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/sched/mm.h>
36 #include <linux/sched/coredump.h>
37 #include <linux/sched/signal.h>
38 #include <linux/sched/numa_balancing.h>
39 #include <linux/sched/task.h>
40 #include <linux/pagemap.h>
41 #include <linux/perf_event.h>
42 #include <linux/highmem.h>
43 #include <linux/spinlock.h>
44 #include <linux/key.h>
45 #include <linux/personality.h>
46 #include <linux/binfmts.h>
47 #include <linux/utsname.h>
48 #include <linux/pid_namespace.h>
49 #include <linux/module.h>
50 #include <linux/namei.h>
51 #include <linux/mount.h>
52 #include <linux/security.h>
53 #include <linux/syscalls.h>
54 #include <linux/tsacct_kern.h>
55 #include <linux/cn_proc.h>
56 #include <linux/audit.h>
57 #include <linux/tracehook.h>
58 #include <linux/kmod.h>
59 #include <linux/fsnotify.h>
60 #include <linux/fs_struct.h>
61 #include <linux/pipe_fs_i.h>
62 #include <linux/oom.h>
63 #include <linux/compat.h>
64 #include <linux/vmalloc.h>
66 #include <trace/events/fs.h>
68 #include <linux/uaccess.h>
69 #include <asm/mmu_context.h>
72 #include <trace/events/task.h>
75 #include <trace/events/sched.h>
77 int suid_dumpable
= 0;
79 static LIST_HEAD(formats
);
80 static DEFINE_RWLOCK(binfmt_lock
);
82 void __register_binfmt(struct linux_binfmt
* fmt
, int insert
)
85 if (WARN_ON(!fmt
->load_binary
))
87 write_lock(&binfmt_lock
);
88 insert
? list_add(&fmt
->lh
, &formats
) :
89 list_add_tail(&fmt
->lh
, &formats
);
90 write_unlock(&binfmt_lock
);
93 EXPORT_SYMBOL(__register_binfmt
);
95 void unregister_binfmt(struct linux_binfmt
* fmt
)
97 write_lock(&binfmt_lock
);
99 write_unlock(&binfmt_lock
);
102 EXPORT_SYMBOL(unregister_binfmt
);
104 static inline void put_binfmt(struct linux_binfmt
* fmt
)
106 module_put(fmt
->module
);
109 bool path_noexec(const struct path
*path
)
111 return (path
->mnt
->mnt_flags
& MNT_NOEXEC
) ||
112 (path
->mnt
->mnt_sb
->s_iflags
& SB_I_NOEXEC
);
117 * Note that a shared library must be both readable and executable due to
120 * Also note that we take the address to load from from the file itself.
122 SYSCALL_DEFINE1(uselib
, const char __user
*, library
)
124 struct linux_binfmt
*fmt
;
126 struct filename
*tmp
= getname(library
);
127 int error
= PTR_ERR(tmp
);
128 static const struct open_flags uselib_flags
= {
129 .open_flag
= O_LARGEFILE
| O_RDONLY
| __FMODE_EXEC
,
130 .acc_mode
= MAY_READ
| MAY_EXEC
,
131 .intent
= LOOKUP_OPEN
,
132 .lookup_flags
= LOOKUP_FOLLOW
,
138 file
= do_filp_open(AT_FDCWD
, tmp
, &uselib_flags
);
140 error
= PTR_ERR(file
);
145 if (!S_ISREG(file_inode(file
)->i_mode
))
149 if (path_noexec(&file
->f_path
))
156 read_lock(&binfmt_lock
);
157 list_for_each_entry(fmt
, &formats
, lh
) {
158 if (!fmt
->load_shlib
)
160 if (!try_module_get(fmt
->module
))
162 read_unlock(&binfmt_lock
);
163 error
= fmt
->load_shlib(file
);
164 read_lock(&binfmt_lock
);
166 if (error
!= -ENOEXEC
)
169 read_unlock(&binfmt_lock
);
175 #endif /* #ifdef CONFIG_USELIB */
179 * The nascent bprm->mm is not visible until exec_mmap() but it can
180 * use a lot of memory, account these pages in current->mm temporary
181 * for oom_badness()->get_mm_rss(). Once exec succeeds or fails, we
182 * change the counter back via acct_arg_size(0).
184 static void acct_arg_size(struct linux_binprm
*bprm
, unsigned long pages
)
186 struct mm_struct
*mm
= current
->mm
;
187 long diff
= (long)(pages
- bprm
->vma_pages
);
192 bprm
->vma_pages
= pages
;
193 add_mm_counter(mm
, MM_ANONPAGES
, diff
);
196 static struct page
*get_arg_page(struct linux_binprm
*bprm
, unsigned long pos
,
201 unsigned int gup_flags
= FOLL_FORCE
;
203 #ifdef CONFIG_STACK_GROWSUP
205 ret
= expand_downwards(bprm
->vma
, pos
);
212 gup_flags
|= FOLL_WRITE
;
215 * We are doing an exec(). 'current' is the process
216 * doing the exec and bprm->mm is the new process's mm.
218 ret
= get_user_pages_remote(current
, bprm
->mm
, pos
, 1, gup_flags
,
224 unsigned long size
= bprm
->vma
->vm_end
- bprm
->vma
->vm_start
;
225 unsigned long ptr_size
, limit
;
228 * Since the stack will hold pointers to the strings, we
229 * must account for them as well.
231 * The size calculation is the entire vma while each arg page is
232 * built, so each time we get here it's calculating how far it
233 * is currently (rather than each call being just the newly
234 * added size from the arg page). As a result, we need to
235 * always add the entire size of the pointers, so that on the
236 * last call to get_arg_page() we'll actually have the entire
239 ptr_size
= (bprm
->argc
+ bprm
->envc
) * sizeof(void *);
240 if (ptr_size
> ULONG_MAX
- size
)
244 acct_arg_size(bprm
, size
/ PAGE_SIZE
);
247 * We've historically supported up to 32 pages (ARG_MAX)
248 * of argument strings even with small stacks
254 * Limit to 1/4 of the max stack size or 3/4 of _STK_LIM
255 * (whichever is smaller) for the argv+env strings.
257 * - the remaining binfmt code will not run out of stack space,
258 * - the program will have a reasonable amount of stack left
261 limit
= _STK_LIM
/ 4 * 3;
262 limit
= min(limit
, rlimit(RLIMIT_STACK
) / 4);
274 static void put_arg_page(struct page
*page
)
279 static void free_arg_pages(struct linux_binprm
*bprm
)
283 static void flush_arg_page(struct linux_binprm
*bprm
, unsigned long pos
,
286 flush_cache_page(bprm
->vma
, pos
, page_to_pfn(page
));
289 static int __bprm_mm_init(struct linux_binprm
*bprm
)
292 struct vm_area_struct
*vma
= NULL
;
293 struct mm_struct
*mm
= bprm
->mm
;
295 bprm
->vma
= vma
= kmem_cache_zalloc(vm_area_cachep
, GFP_KERNEL
);
299 if (down_write_killable(&mm
->mmap_sem
)) {
306 * Place the stack at the largest stack address the architecture
307 * supports. Later, we'll move this to an appropriate place. We don't
308 * use STACK_TOP because that can depend on attributes which aren't
311 BUILD_BUG_ON(VM_STACK_FLAGS
& VM_STACK_INCOMPLETE_SETUP
);
312 vma
->vm_end
= STACK_TOP_MAX
;
313 vma
->vm_start
= vma
->vm_end
- PAGE_SIZE
;
314 vma
->vm_flags
= VM_SOFTDIRTY
| VM_STACK_FLAGS
| VM_STACK_INCOMPLETE_SETUP
;
315 vma
->vm_page_prot
= vm_get_page_prot(vma
->vm_flags
);
316 INIT_LIST_HEAD(&vma
->anon_vma_chain
);
318 err
= insert_vm_struct(mm
, vma
);
322 mm
->stack_vm
= mm
->total_vm
= 1;
323 arch_bprm_mm_init(mm
, vma
);
324 up_write(&mm
->mmap_sem
);
325 bprm
->p
= vma
->vm_end
- sizeof(void *);
328 up_write(&mm
->mmap_sem
);
331 kmem_cache_free(vm_area_cachep
, vma
);
335 static bool valid_arg_len(struct linux_binprm
*bprm
, long len
)
337 return len
<= MAX_ARG_STRLEN
;
342 static inline void acct_arg_size(struct linux_binprm
*bprm
, unsigned long pages
)
346 static struct page
*get_arg_page(struct linux_binprm
*bprm
, unsigned long pos
,
351 page
= bprm
->page
[pos
/ PAGE_SIZE
];
352 if (!page
&& write
) {
353 page
= alloc_page(GFP_HIGHUSER
|__GFP_ZERO
);
356 bprm
->page
[pos
/ PAGE_SIZE
] = page
;
362 static void put_arg_page(struct page
*page
)
366 static void free_arg_page(struct linux_binprm
*bprm
, int i
)
369 __free_page(bprm
->page
[i
]);
370 bprm
->page
[i
] = NULL
;
374 static void free_arg_pages(struct linux_binprm
*bprm
)
378 for (i
= 0; i
< MAX_ARG_PAGES
; i
++)
379 free_arg_page(bprm
, i
);
382 static void flush_arg_page(struct linux_binprm
*bprm
, unsigned long pos
,
387 static int __bprm_mm_init(struct linux_binprm
*bprm
)
389 bprm
->p
= PAGE_SIZE
* MAX_ARG_PAGES
- sizeof(void *);
393 static bool valid_arg_len(struct linux_binprm
*bprm
, long len
)
395 return len
<= bprm
->p
;
398 #endif /* CONFIG_MMU */
401 * Create a new mm_struct and populate it with a temporary stack
402 * vm_area_struct. We don't have enough context at this point to set the stack
403 * flags, permissions, and offset, so we use temporary values. We'll update
404 * them later in setup_arg_pages().
406 static int bprm_mm_init(struct linux_binprm
*bprm
)
409 struct mm_struct
*mm
= NULL
;
411 bprm
->mm
= mm
= mm_alloc();
416 err
= __bprm_mm_init(bprm
);
431 struct user_arg_ptr
{
436 const char __user
*const __user
*native
;
438 const compat_uptr_t __user
*compat
;
443 static const char __user
*get_user_arg_ptr(struct user_arg_ptr argv
, int nr
)
445 const char __user
*native
;
448 if (unlikely(argv
.is_compat
)) {
449 compat_uptr_t compat
;
451 if (get_user(compat
, argv
.ptr
.compat
+ nr
))
452 return ERR_PTR(-EFAULT
);
454 return compat_ptr(compat
);
458 if (get_user(native
, argv
.ptr
.native
+ nr
))
459 return ERR_PTR(-EFAULT
);
465 * count() counts the number of strings in array ARGV.
467 static int count(struct user_arg_ptr argv
, int max
)
471 if (argv
.ptr
.native
!= NULL
) {
473 const char __user
*p
= get_user_arg_ptr(argv
, i
);
485 if (fatal_signal_pending(current
))
486 return -ERESTARTNOHAND
;
494 * 'copy_strings()' copies argument/environment strings from the old
495 * processes's memory to the new process's stack. The call to get_user_pages()
496 * ensures the destination page is created and not swapped out.
498 static int copy_strings(int argc
, struct user_arg_ptr argv
,
499 struct linux_binprm
*bprm
)
501 struct page
*kmapped_page
= NULL
;
503 unsigned long kpos
= 0;
507 const char __user
*str
;
512 str
= get_user_arg_ptr(argv
, argc
);
516 len
= strnlen_user(str
, MAX_ARG_STRLEN
);
521 if (!valid_arg_len(bprm
, len
))
524 /* We're going to work our way backwords. */
530 int offset
, bytes_to_copy
;
532 if (fatal_signal_pending(current
)) {
533 ret
= -ERESTARTNOHAND
;
538 offset
= pos
% PAGE_SIZE
;
542 bytes_to_copy
= offset
;
543 if (bytes_to_copy
> len
)
546 offset
-= bytes_to_copy
;
547 pos
-= bytes_to_copy
;
548 str
-= bytes_to_copy
;
549 len
-= bytes_to_copy
;
551 if (!kmapped_page
|| kpos
!= (pos
& PAGE_MASK
)) {
554 page
= get_arg_page(bprm
, pos
, 1);
561 flush_kernel_dcache_page(kmapped_page
);
562 kunmap(kmapped_page
);
563 put_arg_page(kmapped_page
);
566 kaddr
= kmap(kmapped_page
);
567 kpos
= pos
& PAGE_MASK
;
568 flush_arg_page(bprm
, kpos
, kmapped_page
);
570 if (copy_from_user(kaddr
+offset
, str
, bytes_to_copy
)) {
579 flush_kernel_dcache_page(kmapped_page
);
580 kunmap(kmapped_page
);
581 put_arg_page(kmapped_page
);
587 * Like copy_strings, but get argv and its values from kernel memory.
589 int copy_strings_kernel(int argc
, const char *const *__argv
,
590 struct linux_binprm
*bprm
)
593 mm_segment_t oldfs
= get_fs();
594 struct user_arg_ptr argv
= {
595 .ptr
.native
= (const char __user
*const __user
*)__argv
,
599 r
= copy_strings(argc
, argv
, bprm
);
604 EXPORT_SYMBOL(copy_strings_kernel
);
609 * During bprm_mm_init(), we create a temporary stack at STACK_TOP_MAX. Once
610 * the binfmt code determines where the new stack should reside, we shift it to
611 * its final location. The process proceeds as follows:
613 * 1) Use shift to calculate the new vma endpoints.
614 * 2) Extend vma to cover both the old and new ranges. This ensures the
615 * arguments passed to subsequent functions are consistent.
616 * 3) Move vma's page tables to the new range.
617 * 4) Free up any cleared pgd range.
618 * 5) Shrink the vma to cover only the new range.
620 static int shift_arg_pages(struct vm_area_struct
*vma
, unsigned long shift
)
622 struct mm_struct
*mm
= vma
->vm_mm
;
623 unsigned long old_start
= vma
->vm_start
;
624 unsigned long old_end
= vma
->vm_end
;
625 unsigned long length
= old_end
- old_start
;
626 unsigned long new_start
= old_start
- shift
;
627 unsigned long new_end
= old_end
- shift
;
628 struct mmu_gather tlb
;
630 BUG_ON(new_start
> new_end
);
633 * ensure there are no vmas between where we want to go
636 if (vma
!= find_vma(mm
, new_start
))
640 * cover the whole range: [new_start, old_end)
642 if (vma_adjust(vma
, new_start
, old_end
, vma
->vm_pgoff
, NULL
))
646 * move the page tables downwards, on failure we rely on
647 * process cleanup to remove whatever mess we made.
649 if (length
!= move_page_tables(vma
, old_start
,
650 vma
, new_start
, length
, false))
654 tlb_gather_mmu(&tlb
, mm
, old_start
, old_end
);
655 if (new_end
> old_start
) {
657 * when the old and new regions overlap clear from new_end.
659 free_pgd_range(&tlb
, new_end
, old_end
, new_end
,
660 vma
->vm_next
? vma
->vm_next
->vm_start
: USER_PGTABLES_CEILING
);
663 * otherwise, clean from old_start; this is done to not touch
664 * the address space in [new_end, old_start) some architectures
665 * have constraints on va-space that make this illegal (IA64) -
666 * for the others its just a little faster.
668 free_pgd_range(&tlb
, old_start
, old_end
, new_end
,
669 vma
->vm_next
? vma
->vm_next
->vm_start
: USER_PGTABLES_CEILING
);
671 tlb_finish_mmu(&tlb
, old_start
, old_end
);
674 * Shrink the vma to just the new range. Always succeeds.
676 vma_adjust(vma
, new_start
, new_end
, vma
->vm_pgoff
, NULL
);
682 * Finalizes the stack vm_area_struct. The flags and permissions are updated,
683 * the stack is optionally relocated, and some extra space is added.
685 int setup_arg_pages(struct linux_binprm
*bprm
,
686 unsigned long stack_top
,
687 int executable_stack
)
690 unsigned long stack_shift
;
691 struct mm_struct
*mm
= current
->mm
;
692 struct vm_area_struct
*vma
= bprm
->vma
;
693 struct vm_area_struct
*prev
= NULL
;
694 unsigned long vm_flags
;
695 unsigned long stack_base
;
696 unsigned long stack_size
;
697 unsigned long stack_expand
;
698 unsigned long rlim_stack
;
700 #ifdef CONFIG_STACK_GROWSUP
701 /* Limit stack size */
702 stack_base
= rlimit_max(RLIMIT_STACK
);
703 if (stack_base
> STACK_SIZE_MAX
)
704 stack_base
= STACK_SIZE_MAX
;
706 /* Add space for stack randomization. */
707 stack_base
+= (STACK_RND_MASK
<< PAGE_SHIFT
);
709 /* Make sure we didn't let the argument array grow too large. */
710 if (vma
->vm_end
- vma
->vm_start
> stack_base
)
713 stack_base
= PAGE_ALIGN(stack_top
- stack_base
);
715 stack_shift
= vma
->vm_start
- stack_base
;
716 mm
->arg_start
= bprm
->p
- stack_shift
;
717 bprm
->p
= vma
->vm_end
- stack_shift
;
719 stack_top
= arch_align_stack(stack_top
);
720 stack_top
= PAGE_ALIGN(stack_top
);
722 if (unlikely(stack_top
< mmap_min_addr
) ||
723 unlikely(vma
->vm_end
- vma
->vm_start
>= stack_top
- mmap_min_addr
))
726 stack_shift
= vma
->vm_end
- stack_top
;
728 bprm
->p
-= stack_shift
;
729 mm
->arg_start
= bprm
->p
;
733 bprm
->loader
-= stack_shift
;
734 bprm
->exec
-= stack_shift
;
736 if (down_write_killable(&mm
->mmap_sem
))
739 vm_flags
= VM_STACK_FLAGS
;
742 * Adjust stack execute permissions; explicitly enable for
743 * EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X and leave alone
744 * (arch default) otherwise.
746 if (unlikely(executable_stack
== EXSTACK_ENABLE_X
))
748 else if (executable_stack
== EXSTACK_DISABLE_X
)
749 vm_flags
&= ~VM_EXEC
;
750 vm_flags
|= mm
->def_flags
;
751 vm_flags
|= VM_STACK_INCOMPLETE_SETUP
;
753 ret
= mprotect_fixup(vma
, &prev
, vma
->vm_start
, vma
->vm_end
,
759 /* Move stack pages down in memory. */
761 ret
= shift_arg_pages(vma
, stack_shift
);
766 /* mprotect_fixup is overkill to remove the temporary stack flags */
767 vma
->vm_flags
&= ~VM_STACK_INCOMPLETE_SETUP
;
769 stack_expand
= 131072UL; /* randomly 32*4k (or 2*64k) pages */
770 stack_size
= vma
->vm_end
- vma
->vm_start
;
772 * Align this down to a page boundary as expand_stack
775 rlim_stack
= rlimit(RLIMIT_STACK
) & PAGE_MASK
;
776 #ifdef CONFIG_STACK_GROWSUP
777 if (stack_size
+ stack_expand
> rlim_stack
)
778 stack_base
= vma
->vm_start
+ rlim_stack
;
780 stack_base
= vma
->vm_end
+ stack_expand
;
782 if (stack_size
+ stack_expand
> rlim_stack
)
783 stack_base
= vma
->vm_end
- rlim_stack
;
785 stack_base
= vma
->vm_start
- stack_expand
;
787 current
->mm
->start_stack
= bprm
->p
;
788 ret
= expand_stack(vma
, stack_base
);
793 up_write(&mm
->mmap_sem
);
796 EXPORT_SYMBOL(setup_arg_pages
);
801 * Transfer the program arguments and environment from the holding pages
802 * onto the stack. The provided stack pointer is adjusted accordingly.
804 int transfer_args_to_stack(struct linux_binprm
*bprm
,
805 unsigned long *sp_location
)
807 unsigned long index
, stop
, sp
;
810 stop
= bprm
->p
>> PAGE_SHIFT
;
813 for (index
= MAX_ARG_PAGES
- 1; index
>= stop
; index
--) {
814 unsigned int offset
= index
== stop
? bprm
->p
& ~PAGE_MASK
: 0;
815 char *src
= kmap(bprm
->page
[index
]) + offset
;
816 sp
-= PAGE_SIZE
- offset
;
817 if (copy_to_user((void *) sp
, src
, PAGE_SIZE
- offset
) != 0)
819 kunmap(bprm
->page
[index
]);
829 EXPORT_SYMBOL(transfer_args_to_stack
);
831 #endif /* CONFIG_MMU */
833 static struct file
*do_open_execat(int fd
, struct filename
*name
, int flags
)
837 struct open_flags open_exec_flags
= {
838 .open_flag
= O_LARGEFILE
| O_RDONLY
| __FMODE_EXEC
,
839 .acc_mode
= MAY_EXEC
,
840 .intent
= LOOKUP_OPEN
,
841 .lookup_flags
= LOOKUP_FOLLOW
,
844 if ((flags
& ~(AT_SYMLINK_NOFOLLOW
| AT_EMPTY_PATH
)) != 0)
845 return ERR_PTR(-EINVAL
);
846 if (flags
& AT_SYMLINK_NOFOLLOW
)
847 open_exec_flags
.lookup_flags
&= ~LOOKUP_FOLLOW
;
848 if (flags
& AT_EMPTY_PATH
)
849 open_exec_flags
.lookup_flags
|= LOOKUP_EMPTY
;
851 file
= do_filp_open(fd
, name
, &open_exec_flags
);
856 if (!S_ISREG(file_inode(file
)->i_mode
))
859 if (path_noexec(&file
->f_path
))
862 err
= deny_write_access(file
);
866 if (name
->name
[0] != '\0')
869 trace_open_exec(name
->name
);
879 struct file
*open_exec(const char *name
)
881 struct filename
*filename
= getname_kernel(name
);
882 struct file
*f
= ERR_CAST(filename
);
884 if (!IS_ERR(filename
)) {
885 f
= do_open_execat(AT_FDCWD
, filename
, 0);
890 EXPORT_SYMBOL(open_exec
);
892 int kernel_read(struct file
*file
, loff_t offset
,
893 char *addr
, unsigned long count
)
901 /* The cast to a user pointer is valid due to the set_fs() */
902 result
= vfs_read(file
, (void __user
*)addr
, count
, &pos
);
907 EXPORT_SYMBOL(kernel_read
);
909 int kernel_read_file(struct file
*file
, void **buf
, loff_t
*size
,
910 loff_t max_size
, enum kernel_read_file_id id
)
916 if (!S_ISREG(file_inode(file
)->i_mode
) || max_size
< 0)
919 ret
= security_kernel_read_file(file
, id
);
923 ret
= deny_write_access(file
);
927 i_size
= i_size_read(file_inode(file
));
928 if (max_size
> 0 && i_size
> max_size
) {
937 if (id
!= READING_FIRMWARE_PREALLOC_BUFFER
)
938 *buf
= vmalloc(i_size
);
945 while (pos
< i_size
) {
946 bytes
= kernel_read(file
, pos
, (char *)(*buf
) + pos
,
963 ret
= security_kernel_post_read_file(file
, *buf
, i_size
, id
);
969 if (id
!= READING_FIRMWARE_PREALLOC_BUFFER
) {
976 allow_write_access(file
);
979 EXPORT_SYMBOL_GPL(kernel_read_file
);
981 int kernel_read_file_from_path(char *path
, void **buf
, loff_t
*size
,
982 loff_t max_size
, enum kernel_read_file_id id
)
990 file
= filp_open(path
, O_RDONLY
, 0);
992 return PTR_ERR(file
);
994 ret
= kernel_read_file(file
, buf
, size
, max_size
, id
);
998 EXPORT_SYMBOL_GPL(kernel_read_file_from_path
);
1000 int kernel_read_file_from_fd(int fd
, void **buf
, loff_t
*size
, loff_t max_size
,
1001 enum kernel_read_file_id id
)
1003 struct fd f
= fdget(fd
);
1009 ret
= kernel_read_file(f
.file
, buf
, size
, max_size
, id
);
1014 EXPORT_SYMBOL_GPL(kernel_read_file_from_fd
);
1016 ssize_t
read_code(struct file
*file
, unsigned long addr
, loff_t pos
, size_t len
)
1018 ssize_t res
= vfs_read(file
, (void __user
*)addr
, len
, &pos
);
1020 flush_icache_range(addr
, addr
+ len
);
1023 EXPORT_SYMBOL(read_code
);
1025 static int exec_mmap(struct mm_struct
*mm
)
1027 struct task_struct
*tsk
;
1028 struct mm_struct
*old_mm
, *active_mm
;
1030 /* Notify parent that we're no longer interested in the old VM */
1032 old_mm
= current
->mm
;
1033 mm_release(tsk
, old_mm
);
1036 sync_mm_rss(old_mm
);
1038 * Make sure that if there is a core dump in progress
1039 * for the old mm, we get out and die instead of going
1040 * through with the exec. We must hold mmap_sem around
1041 * checking core_state and changing tsk->mm.
1043 down_read(&old_mm
->mmap_sem
);
1044 if (unlikely(old_mm
->core_state
)) {
1045 up_read(&old_mm
->mmap_sem
);
1050 active_mm
= tsk
->active_mm
;
1052 tsk
->active_mm
= mm
;
1053 activate_mm(active_mm
, mm
);
1054 tsk
->mm
->vmacache_seqnum
= 0;
1055 vmacache_flush(tsk
);
1058 up_read(&old_mm
->mmap_sem
);
1059 BUG_ON(active_mm
!= old_mm
);
1060 setmax_mm_hiwater_rss(&tsk
->signal
->maxrss
, old_mm
);
1061 mm_update_next_owner(old_mm
);
1070 * This function makes sure the current process has its own signal table,
1071 * so that flush_signal_handlers can later reset the handlers without
1072 * disturbing other processes. (Other processes might share the signal
1073 * table via the CLONE_SIGHAND option to clone().)
1075 static int de_thread(struct task_struct
*tsk
)
1077 struct signal_struct
*sig
= tsk
->signal
;
1078 struct sighand_struct
*oldsighand
= tsk
->sighand
;
1079 spinlock_t
*lock
= &oldsighand
->siglock
;
1081 if (thread_group_empty(tsk
))
1082 goto no_thread_group
;
1085 * Kill all other threads in the thread group.
1087 spin_lock_irq(lock
);
1088 if (signal_group_exit(sig
)) {
1090 * Another group action in progress, just
1091 * return so that the signal is processed.
1093 spin_unlock_irq(lock
);
1097 sig
->group_exit_task
= tsk
;
1098 sig
->notify_count
= zap_other_threads(tsk
);
1099 if (!thread_group_leader(tsk
))
1100 sig
->notify_count
--;
1102 while (sig
->notify_count
) {
1103 __set_current_state(TASK_KILLABLE
);
1104 spin_unlock_irq(lock
);
1106 if (unlikely(__fatal_signal_pending(tsk
)))
1108 spin_lock_irq(lock
);
1110 spin_unlock_irq(lock
);
1113 * At this point all other threads have exited, all we have to
1114 * do is to wait for the thread group leader to become inactive,
1115 * and to assume its PID:
1117 if (!thread_group_leader(tsk
)) {
1118 struct task_struct
*leader
= tsk
->group_leader
;
1121 cgroup_threadgroup_change_begin(tsk
);
1122 write_lock_irq(&tasklist_lock
);
1124 * Do this under tasklist_lock to ensure that
1125 * exit_notify() can't miss ->group_exit_task
1127 sig
->notify_count
= -1;
1128 if (likely(leader
->exit_state
))
1130 __set_current_state(TASK_KILLABLE
);
1131 write_unlock_irq(&tasklist_lock
);
1132 cgroup_threadgroup_change_end(tsk
);
1134 if (unlikely(__fatal_signal_pending(tsk
)))
1139 * The only record we have of the real-time age of a
1140 * process, regardless of execs it's done, is start_time.
1141 * All the past CPU time is accumulated in signal_struct
1142 * from sister threads now dead. But in this non-leader
1143 * exec, nothing survives from the original leader thread,
1144 * whose birth marks the true age of this process now.
1145 * When we take on its identity by switching to its PID, we
1146 * also take its birthdate (always earlier than our own).
1148 tsk
->start_time
= leader
->start_time
;
1149 tsk
->real_start_time
= leader
->real_start_time
;
1151 BUG_ON(!same_thread_group(leader
, tsk
));
1152 BUG_ON(has_group_leader_pid(tsk
));
1154 * An exec() starts a new thread group with the
1155 * TGID of the previous thread group. Rehash the
1156 * two threads with a switched PID, and release
1157 * the former thread group leader:
1160 /* Become a process group leader with the old leader's pid.
1161 * The old leader becomes a thread of the this thread group.
1162 * Note: The old leader also uses this pid until release_task
1163 * is called. Odd but simple and correct.
1165 tsk
->pid
= leader
->pid
;
1166 change_pid(tsk
, PIDTYPE_PID
, task_pid(leader
));
1167 transfer_pid(leader
, tsk
, PIDTYPE_PGID
);
1168 transfer_pid(leader
, tsk
, PIDTYPE_SID
);
1170 list_replace_rcu(&leader
->tasks
, &tsk
->tasks
);
1171 list_replace_init(&leader
->sibling
, &tsk
->sibling
);
1173 tsk
->group_leader
= tsk
;
1174 leader
->group_leader
= tsk
;
1176 tsk
->exit_signal
= SIGCHLD
;
1177 leader
->exit_signal
= -1;
1179 BUG_ON(leader
->exit_state
!= EXIT_ZOMBIE
);
1180 leader
->exit_state
= EXIT_DEAD
;
1183 * We are going to release_task()->ptrace_unlink() silently,
1184 * the tracer can sleep in do_wait(). EXIT_DEAD guarantees
1185 * the tracer wont't block again waiting for this thread.
1187 if (unlikely(leader
->ptrace
))
1188 __wake_up_parent(leader
, leader
->parent
);
1189 write_unlock_irq(&tasklist_lock
);
1190 cgroup_threadgroup_change_end(tsk
);
1192 release_task(leader
);
1195 sig
->group_exit_task
= NULL
;
1196 sig
->notify_count
= 0;
1199 /* we have changed execution domain */
1200 tsk
->exit_signal
= SIGCHLD
;
1202 #ifdef CONFIG_POSIX_TIMERS
1204 flush_itimer_signals();
1207 if (atomic_read(&oldsighand
->count
) != 1) {
1208 struct sighand_struct
*newsighand
;
1210 * This ->sighand is shared with the CLONE_SIGHAND
1211 * but not CLONE_THREAD task, switch to the new one.
1213 newsighand
= kmem_cache_alloc(sighand_cachep
, GFP_KERNEL
);
1217 atomic_set(&newsighand
->count
, 1);
1218 memcpy(newsighand
->action
, oldsighand
->action
,
1219 sizeof(newsighand
->action
));
1221 write_lock_irq(&tasklist_lock
);
1222 spin_lock(&oldsighand
->siglock
);
1223 rcu_assign_pointer(tsk
->sighand
, newsighand
);
1224 spin_unlock(&oldsighand
->siglock
);
1225 write_unlock_irq(&tasklist_lock
);
1227 __cleanup_sighand(oldsighand
);
1230 BUG_ON(!thread_group_leader(tsk
));
1234 /* protects against exit_notify() and __exit_signal() */
1235 read_lock(&tasklist_lock
);
1236 sig
->group_exit_task
= NULL
;
1237 sig
->notify_count
= 0;
1238 read_unlock(&tasklist_lock
);
1242 char *get_task_comm(char *buf
, struct task_struct
*tsk
)
1244 /* buf must be at least sizeof(tsk->comm) in size */
1246 strncpy(buf
, tsk
->comm
, sizeof(tsk
->comm
));
1250 EXPORT_SYMBOL_GPL(get_task_comm
);
1253 * These functions flushes out all traces of the currently running executable
1254 * so that a new one can be started
1257 void __set_task_comm(struct task_struct
*tsk
, const char *buf
, bool exec
)
1260 trace_task_rename(tsk
, buf
);
1261 strlcpy(tsk
->comm
, buf
, sizeof(tsk
->comm
));
1263 perf_event_comm(tsk
, exec
);
1266 int flush_old_exec(struct linux_binprm
* bprm
)
1271 * Make sure we have a private signal table and that
1272 * we are unassociated from the previous thread group.
1274 retval
= de_thread(current
);
1279 * Must be called _before_ exec_mmap() as bprm->mm is
1280 * not visibile until then. This also enables the update
1283 set_mm_exe_file(bprm
->mm
, bprm
->file
);
1286 * Release all of the old mmap stuff
1288 acct_arg_size(bprm
, 0);
1289 retval
= exec_mmap(bprm
->mm
);
1293 bprm
->mm
= NULL
; /* We're using it now */
1296 current
->flags
&= ~(PF_RANDOMIZE
| PF_FORKNOEXEC
| PF_KTHREAD
|
1297 PF_NOFREEZE
| PF_NO_SETAFFINITY
);
1299 current
->personality
&= ~bprm
->per_clear
;
1302 * We have to apply CLOEXEC before we change whether the process is
1303 * dumpable (in setup_new_exec) to avoid a race with a process in userspace
1304 * trying to access the should-be-closed file descriptors of a process
1305 * undergoing exec(2).
1307 do_close_on_exec(current
->files
);
1313 EXPORT_SYMBOL(flush_old_exec
);
1315 void would_dump(struct linux_binprm
*bprm
, struct file
*file
)
1317 struct inode
*inode
= file_inode(file
);
1318 if (inode_permission(inode
, MAY_READ
) < 0) {
1319 struct user_namespace
*old
, *user_ns
;
1320 bprm
->interp_flags
|= BINPRM_FLAGS_ENFORCE_NONDUMP
;
1322 /* Ensure mm->user_ns contains the executable */
1323 user_ns
= old
= bprm
->mm
->user_ns
;
1324 while ((user_ns
!= &init_user_ns
) &&
1325 !privileged_wrt_inode_uidgid(user_ns
, inode
))
1326 user_ns
= user_ns
->parent
;
1328 if (old
!= user_ns
) {
1329 bprm
->mm
->user_ns
= get_user_ns(user_ns
);
1334 EXPORT_SYMBOL(would_dump
);
1336 void setup_new_exec(struct linux_binprm
* bprm
)
1338 arch_pick_mmap_layout(current
->mm
);
1340 /* This is the point of no return */
1341 current
->sas_ss_sp
= current
->sas_ss_size
= 0;
1343 if (uid_eq(current_euid(), current_uid()) && gid_eq(current_egid(), current_gid()))
1344 set_dumpable(current
->mm
, SUID_DUMP_USER
);
1346 set_dumpable(current
->mm
, suid_dumpable
);
1348 arch_setup_new_exec();
1350 __set_task_comm(current
, kbasename(bprm
->filename
), true);
1352 /* Set the new mm task size. We have to do that late because it may
1353 * depend on TIF_32BIT which is only updated in flush_thread() on
1354 * some architectures like powerpc
1356 current
->mm
->task_size
= TASK_SIZE
;
1358 /* install the new credentials */
1359 if (!uid_eq(bprm
->cred
->uid
, current_euid()) ||
1360 !gid_eq(bprm
->cred
->gid
, current_egid())) {
1361 current
->pdeath_signal
= 0;
1363 if (bprm
->interp_flags
& BINPRM_FLAGS_ENFORCE_NONDUMP
)
1364 set_dumpable(current
->mm
, suid_dumpable
);
1367 /* An exec changes our domain. We are no longer part of the thread
1369 current
->self_exec_id
++;
1370 flush_signal_handlers(current
, 0);
1372 EXPORT_SYMBOL(setup_new_exec
);
1375 * Prepare credentials and lock ->cred_guard_mutex.
1376 * install_exec_creds() commits the new creds and drops the lock.
1377 * Or, if exec fails before, free_bprm() should release ->cred and
1380 int prepare_bprm_creds(struct linux_binprm
*bprm
)
1382 if (mutex_lock_interruptible(¤t
->signal
->cred_guard_mutex
))
1383 return -ERESTARTNOINTR
;
1385 bprm
->cred
= prepare_exec_creds();
1386 if (likely(bprm
->cred
))
1389 mutex_unlock(¤t
->signal
->cred_guard_mutex
);
1393 static void free_bprm(struct linux_binprm
*bprm
)
1395 free_arg_pages(bprm
);
1397 mutex_unlock(¤t
->signal
->cred_guard_mutex
);
1398 abort_creds(bprm
->cred
);
1401 allow_write_access(bprm
->file
);
1404 /* If a binfmt changed the interp, free it. */
1405 if (bprm
->interp
!= bprm
->filename
)
1406 kfree(bprm
->interp
);
1410 int bprm_change_interp(char *interp
, struct linux_binprm
*bprm
)
1412 /* If a binfmt changed the interp, free it first. */
1413 if (bprm
->interp
!= bprm
->filename
)
1414 kfree(bprm
->interp
);
1415 bprm
->interp
= kstrdup(interp
, GFP_KERNEL
);
1420 EXPORT_SYMBOL(bprm_change_interp
);
1423 * install the new credentials for this executable
1425 void install_exec_creds(struct linux_binprm
*bprm
)
1427 security_bprm_committing_creds(bprm
);
1429 commit_creds(bprm
->cred
);
1433 * Disable monitoring for regular users
1434 * when executing setuid binaries. Must
1435 * wait until new credentials are committed
1436 * by commit_creds() above
1438 if (get_dumpable(current
->mm
) != SUID_DUMP_USER
)
1439 perf_event_exit_task(current
);
1441 * cred_guard_mutex must be held at least to this point to prevent
1442 * ptrace_attach() from altering our determination of the task's
1443 * credentials; any time after this it may be unlocked.
1445 security_bprm_committed_creds(bprm
);
1446 mutex_unlock(¤t
->signal
->cred_guard_mutex
);
1448 EXPORT_SYMBOL(install_exec_creds
);
1451 * determine how safe it is to execute the proposed program
1452 * - the caller must hold ->cred_guard_mutex to protect against
1453 * PTRACE_ATTACH or seccomp thread-sync
1455 static void check_unsafe_exec(struct linux_binprm
*bprm
)
1457 struct task_struct
*p
= current
, *t
;
1461 bprm
->unsafe
|= LSM_UNSAFE_PTRACE
;
1464 * This isn't strictly necessary, but it makes it harder for LSMs to
1467 if (task_no_new_privs(current
))
1468 bprm
->unsafe
|= LSM_UNSAFE_NO_NEW_PRIVS
;
1472 spin_lock(&p
->fs
->lock
);
1474 while_each_thread(p
, t
) {
1480 if (p
->fs
->users
> n_fs
)
1481 bprm
->unsafe
|= LSM_UNSAFE_SHARE
;
1484 spin_unlock(&p
->fs
->lock
);
1487 static void bprm_fill_uid(struct linux_binprm
*bprm
)
1489 struct inode
*inode
;
1495 * Since this can be called multiple times (via prepare_binprm),
1496 * we must clear any previous work done when setting set[ug]id
1497 * bits from any earlier bprm->file uses (for example when run
1498 * first for a setuid script then again for its interpreter).
1500 bprm
->cred
->euid
= current_euid();
1501 bprm
->cred
->egid
= current_egid();
1503 if (!mnt_may_suid(bprm
->file
->f_path
.mnt
))
1506 if (task_no_new_privs(current
))
1509 inode
= bprm
->file
->f_path
.dentry
->d_inode
;
1510 mode
= READ_ONCE(inode
->i_mode
);
1511 if (!(mode
& (S_ISUID
|S_ISGID
)))
1514 /* Be careful if suid/sgid is set */
1517 /* reload atomically mode/uid/gid now that lock held */
1518 mode
= inode
->i_mode
;
1521 inode_unlock(inode
);
1523 /* We ignore suid/sgid if there are no mappings for them in the ns */
1524 if (!kuid_has_mapping(bprm
->cred
->user_ns
, uid
) ||
1525 !kgid_has_mapping(bprm
->cred
->user_ns
, gid
))
1528 if (mode
& S_ISUID
) {
1529 bprm
->per_clear
|= PER_CLEAR_ON_SETID
;
1530 bprm
->cred
->euid
= uid
;
1533 if ((mode
& (S_ISGID
| S_IXGRP
)) == (S_ISGID
| S_IXGRP
)) {
1534 bprm
->per_clear
|= PER_CLEAR_ON_SETID
;
1535 bprm
->cred
->egid
= gid
;
1540 * Fill the binprm structure from the inode.
1541 * Check permissions, then read the first 128 (BINPRM_BUF_SIZE) bytes
1543 * This may be called multiple times for binary chains (scripts for example).
1545 int prepare_binprm(struct linux_binprm
*bprm
)
1549 bprm_fill_uid(bprm
);
1551 /* fill in binprm security blob */
1552 retval
= security_bprm_set_creds(bprm
);
1555 bprm
->cred_prepared
= 1;
1557 memset(bprm
->buf
, 0, BINPRM_BUF_SIZE
);
1558 return kernel_read(bprm
->file
, 0, bprm
->buf
, BINPRM_BUF_SIZE
);
1561 EXPORT_SYMBOL(prepare_binprm
);
1564 * Arguments are '\0' separated strings found at the location bprm->p
1565 * points to; chop off the first by relocating brpm->p to right after
1566 * the first '\0' encountered.
1568 int remove_arg_zero(struct linux_binprm
*bprm
)
1571 unsigned long offset
;
1579 offset
= bprm
->p
& ~PAGE_MASK
;
1580 page
= get_arg_page(bprm
, bprm
->p
, 0);
1585 kaddr
= kmap_atomic(page
);
1587 for (; offset
< PAGE_SIZE
&& kaddr
[offset
];
1588 offset
++, bprm
->p
++)
1591 kunmap_atomic(kaddr
);
1593 } while (offset
== PAGE_SIZE
);
1602 EXPORT_SYMBOL(remove_arg_zero
);
1604 #define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e))
1606 * cycle the list of binary formats handler, until one recognizes the image
1608 int search_binary_handler(struct linux_binprm
*bprm
)
1610 bool need_retry
= IS_ENABLED(CONFIG_MODULES
);
1611 struct linux_binfmt
*fmt
;
1614 /* This allows 4 levels of binfmt rewrites before failing hard. */
1615 if (bprm
->recursion_depth
> 5)
1618 retval
= security_bprm_check(bprm
);
1624 read_lock(&binfmt_lock
);
1625 list_for_each_entry(fmt
, &formats
, lh
) {
1626 if (!try_module_get(fmt
->module
))
1628 read_unlock(&binfmt_lock
);
1629 bprm
->recursion_depth
++;
1630 retval
= fmt
->load_binary(bprm
);
1631 read_lock(&binfmt_lock
);
1633 bprm
->recursion_depth
--;
1634 if (retval
< 0 && !bprm
->mm
) {
1635 /* we got to flush_old_exec() and failed after it */
1636 read_unlock(&binfmt_lock
);
1637 force_sigsegv(SIGSEGV
, current
);
1640 if (retval
!= -ENOEXEC
|| !bprm
->file
) {
1641 read_unlock(&binfmt_lock
);
1645 read_unlock(&binfmt_lock
);
1648 if (printable(bprm
->buf
[0]) && printable(bprm
->buf
[1]) &&
1649 printable(bprm
->buf
[2]) && printable(bprm
->buf
[3]))
1651 if (request_module("binfmt-%04x", *(ushort
*)(bprm
->buf
+ 2)) < 0)
1659 EXPORT_SYMBOL(search_binary_handler
);
1661 static int exec_binprm(struct linux_binprm
*bprm
)
1663 pid_t old_pid
, old_vpid
;
1666 /* Need to fetch pid before load_binary changes it */
1667 old_pid
= current
->pid
;
1669 old_vpid
= task_pid_nr_ns(current
, task_active_pid_ns(current
->parent
));
1672 ret
= search_binary_handler(bprm
);
1675 trace_sched_process_exec(current
, old_pid
, bprm
);
1676 ptrace_event(PTRACE_EVENT_EXEC
, old_vpid
);
1677 proc_exec_connector(current
);
1684 * sys_execve() executes a new program.
1686 static int do_execveat_common(int fd
, struct filename
*filename
,
1687 struct user_arg_ptr argv
,
1688 struct user_arg_ptr envp
,
1691 char *pathbuf
= NULL
;
1692 struct linux_binprm
*bprm
;
1694 struct files_struct
*displaced
;
1697 if (IS_ERR(filename
))
1698 return PTR_ERR(filename
);
1701 * We move the actual failure in case of RLIMIT_NPROC excess from
1702 * set*uid() to execve() because too many poorly written programs
1703 * don't check setuid() return code. Here we additionally recheck
1704 * whether NPROC limit is still exceeded.
1706 if ((current
->flags
& PF_NPROC_EXCEEDED
) &&
1707 atomic_read(¤t_user()->processes
) > rlimit(RLIMIT_NPROC
)) {
1712 /* We're below the limit (still or again), so we don't want to make
1713 * further execve() calls fail. */
1714 current
->flags
&= ~PF_NPROC_EXCEEDED
;
1716 retval
= unshare_files(&displaced
);
1721 bprm
= kzalloc(sizeof(*bprm
), GFP_KERNEL
);
1725 retval
= prepare_bprm_creds(bprm
);
1729 check_unsafe_exec(bprm
);
1730 current
->in_execve
= 1;
1732 file
= do_open_execat(fd
, filename
, flags
);
1733 retval
= PTR_ERR(file
);
1740 if (fd
== AT_FDCWD
|| filename
->name
[0] == '/') {
1741 bprm
->filename
= filename
->name
;
1743 if (filename
->name
[0] == '\0')
1744 pathbuf
= kasprintf(GFP_TEMPORARY
, "/dev/fd/%d", fd
);
1746 pathbuf
= kasprintf(GFP_TEMPORARY
, "/dev/fd/%d/%s",
1747 fd
, filename
->name
);
1753 * Record that a name derived from an O_CLOEXEC fd will be
1754 * inaccessible after exec. Relies on having exclusive access to
1755 * current->files (due to unshare_files above).
1757 if (close_on_exec(fd
, rcu_dereference_raw(current
->files
->fdt
)))
1758 bprm
->interp_flags
|= BINPRM_FLAGS_PATH_INACCESSIBLE
;
1759 bprm
->filename
= pathbuf
;
1761 bprm
->interp
= bprm
->filename
;
1763 retval
= bprm_mm_init(bprm
);
1767 bprm
->argc
= count(argv
, MAX_ARG_STRINGS
);
1768 if ((retval
= bprm
->argc
) < 0)
1771 bprm
->envc
= count(envp
, MAX_ARG_STRINGS
);
1772 if ((retval
= bprm
->envc
) < 0)
1775 retval
= prepare_binprm(bprm
);
1779 retval
= copy_strings_kernel(1, &bprm
->filename
, bprm
);
1783 bprm
->exec
= bprm
->p
;
1784 retval
= copy_strings(bprm
->envc
, envp
, bprm
);
1788 retval
= copy_strings(bprm
->argc
, argv
, bprm
);
1792 would_dump(bprm
, bprm
->file
);
1794 retval
= exec_binprm(bprm
);
1798 /* execve succeeded */
1799 current
->fs
->in_exec
= 0;
1800 current
->in_execve
= 0;
1801 acct_update_integrals(current
);
1802 task_numa_free(current
);
1807 put_files_struct(displaced
);
1812 acct_arg_size(bprm
, 0);
1817 current
->fs
->in_exec
= 0;
1818 current
->in_execve
= 0;
1826 reset_files_struct(displaced
);
1832 int do_execve(struct filename
*filename
,
1833 const char __user
*const __user
*__argv
,
1834 const char __user
*const __user
*__envp
)
1836 struct user_arg_ptr argv
= { .ptr
.native
= __argv
};
1837 struct user_arg_ptr envp
= { .ptr
.native
= __envp
};
1838 return do_execveat_common(AT_FDCWD
, filename
, argv
, envp
, 0);
1841 int do_execveat(int fd
, struct filename
*filename
,
1842 const char __user
*const __user
*__argv
,
1843 const char __user
*const __user
*__envp
,
1846 struct user_arg_ptr argv
= { .ptr
.native
= __argv
};
1847 struct user_arg_ptr envp
= { .ptr
.native
= __envp
};
1849 return do_execveat_common(fd
, filename
, argv
, envp
, flags
);
1852 #ifdef CONFIG_COMPAT
1853 static int compat_do_execve(struct filename
*filename
,
1854 const compat_uptr_t __user
*__argv
,
1855 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(AT_FDCWD
, filename
, argv
, envp
, 0);
1868 static int compat_do_execveat(int fd
, struct filename
*filename
,
1869 const compat_uptr_t __user
*__argv
,
1870 const compat_uptr_t __user
*__envp
,
1873 struct user_arg_ptr argv
= {
1875 .ptr
.compat
= __argv
,
1877 struct user_arg_ptr envp
= {
1879 .ptr
.compat
= __envp
,
1881 return do_execveat_common(fd
, filename
, argv
, envp
, flags
);
1885 void set_binfmt(struct linux_binfmt
*new)
1887 struct mm_struct
*mm
= current
->mm
;
1890 module_put(mm
->binfmt
->module
);
1894 __module_get(new->module
);
1896 EXPORT_SYMBOL(set_binfmt
);
1899 * set_dumpable stores three-value SUID_DUMP_* into mm->flags.
1901 void set_dumpable(struct mm_struct
*mm
, int value
)
1903 unsigned long old
, new;
1905 if (WARN_ON((unsigned)value
> SUID_DUMP_ROOT
))
1909 old
= ACCESS_ONCE(mm
->flags
);
1910 new = (old
& ~MMF_DUMPABLE_MASK
) | value
;
1911 } while (cmpxchg(&mm
->flags
, old
, new) != old
);
1914 SYSCALL_DEFINE3(execve
,
1915 const char __user
*, filename
,
1916 const char __user
*const __user
*, argv
,
1917 const char __user
*const __user
*, envp
)
1919 return do_execve(getname(filename
), argv
, envp
);
1922 SYSCALL_DEFINE5(execveat
,
1923 int, fd
, const char __user
*, filename
,
1924 const char __user
*const __user
*, argv
,
1925 const char __user
*const __user
*, envp
,
1928 int lookup_flags
= (flags
& AT_EMPTY_PATH
) ? LOOKUP_EMPTY
: 0;
1930 return do_execveat(fd
,
1931 getname_flags(filename
, lookup_flags
, NULL
),
1935 #ifdef CONFIG_COMPAT
1936 COMPAT_SYSCALL_DEFINE3(execve
, const char __user
*, filename
,
1937 const compat_uptr_t __user
*, argv
,
1938 const compat_uptr_t __user
*, envp
)
1940 return compat_do_execve(getname(filename
), argv
, envp
);
1943 COMPAT_SYSCALL_DEFINE5(execveat
, int, fd
,
1944 const char __user
*, filename
,
1945 const compat_uptr_t __user
*, argv
,
1946 const compat_uptr_t __user
*, envp
,
1949 int lookup_flags
= (flags
& AT_EMPTY_PATH
) ? LOOKUP_EMPTY
: 0;
1951 return compat_do_execveat(fd
,
1952 getname_flags(filename
, lookup_flags
, NULL
),