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 1.64 2008/10/26 04:29:19 sephe 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>
46 #include <sys/ktrace.h>
47 #include <sys/signalvar.h>
48 #include <sys/pioctl.h>
49 #include <sys/nlookup.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>
58 #include <cpu/lwbuf.h>
61 #include <vm/vm_param.h>
64 #include <vm/vm_page.h>
65 #include <vm/vm_map.h>
66 #include <vm/vm_kern.h>
67 #include <vm/vm_extern.h>
68 #include <vm/vm_object.h>
69 #include <vm/vnode_pager.h>
70 #include <vm/vm_pager.h>
75 #include <sys/thread2.h>
76 #include <sys/mplock2.h>
78 MALLOC_DEFINE(M_PARGS, "proc-args", "Process arguments");
79 MALLOC_DEFINE(M_EXECARGS, "exec-args", "Exec arguments");
81 static register_t *exec_copyout_strings (struct image_params *);
83 /* XXX This should be vm_size_t. */
84 static u_long ps_strings = PS_STRINGS;
85 SYSCTL_ULONG(_kern, KERN_PS_STRINGS, ps_strings, CTLFLAG_RD, &ps_strings, 0, "");
87 /* XXX This should be vm_size_t. */
88 static u_long usrstack = USRSTACK;
89 SYSCTL_ULONG(_kern, KERN_USRSTACK, usrstack, CTLFLAG_RD, &usrstack, 0, "");
91 u_long ps_arg_cache_limit = PAGE_SIZE / 16;
92 SYSCTL_LONG(_kern, OID_AUTO, ps_arg_cache_limit, CTLFLAG_RW,
93 &ps_arg_cache_limit, 0, "");
96 SYSCTL_INT(_kern, OID_AUTO, ps_argsopen, CTLFLAG_RW, &ps_argsopen, 0, "");
98 static int ktrace_suid = 0;
99 SYSCTL_INT(_kern, OID_AUTO, ktrace_suid, CTLFLAG_RW, &ktrace_suid, 0, "");
101 void print_execve_args(struct image_args *args);
102 int debug_execve_args = 0;
103 SYSCTL_INT(_kern, OID_AUTO, debug_execve_args, CTLFLAG_RW, &debug_execve_args,
107 * Exec arguments object cache
109 static struct objcache *exec_objcache;
113 exec_objcache_init(void *arg __unused)
117 cluster_limit = 16; /* up to this many objects */
118 exec_objcache = objcache_create_mbacked(
119 M_EXECARGS, PATH_MAX + ARG_MAX,
121 2, /* minimal magazine capacity */
124 SYSINIT(exec_objcache, SI_BOOT2_MACHDEP, SI_ORDER_ANY, exec_objcache_init, 0);
127 * stackgap_random specifies if the stackgap should have a random size added
128 * to it. It must be a power of 2. If non-zero, the stack gap will be
129 * calculated as: ALIGN(karc4random() & (stackgap_random - 1)).
131 static int stackgap_random = 1024;
133 sysctl_kern_stackgap(SYSCTL_HANDLER_ARGS)
136 new_val = stackgap_random;
137 error = sysctl_handle_int(oidp, &new_val, 0, req);
138 if (error != 0 || req->newptr == NULL)
140 if ((new_val < 0) || (new_val > 16 * PAGE_SIZE) || ! powerof2(new_val))
142 stackgap_random = new_val;
147 SYSCTL_PROC(_kern, OID_AUTO, stackgap_random, CTLFLAG_RW|CTLTYPE_UINT,
148 0, 0, sysctl_kern_stackgap, "IU", "Max random stack gap (power of 2)");
151 print_execve_args(struct image_args *args)
156 cp = args->begin_argv;
157 for (ndx = 0; ndx < args->argc; ndx++) {
158 kprintf("\targv[%d]: %s\n", ndx, cp);
159 while (*cp++ != '\0');
161 for (ndx = 0; ndx < args->envc; ndx++) {
162 kprintf("\tenvv[%d]: %s\n", ndx, cp);
163 while (*cp++ != '\0');
168 * Each of the items is a pointer to a `const struct execsw', hence the
169 * double pointer here.
171 static const struct execsw **execsw;
174 * Replace current vmspace with a new binary.
175 * Returns 0 on success, > 0 on recoverable error (use as errno).
176 * Returns -1 on lethal error which demands killing of the current
180 kern_execve(struct nlookupdata *nd, struct image_args *args)
182 struct thread *td = curthread;
183 struct lwp *lp = td->td_lwp;
184 struct proc *p = td->td_proc;
185 register_t *stack_base;
187 struct image_params image_params, *imgp;
189 int (*img_first) (struct image_params *);
191 if (debug_execve_args) {
192 kprintf("%s()\n", __func__);
193 print_execve_args(args);
197 imgp = &image_params;
200 * NOTE: P_INEXEC is handled by exec_new_vmspace() now. We make
201 * no modifications to the process at all until we get there.
203 * Note that multiple threads may be trying to exec at the same
204 * time. exec_new_vmspace() handles that too.
208 * Initialize part of the common data
213 imgp->entry_addr = 0;
215 imgp->vmspace_destroyed = 0;
216 imgp->interpreted = 0;
217 imgp->interpreter_name[0] = 0;
218 imgp->auxargs = NULL;
220 imgp->firstpage = NULL;
221 imgp->ps_strings = 0;
222 imgp->image_header = NULL;
227 * Translate the file name to a vnode. Unlock the cache entry to
228 * improve parallelism for programs exec'd in parallel.
230 if ((error = nlookup(nd)) != 0)
232 error = cache_vget(&nd->nl_nch, nd->nl_cred, LK_EXCLUSIVE, &imgp->vp);
233 KKASSERT(nd->nl_flags & NLC_NCPISLOCKED);
234 nd->nl_flags &= ~NLC_NCPISLOCKED;
235 cache_unlock(&nd->nl_nch);
240 * Check file permissions (also 'opens' file).
241 * Include also the top level mount in the check.
243 error = exec_check_permissions(imgp, nd->nl_nch.mount);
246 goto exec_fail_dealloc;
249 error = exec_map_first_page(imgp);
252 goto exec_fail_dealloc;
254 if (debug_execve_args && imgp->interpreted) {
255 kprintf(" target is interpreted -- recursive pass\n");
256 kprintf(" interpreter: %s\n", imgp->interpreter_name);
257 print_execve_args(args);
261 * If the current process has a special image activator it
262 * wants to try first, call it. For example, emulating shell
263 * scripts differently.
266 if ((img_first = imgp->proc->p_sysent->sv_imgact_try) != NULL)
267 error = img_first(imgp);
270 * If the vnode has a registered vmspace, exec the vmspace
272 if (error == -1 && imgp->vp->v_resident) {
273 error = exec_resident_imgact(imgp);
277 * Loop through the list of image activators, calling each one.
278 * An activator returns -1 if there is no match, 0 on success,
279 * and an error otherwise.
281 for (i = 0; error == -1 && execsw[i]; ++i) {
282 if (execsw[i]->ex_imgact == NULL ||
283 execsw[i]->ex_imgact == img_first) {
286 error = (*execsw[i]->ex_imgact)(imgp);
292 goto exec_fail_dealloc;
296 * Special interpreter operation, cleanup and loop up to try to
297 * activate the interpreter.
299 if (imgp->interpreted) {
300 exec_unmap_first_page(imgp);
304 error = nlookup_init(nd, imgp->interpreter_name, UIO_SYSSPACE,
312 * Copy out strings (args and env) and initialize stack base
314 stack_base = exec_copyout_strings(imgp);
315 p->p_vmspace->vm_minsaddr = (char *)stack_base;
318 * If custom stack fixup routine present for this process
319 * let it do the stack setup. If we are running a resident
320 * image there is no auxinfo or other image activator context
321 * so don't try to add fixups to the stack.
323 * Else stuff argument count as first item on stack
325 if (p->p_sysent->sv_fixup && imgp->resident == 0)
326 (*p->p_sysent->sv_fixup)(&stack_base, imgp);
328 suword(--stack_base, imgp->args->argc);
331 * For security and other reasons, the file descriptor table cannot
332 * be shared after an exec.
334 if (p->p_fd->fd_refcnt > 1) {
335 struct filedesc *tmp;
342 * For security and other reasons, signal handlers cannot
343 * be shared after an exec. The new proces gets a copy of the old
344 * handlers. In execsigs(), the new process will have its signals
347 if (p->p_sigacts->ps_refcnt > 1) {
348 struct sigacts *newsigacts;
350 newsigacts = (struct sigacts *)kmalloc(sizeof(*newsigacts),
351 M_SUBPROC, M_WAITOK);
352 bcopy(p->p_sigacts, newsigacts, sizeof(*newsigacts));
353 p->p_sigacts->ps_refcnt--;
354 p->p_sigacts = newsigacts;
355 p->p_sigacts->ps_refcnt = 1;
359 * For security and other reasons virtual kernels cannot be
360 * inherited by an exec. This also allows a virtual kernel
361 * to fork/exec unrelated applications.
369 /* close files on exec */
372 /* reset caught signals */
375 /* name this process - nameiexec(p, ndp) */
376 len = min(nd->nl_nch.ncp->nc_nlen, MAXCOMLEN);
377 bcopy(nd->nl_nch.ncp->nc_name, p->p_comm, len);
379 bcopy(p->p_comm, lp->lwp_thread->td_comm, MAXCOMLEN+1);
382 * mark as execed, wakeup the process that vforked (if any) and tell
383 * it that it now has its own resources back
386 if (p->p_pptr && (p->p_flag & P_PPWAIT)) {
387 p->p_flag &= ~P_PPWAIT;
388 wakeup((caddr_t)p->p_pptr);
392 * Implement image setuid/setgid.
394 * Don't honor setuid/setgid if the filesystem prohibits it or if
395 * the process is being traced.
397 if ((((attr.va_mode & VSUID) && p->p_ucred->cr_uid != attr.va_uid) ||
398 ((attr.va_mode & VSGID) && p->p_ucred->cr_gid != attr.va_gid)) &&
399 (imgp->vp->v_mount->mnt_flag & MNT_NOSUID) == 0 &&
400 (p->p_flag & P_TRACED) == 0) {
402 * Turn off syscall tracing for set-id programs, except for
403 * root. Record any set-id flags first to make sure that
404 * we do not regain any tracing during a possible block.
407 if (p->p_tracenode && ktrace_suid == 0 &&
408 priv_check(td, PRIV_ROOT) != 0) {
409 ktrdestroy(&p->p_tracenode);
412 /* Close any file descriptors 0..2 that reference procfs */
414 /* Make sure file descriptors 0..2 are in use. */
415 error = fdcheckstd(lp);
417 goto exec_fail_dealloc;
419 * Set the new credentials.
422 if (attr.va_mode & VSUID)
423 change_euid(attr.va_uid);
424 if (attr.va_mode & VSGID)
425 p->p_ucred->cr_gid = attr.va_gid;
428 * Clear local varsym variables
430 varsymset_clean(&p->p_varsymset);
432 if (p->p_ucred->cr_uid == p->p_ucred->cr_ruid &&
433 p->p_ucred->cr_gid == p->p_ucred->cr_rgid)
434 p->p_flag &= ~P_SUGID;
438 * Implement correct POSIX saved-id behavior.
440 if (p->p_ucred->cr_svuid != p->p_ucred->cr_uid ||
441 p->p_ucred->cr_svgid != p->p_ucred->cr_gid) {
443 p->p_ucred->cr_svuid = p->p_ucred->cr_uid;
444 p->p_ucred->cr_svgid = p->p_ucred->cr_gid;
448 * Store the vp for use in procfs
450 if (p->p_textvp) /* release old reference */
452 p->p_textvp = imgp->vp;
455 /* Release old namecache handle to text file */
456 if (p->p_textnch.ncp)
457 cache_drop(&p->p_textnch);
459 if (nd->nl_nch.mount)
460 cache_copy(&nd->nl_nch, &p->p_textnch);
463 * Notify others that we exec'd, and clear the P_INEXEC flag
464 * as we're now a bona fide freshly-execed process.
466 KNOTE(&p->p_klist, NOTE_EXEC);
467 p->p_flag &= ~P_INEXEC;
470 * If tracing the process, trap to debugger so breakpoints
471 * can be set before the program executes.
473 STOPEVENT(p, S_EXEC, 0);
475 if (p->p_flag & P_TRACED)
478 /* clear "fork but no exec" flag, as we _are_ execing */
479 p->p_acflag &= ~AFORK;
481 /* Set values passed into the program in registers. */
482 exec_setregs(imgp->entry_addr, (u_long)(uintptr_t)stack_base,
485 /* Set the access time on the vnode */
486 vn_mark_atime(imgp->vp, td);
488 /* Free any previous argument cache */
489 if (p->p_args && --p->p_args->ar_ref == 0)
490 FREE(p->p_args, M_PARGS);
493 /* Cache arguments if they fit inside our allowance */
494 i = imgp->args->begin_envv - imgp->args->begin_argv;
495 if (ps_arg_cache_limit >= i + sizeof(struct pargs)) {
496 MALLOC(p->p_args, struct pargs *, sizeof(struct pargs) + i,
498 p->p_args->ar_ref = 1;
499 p->p_args->ar_length = i;
500 bcopy(imgp->args->begin_argv, p->p_args->ar_args, i);
506 * free various allocated resources
509 exec_unmap_first_page(imgp);
517 ++mycpu->gd_cnt.v_exec;
523 * we're done here, clear P_INEXEC if we were the ones that
524 * set it. Otherwise if vmspace_destroyed is still set we
525 * raced another thread and that thread is responsible for
528 if (imgp->vmspace_destroyed & 2)
529 p->p_flag &= ~P_INEXEC;
530 if (imgp->vmspace_destroyed) {
532 * Sorry, no more process anymore. exit gracefully.
533 * However we can't die right here, because our
534 * caller might have to clean up, so indicate a
535 * lethal error by returning -1.
544 * execve() system call.
549 sys_execve(struct execve_args *uap)
551 struct nlookupdata nd;
552 struct image_args args;
555 bzero(&args, sizeof(args));
558 error = nlookup_init(&nd, uap->fname, UIO_USERSPACE, NLC_FOLLOW);
560 error = exec_copyin_args(&args, uap->fname, PATH_USERSPACE,
561 uap->argv, uap->envv);
564 error = kern_execve(&nd, &args);
566 exec_free_args(&args);
569 /* We hit a lethal error condition. Let's die now. */
570 exit1(W_EXITCODE(0, SIGABRT));
576 * The syscall result is returned in registers to the new program.
577 * Linux will register %edx as an atexit function and we must be
578 * sure to set it to 0. XXX
581 uap->sysmsg_result64 = 0;
587 exec_map_page(struct image_params *imgp, vm_pindex_t pageno,
588 struct lwbuf **plwb, const char **pdata)
596 * The file has to be mappable.
598 if ((object = imgp->vp->v_object) == NULL)
601 if (pageno >= object->size)
604 m = vm_page_grab(object, pageno, VM_ALLOC_NORMAL | VM_ALLOC_RETRY);
606 lwkt_gettoken(&vm_token);
607 while ((m->valid & VM_PAGE_BITS_ALL) != VM_PAGE_BITS_ALL) {
611 * get_pages unbusies all the requested pages except the
612 * primary page (at index 0 in this case). The primary
613 * page may have been wired during the pagein (e.g. by
614 * the buffer cache) so vnode_pager_freepage() must be
615 * used to properly release it.
617 rv = vm_pager_get_page(object, &ma, 1);
618 m = vm_page_lookup(object, pageno);
620 if (rv != VM_PAGER_OK || m == NULL || m->valid == 0) {
622 vm_page_protect(m, VM_PROT_NONE);
623 vnode_pager_freepage(m);
625 lwkt_reltoken(&vm_token);
629 vm_page_hold(m); /* requires vm_token to be held */
630 vm_page_wakeup(m); /* unbusy the page */
631 lwkt_reltoken(&vm_token);
633 *plwb = lwbuf_alloc(m, *plwb);
634 *pdata = (void *)lwbuf_kva(*plwb);
640 exec_map_first_page(struct image_params *imgp)
645 exec_unmap_first_page(imgp);
647 imgp->firstpage = &imgp->firstpage_cache;
648 err = exec_map_page(imgp, 0, &imgp->firstpage, &imgp->image_header);
657 exec_unmap_page(struct lwbuf *lwb)
671 exec_unmap_first_page(struct image_params *imgp)
673 exec_unmap_page(imgp->firstpage);
674 imgp->firstpage = NULL;
675 imgp->image_header = NULL;
679 * Destroy old address space, and allocate a new stack
680 * The new stack is only SGROWSIZ large because it is grown
681 * automatically in trap.c.
683 * This is the point of no return.
686 exec_new_vmspace(struct image_params *imgp, struct vmspace *vmcopy)
688 struct vmspace *vmspace = imgp->proc->p_vmspace;
689 vm_offset_t stack_addr = USRSTACK - maxssiz;
695 * Indicate that we cannot gracefully error out any more, kill
696 * any other threads present, and set P_INEXEC to indicate that
697 * we are now messing with the process structure proper.
699 * If killalllwps() races return an error which coupled with
700 * vmspace_destroyed will cause us to exit. This is what we
701 * want since another thread is patiently waiting for us to exit
705 imgp->vmspace_destroyed = 1;
707 if (curthread->td_proc->p_nthreads > 1) {
708 error = killalllwps(1);
712 imgp->vmspace_destroyed |= 2; /* we are responsible for P_INEXEC */
713 p->p_flag |= P_INEXEC;
716 * Prevent a pending AIO from modifying the new address space.
718 aio_proc_rundown(imgp->proc);
721 * Blow away entire process VM, if address space not shared,
722 * otherwise, create a new VM space so that other threads are
723 * not disrupted. If we are execing a resident vmspace we
724 * create a duplicate of it and remap the stack.
726 * The exitingcnt test is not strictly necessary but has been
727 * included for code sanity (to make the code more deterministic).
729 map = &vmspace->vm_map;
731 vmspace_exec(imgp->proc, vmcopy);
732 vmspace = imgp->proc->p_vmspace;
733 pmap_remove_pages(vmspace_pmap(vmspace), stack_addr, USRSTACK);
734 map = &vmspace->vm_map;
735 } else if (vmspace->vm_sysref.refcnt == 1 &&
736 vmspace->vm_exitingcnt == 0) {
738 if (vmspace->vm_upcalls)
739 upc_release(vmspace, ONLY_LWP_IN_PROC(imgp->proc));
740 pmap_remove_pages(vmspace_pmap(vmspace),
741 0, VM_MAX_USER_ADDRESS);
742 vm_map_remove(map, 0, VM_MAX_USER_ADDRESS);
744 vmspace_exec(imgp->proc, NULL);
745 vmspace = imgp->proc->p_vmspace;
746 map = &vmspace->vm_map;
749 /* Allocate a new stack */
750 error = vm_map_stack(&vmspace->vm_map, stack_addr, (vm_size_t)maxssiz,
751 0, VM_PROT_ALL, VM_PROT_ALL, 0);
755 /* vm_ssize and vm_maxsaddr are somewhat antiquated concepts in the
756 * VM_STACK case, but they are still used to monitor the size of the
757 * process stack so we can check the stack rlimit.
759 vmspace->vm_ssize = sgrowsiz >> PAGE_SHIFT;
760 vmspace->vm_maxsaddr = (char *)USRSTACK - maxssiz;
766 * Copy out argument and environment strings from the old process
767 * address space into the temporary string buffer.
770 exec_copyin_args(struct image_args *args, char *fname,
771 enum exec_path_segflg segflg, char **argv, char **envv)
777 args->buf = objcache_get(exec_objcache, M_WAITOK);
778 if (args->buf == NULL)
780 args->begin_argv = args->buf;
781 args->endp = args->begin_argv;
782 args->space = ARG_MAX;
784 args->fname = args->buf + ARG_MAX;
787 * Copy the file name.
789 if (segflg == PATH_SYSSPACE) {
790 error = copystr(fname, args->fname, PATH_MAX, &length);
791 } else if (segflg == PATH_USERSPACE) {
792 error = copyinstr(fname, args->fname, PATH_MAX, &length);
796 * Extract argument strings. argv may not be NULL. The argv
797 * array is terminated by a NULL entry. We special-case the
798 * situation where argv[0] is NULL by passing { filename, NULL }
799 * to the new program to guarentee that the interpreter knows what
800 * file to open in case we exec an interpreted file. Note that
801 * a NULL argv[0] terminates the argv[] array.
803 * XXX the special-casing of argv[0] is historical and needs to be
809 while ((argp = (caddr_t)(intptr_t)fuword(argv++)) != NULL) {
810 if (argp == (caddr_t)-1) {
814 error = copyinstr(argp, args->endp,
815 args->space, &length);
817 if (error == ENAMETOOLONG)
821 args->space -= length;
822 args->endp += length;
825 if (args->argc == 0 && error == 0) {
826 length = strlen(args->fname) + 1;
827 if (length > args->space) {
830 bcopy(args->fname, args->endp, length);
831 args->space -= length;
832 args->endp += length;
838 args->begin_envv = args->endp;
841 * extract environment strings. envv may be NULL.
843 if (envv && error == 0) {
844 while ((envp = (caddr_t) (intptr_t) fuword(envv++))) {
845 if (envp == (caddr_t) -1) {
849 error = copyinstr(envp, args->endp, args->space,
852 if (error == ENAMETOOLONG)
856 args->space -= length;
857 args->endp += length;
865 exec_free_args(struct image_args *args)
868 objcache_put(exec_objcache, args->buf);
874 * Copy strings out to the new process address space, constructing
875 * new arg and env vector tables. Return a pointer to the base
876 * so that it can be used as the initial stack pointer.
879 exec_copyout_strings(struct image_params *imgp)
881 int argc, envc, sgap;
883 char *stringp, *destp;
884 register_t *stack_base;
885 struct ps_strings *arginfo;
889 * Calculate string base and vector table pointers.
890 * Also deal with signal trampoline code for this exec type.
892 arginfo = (struct ps_strings *)PS_STRINGS;
893 szsigcode = *(imgp->proc->p_sysent->sv_szsigcode);
894 if (stackgap_random != 0)
895 sgap = ALIGN(karc4random() & (stackgap_random - 1));
898 destp = (caddr_t)arginfo - szsigcode - SPARE_USRSPACE - sgap -
899 roundup((ARG_MAX - imgp->args->space), sizeof(char *));
905 copyout(imgp->proc->p_sysent->sv_sigcode,
906 ((caddr_t)arginfo - szsigcode), szsigcode);
909 * If we have a valid auxargs ptr, prepare some room
912 * The '+ 2' is for the null pointers at the end of each of the
913 * arg and env vector sets, and 'AT_COUNT*2' is room for the
917 vectp = (char **)(destp - (imgp->args->argc +
918 imgp->args->envc + 2 + AT_COUNT * 2) * sizeof(char*));
920 vectp = (char **)(destp - (imgp->args->argc +
921 imgp->args->envc + 2) * sizeof(char*));
925 * NOTE: don't bother aligning the stack here for GCC 2.x, it will
926 * be done in crt1.o. Note that GCC 3.x aligns the stack in main.
930 * vectp also becomes our initial stack base
932 stack_base = (register_t *)vectp;
934 stringp = imgp->args->begin_argv;
935 argc = imgp->args->argc;
936 envc = imgp->args->envc;
939 * Copy out strings - arguments and environment.
941 copyout(stringp, destp, ARG_MAX - imgp->args->space);
944 * Fill in "ps_strings" struct for ps, w, etc.
946 suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp);
947 suword(&arginfo->ps_nargvstr, argc);
950 * Fill in argument portion of vector table.
952 for (; argc > 0; --argc) {
953 suword(vectp++, (long)(intptr_t)destp);
954 while (*stringp++ != 0)
959 /* a null vector table pointer separates the argp's from the envp's */
962 suword(&arginfo->ps_envstr, (long)(intptr_t)vectp);
963 suword(&arginfo->ps_nenvstr, envc);
966 * Fill in environment portion of vector table.
968 for (; envc > 0; --envc) {
969 suword(vectp++, (long)(intptr_t)destp);
970 while (*stringp++ != 0)
975 /* end of vector table is a null pointer */
982 * Check permissions of file to execute.
983 * Return 0 for success or error code on failure.
986 exec_check_permissions(struct image_params *imgp, struct mount *topmnt)
988 struct proc *p = imgp->proc;
989 struct vnode *vp = imgp->vp;
990 struct vattr *attr = imgp->attr;
993 /* Get file attributes */
994 error = VOP_GETATTR(vp, attr);
999 * 1) Check if file execution is disabled for the filesystem that this
1001 * 2) Insure that at least one execute bit is on - otherwise root
1002 * will always succeed, and we don't want to happen unless the
1003 * file really is executable.
1004 * 3) Insure that the file is a regular file.
1006 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
1007 ((topmnt != NULL) && (topmnt->mnt_flag & MNT_NOEXEC)) ||
1008 ((attr->va_mode & 0111) == 0) ||
1009 (attr->va_type != VREG)) {
1014 * Zero length files can't be exec'd
1016 if (attr->va_size == 0)
1020 * Check for execute permission to file based on current credentials.
1022 error = VOP_EACCESS(vp, VEXEC, p->p_ucred);
1027 * Check number of open-for-writes on the file and deny execution
1030 if (vp->v_writecount)
1034 * Call filesystem specific open routine, which allows us to read,
1035 * write, and mmap the file. Without the VOP_OPEN we can only
1038 error = VOP_OPEN(vp, FREAD, p->p_ucred, NULL);
1046 * Exec handler registration
1049 exec_register(const struct execsw *execsw_arg)
1051 const struct execsw **es, **xs, **newexecsw;
1052 int count = 2; /* New slot and trailing NULL */
1055 for (es = execsw; *es; es++)
1057 newexecsw = kmalloc(count * sizeof(*es), M_TEMP, M_WAITOK);
1060 for (es = execsw; *es; es++)
1065 kfree(execsw, M_TEMP);
1071 exec_unregister(const struct execsw *execsw_arg)
1073 const struct execsw **es, **xs, **newexecsw;
1077 panic("unregister with no handlers left?");
1079 for (es = execsw; *es; es++) {
1080 if (*es == execsw_arg)
1085 for (es = execsw; *es; es++)
1086 if (*es != execsw_arg)
1088 newexecsw = kmalloc(count * sizeof(*es), M_TEMP, M_WAITOK);
1090 for (es = execsw; *es; es++)
1091 if (*es != execsw_arg)
1095 kfree(execsw, M_TEMP);