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>
55 #include <sys/libkern.h>
57 #include <cpu/lwbuf.h>
60 #include <vm/vm_param.h>
63 #include <vm/vm_page.h>
64 #include <vm/vm_map.h>
65 #include <vm/vm_kern.h>
66 #include <vm/vm_extern.h>
67 #include <vm/vm_object.h>
68 #include <vm/vnode_pager.h>
69 #include <vm/vm_pager.h>
74 #include <sys/refcount.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;
190 struct image_params image_params, *imgp;
192 int (*img_first) (struct image_params *);
194 if (debug_execve_args) {
195 kprintf("%s()\n", __func__);
196 print_execve_args(args);
200 lwkt_gettoken(&p->p_token);
201 imgp = &image_params;
204 * NOTE: P_INEXEC is handled by exec_new_vmspace() now. We make
205 * no modifications to the process at all until we get there.
207 * Note that multiple threads may be trying to exec at the same
208 * time. exec_new_vmspace() handles that too.
212 * Initialize part of the common data
217 imgp->entry_addr = 0;
219 imgp->vmspace_destroyed = 0;
220 imgp->interpreted = 0;
221 imgp->interpreter_name[0] = 0;
222 imgp->auxargs = NULL;
224 imgp->firstpage = NULL;
225 imgp->ps_strings = 0;
226 imgp->image_header = NULL;
231 * Translate the file name to a vnode. Unlock the cache entry to
232 * improve parallelism for programs exec'd in parallel.
234 if ((error = nlookup(nd)) != 0)
236 error = cache_vget(&nd->nl_nch, nd->nl_cred, LK_EXCLUSIVE, &imgp->vp);
237 KKASSERT(nd->nl_flags & NLC_NCPISLOCKED);
238 nd->nl_flags &= ~NLC_NCPISLOCKED;
239 cache_unlock(&nd->nl_nch);
244 * Check file permissions (also 'opens' file).
245 * Include also the top level mount in the check.
247 error = exec_check_permissions(imgp, nd->nl_nch.mount);
250 goto exec_fail_dealloc;
253 error = exec_map_first_page(imgp);
256 goto exec_fail_dealloc;
258 if (debug_execve_args && imgp->interpreted) {
259 kprintf(" target is interpreted -- recursive pass\n");
260 kprintf(" interpreter: %s\n", imgp->interpreter_name);
261 print_execve_args(args);
265 * If the current process has a special image activator it
266 * wants to try first, call it. For example, emulating shell
267 * scripts differently.
270 if ((img_first = imgp->proc->p_sysent->sv_imgact_try) != NULL)
271 error = img_first(imgp);
274 * If the vnode has a registered vmspace, exec the vmspace
276 if (error == -1 && imgp->vp->v_resident) {
277 error = exec_resident_imgact(imgp);
281 * Loop through the list of image activators, calling each one.
282 * An activator returns -1 if there is no match, 0 on success,
283 * and an error otherwise.
285 for (i = 0; error == -1 && execsw[i]; ++i) {
286 if (execsw[i]->ex_imgact == NULL ||
287 execsw[i]->ex_imgact == img_first) {
290 error = (*execsw[i]->ex_imgact)(imgp);
296 goto exec_fail_dealloc;
300 * Special interpreter operation, cleanup and loop up to try to
301 * activate the interpreter.
303 if (imgp->interpreted) {
304 exec_unmap_first_page(imgp);
308 error = nlookup_init(nd, imgp->interpreter_name, UIO_SYSSPACE,
316 * Copy out strings (args and env) and initialize stack base
318 stack_base = exec_copyout_strings(imgp);
319 p->p_vmspace->vm_minsaddr = (char *)stack_base;
322 * If custom stack fixup routine present for this process
323 * let it do the stack setup. If we are running a resident
324 * image there is no auxinfo or other image activator context
325 * so don't try to add fixups to the stack.
327 * Else stuff argument count as first item on stack
329 if (p->p_sysent->sv_fixup && imgp->resident == 0)
330 (*p->p_sysent->sv_fixup)(&stack_base, imgp);
332 suword(--stack_base, imgp->args->argc);
335 * For security and other reasons, the file descriptor table cannot
336 * be shared after an exec.
338 if (p->p_fd->fd_refcnt > 1) {
339 struct filedesc *tmp;
346 * For security and other reasons, signal handlers cannot
347 * be shared after an exec. The new proces gets a copy of the old
348 * handlers. In execsigs(), the new process will have its signals
352 if (ops->ps_refcnt > 1) {
353 nps = kmalloc(sizeof(*nps), M_SUBPROC, M_WAITOK);
354 bcopy(ops, nps, sizeof(*nps));
355 refcount_init(&nps->ps_refcnt, 1);
357 if (refcount_release(&ops->ps_refcnt)) {
358 kfree(ops, M_SUBPROC);
364 * For security and other reasons virtual kernels cannot be
365 * inherited by an exec. This also allows a virtual kernel
366 * to fork/exec unrelated applications.
374 /* close files on exec */
377 /* reset caught signals */
380 /* name this process - nameiexec(p, ndp) */
381 len = min(nd->nl_nch.ncp->nc_nlen, MAXCOMLEN);
382 bcopy(nd->nl_nch.ncp->nc_name, p->p_comm, len);
384 bcopy(p->p_comm, lp->lwp_thread->td_comm, MAXCOMLEN+1);
387 * mark as execed, wakeup the process that vforked (if any) and tell
388 * it that it now has its own resources back
391 if (p->p_pptr && (p->p_flag & P_PPWAIT)) {
392 p->p_flag &= ~P_PPWAIT;
393 wakeup((caddr_t)p->p_pptr);
397 * Implement image setuid/setgid.
399 * Don't honor setuid/setgid if the filesystem prohibits it or if
400 * the process is being traced.
402 if ((((attr.va_mode & VSUID) && p->p_ucred->cr_uid != attr.va_uid) ||
403 ((attr.va_mode & VSGID) && p->p_ucred->cr_gid != attr.va_gid)) &&
404 (imgp->vp->v_mount->mnt_flag & MNT_NOSUID) == 0 &&
405 (p->p_flag & P_TRACED) == 0) {
407 * Turn off syscall tracing for set-id programs, except for
408 * root. Record any set-id flags first to make sure that
409 * we do not regain any tracing during a possible block.
412 if (p->p_tracenode && ktrace_suid == 0 &&
413 priv_check(td, PRIV_ROOT) != 0) {
414 ktrdestroy(&p->p_tracenode);
417 /* Close any file descriptors 0..2 that reference procfs */
419 /* Make sure file descriptors 0..2 are in use. */
420 error = fdcheckstd(lp);
422 goto exec_fail_dealloc;
424 * Set the new credentials.
427 if (attr.va_mode & VSUID)
428 change_euid(attr.va_uid);
429 if (attr.va_mode & VSGID)
430 p->p_ucred->cr_gid = attr.va_gid;
433 * Clear local varsym variables
435 varsymset_clean(&p->p_varsymset);
437 if (p->p_ucred->cr_uid == p->p_ucred->cr_ruid &&
438 p->p_ucred->cr_gid == p->p_ucred->cr_rgid)
439 p->p_flag &= ~P_SUGID;
443 * Implement correct POSIX saved-id behavior.
445 if (p->p_ucred->cr_svuid != p->p_ucred->cr_uid ||
446 p->p_ucred->cr_svgid != p->p_ucred->cr_gid) {
448 p->p_ucred->cr_svuid = p->p_ucred->cr_uid;
449 p->p_ucred->cr_svgid = p->p_ucred->cr_gid;
453 * Store the vp for use in procfs
455 if (p->p_textvp) /* release old reference */
457 p->p_textvp = imgp->vp;
460 /* Release old namecache handle to text file */
461 if (p->p_textnch.ncp)
462 cache_drop(&p->p_textnch);
464 if (nd->nl_nch.mount)
465 cache_copy(&nd->nl_nch, &p->p_textnch);
468 * Notify others that we exec'd, and clear the P_INEXEC flag
469 * as we're now a bona fide freshly-execed process.
471 KNOTE(&p->p_klist, NOTE_EXEC);
472 p->p_flag &= ~P_INEXEC;
475 * If tracing the process, trap to debugger so breakpoints
476 * can be set before the program executes.
478 STOPEVENT(p, S_EXEC, 0);
480 if (p->p_flag & P_TRACED)
483 /* clear "fork but no exec" flag, as we _are_ execing */
484 p->p_acflag &= ~AFORK;
486 /* Set values passed into the program in registers. */
487 exec_setregs(imgp->entry_addr, (u_long)(uintptr_t)stack_base,
490 /* Set the access time on the vnode */
491 vn_mark_atime(imgp->vp, td);
494 * Free any previous argument cache
498 if (pa && refcount_release(&pa->ar_ref)) {
504 * Cache arguments if they fit inside our allowance
506 i = imgp->args->begin_envv - imgp->args->begin_argv;
507 if (sizeof(struct pargs) + i <= ps_arg_cache_limit) {
508 pa = kmalloc(sizeof(struct pargs) + i, M_PARGS, M_WAITOK);
509 refcount_init(&pa->ar_ref, 1);
511 bcopy(imgp->args->begin_argv, pa->ar_args, i);
512 KKASSERT(p->p_args == NULL);
519 * free various allocated resources
522 exec_unmap_first_page(imgp);
530 ++mycpu->gd_cnt.v_exec;
531 lwkt_reltoken(&p->p_token);
537 * we're done here, clear P_INEXEC if we were the ones that
538 * set it. Otherwise if vmspace_destroyed is still set we
539 * raced another thread and that thread is responsible for
542 if (imgp->vmspace_destroyed & 2)
543 p->p_flag &= ~P_INEXEC;
544 lwkt_reltoken(&p->p_token);
545 if (imgp->vmspace_destroyed) {
547 * Sorry, no more process anymore. exit gracefully.
548 * However we can't die right here, because our
549 * caller might have to clean up, so indicate a
550 * lethal error by returning -1.
559 * execve() system call.
564 sys_execve(struct execve_args *uap)
566 struct nlookupdata nd;
567 struct image_args args;
570 bzero(&args, sizeof(args));
573 error = nlookup_init(&nd, uap->fname, UIO_USERSPACE, NLC_FOLLOW);
575 error = exec_copyin_args(&args, uap->fname, PATH_USERSPACE,
576 uap->argv, uap->envv);
579 error = kern_execve(&nd, &args);
581 exec_free_args(&args);
584 /* We hit a lethal error condition. Let's die now. */
585 exit1(W_EXITCODE(0, SIGABRT));
591 * The syscall result is returned in registers to the new program.
592 * Linux will register %edx as an atexit function and we must be
593 * sure to set it to 0. XXX
596 uap->sysmsg_result64 = 0;
602 exec_map_page(struct image_params *imgp, vm_pindex_t pageno,
603 struct lwbuf **plwb, const char **pdata)
611 * The file has to be mappable.
613 if ((object = imgp->vp->v_object) == NULL)
616 if (pageno >= object->size)
619 m = vm_page_grab(object, pageno, VM_ALLOC_NORMAL | VM_ALLOC_RETRY);
621 lwkt_gettoken(&vm_token);
622 while ((m->valid & VM_PAGE_BITS_ALL) != VM_PAGE_BITS_ALL) {
626 * get_pages unbusies all the requested pages except the
627 * primary page (at index 0 in this case). The primary
628 * page may have been wired during the pagein (e.g. by
629 * the buffer cache) so vnode_pager_freepage() must be
630 * used to properly release it.
632 rv = vm_pager_get_page(object, &ma, 1);
633 m = vm_page_lookup(object, pageno);
635 if (rv != VM_PAGER_OK || m == NULL || m->valid == 0) {
637 vm_page_protect(m, VM_PROT_NONE);
638 vnode_pager_freepage(m);
640 lwkt_reltoken(&vm_token);
644 vm_page_hold(m); /* requires vm_token to be held */
645 vm_page_wakeup(m); /* unbusy the page */
646 lwkt_reltoken(&vm_token);
648 *plwb = lwbuf_alloc(m, *plwb);
649 *pdata = (void *)lwbuf_kva(*plwb);
655 exec_map_first_page(struct image_params *imgp)
660 exec_unmap_first_page(imgp);
662 imgp->firstpage = &imgp->firstpage_cache;
663 err = exec_map_page(imgp, 0, &imgp->firstpage, &imgp->image_header);
672 exec_unmap_page(struct lwbuf *lwb)
686 exec_unmap_first_page(struct image_params *imgp)
688 exec_unmap_page(imgp->firstpage);
689 imgp->firstpage = NULL;
690 imgp->image_header = NULL;
694 * Destroy old address space, and allocate a new stack
695 * The new stack is only SGROWSIZ large because it is grown
696 * automatically in trap.c.
698 * This is the point of no return.
701 exec_new_vmspace(struct image_params *imgp, struct vmspace *vmcopy)
703 struct vmspace *vmspace = imgp->proc->p_vmspace;
704 vm_offset_t stack_addr = USRSTACK - maxssiz;
710 * Indicate that we cannot gracefully error out any more, kill
711 * any other threads present, and set P_INEXEC to indicate that
712 * we are now messing with the process structure proper.
714 * If killalllwps() races return an error which coupled with
715 * vmspace_destroyed will cause us to exit. This is what we
716 * want since another thread is patiently waiting for us to exit
720 imgp->vmspace_destroyed = 1;
722 if (curthread->td_proc->p_nthreads > 1) {
723 error = killalllwps(1);
727 imgp->vmspace_destroyed |= 2; /* we are responsible for P_INEXEC */
728 p->p_flag |= P_INEXEC;
731 * Blow away entire process VM, if address space not shared,
732 * otherwise, create a new VM space so that other threads are
733 * not disrupted. If we are execing a resident vmspace we
734 * create a duplicate of it and remap the stack.
736 * The exitingcnt test is not strictly necessary but has been
737 * included for code sanity (to make the code more deterministic).
739 map = &vmspace->vm_map;
741 vmspace_exec(imgp->proc, vmcopy);
742 vmspace = imgp->proc->p_vmspace;
743 pmap_remove_pages(vmspace_pmap(vmspace), stack_addr, USRSTACK);
744 map = &vmspace->vm_map;
745 } else if (vmspace->vm_sysref.refcnt == 1 &&
746 vmspace->vm_exitingcnt == 0) {
748 if (vmspace->vm_upcalls)
749 upc_release(vmspace, ONLY_LWP_IN_PROC(imgp->proc));
750 pmap_remove_pages(vmspace_pmap(vmspace),
751 0, VM_MAX_USER_ADDRESS);
752 vm_map_remove(map, 0, VM_MAX_USER_ADDRESS);
754 vmspace_exec(imgp->proc, NULL);
755 vmspace = imgp->proc->p_vmspace;
756 map = &vmspace->vm_map;
759 /* Allocate a new stack */
760 error = vm_map_stack(&vmspace->vm_map, stack_addr, (vm_size_t)maxssiz,
761 0, VM_PROT_ALL, VM_PROT_ALL, 0);
765 /* vm_ssize and vm_maxsaddr are somewhat antiquated concepts in the
766 * VM_STACK case, but they are still used to monitor the size of the
767 * process stack so we can check the stack rlimit.
769 vmspace->vm_ssize = sgrowsiz >> PAGE_SHIFT;
770 vmspace->vm_maxsaddr = (char *)USRSTACK - maxssiz;
776 * Copy out argument and environment strings from the old process
777 * address space into the temporary string buffer.
780 exec_copyin_args(struct image_args *args, char *fname,
781 enum exec_path_segflg segflg, char **argv, char **envv)
787 args->buf = objcache_get(exec_objcache, M_WAITOK);
788 if (args->buf == NULL)
790 args->begin_argv = args->buf;
791 args->endp = args->begin_argv;
792 args->space = ARG_MAX;
794 args->fname = args->buf + ARG_MAX;
797 * Copy the file name.
799 if (segflg == PATH_SYSSPACE) {
800 error = copystr(fname, args->fname, PATH_MAX, &length);
801 } else if (segflg == PATH_USERSPACE) {
802 error = copyinstr(fname, args->fname, PATH_MAX, &length);
806 * Extract argument strings. argv may not be NULL. The argv
807 * array is terminated by a NULL entry. We special-case the
808 * situation where argv[0] is NULL by passing { filename, NULL }
809 * to the new program to guarentee that the interpreter knows what
810 * file to open in case we exec an interpreted file. Note that
811 * a NULL argv[0] terminates the argv[] array.
813 * XXX the special-casing of argv[0] is historical and needs to be
819 while ((argp = (caddr_t)(intptr_t)fuword(argv++)) != NULL) {
820 if (argp == (caddr_t)-1) {
824 error = copyinstr(argp, args->endp,
825 args->space, &length);
827 if (error == ENAMETOOLONG)
831 args->space -= length;
832 args->endp += length;
835 if (args->argc == 0 && error == 0) {
836 length = strlen(args->fname) + 1;
837 if (length > args->space) {
840 bcopy(args->fname, args->endp, length);
841 args->space -= length;
842 args->endp += length;
848 args->begin_envv = args->endp;
851 * extract environment strings. envv may be NULL.
853 if (envv && error == 0) {
854 while ((envp = (caddr_t) (intptr_t) fuword(envv++))) {
855 if (envp == (caddr_t) -1) {
859 error = copyinstr(envp, args->endp, args->space,
862 if (error == ENAMETOOLONG)
866 args->space -= length;
867 args->endp += length;
875 exec_free_args(struct image_args *args)
878 objcache_put(exec_objcache, args->buf);
884 * Copy strings out to the new process address space, constructing
885 * new arg and env vector tables. Return a pointer to the base
886 * so that it can be used as the initial stack pointer.
889 exec_copyout_strings(struct image_params *imgp)
891 int argc, envc, sgap;
893 char *stringp, *destp;
894 register_t *stack_base;
895 struct ps_strings *arginfo;
899 * Calculate string base and vector table pointers.
900 * Also deal with signal trampoline code for this exec type.
902 arginfo = (struct ps_strings *)PS_STRINGS;
903 szsigcode = *(imgp->proc->p_sysent->sv_szsigcode);
904 if (stackgap_random != 0)
905 sgap = ALIGN(karc4random() & (stackgap_random - 1));
908 destp = (caddr_t)arginfo - szsigcode - SPARE_USRSPACE - sgap -
909 roundup((ARG_MAX - imgp->args->space), sizeof(char *));
915 copyout(imgp->proc->p_sysent->sv_sigcode,
916 ((caddr_t)arginfo - szsigcode), szsigcode);
919 * If we have a valid auxargs ptr, prepare some room
922 * The '+ 2' is for the null pointers at the end of each of the
923 * arg and env vector sets, and 'AT_COUNT*2' is room for the
927 vectp = (char **)(destp - (imgp->args->argc +
928 imgp->args->envc + 2 + AT_COUNT * 2) * sizeof(char*));
930 vectp = (char **)(destp - (imgp->args->argc +
931 imgp->args->envc + 2) * sizeof(char*));
935 * NOTE: don't bother aligning the stack here for GCC 2.x, it will
936 * be done in crt1.o. Note that GCC 3.x aligns the stack in main.
940 * vectp also becomes our initial stack base
942 stack_base = (register_t *)vectp;
944 stringp = imgp->args->begin_argv;
945 argc = imgp->args->argc;
946 envc = imgp->args->envc;
949 * Copy out strings - arguments and environment.
951 copyout(stringp, destp, ARG_MAX - imgp->args->space);
954 * Fill in "ps_strings" struct for ps, w, etc.
956 suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp);
957 suword(&arginfo->ps_nargvstr, argc);
960 * Fill in argument portion of vector table.
962 for (; argc > 0; --argc) {
963 suword(vectp++, (long)(intptr_t)destp);
964 while (*stringp++ != 0)
969 /* a null vector table pointer separates the argp's from the envp's */
972 suword(&arginfo->ps_envstr, (long)(intptr_t)vectp);
973 suword(&arginfo->ps_nenvstr, envc);
976 * Fill in environment portion of vector table.
978 for (; envc > 0; --envc) {
979 suword(vectp++, (long)(intptr_t)destp);
980 while (*stringp++ != 0)
985 /* end of vector table is a null pointer */
992 * Check permissions of file to execute.
993 * Return 0 for success or error code on failure.
996 exec_check_permissions(struct image_params *imgp, struct mount *topmnt)
998 struct proc *p = imgp->proc;
999 struct vnode *vp = imgp->vp;
1000 struct vattr *attr = imgp->attr;
1003 /* Get file attributes */
1004 error = VOP_GETATTR(vp, attr);
1009 * 1) Check if file execution is disabled for the filesystem that this
1011 * 2) Insure that at least one execute bit is on - otherwise root
1012 * will always succeed, and we don't want to happen unless the
1013 * file really is executable.
1014 * 3) Insure that the file is a regular file.
1016 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
1017 ((topmnt != NULL) && (topmnt->mnt_flag & MNT_NOEXEC)) ||
1018 ((attr->va_mode & 0111) == 0) ||
1019 (attr->va_type != VREG)) {
1024 * Zero length files can't be exec'd
1026 if (attr->va_size == 0)
1030 * Check for execute permission to file based on current credentials.
1032 error = VOP_EACCESS(vp, VEXEC, p->p_ucred);
1037 * Check number of open-for-writes on the file and deny execution
1040 if (vp->v_writecount)
1044 * Call filesystem specific open routine, which allows us to read,
1045 * write, and mmap the file. Without the VOP_OPEN we can only
1048 error = VOP_OPEN(vp, FREAD, p->p_ucred, NULL);
1056 * Exec handler registration
1059 exec_register(const struct execsw *execsw_arg)
1061 const struct execsw **es, **xs, **newexecsw;
1062 int count = 2; /* New slot and trailing NULL */
1065 for (es = execsw; *es; es++)
1067 newexecsw = kmalloc(count * sizeof(*es), M_TEMP, M_WAITOK);
1070 for (es = execsw; *es; es++)
1075 kfree(execsw, M_TEMP);
1081 exec_unregister(const struct execsw *execsw_arg)
1083 const struct execsw **es, **xs, **newexecsw;
1087 panic("unregister with no handlers left?");
1089 for (es = execsw; *es; es++) {
1090 if (*es == execsw_arg)
1095 for (es = execsw; *es; es++)
1096 if (*es != execsw_arg)
1098 newexecsw = kmalloc(count * sizeof(*es), M_TEMP, M_WAITOK);
1100 for (es = execsw; *es; es++)
1101 if (*es != execsw_arg)
1105 kfree(execsw, M_TEMP);