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
);
109 EXPORT_SYMBOL_GPL(path_noexec
);
111 bool path_nosuid(const struct path
*path
)
113 return !mnt_may_suid(path
->mnt
) ||
114 (path
->mnt
->mnt_sb
->s_iflags
& SB_I_NOSUID
);
116 EXPORT_SYMBOL(path_nosuid
);
120 * Note that a shared library must be both readable and executable due to
123 * Also note that we take the address to load from from the file itself.
125 SYSCALL_DEFINE1(uselib
, const char __user
*, library
)
127 struct linux_binfmt
*fmt
;
129 struct filename
*tmp
= getname(library
);
130 int error
= PTR_ERR(tmp
);
131 static const struct open_flags uselib_flags
= {
132 .open_flag
= O_LARGEFILE
| O_RDONLY
| __FMODE_EXEC
,
133 .acc_mode
= MAY_READ
| MAY_EXEC
,
134 .intent
= LOOKUP_OPEN
,
135 .lookup_flags
= LOOKUP_FOLLOW
,
141 file
= do_filp_open(AT_FDCWD
, tmp
, &uselib_flags
);
143 error
= PTR_ERR(file
);
148 if (!S_ISREG(file_inode(file
)->i_mode
))
152 if (path_noexec(&file
->f_path
))
159 read_lock(&binfmt_lock
);
160 list_for_each_entry(fmt
, &formats
, lh
) {
161 if (!fmt
->load_shlib
)
163 if (!try_module_get(fmt
->module
))
165 read_unlock(&binfmt_lock
);
166 error
= fmt
->load_shlib(file
);
167 read_lock(&binfmt_lock
);
169 if (error
!= -ENOEXEC
)
172 read_unlock(&binfmt_lock
);
178 #endif /* #ifdef CONFIG_USELIB */
182 * The nascent bprm->mm is not visible until exec_mmap() but it can
183 * use a lot of memory, account these pages in current->mm temporary
184 * for oom_badness()->get_mm_rss(). Once exec succeeds or fails, we
185 * change the counter back via acct_arg_size(0).
187 static void acct_arg_size(struct linux_binprm
*bprm
, unsigned long pages
)
189 struct mm_struct
*mm
= current
->mm
;
190 long diff
= (long)(pages
- bprm
->vma_pages
);
195 bprm
->vma_pages
= pages
;
196 add_mm_counter(mm
, MM_ANONPAGES
, diff
);
199 static struct page
*get_arg_page(struct linux_binprm
*bprm
, unsigned long pos
,
204 unsigned int gup_flags
= FOLL_FORCE
;
206 #ifdef CONFIG_STACK_GROWSUP
208 ret
= expand_downwards(bprm
->vma
, pos
);
215 gup_flags
|= FOLL_WRITE
;
218 * We are doing an exec(). 'current' is the process
219 * doing the exec and bprm->mm is the new process's mm.
221 ret
= get_user_pages_remote(current
, bprm
->mm
, pos
, 1, gup_flags
,
227 unsigned long size
= bprm
->vma
->vm_end
- bprm
->vma
->vm_start
;
230 acct_arg_size(bprm
, size
/ PAGE_SIZE
);
233 * We've historically supported up to 32 pages (ARG_MAX)
234 * of argument strings even with small stacks
240 * Limit to 1/4-th the stack size for the argv+env strings.
242 * - the remaining binfmt code will not run out of stack space,
243 * - the program will have a reasonable amount of stack left
246 rlim
= current
->signal
->rlim
;
247 if (size
> ACCESS_ONCE(rlim
[RLIMIT_STACK
].rlim_cur
) / 4) {
256 static void put_arg_page(struct page
*page
)
261 static void free_arg_pages(struct linux_binprm
*bprm
)
265 static void flush_arg_page(struct linux_binprm
*bprm
, unsigned long pos
,
268 flush_cache_page(bprm
->vma
, pos
, page_to_pfn(page
));
271 static int __bprm_mm_init(struct linux_binprm
*bprm
)
274 struct vm_area_struct
*vma
= NULL
;
275 struct mm_struct
*mm
= bprm
->mm
;
277 bprm
->vma
= vma
= kmem_cache_zalloc(vm_area_cachep
, GFP_KERNEL
);
281 if (down_write_killable(&mm
->mmap_sem
)) {
288 * Place the stack at the largest stack address the architecture
289 * supports. Later, we'll move this to an appropriate place. We don't
290 * use STACK_TOP because that can depend on attributes which aren't
293 BUILD_BUG_ON(VM_STACK_FLAGS
& VM_STACK_INCOMPLETE_SETUP
);
294 vma
->vm_end
= STACK_TOP_MAX
;
295 vma
->vm_start
= vma
->vm_end
- PAGE_SIZE
;
296 vma
->vm_flags
= VM_SOFTDIRTY
| VM_STACK_FLAGS
| VM_STACK_INCOMPLETE_SETUP
;
297 vma
->vm_page_prot
= vm_get_page_prot(vma
->vm_flags
);
298 INIT_LIST_HEAD(&vma
->anon_vma_chain
);
300 err
= insert_vm_struct(mm
, vma
);
304 mm
->stack_vm
= mm
->total_vm
= 1;
305 arch_bprm_mm_init(mm
, vma
);
306 up_write(&mm
->mmap_sem
);
307 bprm
->p
= vma
->vm_end
- sizeof(void *);
310 up_write(&mm
->mmap_sem
);
313 kmem_cache_free(vm_area_cachep
, vma
);
317 static bool valid_arg_len(struct linux_binprm
*bprm
, long len
)
319 return len
<= MAX_ARG_STRLEN
;
324 static inline void acct_arg_size(struct linux_binprm
*bprm
, unsigned long pages
)
328 static struct page
*get_arg_page(struct linux_binprm
*bprm
, unsigned long pos
,
333 page
= bprm
->page
[pos
/ PAGE_SIZE
];
334 if (!page
&& write
) {
335 page
= alloc_page(GFP_HIGHUSER
|__GFP_ZERO
);
338 bprm
->page
[pos
/ PAGE_SIZE
] = page
;
344 static void put_arg_page(struct page
*page
)
348 static void free_arg_page(struct linux_binprm
*bprm
, int i
)
351 __free_page(bprm
->page
[i
]);
352 bprm
->page
[i
] = NULL
;
356 static void free_arg_pages(struct linux_binprm
*bprm
)
360 for (i
= 0; i
< MAX_ARG_PAGES
; i
++)
361 free_arg_page(bprm
, i
);
364 static void flush_arg_page(struct linux_binprm
*bprm
, unsigned long pos
,
369 static int __bprm_mm_init(struct linux_binprm
*bprm
)
371 bprm
->p
= PAGE_SIZE
* MAX_ARG_PAGES
- sizeof(void *);
375 static bool valid_arg_len(struct linux_binprm
*bprm
, long len
)
377 return len
<= bprm
->p
;
380 #endif /* CONFIG_MMU */
383 * Create a new mm_struct and populate it with a temporary stack
384 * vm_area_struct. We don't have enough context at this point to set the stack
385 * flags, permissions, and offset, so we use temporary values. We'll update
386 * them later in setup_arg_pages().
388 static int bprm_mm_init(struct linux_binprm
*bprm
)
391 struct mm_struct
*mm
= NULL
;
393 bprm
->mm
= mm
= mm_alloc();
398 err
= __bprm_mm_init(bprm
);
413 struct user_arg_ptr
{
418 const char __user
*const __user
*native
;
420 const compat_uptr_t __user
*compat
;
425 static const char __user
*get_user_arg_ptr(struct user_arg_ptr argv
, int nr
)
427 const char __user
*native
;
430 if (unlikely(argv
.is_compat
)) {
431 compat_uptr_t compat
;
433 if (get_user(compat
, argv
.ptr
.compat
+ nr
))
434 return ERR_PTR(-EFAULT
);
436 return compat_ptr(compat
);
440 if (get_user(native
, argv
.ptr
.native
+ nr
))
441 return ERR_PTR(-EFAULT
);
447 * count() counts the number of strings in array ARGV.
449 static int count(struct user_arg_ptr argv
, int max
)
453 if (argv
.ptr
.native
!= NULL
) {
455 const char __user
*p
= get_user_arg_ptr(argv
, i
);
467 if (fatal_signal_pending(current
))
468 return -ERESTARTNOHAND
;
476 * 'copy_strings()' copies argument/environment strings from the old
477 * processes's memory to the new process's stack. The call to get_user_pages()
478 * ensures the destination page is created and not swapped out.
480 static int copy_strings(int argc
, struct user_arg_ptr argv
,
481 struct linux_binprm
*bprm
)
483 struct page
*kmapped_page
= NULL
;
485 unsigned long kpos
= 0;
489 const char __user
*str
;
494 str
= get_user_arg_ptr(argv
, argc
);
498 len
= strnlen_user(str
, MAX_ARG_STRLEN
);
503 if (!valid_arg_len(bprm
, len
))
506 /* We're going to work our way backwords. */
512 int offset
, bytes_to_copy
;
514 if (fatal_signal_pending(current
)) {
515 ret
= -ERESTARTNOHAND
;
520 offset
= pos
% PAGE_SIZE
;
524 bytes_to_copy
= offset
;
525 if (bytes_to_copy
> len
)
528 offset
-= bytes_to_copy
;
529 pos
-= bytes_to_copy
;
530 str
-= bytes_to_copy
;
531 len
-= bytes_to_copy
;
533 if (!kmapped_page
|| kpos
!= (pos
& PAGE_MASK
)) {
536 page
= get_arg_page(bprm
, pos
, 1);
543 flush_kernel_dcache_page(kmapped_page
);
544 kunmap(kmapped_page
);
545 put_arg_page(kmapped_page
);
548 kaddr
= kmap(kmapped_page
);
549 kpos
= pos
& PAGE_MASK
;
550 flush_arg_page(bprm
, kpos
, kmapped_page
);
552 if (copy_from_user(kaddr
+offset
, str
, bytes_to_copy
)) {
561 flush_kernel_dcache_page(kmapped_page
);
562 kunmap(kmapped_page
);
563 put_arg_page(kmapped_page
);
569 * Like copy_strings, but get argv and its values from kernel memory.
571 int copy_strings_kernel(int argc
, const char *const *__argv
,
572 struct linux_binprm
*bprm
)
575 mm_segment_t oldfs
= get_fs();
576 struct user_arg_ptr argv
= {
577 .ptr
.native
= (const char __user
*const __user
*)__argv
,
581 r
= copy_strings(argc
, argv
, bprm
);
586 EXPORT_SYMBOL(copy_strings_kernel
);
591 * During bprm_mm_init(), we create a temporary stack at STACK_TOP_MAX. Once
592 * the binfmt code determines where the new stack should reside, we shift it to
593 * its final location. The process proceeds as follows:
595 * 1) Use shift to calculate the new vma endpoints.
596 * 2) Extend vma to cover both the old and new ranges. This ensures the
597 * arguments passed to subsequent functions are consistent.
598 * 3) Move vma's page tables to the new range.
599 * 4) Free up any cleared pgd range.
600 * 5) Shrink the vma to cover only the new range.
602 static int shift_arg_pages(struct vm_area_struct
*vma
, unsigned long shift
)
604 struct mm_struct
*mm
= vma
->vm_mm
;
605 unsigned long old_start
= vma
->vm_start
;
606 unsigned long old_end
= vma
->vm_end
;
607 unsigned long length
= old_end
- old_start
;
608 unsigned long new_start
= old_start
- shift
;
609 unsigned long new_end
= old_end
- shift
;
610 struct mmu_gather tlb
;
612 BUG_ON(new_start
> new_end
);
615 * ensure there are no vmas between where we want to go
618 if (vma
!= find_vma(mm
, new_start
))
622 * cover the whole range: [new_start, old_end)
624 if (vma_adjust(vma
, new_start
, old_end
, vma
->vm_pgoff
, NULL
))
628 * move the page tables downwards, on failure we rely on
629 * process cleanup to remove whatever mess we made.
631 if (length
!= move_page_tables(vma
, old_start
,
632 vma
, new_start
, length
, false))
636 tlb_gather_mmu(&tlb
, mm
, old_start
, old_end
);
637 if (new_end
> old_start
) {
639 * when the old and new regions overlap clear from new_end.
641 free_pgd_range(&tlb
, new_end
, old_end
, new_end
,
642 vma
->vm_next
? vma
->vm_next
->vm_start
: USER_PGTABLES_CEILING
);
645 * otherwise, clean from old_start; this is done to not touch
646 * the address space in [new_end, old_start) some architectures
647 * have constraints on va-space that make this illegal (IA64) -
648 * for the others its just a little faster.
650 free_pgd_range(&tlb
, old_start
, old_end
, new_end
,
651 vma
->vm_next
? vma
->vm_next
->vm_start
: USER_PGTABLES_CEILING
);
653 tlb_finish_mmu(&tlb
, old_start
, old_end
);
656 * Shrink the vma to just the new range. Always succeeds.
658 vma_adjust(vma
, new_start
, new_end
, vma
->vm_pgoff
, NULL
);
664 * Finalizes the stack vm_area_struct. The flags and permissions are updated,
665 * the stack is optionally relocated, and some extra space is added.
667 int setup_arg_pages(struct linux_binprm
*bprm
,
668 unsigned long stack_top
,
669 int executable_stack
)
672 unsigned long stack_shift
;
673 struct mm_struct
*mm
= current
->mm
;
674 struct vm_area_struct
*vma
= bprm
->vma
;
675 struct vm_area_struct
*prev
= NULL
;
676 unsigned long vm_flags
;
677 unsigned long stack_base
;
678 unsigned long stack_size
;
679 unsigned long stack_expand
;
680 unsigned long rlim_stack
;
682 #ifdef CONFIG_STACK_GROWSUP
683 /* Limit stack size */
684 stack_base
= rlimit_max(RLIMIT_STACK
);
685 if (stack_base
> STACK_SIZE_MAX
)
686 stack_base
= STACK_SIZE_MAX
;
688 /* Add space for stack randomization. */
689 stack_base
+= (STACK_RND_MASK
<< PAGE_SHIFT
);
691 /* Make sure we didn't let the argument array grow too large. */
692 if (vma
->vm_end
- vma
->vm_start
> stack_base
)
695 stack_base
= PAGE_ALIGN(stack_top
- stack_base
);
697 stack_shift
= vma
->vm_start
- stack_base
;
698 mm
->arg_start
= bprm
->p
- stack_shift
;
699 bprm
->p
= vma
->vm_end
- stack_shift
;
701 stack_top
= arch_align_stack(stack_top
);
702 stack_top
= PAGE_ALIGN(stack_top
);
704 if (unlikely(stack_top
< mmap_min_addr
) ||
705 unlikely(vma
->vm_end
- vma
->vm_start
>= stack_top
- mmap_min_addr
))
708 stack_shift
= vma
->vm_end
- stack_top
;
710 bprm
->p
-= stack_shift
;
711 mm
->arg_start
= bprm
->p
;
715 bprm
->loader
-= stack_shift
;
716 bprm
->exec
-= stack_shift
;
718 if (down_write_killable(&mm
->mmap_sem
))
721 vm_flags
= VM_STACK_FLAGS
;
724 * Adjust stack execute permissions; explicitly enable for
725 * EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X and leave alone
726 * (arch default) otherwise.
728 if (unlikely(executable_stack
== EXSTACK_ENABLE_X
))
730 else if (executable_stack
== EXSTACK_DISABLE_X
)
731 vm_flags
&= ~VM_EXEC
;
732 vm_flags
|= mm
->def_flags
;
733 vm_flags
|= VM_STACK_INCOMPLETE_SETUP
;
735 ret
= mprotect_fixup(vma
, &prev
, vma
->vm_start
, vma
->vm_end
,
741 /* Move stack pages down in memory. */
743 ret
= shift_arg_pages(vma
, stack_shift
);
748 /* mprotect_fixup is overkill to remove the temporary stack flags */
749 vma
->vm_flags
&= ~VM_STACK_INCOMPLETE_SETUP
;
751 stack_expand
= 131072UL; /* randomly 32*4k (or 2*64k) pages */
752 stack_size
= vma
->vm_end
- vma
->vm_start
;
754 * Align this down to a page boundary as expand_stack
757 rlim_stack
= rlimit(RLIMIT_STACK
) & PAGE_MASK
;
758 #ifdef CONFIG_STACK_GROWSUP
759 if (stack_size
+ stack_expand
> rlim_stack
)
760 stack_base
= vma
->vm_start
+ rlim_stack
;
762 stack_base
= vma
->vm_end
+ stack_expand
;
764 if (stack_size
+ stack_expand
> rlim_stack
)
765 stack_base
= vma
->vm_end
- rlim_stack
;
767 stack_base
= vma
->vm_start
- stack_expand
;
769 current
->mm
->start_stack
= bprm
->p
;
770 ret
= expand_stack(vma
, stack_base
);
775 up_write(&mm
->mmap_sem
);
778 EXPORT_SYMBOL(setup_arg_pages
);
783 * Transfer the program arguments and environment from the holding pages
784 * onto the stack. The provided stack pointer is adjusted accordingly.
786 int transfer_args_to_stack(struct linux_binprm
*bprm
,
787 unsigned long *sp_location
)
789 unsigned long index
, stop
, sp
;
792 stop
= bprm
->p
>> PAGE_SHIFT
;
795 for (index
= MAX_ARG_PAGES
- 1; index
>= stop
; index
--) {
796 unsigned int offset
= index
== stop
? bprm
->p
& ~PAGE_MASK
: 0;
797 char *src
= kmap(bprm
->page
[index
]) + offset
;
798 sp
-= PAGE_SIZE
- offset
;
799 if (copy_to_user((void *) sp
, src
, PAGE_SIZE
- offset
) != 0)
801 kunmap(bprm
->page
[index
]);
811 EXPORT_SYMBOL(transfer_args_to_stack
);
813 #endif /* CONFIG_MMU */
815 static struct file
*do_open_execat(int fd
, struct filename
*name
, int flags
)
819 struct open_flags open_exec_flags
= {
820 .open_flag
= O_LARGEFILE
| O_RDONLY
| __FMODE_EXEC
,
821 .acc_mode
= MAY_EXEC
,
822 .intent
= LOOKUP_OPEN
,
823 .lookup_flags
= LOOKUP_FOLLOW
,
826 if ((flags
& ~(AT_SYMLINK_NOFOLLOW
| AT_EMPTY_PATH
)) != 0)
827 return ERR_PTR(-EINVAL
);
828 if (flags
& AT_SYMLINK_NOFOLLOW
)
829 open_exec_flags
.lookup_flags
&= ~LOOKUP_FOLLOW
;
830 if (flags
& AT_EMPTY_PATH
)
831 open_exec_flags
.lookup_flags
|= LOOKUP_EMPTY
;
833 file
= do_filp_open(fd
, name
, &open_exec_flags
);
838 if (!S_ISREG(file_inode(file
)->i_mode
))
841 if (path_noexec(&file
->f_path
))
844 err
= deny_write_access(file
);
848 if (name
->name
[0] != '\0')
851 trace_open_exec(name
->name
);
861 struct file
*open_exec(const char *name
)
863 struct filename
*filename
= getname_kernel(name
);
864 struct file
*f
= ERR_CAST(filename
);
866 if (!IS_ERR(filename
)) {
867 f
= do_open_execat(AT_FDCWD
, filename
, 0);
872 EXPORT_SYMBOL(open_exec
);
874 int kernel_read(struct file
*file
, loff_t offset
,
875 char *addr
, unsigned long count
)
883 /* The cast to a user pointer is valid due to the set_fs() */
884 result
= vfs_read(file
, (void __user
*)addr
, count
, &pos
);
889 EXPORT_SYMBOL(kernel_read
);
891 int kernel_read_file(struct file
*file
, void **buf
, loff_t
*size
,
892 loff_t max_size
, enum kernel_read_file_id id
)
898 if (!S_ISREG(file_inode(file
)->i_mode
) || max_size
< 0)
901 ret
= security_kernel_read_file(file
, id
);
905 ret
= deny_write_access(file
);
909 i_size
= i_size_read(file_inode(file
));
910 if (max_size
> 0 && i_size
> max_size
) {
919 if (id
!= READING_FIRMWARE_PREALLOC_BUFFER
)
920 *buf
= vmalloc(i_size
);
927 while (pos
< i_size
) {
928 bytes
= kernel_read(file
, pos
, (char *)(*buf
) + pos
,
945 ret
= security_kernel_post_read_file(file
, *buf
, i_size
, id
);
951 if (id
!= READING_FIRMWARE_PREALLOC_BUFFER
) {
958 allow_write_access(file
);
961 EXPORT_SYMBOL_GPL(kernel_read_file
);
963 int kernel_read_file_from_path(char *path
, void **buf
, loff_t
*size
,
964 loff_t max_size
, enum kernel_read_file_id id
)
972 file
= filp_open(path
, O_RDONLY
, 0);
974 return PTR_ERR(file
);
976 ret
= kernel_read_file(file
, buf
, size
, max_size
, id
);
980 EXPORT_SYMBOL_GPL(kernel_read_file_from_path
);
982 int kernel_read_file_from_fd(int fd
, void **buf
, loff_t
*size
, loff_t max_size
,
983 enum kernel_read_file_id id
)
985 struct fd f
= fdget(fd
);
991 ret
= kernel_read_file(f
.file
, buf
, size
, max_size
, id
);
996 EXPORT_SYMBOL_GPL(kernel_read_file_from_fd
);
998 ssize_t
read_code(struct file
*file
, unsigned long addr
, loff_t pos
, size_t len
)
1000 ssize_t res
= vfs_read(file
, (void __user
*)addr
, len
, &pos
);
1002 flush_icache_range(addr
, addr
+ len
);
1005 EXPORT_SYMBOL(read_code
);
1007 static int exec_mmap(struct mm_struct
*mm
)
1009 struct task_struct
*tsk
;
1010 struct mm_struct
*old_mm
, *active_mm
;
1012 /* Notify parent that we're no longer interested in the old VM */
1014 old_mm
= current
->mm
;
1015 mm_release(tsk
, old_mm
);
1018 sync_mm_rss(old_mm
);
1020 * Make sure that if there is a core dump in progress
1021 * for the old mm, we get out and die instead of going
1022 * through with the exec. We must hold mmap_sem around
1023 * checking core_state and changing tsk->mm.
1025 down_read(&old_mm
->mmap_sem
);
1026 if (unlikely(old_mm
->core_state
)) {
1027 up_read(&old_mm
->mmap_sem
);
1032 active_mm
= tsk
->active_mm
;
1034 tsk
->active_mm
= mm
;
1035 activate_mm(active_mm
, mm
);
1036 tsk
->mm
->vmacache_seqnum
= 0;
1037 vmacache_flush(tsk
);
1040 up_read(&old_mm
->mmap_sem
);
1041 BUG_ON(active_mm
!= old_mm
);
1042 setmax_mm_hiwater_rss(&tsk
->signal
->maxrss
, old_mm
);
1043 mm_update_next_owner(old_mm
);
1052 * This function makes sure the current process has its own signal table,
1053 * so that flush_signal_handlers can later reset the handlers without
1054 * disturbing other processes. (Other processes might share the signal
1055 * table via the CLONE_SIGHAND option to clone().)
1057 static int de_thread(struct task_struct
*tsk
)
1059 struct signal_struct
*sig
= tsk
->signal
;
1060 struct sighand_struct
*oldsighand
= tsk
->sighand
;
1061 spinlock_t
*lock
= &oldsighand
->siglock
;
1063 if (thread_group_empty(tsk
))
1064 goto no_thread_group
;
1067 * Kill all other threads in the thread group.
1069 spin_lock_irq(lock
);
1070 if (signal_group_exit(sig
)) {
1072 * Another group action in progress, just
1073 * return so that the signal is processed.
1075 spin_unlock_irq(lock
);
1079 sig
->group_exit_task
= tsk
;
1080 sig
->notify_count
= zap_other_threads(tsk
);
1081 if (!thread_group_leader(tsk
))
1082 sig
->notify_count
--;
1084 while (sig
->notify_count
) {
1085 __set_current_state(TASK_KILLABLE
);
1086 spin_unlock_irq(lock
);
1088 if (unlikely(__fatal_signal_pending(tsk
)))
1090 spin_lock_irq(lock
);
1092 spin_unlock_irq(lock
);
1095 * At this point all other threads have exited, all we have to
1096 * do is to wait for the thread group leader to become inactive,
1097 * and to assume its PID:
1099 if (!thread_group_leader(tsk
)) {
1100 struct task_struct
*leader
= tsk
->group_leader
;
1103 threadgroup_change_begin(tsk
);
1104 write_lock_irq(&tasklist_lock
);
1106 * Do this under tasklist_lock to ensure that
1107 * exit_notify() can't miss ->group_exit_task
1109 sig
->notify_count
= -1;
1110 if (likely(leader
->exit_state
))
1112 __set_current_state(TASK_KILLABLE
);
1113 write_unlock_irq(&tasklist_lock
);
1114 threadgroup_change_end(tsk
);
1116 if (unlikely(__fatal_signal_pending(tsk
)))
1121 * The only record we have of the real-time age of a
1122 * process, regardless of execs it's done, is start_time.
1123 * All the past CPU time is accumulated in signal_struct
1124 * from sister threads now dead. But in this non-leader
1125 * exec, nothing survives from the original leader thread,
1126 * whose birth marks the true age of this process now.
1127 * When we take on its identity by switching to its PID, we
1128 * also take its birthdate (always earlier than our own).
1130 tsk
->start_time
= leader
->start_time
;
1131 tsk
->real_start_time
= leader
->real_start_time
;
1133 BUG_ON(!same_thread_group(leader
, tsk
));
1134 BUG_ON(has_group_leader_pid(tsk
));
1136 * An exec() starts a new thread group with the
1137 * TGID of the previous thread group. Rehash the
1138 * two threads with a switched PID, and release
1139 * the former thread group leader:
1142 /* Become a process group leader with the old leader's pid.
1143 * The old leader becomes a thread of the this thread group.
1144 * Note: The old leader also uses this pid until release_task
1145 * is called. Odd but simple and correct.
1147 tsk
->pid
= leader
->pid
;
1148 change_pid(tsk
, PIDTYPE_PID
, task_pid(leader
));
1149 transfer_pid(leader
, tsk
, PIDTYPE_PGID
);
1150 transfer_pid(leader
, tsk
, PIDTYPE_SID
);
1152 list_replace_rcu(&leader
->tasks
, &tsk
->tasks
);
1153 list_replace_init(&leader
->sibling
, &tsk
->sibling
);
1155 tsk
->group_leader
= tsk
;
1156 leader
->group_leader
= tsk
;
1158 tsk
->exit_signal
= SIGCHLD
;
1159 leader
->exit_signal
= -1;
1161 BUG_ON(leader
->exit_state
!= EXIT_ZOMBIE
);
1162 leader
->exit_state
= EXIT_DEAD
;
1165 * We are going to release_task()->ptrace_unlink() silently,
1166 * the tracer can sleep in do_wait(). EXIT_DEAD guarantees
1167 * the tracer wont't block again waiting for this thread.
1169 if (unlikely(leader
->ptrace
))
1170 __wake_up_parent(leader
, leader
->parent
);
1171 write_unlock_irq(&tasklist_lock
);
1172 threadgroup_change_end(tsk
);
1174 release_task(leader
);
1177 sig
->group_exit_task
= NULL
;
1178 sig
->notify_count
= 0;
1181 /* we have changed execution domain */
1182 tsk
->exit_signal
= SIGCHLD
;
1184 #ifdef CONFIG_POSIX_TIMERS
1186 flush_itimer_signals();
1189 if (atomic_read(&oldsighand
->count
) != 1) {
1190 struct sighand_struct
*newsighand
;
1192 * This ->sighand is shared with the CLONE_SIGHAND
1193 * but not CLONE_THREAD task, switch to the new one.
1195 newsighand
= kmem_cache_alloc(sighand_cachep
, GFP_KERNEL
);
1199 atomic_set(&newsighand
->count
, 1);
1200 memcpy(newsighand
->action
, oldsighand
->action
,
1201 sizeof(newsighand
->action
));
1203 write_lock_irq(&tasklist_lock
);
1204 spin_lock(&oldsighand
->siglock
);
1205 rcu_assign_pointer(tsk
->sighand
, newsighand
);
1206 spin_unlock(&oldsighand
->siglock
);
1207 write_unlock_irq(&tasklist_lock
);
1209 __cleanup_sighand(oldsighand
);
1212 BUG_ON(!thread_group_leader(tsk
));
1216 /* protects against exit_notify() and __exit_signal() */
1217 read_lock(&tasklist_lock
);
1218 sig
->group_exit_task
= NULL
;
1219 sig
->notify_count
= 0;
1220 read_unlock(&tasklist_lock
);
1224 char *get_task_comm(char *buf
, struct task_struct
*tsk
)
1226 /* buf must be at least sizeof(tsk->comm) in size */
1228 strncpy(buf
, tsk
->comm
, sizeof(tsk
->comm
));
1232 EXPORT_SYMBOL_GPL(get_task_comm
);
1235 * These functions flushes out all traces of the currently running executable
1236 * so that a new one can be started
1239 void __set_task_comm(struct task_struct
*tsk
, const char *buf
, bool exec
)
1242 trace_task_rename(tsk
, buf
);
1243 strlcpy(tsk
->comm
, buf
, sizeof(tsk
->comm
));
1245 perf_event_comm(tsk
, exec
);
1248 int flush_old_exec(struct linux_binprm
* bprm
)
1253 * Make sure we have a private signal table and that
1254 * we are unassociated from the previous thread group.
1256 retval
= de_thread(current
);
1261 * Must be called _before_ exec_mmap() as bprm->mm is
1262 * not visibile until then. This also enables the update
1265 set_mm_exe_file(bprm
->mm
, bprm
->file
);
1268 * Release all of the old mmap stuff
1270 acct_arg_size(bprm
, 0);
1271 retval
= exec_mmap(bprm
->mm
);
1275 bprm
->mm
= NULL
; /* We're using it now */
1278 current
->flags
&= ~(PF_RANDOMIZE
| PF_FORKNOEXEC
| PF_KTHREAD
|
1279 PF_NOFREEZE
| PF_NO_SETAFFINITY
);
1281 current
->personality
&= ~bprm
->per_clear
;
1284 * We have to apply CLOEXEC before we change whether the process is
1285 * dumpable (in setup_new_exec) to avoid a race with a process in userspace
1286 * trying to access the should-be-closed file descriptors of a process
1287 * undergoing exec(2).
1289 do_close_on_exec(current
->files
);
1295 EXPORT_SYMBOL(flush_old_exec
);
1297 void would_dump(struct linux_binprm
*bprm
, struct file
*file
)
1299 struct inode
*inode
= file_inode(file
);
1300 if (inode_permission(inode
, MAY_READ
) < 0) {
1301 struct user_namespace
*old
, *user_ns
;
1302 bprm
->interp_flags
|= BINPRM_FLAGS_ENFORCE_NONDUMP
;
1304 /* Ensure mm->user_ns contains the executable */
1305 user_ns
= old
= bprm
->mm
->user_ns
;
1306 while ((user_ns
!= &init_user_ns
) &&
1307 !privileged_wrt_inode_uidgid(user_ns
, inode
))
1308 user_ns
= user_ns
->parent
;
1310 if (old
!= user_ns
) {
1311 bprm
->mm
->user_ns
= get_user_ns(user_ns
);
1316 EXPORT_SYMBOL(would_dump
);
1318 void setup_new_exec(struct linux_binprm
* bprm
)
1320 arch_pick_mmap_layout(current
->mm
);
1322 /* This is the point of no return */
1323 current
->sas_ss_sp
= current
->sas_ss_size
= 0;
1325 if (uid_eq(current_euid(), current_uid()) && gid_eq(current_egid(), current_gid()))
1326 set_dumpable(current
->mm
, SUID_DUMP_USER
);
1328 set_dumpable(current
->mm
, suid_dumpable
);
1331 __set_task_comm(current
, kbasename(bprm
->filename
), true);
1333 /* Set the new mm task size. We have to do that late because it may
1334 * depend on TIF_32BIT which is only updated in flush_thread() on
1335 * some architectures like powerpc
1337 current
->mm
->task_size
= TASK_SIZE
;
1339 /* install the new credentials */
1340 if (!uid_eq(bprm
->cred
->uid
, current_euid()) ||
1341 !gid_eq(bprm
->cred
->gid
, current_egid())) {
1342 current
->pdeath_signal
= 0;
1344 if (bprm
->interp_flags
& BINPRM_FLAGS_ENFORCE_NONDUMP
)
1345 set_dumpable(current
->mm
, suid_dumpable
);
1348 /* An exec changes our domain. We are no longer part of the thread
1350 current
->self_exec_id
++;
1351 flush_signal_handlers(current
, 0);
1353 EXPORT_SYMBOL(setup_new_exec
);
1356 * Prepare credentials and lock ->cred_guard_mutex.
1357 * install_exec_creds() commits the new creds and drops the lock.
1358 * Or, if exec fails before, free_bprm() should release ->cred and
1361 int prepare_bprm_creds(struct linux_binprm
*bprm
)
1363 if (mutex_lock_interruptible(¤t
->signal
->cred_guard_mutex
))
1364 return -ERESTARTNOINTR
;
1366 bprm
->cred
= prepare_exec_creds();
1367 if (likely(bprm
->cred
))
1370 mutex_unlock(¤t
->signal
->cred_guard_mutex
);
1374 static void free_bprm(struct linux_binprm
*bprm
)
1376 free_arg_pages(bprm
);
1378 mutex_unlock(¤t
->signal
->cred_guard_mutex
);
1379 abort_creds(bprm
->cred
);
1382 allow_write_access(bprm
->file
);
1385 /* If a binfmt changed the interp, free it. */
1386 if (bprm
->interp
!= bprm
->filename
)
1387 kfree(bprm
->interp
);
1391 int bprm_change_interp(char *interp
, struct linux_binprm
*bprm
)
1393 /* If a binfmt changed the interp, free it first. */
1394 if (bprm
->interp
!= bprm
->filename
)
1395 kfree(bprm
->interp
);
1396 bprm
->interp
= kstrdup(interp
, GFP_KERNEL
);
1401 EXPORT_SYMBOL(bprm_change_interp
);
1404 * install the new credentials for this executable
1406 void install_exec_creds(struct linux_binprm
*bprm
)
1408 security_bprm_committing_creds(bprm
);
1410 commit_creds(bprm
->cred
);
1414 * Disable monitoring for regular users
1415 * when executing setuid binaries. Must
1416 * wait until new credentials are committed
1417 * by commit_creds() above
1419 if (get_dumpable(current
->mm
) != SUID_DUMP_USER
)
1420 perf_event_exit_task(current
);
1422 * cred_guard_mutex must be held at least to this point to prevent
1423 * ptrace_attach() from altering our determination of the task's
1424 * credentials; any time after this it may be unlocked.
1426 security_bprm_committed_creds(bprm
);
1427 mutex_unlock(¤t
->signal
->cred_guard_mutex
);
1429 EXPORT_SYMBOL(install_exec_creds
);
1432 * determine how safe it is to execute the proposed program
1433 * - the caller must hold ->cred_guard_mutex to protect against
1434 * PTRACE_ATTACH or seccomp thread-sync
1436 static void check_unsafe_exec(struct linux_binprm
*bprm
)
1438 struct task_struct
*p
= current
, *t
;
1442 if (ptracer_capable(p
, current_user_ns()))
1443 bprm
->unsafe
|= LSM_UNSAFE_PTRACE_CAP
;
1445 bprm
->unsafe
|= LSM_UNSAFE_PTRACE
;
1449 * This isn't strictly necessary, but it makes it harder for LSMs to
1452 if (task_no_new_privs(current
))
1453 bprm
->unsafe
|= LSM_UNSAFE_NO_NEW_PRIVS
;
1457 spin_lock(&p
->fs
->lock
);
1459 while_each_thread(p
, t
) {
1465 if (p
->fs
->users
> n_fs
)
1466 bprm
->unsafe
|= LSM_UNSAFE_SHARE
;
1469 spin_unlock(&p
->fs
->lock
);
1472 static void bprm_fill_uid(struct linux_binprm
*bprm
)
1474 struct inode
*inode
;
1480 * Since this can be called multiple times (via prepare_binprm),
1481 * we must clear any previous work done when setting set[ug]id
1482 * bits from any earlier bprm->file uses (for example when run
1483 * first for a setuid script then again for its interpreter).
1485 bprm
->cred
->euid
= current_euid();
1486 bprm
->cred
->egid
= current_egid();
1488 if (path_nosuid(&bprm
->file
->f_path
))
1491 if (task_no_new_privs(current
))
1494 inode
= file_inode(bprm
->file
);
1495 mode
= READ_ONCE(inode
->i_mode
);
1496 if (!(mode
& (S_ISUID
|S_ISGID
)))
1499 /* Be careful if suid/sgid is set */
1502 /* reload atomically mode/uid/gid now that lock held */
1503 mode
= inode
->i_mode
;
1506 inode_unlock(inode
);
1508 /* We ignore suid/sgid if there are no mappings for them in the ns */
1509 if (!kuid_has_mapping(bprm
->cred
->user_ns
, uid
) ||
1510 !kgid_has_mapping(bprm
->cred
->user_ns
, gid
))
1513 if (mode
& S_ISUID
) {
1514 bprm
->per_clear
|= PER_CLEAR_ON_SETID
;
1515 bprm
->cred
->euid
= uid
;
1518 if ((mode
& (S_ISGID
| S_IXGRP
)) == (S_ISGID
| S_IXGRP
)) {
1519 bprm
->per_clear
|= PER_CLEAR_ON_SETID
;
1520 bprm
->cred
->egid
= gid
;
1525 * Fill the binprm structure from the inode.
1526 * Check permissions, then read the first 128 (BINPRM_BUF_SIZE) bytes
1528 * This may be called multiple times for binary chains (scripts for example).
1530 int prepare_binprm(struct linux_binprm
*bprm
)
1534 bprm_fill_uid(bprm
);
1536 /* fill in binprm security blob */
1537 retval
= security_bprm_set_creds(bprm
);
1540 bprm
->cred_prepared
= 1;
1542 memset(bprm
->buf
, 0, BINPRM_BUF_SIZE
);
1543 return kernel_read(bprm
->file
, 0, bprm
->buf
, BINPRM_BUF_SIZE
);
1546 EXPORT_SYMBOL(prepare_binprm
);
1549 * Arguments are '\0' separated strings found at the location bprm->p
1550 * points to; chop off the first by relocating brpm->p to right after
1551 * the first '\0' encountered.
1553 int remove_arg_zero(struct linux_binprm
*bprm
)
1556 unsigned long offset
;
1564 offset
= bprm
->p
& ~PAGE_MASK
;
1565 page
= get_arg_page(bprm
, bprm
->p
, 0);
1570 kaddr
= kmap_atomic(page
);
1572 for (; offset
< PAGE_SIZE
&& kaddr
[offset
];
1573 offset
++, bprm
->p
++)
1576 kunmap_atomic(kaddr
);
1578 } while (offset
== PAGE_SIZE
);
1587 EXPORT_SYMBOL(remove_arg_zero
);
1589 #define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e))
1591 * cycle the list of binary formats handler, until one recognizes the image
1593 int search_binary_handler(struct linux_binprm
*bprm
)
1595 bool need_retry
= IS_ENABLED(CONFIG_MODULES
);
1596 struct linux_binfmt
*fmt
;
1599 /* This allows 4 levels of binfmt rewrites before failing hard. */
1600 if (bprm
->recursion_depth
> 5)
1603 retval
= security_bprm_check(bprm
);
1609 read_lock(&binfmt_lock
);
1610 list_for_each_entry(fmt
, &formats
, lh
) {
1611 if (!try_module_get(fmt
->module
))
1613 read_unlock(&binfmt_lock
);
1614 bprm
->recursion_depth
++;
1615 retval
= fmt
->load_binary(bprm
);
1616 read_lock(&binfmt_lock
);
1618 bprm
->recursion_depth
--;
1619 if (retval
< 0 && !bprm
->mm
) {
1620 /* we got to flush_old_exec() and failed after it */
1621 read_unlock(&binfmt_lock
);
1622 force_sigsegv(SIGSEGV
, current
);
1625 if (retval
!= -ENOEXEC
|| !bprm
->file
) {
1626 read_unlock(&binfmt_lock
);
1630 read_unlock(&binfmt_lock
);
1633 if (printable(bprm
->buf
[0]) && printable(bprm
->buf
[1]) &&
1634 printable(bprm
->buf
[2]) && printable(bprm
->buf
[3]))
1636 if (request_module("binfmt-%04x", *(ushort
*)(bprm
->buf
+ 2)) < 0)
1644 EXPORT_SYMBOL(search_binary_handler
);
1646 static int exec_binprm(struct linux_binprm
*bprm
)
1648 pid_t old_pid
, old_vpid
;
1651 /* Need to fetch pid before load_binary changes it */
1652 old_pid
= current
->pid
;
1654 old_vpid
= task_pid_nr_ns(current
, task_active_pid_ns(current
->parent
));
1657 ret
= search_binary_handler(bprm
);
1660 trace_sched_process_exec(current
, old_pid
, bprm
);
1661 ptrace_event(PTRACE_EVENT_EXEC
, old_vpid
);
1662 proc_exec_connector(current
);
1669 * sys_execve() executes a new program.
1671 static int do_execveat_common(int fd
, struct filename
*filename
,
1672 struct user_arg_ptr argv
,
1673 struct user_arg_ptr envp
,
1676 char *pathbuf
= NULL
;
1677 struct linux_binprm
*bprm
;
1679 struct files_struct
*displaced
;
1682 if (IS_ERR(filename
))
1683 return PTR_ERR(filename
);
1686 * We move the actual failure in case of RLIMIT_NPROC excess from
1687 * set*uid() to execve() because too many poorly written programs
1688 * don't check setuid() return code. Here we additionally recheck
1689 * whether NPROC limit is still exceeded.
1691 if ((current
->flags
& PF_NPROC_EXCEEDED
) &&
1692 atomic_read(¤t_user()->processes
) > rlimit(RLIMIT_NPROC
)) {
1697 /* We're below the limit (still or again), so we don't want to make
1698 * further execve() calls fail. */
1699 current
->flags
&= ~PF_NPROC_EXCEEDED
;
1701 retval
= unshare_files(&displaced
);
1706 bprm
= kzalloc(sizeof(*bprm
), GFP_KERNEL
);
1710 retval
= prepare_bprm_creds(bprm
);
1714 check_unsafe_exec(bprm
);
1715 current
->in_execve
= 1;
1717 file
= do_open_execat(fd
, filename
, flags
);
1718 retval
= PTR_ERR(file
);
1725 if (fd
== AT_FDCWD
|| filename
->name
[0] == '/') {
1726 bprm
->filename
= filename
->name
;
1728 if (filename
->name
[0] == '\0')
1729 pathbuf
= kasprintf(GFP_TEMPORARY
, "/dev/fd/%d", fd
);
1731 pathbuf
= kasprintf(GFP_TEMPORARY
, "/dev/fd/%d/%s",
1732 fd
, filename
->name
);
1738 * Record that a name derived from an O_CLOEXEC fd will be
1739 * inaccessible after exec. Relies on having exclusive access to
1740 * current->files (due to unshare_files above).
1742 if (close_on_exec(fd
, rcu_dereference_raw(current
->files
->fdt
)))
1743 bprm
->interp_flags
|= BINPRM_FLAGS_PATH_INACCESSIBLE
;
1744 bprm
->filename
= pathbuf
;
1746 bprm
->interp
= bprm
->filename
;
1748 retval
= bprm_mm_init(bprm
);
1752 bprm
->argc
= count(argv
, MAX_ARG_STRINGS
);
1753 if ((retval
= bprm
->argc
) < 0)
1756 bprm
->envc
= count(envp
, MAX_ARG_STRINGS
);
1757 if ((retval
= bprm
->envc
) < 0)
1760 retval
= prepare_binprm(bprm
);
1764 retval
= copy_strings_kernel(1, &bprm
->filename
, bprm
);
1768 bprm
->exec
= bprm
->p
;
1769 retval
= copy_strings(bprm
->envc
, envp
, bprm
);
1773 retval
= copy_strings(bprm
->argc
, argv
, bprm
);
1777 would_dump(bprm
, bprm
->file
);
1779 retval
= exec_binprm(bprm
);
1783 /* execve succeeded */
1784 current
->fs
->in_exec
= 0;
1785 current
->in_execve
= 0;
1786 acct_update_integrals(current
);
1787 task_numa_free(current
);
1792 put_files_struct(displaced
);
1797 acct_arg_size(bprm
, 0);
1802 current
->fs
->in_exec
= 0;
1803 current
->in_execve
= 0;
1811 reset_files_struct(displaced
);
1817 int do_execve(struct filename
*filename
,
1818 const char __user
*const __user
*__argv
,
1819 const char __user
*const __user
*__envp
)
1821 struct user_arg_ptr argv
= { .ptr
.native
= __argv
};
1822 struct user_arg_ptr envp
= { .ptr
.native
= __envp
};
1823 return do_execveat_common(AT_FDCWD
, filename
, argv
, envp
, 0);
1826 int do_execveat(int fd
, struct filename
*filename
,
1827 const char __user
*const __user
*__argv
,
1828 const char __user
*const __user
*__envp
,
1831 struct user_arg_ptr argv
= { .ptr
.native
= __argv
};
1832 struct user_arg_ptr envp
= { .ptr
.native
= __envp
};
1834 return do_execveat_common(fd
, filename
, argv
, envp
, flags
);
1837 #ifdef CONFIG_COMPAT
1838 static int compat_do_execve(struct filename
*filename
,
1839 const compat_uptr_t __user
*__argv
,
1840 const compat_uptr_t __user
*__envp
)
1842 struct user_arg_ptr argv
= {
1844 .ptr
.compat
= __argv
,
1846 struct user_arg_ptr envp
= {
1848 .ptr
.compat
= __envp
,
1850 return do_execveat_common(AT_FDCWD
, filename
, argv
, envp
, 0);
1853 static int compat_do_execveat(int fd
, struct filename
*filename
,
1854 const compat_uptr_t __user
*__argv
,
1855 const compat_uptr_t __user
*__envp
,
1858 struct user_arg_ptr argv
= {
1860 .ptr
.compat
= __argv
,
1862 struct user_arg_ptr envp
= {
1864 .ptr
.compat
= __envp
,
1866 return do_execveat_common(fd
, filename
, argv
, envp
, flags
);
1870 void set_binfmt(struct linux_binfmt
*new)
1872 struct mm_struct
*mm
= current
->mm
;
1875 module_put(mm
->binfmt
->module
);
1879 __module_get(new->module
);
1881 EXPORT_SYMBOL(set_binfmt
);
1884 * set_dumpable stores three-value SUID_DUMP_* into mm->flags.
1886 void set_dumpable(struct mm_struct
*mm
, int value
)
1888 unsigned long old
, new;
1890 if (WARN_ON((unsigned)value
> SUID_DUMP_ROOT
))
1894 old
= ACCESS_ONCE(mm
->flags
);
1895 new = (old
& ~MMF_DUMPABLE_MASK
) | value
;
1896 } while (cmpxchg(&mm
->flags
, old
, new) != old
);
1899 SYSCALL_DEFINE3(execve
,
1900 const char __user
*, filename
,
1901 const char __user
*const __user
*, argv
,
1902 const char __user
*const __user
*, envp
)
1904 return do_execve(getname(filename
), argv
, envp
);
1907 SYSCALL_DEFINE5(execveat
,
1908 int, fd
, const char __user
*, filename
,
1909 const char __user
*const __user
*, argv
,
1910 const char __user
*const __user
*, envp
,
1913 int lookup_flags
= (flags
& AT_EMPTY_PATH
) ? LOOKUP_EMPTY
: 0;
1915 return do_execveat(fd
,
1916 getname_flags(filename
, lookup_flags
, NULL
),
1920 #ifdef CONFIG_COMPAT
1921 COMPAT_SYSCALL_DEFINE3(execve
, const char __user
*, filename
,
1922 const compat_uptr_t __user
*, argv
,
1923 const compat_uptr_t __user
*, envp
)
1925 return compat_do_execve(getname(filename
), argv
, envp
);
1928 COMPAT_SYSCALL_DEFINE5(execveat
, int, fd
,
1929 const char __user
*, filename
,
1930 const compat_uptr_t __user
*, argv
,
1931 const compat_uptr_t __user
*, envp
,
1934 int lookup_flags
= (flags
& AT_EMPTY_PATH
) ? LOOKUP_EMPTY
: 0;
1936 return compat_do_execveat(fd
,
1937 getname_flags(filename
, lookup_flags
, NULL
),