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)
242 error = exec_check_permissions(imgp);
245 goto exec_fail_dealloc;
248 error = exec_map_first_page(imgp);
251 goto exec_fail_dealloc;
253 if (debug_execve_args && imgp->interpreted) {
254 kprintf(" target is interpreted -- recursive pass\n");
255 kprintf(" interpreter: %s\n", imgp->interpreter_name);
256 print_execve_args(args);
260 * If the current process has a special image activator it
261 * wants to try first, call it. For example, emulating shell
262 * scripts differently.
265 if ((img_first = imgp->proc->p_sysent->sv_imgact_try) != NULL)
266 error = img_first(imgp);
269 * If the vnode has a registered vmspace, exec the vmspace
271 if (error == -1 && imgp->vp->v_resident) {
272 error = exec_resident_imgact(imgp);
276 * Loop through the list of image activators, calling each one.
277 * An activator returns -1 if there is no match, 0 on success,
278 * and an error otherwise.
280 for (i = 0; error == -1 && execsw[i]; ++i) {
281 if (execsw[i]->ex_imgact == NULL ||
282 execsw[i]->ex_imgact == img_first) {
285 error = (*execsw[i]->ex_imgact)(imgp);
291 goto exec_fail_dealloc;
295 * Special interpreter operation, cleanup and loop up to try to
296 * activate the interpreter.
298 if (imgp->interpreted) {
299 exec_unmap_first_page(imgp);
303 error = nlookup_init(nd, imgp->interpreter_name, UIO_SYSSPACE,
311 * Copy out strings (args and env) and initialize stack base
313 stack_base = exec_copyout_strings(imgp);
314 p->p_vmspace->vm_minsaddr = (char *)stack_base;
317 * If custom stack fixup routine present for this process
318 * let it do the stack setup. If we are running a resident
319 * image there is no auxinfo or other image activator context
320 * so don't try to add fixups to the stack.
322 * Else stuff argument count as first item on stack
324 if (p->p_sysent->sv_fixup && imgp->resident == 0)
325 (*p->p_sysent->sv_fixup)(&stack_base, imgp);
327 suword(--stack_base, imgp->args->argc);
330 * For security and other reasons, the file descriptor table cannot
331 * be shared after an exec.
333 if (p->p_fd->fd_refcnt > 1) {
334 struct filedesc *tmp;
341 * For security and other reasons, signal handlers cannot
342 * be shared after an exec. The new proces gets a copy of the old
343 * handlers. In execsigs(), the new process will have its signals
346 if (p->p_sigacts->ps_refcnt > 1) {
347 struct sigacts *newsigacts;
349 newsigacts = (struct sigacts *)kmalloc(sizeof(*newsigacts),
350 M_SUBPROC, M_WAITOK);
351 bcopy(p->p_sigacts, newsigacts, sizeof(*newsigacts));
352 p->p_sigacts->ps_refcnt--;
353 p->p_sigacts = newsigacts;
354 p->p_sigacts->ps_refcnt = 1;
358 * For security and other reasons virtual kernels cannot be
359 * inherited by an exec. This also allows a virtual kernel
360 * to fork/exec unrelated applications.
368 /* close files on exec */
371 /* reset caught signals */
374 /* name this process - nameiexec(p, ndp) */
375 len = min(nd->nl_nch.ncp->nc_nlen, MAXCOMLEN);
376 bcopy(nd->nl_nch.ncp->nc_name, p->p_comm, len);
378 bcopy(p->p_comm, lp->lwp_thread->td_comm, MAXCOMLEN+1);
381 * mark as execed, wakeup the process that vforked (if any) and tell
382 * it that it now has its own resources back
385 if (p->p_pptr && (p->p_flag & P_PPWAIT)) {
386 p->p_flag &= ~P_PPWAIT;
387 wakeup((caddr_t)p->p_pptr);
391 * Implement image setuid/setgid.
393 * Don't honor setuid/setgid if the filesystem prohibits it or if
394 * the process is being traced.
396 if ((((attr.va_mode & VSUID) && p->p_ucred->cr_uid != attr.va_uid) ||
397 ((attr.va_mode & VSGID) && p->p_ucred->cr_gid != attr.va_gid)) &&
398 (imgp->vp->v_mount->mnt_flag & MNT_NOSUID) == 0 &&
399 (p->p_flag & P_TRACED) == 0) {
401 * Turn off syscall tracing for set-id programs, except for
402 * root. Record any set-id flags first to make sure that
403 * we do not regain any tracing during a possible block.
406 if (p->p_tracenode && ktrace_suid == 0 &&
407 priv_check(td, PRIV_ROOT) != 0) {
408 ktrdestroy(&p->p_tracenode);
411 /* Close any file descriptors 0..2 that reference procfs */
413 /* Make sure file descriptors 0..2 are in use. */
414 error = fdcheckstd(lp);
416 goto exec_fail_dealloc;
418 * Set the new credentials.
421 if (attr.va_mode & VSUID)
422 change_euid(attr.va_uid);
423 if (attr.va_mode & VSGID)
424 p->p_ucred->cr_gid = attr.va_gid;
427 * Clear local varsym variables
429 varsymset_clean(&p->p_varsymset);
431 if (p->p_ucred->cr_uid == p->p_ucred->cr_ruid &&
432 p->p_ucred->cr_gid == p->p_ucred->cr_rgid)
433 p->p_flag &= ~P_SUGID;
437 * Implement correct POSIX saved-id behavior.
439 if (p->p_ucred->cr_svuid != p->p_ucred->cr_uid ||
440 p->p_ucred->cr_svgid != p->p_ucred->cr_gid) {
442 p->p_ucred->cr_svuid = p->p_ucred->cr_uid;
443 p->p_ucred->cr_svgid = p->p_ucred->cr_gid;
447 * Store the vp for use in procfs
449 if (p->p_textvp) /* release old reference */
451 p->p_textvp = imgp->vp;
454 /* Release old namecache handle to text file */
455 if (p->p_textnch.ncp)
456 cache_drop(&p->p_textnch);
458 if (nd->nl_nch.mount)
459 cache_copy(&nd->nl_nch, &p->p_textnch);
462 * Notify others that we exec'd, and clear the P_INEXEC flag
463 * as we're now a bona fide freshly-execed process.
465 KNOTE(&p->p_klist, NOTE_EXEC);
466 p->p_flag &= ~P_INEXEC;
469 * If tracing the process, trap to debugger so breakpoints
470 * can be set before the program executes.
472 STOPEVENT(p, S_EXEC, 0);
474 if (p->p_flag & P_TRACED)
477 /* clear "fork but no exec" flag, as we _are_ execing */
478 p->p_acflag &= ~AFORK;
480 /* Set values passed into the program in registers. */
481 exec_setregs(imgp->entry_addr, (u_long)(uintptr_t)stack_base,
484 /* Set the access time on the vnode */
485 vn_mark_atime(imgp->vp, td);
487 /* Free any previous argument cache */
488 if (p->p_args && --p->p_args->ar_ref == 0)
489 FREE(p->p_args, M_PARGS);
492 /* Cache arguments if they fit inside our allowance */
493 i = imgp->args->begin_envv - imgp->args->begin_argv;
494 if (ps_arg_cache_limit >= i + sizeof(struct pargs)) {
495 MALLOC(p->p_args, struct pargs *, sizeof(struct pargs) + i,
497 p->p_args->ar_ref = 1;
498 p->p_args->ar_length = i;
499 bcopy(imgp->args->begin_argv, p->p_args->ar_args, i);
505 * free various allocated resources
508 exec_unmap_first_page(imgp);
516 ++mycpu->gd_cnt.v_exec;
522 * we're done here, clear P_INEXEC if we were the ones that
523 * set it. Otherwise if vmspace_destroyed is still set we
524 * raced another thread and that thread is responsible for
527 if (imgp->vmspace_destroyed & 2)
528 p->p_flag &= ~P_INEXEC;
529 if (imgp->vmspace_destroyed) {
531 * Sorry, no more process anymore. exit gracefully.
532 * However we can't die right here, because our
533 * caller might have to clean up, so indicate a
534 * lethal error by returning -1.
543 * execve() system call.
548 sys_execve(struct execve_args *uap)
550 struct nlookupdata nd;
551 struct image_args args;
554 bzero(&args, sizeof(args));
557 error = nlookup_init(&nd, uap->fname, UIO_USERSPACE, NLC_FOLLOW);
559 error = exec_copyin_args(&args, uap->fname, PATH_USERSPACE,
560 uap->argv, uap->envv);
563 error = kern_execve(&nd, &args);
565 exec_free_args(&args);
568 /* We hit a lethal error condition. Let's die now. */
569 exit1(W_EXITCODE(0, SIGABRT));
575 * The syscall result is returned in registers to the new program.
576 * Linux will register %edx as an atexit function and we must be
577 * sure to set it to 0. XXX
580 uap->sysmsg_result64 = 0;
586 exec_map_page(struct image_params *imgp, vm_pindex_t pageno,
587 struct lwbuf **plwb, const char **pdata)
595 * The file has to be mappable.
597 if ((object = imgp->vp->v_object) == NULL)
600 if (pageno >= object->size)
604 * We shouldn't need protection for vm_page_grab() but we certainly
605 * need it for the lookup loop below (lookup/busy race), since
606 * an interrupt can unbusy and free the page before our busy check.
608 m = vm_page_grab(object, pageno, VM_ALLOC_NORMAL | VM_ALLOC_RETRY);
610 while ((m->valid & VM_PAGE_BITS_ALL) != VM_PAGE_BITS_ALL) {
614 * get_pages unbusies all the requested pages except the
615 * primary page (at index 0 in this case). The primary
616 * page may have been wired during the pagein (e.g. by
617 * the buffer cache) so vnode_pager_freepage() must be
618 * used to properly release it.
620 rv = vm_pager_get_page(object, &ma, 1);
621 m = vm_page_lookup(object, pageno);
623 if (rv != VM_PAGER_OK || m == NULL || m->valid == 0) {
625 vm_page_protect(m, VM_PROT_NONE);
626 vnode_pager_freepage(m);
633 vm_page_wakeup(m); /* unbusy the page */
636 *plwb = lwbuf_alloc(m);
637 *pdata = (void *)lwbuf_kva(*plwb);
643 exec_map_first_page(struct image_params *imgp)
648 exec_unmap_first_page(imgp);
650 err = exec_map_page(imgp, 0, &imgp->firstpage, &imgp->image_header);
659 exec_unmap_page(struct lwbuf *lwb)
673 exec_unmap_first_page(struct image_params *imgp)
675 exec_unmap_page(imgp->firstpage);
676 imgp->firstpage = NULL;
677 imgp->image_header = NULL;
681 * Destroy old address space, and allocate a new stack
682 * The new stack is only SGROWSIZ large because it is grown
683 * automatically in trap.c.
685 * This is the point of no return.
688 exec_new_vmspace(struct image_params *imgp, struct vmspace *vmcopy)
690 struct vmspace *vmspace = imgp->proc->p_vmspace;
691 vm_offset_t stack_addr = USRSTACK - maxssiz;
697 * Indicate that we cannot gracefully error out any more, kill
698 * any other threads present, and set P_INEXEC to indicate that
699 * we are now messing with the process structure proper.
701 * If killalllwps() races return an error which coupled with
702 * vmspace_destroyed will cause us to exit. This is what we
703 * want since another thread is patiently waiting for us to exit
707 imgp->vmspace_destroyed = 1;
709 if (curthread->td_proc->p_nthreads > 1) {
710 error = killalllwps(1);
714 imgp->vmspace_destroyed |= 2; /* we are responsible for P_INEXEC */
715 p->p_flag |= P_INEXEC;
718 * Prevent a pending AIO from modifying the new address space.
720 aio_proc_rundown(imgp->proc);
723 * Blow away entire process VM, if address space not shared,
724 * otherwise, create a new VM space so that other threads are
725 * not disrupted. If we are execing a resident vmspace we
726 * create a duplicate of it and remap the stack.
728 * The exitingcnt test is not strictly necessary but has been
729 * included for code sanity (to make the code more deterministic).
731 map = &vmspace->vm_map;
733 vmspace_exec(imgp->proc, vmcopy);
734 vmspace = imgp->proc->p_vmspace;
735 pmap_remove_pages(vmspace_pmap(vmspace), stack_addr, USRSTACK);
736 map = &vmspace->vm_map;
737 } else if (vmspace->vm_sysref.refcnt == 1 &&
738 vmspace->vm_exitingcnt == 0) {
740 if (vmspace->vm_upcalls)
741 upc_release(vmspace, ONLY_LWP_IN_PROC(imgp->proc));
742 pmap_remove_pages(vmspace_pmap(vmspace),
743 0, VM_MAX_USER_ADDRESS);
744 vm_map_remove(map, 0, VM_MAX_USER_ADDRESS);
746 vmspace_exec(imgp->proc, NULL);
747 vmspace = imgp->proc->p_vmspace;
748 map = &vmspace->vm_map;
751 /* Allocate a new stack */
752 error = vm_map_stack(&vmspace->vm_map, stack_addr, (vm_size_t)maxssiz,
753 0, VM_PROT_ALL, VM_PROT_ALL, 0);
757 /* vm_ssize and vm_maxsaddr are somewhat antiquated concepts in the
758 * VM_STACK case, but they are still used to monitor the size of the
759 * process stack so we can check the stack rlimit.
761 vmspace->vm_ssize = sgrowsiz >> PAGE_SHIFT;
762 vmspace->vm_maxsaddr = (char *)USRSTACK - maxssiz;
768 * Copy out argument and environment strings from the old process
769 * address space into the temporary string buffer.
772 exec_copyin_args(struct image_args *args, char *fname,
773 enum exec_path_segflg segflg, char **argv, char **envv)
779 args->buf = objcache_get(exec_objcache, M_WAITOK);
780 if (args->buf == NULL)
782 args->begin_argv = args->buf;
783 args->endp = args->begin_argv;
784 args->space = ARG_MAX;
786 args->fname = args->buf + ARG_MAX;
789 * Copy the file name.
791 if (segflg == PATH_SYSSPACE) {
792 error = copystr(fname, args->fname, PATH_MAX, &length);
793 } else if (segflg == PATH_USERSPACE) {
794 error = copyinstr(fname, args->fname, PATH_MAX, &length);
798 * Extract argument strings. argv may not be NULL. The argv
799 * array is terminated by a NULL entry. We special-case the
800 * situation where argv[0] is NULL by passing { filename, NULL }
801 * to the new program to guarentee that the interpreter knows what
802 * file to open in case we exec an interpreted file. Note that
803 * a NULL argv[0] terminates the argv[] array.
805 * XXX the special-casing of argv[0] is historical and needs to be
811 while ((argp = (caddr_t)(intptr_t)fuword(argv++)) != NULL) {
812 if (argp == (caddr_t)-1) {
816 error = copyinstr(argp, args->endp,
817 args->space, &length);
819 if (error == ENAMETOOLONG)
823 args->space -= length;
824 args->endp += length;
827 if (args->argc == 0 && error == 0) {
828 length = strlen(args->fname) + 1;
829 if (length > args->space) {
832 bcopy(args->fname, args->endp, length);
833 args->space -= length;
834 args->endp += length;
840 args->begin_envv = args->endp;
843 * extract environment strings. envv may be NULL.
845 if (envv && error == 0) {
846 while ((envp = (caddr_t) (intptr_t) fuword(envv++))) {
847 if (envp == (caddr_t) -1) {
851 error = copyinstr(envp, args->endp, args->space,
854 if (error == ENAMETOOLONG)
858 args->space -= length;
859 args->endp += length;
867 exec_free_args(struct image_args *args)
870 objcache_put(exec_objcache, args->buf);
876 * Copy strings out to the new process address space, constructing
877 * new arg and env vector tables. Return a pointer to the base
878 * so that it can be used as the initial stack pointer.
881 exec_copyout_strings(struct image_params *imgp)
883 int argc, envc, sgap;
885 char *stringp, *destp;
886 register_t *stack_base;
887 struct ps_strings *arginfo;
891 * Calculate string base and vector table pointers.
892 * Also deal with signal trampoline code for this exec type.
894 arginfo = (struct ps_strings *)PS_STRINGS;
895 szsigcode = *(imgp->proc->p_sysent->sv_szsigcode);
896 if (stackgap_random != 0)
897 sgap = ALIGN(karc4random() & (stackgap_random - 1));
900 destp = (caddr_t)arginfo - szsigcode - SPARE_USRSPACE - sgap -
901 roundup((ARG_MAX - imgp->args->space), sizeof(char *));
907 copyout(imgp->proc->p_sysent->sv_sigcode,
908 ((caddr_t)arginfo - szsigcode), szsigcode);
911 * If we have a valid auxargs ptr, prepare some room
914 * The '+ 2' is for the null pointers at the end of each of the
915 * arg and env vector sets, and 'AT_COUNT*2' is room for the
919 vectp = (char **)(destp - (imgp->args->argc +
920 imgp->args->envc + 2 + AT_COUNT * 2) * sizeof(char*));
922 vectp = (char **)(destp - (imgp->args->argc +
923 imgp->args->envc + 2) * sizeof(char*));
927 * NOTE: don't bother aligning the stack here for GCC 2.x, it will
928 * be done in crt1.o. Note that GCC 3.x aligns the stack in main.
932 * vectp also becomes our initial stack base
934 stack_base = (register_t *)vectp;
936 stringp = imgp->args->begin_argv;
937 argc = imgp->args->argc;
938 envc = imgp->args->envc;
941 * Copy out strings - arguments and environment.
943 copyout(stringp, destp, ARG_MAX - imgp->args->space);
946 * Fill in "ps_strings" struct for ps, w, etc.
948 suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp);
949 suword(&arginfo->ps_nargvstr, argc);
952 * Fill in argument portion of vector table.
954 for (; argc > 0; --argc) {
955 suword(vectp++, (long)(intptr_t)destp);
956 while (*stringp++ != 0)
961 /* a null vector table pointer separates the argp's from the envp's */
964 suword(&arginfo->ps_envstr, (long)(intptr_t)vectp);
965 suword(&arginfo->ps_nenvstr, envc);
968 * Fill in environment portion of vector table.
970 for (; envc > 0; --envc) {
971 suword(vectp++, (long)(intptr_t)destp);
972 while (*stringp++ != 0)
977 /* end of vector table is a null pointer */
984 * Check permissions of file to execute.
985 * Return 0 for success or error code on failure.
988 exec_check_permissions(struct image_params *imgp)
990 struct proc *p = imgp->proc;
991 struct vnode *vp = imgp->vp;
992 struct vattr *attr = imgp->attr;
995 /* Get file attributes */
996 error = VOP_GETATTR(vp, attr);
1001 * 1) Check if file execution is disabled for the filesystem that this
1003 * 2) Insure that at least one execute bit is on - otherwise root
1004 * will always succeed, and we don't want to happen unless the
1005 * file really is executable.
1006 * 3) Insure that the file is a regular file.
1008 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
1009 ((attr->va_mode & 0111) == 0) ||
1010 (attr->va_type != VREG)) {
1015 * Zero length files can't be exec'd
1017 if (attr->va_size == 0)
1021 * Check for execute permission to file based on current credentials.
1023 error = VOP_EACCESS(vp, VEXEC, p->p_ucred);
1028 * Check number of open-for-writes on the file and deny execution
1031 if (vp->v_writecount)
1035 * Call filesystem specific open routine, which allows us to read,
1036 * write, and mmap the file. Without the VOP_OPEN we can only
1039 error = VOP_OPEN(vp, FREAD, p->p_ucred, NULL);
1047 * Exec handler registration
1050 exec_register(const struct execsw *execsw_arg)
1052 const struct execsw **es, **xs, **newexecsw;
1053 int count = 2; /* New slot and trailing NULL */
1056 for (es = execsw; *es; es++)
1058 newexecsw = kmalloc(count * sizeof(*es), M_TEMP, M_WAITOK);
1061 for (es = execsw; *es; es++)
1066 kfree(execsw, M_TEMP);
1072 exec_unregister(const struct execsw *execsw_arg)
1074 const struct execsw **es, **xs, **newexecsw;
1078 panic("unregister with no handlers left?");
1080 for (es = execsw; *es; es++) {
1081 if (*es == execsw_arg)
1086 for (es = execsw; *es; es++)
1087 if (*es != execsw_arg)
1089 newexecsw = kmalloc(count * sizeof(*es), M_TEMP, M_WAITOK);
1091 for (es = execsw; *es; es++)
1092 if (*es != execsw_arg)
1096 kfree(execsw, M_TEMP);