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,
123 SYSINIT(exec_objcache, SI_BOOT2_MACHDEP, SI_ORDER_ANY, exec_objcache_init, 0);
126 * stackgap_random specifies if the stackgap should have a random size added
127 * to it. It must be a power of 2. If non-zero, the stack gap will be
128 * calculated as: ALIGN(karc4random() & (stackgap_random - 1)).
130 static int stackgap_random = 1024;
132 sysctl_kern_stackgap(SYSCTL_HANDLER_ARGS)
135 new_val = stackgap_random;
136 error = sysctl_handle_int(oidp, &new_val, 0, req);
137 if (error != 0 || req->newptr == NULL)
139 if ((new_val < 0) || (new_val > 16 * PAGE_SIZE) || ! powerof2(new_val))
141 stackgap_random = new_val;
146 SYSCTL_PROC(_kern, OID_AUTO, stackgap_random, CTLFLAG_RW|CTLTYPE_UINT,
147 0, 0, sysctl_kern_stackgap, "IU", "Max random stack gap (power of 2)");
150 print_execve_args(struct image_args *args)
155 cp = args->begin_argv;
156 for (ndx = 0; ndx < args->argc; ndx++) {
157 kprintf("\targv[%d]: %s\n", ndx, cp);
158 while (*cp++ != '\0');
160 for (ndx = 0; ndx < args->envc; ndx++) {
161 kprintf("\tenvv[%d]: %s\n", ndx, cp);
162 while (*cp++ != '\0');
167 * Each of the items is a pointer to a `const struct execsw', hence the
168 * double pointer here.
170 static const struct execsw **execsw;
173 * Replace current vmspace with a new binary.
174 * Returns 0 on success, > 0 on recoverable error (use as errno).
175 * Returns -1 on lethal error which demands killing of the current
179 kern_execve(struct nlookupdata *nd, struct image_args *args)
181 struct thread *td = curthread;
182 struct lwp *lp = td->td_lwp;
183 struct proc *p = td->td_proc;
184 register_t *stack_base;
189 struct image_params image_params, *imgp;
191 int (*img_first) (struct image_params *);
193 if (debug_execve_args) {
194 kprintf("%s()\n", __func__);
195 print_execve_args(args);
199 lwkt_gettoken(&p->p_token);
200 imgp = &image_params;
203 * NOTE: P_INEXEC is handled by exec_new_vmspace() now. We make
204 * no modifications to the process at all until we get there.
206 * Note that multiple threads may be trying to exec at the same
207 * time. exec_new_vmspace() handles that too.
211 * Initialize part of the common data
216 imgp->entry_addr = 0;
218 imgp->vmspace_destroyed = 0;
219 imgp->interpreted = 0;
220 imgp->interpreter_name[0] = 0;
221 imgp->auxargs = NULL;
223 imgp->firstpage = NULL;
224 imgp->ps_strings = 0;
225 imgp->image_header = NULL;
230 * Translate the file name to a vnode. Unlock the cache entry to
231 * improve parallelism for programs exec'd in parallel.
233 if ((error = nlookup(nd)) != 0)
235 error = cache_vget(&nd->nl_nch, nd->nl_cred, LK_EXCLUSIVE, &imgp->vp);
236 KKASSERT(nd->nl_flags & NLC_NCPISLOCKED);
237 nd->nl_flags &= ~NLC_NCPISLOCKED;
238 cache_unlock(&nd->nl_nch);
243 * Check file permissions (also 'opens' file).
244 * Include also the top level mount in the check.
246 error = exec_check_permissions(imgp, nd->nl_nch.mount);
249 goto exec_fail_dealloc;
252 error = exec_map_first_page(imgp);
255 goto exec_fail_dealloc;
257 if (debug_execve_args && imgp->interpreted) {
258 kprintf(" target is interpreted -- recursive pass\n");
259 kprintf(" interpreter: %s\n", imgp->interpreter_name);
260 print_execve_args(args);
264 * If the current process has a special image activator it
265 * wants to try first, call it. For example, emulating shell
266 * scripts differently.
269 if ((img_first = imgp->proc->p_sysent->sv_imgact_try) != NULL)
270 error = img_first(imgp);
273 * If the vnode has a registered vmspace, exec the vmspace
275 if (error == -1 && imgp->vp->v_resident) {
276 error = exec_resident_imgact(imgp);
280 * Loop through the list of image activators, calling each one.
281 * An activator returns -1 if there is no match, 0 on success,
282 * and an error otherwise.
284 for (i = 0; error == -1 && execsw[i]; ++i) {
285 if (execsw[i]->ex_imgact == NULL ||
286 execsw[i]->ex_imgact == img_first) {
289 error = (*execsw[i]->ex_imgact)(imgp);
295 goto exec_fail_dealloc;
299 * Special interpreter operation, cleanup and loop up to try to
300 * activate the interpreter.
302 if (imgp->interpreted) {
303 exec_unmap_first_page(imgp);
307 error = nlookup_init(nd, imgp->interpreter_name, UIO_SYSSPACE,
315 * Copy out strings (args and env) and initialize stack base
317 stack_base = exec_copyout_strings(imgp);
318 p->p_vmspace->vm_minsaddr = (char *)stack_base;
321 * If custom stack fixup routine present for this process
322 * let it do the stack setup. If we are running a resident
323 * image there is no auxinfo or other image activator context
324 * so don't try to add fixups to the stack.
326 * Else stuff argument count as first item on stack
328 if (p->p_sysent->sv_fixup && imgp->resident == 0)
329 (*p->p_sysent->sv_fixup)(&stack_base, imgp);
331 suword(--stack_base, imgp->args->argc);
334 * For security and other reasons, the file descriptor table cannot
335 * be shared after an exec.
337 if (p->p_fd->fd_refcnt > 1) {
338 struct filedesc *tmp;
340 error = fdcopy(p, &tmp);
347 * For security and other reasons, signal handlers cannot
348 * be shared after an exec. The new proces gets a copy of the old
349 * handlers. In execsigs(), the new process will have its signals
353 if (ops->ps_refcnt > 1) {
354 nps = kmalloc(sizeof(*nps), M_SUBPROC, M_WAITOK);
355 bcopy(ops, nps, sizeof(*nps));
356 refcount_init(&nps->ps_refcnt, 1);
358 if (refcount_release(&ops->ps_refcnt)) {
359 kfree(ops, M_SUBPROC);
365 * For security and other reasons virtual kernels cannot be
366 * inherited by an exec. This also allows a virtual kernel
367 * to fork/exec unrelated applications.
375 /* close files on exec */
378 /* reset caught signals */
381 /* name this process - nameiexec(p, ndp) */
382 len = min(nd->nl_nch.ncp->nc_nlen, MAXCOMLEN);
383 bcopy(nd->nl_nch.ncp->nc_name, p->p_comm, len);
385 bcopy(p->p_comm, lp->lwp_thread->td_comm, MAXCOMLEN+1);
388 * mark as execed, wakeup the process that vforked (if any) and tell
389 * it that it now has its own resources back
392 if (p->p_pptr && (p->p_flag & P_PPWAIT)) {
393 p->p_flag &= ~P_PPWAIT;
394 wakeup((caddr_t)p->p_pptr);
398 * Implement image setuid/setgid.
400 * Don't honor setuid/setgid if the filesystem prohibits it or if
401 * the process is being traced.
403 if ((((attr.va_mode & VSUID) && p->p_ucred->cr_uid != attr.va_uid) ||
404 ((attr.va_mode & VSGID) && p->p_ucred->cr_gid != attr.va_gid)) &&
405 (imgp->vp->v_mount->mnt_flag & MNT_NOSUID) == 0 &&
406 (p->p_flag & P_TRACED) == 0) {
408 * Turn off syscall tracing for set-id programs, except for
409 * root. Record any set-id flags first to make sure that
410 * we do not regain any tracing during a possible block.
413 if (p->p_tracenode && ktrace_suid == 0 &&
414 priv_check(td, PRIV_ROOT) != 0) {
415 ktrdestroy(&p->p_tracenode);
418 /* Close any file descriptors 0..2 that reference procfs */
420 /* Make sure file descriptors 0..2 are in use. */
421 error = fdcheckstd(lp);
423 goto exec_fail_dealloc;
425 * Set the new credentials.
428 if (attr.va_mode & VSUID)
429 change_euid(attr.va_uid);
430 if (attr.va_mode & VSGID)
431 p->p_ucred->cr_gid = attr.va_gid;
434 * Clear local varsym variables
436 varsymset_clean(&p->p_varsymset);
438 if (p->p_ucred->cr_uid == p->p_ucred->cr_ruid &&
439 p->p_ucred->cr_gid == p->p_ucred->cr_rgid)
440 p->p_flag &= ~P_SUGID;
444 * Implement correct POSIX saved-id behavior.
446 if (p->p_ucred->cr_svuid != p->p_ucred->cr_uid ||
447 p->p_ucred->cr_svgid != p->p_ucred->cr_gid) {
449 p->p_ucred->cr_svuid = p->p_ucred->cr_uid;
450 p->p_ucred->cr_svgid = p->p_ucred->cr_gid;
454 * Store the vp for use in procfs
456 if (p->p_textvp) /* release old reference */
458 p->p_textvp = imgp->vp;
461 /* Release old namecache handle to text file */
462 if (p->p_textnch.ncp)
463 cache_drop(&p->p_textnch);
465 if (nd->nl_nch.mount)
466 cache_copy(&nd->nl_nch, &p->p_textnch);
469 * Notify others that we exec'd, and clear the P_INEXEC flag
470 * as we're now a bona fide freshly-execed process.
472 KNOTE(&p->p_klist, NOTE_EXEC);
473 p->p_flag &= ~P_INEXEC;
476 * If tracing the process, trap to debugger so breakpoints
477 * can be set before the program executes.
479 STOPEVENT(p, S_EXEC, 0);
481 if (p->p_flag & P_TRACED)
484 /* clear "fork but no exec" flag, as we _are_ execing */
485 p->p_acflag &= ~AFORK;
487 /* Set values passed into the program in registers. */
488 exec_setregs(imgp->entry_addr, (u_long)(uintptr_t)stack_base,
491 /* Set the access time on the vnode */
492 vn_mark_atime(imgp->vp, td);
495 * Free any previous argument cache
499 if (pa && refcount_release(&pa->ar_ref)) {
505 * Cache arguments if they fit inside our allowance
507 i = imgp->args->begin_envv - imgp->args->begin_argv;
508 if (sizeof(struct pargs) + i <= ps_arg_cache_limit) {
509 pa = kmalloc(sizeof(struct pargs) + i, M_PARGS, M_WAITOK);
510 refcount_init(&pa->ar_ref, 1);
512 bcopy(imgp->args->begin_argv, pa->ar_args, i);
513 KKASSERT(p->p_args == NULL);
520 * free various allocated resources
523 exec_unmap_first_page(imgp);
531 ++mycpu->gd_cnt.v_exec;
532 lwkt_reltoken(&p->p_token);
538 * we're done here, clear P_INEXEC if we were the ones that
539 * set it. Otherwise if vmspace_destroyed is still set we
540 * raced another thread and that thread is responsible for
543 if (imgp->vmspace_destroyed & 2)
544 p->p_flag &= ~P_INEXEC;
545 lwkt_reltoken(&p->p_token);
546 if (imgp->vmspace_destroyed) {
548 * Sorry, no more process anymore. exit gracefully.
549 * However we can't die right here, because our
550 * caller might have to clean up, so indicate a
551 * lethal error by returning -1.
560 * execve() system call.
565 sys_execve(struct execve_args *uap)
567 struct nlookupdata nd;
568 struct image_args args;
571 bzero(&args, sizeof(args));
574 error = nlookup_init(&nd, uap->fname, UIO_USERSPACE, NLC_FOLLOW);
576 error = exec_copyin_args(&args, uap->fname, PATH_USERSPACE,
577 uap->argv, uap->envv);
580 error = kern_execve(&nd, &args);
582 exec_free_args(&args);
585 /* We hit a lethal error condition. Let's die now. */
586 exit1(W_EXITCODE(0, SIGABRT));
592 * The syscall result is returned in registers to the new program.
593 * Linux will register %edx as an atexit function and we must be
594 * sure to set it to 0. XXX
597 uap->sysmsg_result64 = 0;
603 exec_map_page(struct image_params *imgp, vm_pindex_t pageno,
604 struct lwbuf **plwb, const char **pdata)
612 * The file has to be mappable.
614 if ((object = imgp->vp->v_object) == NULL)
617 if (pageno >= object->size)
620 m = vm_page_grab(object, pageno, VM_ALLOC_NORMAL | VM_ALLOC_RETRY);
622 lwkt_gettoken(&vm_token);
623 while ((m->valid & VM_PAGE_BITS_ALL) != VM_PAGE_BITS_ALL) {
627 * get_pages unbusies all the requested pages except the
628 * primary page (at index 0 in this case). The primary
629 * page may have been wired during the pagein (e.g. by
630 * the buffer cache) so vnode_pager_freepage() must be
631 * used to properly release it.
633 rv = vm_pager_get_page(object, &ma, 1);
634 m = vm_page_lookup(object, pageno);
636 if (rv != VM_PAGER_OK || m == NULL || m->valid == 0) {
638 vm_page_protect(m, VM_PROT_NONE);
639 vnode_pager_freepage(m);
641 lwkt_reltoken(&vm_token);
645 vm_page_hold(m); /* requires vm_token to be held */
646 vm_page_wakeup(m); /* unbusy the page */
647 lwkt_reltoken(&vm_token);
649 *plwb = lwbuf_alloc(m, *plwb);
650 *pdata = (void *)lwbuf_kva(*plwb);
656 exec_map_first_page(struct image_params *imgp)
661 exec_unmap_first_page(imgp);
663 imgp->firstpage = &imgp->firstpage_cache;
664 err = exec_map_page(imgp, 0, &imgp->firstpage, &imgp->image_header);
673 exec_unmap_page(struct lwbuf *lwb)
687 exec_unmap_first_page(struct image_params *imgp)
689 exec_unmap_page(imgp->firstpage);
690 imgp->firstpage = NULL;
691 imgp->image_header = NULL;
695 * Destroy old address space, and allocate a new stack
696 * The new stack is only SGROWSIZ large because it is grown
697 * automatically in trap.c.
699 * This is the point of no return.
702 exec_new_vmspace(struct image_params *imgp, struct vmspace *vmcopy)
704 struct vmspace *vmspace = imgp->proc->p_vmspace;
705 vm_offset_t stack_addr = USRSTACK - maxssiz;
711 * Indicate that we cannot gracefully error out any more, kill
712 * any other threads present, and set P_INEXEC to indicate that
713 * we are now messing with the process structure proper.
715 * If killalllwps() races return an error which coupled with
716 * vmspace_destroyed will cause us to exit. This is what we
717 * want since another thread is patiently waiting for us to exit
721 imgp->vmspace_destroyed = 1;
723 if (curthread->td_proc->p_nthreads > 1) {
724 error = killalllwps(1);
728 imgp->vmspace_destroyed |= 2; /* we are responsible for P_INEXEC */
729 p->p_flag |= P_INEXEC;
732 * Blow away entire process VM, if address space not shared,
733 * otherwise, create a new VM space so that other threads are
734 * not disrupted. If we are execing a resident vmspace we
735 * create a duplicate of it and remap the stack.
737 * The exitingcnt test is not strictly necessary but has been
738 * included for code sanity (to make the code more deterministic).
740 map = &vmspace->vm_map;
742 vmspace_exec(imgp->proc, vmcopy);
743 vmspace = imgp->proc->p_vmspace;
744 pmap_remove_pages(vmspace_pmap(vmspace), stack_addr, USRSTACK);
745 map = &vmspace->vm_map;
746 } else if (vmspace->vm_sysref.refcnt == 1 &&
747 vmspace->vm_exitingcnt == 0) {
749 if (vmspace->vm_upcalls)
750 upc_release(vmspace, ONLY_LWP_IN_PROC(imgp->proc));
751 pmap_remove_pages(vmspace_pmap(vmspace),
752 0, VM_MAX_USER_ADDRESS);
753 vm_map_remove(map, 0, VM_MAX_USER_ADDRESS);
755 vmspace_exec(imgp->proc, NULL);
756 vmspace = imgp->proc->p_vmspace;
757 map = &vmspace->vm_map;
760 /* Allocate a new stack */
761 error = vm_map_stack(&vmspace->vm_map, stack_addr, (vm_size_t)maxssiz,
762 0, VM_PROT_ALL, VM_PROT_ALL, 0);
766 /* vm_ssize and vm_maxsaddr are somewhat antiquated concepts in the
767 * VM_STACK case, but they are still used to monitor the size of the
768 * process stack so we can check the stack rlimit.
770 vmspace->vm_ssize = sgrowsiz >> PAGE_SHIFT;
771 vmspace->vm_maxsaddr = (char *)USRSTACK - maxssiz;
777 * Copy out argument and environment strings from the old process
778 * address space into the temporary string buffer.
781 exec_copyin_args(struct image_args *args, char *fname,
782 enum exec_path_segflg segflg, char **argv, char **envv)
788 args->buf = objcache_get(exec_objcache, M_WAITOK);
789 if (args->buf == NULL)
791 args->begin_argv = args->buf;
792 args->endp = args->begin_argv;
793 args->space = ARG_MAX;
795 args->fname = args->buf + ARG_MAX;
798 * Copy the file name.
800 if (segflg == PATH_SYSSPACE) {
801 error = copystr(fname, args->fname, PATH_MAX, &length);
802 } else if (segflg == PATH_USERSPACE) {
803 error = copyinstr(fname, args->fname, PATH_MAX, &length);
807 * Extract argument strings. argv may not be NULL. The argv
808 * array is terminated by a NULL entry. We special-case the
809 * situation where argv[0] is NULL by passing { filename, NULL }
810 * to the new program to guarentee that the interpreter knows what
811 * file to open in case we exec an interpreted file. Note that
812 * a NULL argv[0] terminates the argv[] array.
814 * XXX the special-casing of argv[0] is historical and needs to be
820 while ((argp = (caddr_t)(intptr_t)fuword(argv++)) != NULL) {
821 if (argp == (caddr_t)-1) {
825 error = copyinstr(argp, args->endp,
826 args->space, &length);
828 if (error == ENAMETOOLONG)
832 args->space -= length;
833 args->endp += length;
836 if (args->argc == 0 && error == 0) {
837 length = strlen(args->fname) + 1;
838 if (length > args->space) {
841 bcopy(args->fname, args->endp, length);
842 args->space -= length;
843 args->endp += length;
849 args->begin_envv = args->endp;
852 * extract environment strings. envv may be NULL.
854 if (envv && error == 0) {
855 while ((envp = (caddr_t) (intptr_t) fuword(envv++))) {
856 if (envp == (caddr_t) -1) {
860 error = copyinstr(envp, args->endp, args->space,
863 if (error == ENAMETOOLONG)
867 args->space -= length;
868 args->endp += length;
876 exec_free_args(struct image_args *args)
879 objcache_put(exec_objcache, args->buf);
885 * Copy strings out to the new process address space, constructing
886 * new arg and env vector tables. Return a pointer to the base
887 * so that it can be used as the initial stack pointer.
890 exec_copyout_strings(struct image_params *imgp)
892 int argc, envc, sgap;
894 char *stringp, *destp;
895 register_t *stack_base;
896 struct ps_strings *arginfo;
900 * Calculate string base and vector table pointers.
901 * Also deal with signal trampoline code for this exec type.
903 arginfo = (struct ps_strings *)PS_STRINGS;
904 szsigcode = *(imgp->proc->p_sysent->sv_szsigcode);
905 if (stackgap_random != 0)
906 sgap = ALIGN(karc4random() & (stackgap_random - 1));
909 destp = (caddr_t)arginfo - szsigcode - SPARE_USRSPACE - sgap -
910 roundup((ARG_MAX - imgp->args->space), sizeof(char *));
916 copyout(imgp->proc->p_sysent->sv_sigcode,
917 ((caddr_t)arginfo - szsigcode), szsigcode);
920 * If we have a valid auxargs ptr, prepare some room
923 * The '+ 2' is for the null pointers at the end of each of the
924 * arg and env vector sets, and 'AT_COUNT*2' is room for the
928 vectp = (char **)(destp - (imgp->args->argc +
929 imgp->args->envc + 2 + AT_COUNT * 2) * sizeof(char*));
931 vectp = (char **)(destp - (imgp->args->argc +
932 imgp->args->envc + 2) * sizeof(char*));
936 * NOTE: don't bother aligning the stack here for GCC 2.x, it will
937 * be done in crt1.o. Note that GCC 3.x aligns the stack in main.
941 * vectp also becomes our initial stack base
943 stack_base = (register_t *)vectp;
945 stringp = imgp->args->begin_argv;
946 argc = imgp->args->argc;
947 envc = imgp->args->envc;
950 * Copy out strings - arguments and environment.
952 copyout(stringp, destp, ARG_MAX - imgp->args->space);
955 * Fill in "ps_strings" struct for ps, w, etc.
957 suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp);
958 suword(&arginfo->ps_nargvstr, argc);
961 * Fill in argument portion of vector table.
963 for (; argc > 0; --argc) {
964 suword(vectp++, (long)(intptr_t)destp);
965 while (*stringp++ != 0)
970 /* a null vector table pointer separates the argp's from the envp's */
973 suword(&arginfo->ps_envstr, (long)(intptr_t)vectp);
974 suword(&arginfo->ps_nenvstr, envc);
977 * Fill in environment portion of vector table.
979 for (; envc > 0; --envc) {
980 suword(vectp++, (long)(intptr_t)destp);
981 while (*stringp++ != 0)
986 /* end of vector table is a null pointer */
993 * Check permissions of file to execute.
994 * Return 0 for success or error code on failure.
997 exec_check_permissions(struct image_params *imgp, struct mount *topmnt)
999 struct proc *p = imgp->proc;
1000 struct vnode *vp = imgp->vp;
1001 struct vattr *attr = imgp->attr;
1004 /* Get file attributes */
1005 error = VOP_GETATTR(vp, attr);
1010 * 1) Check if file execution is disabled for the filesystem that this
1012 * 2) Insure that at least one execute bit is on - otherwise root
1013 * will always succeed, and we don't want to happen unless the
1014 * file really is executable.
1015 * 3) Insure that the file is a regular file.
1017 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
1018 ((topmnt != NULL) && (topmnt->mnt_flag & MNT_NOEXEC)) ||
1019 ((attr->va_mode & 0111) == 0) ||
1020 (attr->va_type != VREG)) {
1025 * Zero length files can't be exec'd
1027 if (attr->va_size == 0)
1031 * Check for execute permission to file based on current credentials.
1033 error = VOP_EACCESS(vp, VEXEC, p->p_ucred);
1038 * Check number of open-for-writes on the file and deny execution
1041 if (vp->v_writecount)
1045 * Call filesystem specific open routine, which allows us to read,
1046 * write, and mmap the file. Without the VOP_OPEN we can only
1049 error = VOP_OPEN(vp, FREAD, p->p_ucred, NULL);
1057 * Exec handler registration
1060 exec_register(const struct execsw *execsw_arg)
1062 const struct execsw **es, **xs, **newexecsw;
1063 int count = 2; /* New slot and trailing NULL */
1066 for (es = execsw; *es; es++)
1068 newexecsw = kmalloc(count * sizeof(*es), M_TEMP, M_WAITOK);
1071 for (es = execsw; *es; es++)
1076 kfree(execsw, M_TEMP);
1082 exec_unregister(const struct execsw *execsw_arg)
1084 const struct execsw **es, **xs, **newexecsw;
1088 panic("unregister with no handlers left?");
1090 for (es = execsw; *es; es++) {
1091 if (*es == execsw_arg)
1096 for (es = execsw; *es; es++)
1097 if (*es != execsw_arg)
1099 newexecsw = kmalloc(count * sizeof(*es), M_TEMP, M_WAITOK);
1101 for (es = execsw; *es; es++)
1102 if (*es != execsw_arg)
1106 kfree(execsw, M_TEMP);