1 // SPDX-License-Identifier: GPL-2.0-only
5 * Copyright (C) 1991, 1992 Linus Torvalds
9 * #!-checking implemented by tytso.
12 * Demand-loading implemented 01.12.91 - no need to read anything but
13 * the header into memory. The inode of the executable is put into
14 * "current->executable", and page faults do the actual loading. Clean.
16 * Once more I can proudly say that linux stood up to being changed: it
17 * was less than 2 hours work to get demand-loading completely implemented.
19 * Demand loading changed July 1993 by Eric Youngdale. Use mmap instead,
20 * current->executable is only used by the procfs. This allows a dispatch
21 * table to check for several different types of binary formats. We keep
22 * trying until we recognize the file or we run out of supported binary
26 #include <linux/slab.h>
27 #include <linux/file.h>
28 #include <linux/fdtable.h>
30 #include <linux/vmacache.h>
31 #include <linux/stat.h>
32 #include <linux/fcntl.h>
33 #include <linux/swap.h>
34 #include <linux/string.h>
35 #include <linux/init.h>
36 #include <linux/sched/mm.h>
37 #include <linux/sched/coredump.h>
38 #include <linux/sched/signal.h>
39 #include <linux/sched/numa_balancing.h>
40 #include <linux/sched/task.h>
41 #include <linux/pagemap.h>
42 #include <linux/perf_event.h>
43 #include <linux/highmem.h>
44 #include <linux/spinlock.h>
45 #include <linux/key.h>
46 #include <linux/personality.h>
47 #include <linux/binfmts.h>
48 #include <linux/utsname.h>
49 #include <linux/pid_namespace.h>
50 #include <linux/module.h>
51 #include <linux/namei.h>
52 #include <linux/mount.h>
53 #include <linux/security.h>
54 #include <linux/syscalls.h>
55 #include <linux/tsacct_kern.h>
56 #include <linux/cn_proc.h>
57 #include <linux/audit.h>
58 #include <linux/tracehook.h>
59 #include <linux/kmod.h>
60 #include <linux/fsnotify.h>
61 #include <linux/fs_struct.h>
62 #include <linux/oom.h>
63 #include <linux/compat.h>
64 #include <linux/vmalloc.h>
66 #include <linux/uaccess.h>
67 #include <asm/mmu_context.h>
70 #include <trace/events/task.h>
73 #include <trace/events/sched.h>
75 int suid_dumpable
= 0;
77 static LIST_HEAD(formats
);
78 static DEFINE_RWLOCK(binfmt_lock
);
80 void __register_binfmt(struct linux_binfmt
* fmt
, int insert
)
83 if (WARN_ON(!fmt
->load_binary
))
85 write_lock(&binfmt_lock
);
86 insert
? list_add(&fmt
->lh
, &formats
) :
87 list_add_tail(&fmt
->lh
, &formats
);
88 write_unlock(&binfmt_lock
);
91 EXPORT_SYMBOL(__register_binfmt
);
93 void unregister_binfmt(struct linux_binfmt
* fmt
)
95 write_lock(&binfmt_lock
);
97 write_unlock(&binfmt_lock
);
100 EXPORT_SYMBOL(unregister_binfmt
);
102 static inline void put_binfmt(struct linux_binfmt
* fmt
)
104 module_put(fmt
->module
);
107 bool path_noexec(const struct path
*path
)
109 return (path
->mnt
->mnt_flags
& MNT_NOEXEC
) ||
110 (path
->mnt
->mnt_sb
->s_iflags
& SB_I_NOEXEC
);
115 * Note that a shared library must be both readable and executable due to
118 * Also note that we take the address to load from from the file itself.
120 SYSCALL_DEFINE1(uselib
, const char __user
*, library
)
122 struct linux_binfmt
*fmt
;
124 struct filename
*tmp
= getname(library
);
125 int error
= PTR_ERR(tmp
);
126 static const struct open_flags uselib_flags
= {
127 .open_flag
= O_LARGEFILE
| O_RDONLY
| __FMODE_EXEC
,
128 .acc_mode
= MAY_READ
| MAY_EXEC
,
129 .intent
= LOOKUP_OPEN
,
130 .lookup_flags
= LOOKUP_FOLLOW
,
136 file
= do_filp_open(AT_FDCWD
, tmp
, &uselib_flags
);
138 error
= PTR_ERR(file
);
143 if (!S_ISREG(file_inode(file
)->i_mode
))
147 if (path_noexec(&file
->f_path
))
154 read_lock(&binfmt_lock
);
155 list_for_each_entry(fmt
, &formats
, lh
) {
156 if (!fmt
->load_shlib
)
158 if (!try_module_get(fmt
->module
))
160 read_unlock(&binfmt_lock
);
161 error
= fmt
->load_shlib(file
);
162 read_lock(&binfmt_lock
);
164 if (error
!= -ENOEXEC
)
167 read_unlock(&binfmt_lock
);
173 #endif /* #ifdef CONFIG_USELIB */
177 * The nascent bprm->mm is not visible until exec_mmap() but it can
178 * use a lot of memory, account these pages in current->mm temporary
179 * for oom_badness()->get_mm_rss(). Once exec succeeds or fails, we
180 * change the counter back via acct_arg_size(0).
182 static void acct_arg_size(struct linux_binprm
*bprm
, unsigned long pages
)
184 struct mm_struct
*mm
= current
->mm
;
185 long diff
= (long)(pages
- bprm
->vma_pages
);
190 bprm
->vma_pages
= pages
;
191 add_mm_counter(mm
, MM_ANONPAGES
, diff
);
194 static struct page
*get_arg_page(struct linux_binprm
*bprm
, unsigned long pos
,
199 unsigned int gup_flags
= FOLL_FORCE
;
201 #ifdef CONFIG_STACK_GROWSUP
203 ret
= expand_downwards(bprm
->vma
, pos
);
210 gup_flags
|= FOLL_WRITE
;
213 * We are doing an exec(). 'current' is the process
214 * doing the exec and bprm->mm is the new process's mm.
216 ret
= get_user_pages_remote(current
, bprm
->mm
, pos
, 1, gup_flags
,
222 acct_arg_size(bprm
, vma_pages(bprm
->vma
));
227 static void put_arg_page(struct page
*page
)
232 static void free_arg_pages(struct linux_binprm
*bprm
)
236 static void flush_arg_page(struct linux_binprm
*bprm
, unsigned long pos
,
239 flush_cache_page(bprm
->vma
, pos
, page_to_pfn(page
));
242 static int __bprm_mm_init(struct linux_binprm
*bprm
)
245 struct vm_area_struct
*vma
= NULL
;
246 struct mm_struct
*mm
= bprm
->mm
;
248 bprm
->vma
= vma
= vm_area_alloc(mm
);
251 vma_set_anonymous(vma
);
253 if (down_write_killable(&mm
->mmap_sem
)) {
259 * Place the stack at the largest stack address the architecture
260 * supports. Later, we'll move this to an appropriate place. We don't
261 * use STACK_TOP because that can depend on attributes which aren't
264 BUILD_BUG_ON(VM_STACK_FLAGS
& VM_STACK_INCOMPLETE_SETUP
);
265 vma
->vm_end
= STACK_TOP_MAX
;
266 vma
->vm_start
= vma
->vm_end
- PAGE_SIZE
;
267 vma
->vm_flags
= VM_SOFTDIRTY
| VM_STACK_FLAGS
| VM_STACK_INCOMPLETE_SETUP
;
268 vma
->vm_page_prot
= vm_get_page_prot(vma
->vm_flags
);
270 err
= insert_vm_struct(mm
, vma
);
274 mm
->stack_vm
= mm
->total_vm
= 1;
275 up_write(&mm
->mmap_sem
);
276 bprm
->p
= vma
->vm_end
- sizeof(void *);
279 up_write(&mm
->mmap_sem
);
286 static bool valid_arg_len(struct linux_binprm
*bprm
, long len
)
288 return len
<= MAX_ARG_STRLEN
;
293 static inline void acct_arg_size(struct linux_binprm
*bprm
, unsigned long pages
)
297 static struct page
*get_arg_page(struct linux_binprm
*bprm
, unsigned long pos
,
302 page
= bprm
->page
[pos
/ PAGE_SIZE
];
303 if (!page
&& write
) {
304 page
= alloc_page(GFP_HIGHUSER
|__GFP_ZERO
);
307 bprm
->page
[pos
/ PAGE_SIZE
] = page
;
313 static void put_arg_page(struct page
*page
)
317 static void free_arg_page(struct linux_binprm
*bprm
, int i
)
320 __free_page(bprm
->page
[i
]);
321 bprm
->page
[i
] = NULL
;
325 static void free_arg_pages(struct linux_binprm
*bprm
)
329 for (i
= 0; i
< MAX_ARG_PAGES
; i
++)
330 free_arg_page(bprm
, i
);
333 static void flush_arg_page(struct linux_binprm
*bprm
, unsigned long pos
,
338 static int __bprm_mm_init(struct linux_binprm
*bprm
)
340 bprm
->p
= PAGE_SIZE
* MAX_ARG_PAGES
- sizeof(void *);
344 static bool valid_arg_len(struct linux_binprm
*bprm
, long len
)
346 return len
<= bprm
->p
;
349 #endif /* CONFIG_MMU */
352 * Create a new mm_struct and populate it with a temporary stack
353 * vm_area_struct. We don't have enough context at this point to set the stack
354 * flags, permissions, and offset, so we use temporary values. We'll update
355 * them later in setup_arg_pages().
357 static int bprm_mm_init(struct linux_binprm
*bprm
)
360 struct mm_struct
*mm
= NULL
;
362 bprm
->mm
= mm
= mm_alloc();
367 /* Save current stack limit for all calculations made during exec. */
368 task_lock(current
->group_leader
);
369 bprm
->rlim_stack
= current
->signal
->rlim
[RLIMIT_STACK
];
370 task_unlock(current
->group_leader
);
372 err
= __bprm_mm_init(bprm
);
387 struct user_arg_ptr
{
392 const char __user
*const __user
*native
;
394 const compat_uptr_t __user
*compat
;
399 static const char __user
*get_user_arg_ptr(struct user_arg_ptr argv
, int nr
)
401 const char __user
*native
;
404 if (unlikely(argv
.is_compat
)) {
405 compat_uptr_t compat
;
407 if (get_user(compat
, argv
.ptr
.compat
+ nr
))
408 return ERR_PTR(-EFAULT
);
410 return compat_ptr(compat
);
414 if (get_user(native
, argv
.ptr
.native
+ nr
))
415 return ERR_PTR(-EFAULT
);
421 * count() counts the number of strings in array ARGV.
423 static int count(struct user_arg_ptr argv
, int max
)
427 if (argv
.ptr
.native
!= NULL
) {
429 const char __user
*p
= get_user_arg_ptr(argv
, i
);
441 if (fatal_signal_pending(current
))
442 return -ERESTARTNOHAND
;
449 static int prepare_arg_pages(struct linux_binprm
*bprm
,
450 struct user_arg_ptr argv
, struct user_arg_ptr envp
)
452 unsigned long limit
, ptr_size
;
454 bprm
->argc
= count(argv
, MAX_ARG_STRINGS
);
458 bprm
->envc
= count(envp
, MAX_ARG_STRINGS
);
463 * Limit to 1/4 of the max stack size or 3/4 of _STK_LIM
464 * (whichever is smaller) for the argv+env strings.
466 * - the remaining binfmt code will not run out of stack space,
467 * - the program will have a reasonable amount of stack left
470 limit
= _STK_LIM
/ 4 * 3;
471 limit
= min(limit
, bprm
->rlim_stack
.rlim_cur
/ 4);
473 * We've historically supported up to 32 pages (ARG_MAX)
474 * of argument strings even with small stacks
476 limit
= max_t(unsigned long, limit
, ARG_MAX
);
478 * We must account for the size of all the argv and envp pointers to
479 * the argv and envp strings, since they will also take up space in
480 * the stack. They aren't stored until much later when we can't
481 * signal to the parent that the child has run out of stack space.
482 * Instead, calculate it here so it's possible to fail gracefully.
484 ptr_size
= (bprm
->argc
+ bprm
->envc
) * sizeof(void *);
485 if (limit
<= ptr_size
)
489 bprm
->argmin
= bprm
->p
- limit
;
494 * 'copy_strings()' copies argument/environment strings from the old
495 * processes's memory to the new process's stack. The call to get_user_pages()
496 * ensures the destination page is created and not swapped out.
498 static int copy_strings(int argc
, struct user_arg_ptr argv
,
499 struct linux_binprm
*bprm
)
501 struct page
*kmapped_page
= NULL
;
503 unsigned long kpos
= 0;
507 const char __user
*str
;
512 str
= get_user_arg_ptr(argv
, argc
);
516 len
= strnlen_user(str
, MAX_ARG_STRLEN
);
521 if (!valid_arg_len(bprm
, len
))
524 /* We're going to work our way backwords. */
529 if (bprm
->p
< bprm
->argmin
)
534 int offset
, bytes_to_copy
;
536 if (fatal_signal_pending(current
)) {
537 ret
= -ERESTARTNOHAND
;
542 offset
= pos
% PAGE_SIZE
;
546 bytes_to_copy
= offset
;
547 if (bytes_to_copy
> len
)
550 offset
-= bytes_to_copy
;
551 pos
-= bytes_to_copy
;
552 str
-= bytes_to_copy
;
553 len
-= bytes_to_copy
;
555 if (!kmapped_page
|| kpos
!= (pos
& PAGE_MASK
)) {
558 page
= get_arg_page(bprm
, pos
, 1);
565 flush_kernel_dcache_page(kmapped_page
);
566 kunmap(kmapped_page
);
567 put_arg_page(kmapped_page
);
570 kaddr
= kmap(kmapped_page
);
571 kpos
= pos
& PAGE_MASK
;
572 flush_arg_page(bprm
, kpos
, kmapped_page
);
574 if (copy_from_user(kaddr
+offset
, str
, bytes_to_copy
)) {
583 flush_kernel_dcache_page(kmapped_page
);
584 kunmap(kmapped_page
);
585 put_arg_page(kmapped_page
);
591 * Like copy_strings, but get argv and its values from kernel memory.
593 int copy_strings_kernel(int argc
, const char *const *__argv
,
594 struct linux_binprm
*bprm
)
597 mm_segment_t oldfs
= get_fs();
598 struct user_arg_ptr argv
= {
599 .ptr
.native
= (const char __user
*const __user
*)__argv
,
603 r
= copy_strings(argc
, argv
, bprm
);
608 EXPORT_SYMBOL(copy_strings_kernel
);
613 * During bprm_mm_init(), we create a temporary stack at STACK_TOP_MAX. Once
614 * the binfmt code determines where the new stack should reside, we shift it to
615 * its final location. The process proceeds as follows:
617 * 1) Use shift to calculate the new vma endpoints.
618 * 2) Extend vma to cover both the old and new ranges. This ensures the
619 * arguments passed to subsequent functions are consistent.
620 * 3) Move vma's page tables to the new range.
621 * 4) Free up any cleared pgd range.
622 * 5) Shrink the vma to cover only the new range.
624 static int shift_arg_pages(struct vm_area_struct
*vma
, unsigned long shift
)
626 struct mm_struct
*mm
= vma
->vm_mm
;
627 unsigned long old_start
= vma
->vm_start
;
628 unsigned long old_end
= vma
->vm_end
;
629 unsigned long length
= old_end
- old_start
;
630 unsigned long new_start
= old_start
- shift
;
631 unsigned long new_end
= old_end
- shift
;
632 struct mmu_gather tlb
;
634 BUG_ON(new_start
> new_end
);
637 * ensure there are no vmas between where we want to go
640 if (vma
!= find_vma(mm
, new_start
))
644 * cover the whole range: [new_start, old_end)
646 if (vma_adjust(vma
, new_start
, old_end
, vma
->vm_pgoff
, NULL
))
650 * move the page tables downwards, on failure we rely on
651 * process cleanup to remove whatever mess we made.
653 if (length
!= move_page_tables(vma
, old_start
,
654 vma
, new_start
, length
, false))
658 tlb_gather_mmu(&tlb
, mm
, old_start
, old_end
);
659 if (new_end
> old_start
) {
661 * when the old and new regions overlap clear from new_end.
663 free_pgd_range(&tlb
, new_end
, old_end
, new_end
,
664 vma
->vm_next
? vma
->vm_next
->vm_start
: USER_PGTABLES_CEILING
);
667 * otherwise, clean from old_start; this is done to not touch
668 * the address space in [new_end, old_start) some architectures
669 * have constraints on va-space that make this illegal (IA64) -
670 * for the others its just a little faster.
672 free_pgd_range(&tlb
, old_start
, old_end
, new_end
,
673 vma
->vm_next
? vma
->vm_next
->vm_start
: USER_PGTABLES_CEILING
);
675 tlb_finish_mmu(&tlb
, old_start
, old_end
);
678 * Shrink the vma to just the new range. Always succeeds.
680 vma_adjust(vma
, new_start
, new_end
, vma
->vm_pgoff
, NULL
);
686 * Finalizes the stack vm_area_struct. The flags and permissions are updated,
687 * the stack is optionally relocated, and some extra space is added.
689 int setup_arg_pages(struct linux_binprm
*bprm
,
690 unsigned long stack_top
,
691 int executable_stack
)
694 unsigned long stack_shift
;
695 struct mm_struct
*mm
= current
->mm
;
696 struct vm_area_struct
*vma
= bprm
->vma
;
697 struct vm_area_struct
*prev
= NULL
;
698 unsigned long vm_flags
;
699 unsigned long stack_base
;
700 unsigned long stack_size
;
701 unsigned long stack_expand
;
702 unsigned long rlim_stack
;
704 #ifdef CONFIG_STACK_GROWSUP
705 /* Limit stack size */
706 stack_base
= bprm
->rlim_stack
.rlim_max
;
707 if (stack_base
> STACK_SIZE_MAX
)
708 stack_base
= STACK_SIZE_MAX
;
710 /* Add space for stack randomization. */
711 stack_base
+= (STACK_RND_MASK
<< PAGE_SHIFT
);
713 /* Make sure we didn't let the argument array grow too large. */
714 if (vma
->vm_end
- vma
->vm_start
> stack_base
)
717 stack_base
= PAGE_ALIGN(stack_top
- stack_base
);
719 stack_shift
= vma
->vm_start
- stack_base
;
720 mm
->arg_start
= bprm
->p
- stack_shift
;
721 bprm
->p
= vma
->vm_end
- stack_shift
;
723 stack_top
= arch_align_stack(stack_top
);
724 stack_top
= PAGE_ALIGN(stack_top
);
726 if (unlikely(stack_top
< mmap_min_addr
) ||
727 unlikely(vma
->vm_end
- vma
->vm_start
>= stack_top
- mmap_min_addr
))
730 stack_shift
= vma
->vm_end
- stack_top
;
732 bprm
->p
-= stack_shift
;
733 mm
->arg_start
= bprm
->p
;
737 bprm
->loader
-= stack_shift
;
738 bprm
->exec
-= stack_shift
;
740 if (down_write_killable(&mm
->mmap_sem
))
743 vm_flags
= VM_STACK_FLAGS
;
746 * Adjust stack execute permissions; explicitly enable for
747 * EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X and leave alone
748 * (arch default) otherwise.
750 if (unlikely(executable_stack
== EXSTACK_ENABLE_X
))
752 else if (executable_stack
== EXSTACK_DISABLE_X
)
753 vm_flags
&= ~VM_EXEC
;
754 vm_flags
|= mm
->def_flags
;
755 vm_flags
|= VM_STACK_INCOMPLETE_SETUP
;
757 ret
= mprotect_fixup(vma
, &prev
, vma
->vm_start
, vma
->vm_end
,
763 if (unlikely(vm_flags
& VM_EXEC
)) {
764 pr_warn_once("process '%pD4' started with executable stack\n",
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
= bprm
->rlim_stack
.rlim_cur
& 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')
886 struct file
*open_exec(const char *name
)
888 struct filename
*filename
= getname_kernel(name
);
889 struct file
*f
= ERR_CAST(filename
);
891 if (!IS_ERR(filename
)) {
892 f
= do_open_execat(AT_FDCWD
, filename
, 0);
897 EXPORT_SYMBOL(open_exec
);
899 int kernel_read_file(struct file
*file
, void **buf
, loff_t
*size
,
900 loff_t max_size
, enum kernel_read_file_id id
)
906 if (!S_ISREG(file_inode(file
)->i_mode
) || max_size
< 0)
909 ret
= deny_write_access(file
);
913 ret
= security_kernel_read_file(file
, id
);
917 i_size
= i_size_read(file_inode(file
));
922 if (i_size
> SIZE_MAX
|| (max_size
> 0 && i_size
> max_size
)) {
927 if (id
!= READING_FIRMWARE_PREALLOC_BUFFER
)
928 *buf
= vmalloc(i_size
);
935 while (pos
< i_size
) {
936 bytes
= kernel_read(file
, *buf
+ pos
, i_size
- pos
, &pos
);
951 ret
= security_kernel_post_read_file(file
, *buf
, i_size
, id
);
957 if (id
!= READING_FIRMWARE_PREALLOC_BUFFER
) {
964 allow_write_access(file
);
967 EXPORT_SYMBOL_GPL(kernel_read_file
);
969 int kernel_read_file_from_path(const char *path
, void **buf
, loff_t
*size
,
970 loff_t max_size
, enum kernel_read_file_id id
)
978 file
= filp_open(path
, O_RDONLY
, 0);
980 return PTR_ERR(file
);
982 ret
= kernel_read_file(file
, buf
, size
, max_size
, id
);
986 EXPORT_SYMBOL_GPL(kernel_read_file_from_path
);
988 int kernel_read_file_from_path_initns(const char *path
, void **buf
,
989 loff_t
*size
, loff_t max_size
,
990 enum kernel_read_file_id id
)
999 task_lock(&init_task
);
1000 get_fs_root(init_task
.fs
, &root
);
1001 task_unlock(&init_task
);
1003 file
= file_open_root(root
.dentry
, root
.mnt
, path
, O_RDONLY
, 0);
1006 return PTR_ERR(file
);
1008 ret
= kernel_read_file(file
, buf
, size
, max_size
, id
);
1012 EXPORT_SYMBOL_GPL(kernel_read_file_from_path_initns
);
1014 int kernel_read_file_from_fd(int fd
, void **buf
, loff_t
*size
, loff_t max_size
,
1015 enum kernel_read_file_id id
)
1017 struct fd f
= fdget(fd
);
1023 ret
= kernel_read_file(f
.file
, buf
, size
, max_size
, id
);
1028 EXPORT_SYMBOL_GPL(kernel_read_file_from_fd
);
1030 ssize_t
read_code(struct file
*file
, unsigned long addr
, loff_t pos
, size_t len
)
1032 ssize_t res
= vfs_read(file
, (void __user
*)addr
, len
, &pos
);
1034 flush_icache_range(addr
, addr
+ len
);
1037 EXPORT_SYMBOL(read_code
);
1039 static int exec_mmap(struct mm_struct
*mm
)
1041 struct task_struct
*tsk
;
1042 struct mm_struct
*old_mm
, *active_mm
;
1044 /* Notify parent that we're no longer interested in the old VM */
1046 old_mm
= current
->mm
;
1047 exec_mm_release(tsk
, old_mm
);
1050 sync_mm_rss(old_mm
);
1052 * Make sure that if there is a core dump in progress
1053 * for the old mm, we get out and die instead of going
1054 * through with the exec. We must hold mmap_sem around
1055 * checking core_state and changing tsk->mm.
1057 down_read(&old_mm
->mmap_sem
);
1058 if (unlikely(old_mm
->core_state
)) {
1059 up_read(&old_mm
->mmap_sem
);
1064 active_mm
= tsk
->active_mm
;
1065 membarrier_exec_mmap(mm
);
1067 tsk
->active_mm
= mm
;
1068 activate_mm(active_mm
, mm
);
1069 tsk
->mm
->vmacache_seqnum
= 0;
1070 vmacache_flush(tsk
);
1073 up_read(&old_mm
->mmap_sem
);
1074 BUG_ON(active_mm
!= old_mm
);
1075 setmax_mm_hiwater_rss(&tsk
->signal
->maxrss
, old_mm
);
1076 mm_update_next_owner(old_mm
);
1085 * This function makes sure the current process has its own signal table,
1086 * so that flush_signal_handlers can later reset the handlers without
1087 * disturbing other processes. (Other processes might share the signal
1088 * table via the CLONE_SIGHAND option to clone().)
1090 static int de_thread(struct task_struct
*tsk
)
1092 struct signal_struct
*sig
= tsk
->signal
;
1093 struct sighand_struct
*oldsighand
= tsk
->sighand
;
1094 spinlock_t
*lock
= &oldsighand
->siglock
;
1096 if (thread_group_empty(tsk
))
1097 goto no_thread_group
;
1100 * Kill all other threads in the thread group.
1102 spin_lock_irq(lock
);
1103 if (signal_group_exit(sig
)) {
1105 * Another group action in progress, just
1106 * return so that the signal is processed.
1108 spin_unlock_irq(lock
);
1112 sig
->group_exit_task
= tsk
;
1113 sig
->notify_count
= zap_other_threads(tsk
);
1114 if (!thread_group_leader(tsk
))
1115 sig
->notify_count
--;
1117 while (sig
->notify_count
) {
1118 __set_current_state(TASK_KILLABLE
);
1119 spin_unlock_irq(lock
);
1121 if (__fatal_signal_pending(tsk
))
1123 spin_lock_irq(lock
);
1125 spin_unlock_irq(lock
);
1128 * At this point all other threads have exited, all we have to
1129 * do is to wait for the thread group leader to become inactive,
1130 * and to assume its PID:
1132 if (!thread_group_leader(tsk
)) {
1133 struct task_struct
*leader
= tsk
->group_leader
;
1136 cgroup_threadgroup_change_begin(tsk
);
1137 write_lock_irq(&tasklist_lock
);
1139 * Do this under tasklist_lock to ensure that
1140 * exit_notify() can't miss ->group_exit_task
1142 sig
->notify_count
= -1;
1143 if (likely(leader
->exit_state
))
1145 __set_current_state(TASK_KILLABLE
);
1146 write_unlock_irq(&tasklist_lock
);
1147 cgroup_threadgroup_change_end(tsk
);
1149 if (__fatal_signal_pending(tsk
))
1154 * The only record we have of the real-time age of a
1155 * process, regardless of execs it's done, is start_time.
1156 * All the past CPU time is accumulated in signal_struct
1157 * from sister threads now dead. But in this non-leader
1158 * exec, nothing survives from the original leader thread,
1159 * whose birth marks the true age of this process now.
1160 * When we take on its identity by switching to its PID, we
1161 * also take its birthdate (always earlier than our own).
1163 tsk
->start_time
= leader
->start_time
;
1164 tsk
->start_boottime
= leader
->start_boottime
;
1166 BUG_ON(!same_thread_group(leader
, tsk
));
1167 BUG_ON(has_group_leader_pid(tsk
));
1169 * An exec() starts a new thread group with the
1170 * TGID of the previous thread group. Rehash the
1171 * two threads with a switched PID, and release
1172 * the former thread group leader:
1175 /* Become a process group leader with the old leader's pid.
1176 * The old leader becomes a thread of the this thread group.
1177 * Note: The old leader also uses this pid until release_task
1178 * is called. Odd but simple and correct.
1180 tsk
->pid
= leader
->pid
;
1181 change_pid(tsk
, PIDTYPE_PID
, task_pid(leader
));
1182 transfer_pid(leader
, tsk
, PIDTYPE_TGID
);
1183 transfer_pid(leader
, tsk
, PIDTYPE_PGID
);
1184 transfer_pid(leader
, tsk
, PIDTYPE_SID
);
1186 list_replace_rcu(&leader
->tasks
, &tsk
->tasks
);
1187 list_replace_init(&leader
->sibling
, &tsk
->sibling
);
1189 tsk
->group_leader
= tsk
;
1190 leader
->group_leader
= tsk
;
1192 tsk
->exit_signal
= SIGCHLD
;
1193 leader
->exit_signal
= -1;
1195 BUG_ON(leader
->exit_state
!= EXIT_ZOMBIE
);
1196 leader
->exit_state
= EXIT_DEAD
;
1199 * We are going to release_task()->ptrace_unlink() silently,
1200 * the tracer can sleep in do_wait(). EXIT_DEAD guarantees
1201 * the tracer wont't block again waiting for this thread.
1203 if (unlikely(leader
->ptrace
))
1204 __wake_up_parent(leader
, leader
->parent
);
1205 write_unlock_irq(&tasklist_lock
);
1206 cgroup_threadgroup_change_end(tsk
);
1208 release_task(leader
);
1211 sig
->group_exit_task
= NULL
;
1212 sig
->notify_count
= 0;
1215 /* we have changed execution domain */
1216 tsk
->exit_signal
= SIGCHLD
;
1218 #ifdef CONFIG_POSIX_TIMERS
1220 flush_itimer_signals();
1223 if (refcount_read(&oldsighand
->count
) != 1) {
1224 struct sighand_struct
*newsighand
;
1226 * This ->sighand is shared with the CLONE_SIGHAND
1227 * but not CLONE_THREAD task, switch to the new one.
1229 newsighand
= kmem_cache_alloc(sighand_cachep
, GFP_KERNEL
);
1233 refcount_set(&newsighand
->count
, 1);
1234 memcpy(newsighand
->action
, oldsighand
->action
,
1235 sizeof(newsighand
->action
));
1237 write_lock_irq(&tasklist_lock
);
1238 spin_lock(&oldsighand
->siglock
);
1239 rcu_assign_pointer(tsk
->sighand
, newsighand
);
1240 spin_unlock(&oldsighand
->siglock
);
1241 write_unlock_irq(&tasklist_lock
);
1243 __cleanup_sighand(oldsighand
);
1246 BUG_ON(!thread_group_leader(tsk
));
1250 /* protects against exit_notify() and __exit_signal() */
1251 read_lock(&tasklist_lock
);
1252 sig
->group_exit_task
= NULL
;
1253 sig
->notify_count
= 0;
1254 read_unlock(&tasklist_lock
);
1258 char *__get_task_comm(char *buf
, size_t buf_size
, struct task_struct
*tsk
)
1261 strncpy(buf
, tsk
->comm
, buf_size
);
1265 EXPORT_SYMBOL_GPL(__get_task_comm
);
1268 * These functions flushes out all traces of the currently running executable
1269 * so that a new one can be started
1272 void __set_task_comm(struct task_struct
*tsk
, const char *buf
, bool exec
)
1275 trace_task_rename(tsk
, buf
);
1276 strlcpy(tsk
->comm
, buf
, sizeof(tsk
->comm
));
1278 perf_event_comm(tsk
, exec
);
1282 * Calling this is the point of no return. None of the failures will be
1283 * seen by userspace since either the process is already taking a fatal
1284 * signal (via de_thread() or coredump), or will have SEGV raised
1285 * (after exec_mmap()) by search_binary_handlers (see below).
1287 int flush_old_exec(struct linux_binprm
* bprm
)
1292 * Make sure we have a private signal table and that
1293 * we are unassociated from the previous thread group.
1295 retval
= de_thread(current
);
1300 * Must be called _before_ exec_mmap() as bprm->mm is
1301 * not visibile until then. This also enables the update
1304 set_mm_exe_file(bprm
->mm
, bprm
->file
);
1307 * Release all of the old mmap stuff
1309 acct_arg_size(bprm
, 0);
1310 retval
= exec_mmap(bprm
->mm
);
1315 * After clearing bprm->mm (to mark that current is using the
1316 * prepared mm now), we have nothing left of the original
1317 * process. If anything from here on returns an error, the check
1318 * in search_binary_handler() will SEGV current.
1323 current
->flags
&= ~(PF_RANDOMIZE
| PF_FORKNOEXEC
| PF_KTHREAD
|
1324 PF_NOFREEZE
| PF_NO_SETAFFINITY
);
1326 current
->personality
&= ~bprm
->per_clear
;
1329 * We have to apply CLOEXEC before we change whether the process is
1330 * dumpable (in setup_new_exec) to avoid a race with a process in userspace
1331 * trying to access the should-be-closed file descriptors of a process
1332 * undergoing exec(2).
1334 do_close_on_exec(current
->files
);
1340 EXPORT_SYMBOL(flush_old_exec
);
1342 void would_dump(struct linux_binprm
*bprm
, struct file
*file
)
1344 struct inode
*inode
= file_inode(file
);
1345 if (inode_permission(inode
, MAY_READ
) < 0) {
1346 struct user_namespace
*old
, *user_ns
;
1347 bprm
->interp_flags
|= BINPRM_FLAGS_ENFORCE_NONDUMP
;
1349 /* Ensure mm->user_ns contains the executable */
1350 user_ns
= old
= bprm
->mm
->user_ns
;
1351 while ((user_ns
!= &init_user_ns
) &&
1352 !privileged_wrt_inode_uidgid(user_ns
, inode
))
1353 user_ns
= user_ns
->parent
;
1355 if (old
!= user_ns
) {
1356 bprm
->mm
->user_ns
= get_user_ns(user_ns
);
1361 EXPORT_SYMBOL(would_dump
);
1363 void setup_new_exec(struct linux_binprm
* bprm
)
1366 * Once here, prepare_binrpm() will not be called any more, so
1367 * the final state of setuid/setgid/fscaps can be merged into the
1370 bprm
->secureexec
|= bprm
->cap_elevated
;
1372 if (bprm
->secureexec
) {
1373 /* Make sure parent cannot signal privileged process. */
1374 current
->pdeath_signal
= 0;
1377 * For secureexec, reset the stack limit to sane default to
1378 * avoid bad behavior from the prior rlimits. This has to
1379 * happen before arch_pick_mmap_layout(), which examines
1380 * RLIMIT_STACK, but after the point of no return to avoid
1381 * needing to clean up the change on failure.
1383 if (bprm
->rlim_stack
.rlim_cur
> _STK_LIM
)
1384 bprm
->rlim_stack
.rlim_cur
= _STK_LIM
;
1387 arch_pick_mmap_layout(current
->mm
, &bprm
->rlim_stack
);
1389 current
->sas_ss_sp
= current
->sas_ss_size
= 0;
1392 * Figure out dumpability. Note that this checking only of current
1393 * is wrong, but userspace depends on it. This should be testing
1394 * bprm->secureexec instead.
1396 if (bprm
->interp_flags
& BINPRM_FLAGS_ENFORCE_NONDUMP
||
1397 !(uid_eq(current_euid(), current_uid()) &&
1398 gid_eq(current_egid(), current_gid())))
1399 set_dumpable(current
->mm
, suid_dumpable
);
1401 set_dumpable(current
->mm
, SUID_DUMP_USER
);
1403 arch_setup_new_exec();
1405 __set_task_comm(current
, kbasename(bprm
->filename
), true);
1407 /* Set the new mm task size. We have to do that late because it may
1408 * depend on TIF_32BIT which is only updated in flush_thread() on
1409 * some architectures like powerpc
1411 current
->mm
->task_size
= TASK_SIZE
;
1413 /* An exec changes our domain. We are no longer part of the thread
1415 current
->self_exec_id
++;
1416 flush_signal_handlers(current
, 0);
1418 EXPORT_SYMBOL(setup_new_exec
);
1420 /* Runs immediately before start_thread() takes over. */
1421 void finalize_exec(struct linux_binprm
*bprm
)
1423 /* Store any stack rlimit changes before starting thread. */
1424 task_lock(current
->group_leader
);
1425 current
->signal
->rlim
[RLIMIT_STACK
] = bprm
->rlim_stack
;
1426 task_unlock(current
->group_leader
);
1428 EXPORT_SYMBOL(finalize_exec
);
1431 * Prepare credentials and lock ->cred_guard_mutex.
1432 * install_exec_creds() commits the new creds and drops the lock.
1433 * Or, if exec fails before, free_bprm() should release ->cred and
1436 static int prepare_bprm_creds(struct linux_binprm
*bprm
)
1438 if (mutex_lock_interruptible(¤t
->signal
->cred_guard_mutex
))
1439 return -ERESTARTNOINTR
;
1441 bprm
->cred
= prepare_exec_creds();
1442 if (likely(bprm
->cred
))
1445 mutex_unlock(¤t
->signal
->cred_guard_mutex
);
1449 static void free_bprm(struct linux_binprm
*bprm
)
1451 free_arg_pages(bprm
);
1453 mutex_unlock(¤t
->signal
->cred_guard_mutex
);
1454 abort_creds(bprm
->cred
);
1457 allow_write_access(bprm
->file
);
1460 /* If a binfmt changed the interp, free it. */
1461 if (bprm
->interp
!= bprm
->filename
)
1462 kfree(bprm
->interp
);
1466 int bprm_change_interp(const char *interp
, struct linux_binprm
*bprm
)
1468 /* If a binfmt changed the interp, free it first. */
1469 if (bprm
->interp
!= bprm
->filename
)
1470 kfree(bprm
->interp
);
1471 bprm
->interp
= kstrdup(interp
, GFP_KERNEL
);
1476 EXPORT_SYMBOL(bprm_change_interp
);
1479 * install the new credentials for this executable
1481 void install_exec_creds(struct linux_binprm
*bprm
)
1483 security_bprm_committing_creds(bprm
);
1485 commit_creds(bprm
->cred
);
1489 * Disable monitoring for regular users
1490 * when executing setuid binaries. Must
1491 * wait until new credentials are committed
1492 * by commit_creds() above
1494 if (get_dumpable(current
->mm
) != SUID_DUMP_USER
)
1495 perf_event_exit_task(current
);
1497 * cred_guard_mutex must be held at least to this point to prevent
1498 * ptrace_attach() from altering our determination of the task's
1499 * credentials; any time after this it may be unlocked.
1501 security_bprm_committed_creds(bprm
);
1502 mutex_unlock(¤t
->signal
->cred_guard_mutex
);
1504 EXPORT_SYMBOL(install_exec_creds
);
1507 * determine how safe it is to execute the proposed program
1508 * - the caller must hold ->cred_guard_mutex to protect against
1509 * PTRACE_ATTACH or seccomp thread-sync
1511 static void check_unsafe_exec(struct linux_binprm
*bprm
)
1513 struct task_struct
*p
= current
, *t
;
1517 bprm
->unsafe
|= LSM_UNSAFE_PTRACE
;
1520 * This isn't strictly necessary, but it makes it harder for LSMs to
1523 if (task_no_new_privs(current
))
1524 bprm
->unsafe
|= LSM_UNSAFE_NO_NEW_PRIVS
;
1528 spin_lock(&p
->fs
->lock
);
1530 while_each_thread(p
, t
) {
1536 if (p
->fs
->users
> n_fs
)
1537 bprm
->unsafe
|= LSM_UNSAFE_SHARE
;
1540 spin_unlock(&p
->fs
->lock
);
1543 static void bprm_fill_uid(struct linux_binprm
*bprm
)
1545 struct inode
*inode
;
1551 * Since this can be called multiple times (via prepare_binprm),
1552 * we must clear any previous work done when setting set[ug]id
1553 * bits from any earlier bprm->file uses (for example when run
1554 * first for a setuid script then again for its interpreter).
1556 bprm
->cred
->euid
= current_euid();
1557 bprm
->cred
->egid
= current_egid();
1559 if (!mnt_may_suid(bprm
->file
->f_path
.mnt
))
1562 if (task_no_new_privs(current
))
1565 inode
= bprm
->file
->f_path
.dentry
->d_inode
;
1566 mode
= READ_ONCE(inode
->i_mode
);
1567 if (!(mode
& (S_ISUID
|S_ISGID
)))
1570 /* Be careful if suid/sgid is set */
1573 /* reload atomically mode/uid/gid now that lock held */
1574 mode
= inode
->i_mode
;
1577 inode_unlock(inode
);
1579 /* We ignore suid/sgid if there are no mappings for them in the ns */
1580 if (!kuid_has_mapping(bprm
->cred
->user_ns
, uid
) ||
1581 !kgid_has_mapping(bprm
->cred
->user_ns
, gid
))
1584 if (mode
& S_ISUID
) {
1585 bprm
->per_clear
|= PER_CLEAR_ON_SETID
;
1586 bprm
->cred
->euid
= uid
;
1589 if ((mode
& (S_ISGID
| S_IXGRP
)) == (S_ISGID
| S_IXGRP
)) {
1590 bprm
->per_clear
|= PER_CLEAR_ON_SETID
;
1591 bprm
->cred
->egid
= gid
;
1596 * Fill the binprm structure from the inode.
1597 * Check permissions, then read the first BINPRM_BUF_SIZE bytes
1599 * This may be called multiple times for binary chains (scripts for example).
1601 int prepare_binprm(struct linux_binprm
*bprm
)
1606 bprm_fill_uid(bprm
);
1608 /* fill in binprm security blob */
1609 retval
= security_bprm_set_creds(bprm
);
1612 bprm
->called_set_creds
= 1;
1614 memset(bprm
->buf
, 0, BINPRM_BUF_SIZE
);
1615 return kernel_read(bprm
->file
, bprm
->buf
, BINPRM_BUF_SIZE
, &pos
);
1618 EXPORT_SYMBOL(prepare_binprm
);
1621 * Arguments are '\0' separated strings found at the location bprm->p
1622 * points to; chop off the first by relocating brpm->p to right after
1623 * the first '\0' encountered.
1625 int remove_arg_zero(struct linux_binprm
*bprm
)
1628 unsigned long offset
;
1636 offset
= bprm
->p
& ~PAGE_MASK
;
1637 page
= get_arg_page(bprm
, bprm
->p
, 0);
1642 kaddr
= kmap_atomic(page
);
1644 for (; offset
< PAGE_SIZE
&& kaddr
[offset
];
1645 offset
++, bprm
->p
++)
1648 kunmap_atomic(kaddr
);
1650 } while (offset
== PAGE_SIZE
);
1659 EXPORT_SYMBOL(remove_arg_zero
);
1661 #define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e))
1663 * cycle the list of binary formats handler, until one recognizes the image
1665 int search_binary_handler(struct linux_binprm
*bprm
)
1667 bool need_retry
= IS_ENABLED(CONFIG_MODULES
);
1668 struct linux_binfmt
*fmt
;
1671 /* This allows 4 levels of binfmt rewrites before failing hard. */
1672 if (bprm
->recursion_depth
> 5)
1675 retval
= security_bprm_check(bprm
);
1681 read_lock(&binfmt_lock
);
1682 list_for_each_entry(fmt
, &formats
, lh
) {
1683 if (!try_module_get(fmt
->module
))
1685 read_unlock(&binfmt_lock
);
1687 bprm
->recursion_depth
++;
1688 retval
= fmt
->load_binary(bprm
);
1689 bprm
->recursion_depth
--;
1691 read_lock(&binfmt_lock
);
1693 if (retval
< 0 && !bprm
->mm
) {
1694 /* we got to flush_old_exec() and failed after it */
1695 read_unlock(&binfmt_lock
);
1696 force_sigsegv(SIGSEGV
);
1699 if (retval
!= -ENOEXEC
|| !bprm
->file
) {
1700 read_unlock(&binfmt_lock
);
1704 read_unlock(&binfmt_lock
);
1707 if (printable(bprm
->buf
[0]) && printable(bprm
->buf
[1]) &&
1708 printable(bprm
->buf
[2]) && printable(bprm
->buf
[3]))
1710 if (request_module("binfmt-%04x", *(ushort
*)(bprm
->buf
+ 2)) < 0)
1718 EXPORT_SYMBOL(search_binary_handler
);
1720 static int exec_binprm(struct linux_binprm
*bprm
)
1722 pid_t old_pid
, old_vpid
;
1725 /* Need to fetch pid before load_binary changes it */
1726 old_pid
= current
->pid
;
1728 old_vpid
= task_pid_nr_ns(current
, task_active_pid_ns(current
->parent
));
1731 ret
= search_binary_handler(bprm
);
1734 trace_sched_process_exec(current
, old_pid
, bprm
);
1735 ptrace_event(PTRACE_EVENT_EXEC
, old_vpid
);
1736 proc_exec_connector(current
);
1743 * sys_execve() executes a new program.
1745 static int __do_execve_file(int fd
, struct filename
*filename
,
1746 struct user_arg_ptr argv
,
1747 struct user_arg_ptr envp
,
1748 int flags
, struct file
*file
)
1750 char *pathbuf
= NULL
;
1751 struct linux_binprm
*bprm
;
1752 struct files_struct
*displaced
;
1755 if (IS_ERR(filename
))
1756 return PTR_ERR(filename
);
1759 * We move the actual failure in case of RLIMIT_NPROC excess from
1760 * set*uid() to execve() because too many poorly written programs
1761 * don't check setuid() return code. Here we additionally recheck
1762 * whether NPROC limit is still exceeded.
1764 if ((current
->flags
& PF_NPROC_EXCEEDED
) &&
1765 atomic_read(¤t_user()->processes
) > rlimit(RLIMIT_NPROC
)) {
1770 /* We're below the limit (still or again), so we don't want to make
1771 * further execve() calls fail. */
1772 current
->flags
&= ~PF_NPROC_EXCEEDED
;
1774 retval
= unshare_files(&displaced
);
1779 bprm
= kzalloc(sizeof(*bprm
), GFP_KERNEL
);
1783 retval
= prepare_bprm_creds(bprm
);
1787 check_unsafe_exec(bprm
);
1788 current
->in_execve
= 1;
1791 file
= do_open_execat(fd
, filename
, flags
);
1792 retval
= PTR_ERR(file
);
1800 bprm
->filename
= "none";
1801 } else if (fd
== AT_FDCWD
|| filename
->name
[0] == '/') {
1802 bprm
->filename
= filename
->name
;
1804 if (filename
->name
[0] == '\0')
1805 pathbuf
= kasprintf(GFP_KERNEL
, "/dev/fd/%d", fd
);
1807 pathbuf
= kasprintf(GFP_KERNEL
, "/dev/fd/%d/%s",
1808 fd
, filename
->name
);
1814 * Record that a name derived from an O_CLOEXEC fd will be
1815 * inaccessible after exec. Relies on having exclusive access to
1816 * current->files (due to unshare_files above).
1818 if (close_on_exec(fd
, rcu_dereference_raw(current
->files
->fdt
)))
1819 bprm
->interp_flags
|= BINPRM_FLAGS_PATH_INACCESSIBLE
;
1820 bprm
->filename
= pathbuf
;
1822 bprm
->interp
= bprm
->filename
;
1824 retval
= bprm_mm_init(bprm
);
1828 retval
= prepare_arg_pages(bprm
, argv
, envp
);
1832 retval
= prepare_binprm(bprm
);
1836 retval
= copy_strings_kernel(1, &bprm
->filename
, bprm
);
1840 bprm
->exec
= bprm
->p
;
1841 retval
= copy_strings(bprm
->envc
, envp
, bprm
);
1845 retval
= copy_strings(bprm
->argc
, argv
, bprm
);
1849 would_dump(bprm
, bprm
->file
);
1851 retval
= exec_binprm(bprm
);
1855 /* execve succeeded */
1856 current
->fs
->in_exec
= 0;
1857 current
->in_execve
= 0;
1858 rseq_execve(current
);
1859 acct_update_integrals(current
);
1860 task_numa_free(current
, false);
1866 put_files_struct(displaced
);
1871 acct_arg_size(bprm
, 0);
1876 current
->fs
->in_exec
= 0;
1877 current
->in_execve
= 0;
1885 reset_files_struct(displaced
);
1892 static int do_execveat_common(int fd
, struct filename
*filename
,
1893 struct user_arg_ptr argv
,
1894 struct user_arg_ptr envp
,
1897 return __do_execve_file(fd
, filename
, argv
, envp
, flags
, NULL
);
1900 int do_execve_file(struct file
*file
, void *__argv
, void *__envp
)
1902 struct user_arg_ptr argv
= { .ptr
.native
= __argv
};
1903 struct user_arg_ptr envp
= { .ptr
.native
= __envp
};
1905 return __do_execve_file(AT_FDCWD
, NULL
, argv
, envp
, 0, file
);
1908 int do_execve(struct filename
*filename
,
1909 const char __user
*const __user
*__argv
,
1910 const char __user
*const __user
*__envp
)
1912 struct user_arg_ptr argv
= { .ptr
.native
= __argv
};
1913 struct user_arg_ptr envp
= { .ptr
.native
= __envp
};
1914 return do_execveat_common(AT_FDCWD
, filename
, argv
, envp
, 0);
1917 int do_execveat(int fd
, struct filename
*filename
,
1918 const char __user
*const __user
*__argv
,
1919 const char __user
*const __user
*__envp
,
1922 struct user_arg_ptr argv
= { .ptr
.native
= __argv
};
1923 struct user_arg_ptr envp
= { .ptr
.native
= __envp
};
1925 return do_execveat_common(fd
, filename
, argv
, envp
, flags
);
1928 #ifdef CONFIG_COMPAT
1929 static int compat_do_execve(struct filename
*filename
,
1930 const compat_uptr_t __user
*__argv
,
1931 const compat_uptr_t __user
*__envp
)
1933 struct user_arg_ptr argv
= {
1935 .ptr
.compat
= __argv
,
1937 struct user_arg_ptr envp
= {
1939 .ptr
.compat
= __envp
,
1941 return do_execveat_common(AT_FDCWD
, filename
, argv
, envp
, 0);
1944 static int compat_do_execveat(int fd
, struct filename
*filename
,
1945 const compat_uptr_t __user
*__argv
,
1946 const compat_uptr_t __user
*__envp
,
1949 struct user_arg_ptr argv
= {
1951 .ptr
.compat
= __argv
,
1953 struct user_arg_ptr envp
= {
1955 .ptr
.compat
= __envp
,
1957 return do_execveat_common(fd
, filename
, argv
, envp
, flags
);
1961 void set_binfmt(struct linux_binfmt
*new)
1963 struct mm_struct
*mm
= current
->mm
;
1966 module_put(mm
->binfmt
->module
);
1970 __module_get(new->module
);
1972 EXPORT_SYMBOL(set_binfmt
);
1975 * set_dumpable stores three-value SUID_DUMP_* into mm->flags.
1977 void set_dumpable(struct mm_struct
*mm
, int value
)
1979 if (WARN_ON((unsigned)value
> SUID_DUMP_ROOT
))
1982 set_mask_bits(&mm
->flags
, MMF_DUMPABLE_MASK
, value
);
1985 SYSCALL_DEFINE3(execve
,
1986 const char __user
*, filename
,
1987 const char __user
*const __user
*, argv
,
1988 const char __user
*const __user
*, envp
)
1990 return do_execve(getname(filename
), argv
, envp
);
1993 SYSCALL_DEFINE5(execveat
,
1994 int, fd
, const char __user
*, filename
,
1995 const char __user
*const __user
*, argv
,
1996 const char __user
*const __user
*, envp
,
1999 int lookup_flags
= (flags
& AT_EMPTY_PATH
) ? LOOKUP_EMPTY
: 0;
2001 return do_execveat(fd
,
2002 getname_flags(filename
, lookup_flags
, NULL
),
2006 #ifdef CONFIG_COMPAT
2007 COMPAT_SYSCALL_DEFINE3(execve
, const char __user
*, filename
,
2008 const compat_uptr_t __user
*, argv
,
2009 const compat_uptr_t __user
*, envp
)
2011 return compat_do_execve(getname(filename
), argv
, envp
);
2014 COMPAT_SYSCALL_DEFINE5(execveat
, int, fd
,
2015 const char __user
*, filename
,
2016 const compat_uptr_t __user
*, argv
,
2017 const compat_uptr_t __user
*, envp
,
2020 int lookup_flags
= (flags
& AT_EMPTY_PATH
) ? LOOKUP_EMPTY
: 0;
2022 return compat_do_execveat(fd
,
2023 getname_flags(filename
, lookup_flags
, NULL
),