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
= vm_area_alloc(mm
);
307 if (down_write_killable(&mm
->mmap_sem
)) {
313 * Place the stack at the largest stack address the architecture
314 * supports. Later, we'll move this to an appropriate place. We don't
315 * use STACK_TOP because that can depend on attributes which aren't
318 BUILD_BUG_ON(VM_STACK_FLAGS
& VM_STACK_INCOMPLETE_SETUP
);
319 vma
->vm_end
= STACK_TOP_MAX
;
320 vma
->vm_start
= vma
->vm_end
- PAGE_SIZE
;
321 vma
->vm_flags
= VM_SOFTDIRTY
| VM_STACK_FLAGS
| VM_STACK_INCOMPLETE_SETUP
;
322 vma
->vm_page_prot
= vm_get_page_prot(vma
->vm_flags
);
324 err
= insert_vm_struct(mm
, vma
);
328 mm
->stack_vm
= mm
->total_vm
= 1;
329 arch_bprm_mm_init(mm
, vma
);
330 up_write(&mm
->mmap_sem
);
331 bprm
->p
= vma
->vm_end
- sizeof(void *);
334 up_write(&mm
->mmap_sem
);
341 static bool valid_arg_len(struct linux_binprm
*bprm
, long len
)
343 return len
<= MAX_ARG_STRLEN
;
348 static inline void acct_arg_size(struct linux_binprm
*bprm
, unsigned long pages
)
352 static struct page
*get_arg_page(struct linux_binprm
*bprm
, unsigned long pos
,
357 page
= bprm
->page
[pos
/ PAGE_SIZE
];
358 if (!page
&& write
) {
359 page
= alloc_page(GFP_HIGHUSER
|__GFP_ZERO
);
362 bprm
->page
[pos
/ PAGE_SIZE
] = page
;
368 static void put_arg_page(struct page
*page
)
372 static void free_arg_page(struct linux_binprm
*bprm
, int i
)
375 __free_page(bprm
->page
[i
]);
376 bprm
->page
[i
] = NULL
;
380 static void free_arg_pages(struct linux_binprm
*bprm
)
384 for (i
= 0; i
< MAX_ARG_PAGES
; i
++)
385 free_arg_page(bprm
, i
);
388 static void flush_arg_page(struct linux_binprm
*bprm
, unsigned long pos
,
393 static int __bprm_mm_init(struct linux_binprm
*bprm
)
395 bprm
->p
= PAGE_SIZE
* MAX_ARG_PAGES
- sizeof(void *);
399 static bool valid_arg_len(struct linux_binprm
*bprm
, long len
)
401 return len
<= bprm
->p
;
404 #endif /* CONFIG_MMU */
407 * Create a new mm_struct and populate it with a temporary stack
408 * vm_area_struct. We don't have enough context at this point to set the stack
409 * flags, permissions, and offset, so we use temporary values. We'll update
410 * them later in setup_arg_pages().
412 static int bprm_mm_init(struct linux_binprm
*bprm
)
415 struct mm_struct
*mm
= NULL
;
417 bprm
->mm
= mm
= mm_alloc();
422 err
= __bprm_mm_init(bprm
);
437 struct user_arg_ptr
{
442 const char __user
*const __user
*native
;
444 const compat_uptr_t __user
*compat
;
449 static const char __user
*get_user_arg_ptr(struct user_arg_ptr argv
, int nr
)
451 const char __user
*native
;
454 if (unlikely(argv
.is_compat
)) {
455 compat_uptr_t compat
;
457 if (get_user(compat
, argv
.ptr
.compat
+ nr
))
458 return ERR_PTR(-EFAULT
);
460 return compat_ptr(compat
);
464 if (get_user(native
, argv
.ptr
.native
+ nr
))
465 return ERR_PTR(-EFAULT
);
471 * count() counts the number of strings in array ARGV.
473 static int count(struct user_arg_ptr argv
, int max
)
477 if (argv
.ptr
.native
!= NULL
) {
479 const char __user
*p
= get_user_arg_ptr(argv
, i
);
491 if (fatal_signal_pending(current
))
492 return -ERESTARTNOHAND
;
500 * 'copy_strings()' copies argument/environment strings from the old
501 * processes's memory to the new process's stack. The call to get_user_pages()
502 * ensures the destination page is created and not swapped out.
504 static int copy_strings(int argc
, struct user_arg_ptr argv
,
505 struct linux_binprm
*bprm
)
507 struct page
*kmapped_page
= NULL
;
509 unsigned long kpos
= 0;
513 const char __user
*str
;
518 str
= get_user_arg_ptr(argv
, argc
);
522 len
= strnlen_user(str
, MAX_ARG_STRLEN
);
527 if (!valid_arg_len(bprm
, len
))
530 /* We're going to work our way backwords. */
536 int offset
, bytes_to_copy
;
538 if (fatal_signal_pending(current
)) {
539 ret
= -ERESTARTNOHAND
;
544 offset
= pos
% PAGE_SIZE
;
548 bytes_to_copy
= offset
;
549 if (bytes_to_copy
> len
)
552 offset
-= bytes_to_copy
;
553 pos
-= bytes_to_copy
;
554 str
-= bytes_to_copy
;
555 len
-= bytes_to_copy
;
557 if (!kmapped_page
|| kpos
!= (pos
& PAGE_MASK
)) {
560 page
= get_arg_page(bprm
, pos
, 1);
567 flush_kernel_dcache_page(kmapped_page
);
568 kunmap(kmapped_page
);
569 put_arg_page(kmapped_page
);
572 kaddr
= kmap(kmapped_page
);
573 kpos
= pos
& PAGE_MASK
;
574 flush_arg_page(bprm
, kpos
, kmapped_page
);
576 if (copy_from_user(kaddr
+offset
, str
, bytes_to_copy
)) {
585 flush_kernel_dcache_page(kmapped_page
);
586 kunmap(kmapped_page
);
587 put_arg_page(kmapped_page
);
593 * Like copy_strings, but get argv and its values from kernel memory.
595 int copy_strings_kernel(int argc
, const char *const *__argv
,
596 struct linux_binprm
*bprm
)
599 mm_segment_t oldfs
= get_fs();
600 struct user_arg_ptr argv
= {
601 .ptr
.native
= (const char __user
*const __user
*)__argv
,
605 r
= copy_strings(argc
, argv
, bprm
);
610 EXPORT_SYMBOL(copy_strings_kernel
);
615 * During bprm_mm_init(), we create a temporary stack at STACK_TOP_MAX. Once
616 * the binfmt code determines where the new stack should reside, we shift it to
617 * its final location. The process proceeds as follows:
619 * 1) Use shift to calculate the new vma endpoints.
620 * 2) Extend vma to cover both the old and new ranges. This ensures the
621 * arguments passed to subsequent functions are consistent.
622 * 3) Move vma's page tables to the new range.
623 * 4) Free up any cleared pgd range.
624 * 5) Shrink the vma to cover only the new range.
626 static int shift_arg_pages(struct vm_area_struct
*vma
, unsigned long shift
)
628 struct mm_struct
*mm
= vma
->vm_mm
;
629 unsigned long old_start
= vma
->vm_start
;
630 unsigned long old_end
= vma
->vm_end
;
631 unsigned long length
= old_end
- old_start
;
632 unsigned long new_start
= old_start
- shift
;
633 unsigned long new_end
= old_end
- shift
;
634 struct mmu_gather tlb
;
636 BUG_ON(new_start
> new_end
);
639 * ensure there are no vmas between where we want to go
642 if (vma
!= find_vma(mm
, new_start
))
646 * cover the whole range: [new_start, old_end)
648 if (vma_adjust(vma
, new_start
, old_end
, vma
->vm_pgoff
, NULL
))
652 * move the page tables downwards, on failure we rely on
653 * process cleanup to remove whatever mess we made.
655 if (length
!= move_page_tables(vma
, old_start
,
656 vma
, new_start
, length
, false))
660 tlb_gather_mmu(&tlb
, mm
, old_start
, old_end
);
661 if (new_end
> old_start
) {
663 * when the old and new regions overlap clear from new_end.
665 free_pgd_range(&tlb
, new_end
, old_end
, new_end
,
666 vma
->vm_next
? vma
->vm_next
->vm_start
: USER_PGTABLES_CEILING
);
669 * otherwise, clean from old_start; this is done to not touch
670 * the address space in [new_end, old_start) some architectures
671 * have constraints on va-space that make this illegal (IA64) -
672 * for the others its just a little faster.
674 free_pgd_range(&tlb
, old_start
, old_end
, new_end
,
675 vma
->vm_next
? vma
->vm_next
->vm_start
: USER_PGTABLES_CEILING
);
677 tlb_finish_mmu(&tlb
, old_start
, old_end
);
680 * Shrink the vma to just the new range. Always succeeds.
682 vma_adjust(vma
, new_start
, new_end
, vma
->vm_pgoff
, NULL
);
688 * Finalizes the stack vm_area_struct. The flags and permissions are updated,
689 * the stack is optionally relocated, and some extra space is added.
691 int setup_arg_pages(struct linux_binprm
*bprm
,
692 unsigned long stack_top
,
693 int executable_stack
)
696 unsigned long stack_shift
;
697 struct mm_struct
*mm
= current
->mm
;
698 struct vm_area_struct
*vma
= bprm
->vma
;
699 struct vm_area_struct
*prev
= NULL
;
700 unsigned long vm_flags
;
701 unsigned long stack_base
;
702 unsigned long stack_size
;
703 unsigned long stack_expand
;
704 unsigned long rlim_stack
;
706 #ifdef CONFIG_STACK_GROWSUP
707 /* Limit stack size */
708 stack_base
= rlimit_max(RLIMIT_STACK
);
709 if (stack_base
> STACK_SIZE_MAX
)
710 stack_base
= STACK_SIZE_MAX
;
712 /* Add space for stack randomization. */
713 stack_base
+= (STACK_RND_MASK
<< PAGE_SHIFT
);
715 /* Make sure we didn't let the argument array grow too large. */
716 if (vma
->vm_end
- vma
->vm_start
> stack_base
)
719 stack_base
= PAGE_ALIGN(stack_top
- stack_base
);
721 stack_shift
= vma
->vm_start
- stack_base
;
722 mm
->arg_start
= bprm
->p
- stack_shift
;
723 bprm
->p
= vma
->vm_end
- stack_shift
;
725 stack_top
= arch_align_stack(stack_top
);
726 stack_top
= PAGE_ALIGN(stack_top
);
728 if (unlikely(stack_top
< mmap_min_addr
) ||
729 unlikely(vma
->vm_end
- vma
->vm_start
>= stack_top
- mmap_min_addr
))
732 stack_shift
= vma
->vm_end
- stack_top
;
734 bprm
->p
-= stack_shift
;
735 mm
->arg_start
= bprm
->p
;
739 bprm
->loader
-= stack_shift
;
740 bprm
->exec
-= stack_shift
;
742 if (down_write_killable(&mm
->mmap_sem
))
745 vm_flags
= VM_STACK_FLAGS
;
748 * Adjust stack execute permissions; explicitly enable for
749 * EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X and leave alone
750 * (arch default) otherwise.
752 if (unlikely(executable_stack
== EXSTACK_ENABLE_X
))
754 else if (executable_stack
== EXSTACK_DISABLE_X
)
755 vm_flags
&= ~VM_EXEC
;
756 vm_flags
|= mm
->def_flags
;
757 vm_flags
|= VM_STACK_INCOMPLETE_SETUP
;
759 ret
= mprotect_fixup(vma
, &prev
, vma
->vm_start
, vma
->vm_end
,
765 /* Move stack pages down in memory. */
767 ret
= shift_arg_pages(vma
, stack_shift
);
772 /* mprotect_fixup is overkill to remove the temporary stack flags */
773 vma
->vm_flags
&= ~VM_STACK_INCOMPLETE_SETUP
;
775 stack_expand
= 131072UL; /* randomly 32*4k (or 2*64k) pages */
776 stack_size
= vma
->vm_end
- vma
->vm_start
;
778 * Align this down to a page boundary as expand_stack
781 rlim_stack
= rlimit(RLIMIT_STACK
) & PAGE_MASK
;
782 #ifdef CONFIG_STACK_GROWSUP
783 if (stack_size
+ stack_expand
> rlim_stack
)
784 stack_base
= vma
->vm_start
+ rlim_stack
;
786 stack_base
= vma
->vm_end
+ stack_expand
;
788 if (stack_size
+ stack_expand
> rlim_stack
)
789 stack_base
= vma
->vm_end
- rlim_stack
;
791 stack_base
= vma
->vm_start
- stack_expand
;
793 current
->mm
->start_stack
= bprm
->p
;
794 ret
= expand_stack(vma
, stack_base
);
799 up_write(&mm
->mmap_sem
);
802 EXPORT_SYMBOL(setup_arg_pages
);
807 * Transfer the program arguments and environment from the holding pages
808 * onto the stack. The provided stack pointer is adjusted accordingly.
810 int transfer_args_to_stack(struct linux_binprm
*bprm
,
811 unsigned long *sp_location
)
813 unsigned long index
, stop
, sp
;
816 stop
= bprm
->p
>> PAGE_SHIFT
;
819 for (index
= MAX_ARG_PAGES
- 1; index
>= stop
; index
--) {
820 unsigned int offset
= index
== stop
? bprm
->p
& ~PAGE_MASK
: 0;
821 char *src
= kmap(bprm
->page
[index
]) + offset
;
822 sp
-= PAGE_SIZE
- offset
;
823 if (copy_to_user((void *) sp
, src
, PAGE_SIZE
- offset
) != 0)
825 kunmap(bprm
->page
[index
]);
835 EXPORT_SYMBOL(transfer_args_to_stack
);
837 #endif /* CONFIG_MMU */
839 static struct file
*do_open_execat(int fd
, struct filename
*name
, int flags
)
843 struct open_flags open_exec_flags
= {
844 .open_flag
= O_LARGEFILE
| O_RDONLY
| __FMODE_EXEC
,
845 .acc_mode
= MAY_EXEC
,
846 .intent
= LOOKUP_OPEN
,
847 .lookup_flags
= LOOKUP_FOLLOW
,
850 if ((flags
& ~(AT_SYMLINK_NOFOLLOW
| AT_EMPTY_PATH
)) != 0)
851 return ERR_PTR(-EINVAL
);
852 if (flags
& AT_SYMLINK_NOFOLLOW
)
853 open_exec_flags
.lookup_flags
&= ~LOOKUP_FOLLOW
;
854 if (flags
& AT_EMPTY_PATH
)
855 open_exec_flags
.lookup_flags
|= LOOKUP_EMPTY
;
857 file
= do_filp_open(fd
, name
, &open_exec_flags
);
862 if (!S_ISREG(file_inode(file
)->i_mode
))
865 if (path_noexec(&file
->f_path
))
868 err
= deny_write_access(file
);
872 if (name
->name
[0] != '\0')
875 trace_open_exec(name
->name
);
885 struct file
*open_exec(const char *name
)
887 struct filename
*filename
= getname_kernel(name
);
888 struct file
*f
= ERR_CAST(filename
);
890 if (!IS_ERR(filename
)) {
891 f
= do_open_execat(AT_FDCWD
, filename
, 0);
896 EXPORT_SYMBOL(open_exec
);
898 int kernel_read_file(struct file
*file
, void **buf
, loff_t
*size
,
899 loff_t max_size
, enum kernel_read_file_id id
)
905 if (!S_ISREG(file_inode(file
)->i_mode
) || max_size
< 0)
908 ret
= security_kernel_read_file(file
, id
);
912 ret
= deny_write_access(file
);
916 i_size
= i_size_read(file_inode(file
));
917 if (max_size
> 0 && i_size
> max_size
) {
926 if (id
!= READING_FIRMWARE_PREALLOC_BUFFER
)
927 *buf
= vmalloc(i_size
);
934 while (pos
< i_size
) {
935 bytes
= kernel_read(file
, *buf
+ pos
, i_size
- pos
, &pos
);
950 ret
= security_kernel_post_read_file(file
, *buf
, i_size
, id
);
956 if (id
!= READING_FIRMWARE_PREALLOC_BUFFER
) {
963 allow_write_access(file
);
966 EXPORT_SYMBOL_GPL(kernel_read_file
);
968 int kernel_read_file_from_path(const char *path
, void **buf
, loff_t
*size
,
969 loff_t max_size
, enum kernel_read_file_id id
)
977 file
= filp_open(path
, O_RDONLY
, 0);
979 return PTR_ERR(file
);
981 ret
= kernel_read_file(file
, buf
, size
, max_size
, id
);
985 EXPORT_SYMBOL_GPL(kernel_read_file_from_path
);
987 int kernel_read_file_from_fd(int fd
, void **buf
, loff_t
*size
, loff_t max_size
,
988 enum kernel_read_file_id id
)
990 struct fd f
= fdget(fd
);
996 ret
= kernel_read_file(f
.file
, buf
, size
, max_size
, id
);
1001 EXPORT_SYMBOL_GPL(kernel_read_file_from_fd
);
1003 ssize_t
read_code(struct file
*file
, unsigned long addr
, loff_t pos
, size_t len
)
1005 ssize_t res
= vfs_read(file
, (void __user
*)addr
, len
, &pos
);
1007 flush_icache_range(addr
, addr
+ len
);
1010 EXPORT_SYMBOL(read_code
);
1012 static int exec_mmap(struct mm_struct
*mm
)
1014 struct task_struct
*tsk
;
1015 struct mm_struct
*old_mm
, *active_mm
;
1017 /* Notify parent that we're no longer interested in the old VM */
1019 old_mm
= current
->mm
;
1020 mm_release(tsk
, old_mm
);
1023 sync_mm_rss(old_mm
);
1025 * Make sure that if there is a core dump in progress
1026 * for the old mm, we get out and die instead of going
1027 * through with the exec. We must hold mmap_sem around
1028 * checking core_state and changing tsk->mm.
1030 down_read(&old_mm
->mmap_sem
);
1031 if (unlikely(old_mm
->core_state
)) {
1032 up_read(&old_mm
->mmap_sem
);
1037 active_mm
= tsk
->active_mm
;
1039 tsk
->active_mm
= mm
;
1040 activate_mm(active_mm
, mm
);
1041 tsk
->mm
->vmacache_seqnum
= 0;
1042 vmacache_flush(tsk
);
1045 up_read(&old_mm
->mmap_sem
);
1046 BUG_ON(active_mm
!= old_mm
);
1047 setmax_mm_hiwater_rss(&tsk
->signal
->maxrss
, old_mm
);
1048 mm_update_next_owner(old_mm
);
1057 * This function makes sure the current process has its own signal table,
1058 * so that flush_signal_handlers can later reset the handlers without
1059 * disturbing other processes. (Other processes might share the signal
1060 * table via the CLONE_SIGHAND option to clone().)
1062 static int de_thread(struct task_struct
*tsk
)
1064 struct signal_struct
*sig
= tsk
->signal
;
1065 struct sighand_struct
*oldsighand
= tsk
->sighand
;
1066 spinlock_t
*lock
= &oldsighand
->siglock
;
1068 if (thread_group_empty(tsk
))
1069 goto no_thread_group
;
1072 * Kill all other threads in the thread group.
1074 spin_lock_irq(lock
);
1075 if (signal_group_exit(sig
)) {
1077 * Another group action in progress, just
1078 * return so that the signal is processed.
1080 spin_unlock_irq(lock
);
1084 sig
->group_exit_task
= tsk
;
1085 sig
->notify_count
= zap_other_threads(tsk
);
1086 if (!thread_group_leader(tsk
))
1087 sig
->notify_count
--;
1089 while (sig
->notify_count
) {
1090 __set_current_state(TASK_KILLABLE
);
1091 spin_unlock_irq(lock
);
1093 if (unlikely(__fatal_signal_pending(tsk
)))
1095 spin_lock_irq(lock
);
1097 spin_unlock_irq(lock
);
1100 * At this point all other threads have exited, all we have to
1101 * do is to wait for the thread group leader to become inactive,
1102 * and to assume its PID:
1104 if (!thread_group_leader(tsk
)) {
1105 struct task_struct
*leader
= tsk
->group_leader
;
1108 cgroup_threadgroup_change_begin(tsk
);
1109 write_lock_irq(&tasklist_lock
);
1111 * Do this under tasklist_lock to ensure that
1112 * exit_notify() can't miss ->group_exit_task
1114 sig
->notify_count
= -1;
1115 if (likely(leader
->exit_state
))
1117 __set_current_state(TASK_KILLABLE
);
1118 write_unlock_irq(&tasklist_lock
);
1119 cgroup_threadgroup_change_end(tsk
);
1121 if (unlikely(__fatal_signal_pending(tsk
)))
1126 * The only record we have of the real-time age of a
1127 * process, regardless of execs it's done, is start_time.
1128 * All the past CPU time is accumulated in signal_struct
1129 * from sister threads now dead. But in this non-leader
1130 * exec, nothing survives from the original leader thread,
1131 * whose birth marks the true age of this process now.
1132 * When we take on its identity by switching to its PID, we
1133 * also take its birthdate (always earlier than our own).
1135 tsk
->start_time
= leader
->start_time
;
1136 tsk
->real_start_time
= leader
->real_start_time
;
1138 BUG_ON(!same_thread_group(leader
, tsk
));
1139 BUG_ON(has_group_leader_pid(tsk
));
1141 * An exec() starts a new thread group with the
1142 * TGID of the previous thread group. Rehash the
1143 * two threads with a switched PID, and release
1144 * the former thread group leader:
1147 /* Become a process group leader with the old leader's pid.
1148 * The old leader becomes a thread of the this thread group.
1149 * Note: The old leader also uses this pid until release_task
1150 * is called. Odd but simple and correct.
1152 tsk
->pid
= leader
->pid
;
1153 change_pid(tsk
, PIDTYPE_PID
, task_pid(leader
));
1154 transfer_pid(leader
, tsk
, PIDTYPE_PGID
);
1155 transfer_pid(leader
, tsk
, PIDTYPE_SID
);
1157 list_replace_rcu(&leader
->tasks
, &tsk
->tasks
);
1158 list_replace_init(&leader
->sibling
, &tsk
->sibling
);
1160 tsk
->group_leader
= tsk
;
1161 leader
->group_leader
= tsk
;
1163 tsk
->exit_signal
= SIGCHLD
;
1164 leader
->exit_signal
= -1;
1166 BUG_ON(leader
->exit_state
!= EXIT_ZOMBIE
);
1167 leader
->exit_state
= EXIT_DEAD
;
1170 * We are going to release_task()->ptrace_unlink() silently,
1171 * the tracer can sleep in do_wait(). EXIT_DEAD guarantees
1172 * the tracer wont't block again waiting for this thread.
1174 if (unlikely(leader
->ptrace
))
1175 __wake_up_parent(leader
, leader
->parent
);
1176 write_unlock_irq(&tasklist_lock
);
1177 cgroup_threadgroup_change_end(tsk
);
1179 release_task(leader
);
1182 sig
->group_exit_task
= NULL
;
1183 sig
->notify_count
= 0;
1186 /* we have changed execution domain */
1187 tsk
->exit_signal
= SIGCHLD
;
1189 #ifdef CONFIG_POSIX_TIMERS
1191 flush_itimer_signals();
1194 if (atomic_read(&oldsighand
->count
) != 1) {
1195 struct sighand_struct
*newsighand
;
1197 * This ->sighand is shared with the CLONE_SIGHAND
1198 * but not CLONE_THREAD task, switch to the new one.
1200 newsighand
= kmem_cache_alloc(sighand_cachep
, GFP_KERNEL
);
1204 atomic_set(&newsighand
->count
, 1);
1205 memcpy(newsighand
->action
, oldsighand
->action
,
1206 sizeof(newsighand
->action
));
1208 write_lock_irq(&tasklist_lock
);
1209 spin_lock(&oldsighand
->siglock
);
1210 rcu_assign_pointer(tsk
->sighand
, newsighand
);
1211 spin_unlock(&oldsighand
->siglock
);
1212 write_unlock_irq(&tasklist_lock
);
1214 __cleanup_sighand(oldsighand
);
1217 BUG_ON(!thread_group_leader(tsk
));
1221 /* protects against exit_notify() and __exit_signal() */
1222 read_lock(&tasklist_lock
);
1223 sig
->group_exit_task
= NULL
;
1224 sig
->notify_count
= 0;
1225 read_unlock(&tasklist_lock
);
1229 char *__get_task_comm(char *buf
, size_t buf_size
, struct task_struct
*tsk
)
1232 strncpy(buf
, tsk
->comm
, buf_size
);
1236 EXPORT_SYMBOL_GPL(__get_task_comm
);
1239 * These functions flushes out all traces of the currently running executable
1240 * so that a new one can be started
1243 void __set_task_comm(struct task_struct
*tsk
, const char *buf
, bool exec
)
1246 trace_task_rename(tsk
, buf
);
1247 strlcpy(tsk
->comm
, buf
, sizeof(tsk
->comm
));
1249 perf_event_comm(tsk
, exec
);
1253 * Calling this is the point of no return. None of the failures will be
1254 * seen by userspace since either the process is already taking a fatal
1255 * signal (via de_thread() or coredump), or will have SEGV raised
1256 * (after exec_mmap()) by search_binary_handlers (see below).
1258 int flush_old_exec(struct linux_binprm
* bprm
)
1263 * Make sure we have a private signal table and that
1264 * we are unassociated from the previous thread group.
1266 retval
= de_thread(current
);
1271 * Must be called _before_ exec_mmap() as bprm->mm is
1272 * not visibile until then. This also enables the update
1275 set_mm_exe_file(bprm
->mm
, bprm
->file
);
1278 * Release all of the old mmap stuff
1280 acct_arg_size(bprm
, 0);
1281 retval
= exec_mmap(bprm
->mm
);
1286 * After clearing bprm->mm (to mark that current is using the
1287 * prepared mm now), we have nothing left of the original
1288 * process. If anything from here on returns an error, the check
1289 * in search_binary_handler() will SEGV current.
1294 current
->flags
&= ~(PF_RANDOMIZE
| PF_FORKNOEXEC
| PF_KTHREAD
|
1295 PF_NOFREEZE
| PF_NO_SETAFFINITY
);
1297 current
->personality
&= ~bprm
->per_clear
;
1300 * We have to apply CLOEXEC before we change whether the process is
1301 * dumpable (in setup_new_exec) to avoid a race with a process in userspace
1302 * trying to access the should-be-closed file descriptors of a process
1303 * undergoing exec(2).
1305 do_close_on_exec(current
->files
);
1311 EXPORT_SYMBOL(flush_old_exec
);
1313 void would_dump(struct linux_binprm
*bprm
, struct file
*file
)
1315 struct inode
*inode
= file_inode(file
);
1316 if (inode_permission(inode
, MAY_READ
) < 0) {
1317 struct user_namespace
*old
, *user_ns
;
1318 bprm
->interp_flags
|= BINPRM_FLAGS_ENFORCE_NONDUMP
;
1320 /* Ensure mm->user_ns contains the executable */
1321 user_ns
= old
= bprm
->mm
->user_ns
;
1322 while ((user_ns
!= &init_user_ns
) &&
1323 !privileged_wrt_inode_uidgid(user_ns
, inode
))
1324 user_ns
= user_ns
->parent
;
1326 if (old
!= user_ns
) {
1327 bprm
->mm
->user_ns
= get_user_ns(user_ns
);
1332 EXPORT_SYMBOL(would_dump
);
1334 void setup_new_exec(struct linux_binprm
* bprm
)
1337 * Once here, prepare_binrpm() will not be called any more, so
1338 * the final state of setuid/setgid/fscaps can be merged into the
1341 bprm
->secureexec
|= bprm
->cap_elevated
;
1343 if (bprm
->secureexec
) {
1344 /* Make sure parent cannot signal privileged process. */
1345 current
->pdeath_signal
= 0;
1348 * For secureexec, reset the stack limit to sane default to
1349 * avoid bad behavior from the prior rlimits. This has to
1350 * happen before arch_pick_mmap_layout(), which examines
1351 * RLIMIT_STACK, but after the point of no return to avoid
1352 * needing to clean up the change on failure.
1354 if (current
->signal
->rlim
[RLIMIT_STACK
].rlim_cur
> _STK_LIM
)
1355 current
->signal
->rlim
[RLIMIT_STACK
].rlim_cur
= _STK_LIM
;
1358 arch_pick_mmap_layout(current
->mm
);
1360 current
->sas_ss_sp
= current
->sas_ss_size
= 0;
1363 * Figure out dumpability. Note that this checking only of current
1364 * is wrong, but userspace depends on it. This should be testing
1365 * bprm->secureexec instead.
1367 if (bprm
->interp_flags
& BINPRM_FLAGS_ENFORCE_NONDUMP
||
1368 !(uid_eq(current_euid(), current_uid()) &&
1369 gid_eq(current_egid(), current_gid())))
1370 set_dumpable(current
->mm
, suid_dumpable
);
1372 set_dumpable(current
->mm
, SUID_DUMP_USER
);
1374 arch_setup_new_exec();
1376 __set_task_comm(current
, kbasename(bprm
->filename
), true);
1378 /* Set the new mm task size. We have to do that late because it may
1379 * depend on TIF_32BIT which is only updated in flush_thread() on
1380 * some architectures like powerpc
1382 current
->mm
->task_size
= TASK_SIZE
;
1384 /* An exec changes our domain. We are no longer part of the thread
1386 current
->self_exec_id
++;
1387 flush_signal_handlers(current
, 0);
1389 EXPORT_SYMBOL(setup_new_exec
);
1392 * Prepare credentials and lock ->cred_guard_mutex.
1393 * install_exec_creds() commits the new creds and drops the lock.
1394 * Or, if exec fails before, free_bprm() should release ->cred and
1397 int prepare_bprm_creds(struct linux_binprm
*bprm
)
1399 if (mutex_lock_interruptible(¤t
->signal
->cred_guard_mutex
))
1400 return -ERESTARTNOINTR
;
1402 bprm
->cred
= prepare_exec_creds();
1403 if (likely(bprm
->cred
))
1406 mutex_unlock(¤t
->signal
->cred_guard_mutex
);
1410 static void free_bprm(struct linux_binprm
*bprm
)
1412 free_arg_pages(bprm
);
1414 mutex_unlock(¤t
->signal
->cred_guard_mutex
);
1415 abort_creds(bprm
->cred
);
1418 allow_write_access(bprm
->file
);
1421 /* If a binfmt changed the interp, free it. */
1422 if (bprm
->interp
!= bprm
->filename
)
1423 kfree(bprm
->interp
);
1427 int bprm_change_interp(const char *interp
, struct linux_binprm
*bprm
)
1429 /* If a binfmt changed the interp, free it first. */
1430 if (bprm
->interp
!= bprm
->filename
)
1431 kfree(bprm
->interp
);
1432 bprm
->interp
= kstrdup(interp
, GFP_KERNEL
);
1437 EXPORT_SYMBOL(bprm_change_interp
);
1440 * install the new credentials for this executable
1442 void install_exec_creds(struct linux_binprm
*bprm
)
1444 security_bprm_committing_creds(bprm
);
1446 commit_creds(bprm
->cred
);
1450 * Disable monitoring for regular users
1451 * when executing setuid binaries. Must
1452 * wait until new credentials are committed
1453 * by commit_creds() above
1455 if (get_dumpable(current
->mm
) != SUID_DUMP_USER
)
1456 perf_event_exit_task(current
);
1458 * cred_guard_mutex must be held at least to this point to prevent
1459 * ptrace_attach() from altering our determination of the task's
1460 * credentials; any time after this it may be unlocked.
1462 security_bprm_committed_creds(bprm
);
1463 mutex_unlock(¤t
->signal
->cred_guard_mutex
);
1465 EXPORT_SYMBOL(install_exec_creds
);
1468 * determine how safe it is to execute the proposed program
1469 * - the caller must hold ->cred_guard_mutex to protect against
1470 * PTRACE_ATTACH or seccomp thread-sync
1472 static void check_unsafe_exec(struct linux_binprm
*bprm
)
1474 struct task_struct
*p
= current
, *t
;
1478 bprm
->unsafe
|= LSM_UNSAFE_PTRACE
;
1481 * This isn't strictly necessary, but it makes it harder for LSMs to
1484 if (task_no_new_privs(current
))
1485 bprm
->unsafe
|= LSM_UNSAFE_NO_NEW_PRIVS
;
1489 spin_lock(&p
->fs
->lock
);
1491 while_each_thread(p
, t
) {
1497 if (p
->fs
->users
> n_fs
)
1498 bprm
->unsafe
|= LSM_UNSAFE_SHARE
;
1501 spin_unlock(&p
->fs
->lock
);
1504 static void bprm_fill_uid(struct linux_binprm
*bprm
)
1506 struct inode
*inode
;
1512 * Since this can be called multiple times (via prepare_binprm),
1513 * we must clear any previous work done when setting set[ug]id
1514 * bits from any earlier bprm->file uses (for example when run
1515 * first for a setuid script then again for its interpreter).
1517 bprm
->cred
->euid
= current_euid();
1518 bprm
->cred
->egid
= current_egid();
1520 if (path_nosuid(&bprm
->file
->f_path
))
1523 if (task_no_new_privs(current
))
1526 inode
= bprm
->file
->f_path
.dentry
->d_inode
;
1527 mode
= READ_ONCE(inode
->i_mode
);
1528 if (!(mode
& (S_ISUID
|S_ISGID
)))
1531 /* Be careful if suid/sgid is set */
1534 /* reload atomically mode/uid/gid now that lock held */
1535 mode
= inode
->i_mode
;
1538 inode_unlock(inode
);
1540 /* We ignore suid/sgid if there are no mappings for them in the ns */
1541 if (!kuid_has_mapping(bprm
->cred
->user_ns
, uid
) ||
1542 !kgid_has_mapping(bprm
->cred
->user_ns
, gid
))
1545 if (mode
& S_ISUID
) {
1546 bprm
->per_clear
|= PER_CLEAR_ON_SETID
;
1547 bprm
->cred
->euid
= uid
;
1550 if ((mode
& (S_ISGID
| S_IXGRP
)) == (S_ISGID
| S_IXGRP
)) {
1551 bprm
->per_clear
|= PER_CLEAR_ON_SETID
;
1552 bprm
->cred
->egid
= gid
;
1557 * Fill the binprm structure from the inode.
1558 * Check permissions, then read the first 128 (BINPRM_BUF_SIZE) bytes
1560 * This may be called multiple times for binary chains (scripts for example).
1562 int prepare_binprm(struct linux_binprm
*bprm
)
1567 bprm_fill_uid(bprm
);
1569 /* fill in binprm security blob */
1570 retval
= security_bprm_set_creds(bprm
);
1573 bprm
->called_set_creds
= 1;
1575 memset(bprm
->buf
, 0, BINPRM_BUF_SIZE
);
1576 return kernel_read(bprm
->file
, bprm
->buf
, BINPRM_BUF_SIZE
, &pos
);
1579 EXPORT_SYMBOL(prepare_binprm
);
1582 * Arguments are '\0' separated strings found at the location bprm->p
1583 * points to; chop off the first by relocating brpm->p to right after
1584 * the first '\0' encountered.
1586 int remove_arg_zero(struct linux_binprm
*bprm
)
1589 unsigned long offset
;
1597 offset
= bprm
->p
& ~PAGE_MASK
;
1598 page
= get_arg_page(bprm
, bprm
->p
, 0);
1603 kaddr
= kmap_atomic(page
);
1605 for (; offset
< PAGE_SIZE
&& kaddr
[offset
];
1606 offset
++, bprm
->p
++)
1609 kunmap_atomic(kaddr
);
1611 } while (offset
== PAGE_SIZE
);
1620 EXPORT_SYMBOL(remove_arg_zero
);
1622 #define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e))
1624 * cycle the list of binary formats handler, until one recognizes the image
1626 int search_binary_handler(struct linux_binprm
*bprm
)
1628 bool need_retry
= IS_ENABLED(CONFIG_MODULES
);
1629 struct linux_binfmt
*fmt
;
1632 /* This allows 4 levels of binfmt rewrites before failing hard. */
1633 if (bprm
->recursion_depth
> 5)
1636 retval
= security_bprm_check(bprm
);
1642 read_lock(&binfmt_lock
);
1643 list_for_each_entry(fmt
, &formats
, lh
) {
1644 if (!try_module_get(fmt
->module
))
1646 read_unlock(&binfmt_lock
);
1647 bprm
->recursion_depth
++;
1648 retval
= fmt
->load_binary(bprm
);
1649 read_lock(&binfmt_lock
);
1651 bprm
->recursion_depth
--;
1652 if (retval
< 0 && !bprm
->mm
) {
1653 /* we got to flush_old_exec() and failed after it */
1654 read_unlock(&binfmt_lock
);
1655 force_sigsegv(SIGSEGV
, current
);
1658 if (retval
!= -ENOEXEC
|| !bprm
->file
) {
1659 read_unlock(&binfmt_lock
);
1663 read_unlock(&binfmt_lock
);
1666 if (printable(bprm
->buf
[0]) && printable(bprm
->buf
[1]) &&
1667 printable(bprm
->buf
[2]) && printable(bprm
->buf
[3]))
1669 if (request_module("binfmt-%04x", *(ushort
*)(bprm
->buf
+ 2)) < 0)
1677 EXPORT_SYMBOL(search_binary_handler
);
1679 static int exec_binprm(struct linux_binprm
*bprm
)
1681 pid_t old_pid
, old_vpid
;
1684 /* Need to fetch pid before load_binary changes it */
1685 old_pid
= current
->pid
;
1687 old_vpid
= task_pid_nr_ns(current
, task_active_pid_ns(current
->parent
));
1690 ret
= search_binary_handler(bprm
);
1693 trace_sched_process_exec(current
, old_pid
, bprm
);
1694 ptrace_event(PTRACE_EVENT_EXEC
, old_vpid
);
1695 proc_exec_connector(current
);
1702 * sys_execve() executes a new program.
1704 static int do_execveat_common(int fd
, struct filename
*filename
,
1705 struct user_arg_ptr argv
,
1706 struct user_arg_ptr envp
,
1709 char *pathbuf
= NULL
;
1710 struct linux_binprm
*bprm
;
1712 struct files_struct
*displaced
;
1715 if (IS_ERR(filename
))
1716 return PTR_ERR(filename
);
1719 * We move the actual failure in case of RLIMIT_NPROC excess from
1720 * set*uid() to execve() because too many poorly written programs
1721 * don't check setuid() return code. Here we additionally recheck
1722 * whether NPROC limit is still exceeded.
1724 if ((current
->flags
& PF_NPROC_EXCEEDED
) &&
1725 atomic_read(¤t_user()->processes
) > rlimit(RLIMIT_NPROC
)) {
1730 /* We're below the limit (still or again), so we don't want to make
1731 * further execve() calls fail. */
1732 current
->flags
&= ~PF_NPROC_EXCEEDED
;
1734 retval
= unshare_files(&displaced
);
1739 bprm
= kzalloc(sizeof(*bprm
), GFP_KERNEL
);
1743 retval
= prepare_bprm_creds(bprm
);
1747 check_unsafe_exec(bprm
);
1748 current
->in_execve
= 1;
1750 file
= do_open_execat(fd
, filename
, flags
);
1751 retval
= PTR_ERR(file
);
1758 if (fd
== AT_FDCWD
|| filename
->name
[0] == '/') {
1759 bprm
->filename
= filename
->name
;
1761 if (filename
->name
[0] == '\0')
1762 pathbuf
= kasprintf(GFP_KERNEL
, "/dev/fd/%d", fd
);
1764 pathbuf
= kasprintf(GFP_KERNEL
, "/dev/fd/%d/%s",
1765 fd
, filename
->name
);
1771 * Record that a name derived from an O_CLOEXEC fd will be
1772 * inaccessible after exec. Relies on having exclusive access to
1773 * current->files (due to unshare_files above).
1775 if (close_on_exec(fd
, rcu_dereference_raw(current
->files
->fdt
)))
1776 bprm
->interp_flags
|= BINPRM_FLAGS_PATH_INACCESSIBLE
;
1777 bprm
->filename
= pathbuf
;
1779 bprm
->interp
= bprm
->filename
;
1781 retval
= bprm_mm_init(bprm
);
1785 bprm
->argc
= count(argv
, MAX_ARG_STRINGS
);
1786 if ((retval
= bprm
->argc
) < 0)
1789 bprm
->envc
= count(envp
, MAX_ARG_STRINGS
);
1790 if ((retval
= bprm
->envc
) < 0)
1793 retval
= prepare_binprm(bprm
);
1797 retval
= copy_strings_kernel(1, &bprm
->filename
, bprm
);
1801 bprm
->exec
= bprm
->p
;
1802 retval
= copy_strings(bprm
->envc
, envp
, bprm
);
1806 retval
= copy_strings(bprm
->argc
, argv
, bprm
);
1810 would_dump(bprm
, bprm
->file
);
1812 retval
= exec_binprm(bprm
);
1816 /* execve succeeded */
1817 current
->fs
->in_exec
= 0;
1818 current
->in_execve
= 0;
1819 membarrier_execve(current
);
1820 acct_update_integrals(current
);
1821 task_numa_free(current
);
1826 put_files_struct(displaced
);
1831 acct_arg_size(bprm
, 0);
1836 current
->fs
->in_exec
= 0;
1837 current
->in_execve
= 0;
1845 reset_files_struct(displaced
);
1851 int do_execve(struct filename
*filename
,
1852 const char __user
*const __user
*__argv
,
1853 const char __user
*const __user
*__envp
)
1855 struct user_arg_ptr argv
= { .ptr
.native
= __argv
};
1856 struct user_arg_ptr envp
= { .ptr
.native
= __envp
};
1857 return do_execveat_common(AT_FDCWD
, filename
, argv
, envp
, 0);
1860 int do_execveat(int fd
, struct filename
*filename
,
1861 const char __user
*const __user
*__argv
,
1862 const char __user
*const __user
*__envp
,
1865 struct user_arg_ptr argv
= { .ptr
.native
= __argv
};
1866 struct user_arg_ptr envp
= { .ptr
.native
= __envp
};
1868 return do_execveat_common(fd
, filename
, argv
, envp
, flags
);
1871 #ifdef CONFIG_COMPAT
1872 static int compat_do_execve(struct filename
*filename
,
1873 const compat_uptr_t __user
*__argv
,
1874 const compat_uptr_t __user
*__envp
)
1876 struct user_arg_ptr argv
= {
1878 .ptr
.compat
= __argv
,
1880 struct user_arg_ptr envp
= {
1882 .ptr
.compat
= __envp
,
1884 return do_execveat_common(AT_FDCWD
, filename
, argv
, envp
, 0);
1887 static int compat_do_execveat(int fd
, struct filename
*filename
,
1888 const compat_uptr_t __user
*__argv
,
1889 const compat_uptr_t __user
*__envp
,
1892 struct user_arg_ptr argv
= {
1894 .ptr
.compat
= __argv
,
1896 struct user_arg_ptr envp
= {
1898 .ptr
.compat
= __envp
,
1900 return do_execveat_common(fd
, filename
, argv
, envp
, flags
);
1904 void set_binfmt(struct linux_binfmt
*new)
1906 struct mm_struct
*mm
= current
->mm
;
1909 module_put(mm
->binfmt
->module
);
1913 __module_get(new->module
);
1915 EXPORT_SYMBOL(set_binfmt
);
1918 * set_dumpable stores three-value SUID_DUMP_* into mm->flags.
1920 void set_dumpable(struct mm_struct
*mm
, int value
)
1922 unsigned long old
, new;
1924 if (WARN_ON((unsigned)value
> SUID_DUMP_ROOT
))
1928 old
= READ_ONCE(mm
->flags
);
1929 new = (old
& ~MMF_DUMPABLE_MASK
) | value
;
1930 } while (cmpxchg(&mm
->flags
, old
, new) != old
);
1933 SYSCALL_DEFINE3(execve
,
1934 const char __user
*, filename
,
1935 const char __user
*const __user
*, argv
,
1936 const char __user
*const __user
*, envp
)
1938 return do_execve(getname(filename
), argv
, envp
);
1941 SYSCALL_DEFINE5(execveat
,
1942 int, fd
, const char __user
*, filename
,
1943 const char __user
*const __user
*, argv
,
1944 const char __user
*const __user
*, envp
,
1947 int lookup_flags
= (flags
& AT_EMPTY_PATH
) ? LOOKUP_EMPTY
: 0;
1949 return do_execveat(fd
,
1950 getname_flags(filename
, lookup_flags
, NULL
),
1954 #ifdef CONFIG_COMPAT
1955 COMPAT_SYSCALL_DEFINE3(execve
, const char __user
*, filename
,
1956 const compat_uptr_t __user
*, argv
,
1957 const compat_uptr_t __user
*, envp
)
1959 return compat_do_execve(getname(filename
), argv
, envp
);
1962 COMPAT_SYSCALL_DEFINE5(execveat
, int, fd
,
1963 const char __user
*, filename
,
1964 const compat_uptr_t __user
*, argv
,
1965 const compat_uptr_t __user
*, envp
,
1968 int lookup_flags
= (flags
& AT_EMPTY_PATH
) ? LOOKUP_EMPTY
: 0;
1970 return compat_do_execveat(fd
,
1971 getname_flags(filename
, lookup_flags
, NULL
),