4 * Copyright (C) 1991, 1992 Linus Torvalds
8 * #!-checking implemented by tytso.
11 * Demand-loading implemented 01.12.91 - no need to read anything but
12 * the header into memory. The inode of the executable is put into
13 * "current->executable", and page faults do the actual loading. Clean.
15 * Once more I can proudly say that linux stood up to being changed: it
16 * was less than 2 hours work to get demand-loading completely implemented.
18 * Demand loading changed July 1993 by Eric Youngdale. Use mmap instead,
19 * current->executable is only used by the procfs. This allows a dispatch
20 * table to check for several different types of binary formats. We keep
21 * trying until we recognize the file or we run out of supported binary
25 #include <linux/slab.h>
26 #include <linux/file.h>
27 #include <linux/fdtable.h>
29 #include <linux/vmacache.h>
30 #include <linux/stat.h>
31 #include <linux/fcntl.h>
32 #include <linux/swap.h>
33 #include <linux/string.h>
34 #include <linux/init.h>
35 #include <linux/pagemap.h>
36 #include <linux/perf_event.h>
37 #include <linux/highmem.h>
38 #include <linux/spinlock.h>
39 #include <linux/key.h>
40 #include <linux/personality.h>
41 #include <linux/binfmts.h>
42 #include <linux/utsname.h>
43 #include <linux/pid_namespace.h>
44 #include <linux/module.h>
45 #include <linux/namei.h>
46 #include <linux/mount.h>
47 #include <linux/security.h>
48 #include <linux/syscalls.h>
49 #include <linux/tsacct_kern.h>
50 #include <linux/cn_proc.h>
51 #include <linux/audit.h>
52 #include <linux/tracehook.h>
53 #include <linux/kmod.h>
54 #include <linux/fsnotify.h>
55 #include <linux/fs_struct.h>
56 #include <linux/pipe_fs_i.h>
57 #include <linux/oom.h>
58 #include <linux/compat.h>
59 #include <linux/vmalloc.h>
61 #include <trace/events/fs.h>
63 #include <linux/uaccess.h>
64 #include <asm/mmu_context.h>
67 #include <trace/events/task.h>
70 #include <trace/events/sched.h>
72 int suid_dumpable
= 0;
74 static LIST_HEAD(formats
);
75 static DEFINE_RWLOCK(binfmt_lock
);
77 void __register_binfmt(struct linux_binfmt
* fmt
, int insert
)
80 if (WARN_ON(!fmt
->load_binary
))
82 write_lock(&binfmt_lock
);
83 insert
? list_add(&fmt
->lh
, &formats
) :
84 list_add_tail(&fmt
->lh
, &formats
);
85 write_unlock(&binfmt_lock
);
88 EXPORT_SYMBOL(__register_binfmt
);
90 void unregister_binfmt(struct linux_binfmt
* fmt
)
92 write_lock(&binfmt_lock
);
94 write_unlock(&binfmt_lock
);
97 EXPORT_SYMBOL(unregister_binfmt
);
99 static inline void put_binfmt(struct linux_binfmt
* fmt
)
101 module_put(fmt
->module
);
104 bool path_noexec(const struct path
*path
)
106 return (path
->mnt
->mnt_flags
& MNT_NOEXEC
) ||
107 (path
->mnt
->mnt_sb
->s_iflags
& SB_I_NOEXEC
);
110 bool path_nosuid(const struct path
*path
)
112 return !mnt_may_suid(path
->mnt
) ||
113 (path
->mnt
->mnt_sb
->s_iflags
& SB_I_NOSUID
);
115 EXPORT_SYMBOL(path_nosuid
);
119 * Note that a shared library must be both readable and executable due to
122 * Also note that we take the address to load from from the file itself.
124 SYSCALL_DEFINE1(uselib
, const char __user
*, library
)
126 struct linux_binfmt
*fmt
;
128 struct filename
*tmp
= getname(library
);
129 int error
= PTR_ERR(tmp
);
130 static const struct open_flags uselib_flags
= {
131 .open_flag
= O_LARGEFILE
| O_RDONLY
| __FMODE_EXEC
,
132 .acc_mode
= MAY_READ
| MAY_EXEC
,
133 .intent
= LOOKUP_OPEN
,
134 .lookup_flags
= LOOKUP_FOLLOW
,
140 file
= do_filp_open(AT_FDCWD
, tmp
, &uselib_flags
);
142 error
= PTR_ERR(file
);
147 if (!S_ISREG(file_inode(file
)->i_mode
))
151 if (path_noexec(&file
->f_path
))
158 read_lock(&binfmt_lock
);
159 list_for_each_entry(fmt
, &formats
, lh
) {
160 if (!fmt
->load_shlib
)
162 if (!try_module_get(fmt
->module
))
164 read_unlock(&binfmt_lock
);
165 error
= fmt
->load_shlib(file
);
166 read_lock(&binfmt_lock
);
168 if (error
!= -ENOEXEC
)
171 read_unlock(&binfmt_lock
);
177 #endif /* #ifdef CONFIG_USELIB */
181 * The nascent bprm->mm is not visible until exec_mmap() but it can
182 * use a lot of memory, account these pages in current->mm temporary
183 * for oom_badness()->get_mm_rss(). Once exec succeeds or fails, we
184 * change the counter back via acct_arg_size(0).
186 static void acct_arg_size(struct linux_binprm
*bprm
, unsigned long pages
)
188 struct mm_struct
*mm
= current
->mm
;
189 long diff
= (long)(pages
- bprm
->vma_pages
);
194 bprm
->vma_pages
= pages
;
195 add_mm_counter(mm
, MM_ANONPAGES
, diff
);
198 static struct page
*get_arg_page(struct linux_binprm
*bprm
, unsigned long pos
,
203 unsigned int gup_flags
= FOLL_FORCE
;
205 #ifdef CONFIG_STACK_GROWSUP
207 ret
= expand_downwards(bprm
->vma
, pos
);
214 gup_flags
|= FOLL_WRITE
;
217 * We are doing an exec(). 'current' is the process
218 * doing the exec and bprm->mm is the new process's mm.
220 ret
= get_user_pages_remote(current
, bprm
->mm
, pos
, 1, gup_flags
,
226 unsigned long size
= bprm
->vma
->vm_end
- bprm
->vma
->vm_start
;
229 acct_arg_size(bprm
, size
/ PAGE_SIZE
);
232 * We've historically supported up to 32 pages (ARG_MAX)
233 * of argument strings even with small stacks
239 * Limit to 1/4-th the stack size for the argv+env strings.
241 * - the remaining binfmt code will not run out of stack space,
242 * - the program will have a reasonable amount of stack left
245 rlim
= current
->signal
->rlim
;
246 if (size
> ACCESS_ONCE(rlim
[RLIMIT_STACK
].rlim_cur
) / 4) {
255 static void put_arg_page(struct page
*page
)
260 static void free_arg_pages(struct linux_binprm
*bprm
)
264 static void flush_arg_page(struct linux_binprm
*bprm
, unsigned long pos
,
267 flush_cache_page(bprm
->vma
, pos
, page_to_pfn(page
));
270 static int __bprm_mm_init(struct linux_binprm
*bprm
)
273 struct vm_area_struct
*vma
= NULL
;
274 struct mm_struct
*mm
= bprm
->mm
;
276 bprm
->vma
= vma
= kmem_cache_zalloc(vm_area_cachep
, GFP_KERNEL
);
280 if (down_write_killable(&mm
->mmap_sem
)) {
287 * Place the stack at the largest stack address the architecture
288 * supports. Later, we'll move this to an appropriate place. We don't
289 * use STACK_TOP because that can depend on attributes which aren't
292 BUILD_BUG_ON(VM_STACK_FLAGS
& VM_STACK_INCOMPLETE_SETUP
);
293 vma
->vm_end
= STACK_TOP_MAX
;
294 vma
->vm_start
= vma
->vm_end
- PAGE_SIZE
;
295 vma
->vm_flags
= VM_SOFTDIRTY
| VM_STACK_FLAGS
| VM_STACK_INCOMPLETE_SETUP
;
296 vma
->vm_page_prot
= vm_get_page_prot(vma
->vm_flags
);
297 INIT_LIST_HEAD(&vma
->anon_vma_chain
);
299 err
= insert_vm_struct(mm
, vma
);
303 mm
->stack_vm
= mm
->total_vm
= 1;
304 arch_bprm_mm_init(mm
, vma
);
305 up_write(&mm
->mmap_sem
);
306 bprm
->p
= vma
->vm_end
- sizeof(void *);
309 up_write(&mm
->mmap_sem
);
312 kmem_cache_free(vm_area_cachep
, vma
);
316 static bool valid_arg_len(struct linux_binprm
*bprm
, long len
)
318 return len
<= MAX_ARG_STRLEN
;
323 static inline void acct_arg_size(struct linux_binprm
*bprm
, unsigned long pages
)
327 static struct page
*get_arg_page(struct linux_binprm
*bprm
, unsigned long pos
,
332 page
= bprm
->page
[pos
/ PAGE_SIZE
];
333 if (!page
&& write
) {
334 page
= alloc_page(GFP_HIGHUSER
|__GFP_ZERO
);
337 bprm
->page
[pos
/ PAGE_SIZE
] = page
;
343 static void put_arg_page(struct page
*page
)
347 static void free_arg_page(struct linux_binprm
*bprm
, int i
)
350 __free_page(bprm
->page
[i
]);
351 bprm
->page
[i
] = NULL
;
355 static void free_arg_pages(struct linux_binprm
*bprm
)
359 for (i
= 0; i
< MAX_ARG_PAGES
; i
++)
360 free_arg_page(bprm
, i
);
363 static void flush_arg_page(struct linux_binprm
*bprm
, unsigned long pos
,
368 static int __bprm_mm_init(struct linux_binprm
*bprm
)
370 bprm
->p
= PAGE_SIZE
* MAX_ARG_PAGES
- sizeof(void *);
374 static bool valid_arg_len(struct linux_binprm
*bprm
, long len
)
376 return len
<= bprm
->p
;
379 #endif /* CONFIG_MMU */
382 * Create a new mm_struct and populate it with a temporary stack
383 * vm_area_struct. We don't have enough context at this point to set the stack
384 * flags, permissions, and offset, so we use temporary values. We'll update
385 * them later in setup_arg_pages().
387 static int bprm_mm_init(struct linux_binprm
*bprm
)
390 struct mm_struct
*mm
= NULL
;
392 bprm
->mm
= mm
= mm_alloc();
397 err
= __bprm_mm_init(bprm
);
412 struct user_arg_ptr
{
417 const char __user
*const __user
*native
;
419 const compat_uptr_t __user
*compat
;
424 static const char __user
*get_user_arg_ptr(struct user_arg_ptr argv
, int nr
)
426 const char __user
*native
;
429 if (unlikely(argv
.is_compat
)) {
430 compat_uptr_t compat
;
432 if (get_user(compat
, argv
.ptr
.compat
+ nr
))
433 return ERR_PTR(-EFAULT
);
435 return compat_ptr(compat
);
439 if (get_user(native
, argv
.ptr
.native
+ nr
))
440 return ERR_PTR(-EFAULT
);
446 * count() counts the number of strings in array ARGV.
448 static int count(struct user_arg_ptr argv
, int max
)
452 if (argv
.ptr
.native
!= NULL
) {
454 const char __user
*p
= get_user_arg_ptr(argv
, i
);
466 if (fatal_signal_pending(current
))
467 return -ERESTARTNOHAND
;
475 * 'copy_strings()' copies argument/environment strings from the old
476 * processes's memory to the new process's stack. The call to get_user_pages()
477 * ensures the destination page is created and not swapped out.
479 static int copy_strings(int argc
, struct user_arg_ptr argv
,
480 struct linux_binprm
*bprm
)
482 struct page
*kmapped_page
= NULL
;
484 unsigned long kpos
= 0;
488 const char __user
*str
;
493 str
= get_user_arg_ptr(argv
, argc
);
497 len
= strnlen_user(str
, MAX_ARG_STRLEN
);
502 if (!valid_arg_len(bprm
, len
))
505 /* We're going to work our way backwords. */
511 int offset
, bytes_to_copy
;
513 if (fatal_signal_pending(current
)) {
514 ret
= -ERESTARTNOHAND
;
519 offset
= pos
% PAGE_SIZE
;
523 bytes_to_copy
= offset
;
524 if (bytes_to_copy
> len
)
527 offset
-= bytes_to_copy
;
528 pos
-= bytes_to_copy
;
529 str
-= bytes_to_copy
;
530 len
-= bytes_to_copy
;
532 if (!kmapped_page
|| kpos
!= (pos
& PAGE_MASK
)) {
535 page
= get_arg_page(bprm
, pos
, 1);
542 flush_kernel_dcache_page(kmapped_page
);
543 kunmap(kmapped_page
);
544 put_arg_page(kmapped_page
);
547 kaddr
= kmap(kmapped_page
);
548 kpos
= pos
& PAGE_MASK
;
549 flush_arg_page(bprm
, kpos
, kmapped_page
);
551 if (copy_from_user(kaddr
+offset
, str
, bytes_to_copy
)) {
560 flush_kernel_dcache_page(kmapped_page
);
561 kunmap(kmapped_page
);
562 put_arg_page(kmapped_page
);
568 * Like copy_strings, but get argv and its values from kernel memory.
570 int copy_strings_kernel(int argc
, const char *const *__argv
,
571 struct linux_binprm
*bprm
)
574 mm_segment_t oldfs
= get_fs();
575 struct user_arg_ptr argv
= {
576 .ptr
.native
= (const char __user
*const __user
*)__argv
,
580 r
= copy_strings(argc
, argv
, bprm
);
585 EXPORT_SYMBOL(copy_strings_kernel
);
590 * During bprm_mm_init(), we create a temporary stack at STACK_TOP_MAX. Once
591 * the binfmt code determines where the new stack should reside, we shift it to
592 * its final location. The process proceeds as follows:
594 * 1) Use shift to calculate the new vma endpoints.
595 * 2) Extend vma to cover both the old and new ranges. This ensures the
596 * arguments passed to subsequent functions are consistent.
597 * 3) Move vma's page tables to the new range.
598 * 4) Free up any cleared pgd range.
599 * 5) Shrink the vma to cover only the new range.
601 static int shift_arg_pages(struct vm_area_struct
*vma
, unsigned long shift
)
603 struct mm_struct
*mm
= vma
->vm_mm
;
604 unsigned long old_start
= vma
->vm_start
;
605 unsigned long old_end
= vma
->vm_end
;
606 unsigned long length
= old_end
- old_start
;
607 unsigned long new_start
= old_start
- shift
;
608 unsigned long new_end
= old_end
- shift
;
609 struct mmu_gather tlb
;
611 BUG_ON(new_start
> new_end
);
614 * ensure there are no vmas between where we want to go
617 if (vma
!= find_vma(mm
, new_start
))
621 * cover the whole range: [new_start, old_end)
623 if (vma_adjust(vma
, new_start
, old_end
, vma
->vm_pgoff
, NULL
))
627 * move the page tables downwards, on failure we rely on
628 * process cleanup to remove whatever mess we made.
630 if (length
!= move_page_tables(vma
, old_start
,
631 vma
, new_start
, length
, false))
635 tlb_gather_mmu(&tlb
, mm
, old_start
, old_end
);
636 if (new_end
> old_start
) {
638 * when the old and new regions overlap clear from new_end.
640 free_pgd_range(&tlb
, new_end
, old_end
, new_end
,
641 vma
->vm_next
? vma
->vm_next
->vm_start
: USER_PGTABLES_CEILING
);
644 * otherwise, clean from old_start; this is done to not touch
645 * the address space in [new_end, old_start) some architectures
646 * have constraints on va-space that make this illegal (IA64) -
647 * for the others its just a little faster.
649 free_pgd_range(&tlb
, old_start
, old_end
, new_end
,
650 vma
->vm_next
? vma
->vm_next
->vm_start
: USER_PGTABLES_CEILING
);
652 tlb_finish_mmu(&tlb
, old_start
, old_end
);
655 * Shrink the vma to just the new range. Always succeeds.
657 vma_adjust(vma
, new_start
, new_end
, vma
->vm_pgoff
, NULL
);
663 * Finalizes the stack vm_area_struct. The flags and permissions are updated,
664 * the stack is optionally relocated, and some extra space is added.
666 int setup_arg_pages(struct linux_binprm
*bprm
,
667 unsigned long stack_top
,
668 int executable_stack
)
671 unsigned long stack_shift
;
672 struct mm_struct
*mm
= current
->mm
;
673 struct vm_area_struct
*vma
= bprm
->vma
;
674 struct vm_area_struct
*prev
= NULL
;
675 unsigned long vm_flags
;
676 unsigned long stack_base
;
677 unsigned long stack_size
;
678 unsigned long stack_expand
;
679 unsigned long rlim_stack
;
681 #ifdef CONFIG_STACK_GROWSUP
682 /* Limit stack size */
683 stack_base
= rlimit_max(RLIMIT_STACK
);
684 if (stack_base
> STACK_SIZE_MAX
)
685 stack_base
= STACK_SIZE_MAX
;
687 /* Add space for stack randomization. */
688 stack_base
+= (STACK_RND_MASK
<< PAGE_SHIFT
);
690 /* Make sure we didn't let the argument array grow too large. */
691 if (vma
->vm_end
- vma
->vm_start
> stack_base
)
694 stack_base
= PAGE_ALIGN(stack_top
- stack_base
);
696 stack_shift
= vma
->vm_start
- stack_base
;
697 mm
->arg_start
= bprm
->p
- stack_shift
;
698 bprm
->p
= vma
->vm_end
- stack_shift
;
700 stack_top
= arch_align_stack(stack_top
);
701 stack_top
= PAGE_ALIGN(stack_top
);
703 if (unlikely(stack_top
< mmap_min_addr
) ||
704 unlikely(vma
->vm_end
- vma
->vm_start
>= stack_top
- mmap_min_addr
))
707 stack_shift
= vma
->vm_end
- stack_top
;
709 bprm
->p
-= stack_shift
;
710 mm
->arg_start
= bprm
->p
;
714 bprm
->loader
-= stack_shift
;
715 bprm
->exec
-= stack_shift
;
717 if (down_write_killable(&mm
->mmap_sem
))
720 vm_flags
= VM_STACK_FLAGS
;
723 * Adjust stack execute permissions; explicitly enable for
724 * EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X and leave alone
725 * (arch default) otherwise.
727 if (unlikely(executable_stack
== EXSTACK_ENABLE_X
))
729 else if (executable_stack
== EXSTACK_DISABLE_X
)
730 vm_flags
&= ~VM_EXEC
;
731 vm_flags
|= mm
->def_flags
;
732 vm_flags
|= VM_STACK_INCOMPLETE_SETUP
;
734 ret
= mprotect_fixup(vma
, &prev
, vma
->vm_start
, vma
->vm_end
,
740 /* Move stack pages down in memory. */
742 ret
= shift_arg_pages(vma
, stack_shift
);
747 /* mprotect_fixup is overkill to remove the temporary stack flags */
748 vma
->vm_flags
&= ~VM_STACK_INCOMPLETE_SETUP
;
750 stack_expand
= 131072UL; /* randomly 32*4k (or 2*64k) pages */
751 stack_size
= vma
->vm_end
- vma
->vm_start
;
753 * Align this down to a page boundary as expand_stack
756 rlim_stack
= rlimit(RLIMIT_STACK
) & PAGE_MASK
;
757 #ifdef CONFIG_STACK_GROWSUP
758 if (stack_size
+ stack_expand
> rlim_stack
)
759 stack_base
= vma
->vm_start
+ rlim_stack
;
761 stack_base
= vma
->vm_end
+ stack_expand
;
763 if (stack_size
+ stack_expand
> rlim_stack
)
764 stack_base
= vma
->vm_end
- rlim_stack
;
766 stack_base
= vma
->vm_start
- stack_expand
;
768 current
->mm
->start_stack
= bprm
->p
;
769 ret
= expand_stack(vma
, stack_base
);
774 up_write(&mm
->mmap_sem
);
777 EXPORT_SYMBOL(setup_arg_pages
);
782 * Transfer the program arguments and environment from the holding pages
783 * onto the stack. The provided stack pointer is adjusted accordingly.
785 int transfer_args_to_stack(struct linux_binprm
*bprm
,
786 unsigned long *sp_location
)
788 unsigned long index
, stop
, sp
;
791 stop
= bprm
->p
>> PAGE_SHIFT
;
794 for (index
= MAX_ARG_PAGES
- 1; index
>= stop
; index
--) {
795 unsigned int offset
= index
== stop
? bprm
->p
& ~PAGE_MASK
: 0;
796 char *src
= kmap(bprm
->page
[index
]) + offset
;
797 sp
-= PAGE_SIZE
- offset
;
798 if (copy_to_user((void *) sp
, src
, PAGE_SIZE
- offset
) != 0)
800 kunmap(bprm
->page
[index
]);
810 EXPORT_SYMBOL(transfer_args_to_stack
);
812 #endif /* CONFIG_MMU */
814 static struct file
*do_open_execat(int fd
, struct filename
*name
, int flags
)
818 struct open_flags open_exec_flags
= {
819 .open_flag
= O_LARGEFILE
| O_RDONLY
| __FMODE_EXEC
,
820 .acc_mode
= MAY_EXEC
,
821 .intent
= LOOKUP_OPEN
,
822 .lookup_flags
= LOOKUP_FOLLOW
,
825 if ((flags
& ~(AT_SYMLINK_NOFOLLOW
| AT_EMPTY_PATH
)) != 0)
826 return ERR_PTR(-EINVAL
);
827 if (flags
& AT_SYMLINK_NOFOLLOW
)
828 open_exec_flags
.lookup_flags
&= ~LOOKUP_FOLLOW
;
829 if (flags
& AT_EMPTY_PATH
)
830 open_exec_flags
.lookup_flags
|= LOOKUP_EMPTY
;
832 file
= do_filp_open(fd
, name
, &open_exec_flags
);
837 if (!S_ISREG(file_inode(file
)->i_mode
))
840 if (path_noexec(&file
->f_path
))
843 err
= deny_write_access(file
);
847 if (name
->name
[0] != '\0')
850 trace_open_exec(name
->name
);
860 struct file
*open_exec(const char *name
)
862 struct filename
*filename
= getname_kernel(name
);
863 struct file
*f
= ERR_CAST(filename
);
865 if (!IS_ERR(filename
)) {
866 f
= do_open_execat(AT_FDCWD
, filename
, 0);
871 EXPORT_SYMBOL(open_exec
);
873 int kernel_read(struct file
*file
, loff_t offset
,
874 char *addr
, unsigned long count
)
882 /* The cast to a user pointer is valid due to the set_fs() */
883 result
= vfs_read(file
, (void __user
*)addr
, count
, &pos
);
888 EXPORT_SYMBOL(kernel_read
);
890 int kernel_read_file(struct file
*file
, void **buf
, loff_t
*size
,
891 loff_t max_size
, enum kernel_read_file_id id
)
897 if (!S_ISREG(file_inode(file
)->i_mode
) || max_size
< 0)
900 ret
= security_kernel_read_file(file
, id
);
904 ret
= deny_write_access(file
);
908 i_size
= i_size_read(file_inode(file
));
909 if (max_size
> 0 && i_size
> max_size
) {
918 if (id
!= READING_FIRMWARE_PREALLOC_BUFFER
)
919 *buf
= vmalloc(i_size
);
926 while (pos
< i_size
) {
927 bytes
= kernel_read(file
, pos
, (char *)(*buf
) + pos
,
944 ret
= security_kernel_post_read_file(file
, *buf
, i_size
, id
);
950 if (id
!= READING_FIRMWARE_PREALLOC_BUFFER
) {
957 allow_write_access(file
);
960 EXPORT_SYMBOL_GPL(kernel_read_file
);
962 int kernel_read_file_from_path(char *path
, void **buf
, loff_t
*size
,
963 loff_t max_size
, enum kernel_read_file_id id
)
971 file
= filp_open(path
, O_RDONLY
, 0);
973 return PTR_ERR(file
);
975 ret
= kernel_read_file(file
, buf
, size
, max_size
, id
);
979 EXPORT_SYMBOL_GPL(kernel_read_file_from_path
);
981 int kernel_read_file_from_fd(int fd
, void **buf
, loff_t
*size
, loff_t max_size
,
982 enum kernel_read_file_id id
)
984 struct fd f
= fdget(fd
);
990 ret
= kernel_read_file(f
.file
, buf
, size
, max_size
, id
);
995 EXPORT_SYMBOL_GPL(kernel_read_file_from_fd
);
997 ssize_t
read_code(struct file
*file
, unsigned long addr
, loff_t pos
, size_t len
)
999 ssize_t res
= vfs_read(file
, (void __user
*)addr
, len
, &pos
);
1001 flush_icache_range(addr
, addr
+ len
);
1004 EXPORT_SYMBOL(read_code
);
1006 static int exec_mmap(struct mm_struct
*mm
)
1008 struct task_struct
*tsk
;
1009 struct mm_struct
*old_mm
, *active_mm
;
1011 /* Notify parent that we're no longer interested in the old VM */
1013 old_mm
= current
->mm
;
1014 mm_release(tsk
, old_mm
);
1017 sync_mm_rss(old_mm
);
1019 * Make sure that if there is a core dump in progress
1020 * for the old mm, we get out and die instead of going
1021 * through with the exec. We must hold mmap_sem around
1022 * checking core_state and changing tsk->mm.
1024 down_read(&old_mm
->mmap_sem
);
1025 if (unlikely(old_mm
->core_state
)) {
1026 up_read(&old_mm
->mmap_sem
);
1031 active_mm
= tsk
->active_mm
;
1033 tsk
->active_mm
= mm
;
1034 activate_mm(active_mm
, mm
);
1035 tsk
->mm
->vmacache_seqnum
= 0;
1036 vmacache_flush(tsk
);
1039 up_read(&old_mm
->mmap_sem
);
1040 BUG_ON(active_mm
!= old_mm
);
1041 setmax_mm_hiwater_rss(&tsk
->signal
->maxrss
, old_mm
);
1042 mm_update_next_owner(old_mm
);
1051 * This function makes sure the current process has its own signal table,
1052 * so that flush_signal_handlers can later reset the handlers without
1053 * disturbing other processes. (Other processes might share the signal
1054 * table via the CLONE_SIGHAND option to clone().)
1056 static int de_thread(struct task_struct
*tsk
)
1058 struct signal_struct
*sig
= tsk
->signal
;
1059 struct sighand_struct
*oldsighand
= tsk
->sighand
;
1060 spinlock_t
*lock
= &oldsighand
->siglock
;
1062 if (thread_group_empty(tsk
))
1063 goto no_thread_group
;
1066 * Kill all other threads in the thread group.
1068 spin_lock_irq(lock
);
1069 if (signal_group_exit(sig
)) {
1071 * Another group action in progress, just
1072 * return so that the signal is processed.
1074 spin_unlock_irq(lock
);
1078 sig
->group_exit_task
= tsk
;
1079 sig
->notify_count
= zap_other_threads(tsk
);
1080 if (!thread_group_leader(tsk
))
1081 sig
->notify_count
--;
1083 while (sig
->notify_count
) {
1084 __set_current_state(TASK_KILLABLE
);
1085 spin_unlock_irq(lock
);
1087 if (unlikely(__fatal_signal_pending(tsk
)))
1089 spin_lock_irq(lock
);
1091 spin_unlock_irq(lock
);
1094 * At this point all other threads have exited, all we have to
1095 * do is to wait for the thread group leader to become inactive,
1096 * and to assume its PID:
1098 if (!thread_group_leader(tsk
)) {
1099 struct task_struct
*leader
= tsk
->group_leader
;
1102 threadgroup_change_begin(tsk
);
1103 write_lock_irq(&tasklist_lock
);
1105 * Do this under tasklist_lock to ensure that
1106 * exit_notify() can't miss ->group_exit_task
1108 sig
->notify_count
= -1;
1109 if (likely(leader
->exit_state
))
1111 __set_current_state(TASK_KILLABLE
);
1112 write_unlock_irq(&tasklist_lock
);
1113 threadgroup_change_end(tsk
);
1115 if (unlikely(__fatal_signal_pending(tsk
)))
1120 * The only record we have of the real-time age of a
1121 * process, regardless of execs it's done, is start_time.
1122 * All the past CPU time is accumulated in signal_struct
1123 * from sister threads now dead. But in this non-leader
1124 * exec, nothing survives from the original leader thread,
1125 * whose birth marks the true age of this process now.
1126 * When we take on its identity by switching to its PID, we
1127 * also take its birthdate (always earlier than our own).
1129 tsk
->start_time
= leader
->start_time
;
1130 tsk
->real_start_time
= leader
->real_start_time
;
1132 BUG_ON(!same_thread_group(leader
, tsk
));
1133 BUG_ON(has_group_leader_pid(tsk
));
1135 * An exec() starts a new thread group with the
1136 * TGID of the previous thread group. Rehash the
1137 * two threads with a switched PID, and release
1138 * the former thread group leader:
1141 /* Become a process group leader with the old leader's pid.
1142 * The old leader becomes a thread of the this thread group.
1143 * Note: The old leader also uses this pid until release_task
1144 * is called. Odd but simple and correct.
1146 tsk
->pid
= leader
->pid
;
1147 change_pid(tsk
, PIDTYPE_PID
, task_pid(leader
));
1148 transfer_pid(leader
, tsk
, PIDTYPE_PGID
);
1149 transfer_pid(leader
, tsk
, PIDTYPE_SID
);
1151 list_replace_rcu(&leader
->tasks
, &tsk
->tasks
);
1152 list_replace_init(&leader
->sibling
, &tsk
->sibling
);
1154 tsk
->group_leader
= tsk
;
1155 leader
->group_leader
= tsk
;
1157 tsk
->exit_signal
= SIGCHLD
;
1158 leader
->exit_signal
= -1;
1160 BUG_ON(leader
->exit_state
!= EXIT_ZOMBIE
);
1161 leader
->exit_state
= EXIT_DEAD
;
1164 * We are going to release_task()->ptrace_unlink() silently,
1165 * the tracer can sleep in do_wait(). EXIT_DEAD guarantees
1166 * the tracer wont't block again waiting for this thread.
1168 if (unlikely(leader
->ptrace
))
1169 __wake_up_parent(leader
, leader
->parent
);
1170 write_unlock_irq(&tasklist_lock
);
1171 threadgroup_change_end(tsk
);
1173 release_task(leader
);
1176 sig
->group_exit_task
= NULL
;
1177 sig
->notify_count
= 0;
1180 /* we have changed execution domain */
1181 tsk
->exit_signal
= SIGCHLD
;
1183 #ifdef CONFIG_POSIX_TIMERS
1185 flush_itimer_signals();
1188 if (atomic_read(&oldsighand
->count
) != 1) {
1189 struct sighand_struct
*newsighand
;
1191 * This ->sighand is shared with the CLONE_SIGHAND
1192 * but not CLONE_THREAD task, switch to the new one.
1194 newsighand
= kmem_cache_alloc(sighand_cachep
, GFP_KERNEL
);
1198 atomic_set(&newsighand
->count
, 1);
1199 memcpy(newsighand
->action
, oldsighand
->action
,
1200 sizeof(newsighand
->action
));
1202 write_lock_irq(&tasklist_lock
);
1203 spin_lock(&oldsighand
->siglock
);
1204 rcu_assign_pointer(tsk
->sighand
, newsighand
);
1205 spin_unlock(&oldsighand
->siglock
);
1206 write_unlock_irq(&tasklist_lock
);
1208 __cleanup_sighand(oldsighand
);
1211 BUG_ON(!thread_group_leader(tsk
));
1215 /* protects against exit_notify() and __exit_signal() */
1216 read_lock(&tasklist_lock
);
1217 sig
->group_exit_task
= NULL
;
1218 sig
->notify_count
= 0;
1219 read_unlock(&tasklist_lock
);
1223 char *get_task_comm(char *buf
, struct task_struct
*tsk
)
1225 /* buf must be at least sizeof(tsk->comm) in size */
1227 strncpy(buf
, tsk
->comm
, sizeof(tsk
->comm
));
1231 EXPORT_SYMBOL_GPL(get_task_comm
);
1234 * These functions flushes out all traces of the currently running executable
1235 * so that a new one can be started
1238 void __set_task_comm(struct task_struct
*tsk
, const char *buf
, bool exec
)
1241 trace_task_rename(tsk
, buf
);
1242 strlcpy(tsk
->comm
, buf
, sizeof(tsk
->comm
));
1244 perf_event_comm(tsk
, exec
);
1247 int flush_old_exec(struct linux_binprm
* bprm
)
1252 * Make sure we have a private signal table and that
1253 * we are unassociated from the previous thread group.
1255 retval
= de_thread(current
);
1260 * Must be called _before_ exec_mmap() as bprm->mm is
1261 * not visibile until then. This also enables the update
1264 set_mm_exe_file(bprm
->mm
, bprm
->file
);
1267 * Release all of the old mmap stuff
1269 acct_arg_size(bprm
, 0);
1270 retval
= exec_mmap(bprm
->mm
);
1274 bprm
->mm
= NULL
; /* We're using it now */
1277 current
->flags
&= ~(PF_RANDOMIZE
| PF_FORKNOEXEC
| PF_KTHREAD
|
1278 PF_NOFREEZE
| PF_NO_SETAFFINITY
);
1280 current
->personality
&= ~bprm
->per_clear
;
1283 * We have to apply CLOEXEC before we change whether the process is
1284 * dumpable (in setup_new_exec) to avoid a race with a process in userspace
1285 * trying to access the should-be-closed file descriptors of a process
1286 * undergoing exec(2).
1288 do_close_on_exec(current
->files
);
1294 EXPORT_SYMBOL(flush_old_exec
);
1296 void would_dump(struct linux_binprm
*bprm
, struct file
*file
)
1298 struct inode
*inode
= file_inode(file
);
1299 if (inode_permission(inode
, MAY_READ
) < 0) {
1300 struct user_namespace
*old
, *user_ns
;
1301 bprm
->interp_flags
|= BINPRM_FLAGS_ENFORCE_NONDUMP
;
1303 /* Ensure mm->user_ns contains the executable */
1304 user_ns
= old
= bprm
->mm
->user_ns
;
1305 while ((user_ns
!= &init_user_ns
) &&
1306 !privileged_wrt_inode_uidgid(user_ns
, inode
))
1307 user_ns
= user_ns
->parent
;
1309 if (old
!= user_ns
) {
1310 bprm
->mm
->user_ns
= get_user_ns(user_ns
);
1315 EXPORT_SYMBOL(would_dump
);
1317 void setup_new_exec(struct linux_binprm
* bprm
)
1319 arch_pick_mmap_layout(current
->mm
);
1321 /* This is the point of no return */
1322 current
->sas_ss_sp
= current
->sas_ss_size
= 0;
1324 if (uid_eq(current_euid(), current_uid()) && gid_eq(current_egid(), current_gid()))
1325 set_dumpable(current
->mm
, SUID_DUMP_USER
);
1327 set_dumpable(current
->mm
, suid_dumpable
);
1330 __set_task_comm(current
, kbasename(bprm
->filename
), true);
1332 /* Set the new mm task size. We have to do that late because it may
1333 * depend on TIF_32BIT which is only updated in flush_thread() on
1334 * some architectures like powerpc
1336 current
->mm
->task_size
= TASK_SIZE
;
1338 /* install the new credentials */
1339 if (!uid_eq(bprm
->cred
->uid
, current_euid()) ||
1340 !gid_eq(bprm
->cred
->gid
, current_egid())) {
1341 current
->pdeath_signal
= 0;
1343 if (bprm
->interp_flags
& BINPRM_FLAGS_ENFORCE_NONDUMP
)
1344 set_dumpable(current
->mm
, suid_dumpable
);
1347 /* An exec changes our domain. We are no longer part of the thread
1349 current
->self_exec_id
++;
1350 flush_signal_handlers(current
, 0);
1352 EXPORT_SYMBOL(setup_new_exec
);
1355 * Prepare credentials and lock ->cred_guard_mutex.
1356 * install_exec_creds() commits the new creds and drops the lock.
1357 * Or, if exec fails before, free_bprm() should release ->cred and
1360 int prepare_bprm_creds(struct linux_binprm
*bprm
)
1362 if (mutex_lock_interruptible(¤t
->signal
->cred_guard_mutex
))
1363 return -ERESTARTNOINTR
;
1365 bprm
->cred
= prepare_exec_creds();
1366 if (likely(bprm
->cred
))
1369 mutex_unlock(¤t
->signal
->cred_guard_mutex
);
1373 static void free_bprm(struct linux_binprm
*bprm
)
1375 free_arg_pages(bprm
);
1377 mutex_unlock(¤t
->signal
->cred_guard_mutex
);
1378 abort_creds(bprm
->cred
);
1381 allow_write_access(bprm
->file
);
1384 /* If a binfmt changed the interp, free it. */
1385 if (bprm
->interp
!= bprm
->filename
)
1386 kfree(bprm
->interp
);
1390 int bprm_change_interp(char *interp
, struct linux_binprm
*bprm
)
1392 /* If a binfmt changed the interp, free it first. */
1393 if (bprm
->interp
!= bprm
->filename
)
1394 kfree(bprm
->interp
);
1395 bprm
->interp
= kstrdup(interp
, GFP_KERNEL
);
1400 EXPORT_SYMBOL(bprm_change_interp
);
1403 * install the new credentials for this executable
1405 void install_exec_creds(struct linux_binprm
*bprm
)
1407 security_bprm_committing_creds(bprm
);
1409 commit_creds(bprm
->cred
);
1413 * Disable monitoring for regular users
1414 * when executing setuid binaries. Must
1415 * wait until new credentials are committed
1416 * by commit_creds() above
1418 if (get_dumpable(current
->mm
) != SUID_DUMP_USER
)
1419 perf_event_exit_task(current
);
1421 * cred_guard_mutex must be held at least to this point to prevent
1422 * ptrace_attach() from altering our determination of the task's
1423 * credentials; any time after this it may be unlocked.
1425 security_bprm_committed_creds(bprm
);
1426 mutex_unlock(¤t
->signal
->cred_guard_mutex
);
1428 EXPORT_SYMBOL(install_exec_creds
);
1431 * determine how safe it is to execute the proposed program
1432 * - the caller must hold ->cred_guard_mutex to protect against
1433 * PTRACE_ATTACH or seccomp thread-sync
1435 static void check_unsafe_exec(struct linux_binprm
*bprm
)
1437 struct task_struct
*p
= current
, *t
;
1441 if (ptracer_capable(p
, current_user_ns()))
1442 bprm
->unsafe
|= LSM_UNSAFE_PTRACE_CAP
;
1444 bprm
->unsafe
|= LSM_UNSAFE_PTRACE
;
1448 * This isn't strictly necessary, but it makes it harder for LSMs to
1451 if (task_no_new_privs(current
))
1452 bprm
->unsafe
|= LSM_UNSAFE_NO_NEW_PRIVS
;
1456 spin_lock(&p
->fs
->lock
);
1458 while_each_thread(p
, t
) {
1464 if (p
->fs
->users
> n_fs
)
1465 bprm
->unsafe
|= LSM_UNSAFE_SHARE
;
1468 spin_unlock(&p
->fs
->lock
);
1471 static void bprm_fill_uid(struct linux_binprm
*bprm
)
1473 struct inode
*inode
;
1479 * Since this can be called multiple times (via prepare_binprm),
1480 * we must clear any previous work done when setting set[ug]id
1481 * bits from any earlier bprm->file uses (for example when run
1482 * first for a setuid script then again for its interpreter).
1484 bprm
->cred
->euid
= current_euid();
1485 bprm
->cred
->egid
= current_egid();
1487 if (path_nosuid(&bprm
->file
->f_path
))
1490 if (task_no_new_privs(current
))
1493 inode
= file_inode(bprm
->file
);
1494 mode
= READ_ONCE(inode
->i_mode
);
1495 if (!(mode
& (S_ISUID
|S_ISGID
)))
1498 /* Be careful if suid/sgid is set */
1501 /* reload atomically mode/uid/gid now that lock held */
1502 mode
= inode
->i_mode
;
1505 inode_unlock(inode
);
1507 /* We ignore suid/sgid if there are no mappings for them in the ns */
1508 if (!kuid_has_mapping(bprm
->cred
->user_ns
, uid
) ||
1509 !kgid_has_mapping(bprm
->cred
->user_ns
, gid
))
1512 if (mode
& S_ISUID
) {
1513 bprm
->per_clear
|= PER_CLEAR_ON_SETID
;
1514 bprm
->cred
->euid
= uid
;
1517 if ((mode
& (S_ISGID
| S_IXGRP
)) == (S_ISGID
| S_IXGRP
)) {
1518 bprm
->per_clear
|= PER_CLEAR_ON_SETID
;
1519 bprm
->cred
->egid
= gid
;
1524 * Fill the binprm structure from the inode.
1525 * Check permissions, then read the first 128 (BINPRM_BUF_SIZE) bytes
1527 * This may be called multiple times for binary chains (scripts for example).
1529 int prepare_binprm(struct linux_binprm
*bprm
)
1533 bprm_fill_uid(bprm
);
1535 /* fill in binprm security blob */
1536 retval
= security_bprm_set_creds(bprm
);
1539 bprm
->cred_prepared
= 1;
1541 memset(bprm
->buf
, 0, BINPRM_BUF_SIZE
);
1542 return kernel_read(bprm
->file
, 0, bprm
->buf
, BINPRM_BUF_SIZE
);
1545 EXPORT_SYMBOL(prepare_binprm
);
1548 * Arguments are '\0' separated strings found at the location bprm->p
1549 * points to; chop off the first by relocating brpm->p to right after
1550 * the first '\0' encountered.
1552 int remove_arg_zero(struct linux_binprm
*bprm
)
1555 unsigned long offset
;
1563 offset
= bprm
->p
& ~PAGE_MASK
;
1564 page
= get_arg_page(bprm
, bprm
->p
, 0);
1569 kaddr
= kmap_atomic(page
);
1571 for (; offset
< PAGE_SIZE
&& kaddr
[offset
];
1572 offset
++, bprm
->p
++)
1575 kunmap_atomic(kaddr
);
1577 } while (offset
== PAGE_SIZE
);
1586 EXPORT_SYMBOL(remove_arg_zero
);
1588 #define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e))
1590 * cycle the list of binary formats handler, until one recognizes the image
1592 int search_binary_handler(struct linux_binprm
*bprm
)
1594 bool need_retry
= IS_ENABLED(CONFIG_MODULES
);
1595 struct linux_binfmt
*fmt
;
1598 /* This allows 4 levels of binfmt rewrites before failing hard. */
1599 if (bprm
->recursion_depth
> 5)
1602 retval
= security_bprm_check(bprm
);
1608 read_lock(&binfmt_lock
);
1609 list_for_each_entry(fmt
, &formats
, lh
) {
1610 if (!try_module_get(fmt
->module
))
1612 read_unlock(&binfmt_lock
);
1613 bprm
->recursion_depth
++;
1614 retval
= fmt
->load_binary(bprm
);
1615 read_lock(&binfmt_lock
);
1617 bprm
->recursion_depth
--;
1618 if (retval
< 0 && !bprm
->mm
) {
1619 /* we got to flush_old_exec() and failed after it */
1620 read_unlock(&binfmt_lock
);
1621 force_sigsegv(SIGSEGV
, current
);
1624 if (retval
!= -ENOEXEC
|| !bprm
->file
) {
1625 read_unlock(&binfmt_lock
);
1629 read_unlock(&binfmt_lock
);
1632 if (printable(bprm
->buf
[0]) && printable(bprm
->buf
[1]) &&
1633 printable(bprm
->buf
[2]) && printable(bprm
->buf
[3]))
1635 if (request_module("binfmt-%04x", *(ushort
*)(bprm
->buf
+ 2)) < 0)
1643 EXPORT_SYMBOL(search_binary_handler
);
1645 static int exec_binprm(struct linux_binprm
*bprm
)
1647 pid_t old_pid
, old_vpid
;
1650 /* Need to fetch pid before load_binary changes it */
1651 old_pid
= current
->pid
;
1653 old_vpid
= task_pid_nr_ns(current
, task_active_pid_ns(current
->parent
));
1656 ret
= search_binary_handler(bprm
);
1659 trace_sched_process_exec(current
, old_pid
, bprm
);
1660 ptrace_event(PTRACE_EVENT_EXEC
, old_vpid
);
1661 proc_exec_connector(current
);
1668 * sys_execve() executes a new program.
1670 static int do_execveat_common(int fd
, struct filename
*filename
,
1671 struct user_arg_ptr argv
,
1672 struct user_arg_ptr envp
,
1675 char *pathbuf
= NULL
;
1676 struct linux_binprm
*bprm
;
1678 struct files_struct
*displaced
;
1681 if (IS_ERR(filename
))
1682 return PTR_ERR(filename
);
1685 * We move the actual failure in case of RLIMIT_NPROC excess from
1686 * set*uid() to execve() because too many poorly written programs
1687 * don't check setuid() return code. Here we additionally recheck
1688 * whether NPROC limit is still exceeded.
1690 if ((current
->flags
& PF_NPROC_EXCEEDED
) &&
1691 atomic_read(¤t_user()->processes
) > rlimit(RLIMIT_NPROC
)) {
1696 /* We're below the limit (still or again), so we don't want to make
1697 * further execve() calls fail. */
1698 current
->flags
&= ~PF_NPROC_EXCEEDED
;
1700 retval
= unshare_files(&displaced
);
1705 bprm
= kzalloc(sizeof(*bprm
), GFP_KERNEL
);
1709 retval
= prepare_bprm_creds(bprm
);
1713 check_unsafe_exec(bprm
);
1714 current
->in_execve
= 1;
1716 file
= do_open_execat(fd
, filename
, flags
);
1717 retval
= PTR_ERR(file
);
1724 if (fd
== AT_FDCWD
|| filename
->name
[0] == '/') {
1725 bprm
->filename
= filename
->name
;
1727 if (filename
->name
[0] == '\0')
1728 pathbuf
= kasprintf(GFP_TEMPORARY
, "/dev/fd/%d", fd
);
1730 pathbuf
= kasprintf(GFP_TEMPORARY
, "/dev/fd/%d/%s",
1731 fd
, filename
->name
);
1737 * Record that a name derived from an O_CLOEXEC fd will be
1738 * inaccessible after exec. Relies on having exclusive access to
1739 * current->files (due to unshare_files above).
1741 if (close_on_exec(fd
, rcu_dereference_raw(current
->files
->fdt
)))
1742 bprm
->interp_flags
|= BINPRM_FLAGS_PATH_INACCESSIBLE
;
1743 bprm
->filename
= pathbuf
;
1745 bprm
->interp
= bprm
->filename
;
1747 retval
= bprm_mm_init(bprm
);
1751 bprm
->argc
= count(argv
, MAX_ARG_STRINGS
);
1752 if ((retval
= bprm
->argc
) < 0)
1755 bprm
->envc
= count(envp
, MAX_ARG_STRINGS
);
1756 if ((retval
= bprm
->envc
) < 0)
1759 retval
= prepare_binprm(bprm
);
1763 retval
= copy_strings_kernel(1, &bprm
->filename
, bprm
);
1767 bprm
->exec
= bprm
->p
;
1768 retval
= copy_strings(bprm
->envc
, envp
, bprm
);
1772 retval
= copy_strings(bprm
->argc
, argv
, bprm
);
1776 would_dump(bprm
, bprm
->file
);
1778 retval
= exec_binprm(bprm
);
1782 /* execve succeeded */
1783 current
->fs
->in_exec
= 0;
1784 current
->in_execve
= 0;
1785 acct_update_integrals(current
);
1786 task_numa_free(current
);
1791 put_files_struct(displaced
);
1796 acct_arg_size(bprm
, 0);
1801 current
->fs
->in_exec
= 0;
1802 current
->in_execve
= 0;
1810 reset_files_struct(displaced
);
1816 int do_execve(struct filename
*filename
,
1817 const char __user
*const __user
*__argv
,
1818 const char __user
*const __user
*__envp
)
1820 struct user_arg_ptr argv
= { .ptr
.native
= __argv
};
1821 struct user_arg_ptr envp
= { .ptr
.native
= __envp
};
1822 return do_execveat_common(AT_FDCWD
, filename
, argv
, envp
, 0);
1825 int do_execveat(int fd
, struct filename
*filename
,
1826 const char __user
*const __user
*__argv
,
1827 const char __user
*const __user
*__envp
,
1830 struct user_arg_ptr argv
= { .ptr
.native
= __argv
};
1831 struct user_arg_ptr envp
= { .ptr
.native
= __envp
};
1833 return do_execveat_common(fd
, filename
, argv
, envp
, flags
);
1836 #ifdef CONFIG_COMPAT
1837 static int compat_do_execve(struct filename
*filename
,
1838 const compat_uptr_t __user
*__argv
,
1839 const compat_uptr_t __user
*__envp
)
1841 struct user_arg_ptr argv
= {
1843 .ptr
.compat
= __argv
,
1845 struct user_arg_ptr envp
= {
1847 .ptr
.compat
= __envp
,
1849 return do_execveat_common(AT_FDCWD
, filename
, argv
, envp
, 0);
1852 static int compat_do_execveat(int fd
, struct filename
*filename
,
1853 const compat_uptr_t __user
*__argv
,
1854 const compat_uptr_t __user
*__envp
,
1857 struct user_arg_ptr argv
= {
1859 .ptr
.compat
= __argv
,
1861 struct user_arg_ptr envp
= {
1863 .ptr
.compat
= __envp
,
1865 return do_execveat_common(fd
, filename
, argv
, envp
, flags
);
1869 void set_binfmt(struct linux_binfmt
*new)
1871 struct mm_struct
*mm
= current
->mm
;
1874 module_put(mm
->binfmt
->module
);
1878 __module_get(new->module
);
1880 EXPORT_SYMBOL(set_binfmt
);
1883 * set_dumpable stores three-value SUID_DUMP_* into mm->flags.
1885 void set_dumpable(struct mm_struct
*mm
, int value
)
1887 unsigned long old
, new;
1889 if (WARN_ON((unsigned)value
> SUID_DUMP_ROOT
))
1893 old
= ACCESS_ONCE(mm
->flags
);
1894 new = (old
& ~MMF_DUMPABLE_MASK
) | value
;
1895 } while (cmpxchg(&mm
->flags
, old
, new) != old
);
1898 SYSCALL_DEFINE3(execve
,
1899 const char __user
*, filename
,
1900 const char __user
*const __user
*, argv
,
1901 const char __user
*const __user
*, envp
)
1903 return do_execve(getname(filename
), argv
, envp
);
1906 SYSCALL_DEFINE5(execveat
,
1907 int, fd
, const char __user
*, filename
,
1908 const char __user
*const __user
*, argv
,
1909 const char __user
*const __user
*, envp
,
1912 int lookup_flags
= (flags
& AT_EMPTY_PATH
) ? LOOKUP_EMPTY
: 0;
1914 return do_execveat(fd
,
1915 getname_flags(filename
, lookup_flags
, NULL
),
1919 #ifdef CONFIG_COMPAT
1920 COMPAT_SYSCALL_DEFINE3(execve
, const char __user
*, filename
,
1921 const compat_uptr_t __user
*, argv
,
1922 const compat_uptr_t __user
*, envp
)
1924 return compat_do_execve(getname(filename
), argv
, envp
);
1927 COMPAT_SYSCALL_DEFINE5(execveat
, int, fd
,
1928 const char __user
*, filename
,
1929 const compat_uptr_t __user
*, argv
,
1930 const compat_uptr_t __user
*, envp
,
1933 int lookup_flags
= (flags
& AT_EMPTY_PATH
) ? LOOKUP_EMPTY
: 0;
1935 return compat_do_execveat(fd
,
1936 getname_flags(filename
, lookup_flags
, NULL
),