc(pgrp, entering)
 525         struct pgrp *pgrp;
 526         int entering;
 527 {
 528
 529         PGRP_LOCK(pgrp);
 530         if (entering)
 531                 pgrp->pg_jobc++;
 532         else {
 533                 --pgrp->pg_jobc;
 534                 if (pgrp->pg_jobc == 0)
 535                         orphanpg(pgrp);
 536         }
 537         PGRP_UNLOCK(pgrp);
 538 }
 539
 540 /*
 541  * Adjust pgrp jobc counters when specified process changes process group.
 542  * We count the number of processes in each process group that "qualify"
 543  * the group for terminal job control (those with a parent in a different
 544  * process group of the same session).  If that count reaches zero, the
 545  * process group becomes orphaned.  Check both the specified process'
 546  * process group and that of its children.
 547  * entering == 0 => p is leaving specified group.
 548  * entering == 1 => p is entering specified group.
 549  */
 550 void
 551 fixjobc(p, pgrp, entering)
 552         register struct proc *p;
 553         register struct pgrp *pgrp;
 554         int entering;
 555 {
 556         register struct pgrp *hispgrp;
 557         register struct session *mysession;
 558
 559         sx_assert(&proctree_lock, SX_LOCKED);
 560         PROC_LOCK_ASSERT(p, MA_NOTOWNED);
 561         PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
 562         SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
 563
 564         /*
 565          * Check p's parent to see whether p qualifies its own process
 566          * group; if so, adjust count for p's process group.
 567          */
 568         mysession = pgrp->pg_session;
 569         if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
 570             hispgrp->pg_session == mysession)
 571                 pgadjustjobc(pgrp, entering);
 572
 573         /*
 574          * Check this process' children to see whether they qualify
 575          * their process groups; if so, adjust counts for children's
 576          * process groups.
 577          */
 578         LIST_FOREACH(p, &p->p_children, p_sibling) {
 579                 hispgrp = p->p_pgrp;
 580                 if (hispgrp == pgrp ||
 581                     hispgrp->pg_session != mysession)
 582                         continue;
 583                 PROC_LOCK(p);
 584                 if (p->p_state == PRS_ZOMBIE) {
 585                         PROC_UNLOCK(p);
 586                         continue;
 587                 }
 588                 PROC_UNLOCK(p);
 589                 pgadjustjobc(hispgrp, entering);
 590         }
 591 }
 592
 593 /*
 594  * A process group has become orphaned;
 595  * if there are any stopped processes in the group,
 596  * hang-up all process in that group.
 597  */
 598 static void
 599 orphanpg(pg)
 600         struct pgrp *pg;
 601 {
 602         register struct proc *p;
 603
 604         PGRP_LOCK_ASSERT(pg, MA_OWNED);
 605
 606         LIST_FOREACH(p, &pg->pg_members, p_pglist) {
 607                 PROC_LOCK(p);
 608                 if (P_SHOULDSTOP(p)) {
 609                         PROC_UNLOCK(p);
 610                         LIST_FOREACH(p, &pg->pg_members, p_pglist) {
 611                                 PROC_LOCK(p);
 612                                 psignal(p, SIGHUP);
 613                                 psignal(p, SIGCONT);
 614                                 PROC_UNLOCK(p);
 615                         }
 616                         return;
 617                 }
 618                 PROC_UNLOCK(p);
 619         }
 620 }
 621
 622 void
 623 sess_hold(struct session *s)
 624 {
 625
 626         refcount_acquire(&s->s_count);
 627 }
 628
 629 void
 630 sess_release(struct session *s)
 631 {
 632
 633         if (refcount_release(&s->s_count)) {
 634                 if (s->s_ttyp != NULL) {
 635                         tty_lock(s->s_ttyp);
 636                         tty_rel_sess(s->s_ttyp, s);
 637                 }
 638                 mtx_destroy(&s->s_mtx);
 639                 free(s, M_SESSION);
 640         }
 641 }
 642
 643 #include "opt_ddb.h"
 644 #ifdef DDB
 645 #include <ddb/ddb.h>
 646
 647 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
 648 {
 649         register struct pgrp *pgrp;
 650         register struct proc *p;
 651         register int i;
 652
 653         for (i = 0; i <= pgrphash; i++) {
 654                 if (!LIST_EMPTY(&pgrphashtbl[i])) {
 655                         printf("\tindx %d\n", i);
 656                         LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
 657                                 printf(
 658                         "\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
 659                                     (void *)pgrp, (long)pgrp->pg_id,
 660                                     (void *)pgrp->pg_session,
 661                                     pgrp->pg_session->s_count,
 662                                     (void *)LIST_FIRST(&pgrp->pg_members));
 663                                 LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
 664                                         printf("\t\tpid %ld addr %p pgrp %p\n",
 665                                             (long)p->p_pid, (void *)p,
 666                                             (void *)p->p_pgrp);
 667                                 }
 668                         }
 669                 }
 670         }
 671 }
 672 #endif /* DDB */
 673
 674 /*
 675  * Calculate the kinfo_proc members which contain process-wide
 676  * informations.
 677  * Must be called with the target process locked.
 678  */
 679 static void
 680 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
 681 {
 682         struct thread *td;
 683
 684         PROC_LOCK_ASSERT(p, MA_OWNED);
 685
 686         kp->ki_estcpu = 0;
 687         kp->ki_pctcpu = 0;
 688         FOREACH_THREAD_IN_PROC(p, td) {
 689                 thread_lock(td);
 690                 kp->ki_pctcpu += sched_pctcpu(td);
 691                 kp->ki_estcpu += td->td_estcpu;
 692                 thread_unlock(td);
 693         }
 694 }
 695
 696 /*
 697  * Clear kinfo_proc and fill in any information that is common
 698  * to all threads in the process.
 699  * Must be called with the target process locked.
 700  */
 701 static void
 702 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
 703 {
 704         struct thread *td0;
 705         struct tty *tp;
 706         struct session *sp;
 707         struct ucred *cred;
 708         struct sigacts *ps;
 709
 710         PROC_LOCK_ASSERT(p, MA_OWNED);
 711         bzero(kp, sizeof(*kp));
 712
 713         kp->ki_structsize = sizeof(*kp);
 714         kp->ki_paddr = p;
 715         kp->ki_addr =/* p->p_addr; */0; /* XXX */
 716         kp->ki_args = p->p_args;
 717         kp->ki_textvp = p->p_textvp;
 718 #ifdef KTRACE
 719         kp->ki_tracep = p->p_tracevp;
 720         mtx_lock(&ktrace_mtx);
 721         kp->ki_traceflag = p->p_traceflag;
 722         mtx_unlock(&ktrace_mtx);
 723 #endif
 724         kp->ki_fd = p->p_fd;
 725         kp->ki_vmspace = p->p_vmspace;
 726         kp->ki_flag = p->p_flag;
 727         cred = p->p_ucred;
 728         if (cred) {
 729                 kp->ki_uid = cred->cr_uid;
 730                 kp->ki_ruid = cred->cr_ruid;
 731                 kp->ki_svuid = cred->cr_svuid;
 732                 kp->ki_cr_flags = cred->cr_flags;
 733                 /* XXX bde doesn't like KI_NGROUPS */
 734                 if (cred->cr_ngroups > KI_NGROUPS) {
 735                         kp->ki_ngroups = KI_NGROUPS;
 736                         kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
 737                 } else
 738                         kp->ki_ngroups = cred->cr_ngroups;
 739                 bcopy(cred->cr_groups, kp->ki_groups,
 740                     kp->ki_ngroups * sizeof(gid_t));
 741                 kp->ki_rgid = cred->cr_rgid;
 742                 kp->ki_svgid = cred->cr_svgid;
 743                 /* If jailed(cred), emulate the old P_JAILED flag. */
 744                 if (jailed(cred)) {
 745                         kp->ki_flag |= P_JAILED;
 746                         /* If inside the jail, use 0 as a jail ID. */
 747                         if (cred->cr_prison != curthread->td_ucred->cr_prison)
 748                                 kp->ki_jid = cred->cr_prison->pr_id;
 749                 }
 750         }
 751         ps = p->p_sigacts;
 752         if (ps) {
 753                 mtx_lock(&ps->ps_mtx);
 754                 kp->ki_sigignore = ps->ps_sigignore;
 755                 kp->ki_sigcatch = ps->ps_sigcatch;
 756                 mtx_unlock(&ps->ps_mtx);
 757         }
 758         PROC_SLOCK(p);
 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         rufetch(p, &kp->ki_rusage);
 785         kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
 786         PROC_SUNLOCK(p);
 787         if ((p->p_flag & P_INMEM) && p->p_stats != NULL) {
 788                 kp->ki_start = p->p_stats->p_start;
 789                 timevaladd(&kp->ki_start, &boottime);
 790                 PROC_SLOCK(p);
 791                 calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
 792                 PROC_SUNLOCK(p);
 793                 calccru(p, &kp->ki_childutime, &kp->ki_childstime);
 794
 795                 /* Some callers want child-times in a single value */
 796                 kp->ki_childtime = kp->ki_childstime;
 797                 timevaladd(&kp->ki_childtime, &kp->ki_childutime);
 798         }
 799         tp = NULL;
 800         if (p->p_pgrp) {
 801                 kp->ki_pgid = p->p_pgrp->pg_id;
 802                 kp->ki_jobc = p->p_pgrp->pg_jobc;
 803                 sp = p->p_pgrp->pg_session;
 804
 805                 if (sp != NULL) {
 806                         kp->ki_sid = sp->s_sid;
 807                         SESS_LOCK(sp);
 808                         strlcpy(kp->ki_login, sp->s_login,
 809                             sizeof(kp->ki_login));
 810                         if (sp->s_ttyvp)
 811                                 kp->ki_kiflag |= KI_CTTY;
 812                         if (SESS_LEADER(p))
 813                                 kp->ki_kiflag |= KI_SLEADER;
 814                         /* XXX proctree_lock */
 815                         tp = sp->s_ttyp;
 816                         SESS_UNLOCK(sp);
 817                 }
 818         }
 819         if ((p->p_flag & P_CONTROLT) && tp != NULL) {
 820                 kp->ki_tdev = tty_udev(tp);
 821                 kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
 822                 if (tp->t_session)
 823                         kp->ki_tsid = tp->t_session->s_sid;
 824         } else
 825                 kp->ki_tdev = NODEV;
 826         if (p->p_comm[0] != '\0')
 827                 strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
 828         if (p->p_sysent && p->p_sysent->sv_name != NULL &&
 829             p->p_sysent->sv_name[0] != '\0')
 830                 strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
 831         kp->ki_siglist = p->p_siglist;
 832         kp->ki_xstat = p->p_xstat;
 833         kp->ki_acflag = p->p_acflag;
 834         kp->ki_lock = p->p_lock;
 835         if (p->p_pptr)
 836                 kp->ki_ppid = p->p_pptr->p_pid;
 837 }
 838
 839 /*
 840  * Fill in information that is thread specific.  Must be called with
 841  * target process locked.  If 'preferthread' is set, overwrite certain
 842  * process-related fields that are maintained for both threads and
 843  * processes.
 844  */
 845 static void
 846 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
 847 {
 848         struct proc *p;
 849
 850         p = td->td_proc;
 851         kp->ki_tdaddr = td;
 852         PROC_LOCK_ASSERT(p, MA_OWNED);
 853
 854         thread_lock(td);
 855         if (td->td_wmesg != NULL)
 856                 strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
 857         else
 858                 bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
 859         strlcpy(kp->ki_ocomm, td->td_name, sizeof(kp->ki_ocomm));
 860         if (TD_ON_LOCK(td)) {
 861                 kp->ki_kiflag |= KI_LOCKBLOCK;
 862                 strlcpy(kp->ki_lockname, td->td_lockname,
 863                     sizeof(kp->ki_lockname));
 864         } else {
 865                 kp->ki_kiflag &= ~KI_LOCKBLOCK;
 866                 bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
 867         }
 868
 869         if (p->p_state == PRS_NORMAL) { /* approximate. */
 870                 if (TD_ON_RUNQ(td) ||
 871                     TD_CAN_RUN(td) ||
 872                     TD_IS_RUNNING(td)) {
 873                         kp->ki_stat = SRUN;
 874                 } else if (P_SHOULDSTOP(p)) {
 875                         kp->ki_stat = SSTOP;
 876                 } else if (TD_IS_SLEEPING(td)) {
 877                         kp->ki_stat = SSLEEP;
 878                 } else if (TD_ON_LOCK(td)) {
 879                         kp->ki_stat = SLOCK;
 880                 } else {
 881                         kp->ki_stat = SWAIT;
 882                 }
 883         } else if (p->p_state == PRS_ZOMBIE) {
 884                 kp->ki_stat = SZOMB;
 885         } else {
 886                 kp->ki_stat = SIDL;
 887         }
 888
 889         /* Things in the thread */
 890         kp->ki_wchan = td->td_wchan;
 891         kp->ki_pri.pri_level = td->td_priority;
 892         kp->ki_pri.pri_native = td->td_base_pri;
 893         kp->ki_lastcpu = td->td_lastcpu;
 894         kp->ki_oncpu = td->td_oncpu;
 895         kp->ki_tdflags = td->td_flags;
 896         kp->ki_tid = td->td_tid;
 897         kp->ki_numthreads = p->p_numthreads;
 898         kp->ki_pcb = td->td_pcb;
 899         kp->ki_kstack = (void *)td->td_kstack;
 900         kp->ki_slptime = (ticks - td->td_slptick) / hz;
 901         kp->ki_pri.pri_class = td->td_pri_class;
 902         kp->ki_pri.pri_user = td->td_user_pri;
 903
 904         if (preferthread) {
 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 }
 918
 919 /*
 920  * Fill in a kinfo_proc structure for the specified process.
 921  * Must be called with the target process locked.
 922  */
 923 void
 924 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
 925 {
 926
 927         MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
 928
 929         fill_kinfo_proc_only(p, kp);
 930         fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
 931         fill_kinfo_aggregate(p, kp);
 932 }
 933
 934 struct pstats *
 935 pstats_alloc(void)
 936 {
 937
 938         return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
 939 }
 940
 941 /*
 942  * Copy parts of p_stats; zero the rest of p_stats (statistics).
 943  */
 944 void
 945 pstats_fork(struct pstats *src, struct pstats *dst)
 946 {
 947
 948         bzero(&dst->pstat_startzero,
 949             __rangeof(struct pstats, pstat_startzero, pstat_endzero));
 950         bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
 951             __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
 952 }
 953
 954 void
 955 pstats_free(struct pstats *ps)
 956 {
 957
 958         free(ps, M_SUBPROC);
 959 }
 960
 961 /*
 962  * Locate a zombie process by number
 963  */
 964 struct proc *
 965 zpfind(pid_t pid)
 966 {
 967         struct proc *p;
 968
 969         sx_slock(&allproc_lock);
 970         LIST_FOREACH(p, &zombproc, p_list)
 971                 if (p->p_pid == pid) {
 972                         PROC_LOCK(p);
 973                         break;
 974                 }
 975         sx_sunlock(&allproc_lock);
 976         return (p);
 977 }
 978
 979 #define KERN_PROC_ZOMBMASK      0x3
 980 #define KERN_PROC_NOTHREADS     0x4
 981
 982 #ifdef COMPAT_FREEBSD32
 983
 984 /*
 985  * This function is typically used to copy out the kernel address, so
 986  * it can be replaced by assignment of zero.
 987  */
 988 static inline uint32_t
 989 ptr32_trim(void *ptr)
 990 {
 991         uintptr_t uptr;
 992
 993         uptr = (uintptr_t)ptr;
 994         return ((uptr > UINT_MAX) ? 0 : uptr);
 995 }
 996
 997 #define PTRTRIM_CP(src,dst,fld) \
 998         do { (dst).fld = ptr32_trim((src).fld); } while (0)
 999
1000 static void
1001 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1002 {
1003         int i;
1004
1005         bzero(ki32, sizeof(struct kinfo_proc32));
1006         ki32->ki_structsize = sizeof(struct kinfo_proc32);
1007         CP(*ki, *ki32, ki_layout);
1008         PTRTRIM_CP(*ki, *ki32, ki_args);
1009         PTRTRIM_CP(*ki, *ki32, ki_paddr);
1010         PTRTRIM_CP(*ki, *ki32, ki_addr);
1011         PTRTRIM_CP(*ki, *ki32, ki_tracep);
1012         PTRTRIM_CP(*ki, *ki32, ki_textvp);
1013         PTRTRIM_CP(*ki, *ki32, ki_fd);
1014         PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1015         PTRTRIM_CP(*ki, *ki32, ki_wchan);
1016         CP(*ki, *ki32, ki_pid);
1017         CP(*ki, *ki32, ki_ppid);
1018         CP(*ki, *ki32, ki_pgid);
1019         CP(*ki, *ki32, ki_tpgid);
1020         CP(*ki, *ki32, ki_sid);
1021         CP(*ki, *ki32, ki_tsid);
1022         CP(*ki, *ki32, ki_jobc);
1023         CP(*ki, *ki32, ki_tdev);
1024         CP(*ki, *ki32, ki_siglist);
1025         CP(*ki, *ki32, ki_sigmask);
1026         CP(*ki, *ki32, ki_sigignore);
1027         CP(*ki, *ki32, ki_sigcatch);
1028         CP(*ki, *ki32, ki_uid);
1029         CP(*ki, *ki32, ki_ruid);
1030         CP(*ki, *ki32, ki_svuid);
1031         CP(*ki, *ki32, ki_rgid);
1032         CP(*ki, *ki32, ki_svgid);
1033         CP(*ki, *ki32, ki_ngroups);
1034         for (i = 0; i < KI_NGROUPS; i++)
1035                 CP(*ki, *ki32, ki_groups[i]);
1036         CP(*ki, *ki32, ki_size);
1037         CP(*ki, *ki32, ki_rssize);
1038         CP(*ki, *ki32, ki_swrss);
1039         CP(*ki, *ki32, ki_tsize);
1040         CP(*ki, *ki32, ki_dsize);
1041         CP(*ki, *ki32, ki_ssize);
1042         CP(*ki, *ki32, ki_xstat);
1043         CP(*ki, *ki32, ki_acflag);
1044         CP(*ki, *ki32, ki_pctcpu);
1045         CP(*ki, *ki32, ki_estcpu);
1046         CP(*ki, *ki32, ki_slptime);
1047         CP(*ki, *ki32, ki_swtime);
1048         CP(*ki, *ki32, ki_runtime);
1049         TV_CP(*ki, *ki32, ki_start);
1050         TV_CP(*ki, *ki32, ki_childtime);
1051         CP(*ki, *ki32, ki_flag);
1052         CP(*ki, *ki32, ki_kiflag);
1053         CP(*ki, *ki32, ki_traceflag);
1054         CP(*ki, *ki32, ki_stat);
1055         CP(*ki, *ki32, ki_nice);
1056         CP(*ki, *ki32, ki_lock);
1057         CP(*ki, *ki32, ki_rqindex);
1058         CP(*ki, *ki32, ki_oncpu);
1059         CP(*ki, *ki32, ki_lastcpu);
1060         bcopy(ki->ki_ocomm, ki32->ki_ocomm, OCOMMLEN + 1);
1061         bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1062         bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1063         bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1064         bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1065         bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1066         CP(*ki, *ki32, ki_cr_flags);
1067         CP(*ki, *ki32, ki_jid);
1068         CP(*ki, *ki32, ki_numthreads);
1069         CP(*ki, *ki32, ki_tid);
1070         CP(*ki, *ki32, ki_pri);
1071         freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1072         freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1073         PTRTRIM_CP(*ki, *ki32, ki_pcb);
1074         PTRTRIM_CP(*ki, *ki32, ki_kstack);
1075         PTRTRIM_CP(*ki, *ki32, ki_udata);
1076         CP(*ki, *ki32, ki_sflag);
1077         CP(*ki, *ki32, ki_tdflags);
1078 }
1079
1080 static int
1081 sysctl_out_proc_copyout(struct kinfo_proc *ki, struct sysctl_req *req)
1082 {
1083         struct kinfo_proc32 ki32;
1084         int error;
1085
1086         if (req->flags & SCTL_MASK32) {
1087                 freebsd32_kinfo_proc_out(ki, &ki32);
1088                 error = SYSCTL_OUT(req, &ki32, sizeof(struct kinfo_proc32));
1089         } else
1090                 error = SYSCTL_OUT(req, ki, sizeof(struct kinfo_proc));
1091         return (error);
1092 }
1093 #else
1094 static int
1095 sysctl_out_proc_copyout(struct kinfo_proc *ki, struct sysctl_req *req)
1096 {
1097
1098         return (SYSCTL_OUT(req, ki, sizeof(struct kinfo_proc)));
1099 }
1100 #endif
1101
1102 /*
1103  * Must be called with the process locked and will return with it unlocked.
1104  */
1105 static int
1106 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
1107 {
1108         struct thread *td;
1109         struct kinfo_proc kinfo_proc;
1110         int error = 0;
1111         struct proc *np;
1112         pid_t pid = p->p_pid;
1113
1114         PROC_LOCK_ASSERT(p, MA_OWNED);
1115         MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1116
1117         fill_kinfo_proc(p, &kinfo_proc);
1118         if (flags & KERN_PROC_NOTHREADS)
1119                 error = sysctl_out_proc_copyout(&kinfo_proc, req);
1120         else {
1121                 FOREACH_THREAD_IN_PROC(p, td) {
1122                         fill_kinfo_thread(td, &kinfo_proc, 1);
1123                         error = sysctl_out_proc_copyout(&kinfo_proc, req);
1124                         if (error)
1125                                 break;
1126                 }
1127         }
1128         PROC_UNLOCK(p);
1129         if (error)
1130                 return (error);
1131         if (flags & KERN_PROC_ZOMBMASK)
1132                 np = zpfind(pid);
1133         else {
1134                 if (pid == 0)
1135                         return (0);
1136                 np = pfind(pid);
1137         }
1138         if (np == NULL)
1139                 return (ESRCH);
1140         if (np != p) {
1141                 PROC_UNLOCK(np);
1142                 return (ESRCH);
1143         }
1144         PROC_UNLOCK(np);
1145         return (0);
1146 }
1147
1148 static int
1149 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1150 {
1151         int *name = (int*) arg1;
1152         u_int namelen = arg2;
1153         struct proc *p;
1154         int flags, doingzomb, oid_number;
1155         int error = 0;
1156
1157         oid_number = oidp->oid_number;
1158         if (oid_number != KERN_PROC_ALL &&
1159             (oid_number & KERN_PROC_INC_THREAD) == 0)
1160                 flags = KERN_PROC_NOTHREADS;
1161         else {
1162                 flags = 0;
1163                 oid_number &= ~KERN_PROC_INC_THREAD;
1164         }
1165         if (oid_number == KERN_PROC_PID) {
1166                 if (namelen != 1)
1167                         return (EINVAL);
1168                 error = sysctl_wire_old_buffer(req, 0);
1169                 if (error)
1170                         return (error);
1171                 p = pfind((pid_t)name[0]);
1172                 if (!p)
1173                         return (ESRCH);
1174                 if ((error = p_cansee(curthread, p))) {
1175                         PROC_UNLOCK(p);
1176                         return (error);
1177                 }
1178                 error = sysctl_out_proc(p, req, flags);
1179                 return (error);
1180         }
1181
1182         switch (oid_number) {
1183         case KERN_PROC_ALL:
1184                 if (namelen != 0)
1185                         return (EINVAL);
1186                 break;
1187         case KERN_PROC_PROC:
1188                 if (namelen != 0 && namelen != 1)
1189                         return (EINVAL);
1190                 break;
1191         default:
1192                 if (namelen != 1)
1193                         return (EINVAL);
1194                 break;
1195         }
1196
1197         if (!req->oldptr) {
1198                 /* overestimate by 5 procs */
1199                 error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1200                 if (error)
1201                         return (error);
1202         }
1203         error = sysctl_wire_old_buffer(req, 0);
1204         if (error != 0)
1205                 return (error);
1206         sx_slock(&allproc_lock);
1207         for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) {
1208                 if (!doingzomb)
1209                         p = LIST_FIRST(&allproc);
1210                 else
1211                         p = LIST_FIRST(&zombproc);
1212                 for (; p != 0; p = LIST_NEXT(p, p_list)) {
1213                         /*
1214                          * Skip embryonic processes.
1215                          */
1216                         PROC_SLOCK(p);
1217                         if (p->p_state == PRS_NEW) {
1218                                 PROC_SUNLOCK(p);
1219                                 continue;
1220                         }
1221                         PROC_SUNLOCK(p);
1222                         PROC_LOCK(p);
1223                         KASSERT(p->p_ucred != NULL,
1224                             ("process credential is NULL for non-NEW proc"));
1225                         /*
1226                          * Show a user only appropriate processes.
1227                          */
1228                         if (p_cansee(curthread, p)) {
1229                                 PROC_UNLOCK(p);
1230                                 continue;
1231                         }
1232                         /*
1233                          * TODO - make more efficient (see notes below).
1234                          * do by session.
1235                          */
1236                         switch (oid_number) {
1237
1238                         case KERN_PROC_GID:
1239                                 if (p->p_ucred->cr_gid != (gid_t)name[0]) {
1240                                         PROC_UNLOCK(p);
1241                                         continue;
1242                                 }
1243                                 break;
1244
1245                         case KERN_PROC_PGRP:
1246                                 /* could do this by traversing pgrp */
1247                                 if (p->p_pgrp == NULL ||
1248                                     p->p_pgrp->pg_id != (pid_t)name[0]) {
1249                                         PROC_UNLOCK(p);
1250                                         continue;
1251                                 }
1252                                 break;
1253
1254                         case KERN_PROC_RGID:
1255                                 if (p->p_ucred->cr_rgid != (gid_t)name[0]) {
1256                                         PROC_UNLOCK(p);
1257                                         continue;
1258                                 }
1259                                 break;
1260
1261                         case KERN_PROC_SESSION:
1262                                 if (p->p_session == NULL ||
1263                                     p->p_session->s_sid != (pid_t)name[0]) {
1264                                         PROC_UNLOCK(p);
1265                                         continue;
1266                                 }
1267                                 break;
1268
1269                         case KERN_PROC_TTY:
1270                                 if ((p->p_flag & P_CONTROLT) == 0 ||
1271                                     p->p_session == NULL) {
1272                                         PROC_UNLOCK(p);
1273                                         continue;
1274                                 }
1275                                 /* XXX proctree_lock */
1276                                 SESS_LOCK(p->p_session);
1277                                 if (p->p_session->s_ttyp == NULL ||
1278                                     tty_udev(p->p_session->s_ttyp) !=
1279                                     (dev_t)name[0]) {
1280                                         SESS_UNLOCK(p->p_session);
1281                                         PROC_UNLOCK(p);
1282                                         continue;
1283                                 }
1284                                 SESS_UNLOCK(p->p_session);
1285                                 break;
1286
1287                         case KERN_PROC_UID:
1288                                 if (p->p_ucred->cr_uid != (uid_t)name[0]) {
1289                                         PROC_UNLOCK(p);
1290                                         continue;
1291                                 }
1292                                 break;
1293
1294                         case KERN_PROC_RUID:
1295                                 if (p->p_ucred->cr_ruid != (uid_t)name[0]) {
1296                                         PROC_UNLOCK(p);
1297                                         continue;
1298                                 }
1299                                 break;
1300
1301                         case KERN_PROC_PROC:
1302                                 break;
1303
1304                         default:
1305                                 break;
1306
1307                         }
1308
1309                         error = sysctl_out_proc(p, req, flags | doingzomb);
1310                         if (error) {
1311                                 sx_sunlock(&allproc_lock);
1312                                 return (error);
1313                         }
1314                 }
1315         }
1316         sx_sunlock(&allproc_lock);
1317         return (0);
1318 }
1319
1320 struct pargs *
1321 pargs_alloc(int len)
1322 {
1323         struct pargs *pa;
1324
1325         pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1326                 M_WAITOK);
1327         refcount_init(&pa->ar_ref, 1);
1328         pa->ar_length = len;
1329         return (pa);
1330 }
1331
1332 static void
1333 pargs_free(struct pargs *pa)
1334 {
1335
1336         free(pa, M_PARGS);
1337 }
1338
1339 void
1340 pargs_hold(struct pargs *pa)
1341 {
1342
1343         if (pa == NULL)
1344                 return;
1345         refcount_acquire(&pa->ar_ref);
1346 }
1347
1348 void
1349 pargs_drop(struct pargs *pa)
1350 {
1351
1352         if (pa == NULL)
1353                 return;
1354         if (refcount_release(&pa->ar_ref))
1355                 pargs_free(pa);
1356 }
1357
1358 /*
1359  * This sysctl allows a process to retrieve the argument list or process
1360  * title for another process without groping around in the address space
1361  * of the other process.  It also allow a process to set its own "process
1362  * title to a string of its own choice.
1363  */
1364 static int
1365 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
1366 {
1367         int *name = (int*) arg1;
1368         u_int namelen = arg2;
1369         struct pargs *newpa, *pa;
1370         struct proc *p;
1371         int error = 0;
1372
1373         if (namelen != 1)
1374                 return (EINVAL);
1375
1376         p = pfind((pid_t)name[0]);
1377         if (!p)
1378                 return (ESRCH);
1379
1380         if ((error = p_cansee(curthread, p)) != 0) {
1381                 PROC_UNLOCK(p);
1382                 return (error);
1383         }
1384
1385         if (req->newptr && curproc != p) {
1386                 PROC_UNLOCK(p);
1387                 return (EPERM);
1388         }
1389
1390         pa = p->p_args;
1391         pargs_hold(pa);
1392         PROC_UNLOCK(p);
1393         if (req->oldptr != NULL && pa != NULL)
1394                 error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
1395         pargs_drop(pa);
1396         if (error != 0 || req->newptr == NULL)
1397                 return (error);
1398
1399         if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit)
1400                 return (ENOMEM);
1401         newpa = pargs_alloc(req->newlen);
1402         error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
1403         if (error != 0) {
1404                 pargs_free(newpa);
1405                 return (error);
1406         }
1407         PROC_LOCK(p);
1408         pa = p->p_args;
1409         p->p_args = newpa;
1410         PROC_UNLOCK(p);
1411         pargs_drop(pa);
1412         return (0);
1413 }
1414
1415 /*
1416  * This sysctl allows a process to retrieve the path of the executable for
1417  * itself or another process.
1418  */
1419 static int
1420 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
1421 {
1422         pid_t *pidp = (pid_t *)arg1;
1423         unsigned int arglen = arg2;
1424         struct proc *p;
1425         struct vnode *vp;
1426         char *retbuf, *freebuf;
1427         int error, vfslocked;
1428
1429         if (arglen != 1)
1430                 return (EINVAL);
1431         if (*pidp == -1) {      /* -1 means this process */
1432                 p = req->td->td_proc;
1433         } else {
1434                 p = pfind(*pidp);
1435                 if (p == NULL)
1436                         return (ESRCH);
1437                 if ((error = p_cansee(curthread, p)) != 0) {
1438                         PROC_UNLOCK(p);
1439                         return (error);
1440                 }
1441         }
1442
1443         vp = p->p_textvp;
1444         if (vp == NULL) {
1445                 if (*pidp != -1)
1446                         PROC_UNLOCK(p);
1447                 return (0);
1448         }
1449         vref(vp);
1450         if (*pidp != -1)
1451                 PROC_UNLOCK(p);
1452         error = vn_fullpath(req->td, vp, &retbuf, &freebuf);
1453         vfslocked = VFS_LOCK_GIANT(vp->v_mount);
1454         vrele(vp);
1455         VFS_UNLOCK_GIANT(vfslocked);
1456         if (error)
1457                 return (error);
1458         error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
1459         free(freebuf, M_TEMP);
1460         return (error);
1461 }
1462
1463 static int
1464 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
1465 {
1466         struct proc *p;
1467         char *sv_name;
1468         int *name;
1469         int namelen;
1470         int error;
1471
1472         namelen = arg2;
1473         if (namelen != 1)
1474                 return (EINVAL);
1475
1476         name = (int *)arg1;
1477         if ((p = pfind((pid_t)name[0])) == NULL)
1478                 return (ESRCH);
1479         if ((error = p_cansee(curthread, p))) {
1480                 PROC_UNLOCK(p);
1481                 return (error);
1482         }
1483         sv_name = p->p_sysent->sv_name;
1484         PROC_UNLOCK(p);
1485         return (sysctl_handle_string(oidp, sv_name, 0, req));
1486 }
1487
1488 #ifdef KINFO_OVMENTRY_SIZE
1489 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
1490 #endif
1491
1492 #ifdef COMPAT_FREEBSD7
1493 static int
1494 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
1495 {
1496         vm_map_entry_t entry, tmp_entry;
1497         unsigned int last_timestamp;
1498         char *fullpath, *freepath;
1499         struct kinfo_ovmentry *kve;
1500         struct vattr va;
1501         struct ucred *cred;
1502         int error, *name;
1503         struct vnode *vp;
1504         struct proc *p;
1505         vm_map_t map;
1506         struct vmspace *vm;
1507
1508         name = (int *)arg1;
1509         if ((p = pfind((pid_t)name[0])) == NULL)
1510                 return (ESRCH);
1511         if (p->p_flag & P_WEXIT) {
1512                 PROC_UNLOCK(p);
1513                 return (ESRCH);
1514         }
1515         if ((error = p_candebug(curthread, p))) {
1516                 PROC_UNLOCK(p);
1517                 return (error);
1518         }
1519         _PHOLD(p);
1520         PROC_UNLOCK(p);
1521         vm = vmspace_acquire_ref(p);
1522         if (vm == NULL) {
1523                 PRELE(p);
1524                 return (ESRCH);
1525         }
1526         kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
1527
1528         map = &p->p_vmspace->vm_map;    /* XXXRW: More locking required? */
1529         vm_map_lock_read(map);
1530         for (entry = map->header.next; entry != &map->header;
1531             entry = entry->next) {
1532                 vm_object_t obj, tobj, lobj;
1533                 vm_offset_t addr;
1534                 int vfslocked;
1535
1536                 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
1537                         continue;
1538
1539                 bzero(kve, sizeof(*kve));
1540                 kve->kve_structsize = sizeof(*kve);
1541
1542                 kve->kve_private_resident = 0;
1543                 obj = entry->object.vm_object;
1544                 if (obj != NULL) {
1545                         VM_OBJECT_LOCK(obj);
1546                         if (obj->shadow_count == 1)
1547                                 kve->kve_private_resident =
1548                                     obj->resident_page_count;
1549                 }
1550                 kve->kve_resident = 0;
1551                 addr = entry->start;
1552                 while (addr < entry->end) {
1553                         if (pmap_extract(map->pmap, addr))
1554                                 kve->kve_resident++;
1555                         addr += PAGE_SIZE;
1556                 }
1557
1558                 for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
1559                         if (tobj != obj)
1560                                 VM_OBJECT_LOCK(tobj);
1561                         if (lobj != obj)
1562                                 VM_OBJECT_UNLOCK(lobj);
1563                         lobj = tobj;
1564                 }
1565
1566                 kve->kve_start = (void*)entry->start;
1567                 kve->kve_end = (void*)entry->end;
1568                 kve->kve_offset = (off_t)entry->offset;
1569
1570                 if (entry->protection & VM_PROT_READ)
1571                         kve->kve_protection |= KVME_PROT_READ;
1572                 if (entry->protection & VM_PROT_WRITE)
1573                         kve->kve_protection |= KVME_PROT_WRITE;
1574                 if (entry->protection & VM_PROT_EXECUTE)
1575                         kve->kve_protection |= KVME_PROT_EXEC;
1576
1577                 if (entry->eflags & MAP_ENTRY_COW)
1578                         kve->kve_flags |= KVME_FLAG_COW;
1579                 if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
1580                         kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
1581                 if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
1582                         kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
1583
1584                 last_timestamp = map->timestamp;
1585                 vm_map_unlock_read(map);
1586
1587                 kve->kve_fileid = 0;
1588                 kve->kve_fsid = 0;
1589                 freepath = NULL;
1590                 fullpath = "";
1591                 if (lobj) {
1592                         vp = NULL;
1593                         switch (lobj->type) {
1594                         case OBJT_DEFAULT:
1595                                 kve->kve_type = KVME_TYPE_DEFAULT;
1596                                 break;
1597                         case OBJT_VNODE:
1598                                 kve->kve_type = KVME_TYPE_VNODE;
1599                                 vp = lobj->handle;
1600                                 vref(vp);
1601                                 break;
1602                         case OBJT_SWAP:
1603                                 kve->kve_type = KVME_TYPE_SWAP;
1604                                 break;
1605                         case OBJT_DEVICE:
1606                                 kve->kve_type = KVME_TYPE_DEVICE;
1607                                 break;
1608                         case OBJT_PHYS:
1609                                 kve->kve_type = KVME_TYPE_PHYS;
1610                                 break;
1611                         case OBJT_DEAD:
1612                                 kve->kve_type = KVME_TYPE_DEAD;
1613                                 break;
1614                         case OBJT_SG:
1615                                 kve->kve_type = KVME_TYPE_SG;
1616                                 break;
1617                         default:
1618                                 kve->kve_type = KVME_TYPE_UNKNOWN;
1619                                 break;
1620                         }
1621                         if (lobj != obj)
1622                                 VM_OBJECT_UNLOCK(lobj);
1623
1624                         kve->kve_ref_count = obj->ref_count;
1625                         kve->kve_shadow_count = obj->shadow_count;
1626                         VM_OBJECT_UNLOCK(obj);
1627                         if (vp != NULL) {
1628                                 vn_fullpath(curthread, vp, &fullpath,
1629                                     &freepath);
1630                                 cred = curthread->td_ucred;
1631                                 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
1632                                 vn_lock(vp, LK_SHARED | LK_RETRY);
1633                                 if (VOP_GETATTR(vp, &va, cred) == 0) {
1634                                         kve->kve_fileid = va.va_fileid;
1635                                         kve->kve_fsid = va.va_fsid;
1636                                 }
1637                                 vput(vp);
1638                                 VFS_UNLOCK_GIANT(vfslocked);
1639                         }
1640                 } else {
1641                         kve->kve_type = KVME_TYPE_NONE;
1642                         kve->kve_ref_count = 0;
1643                         kve->kve_shadow_count = 0;
1644                 }
1645
1646                 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
1647                 if (freepath != NULL)
1648                         free(freepath, M_TEMP);
1649
1650                 error = SYSCTL_OUT(req, kve, sizeof(*kve));
1651                 vm_map_lock_read(map);
1652                 if (error)
1653                         break;
1654                 if (last_timestamp != map->timestamp) {
1655                         vm_map_lookup_entry(map, addr - 1, &tmp_entry);
1656                         entry = tmp_entry;
1657                 }
1658         }
1659         vm_map_unlock_read(map);
1660         vmspace_free(vm);
1661         PRELE(p);
1662         free(kve, M_TEMP);
1663         return (error);
1664 }
1665 #endif  /* COMPAT_FREEBSD7 */
1666
1667 #ifdef KINFO_VMENTRY_SIZE
1668 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
1669 #endif
1670
1671 static int
1672 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
1673 {
1674         vm_map_entry_t entry, tmp_entry;
1675         unsigned int last_timestamp;
1676         char *fullpath, *freepath;
1677         struct kinfo_vmentry *kve;
1678         struct vattr va;
1679         struct ucred *cred;
1680         int error, *name;
1681         struct vnode *vp;
1682         struct proc *p;
1683         struct vmspace *vm;
1684         vm_map_t map;
1685
1686         name = (int *)arg1;
1687         if ((p = pfind((pid_t)name[0])) == NULL)
1688                 return (ESRCH);
1689         if (p->p_flag & P_WEXIT) {
1690                 PROC_UNLOCK(p);
1691                 return (ESRCH);
1692         }
1693         if ((error = p_candebug(curthread, p))) {
1694                 PROC_UNLOCK(p);
1695                 return (error);
1696         }
1697         _PHOLD(p);
1698         PROC_UNLOCK(p);
1699         vm = vmspace_acquire_ref(p);
1700         if (vm == NULL) {
1701                 PRELE(p);
1702                 return (ESRCH);
1703         }
1704         kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
1705
1706         map = &vm->vm_map;      /* XXXRW: More locking required? */
1707         vm_map_lock_read(map);
1708         for (entry = map->header.next; entry != &map->header;
1709             entry = entry->next) {
1710                 vm_object_t obj, tobj, lobj;
1711                 vm_offset_t addr;
1712                 int vfslocked;
1713
1714                 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
1715                         continue;
1716
1717                 bzero(kve, sizeof(*kve));
1718
1719                 kve->kve_private_resident = 0;
1720                 obj = entry->object.vm_object;
1721                 if (obj != NULL) {
1722                         VM_OBJECT_LOCK(obj);
1723                         if (obj->shadow_count == 1)
1724                                 kve->kve_private_resident =
1725                                     obj->resident_page_count;
1726                 }
1727                 kve->kve_resident = 0;
1728                 addr = entry->start;
1729                 while (addr < entry->end) {
1730                         if (pmap_extract(map->pmap, addr))
1731                                 kve->kve_resident++;
1732                         addr += PAGE_SIZE;
1733                 }
1734
1735                 for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
1736                         if (tobj != obj)
1737                                 VM_OBJECT_LOCK(tobj);
1738                         if (lobj != obj)
1739                                 VM_OBJECT_UNLOCK(lobj);
1740                         lobj = tobj;
1741                 }
1742
1743                 kve->kve_start = entry->start;
1744                 kve->kve_end = entry->end;
1745                 kve->kve_offset = entry->offset;
1746
1747                 if (entry->protection & VM_PROT_READ)
1748                         kve->kve_protection |= KVME_PROT_READ;
1749                 if (entry->protection & VM_PROT_WRITE)
1750                         kve->kve_protection |= KVME_PROT_WRITE;
1751                 if (entry->protection & VM_PROT_EXECUTE)
1752                         kve->kve_protection |= KVME_PROT_EXEC;
1753
1754                 if (entry->eflags & MAP_ENTRY_COW)
1755                         kve->kve_flags |= KVME_FLAG_COW;
1756                 if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
1757                         kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
1758                 if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
1759                         kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
1760
1761                 last_timestamp = map->timestamp;
1762                 vm_map_unlock_read(map);
1763
1764                 kve->kve_fileid = 0;
1765                 kve->kve_fsid = 0;
1766                 freepath = NULL;
1767                 fullpath = "";
1768                 if (lobj) {
1769                         vp = NULL;
1770                         switch (lobj->type) {
1771                         case OBJT_DEFAULT:
1772                                 kve->kve_type = KVME_TYPE_DEFAULT;
1773                                 break;
1774                         case OBJT_VNODE:
1775                                 kve->kve_type = KVME_TYPE_VNODE;
1776                                 vp = lobj->handle;
1777                                 vref(vp);
1778                                 break;
1779                         case OBJT_SWAP:
1780                                 kve->kve_type = KVME_TYPE_SWAP;
1781                                 break;
1782                         case OBJT_DEVICE:
1783                                 kve->kve_type = KVME_TYPE_DEVICE;
1784                                 break;
1785                         case OBJT_PHYS:
1786                                 kve->kve_type = KVME_TYPE_PHYS;
1787                                 break;
1788                         case OBJT_DEAD:
1789                                 kve->kve_type = KVME_TYPE_DEAD;
1790                                 break;
1791                         case OBJT_SG:
1792                                 kve->kve_type = KVME_TYPE_SG;
1793                                 break;
1794                         default:
1795                                 kve->kve_type = KVME_TYPE_UNKNOWN;
1796                                 break;
1797                         }
1798                         if (lobj != obj)
1799                                 VM_OBJECT_UNLOCK(lobj);
1800
1801                         kve->kve_ref_count = obj->ref_count;
1802                         kve->kve_shadow_count = obj->shadow_count;
1803                         VM_OBJECT_UNLOCK(obj);
1804                         if (vp != NULL) {
1805                                 vn_fullpath(curthread, vp, &fullpath,
1806                                     &freepath);
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_fileid = va.va_fileid;
1812                                         kve->kve_fsid = va.va_fsid;
1813                                 }
1814                                 vput(vp);
1815                                 VFS_UNLOCK_GIANT(vfslocked);
1816                         }
1817                 } else {
1818                         kve->kve_type = KVME_TYPE_NONE;
1819                         kve->kve_ref_count = 0;
1820                         kve->kve_shadow_count = 0;
1821                 }
1822
1823                 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
1824                 if (freepath != NULL)
1825                         free(freepath, M_TEMP);
1826
1827                 /* Pack record size down */
1828                 kve->kve_structsize = offsetof(struct kinfo_vmentry, kve_path) +
1829                     strlen(kve->kve_path) + 1;
1830                 kve->kve_structsize = roundup(kve->kve_structsize,
1831                     sizeof(uint64_t));
1832                 error = SYSCTL_OUT(req, kve, kve->kve_structsize);
1833                 vm_map_lock_read(map);
1834                 if (error)
1835                         break;
1836                 if (last_timestamp != map->timestamp) {
1837                         vm_map_lookup_entry(map, addr - 1, &tmp_entry);
1838                         entry = tmp_entry;
1839                 }
1840         }
1841         vm_map_unlock_read(map);
1842         vmspace_free(vm);
1843         PRELE(p);
1844         free(kve, M_TEMP);
1845         return (error);
1846 }
1847
1848 #if defined(STACK) || defined(DDB)
1849 static int
1850 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
1851 {
1852         struct kinfo_kstack *kkstp;
1853         int error, i, *name, numthreads;
1854         lwpid_t *lwpidarray;
1855         struct thread *td;
1856         struct stack *st;
1857         struct sbuf sb;
1858         struct proc *p;
1859
1860         name = (int *)arg1;
1861         if ((p = pfind((pid_t)name[0])) == NULL)
1862                 return (ESRCH);
1863         /* XXXRW: Not clear ESRCH is the right error during proc execve(). */
1864         if (p->p_flag & P_WEXIT || p->p_flag & P_INEXEC) {
1865                 PROC_UNLOCK(p);
1866                 return (ESRCH);
1867         }
1868         if ((error = p_candebug(curthread, p))) {
1869                 PROC_UNLOCK(p);
1870                 return (error);
1871         }
1872         _PHOLD(p);
1873         PROC_UNLOCK(p);
1874
1875         kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
1876         st = stack_create();
1877
1878         lwpidarray = NULL;
1879         numthreads = 0;
1880         PROC_LOCK(p);
1881 repeat:
1882         if (numthreads < p->p_numthreads) {
1883                 if (lwpidarray != NULL) {
1884                         free(lwpidarray, M_TEMP);
1885                         lwpidarray = NULL;
1886                 }
1887                 numthreads = p->p_numthreads;
1888                 PROC_UNLOCK(p);
1889                 lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
1890                     M_WAITOK | M_ZERO);
1891                 PROC_LOCK(p);
1892                 goto repeat;
1893         }
1894         i = 0;
1895
1896         /*
1897          * XXXRW: During the below loop, execve(2) and countless other sorts
1898          * of changes could have taken place.  Should we check to see if the
1899          * vmspace has been replaced, or the like, in order to prevent
1900          * giving a snapshot that spans, say, execve(2), with some threads
1901          * before and some after?  Among other things, the credentials could
1902          * have changed, in which case the right to extract debug info might
1903          * no longer be assured.
1904          */
1905         FOREACH_THREAD_IN_PROC(p, td) {
1906                 KASSERT(i < numthreads,
1907                     ("sysctl_kern_proc_kstack: numthreads"));
1908                 lwpidarray[i] = td->td_tid;
1909                 i++;
1910         }
1911         numthreads = i;
1912         for (i = 0; i < numthreads; i++) {
1913                 td = thread_find(p, lwpidarray[i]);
1914                 if (td == NULL) {
1915                         continue;
1916                 }
1917                 bzero(kkstp, sizeof(*kkstp));
1918                 (void)sbuf_new(&sb, kkstp->kkst_trace,
1919                     sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
1920                 thread_lock(td);
1921                 kkstp->kkst_tid = td->td_tid;
1922                 if (TD_IS_SWAPPED(td))
1923                         kkstp->kkst_state = KKST_STATE_SWAPPED;
1924                 else if (TD_IS_RUNNING(td))
1925                         kkstp->kkst_state = KKST_STATE_RUNNING;
1926                 else {
1927                         kkstp->kkst_state = KKST_STATE_STACKOK;
1928                         stack_save_td(st, td);
1929                 }
1930                 thread_unlock(td);
1931                 PROC_UNLOCK(p);
1932                 stack_sbuf_print(&sb, st);
1933                 sbuf_finish(&sb);
1934                 sbuf_delete(&sb);
1935                 error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
1936                 PROC_LOCK(p);
1937                 if (error)
1938                         break;
1939         }
1940         _PRELE(p);
1941         PROC_UNLOCK(p);
1942         if (lwpidarray != NULL)
1943                 free(lwpidarray, M_TEMP);
1944         stack_destroy(st);
1945         free(kkstp, M_TEMP);
1946         return (error);
1947 }
1948 #endif
1949
1950 /*
1951  * This sysctl allows a process to retrieve the full list of groups from
1952  * itself or another process.
1953  */
1954 static int
1955 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
1956 {
1957         pid_t *pidp = (pid_t *)arg1;
1958         unsigned int arglen = arg2;
1959         struct proc *p;
1960         struct ucred *cred;
1961         int error;
1962
1963         if (arglen != 1)
1964                 return (EINVAL);
1965         if (*pidp == -1) {      /* -1 means this process */
1966                 p = req->td->td_proc;
1967         } else {
1968                 p = pfind(*pidp);
1969                 if (p == NULL)
1970                         return (ESRCH);
1971                 if ((error = p_cansee(curthread, p)) != 0) {
1972                         PROC_UNLOCK(p);
1973                         return (error);
1974                 }
1975         }
1976
1977         cred = crhold(p->p_ucred);
1978         if (*pidp != -1)
1979                 PROC_UNLOCK(p);
1980
1981         error = SYSCTL_OUT(req, cred->cr_groups,
1982             cred->cr_ngroups * sizeof(gid_t));
1983         crfree(cred);
1984         return (error);
1985 }
1986
1987 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD,  0, "Process table");
1988
1989 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
1990         CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
1991         "Return entire process table");
1992
1993 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
1994         sysctl_kern_proc, "Process table");
1995
1996 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
1997         sysctl_kern_proc, "Process table");
1998
1999 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2000         sysctl_kern_proc, "Process table");
2001
2002 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
2003         CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2004
2005 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
2006         sysctl_kern_proc, "Process table");
2007
2008 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2009         sysctl_kern_proc, "Process table");
2010
2011 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2012         sysctl_kern_proc, "Process table");
2013
2014 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2015         sysctl_kern_proc, "Process table");
2016
2017 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
2018         sysctl_kern_proc, "Return process table, no threads");
2019
2020 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
2021         CTLFLAG_RW | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
2022         sysctl_kern_proc_args, "Process argument list");
2023
2024 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
2025         CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
2026
2027 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
2028         CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
2029         "Process syscall vector name (ABI type)");
2030
2031 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
2032         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2033
2034 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
2035         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2036
2037 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
2038         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2039
2040 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
2041         sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2042
2043 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
2044         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2045
2046 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
2047         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2048
2049 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
2050         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2051
2052 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
2053         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2054
2055 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
2056         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
2057         "Return process table, no threads");
2058
2059 #ifdef COMPAT_FREEBSD7
2060 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
2061         CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
2062 #endif
2063
2064 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
2065         CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
2066
2067 #if defined(STACK) || defined(DDB)
2068 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
2069         CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
2070 #endif
2071
2072 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
2073         CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");