Commit | Line | Data |
---|---|---|
457c8996 | 1 | // SPDX-License-Identifier: GPL-2.0-only |
1da177e4 LT |
2 | /* |
3 | * linux/fs/exec.c | |
4 | * | |
5 | * Copyright (C) 1991, 1992 Linus Torvalds | |
6 | */ | |
7 | ||
8 | /* | |
9 | * #!-checking implemented by tytso. | |
10 | */ | |
11 | /* | |
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. | |
15 | * | |
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. | |
18 | * | |
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 | |
613cc2b6 | 23 | * formats. |
1da177e4 LT |
24 | */ |
25 | ||
1da177e4 LT |
26 | #include <linux/slab.h> |
27 | #include <linux/file.h> | |
9f3acc31 | 28 | #include <linux/fdtable.h> |
ba92a43d | 29 | #include <linux/mm.h> |
615d6e87 | 30 | #include <linux/vmacache.h> |
1da177e4 LT |
31 | #include <linux/stat.h> |
32 | #include <linux/fcntl.h> | |
ba92a43d | 33 | #include <linux/swap.h> |
74aadce9 | 34 | #include <linux/string.h> |
1da177e4 | 35 | #include <linux/init.h> |
6e84f315 | 36 | #include <linux/sched/mm.h> |
f7ccbae4 | 37 | #include <linux/sched/coredump.h> |
3f07c014 | 38 | #include <linux/sched/signal.h> |
6a3827d7 | 39 | #include <linux/sched/numa_balancing.h> |
29930025 | 40 | #include <linux/sched/task.h> |
ca5b172b | 41 | #include <linux/pagemap.h> |
cdd6c482 | 42 | #include <linux/perf_event.h> |
1da177e4 LT |
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> | |
1da177e4 | 48 | #include <linux/utsname.h> |
84d73786 | 49 | #include <linux/pid_namespace.h> |
1da177e4 LT |
50 | #include <linux/module.h> |
51 | #include <linux/namei.h> | |
1da177e4 LT |
52 | #include <linux/mount.h> |
53 | #include <linux/security.h> | |
54 | #include <linux/syscalls.h> | |
8f0ab514 | 55 | #include <linux/tsacct_kern.h> |
9f46080c | 56 | #include <linux/cn_proc.h> |
473ae30b | 57 | #include <linux/audit.h> |
6341c393 | 58 | #include <linux/tracehook.h> |
5f4123be | 59 | #include <linux/kmod.h> |
6110e3ab | 60 | #include <linux/fsnotify.h> |
5ad4e53b | 61 | #include <linux/fs_struct.h> |
3d5992d2 | 62 | #include <linux/oom.h> |
0e028465 | 63 | #include <linux/compat.h> |
b44a7dfc | 64 | #include <linux/vmalloc.h> |
1da177e4 | 65 | |
7c0f6ba6 | 66 | #include <linux/uaccess.h> |
1da177e4 | 67 | #include <asm/mmu_context.h> |
b6a2fea3 | 68 | #include <asm/tlb.h> |
43d2b113 KH |
69 | |
70 | #include <trace/events/task.h> | |
a6f76f23 | 71 | #include "internal.h" |
1da177e4 | 72 | |
4ff16c25 DS |
73 | #include <trace/events/sched.h> |
74 | ||
d6e71144 AC |
75 | int suid_dumpable = 0; |
76 | ||
e4dc1b14 | 77 | static LIST_HEAD(formats); |
1da177e4 LT |
78 | static DEFINE_RWLOCK(binfmt_lock); |
79 | ||
8fc3dc5a | 80 | void __register_binfmt(struct linux_binfmt * fmt, int insert) |
1da177e4 | 81 | { |
8fc3dc5a | 82 | BUG_ON(!fmt); |
92eaa565 ON |
83 | if (WARN_ON(!fmt->load_binary)) |
84 | return; | |
1da177e4 | 85 | write_lock(&binfmt_lock); |
74641f58 IK |
86 | insert ? list_add(&fmt->lh, &formats) : |
87 | list_add_tail(&fmt->lh, &formats); | |
1da177e4 | 88 | write_unlock(&binfmt_lock); |
1da177e4 LT |
89 | } |
90 | ||
74641f58 | 91 | EXPORT_SYMBOL(__register_binfmt); |
1da177e4 | 92 | |
f6b450d4 | 93 | void unregister_binfmt(struct linux_binfmt * fmt) |
1da177e4 | 94 | { |
1da177e4 | 95 | write_lock(&binfmt_lock); |
e4dc1b14 | 96 | list_del(&fmt->lh); |
1da177e4 | 97 | write_unlock(&binfmt_lock); |
1da177e4 LT |
98 | } |
99 | ||
100 | EXPORT_SYMBOL(unregister_binfmt); | |
101 | ||
102 | static inline void put_binfmt(struct linux_binfmt * fmt) | |
103 | { | |
104 | module_put(fmt->module); | |
105 | } | |
106 | ||
90f8572b EB |
107 | bool path_noexec(const struct path *path) |
108 | { | |
109 | return (path->mnt->mnt_flags & MNT_NOEXEC) || | |
110 | (path->mnt->mnt_sb->s_iflags & SB_I_NOEXEC); | |
111 | } | |
112 | ||
69369a70 | 113 | #ifdef CONFIG_USELIB |
1da177e4 LT |
114 | /* |
115 | * Note that a shared library must be both readable and executable due to | |
116 | * security reasons. | |
117 | * | |
118 | * Also note that we take the address to load from from the file itself. | |
119 | */ | |
1e7bfb21 | 120 | SYSCALL_DEFINE1(uselib, const char __user *, library) |
1da177e4 | 121 | { |
72c2d531 | 122 | struct linux_binfmt *fmt; |
964bd183 | 123 | struct file *file; |
91a27b2a | 124 | struct filename *tmp = getname(library); |
964bd183 | 125 | int error = PTR_ERR(tmp); |
47c805dc AV |
126 | static const struct open_flags uselib_flags = { |
127 | .open_flag = O_LARGEFILE | O_RDONLY | __FMODE_EXEC, | |
62fb4a15 | 128 | .acc_mode = MAY_READ | MAY_EXEC, |
f9652e10 AV |
129 | .intent = LOOKUP_OPEN, |
130 | .lookup_flags = LOOKUP_FOLLOW, | |
47c805dc | 131 | }; |
964bd183 | 132 | |
6e8341a1 AV |
133 | if (IS_ERR(tmp)) |
134 | goto out; | |
135 | ||
f9652e10 | 136 | file = do_filp_open(AT_FDCWD, tmp, &uselib_flags); |
6e8341a1 AV |
137 | putname(tmp); |
138 | error = PTR_ERR(file); | |
139 | if (IS_ERR(file)) | |
1da177e4 LT |
140 | goto out; |
141 | ||
142 | error = -EINVAL; | |
496ad9aa | 143 | if (!S_ISREG(file_inode(file)->i_mode)) |
1da177e4 LT |
144 | goto exit; |
145 | ||
30524472 | 146 | error = -EACCES; |
90f8572b | 147 | if (path_noexec(&file->f_path)) |
1da177e4 LT |
148 | goto exit; |
149 | ||
2a12a9d7 | 150 | fsnotify_open(file); |
6110e3ab | 151 | |
1da177e4 | 152 | error = -ENOEXEC; |
1da177e4 | 153 | |
72c2d531 AV |
154 | read_lock(&binfmt_lock); |
155 | list_for_each_entry(fmt, &formats, lh) { | |
156 | if (!fmt->load_shlib) | |
157 | continue; | |
158 | if (!try_module_get(fmt->module)) | |
159 | continue; | |
1da177e4 | 160 | read_unlock(&binfmt_lock); |
72c2d531 AV |
161 | error = fmt->load_shlib(file); |
162 | read_lock(&binfmt_lock); | |
163 | put_binfmt(fmt); | |
164 | if (error != -ENOEXEC) | |
165 | break; | |
1da177e4 | 166 | } |
72c2d531 | 167 | read_unlock(&binfmt_lock); |
6e8341a1 | 168 | exit: |
1da177e4 LT |
169 | fput(file); |
170 | out: | |
171 | return error; | |
1da177e4 | 172 | } |
69369a70 | 173 | #endif /* #ifdef CONFIG_USELIB */ |
1da177e4 | 174 | |
b6a2fea3 | 175 | #ifdef CONFIG_MMU |
ae6b585e ON |
176 | /* |
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). | |
181 | */ | |
0e028465 | 182 | static void acct_arg_size(struct linux_binprm *bprm, unsigned long pages) |
3c77f845 ON |
183 | { |
184 | struct mm_struct *mm = current->mm; | |
185 | long diff = (long)(pages - bprm->vma_pages); | |
186 | ||
187 | if (!mm || !diff) | |
188 | return; | |
189 | ||
190 | bprm->vma_pages = pages; | |
3c77f845 | 191 | add_mm_counter(mm, MM_ANONPAGES, diff); |
3c77f845 ON |
192 | } |
193 | ||
0e028465 | 194 | static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos, |
b6a2fea3 OW |
195 | int write) |
196 | { | |
197 | struct page *page; | |
198 | int ret; | |
9beae1ea | 199 | unsigned int gup_flags = FOLL_FORCE; |
b6a2fea3 OW |
200 | |
201 | #ifdef CONFIG_STACK_GROWSUP | |
202 | if (write) { | |
d05f3169 | 203 | ret = expand_downwards(bprm->vma, pos); |
b6a2fea3 OW |
204 | if (ret < 0) |
205 | return NULL; | |
206 | } | |
207 | #endif | |
9beae1ea LS |
208 | |
209 | if (write) | |
210 | gup_flags |= FOLL_WRITE; | |
211 | ||
1e987790 DH |
212 | /* |
213 | * We are doing an exec(). 'current' is the process | |
214 | * doing the exec and bprm->mm is the new process's mm. | |
215 | */ | |
9beae1ea | 216 | ret = get_user_pages_remote(current, bprm->mm, pos, 1, gup_flags, |
5b56d49f | 217 | &page, NULL, NULL); |
b6a2fea3 OW |
218 | if (ret <= 0) |
219 | return NULL; | |
220 | ||
655c16a8 ON |
221 | if (write) |
222 | acct_arg_size(bprm, vma_pages(bprm->vma)); | |
b6a2fea3 OW |
223 | |
224 | return page; | |
225 | } | |
226 | ||
227 | static void put_arg_page(struct page *page) | |
228 | { | |
229 | put_page(page); | |
230 | } | |
231 | ||
b6a2fea3 OW |
232 | static void free_arg_pages(struct linux_binprm *bprm) |
233 | { | |
234 | } | |
235 | ||
236 | static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos, | |
237 | struct page *page) | |
238 | { | |
239 | flush_cache_page(bprm->vma, pos, page_to_pfn(page)); | |
240 | } | |
241 | ||
242 | static int __bprm_mm_init(struct linux_binprm *bprm) | |
243 | { | |
eaccbfa5 | 244 | int err; |
b6a2fea3 OW |
245 | struct vm_area_struct *vma = NULL; |
246 | struct mm_struct *mm = bprm->mm; | |
247 | ||
490fc053 | 248 | bprm->vma = vma = vm_area_alloc(mm); |
b6a2fea3 | 249 | if (!vma) |
eaccbfa5 | 250 | return -ENOMEM; |
bfd40eaf | 251 | vma_set_anonymous(vma); |
b6a2fea3 | 252 | |
f268dfe9 MH |
253 | if (down_write_killable(&mm->mmap_sem)) { |
254 | err = -EINTR; | |
255 | goto err_free; | |
256 | } | |
b6a2fea3 OW |
257 | |
258 | /* | |
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 | |
262 | * configured yet. | |
263 | */ | |
aacb3d17 | 264 | BUILD_BUG_ON(VM_STACK_FLAGS & VM_STACK_INCOMPLETE_SETUP); |
b6a2fea3 OW |
265 | vma->vm_end = STACK_TOP_MAX; |
266 | vma->vm_start = vma->vm_end - PAGE_SIZE; | |
d9104d1c | 267 | vma->vm_flags = VM_SOFTDIRTY | VM_STACK_FLAGS | VM_STACK_INCOMPLETE_SETUP; |
3ed75eb8 | 268 | vma->vm_page_prot = vm_get_page_prot(vma->vm_flags); |
462e635e | 269 | |
b6a2fea3 | 270 | err = insert_vm_struct(mm, vma); |
eaccbfa5 | 271 | if (err) |
b6a2fea3 | 272 | goto err; |
b6a2fea3 OW |
273 | |
274 | mm->stack_vm = mm->total_vm = 1; | |
275 | up_write(&mm->mmap_sem); | |
b6a2fea3 | 276 | bprm->p = vma->vm_end - sizeof(void *); |
b6a2fea3 | 277 | return 0; |
b6a2fea3 | 278 | err: |
eaccbfa5 | 279 | up_write(&mm->mmap_sem); |
f268dfe9 | 280 | err_free: |
eaccbfa5 | 281 | bprm->vma = NULL; |
3928d4f5 | 282 | vm_area_free(vma); |
b6a2fea3 OW |
283 | return err; |
284 | } | |
285 | ||
286 | static bool valid_arg_len(struct linux_binprm *bprm, long len) | |
287 | { | |
288 | return len <= MAX_ARG_STRLEN; | |
289 | } | |
290 | ||
291 | #else | |
292 | ||
0e028465 | 293 | static inline void acct_arg_size(struct linux_binprm *bprm, unsigned long pages) |
3c77f845 ON |
294 | { |
295 | } | |
296 | ||
0e028465 | 297 | static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos, |
b6a2fea3 OW |
298 | int write) |
299 | { | |
300 | struct page *page; | |
301 | ||
302 | page = bprm->page[pos / PAGE_SIZE]; | |
303 | if (!page && write) { | |
304 | page = alloc_page(GFP_HIGHUSER|__GFP_ZERO); | |
305 | if (!page) | |
306 | return NULL; | |
307 | bprm->page[pos / PAGE_SIZE] = page; | |
308 | } | |
309 | ||
310 | return page; | |
311 | } | |
312 | ||
313 | static void put_arg_page(struct page *page) | |
314 | { | |
315 | } | |
316 | ||
317 | static void free_arg_page(struct linux_binprm *bprm, int i) | |
318 | { | |
319 | if (bprm->page[i]) { | |
320 | __free_page(bprm->page[i]); | |
321 | bprm->page[i] = NULL; | |
322 | } | |
323 | } | |
324 | ||
325 | static void free_arg_pages(struct linux_binprm *bprm) | |
326 | { | |
327 | int i; | |
328 | ||
329 | for (i = 0; i < MAX_ARG_PAGES; i++) | |
330 | free_arg_page(bprm, i); | |
331 | } | |
332 | ||
333 | static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos, | |
334 | struct page *page) | |
335 | { | |
336 | } | |
337 | ||
338 | static int __bprm_mm_init(struct linux_binprm *bprm) | |
339 | { | |
340 | bprm->p = PAGE_SIZE * MAX_ARG_PAGES - sizeof(void *); | |
341 | return 0; | |
342 | } | |
343 | ||
344 | static bool valid_arg_len(struct linux_binprm *bprm, long len) | |
345 | { | |
346 | return len <= bprm->p; | |
347 | } | |
348 | ||
349 | #endif /* CONFIG_MMU */ | |
350 | ||
351 | /* | |
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(). | |
356 | */ | |
9cc64cea | 357 | static int bprm_mm_init(struct linux_binprm *bprm) |
b6a2fea3 OW |
358 | { |
359 | int err; | |
360 | struct mm_struct *mm = NULL; | |
361 | ||
362 | bprm->mm = mm = mm_alloc(); | |
363 | err = -ENOMEM; | |
364 | if (!mm) | |
365 | goto err; | |
366 | ||
c31dbb14 KC |
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); | |
371 | ||
b6a2fea3 OW |
372 | err = __bprm_mm_init(bprm); |
373 | if (err) | |
374 | goto err; | |
375 | ||
376 | return 0; | |
377 | ||
378 | err: | |
379 | if (mm) { | |
380 | bprm->mm = NULL; | |
381 | mmdrop(mm); | |
382 | } | |
383 | ||
384 | return err; | |
385 | } | |
386 | ||
ba2d0162 | 387 | struct user_arg_ptr { |
0e028465 ON |
388 | #ifdef CONFIG_COMPAT |
389 | bool is_compat; | |
390 | #endif | |
391 | union { | |
392 | const char __user *const __user *native; | |
393 | #ifdef CONFIG_COMPAT | |
38b983b3 | 394 | const compat_uptr_t __user *compat; |
0e028465 ON |
395 | #endif |
396 | } ptr; | |
ba2d0162 ON |
397 | }; |
398 | ||
399 | static const char __user *get_user_arg_ptr(struct user_arg_ptr argv, int nr) | |
1d1dbf81 | 400 | { |
0e028465 ON |
401 | const char __user *native; |
402 | ||
403 | #ifdef CONFIG_COMPAT | |
404 | if (unlikely(argv.is_compat)) { | |
405 | compat_uptr_t compat; | |
406 | ||
407 | if (get_user(compat, argv.ptr.compat + nr)) | |
408 | return ERR_PTR(-EFAULT); | |
1d1dbf81 | 409 | |
0e028465 ON |
410 | return compat_ptr(compat); |
411 | } | |
412 | #endif | |
413 | ||
414 | if (get_user(native, argv.ptr.native + nr)) | |
1d1dbf81 ON |
415 | return ERR_PTR(-EFAULT); |
416 | ||
0e028465 | 417 | return native; |
1d1dbf81 ON |
418 | } |
419 | ||
1da177e4 LT |
420 | /* |
421 | * count() counts the number of strings in array ARGV. | |
422 | */ | |
ba2d0162 | 423 | static int count(struct user_arg_ptr argv, int max) |
1da177e4 LT |
424 | { |
425 | int i = 0; | |
426 | ||
0e028465 | 427 | if (argv.ptr.native != NULL) { |
1da177e4 | 428 | for (;;) { |
1d1dbf81 | 429 | const char __user *p = get_user_arg_ptr(argv, i); |
1da177e4 | 430 | |
1da177e4 LT |
431 | if (!p) |
432 | break; | |
1d1dbf81 ON |
433 | |
434 | if (IS_ERR(p)) | |
435 | return -EFAULT; | |
436 | ||
6d92d4f6 | 437 | if (i >= max) |
1da177e4 | 438 | return -E2BIG; |
6d92d4f6 | 439 | ++i; |
9aea5a65 RM |
440 | |
441 | if (fatal_signal_pending(current)) | |
442 | return -ERESTARTNOHAND; | |
1da177e4 LT |
443 | cond_resched(); |
444 | } | |
445 | } | |
446 | return i; | |
447 | } | |
448 | ||
655c16a8 ON |
449 | static int prepare_arg_pages(struct linux_binprm *bprm, |
450 | struct user_arg_ptr argv, struct user_arg_ptr envp) | |
451 | { | |
452 | unsigned long limit, ptr_size; | |
453 | ||
454 | bprm->argc = count(argv, MAX_ARG_STRINGS); | |
455 | if (bprm->argc < 0) | |
456 | return bprm->argc; | |
457 | ||
458 | bprm->envc = count(envp, MAX_ARG_STRINGS); | |
459 | if (bprm->envc < 0) | |
460 | return bprm->envc; | |
461 | ||
462 | /* | |
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. | |
465 | * This ensures that: | |
466 | * - the remaining binfmt code will not run out of stack space, | |
467 | * - the program will have a reasonable amount of stack left | |
468 | * to work from. | |
469 | */ | |
470 | limit = _STK_LIM / 4 * 3; | |
471 | limit = min(limit, bprm->rlim_stack.rlim_cur / 4); | |
472 | /* | |
473 | * We've historically supported up to 32 pages (ARG_MAX) | |
474 | * of argument strings even with small stacks | |
475 | */ | |
476 | limit = max_t(unsigned long, limit, ARG_MAX); | |
477 | /* | |
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. | |
483 | */ | |
484 | ptr_size = (bprm->argc + bprm->envc) * sizeof(void *); | |
485 | if (limit <= ptr_size) | |
486 | return -E2BIG; | |
487 | limit -= ptr_size; | |
488 | ||
489 | bprm->argmin = bprm->p - limit; | |
490 | return 0; | |
491 | } | |
492 | ||
1da177e4 | 493 | /* |
b6a2fea3 OW |
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. | |
1da177e4 | 497 | */ |
ba2d0162 | 498 | static int copy_strings(int argc, struct user_arg_ptr argv, |
75c96f85 | 499 | struct linux_binprm *bprm) |
1da177e4 LT |
500 | { |
501 | struct page *kmapped_page = NULL; | |
502 | char *kaddr = NULL; | |
b6a2fea3 | 503 | unsigned long kpos = 0; |
1da177e4 LT |
504 | int ret; |
505 | ||
506 | while (argc-- > 0) { | |
d7627467 | 507 | const char __user *str; |
1da177e4 LT |
508 | int len; |
509 | unsigned long pos; | |
510 | ||
1d1dbf81 ON |
511 | ret = -EFAULT; |
512 | str = get_user_arg_ptr(argv, argc); | |
513 | if (IS_ERR(str)) | |
1da177e4 | 514 | goto out; |
1da177e4 | 515 | |
1d1dbf81 ON |
516 | len = strnlen_user(str, MAX_ARG_STRLEN); |
517 | if (!len) | |
518 | goto out; | |
519 | ||
520 | ret = -E2BIG; | |
521 | if (!valid_arg_len(bprm, len)) | |
1da177e4 | 522 | goto out; |
1da177e4 | 523 | |
b6a2fea3 | 524 | /* We're going to work our way backwords. */ |
1da177e4 | 525 | pos = bprm->p; |
b6a2fea3 OW |
526 | str += len; |
527 | bprm->p -= len; | |
655c16a8 ON |
528 | #ifdef CONFIG_MMU |
529 | if (bprm->p < bprm->argmin) | |
530 | goto out; | |
531 | #endif | |
1da177e4 LT |
532 | |
533 | while (len > 0) { | |
1da177e4 | 534 | int offset, bytes_to_copy; |
1da177e4 | 535 | |
9aea5a65 RM |
536 | if (fatal_signal_pending(current)) { |
537 | ret = -ERESTARTNOHAND; | |
538 | goto out; | |
539 | } | |
7993bc1f RM |
540 | cond_resched(); |
541 | ||
1da177e4 | 542 | offset = pos % PAGE_SIZE; |
b6a2fea3 OW |
543 | if (offset == 0) |
544 | offset = PAGE_SIZE; | |
545 | ||
546 | bytes_to_copy = offset; | |
547 | if (bytes_to_copy > len) | |
548 | bytes_to_copy = len; | |
549 | ||
550 | offset -= bytes_to_copy; | |
551 | pos -= bytes_to_copy; | |
552 | str -= bytes_to_copy; | |
553 | len -= bytes_to_copy; | |
554 | ||
555 | if (!kmapped_page || kpos != (pos & PAGE_MASK)) { | |
556 | struct page *page; | |
557 | ||
558 | page = get_arg_page(bprm, pos, 1); | |
1da177e4 | 559 | if (!page) { |
b6a2fea3 | 560 | ret = -E2BIG; |
1da177e4 LT |
561 | goto out; |
562 | } | |
1da177e4 | 563 | |
b6a2fea3 OW |
564 | if (kmapped_page) { |
565 | flush_kernel_dcache_page(kmapped_page); | |
1da177e4 | 566 | kunmap(kmapped_page); |
b6a2fea3 OW |
567 | put_arg_page(kmapped_page); |
568 | } | |
1da177e4 LT |
569 | kmapped_page = page; |
570 | kaddr = kmap(kmapped_page); | |
b6a2fea3 OW |
571 | kpos = pos & PAGE_MASK; |
572 | flush_arg_page(bprm, kpos, kmapped_page); | |
1da177e4 | 573 | } |
b6a2fea3 | 574 | if (copy_from_user(kaddr+offset, str, bytes_to_copy)) { |
1da177e4 LT |
575 | ret = -EFAULT; |
576 | goto out; | |
577 | } | |
1da177e4 LT |
578 | } |
579 | } | |
580 | ret = 0; | |
581 | out: | |
b6a2fea3 OW |
582 | if (kmapped_page) { |
583 | flush_kernel_dcache_page(kmapped_page); | |
1da177e4 | 584 | kunmap(kmapped_page); |
b6a2fea3 OW |
585 | put_arg_page(kmapped_page); |
586 | } | |
1da177e4 LT |
587 | return ret; |
588 | } | |
589 | ||
590 | /* | |
986db2d1 | 591 | * Copy and argument/environment string from the kernel to the processes stack. |
1da177e4 | 592 | */ |
986db2d1 | 593 | int copy_string_kernel(const char *arg, struct linux_binprm *bprm) |
1da177e4 LT |
594 | { |
595 | int r; | |
596 | mm_segment_t oldfs = get_fs(); | |
ba2d0162 | 597 | struct user_arg_ptr argv = { |
986db2d1 | 598 | .ptr.native = (const char __user *const __user *)&arg, |
ba2d0162 ON |
599 | }; |
600 | ||
1da177e4 | 601 | set_fs(KERNEL_DS); |
986db2d1 | 602 | r = copy_strings(1, argv, bprm); |
1da177e4 | 603 | set_fs(oldfs); |
ba2d0162 | 604 | |
1da177e4 LT |
605 | return r; |
606 | } | |
986db2d1 | 607 | EXPORT_SYMBOL(copy_string_kernel); |
1da177e4 LT |
608 | |
609 | #ifdef CONFIG_MMU | |
b6a2fea3 | 610 | |
1da177e4 | 611 | /* |
b6a2fea3 OW |
612 | * During bprm_mm_init(), we create a temporary stack at STACK_TOP_MAX. Once |
613 | * the binfmt code determines where the new stack should reside, we shift it to | |
614 | * its final location. The process proceeds as follows: | |
1da177e4 | 615 | * |
b6a2fea3 OW |
616 | * 1) Use shift to calculate the new vma endpoints. |
617 | * 2) Extend vma to cover both the old and new ranges. This ensures the | |
618 | * arguments passed to subsequent functions are consistent. | |
619 | * 3) Move vma's page tables to the new range. | |
620 | * 4) Free up any cleared pgd range. | |
621 | * 5) Shrink the vma to cover only the new range. | |
1da177e4 | 622 | */ |
b6a2fea3 | 623 | static int shift_arg_pages(struct vm_area_struct *vma, unsigned long shift) |
1da177e4 LT |
624 | { |
625 | struct mm_struct *mm = vma->vm_mm; | |
b6a2fea3 OW |
626 | unsigned long old_start = vma->vm_start; |
627 | unsigned long old_end = vma->vm_end; | |
628 | unsigned long length = old_end - old_start; | |
629 | unsigned long new_start = old_start - shift; | |
630 | unsigned long new_end = old_end - shift; | |
d16dfc55 | 631 | struct mmu_gather tlb; |
1da177e4 | 632 | |
b6a2fea3 | 633 | BUG_ON(new_start > new_end); |
1da177e4 | 634 | |
b6a2fea3 OW |
635 | /* |
636 | * ensure there are no vmas between where we want to go | |
637 | * and where we are | |
638 | */ | |
639 | if (vma != find_vma(mm, new_start)) | |
640 | return -EFAULT; | |
641 | ||
642 | /* | |
643 | * cover the whole range: [new_start, old_end) | |
644 | */ | |
5beb4930 RR |
645 | if (vma_adjust(vma, new_start, old_end, vma->vm_pgoff, NULL)) |
646 | return -ENOMEM; | |
b6a2fea3 OW |
647 | |
648 | /* | |
649 | * move the page tables downwards, on failure we rely on | |
650 | * process cleanup to remove whatever mess we made. | |
651 | */ | |
652 | if (length != move_page_tables(vma, old_start, | |
38a76013 | 653 | vma, new_start, length, false)) |
b6a2fea3 OW |
654 | return -ENOMEM; |
655 | ||
656 | lru_add_drain(); | |
2b047252 | 657 | tlb_gather_mmu(&tlb, mm, old_start, old_end); |
b6a2fea3 OW |
658 | if (new_end > old_start) { |
659 | /* | |
660 | * when the old and new regions overlap clear from new_end. | |
661 | */ | |
d16dfc55 | 662 | free_pgd_range(&tlb, new_end, old_end, new_end, |
6ee8630e | 663 | vma->vm_next ? vma->vm_next->vm_start : USER_PGTABLES_CEILING); |
b6a2fea3 OW |
664 | } else { |
665 | /* | |
666 | * otherwise, clean from old_start; this is done to not touch | |
667 | * the address space in [new_end, old_start) some architectures | |
668 | * have constraints on va-space that make this illegal (IA64) - | |
669 | * for the others its just a little faster. | |
670 | */ | |
d16dfc55 | 671 | free_pgd_range(&tlb, old_start, old_end, new_end, |
6ee8630e | 672 | vma->vm_next ? vma->vm_next->vm_start : USER_PGTABLES_CEILING); |
1da177e4 | 673 | } |
2b047252 | 674 | tlb_finish_mmu(&tlb, old_start, old_end); |
b6a2fea3 OW |
675 | |
676 | /* | |
5beb4930 | 677 | * Shrink the vma to just the new range. Always succeeds. |
b6a2fea3 OW |
678 | */ |
679 | vma_adjust(vma, new_start, new_end, vma->vm_pgoff, NULL); | |
680 | ||
681 | return 0; | |
1da177e4 LT |
682 | } |
683 | ||
b6a2fea3 OW |
684 | /* |
685 | * Finalizes the stack vm_area_struct. The flags and permissions are updated, | |
686 | * the stack is optionally relocated, and some extra space is added. | |
687 | */ | |
1da177e4 LT |
688 | int setup_arg_pages(struct linux_binprm *bprm, |
689 | unsigned long stack_top, | |
690 | int executable_stack) | |
691 | { | |
b6a2fea3 OW |
692 | unsigned long ret; |
693 | unsigned long stack_shift; | |
1da177e4 | 694 | struct mm_struct *mm = current->mm; |
b6a2fea3 OW |
695 | struct vm_area_struct *vma = bprm->vma; |
696 | struct vm_area_struct *prev = NULL; | |
697 | unsigned long vm_flags; | |
698 | unsigned long stack_base; | |
803bf5ec MN |
699 | unsigned long stack_size; |
700 | unsigned long stack_expand; | |
701 | unsigned long rlim_stack; | |
1da177e4 LT |
702 | |
703 | #ifdef CONFIG_STACK_GROWSUP | |
d71f290b | 704 | /* Limit stack size */ |
c31dbb14 | 705 | stack_base = bprm->rlim_stack.rlim_max; |
d71f290b JH |
706 | if (stack_base > STACK_SIZE_MAX) |
707 | stack_base = STACK_SIZE_MAX; | |
1da177e4 | 708 | |
d045c77c HD |
709 | /* Add space for stack randomization. */ |
710 | stack_base += (STACK_RND_MASK << PAGE_SHIFT); | |
711 | ||
b6a2fea3 OW |
712 | /* Make sure we didn't let the argument array grow too large. */ |
713 | if (vma->vm_end - vma->vm_start > stack_base) | |
714 | return -ENOMEM; | |
1da177e4 | 715 | |
b6a2fea3 | 716 | stack_base = PAGE_ALIGN(stack_top - stack_base); |
1da177e4 | 717 | |
b6a2fea3 OW |
718 | stack_shift = vma->vm_start - stack_base; |
719 | mm->arg_start = bprm->p - stack_shift; | |
720 | bprm->p = vma->vm_end - stack_shift; | |
1da177e4 | 721 | #else |
b6a2fea3 OW |
722 | stack_top = arch_align_stack(stack_top); |
723 | stack_top = PAGE_ALIGN(stack_top); | |
1b528181 RM |
724 | |
725 | if (unlikely(stack_top < mmap_min_addr) || | |
726 | unlikely(vma->vm_end - vma->vm_start >= stack_top - mmap_min_addr)) | |
727 | return -ENOMEM; | |
728 | ||
b6a2fea3 OW |
729 | stack_shift = vma->vm_end - stack_top; |
730 | ||
731 | bprm->p -= stack_shift; | |
1da177e4 | 732 | mm->arg_start = bprm->p; |
1da177e4 LT |
733 | #endif |
734 | ||
1da177e4 | 735 | if (bprm->loader) |
b6a2fea3 OW |
736 | bprm->loader -= stack_shift; |
737 | bprm->exec -= stack_shift; | |
1da177e4 | 738 | |
f268dfe9 MH |
739 | if (down_write_killable(&mm->mmap_sem)) |
740 | return -EINTR; | |
741 | ||
96a8e13e | 742 | vm_flags = VM_STACK_FLAGS; |
b6a2fea3 OW |
743 | |
744 | /* | |
745 | * Adjust stack execute permissions; explicitly enable for | |
746 | * EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X and leave alone | |
747 | * (arch default) otherwise. | |
748 | */ | |
749 | if (unlikely(executable_stack == EXSTACK_ENABLE_X)) | |
750 | vm_flags |= VM_EXEC; | |
751 | else if (executable_stack == EXSTACK_DISABLE_X) | |
752 | vm_flags &= ~VM_EXEC; | |
753 | vm_flags |= mm->def_flags; | |
a8bef8ff | 754 | vm_flags |= VM_STACK_INCOMPLETE_SETUP; |
b6a2fea3 OW |
755 | |
756 | ret = mprotect_fixup(vma, &prev, vma->vm_start, vma->vm_end, | |
757 | vm_flags); | |
758 | if (ret) | |
759 | goto out_unlock; | |
760 | BUG_ON(prev != vma); | |
761 | ||
47a2ebb7 AD |
762 | if (unlikely(vm_flags & VM_EXEC)) { |
763 | pr_warn_once("process '%pD4' started with executable stack\n", | |
764 | bprm->file); | |
765 | } | |
766 | ||
b6a2fea3 OW |
767 | /* Move stack pages down in memory. */ |
768 | if (stack_shift) { | |
769 | ret = shift_arg_pages(vma, stack_shift); | |
fc63cf23 AB |
770 | if (ret) |
771 | goto out_unlock; | |
1da177e4 LT |
772 | } |
773 | ||
a8bef8ff MG |
774 | /* mprotect_fixup is overkill to remove the temporary stack flags */ |
775 | vma->vm_flags &= ~VM_STACK_INCOMPLETE_SETUP; | |
776 | ||
5ef097dd | 777 | stack_expand = 131072UL; /* randomly 32*4k (or 2*64k) pages */ |
803bf5ec MN |
778 | stack_size = vma->vm_end - vma->vm_start; |
779 | /* | |
780 | * Align this down to a page boundary as expand_stack | |
781 | * will align it up. | |
782 | */ | |
c31dbb14 | 783 | rlim_stack = bprm->rlim_stack.rlim_cur & PAGE_MASK; |
b6a2fea3 | 784 | #ifdef CONFIG_STACK_GROWSUP |
803bf5ec MN |
785 | if (stack_size + stack_expand > rlim_stack) |
786 | stack_base = vma->vm_start + rlim_stack; | |
787 | else | |
788 | stack_base = vma->vm_end + stack_expand; | |
b6a2fea3 | 789 | #else |
803bf5ec MN |
790 | if (stack_size + stack_expand > rlim_stack) |
791 | stack_base = vma->vm_end - rlim_stack; | |
792 | else | |
793 | stack_base = vma->vm_start - stack_expand; | |
b6a2fea3 | 794 | #endif |
3af9e859 | 795 | current->mm->start_stack = bprm->p; |
b6a2fea3 OW |
796 | ret = expand_stack(vma, stack_base); |
797 | if (ret) | |
798 | ret = -EFAULT; | |
799 | ||
800 | out_unlock: | |
1da177e4 | 801 | up_write(&mm->mmap_sem); |
fc63cf23 | 802 | return ret; |
1da177e4 | 803 | } |
1da177e4 LT |
804 | EXPORT_SYMBOL(setup_arg_pages); |
805 | ||
7e7ec6a9 NP |
806 | #else |
807 | ||
808 | /* | |
809 | * Transfer the program arguments and environment from the holding pages | |
810 | * onto the stack. The provided stack pointer is adjusted accordingly. | |
811 | */ | |
812 | int transfer_args_to_stack(struct linux_binprm *bprm, | |
813 | unsigned long *sp_location) | |
814 | { | |
815 | unsigned long index, stop, sp; | |
816 | int ret = 0; | |
817 | ||
818 | stop = bprm->p >> PAGE_SHIFT; | |
819 | sp = *sp_location; | |
820 | ||
821 | for (index = MAX_ARG_PAGES - 1; index >= stop; index--) { | |
822 | unsigned int offset = index == stop ? bprm->p & ~PAGE_MASK : 0; | |
823 | char *src = kmap(bprm->page[index]) + offset; | |
824 | sp -= PAGE_SIZE - offset; | |
825 | if (copy_to_user((void *) sp, src, PAGE_SIZE - offset) != 0) | |
826 | ret = -EFAULT; | |
827 | kunmap(bprm->page[index]); | |
828 | if (ret) | |
829 | goto out; | |
830 | } | |
831 | ||
832 | *sp_location = sp; | |
833 | ||
834 | out: | |
835 | return ret; | |
836 | } | |
837 | EXPORT_SYMBOL(transfer_args_to_stack); | |
838 | ||
1da177e4 LT |
839 | #endif /* CONFIG_MMU */ |
840 | ||
51f39a1f | 841 | static struct file *do_open_execat(int fd, struct filename *name, int flags) |
1da177e4 | 842 | { |
1da177e4 | 843 | struct file *file; |
e56b6a5d | 844 | int err; |
51f39a1f | 845 | struct open_flags open_exec_flags = { |
47c805dc | 846 | .open_flag = O_LARGEFILE | O_RDONLY | __FMODE_EXEC, |
62fb4a15 | 847 | .acc_mode = MAY_EXEC, |
f9652e10 AV |
848 | .intent = LOOKUP_OPEN, |
849 | .lookup_flags = LOOKUP_FOLLOW, | |
47c805dc | 850 | }; |
1da177e4 | 851 | |
51f39a1f DD |
852 | if ((flags & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0) |
853 | return ERR_PTR(-EINVAL); | |
854 | if (flags & AT_SYMLINK_NOFOLLOW) | |
855 | open_exec_flags.lookup_flags &= ~LOOKUP_FOLLOW; | |
856 | if (flags & AT_EMPTY_PATH) | |
857 | open_exec_flags.lookup_flags |= LOOKUP_EMPTY; | |
858 | ||
859 | file = do_filp_open(fd, name, &open_exec_flags); | |
6e8341a1 | 860 | if (IS_ERR(file)) |
e56b6a5d CH |
861 | goto out; |
862 | ||
863 | err = -EACCES; | |
496ad9aa | 864 | if (!S_ISREG(file_inode(file)->i_mode)) |
6e8341a1 | 865 | goto exit; |
e56b6a5d | 866 | |
90f8572b | 867 | if (path_noexec(&file->f_path)) |
6e8341a1 | 868 | goto exit; |
e56b6a5d CH |
869 | |
870 | err = deny_write_access(file); | |
6e8341a1 AV |
871 | if (err) |
872 | goto exit; | |
1da177e4 | 873 | |
51f39a1f DD |
874 | if (name->name[0] != '\0') |
875 | fsnotify_open(file); | |
876 | ||
6e8341a1 | 877 | out: |
e56b6a5d CH |
878 | return file; |
879 | ||
6e8341a1 AV |
880 | exit: |
881 | fput(file); | |
e56b6a5d CH |
882 | return ERR_PTR(err); |
883 | } | |
c4ad8f98 LT |
884 | |
885 | struct file *open_exec(const char *name) | |
886 | { | |
51689104 PM |
887 | struct filename *filename = getname_kernel(name); |
888 | struct file *f = ERR_CAST(filename); | |
889 | ||
890 | if (!IS_ERR(filename)) { | |
891 | f = do_open_execat(AT_FDCWD, filename, 0); | |
892 | putname(filename); | |
893 | } | |
894 | return f; | |
c4ad8f98 | 895 | } |
1da177e4 LT |
896 | EXPORT_SYMBOL(open_exec); |
897 | ||
b44a7dfc | 898 | int kernel_read_file(struct file *file, void **buf, loff_t *size, |
bc8ca5b9 | 899 | loff_t max_size, enum kernel_read_file_id id) |
b44a7dfc MZ |
900 | { |
901 | loff_t i_size, pos; | |
902 | ssize_t bytes = 0; | |
903 | int ret; | |
904 | ||
905 | if (!S_ISREG(file_inode(file)->i_mode) || max_size < 0) | |
906 | return -EINVAL; | |
907 | ||
7bd698b3 | 908 | ret = deny_write_access(file); |
39eeb4fb MZ |
909 | if (ret) |
910 | return ret; | |
911 | ||
7bd698b3 | 912 | ret = security_kernel_read_file(file, id); |
39d637af | 913 | if (ret) |
7bd698b3 | 914 | goto out; |
39d637af | 915 | |
b44a7dfc | 916 | i_size = i_size_read(file_inode(file)); |
39d637af DK |
917 | if (i_size <= 0) { |
918 | ret = -EINVAL; | |
919 | goto out; | |
920 | } | |
691115c3 EB |
921 | if (i_size > SIZE_MAX || (max_size > 0 && i_size > max_size)) { |
922 | ret = -EFBIG; | |
923 | goto out; | |
924 | } | |
b44a7dfc | 925 | |
a098ecd2 SB |
926 | if (id != READING_FIRMWARE_PREALLOC_BUFFER) |
927 | *buf = vmalloc(i_size); | |
39d637af DK |
928 | if (!*buf) { |
929 | ret = -ENOMEM; | |
930 | goto out; | |
931 | } | |
b44a7dfc MZ |
932 | |
933 | pos = 0; | |
934 | while (pos < i_size) { | |
bdd1d2d3 | 935 | bytes = kernel_read(file, *buf + pos, i_size - pos, &pos); |
b44a7dfc MZ |
936 | if (bytes < 0) { |
937 | ret = bytes; | |
f612acfa | 938 | goto out_free; |
b44a7dfc MZ |
939 | } |
940 | ||
941 | if (bytes == 0) | |
942 | break; | |
b44a7dfc MZ |
943 | } |
944 | ||
945 | if (pos != i_size) { | |
946 | ret = -EIO; | |
39d637af | 947 | goto out_free; |
b44a7dfc MZ |
948 | } |
949 | ||
bc8ca5b9 | 950 | ret = security_kernel_post_read_file(file, *buf, i_size, id); |
b44a7dfc MZ |
951 | if (!ret) |
952 | *size = pos; | |
953 | ||
39d637af | 954 | out_free: |
b44a7dfc | 955 | if (ret < 0) { |
a098ecd2 SB |
956 | if (id != READING_FIRMWARE_PREALLOC_BUFFER) { |
957 | vfree(*buf); | |
958 | *buf = NULL; | |
959 | } | |
b44a7dfc | 960 | } |
39d637af DK |
961 | |
962 | out: | |
963 | allow_write_access(file); | |
b44a7dfc MZ |
964 | return ret; |
965 | } | |
966 | EXPORT_SYMBOL_GPL(kernel_read_file); | |
967 | ||
711aab1d | 968 | int kernel_read_file_from_path(const char *path, void **buf, loff_t *size, |
09596b94 MZ |
969 | loff_t max_size, enum kernel_read_file_id id) |
970 | { | |
971 | struct file *file; | |
972 | int ret; | |
973 | ||
974 | if (!path || !*path) | |
975 | return -EINVAL; | |
976 | ||
977 | file = filp_open(path, O_RDONLY, 0); | |
978 | if (IS_ERR(file)) | |
979 | return PTR_ERR(file); | |
980 | ||
981 | ret = kernel_read_file(file, buf, size, max_size, id); | |
982 | fput(file); | |
983 | return ret; | |
984 | } | |
985 | EXPORT_SYMBOL_GPL(kernel_read_file_from_path); | |
986 | ||
901cff7c TM |
987 | int kernel_read_file_from_path_initns(const char *path, void **buf, |
988 | loff_t *size, loff_t max_size, | |
989 | enum kernel_read_file_id id) | |
990 | { | |
991 | struct file *file; | |
992 | struct path root; | |
993 | int ret; | |
994 | ||
995 | if (!path || !*path) | |
996 | return -EINVAL; | |
997 | ||
998 | task_lock(&init_task); | |
999 | get_fs_root(init_task.fs, &root); | |
1000 | task_unlock(&init_task); | |
1001 | ||
1002 | file = file_open_root(root.dentry, root.mnt, path, O_RDONLY, 0); | |
1003 | path_put(&root); | |
1004 | if (IS_ERR(file)) | |
1005 | return PTR_ERR(file); | |
1006 | ||
1007 | ret = kernel_read_file(file, buf, size, max_size, id); | |
1008 | fput(file); | |
1009 | return ret; | |
1010 | } | |
1011 | EXPORT_SYMBOL_GPL(kernel_read_file_from_path_initns); | |
1012 | ||
b844f0ec MZ |
1013 | int kernel_read_file_from_fd(int fd, void **buf, loff_t *size, loff_t max_size, |
1014 | enum kernel_read_file_id id) | |
1015 | { | |
1016 | struct fd f = fdget(fd); | |
1017 | int ret = -EBADF; | |
1018 | ||
1019 | if (!f.file) | |
1020 | goto out; | |
1021 | ||
1022 | ret = kernel_read_file(f.file, buf, size, max_size, id); | |
1023 | out: | |
1024 | fdput(f); | |
1025 | return ret; | |
1026 | } | |
1027 | EXPORT_SYMBOL_GPL(kernel_read_file_from_fd); | |
1028 | ||
3dc20cb2 AV |
1029 | ssize_t read_code(struct file *file, unsigned long addr, loff_t pos, size_t len) |
1030 | { | |
ec695579 | 1031 | ssize_t res = vfs_read(file, (void __user *)addr, len, &pos); |
3dc20cb2 AV |
1032 | if (res > 0) |
1033 | flush_icache_range(addr, addr + len); | |
1034 | return res; | |
1035 | } | |
1036 | EXPORT_SYMBOL(read_code); | |
1037 | ||
eea96732 EB |
1038 | /* |
1039 | * Maps the mm_struct mm into the current task struct. | |
1040 | * On success, this function returns with the mutex | |
1041 | * exec_update_mutex locked. | |
1042 | */ | |
1da177e4 LT |
1043 | static int exec_mmap(struct mm_struct *mm) |
1044 | { | |
1045 | struct task_struct *tsk; | |
615d6e87 | 1046 | struct mm_struct *old_mm, *active_mm; |
eea96732 | 1047 | int ret; |
1da177e4 LT |
1048 | |
1049 | /* Notify parent that we're no longer interested in the old VM */ | |
1050 | tsk = current; | |
1051 | old_mm = current->mm; | |
4610ba7a | 1052 | exec_mm_release(tsk, old_mm); |
1da177e4 | 1053 | |
eea96732 EB |
1054 | ret = mutex_lock_killable(&tsk->signal->exec_update_mutex); |
1055 | if (ret) | |
1056 | return ret; | |
1057 | ||
1da177e4 | 1058 | if (old_mm) { |
4fe7efdb | 1059 | sync_mm_rss(old_mm); |
1da177e4 LT |
1060 | /* |
1061 | * Make sure that if there is a core dump in progress | |
1062 | * for the old mm, we get out and die instead of going | |
1063 | * through with the exec. We must hold mmap_sem around | |
999d9fc1 | 1064 | * checking core_state and changing tsk->mm. |
1da177e4 LT |
1065 | */ |
1066 | down_read(&old_mm->mmap_sem); | |
999d9fc1 | 1067 | if (unlikely(old_mm->core_state)) { |
1da177e4 | 1068 | up_read(&old_mm->mmap_sem); |
eea96732 | 1069 | mutex_unlock(&tsk->signal->exec_update_mutex); |
1da177e4 LT |
1070 | return -EINTR; |
1071 | } | |
1072 | } | |
eea96732 | 1073 | |
1da177e4 LT |
1074 | task_lock(tsk); |
1075 | active_mm = tsk->active_mm; | |
227a4aad | 1076 | membarrier_exec_mmap(mm); |
1da177e4 LT |
1077 | tsk->mm = mm; |
1078 | tsk->active_mm = mm; | |
1079 | activate_mm(active_mm, mm); | |
615d6e87 DB |
1080 | tsk->mm->vmacache_seqnum = 0; |
1081 | vmacache_flush(tsk); | |
1da177e4 | 1082 | task_unlock(tsk); |
1da177e4 LT |
1083 | if (old_mm) { |
1084 | up_read(&old_mm->mmap_sem); | |
7dddb12c | 1085 | BUG_ON(active_mm != old_mm); |
701085b2 | 1086 | setmax_mm_hiwater_rss(&tsk->signal->maxrss, old_mm); |
31a78f23 | 1087 | mm_update_next_owner(old_mm); |
1da177e4 LT |
1088 | mmput(old_mm); |
1089 | return 0; | |
1090 | } | |
1091 | mmdrop(active_mm); | |
1092 | return 0; | |
1093 | } | |
1094 | ||
1095 | /* | |
1096 | * This function makes sure the current process has its own signal table, | |
1097 | * so that flush_signal_handlers can later reset the handlers without | |
1098 | * disturbing other processes. (Other processes might share the signal | |
1099 | * table via the CLONE_SIGHAND option to clone().) | |
1100 | */ | |
858119e1 | 1101 | static int de_thread(struct task_struct *tsk) |
1da177e4 LT |
1102 | { |
1103 | struct signal_struct *sig = tsk->signal; | |
b2c903b8 | 1104 | struct sighand_struct *oldsighand = tsk->sighand; |
1da177e4 | 1105 | spinlock_t *lock = &oldsighand->siglock; |
1da177e4 | 1106 | |
aafe6c2a | 1107 | if (thread_group_empty(tsk)) |
1da177e4 LT |
1108 | goto no_thread_group; |
1109 | ||
1110 | /* | |
1111 | * Kill all other threads in the thread group. | |
1da177e4 | 1112 | */ |
1da177e4 | 1113 | spin_lock_irq(lock); |
ed5d2cac | 1114 | if (signal_group_exit(sig)) { |
1da177e4 LT |
1115 | /* |
1116 | * Another group action in progress, just | |
1117 | * return so that the signal is processed. | |
1118 | */ | |
1119 | spin_unlock_irq(lock); | |
1da177e4 LT |
1120 | return -EAGAIN; |
1121 | } | |
d344193a | 1122 | |
ed5d2cac | 1123 | sig->group_exit_task = tsk; |
d344193a ON |
1124 | sig->notify_count = zap_other_threads(tsk); |
1125 | if (!thread_group_leader(tsk)) | |
1126 | sig->notify_count--; | |
1da177e4 | 1127 | |
d344193a | 1128 | while (sig->notify_count) { |
d5bbd43d | 1129 | __set_current_state(TASK_KILLABLE); |
1da177e4 | 1130 | spin_unlock_irq(lock); |
a72173ec | 1131 | schedule(); |
08d405c8 | 1132 | if (__fatal_signal_pending(tsk)) |
d5bbd43d | 1133 | goto killed; |
1da177e4 LT |
1134 | spin_lock_irq(lock); |
1135 | } | |
1da177e4 LT |
1136 | spin_unlock_irq(lock); |
1137 | ||
1138 | /* | |
1139 | * At this point all other threads have exited, all we have to | |
1140 | * do is to wait for the thread group leader to become inactive, | |
1141 | * and to assume its PID: | |
1142 | */ | |
aafe6c2a | 1143 | if (!thread_group_leader(tsk)) { |
8187926b | 1144 | struct task_struct *leader = tsk->group_leader; |
6db840fa | 1145 | |
6db840fa | 1146 | for (;;) { |
780de9dd | 1147 | cgroup_threadgroup_change_begin(tsk); |
6db840fa | 1148 | write_lock_irq(&tasklist_lock); |
dfcce791 KT |
1149 | /* |
1150 | * Do this under tasklist_lock to ensure that | |
1151 | * exit_notify() can't miss ->group_exit_task | |
1152 | */ | |
1153 | sig->notify_count = -1; | |
6db840fa ON |
1154 | if (likely(leader->exit_state)) |
1155 | break; | |
d5bbd43d | 1156 | __set_current_state(TASK_KILLABLE); |
6db840fa | 1157 | write_unlock_irq(&tasklist_lock); |
780de9dd | 1158 | cgroup_threadgroup_change_end(tsk); |
a72173ec | 1159 | schedule(); |
08d405c8 | 1160 | if (__fatal_signal_pending(tsk)) |
d5bbd43d | 1161 | goto killed; |
6db840fa | 1162 | } |
1da177e4 | 1163 | |
f5e90281 RM |
1164 | /* |
1165 | * The only record we have of the real-time age of a | |
1166 | * process, regardless of execs it's done, is start_time. | |
1167 | * All the past CPU time is accumulated in signal_struct | |
1168 | * from sister threads now dead. But in this non-leader | |
1169 | * exec, nothing survives from the original leader thread, | |
1170 | * whose birth marks the true age of this process now. | |
1171 | * When we take on its identity by switching to its PID, we | |
1172 | * also take its birthdate (always earlier than our own). | |
1173 | */ | |
aafe6c2a | 1174 | tsk->start_time = leader->start_time; |
cf25e24d | 1175 | tsk->start_boottime = leader->start_boottime; |
f5e90281 | 1176 | |
bac0abd6 PE |
1177 | BUG_ON(!same_thread_group(leader, tsk)); |
1178 | BUG_ON(has_group_leader_pid(tsk)); | |
1da177e4 LT |
1179 | /* |
1180 | * An exec() starts a new thread group with the | |
1181 | * TGID of the previous thread group. Rehash the | |
1182 | * two threads with a switched PID, and release | |
1183 | * the former thread group leader: | |
1184 | */ | |
d73d6529 EB |
1185 | |
1186 | /* Become a process group leader with the old leader's pid. | |
c18258c6 EB |
1187 | * The old leader becomes a thread of the this thread group. |
1188 | * Note: The old leader also uses this pid until release_task | |
d73d6529 EB |
1189 | * is called. Odd but simple and correct. |
1190 | */ | |
aafe6c2a | 1191 | tsk->pid = leader->pid; |
3f418548 | 1192 | change_pid(tsk, PIDTYPE_PID, task_pid(leader)); |
6883f81a | 1193 | transfer_pid(leader, tsk, PIDTYPE_TGID); |
aafe6c2a EB |
1194 | transfer_pid(leader, tsk, PIDTYPE_PGID); |
1195 | transfer_pid(leader, tsk, PIDTYPE_SID); | |
9cd80bbb | 1196 | |
aafe6c2a | 1197 | list_replace_rcu(&leader->tasks, &tsk->tasks); |
9cd80bbb | 1198 | list_replace_init(&leader->sibling, &tsk->sibling); |
1da177e4 | 1199 | |
aafe6c2a EB |
1200 | tsk->group_leader = tsk; |
1201 | leader->group_leader = tsk; | |
de12a787 | 1202 | |
aafe6c2a | 1203 | tsk->exit_signal = SIGCHLD; |
087806b1 | 1204 | leader->exit_signal = -1; |
962b564c ON |
1205 | |
1206 | BUG_ON(leader->exit_state != EXIT_ZOMBIE); | |
1207 | leader->exit_state = EXIT_DEAD; | |
eac1b5e5 ON |
1208 | |
1209 | /* | |
1210 | * We are going to release_task()->ptrace_unlink() silently, | |
1211 | * the tracer can sleep in do_wait(). EXIT_DEAD guarantees | |
1212 | * the tracer wont't block again waiting for this thread. | |
1213 | */ | |
1214 | if (unlikely(leader->ptrace)) | |
1215 | __wake_up_parent(leader, leader->parent); | |
1da177e4 | 1216 | write_unlock_irq(&tasklist_lock); |
780de9dd | 1217 | cgroup_threadgroup_change_end(tsk); |
8187926b ON |
1218 | |
1219 | release_task(leader); | |
ed5d2cac | 1220 | } |
1da177e4 | 1221 | |
6db840fa ON |
1222 | sig->group_exit_task = NULL; |
1223 | sig->notify_count = 0; | |
1da177e4 LT |
1224 | |
1225 | no_thread_group: | |
e6368253 ON |
1226 | /* we have changed execution domain */ |
1227 | tsk->exit_signal = SIGCHLD; | |
1228 | ||
02169155 EB |
1229 | BUG_ON(!thread_group_leader(tsk)); |
1230 | return 0; | |
1231 | ||
1232 | killed: | |
1233 | /* protects against exit_notify() and __exit_signal() */ | |
1234 | read_lock(&tasklist_lock); | |
1235 | sig->group_exit_task = NULL; | |
1236 | sig->notify_count = 0; | |
1237 | read_unlock(&tasklist_lock); | |
1238 | return -EAGAIN; | |
1239 | } | |
1240 | ||
1241 | ||
1242 | static int unshare_sighand(struct task_struct *me) | |
1243 | { | |
1244 | struct sighand_struct *oldsighand = me->sighand; | |
329f7dba | 1245 | |
d036bda7 | 1246 | if (refcount_read(&oldsighand->count) != 1) { |
b2c903b8 | 1247 | struct sighand_struct *newsighand; |
1da177e4 | 1248 | /* |
b2c903b8 ON |
1249 | * This ->sighand is shared with the CLONE_SIGHAND |
1250 | * but not CLONE_THREAD task, switch to the new one. | |
1da177e4 | 1251 | */ |
b2c903b8 ON |
1252 | newsighand = kmem_cache_alloc(sighand_cachep, GFP_KERNEL); |
1253 | if (!newsighand) | |
1254 | return -ENOMEM; | |
1255 | ||
d036bda7 | 1256 | refcount_set(&newsighand->count, 1); |
1da177e4 LT |
1257 | memcpy(newsighand->action, oldsighand->action, |
1258 | sizeof(newsighand->action)); | |
1259 | ||
1260 | write_lock_irq(&tasklist_lock); | |
1261 | spin_lock(&oldsighand->siglock); | |
02169155 | 1262 | rcu_assign_pointer(me->sighand, newsighand); |
1da177e4 LT |
1263 | spin_unlock(&oldsighand->siglock); |
1264 | write_unlock_irq(&tasklist_lock); | |
1265 | ||
fba2afaa | 1266 | __cleanup_sighand(oldsighand); |
1da177e4 | 1267 | } |
1da177e4 LT |
1268 | return 0; |
1269 | } | |
0840a90d | 1270 | |
3756f640 | 1271 | char *__get_task_comm(char *buf, size_t buf_size, struct task_struct *tsk) |
1da177e4 | 1272 | { |
1da177e4 | 1273 | task_lock(tsk); |
3756f640 | 1274 | strncpy(buf, tsk->comm, buf_size); |
1da177e4 | 1275 | task_unlock(tsk); |
59714d65 | 1276 | return buf; |
1da177e4 | 1277 | } |
3756f640 | 1278 | EXPORT_SYMBOL_GPL(__get_task_comm); |
1da177e4 | 1279 | |
6a6d27de AV |
1280 | /* |
1281 | * These functions flushes out all traces of the currently running executable | |
1282 | * so that a new one can be started | |
1283 | */ | |
1284 | ||
82b89778 | 1285 | void __set_task_comm(struct task_struct *tsk, const char *buf, bool exec) |
1da177e4 LT |
1286 | { |
1287 | task_lock(tsk); | |
43d2b113 | 1288 | trace_task_rename(tsk, buf); |
1da177e4 LT |
1289 | strlcpy(tsk->comm, buf, sizeof(tsk->comm)); |
1290 | task_unlock(tsk); | |
82b89778 | 1291 | perf_event_comm(tsk, exec); |
1da177e4 LT |
1292 | } |
1293 | ||
a9208e42 KC |
1294 | /* |
1295 | * Calling this is the point of no return. None of the failures will be | |
1296 | * seen by userspace since either the process is already taking a fatal | |
1297 | * signal (via de_thread() or coredump), or will have SEGV raised | |
1298 | * (after exec_mmap()) by search_binary_handlers (see below). | |
1299 | */ | |
1da177e4 LT |
1300 | int flush_old_exec(struct linux_binprm * bprm) |
1301 | { | |
2ca7be7d | 1302 | struct task_struct *me = current; |
221af7f8 | 1303 | int retval; |
1da177e4 LT |
1304 | |
1305 | /* | |
02169155 | 1306 | * Make this the only thread in the thread group. |
1da177e4 | 1307 | */ |
2ca7be7d | 1308 | retval = de_thread(me); |
1da177e4 LT |
1309 | if (retval) |
1310 | goto out; | |
1311 | ||
6e399cd1 DB |
1312 | /* |
1313 | * Must be called _before_ exec_mmap() as bprm->mm is | |
1314 | * not visibile until then. This also enables the update | |
1315 | * to be lockless. | |
1316 | */ | |
925d1c40 | 1317 | set_mm_exe_file(bprm->mm, bprm->file); |
6e399cd1 | 1318 | |
f87d1c95 EB |
1319 | would_dump(bprm, bprm->file); |
1320 | ||
1da177e4 LT |
1321 | /* |
1322 | * Release all of the old mmap stuff | |
1323 | */ | |
3c77f845 | 1324 | acct_arg_size(bprm, 0); |
1da177e4 LT |
1325 | retval = exec_mmap(bprm->mm); |
1326 | if (retval) | |
fd8328be | 1327 | goto out; |
1da177e4 | 1328 | |
a9208e42 | 1329 | /* |
eea96732 EB |
1330 | * After setting bprm->called_exec_mmap (to mark that current is |
1331 | * using the prepared mm now), we have nothing left of the original | |
a9208e42 KC |
1332 | * process. If anything from here on returns an error, the check |
1333 | * in search_binary_handler() will SEGV current. | |
1334 | */ | |
eea96732 | 1335 | bprm->called_exec_mmap = 1; |
a9208e42 | 1336 | bprm->mm = NULL; |
7ab02af4 | 1337 | |
ccf0fa6b EB |
1338 | #ifdef CONFIG_POSIX_TIMERS |
1339 | exit_itimers(me->signal); | |
1340 | flush_itimer_signals(); | |
1341 | #endif | |
1342 | ||
1343 | /* | |
1344 | * Make the signal table private. | |
1345 | */ | |
1346 | retval = unshare_sighand(me); | |
1347 | if (retval) | |
1348 | goto out; | |
1349 | ||
dac853ae | 1350 | set_fs(USER_DS); |
2ca7be7d | 1351 | me->flags &= ~(PF_RANDOMIZE | PF_FORKNOEXEC | PF_KTHREAD | |
b88fae64 | 1352 | PF_NOFREEZE | PF_NO_SETAFFINITY); |
7ab02af4 | 1353 | flush_thread(); |
2ca7be7d | 1354 | me->personality &= ~bprm->per_clear; |
7ab02af4 | 1355 | |
613cc2b6 AS |
1356 | /* |
1357 | * We have to apply CLOEXEC before we change whether the process is | |
1358 | * dumpable (in setup_new_exec) to avoid a race with a process in userspace | |
1359 | * trying to access the should-be-closed file descriptors of a process | |
1360 | * undergoing exec(2). | |
1361 | */ | |
2ca7be7d | 1362 | do_close_on_exec(me->files); |
221af7f8 LT |
1363 | return 0; |
1364 | ||
1365 | out: | |
1366 | return retval; | |
1367 | } | |
1368 | EXPORT_SYMBOL(flush_old_exec); | |
1369 | ||
1b5d783c AV |
1370 | void would_dump(struct linux_binprm *bprm, struct file *file) |
1371 | { | |
f84df2a6 EB |
1372 | struct inode *inode = file_inode(file); |
1373 | if (inode_permission(inode, MAY_READ) < 0) { | |
1374 | struct user_namespace *old, *user_ns; | |
1b5d783c | 1375 | bprm->interp_flags |= BINPRM_FLAGS_ENFORCE_NONDUMP; |
f84df2a6 EB |
1376 | |
1377 | /* Ensure mm->user_ns contains the executable */ | |
1378 | user_ns = old = bprm->mm->user_ns; | |
1379 | while ((user_ns != &init_user_ns) && | |
1380 | !privileged_wrt_inode_uidgid(user_ns, inode)) | |
1381 | user_ns = user_ns->parent; | |
1382 | ||
1383 | if (old != user_ns) { | |
1384 | bprm->mm->user_ns = get_user_ns(user_ns); | |
1385 | put_user_ns(old); | |
1386 | } | |
1387 | } | |
1b5d783c AV |
1388 | } |
1389 | EXPORT_SYMBOL(would_dump); | |
1390 | ||
221af7f8 LT |
1391 | void setup_new_exec(struct linux_binprm * bprm) |
1392 | { | |
46d98eb4 KC |
1393 | /* |
1394 | * Once here, prepare_binrpm() will not be called any more, so | |
1395 | * the final state of setuid/setgid/fscaps can be merged into the | |
1396 | * secureexec flag. | |
1397 | */ | |
1398 | bprm->secureexec |= bprm->cap_elevated; | |
1399 | ||
64701dee | 1400 | if (bprm->secureexec) { |
fe8993b3 KC |
1401 | /* Make sure parent cannot signal privileged process. */ |
1402 | current->pdeath_signal = 0; | |
1403 | ||
64701dee KC |
1404 | /* |
1405 | * For secureexec, reset the stack limit to sane default to | |
1406 | * avoid bad behavior from the prior rlimits. This has to | |
1407 | * happen before arch_pick_mmap_layout(), which examines | |
1408 | * RLIMIT_STACK, but after the point of no return to avoid | |
779f4e1c | 1409 | * needing to clean up the change on failure. |
64701dee | 1410 | */ |
c31dbb14 KC |
1411 | if (bprm->rlim_stack.rlim_cur > _STK_LIM) |
1412 | bprm->rlim_stack.rlim_cur = _STK_LIM; | |
64701dee KC |
1413 | } |
1414 | ||
c31dbb14 | 1415 | arch_pick_mmap_layout(current->mm, &bprm->rlim_stack); |
1da177e4 | 1416 | |
1da177e4 LT |
1417 | current->sas_ss_sp = current->sas_ss_size = 0; |
1418 | ||
e816c201 KC |
1419 | /* |
1420 | * Figure out dumpability. Note that this checking only of current | |
1421 | * is wrong, but userspace depends on it. This should be testing | |
1422 | * bprm->secureexec instead. | |
1423 | */ | |
473d8963 | 1424 | if (bprm->interp_flags & BINPRM_FLAGS_ENFORCE_NONDUMP || |
e816c201 KC |
1425 | !(uid_eq(current_euid(), current_uid()) && |
1426 | gid_eq(current_egid(), current_gid()))) | |
6c5d5238 | 1427 | set_dumpable(current->mm, suid_dumpable); |
473d8963 KC |
1428 | else |
1429 | set_dumpable(current->mm, SUID_DUMP_USER); | |
d6e71144 | 1430 | |
e9ea1e7f | 1431 | arch_setup_new_exec(); |
e041e328 | 1432 | perf_event_exec(); |
82b89778 | 1433 | __set_task_comm(current, kbasename(bprm->filename), true); |
1da177e4 | 1434 | |
0551fbd2 BH |
1435 | /* Set the new mm task size. We have to do that late because it may |
1436 | * depend on TIF_32BIT which is only updated in flush_thread() on | |
1437 | * some architectures like powerpc | |
1438 | */ | |
1439 | current->mm->task_size = TASK_SIZE; | |
1440 | ||
1da177e4 LT |
1441 | /* An exec changes our domain. We are no longer part of the thread |
1442 | group */ | |
d1e7fd64 | 1443 | WRITE_ONCE(current->self_exec_id, current->self_exec_id + 1); |
1da177e4 | 1444 | flush_signal_handlers(current, 0); |
1da177e4 | 1445 | } |
221af7f8 | 1446 | EXPORT_SYMBOL(setup_new_exec); |
1da177e4 | 1447 | |
b8383831 KC |
1448 | /* Runs immediately before start_thread() takes over. */ |
1449 | void finalize_exec(struct linux_binprm *bprm) | |
1450 | { | |
c31dbb14 KC |
1451 | /* Store any stack rlimit changes before starting thread. */ |
1452 | task_lock(current->group_leader); | |
1453 | current->signal->rlim[RLIMIT_STACK] = bprm->rlim_stack; | |
1454 | task_unlock(current->group_leader); | |
b8383831 KC |
1455 | } |
1456 | EXPORT_SYMBOL(finalize_exec); | |
1457 | ||
a2a8474c ON |
1458 | /* |
1459 | * Prepare credentials and lock ->cred_guard_mutex. | |
1460 | * install_exec_creds() commits the new creds and drops the lock. | |
1461 | * Or, if exec fails before, free_bprm() should release ->cred and | |
1462 | * and unlock. | |
1463 | */ | |
4addd264 | 1464 | static int prepare_bprm_creds(struct linux_binprm *bprm) |
a2a8474c | 1465 | { |
9b1bf12d | 1466 | if (mutex_lock_interruptible(¤t->signal->cred_guard_mutex)) |
a2a8474c ON |
1467 | return -ERESTARTNOINTR; |
1468 | ||
1469 | bprm->cred = prepare_exec_creds(); | |
1470 | if (likely(bprm->cred)) | |
1471 | return 0; | |
1472 | ||
9b1bf12d | 1473 | mutex_unlock(¤t->signal->cred_guard_mutex); |
a2a8474c ON |
1474 | return -ENOMEM; |
1475 | } | |
1476 | ||
c4ad8f98 | 1477 | static void free_bprm(struct linux_binprm *bprm) |
a2a8474c ON |
1478 | { |
1479 | free_arg_pages(bprm); | |
1480 | if (bprm->cred) { | |
eea96732 EB |
1481 | if (bprm->called_exec_mmap) |
1482 | mutex_unlock(¤t->signal->exec_update_mutex); | |
9b1bf12d | 1483 | mutex_unlock(¤t->signal->cred_guard_mutex); |
a2a8474c ON |
1484 | abort_creds(bprm->cred); |
1485 | } | |
63e46b95 ON |
1486 | if (bprm->file) { |
1487 | allow_write_access(bprm->file); | |
1488 | fput(bprm->file); | |
1489 | } | |
b66c5984 KC |
1490 | /* If a binfmt changed the interp, free it. */ |
1491 | if (bprm->interp != bprm->filename) | |
1492 | kfree(bprm->interp); | |
a2a8474c ON |
1493 | kfree(bprm); |
1494 | } | |
1495 | ||
c2315c18 | 1496 | int bprm_change_interp(const char *interp, struct linux_binprm *bprm) |
b66c5984 KC |
1497 | { |
1498 | /* If a binfmt changed the interp, free it first. */ | |
1499 | if (bprm->interp != bprm->filename) | |
1500 | kfree(bprm->interp); | |
1501 | bprm->interp = kstrdup(interp, GFP_KERNEL); | |
1502 | if (!bprm->interp) | |
1503 | return -ENOMEM; | |
1504 | return 0; | |
1505 | } | |
1506 | EXPORT_SYMBOL(bprm_change_interp); | |
1507 | ||
a6f76f23 DH |
1508 | /* |
1509 | * install the new credentials for this executable | |
1510 | */ | |
1511 | void install_exec_creds(struct linux_binprm *bprm) | |
1512 | { | |
1513 | security_bprm_committing_creds(bprm); | |
1514 | ||
1515 | commit_creds(bprm->cred); | |
1516 | bprm->cred = NULL; | |
2976b10f SE |
1517 | |
1518 | /* | |
1519 | * Disable monitoring for regular users | |
1520 | * when executing setuid binaries. Must | |
1521 | * wait until new credentials are committed | |
1522 | * by commit_creds() above | |
1523 | */ | |
1524 | if (get_dumpable(current->mm) != SUID_DUMP_USER) | |
1525 | perf_event_exit_task(current); | |
a2a8474c ON |
1526 | /* |
1527 | * cred_guard_mutex must be held at least to this point to prevent | |
a6f76f23 | 1528 | * ptrace_attach() from altering our determination of the task's |
a2a8474c ON |
1529 | * credentials; any time after this it may be unlocked. |
1530 | */ | |
a6f76f23 | 1531 | security_bprm_committed_creds(bprm); |
eea96732 | 1532 | mutex_unlock(¤t->signal->exec_update_mutex); |
9b1bf12d | 1533 | mutex_unlock(¤t->signal->cred_guard_mutex); |
a6f76f23 DH |
1534 | } |
1535 | EXPORT_SYMBOL(install_exec_creds); | |
1536 | ||
1537 | /* | |
1538 | * determine how safe it is to execute the proposed program | |
9b1bf12d | 1539 | * - the caller must hold ->cred_guard_mutex to protect against |
c2e1f2e3 | 1540 | * PTRACE_ATTACH or seccomp thread-sync |
a6f76f23 | 1541 | */ |
9e00cdb0 | 1542 | static void check_unsafe_exec(struct linux_binprm *bprm) |
a6f76f23 | 1543 | { |
0bf2f3ae | 1544 | struct task_struct *p = current, *t; |
f1191b50 | 1545 | unsigned n_fs; |
a6f76f23 | 1546 | |
9227dd2a EB |
1547 | if (p->ptrace) |
1548 | bprm->unsafe |= LSM_UNSAFE_PTRACE; | |
a6f76f23 | 1549 | |
259e5e6c AL |
1550 | /* |
1551 | * This isn't strictly necessary, but it makes it harder for LSMs to | |
1552 | * mess up. | |
1553 | */ | |
1d4457f9 | 1554 | if (task_no_new_privs(current)) |
259e5e6c AL |
1555 | bprm->unsafe |= LSM_UNSAFE_NO_NEW_PRIVS; |
1556 | ||
83f62a2e | 1557 | t = p; |
0bf2f3ae | 1558 | n_fs = 1; |
2a4419b5 | 1559 | spin_lock(&p->fs->lock); |
437f7fdb | 1560 | rcu_read_lock(); |
83f62a2e | 1561 | while_each_thread(p, t) { |
0bf2f3ae DH |
1562 | if (t->fs == p->fs) |
1563 | n_fs++; | |
0bf2f3ae | 1564 | } |
437f7fdb | 1565 | rcu_read_unlock(); |
0bf2f3ae | 1566 | |
9e00cdb0 | 1567 | if (p->fs->users > n_fs) |
a6f76f23 | 1568 | bprm->unsafe |= LSM_UNSAFE_SHARE; |
9e00cdb0 ON |
1569 | else |
1570 | p->fs->in_exec = 1; | |
2a4419b5 | 1571 | spin_unlock(&p->fs->lock); |
a6f76f23 DH |
1572 | } |
1573 | ||
8b01fc86 JH |
1574 | static void bprm_fill_uid(struct linux_binprm *bprm) |
1575 | { | |
1576 | struct inode *inode; | |
1577 | unsigned int mode; | |
1578 | kuid_t uid; | |
1579 | kgid_t gid; | |
1580 | ||
cb6fd68f KC |
1581 | /* |
1582 | * Since this can be called multiple times (via prepare_binprm), | |
1583 | * we must clear any previous work done when setting set[ug]id | |
1584 | * bits from any earlier bprm->file uses (for example when run | |
1585 | * first for a setuid script then again for its interpreter). | |
1586 | */ | |
8b01fc86 JH |
1587 | bprm->cred->euid = current_euid(); |
1588 | bprm->cred->egid = current_egid(); | |
1589 | ||
380cf5ba | 1590 | if (!mnt_may_suid(bprm->file->f_path.mnt)) |
8b01fc86 JH |
1591 | return; |
1592 | ||
1593 | if (task_no_new_privs(current)) | |
1594 | return; | |
1595 | ||
fea6d2a6 | 1596 | inode = bprm->file->f_path.dentry->d_inode; |
8b01fc86 JH |
1597 | mode = READ_ONCE(inode->i_mode); |
1598 | if (!(mode & (S_ISUID|S_ISGID))) | |
1599 | return; | |
1600 | ||
1601 | /* Be careful if suid/sgid is set */ | |
5955102c | 1602 | inode_lock(inode); |
8b01fc86 JH |
1603 | |
1604 | /* reload atomically mode/uid/gid now that lock held */ | |
1605 | mode = inode->i_mode; | |
1606 | uid = inode->i_uid; | |
1607 | gid = inode->i_gid; | |
5955102c | 1608 | inode_unlock(inode); |
8b01fc86 JH |
1609 | |
1610 | /* We ignore suid/sgid if there are no mappings for them in the ns */ | |
1611 | if (!kuid_has_mapping(bprm->cred->user_ns, uid) || | |
1612 | !kgid_has_mapping(bprm->cred->user_ns, gid)) | |
1613 | return; | |
1614 | ||
1615 | if (mode & S_ISUID) { | |
1616 | bprm->per_clear |= PER_CLEAR_ON_SETID; | |
1617 | bprm->cred->euid = uid; | |
1618 | } | |
1619 | ||
1620 | if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) { | |
1621 | bprm->per_clear |= PER_CLEAR_ON_SETID; | |
1622 | bprm->cred->egid = gid; | |
1623 | } | |
1624 | } | |
1625 | ||
9e00cdb0 ON |
1626 | /* |
1627 | * Fill the binprm structure from the inode. | |
6eb3c3d0 | 1628 | * Check permissions, then read the first BINPRM_BUF_SIZE bytes |
a6f76f23 DH |
1629 | * |
1630 | * This may be called multiple times for binary chains (scripts for example). | |
1da177e4 LT |
1631 | */ |
1632 | int prepare_binprm(struct linux_binprm *bprm) | |
1633 | { | |
1da177e4 | 1634 | int retval; |
bdd1d2d3 | 1635 | loff_t pos = 0; |
1da177e4 | 1636 | |
8b01fc86 | 1637 | bprm_fill_uid(bprm); |
1da177e4 LT |
1638 | |
1639 | /* fill in binprm security blob */ | |
a6f76f23 | 1640 | retval = security_bprm_set_creds(bprm); |
1da177e4 LT |
1641 | if (retval) |
1642 | return retval; | |
ddb4a144 | 1643 | bprm->called_set_creds = 1; |
1da177e4 | 1644 | |
a6f76f23 | 1645 | memset(bprm->buf, 0, BINPRM_BUF_SIZE); |
bdd1d2d3 | 1646 | return kernel_read(bprm->file, bprm->buf, BINPRM_BUF_SIZE, &pos); |
1da177e4 LT |
1647 | } |
1648 | ||
1649 | EXPORT_SYMBOL(prepare_binprm); | |
1650 | ||
4fc75ff4 NP |
1651 | /* |
1652 | * Arguments are '\0' separated strings found at the location bprm->p | |
1653 | * points to; chop off the first by relocating brpm->p to right after | |
1654 | * the first '\0' encountered. | |
1655 | */ | |
b6a2fea3 | 1656 | int remove_arg_zero(struct linux_binprm *bprm) |
1da177e4 | 1657 | { |
b6a2fea3 OW |
1658 | int ret = 0; |
1659 | unsigned long offset; | |
1660 | char *kaddr; | |
1661 | struct page *page; | |
4fc75ff4 | 1662 | |
b6a2fea3 OW |
1663 | if (!bprm->argc) |
1664 | return 0; | |
1da177e4 | 1665 | |
b6a2fea3 OW |
1666 | do { |
1667 | offset = bprm->p & ~PAGE_MASK; | |
1668 | page = get_arg_page(bprm, bprm->p, 0); | |
1669 | if (!page) { | |
1670 | ret = -EFAULT; | |
1671 | goto out; | |
1672 | } | |
e8e3c3d6 | 1673 | kaddr = kmap_atomic(page); |
4fc75ff4 | 1674 | |
b6a2fea3 OW |
1675 | for (; offset < PAGE_SIZE && kaddr[offset]; |
1676 | offset++, bprm->p++) | |
1677 | ; | |
4fc75ff4 | 1678 | |
e8e3c3d6 | 1679 | kunmap_atomic(kaddr); |
b6a2fea3 | 1680 | put_arg_page(page); |
b6a2fea3 | 1681 | } while (offset == PAGE_SIZE); |
4fc75ff4 | 1682 | |
b6a2fea3 OW |
1683 | bprm->p++; |
1684 | bprm->argc--; | |
1685 | ret = 0; | |
4fc75ff4 | 1686 | |
b6a2fea3 OW |
1687 | out: |
1688 | return ret; | |
1da177e4 | 1689 | } |
1da177e4 LT |
1690 | EXPORT_SYMBOL(remove_arg_zero); |
1691 | ||
cb7b6b1c | 1692 | #define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e)) |
1da177e4 LT |
1693 | /* |
1694 | * cycle the list of binary formats handler, until one recognizes the image | |
1695 | */ | |
3c456bfc | 1696 | int search_binary_handler(struct linux_binprm *bprm) |
1da177e4 | 1697 | { |
cb7b6b1c | 1698 | bool need_retry = IS_ENABLED(CONFIG_MODULES); |
1da177e4 | 1699 | struct linux_binfmt *fmt; |
cb7b6b1c | 1700 | int retval; |
1da177e4 | 1701 | |
d7402698 | 1702 | /* This allows 4 levels of binfmt rewrites before failing hard. */ |
131b2f9f | 1703 | if (bprm->recursion_depth > 5) |
d7402698 KC |
1704 | return -ELOOP; |
1705 | ||
1da177e4 LT |
1706 | retval = security_bprm_check(bprm); |
1707 | if (retval) | |
1708 | return retval; | |
1709 | ||
1da177e4 | 1710 | retval = -ENOENT; |
cb7b6b1c ON |
1711 | retry: |
1712 | read_lock(&binfmt_lock); | |
1713 | list_for_each_entry(fmt, &formats, lh) { | |
1714 | if (!try_module_get(fmt->module)) | |
1715 | continue; | |
1716 | read_unlock(&binfmt_lock); | |
d53ddd01 | 1717 | |
cb7b6b1c ON |
1718 | bprm->recursion_depth++; |
1719 | retval = fmt->load_binary(bprm); | |
d53ddd01 AD |
1720 | bprm->recursion_depth--; |
1721 | ||
19d860a1 AV |
1722 | read_lock(&binfmt_lock); |
1723 | put_binfmt(fmt); | |
eea96732 | 1724 | if (retval < 0 && bprm->called_exec_mmap) { |
19d860a1 AV |
1725 | /* we got to flush_old_exec() and failed after it */ |
1726 | read_unlock(&binfmt_lock); | |
cb44c9a0 | 1727 | force_sigsegv(SIGSEGV); |
19d860a1 AV |
1728 | return retval; |
1729 | } | |
1730 | if (retval != -ENOEXEC || !bprm->file) { | |
1731 | read_unlock(&binfmt_lock); | |
cb7b6b1c | 1732 | return retval; |
1da177e4 | 1733 | } |
1da177e4 | 1734 | } |
cb7b6b1c ON |
1735 | read_unlock(&binfmt_lock); |
1736 | ||
19d860a1 | 1737 | if (need_retry) { |
cb7b6b1c ON |
1738 | if (printable(bprm->buf[0]) && printable(bprm->buf[1]) && |
1739 | printable(bprm->buf[2]) && printable(bprm->buf[3])) | |
1740 | return retval; | |
4e0621a0 ON |
1741 | if (request_module("binfmt-%04x", *(ushort *)(bprm->buf + 2)) < 0) |
1742 | return retval; | |
cb7b6b1c ON |
1743 | need_retry = false; |
1744 | goto retry; | |
1745 | } | |
1746 | ||
1da177e4 LT |
1747 | return retval; |
1748 | } | |
1da177e4 LT |
1749 | EXPORT_SYMBOL(search_binary_handler); |
1750 | ||
5d1baf3b ON |
1751 | static int exec_binprm(struct linux_binprm *bprm) |
1752 | { | |
1753 | pid_t old_pid, old_vpid; | |
1754 | int ret; | |
1755 | ||
1756 | /* Need to fetch pid before load_binary changes it */ | |
1757 | old_pid = current->pid; | |
1758 | rcu_read_lock(); | |
1759 | old_vpid = task_pid_nr_ns(current, task_active_pid_ns(current->parent)); | |
1760 | rcu_read_unlock(); | |
1761 | ||
1762 | ret = search_binary_handler(bprm); | |
1763 | if (ret >= 0) { | |
3eaded86 | 1764 | audit_bprm(bprm); |
5d1baf3b ON |
1765 | trace_sched_process_exec(current, old_pid, bprm); |
1766 | ptrace_event(PTRACE_EVENT_EXEC, old_vpid); | |
9beb266f | 1767 | proc_exec_connector(current); |
5d1baf3b ON |
1768 | } |
1769 | ||
1770 | return ret; | |
1771 | } | |
1772 | ||
1da177e4 LT |
1773 | /* |
1774 | * sys_execve() executes a new program. | |
1775 | */ | |
449325b5 AS |
1776 | static int __do_execve_file(int fd, struct filename *filename, |
1777 | struct user_arg_ptr argv, | |
1778 | struct user_arg_ptr envp, | |
1779 | int flags, struct file *file) | |
1da177e4 | 1780 | { |
51f39a1f | 1781 | char *pathbuf = NULL; |
1da177e4 | 1782 | struct linux_binprm *bprm; |
3b125388 | 1783 | struct files_struct *displaced; |
1da177e4 | 1784 | int retval; |
72fa5997 | 1785 | |
c4ad8f98 LT |
1786 | if (IS_ERR(filename)) |
1787 | return PTR_ERR(filename); | |
1788 | ||
72fa5997 VK |
1789 | /* |
1790 | * We move the actual failure in case of RLIMIT_NPROC excess from | |
1791 | * set*uid() to execve() because too many poorly written programs | |
1792 | * don't check setuid() return code. Here we additionally recheck | |
1793 | * whether NPROC limit is still exceeded. | |
1794 | */ | |
1795 | if ((current->flags & PF_NPROC_EXCEEDED) && | |
bd9d43f4 | 1796 | atomic_read(¤t_user()->processes) > rlimit(RLIMIT_NPROC)) { |
72fa5997 VK |
1797 | retval = -EAGAIN; |
1798 | goto out_ret; | |
1799 | } | |
1800 | ||
1801 | /* We're below the limit (still or again), so we don't want to make | |
1802 | * further execve() calls fail. */ | |
1803 | current->flags &= ~PF_NPROC_EXCEEDED; | |
1da177e4 | 1804 | |
3b125388 | 1805 | retval = unshare_files(&displaced); |
fd8328be AV |
1806 | if (retval) |
1807 | goto out_ret; | |
1808 | ||
1da177e4 | 1809 | retval = -ENOMEM; |
11b0b5ab | 1810 | bprm = kzalloc(sizeof(*bprm), GFP_KERNEL); |
1da177e4 | 1811 | if (!bprm) |
fd8328be | 1812 | goto out_files; |
1da177e4 | 1813 | |
a2a8474c ON |
1814 | retval = prepare_bprm_creds(bprm); |
1815 | if (retval) | |
a6f76f23 | 1816 | goto out_free; |
498052bb | 1817 | |
9e00cdb0 | 1818 | check_unsafe_exec(bprm); |
a2a8474c | 1819 | current->in_execve = 1; |
a6f76f23 | 1820 | |
449325b5 AS |
1821 | if (!file) |
1822 | file = do_open_execat(fd, filename, flags); | |
1da177e4 LT |
1823 | retval = PTR_ERR(file); |
1824 | if (IS_ERR(file)) | |
498052bb | 1825 | goto out_unmark; |
1da177e4 LT |
1826 | |
1827 | sched_exec(); | |
1828 | ||
1da177e4 | 1829 | bprm->file = file; |
449325b5 AS |
1830 | if (!filename) { |
1831 | bprm->filename = "none"; | |
1832 | } else if (fd == AT_FDCWD || filename->name[0] == '/') { | |
51f39a1f DD |
1833 | bprm->filename = filename->name; |
1834 | } else { | |
1835 | if (filename->name[0] == '\0') | |
0ee931c4 | 1836 | pathbuf = kasprintf(GFP_KERNEL, "/dev/fd/%d", fd); |
51f39a1f | 1837 | else |
0ee931c4 | 1838 | pathbuf = kasprintf(GFP_KERNEL, "/dev/fd/%d/%s", |
51f39a1f DD |
1839 | fd, filename->name); |
1840 | if (!pathbuf) { | |
1841 | retval = -ENOMEM; | |
1842 | goto out_unmark; | |
1843 | } | |
1844 | /* | |
1845 | * Record that a name derived from an O_CLOEXEC fd will be | |
1846 | * inaccessible after exec. Relies on having exclusive access to | |
1847 | * current->files (due to unshare_files above). | |
1848 | */ | |
1849 | if (close_on_exec(fd, rcu_dereference_raw(current->files->fdt))) | |
1850 | bprm->interp_flags |= BINPRM_FLAGS_PATH_INACCESSIBLE; | |
1851 | bprm->filename = pathbuf; | |
1852 | } | |
1853 | bprm->interp = bprm->filename; | |
1da177e4 | 1854 | |
b6a2fea3 OW |
1855 | retval = bprm_mm_init(bprm); |
1856 | if (retval) | |
63e46b95 | 1857 | goto out_unmark; |
1da177e4 | 1858 | |
655c16a8 ON |
1859 | retval = prepare_arg_pages(bprm, argv, envp); |
1860 | if (retval < 0) | |
1da177e4 LT |
1861 | goto out; |
1862 | ||
1863 | retval = prepare_binprm(bprm); | |
1864 | if (retval < 0) | |
1865 | goto out; | |
1866 | ||
986db2d1 | 1867 | retval = copy_string_kernel(bprm->filename, bprm); |
1da177e4 LT |
1868 | if (retval < 0) |
1869 | goto out; | |
1870 | ||
1871 | bprm->exec = bprm->p; | |
1872 | retval = copy_strings(bprm->envc, envp, bprm); | |
1873 | if (retval < 0) | |
1874 | goto out; | |
1875 | ||
1876 | retval = copy_strings(bprm->argc, argv, bprm); | |
1877 | if (retval < 0) | |
1878 | goto out; | |
1879 | ||
5d1baf3b | 1880 | retval = exec_binprm(bprm); |
a6f76f23 DH |
1881 | if (retval < 0) |
1882 | goto out; | |
1da177e4 | 1883 | |
a6f76f23 | 1884 | /* execve succeeded */ |
498052bb | 1885 | current->fs->in_exec = 0; |
f9ce1f1c | 1886 | current->in_execve = 0; |
d7822b1e | 1887 | rseq_execve(current); |
a6f76f23 | 1888 | acct_update_integrals(current); |
16d51a59 | 1889 | task_numa_free(current, false); |
a6f76f23 | 1890 | free_bprm(bprm); |
51f39a1f | 1891 | kfree(pathbuf); |
449325b5 AS |
1892 | if (filename) |
1893 | putname(filename); | |
a6f76f23 DH |
1894 | if (displaced) |
1895 | put_files_struct(displaced); | |
1896 | return retval; | |
1da177e4 | 1897 | |
a6f76f23 | 1898 | out: |
3c77f845 ON |
1899 | if (bprm->mm) { |
1900 | acct_arg_size(bprm, 0); | |
1901 | mmput(bprm->mm); | |
1902 | } | |
1da177e4 | 1903 | |
498052bb | 1904 | out_unmark: |
9e00cdb0 | 1905 | current->fs->in_exec = 0; |
f9ce1f1c | 1906 | current->in_execve = 0; |
a6f76f23 DH |
1907 | |
1908 | out_free: | |
08a6fac1 | 1909 | free_bprm(bprm); |
51f39a1f | 1910 | kfree(pathbuf); |
1da177e4 | 1911 | |
fd8328be | 1912 | out_files: |
3b125388 AV |
1913 | if (displaced) |
1914 | reset_files_struct(displaced); | |
1da177e4 | 1915 | out_ret: |
449325b5 AS |
1916 | if (filename) |
1917 | putname(filename); | |
1da177e4 LT |
1918 | return retval; |
1919 | } | |
1920 | ||
449325b5 AS |
1921 | static int do_execveat_common(int fd, struct filename *filename, |
1922 | struct user_arg_ptr argv, | |
1923 | struct user_arg_ptr envp, | |
1924 | int flags) | |
1925 | { | |
1926 | return __do_execve_file(fd, filename, argv, envp, flags, NULL); | |
1927 | } | |
1928 | ||
1929 | int do_execve_file(struct file *file, void *__argv, void *__envp) | |
1930 | { | |
1931 | struct user_arg_ptr argv = { .ptr.native = __argv }; | |
1932 | struct user_arg_ptr envp = { .ptr.native = __envp }; | |
1933 | ||
1934 | return __do_execve_file(AT_FDCWD, NULL, argv, envp, 0, file); | |
1935 | } | |
1936 | ||
c4ad8f98 | 1937 | int do_execve(struct filename *filename, |
ba2d0162 | 1938 | const char __user *const __user *__argv, |
da3d4c5f | 1939 | const char __user *const __user *__envp) |
ba2d0162 | 1940 | { |
0e028465 ON |
1941 | struct user_arg_ptr argv = { .ptr.native = __argv }; |
1942 | struct user_arg_ptr envp = { .ptr.native = __envp }; | |
51f39a1f DD |
1943 | return do_execveat_common(AT_FDCWD, filename, argv, envp, 0); |
1944 | } | |
1945 | ||
1946 | int do_execveat(int fd, struct filename *filename, | |
1947 | const char __user *const __user *__argv, | |
1948 | const char __user *const __user *__envp, | |
1949 | int flags) | |
1950 | { | |
1951 | struct user_arg_ptr argv = { .ptr.native = __argv }; | |
1952 | struct user_arg_ptr envp = { .ptr.native = __envp }; | |
1953 | ||
1954 | return do_execveat_common(fd, filename, argv, envp, flags); | |
0e028465 ON |
1955 | } |
1956 | ||
1957 | #ifdef CONFIG_COMPAT | |
c4ad8f98 | 1958 | static int compat_do_execve(struct filename *filename, |
38b983b3 | 1959 | const compat_uptr_t __user *__argv, |
d03d26e5 | 1960 | const compat_uptr_t __user *__envp) |
0e028465 ON |
1961 | { |
1962 | struct user_arg_ptr argv = { | |
1963 | .is_compat = true, | |
1964 | .ptr.compat = __argv, | |
1965 | }; | |
1966 | struct user_arg_ptr envp = { | |
1967 | .is_compat = true, | |
1968 | .ptr.compat = __envp, | |
1969 | }; | |
51f39a1f DD |
1970 | return do_execveat_common(AT_FDCWD, filename, argv, envp, 0); |
1971 | } | |
1972 | ||
1973 | static int compat_do_execveat(int fd, struct filename *filename, | |
1974 | const compat_uptr_t __user *__argv, | |
1975 | const compat_uptr_t __user *__envp, | |
1976 | int flags) | |
1977 | { | |
1978 | struct user_arg_ptr argv = { | |
1979 | .is_compat = true, | |
1980 | .ptr.compat = __argv, | |
1981 | }; | |
1982 | struct user_arg_ptr envp = { | |
1983 | .is_compat = true, | |
1984 | .ptr.compat = __envp, | |
1985 | }; | |
1986 | return do_execveat_common(fd, filename, argv, envp, flags); | |
ba2d0162 | 1987 | } |
0e028465 | 1988 | #endif |
ba2d0162 | 1989 | |
964ee7df | 1990 | void set_binfmt(struct linux_binfmt *new) |
1da177e4 | 1991 | { |
801460d0 HS |
1992 | struct mm_struct *mm = current->mm; |
1993 | ||
1994 | if (mm->binfmt) | |
1995 | module_put(mm->binfmt->module); | |
1da177e4 | 1996 | |
801460d0 | 1997 | mm->binfmt = new; |
964ee7df ON |
1998 | if (new) |
1999 | __module_get(new->module); | |
1da177e4 | 2000 | } |
1da177e4 LT |
2001 | EXPORT_SYMBOL(set_binfmt); |
2002 | ||
6c5d5238 | 2003 | /* |
7288e118 | 2004 | * set_dumpable stores three-value SUID_DUMP_* into mm->flags. |
6c5d5238 KH |
2005 | */ |
2006 | void set_dumpable(struct mm_struct *mm, int value) | |
2007 | { | |
7288e118 ON |
2008 | if (WARN_ON((unsigned)value > SUID_DUMP_ROOT)) |
2009 | return; | |
2010 | ||
26e15225 | 2011 | set_mask_bits(&mm->flags, MMF_DUMPABLE_MASK, value); |
6c5d5238 | 2012 | } |
6c5d5238 | 2013 | |
38b983b3 AV |
2014 | SYSCALL_DEFINE3(execve, |
2015 | const char __user *, filename, | |
2016 | const char __user *const __user *, argv, | |
2017 | const char __user *const __user *, envp) | |
2018 | { | |
c4ad8f98 | 2019 | return do_execve(getname(filename), argv, envp); |
38b983b3 | 2020 | } |
51f39a1f DD |
2021 | |
2022 | SYSCALL_DEFINE5(execveat, | |
2023 | int, fd, const char __user *, filename, | |
2024 | const char __user *const __user *, argv, | |
2025 | const char __user *const __user *, envp, | |
2026 | int, flags) | |
2027 | { | |
2028 | int lookup_flags = (flags & AT_EMPTY_PATH) ? LOOKUP_EMPTY : 0; | |
2029 | ||
2030 | return do_execveat(fd, | |
2031 | getname_flags(filename, lookup_flags, NULL), | |
2032 | argv, envp, flags); | |
2033 | } | |
2034 | ||
38b983b3 | 2035 | #ifdef CONFIG_COMPAT |
625b1d7e HC |
2036 | COMPAT_SYSCALL_DEFINE3(execve, const char __user *, filename, |
2037 | const compat_uptr_t __user *, argv, | |
2038 | const compat_uptr_t __user *, envp) | |
38b983b3 | 2039 | { |
c4ad8f98 | 2040 | return compat_do_execve(getname(filename), argv, envp); |
38b983b3 | 2041 | } |
51f39a1f DD |
2042 | |
2043 | COMPAT_SYSCALL_DEFINE5(execveat, int, fd, | |
2044 | const char __user *, filename, | |
2045 | const compat_uptr_t __user *, argv, | |
2046 | const compat_uptr_t __user *, envp, | |
2047 | int, flags) | |
2048 | { | |
2049 | int lookup_flags = (flags & AT_EMPTY_PATH) ? LOOKUP_EMPTY : 0; | |
2050 | ||
2051 | return compat_do_execveat(fd, | |
2052 | getname_flags(filename, lookup_flags, NULL), | |
2053 | argv, envp, flags); | |
2054 | } | |
38b983b3 | 2055 | #endif |