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_exec.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  * Copyright (c) 1993, David Greenman
    3  * All rights reserved.
    4  *
    5  * Redistribution and use in source and binary forms, with or without
    6  * modification, are permitted provided that the following conditions
    7  * are met:
    8  * 1. Redistributions of source code must retain the above copyright
    9  *    notice, this list of conditions and the following disclaimer.
   10  * 2. Redistributions in binary form must reproduce the above copyright
   11  *    notice, this list of conditions and the following disclaimer in the
   12  *    documentation and/or other materials provided with the distribution.
   13  *
   14  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
   15  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
   16  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
   17  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
   18  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
   19  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
   20  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
   21  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
   22  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
   23  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   24  * SUCH DAMAGE.
   25  */
   26 
   27 #include <sys/cdefs.h>
   28 __FBSDID("$FreeBSD: releng/8.3/sys/kern/kern_exec.c 226256 2011-10-11 13:54:20Z kib $");
   29 
   30 #include "opt_hwpmc_hooks.h"
   31 #include "opt_kdtrace.h"
   32 #include "opt_ktrace.h"
   33 #include "opt_vm.h"
   34 
   35 #include <sys/param.h>
   36 #include <sys/systm.h>
   37 #include <sys/eventhandler.h>
   38 #include <sys/lock.h>
   39 #include <sys/mutex.h>
   40 #include <sys/sysproto.h>
   41 #include <sys/signalvar.h>
   42 #include <sys/kernel.h>
   43 #include <sys/mount.h>
   44 #include <sys/filedesc.h>
   45 #include <sys/fcntl.h>
   46 #include <sys/acct.h>
   47 #include <sys/exec.h>
   48 #include <sys/imgact.h>
   49 #include <sys/imgact_elf.h>
   50 #include <sys/wait.h>
   51 #include <sys/malloc.h>
   52 #include <sys/priv.h>
   53 #include <sys/proc.h>
   54 #include <sys/pioctl.h>
   55 #include <sys/namei.h>
   56 #include <sys/resourcevar.h>
   57 #include <sys/sdt.h>
   58 #include <sys/sf_buf.h>
   59 #include <sys/syscallsubr.h>
   60 #include <sys/sysent.h>
   61 #include <sys/shm.h>
   62 #include <sys/sysctl.h>
   63 #include <sys/vnode.h>
   64 #include <sys/stat.h>
   65 #ifdef KTRACE
   66 #include <sys/ktrace.h>
   67 #endif
   68 
   69 #include <vm/vm.h>
   70 #include <vm/vm_param.h>
   71 #include <vm/pmap.h>
   72 #include <vm/vm_page.h>
   73 #include <vm/vm_map.h>
   74 #include <vm/vm_kern.h>
   75 #include <vm/vm_extern.h>
   76 #include <vm/vm_object.h>
   77 #include <vm/vm_pager.h>
   78 
   79 #ifdef  HWPMC_HOOKS
   80 #include <sys/pmckern.h>
   81 #endif
   82 
   83 #include <machine/reg.h>
   84 
   85 #include <security/audit/audit.h>
   86 #include <security/mac/mac_framework.h>
   87 
   88 #ifdef KDTRACE_HOOKS
   89 #include <sys/dtrace_bsd.h>
   90 dtrace_execexit_func_t  dtrace_fasttrap_exec;
   91 #endif
   92 
   93 SDT_PROVIDER_DECLARE(proc);
   94 SDT_PROBE_DEFINE(proc, kernel, , exec, exec);
   95 SDT_PROBE_ARGTYPE(proc, kernel, , exec, 0, "char *");
   96 SDT_PROBE_DEFINE(proc, kernel, , exec_failure, exec-failure);
   97 SDT_PROBE_ARGTYPE(proc, kernel, , exec_failure, 0, "int");
   98 SDT_PROBE_DEFINE(proc, kernel, , exec_success, exec-success);
   99 SDT_PROBE_ARGTYPE(proc, kernel, , exec_success, 0, "char *");
  100 
  101 MALLOC_DEFINE(M_PARGS, "proc-args", "Process arguments");
  102 
  103 static int sysctl_kern_ps_strings(SYSCTL_HANDLER_ARGS);
  104 static int sysctl_kern_usrstack(SYSCTL_HANDLER_ARGS);
  105 static int sysctl_kern_stackprot(SYSCTL_HANDLER_ARGS);
  106 static int do_execve(struct thread *td, struct image_args *args,
  107     struct mac *mac_p);
  108 static void exec_free_args(struct image_args *);
  109 
  110 /* XXX This should be vm_size_t. */
  111 SYSCTL_PROC(_kern, KERN_PS_STRINGS, ps_strings, CTLTYPE_ULONG|CTLFLAG_RD,
  112     NULL, 0, sysctl_kern_ps_strings, "LU", "");
  113 
  114 /* XXX This should be vm_size_t. */
  115 SYSCTL_PROC(_kern, KERN_USRSTACK, usrstack, CTLTYPE_ULONG|CTLFLAG_RD,
  116     NULL, 0, sysctl_kern_usrstack, "LU", "");
  117 
  118 SYSCTL_PROC(_kern, OID_AUTO, stackprot, CTLTYPE_INT|CTLFLAG_RD,
  119     NULL, 0, sysctl_kern_stackprot, "I", "");
  120 
  121 u_long ps_arg_cache_limit = PAGE_SIZE / 16;
  122 SYSCTL_ULONG(_kern, OID_AUTO, ps_arg_cache_limit, CTLFLAG_RW, 
  123     &ps_arg_cache_limit, 0, "");
  124 
  125 static int map_at_zero = 0;
  126 TUNABLE_INT("security.bsd.map_at_zero", &map_at_zero);
  127 SYSCTL_INT(_security_bsd, OID_AUTO, map_at_zero, CTLFLAG_RW, &map_at_zero, 0,
  128     "Permit processes to map an object at virtual address 0.");
  129 
  130 static int
  131 sysctl_kern_ps_strings(SYSCTL_HANDLER_ARGS)
  132 {
  133         struct proc *p;
  134         int error;
  135 
  136         p = curproc;
  137 #ifdef SCTL_MASK32
  138         if (req->flags & SCTL_MASK32) {
  139                 unsigned int val;
  140                 val = (unsigned int)p->p_sysent->sv_psstrings;
  141                 error = SYSCTL_OUT(req, &val, sizeof(val));
  142         } else
  143 #endif
  144                 error = SYSCTL_OUT(req, &p->p_sysent->sv_psstrings,
  145                    sizeof(p->p_sysent->sv_psstrings));
  146         return error;
  147 }
  148 
  149 static int
  150 sysctl_kern_usrstack(SYSCTL_HANDLER_ARGS)
  151 {
  152         struct proc *p;
  153         int error;
  154 
  155         p = curproc;
  156 #ifdef SCTL_MASK32
  157         if (req->flags & SCTL_MASK32) {
  158                 unsigned int val;
  159                 val = (unsigned int)p->p_sysent->sv_usrstack;
  160                 error = SYSCTL_OUT(req, &val, sizeof(val));
  161         } else
  162 #endif
  163                 error = SYSCTL_OUT(req, &p->p_sysent->sv_usrstack,
  164                     sizeof(p->p_sysent->sv_usrstack));
  165         return error;
  166 }
  167 
  168 static int
  169 sysctl_kern_stackprot(SYSCTL_HANDLER_ARGS)
  170 {
  171         struct proc *p;
  172 
  173         p = curproc;
  174         return (SYSCTL_OUT(req, &p->p_sysent->sv_stackprot,
  175             sizeof(p->p_sysent->sv_stackprot)));
  176 }
  177 
  178 /*
  179  * Each of the items is a pointer to a `const struct execsw', hence the
  180  * double pointer here.
  181  */
  182 static const struct execsw **execsw;
  183 
  184 #ifndef _SYS_SYSPROTO_H_
  185 struct execve_args {
  186         char    *fname; 
  187         char    **argv;
  188         char    **envv; 
  189 };
  190 #endif
  191 
  192 int
  193 execve(td, uap)
  194         struct thread *td;
  195         struct execve_args /* {
  196                 char *fname;
  197                 char **argv;
  198                 char **envv;
  199         } */ *uap;
  200 {
  201         int error;
  202         struct image_args args;
  203 
  204         error = exec_copyin_args(&args, uap->fname, UIO_USERSPACE,
  205             uap->argv, uap->envv);
  206         if (error == 0)
  207                 error = kern_execve(td, &args, NULL);
  208         return (error);
  209 }
  210 
  211 #ifndef _SYS_SYSPROTO_H_
  212 struct fexecve_args {
  213         int     fd;
  214         char    **argv;
  215         char    **envv;
  216 }
  217 #endif
  218 int
  219 fexecve(struct thread *td, struct fexecve_args *uap)
  220 {
  221         int error;
  222         struct image_args args;
  223 
  224         error = exec_copyin_args(&args, NULL, UIO_SYSSPACE,
  225             uap->argv, uap->envv);
  226         if (error == 0) {
  227                 args.fd = uap->fd;
  228                 error = kern_execve(td, &args, NULL);
  229         }
  230         return (error);
  231 }
  232 
  233 #ifndef _SYS_SYSPROTO_H_
  234 struct __mac_execve_args {
  235         char    *fname;
  236         char    **argv;
  237         char    **envv;
  238         struct mac      *mac_p;
  239 };
  240 #endif
  241 
  242 int
  243 __mac_execve(td, uap)
  244         struct thread *td;
  245         struct __mac_execve_args /* {
  246                 char *fname;
  247                 char **argv;
  248                 char **envv;
  249                 struct mac *mac_p;
  250         } */ *uap;
  251 {
  252 #ifdef MAC
  253         int error;
  254         struct image_args args;
  255 
  256         error = exec_copyin_args(&args, uap->fname, UIO_USERSPACE,
  257             uap->argv, uap->envv);
  258         if (error == 0)
  259                 error = kern_execve(td, &args, uap->mac_p);
  260         return (error);
  261 #else
  262         return (ENOSYS);
  263 #endif
  264 }
  265 
  266 /*
  267  * XXX: kern_execve has the astonishing property of not always returning to
  268  * the caller.  If sufficiently bad things happen during the call to
  269  * do_execve(), it can end up calling exit1(); as a result, callers must
  270  * avoid doing anything which they might need to undo (e.g., allocating
  271  * memory).
  272  */
  273 int
  274 kern_execve(td, args, mac_p)
  275         struct thread *td;
  276         struct image_args *args;
  277         struct mac *mac_p;
  278 {
  279         struct proc *p = td->td_proc;
  280         int error;
  281 
  282         AUDIT_ARG_ARGV(args->begin_argv, args->argc,
  283             args->begin_envv - args->begin_argv);
  284         AUDIT_ARG_ENVV(args->begin_envv, args->envc,
  285             args->endp - args->begin_envv);
  286         if (p->p_flag & P_HADTHREADS) {
  287                 PROC_LOCK(p);
  288                 if (thread_single(SINGLE_BOUNDARY)) {
  289                         PROC_UNLOCK(p);
  290                         exec_free_args(args);
  291                         return (ERESTART);      /* Try again later. */
  292                 }
  293                 PROC_UNLOCK(p);
  294         }
  295 
  296         error = do_execve(td, args, mac_p);
  297 
  298         if (p->p_flag & P_HADTHREADS) {
  299                 PROC_LOCK(p);
  300                 /*
  301                  * If success, we upgrade to SINGLE_EXIT state to
  302                  * force other threads to suicide.
  303                  */
  304                 if (error == 0)
  305                         thread_single(SINGLE_EXIT);
  306                 else
  307                         thread_single_end();
  308                 PROC_UNLOCK(p);
  309         }
  310 
  311         return (error);
  312 }
  313 
  314 /*
  315  * In-kernel implementation of execve().  All arguments are assumed to be
  316  * userspace pointers from the passed thread.
  317  */
  318 static int
  319 do_execve(td, args, mac_p)
  320         struct thread *td;
  321         struct image_args *args;
  322         struct mac *mac_p;
  323 {
  324         struct proc *p = td->td_proc;
  325         struct nameidata nd;
  326         struct ucred *newcred = NULL, *oldcred;
  327         struct uidinfo *euip;
  328         register_t *stack_base;
  329         int error, i;
  330         struct image_params image_params, *imgp;
  331         struct vattr attr;
  332         int (*img_first)(struct image_params *);
  333         struct pargs *oldargs = NULL, *newargs = NULL;
  334         struct sigacts *oldsigacts, *newsigacts;
  335 #ifdef KTRACE
  336         struct vnode *tracevp = NULL;
  337         struct ucred *tracecred = NULL;
  338 #endif
  339         struct vnode *textvp = NULL, *binvp = NULL;
  340         int credential_changing;
  341         int vfslocked;
  342         int textset;
  343 #ifdef MAC
  344         struct label *interpvplabel = NULL;
  345         int will_transition;
  346 #endif
  347 #ifdef HWPMC_HOOKS
  348         struct pmckern_procexec pe;
  349 #endif
  350         static const char fexecv_proc_title[] = "(fexecv)";
  351 
  352         vfslocked = 0;
  353         imgp = &image_params;
  354 
  355         /*
  356          * Lock the process and set the P_INEXEC flag to indicate that
  357          * it should be left alone until we're done here.  This is
  358          * necessary to avoid race conditions - e.g. in ptrace() -
  359          * that might allow a local user to illicitly obtain elevated
  360          * privileges.
  361          */
  362         PROC_LOCK(p);
  363         KASSERT((p->p_flag & P_INEXEC) == 0,
  364             ("%s(): process already has P_INEXEC flag", __func__));
  365         p->p_flag |= P_INEXEC;
  366         PROC_UNLOCK(p);
  367 
  368         /*
  369          * Initialize part of the common data
  370          */
  371         imgp->proc = p;
  372         imgp->execlabel = NULL;
  373         imgp->attr = &attr;
  374         imgp->entry_addr = 0;
  375         imgp->vmspace_destroyed = 0;
  376         imgp->interpreted = 0;
  377         imgp->opened = 0;
  378         imgp->interpreter_name = args->buf + PATH_MAX + ARG_MAX;
  379         imgp->auxargs = NULL;
  380         imgp->vp = NULL;
  381         imgp->object = NULL;
  382         imgp->firstpage = NULL;
  383         imgp->ps_strings = 0;
  384         imgp->auxarg_size = 0;
  385         imgp->args = args;
  386         imgp->execpath = imgp->freepath = NULL;
  387         imgp->execpathp = 0;
  388 
  389 #ifdef MAC
  390         error = mac_execve_enter(imgp, mac_p);
  391         if (error)
  392                 goto exec_fail;
  393 #endif
  394 
  395         imgp->image_header = NULL;
  396 
  397         /*
  398          * Translate the file name. namei() returns a vnode pointer
  399          *      in ni_vp amoung other things.
  400          *
  401          * XXXAUDIT: It would be desirable to also audit the name of the
  402          * interpreter if this is an interpreted binary.
  403          */
  404         if (args->fname != NULL) {
  405                 NDINIT(&nd, LOOKUP, ISOPEN | LOCKLEAF | FOLLOW | SAVENAME
  406                     | MPSAFE | AUDITVNODE1, UIO_SYSSPACE, args->fname, td);
  407         }
  408 
  409         SDT_PROBE(proc, kernel, , exec, args->fname, 0, 0, 0, 0 );
  410 
  411 interpret:
  412         if (args->fname != NULL) {
  413                 error = namei(&nd);
  414                 if (error)
  415                         goto exec_fail;
  416 
  417                 vfslocked = NDHASGIANT(&nd);
  418                 binvp  = nd.ni_vp;
  419                 imgp->vp = binvp;
  420         } else {
  421                 AUDIT_ARG_FD(args->fd);
  422                 error = fgetvp(td, args->fd, &binvp);
  423                 if (error)
  424                         goto exec_fail;
  425                 vfslocked = VFS_LOCK_GIANT(binvp->v_mount);
  426                 vn_lock(binvp, LK_EXCLUSIVE | LK_RETRY);
  427                 AUDIT_ARG_VNODE1(binvp);
  428                 imgp->vp = binvp;
  429         }
  430 
  431         /*
  432          * Check file permissions (also 'opens' file)
  433          */
  434         error = exec_check_permissions(imgp);
  435         if (error)
  436                 goto exec_fail_dealloc;
  437 
  438         imgp->object = imgp->vp->v_object;
  439         if (imgp->object != NULL)
  440                 vm_object_reference(imgp->object);
  441 
  442         /*
  443          * Set VV_TEXT now so no one can write to the executable while we're
  444          * activating it.
  445          *
  446          * Remember if this was set before and unset it in case this is not
  447          * actually an executable image.
  448          */
  449         textset = imgp->vp->v_vflag & VV_TEXT;
  450         imgp->vp->v_vflag |= VV_TEXT;
  451 
  452         error = exec_map_first_page(imgp);
  453         if (error)
  454                 goto exec_fail_dealloc;
  455 
  456         imgp->proc->p_osrel = 0;
  457         /*
  458          *      If the current process has a special image activator it
  459          *      wants to try first, call it.   For example, emulating shell
  460          *      scripts differently.
  461          */
  462         error = -1;
  463         if ((img_first = imgp->proc->p_sysent->sv_imgact_try) != NULL)
  464                 error = img_first(imgp);
  465 
  466         /*
  467          *      Loop through the list of image activators, calling each one.
  468          *      An activator returns -1 if there is no match, 0 on success,
  469          *      and an error otherwise.
  470          */
  471         for (i = 0; error == -1 && execsw[i]; ++i) {
  472                 if (execsw[i]->ex_imgact == NULL ||
  473                     execsw[i]->ex_imgact == img_first) {
  474                         continue;
  475                 }
  476                 error = (*execsw[i]->ex_imgact)(imgp);
  477         }
  478 
  479         if (error) {
  480                 if (error == -1) {
  481                         if (textset == 0)
  482                                 imgp->vp->v_vflag &= ~VV_TEXT;
  483                         error = ENOEXEC;
  484                 }
  485                 goto exec_fail_dealloc;
  486         }
  487 
  488         /*
  489          * Special interpreter operation, cleanup and loop up to try to
  490          * activate the interpreter.
  491          */
  492         if (imgp->interpreted) {
  493                 exec_unmap_first_page(imgp);
  494                 /*
  495                  * VV_TEXT needs to be unset for scripts.  There is a short
  496                  * period before we determine that something is a script where
  497                  * VV_TEXT will be set. The vnode lock is held over this
  498                  * entire period so nothing should illegitimately be blocked.
  499                  */
  500                 imgp->vp->v_vflag &= ~VV_TEXT;
  501                 /* free name buffer and old vnode */
  502                 if (args->fname != NULL)
  503                         NDFREE(&nd, NDF_ONLY_PNBUF);
  504 #ifdef MAC
  505                 mac_execve_interpreter_enter(binvp, &interpvplabel);
  506 #endif
  507                 if (imgp->opened) {
  508                         VOP_CLOSE(binvp, FREAD, td->td_ucred, td);
  509                         imgp->opened = 0;
  510                 }
  511                 vput(binvp);
  512                 vm_object_deallocate(imgp->object);
  513                 imgp->object = NULL;
  514                 VFS_UNLOCK_GIANT(vfslocked);
  515                 vfslocked = 0;
  516                 /* set new name to that of the interpreter */
  517                 NDINIT(&nd, LOOKUP, LOCKLEAF | FOLLOW | SAVENAME | MPSAFE,
  518                     UIO_SYSSPACE, imgp->interpreter_name, td);
  519                 args->fname = imgp->interpreter_name;
  520                 goto interpret;
  521         }
  522 
  523         /*
  524          * NB: We unlock the vnode here because it is believed that none
  525          * of the sv_copyout_strings/sv_fixup operations require the vnode.
  526          */
  527         VOP_UNLOCK(imgp->vp, 0);
  528 
  529         /*
  530          * Do the best to calculate the full path to the image file.
  531          */
  532         if (imgp->auxargs != NULL &&
  533             ((args->fname != NULL && args->fname[0] == '/') ||
  534              vn_fullpath(td, imgp->vp, &imgp->execpath, &imgp->freepath) != 0))
  535                 imgp->execpath = args->fname;
  536 
  537         /*
  538          * Copy out strings (args and env) and initialize stack base
  539          */
  540         if (p->p_sysent->sv_copyout_strings)
  541                 stack_base = (*p->p_sysent->sv_copyout_strings)(imgp);
  542         else
  543                 stack_base = exec_copyout_strings(imgp);
  544 
  545         /*
  546          * If custom stack fixup routine present for this process
  547          * let it do the stack setup.
  548          * Else stuff argument count as first item on stack
  549          */
  550         if (p->p_sysent->sv_fixup != NULL)
  551                 (*p->p_sysent->sv_fixup)(&stack_base, imgp);
  552         else
  553                 suword(--stack_base, imgp->args->argc);
  554 
  555         /*
  556          * For security and other reasons, the file descriptor table cannot
  557          * be shared after an exec.
  558          */
  559         fdunshare(p, td);
  560 
  561         /*
  562          * Malloc things before we need locks.
  563          */
  564         newcred = crget();
  565         euip = uifind(attr.va_uid);
  566         i = imgp->args->begin_envv - imgp->args->begin_argv;
  567         /* Cache arguments if they fit inside our allowance */
  568         if (ps_arg_cache_limit >= i + sizeof(struct pargs)) {
  569                 newargs = pargs_alloc(i);
  570                 bcopy(imgp->args->begin_argv, newargs->ar_args, i);
  571         }
  572 
  573         /* close files on exec */
  574         fdcloseexec(td);
  575         vn_lock(imgp->vp, LK_EXCLUSIVE | LK_RETRY);
  576 
  577         /* Get a reference to the vnode prior to locking the proc */
  578         VREF(binvp);
  579 
  580         /*
  581          * For security and other reasons, signal handlers cannot
  582          * be shared after an exec. The new process gets a copy of the old
  583          * handlers. In execsigs(), the new process will have its signals
  584          * reset.
  585          */
  586         PROC_LOCK(p);
  587         oldcred = crcopysafe(p, newcred);
  588         if (sigacts_shared(p->p_sigacts)) {
  589                 oldsigacts = p->p_sigacts;
  590                 PROC_UNLOCK(p);
  591                 newsigacts = sigacts_alloc();
  592                 sigacts_copy(newsigacts, oldsigacts);
  593                 PROC_LOCK(p);
  594                 p->p_sigacts = newsigacts;
  595         } else
  596                 oldsigacts = NULL;
  597 
  598         /* Stop profiling */
  599         stopprofclock(p);
  600 
  601         /* reset caught signals */
  602         execsigs(p);
  603 
  604         /* name this process - nameiexec(p, ndp) */
  605         bzero(p->p_comm, sizeof(p->p_comm));
  606         if (args->fname)
  607                 bcopy(nd.ni_cnd.cn_nameptr, p->p_comm,
  608                     min(nd.ni_cnd.cn_namelen, MAXCOMLEN));
  609         else if (vn_commname(binvp, p->p_comm, sizeof(p->p_comm)) != 0)
  610                 bcopy(fexecv_proc_title, p->p_comm, sizeof(fexecv_proc_title));
  611         bcopy(p->p_comm, td->td_name, sizeof(td->td_name));
  612 
  613         /*
  614          * mark as execed, wakeup the process that vforked (if any) and tell
  615          * it that it now has its own resources back
  616          */
  617         p->p_flag |= P_EXEC;
  618         if (p->p_pptr && (p->p_flag & P_PPWAIT)) {
  619                 p->p_flag &= ~P_PPWAIT;
  620                 cv_broadcast(&p->p_pwait);
  621         }
  622 
  623         /*
  624          * Implement image setuid/setgid.
  625          *
  626          * Don't honor setuid/setgid if the filesystem prohibits it or if
  627          * the process is being traced.
  628          *
  629          * XXXMAC: For the time being, use NOSUID to also prohibit
  630          * transitions on the file system.
  631          */
  632         credential_changing = 0;
  633         credential_changing |= (attr.va_mode & S_ISUID) && oldcred->cr_uid !=
  634             attr.va_uid;
  635         credential_changing |= (attr.va_mode & S_ISGID) && oldcred->cr_gid !=
  636             attr.va_gid;
  637 #ifdef MAC
  638         will_transition = mac_vnode_execve_will_transition(oldcred, imgp->vp,
  639             interpvplabel, imgp);
  640         credential_changing |= will_transition;
  641 #endif
  642 
  643         if (credential_changing &&
  644             (imgp->vp->v_mount->mnt_flag & MNT_NOSUID) == 0 &&
  645             (p->p_flag & P_TRACED) == 0) {
  646                 /*
  647                  * Turn off syscall tracing for set-id programs, except for
  648                  * root.  Record any set-id flags first to make sure that
  649                  * we do not regain any tracing during a possible block.
  650                  */
  651                 setsugid(p);
  652 
  653 #ifdef KTRACE
  654                 if (priv_check_cred(oldcred, PRIV_DEBUG_DIFFCRED, 0))
  655                         ktrprocexec(p, &tracecred, &tracevp);
  656 #endif
  657                 /*
  658                  * Close any file descriptors 0..2 that reference procfs,
  659                  * then make sure file descriptors 0..2 are in use.
  660                  *
  661                  * setugidsafety() may call closef() and then pfind()
  662                  * which may grab the process lock.
  663                  * fdcheckstd() may call falloc() which may block to
  664                  * allocate memory, so temporarily drop the process lock.
  665                  */
  666                 PROC_UNLOCK(p);
  667                 VOP_UNLOCK(imgp->vp, 0);
  668                 setugidsafety(td);
  669                 error = fdcheckstd(td);
  670                 vn_lock(imgp->vp, LK_EXCLUSIVE | LK_RETRY);
  671                 if (error != 0)
  672                         goto done1;
  673                 PROC_LOCK(p);
  674                 /*
  675                  * Set the new credentials.
  676                  */
  677                 if (attr.va_mode & S_ISUID)
  678                         change_euid(newcred, euip);
  679                 if (attr.va_mode & S_ISGID)
  680                         change_egid(newcred, attr.va_gid);
  681 #ifdef MAC
  682                 if (will_transition) {
  683                         mac_vnode_execve_transition(oldcred, newcred, imgp->vp,
  684                             interpvplabel, imgp);
  685                 }
  686 #endif
  687                 /*
  688                  * Implement correct POSIX saved-id behavior.
  689                  *
  690                  * XXXMAC: Note that the current logic will save the
  691                  * uid and gid if a MAC domain transition occurs, even
  692                  * though maybe it shouldn't.
  693                  */
  694                 change_svuid(newcred, newcred->cr_uid);
  695                 change_svgid(newcred, newcred->cr_gid);
  696                 p->p_ucred = newcred;
  697                 newcred = NULL;
  698         } else {
  699                 if (oldcred->cr_uid == oldcred->cr_ruid &&
  700                     oldcred->cr_gid == oldcred->cr_rgid)
  701                         p->p_flag &= ~P_SUGID;
  702                 /*
  703                  * Implement correct POSIX saved-id behavior.
  704                  *
  705                  * XXX: It's not clear that the existing behavior is
  706                  * POSIX-compliant.  A number of sources indicate that the
  707                  * saved uid/gid should only be updated if the new ruid is
  708                  * not equal to the old ruid, or the new euid is not equal
  709                  * to the old euid and the new euid is not equal to the old
  710                  * ruid.  The FreeBSD code always updates the saved uid/gid.
  711                  * Also, this code uses the new (replaced) euid and egid as
  712                  * the source, which may or may not be the right ones to use.
  713                  */
  714                 if (oldcred->cr_svuid != oldcred->cr_uid ||
  715                     oldcred->cr_svgid != oldcred->cr_gid) {
  716                         change_svuid(newcred, newcred->cr_uid);
  717                         change_svgid(newcred, newcred->cr_gid);
  718                         p->p_ucred = newcred;
  719                         newcred = NULL;
  720                 }
  721         }
  722 
  723         /*
  724          * Store the vp for use in procfs.  This vnode was referenced prior
  725          * to locking the proc lock.
  726          */
  727         textvp = p->p_textvp;
  728         p->p_textvp = binvp;
  729 
  730 #ifdef KDTRACE_HOOKS
  731         /*
  732          * Tell the DTrace fasttrap provider about the exec if it
  733          * has declared an interest.
  734          */
  735         if (dtrace_fasttrap_exec)
  736                 dtrace_fasttrap_exec(p);
  737 #endif
  738 
  739         /*
  740          * Notify others that we exec'd, and clear the P_INEXEC flag
  741          * as we're now a bona fide freshly-execed process.
  742          */
  743         KNOTE_LOCKED(&p->p_klist, NOTE_EXEC);
  744         p->p_flag &= ~P_INEXEC;
  745 
  746         /* clear "fork but no exec" flag, as we _are_ execing */
  747         p->p_acflag &= ~AFORK;
  748 
  749         /*
  750          * Free any previous argument cache and replace it with
  751          * the new argument cache, if any.
  752          */
  753         oldargs = p->p_args;
  754         p->p_args = newargs;
  755         newargs = NULL;
  756 
  757 #ifdef  HWPMC_HOOKS
  758         /*
  759          * Check if system-wide sampling is in effect or if the
  760          * current process is using PMCs.  If so, do exec() time
  761          * processing.  This processing needs to happen AFTER the
  762          * P_INEXEC flag is cleared.
  763          *
  764          * The proc lock needs to be released before taking the PMC
  765          * SX.
  766          */
  767         if (PMC_SYSTEM_SAMPLING_ACTIVE() || PMC_PROC_IS_USING_PMCS(p)) {
  768                 PROC_UNLOCK(p);
  769                 VOP_UNLOCK(imgp->vp, 0);
  770                 pe.pm_credentialschanged = credential_changing;
  771                 pe.pm_entryaddr = imgp->entry_addr;
  772 
  773                 PMC_CALL_HOOK_X(td, PMC_FN_PROCESS_EXEC, (void *) &pe);
  774                 vn_lock(imgp->vp, LK_EXCLUSIVE | LK_RETRY);
  775         } else
  776                 PROC_UNLOCK(p);
  777 #else  /* !HWPMC_HOOKS */
  778         PROC_UNLOCK(p);
  779 #endif
  780 
  781         /* Set values passed into the program in registers. */
  782         if (p->p_sysent->sv_setregs)
  783                 (*p->p_sysent->sv_setregs)(td, imgp->entry_addr,
  784                     (u_long)(uintptr_t)stack_base, imgp->ps_strings);
  785         else
  786                 exec_setregs(td, imgp->entry_addr,
  787                     (u_long)(uintptr_t)stack_base, imgp->ps_strings);
  788 
  789         vfs_mark_atime(imgp->vp, td->td_ucred);
  790 
  791         SDT_PROBE(proc, kernel, , exec_success, args->fname, 0, 0, 0, 0);
  792 
  793 done1:
  794         /*
  795          * Free any resources malloc'd earlier that we didn't use.
  796          */
  797         uifree(euip);
  798         if (newcred == NULL)
  799                 crfree(oldcred);
  800         else
  801                 crfree(newcred);
  802         VOP_UNLOCK(imgp->vp, 0);
  803 
  804         /*
  805          * Handle deferred decrement of ref counts.
  806          */
  807         if (textvp != NULL) {
  808                 int tvfslocked;
  809 
  810                 tvfslocked = VFS_LOCK_GIANT(textvp->v_mount);
  811                 vrele(textvp);
  812                 VFS_UNLOCK_GIANT(tvfslocked);
  813         }
  814         if (binvp && error != 0)
  815                 vrele(binvp);
  816 #ifdef KTRACE
  817         if (tracevp != NULL) {
  818                 int tvfslocked;
  819 
  820                 tvfslocked = VFS_LOCK_GIANT(tracevp->v_mount);
  821                 vrele(tracevp);
  822                 VFS_UNLOCK_GIANT(tvfslocked);
  823         }
  824         if (tracecred != NULL)
  825                 crfree(tracecred);
  826 #endif
  827         vn_lock(imgp->vp, LK_EXCLUSIVE | LK_RETRY);
  828         pargs_drop(oldargs);
  829         pargs_drop(newargs);
  830         if (oldsigacts != NULL)
  831                 sigacts_free(oldsigacts);
  832 
  833 exec_fail_dealloc:
  834 
  835         /*
  836          * free various allocated resources
  837          */
  838         if (imgp->firstpage != NULL)
  839                 exec_unmap_first_page(imgp);
  840 
  841         if (imgp->vp != NULL) {
  842                 if (args->fname)
  843                         NDFREE(&nd, NDF_ONLY_PNBUF);
  844                 if (imgp->opened)
  845                         VOP_CLOSE(imgp->vp, FREAD, td->td_ucred, td);
  846                 vput(imgp->vp);
  847         }
  848 
  849         if (imgp->object != NULL)
  850                 vm_object_deallocate(imgp->object);
  851 
  852         free(imgp->freepath, M_TEMP);
  853 
  854         if (error == 0) {
  855                 PROC_LOCK(p);
  856                 td->td_dbgflags |= TDB_EXEC;
  857                 PROC_UNLOCK(p);
  858 
  859                 /*
  860                  * Stop the process here if its stop event mask has
  861                  * the S_EXEC bit set.
  862                  */
  863                 STOPEVENT(p, S_EXEC, 0);
  864                 goto done2;
  865         }
  866 
  867 exec_fail:
  868         /* we're done here, clear P_INEXEC */
  869         PROC_LOCK(p);
  870         p->p_flag &= ~P_INEXEC;
  871         PROC_UNLOCK(p);
  872 
  873         SDT_PROBE(proc, kernel, , exec_failure, error, 0, 0, 0, 0);
  874 
  875 done2:
  876 #ifdef MAC
  877         mac_execve_exit(imgp);
  878         mac_execve_interpreter_exit(interpvplabel);
  879 #endif
  880         VFS_UNLOCK_GIANT(vfslocked);
  881         exec_free_args(args);
  882 
  883         if (error && imgp->vmspace_destroyed) {
  884                 /* sorry, no more process anymore. exit gracefully */
  885                 exit1(td, W_EXITCODE(0, SIGABRT));
  886                 /* NOT REACHED */
  887         }
  888 
  889 #ifdef KTRACE
  890         if (error == 0)
  891                 ktrprocctor(p);
  892 #endif
  893 
  894         return (error);
  895 }
  896 
  897 int
  898 exec_map_first_page(imgp)
  899         struct image_params *imgp;
  900 {
  901         int rv, i;
  902         int initial_pagein;
  903         vm_page_t ma[VM_INITIAL_PAGEIN];
  904         vm_object_t object;
  905 
  906         if (imgp->firstpage != NULL)
  907                 exec_unmap_first_page(imgp);
  908 
  909         object = imgp->vp->v_object;
  910         if (object == NULL)
  911                 return (EACCES);
  912         VM_OBJECT_LOCK(object);
  913 #if VM_NRESERVLEVEL > 0
  914         if ((object->flags & OBJ_COLORED) == 0) {
  915                 object->flags |= OBJ_COLORED;
  916                 object->pg_color = 0;
  917         }
  918 #endif
  919         ma[0] = vm_page_grab(object, 0, VM_ALLOC_NORMAL | VM_ALLOC_RETRY);
  920         if (ma[0]->valid != VM_PAGE_BITS_ALL) {
  921                 initial_pagein = VM_INITIAL_PAGEIN;
  922                 if (initial_pagein > object->size)
  923                         initial_pagein = object->size;
  924                 for (i = 1; i < initial_pagein; i++) {
  925                         if ((ma[i] = vm_page_next(ma[i - 1])) != NULL) {
  926                                 if (ma[i]->valid)
  927                                         break;
  928                                 if ((ma[i]->oflags & VPO_BUSY) || ma[i]->busy)
  929                                         break;
  930                                 vm_page_busy(ma[i]);
  931                         } else {
  932                                 ma[i] = vm_page_alloc(object, i,
  933                                     VM_ALLOC_NORMAL | VM_ALLOC_IFNOTCACHED);
  934                                 if (ma[i] == NULL)
  935                                         break;
  936                         }
  937                 }
  938                 initial_pagein = i;
  939                 rv = vm_pager_get_pages(object, ma, initial_pagein, 0);
  940                 ma[0] = vm_page_lookup(object, 0);
  941                 if ((rv != VM_PAGER_OK) || (ma[0] == NULL)) {
  942                         if (ma[0]) {
  943                                 vm_page_lock_queues();
  944                                 vm_page_free(ma[0]);
  945                                 vm_page_unlock_queues();
  946                         }
  947                         VM_OBJECT_UNLOCK(object);
  948                         return (EIO);
  949                 }
  950         }
  951         vm_page_lock_queues();
  952         vm_page_hold(ma[0]);
  953         vm_page_unlock_queues();
  954         vm_page_wakeup(ma[0]);
  955         VM_OBJECT_UNLOCK(object);
  956 
  957         imgp->firstpage = sf_buf_alloc(ma[0], 0);
  958         imgp->image_header = (char *)sf_buf_kva(imgp->firstpage);
  959 
  960         return (0);
  961 }
  962 
  963 void
  964 exec_unmap_first_page(imgp)
  965         struct image_params *imgp;
  966 {
  967         vm_page_t m;
  968 
  969         if (imgp->firstpage != NULL) {
  970                 m = sf_buf_page(imgp->firstpage);
  971                 sf_buf_free(imgp->firstpage);
  972                 imgp->firstpage = NULL;
  973                 vm_page_lock_queues();
  974                 vm_page_unhold(m);
  975                 vm_page_unlock_queues();
  976         }
  977 }
  978 
  979 /*
  980  * Destroy old address space, and allocate a new stack
  981  *      The new stack is only SGROWSIZ large because it is grown
  982  *      automatically in trap.c.
  983  */
  984 int
  985 exec_new_vmspace(imgp, sv)
  986         struct image_params *imgp;
  987         struct sysentvec *sv;
  988 {
  989         int error;
  990         struct proc *p = imgp->proc;
  991         struct vmspace *vmspace = p->p_vmspace;
  992         vm_offset_t sv_minuser, stack_addr;
  993         vm_map_t map;
  994         u_long ssiz;
  995 
  996         imgp->vmspace_destroyed = 1;
  997         imgp->sysent = sv;
  998 
  999         /* May be called with Giant held */
 1000         EVENTHANDLER_INVOKE(process_exec, p, imgp);
 1001 
 1002         /*
 1003          * Blow away entire process VM, if address space not shared,
 1004          * otherwise, create a new VM space so that other threads are
 1005          * not disrupted
 1006          */
 1007         map = &vmspace->vm_map;
 1008         if (map_at_zero)
 1009                 sv_minuser = sv->sv_minuser;
 1010         else
 1011                 sv_minuser = MAX(sv->sv_minuser, PAGE_SIZE);
 1012         if (vmspace->vm_refcnt == 1 && vm_map_min(map) == sv_minuser &&
 1013             vm_map_max(map) == sv->sv_maxuser) {
 1014                 shmexit(vmspace);
 1015                 pmap_remove_pages(vmspace_pmap(vmspace));
 1016                 vm_map_remove(map, vm_map_min(map), vm_map_max(map));
 1017         } else {
 1018                 error = vmspace_exec(p, sv_minuser, sv->sv_maxuser);
 1019                 if (error)
 1020                         return (error);
 1021                 vmspace = p->p_vmspace;
 1022                 map = &vmspace->vm_map;
 1023         }
 1024 
 1025         /* Allocate a new stack */
 1026         if (sv->sv_maxssiz != NULL)
 1027                 ssiz = *sv->sv_maxssiz;
 1028         else
 1029                 ssiz = maxssiz;
 1030         stack_addr = sv->sv_usrstack - ssiz;
 1031         error = vm_map_stack(map, stack_addr, (vm_size_t)ssiz,
 1032             sv->sv_stackprot, VM_PROT_ALL, MAP_STACK_GROWS_DOWN);
 1033         if (error)
 1034                 return (error);
 1035 
 1036 #ifdef __ia64__
 1037         /* Allocate a new register stack */
 1038         stack_addr = IA64_BACKINGSTORE;
 1039         error = vm_map_stack(map, stack_addr, (vm_size_t)ssiz,
 1040             sv->sv_stackprot, VM_PROT_ALL, MAP_STACK_GROWS_UP);
 1041         if (error)
 1042                 return (error);
 1043 #endif
 1044 
 1045         /* vm_ssize and vm_maxsaddr are somewhat antiquated concepts in the
 1046          * VM_STACK case, but they are still used to monitor the size of the
 1047          * process stack so we can check the stack rlimit.
 1048          */
 1049         vmspace->vm_ssize = sgrowsiz >> PAGE_SHIFT;
 1050         vmspace->vm_maxsaddr = (char *)sv->sv_usrstack - ssiz;
 1051 
 1052         return (0);
 1053 }
 1054 
 1055 /*
 1056  * Copy out argument and environment strings from the old process address
 1057  * space into the temporary string buffer.
 1058  */
 1059 int
 1060 exec_copyin_args(struct image_args *args, char *fname,
 1061     enum uio_seg segflg, char **argv, char **envv)
 1062 {
 1063         char *argp, *envp;
 1064         int error;
 1065         size_t length;
 1066 
 1067         bzero(args, sizeof(*args));
 1068         if (argv == NULL)
 1069                 return (EFAULT);
 1070         /*
 1071          * Allocate temporary demand zeroed space for argument and
 1072          *      environment strings:
 1073          *
 1074          * o ARG_MAX for argument and environment;
 1075          * o MAXSHELLCMDLEN for the name of interpreters.
 1076          */
 1077         args->buf = (char *) kmem_alloc_wait(exec_map,
 1078             PATH_MAX + ARG_MAX + MAXSHELLCMDLEN);
 1079         if (args->buf == NULL)
 1080                 return (ENOMEM);
 1081         args->begin_argv = args->buf;
 1082         args->endp = args->begin_argv;
 1083         args->stringspace = ARG_MAX;
 1084         /*
 1085          * Copy the file name.
 1086          */
 1087         if (fname != NULL) {
 1088                 args->fname = args->buf + ARG_MAX;
 1089                 error = (segflg == UIO_SYSSPACE) ?
 1090                     copystr(fname, args->fname, PATH_MAX, &length) :
 1091                     copyinstr(fname, args->fname, PATH_MAX, &length);
 1092                 if (error != 0)
 1093                         goto err_exit;
 1094         } else
 1095                 args->fname = NULL;
 1096 
 1097         /*
 1098          * extract arguments first
 1099          */
 1100         while ((argp = (caddr_t) (intptr_t) fuword(argv++))) {
 1101                 if (argp == (caddr_t) -1) {
 1102                         error = EFAULT;
 1103                         goto err_exit;
 1104                 }
 1105                 if ((error = copyinstr(argp, args->endp,
 1106                     args->stringspace, &length))) {
 1107                         if (error == ENAMETOOLONG) 
 1108                                 error = E2BIG;
 1109                         goto err_exit;
 1110                 }
 1111                 args->stringspace -= length;
 1112                 args->endp += length;
 1113                 args->argc++;
 1114         }
 1115 
 1116         args->begin_envv = args->endp;
 1117 
 1118         /*
 1119          * extract environment strings
 1120          */
 1121         if (envv) {
 1122                 while ((envp = (caddr_t)(intptr_t)fuword(envv++))) {
 1123                         if (envp == (caddr_t)-1) {
 1124                                 error = EFAULT;
 1125                                 goto err_exit;
 1126                         }
 1127                         if ((error = copyinstr(envp, args->endp,
 1128                             args->stringspace, &length))) {
 1129                                 if (error == ENAMETOOLONG)
 1130                                         error = E2BIG;
 1131                                 goto err_exit;
 1132                         }
 1133                         args->stringspace -= length;
 1134                         args->endp += length;
 1135                         args->envc++;
 1136                 }
 1137         }
 1138 
 1139         return (0);
 1140 
 1141 err_exit:
 1142         exec_free_args(args);
 1143         return (error);
 1144 }
 1145 
 1146 static void
 1147 exec_free_args(struct image_args *args)
 1148 {
 1149 
 1150         if (args->buf) {
 1151                 kmem_free_wakeup(exec_map, (vm_offset_t)args->buf,
 1152                     PATH_MAX + ARG_MAX + MAXSHELLCMDLEN);
 1153                 args->buf = NULL;
 1154         }
 1155 }
 1156 
 1157 /*
 1158  * Copy strings out to the new process address space, constructing new arg
 1159  * and env vector tables. Return a pointer to the base so that it can be used
 1160  * as the initial stack pointer.
 1161  */
 1162 register_t *
 1163 exec_copyout_strings(imgp)
 1164         struct image_params *imgp;
 1165 {
 1166         int argc, envc;
 1167         char **vectp;
 1168         char *stringp, *destp;
 1169         register_t *stack_base;
 1170         struct ps_strings *arginfo;
 1171         struct proc *p;
 1172         size_t execpath_len;
 1173         int szsigcode;
 1174 
 1175         /*
 1176          * Calculate string base and vector table pointers.
 1177          * Also deal with signal trampoline code for this exec type.
 1178          */
 1179         if (imgp->execpath != NULL && imgp->auxargs != NULL)
 1180                 execpath_len = strlen(imgp->execpath) + 1;
 1181         else
 1182                 execpath_len = 0;
 1183         p = imgp->proc;
 1184         szsigcode = 0;
 1185         arginfo = (struct ps_strings *)p->p_sysent->sv_psstrings;
 1186         if (p->p_sysent->sv_szsigcode != NULL)
 1187                 szsigcode = *(p->p_sysent->sv_szsigcode);
 1188         destp = (caddr_t)arginfo - szsigcode - SPARE_USRSPACE -
 1189             roundup(execpath_len, sizeof(char *)) -
 1190             roundup((ARG_MAX - imgp->args->stringspace), sizeof(char *));
 1191 
 1192         /*
 1193          * install sigcode
 1194          */
 1195         if (szsigcode)
 1196                 copyout(p->p_sysent->sv_sigcode, ((caddr_t)arginfo -
 1197                     szsigcode), szsigcode);
 1198 
 1199         /*
 1200          * Copy the image path for the rtld.
 1201          */
 1202         if (execpath_len != 0) {
 1203                 imgp->execpathp = (uintptr_t)arginfo - szsigcode - execpath_len;
 1204                 copyout(imgp->execpath, (void *)imgp->execpathp,
 1205                     execpath_len);
 1206         }
 1207 
 1208         /*
 1209          * If we have a valid auxargs ptr, prepare some room
 1210          * on the stack.
 1211          */
 1212         if (imgp->auxargs) {
 1213                 /*
 1214                  * 'AT_COUNT*2' is size for the ELF Auxargs data. This is for
 1215                  * lower compatibility.
 1216                  */
 1217                 imgp->auxarg_size = (imgp->auxarg_size) ? imgp->auxarg_size :
 1218                     (AT_COUNT * 2);
 1219                 /*
 1220                  * The '+ 2' is for the null pointers at the end of each of
 1221                  * the arg and env vector sets,and imgp->auxarg_size is room
 1222                  * for argument of Runtime loader.
 1223                  */
 1224                 vectp = (char **)(destp - (imgp->args->argc +
 1225                     imgp->args->envc + 2 + imgp->auxarg_size + execpath_len) *
 1226                     sizeof(char *));
 1227         } else {
 1228                 /*
 1229                  * The '+ 2' is for the null pointers at the end of each of
 1230                  * the arg and env vector sets
 1231                  */
 1232                 vectp = (char **)(destp - (imgp->args->argc + imgp->args->envc + 2) *
 1233                     sizeof(char *));
 1234         }
 1235 
 1236         /*
 1237          * vectp also becomes our initial stack base
 1238          */
 1239         stack_base = (register_t *)vectp;
 1240 
 1241         stringp = imgp->args->begin_argv;
 1242         argc = imgp->args->argc;
 1243         envc = imgp->args->envc;
 1244 
 1245         /*
 1246          * Copy out strings - arguments and environment.
 1247          */
 1248         copyout(stringp, destp, ARG_MAX - imgp->args->stringspace);
 1249 
 1250         /*
 1251          * Fill in "ps_strings" struct for ps, w, etc.
 1252          */
 1253         suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp);
 1254         suword(&arginfo->ps_nargvstr, argc);
 1255 
 1256         /*
 1257          * Fill in argument portion of vector table.
 1258          */
 1259         for (; argc > 0; --argc) {
 1260                 suword(vectp++, (long)(intptr_t)destp);
 1261                 while (*stringp++ != 0)
 1262                         destp++;
 1263                 destp++;
 1264         }
 1265 
 1266         /* a null vector table pointer separates the argp's from the envp's */
 1267         suword(vectp++, 0);
 1268 
 1269         suword(&arginfo->ps_envstr, (long)(intptr_t)vectp);
 1270         suword(&arginfo->ps_nenvstr, envc);
 1271 
 1272         /*
 1273          * Fill in environment portion of vector table.
 1274          */
 1275         for (; envc > 0; --envc) {
 1276                 suword(vectp++, (long)(intptr_t)destp);
 1277                 while (*stringp++ != 0)
 1278                         destp++;
 1279                 destp++;
 1280         }
 1281 
 1282         /* end of vector table is a null pointer */
 1283         suword(vectp, 0);
 1284 
 1285         return (stack_base);
 1286 }
 1287 
 1288 /*
 1289  * Check permissions of file to execute.
 1290  *      Called with imgp->vp locked.
 1291  *      Return 0 for success or error code on failure.
 1292  */
 1293 int
 1294 exec_check_permissions(imgp)
 1295         struct image_params *imgp;
 1296 {
 1297         struct vnode *vp = imgp->vp;
 1298         struct vattr *attr = imgp->attr;
 1299         struct thread *td;
 1300         int error;
 1301 
 1302         td = curthread;
 1303 
 1304         /* Get file attributes */
 1305         error = VOP_GETATTR(vp, attr, td->td_ucred);
 1306         if (error)
 1307                 return (error);
 1308 
 1309 #ifdef MAC
 1310         error = mac_vnode_check_exec(td->td_ucred, imgp->vp, imgp);
 1311         if (error)
 1312                 return (error);
 1313 #endif
 1314         
 1315         /*
 1316          * 1) Check if file execution is disabled for the filesystem that this
 1317          *      file resides on.
 1318          * 2) Insure that at least one execute bit is on - otherwise root
 1319          *      will always succeed, and we don't want to happen unless the
 1320          *      file really is executable.
 1321          * 3) Insure that the file is a regular file.
 1322          */
 1323         if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
 1324             ((attr->va_mode & 0111) == 0) ||
 1325             (attr->va_type != VREG))
 1326                 return (EACCES);
 1327 
 1328         /*
 1329          * Zero length files can't be exec'd
 1330          */
 1331         if (attr->va_size == 0)
 1332                 return (ENOEXEC);
 1333 
 1334         /*
 1335          *  Check for execute permission to file based on current credentials.
 1336          */
 1337         error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
 1338         if (error)
 1339                 return (error);
 1340 
 1341         /*
 1342          * Check number of open-for-writes on the file and deny execution
 1343          * if there are any.
 1344          */
 1345         if (vp->v_writecount)
 1346                 return (ETXTBSY);
 1347 
 1348         /*
 1349          * Call filesystem specific open routine (which does nothing in the
 1350          * general case).
 1351          */
 1352         error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
 1353         if (error == 0)
 1354                 imgp->opened = 1;
 1355         return (error);
 1356 }
 1357 
 1358 /*
 1359  * Exec handler registration
 1360  */
 1361 int
 1362 exec_register(execsw_arg)
 1363         const struct execsw *execsw_arg;
 1364 {
 1365         const struct execsw **es, **xs, **newexecsw;
 1366         int count = 2;  /* New slot and trailing NULL */
 1367 
 1368         if (execsw)
 1369                 for (es = execsw; *es; es++)
 1370                         count++;
 1371         newexecsw = malloc(count * sizeof(*es), M_TEMP, M_WAITOK);
 1372         if (newexecsw == NULL)
 1373                 return (ENOMEM);
 1374         xs = newexecsw;
 1375         if (execsw)
 1376                 for (es = execsw; *es; es++)
 1377                         *xs++ = *es;
 1378         *xs++ = execsw_arg;
 1379         *xs = NULL;
 1380         if (execsw)
 1381                 free(execsw, M_TEMP);
 1382         execsw = newexecsw;
 1383         return (0);
 1384 }
 1385 
 1386 int
 1387 exec_unregister(execsw_arg)
 1388         const struct execsw *execsw_arg;
 1389 {
 1390         const struct execsw **es, **xs, **newexecsw;
 1391         int count = 1;
 1392 
 1393         if (execsw == NULL)
 1394                 panic("unregister with no handlers left?\n");
 1395 
 1396         for (es = execsw; *es; es++) {
 1397                 if (*es == execsw_arg)
 1398                         break;
 1399         }
 1400         if (*es == NULL)
 1401                 return (ENOENT);
 1402         for (es = execsw; *es; es++)
 1403                 if (*es != execsw_arg)
 1404                         count++;
 1405         newexecsw = malloc(count * sizeof(*es), M_TEMP, M_WAITOK);
 1406         if (newexecsw == NULL)
 1407                 return (ENOMEM);
 1408         xs = newexecsw;
 1409         for (es = execsw; *es; es++)
 1410                 if (*es != execsw_arg)
 1411                         *xs++ = *es;
 1412         *xs = NULL;
 1413         if (execsw)
 1414                 free(execsw, M_TEMP);
 1415         execsw = newexecsw;
 1416         return (0);
 1417 }

Cache object: 309f2f3e423df1200f71a5a6a7c91d11


[ 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.