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_fork.c

Version: -  FREEBSD  -  FREEBSD-12-STABLE  -  FREEBSD-12-0  -  FREEBSD-11-STABLE  -  FREEBSD-11-2  -  FREEBSD-11-1  -  FREEBSD-11-0  -  FREEBSD-10-STABLE  -  FREEBSD-10-4  -  FREEBSD-10-3  -  FREEBSD-10-2  -  FREEBSD-10-1  -  FREEBSD-10-0  -  FREEBSD-9-STABLE  -  FREEBSD-9-3  -  FREEBSD-9-2  -  FREEBSD-9-1  -  FREEBSD-9-0  -  FREEBSD-8-STABLE  -  FREEBSD-8-4  -  FREEBSD-8-3  -  FREEBSD-8-2  -  FREEBSD-8-1  -  FREEBSD-8-0  -  FREEBSD-7-STABLE  -  FREEBSD-7-4  -  FREEBSD-7-3  -  FREEBSD-7-2  -  FREEBSD-7-1  -  FREEBSD-7-0  -  FREEBSD-6-STABLE  -  FREEBSD-6-4  -  FREEBSD-6-3  -  FREEBSD-6-2  -  FREEBSD-6-1  -  FREEBSD-6-0  -  FREEBSD-5-STABLE  -  FREEBSD-5-5  -  FREEBSD-5-4  -  FREEBSD-5-3  -  FREEBSD-5-2  -  FREEBSD-5-1  -  FREEBSD-5-0  -  FREEBSD-4-STABLE  -  FREEBSD-3-STABLE  -  FREEBSD22  -  linux-2.6  -  linux-2.4.22  -  MK83  -  MK84  -  PLAN9  -  DFBSD  -  NETBSD  -  NETBSD5  -  NETBSD4  -  NETBSD3  -  NETBSD20  -  OPENBSD  -  xnu-517  -  xnu-792  -  xnu-792.6.70  -  xnu-1228  -  xnu-1456.1.26  -  xnu-1699.24.8  -  xnu-2050.18.24  -  OPENSOLARIS  -  minix-3-1-1 
SearchContext: -  none  -  3  -  10 

    1 /*-
    2  * Copyright (c) 1982, 1986, 1989, 1991, 1993
    3  *      The Regents of the University of California.  All rights reserved.
    4  * (c) UNIX System Laboratories, Inc.
    5  * All or some portions of this file are derived from material licensed
    6  * to the University of California by American Telephone and Telegraph
    7  * Co. or Unix System Laboratories, Inc. and are reproduced herein with
    8  * the permission of UNIX System Laboratories, Inc.
    9  *
   10  * Redistribution and use in source and binary forms, with or without
   11  * modification, are permitted provided that the following conditions
   12  * are met:
   13  * 1. Redistributions of source code must retain the above copyright
   14  *    notice, this list of conditions and the following disclaimer.
   15  * 2. Redistributions in binary form must reproduce the above copyright
   16  *    notice, this list of conditions and the following disclaimer in the
   17  *    documentation and/or other materials provided with the distribution.
   18  * 4. Neither the name of the University nor the names of its contributors
   19  *    may be used to endorse or promote products derived from this software
   20  *    without specific prior written permission.
   21  *
   22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
   23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
   24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
   25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
   26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
   27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
   28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
   29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
   30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
   31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   32  * SUCH DAMAGE.
   33  *
   34  *      @(#)kern_fork.c 8.6 (Berkeley) 4/8/94
   35  */
   36 
   37 #include <sys/cdefs.h>
   38 __FBSDID("$FreeBSD: releng/7.3/sys/kern/kern_fork.c 202814 2010-01-22 17:02:07Z jhb $");
   39 
   40 #include "opt_kdtrace.h"
   41 #include "opt_ktrace.h"
   42 #include "opt_mac.h"
   43 
   44 #include <sys/param.h>
   45 #include <sys/systm.h>
   46 #include <sys/sysproto.h>
   47 #include <sys/eventhandler.h>
   48 #include <sys/filedesc.h>
   49 #include <sys/kernel.h>
   50 #include <sys/kthread.h>
   51 #include <sys/sysctl.h>
   52 #include <sys/lock.h>
   53 #include <sys/malloc.h>
   54 #include <sys/mutex.h>
   55 #include <sys/priv.h>
   56 #include <sys/proc.h>
   57 #include <sys/jail.h>
   58 #include <sys/pioctl.h>
   59 #include <sys/resourcevar.h>
   60 #include <sys/sched.h>
   61 #include <sys/syscall.h>
   62 #include <sys/vmmeter.h>
   63 #include <sys/vnode.h>
   64 #include <sys/acct.h>
   65 #include <sys/ktr.h>
   66 #include <sys/ktrace.h>
   67 #include <sys/unistd.h> 
   68 #include <sys/sdt.h>
   69 #include <sys/sx.h>
   70 #include <sys/signalvar.h>
   71 
   72 #include <security/audit/audit.h>
   73 #include <security/mac/mac_framework.h>
   74 
   75 #include <vm/vm.h>
   76 #include <vm/pmap.h>
   77 #include <vm/vm_map.h>
   78 #include <vm/vm_extern.h>
   79 #include <vm/uma.h>
   80 
   81 #ifdef KDTRACE_HOOKS
   82 #include <sys/dtrace_bsd.h>
   83 dtrace_fork_func_t      dtrace_fasttrap_fork;
   84 #endif
   85 
   86 SDT_PROVIDER_DECLARE(proc);
   87 SDT_PROBE_DEFINE(proc, kernel, , create);
   88 SDT_PROBE_ARGTYPE(proc, kernel, , create, 0, "struct proc *");
   89 SDT_PROBE_ARGTYPE(proc, kernel, , create, 1, "struct proc *");
   90 SDT_PROBE_ARGTYPE(proc, kernel, , create, 2, "int");
   91 
   92 #ifndef _SYS_SYSPROTO_H_
   93 struct fork_args {
   94         int     dummy;
   95 };
   96 #endif
   97 
   98 /* ARGSUSED */
   99 int
  100 fork(td, uap)
  101         struct thread *td;
  102         struct fork_args *uap;
  103 {
  104         int error;
  105         struct proc *p2;
  106 
  107         error = fork1(td, RFFDG | RFPROC, 0, &p2);
  108         if (error == 0) {
  109                 td->td_retval[0] = p2->p_pid;
  110                 td->td_retval[1] = 0;
  111         }
  112         return (error);
  113 }
  114 
  115 /* ARGSUSED */
  116 int
  117 vfork(td, uap)
  118         struct thread *td;
  119         struct vfork_args *uap;
  120 {
  121         int error;
  122         struct proc *p2;
  123 
  124         error = fork1(td, RFFDG | RFPROC | RFPPWAIT | RFMEM, 0, &p2);
  125         if (error == 0) {
  126                 td->td_retval[0] = p2->p_pid;
  127                 td->td_retval[1] = 0;
  128         }
  129         return (error);
  130 }
  131 
  132 int
  133 rfork(td, uap)
  134         struct thread *td;
  135         struct rfork_args *uap;
  136 {
  137         struct proc *p2;
  138         int error;
  139 
  140         /* Don't allow kernel-only flags. */
  141         if ((uap->flags & RFKERNELONLY) != 0)
  142                 return (EINVAL);
  143 
  144         AUDIT_ARG(fflags, uap->flags);
  145         error = fork1(td, uap->flags, 0, &p2);
  146         if (error == 0) {
  147                 td->td_retval[0] = p2 ? p2->p_pid : 0;
  148                 td->td_retval[1] = 0;
  149         }
  150         return (error);
  151 }
  152 
  153 int     nprocs = 1;             /* process 0 */
  154 int     lastpid = 0;
  155 SYSCTL_INT(_kern, OID_AUTO, lastpid, CTLFLAG_RD, &lastpid, 0, 
  156     "Last used PID");
  157 
  158 /*
  159  * Random component to lastpid generation.  We mix in a random factor to make
  160  * it a little harder to predict.  We sanity check the modulus value to avoid
  161  * doing it in critical paths.  Don't let it be too small or we pointlessly
  162  * waste randomness entropy, and don't let it be impossibly large.  Using a
  163  * modulus that is too big causes a LOT more process table scans and slows
  164  * down fork processing as the pidchecked caching is defeated.
  165  */
  166 static int randompid = 0;
  167 
  168 static int
  169 sysctl_kern_randompid(SYSCTL_HANDLER_ARGS)
  170 {
  171         int error, pid;
  172 
  173         error = sysctl_wire_old_buffer(req, sizeof(int));
  174         if (error != 0)
  175                 return(error);
  176         sx_xlock(&allproc_lock);
  177         pid = randompid;
  178         error = sysctl_handle_int(oidp, &pid, 0, req);
  179         if (error == 0 && req->newptr != NULL) {
  180                 if (pid < 0 || pid > PID_MAX - 100)     /* out of range */
  181                         pid = PID_MAX - 100;
  182                 else if (pid < 2)                       /* NOP */
  183                         pid = 0;
  184                 else if (pid < 100)                     /* Make it reasonable */
  185                         pid = 100;
  186                 randompid = pid;
  187         }
  188         sx_xunlock(&allproc_lock);
  189         return (error);
  190 }
  191 
  192 SYSCTL_PROC(_kern, OID_AUTO, randompid, CTLTYPE_INT|CTLFLAG_RW,
  193     0, 0, sysctl_kern_randompid, "I", "Random PID modulus");
  194 
  195 int
  196 fork1(td, flags, pages, procp)
  197         struct thread *td;
  198         int flags;
  199         int pages;
  200         struct proc **procp;
  201 {
  202         struct proc *p1, *p2, *pptr;
  203         struct proc *newproc;
  204         int ok, trypid;
  205         static int curfail, pidchecked = 0;
  206         static struct timeval lastfail;
  207         struct filedesc *fd;
  208         struct filedesc_to_leader *fdtol;
  209         struct thread *td2;
  210         struct sigacts *newsigacts;
  211         struct vmspace *vm2;
  212         int error;
  213 
  214         /* Can't copy and clear. */
  215         if ((flags & (RFFDG|RFCFDG)) == (RFFDG|RFCFDG))
  216                 return (EINVAL);
  217 
  218         p1 = td->td_proc;
  219 
  220         /*
  221          * Here we don't create a new process, but we divorce
  222          * certain parts of a process from itself.
  223          */
  224         if ((flags & RFPROC) == 0) {
  225                 if ((p1->p_flag & P_HADTHREADS) &&
  226                     (flags & (RFCFDG | RFFDG))) {
  227                         PROC_LOCK(p1);
  228                         if (thread_single(SINGLE_BOUNDARY)) {
  229                                 PROC_UNLOCK(p1);
  230                                 return (ERESTART);
  231                         }
  232                         PROC_UNLOCK(p1);
  233                 }
  234 
  235                 error = vm_forkproc(td, NULL, NULL, NULL, flags);
  236                 if (error)
  237                         goto norfproc_fail;
  238 
  239                 /*
  240                  * Close all file descriptors.
  241                  */
  242                 if (flags & RFCFDG) {
  243                         struct filedesc *fdtmp;
  244                         fdtmp = fdinit(td->td_proc->p_fd);
  245                         fdfree(td);
  246                         p1->p_fd = fdtmp;
  247                 }
  248 
  249                 /*
  250                  * Unshare file descriptors (from parent).
  251                  */
  252                 if (flags & RFFDG) 
  253                         fdunshare(p1, td);
  254 
  255 norfproc_fail:
  256                 if ((p1->p_flag & P_HADTHREADS) &&
  257                     (flags & (RFCFDG | RFFDG))) {
  258                         PROC_LOCK(p1);
  259                         thread_single_end();
  260                         PROC_UNLOCK(p1);
  261                 }
  262                 *procp = NULL;
  263                 return (error);
  264         }
  265 
  266         vm2 = NULL;
  267         /* Allocate new proc. */
  268         newproc = uma_zalloc(proc_zone, M_WAITOK);
  269         if (TAILQ_EMPTY(&newproc->p_threads)) {
  270                 td2 = thread_alloc();
  271                 if (td2 == NULL) {
  272                         error = ENOMEM;
  273                         goto fail1;
  274                 }
  275                 proc_linkup(newproc, td2);
  276                 sched_newproc(newproc, td2);
  277         } else
  278                 td2 = FIRST_THREAD_IN_PROC(newproc);
  279 
  280         /* Allocate and switch to an alternate kstack if specified. */
  281         if (pages != 0) {
  282                 if (!vm_thread_new_altkstack(td2, pages)) {
  283                         error = ENOMEM;
  284                         goto fail1;
  285                 }
  286         }
  287         if ((flags & RFMEM) == 0) {
  288                 vm2 = vmspace_fork(p1->p_vmspace);
  289                 if (vm2 == NULL) {
  290                         error = ENOMEM;
  291                         goto fail1;
  292                 }
  293         }
  294 #ifdef MAC
  295         mac_init_proc(newproc);
  296 #endif
  297         knlist_init_mtx(&newproc->p_klist, &newproc->p_mtx);
  298         STAILQ_INIT(&newproc->p_ktr);
  299 
  300         /* We have to lock the process tree while we look for a pid. */
  301         sx_slock(&proctree_lock);
  302 
  303         /*
  304          * Although process entries are dynamically created, we still keep
  305          * a global limit on the maximum number we will create.  Don't allow
  306          * a nonprivileged user to use the last ten processes; don't let root
  307          * exceed the limit. The variable nprocs is the current number of
  308          * processes, maxproc is the limit.
  309          */
  310         sx_xlock(&allproc_lock);
  311         if ((nprocs >= maxproc - 10 && priv_check_cred(td->td_ucred,
  312             PRIV_MAXPROC, 0) != 0) || nprocs >= maxproc) {
  313                 error = EAGAIN;
  314                 goto fail;
  315         }
  316 
  317         /*
  318          * Increment the count of procs running with this uid. Don't allow
  319          * a nonprivileged user to exceed their current limit.
  320          *
  321          * XXXRW: Can we avoid privilege here if it's not needed?
  322          */
  323         error = priv_check_cred(td->td_ucred, PRIV_PROC_LIMIT, 0);
  324         if (error == 0)
  325                 ok = chgproccnt(td->td_ucred->cr_ruidinfo, 1, 0);
  326         else {
  327                 PROC_LOCK(p1);
  328                 ok = chgproccnt(td->td_ucred->cr_ruidinfo, 1,
  329                     lim_cur(p1, RLIMIT_NPROC));
  330                 PROC_UNLOCK(p1);
  331         }
  332         if (!ok) {
  333                 error = EAGAIN;
  334                 goto fail;
  335         }
  336 
  337         /*
  338          * Increment the nprocs resource before blocking can occur.  There
  339          * are hard-limits as to the number of processes that can run.
  340          */
  341         nprocs++;
  342 
  343         /*
  344          * Find an unused process ID.  We remember a range of unused IDs
  345          * ready to use (from lastpid+1 through pidchecked-1).
  346          *
  347          * If RFHIGHPID is set (used during system boot), do not allocate
  348          * low-numbered pids.
  349          */
  350         trypid = lastpid + 1;
  351         if (flags & RFHIGHPID) {
  352                 if (trypid < 10)
  353                         trypid = 10;
  354         } else {
  355                 if (randompid)
  356                         trypid += arc4random() % randompid;
  357         }
  358 retry:
  359         /*
  360          * If the process ID prototype has wrapped around,
  361          * restart somewhat above 0, as the low-numbered procs
  362          * tend to include daemons that don't exit.
  363          */
  364         if (trypid >= PID_MAX) {
  365                 trypid = trypid % PID_MAX;
  366                 if (trypid < 100)
  367                         trypid += 100;
  368                 pidchecked = 0;
  369         }
  370         if (trypid >= pidchecked) {
  371                 int doingzomb = 0;
  372 
  373                 pidchecked = PID_MAX;
  374                 /*
  375                  * Scan the active and zombie procs to check whether this pid
  376                  * is in use.  Remember the lowest pid that's greater
  377                  * than trypid, so we can avoid checking for a while.
  378                  */
  379                 p2 = LIST_FIRST(&allproc);
  380 again:
  381                 for (; p2 != NULL; p2 = LIST_NEXT(p2, p_list)) {
  382                         while (p2->p_pid == trypid ||
  383                             (p2->p_pgrp != NULL &&
  384                             (p2->p_pgrp->pg_id == trypid ||
  385                             (p2->p_session != NULL &&
  386                             p2->p_session->s_sid == trypid)))) {
  387                                 trypid++;
  388                                 if (trypid >= pidchecked)
  389                                         goto retry;
  390                         }
  391                         if (p2->p_pid > trypid && pidchecked > p2->p_pid)
  392                                 pidchecked = p2->p_pid;
  393                         if (p2->p_pgrp != NULL) {
  394                                 if (p2->p_pgrp->pg_id > trypid &&
  395                                     pidchecked > p2->p_pgrp->pg_id)
  396                                         pidchecked = p2->p_pgrp->pg_id;
  397                                 if (p2->p_session != NULL &&
  398                                     p2->p_session->s_sid > trypid &&
  399                                     pidchecked > p2->p_session->s_sid)
  400                                         pidchecked = p2->p_session->s_sid;
  401                         }
  402                 }
  403                 if (!doingzomb) {
  404                         doingzomb = 1;
  405                         p2 = LIST_FIRST(&zombproc);
  406                         goto again;
  407                 }
  408         }
  409         sx_sunlock(&proctree_lock);
  410 
  411         /*
  412          * RFHIGHPID does not mess with the lastpid counter during boot.
  413          */
  414         if (flags & RFHIGHPID)
  415                 pidchecked = 0;
  416         else
  417                 lastpid = trypid;
  418 
  419         p2 = newproc;
  420         p2->p_state = PRS_NEW;          /* protect against others */
  421         p2->p_pid = trypid;
  422         /*
  423          * Allow the scheduler to initialize the child.
  424          */
  425         thread_lock(td);
  426         sched_fork(td, td2);
  427         thread_unlock(td);
  428         AUDIT_ARG(pid, p2->p_pid);
  429         LIST_INSERT_HEAD(&allproc, p2, p_list);
  430         LIST_INSERT_HEAD(PIDHASH(p2->p_pid), p2, p_hash);
  431 
  432         PROC_LOCK(p2);
  433         PROC_LOCK(p1);
  434 
  435         sx_xunlock(&allproc_lock);
  436 
  437         bcopy(&p1->p_startcopy, &p2->p_startcopy,
  438             __rangeof(struct proc, p_startcopy, p_endcopy));
  439         pargs_hold(p2->p_args);
  440         PROC_UNLOCK(p1);
  441 
  442         bzero(&p2->p_startzero,
  443             __rangeof(struct proc, p_startzero, p_endzero));
  444 
  445         p2->p_ucred = crhold(td->td_ucred);
  446 
  447         /* In case we are jailed tell the prison that we exist. */
  448         if (jailed(p2->p_ucred))
  449                 prison_proc_hold(p2->p_ucred->cr_prison);
  450 
  451         PROC_UNLOCK(p2);
  452 
  453         /*
  454          * Malloc things while we don't hold any locks.
  455          */
  456         if (flags & RFSIGSHARE)
  457                 newsigacts = NULL;
  458         else
  459                 newsigacts = sigacts_alloc();
  460 
  461         /*
  462          * Copy filedesc.
  463          */
  464         if (flags & RFCFDG) {
  465                 fd = fdinit(p1->p_fd);
  466                 fdtol = NULL;
  467         } else if (flags & RFFDG) {
  468                 fd = fdcopy(p1->p_fd);
  469                 fdtol = NULL;
  470         } else {
  471                 fd = fdshare(p1->p_fd);
  472                 if (p1->p_fdtol == NULL)
  473                         p1->p_fdtol =
  474                                 filedesc_to_leader_alloc(NULL,
  475                                                          NULL,
  476                                                          p1->p_leader);
  477                 if ((flags & RFTHREAD) != 0) {
  478                         /*
  479                          * Shared file descriptor table and
  480                          * shared process leaders.
  481                          */
  482                         fdtol = p1->p_fdtol;
  483                         FILEDESC_XLOCK(p1->p_fd);
  484                         fdtol->fdl_refcount++;
  485                         FILEDESC_XUNLOCK(p1->p_fd);
  486                 } else {
  487                         /* 
  488                          * Shared file descriptor table, and
  489                          * different process leaders 
  490                          */
  491                         fdtol = filedesc_to_leader_alloc(p1->p_fdtol,
  492                                                          p1->p_fd,
  493                                                          p2);
  494                 }
  495         }
  496         /*
  497          * Make a proc table entry for the new process.
  498          * Start by zeroing the section of proc that is zero-initialized,
  499          * then copy the section that is copied directly from the parent.
  500          */
  501 
  502         PROC_LOCK(p2);
  503         PROC_LOCK(p1);
  504 
  505         bzero(&td2->td_startzero,
  506             __rangeof(struct thread, td_startzero, td_endzero));
  507 
  508         bcopy(&td->td_startcopy, &td2->td_startcopy,
  509             __rangeof(struct thread, td_startcopy, td_endcopy));
  510         bzero(&td2->td_osd, sizeof(struct osd));
  511 
  512         td2->td_sigstk = td->td_sigstk;
  513         td2->td_sigmask = td->td_sigmask;
  514         td2->td_flags = TDF_INMEM;
  515 
  516         /*
  517          * Duplicate sub-structures as needed.
  518          * Increase reference counts on shared objects.
  519          */
  520         p2->p_flag = P_INMEM;
  521         p2->p_swtick = ticks;
  522         if (p1->p_flag & P_PROFIL)
  523                 startprofclock(p2);
  524         td2->td_ucred = crhold(p2->p_ucred);
  525 
  526         if (flags & RFSIGSHARE) {
  527                 p2->p_sigacts = sigacts_hold(p1->p_sigacts);
  528         } else {
  529                 sigacts_copy(newsigacts, p1->p_sigacts);
  530                 p2->p_sigacts = newsigacts;
  531         }
  532         if (flags & RFLINUXTHPN) 
  533                 p2->p_sigparent = SIGUSR1;
  534         else
  535                 p2->p_sigparent = SIGCHLD;
  536 
  537         p2->p_textvp = p1->p_textvp;
  538         p2->p_fd = fd;
  539         p2->p_fdtol = fdtol;
  540 
  541         /*
  542          * p_limit is copy-on-write.  Bump its refcount.
  543          */
  544         lim_fork(p1, p2);
  545 
  546         pstats_fork(p1->p_stats, p2->p_stats);
  547 
  548         PROC_UNLOCK(p1);
  549         PROC_UNLOCK(p2);
  550 
  551         /* Bump references to the text vnode (for procfs) */
  552         if (p2->p_textvp)
  553                 vref(p2->p_textvp);
  554 
  555         /*
  556          * Set up linkage for kernel based threading.
  557          */
  558         if ((flags & RFTHREAD) != 0) {
  559                 mtx_lock(&ppeers_lock);
  560                 p2->p_peers = p1->p_peers;
  561                 p1->p_peers = p2;
  562                 p2->p_leader = p1->p_leader;
  563                 mtx_unlock(&ppeers_lock);
  564                 PROC_LOCK(p1->p_leader);
  565                 if ((p1->p_leader->p_flag & P_WEXIT) != 0) {
  566                         PROC_UNLOCK(p1->p_leader);
  567                         /*
  568                          * The task leader is exiting, so process p1 is
  569                          * going to be killed shortly.  Since p1 obviously
  570                          * isn't dead yet, we know that the leader is either
  571                          * sending SIGKILL's to all the processes in this
  572                          * task or is sleeping waiting for all the peers to
  573                          * exit.  We let p1 complete the fork, but we need
  574                          * to go ahead and kill the new process p2 since
  575                          * the task leader may not get a chance to send
  576                          * SIGKILL to it.  We leave it on the list so that
  577                          * the task leader will wait for this new process
  578                          * to commit suicide.
  579                          */
  580                         PROC_LOCK(p2);
  581                         psignal(p2, SIGKILL);
  582                         PROC_UNLOCK(p2);
  583                 } else
  584                         PROC_UNLOCK(p1->p_leader);
  585         } else {
  586                 p2->p_peers = NULL;
  587                 p2->p_leader = p2;
  588         }
  589 
  590         sx_xlock(&proctree_lock);
  591         PGRP_LOCK(p1->p_pgrp);
  592         PROC_LOCK(p2);
  593         PROC_LOCK(p1);
  594 
  595         /*
  596          * Preserve some more flags in subprocess.  P_PROFIL has already
  597          * been preserved.
  598          */
  599         p2->p_flag |= p1->p_flag & P_SUGID;
  600         td2->td_pflags |= td->td_pflags & TDP_ALTSTACK;
  601         SESS_LOCK(p1->p_session);
  602         if (p1->p_session->s_ttyvp != NULL && p1->p_flag & P_CONTROLT)
  603                 p2->p_flag |= P_CONTROLT;
  604         SESS_UNLOCK(p1->p_session);
  605         if (flags & RFPPWAIT)
  606                 p2->p_flag |= P_PPWAIT;
  607 
  608         p2->p_pgrp = p1->p_pgrp;
  609         LIST_INSERT_AFTER(p1, p2, p_pglist);
  610         PGRP_UNLOCK(p1->p_pgrp);
  611         LIST_INIT(&p2->p_children);
  612 
  613         callout_init(&p2->p_itcallout, CALLOUT_MPSAFE);
  614 
  615 #ifdef KTRACE
  616         /*
  617          * Copy traceflag and tracefile if enabled.
  618          */
  619         mtx_lock(&ktrace_mtx);
  620         KASSERT(p2->p_tracevp == NULL, ("new process has a ktrace vnode"));
  621         if (p1->p_traceflag & KTRFAC_INHERIT) {
  622                 p2->p_traceflag = p1->p_traceflag;
  623                 if ((p2->p_tracevp = p1->p_tracevp) != NULL) {
  624                         VREF(p2->p_tracevp);
  625                         KASSERT(p1->p_tracecred != NULL,
  626                             ("ktrace vnode with no cred"));
  627                         p2->p_tracecred = crhold(p1->p_tracecred);
  628                 }
  629         }
  630         mtx_unlock(&ktrace_mtx);
  631 #endif
  632 
  633         /*
  634          * If PF_FORK is set, the child process inherits the
  635          * procfs ioctl flags from its parent.
  636          */
  637         if (p1->p_pfsflags & PF_FORK) {
  638                 p2->p_stops = p1->p_stops;
  639                 p2->p_pfsflags = p1->p_pfsflags;
  640         }
  641 
  642 #ifdef KDTRACE_HOOKS
  643         /*
  644          * Tell the DTrace fasttrap provider about the new process
  645          * if it has registered an interest.
  646          */
  647         if (dtrace_fasttrap_fork)
  648                 dtrace_fasttrap_fork(p1, p2);
  649 #endif
  650 
  651         /*
  652          * This begins the section where we must prevent the parent
  653          * from being swapped.
  654          */
  655         _PHOLD(p1);
  656         PROC_UNLOCK(p1);
  657 
  658         /*
  659          * Attach the new process to its parent.
  660          *
  661          * If RFNOWAIT is set, the newly created process becomes a child
  662          * of init.  This effectively disassociates the child from the
  663          * parent.
  664          */
  665         if (flags & RFNOWAIT)
  666                 pptr = initproc;
  667         else
  668                 pptr = p1;
  669         p2->p_pptr = pptr;
  670         LIST_INSERT_HEAD(&pptr->p_children, p2, p_sibling);
  671         sx_xunlock(&proctree_lock);
  672 
  673         /* Inform accounting that we have forked. */
  674         p2->p_acflag = AFORK;
  675         PROC_UNLOCK(p2);
  676 
  677         /*
  678          * Finish creating the child process.  It will return via a different
  679          * execution path later.  (ie: directly into user mode)
  680          */
  681         vm_forkproc(td, p2, td2, vm2, flags);
  682 
  683         if (flags == (RFFDG | RFPROC)) {
  684                 PCPU_INC(cnt.v_forks);
  685                 PCPU_ADD(cnt.v_forkpages, p2->p_vmspace->vm_dsize +
  686                     p2->p_vmspace->vm_ssize);
  687         } else if (flags == (RFFDG | RFPROC | RFPPWAIT | RFMEM)) {
  688                 PCPU_INC(cnt.v_vforks);
  689                 PCPU_ADD(cnt.v_vforkpages, p2->p_vmspace->vm_dsize +
  690                     p2->p_vmspace->vm_ssize);
  691         } else if (p1 == &proc0) {
  692                 PCPU_INC(cnt.v_kthreads);
  693                 PCPU_ADD(cnt.v_kthreadpages, p2->p_vmspace->vm_dsize +
  694                     p2->p_vmspace->vm_ssize);
  695         } else {
  696                 PCPU_INC(cnt.v_rforks);
  697                 PCPU_ADD(cnt.v_rforkpages, p2->p_vmspace->vm_dsize +
  698                     p2->p_vmspace->vm_ssize);
  699         }
  700 
  701         /*
  702          * Both processes are set up, now check if any loadable modules want
  703          * to adjust anything.
  704          *   What if they have an error? XXX
  705          */
  706         EVENTHANDLER_INVOKE(process_fork, p1, p2, flags);
  707 
  708         /*
  709          * Set the child start time and mark the process as being complete.
  710          */
  711         microuptime(&p2->p_stats->p_start);
  712         PROC_SLOCK(p2);
  713         p2->p_state = PRS_NORMAL;
  714         PROC_SUNLOCK(p2);
  715 
  716         /*
  717          * If RFSTOPPED not requested, make child runnable and add to
  718          * run queue.
  719          */
  720         if ((flags & RFSTOPPED) == 0) {
  721                 thread_lock(td2);
  722                 TD_SET_CAN_RUN(td2);
  723                 sched_add(td2, SRQ_BORING);
  724                 thread_unlock(td2);
  725         }
  726 
  727         /*
  728          * Now can be swapped.
  729          */
  730         PROC_LOCK(p1);
  731         _PRELE(p1);
  732         PROC_UNLOCK(p1);
  733 
  734         /*
  735          * Tell any interested parties about the new process.
  736          */
  737         knote_fork(&p1->p_klist, p2->p_pid);
  738         SDT_PROBE(proc, kernel, , create, p2, p1, flags, 0, 0);
  739 
  740         /*
  741          * Preserve synchronization semantics of vfork.  If waiting for
  742          * child to exec or exit, set P_PPWAIT on child, and sleep on our
  743          * proc (in case of exit).
  744          */
  745         PROC_LOCK(p2);
  746         while (p2->p_flag & P_PPWAIT)
  747                 cv_wait(&p2->p_pwait, &p2->p_mtx);
  748         PROC_UNLOCK(p2);
  749 
  750         /*
  751          * Return child proc pointer to parent.
  752          */
  753         *procp = p2;
  754         return (0);
  755 fail:
  756         sx_sunlock(&proctree_lock);
  757         if (ppsratecheck(&lastfail, &curfail, 1))
  758                 printf("maxproc limit exceeded by uid %i, please see tuning(7) and login.conf(5).\n",
  759                     td->td_ucred->cr_ruid);
  760         sx_xunlock(&allproc_lock);
  761 #ifdef MAC
  762         mac_destroy_proc(newproc);
  763 #endif
  764 fail1:
  765         if (vm2 != NULL)
  766                 vmspace_free(vm2);
  767         uma_zfree(proc_zone, newproc);
  768         pause("fork", hz / 2);
  769         return (error);
  770 }
  771 
  772 /*
  773  * Handle the return of a child process from fork1().  This function
  774  * is called from the MD fork_trampoline() entry point.
  775  */
  776 void
  777 fork_exit(callout, arg, frame)
  778         void (*callout)(void *, struct trapframe *);
  779         void *arg;
  780         struct trapframe *frame;
  781 {
  782         struct proc *p;
  783         struct thread *td;
  784         struct thread *dtd;
  785 
  786         td = curthread;
  787         p = td->td_proc;
  788         KASSERT(p->p_state == PRS_NORMAL, ("executing process is still new"));
  789 
  790         CTR4(KTR_PROC, "fork_exit: new thread %p (kse %p, pid %d, %s)",
  791                 td, td->td_sched, p->p_pid, p->p_comm);
  792 
  793         sched_fork_exit(td);
  794         /*
  795         * Processes normally resume in mi_switch() after being
  796         * cpu_switch()'ed to, but when children start up they arrive here
  797         * instead, so we must do much the same things as mi_switch() would.
  798         */
  799         if ((dtd = PCPU_GET(deadthread))) {
  800                 PCPU_SET(deadthread, NULL);
  801                 thread_stash(dtd);
  802         }
  803         thread_unlock(td);
  804 
  805         /*
  806          * cpu_set_fork_handler intercepts this function call to
  807          * have this call a non-return function to stay in kernel mode.
  808          * initproc has its own fork handler, but it does return.
  809          */
  810         KASSERT(callout != NULL, ("NULL callout in fork_exit"));
  811         callout(arg, frame);
  812 
  813         /*
  814          * Check if a kernel thread misbehaved and returned from its main
  815          * function.
  816          */
  817         if (p->p_flag & P_KTHREAD) {
  818                 printf("Kernel thread \"%s\" (pid %d) exited prematurely.\n",
  819                     p->p_comm, p->p_pid);
  820                 kthread_exit(0);
  821         }
  822         mtx_assert(&Giant, MA_NOTOWNED);
  823 
  824         EVENTHANDLER_INVOKE(schedtail, p);
  825 }
  826 
  827 /*
  828  * Simplified back end of syscall(), used when returning from fork()
  829  * directly into user mode.  Giant is not held on entry, and must not
  830  * be held on return.  This function is passed in to fork_exit() as the
  831  * first parameter and is called when returning to a new userland process.
  832  */
  833 void
  834 fork_return(td, frame)
  835         struct thread *td;
  836         struct trapframe *frame;
  837 {
  838 
  839         userret(td, frame);
  840 #ifdef KTRACE
  841         if (KTRPOINT(td, KTR_SYSRET))
  842                 ktrsysret(SYS_fork, 0, 0);
  843 #endif
  844         mtx_assert(&Giant, MA_NOTOWNED);
  845 }

Cache object: 9ad97717d8ec4845967efde606d71701


[ 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.