The Design and Implementation of the FreeBSD Operating System, Second Edition
Now available: The Design and Implementation of the FreeBSD Operating System (Second Edition)


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]

FreeBSD/Linux Kernel Cross Reference
sys/kern/kern_exit.c

Version: -  FREEBSD  -  FREEBSD-13-STABLE  -  FREEBSD-13-0  -  FREEBSD-12-STABLE  -  FREEBSD-12-0  -  FREEBSD-11-STABLE  -  FREEBSD-11-0  -  FREEBSD-10-STABLE  -  FREEBSD-10-0  -  FREEBSD-9-STABLE  -  FREEBSD-9-0  -  FREEBSD-8-STABLE  -  FREEBSD-8-0  -  FREEBSD-7-STABLE  -  FREEBSD-7-0  -  FREEBSD-6-STABLE  -  FREEBSD-6-0  -  FREEBSD-5-STABLE  -  FREEBSD-5-0  -  FREEBSD-4-STABLE  -  FREEBSD-3-STABLE  -  FREEBSD22  -  l41  -  OPENBSD  -  linux-2.6  -  MK84  -  PLAN9  -  xnu-8792 
SearchContext: -  none  -  3  -  10 

    1 /*-
    2  * SPDX-License-Identifier: BSD-3-Clause
    3  *
    4  * Copyright (c) 1982, 1986, 1989, 1991, 1993
    5  *      The Regents of the University of California.  All rights reserved.
    6  * (c) UNIX System Laboratories, Inc.
    7  * All or some portions of this file are derived from material licensed
    8  * to the University of California by American Telephone and Telegraph
    9  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
   10  * the permission of UNIX System Laboratories, Inc.
   11  *
   12  * Redistribution and use in source and binary forms, with or without
   13  * modification, are permitted provided that the following conditions
   14  * are met:
   15  * 1. Redistributions of source code must retain the above copyright
   16  *    notice, this list of conditions and the following disclaimer.
   17  * 2. Redistributions in binary form must reproduce the above copyright
   18  *    notice, this list of conditions and the following disclaimer in the
   19  *    documentation and/or other materials provided with the distribution.
   20  * 3. Neither the name of the University nor the names of its contributors
   21  *    may be used to endorse or promote products derived from this software
   22  *    without specific prior written permission.
   23  *
   24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
   25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
   26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
   27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
   28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
   29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
   30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
   31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
   32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
   33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   34  * SUCH DAMAGE.
   35  *
   36  *      @(#)kern_exit.c 8.7 (Berkeley) 2/12/94
   37  */
   38 
   39 #include <sys/cdefs.h>
   40 __FBSDID("$FreeBSD$");
   41 
   42 #include "opt_ddb.h"
   43 #include "opt_ktrace.h"
   44 
   45 #include <sys/param.h>
   46 #include <sys/systm.h>
   47 #include <sys/sysproto.h>
   48 #include <sys/capsicum.h>
   49 #include <sys/eventhandler.h>
   50 #include <sys/kernel.h>
   51 #include <sys/ktr.h>
   52 #include <sys/malloc.h>
   53 #include <sys/lock.h>
   54 #include <sys/mutex.h>
   55 #include <sys/proc.h>
   56 #include <sys/procdesc.h>
   57 #include <sys/jail.h>
   58 #include <sys/tty.h>
   59 #include <sys/wait.h>
   60 #include <sys/vmmeter.h>
   61 #include <sys/vnode.h>
   62 #include <sys/racct.h>
   63 #include <sys/resourcevar.h>
   64 #include <sys/sbuf.h>
   65 #include <sys/signalvar.h>
   66 #include <sys/sched.h>
   67 #include <sys/sx.h>
   68 #include <sys/syscallsubr.h>
   69 #include <sys/syslog.h>
   70 #include <sys/ptrace.h>
   71 #include <sys/acct.h>           /* for acct_process() function prototype */
   72 #include <sys/filedesc.h>
   73 #include <sys/sdt.h>
   74 #include <sys/shm.h>
   75 #include <sys/sem.h>
   76 #include <sys/sysent.h>
   77 #include <sys/timers.h>
   78 #include <sys/umtx.h>
   79 #ifdef KTRACE
   80 #include <sys/ktrace.h>
   81 #endif
   82 
   83 #include <security/audit/audit.h>
   84 #include <security/mac/mac_framework.h>
   85 
   86 #include <vm/vm.h>
   87 #include <vm/vm_extern.h>
   88 #include <vm/vm_param.h>
   89 #include <vm/pmap.h>
   90 #include <vm/vm_map.h>
   91 #include <vm/vm_page.h>
   92 #include <vm/uma.h>
   93 
   94 #ifdef KDTRACE_HOOKS
   95 #include <sys/dtrace_bsd.h>
   96 dtrace_execexit_func_t  dtrace_fasttrap_exit;
   97 #endif
   98 
   99 SDT_PROVIDER_DECLARE(proc);
  100 SDT_PROBE_DEFINE1(proc, , , exit, "int");
  101 
  102 struct proc *
  103 proc_realparent(struct proc *child)
  104 {
  105         struct proc *p, *parent;
  106 
  107         sx_assert(&proctree_lock, SX_LOCKED);
  108         if ((child->p_treeflag & P_TREE_ORPHANED) == 0)
  109                 return (child->p_pptr->p_pid == child->p_oppid ?
  110                     child->p_pptr : child->p_reaper);
  111         for (p = child; (p->p_treeflag & P_TREE_FIRST_ORPHAN) == 0;) {
  112                 /* Cannot use LIST_PREV(), since the list head is not known. */
  113                 p = __containerof(p->p_orphan.le_prev, struct proc,
  114                     p_orphan.le_next);
  115                 KASSERT((p->p_treeflag & P_TREE_ORPHANED) != 0,
  116                     ("missing P_ORPHAN %p", p));
  117         }
  118         parent = __containerof(p->p_orphan.le_prev, struct proc,
  119             p_orphans.lh_first);
  120         return (parent);
  121 }
  122 
  123 void
  124 reaper_abandon_children(struct proc *p, bool exiting)
  125 {
  126         struct proc *p1, *p2, *ptmp;
  127 
  128         sx_assert(&proctree_lock, SX_LOCKED);
  129         KASSERT(p != initproc, ("reaper_abandon_children for initproc"));
  130         if ((p->p_treeflag & P_TREE_REAPER) == 0)
  131                 return;
  132         p1 = p->p_reaper;
  133         LIST_FOREACH_SAFE(p2, &p->p_reaplist, p_reapsibling, ptmp) {
  134                 LIST_REMOVE(p2, p_reapsibling);
  135                 p2->p_reaper = p1;
  136                 p2->p_reapsubtree = p->p_reapsubtree;
  137                 LIST_INSERT_HEAD(&p1->p_reaplist, p2, p_reapsibling);
  138                 if (exiting && p2->p_pptr == p) {
  139                         PROC_LOCK(p2);
  140                         proc_reparent(p2, p1, true);
  141                         PROC_UNLOCK(p2);
  142                 }
  143         }
  144         KASSERT(LIST_EMPTY(&p->p_reaplist), ("p_reaplist not empty"));
  145         p->p_treeflag &= ~P_TREE_REAPER;
  146 }
  147 
  148 static void
  149 reaper_clear(struct proc *p)
  150 {
  151         struct proc *p1;
  152         bool clear;
  153 
  154         sx_assert(&proctree_lock, SX_LOCKED);
  155         LIST_REMOVE(p, p_reapsibling);
  156         if (p->p_reapsubtree == 1)
  157                 return;
  158         clear = true;
  159         LIST_FOREACH(p1, &p->p_reaper->p_reaplist, p_reapsibling) {
  160                 if (p1->p_reapsubtree == p->p_reapsubtree) {
  161                         clear = false;
  162                         break;
  163                 }
  164         }
  165         if (clear)
  166                 proc_id_clear(PROC_ID_REAP, p->p_reapsubtree);
  167 }
  168 
  169 void
  170 proc_clear_orphan(struct proc *p)
  171 {
  172         struct proc *p1;
  173 
  174         sx_assert(&proctree_lock, SA_XLOCKED);
  175         if ((p->p_treeflag & P_TREE_ORPHANED) == 0)
  176                 return;
  177         if ((p->p_treeflag & P_TREE_FIRST_ORPHAN) != 0) {
  178                 p1 = LIST_NEXT(p, p_orphan);
  179                 if (p1 != NULL)
  180                         p1->p_treeflag |= P_TREE_FIRST_ORPHAN;
  181                 p->p_treeflag &= ~P_TREE_FIRST_ORPHAN;
  182         }
  183         LIST_REMOVE(p, p_orphan);
  184         p->p_treeflag &= ~P_TREE_ORPHANED;
  185 }
  186 
  187 /*
  188  * exit -- death of process.
  189  */
  190 void
  191 sys_sys_exit(struct thread *td, struct sys_exit_args *uap)
  192 {
  193 
  194         exit1(td, uap->rval, 0);
  195         /* NOTREACHED */
  196 }
  197 
  198 /*
  199  * Exit: deallocate address space and other resources, change proc state to
  200  * zombie, and unlink proc from allproc and parent's lists.  Save exit status
  201  * and rusage for wait().  Check for child processes and orphan them.
  202  */
  203 void
  204 exit1(struct thread *td, int rval, int signo)
  205 {
  206         struct proc *p, *nq, *q, *t;
  207         struct thread *tdt;
  208         ksiginfo_t *ksi, *ksi1;
  209         int signal_parent;
  210 
  211         mtx_assert(&Giant, MA_NOTOWNED);
  212         KASSERT(rval == 0 || signo == 0, ("exit1 rv %d sig %d", rval, signo));
  213 
  214         p = td->td_proc;
  215         /*
  216          * XXX in case we're rebooting we just let init die in order to
  217          * work around an unsolved stack overflow seen very late during
  218          * shutdown on sparc64 when the gmirror worker process exists.
  219          * XXX what to do now that sparc64 is gone... remove if?
  220          */
  221         if (p == initproc && rebooting == 0) {
  222                 printf("init died (signal %d, exit %d)\n", signo, rval);
  223                 panic("Going nowhere without my init!");
  224         }
  225 
  226         /*
  227          * Deref SU mp, since the thread does not return to userspace.
  228          */
  229         td_softdep_cleanup(td);
  230 
  231         /*
  232          * MUST abort all other threads before proceeding past here.
  233          */
  234         PROC_LOCK(p);
  235         /*
  236          * First check if some other thread or external request got
  237          * here before us.  If so, act appropriately: exit or suspend.
  238          * We must ensure that stop requests are handled before we set
  239          * P_WEXIT.
  240          */
  241         thread_suspend_check(0);
  242         while (p->p_flag & P_HADTHREADS) {
  243                 /*
  244                  * Kill off the other threads. This requires
  245                  * some co-operation from other parts of the kernel
  246                  * so it may not be instantaneous.  With this state set
  247                  * any thread entering the kernel from userspace will
  248                  * thread_exit() in trap().  Any thread attempting to
  249                  * sleep will return immediately with EINTR or EWOULDBLOCK
  250                  * which will hopefully force them to back out to userland
  251                  * freeing resources as they go.  Any thread attempting
  252                  * to return to userland will thread_exit() from userret().
  253                  * thread_exit() will unsuspend us when the last of the
  254                  * other threads exits.
  255                  * If there is already a thread singler after resumption,
  256                  * calling thread_single will fail; in that case, we just
  257                  * re-check all suspension request, the thread should
  258                  * either be suspended there or exit.
  259                  */
  260                 if (!thread_single(p, SINGLE_EXIT))
  261                         /*
  262                          * All other activity in this process is now
  263                          * stopped.  Threading support has been turned
  264                          * off.
  265                          */
  266                         break;
  267                 /*
  268                  * Recheck for new stop or suspend requests which
  269                  * might appear while process lock was dropped in
  270                  * thread_single().
  271                  */
  272                 thread_suspend_check(0);
  273         }
  274         KASSERT(p->p_numthreads == 1,
  275             ("exit1: proc %p exiting with %d threads", p, p->p_numthreads));
  276         racct_sub(p, RACCT_NTHR, 1);
  277 
  278         /* Let event handler change exit status */
  279         p->p_xexit = rval;
  280         p->p_xsig = signo;
  281 
  282         /*
  283          * Ignore any pending request to stop due to a stop signal.
  284          * Once P_WEXIT is set, future requests will be ignored as
  285          * well.
  286          */
  287         p->p_flag &= ~P_STOPPED_SIG;
  288         KASSERT(!P_SHOULDSTOP(p), ("exiting process is stopped"));
  289 
  290         /* Note that we are exiting. */
  291         p->p_flag |= P_WEXIT;
  292 
  293         /*
  294          * Wait for any processes that have a hold on our vmspace to
  295          * release their reference.
  296          */
  297         while (p->p_lock > 0)
  298                 msleep(&p->p_lock, &p->p_mtx, PWAIT, "exithold", 0);
  299 
  300         PROC_UNLOCK(p);
  301         /* Drain the limit callout while we don't have the proc locked */
  302         callout_drain(&p->p_limco);
  303 
  304 #ifdef AUDIT
  305         /*
  306          * The Sun BSM exit token contains two components: an exit status as
  307          * passed to exit(), and a return value to indicate what sort of exit
  308          * it was.  The exit status is WEXITSTATUS(rv), but it's not clear
  309          * what the return value is.
  310          */
  311         AUDIT_ARG_EXIT(rval, 0);
  312         AUDIT_SYSCALL_EXIT(0, td);
  313 #endif
  314 
  315         /* Are we a task leader with peers? */
  316         if (p->p_peers != NULL && p == p->p_leader) {
  317                 mtx_lock(&ppeers_lock);
  318                 q = p->p_peers;
  319                 while (q != NULL) {
  320                         PROC_LOCK(q);
  321                         kern_psignal(q, SIGKILL);
  322                         PROC_UNLOCK(q);
  323                         q = q->p_peers;
  324                 }
  325                 while (p->p_peers != NULL)
  326                         msleep(p, &ppeers_lock, PWAIT, "exit1", 0);
  327                 mtx_unlock(&ppeers_lock);
  328         }
  329 
  330         itimers_exit(p);
  331 
  332         if (p->p_sysent->sv_onexit != NULL)
  333                 p->p_sysent->sv_onexit(p);
  334 
  335         /*
  336          * Check if any loadable modules need anything done at process exit.
  337          * E.g. SYSV IPC stuff.
  338          * Event handler could change exit status.
  339          * XXX what if one of these generates an error?
  340          */
  341         EVENTHANDLER_DIRECT_INVOKE(process_exit, p);
  342 
  343         /*
  344          * If parent is waiting for us to exit or exec,
  345          * P_PPWAIT is set; we will wakeup the parent below.
  346          */
  347         PROC_LOCK(p);
  348         stopprofclock(p);
  349         p->p_ptevents = 0;
  350 
  351         /*
  352          * Stop the real interval timer.  If the handler is currently
  353          * executing, prevent it from rearming itself and let it finish.
  354          */
  355         if (timevalisset(&p->p_realtimer.it_value) &&
  356             _callout_stop_safe(&p->p_itcallout, CS_EXECUTING, NULL) == 0) {
  357                 timevalclear(&p->p_realtimer.it_interval);
  358                 msleep(&p->p_itcallout, &p->p_mtx, PWAIT, "ritwait", 0);
  359                 KASSERT(!timevalisset(&p->p_realtimer.it_value),
  360                     ("realtime timer is still armed"));
  361         }
  362 
  363         PROC_UNLOCK(p);
  364 
  365         umtx_thread_exit(td);
  366         seltdfini(td);
  367 
  368         /*
  369          * Reset any sigio structures pointing to us as a result of
  370          * F_SETOWN with our pid.  The P_WEXIT flag interlocks with fsetown().
  371          */
  372         funsetownlst(&p->p_sigiolst);
  373 
  374         /*
  375          * Close open files and release open-file table.
  376          * This may block!
  377          */
  378         pdescfree(td);
  379         fdescfree(td);
  380 
  381         /*
  382          * If this thread tickled GEOM, we need to wait for the giggling to
  383          * stop before we return to userland
  384          */
  385         if (td->td_pflags & TDP_GEOM)
  386                 g_waitidle();
  387 
  388         /*
  389          * Remove ourself from our leader's peer list and wake our leader.
  390          */
  391         if (p->p_leader->p_peers != NULL) {
  392                 mtx_lock(&ppeers_lock);
  393                 if (p->p_leader->p_peers != NULL) {
  394                         q = p->p_leader;
  395                         while (q->p_peers != p)
  396                                 q = q->p_peers;
  397                         q->p_peers = p->p_peers;
  398                         wakeup(p->p_leader);
  399                 }
  400                 mtx_unlock(&ppeers_lock);
  401         }
  402 
  403         vmspace_exit(td);
  404         (void)acct_process(td);
  405 
  406 #ifdef KTRACE
  407         ktrprocexit(td);
  408 #endif
  409         /*
  410          * Release reference to text vnode
  411          */
  412         if (p->p_textvp != NULL) {
  413                 vrele(p->p_textvp);
  414                 p->p_textvp = NULL;
  415         }
  416 
  417         /*
  418          * Release our limits structure.
  419          */
  420         lim_free(p->p_limit);
  421         p->p_limit = NULL;
  422 
  423         tidhash_remove(td);
  424 
  425         /*
  426          * Call machine-dependent code to release any
  427          * machine-dependent resources other than the address space.
  428          * The address space is released by "vmspace_exitfree(p)" in
  429          * vm_waitproc().
  430          */
  431         cpu_exit(td);
  432 
  433         WITNESS_WARN(WARN_PANIC, NULL, "process (pid %d) exiting", p->p_pid);
  434 
  435         /*
  436          * Remove from allproc. It still sits in the hash.
  437          */
  438         sx_xlock(&allproc_lock);
  439         LIST_REMOVE(p, p_list);
  440 
  441 #ifdef DDB
  442         /*
  443          * Used by ddb's 'ps' command to find this process via the
  444          * pidhash.
  445          */
  446         p->p_list.le_prev = NULL;
  447 #endif
  448         sx_xunlock(&allproc_lock);
  449 
  450         sx_xlock(&proctree_lock);
  451         PROC_LOCK(p);
  452         p->p_flag &= ~(P_TRACED | P_PPWAIT | P_PPTRACE);
  453         PROC_UNLOCK(p);
  454 
  455         /*
  456          * killjobc() might drop and re-acquire proctree_lock to
  457          * revoke control tty if exiting process was a session leader.
  458          */
  459         killjobc();
  460 
  461         /*
  462          * Reparent all children processes:
  463          * - traced ones to the original parent (or init if we are that parent)
  464          * - the rest to init
  465          */
  466         q = LIST_FIRST(&p->p_children);
  467         if (q != NULL)          /* only need this if any child is S_ZOMB */
  468                 wakeup(q->p_reaper);
  469         for (; q != NULL; q = nq) {
  470                 nq = LIST_NEXT(q, p_sibling);
  471                 ksi = ksiginfo_alloc(TRUE);
  472                 PROC_LOCK(q);
  473                 q->p_sigparent = SIGCHLD;
  474 
  475                 if ((q->p_flag & P_TRACED) == 0) {
  476                         proc_reparent(q, q->p_reaper, true);
  477                         if (q->p_state == PRS_ZOMBIE) {
  478                                 /*
  479                                  * Inform reaper about the reparented
  480                                  * zombie, since wait(2) has something
  481                                  * new to report.  Guarantee queueing
  482                                  * of the SIGCHLD signal, similar to
  483                                  * the _exit() behaviour, by providing
  484                                  * our ksiginfo.  Ksi is freed by the
  485                                  * signal delivery.
  486                                  */
  487                                 if (q->p_ksi == NULL) {
  488                                         ksi1 = NULL;
  489                                 } else {
  490                                         ksiginfo_copy(q->p_ksi, ksi);
  491                                         ksi->ksi_flags |= KSI_INS;
  492                                         ksi1 = ksi;
  493                                         ksi = NULL;
  494                                 }
  495                                 PROC_LOCK(q->p_reaper);
  496                                 pksignal(q->p_reaper, SIGCHLD, ksi1);
  497                                 PROC_UNLOCK(q->p_reaper);
  498                         } else if (q->p_pdeathsig > 0) {
  499                                 /*
  500                                  * The child asked to received a signal
  501                                  * when we exit.
  502                                  */
  503                                 kern_psignal(q, q->p_pdeathsig);
  504                         }
  505                 } else {
  506                         /*
  507                          * Traced processes are killed since their existence
  508                          * means someone is screwing up.
  509                          */
  510                         t = proc_realparent(q);
  511                         if (t == p) {
  512                                 proc_reparent(q, q->p_reaper, true);
  513                         } else {
  514                                 PROC_LOCK(t);
  515                                 proc_reparent(q, t, true);
  516                                 PROC_UNLOCK(t);
  517                         }
  518                         /*
  519                          * Since q was found on our children list, the
  520                          * proc_reparent() call moved q to the orphan
  521                          * list due to present P_TRACED flag. Clear
  522                          * orphan link for q now while q is locked.
  523                          */
  524                         proc_clear_orphan(q);
  525                         q->p_flag &= ~(P_TRACED | P_STOPPED_TRACE);
  526                         q->p_flag2 &= ~P2_PTRACE_FSTP;
  527                         q->p_ptevents = 0;
  528                         FOREACH_THREAD_IN_PROC(q, tdt) {
  529                                 tdt->td_dbgflags &= ~(TDB_SUSPEND | TDB_XSIG |
  530                                     TDB_FSTP);
  531                         }
  532                         kern_psignal(q, SIGKILL);
  533                 }
  534                 PROC_UNLOCK(q);
  535                 if (ksi != NULL)
  536                         ksiginfo_free(ksi);
  537         }
  538 
  539         /*
  540          * Also get rid of our orphans.
  541          */
  542         while ((q = LIST_FIRST(&p->p_orphans)) != NULL) {
  543                 PROC_LOCK(q);
  544                 KASSERT(q->p_oppid == p->p_pid,
  545                     ("orphan %p of %p has unexpected oppid %d", q, p,
  546                     q->p_oppid));
  547                 q->p_oppid = q->p_reaper->p_pid;
  548 
  549                 /*
  550                  * If we are the real parent of this process
  551                  * but it has been reparented to a debugger, then
  552                  * check if it asked for a signal when we exit.
  553                  */
  554                 if (q->p_pdeathsig > 0)
  555                         kern_psignal(q, q->p_pdeathsig);
  556                 CTR2(KTR_PTRACE, "exit: pid %d, clearing orphan %d", p->p_pid,
  557                     q->p_pid);
  558                 proc_clear_orphan(q);
  559                 PROC_UNLOCK(q);
  560         }
  561 
  562 #ifdef KDTRACE_HOOKS
  563         if (SDT_PROBES_ENABLED()) {
  564                 int reason = CLD_EXITED;
  565                 if (WCOREDUMP(signo))
  566                         reason = CLD_DUMPED;
  567                 else if (WIFSIGNALED(signo))
  568                         reason = CLD_KILLED;
  569                 SDT_PROBE1(proc, , , exit, reason);
  570         }
  571 #endif
  572 
  573         /* Save exit status. */
  574         PROC_LOCK(p);
  575         p->p_xthread = td;
  576 
  577         if (p->p_sysent->sv_ontdexit != NULL)
  578                 p->p_sysent->sv_ontdexit(td);
  579 
  580 #ifdef KDTRACE_HOOKS
  581         /*
  582          * Tell the DTrace fasttrap provider about the exit if it
  583          * has declared an interest.
  584          */
  585         if (dtrace_fasttrap_exit)
  586                 dtrace_fasttrap_exit(p);
  587 #endif
  588 
  589         /*
  590          * Notify interested parties of our demise.
  591          */
  592         KNOTE_LOCKED(p->p_klist, NOTE_EXIT);
  593 
  594         /*
  595          * If this is a process with a descriptor, we may not need to deliver
  596          * a signal to the parent.  proctree_lock is held over
  597          * procdesc_exit() to serialize concurrent calls to close() and
  598          * exit().
  599          */
  600         signal_parent = 0;
  601         if (p->p_procdesc == NULL || procdesc_exit(p)) {
  602                 /*
  603                  * Notify parent that we're gone.  If parent has the
  604                  * PS_NOCLDWAIT flag set, or if the handler is set to SIG_IGN,
  605                  * notify process 1 instead (and hope it will handle this
  606                  * situation).
  607                  */
  608                 PROC_LOCK(p->p_pptr);
  609                 mtx_lock(&p->p_pptr->p_sigacts->ps_mtx);
  610                 if (p->p_pptr->p_sigacts->ps_flag &
  611                     (PS_NOCLDWAIT | PS_CLDSIGIGN)) {
  612                         struct proc *pp;
  613 
  614                         mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
  615                         pp = p->p_pptr;
  616                         PROC_UNLOCK(pp);
  617                         proc_reparent(p, p->p_reaper, true);
  618                         p->p_sigparent = SIGCHLD;
  619                         PROC_LOCK(p->p_pptr);
  620 
  621                         /*
  622                          * Notify parent, so in case he was wait(2)ing or
  623                          * executing waitpid(2) with our pid, he will
  624                          * continue.
  625                          */
  626                         wakeup(pp);
  627                 } else
  628                         mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
  629 
  630                 if (p->p_pptr == p->p_reaper || p->p_pptr == initproc) {
  631                         signal_parent = 1;
  632                 } else if (p->p_sigparent != 0) {
  633                         if (p->p_sigparent == SIGCHLD) {
  634                                 signal_parent = 1;
  635                         } else { /* LINUX thread */
  636                                 signal_parent = 2;
  637                         }
  638                 }
  639         } else
  640                 PROC_LOCK(p->p_pptr);
  641         sx_xunlock(&proctree_lock);
  642 
  643         if (signal_parent == 1) {
  644                 childproc_exited(p);
  645         } else if (signal_parent == 2) {
  646                 kern_psignal(p->p_pptr, p->p_sigparent);
  647         }
  648 
  649         /* Tell the prison that we are gone. */
  650         prison_proc_free(p->p_ucred->cr_prison);
  651 
  652         /*
  653          * The state PRS_ZOMBIE prevents other proesses from sending
  654          * signal to the process, to avoid memory leak, we free memory
  655          * for signal queue at the time when the state is set.
  656          */
  657         sigqueue_flush(&p->p_sigqueue);
  658         sigqueue_flush(&td->td_sigqueue);
  659 
  660         /*
  661          * We have to wait until after acquiring all locks before
  662          * changing p_state.  We need to avoid all possible context
  663          * switches (including ones from blocking on a mutex) while
  664          * marked as a zombie.  We also have to set the zombie state
  665          * before we release the parent process' proc lock to avoid
  666          * a lost wakeup.  So, we first call wakeup, then we grab the
  667          * sched lock, update the state, and release the parent process'
  668          * proc lock.
  669          */
  670         wakeup(p->p_pptr);
  671         cv_broadcast(&p->p_pwait);
  672         sched_exit(p->p_pptr, td);
  673         PROC_SLOCK(p);
  674         p->p_state = PRS_ZOMBIE;
  675         PROC_UNLOCK(p->p_pptr);
  676 
  677         /*
  678          * Save our children's rusage information in our exit rusage.
  679          */
  680         PROC_STATLOCK(p);
  681         ruadd(&p->p_ru, &p->p_rux, &p->p_stats->p_cru, &p->p_crux);
  682         PROC_STATUNLOCK(p);
  683 
  684         /*
  685          * Make sure the scheduler takes this thread out of its tables etc.
  686          * This will also release this thread's reference to the ucred.
  687          * Other thread parts to release include pcb bits and such.
  688          */
  689         thread_exit();
  690 }
  691 
  692 #ifndef _SYS_SYSPROTO_H_
  693 struct abort2_args {
  694         char *why;
  695         int nargs;
  696         void **args;
  697 };
  698 #endif
  699 
  700 int
  701 sys_abort2(struct thread *td, struct abort2_args *uap)
  702 {
  703         struct proc *p = td->td_proc;
  704         struct sbuf *sb;
  705         void *uargs[16];
  706         int error, i, sig;
  707 
  708         /*
  709          * Do it right now so we can log either proper call of abort2(), or
  710          * note, that invalid argument was passed. 512 is big enough to
  711          * handle 16 arguments' descriptions with additional comments.
  712          */
  713         sb = sbuf_new(NULL, NULL, 512, SBUF_FIXEDLEN);
  714         sbuf_clear(sb);
  715         sbuf_printf(sb, "%s(pid %d uid %d) aborted: ",
  716             p->p_comm, p->p_pid, td->td_ucred->cr_uid);
  717         /*
  718          * Since we can't return from abort2(), send SIGKILL in cases, where
  719          * abort2() was called improperly
  720          */
  721         sig = SIGKILL;
  722         /* Prevent from DoSes from user-space. */
  723         if (uap->nargs < 0 || uap->nargs > 16)
  724                 goto out;
  725         if (uap->nargs > 0) {
  726                 if (uap->args == NULL)
  727                         goto out;
  728                 error = copyin(uap->args, uargs, uap->nargs * sizeof(void *));
  729                 if (error != 0)
  730                         goto out;
  731         }
  732         /*
  733          * Limit size of 'reason' string to 128. Will fit even when
  734          * maximal number of arguments was chosen to be logged.
  735          */
  736         if (uap->why != NULL) {
  737                 error = sbuf_copyin(sb, uap->why, 128);
  738                 if (error < 0)
  739                         goto out;
  740         } else {
  741                 sbuf_printf(sb, "(null)");
  742         }
  743         if (uap->nargs > 0) {
  744                 sbuf_printf(sb, "(");
  745                 for (i = 0;i < uap->nargs; i++)
  746                         sbuf_printf(sb, "%s%p", i == 0 ? "" : ", ", uargs[i]);
  747                 sbuf_printf(sb, ")");
  748         }
  749         /*
  750          * Final stage: arguments were proper, string has been
  751          * successfully copied from userspace, and copying pointers
  752          * from user-space succeed.
  753          */
  754         sig = SIGABRT;
  755 out:
  756         if (sig == SIGKILL) {
  757                 sbuf_trim(sb);
  758                 sbuf_printf(sb, " (Reason text inaccessible)");
  759         }
  760         sbuf_cat(sb, "\n");
  761         sbuf_finish(sb);
  762         log(LOG_INFO, "%s", sbuf_data(sb));
  763         sbuf_delete(sb);
  764         exit1(td, 0, sig);
  765         return (0);
  766 }
  767 
  768 #ifdef COMPAT_43
  769 /*
  770  * The dirty work is handled by kern_wait().
  771  */
  772 int
  773 owait(struct thread *td, struct owait_args *uap __unused)
  774 {
  775         int error, status;
  776 
  777         error = kern_wait(td, WAIT_ANY, &status, 0, NULL);
  778         if (error == 0)
  779                 td->td_retval[1] = status;
  780         return (error);
  781 }
  782 #endif /* COMPAT_43 */
  783 
  784 /*
  785  * The dirty work is handled by kern_wait().
  786  */
  787 int
  788 sys_wait4(struct thread *td, struct wait4_args *uap)
  789 {
  790         struct rusage ru, *rup;
  791         int error, status;
  792 
  793         if (uap->rusage != NULL)
  794                 rup = &ru;
  795         else
  796                 rup = NULL;
  797         error = kern_wait(td, uap->pid, &status, uap->options, rup);
  798         if (uap->status != NULL && error == 0 && td->td_retval[0] != 0)
  799                 error = copyout(&status, uap->status, sizeof(status));
  800         if (uap->rusage != NULL && error == 0 && td->td_retval[0] != 0)
  801                 error = copyout(&ru, uap->rusage, sizeof(struct rusage));
  802         return (error);
  803 }
  804 
  805 int
  806 sys_wait6(struct thread *td, struct wait6_args *uap)
  807 {
  808         struct __wrusage wru, *wrup;
  809         siginfo_t si, *sip;
  810         idtype_t idtype;
  811         id_t id;
  812         int error, status;
  813 
  814         idtype = uap->idtype;
  815         id = uap->id;
  816 
  817         if (uap->wrusage != NULL)
  818                 wrup = &wru;
  819         else
  820                 wrup = NULL;
  821 
  822         if (uap->info != NULL) {
  823                 sip = &si;
  824                 bzero(sip, sizeof(*sip));
  825         } else
  826                 sip = NULL;
  827 
  828         /*
  829          *  We expect all callers of wait6() to know about WEXITED and
  830          *  WTRAPPED.
  831          */
  832         error = kern_wait6(td, idtype, id, &status, uap->options, wrup, sip);
  833 
  834         if (uap->status != NULL && error == 0 && td->td_retval[0] != 0)
  835                 error = copyout(&status, uap->status, sizeof(status));
  836         if (uap->wrusage != NULL && error == 0 && td->td_retval[0] != 0)
  837                 error = copyout(&wru, uap->wrusage, sizeof(wru));
  838         if (uap->info != NULL && error == 0)
  839                 error = copyout(&si, uap->info, sizeof(si));
  840         return (error);
  841 }
  842 
  843 /*
  844  * Reap the remains of a zombie process and optionally return status and
  845  * rusage.  Asserts and will release both the proctree_lock and the process
  846  * lock as part of its work.
  847  */
  848 void
  849 proc_reap(struct thread *td, struct proc *p, int *status, int options)
  850 {
  851         struct proc *q, *t;
  852 
  853         sx_assert(&proctree_lock, SA_XLOCKED);
  854         PROC_LOCK_ASSERT(p, MA_OWNED);
  855         KASSERT(p->p_state == PRS_ZOMBIE, ("proc_reap: !PRS_ZOMBIE"));
  856 
  857         mtx_spin_wait_unlocked(&p->p_slock);
  858 
  859         q = td->td_proc;
  860 
  861         if (status)
  862                 *status = KW_EXITCODE(p->p_xexit, p->p_xsig);
  863         if (options & WNOWAIT) {
  864                 /*
  865                  *  Only poll, returning the status.  Caller does not wish to
  866                  * release the proc struct just yet.
  867                  */
  868                 PROC_UNLOCK(p);
  869                 sx_xunlock(&proctree_lock);
  870                 return;
  871         }
  872 
  873         PROC_LOCK(q);
  874         sigqueue_take(p->p_ksi);
  875         PROC_UNLOCK(q);
  876 
  877         /*
  878          * If we got the child via a ptrace 'attach', we need to give it back
  879          * to the old parent.
  880          */
  881         if (p->p_oppid != p->p_pptr->p_pid) {
  882                 PROC_UNLOCK(p);
  883                 t = proc_realparent(p);
  884                 PROC_LOCK(t);
  885                 PROC_LOCK(p);
  886                 CTR2(KTR_PTRACE,
  887                     "wait: traced child %d moved back to parent %d", p->p_pid,
  888                     t->p_pid);
  889                 proc_reparent(p, t, false);
  890                 PROC_UNLOCK(p);
  891                 pksignal(t, SIGCHLD, p->p_ksi);
  892                 wakeup(t);
  893                 cv_broadcast(&p->p_pwait);
  894                 PROC_UNLOCK(t);
  895                 sx_xunlock(&proctree_lock);
  896                 return;
  897         }
  898         PROC_UNLOCK(p);
  899 
  900         /*
  901          * Remove other references to this process to ensure we have an
  902          * exclusive reference.
  903          */
  904         sx_xlock(PIDHASHLOCK(p->p_pid));
  905         LIST_REMOVE(p, p_hash);
  906         sx_xunlock(PIDHASHLOCK(p->p_pid));
  907         LIST_REMOVE(p, p_sibling);
  908         reaper_abandon_children(p, true);
  909         reaper_clear(p);
  910         PROC_LOCK(p);
  911         proc_clear_orphan(p);
  912         PROC_UNLOCK(p);
  913         leavepgrp(p);
  914         if (p->p_procdesc != NULL)
  915                 procdesc_reap(p);
  916         sx_xunlock(&proctree_lock);
  917 
  918         proc_id_clear(PROC_ID_PID, p->p_pid);
  919 
  920         PROC_LOCK(p);
  921         knlist_detach(p->p_klist);
  922         p->p_klist = NULL;
  923         PROC_UNLOCK(p);
  924 
  925         /*
  926          * Removal from allproc list and process group list paired with
  927          * PROC_LOCK which was executed during that time should guarantee
  928          * nothing can reach this process anymore. As such further locking
  929          * is unnecessary.
  930          */
  931         p->p_xexit = p->p_xsig = 0;             /* XXX: why? */
  932 
  933         PROC_LOCK(q);
  934         ruadd(&q->p_stats->p_cru, &q->p_crux, &p->p_ru, &p->p_rux);
  935         PROC_UNLOCK(q);
  936 
  937         /*
  938          * Decrement the count of procs running with this uid.
  939          */
  940         (void)chgproccnt(p->p_ucred->cr_ruidinfo, -1, 0);
  941 
  942         /*
  943          * Destroy resource accounting information associated with the process.
  944          */
  945 #ifdef RACCT
  946         if (racct_enable) {
  947                 PROC_LOCK(p);
  948                 racct_sub(p, RACCT_NPROC, 1);
  949                 PROC_UNLOCK(p);
  950         }
  951 #endif
  952         racct_proc_exit(p);
  953 
  954         /*
  955          * Free credentials, arguments, and sigacts.
  956          */
  957         proc_unset_cred(p);
  958         pargs_drop(p->p_args);
  959         p->p_args = NULL;
  960         sigacts_free(p->p_sigacts);
  961         p->p_sigacts = NULL;
  962 
  963         /*
  964          * Do any thread-system specific cleanups.
  965          */
  966         thread_wait(p);
  967 
  968         /*
  969          * Give vm and machine-dependent layer a chance to free anything that
  970          * cpu_exit couldn't release while still running in process context.
  971          */
  972         vm_waitproc(p);
  973 #ifdef MAC
  974         mac_proc_destroy(p);
  975 #endif
  976 
  977         KASSERT(FIRST_THREAD_IN_PROC(p),
  978             ("proc_reap: no residual thread!"));
  979         uma_zfree(proc_zone, p);
  980         atomic_add_int(&nprocs, -1);
  981 }
  982 
  983 static int
  984 proc_to_reap(struct thread *td, struct proc *p, idtype_t idtype, id_t id,
  985     int *status, int options, struct __wrusage *wrusage, siginfo_t *siginfo,
  986     int check_only)
  987 {
  988         struct rusage *rup;
  989 
  990         sx_assert(&proctree_lock, SA_XLOCKED);
  991 
  992         PROC_LOCK(p);
  993 
  994         switch (idtype) {
  995         case P_ALL:
  996                 if (p->p_procdesc == NULL ||
  997                    (p->p_pptr == td->td_proc &&
  998                    (p->p_flag & P_TRACED) != 0)) {
  999                         break;
 1000                 }
 1001 
 1002                 PROC_UNLOCK(p);
 1003                 return (0);
 1004         case P_PID:
 1005                 if (p->p_pid != (pid_t)id) {
 1006                         PROC_UNLOCK(p);
 1007                         return (0);
 1008                 }
 1009                 break;
 1010         case P_PGID:
 1011                 if (p->p_pgid != (pid_t)id) {
 1012                         PROC_UNLOCK(p);
 1013                         return (0);
 1014                 }
 1015                 break;
 1016         case P_SID:
 1017                 if (p->p_session->s_sid != (pid_t)id) {
 1018                         PROC_UNLOCK(p);
 1019                         return (0);
 1020                 }
 1021                 break;
 1022         case P_UID:
 1023                 if (p->p_ucred->cr_uid != (uid_t)id) {
 1024                         PROC_UNLOCK(p);
 1025                         return (0);
 1026                 }
 1027                 break;
 1028         case P_GID:
 1029                 if (p->p_ucred->cr_gid != (gid_t)id) {
 1030                         PROC_UNLOCK(p);
 1031                         return (0);
 1032                 }
 1033                 break;
 1034         case P_JAILID:
 1035                 if (p->p_ucred->cr_prison->pr_id != (int)id) {
 1036                         PROC_UNLOCK(p);
 1037                         return (0);
 1038                 }
 1039                 break;
 1040         /*
 1041          * It seems that the thread structures get zeroed out
 1042          * at process exit.  This makes it impossible to
 1043          * support P_SETID, P_CID or P_CPUID.
 1044          */
 1045         default:
 1046                 PROC_UNLOCK(p);
 1047                 return (0);
 1048         }
 1049 
 1050         if (p_canwait(td, p)) {
 1051                 PROC_UNLOCK(p);
 1052                 return (0);
 1053         }
 1054 
 1055         if (((options & WEXITED) == 0) && (p->p_state == PRS_ZOMBIE)) {
 1056                 PROC_UNLOCK(p);
 1057                 return (0);
 1058         }
 1059 
 1060         /*
 1061          * This special case handles a kthread spawned by linux_clone
 1062          * (see linux_misc.c).  The linux_wait4 and linux_waitpid
 1063          * functions need to be able to distinguish between waiting
 1064          * on a process and waiting on a thread.  It is a thread if
 1065          * p_sigparent is not SIGCHLD, and the WLINUXCLONE option
 1066          * signifies we want to wait for threads and not processes.
 1067          */
 1068         if ((p->p_sigparent != SIGCHLD) ^
 1069             ((options & WLINUXCLONE) != 0)) {
 1070                 PROC_UNLOCK(p);
 1071                 return (0);
 1072         }
 1073 
 1074         if (siginfo != NULL) {
 1075                 bzero(siginfo, sizeof(*siginfo));
 1076                 siginfo->si_errno = 0;
 1077 
 1078                 /*
 1079                  * SUSv4 requires that the si_signo value is always
 1080                  * SIGCHLD. Obey it despite the rfork(2) interface
 1081                  * allows to request other signal for child exit
 1082                  * notification.
 1083                  */
 1084                 siginfo->si_signo = SIGCHLD;
 1085 
 1086                 /*
 1087                  *  This is still a rough estimate.  We will fix the
 1088                  *  cases TRAPPED, STOPPED, and CONTINUED later.
 1089                  */
 1090                 if (WCOREDUMP(p->p_xsig)) {
 1091                         siginfo->si_code = CLD_DUMPED;
 1092                         siginfo->si_status = WTERMSIG(p->p_xsig);
 1093                 } else if (WIFSIGNALED(p->p_xsig)) {
 1094                         siginfo->si_code = CLD_KILLED;
 1095                         siginfo->si_status = WTERMSIG(p->p_xsig);
 1096                 } else {
 1097                         siginfo->si_code = CLD_EXITED;
 1098                         siginfo->si_status = p->p_xexit;
 1099                 }
 1100 
 1101                 siginfo->si_pid = p->p_pid;
 1102                 siginfo->si_uid = p->p_ucred->cr_uid;
 1103 
 1104                 /*
 1105                  * The si_addr field would be useful additional
 1106                  * detail, but apparently the PC value may be lost
 1107                  * when we reach this point.  bzero() above sets
 1108                  * siginfo->si_addr to NULL.
 1109                  */
 1110         }
 1111 
 1112         /*
 1113          * There should be no reason to limit resources usage info to
 1114          * exited processes only.  A snapshot about any resources used
 1115          * by a stopped process may be exactly what is needed.
 1116          */
 1117         if (wrusage != NULL) {
 1118                 rup = &wrusage->wru_self;
 1119                 *rup = p->p_ru;
 1120                 PROC_STATLOCK(p);
 1121                 calcru(p, &rup->ru_utime, &rup->ru_stime);
 1122                 PROC_STATUNLOCK(p);
 1123 
 1124                 rup = &wrusage->wru_children;
 1125                 *rup = p->p_stats->p_cru;
 1126                 calccru(p, &rup->ru_utime, &rup->ru_stime);
 1127         }
 1128 
 1129         if (p->p_state == PRS_ZOMBIE && !check_only) {
 1130                 proc_reap(td, p, status, options);
 1131                 return (-1);
 1132         }
 1133         return (1);
 1134 }
 1135 
 1136 int
 1137 kern_wait(struct thread *td, pid_t pid, int *status, int options,
 1138     struct rusage *rusage)
 1139 {
 1140         struct __wrusage wru, *wrup;
 1141         idtype_t idtype;
 1142         id_t id;
 1143         int ret;
 1144 
 1145         /*
 1146          * Translate the special pid values into the (idtype, pid)
 1147          * pair for kern_wait6.  The WAIT_MYPGRP case is handled by
 1148          * kern_wait6() on its own.
 1149          */
 1150         if (pid == WAIT_ANY) {
 1151                 idtype = P_ALL;
 1152                 id = 0;
 1153         } else if (pid < 0) {
 1154                 idtype = P_PGID;
 1155                 id = (id_t)-pid;
 1156         } else {
 1157                 idtype = P_PID;
 1158                 id = (id_t)pid;
 1159         }
 1160 
 1161         if (rusage != NULL)
 1162                 wrup = &wru;
 1163         else
 1164                 wrup = NULL;
 1165 
 1166         /*
 1167          * For backward compatibility we implicitly add flags WEXITED
 1168          * and WTRAPPED here.
 1169          */
 1170         options |= WEXITED | WTRAPPED;
 1171         ret = kern_wait6(td, idtype, id, status, options, wrup, NULL);
 1172         if (rusage != NULL)
 1173                 *rusage = wru.wru_self;
 1174         return (ret);
 1175 }
 1176 
 1177 static void
 1178 report_alive_proc(struct thread *td, struct proc *p, siginfo_t *siginfo,
 1179     int *status, int options, int si_code)
 1180 {
 1181         bool cont;
 1182 
 1183         PROC_LOCK_ASSERT(p, MA_OWNED);
 1184         sx_assert(&proctree_lock, SA_XLOCKED);
 1185         MPASS(si_code == CLD_TRAPPED || si_code == CLD_STOPPED ||
 1186             si_code == CLD_CONTINUED);
 1187 
 1188         cont = si_code == CLD_CONTINUED;
 1189         if ((options & WNOWAIT) == 0) {
 1190                 if (cont)
 1191                         p->p_flag &= ~P_CONTINUED;
 1192                 else
 1193                         p->p_flag |= P_WAITED;
 1194                 PROC_LOCK(td->td_proc);
 1195                 sigqueue_take(p->p_ksi);
 1196                 PROC_UNLOCK(td->td_proc);
 1197         }
 1198         sx_xunlock(&proctree_lock);
 1199         if (siginfo != NULL) {
 1200                 siginfo->si_code = si_code;
 1201                 siginfo->si_status = cont ? SIGCONT : p->p_xsig;
 1202         }
 1203         if (status != NULL)
 1204                 *status = cont ? SIGCONT : W_STOPCODE(p->p_xsig);
 1205         PROC_UNLOCK(p);
 1206         td->td_retval[0] = p->p_pid;
 1207 }
 1208 
 1209 int
 1210 kern_wait6(struct thread *td, idtype_t idtype, id_t id, int *status,
 1211     int options, struct __wrusage *wrusage, siginfo_t *siginfo)
 1212 {
 1213         struct proc *p, *q;
 1214         pid_t pid;
 1215         int error, nfound, ret;
 1216         bool report;
 1217 
 1218         AUDIT_ARG_VALUE((int)idtype);   /* XXX - This is likely wrong! */
 1219         AUDIT_ARG_PID((pid_t)id);       /* XXX - This may be wrong! */
 1220         AUDIT_ARG_VALUE(options);
 1221 
 1222         q = td->td_proc;
 1223 
 1224         if ((pid_t)id == WAIT_MYPGRP && (idtype == P_PID || idtype == P_PGID)) {
 1225                 PROC_LOCK(q);
 1226                 id = (id_t)q->p_pgid;
 1227                 PROC_UNLOCK(q);
 1228                 idtype = P_PGID;
 1229         }
 1230 
 1231         /* If we don't know the option, just return. */
 1232         if ((options & ~(WUNTRACED | WNOHANG | WCONTINUED | WNOWAIT |
 1233             WEXITED | WTRAPPED | WLINUXCLONE)) != 0)
 1234                 return (EINVAL);
 1235         if ((options & (WEXITED | WUNTRACED | WCONTINUED | WTRAPPED)) == 0) {
 1236                 /*
 1237                  * We will be unable to find any matching processes,
 1238                  * because there are no known events to look for.
 1239                  * Prefer to return error instead of blocking
 1240                  * indefinitely.
 1241                  */
 1242                 return (EINVAL);
 1243         }
 1244 
 1245 loop:
 1246         if (q->p_flag & P_STATCHILD) {
 1247                 PROC_LOCK(q);
 1248                 q->p_flag &= ~P_STATCHILD;
 1249                 PROC_UNLOCK(q);
 1250         }
 1251         sx_xlock(&proctree_lock);
 1252 loop_locked:
 1253         nfound = 0;
 1254         LIST_FOREACH(p, &q->p_children, p_sibling) {
 1255                 pid = p->p_pid;
 1256                 ret = proc_to_reap(td, p, idtype, id, status, options,
 1257                     wrusage, siginfo, 0);
 1258                 if (ret == 0)
 1259                         continue;
 1260                 else if (ret != 1) {
 1261                         td->td_retval[0] = pid;
 1262                         return (0);
 1263                 }
 1264 
 1265                 nfound++;
 1266                 PROC_LOCK_ASSERT(p, MA_OWNED);
 1267 
 1268                 if ((options & WTRAPPED) != 0 &&
 1269                     (p->p_flag & P_TRACED) != 0) {
 1270                         PROC_SLOCK(p);
 1271                         report =
 1272                             ((p->p_flag & (P_STOPPED_TRACE | P_STOPPED_SIG)) &&
 1273                             p->p_suspcount == p->p_numthreads &&
 1274                             (p->p_flag & P_WAITED) == 0);
 1275                         PROC_SUNLOCK(p);
 1276                         if (report) {
 1277                         CTR4(KTR_PTRACE,
 1278                             "wait: returning trapped pid %d status %#x "
 1279                             "(xstat %d) xthread %d",
 1280                             p->p_pid, W_STOPCODE(p->p_xsig), p->p_xsig,
 1281                             p->p_xthread != NULL ?
 1282                             p->p_xthread->td_tid : -1);
 1283                                 report_alive_proc(td, p, siginfo, status,
 1284                                     options, CLD_TRAPPED);
 1285                                 return (0);
 1286                         }
 1287                 }
 1288                 if ((options & WUNTRACED) != 0 &&
 1289                     (p->p_flag & P_STOPPED_SIG) != 0) {
 1290                         PROC_SLOCK(p);
 1291                         report = (p->p_suspcount == p->p_numthreads &&
 1292                             ((p->p_flag & P_WAITED) == 0));
 1293                         PROC_SUNLOCK(p);
 1294                         if (report) {
 1295                                 report_alive_proc(td, p, siginfo, status,
 1296                                     options, CLD_STOPPED);
 1297                                 return (0);
 1298                         }
 1299                 }
 1300                 if ((options & WCONTINUED) != 0 &&
 1301                     (p->p_flag & P_CONTINUED) != 0) {
 1302                         report_alive_proc(td, p, siginfo, status, options,
 1303                             CLD_CONTINUED);
 1304                         return (0);
 1305                 }
 1306                 PROC_UNLOCK(p);
 1307         }
 1308 
 1309         /*
 1310          * Look in the orphans list too, to allow the parent to
 1311          * collect it's child exit status even if child is being
 1312          * debugged.
 1313          *
 1314          * Debugger detaches from the parent upon successful
 1315          * switch-over from parent to child.  At this point due to
 1316          * re-parenting the parent loses the child to debugger and a
 1317          * wait4(2) call would report that it has no children to wait
 1318          * for.  By maintaining a list of orphans we allow the parent
 1319          * to successfully wait until the child becomes a zombie.
 1320          */
 1321         if (nfound == 0) {
 1322                 LIST_FOREACH(p, &q->p_orphans, p_orphan) {
 1323                         ret = proc_to_reap(td, p, idtype, id, NULL, options,
 1324                             NULL, NULL, 1);
 1325                         if (ret != 0) {
 1326                                 KASSERT(ret != -1, ("reaped an orphan (pid %d)",
 1327                                     (int)td->td_retval[0]));
 1328                                 PROC_UNLOCK(p);
 1329                                 nfound++;
 1330                                 break;
 1331                         }
 1332                 }
 1333         }
 1334         if (nfound == 0) {
 1335                 sx_xunlock(&proctree_lock);
 1336                 return (ECHILD);
 1337         }
 1338         if (options & WNOHANG) {
 1339                 sx_xunlock(&proctree_lock);
 1340                 td->td_retval[0] = 0;
 1341                 return (0);
 1342         }
 1343         PROC_LOCK(q);
 1344         if (q->p_flag & P_STATCHILD) {
 1345                 q->p_flag &= ~P_STATCHILD;
 1346                 PROC_UNLOCK(q);
 1347                 goto loop_locked;
 1348         }
 1349         sx_xunlock(&proctree_lock);
 1350         error = msleep(q, &q->p_mtx, PWAIT | PCATCH | PDROP, "wait", 0);
 1351         if (error)
 1352                 return (error);
 1353         goto loop;
 1354 }
 1355 
 1356 void
 1357 proc_add_orphan(struct proc *child, struct proc *parent)
 1358 {
 1359 
 1360         sx_assert(&proctree_lock, SX_XLOCKED);
 1361         KASSERT((child->p_flag & P_TRACED) != 0,
 1362             ("proc_add_orphan: not traced"));
 1363 
 1364         if (LIST_EMPTY(&parent->p_orphans)) {
 1365                 child->p_treeflag |= P_TREE_FIRST_ORPHAN;
 1366                 LIST_INSERT_HEAD(&parent->p_orphans, child, p_orphan);
 1367         } else {
 1368                 LIST_INSERT_AFTER(LIST_FIRST(&parent->p_orphans),
 1369                     child, p_orphan);
 1370         }
 1371         child->p_treeflag |= P_TREE_ORPHANED;
 1372 }
 1373 
 1374 /*
 1375  * Make process 'parent' the new parent of process 'child'.
 1376  * Must be called with an exclusive hold of proctree lock.
 1377  */
 1378 void
 1379 proc_reparent(struct proc *child, struct proc *parent, bool set_oppid)
 1380 {
 1381 
 1382         sx_assert(&proctree_lock, SX_XLOCKED);
 1383         PROC_LOCK_ASSERT(child, MA_OWNED);
 1384         if (child->p_pptr == parent)
 1385                 return;
 1386 
 1387         PROC_LOCK(child->p_pptr);
 1388         sigqueue_take(child->p_ksi);
 1389         PROC_UNLOCK(child->p_pptr);
 1390         LIST_REMOVE(child, p_sibling);
 1391         LIST_INSERT_HEAD(&parent->p_children, child, p_sibling);
 1392 
 1393         proc_clear_orphan(child);
 1394         if ((child->p_flag & P_TRACED) != 0) {
 1395                 proc_add_orphan(child, child->p_pptr);
 1396         }
 1397 
 1398         child->p_pptr = parent;
 1399         if (set_oppid)
 1400                 child->p_oppid = parent->p_pid;
 1401 }

Cache object: 6d67371ceb6c05b106b44a907bacb296


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]


This page is part of the FreeBSD/Linux Linux Kernel Cross-Reference, and was automatically generated using a modified version of the LXR engine.