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/8.4/sys/kern/kern_exec.c 267019 2014-06-03 19:03:23Z delphij $");
   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/sched.h>
   58 #include <sys/sdt.h>
   59 #include <sys/sf_buf.h>
   60 #include <sys/syscallsubr.h>
   61 #include <sys/sysent.h>
   62 #include <sys/shm.h>
   63 #include <sys/sysctl.h>
   64 #include <sys/vnode.h>
   65 #include <sys/stat.h>
   66 #ifdef KTRACE
   67 #include <sys/ktrace.h>
   68 #endif
   69 
   70 #include <vm/vm.h>
   71 #include <vm/vm_param.h>
   72 #include <vm/pmap.h>
   73 #include <vm/vm_page.h>
   74 #include <vm/vm_map.h>
   75 #include <vm/vm_kern.h>
   76 #include <vm/vm_extern.h>
   77 #include <vm/vm_object.h>
   78 #include <vm/vm_pager.h>
   79 
   80 #ifdef  HWPMC_HOOKS
   81 #include <sys/pmckern.h>
   82 #endif
   83 
   84 #include <machine/reg.h>
   85 
   86 #include <security/audit/audit.h>
   87 #include <security/mac/mac_framework.h>
   88 
   89 #ifdef KDTRACE_HOOKS
   90 #include <sys/dtrace_bsd.h>
   91 dtrace_execexit_func_t  dtrace_fasttrap_exec;
   92 #endif
   93 
   94 SDT_PROVIDER_DECLARE(proc);
   95 SDT_PROBE_DEFINE(proc, kernel, , exec, exec);
   96 SDT_PROBE_ARGTYPE(proc, kernel, , exec, 0, "char *");
   97 SDT_PROBE_DEFINE(proc, kernel, , exec_failure, exec-failure);
   98 SDT_PROBE_ARGTYPE(proc, kernel, , exec_failure, 0, "int");
   99 SDT_PROBE_DEFINE(proc, kernel, , exec_success, exec-success);
  100 SDT_PROBE_ARGTYPE(proc, kernel, , exec_success, 0, "char *");
  101 
  102 MALLOC_DEFINE(M_PARGS, "proc-args", "Process arguments");
  103 
  104 static int sysctl_kern_ps_strings(SYSCTL_HANDLER_ARGS);
  105 static int sysctl_kern_usrstack(SYSCTL_HANDLER_ARGS);
  106 static int sysctl_kern_stackprot(SYSCTL_HANDLER_ARGS);
  107 static int do_execve(struct thread *td, struct image_args *args,
  108     struct mac *mac_p);
  109 static void exec_free_args(struct image_args *);
  110 
  111 /* XXX This should be vm_size_t. */
  112 SYSCTL_PROC(_kern, KERN_PS_STRINGS, ps_strings, CTLTYPE_ULONG|CTLFLAG_RD,
  113     NULL, 0, sysctl_kern_ps_strings, "LU", "");
  114 
  115 /* XXX This should be vm_size_t. */
  116 SYSCTL_PROC(_kern, KERN_USRSTACK, usrstack, CTLTYPE_ULONG|CTLFLAG_RD,
  117     NULL, 0, sysctl_kern_usrstack, "LU", "");
  118 
  119 SYSCTL_PROC(_kern, OID_AUTO, stackprot, CTLTYPE_INT|CTLFLAG_RD,
  120     NULL, 0, sysctl_kern_stackprot, "I", "");
  121 
  122 u_long ps_arg_cache_limit = PAGE_SIZE / 16;
  123 SYSCTL_ULONG(_kern, OID_AUTO, ps_arg_cache_limit, CTLFLAG_RW, 
  124     &ps_arg_cache_limit, 0, "");
  125 
  126 static int map_at_zero = 0;
  127 TUNABLE_INT("security.bsd.map_at_zero", &map_at_zero);
  128 SYSCTL_INT(_security_bsd, OID_AUTO, map_at_zero, CTLFLAG_RW, &map_at_zero, 0,
  129     "Permit processes to map an object at virtual address 0.");
  130 
  131 static int
  132 sysctl_kern_ps_strings(SYSCTL_HANDLER_ARGS)
  133 {
  134         struct proc *p;
  135         int error;
  136 
  137         p = curproc;
  138 #ifdef SCTL_MASK32
  139         if (req->flags & SCTL_MASK32) {
  140                 unsigned int val;
  141                 val = (unsigned int)p->p_sysent->sv_psstrings;
  142                 error = SYSCTL_OUT(req, &val, sizeof(val));
  143         } else
  144 #endif
  145                 error = SYSCTL_OUT(req, &p->p_sysent->sv_psstrings,
  146                    sizeof(p->p_sysent->sv_psstrings));
  147         return error;
  148 }
  149 
  150 static int
  151 sysctl_kern_usrstack(SYSCTL_HANDLER_ARGS)
  152 {
  153         struct proc *p;
  154         int error;
  155 
  156         p = curproc;
  157 #ifdef SCTL_MASK32
  158         if (req->flags & SCTL_MASK32) {
  159                 unsigned int val;
  160                 val = (unsigned int)p->p_sysent->sv_usrstack;
  161                 error = SYSCTL_OUT(req, &val, sizeof(val));
  162         } else
  163 #endif
  164                 error = SYSCTL_OUT(req, &p->p_sysent->sv_usrstack,
  165                     sizeof(p->p_sysent->sv_usrstack));
  166         return error;
  167 }
  168 
  169 static int
  170 sysctl_kern_stackprot(SYSCTL_HANDLER_ARGS)
  171 {
  172         struct proc *p;
  173 
  174         p = curproc;
  175         return (SYSCTL_OUT(req, &p->p_sysent->sv_stackprot,
  176             sizeof(p->p_sysent->sv_stackprot)));
  177 }
  178 
  179 /*
  180  * Each of the items is a pointer to a `const struct execsw', hence the
  181  * double pointer here.
  182  */
  183 static const struct execsw **execsw;
  184 
  185 #ifndef _SYS_SYSPROTO_H_
  186 struct execve_args {
  187         char    *fname; 
  188         char    **argv;
  189         char    **envv; 
  190 };
  191 #endif
  192 
  193 int
  194 execve(td, uap)
  195         struct thread *td;
  196         struct execve_args /* {
  197                 char *fname;
  198                 char **argv;
  199                 char **envv;
  200         } */ *uap;
  201 {
  202         int error;
  203         struct image_args args;
  204 
  205         error = exec_copyin_args(&args, uap->fname, UIO_USERSPACE,
  206             uap->argv, uap->envv);
  207         if (error == 0)
  208                 error = kern_execve(td, &args, NULL);
  209         return (error);
  210 }
  211 
  212 #ifndef _SYS_SYSPROTO_H_
  213 struct fexecve_args {
  214         int     fd;
  215         char    **argv;
  216         char    **envv;
  217 }
  218 #endif
  219 int
  220 fexecve(struct thread *td, struct fexecve_args *uap)
  221 {
  222         int error;
  223         struct image_args args;
  224 
  225         error = exec_copyin_args(&args, NULL, UIO_SYSSPACE,
  226             uap->argv, uap->envv);
  227         if (error == 0) {
  228                 args.fd = uap->fd;
  229                 error = kern_execve(td, &args, NULL);
  230         }
  231         return (error);
  232 }
  233 
  234 #ifndef _SYS_SYSPROTO_H_
  235 struct __mac_execve_args {
  236         char    *fname;
  237         char    **argv;
  238         char    **envv;
  239         struct mac      *mac_p;
  240 };
  241 #endif
  242 
  243 int
  244 __mac_execve(td, uap)
  245         struct thread *td;
  246         struct __mac_execve_args /* {
  247                 char *fname;
  248                 char **argv;
  249                 char **envv;
  250                 struct mac *mac_p;
  251         } */ *uap;
  252 {
  253 #ifdef MAC
  254         int error;
  255         struct image_args args;
  256 
  257         error = exec_copyin_args(&args, uap->fname, UIO_USERSPACE,
  258             uap->argv, uap->envv);
  259         if (error == 0)
  260                 error = kern_execve(td, &args, uap->mac_p);
  261         return (error);
  262 #else
  263         return (ENOSYS);
  264 #endif
  265 }
  266 
  267 /*
  268  * XXX: kern_execve has the astonishing property of not always returning to
  269  * the caller.  If sufficiently bad things happen during the call to
  270  * do_execve(), it can end up calling exit1(); as a result, callers must
  271  * avoid doing anything which they might need to undo (e.g., allocating
  272  * memory).
  273  */
  274 int
  275 kern_execve(td, args, mac_p)
  276         struct thread *td;
  277         struct image_args *args;
  278         struct mac *mac_p;
  279 {
  280         struct proc *p = td->td_proc;
  281         struct vmspace *oldvmspace;
  282         int error;
  283 
  284         AUDIT_ARG_ARGV(args->begin_argv, args->argc,
  285             args->begin_envv - args->begin_argv);
  286         AUDIT_ARG_ENVV(args->begin_envv, args->envc,
  287             args->endp - args->begin_envv);
  288         if (p->p_flag & P_HADTHREADS) {
  289                 PROC_LOCK(p);
  290                 if (thread_single(SINGLE_BOUNDARY)) {
  291                         PROC_UNLOCK(p);
  292                         exec_free_args(args);
  293                         return (ERESTART);      /* Try again later. */
  294                 }
  295                 PROC_UNLOCK(p);
  296         }
  297 
  298         KASSERT((td->td_pflags & TDP_EXECVMSPC) == 0, ("nested execve"));
  299         oldvmspace = td->td_proc->p_vmspace;
  300         error = do_execve(td, args, mac_p);
  301 
  302         if (p->p_flag & P_HADTHREADS) {
  303                 PROC_LOCK(p);
  304                 /*
  305                  * If success, we upgrade to SINGLE_EXIT state to
  306                  * force other threads to suicide.
  307                  */
  308                 if (error == 0)
  309                         thread_single(SINGLE_EXIT);
  310                 else
  311                         thread_single_end();
  312                 PROC_UNLOCK(p);
  313         }
  314         if ((td->td_pflags & TDP_EXECVMSPC) != 0) {
  315                 KASSERT(td->td_proc->p_vmspace != oldvmspace,
  316                     ("oldvmspace still used"));
  317                 vmspace_free(oldvmspace);
  318                 td->td_pflags &= ~TDP_EXECVMSPC;
  319         }
  320 
  321         return (error);
  322 }
  323 
  324 /*
  325  * In-kernel implementation of execve().  All arguments are assumed to be
  326  * userspace pointers from the passed thread.
  327  */
  328 static int
  329 do_execve(td, args, mac_p)
  330         struct thread *td;
  331         struct image_args *args;
  332         struct mac *mac_p;
  333 {
  334         struct proc *p = td->td_proc;
  335         struct nameidata nd;
  336         struct ucred *newcred = NULL, *oldcred;
  337         struct uidinfo *euip;
  338         register_t *stack_base;
  339         int error, i;
  340         struct image_params image_params, *imgp;
  341         struct vattr attr;
  342         int (*img_first)(struct image_params *);
  343         struct pargs *oldargs = NULL, *newargs = NULL;
  344         struct sigacts *oldsigacts, *newsigacts;
  345 #ifdef KTRACE
  346         struct vnode *tracevp = NULL;
  347         struct ucred *tracecred = NULL;
  348 #endif
  349         struct vnode *textvp = NULL, *binvp = NULL;
  350         int credential_changing;
  351         int vfslocked;
  352         int textset;
  353 #ifdef MAC
  354         struct label *interpvplabel = NULL;
  355         int will_transition;
  356 #endif
  357 #ifdef HWPMC_HOOKS
  358         struct pmckern_procexec pe;
  359 #endif
  360         static const char fexecv_proc_title[] = "(fexecv)";
  361 
  362         vfslocked = 0;
  363         imgp = &image_params;
  364 
  365         /*
  366          * Lock the process and set the P_INEXEC flag to indicate that
  367          * it should be left alone until we're done here.  This is
  368          * necessary to avoid race conditions - e.g. in ptrace() -
  369          * that might allow a local user to illicitly obtain elevated
  370          * privileges.
  371          */
  372         PROC_LOCK(p);
  373         KASSERT((p->p_flag & P_INEXEC) == 0,
  374             ("%s(): process already has P_INEXEC flag", __func__));
  375         p->p_flag |= P_INEXEC;
  376         PROC_UNLOCK(p);
  377 
  378         /*
  379          * Initialize part of the common data
  380          */
  381         imgp->proc = p;
  382         imgp->execlabel = NULL;
  383         imgp->attr = &attr;
  384         imgp->entry_addr = 0;
  385         imgp->vmspace_destroyed = 0;
  386         imgp->interpreted = 0;
  387         imgp->opened = 0;
  388         imgp->interpreter_name = args->buf + PATH_MAX + ARG_MAX;
  389         imgp->auxargs = NULL;
  390         imgp->vp = NULL;
  391         imgp->object = NULL;
  392         imgp->firstpage = NULL;
  393         imgp->ps_strings = 0;
  394         imgp->auxarg_size = 0;
  395         imgp->args = args;
  396         imgp->execpath = imgp->freepath = NULL;
  397         imgp->execpathp = 0;
  398 
  399 #ifdef MAC
  400         error = mac_execve_enter(imgp, mac_p);
  401         if (error)
  402                 goto exec_fail;
  403 #endif
  404 
  405         imgp->image_header = NULL;
  406 
  407         /*
  408          * Translate the file name. namei() returns a vnode pointer
  409          *      in ni_vp amoung other things.
  410          *
  411          * XXXAUDIT: It would be desirable to also audit the name of the
  412          * interpreter if this is an interpreted binary.
  413          */
  414         if (args->fname != NULL) {
  415                 NDINIT(&nd, LOOKUP, ISOPEN | LOCKLEAF | FOLLOW | SAVENAME
  416                     | MPSAFE | AUDITVNODE1, UIO_SYSSPACE, args->fname, td);
  417         }
  418 
  419         SDT_PROBE(proc, kernel, , exec, args->fname, 0, 0, 0, 0 );
  420 
  421 interpret:
  422         if (args->fname != NULL) {
  423                 error = namei(&nd);
  424                 if (error)
  425                         goto exec_fail;
  426 
  427                 vfslocked = NDHASGIANT(&nd);
  428                 binvp  = nd.ni_vp;
  429                 imgp->vp = binvp;
  430         } else {
  431                 AUDIT_ARG_FD(args->fd);
  432                 error = fgetvp(td, args->fd, &binvp);
  433                 if (error)
  434                         goto exec_fail;
  435                 vfslocked = VFS_LOCK_GIANT(binvp->v_mount);
  436                 vn_lock(binvp, LK_EXCLUSIVE | LK_RETRY);
  437                 AUDIT_ARG_VNODE1(binvp);
  438                 imgp->vp = binvp;
  439         }
  440 
  441         /*
  442          * Check file permissions (also 'opens' file)
  443          */
  444         error = exec_check_permissions(imgp);
  445         if (error)
  446                 goto exec_fail_dealloc;
  447 
  448         imgp->object = imgp->vp->v_object;
  449         if (imgp->object != NULL)
  450                 vm_object_reference(imgp->object);
  451 
  452         /*
  453          * Set VV_TEXT now so no one can write to the executable while we're
  454          * activating it.
  455          *
  456          * Remember if this was set before and unset it in case this is not
  457          * actually an executable image.
  458          */
  459         textset = imgp->vp->v_vflag & VV_TEXT;
  460         imgp->vp->v_vflag |= VV_TEXT;
  461 
  462         error = exec_map_first_page(imgp);
  463         if (error)
  464                 goto exec_fail_dealloc;
  465 
  466         imgp->proc->p_osrel = 0;
  467         /*
  468          *      If the current process has a special image activator it
  469          *      wants to try first, call it.   For example, emulating shell
  470          *      scripts differently.
  471          */
  472         error = -1;
  473         if ((img_first = imgp->proc->p_sysent->sv_imgact_try) != NULL)
  474                 error = img_first(imgp);
  475 
  476         /*
  477          *      Loop through the list of image activators, calling each one.
  478          *      An activator returns -1 if there is no match, 0 on success,
  479          *      and an error otherwise.
  480          */
  481         for (i = 0; error == -1 && execsw[i]; ++i) {
  482                 if (execsw[i]->ex_imgact == NULL ||
  483                     execsw[i]->ex_imgact == img_first) {
  484                         continue;
  485                 }
  486                 error = (*execsw[i]->ex_imgact)(imgp);
  487         }
  488 
  489         if (error) {
  490                 if (error == -1) {
  491                         if (textset == 0)
  492                                 imgp->vp->v_vflag &= ~VV_TEXT;
  493                         error = ENOEXEC;
  494                 }
  495                 goto exec_fail_dealloc;
  496         }
  497 
  498         /*
  499          * Special interpreter operation, cleanup and loop up to try to
  500          * activate the interpreter.
  501          */
  502         if (imgp->interpreted) {
  503                 exec_unmap_first_page(imgp);
  504                 /*
  505                  * VV_TEXT needs to be unset for scripts.  There is a short
  506                  * period before we determine that something is a script where
  507                  * VV_TEXT will be set. The vnode lock is held over this
  508                  * entire period so nothing should illegitimately be blocked.
  509                  */
  510                 imgp->vp->v_vflag &= ~VV_TEXT;
  511                 /* free name buffer and old vnode */
  512                 if (args->fname != NULL)
  513                         NDFREE(&nd, NDF_ONLY_PNBUF);
  514 #ifdef MAC
  515                 mac_execve_interpreter_enter(binvp, &interpvplabel);
  516 #endif
  517                 if (imgp->opened) {
  518                         VOP_CLOSE(binvp, FREAD, td->td_ucred, td);
  519                         imgp->opened = 0;
  520                 }
  521                 vput(binvp);
  522                 vm_object_deallocate(imgp->object);
  523                 imgp->object = NULL;
  524                 VFS_UNLOCK_GIANT(vfslocked);
  525                 vfslocked = 0;
  526                 /* set new name to that of the interpreter */
  527                 NDINIT(&nd, LOOKUP, LOCKLEAF | FOLLOW | SAVENAME | MPSAFE,
  528                     UIO_SYSSPACE, imgp->interpreter_name, td);
  529                 args->fname = imgp->interpreter_name;
  530                 goto interpret;
  531         }
  532 
  533         /*
  534          * NB: We unlock the vnode here because it is believed that none
  535          * of the sv_copyout_strings/sv_fixup operations require the vnode.
  536          */
  537         VOP_UNLOCK(imgp->vp, 0);
  538 
  539         /*
  540          * Do the best to calculate the full path to the image file.
  541          */
  542         if (imgp->auxargs != NULL &&
  543             ((args->fname != NULL && args->fname[0] == '/') ||
  544              vn_fullpath(td, imgp->vp, &imgp->execpath, &imgp->freepath) != 0))
  545                 imgp->execpath = args->fname;
  546 
  547         /*
  548          * Copy out strings (args and env) and initialize stack base
  549          */
  550         if (p->p_sysent->sv_copyout_strings)
  551                 stack_base = (*p->p_sysent->sv_copyout_strings)(imgp);
  552         else
  553                 stack_base = exec_copyout_strings(imgp);
  554 
  555         /*
  556          * If custom stack fixup routine present for this process
  557          * let it do the stack setup.
  558          * Else stuff argument count as first item on stack
  559          */
  560         if (p->p_sysent->sv_fixup != NULL)
  561                 (*p->p_sysent->sv_fixup)(&stack_base, imgp);
  562         else
  563                 suword(--stack_base, imgp->args->argc);
  564 
  565         /*
  566          * For security and other reasons, the file descriptor table cannot
  567          * be shared after an exec.
  568          */
  569         fdunshare(p, td);
  570 
  571         /*
  572          * Malloc things before we need locks.
  573          */
  574         newcred = crget();
  575         euip = uifind(attr.va_uid);
  576         i = imgp->args->begin_envv - imgp->args->begin_argv;
  577         /* Cache arguments if they fit inside our allowance */
  578         if (ps_arg_cache_limit >= i + sizeof(struct pargs)) {
  579                 newargs = pargs_alloc(i);
  580                 bcopy(imgp->args->begin_argv, newargs->ar_args, i);
  581         }
  582 
  583         /* close files on exec */
  584         fdcloseexec(td);
  585         vn_lock(imgp->vp, LK_EXCLUSIVE | LK_RETRY);
  586 
  587         /* Get a reference to the vnode prior to locking the proc */
  588         VREF(binvp);
  589 
  590         /*
  591          * For security and other reasons, signal handlers cannot
  592          * be shared after an exec. The new process gets a copy of the old
  593          * handlers. In execsigs(), the new process will have its signals
  594          * reset.
  595          */
  596         PROC_LOCK(p);
  597         oldcred = crcopysafe(p, newcred);
  598         if (sigacts_shared(p->p_sigacts)) {
  599                 oldsigacts = p->p_sigacts;
  600                 PROC_UNLOCK(p);
  601                 newsigacts = sigacts_alloc();
  602                 sigacts_copy(newsigacts, oldsigacts);
  603                 PROC_LOCK(p);
  604                 p->p_sigacts = newsigacts;
  605         } else
  606                 oldsigacts = NULL;
  607 
  608         /* Stop profiling */
  609         stopprofclock(p);
  610 
  611         /* reset caught signals */
  612         execsigs(p);
  613 
  614         /* name this process - nameiexec(p, ndp) */
  615         bzero(p->p_comm, sizeof(p->p_comm));
  616         if (args->fname)
  617                 bcopy(nd.ni_cnd.cn_nameptr, p->p_comm,
  618                     min(nd.ni_cnd.cn_namelen, MAXCOMLEN));
  619         else if (vn_commname(binvp, p->p_comm, sizeof(p->p_comm)) != 0)
  620                 bcopy(fexecv_proc_title, p->p_comm, sizeof(fexecv_proc_title));
  621         bcopy(p->p_comm, td->td_name, sizeof(td->td_name));
  622 #ifdef KTR
  623         sched_clear_tdname(td);
  624 #endif
  625 
  626         /*
  627          * mark as execed, wakeup the process that vforked (if any) and tell
  628          * it that it now has its own resources back
  629          */
  630         p->p_flag |= P_EXEC;
  631         if (p->p_pptr && (p->p_flag & P_PPWAIT)) {
  632                 p->p_flag &= ~P_PPWAIT;
  633                 cv_broadcast(&p->p_pwait);
  634         }
  635 
  636         /*
  637          * Implement image setuid/setgid.
  638          *
  639          * Don't honor setuid/setgid if the filesystem prohibits it or if
  640          * the process is being traced.
  641          *
  642          * XXXMAC: For the time being, use NOSUID to also prohibit
  643          * transitions on the file system.
  644          */
  645         credential_changing = 0;
  646         credential_changing |= (attr.va_mode & S_ISUID) && oldcred->cr_uid !=
  647             attr.va_uid;
  648         credential_changing |= (attr.va_mode & S_ISGID) && oldcred->cr_gid !=
  649             attr.va_gid;
  650 #ifdef MAC
  651         will_transition = mac_vnode_execve_will_transition(oldcred, imgp->vp,
  652             interpvplabel, imgp);
  653         credential_changing |= will_transition;
  654 #endif
  655 
  656         if (credential_changing &&
  657             (imgp->vp->v_mount->mnt_flag & MNT_NOSUID) == 0 &&
  658             (p->p_flag & P_TRACED) == 0) {
  659                 /*
  660                  * Turn off syscall tracing for set-id programs, except for
  661                  * root.  Record any set-id flags first to make sure that
  662                  * we do not regain any tracing during a possible block.
  663                  */
  664                 setsugid(p);
  665 
  666 #ifdef KTRACE
  667                 if (p->p_tracecred != NULL &&
  668                     priv_check_cred(p->p_tracecred, PRIV_DEBUG_DIFFCRED, 0))
  669                         ktrprocexec(p, &tracecred, &tracevp);
  670 #endif
  671                 /*
  672                  * Close any file descriptors 0..2 that reference procfs,
  673                  * then make sure file descriptors 0..2 are in use.
  674                  *
  675                  * setugidsafety() may call closef() and then pfind()
  676                  * which may grab the process lock.
  677                  * fdcheckstd() may call falloc() which may block to
  678                  * allocate memory, so temporarily drop the process lock.
  679                  */
  680                 PROC_UNLOCK(p);
  681                 VOP_UNLOCK(imgp->vp, 0);
  682                 setugidsafety(td);
  683                 error = fdcheckstd(td);
  684                 vn_lock(imgp->vp, LK_EXCLUSIVE | LK_RETRY);
  685                 if (error != 0)
  686                         goto done1;
  687                 PROC_LOCK(p);
  688                 /*
  689                  * Set the new credentials.
  690                  */
  691                 if (attr.va_mode & S_ISUID)
  692                         change_euid(newcred, euip);
  693                 if (attr.va_mode & S_ISGID)
  694                         change_egid(newcred, attr.va_gid);
  695 #ifdef MAC
  696                 if (will_transition) {
  697                         mac_vnode_execve_transition(oldcred, newcred, imgp->vp,
  698                             interpvplabel, imgp);
  699                 }
  700 #endif
  701                 /*
  702                  * Implement correct POSIX saved-id behavior.
  703                  *
  704                  * XXXMAC: Note that the current logic will save the
  705                  * uid and gid if a MAC domain transition occurs, even
  706                  * though maybe it shouldn't.
  707                  */
  708                 change_svuid(newcred, newcred->cr_uid);
  709                 change_svgid(newcred, newcred->cr_gid);
  710                 p->p_ucred = newcred;
  711                 newcred = NULL;
  712         } else {
  713                 if (oldcred->cr_uid == oldcred->cr_ruid &&
  714                     oldcred->cr_gid == oldcred->cr_rgid)
  715                         p->p_flag &= ~P_SUGID;
  716                 /*
  717                  * Implement correct POSIX saved-id behavior.
  718                  *
  719                  * XXX: It's not clear that the existing behavior is
  720                  * POSIX-compliant.  A number of sources indicate that the
  721                  * saved uid/gid should only be updated if the new ruid is
  722                  * not equal to the old ruid, or the new euid is not equal
  723                  * to the old euid and the new euid is not equal to the old
  724                  * ruid.  The FreeBSD code always updates the saved uid/gid.
  725                  * Also, this code uses the new (replaced) euid and egid as
  726                  * the source, which may or may not be the right ones to use.
  727                  */
  728                 if (oldcred->cr_svuid != oldcred->cr_uid ||
  729                     oldcred->cr_svgid != oldcred->cr_gid) {
  730                         change_svuid(newcred, newcred->cr_uid);
  731                         change_svgid(newcred, newcred->cr_gid);
  732                         p->p_ucred = newcred;
  733                         newcred = NULL;
  734                 }
  735         }
  736 
  737         /*
  738          * Store the vp for use in procfs.  This vnode was referenced prior
  739          * to locking the proc lock.
  740          */
  741         textvp = p->p_textvp;
  742         p->p_textvp = binvp;
  743 
  744 #ifdef KDTRACE_HOOKS
  745         /*
  746          * Tell the DTrace fasttrap provider about the exec if it
  747          * has declared an interest.
  748          */
  749         if (dtrace_fasttrap_exec)
  750                 dtrace_fasttrap_exec(p);
  751 #endif
  752 
  753         /*
  754          * Notify others that we exec'd, and clear the P_INEXEC flag
  755          * as we're now a bona fide freshly-execed process.
  756          */
  757         KNOTE_LOCKED(&p->p_klist, NOTE_EXEC);
  758         p->p_flag &= ~P_INEXEC;
  759 
  760         /* clear "fork but no exec" flag, as we _are_ execing */
  761         p->p_acflag &= ~AFORK;
  762 
  763         /*
  764          * Free any previous argument cache and replace it with
  765          * the new argument cache, if any.
  766          */
  767         oldargs = p->p_args;
  768         p->p_args = newargs;
  769         newargs = NULL;
  770 
  771 #ifdef  HWPMC_HOOKS
  772         /*
  773          * Check if system-wide sampling is in effect or if the
  774          * current process is using PMCs.  If so, do exec() time
  775          * processing.  This processing needs to happen AFTER the
  776          * P_INEXEC flag is cleared.
  777          *
  778          * The proc lock needs to be released before taking the PMC
  779          * SX.
  780          */
  781         if (PMC_SYSTEM_SAMPLING_ACTIVE() || PMC_PROC_IS_USING_PMCS(p)) {
  782                 PROC_UNLOCK(p);
  783                 VOP_UNLOCK(imgp->vp, 0);
  784                 pe.pm_credentialschanged = credential_changing;
  785                 pe.pm_entryaddr = imgp->entry_addr;
  786 
  787                 PMC_CALL_HOOK_X(td, PMC_FN_PROCESS_EXEC, (void *) &pe);
  788                 vn_lock(imgp->vp, LK_EXCLUSIVE | LK_RETRY);
  789         } else
  790                 PROC_UNLOCK(p);
  791 #else  /* !HWPMC_HOOKS */
  792         PROC_UNLOCK(p);
  793 #endif
  794 
  795         /* Set values passed into the program in registers. */
  796         if (p->p_sysent->sv_setregs)
  797                 (*p->p_sysent->sv_setregs)(td, imgp->entry_addr,
  798                     (u_long)(uintptr_t)stack_base, imgp->ps_strings);
  799         else
  800                 exec_setregs(td, imgp->entry_addr,
  801                     (u_long)(uintptr_t)stack_base, imgp->ps_strings);
  802 
  803         vfs_mark_atime(imgp->vp, td->td_ucred);
  804 
  805         SDT_PROBE(proc, kernel, , exec_success, args->fname, 0, 0, 0, 0);
  806 
  807 done1:
  808         /*
  809          * Free any resources malloc'd earlier that we didn't use.
  810          */
  811         uifree(euip);
  812         if (newcred == NULL)
  813                 crfree(oldcred);
  814         else
  815                 crfree(newcred);
  816         VOP_UNLOCK(imgp->vp, 0);
  817 
  818         /*
  819          * Handle deferred decrement of ref counts.
  820          */
  821         if (textvp != NULL) {
  822                 int tvfslocked;
  823 
  824                 tvfslocked = VFS_LOCK_GIANT(textvp->v_mount);
  825                 vrele(textvp);
  826                 VFS_UNLOCK_GIANT(tvfslocked);
  827         }
  828         if (binvp && error != 0)
  829                 vrele(binvp);
  830 #ifdef KTRACE
  831         if (tracevp != NULL) {
  832                 int tvfslocked;
  833 
  834                 tvfslocked = VFS_LOCK_GIANT(tracevp->v_mount);
  835                 vrele(tracevp);
  836                 VFS_UNLOCK_GIANT(tvfslocked);
  837         }
  838         if (tracecred != NULL)
  839                 crfree(tracecred);
  840 #endif
  841         vn_lock(imgp->vp, LK_EXCLUSIVE | LK_RETRY);
  842         pargs_drop(oldargs);
  843         pargs_drop(newargs);
  844         if (oldsigacts != NULL)
  845                 sigacts_free(oldsigacts);
  846 
  847 exec_fail_dealloc:
  848 
  849         /*
  850          * free various allocated resources
  851          */
  852         if (imgp->firstpage != NULL)
  853                 exec_unmap_first_page(imgp);
  854 
  855         if (imgp->vp != NULL) {
  856                 if (args->fname)
  857                         NDFREE(&nd, NDF_ONLY_PNBUF);
  858                 if (imgp->opened)
  859                         VOP_CLOSE(imgp->vp, FREAD, td->td_ucred, td);
  860                 vput(imgp->vp);
  861         }
  862 
  863         if (imgp->object != NULL)
  864                 vm_object_deallocate(imgp->object);
  865 
  866         free(imgp->freepath, M_TEMP);
  867 
  868         if (error == 0) {
  869                 PROC_LOCK(p);
  870                 td->td_dbgflags |= TDB_EXEC;
  871                 PROC_UNLOCK(p);
  872 
  873                 /*
  874                  * Stop the process here if its stop event mask has
  875                  * the S_EXEC bit set.
  876                  */
  877                 STOPEVENT(p, S_EXEC, 0);
  878                 goto done2;
  879         }
  880 
  881 exec_fail:
  882         /* we're done here, clear P_INEXEC */
  883         PROC_LOCK(p);
  884         p->p_flag &= ~P_INEXEC;
  885         PROC_UNLOCK(p);
  886 
  887         SDT_PROBE(proc, kernel, , exec_failure, error, 0, 0, 0, 0);
  888 
  889 done2:
  890 #ifdef MAC
  891         mac_execve_exit(imgp);
  892         mac_execve_interpreter_exit(interpvplabel);
  893 #endif
  894         VFS_UNLOCK_GIANT(vfslocked);
  895         exec_free_args(args);
  896 
  897         if (error && imgp->vmspace_destroyed) {
  898                 /* sorry, no more process anymore. exit gracefully */
  899                 exit1(td, W_EXITCODE(0, SIGABRT));
  900                 /* NOT REACHED */
  901         }
  902 
  903 #ifdef KTRACE
  904         if (error == 0)
  905                 ktrprocctor(p);
  906 #endif
  907 
  908         return (error);
  909 }
  910 
  911 int
  912 exec_map_first_page(imgp)
  913         struct image_params *imgp;
  914 {
  915         int rv, i;
  916         int initial_pagein;
  917         vm_page_t ma[VM_INITIAL_PAGEIN];
  918         vm_object_t object;
  919 
  920         if (imgp->firstpage != NULL)
  921                 exec_unmap_first_page(imgp);
  922 
  923         object = imgp->vp->v_object;
  924         if (object == NULL)
  925                 return (EACCES);
  926         VM_OBJECT_LOCK(object);
  927 #if VM_NRESERVLEVEL > 0
  928         if ((object->flags & OBJ_COLORED) == 0) {
  929                 object->flags |= OBJ_COLORED;
  930                 object->pg_color = 0;
  931         }
  932 #endif
  933         ma[0] = vm_page_grab(object, 0, VM_ALLOC_NORMAL | VM_ALLOC_RETRY);
  934         if (ma[0]->valid != VM_PAGE_BITS_ALL) {
  935                 initial_pagein = VM_INITIAL_PAGEIN;
  936                 if (initial_pagein > object->size)
  937                         initial_pagein = object->size;
  938                 for (i = 1; i < initial_pagein; i++) {
  939                         if ((ma[i] = vm_page_next(ma[i - 1])) != NULL) {
  940                                 if (ma[i]->valid)
  941                                         break;
  942                                 if ((ma[i]->oflags & VPO_BUSY) || ma[i]->busy)
  943                                         break;
  944                                 vm_page_busy(ma[i]);
  945                         } else {
  946                                 ma[i] = vm_page_alloc(object, i,
  947                                     VM_ALLOC_NORMAL | VM_ALLOC_IFNOTCACHED);
  948                                 if (ma[i] == NULL)
  949                                         break;
  950                         }
  951                 }
  952                 initial_pagein = i;
  953                 rv = vm_pager_get_pages(object, ma, initial_pagein, 0);
  954                 ma[0] = vm_page_lookup(object, 0);
  955                 if ((rv != VM_PAGER_OK) || (ma[0] == NULL)) {
  956                         if (ma[0]) {
  957                                 vm_page_lock_queues();
  958                                 vm_page_free(ma[0]);
  959                                 vm_page_unlock_queues();
  960                         }
  961                         VM_OBJECT_UNLOCK(object);
  962                         return (EIO);
  963                 }
  964         }
  965         vm_page_lock_queues();
  966         vm_page_hold(ma[0]);
  967         vm_page_unlock_queues();
  968         vm_page_wakeup(ma[0]);
  969         VM_OBJECT_UNLOCK(object);
  970 
  971         imgp->firstpage = sf_buf_alloc(ma[0], 0);
  972         imgp->image_header = (char *)sf_buf_kva(imgp->firstpage);
  973 
  974         return (0);
  975 }
  976 
  977 void
  978 exec_unmap_first_page(imgp)
  979         struct image_params *imgp;
  980 {
  981         vm_page_t m;
  982 
  983         if (imgp->firstpage != NULL) {
  984                 m = sf_buf_page(imgp->firstpage);
  985                 sf_buf_free(imgp->firstpage);
  986                 imgp->firstpage = NULL;
  987                 vm_page_lock_queues();
  988                 vm_page_unhold(m);
  989                 vm_page_unlock_queues();
  990         }
  991 }
  992 
  993 /*
  994  * Destroy old address space, and allocate a new stack
  995  *      The new stack is only SGROWSIZ large because it is grown
  996  *      automatically in trap.c.
  997  */
  998 int
  999 exec_new_vmspace(imgp, sv)
 1000         struct image_params *imgp;
 1001         struct sysentvec *sv;
 1002 {
 1003         int error;
 1004         struct proc *p = imgp->proc;
 1005         struct vmspace *vmspace = p->p_vmspace;
 1006         vm_offset_t sv_minuser, stack_addr;
 1007         vm_map_t map;
 1008         u_long ssiz;
 1009 
 1010         imgp->vmspace_destroyed = 1;
 1011         imgp->sysent = sv;
 1012 
 1013         /* May be called with Giant held */
 1014         EVENTHANDLER_INVOKE(process_exec, p, imgp);
 1015 
 1016         /*
 1017          * Blow away entire process VM, if address space not shared,
 1018          * otherwise, create a new VM space so that other threads are
 1019          * not disrupted
 1020          */
 1021         map = &vmspace->vm_map;
 1022         if (map_at_zero)
 1023                 sv_minuser = sv->sv_minuser;
 1024         else
 1025                 sv_minuser = MAX(sv->sv_minuser, PAGE_SIZE);
 1026         if (vmspace->vm_refcnt == 1 && vm_map_min(map) == sv_minuser &&
 1027             vm_map_max(map) == sv->sv_maxuser) {
 1028                 shmexit(vmspace);
 1029                 pmap_remove_pages(vmspace_pmap(vmspace));
 1030                 vm_map_remove(map, vm_map_min(map), vm_map_max(map));
 1031         } else {
 1032                 error = vmspace_exec(p, sv_minuser, sv->sv_maxuser);
 1033                 if (error)
 1034                         return (error);
 1035                 vmspace = p->p_vmspace;
 1036                 map = &vmspace->vm_map;
 1037         }
 1038 
 1039         /* Allocate a new stack */
 1040         if (sv->sv_maxssiz != NULL)
 1041                 ssiz = *sv->sv_maxssiz;
 1042         else
 1043                 ssiz = maxssiz;
 1044         stack_addr = sv->sv_usrstack - ssiz;
 1045         error = vm_map_stack(map, stack_addr, (vm_size_t)ssiz,
 1046             sv->sv_stackprot, VM_PROT_ALL, MAP_STACK_GROWS_DOWN);
 1047         if (error)
 1048                 return (error);
 1049 
 1050 #ifdef __ia64__
 1051         /* Allocate a new register stack */
 1052         stack_addr = IA64_BACKINGSTORE;
 1053         error = vm_map_stack(map, stack_addr, (vm_size_t)ssiz,
 1054             sv->sv_stackprot, VM_PROT_ALL, MAP_STACK_GROWS_UP);
 1055         if (error)
 1056                 return (error);
 1057 #endif
 1058 
 1059         /* vm_ssize and vm_maxsaddr are somewhat antiquated concepts in the
 1060          * VM_STACK case, but they are still used to monitor the size of the
 1061          * process stack so we can check the stack rlimit.
 1062          */
 1063         vmspace->vm_ssize = sgrowsiz >> PAGE_SHIFT;
 1064         vmspace->vm_maxsaddr = (char *)sv->sv_usrstack - ssiz;
 1065 
 1066         return (0);
 1067 }
 1068 
 1069 /*
 1070  * Copy out argument and environment strings from the old process address
 1071  * space into the temporary string buffer.
 1072  */
 1073 int
 1074 exec_copyin_args(struct image_args *args, char *fname,
 1075     enum uio_seg segflg, char **argv, char **envv)
 1076 {
 1077         char *argp, *envp;
 1078         int error;
 1079         size_t length;
 1080 
 1081         bzero(args, sizeof(*args));
 1082         if (argv == NULL)
 1083                 return (EFAULT);
 1084         /*
 1085          * Allocate temporary demand zeroed space for argument and
 1086          *      environment strings:
 1087          *
 1088          * o ARG_MAX for argument and environment;
 1089          * o MAXSHELLCMDLEN for the name of interpreters.
 1090          */
 1091         args->buf = (char *) kmem_alloc_wait(exec_map,
 1092             PATH_MAX + ARG_MAX + MAXSHELLCMDLEN);
 1093         if (args->buf == NULL)
 1094                 return (ENOMEM);
 1095         args->begin_argv = args->buf;
 1096         args->endp = args->begin_argv;
 1097         args->stringspace = ARG_MAX;
 1098         /*
 1099          * Copy the file name.
 1100          */
 1101         if (fname != NULL) {
 1102                 args->fname = args->buf + ARG_MAX;
 1103                 error = (segflg == UIO_SYSSPACE) ?
 1104                     copystr(fname, args->fname, PATH_MAX, &length) :
 1105                     copyinstr(fname, args->fname, PATH_MAX, &length);
 1106                 if (error != 0)
 1107                         goto err_exit;
 1108         } else
 1109                 args->fname = NULL;
 1110 
 1111         /*
 1112          * extract arguments first
 1113          */
 1114         while ((argp = (caddr_t) (intptr_t) fuword(argv++))) {
 1115                 if (argp == (caddr_t) -1) {
 1116                         error = EFAULT;
 1117                         goto err_exit;
 1118                 }
 1119                 if ((error = copyinstr(argp, args->endp,
 1120                     args->stringspace, &length))) {
 1121                         if (error == ENAMETOOLONG) 
 1122                                 error = E2BIG;
 1123                         goto err_exit;
 1124                 }
 1125                 args->stringspace -= length;
 1126                 args->endp += length;
 1127                 args->argc++;
 1128         }
 1129 
 1130         args->begin_envv = args->endp;
 1131 
 1132         /*
 1133          * extract environment strings
 1134          */
 1135         if (envv) {
 1136                 while ((envp = (caddr_t)(intptr_t)fuword(envv++))) {
 1137                         if (envp == (caddr_t)-1) {
 1138                                 error = EFAULT;
 1139                                 goto err_exit;
 1140                         }
 1141                         if ((error = copyinstr(envp, args->endp,
 1142                             args->stringspace, &length))) {
 1143                                 if (error == ENAMETOOLONG)
 1144                                         error = E2BIG;
 1145                                 goto err_exit;
 1146                         }
 1147                         args->stringspace -= length;
 1148                         args->endp += length;
 1149                         args->envc++;
 1150                 }
 1151         }
 1152 
 1153         return (0);
 1154 
 1155 err_exit:
 1156         exec_free_args(args);
 1157         return (error);
 1158 }
 1159 
 1160 static void
 1161 exec_free_args(struct image_args *args)
 1162 {
 1163 
 1164         if (args->buf) {
 1165                 kmem_free_wakeup(exec_map, (vm_offset_t)args->buf,
 1166                     PATH_MAX + ARG_MAX + MAXSHELLCMDLEN);
 1167                 args->buf = NULL;
 1168         }
 1169 }
 1170 
 1171 /*
 1172  * Copy strings out to the new process address space, constructing new arg
 1173  * and env vector tables. Return a pointer to the base so that it can be used
 1174  * as the initial stack pointer.
 1175  */
 1176 register_t *
 1177 exec_copyout_strings(imgp)
 1178         struct image_params *imgp;
 1179 {
 1180         int argc, envc;
 1181         char **vectp;
 1182         char *stringp, *destp;
 1183         register_t *stack_base;
 1184         struct ps_strings *arginfo;
 1185         struct proc *p;
 1186         size_t execpath_len;
 1187         int szsigcode;
 1188 
 1189         /*
 1190          * Calculate string base and vector table pointers.
 1191          * Also deal with signal trampoline code for this exec type.
 1192          */
 1193         if (imgp->execpath != NULL && imgp->auxargs != NULL)
 1194                 execpath_len = strlen(imgp->execpath) + 1;
 1195         else
 1196                 execpath_len = 0;
 1197         p = imgp->proc;
 1198         szsigcode = 0;
 1199         arginfo = (struct ps_strings *)p->p_sysent->sv_psstrings;
 1200         if (p->p_sysent->sv_szsigcode != NULL)
 1201                 szsigcode = *(p->p_sysent->sv_szsigcode);
 1202         destp = (caddr_t)arginfo - szsigcode - SPARE_USRSPACE -
 1203             roundup(execpath_len, sizeof(char *)) -
 1204             roundup((ARG_MAX - imgp->args->stringspace), sizeof(char *));
 1205 
 1206         /*
 1207          * install sigcode
 1208          */
 1209         if (szsigcode)
 1210                 copyout(p->p_sysent->sv_sigcode, ((caddr_t)arginfo -
 1211                     szsigcode), szsigcode);
 1212 
 1213         /*
 1214          * Copy the image path for the rtld.
 1215          */
 1216         if (execpath_len != 0) {
 1217                 imgp->execpathp = (uintptr_t)arginfo - szsigcode - execpath_len;
 1218                 copyout(imgp->execpath, (void *)imgp->execpathp,
 1219                     execpath_len);
 1220         }
 1221 
 1222         /*
 1223          * If we have a valid auxargs ptr, prepare some room
 1224          * on the stack.
 1225          */
 1226         if (imgp->auxargs) {
 1227                 /*
 1228                  * 'AT_COUNT*2' is size for the ELF Auxargs data. This is for
 1229                  * lower compatibility.
 1230                  */
 1231                 imgp->auxarg_size = (imgp->auxarg_size) ? imgp->auxarg_size :
 1232                     (AT_COUNT * 2);
 1233                 /*
 1234                  * The '+ 2' is for the null pointers at the end of each of
 1235                  * the arg and env vector sets,and imgp->auxarg_size is room
 1236                  * for argument of Runtime loader.
 1237                  */
 1238                 vectp = (char **)(destp - (imgp->args->argc +
 1239                     imgp->args->envc + 2 + imgp->auxarg_size + execpath_len) *
 1240                     sizeof(char *));
 1241         } else {
 1242                 /*
 1243                  * The '+ 2' is for the null pointers at the end of each of
 1244                  * the arg and env vector sets
 1245                  */
 1246                 vectp = (char **)(destp - (imgp->args->argc + imgp->args->envc + 2) *
 1247                     sizeof(char *));
 1248         }
 1249 
 1250         /*
 1251          * vectp also becomes our initial stack base
 1252          */
 1253         stack_base = (register_t *)vectp;
 1254 
 1255         stringp = imgp->args->begin_argv;
 1256         argc = imgp->args->argc;
 1257         envc = imgp->args->envc;
 1258 
 1259         /*
 1260          * Copy out strings - arguments and environment.
 1261          */
 1262         copyout(stringp, destp, ARG_MAX - imgp->args->stringspace);
 1263 
 1264         /*
 1265          * Fill in "ps_strings" struct for ps, w, etc.
 1266          */
 1267         suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp);
 1268         suword(&arginfo->ps_nargvstr, argc);
 1269 
 1270         /*
 1271          * Fill in argument portion of vector table.
 1272          */
 1273         for (; argc > 0; --argc) {
 1274                 suword(vectp++, (long)(intptr_t)destp);
 1275                 while (*stringp++ != 0)
 1276                         destp++;
 1277                 destp++;
 1278         }
 1279 
 1280         /* a null vector table pointer separates the argp's from the envp's */
 1281         suword(vectp++, 0);
 1282 
 1283         suword(&arginfo->ps_envstr, (long)(intptr_t)vectp);
 1284         suword(&arginfo->ps_nenvstr, envc);
 1285 
 1286         /*
 1287          * Fill in environment portion of vector table.
 1288          */
 1289         for (; envc > 0; --envc) {
 1290                 suword(vectp++, (long)(intptr_t)destp);
 1291                 while (*stringp++ != 0)
 1292                         destp++;
 1293                 destp++;
 1294         }
 1295 
 1296         /* end of vector table is a null pointer */
 1297         suword(vectp, 0);
 1298 
 1299         return (stack_base);
 1300 }
 1301 
 1302 /*
 1303  * Check permissions of file to execute.
 1304  *      Called with imgp->vp locked.
 1305  *      Return 0 for success or error code on failure.
 1306  */
 1307 int
 1308 exec_check_permissions(imgp)
 1309         struct image_params *imgp;
 1310 {
 1311         struct vnode *vp = imgp->vp;
 1312         struct vattr *attr = imgp->attr;
 1313         struct thread *td;
 1314         int error;
 1315 
 1316         td = curthread;
 1317 
 1318         /* Get file attributes */
 1319         error = VOP_GETATTR(vp, attr, td->td_ucred);
 1320         if (error)
 1321                 return (error);
 1322 
 1323 #ifdef MAC
 1324         error = mac_vnode_check_exec(td->td_ucred, imgp->vp, imgp);
 1325         if (error)
 1326                 return (error);
 1327 #endif
 1328         
 1329         /*
 1330          * 1) Check if file execution is disabled for the filesystem that this
 1331          *      file resides on.
 1332          * 2) Insure that at least one execute bit is on - otherwise root
 1333          *      will always succeed, and we don't want to happen unless the
 1334          *      file really is executable.
 1335          * 3) Insure that the file is a regular file.
 1336          */
 1337         if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
 1338             ((attr->va_mode & 0111) == 0) ||
 1339             (attr->va_type != VREG))
 1340                 return (EACCES);
 1341 
 1342         /*
 1343          * Zero length files can't be exec'd
 1344          */
 1345         if (attr->va_size == 0)
 1346                 return (ENOEXEC);
 1347 
 1348         /*
 1349          *  Check for execute permission to file based on current credentials.
 1350          */
 1351         error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
 1352         if (error)
 1353                 return (error);
 1354 
 1355         /*
 1356          * Check number of open-for-writes on the file and deny execution
 1357          * if there are any.
 1358          */
 1359         if (vp->v_writecount)
 1360                 return (ETXTBSY);
 1361 
 1362         /*
 1363          * Call filesystem specific open routine (which does nothing in the
 1364          * general case).
 1365          */
 1366         error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
 1367         if (error == 0)
 1368                 imgp->opened = 1;
 1369         return (error);
 1370 }
 1371 
 1372 /*
 1373  * Exec handler registration
 1374  */
 1375 int
 1376 exec_register(execsw_arg)
 1377         const struct execsw *execsw_arg;
 1378 {
 1379         const struct execsw **es, **xs, **newexecsw;
 1380         int count = 2;  /* New slot and trailing NULL */
 1381 
 1382         if (execsw)
 1383                 for (es = execsw; *es; es++)
 1384                         count++;
 1385         newexecsw = malloc(count * sizeof(*es), M_TEMP, M_WAITOK);
 1386         if (newexecsw == NULL)
 1387                 return (ENOMEM);
 1388         xs = newexecsw;
 1389         if (execsw)
 1390                 for (es = execsw; *es; es++)
 1391                         *xs++ = *es;
 1392         *xs++ = execsw_arg;
 1393         *xs = NULL;
 1394         if (execsw)
 1395                 free(execsw, M_TEMP);
 1396         execsw = newexecsw;
 1397         return (0);
 1398 }
 1399 
 1400 int
 1401 exec_unregister(execsw_arg)
 1402         const struct execsw *execsw_arg;
 1403 {
 1404         const struct execsw **es, **xs, **newexecsw;
 1405         int count = 1;
 1406 
 1407         if (execsw == NULL)
 1408                 panic("unregister with no handlers left?\n");
 1409 
 1410         for (es = execsw; *es; es++) {
 1411                 if (*es == execsw_arg)
 1412                         break;
 1413         }
 1414         if (*es == NULL)
 1415                 return (ENOENT);
 1416         for (es = execsw; *es; es++)
 1417                 if (*es != execsw_arg)
 1418                         count++;
 1419         newexecsw = malloc(count * sizeof(*es), M_TEMP, M_WAITOK);
 1420         if (newexecsw == NULL)
 1421                 return (ENOMEM);
 1422         xs = newexecsw;
 1423         for (es = execsw; *es; es++)
 1424                 if (*es != execsw_arg)
 1425                         *xs++ = *es;
 1426         *xs = NULL;
 1427         if (execsw)
 1428                 free(execsw, M_TEMP);
 1429         execsw = newexecsw;
 1430         return (0);
 1431 }

Cache object: de5aed7d438d9c891d6cdb234847415d


[ 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.