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 $
29 #include <sys/param.h>
30 #include <sys/systm.h>
31 #include <sys/sysproto.h>
32 #include <sys/kernel.h>
33 #include <sys/mount.h>
34 #include <sys/filedesc.h>
35 #include <sys/fcntl.h>
38 #include <sys/imgact.h>
39 #include <sys/imgact_elf.h>
40 #include <sys/kern_syscall.h>
42 #include <sys/malloc.h>
45 #include <sys/ktrace.h>
46 #include <sys/signalvar.h>
47 #include <sys/pioctl.h>
48 #include <sys/nlookup.h>
49 #include <sys/sysent.h>
51 #include <sys/sysctl.h>
52 #include <sys/vnode.h>
53 #include <sys/vmmeter.h>
54 #include <sys/libkern.h>
56 #include <cpu/lwbuf.h>
59 #include <vm/vm_param.h>
62 #include <vm/vm_page.h>
63 #include <vm/vm_map.h>
64 #include <vm/vm_kern.h>
65 #include <vm/vm_extern.h>
66 #include <vm/vm_object.h>
67 #include <vm/vnode_pager.h>
68 #include <vm/vm_pager.h>
73 #include <sys/refcount.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,
122 SYSINIT(exec_objcache, SI_BOOT2_MACHDEP, SI_ORDER_ANY, exec_objcache_init, 0);
125 * stackgap_random specifies if the stackgap should have a random size added
126 * to it. It must be a power of 2. If non-zero, the stack gap will be
127 * calculated as: ALIGN(karc4random() & (stackgap_random - 1)).
129 static int stackgap_random = 1024;
131 sysctl_kern_stackgap(SYSCTL_HANDLER_ARGS)
134 new_val = stackgap_random;
135 error = sysctl_handle_int(oidp, &new_val, 0, req);
136 if (error != 0 || req->newptr == NULL)
138 if ((new_val < 0) || (new_val > 16 * PAGE_SIZE) || ! powerof2(new_val))
140 stackgap_random = new_val;
145 SYSCTL_PROC(_kern, OID_AUTO, stackgap_random, CTLFLAG_RW|CTLTYPE_UINT,
146 0, 0, sysctl_kern_stackgap, "IU", "Max random stack gap (power of 2)");
149 print_execve_args(struct image_args *args)
154 cp = args->begin_argv;
155 for (ndx = 0; ndx < args->argc; ndx++) {
156 kprintf("\targv[%d]: %s\n", ndx, cp);
157 while (*cp++ != '\0');
159 for (ndx = 0; ndx < args->envc; ndx++) {
160 kprintf("\tenvv[%d]: %s\n", ndx, cp);
161 while (*cp++ != '\0');
166 * Each of the items is a pointer to a `const struct execsw', hence the
167 * double pointer here.
169 static const struct execsw **execsw;
172 * Replace current vmspace with a new binary.
173 * Returns 0 on success, > 0 on recoverable error (use as errno).
174 * Returns -1 on lethal error which demands killing of the current
178 kern_execve(struct nlookupdata *nd, struct image_args *args)
180 struct thread *td = curthread;
181 struct lwp *lp = td->td_lwp;
182 struct proc *p = td->td_proc;
183 register_t *stack_base;
188 struct image_params image_params, *imgp;
190 int (*img_first) (struct image_params *);
192 if (debug_execve_args) {
193 kprintf("%s()\n", __func__);
194 print_execve_args(args);
198 lwkt_gettoken(&p->p_token);
199 imgp = &image_params;
202 * NOTE: P_INEXEC is handled by exec_new_vmspace() now. We make
203 * no modifications to the process at all until we get there.
205 * Note that multiple threads may be trying to exec at the same
206 * time. exec_new_vmspace() handles that too.
210 * Initialize part of the common data
215 imgp->entry_addr = 0;
217 imgp->vmspace_destroyed = 0;
218 imgp->interpreted = 0;
219 imgp->interpreter_name[0] = 0;
220 imgp->auxargs = NULL;
222 imgp->firstpage = NULL;
223 imgp->ps_strings = 0;
224 imgp->execpath = imgp->freepath = NULL;
226 imgp->image_header = NULL;
231 * Translate the file name to a vnode. Unlock the cache entry to
232 * improve parallelism for programs exec'd in parallel.
234 if ((error = nlookup(nd)) != 0)
236 error = cache_vget(&nd->nl_nch, nd->nl_cred, LK_EXCLUSIVE, &imgp->vp);
237 KKASSERT(nd->nl_flags & NLC_NCPISLOCKED);
238 nd->nl_flags &= ~NLC_NCPISLOCKED;
239 cache_unlock(&nd->nl_nch);
244 * Check file permissions (also 'opens' file).
245 * Include also the top level mount in the check.
247 error = exec_check_permissions(imgp, nd->nl_nch.mount);
250 goto exec_fail_dealloc;
253 error = exec_map_first_page(imgp);
256 goto exec_fail_dealloc;
258 imgp->proc->p_osrel = 0;
260 if (debug_execve_args && imgp->interpreted) {
261 kprintf(" target is interpreted -- recursive pass\n");
262 kprintf(" interpreter: %s\n", imgp->interpreter_name);
263 print_execve_args(args);
267 * If the current process has a special image activator it
268 * wants to try first, call it. For example, emulating shell
269 * scripts differently.
272 if ((img_first = imgp->proc->p_sysent->sv_imgact_try) != NULL)
273 error = img_first(imgp);
276 * If the vnode has a registered vmspace, exec the vmspace
278 if (error == -1 && imgp->vp->v_resident) {
279 error = exec_resident_imgact(imgp);
283 * Loop through the list of image activators, calling each one.
284 * An activator returns -1 if there is no match, 0 on success,
285 * and an error otherwise.
287 for (i = 0; error == -1 && execsw[i]; ++i) {
288 if (execsw[i]->ex_imgact == NULL ||
289 execsw[i]->ex_imgact == img_first) {
292 error = (*execsw[i]->ex_imgact)(imgp);
298 goto exec_fail_dealloc;
302 * Special interpreter operation, cleanup and loop up to try to
303 * activate the interpreter.
305 if (imgp->interpreted) {
306 exec_unmap_first_page(imgp);
310 error = nlookup_init(nd, imgp->interpreter_name, UIO_SYSSPACE,
318 * Do the best to calculate the full path to the image file
320 if (imgp->auxargs != NULL &&
321 ((args->fname != NULL && args->fname[0] == '/') ||
322 vn_fullpath(imgp->proc,
327 imgp->execpath = args->fname;
330 * Copy out strings (args and env) and initialize stack base
332 stack_base = exec_copyout_strings(imgp);
333 p->p_vmspace->vm_minsaddr = (char *)stack_base;
336 * If custom stack fixup routine present for this process
337 * let it do the stack setup. If we are running a resident
338 * image there is no auxinfo or other image activator context
339 * so don't try to add fixups to the stack.
341 * Else stuff argument count as first item on stack
343 if (p->p_sysent->sv_fixup && imgp->resident == 0)
344 (*p->p_sysent->sv_fixup)(&stack_base, imgp);
346 suword(--stack_base, imgp->args->argc);
349 * For security and other reasons, the file descriptor table cannot
350 * be shared after an exec.
352 if (p->p_fd->fd_refcnt > 1) {
353 struct filedesc *tmp;
355 error = fdcopy(p, &tmp);
362 * For security and other reasons, signal handlers cannot
363 * be shared after an exec. The new proces gets a copy of the old
364 * handlers. In execsigs(), the new process will have its signals
368 if (ops->ps_refcnt > 1) {
369 nps = kmalloc(sizeof(*nps), M_SUBPROC, M_WAITOK);
370 bcopy(ops, nps, sizeof(*nps));
371 refcount_init(&nps->ps_refcnt, 1);
373 if (refcount_release(&ops->ps_refcnt)) {
374 kfree(ops, M_SUBPROC);
380 * For security and other reasons virtual kernels cannot be
381 * inherited by an exec. This also allows a virtual kernel
382 * to fork/exec unrelated applications.
390 /* close files on exec */
393 /* reset caught signals */
396 /* name this process - nameiexec(p, ndp) */
397 len = min(nd->nl_nch.ncp->nc_nlen, MAXCOMLEN);
398 bcopy(nd->nl_nch.ncp->nc_name, p->p_comm, len);
400 bcopy(p->p_comm, lp->lwp_thread->td_comm, MAXCOMLEN+1);
403 * mark as execed, wakeup the process that vforked (if any) and tell
404 * it that it now has its own resources back
407 if (p->p_pptr && (p->p_flag & P_PPWAIT)) {
408 p->p_flag &= ~P_PPWAIT;
409 wakeup((caddr_t)p->p_pptr);
413 * Implement image setuid/setgid.
415 * Don't honor setuid/setgid if the filesystem prohibits it or if
416 * the process is being traced.
418 if ((((attr.va_mode & VSUID) && p->p_ucred->cr_uid != attr.va_uid) ||
419 ((attr.va_mode & VSGID) && p->p_ucred->cr_gid != attr.va_gid)) &&
420 (imgp->vp->v_mount->mnt_flag & MNT_NOSUID) == 0 &&
421 (p->p_flag & P_TRACED) == 0) {
423 * Turn off syscall tracing for set-id programs, except for
424 * root. Record any set-id flags first to make sure that
425 * we do not regain any tracing during a possible block.
428 if (p->p_tracenode && ktrace_suid == 0 &&
429 priv_check(td, PRIV_ROOT) != 0) {
430 ktrdestroy(&p->p_tracenode);
433 /* Close any file descriptors 0..2 that reference procfs */
435 /* Make sure file descriptors 0..2 are in use. */
436 error = fdcheckstd(lp);
438 goto exec_fail_dealloc;
440 * Set the new credentials.
443 if (attr.va_mode & VSUID)
444 change_euid(attr.va_uid);
445 if (attr.va_mode & VSGID)
446 p->p_ucred->cr_gid = attr.va_gid;
449 * Clear local varsym variables
451 varsymset_clean(&p->p_varsymset);
453 if (p->p_ucred->cr_uid == p->p_ucred->cr_ruid &&
454 p->p_ucred->cr_gid == p->p_ucred->cr_rgid)
455 p->p_flag &= ~P_SUGID;
459 * Implement correct POSIX saved-id behavior.
461 if (p->p_ucred->cr_svuid != p->p_ucred->cr_uid ||
462 p->p_ucred->cr_svgid != p->p_ucred->cr_gid) {
464 p->p_ucred->cr_svuid = p->p_ucred->cr_uid;
465 p->p_ucred->cr_svgid = p->p_ucred->cr_gid;
469 * Store the vp for use in procfs
471 if (p->p_textvp) /* release old reference */
473 p->p_textvp = imgp->vp;
476 /* Release old namecache handle to text file */
477 if (p->p_textnch.ncp)
478 cache_drop(&p->p_textnch);
480 if (nd->nl_nch.mount)
481 cache_copy(&nd->nl_nch, &p->p_textnch);
484 * Notify others that we exec'd, and clear the P_INEXEC flag
485 * as we're now a bona fide freshly-execed process.
487 KNOTE(&p->p_klist, NOTE_EXEC);
488 p->p_flag &= ~P_INEXEC;
491 * If tracing the process, trap to debugger so breakpoints
492 * can be set before the program executes.
494 STOPEVENT(p, S_EXEC, 0);
496 if (p->p_flag & P_TRACED)
499 /* clear "fork but no exec" flag, as we _are_ execing */
500 p->p_acflag &= ~AFORK;
502 /* Set values passed into the program in registers. */
503 exec_setregs(imgp->entry_addr, (u_long)(uintptr_t)stack_base,
506 /* Set the access time on the vnode */
507 vn_mark_atime(imgp->vp, td);
510 * Free any previous argument cache
514 if (pa && refcount_release(&pa->ar_ref)) {
520 * Cache arguments if they fit inside our allowance
522 i = imgp->args->begin_envv - imgp->args->begin_argv;
523 if (sizeof(struct pargs) + i <= ps_arg_cache_limit) {
524 pa = kmalloc(sizeof(struct pargs) + i, M_PARGS, M_WAITOK);
525 refcount_init(&pa->ar_ref, 1);
527 bcopy(imgp->args->begin_argv, pa->ar_args, i);
528 KKASSERT(p->p_args == NULL);
535 * free various allocated resources
538 exec_unmap_first_page(imgp);
546 ++mycpu->gd_cnt.v_exec;
547 lwkt_reltoken(&p->p_token);
552 kfree(imgp->freepath, M_TEMP);
556 * we're done here, clear P_INEXEC if we were the ones that
557 * set it. Otherwise if vmspace_destroyed is still set we
558 * raced another thread and that thread is responsible for
561 if (imgp->vmspace_destroyed & 2)
562 p->p_flag &= ~P_INEXEC;
563 lwkt_reltoken(&p->p_token);
564 if (imgp->vmspace_destroyed) {
566 * Sorry, no more process anymore. exit gracefully.
567 * However we can't die right here, because our
568 * caller might have to clean up, so indicate a
569 * lethal error by returning -1.
578 * execve() system call.
583 sys_execve(struct execve_args *uap)
585 struct nlookupdata nd;
586 struct image_args args;
589 bzero(&args, sizeof(args));
592 error = nlookup_init(&nd, uap->fname, UIO_USERSPACE, NLC_FOLLOW);
594 error = exec_copyin_args(&args, uap->fname, PATH_USERSPACE,
595 uap->argv, uap->envv);
598 error = kern_execve(&nd, &args);
600 exec_free_args(&args);
603 /* We hit a lethal error condition. Let's die now. */
604 exit1(W_EXITCODE(0, SIGABRT));
610 * The syscall result is returned in registers to the new program.
611 * Linux will register %edx as an atexit function and we must be
612 * sure to set it to 0. XXX
615 uap->sysmsg_result64 = 0;
621 exec_map_page(struct image_params *imgp, vm_pindex_t pageno,
622 struct lwbuf **plwb, const char **pdata)
630 * The file has to be mappable.
632 if ((object = imgp->vp->v_object) == NULL)
635 if (pageno >= object->size)
638 m = vm_page_grab(object, pageno, VM_ALLOC_NORMAL | VM_ALLOC_RETRY);
640 lwkt_gettoken(&vm_token);
641 while ((m->valid & VM_PAGE_BITS_ALL) != VM_PAGE_BITS_ALL) {
645 * get_pages unbusies all the requested pages except the
646 * primary page (at index 0 in this case). The primary
647 * page may have been wired during the pagein (e.g. by
648 * the buffer cache) so vnode_pager_freepage() must be
649 * used to properly release it.
651 rv = vm_pager_get_page(object, &ma, 1);
652 m = vm_page_lookup(object, pageno);
654 if (rv != VM_PAGER_OK || m == NULL || m->valid == 0) {
656 vm_page_protect(m, VM_PROT_NONE);
657 vnode_pager_freepage(m);
659 lwkt_reltoken(&vm_token);
663 vm_page_hold(m); /* requires vm_token to be held */
664 vm_page_wakeup(m); /* unbusy the page */
665 lwkt_reltoken(&vm_token);
667 *plwb = lwbuf_alloc(m, *plwb);
668 *pdata = (void *)lwbuf_kva(*plwb);
674 exec_map_first_page(struct image_params *imgp)
679 exec_unmap_first_page(imgp);
681 imgp->firstpage = &imgp->firstpage_cache;
682 err = exec_map_page(imgp, 0, &imgp->firstpage, &imgp->image_header);
691 exec_unmap_page(struct lwbuf *lwb)
705 exec_unmap_first_page(struct image_params *imgp)
707 exec_unmap_page(imgp->firstpage);
708 imgp->firstpage = NULL;
709 imgp->image_header = NULL;
713 * Destroy old address space, and allocate a new stack
714 * The new stack is only SGROWSIZ large because it is grown
715 * automatically in trap.c.
717 * This is the point of no return.
720 exec_new_vmspace(struct image_params *imgp, struct vmspace *vmcopy)
722 struct vmspace *vmspace = imgp->proc->p_vmspace;
723 vm_offset_t stack_addr = USRSTACK - maxssiz;
729 * Indicate that we cannot gracefully error out any more, kill
730 * any other threads present, and set P_INEXEC to indicate that
731 * we are now messing with the process structure proper.
733 * If killalllwps() races return an error which coupled with
734 * vmspace_destroyed will cause us to exit. This is what we
735 * want since another thread is patiently waiting for us to exit
739 imgp->vmspace_destroyed = 1;
741 if (curthread->td_proc->p_nthreads > 1) {
742 error = killalllwps(1);
746 imgp->vmspace_destroyed |= 2; /* we are responsible for P_INEXEC */
747 p->p_flag |= P_INEXEC;
750 * Blow away entire process VM, if address space not shared,
751 * otherwise, create a new VM space so that other threads are
752 * not disrupted. If we are execing a resident vmspace we
753 * create a duplicate of it and remap the stack.
755 * The exitingcnt test is not strictly necessary but has been
756 * included for code sanity (to make the code more deterministic).
758 map = &vmspace->vm_map;
760 vmspace_exec(imgp->proc, vmcopy);
761 vmspace = imgp->proc->p_vmspace;
762 pmap_remove_pages(vmspace_pmap(vmspace), stack_addr, USRSTACK);
763 map = &vmspace->vm_map;
764 } else if (vmspace->vm_sysref.refcnt == 1 &&
765 vmspace->vm_exitingcnt == 0) {
767 if (vmspace->vm_upcalls)
768 upc_release(vmspace, ONLY_LWP_IN_PROC(imgp->proc));
769 pmap_remove_pages(vmspace_pmap(vmspace),
770 0, VM_MAX_USER_ADDRESS);
771 vm_map_remove(map, 0, VM_MAX_USER_ADDRESS);
773 vmspace_exec(imgp->proc, NULL);
774 vmspace = imgp->proc->p_vmspace;
775 map = &vmspace->vm_map;
778 /* Allocate a new stack */
779 error = vm_map_stack(&vmspace->vm_map, stack_addr, (vm_size_t)maxssiz,
780 0, VM_PROT_ALL, VM_PROT_ALL, 0);
784 /* vm_ssize and vm_maxsaddr are somewhat antiquated concepts in the
785 * VM_STACK case, but they are still used to monitor the size of the
786 * process stack so we can check the stack rlimit.
788 vmspace->vm_ssize = sgrowsiz >> PAGE_SHIFT;
789 vmspace->vm_maxsaddr = (char *)USRSTACK - maxssiz;
795 * Copy out argument and environment strings from the old process
796 * address space into the temporary string buffer.
799 exec_copyin_args(struct image_args *args, char *fname,
800 enum exec_path_segflg segflg, char **argv, char **envv)
806 args->buf = objcache_get(exec_objcache, M_WAITOK);
807 if (args->buf == NULL)
809 args->begin_argv = args->buf;
810 args->endp = args->begin_argv;
811 args->space = ARG_MAX;
813 args->fname = args->buf + ARG_MAX;
816 * Copy the file name.
818 if (segflg == PATH_SYSSPACE) {
819 error = copystr(fname, args->fname, PATH_MAX, &length);
820 } else if (segflg == PATH_USERSPACE) {
821 error = copyinstr(fname, args->fname, PATH_MAX, &length);
825 * Extract argument strings. argv may not be NULL. The argv
826 * array is terminated by a NULL entry. We special-case the
827 * situation where argv[0] is NULL by passing { filename, NULL }
828 * to the new program to guarentee that the interpreter knows what
829 * file to open in case we exec an interpreted file. Note that
830 * a NULL argv[0] terminates the argv[] array.
832 * XXX the special-casing of argv[0] is historical and needs to be
838 while ((argp = (caddr_t)(intptr_t)fuword(argv++)) != NULL) {
839 if (argp == (caddr_t)-1) {
843 error = copyinstr(argp, args->endp,
844 args->space, &length);
846 if (error == ENAMETOOLONG)
850 args->space -= length;
851 args->endp += length;
854 if (args->argc == 0 && error == 0) {
855 length = strlen(args->fname) + 1;
856 if (length > args->space) {
859 bcopy(args->fname, args->endp, length);
860 args->space -= length;
861 args->endp += length;
867 args->begin_envv = args->endp;
870 * extract environment strings. envv may be NULL.
872 if (envv && error == 0) {
873 while ((envp = (caddr_t) (intptr_t) fuword(envv++))) {
874 if (envp == (caddr_t) -1) {
878 error = copyinstr(envp, args->endp, args->space,
881 if (error == ENAMETOOLONG)
885 args->space -= length;
886 args->endp += length;
894 exec_free_args(struct image_args *args)
897 objcache_put(exec_objcache, args->buf);
903 * Copy strings out to the new process address space, constructing
904 * new arg and env vector tables. Return a pointer to the base
905 * so that it can be used as the initial stack pointer.
908 exec_copyout_strings(struct image_params *imgp)
910 int argc, envc, sgap;
912 char *stringp, *destp;
913 register_t *stack_base;
914 struct ps_strings *arginfo;
919 * Calculate string base and vector table pointers.
920 * Also deal with signal trampoline code for this exec type.
922 if (imgp->execpath != NULL && imgp->auxargs != NULL)
923 execpath_len = strlen(imgp->execpath) + 1;
926 arginfo = (struct ps_strings *)PS_STRINGS;
927 szsigcode = *(imgp->proc->p_sysent->sv_szsigcode);
928 if (stackgap_random != 0)
929 sgap = ALIGN(karc4random() & (stackgap_random - 1));
932 destp = (caddr_t)arginfo - szsigcode - SPARE_USRSPACE - sgap -
933 roundup(execpath_len, sizeof(char *)) -
934 roundup((ARG_MAX - imgp->args->space), sizeof(char *));
940 copyout(imgp->proc->p_sysent->sv_sigcode,
941 ((caddr_t)arginfo - szsigcode), szsigcode);
944 * Copy the image path for the rtld
946 if (execpath_len != 0) {
947 imgp->execpathp = (uintptr_t)arginfo
950 copyout(imgp->execpath, (void *)imgp->execpathp, execpath_len);
954 * If we have a valid auxargs ptr, prepare some room
957 * The '+ 2' is for the null pointers at the end of each of the
958 * arg and env vector sets, and 'AT_COUNT*2' is room for the
962 vectp = (char **)(destp - (imgp->args->argc +
963 imgp->args->envc + 2 + (AT_COUNT * 2) + execpath_len) *
966 vectp = (char **)(destp - (imgp->args->argc +
967 imgp->args->envc + 2) * sizeof(char*));
971 * NOTE: don't bother aligning the stack here for GCC 2.x, it will
972 * be done in crt1.o. Note that GCC 3.x aligns the stack in main.
976 * vectp also becomes our initial stack base
978 stack_base = (register_t *)vectp;
980 stringp = imgp->args->begin_argv;
981 argc = imgp->args->argc;
982 envc = imgp->args->envc;
985 * Copy out strings - arguments and environment.
987 copyout(stringp, destp, ARG_MAX - imgp->args->space);
990 * Fill in "ps_strings" struct for ps, w, etc.
992 suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp);
993 suword(&arginfo->ps_nargvstr, argc);
996 * Fill in argument portion of vector table.
998 for (; argc > 0; --argc) {
999 suword(vectp++, (long)(intptr_t)destp);
1000 while (*stringp++ != 0)
1005 /* a null vector table pointer separates the argp's from the envp's */
1008 suword(&arginfo->ps_envstr, (long)(intptr_t)vectp);
1009 suword(&arginfo->ps_nenvstr, envc);
1012 * Fill in environment portion of vector table.
1014 for (; envc > 0; --envc) {
1015 suword(vectp++, (long)(intptr_t)destp);
1016 while (*stringp++ != 0)
1021 /* end of vector table is a null pointer */
1024 return (stack_base);
1028 * Check permissions of file to execute.
1029 * Return 0 for success or error code on failure.
1032 exec_check_permissions(struct image_params *imgp, struct mount *topmnt)
1034 struct proc *p = imgp->proc;
1035 struct vnode *vp = imgp->vp;
1036 struct vattr *attr = imgp->attr;
1039 /* Get file attributes */
1040 error = VOP_GETATTR(vp, attr);
1045 * 1) Check if file execution is disabled for the filesystem that this
1047 * 2) Insure that at least one execute bit is on - otherwise root
1048 * will always succeed, and we don't want to happen unless the
1049 * file really is executable.
1050 * 3) Insure that the file is a regular file.
1052 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
1053 ((topmnt != NULL) && (topmnt->mnt_flag & MNT_NOEXEC)) ||
1054 ((attr->va_mode & 0111) == 0) ||
1055 (attr->va_type != VREG)) {
1060 * Zero length files can't be exec'd
1062 if (attr->va_size == 0)
1066 * Check for execute permission to file based on current credentials.
1068 error = VOP_EACCESS(vp, VEXEC, p->p_ucred);
1073 * Check number of open-for-writes on the file and deny execution
1076 if (vp->v_writecount)
1080 * Call filesystem specific open routine, which allows us to read,
1081 * write, and mmap the file. Without the VOP_OPEN we can only
1084 error = VOP_OPEN(vp, FREAD, p->p_ucred, NULL);
1092 * Exec handler registration
1095 exec_register(const struct execsw *execsw_arg)
1097 const struct execsw **es, **xs, **newexecsw;
1098 int count = 2; /* New slot and trailing NULL */
1101 for (es = execsw; *es; es++)
1103 newexecsw = kmalloc(count * sizeof(*es), M_TEMP, M_WAITOK);
1106 for (es = execsw; *es; es++)
1111 kfree(execsw, M_TEMP);
1117 exec_unregister(const struct execsw *execsw_arg)
1119 const struct execsw **es, **xs, **newexecsw;
1123 panic("unregister with no handlers left?");
1125 for (es = execsw; *es; es++) {
1126 if (*es == execsw_arg)
1131 for (es = execsw; *es; es++)
1132 if (*es != execsw_arg)
1134 newexecsw = kmalloc(count * sizeof(*es), M_TEMP, M_WAITOK);
1136 for (es = execsw; *es; es++)
1137 if (*es != execsw_arg)
1141 kfree(execsw, M_TEMP);