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
);
114 EXPORT_SYMBOL_GPL(path_noexec
);
116 bool path_nosuid(const struct path
*path
)
118 return !mnt_may_suid(path
->mnt
) ||
119 (path
->mnt
->mnt_sb
->s_iflags
& SB_I_NOSUID
);
121 EXPORT_SYMBOL(path_nosuid
);
125 * Note that a shared library must be both readable and executable due to
128 * Also note that we take the address to load from from the file itself.
130 SYSCALL_DEFINE1(uselib
, const char __user
*, library
)
132 struct linux_binfmt
*fmt
;
134 struct filename
*tmp
= getname(library
);
135 int error
= PTR_ERR(tmp
);
136 static const struct open_flags uselib_flags
= {
137 .open_flag
= O_LARGEFILE
| O_RDONLY
| __FMODE_EXEC
,
138 .acc_mode
= MAY_READ
| MAY_EXEC
,
139 .intent
= LOOKUP_OPEN
,
140 .lookup_flags
= LOOKUP_FOLLOW
,
146 file
= do_filp_open(AT_FDCWD
, tmp
, &uselib_flags
);
148 error
= PTR_ERR(file
);
153 if (!S_ISREG(file_inode(file
)->i_mode
))
157 if (path_noexec(&file
->f_path
))
164 read_lock(&binfmt_lock
);
165 list_for_each_entry(fmt
, &formats
, lh
) {
166 if (!fmt
->load_shlib
)
168 if (!try_module_get(fmt
->module
))
170 read_unlock(&binfmt_lock
);
171 error
= fmt
->load_shlib(file
);
172 read_lock(&binfmt_lock
);
174 if (error
!= -ENOEXEC
)
177 read_unlock(&binfmt_lock
);
183 #endif /* #ifdef CONFIG_USELIB */
187 * The nascent bprm->mm is not visible until exec_mmap() but it can
188 * use a lot of memory, account these pages in current->mm temporary
189 * for oom_badness()->get_mm_rss(). Once exec succeeds or fails, we
190 * change the counter back via acct_arg_size(0).
192 static void acct_arg_size(struct linux_binprm
*bprm
, unsigned long pages
)
194 struct mm_struct
*mm
= current
->mm
;
195 long diff
= (long)(pages
- bprm
->vma_pages
);
200 bprm
->vma_pages
= pages
;
201 add_mm_counter(mm
, MM_ANONPAGES
, diff
);
204 static struct page
*get_arg_page(struct linux_binprm
*bprm
, unsigned long pos
,
209 unsigned int gup_flags
= FOLL_FORCE
;
211 #ifdef CONFIG_STACK_GROWSUP
213 ret
= expand_downwards(bprm
->vma
, pos
);
220 gup_flags
|= FOLL_WRITE
;
223 * We are doing an exec(). 'current' is the process
224 * doing the exec and bprm->mm is the new process's mm.
226 ret
= get_user_pages_remote(current
, bprm
->mm
, pos
, 1, gup_flags
,
232 unsigned long size
= bprm
->vma
->vm_end
- bprm
->vma
->vm_start
;
233 unsigned long ptr_size
, limit
;
236 * Since the stack will hold pointers to the strings, we
237 * must account for them as well.
239 * The size calculation is the entire vma while each arg page is
240 * built, so each time we get here it's calculating how far it
241 * is currently (rather than each call being just the newly
242 * added size from the arg page). As a result, we need to
243 * always add the entire size of the pointers, so that on the
244 * last call to get_arg_page() we'll actually have the entire
247 ptr_size
= (bprm
->argc
+ bprm
->envc
) * sizeof(void *);
248 if (ptr_size
> ULONG_MAX
- size
)
252 acct_arg_size(bprm
, size
/ PAGE_SIZE
);
255 * We've historically supported up to 32 pages (ARG_MAX)
256 * of argument strings even with small stacks
262 * Limit to 1/4 of the max stack size or 3/4 of _STK_LIM
263 * (whichever is smaller) for the argv+env strings.
265 * - the remaining binfmt code will not run out of stack space,
266 * - the program will have a reasonable amount of stack left
269 limit
= _STK_LIM
/ 4 * 3;
270 limit
= min(limit
, rlimit(RLIMIT_STACK
) / 4);
282 static void put_arg_page(struct page
*page
)
287 static void free_arg_pages(struct linux_binprm
*bprm
)
291 static void flush_arg_page(struct linux_binprm
*bprm
, unsigned long pos
,
294 flush_cache_page(bprm
->vma
, pos
, page_to_pfn(page
));
297 static int __bprm_mm_init(struct linux_binprm
*bprm
)
300 struct vm_area_struct
*vma
= NULL
;
301 struct mm_struct
*mm
= bprm
->mm
;
303 bprm
->vma
= vma
= kmem_cache_zalloc(vm_area_cachep
, GFP_KERNEL
);
307 if (down_write_killable(&mm
->mmap_sem
)) {
314 * Place the stack at the largest stack address the architecture
315 * supports. Later, we'll move this to an appropriate place. We don't
316 * use STACK_TOP because that can depend on attributes which aren't
319 BUILD_BUG_ON(VM_STACK_FLAGS
& VM_STACK_INCOMPLETE_SETUP
);
320 vma
->vm_end
= STACK_TOP_MAX
;
321 vma
->vm_start
= vma
->vm_end
- PAGE_SIZE
;
322 vma
->vm_flags
= VM_SOFTDIRTY
| VM_STACK_FLAGS
| VM_STACK_INCOMPLETE_SETUP
;
323 vma
->vm_page_prot
= vm_get_page_prot(vma
->vm_flags
);
324 INIT_LIST_HEAD(&vma
->anon_vma_chain
);
326 err
= insert_vm_struct(mm
, vma
);
330 mm
->stack_vm
= mm
->total_vm
= 1;
331 arch_bprm_mm_init(mm
, vma
);
332 up_write(&mm
->mmap_sem
);
333 bprm
->p
= vma
->vm_end
- sizeof(void *);
336 up_write(&mm
->mmap_sem
);
339 kmem_cache_free(vm_area_cachep
, vma
);
343 static bool valid_arg_len(struct linux_binprm
*bprm
, long len
)
345 return len
<= MAX_ARG_STRLEN
;
350 static inline void acct_arg_size(struct linux_binprm
*bprm
, unsigned long pages
)
354 static struct page
*get_arg_page(struct linux_binprm
*bprm
, unsigned long pos
,
359 page
= bprm
->page
[pos
/ PAGE_SIZE
];
360 if (!page
&& write
) {
361 page
= alloc_page(GFP_HIGHUSER
|__GFP_ZERO
);
364 bprm
->page
[pos
/ PAGE_SIZE
] = page
;
370 static void put_arg_page(struct page
*page
)
374 static void free_arg_page(struct linux_binprm
*bprm
, int i
)
377 __free_page(bprm
->page
[i
]);
378 bprm
->page
[i
] = NULL
;
382 static void free_arg_pages(struct linux_binprm
*bprm
)
386 for (i
= 0; i
< MAX_ARG_PAGES
; i
++)
387 free_arg_page(bprm
, i
);
390 static void flush_arg_page(struct linux_binprm
*bprm
, unsigned long pos
,
395 static int __bprm_mm_init(struct linux_binprm
*bprm
)
397 bprm
->p
= PAGE_SIZE
* MAX_ARG_PAGES
- sizeof(void *);
401 static bool valid_arg_len(struct linux_binprm
*bprm
, long len
)
403 return len
<= bprm
->p
;
406 #endif /* CONFIG_MMU */
409 * Create a new mm_struct and populate it with a temporary stack
410 * vm_area_struct. We don't have enough context at this point to set the stack
411 * flags, permissions, and offset, so we use temporary values. We'll update
412 * them later in setup_arg_pages().
414 static int bprm_mm_init(struct linux_binprm
*bprm
)
417 struct mm_struct
*mm
= NULL
;
419 bprm
->mm
= mm
= mm_alloc();
424 err
= __bprm_mm_init(bprm
);
439 struct user_arg_ptr
{
444 const char __user
*const __user
*native
;
446 const compat_uptr_t __user
*compat
;
451 static const char __user
*get_user_arg_ptr(struct user_arg_ptr argv
, int nr
)
453 const char __user
*native
;
456 if (unlikely(argv
.is_compat
)) {
457 compat_uptr_t compat
;
459 if (get_user(compat
, argv
.ptr
.compat
+ nr
))
460 return ERR_PTR(-EFAULT
);
462 return compat_ptr(compat
);
466 if (get_user(native
, argv
.ptr
.native
+ nr
))
467 return ERR_PTR(-EFAULT
);
473 * count() counts the number of strings in array ARGV.
475 static int count(struct user_arg_ptr argv
, int max
)
479 if (argv
.ptr
.native
!= NULL
) {
481 const char __user
*p
= get_user_arg_ptr(argv
, i
);
493 if (fatal_signal_pending(current
))
494 return -ERESTARTNOHAND
;
502 * 'copy_strings()' copies argument/environment strings from the old
503 * processes's memory to the new process's stack. The call to get_user_pages()
504 * ensures the destination page is created and not swapped out.
506 static int copy_strings(int argc
, struct user_arg_ptr argv
,
507 struct linux_binprm
*bprm
)
509 struct page
*kmapped_page
= NULL
;
511 unsigned long kpos
= 0;
515 const char __user
*str
;
520 str
= get_user_arg_ptr(argv
, argc
);
524 len
= strnlen_user(str
, MAX_ARG_STRLEN
);
529 if (!valid_arg_len(bprm
, len
))
532 /* We're going to work our way backwords. */
538 int offset
, bytes_to_copy
;
540 if (fatal_signal_pending(current
)) {
541 ret
= -ERESTARTNOHAND
;
546 offset
= pos
% PAGE_SIZE
;
550 bytes_to_copy
= offset
;
551 if (bytes_to_copy
> len
)
554 offset
-= bytes_to_copy
;
555 pos
-= bytes_to_copy
;
556 str
-= bytes_to_copy
;
557 len
-= bytes_to_copy
;
559 if (!kmapped_page
|| kpos
!= (pos
& PAGE_MASK
)) {
562 page
= get_arg_page(bprm
, pos
, 1);
569 flush_kernel_dcache_page(kmapped_page
);
570 kunmap(kmapped_page
);
571 put_arg_page(kmapped_page
);
574 kaddr
= kmap(kmapped_page
);
575 kpos
= pos
& PAGE_MASK
;
576 flush_arg_page(bprm
, kpos
, kmapped_page
);
578 if (copy_from_user(kaddr
+offset
, str
, bytes_to_copy
)) {
587 flush_kernel_dcache_page(kmapped_page
);
588 kunmap(kmapped_page
);
589 put_arg_page(kmapped_page
);
595 * Like copy_strings, but get argv and its values from kernel memory.
597 int copy_strings_kernel(int argc
, const char *const *__argv
,
598 struct linux_binprm
*bprm
)
601 mm_segment_t oldfs
= get_fs();
602 struct user_arg_ptr argv
= {
603 .ptr
.native
= (const char __user
*const __user
*)__argv
,
607 r
= copy_strings(argc
, argv
, bprm
);
612 EXPORT_SYMBOL(copy_strings_kernel
);
617 * During bprm_mm_init(), we create a temporary stack at STACK_TOP_MAX. Once
618 * the binfmt code determines where the new stack should reside, we shift it to
619 * its final location. The process proceeds as follows:
621 * 1) Use shift to calculate the new vma endpoints.
622 * 2) Extend vma to cover both the old and new ranges. This ensures the
623 * arguments passed to subsequent functions are consistent.
624 * 3) Move vma's page tables to the new range.
625 * 4) Free up any cleared pgd range.
626 * 5) Shrink the vma to cover only the new range.
628 static int shift_arg_pages(struct vm_area_struct
*vma
, unsigned long shift
)
630 struct mm_struct
*mm
= vma
->vm_mm
;
631 unsigned long old_start
= vma
->vm_start
;
632 unsigned long old_end
= vma
->vm_end
;
633 unsigned long length
= old_end
- old_start
;
634 unsigned long new_start
= old_start
- shift
;
635 unsigned long new_end
= old_end
- shift
;
636 struct mmu_gather tlb
;
638 BUG_ON(new_start
> new_end
);
641 * ensure there are no vmas between where we want to go
644 if (vma
!= find_vma(mm
, new_start
))
648 * cover the whole range: [new_start, old_end)
650 if (vma_adjust(vma
, new_start
, old_end
, vma
->vm_pgoff
, NULL
))
654 * move the page tables downwards, on failure we rely on
655 * process cleanup to remove whatever mess we made.
657 if (length
!= move_page_tables(vma
, old_start
,
658 vma
, new_start
, length
, false))
662 tlb_gather_mmu(&tlb
, mm
, old_start
, old_end
);
663 if (new_end
> old_start
) {
665 * when the old and new regions overlap clear from new_end.
667 free_pgd_range(&tlb
, new_end
, old_end
, new_end
,
668 vma
->vm_next
? vma
->vm_next
->vm_start
: USER_PGTABLES_CEILING
);
671 * otherwise, clean from old_start; this is done to not touch
672 * the address space in [new_end, old_start) some architectures
673 * have constraints on va-space that make this illegal (IA64) -
674 * for the others its just a little faster.
676 free_pgd_range(&tlb
, old_start
, old_end
, new_end
,
677 vma
->vm_next
? vma
->vm_next
->vm_start
: USER_PGTABLES_CEILING
);
679 tlb_finish_mmu(&tlb
, old_start
, old_end
);
682 * Shrink the vma to just the new range. Always succeeds.
684 vma_adjust(vma
, new_start
, new_end
, vma
->vm_pgoff
, NULL
);
690 * Finalizes the stack vm_area_struct. The flags and permissions are updated,
691 * the stack is optionally relocated, and some extra space is added.
693 int setup_arg_pages(struct linux_binprm
*bprm
,
694 unsigned long stack_top
,
695 int executable_stack
)
698 unsigned long stack_shift
;
699 struct mm_struct
*mm
= current
->mm
;
700 struct vm_area_struct
*vma
= bprm
->vma
;
701 struct vm_area_struct
*prev
= NULL
;
702 unsigned long vm_flags
;
703 unsigned long stack_base
;
704 unsigned long stack_size
;
705 unsigned long stack_expand
;
706 unsigned long rlim_stack
;
708 #ifdef CONFIG_STACK_GROWSUP
709 /* Limit stack size */
710 stack_base
= rlimit_max(RLIMIT_STACK
);
711 if (stack_base
> STACK_SIZE_MAX
)
712 stack_base
= STACK_SIZE_MAX
;
714 /* Add space for stack randomization. */
715 stack_base
+= (STACK_RND_MASK
<< PAGE_SHIFT
);
717 /* Make sure we didn't let the argument array grow too large. */
718 if (vma
->vm_end
- vma
->vm_start
> stack_base
)
721 stack_base
= PAGE_ALIGN(stack_top
- stack_base
);
723 stack_shift
= vma
->vm_start
- stack_base
;
724 mm
->arg_start
= bprm
->p
- stack_shift
;
725 bprm
->p
= vma
->vm_end
- stack_shift
;
727 stack_top
= arch_align_stack(stack_top
);
728 stack_top
= PAGE_ALIGN(stack_top
);
730 if (unlikely(stack_top
< mmap_min_addr
) ||
731 unlikely(vma
->vm_end
- vma
->vm_start
>= stack_top
- mmap_min_addr
))
734 stack_shift
= vma
->vm_end
- stack_top
;
736 bprm
->p
-= stack_shift
;
737 mm
->arg_start
= bprm
->p
;
741 bprm
->loader
-= stack_shift
;
742 bprm
->exec
-= stack_shift
;
744 if (down_write_killable(&mm
->mmap_sem
))
747 vm_flags
= VM_STACK_FLAGS
;
750 * Adjust stack execute permissions; explicitly enable for
751 * EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X and leave alone
752 * (arch default) otherwise.
754 if (unlikely(executable_stack
== EXSTACK_ENABLE_X
))
756 else if (executable_stack
== EXSTACK_DISABLE_X
)
757 vm_flags
&= ~VM_EXEC
;
758 vm_flags
|= mm
->def_flags
;
759 vm_flags
|= VM_STACK_INCOMPLETE_SETUP
;
761 ret
= mprotect_fixup(vma
, &prev
, vma
->vm_start
, vma
->vm_end
,
767 /* Move stack pages down in memory. */
769 ret
= shift_arg_pages(vma
, stack_shift
);
774 /* mprotect_fixup is overkill to remove the temporary stack flags */
775 vma
->vm_flags
&= ~VM_STACK_INCOMPLETE_SETUP
;
777 stack_expand
= 131072UL; /* randomly 32*4k (or 2*64k) pages */
778 stack_size
= vma
->vm_end
- vma
->vm_start
;
780 * Align this down to a page boundary as expand_stack
783 rlim_stack
= rlimit(RLIMIT_STACK
) & PAGE_MASK
;
784 #ifdef CONFIG_STACK_GROWSUP
785 if (stack_size
+ stack_expand
> rlim_stack
)
786 stack_base
= vma
->vm_start
+ rlim_stack
;
788 stack_base
= vma
->vm_end
+ stack_expand
;
790 if (stack_size
+ stack_expand
> rlim_stack
)
791 stack_base
= vma
->vm_end
- rlim_stack
;
793 stack_base
= vma
->vm_start
- stack_expand
;
795 current
->mm
->start_stack
= bprm
->p
;
796 ret
= expand_stack(vma
, stack_base
);
801 up_write(&mm
->mmap_sem
);
804 EXPORT_SYMBOL(setup_arg_pages
);
809 * Transfer the program arguments and environment from the holding pages
810 * onto the stack. The provided stack pointer is adjusted accordingly.
812 int transfer_args_to_stack(struct linux_binprm
*bprm
,
813 unsigned long *sp_location
)
815 unsigned long index
, stop
, sp
;
818 stop
= bprm
->p
>> PAGE_SHIFT
;
821 for (index
= MAX_ARG_PAGES
- 1; index
>= stop
; index
--) {
822 unsigned int offset
= index
== stop
? bprm
->p
& ~PAGE_MASK
: 0;
823 char *src
= kmap(bprm
->page
[index
]) + offset
;
824 sp
-= PAGE_SIZE
- offset
;
825 if (copy_to_user((void *) sp
, src
, PAGE_SIZE
- offset
) != 0)
827 kunmap(bprm
->page
[index
]);
837 EXPORT_SYMBOL(transfer_args_to_stack
);
839 #endif /* CONFIG_MMU */
841 static struct file
*do_open_execat(int fd
, struct filename
*name
, int flags
)
845 struct open_flags open_exec_flags
= {
846 .open_flag
= O_LARGEFILE
| O_RDONLY
| __FMODE_EXEC
,
847 .acc_mode
= MAY_EXEC
,
848 .intent
= LOOKUP_OPEN
,
849 .lookup_flags
= LOOKUP_FOLLOW
,
852 if ((flags
& ~(AT_SYMLINK_NOFOLLOW
| AT_EMPTY_PATH
)) != 0)
853 return ERR_PTR(-EINVAL
);
854 if (flags
& AT_SYMLINK_NOFOLLOW
)
855 open_exec_flags
.lookup_flags
&= ~LOOKUP_FOLLOW
;
856 if (flags
& AT_EMPTY_PATH
)
857 open_exec_flags
.lookup_flags
|= LOOKUP_EMPTY
;
859 file
= do_filp_open(fd
, name
, &open_exec_flags
);
864 if (!S_ISREG(file_inode(file
)->i_mode
))
867 if (path_noexec(&file
->f_path
))
870 err
= deny_write_access(file
);
874 if (name
->name
[0] != '\0')
877 trace_open_exec(name
->name
);
887 struct file
*open_exec(const char *name
)
889 struct filename
*filename
= getname_kernel(name
);
890 struct file
*f
= ERR_CAST(filename
);
892 if (!IS_ERR(filename
)) {
893 f
= do_open_execat(AT_FDCWD
, filename
, 0);
898 EXPORT_SYMBOL(open_exec
);
900 int kernel_read_file(struct file
*file
, void **buf
, loff_t
*size
,
901 loff_t max_size
, enum kernel_read_file_id id
)
907 if (!S_ISREG(file_inode(file
)->i_mode
) || max_size
< 0)
910 ret
= security_kernel_read_file(file
, id
);
914 ret
= deny_write_access(file
);
918 i_size
= i_size_read(file_inode(file
));
919 if (max_size
> 0 && i_size
> max_size
) {
928 if (id
!= READING_FIRMWARE_PREALLOC_BUFFER
)
929 *buf
= vmalloc(i_size
);
936 while (pos
< i_size
) {
937 bytes
= kernel_read(file
, *buf
+ pos
, i_size
- pos
, &pos
);
952 ret
= security_kernel_post_read_file(file
, *buf
, i_size
, id
);
958 if (id
!= READING_FIRMWARE_PREALLOC_BUFFER
) {
965 allow_write_access(file
);
968 EXPORT_SYMBOL_GPL(kernel_read_file
);
970 int kernel_read_file_from_path(const char *path
, void **buf
, loff_t
*size
,
971 loff_t max_size
, enum kernel_read_file_id id
)
979 file
= filp_open(path
, O_RDONLY
, 0);
981 return PTR_ERR(file
);
983 ret
= kernel_read_file(file
, buf
, size
, max_size
, id
);
987 EXPORT_SYMBOL_GPL(kernel_read_file_from_path
);
989 int kernel_read_file_from_fd(int fd
, void **buf
, loff_t
*size
, loff_t max_size
,
990 enum kernel_read_file_id id
)
992 struct fd f
= fdget(fd
);
998 ret
= kernel_read_file(f
.file
, buf
, size
, max_size
, id
);
1003 EXPORT_SYMBOL_GPL(kernel_read_file_from_fd
);
1005 ssize_t
read_code(struct file
*file
, unsigned long addr
, loff_t pos
, size_t len
)
1007 ssize_t res
= vfs_read(file
, (void __user
*)addr
, len
, &pos
);
1009 flush_icache_range(addr
, addr
+ len
);
1012 EXPORT_SYMBOL(read_code
);
1014 static int exec_mmap(struct mm_struct
*mm
)
1016 struct task_struct
*tsk
;
1017 struct mm_struct
*old_mm
, *active_mm
;
1019 /* Notify parent that we're no longer interested in the old VM */
1021 old_mm
= current
->mm
;
1022 mm_release(tsk
, old_mm
);
1025 sync_mm_rss(old_mm
);
1027 * Make sure that if there is a core dump in progress
1028 * for the old mm, we get out and die instead of going
1029 * through with the exec. We must hold mmap_sem around
1030 * checking core_state and changing tsk->mm.
1032 down_read(&old_mm
->mmap_sem
);
1033 if (unlikely(old_mm
->core_state
)) {
1034 up_read(&old_mm
->mmap_sem
);
1039 active_mm
= tsk
->active_mm
;
1041 tsk
->active_mm
= mm
;
1042 activate_mm(active_mm
, mm
);
1043 tsk
->mm
->vmacache_seqnum
= 0;
1044 vmacache_flush(tsk
);
1047 up_read(&old_mm
->mmap_sem
);
1048 BUG_ON(active_mm
!= old_mm
);
1049 setmax_mm_hiwater_rss(&tsk
->signal
->maxrss
, old_mm
);
1050 mm_update_next_owner(old_mm
);
1059 * This function makes sure the current process has its own signal table,
1060 * so that flush_signal_handlers can later reset the handlers without
1061 * disturbing other processes. (Other processes might share the signal
1062 * table via the CLONE_SIGHAND option to clone().)
1064 static int de_thread(struct task_struct
*tsk
)
1066 struct signal_struct
*sig
= tsk
->signal
;
1067 struct sighand_struct
*oldsighand
= tsk
->sighand
;
1068 spinlock_t
*lock
= &oldsighand
->siglock
;
1070 if (thread_group_empty(tsk
))
1071 goto no_thread_group
;
1074 * Kill all other threads in the thread group.
1076 spin_lock_irq(lock
);
1077 if (signal_group_exit(sig
)) {
1079 * Another group action in progress, just
1080 * return so that the signal is processed.
1082 spin_unlock_irq(lock
);
1086 sig
->group_exit_task
= tsk
;
1087 sig
->notify_count
= zap_other_threads(tsk
);
1088 if (!thread_group_leader(tsk
))
1089 sig
->notify_count
--;
1091 while (sig
->notify_count
) {
1092 __set_current_state(TASK_KILLABLE
);
1093 spin_unlock_irq(lock
);
1095 if (unlikely(__fatal_signal_pending(tsk
)))
1097 spin_lock_irq(lock
);
1099 spin_unlock_irq(lock
);
1102 * At this point all other threads have exited, all we have to
1103 * do is to wait for the thread group leader to become inactive,
1104 * and to assume its PID:
1106 if (!thread_group_leader(tsk
)) {
1107 struct task_struct
*leader
= tsk
->group_leader
;
1110 cgroup_threadgroup_change_begin(tsk
);
1111 write_lock_irq(&tasklist_lock
);
1113 * Do this under tasklist_lock to ensure that
1114 * exit_notify() can't miss ->group_exit_task
1116 sig
->notify_count
= -1;
1117 if (likely(leader
->exit_state
))
1119 __set_current_state(TASK_KILLABLE
);
1120 write_unlock_irq(&tasklist_lock
);
1121 cgroup_threadgroup_change_end(tsk
);
1123 if (unlikely(__fatal_signal_pending(tsk
)))
1128 * The only record we have of the real-time age of a
1129 * process, regardless of execs it's done, is start_time.
1130 * All the past CPU time is accumulated in signal_struct
1131 * from sister threads now dead. But in this non-leader
1132 * exec, nothing survives from the original leader thread,
1133 * whose birth marks the true age of this process now.
1134 * When we take on its identity by switching to its PID, we
1135 * also take its birthdate (always earlier than our own).
1137 tsk
->start_time
= leader
->start_time
;
1138 tsk
->real_start_time
= leader
->real_start_time
;
1140 BUG_ON(!same_thread_group(leader
, tsk
));
1141 BUG_ON(has_group_leader_pid(tsk
));
1143 * An exec() starts a new thread group with the
1144 * TGID of the previous thread group. Rehash the
1145 * two threads with a switched PID, and release
1146 * the former thread group leader:
1149 /* Become a process group leader with the old leader's pid.
1150 * The old leader becomes a thread of the this thread group.
1151 * Note: The old leader also uses this pid until release_task
1152 * is called. Odd but simple and correct.
1154 tsk
->pid
= leader
->pid
;
1155 change_pid(tsk
, PIDTYPE_PID
, task_pid(leader
));
1156 transfer_pid(leader
, tsk
, PIDTYPE_PGID
);
1157 transfer_pid(leader
, tsk
, PIDTYPE_SID
);
1159 list_replace_rcu(&leader
->tasks
, &tsk
->tasks
);
1160 list_replace_init(&leader
->sibling
, &tsk
->sibling
);
1162 tsk
->group_leader
= tsk
;
1163 leader
->group_leader
= tsk
;
1165 tsk
->exit_signal
= SIGCHLD
;
1166 leader
->exit_signal
= -1;
1168 BUG_ON(leader
->exit_state
!= EXIT_ZOMBIE
);
1169 leader
->exit_state
= EXIT_DEAD
;
1172 * We are going to release_task()->ptrace_unlink() silently,
1173 * the tracer can sleep in do_wait(). EXIT_DEAD guarantees
1174 * the tracer wont't block again waiting for this thread.
1176 if (unlikely(leader
->ptrace
))
1177 __wake_up_parent(leader
, leader
->parent
);
1178 write_unlock_irq(&tasklist_lock
);
1179 cgroup_threadgroup_change_end(tsk
);
1181 release_task(leader
);
1184 sig
->group_exit_task
= NULL
;
1185 sig
->notify_count
= 0;
1188 /* we have changed execution domain */
1189 tsk
->exit_signal
= SIGCHLD
;
1191 #ifdef CONFIG_POSIX_TIMERS
1193 flush_itimer_signals();
1196 if (atomic_read(&oldsighand
->count
) != 1) {
1197 struct sighand_struct
*newsighand
;
1199 * This ->sighand is shared with the CLONE_SIGHAND
1200 * but not CLONE_THREAD task, switch to the new one.
1202 newsighand
= kmem_cache_alloc(sighand_cachep
, GFP_KERNEL
);
1206 atomic_set(&newsighand
->count
, 1);
1207 memcpy(newsighand
->action
, oldsighand
->action
,
1208 sizeof(newsighand
->action
));
1210 write_lock_irq(&tasklist_lock
);
1211 spin_lock(&oldsighand
->siglock
);
1212 rcu_assign_pointer(tsk
->sighand
, newsighand
);
1213 spin_unlock(&oldsighand
->siglock
);
1214 write_unlock_irq(&tasklist_lock
);
1216 __cleanup_sighand(oldsighand
);
1219 BUG_ON(!thread_group_leader(tsk
));
1223 /* protects against exit_notify() and __exit_signal() */
1224 read_lock(&tasklist_lock
);
1225 sig
->group_exit_task
= NULL
;
1226 sig
->notify_count
= 0;
1227 read_unlock(&tasklist_lock
);
1231 char *__get_task_comm(char *buf
, size_t buf_size
, struct task_struct
*tsk
)
1234 strncpy(buf
, tsk
->comm
, buf_size
);
1238 EXPORT_SYMBOL_GPL(__get_task_comm
);
1241 * These functions flushes out all traces of the currently running executable
1242 * so that a new one can be started
1245 void __set_task_comm(struct task_struct
*tsk
, const char *buf
, bool exec
)
1248 trace_task_rename(tsk
, buf
);
1249 strlcpy(tsk
->comm
, buf
, sizeof(tsk
->comm
));
1251 perf_event_comm(tsk
, exec
);
1255 * Calling this is the point of no return. None of the failures will be
1256 * seen by userspace since either the process is already taking a fatal
1257 * signal (via de_thread() or coredump), or will have SEGV raised
1258 * (after exec_mmap()) by search_binary_handlers (see below).
1260 int flush_old_exec(struct linux_binprm
* bprm
)
1265 * Make sure we have a private signal table and that
1266 * we are unassociated from the previous thread group.
1268 retval
= de_thread(current
);
1273 * Must be called _before_ exec_mmap() as bprm->mm is
1274 * not visibile until then. This also enables the update
1277 set_mm_exe_file(bprm
->mm
, bprm
->file
);
1280 * Release all of the old mmap stuff
1282 acct_arg_size(bprm
, 0);
1283 retval
= exec_mmap(bprm
->mm
);
1288 * After clearing bprm->mm (to mark that current is using the
1289 * prepared mm now), we have nothing left of the original
1290 * process. If anything from here on returns an error, the check
1291 * in search_binary_handler() will SEGV current.
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
)
1339 * Once here, prepare_binrpm() will not be called any more, so
1340 * the final state of setuid/setgid/fscaps can be merged into the
1343 bprm
->secureexec
|= bprm
->cap_elevated
;
1345 if (bprm
->secureexec
) {
1346 /* Make sure parent cannot signal privileged process. */
1347 current
->pdeath_signal
= 0;
1350 * For secureexec, reset the stack limit to sane default to
1351 * avoid bad behavior from the prior rlimits. This has to
1352 * happen before arch_pick_mmap_layout(), which examines
1353 * RLIMIT_STACK, but after the point of no return to avoid
1354 * needing to clean up the change on failure.
1356 if (current
->signal
->rlim
[RLIMIT_STACK
].rlim_cur
> _STK_LIM
)
1357 current
->signal
->rlim
[RLIMIT_STACK
].rlim_cur
= _STK_LIM
;
1360 arch_pick_mmap_layout(current
->mm
);
1362 current
->sas_ss_sp
= current
->sas_ss_size
= 0;
1365 * Figure out dumpability. Note that this checking only of current
1366 * is wrong, but userspace depends on it. This should be testing
1367 * bprm->secureexec instead.
1369 if (bprm
->interp_flags
& BINPRM_FLAGS_ENFORCE_NONDUMP
||
1370 !(uid_eq(current_euid(), current_uid()) &&
1371 gid_eq(current_egid(), current_gid())))
1372 set_dumpable(current
->mm
, suid_dumpable
);
1374 set_dumpable(current
->mm
, SUID_DUMP_USER
);
1376 arch_setup_new_exec();
1378 __set_task_comm(current
, kbasename(bprm
->filename
), true);
1380 /* Set the new mm task size. We have to do that late because it may
1381 * depend on TIF_32BIT which is only updated in flush_thread() on
1382 * some architectures like powerpc
1384 current
->mm
->task_size
= TASK_SIZE
;
1386 /* An exec changes our domain. We are no longer part of the thread
1388 current
->self_exec_id
++;
1389 flush_signal_handlers(current
, 0);
1391 EXPORT_SYMBOL(setup_new_exec
);
1394 * Prepare credentials and lock ->cred_guard_mutex.
1395 * install_exec_creds() commits the new creds and drops the lock.
1396 * Or, if exec fails before, free_bprm() should release ->cred and
1399 int prepare_bprm_creds(struct linux_binprm
*bprm
)
1401 if (mutex_lock_interruptible(¤t
->signal
->cred_guard_mutex
))
1402 return -ERESTARTNOINTR
;
1404 bprm
->cred
= prepare_exec_creds();
1405 if (likely(bprm
->cred
))
1408 mutex_unlock(¤t
->signal
->cred_guard_mutex
);
1412 static void free_bprm(struct linux_binprm
*bprm
)
1414 free_arg_pages(bprm
);
1416 mutex_unlock(¤t
->signal
->cred_guard_mutex
);
1417 abort_creds(bprm
->cred
);
1420 allow_write_access(bprm
->file
);
1423 /* If a binfmt changed the interp, free it. */
1424 if (bprm
->interp
!= bprm
->filename
)
1425 kfree(bprm
->interp
);
1429 int bprm_change_interp(const char *interp
, struct linux_binprm
*bprm
)
1431 /* If a binfmt changed the interp, free it first. */
1432 if (bprm
->interp
!= bprm
->filename
)
1433 kfree(bprm
->interp
);
1434 bprm
->interp
= kstrdup(interp
, GFP_KERNEL
);
1439 EXPORT_SYMBOL(bprm_change_interp
);
1442 * install the new credentials for this executable
1444 void install_exec_creds(struct linux_binprm
*bprm
)
1446 security_bprm_committing_creds(bprm
);
1448 commit_creds(bprm
->cred
);
1452 * Disable monitoring for regular users
1453 * when executing setuid binaries. Must
1454 * wait until new credentials are committed
1455 * by commit_creds() above
1457 if (get_dumpable(current
->mm
) != SUID_DUMP_USER
)
1458 perf_event_exit_task(current
);
1460 * cred_guard_mutex must be held at least to this point to prevent
1461 * ptrace_attach() from altering our determination of the task's
1462 * credentials; any time after this it may be unlocked.
1464 security_bprm_committed_creds(bprm
);
1465 mutex_unlock(¤t
->signal
->cred_guard_mutex
);
1467 EXPORT_SYMBOL(install_exec_creds
);
1470 * determine how safe it is to execute the proposed program
1471 * - the caller must hold ->cred_guard_mutex to protect against
1472 * PTRACE_ATTACH or seccomp thread-sync
1474 static void check_unsafe_exec(struct linux_binprm
*bprm
)
1476 struct task_struct
*p
= current
, *t
;
1480 bprm
->unsafe
|= LSM_UNSAFE_PTRACE
;
1483 * This isn't strictly necessary, but it makes it harder for LSMs to
1486 if (task_no_new_privs(current
))
1487 bprm
->unsafe
|= LSM_UNSAFE_NO_NEW_PRIVS
;
1491 spin_lock(&p
->fs
->lock
);
1493 while_each_thread(p
, t
) {
1499 if (p
->fs
->users
> n_fs
)
1500 bprm
->unsafe
|= LSM_UNSAFE_SHARE
;
1503 spin_unlock(&p
->fs
->lock
);
1506 static void bprm_fill_uid(struct linux_binprm
*bprm
)
1508 struct inode
*inode
;
1514 * Since this can be called multiple times (via prepare_binprm),
1515 * we must clear any previous work done when setting set[ug]id
1516 * bits from any earlier bprm->file uses (for example when run
1517 * first for a setuid script then again for its interpreter).
1519 bprm
->cred
->euid
= current_euid();
1520 bprm
->cred
->egid
= current_egid();
1522 if (path_nosuid(&bprm
->file
->f_path
))
1525 if (task_no_new_privs(current
))
1528 inode
= bprm
->file
->f_path
.dentry
->d_inode
;
1529 mode
= READ_ONCE(inode
->i_mode
);
1530 if (!(mode
& (S_ISUID
|S_ISGID
)))
1533 /* Be careful if suid/sgid is set */
1536 /* reload atomically mode/uid/gid now that lock held */
1537 mode
= inode
->i_mode
;
1540 inode_unlock(inode
);
1542 /* We ignore suid/sgid if there are no mappings for them in the ns */
1543 if (!kuid_has_mapping(bprm
->cred
->user_ns
, uid
) ||
1544 !kgid_has_mapping(bprm
->cred
->user_ns
, gid
))
1547 if (mode
& S_ISUID
) {
1548 bprm
->per_clear
|= PER_CLEAR_ON_SETID
;
1549 bprm
->cred
->euid
= uid
;
1552 if ((mode
& (S_ISGID
| S_IXGRP
)) == (S_ISGID
| S_IXGRP
)) {
1553 bprm
->per_clear
|= PER_CLEAR_ON_SETID
;
1554 bprm
->cred
->egid
= gid
;
1559 * Fill the binprm structure from the inode.
1560 * Check permissions, then read the first 128 (BINPRM_BUF_SIZE) bytes
1562 * This may be called multiple times for binary chains (scripts for example).
1564 int prepare_binprm(struct linux_binprm
*bprm
)
1569 bprm_fill_uid(bprm
);
1571 /* fill in binprm security blob */
1572 retval
= security_bprm_set_creds(bprm
);
1575 bprm
->called_set_creds
= 1;
1577 memset(bprm
->buf
, 0, BINPRM_BUF_SIZE
);
1578 return kernel_read(bprm
->file
, bprm
->buf
, BINPRM_BUF_SIZE
, &pos
);
1581 EXPORT_SYMBOL(prepare_binprm
);
1584 * Arguments are '\0' separated strings found at the location bprm->p
1585 * points to; chop off the first by relocating brpm->p to right after
1586 * the first '\0' encountered.
1588 int remove_arg_zero(struct linux_binprm
*bprm
)
1591 unsigned long offset
;
1599 offset
= bprm
->p
& ~PAGE_MASK
;
1600 page
= get_arg_page(bprm
, bprm
->p
, 0);
1605 kaddr
= kmap_atomic(page
);
1607 for (; offset
< PAGE_SIZE
&& kaddr
[offset
];
1608 offset
++, bprm
->p
++)
1611 kunmap_atomic(kaddr
);
1613 } while (offset
== PAGE_SIZE
);
1622 EXPORT_SYMBOL(remove_arg_zero
);
1624 #define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e))
1626 * cycle the list of binary formats handler, until one recognizes the image
1628 int search_binary_handler(struct linux_binprm
*bprm
)
1630 bool need_retry
= IS_ENABLED(CONFIG_MODULES
);
1631 struct linux_binfmt
*fmt
;
1634 /* This allows 4 levels of binfmt rewrites before failing hard. */
1635 if (bprm
->recursion_depth
> 5)
1638 retval
= security_bprm_check(bprm
);
1644 read_lock(&binfmt_lock
);
1645 list_for_each_entry(fmt
, &formats
, lh
) {
1646 if (!try_module_get(fmt
->module
))
1648 read_unlock(&binfmt_lock
);
1649 bprm
->recursion_depth
++;
1650 retval
= fmt
->load_binary(bprm
);
1651 read_lock(&binfmt_lock
);
1653 bprm
->recursion_depth
--;
1654 if (retval
< 0 && !bprm
->mm
) {
1655 /* we got to flush_old_exec() and failed after it */
1656 read_unlock(&binfmt_lock
);
1657 force_sigsegv(SIGSEGV
, current
);
1660 if (retval
!= -ENOEXEC
|| !bprm
->file
) {
1661 read_unlock(&binfmt_lock
);
1665 read_unlock(&binfmt_lock
);
1668 if (printable(bprm
->buf
[0]) && printable(bprm
->buf
[1]) &&
1669 printable(bprm
->buf
[2]) && printable(bprm
->buf
[3]))
1671 if (request_module("binfmt-%04x", *(ushort
*)(bprm
->buf
+ 2)) < 0)
1679 EXPORT_SYMBOL(search_binary_handler
);
1681 static int exec_binprm(struct linux_binprm
*bprm
)
1683 pid_t old_pid
, old_vpid
;
1686 /* Need to fetch pid before load_binary changes it */
1687 old_pid
= current
->pid
;
1689 old_vpid
= task_pid_nr_ns(current
, task_active_pid_ns(current
->parent
));
1692 ret
= search_binary_handler(bprm
);
1695 trace_sched_process_exec(current
, old_pid
, bprm
);
1696 ptrace_event(PTRACE_EVENT_EXEC
, old_vpid
);
1697 proc_exec_connector(current
);
1704 * sys_execve() executes a new program.
1706 static int do_execveat_common(int fd
, struct filename
*filename
,
1707 struct user_arg_ptr argv
,
1708 struct user_arg_ptr envp
,
1711 char *pathbuf
= NULL
;
1712 struct linux_binprm
*bprm
;
1714 struct files_struct
*displaced
;
1717 if (IS_ERR(filename
))
1718 return PTR_ERR(filename
);
1721 * We move the actual failure in case of RLIMIT_NPROC excess from
1722 * set*uid() to execve() because too many poorly written programs
1723 * don't check setuid() return code. Here we additionally recheck
1724 * whether NPROC limit is still exceeded.
1726 if ((current
->flags
& PF_NPROC_EXCEEDED
) &&
1727 atomic_read(¤t_user()->processes
) > rlimit(RLIMIT_NPROC
)) {
1732 /* We're below the limit (still or again), so we don't want to make
1733 * further execve() calls fail. */
1734 current
->flags
&= ~PF_NPROC_EXCEEDED
;
1736 retval
= unshare_files(&displaced
);
1741 bprm
= kzalloc(sizeof(*bprm
), GFP_KERNEL
);
1745 retval
= prepare_bprm_creds(bprm
);
1749 check_unsafe_exec(bprm
);
1750 current
->in_execve
= 1;
1752 file
= do_open_execat(fd
, filename
, flags
);
1753 retval
= PTR_ERR(file
);
1760 if (fd
== AT_FDCWD
|| filename
->name
[0] == '/') {
1761 bprm
->filename
= filename
->name
;
1763 if (filename
->name
[0] == '\0')
1764 pathbuf
= kasprintf(GFP_KERNEL
, "/dev/fd/%d", fd
);
1766 pathbuf
= kasprintf(GFP_KERNEL
, "/dev/fd/%d/%s",
1767 fd
, filename
->name
);
1773 * Record that a name derived from an O_CLOEXEC fd will be
1774 * inaccessible after exec. Relies on having exclusive access to
1775 * current->files (due to unshare_files above).
1777 if (close_on_exec(fd
, rcu_dereference_raw(current
->files
->fdt
)))
1778 bprm
->interp_flags
|= BINPRM_FLAGS_PATH_INACCESSIBLE
;
1779 bprm
->filename
= pathbuf
;
1781 bprm
->interp
= bprm
->filename
;
1783 retval
= bprm_mm_init(bprm
);
1787 bprm
->argc
= count(argv
, MAX_ARG_STRINGS
);
1788 if ((retval
= bprm
->argc
) < 0)
1791 bprm
->envc
= count(envp
, MAX_ARG_STRINGS
);
1792 if ((retval
= bprm
->envc
) < 0)
1795 retval
= prepare_binprm(bprm
);
1799 retval
= copy_strings_kernel(1, &bprm
->filename
, bprm
);
1803 bprm
->exec
= bprm
->p
;
1804 retval
= copy_strings(bprm
->envc
, envp
, bprm
);
1808 retval
= copy_strings(bprm
->argc
, argv
, bprm
);
1812 would_dump(bprm
, bprm
->file
);
1814 retval
= exec_binprm(bprm
);
1818 /* execve succeeded */
1819 current
->fs
->in_exec
= 0;
1820 current
->in_execve
= 0;
1821 membarrier_execve(current
);
1822 acct_update_integrals(current
);
1823 task_numa_free(current
);
1828 put_files_struct(displaced
);
1833 acct_arg_size(bprm
, 0);
1838 current
->fs
->in_exec
= 0;
1839 current
->in_execve
= 0;
1847 reset_files_struct(displaced
);
1853 int do_execve(struct filename
*filename
,
1854 const char __user
*const __user
*__argv
,
1855 const char __user
*const __user
*__envp
)
1857 struct user_arg_ptr argv
= { .ptr
.native
= __argv
};
1858 struct user_arg_ptr envp
= { .ptr
.native
= __envp
};
1859 return do_execveat_common(AT_FDCWD
, filename
, argv
, envp
, 0);
1862 int do_execveat(int fd
, struct filename
*filename
,
1863 const char __user
*const __user
*__argv
,
1864 const char __user
*const __user
*__envp
,
1867 struct user_arg_ptr argv
= { .ptr
.native
= __argv
};
1868 struct user_arg_ptr envp
= { .ptr
.native
= __envp
};
1870 return do_execveat_common(fd
, filename
, argv
, envp
, flags
);
1873 #ifdef CONFIG_COMPAT
1874 static int compat_do_execve(struct filename
*filename
,
1875 const compat_uptr_t __user
*__argv
,
1876 const compat_uptr_t __user
*__envp
)
1878 struct user_arg_ptr argv
= {
1880 .ptr
.compat
= __argv
,
1882 struct user_arg_ptr envp
= {
1884 .ptr
.compat
= __envp
,
1886 return do_execveat_common(AT_FDCWD
, filename
, argv
, envp
, 0);
1889 static int compat_do_execveat(int fd
, struct filename
*filename
,
1890 const compat_uptr_t __user
*__argv
,
1891 const compat_uptr_t __user
*__envp
,
1894 struct user_arg_ptr argv
= {
1896 .ptr
.compat
= __argv
,
1898 struct user_arg_ptr envp
= {
1900 .ptr
.compat
= __envp
,
1902 return do_execveat_common(fd
, filename
, argv
, envp
, flags
);
1906 void set_binfmt(struct linux_binfmt
*new)
1908 struct mm_struct
*mm
= current
->mm
;
1911 module_put(mm
->binfmt
->module
);
1915 __module_get(new->module
);
1917 EXPORT_SYMBOL(set_binfmt
);
1920 * set_dumpable stores three-value SUID_DUMP_* into mm->flags.
1922 void set_dumpable(struct mm_struct
*mm
, int value
)
1924 unsigned long old
, new;
1926 if (WARN_ON((unsigned)value
> SUID_DUMP_ROOT
))
1930 old
= READ_ONCE(mm
->flags
);
1931 new = (old
& ~MMF_DUMPABLE_MASK
) | value
;
1932 } while (cmpxchg(&mm
->flags
, old
, new) != old
);
1935 SYSCALL_DEFINE3(execve
,
1936 const char __user
*, filename
,
1937 const char __user
*const __user
*, argv
,
1938 const char __user
*const __user
*, envp
)
1940 return do_execve(getname(filename
), argv
, envp
);
1943 SYSCALL_DEFINE5(execveat
,
1944 int, fd
, const char __user
*, filename
,
1945 const char __user
*const __user
*, argv
,
1946 const char __user
*const __user
*, envp
,
1949 int lookup_flags
= (flags
& AT_EMPTY_PATH
) ? LOOKUP_EMPTY
: 0;
1951 return do_execveat(fd
,
1952 getname_flags(filename
, lookup_flags
, NULL
),
1956 #ifdef CONFIG_COMPAT
1957 COMPAT_SYSCALL_DEFINE3(execve
, const char __user
*, filename
,
1958 const compat_uptr_t __user
*, argv
,
1959 const compat_uptr_t __user
*, envp
)
1961 return compat_do_execve(getname(filename
), argv
, envp
);
1964 COMPAT_SYSCALL_DEFINE5(execveat
, int, fd
,
1965 const char __user
*, filename
,
1966 const compat_uptr_t __user
*, argv
,
1967 const compat_uptr_t __user
*, envp
,
1970 int lookup_flags
= (flags
& AT_EMPTY_PATH
) ? LOOKUP_EMPTY
: 0;
1972 return compat_do_execveat(fd
,
1973 getname_flags(filename
, lookup_flags
, NULL
),