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
;
228 unsigned long ptr_size
;
232 * Since the stack will hold pointers to the strings, we
233 * must account for them as well.
235 * The size calculation is the entire vma while each arg page is
236 * built, so each time we get here it's calculating how far it
237 * is currently (rather than each call being just the newly
238 * added size from the arg page). As a result, we need to
239 * always add the entire size of the pointers, so that on the
240 * last call to get_arg_page() we'll actually have the entire
243 ptr_size
= (bprm
->argc
+ bprm
->envc
) * sizeof(void *);
244 if (ptr_size
> ULONG_MAX
- size
)
248 acct_arg_size(bprm
, size
/ PAGE_SIZE
);
251 * We've historically supported up to 32 pages (ARG_MAX)
252 * of argument strings even with small stacks
258 * Limit to 1/4-th the stack size for the argv+env strings.
260 * - the remaining binfmt code will not run out of stack space,
261 * - the program will have a reasonable amount of stack left
264 rlim
= current
->signal
->rlim
;
265 if (size
> READ_ONCE(rlim
[RLIMIT_STACK
].rlim_cur
) / 4)
276 static void put_arg_page(struct page
*page
)
281 static void free_arg_pages(struct linux_binprm
*bprm
)
285 static void flush_arg_page(struct linux_binprm
*bprm
, unsigned long pos
,
288 flush_cache_page(bprm
->vma
, pos
, page_to_pfn(page
));
291 static int __bprm_mm_init(struct linux_binprm
*bprm
)
294 struct vm_area_struct
*vma
= NULL
;
295 struct mm_struct
*mm
= bprm
->mm
;
297 bprm
->vma
= vma
= kmem_cache_zalloc(vm_area_cachep
, GFP_KERNEL
);
301 if (down_write_killable(&mm
->mmap_sem
)) {
308 * Place the stack at the largest stack address the architecture
309 * supports. Later, we'll move this to an appropriate place. We don't
310 * use STACK_TOP because that can depend on attributes which aren't
313 BUILD_BUG_ON(VM_STACK_FLAGS
& VM_STACK_INCOMPLETE_SETUP
);
314 vma
->vm_end
= STACK_TOP_MAX
;
315 vma
->vm_start
= vma
->vm_end
- PAGE_SIZE
;
316 vma
->vm_flags
= VM_SOFTDIRTY
| VM_STACK_FLAGS
| VM_STACK_INCOMPLETE_SETUP
;
317 vma
->vm_page_prot
= vm_get_page_prot(vma
->vm_flags
);
318 INIT_LIST_HEAD(&vma
->anon_vma_chain
);
320 err
= insert_vm_struct(mm
, vma
);
324 mm
->stack_vm
= mm
->total_vm
= 1;
325 arch_bprm_mm_init(mm
, vma
);
326 up_write(&mm
->mmap_sem
);
327 bprm
->p
= vma
->vm_end
- sizeof(void *);
330 up_write(&mm
->mmap_sem
);
333 kmem_cache_free(vm_area_cachep
, vma
);
337 static bool valid_arg_len(struct linux_binprm
*bprm
, long len
)
339 return len
<= MAX_ARG_STRLEN
;
344 static inline void acct_arg_size(struct linux_binprm
*bprm
, unsigned long pages
)
348 static struct page
*get_arg_page(struct linux_binprm
*bprm
, unsigned long pos
,
353 page
= bprm
->page
[pos
/ PAGE_SIZE
];
354 if (!page
&& write
) {
355 page
= alloc_page(GFP_HIGHUSER
|__GFP_ZERO
);
358 bprm
->page
[pos
/ PAGE_SIZE
] = page
;
364 static void put_arg_page(struct page
*page
)
368 static void free_arg_page(struct linux_binprm
*bprm
, int i
)
371 __free_page(bprm
->page
[i
]);
372 bprm
->page
[i
] = NULL
;
376 static void free_arg_pages(struct linux_binprm
*bprm
)
380 for (i
= 0; i
< MAX_ARG_PAGES
; i
++)
381 free_arg_page(bprm
, i
);
384 static void flush_arg_page(struct linux_binprm
*bprm
, unsigned long pos
,
389 static int __bprm_mm_init(struct linux_binprm
*bprm
)
391 bprm
->p
= PAGE_SIZE
* MAX_ARG_PAGES
- sizeof(void *);
395 static bool valid_arg_len(struct linux_binprm
*bprm
, long len
)
397 return len
<= bprm
->p
;
400 #endif /* CONFIG_MMU */
403 * Create a new mm_struct and populate it with a temporary stack
404 * vm_area_struct. We don't have enough context at this point to set the stack
405 * flags, permissions, and offset, so we use temporary values. We'll update
406 * them later in setup_arg_pages().
408 static int bprm_mm_init(struct linux_binprm
*bprm
)
411 struct mm_struct
*mm
= NULL
;
413 bprm
->mm
= mm
= mm_alloc();
418 err
= __bprm_mm_init(bprm
);
433 struct user_arg_ptr
{
438 const char __user
*const __user
*native
;
440 const compat_uptr_t __user
*compat
;
445 static const char __user
*get_user_arg_ptr(struct user_arg_ptr argv
, int nr
)
447 const char __user
*native
;
450 if (unlikely(argv
.is_compat
)) {
451 compat_uptr_t compat
;
453 if (get_user(compat
, argv
.ptr
.compat
+ nr
))
454 return ERR_PTR(-EFAULT
);
456 return compat_ptr(compat
);
460 if (get_user(native
, argv
.ptr
.native
+ nr
))
461 return ERR_PTR(-EFAULT
);
467 * count() counts the number of strings in array ARGV.
469 static int count(struct user_arg_ptr argv
, int max
)
473 if (argv
.ptr
.native
!= NULL
) {
475 const char __user
*p
= get_user_arg_ptr(argv
, i
);
487 if (fatal_signal_pending(current
))
488 return -ERESTARTNOHAND
;
496 * 'copy_strings()' copies argument/environment strings from the old
497 * processes's memory to the new process's stack. The call to get_user_pages()
498 * ensures the destination page is created and not swapped out.
500 static int copy_strings(int argc
, struct user_arg_ptr argv
,
501 struct linux_binprm
*bprm
)
503 struct page
*kmapped_page
= NULL
;
505 unsigned long kpos
= 0;
509 const char __user
*str
;
514 str
= get_user_arg_ptr(argv
, argc
);
518 len
= strnlen_user(str
, MAX_ARG_STRLEN
);
523 if (!valid_arg_len(bprm
, len
))
526 /* We're going to work our way backwords. */
532 int offset
, bytes_to_copy
;
534 if (fatal_signal_pending(current
)) {
535 ret
= -ERESTARTNOHAND
;
540 offset
= pos
% PAGE_SIZE
;
544 bytes_to_copy
= offset
;
545 if (bytes_to_copy
> len
)
548 offset
-= bytes_to_copy
;
549 pos
-= bytes_to_copy
;
550 str
-= bytes_to_copy
;
551 len
-= bytes_to_copy
;
553 if (!kmapped_page
|| kpos
!= (pos
& PAGE_MASK
)) {
556 page
= get_arg_page(bprm
, pos
, 1);
563 flush_kernel_dcache_page(kmapped_page
);
564 kunmap(kmapped_page
);
565 put_arg_page(kmapped_page
);
568 kaddr
= kmap(kmapped_page
);
569 kpos
= pos
& PAGE_MASK
;
570 flush_arg_page(bprm
, kpos
, kmapped_page
);
572 if (copy_from_user(kaddr
+offset
, str
, bytes_to_copy
)) {
581 flush_kernel_dcache_page(kmapped_page
);
582 kunmap(kmapped_page
);
583 put_arg_page(kmapped_page
);
589 * Like copy_strings, but get argv and its values from kernel memory.
591 int copy_strings_kernel(int argc
, const char *const *__argv
,
592 struct linux_binprm
*bprm
)
595 mm_segment_t oldfs
= get_fs();
596 struct user_arg_ptr argv
= {
597 .ptr
.native
= (const char __user
*const __user
*)__argv
,
601 r
= copy_strings(argc
, argv
, bprm
);
606 EXPORT_SYMBOL(copy_strings_kernel
);
611 * During bprm_mm_init(), we create a temporary stack at STACK_TOP_MAX. Once
612 * the binfmt code determines where the new stack should reside, we shift it to
613 * its final location. The process proceeds as follows:
615 * 1) Use shift to calculate the new vma endpoints.
616 * 2) Extend vma to cover both the old and new ranges. This ensures the
617 * arguments passed to subsequent functions are consistent.
618 * 3) Move vma's page tables to the new range.
619 * 4) Free up any cleared pgd range.
620 * 5) Shrink the vma to cover only the new range.
622 static int shift_arg_pages(struct vm_area_struct
*vma
, unsigned long shift
)
624 struct mm_struct
*mm
= vma
->vm_mm
;
625 unsigned long old_start
= vma
->vm_start
;
626 unsigned long old_end
= vma
->vm_end
;
627 unsigned long length
= old_end
- old_start
;
628 unsigned long new_start
= old_start
- shift
;
629 unsigned long new_end
= old_end
- shift
;
630 struct mmu_gather tlb
;
632 BUG_ON(new_start
> new_end
);
635 * ensure there are no vmas between where we want to go
638 if (vma
!= find_vma(mm
, new_start
))
642 * cover the whole range: [new_start, old_end)
644 if (vma_adjust(vma
, new_start
, old_end
, vma
->vm_pgoff
, NULL
))
648 * move the page tables downwards, on failure we rely on
649 * process cleanup to remove whatever mess we made.
651 if (length
!= move_page_tables(vma
, old_start
,
652 vma
, new_start
, length
, false))
656 tlb_gather_mmu(&tlb
, mm
, old_start
, old_end
);
657 if (new_end
> old_start
) {
659 * when the old and new regions overlap clear from new_end.
661 free_pgd_range(&tlb
, new_end
, old_end
, new_end
,
662 vma
->vm_next
? vma
->vm_next
->vm_start
: USER_PGTABLES_CEILING
);
665 * otherwise, clean from old_start; this is done to not touch
666 * the address space in [new_end, old_start) some architectures
667 * have constraints on va-space that make this illegal (IA64) -
668 * for the others its just a little faster.
670 free_pgd_range(&tlb
, old_start
, old_end
, new_end
,
671 vma
->vm_next
? vma
->vm_next
->vm_start
: USER_PGTABLES_CEILING
);
673 tlb_finish_mmu(&tlb
, old_start
, old_end
);
676 * Shrink the vma to just the new range. Always succeeds.
678 vma_adjust(vma
, new_start
, new_end
, vma
->vm_pgoff
, NULL
);
684 * Finalizes the stack vm_area_struct. The flags and permissions are updated,
685 * the stack is optionally relocated, and some extra space is added.
687 int setup_arg_pages(struct linux_binprm
*bprm
,
688 unsigned long stack_top
,
689 int executable_stack
)
692 unsigned long stack_shift
;
693 struct mm_struct
*mm
= current
->mm
;
694 struct vm_area_struct
*vma
= bprm
->vma
;
695 struct vm_area_struct
*prev
= NULL
;
696 unsigned long vm_flags
;
697 unsigned long stack_base
;
698 unsigned long stack_size
;
699 unsigned long stack_expand
;
700 unsigned long rlim_stack
;
702 #ifdef CONFIG_STACK_GROWSUP
703 /* Limit stack size */
704 stack_base
= rlimit_max(RLIMIT_STACK
);
705 if (stack_base
> STACK_SIZE_MAX
)
706 stack_base
= STACK_SIZE_MAX
;
708 /* Add space for stack randomization. */
709 stack_base
+= (STACK_RND_MASK
<< PAGE_SHIFT
);
711 /* Make sure we didn't let the argument array grow too large. */
712 if (vma
->vm_end
- vma
->vm_start
> stack_base
)
715 stack_base
= PAGE_ALIGN(stack_top
- stack_base
);
717 stack_shift
= vma
->vm_start
- stack_base
;
718 mm
->arg_start
= bprm
->p
- stack_shift
;
719 bprm
->p
= vma
->vm_end
- stack_shift
;
721 stack_top
= arch_align_stack(stack_top
);
722 stack_top
= PAGE_ALIGN(stack_top
);
724 if (unlikely(stack_top
< mmap_min_addr
) ||
725 unlikely(vma
->vm_end
- vma
->vm_start
>= stack_top
- mmap_min_addr
))
728 stack_shift
= vma
->vm_end
- stack_top
;
730 bprm
->p
-= stack_shift
;
731 mm
->arg_start
= bprm
->p
;
735 bprm
->loader
-= stack_shift
;
736 bprm
->exec
-= stack_shift
;
738 if (down_write_killable(&mm
->mmap_sem
))
741 vm_flags
= VM_STACK_FLAGS
;
744 * Adjust stack execute permissions; explicitly enable for
745 * EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X and leave alone
746 * (arch default) otherwise.
748 if (unlikely(executable_stack
== EXSTACK_ENABLE_X
))
750 else if (executable_stack
== EXSTACK_DISABLE_X
)
751 vm_flags
&= ~VM_EXEC
;
752 vm_flags
|= mm
->def_flags
;
753 vm_flags
|= VM_STACK_INCOMPLETE_SETUP
;
755 ret
= mprotect_fixup(vma
, &prev
, vma
->vm_start
, vma
->vm_end
,
761 /* Move stack pages down in memory. */
763 ret
= shift_arg_pages(vma
, stack_shift
);
768 /* mprotect_fixup is overkill to remove the temporary stack flags */
769 vma
->vm_flags
&= ~VM_STACK_INCOMPLETE_SETUP
;
771 stack_expand
= 131072UL; /* randomly 32*4k (or 2*64k) pages */
772 stack_size
= vma
->vm_end
- vma
->vm_start
;
774 * Align this down to a page boundary as expand_stack
777 rlim_stack
= rlimit(RLIMIT_STACK
) & PAGE_MASK
;
778 #ifdef CONFIG_STACK_GROWSUP
779 if (stack_size
+ stack_expand
> rlim_stack
)
780 stack_base
= vma
->vm_start
+ rlim_stack
;
782 stack_base
= vma
->vm_end
+ stack_expand
;
784 if (stack_size
+ stack_expand
> rlim_stack
)
785 stack_base
= vma
->vm_end
- rlim_stack
;
787 stack_base
= vma
->vm_start
- stack_expand
;
789 current
->mm
->start_stack
= bprm
->p
;
790 ret
= expand_stack(vma
, stack_base
);
795 up_write(&mm
->mmap_sem
);
798 EXPORT_SYMBOL(setup_arg_pages
);
803 * Transfer the program arguments and environment from the holding pages
804 * onto the stack. The provided stack pointer is adjusted accordingly.
806 int transfer_args_to_stack(struct linux_binprm
*bprm
,
807 unsigned long *sp_location
)
809 unsigned long index
, stop
, sp
;
812 stop
= bprm
->p
>> PAGE_SHIFT
;
815 for (index
= MAX_ARG_PAGES
- 1; index
>= stop
; index
--) {
816 unsigned int offset
= index
== stop
? bprm
->p
& ~PAGE_MASK
: 0;
817 char *src
= kmap(bprm
->page
[index
]) + offset
;
818 sp
-= PAGE_SIZE
- offset
;
819 if (copy_to_user((void *) sp
, src
, PAGE_SIZE
- offset
) != 0)
821 kunmap(bprm
->page
[index
]);
831 EXPORT_SYMBOL(transfer_args_to_stack
);
833 #endif /* CONFIG_MMU */
835 static struct file
*do_open_execat(int fd
, struct filename
*name
, int flags
)
839 struct open_flags open_exec_flags
= {
840 .open_flag
= O_LARGEFILE
| O_RDONLY
| __FMODE_EXEC
,
841 .acc_mode
= MAY_EXEC
,
842 .intent
= LOOKUP_OPEN
,
843 .lookup_flags
= LOOKUP_FOLLOW
,
846 if ((flags
& ~(AT_SYMLINK_NOFOLLOW
| AT_EMPTY_PATH
)) != 0)
847 return ERR_PTR(-EINVAL
);
848 if (flags
& AT_SYMLINK_NOFOLLOW
)
849 open_exec_flags
.lookup_flags
&= ~LOOKUP_FOLLOW
;
850 if (flags
& AT_EMPTY_PATH
)
851 open_exec_flags
.lookup_flags
|= LOOKUP_EMPTY
;
853 file
= do_filp_open(fd
, name
, &open_exec_flags
);
858 if (!S_ISREG(file_inode(file
)->i_mode
))
861 if (path_noexec(&file
->f_path
))
864 err
= deny_write_access(file
);
868 if (name
->name
[0] != '\0')
871 trace_open_exec(name
->name
);
881 struct file
*open_exec(const char *name
)
883 struct filename
*filename
= getname_kernel(name
);
884 struct file
*f
= ERR_CAST(filename
);
886 if (!IS_ERR(filename
)) {
887 f
= do_open_execat(AT_FDCWD
, filename
, 0);
892 EXPORT_SYMBOL(open_exec
);
894 int kernel_read(struct file
*file
, loff_t offset
,
895 char *addr
, unsigned long count
)
903 /* The cast to a user pointer is valid due to the set_fs() */
904 result
= vfs_read(file
, (void __user
*)addr
, count
, &pos
);
909 EXPORT_SYMBOL(kernel_read
);
911 int kernel_read_file(struct file
*file
, void **buf
, loff_t
*size
,
912 loff_t max_size
, enum kernel_read_file_id id
)
918 if (!S_ISREG(file_inode(file
)->i_mode
) || max_size
< 0)
921 ret
= security_kernel_read_file(file
, id
);
925 ret
= deny_write_access(file
);
929 i_size
= i_size_read(file_inode(file
));
930 if (max_size
> 0 && i_size
> max_size
) {
939 if (id
!= READING_FIRMWARE_PREALLOC_BUFFER
)
940 *buf
= vmalloc(i_size
);
947 while (pos
< i_size
) {
948 bytes
= kernel_read(file
, pos
, (char *)(*buf
) + pos
,
965 ret
= security_kernel_post_read_file(file
, *buf
, i_size
, id
);
971 if (id
!= READING_FIRMWARE_PREALLOC_BUFFER
) {
978 allow_write_access(file
);
981 EXPORT_SYMBOL_GPL(kernel_read_file
);
983 int kernel_read_file_from_path(char *path
, void **buf
, loff_t
*size
,
984 loff_t max_size
, enum kernel_read_file_id id
)
992 file
= filp_open(path
, O_RDONLY
, 0);
994 return PTR_ERR(file
);
996 ret
= kernel_read_file(file
, buf
, size
, max_size
, id
);
1000 EXPORT_SYMBOL_GPL(kernel_read_file_from_path
);
1002 int kernel_read_file_from_fd(int fd
, void **buf
, loff_t
*size
, loff_t max_size
,
1003 enum kernel_read_file_id id
)
1005 struct fd f
= fdget(fd
);
1011 ret
= kernel_read_file(f
.file
, buf
, size
, max_size
, id
);
1016 EXPORT_SYMBOL_GPL(kernel_read_file_from_fd
);
1018 ssize_t
read_code(struct file
*file
, unsigned long addr
, loff_t pos
, size_t len
)
1020 ssize_t res
= vfs_read(file
, (void __user
*)addr
, len
, &pos
);
1022 flush_icache_range(addr
, addr
+ len
);
1025 EXPORT_SYMBOL(read_code
);
1027 static int exec_mmap(struct mm_struct
*mm
)
1029 struct task_struct
*tsk
;
1030 struct mm_struct
*old_mm
, *active_mm
;
1032 /* Notify parent that we're no longer interested in the old VM */
1034 old_mm
= current
->mm
;
1035 mm_release(tsk
, old_mm
);
1038 sync_mm_rss(old_mm
);
1040 * Make sure that if there is a core dump in progress
1041 * for the old mm, we get out and die instead of going
1042 * through with the exec. We must hold mmap_sem around
1043 * checking core_state and changing tsk->mm.
1045 down_read(&old_mm
->mmap_sem
);
1046 if (unlikely(old_mm
->core_state
)) {
1047 up_read(&old_mm
->mmap_sem
);
1052 active_mm
= tsk
->active_mm
;
1054 tsk
->active_mm
= mm
;
1055 activate_mm(active_mm
, mm
);
1056 tsk
->mm
->vmacache_seqnum
= 0;
1057 vmacache_flush(tsk
);
1060 up_read(&old_mm
->mmap_sem
);
1061 BUG_ON(active_mm
!= old_mm
);
1062 setmax_mm_hiwater_rss(&tsk
->signal
->maxrss
, old_mm
);
1063 mm_update_next_owner(old_mm
);
1072 * This function makes sure the current process has its own signal table,
1073 * so that flush_signal_handlers can later reset the handlers without
1074 * disturbing other processes. (Other processes might share the signal
1075 * table via the CLONE_SIGHAND option to clone().)
1077 static int de_thread(struct task_struct
*tsk
)
1079 struct signal_struct
*sig
= tsk
->signal
;
1080 struct sighand_struct
*oldsighand
= tsk
->sighand
;
1081 spinlock_t
*lock
= &oldsighand
->siglock
;
1083 if (thread_group_empty(tsk
))
1084 goto no_thread_group
;
1087 * Kill all other threads in the thread group.
1089 spin_lock_irq(lock
);
1090 if (signal_group_exit(sig
)) {
1092 * Another group action in progress, just
1093 * return so that the signal is processed.
1095 spin_unlock_irq(lock
);
1099 sig
->group_exit_task
= tsk
;
1100 sig
->notify_count
= zap_other_threads(tsk
);
1101 if (!thread_group_leader(tsk
))
1102 sig
->notify_count
--;
1104 while (sig
->notify_count
) {
1105 __set_current_state(TASK_KILLABLE
);
1106 spin_unlock_irq(lock
);
1108 if (unlikely(__fatal_signal_pending(tsk
)))
1110 spin_lock_irq(lock
);
1112 spin_unlock_irq(lock
);
1115 * At this point all other threads have exited, all we have to
1116 * do is to wait for the thread group leader to become inactive,
1117 * and to assume its PID:
1119 if (!thread_group_leader(tsk
)) {
1120 struct task_struct
*leader
= tsk
->group_leader
;
1123 threadgroup_change_begin(tsk
);
1124 write_lock_irq(&tasklist_lock
);
1126 * Do this under tasklist_lock to ensure that
1127 * exit_notify() can't miss ->group_exit_task
1129 sig
->notify_count
= -1;
1130 if (likely(leader
->exit_state
))
1132 __set_current_state(TASK_KILLABLE
);
1133 write_unlock_irq(&tasklist_lock
);
1134 threadgroup_change_end(tsk
);
1136 if (unlikely(__fatal_signal_pending(tsk
)))
1141 * The only record we have of the real-time age of a
1142 * process, regardless of execs it's done, is start_time.
1143 * All the past CPU time is accumulated in signal_struct
1144 * from sister threads now dead. But in this non-leader
1145 * exec, nothing survives from the original leader thread,
1146 * whose birth marks the true age of this process now.
1147 * When we take on its identity by switching to its PID, we
1148 * also take its birthdate (always earlier than our own).
1150 tsk
->start_time
= leader
->start_time
;
1151 tsk
->real_start_time
= leader
->real_start_time
;
1153 BUG_ON(!same_thread_group(leader
, tsk
));
1154 BUG_ON(has_group_leader_pid(tsk
));
1156 * An exec() starts a new thread group with the
1157 * TGID of the previous thread group. Rehash the
1158 * two threads with a switched PID, and release
1159 * the former thread group leader:
1162 /* Become a process group leader with the old leader's pid.
1163 * The old leader becomes a thread of the this thread group.
1164 * Note: The old leader also uses this pid until release_task
1165 * is called. Odd but simple and correct.
1167 tsk
->pid
= leader
->pid
;
1168 change_pid(tsk
, PIDTYPE_PID
, task_pid(leader
));
1169 transfer_pid(leader
, tsk
, PIDTYPE_PGID
);
1170 transfer_pid(leader
, tsk
, PIDTYPE_SID
);
1172 list_replace_rcu(&leader
->tasks
, &tsk
->tasks
);
1173 list_replace_init(&leader
->sibling
, &tsk
->sibling
);
1175 tsk
->group_leader
= tsk
;
1176 leader
->group_leader
= tsk
;
1178 tsk
->exit_signal
= SIGCHLD
;
1179 leader
->exit_signal
= -1;
1181 BUG_ON(leader
->exit_state
!= EXIT_ZOMBIE
);
1182 leader
->exit_state
= EXIT_DEAD
;
1185 * We are going to release_task()->ptrace_unlink() silently,
1186 * the tracer can sleep in do_wait(). EXIT_DEAD guarantees
1187 * the tracer wont't block again waiting for this thread.
1189 if (unlikely(leader
->ptrace
))
1190 __wake_up_parent(leader
, leader
->parent
);
1191 write_unlock_irq(&tasklist_lock
);
1192 threadgroup_change_end(tsk
);
1194 release_task(leader
);
1197 sig
->group_exit_task
= NULL
;
1198 sig
->notify_count
= 0;
1201 /* we have changed execution domain */
1202 tsk
->exit_signal
= SIGCHLD
;
1204 #ifdef CONFIG_POSIX_TIMERS
1206 flush_itimer_signals();
1209 if (atomic_read(&oldsighand
->count
) != 1) {
1210 struct sighand_struct
*newsighand
;
1212 * This ->sighand is shared with the CLONE_SIGHAND
1213 * but not CLONE_THREAD task, switch to the new one.
1215 newsighand
= kmem_cache_alloc(sighand_cachep
, GFP_KERNEL
);
1219 atomic_set(&newsighand
->count
, 1);
1220 memcpy(newsighand
->action
, oldsighand
->action
,
1221 sizeof(newsighand
->action
));
1223 write_lock_irq(&tasklist_lock
);
1224 spin_lock(&oldsighand
->siglock
);
1225 rcu_assign_pointer(tsk
->sighand
, newsighand
);
1226 spin_unlock(&oldsighand
->siglock
);
1227 write_unlock_irq(&tasklist_lock
);
1229 __cleanup_sighand(oldsighand
);
1232 BUG_ON(!thread_group_leader(tsk
));
1236 /* protects against exit_notify() and __exit_signal() */
1237 read_lock(&tasklist_lock
);
1238 sig
->group_exit_task
= NULL
;
1239 sig
->notify_count
= 0;
1240 read_unlock(&tasklist_lock
);
1244 char *get_task_comm(char *buf
, struct task_struct
*tsk
)
1246 /* buf must be at least sizeof(tsk->comm) in size */
1248 strncpy(buf
, tsk
->comm
, sizeof(tsk
->comm
));
1252 EXPORT_SYMBOL_GPL(get_task_comm
);
1255 * These functions flushes out all traces of the currently running executable
1256 * so that a new one can be started
1259 void __set_task_comm(struct task_struct
*tsk
, const char *buf
, bool exec
)
1262 trace_task_rename(tsk
, buf
);
1263 strlcpy(tsk
->comm
, buf
, sizeof(tsk
->comm
));
1265 perf_event_comm(tsk
, exec
);
1268 int flush_old_exec(struct linux_binprm
* bprm
)
1273 * Make sure we have a private signal table and that
1274 * we are unassociated from the previous thread group.
1276 retval
= de_thread(current
);
1281 * Must be called _before_ exec_mmap() as bprm->mm is
1282 * not visibile until then. This also enables the update
1285 set_mm_exe_file(bprm
->mm
, bprm
->file
);
1288 * Release all of the old mmap stuff
1290 acct_arg_size(bprm
, 0);
1291 retval
= exec_mmap(bprm
->mm
);
1295 bprm
->mm
= NULL
; /* We're using it now */
1298 current
->flags
&= ~(PF_RANDOMIZE
| PF_FORKNOEXEC
| PF_KTHREAD
|
1299 PF_NOFREEZE
| PF_NO_SETAFFINITY
);
1301 current
->personality
&= ~bprm
->per_clear
;
1304 * We have to apply CLOEXEC before we change whether the process is
1305 * dumpable (in setup_new_exec) to avoid a race with a process in userspace
1306 * trying to access the should-be-closed file descriptors of a process
1307 * undergoing exec(2).
1309 do_close_on_exec(current
->files
);
1315 EXPORT_SYMBOL(flush_old_exec
);
1317 void would_dump(struct linux_binprm
*bprm
, struct file
*file
)
1319 struct inode
*inode
= file_inode(file
);
1320 if (inode_permission(inode
, MAY_READ
) < 0) {
1321 struct user_namespace
*old
, *user_ns
;
1322 bprm
->interp_flags
|= BINPRM_FLAGS_ENFORCE_NONDUMP
;
1324 /* Ensure mm->user_ns contains the executable */
1325 user_ns
= old
= bprm
->mm
->user_ns
;
1326 while ((user_ns
!= &init_user_ns
) &&
1327 !privileged_wrt_inode_uidgid(user_ns
, inode
))
1328 user_ns
= user_ns
->parent
;
1330 if (old
!= user_ns
) {
1331 bprm
->mm
->user_ns
= get_user_ns(user_ns
);
1336 EXPORT_SYMBOL(would_dump
);
1338 void setup_new_exec(struct linux_binprm
* bprm
)
1340 arch_pick_mmap_layout(current
->mm
);
1342 /* This is the point of no return */
1343 current
->sas_ss_sp
= current
->sas_ss_size
= 0;
1345 if (uid_eq(current_euid(), current_uid()) && gid_eq(current_egid(), current_gid()))
1346 set_dumpable(current
->mm
, SUID_DUMP_USER
);
1348 set_dumpable(current
->mm
, suid_dumpable
);
1351 __set_task_comm(current
, kbasename(bprm
->filename
), true);
1353 /* Set the new mm task size. We have to do that late because it may
1354 * depend on TIF_32BIT which is only updated in flush_thread() on
1355 * some architectures like powerpc
1357 current
->mm
->task_size
= TASK_SIZE
;
1359 /* install the new credentials */
1360 if (!uid_eq(bprm
->cred
->uid
, current_euid()) ||
1361 !gid_eq(bprm
->cred
->gid
, current_egid())) {
1362 current
->pdeath_signal
= 0;
1364 if (bprm
->interp_flags
& BINPRM_FLAGS_ENFORCE_NONDUMP
)
1365 set_dumpable(current
->mm
, suid_dumpable
);
1368 /* An exec changes our domain. We are no longer part of the thread
1370 current
->self_exec_id
++;
1371 flush_signal_handlers(current
, 0);
1373 EXPORT_SYMBOL(setup_new_exec
);
1376 * Prepare credentials and lock ->cred_guard_mutex.
1377 * install_exec_creds() commits the new creds and drops the lock.
1378 * Or, if exec fails before, free_bprm() should release ->cred and
1381 int prepare_bprm_creds(struct linux_binprm
*bprm
)
1383 if (mutex_lock_interruptible(¤t
->signal
->cred_guard_mutex
))
1384 return -ERESTARTNOINTR
;
1386 bprm
->cred
= prepare_exec_creds();
1387 if (likely(bprm
->cred
))
1390 mutex_unlock(¤t
->signal
->cred_guard_mutex
);
1394 static void free_bprm(struct linux_binprm
*bprm
)
1396 free_arg_pages(bprm
);
1398 mutex_unlock(¤t
->signal
->cred_guard_mutex
);
1399 abort_creds(bprm
->cred
);
1402 allow_write_access(bprm
->file
);
1405 /* If a binfmt changed the interp, free it. */
1406 if (bprm
->interp
!= bprm
->filename
)
1407 kfree(bprm
->interp
);
1411 int bprm_change_interp(char *interp
, struct linux_binprm
*bprm
)
1413 /* If a binfmt changed the interp, free it first. */
1414 if (bprm
->interp
!= bprm
->filename
)
1415 kfree(bprm
->interp
);
1416 bprm
->interp
= kstrdup(interp
, GFP_KERNEL
);
1421 EXPORT_SYMBOL(bprm_change_interp
);
1424 * install the new credentials for this executable
1426 void install_exec_creds(struct linux_binprm
*bprm
)
1428 security_bprm_committing_creds(bprm
);
1430 commit_creds(bprm
->cred
);
1434 * Disable monitoring for regular users
1435 * when executing setuid binaries. Must
1436 * wait until new credentials are committed
1437 * by commit_creds() above
1439 if (get_dumpable(current
->mm
) != SUID_DUMP_USER
)
1440 perf_event_exit_task(current
);
1442 * cred_guard_mutex must be held at least to this point to prevent
1443 * ptrace_attach() from altering our determination of the task's
1444 * credentials; any time after this it may be unlocked.
1446 security_bprm_committed_creds(bprm
);
1447 mutex_unlock(¤t
->signal
->cred_guard_mutex
);
1449 EXPORT_SYMBOL(install_exec_creds
);
1452 * determine how safe it is to execute the proposed program
1453 * - the caller must hold ->cred_guard_mutex to protect against
1454 * PTRACE_ATTACH or seccomp thread-sync
1456 static void check_unsafe_exec(struct linux_binprm
*bprm
)
1458 struct task_struct
*p
= current
, *t
;
1463 if (ptracer_capable(p
, current_user_ns()))
1464 bprm
->unsafe
|= LSM_UNSAFE_PTRACE_CAP
;
1466 bprm
->unsafe
|= LSM_UNSAFE_PTRACE
;
1470 * This isn't strictly necessary, but it makes it harder for LSMs to
1473 if (task_no_new_privs(current
))
1474 bprm
->unsafe
|= LSM_UNSAFE_NO_NEW_PRIVS
;
1480 spin_lock(&p
->fs
->lock
);
1482 while_each_thread(p
, t
) {
1485 if (t
->flags
& (PF_EXITING
| PF_FORKNOEXEC
))
1490 if (p
->fs
->users
> n_fs
) {
1492 spin_unlock(&p
->fs
->lock
);
1495 bprm
->unsafe
|= LSM_UNSAFE_SHARE
;
1498 spin_unlock(&p
->fs
->lock
);
1501 static void bprm_fill_uid(struct linux_binprm
*bprm
)
1503 struct inode
*inode
;
1509 * Since this can be called multiple times (via prepare_binprm),
1510 * we must clear any previous work done when setting set[ug]id
1511 * bits from any earlier bprm->file uses (for example when run
1512 * first for a setuid script then again for its interpreter).
1514 bprm
->cred
->euid
= current_euid();
1515 bprm
->cred
->egid
= current_egid();
1517 if (path_nosuid(&bprm
->file
->f_path
))
1520 if (task_no_new_privs(current
))
1523 inode
= file_inode(bprm
->file
);
1524 mode
= READ_ONCE(inode
->i_mode
);
1525 if (!(mode
& (S_ISUID
|S_ISGID
)))
1528 /* Be careful if suid/sgid is set */
1531 /* reload atomically mode/uid/gid now that lock held */
1532 mode
= inode
->i_mode
;
1535 inode_unlock(inode
);
1537 /* We ignore suid/sgid if there are no mappings for them in the ns */
1538 if (!kuid_has_mapping(bprm
->cred
->user_ns
, uid
) ||
1539 !kgid_has_mapping(bprm
->cred
->user_ns
, gid
))
1542 if (mode
& S_ISUID
) {
1543 bprm
->per_clear
|= PER_CLEAR_ON_SETID
;
1544 bprm
->cred
->euid
= uid
;
1547 if ((mode
& (S_ISGID
| S_IXGRP
)) == (S_ISGID
| S_IXGRP
)) {
1548 bprm
->per_clear
|= PER_CLEAR_ON_SETID
;
1549 bprm
->cred
->egid
= gid
;
1554 * Fill the binprm structure from the inode.
1555 * Check permissions, then read the first 128 (BINPRM_BUF_SIZE) bytes
1557 * This may be called multiple times for binary chains (scripts for example).
1559 int prepare_binprm(struct linux_binprm
*bprm
)
1563 bprm_fill_uid(bprm
);
1565 /* fill in binprm security blob */
1566 retval
= security_bprm_set_creds(bprm
);
1569 bprm
->cred_prepared
= 1;
1571 memset(bprm
->buf
, 0, BINPRM_BUF_SIZE
);
1572 return kernel_read(bprm
->file
, 0, bprm
->buf
, BINPRM_BUF_SIZE
);
1575 EXPORT_SYMBOL(prepare_binprm
);
1578 * Arguments are '\0' separated strings found at the location bprm->p
1579 * points to; chop off the first by relocating brpm->p to right after
1580 * the first '\0' encountered.
1582 int remove_arg_zero(struct linux_binprm
*bprm
)
1585 unsigned long offset
;
1593 offset
= bprm
->p
& ~PAGE_MASK
;
1594 page
= get_arg_page(bprm
, bprm
->p
, 0);
1599 kaddr
= kmap_atomic(page
);
1601 for (; offset
< PAGE_SIZE
&& kaddr
[offset
];
1602 offset
++, bprm
->p
++)
1605 kunmap_atomic(kaddr
);
1607 } while (offset
== PAGE_SIZE
);
1616 EXPORT_SYMBOL(remove_arg_zero
);
1618 #define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e))
1620 * cycle the list of binary formats handler, until one recognizes the image
1622 int search_binary_handler(struct linux_binprm
*bprm
)
1624 bool need_retry
= IS_ENABLED(CONFIG_MODULES
);
1625 struct linux_binfmt
*fmt
;
1628 /* This allows 4 levels of binfmt rewrites before failing hard. */
1629 if (bprm
->recursion_depth
> 5)
1632 retval
= security_bprm_check(bprm
);
1638 read_lock(&binfmt_lock
);
1639 list_for_each_entry(fmt
, &formats
, lh
) {
1640 if (!try_module_get(fmt
->module
))
1642 read_unlock(&binfmt_lock
);
1643 bprm
->recursion_depth
++;
1644 retval
= fmt
->load_binary(bprm
);
1645 read_lock(&binfmt_lock
);
1647 bprm
->recursion_depth
--;
1648 if (retval
< 0 && !bprm
->mm
) {
1649 /* we got to flush_old_exec() and failed after it */
1650 read_unlock(&binfmt_lock
);
1651 force_sigsegv(SIGSEGV
, current
);
1654 if (retval
!= -ENOEXEC
|| !bprm
->file
) {
1655 read_unlock(&binfmt_lock
);
1659 read_unlock(&binfmt_lock
);
1662 if (printable(bprm
->buf
[0]) && printable(bprm
->buf
[1]) &&
1663 printable(bprm
->buf
[2]) && printable(bprm
->buf
[3]))
1665 if (request_module("binfmt-%04x", *(ushort
*)(bprm
->buf
+ 2)) < 0)
1673 EXPORT_SYMBOL(search_binary_handler
);
1675 static int exec_binprm(struct linux_binprm
*bprm
)
1677 pid_t old_pid
, old_vpid
;
1680 /* Need to fetch pid before load_binary changes it */
1681 old_pid
= current
->pid
;
1683 old_vpid
= task_pid_nr_ns(current
, task_active_pid_ns(current
->parent
));
1686 ret
= search_binary_handler(bprm
);
1689 trace_sched_process_exec(current
, old_pid
, bprm
);
1690 ptrace_event(PTRACE_EVENT_EXEC
, old_vpid
);
1691 proc_exec_connector(current
);
1698 * sys_execve() executes a new program.
1700 static int do_execveat_common(int fd
, struct filename
*filename
,
1701 struct user_arg_ptr argv
,
1702 struct user_arg_ptr envp
,
1705 char *pathbuf
= NULL
;
1706 struct linux_binprm
*bprm
;
1708 struct files_struct
*displaced
;
1711 if (IS_ERR(filename
))
1712 return PTR_ERR(filename
);
1715 * We move the actual failure in case of RLIMIT_NPROC excess from
1716 * set*uid() to execve() because too many poorly written programs
1717 * don't check setuid() return code. Here we additionally recheck
1718 * whether NPROC limit is still exceeded.
1720 if ((current
->flags
& PF_NPROC_EXCEEDED
) &&
1721 atomic_read(¤t_user()->processes
) > rlimit(RLIMIT_NPROC
)) {
1726 /* We're below the limit (still or again), so we don't want to make
1727 * further execve() calls fail. */
1728 current
->flags
&= ~PF_NPROC_EXCEEDED
;
1730 retval
= unshare_files(&displaced
);
1735 bprm
= kzalloc(sizeof(*bprm
), GFP_KERNEL
);
1739 retval
= prepare_bprm_creds(bprm
);
1743 check_unsafe_exec(bprm
);
1744 current
->in_execve
= 1;
1746 file
= do_open_execat(fd
, filename
, flags
);
1747 retval
= PTR_ERR(file
);
1754 if (fd
== AT_FDCWD
|| filename
->name
[0] == '/') {
1755 bprm
->filename
= filename
->name
;
1757 if (filename
->name
[0] == '\0')
1758 pathbuf
= kasprintf(GFP_TEMPORARY
, "/dev/fd/%d", fd
);
1760 pathbuf
= kasprintf(GFP_TEMPORARY
, "/dev/fd/%d/%s",
1761 fd
, filename
->name
);
1767 * Record that a name derived from an O_CLOEXEC fd will be
1768 * inaccessible after exec. Relies on having exclusive access to
1769 * current->files (due to unshare_files above).
1771 if (close_on_exec(fd
, rcu_dereference_raw(current
->files
->fdt
)))
1772 bprm
->interp_flags
|= BINPRM_FLAGS_PATH_INACCESSIBLE
;
1773 bprm
->filename
= pathbuf
;
1775 bprm
->interp
= bprm
->filename
;
1777 retval
= bprm_mm_init(bprm
);
1781 bprm
->argc
= count(argv
, MAX_ARG_STRINGS
);
1782 if ((retval
= bprm
->argc
) < 0)
1785 bprm
->envc
= count(envp
, MAX_ARG_STRINGS
);
1786 if ((retval
= bprm
->envc
) < 0)
1789 retval
= prepare_binprm(bprm
);
1793 retval
= copy_strings_kernel(1, &bprm
->filename
, bprm
);
1797 bprm
->exec
= bprm
->p
;
1798 retval
= copy_strings(bprm
->envc
, envp
, bprm
);
1802 retval
= copy_strings(bprm
->argc
, argv
, bprm
);
1806 would_dump(bprm
, bprm
->file
);
1808 retval
= exec_binprm(bprm
);
1812 /* execve succeeded */
1813 current
->fs
->in_exec
= 0;
1814 current
->in_execve
= 0;
1815 acct_update_integrals(current
);
1816 task_numa_free(current
);
1821 put_files_struct(displaced
);
1826 acct_arg_size(bprm
, 0);
1831 current
->fs
->in_exec
= 0;
1832 current
->in_execve
= 0;
1840 reset_files_struct(displaced
);
1846 int do_execve(struct filename
*filename
,
1847 const char __user
*const __user
*__argv
,
1848 const char __user
*const __user
*__envp
)
1850 struct user_arg_ptr argv
= { .ptr
.native
= __argv
};
1851 struct user_arg_ptr envp
= { .ptr
.native
= __envp
};
1852 return do_execveat_common(AT_FDCWD
, filename
, argv
, envp
, 0);
1855 int do_execveat(int fd
, struct filename
*filename
,
1856 const char __user
*const __user
*__argv
,
1857 const char __user
*const __user
*__envp
,
1860 struct user_arg_ptr argv
= { .ptr
.native
= __argv
};
1861 struct user_arg_ptr envp
= { .ptr
.native
= __envp
};
1863 return do_execveat_common(fd
, filename
, argv
, envp
, flags
);
1866 #ifdef CONFIG_COMPAT
1867 static int compat_do_execve(struct filename
*filename
,
1868 const compat_uptr_t __user
*__argv
,
1869 const compat_uptr_t __user
*__envp
)
1871 struct user_arg_ptr argv
= {
1873 .ptr
.compat
= __argv
,
1875 struct user_arg_ptr envp
= {
1877 .ptr
.compat
= __envp
,
1879 return do_execveat_common(AT_FDCWD
, filename
, argv
, envp
, 0);
1882 static int compat_do_execveat(int fd
, struct filename
*filename
,
1883 const compat_uptr_t __user
*__argv
,
1884 const compat_uptr_t __user
*__envp
,
1887 struct user_arg_ptr argv
= {
1889 .ptr
.compat
= __argv
,
1891 struct user_arg_ptr envp
= {
1893 .ptr
.compat
= __envp
,
1895 return do_execveat_common(fd
, filename
, argv
, envp
, flags
);
1899 void set_binfmt(struct linux_binfmt
*new)
1901 struct mm_struct
*mm
= current
->mm
;
1904 module_put(mm
->binfmt
->module
);
1908 __module_get(new->module
);
1910 EXPORT_SYMBOL(set_binfmt
);
1913 * set_dumpable stores three-value SUID_DUMP_* into mm->flags.
1915 void set_dumpable(struct mm_struct
*mm
, int value
)
1917 unsigned long old
, new;
1919 if (WARN_ON((unsigned)value
> SUID_DUMP_ROOT
))
1923 old
= ACCESS_ONCE(mm
->flags
);
1924 new = (old
& ~MMF_DUMPABLE_MASK
) | value
;
1925 } while (cmpxchg(&mm
->flags
, old
, new) != old
);
1928 SYSCALL_DEFINE3(execve
,
1929 const char __user
*, filename
,
1930 const char __user
*const __user
*, argv
,
1931 const char __user
*const __user
*, envp
)
1933 return do_execve(getname(filename
), argv
, envp
);
1936 SYSCALL_DEFINE5(execveat
,
1937 int, fd
, const char __user
*, filename
,
1938 const char __user
*const __user
*, argv
,
1939 const char __user
*const __user
*, envp
,
1942 int lookup_flags
= (flags
& AT_EMPTY_PATH
) ? LOOKUP_EMPTY
: 0;
1944 return do_execveat(fd
,
1945 getname_flags(filename
, lookup_flags
, NULL
),
1949 #ifdef CONFIG_COMPAT
1950 COMPAT_SYSCALL_DEFINE3(execve
, const char __user
*, filename
,
1951 const compat_uptr_t __user
*, argv
,
1952 const compat_uptr_t __user
*, envp
)
1954 return compat_do_execve(getname(filename
), argv
, envp
);
1957 COMPAT_SYSCALL_DEFINE5(execveat
, int, fd
,
1958 const char __user
*, filename
,
1959 const compat_uptr_t __user
*, argv
,
1960 const compat_uptr_t __user
*, envp
,
1963 int lookup_flags
= (flags
& AT_EMPTY_PATH
) ? LOOKUP_EMPTY
: 0;
1965 return compat_do_execveat(fd
,
1966 getname_flags(filename
, lookup_flags
, NULL
),