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_ktrace.h"
  38 #include "opt_kstack_pages.h"
  39 #include "opt_stack.h"
  40
  41 #include <sys/param.h>
  42 #include <sys/systm.h>
  43 #include <sys/elf.h>
  44 #include <sys/eventhandler.h>
  45 #include <sys/exec.h>
  46 #include <sys/jail.h>
  47 #include <sys/kernel.h>
  48 #include <sys/limits.h>
  49 #include <sys/lock.h>
  50 #include <sys/loginclass.h>
  51 #include <sys/malloc.h>
  52 #include <sys/mman.h>
  53 #include <sys/mount.h>
  54 #include <sys/mutex.h>
  55 #include <sys/proc.h>
  56 #include <sys/ptrace.h>
  57 #include <sys/refcount.h>
  58 #include <sys/resourcevar.h>
  59 #include <sys/rwlock.h>
  60 #include <sys/sbuf.h>
  61 #include <sys/sysent.h>
  62 #include <sys/sched.h>
  63 #include <sys/smp.h>
  64 #include <sys/stack.h>
  65 #include <sys/stat.h>
  66 #include <sys/sysctl.h>
  67 #include <sys/filedesc.h>
  68 #include <sys/tty.h>
  69 #include <sys/signalvar.h>
  70 #include <sys/sdt.h>
  71 #include <sys/sx.h>
  72 #include <sys/user.h>
  73 #include <sys/vnode.h>
  74 #include <sys/wait.h>
  75
  76 #ifdef DDB
  77 #include <ddb/ddb.h>
  78 #endif
  79
  80 #include <vm/vm.h>
  81 #include <vm/vm_param.h>
  82 #include <vm/vm_extern.h>
  83 #include <vm/pmap.h>
  84 #include <vm/vm_map.h>
  85 #include <vm/vm_object.h>
  86 #include <vm/vm_page.h>
  87 #include <vm/uma.h>
  88
  89 #ifdef COMPAT_FREEBSD32
  90 #include <compat/freebsd32/freebsd32.h>
  91 #include <compat/freebsd32/freebsd32_util.h>
  92 #endif
  93
  94 SDT_PROVIDER_DEFINE(proc);
  95 SDT_PROBE_DEFINE4(proc, , ctor, entry, "struct proc *", "int", "void *",
  96     "int");
  97 SDT_PROBE_DEFINE4(proc, , ctor, return, "struct proc *", "int", "void *",
  98     "int");
  99 SDT_PROBE_DEFINE4(proc, , dtor, entry, "struct proc *", "int", "void *",
 100     "struct thread *");
 101 SDT_PROBE_DEFINE3(proc, , dtor, return, "struct proc *", "int", "void *");
 102 SDT_PROBE_DEFINE3(proc, , init, entry, "struct proc *", "int", "int");
 103 SDT_PROBE_DEFINE3(proc, , init, return, "struct proc *", "int", "int");
 104
 105 MALLOC_DEFINE(M_PGRP, "pgrp", "process group header");
 106 MALLOC_DEFINE(M_SESSION, "session", "session header");
 107 static MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
 108 MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
 109
 110 static void doenterpgrp(struct proc *, struct pgrp *);
 111 static void orphanpg(struct pgrp *pg);
 112 static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp);
 113 static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp);
 114 static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp,
 115     int preferthread);
 116 static void pgadjustjobc(struct pgrp *pgrp, int entering);
 117 static void pgdelete(struct pgrp *);
 118 static int proc_ctor(void *mem, int size, void *arg, int flags);
 119 static void proc_dtor(void *mem, int size, void *arg);
 120 static int proc_init(void *mem, int size, int flags);
 121 static void proc_fini(void *mem, int size);
 122 static void pargs_free(struct pargs *pa);
 123 static struct proc *zpfind_locked(pid_t pid);
 124
 125 /*
 126  * Other process lists
 127  */
 128 struct pidhashhead *pidhashtbl;
 129 u_long pidhash;
 130 struct pgrphashhead *pgrphashtbl;
 131 u_long pgrphash;
 132 struct proclist allproc;
 133 struct proclist zombproc;
 134 struct sx __exclusive_cache_line allproc_lock;
 135 struct sx __exclusive_cache_line proctree_lock;
 136 struct mtx __exclusive_cache_line ppeers_lock;
 137 uma_zone_t proc_zone;
 138
 139 /*
 140  * The offset of various fields in struct proc and struct thread.
 141  * These are used by kernel debuggers to enumerate kernel threads and
 142  * processes.
 143  */
 144 const int proc_off_p_pid = offsetof(struct proc, p_pid);
 145 const int proc_off_p_comm = offsetof(struct proc, p_comm);
 146 const int proc_off_p_list = offsetof(struct proc, p_list);
 147 const int proc_off_p_threads = offsetof(struct proc, p_threads);
 148 const int thread_off_td_tid = offsetof(struct thread, td_tid);
 149 const int thread_off_td_name = offsetof(struct thread, td_name);
 150 const int thread_off_td_oncpu = offsetof(struct thread, td_oncpu);
 151 const int thread_off_td_pcb = offsetof(struct thread, td_pcb);
 152 const int thread_off_td_plist = offsetof(struct thread, td_plist);
 153
 154 EVENTHANDLER_LIST_DEFINE(process_ctor);
 155 EVENTHANDLER_LIST_DEFINE(process_dtor);
 156 EVENTHANDLER_LIST_DEFINE(process_init);
 157 EVENTHANDLER_LIST_DEFINE(process_fini);
 158 EVENTHANDLER_LIST_DEFINE(process_exit);
 159 EVENTHANDLER_LIST_DEFINE(process_fork);
 160 EVENTHANDLER_LIST_DEFINE(process_exec);
 161
 162 EVENTHANDLER_LIST_DECLARE(thread_ctor);
 163 EVENTHANDLER_LIST_DECLARE(thread_dtor);
 164
 165 int kstack_pages = KSTACK_PAGES;
 166 SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0,
 167     "Kernel stack size in pages");
 168 static int vmmap_skip_res_cnt = 0;
 169 SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW,
 170     &vmmap_skip_res_cnt, 0,
 171     "Skip calculation of the pages resident count in kern.proc.vmmap");
 172
 173 CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
 174 #ifdef COMPAT_FREEBSD32
 175 CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
 176 #endif
 177
 178 /*
 179  * Initialize global process hashing structures.
 180  */
 181 void
 182 procinit(void)
 183 {
 184
 185         sx_init(&allproc_lock, "allproc");
 186         sx_init(&proctree_lock, "proctree");
 187         mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
 188         LIST_INIT(&allproc);
 189         LIST_INIT(&zombproc);
 190         pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
 191         pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
 192         proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
 193             proc_ctor, proc_dtor, proc_init, proc_fini,
 194             UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
 195         uihashinit();
 196 }
 197
 198 /*
 199  * Prepare a proc for use.
 200  */
 201 static int
 202 proc_ctor(void *mem, int size, void *arg, int flags)
 203 {
 204         struct proc *p;
 205         struct thread *td;
 206
 207         p = (struct proc *)mem;
 208         SDT_PROBE4(proc, , ctor , entry, p, size, arg, flags);
 209         EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
 210         SDT_PROBE4(proc, , ctor , return, p, size, arg, flags);
 211         td = FIRST_THREAD_IN_PROC(p);
 212         if (td != NULL) {
 213                 /* Make sure all thread constructors are executed */
 214                 EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
 215         }
 216         return (0);
 217 }
 218
 219 /*
 220  * Reclaim a proc after use.
 221  */
 222 static void
 223 proc_dtor(void *mem, int size, void *arg)
 224 {
 225         struct proc *p;
 226         struct thread *td;
 227
 228         /* INVARIANTS checks go here */
 229         p = (struct proc *)mem;
 230         td = FIRST_THREAD_IN_PROC(p);
 231         SDT_PROBE4(proc, , dtor, entry, p, size, arg, td);
 232         if (td != NULL) {
 233 #ifdef INVARIANTS
 234                 KASSERT((p->p_numthreads == 1),
 235                     ("bad number of threads in exiting process"));
 236                 KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
 237 #endif
 238                 /* Free all OSD associated to this thread. */
 239                 osd_thread_exit(td);
 240                 td_softdep_cleanup(td);
 241                 MPASS(td->td_su == NULL);
 242
 243                 /* Make sure all thread destructors are executed */
 244                 EVENTHANDLER_DIRECT_INVOKE(thread_dtor, td);
 245         }
 246         EVENTHANDLER_DIRECT_INVOKE(process_dtor, p);
 247         if (p->p_ksi != NULL)
 248                 KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
 249         SDT_PROBE3(proc, , dtor, return, p, size, arg);
 250 }
 251
 252 /*
 253  * Initialize type-stable parts of a proc (when newly created).
 254  */
 255 static int
 256 proc_init(void *mem, int size, int flags)
 257 {
 258         struct proc *p;
 259
 260         p = (struct proc *)mem;
 261         SDT_PROBE3(proc, , init, entry, p, size, flags);
 262         mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK | MTX_NEW);
 263         mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_NEW);
 264         mtx_init(&p->p_statmtx, "pstatl", NULL, MTX_SPIN | MTX_NEW);
 265         mtx_init(&p->p_itimmtx, "pitiml", NULL, MTX_SPIN | MTX_NEW);
 266         mtx_init(&p->p_profmtx, "pprofl", NULL, MTX_SPIN | MTX_NEW);
 267         cv_init(&p->p_pwait, "ppwait");
 268         TAILQ_INIT(&p->p_threads);           /* all threads in proc */
 269         EVENTHANDLER_DIRECT_INVOKE(process_init, p);
 270         p->p_stats = pstats_alloc();
 271         p->p_pgrp = NULL;
 272         SDT_PROBE3(proc, , init, return, p, size, flags);
 273         return (0);
 274 }
 275
 276 /*
 277  * UMA should ensure that this function is never called.
 278  * Freeing a proc structure would violate type stability.
 279  */
 280 static void
 281 proc_fini(void *mem, int size)
 282 {
 283 #ifdef notnow
 284         struct proc *p;
 285
 286         p = (struct proc *)mem;
 287         EVENTHANDLER_DIRECT_INVOKE(process_fini, p);
 288         pstats_free(p->p_stats);
 289         thread_free(FIRST_THREAD_IN_PROC(p));
 290         mtx_destroy(&p->p_mtx);
 291         if (p->p_ksi != NULL)
 292                 ksiginfo_free(p->p_ksi);
 293 #else
 294         panic("proc reclaimed");
 295 #endif
 296 }
 297
 298 /*
 299  * Is p an inferior of the current process?
 300  */
 301 int
 302 inferior(struct proc *p)
 303 {
 304
 305         sx_assert(&proctree_lock, SX_LOCKED);
 306         PROC_LOCK_ASSERT(p, MA_OWNED);
 307         for (; p != curproc; p = proc_realparent(p)) {
 308                 if (p->p_pid == 0)
 309                         return (0);
 310         }
 311         return (1);
 312 }
 313
 314 struct proc *
 315 pfind_locked(pid_t pid)
 316 {
 317         struct proc *p;
 318
 319         sx_assert(&allproc_lock, SX_LOCKED);
 320         LIST_FOREACH(p, PIDHASH(pid), p_hash) {
 321                 if (p->p_pid == pid) {
 322                         PROC_LOCK(p);
 323                         if (p->p_state == PRS_NEW) {
 324                                 PROC_UNLOCK(p);
 325                                 p = NULL;
 326                         }
 327                         break;
 328                 }
 329         }
 330         return (p);
 331 }
 332
 333 /*
 334  * Locate a process by number; return only "live" processes -- i.e., neither
 335  * zombies nor newly born but incompletely initialized processes.  By not
 336  * returning processes in the PRS_NEW state, we allow callers to avoid
 337  * testing for that condition to avoid dereferencing p_ucred, et al.
 338  */
 339 struct proc *
 340 pfind(pid_t pid)
 341 {
 342         struct proc *p;
 343
 344         sx_slock(&allproc_lock);
 345         p = pfind_locked(pid);
 346         sx_sunlock(&allproc_lock);
 347         return (p);
 348 }
 349
 350 static struct proc *
 351 pfind_tid_locked(pid_t tid)
 352 {
 353         struct proc *p;
 354         struct thread *td;
 355
 356         sx_assert(&allproc_lock, SX_LOCKED);
 357         FOREACH_PROC_IN_SYSTEM(p) {
 358                 PROC_LOCK(p);
 359                 if (p->p_state == PRS_NEW) {
 360                         PROC_UNLOCK(p);
 361                         continue;
 362                 }
 363                 FOREACH_THREAD_IN_PROC(p, td) {
 364                         if (td->td_tid == tid)
 365                                 goto found;
 366                 }
 367                 PROC_UNLOCK(p);
 368         }
 369 found:
 370         return (p);
 371 }
 372
 373 /*
 374  * Locate a process group by number.
 375  * The caller must hold proctree_lock.
 376  */
 377 struct pgrp *
 378 pgfind(pgid)
 379         register pid_t pgid;
 380 {
 381         register struct pgrp *pgrp;
 382
 383         sx_assert(&proctree_lock, SX_LOCKED);
 384
 385         LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
 386                 if (pgrp->pg_id == pgid) {
 387                         PGRP_LOCK(pgrp);
 388                         return (pgrp);
 389                 }
 390         }
 391         return (NULL);
 392 }
 393
 394 /*
 395  * Locate process and do additional manipulations, depending on flags.
 396  */
 397 int
 398 pget(pid_t pid, int flags, struct proc **pp)
 399 {
 400         struct proc *p;
 401         int error;
 402
 403         sx_slock(&allproc_lock);
 404         if (pid <= PID_MAX) {
 405                 p = pfind_locked(pid);
 406                 if (p == NULL && (flags & PGET_NOTWEXIT) == 0)
 407                         p = zpfind_locked(pid);
 408         } else if ((flags & PGET_NOTID) == 0) {
 409                 p = pfind_tid_locked(pid);
 410         } else {
 411                 p = NULL;
 412         }
 413         sx_sunlock(&allproc_lock);
 414         if (p == NULL)
 415                 return (ESRCH);
 416         if ((flags & PGET_CANSEE) != 0) {
 417                 error = p_cansee(curthread, p);
 418                 if (error != 0)
 419                         goto errout;
 420         }
 421         if ((flags & PGET_CANDEBUG) != 0) {
 422                 error = p_candebug(curthread, p);
 423                 if (error != 0)
 424                         goto errout;
 425         }
 426         if ((flags & PGET_ISCURRENT) != 0 && curproc != p) {
 427                 error = EPERM;
 428                 goto errout;
 429         }
 430         if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) {
 431                 error = ESRCH;
 432                 goto errout;
 433         }
 434         if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) {
 435                 /*
 436                  * XXXRW: Not clear ESRCH is the right error during proc
 437                  * execve().
 438                  */
 439                 error = ESRCH;
 440                 goto errout;
 441         }
 442         if ((flags & PGET_HOLD) != 0) {
 443                 _PHOLD(p);
 444                 PROC_UNLOCK(p);
 445         }
 446         *pp = p;
 447         return (0);
 448 errout:
 449         PROC_UNLOCK(p);
 450         return (error);
 451 }
 452
 453 /*
 454  * Create a new process group.
 455  * pgid must be equal to the pid of p.
 456  * Begin a new session if required.
 457  */
 458 int
 459 enterpgrp(p, pgid, pgrp, sess)
 460         register struct proc *p;
 461         pid_t pgid;
 462         struct pgrp *pgrp;
 463         struct session *sess;
 464 {
 465
 466         sx_assert(&proctree_lock, SX_XLOCKED);
 467
 468         KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
 469         KASSERT(p->p_pid == pgid,
 470             ("enterpgrp: new pgrp and pid != pgid"));
 471         KASSERT(pgfind(pgid) == NULL,
 472             ("enterpgrp: pgrp with pgid exists"));
 473         KASSERT(!SESS_LEADER(p),
 474             ("enterpgrp: session leader attempted setpgrp"));
 475
 476         mtx_init(&pgrp->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
 477
 478         if (sess != NULL) {
 479                 /*
 480                  * new session
 481                  */
 482                 mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
 483                 PROC_LOCK(p);
 484                 p->p_flag &= ~P_CONTROLT;
 485                 PROC_UNLOCK(p);
 486                 PGRP_LOCK(pgrp);
 487                 sess->s_leader = p;
 488                 sess->s_sid = p->p_pid;
 489                 refcount_init(&sess->s_count, 1);
 490                 sess->s_ttyvp = NULL;
 491                 sess->s_ttydp = NULL;
 492                 sess->s_ttyp = NULL;
 493                 bcopy(p->p_session->s_login, sess->s_login,
 494                             sizeof(sess->s_login));
 495                 pgrp->pg_session = sess;
 496                 KASSERT(p == curproc,
 497                     ("enterpgrp: mksession and p != curproc"));
 498         } else {
 499                 pgrp->pg_session = p->p_session;
 500                 sess_hold(pgrp->pg_session);
 501                 PGRP_LOCK(pgrp);
 502         }
 503         pgrp->pg_id = pgid;
 504         LIST_INIT(&pgrp->pg_members);
 505
 506         /*
 507          * As we have an exclusive lock of proctree_lock,
 508          * this should not deadlock.
 509          */
 510         LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
 511         pgrp->pg_jobc = 0;
 512         SLIST_INIT(&pgrp->pg_sigiolst);
 513         PGRP_UNLOCK(pgrp);
 514
 515         doenterpgrp(p, pgrp);
 516
 517         return (0);
 518 }
 519
 520 /*
 521  * Move p to an existing process group
 522  */
 523 int
 524 enterthispgrp(p, pgrp)
 525         register struct proc *p;
 526         struct pgrp *pgrp;
 527 {
 528
 529         sx_assert(&proctree_lock, SX_XLOCKED);
 530         PROC_LOCK_ASSERT(p, MA_NOTOWNED);
 531         PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
 532         PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
 533         SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
 534         KASSERT(pgrp->pg_session == p->p_session,
 535                 ("%s: pgrp's session %p, p->p_session %p.\n",
 536                 __func__,
 537                 pgrp->pg_session,
 538                 p->p_session));
 539         KASSERT(pgrp != p->p_pgrp,
 540                 ("%s: p belongs to pgrp.", __func__));
 541
 542         doenterpgrp(p, pgrp);
 543
 544         return (0);
 545 }
 546
 547 /*
 548  * Move p to a process group
 549  */
 550 static void
 551 doenterpgrp(p, pgrp)
 552         struct proc *p;
 553         struct pgrp *pgrp;
 554 {
 555         struct pgrp *savepgrp;
 556
 557         sx_assert(&proctree_lock, SX_XLOCKED);
 558         PROC_LOCK_ASSERT(p, MA_NOTOWNED);
 559         PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
 560         PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
 561         SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
 562
 563         savepgrp = p->p_pgrp;
 564
 565         /*
 566          * Adjust eligibility of affected pgrps to participate in job control.
 567          * Increment eligibility counts before decrementing, otherwise we
 568          * could reach 0 spuriously during the first call.
 569          */
 570         fixjobc(p, pgrp, 1);
 571         fixjobc(p, p->p_pgrp, 0);
 572
 573         PGRP_LOCK(pgrp);
 574         PGRP_LOCK(savepgrp);
 575         PROC_LOCK(p);
 576         LIST_REMOVE(p, p_pglist);
 577         p->p_pgrp = pgrp;
 578         PROC_UNLOCK(p);
 579         LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
 580         PGRP_UNLOCK(savepgrp);
 581         PGRP_UNLOCK(pgrp);
 582         if (LIST_EMPTY(&savepgrp->pg_members))
 583                 pgdelete(savepgrp);
 584 }
 585
 586 /*
 587  * remove process from process group
 588  */
 589 int
 590 leavepgrp(p)
 591         register struct proc *p;
 592 {
 593         struct pgrp *savepgrp;
 594
 595         sx_assert(&proctree_lock, SX_XLOCKED);
 596         savepgrp = p->p_pgrp;
 597         PGRP_LOCK(savepgrp);
 598         PROC_LOCK(p);
 599         LIST_REMOVE(p, p_pglist);
 600         p->p_pgrp = NULL;
 601         PROC_UNLOCK(p);
 602         PGRP_UNLOCK(savepgrp);
 603         if (LIST_EMPTY(&savepgrp->pg_members))
 604                 pgdelete(savepgrp);
 605         return (0);
 606 }
 607
 608 /*
 609  * delete a process group
 610  */
 611 static void
 612 pgdelete(pgrp)
 613         register struct pgrp *pgrp;
 614 {
 615         struct session *savesess;
 616         struct tty *tp;
 617
 618         sx_assert(&proctree_lock, SX_XLOCKED);
 619         PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
 620         SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
 621
 622         /*
 623          * Reset any sigio structures pointing to us as a result of
 624          * F_SETOWN with our pgid.
 625          */
 626         funsetownlst(&pgrp->pg_sigiolst);
 627
 628         PGRP_LOCK(pgrp);
 629         tp = pgrp->pg_session->s_ttyp;
 630         LIST_REMOVE(pgrp, pg_hash);
 631         savesess = pgrp->pg_session;
 632         PGRP_UNLOCK(pgrp);
 633
 634         /* Remove the reference to the pgrp before deallocating it. */
 635         if (tp != NULL) {
 636                 tty_lock(tp);
 637                 tty_rel_pgrp(tp, pgrp);
 638         }
 639
 640         mtx_destroy(&pgrp->pg_mtx);
 641         free(pgrp, M_PGRP);
 642         sess_release(savesess);
 643 }
 644
 645 static void
 646 pgadjustjobc(pgrp, entering)
 647         struct pgrp *pgrp;
 648         int entering;
 649 {
 650
 651         PGRP_LOCK(pgrp);
 652         if (entering)
 653                 pgrp->pg_jobc++;
 654         else {
 655                 --pgrp->pg_jobc;
 656                 if (pgrp->pg_jobc == 0)
 657                         orphanpg(pgrp);
 658         }
 659         PGRP_UNLOCK(pgrp);
 660 }
 661
 662 /*
 663  * Adjust pgrp jobc counters when specified process changes process group.
 664  * We count the number of processes in each process group that "qualify"
 665  * the group for terminal job control (those with a parent in a different
 666  * process group of the same session).  If that count reaches zero, the
 667  * process group becomes orphaned.  Check both the specified process'
 668  * process group and that of its children.
 669  * entering == 0 => p is leaving specified group.
 670  * entering == 1 => p is entering specified group.
 671  */
 672 void
 673 fixjobc(struct proc *p, struct pgrp *pgrp, int entering)
 674 {
 675         struct pgrp *hispgrp;
 676         struct session *mysession;
 677         struct proc *q;
 678
 679         sx_assert(&proctree_lock, SX_LOCKED);
 680         PROC_LOCK_ASSERT(p, MA_NOTOWNED);
 681         PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
 682         SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
 683
 684         /*
 685          * Check p's parent to see whether p qualifies its own process
 686          * group; if so, adjust count for p's process group.
 687          */
 688         mysession = pgrp->pg_session;
 689         if ((hispgrp = p->p_pptr->p_pgrp) != pgrp &&
 690             hispgrp->pg_session == mysession)
 691                 pgadjustjobc(pgrp, entering);
 692
 693         /*
 694          * Check this process' children to see whether they qualify
 695          * their process groups; if so, adjust counts for children's
 696          * process groups.
 697          */
 698         LIST_FOREACH(q, &p->p_children, p_sibling) {
 699                 hispgrp = q->p_pgrp;
 700                 if (hispgrp == pgrp ||
 701                     hispgrp->pg_session != mysession)
 702                         continue;
 703                 if (q->p_state == PRS_ZOMBIE)
 704                         continue;
 705                 pgadjustjobc(hispgrp, entering);
 706         }
 707 }
 708
 709 void
 710 killjobc(void)
 711 {
 712         struct session *sp;
 713         struct tty *tp;
 714         struct proc *p;
 715         struct vnode *ttyvp;
 716
 717         p = curproc;
 718         MPASS(p->p_flag & P_WEXIT);
 719         /*
 720          * Do a quick check to see if there is anything to do with the
 721          * proctree_lock held. pgrp and LIST_EMPTY checks are for fixjobc().
 722          */
 723         PROC_LOCK(p);
 724         if (!SESS_LEADER(p) &&
 725             (p->p_pgrp == p->p_pptr->p_pgrp) &&
 726             LIST_EMPTY(&p->p_children)) {
 727                 PROC_UNLOCK(p);
 728                 return;
 729         }
 730         PROC_UNLOCK(p);
 731
 732         sx_xlock(&proctree_lock);
 733         if (SESS_LEADER(p)) {
 734                 sp = p->p_session;
 735
 736                 /*
 737                  * s_ttyp is not zero'd; we use this to indicate that
 738                  * the session once had a controlling terminal. (for
 739                  * logging and informational purposes)
 740                  */
 741                 SESS_LOCK(sp);
 742                 ttyvp = sp->s_ttyvp;
 743                 tp = sp->s_ttyp;
 744                 sp->s_ttyvp = NULL;
 745                 sp->s_ttydp = NULL;
 746                 sp->s_leader = NULL;
 747                 SESS_UNLOCK(sp);
 748
 749                 /*
 750                  * Signal foreground pgrp and revoke access to
 751                  * controlling terminal if it has not been revoked
 752                  * already.
 753                  *
 754                  * Because the TTY may have been revoked in the mean
 755                  * time and could already have a new session associated
 756                  * with it, make sure we don't send a SIGHUP to a
 757                  * foreground process group that does not belong to this
 758                  * session.
 759                  */
 760
 761                 if (tp != NULL) {
 762                         tty_lock(tp);
 763                         if (tp->t_session == sp)
 764                                 tty_signal_pgrp(tp, SIGHUP);
 765                         tty_unlock(tp);
 766                 }
 767
 768                 if (ttyvp != NULL) {
 769                         sx_xunlock(&proctree_lock);
 770                         if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) {
 771                                 VOP_REVOKE(ttyvp, REVOKEALL);
 772                                 VOP_UNLOCK(ttyvp, 0);
 773                         }
 774                         vrele(ttyvp);
 775                         sx_xlock(&proctree_lock);
 776                 }
 777         }
 778         fixjobc(p, p->p_pgrp, 0);
 779         sx_xunlock(&proctree_lock);
 780 }
 781
 782 /*
 783  * A process group has become orphaned;
 784  * if there are any stopped processes in the group,
 785  * hang-up all process in that group.
 786  */
 787 static void
 788 orphanpg(pg)
 789         struct pgrp *pg;
 790 {
 791         register struct proc *p;
 792
 793         PGRP_LOCK_ASSERT(pg, MA_OWNED);
 794
 795         LIST_FOREACH(p, &pg->pg_members, p_pglist) {
 796                 PROC_LOCK(p);
 797                 if (P_SHOULDSTOP(p) == P_STOPPED_SIG) {
 798                         PROC_UNLOCK(p);
 799                         LIST_FOREACH(p, &pg->pg_members, p_pglist) {
 800                                 PROC_LOCK(p);
 801                                 kern_psignal(p, SIGHUP);
 802                                 kern_psignal(p, SIGCONT);
 803                                 PROC_UNLOCK(p);
 804                         }
 805                         return;
 806                 }
 807                 PROC_UNLOCK(p);
 808         }
 809 }
 810
 811 void
 812 sess_hold(struct session *s)
 813 {
 814
 815         refcount_acquire(&s->s_count);
 816 }
 817
 818 void
 819 sess_release(struct session *s)
 820 {
 821
 822         if (refcount_release(&s->s_count)) {
 823                 if (s->s_ttyp != NULL) {
 824                         tty_lock(s->s_ttyp);
 825                         tty_rel_sess(s->s_ttyp, s);
 826                 }
 827                 mtx_destroy(&s->s_mtx);
 828                 free(s, M_SESSION);
 829         }
 830 }
 831
 832 #ifdef DDB
 833
 834 DB_SHOW_COMMAND(pgrpdump, pgrpdump)
 835 {
 836         register struct pgrp *pgrp;
 837         register struct proc *p;
 838         register int i;
 839
 840         for (i = 0; i <= pgrphash; i++) {
 841                 if (!LIST_EMPTY(&pgrphashtbl[i])) {
 842                         printf("\tindx %d\n", i);
 843                         LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
 844                                 printf(
 845                         "\tpgrp %p, pgid %ld, sess %p, sesscnt %d, mem %p\n",
 846                                     (void *)pgrp, (long)pgrp->pg_id,
 847                                     (void *)pgrp->pg_session,
 848                                     pgrp->pg_session->s_count,
 849                                     (void *)LIST_FIRST(&pgrp->pg_members));
 850                                 LIST_FOREACH(p, &pgrp->pg_members, p_pglist) {
 851                                         printf("\t\tpid %ld addr %p pgrp %p\n",
 852                                             (long)p->p_pid, (void *)p,
 853                                             (void *)p->p_pgrp);
 854                                 }
 855                         }
 856                 }
 857         }
 858 }
 859 #endif /* DDB */
 860
 861 /*
 862  * Calculate the kinfo_proc members which contain process-wide
 863  * informations.
 864  * Must be called with the target process locked.
 865  */
 866 static void
 867 fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
 868 {
 869         struct thread *td;
 870
 871         PROC_LOCK_ASSERT(p, MA_OWNED);
 872
 873         kp->ki_estcpu = 0;
 874         kp->ki_pctcpu = 0;
 875         FOREACH_THREAD_IN_PROC(p, td) {
 876                 thread_lock(td);
 877                 kp->ki_pctcpu += sched_pctcpu(td);
 878                 kp->ki_estcpu += sched_estcpu(td);
 879                 thread_unlock(td);
 880         }
 881 }
 882
 883 /*
 884  * Clear kinfo_proc and fill in any information that is common
 885  * to all threads in the process.
 886  * Must be called with the target process locked.
 887  */
 888 static void
 889 fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
 890 {
 891         struct thread *td0;
 892         struct tty *tp;
 893         struct session *sp;
 894         struct ucred *cred;
 895         struct sigacts *ps;
 896         struct timeval boottime;
 897
 898         /* For proc_realparent. */
 899         sx_assert(&proctree_lock, SX_LOCKED);
 900         PROC_LOCK_ASSERT(p, MA_OWNED);
 901         bzero(kp, sizeof(*kp));
 902
 903         kp->ki_structsize = sizeof(*kp);
 904         kp->ki_paddr = p;
 905         kp->ki_addr =/* p->p_addr; */0; /* XXX */
 906         kp->ki_args = p->p_args;
 907         kp->ki_textvp = p->p_textvp;
 908 #ifdef KTRACE
 909         kp->ki_tracep = p->p_tracevp;
 910         kp->ki_traceflag = p->p_traceflag;
 911 #endif
 912         kp->ki_fd = p->p_fd;
 913         kp->ki_vmspace = p->p_vmspace;
 914         kp->ki_flag = p->p_flag;
 915         kp->ki_flag2 = p->p_flag2;
 916         cred = p->p_ucred;
 917         if (cred) {
 918                 kp->ki_uid = cred->cr_uid;
 919                 kp->ki_ruid = cred->cr_ruid;
 920                 kp->ki_svuid = cred->cr_svuid;
 921                 kp->ki_cr_flags = 0;
 922                 if (cred->cr_flags & CRED_FLAG_CAPMODE)
 923                         kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
 924                 /* XXX bde doesn't like KI_NGROUPS */
 925                 if (cred->cr_ngroups > KI_NGROUPS) {
 926                         kp->ki_ngroups = KI_NGROUPS;
 927                         kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
 928                 } else
 929                         kp->ki_ngroups = cred->cr_ngroups;
 930                 bcopy(cred->cr_groups, kp->ki_groups,
 931                     kp->ki_ngroups * sizeof(gid_t));
 932                 kp->ki_rgid = cred->cr_rgid;
 933                 kp->ki_svgid = cred->cr_svgid;
 934                 /* If jailed(cred), emulate the old P_JAILED flag. */
 935                 if (jailed(cred)) {
 936                         kp->ki_flag |= P_JAILED;
 937                         /* If inside the jail, use 0 as a jail ID. */
 938                         if (cred->cr_prison != curthread->td_ucred->cr_prison)
 939                                 kp->ki_jid = cred->cr_prison->pr_id;
 940                 }
 941                 strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
 942                     sizeof(kp->ki_loginclass));
 943         }
 944         ps = p->p_sigacts;
 945         if (ps) {
 946                 mtx_lock(&ps->ps_mtx);
 947                 kp->ki_sigignore = ps->ps_sigignore;
 948                 kp->ki_sigcatch = ps->ps_sigcatch;
 949                 mtx_unlock(&ps->ps_mtx);
 950         }
 951         if (p->p_state != PRS_NEW &&
 952             p->p_state != PRS_ZOMBIE &&
 953             p->p_vmspace != NULL) {
 954                 struct vmspace *vm = p->p_vmspace;
 955
 956                 kp->ki_size = vm->vm_map.size;
 957                 kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
 958                 FOREACH_THREAD_IN_PROC(p, td0) {
 959                         if (!TD_IS_SWAPPED(td0))
 960                                 kp->ki_rssize += td0->td_kstack_pages;
 961                 }
 962                 kp->ki_swrss = vm->vm_swrss;
 963                 kp->ki_tsize = vm->vm_tsize;
 964                 kp->ki_dsize = vm->vm_dsize;
 965                 kp->ki_ssize = vm->vm_ssize;
 966         } else if (p->p_state == PRS_ZOMBIE)
 967                 kp->ki_stat = SZOMB;
 968         if (kp->ki_flag & P_INMEM)
 969                 kp->ki_sflag = PS_INMEM;
 970         else
 971                 kp->ki_sflag = 0;
 972         /* Calculate legacy swtime as seconds since 'swtick'. */
 973         kp->ki_swtime = (ticks - p->p_swtick) / hz;
 974         kp->ki_pid = p->p_pid;
 975         kp->ki_nice = p->p_nice;
 976         kp->ki_fibnum = p->p_fibnum;
 977         kp->ki_start = p->p_stats->p_start;
 978         getboottime(&boottime);
 979         timevaladd(&kp->ki_start, &boottime);
 980         PROC_STATLOCK(p);
 981         rufetch(p, &kp->ki_rusage);
 982         kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
 983         calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
 984         PROC_STATUNLOCK(p);
 985         calccru(p, &kp->ki_childutime, &kp->ki_childstime);
 986         /* Some callers want child times in a single value. */
 987         kp->ki_childtime = kp->ki_childstime;
 988         timevaladd(&kp->ki_childtime, &kp->ki_childutime);
 989
 990         FOREACH_THREAD_IN_PROC(p, td0)
 991                 kp->ki_cow += td0->td_cow;
 992
 993         tp = NULL;
 994         if (p->p_pgrp) {
 995                 kp->ki_pgid = p->p_pgrp->pg_id;
 996                 kp->ki_jobc = p->p_pgrp->pg_jobc;
 997                 sp = p->p_pgrp->pg_session;
 998
 999                 if (sp != NULL) {
1000                         kp->ki_sid = sp->s_sid;
1001                         SESS_LOCK(sp);
1002                         strlcpy(kp->ki_login, sp->s_login,
1003                             sizeof(kp->ki_login));
1004                         if (sp->s_ttyvp)
1005                                 kp->ki_kiflag |= KI_CTTY;
1006                         if (SESS_LEADER(p))
1007                                 kp->ki_kiflag |= KI_SLEADER;
1008                         /* XXX proctree_lock */
1009                         tp = sp->s_ttyp;
1010                         SESS_UNLOCK(sp);
1011                 }
1012         }
1013         if ((p->p_flag & P_CONTROLT) && tp != NULL) {
1014                 kp->ki_tdev = tty_udev(tp);
1015                 kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
1016                 if (tp->t_session)
1017                         kp->ki_tsid = tp->t_session->s_sid;
1018         } else
1019                 kp->ki_tdev = NODEV;
1020         if (p->p_comm[0] != '\0')
1021                 strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
1022         if (p->p_sysent && p->p_sysent->sv_name != NULL &&
1023             p->p_sysent->sv_name[0] != '\0')
1024                 strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
1025         kp->ki_siglist = p->p_siglist;
1026         kp->ki_xstat = KW_EXITCODE(p->p_xexit, p->p_xsig);
1027         kp->ki_acflag = p->p_acflag;
1028         kp->ki_lock = p->p_lock;
1029         if (p->p_pptr) {
1030                 kp->ki_ppid = proc_realparent(p)->p_pid;
1031                 if (p->p_flag & P_TRACED)
1032                         kp->ki_tracer = p->p_pptr->p_pid;
1033         }
1034 }
1035
1036 /*
1037  * Fill in information that is thread specific.  Must be called with
1038  * target process locked.  If 'preferthread' is set, overwrite certain
1039  * process-related fields that are maintained for both threads and
1040  * processes.
1041  */
1042 static void
1043 fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
1044 {
1045         struct proc *p;
1046
1047         p = td->td_proc;
1048         kp->ki_tdaddr = td;
1049         PROC_LOCK_ASSERT(p, MA_OWNED);
1050
1051         if (preferthread)
1052                 PROC_STATLOCK(p);
1053         thread_lock(td);
1054         if (td->td_wmesg != NULL)
1055                 strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
1056         else
1057                 bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
1058         if (strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname)) >=
1059             sizeof(kp->ki_tdname)) {
1060                 strlcpy(kp->ki_moretdname,
1061                     td->td_name + sizeof(kp->ki_tdname) - 1,
1062                     sizeof(kp->ki_moretdname));
1063         } else {
1064                 bzero(kp->ki_moretdname, sizeof(kp->ki_moretdname));
1065         }
1066         if (TD_ON_LOCK(td)) {
1067                 kp->ki_kiflag |= KI_LOCKBLOCK;
1068                 strlcpy(kp->ki_lockname, td->td_lockname,
1069                     sizeof(kp->ki_lockname));
1070         } else {
1071                 kp->ki_kiflag &= ~KI_LOCKBLOCK;
1072                 bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
1073         }
1074
1075         if (p->p_state == PRS_NORMAL) { /* approximate. */
1076                 if (TD_ON_RUNQ(td) ||
1077                     TD_CAN_RUN(td) ||
1078                     TD_IS_RUNNING(td)) {
1079                         kp->ki_stat = SRUN;
1080                 } else if (P_SHOULDSTOP(p)) {
1081                         kp->ki_stat = SSTOP;
1082                 } else if (TD_IS_SLEEPING(td)) {
1083                         kp->ki_stat = SSLEEP;
1084                 } else if (TD_ON_LOCK(td)) {
1085                         kp->ki_stat = SLOCK;
1086                 } else {
1087                         kp->ki_stat = SWAIT;
1088                 }
1089         } else if (p->p_state == PRS_ZOMBIE) {
1090                 kp->ki_stat = SZOMB;
1091         } else {
1092                 kp->ki_stat = SIDL;
1093         }
1094
1095         /* Things in the thread */
1096         kp->ki_wchan = td->td_wchan;
1097         kp->ki_pri.pri_level = td->td_priority;
1098         kp->ki_pri.pri_native = td->td_base_pri;
1099
1100         /*
1101          * Note: legacy fields; clamp at the old NOCPU value and/or
1102          * the maximum u_char CPU value.
1103          */
1104         if (td->td_lastcpu == NOCPU)
1105                 kp->ki_lastcpu_old = NOCPU_OLD;
1106         else if (td->td_lastcpu > MAXCPU_OLD)
1107                 kp->ki_lastcpu_old = MAXCPU_OLD;
1108         else
1109                 kp->ki_lastcpu_old = td->td_lastcpu;
1110
1111         if (td->td_oncpu == NOCPU)
1112                 kp->ki_oncpu_old = NOCPU_OLD;
1113         else if (td->td_oncpu > MAXCPU_OLD)
1114                 kp->ki_oncpu_old = MAXCPU_OLD;
1115         else
1116                 kp->ki_oncpu_old = td->td_oncpu;
1117
1118         kp->ki_lastcpu = td->td_lastcpu;
1119         kp->ki_oncpu = td->td_oncpu;
1120         kp->ki_tdflags = td->td_flags;
1121         kp->ki_tid = td->td_tid;
1122         kp->ki_numthreads = p->p_numthreads;
1123         kp->ki_pcb = td->td_pcb;
1124         kp->ki_kstack = (void *)td->td_kstack;
1125         kp->ki_slptime = (ticks - td->td_slptick) / hz;
1126         kp->ki_pri.pri_class = td->td_pri_class;
1127         kp->ki_pri.pri_user = td->td_user_pri;
1128
1129         if (preferthread) {
1130                 rufetchtd(td, &kp->ki_rusage);
1131                 kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
1132                 kp->ki_pctcpu = sched_pctcpu(td);
1133                 kp->ki_estcpu = sched_estcpu(td);
1134                 kp->ki_cow = td->td_cow;
1135         }
1136
1137         /* We can't get this anymore but ps etc never used it anyway. */
1138         kp->ki_rqindex = 0;
1139
1140         if (preferthread)
1141                 kp->ki_siglist = td->td_siglist;
1142         kp->ki_sigmask = td->td_sigmask;
1143         thread_unlock(td);
1144         if (preferthread)
1145                 PROC_STATUNLOCK(p);
1146 }
1147
1148 /*
1149  * Fill in a kinfo_proc structure for the specified process.
1150  * Must be called with the target process locked.
1151  */
1152 void
1153 fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
1154 {
1155
1156         MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1157
1158         fill_kinfo_proc_only(p, kp);
1159         fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
1160         fill_kinfo_aggregate(p, kp);
1161 }
1162
1163 struct pstats *
1164 pstats_alloc(void)
1165 {
1166
1167         return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
1168 }
1169
1170 /*
1171  * Copy parts of p_stats; zero the rest of p_stats (statistics).
1172  */
1173 void
1174 pstats_fork(struct pstats *src, struct pstats *dst)
1175 {
1176
1177         bzero(&dst->pstat_startzero,
1178             __rangeof(struct pstats, pstat_startzero, pstat_endzero));
1179         bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
1180             __rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
1181 }
1182
1183 void
1184 pstats_free(struct pstats *ps)
1185 {
1186
1187         free(ps, M_SUBPROC);
1188 }
1189
1190 static struct proc *
1191 zpfind_locked(pid_t pid)
1192 {
1193         struct proc *p;
1194
1195         sx_assert(&allproc_lock, SX_LOCKED);
1196         LIST_FOREACH(p, &zombproc, p_list) {
1197                 if (p->p_pid == pid) {
1198                         PROC_LOCK(p);
1199                         break;
1200                 }
1201         }
1202         return (p);
1203 }
1204
1205 /*
1206  * Locate a zombie process by number
1207  */
1208 struct proc *
1209 zpfind(pid_t pid)
1210 {
1211         struct proc *p;
1212
1213         sx_slock(&allproc_lock);
1214         p = zpfind_locked(pid);
1215         sx_sunlock(&allproc_lock);
1216         return (p);
1217 }
1218
1219 #ifdef COMPAT_FREEBSD32
1220
1221 /*
1222  * This function is typically used to copy out the kernel address, so
1223  * it can be replaced by assignment of zero.
1224  */
1225 static inline uint32_t
1226 ptr32_trim(void *ptr)
1227 {
1228         uintptr_t uptr;
1229
1230         uptr = (uintptr_t)ptr;
1231         return ((uptr > UINT_MAX) ? 0 : uptr);
1232 }
1233
1234 #define PTRTRIM_CP(src,dst,fld) \
1235         do { (dst).fld = ptr32_trim((src).fld); } while (0)
1236
1237 static void
1238 freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
1239 {
1240         int i;
1241
1242         bzero(ki32, sizeof(struct kinfo_proc32));
1243         ki32->ki_structsize = sizeof(struct kinfo_proc32);
1244         CP(*ki, *ki32, ki_layout);
1245         PTRTRIM_CP(*ki, *ki32, ki_args);
1246         PTRTRIM_CP(*ki, *ki32, ki_paddr);
1247         PTRTRIM_CP(*ki, *ki32, ki_addr);
1248         PTRTRIM_CP(*ki, *ki32, ki_tracep);
1249         PTRTRIM_CP(*ki, *ki32, ki_textvp);
1250         PTRTRIM_CP(*ki, *ki32, ki_fd);
1251         PTRTRIM_CP(*ki, *ki32, ki_vmspace);
1252         PTRTRIM_CP(*ki, *ki32, ki_wchan);
1253         CP(*ki, *ki32, ki_pid);
1254         CP(*ki, *ki32, ki_ppid);
1255         CP(*ki, *ki32, ki_pgid);
1256         CP(*ki, *ki32, ki_tpgid);
1257         CP(*ki, *ki32, ki_sid);
1258         CP(*ki, *ki32, ki_tsid);
1259         CP(*ki, *ki32, ki_jobc);
1260         CP(*ki, *ki32, ki_tdev);
1261         CP(*ki, *ki32, ki_siglist);
1262         CP(*ki, *ki32, ki_sigmask);
1263         CP(*ki, *ki32, ki_sigignore);
1264         CP(*ki, *ki32, ki_sigcatch);
1265         CP(*ki, *ki32, ki_uid);
1266         CP(*ki, *ki32, ki_ruid);
1267         CP(*ki, *ki32, ki_svuid);
1268         CP(*ki, *ki32, ki_rgid);
1269         CP(*ki, *ki32, ki_svgid);
1270         CP(*ki, *ki32, ki_ngroups);
1271         for (i = 0; i < KI_NGROUPS; i++)
1272                 CP(*ki, *ki32, ki_groups[i]);
1273         CP(*ki, *ki32, ki_size);
1274         CP(*ki, *ki32, ki_rssize);
1275         CP(*ki, *ki32, ki_swrss);
1276         CP(*ki, *ki32, ki_tsize);
1277         CP(*ki, *ki32, ki_dsize);
1278         CP(*ki, *ki32, ki_ssize);
1279         CP(*ki, *ki32, ki_xstat);
1280         CP(*ki, *ki32, ki_acflag);
1281         CP(*ki, *ki32, ki_pctcpu);
1282         CP(*ki, *ki32, ki_estcpu);
1283         CP(*ki, *ki32, ki_slptime);
1284         CP(*ki, *ki32, ki_swtime);
1285         CP(*ki, *ki32, ki_cow);
1286         CP(*ki, *ki32, ki_runtime);
1287         TV_CP(*ki, *ki32, ki_start);
1288         TV_CP(*ki, *ki32, ki_childtime);
1289         CP(*ki, *ki32, ki_flag);
1290         CP(*ki, *ki32, ki_kiflag);
1291         CP(*ki, *ki32, ki_traceflag);
1292         CP(*ki, *ki32, ki_stat);
1293         CP(*ki, *ki32, ki_nice);
1294         CP(*ki, *ki32, ki_lock);
1295         CP(*ki, *ki32, ki_rqindex);
1296         CP(*ki, *ki32, ki_oncpu);
1297         CP(*ki, *ki32, ki_lastcpu);
1298
1299         /* XXX TODO: wrap cpu value as appropriate */
1300         CP(*ki, *ki32, ki_oncpu_old);
1301         CP(*ki, *ki32, ki_lastcpu_old);
1302
1303         bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
1304         bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
1305         bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
1306         bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
1307         bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
1308         bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
1309         bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
1310         bcopy(ki->ki_moretdname, ki32->ki_moretdname, MAXCOMLEN - TDNAMLEN + 1);
1311         CP(*ki, *ki32, ki_tracer);
1312         CP(*ki, *ki32, ki_flag2);
1313         CP(*ki, *ki32, ki_fibnum);
1314         CP(*ki, *ki32, ki_cr_flags);
1315         CP(*ki, *ki32, ki_jid);
1316         CP(*ki, *ki32, ki_numthreads);
1317         CP(*ki, *ki32, ki_tid);
1318         CP(*ki, *ki32, ki_pri);
1319         freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
1320         freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
1321         PTRTRIM_CP(*ki, *ki32, ki_pcb);
1322         PTRTRIM_CP(*ki, *ki32, ki_kstack);
1323         PTRTRIM_CP(*ki, *ki32, ki_udata);
1324         PTRTRIM_CP(*ki, *ki32, ki_tdaddr);
1325         CP(*ki, *ki32, ki_sflag);
1326         CP(*ki, *ki32, ki_tdflags);
1327 }
1328 #endif
1329
1330 int
1331 kern_proc_out(struct proc *p, struct sbuf *sb, int flags)
1332 {
1333         struct thread *td;
1334         struct kinfo_proc ki;
1335 #ifdef COMPAT_FREEBSD32
1336         struct kinfo_proc32 ki32;
1337 #endif
1338         int error;
1339
1340         PROC_LOCK_ASSERT(p, MA_OWNED);
1341         MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
1342
1343         error = 0;
1344         fill_kinfo_proc(p, &ki);
1345         if ((flags & KERN_PROC_NOTHREADS) != 0) {
1346 #ifdef COMPAT_FREEBSD32
1347                 if ((flags & KERN_PROC_MASK32) != 0) {
1348                         freebsd32_kinfo_proc_out(&ki, &ki32);
1349                         if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1350                                 error = ENOMEM;
1351                 } else
1352 #endif
1353                         if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1354                                 error = ENOMEM;
1355         } else {
1356                 FOREACH_THREAD_IN_PROC(p, td) {
1357                         fill_kinfo_thread(td, &ki, 1);
1358 #ifdef COMPAT_FREEBSD32
1359                         if ((flags & KERN_PROC_MASK32) != 0) {
1360                                 freebsd32_kinfo_proc_out(&ki, &ki32);
1361                                 if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
1362                                         error = ENOMEM;
1363                         } else
1364 #endif
1365                                 if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
1366                                         error = ENOMEM;
1367                         if (error != 0)
1368                                 break;
1369                 }
1370         }
1371         PROC_UNLOCK(p);
1372         return (error);
1373 }
1374
1375 static int
1376 sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags,
1377     int doingzomb)
1378 {
1379         struct sbuf sb;
1380         struct kinfo_proc ki;
1381         struct proc *np;
1382         int error, error2;
1383         pid_t pid;
1384
1385         pid = p->p_pid;
1386         sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req);
1387         sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1388         error = kern_proc_out(p, &sb, flags);
1389         error2 = sbuf_finish(&sb);
1390         sbuf_delete(&sb);
1391         if (error != 0)
1392                 return (error);
1393         else if (error2 != 0)
1394                 return (error2);
1395         if (doingzomb)
1396                 np = zpfind(pid);
1397         else {
1398                 if (pid == 0)
1399                         return (0);
1400                 np = pfind(pid);
1401         }
1402         if (np == NULL)
1403                 return (ESRCH);
1404         if (np != p) {
1405                 PROC_UNLOCK(np);
1406                 return (ESRCH);
1407         }
1408         PROC_UNLOCK(np);
1409         return (0);
1410 }
1411
1412 static int
1413 sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
1414 {
1415         int *name = (int *)arg1;
1416         u_int namelen = arg2;
1417         struct proc *p;
1418         int flags, doingzomb, oid_number;
1419         int error = 0;
1420
1421         oid_number = oidp->oid_number;
1422         if (oid_number != KERN_PROC_ALL &&
1423             (oid_number & KERN_PROC_INC_THREAD) == 0)
1424                 flags = KERN_PROC_NOTHREADS;
1425         else {
1426                 flags = 0;
1427                 oid_number &= ~KERN_PROC_INC_THREAD;
1428         }
1429 #ifdef COMPAT_FREEBSD32
1430         if (req->flags & SCTL_MASK32)
1431                 flags |= KERN_PROC_MASK32;
1432 #endif
1433         if (oid_number == KERN_PROC_PID) {
1434                 if (namelen != 1)
1435                         return (EINVAL);
1436                 error = sysctl_wire_old_buffer(req, 0);
1437                 if (error)
1438                         return (error);
1439                 sx_slock(&proctree_lock);
1440                 error = pget((pid_t)name[0], PGET_CANSEE, &p);
1441                 if (error == 0)
1442                         error = sysctl_out_proc(p, req, flags, 0);
1443                 sx_sunlock(&proctree_lock);
1444                 return (error);
1445         }
1446
1447         switch (oid_number) {
1448         case KERN_PROC_ALL:
1449                 if (namelen != 0)
1450                         return (EINVAL);
1451                 break;
1452         case KERN_PROC_PROC:
1453                 if (namelen != 0 && namelen != 1)
1454                         return (EINVAL);
1455                 break;
1456         default:
1457                 if (namelen != 1)
1458                         return (EINVAL);
1459                 break;
1460         }
1461
1462         if (!req->oldptr) {
1463                 /* overestimate by 5 procs */
1464                 error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
1465                 if (error)
1466                         return (error);
1467         }
1468         error = sysctl_wire_old_buffer(req, 0);
1469         if (error != 0)
1470                 return (error);
1471         sx_slock(&proctree_lock);
1472         sx_slock(&allproc_lock);
1473         for (doingzomb=0 ; doingzomb < 2 ; doingzomb++) {
1474                 if (!doingzomb)
1475                         p = LIST_FIRST(&allproc);
1476                 else
1477                         p = LIST_FIRST(&zombproc);
1478                 for (; p != NULL; p = LIST_NEXT(p, p_list)) {
1479                         /*
1480                          * Skip embryonic processes.
1481                          */
1482                         PROC_LOCK(p);
1483                         if (p->p_state == PRS_NEW) {
1484                                 PROC_UNLOCK(p);
1485                                 continue;
1486                         }
1487                         KASSERT(p->p_ucred != NULL,
1488                             ("process credential is NULL for non-NEW proc"));
1489                         /*
1490                          * Show a user only appropriate processes.
1491                          */
1492                         if (p_cansee(curthread, p)) {
1493                                 PROC_UNLOCK(p);
1494                                 continue;
1495                         }
1496                         /*
1497                          * TODO - make more efficient (see notes below).
1498                          * do by session.
1499                          */
1500                         switch (oid_number) {
1501
1502                         case KERN_PROC_GID:
1503                                 if (p->p_ucred->cr_gid != (gid_t)name[0]) {
1504                                         PROC_UNLOCK(p);
1505                                         continue;
1506                                 }
1507                                 break;
1508
1509                         case KERN_PROC_PGRP:
1510                                 /* could do this by traversing pgrp */
1511                                 if (p->p_pgrp == NULL ||
1512                                     p->p_pgrp->pg_id != (pid_t)name[0]) {
1513                                         PROC_UNLOCK(p);
1514                                         continue;
1515                                 }
1516                                 break;
1517
1518                         case KERN_PROC_RGID:
1519                                 if (p->p_ucred->cr_rgid != (gid_t)name[0]) {
1520                                         PROC_UNLOCK(p);
1521                                         continue;
1522                                 }
1523                                 break;
1524
1525                         case KERN_PROC_SESSION:
1526                                 if (p->p_session == NULL ||
1527                                     p->p_session->s_sid != (pid_t)name[0]) {
1528                                         PROC_UNLOCK(p);
1529                                         continue;
1530                                 }
1531                                 break;
1532
1533                         case KERN_PROC_TTY:
1534                                 if ((p->p_flag & P_CONTROLT) == 0 ||
1535                                     p->p_session == NULL) {
1536                                         PROC_UNLOCK(p);
1537                                         continue;
1538                                 }
1539                                 /* XXX proctree_lock */
1540                                 SESS_LOCK(p->p_session);
1541                                 if (p->p_session->s_ttyp == NULL ||
1542                                     tty_udev(p->p_session->s_ttyp) !=
1543                                     (dev_t)name[0]) {
1544                                         SESS_UNLOCK(p->p_session);
1545                                         PROC_UNLOCK(p);
1546                                         continue;
1547                                 }
1548                                 SESS_UNLOCK(p->p_session);
1549                                 break;
1550
1551                         case KERN_PROC_UID:
1552                                 if (p->p_ucred->cr_uid != (uid_t)name[0]) {
1553                                         PROC_UNLOCK(p);
1554                                         continue;
1555                                 }
1556                                 break;
1557
1558                         case KERN_PROC_RUID:
1559                                 if (p->p_ucred->cr_ruid != (uid_t)name[0]) {
1560                                         PROC_UNLOCK(p);
1561                                         continue;
1562                                 }
1563                                 break;
1564
1565                         case KERN_PROC_PROC:
1566                                 break;
1567
1568                         default:
1569                                 break;
1570
1571                         }
1572
1573                         error = sysctl_out_proc(p, req, flags, doingzomb);
1574                         if (error) {
1575                                 sx_sunlock(&allproc_lock);
1576                                 sx_sunlock(&proctree_lock);
1577                                 return (error);
1578                         }
1579                 }
1580         }
1581         sx_sunlock(&allproc_lock);
1582         sx_sunlock(&proctree_lock);
1583         return (0);
1584 }
1585
1586 struct pargs *
1587 pargs_alloc(int len)
1588 {
1589         struct pargs *pa;
1590
1591         pa = malloc(sizeof(struct pargs) + len, M_PARGS,
1592                 M_WAITOK);
1593         refcount_init(&pa->ar_ref, 1);
1594         pa->ar_length = len;
1595         return (pa);
1596 }
1597
1598 static void
1599 pargs_free(struct pargs *pa)
1600 {
1601
1602         free(pa, M_PARGS);
1603 }
1604
1605 void
1606 pargs_hold(struct pargs *pa)
1607 {
1608
1609         if (pa == NULL)
1610                 return;
1611         refcount_acquire(&pa->ar_ref);
1612 }
1613
1614 void
1615 pargs_drop(struct pargs *pa)
1616 {
1617
1618         if (pa == NULL)
1619                 return;
1620         if (refcount_release(&pa->ar_ref))
1621                 pargs_free(pa);
1622 }
1623
1624 static int
1625 proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
1626     size_t len)
1627 {
1628         ssize_t n;
1629
1630         /*
1631          * This may return a short read if the string is shorter than the chunk
1632          * and is aligned at the end of the page, and the following page is not
1633          * mapped.
1634          */
1635         n = proc_readmem(td, p, (vm_offset_t)sptr, buf, len);
1636         if (n <= 0)
1637                 return (ENOMEM);
1638         return (0);
1639 }
1640
1641 #define PROC_AUXV_MAX   256     /* Safety limit on auxv size. */
1642
1643 enum proc_vector_type {
1644         PROC_ARG,
1645         PROC_ENV,
1646         PROC_AUX,
1647 };
1648
1649 #ifdef COMPAT_FREEBSD32
1650 static int
1651 get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
1652     size_t *vsizep, enum proc_vector_type type)
1653 {
1654         struct freebsd32_ps_strings pss;
1655         Elf32_Auxinfo aux;
1656         vm_offset_t vptr, ptr;
1657         uint32_t *proc_vector32;
1658         char **proc_vector;
1659         size_t vsize, size;
1660         int i, error;
1661
1662         error = 0;
1663         if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss,
1664             sizeof(pss)) != sizeof(pss))
1665                 return (ENOMEM);
1666         switch (type) {
1667         case PROC_ARG:
1668                 vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
1669                 vsize = pss.ps_nargvstr;
1670                 if (vsize > ARG_MAX)
1671                         return (ENOEXEC);
1672                 size = vsize * sizeof(int32_t);
1673                 break;
1674         case PROC_ENV:
1675                 vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
1676                 vsize = pss.ps_nenvstr;
1677                 if (vsize > ARG_MAX)
1678                         return (ENOEXEC);
1679                 size = vsize * sizeof(int32_t);
1680                 break;
1681         case PROC_AUX:
1682                 vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
1683                     (pss.ps_nenvstr + 1) * sizeof(int32_t);
1684                 if (vptr % 4 != 0)
1685                         return (ENOEXEC);
1686                 for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1687                         if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1688                             sizeof(aux))
1689                                 return (ENOMEM);
1690                         if (aux.a_type == AT_NULL)
1691                                 break;
1692                         ptr += sizeof(aux);
1693                 }
1694                 if (aux.a_type != AT_NULL)
1695                         return (ENOEXEC);
1696                 vsize = i + 1;
1697                 size = vsize * sizeof(aux);
1698                 break;
1699         default:
1700                 KASSERT(0, ("Wrong proc vector type: %d", type));
1701                 return (EINVAL);
1702         }
1703         proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
1704         if (proc_readmem(td, p, vptr, proc_vector32, size) != size) {
1705                 error = ENOMEM;
1706                 goto done;
1707         }
1708         if (type == PROC_AUX) {
1709                 *proc_vectorp = (char **)proc_vector32;
1710                 *vsizep = vsize;
1711                 return (0);
1712         }
1713         proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
1714         for (i = 0; i < (int)vsize; i++)
1715                 proc_vector[i] = PTRIN(proc_vector32[i]);
1716         *proc_vectorp = proc_vector;
1717         *vsizep = vsize;
1718 done:
1719         free(proc_vector32, M_TEMP);
1720         return (error);
1721 }
1722 #endif
1723
1724 static int
1725 get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
1726     size_t *vsizep, enum proc_vector_type type)
1727 {
1728         struct ps_strings pss;
1729         Elf_Auxinfo aux;
1730         vm_offset_t vptr, ptr;
1731         char **proc_vector;
1732         size_t vsize, size;
1733         int i;
1734
1735 #ifdef COMPAT_FREEBSD32
1736         if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1737                 return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
1738 #endif
1739         if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss,
1740             sizeof(pss)) != sizeof(pss))
1741                 return (ENOMEM);
1742         switch (type) {
1743         case PROC_ARG:
1744                 vptr = (vm_offset_t)pss.ps_argvstr;
1745                 vsize = pss.ps_nargvstr;
1746                 if (vsize > ARG_MAX)
1747                         return (ENOEXEC);
1748                 size = vsize * sizeof(char *);
1749                 break;
1750         case PROC_ENV:
1751                 vptr = (vm_offset_t)pss.ps_envstr;
1752                 vsize = pss.ps_nenvstr;
1753                 if (vsize > ARG_MAX)
1754                         return (ENOEXEC);
1755                 size = vsize * sizeof(char *);
1756                 break;
1757         case PROC_AUX:
1758                 /*
1759                  * The aux array is just above env array on the stack. Check
1760                  * that the address is naturally aligned.
1761                  */
1762                 vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
1763                     * sizeof(char *);
1764 #if __ELF_WORD_SIZE == 64
1765                 if (vptr % sizeof(uint64_t) != 0)
1766 #else
1767                 if (vptr % sizeof(uint32_t) != 0)
1768 #endif
1769                         return (ENOEXEC);
1770                 /*
1771                  * We count the array size reading the aux vectors from the
1772                  * stack until AT_NULL vector is returned.  So (to keep the code
1773                  * simple) we read the process stack twice: the first time here
1774                  * to find the size and the second time when copying the vectors
1775                  * to the allocated proc_vector.
1776                  */
1777                 for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
1778                         if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
1779                             sizeof(aux))
1780                                 return (ENOMEM);
1781                         if (aux.a_type == AT_NULL)
1782                                 break;
1783                         ptr += sizeof(aux);
1784                 }
1785                 /*
1786                  * If the PROC_AUXV_MAX entries are iterated over, and we have
1787                  * not reached AT_NULL, it is most likely we are reading wrong
1788                  * data: either the process doesn't have auxv array or data has
1789                  * been modified. Return the error in this case.
1790                  */
1791                 if (aux.a_type != AT_NULL)
1792                         return (ENOEXEC);
1793                 vsize = i + 1;
1794                 size = vsize * sizeof(aux);
1795                 break;
1796         default:
1797                 KASSERT(0, ("Wrong proc vector type: %d", type));
1798                 return (EINVAL); /* In case we are built without INVARIANTS. */
1799         }
1800         proc_vector = malloc(size, M_TEMP, M_WAITOK);
1801         if (proc_readmem(td, p, vptr, proc_vector, size) != size) {
1802                 free(proc_vector, M_TEMP);
1803                 return (ENOMEM);
1804         }
1805         *proc_vectorp = proc_vector;
1806         *vsizep = vsize;
1807
1808         return (0);
1809 }
1810
1811 #define GET_PS_STRINGS_CHUNK_SZ 256     /* Chunk size (bytes) for ps_strings operations. */
1812
1813 static int
1814 get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
1815     enum proc_vector_type type)
1816 {
1817         size_t done, len, nchr, vsize;
1818         int error, i;
1819         char **proc_vector, *sptr;
1820         char pss_string[GET_PS_STRINGS_CHUNK_SZ];
1821
1822         PROC_ASSERT_HELD(p);
1823
1824         /*
1825          * We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
1826          */
1827         nchr = 2 * (PATH_MAX + ARG_MAX);
1828
1829         error = get_proc_vector(td, p, &proc_vector, &vsize, type);
1830         if (error != 0)
1831                 return (error);
1832         for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
1833                 /*
1834                  * The program may have scribbled into its argv array, e.g. to
1835                  * remove some arguments.  If that has happened, break out
1836                  * before trying to read from NULL.
1837                  */
1838                 if (proc_vector[i] == NULL)
1839                         break;
1840                 for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
1841                         error = proc_read_string(td, p, sptr, pss_string,
1842                             sizeof(pss_string));
1843                         if (error != 0)
1844                                 goto done;
1845                         len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
1846                         if (done + len >= nchr)
1847                                 len = nchr - done - 1;
1848                         sbuf_bcat(sb, pss_string, len);
1849                         if (len != GET_PS_STRINGS_CHUNK_SZ)
1850                                 break;
1851                         done += GET_PS_STRINGS_CHUNK_SZ;
1852                 }
1853                 sbuf_bcat(sb, "", 1);
1854                 done += len + 1;
1855         }
1856 done:
1857         free(proc_vector, M_TEMP);
1858         return (error);
1859 }
1860
1861 int
1862 proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
1863 {
1864
1865         return (get_ps_strings(curthread, p, sb, PROC_ARG));
1866 }
1867
1868 int
1869 proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
1870 {
1871
1872         return (get_ps_strings(curthread, p, sb, PROC_ENV));
1873 }
1874
1875 int
1876 proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb)
1877 {
1878         size_t vsize, size;
1879         char **auxv;
1880         int error;
1881
1882         error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX);
1883         if (error == 0) {
1884 #ifdef COMPAT_FREEBSD32
1885                 if (SV_PROC_FLAG(p, SV_ILP32) != 0)
1886                         size = vsize * sizeof(Elf32_Auxinfo);
1887                 else
1888 #endif
1889                         size = vsize * sizeof(Elf_Auxinfo);
1890                 if (sbuf_bcat(sb, auxv, size) != 0)
1891                         error = ENOMEM;
1892                 free(auxv, M_TEMP);
1893         }
1894         return (error);
1895 }
1896
1897 /*
1898  * This sysctl allows a process to retrieve the argument list or process
1899  * title for another process without groping around in the address space
1900  * of the other process.  It also allow a process to set its own "process
1901  * title to a string of its own choice.
1902  */
1903 static int
1904 sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
1905 {
1906         int *name = (int *)arg1;
1907         u_int namelen = arg2;
1908         struct pargs *newpa, *pa;
1909         struct proc *p;
1910         struct sbuf sb;
1911         int flags, error = 0, error2;
1912
1913         if (namelen != 1)
1914                 return (EINVAL);
1915
1916         flags = PGET_CANSEE;
1917         if (req->newptr != NULL)
1918                 flags |= PGET_ISCURRENT;
1919         error = pget((pid_t)name[0], flags, &p);
1920         if (error)
1921                 return (error);
1922
1923         pa = p->p_args;
1924         if (pa != NULL) {
1925                 pargs_hold(pa);
1926                 PROC_UNLOCK(p);
1927                 error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
1928                 pargs_drop(pa);
1929         } else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
1930                 _PHOLD(p);
1931                 PROC_UNLOCK(p);
1932                 sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1933                 sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1934                 error = proc_getargv(curthread, p, &sb);
1935                 error2 = sbuf_finish(&sb);
1936                 PRELE(p);
1937                 sbuf_delete(&sb);
1938                 if (error == 0 && error2 != 0)
1939                         error = error2;
1940         } else {
1941                 PROC_UNLOCK(p);
1942         }
1943         if (error != 0 || req->newptr == NULL)
1944                 return (error);
1945
1946         if (req->newlen > ps_arg_cache_limit - sizeof(struct pargs))
1947                 return (ENOMEM);
1948
1949         if (req->newlen == 0) {
1950                 /*
1951                  * Clear the argument pointer, so that we'll fetch arguments
1952                  * with proc_getargv() until further notice.
1953                  */
1954                 newpa = NULL;
1955         } else {
1956                 newpa = pargs_alloc(req->newlen);
1957                 error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
1958                 if (error != 0) {
1959                         pargs_free(newpa);
1960                         return (error);
1961                 }
1962         }
1963         PROC_LOCK(p);
1964         pa = p->p_args;
1965         p->p_args = newpa;
1966         PROC_UNLOCK(p);
1967         pargs_drop(pa);
1968         return (0);
1969 }
1970
1971 /*
1972  * This sysctl allows a process to retrieve environment of another process.
1973  */
1974 static int
1975 sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
1976 {
1977         int *name = (int *)arg1;
1978         u_int namelen = arg2;
1979         struct proc *p;
1980         struct sbuf sb;
1981         int error, error2;
1982
1983         if (namelen != 1)
1984                 return (EINVAL);
1985
1986         error = pget((pid_t)name[0], PGET_WANTREAD, &p);
1987         if (error != 0)
1988                 return (error);
1989         if ((p->p_flag & P_SYSTEM) != 0) {
1990                 PRELE(p);
1991                 return (0);
1992         }
1993
1994         sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
1995         sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
1996         error = proc_getenvv(curthread, p, &sb);
1997         error2 = sbuf_finish(&sb);
1998         PRELE(p);
1999         sbuf_delete(&sb);
2000         return (error != 0 ? error : error2);
2001 }
2002
2003 /*
2004  * This sysctl allows a process to retrieve ELF auxiliary vector of
2005  * another process.
2006  */
2007 static int
2008 sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
2009 {
2010         int *name = (int *)arg1;
2011         u_int namelen = arg2;
2012         struct proc *p;
2013         struct sbuf sb;
2014         int error, error2;
2015
2016         if (namelen != 1)
2017                 return (EINVAL);
2018
2019         error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2020         if (error != 0)
2021                 return (error);
2022         if ((p->p_flag & P_SYSTEM) != 0) {
2023                 PRELE(p);
2024                 return (0);
2025         }
2026         sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
2027         sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2028         error = proc_getauxv(curthread, p, &sb);
2029         error2 = sbuf_finish(&sb);
2030         PRELE(p);
2031         sbuf_delete(&sb);
2032         return (error != 0 ? error : error2);
2033 }
2034
2035 /*
2036  * This sysctl allows a process to retrieve the path of the executable for
2037  * itself or another process.
2038  */
2039 static int
2040 sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
2041 {
2042         pid_t *pidp = (pid_t *)arg1;
2043         unsigned int arglen = arg2;
2044         struct proc *p;
2045         struct vnode *vp;
2046         char *retbuf, *freebuf;
2047         int error;
2048
2049         if (arglen != 1)
2050                 return (EINVAL);
2051         if (*pidp == -1) {      /* -1 means this process */
2052                 p = req->td->td_proc;
2053         } else {
2054                 error = pget(*pidp, PGET_CANSEE, &p);
2055                 if (error != 0)
2056                         return (error);
2057         }
2058
2059         vp = p->p_textvp;
2060         if (vp == NULL) {
2061                 if (*pidp != -1)
2062                         PROC_UNLOCK(p);
2063                 return (0);
2064         }
2065         vref(vp);
2066         if (*pidp != -1)
2067                 PROC_UNLOCK(p);
2068         error = vn_fullpath(req->td, vp, &retbuf, &freebuf);
2069         vrele(vp);
2070         if (error)
2071                 return (error);
2072         error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
2073         free(freebuf, M_TEMP);
2074         return (error);
2075 }
2076
2077 static int
2078 sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
2079 {
2080         struct proc *p;
2081         char *sv_name;
2082         int *name;
2083         int namelen;
2084         int error;
2085
2086         namelen = arg2;
2087         if (namelen != 1)
2088                 return (EINVAL);
2089
2090         name = (int *)arg1;
2091         error = pget((pid_t)name[0], PGET_CANSEE, &p);
2092         if (error != 0)
2093                 return (error);
2094         sv_name = p->p_sysent->sv_name;
2095         PROC_UNLOCK(p);
2096         return (sysctl_handle_string(oidp, sv_name, 0, req));
2097 }
2098
2099 #ifdef KINFO_OVMENTRY_SIZE
2100 CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
2101 #endif
2102
2103 #ifdef COMPAT_FREEBSD7
2104 static int
2105 sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
2106 {
2107         vm_map_entry_t entry, tmp_entry;
2108         unsigned int last_timestamp;
2109         char *fullpath, *freepath;
2110         struct kinfo_ovmentry *kve;
2111         struct vattr va;
2112         struct ucred *cred;
2113         int error, *name;
2114         struct vnode *vp;
2115         struct proc *p;
2116         vm_map_t map;
2117         struct vmspace *vm;
2118
2119         name = (int *)arg1;
2120         error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2121         if (error != 0)
2122                 return (error);
2123         vm = vmspace_acquire_ref(p);
2124         if (vm == NULL) {
2125                 PRELE(p);
2126                 return (ESRCH);
2127         }
2128         kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
2129
2130         map = &vm->vm_map;
2131         vm_map_lock_read(map);
2132         for (entry = map->header.next; entry != &map->header;
2133             entry = entry->next) {
2134                 vm_object_t obj, tobj, lobj;
2135                 vm_offset_t addr;
2136
2137                 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2138                         continue;
2139
2140                 bzero(kve, sizeof(*kve));
2141                 kve->kve_structsize = sizeof(*kve);
2142
2143                 kve->kve_private_resident = 0;
2144                 obj = entry->object.vm_object;
2145                 if (obj != NULL) {
2146                         VM_OBJECT_RLOCK(obj);
2147                         if (obj->shadow_count == 1)
2148                                 kve->kve_private_resident =
2149                                     obj->resident_page_count;
2150                 }
2151                 kve->kve_resident = 0;
2152                 addr = entry->start;
2153                 while (addr < entry->end) {
2154                         if (pmap_extract(map->pmap, addr))
2155                                 kve->kve_resident++;
2156                         addr += PAGE_SIZE;
2157                 }
2158
2159                 for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
2160                         if (tobj != obj) {
2161                                 VM_OBJECT_RLOCK(tobj);
2162                                 kve->kve_offset += tobj->backing_object_offset;
2163                         }
2164                         if (lobj != obj)
2165                                 VM_OBJECT_RUNLOCK(lobj);
2166                         lobj = tobj;
2167                 }
2168
2169                 kve->kve_start = (void*)entry->start;
2170                 kve->kve_end = (void*)entry->end;
2171                 kve->kve_offset += (off_t)entry->offset;
2172
2173                 if (entry->protection & VM_PROT_READ)
2174                         kve->kve_protection |= KVME_PROT_READ;
2175                 if (entry->protection & VM_PROT_WRITE)
2176                         kve->kve_protection |= KVME_PROT_WRITE;
2177                 if (entry->protection & VM_PROT_EXECUTE)
2178                         kve->kve_protection |= KVME_PROT_EXEC;
2179
2180                 if (entry->eflags & MAP_ENTRY_COW)
2181                         kve->kve_flags |= KVME_FLAG_COW;
2182                 if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2183                         kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2184                 if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2185                         kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2186
2187                 last_timestamp = map->timestamp;
2188                 vm_map_unlock_read(map);
2189
2190                 kve->kve_fileid = 0;
2191                 kve->kve_fsid = 0;
2192                 freepath = NULL;
2193                 fullpath = "";
2194                 if (lobj) {
2195                         vp = NULL;
2196                         switch (lobj->type) {
2197                         case OBJT_DEFAULT:
2198                                 kve->kve_type = KVME_TYPE_DEFAULT;
2199                                 break;
2200                         case OBJT_VNODE:
2201                                 kve->kve_type = KVME_TYPE_VNODE;
2202                                 vp = lobj->handle;
2203                                 vref(vp);
2204                                 break;
2205                         case OBJT_SWAP:
2206                                 if ((lobj->flags & OBJ_TMPFS_NODE) != 0) {
2207                                         kve->kve_type = KVME_TYPE_VNODE;
2208                                         if ((lobj->flags & OBJ_TMPFS) != 0) {
2209                                                 vp = lobj->un_pager.swp.swp_tmpfs;
2210                                                 vref(vp);
2211                                         }
2212                                 } else {
2213                                         kve->kve_type = KVME_TYPE_SWAP;
2214                                 }
2215                                 break;
2216                         case OBJT_DEVICE:
2217                                 kve->kve_type = KVME_TYPE_DEVICE;
2218                                 break;
2219                         case OBJT_PHYS:
2220                                 kve->kve_type = KVME_TYPE_PHYS;
2221                                 break;
2222                         case OBJT_DEAD:
2223                                 kve->kve_type = KVME_TYPE_DEAD;
2224                                 break;
2225                         case OBJT_SG:
2226                                 kve->kve_type = KVME_TYPE_SG;
2227                                 break;
2228                         default:
2229                                 kve->kve_type = KVME_TYPE_UNKNOWN;
2230                                 break;
2231                         }
2232                         if (lobj != obj)
2233                                 VM_OBJECT_RUNLOCK(lobj);
2234
2235                         kve->kve_ref_count = obj->ref_count;
2236                         kve->kve_shadow_count = obj->shadow_count;
2237                         VM_OBJECT_RUNLOCK(obj);
2238                         if (vp != NULL) {
2239                                 vn_fullpath(curthread, vp, &fullpath,
2240                                     &freepath);
2241                                 cred = curthread->td_ucred;
2242                                 vn_lock(vp, LK_SHARED | LK_RETRY);
2243                                 if (VOP_GETATTR(vp, &va, cred) == 0) {
2244                                         kve->kve_fileid = va.va_fileid;
2245                                         kve->kve_fsid = va.va_fsid;
2246                                 }
2247                                 vput(vp);
2248                         }
2249                 } else {
2250                         kve->kve_type = KVME_TYPE_NONE;
2251                         kve->kve_ref_count = 0;
2252                         kve->kve_shadow_count = 0;
2253                 }
2254
2255                 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2256                 if (freepath != NULL)
2257                         free(freepath, M_TEMP);
2258
2259                 error = SYSCTL_OUT(req, kve, sizeof(*kve));
2260                 vm_map_lock_read(map);
2261                 if (error)
2262                         break;
2263                 if (last_timestamp != map->timestamp) {
2264                         vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2265                         entry = tmp_entry;
2266                 }
2267         }
2268         vm_map_unlock_read(map);
2269         vmspace_free(vm);
2270         PRELE(p);
2271         free(kve, M_TEMP);
2272         return (error);
2273 }
2274 #endif  /* COMPAT_FREEBSD7 */
2275
2276 #ifdef KINFO_VMENTRY_SIZE
2277 CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
2278 #endif
2279
2280 void
2281 kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry,
2282     int *resident_count, bool *super)
2283 {
2284         vm_object_t obj, tobj;
2285         vm_page_t m, m_adv;
2286         vm_offset_t addr;
2287         vm_paddr_t locked_pa;
2288         vm_pindex_t pi, pi_adv, pindex;
2289
2290         *super = false;
2291         *resident_count = 0;
2292         if (vmmap_skip_res_cnt)
2293                 return;
2294
2295         locked_pa = 0;
2296         obj = entry->object.vm_object;
2297         addr = entry->start;
2298         m_adv = NULL;
2299         pi = OFF_TO_IDX(entry->offset);
2300         for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) {
2301                 if (m_adv != NULL) {
2302                         m = m_adv;
2303                 } else {
2304                         pi_adv = atop(entry->end - addr);
2305                         pindex = pi;
2306                         for (tobj = obj;; tobj = tobj->backing_object) {
2307                                 m = vm_page_find_least(tobj, pindex);
2308                                 if (m != NULL) {
2309                                         if (m->pindex == pindex)
2310                                                 break;
2311                                         if (pi_adv > m->pindex - pindex) {
2312                                                 pi_adv = m->pindex - pindex;
2313                                                 m_adv = m;
2314                                         }
2315                                 }
2316                                 if (tobj->backing_object == NULL)
2317                                         goto next;
2318                                 pindex += OFF_TO_IDX(tobj->
2319                                     backing_object_offset);
2320                         }
2321                 }
2322                 m_adv = NULL;
2323                 if (m->psind != 0 && addr + pagesizes[1] <= entry->end &&
2324                     (addr & (pagesizes[1] - 1)) == 0 &&
2325                     (pmap_mincore(map->pmap, addr, &locked_pa) &
2326                     MINCORE_SUPER) != 0) {
2327                         *super = true;
2328                         pi_adv = atop(pagesizes[1]);
2329                 } else {
2330                         /*
2331                          * We do not test the found page on validity.
2332                          * Either the page is busy and being paged in,
2333                          * or it was invalidated.  The first case
2334                          * should be counted as resident, the second
2335                          * is not so clear; we do account both.
2336                          */
2337                         pi_adv = 1;
2338                 }
2339                 *resident_count += pi_adv;
2340 next:;
2341         }
2342         PA_UNLOCK_COND(locked_pa);
2343 }
2344
2345 /*
2346  * Must be called with the process locked and will return unlocked.
2347  */
2348 int
2349 kern_proc_vmmap_out(struct proc *p, struct sbuf *sb, ssize_t maxlen, int flags)
2350 {
2351         vm_map_entry_t entry, tmp_entry;
2352         struct vattr va;
2353         vm_map_t map;
2354         vm_object_t obj, tobj, lobj;
2355         char *fullpath, *freepath;
2356         struct kinfo_vmentry *kve;
2357         struct ucred *cred;
2358         struct vnode *vp;
2359         struct vmspace *vm;
2360         vm_offset_t addr;
2361         unsigned int last_timestamp;
2362         int error;
2363         bool super;
2364
2365         PROC_LOCK_ASSERT(p, MA_OWNED);
2366
2367         _PHOLD(p);
2368         PROC_UNLOCK(p);
2369         vm = vmspace_acquire_ref(p);
2370         if (vm == NULL) {
2371                 PRELE(p);
2372                 return (ESRCH);
2373         }
2374         kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK | M_ZERO);
2375
2376         error = 0;
2377         map = &vm->vm_map;
2378         vm_map_lock_read(map);
2379         for (entry = map->header.next; entry != &map->header;
2380             entry = entry->next) {
2381                 if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
2382                         continue;
2383
2384                 addr = entry->end;
2385                 bzero(kve, sizeof(*kve));
2386                 obj = entry->object.vm_object;
2387                 if (obj != NULL) {
2388                         for (tobj = obj; tobj != NULL;
2389                             tobj = tobj->backing_object) {
2390                                 VM_OBJECT_RLOCK(tobj);
2391                                 kve->kve_offset += tobj->backing_object_offset;
2392                                 lobj = tobj;
2393                         }
2394                         if (obj->backing_object == NULL)
2395                                 kve->kve_private_resident =
2396                                     obj->resident_page_count;
2397                         kern_proc_vmmap_resident(map, entry,
2398                             &kve->kve_resident, &super);
2399                         if (super)
2400                                 kve->kve_flags |= KVME_FLAG_SUPER;
2401                         for (tobj = obj; tobj != NULL;
2402                             tobj = tobj->backing_object) {
2403                                 if (tobj != obj && tobj != lobj)
2404                                         VM_OBJECT_RUNLOCK(tobj);
2405                         }
2406                 } else {
2407                         lobj = NULL;
2408                 }
2409
2410                 kve->kve_start = entry->start;
2411                 kve->kve_end = entry->end;
2412                 kve->kve_offset += entry->offset;
2413
2414                 if (entry->protection & VM_PROT_READ)
2415                         kve->kve_protection |= KVME_PROT_READ;
2416                 if (entry->protection & VM_PROT_WRITE)
2417                         kve->kve_protection |= KVME_PROT_WRITE;
2418                 if (entry->protection & VM_PROT_EXECUTE)
2419                         kve->kve_protection |= KVME_PROT_EXEC;
2420
2421                 if (entry->eflags & MAP_ENTRY_COW)
2422                         kve->kve_flags |= KVME_FLAG_COW;
2423                 if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
2424                         kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
2425                 if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
2426                         kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
2427                 if (entry->eflags & MAP_ENTRY_GROWS_UP)
2428                         kve->kve_flags |= KVME_FLAG_GROWS_UP;
2429                 if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
2430                         kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
2431
2432                 last_timestamp = map->timestamp;
2433                 vm_map_unlock_read(map);
2434
2435                 freepath = NULL;
2436                 fullpath = "";
2437                 if (lobj != NULL) {
2438                         vp = NULL;
2439                         switch (lobj->type) {
2440                         case OBJT_DEFAULT:
2441                                 kve->kve_type = KVME_TYPE_DEFAULT;
2442                                 break;
2443                         case OBJT_VNODE:
2444                                 kve->kve_type = KVME_TYPE_VNODE;
2445                                 vp = lobj->handle;
2446                                 vref(vp);
2447                                 break;
2448                         case OBJT_SWAP:
2449                                 if ((lobj->flags & OBJ_TMPFS_NODE) != 0) {
2450                                         kve->kve_type = KVME_TYPE_VNODE;
2451                                         if ((lobj->flags & OBJ_TMPFS) != 0) {
2452                                                 vp = lobj->un_pager.swp.swp_tmpfs;
2453                                                 vref(vp);
2454                                         }
2455                                 } else {
2456                                         kve->kve_type = KVME_TYPE_SWAP;
2457                                 }
2458                                 break;
2459                         case OBJT_DEVICE:
2460                                 kve->kve_type = KVME_TYPE_DEVICE;
2461                                 break;
2462                         case OBJT_PHYS:
2463                                 kve->kve_type = KVME_TYPE_PHYS;
2464                                 break;
2465                         case OBJT_DEAD:
2466                                 kve->kve_type = KVME_TYPE_DEAD;
2467                                 break;
2468                         case OBJT_SG:
2469                                 kve->kve_type = KVME_TYPE_SG;
2470                                 break;
2471                         case OBJT_MGTDEVICE:
2472                                 kve->kve_type = KVME_TYPE_MGTDEVICE;
2473                                 break;
2474                         default:
2475                                 kve->kve_type = KVME_TYPE_UNKNOWN;
2476                                 break;
2477                         }
2478                         if (lobj != obj)
2479                                 VM_OBJECT_RUNLOCK(lobj);
2480
2481                         kve->kve_ref_count = obj->ref_count;
2482                         kve->kve_shadow_count = obj->shadow_count;
2483                         VM_OBJECT_RUNLOCK(obj);
2484                         if (vp != NULL) {
2485                                 vn_fullpath(curthread, vp, &fullpath,
2486                                     &freepath);
2487                                 kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
2488                                 cred = curthread->td_ucred;
2489                                 vn_lock(vp, LK_SHARED | LK_RETRY);
2490                                 if (VOP_GETATTR(vp, &va, cred) == 0) {
2491                                         kve->kve_vn_fileid = va.va_fileid;
2492                                         kve->kve_vn_fsid = va.va_fsid;
2493                                         kve->kve_vn_mode =
2494                                             MAKEIMODE(va.va_type, va.va_mode);
2495                                         kve->kve_vn_size = va.va_size;
2496                                         kve->kve_vn_rdev = va.va_rdev;
2497                                         kve->kve_status = KF_ATTR_VALID;
2498                                 }
2499                                 vput(vp);
2500                         }
2501                 } else {
2502                         kve->kve_type = KVME_TYPE_NONE;
2503                         kve->kve_ref_count = 0;
2504                         kve->kve_shadow_count = 0;
2505                 }
2506
2507                 strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
2508                 if (freepath != NULL)
2509                         free(freepath, M_TEMP);
2510
2511                 /* Pack record size down */
2512                 if ((flags & KERN_VMMAP_PACK_KINFO) != 0)
2513                         kve->kve_structsize =
2514                             offsetof(struct kinfo_vmentry, kve_path) +
2515                             strlen(kve->kve_path) + 1;
2516                 else
2517                         kve->kve_structsize = sizeof(*kve);
2518                 kve->kve_structsize = roundup(kve->kve_structsize,
2519                     sizeof(uint64_t));
2520
2521                 /* Halt filling and truncate rather than exceeding maxlen */
2522                 if (maxlen != -1 && maxlen < kve->kve_structsize) {
2523                         error = 0;
2524                         vm_map_lock_read(map);
2525                         break;
2526                 } else if (maxlen != -1)
2527                         maxlen -= kve->kve_structsize;
2528
2529                 if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0)
2530                         error = ENOMEM;
2531                 vm_map_lock_read(map);
2532                 if (error != 0)
2533                         break;
2534                 if (last_timestamp != map->timestamp) {
2535                         vm_map_lookup_entry(map, addr - 1, &tmp_entry);
2536                         entry = tmp_entry;
2537                 }
2538         }
2539         vm_map_unlock_read(map);
2540         vmspace_free(vm);
2541         PRELE(p);
2542         free(kve, M_TEMP);
2543         return (error);
2544 }
2545
2546 static int
2547 sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
2548 {
2549         struct proc *p;
2550         struct sbuf sb;
2551         int error, error2, *name;
2552
2553         name = (int *)arg1;
2554         sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req);
2555         sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
2556         error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
2557         if (error != 0) {
2558                 sbuf_delete(&sb);
2559                 return (error);
2560         }
2561         error = kern_proc_vmmap_out(p, &sb, -1, KERN_VMMAP_PACK_KINFO);
2562         error2 = sbuf_finish(&sb);
2563         sbuf_delete(&sb);
2564         return (error != 0 ? error : error2);
2565 }
2566
2567 #if defined(STACK) || defined(DDB)
2568 static int
2569 sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
2570 {
2571         struct kinfo_kstack *kkstp;
2572         int error, i, *name, numthreads;
2573         lwpid_t *lwpidarray;
2574         struct thread *td;
2575         struct stack *st;
2576         struct sbuf sb;
2577         struct proc *p;
2578
2579         name = (int *)arg1;
2580         error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
2581         if (error != 0)
2582                 return (error);
2583
2584         kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
2585         st = stack_create();
2586
2587         lwpidarray = NULL;
2588         PROC_LOCK(p);
2589         do {
2590                 if (lwpidarray != NULL) {
2591                         free(lwpidarray, M_TEMP);
2592                         lwpidarray = NULL;
2593                 }
2594                 numthreads = p->p_numthreads;
2595                 PROC_UNLOCK(p);
2596                 lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
2597                     M_WAITOK | M_ZERO);
2598                 PROC_LOCK(p);
2599         } while (numthreads < p->p_numthreads);
2600
2601         /*
2602          * XXXRW: During the below loop, execve(2) and countless other sorts
2603          * of changes could have taken place.  Should we check to see if the
2604          * vmspace has been replaced, or the like, in order to prevent
2605          * giving a snapshot that spans, say, execve(2), with some threads
2606          * before and some after?  Among other things, the credentials could
2607          * have changed, in which case the right to extract debug info might
2608          * no longer be assured.
2609          */
2610         i = 0;
2611         FOREACH_THREAD_IN_PROC(p, td) {
2612                 KASSERT(i < numthreads,
2613                     ("sysctl_kern_proc_kstack: numthreads"));
2614                 lwpidarray[i] = td->td_tid;
2615                 i++;
2616         }
2617         numthreads = i;
2618         for (i = 0; i < numthreads; i++) {
2619                 td = thread_find(p, lwpidarray[i]);
2620                 if (td == NULL) {
2621                         continue;
2622                 }
2623                 bzero(kkstp, sizeof(*kkstp));
2624                 (void)sbuf_new(&sb, kkstp->kkst_trace,
2625                     sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
2626                 thread_lock(td);
2627                 kkstp->kkst_tid = td->td_tid;
2628                 if (TD_IS_SWAPPED(td)) {
2629                         kkstp->kkst_state = KKST_STATE_SWAPPED;
2630                 } else if (TD_IS_RUNNING(td)) {
2631                         if (stack_save_td_running(st, td) == 0)
2632                                 kkstp->kkst_state = KKST_STATE_STACKOK;
2633                         else
2634                                 kkstp->kkst_state = KKST_STATE_RUNNING;
2635                 } else {
2636                         kkstp->kkst_state = KKST_STATE_STACKOK;
2637                         stack_save_td(st, td);
2638                 }
2639                 thread_unlock(td);
2640                 PROC_UNLOCK(p);
2641                 stack_sbuf_print(&sb, st);
2642                 sbuf_finish(&sb);
2643                 sbuf_delete(&sb);
2644                 error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
2645                 PROC_LOCK(p);
2646                 if (error)
2647                         break;
2648         }
2649         _PRELE(p);
2650         PROC_UNLOCK(p);
2651         if (lwpidarray != NULL)
2652                 free(lwpidarray, M_TEMP);
2653         stack_destroy(st);
2654         free(kkstp, M_TEMP);
2655         return (error);
2656 }
2657 #endif
2658
2659 /*
2660  * This sysctl allows a process to retrieve the full list of groups from
2661  * itself or another process.
2662  */
2663 static int
2664 sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
2665 {
2666         pid_t *pidp = (pid_t *)arg1;
2667         unsigned int arglen = arg2;
2668         struct proc *p;
2669         struct ucred *cred;
2670         int error;
2671
2672         if (arglen != 1)
2673                 return (EINVAL);
2674         if (*pidp == -1) {      /* -1 means this process */
2675                 p = req->td->td_proc;
2676                 PROC_LOCK(p);
2677         } else {
2678                 error = pget(*pidp, PGET_CANSEE, &p);
2679                 if (error != 0)
2680                         return (error);
2681         }
2682
2683         cred = crhold(p->p_ucred);
2684         PROC_UNLOCK(p);
2685
2686         error = SYSCTL_OUT(req, cred->cr_groups,
2687             cred->cr_ngroups * sizeof(gid_t));
2688         crfree(cred);
2689         return (error);
2690 }
2691
2692 /*
2693  * This sysctl allows a process to retrieve or/and set the resource limit for
2694  * another process.
2695  */
2696 static int
2697 sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
2698 {
2699         int *name = (int *)arg1;
2700         u_int namelen = arg2;
2701         struct rlimit rlim;
2702         struct proc *p;
2703         u_int which;
2704         int flags, error;
2705
2706         if (namelen != 2)
2707                 return (EINVAL);
2708
2709         which = (u_int)name[1];
2710         if (which >= RLIM_NLIMITS)
2711                 return (EINVAL);
2712
2713         if (req->newptr != NULL && req->newlen != sizeof(rlim))
2714                 return (EINVAL);
2715
2716         flags = PGET_HOLD | PGET_NOTWEXIT;
2717         if (req->newptr != NULL)
2718                 flags |= PGET_CANDEBUG;
2719         else
2720                 flags |= PGET_CANSEE;
2721         error = pget((pid_t)name[0], flags, &p);
2722         if (error != 0)
2723                 return (error);
2724
2725         /*
2726          * Retrieve limit.
2727          */
2728         if (req->oldptr != NULL) {
2729                 PROC_LOCK(p);
2730                 lim_rlimit_proc(p, which, &rlim);
2731                 PROC_UNLOCK(p);
2732         }
2733         error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
2734         if (error != 0)
2735                 goto errout;
2736
2737         /*
2738          * Set limit.
2739          */
2740         if (req->newptr != NULL) {
2741                 error = SYSCTL_IN(req, &rlim, sizeof(rlim));
2742                 if (error == 0)
2743                         error = kern_proc_setrlimit(curthread, p, which, &rlim);
2744         }
2745
2746 errout:
2747         PRELE(p);
2748         return (error);
2749 }
2750
2751 /*
2752  * This sysctl allows a process to retrieve ps_strings structure location of
2753  * another process.
2754  */
2755 static int
2756 sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
2757 {
2758         int *name = (int *)arg1;
2759         u_int namelen = arg2;
2760         struct proc *p;
2761         vm_offset_t ps_strings;
2762         int error;
2763 #ifdef COMPAT_FREEBSD32
2764         uint32_t ps_strings32;
2765 #endif
2766
2767         if (namelen != 1)
2768                 return (EINVAL);
2769
2770         error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2771         if (error != 0)
2772                 return (error);
2773 #ifdef COMPAT_FREEBSD32
2774         if ((req->flags & SCTL_MASK32) != 0) {
2775                 /*
2776                  * We return 0 if the 32 bit emulation request is for a 64 bit
2777                  * process.
2778                  */
2779                 ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
2780                     PTROUT(p->p_sysent->sv_psstrings) : 0;
2781                 PROC_UNLOCK(p);
2782                 error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
2783                 return (error);
2784         }
2785 #endif
2786         ps_strings = p->p_sysent->sv_psstrings;
2787         PROC_UNLOCK(p);
2788         error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
2789         return (error);
2790 }
2791
2792 /*
2793  * This sysctl allows a process to retrieve umask of another process.
2794  */
2795 static int
2796 sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
2797 {
2798         int *name = (int *)arg1;
2799         u_int namelen = arg2;
2800         struct proc *p;
2801         int error;
2802         u_short fd_cmask;
2803
2804         if (namelen != 1)
2805                 return (EINVAL);
2806
2807         error = pget((pid_t)name[0], PGET_WANTREAD, &p);
2808         if (error != 0)
2809                 return (error);
2810
2811         FILEDESC_SLOCK(p->p_fd);
2812         fd_cmask = p->p_fd->fd_cmask;
2813         FILEDESC_SUNLOCK(p->p_fd);
2814         PRELE(p);
2815         error = SYSCTL_OUT(req, &fd_cmask, sizeof(fd_cmask));
2816         return (error);
2817 }
2818
2819 /*
2820  * This sysctl allows a process to set and retrieve binary osreldate of
2821  * another process.
2822  */
2823 static int
2824 sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
2825 {
2826         int *name = (int *)arg1;
2827         u_int namelen = arg2;
2828         struct proc *p;
2829         int flags, error, osrel;
2830
2831         if (namelen != 1)
2832                 return (EINVAL);
2833
2834         if (req->newptr != NULL && req->newlen != sizeof(osrel))
2835                 return (EINVAL);
2836
2837         flags = PGET_HOLD | PGET_NOTWEXIT;
2838         if (req->newptr != NULL)
2839                 flags |= PGET_CANDEBUG;
2840         else
2841                 flags |= PGET_CANSEE;
2842         error = pget((pid_t)name[0], flags, &p);
2843         if (error != 0)
2844                 return (error);
2845
2846         error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
2847         if (error != 0)
2848                 goto errout;
2849
2850         if (req->newptr != NULL) {
2851                 error = SYSCTL_IN(req, &osrel, sizeof(osrel));
2852                 if (error != 0)
2853                         goto errout;
2854                 if (osrel < 0) {
2855                         error = EINVAL;
2856                         goto errout;
2857                 }
2858                 p->p_osrel = osrel;
2859         }
2860 errout:
2861         PRELE(p);
2862         return (error);
2863 }
2864
2865 static int
2866 sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS)
2867 {
2868         int *name = (int *)arg1;
2869         u_int namelen = arg2;
2870         struct proc *p;
2871         struct kinfo_sigtramp kst;
2872         const struct sysentvec *sv;
2873         int error;
2874 #ifdef COMPAT_FREEBSD32
2875         struct kinfo_sigtramp32 kst32;
2876 #endif
2877
2878         if (namelen != 1)
2879                 return (EINVAL);
2880
2881         error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
2882         if (error != 0)
2883                 return (error);
2884         sv = p->p_sysent;
2885 #ifdef COMPAT_FREEBSD32
2886         if ((req->flags & SCTL_MASK32) != 0) {
2887                 bzero(&kst32, sizeof(kst32));
2888                 if (SV_PROC_FLAG(p, SV_ILP32)) {
2889                         if (sv->sv_sigcode_base != 0) {
2890                                 kst32.ksigtramp_start = sv->sv_sigcode_base;
2891                                 kst32.ksigtramp_end = sv->sv_sigcode_base +
2892                                     *sv->sv_szsigcode;
2893                         } else {
2894                                 kst32.ksigtramp_start = sv->sv_psstrings -
2895                                     *sv->sv_szsigcode;
2896                                 kst32.ksigtramp_end = sv->sv_psstrings;
2897                         }
2898                 }
2899                 PROC_UNLOCK(p);
2900                 error = SYSCTL_OUT(req, &kst32, sizeof(kst32));
2901                 return (error);
2902         }
2903 #endif
2904         bzero(&kst, sizeof(kst));
2905         if (sv->sv_sigcode_base != 0) {
2906                 kst.ksigtramp_start = (char *)sv->sv_sigcode_base;
2907                 kst.ksigtramp_end = (char *)sv->sv_sigcode_base +
2908                     *sv->sv_szsigcode;
2909         } else {
2910                 kst.ksigtramp_start = (char *)sv->sv_psstrings -
2911                     *sv->sv_szsigcode;
2912                 kst.ksigtramp_end = (char *)sv->sv_psstrings;
2913         }
2914         PROC_UNLOCK(p);
2915         error = SYSCTL_OUT(req, &kst, sizeof(kst));
2916         return (error);
2917 }
2918
2919 SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD,  0, "Process table");
2920
2921 SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
2922         CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
2923         "Return entire process table");
2924
2925 static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2926         sysctl_kern_proc, "Process table");
2927
2928 static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
2929         sysctl_kern_proc, "Process table");
2930
2931 static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2932         sysctl_kern_proc, "Process table");
2933
2934 static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
2935         CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2936
2937 static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
2938         sysctl_kern_proc, "Process table");
2939
2940 static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2941         sysctl_kern_proc, "Process table");
2942
2943 static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2944         sysctl_kern_proc, "Process table");
2945
2946 static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
2947         sysctl_kern_proc, "Process table");
2948
2949 static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
2950         sysctl_kern_proc, "Return process table, no threads");
2951
2952 static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
2953         CTLFLAG_RW | CTLFLAG_CAPWR | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
2954         sysctl_kern_proc_args, "Process argument list");
2955
2956 static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
2957         sysctl_kern_proc_env, "Process environment");
2958
2959 static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
2960         CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
2961
2962 static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
2963         CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
2964
2965 static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
2966         CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
2967         "Process syscall vector name (ABI type)");
2968
2969 static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
2970         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2971
2972 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
2973         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2974
2975 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
2976         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2977
2978 static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
2979         sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2980
2981 static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
2982         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2983
2984 static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
2985         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2986
2987 static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
2988         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2989
2990 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
2991         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
2992
2993 static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
2994         CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
2995         "Return process table, no threads");
2996
2997 #ifdef COMPAT_FREEBSD7
2998 static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
2999         CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
3000 #endif
3001
3002 static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
3003         CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
3004
3005 #if defined(STACK) || defined(DDB)
3006 static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
3007         CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
3008 #endif
3009
3010 static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
3011         CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
3012
3013 static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
3014         CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
3015         "Process resource limits");
3016
3017 static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
3018         CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
3019         "Process ps_strings location");
3020
3021 static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
3022         CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
3023
3024 static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
3025         CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
3026         "Process binary osreldate");
3027
3028 static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD |
3029         CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp,
3030         "Process signal trampoline location");
3031
3032 int allproc_gen;
3033
3034 /*
3035  * stop_all_proc() purpose is to stop all process which have usermode,
3036  * except current process for obvious reasons.  This makes it somewhat
3037  * unreliable when invoked from multithreaded process.  The service
3038  * must not be user-callable anyway.
3039  */
3040 void
3041 stop_all_proc(void)
3042 {
3043         struct proc *cp, *p;
3044         int r, gen;
3045         bool restart, seen_stopped, seen_exiting, stopped_some;
3046
3047         cp = curproc;
3048 allproc_loop:
3049         sx_xlock(&allproc_lock);
3050         gen = allproc_gen;
3051         seen_exiting = seen_stopped = stopped_some = restart = false;
3052         LIST_REMOVE(cp, p_list);
3053         LIST_INSERT_HEAD(&allproc, cp, p_list);
3054         for (;;) {
3055                 p = LIST_NEXT(cp, p_list);
3056                 if (p == NULL)
3057                         break;
3058                 LIST_REMOVE(cp, p_list);
3059                 LIST_INSERT_AFTER(p, cp, p_list);
3060                 PROC_LOCK(p);
3061                 if ((p->p_flag & (P_KPROC | P_SYSTEM | P_TOTAL_STOP)) != 0) {
3062                         PROC_UNLOCK(p);
3063                         continue;
3064                 }
3065                 if ((p->p_flag & P_WEXIT) != 0) {
3066                         seen_exiting = true;
3067                         PROC_UNLOCK(p);
3068                         continue;
3069                 }
3070                 if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) {
3071                         /*
3072                          * Stopped processes are tolerated when there
3073                          * are no other processes which might continue
3074                          * them.  P_STOPPED_SINGLE but not
3075                          * P_TOTAL_STOP process still has at least one
3076                          * thread running.
3077                          */
3078                         seen_stopped = true;
3079                         PROC_UNLOCK(p);
3080                         continue;
3081                 }
3082                 sx_xunlock(&allproc_lock);
3083                 _PHOLD(p);
3084                 r = thread_single(p, SINGLE_ALLPROC);
3085                 if (r != 0)
3086                         restart = true;
3087                 else
3088                         stopped_some = true;
3089                 _PRELE(p);
3090                 PROC_UNLOCK(p);
3091                 sx_xlock(&allproc_lock);
3092         }
3093         /* Catch forked children we did not see in iteration. */
3094         if (gen != allproc_gen)
3095                 restart = true;
3096         sx_xunlock(&allproc_lock);
3097         if (restart || stopped_some || seen_exiting || seen_stopped) {
3098                 kern_yield(PRI_USER);
3099                 goto allproc_loop;
3100         }
3101 }
3102
3103 void
3104 resume_all_proc(void)
3105 {
3106         struct proc *cp, *p;
3107
3108         cp = curproc;
3109         sx_xlock(&allproc_lock);
3110 again:
3111         LIST_REMOVE(cp, p_list);
3112         LIST_INSERT_HEAD(&allproc, cp, p_list);
3113         for (;;) {
3114                 p = LIST_NEXT(cp, p_list);
3115                 if (p == NULL)
3116                         break;
3117                 LIST_REMOVE(cp, p_list);
3118                 LIST_INSERT_AFTER(p, cp, p_list);
3119                 PROC_LOCK(p);
3120                 if ((p->p_flag & P_TOTAL_STOP) != 0) {
3121                         sx_xunlock(&allproc_lock);
3122                         _PHOLD(p);
3123                         thread_single_end(p, SINGLE_ALLPROC);
3124                         _PRELE(p);
3125                         PROC_UNLOCK(p);
3126                         sx_xlock(&allproc_lock);
3127                 } else {
3128                         PROC_UNLOCK(p);
3129                 }
3130         }
3131         /*  Did the loop above missed any stopped process ? */
3132         LIST_FOREACH(p, &allproc, p_list) {
3133                 /* No need for proc lock. */
3134                 if ((p->p_flag & P_TOTAL_STOP) != 0)
3135                         goto again;
3136         }
3137         sx_xunlock(&allproc_lock);
3138 }
3139
3140 /* #define      TOTAL_STOP_DEBUG        1 */
3141 #ifdef TOTAL_STOP_DEBUG
3142 volatile static int ap_resume;
3143 #include <sys/mount.h>
3144
3145 static int
3146 sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS)
3147 {
3148         int error, val;
3149
3150         val = 0;
3151         ap_resume = 0;
3152         error = sysctl_handle_int(oidp, &val, 0, req);
3153         if (error != 0 || req->newptr == NULL)
3154                 return (error);
3155         if (val != 0) {
3156                 stop_all_proc();
3157                 syncer_suspend();
3158                 while (ap_resume == 0)
3159                         ;
3160                 syncer_resume();
3161                 resume_all_proc();
3162         }
3163         return (0);
3164 }
3165
3166 SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW |
3167     CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0,
3168     sysctl_debug_stop_all_proc, "I",
3169     "");
3170 #endif