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>
65 #include <linux/delay.h>
67 #include <trace/events/fs.h>
69 #include <linux/uaccess.h>
70 #include <asm/mmu_context.h>
73 #include <trace/events/task.h>
76 #include <trace/events/sched.h>
78 int suid_dumpable
= 0;
80 static LIST_HEAD(formats
);
81 static DEFINE_RWLOCK(binfmt_lock
);
83 void __register_binfmt(struct linux_binfmt
* fmt
, int insert
)
86 if (WARN_ON(!fmt
->load_binary
))
88 write_lock(&binfmt_lock
);
89 insert
? list_add(&fmt
->lh
, &formats
) :
90 list_add_tail(&fmt
->lh
, &formats
);
91 write_unlock(&binfmt_lock
);
94 EXPORT_SYMBOL(__register_binfmt
);
96 void unregister_binfmt(struct linux_binfmt
* fmt
)
98 write_lock(&binfmt_lock
);
100 write_unlock(&binfmt_lock
);
103 EXPORT_SYMBOL(unregister_binfmt
);
105 static inline void put_binfmt(struct linux_binfmt
* fmt
)
107 module_put(fmt
->module
);
110 bool path_noexec(const struct path
*path
)
112 return (path
->mnt
->mnt_flags
& MNT_NOEXEC
) ||
113 (path
->mnt
->mnt_sb
->s_iflags
& SB_I_NOEXEC
);
115 EXPORT_SYMBOL_GPL(path_noexec
);
117 bool path_nosuid(const struct path
*path
)
119 return !mnt_may_suid(path
->mnt
) ||
120 (path
->mnt
->mnt_sb
->s_iflags
& SB_I_NOSUID
);
122 EXPORT_SYMBOL(path_nosuid
);
126 * Note that a shared library must be both readable and executable due to
129 * Also note that we take the address to load from from the file itself.
131 SYSCALL_DEFINE1(uselib
, const char __user
*, library
)
133 struct linux_binfmt
*fmt
;
135 struct filename
*tmp
= getname(library
);
136 int error
= PTR_ERR(tmp
);
137 static const struct open_flags uselib_flags
= {
138 .open_flag
= O_LARGEFILE
| O_RDONLY
| __FMODE_EXEC
,
139 .acc_mode
= MAY_READ
| MAY_EXEC
,
140 .intent
= LOOKUP_OPEN
,
141 .lookup_flags
= LOOKUP_FOLLOW
,
147 file
= do_filp_open(AT_FDCWD
, tmp
, &uselib_flags
);
149 error
= PTR_ERR(file
);
154 if (!S_ISREG(file_inode(file
)->i_mode
))
158 if (path_noexec(&file
->f_path
))
165 read_lock(&binfmt_lock
);
166 list_for_each_entry(fmt
, &formats
, lh
) {
167 if (!fmt
->load_shlib
)
169 if (!try_module_get(fmt
->module
))
171 read_unlock(&binfmt_lock
);
172 error
= fmt
->load_shlib(file
);
173 read_lock(&binfmt_lock
);
175 if (error
!= -ENOEXEC
)
178 read_unlock(&binfmt_lock
);
184 #endif /* #ifdef CONFIG_USELIB */
188 * The nascent bprm->mm is not visible until exec_mmap() but it can
189 * use a lot of memory, account these pages in current->mm temporary
190 * for oom_badness()->get_mm_rss(). Once exec succeeds or fails, we
191 * change the counter back via acct_arg_size(0).
193 static void acct_arg_size(struct linux_binprm
*bprm
, unsigned long pages
)
195 struct mm_struct
*mm
= current
->mm
;
196 long diff
= (long)(pages
- bprm
->vma_pages
);
201 bprm
->vma_pages
= pages
;
202 add_mm_counter(mm
, MM_ANONPAGES
, diff
);
205 static struct page
*get_arg_page(struct linux_binprm
*bprm
, unsigned long pos
,
210 unsigned int gup_flags
= FOLL_FORCE
;
212 #ifdef CONFIG_STACK_GROWSUP
214 ret
= expand_downwards(bprm
->vma
, pos
);
221 gup_flags
|= FOLL_WRITE
;
224 * We are doing an exec(). 'current' is the process
225 * doing the exec and bprm->mm is the new process's mm.
227 ret
= get_user_pages_remote(current
, bprm
->mm
, pos
, 1, gup_flags
,
233 unsigned long size
= bprm
->vma
->vm_end
- bprm
->vma
->vm_start
;
234 unsigned long ptr_size
, limit
;
237 * Since the stack will hold pointers to the strings, we
238 * must account for them as well.
240 * The size calculation is the entire vma while each arg page is
241 * built, so each time we get here it's calculating how far it
242 * is currently (rather than each call being just the newly
243 * added size from the arg page). As a result, we need to
244 * always add the entire size of the pointers, so that on the
245 * last call to get_arg_page() we'll actually have the entire
248 ptr_size
= (bprm
->argc
+ bprm
->envc
) * sizeof(void *);
249 if (ptr_size
> ULONG_MAX
- size
)
253 acct_arg_size(bprm
, size
/ PAGE_SIZE
);
256 * We've historically supported up to 32 pages (ARG_MAX)
257 * of argument strings even with small stacks
263 * Limit to 1/4 of the max stack size or 3/4 of _STK_LIM
264 * (whichever is smaller) for the argv+env strings.
266 * - the remaining binfmt code will not run out of stack space,
267 * - the program will have a reasonable amount of stack left
270 limit
= _STK_LIM
/ 4 * 3;
271 limit
= min(limit
, rlimit(RLIMIT_STACK
) / 4);
283 static void put_arg_page(struct page
*page
)
288 static void free_arg_pages(struct linux_binprm
*bprm
)
292 static void flush_arg_page(struct linux_binprm
*bprm
, unsigned long pos
,
295 flush_cache_page(bprm
->vma
, pos
, page_to_pfn(page
));
298 static int __bprm_mm_init(struct linux_binprm
*bprm
)
301 struct vm_area_struct
*vma
= NULL
;
302 struct mm_struct
*mm
= bprm
->mm
;
304 bprm
->vma
= vma
= kmem_cache_zalloc(vm_area_cachep
, GFP_KERNEL
);
308 if (down_write_killable(&mm
->mmap_sem
)) {
315 * Place the stack at the largest stack address the architecture
316 * supports. Later, we'll move this to an appropriate place. We don't
317 * use STACK_TOP because that can depend on attributes which aren't
320 BUILD_BUG_ON(VM_STACK_FLAGS
& VM_STACK_INCOMPLETE_SETUP
);
321 vma
->vm_end
= STACK_TOP_MAX
;
322 vma
->vm_start
= vma
->vm_end
- PAGE_SIZE
;
323 vma
->vm_flags
= VM_SOFTDIRTY
| VM_STACK_FLAGS
| VM_STACK_INCOMPLETE_SETUP
;
324 vma
->vm_page_prot
= vm_get_page_prot(vma
->vm_flags
);
325 INIT_LIST_HEAD(&vma
->anon_vma_chain
);
327 err
= insert_vm_struct(mm
, vma
);
331 mm
->stack_vm
= mm
->total_vm
= 1;
332 arch_bprm_mm_init(mm
, vma
);
333 up_write(&mm
->mmap_sem
);
334 bprm
->p
= vma
->vm_end
- sizeof(void *);
337 up_write(&mm
->mmap_sem
);
340 kmem_cache_free(vm_area_cachep
, vma
);
344 static bool valid_arg_len(struct linux_binprm
*bprm
, long len
)
346 return len
<= MAX_ARG_STRLEN
;
351 static inline void acct_arg_size(struct linux_binprm
*bprm
, unsigned long pages
)
355 static struct page
*get_arg_page(struct linux_binprm
*bprm
, unsigned long pos
,
360 page
= bprm
->page
[pos
/ PAGE_SIZE
];
361 if (!page
&& write
) {
362 page
= alloc_page(GFP_HIGHUSER
|__GFP_ZERO
);
365 bprm
->page
[pos
/ PAGE_SIZE
] = page
;
371 static void put_arg_page(struct page
*page
)
375 static void free_arg_page(struct linux_binprm
*bprm
, int i
)
378 __free_page(bprm
->page
[i
]);
379 bprm
->page
[i
] = NULL
;
383 static void free_arg_pages(struct linux_binprm
*bprm
)
387 for (i
= 0; i
< MAX_ARG_PAGES
; i
++)
388 free_arg_page(bprm
, i
);
391 static void flush_arg_page(struct linux_binprm
*bprm
, unsigned long pos
,
396 static int __bprm_mm_init(struct linux_binprm
*bprm
)
398 bprm
->p
= PAGE_SIZE
* MAX_ARG_PAGES
- sizeof(void *);
402 static bool valid_arg_len(struct linux_binprm
*bprm
, long len
)
404 return len
<= bprm
->p
;
407 #endif /* CONFIG_MMU */
410 * Create a new mm_struct and populate it with a temporary stack
411 * vm_area_struct. We don't have enough context at this point to set the stack
412 * flags, permissions, and offset, so we use temporary values. We'll update
413 * them later in setup_arg_pages().
415 static int bprm_mm_init(struct linux_binprm
*bprm
)
418 struct mm_struct
*mm
= NULL
;
420 bprm
->mm
= mm
= mm_alloc();
425 err
= __bprm_mm_init(bprm
);
440 struct user_arg_ptr
{
445 const char __user
*const __user
*native
;
447 const compat_uptr_t __user
*compat
;
452 static const char __user
*get_user_arg_ptr(struct user_arg_ptr argv
, int nr
)
454 const char __user
*native
;
457 if (unlikely(argv
.is_compat
)) {
458 compat_uptr_t compat
;
460 if (get_user(compat
, argv
.ptr
.compat
+ nr
))
461 return ERR_PTR(-EFAULT
);
463 return compat_ptr(compat
);
467 if (get_user(native
, argv
.ptr
.native
+ nr
))
468 return ERR_PTR(-EFAULT
);
474 * count() counts the number of strings in array ARGV.
476 static int count(struct user_arg_ptr argv
, int max
)
480 if (argv
.ptr
.native
!= NULL
) {
482 const char __user
*p
= get_user_arg_ptr(argv
, i
);
494 if (fatal_signal_pending(current
))
495 return -ERESTARTNOHAND
;
503 * 'copy_strings()' copies argument/environment strings from the old
504 * processes's memory to the new process's stack. The call to get_user_pages()
505 * ensures the destination page is created and not swapped out.
507 static int copy_strings(int argc
, struct user_arg_ptr argv
,
508 struct linux_binprm
*bprm
)
510 struct page
*kmapped_page
= NULL
;
512 unsigned long kpos
= 0;
516 const char __user
*str
;
521 str
= get_user_arg_ptr(argv
, argc
);
525 len
= strnlen_user(str
, MAX_ARG_STRLEN
);
530 if (!valid_arg_len(bprm
, len
))
533 /* We're going to work our way backwords. */
539 int offset
, bytes_to_copy
;
541 if (fatal_signal_pending(current
)) {
542 ret
= -ERESTARTNOHAND
;
547 offset
= pos
% PAGE_SIZE
;
551 bytes_to_copy
= offset
;
552 if (bytes_to_copy
> len
)
555 offset
-= bytes_to_copy
;
556 pos
-= bytes_to_copy
;
557 str
-= bytes_to_copy
;
558 len
-= bytes_to_copy
;
560 if (!kmapped_page
|| kpos
!= (pos
& PAGE_MASK
)) {
563 page
= get_arg_page(bprm
, pos
, 1);
570 flush_kernel_dcache_page(kmapped_page
);
571 kunmap(kmapped_page
);
572 put_arg_page(kmapped_page
);
575 kaddr
= kmap(kmapped_page
);
576 kpos
= pos
& PAGE_MASK
;
577 flush_arg_page(bprm
, kpos
, kmapped_page
);
579 if (copy_from_user(kaddr
+offset
, str
, bytes_to_copy
)) {
588 flush_kernel_dcache_page(kmapped_page
);
589 kunmap(kmapped_page
);
590 put_arg_page(kmapped_page
);
596 * Like copy_strings, but get argv and its values from kernel memory.
598 int copy_strings_kernel(int argc
, const char *const *__argv
,
599 struct linux_binprm
*bprm
)
602 mm_segment_t oldfs
= get_fs();
603 struct user_arg_ptr argv
= {
604 .ptr
.native
= (const char __user
*const __user
*)__argv
,
608 r
= copy_strings(argc
, argv
, bprm
);
613 EXPORT_SYMBOL(copy_strings_kernel
);
618 * During bprm_mm_init(), we create a temporary stack at STACK_TOP_MAX. Once
619 * the binfmt code determines where the new stack should reside, we shift it to
620 * its final location. The process proceeds as follows:
622 * 1) Use shift to calculate the new vma endpoints.
623 * 2) Extend vma to cover both the old and new ranges. This ensures the
624 * arguments passed to subsequent functions are consistent.
625 * 3) Move vma's page tables to the new range.
626 * 4) Free up any cleared pgd range.
627 * 5) Shrink the vma to cover only the new range.
629 static int shift_arg_pages(struct vm_area_struct
*vma
, unsigned long shift
)
631 struct mm_struct
*mm
= vma
->vm_mm
;
632 unsigned long old_start
= vma
->vm_start
;
633 unsigned long old_end
= vma
->vm_end
;
634 unsigned long length
= old_end
- old_start
;
635 unsigned long new_start
= old_start
- shift
;
636 unsigned long new_end
= old_end
- shift
;
637 struct mmu_gather tlb
;
639 BUG_ON(new_start
> new_end
);
642 * ensure there are no vmas between where we want to go
645 if (vma
!= find_vma(mm
, new_start
))
649 * cover the whole range: [new_start, old_end)
651 if (vma_adjust(vma
, new_start
, old_end
, vma
->vm_pgoff
, NULL
))
655 * move the page tables downwards, on failure we rely on
656 * process cleanup to remove whatever mess we made.
658 if (length
!= move_page_tables(vma
, old_start
,
659 vma
, new_start
, length
, false))
663 tlb_gather_mmu(&tlb
, mm
, old_start
, old_end
);
664 if (new_end
> old_start
) {
666 * when the old and new regions overlap clear from new_end.
668 free_pgd_range(&tlb
, new_end
, old_end
, new_end
,
669 vma
->vm_next
? vma
->vm_next
->vm_start
: USER_PGTABLES_CEILING
);
672 * otherwise, clean from old_start; this is done to not touch
673 * the address space in [new_end, old_start) some architectures
674 * have constraints on va-space that make this illegal (IA64) -
675 * for the others its just a little faster.
677 free_pgd_range(&tlb
, old_start
, old_end
, new_end
,
678 vma
->vm_next
? vma
->vm_next
->vm_start
: USER_PGTABLES_CEILING
);
680 tlb_finish_mmu(&tlb
, old_start
, old_end
);
683 * Shrink the vma to just the new range. Always succeeds.
685 vma_adjust(vma
, new_start
, new_end
, vma
->vm_pgoff
, NULL
);
691 * Finalizes the stack vm_area_struct. The flags and permissions are updated,
692 * the stack is optionally relocated, and some extra space is added.
694 int setup_arg_pages(struct linux_binprm
*bprm
,
695 unsigned long stack_top
,
696 int executable_stack
)
699 unsigned long stack_shift
;
700 struct mm_struct
*mm
= current
->mm
;
701 struct vm_area_struct
*vma
= bprm
->vma
;
702 struct vm_area_struct
*prev
= NULL
;
703 unsigned long vm_flags
;
704 unsigned long stack_base
;
705 unsigned long stack_size
;
706 unsigned long stack_expand
;
707 unsigned long rlim_stack
;
709 #ifdef CONFIG_STACK_GROWSUP
710 /* Limit stack size */
711 stack_base
= rlimit_max(RLIMIT_STACK
);
712 if (stack_base
> STACK_SIZE_MAX
)
713 stack_base
= STACK_SIZE_MAX
;
715 /* Add space for stack randomization. */
716 stack_base
+= (STACK_RND_MASK
<< PAGE_SHIFT
);
718 /* Make sure we didn't let the argument array grow too large. */
719 if (vma
->vm_end
- vma
->vm_start
> stack_base
)
722 stack_base
= PAGE_ALIGN(stack_top
- stack_base
);
724 stack_shift
= vma
->vm_start
- stack_base
;
725 mm
->arg_start
= bprm
->p
- stack_shift
;
726 bprm
->p
= vma
->vm_end
- stack_shift
;
728 stack_top
= arch_align_stack(stack_top
);
729 stack_top
= PAGE_ALIGN(stack_top
);
731 if (unlikely(stack_top
< mmap_min_addr
) ||
732 unlikely(vma
->vm_end
- vma
->vm_start
>= stack_top
- mmap_min_addr
))
735 stack_shift
= vma
->vm_end
- stack_top
;
737 bprm
->p
-= stack_shift
;
738 mm
->arg_start
= bprm
->p
;
742 bprm
->loader
-= stack_shift
;
743 bprm
->exec
-= stack_shift
;
745 if (down_write_killable(&mm
->mmap_sem
))
748 vm_flags
= VM_STACK_FLAGS
;
751 * Adjust stack execute permissions; explicitly enable for
752 * EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X and leave alone
753 * (arch default) otherwise.
755 if (unlikely(executable_stack
== EXSTACK_ENABLE_X
))
757 else if (executable_stack
== EXSTACK_DISABLE_X
)
758 vm_flags
&= ~VM_EXEC
;
759 vm_flags
|= mm
->def_flags
;
760 vm_flags
|= VM_STACK_INCOMPLETE_SETUP
;
762 ret
= mprotect_fixup(vma
, &prev
, vma
->vm_start
, vma
->vm_end
,
768 /* Move stack pages down in memory. */
770 ret
= shift_arg_pages(vma
, stack_shift
);
775 /* mprotect_fixup is overkill to remove the temporary stack flags */
776 vma
->vm_flags
&= ~VM_STACK_INCOMPLETE_SETUP
;
778 stack_expand
= 131072UL; /* randomly 32*4k (or 2*64k) pages */
779 stack_size
= vma
->vm_end
- vma
->vm_start
;
781 * Align this down to a page boundary as expand_stack
784 rlim_stack
= rlimit(RLIMIT_STACK
) & PAGE_MASK
;
785 #ifdef CONFIG_STACK_GROWSUP
786 if (stack_size
+ stack_expand
> rlim_stack
)
787 stack_base
= vma
->vm_start
+ rlim_stack
;
789 stack_base
= vma
->vm_end
+ stack_expand
;
791 if (stack_size
+ stack_expand
> rlim_stack
)
792 stack_base
= vma
->vm_end
- rlim_stack
;
794 stack_base
= vma
->vm_start
- stack_expand
;
796 current
->mm
->start_stack
= bprm
->p
;
797 ret
= expand_stack(vma
, stack_base
);
802 up_write(&mm
->mmap_sem
);
805 EXPORT_SYMBOL(setup_arg_pages
);
810 * Transfer the program arguments and environment from the holding pages
811 * onto the stack. The provided stack pointer is adjusted accordingly.
813 int transfer_args_to_stack(struct linux_binprm
*bprm
,
814 unsigned long *sp_location
)
816 unsigned long index
, stop
, sp
;
819 stop
= bprm
->p
>> PAGE_SHIFT
;
822 for (index
= MAX_ARG_PAGES
- 1; index
>= stop
; index
--) {
823 unsigned int offset
= index
== stop
? bprm
->p
& ~PAGE_MASK
: 0;
824 char *src
= kmap(bprm
->page
[index
]) + offset
;
825 sp
-= PAGE_SIZE
- offset
;
826 if (copy_to_user((void *) sp
, src
, PAGE_SIZE
- offset
) != 0)
828 kunmap(bprm
->page
[index
]);
838 EXPORT_SYMBOL(transfer_args_to_stack
);
840 #endif /* CONFIG_MMU */
842 static struct file
*do_open_execat(int fd
, struct filename
*name
, int flags
)
846 struct open_flags open_exec_flags
= {
847 .open_flag
= O_LARGEFILE
| O_RDONLY
| __FMODE_EXEC
,
848 .acc_mode
= MAY_EXEC
,
849 .intent
= LOOKUP_OPEN
,
850 .lookup_flags
= LOOKUP_FOLLOW
,
853 if ((flags
& ~(AT_SYMLINK_NOFOLLOW
| AT_EMPTY_PATH
)) != 0)
854 return ERR_PTR(-EINVAL
);
855 if (flags
& AT_SYMLINK_NOFOLLOW
)
856 open_exec_flags
.lookup_flags
&= ~LOOKUP_FOLLOW
;
857 if (flags
& AT_EMPTY_PATH
)
858 open_exec_flags
.lookup_flags
|= LOOKUP_EMPTY
;
860 file
= do_filp_open(fd
, name
, &open_exec_flags
);
865 if (!S_ISREG(file_inode(file
)->i_mode
))
868 if (path_noexec(&file
->f_path
))
871 err
= deny_write_access(file
);
875 if (name
->name
[0] != '\0')
878 trace_open_exec(name
->name
);
888 struct file
*open_exec(const char *name
)
890 struct filename
*filename
= getname_kernel(name
);
891 struct file
*f
= ERR_CAST(filename
);
893 if (!IS_ERR(filename
)) {
894 f
= do_open_execat(AT_FDCWD
, filename
, 0);
899 EXPORT_SYMBOL(open_exec
);
901 int kernel_read(struct file
*file
, loff_t offset
,
902 char *addr
, unsigned long count
)
910 /* The cast to a user pointer is valid due to the set_fs() */
911 result
= vfs_read(file
, (void __user
*)addr
, count
, &pos
);
916 EXPORT_SYMBOL(kernel_read
);
918 int kernel_read_file(struct file
*file
, void **buf
, loff_t
*size
,
919 loff_t max_size
, enum kernel_read_file_id id
)
925 if (!S_ISREG(file_inode(file
)->i_mode
) || max_size
< 0)
928 ret
= security_kernel_read_file(file
, id
);
932 ret
= deny_write_access(file
);
936 i_size
= i_size_read(file_inode(file
));
937 if (max_size
> 0 && i_size
> max_size
) {
946 if (id
!= READING_FIRMWARE_PREALLOC_BUFFER
)
947 *buf
= vmalloc(i_size
);
954 while (pos
< i_size
) {
955 bytes
= kernel_read(file
, pos
, (char *)(*buf
) + pos
,
972 ret
= security_kernel_post_read_file(file
, *buf
, i_size
, id
);
978 if (id
!= READING_FIRMWARE_PREALLOC_BUFFER
) {
985 allow_write_access(file
);
988 EXPORT_SYMBOL_GPL(kernel_read_file
);
990 int kernel_read_file_from_path(char *path
, void **buf
, loff_t
*size
,
991 loff_t max_size
, enum kernel_read_file_id id
)
999 file
= filp_open(path
, O_RDONLY
, 0);
1001 return PTR_ERR(file
);
1003 ret
= kernel_read_file(file
, buf
, size
, max_size
, id
);
1007 EXPORT_SYMBOL_GPL(kernel_read_file_from_path
);
1009 int kernel_read_file_from_fd(int fd
, void **buf
, loff_t
*size
, loff_t max_size
,
1010 enum kernel_read_file_id id
)
1012 struct fd f
= fdget(fd
);
1018 ret
= kernel_read_file(f
.file
, buf
, size
, max_size
, id
);
1023 EXPORT_SYMBOL_GPL(kernel_read_file_from_fd
);
1025 ssize_t
read_code(struct file
*file
, unsigned long addr
, loff_t pos
, size_t len
)
1027 ssize_t res
= vfs_read(file
, (void __user
*)addr
, len
, &pos
);
1029 flush_icache_range(addr
, addr
+ len
);
1032 EXPORT_SYMBOL(read_code
);
1034 static int exec_mmap(struct mm_struct
*mm
)
1036 struct task_struct
*tsk
;
1037 struct mm_struct
*old_mm
, *active_mm
;
1039 /* Notify parent that we're no longer interested in the old VM */
1041 old_mm
= current
->mm
;
1042 mm_release(tsk
, old_mm
);
1045 sync_mm_rss(old_mm
);
1047 * Make sure that if there is a core dump in progress
1048 * for the old mm, we get out and die instead of going
1049 * through with the exec. We must hold mmap_sem around
1050 * checking core_state and changing tsk->mm.
1052 down_read(&old_mm
->mmap_sem
);
1053 if (unlikely(old_mm
->core_state
)) {
1054 up_read(&old_mm
->mmap_sem
);
1059 active_mm
= tsk
->active_mm
;
1061 tsk
->active_mm
= mm
;
1062 activate_mm(active_mm
, mm
);
1063 tsk
->mm
->vmacache_seqnum
= 0;
1064 vmacache_flush(tsk
);
1067 up_read(&old_mm
->mmap_sem
);
1068 BUG_ON(active_mm
!= old_mm
);
1069 setmax_mm_hiwater_rss(&tsk
->signal
->maxrss
, old_mm
);
1070 mm_update_next_owner(old_mm
);
1079 * This function makes sure the current process has its own signal table,
1080 * so that flush_signal_handlers can later reset the handlers without
1081 * disturbing other processes. (Other processes might share the signal
1082 * table via the CLONE_SIGHAND option to clone().)
1084 static int de_thread(struct task_struct
*tsk
)
1086 struct signal_struct
*sig
= tsk
->signal
;
1087 struct sighand_struct
*oldsighand
= tsk
->sighand
;
1088 spinlock_t
*lock
= &oldsighand
->siglock
;
1090 if (thread_group_empty(tsk
))
1091 goto no_thread_group
;
1094 * Kill all other threads in the thread group.
1096 spin_lock_irq(lock
);
1097 if (signal_group_exit(sig
)) {
1099 * Another group action in progress, just
1100 * return so that the signal is processed.
1102 spin_unlock_irq(lock
);
1106 sig
->group_exit_task
= tsk
;
1107 sig
->notify_count
= zap_other_threads(tsk
);
1108 if (!thread_group_leader(tsk
))
1109 sig
->notify_count
--;
1111 while (sig
->notify_count
) {
1112 __set_current_state(TASK_KILLABLE
);
1113 spin_unlock_irq(lock
);
1115 if (unlikely(__fatal_signal_pending(tsk
)))
1117 spin_lock_irq(lock
);
1119 spin_unlock_irq(lock
);
1122 * At this point all other threads have exited, all we have to
1123 * do is to wait for the thread group leader to become inactive,
1124 * and to assume its PID:
1126 if (!thread_group_leader(tsk
)) {
1127 struct task_struct
*leader
= tsk
->group_leader
;
1130 cgroup_threadgroup_change_begin(tsk
);
1131 write_lock_irq(&tasklist_lock
);
1133 * Do this under tasklist_lock to ensure that
1134 * exit_notify() can't miss ->group_exit_task
1136 sig
->notify_count
= -1;
1137 if (likely(leader
->exit_state
))
1139 __set_current_state(TASK_KILLABLE
);
1140 write_unlock_irq(&tasklist_lock
);
1141 cgroup_threadgroup_change_end(tsk
);
1143 if (unlikely(__fatal_signal_pending(tsk
)))
1148 * The only record we have of the real-time age of a
1149 * process, regardless of execs it's done, is start_time.
1150 * All the past CPU time is accumulated in signal_struct
1151 * from sister threads now dead. But in this non-leader
1152 * exec, nothing survives from the original leader thread,
1153 * whose birth marks the true age of this process now.
1154 * When we take on its identity by switching to its PID, we
1155 * also take its birthdate (always earlier than our own).
1157 tsk
->start_time
= leader
->start_time
;
1158 tsk
->real_start_time
= leader
->real_start_time
;
1160 BUG_ON(!same_thread_group(leader
, tsk
));
1161 BUG_ON(has_group_leader_pid(tsk
));
1163 * An exec() starts a new thread group with the
1164 * TGID of the previous thread group. Rehash the
1165 * two threads with a switched PID, and release
1166 * the former thread group leader:
1169 /* Become a process group leader with the old leader's pid.
1170 * The old leader becomes a thread of the this thread group.
1171 * Note: The old leader also uses this pid until release_task
1172 * is called. Odd but simple and correct.
1174 tsk
->pid
= leader
->pid
;
1175 change_pid(tsk
, PIDTYPE_PID
, task_pid(leader
));
1176 transfer_pid(leader
, tsk
, PIDTYPE_PGID
);
1177 transfer_pid(leader
, tsk
, PIDTYPE_SID
);
1179 list_replace_rcu(&leader
->tasks
, &tsk
->tasks
);
1180 list_replace_init(&leader
->sibling
, &tsk
->sibling
);
1182 tsk
->group_leader
= tsk
;
1183 leader
->group_leader
= tsk
;
1185 tsk
->exit_signal
= SIGCHLD
;
1186 leader
->exit_signal
= -1;
1188 BUG_ON(leader
->exit_state
!= EXIT_ZOMBIE
);
1189 leader
->exit_state
= EXIT_DEAD
;
1192 * We are going to release_task()->ptrace_unlink() silently,
1193 * the tracer can sleep in do_wait(). EXIT_DEAD guarantees
1194 * the tracer wont't block again waiting for this thread.
1196 if (unlikely(leader
->ptrace
))
1197 __wake_up_parent(leader
, leader
->parent
);
1198 write_unlock_irq(&tasklist_lock
);
1199 cgroup_threadgroup_change_end(tsk
);
1201 release_task(leader
);
1204 sig
->group_exit_task
= NULL
;
1205 sig
->notify_count
= 0;
1208 /* we have changed execution domain */
1209 tsk
->exit_signal
= SIGCHLD
;
1211 #ifdef CONFIG_POSIX_TIMERS
1213 flush_itimer_signals();
1216 if (atomic_read(&oldsighand
->count
) != 1) {
1217 struct sighand_struct
*newsighand
;
1219 * This ->sighand is shared with the CLONE_SIGHAND
1220 * but not CLONE_THREAD task, switch to the new one.
1222 newsighand
= kmem_cache_alloc(sighand_cachep
, GFP_KERNEL
);
1226 atomic_set(&newsighand
->count
, 1);
1227 memcpy(newsighand
->action
, oldsighand
->action
,
1228 sizeof(newsighand
->action
));
1230 write_lock_irq(&tasklist_lock
);
1231 spin_lock(&oldsighand
->siglock
);
1232 rcu_assign_pointer(tsk
->sighand
, newsighand
);
1233 spin_unlock(&oldsighand
->siglock
);
1234 write_unlock_irq(&tasklist_lock
);
1236 __cleanup_sighand(oldsighand
);
1239 BUG_ON(!thread_group_leader(tsk
));
1243 /* protects against exit_notify() and __exit_signal() */
1244 read_lock(&tasklist_lock
);
1245 sig
->group_exit_task
= NULL
;
1246 sig
->notify_count
= 0;
1247 read_unlock(&tasklist_lock
);
1251 char *get_task_comm(char *buf
, struct task_struct
*tsk
)
1253 /* buf must be at least sizeof(tsk->comm) in size */
1255 strncpy(buf
, tsk
->comm
, sizeof(tsk
->comm
));
1259 EXPORT_SYMBOL_GPL(get_task_comm
);
1262 * These functions flushes out all traces of the currently running executable
1263 * so that a new one can be started
1266 void __set_task_comm(struct task_struct
*tsk
, const char *buf
, bool exec
)
1269 trace_task_rename(tsk
, buf
);
1270 strlcpy(tsk
->comm
, buf
, sizeof(tsk
->comm
));
1272 perf_event_comm(tsk
, exec
);
1275 int flush_old_exec(struct linux_binprm
* bprm
)
1280 * Make sure we have a private signal table and that
1281 * we are unassociated from the previous thread group.
1283 retval
= de_thread(current
);
1288 * Must be called _before_ exec_mmap() as bprm->mm is
1289 * not visibile until then. This also enables the update
1292 set_mm_exe_file(bprm
->mm
, bprm
->file
);
1295 * Release all of the old mmap stuff
1297 acct_arg_size(bprm
, 0);
1298 retval
= exec_mmap(bprm
->mm
);
1302 bprm
->mm
= NULL
; /* We're using it now */
1305 current
->flags
&= ~(PF_RANDOMIZE
| PF_FORKNOEXEC
| PF_KTHREAD
|
1306 PF_NOFREEZE
| PF_NO_SETAFFINITY
);
1308 current
->personality
&= ~bprm
->per_clear
;
1311 * We have to apply CLOEXEC before we change whether the process is
1312 * dumpable (in setup_new_exec) to avoid a race with a process in userspace
1313 * trying to access the should-be-closed file descriptors of a process
1314 * undergoing exec(2).
1316 do_close_on_exec(current
->files
);
1322 EXPORT_SYMBOL(flush_old_exec
);
1324 void would_dump(struct linux_binprm
*bprm
, struct file
*file
)
1326 struct inode
*inode
= file_inode(file
);
1327 if (inode_permission(inode
, MAY_READ
) < 0) {
1328 struct user_namespace
*old
, *user_ns
;
1329 bprm
->interp_flags
|= BINPRM_FLAGS_ENFORCE_NONDUMP
;
1331 /* Ensure mm->user_ns contains the executable */
1332 user_ns
= old
= bprm
->mm
->user_ns
;
1333 while ((user_ns
!= &init_user_ns
) &&
1334 !privileged_wrt_inode_uidgid(user_ns
, inode
))
1335 user_ns
= user_ns
->parent
;
1337 if (old
!= user_ns
) {
1338 bprm
->mm
->user_ns
= get_user_ns(user_ns
);
1343 EXPORT_SYMBOL(would_dump
);
1345 void setup_new_exec(struct linux_binprm
* bprm
)
1347 arch_pick_mmap_layout(current
->mm
);
1349 /* This is the point of no return */
1350 current
->sas_ss_sp
= current
->sas_ss_size
= 0;
1352 if (uid_eq(current_euid(), current_uid()) && gid_eq(current_egid(), current_gid()))
1353 set_dumpable(current
->mm
, SUID_DUMP_USER
);
1355 set_dumpable(current
->mm
, suid_dumpable
);
1357 arch_setup_new_exec();
1359 __set_task_comm(current
, kbasename(bprm
->filename
), true);
1361 /* Set the new mm task size. We have to do that late because it may
1362 * depend on TIF_32BIT which is only updated in flush_thread() on
1363 * some architectures like powerpc
1365 current
->mm
->task_size
= TASK_SIZE
;
1367 /* install the new credentials */
1368 if (!uid_eq(bprm
->cred
->uid
, current_euid()) ||
1369 !gid_eq(bprm
->cred
->gid
, current_egid())) {
1370 current
->pdeath_signal
= 0;
1372 if (bprm
->interp_flags
& BINPRM_FLAGS_ENFORCE_NONDUMP
)
1373 set_dumpable(current
->mm
, suid_dumpable
);
1376 /* An exec changes our domain. We are no longer part of the thread
1378 current
->self_exec_id
++;
1379 flush_signal_handlers(current
, 0);
1381 EXPORT_SYMBOL(setup_new_exec
);
1384 * Prepare credentials and lock ->cred_guard_mutex.
1385 * install_exec_creds() commits the new creds and drops the lock.
1386 * Or, if exec fails before, free_bprm() should release ->cred and
1389 int prepare_bprm_creds(struct linux_binprm
*bprm
)
1391 if (mutex_lock_interruptible(¤t
->signal
->cred_guard_mutex
))
1392 return -ERESTARTNOINTR
;
1394 bprm
->cred
= prepare_exec_creds();
1395 if (likely(bprm
->cred
))
1398 mutex_unlock(¤t
->signal
->cred_guard_mutex
);
1402 static void free_bprm(struct linux_binprm
*bprm
)
1404 free_arg_pages(bprm
);
1406 mutex_unlock(¤t
->signal
->cred_guard_mutex
);
1407 abort_creds(bprm
->cred
);
1410 allow_write_access(bprm
->file
);
1413 /* If a binfmt changed the interp, free it. */
1414 if (bprm
->interp
!= bprm
->filename
)
1415 kfree(bprm
->interp
);
1419 int bprm_change_interp(char *interp
, struct linux_binprm
*bprm
)
1421 /* If a binfmt changed the interp, free it first. */
1422 if (bprm
->interp
!= bprm
->filename
)
1423 kfree(bprm
->interp
);
1424 bprm
->interp
= kstrdup(interp
, GFP_KERNEL
);
1429 EXPORT_SYMBOL(bprm_change_interp
);
1432 * install the new credentials for this executable
1434 void install_exec_creds(struct linux_binprm
*bprm
)
1436 security_bprm_committing_creds(bprm
);
1438 commit_creds(bprm
->cred
);
1442 * Disable monitoring for regular users
1443 * when executing setuid binaries. Must
1444 * wait until new credentials are committed
1445 * by commit_creds() above
1447 if (get_dumpable(current
->mm
) != SUID_DUMP_USER
)
1448 perf_event_exit_task(current
);
1450 * cred_guard_mutex must be held at least to this point to prevent
1451 * ptrace_attach() from altering our determination of the task's
1452 * credentials; any time after this it may be unlocked.
1454 security_bprm_committed_creds(bprm
);
1455 mutex_unlock(¤t
->signal
->cred_guard_mutex
);
1457 EXPORT_SYMBOL(install_exec_creds
);
1460 * determine how safe it is to execute the proposed program
1461 * - the caller must hold ->cred_guard_mutex to protect against
1462 * PTRACE_ATTACH or seccomp thread-sync
1464 static void check_unsafe_exec(struct linux_binprm
*bprm
)
1466 struct task_struct
*p
= current
, *t
;
1472 bprm
->unsafe
|= LSM_UNSAFE_PTRACE
;
1475 * This isn't strictly necessary, but it makes it harder for LSMs to
1478 if (task_no_new_privs(current
))
1479 bprm
->unsafe
|= LSM_UNSAFE_NO_NEW_PRIVS
;
1485 spin_lock(&p
->fs
->lock
);
1487 while_each_thread(p
, t
) {
1490 if (t
->flags
& (PF_EXITING
| PF_FORKNOEXEC
))
1495 if (p
->fs
->users
> n_fs
) {
1498 spin_unlock(&p
->fs
->lock
);
1499 msleep_interruptible(1);
1503 bprm
->unsafe
|= LSM_UNSAFE_SHARE
;
1506 spin_unlock(&p
->fs
->lock
);
1509 static void bprm_fill_uid(struct linux_binprm
*bprm
)
1511 struct inode
*inode
;
1517 * Since this can be called multiple times (via prepare_binprm),
1518 * we must clear any previous work done when setting set[ug]id
1519 * bits from any earlier bprm->file uses (for example when run
1520 * first for a setuid script then again for its interpreter).
1522 bprm
->cred
->euid
= current_euid();
1523 bprm
->cred
->egid
= current_egid();
1525 if (path_nosuid(&bprm
->file
->f_path
))
1528 if (task_no_new_privs(current
))
1531 inode
= bprm
->file
->f_path
.dentry
->d_inode
;
1532 mode
= READ_ONCE(inode
->i_mode
);
1533 if (!(mode
& (S_ISUID
|S_ISGID
)))
1536 /* Be careful if suid/sgid is set */
1539 /* reload atomically mode/uid/gid now that lock held */
1540 mode
= inode
->i_mode
;
1543 inode_unlock(inode
);
1545 /* We ignore suid/sgid if there are no mappings for them in the ns */
1546 if (!kuid_has_mapping(bprm
->cred
->user_ns
, uid
) ||
1547 !kgid_has_mapping(bprm
->cred
->user_ns
, gid
))
1550 if (mode
& S_ISUID
) {
1551 bprm
->per_clear
|= PER_CLEAR_ON_SETID
;
1552 bprm
->cred
->euid
= uid
;
1555 if ((mode
& (S_ISGID
| S_IXGRP
)) == (S_ISGID
| S_IXGRP
)) {
1556 bprm
->per_clear
|= PER_CLEAR_ON_SETID
;
1557 bprm
->cred
->egid
= gid
;
1562 * Fill the binprm structure from the inode.
1563 * Check permissions, then read the first 128 (BINPRM_BUF_SIZE) bytes
1565 * This may be called multiple times for binary chains (scripts for example).
1567 int prepare_binprm(struct linux_binprm
*bprm
)
1571 bprm_fill_uid(bprm
);
1573 /* fill in binprm security blob */
1574 retval
= security_bprm_set_creds(bprm
);
1577 bprm
->cred_prepared
= 1;
1579 memset(bprm
->buf
, 0, BINPRM_BUF_SIZE
);
1580 return kernel_read(bprm
->file
, 0, bprm
->buf
, BINPRM_BUF_SIZE
);
1583 EXPORT_SYMBOL(prepare_binprm
);
1586 * Arguments are '\0' separated strings found at the location bprm->p
1587 * points to; chop off the first by relocating brpm->p to right after
1588 * the first '\0' encountered.
1590 int remove_arg_zero(struct linux_binprm
*bprm
)
1593 unsigned long offset
;
1601 offset
= bprm
->p
& ~PAGE_MASK
;
1602 page
= get_arg_page(bprm
, bprm
->p
, 0);
1607 kaddr
= kmap_atomic(page
);
1609 for (; offset
< PAGE_SIZE
&& kaddr
[offset
];
1610 offset
++, bprm
->p
++)
1613 kunmap_atomic(kaddr
);
1615 } while (offset
== PAGE_SIZE
);
1624 EXPORT_SYMBOL(remove_arg_zero
);
1626 #define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e))
1628 * cycle the list of binary formats handler, until one recognizes the image
1630 int search_binary_handler(struct linux_binprm
*bprm
)
1632 bool need_retry
= IS_ENABLED(CONFIG_MODULES
);
1633 struct linux_binfmt
*fmt
;
1636 /* This allows 4 levels of binfmt rewrites before failing hard. */
1637 if (bprm
->recursion_depth
> 5)
1640 retval
= security_bprm_check(bprm
);
1646 read_lock(&binfmt_lock
);
1647 list_for_each_entry(fmt
, &formats
, lh
) {
1648 if (!try_module_get(fmt
->module
))
1650 read_unlock(&binfmt_lock
);
1651 bprm
->recursion_depth
++;
1652 retval
= fmt
->load_binary(bprm
);
1653 read_lock(&binfmt_lock
);
1655 bprm
->recursion_depth
--;
1656 if (retval
< 0 && !bprm
->mm
) {
1657 /* we got to flush_old_exec() and failed after it */
1658 read_unlock(&binfmt_lock
);
1659 force_sigsegv(SIGSEGV
, current
);
1662 if (retval
!= -ENOEXEC
|| !bprm
->file
) {
1663 read_unlock(&binfmt_lock
);
1667 read_unlock(&binfmt_lock
);
1670 if (printable(bprm
->buf
[0]) && printable(bprm
->buf
[1]) &&
1671 printable(bprm
->buf
[2]) && printable(bprm
->buf
[3]))
1673 if (request_module("binfmt-%04x", *(ushort
*)(bprm
->buf
+ 2)) < 0)
1681 EXPORT_SYMBOL(search_binary_handler
);
1683 static int exec_binprm(struct linux_binprm
*bprm
)
1685 pid_t old_pid
, old_vpid
;
1688 /* Need to fetch pid before load_binary changes it */
1689 old_pid
= current
->pid
;
1691 old_vpid
= task_pid_nr_ns(current
, task_active_pid_ns(current
->parent
));
1694 ret
= search_binary_handler(bprm
);
1697 trace_sched_process_exec(current
, old_pid
, bprm
);
1698 ptrace_event(PTRACE_EVENT_EXEC
, old_vpid
);
1699 proc_exec_connector(current
);
1706 * sys_execve() executes a new program.
1708 static int do_execveat_common(int fd
, struct filename
*filename
,
1709 struct user_arg_ptr argv
,
1710 struct user_arg_ptr envp
,
1713 char *pathbuf
= NULL
;
1714 struct linux_binprm
*bprm
;
1716 struct files_struct
*displaced
;
1719 if (IS_ERR(filename
))
1720 return PTR_ERR(filename
);
1723 * We move the actual failure in case of RLIMIT_NPROC excess from
1724 * set*uid() to execve() because too many poorly written programs
1725 * don't check setuid() return code. Here we additionally recheck
1726 * whether NPROC limit is still exceeded.
1728 if ((current
->flags
& PF_NPROC_EXCEEDED
) &&
1729 atomic_read(¤t_user()->processes
) > rlimit(RLIMIT_NPROC
)) {
1734 /* We're below the limit (still or again), so we don't want to make
1735 * further execve() calls fail. */
1736 current
->flags
&= ~PF_NPROC_EXCEEDED
;
1738 retval
= unshare_files(&displaced
);
1743 bprm
= kzalloc(sizeof(*bprm
), GFP_KERNEL
);
1747 retval
= prepare_bprm_creds(bprm
);
1751 check_unsafe_exec(bprm
);
1752 current
->in_execve
= 1;
1754 file
= do_open_execat(fd
, filename
, flags
);
1755 retval
= PTR_ERR(file
);
1762 if (fd
== AT_FDCWD
|| filename
->name
[0] == '/') {
1763 bprm
->filename
= filename
->name
;
1765 if (filename
->name
[0] == '\0')
1766 pathbuf
= kasprintf(GFP_TEMPORARY
, "/dev/fd/%d", fd
);
1768 pathbuf
= kasprintf(GFP_TEMPORARY
, "/dev/fd/%d/%s",
1769 fd
, filename
->name
);
1775 * Record that a name derived from an O_CLOEXEC fd will be
1776 * inaccessible after exec. Relies on having exclusive access to
1777 * current->files (due to unshare_files above).
1779 if (close_on_exec(fd
, rcu_dereference_raw(current
->files
->fdt
)))
1780 bprm
->interp_flags
|= BINPRM_FLAGS_PATH_INACCESSIBLE
;
1781 bprm
->filename
= pathbuf
;
1783 bprm
->interp
= bprm
->filename
;
1785 retval
= bprm_mm_init(bprm
);
1789 bprm
->argc
= count(argv
, MAX_ARG_STRINGS
);
1790 if ((retval
= bprm
->argc
) < 0)
1793 bprm
->envc
= count(envp
, MAX_ARG_STRINGS
);
1794 if ((retval
= bprm
->envc
) < 0)
1797 retval
= prepare_binprm(bprm
);
1801 retval
= copy_strings_kernel(1, &bprm
->filename
, bprm
);
1805 bprm
->exec
= bprm
->p
;
1806 retval
= copy_strings(bprm
->envc
, envp
, bprm
);
1810 retval
= copy_strings(bprm
->argc
, argv
, bprm
);
1814 would_dump(bprm
, bprm
->file
);
1816 retval
= exec_binprm(bprm
);
1820 /* execve succeeded */
1821 current
->fs
->in_exec
= 0;
1822 current
->in_execve
= 0;
1823 acct_update_integrals(current
);
1824 task_numa_free(current
);
1829 put_files_struct(displaced
);
1834 acct_arg_size(bprm
, 0);
1839 current
->fs
->in_exec
= 0;
1840 current
->in_execve
= 0;
1848 reset_files_struct(displaced
);
1854 int do_execve(struct filename
*filename
,
1855 const char __user
*const __user
*__argv
,
1856 const char __user
*const __user
*__envp
)
1858 struct user_arg_ptr argv
= { .ptr
.native
= __argv
};
1859 struct user_arg_ptr envp
= { .ptr
.native
= __envp
};
1860 return do_execveat_common(AT_FDCWD
, filename
, argv
, envp
, 0);
1863 int do_execveat(int fd
, struct filename
*filename
,
1864 const char __user
*const __user
*__argv
,
1865 const char __user
*const __user
*__envp
,
1868 struct user_arg_ptr argv
= { .ptr
.native
= __argv
};
1869 struct user_arg_ptr envp
= { .ptr
.native
= __envp
};
1871 return do_execveat_common(fd
, filename
, argv
, envp
, flags
);
1874 #ifdef CONFIG_COMPAT
1875 static int compat_do_execve(struct filename
*filename
,
1876 const compat_uptr_t __user
*__argv
,
1877 const compat_uptr_t __user
*__envp
)
1879 struct user_arg_ptr argv
= {
1881 .ptr
.compat
= __argv
,
1883 struct user_arg_ptr envp
= {
1885 .ptr
.compat
= __envp
,
1887 return do_execveat_common(AT_FDCWD
, filename
, argv
, envp
, 0);
1890 static int compat_do_execveat(int fd
, struct filename
*filename
,
1891 const compat_uptr_t __user
*__argv
,
1892 const compat_uptr_t __user
*__envp
,
1895 struct user_arg_ptr argv
= {
1897 .ptr
.compat
= __argv
,
1899 struct user_arg_ptr envp
= {
1901 .ptr
.compat
= __envp
,
1903 return do_execveat_common(fd
, filename
, argv
, envp
, flags
);
1907 void set_binfmt(struct linux_binfmt
*new)
1909 struct mm_struct
*mm
= current
->mm
;
1912 module_put(mm
->binfmt
->module
);
1916 __module_get(new->module
);
1918 EXPORT_SYMBOL(set_binfmt
);
1921 * set_dumpable stores three-value SUID_DUMP_* into mm->flags.
1923 void set_dumpable(struct mm_struct
*mm
, int value
)
1925 unsigned long old
, new;
1927 if (WARN_ON((unsigned)value
> SUID_DUMP_ROOT
))
1931 old
= ACCESS_ONCE(mm
->flags
);
1932 new = (old
& ~MMF_DUMPABLE_MASK
) | value
;
1933 } while (cmpxchg(&mm
->flags
, old
, new) != old
);
1936 SYSCALL_DEFINE3(execve
,
1937 const char __user
*, filename
,
1938 const char __user
*const __user
*, argv
,
1939 const char __user
*const __user
*, envp
)
1941 return do_execve(getname(filename
), argv
, envp
);
1944 SYSCALL_DEFINE5(execveat
,
1945 int, fd
, const char __user
*, filename
,
1946 const char __user
*const __user
*, argv
,
1947 const char __user
*const __user
*, envp
,
1950 int lookup_flags
= (flags
& AT_EMPTY_PATH
) ? LOOKUP_EMPTY
: 0;
1952 return do_execveat(fd
,
1953 getname_flags(filename
, lookup_flags
, NULL
),
1957 #ifdef CONFIG_COMPAT
1958 COMPAT_SYSCALL_DEFINE3(execve
, const char __user
*, filename
,
1959 const compat_uptr_t __user
*, argv
,
1960 const compat_uptr_t __user
*, envp
)
1962 return compat_do_execve(getname(filename
), argv
, envp
);
1965 COMPAT_SYSCALL_DEFINE5(execveat
, int, fd
,
1966 const char __user
*, filename
,
1967 const compat_uptr_t __user
*, argv
,
1968 const compat_uptr_t __user
*, envp
,
1971 int lookup_flags
= (flags
& AT_EMPTY_PATH
) ? LOOKUP_EMPTY
: 0;
1973 return compat_do_execveat(fd
,
1974 getname_flags(filename
, lookup_flags
, NULL
),