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 1.57 2007/06/07 23:14:25 dillon 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");
75 MALLOC_DEFINE(M_EXECARGS
, "exec-args", "Exec arguments");
77 static register_t
*exec_copyout_strings (struct image_params
*);
79 /* XXX This should be vm_size_t. */
80 static u_long ps_strings
= PS_STRINGS
;
81 SYSCTL_ULONG(_kern
, KERN_PS_STRINGS
, ps_strings
, CTLFLAG_RD
, &ps_strings
, 0, "");
83 /* XXX This should be vm_size_t. */
84 static u_long usrstack
= USRSTACK
;
85 SYSCTL_ULONG(_kern
, KERN_USRSTACK
, usrstack
, CTLFLAG_RD
, &usrstack
, 0, "");
87 u_long ps_arg_cache_limit
= PAGE_SIZE
/ 16;
88 SYSCTL_LONG(_kern
, OID_AUTO
, ps_arg_cache_limit
, CTLFLAG_RW
,
89 &ps_arg_cache_limit
, 0, "");
92 SYSCTL_INT(_kern
, OID_AUTO
, ps_argsopen
, CTLFLAG_RW
, &ps_argsopen
, 0, "");
94 void print_execve_args(struct image_args
*args
);
95 int debug_execve_args
= 0;
96 SYSCTL_INT(_kern
, OID_AUTO
, debug_execve_args
, CTLFLAG_RW
, &debug_execve_args
,
100 * Exec arguments object cache
102 static struct objcache
*exec_objcache
;
106 exec_objcache_init(void *arg __unused
)
108 exec_objcache
= objcache_create_mbacked(
109 M_EXECARGS
, PATH_MAX
+ ARG_MAX
,
110 16, /* up to this many objects */
111 2, /* minimal magazine capacity */
114 SYSINIT(exec_objcache
, SI_BOOT2_MACHDEP
, SI_ORDER_ANY
, exec_objcache_init
, 0);
117 * stackgap_random specifies if the stackgap should have a random size added
118 * to it. It must be a power of 2. If non-zero, the stack gap will be
119 * calculated as: ALIGN(karc4random() & (stackgap_random - 1)).
121 static int stackgap_random
= 1024;
123 sysctl_kern_stackgap(SYSCTL_HANDLER_ARGS
)
126 new_val
= stackgap_random
;
127 error
= sysctl_handle_int(oidp
, &new_val
, 0, req
);
128 if (error
!= 0 || req
->newptr
== NULL
)
130 if ((new_val
< 0) || (new_val
> 16 * PAGE_SIZE
) || ! powerof2(new_val
))
132 stackgap_random
= new_val
;
137 SYSCTL_PROC(_kern
, OID_AUTO
, stackgap_random
, CTLFLAG_RW
|CTLTYPE_UINT
,
138 0, 0, sysctl_kern_stackgap
, "IU", "Max random stack gap (power of 2)");
141 print_execve_args(struct image_args
*args
)
146 cp
= args
->begin_argv
;
147 for (ndx
= 0; ndx
< args
->argc
; ndx
++) {
148 kprintf("\targv[%d]: %s\n", ndx
, cp
);
149 while (*cp
++ != '\0');
151 for (ndx
= 0; ndx
< args
->envc
; ndx
++) {
152 kprintf("\tenvv[%d]: %s\n", ndx
, cp
);
153 while (*cp
++ != '\0');
158 * Each of the items is a pointer to a `const struct execsw', hence the
159 * double pointer here.
161 static const struct execsw
**execsw
;
164 kern_execve(struct nlookupdata
*nd
, struct image_args
*args
)
166 struct thread
*td
= curthread
;
167 struct lwp
*lp
= td
->td_lwp
;
168 struct proc
*p
= td
->td_proc
;
169 register_t
*stack_base
;
171 struct image_params image_params
, *imgp
;
173 int (*img_first
) (struct image_params
*);
175 if (debug_execve_args
) {
176 kprintf("%s()\n", __func__
);
177 print_execve_args(args
);
181 imgp
= &image_params
;
184 * Lock the process and set the P_INEXEC flag to indicate that
185 * it should be left alone until we're done here. This is
186 * necessary to avoid race conditions - e.g. in ptrace() -
187 * that might allow a local user to illicitly obtain elevated
190 p
->p_flag
|= P_INEXEC
;
193 * Initialize part of the common data
198 imgp
->entry_addr
= 0;
200 imgp
->vmspace_destroyed
= 0;
201 imgp
->interpreted
= 0;
202 imgp
->interpreter_name
[0] = 0;
203 imgp
->auxargs
= NULL
;
205 imgp
->firstpage
= NULL
;
206 imgp
->ps_strings
= 0;
207 imgp
->image_header
= NULL
;
212 * Translate the file name to a vnode. Unlock the cache entry to
213 * improve parallelism for programs exec'd in parallel.
215 if ((error
= nlookup(nd
)) != 0)
217 error
= cache_vget(&nd
->nl_nch
, nd
->nl_cred
, LK_EXCLUSIVE
, &imgp
->vp
);
218 KKASSERT(nd
->nl_flags
& NLC_NCPISLOCKED
);
219 nd
->nl_flags
&= ~NLC_NCPISLOCKED
;
220 cache_unlock(&nd
->nl_nch
);
225 * Check file permissions (also 'opens' file)
227 error
= exec_check_permissions(imgp
);
230 goto exec_fail_dealloc
;
233 error
= exec_map_first_page(imgp
);
236 goto exec_fail_dealloc
;
238 if (debug_execve_args
&& imgp
->interpreted
) {
239 kprintf(" target is interpreted -- recursive pass\n");
240 kprintf(" interpreter: %s\n", imgp
->interpreter_name
);
241 print_execve_args(args
);
245 * If the current process has a special image activator it
246 * wants to try first, call it. For example, emulating shell
247 * scripts differently.
250 if ((img_first
= imgp
->proc
->p_sysent
->sv_imgact_try
) != NULL
)
251 error
= img_first(imgp
);
254 * If the vnode has a registered vmspace, exec the vmspace
256 if (error
== -1 && imgp
->vp
->v_resident
) {
257 error
= exec_resident_imgact(imgp
);
261 * Loop through the list of image activators, calling each one.
262 * An activator returns -1 if there is no match, 0 on success,
263 * and an error otherwise.
265 for (i
= 0; error
== -1 && execsw
[i
]; ++i
) {
266 if (execsw
[i
]->ex_imgact
== NULL
||
267 execsw
[i
]->ex_imgact
== img_first
) {
270 error
= (*execsw
[i
]->ex_imgact
)(imgp
);
276 goto exec_fail_dealloc
;
280 * Special interpreter operation, cleanup and loop up to try to
281 * activate the interpreter.
283 if (imgp
->interpreted
) {
284 exec_unmap_first_page(imgp
);
288 error
= nlookup_init(nd
, imgp
->interpreter_name
, UIO_SYSSPACE
,
296 * Copy out strings (args and env) and initialize stack base
298 stack_base
= exec_copyout_strings(imgp
);
299 p
->p_vmspace
->vm_minsaddr
= (char *)stack_base
;
302 * If custom stack fixup routine present for this process
303 * let it do the stack setup. If we are running a resident
304 * image there is no auxinfo or other image activator context
305 * so don't try to add fixups to the stack.
307 * Else stuff argument count as first item on stack
309 if (p
->p_sysent
->sv_fixup
&& imgp
->resident
== 0)
310 (*p
->p_sysent
->sv_fixup
)(&stack_base
, imgp
);
312 suword(--stack_base
, imgp
->args
->argc
);
315 * For security and other reasons, the file descriptor table cannot
316 * be shared after an exec.
318 if (p
->p_fd
->fd_refcnt
> 1) {
319 struct filedesc
*tmp
;
327 * For security and other reasons, signal handlers cannot
328 * be shared after an exec. The new proces gets a copy of the old
329 * handlers. In execsigs(), the new process will have its signals
332 if (p
->p_sigacts
->ps_refcnt
> 1) {
333 struct sigacts
*newsigacts
;
335 newsigacts
= (struct sigacts
*)kmalloc(sizeof(*newsigacts
),
336 M_SUBPROC
, M_WAITOK
);
337 bcopy(p
->p_sigacts
, newsigacts
, sizeof(*newsigacts
));
338 p
->p_sigacts
->ps_refcnt
--;
339 p
->p_sigacts
= newsigacts
;
340 p
->p_sigacts
->ps_refcnt
= 1;
344 * For security and other reasons virtual kernels cannot be
345 * inherited by an exec. This also allows a virtual kernel
346 * to fork/exec unrelated applications.
354 /* close files on exec */
357 /* reset caught signals */
360 /* name this process - nameiexec(p, ndp) */
361 len
= min(nd
->nl_nch
.ncp
->nc_nlen
, MAXCOMLEN
);
362 bcopy(nd
->nl_nch
.ncp
->nc_name
, p
->p_comm
, len
);
364 bcopy(p
->p_comm
, lp
->lwp_thread
->td_comm
, MAXCOMLEN
+1);
367 * mark as execed, wakeup the process that vforked (if any) and tell
368 * it that it now has its own resources back
371 if (p
->p_pptr
&& (p
->p_flag
& P_PPWAIT
)) {
372 p
->p_flag
&= ~P_PPWAIT
;
373 wakeup((caddr_t
)p
->p_pptr
);
377 * Implement image setuid/setgid.
379 * Don't honor setuid/setgid if the filesystem prohibits it or if
380 * the process is being traced.
382 if ((((attr
.va_mode
& VSUID
) && p
->p_ucred
->cr_uid
!= attr
.va_uid
) ||
383 ((attr
.va_mode
& VSGID
) && p
->p_ucred
->cr_gid
!= attr
.va_gid
)) &&
384 (imgp
->vp
->v_mount
->mnt_flag
& MNT_NOSUID
) == 0 &&
385 (p
->p_flag
& P_TRACED
) == 0) {
387 * Turn off syscall tracing for set-id programs, except for
388 * root. Record any set-id flags first to make sure that
389 * we do not regain any tracing during a possible block.
392 if (p
->p_tracenode
&& suser(td
) != 0) {
393 ktrdestroy(&p
->p_tracenode
);
396 /* Close any file descriptors 0..2 that reference procfs */
398 /* Make sure file descriptors 0..2 are in use. */
399 error
= fdcheckstd(p
);
401 goto exec_fail_dealloc
;
403 * Set the new credentials.
406 if (attr
.va_mode
& VSUID
)
407 change_euid(attr
.va_uid
);
408 if (attr
.va_mode
& VSGID
)
409 p
->p_ucred
->cr_gid
= attr
.va_gid
;
412 * Clear local varsym variables
414 varsymset_clean(&p
->p_varsymset
);
416 if (p
->p_ucred
->cr_uid
== p
->p_ucred
->cr_ruid
&&
417 p
->p_ucred
->cr_gid
== p
->p_ucred
->cr_rgid
)
418 p
->p_flag
&= ~P_SUGID
;
422 * Implement correct POSIX saved-id behavior.
424 if (p
->p_ucred
->cr_svuid
!= p
->p_ucred
->cr_uid
||
425 p
->p_ucred
->cr_svgid
!= p
->p_ucred
->cr_gid
) {
427 p
->p_ucred
->cr_svuid
= p
->p_ucred
->cr_uid
;
428 p
->p_ucred
->cr_svgid
= p
->p_ucred
->cr_gid
;
432 * Store the vp for use in procfs
434 if (p
->p_textvp
) /* release old reference */
436 p
->p_textvp
= imgp
->vp
;
440 * Notify others that we exec'd, and clear the P_INEXEC flag
441 * as we're now a bona fide freshly-execed process.
443 KNOTE(&p
->p_klist
, NOTE_EXEC
);
444 p
->p_flag
&= ~P_INEXEC
;
447 * If tracing the process, trap to debugger so breakpoints
448 * can be set before the program executes.
450 STOPEVENT(p
, S_EXEC
, 0);
452 if (p
->p_flag
& P_TRACED
)
455 /* clear "fork but no exec" flag, as we _are_ execing */
456 p
->p_acflag
&= ~AFORK
;
458 /* Set values passed into the program in registers. */
459 exec_setregs(imgp
->entry_addr
, (u_long
)(uintptr_t)stack_base
,
462 /* Free any previous argument cache */
463 if (p
->p_args
&& --p
->p_args
->ar_ref
== 0)
464 FREE(p
->p_args
, M_PARGS
);
467 /* Cache arguments if they fit inside our allowance */
468 i
= imgp
->args
->begin_envv
- imgp
->args
->begin_argv
;
469 if (ps_arg_cache_limit
>= i
+ sizeof(struct pargs
)) {
470 MALLOC(p
->p_args
, struct pargs
*, sizeof(struct pargs
) + i
,
472 p
->p_args
->ar_ref
= 1;
473 p
->p_args
->ar_length
= i
;
474 bcopy(imgp
->args
->begin_argv
, p
->p_args
->ar_args
, i
);
480 * free various allocated resources
483 exec_unmap_first_page(imgp
);
491 ++mycpu
->gd_cnt
.v_exec
;
496 /* we're done here, clear P_INEXEC */
497 p
->p_flag
&= ~P_INEXEC
;
498 if (imgp
->vmspace_destroyed
) {
499 /* sorry, no more process anymore. exit gracefully */
500 exit1(W_EXITCODE(0, SIGABRT
));
509 * execve() system call.
512 sys_execve(struct execve_args
*uap
)
514 struct nlookupdata nd
;
515 struct image_args args
;
518 error
= nlookup_init(&nd
, uap
->fname
, UIO_USERSPACE
, NLC_FOLLOW
);
520 error
= exec_copyin_args(&args
, uap
->fname
, PATH_USERSPACE
,
521 uap
->argv
, uap
->envv
);
524 error
= kern_execve(&nd
, &args
);
526 exec_free_args(&args
);
529 * The syscall result is returned in registers to the new program.
530 * Linux will register %edx as an atexit function and we must be
531 * sure to set it to 0. XXX
534 uap
->sysmsg_result64
= 0;
540 exec_map_first_page(struct image_params
*imgp
)
544 vm_page_t ma
[VM_INITIAL_PAGEIN
];
549 exec_unmap_first_page(imgp
);
552 * The file has to be mappable.
554 if ((object
= imgp
->vp
->v_object
) == NULL
)
558 * We shouldn't need protection for vm_page_grab() but we certainly
559 * need it for the lookup loop below (lookup/busy race), since
560 * an interrupt can unbusy and free the page before our busy check.
563 m
= vm_page_grab(object
, 0, VM_ALLOC_NORMAL
| VM_ALLOC_RETRY
);
565 if ((m
->valid
& VM_PAGE_BITS_ALL
) != VM_PAGE_BITS_ALL
) {
567 initial_pagein
= VM_INITIAL_PAGEIN
;
568 if (initial_pagein
> object
->size
)
569 initial_pagein
= object
->size
;
570 for (i
= 1; i
< initial_pagein
; i
++) {
571 if ((m
= vm_page_lookup(object
, i
)) != NULL
) {
572 if ((m
->flags
& PG_BUSY
) || m
->busy
)
578 m
= vm_page_alloc(object
, i
, VM_ALLOC_NORMAL
);
587 * get_pages unbusies all the requested pages except the
588 * primary page (at index 0 in this case).
590 rv
= vm_pager_get_pages(object
, ma
, initial_pagein
, 0);
591 m
= vm_page_lookup(object
, 0);
593 if (rv
!= VM_PAGER_OK
|| m
== NULL
|| m
->valid
== 0) {
595 vm_page_protect(m
, VM_PROT_NONE
);
603 vm_page_wakeup(m
); /* unbusy the page */
606 imgp
->firstpage
= sf_buf_alloc(m
, SFB_CPUPRIVATE
);
607 imgp
->image_header
= (void *)sf_buf_kva(imgp
->firstpage
);
613 exec_unmap_first_page(struct image_params
*imgp
)
618 if (imgp
->firstpage
!= NULL
) {
619 m
= sf_buf_page(imgp
->firstpage
);
620 sf_buf_free(imgp
->firstpage
);
621 imgp
->firstpage
= NULL
;
622 imgp
->image_header
= NULL
;
629 * Destroy old address space, and allocate a new stack
630 * The new stack is only SGROWSIZ large because it is grown
631 * automatically in trap.c.
634 exec_new_vmspace(struct image_params
*imgp
, struct vmspace
*vmcopy
)
637 struct vmspace
*vmspace
= imgp
->proc
->p_vmspace
;
638 vm_offset_t stack_addr
= USRSTACK
- maxssiz
;
641 imgp
->vmspace_destroyed
= 1;
643 if (curthread
->td_proc
->p_nthreads
> 1)
644 killlwps(curthread
->td_lwp
);
647 * Prevent a pending AIO from modifying the new address space.
649 aio_proc_rundown(imgp
->proc
);
652 * Blow away entire process VM, if address space not shared,
653 * otherwise, create a new VM space so that other threads are
654 * not disrupted. If we are execing a resident vmspace we
655 * create a duplicate of it and remap the stack.
657 * The exitingcnt test is not strictly necessary but has been
658 * included for code sanity (to make the code more deterministic).
660 map
= &vmspace
->vm_map
;
662 vmspace_exec(imgp
->proc
, vmcopy
);
663 vmspace
= imgp
->proc
->p_vmspace
;
664 pmap_remove_pages(vmspace_pmap(vmspace
), stack_addr
, USRSTACK
);
665 map
= &vmspace
->vm_map
;
666 } else if (vmspace
->vm_sysref
.refcnt
== 1 &&
667 vmspace
->vm_exitingcnt
== 0) {
669 if (vmspace
->vm_upcalls
)
670 upc_release(vmspace
, ONLY_LWP_IN_PROC(imgp
->proc
));
671 pmap_remove_pages(vmspace_pmap(vmspace
),
672 0, VM_MAX_USER_ADDRESS
);
673 vm_map_remove(map
, 0, VM_MAX_USER_ADDRESS
);
675 vmspace_exec(imgp
->proc
, NULL
);
676 vmspace
= imgp
->proc
->p_vmspace
;
677 map
= &vmspace
->vm_map
;
680 /* Allocate a new stack */
681 error
= vm_map_stack(&vmspace
->vm_map
, stack_addr
, (vm_size_t
)maxssiz
,
682 VM_PROT_ALL
, VM_PROT_ALL
, 0);
686 /* vm_ssize and vm_maxsaddr are somewhat antiquated concepts in the
687 * VM_STACK case, but they are still used to monitor the size of the
688 * process stack so we can check the stack rlimit.
690 vmspace
->vm_ssize
= sgrowsiz
>> PAGE_SHIFT
;
691 vmspace
->vm_maxsaddr
= (char *)USRSTACK
- maxssiz
;
697 * Copy out argument and environment strings from the old process
698 * address space into the temporary string buffer.
701 exec_copyin_args(struct image_args
*args
, char *fname
,
702 enum exec_path_segflg segflg
, char **argv
, char **envv
)
708 bzero(args
, sizeof(*args
));
710 args
->buf
= objcache_get(exec_objcache
, M_WAITOK
);
711 if (args
->buf
== NULL
)
713 args
->begin_argv
= args
->buf
;
714 args
->endp
= args
->begin_argv
;
715 args
->space
= ARG_MAX
;
717 args
->fname
= args
->buf
+ ARG_MAX
;
720 * Copy the file name.
722 if (segflg
== PATH_SYSSPACE
) {
723 error
= copystr(fname
, args
->fname
, PATH_MAX
, &length
);
724 } else if (segflg
== PATH_USERSPACE
) {
725 error
= copyinstr(fname
, args
->fname
, PATH_MAX
, &length
);
729 * Extract argument strings. argv may not be NULL. The argv
730 * array is terminated by a NULL entry. We special-case the
731 * situation where argv[0] is NULL by passing { filename, NULL }
732 * to the new program to guarentee that the interpreter knows what
733 * file to open in case we exec an interpreted file. Note that
734 * a NULL argv[0] terminates the argv[] array.
736 * XXX the special-casing of argv[0] is historical and needs to be
742 while ((argp
= (caddr_t
)(intptr_t)fuword(argv
++)) != NULL
) {
743 if (argp
== (caddr_t
)-1) {
747 error
= copyinstr(argp
, args
->endp
,
748 args
->space
, &length
);
750 if (error
== ENAMETOOLONG
)
754 args
->space
-= length
;
755 args
->endp
+= length
;
758 if (args
->argc
== 0 && error
== 0) {
759 length
= strlen(args
->fname
) + 1;
760 if (length
> args
->space
) {
763 bcopy(args
->fname
, args
->endp
, length
);
764 args
->space
-= length
;
765 args
->endp
+= length
;
771 args
->begin_envv
= args
->endp
;
774 * extract environment strings. envv may be NULL.
776 if (envv
&& error
== 0) {
777 while ((envp
= (caddr_t
) (intptr_t) fuword(envv
++))) {
778 if (envp
== (caddr_t
) -1) {
782 error
= copyinstr(envp
, args
->endp
, args
->space
,
785 if (error
== ENAMETOOLONG
)
789 args
->space
-= length
;
790 args
->endp
+= length
;
798 exec_free_args(struct image_args
*args
)
801 objcache_put(exec_objcache
, args
->buf
);
807 * Copy strings out to the new process address space, constructing
808 * new arg and env vector tables. Return a pointer to the base
809 * so that it can be used as the initial stack pointer.
812 exec_copyout_strings(struct image_params
*imgp
)
814 int argc
, envc
, sgap
;
816 char *stringp
, *destp
;
817 register_t
*stack_base
;
818 struct ps_strings
*arginfo
;
822 * Calculate string base and vector table pointers.
823 * Also deal with signal trampoline code for this exec type.
825 arginfo
= (struct ps_strings
*)PS_STRINGS
;
826 szsigcode
= *(imgp
->proc
->p_sysent
->sv_szsigcode
);
827 if (stackgap_random
!= 0)
828 sgap
= ALIGN(karc4random() & (stackgap_random
- 1));
831 destp
= (caddr_t
)arginfo
- szsigcode
- SPARE_USRSPACE
- sgap
-
832 roundup((ARG_MAX
- imgp
->args
->space
), sizeof(char *));
838 copyout(imgp
->proc
->p_sysent
->sv_sigcode
,
839 ((caddr_t
)arginfo
- szsigcode
), szsigcode
);
842 * If we have a valid auxargs ptr, prepare some room
845 * The '+ 2' is for the null pointers at the end of each of the
846 * arg and env vector sets, and 'AT_COUNT*2' is room for the
850 vectp
= (char **)(destp
- (imgp
->args
->argc
+
851 imgp
->args
->envc
+ 2 + AT_COUNT
* 2) * sizeof(char*));
853 vectp
= (char **)(destp
- (imgp
->args
->argc
+
854 imgp
->args
->envc
+ 2) * sizeof(char*));
858 * NOTE: don't bother aligning the stack here for GCC 2.x, it will
859 * be done in crt1.o. Note that GCC 3.x aligns the stack in main.
863 * vectp also becomes our initial stack base
865 stack_base
= (register_t
*)vectp
;
867 stringp
= imgp
->args
->begin_argv
;
868 argc
= imgp
->args
->argc
;
869 envc
= imgp
->args
->envc
;
872 * Copy out strings - arguments and environment.
874 copyout(stringp
, destp
, ARG_MAX
- imgp
->args
->space
);
877 * Fill in "ps_strings" struct for ps, w, etc.
879 suword(&arginfo
->ps_argvstr
, (long)(intptr_t)vectp
);
880 suword(&arginfo
->ps_nargvstr
, argc
);
883 * Fill in argument portion of vector table.
885 for (; argc
> 0; --argc
) {
886 suword(vectp
++, (long)(intptr_t)destp
);
887 while (*stringp
++ != 0)
892 /* a null vector table pointer separates the argp's from the envp's */
895 suword(&arginfo
->ps_envstr
, (long)(intptr_t)vectp
);
896 suword(&arginfo
->ps_nenvstr
, envc
);
899 * Fill in environment portion of vector table.
901 for (; envc
> 0; --envc
) {
902 suword(vectp
++, (long)(intptr_t)destp
);
903 while (*stringp
++ != 0)
908 /* end of vector table is a null pointer */
915 * Check permissions of file to execute.
916 * Return 0 for success or error code on failure.
919 exec_check_permissions(struct image_params
*imgp
)
921 struct proc
*p
= imgp
->proc
;
922 struct vnode
*vp
= imgp
->vp
;
923 struct vattr
*attr
= imgp
->attr
;
926 /* Get file attributes */
927 error
= VOP_GETATTR(vp
, attr
);
932 * 1) Check if file execution is disabled for the filesystem that this
934 * 2) Insure that at least one execute bit is on - otherwise root
935 * will always succeed, and we don't want to happen unless the
936 * file really is executable.
937 * 3) Insure that the file is a regular file.
939 if ((vp
->v_mount
->mnt_flag
& MNT_NOEXEC
) ||
940 ((attr
->va_mode
& 0111) == 0) ||
941 (attr
->va_type
!= VREG
)) {
946 * Zero length files can't be exec'd
948 if (attr
->va_size
== 0)
952 * Check for execute permission to file based on current credentials.
954 error
= VOP_ACCESS(vp
, VEXEC
, p
->p_ucred
);
959 * Check number of open-for-writes on the file and deny execution
962 if (vp
->v_writecount
)
966 * Call filesystem specific open routine, which allows us to read,
967 * write, and mmap the file. Without the VOP_OPEN we can only
970 error
= VOP_OPEN(vp
, FREAD
, p
->p_ucred
, NULL
);
978 * Exec handler registration
981 exec_register(const struct execsw
*execsw_arg
)
983 const struct execsw
**es
, **xs
, **newexecsw
;
984 int count
= 2; /* New slot and trailing NULL */
987 for (es
= execsw
; *es
; es
++)
989 newexecsw
= kmalloc(count
* sizeof(*es
), M_TEMP
, M_WAITOK
);
990 if (newexecsw
== NULL
)
994 for (es
= execsw
; *es
; es
++)
999 kfree(execsw
, M_TEMP
);
1005 exec_unregister(const struct execsw
*execsw_arg
)
1007 const struct execsw
**es
, **xs
, **newexecsw
;
1011 panic("unregister with no handlers left?");
1013 for (es
= execsw
; *es
; es
++) {
1014 if (*es
== execsw_arg
)
1019 for (es
= execsw
; *es
; es
++)
1020 if (*es
!= execsw_arg
)
1022 newexecsw
= kmalloc(count
* sizeof(*es
), M_TEMP
, M_WAITOK
);
1023 if (newexecsw
== NULL
)
1026 for (es
= execsw
; *es
; es
++)
1027 if (*es
!= execsw_arg
)
1031 kfree(execsw
, M_TEMP
);