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 static int ktrace_suid
= 0;
97 SYSCTL_INT(_kern
, OID_AUTO
, ktrace_suid
, CTLFLAG_RW
, &ktrace_suid
, 0, "");
99 void print_execve_args(struct image_args
*args
);
100 int debug_execve_args
= 0;
101 SYSCTL_INT(_kern
, OID_AUTO
, debug_execve_args
, CTLFLAG_RW
, &debug_execve_args
,
105 * Exec arguments object cache
107 static struct objcache
*exec_objcache
;
111 exec_objcache_init(void *arg __unused
)
115 cluster_limit
= 16; /* up to this many objects */
116 exec_objcache
= objcache_create_mbacked(
117 M_EXECARGS
, PATH_MAX
+ ARG_MAX
,
119 2, /* minimal magazine capacity */
122 SYSINIT(exec_objcache
, SI_BOOT2_MACHDEP
, SI_ORDER_ANY
, exec_objcache_init
, 0);
125 * stackgap_random specifies if the stackgap should have a random size added
126 * to it. It must be a power of 2. If non-zero, the stack gap will be
127 * calculated as: ALIGN(karc4random() & (stackgap_random - 1)).
129 static int stackgap_random
= 1024;
131 sysctl_kern_stackgap(SYSCTL_HANDLER_ARGS
)
134 new_val
= stackgap_random
;
135 error
= sysctl_handle_int(oidp
, &new_val
, 0, req
);
136 if (error
!= 0 || req
->newptr
== NULL
)
138 if ((new_val
< 0) || (new_val
> 16 * PAGE_SIZE
) || ! powerof2(new_val
))
140 stackgap_random
= new_val
;
145 SYSCTL_PROC(_kern
, OID_AUTO
, stackgap_random
, CTLFLAG_RW
|CTLTYPE_UINT
,
146 0, 0, sysctl_kern_stackgap
, "IU", "Max random stack gap (power of 2)");
149 print_execve_args(struct image_args
*args
)
154 cp
= args
->begin_argv
;
155 for (ndx
= 0; ndx
< args
->argc
; ndx
++) {
156 kprintf("\targv[%d]: %s\n", ndx
, cp
);
157 while (*cp
++ != '\0');
159 for (ndx
= 0; ndx
< args
->envc
; ndx
++) {
160 kprintf("\tenvv[%d]: %s\n", ndx
, cp
);
161 while (*cp
++ != '\0');
166 * Each of the items is a pointer to a `const struct execsw', hence the
167 * double pointer here.
169 static const struct execsw
**execsw
;
172 * Replace current vmspace with a new binary.
173 * Returns 0 on success, > 0 on recoverable error (use as errno).
174 * Returns -1 on lethal error which demands killing of the current
178 kern_execve(struct nlookupdata
*nd
, struct image_args
*args
)
180 struct thread
*td
= curthread
;
181 struct lwp
*lp
= td
->td_lwp
;
182 struct proc
*p
= td
->td_proc
;
183 register_t
*stack_base
;
185 struct image_params image_params
, *imgp
;
187 int (*img_first
) (struct image_params
*);
189 if (debug_execve_args
) {
190 kprintf("%s()\n", __func__
);
191 print_execve_args(args
);
195 imgp
= &image_params
;
198 * NOTE: P_INEXEC is handled by exec_new_vmspace() now. We make
199 * no modifications to the process at all until we get there.
201 * Note that multiple threads may be trying to exec at the same
202 * time. exec_new_vmspace() handles that too.
206 * Initialize part of the common data
211 imgp
->entry_addr
= 0;
213 imgp
->vmspace_destroyed
= 0;
214 imgp
->interpreted
= 0;
215 imgp
->interpreter_name
[0] = 0;
216 imgp
->auxargs
= NULL
;
218 imgp
->firstpage
= NULL
;
219 imgp
->ps_strings
= 0;
220 imgp
->image_header
= NULL
;
225 * Translate the file name to a vnode. Unlock the cache entry to
226 * improve parallelism for programs exec'd in parallel.
228 if ((error
= nlookup(nd
)) != 0)
230 error
= cache_vget(&nd
->nl_nch
, nd
->nl_cred
, LK_EXCLUSIVE
, &imgp
->vp
);
231 KKASSERT(nd
->nl_flags
& NLC_NCPISLOCKED
);
232 nd
->nl_flags
&= ~NLC_NCPISLOCKED
;
233 cache_unlock(&nd
->nl_nch
);
238 * Check file permissions (also 'opens' file)
240 error
= exec_check_permissions(imgp
);
243 goto exec_fail_dealloc
;
246 error
= exec_map_first_page(imgp
);
249 goto exec_fail_dealloc
;
251 if (debug_execve_args
&& imgp
->interpreted
) {
252 kprintf(" target is interpreted -- recursive pass\n");
253 kprintf(" interpreter: %s\n", imgp
->interpreter_name
);
254 print_execve_args(args
);
258 * If the current process has a special image activator it
259 * wants to try first, call it. For example, emulating shell
260 * scripts differently.
263 if ((img_first
= imgp
->proc
->p_sysent
->sv_imgact_try
) != NULL
)
264 error
= img_first(imgp
);
267 * If the vnode has a registered vmspace, exec the vmspace
269 if (error
== -1 && imgp
->vp
->v_resident
) {
270 error
= exec_resident_imgact(imgp
);
274 * Loop through the list of image activators, calling each one.
275 * An activator returns -1 if there is no match, 0 on success,
276 * and an error otherwise.
278 for (i
= 0; error
== -1 && execsw
[i
]; ++i
) {
279 if (execsw
[i
]->ex_imgact
== NULL
||
280 execsw
[i
]->ex_imgact
== img_first
) {
283 error
= (*execsw
[i
]->ex_imgact
)(imgp
);
289 goto exec_fail_dealloc
;
293 * Special interpreter operation, cleanup and loop up to try to
294 * activate the interpreter.
296 if (imgp
->interpreted
) {
297 exec_unmap_first_page(imgp
);
301 error
= nlookup_init(nd
, imgp
->interpreter_name
, UIO_SYSSPACE
,
309 * Copy out strings (args and env) and initialize stack base
311 stack_base
= exec_copyout_strings(imgp
);
312 p
->p_vmspace
->vm_minsaddr
= (char *)stack_base
;
315 * If custom stack fixup routine present for this process
316 * let it do the stack setup. If we are running a resident
317 * image there is no auxinfo or other image activator context
318 * so don't try to add fixups to the stack.
320 * Else stuff argument count as first item on stack
322 if (p
->p_sysent
->sv_fixup
&& imgp
->resident
== 0)
323 (*p
->p_sysent
->sv_fixup
)(&stack_base
, imgp
);
325 suword(--stack_base
, imgp
->args
->argc
);
328 * For security and other reasons, the file descriptor table cannot
329 * be shared after an exec.
331 if (p
->p_fd
->fd_refcnt
> 1) {
332 struct filedesc
*tmp
;
339 * For security and other reasons, signal handlers cannot
340 * be shared after an exec. The new proces gets a copy of the old
341 * handlers. In execsigs(), the new process will have its signals
344 if (p
->p_sigacts
->ps_refcnt
> 1) {
345 struct sigacts
*newsigacts
;
347 newsigacts
= (struct sigacts
*)kmalloc(sizeof(*newsigacts
),
348 M_SUBPROC
, M_WAITOK
);
349 bcopy(p
->p_sigacts
, newsigacts
, sizeof(*newsigacts
));
350 p
->p_sigacts
->ps_refcnt
--;
351 p
->p_sigacts
= newsigacts
;
352 p
->p_sigacts
->ps_refcnt
= 1;
356 * For security and other reasons virtual kernels cannot be
357 * inherited by an exec. This also allows a virtual kernel
358 * to fork/exec unrelated applications.
366 /* close files on exec */
369 /* reset caught signals */
372 /* name this process - nameiexec(p, ndp) */
373 len
= min(nd
->nl_nch
.ncp
->nc_nlen
, MAXCOMLEN
);
374 bcopy(nd
->nl_nch
.ncp
->nc_name
, p
->p_comm
, len
);
376 bcopy(p
->p_comm
, lp
->lwp_thread
->td_comm
, MAXCOMLEN
+1);
379 * mark as execed, wakeup the process that vforked (if any) and tell
380 * it that it now has its own resources back
383 if (p
->p_pptr
&& (p
->p_flag
& P_PPWAIT
)) {
384 p
->p_flag
&= ~P_PPWAIT
;
385 wakeup((caddr_t
)p
->p_pptr
);
389 * Implement image setuid/setgid.
391 * Don't honor setuid/setgid if the filesystem prohibits it or if
392 * the process is being traced.
394 if ((((attr
.va_mode
& VSUID
) && p
->p_ucred
->cr_uid
!= attr
.va_uid
) ||
395 ((attr
.va_mode
& VSGID
) && p
->p_ucred
->cr_gid
!= attr
.va_gid
)) &&
396 (imgp
->vp
->v_mount
->mnt_flag
& MNT_NOSUID
) == 0 &&
397 (p
->p_flag
& P_TRACED
) == 0) {
399 * Turn off syscall tracing for set-id programs, except for
400 * root. Record any set-id flags first to make sure that
401 * we do not regain any tracing during a possible block.
404 if (p
->p_tracenode
&& ktrace_suid
== 0 &&
405 priv_check(td
, PRIV_ROOT
) != 0) {
406 ktrdestroy(&p
->p_tracenode
);
409 /* Close any file descriptors 0..2 that reference procfs */
411 /* Make sure file descriptors 0..2 are in use. */
412 error
= fdcheckstd(p
);
414 goto exec_fail_dealloc
;
416 * Set the new credentials.
419 if (attr
.va_mode
& VSUID
)
420 change_euid(attr
.va_uid
);
421 if (attr
.va_mode
& VSGID
)
422 p
->p_ucred
->cr_gid
= attr
.va_gid
;
425 * Clear local varsym variables
427 varsymset_clean(&p
->p_varsymset
);
429 if (p
->p_ucred
->cr_uid
== p
->p_ucred
->cr_ruid
&&
430 p
->p_ucred
->cr_gid
== p
->p_ucred
->cr_rgid
)
431 p
->p_flag
&= ~P_SUGID
;
435 * Implement correct POSIX saved-id behavior.
437 if (p
->p_ucred
->cr_svuid
!= p
->p_ucred
->cr_uid
||
438 p
->p_ucred
->cr_svgid
!= p
->p_ucred
->cr_gid
) {
440 p
->p_ucred
->cr_svuid
= p
->p_ucred
->cr_uid
;
441 p
->p_ucred
->cr_svgid
= p
->p_ucred
->cr_gid
;
445 * Store the vp for use in procfs
447 if (p
->p_textvp
) /* release old reference */
449 p
->p_textvp
= imgp
->vp
;
453 * Notify others that we exec'd, and clear the P_INEXEC flag
454 * as we're now a bona fide freshly-execed process.
456 KNOTE(&p
->p_klist
, NOTE_EXEC
);
457 p
->p_flag
&= ~P_INEXEC
;
460 * If tracing the process, trap to debugger so breakpoints
461 * can be set before the program executes.
463 STOPEVENT(p
, S_EXEC
, 0);
465 if (p
->p_flag
& P_TRACED
)
468 /* clear "fork but no exec" flag, as we _are_ execing */
469 p
->p_acflag
&= ~AFORK
;
471 /* Set values passed into the program in registers. */
472 exec_setregs(imgp
->entry_addr
, (u_long
)(uintptr_t)stack_base
,
475 /* Set the access time on the vnode */
476 vn_mark_atime(imgp
->vp
, td
);
478 /* Free any previous argument cache */
479 if (p
->p_args
&& --p
->p_args
->ar_ref
== 0)
480 FREE(p
->p_args
, M_PARGS
);
483 /* Cache arguments if they fit inside our allowance */
484 i
= imgp
->args
->begin_envv
- imgp
->args
->begin_argv
;
485 if (ps_arg_cache_limit
>= i
+ sizeof(struct pargs
)) {
486 MALLOC(p
->p_args
, struct pargs
*, sizeof(struct pargs
) + i
,
488 p
->p_args
->ar_ref
= 1;
489 p
->p_args
->ar_length
= i
;
490 bcopy(imgp
->args
->begin_argv
, p
->p_args
->ar_args
, i
);
496 * free various allocated resources
499 exec_unmap_first_page(imgp
);
507 ++mycpu
->gd_cnt
.v_exec
;
513 * we're done here, clear P_INEXEC if we were the ones that
514 * set it. Otherwise if vmspace_destroyed is still set we
515 * raced another thread and that thread is responsible for
518 if (imgp
->vmspace_destroyed
& 2)
519 p
->p_flag
&= ~P_INEXEC
;
520 if (imgp
->vmspace_destroyed
) {
522 * Sorry, no more process anymore. exit gracefully.
523 * However we can't die right here, because our
524 * caller might have to clean up, so indicate a
525 * lethal error by returning -1.
534 * execve() system call.
537 sys_execve(struct execve_args
*uap
)
539 struct nlookupdata nd
;
540 struct image_args args
;
543 error
= nlookup_init(&nd
, uap
->fname
, UIO_USERSPACE
, NLC_FOLLOW
);
544 bzero(&args
, sizeof(args
));
546 error
= exec_copyin_args(&args
, uap
->fname
, PATH_USERSPACE
,
547 uap
->argv
, uap
->envv
);
550 error
= kern_execve(&nd
, &args
);
552 exec_free_args(&args
);
555 /* We hit a lethal error condition. Let's die now. */
556 exit1(W_EXITCODE(0, SIGABRT
));
561 * The syscall result is returned in registers to the new program.
562 * Linux will register %edx as an atexit function and we must be
563 * sure to set it to 0. XXX
566 uap
->sysmsg_result64
= 0;
572 exec_map_first_page(struct image_params
*imgp
)
576 vm_page_t ma
[VM_INITIAL_PAGEIN
];
581 exec_unmap_first_page(imgp
);
584 * The file has to be mappable.
586 if ((object
= imgp
->vp
->v_object
) == NULL
)
590 * We shouldn't need protection for vm_page_grab() but we certainly
591 * need it for the lookup loop below (lookup/busy race), since
592 * an interrupt can unbusy and free the page before our busy check.
595 m
= vm_page_grab(object
, 0, VM_ALLOC_NORMAL
| VM_ALLOC_RETRY
);
597 if ((m
->valid
& VM_PAGE_BITS_ALL
) != VM_PAGE_BITS_ALL
) {
599 initial_pagein
= VM_INITIAL_PAGEIN
;
600 if (initial_pagein
> object
->size
)
601 initial_pagein
= object
->size
;
602 for (i
= 1; i
< initial_pagein
; i
++) {
603 if ((m
= vm_page_lookup(object
, i
)) != NULL
) {
604 if ((m
->flags
& PG_BUSY
) || m
->busy
)
610 m
= vm_page_alloc(object
, i
, VM_ALLOC_NORMAL
);
619 * get_pages unbusies all the requested pages except the
620 * primary page (at index 0 in this case). The primary
621 * page may have been wired during the pagein (e.g. by
622 * the buffer cache) so vnode_pager_freepage() must be
623 * used to properly release it.
625 rv
= vm_pager_get_pages(object
, ma
, initial_pagein
, 0);
626 m
= vm_page_lookup(object
, 0);
628 if (rv
!= VM_PAGER_OK
|| m
== NULL
|| m
->valid
== 0) {
630 vm_page_protect(m
, VM_PROT_NONE
);
631 vnode_pager_freepage(m
);
638 vm_page_wakeup(m
); /* unbusy the page */
641 imgp
->firstpage
= sf_buf_alloc(m
, SFB_CPUPRIVATE
);
642 imgp
->image_header
= (void *)sf_buf_kva(imgp
->firstpage
);
648 exec_unmap_first_page(struct image_params
*imgp
)
653 if (imgp
->firstpage
!= NULL
) {
654 m
= sf_buf_page(imgp
->firstpage
);
655 sf_buf_free(imgp
->firstpage
);
656 imgp
->firstpage
= NULL
;
657 imgp
->image_header
= NULL
;
664 * Destroy old address space, and allocate a new stack
665 * The new stack is only SGROWSIZ large because it is grown
666 * automatically in trap.c.
668 * This is the point of no return.
671 exec_new_vmspace(struct image_params
*imgp
, struct vmspace
*vmcopy
)
673 struct vmspace
*vmspace
= imgp
->proc
->p_vmspace
;
674 vm_offset_t stack_addr
= USRSTACK
- maxssiz
;
680 * Indicate that we cannot gracefully error out any more, kill
681 * any other threads present, and set P_INEXEC to indicate that
682 * we are now messing with the process structure proper.
684 * If killalllwps() races return an error which coupled with
685 * vmspace_destroyed will cause us to exit. This is what we
686 * want since another thread is patiently waiting for us to exit
690 imgp
->vmspace_destroyed
= 1;
692 if (curthread
->td_proc
->p_nthreads
> 1) {
693 error
= killalllwps(1);
697 imgp
->vmspace_destroyed
|= 2; /* we are responsible for P_INEXEC */
698 p
->p_flag
|= P_INEXEC
;
701 * Prevent a pending AIO from modifying the new address space.
703 aio_proc_rundown(imgp
->proc
);
706 * Blow away entire process VM, if address space not shared,
707 * otherwise, create a new VM space so that other threads are
708 * not disrupted. If we are execing a resident vmspace we
709 * create a duplicate of it and remap the stack.
711 * The exitingcnt test is not strictly necessary but has been
712 * included for code sanity (to make the code more deterministic).
714 map
= &vmspace
->vm_map
;
716 vmspace_exec(imgp
->proc
, vmcopy
);
717 vmspace
= imgp
->proc
->p_vmspace
;
718 pmap_remove_pages(vmspace_pmap(vmspace
), stack_addr
, USRSTACK
);
719 map
= &vmspace
->vm_map
;
720 } else if (vmspace
->vm_sysref
.refcnt
== 1 &&
721 vmspace
->vm_exitingcnt
== 0) {
723 if (vmspace
->vm_upcalls
)
724 upc_release(vmspace
, ONLY_LWP_IN_PROC(imgp
->proc
));
725 pmap_remove_pages(vmspace_pmap(vmspace
),
726 0, VM_MAX_USER_ADDRESS
);
727 vm_map_remove(map
, 0, VM_MAX_USER_ADDRESS
);
729 vmspace_exec(imgp
->proc
, NULL
);
730 vmspace
= imgp
->proc
->p_vmspace
;
731 map
= &vmspace
->vm_map
;
734 /* Allocate a new stack */
735 error
= vm_map_stack(&vmspace
->vm_map
, stack_addr
, (vm_size_t
)maxssiz
,
736 0, VM_PROT_ALL
, VM_PROT_ALL
, 0);
740 /* vm_ssize and vm_maxsaddr are somewhat antiquated concepts in the
741 * VM_STACK case, but they are still used to monitor the size of the
742 * process stack so we can check the stack rlimit.
744 vmspace
->vm_ssize
= sgrowsiz
>> PAGE_SHIFT
;
745 vmspace
->vm_maxsaddr
= (char *)USRSTACK
- maxssiz
;
751 * Copy out argument and environment strings from the old process
752 * address space into the temporary string buffer.
755 exec_copyin_args(struct image_args
*args
, char *fname
,
756 enum exec_path_segflg segflg
, char **argv
, char **envv
)
762 args
->buf
= objcache_get(exec_objcache
, M_WAITOK
);
763 if (args
->buf
== NULL
)
765 args
->begin_argv
= args
->buf
;
766 args
->endp
= args
->begin_argv
;
767 args
->space
= ARG_MAX
;
769 args
->fname
= args
->buf
+ ARG_MAX
;
772 * Copy the file name.
774 if (segflg
== PATH_SYSSPACE
) {
775 error
= copystr(fname
, args
->fname
, PATH_MAX
, &length
);
776 } else if (segflg
== PATH_USERSPACE
) {
777 error
= copyinstr(fname
, args
->fname
, PATH_MAX
, &length
);
781 * Extract argument strings. argv may not be NULL. The argv
782 * array is terminated by a NULL entry. We special-case the
783 * situation where argv[0] is NULL by passing { filename, NULL }
784 * to the new program to guarentee that the interpreter knows what
785 * file to open in case we exec an interpreted file. Note that
786 * a NULL argv[0] terminates the argv[] array.
788 * XXX the special-casing of argv[0] is historical and needs to be
794 while ((argp
= (caddr_t
)(intptr_t)fuword(argv
++)) != NULL
) {
795 if (argp
== (caddr_t
)-1) {
799 error
= copyinstr(argp
, args
->endp
,
800 args
->space
, &length
);
802 if (error
== ENAMETOOLONG
)
806 args
->space
-= length
;
807 args
->endp
+= length
;
810 if (args
->argc
== 0 && error
== 0) {
811 length
= strlen(args
->fname
) + 1;
812 if (length
> args
->space
) {
815 bcopy(args
->fname
, args
->endp
, length
);
816 args
->space
-= length
;
817 args
->endp
+= length
;
823 args
->begin_envv
= args
->endp
;
826 * extract environment strings. envv may be NULL.
828 if (envv
&& error
== 0) {
829 while ((envp
= (caddr_t
) (intptr_t) fuword(envv
++))) {
830 if (envp
== (caddr_t
) -1) {
834 error
= copyinstr(envp
, args
->endp
, args
->space
,
837 if (error
== ENAMETOOLONG
)
841 args
->space
-= length
;
842 args
->endp
+= length
;
850 exec_free_args(struct image_args
*args
)
853 objcache_put(exec_objcache
, args
->buf
);
859 * Copy strings out to the new process address space, constructing
860 * new arg and env vector tables. Return a pointer to the base
861 * so that it can be used as the initial stack pointer.
864 exec_copyout_strings(struct image_params
*imgp
)
866 int argc
, envc
, sgap
;
868 char *stringp
, *destp
;
869 register_t
*stack_base
;
870 struct ps_strings
*arginfo
;
874 * Calculate string base and vector table pointers.
875 * Also deal with signal trampoline code for this exec type.
877 arginfo
= (struct ps_strings
*)PS_STRINGS
;
878 szsigcode
= *(imgp
->proc
->p_sysent
->sv_szsigcode
);
879 if (stackgap_random
!= 0)
880 sgap
= ALIGN(karc4random() & (stackgap_random
- 1));
883 destp
= (caddr_t
)arginfo
- szsigcode
- SPARE_USRSPACE
- sgap
-
884 roundup((ARG_MAX
- imgp
->args
->space
), sizeof(char *));
890 copyout(imgp
->proc
->p_sysent
->sv_sigcode
,
891 ((caddr_t
)arginfo
- szsigcode
), szsigcode
);
894 * If we have a valid auxargs ptr, prepare some room
897 * The '+ 2' is for the null pointers at the end of each of the
898 * arg and env vector sets, and 'AT_COUNT*2' is room for the
902 vectp
= (char **)(destp
- (imgp
->args
->argc
+
903 imgp
->args
->envc
+ 2 + AT_COUNT
* 2) * sizeof(char*));
905 vectp
= (char **)(destp
- (imgp
->args
->argc
+
906 imgp
->args
->envc
+ 2) * sizeof(char*));
910 * NOTE: don't bother aligning the stack here for GCC 2.x, it will
911 * be done in crt1.o. Note that GCC 3.x aligns the stack in main.
915 * vectp also becomes our initial stack base
917 stack_base
= (register_t
*)vectp
;
919 stringp
= imgp
->args
->begin_argv
;
920 argc
= imgp
->args
->argc
;
921 envc
= imgp
->args
->envc
;
924 * Copy out strings - arguments and environment.
926 copyout(stringp
, destp
, ARG_MAX
- imgp
->args
->space
);
929 * Fill in "ps_strings" struct for ps, w, etc.
931 suword(&arginfo
->ps_argvstr
, (long)(intptr_t)vectp
);
932 suword(&arginfo
->ps_nargvstr
, argc
);
935 * Fill in argument portion of vector table.
937 for (; argc
> 0; --argc
) {
938 suword(vectp
++, (long)(intptr_t)destp
);
939 while (*stringp
++ != 0)
944 /* a null vector table pointer separates the argp's from the envp's */
947 suword(&arginfo
->ps_envstr
, (long)(intptr_t)vectp
);
948 suword(&arginfo
->ps_nenvstr
, envc
);
951 * Fill in environment portion of vector table.
953 for (; envc
> 0; --envc
) {
954 suword(vectp
++, (long)(intptr_t)destp
);
955 while (*stringp
++ != 0)
960 /* end of vector table is a null pointer */
967 * Check permissions of file to execute.
968 * Return 0 for success or error code on failure.
971 exec_check_permissions(struct image_params
*imgp
)
973 struct proc
*p
= imgp
->proc
;
974 struct vnode
*vp
= imgp
->vp
;
975 struct vattr
*attr
= imgp
->attr
;
978 /* Get file attributes */
979 error
= VOP_GETATTR(vp
, attr
);
984 * 1) Check if file execution is disabled for the filesystem that this
986 * 2) Insure that at least one execute bit is on - otherwise root
987 * will always succeed, and we don't want to happen unless the
988 * file really is executable.
989 * 3) Insure that the file is a regular file.
991 if ((vp
->v_mount
->mnt_flag
& MNT_NOEXEC
) ||
992 ((attr
->va_mode
& 0111) == 0) ||
993 (attr
->va_type
!= VREG
)) {
998 * Zero length files can't be exec'd
1000 if (attr
->va_size
== 0)
1004 * Check for execute permission to file based on current credentials.
1006 error
= VOP_EACCESS(vp
, VEXEC
, p
->p_ucred
);
1011 * Check number of open-for-writes on the file and deny execution
1014 if (vp
->v_writecount
)
1018 * Call filesystem specific open routine, which allows us to read,
1019 * write, and mmap the file. Without the VOP_OPEN we can only
1022 error
= VOP_OPEN(vp
, FREAD
, p
->p_ucred
, NULL
);
1030 * Exec handler registration
1033 exec_register(const struct execsw
*execsw_arg
)
1035 const struct execsw
**es
, **xs
, **newexecsw
;
1036 int count
= 2; /* New slot and trailing NULL */
1039 for (es
= execsw
; *es
; es
++)
1041 newexecsw
= kmalloc(count
* sizeof(*es
), M_TEMP
, M_WAITOK
);
1044 for (es
= execsw
; *es
; es
++)
1049 kfree(execsw
, M_TEMP
);
1055 exec_unregister(const struct execsw
*execsw_arg
)
1057 const struct execsw
**es
, **xs
, **newexecsw
;
1061 panic("unregister with no handlers left?");
1063 for (es
= execsw
; *es
; es
++) {
1064 if (*es
== execsw_arg
)
1069 for (es
= execsw
; *es
; es
++)
1070 if (*es
!= execsw_arg
)
1072 newexecsw
= kmalloc(count
* sizeof(*es
), M_TEMP
, M_WAITOK
);
1074 for (es
= execsw
; *es
; es
++)
1075 if (*es
!= execsw_arg
)
1079 kfree(execsw
, M_TEMP
);