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 1.55 2007/02/25 23:17:12 corecode 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>
45 #include <sys/ktrace.h>
46 #include <sys/signalvar.h>
47 #include <sys/pioctl.h>
48 #include <sys/nlookup.h>
49 #include <sys/sfbuf.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>
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/vm_pager.h>
72 #include <sys/thread2.h>
74 MALLOC_DEFINE(M_PARGS
, "proc-args", "Process arguments");
76 static register_t
*exec_copyout_strings (struct image_params
*);
78 /* XXX This should be vm_size_t. */
79 static u_long ps_strings
= PS_STRINGS
;
80 SYSCTL_ULONG(_kern
, KERN_PS_STRINGS
, ps_strings
, CTLFLAG_RD
, &ps_strings
, 0, "");
82 /* XXX This should be vm_size_t. */
83 static u_long usrstack
= USRSTACK
;
84 SYSCTL_ULONG(_kern
, KERN_USRSTACK
, usrstack
, CTLFLAG_RD
, &usrstack
, 0, "");
86 u_long ps_arg_cache_limit
= PAGE_SIZE
/ 16;
87 SYSCTL_LONG(_kern
, OID_AUTO
, ps_arg_cache_limit
, CTLFLAG_RW
,
88 &ps_arg_cache_limit
, 0, "");
91 SYSCTL_INT(_kern
, OID_AUTO
, ps_argsopen
, CTLFLAG_RW
, &ps_argsopen
, 0, "");
93 void print_execve_args(struct image_args
*args
);
94 int debug_execve_args
= 0;
95 SYSCTL_INT(_kern
, OID_AUTO
, debug_execve_args
, CTLFLAG_RW
, &debug_execve_args
,
99 * stackgap_random specifies if the stackgap should have a random size added
100 * to it. It must be a power of 2. If non-zero, the stack gap will be
101 * calculated as: ALIGN(karc4random() & (stackgap_random - 1)).
103 static int stackgap_random
= 1024;
105 sysctl_kern_stackgap(SYSCTL_HANDLER_ARGS
)
108 new_val
= stackgap_random
;
109 error
= sysctl_handle_int(oidp
, &new_val
, 0, req
);
110 if (error
!= 0 || req
->newptr
== NULL
)
112 if ((new_val
< 0) || (new_val
> 16 * PAGE_SIZE
) || ! powerof2(new_val
))
114 stackgap_random
= new_val
;
119 SYSCTL_PROC(_kern
, OID_AUTO
, stackgap_random
, CTLFLAG_RW
|CTLTYPE_UINT
,
120 0, 0, sysctl_kern_stackgap
, "IU", "Max random stack gap (power of 2)");
123 print_execve_args(struct image_args
*args
)
128 cp
= args
->begin_argv
;
129 for (ndx
= 0; ndx
< args
->argc
; ndx
++) {
130 kprintf("\targv[%d]: %s\n", ndx
, cp
);
131 while (*cp
++ != '\0');
133 for (ndx
= 0; ndx
< args
->envc
; ndx
++) {
134 kprintf("\tenvv[%d]: %s\n", ndx
, cp
);
135 while (*cp
++ != '\0');
140 * Each of the items is a pointer to a `const struct execsw', hence the
141 * double pointer here.
143 static const struct execsw
**execsw
;
146 kern_execve(struct nlookupdata
*nd
, struct image_args
*args
)
148 struct thread
*td
= curthread
;
149 struct lwp
*lp
= td
->td_lwp
;
150 struct proc
*p
= td
->td_proc
;
151 register_t
*stack_base
;
153 struct image_params image_params
, *imgp
;
155 int (*img_first
) (struct image_params
*);
157 if (debug_execve_args
) {
158 kprintf("%s()\n", __func__
);
159 print_execve_args(args
);
163 imgp
= &image_params
;
166 * Lock the process and set the P_INEXEC flag to indicate that
167 * it should be left alone until we're done here. This is
168 * necessary to avoid race conditions - e.g. in ptrace() -
169 * that might allow a local user to illicitly obtain elevated
172 p
->p_flag
|= P_INEXEC
;
175 * Initialize part of the common data
180 imgp
->entry_addr
= 0;
182 imgp
->vmspace_destroyed
= 0;
183 imgp
->interpreted
= 0;
184 imgp
->interpreter_name
[0] = 0;
185 imgp
->auxargs
= NULL
;
187 imgp
->firstpage
= NULL
;
188 imgp
->ps_strings
= 0;
189 imgp
->image_header
= NULL
;
194 * Translate the file name to a vnode. Unlock the cache entry to
195 * improve parallelism for programs exec'd in parallel.
197 if ((error
= nlookup(nd
)) != 0)
199 error
= cache_vget(&nd
->nl_nch
, nd
->nl_cred
, LK_EXCLUSIVE
, &imgp
->vp
);
200 KKASSERT(nd
->nl_flags
& NLC_NCPISLOCKED
);
201 nd
->nl_flags
&= ~NLC_NCPISLOCKED
;
202 cache_unlock(&nd
->nl_nch
);
207 * Check file permissions (also 'opens' file)
209 error
= exec_check_permissions(imgp
);
212 goto exec_fail_dealloc
;
215 error
= exec_map_first_page(imgp
);
218 goto exec_fail_dealloc
;
220 if (debug_execve_args
&& imgp
->interpreted
) {
221 kprintf(" target is interpreted -- recursive pass\n");
222 kprintf(" interpreter: %s\n", imgp
->interpreter_name
);
223 print_execve_args(args
);
227 * If the current process has a special image activator it
228 * wants to try first, call it. For example, emulating shell
229 * scripts differently.
232 if ((img_first
= imgp
->proc
->p_sysent
->sv_imgact_try
) != NULL
)
233 error
= img_first(imgp
);
236 * If the vnode has a registered vmspace, exec the vmspace
238 if (error
== -1 && imgp
->vp
->v_resident
) {
239 error
= exec_resident_imgact(imgp
);
243 * Loop through the list of image activators, calling each one.
244 * An activator returns -1 if there is no match, 0 on success,
245 * and an error otherwise.
247 for (i
= 0; error
== -1 && execsw
[i
]; ++i
) {
248 if (execsw
[i
]->ex_imgact
== NULL
||
249 execsw
[i
]->ex_imgact
== img_first
) {
252 error
= (*execsw
[i
]->ex_imgact
)(imgp
);
258 goto exec_fail_dealloc
;
262 * Special interpreter operation, cleanup and loop up to try to
263 * activate the interpreter.
265 if (imgp
->interpreted
) {
266 exec_unmap_first_page(imgp
);
270 error
= nlookup_init(nd
, imgp
->interpreter_name
, UIO_SYSSPACE
,
278 * Copy out strings (args and env) and initialize stack base
280 stack_base
= exec_copyout_strings(imgp
);
281 p
->p_vmspace
->vm_minsaddr
= (char *)stack_base
;
284 * If custom stack fixup routine present for this process
285 * let it do the stack setup. If we are running a resident
286 * image there is no auxinfo or other image activator context
287 * so don't try to add fixups to the stack.
289 * Else stuff argument count as first item on stack
291 if (p
->p_sysent
->sv_fixup
&& imgp
->resident
== 0)
292 (*p
->p_sysent
->sv_fixup
)(&stack_base
, imgp
);
294 suword(--stack_base
, imgp
->args
->argc
);
297 * For security and other reasons, the file descriptor table cannot
298 * be shared after an exec.
300 if (p
->p_fd
->fd_refcnt
> 1) {
301 struct filedesc
*tmp
;
309 * For security and other reasons, signal handlers cannot
310 * be shared after an exec. The new proces gets a copy of the old
311 * handlers. In execsigs(), the new process will have its signals
314 if (p
->p_sigacts
->ps_refcnt
> 1) {
315 struct sigacts
*newsigacts
;
317 newsigacts
= (struct sigacts
*)kmalloc(sizeof(*newsigacts
),
318 M_SUBPROC
, M_WAITOK
);
319 bcopy(p
->p_sigacts
, newsigacts
, sizeof(*newsigacts
));
320 p
->p_sigacts
->ps_refcnt
--;
321 p
->p_sigacts
= newsigacts
;
322 p
->p_sigacts
->ps_refcnt
= 1;
326 * For security and other reasons virtual kernels cannot be
327 * inherited by an exec. This also allows a virtual kernel
328 * to fork/exec unrelated applications.
336 /* close files on exec */
339 /* reset caught signals */
342 /* name this process - nameiexec(p, ndp) */
343 len
= min(nd
->nl_nch
.ncp
->nc_nlen
, MAXCOMLEN
);
344 bcopy(nd
->nl_nch
.ncp
->nc_name
, p
->p_comm
, len
);
346 bcopy(p
->p_comm
, lp
->lwp_thread
->td_comm
, MAXCOMLEN
+1);
349 * mark as execed, wakeup the process that vforked (if any) and tell
350 * it that it now has its own resources back
353 if (p
->p_pptr
&& (p
->p_flag
& P_PPWAIT
)) {
354 p
->p_flag
&= ~P_PPWAIT
;
355 wakeup((caddr_t
)p
->p_pptr
);
359 * Implement image setuid/setgid.
361 * Don't honor setuid/setgid if the filesystem prohibits it or if
362 * the process is being traced.
364 if ((((attr
.va_mode
& VSUID
) && p
->p_ucred
->cr_uid
!= attr
.va_uid
) ||
365 ((attr
.va_mode
& VSGID
) && p
->p_ucred
->cr_gid
!= attr
.va_gid
)) &&
366 (imgp
->vp
->v_mount
->mnt_flag
& MNT_NOSUID
) == 0 &&
367 (p
->p_flag
& P_TRACED
) == 0) {
369 * Turn off syscall tracing for set-id programs, except for
370 * root. Record any set-id flags first to make sure that
371 * we do not regain any tracing during a possible block.
374 if (p
->p_tracenode
&& suser(td
) != 0) {
375 ktrdestroy(&p
->p_tracenode
);
378 /* Close any file descriptors 0..2 that reference procfs */
380 /* Make sure file descriptors 0..2 are in use. */
381 error
= fdcheckstd(p
);
383 goto exec_fail_dealloc
;
385 * Set the new credentials.
388 if (attr
.va_mode
& VSUID
)
389 change_euid(attr
.va_uid
);
390 if (attr
.va_mode
& VSGID
)
391 p
->p_ucred
->cr_gid
= attr
.va_gid
;
394 * Clear local varsym variables
396 varsymset_clean(&p
->p_varsymset
);
398 if (p
->p_ucred
->cr_uid
== p
->p_ucred
->cr_ruid
&&
399 p
->p_ucred
->cr_gid
== p
->p_ucred
->cr_rgid
)
400 p
->p_flag
&= ~P_SUGID
;
404 * Implement correct POSIX saved-id behavior.
406 if (p
->p_ucred
->cr_svuid
!= p
->p_ucred
->cr_uid
||
407 p
->p_ucred
->cr_svgid
!= p
->p_ucred
->cr_gid
) {
409 p
->p_ucred
->cr_svuid
= p
->p_ucred
->cr_uid
;
410 p
->p_ucred
->cr_svgid
= p
->p_ucred
->cr_gid
;
414 * Store the vp for use in procfs
416 if (p
->p_textvp
) /* release old reference */
418 p
->p_textvp
= imgp
->vp
;
422 * Notify others that we exec'd, and clear the P_INEXEC flag
423 * as we're now a bona fide freshly-execed process.
425 KNOTE(&p
->p_klist
, NOTE_EXEC
);
426 p
->p_flag
&= ~P_INEXEC
;
429 * If tracing the process, trap to debugger so breakpoints
430 * can be set before the program executes.
432 STOPEVENT(p
, S_EXEC
, 0);
434 if (p
->p_flag
& P_TRACED
)
437 /* clear "fork but no exec" flag, as we _are_ execing */
438 p
->p_acflag
&= ~AFORK
;
440 /* Set values passed into the program in registers. */
441 exec_setregs(imgp
->entry_addr
, (u_long
)(uintptr_t)stack_base
,
444 /* Free any previous argument cache */
445 if (p
->p_args
&& --p
->p_args
->ar_ref
== 0)
446 FREE(p
->p_args
, M_PARGS
);
449 /* Cache arguments if they fit inside our allowance */
450 i
= imgp
->args
->begin_envv
- imgp
->args
->begin_argv
;
451 if (ps_arg_cache_limit
>= i
+ sizeof(struct pargs
)) {
452 MALLOC(p
->p_args
, struct pargs
*, sizeof(struct pargs
) + i
,
454 p
->p_args
->ar_ref
= 1;
455 p
->p_args
->ar_length
= i
;
456 bcopy(imgp
->args
->begin_argv
, p
->p_args
->ar_args
, i
);
462 * free various allocated resources
465 exec_unmap_first_page(imgp
);
473 ++mycpu
->gd_cnt
.v_exec
;
478 /* we're done here, clear P_INEXEC */
479 p
->p_flag
&= ~P_INEXEC
;
480 if (imgp
->vmspace_destroyed
) {
481 /* sorry, no more process anymore. exit gracefully */
482 exit1(W_EXITCODE(0, SIGABRT
));
491 * execve() system call.
494 sys_execve(struct execve_args
*uap
)
496 struct nlookupdata nd
;
497 struct image_args args
;
500 error
= nlookup_init(&nd
, uap
->fname
, UIO_USERSPACE
, NLC_FOLLOW
);
502 error
= exec_copyin_args(&args
, uap
->fname
, PATH_USERSPACE
,
503 uap
->argv
, uap
->envv
);
506 error
= kern_execve(&nd
, &args
);
508 exec_free_args(&args
);
511 * The syscall result is returned in registers to the new program.
512 * Linux will register %edx as an atexit function and we must be
513 * sure to set it to 0. XXX
516 uap
->sysmsg_result64
= 0;
522 exec_map_first_page(struct image_params
*imgp
)
526 vm_page_t ma
[VM_INITIAL_PAGEIN
];
531 exec_unmap_first_page(imgp
);
534 * The file has to be mappable.
536 if ((object
= imgp
->vp
->v_object
) == NULL
)
540 * We shouldn't need protection for vm_page_grab() but we certainly
541 * need it for the lookup loop below (lookup/busy race), since
542 * an interrupt can unbusy and free the page before our busy check.
545 m
= vm_page_grab(object
, 0, VM_ALLOC_NORMAL
| VM_ALLOC_RETRY
);
547 if ((m
->valid
& VM_PAGE_BITS_ALL
) != VM_PAGE_BITS_ALL
) {
549 initial_pagein
= VM_INITIAL_PAGEIN
;
550 if (initial_pagein
> object
->size
)
551 initial_pagein
= object
->size
;
552 for (i
= 1; i
< initial_pagein
; i
++) {
553 if ((m
= vm_page_lookup(object
, i
)) != NULL
) {
554 if ((m
->flags
& PG_BUSY
) || m
->busy
)
560 m
= vm_page_alloc(object
, i
, VM_ALLOC_NORMAL
);
569 * get_pages unbusies all the requested pages except the
570 * primary page (at index 0 in this case).
572 rv
= vm_pager_get_pages(object
, ma
, initial_pagein
, 0);
573 m
= vm_page_lookup(object
, 0);
575 if (rv
!= VM_PAGER_OK
|| m
== NULL
|| m
->valid
== 0) {
577 vm_page_protect(m
, VM_PROT_NONE
);
585 vm_page_wakeup(m
); /* unbusy the page */
588 imgp
->firstpage
= sf_buf_alloc(m
, SFB_CPUPRIVATE
);
589 imgp
->image_header
= (void *)sf_buf_kva(imgp
->firstpage
);
595 exec_unmap_first_page(struct image_params
*imgp
)
600 if (imgp
->firstpage
!= NULL
) {
601 m
= sf_buf_page(imgp
->firstpage
);
602 sf_buf_free(imgp
->firstpage
);
603 imgp
->firstpage
= NULL
;
604 imgp
->image_header
= NULL
;
611 * Destroy old address space, and allocate a new stack
612 * The new stack is only SGROWSIZ large because it is grown
613 * automatically in trap.c.
616 exec_new_vmspace(struct image_params
*imgp
, struct vmspace
*vmcopy
)
619 struct vmspace
*vmspace
= imgp
->proc
->p_vmspace
;
620 vm_offset_t stack_addr
= USRSTACK
- maxssiz
;
623 imgp
->vmspace_destroyed
= 1;
625 if (curthread
->td_proc
->p_nthreads
> 1)
626 killlwps(curthread
->td_lwp
);
629 * Prevent a pending AIO from modifying the new address space.
631 aio_proc_rundown(imgp
->proc
);
634 * Blow away entire process VM, if address space not shared,
635 * otherwise, create a new VM space so that other threads are
636 * not disrupted. If we are execing a resident vmspace we
637 * create a duplicate of it and remap the stack.
639 * The exitingcnt test is not strictly necessary but has been
640 * included for code sanity (to make the code more deterministic).
642 map
= &vmspace
->vm_map
;
644 vmspace_exec(imgp
->proc
, vmcopy
);
645 vmspace
= imgp
->proc
->p_vmspace
;
646 pmap_remove_pages(vmspace_pmap(vmspace
), stack_addr
, USRSTACK
);
647 map
= &vmspace
->vm_map
;
648 } else if (vmspace
->vm_refcnt
== 1 && vmspace
->vm_exitingcnt
== 0) {
650 if (vmspace
->vm_upcalls
)
651 upc_release(vmspace
, ONLY_LWP_IN_PROC(imgp
->proc
));
652 pmap_remove_pages(vmspace_pmap(vmspace
),
653 0, VM_MAX_USER_ADDRESS
);
654 vm_map_remove(map
, 0, VM_MAX_USER_ADDRESS
);
656 vmspace_exec(imgp
->proc
, NULL
);
657 vmspace
= imgp
->proc
->p_vmspace
;
658 map
= &vmspace
->vm_map
;
661 /* Allocate a new stack */
662 error
= vm_map_stack(&vmspace
->vm_map
, stack_addr
, (vm_size_t
)maxssiz
,
663 VM_PROT_ALL
, VM_PROT_ALL
, 0);
667 /* vm_ssize and vm_maxsaddr are somewhat antiquated concepts in the
668 * VM_STACK case, but they are still used to monitor the size of the
669 * process stack so we can check the stack rlimit.
671 vmspace
->vm_ssize
= sgrowsiz
>> PAGE_SHIFT
;
672 vmspace
->vm_maxsaddr
= (char *)USRSTACK
- maxssiz
;
678 * Copy out argument and environment strings from the old process
679 * address space into the temporary string buffer.
682 exec_copyin_args(struct image_args
*args
, char *fname
,
683 enum exec_path_segflg segflg
, char **argv
, char **envv
)
689 bzero(args
, sizeof(*args
));
690 args
->buf
= (char *) kmem_alloc_wait(&exec_map
, PATH_MAX
+ ARG_MAX
);
691 if (args
->buf
== NULL
)
693 args
->begin_argv
= args
->buf
;
694 args
->endp
= args
->begin_argv
;
695 args
->space
= ARG_MAX
;
697 args
->fname
= args
->buf
+ ARG_MAX
;
700 * Copy the file name.
702 if (segflg
== PATH_SYSSPACE
) {
703 error
= copystr(fname
, args
->fname
, PATH_MAX
, &length
);
704 } else if (segflg
== PATH_USERSPACE
) {
705 error
= copyinstr(fname
, args
->fname
, PATH_MAX
, &length
);
709 * Extract argument strings. argv may not be NULL. The argv
710 * array is terminated by a NULL entry. We special-case the
711 * situation where argv[0] is NULL by passing { filename, NULL }
712 * to the new program to guarentee that the interpreter knows what
713 * file to open in case we exec an interpreted file. Note that
714 * a NULL argv[0] terminates the argv[] array.
716 * XXX the special-casing of argv[0] is historical and needs to be
722 while ((argp
= (caddr_t
)(intptr_t)fuword(argv
++)) != NULL
) {
723 if (argp
== (caddr_t
)-1) {
727 error
= copyinstr(argp
, args
->endp
,
728 args
->space
, &length
);
730 if (error
== ENAMETOOLONG
)
734 args
->space
-= length
;
735 args
->endp
+= length
;
738 if (args
->argc
== 0 && error
== 0) {
739 length
= strlen(args
->fname
) + 1;
740 if (length
> args
->space
) {
743 bcopy(args
->fname
, args
->endp
, length
);
744 args
->space
-= length
;
745 args
->endp
+= length
;
751 args
->begin_envv
= args
->endp
;
754 * extract environment strings. envv may be NULL.
756 if (envv
&& error
== 0) {
757 while ((envp
= (caddr_t
) (intptr_t) fuword(envv
++))) {
758 if (envp
== (caddr_t
) -1) {
762 error
= copyinstr(envp
, args
->endp
, args
->space
,
765 if (error
== ENAMETOOLONG
)
769 args
->space
-= length
;
770 args
->endp
+= length
;
778 exec_free_args(struct image_args
*args
)
781 kmem_free_wakeup(&exec_map
,
782 (vm_offset_t
)args
->buf
, PATH_MAX
+ ARG_MAX
);
788 * Copy strings out to the new process address space, constructing
789 * new arg and env vector tables. Return a pointer to the base
790 * so that it can be used as the initial stack pointer.
793 exec_copyout_strings(struct image_params
*imgp
)
795 int argc
, envc
, sgap
;
797 char *stringp
, *destp
;
798 register_t
*stack_base
;
799 struct ps_strings
*arginfo
;
803 * Calculate string base and vector table pointers.
804 * Also deal with signal trampoline code for this exec type.
806 arginfo
= (struct ps_strings
*)PS_STRINGS
;
807 szsigcode
= *(imgp
->proc
->p_sysent
->sv_szsigcode
);
808 if (stackgap_random
!= 0)
809 sgap
= ALIGN(karc4random() & (stackgap_random
- 1));
812 destp
= (caddr_t
)arginfo
- szsigcode
- SPARE_USRSPACE
- sgap
-
813 roundup((ARG_MAX
- imgp
->args
->space
), sizeof(char *));
819 copyout(imgp
->proc
->p_sysent
->sv_sigcode
,
820 ((caddr_t
)arginfo
- szsigcode
), szsigcode
);
823 * If we have a valid auxargs ptr, prepare some room
826 * The '+ 2' is for the null pointers at the end of each of the
827 * arg and env vector sets, and 'AT_COUNT*2' is room for the
831 vectp
= (char **)(destp
- (imgp
->args
->argc
+
832 imgp
->args
->envc
+ 2 + AT_COUNT
* 2) * sizeof(char*));
834 vectp
= (char **)(destp
- (imgp
->args
->argc
+
835 imgp
->args
->envc
+ 2) * sizeof(char*));
839 * NOTE: don't bother aligning the stack here for GCC 2.x, it will
840 * be done in crt1.o. Note that GCC 3.x aligns the stack in main.
844 * vectp also becomes our initial stack base
846 stack_base
= (register_t
*)vectp
;
848 stringp
= imgp
->args
->begin_argv
;
849 argc
= imgp
->args
->argc
;
850 envc
= imgp
->args
->envc
;
853 * Copy out strings - arguments and environment.
855 copyout(stringp
, destp
, ARG_MAX
- imgp
->args
->space
);
858 * Fill in "ps_strings" struct for ps, w, etc.
860 suword(&arginfo
->ps_argvstr
, (long)(intptr_t)vectp
);
861 suword(&arginfo
->ps_nargvstr
, argc
);
864 * Fill in argument portion of vector table.
866 for (; argc
> 0; --argc
) {
867 suword(vectp
++, (long)(intptr_t)destp
);
868 while (*stringp
++ != 0)
873 /* a null vector table pointer separates the argp's from the envp's */
876 suword(&arginfo
->ps_envstr
, (long)(intptr_t)vectp
);
877 suword(&arginfo
->ps_nenvstr
, envc
);
880 * Fill in environment portion of vector table.
882 for (; envc
> 0; --envc
) {
883 suword(vectp
++, (long)(intptr_t)destp
);
884 while (*stringp
++ != 0)
889 /* end of vector table is a null pointer */
896 * Check permissions of file to execute.
897 * Return 0 for success or error code on failure.
900 exec_check_permissions(struct image_params
*imgp
)
902 struct proc
*p
= imgp
->proc
;
903 struct vnode
*vp
= imgp
->vp
;
904 struct vattr
*attr
= imgp
->attr
;
907 /* Get file attributes */
908 error
= VOP_GETATTR(vp
, attr
);
913 * 1) Check if file execution is disabled for the filesystem that this
915 * 2) Insure that at least one execute bit is on - otherwise root
916 * will always succeed, and we don't want to happen unless the
917 * file really is executable.
918 * 3) Insure that the file is a regular file.
920 if ((vp
->v_mount
->mnt_flag
& MNT_NOEXEC
) ||
921 ((attr
->va_mode
& 0111) == 0) ||
922 (attr
->va_type
!= VREG
)) {
927 * Zero length files can't be exec'd
929 if (attr
->va_size
== 0)
933 * Check for execute permission to file based on current credentials.
935 error
= VOP_ACCESS(vp
, VEXEC
, p
->p_ucred
);
940 * Check number of open-for-writes on the file and deny execution
943 if (vp
->v_writecount
)
947 * Call filesystem specific open routine, which allows us to read,
948 * write, and mmap the file. Without the VOP_OPEN we can only
951 error
= VOP_OPEN(vp
, FREAD
, p
->p_ucred
, NULL
);
959 * Exec handler registration
962 exec_register(const struct execsw
*execsw_arg
)
964 const struct execsw
**es
, **xs
, **newexecsw
;
965 int count
= 2; /* New slot and trailing NULL */
968 for (es
= execsw
; *es
; es
++)
970 newexecsw
= kmalloc(count
* sizeof(*es
), M_TEMP
, M_WAITOK
);
971 if (newexecsw
== NULL
)
975 for (es
= execsw
; *es
; es
++)
980 kfree(execsw
, M_TEMP
);
986 exec_unregister(const struct execsw
*execsw_arg
)
988 const struct execsw
**es
, **xs
, **newexecsw
;
992 panic("unregister with no handlers left?");
994 for (es
= execsw
; *es
; es
++) {
995 if (*es
== execsw_arg
)
1000 for (es
= execsw
; *es
; es
++)
1001 if (*es
!= execsw_arg
)
1003 newexecsw
= kmalloc(count
* sizeof(*es
), M_TEMP
, M_WAITOK
);
1004 if (newexecsw
== NULL
)
1007 for (es
= execsw
; *es
; es
++)
1008 if (*es
!= execsw_arg
)
1012 kfree(execsw
, M_TEMP
);