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/mman.h>
  50 #include <sys/mount.h>
  51 #include <sys/mutex.h>
  52 #include <sys/proc.h>
  53 #include <sys/refcount.h>
  54 #include <sys/sbuf.h>
  55 #include <sys/sysent.h>
  56 #include <sys/sched.h>
  57 #include <sys/smp.h>
  58 #include <sys/stack.h>
  59 #include <sys/sysctl.h>
  60 #include <sys/filedesc.h>
  61 #include <sys/tty.h>
  62 #include <sys/signalvar.h>
  63 #include <sys/sdt.h>
  64 #include <sys/sx.h>
  65 #include <sys/user.h>
  66 #include <sys/jail.h>
  67 #include <sys/vnode.h>
  68 #include <sys/eventhandler.h>
  69
  70 #ifdef DDB
  71 #include <ddb/ddb.h>
  72 #endif
  73
  74 #include <vm/vm.h>
  75 #include <vm/vm_extern.h>
  76 #include <vm/pmap.h>
  77 #include <vm/vm_map.h>
  78 #include <vm/vm_object.h>
  79 #include <vm/vm_page.h>
  80 #include <vm/uma.h>
  81
  82 #ifdef COMPAT_FREEBSD32
  83 #include <compat/freebsd32/freebsd32.h>
  84 #include <compat/freebsd32/freebsd32_util.h>
  85 #endif
  86
  87 SDT_PROVIDER_DEFINE(proc);
  88 SDT_PROBE_DEFINE(proc, kernel, ctor, entry, entry);
  89 SDT_PROBE_ARGTYPE(proc, kernel, ctor, entry, 0, "struct proc *");
  90 SDT_PROBE_ARGTYPE(proc, kernel, ctor, entry, 1, "int");
  91 SDT_PROBE_ARGTYPE(proc, kernel, ctor, entry, 2, "void *");
  92 SDT_PROBE_ARGTYPE(proc, kernel, ctor, entry, 3, "int");
  93 SDT_PROBE_DEFINE(proc, kernel, ctor, return, return);
  94 SDT_PROBE_ARGTYPE(proc, kernel, ctor, return, 0, "struct proc *");
  95 SDT_PROBE_ARGTYPE(proc, kernel, ctor, return, 1, "int");
  96 SDT_PROBE_ARGTYPE(proc, kernel, ctor, return, 2, "void *");
  97 SDT_PROBE_ARGTYPE(proc, kernel, ctor, return, 3, "int");
  98 SDT_PROBE_DEFINE(proc, kernel, dtor, entry, entry);
  99 SDT_PROBE_ARGTYPE(proc, kernel, dtor, entry, 0, "struct proc *");
 100 SDT_PROBE_ARGTYPE(proc, kernel, dtor, entry, 1, "int");
 101 SDT_PROBE_ARGTYPE(proc, kernel, dtor, entry, 2, "void *");
 102 SDT_PROBE_ARGTYPE(proc, kernel, dtor, entry, 3, "struct thread *");
 103 SDT_PROBE_DEFINE(proc, kernel, dtor, return, return);
 104 SDT_PROBE_ARGTYPE(proc, kernel, dtor, return, 0, "struct proc *");
 105 SDT_PROBE_ARGTYPE(proc, kernel, dtor, return, 1, "int");
 106 SDT_PROBE_ARGTYPE(proc, kernel, dtor, return, 2, "void *");
 107 SDT_PROBE_DEFINE(proc, kernel, init, entry, entry);
 108 SDT_PROBE_ARGTYPE(proc, kernel, init, entry, 0, "struct proc *");
 109 SDT_PROBE_ARGTYPE(proc, kernel, init, entry, 1, "int");
 110 SDT_PROBE_ARGTYPE(proc, kernel, init, entry, 2, "int");
 111 SDT_PROBE_DEFINE(proc, kernel, init, return, return);
 112 SDT_PROBE_ARGTYPE(proc, kernel, init, return, 0, "struct proc *");
 113 SDT_PROBE_ARGTYPE(proc, kernel, init, return, 1, "int");
 114 SDT_PROBE_ARGTYPE(proc, kernel, init, return, 2, "int");
 115
 116 MALLOC_DEFINE(M_PGRP, "pgrp", "process group header");
 117 MALLOC_DEFINE(M_SESSION, "session", "session header");
 118 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
 119 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
 120
 121 static void doenterpgrp(struct proc *, struct pgrp *);
 122 static void orphanpg(struct pgrp *pg);
 123 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp);
 124 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp);
 125 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp,
 126     int preferthread);
 127 static void pgadjustjobc(struct pgrp *pgrp, int entering);
 128 static void pgdelete(struct pgrp *);
 129 static int proc_ctor(void *mem, int size, void *arg, int flags);
 130 static void proc_dtor(void *mem, int size, void *arg);
 131 static int proc_init(void *mem, int size, int flags);
 132 static void proc_fini(void *mem, int size);
 133 static void pargs_free(struct pargs *pa);
 134
 135 /*
 136  * Other process lists
 137  */
 138 struct pidhashhead *pidhashtbl;
 139 u_long pidhash;
 140 struct pgrphashhead *pgrphashtbl;
 141 u_long pgrphash;
 142 struct proclist allproc;
 143 struct proclist zombproc;
 144 struct sx allproc_lock;
 145 struct sx proctree_lock;
 146 struct mtx ppeers_lock;
 147 uma_zone_t proc_zone;
 148
 149 int kstack_pages = KSTACK_PAGES;
 150 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0,
 151     "Kernel stack size in pages");
 152
 153 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
 154 #ifdef COMPAT_FREEBSD32
 155 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
 156 #endif
 157
 158 /*
 159  * Initialize global process hashing structures.
 160  */
 161 void
 162 procinit()
 163 {
 164
 165         sx_init(&allproc_lock, "allproc");
 166         sx_init(&proctree_lock, "proctree");
 167         mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
 168         LIST_INIT(&allproc);
 169         LIST_INIT(&zombproc);
 170         pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
 171         pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
 172         proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
 173             proc_ctor, proc_dtor, proc_init, proc_fini,
 174             UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
 175         uihashinit();
 176 }
 177
 178 /*
 179  * Prepare a proc for use.
 180  */
 181 static int
 182 proc_ctor(void *mem, int size, void *arg, int flags)
 183 {
 184         struct proc *p;
 185
 186         p = (struct proc *)mem;
 187         SDT_PROBE(proc, kernel, ctor , entry, p, size, arg, flags, 0);
 188         EVENTHANDLER_INVOKE(process_ctor, p);
 189         SDT_PROBE(proc, kernel, ctor , return, p, size, arg, flags, 0);
 190         return (0);
 191 }
 192
 193 /*
 194  * Reclaim a proc after use.
 195  */
 196 static void
 197 proc_dtor(void *mem, int size, void *arg)
 198 {
 199         struct proc *p;
 200         struct thread *td;
 201
 202         /* INVARIANTS checks go here */
 203         p = (struct proc *)mem;
 204         td = FIRST_THREAD_IN_PROC(p);
 205         SDT_PROBE(proc, kernel, dtor, entry, p, size, arg, td, 0);
 206         if (td != NULL) {
 207 #ifdef INVARIANTS
 208                 KASSERT((p->p_numthreads == 1),
 209                     ("bad number of threads in exiting process"));
 210                 KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
 211 #endif
 212                 /* Free all OSD associated to this thread. */
 213                 osd_thread_exit(td);
 214         }
 215         EVENTHANDLER_INVOKE(process_dtor, p);
 216         if (p->p_ksi != NULL)
 217                 KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
 218         SDT_PROBE(proc, kernel, dtor, return, p, size, arg, 0, 0);
 219 }
 220
 221 /*
 222  * Initialize type-stable parts of a proc (when newly created).
 223  */
 224 static int
 225 proc_init(void *mem, int size, int flags)
 226 {
 227         struct proc *p;
 228
 229         p = (struct proc *)mem;
 230         SDT_PROBE(proc, kernel, init, entry, p, size, flags, 0, 0);
 231         p->p_sched = (struct p_sched *)&p[1];
 232         bzero(&p->p_mtx, sizeof(struct mtx));
 233         mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK);
 234         mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_RECURSE);
 235         cv_init(&p->p_pwait, "ppwait");
 236         cv_init(&p->p_dbgwait, "dbgwait");
 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                         PROC_LOCK(p);
 297                         if (p->p_state == PRS_NEW) {
 298                                 PROC_UNLOCK(p);
 299                                 p = NULL;
 300                         }
 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_ttydp = NULL;
 371                 sess->s_ttyp = NULL;
 372                 bcopy(p->p_session->s_login, sess->s_login,
 373                             sizeof(sess->s_login));
 374                 pgrp->pg_session = sess;
 375                 KASSERT(p == curproc,
 376                     ("enterpgrp: mksession and p != curproc"));
 377         } else {
 378                 pgrp->pg_session = p->p_session;
 379                 sess_hold(pgrp->pg_session);
 380                 PGRP_LOCK(pgrp);
 381         }
 382         pgrp->pg_id = pgid;
 383         LIST_INIT(&pgrp->pg_members);
 384
 385         /*
 386          * As we have an exclusive lock of proctree_lock,
 387          * this should not deadlock.
 388          */
 389         LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
 390         pgrp->pg_jobc = 0;
 391         SLIST_INIT(&pgrp->pg_sigiolst);
 392         PGRP_UNLOCK(pgrp);
 393
 394         doenterpgrp(p, pgrp);
 395
 396         return (0);
 397 }
 398
 399 /*
 400  * Move p to an existing process group
 401  */
 402 int
 403 enterthispgrp(p, pgrp)
 404         register struct proc *p;
 405         struct pgrp *pgrp;
 406 {
 407
 408         sx_assert(&proctree_lock, SX_XLOCKED);
 409         PROC_LOCK_ASSERT(p, MA_NOTOWNED);
 410         PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
 411         PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
 412         SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
 413         KASSERT(pgrp->pg_session == p->p_session,
 414                 ("%s: pgrp's session %p, p->p_session %p.\n",
 415                 __func__,
 416                 pgrp->pg_session,
 417                 p->p_session));
 418         KASSERT(pgrp != p->p_pgrp,
 419                 ("%s: p belongs to pgrp.", __func__));
 420
 421         doenterpgrp(p, pgrp);
 422
 423         return (0);
 424 }
 425
 426 /*
 427  * Move p to a process group
 428  */
 429 static void
 430 doenterpgrp(p, pgrp)
 431         struct proc *p;
 432         struct pgrp *pgrp;
 433 {
 434         struct pgrp *savepgrp;
 435
 436         sx_assert(&proctree_lock, SX_XLOCKED);
 437         PROC_LOCK_ASSERT(p, MA_NOTOWNED);
 438         PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
 439         PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
 440         SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
 441
 442         savepgrp = p->p_pgrp;
 443
 444         /*
 445          * Adjust eligibility of affected pgrps to participate in job control.
 446          * Increment eligibility counts before decrementing, otherwise we
 447          * could reach 0 spuriously during the first call.
 448          */
 449         fixjobc(p, pgrp, 1);
 450         fixjobc(p, p->p_pgrp, 0);
 451
 452         PGRP_LOCK(pgrp);
 453         PGRP_LOCK(savepgrp);
 454         PROC_LOCK(p);
 455         LIST_REMOVE(p, p_pglist);
 456         p->p_pgrp = pgrp;
 457         PROC_UNLOCK(p);
 458         LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
 459         PGRP_UNLOCK(savepgrp);
 460         PGRP_UNLOCK(pgrp);
 461         if (LIST_EMPTY(&savepgrp->pg_members))
 462                 pgdelete(savepgrp);
 463 }
 464
 465 /*
 466  * remove process from process group
 467  */
 468 int
 469 leavepgrp(p)
 470         register struct proc *p;
 471 {
 472         struct pgrp *savepgrp;
 473
 474         sx_assert(&proctree_lock, SX_XLOCKED);
 475         savepgrp = p->p_pgrp;
 476         PGRP_LOCK(savepgrp);
 477         PROC_LOCK(p);
 478         LIST_REMOVE(p, p_pglist);
 479         p->p_pgrp = NULL;
 480         PROC_UNLOCK(p);
 481         PGRP_UNLOCK(savepgrp);
 482         if (LIST_EMPTY(&savepgrp->pg_members))
 483                 pgdelete(savepgrp);
 484         return (0);
 485 }
 486
 487 /*
 488  * delete a process group
 489  */
 490 static void
 491 pgdelete(pgrp)
 492         register struct pgrp *pgrp;
 493 {
 494         struct session *savesess;
 495         struct tty *tp;
 496
 497         sx_assert(&proctree_lock, SX_XLOCKED);
 498         PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
 499         SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
 500
 501         /*
 502          * Reset any sigio structures pointing to us as a result of
 503          * F_SETOWN with our pgid.
 504          */
 505         funsetownlst(&pgrp->pg_sigiolst);
 506
 507         PGRP_LOCK(pgrp);
 508         tp = pgrp->pg_session->s_ttyp;
 509         LIST_REMOVE(pgrp, pg_hash);
 510         savesess = pgrp->pg_session;
 511         PGRP_UNLOCK(pgrp);
 512
 513         /* Remove the reference to the pgrp before deallocating it. */
 514         if (tp != NULL) {
 515                 tty_lock(tp);
 516                 tty_rel_pgrp(tp, pgrp);
 517         }
 518
 519         mtx_destroy(&pgrp->pg_mtx);
 520         free(pgrp, M_PGRP);
 521         sess_release(savesess);
 522 }
 523
 524 static void
 525 pgadjustjobc(pgrp, entering)
 526         struct pgrp *pgrp;
 527         int entering;
 528 {
 529
 530         PGRP_LOCK(pgrp);
 531         if (entering)
 532                 pgrp->pg_jobc++;
 533         else {
 534                 --pgrp->pg_jobc;
 535                 if (pgrp->pg_jobc == 0)
 536                         orphanpg(pgrp);
 537         }
 538         PGRP_UNLOCK(pgrp);
 539 }
 540
 541 /*
 542  * Adjust pgrp jobc counters when specified process changes process group.
 543  * We count the number of processes in each process group that "qualify"
 544  * the group for terminal job control (those with a parent in a different
 545  * process group of the same session).  If that count reaches zero, the
 546  * process group becomes orphaned.  Check both the specified process'
 547  * process group and that of its children.
 548  * entering == 0 => p is leaving specified group.
 549  * entering == 1 => p is entering specified group.
 550  */
 551 void
 552 fixjobc(p, pgrp, entering)
 553         register struct proc *p;
 554         register struct pgrp *pgrp;
 555         int entering;
 556 {
 557         register struct pgrp *hispgrp;
 558         register struct session *mysession;
 559
 560         sx_assert(&proctree_lock, SX_LOCKED);
 561         PROC_LOCK_ASSERT(p, MA_NOTOWNED);
 562         PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
 563         SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
 564
 565         /*
 566          * Check p's parent to see whether p qualifies its own process
 567          * group; if so, adjust count for p's process group.
 568          */
 569         mysession = pgrp->pg_session;
 570         if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
 571             hispgrp->pg_session == mysession)
 572                 pgadjustjobc(pgrp, entering);
 573
 574         /*
 575          * Check this process' children to see whether they qualify
 576          * their process groups; if so, adjust counts for children's
 577          * process groups.
 578          */
 579         LIST_FOREACH(p, &p->p_children, p_sibling) {
 580                 hispgrp = p->p_pgrp;
 581                 if (hispgrp == pgrp ||
 582                     hispgrp->pg_session != mysession)
 583                         continue;
 584                 PROC_LOCK(p);
 585                 if (p->p_state == PRS_ZOMBIE) {
 586                         PROC_UNLOCK(p);
 587                         continue;
 588                 }
 589                 PROC_UNLOCK(p);
 590                 pgadjustjobc(hispgrp, entering);
 591         }
 592 }
 593
 594 /*
 595  * A process group has become orphaned;
 596  * if there are any stopped processes in the group,
 597  * hang-up all process in that group.
 598  */
 599 static void
 600 orphanpg(pg)
 601         struct pgrp *pg;
 602 {
 603         register struct proc *p;
 604
 605         PGRP_LOCK_ASSERT(pg, MA_OWNED);
 606
 607         LIST_FOREACH(p, &pg->pg_members, p_pglist) {
 608                 PROC_LOCK(p);
 609                 if (P_SHOULDSTOP(p)) {
 610                         PROC_UNLOCK(p);
 611                         LIST_FOREACH(p, &pg->pg_members, p_pglist) {
 612                                 PROC_LOCK(p);
 613                                 kern_psignal(p, SIGHUP);
 614                                 kern_psignal(p, SIGCONT);
 615                                 PROC_UNLOCK(p);
 616                         }
 617                         return;
 618                 }
 619                 PROC_UNLOCK(p);
 620         }
 621 }
 622
 623 void
 624 sess_hold(struct session *s)
 625 {
 626
 627         refcount_acquire(&s->s_count);
 628 }
 629
 630 void
 631 sess_release(struct session *s)
 632 {
 633
 634         if (refcount_release(&s->s_count)) {
 635                 if (s->s_ttyp != NULL) {
 636                         tty_lock(s->s_ttyp);
 637                         tty_rel_sess(s->s_ttyp, s);
 638                 }
 639                 mtx_destroy(&s->s_mtx);
 640                 free(s, M_SESSION);
 641         }
 642 }
 643
 644 #include "opt_ddb.h"
 645 #ifdef DDB
 646 #include <ddb/ddb.h>
 647
 648 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
 649 {
 650         register struct pgrp *pgrp;
 651         register struct proc *p;
 652         register int i;
 653
 654         for (i = 0; i <= pgrphash; i++) {
 655                 if (!LIST_EMPTY(&pgrphashtbl[i])) {
 656                         printf("\tindx %d\n", i);
 657                         LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
 658                                 printf(
 659                         "\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
 660                                     (void *)pgrp, (long)pgrp->pg_id,
 661                                     (void *)pgrp->pg_session,
 662                                     pgrp->pg_session->s_count,
 663                                     (void *)LIST_FIRST(&pgrp->pg_members));
 664                                 LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
 665                                         printf("\t\tpid %ld addr %p pgrp %p\n",
 666                                             (long)p->p_pid, (void *)p,
 667                                             (void *)p->p_pgrp);
 668                                 }
 669                         }
 670                 }
 671         }
 672 }
 673 #endif /* DDB */
 674
 675 /*
 676  * Calculate the kinfo_proc members which contain process-wide
 677  * informations.
 678  * Must be called with the target process locked.
 679  */
 680 static void
 681 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
 682 {
 683         struct thread *td;
 684
 685         PROC_LOCK_ASSERT(p, MA_OWNED);
 686
 687         kp->ki_estcpu = 0;
 688         kp->ki_pctcpu = 0;
 689         FOREACH_THREAD_IN_PROC(p, td) {
 690                 thread_lock(td);
 691                 kp->ki_pctcpu += sched_pctcpu(td);
 692                 kp->ki_estcpu += td->td_estcpu;
 693                 thread_unlock(td);
 694         }
 695 }
 696
 697 /*
 698  * Clear kinfo_proc and fill in any information that is common
 699  * to all threads in the process.
 700  * Must be called with the target process locked.
 701  */
 702 static void
 703 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
 704 {
 705         struct thread *td0;
 706         struct tty *tp;
 707         struct session *sp;
 708         struct ucred *cred;
 709         struct sigacts *ps;
 710
 711         PROC_LOCK_ASSERT(p, MA_OWNED);
 712         bzero(kp, sizeof(*kp));
 713
 714         kp->ki_structsize = sizeof(*kp);
 715         kp->ki_paddr = p;
 716         kp->ki_addr =/* p->p_addr; */0; /* XXX */
 717         kp->ki_args = p->p_args;
 718         kp->ki_textvp = p->p_textvp;
 719 #ifdef KTRACE
 720         kp->ki_tracep = p->p_tracevp;
 721         kp->ki_traceflag = p->p_traceflag;
 722 #endif
 723         kp->ki_fd = p->p_fd;
 724         kp->ki_vmspace = p->p_vmspace;
 725         kp->ki_flag = p->p_flag;
 726         cred = p->p_ucred;
 727         if (cred) {
 728                 kp->ki_uid = cred->cr_uid;
 729                 kp->ki_ruid = cred->cr_ruid;
 730                 kp->ki_svuid = cred->cr_svuid;
 731                 kp->ki_cr_flags = 0;
 732                 if (cred->cr_flags & CRED_FLAG_CAPMODE)
 733                         kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
 734                 /* XXX bde doesn't like KI_NGROUPS */
 735                 if (cred->cr_ngroups > KI_NGROUPS) {
 736                         kp->ki_ngroups = KI_NGROUPS;
 737                         kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
 738                 } else
 739                         kp->ki_ngroups = cred->cr_ngroups;
 740                 bcopy(cred->cr_groups, kp->ki_groups,
 741                     kp->ki_ngroups * sizeof(gid_t));
 742                 kp->ki_rgid = cred->cr_rgid;
 743                 kp->ki_svgid = cred->cr_svgid;
 744                 /* If jailed(cred), emulate the old P_JAILED flag. */
 745                 if (jailed(cred)) {
 746                         kp->ki_flag |= P_JAILED;
 747                         /* If inside the jail, use 0 as a jail ID. */
 748                         if (cred->cr_prison != curthread->td_ucred->cr_prison)
 749                                 kp->ki_jid = cred->cr_prison->pr_id;
 750                 }
 751                 strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
 752                     sizeof(kp->ki_loginclass));
 753         }
 754         ps = p->p_sigacts;
 755         if (ps) {
 756                 mtx_lock(&ps->ps_mtx);
 757                 kp->ki_sigignore = ps->ps_sigignore;
 758                 kp->ki_sigcatch = ps->ps_sigcatch;
 759                 mtx_unlock(&ps->ps_mtx);
 760         }
 761         if (p->p_state != PRS_NEW &&
 762             p->p_state != PRS_ZOMBIE &&
 763             p->p_vmspace != NULL) {
 764                 struct vmspace *vm = p->p_vmspace;
 765
 766                 kp->ki_size = vm->vm_map.size;
 767                 kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
 768                 FOREACH_THREAD_IN_PROC(p, td0) {
 769                         if (!TD_IS_SWAPPED(td0))
 770                                 kp->ki_rssize += td0->td_kstack_pages;
 771                 }
 772                 kp->ki_swrss = vm->vm_swrss;
 773                 kp->ki_tsize = vm->vm_tsize;
 774                 kp->ki_dsize = vm->vm_dsize;
 775                 kp->ki_ssize = vm->vm_ssize;
 776         } else if (p->p_state == PRS_ZOMBIE)
 777                 kp->ki_stat = SZOMB;
 778         if (kp->ki_flag & P_INMEM)
 779                 kp->ki_sflag = PS_INMEM;
 780         else
 781                 kp->ki_sflag = 0;
 782         /* Calculate legacy swtime as seconds since 'swtick'. */
 783         kp->ki_swtime = (ticks - p->p_swtick) / hz;
 784         kp->ki_pid = p->p_pid;
 785         kp->ki_nice = p->p_nice;
 786         kp->ki_start = p->p_stats->p_start;
 787         timevaladd(&kp->ki_start, &boottime);
 788         PROC_SLOCK(p);
 789         rufetch(p, &kp->ki_rusage);
 790         kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
 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         /* Some callers want child times in a single value. */
 795         kp->ki_childtime = kp->ki_childstime;
 796         timevaladd(&kp->ki_childtime, &kp->ki_childutime);
 797
 798         tp = NULL;
 799         if (p->p_pgrp) {
 800                 kp->ki_pgid = p->p_pgrp->pg_id;
 801                 kp->ki_jobc = p->p_pgrp->pg_jobc;
 802                 sp = p->p_pgrp->pg_session;
 803
 804                 if (sp != NULL) {
 805                         kp->ki_sid = sp->s_sid;
 806                         SESS_LOCK(sp);
 807                         strlcpy(kp->ki_login, sp->s_login,
 808                             sizeof(kp->ki_login));
 809                         if (sp->s_ttyvp)
 810                                 kp->ki_kiflag |= KI_CTTY;
 811                         if (SESS_LEADER(p))
 812                                 kp->ki_kiflag |= KI_SLEADER;
 813                         /* XXX proctree_lock */
 814                         tp = sp->s_ttyp;
 815                         SESS_UNLOCK(sp);
 816                 }
 817         }
 818         if ((p->p_flag & P_CONTROLT) && tp != NULL) {
 819                 kp->ki_tdev = tty_udev(tp);
 820                 kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
 821                 if (tp->t_session)
 822                         kp->ki_tsid = tp->t_session->s_sid;
 823         } else
 824                 kp->ki_tdev = NODEV;
 825         if (p->p_comm[0] != '\0')
 826                 strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
 827         if (p->p_sysent && p->p_sysent->sv_name != NULL &&
 828             p->p_sysent->sv_name[0] != '\0')
 829                 strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
 830         kp->ki_siglist = p->p_siglist;
 831         kp->ki_xstat = p->p_xstat;
 832         kp->ki_acflag = p->p_acflag;
 833         kp->ki_lock = p->p_lock;
 834         if (p->p_pptr)
 835                 kp->ki_ppid = p->p_pptr->p_pid;
 836 }
 837
 838 /*
 839  * Fill in information that is thread specific.  Must be called with
 840  * target process locked.  If 'preferthread' is set, overwrite certain
 841  * process-related fields that are maintained for both threads and
 842  * processes.
 843  */
 844 static void
 845 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
 846 {
 847         struct proc *p;
 848
 849         p = td->td_proc;
 850         kp->ki_tdaddr = td;
 851         PROC_LOCK_ASSERT(p, MA_OWNED);
 852
 853         if (preferthread)
 854                 PROC_SLOCK(p);
 855         thread_lock(td);
 856         if (td->td_wmesg != NULL)
 857                 strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
 858         else
 859                 bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
 860         strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname));
 861         if (TD_ON_LOCK(td)) {
 862                 kp->ki_kiflag |= KI_LOCKBLOCK;
 863                 strlcpy(kp->ki_lockname, td->td_lockname,
 864                     sizeof(kp->ki_lockname));
 865         } else {
 866                 kp->ki_kiflag &= ~KI_LOCKBLOCK;
 867                 bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
 868         }
 869
 870         if (p->p_state == PRS_NORMAL) { /* approximate. */
 871                 if (TD_ON_RUNQ(td) ||
 872                     TD_CAN_RUN(td) ||
 873                     TD_IS_RUNNING(td)) {
 874                         kp->ki_stat = SRUN;
 875                 } else if (P_SHOULDSTOP(p)) {
 876                         kp->ki_stat = SSTOP;
 877                 } else if (TD_IS_SLEEPING(td)) {
 878                         kp->ki_stat = SSLEEP;
 879                 } else if (TD_ON_LOCK(td)) {
 880                         kp->ki_stat = SLOCK;
 881                 } else {
 882                         kp->ki_stat = SWAIT;
 883                 }
 884         } else if (p->p_state == PRS_ZOMBIE) {
 885                 kp->ki_stat = SZOMB;
 886         } else {
 887                 kp->ki_stat = SIDL;
 888         }
 889
 890         /* Things in the thread */
 891         kp->ki_wchan = td->td_wchan;
 892         kp->ki_pri.pri_level = td->td_priority;
 893         kp->ki_pri.pri_native = td->td_base_pri;
 894         kp->ki_lastcpu = td->td_lastcpu;
 895         kp->ki_oncpu = td->td_oncpu;
 896         kp->ki_tdflags = td->td_flags;
 897         kp->ki_tid = td->td_tid;
 898         kp->ki_numthreads = p->p_numthreads;
 899         kp->ki_pcb = td->td_pcb;
 900         kp->ki_kstack = (void *)td->td_kstack;
 901         kp->ki_slptime = (ticks - td->td_slptick) / hz;
 902         kp->ki_pri.pri_class = td->td_pri_class;
 903         kp->ki_pri.pri_user = td->td_user_pri;
 904
 905         if (preferthread) {
 906                 rufetchtd(td, &kp->ki_rusage);
 907                 kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
 908                 kp->ki_pctcpu = sched_pctcpu(td);
 909                 kp->ki_estcpu = td->td_estcpu;
 910         }
 911
 912         /* We can't get this anymore but ps etc never used it anyway. */
 913         kp->ki_rqindex = 0;
 914
 915         if (preferthread)
 916                 kp->ki_siglist = td->td_siglist;
 917         kp->ki_sigmask = td->td_sigmask;
 918         thread_unlock(td);
 919         if (preferthread)
 920                 PROC_SUNLOCK(p);
 921 }
 922
 923 /*
 924  * Fill in a kinfo_proc structure for the specified process.
 925  * Must be called with the target process locked.
 926  */
 927 void
 928 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
 929 {
 930
 931         MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
 932
 933         fill_kinfo_proc_only(p, kp);
 934         fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
 935         fill_kinfo_aggregate(p, kp);
 936 }
 937
 938 struct pstats *
 939 pstats_alloc(void)
 940 {
 941
 942         return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
 943 }
 944
 945 /*
 946  * Copy parts of p_stats; zero the rest of p_stats (statistics).
 947  */
 948 void
 949 pstats_fork(struct pstats *src, struct pstats *dst)
 950 {
 951
 952         bzero(&dst->pstat_startzero,
 953             __rangeof(struct pstats, pstat_startzero, pstat_endzero));
 954         bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
 955             __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
 956 }
 957
 958 void
 959 pstats_free(struct pstats *ps)
 960 {
 961
 962         free(ps, M_SUBPROC);
 963 }
 964
 965 /*
 966  * Locate a zombie process by number
 967  */
 968 struct proc *
 969 zpfind(pid_t pid)
 970 {
 971         struct proc *p;
 972
 973         sx_slock(&allproc_lock);
 974         LIST_FOREACH(p, &zombproc, p_list)
 975                 if (p->p_pid == pid) {
 976                         PROC_LOCK(p);
 977                         break;
 978                 }
 979         sx_sunlock(&allproc_lock);
 980         return (p);
 981 }
 982
 983 #define KERN_PROC_ZOMBMASK      0x3
 984 #define KERN_PROC_NOTHREADS     0x4
 985
 986 #ifdef COMPAT_FREEBSD32
 987
 988 /*
 989  * This function is typically used to copy out the kernel address, so
 990  * it can be replaced by assignment of zero.
 991  */
 992 static inline uint32_t
 993 ptr32_trim(void *ptr)
 994 {
 995         uintptr_t uptr;
 996
 997         uptr = (uintptr_t)ptr;
 998         return ((uptr > UINT_MAX) ? 0 : uptr);
 999 }
1000
1001 #define PTRTRIM_CP(src,dst,fld) \
1002         do { (dst).fld = ptr32_trim((src).fld); } while (0)
1003
1004 static void
1005 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1006 {
1007         int i;
1008
1009         bzero(ki32, sizeof(struct kinfo_proc32));
1010         ki32->ki_structsize = sizeof(struct kinfo_proc32);
1011         CP(*ki, *ki32, ki_layout);
1012         PTRTRIM_CP(*ki, *ki32, ki_args);
1013         PTRTRIM_CP(*ki, *ki32, ki_paddr);
1014         PTRTRIM_CP(*ki, *ki32, ki_addr);
1015         PTRTRIM_CP(*ki, *ki32, ki_tracep);
1016         PTRTRIM_CP(*ki, *ki32, ki_textvp);
1017         PTRTRIM_CP(*ki, *ki32, ki_fd);
1018         PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1019         PTRTRIM_CP(*ki, *ki32, ki_wchan);
1020         CP(*ki, *ki32, ki_pid);
1021         CP(*ki, *ki32, ki_ppid);
1022         CP(*ki, *ki32, ki_pgid);
1023         CP(*ki, *ki32, ki_tpgid);
1024         CP(*ki, *ki32, ki_sid);
1025         CP(*ki, *ki32, ki_tsid);
1026         CP(*ki, *ki32, ki_jobc);
1027         CP(*ki, *ki32, ki_tdev);
1028         CP(*ki, *ki32, ki_siglist);
1029         CP(*ki, *ki32, ki_sigmask);
1030         CP(*ki, *ki32, ki_sigignore);
1031         CP(*ki, *ki32, ki_sigcatch);
1032         CP(*ki, *ki32, ki_uid);
1033         CP(*ki, *ki32, ki_ruid);
1034         CP(*ki, *ki32, ki_svuid);
1035         CP(*ki, *ki32, ki_rgid);
1036         CP(*ki, *ki32, ki_svgid);
1037         CP(*ki, *ki32, ki_ngroups);
1038         for (i = 0; i < KI_NGROUPS; i++)
1039                 CP(*ki, *ki32, ki_groups[i]);
1040         CP(*ki, *ki32, ki_size);
1041         CP(*ki, *ki32, ki_rssize);
1042         CP(*ki, *ki32, ki_swrss);
1043         CP(*ki, *ki32, ki_tsize);
1044         CP(*ki, *ki32, ki_dsize);
1045         CP(*ki, *ki32, ki_ssize);
1046         CP(*ki, *ki32, ki_xstat);
1047         CP(*ki, *ki32, ki_acflag);
1048         CP(*ki, *ki32, ki_pctcpu);
1049         CP(*ki, *ki32, ki_estcpu);
1050         CP(*ki, *ki32, ki_slptime);
1051         CP(*ki, *ki32, ki_swtime);
1052         CP(*ki, *ki32, ki_runtime);
1053         TV_CP(*ki, *ki32, ki_start);
1054         TV_CP(*ki, *ki32, ki_childtime);
1055         CP(*ki, *ki32, ki_flag);
1056         CP(*ki, *ki32, ki_kiflag);
1057         CP(*ki, *ki32, ki_traceflag);
1058         CP(*ki, *ki32, ki_stat);
1059         CP(*ki, *ki32, ki_nice);
1060         CP(*ki, *ki32, ki_lock);
1061         CP(*ki, *ki32, ki_rqindex);
1062         CP(*ki, *ki32, ki_oncpu);
1063         CP(*ki, *ki32, ki_lastcpu);
1064         bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1065         bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1066         bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1067         bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1068         bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1069         bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1070         bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1071         CP(*ki, *ki32, ki_cr_flags);
1072         CP(*ki, *ki32, ki_jid);
1073         CP(*ki, *ki32, ki_numthreads);
1074         CP(*ki, *ki32, ki_tid);
1075         CP(*ki, *ki32, ki_pri);
1076         freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1077         freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1078         PTRTRIM_CP(*ki, *ki32, ki_pcb);
1079         PTRTRIM_CP(*ki, *ki32, ki_kstack);
1080         PTRTRIM_CP(*ki, *ki32, ki_udata);
1081         CP(*ki, *ki32, ki_sflag);
1082         CP(*ki, *ki32, ki_tdflags);
1083 }
1084
1085 static int
1086 sysctl_out_proc_copyout(struct kinfo_proc *ki, struct sysctl_req *req)
1087 {
1088         struct kinfo_proc32 ki32;
1089         int error;
1090
1091         if (req->flags & SCTL_MASK32) {
1092                 freebsd32_kinfo_proc_out(ki, &ki32);
1093                 error = SYSCTL_OUT(req, &ki32, sizeof(struct kinfo_proc32));
1094         } else
1095                 error = SYSCTL_OUT(req, ki, sizeof(struct kinfo_proc));
1096         return (error);
1097 }
1098 #else
1099 static int
1100 sysctl_out_proc_copyout(struct kinfo_proc *ki, struct sysctl_req *req)
1101 {
1102
1103         return (SYSCTL_OUT(req, ki, sizeof(struct kinfo_proc)));
1104 }
1105 #endif
1106
1107 /*
1108  * Must be called with the process locked and will return with it unlocked.
1109  */
1110 static int
1111 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
1112 {
1113         struct thread *td;
1114         struct kinfo_proc kinfo_proc;
1115         int error = 0;
1116         struct proc *np;
1117         pid_t pid = p->p_pid;
1118
1119         PROC_LOCK_ASSERT(p, MA_OWNED);
1120         MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1121
1122         fill_kinfo_proc(p, &kinfo_proc);
1123         if (flags & KERN_PROC_NOTHREADS)
1124                 error = sysctl_out_proc_copyout(&kinfo_proc, req);
1125         else {
1126                 FOREACH_THREAD_IN_PROC(p, td) {
1127                         fill_kinfo_thread(td, &kinfo_proc, 1);
1128                         error = sysctl_out_proc_copyout(&kinfo_proc, req);
1129                         if (error)
1130                                 break;
1131                 }
1132         }
1133         PROC_UNLOCK(p);
1134         if (error)
1135                 return (error);
1136         if (flags & KERN_PROC_ZOMBMASK)
1137                 np = zpfind(pid);
1138         else {
1139                 if (pid == 0)
1140                         return (0);
1141                 np = pfind(pid);
1142         }
1143         if (np == NULL)
1144                 return (ESRCH);
1145         if (np != p) {
1146                 PROC_UNLOCK(np);
1147                 return (ESRCH);
1148         }
1149         PROC_UNLOCK(np);
1150         return (0);
1151 }
1152
1153 static int
1154 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1155 {
1156         int *name = (int*) arg1;
1157         u_int namelen = arg2;
1158         struct proc *p;
1159         int flags, doingzomb, oid_number;
1160         int error = 0;
1161
1162         oid_number = oidp->oid_number;
1163         if (oid_number != KERN_PROC_ALL &&
1164             (oid_number & KERN_PROC_INC_THREAD) == 0)
1165                 flags = KERN_PROC_NOTHREADS;
1166         else {
1167                 flags = 0;
1168                 oid_number &= ~KERN_PROC_INC_THREAD;
1169         }
1170         if (oid_number == KERN_PROC_PID) {
1171                 if (namelen != 1)
1172                         return (EINVAL);
1173                 error = sysctl_wire_old_buffer(req, 0);
1174                 if (error)
1175                         return (error);
1176                 p = pfind((pid_t)name[0]);
1177                 if (!p)
1178                         return (ESRCH);
1179                 if ((error = p_cansee(curthread, p))) {
1180                         PROC_UNLOCK(p);
1181                         return (error);
1182                 }
1183                 error = sysctl_out_proc(p, req, flags);
1184                 return (error);
1185         }
1186
1187         switch (oid_number) {
1188         case KERN_PROC_ALL:
1189                 if (namelen != 0)
1190                         return (EINVAL);
1191                 break;
1192         case KERN_PROC_PROC:
1193                 if (namelen != 0 && namelen != 1)
1194                         return (EINVAL);
1195                 break;
1196         default:
1197                 if (namelen != 1)
1198                         return (EINVAL);
1199                 break;
1200         }
1201
1202         if (!req->oldptr) {
1203                 /* overestimate by 5 procs */
1204                 error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1205                 if (error)
1206                         return (error);
1207         }
1208         error = sysctl_wire_old_buffer(req, 0);
1209         if (error != 0)
1210                 return (error);
1211         sx_slock(&allproc_lock);
1212         for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) {
1213                 if (!doingzomb)
1214                         p = LIST_FIRST(&allproc);
1215                 else
1216                         p = LIST_FIRST(&zombproc);
1217                 for (; p != 0; p = LIST_NEXT(p, p_list)) {
1218                         /*
1219                          * Skip embryonic processes.
1220                          */
1221                         PROC_LOCK(p);
1222                         if (p->p_state == PRS_NEW) {
1223                                 PROC_UNLOCK(p);
1224                                 continue;
1225                         }
1226                         KASSERT(p->p_ucred != NULL,
1227                             ("process credential is NULL for non-NEW proc"));
1228                         /*
1229                          * Show a user only appropriate processes.
1230                          */
1231                         if (p_cansee(curthread, p)) {
1232                                 PROC_UNLOCK(p);
1233                                 continue;
1234                         }
1235                         /*
1236                          * TODO - make more efficient (see notes below).
1237                          * do by session.
1238                          */
1239                         switch (oid_number) {
1240
1241                         case KERN_PROC_GID:
1242                                 if (p->p_ucred->cr_gid != (gid_t)name[0]) {
1243                                         PROC_UNLOCK(p);
1244                                         continue;
1245                                 }
1246                                 break;
1247
1248                         case KERN_PROC_PGRP:
1249                                 /* could do this by traversing pgrp */
1250                                 if (p->p_pgrp == NULL ||
1251                                     p->p_pgrp->pg_id != (pid_t)name[0]) {
1252                                         PROC_UNLOCK(p);
1253                                         continue;
1254                                 }
1255                                 break;
1256
1257                         case KERN_PROC_RGID:
1258                                 if (p->p_ucred->cr_rgid != (gid_t)name[0]) {
1259                                         PROC_UNLOCK(p);
1260                                         continue;
1261                                 }
1262                                 break;
1263
1264                         case KERN_PROC_SESSION:
1265                                 if (p->p_session == NULL ||
1266                                     p->p_session->s_sid != (pid_t)name[0]) {
1267                                         PROC_UNLOCK(p);
1268                                         continue;
1269                                 }
1270                                 break;
1271
1272                         case KERN_PROC_TTY:
1273                                 if ((p->p_flag & P_CONTROLT) == 0 ||
1274                                     p->p_session == NULL) {
1275                                         PROC_UNLOCK(p);
1276                                         continue;
1277                                 }
1278                                 /* XXX proctree_lock */
1279                                 SESS_LOCK(p->p_session);
1280                                 if (p->p_session->s_ttyp == NULL ||
1281                                     tty_udev(p->p_session->s_ttyp) !=
1282                                     (dev_t)name[0]) {
1283                                         SESS_UNLOCK(p->p_session);
1284                                         PROC_UNLOCK(p);
1285                                         continue;
1286                                 }
1287                                 SESS_UNLOCK(p->p_session);
1288                                 break;
1289
1290                         case KERN_PROC_UID:
1291                                 if (p->p_ucred->cr_uid != (uid_t)name[0]) {
1292                                         PROC_UNLOCK(p);
1293                                         continue;
1294                                 }
1295                                 break;
1296
1297                         case KERN_PROC_RUID:
1298                                 if (p->p_ucred->cr_ruid != (uid_t)name[0]) {
1299                                         PROC_UNLOCK(p);
1300                                         continue;
1301                                 }
1302                                 break;
1303
1304                         case KERN_PROC_PROC:
1305                                 break;
1306
1307                         default:
1308                                 break;
1309
1310                         }
1311
1312                         error = sysctl_out_proc(p, req, flags | doingzomb);
1313                         if (error) {
1314                                 sx_sunlock(&allproc_lock);
1315                                 return (error);
1316                         }
1317                 }
1318         }
1319         sx_sunlock(&allproc_lock);
1320         return (0);
1321 }
1322
1323 struct pargs *
1324 pargs_alloc(int len)
1325 {
1326         struct pargs *pa;
1327
1328         pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1329                 M_WAITOK);
1330         refcount_init(&pa->ar_ref, 1);
1331         pa->ar_length = len;
1332         return (pa);
1333 }
1334
1335 static void
1336 pargs_free(struct pargs *pa)
1337 {
1338
1339         free(pa, M_PARGS);
1340 }
1341
1342 void
1343 pargs_hold(struct pargs *pa)
1344 {
1345
1346         if (pa == NULL)
1347                 return;
1348         refcount_acquire(&pa->ar_ref);
1349 }
1350
1351 void
1352 pargs_drop(struct pargs *pa)
1353 {
1354
1355         if (pa == NULL)
1356                 return;
1357         if (refcount_release(&pa->ar_ref))
1358                 pargs_free(pa);
1359 }
1360
1361 /*
1362  * This sysctl allows a process to retrieve the argument list or process
1363  * title for another process without groping around in the address space
1364  * of the other process.  It also allow a process to set its own "process
1365  * title to a string of its own choice.
1366  */
1367 static int
1368 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
1369 {
1370         int *name = (int*) arg1;
1371         u_int namelen = arg2;
1372         struct pargs *newpa, *pa;
1373         struct proc *p;
1374         int error = 0;
1375
1376         if (namelen != 1)
1377                 return (EINVAL);
1378
1379         p = pfind((pid_t)name[0]);
1380         if (!p)
1381                 return (ESRCH);
1382
1383         if ((error = p_cansee(curthread, p)) != 0) {
1384                 PROC_UNLOCK(p);
1385                 return (error);
1386         }
1387
1388         if (req->newptr && curproc != p) {
1389                 PROC_UNLOCK(p);
1390                 return (EPERM);
1391         }
1392
1393         pa = p->p_args;
1394         pargs_hold(pa);
1395         PROC_UNLOCK(p);
1396         if (pa != NULL)
1397                 error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
1398         pargs_drop(pa);
1399         if (error != 0 || req->newptr == NULL)
1400                 return (error);
1401
1402         if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit)
1403                 return (ENOMEM);
1404         newpa = pargs_alloc(req->newlen);
1405         error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
1406         if (error != 0) {
1407                 pargs_free(newpa);
1408                 return (error);
1409         }
1410         PROC_LOCK(p);
1411         pa = p->p_args;
1412         p->p_args = newpa;
1413         PROC_UNLOCK(p);
1414         pargs_drop(pa);
1415         return (0);
1416 }
1417
1418 /*
1419  * This sysctl allows a process to retrieve the path of the executable for
1420  * itself or another process.
1421  */
1422 static int
1423 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
1424 {
1425         pid_t *pidp = (pid_t *)arg1;
1426         unsigned int arglen = arg2;
1427         struct proc *p;
1428         struct vnode *vp;
1429         char *retbuf, *freebuf;
1430         int error, vfslocked;
1431
1432         if (arglen != 1)
1433                 return (EINVAL);
1434         if (*pidp == -1) {      /* -1 means this process */
1435                 p = req->td->td_proc;
1436         } else {
1437                 p = pfind(*pidp);
1438                 if (p == NULL)
1439                         return (ESRCH);
1440                 if ((error = p_cansee(curthread, p)) != 0) {
1441                         PROC_UNLOCK(p);
1442                         return (error);
1443                 }
1444         }
1445
1446         vp = p->p_textvp;
1447         if (vp == NULL) {
1448                 if (*pidp != -1)
1449                         PROC_UNLOCK(p);
1450                 return (0);
1451         }
1452         vref(vp);
1453         if (*pidp != -1)
1454                 PROC_UNLOCK(p);
1455         error = vn_fullpath(req->td, vp, &retbuf, &freebuf);
1456         vfslocked = VFS_LOCK_GIANT(vp->v_mount);
1457         vrele(vp);
1458         VFS_UNLOCK_GIANT(vfslocked);
1459         if (error)
1460                 return (error);
1461         error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
1462         free(freebuf, M_TEMP);
1463         return (error);
1464 }
1465
1466 static int
1467 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
1468 {
1469         struct proc *p;
1470         char *sv_name;
1471         int *name;
1472         int namelen;
1473         int error;
1474
1475         namelen = arg2;
1476         if (namelen != 1)
1477                 return (EINVAL);
1478
1479         name = (int *)arg1;
1480         if ((p = pfind((pid_t)name[0])) == NULL)
1481                 return (ESRCH);
1482         if ((error = p_cansee(curthread, p))) {
1483                 PROC_UNLOCK(p);
1484                 return (error);
1485         }
1486         sv_name = p->p_sysent->sv_name;
1487         PROC_UNLOCK(p);
1488         return (sysctl_handle_string(oidp, sv_name, 0, req));
1489 }
1490
1491 #ifdef KINFO_OVMENTRY_SIZE
1492 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
1493 #endif
1494
1495 #ifdef COMPAT_FREEBSD7
1496 static int
1497 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
1498 {
1499         vm_map_entry_t entry, tmp_entry;
1500         unsigned int last_timestamp;
1501         char *fullpath, *freepath;
1502         struct kinfo_ovmentry *kve;
1503         struct vattr va;
1504         struct ucred *cred;
1505         int error, *name;
1506         struct vnode *vp;
1507         struct proc *p;
1508         vm_map_t map;
1509         struct vmspace *vm;
1510
1511         name = (int *)arg1;
1512         if ((p = pfind((pid_t)name[0])) == NULL)
1513                 return (ESRCH);
1514         if (p->p_flag & P_WEXIT) {
1515                 PROC_UNLOCK(p);
1516                 return (ESRCH);
1517         }
1518         if ((error = p_candebug(curthread, p))) {
1519                 PROC_UNLOCK(p);
1520                 return (error);
1521         }
1522         _PHOLD(p);
1523         PROC_UNLOCK(p);
1524         vm = vmspace_acquire_ref(p);
1525         if (vm == NULL) {
1526                 PRELE(p);
1527                 return (ESRCH);
1528         }
1529         kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
1530
1531         map = &p->p_vmspace->vm_map;    /* XXXRW: More locking required? */
1532         vm_map_lock_read(map);
1533         for (entry = map->header.next; entry != &map->header;
1534             entry = entry->next) {
1535                 vm_object_t obj, tobj, lobj;
1536                 vm_offset_t addr;
1537                 int vfslocked;
1538
1539                 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
1540                         continue;
1541
1542                 bzero(kve, sizeof(*kve));
1543                 kve->kve_structsize = sizeof(*kve);
1544
1545                 kve->kve_private_resident = 0;
1546                 obj = entry->object.vm_object;
1547                 if (obj != NULL) {
1548                         VM_OBJECT_LOCK(obj);
1549                         if (obj->shadow_count == 1)
1550                                 kve->kve_private_resident =
1551                                     obj->resident_page_count;
1552                 }
1553                 kve->kve_resident = 0;
1554                 addr = entry->start;
1555                 while (addr < entry->end) {
1556                         if (pmap_extract(map->pmap, addr))
1557                                 kve->kve_resident++;
1558                         addr += PAGE_SIZE;
1559                 }
1560
1561                 for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
1562                         if (tobj != obj)
1563                                 VM_OBJECT_LOCK(tobj);
1564                         if (lobj != obj)
1565                                 VM_OBJECT_UNLOCK(lobj);
1566                         lobj = tobj;
1567                 }
1568
1569                 kve->kve_start = (void*)entry->start;
1570                 kve->kve_end = (void*)entry->end;
1571                 kve->kve_offset = (off_t)entry->offset;
1572
1573                 if (entry->protection & VM_PROT_READ)
1574                         kve->kve_protection |= KVME_PROT_READ;
1575                 if (entry->protection & VM_PROT_WRITE)
1576                         kve->kve_protection |= KVME_PROT_WRITE;
1577                 if (entry->protection & VM_PROT_EXECUTE)
1578                         kve->kve_protection |= KVME_PROT_EXEC;
1579
1580                 if (entry->eflags & MAP_ENTRY_COW)
1581                         kve->kve_flags |= KVME_FLAG_COW;
1582                 if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
1583                         kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
1584                 if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
1585                         kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
1586
1587                 last_timestamp = map->timestamp;
1588                 vm_map_unlock_read(map);
1589
1590                 kve->kve_fileid = 0;
1591                 kve->kve_fsid = 0;
1592                 freepath = NULL;
1593                 fullpath = "";
1594                 if (lobj) {
1595                         vp = NULL;
1596                         switch (lobj->type) {
1597                         case OBJT_DEFAULT:
1598                                 kve->kve_type = KVME_TYPE_DEFAULT;
1599                                 break;
1600                         case OBJT_VNODE:
1601                                 kve->kve_type = KVME_TYPE_VNODE;
1602                                 vp = lobj->handle;
1603                                 vref(vp);
1604                                 break;
1605                         case OBJT_SWAP:
1606                                 kve->kve_type = KVME_TYPE_SWAP;
1607                                 break;
1608                         case OBJT_DEVICE:
1609                                 kve->kve_type = KVME_TYPE_DEVICE;
1610                                 break;
1611                         case OBJT_PHYS:
1612                                 kve->kve_type = KVME_TYPE_PHYS;
1613                                 break;
1614                         case OBJT_DEAD:
1615                                 kve->kve_type = KVME_TYPE_DEAD;
1616                                 break;
1617                         case OBJT_SG:
1618                                 kve->kve_type = KVME_TYPE_SG;
1619                                 break;
1620                         default:
1621                                 kve->kve_type = KVME_TYPE_UNKNOWN;
1622                                 break;
1623                         }
1624                         if (lobj != obj)
1625                                 VM_OBJECT_UNLOCK(lobj);
1626
1627                         kve->kve_ref_count = obj->ref_count;
1628                         kve->kve_shadow_count = obj->shadow_count;
1629                         VM_OBJECT_UNLOCK(obj);
1630                         if (vp != NULL) {
1631                                 vn_fullpath(curthread, vp, &fullpath,
1632                                     &freepath);
1633                                 cred = curthread->td_ucred;
1634                                 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
1635                                 vn_lock(vp, LK_SHARED | LK_RETRY);
1636                                 if (VOP_GETATTR(vp, &va, cred) == 0) {
1637                                         kve->kve_fileid = va.va_fileid;
1638                                         kve->kve_fsid = va.va_fsid;
1639                                 }
1640                                 vput(vp);
1641                                 VFS_UNLOCK_GIANT(vfslocked);
1642                         }
1643                 } else {
1644                         kve->kve_type = KVME_TYPE_NONE;
1645                         kve->kve_ref_count = 0;
1646                         kve->kve_shadow_count = 0;
1647                 }
1648
1649                 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
1650                 if (freepath != NULL)
1651                         free(freepath, M_TEMP);
1652
1653                 error = SYSCTL_OUT(req, kve, sizeof(*kve));
1654                 vm_map_lock_read(map);
1655                 if (error)
1656                         break;
1657                 if (last_timestamp != map->timestamp) {
1658                         vm_map_lookup_entry(map, addr - 1, &tmp_entry);
1659                         entry = tmp_entry;
1660                 }
1661         }
1662         vm_map_unlock_read(map);
1663         vmspace_free(vm);
1664         PRELE(p);
1665         free(kve, M_TEMP);
1666         return (error);
1667 }
1668 #endif  /* COMPAT_FREEBSD7 */
1669
1670 #ifdef KINFO_VMENTRY_SIZE
1671 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
1672 #endif
1673
1674 static int
1675 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
1676 {
1677         vm_map_entry_t entry, tmp_entry;
1678         unsigned int last_timestamp;
1679         char *fullpath, *freepath;
1680         struct kinfo_vmentry *kve;
1681         struct vattr va;
1682         struct ucred *cred;
1683         int error, *name;
1684         struct vnode *vp;
1685         struct proc *p;
1686         struct vmspace *vm;
1687         vm_map_t map;
1688
1689         name = (int *)arg1;
1690         if ((p = pfind((pid_t)name[0])) == NULL)
1691                 return (ESRCH);
1692         if (p->p_flag & P_WEXIT) {
1693                 PROC_UNLOCK(p);
1694                 return (ESRCH);
1695         }
1696         if ((error = p_candebug(curthread, p))) {
1697                 PROC_UNLOCK(p);
1698                 return (error);
1699         }
1700         _PHOLD(p);
1701         PROC_UNLOCK(p);
1702         vm = vmspace_acquire_ref(p);
1703         if (vm == NULL) {
1704                 PRELE(p);
1705                 return (ESRCH);
1706         }
1707         kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
1708
1709         map = &vm->vm_map;      /* XXXRW: More locking required? */
1710         vm_map_lock_read(map);
1711         for (entry = map->header.next; entry != &map->header;
1712             entry = entry->next) {
1713                 vm_object_t obj, tobj, lobj;
1714                 vm_offset_t addr;
1715                 vm_paddr_t locked_pa;
1716                 int vfslocked, mincoreinfo;
1717
1718                 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
1719                         continue;
1720
1721                 bzero(kve, sizeof(*kve));
1722
1723                 kve->kve_private_resident = 0;
1724                 obj = entry->object.vm_object;
1725                 if (obj != NULL) {
1726                         VM_OBJECT_LOCK(obj);
1727                         if (obj->shadow_count == 1)
1728                                 kve->kve_private_resident =
1729                                     obj->resident_page_count;
1730                 }
1731                 kve->kve_resident = 0;
1732                 addr = entry->start;
1733                 while (addr < entry->end) {
1734                         locked_pa = 0;
1735                         mincoreinfo = pmap_mincore(map->pmap, addr, &locked_pa);
1736                         if (locked_pa != 0)
1737                                 vm_page_unlock(PHYS_TO_VM_PAGE(locked_pa));
1738                         if (mincoreinfo & MINCORE_INCORE)
1739                                 kve->kve_resident++;
1740                         if (mincoreinfo & MINCORE_SUPER)
1741                                 kve->kve_flags |= KVME_FLAG_SUPER;
1742                         addr += PAGE_SIZE;
1743                 }
1744
1745                 for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
1746                         if (tobj != obj)
1747                                 VM_OBJECT_LOCK(tobj);
1748                         if (lobj != obj)
1749                                 VM_OBJECT_UNLOCK(lobj);
1750                         lobj = tobj;
1751                 }
1752
1753                 kve->kve_start = entry->start;
1754                 kve->kve_end = entry->end;
1755                 kve->kve_offset = entry->offset;
1756
1757                 if (entry->protection & VM_PROT_READ)
1758                         kve->kve_protection |= KVME_PROT_READ;
1759                 if (entry->protection & VM_PROT_WRITE)
1760                         kve->kve_protection |= KVME_PROT_WRITE;
1761                 if (entry->protection & VM_PROT_EXECUTE)
1762                         kve->kve_protection |= KVME_PROT_EXEC;
1763
1764                 if (entry->eflags & MAP_ENTRY_COW)
1765                         kve->kve_flags |= KVME_FLAG_COW;
1766                 if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
1767                         kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
1768                 if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
1769                         kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
1770
1771                 last_timestamp = map->timestamp;
1772                 vm_map_unlock_read(map);
1773
1774                 freepath = NULL;
1775                 fullpath = "";
1776                 if (lobj) {
1777                         vp = NULL;
1778                         switch (lobj->type) {
1779                         case OBJT_DEFAULT:
1780                                 kve->kve_type = KVME_TYPE_DEFAULT;
1781                                 break;
1782                         case OBJT_VNODE:
1783                                 kve->kve_type = KVME_TYPE_VNODE;
1784                                 vp = lobj->handle;
1785                                 vref(vp);
1786                                 break;
1787                         case OBJT_SWAP:
1788                                 kve->kve_type = KVME_TYPE_SWAP;
1789                                 break;
1790                         case OBJT_DEVICE:
1791                                 kve->kve_type = KVME_TYPE_DEVICE;
1792                                 break;
1793                         case OBJT_PHYS:
1794                                 kve->kve_type = KVME_TYPE_PHYS;
1795                                 break;
1796                         case OBJT_DEAD:
1797                                 kve->kve_type = KVME_TYPE_DEAD;
1798                                 break;
1799                         case OBJT_SG:
1800                                 kve->kve_type = KVME_TYPE_SG;
1801                                 break;
1802                         default:
1803                                 kve->kve_type = KVME_TYPE_UNKNOWN;
1804                                 break;
1805                         }
1806                         if (lobj != obj)
1807                                 VM_OBJECT_UNLOCK(lobj);
1808
1809                         kve->kve_ref_count = obj->ref_count;
1810                         kve->kve_shadow_count = obj->shadow_count;
1811                         VM_OBJECT_UNLOCK(obj);
1812                         if (vp != NULL) {
1813                                 vn_fullpath(curthread, vp, &fullpath,
1814                                     &freepath);
1815                                 kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
1816                                 cred = curthread->td_ucred;
1817                                 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
1818                                 vn_lock(vp, LK_SHARED | LK_RETRY);
1819                                 if (VOP_GETATTR(vp, &va, cred) == 0) {
1820                                         kve->kve_vn_fileid = va.va_fileid;
1821                                         kve->kve_vn_fsid = va.va_fsid;
1822                                         kve->kve_vn_mode =
1823                                             MAKEIMODE(va.va_type, va.va_mode);
1824                                         kve->kve_vn_size = va.va_size;
1825                                         kve->kve_vn_rdev = va.va_rdev;
1826                                         kve->kve_status = KF_ATTR_VALID;
1827                                 }
1828                                 vput(vp);
1829                                 VFS_UNLOCK_GIANT(vfslocked);
1830                         }
1831                 } else {
1832                         kve->kve_type = KVME_TYPE_NONE;
1833                         kve->kve_ref_count = 0;
1834                         kve->kve_shadow_count = 0;
1835                 }
1836
1837                 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
1838                 if (freepath != NULL)
1839                         free(freepath, M_TEMP);
1840
1841                 /* Pack record size down */
1842                 kve->kve_structsize = offsetof(struct kinfo_vmentry, kve_path) +
1843                     strlen(kve->kve_path) + 1;
1844                 kve->kve_structsize = roundup(kve->kve_structsize,
1845                     sizeof(uint64_t));
1846                 error = SYSCTL_OUT(req, kve, kve->kve_structsize);
1847                 vm_map_lock_read(map);
1848                 if (error)
1849                         break;
1850                 if (last_timestamp != map->timestamp) {
1851                         vm_map_lookup_entry(map, addr - 1, &tmp_entry);
1852                         entry = tmp_entry;
1853                 }
1854         }
1855         vm_map_unlock_read(map);
1856         vmspace_free(vm);
1857         PRELE(p);
1858         free(kve, M_TEMP);
1859         return (error);
1860 }
1861
1862 #if defined(STACK) || defined(DDB)
1863 static int
1864 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
1865 {
1866         struct kinfo_kstack *kkstp;
1867         int error, i, *name, numthreads;
1868         lwpid_t *lwpidarray;
1869         struct thread *td;
1870         struct stack *st;
1871         struct sbuf sb;
1872         struct proc *p;
1873
1874         name = (int *)arg1;
1875         if ((p = pfind((pid_t)name[0])) == NULL)
1876                 return (ESRCH);
1877         /* XXXRW: Not clear ESRCH is the right error during proc execve(). */
1878         if (p->p_flag & P_WEXIT || p->p_flag & P_INEXEC) {
1879                 PROC_UNLOCK(p);
1880                 return (ESRCH);
1881         }
1882         if ((error = p_candebug(curthread, p))) {
1883                 PROC_UNLOCK(p);
1884                 return (error);
1885         }
1886         _PHOLD(p);
1887         PROC_UNLOCK(p);
1888
1889         kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
1890         st = stack_create();
1891
1892         lwpidarray = NULL;
1893         numthreads = 0;
1894         PROC_LOCK(p);
1895 repeat:
1896         if (numthreads < p->p_numthreads) {
1897                 if (lwpidarray != NULL) {
1898                         free(lwpidarray, M_TEMP);
1899                         lwpidarray = NULL;
1900                 }
1901                 numthreads = p->p_numthreads;
1902                 PROC_UNLOCK(p);
1903                 lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
1904                     M_WAITOK | M_ZERO);
1905                 PROC_LOCK(p);
1906                 goto repeat;
1907         }
1908         i = 0;
1909
1910         /*
1911          * XXXRW: During the below loop, execve(2) and countless other sorts
1912          * of changes could have taken place.  Should we check to see if the
1913          * vmspace has been replaced, or the like, in order to prevent
1914          * giving a snapshot that spans, say, execve(2), with some threads
1915          * before and some after?  Among other things, the credentials could
1916          * have changed, in which case the right to extract debug info might
1917          * no longer be assured.
1918          */
1919         FOREACH_THREAD_IN_PROC(p, td) {
1920                 KASSERT(i < numthreads,
1921                     ("sysctl_kern_proc_kstack: numthreads"));
1922                 lwpidarray[i] = td->td_tid;
1923                 i++;
1924         }
1925         numthreads = i;
1926         for (i = 0; i < numthreads; i++) {
1927                 td = thread_find(p, lwpidarray[i]);
1928                 if (td == NULL) {
1929                         continue;
1930                 }
1931                 bzero(kkstp, sizeof(*kkstp));
1932                 (void)sbuf_new(&sb, kkstp->kkst_trace,
1933                     sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
1934                 thread_lock(td);
1935                 kkstp->kkst_tid = td->td_tid;
1936                 if (TD_IS_SWAPPED(td))
1937                         kkstp->kkst_state = KKST_STATE_SWAPPED;
1938                 else if (TD_IS_RUNNING(td))
1939                         kkstp->kkst_state = KKST_STATE_RUNNING;
1940                 else {
1941                         kkstp->kkst_state = KKST_STATE_STACKOK;
1942                         stack_save_td(st, td);
1943                 }
1944                 thread_unlock(td);
1945                 PROC_UNLOCK(p);
1946                 stack_sbuf_print(&sb, st);
1947                 sbuf_finish(&sb);
1948                 sbuf_delete(&sb);
1949                 error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
1950                 PROC_LOCK(p);
1951                 if (error)
1952                         break;
1953         }
1954         _PRELE(p);
1955         PROC_UNLOCK(p);
1956         if (lwpidarray != NULL)
1957                 free(lwpidarray, M_TEMP);
1958         stack_destroy(st);
1959         free(kkstp, M_TEMP);
1960         return (error);
1961 }
1962 #endif
1963
1964 /*
1965  * This sysctl allows a process to retrieve the full list of groups from
1966  * itself or another process.
1967  */
1968 static int
1969 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
1970 {
1971         pid_t *pidp = (pid_t *)arg1;
1972         unsigned int arglen = arg2;
1973         struct proc *p;
1974         struct ucred *cred;
1975         int error;
1976
1977         if (arglen != 1)
1978                 return (EINVAL);
1979         if (*pidp == -1) {      /* -1 means this process */
1980                 p = req->td->td_proc;
1981         } else {
1982                 p = pfind(*pidp);
1983                 if (p == NULL)
1984                         return (ESRCH);
1985                 if ((error = p_cansee(curthread, p)) != 0) {
1986                         PROC_UNLOCK(p);
1987                         return (error);
1988                 }
1989         }
1990
1991         cred = crhold(p->p_ucred);
1992         if (*pidp != -1)
1993                 PROC_UNLOCK(p);
1994
1995         error = SYSCTL_OUT(req, cred->cr_groups,
1996             cred->cr_ngroups * sizeof(gid_t));
1997         crfree(cred);
1998         return (error);
1999 }
2000
2001 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD,  0, "Process table");
2002
2003 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
2004         CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
2005         "Return entire process table");
2006
2007 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2008         sysctl_kern_proc, "Process table");
2009
2010 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
2011         sysctl_kern_proc, "Process table");
2012
2013 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2014         sysctl_kern_proc, "Process table");
2015
2016 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
2017         CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2018
2019 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
2020         sysctl_kern_proc, "Process table");
2021
2022 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2023         sysctl_kern_proc, "Process table");
2024
2025 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2026         sysctl_kern_proc, "Process table");
2027
2028 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2029         sysctl_kern_proc, "Process table");
2030
2031 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
2032         sysctl_kern_proc, "Return process table, no threads");
2033
2034 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
2035         CTLFLAG_RW | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
2036         sysctl_kern_proc_args, "Process argument list");
2037
2038 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
2039         CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
2040
2041 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
2042         CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
2043         "Process syscall vector name (ABI type)");
2044
2045 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
2046         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2047
2048 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
2049         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2050
2051 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
2052         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2053
2054 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
2055         sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2056
2057 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
2058         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2059
2060 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
2061         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2062
2063 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
2064         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2065
2066 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
2067         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2068
2069 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
2070         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
2071         "Return process table, no threads");
2072
2073 #ifdef COMPAT_FREEBSD7
2074 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
2075         CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
2076 #endif
2077
2078 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
2079         CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
2080
2081 #if defined(STACK) || defined(DDB)
2082 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
2083         CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
2084 #endif
2085
2086 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
2087         CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");