2 * Copyright (c) 1993, David Greenman
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
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.
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
26 * $FreeBSD: src/sys/kern/kern_exec.c,v 1.107.2.15 2002/07/30 15:40:46 nectar Exp $
27 * $DragonFly: src/sys/kern/kern_exec.c,v 1.45 2006/09/17 21:07:32 dillon Exp $
30 #include <sys/param.h>
31 #include <sys/systm.h>
32 #include <sys/sysproto.h>
33 #include <sys/kernel.h>
34 #include <sys/mount.h>
35 #include <sys/filedesc.h>
36 #include <sys/fcntl.h>
39 #include <sys/imgact.h>
40 #include <sys/imgact_elf.h>
41 #include <sys/kern_syscall.h>
43 #include <sys/malloc.h>
45 #include <sys/ktrace.h>
46 #include <sys/signalvar.h>
47 #include <sys/pioctl.h>
48 #include <sys/nlookup.h>
49 #include <sys/sfbuf.h>
50 #include <sys/sysent.h>
52 #include <sys/sysctl.h>
53 #include <sys/vnode.h>
54 #include <sys/vmmeter.h>
56 #include <sys/libkern.h>
59 #include <vm/vm_param.h>
62 #include <vm/vm_page.h>
63 #include <vm/vm_map.h>
64 #include <vm/vm_kern.h>
65 #include <vm/vm_extern.h>
66 #include <vm/vm_object.h>
67 #include <vm/vm_pager.h>
70 #include <machine/reg.h>
72 #include <sys/thread2.h>
74 MALLOC_DEFINE(M_PARGS, "proc-args", "Process arguments");
76 static register_t *exec_copyout_strings (struct image_params *);
78 /* XXX This should be vm_size_t. */
79 static u_long ps_strings = PS_STRINGS;
80 SYSCTL_ULONG(_kern, KERN_PS_STRINGS, ps_strings, CTLFLAG_RD, &ps_strings, 0, "");
82 /* XXX This should be vm_size_t. */
83 static u_long usrstack = USRSTACK;
84 SYSCTL_ULONG(_kern, KERN_USRSTACK, usrstack, CTLFLAG_RD, &usrstack, 0, "");
86 u_long ps_arg_cache_limit = PAGE_SIZE / 16;
87 SYSCTL_LONG(_kern, OID_AUTO, ps_arg_cache_limit, CTLFLAG_RW,
88 &ps_arg_cache_limit, 0, "");
91 SYSCTL_INT(_kern, OID_AUTO, ps_argsopen, CTLFLAG_RW, &ps_argsopen, 0, "");
93 void print_execve_args(struct image_args *args);
94 int debug_execve_args = 0;
95 SYSCTL_INT(_kern, OID_AUTO, debug_execve_args, CTLFLAG_RW, &debug_execve_args,
99 * stackgap_random specifies if the stackgap should have a random size added
100 * to it. It must be a power of 2. If non-zero, the stack gap will be
101 * calculated as: ALIGN(karc4random() & (stackgap_random - 1)).
103 static int stackgap_random = 1024;
105 sysctl_kern_stackgap(SYSCTL_HANDLER_ARGS)
108 new_val = stackgap_random;
109 error = sysctl_handle_int(oidp, &new_val, 0, req);
110 if (error != 0 || req->newptr == NULL)
112 if ((new_val < 0) || (new_val > 16 * PAGE_SIZE) || ! powerof2(new_val))
114 stackgap_random = new_val;
119 SYSCTL_PROC(_kern, OID_AUTO, stackgap_random, CTLFLAG_RW|CTLTYPE_UINT,
120 0, 0, sysctl_kern_stackgap, "IU", "Max random stack gap (power of 2)");
123 print_execve_args(struct image_args *args)
128 cp = args->begin_argv;
129 for (ndx = 0; ndx < args->argc; ndx++) {
130 printf("\targv[%d]: %s\n", ndx, cp);
131 while (*cp++ != '\0');
133 for (ndx = 0; ndx < args->envc; ndx++) {
134 printf("\tenvv[%d]: %s\n", ndx, cp);
135 while (*cp++ != '\0');
140 * Each of the items is a pointer to a `const struct execsw', hence the
141 * double pointer here.
143 static const struct execsw **execsw;
146 kern_execve(struct nlookupdata *nd, struct image_args *args)
148 struct thread *td = curthread;
149 struct proc *p = td->td_proc;
150 register_t *stack_base;
152 struct image_params image_params, *imgp;
154 int (*img_first) (struct image_params *);
156 if (debug_execve_args) {
157 printf("%s()\n", __func__);
158 print_execve_args(args);
162 imgp = &image_params;
165 * Lock the process and set the P_INEXEC flag to indicate that
166 * it should be left alone until we're done here. This is
167 * necessary to avoid race conditions - e.g. in ptrace() -
168 * that might allow a local user to illicitly obtain elevated
171 p->p_flag |= P_INEXEC;
174 * Initialize part of the common data
179 imgp->entry_addr = 0;
181 imgp->vmspace_destroyed = 0;
182 imgp->interpreted = 0;
183 imgp->interpreter_name[0] = 0;
184 imgp->auxargs = NULL;
186 imgp->firstpage = NULL;
187 imgp->ps_strings = 0;
188 imgp->image_header = NULL;
193 * Translate the file name to a vnode. Unlock the cache entry to
194 * improve parallelism for programs exec'd in parallel.
196 if ((error = nlookup(nd)) != 0)
198 error = cache_vget(nd->nl_ncp, nd->nl_cred, LK_EXCLUSIVE, &imgp->vp);
199 KKASSERT(nd->nl_flags & NLC_NCPISLOCKED);
200 nd->nl_flags &= ~NLC_NCPISLOCKED;
201 cache_unlock(nd->nl_ncp);
206 * Check file permissions (also 'opens' file)
208 error = exec_check_permissions(imgp);
211 goto exec_fail_dealloc;
214 error = exec_map_first_page(imgp);
217 goto exec_fail_dealloc;
219 if (debug_execve_args && imgp->interpreted) {
220 printf(" target is interpreted -- recursive pass\n");
221 printf(" interpreter: %s\n", imgp->interpreter_name);
222 print_execve_args(args);
226 * If the current process has a special image activator it
227 * wants to try first, call it. For example, emulating shell
228 * scripts differently.
231 if ((img_first = imgp->proc->p_sysent->sv_imgact_try) != NULL)
232 error = img_first(imgp);
235 * If the vnode has a registered vmspace, exec the vmspace
237 if (error == -1 && imgp->vp->v_resident) {
238 error = exec_resident_imgact(imgp);
242 * Loop through the list of image activators, calling each one.
243 * An activator returns -1 if there is no match, 0 on success,
244 * and an error otherwise.
246 for (i = 0; error == -1 && execsw[i]; ++i) {
247 if (execsw[i]->ex_imgact == NULL ||
248 execsw[i]->ex_imgact == img_first) {
251 error = (*execsw[i]->ex_imgact)(imgp);
257 goto exec_fail_dealloc;
261 * Special interpreter operation, cleanup and loop up to try to
262 * activate the interpreter.
264 if (imgp->interpreted) {
265 exec_unmap_first_page(imgp);
269 error = nlookup_init(nd, imgp->interpreter_name, UIO_SYSSPACE,
277 * Copy out strings (args and env) and initialize stack base
279 stack_base = exec_copyout_strings(imgp);
280 p->p_vmspace->vm_minsaddr = (char *)stack_base;
283 * If custom stack fixup routine present for this process
284 * let it do the stack setup. If we are running a resident
285 * image there is no auxinfo or other image activator context
286 * so don't try to add fixups to the stack.
288 * Else stuff argument count as first item on stack
290 if (p->p_sysent->sv_fixup && imgp->resident == 0)
291 (*p->p_sysent->sv_fixup)(&stack_base, imgp);
293 suword(--stack_base, imgp->args->argc);
296 * For security and other reasons, the file descriptor table cannot
297 * be shared after an exec.
299 if (p->p_fd->fd_refcnt > 1) {
300 struct filedesc *tmp;
308 * For security and other reasons, signal handlers cannot
309 * be shared after an exec. The new proces gets a copy of the old
310 * handlers. In execsigs(), the new process will have its signals
313 if (p->p_procsig->ps_refcnt > 1) {
314 struct procsig *newprocsig;
316 MALLOC(newprocsig, struct procsig *, sizeof(struct procsig),
317 M_SUBPROC, M_WAITOK);
318 bcopy(p->p_procsig, newprocsig, sizeof(*newprocsig));
319 p->p_procsig->ps_refcnt--;
320 p->p_procsig = newprocsig;
321 p->p_procsig->ps_refcnt = 1;
322 if (p->p_sigacts == &p->p_addr->u_sigacts)
323 panic("shared procsig but private sigacts?");
325 p->p_addr->u_sigacts = *p->p_sigacts;
326 p->p_sigacts = &p->p_addr->u_sigacts;
330 * For security and other reasons virtual kernels cannot be
331 * inherited by an exec. This also allows a virtual kernel
332 * to fork/exec unrelated applications.
335 vkernel_drop(p->p_vkernel);
342 /* close files on exec */
345 /* reset caught signals */
348 /* name this process - nameiexec(p, ndp) */
349 len = min(nd->nl_ncp->nc_nlen, MAXCOMLEN);
350 bcopy(nd->nl_ncp->nc_name, p->p_comm, len);
352 bcopy(p->p_comm, p->p_lwp.lwp_thread->td_comm, MAXCOMLEN+1);
355 * mark as execed, wakeup the process that vforked (if any) and tell
356 * it that it now has its own resources back
359 if (p->p_pptr && (p->p_flag & P_PPWAIT)) {
360 p->p_flag &= ~P_PPWAIT;
361 wakeup((caddr_t)p->p_pptr);
365 * Implement image setuid/setgid.
367 * Don't honor setuid/setgid if the filesystem prohibits it or if
368 * the process is being traced.
370 if ((((attr.va_mode & VSUID) && p->p_ucred->cr_uid != attr.va_uid) ||
371 ((attr.va_mode & VSGID) && p->p_ucred->cr_gid != attr.va_gid)) &&
372 (imgp->vp->v_mount->mnt_flag & MNT_NOSUID) == 0 &&
373 (p->p_flag & P_TRACED) == 0) {
375 * Turn off syscall tracing for set-id programs, except for
376 * root. Record any set-id flags first to make sure that
377 * we do not regain any tracing during a possible block.
380 if (p->p_tracenode && suser(td) != 0) {
381 ktrdestroy(&p->p_tracenode);
384 /* Close any file descriptors 0..2 that reference procfs */
386 /* Make sure file descriptors 0..2 are in use. */
387 error = fdcheckstd(p);
389 goto exec_fail_dealloc;
391 * Set the new credentials.
394 if (attr.va_mode & VSUID)
395 change_euid(attr.va_uid);
396 if (attr.va_mode & VSGID)
397 p->p_ucred->cr_gid = attr.va_gid;
400 * Clear local varsym variables
402 varsymset_clean(&p->p_varsymset);
404 if (p->p_ucred->cr_uid == p->p_ucred->cr_ruid &&
405 p->p_ucred->cr_gid == p->p_ucred->cr_rgid)
406 p->p_flag &= ~P_SUGID;
410 * Implement correct POSIX saved-id behavior.
412 if (p->p_ucred->cr_svuid != p->p_ucred->cr_uid ||
413 p->p_ucred->cr_svgid != p->p_ucred->cr_gid) {
415 p->p_ucred->cr_svuid = p->p_ucred->cr_uid;
416 p->p_ucred->cr_svgid = p->p_ucred->cr_gid;
420 * Store the vp for use in procfs
422 if (p->p_textvp) /* release old reference */
424 p->p_textvp = imgp->vp;
428 * Notify others that we exec'd, and clear the P_INEXEC flag
429 * as we're now a bona fide freshly-execed process.
431 KNOTE(&p->p_klist, NOTE_EXEC);
432 p->p_flag &= ~P_INEXEC;
435 * If tracing the process, trap to debugger so breakpoints
436 * can be set before the program executes.
438 STOPEVENT(p, S_EXEC, 0);
440 if (p->p_flag & P_TRACED)
443 /* clear "fork but no exec" flag, as we _are_ execing */
444 p->p_acflag &= ~AFORK;
446 /* Set values passed into the program in registers. */
447 setregs(p, imgp->entry_addr, (u_long)(uintptr_t)stack_base,
450 /* Free any previous argument cache */
451 if (p->p_args && --p->p_args->ar_ref == 0)
452 FREE(p->p_args, M_PARGS);
455 /* Cache arguments if they fit inside our allowance */
456 i = imgp->args->begin_envv - imgp->args->begin_argv;
457 if (ps_arg_cache_limit >= i + sizeof(struct pargs)) {
458 MALLOC(p->p_args, struct pargs *, sizeof(struct pargs) + i,
460 p->p_args->ar_ref = 1;
461 p->p_args->ar_length = i;
462 bcopy(imgp->args->begin_argv, p->p_args->ar_args, i);
468 * free various allocated resources
471 exec_unmap_first_page(imgp);
479 ++mycpu->gd_cnt.v_exec;
484 /* we're done here, clear P_INEXEC */
485 p->p_flag &= ~P_INEXEC;
486 if (imgp->vmspace_destroyed) {
487 /* sorry, no more process anymore. exit gracefully */
488 exit1(W_EXITCODE(0, SIGABRT));
497 * execve() system call.
500 sys_execve(struct execve_args *uap)
502 struct nlookupdata nd;
503 struct image_args args;
506 error = nlookup_init(&nd, uap->fname, UIO_USERSPACE, NLC_FOLLOW);
508 error = exec_copyin_args(&args, uap->fname, PATH_USERSPACE,
509 uap->argv, uap->envv);
512 error = kern_execve(&nd, &args);
514 exec_free_args(&args);
517 * The syscall result is returned in registers to the new program.
518 * Linux will register %edx as an atexit function and we must be
519 * sure to set it to 0. XXX
522 uap->sysmsg_result64 = 0;
528 exec_map_first_page(struct image_params *imgp)
532 vm_page_t ma[VM_INITIAL_PAGEIN];
537 exec_unmap_first_page(imgp);
540 * The file has to be mappable.
542 if ((object = imgp->vp->v_object) == NULL)
546 * We shouldn't need protection for vm_page_grab() but we certainly
547 * need it for the lookup loop below (lookup/busy race), since
548 * an interrupt can unbusy and free the page before our busy check.
551 m = vm_page_grab(object, 0, VM_ALLOC_NORMAL | VM_ALLOC_RETRY);
553 if ((m->valid & VM_PAGE_BITS_ALL) != VM_PAGE_BITS_ALL) {
555 initial_pagein = VM_INITIAL_PAGEIN;
556 if (initial_pagein > object->size)
557 initial_pagein = object->size;
558 for (i = 1; i < initial_pagein; i++) {
559 if ((m = vm_page_lookup(object, i)) != NULL) {
560 if ((m->flags & PG_BUSY) || m->busy)
566 m = vm_page_alloc(object, i, VM_ALLOC_NORMAL);
575 * get_pages unbusies all the requested pages except the
576 * primary page (at index 0 in this case).
578 rv = vm_pager_get_pages(object, ma, initial_pagein, 0);
579 m = vm_page_lookup(object, 0);
581 if (rv != VM_PAGER_OK || m == NULL || m->valid == 0) {
583 vm_page_protect(m, VM_PROT_NONE);
591 vm_page_wakeup(m); /* unbusy the page */
594 imgp->firstpage = sf_buf_alloc(m, SFB_CPUPRIVATE);
595 imgp->image_header = (void *)sf_buf_kva(imgp->firstpage);
601 exec_unmap_first_page(struct image_params *imgp)
606 if (imgp->firstpage != NULL) {
607 m = sf_buf_page(imgp->firstpage);
608 sf_buf_free(imgp->firstpage);
609 imgp->firstpage = NULL;
610 imgp->image_header = NULL;
617 * Destroy old address space, and allocate a new stack
618 * The new stack is only SGROWSIZ large because it is grown
619 * automatically in trap.c.
622 exec_new_vmspace(struct image_params *imgp, struct vmspace *vmcopy)
625 struct vmspace *vmspace = imgp->proc->p_vmspace;
626 vm_offset_t stack_addr = USRSTACK - maxssiz;
629 imgp->vmspace_destroyed = 1;
632 * Prevent a pending AIO from modifying the new address space.
634 aio_proc_rundown(imgp->proc);
637 * Blow away entire process VM, if address space not shared,
638 * otherwise, create a new VM space so that other threads are
639 * not disrupted. If we are execing a resident vmspace we
640 * create a duplicate of it and remap the stack.
642 * The exitingcnt test is not strictly necessary but has been
643 * included for code sanity (to make the code more deterministic).
645 map = &vmspace->vm_map;
647 vmspace_exec(imgp->proc, vmcopy);
648 vmspace = imgp->proc->p_vmspace;
649 pmap_remove_pages(vmspace_pmap(vmspace), stack_addr, USRSTACK);
650 map = &vmspace->vm_map;
651 } else if (vmspace->vm_refcnt == 1 && vmspace->vm_exitingcnt == 0) {
653 if (vmspace->vm_upcalls)
654 upc_release(vmspace, &imgp->proc->p_lwp);
655 pmap_remove_pages(vmspace_pmap(vmspace), 0, VM_MAXUSER_ADDRESS);
656 vm_map_remove(map, 0, VM_MAXUSER_ADDRESS);
658 vmspace_exec(imgp->proc, NULL);
659 vmspace = imgp->proc->p_vmspace;
660 map = &vmspace->vm_map;
663 /* Allocate a new stack */
664 error = vm_map_stack(&vmspace->vm_map, stack_addr, (vm_size_t)maxssiz,
665 VM_PROT_ALL, VM_PROT_ALL, 0);
669 /* vm_ssize and vm_maxsaddr are somewhat antiquated concepts in the
670 * VM_STACK case, but they are still used to monitor the size of the
671 * process stack so we can check the stack rlimit.
673 vmspace->vm_ssize = sgrowsiz >> PAGE_SHIFT;
674 vmspace->vm_maxsaddr = (char *)USRSTACK - maxssiz;
680 * Copy out argument and environment strings from the old process
681 * address space into the temporary string buffer.
684 exec_copyin_args(struct image_args *args, char *fname,
685 enum exec_path_segflg segflg, char **argv, char **envv)
691 bzero(args, sizeof(*args));
692 args->buf = (char *) kmem_alloc_wait(exec_map, PATH_MAX + ARG_MAX);
693 if (args->buf == NULL)
695 args->begin_argv = args->buf;
696 args->endp = args->begin_argv;
697 args->space = ARG_MAX;
699 args->fname = args->buf + ARG_MAX;
702 * Copy the file name.
704 if (segflg == PATH_SYSSPACE) {
705 error = copystr(fname, args->fname, PATH_MAX, &length);
706 } else if (segflg == PATH_USERSPACE) {
707 error = copyinstr(fname, args->fname, PATH_MAX, &length);
711 * Extract argument strings. argv may not be NULL. The argv
712 * array is terminated by a NULL entry. We special-case the
713 * situation where argv[0] is NULL by passing { filename, NULL }
714 * to the new program to guarentee that the interpreter knows what
715 * file to open in case we exec an interpreted file. Note that
716 * a NULL argv[0] terminates the argv[] array.
718 * XXX the special-casing of argv[0] is historical and needs to be
724 while ((argp = (caddr_t)(intptr_t)fuword(argv++)) != NULL) {
725 if (argp == (caddr_t)-1) {
729 error = copyinstr(argp, args->endp,
730 args->space, &length);
732 if (error == ENAMETOOLONG)
736 args->space -= length;
737 args->endp += length;
740 if (args->argc == 0 && error == 0) {
741 length = strlen(args->fname) + 1;
742 if (length > args->space) {
745 bcopy(args->fname, args->endp, length);
746 args->space -= length;
747 args->endp += length;
753 args->begin_envv = args->endp;
756 * extract environment strings. envv may be NULL.
758 if (envv && error == 0) {
759 while ((envp = (caddr_t) (intptr_t) fuword(envv++))) {
760 if (envp == (caddr_t) -1) {
764 error = copyinstr(envp, args->endp, args->space,
767 if (error == ENAMETOOLONG)
771 args->space -= length;
772 args->endp += length;
780 exec_free_args(struct image_args *args)
783 kmem_free_wakeup(exec_map,
784 (vm_offset_t)args->buf, PATH_MAX + ARG_MAX);
790 * Copy strings out to the new process address space, constructing
791 * new arg and env vector tables. Return a pointer to the base
792 * so that it can be used as the initial stack pointer.
795 exec_copyout_strings(struct image_params *imgp)
797 int argc, envc, sgap;
799 char *stringp, *destp;
800 register_t *stack_base;
801 struct ps_strings *arginfo;
805 * Calculate string base and vector table pointers.
806 * Also deal with signal trampoline code for this exec type.
808 arginfo = (struct ps_strings *)PS_STRINGS;
809 szsigcode = *(imgp->proc->p_sysent->sv_szsigcode);
810 if (stackgap_random != 0)
811 sgap = ALIGN(karc4random() & (stackgap_random - 1));
814 destp = (caddr_t)arginfo - szsigcode - SPARE_USRSPACE - sgap -
815 roundup((ARG_MAX - imgp->args->space), sizeof(char *));
821 copyout(imgp->proc->p_sysent->sv_sigcode,
822 ((caddr_t)arginfo - szsigcode), szsigcode);
825 * If we have a valid auxargs ptr, prepare some room
828 * The '+ 2' is for the null pointers at the end of each of the
829 * arg and env vector sets, and 'AT_COUNT*2' is room for the
833 vectp = (char **)(destp - (imgp->args->argc +
834 imgp->args->envc + 2 + AT_COUNT * 2) * sizeof(char*));
836 vectp = (char **)(destp - (imgp->args->argc +
837 imgp->args->envc + 2) * sizeof(char*));
841 * NOTE: don't bother aligning the stack here for GCC 2.x, it will
842 * be done in crt1.o. Note that GCC 3.x aligns the stack in main.
846 * vectp also becomes our initial stack base
848 stack_base = (register_t *)vectp;
850 stringp = imgp->args->begin_argv;
851 argc = imgp->args->argc;
852 envc = imgp->args->envc;
855 * Copy out strings - arguments and environment.
857 copyout(stringp, destp, ARG_MAX - imgp->args->space);
860 * Fill in "ps_strings" struct for ps, w, etc.
862 suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp);
863 suword(&arginfo->ps_nargvstr, argc);
866 * Fill in argument portion of vector table.
868 for (; argc > 0; --argc) {
869 suword(vectp++, (long)(intptr_t)destp);
870 while (*stringp++ != 0)
875 /* a null vector table pointer separates the argp's from the envp's */
878 suword(&arginfo->ps_envstr, (long)(intptr_t)vectp);
879 suword(&arginfo->ps_nenvstr, envc);
882 * Fill in environment portion of vector table.
884 for (; envc > 0; --envc) {
885 suword(vectp++, (long)(intptr_t)destp);
886 while (*stringp++ != 0)
891 /* end of vector table is a null pointer */
898 * Check permissions of file to execute.
899 * Return 0 for success or error code on failure.
902 exec_check_permissions(struct image_params *imgp)
904 struct proc *p = imgp->proc;
905 struct vnode *vp = imgp->vp;
906 struct vattr *attr = imgp->attr;
909 /* Get file attributes */
910 error = VOP_GETATTR(vp, attr);
915 * 1) Check if file execution is disabled for the filesystem that this
917 * 2) Insure that at least one execute bit is on - otherwise root
918 * will always succeed, and we don't want to happen unless the
919 * file really is executable.
920 * 3) Insure that the file is a regular file.
922 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
923 ((attr->va_mode & 0111) == 0) ||
924 (attr->va_type != VREG)) {
929 * Zero length files can't be exec'd
931 if (attr->va_size == 0)
935 * Check for execute permission to file based on current credentials.
937 error = VOP_ACCESS(vp, VEXEC, p->p_ucred);
942 * Check number of open-for-writes on the file and deny execution
945 if (vp->v_writecount)
949 * Call filesystem specific open routine, which allows us to read,
950 * write, and mmap the file. Without the VOP_OPEN we can only
953 error = VOP_OPEN(vp, FREAD, p->p_ucred, NULL);
961 * Exec handler registration
964 exec_register(const struct execsw *execsw_arg)
966 const struct execsw **es, **xs, **newexecsw;
967 int count = 2; /* New slot and trailing NULL */
970 for (es = execsw; *es; es++)
972 newexecsw = kmalloc(count * sizeof(*es), M_TEMP, M_WAITOK);
973 if (newexecsw == NULL)
977 for (es = execsw; *es; es++)
982 kfree(execsw, M_TEMP);
988 exec_unregister(const struct execsw *execsw_arg)
990 const struct execsw **es, **xs, **newexecsw;
994 panic("unregister with no handlers left?");
996 for (es = execsw; *es; es++) {
997 if (*es == execsw_arg)
1002 for (es = execsw; *es; es++)
1003 if (*es != execsw_arg)
1005 newexecsw = kmalloc(count * sizeof(*es), M_TEMP, M_WAITOK);
1006 if (newexecsw == NULL)
1009 for (es = execsw; *es; es++)
1010 if (*es != execsw_arg)
1014 kfree(execsw, M_TEMP);