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/sfbuf.h>
51 #include <sys/sysent.h>
53 #include <sys/sysctl.h>
54 #include <sys/vnode.h>
55 #include <sys/vmmeter.h>
57 #include <sys/libkern.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/thread2.h>
75 #include <sys/mplock2.h>
77 MALLOC_DEFINE(M_PARGS, "proc-args", "Process arguments");
78 MALLOC_DEFINE(M_EXECARGS, "exec-args", "Exec arguments");
80 static register_t *exec_copyout_strings (struct image_params *);
82 /* XXX This should be vm_size_t. */
83 static u_long ps_strings = PS_STRINGS;
84 SYSCTL_ULONG(_kern, KERN_PS_STRINGS, ps_strings, CTLFLAG_RD, &ps_strings, 0, "");
86 /* XXX This should be vm_size_t. */
87 static u_long usrstack = USRSTACK;
88 SYSCTL_ULONG(_kern, KERN_USRSTACK, usrstack, CTLFLAG_RD, &usrstack, 0, "");
90 u_long ps_arg_cache_limit = PAGE_SIZE / 16;
91 SYSCTL_LONG(_kern, OID_AUTO, ps_arg_cache_limit, CTLFLAG_RW,
92 &ps_arg_cache_limit, 0, "");
95 SYSCTL_INT(_kern, OID_AUTO, ps_argsopen, CTLFLAG_RW, &ps_argsopen, 0, "");
97 static int ktrace_suid = 0;
98 SYSCTL_INT(_kern, OID_AUTO, ktrace_suid, CTLFLAG_RW, &ktrace_suid, 0, "");
100 void print_execve_args(struct image_args *args);
101 int debug_execve_args = 0;
102 SYSCTL_INT(_kern, OID_AUTO, debug_execve_args, CTLFLAG_RW, &debug_execve_args,
106 * Exec arguments object cache
108 static struct objcache *exec_objcache;
112 exec_objcache_init(void *arg __unused)
116 cluster_limit = 16; /* up to this many objects */
117 exec_objcache = objcache_create_mbacked(
118 M_EXECARGS, PATH_MAX + ARG_MAX,
120 2, /* minimal magazine capacity */
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;
186 struct image_params image_params, *imgp;
188 int (*img_first) (struct image_params *);
190 if (debug_execve_args) {
191 kprintf("%s()\n", __func__);
192 print_execve_args(args);
196 imgp = &image_params;
199 * NOTE: P_INEXEC is handled by exec_new_vmspace() now. We make
200 * no modifications to the process at all until we get there.
202 * Note that multiple threads may be trying to exec at the same
203 * time. exec_new_vmspace() handles that too.
207 * Initialize part of the common data
212 imgp->entry_addr = 0;
214 imgp->vmspace_destroyed = 0;
215 imgp->interpreted = 0;
216 imgp->interpreter_name[0] = 0;
217 imgp->auxargs = NULL;
219 imgp->firstpage = NULL;
220 imgp->ps_strings = 0;
221 imgp->image_header = NULL;
226 * Translate the file name to a vnode. Unlock the cache entry to
227 * improve parallelism for programs exec'd in parallel.
229 if ((error = nlookup(nd)) != 0)
231 error = cache_vget(&nd->nl_nch, nd->nl_cred, LK_EXCLUSIVE, &imgp->vp);
232 KKASSERT(nd->nl_flags & NLC_NCPISLOCKED);
233 nd->nl_flags &= ~NLC_NCPISLOCKED;
234 cache_unlock(&nd->nl_nch);
239 * Check file permissions (also 'opens' file)
241 error = exec_check_permissions(imgp);
244 goto exec_fail_dealloc;
247 error = exec_map_first_page(imgp);
250 goto exec_fail_dealloc;
252 if (debug_execve_args && imgp->interpreted) {
253 kprintf(" target is interpreted -- recursive pass\n");
254 kprintf(" interpreter: %s\n", imgp->interpreter_name);
255 print_execve_args(args);
259 * If the current process has a special image activator it
260 * wants to try first, call it. For example, emulating shell
261 * scripts differently.
264 if ((img_first = imgp->proc->p_sysent->sv_imgact_try) != NULL)
265 error = img_first(imgp);
268 * If the vnode has a registered vmspace, exec the vmspace
270 if (error == -1 && imgp->vp->v_resident) {
271 error = exec_resident_imgact(imgp);
275 * Loop through the list of image activators, calling each one.
276 * An activator returns -1 if there is no match, 0 on success,
277 * and an error otherwise.
279 for (i = 0; error == -1 && execsw[i]; ++i) {
280 if (execsw[i]->ex_imgact == NULL ||
281 execsw[i]->ex_imgact == img_first) {
284 error = (*execsw[i]->ex_imgact)(imgp);
290 goto exec_fail_dealloc;
294 * Special interpreter operation, cleanup and loop up to try to
295 * activate the interpreter.
297 if (imgp->interpreted) {
298 exec_unmap_first_page(imgp);
302 error = nlookup_init(nd, imgp->interpreter_name, UIO_SYSSPACE,
310 * Copy out strings (args and env) and initialize stack base
312 stack_base = exec_copyout_strings(imgp);
313 p->p_vmspace->vm_minsaddr = (char *)stack_base;
316 * If custom stack fixup routine present for this process
317 * let it do the stack setup. If we are running a resident
318 * image there is no auxinfo or other image activator context
319 * so don't try to add fixups to the stack.
321 * Else stuff argument count as first item on stack
323 if (p->p_sysent->sv_fixup && imgp->resident == 0)
324 (*p->p_sysent->sv_fixup)(&stack_base, imgp);
326 suword(--stack_base, imgp->args->argc);
329 * For security and other reasons, the file descriptor table cannot
330 * be shared after an exec.
332 if (p->p_fd->fd_refcnt > 1) {
333 struct filedesc *tmp;
340 * For security and other reasons, signal handlers cannot
341 * be shared after an exec. The new proces gets a copy of the old
342 * handlers. In execsigs(), the new process will have its signals
345 if (p->p_sigacts->ps_refcnt > 1) {
346 struct sigacts *newsigacts;
348 newsigacts = (struct sigacts *)kmalloc(sizeof(*newsigacts),
349 M_SUBPROC, M_WAITOK);
350 bcopy(p->p_sigacts, newsigacts, sizeof(*newsigacts));
351 p->p_sigacts->ps_refcnt--;
352 p->p_sigacts = newsigacts;
353 p->p_sigacts->ps_refcnt = 1;
357 * For security and other reasons virtual kernels cannot be
358 * inherited by an exec. This also allows a virtual kernel
359 * to fork/exec unrelated applications.
367 /* close files on exec */
370 /* reset caught signals */
373 /* name this process - nameiexec(p, ndp) */
374 len = min(nd->nl_nch.ncp->nc_nlen, MAXCOMLEN);
375 bcopy(nd->nl_nch.ncp->nc_name, p->p_comm, len);
377 bcopy(p->p_comm, lp->lwp_thread->td_comm, MAXCOMLEN+1);
380 * mark as execed, wakeup the process that vforked (if any) and tell
381 * it that it now has its own resources back
384 if (p->p_pptr && (p->p_flag & P_PPWAIT)) {
385 p->p_flag &= ~P_PPWAIT;
386 wakeup((caddr_t)p->p_pptr);
390 * Implement image setuid/setgid.
392 * Don't honor setuid/setgid if the filesystem prohibits it or if
393 * the process is being traced.
395 if ((((attr.va_mode & VSUID) && p->p_ucred->cr_uid != attr.va_uid) ||
396 ((attr.va_mode & VSGID) && p->p_ucred->cr_gid != attr.va_gid)) &&
397 (imgp->vp->v_mount->mnt_flag & MNT_NOSUID) == 0 &&
398 (p->p_flag & P_TRACED) == 0) {
400 * Turn off syscall tracing for set-id programs, except for
401 * root. Record any set-id flags first to make sure that
402 * we do not regain any tracing during a possible block.
405 if (p->p_tracenode && ktrace_suid == 0 &&
406 priv_check(td, PRIV_ROOT) != 0) {
407 ktrdestroy(&p->p_tracenode);
410 /* Close any file descriptors 0..2 that reference procfs */
412 /* Make sure file descriptors 0..2 are in use. */
413 error = fdcheckstd(lp);
415 goto exec_fail_dealloc;
417 * Set the new credentials.
420 if (attr.va_mode & VSUID)
421 change_euid(attr.va_uid);
422 if (attr.va_mode & VSGID)
423 p->p_ucred->cr_gid = attr.va_gid;
426 * Clear local varsym variables
428 varsymset_clean(&p->p_varsymset);
430 if (p->p_ucred->cr_uid == p->p_ucred->cr_ruid &&
431 p->p_ucred->cr_gid == p->p_ucred->cr_rgid)
432 p->p_flag &= ~P_SUGID;
436 * Implement correct POSIX saved-id behavior.
438 if (p->p_ucred->cr_svuid != p->p_ucred->cr_uid ||
439 p->p_ucred->cr_svgid != p->p_ucred->cr_gid) {
441 p->p_ucred->cr_svuid = p->p_ucred->cr_uid;
442 p->p_ucred->cr_svgid = p->p_ucred->cr_gid;
446 * Store the vp for use in procfs
448 if (p->p_textvp) /* release old reference */
450 p->p_textvp = imgp->vp;
454 * Notify others that we exec'd, and clear the P_INEXEC flag
455 * as we're now a bona fide freshly-execed process.
457 KNOTE(&p->p_klist, NOTE_EXEC);
458 p->p_flag &= ~P_INEXEC;
461 * If tracing the process, trap to debugger so breakpoints
462 * can be set before the program executes.
464 STOPEVENT(p, S_EXEC, 0);
466 if (p->p_flag & P_TRACED)
469 /* clear "fork but no exec" flag, as we _are_ execing */
470 p->p_acflag &= ~AFORK;
472 /* Set values passed into the program in registers. */
473 exec_setregs(imgp->entry_addr, (u_long)(uintptr_t)stack_base,
476 /* Set the access time on the vnode */
477 vn_mark_atime(imgp->vp, td);
479 /* Free any previous argument cache */
480 if (p->p_args && --p->p_args->ar_ref == 0)
481 FREE(p->p_args, M_PARGS);
484 /* Cache arguments if they fit inside our allowance */
485 i = imgp->args->begin_envv - imgp->args->begin_argv;
486 if (ps_arg_cache_limit >= i + sizeof(struct pargs)) {
487 MALLOC(p->p_args, struct pargs *, sizeof(struct pargs) + i,
489 p->p_args->ar_ref = 1;
490 p->p_args->ar_length = i;
491 bcopy(imgp->args->begin_argv, p->p_args->ar_args, i);
497 * free various allocated resources
500 exec_unmap_first_page(imgp);
508 ++mycpu->gd_cnt.v_exec;
514 * we're done here, clear P_INEXEC if we were the ones that
515 * set it. Otherwise if vmspace_destroyed is still set we
516 * raced another thread and that thread is responsible for
519 if (imgp->vmspace_destroyed & 2)
520 p->p_flag &= ~P_INEXEC;
521 if (imgp->vmspace_destroyed) {
523 * Sorry, no more process anymore. exit gracefully.
524 * However we can't die right here, because our
525 * caller might have to clean up, so indicate a
526 * lethal error by returning -1.
535 * execve() system call.
540 sys_execve(struct execve_args *uap)
542 struct nlookupdata nd;
543 struct image_args args;
546 bzero(&args, sizeof(args));
549 error = nlookup_init(&nd, uap->fname, UIO_USERSPACE, NLC_FOLLOW);
551 error = exec_copyin_args(&args, uap->fname, PATH_USERSPACE,
552 uap->argv, uap->envv);
555 error = kern_execve(&nd, &args);
557 exec_free_args(&args);
560 /* We hit a lethal error condition. Let's die now. */
561 exit1(W_EXITCODE(0, SIGABRT));
567 * The syscall result is returned in registers to the new program.
568 * Linux will register %edx as an atexit function and we must be
569 * sure to set it to 0. XXX
572 uap->sysmsg_result64 = 0;
578 exec_map_page(struct image_params *imgp, vm_pindex_t pageno,
579 struct sf_buf **psfb, const char **pdata)
587 * The file has to be mappable.
589 if ((object = imgp->vp->v_object) == NULL)
592 if (pageno >= object->size)
596 * We shouldn't need protection for vm_page_grab() but we certainly
597 * need it for the lookup loop below (lookup/busy race), since
598 * an interrupt can unbusy and free the page before our busy check.
600 m = vm_page_grab(object, pageno, VM_ALLOC_NORMAL | VM_ALLOC_RETRY);
602 while ((m->valid & VM_PAGE_BITS_ALL) != VM_PAGE_BITS_ALL) {
606 * get_pages unbusies all the requested pages except the
607 * primary page (at index 0 in this case). The primary
608 * page may have been wired during the pagein (e.g. by
609 * the buffer cache) so vnode_pager_freepage() must be
610 * used to properly release it.
612 rv = vm_pager_get_page(object, &ma, 1);
613 m = vm_page_lookup(object, pageno);
615 if (rv != VM_PAGER_OK || m == NULL || m->valid == 0) {
617 vm_page_protect(m, VM_PROT_NONE);
618 vnode_pager_freepage(m);
625 vm_page_wakeup(m); /* unbusy the page */
628 *psfb = sf_buf_alloc(m, SFB_CPUPRIVATE);
629 *pdata = (void *)sf_buf_kva(*psfb);
635 exec_map_first_page(struct image_params *imgp)
640 exec_unmap_first_page(imgp);
642 err = exec_map_page(imgp, 0, &imgp->firstpage, &imgp->image_header);
651 exec_unmap_page(struct sf_buf *sfb)
657 m = sf_buf_page(sfb);
665 exec_unmap_first_page(struct image_params *imgp)
667 exec_unmap_page(imgp->firstpage);
668 imgp->firstpage = NULL;
669 imgp->image_header = NULL;
673 * Destroy old address space, and allocate a new stack
674 * The new stack is only SGROWSIZ large because it is grown
675 * automatically in trap.c.
677 * This is the point of no return.
680 exec_new_vmspace(struct image_params *imgp, struct vmspace *vmcopy)
682 struct vmspace *vmspace = imgp->proc->p_vmspace;
683 vm_offset_t stack_addr = USRSTACK - maxssiz;
689 * Indicate that we cannot gracefully error out any more, kill
690 * any other threads present, and set P_INEXEC to indicate that
691 * we are now messing with the process structure proper.
693 * If killalllwps() races return an error which coupled with
694 * vmspace_destroyed will cause us to exit. This is what we
695 * want since another thread is patiently waiting for us to exit
699 imgp->vmspace_destroyed = 1;
701 if (curthread->td_proc->p_nthreads > 1) {
702 error = killalllwps(1);
706 imgp->vmspace_destroyed |= 2; /* we are responsible for P_INEXEC */
707 p->p_flag |= P_INEXEC;
710 * Prevent a pending AIO from modifying the new address space.
712 aio_proc_rundown(imgp->proc);
715 * Blow away entire process VM, if address space not shared,
716 * otherwise, create a new VM space so that other threads are
717 * not disrupted. If we are execing a resident vmspace we
718 * create a duplicate of it and remap the stack.
720 * The exitingcnt test is not strictly necessary but has been
721 * included for code sanity (to make the code more deterministic).
723 map = &vmspace->vm_map;
725 vmspace_exec(imgp->proc, vmcopy);
726 vmspace = imgp->proc->p_vmspace;
727 pmap_remove_pages(vmspace_pmap(vmspace), stack_addr, USRSTACK);
728 map = &vmspace->vm_map;
729 } else if (vmspace->vm_sysref.refcnt == 1 &&
730 vmspace->vm_exitingcnt == 0) {
732 if (vmspace->vm_upcalls)
733 upc_release(vmspace, ONLY_LWP_IN_PROC(imgp->proc));
734 pmap_remove_pages(vmspace_pmap(vmspace),
735 0, VM_MAX_USER_ADDRESS);
736 vm_map_remove(map, 0, VM_MAX_USER_ADDRESS);
738 vmspace_exec(imgp->proc, NULL);
739 vmspace = imgp->proc->p_vmspace;
740 map = &vmspace->vm_map;
743 /* Allocate a new stack */
744 error = vm_map_stack(&vmspace->vm_map, stack_addr, (vm_size_t)maxssiz,
745 0, VM_PROT_ALL, VM_PROT_ALL, 0);
749 /* vm_ssize and vm_maxsaddr are somewhat antiquated concepts in the
750 * VM_STACK case, but they are still used to monitor the size of the
751 * process stack so we can check the stack rlimit.
753 vmspace->vm_ssize = sgrowsiz >> PAGE_SHIFT;
754 vmspace->vm_maxsaddr = (char *)USRSTACK - maxssiz;
760 * Copy out argument and environment strings from the old process
761 * address space into the temporary string buffer.
764 exec_copyin_args(struct image_args *args, char *fname,
765 enum exec_path_segflg segflg, char **argv, char **envv)
771 args->buf = objcache_get(exec_objcache, M_WAITOK);
772 if (args->buf == NULL)
774 args->begin_argv = args->buf;
775 args->endp = args->begin_argv;
776 args->space = ARG_MAX;
778 args->fname = args->buf + ARG_MAX;
781 * Copy the file name.
783 if (segflg == PATH_SYSSPACE) {
784 error = copystr(fname, args->fname, PATH_MAX, &length);
785 } else if (segflg == PATH_USERSPACE) {
786 error = copyinstr(fname, args->fname, PATH_MAX, &length);
790 * Extract argument strings. argv may not be NULL. The argv
791 * array is terminated by a NULL entry. We special-case the
792 * situation where argv[0] is NULL by passing { filename, NULL }
793 * to the new program to guarentee that the interpreter knows what
794 * file to open in case we exec an interpreted file. Note that
795 * a NULL argv[0] terminates the argv[] array.
797 * XXX the special-casing of argv[0] is historical and needs to be
803 while ((argp = (caddr_t)(intptr_t)fuword(argv++)) != NULL) {
804 if (argp == (caddr_t)-1) {
808 error = copyinstr(argp, args->endp,
809 args->space, &length);
811 if (error == ENAMETOOLONG)
815 args->space -= length;
816 args->endp += length;
819 if (args->argc == 0 && error == 0) {
820 length = strlen(args->fname) + 1;
821 if (length > args->space) {
824 bcopy(args->fname, args->endp, length);
825 args->space -= length;
826 args->endp += length;
832 args->begin_envv = args->endp;
835 * extract environment strings. envv may be NULL.
837 if (envv && error == 0) {
838 while ((envp = (caddr_t) (intptr_t) fuword(envv++))) {
839 if (envp == (caddr_t) -1) {
843 error = copyinstr(envp, args->endp, args->space,
846 if (error == ENAMETOOLONG)
850 args->space -= length;
851 args->endp += length;
859 exec_free_args(struct image_args *args)
862 objcache_put(exec_objcache, args->buf);
868 * Copy strings out to the new process address space, constructing
869 * new arg and env vector tables. Return a pointer to the base
870 * so that it can be used as the initial stack pointer.
873 exec_copyout_strings(struct image_params *imgp)
875 int argc, envc, sgap;
877 char *stringp, *destp;
878 register_t *stack_base;
879 struct ps_strings *arginfo;
883 * Calculate string base and vector table pointers.
884 * Also deal with signal trampoline code for this exec type.
886 arginfo = (struct ps_strings *)PS_STRINGS;
887 szsigcode = *(imgp->proc->p_sysent->sv_szsigcode);
888 if (stackgap_random != 0)
889 sgap = ALIGN(karc4random() & (stackgap_random - 1));
892 destp = (caddr_t)arginfo - szsigcode - SPARE_USRSPACE - sgap -
893 roundup((ARG_MAX - imgp->args->space), sizeof(char *));
899 copyout(imgp->proc->p_sysent->sv_sigcode,
900 ((caddr_t)arginfo - szsigcode), szsigcode);
903 * If we have a valid auxargs ptr, prepare some room
906 * The '+ 2' is for the null pointers at the end of each of the
907 * arg and env vector sets, and 'AT_COUNT*2' is room for the
911 vectp = (char **)(destp - (imgp->args->argc +
912 imgp->args->envc + 2 + AT_COUNT * 2) * sizeof(char*));
914 vectp = (char **)(destp - (imgp->args->argc +
915 imgp->args->envc + 2) * sizeof(char*));
919 * NOTE: don't bother aligning the stack here for GCC 2.x, it will
920 * be done in crt1.o. Note that GCC 3.x aligns the stack in main.
924 * vectp also becomes our initial stack base
926 stack_base = (register_t *)vectp;
928 stringp = imgp->args->begin_argv;
929 argc = imgp->args->argc;
930 envc = imgp->args->envc;
933 * Copy out strings - arguments and environment.
935 copyout(stringp, destp, ARG_MAX - imgp->args->space);
938 * Fill in "ps_strings" struct for ps, w, etc.
940 suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp);
941 suword(&arginfo->ps_nargvstr, argc);
944 * Fill in argument portion of vector table.
946 for (; argc > 0; --argc) {
947 suword(vectp++, (long)(intptr_t)destp);
948 while (*stringp++ != 0)
953 /* a null vector table pointer separates the argp's from the envp's */
956 suword(&arginfo->ps_envstr, (long)(intptr_t)vectp);
957 suword(&arginfo->ps_nenvstr, envc);
960 * Fill in environment portion of vector table.
962 for (; envc > 0; --envc) {
963 suword(vectp++, (long)(intptr_t)destp);
964 while (*stringp++ != 0)
969 /* end of vector table is a null pointer */
976 * Check permissions of file to execute.
977 * Return 0 for success or error code on failure.
980 exec_check_permissions(struct image_params *imgp)
982 struct proc *p = imgp->proc;
983 struct vnode *vp = imgp->vp;
984 struct vattr *attr = imgp->attr;
987 /* Get file attributes */
988 error = VOP_GETATTR(vp, attr);
993 * 1) Check if file execution is disabled for the filesystem that this
995 * 2) Insure that at least one execute bit is on - otherwise root
996 * will always succeed, and we don't want to happen unless the
997 * file really is executable.
998 * 3) Insure that the file is a regular file.
1000 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
1001 ((attr->va_mode & 0111) == 0) ||
1002 (attr->va_type != VREG)) {
1007 * Zero length files can't be exec'd
1009 if (attr->va_size == 0)
1013 * Check for execute permission to file based on current credentials.
1015 error = VOP_EACCESS(vp, VEXEC, p->p_ucred);
1020 * Check number of open-for-writes on the file and deny execution
1023 if (vp->v_writecount)
1027 * Call filesystem specific open routine, which allows us to read,
1028 * write, and mmap the file. Without the VOP_OPEN we can only
1031 error = VOP_OPEN(vp, FREAD, p->p_ucred, NULL);
1039 * Exec handler registration
1042 exec_register(const struct execsw *execsw_arg)
1044 const struct execsw **es, **xs, **newexecsw;
1045 int count = 2; /* New slot and trailing NULL */
1048 for (es = execsw; *es; es++)
1050 newexecsw = kmalloc(count * sizeof(*es), M_TEMP, M_WAITOK);
1053 for (es = execsw; *es; es++)
1058 kfree(execsw, M_TEMP);
1064 exec_unregister(const struct execsw *execsw_arg)
1066 const struct execsw **es, **xs, **newexecsw;
1070 panic("unregister with no handlers left?");
1072 for (es = execsw; *es; es++) {
1073 if (*es == execsw_arg)
1078 for (es = execsw; *es; es++)
1079 if (*es != execsw_arg)
1081 newexecsw = kmalloc(count * sizeof(*es), M_TEMP, M_WAITOK);
1083 for (es = execsw; *es; es++)
1084 if (*es != execsw_arg)
1088 kfree(execsw, M_TEMP);