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/mman.h>
  49 #include <sys/mount.h>
  50 #include <sys/mutex.h>
  51 #include <sys/proc.h>
  52 #include <sys/refcount.h>
  53 #include <sys/sbuf.h>
  54 #include <sys/sysent.h>
  55 #include <sys/sched.h>
  56 #include <sys/smp.h>
  57 #include <sys/stack.h>
  58 #include <sys/sysctl.h>
  59 #include <sys/filedesc.h>
  60 #include <sys/tty.h>
  61 #include <sys/signalvar.h>
  62 #include <sys/sdt.h>
  63 #include <sys/sx.h>
  64 #include <sys/user.h>
  65 #include <sys/jail.h>
  66 #include <sys/vnode.h>
  67 #include <sys/eventhandler.h>
  68
  69 #ifdef DDB
  70 #include <ddb/ddb.h>
  71 #endif
  72
  73 #include <vm/vm.h>
  74 #include <vm/vm_extern.h>
  75 #include <vm/pmap.h>
  76 #include <vm/vm_map.h>
  77 #include <vm/vm_object.h>
  78 #include <vm/vm_page.h>
  79 #include <vm/uma.h>
  80
  81 #ifdef COMPAT_FREEBSD32
  82 #include <compat/freebsd32/freebsd32.h>
  83 #include <compat/freebsd32/freebsd32_util.h>
  84 #endif
  85
  86 SDT_PROVIDER_DEFINE(proc);
  87 SDT_PROBE_DEFINE(proc, kernel, ctor, entry, entry);
  88 SDT_PROBE_ARGTYPE(proc, kernel, ctor, entry, 0, "struct proc *");
  89 SDT_PROBE_ARGTYPE(proc, kernel, ctor, entry, 1, "int");
  90 SDT_PROBE_ARGTYPE(proc, kernel, ctor, entry, 2, "void *");
  91 SDT_PROBE_ARGTYPE(proc, kernel, ctor, entry, 3, "int");
  92 SDT_PROBE_DEFINE(proc, kernel, ctor, return, return);
  93 SDT_PROBE_ARGTYPE(proc, kernel, ctor, return, 0, "struct proc *");
  94 SDT_PROBE_ARGTYPE(proc, kernel, ctor, return, 1, "int");
  95 SDT_PROBE_ARGTYPE(proc, kernel, ctor, return, 2, "void *");
  96 SDT_PROBE_ARGTYPE(proc, kernel, ctor, return, 3, "int");
  97 SDT_PROBE_DEFINE(proc, kernel, dtor, entry, entry);
  98 SDT_PROBE_ARGTYPE(proc, kernel, dtor, entry, 0, "struct proc *");
  99 SDT_PROBE_ARGTYPE(proc, kernel, dtor, entry, 1, "int");
 100 SDT_PROBE_ARGTYPE(proc, kernel, dtor, entry, 2, "void *");
 101 SDT_PROBE_ARGTYPE(proc, kernel, dtor, entry, 3, "struct thread *");
 102 SDT_PROBE_DEFINE(proc, kernel, dtor, return, return);
 103 SDT_PROBE_ARGTYPE(proc, kernel, dtor, return, 0, "struct proc *");
 104 SDT_PROBE_ARGTYPE(proc, kernel, dtor, return, 1, "int");
 105 SDT_PROBE_ARGTYPE(proc, kernel, dtor, return, 2, "void *");
 106 SDT_PROBE_DEFINE(proc, kernel, init, entry, entry);
 107 SDT_PROBE_ARGTYPE(proc, kernel, init, entry, 0, "struct proc *");
 108 SDT_PROBE_ARGTYPE(proc, kernel, init, entry, 1, "int");
 109 SDT_PROBE_ARGTYPE(proc, kernel, init, entry, 2, "int");
 110 SDT_PROBE_DEFINE(proc, kernel, init, return, return);
 111 SDT_PROBE_ARGTYPE(proc, kernel, init, return, 0, "struct proc *");
 112 SDT_PROBE_ARGTYPE(proc, kernel, init, return, 1, "int");
 113 SDT_PROBE_ARGTYPE(proc, kernel, init, return, 2, "int");
 114
 115 MALLOC_DEFINE(M_PGRP, "pgrp", "process group header");
 116 MALLOC_DEFINE(M_SESSION, "session", "session header");
 117 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
 118 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
 119
 120 static void doenterpgrp(struct proc *, struct pgrp *);
 121 static void orphanpg(struct pgrp *pg);
 122 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp);
 123 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp);
 124 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp,
 125     int preferthread);
 126 static void pgadjustjobc(struct pgrp *pgrp, int entering);
 127 static void pgdelete(struct pgrp *);
 128 static int proc_ctor(void *mem, int size, void *arg, int flags);
 129 static void proc_dtor(void *mem, int size, void *arg);
 130 static int proc_init(void *mem, int size, int flags);
 131 static void proc_fini(void *mem, int size);
 132 static void pargs_free(struct pargs *pa);
 133
 134 /*
 135  * Other process lists
 136  */
 137 struct pidhashhead *pidhashtbl;
 138 u_long pidhash;
 139 struct pgrphashhead *pgrphashtbl;
 140 u_long pgrphash;
 141 struct proclist allproc;
 142 struct proclist zombproc;
 143 struct sx allproc_lock;
 144 struct sx proctree_lock;
 145 struct mtx ppeers_lock;
 146 uma_zone_t proc_zone;
 147
 148 int kstack_pages = KSTACK_PAGES;
 149 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0,
 150     "Kernel stack size in pages");
 151
 152 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
 153 #ifdef COMPAT_FREEBSD32
 154 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
 155 #endif
 156
 157 /*
 158  * Initialize global process hashing structures.
 159  */
 160 void
 161 procinit()
 162 {
 163
 164         sx_init(&allproc_lock, "allproc");
 165         sx_init(&proctree_lock, "proctree");
 166         mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
 167         LIST_INIT(&allproc);
 168         LIST_INIT(&zombproc);
 169         pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
 170         pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
 171         proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
 172             proc_ctor, proc_dtor, proc_init, proc_fini,
 173             UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
 174         uihashinit();
 175 }
 176
 177 /*
 178  * Prepare a proc for use.
 179  */
 180 static int
 181 proc_ctor(void *mem, int size, void *arg, int flags)
 182 {
 183         struct proc *p;
 184
 185         p = (struct proc *)mem;
 186         SDT_PROBE(proc, kernel, ctor , entry, p, size, arg, flags, 0);
 187         EVENTHANDLER_INVOKE(process_ctor, p);
 188         SDT_PROBE(proc, kernel, ctor , return, p, size, arg, flags, 0);
 189         return (0);
 190 }
 191
 192 /*
 193  * Reclaim a proc after use.
 194  */
 195 static void
 196 proc_dtor(void *mem, int size, void *arg)
 197 {
 198         struct proc *p;
 199         struct thread *td;
 200
 201         /* INVARIANTS checks go here */
 202         p = (struct proc *)mem;
 203         td = FIRST_THREAD_IN_PROC(p);
 204         SDT_PROBE(proc, kernel, dtor, entry, p, size, arg, td, 0);
 205         if (td != NULL) {
 206 #ifdef INVARIANTS
 207                 KASSERT((p->p_numthreads == 1),
 208                     ("bad number of threads in exiting process"));
 209                 KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
 210 #endif
 211                 /* Free all OSD associated to this thread. */
 212                 osd_thread_exit(td);
 213         }
 214         EVENTHANDLER_INVOKE(process_dtor, p);
 215         if (p->p_ksi != NULL)
 216                 KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
 217         SDT_PROBE(proc, kernel, dtor, return, p, size, arg, 0, 0);
 218 }
 219
 220 /*
 221  * Initialize type-stable parts of a proc (when newly created).
 222  */
 223 static int
 224 proc_init(void *mem, int size, int flags)
 225 {
 226         struct proc *p;
 227
 228         p = (struct proc *)mem;
 229         SDT_PROBE(proc, kernel, init, entry, p, size, flags, 0, 0);
 230         p->p_sched = (struct p_sched *)&p[1];
 231         bzero(&p->p_mtx, sizeof(struct mtx));
 232         mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK);
 233         mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_RECURSE);
 234         cv_init(&p->p_pwait, "ppwait");
 235         cv_init(&p->p_dbgwait, "dbgwait");
 236         TAILQ_INIT(&p->p_threads);           /* all threads in proc */
 237         EVENTHANDLER_INVOKE(process_init, p);
 238         p->p_stats = pstats_alloc();
 239         SDT_PROBE(proc, kernel, init, return, p, size, flags, 0, 0);
 240         return (0);
 241 }
 242
 243 /*
 244  * UMA should ensure that this function is never called.
 245  * Freeing a proc structure would violate type stability.
 246  */
 247 static void
 248 proc_fini(void *mem, int size)
 249 {
 250 #ifdef notnow
 251         struct proc *p;
 252
 253         p = (struct proc *)mem;
 254         EVENTHANDLER_INVOKE(process_fini, p);
 255         pstats_free(p->p_stats);
 256         thread_free(FIRST_THREAD_IN_PROC(p));
 257         mtx_destroy(&p->p_mtx);
 258         if (p->p_ksi != NULL)
 259                 ksiginfo_free(p->p_ksi);
 260 #else
 261         panic("proc reclaimed");
 262 #endif
 263 }
 264
 265 /*
 266  * Is p an inferior of the current process?
 267  */
 268 int
 269 inferior(p)
 270         register struct proc *p;
 271 {
 272
 273         sx_assert(&proctree_lock, SX_LOCKED);
 274         for (; p != curproc; p = p->p_pptr)
 275                 if (p->p_pid == 0)
 276                         return (0);
 277         return (1);
 278 }
 279
 280 /*
 281  * Locate a process by number; return only "live" processes -- i.e., neither
 282  * zombies nor newly born but incompletely initialized processes.  By not
 283  * returning processes in the PRS_NEW state, we allow callers to avoid
 284  * testing for that condition to avoid dereferencing p_ucred, et al.
 285  */
 286 struct proc *
 287 pfind(pid)
 288         register pid_t pid;
 289 {
 290         register struct proc *p;
 291
 292         sx_slock(&allproc_lock);
 293         LIST_FOREACH(p, PIDHASH(pid), p_hash)
 294                 if (p->p_pid == pid) {
 295                         PROC_LOCK(p);
 296                         if (p->p_state == PRS_NEW) {
 297                                 PROC_UNLOCK(p);
 298                                 p = NULL;
 299                         }
 300                         break;
 301                 }
 302         sx_sunlock(&allproc_lock);
 303         return (p);
 304 }
 305
 306 static struct proc *
 307 pfind_tid(pid_t tid)
 308 {
 309         struct proc *p;
 310         struct thread *td;
 311
 312         sx_slock(&allproc_lock);
 313         FOREACH_PROC_IN_SYSTEM(p) {
 314                 PROC_LOCK(p);
 315                 if (p->p_state == PRS_NEW) {
 316                         PROC_UNLOCK(p);
 317                         continue;
 318                 }
 319                 FOREACH_THREAD_IN_PROC(p, td) {
 320                         if (td->td_tid == tid)
 321                                 goto found;
 322                 }
 323                 PROC_UNLOCK(p);
 324         }
 325 found:
 326         sx_sunlock(&allproc_lock);
 327         return (p);
 328 }
 329
 330 /*
 331  * Locate a process group by number.
 332  * The caller must hold proctree_lock.
 333  */
 334 struct pgrp *
 335 pgfind(pgid)
 336         register pid_t pgid;
 337 {
 338         register struct pgrp *pgrp;
 339
 340         sx_assert(&proctree_lock, SX_LOCKED);
 341
 342         LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
 343                 if (pgrp->pg_id == pgid) {
 344                         PGRP_LOCK(pgrp);
 345                         return (pgrp);
 346                 }
 347         }
 348         return (NULL);
 349 }
 350
 351 /*
 352  * Locate process and do additional manipulations, depending on flags.
 353  */
 354 int
 355 pget(pid_t pid, int flags, struct proc **pp)
 356 {
 357         struct proc *p;
 358         int error;
 359
 360         if (pid <= PID_MAX)
 361                 p = pfind(pid);
 362         else if ((flags & PGET_NOTID) == 0)
 363                 p = pfind_tid(pid);
 364         else
 365                 p = NULL;
 366         if (p == NULL)
 367                 return (ESRCH);
 368         if ((flags & PGET_CANSEE) != 0) {
 369                 error = p_cansee(curthread, p);
 370                 if (error != 0)
 371                         goto errout;
 372         }
 373         if ((flags & PGET_CANDEBUG) != 0) {
 374                 error = p_candebug(curthread, p);
 375                 if (error != 0)
 376                         goto errout;
 377         }
 378         if ((flags & PGET_ISCURRENT) != 0 && curproc != p) {
 379                 error = EPERM;
 380                 goto errout;
 381         }
 382         if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) {
 383                 error = ESRCH;
 384                 goto errout;
 385         }
 386         if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) {
 387                 /*
 388                  * XXXRW: Not clear ESRCH is the right error during proc
 389                  * execve().
 390                  */
 391                 error = ESRCH;
 392                 goto errout;
 393         }
 394         if ((flags & PGET_HOLD) != 0) {
 395                 _PHOLD(p);
 396                 PROC_UNLOCK(p);
 397         }
 398         *pp = p;
 399         return (0);
 400 errout:
 401         PROC_UNLOCK(p);
 402         return (error);
 403 }
 404
 405 /*
 406  * Create a new process group.
 407  * pgid must be equal to the pid of p.
 408  * Begin a new session if required.
 409  */
 410 int
 411 enterpgrp(p, pgid, pgrp, sess)
 412         register struct proc *p;
 413         pid_t pgid;
 414         struct pgrp *pgrp;
 415         struct session *sess;
 416 {
 417
 418         sx_assert(&proctree_lock, SX_XLOCKED);
 419
 420         KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
 421         KASSERT(p->p_pid == pgid,
 422             ("enterpgrp: new pgrp and pid != pgid"));
 423         KASSERT(pgfind(pgid) == NULL,
 424             ("enterpgrp: pgrp with pgid exists"));
 425         KASSERT(!SESS_LEADER(p),
 426             ("enterpgrp: session leader attempted setpgrp"));
 427
 428         mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
 429
 430         if (sess != NULL) {
 431                 /*
 432                  * new session
 433                  */
 434                 mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
 435                 PROC_LOCK(p);
 436                 p->p_flag &= ~P_CONTROLT;
 437                 PROC_UNLOCK(p);
 438                 PGRP_LOCK(pgrp);
 439                 sess->s_leader = p;
 440                 sess->s_sid = p->p_pid;
 441                 refcount_init(&sess->s_count, 1);
 442                 sess->s_ttyvp = NULL;
 443                 sess->s_ttyp = NULL;
 444                 bcopy(p->p_session->s_login, sess->s_login,
 445                             sizeof(sess->s_login));
 446                 pgrp->pg_session = sess;
 447                 KASSERT(p == curproc,
 448                     ("enterpgrp: mksession and p != curproc"));
 449         } else {
 450                 pgrp->pg_session = p->p_session;
 451                 sess_hold(pgrp->pg_session);
 452                 PGRP_LOCK(pgrp);
 453         }
 454         pgrp->pg_id = pgid;
 455         LIST_INIT(&pgrp->pg_members);
 456
 457         /*
 458          * As we have an exclusive lock of proctree_lock,
 459          * this should not deadlock.
 460          */
 461         LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
 462         pgrp->pg_jobc = 0;
 463         SLIST_INIT(&pgrp->pg_sigiolst);
 464         PGRP_UNLOCK(pgrp);
 465
 466         doenterpgrp(p, pgrp);
 467
 468         return (0);
 469 }
 470
 471 /*
 472  * Move p to an existing process group
 473  */
 474 int
 475 enterthispgrp(p, pgrp)
 476         register struct proc *p;
 477         struct pgrp *pgrp;
 478 {
 479
 480         sx_assert(&proctree_lock, SX_XLOCKED);
 481         PROC_LOCK_ASSERT(p, MA_NOTOWNED);
 482         PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
 483         PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
 484         SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
 485         KASSERT(pgrp->pg_session == p->p_session,
 486                 ("%s: pgrp's session %p, p->p_session %p.\n",
 487                 __func__,
 488                 pgrp->pg_session,
 489                 p->p_session));
 490         KASSERT(pgrp != p->p_pgrp,
 491                 ("%s: p belongs to pgrp.", __func__));
 492
 493         doenterpgrp(p, pgrp);
 494
 495         return (0);
 496 }
 497
 498 /*
 499  * Move p to a process group
 500  */
 501 static void
 502 doenterpgrp(p, pgrp)
 503         struct proc *p;
 504         struct pgrp *pgrp;
 505 {
 506         struct pgrp *savepgrp;
 507
 508         sx_assert(&proctree_lock, SX_XLOCKED);
 509         PROC_LOCK_ASSERT(p, MA_NOTOWNED);
 510         PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
 511         PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
 512         SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
 513
 514         savepgrp = p->p_pgrp;
 515
 516         /*
 517          * Adjust eligibility of affected pgrps to participate in job control.
 518          * Increment eligibility counts before decrementing, otherwise we
 519          * could reach 0 spuriously during the first call.
 520          */
 521         fixjobc(p, pgrp, 1);
 522         fixjobc(p, p->p_pgrp, 0);
 523
 524         PGRP_LOCK(pgrp);
 525         PGRP_LOCK(savepgrp);
 526         PROC_LOCK(p);
 527         LIST_REMOVE(p, p_pglist);
 528         p->p_pgrp = pgrp;
 529         PROC_UNLOCK(p);
 530         LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
 531         PGRP_UNLOCK(savepgrp);
 532         PGRP_UNLOCK(pgrp);
 533         if (LIST_EMPTY(&savepgrp->pg_members))
 534                 pgdelete(savepgrp);
 535 }
 536
 537 /*
 538  * remove process from process group
 539  */
 540 int
 541 leavepgrp(p)
 542         register struct proc *p;
 543 {
 544         struct pgrp *savepgrp;
 545
 546         sx_assert(&proctree_lock, SX_XLOCKED);
 547         savepgrp = p->p_pgrp;
 548         PGRP_LOCK(savepgrp);
 549         PROC_LOCK(p);
 550         LIST_REMOVE(p, p_pglist);
 551         p->p_pgrp = NULL;
 552         PROC_UNLOCK(p);
 553         PGRP_UNLOCK(savepgrp);
 554         if (LIST_EMPTY(&savepgrp->pg_members))
 555                 pgdelete(savepgrp);
 556         return (0);
 557 }
 558
 559 /*
 560  * delete a process group
 561  */
 562 static void
 563 pgdelete(pgrp)
 564         register struct pgrp *pgrp;
 565 {
 566         struct session *savesess;
 567         struct tty *tp;
 568
 569         sx_assert(&proctree_lock, SX_XLOCKED);
 570         PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
 571         SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
 572
 573         /*
 574          * Reset any sigio structures pointing to us as a result of
 575          * F_SETOWN with our pgid.
 576          */
 577         funsetownlst(&pgrp->pg_sigiolst);
 578
 579         PGRP_LOCK(pgrp);
 580         tp = pgrp->pg_session->s_ttyp;
 581         LIST_REMOVE(pgrp, pg_hash);
 582         savesess = pgrp->pg_session;
 583         PGRP_UNLOCK(pgrp);
 584
 585         /* Remove the reference to the pgrp before deallocating it. */
 586         if (tp != NULL) {
 587                 tty_lock(tp);
 588                 tty_rel_pgrp(tp, pgrp);
 589         }
 590
 591         mtx_destroy(&pgrp->pg_mtx);
 592         free(pgrp, M_PGRP);
 593         sess_release(savesess);
 594 }
 595
 596 static void
 597 pgadjustjobc(pgrp, entering)
 598         struct pgrp *pgrp;
 599         int entering;
 600 {
 601
 602         PGRP_LOCK(pgrp);
 603         if (entering)
 604                 pgrp->pg_jobc++;
 605         else {
 606                 --pgrp->pg_jobc;
 607                 if (pgrp->pg_jobc == 0)
 608                         orphanpg(pgrp);
 609         }
 610         PGRP_UNLOCK(pgrp);
 611 }
 612
 613 /*
 614  * Adjust pgrp jobc counters when specified process changes process group.
 615  * We count the number of processes in each process group that "qualify"
 616  * the group for terminal job control (those with a parent in a different
 617  * process group of the same session).  If that count reaches zero, the
 618  * process group becomes orphaned.  Check both the specified process'
 619  * process group and that of its children.
 620  * entering == 0 => p is leaving specified group.
 621  * entering == 1 => p is entering specified group.
 622  */
 623 void
 624 fixjobc(p, pgrp, entering)
 625         register struct proc *p;
 626         register struct pgrp *pgrp;
 627         int entering;
 628 {
 629         register struct pgrp *hispgrp;
 630         register struct session *mysession;
 631
 632         sx_assert(&proctree_lock, SX_LOCKED);
 633         PROC_LOCK_ASSERT(p, MA_NOTOWNED);
 634         PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
 635         SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
 636
 637         /*
 638          * Check p's parent to see whether p qualifies its own process
 639          * group; if so, adjust count for p's process group.
 640          */
 641         mysession = pgrp->pg_session;
 642         if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
 643             hispgrp->pg_session == mysession)
 644                 pgadjustjobc(pgrp, entering);
 645
 646         /*
 647          * Check this process' children to see whether they qualify
 648          * their process groups; if so, adjust counts for children's
 649          * process groups.
 650          */
 651         LIST_FOREACH(p, &p->p_children, p_sibling) {
 652                 hispgrp = p->p_pgrp;
 653                 if (hispgrp == pgrp ||
 654                     hispgrp->pg_session != mysession)
 655                         continue;
 656                 PROC_LOCK(p);
 657                 if (p->p_state == PRS_ZOMBIE) {
 658                         PROC_UNLOCK(p);
 659                         continue;
 660                 }
 661                 PROC_UNLOCK(p);
 662                 pgadjustjobc(hispgrp, entering);
 663         }
 664 }
 665
 666 /*
 667  * A process group has become orphaned;
 668  * if there are any stopped processes in the group,
 669  * hang-up all process in that group.
 670  */
 671 static void
 672 orphanpg(pg)
 673         struct pgrp *pg;
 674 {
 675         register struct proc *p;
 676
 677         PGRP_LOCK_ASSERT(pg, MA_OWNED);
 678
 679         LIST_FOREACH(p, &pg->pg_members, p_pglist) {
 680                 PROC_LOCK(p);
 681                 if (P_SHOULDSTOP(p)) {
 682                         PROC_UNLOCK(p);
 683                         LIST_FOREACH(p, &pg->pg_members, p_pglist) {
 684                                 PROC_LOCK(p);
 685                                 psignal(p, SIGHUP);
 686                                 psignal(p, SIGCONT);
 687                                 PROC_UNLOCK(p);
 688                         }
 689                         return;
 690                 }
 691                 PROC_UNLOCK(p);
 692         }
 693 }
 694
 695 void
 696 sess_hold(struct session *s)
 697 {
 698
 699         refcount_acquire(&s->s_count);
 700 }
 701
 702 void
 703 sess_release(struct session *s)
 704 {
 705
 706         if (refcount_release(&s->s_count)) {
 707                 if (s->s_ttyp != NULL) {
 708                         tty_lock(s->s_ttyp);
 709                         tty_rel_sess(s->s_ttyp, s);
 710                 }
 711                 mtx_destroy(&s->s_mtx);
 712                 free(s, M_SESSION);
 713         }
 714 }
 715
 716 #include "opt_ddb.h"
 717 #ifdef DDB
 718 #include <ddb/ddb.h>
 719
 720 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
 721 {
 722         register struct pgrp *pgrp;
 723         register struct proc *p;
 724         register int i;
 725
 726         for (i = 0; i <= pgrphash; i++) {
 727                 if (!LIST_EMPTY(&pgrphashtbl[i])) {
 728                         printf("\tindx %d\n", i);
 729                         LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
 730                                 printf(
 731                         "\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
 732                                     (void *)pgrp, (long)pgrp->pg_id,
 733                                     (void *)pgrp->pg_session,
 734                                     pgrp->pg_session->s_count,
 735                                     (void *)LIST_FIRST(&pgrp->pg_members));
 736                                 LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
 737                                         printf("\t\tpid %ld addr %p pgrp %p\n",
 738                                             (long)p->p_pid, (void *)p,
 739                                             (void *)p->p_pgrp);
 740                                 }
 741                         }
 742                 }
 743         }
 744 }
 745 #endif /* DDB */
 746
 747 /*
 748  * Calculate the kinfo_proc members which contain process-wide
 749  * informations.
 750  * Must be called with the target process locked.
 751  */
 752 static void
 753 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
 754 {
 755         struct thread *td;
 756
 757         PROC_LOCK_ASSERT(p, MA_OWNED);
 758
 759         kp->ki_estcpu = 0;
 760         kp->ki_pctcpu = 0;
 761         FOREACH_THREAD_IN_PROC(p, td) {
 762                 thread_lock(td);
 763                 kp->ki_pctcpu += sched_pctcpu(td);
 764                 kp->ki_estcpu += td->td_estcpu;
 765                 thread_unlock(td);
 766         }
 767 }
 768
 769 /*
 770  * Clear kinfo_proc and fill in any information that is common
 771  * to all threads in the process.
 772  * Must be called with the target process locked.
 773  */
 774 static void
 775 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
 776 {
 777         struct thread *td0;
 778         struct tty *tp;
 779         struct session *sp;
 780         struct ucred *cred;
 781         struct sigacts *ps;
 782
 783         PROC_LOCK_ASSERT(p, MA_OWNED);
 784         bzero(kp, sizeof(*kp));
 785
 786         kp->ki_structsize = sizeof(*kp);
 787         kp->ki_paddr = p;
 788         kp->ki_addr =/* p->p_addr; */0; /* XXX */
 789         kp->ki_args = p->p_args;
 790         kp->ki_textvp = p->p_textvp;
 791 #ifdef KTRACE
 792         kp->ki_tracep = p->p_tracevp;
 793         kp->ki_traceflag = p->p_traceflag;
 794 #endif
 795         kp->ki_fd = p->p_fd;
 796         kp->ki_vmspace = p->p_vmspace;
 797         kp->ki_flag = p->p_flag;
 798         cred = p->p_ucred;
 799         if (cred) {
 800                 kp->ki_uid = cred->cr_uid;
 801                 kp->ki_ruid = cred->cr_ruid;
 802                 kp->ki_svuid = cred->cr_svuid;
 803                 kp->ki_cr_flags = cred->cr_flags;
 804                 /* XXX bde doesn't like KI_NGROUPS */
 805                 if (cred->cr_ngroups > KI_NGROUPS) {
 806                         kp->ki_ngroups = KI_NGROUPS;
 807                         kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
 808                 } else
 809                         kp->ki_ngroups = cred->cr_ngroups;
 810                 bcopy(cred->cr_groups, kp->ki_groups,
 811                     kp->ki_ngroups * sizeof(gid_t));
 812                 kp->ki_rgid = cred->cr_rgid;
 813                 kp->ki_svgid = cred->cr_svgid;
 814                 /* If jailed(cred), emulate the old P_JAILED flag. */
 815                 if (jailed(cred)) {
 816                         kp->ki_flag |= P_JAILED;
 817                         /* If inside the jail, use 0 as a jail ID. */
 818                         if (cred->cr_prison != curthread->td_ucred->cr_prison)
 819                                 kp->ki_jid = cred->cr_prison->pr_id;
 820                 }
 821         }
 822         ps = p->p_sigacts;
 823         if (ps) {
 824                 mtx_lock(&ps->ps_mtx);
 825                 kp->ki_sigignore = ps->ps_sigignore;
 826                 kp->ki_sigcatch = ps->ps_sigcatch;
 827                 mtx_unlock(&ps->ps_mtx);
 828         }
 829         if (p->p_state != PRS_NEW &&
 830             p->p_state != PRS_ZOMBIE &&
 831             p->p_vmspace != NULL) {
 832                 struct vmspace *vm = p->p_vmspace;
 833
 834                 kp->ki_size = vm->vm_map.size;
 835                 kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
 836                 FOREACH_THREAD_IN_PROC(p, td0) {
 837                         if (!TD_IS_SWAPPED(td0))
 838                                 kp->ki_rssize += td0->td_kstack_pages;
 839                 }
 840                 kp->ki_swrss = vm->vm_swrss;
 841                 kp->ki_tsize = vm->vm_tsize;
 842                 kp->ki_dsize = vm->vm_dsize;
 843                 kp->ki_ssize = vm->vm_ssize;
 844         } else if (p->p_state == PRS_ZOMBIE)
 845                 kp->ki_stat = SZOMB;
 846         if (kp->ki_flag & P_INMEM)
 847                 kp->ki_sflag = PS_INMEM;
 848         else
 849                 kp->ki_sflag = 0;
 850         /* Calculate legacy swtime as seconds since 'swtick'. */
 851         kp->ki_swtime = (ticks - p->p_swtick) / hz;
 852         kp->ki_pid = p->p_pid;
 853         kp->ki_nice = p->p_nice;
 854         PROC_SLOCK(p);
 855         rufetch(p, &kp->ki_rusage);
 856         kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
 857         PROC_SUNLOCK(p);
 858         if ((p->p_flag & P_INMEM) && p->p_stats != NULL) {
 859                 kp->ki_start = p->p_stats->p_start;
 860                 timevaladd(&kp->ki_start, &boottime);
 861                 PROC_SLOCK(p);
 862                 calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
 863                 PROC_SUNLOCK(p);
 864                 calccru(p, &kp->ki_childutime, &kp->ki_childstime);
 865
 866                 /* Some callers want child-times in a single value */
 867                 kp->ki_childtime = kp->ki_childstime;
 868                 timevaladd(&kp->ki_childtime, &kp->ki_childutime);
 869         }
 870
 871         FOREACH_THREAD_IN_PROC(p, td0)
 872                 kp->ki_cow += td0->td_cow;
 873
 874         tp = NULL;
 875         if (p->p_pgrp) {
 876                 kp->ki_pgid = p->p_pgrp->pg_id;
 877                 kp->ki_jobc = p->p_pgrp->pg_jobc;
 878                 sp = p->p_pgrp->pg_session;
 879
 880                 if (sp != NULL) {
 881                         kp->ki_sid = sp->s_sid;
 882                         SESS_LOCK(sp);
 883                         strlcpy(kp->ki_login, sp->s_login,
 884                             sizeof(kp->ki_login));
 885                         if (sp->s_ttyvp)
 886                                 kp->ki_kiflag |= KI_CTTY;
 887                         if (SESS_LEADER(p))
 888                                 kp->ki_kiflag |= KI_SLEADER;
 889                         /* XXX proctree_lock */
 890                         tp = sp->s_ttyp;
 891                         SESS_UNLOCK(sp);
 892                 }
 893         }
 894         if ((p->p_flag & P_CONTROLT) && tp != NULL) {
 895                 kp->ki_tdev = tty_udev(tp);
 896                 kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
 897                 if (tp->t_session)
 898                         kp->ki_tsid = tp->t_session->s_sid;
 899         } else
 900                 kp->ki_tdev = NODEV;
 901         if (p->p_comm[0] != '\0')
 902                 strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
 903         if (p->p_sysent && p->p_sysent->sv_name != NULL &&
 904             p->p_sysent->sv_name[0] != '\0')
 905                 strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
 906         kp->ki_siglist = p->p_siglist;
 907         kp->ki_xstat = p->p_xstat;
 908         kp->ki_acflag = p->p_acflag;
 909         kp->ki_lock = p->p_lock;
 910         if (p->p_pptr)
 911                 kp->ki_ppid = p->p_pptr->p_pid;
 912 }
 913
 914 /*
 915  * Fill in information that is thread specific.  Must be called with
 916  * target process locked.  If 'preferthread' is set, overwrite certain
 917  * process-related fields that are maintained for both threads and
 918  * processes.
 919  */
 920 static void
 921 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
 922 {
 923         struct proc *p;
 924
 925         p = td->td_proc;
 926         kp->ki_tdaddr = td;
 927         PROC_LOCK_ASSERT(p, MA_OWNED);
 928
 929         thread_lock(td);
 930         if (td->td_wmesg != NULL)
 931                 strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
 932         else
 933                 bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
 934         strlcpy(kp->ki_ocomm, td->td_name, sizeof(kp->ki_ocomm));
 935         if (TD_ON_LOCK(td)) {
 936                 kp->ki_kiflag |= KI_LOCKBLOCK;
 937                 strlcpy(kp->ki_lockname, td->td_lockname,
 938                     sizeof(kp->ki_lockname));
 939         } else {
 940                 kp->ki_kiflag &= ~KI_LOCKBLOCK;
 941                 bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
 942         }
 943
 944         if (p->p_state == PRS_NORMAL) { /* approximate. */
 945                 if (TD_ON_RUNQ(td) ||
 946                     TD_CAN_RUN(td) ||
 947                     TD_IS_RUNNING(td)) {
 948                         kp->ki_stat = SRUN;
 949                 } else if (P_SHOULDSTOP(p)) {
 950                         kp->ki_stat = SSTOP;
 951                 } else if (TD_IS_SLEEPING(td)) {
 952                         kp->ki_stat = SSLEEP;
 953                 } else if (TD_ON_LOCK(td)) {
 954                         kp->ki_stat = SLOCK;
 955                 } else {
 956                         kp->ki_stat = SWAIT;
 957                 }
 958         } else if (p->p_state == PRS_ZOMBIE) {
 959                 kp->ki_stat = SZOMB;
 960         } else {
 961                 kp->ki_stat = SIDL;
 962         }
 963
 964         /* Things in the thread */
 965         kp->ki_wchan = td->td_wchan;
 966         kp->ki_pri.pri_level = td->td_priority;
 967         kp->ki_pri.pri_native = td->td_base_pri;
 968         kp->ki_lastcpu = td->td_lastcpu;
 969         kp->ki_oncpu = td->td_oncpu;
 970         kp->ki_tdflags = td->td_flags;
 971         kp->ki_tid = td->td_tid;
 972         kp->ki_numthreads = p->p_numthreads;
 973         kp->ki_pcb = td->td_pcb;
 974         kp->ki_kstack = (void *)td->td_kstack;
 975         kp->ki_slptime = (ticks - td->td_slptick) / hz;
 976         kp->ki_pri.pri_class = td->td_pri_class;
 977         kp->ki_pri.pri_user = td->td_user_pri;
 978
 979         if (preferthread) {
 980                 kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
 981                 kp->ki_pctcpu = sched_pctcpu(td);
 982                 kp->ki_estcpu = td->td_estcpu;
 983                 kp->ki_cow = td->td_cow;
 984         }
 985
 986         /* We can't get this anymore but ps etc never used it anyway. */
 987         kp->ki_rqindex = 0;
 988
 989         if (preferthread)
 990                 kp->ki_siglist = td->td_siglist;
 991         kp->ki_sigmask = td->td_sigmask;
 992         thread_unlock(td);
 993 }
 994
 995 /*
 996  * Fill in a kinfo_proc structure for the specified process.
 997  * Must be called with the target process locked.
 998  */
 999 void
1000 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1001 {
1002
1003         MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1004
1005         fill_kinfo_proc_only(p, kp);
1006         fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1007         fill_kinfo_aggregate(p, kp);
1008 }
1009
1010 struct pstats *
1011 pstats_alloc(void)
1012 {
1013
1014         return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1015 }
1016
1017 /*
1018  * Copy parts of p_stats; zero the rest of p_stats (statistics).
1019  */
1020 void
1021 pstats_fork(struct pstats *src, struct pstats *dst)
1022 {
1023
1024         bzero(&dst->pstat_startzero,
1025             __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1026         bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1027             __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1028 }
1029
1030 void
1031 pstats_free(struct pstats *ps)
1032 {
1033
1034         free(ps, M_SUBPROC);
1035 }
1036
1037 /*
1038  * Locate a zombie process by number
1039  */
1040 struct proc *
1041 zpfind(pid_t pid)
1042 {
1043         struct proc *p;
1044
1045         sx_slock(&allproc_lock);
1046         LIST_FOREACH(p, &zombproc, p_list)
1047                 if (p->p_pid == pid) {
1048                         PROC_LOCK(p);
1049                         break;
1050                 }
1051         sx_sunlock(&allproc_lock);
1052         return (p);
1053 }
1054
1055 #define KERN_PROC_ZOMBMASK      0x3
1056 #define KERN_PROC_NOTHREADS     0x4
1057
1058 #ifdef COMPAT_FREEBSD32
1059
1060 /*
1061  * This function is typically used to copy out the kernel address, so
1062  * it can be replaced by assignment of zero.
1063  */
1064 static inline uint32_t
1065 ptr32_trim(void *ptr)
1066 {
1067         uintptr_t uptr;
1068
1069         uptr = (uintptr_t)ptr;
1070         return ((uptr > UINT_MAX) ? 0 : uptr);
1071 }
1072
1073 #define PTRTRIM_CP(src,dst,fld) \
1074         do { (dst).fld = ptr32_trim((src).fld); } while (0)
1075
1076 static void
1077 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1078 {
1079         int i;
1080
1081         bzero(ki32, sizeof(struct kinfo_proc32));
1082         ki32->ki_structsize = sizeof(struct kinfo_proc32);
1083         CP(*ki, *ki32, ki_layout);
1084         PTRTRIM_CP(*ki, *ki32, ki_args);
1085         PTRTRIM_CP(*ki, *ki32, ki_paddr);
1086         PTRTRIM_CP(*ki, *ki32, ki_addr);
1087         PTRTRIM_CP(*ki, *ki32, ki_tracep);
1088         PTRTRIM_CP(*ki, *ki32, ki_textvp);
1089         PTRTRIM_CP(*ki, *ki32, ki_fd);
1090         PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1091         PTRTRIM_CP(*ki, *ki32, ki_wchan);
1092         CP(*ki, *ki32, ki_pid);
1093         CP(*ki, *ki32, ki_ppid);
1094         CP(*ki, *ki32, ki_pgid);
1095         CP(*ki, *ki32, ki_tpgid);
1096         CP(*ki, *ki32, ki_sid);
1097         CP(*ki, *ki32, ki_tsid);
1098         CP(*ki, *ki32, ki_jobc);
1099         CP(*ki, *ki32, ki_tdev);
1100         CP(*ki, *ki32, ki_siglist);
1101         CP(*ki, *ki32, ki_sigmask);
1102         CP(*ki, *ki32, ki_sigignore);
1103         CP(*ki, *ki32, ki_sigcatch);
1104         CP(*ki, *ki32, ki_uid);
1105         CP(*ki, *ki32, ki_ruid);
1106         CP(*ki, *ki32, ki_svuid);
1107         CP(*ki, *ki32, ki_rgid);
1108         CP(*ki, *ki32, ki_svgid);
1109         CP(*ki, *ki32, ki_ngroups);
1110         for (i = 0; i < KI_NGROUPS; i++)
1111                 CP(*ki, *ki32, ki_groups[i]);
1112         CP(*ki, *ki32, ki_size);
1113         CP(*ki, *ki32, ki_rssize);
1114         CP(*ki, *ki32, ki_swrss);
1115         CP(*ki, *ki32, ki_tsize);
1116         CP(*ki, *ki32, ki_dsize);
1117         CP(*ki, *ki32, ki_ssize);
1118         CP(*ki, *ki32, ki_xstat);
1119         CP(*ki, *ki32, ki_acflag);
1120         CP(*ki, *ki32, ki_pctcpu);
1121         CP(*ki, *ki32, ki_estcpu);
1122         CP(*ki, *ki32, ki_slptime);
1123         CP(*ki, *ki32, ki_swtime);
1124         CP(*ki, *ki32, ki_cow);
1125         CP(*ki, *ki32, ki_runtime);
1126         TV_CP(*ki, *ki32, ki_start);
1127         TV_CP(*ki, *ki32, ki_childtime);
1128         CP(*ki, *ki32, ki_flag);
1129         CP(*ki, *ki32, ki_kiflag);
1130         CP(*ki, *ki32, ki_traceflag);
1131         CP(*ki, *ki32, ki_stat);
1132         CP(*ki, *ki32, ki_nice);
1133         CP(*ki, *ki32, ki_lock);
1134         CP(*ki, *ki32, ki_rqindex);
1135         CP(*ki, *ki32, ki_oncpu);
1136         CP(*ki, *ki32, ki_lastcpu);
1137         bcopy(ki->ki_ocomm, ki32->ki_ocomm, OCOMMLEN + 1);
1138         bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1139         bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1140         bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1141         bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1142         bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1143         CP(*ki, *ki32, ki_cr_flags);
1144         CP(*ki, *ki32, ki_jid);
1145         CP(*ki, *ki32, ki_numthreads);
1146         CP(*ki, *ki32, ki_tid);
1147         CP(*ki, *ki32, ki_pri);
1148         freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1149         freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1150         PTRTRIM_CP(*ki, *ki32, ki_pcb);
1151         PTRTRIM_CP(*ki, *ki32, ki_kstack);
1152         PTRTRIM_CP(*ki, *ki32, ki_udata);
1153         CP(*ki, *ki32, ki_sflag);
1154         CP(*ki, *ki32, ki_tdflags);
1155 }
1156
1157 static int
1158 sysctl_out_proc_copyout(struct kinfo_proc *ki, struct sysctl_req *req)
1159 {
1160         struct kinfo_proc32 ki32;
1161         int error;
1162
1163         if (req->flags & SCTL_MASK32) {
1164                 freebsd32_kinfo_proc_out(ki, &ki32);
1165                 error = SYSCTL_OUT(req, &ki32, sizeof(struct kinfo_proc32));
1166         } else
1167                 error = SYSCTL_OUT(req, ki, sizeof(struct kinfo_proc));
1168         return (error);
1169 }
1170 #else
1171 static int
1172 sysctl_out_proc_copyout(struct kinfo_proc *ki, struct sysctl_req *req)
1173 {
1174
1175         return (SYSCTL_OUT(req, ki, sizeof(struct kinfo_proc)));
1176 }
1177 #endif
1178
1179 /*
1180  * Must be called with the process locked and will return with it unlocked.
1181  */
1182 static int
1183 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
1184 {
1185         struct thread *td;
1186         struct kinfo_proc kinfo_proc;
1187         int error = 0;
1188         struct proc *np;
1189         pid_t pid = p->p_pid;
1190
1191         PROC_LOCK_ASSERT(p, MA_OWNED);
1192         MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1193
1194         fill_kinfo_proc(p, &kinfo_proc);
1195         if (flags & KERN_PROC_NOTHREADS)
1196                 error = sysctl_out_proc_copyout(&kinfo_proc, req);
1197         else {
1198                 FOREACH_THREAD_IN_PROC(p, td) {
1199                         fill_kinfo_thread(td, &kinfo_proc, 1);
1200                         error = sysctl_out_proc_copyout(&kinfo_proc, req);
1201                         if (error)
1202                                 break;
1203                 }
1204         }
1205         PROC_UNLOCK(p);
1206         if (error)
1207                 return (error);
1208         if (flags & KERN_PROC_ZOMBMASK)
1209                 np = zpfind(pid);
1210         else {
1211                 if (pid == 0)
1212                         return (0);
1213                 np = pfind(pid);
1214         }
1215         if (np == NULL)
1216                 return (ESRCH);
1217         if (np != p) {
1218                 PROC_UNLOCK(np);
1219                 return (ESRCH);
1220         }
1221         PROC_UNLOCK(np);
1222         return (0);
1223 }
1224
1225 static int
1226 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1227 {
1228         int *name = (int*) arg1;
1229         u_int namelen = arg2;
1230         struct proc *p;
1231         int flags, doingzomb, oid_number;
1232         int error = 0;
1233
1234         oid_number = oidp->oid_number;
1235         if (oid_number != KERN_PROC_ALL &&
1236             (oid_number & KERN_PROC_INC_THREAD) == 0)
1237                 flags = KERN_PROC_NOTHREADS;
1238         else {
1239                 flags = 0;
1240                 oid_number &= ~KERN_PROC_INC_THREAD;
1241         }
1242         if (oid_number == KERN_PROC_PID) {
1243                 if (namelen != 1)
1244                         return (EINVAL);
1245                 error = sysctl_wire_old_buffer(req, 0);
1246                 if (error)
1247                         return (error);
1248                 error = pget((pid_t)name[0], PGET_CANSEE, &p);
1249                 if (error != 0)
1250                         return (error);
1251                 error = sysctl_out_proc(p, req, flags);
1252                 return (error);
1253         }
1254
1255         switch (oid_number) {
1256         case KERN_PROC_ALL:
1257                 if (namelen != 0)
1258                         return (EINVAL);
1259                 break;
1260         case KERN_PROC_PROC:
1261                 if (namelen != 0 && namelen != 1)
1262                         return (EINVAL);
1263                 break;
1264         default:
1265                 if (namelen != 1)
1266                         return (EINVAL);
1267                 break;
1268         }
1269
1270         if (!req->oldptr) {
1271                 /* overestimate by 5 procs */
1272                 error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1273                 if (error)
1274                         return (error);
1275         }
1276         error = sysctl_wire_old_buffer(req, 0);
1277         if (error != 0)
1278                 return (error);
1279         sx_slock(&allproc_lock);
1280         for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) {
1281                 if (!doingzomb)
1282                         p = LIST_FIRST(&allproc);
1283                 else
1284                         p = LIST_FIRST(&zombproc);
1285                 for (; p != 0; p = LIST_NEXT(p, p_list)) {
1286                         /*
1287                          * Skip embryonic processes.
1288                          */
1289                         PROC_LOCK(p);
1290                         if (p->p_state == PRS_NEW) {
1291                                 PROC_UNLOCK(p);
1292                                 continue;
1293                         }
1294                         KASSERT(p->p_ucred != NULL,
1295                             ("process credential is NULL for non-NEW proc"));
1296                         /*
1297                          * Show a user only appropriate processes.
1298                          */
1299                         if (p_cansee(curthread, p)) {
1300                                 PROC_UNLOCK(p);
1301                                 continue;
1302                         }
1303                         /*
1304                          * TODO - make more efficient (see notes below).
1305                          * do by session.
1306                          */
1307                         switch (oid_number) {
1308
1309                         case KERN_PROC_GID:
1310                                 if (p->p_ucred->cr_gid != (gid_t)name[0]) {
1311                                         PROC_UNLOCK(p);
1312                                         continue;
1313                                 }
1314                                 break;
1315
1316                         case KERN_PROC_PGRP:
1317                                 /* could do this by traversing pgrp */
1318                                 if (p->p_pgrp == NULL ||
1319                                     p->p_pgrp->pg_id != (pid_t)name[0]) {
1320                                         PROC_UNLOCK(p);
1321                                         continue;
1322                                 }
1323                                 break;
1324
1325                         case KERN_PROC_RGID:
1326                                 if (p->p_ucred->cr_rgid != (gid_t)name[0]) {
1327                                         PROC_UNLOCK(p);
1328                                         continue;
1329                                 }
1330                                 break;
1331
1332                         case KERN_PROC_SESSION:
1333                                 if (p->p_session == NULL ||
1334                                     p->p_session->s_sid != (pid_t)name[0]) {
1335                                         PROC_UNLOCK(p);
1336                                         continue;
1337                                 }
1338                                 break;
1339
1340                         case KERN_PROC_TTY:
1341                                 if ((p->p_flag & P_CONTROLT) == 0 ||
1342                                     p->p_session == NULL) {
1343                                         PROC_UNLOCK(p);
1344                                         continue;
1345                                 }
1346                                 /* XXX proctree_lock */
1347                                 SESS_LOCK(p->p_session);
1348                                 if (p->p_session->s_ttyp == NULL ||
1349                                     tty_udev(p->p_session->s_ttyp) !=
1350                                     (dev_t)name[0]) {
1351                                         SESS_UNLOCK(p->p_session);
1352                                         PROC_UNLOCK(p);
1353                                         continue;
1354                                 }
1355                                 SESS_UNLOCK(p->p_session);
1356                                 break;
1357
1358                         case KERN_PROC_UID:
1359                                 if (p->p_ucred->cr_uid != (uid_t)name[0]) {
1360                                         PROC_UNLOCK(p);
1361                                         continue;
1362                                 }
1363                                 break;
1364
1365                         case KERN_PROC_RUID:
1366                                 if (p->p_ucred->cr_ruid != (uid_t)name[0]) {
1367                                         PROC_UNLOCK(p);
1368                                         continue;
1369                                 }
1370                                 break;
1371
1372                         case KERN_PROC_PROC:
1373                                 break;
1374
1375                         default:
1376                                 break;
1377
1378                         }
1379
1380                         error = sysctl_out_proc(p, req, flags | doingzomb);
1381                         if (error) {
1382                                 sx_sunlock(&allproc_lock);
1383                                 return (error);
1384                         }
1385                 }
1386         }
1387         sx_sunlock(&allproc_lock);
1388         return (0);
1389 }
1390
1391 struct pargs *
1392 pargs_alloc(int len)
1393 {
1394         struct pargs *pa;
1395
1396         pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1397                 M_WAITOK);
1398         refcount_init(&pa->ar_ref, 1);
1399         pa->ar_length = len;
1400         return (pa);
1401 }
1402
1403 static void
1404 pargs_free(struct pargs *pa)
1405 {
1406
1407         free(pa, M_PARGS);
1408 }
1409
1410 void
1411 pargs_hold(struct pargs *pa)
1412 {
1413
1414         if (pa == NULL)
1415                 return;
1416         refcount_acquire(&pa->ar_ref);
1417 }
1418
1419 void
1420 pargs_drop(struct pargs *pa)
1421 {
1422
1423         if (pa == NULL)
1424                 return;
1425         if (refcount_release(&pa->ar_ref))
1426                 pargs_free(pa);
1427 }
1428
1429 /*
1430  * This sysctl allows a process to retrieve the argument list or process
1431  * title for another process without groping around in the address space
1432  * of the other process.  It also allow a process to set its own "process
1433  * title to a string of its own choice.
1434  */
1435 static int
1436 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
1437 {
1438         int *name = (int*) arg1;
1439         u_int namelen = arg2;
1440         struct pargs *newpa, *pa;
1441         struct proc *p;
1442         int flags, error = 0;
1443
1444         if (namelen != 1)
1445                 return (EINVAL);
1446
1447         flags = PGET_CANSEE;
1448         if (req->newptr != NULL)
1449                 flags |= PGET_ISCURRENT;
1450         error = pget((pid_t)name[0], flags, &p);
1451         if (error)
1452                 return (error);
1453
1454         pa = p->p_args;
1455         pargs_hold(pa);
1456         PROC_UNLOCK(p);
1457         if (pa != NULL)
1458                 error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
1459         pargs_drop(pa);
1460         if (error != 0 || req->newptr == NULL)
1461                 return (error);
1462
1463         if (req->newlen + sizeof(struct pargs) > ps_arg_cache_limit)
1464                 return (ENOMEM);
1465         newpa = pargs_alloc(req->newlen);
1466         error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
1467         if (error != 0) {
1468                 pargs_free(newpa);
1469                 return (error);
1470         }
1471         PROC_LOCK(p);
1472         pa = p->p_args;
1473         p->p_args = newpa;
1474         PROC_UNLOCK(p);
1475         pargs_drop(pa);
1476         return (0);
1477 }
1478
1479 /*
1480  * This sysctl allows a process to retrieve the path of the executable for
1481  * itself or another process.
1482  */
1483 static int
1484 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
1485 {
1486         pid_t *pidp = (pid_t *)arg1;
1487         unsigned int arglen = arg2;
1488         struct proc *p;
1489         struct vnode *vp;
1490         char *retbuf, *freebuf;
1491         int error, vfslocked;
1492
1493         if (arglen != 1)
1494                 return (EINVAL);
1495         if (*pidp == -1) {      /* -1 means this process */
1496                 p = req->td->td_proc;
1497         } else {
1498                 error = pget(*pidp, PGET_CANSEE, &p);
1499                 if (error != 0)
1500                         return (error);
1501         }
1502
1503         vp = p->p_textvp;
1504         if (vp == NULL) {
1505                 if (*pidp != -1)
1506                         PROC_UNLOCK(p);
1507                 return (0);
1508         }
1509         vref(vp);
1510         if (*pidp != -1)
1511                 PROC_UNLOCK(p);
1512         error = vn_fullpath(req->td, vp, &retbuf, &freebuf);
1513         vfslocked = VFS_LOCK_GIANT(vp->v_mount);
1514         vrele(vp);
1515         VFS_UNLOCK_GIANT(vfslocked);
1516         if (error)
1517                 return (error);
1518         error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
1519         free(freebuf, M_TEMP);
1520         return (error);
1521 }
1522
1523 static int
1524 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
1525 {
1526         struct proc *p;
1527         char *sv_name;
1528         int *name;
1529         int namelen;
1530         int error;
1531
1532         namelen = arg2;
1533         if (namelen != 1)
1534                 return (EINVAL);
1535
1536         name = (int *)arg1;
1537         error = pget((pid_t)name[0], PGET_CANSEE, &p);
1538         if (error != 0)
1539                 return (error);
1540         sv_name = p->p_sysent->sv_name;
1541         PROC_UNLOCK(p);
1542         return (sysctl_handle_string(oidp, sv_name, 0, req));
1543 }
1544
1545 #ifdef KINFO_OVMENTRY_SIZE
1546 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
1547 #endif
1548
1549 #ifdef COMPAT_FREEBSD7
1550 static int
1551 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
1552 {
1553         vm_map_entry_t entry, tmp_entry;
1554         unsigned int last_timestamp;
1555         char *fullpath, *freepath;
1556         struct kinfo_ovmentry *kve;
1557         struct vattr va;
1558         struct ucred *cred;
1559         int error, *name;
1560         struct vnode *vp;
1561         struct proc *p;
1562         vm_map_t map;
1563         struct vmspace *vm;
1564
1565         name = (int *)arg1;
1566         error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1567         if (error != 0)
1568                 return (error);
1569         vm = vmspace_acquire_ref(p);
1570         if (vm == NULL) {
1571                 PRELE(p);
1572                 return (ESRCH);
1573         }
1574         kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
1575
1576         map = &p->p_vmspace->vm_map;    /* XXXRW: More locking required? */
1577         vm_map_lock_read(map);
1578         for (entry = map->header.next; entry != &map->header;
1579             entry = entry->next) {
1580                 vm_object_t obj, tobj, lobj;
1581                 vm_offset_t addr;
1582                 int vfslocked;
1583
1584                 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
1585                         continue;
1586
1587                 bzero(kve, sizeof(*kve));
1588                 kve->kve_structsize = sizeof(*kve);
1589
1590                 kve->kve_private_resident = 0;
1591                 obj = entry->object.vm_object;
1592                 if (obj != NULL) {
1593                         VM_OBJECT_LOCK(obj);
1594                         if (obj->shadow_count == 1)
1595                                 kve->kve_private_resident =
1596                                     obj->resident_page_count;
1597                 }
1598                 kve->kve_resident = 0;
1599                 addr = entry->start;
1600                 while (addr < entry->end) {
1601                         if (pmap_extract(map->pmap, addr))
1602                                 kve->kve_resident++;
1603                         addr += PAGE_SIZE;
1604                 }
1605
1606                 for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
1607                         if (tobj != obj)
1608                                 VM_OBJECT_LOCK(tobj);
1609                         if (lobj != obj)
1610                                 VM_OBJECT_UNLOCK(lobj);
1611                         lobj = tobj;
1612                 }
1613
1614                 kve->kve_start = (void*)entry->start;
1615                 kve->kve_end = (void*)entry->end;
1616                 kve->kve_offset = (off_t)entry->offset;
1617
1618                 if (entry->protection & VM_PROT_READ)
1619                         kve->kve_protection |= KVME_PROT_READ;
1620                 if (entry->protection & VM_PROT_WRITE)
1621                         kve->kve_protection |= KVME_PROT_WRITE;
1622                 if (entry->protection & VM_PROT_EXECUTE)
1623                         kve->kve_protection |= KVME_PROT_EXEC;
1624
1625                 if (entry->eflags & MAP_ENTRY_COW)
1626                         kve->kve_flags |= KVME_FLAG_COW;
1627                 if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
1628                         kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
1629                 if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
1630                         kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
1631
1632                 last_timestamp = map->timestamp;
1633                 vm_map_unlock_read(map);
1634
1635                 kve->kve_fileid = 0;
1636                 kve->kve_fsid = 0;
1637                 freepath = NULL;
1638                 fullpath = "";
1639                 if (lobj) {
1640                         vp = NULL;
1641                         switch (lobj->type) {
1642                         case OBJT_DEFAULT:
1643                                 kve->kve_type = KVME_TYPE_DEFAULT;
1644                                 break;
1645                         case OBJT_VNODE:
1646                                 kve->kve_type = KVME_TYPE_VNODE;
1647                                 vp = lobj->handle;
1648                                 vref(vp);
1649                                 break;
1650                         case OBJT_SWAP:
1651                                 kve->kve_type = KVME_TYPE_SWAP;
1652                                 break;
1653                         case OBJT_DEVICE:
1654                                 kve->kve_type = KVME_TYPE_DEVICE;
1655                                 break;
1656                         case OBJT_PHYS:
1657                                 kve->kve_type = KVME_TYPE_PHYS;
1658                                 break;
1659                         case OBJT_DEAD:
1660                                 kve->kve_type = KVME_TYPE_DEAD;
1661                                 break;
1662                         case OBJT_SG:
1663                                 kve->kve_type = KVME_TYPE_SG;
1664                                 break;
1665                         default:
1666                                 kve->kve_type = KVME_TYPE_UNKNOWN;
1667                                 break;
1668                         }
1669                         if (lobj != obj)
1670                                 VM_OBJECT_UNLOCK(lobj);
1671
1672                         kve->kve_ref_count = obj->ref_count;
1673                         kve->kve_shadow_count = obj->shadow_count;
1674                         VM_OBJECT_UNLOCK(obj);
1675                         if (vp != NULL) {
1676                                 vn_fullpath(curthread, vp, &fullpath,
1677                                     &freepath);
1678                                 cred = curthread->td_ucred;
1679                                 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
1680                                 vn_lock(vp, LK_SHARED | LK_RETRY);
1681                                 if (VOP_GETATTR(vp, &va, cred) == 0) {
1682                                         kve->kve_fileid = va.va_fileid;
1683                                         kve->kve_fsid = va.va_fsid;
1684                                 }
1685                                 vput(vp);
1686                                 VFS_UNLOCK_GIANT(vfslocked);
1687                         }
1688                 } else {
1689                         kve->kve_type = KVME_TYPE_NONE;
1690                         kve->kve_ref_count = 0;
1691                         kve->kve_shadow_count = 0;
1692                 }
1693
1694                 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
1695                 if (freepath != NULL)
1696                         free(freepath, M_TEMP);
1697
1698                 error = SYSCTL_OUT(req, kve, sizeof(*kve));
1699                 vm_map_lock_read(map);
1700                 if (error)
1701                         break;
1702                 if (last_timestamp != map->timestamp) {
1703                         vm_map_lookup_entry(map, addr - 1, &tmp_entry);
1704                         entry = tmp_entry;
1705                 }
1706         }
1707         vm_map_unlock_read(map);
1708         vmspace_free(vm);
1709         PRELE(p);
1710         free(kve, M_TEMP);
1711         return (error);
1712 }
1713 #endif  /* COMPAT_FREEBSD7 */
1714
1715 #ifdef KINFO_VMENTRY_SIZE
1716 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
1717 #endif
1718
1719 static int
1720 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
1721 {
1722         vm_map_entry_t entry, tmp_entry;
1723         unsigned int last_timestamp;
1724         char *fullpath, *freepath;
1725         struct kinfo_vmentry *kve;
1726         struct vattr va;
1727         struct ucred *cred;
1728         int error, *name;
1729         struct vnode *vp;
1730         struct proc *p;
1731         struct vmspace *vm;
1732         vm_map_t map;
1733
1734         name = (int *)arg1;
1735         error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1736         if (error != 0)
1737                 return (error);
1738         vm = vmspace_acquire_ref(p);
1739         if (vm == NULL) {
1740                 PRELE(p);
1741                 return (ESRCH);
1742         }
1743         kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
1744
1745         map = &vm->vm_map;      /* XXXRW: More locking required? */
1746         vm_map_lock_read(map);
1747         for (entry = map->header.next; entry != &map->header;
1748             entry = entry->next) {
1749                 vm_object_t obj, tobj, lobj;
1750                 vm_offset_t addr;
1751                 int vfslocked, mincoreinfo;
1752
1753                 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
1754                         continue;
1755
1756                 bzero(kve, sizeof(*kve));
1757
1758                 kve->kve_private_resident = 0;
1759                 obj = entry->object.vm_object;
1760                 if (obj != NULL) {
1761                         VM_OBJECT_LOCK(obj);
1762                         if (obj->shadow_count == 1)
1763                                 kve->kve_private_resident =
1764                                     obj->resident_page_count;
1765                 }
1766                 kve->kve_resident = 0;
1767                 addr = entry->start;
1768                 while (addr < entry->end) {
1769                         mincoreinfo = pmap_mincore(map->pmap, addr);
1770                         if (mincoreinfo & MINCORE_INCORE)
1771                                 kve->kve_resident++;
1772                         if (mincoreinfo & MINCORE_SUPER)
1773                                 kve->kve_flags |= KVME_FLAG_SUPER;
1774                         addr += PAGE_SIZE;
1775                 }
1776
1777                 for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
1778                         if (tobj != obj)
1779                                 VM_OBJECT_LOCK(tobj);
1780                         if (lobj != obj)
1781                                 VM_OBJECT_UNLOCK(lobj);
1782                         lobj = tobj;
1783                 }
1784
1785                 kve->kve_start = entry->start;
1786                 kve->kve_end = entry->end;
1787                 kve->kve_offset = entry->offset;
1788
1789                 if (entry->protection & VM_PROT_READ)
1790                         kve->kve_protection |= KVME_PROT_READ;
1791                 if (entry->protection & VM_PROT_WRITE)
1792                         kve->kve_protection |= KVME_PROT_WRITE;
1793                 if (entry->protection & VM_PROT_EXECUTE)
1794                         kve->kve_protection |= KVME_PROT_EXEC;
1795
1796                 if (entry->eflags & MAP_ENTRY_COW)
1797                         kve->kve_flags |= KVME_FLAG_COW;
1798                 if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
1799                         kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
1800                 if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
1801                         kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
1802
1803                 last_timestamp = map->timestamp;
1804                 vm_map_unlock_read(map);
1805
1806                 kve->kve_fileid = 0;
1807                 kve->kve_fsid = 0;
1808                 freepath = NULL;
1809                 fullpath = "";
1810                 if (lobj) {
1811                         vp = NULL;
1812                         switch (lobj->type) {
1813                         case OBJT_DEFAULT:
1814                                 kve->kve_type = KVME_TYPE_DEFAULT;
1815                                 break;
1816                         case OBJT_VNODE:
1817                                 kve->kve_type = KVME_TYPE_VNODE;
1818                                 vp = lobj->handle;
1819                                 vref(vp);
1820                                 break;
1821                         case OBJT_SWAP:
1822                                 kve->kve_type = KVME_TYPE_SWAP;
1823                                 break;
1824                         case OBJT_DEVICE:
1825                                 kve->kve_type = KVME_TYPE_DEVICE;
1826                                 break;
1827                         case OBJT_PHYS:
1828                                 kve->kve_type = KVME_TYPE_PHYS;
1829                                 break;
1830                         case OBJT_DEAD:
1831                                 kve->kve_type = KVME_TYPE_DEAD;
1832                                 break;
1833                         case OBJT_SG:
1834                                 kve->kve_type = KVME_TYPE_SG;
1835                                 break;
1836                         default:
1837                                 kve->kve_type = KVME_TYPE_UNKNOWN;
1838                                 break;
1839                         }
1840                         if (lobj != obj)
1841                                 VM_OBJECT_UNLOCK(lobj);
1842
1843                         kve->kve_ref_count = obj->ref_count;
1844                         kve->kve_shadow_count = obj->shadow_count;
1845                         VM_OBJECT_UNLOCK(obj);
1846                         if (vp != NULL) {
1847                                 vn_fullpath(curthread, vp, &fullpath,
1848                                     &freepath);
1849                                 cred = curthread->td_ucred;
1850                                 vfslocked = VFS_LOCK_GIANT(vp->v_mount);
1851                                 vn_lock(vp, LK_SHARED | LK_RETRY);
1852                                 if (VOP_GETATTR(vp, &va, cred) == 0) {
1853                                         kve->kve_fileid = va.va_fileid;
1854                                         kve->kve_fsid = va.va_fsid;
1855                                 }
1856                                 vput(vp);
1857                                 VFS_UNLOCK_GIANT(vfslocked);
1858                         }
1859                 } else {
1860                         kve->kve_type = KVME_TYPE_NONE;
1861                         kve->kve_ref_count = 0;
1862                         kve->kve_shadow_count = 0;
1863                 }
1864
1865                 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
1866                 if (freepath != NULL)
1867                         free(freepath, M_TEMP);
1868
1869                 /* Pack record size down */
1870                 kve->kve_structsize = offsetof(struct kinfo_vmentry, kve_path) +
1871                     strlen(kve->kve_path) + 1;
1872                 kve->kve_structsize = roundup(kve->kve_structsize,
1873                     sizeof(uint64_t));
1874                 error = SYSCTL_OUT(req, kve, kve->kve_structsize);
1875                 vm_map_lock_read(map);
1876                 if (error)
1877                         break;
1878                 if (last_timestamp != map->timestamp) {
1879                         vm_map_lookup_entry(map, addr - 1, &tmp_entry);
1880                         entry = tmp_entry;
1881                 }
1882         }
1883         vm_map_unlock_read(map);
1884         vmspace_free(vm);
1885         PRELE(p);
1886         free(kve, M_TEMP);
1887         return (error);
1888 }
1889
1890 #if defined(STACK) || defined(DDB)
1891 static int
1892 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
1893 {
1894         struct kinfo_kstack *kkstp;
1895         int error, i, *name, numthreads;
1896         lwpid_t *lwpidarray;
1897         struct thread *td;
1898         struct stack *st;
1899         struct sbuf sb;
1900         struct proc *p;
1901
1902         name = (int *)arg1;
1903         error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
1904         if (error != 0)
1905                 return (error);
1906
1907         kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
1908         st = stack_create();
1909
1910         lwpidarray = NULL;
1911         numthreads = 0;
1912         PROC_LOCK(p);
1913 repeat:
1914         if (numthreads < p->p_numthreads) {
1915                 if (lwpidarray != NULL) {
1916                         free(lwpidarray, M_TEMP);
1917                         lwpidarray = NULL;
1918                 }
1919                 numthreads = p->p_numthreads;
1920                 PROC_UNLOCK(p);
1921                 lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
1922                     M_WAITOK | M_ZERO);
1923                 PROC_LOCK(p);
1924                 goto repeat;
1925         }
1926         i = 0;
1927
1928         /*
1929          * XXXRW: During the below loop, execve(2) and countless other sorts
1930          * of changes could have taken place.  Should we check to see if the
1931          * vmspace has been replaced, or the like, in order to prevent
1932          * giving a snapshot that spans, say, execve(2), with some threads
1933          * before and some after?  Among other things, the credentials could
1934          * have changed, in which case the right to extract debug info might
1935          * no longer be assured.
1936          */
1937         FOREACH_THREAD_IN_PROC(p, td) {
1938                 KASSERT(i < numthreads,
1939                     ("sysctl_kern_proc_kstack: numthreads"));
1940                 lwpidarray[i] = td->td_tid;
1941                 i++;
1942         }
1943         numthreads = i;
1944         for (i = 0; i < numthreads; i++) {
1945                 td = thread_find(p, lwpidarray[i]);
1946                 if (td == NULL) {
1947                         continue;
1948                 }
1949                 bzero(kkstp, sizeof(*kkstp));
1950                 (void)sbuf_new(&sb, kkstp->kkst_trace,
1951                     sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
1952                 thread_lock(td);
1953                 kkstp->kkst_tid = td->td_tid;
1954                 if (TD_IS_SWAPPED(td))
1955                         kkstp->kkst_state = KKST_STATE_SWAPPED;
1956                 else if (TD_IS_RUNNING(td))
1957                         kkstp->kkst_state = KKST_STATE_RUNNING;
1958                 else {
1959                         kkstp->kkst_state = KKST_STATE_STACKOK;
1960                         stack_save_td(st, td);
1961                 }
1962                 thread_unlock(td);
1963                 PROC_UNLOCK(p);
1964                 stack_sbuf_print(&sb, st);
1965                 sbuf_finish(&sb);
1966                 sbuf_delete(&sb);
1967                 error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
1968                 PROC_LOCK(p);
1969                 if (error)
1970                         break;
1971         }
1972         _PRELE(p);
1973         PROC_UNLOCK(p);
1974         if (lwpidarray != NULL)
1975                 free(lwpidarray, M_TEMP);
1976         stack_destroy(st);
1977         free(kkstp, M_TEMP);
1978         return (error);
1979 }
1980 #endif
1981
1982 /*
1983  * This sysctl allows a process to retrieve the full list of groups from
1984  * itself or another process.
1985  */
1986 static int
1987 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
1988 {
1989         pid_t *pidp = (pid_t *)arg1;
1990         unsigned int arglen = arg2;
1991         struct proc *p;
1992         struct ucred *cred;
1993         int error;
1994
1995         if (arglen != 1)
1996                 return (EINVAL);
1997         if (*pidp == -1) {      /* -1 means this process */
1998                 p = req->td->td_proc;
1999         } else {
2000                 error = pget(*pidp, PGET_CANSEE, &p);
2001                 if (error != 0)
2002                         return (error);
2003         }
2004
2005         cred = crhold(p->p_ucred);
2006         if (*pidp != -1)
2007                 PROC_UNLOCK(p);
2008
2009         error = SYSCTL_OUT(req, cred->cr_groups,
2010             cred->cr_ngroups * sizeof(gid_t));
2011         crfree(cred);
2012         return (error);
2013 }
2014
2015 /*
2016  * This sysctl allows a process to set and retrieve binary osreldate of
2017  * another process.
2018  */
2019 static int
2020 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
2021 {
2022         int *name = (int *)arg1;
2023         u_int namelen = arg2;
2024         struct proc *p;
2025         int flags, error, osrel;
2026
2027         if (namelen != 1)
2028                 return (EINVAL);
2029
2030         if (req->newptr != NULL && req->newlen != sizeof(osrel))
2031                 return (EINVAL);
2032
2033         flags = PGET_HOLD | PGET_NOTWEXIT;
2034         if (req->newptr != NULL)
2035                 flags |= PGET_CANDEBUG;
2036         else
2037                 flags |= PGET_CANSEE;
2038         error = pget((pid_t)name[0], flags, &p);
2039         if (error != 0)
2040                 return (error);
2041
2042         error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
2043         if (error != 0)
2044                 goto errout;
2045
2046         if (req->newptr != NULL) {
2047                 error = SYSCTL_IN(req, &osrel, sizeof(osrel));
2048                 if (error != 0)
2049                         goto errout;
2050                 if (osrel < 0) {
2051                         error = EINVAL;
2052                         goto errout;
2053                 }
2054                 p->p_osrel = osrel;
2055         }
2056 errout:
2057         PRELE(p);
2058         return (error);
2059 }
2060
2061 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD,  0, "Process table");
2062
2063 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
2064         CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
2065         "Return entire process table");
2066
2067 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2068         sysctl_kern_proc, "Process table");
2069
2070 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
2071         sysctl_kern_proc, "Process table");
2072
2073 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2074         sysctl_kern_proc, "Process table");
2075
2076 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
2077         CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2078
2079 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
2080         sysctl_kern_proc, "Process table");
2081
2082 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2083         sysctl_kern_proc, "Process table");
2084
2085 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2086         sysctl_kern_proc, "Process table");
2087
2088 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2089         sysctl_kern_proc, "Process table");
2090
2091 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
2092         sysctl_kern_proc, "Return process table, no threads");
2093
2094 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
2095         CTLFLAG_RW | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
2096         sysctl_kern_proc_args, "Process argument list");
2097
2098 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
2099         CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
2100
2101 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
2102         CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
2103         "Process syscall vector name (ABI type)");
2104
2105 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
2106         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2107
2108 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
2109         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2110
2111 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
2112         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2113
2114 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
2115         sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2116
2117 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
2118         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2119
2120 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
2121         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2122
2123 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
2124         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2125
2126 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
2127         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2128
2129 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
2130         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
2131         "Return process table, no threads");
2132
2133 #ifdef COMPAT_FREEBSD7
2134 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
2135         CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
2136 #endif
2137
2138 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
2139         CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
2140
2141 #if defined(STACK) || defined(DDB)
2142 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
2143         CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
2144 #endif
2145
2146 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
2147         CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
2148
2149 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
2150         CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
2151         "Process binary osreldate");