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/refcount.h>
76 #include <sys/thread2.h>
77 #include <sys/mplock2.h>
79 MALLOC_DEFINE(M_PARGS, "proc-args", "Process arguments");
80 MALLOC_DEFINE(M_EXECARGS, "exec-args", "Exec arguments");
82 static register_t *exec_copyout_strings (struct image_params *);
84 /* XXX This should be vm_size_t. */
85 static u_long ps_strings = PS_STRINGS;
86 SYSCTL_ULONG(_kern, KERN_PS_STRINGS, ps_strings, CTLFLAG_RD, &ps_strings, 0, "");
88 /* XXX This should be vm_size_t. */
89 static u_long usrstack = USRSTACK;
90 SYSCTL_ULONG(_kern, KERN_USRSTACK, usrstack, CTLFLAG_RD, &usrstack, 0, "");
92 u_long ps_arg_cache_limit = PAGE_SIZE / 16;
93 SYSCTL_LONG(_kern, OID_AUTO, ps_arg_cache_limit, CTLFLAG_RW,
94 &ps_arg_cache_limit, 0, "");
97 SYSCTL_INT(_kern, OID_AUTO, ps_argsopen, CTLFLAG_RW, &ps_argsopen, 0, "");
99 static int ktrace_suid = 0;
100 SYSCTL_INT(_kern, OID_AUTO, ktrace_suid, CTLFLAG_RW, &ktrace_suid, 0, "");
102 void print_execve_args(struct image_args *args);
103 int debug_execve_args = 0;
104 SYSCTL_INT(_kern, OID_AUTO, debug_execve_args, CTLFLAG_RW, &debug_execve_args,
108 * Exec arguments object cache
110 static struct objcache *exec_objcache;
114 exec_objcache_init(void *arg __unused)
118 cluster_limit = 16; /* up to this many objects */
119 exec_objcache = objcache_create_mbacked(
120 M_EXECARGS, PATH_MAX + ARG_MAX,
122 2, /* minimal magazine capacity */
125 SYSINIT(exec_objcache, SI_BOOT2_MACHDEP, SI_ORDER_ANY, exec_objcache_init, 0);
128 * stackgap_random specifies if the stackgap should have a random size added
129 * to it. It must be a power of 2. If non-zero, the stack gap will be
130 * calculated as: ALIGN(karc4random() & (stackgap_random - 1)).
132 static int stackgap_random = 1024;
134 sysctl_kern_stackgap(SYSCTL_HANDLER_ARGS)
137 new_val = stackgap_random;
138 error = sysctl_handle_int(oidp, &new_val, 0, req);
139 if (error != 0 || req->newptr == NULL)
141 if ((new_val < 0) || (new_val > 16 * PAGE_SIZE) || ! powerof2(new_val))
143 stackgap_random = new_val;
148 SYSCTL_PROC(_kern, OID_AUTO, stackgap_random, CTLFLAG_RW|CTLTYPE_UINT,
149 0, 0, sysctl_kern_stackgap, "IU", "Max random stack gap (power of 2)");
152 print_execve_args(struct image_args *args)
157 cp = args->begin_argv;
158 for (ndx = 0; ndx < args->argc; ndx++) {
159 kprintf("\targv[%d]: %s\n", ndx, cp);
160 while (*cp++ != '\0');
162 for (ndx = 0; ndx < args->envc; ndx++) {
163 kprintf("\tenvv[%d]: %s\n", ndx, cp);
164 while (*cp++ != '\0');
169 * Each of the items is a pointer to a `const struct execsw', hence the
170 * double pointer here.
172 static const struct execsw **execsw;
175 * Replace current vmspace with a new binary.
176 * Returns 0 on success, > 0 on recoverable error (use as errno).
177 * Returns -1 on lethal error which demands killing of the current
181 kern_execve(struct nlookupdata *nd, struct image_args *args)
183 struct thread *td = curthread;
184 struct lwp *lp = td->td_lwp;
185 struct proc *p = td->td_proc;
186 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;
345 * For security and other reasons, signal handlers cannot
346 * be shared after an exec. The new proces gets a copy of the old
347 * handlers. In execsigs(), the new process will have its signals
350 if (p->p_sigacts->ps_refcnt > 1) {
351 struct sigacts *newsigacts;
353 newsigacts = (struct sigacts *)kmalloc(sizeof(*newsigacts),
354 M_SUBPROC, M_WAITOK);
355 bcopy(p->p_sigacts, newsigacts, sizeof(*newsigacts));
356 p->p_sigacts->ps_refcnt--;
357 p->p_sigacts = newsigacts;
358 p->p_sigacts->ps_refcnt = 1;
362 * For security and other reasons virtual kernels cannot be
363 * inherited by an exec. This also allows a virtual kernel
364 * to fork/exec unrelated applications.
372 /* close files on exec */
375 /* reset caught signals */
378 /* name this process - nameiexec(p, ndp) */
379 len = min(nd->nl_nch.ncp->nc_nlen, MAXCOMLEN);
380 bcopy(nd->nl_nch.ncp->nc_name, p->p_comm, len);
382 bcopy(p->p_comm, lp->lwp_thread->td_comm, MAXCOMLEN+1);
385 * mark as execed, wakeup the process that vforked (if any) and tell
386 * it that it now has its own resources back
389 if (p->p_pptr && (p->p_flag & P_PPWAIT)) {
390 p->p_flag &= ~P_PPWAIT;
391 wakeup((caddr_t)p->p_pptr);
395 * Implement image setuid/setgid.
397 * Don't honor setuid/setgid if the filesystem prohibits it or if
398 * the process is being traced.
400 if ((((attr.va_mode & VSUID) && p->p_ucred->cr_uid != attr.va_uid) ||
401 ((attr.va_mode & VSGID) && p->p_ucred->cr_gid != attr.va_gid)) &&
402 (imgp->vp->v_mount->mnt_flag & MNT_NOSUID) == 0 &&
403 (p->p_flag & P_TRACED) == 0) {
405 * Turn off syscall tracing for set-id programs, except for
406 * root. Record any set-id flags first to make sure that
407 * we do not regain any tracing during a possible block.
410 if (p->p_tracenode && ktrace_suid == 0 &&
411 priv_check(td, PRIV_ROOT) != 0) {
412 ktrdestroy(&p->p_tracenode);
415 /* Close any file descriptors 0..2 that reference procfs */
417 /* Make sure file descriptors 0..2 are in use. */
418 error = fdcheckstd(lp);
420 goto exec_fail_dealloc;
422 * Set the new credentials.
425 if (attr.va_mode & VSUID)
426 change_euid(attr.va_uid);
427 if (attr.va_mode & VSGID)
428 p->p_ucred->cr_gid = attr.va_gid;
431 * Clear local varsym variables
433 varsymset_clean(&p->p_varsymset);
435 if (p->p_ucred->cr_uid == p->p_ucred->cr_ruid &&
436 p->p_ucred->cr_gid == p->p_ucred->cr_rgid)
437 p->p_flag &= ~P_SUGID;
441 * Implement correct POSIX saved-id behavior.
443 if (p->p_ucred->cr_svuid != p->p_ucred->cr_uid ||
444 p->p_ucred->cr_svgid != p->p_ucred->cr_gid) {
446 p->p_ucred->cr_svuid = p->p_ucred->cr_uid;
447 p->p_ucred->cr_svgid = p->p_ucred->cr_gid;
451 * Store the vp for use in procfs
453 if (p->p_textvp) /* release old reference */
455 p->p_textvp = imgp->vp;
458 /* Release old namecache handle to text file */
459 if (p->p_textnch.ncp)
460 cache_drop(&p->p_textnch);
462 if (nd->nl_nch.mount)
463 cache_copy(&nd->nl_nch, &p->p_textnch);
466 * Notify others that we exec'd, and clear the P_INEXEC flag
467 * as we're now a bona fide freshly-execed process.
469 KNOTE(&p->p_klist, NOTE_EXEC);
470 p->p_flag &= ~P_INEXEC;
473 * If tracing the process, trap to debugger so breakpoints
474 * can be set before the program executes.
476 STOPEVENT(p, S_EXEC, 0);
478 if (p->p_flag & P_TRACED)
481 /* clear "fork but no exec" flag, as we _are_ execing */
482 p->p_acflag &= ~AFORK;
484 /* Set values passed into the program in registers. */
485 exec_setregs(imgp->entry_addr, (u_long)(uintptr_t)stack_base,
488 /* Set the access time on the vnode */
489 vn_mark_atime(imgp->vp, td);
492 * Free any previous argument cache
496 if (pa && refcount_release(&pa->ar_ref)) {
502 * Cache arguments if they fit inside our allowance
504 i = imgp->args->begin_envv - imgp->args->begin_argv;
505 if (sizeof(struct pargs) + i <= ps_arg_cache_limit) {
506 pa = kmalloc(sizeof(struct pargs) + i, M_PARGS, M_WAITOK);
507 refcount_init(&pa->ar_ref, 1);
509 bcopy(imgp->args->begin_argv, pa->ar_args, i);
510 KKASSERT(p->p_args == NULL);
517 * free various allocated resources
520 exec_unmap_first_page(imgp);
528 ++mycpu->gd_cnt.v_exec;
529 lwkt_reltoken(&p->p_token);
535 * we're done here, clear P_INEXEC if we were the ones that
536 * set it. Otherwise if vmspace_destroyed is still set we
537 * raced another thread and that thread is responsible for
540 if (imgp->vmspace_destroyed & 2)
541 p->p_flag &= ~P_INEXEC;
542 lwkt_reltoken(&p->p_token);
543 if (imgp->vmspace_destroyed) {
545 * Sorry, no more process anymore. exit gracefully.
546 * However we can't die right here, because our
547 * caller might have to clean up, so indicate a
548 * lethal error by returning -1.
557 * execve() system call.
562 sys_execve(struct execve_args *uap)
564 struct nlookupdata nd;
565 struct image_args args;
568 bzero(&args, sizeof(args));
571 error = nlookup_init(&nd, uap->fname, UIO_USERSPACE, NLC_FOLLOW);
573 error = exec_copyin_args(&args, uap->fname, PATH_USERSPACE,
574 uap->argv, uap->envv);
577 error = kern_execve(&nd, &args);
579 exec_free_args(&args);
582 /* We hit a lethal error condition. Let's die now. */
583 exit1(W_EXITCODE(0, SIGABRT));
589 * The syscall result is returned in registers to the new program.
590 * Linux will register %edx as an atexit function and we must be
591 * sure to set it to 0. XXX
594 uap->sysmsg_result64 = 0;
600 exec_map_page(struct image_params *imgp, vm_pindex_t pageno,
601 struct lwbuf **plwb, const char **pdata)
609 * The file has to be mappable.
611 if ((object = imgp->vp->v_object) == NULL)
614 if (pageno >= object->size)
617 m = vm_page_grab(object, pageno, VM_ALLOC_NORMAL | VM_ALLOC_RETRY);
619 lwkt_gettoken(&vm_token);
620 while ((m->valid & VM_PAGE_BITS_ALL) != VM_PAGE_BITS_ALL) {
624 * get_pages unbusies all the requested pages except the
625 * primary page (at index 0 in this case). The primary
626 * page may have been wired during the pagein (e.g. by
627 * the buffer cache) so vnode_pager_freepage() must be
628 * used to properly release it.
630 rv = vm_pager_get_page(object, &ma, 1);
631 m = vm_page_lookup(object, pageno);
633 if (rv != VM_PAGER_OK || m == NULL || m->valid == 0) {
635 vm_page_protect(m, VM_PROT_NONE);
636 vnode_pager_freepage(m);
638 lwkt_reltoken(&vm_token);
642 vm_page_hold(m); /* requires vm_token to be held */
643 vm_page_wakeup(m); /* unbusy the page */
644 lwkt_reltoken(&vm_token);
646 *plwb = lwbuf_alloc(m, *plwb);
647 *pdata = (void *)lwbuf_kva(*plwb);
653 exec_map_first_page(struct image_params *imgp)
658 exec_unmap_first_page(imgp);
660 imgp->firstpage = &imgp->firstpage_cache;
661 err = exec_map_page(imgp, 0, &imgp->firstpage, &imgp->image_header);
670 exec_unmap_page(struct lwbuf *lwb)
684 exec_unmap_first_page(struct image_params *imgp)
686 exec_unmap_page(imgp->firstpage);
687 imgp->firstpage = NULL;
688 imgp->image_header = NULL;
692 * Destroy old address space, and allocate a new stack
693 * The new stack is only SGROWSIZ large because it is grown
694 * automatically in trap.c.
696 * This is the point of no return.
699 exec_new_vmspace(struct image_params *imgp, struct vmspace *vmcopy)
701 struct vmspace *vmspace = imgp->proc->p_vmspace;
702 vm_offset_t stack_addr = USRSTACK - maxssiz;
708 * Indicate that we cannot gracefully error out any more, kill
709 * any other threads present, and set P_INEXEC to indicate that
710 * we are now messing with the process structure proper.
712 * If killalllwps() races return an error which coupled with
713 * vmspace_destroyed will cause us to exit. This is what we
714 * want since another thread is patiently waiting for us to exit
718 imgp->vmspace_destroyed = 1;
720 if (curthread->td_proc->p_nthreads > 1) {
721 error = killalllwps(1);
725 imgp->vmspace_destroyed |= 2; /* we are responsible for P_INEXEC */
726 p->p_flag |= P_INEXEC;
729 * Prevent a pending AIO from modifying the new address space.
731 aio_proc_rundown(imgp->proc);
734 * Blow away entire process VM, if address space not shared,
735 * otherwise, create a new VM space so that other threads are
736 * not disrupted. If we are execing a resident vmspace we
737 * create a duplicate of it and remap the stack.
739 * The exitingcnt test is not strictly necessary but has been
740 * included for code sanity (to make the code more deterministic).
742 map = &vmspace->vm_map;
744 vmspace_exec(imgp->proc, vmcopy);
745 vmspace = imgp->proc->p_vmspace;
746 pmap_remove_pages(vmspace_pmap(vmspace), stack_addr, USRSTACK);
747 map = &vmspace->vm_map;
748 } else if (vmspace->vm_sysref.refcnt == 1 &&
749 vmspace->vm_exitingcnt == 0) {
751 if (vmspace->vm_upcalls)
752 upc_release(vmspace, ONLY_LWP_IN_PROC(imgp->proc));
753 pmap_remove_pages(vmspace_pmap(vmspace),
754 0, VM_MAX_USER_ADDRESS);
755 vm_map_remove(map, 0, VM_MAX_USER_ADDRESS);
757 vmspace_exec(imgp->proc, NULL);
758 vmspace = imgp->proc->p_vmspace;
759 map = &vmspace->vm_map;
762 /* Allocate a new stack */
763 error = vm_map_stack(&vmspace->vm_map, stack_addr, (vm_size_t)maxssiz,
764 0, VM_PROT_ALL, VM_PROT_ALL, 0);
768 /* vm_ssize and vm_maxsaddr are somewhat antiquated concepts in the
769 * VM_STACK case, but they are still used to monitor the size of the
770 * process stack so we can check the stack rlimit.
772 vmspace->vm_ssize = sgrowsiz >> PAGE_SHIFT;
773 vmspace->vm_maxsaddr = (char *)USRSTACK - maxssiz;
779 * Copy out argument and environment strings from the old process
780 * address space into the temporary string buffer.
783 exec_copyin_args(struct image_args *args, char *fname,
784 enum exec_path_segflg segflg, char **argv, char **envv)
790 args->buf = objcache_get(exec_objcache, M_WAITOK);
791 if (args->buf == NULL)
793 args->begin_argv = args->buf;
794 args->endp = args->begin_argv;
795 args->space = ARG_MAX;
797 args->fname = args->buf + ARG_MAX;
800 * Copy the file name.
802 if (segflg == PATH_SYSSPACE) {
803 error = copystr(fname, args->fname, PATH_MAX, &length);
804 } else if (segflg == PATH_USERSPACE) {
805 error = copyinstr(fname, args->fname, PATH_MAX, &length);
809 * Extract argument strings. argv may not be NULL. The argv
810 * array is terminated by a NULL entry. We special-case the
811 * situation where argv[0] is NULL by passing { filename, NULL }
812 * to the new program to guarentee that the interpreter knows what
813 * file to open in case we exec an interpreted file. Note that
814 * a NULL argv[0] terminates the argv[] array.
816 * XXX the special-casing of argv[0] is historical and needs to be
822 while ((argp = (caddr_t)(intptr_t)fuword(argv++)) != NULL) {
823 if (argp == (caddr_t)-1) {
827 error = copyinstr(argp, args->endp,
828 args->space, &length);
830 if (error == ENAMETOOLONG)
834 args->space -= length;
835 args->endp += length;
838 if (args->argc == 0 && error == 0) {
839 length = strlen(args->fname) + 1;
840 if (length > args->space) {
843 bcopy(args->fname, args->endp, length);
844 args->space -= length;
845 args->endp += length;
851 args->begin_envv = args->endp;
854 * extract environment strings. envv may be NULL.
856 if (envv && error == 0) {
857 while ((envp = (caddr_t) (intptr_t) fuword(envv++))) {
858 if (envp == (caddr_t) -1) {
862 error = copyinstr(envp, args->endp, args->space,
865 if (error == ENAMETOOLONG)
869 args->space -= length;
870 args->endp += length;
878 exec_free_args(struct image_args *args)
881 objcache_put(exec_objcache, args->buf);
887 * Copy strings out to the new process address space, constructing
888 * new arg and env vector tables. Return a pointer to the base
889 * so that it can be used as the initial stack pointer.
892 exec_copyout_strings(struct image_params *imgp)
894 int argc, envc, sgap;
896 char *stringp, *destp;
897 register_t *stack_base;
898 struct ps_strings *arginfo;
902 * Calculate string base and vector table pointers.
903 * Also deal with signal trampoline code for this exec type.
905 arginfo = (struct ps_strings *)PS_STRINGS;
906 szsigcode = *(imgp->proc->p_sysent->sv_szsigcode);
907 if (stackgap_random != 0)
908 sgap = ALIGN(karc4random() & (stackgap_random - 1));
911 destp = (caddr_t)arginfo - szsigcode - SPARE_USRSPACE - sgap -
912 roundup((ARG_MAX - imgp->args->space), sizeof(char *));
918 copyout(imgp->proc->p_sysent->sv_sigcode,
919 ((caddr_t)arginfo - szsigcode), szsigcode);
922 * If we have a valid auxargs ptr, prepare some room
925 * The '+ 2' is for the null pointers at the end of each of the
926 * arg and env vector sets, and 'AT_COUNT*2' is room for the
930 vectp = (char **)(destp - (imgp->args->argc +
931 imgp->args->envc + 2 + AT_COUNT * 2) * sizeof(char*));
933 vectp = (char **)(destp - (imgp->args->argc +
934 imgp->args->envc + 2) * sizeof(char*));
938 * NOTE: don't bother aligning the stack here for GCC 2.x, it will
939 * be done in crt1.o. Note that GCC 3.x aligns the stack in main.
943 * vectp also becomes our initial stack base
945 stack_base = (register_t *)vectp;
947 stringp = imgp->args->begin_argv;
948 argc = imgp->args->argc;
949 envc = imgp->args->envc;
952 * Copy out strings - arguments and environment.
954 copyout(stringp, destp, ARG_MAX - imgp->args->space);
957 * Fill in "ps_strings" struct for ps, w, etc.
959 suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp);
960 suword(&arginfo->ps_nargvstr, argc);
963 * Fill in argument portion of vector table.
965 for (; argc > 0; --argc) {
966 suword(vectp++, (long)(intptr_t)destp);
967 while (*stringp++ != 0)
972 /* a null vector table pointer separates the argp's from the envp's */
975 suword(&arginfo->ps_envstr, (long)(intptr_t)vectp);
976 suword(&arginfo->ps_nenvstr, envc);
979 * Fill in environment portion of vector table.
981 for (; envc > 0; --envc) {
982 suword(vectp++, (long)(intptr_t)destp);
983 while (*stringp++ != 0)
988 /* end of vector table is a null pointer */
995 * Check permissions of file to execute.
996 * Return 0 for success or error code on failure.
999 exec_check_permissions(struct image_params *imgp, struct mount *topmnt)
1001 struct proc *p = imgp->proc;
1002 struct vnode *vp = imgp->vp;
1003 struct vattr *attr = imgp->attr;
1006 /* Get file attributes */
1007 error = VOP_GETATTR(vp, attr);
1012 * 1) Check if file execution is disabled for the filesystem that this
1014 * 2) Insure that at least one execute bit is on - otherwise root
1015 * will always succeed, and we don't want to happen unless the
1016 * file really is executable.
1017 * 3) Insure that the file is a regular file.
1019 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
1020 ((topmnt != NULL) && (topmnt->mnt_flag & MNT_NOEXEC)) ||
1021 ((attr->va_mode & 0111) == 0) ||
1022 (attr->va_type != VREG)) {
1027 * Zero length files can't be exec'd
1029 if (attr->va_size == 0)
1033 * Check for execute permission to file based on current credentials.
1035 error = VOP_EACCESS(vp, VEXEC, p->p_ucred);
1040 * Check number of open-for-writes on the file and deny execution
1043 if (vp->v_writecount)
1047 * Call filesystem specific open routine, which allows us to read,
1048 * write, and mmap the file. Without the VOP_OPEN we can only
1051 error = VOP_OPEN(vp, FREAD, p->p_ucred, NULL);
1059 * Exec handler registration
1062 exec_register(const struct execsw *execsw_arg)
1064 const struct execsw **es, **xs, **newexecsw;
1065 int count = 2; /* New slot and trailing NULL */
1068 for (es = execsw; *es; es++)
1070 newexecsw = kmalloc(count * sizeof(*es), M_TEMP, M_WAITOK);
1073 for (es = execsw; *es; es++)
1078 kfree(execsw, M_TEMP);
1084 exec_unregister(const struct execsw *execsw_arg)
1086 const struct execsw **es, **xs, **newexecsw;
1090 panic("unregister with no handlers left?");
1092 for (es = execsw; *es; es++) {
1093 if (*es == execsw_arg)
1098 for (es = execsw; *es; es++)
1099 if (*es != execsw_arg)
1101 newexecsw = kmalloc(count * sizeof(*es), M_TEMP, M_WAITOK);
1103 for (es = execsw; *es; es++)
1104 if (*es != execsw_arg)
1108 kfree(execsw, M_TEMP);