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>
76 MALLOC_DEFINE(M_PARGS
, "proc-args", "Process arguments");
77 MALLOC_DEFINE(M_EXECARGS
, "exec-args", "Exec arguments");
79 static register_t
*exec_copyout_strings (struct image_params
*);
81 /* XXX This should be vm_size_t. */
82 static u_long ps_strings
= PS_STRINGS
;
83 SYSCTL_ULONG(_kern
, KERN_PS_STRINGS
, ps_strings
, CTLFLAG_RD
, &ps_strings
, 0, "");
85 /* XXX This should be vm_size_t. */
86 static u_long usrstack
= USRSTACK
;
87 SYSCTL_ULONG(_kern
, KERN_USRSTACK
, usrstack
, CTLFLAG_RD
, &usrstack
, 0, "");
89 u_long ps_arg_cache_limit
= PAGE_SIZE
/ 16;
90 SYSCTL_LONG(_kern
, OID_AUTO
, ps_arg_cache_limit
, CTLFLAG_RW
,
91 &ps_arg_cache_limit
, 0, "");
94 SYSCTL_INT(_kern
, OID_AUTO
, ps_argsopen
, CTLFLAG_RW
, &ps_argsopen
, 0, "");
96 void print_execve_args(struct image_args
*args
);
97 int debug_execve_args
= 0;
98 SYSCTL_INT(_kern
, OID_AUTO
, debug_execve_args
, CTLFLAG_RW
, &debug_execve_args
,
102 * Exec arguments object cache
104 static struct objcache
*exec_objcache
;
108 exec_objcache_init(void *arg __unused
)
112 cluster_limit
= 16; /* up to this many objects */
113 exec_objcache
= objcache_create_mbacked(
114 M_EXECARGS
, PATH_MAX
+ ARG_MAX
,
116 2, /* minimal magazine capacity */
119 SYSINIT(exec_objcache
, SI_BOOT2_MACHDEP
, SI_ORDER_ANY
, exec_objcache_init
, 0);
122 * stackgap_random specifies if the stackgap should have a random size added
123 * to it. It must be a power of 2. If non-zero, the stack gap will be
124 * calculated as: ALIGN(karc4random() & (stackgap_random - 1)).
126 static int stackgap_random
= 1024;
128 sysctl_kern_stackgap(SYSCTL_HANDLER_ARGS
)
131 new_val
= stackgap_random
;
132 error
= sysctl_handle_int(oidp
, &new_val
, 0, req
);
133 if (error
!= 0 || req
->newptr
== NULL
)
135 if ((new_val
< 0) || (new_val
> 16 * PAGE_SIZE
) || ! powerof2(new_val
))
137 stackgap_random
= new_val
;
142 SYSCTL_PROC(_kern
, OID_AUTO
, stackgap_random
, CTLFLAG_RW
|CTLTYPE_UINT
,
143 0, 0, sysctl_kern_stackgap
, "IU", "Max random stack gap (power of 2)");
146 print_execve_args(struct image_args
*args
)
151 cp
= args
->begin_argv
;
152 for (ndx
= 0; ndx
< args
->argc
; ndx
++) {
153 kprintf("\targv[%d]: %s\n", ndx
, cp
);
154 while (*cp
++ != '\0');
156 for (ndx
= 0; ndx
< args
->envc
; ndx
++) {
157 kprintf("\tenvv[%d]: %s\n", ndx
, cp
);
158 while (*cp
++ != '\0');
163 * Each of the items is a pointer to a `const struct execsw', hence the
164 * double pointer here.
166 static const struct execsw
**execsw
;
169 * Replace current vmspace with a new binary.
170 * Returns 0 on success, > 0 on recoverable error (use as errno).
171 * Returns -1 on lethal error which demands killing of the current
175 kern_execve(struct nlookupdata
*nd
, struct image_args
*args
)
177 struct thread
*td
= curthread
;
178 struct lwp
*lp
= td
->td_lwp
;
179 struct proc
*p
= td
->td_proc
;
180 register_t
*stack_base
;
182 struct image_params image_params
, *imgp
;
184 int (*img_first
) (struct image_params
*);
186 if (debug_execve_args
) {
187 kprintf("%s()\n", __func__
);
188 print_execve_args(args
);
192 imgp
= &image_params
;
195 * NOTE: P_INEXEC is handled by exec_new_vmspace() now. We make
196 * no modifications to the process at all until we get there.
198 * Note that multiple threads may be trying to exec at the same
199 * time. exec_new_vmspace() handles that too.
203 * Initialize part of the common data
208 imgp
->entry_addr
= 0;
210 imgp
->vmspace_destroyed
= 0;
211 imgp
->interpreted
= 0;
212 imgp
->interpreter_name
[0] = 0;
213 imgp
->auxargs
= NULL
;
215 imgp
->firstpage
= NULL
;
216 imgp
->ps_strings
= 0;
217 imgp
->image_header
= NULL
;
222 * Translate the file name to a vnode. Unlock the cache entry to
223 * improve parallelism for programs exec'd in parallel.
225 if ((error
= nlookup(nd
)) != 0)
227 error
= cache_vget(&nd
->nl_nch
, nd
->nl_cred
, LK_EXCLUSIVE
, &imgp
->vp
);
228 KKASSERT(nd
->nl_flags
& NLC_NCPISLOCKED
);
229 nd
->nl_flags
&= ~NLC_NCPISLOCKED
;
230 cache_unlock(&nd
->nl_nch
);
235 * Check file permissions (also 'opens' file)
237 error
= exec_check_permissions(imgp
);
240 goto exec_fail_dealloc
;
243 error
= exec_map_first_page(imgp
);
246 goto exec_fail_dealloc
;
248 if (debug_execve_args
&& imgp
->interpreted
) {
249 kprintf(" target is interpreted -- recursive pass\n");
250 kprintf(" interpreter: %s\n", imgp
->interpreter_name
);
251 print_execve_args(args
);
255 * If the current process has a special image activator it
256 * wants to try first, call it. For example, emulating shell
257 * scripts differently.
260 if ((img_first
= imgp
->proc
->p_sysent
->sv_imgact_try
) != NULL
)
261 error
= img_first(imgp
);
264 * If the vnode has a registered vmspace, exec the vmspace
266 if (error
== -1 && imgp
->vp
->v_resident
) {
267 error
= exec_resident_imgact(imgp
);
271 * Loop through the list of image activators, calling each one.
272 * An activator returns -1 if there is no match, 0 on success,
273 * and an error otherwise.
275 for (i
= 0; error
== -1 && execsw
[i
]; ++i
) {
276 if (execsw
[i
]->ex_imgact
== NULL
||
277 execsw
[i
]->ex_imgact
== img_first
) {
280 error
= (*execsw
[i
]->ex_imgact
)(imgp
);
286 goto exec_fail_dealloc
;
290 * Special interpreter operation, cleanup and loop up to try to
291 * activate the interpreter.
293 if (imgp
->interpreted
) {
294 exec_unmap_first_page(imgp
);
298 error
= nlookup_init(nd
, imgp
->interpreter_name
, UIO_SYSSPACE
,
306 * Copy out strings (args and env) and initialize stack base
308 stack_base
= exec_copyout_strings(imgp
);
309 p
->p_vmspace
->vm_minsaddr
= (char *)stack_base
;
312 * If custom stack fixup routine present for this process
313 * let it do the stack setup. If we are running a resident
314 * image there is no auxinfo or other image activator context
315 * so don't try to add fixups to the stack.
317 * Else stuff argument count as first item on stack
319 if (p
->p_sysent
->sv_fixup
&& imgp
->resident
== 0)
320 (*p
->p_sysent
->sv_fixup
)(&stack_base
, imgp
);
322 suword(--stack_base
, imgp
->args
->argc
);
325 * For security and other reasons, the file descriptor table cannot
326 * be shared after an exec.
328 if (p
->p_fd
->fd_refcnt
> 1) {
329 struct filedesc
*tmp
;
337 * For security and other reasons, signal handlers cannot
338 * be shared after an exec. The new proces gets a copy of the old
339 * handlers. In execsigs(), the new process will have its signals
342 if (p
->p_sigacts
->ps_refcnt
> 1) {
343 struct sigacts
*newsigacts
;
345 newsigacts
= (struct sigacts
*)kmalloc(sizeof(*newsigacts
),
346 M_SUBPROC
, M_WAITOK
);
347 bcopy(p
->p_sigacts
, newsigacts
, sizeof(*newsigacts
));
348 p
->p_sigacts
->ps_refcnt
--;
349 p
->p_sigacts
= newsigacts
;
350 p
->p_sigacts
->ps_refcnt
= 1;
354 * For security and other reasons virtual kernels cannot be
355 * inherited by an exec. This also allows a virtual kernel
356 * to fork/exec unrelated applications.
364 /* close files on exec */
367 /* reset caught signals */
370 /* name this process - nameiexec(p, ndp) */
371 len
= min(nd
->nl_nch
.ncp
->nc_nlen
, MAXCOMLEN
);
372 bcopy(nd
->nl_nch
.ncp
->nc_name
, p
->p_comm
, len
);
374 bcopy(p
->p_comm
, lp
->lwp_thread
->td_comm
, MAXCOMLEN
+1);
377 * mark as execed, wakeup the process that vforked (if any) and tell
378 * it that it now has its own resources back
381 if (p
->p_pptr
&& (p
->p_flag
& P_PPWAIT
)) {
382 p
->p_flag
&= ~P_PPWAIT
;
383 wakeup((caddr_t
)p
->p_pptr
);
387 * Implement image setuid/setgid.
389 * Don't honor setuid/setgid if the filesystem prohibits it or if
390 * the process is being traced.
392 if ((((attr
.va_mode
& VSUID
) && p
->p_ucred
->cr_uid
!= attr
.va_uid
) ||
393 ((attr
.va_mode
& VSGID
) && p
->p_ucred
->cr_gid
!= attr
.va_gid
)) &&
394 (imgp
->vp
->v_mount
->mnt_flag
& MNT_NOSUID
) == 0 &&
395 (p
->p_flag
& P_TRACED
) == 0) {
397 * Turn off syscall tracing for set-id programs, except for
398 * root. Record any set-id flags first to make sure that
399 * we do not regain any tracing during a possible block.
402 if (p
->p_tracenode
&& priv_check(td
, PRIV_ROOT
) != 0) {
403 ktrdestroy(&p
->p_tracenode
);
406 /* Close any file descriptors 0..2 that reference procfs */
408 /* Make sure file descriptors 0..2 are in use. */
409 error
= fdcheckstd(p
);
411 goto exec_fail_dealloc
;
413 * Set the new credentials.
416 if (attr
.va_mode
& VSUID
)
417 change_euid(attr
.va_uid
);
418 if (attr
.va_mode
& VSGID
)
419 p
->p_ucred
->cr_gid
= attr
.va_gid
;
422 * Clear local varsym variables
424 varsymset_clean(&p
->p_varsymset
);
426 if (p
->p_ucred
->cr_uid
== p
->p_ucred
->cr_ruid
&&
427 p
->p_ucred
->cr_gid
== p
->p_ucred
->cr_rgid
)
428 p
->p_flag
&= ~P_SUGID
;
432 * Implement correct POSIX saved-id behavior.
434 if (p
->p_ucred
->cr_svuid
!= p
->p_ucred
->cr_uid
||
435 p
->p_ucred
->cr_svgid
!= p
->p_ucred
->cr_gid
) {
437 p
->p_ucred
->cr_svuid
= p
->p_ucred
->cr_uid
;
438 p
->p_ucred
->cr_svgid
= p
->p_ucred
->cr_gid
;
442 * Store the vp for use in procfs
444 if (p
->p_textvp
) /* release old reference */
446 p
->p_textvp
= imgp
->vp
;
450 * Notify others that we exec'd, and clear the P_INEXEC flag
451 * as we're now a bona fide freshly-execed process.
453 KNOTE(&p
->p_klist
, NOTE_EXEC
);
454 p
->p_flag
&= ~P_INEXEC
;
457 * If tracing the process, trap to debugger so breakpoints
458 * can be set before the program executes.
460 STOPEVENT(p
, S_EXEC
, 0);
462 if (p
->p_flag
& P_TRACED
)
465 /* clear "fork but no exec" flag, as we _are_ execing */
466 p
->p_acflag
&= ~AFORK
;
468 /* Set values passed into the program in registers. */
469 exec_setregs(imgp
->entry_addr
, (u_long
)(uintptr_t)stack_base
,
472 /* Set the access time on the vnode */
473 vn_mark_atime(imgp
->vp
, td
);
475 /* Free any previous argument cache */
476 if (p
->p_args
&& --p
->p_args
->ar_ref
== 0)
477 FREE(p
->p_args
, M_PARGS
);
480 /* Cache arguments if they fit inside our allowance */
481 i
= imgp
->args
->begin_envv
- imgp
->args
->begin_argv
;
482 if (ps_arg_cache_limit
>= i
+ sizeof(struct pargs
)) {
483 MALLOC(p
->p_args
, struct pargs
*, sizeof(struct pargs
) + i
,
485 p
->p_args
->ar_ref
= 1;
486 p
->p_args
->ar_length
= i
;
487 bcopy(imgp
->args
->begin_argv
, p
->p_args
->ar_args
, i
);
493 * free various allocated resources
496 exec_unmap_first_page(imgp
);
504 ++mycpu
->gd_cnt
.v_exec
;
510 * we're done here, clear P_INEXEC if we were the ones that
511 * set it. Otherwise if vmspace_destroyed is still set we
512 * raced another thread and that thread is responsible for
515 if (imgp
->vmspace_destroyed
& 2)
516 p
->p_flag
&= ~P_INEXEC
;
517 if (imgp
->vmspace_destroyed
) {
519 * Sorry, no more process anymore. exit gracefully.
520 * However we can't die right here, because our
521 * caller might have to clean up, so indicate a
522 * lethal error by returning -1.
531 * execve() system call.
534 sys_execve(struct execve_args
*uap
)
536 struct nlookupdata nd
;
537 struct image_args args
;
540 error
= nlookup_init(&nd
, uap
->fname
, UIO_USERSPACE
, NLC_FOLLOW
);
542 error
= exec_copyin_args(&args
, uap
->fname
, PATH_USERSPACE
,
543 uap
->argv
, uap
->envv
);
546 error
= kern_execve(&nd
, &args
);
548 exec_free_args(&args
);
551 /* We hit a lethal error condition. Let's die now. */
552 exit1(W_EXITCODE(0, SIGABRT
));
557 * The syscall result is returned in registers to the new program.
558 * Linux will register %edx as an atexit function and we must be
559 * sure to set it to 0. XXX
562 uap
->sysmsg_result64
= 0;
568 exec_map_first_page(struct image_params
*imgp
)
572 vm_page_t ma
[VM_INITIAL_PAGEIN
];
577 exec_unmap_first_page(imgp
);
580 * The file has to be mappable.
582 if ((object
= imgp
->vp
->v_object
) == NULL
)
586 * We shouldn't need protection for vm_page_grab() but we certainly
587 * need it for the lookup loop below (lookup/busy race), since
588 * an interrupt can unbusy and free the page before our busy check.
591 m
= vm_page_grab(object
, 0, VM_ALLOC_NORMAL
| VM_ALLOC_RETRY
);
593 if ((m
->valid
& VM_PAGE_BITS_ALL
) != VM_PAGE_BITS_ALL
) {
595 initial_pagein
= VM_INITIAL_PAGEIN
;
596 if (initial_pagein
> object
->size
)
597 initial_pagein
= object
->size
;
598 for (i
= 1; i
< initial_pagein
; i
++) {
599 if ((m
= vm_page_lookup(object
, i
)) != NULL
) {
600 if ((m
->flags
& PG_BUSY
) || m
->busy
)
606 m
= vm_page_alloc(object
, i
, VM_ALLOC_NORMAL
);
615 * get_pages unbusies all the requested pages except the
616 * primary page (at index 0 in this case). The primary
617 * page may have been wired during the pagein (e.g. by
618 * the buffer cache) so vnode_pager_freepage() must be
619 * used to properly release it.
621 rv
= vm_pager_get_pages(object
, ma
, initial_pagein
, 0);
622 m
= vm_page_lookup(object
, 0);
624 if (rv
!= VM_PAGER_OK
|| m
== NULL
|| m
->valid
== 0) {
626 vm_page_protect(m
, VM_PROT_NONE
);
627 vnode_pager_freepage(m
);
634 vm_page_wakeup(m
); /* unbusy the page */
637 imgp
->firstpage
= sf_buf_alloc(m
, SFB_CPUPRIVATE
);
638 imgp
->image_header
= (void *)sf_buf_kva(imgp
->firstpage
);
644 exec_unmap_first_page(struct image_params
*imgp
)
649 if (imgp
->firstpage
!= NULL
) {
650 m
= sf_buf_page(imgp
->firstpage
);
651 sf_buf_free(imgp
->firstpage
);
652 imgp
->firstpage
= NULL
;
653 imgp
->image_header
= NULL
;
660 * Destroy old address space, and allocate a new stack
661 * The new stack is only SGROWSIZ large because it is grown
662 * automatically in trap.c.
664 * This is the point of no return.
667 exec_new_vmspace(struct image_params
*imgp
, struct vmspace
*vmcopy
)
669 struct vmspace
*vmspace
= imgp
->proc
->p_vmspace
;
670 vm_offset_t stack_addr
= USRSTACK
- maxssiz
;
676 * Indicate that we cannot gracefully error out any more, kill
677 * any other threads present, and set P_INEXEC to indicate that
678 * we are now messing with the process structure proper.
680 * If killalllwps() races return an error which coupled with
681 * vmspace_destroyed will cause us to exit. This is what we
682 * want since another thread is patiently waiting for us to exit
686 imgp
->vmspace_destroyed
= 1;
688 if (curthread
->td_proc
->p_nthreads
> 1) {
689 error
= killalllwps(1);
693 imgp
->vmspace_destroyed
|= 2; /* we are responsible for P_INEXEC */
694 p
->p_flag
|= P_INEXEC
;
697 * Prevent a pending AIO from modifying the new address space.
699 aio_proc_rundown(imgp
->proc
);
702 * Blow away entire process VM, if address space not shared,
703 * otherwise, create a new VM space so that other threads are
704 * not disrupted. If we are execing a resident vmspace we
705 * create a duplicate of it and remap the stack.
707 * The exitingcnt test is not strictly necessary but has been
708 * included for code sanity (to make the code more deterministic).
710 map
= &vmspace
->vm_map
;
712 vmspace_exec(imgp
->proc
, vmcopy
);
713 vmspace
= imgp
->proc
->p_vmspace
;
714 pmap_remove_pages(vmspace_pmap(vmspace
), stack_addr
, USRSTACK
);
715 map
= &vmspace
->vm_map
;
716 } else if (vmspace
->vm_sysref
.refcnt
== 1 &&
717 vmspace
->vm_exitingcnt
== 0) {
719 if (vmspace
->vm_upcalls
)
720 upc_release(vmspace
, ONLY_LWP_IN_PROC(imgp
->proc
));
721 pmap_remove_pages(vmspace_pmap(vmspace
),
722 0, VM_MAX_USER_ADDRESS
);
723 vm_map_remove(map
, 0, VM_MAX_USER_ADDRESS
);
725 vmspace_exec(imgp
->proc
, NULL
);
726 vmspace
= imgp
->proc
->p_vmspace
;
727 map
= &vmspace
->vm_map
;
730 /* Allocate a new stack */
731 error
= vm_map_stack(&vmspace
->vm_map
, stack_addr
, (vm_size_t
)maxssiz
,
732 0, VM_PROT_ALL
, VM_PROT_ALL
, 0);
736 /* vm_ssize and vm_maxsaddr are somewhat antiquated concepts in the
737 * VM_STACK case, but they are still used to monitor the size of the
738 * process stack so we can check the stack rlimit.
740 vmspace
->vm_ssize
= sgrowsiz
>> PAGE_SHIFT
;
741 vmspace
->vm_maxsaddr
= (char *)USRSTACK
- maxssiz
;
747 * Copy out argument and environment strings from the old process
748 * address space into the temporary string buffer.
751 exec_copyin_args(struct image_args
*args
, char *fname
,
752 enum exec_path_segflg segflg
, char **argv
, char **envv
)
758 bzero(args
, sizeof(*args
));
760 args
->buf
= objcache_get(exec_objcache
, M_WAITOK
);
761 if (args
->buf
== NULL
)
763 args
->begin_argv
= args
->buf
;
764 args
->endp
= args
->begin_argv
;
765 args
->space
= ARG_MAX
;
767 args
->fname
= args
->buf
+ ARG_MAX
;
770 * Copy the file name.
772 if (segflg
== PATH_SYSSPACE
) {
773 error
= copystr(fname
, args
->fname
, PATH_MAX
, &length
);
774 } else if (segflg
== PATH_USERSPACE
) {
775 error
= copyinstr(fname
, args
->fname
, PATH_MAX
, &length
);
779 * Extract argument strings. argv may not be NULL. The argv
780 * array is terminated by a NULL entry. We special-case the
781 * situation where argv[0] is NULL by passing { filename, NULL }
782 * to the new program to guarentee that the interpreter knows what
783 * file to open in case we exec an interpreted file. Note that
784 * a NULL argv[0] terminates the argv[] array.
786 * XXX the special-casing of argv[0] is historical and needs to be
792 while ((argp
= (caddr_t
)(intptr_t)fuword(argv
++)) != NULL
) {
793 if (argp
== (caddr_t
)-1) {
797 error
= copyinstr(argp
, args
->endp
,
798 args
->space
, &length
);
800 if (error
== ENAMETOOLONG
)
804 args
->space
-= length
;
805 args
->endp
+= length
;
808 if (args
->argc
== 0 && error
== 0) {
809 length
= strlen(args
->fname
) + 1;
810 if (length
> args
->space
) {
813 bcopy(args
->fname
, args
->endp
, length
);
814 args
->space
-= length
;
815 args
->endp
+= length
;
821 args
->begin_envv
= args
->endp
;
824 * extract environment strings. envv may be NULL.
826 if (envv
&& error
== 0) {
827 while ((envp
= (caddr_t
) (intptr_t) fuword(envv
++))) {
828 if (envp
== (caddr_t
) -1) {
832 error
= copyinstr(envp
, args
->endp
, args
->space
,
835 if (error
== ENAMETOOLONG
)
839 args
->space
-= length
;
840 args
->endp
+= length
;
848 exec_free_args(struct image_args
*args
)
851 objcache_put(exec_objcache
, args
->buf
);
857 * Copy strings out to the new process address space, constructing
858 * new arg and env vector tables. Return a pointer to the base
859 * so that it can be used as the initial stack pointer.
862 exec_copyout_strings(struct image_params
*imgp
)
864 int argc
, envc
, sgap
;
866 char *stringp
, *destp
;
867 register_t
*stack_base
;
868 struct ps_strings
*arginfo
;
872 * Calculate string base and vector table pointers.
873 * Also deal with signal trampoline code for this exec type.
875 arginfo
= (struct ps_strings
*)PS_STRINGS
;
876 szsigcode
= *(imgp
->proc
->p_sysent
->sv_szsigcode
);
877 if (stackgap_random
!= 0)
878 sgap
= ALIGN(karc4random() & (stackgap_random
- 1));
881 destp
= (caddr_t
)arginfo
- szsigcode
- SPARE_USRSPACE
- sgap
-
882 roundup((ARG_MAX
- imgp
->args
->space
), sizeof(char *));
888 copyout(imgp
->proc
->p_sysent
->sv_sigcode
,
889 ((caddr_t
)arginfo
- szsigcode
), szsigcode
);
892 * If we have a valid auxargs ptr, prepare some room
895 * The '+ 2' is for the null pointers at the end of each of the
896 * arg and env vector sets, and 'AT_COUNT*2' is room for the
900 vectp
= (char **)(destp
- (imgp
->args
->argc
+
901 imgp
->args
->envc
+ 2 + AT_COUNT
* 2) * sizeof(char*));
903 vectp
= (char **)(destp
- (imgp
->args
->argc
+
904 imgp
->args
->envc
+ 2) * sizeof(char*));
908 * NOTE: don't bother aligning the stack here for GCC 2.x, it will
909 * be done in crt1.o. Note that GCC 3.x aligns the stack in main.
913 * vectp also becomes our initial stack base
915 stack_base
= (register_t
*)vectp
;
917 stringp
= imgp
->args
->begin_argv
;
918 argc
= imgp
->args
->argc
;
919 envc
= imgp
->args
->envc
;
922 * Copy out strings - arguments and environment.
924 copyout(stringp
, destp
, ARG_MAX
- imgp
->args
->space
);
927 * Fill in "ps_strings" struct for ps, w, etc.
929 suword(&arginfo
->ps_argvstr
, (long)(intptr_t)vectp
);
930 suword(&arginfo
->ps_nargvstr
, argc
);
933 * Fill in argument portion of vector table.
935 for (; argc
> 0; --argc
) {
936 suword(vectp
++, (long)(intptr_t)destp
);
937 while (*stringp
++ != 0)
942 /* a null vector table pointer separates the argp's from the envp's */
945 suword(&arginfo
->ps_envstr
, (long)(intptr_t)vectp
);
946 suword(&arginfo
->ps_nenvstr
, envc
);
949 * Fill in environment portion of vector table.
951 for (; envc
> 0; --envc
) {
952 suword(vectp
++, (long)(intptr_t)destp
);
953 while (*stringp
++ != 0)
958 /* end of vector table is a null pointer */
965 * Check permissions of file to execute.
966 * Return 0 for success or error code on failure.
969 exec_check_permissions(struct image_params
*imgp
)
971 struct proc
*p
= imgp
->proc
;
972 struct vnode
*vp
= imgp
->vp
;
973 struct vattr
*attr
= imgp
->attr
;
976 /* Get file attributes */
977 error
= VOP_GETATTR(vp
, attr
);
982 * 1) Check if file execution is disabled for the filesystem that this
984 * 2) Insure that at least one execute bit is on - otherwise root
985 * will always succeed, and we don't want to happen unless the
986 * file really is executable.
987 * 3) Insure that the file is a regular file.
989 if ((vp
->v_mount
->mnt_flag
& MNT_NOEXEC
) ||
990 ((attr
->va_mode
& 0111) == 0) ||
991 (attr
->va_type
!= VREG
)) {
996 * Zero length files can't be exec'd
998 if (attr
->va_size
== 0)
1002 * Check for execute permission to file based on current credentials.
1004 error
= VOP_ACCESS(vp
, VEXEC
, p
->p_ucred
);
1009 * Check number of open-for-writes on the file and deny execution
1012 if (vp
->v_writecount
)
1016 * Call filesystem specific open routine, which allows us to read,
1017 * write, and mmap the file. Without the VOP_OPEN we can only
1020 error
= VOP_OPEN(vp
, FREAD
, p
->p_ucred
, NULL
);
1028 * Exec handler registration
1031 exec_register(const struct execsw
*execsw_arg
)
1033 const struct execsw
**es
, **xs
, **newexecsw
;
1034 int count
= 2; /* New slot and trailing NULL */
1037 for (es
= execsw
; *es
; es
++)
1039 newexecsw
= kmalloc(count
* sizeof(*es
), M_TEMP
, M_WAITOK
);
1042 for (es
= execsw
; *es
; es
++)
1047 kfree(execsw
, M_TEMP
);
1053 exec_unregister(const struct execsw
*execsw_arg
)
1055 const struct execsw
**es
, **xs
, **newexecsw
;
1059 panic("unregister with no handlers left?");
1061 for (es
= execsw
; *es
; es
++) {
1062 if (*es
== execsw_arg
)
1067 for (es
= execsw
; *es
; es
++)
1068 if (*es
!= execsw_arg
)
1070 newexecsw
= kmalloc(count
* sizeof(*es
), M_TEMP
, M_WAITOK
);
1072 for (es
= execsw
; *es
; es
++)
1073 if (*es
!= execsw_arg
)
1077 kfree(execsw
, M_TEMP
);