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
27 #include <sys/cdefs.h>
28 __FBSDID("$FreeBSD$");
30 #include "opt_hwpmc_hooks.h"
31 #include "opt_ktrace.h"
34 #include <sys/param.h>
35 #include <sys/systm.h>
36 #include <sys/eventhandler.h>
38 #include <sys/mutex.h>
39 #include <sys/sysproto.h>
40 #include <sys/signalvar.h>
41 #include <sys/kernel.h>
43 #include <sys/mount.h>
44 #include <sys/filedesc.h>
45 #include <sys/fcntl.h>
48 #include <sys/imgact.h>
49 #include <sys/imgact_elf.h>
51 #include <sys/malloc.h>
53 #include <sys/pioctl.h>
54 #include <sys/namei.h>
55 #include <sys/resourcevar.h>
56 #include <sys/sf_buf.h>
57 #include <sys/syscallsubr.h>
58 #include <sys/sysent.h>
60 #include <sys/sysctl.h>
61 #include <sys/vnode.h>
63 #include <sys/ktrace.h>
67 #include <vm/vm_param.h>
69 #include <vm/vm_page.h>
70 #include <vm/vm_map.h>
71 #include <vm/vm_kern.h>
72 #include <vm/vm_extern.h>
73 #include <vm/vm_object.h>
74 #include <vm/vm_pager.h>
77 #include <sys/pmckern.h>
80 #include <machine/reg.h>
82 MALLOC_DEFINE(M_PARGS, "proc-args", "Process arguments");
84 static int sysctl_kern_ps_strings(SYSCTL_HANDLER_ARGS);
85 static int sysctl_kern_usrstack(SYSCTL_HANDLER_ARGS);
86 static int sysctl_kern_stackprot(SYSCTL_HANDLER_ARGS);
87 static int do_execve(struct thread *td, struct image_args *args,
90 /* XXX This should be vm_size_t. */
91 SYSCTL_PROC(_kern, KERN_PS_STRINGS, ps_strings, CTLTYPE_ULONG|CTLFLAG_RD,
92 NULL, 0, sysctl_kern_ps_strings, "LU", "");
94 /* XXX This should be vm_size_t. */
95 SYSCTL_PROC(_kern, KERN_USRSTACK, usrstack, CTLTYPE_ULONG|CTLFLAG_RD,
96 NULL, 0, sysctl_kern_usrstack, "LU", "");
98 SYSCTL_PROC(_kern, OID_AUTO, stackprot, CTLTYPE_INT|CTLFLAG_RD,
99 NULL, 0, sysctl_kern_stackprot, "I", "");
101 u_long ps_arg_cache_limit = PAGE_SIZE / 16;
102 SYSCTL_ULONG(_kern, OID_AUTO, ps_arg_cache_limit, CTLFLAG_RW,
103 &ps_arg_cache_limit, 0, "");
106 sysctl_kern_ps_strings(SYSCTL_HANDLER_ARGS)
113 if (req->flags & SCTL_MASK32) {
115 val = (unsigned int)p->p_sysent->sv_psstrings;
116 error = SYSCTL_OUT(req, &val, sizeof(val));
119 error = SYSCTL_OUT(req, &p->p_sysent->sv_psstrings,
120 sizeof(p->p_sysent->sv_psstrings));
125 sysctl_kern_usrstack(SYSCTL_HANDLER_ARGS)
132 if (req->flags & SCTL_MASK32) {
134 val = (unsigned int)p->p_sysent->sv_usrstack;
135 error = SYSCTL_OUT(req, &val, sizeof(val));
138 error = SYSCTL_OUT(req, &p->p_sysent->sv_usrstack,
139 sizeof(p->p_sysent->sv_usrstack));
144 sysctl_kern_stackprot(SYSCTL_HANDLER_ARGS)
149 return (SYSCTL_OUT(req, &p->p_sysent->sv_stackprot,
150 sizeof(p->p_sysent->sv_stackprot)));
154 * Each of the items is a pointer to a `const struct execsw', hence the
155 * double pointer here.
157 static const struct execsw **execsw;
159 #ifndef _SYS_SYSPROTO_H_
173 struct execve_args /* {
180 struct image_args args;
182 error = exec_copyin_args(&args, uap->fname, UIO_USERSPACE,
183 uap->argv, uap->envv);
186 error = kern_execve(td, &args, NULL);
188 exec_free_args(&args);
193 #ifndef _SYS_SYSPROTO_H_
194 struct __mac_execve_args {
206 __mac_execve(td, uap)
208 struct __mac_execve_args /* {
217 struct image_args args;
219 error = exec_copyin_args(&args, uap->fname, UIO_USERSPACE,
220 uap->argv, uap->envv);
223 error = kern_execve(td, &args, uap->mac_p);
225 exec_free_args(&args);
234 kern_execve(td, args, mac_p)
236 struct image_args *args;
239 struct proc *p = td->td_proc;
242 if (p->p_flag & P_HADTHREADS) {
244 if (thread_single(SINGLE_BOUNDARY)) {
246 return (ERESTART); /* Try again later. */
251 error = do_execve(td, args, mac_p);
253 if (p->p_flag & P_HADTHREADS) {
256 * If success, we upgrade to SINGLE_EXIT state to
257 * force other threads to suicide.
260 thread_single(SINGLE_EXIT);
270 * In-kernel implementation of execve(). All arguments are assumed to be
271 * userspace pointers from the passed thread.
276 do_execve(td, args, mac_p)
278 struct image_args *args;
281 struct proc *p = td->td_proc;
282 struct nameidata nd, *ndp;
283 struct ucred *newcred = NULL, *oldcred;
284 struct uidinfo *euip;
285 register_t *stack_base;
287 struct image_params image_params, *imgp;
288 struct vattr atimeattr, attr;
289 int (*img_first)(struct image_params *);
290 struct pargs *oldargs = NULL, *newargs = NULL;
291 struct sigacts *oldsigacts, *newsigacts;
293 struct vnode *tracevp = NULL;
294 struct ucred *tracecred = NULL;
296 struct vnode *textvp = NULL;
297 int credential_changing;
301 struct label *interplabel = NULL;
305 struct pmckern_procexec pe;
309 imgp = &image_params;
312 * Lock the process and set the P_INEXEC flag to indicate that
313 * it should be left alone until we're done here. This is
314 * necessary to avoid race conditions - e.g. in ptrace() -
315 * that might allow a local user to illicitly obtain elevated
319 KASSERT((p->p_flag & P_INEXEC) == 0,
320 ("%s(): process already has P_INEXEC flag", __func__));
321 p->p_flag |= P_INEXEC;
325 * Initialize part of the common data
328 imgp->execlabel = NULL;
330 imgp->entry_addr = 0;
331 imgp->vmspace_destroyed = 0;
332 imgp->interpreted = 0;
333 imgp->interpreter_name = args->buf + PATH_MAX + ARG_MAX;
334 imgp->auxargs = NULL;
337 imgp->firstpage = NULL;
338 imgp->ps_strings = 0;
339 imgp->auxarg_size = 0;
343 error = mac_execve_enter(imgp, mac_p);
348 imgp->image_header = NULL;
351 * Translate the file name. namei() returns a vnode pointer
352 * in ni_vp amoung other things.
355 NDINIT(ndp, LOOKUP, ISOPEN | LOCKLEAF | FOLLOW | SAVENAME | MPSAFE,
356 UIO_SYSSPACE, args->fname, td);
363 vfslocked = NDHASGIANT(ndp);
364 imgp->vp = ndp->ni_vp;
367 * Check file permissions (also 'opens' file)
369 error = exec_check_permissions(imgp);
371 goto exec_fail_dealloc;
373 imgp->object = imgp->vp->v_object;
374 if (imgp->object != NULL)
375 vm_object_reference(imgp->object);
378 * Set VV_TEXT now so no one can write to the executable while we're
381 * Remember if this was set before and unset it in case this is not
382 * actually an executable image.
384 textset = imgp->vp->v_vflag & VV_TEXT;
385 imgp->vp->v_vflag |= VV_TEXT;
387 error = exec_map_first_page(imgp);
389 goto exec_fail_dealloc;
392 * If the current process has a special image activator it
393 * wants to try first, call it. For example, emulating shell
394 * scripts differently.
397 if ((img_first = imgp->proc->p_sysent->sv_imgact_try) != NULL)
398 error = img_first(imgp);
401 * Loop through the list of image activators, calling each one.
402 * An activator returns -1 if there is no match, 0 on success,
403 * and an error otherwise.
405 for (i = 0; error == -1 && execsw[i]; ++i) {
406 if (execsw[i]->ex_imgact == NULL ||
407 execsw[i]->ex_imgact == img_first) {
410 error = (*execsw[i]->ex_imgact)(imgp);
416 imgp->vp->v_vflag &= ~VV_TEXT;
419 goto exec_fail_dealloc;
423 * Special interpreter operation, cleanup and loop up to try to
424 * activate the interpreter.
426 if (imgp->interpreted) {
427 exec_unmap_first_page(imgp);
429 * VV_TEXT needs to be unset for scripts. There is a short
430 * period before we determine that something is a script where
431 * VV_TEXT will be set. The vnode lock is held over this
432 * entire period so nothing should illegitimately be blocked.
434 imgp->vp->v_vflag &= ~VV_TEXT;
435 /* free name buffer and old vnode */
436 NDFREE(ndp, NDF_ONLY_PNBUF);
438 interplabel = mac_vnode_label_alloc();
439 mac_copy_vnode_label(ndp->ni_vp->v_label, interplabel);
442 vm_object_deallocate(imgp->object);
444 VFS_UNLOCK_GIANT(vfslocked);
446 /* set new name to that of the interpreter */
447 NDINIT(ndp, LOOKUP, LOCKLEAF | FOLLOW | SAVENAME | MPSAFE,
448 UIO_SYSSPACE, imgp->interpreter_name, td);
453 * Copy out strings (args and env) and initialize stack base
455 if (p->p_sysent->sv_copyout_strings)
456 stack_base = (*p->p_sysent->sv_copyout_strings)(imgp);
458 stack_base = exec_copyout_strings(imgp);
461 * If custom stack fixup routine present for this process
462 * let it do the stack setup.
463 * Else stuff argument count as first item on stack
465 if (p->p_sysent->sv_fixup != NULL)
466 (*p->p_sysent->sv_fixup)(&stack_base, imgp);
468 suword(--stack_base, imgp->args->argc);
471 * For security and other reasons, the file descriptor table cannot
472 * be shared after an exec.
477 * Malloc things before we need locks.
480 euip = uifind(attr.va_uid);
481 i = imgp->args->begin_envv - imgp->args->begin_argv;
482 /* Cache arguments if they fit inside our allowance */
483 if (ps_arg_cache_limit >= i + sizeof(struct pargs)) {
484 newargs = pargs_alloc(i);
485 bcopy(imgp->args->begin_argv, newargs->ar_args, i);
488 /* close files on exec */
491 /* Get a reference to the vnode prior to locking the proc */
495 * For security and other reasons, signal handlers cannot
496 * be shared after an exec. The new process gets a copy of the old
497 * handlers. In execsigs(), the new process will have its signals
501 if (sigacts_shared(p->p_sigacts)) {
502 oldsigacts = p->p_sigacts;
504 newsigacts = sigacts_alloc();
505 sigacts_copy(newsigacts, oldsigacts);
507 p->p_sigacts = newsigacts;
514 /* reset caught signals */
517 /* name this process - nameiexec(p, ndp) */
518 len = min(ndp->ni_cnd.cn_namelen,MAXCOMLEN);
519 bcopy(ndp->ni_cnd.cn_nameptr, p->p_comm, len);
523 * mark as execed, wakeup the process that vforked (if any) and tell
524 * it that it now has its own resources back
527 if (p->p_pptr && (p->p_flag & P_PPWAIT)) {
528 p->p_flag &= ~P_PPWAIT;
533 * Implement image setuid/setgid.
535 * Don't honor setuid/setgid if the filesystem prohibits it or if
536 * the process is being traced.
538 * XXXMAC: For the time being, use NOSUID to also prohibit
539 * transitions on the file system.
541 oldcred = p->p_ucred;
542 credential_changing = 0;
543 credential_changing |= (attr.va_mode & VSUID) && oldcred->cr_uid !=
545 credential_changing |= (attr.va_mode & VSGID) && oldcred->cr_gid !=
548 will_transition = mac_execve_will_transition(oldcred, imgp->vp,
550 credential_changing |= will_transition;
553 if (credential_changing &&
554 (imgp->vp->v_mount->mnt_flag & MNT_NOSUID) == 0 &&
555 (p->p_flag & P_TRACED) == 0) {
557 * Turn off syscall tracing for set-id programs, except for
558 * root. Record any set-id flags first to make sure that
559 * we do not regain any tracing during a possible block.
563 if (p->p_tracevp != NULL && suser_cred(oldcred, SUSER_ALLOWJAIL)) {
564 mtx_lock(&ktrace_mtx);
566 tracevp = p->p_tracevp;
568 tracecred = p->p_tracecred;
569 p->p_tracecred = NULL;
570 mtx_unlock(&ktrace_mtx);
574 * Close any file descriptors 0..2 that reference procfs,
575 * then make sure file descriptors 0..2 are in use.
577 * setugidsafety() may call closef() and then pfind()
578 * which may grab the process lock.
579 * fdcheckstd() may call falloc() which may block to
580 * allocate memory, so temporarily drop the process lock.
584 error = fdcheckstd(td);
589 * Set the new credentials.
591 crcopy(newcred, oldcred);
592 if (attr.va_mode & VSUID)
593 change_euid(newcred, euip);
594 if (attr.va_mode & VSGID)
595 change_egid(newcred, attr.va_gid);
597 if (will_transition) {
598 mac_execve_transition(oldcred, newcred, imgp->vp,
603 * Implement correct POSIX saved-id behavior.
605 * XXXMAC: Note that the current logic will save the
606 * uid and gid if a MAC domain transition occurs, even
607 * though maybe it shouldn't.
609 change_svuid(newcred, newcred->cr_uid);
610 change_svgid(newcred, newcred->cr_gid);
611 p->p_ucred = newcred;
614 if (oldcred->cr_uid == oldcred->cr_ruid &&
615 oldcred->cr_gid == oldcred->cr_rgid)
616 p->p_flag &= ~P_SUGID;
618 * Implement correct POSIX saved-id behavior.
620 * XXX: It's not clear that the existing behavior is
621 * POSIX-compliant. A number of sources indicate that the
622 * saved uid/gid should only be updated if the new ruid is
623 * not equal to the old ruid, or the new euid is not equal
624 * to the old euid and the new euid is not equal to the old
625 * ruid. The FreeBSD code always updates the saved uid/gid.
626 * Also, this code uses the new (replaced) euid and egid as
627 * the source, which may or may not be the right ones to use.
629 if (oldcred->cr_svuid != oldcred->cr_uid ||
630 oldcred->cr_svgid != oldcred->cr_gid) {
631 crcopy(newcred, oldcred);
632 change_svuid(newcred, newcred->cr_uid);
633 change_svgid(newcred, newcred->cr_gid);
634 p->p_ucred = newcred;
640 * Store the vp for use in procfs. This vnode was referenced prior
641 * to locking the proc lock.
643 textvp = p->p_textvp;
644 p->p_textvp = ndp->ni_vp;
647 * Notify others that we exec'd, and clear the P_INEXEC flag
648 * as we're now a bona fide freshly-execed process.
650 KNOTE_LOCKED(&p->p_klist, NOTE_EXEC);
651 p->p_flag &= ~P_INEXEC;
654 * If tracing the process, trap to debugger so breakpoints
655 * can be set before the program executes.
656 * Use tdsignal to deliver signal to current thread, use
657 * psignal may cause the signal to be delivered to wrong thread
658 * because that thread will exit, remember we are going to enter
659 * single thread mode.
661 if (p->p_flag & P_TRACED)
662 tdsignal(td, SIGTRAP, SIGTARGET_TD);
664 /* clear "fork but no exec" flag, as we _are_ execing */
665 p->p_acflag &= ~AFORK;
668 * Free any previous argument cache and it with
669 * the new argument cache, if any.
677 * Check if system-wide sampling is in effect or if the
678 * current process is using PMCs. If so, do exec() time
679 * processing. This processing needs to happen AFTER the
680 * P_INEXEC flag is cleared.
682 * The proc lock needs to be released before taking the PMC
685 if (PMC_SYSTEM_SAMPLING_ACTIVE() || PMC_PROC_IS_USING_PMCS(p)) {
687 pe.pm_credentialschanged = credential_changing;
688 pe.pm_entryaddr = imgp->entry_addr;
690 PMC_CALL_HOOK_X(td, PMC_FN_PROCESS_EXEC, (void *) &pe);
693 #else /* !HWPMC_HOOKS */
697 /* Set values passed into the program in registers. */
698 if (p->p_sysent->sv_setregs)
699 (*p->p_sysent->sv_setregs)(td, imgp->entry_addr,
700 (u_long)(uintptr_t)stack_base, imgp->ps_strings);
702 exec_setregs(td, imgp->entry_addr,
703 (u_long)(uintptr_t)stack_base, imgp->ps_strings);
706 * Here we should update the access time of the file. This must
707 * be implemented by the underlying filesystem in the same way as
708 * access timestamps for a VOP_READ() because we want to avoid
709 * blocking and/or I/O, and have not called vn_start_write().
711 if ((imgp->vp->v_mount->mnt_flag & (MNT_NOATIME | MNT_RDONLY)) == 0) {
712 VATTR_NULL(&atimeattr);
713 atimeattr.va_vaflags |= VA_EXECVE_ATIME;
714 (void)VOP_SETATTR(imgp->vp, &atimeattr, td->td_ucred, td);
719 * Free any resources malloc'd earlier that we didn't use.
727 * Handle deferred decrement of ref counts.
731 if (ndp->ni_vp && error != 0)
736 if (tracecred != NULL)
743 if (oldsigacts != NULL)
744 sigacts_free(oldsigacts);
749 * free various allocated resources
751 if (imgp->firstpage != NULL)
752 exec_unmap_first_page(imgp);
754 if (imgp->vp != NULL) {
755 NDFREE(ndp, NDF_ONLY_PNBUF);
759 if (imgp->object != NULL)
760 vm_object_deallocate(imgp->object);
764 * Stop the process here if its stop event mask has
765 * the S_EXEC bit set.
767 STOPEVENT(p, S_EXEC, 0);
772 /* we're done here, clear P_INEXEC */
774 p->p_flag &= ~P_INEXEC;
777 if (imgp->vmspace_destroyed) {
778 /* sorry, no more process anymore. exit gracefully */
780 mac_execve_exit(imgp);
781 if (interplabel != NULL)
782 mac_vnode_label_free(interplabel);
784 VFS_UNLOCK_GIANT(vfslocked);
785 exit1(td, W_EXITCODE(0, SIGABRT));
791 mac_execve_exit(imgp);
792 if (interplabel != NULL)
793 mac_vnode_label_free(interplabel);
795 VFS_UNLOCK_GIANT(vfslocked);
800 exec_map_first_page(imgp)
801 struct image_params *imgp;
805 vm_page_t ma[VM_INITIAL_PAGEIN];
808 if (imgp->firstpage != NULL)
809 exec_unmap_first_page(imgp);
811 object = imgp->vp->v_object;
814 VM_OBJECT_LOCK(object);
815 ma[0] = vm_page_grab(object, 0, VM_ALLOC_NORMAL | VM_ALLOC_RETRY);
816 if ((ma[0]->valid & VM_PAGE_BITS_ALL) != VM_PAGE_BITS_ALL) {
817 initial_pagein = VM_INITIAL_PAGEIN;
818 if (initial_pagein > object->size)
819 initial_pagein = object->size;
820 for (i = 1; i < initial_pagein; i++) {
821 if ((ma[i] = vm_page_lookup(object, i)) != NULL) {
824 vm_page_lock_queues();
825 if ((ma[i]->flags & PG_BUSY) || ma[i]->busy) {
826 vm_page_unlock_queues();
830 vm_page_unlock_queues();
832 ma[i] = vm_page_alloc(object, i,
839 rv = vm_pager_get_pages(object, ma, initial_pagein, 0);
840 ma[0] = vm_page_lookup(object, 0);
841 if ((rv != VM_PAGER_OK) || (ma[0] == NULL) ||
842 (ma[0]->valid == 0)) {
844 vm_page_lock_queues();
845 pmap_remove_all(ma[0]);
847 vm_page_unlock_queues();
849 VM_OBJECT_UNLOCK(object);
853 vm_page_lock_queues();
855 vm_page_wakeup(ma[0]);
856 vm_page_unlock_queues();
857 VM_OBJECT_UNLOCK(object);
859 imgp->firstpage = sf_buf_alloc(ma[0], 0);
860 imgp->image_header = (char *)sf_buf_kva(imgp->firstpage);
866 exec_unmap_first_page(imgp)
867 struct image_params *imgp;
871 if (imgp->firstpage != NULL) {
872 m = sf_buf_page(imgp->firstpage);
873 sf_buf_free(imgp->firstpage);
874 imgp->firstpage = NULL;
875 vm_page_lock_queues();
877 vm_page_unlock_queues();
882 * Destroy old address space, and allocate a new stack
883 * The new stack is only SGROWSIZ large because it is grown
884 * automatically in trap.c.
887 exec_new_vmspace(imgp, sv)
888 struct image_params *imgp;
889 struct sysentvec *sv;
892 struct proc *p = imgp->proc;
893 struct vmspace *vmspace = p->p_vmspace;
894 vm_offset_t stack_addr;
897 imgp->vmspace_destroyed = 1;
899 /* Called with Giant held, do not depend on it! */
900 EVENTHANDLER_INVOKE(process_exec, p);
903 * Here is as good a place as any to do any resource limit cleanups.
904 * This is needed if a 64 bit binary exec's a 32 bit binary - the
905 * data size limit may need to be changed to a value that makes
906 * sense for the 32 bit binary.
908 if (sv->sv_fixlimits != NULL)
909 sv->sv_fixlimits(imgp);
912 * Blow away entire process VM, if address space not shared,
913 * otherwise, create a new VM space so that other threads are
916 map = &vmspace->vm_map;
917 if (vmspace->vm_refcnt == 1 && vm_map_min(map) == sv->sv_minuser &&
918 vm_map_max(map) == sv->sv_maxuser) {
920 pmap_remove_pages(vmspace_pmap(vmspace), vm_map_min(map),
922 vm_map_remove(map, vm_map_min(map), vm_map_max(map));
924 vmspace_exec(p, sv->sv_minuser, sv->sv_maxuser);
925 vmspace = p->p_vmspace;
926 map = &vmspace->vm_map;
929 /* Allocate a new stack */
930 stack_addr = sv->sv_usrstack - maxssiz;
931 error = vm_map_stack(map, stack_addr, (vm_size_t)maxssiz,
932 sv->sv_stackprot, VM_PROT_ALL, MAP_STACK_GROWS_DOWN);
937 /* Allocate a new register stack */
938 stack_addr = IA64_BACKINGSTORE;
939 error = vm_map_stack(map, stack_addr, (vm_size_t)maxssiz,
940 sv->sv_stackprot, VM_PROT_ALL, MAP_STACK_GROWS_UP);
945 /* vm_ssize and vm_maxsaddr are somewhat antiquated concepts in the
946 * VM_STACK case, but they are still used to monitor the size of the
947 * process stack so we can check the stack rlimit.
949 vmspace->vm_ssize = sgrowsiz >> PAGE_SHIFT;
950 vmspace->vm_maxsaddr = (char *)sv->sv_usrstack - maxssiz;
956 * Copy out argument and environment strings from the old process
957 * address space into the temporary string buffer.
960 exec_copyin_args(struct image_args *args, char *fname,
961 enum uio_seg segflg, char **argv, char **envv)
969 bzero(args, sizeof(*args));
973 * Allocate temporary demand zeroed space for argument and
974 * environment strings:
976 * o ARG_MAX for argument and environment;
977 * o MAXSHELLCMDLEN for the name of interpreters.
979 args->buf = (char *) kmem_alloc_wait(exec_map,
980 PATH_MAX + ARG_MAX + MAXSHELLCMDLEN);
981 if (args->buf == NULL)
983 args->begin_argv = args->buf;
984 args->endp = args->begin_argv;
985 args->stringspace = ARG_MAX;
987 args->fname = args->buf + ARG_MAX;
990 * Copy the file name.
992 error = (segflg == UIO_SYSSPACE) ?
993 copystr(fname, args->fname, PATH_MAX, &length) :
994 copyinstr(fname, args->fname, PATH_MAX, &length);
999 * extract arguments first
1001 while ((argp = (caddr_t) (intptr_t) fuword(argv++))) {
1002 if (argp == (caddr_t) -1)
1004 if ((error = copyinstr(argp, args->endp,
1005 args->stringspace, &length))) {
1006 if (error == ENAMETOOLONG)
1010 args->stringspace -= length;
1011 args->endp += length;
1015 args->begin_envv = args->endp;
1018 * extract environment strings
1021 while ((envp = (caddr_t)(intptr_t)fuword(envv++))) {
1022 if (envp == (caddr_t)-1)
1024 if ((error = copyinstr(envp, args->endp,
1025 args->stringspace, &length))) {
1026 if (error == ENAMETOOLONG)
1030 args->stringspace -= length;
1031 args->endp += length;
1040 exec_free_args(struct image_args *args)
1044 kmem_free_wakeup(exec_map, (vm_offset_t)args->buf,
1045 PATH_MAX + ARG_MAX + MAXSHELLCMDLEN);
1051 * Copy strings out to the new process address space, constructing
1052 * new arg and env vector tables. Return a pointer to the base
1053 * so that it can be used as the initial stack pointer.
1056 exec_copyout_strings(imgp)
1057 struct image_params *imgp;
1061 char *stringp, *destp;
1062 register_t *stack_base;
1063 struct ps_strings *arginfo;
1068 * Calculate string base and vector table pointers.
1069 * Also deal with signal trampoline code for this exec type.
1073 arginfo = (struct ps_strings *)p->p_sysent->sv_psstrings;
1074 if (p->p_sysent->sv_szsigcode != NULL)
1075 szsigcode = *(p->p_sysent->sv_szsigcode);
1076 destp = (caddr_t)arginfo - szsigcode - SPARE_USRSPACE -
1077 roundup((ARG_MAX - imgp->args->stringspace), sizeof(char *));
1083 copyout(p->p_sysent->sv_sigcode, ((caddr_t)arginfo -
1084 szsigcode), szsigcode);
1087 * If we have a valid auxargs ptr, prepare some room
1090 if (imgp->auxargs) {
1092 * 'AT_COUNT*2' is size for the ELF Auxargs data. This is for
1093 * lower compatibility.
1095 imgp->auxarg_size = (imgp->auxarg_size) ? imgp->auxarg_size :
1098 * The '+ 2' is for the null pointers at the end of each of
1099 * the arg and env vector sets,and imgp->auxarg_size is room
1100 * for argument of Runtime loader.
1102 vectp = (char **)(destp - (imgp->args->argc +
1103 imgp->args->envc + 2 + imgp->auxarg_size) *
1108 * The '+ 2' is for the null pointers at the end of each of
1109 * the arg and env vector sets
1111 vectp = (char **)(destp - (imgp->args->argc + imgp->args->envc + 2) *
1116 * vectp also becomes our initial stack base
1118 stack_base = (register_t *)vectp;
1120 stringp = imgp->args->begin_argv;
1121 argc = imgp->args->argc;
1122 envc = imgp->args->envc;
1125 * Copy out strings - arguments and environment.
1127 copyout(stringp, destp, ARG_MAX - imgp->args->stringspace);
1130 * Fill in "ps_strings" struct for ps, w, etc.
1132 suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp);
1133 suword(&arginfo->ps_nargvstr, argc);
1136 * Fill in argument portion of vector table.
1138 for (; argc > 0; --argc) {
1139 suword(vectp++, (long)(intptr_t)destp);
1140 while (*stringp++ != 0)
1145 /* a null vector table pointer separates the argp's from the envp's */
1148 suword(&arginfo->ps_envstr, (long)(intptr_t)vectp);
1149 suword(&arginfo->ps_nenvstr, envc);
1152 * Fill in environment portion of vector table.
1154 for (; envc > 0; --envc) {
1155 suword(vectp++, (long)(intptr_t)destp);
1156 while (*stringp++ != 0)
1161 /* end of vector table is a null pointer */
1164 return (stack_base);
1168 * Check permissions of file to execute.
1169 * Called with imgp->vp locked.
1170 * Return 0 for success or error code on failure.
1173 exec_check_permissions(imgp)
1174 struct image_params *imgp;
1176 struct vnode *vp = imgp->vp;
1177 struct vattr *attr = imgp->attr;
1181 td = curthread; /* XXXKSE */
1183 /* Get file attributes */
1184 error = VOP_GETATTR(vp, attr, td->td_ucred, td);
1189 error = mac_check_vnode_exec(td->td_ucred, imgp->vp, imgp);
1195 * 1) Check if file execution is disabled for the filesystem that this
1197 * 2) Insure that at least one execute bit is on - otherwise root
1198 * will always succeed, and we don't want to happen unless the
1199 * file really is executable.
1200 * 3) Insure that the file is a regular file.
1202 if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
1203 ((attr->va_mode & 0111) == 0) ||
1204 (attr->va_type != VREG))
1208 * Zero length files can't be exec'd
1210 if (attr->va_size == 0)
1214 * Check for execute permission to file based on current credentials.
1216 error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
1221 * Check number of open-for-writes on the file and deny execution
1224 if (vp->v_writecount)
1228 * Call filesystem specific open routine (which does nothing in the
1231 error = VOP_OPEN(vp, FREAD, td->td_ucred, td, -1);
1236 * Exec handler registration
1239 exec_register(execsw_arg)
1240 const struct execsw *execsw_arg;
1242 const struct execsw **es, **xs, **newexecsw;
1243 int count = 2; /* New slot and trailing NULL */
1246 for (es = execsw; *es; es++)
1248 newexecsw = malloc(count * sizeof(*es), M_TEMP, M_WAITOK);
1249 if (newexecsw == NULL)
1253 for (es = execsw; *es; es++)
1258 free(execsw, M_TEMP);
1264 exec_unregister(execsw_arg)
1265 const struct execsw *execsw_arg;
1267 const struct execsw **es, **xs, **newexecsw;
1271 panic("unregister with no handlers left?\n");
1273 for (es = execsw; *es; es++) {
1274 if (*es == execsw_arg)
1279 for (es = execsw; *es; es++)
1280 if (*es != execsw_arg)
1282 newexecsw = malloc(count * sizeof(*es), M_TEMP, M_WAITOK);
1283 if (newexecsw == NULL)
1286 for (es = execsw; *es; es++)
1287 if (*es != execsw_arg)
1291 free(execsw, M_TEMP);