]>
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 | ||
56305aa9 EB |
75 | static int bprm_creds_from_file(struct linux_binprm *bprm); |
76 | ||
d6e71144 AC |
77 | int suid_dumpable = 0; |
78 | ||
e4dc1b14 | 79 | static LIST_HEAD(formats); |
1da177e4 LT |
80 | static DEFINE_RWLOCK(binfmt_lock); |
81 | ||
8fc3dc5a | 82 | void __register_binfmt(struct linux_binfmt * fmt, int insert) |
1da177e4 | 83 | { |
8fc3dc5a | 84 | BUG_ON(!fmt); |
92eaa565 ON |
85 | if (WARN_ON(!fmt->load_binary)) |
86 | return; | |
1da177e4 | 87 | write_lock(&binfmt_lock); |
74641f58 IK |
88 | insert ? list_add(&fmt->lh, &formats) : |
89 | list_add_tail(&fmt->lh, &formats); | |
1da177e4 | 90 | write_unlock(&binfmt_lock); |
1da177e4 LT |
91 | } |
92 | ||
74641f58 | 93 | EXPORT_SYMBOL(__register_binfmt); |
1da177e4 | 94 | |
f6b450d4 | 95 | void unregister_binfmt(struct linux_binfmt * fmt) |
1da177e4 | 96 | { |
1da177e4 | 97 | write_lock(&binfmt_lock); |
e4dc1b14 | 98 | list_del(&fmt->lh); |
1da177e4 | 99 | write_unlock(&binfmt_lock); |
1da177e4 LT |
100 | } |
101 | ||
102 | EXPORT_SYMBOL(unregister_binfmt); | |
103 | ||
104 | static inline void put_binfmt(struct linux_binfmt * fmt) | |
105 | { | |
106 | module_put(fmt->module); | |
107 | } | |
108 | ||
90f8572b EB |
109 | bool path_noexec(const struct path *path) |
110 | { | |
111 | return (path->mnt->mnt_flags & MNT_NOEXEC) || | |
112 | (path->mnt->mnt_sb->s_iflags & SB_I_NOEXEC); | |
113 | } | |
114 | ||
69369a70 | 115 | #ifdef CONFIG_USELIB |
1da177e4 LT |
116 | /* |
117 | * Note that a shared library must be both readable and executable due to | |
118 | * security reasons. | |
119 | * | |
120 | * Also note that we take the address to load from from the file itself. | |
121 | */ | |
1e7bfb21 | 122 | SYSCALL_DEFINE1(uselib, const char __user *, library) |
1da177e4 | 123 | { |
72c2d531 | 124 | struct linux_binfmt *fmt; |
964bd183 | 125 | struct file *file; |
91a27b2a | 126 | struct filename *tmp = getname(library); |
964bd183 | 127 | int error = PTR_ERR(tmp); |
47c805dc AV |
128 | static const struct open_flags uselib_flags = { |
129 | .open_flag = O_LARGEFILE | O_RDONLY | __FMODE_EXEC, | |
62fb4a15 | 130 | .acc_mode = MAY_READ | MAY_EXEC, |
f9652e10 AV |
131 | .intent = LOOKUP_OPEN, |
132 | .lookup_flags = LOOKUP_FOLLOW, | |
47c805dc | 133 | }; |
964bd183 | 134 | |
6e8341a1 AV |
135 | if (IS_ERR(tmp)) |
136 | goto out; | |
137 | ||
f9652e10 | 138 | file = do_filp_open(AT_FDCWD, tmp, &uselib_flags); |
6e8341a1 AV |
139 | putname(tmp); |
140 | error = PTR_ERR(file); | |
141 | if (IS_ERR(file)) | |
1da177e4 LT |
142 | goto out; |
143 | ||
633fb6ac KC |
144 | /* |
145 | * may_open() has already checked for this, so it should be | |
146 | * impossible to trip now. But we need to be extra cautious | |
147 | * and check again at the very end too. | |
148 | */ | |
db19c91c | 149 | error = -EACCES; |
633fb6ac | 150 | if (WARN_ON_ONCE(!S_ISREG(file_inode(file)->i_mode))) |
1da177e4 LT |
151 | goto exit; |
152 | ||
90f8572b | 153 | if (path_noexec(&file->f_path)) |
1da177e4 LT |
154 | goto exit; |
155 | ||
2a12a9d7 | 156 | fsnotify_open(file); |
6110e3ab | 157 | |
1da177e4 | 158 | error = -ENOEXEC; |
1da177e4 | 159 | |
72c2d531 AV |
160 | read_lock(&binfmt_lock); |
161 | list_for_each_entry(fmt, &formats, lh) { | |
162 | if (!fmt->load_shlib) | |
163 | continue; | |
164 | if (!try_module_get(fmt->module)) | |
165 | continue; | |
1da177e4 | 166 | read_unlock(&binfmt_lock); |
72c2d531 AV |
167 | error = fmt->load_shlib(file); |
168 | read_lock(&binfmt_lock); | |
169 | put_binfmt(fmt); | |
170 | if (error != -ENOEXEC) | |
171 | break; | |
1da177e4 | 172 | } |
72c2d531 | 173 | read_unlock(&binfmt_lock); |
6e8341a1 | 174 | exit: |
1da177e4 LT |
175 | fput(file); |
176 | out: | |
177 | return error; | |
1da177e4 | 178 | } |
69369a70 | 179 | #endif /* #ifdef CONFIG_USELIB */ |
1da177e4 | 180 | |
b6a2fea3 | 181 | #ifdef CONFIG_MMU |
ae6b585e ON |
182 | /* |
183 | * The nascent bprm->mm is not visible until exec_mmap() but it can | |
184 | * use a lot of memory, account these pages in current->mm temporary | |
185 | * for oom_badness()->get_mm_rss(). Once exec succeeds or fails, we | |
186 | * change the counter back via acct_arg_size(0). | |
187 | */ | |
0e028465 | 188 | static void acct_arg_size(struct linux_binprm *bprm, unsigned long pages) |
3c77f845 ON |
189 | { |
190 | struct mm_struct *mm = current->mm; | |
191 | long diff = (long)(pages - bprm->vma_pages); | |
192 | ||
193 | if (!mm || !diff) | |
194 | return; | |
195 | ||
196 | bprm->vma_pages = pages; | |
3c77f845 | 197 | add_mm_counter(mm, MM_ANONPAGES, diff); |
3c77f845 ON |
198 | } |
199 | ||
0e028465 | 200 | static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos, |
b6a2fea3 OW |
201 | int write) |
202 | { | |
203 | struct page *page; | |
204 | int ret; | |
9beae1ea | 205 | unsigned int gup_flags = FOLL_FORCE; |
b6a2fea3 OW |
206 | |
207 | #ifdef CONFIG_STACK_GROWSUP | |
208 | if (write) { | |
d05f3169 | 209 | ret = expand_downwards(bprm->vma, pos); |
b6a2fea3 OW |
210 | if (ret < 0) |
211 | return NULL; | |
212 | } | |
213 | #endif | |
9beae1ea LS |
214 | |
215 | if (write) | |
216 | gup_flags |= FOLL_WRITE; | |
217 | ||
1e987790 DH |
218 | /* |
219 | * We are doing an exec(). 'current' is the process | |
220 | * doing the exec and bprm->mm is the new process's mm. | |
221 | */ | |
9beae1ea | 222 | ret = get_user_pages_remote(current, bprm->mm, pos, 1, gup_flags, |
5b56d49f | 223 | &page, NULL, NULL); |
b6a2fea3 OW |
224 | if (ret <= 0) |
225 | return NULL; | |
226 | ||
655c16a8 ON |
227 | if (write) |
228 | acct_arg_size(bprm, vma_pages(bprm->vma)); | |
b6a2fea3 OW |
229 | |
230 | return page; | |
231 | } | |
232 | ||
233 | static void put_arg_page(struct page *page) | |
234 | { | |
235 | put_page(page); | |
236 | } | |
237 | ||
b6a2fea3 OW |
238 | static void free_arg_pages(struct linux_binprm *bprm) |
239 | { | |
240 | } | |
241 | ||
242 | static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos, | |
243 | struct page *page) | |
244 | { | |
245 | flush_cache_page(bprm->vma, pos, page_to_pfn(page)); | |
246 | } | |
247 | ||
248 | static int __bprm_mm_init(struct linux_binprm *bprm) | |
249 | { | |
eaccbfa5 | 250 | int err; |
b6a2fea3 OW |
251 | struct vm_area_struct *vma = NULL; |
252 | struct mm_struct *mm = bprm->mm; | |
253 | ||
490fc053 | 254 | bprm->vma = vma = vm_area_alloc(mm); |
b6a2fea3 | 255 | if (!vma) |
eaccbfa5 | 256 | return -ENOMEM; |
bfd40eaf | 257 | vma_set_anonymous(vma); |
b6a2fea3 | 258 | |
d8ed45c5 | 259 | if (mmap_write_lock_killable(mm)) { |
f268dfe9 MH |
260 | err = -EINTR; |
261 | goto err_free; | |
262 | } | |
b6a2fea3 OW |
263 | |
264 | /* | |
265 | * Place the stack at the largest stack address the architecture | |
266 | * supports. Later, we'll move this to an appropriate place. We don't | |
267 | * use STACK_TOP because that can depend on attributes which aren't | |
268 | * configured yet. | |
269 | */ | |
aacb3d17 | 270 | BUILD_BUG_ON(VM_STACK_FLAGS & VM_STACK_INCOMPLETE_SETUP); |
b6a2fea3 OW |
271 | vma->vm_end = STACK_TOP_MAX; |
272 | vma->vm_start = vma->vm_end - PAGE_SIZE; | |
d9104d1c | 273 | vma->vm_flags = VM_SOFTDIRTY | VM_STACK_FLAGS | VM_STACK_INCOMPLETE_SETUP; |
3ed75eb8 | 274 | vma->vm_page_prot = vm_get_page_prot(vma->vm_flags); |
462e635e | 275 | |
b6a2fea3 | 276 | err = insert_vm_struct(mm, vma); |
eaccbfa5 | 277 | if (err) |
b6a2fea3 | 278 | goto err; |
b6a2fea3 OW |
279 | |
280 | mm->stack_vm = mm->total_vm = 1; | |
d8ed45c5 | 281 | mmap_write_unlock(mm); |
b6a2fea3 | 282 | bprm->p = vma->vm_end - sizeof(void *); |
b6a2fea3 | 283 | return 0; |
b6a2fea3 | 284 | err: |
d8ed45c5 | 285 | mmap_write_unlock(mm); |
f268dfe9 | 286 | err_free: |
eaccbfa5 | 287 | bprm->vma = NULL; |
3928d4f5 | 288 | vm_area_free(vma); |
b6a2fea3 OW |
289 | return err; |
290 | } | |
291 | ||
292 | static bool valid_arg_len(struct linux_binprm *bprm, long len) | |
293 | { | |
294 | return len <= MAX_ARG_STRLEN; | |
295 | } | |
296 | ||
297 | #else | |
298 | ||
0e028465 | 299 | static inline void acct_arg_size(struct linux_binprm *bprm, unsigned long pages) |
3c77f845 ON |
300 | { |
301 | } | |
302 | ||
0e028465 | 303 | static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos, |
b6a2fea3 OW |
304 | int write) |
305 | { | |
306 | struct page *page; | |
307 | ||
308 | page = bprm->page[pos / PAGE_SIZE]; | |
309 | if (!page && write) { | |
310 | page = alloc_page(GFP_HIGHUSER|__GFP_ZERO); | |
311 | if (!page) | |
312 | return NULL; | |
313 | bprm->page[pos / PAGE_SIZE] = page; | |
314 | } | |
315 | ||
316 | return page; | |
317 | } | |
318 | ||
319 | static void put_arg_page(struct page *page) | |
320 | { | |
321 | } | |
322 | ||
323 | static void free_arg_page(struct linux_binprm *bprm, int i) | |
324 | { | |
325 | if (bprm->page[i]) { | |
326 | __free_page(bprm->page[i]); | |
327 | bprm->page[i] = NULL; | |
328 | } | |
329 | } | |
330 | ||
331 | static void free_arg_pages(struct linux_binprm *bprm) | |
332 | { | |
333 | int i; | |
334 | ||
335 | for (i = 0; i < MAX_ARG_PAGES; i++) | |
336 | free_arg_page(bprm, i); | |
337 | } | |
338 | ||
339 | static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos, | |
340 | struct page *page) | |
341 | { | |
342 | } | |
343 | ||
344 | static int __bprm_mm_init(struct linux_binprm *bprm) | |
345 | { | |
346 | bprm->p = PAGE_SIZE * MAX_ARG_PAGES - sizeof(void *); | |
347 | return 0; | |
348 | } | |
349 | ||
350 | static bool valid_arg_len(struct linux_binprm *bprm, long len) | |
351 | { | |
352 | return len <= bprm->p; | |
353 | } | |
354 | ||
355 | #endif /* CONFIG_MMU */ | |
356 | ||
357 | /* | |
358 | * Create a new mm_struct and populate it with a temporary stack | |
359 | * vm_area_struct. We don't have enough context at this point to set the stack | |
360 | * flags, permissions, and offset, so we use temporary values. We'll update | |
361 | * them later in setup_arg_pages(). | |
362 | */ | |
9cc64cea | 363 | static int bprm_mm_init(struct linux_binprm *bprm) |
b6a2fea3 OW |
364 | { |
365 | int err; | |
366 | struct mm_struct *mm = NULL; | |
367 | ||
368 | bprm->mm = mm = mm_alloc(); | |
369 | err = -ENOMEM; | |
370 | if (!mm) | |
371 | goto err; | |
372 | ||
c31dbb14 KC |
373 | /* Save current stack limit for all calculations made during exec. */ |
374 | task_lock(current->group_leader); | |
375 | bprm->rlim_stack = current->signal->rlim[RLIMIT_STACK]; | |
376 | task_unlock(current->group_leader); | |
377 | ||
b6a2fea3 OW |
378 | err = __bprm_mm_init(bprm); |
379 | if (err) | |
380 | goto err; | |
381 | ||
382 | return 0; | |
383 | ||
384 | err: | |
385 | if (mm) { | |
386 | bprm->mm = NULL; | |
387 | mmdrop(mm); | |
388 | } | |
389 | ||
390 | return err; | |
391 | } | |
392 | ||
ba2d0162 | 393 | struct user_arg_ptr { |
0e028465 ON |
394 | #ifdef CONFIG_COMPAT |
395 | bool is_compat; | |
396 | #endif | |
397 | union { | |
398 | const char __user *const __user *native; | |
399 | #ifdef CONFIG_COMPAT | |
38b983b3 | 400 | const compat_uptr_t __user *compat; |
0e028465 ON |
401 | #endif |
402 | } ptr; | |
ba2d0162 ON |
403 | }; |
404 | ||
405 | static const char __user *get_user_arg_ptr(struct user_arg_ptr argv, int nr) | |
1d1dbf81 | 406 | { |
0e028465 ON |
407 | const char __user *native; |
408 | ||
409 | #ifdef CONFIG_COMPAT | |
410 | if (unlikely(argv.is_compat)) { | |
411 | compat_uptr_t compat; | |
412 | ||
413 | if (get_user(compat, argv.ptr.compat + nr)) | |
414 | return ERR_PTR(-EFAULT); | |
1d1dbf81 | 415 | |
0e028465 ON |
416 | return compat_ptr(compat); |
417 | } | |
418 | #endif | |
419 | ||
420 | if (get_user(native, argv.ptr.native + nr)) | |
1d1dbf81 ON |
421 | return ERR_PTR(-EFAULT); |
422 | ||
0e028465 | 423 | return native; |
1d1dbf81 ON |
424 | } |
425 | ||
1da177e4 LT |
426 | /* |
427 | * count() counts the number of strings in array ARGV. | |
428 | */ | |
ba2d0162 | 429 | static int count(struct user_arg_ptr argv, int max) |
1da177e4 LT |
430 | { |
431 | int i = 0; | |
432 | ||
0e028465 | 433 | if (argv.ptr.native != NULL) { |
1da177e4 | 434 | for (;;) { |
1d1dbf81 | 435 | const char __user *p = get_user_arg_ptr(argv, i); |
1da177e4 | 436 | |
1da177e4 LT |
437 | if (!p) |
438 | break; | |
1d1dbf81 ON |
439 | |
440 | if (IS_ERR(p)) | |
441 | return -EFAULT; | |
442 | ||
6d92d4f6 | 443 | if (i >= max) |
1da177e4 | 444 | return -E2BIG; |
6d92d4f6 | 445 | ++i; |
9aea5a65 RM |
446 | |
447 | if (fatal_signal_pending(current)) | |
448 | return -ERESTARTNOHAND; | |
1da177e4 LT |
449 | cond_resched(); |
450 | } | |
451 | } | |
452 | return i; | |
453 | } | |
454 | ||
be619f7f EB |
455 | static int count_strings_kernel(const char *const *argv) |
456 | { | |
457 | int i; | |
458 | ||
459 | if (!argv) | |
460 | return 0; | |
461 | ||
462 | for (i = 0; argv[i]; ++i) { | |
463 | if (i >= MAX_ARG_STRINGS) | |
464 | return -E2BIG; | |
465 | if (fatal_signal_pending(current)) | |
466 | return -ERESTARTNOHAND; | |
467 | cond_resched(); | |
468 | } | |
469 | return i; | |
470 | } | |
471 | ||
d8b9cd54 | 472 | static int bprm_stack_limits(struct linux_binprm *bprm) |
655c16a8 ON |
473 | { |
474 | unsigned long limit, ptr_size; | |
475 | ||
655c16a8 ON |
476 | /* |
477 | * Limit to 1/4 of the max stack size or 3/4 of _STK_LIM | |
478 | * (whichever is smaller) for the argv+env strings. | |
479 | * This ensures that: | |
480 | * - the remaining binfmt code will not run out of stack space, | |
481 | * - the program will have a reasonable amount of stack left | |
482 | * to work from. | |
483 | */ | |
484 | limit = _STK_LIM / 4 * 3; | |
485 | limit = min(limit, bprm->rlim_stack.rlim_cur / 4); | |
486 | /* | |
487 | * We've historically supported up to 32 pages (ARG_MAX) | |
488 | * of argument strings even with small stacks | |
489 | */ | |
490 | limit = max_t(unsigned long, limit, ARG_MAX); | |
491 | /* | |
492 | * We must account for the size of all the argv and envp pointers to | |
493 | * the argv and envp strings, since they will also take up space in | |
494 | * the stack. They aren't stored until much later when we can't | |
495 | * signal to the parent that the child has run out of stack space. | |
496 | * Instead, calculate it here so it's possible to fail gracefully. | |
497 | */ | |
498 | ptr_size = (bprm->argc + bprm->envc) * sizeof(void *); | |
499 | if (limit <= ptr_size) | |
500 | return -E2BIG; | |
501 | limit -= ptr_size; | |
502 | ||
503 | bprm->argmin = bprm->p - limit; | |
504 | return 0; | |
505 | } | |
506 | ||
1da177e4 | 507 | /* |
b6a2fea3 OW |
508 | * 'copy_strings()' copies argument/environment strings from the old |
509 | * processes's memory to the new process's stack. The call to get_user_pages() | |
510 | * ensures the destination page is created and not swapped out. | |
1da177e4 | 511 | */ |
ba2d0162 | 512 | static int copy_strings(int argc, struct user_arg_ptr argv, |
75c96f85 | 513 | struct linux_binprm *bprm) |
1da177e4 LT |
514 | { |
515 | struct page *kmapped_page = NULL; | |
516 | char *kaddr = NULL; | |
b6a2fea3 | 517 | unsigned long kpos = 0; |
1da177e4 LT |
518 | int ret; |
519 | ||
520 | while (argc-- > 0) { | |
d7627467 | 521 | const char __user *str; |
1da177e4 LT |
522 | int len; |
523 | unsigned long pos; | |
524 | ||
1d1dbf81 ON |
525 | ret = -EFAULT; |
526 | str = get_user_arg_ptr(argv, argc); | |
527 | if (IS_ERR(str)) | |
1da177e4 | 528 | goto out; |
1da177e4 | 529 | |
1d1dbf81 ON |
530 | len = strnlen_user(str, MAX_ARG_STRLEN); |
531 | if (!len) | |
532 | goto out; | |
533 | ||
534 | ret = -E2BIG; | |
535 | if (!valid_arg_len(bprm, len)) | |
1da177e4 | 536 | goto out; |
1da177e4 | 537 | |
b6a2fea3 | 538 | /* We're going to work our way backwords. */ |
1da177e4 | 539 | pos = bprm->p; |
b6a2fea3 OW |
540 | str += len; |
541 | bprm->p -= len; | |
655c16a8 ON |
542 | #ifdef CONFIG_MMU |
543 | if (bprm->p < bprm->argmin) | |
544 | goto out; | |
545 | #endif | |
1da177e4 LT |
546 | |
547 | while (len > 0) { | |
1da177e4 | 548 | int offset, bytes_to_copy; |
1da177e4 | 549 | |
9aea5a65 RM |
550 | if (fatal_signal_pending(current)) { |
551 | ret = -ERESTARTNOHAND; | |
552 | goto out; | |
553 | } | |
7993bc1f RM |
554 | cond_resched(); |
555 | ||
1da177e4 | 556 | offset = pos % PAGE_SIZE; |
b6a2fea3 OW |
557 | if (offset == 0) |
558 | offset = PAGE_SIZE; | |
559 | ||
560 | bytes_to_copy = offset; | |
561 | if (bytes_to_copy > len) | |
562 | bytes_to_copy = len; | |
563 | ||
564 | offset -= bytes_to_copy; | |
565 | pos -= bytes_to_copy; | |
566 | str -= bytes_to_copy; | |
567 | len -= bytes_to_copy; | |
568 | ||
569 | if (!kmapped_page || kpos != (pos & PAGE_MASK)) { | |
570 | struct page *page; | |
571 | ||
572 | page = get_arg_page(bprm, pos, 1); | |
1da177e4 | 573 | if (!page) { |
b6a2fea3 | 574 | ret = -E2BIG; |
1da177e4 LT |
575 | goto out; |
576 | } | |
1da177e4 | 577 | |
b6a2fea3 OW |
578 | if (kmapped_page) { |
579 | flush_kernel_dcache_page(kmapped_page); | |
1da177e4 | 580 | kunmap(kmapped_page); |
b6a2fea3 OW |
581 | put_arg_page(kmapped_page); |
582 | } | |
1da177e4 LT |
583 | kmapped_page = page; |
584 | kaddr = kmap(kmapped_page); | |
b6a2fea3 OW |
585 | kpos = pos & PAGE_MASK; |
586 | flush_arg_page(bprm, kpos, kmapped_page); | |
1da177e4 | 587 | } |
b6a2fea3 | 588 | if (copy_from_user(kaddr+offset, str, bytes_to_copy)) { |
1da177e4 LT |
589 | ret = -EFAULT; |
590 | goto out; | |
591 | } | |
1da177e4 LT |
592 | } |
593 | } | |
594 | ret = 0; | |
595 | out: | |
b6a2fea3 OW |
596 | if (kmapped_page) { |
597 | flush_kernel_dcache_page(kmapped_page); | |
1da177e4 | 598 | kunmap(kmapped_page); |
b6a2fea3 OW |
599 | put_arg_page(kmapped_page); |
600 | } | |
1da177e4 LT |
601 | return ret; |
602 | } | |
603 | ||
604 | /* | |
986db2d1 | 605 | * Copy and argument/environment string from the kernel to the processes stack. |
1da177e4 | 606 | */ |
986db2d1 | 607 | int copy_string_kernel(const char *arg, struct linux_binprm *bprm) |
1da177e4 | 608 | { |
762a3af6 CH |
609 | int len = strnlen(arg, MAX_ARG_STRLEN) + 1 /* terminating NUL */; |
610 | unsigned long pos = bprm->p; | |
611 | ||
612 | if (len == 0) | |
613 | return -EFAULT; | |
614 | if (!valid_arg_len(bprm, len)) | |
615 | return -E2BIG; | |
616 | ||
617 | /* We're going to work our way backwards. */ | |
618 | arg += len; | |
619 | bprm->p -= len; | |
620 | if (IS_ENABLED(CONFIG_MMU) && bprm->p < bprm->argmin) | |
621 | return -E2BIG; | |
ba2d0162 | 622 | |
762a3af6 CH |
623 | while (len > 0) { |
624 | unsigned int bytes_to_copy = min_t(unsigned int, len, | |
625 | min_not_zero(offset_in_page(pos), PAGE_SIZE)); | |
626 | struct page *page; | |
627 | char *kaddr; | |
ba2d0162 | 628 | |
762a3af6 CH |
629 | pos -= bytes_to_copy; |
630 | arg -= bytes_to_copy; | |
631 | len -= bytes_to_copy; | |
ba2d0162 | 632 | |
762a3af6 CH |
633 | page = get_arg_page(bprm, pos, 1); |
634 | if (!page) | |
635 | return -E2BIG; | |
636 | kaddr = kmap_atomic(page); | |
637 | flush_arg_page(bprm, pos & PAGE_MASK, page); | |
638 | memcpy(kaddr + offset_in_page(pos), arg, bytes_to_copy); | |
639 | flush_kernel_dcache_page(page); | |
640 | kunmap_atomic(kaddr); | |
641 | put_arg_page(page); | |
642 | } | |
643 | ||
644 | return 0; | |
1da177e4 | 645 | } |
986db2d1 | 646 | EXPORT_SYMBOL(copy_string_kernel); |
1da177e4 | 647 | |
be619f7f EB |
648 | static int copy_strings_kernel(int argc, const char *const *argv, |
649 | struct linux_binprm *bprm) | |
650 | { | |
651 | while (argc-- > 0) { | |
652 | int ret = copy_string_kernel(argv[argc], bprm); | |
653 | if (ret < 0) | |
654 | return ret; | |
655 | if (fatal_signal_pending(current)) | |
656 | return -ERESTARTNOHAND; | |
657 | cond_resched(); | |
658 | } | |
659 | return 0; | |
660 | } | |
661 | ||
1da177e4 | 662 | #ifdef CONFIG_MMU |
b6a2fea3 | 663 | |
1da177e4 | 664 | /* |
b6a2fea3 OW |
665 | * During bprm_mm_init(), we create a temporary stack at STACK_TOP_MAX. Once |
666 | * the binfmt code determines where the new stack should reside, we shift it to | |
667 | * its final location. The process proceeds as follows: | |
1da177e4 | 668 | * |
b6a2fea3 OW |
669 | * 1) Use shift to calculate the new vma endpoints. |
670 | * 2) Extend vma to cover both the old and new ranges. This ensures the | |
671 | * arguments passed to subsequent functions are consistent. | |
672 | * 3) Move vma's page tables to the new range. | |
673 | * 4) Free up any cleared pgd range. | |
674 | * 5) Shrink the vma to cover only the new range. | |
1da177e4 | 675 | */ |
b6a2fea3 | 676 | static int shift_arg_pages(struct vm_area_struct *vma, unsigned long shift) |
1da177e4 LT |
677 | { |
678 | struct mm_struct *mm = vma->vm_mm; | |
b6a2fea3 OW |
679 | unsigned long old_start = vma->vm_start; |
680 | unsigned long old_end = vma->vm_end; | |
681 | unsigned long length = old_end - old_start; | |
682 | unsigned long new_start = old_start - shift; | |
683 | unsigned long new_end = old_end - shift; | |
d16dfc55 | 684 | struct mmu_gather tlb; |
1da177e4 | 685 | |
b6a2fea3 | 686 | BUG_ON(new_start > new_end); |
1da177e4 | 687 | |
b6a2fea3 OW |
688 | /* |
689 | * ensure there are no vmas between where we want to go | |
690 | * and where we are | |
691 | */ | |
692 | if (vma != find_vma(mm, new_start)) | |
693 | return -EFAULT; | |
694 | ||
695 | /* | |
696 | * cover the whole range: [new_start, old_end) | |
697 | */ | |
5beb4930 RR |
698 | if (vma_adjust(vma, new_start, old_end, vma->vm_pgoff, NULL)) |
699 | return -ENOMEM; | |
b6a2fea3 OW |
700 | |
701 | /* | |
702 | * move the page tables downwards, on failure we rely on | |
703 | * process cleanup to remove whatever mess we made. | |
704 | */ | |
705 | if (length != move_page_tables(vma, old_start, | |
38a76013 | 706 | vma, new_start, length, false)) |
b6a2fea3 OW |
707 | return -ENOMEM; |
708 | ||
709 | lru_add_drain(); | |
2b047252 | 710 | tlb_gather_mmu(&tlb, mm, old_start, old_end); |
b6a2fea3 OW |
711 | if (new_end > old_start) { |
712 | /* | |
713 | * when the old and new regions overlap clear from new_end. | |
714 | */ | |
d16dfc55 | 715 | free_pgd_range(&tlb, new_end, old_end, new_end, |
6ee8630e | 716 | vma->vm_next ? vma->vm_next->vm_start : USER_PGTABLES_CEILING); |
b6a2fea3 OW |
717 | } else { |
718 | /* | |
719 | * otherwise, clean from old_start; this is done to not touch | |
720 | * the address space in [new_end, old_start) some architectures | |
721 | * have constraints on va-space that make this illegal (IA64) - | |
722 | * for the others its just a little faster. | |
723 | */ | |
d16dfc55 | 724 | free_pgd_range(&tlb, old_start, old_end, new_end, |
6ee8630e | 725 | vma->vm_next ? vma->vm_next->vm_start : USER_PGTABLES_CEILING); |
1da177e4 | 726 | } |
2b047252 | 727 | tlb_finish_mmu(&tlb, old_start, old_end); |
b6a2fea3 OW |
728 | |
729 | /* | |
5beb4930 | 730 | * Shrink the vma to just the new range. Always succeeds. |
b6a2fea3 OW |
731 | */ |
732 | vma_adjust(vma, new_start, new_end, vma->vm_pgoff, NULL); | |
733 | ||
734 | return 0; | |
1da177e4 LT |
735 | } |
736 | ||
b6a2fea3 OW |
737 | /* |
738 | * Finalizes the stack vm_area_struct. The flags and permissions are updated, | |
739 | * the stack is optionally relocated, and some extra space is added. | |
740 | */ | |
1da177e4 LT |
741 | int setup_arg_pages(struct linux_binprm *bprm, |
742 | unsigned long stack_top, | |
743 | int executable_stack) | |
744 | { | |
b6a2fea3 OW |
745 | unsigned long ret; |
746 | unsigned long stack_shift; | |
1da177e4 | 747 | struct mm_struct *mm = current->mm; |
b6a2fea3 OW |
748 | struct vm_area_struct *vma = bprm->vma; |
749 | struct vm_area_struct *prev = NULL; | |
750 | unsigned long vm_flags; | |
751 | unsigned long stack_base; | |
803bf5ec MN |
752 | unsigned long stack_size; |
753 | unsigned long stack_expand; | |
754 | unsigned long rlim_stack; | |
1da177e4 LT |
755 | |
756 | #ifdef CONFIG_STACK_GROWSUP | |
d71f290b | 757 | /* Limit stack size */ |
c31dbb14 | 758 | stack_base = bprm->rlim_stack.rlim_max; |
d71f290b JH |
759 | if (stack_base > STACK_SIZE_MAX) |
760 | stack_base = STACK_SIZE_MAX; | |
1da177e4 | 761 | |
d045c77c HD |
762 | /* Add space for stack randomization. */ |
763 | stack_base += (STACK_RND_MASK << PAGE_SHIFT); | |
764 | ||
b6a2fea3 OW |
765 | /* Make sure we didn't let the argument array grow too large. */ |
766 | if (vma->vm_end - vma->vm_start > stack_base) | |
767 | return -ENOMEM; | |
1da177e4 | 768 | |
b6a2fea3 | 769 | stack_base = PAGE_ALIGN(stack_top - stack_base); |
1da177e4 | 770 | |
b6a2fea3 OW |
771 | stack_shift = vma->vm_start - stack_base; |
772 | mm->arg_start = bprm->p - stack_shift; | |
773 | bprm->p = vma->vm_end - stack_shift; | |
1da177e4 | 774 | #else |
b6a2fea3 OW |
775 | stack_top = arch_align_stack(stack_top); |
776 | stack_top = PAGE_ALIGN(stack_top); | |
1b528181 RM |
777 | |
778 | if (unlikely(stack_top < mmap_min_addr) || | |
779 | unlikely(vma->vm_end - vma->vm_start >= stack_top - mmap_min_addr)) | |
780 | return -ENOMEM; | |
781 | ||
b6a2fea3 OW |
782 | stack_shift = vma->vm_end - stack_top; |
783 | ||
784 | bprm->p -= stack_shift; | |
1da177e4 | 785 | mm->arg_start = bprm->p; |
1da177e4 LT |
786 | #endif |
787 | ||
1da177e4 | 788 | if (bprm->loader) |
b6a2fea3 OW |
789 | bprm->loader -= stack_shift; |
790 | bprm->exec -= stack_shift; | |
1da177e4 | 791 | |
d8ed45c5 | 792 | if (mmap_write_lock_killable(mm)) |
f268dfe9 MH |
793 | return -EINTR; |
794 | ||
96a8e13e | 795 | vm_flags = VM_STACK_FLAGS; |
b6a2fea3 OW |
796 | |
797 | /* | |
798 | * Adjust stack execute permissions; explicitly enable for | |
799 | * EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X and leave alone | |
800 | * (arch default) otherwise. | |
801 | */ | |
802 | if (unlikely(executable_stack == EXSTACK_ENABLE_X)) | |
803 | vm_flags |= VM_EXEC; | |
804 | else if (executable_stack == EXSTACK_DISABLE_X) | |
805 | vm_flags &= ~VM_EXEC; | |
806 | vm_flags |= mm->def_flags; | |
a8bef8ff | 807 | vm_flags |= VM_STACK_INCOMPLETE_SETUP; |
b6a2fea3 OW |
808 | |
809 | ret = mprotect_fixup(vma, &prev, vma->vm_start, vma->vm_end, | |
810 | vm_flags); | |
811 | if (ret) | |
812 | goto out_unlock; | |
813 | BUG_ON(prev != vma); | |
814 | ||
47a2ebb7 AD |
815 | if (unlikely(vm_flags & VM_EXEC)) { |
816 | pr_warn_once("process '%pD4' started with executable stack\n", | |
817 | bprm->file); | |
818 | } | |
819 | ||
b6a2fea3 OW |
820 | /* Move stack pages down in memory. */ |
821 | if (stack_shift) { | |
822 | ret = shift_arg_pages(vma, stack_shift); | |
fc63cf23 AB |
823 | if (ret) |
824 | goto out_unlock; | |
1da177e4 LT |
825 | } |
826 | ||
a8bef8ff MG |
827 | /* mprotect_fixup is overkill to remove the temporary stack flags */ |
828 | vma->vm_flags &= ~VM_STACK_INCOMPLETE_SETUP; | |
829 | ||
5ef097dd | 830 | stack_expand = 131072UL; /* randomly 32*4k (or 2*64k) pages */ |
803bf5ec MN |
831 | stack_size = vma->vm_end - vma->vm_start; |
832 | /* | |
833 | * Align this down to a page boundary as expand_stack | |
834 | * will align it up. | |
835 | */ | |
c31dbb14 | 836 | rlim_stack = bprm->rlim_stack.rlim_cur & PAGE_MASK; |
b6a2fea3 | 837 | #ifdef CONFIG_STACK_GROWSUP |
803bf5ec MN |
838 | if (stack_size + stack_expand > rlim_stack) |
839 | stack_base = vma->vm_start + rlim_stack; | |
840 | else | |
841 | stack_base = vma->vm_end + stack_expand; | |
b6a2fea3 | 842 | #else |
803bf5ec MN |
843 | if (stack_size + stack_expand > rlim_stack) |
844 | stack_base = vma->vm_end - rlim_stack; | |
845 | else | |
846 | stack_base = vma->vm_start - stack_expand; | |
b6a2fea3 | 847 | #endif |
3af9e859 | 848 | current->mm->start_stack = bprm->p; |
b6a2fea3 OW |
849 | ret = expand_stack(vma, stack_base); |
850 | if (ret) | |
851 | ret = -EFAULT; | |
852 | ||
853 | out_unlock: | |
d8ed45c5 | 854 | mmap_write_unlock(mm); |
fc63cf23 | 855 | return ret; |
1da177e4 | 856 | } |
1da177e4 LT |
857 | EXPORT_SYMBOL(setup_arg_pages); |
858 | ||
7e7ec6a9 NP |
859 | #else |
860 | ||
861 | /* | |
862 | * Transfer the program arguments and environment from the holding pages | |
863 | * onto the stack. The provided stack pointer is adjusted accordingly. | |
864 | */ | |
865 | int transfer_args_to_stack(struct linux_binprm *bprm, | |
866 | unsigned long *sp_location) | |
867 | { | |
868 | unsigned long index, stop, sp; | |
869 | int ret = 0; | |
870 | ||
871 | stop = bprm->p >> PAGE_SHIFT; | |
872 | sp = *sp_location; | |
873 | ||
874 | for (index = MAX_ARG_PAGES - 1; index >= stop; index--) { | |
875 | unsigned int offset = index == stop ? bprm->p & ~PAGE_MASK : 0; | |
876 | char *src = kmap(bprm->page[index]) + offset; | |
877 | sp -= PAGE_SIZE - offset; | |
878 | if (copy_to_user((void *) sp, src, PAGE_SIZE - offset) != 0) | |
879 | ret = -EFAULT; | |
880 | kunmap(bprm->page[index]); | |
881 | if (ret) | |
882 | goto out; | |
883 | } | |
884 | ||
885 | *sp_location = sp; | |
886 | ||
887 | out: | |
888 | return ret; | |
889 | } | |
890 | EXPORT_SYMBOL(transfer_args_to_stack); | |
891 | ||
1da177e4 LT |
892 | #endif /* CONFIG_MMU */ |
893 | ||
51f39a1f | 894 | static struct file *do_open_execat(int fd, struct filename *name, int flags) |
1da177e4 | 895 | { |
1da177e4 | 896 | struct file *file; |
e56b6a5d | 897 | int err; |
51f39a1f | 898 | struct open_flags open_exec_flags = { |
47c805dc | 899 | .open_flag = O_LARGEFILE | O_RDONLY | __FMODE_EXEC, |
62fb4a15 | 900 | .acc_mode = MAY_EXEC, |
f9652e10 AV |
901 | .intent = LOOKUP_OPEN, |
902 | .lookup_flags = LOOKUP_FOLLOW, | |
47c805dc | 903 | }; |
1da177e4 | 904 | |
51f39a1f DD |
905 | if ((flags & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0) |
906 | return ERR_PTR(-EINVAL); | |
907 | if (flags & AT_SYMLINK_NOFOLLOW) | |
908 | open_exec_flags.lookup_flags &= ~LOOKUP_FOLLOW; | |
909 | if (flags & AT_EMPTY_PATH) | |
910 | open_exec_flags.lookup_flags |= LOOKUP_EMPTY; | |
911 | ||
912 | file = do_filp_open(fd, name, &open_exec_flags); | |
6e8341a1 | 913 | if (IS_ERR(file)) |
e56b6a5d CH |
914 | goto out; |
915 | ||
633fb6ac KC |
916 | /* |
917 | * may_open() has already checked for this, so it should be | |
918 | * impossible to trip now. But we need to be extra cautious | |
919 | * and check again at the very end too. | |
920 | */ | |
e56b6a5d | 921 | err = -EACCES; |
633fb6ac | 922 | if (WARN_ON_ONCE(!S_ISREG(file_inode(file)->i_mode))) |
6e8341a1 | 923 | goto exit; |
e56b6a5d | 924 | |
90f8572b | 925 | if (path_noexec(&file->f_path)) |
6e8341a1 | 926 | goto exit; |
e56b6a5d CH |
927 | |
928 | err = deny_write_access(file); | |
6e8341a1 AV |
929 | if (err) |
930 | goto exit; | |
1da177e4 | 931 | |
51f39a1f DD |
932 | if (name->name[0] != '\0') |
933 | fsnotify_open(file); | |
934 | ||
6e8341a1 | 935 | out: |
e56b6a5d CH |
936 | return file; |
937 | ||
6e8341a1 AV |
938 | exit: |
939 | fput(file); | |
e56b6a5d CH |
940 | return ERR_PTR(err); |
941 | } | |
c4ad8f98 LT |
942 | |
943 | struct file *open_exec(const char *name) | |
944 | { | |
51689104 PM |
945 | struct filename *filename = getname_kernel(name); |
946 | struct file *f = ERR_CAST(filename); | |
947 | ||
948 | if (!IS_ERR(filename)) { | |
949 | f = do_open_execat(AT_FDCWD, filename, 0); | |
950 | putname(filename); | |
951 | } | |
952 | return f; | |
c4ad8f98 | 953 | } |
1da177e4 LT |
954 | EXPORT_SYMBOL(open_exec); |
955 | ||
b44a7dfc | 956 | int kernel_read_file(struct file *file, void **buf, loff_t *size, |
bc8ca5b9 | 957 | loff_t max_size, enum kernel_read_file_id id) |
b44a7dfc MZ |
958 | { |
959 | loff_t i_size, pos; | |
960 | ssize_t bytes = 0; | |
961 | int ret; | |
962 | ||
963 | if (!S_ISREG(file_inode(file)->i_mode) || max_size < 0) | |
964 | return -EINVAL; | |
965 | ||
7bd698b3 | 966 | ret = deny_write_access(file); |
39eeb4fb MZ |
967 | if (ret) |
968 | return ret; | |
969 | ||
7bd698b3 | 970 | ret = security_kernel_read_file(file, id); |
39d637af | 971 | if (ret) |
7bd698b3 | 972 | goto out; |
39d637af | 973 | |
b44a7dfc | 974 | i_size = i_size_read(file_inode(file)); |
39d637af DK |
975 | if (i_size <= 0) { |
976 | ret = -EINVAL; | |
977 | goto out; | |
978 | } | |
691115c3 EB |
979 | if (i_size > SIZE_MAX || (max_size > 0 && i_size > max_size)) { |
980 | ret = -EFBIG; | |
981 | goto out; | |
982 | } | |
b44a7dfc | 983 | |
a098ecd2 SB |
984 | if (id != READING_FIRMWARE_PREALLOC_BUFFER) |
985 | *buf = vmalloc(i_size); | |
39d637af DK |
986 | if (!*buf) { |
987 | ret = -ENOMEM; | |
988 | goto out; | |
989 | } | |
b44a7dfc MZ |
990 | |
991 | pos = 0; | |
992 | while (pos < i_size) { | |
bdd1d2d3 | 993 | bytes = kernel_read(file, *buf + pos, i_size - pos, &pos); |
b44a7dfc MZ |
994 | if (bytes < 0) { |
995 | ret = bytes; | |
f612acfa | 996 | goto out_free; |
b44a7dfc MZ |
997 | } |
998 | ||
999 | if (bytes == 0) | |
1000 | break; | |
b44a7dfc MZ |
1001 | } |
1002 | ||
1003 | if (pos != i_size) { | |
1004 | ret = -EIO; | |
39d637af | 1005 | goto out_free; |
b44a7dfc MZ |
1006 | } |
1007 | ||
bc8ca5b9 | 1008 | ret = security_kernel_post_read_file(file, *buf, i_size, id); |
b44a7dfc MZ |
1009 | if (!ret) |
1010 | *size = pos; | |
1011 | ||
39d637af | 1012 | out_free: |
b44a7dfc | 1013 | if (ret < 0) { |
a098ecd2 SB |
1014 | if (id != READING_FIRMWARE_PREALLOC_BUFFER) { |
1015 | vfree(*buf); | |
1016 | *buf = NULL; | |
1017 | } | |
b44a7dfc | 1018 | } |
39d637af DK |
1019 | |
1020 | out: | |
1021 | allow_write_access(file); | |
b44a7dfc MZ |
1022 | return ret; |
1023 | } | |
1024 | EXPORT_SYMBOL_GPL(kernel_read_file); | |
1025 | ||
711aab1d | 1026 | int kernel_read_file_from_path(const char *path, void **buf, loff_t *size, |
09596b94 MZ |
1027 | loff_t max_size, enum kernel_read_file_id id) |
1028 | { | |
1029 | struct file *file; | |
1030 | int ret; | |
1031 | ||
1032 | if (!path || !*path) | |
1033 | return -EINVAL; | |
1034 | ||
1035 | file = filp_open(path, O_RDONLY, 0); | |
1036 | if (IS_ERR(file)) | |
1037 | return PTR_ERR(file); | |
1038 | ||
1039 | ret = kernel_read_file(file, buf, size, max_size, id); | |
1040 | fput(file); | |
1041 | return ret; | |
1042 | } | |
1043 | EXPORT_SYMBOL_GPL(kernel_read_file_from_path); | |
1044 | ||
901cff7c TM |
1045 | int kernel_read_file_from_path_initns(const char *path, void **buf, |
1046 | loff_t *size, loff_t max_size, | |
1047 | enum kernel_read_file_id id) | |
1048 | { | |
1049 | struct file *file; | |
1050 | struct path root; | |
1051 | int ret; | |
1052 | ||
1053 | if (!path || !*path) | |
1054 | return -EINVAL; | |
1055 | ||
1056 | task_lock(&init_task); | |
1057 | get_fs_root(init_task.fs, &root); | |
1058 | task_unlock(&init_task); | |
1059 | ||
1060 | file = file_open_root(root.dentry, root.mnt, path, O_RDONLY, 0); | |
1061 | path_put(&root); | |
1062 | if (IS_ERR(file)) | |
1063 | return PTR_ERR(file); | |
1064 | ||
1065 | ret = kernel_read_file(file, buf, size, max_size, id); | |
1066 | fput(file); | |
1067 | return ret; | |
1068 | } | |
1069 | EXPORT_SYMBOL_GPL(kernel_read_file_from_path_initns); | |
1070 | ||
b844f0ec MZ |
1071 | int kernel_read_file_from_fd(int fd, void **buf, loff_t *size, loff_t max_size, |
1072 | enum kernel_read_file_id id) | |
1073 | { | |
1074 | struct fd f = fdget(fd); | |
1075 | int ret = -EBADF; | |
1076 | ||
1077 | if (!f.file) | |
1078 | goto out; | |
1079 | ||
1080 | ret = kernel_read_file(f.file, buf, size, max_size, id); | |
1081 | out: | |
1082 | fdput(f); | |
1083 | return ret; | |
1084 | } | |
1085 | EXPORT_SYMBOL_GPL(kernel_read_file_from_fd); | |
1086 | ||
48304f79 CH |
1087 | #if defined(CONFIG_HAVE_AOUT) || defined(CONFIG_BINFMT_FLAT) || \ |
1088 | defined(CONFIG_BINFMT_ELF_FDPIC) | |
3dc20cb2 AV |
1089 | ssize_t read_code(struct file *file, unsigned long addr, loff_t pos, size_t len) |
1090 | { | |
ec695579 | 1091 | ssize_t res = vfs_read(file, (void __user *)addr, len, &pos); |
3dc20cb2 | 1092 | if (res > 0) |
bce2b68b | 1093 | flush_icache_user_range(addr, addr + len); |
3dc20cb2 AV |
1094 | return res; |
1095 | } | |
1096 | EXPORT_SYMBOL(read_code); | |
48304f79 | 1097 | #endif |
3dc20cb2 | 1098 | |
eea96732 EB |
1099 | /* |
1100 | * Maps the mm_struct mm into the current task struct. | |
1101 | * On success, this function returns with the mutex | |
1102 | * exec_update_mutex locked. | |
1103 | */ | |
1da177e4 LT |
1104 | static int exec_mmap(struct mm_struct *mm) |
1105 | { | |
1106 | struct task_struct *tsk; | |
615d6e87 | 1107 | struct mm_struct *old_mm, *active_mm; |
eea96732 | 1108 | int ret; |
1da177e4 LT |
1109 | |
1110 | /* Notify parent that we're no longer interested in the old VM */ | |
1111 | tsk = current; | |
1112 | old_mm = current->mm; | |
4610ba7a | 1113 | exec_mm_release(tsk, old_mm); |
a28bf136 EB |
1114 | if (old_mm) |
1115 | sync_mm_rss(old_mm); | |
1da177e4 | 1116 | |
eea96732 EB |
1117 | ret = mutex_lock_killable(&tsk->signal->exec_update_mutex); |
1118 | if (ret) | |
1119 | return ret; | |
1120 | ||
1da177e4 LT |
1121 | if (old_mm) { |
1122 | /* | |
1123 | * Make sure that if there is a core dump in progress | |
1124 | * for the old mm, we get out and die instead of going | |
c1e8d7c6 | 1125 | * through with the exec. We must hold mmap_lock around |
999d9fc1 | 1126 | * checking core_state and changing tsk->mm. |
1da177e4 | 1127 | */ |
d8ed45c5 | 1128 | mmap_read_lock(old_mm); |
999d9fc1 | 1129 | if (unlikely(old_mm->core_state)) { |
d8ed45c5 | 1130 | mmap_read_unlock(old_mm); |
eea96732 | 1131 | mutex_unlock(&tsk->signal->exec_update_mutex); |
1da177e4 LT |
1132 | return -EINTR; |
1133 | } | |
1134 | } | |
eea96732 | 1135 | |
1da177e4 LT |
1136 | task_lock(tsk); |
1137 | active_mm = tsk->active_mm; | |
227a4aad | 1138 | membarrier_exec_mmap(mm); |
1da177e4 LT |
1139 | tsk->mm = mm; |
1140 | tsk->active_mm = mm; | |
1141 | activate_mm(active_mm, mm); | |
615d6e87 DB |
1142 | tsk->mm->vmacache_seqnum = 0; |
1143 | vmacache_flush(tsk); | |
1da177e4 | 1144 | task_unlock(tsk); |
1da177e4 | 1145 | if (old_mm) { |
d8ed45c5 | 1146 | mmap_read_unlock(old_mm); |
7dddb12c | 1147 | BUG_ON(active_mm != old_mm); |
701085b2 | 1148 | setmax_mm_hiwater_rss(&tsk->signal->maxrss, old_mm); |
31a78f23 | 1149 | mm_update_next_owner(old_mm); |
1da177e4 LT |
1150 | mmput(old_mm); |
1151 | return 0; | |
1152 | } | |
1153 | mmdrop(active_mm); | |
1154 | return 0; | |
1155 | } | |
1156 | ||
858119e1 | 1157 | static int de_thread(struct task_struct *tsk) |
1da177e4 LT |
1158 | { |
1159 | struct signal_struct *sig = tsk->signal; | |
b2c903b8 | 1160 | struct sighand_struct *oldsighand = tsk->sighand; |
1da177e4 | 1161 | spinlock_t *lock = &oldsighand->siglock; |
1da177e4 | 1162 | |
aafe6c2a | 1163 | if (thread_group_empty(tsk)) |
1da177e4 LT |
1164 | goto no_thread_group; |
1165 | ||
1166 | /* | |
1167 | * Kill all other threads in the thread group. | |
1da177e4 | 1168 | */ |
1da177e4 | 1169 | spin_lock_irq(lock); |
ed5d2cac | 1170 | if (signal_group_exit(sig)) { |
1da177e4 LT |
1171 | /* |
1172 | * Another group action in progress, just | |
1173 | * return so that the signal is processed. | |
1174 | */ | |
1175 | spin_unlock_irq(lock); | |
1da177e4 LT |
1176 | return -EAGAIN; |
1177 | } | |
d344193a | 1178 | |
ed5d2cac | 1179 | sig->group_exit_task = tsk; |
d344193a ON |
1180 | sig->notify_count = zap_other_threads(tsk); |
1181 | if (!thread_group_leader(tsk)) | |
1182 | sig->notify_count--; | |
1da177e4 | 1183 | |
d344193a | 1184 | while (sig->notify_count) { |
d5bbd43d | 1185 | __set_current_state(TASK_KILLABLE); |
1da177e4 | 1186 | spin_unlock_irq(lock); |
a72173ec | 1187 | schedule(); |
08d405c8 | 1188 | if (__fatal_signal_pending(tsk)) |
d5bbd43d | 1189 | goto killed; |
1da177e4 LT |
1190 | spin_lock_irq(lock); |
1191 | } | |
1da177e4 LT |
1192 | spin_unlock_irq(lock); |
1193 | ||
1194 | /* | |
1195 | * At this point all other threads have exited, all we have to | |
1196 | * do is to wait for the thread group leader to become inactive, | |
1197 | * and to assume its PID: | |
1198 | */ | |
aafe6c2a | 1199 | if (!thread_group_leader(tsk)) { |
8187926b | 1200 | struct task_struct *leader = tsk->group_leader; |
6db840fa | 1201 | |
6db840fa | 1202 | for (;;) { |
780de9dd | 1203 | cgroup_threadgroup_change_begin(tsk); |
6db840fa | 1204 | write_lock_irq(&tasklist_lock); |
dfcce791 KT |
1205 | /* |
1206 | * Do this under tasklist_lock to ensure that | |
1207 | * exit_notify() can't miss ->group_exit_task | |
1208 | */ | |
1209 | sig->notify_count = -1; | |
6db840fa ON |
1210 | if (likely(leader->exit_state)) |
1211 | break; | |
d5bbd43d | 1212 | __set_current_state(TASK_KILLABLE); |
6db840fa | 1213 | write_unlock_irq(&tasklist_lock); |
780de9dd | 1214 | cgroup_threadgroup_change_end(tsk); |
a72173ec | 1215 | schedule(); |
08d405c8 | 1216 | if (__fatal_signal_pending(tsk)) |
d5bbd43d | 1217 | goto killed; |
6db840fa | 1218 | } |
1da177e4 | 1219 | |
f5e90281 RM |
1220 | /* |
1221 | * The only record we have of the real-time age of a | |
1222 | * process, regardless of execs it's done, is start_time. | |
1223 | * All the past CPU time is accumulated in signal_struct | |
1224 | * from sister threads now dead. But in this non-leader | |
1225 | * exec, nothing survives from the original leader thread, | |
1226 | * whose birth marks the true age of this process now. | |
1227 | * When we take on its identity by switching to its PID, we | |
1228 | * also take its birthdate (always earlier than our own). | |
1229 | */ | |
aafe6c2a | 1230 | tsk->start_time = leader->start_time; |
cf25e24d | 1231 | tsk->start_boottime = leader->start_boottime; |
f5e90281 | 1232 | |
bac0abd6 | 1233 | BUG_ON(!same_thread_group(leader, tsk)); |
1da177e4 LT |
1234 | /* |
1235 | * An exec() starts a new thread group with the | |
1236 | * TGID of the previous thread group. Rehash the | |
1237 | * two threads with a switched PID, and release | |
1238 | * the former thread group leader: | |
1239 | */ | |
d73d6529 EB |
1240 | |
1241 | /* Become a process group leader with the old leader's pid. | |
c18258c6 | 1242 | * The old leader becomes a thread of the this thread group. |
d73d6529 | 1243 | */ |
6b03d130 | 1244 | exchange_tids(tsk, leader); |
6883f81a | 1245 | transfer_pid(leader, tsk, PIDTYPE_TGID); |
aafe6c2a EB |
1246 | transfer_pid(leader, tsk, PIDTYPE_PGID); |
1247 | transfer_pid(leader, tsk, PIDTYPE_SID); | |
9cd80bbb | 1248 | |
aafe6c2a | 1249 | list_replace_rcu(&leader->tasks, &tsk->tasks); |
9cd80bbb | 1250 | list_replace_init(&leader->sibling, &tsk->sibling); |
1da177e4 | 1251 | |
aafe6c2a EB |
1252 | tsk->group_leader = tsk; |
1253 | leader->group_leader = tsk; | |
de12a787 | 1254 | |
aafe6c2a | 1255 | tsk->exit_signal = SIGCHLD; |
087806b1 | 1256 | leader->exit_signal = -1; |
962b564c ON |
1257 | |
1258 | BUG_ON(leader->exit_state != EXIT_ZOMBIE); | |
1259 | leader->exit_state = EXIT_DEAD; | |
eac1b5e5 ON |
1260 | |
1261 | /* | |
1262 | * We are going to release_task()->ptrace_unlink() silently, | |
1263 | * the tracer can sleep in do_wait(). EXIT_DEAD guarantees | |
1264 | * the tracer wont't block again waiting for this thread. | |
1265 | */ | |
1266 | if (unlikely(leader->ptrace)) | |
1267 | __wake_up_parent(leader, leader->parent); | |
1da177e4 | 1268 | write_unlock_irq(&tasklist_lock); |
780de9dd | 1269 | cgroup_threadgroup_change_end(tsk); |
8187926b ON |
1270 | |
1271 | release_task(leader); | |
ed5d2cac | 1272 | } |
1da177e4 | 1273 | |
6db840fa ON |
1274 | sig->group_exit_task = NULL; |
1275 | sig->notify_count = 0; | |
1da177e4 LT |
1276 | |
1277 | no_thread_group: | |
e6368253 ON |
1278 | /* we have changed execution domain */ |
1279 | tsk->exit_signal = SIGCHLD; | |
1280 | ||
02169155 EB |
1281 | BUG_ON(!thread_group_leader(tsk)); |
1282 | return 0; | |
1283 | ||
1284 | killed: | |
1285 | /* protects against exit_notify() and __exit_signal() */ | |
1286 | read_lock(&tasklist_lock); | |
1287 | sig->group_exit_task = NULL; | |
1288 | sig->notify_count = 0; | |
1289 | read_unlock(&tasklist_lock); | |
1290 | return -EAGAIN; | |
1291 | } | |
1292 | ||
1293 | ||
7a60ef48 EB |
1294 | /* |
1295 | * This function makes sure the current process has its own signal table, | |
1296 | * so that flush_signal_handlers can later reset the handlers without | |
1297 | * disturbing other processes. (Other processes might share the signal | |
1298 | * table via the CLONE_SIGHAND option to clone().) | |
1299 | */ | |
02169155 EB |
1300 | static int unshare_sighand(struct task_struct *me) |
1301 | { | |
1302 | struct sighand_struct *oldsighand = me->sighand; | |
329f7dba | 1303 | |
d036bda7 | 1304 | if (refcount_read(&oldsighand->count) != 1) { |
b2c903b8 | 1305 | struct sighand_struct *newsighand; |
1da177e4 | 1306 | /* |
b2c903b8 ON |
1307 | * This ->sighand is shared with the CLONE_SIGHAND |
1308 | * but not CLONE_THREAD task, switch to the new one. | |
1da177e4 | 1309 | */ |
b2c903b8 ON |
1310 | newsighand = kmem_cache_alloc(sighand_cachep, GFP_KERNEL); |
1311 | if (!newsighand) | |
1312 | return -ENOMEM; | |
1313 | ||
d036bda7 | 1314 | refcount_set(&newsighand->count, 1); |
1da177e4 LT |
1315 | memcpy(newsighand->action, oldsighand->action, |
1316 | sizeof(newsighand->action)); | |
1317 | ||
1318 | write_lock_irq(&tasklist_lock); | |
1319 | spin_lock(&oldsighand->siglock); | |
02169155 | 1320 | rcu_assign_pointer(me->sighand, newsighand); |
1da177e4 LT |
1321 | spin_unlock(&oldsighand->siglock); |
1322 | write_unlock_irq(&tasklist_lock); | |
1323 | ||
fba2afaa | 1324 | __cleanup_sighand(oldsighand); |
1da177e4 | 1325 | } |
1da177e4 LT |
1326 | return 0; |
1327 | } | |
0840a90d | 1328 | |
3756f640 | 1329 | char *__get_task_comm(char *buf, size_t buf_size, struct task_struct *tsk) |
1da177e4 | 1330 | { |
1da177e4 | 1331 | task_lock(tsk); |
3756f640 | 1332 | strncpy(buf, tsk->comm, buf_size); |
1da177e4 | 1333 | task_unlock(tsk); |
59714d65 | 1334 | return buf; |
1da177e4 | 1335 | } |
3756f640 | 1336 | EXPORT_SYMBOL_GPL(__get_task_comm); |
1da177e4 | 1337 | |
6a6d27de AV |
1338 | /* |
1339 | * These functions flushes out all traces of the currently running executable | |
1340 | * so that a new one can be started | |
1341 | */ | |
1342 | ||
82b89778 | 1343 | void __set_task_comm(struct task_struct *tsk, const char *buf, bool exec) |
1da177e4 LT |
1344 | { |
1345 | task_lock(tsk); | |
43d2b113 | 1346 | trace_task_rename(tsk, buf); |
1da177e4 LT |
1347 | strlcpy(tsk->comm, buf, sizeof(tsk->comm)); |
1348 | task_unlock(tsk); | |
82b89778 | 1349 | perf_event_comm(tsk, exec); |
1da177e4 LT |
1350 | } |
1351 | ||
a9208e42 KC |
1352 | /* |
1353 | * Calling this is the point of no return. None of the failures will be | |
1354 | * seen by userspace since either the process is already taking a fatal | |
1355 | * signal (via de_thread() or coredump), or will have SEGV raised | |
13c432b5 | 1356 | * (after exec_mmap()) by search_binary_handler (see below). |
a9208e42 | 1357 | */ |
2388777a | 1358 | int begin_new_exec(struct linux_binprm * bprm) |
1da177e4 | 1359 | { |
2ca7be7d | 1360 | struct task_struct *me = current; |
221af7f8 | 1361 | int retval; |
1da177e4 | 1362 | |
56305aa9 EB |
1363 | /* Once we are committed compute the creds */ |
1364 | retval = bprm_creds_from_file(bprm); | |
1365 | if (retval) | |
1366 | return retval; | |
1367 | ||
6834e0bb EB |
1368 | /* |
1369 | * Ensure all future errors are fatal. | |
1370 | */ | |
1371 | bprm->point_of_no_return = true; | |
1372 | ||
1da177e4 | 1373 | /* |
02169155 | 1374 | * Make this the only thread in the thread group. |
1da177e4 | 1375 | */ |
2ca7be7d | 1376 | retval = de_thread(me); |
1da177e4 LT |
1377 | if (retval) |
1378 | goto out; | |
1379 | ||
6e399cd1 DB |
1380 | /* |
1381 | * Must be called _before_ exec_mmap() as bprm->mm is | |
1382 | * not visibile until then. This also enables the update | |
1383 | * to be lockless. | |
1384 | */ | |
925d1c40 | 1385 | set_mm_exe_file(bprm->mm, bprm->file); |
6e399cd1 | 1386 | |
b8a61c9e | 1387 | /* If the binary is not readable then enforce mm->dumpable=0 */ |
f87d1c95 | 1388 | would_dump(bprm, bprm->file); |
b8a61c9e EB |
1389 | if (bprm->have_execfd) |
1390 | would_dump(bprm, bprm->executable); | |
f87d1c95 | 1391 | |
1da177e4 LT |
1392 | /* |
1393 | * Release all of the old mmap stuff | |
1394 | */ | |
3c77f845 | 1395 | acct_arg_size(bprm, 0); |
1da177e4 LT |
1396 | retval = exec_mmap(bprm->mm); |
1397 | if (retval) | |
fd8328be | 1398 | goto out; |
1da177e4 | 1399 | |
a9208e42 | 1400 | bprm->mm = NULL; |
7ab02af4 | 1401 | |
ccf0fa6b EB |
1402 | #ifdef CONFIG_POSIX_TIMERS |
1403 | exit_itimers(me->signal); | |
1404 | flush_itimer_signals(); | |
1405 | #endif | |
1406 | ||
1407 | /* | |
1408 | * Make the signal table private. | |
1409 | */ | |
1410 | retval = unshare_sighand(me); | |
1411 | if (retval) | |
89826cce | 1412 | goto out_unlock; |
ccf0fa6b | 1413 | |
fe814175 CH |
1414 | /* |
1415 | * Ensure that the uaccess routines can actually operate on userspace | |
1416 | * pointers: | |
1417 | */ | |
1418 | force_uaccess_begin(); | |
1419 | ||
2ca7be7d | 1420 | me->flags &= ~(PF_RANDOMIZE | PF_FORKNOEXEC | PF_KTHREAD | |
b88fae64 | 1421 | PF_NOFREEZE | PF_NO_SETAFFINITY); |
7ab02af4 | 1422 | flush_thread(); |
2ca7be7d | 1423 | me->personality &= ~bprm->per_clear; |
7ab02af4 | 1424 | |
613cc2b6 AS |
1425 | /* |
1426 | * We have to apply CLOEXEC before we change whether the process is | |
1427 | * dumpable (in setup_new_exec) to avoid a race with a process in userspace | |
1428 | * trying to access the should-be-closed file descriptors of a process | |
1429 | * undergoing exec(2). | |
1430 | */ | |
2ca7be7d | 1431 | do_close_on_exec(me->files); |
f84df2a6 | 1432 | |
64701dee | 1433 | if (bprm->secureexec) { |
fe8993b3 | 1434 | /* Make sure parent cannot signal privileged process. */ |
7d503feb | 1435 | me->pdeath_signal = 0; |
fe8993b3 | 1436 | |
64701dee KC |
1437 | /* |
1438 | * For secureexec, reset the stack limit to sane default to | |
1439 | * avoid bad behavior from the prior rlimits. This has to | |
1440 | * happen before arch_pick_mmap_layout(), which examines | |
1441 | * RLIMIT_STACK, but after the point of no return to avoid | |
779f4e1c | 1442 | * needing to clean up the change on failure. |
64701dee | 1443 | */ |
c31dbb14 KC |
1444 | if (bprm->rlim_stack.rlim_cur > _STK_LIM) |
1445 | bprm->rlim_stack.rlim_cur = _STK_LIM; | |
64701dee KC |
1446 | } |
1447 | ||
7d503feb | 1448 | me->sas_ss_sp = me->sas_ss_size = 0; |
1da177e4 | 1449 | |
e816c201 KC |
1450 | /* |
1451 | * Figure out dumpability. Note that this checking only of current | |
1452 | * is wrong, but userspace depends on it. This should be testing | |
1453 | * bprm->secureexec instead. | |
1454 | */ | |
473d8963 | 1455 | if (bprm->interp_flags & BINPRM_FLAGS_ENFORCE_NONDUMP || |
e816c201 KC |
1456 | !(uid_eq(current_euid(), current_uid()) && |
1457 | gid_eq(current_egid(), current_gid()))) | |
6c5d5238 | 1458 | set_dumpable(current->mm, suid_dumpable); |
473d8963 KC |
1459 | else |
1460 | set_dumpable(current->mm, SUID_DUMP_USER); | |
d6e71144 | 1461 | |
e041e328 | 1462 | perf_event_exec(); |
7d503feb | 1463 | __set_task_comm(me, kbasename(bprm->filename), true); |
1da177e4 | 1464 | |
1da177e4 LT |
1465 | /* An exec changes our domain. We are no longer part of the thread |
1466 | group */ | |
7d503feb EB |
1467 | WRITE_ONCE(me->self_exec_id, me->self_exec_id + 1); |
1468 | flush_signal_handlers(me, 0); | |
96ecee29 EB |
1469 | |
1470 | /* | |
1471 | * install the new credentials for this executable | |
1472 | */ | |
1473 | security_bprm_committing_creds(bprm); | |
1474 | ||
1475 | commit_creds(bprm->cred); | |
1476 | bprm->cred = NULL; | |
1477 | ||
1478 | /* | |
1479 | * Disable monitoring for regular users | |
1480 | * when executing setuid binaries. Must | |
1481 | * wait until new credentials are committed | |
1482 | * by commit_creds() above | |
1483 | */ | |
7d503feb EB |
1484 | if (get_dumpable(me->mm) != SUID_DUMP_USER) |
1485 | perf_event_exit_task(me); | |
96ecee29 EB |
1486 | /* |
1487 | * cred_guard_mutex must be held at least to this point to prevent | |
1488 | * ptrace_attach() from altering our determination of the task's | |
1489 | * credentials; any time after this it may be unlocked. | |
1490 | */ | |
1491 | security_bprm_committed_creds(bprm); | |
b8a61c9e EB |
1492 | |
1493 | /* Pass the opened binary to the interpreter. */ | |
1494 | if (bprm->have_execfd) { | |
1495 | retval = get_unused_fd_flags(0); | |
1496 | if (retval < 0) | |
1497 | goto out_unlock; | |
1498 | fd_install(retval, bprm->executable); | |
1499 | bprm->executable = NULL; | |
1500 | bprm->execfd = retval; | |
1501 | } | |
221af7f8 LT |
1502 | return 0; |
1503 | ||
df9e4d2c EB |
1504 | out_unlock: |
1505 | mutex_unlock(&me->signal->exec_update_mutex); | |
221af7f8 LT |
1506 | out: |
1507 | return retval; | |
1508 | } | |
2388777a | 1509 | EXPORT_SYMBOL(begin_new_exec); |
221af7f8 | 1510 | |
1b5d783c AV |
1511 | void would_dump(struct linux_binprm *bprm, struct file *file) |
1512 | { | |
f84df2a6 EB |
1513 | struct inode *inode = file_inode(file); |
1514 | if (inode_permission(inode, MAY_READ) < 0) { | |
1515 | struct user_namespace *old, *user_ns; | |
1b5d783c | 1516 | bprm->interp_flags |= BINPRM_FLAGS_ENFORCE_NONDUMP; |
f84df2a6 EB |
1517 | |
1518 | /* Ensure mm->user_ns contains the executable */ | |
1519 | user_ns = old = bprm->mm->user_ns; | |
1520 | while ((user_ns != &init_user_ns) && | |
1521 | !privileged_wrt_inode_uidgid(user_ns, inode)) | |
1522 | user_ns = user_ns->parent; | |
1523 | ||
1524 | if (old != user_ns) { | |
1525 | bprm->mm->user_ns = get_user_ns(user_ns); | |
1526 | put_user_ns(old); | |
1527 | } | |
1528 | } | |
1b5d783c AV |
1529 | } |
1530 | EXPORT_SYMBOL(would_dump); | |
1531 | ||
221af7f8 LT |
1532 | void setup_new_exec(struct linux_binprm * bprm) |
1533 | { | |
df9e4d2c EB |
1534 | /* Setup things that can depend upon the personality */ |
1535 | struct task_struct *me = current; | |
1da177e4 | 1536 | |
df9e4d2c | 1537 | arch_pick_mmap_layout(me->mm, &bprm->rlim_stack); |
d6e71144 | 1538 | |
e9ea1e7f | 1539 | arch_setup_new_exec(); |
1da177e4 | 1540 | |
0551fbd2 BH |
1541 | /* Set the new mm task size. We have to do that late because it may |
1542 | * depend on TIF_32BIT which is only updated in flush_thread() on | |
1543 | * some architectures like powerpc | |
1544 | */ | |
df9e4d2c | 1545 | me->mm->task_size = TASK_SIZE; |
7d503feb EB |
1546 | mutex_unlock(&me->signal->exec_update_mutex); |
1547 | mutex_unlock(&me->signal->cred_guard_mutex); | |
1da177e4 | 1548 | } |
221af7f8 | 1549 | EXPORT_SYMBOL(setup_new_exec); |
1da177e4 | 1550 | |
b8383831 KC |
1551 | /* Runs immediately before start_thread() takes over. */ |
1552 | void finalize_exec(struct linux_binprm *bprm) | |
1553 | { | |
c31dbb14 KC |
1554 | /* Store any stack rlimit changes before starting thread. */ |
1555 | task_lock(current->group_leader); | |
1556 | current->signal->rlim[RLIMIT_STACK] = bprm->rlim_stack; | |
1557 | task_unlock(current->group_leader); | |
b8383831 KC |
1558 | } |
1559 | EXPORT_SYMBOL(finalize_exec); | |
1560 | ||
a2a8474c ON |
1561 | /* |
1562 | * Prepare credentials and lock ->cred_guard_mutex. | |
96ecee29 | 1563 | * setup_new_exec() commits the new creds and drops the lock. |
a2a8474c ON |
1564 | * Or, if exec fails before, free_bprm() should release ->cred and |
1565 | * and unlock. | |
1566 | */ | |
4addd264 | 1567 | static int prepare_bprm_creds(struct linux_binprm *bprm) |
a2a8474c | 1568 | { |
9b1bf12d | 1569 | if (mutex_lock_interruptible(¤t->signal->cred_guard_mutex)) |
a2a8474c ON |
1570 | return -ERESTARTNOINTR; |
1571 | ||
1572 | bprm->cred = prepare_exec_creds(); | |
1573 | if (likely(bprm->cred)) | |
1574 | return 0; | |
1575 | ||
9b1bf12d | 1576 | mutex_unlock(¤t->signal->cred_guard_mutex); |
a2a8474c ON |
1577 | return -ENOMEM; |
1578 | } | |
1579 | ||
c4ad8f98 | 1580 | static void free_bprm(struct linux_binprm *bprm) |
a2a8474c | 1581 | { |
f18ac551 EB |
1582 | if (bprm->mm) { |
1583 | acct_arg_size(bprm, 0); | |
1584 | mmput(bprm->mm); | |
1585 | } | |
a2a8474c ON |
1586 | free_arg_pages(bprm); |
1587 | if (bprm->cred) { | |
9b1bf12d | 1588 | mutex_unlock(¤t->signal->cred_guard_mutex); |
a2a8474c ON |
1589 | abort_creds(bprm->cred); |
1590 | } | |
63e46b95 ON |
1591 | if (bprm->file) { |
1592 | allow_write_access(bprm->file); | |
1593 | fput(bprm->file); | |
1594 | } | |
b8a61c9e EB |
1595 | if (bprm->executable) |
1596 | fput(bprm->executable); | |
b66c5984 KC |
1597 | /* If a binfmt changed the interp, free it. */ |
1598 | if (bprm->interp != bprm->filename) | |
1599 | kfree(bprm->interp); | |
60d9ad1d | 1600 | kfree(bprm->fdpath); |
a2a8474c ON |
1601 | kfree(bprm); |
1602 | } | |
1603 | ||
60d9ad1d | 1604 | static struct linux_binprm *alloc_bprm(int fd, struct filename *filename) |
0a8f36eb EB |
1605 | { |
1606 | struct linux_binprm *bprm = kzalloc(sizeof(*bprm), GFP_KERNEL); | |
60d9ad1d | 1607 | int retval = -ENOMEM; |
0a8f36eb | 1608 | if (!bprm) |
60d9ad1d EB |
1609 | goto out; |
1610 | ||
1611 | if (fd == AT_FDCWD || filename->name[0] == '/') { | |
1612 | bprm->filename = filename->name; | |
1613 | } else { | |
1614 | if (filename->name[0] == '\0') | |
1615 | bprm->fdpath = kasprintf(GFP_KERNEL, "/dev/fd/%d", fd); | |
1616 | else | |
1617 | bprm->fdpath = kasprintf(GFP_KERNEL, "/dev/fd/%d/%s", | |
1618 | fd, filename->name); | |
1619 | if (!bprm->fdpath) | |
1620 | goto out_free; | |
1621 | ||
1622 | bprm->filename = bprm->fdpath; | |
1623 | } | |
1624 | bprm->interp = bprm->filename; | |
f18ac551 EB |
1625 | |
1626 | retval = bprm_mm_init(bprm); | |
1627 | if (retval) | |
1628 | goto out_free; | |
0a8f36eb | 1629 | return bprm; |
60d9ad1d EB |
1630 | |
1631 | out_free: | |
1632 | free_bprm(bprm); | |
1633 | out: | |
1634 | return ERR_PTR(retval); | |
0a8f36eb EB |
1635 | } |
1636 | ||
c2315c18 | 1637 | int bprm_change_interp(const char *interp, struct linux_binprm *bprm) |
b66c5984 KC |
1638 | { |
1639 | /* If a binfmt changed the interp, free it first. */ | |
1640 | if (bprm->interp != bprm->filename) | |
1641 | kfree(bprm->interp); | |
1642 | bprm->interp = kstrdup(interp, GFP_KERNEL); | |
1643 | if (!bprm->interp) | |
1644 | return -ENOMEM; | |
1645 | return 0; | |
1646 | } | |
1647 | EXPORT_SYMBOL(bprm_change_interp); | |
1648 | ||
a6f76f23 DH |
1649 | /* |
1650 | * determine how safe it is to execute the proposed program | |
9b1bf12d | 1651 | * - the caller must hold ->cred_guard_mutex to protect against |
c2e1f2e3 | 1652 | * PTRACE_ATTACH or seccomp thread-sync |
a6f76f23 | 1653 | */ |
9e00cdb0 | 1654 | static void check_unsafe_exec(struct linux_binprm *bprm) |
a6f76f23 | 1655 | { |
0bf2f3ae | 1656 | struct task_struct *p = current, *t; |
f1191b50 | 1657 | unsigned n_fs; |
a6f76f23 | 1658 | |
9227dd2a EB |
1659 | if (p->ptrace) |
1660 | bprm->unsafe |= LSM_UNSAFE_PTRACE; | |
a6f76f23 | 1661 | |
259e5e6c AL |
1662 | /* |
1663 | * This isn't strictly necessary, but it makes it harder for LSMs to | |
1664 | * mess up. | |
1665 | */ | |
1d4457f9 | 1666 | if (task_no_new_privs(current)) |
259e5e6c AL |
1667 | bprm->unsafe |= LSM_UNSAFE_NO_NEW_PRIVS; |
1668 | ||
83f62a2e | 1669 | t = p; |
0bf2f3ae | 1670 | n_fs = 1; |
2a4419b5 | 1671 | spin_lock(&p->fs->lock); |
437f7fdb | 1672 | rcu_read_lock(); |
83f62a2e | 1673 | while_each_thread(p, t) { |
0bf2f3ae DH |
1674 | if (t->fs == p->fs) |
1675 | n_fs++; | |
0bf2f3ae | 1676 | } |
437f7fdb | 1677 | rcu_read_unlock(); |
0bf2f3ae | 1678 | |
9e00cdb0 | 1679 | if (p->fs->users > n_fs) |
a6f76f23 | 1680 | bprm->unsafe |= LSM_UNSAFE_SHARE; |
9e00cdb0 ON |
1681 | else |
1682 | p->fs->in_exec = 1; | |
2a4419b5 | 1683 | spin_unlock(&p->fs->lock); |
a6f76f23 DH |
1684 | } |
1685 | ||
56305aa9 | 1686 | static void bprm_fill_uid(struct linux_binprm *bprm, struct file *file) |
8b01fc86 | 1687 | { |
56305aa9 | 1688 | /* Handle suid and sgid on files */ |
8b01fc86 JH |
1689 | struct inode *inode; |
1690 | unsigned int mode; | |
1691 | kuid_t uid; | |
1692 | kgid_t gid; | |
1693 | ||
56305aa9 | 1694 | if (!mnt_may_suid(file->f_path.mnt)) |
8b01fc86 JH |
1695 | return; |
1696 | ||
1697 | if (task_no_new_privs(current)) | |
1698 | return; | |
1699 | ||
56305aa9 | 1700 | inode = file->f_path.dentry->d_inode; |
8b01fc86 JH |
1701 | mode = READ_ONCE(inode->i_mode); |
1702 | if (!(mode & (S_ISUID|S_ISGID))) | |
1703 | return; | |
1704 | ||
1705 | /* Be careful if suid/sgid is set */ | |
5955102c | 1706 | inode_lock(inode); |
8b01fc86 JH |
1707 | |
1708 | /* reload atomically mode/uid/gid now that lock held */ | |
1709 | mode = inode->i_mode; | |
1710 | uid = inode->i_uid; | |
1711 | gid = inode->i_gid; | |
5955102c | 1712 | inode_unlock(inode); |
8b01fc86 JH |
1713 | |
1714 | /* We ignore suid/sgid if there are no mappings for them in the ns */ | |
1715 | if (!kuid_has_mapping(bprm->cred->user_ns, uid) || | |
1716 | !kgid_has_mapping(bprm->cred->user_ns, gid)) | |
1717 | return; | |
1718 | ||
1719 | if (mode & S_ISUID) { | |
1720 | bprm->per_clear |= PER_CLEAR_ON_SETID; | |
1721 | bprm->cred->euid = uid; | |
1722 | } | |
1723 | ||
1724 | if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) { | |
1725 | bprm->per_clear |= PER_CLEAR_ON_SETID; | |
1726 | bprm->cred->egid = gid; | |
1727 | } | |
1728 | } | |
1729 | ||
56305aa9 EB |
1730 | /* |
1731 | * Compute brpm->cred based upon the final binary. | |
1732 | */ | |
1733 | static int bprm_creds_from_file(struct linux_binprm *bprm) | |
1734 | { | |
1735 | /* Compute creds based on which file? */ | |
1736 | struct file *file = bprm->execfd_creds ? bprm->executable : bprm->file; | |
1737 | ||
1738 | bprm_fill_uid(bprm, file); | |
1739 | return security_bprm_creds_from_file(bprm, file); | |
1740 | } | |
1741 | ||
9e00cdb0 ON |
1742 | /* |
1743 | * Fill the binprm structure from the inode. | |
56305aa9 | 1744 | * Read the first BINPRM_BUF_SIZE bytes |
a6f76f23 DH |
1745 | * |
1746 | * This may be called multiple times for binary chains (scripts for example). | |
1da177e4 | 1747 | */ |
8b72ca90 | 1748 | static int prepare_binprm(struct linux_binprm *bprm) |
1da177e4 | 1749 | { |
bdd1d2d3 | 1750 | loff_t pos = 0; |
1da177e4 | 1751 | |
a6f76f23 | 1752 | memset(bprm->buf, 0, BINPRM_BUF_SIZE); |
bdd1d2d3 | 1753 | return kernel_read(bprm->file, bprm->buf, BINPRM_BUF_SIZE, &pos); |
1da177e4 LT |
1754 | } |
1755 | ||
4fc75ff4 NP |
1756 | /* |
1757 | * Arguments are '\0' separated strings found at the location bprm->p | |
1758 | * points to; chop off the first by relocating brpm->p to right after | |
1759 | * the first '\0' encountered. | |
1760 | */ | |
b6a2fea3 | 1761 | int remove_arg_zero(struct linux_binprm *bprm) |
1da177e4 | 1762 | { |
b6a2fea3 OW |
1763 | int ret = 0; |
1764 | unsigned long offset; | |
1765 | char *kaddr; | |
1766 | struct page *page; | |
4fc75ff4 | 1767 | |
b6a2fea3 OW |
1768 | if (!bprm->argc) |
1769 | return 0; | |
1da177e4 | 1770 | |
b6a2fea3 OW |
1771 | do { |
1772 | offset = bprm->p & ~PAGE_MASK; | |
1773 | page = get_arg_page(bprm, bprm->p, 0); | |
1774 | if (!page) { | |
1775 | ret = -EFAULT; | |
1776 | goto out; | |
1777 | } | |
e8e3c3d6 | 1778 | kaddr = kmap_atomic(page); |
4fc75ff4 | 1779 | |
b6a2fea3 OW |
1780 | for (; offset < PAGE_SIZE && kaddr[offset]; |
1781 | offset++, bprm->p++) | |
1782 | ; | |
4fc75ff4 | 1783 | |
e8e3c3d6 | 1784 | kunmap_atomic(kaddr); |
b6a2fea3 | 1785 | put_arg_page(page); |
b6a2fea3 | 1786 | } while (offset == PAGE_SIZE); |
4fc75ff4 | 1787 | |
b6a2fea3 OW |
1788 | bprm->p++; |
1789 | bprm->argc--; | |
1790 | ret = 0; | |
4fc75ff4 | 1791 | |
b6a2fea3 OW |
1792 | out: |
1793 | return ret; | |
1da177e4 | 1794 | } |
1da177e4 LT |
1795 | EXPORT_SYMBOL(remove_arg_zero); |
1796 | ||
cb7b6b1c | 1797 | #define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e)) |
1da177e4 LT |
1798 | /* |
1799 | * cycle the list of binary formats handler, until one recognizes the image | |
1800 | */ | |
bc2bf338 | 1801 | static int search_binary_handler(struct linux_binprm *bprm) |
1da177e4 | 1802 | { |
cb7b6b1c | 1803 | bool need_retry = IS_ENABLED(CONFIG_MODULES); |
1da177e4 | 1804 | struct linux_binfmt *fmt; |
cb7b6b1c | 1805 | int retval; |
1da177e4 | 1806 | |
8b72ca90 EB |
1807 | retval = prepare_binprm(bprm); |
1808 | if (retval < 0) | |
1809 | return retval; | |
d7402698 | 1810 | |
1da177e4 LT |
1811 | retval = security_bprm_check(bprm); |
1812 | if (retval) | |
1813 | return retval; | |
1814 | ||
1da177e4 | 1815 | retval = -ENOENT; |
cb7b6b1c ON |
1816 | retry: |
1817 | read_lock(&binfmt_lock); | |
1818 | list_for_each_entry(fmt, &formats, lh) { | |
1819 | if (!try_module_get(fmt->module)) | |
1820 | continue; | |
1821 | read_unlock(&binfmt_lock); | |
d53ddd01 | 1822 | |
cb7b6b1c | 1823 | retval = fmt->load_binary(bprm); |
d53ddd01 | 1824 | |
19d860a1 AV |
1825 | read_lock(&binfmt_lock); |
1826 | put_binfmt(fmt); | |
bc2bf338 | 1827 | if (bprm->point_of_no_return || (retval != -ENOEXEC)) { |
19d860a1 | 1828 | read_unlock(&binfmt_lock); |
cb7b6b1c | 1829 | return retval; |
1da177e4 | 1830 | } |
1da177e4 | 1831 | } |
cb7b6b1c ON |
1832 | read_unlock(&binfmt_lock); |
1833 | ||
19d860a1 | 1834 | if (need_retry) { |
cb7b6b1c ON |
1835 | if (printable(bprm->buf[0]) && printable(bprm->buf[1]) && |
1836 | printable(bprm->buf[2]) && printable(bprm->buf[3])) | |
1837 | return retval; | |
4e0621a0 ON |
1838 | if (request_module("binfmt-%04x", *(ushort *)(bprm->buf + 2)) < 0) |
1839 | return retval; | |
cb7b6b1c ON |
1840 | need_retry = false; |
1841 | goto retry; | |
1842 | } | |
1843 | ||
1da177e4 LT |
1844 | return retval; |
1845 | } | |
1da177e4 | 1846 | |
5d1baf3b ON |
1847 | static int exec_binprm(struct linux_binprm *bprm) |
1848 | { | |
1849 | pid_t old_pid, old_vpid; | |
bc2bf338 | 1850 | int ret, depth; |
5d1baf3b ON |
1851 | |
1852 | /* Need to fetch pid before load_binary changes it */ | |
1853 | old_pid = current->pid; | |
1854 | rcu_read_lock(); | |
1855 | old_vpid = task_pid_nr_ns(current, task_active_pid_ns(current->parent)); | |
1856 | rcu_read_unlock(); | |
1857 | ||
bc2bf338 EB |
1858 | /* This allows 4 levels of binfmt rewrites before failing hard. */ |
1859 | for (depth = 0;; depth++) { | |
1860 | struct file *exec; | |
1861 | if (depth > 5) | |
1862 | return -ELOOP; | |
1863 | ||
1864 | ret = search_binary_handler(bprm); | |
1865 | if (ret < 0) | |
1866 | return ret; | |
1867 | if (!bprm->interpreter) | |
1868 | break; | |
1869 | ||
1870 | exec = bprm->file; | |
1871 | bprm->file = bprm->interpreter; | |
1872 | bprm->interpreter = NULL; | |
1873 | ||
1874 | allow_write_access(exec); | |
1875 | if (unlikely(bprm->have_execfd)) { | |
1876 | if (bprm->executable) { | |
1877 | fput(exec); | |
1878 | return -ENOEXEC; | |
1879 | } | |
1880 | bprm->executable = exec; | |
1881 | } else | |
1882 | fput(exec); | |
5d1baf3b ON |
1883 | } |
1884 | ||
bc2bf338 EB |
1885 | audit_bprm(bprm); |
1886 | trace_sched_process_exec(current, old_pid, bprm); | |
1887 | ptrace_event(PTRACE_EVENT_EXEC, old_vpid); | |
1888 | proc_exec_connector(current); | |
1889 | return 0; | |
5d1baf3b ON |
1890 | } |
1891 | ||
1da177e4 LT |
1892 | /* |
1893 | * sys_execve() executes a new program. | |
1894 | */ | |
0c9cdff0 EB |
1895 | static int bprm_execve(struct linux_binprm *bprm, |
1896 | int fd, struct filename *filename, int flags) | |
1da177e4 | 1897 | { |
25cf336d | 1898 | struct file *file; |
3b125388 | 1899 | struct files_struct *displaced; |
1da177e4 | 1900 | int retval; |
72fa5997 | 1901 | |
0a8f36eb EB |
1902 | retval = unshare_files(&displaced); |
1903 | if (retval) | |
0c9cdff0 | 1904 | return retval; |
1da177e4 | 1905 | |
a2a8474c ON |
1906 | retval = prepare_bprm_creds(bprm); |
1907 | if (retval) | |
0a8f36eb | 1908 | goto out_files; |
498052bb | 1909 | |
9e00cdb0 | 1910 | check_unsafe_exec(bprm); |
a2a8474c | 1911 | current->in_execve = 1; |
a6f76f23 | 1912 | |
25cf336d | 1913 | file = do_open_execat(fd, filename, flags); |
1da177e4 LT |
1914 | retval = PTR_ERR(file); |
1915 | if (IS_ERR(file)) | |
498052bb | 1916 | goto out_unmark; |
1da177e4 LT |
1917 | |
1918 | sched_exec(); | |
1919 | ||
1da177e4 | 1920 | bprm->file = file; |
60d9ad1d EB |
1921 | /* |
1922 | * Record that a name derived from an O_CLOEXEC fd will be | |
1923 | * inaccessible after exec. Relies on having exclusive access to | |
1924 | * current->files (due to unshare_files above). | |
1925 | */ | |
1926 | if (bprm->fdpath && | |
1927 | close_on_exec(fd, rcu_dereference_raw(current->files->fdt))) | |
1928 | bprm->interp_flags |= BINPRM_FLAGS_PATH_INACCESSIBLE; | |
1da177e4 | 1929 | |
b8bff599 EB |
1930 | /* Set the unchanging part of bprm->cred */ |
1931 | retval = security_bprm_creds_for_exec(bprm); | |
1932 | if (retval) | |
1da177e4 LT |
1933 | goto out; |
1934 | ||
5d1baf3b | 1935 | retval = exec_binprm(bprm); |
a6f76f23 DH |
1936 | if (retval < 0) |
1937 | goto out; | |
1da177e4 | 1938 | |
a6f76f23 | 1939 | /* execve succeeded */ |
498052bb | 1940 | current->fs->in_exec = 0; |
f9ce1f1c | 1941 | current->in_execve = 0; |
d7822b1e | 1942 | rseq_execve(current); |
a6f76f23 | 1943 | acct_update_integrals(current); |
16d51a59 | 1944 | task_numa_free(current, false); |
a6f76f23 DH |
1945 | if (displaced) |
1946 | put_files_struct(displaced); | |
1947 | return retval; | |
1da177e4 | 1948 | |
a6f76f23 | 1949 | out: |
8890b293 EB |
1950 | /* |
1951 | * If past the point of no return ensure the the code never | |
1952 | * returns to the userspace process. Use an existing fatal | |
1953 | * signal if present otherwise terminate the process with | |
1954 | * SIGSEGV. | |
1955 | */ | |
1956 | if (bprm->point_of_no_return && !fatal_signal_pending(current)) | |
1957 | force_sigsegv(SIGSEGV); | |
1da177e4 | 1958 | |
498052bb | 1959 | out_unmark: |
9e00cdb0 | 1960 | current->fs->in_exec = 0; |
f9ce1f1c | 1961 | current->in_execve = 0; |
a6f76f23 | 1962 | |
0a8f36eb EB |
1963 | out_files: |
1964 | if (displaced) | |
1965 | reset_files_struct(displaced); | |
0c9cdff0 EB |
1966 | |
1967 | return retval; | |
1968 | } | |
1969 | ||
1970 | static int do_execveat_common(int fd, struct filename *filename, | |
1971 | struct user_arg_ptr argv, | |
1972 | struct user_arg_ptr envp, | |
1973 | int flags) | |
1974 | { | |
1975 | struct linux_binprm *bprm; | |
1976 | int retval; | |
1977 | ||
1978 | if (IS_ERR(filename)) | |
1979 | return PTR_ERR(filename); | |
1980 | ||
1981 | /* | |
1982 | * We move the actual failure in case of RLIMIT_NPROC excess from | |
1983 | * set*uid() to execve() because too many poorly written programs | |
1984 | * don't check setuid() return code. Here we additionally recheck | |
1985 | * whether NPROC limit is still exceeded. | |
1986 | */ | |
1987 | if ((current->flags & PF_NPROC_EXCEEDED) && | |
1988 | atomic_read(¤t_user()->processes) > rlimit(RLIMIT_NPROC)) { | |
1989 | retval = -EAGAIN; | |
1990 | goto out_ret; | |
1991 | } | |
1992 | ||
1993 | /* We're below the limit (still or again), so we don't want to make | |
1994 | * further execve() calls fail. */ | |
1995 | current->flags &= ~PF_NPROC_EXCEEDED; | |
1996 | ||
1997 | bprm = alloc_bprm(fd, filename); | |
1998 | if (IS_ERR(bprm)) { | |
1999 | retval = PTR_ERR(bprm); | |
2000 | goto out_ret; | |
2001 | } | |
2002 | ||
d8b9cd54 EB |
2003 | retval = count(argv, MAX_ARG_STRINGS); |
2004 | if (retval < 0) | |
2005 | goto out_free; | |
2006 | bprm->argc = retval; | |
2007 | ||
2008 | retval = count(envp, MAX_ARG_STRINGS); | |
2009 | if (retval < 0) | |
2010 | goto out_free; | |
2011 | bprm->envc = retval; | |
2012 | ||
2013 | retval = bprm_stack_limits(bprm); | |
0c9cdff0 EB |
2014 | if (retval < 0) |
2015 | goto out_free; | |
2016 | ||
2017 | retval = copy_string_kernel(bprm->filename, bprm); | |
2018 | if (retval < 0) | |
2019 | goto out_free; | |
2020 | bprm->exec = bprm->p; | |
2021 | ||
2022 | retval = copy_strings(bprm->envc, envp, bprm); | |
2023 | if (retval < 0) | |
2024 | goto out_free; | |
2025 | ||
2026 | retval = copy_strings(bprm->argc, argv, bprm); | |
2027 | if (retval < 0) | |
2028 | goto out_free; | |
2029 | ||
2030 | retval = bprm_execve(bprm, fd, filename, flags); | |
a6f76f23 | 2031 | out_free: |
08a6fac1 | 2032 | free_bprm(bprm); |
1da177e4 LT |
2033 | |
2034 | out_ret: | |
25cf336d | 2035 | putname(filename); |
1da177e4 LT |
2036 | return retval; |
2037 | } | |
2038 | ||
be619f7f EB |
2039 | int kernel_execve(const char *kernel_filename, |
2040 | const char *const *argv, const char *const *envp) | |
2041 | { | |
2042 | struct filename *filename; | |
2043 | struct linux_binprm *bprm; | |
2044 | int fd = AT_FDCWD; | |
2045 | int retval; | |
2046 | ||
2047 | filename = getname_kernel(kernel_filename); | |
2048 | if (IS_ERR(filename)) | |
2049 | return PTR_ERR(filename); | |
2050 | ||
2051 | bprm = alloc_bprm(fd, filename); | |
2052 | if (IS_ERR(bprm)) { | |
2053 | retval = PTR_ERR(bprm); | |
2054 | goto out_ret; | |
2055 | } | |
2056 | ||
2057 | retval = count_strings_kernel(argv); | |
2058 | if (retval < 0) | |
2059 | goto out_free; | |
2060 | bprm->argc = retval; | |
2061 | ||
2062 | retval = count_strings_kernel(envp); | |
2063 | if (retval < 0) | |
2064 | goto out_free; | |
2065 | bprm->envc = retval; | |
2066 | ||
2067 | retval = bprm_stack_limits(bprm); | |
2068 | if (retval < 0) | |
2069 | goto out_free; | |
2070 | ||
2071 | retval = copy_string_kernel(bprm->filename, bprm); | |
2072 | if (retval < 0) | |
2073 | goto out_free; | |
2074 | bprm->exec = bprm->p; | |
2075 | ||
2076 | retval = copy_strings_kernel(bprm->envc, envp, bprm); | |
2077 | if (retval < 0) | |
2078 | goto out_free; | |
2079 | ||
2080 | retval = copy_strings_kernel(bprm->argc, argv, bprm); | |
2081 | if (retval < 0) | |
2082 | goto out_free; | |
2083 | ||
2084 | retval = bprm_execve(bprm, fd, filename, 0); | |
2085 | out_free: | |
2086 | free_bprm(bprm); | |
2087 | out_ret: | |
2088 | putname(filename); | |
2089 | return retval; | |
2090 | } | |
2091 | ||
2092 | static int do_execve(struct filename *filename, | |
ba2d0162 | 2093 | const char __user *const __user *__argv, |
da3d4c5f | 2094 | const char __user *const __user *__envp) |
ba2d0162 | 2095 | { |
0e028465 ON |
2096 | struct user_arg_ptr argv = { .ptr.native = __argv }; |
2097 | struct user_arg_ptr envp = { .ptr.native = __envp }; | |
51f39a1f DD |
2098 | return do_execveat_common(AT_FDCWD, filename, argv, envp, 0); |
2099 | } | |
2100 | ||
be619f7f | 2101 | static int do_execveat(int fd, struct filename *filename, |
51f39a1f DD |
2102 | const char __user *const __user *__argv, |
2103 | const char __user *const __user *__envp, | |
2104 | int flags) | |
2105 | { | |
2106 | struct user_arg_ptr argv = { .ptr.native = __argv }; | |
2107 | struct user_arg_ptr envp = { .ptr.native = __envp }; | |
2108 | ||
2109 | return do_execveat_common(fd, filename, argv, envp, flags); | |
0e028465 ON |
2110 | } |
2111 | ||
2112 | #ifdef CONFIG_COMPAT | |
c4ad8f98 | 2113 | static int compat_do_execve(struct filename *filename, |
38b983b3 | 2114 | const compat_uptr_t __user *__argv, |
d03d26e5 | 2115 | const compat_uptr_t __user *__envp) |
0e028465 ON |
2116 | { |
2117 | struct user_arg_ptr argv = { | |
2118 | .is_compat = true, | |
2119 | .ptr.compat = __argv, | |
2120 | }; | |
2121 | struct user_arg_ptr envp = { | |
2122 | .is_compat = true, | |
2123 | .ptr.compat = __envp, | |
2124 | }; | |
51f39a1f DD |
2125 | return do_execveat_common(AT_FDCWD, filename, argv, envp, 0); |
2126 | } | |
2127 | ||
2128 | static int compat_do_execveat(int fd, struct filename *filename, | |
2129 | const compat_uptr_t __user *__argv, | |
2130 | const compat_uptr_t __user *__envp, | |
2131 | int flags) | |
2132 | { | |
2133 | struct user_arg_ptr argv = { | |
2134 | .is_compat = true, | |
2135 | .ptr.compat = __argv, | |
2136 | }; | |
2137 | struct user_arg_ptr envp = { | |
2138 | .is_compat = true, | |
2139 | .ptr.compat = __envp, | |
2140 | }; | |
2141 | return do_execveat_common(fd, filename, argv, envp, flags); | |
ba2d0162 | 2142 | } |
0e028465 | 2143 | #endif |
ba2d0162 | 2144 | |
964ee7df | 2145 | void set_binfmt(struct linux_binfmt *new) |
1da177e4 | 2146 | { |
801460d0 HS |
2147 | struct mm_struct *mm = current->mm; |
2148 | ||
2149 | if (mm->binfmt) | |
2150 | module_put(mm->binfmt->module); | |
1da177e4 | 2151 | |
801460d0 | 2152 | mm->binfmt = new; |
964ee7df ON |
2153 | if (new) |
2154 | __module_get(new->module); | |
1da177e4 | 2155 | } |
1da177e4 LT |
2156 | EXPORT_SYMBOL(set_binfmt); |
2157 | ||
6c5d5238 | 2158 | /* |
7288e118 | 2159 | * set_dumpable stores three-value SUID_DUMP_* into mm->flags. |
6c5d5238 KH |
2160 | */ |
2161 | void set_dumpable(struct mm_struct *mm, int value) | |
2162 | { | |
7288e118 ON |
2163 | if (WARN_ON((unsigned)value > SUID_DUMP_ROOT)) |
2164 | return; | |
2165 | ||
26e15225 | 2166 | set_mask_bits(&mm->flags, MMF_DUMPABLE_MASK, value); |
6c5d5238 | 2167 | } |
6c5d5238 | 2168 | |
38b983b3 AV |
2169 | SYSCALL_DEFINE3(execve, |
2170 | const char __user *, filename, | |
2171 | const char __user *const __user *, argv, | |
2172 | const char __user *const __user *, envp) | |
2173 | { | |
c4ad8f98 | 2174 | return do_execve(getname(filename), argv, envp); |
38b983b3 | 2175 | } |
51f39a1f DD |
2176 | |
2177 | SYSCALL_DEFINE5(execveat, | |
2178 | int, fd, const char __user *, filename, | |
2179 | const char __user *const __user *, argv, | |
2180 | const char __user *const __user *, envp, | |
2181 | int, flags) | |
2182 | { | |
2183 | int lookup_flags = (flags & AT_EMPTY_PATH) ? LOOKUP_EMPTY : 0; | |
2184 | ||
2185 | return do_execveat(fd, | |
2186 | getname_flags(filename, lookup_flags, NULL), | |
2187 | argv, envp, flags); | |
2188 | } | |
2189 | ||
38b983b3 | 2190 | #ifdef CONFIG_COMPAT |
625b1d7e HC |
2191 | COMPAT_SYSCALL_DEFINE3(execve, const char __user *, filename, |
2192 | const compat_uptr_t __user *, argv, | |
2193 | const compat_uptr_t __user *, envp) | |
38b983b3 | 2194 | { |
c4ad8f98 | 2195 | return compat_do_execve(getname(filename), argv, envp); |
38b983b3 | 2196 | } |
51f39a1f DD |
2197 | |
2198 | COMPAT_SYSCALL_DEFINE5(execveat, int, fd, | |
2199 | const char __user *, filename, | |
2200 | const compat_uptr_t __user *, argv, | |
2201 | const compat_uptr_t __user *, envp, | |
2202 | int, flags) | |
2203 | { | |
2204 | int lookup_flags = (flags & AT_EMPTY_PATH) ? LOOKUP_EMPTY : 0; | |
2205 | ||
2206 | return compat_do_execveat(fd, | |
2207 | getname_flags(filename, lookup_flags, NULL), | |
2208 | argv, envp, flags); | |
2209 | } | |
38b983b3 | 2210 | #endif |