]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/fs/exec.c | |
3 | * | |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | |
5 | */ | |
6 | ||
7 | /* | |
8 | * #!-checking implemented by tytso. | |
9 | */ | |
10 | /* | |
11 | * Demand-loading implemented 01.12.91 - no need to read anything but | |
12 | * the header into memory. The inode of the executable is put into | |
13 | * "current->executable", and page faults do the actual loading. Clean. | |
14 | * | |
15 | * Once more I can proudly say that linux stood up to being changed: it | |
16 | * was less than 2 hours work to get demand-loading completely implemented. | |
17 | * | |
18 | * Demand loading changed July 1993 by Eric Youngdale. Use mmap instead, | |
19 | * current->executable is only used by the procfs. This allows a dispatch | |
20 | * table to check for several different types of binary formats. We keep | |
21 | * trying until we recognize the file or we run out of supported binary | |
22 | * formats. | |
23 | */ | |
24 | ||
1da177e4 LT |
25 | #include <linux/slab.h> |
26 | #include <linux/file.h> | |
9f3acc31 | 27 | #include <linux/fdtable.h> |
ba92a43d | 28 | #include <linux/mm.h> |
615d6e87 | 29 | #include <linux/vmacache.h> |
1da177e4 LT |
30 | #include <linux/stat.h> |
31 | #include <linux/fcntl.h> | |
ba92a43d | 32 | #include <linux/swap.h> |
74aadce9 | 33 | #include <linux/string.h> |
1da177e4 | 34 | #include <linux/init.h> |
ca5b172b | 35 | #include <linux/pagemap.h> |
cdd6c482 | 36 | #include <linux/perf_event.h> |
1da177e4 LT |
37 | #include <linux/highmem.h> |
38 | #include <linux/spinlock.h> | |
39 | #include <linux/key.h> | |
40 | #include <linux/personality.h> | |
41 | #include <linux/binfmts.h> | |
1da177e4 | 42 | #include <linux/utsname.h> |
84d73786 | 43 | #include <linux/pid_namespace.h> |
1da177e4 LT |
44 | #include <linux/module.h> |
45 | #include <linux/namei.h> | |
1da177e4 LT |
46 | #include <linux/mount.h> |
47 | #include <linux/security.h> | |
48 | #include <linux/syscalls.h> | |
8f0ab514 | 49 | #include <linux/tsacct_kern.h> |
9f46080c | 50 | #include <linux/cn_proc.h> |
473ae30b | 51 | #include <linux/audit.h> |
6341c393 | 52 | #include <linux/tracehook.h> |
5f4123be | 53 | #include <linux/kmod.h> |
6110e3ab | 54 | #include <linux/fsnotify.h> |
5ad4e53b | 55 | #include <linux/fs_struct.h> |
61be228a | 56 | #include <linux/pipe_fs_i.h> |
3d5992d2 | 57 | #include <linux/oom.h> |
0e028465 | 58 | #include <linux/compat.h> |
1da177e4 LT |
59 | |
60 | #include <asm/uaccess.h> | |
61 | #include <asm/mmu_context.h> | |
b6a2fea3 | 62 | #include <asm/tlb.h> |
fe3d197f | 63 | #include <asm/mpx.h> |
43d2b113 KH |
64 | |
65 | #include <trace/events/task.h> | |
a6f76f23 | 66 | #include "internal.h" |
1da177e4 | 67 | |
4ff16c25 DS |
68 | #include <trace/events/sched.h> |
69 | ||
d6e71144 AC |
70 | int suid_dumpable = 0; |
71 | ||
e4dc1b14 | 72 | static LIST_HEAD(formats); |
1da177e4 LT |
73 | static DEFINE_RWLOCK(binfmt_lock); |
74 | ||
8fc3dc5a | 75 | void __register_binfmt(struct linux_binfmt * fmt, int insert) |
1da177e4 | 76 | { |
8fc3dc5a | 77 | BUG_ON(!fmt); |
92eaa565 ON |
78 | if (WARN_ON(!fmt->load_binary)) |
79 | return; | |
1da177e4 | 80 | write_lock(&binfmt_lock); |
74641f58 IK |
81 | insert ? list_add(&fmt->lh, &formats) : |
82 | list_add_tail(&fmt->lh, &formats); | |
1da177e4 | 83 | write_unlock(&binfmt_lock); |
1da177e4 LT |
84 | } |
85 | ||
74641f58 | 86 | EXPORT_SYMBOL(__register_binfmt); |
1da177e4 | 87 | |
f6b450d4 | 88 | void unregister_binfmt(struct linux_binfmt * fmt) |
1da177e4 | 89 | { |
1da177e4 | 90 | write_lock(&binfmt_lock); |
e4dc1b14 | 91 | list_del(&fmt->lh); |
1da177e4 | 92 | write_unlock(&binfmt_lock); |
1da177e4 LT |
93 | } |
94 | ||
95 | EXPORT_SYMBOL(unregister_binfmt); | |
96 | ||
97 | static inline void put_binfmt(struct linux_binfmt * fmt) | |
98 | { | |
99 | module_put(fmt->module); | |
100 | } | |
101 | ||
69369a70 | 102 | #ifdef CONFIG_USELIB |
1da177e4 LT |
103 | /* |
104 | * Note that a shared library must be both readable and executable due to | |
105 | * security reasons. | |
106 | * | |
107 | * Also note that we take the address to load from from the file itself. | |
108 | */ | |
1e7bfb21 | 109 | SYSCALL_DEFINE1(uselib, const char __user *, library) |
1da177e4 | 110 | { |
72c2d531 | 111 | struct linux_binfmt *fmt; |
964bd183 | 112 | struct file *file; |
91a27b2a | 113 | struct filename *tmp = getname(library); |
964bd183 | 114 | int error = PTR_ERR(tmp); |
47c805dc AV |
115 | static const struct open_flags uselib_flags = { |
116 | .open_flag = O_LARGEFILE | O_RDONLY | __FMODE_EXEC, | |
117 | .acc_mode = MAY_READ | MAY_EXEC | MAY_OPEN, | |
f9652e10 AV |
118 | .intent = LOOKUP_OPEN, |
119 | .lookup_flags = LOOKUP_FOLLOW, | |
47c805dc | 120 | }; |
964bd183 | 121 | |
6e8341a1 AV |
122 | if (IS_ERR(tmp)) |
123 | goto out; | |
124 | ||
f9652e10 | 125 | file = do_filp_open(AT_FDCWD, tmp, &uselib_flags); |
6e8341a1 AV |
126 | putname(tmp); |
127 | error = PTR_ERR(file); | |
128 | if (IS_ERR(file)) | |
1da177e4 LT |
129 | goto out; |
130 | ||
131 | error = -EINVAL; | |
496ad9aa | 132 | if (!S_ISREG(file_inode(file)->i_mode)) |
1da177e4 LT |
133 | goto exit; |
134 | ||
30524472 | 135 | error = -EACCES; |
6e8341a1 | 136 | if (file->f_path.mnt->mnt_flags & MNT_NOEXEC) |
1da177e4 LT |
137 | goto exit; |
138 | ||
2a12a9d7 | 139 | fsnotify_open(file); |
6110e3ab | 140 | |
1da177e4 | 141 | error = -ENOEXEC; |
1da177e4 | 142 | |
72c2d531 AV |
143 | read_lock(&binfmt_lock); |
144 | list_for_each_entry(fmt, &formats, lh) { | |
145 | if (!fmt->load_shlib) | |
146 | continue; | |
147 | if (!try_module_get(fmt->module)) | |
148 | continue; | |
1da177e4 | 149 | read_unlock(&binfmt_lock); |
72c2d531 AV |
150 | error = fmt->load_shlib(file); |
151 | read_lock(&binfmt_lock); | |
152 | put_binfmt(fmt); | |
153 | if (error != -ENOEXEC) | |
154 | break; | |
1da177e4 | 155 | } |
72c2d531 | 156 | read_unlock(&binfmt_lock); |
6e8341a1 | 157 | exit: |
1da177e4 LT |
158 | fput(file); |
159 | out: | |
160 | return error; | |
1da177e4 | 161 | } |
69369a70 | 162 | #endif /* #ifdef CONFIG_USELIB */ |
1da177e4 | 163 | |
b6a2fea3 | 164 | #ifdef CONFIG_MMU |
ae6b585e ON |
165 | /* |
166 | * The nascent bprm->mm is not visible until exec_mmap() but it can | |
167 | * use a lot of memory, account these pages in current->mm temporary | |
168 | * for oom_badness()->get_mm_rss(). Once exec succeeds or fails, we | |
169 | * change the counter back via acct_arg_size(0). | |
170 | */ | |
0e028465 | 171 | static void acct_arg_size(struct linux_binprm *bprm, unsigned long pages) |
3c77f845 ON |
172 | { |
173 | struct mm_struct *mm = current->mm; | |
174 | long diff = (long)(pages - bprm->vma_pages); | |
175 | ||
176 | if (!mm || !diff) | |
177 | return; | |
178 | ||
179 | bprm->vma_pages = pages; | |
3c77f845 | 180 | add_mm_counter(mm, MM_ANONPAGES, diff); |
3c77f845 ON |
181 | } |
182 | ||
0e028465 | 183 | static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos, |
b6a2fea3 OW |
184 | int write) |
185 | { | |
186 | struct page *page; | |
187 | int ret; | |
188 | ||
189 | #ifdef CONFIG_STACK_GROWSUP | |
190 | if (write) { | |
d05f3169 | 191 | ret = expand_downwards(bprm->vma, pos); |
b6a2fea3 OW |
192 | if (ret < 0) |
193 | return NULL; | |
194 | } | |
195 | #endif | |
196 | ret = get_user_pages(current, bprm->mm, pos, | |
197 | 1, write, 1, &page, NULL); | |
198 | if (ret <= 0) | |
199 | return NULL; | |
200 | ||
201 | if (write) { | |
b6a2fea3 | 202 | unsigned long size = bprm->vma->vm_end - bprm->vma->vm_start; |
a64e715f LT |
203 | struct rlimit *rlim; |
204 | ||
3c77f845 ON |
205 | acct_arg_size(bprm, size / PAGE_SIZE); |
206 | ||
a64e715f LT |
207 | /* |
208 | * We've historically supported up to 32 pages (ARG_MAX) | |
209 | * of argument strings even with small stacks | |
210 | */ | |
211 | if (size <= ARG_MAX) | |
212 | return page; | |
b6a2fea3 OW |
213 | |
214 | /* | |
215 | * Limit to 1/4-th the stack size for the argv+env strings. | |
216 | * This ensures that: | |
217 | * - the remaining binfmt code will not run out of stack space, | |
218 | * - the program will have a reasonable amount of stack left | |
219 | * to work from. | |
220 | */ | |
a64e715f | 221 | rlim = current->signal->rlim; |
d554ed89 | 222 | if (size > ACCESS_ONCE(rlim[RLIMIT_STACK].rlim_cur) / 4) { |
b6a2fea3 OW |
223 | put_page(page); |
224 | return NULL; | |
225 | } | |
226 | } | |
227 | ||
228 | return page; | |
229 | } | |
230 | ||
231 | static void put_arg_page(struct page *page) | |
232 | { | |
233 | put_page(page); | |
234 | } | |
235 | ||
236 | static void free_arg_page(struct linux_binprm *bprm, int i) | |
237 | { | |
238 | } | |
239 | ||
240 | static void free_arg_pages(struct linux_binprm *bprm) | |
241 | { | |
242 | } | |
243 | ||
244 | static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos, | |
245 | struct page *page) | |
246 | { | |
247 | flush_cache_page(bprm->vma, pos, page_to_pfn(page)); | |
248 | } | |
249 | ||
250 | static int __bprm_mm_init(struct linux_binprm *bprm) | |
251 | { | |
eaccbfa5 | 252 | int err; |
b6a2fea3 OW |
253 | struct vm_area_struct *vma = NULL; |
254 | struct mm_struct *mm = bprm->mm; | |
255 | ||
256 | bprm->vma = vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL); | |
257 | if (!vma) | |
eaccbfa5 | 258 | return -ENOMEM; |
b6a2fea3 OW |
259 | |
260 | down_write(&mm->mmap_sem); | |
261 | vma->vm_mm = mm; | |
262 | ||
263 | /* | |
264 | * Place the stack at the largest stack address the architecture | |
265 | * supports. Later, we'll move this to an appropriate place. We don't | |
266 | * use STACK_TOP because that can depend on attributes which aren't | |
267 | * configured yet. | |
268 | */ | |
aacb3d17 | 269 | BUILD_BUG_ON(VM_STACK_FLAGS & VM_STACK_INCOMPLETE_SETUP); |
b6a2fea3 OW |
270 | vma->vm_end = STACK_TOP_MAX; |
271 | vma->vm_start = vma->vm_end - PAGE_SIZE; | |
d9104d1c | 272 | vma->vm_flags = VM_SOFTDIRTY | VM_STACK_FLAGS | VM_STACK_INCOMPLETE_SETUP; |
3ed75eb8 | 273 | vma->vm_page_prot = vm_get_page_prot(vma->vm_flags); |
5beb4930 | 274 | INIT_LIST_HEAD(&vma->anon_vma_chain); |
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; | |
fe3d197f | 281 | arch_bprm_mm_init(mm, vma); |
b6a2fea3 | 282 | up_write(&mm->mmap_sem); |
b6a2fea3 | 283 | bprm->p = vma->vm_end - sizeof(void *); |
b6a2fea3 | 284 | return 0; |
b6a2fea3 | 285 | err: |
eaccbfa5 LFC |
286 | up_write(&mm->mmap_sem); |
287 | bprm->vma = NULL; | |
288 | kmem_cache_free(vm_area_cachep, 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 | ||
b6a2fea3 OW |
373 | err = __bprm_mm_init(bprm); |
374 | if (err) | |
375 | goto err; | |
376 | ||
377 | return 0; | |
378 | ||
379 | err: | |
380 | if (mm) { | |
381 | bprm->mm = NULL; | |
382 | mmdrop(mm); | |
383 | } | |
384 | ||
385 | return err; | |
386 | } | |
387 | ||
ba2d0162 | 388 | struct user_arg_ptr { |
0e028465 ON |
389 | #ifdef CONFIG_COMPAT |
390 | bool is_compat; | |
391 | #endif | |
392 | union { | |
393 | const char __user *const __user *native; | |
394 | #ifdef CONFIG_COMPAT | |
38b983b3 | 395 | const compat_uptr_t __user *compat; |
0e028465 ON |
396 | #endif |
397 | } ptr; | |
ba2d0162 ON |
398 | }; |
399 | ||
400 | static const char __user *get_user_arg_ptr(struct user_arg_ptr argv, int nr) | |
1d1dbf81 | 401 | { |
0e028465 ON |
402 | const char __user *native; |
403 | ||
404 | #ifdef CONFIG_COMPAT | |
405 | if (unlikely(argv.is_compat)) { | |
406 | compat_uptr_t compat; | |
407 | ||
408 | if (get_user(compat, argv.ptr.compat + nr)) | |
409 | return ERR_PTR(-EFAULT); | |
1d1dbf81 | 410 | |
0e028465 ON |
411 | return compat_ptr(compat); |
412 | } | |
413 | #endif | |
414 | ||
415 | if (get_user(native, argv.ptr.native + nr)) | |
1d1dbf81 ON |
416 | return ERR_PTR(-EFAULT); |
417 | ||
0e028465 | 418 | return native; |
1d1dbf81 ON |
419 | } |
420 | ||
1da177e4 LT |
421 | /* |
422 | * count() counts the number of strings in array ARGV. | |
423 | */ | |
ba2d0162 | 424 | static int count(struct user_arg_ptr argv, int max) |
1da177e4 LT |
425 | { |
426 | int i = 0; | |
427 | ||
0e028465 | 428 | if (argv.ptr.native != NULL) { |
1da177e4 | 429 | for (;;) { |
1d1dbf81 | 430 | const char __user *p = get_user_arg_ptr(argv, i); |
1da177e4 | 431 | |
1da177e4 LT |
432 | if (!p) |
433 | break; | |
1d1dbf81 ON |
434 | |
435 | if (IS_ERR(p)) | |
436 | return -EFAULT; | |
437 | ||
6d92d4f6 | 438 | if (i >= max) |
1da177e4 | 439 | return -E2BIG; |
6d92d4f6 | 440 | ++i; |
9aea5a65 RM |
441 | |
442 | if (fatal_signal_pending(current)) | |
443 | return -ERESTARTNOHAND; | |
1da177e4 LT |
444 | cond_resched(); |
445 | } | |
446 | } | |
447 | return i; | |
448 | } | |
449 | ||
450 | /* | |
b6a2fea3 OW |
451 | * 'copy_strings()' copies argument/environment strings from the old |
452 | * processes's memory to the new process's stack. The call to get_user_pages() | |
453 | * ensures the destination page is created and not swapped out. | |
1da177e4 | 454 | */ |
ba2d0162 | 455 | static int copy_strings(int argc, struct user_arg_ptr argv, |
75c96f85 | 456 | struct linux_binprm *bprm) |
1da177e4 LT |
457 | { |
458 | struct page *kmapped_page = NULL; | |
459 | char *kaddr = NULL; | |
b6a2fea3 | 460 | unsigned long kpos = 0; |
1da177e4 LT |
461 | int ret; |
462 | ||
463 | while (argc-- > 0) { | |
d7627467 | 464 | const char __user *str; |
1da177e4 LT |
465 | int len; |
466 | unsigned long pos; | |
467 | ||
1d1dbf81 ON |
468 | ret = -EFAULT; |
469 | str = get_user_arg_ptr(argv, argc); | |
470 | if (IS_ERR(str)) | |
1da177e4 | 471 | goto out; |
1da177e4 | 472 | |
1d1dbf81 ON |
473 | len = strnlen_user(str, MAX_ARG_STRLEN); |
474 | if (!len) | |
475 | goto out; | |
476 | ||
477 | ret = -E2BIG; | |
478 | if (!valid_arg_len(bprm, len)) | |
1da177e4 | 479 | goto out; |
1da177e4 | 480 | |
b6a2fea3 | 481 | /* We're going to work our way backwords. */ |
1da177e4 | 482 | pos = bprm->p; |
b6a2fea3 OW |
483 | str += len; |
484 | bprm->p -= len; | |
1da177e4 LT |
485 | |
486 | while (len > 0) { | |
1da177e4 | 487 | int offset, bytes_to_copy; |
1da177e4 | 488 | |
9aea5a65 RM |
489 | if (fatal_signal_pending(current)) { |
490 | ret = -ERESTARTNOHAND; | |
491 | goto out; | |
492 | } | |
7993bc1f RM |
493 | cond_resched(); |
494 | ||
1da177e4 | 495 | offset = pos % PAGE_SIZE; |
b6a2fea3 OW |
496 | if (offset == 0) |
497 | offset = PAGE_SIZE; | |
498 | ||
499 | bytes_to_copy = offset; | |
500 | if (bytes_to_copy > len) | |
501 | bytes_to_copy = len; | |
502 | ||
503 | offset -= bytes_to_copy; | |
504 | pos -= bytes_to_copy; | |
505 | str -= bytes_to_copy; | |
506 | len -= bytes_to_copy; | |
507 | ||
508 | if (!kmapped_page || kpos != (pos & PAGE_MASK)) { | |
509 | struct page *page; | |
510 | ||
511 | page = get_arg_page(bprm, pos, 1); | |
1da177e4 | 512 | if (!page) { |
b6a2fea3 | 513 | ret = -E2BIG; |
1da177e4 LT |
514 | goto out; |
515 | } | |
1da177e4 | 516 | |
b6a2fea3 OW |
517 | if (kmapped_page) { |
518 | flush_kernel_dcache_page(kmapped_page); | |
1da177e4 | 519 | kunmap(kmapped_page); |
b6a2fea3 OW |
520 | put_arg_page(kmapped_page); |
521 | } | |
1da177e4 LT |
522 | kmapped_page = page; |
523 | kaddr = kmap(kmapped_page); | |
b6a2fea3 OW |
524 | kpos = pos & PAGE_MASK; |
525 | flush_arg_page(bprm, kpos, kmapped_page); | |
1da177e4 | 526 | } |
b6a2fea3 | 527 | if (copy_from_user(kaddr+offset, str, bytes_to_copy)) { |
1da177e4 LT |
528 | ret = -EFAULT; |
529 | goto out; | |
530 | } | |
1da177e4 LT |
531 | } |
532 | } | |
533 | ret = 0; | |
534 | out: | |
b6a2fea3 OW |
535 | if (kmapped_page) { |
536 | flush_kernel_dcache_page(kmapped_page); | |
1da177e4 | 537 | kunmap(kmapped_page); |
b6a2fea3 OW |
538 | put_arg_page(kmapped_page); |
539 | } | |
1da177e4 LT |
540 | return ret; |
541 | } | |
542 | ||
543 | /* | |
544 | * Like copy_strings, but get argv and its values from kernel memory. | |
545 | */ | |
ba2d0162 | 546 | int copy_strings_kernel(int argc, const char *const *__argv, |
d7627467 | 547 | struct linux_binprm *bprm) |
1da177e4 LT |
548 | { |
549 | int r; | |
550 | mm_segment_t oldfs = get_fs(); | |
ba2d0162 | 551 | struct user_arg_ptr argv = { |
0e028465 | 552 | .ptr.native = (const char __user *const __user *)__argv, |
ba2d0162 ON |
553 | }; |
554 | ||
1da177e4 | 555 | set_fs(KERNEL_DS); |
ba2d0162 | 556 | r = copy_strings(argc, argv, bprm); |
1da177e4 | 557 | set_fs(oldfs); |
ba2d0162 | 558 | |
1da177e4 LT |
559 | return r; |
560 | } | |
1da177e4 LT |
561 | EXPORT_SYMBOL(copy_strings_kernel); |
562 | ||
563 | #ifdef CONFIG_MMU | |
b6a2fea3 | 564 | |
1da177e4 | 565 | /* |
b6a2fea3 OW |
566 | * During bprm_mm_init(), we create a temporary stack at STACK_TOP_MAX. Once |
567 | * the binfmt code determines where the new stack should reside, we shift it to | |
568 | * its final location. The process proceeds as follows: | |
1da177e4 | 569 | * |
b6a2fea3 OW |
570 | * 1) Use shift to calculate the new vma endpoints. |
571 | * 2) Extend vma to cover both the old and new ranges. This ensures the | |
572 | * arguments passed to subsequent functions are consistent. | |
573 | * 3) Move vma's page tables to the new range. | |
574 | * 4) Free up any cleared pgd range. | |
575 | * 5) Shrink the vma to cover only the new range. | |
1da177e4 | 576 | */ |
b6a2fea3 | 577 | static int shift_arg_pages(struct vm_area_struct *vma, unsigned long shift) |
1da177e4 LT |
578 | { |
579 | struct mm_struct *mm = vma->vm_mm; | |
b6a2fea3 OW |
580 | unsigned long old_start = vma->vm_start; |
581 | unsigned long old_end = vma->vm_end; | |
582 | unsigned long length = old_end - old_start; | |
583 | unsigned long new_start = old_start - shift; | |
584 | unsigned long new_end = old_end - shift; | |
d16dfc55 | 585 | struct mmu_gather tlb; |
1da177e4 | 586 | |
b6a2fea3 | 587 | BUG_ON(new_start > new_end); |
1da177e4 | 588 | |
b6a2fea3 OW |
589 | /* |
590 | * ensure there are no vmas between where we want to go | |
591 | * and where we are | |
592 | */ | |
593 | if (vma != find_vma(mm, new_start)) | |
594 | return -EFAULT; | |
595 | ||
596 | /* | |
597 | * cover the whole range: [new_start, old_end) | |
598 | */ | |
5beb4930 RR |
599 | if (vma_adjust(vma, new_start, old_end, vma->vm_pgoff, NULL)) |
600 | return -ENOMEM; | |
b6a2fea3 OW |
601 | |
602 | /* | |
603 | * move the page tables downwards, on failure we rely on | |
604 | * process cleanup to remove whatever mess we made. | |
605 | */ | |
606 | if (length != move_page_tables(vma, old_start, | |
38a76013 | 607 | vma, new_start, length, false)) |
b6a2fea3 OW |
608 | return -ENOMEM; |
609 | ||
610 | lru_add_drain(); | |
2b047252 | 611 | tlb_gather_mmu(&tlb, mm, old_start, old_end); |
b6a2fea3 OW |
612 | if (new_end > old_start) { |
613 | /* | |
614 | * when the old and new regions overlap clear from new_end. | |
615 | */ | |
d16dfc55 | 616 | free_pgd_range(&tlb, new_end, old_end, new_end, |
6ee8630e | 617 | vma->vm_next ? vma->vm_next->vm_start : USER_PGTABLES_CEILING); |
b6a2fea3 OW |
618 | } else { |
619 | /* | |
620 | * otherwise, clean from old_start; this is done to not touch | |
621 | * the address space in [new_end, old_start) some architectures | |
622 | * have constraints on va-space that make this illegal (IA64) - | |
623 | * for the others its just a little faster. | |
624 | */ | |
d16dfc55 | 625 | free_pgd_range(&tlb, old_start, old_end, new_end, |
6ee8630e | 626 | vma->vm_next ? vma->vm_next->vm_start : USER_PGTABLES_CEILING); |
1da177e4 | 627 | } |
2b047252 | 628 | tlb_finish_mmu(&tlb, old_start, old_end); |
b6a2fea3 OW |
629 | |
630 | /* | |
5beb4930 | 631 | * Shrink the vma to just the new range. Always succeeds. |
b6a2fea3 OW |
632 | */ |
633 | vma_adjust(vma, new_start, new_end, vma->vm_pgoff, NULL); | |
634 | ||
635 | return 0; | |
1da177e4 LT |
636 | } |
637 | ||
b6a2fea3 OW |
638 | /* |
639 | * Finalizes the stack vm_area_struct. The flags and permissions are updated, | |
640 | * the stack is optionally relocated, and some extra space is added. | |
641 | */ | |
1da177e4 LT |
642 | int setup_arg_pages(struct linux_binprm *bprm, |
643 | unsigned long stack_top, | |
644 | int executable_stack) | |
645 | { | |
b6a2fea3 OW |
646 | unsigned long ret; |
647 | unsigned long stack_shift; | |
1da177e4 | 648 | struct mm_struct *mm = current->mm; |
b6a2fea3 OW |
649 | struct vm_area_struct *vma = bprm->vma; |
650 | struct vm_area_struct *prev = NULL; | |
651 | unsigned long vm_flags; | |
652 | unsigned long stack_base; | |
803bf5ec MN |
653 | unsigned long stack_size; |
654 | unsigned long stack_expand; | |
655 | unsigned long rlim_stack; | |
1da177e4 LT |
656 | |
657 | #ifdef CONFIG_STACK_GROWSUP | |
d71f290b | 658 | /* Limit stack size */ |
d554ed89 | 659 | stack_base = rlimit_max(RLIMIT_STACK); |
d71f290b JH |
660 | if (stack_base > STACK_SIZE_MAX) |
661 | stack_base = STACK_SIZE_MAX; | |
1da177e4 | 662 | |
b6a2fea3 OW |
663 | /* Make sure we didn't let the argument array grow too large. */ |
664 | if (vma->vm_end - vma->vm_start > stack_base) | |
665 | return -ENOMEM; | |
1da177e4 | 666 | |
b6a2fea3 | 667 | stack_base = PAGE_ALIGN(stack_top - stack_base); |
1da177e4 | 668 | |
b6a2fea3 OW |
669 | stack_shift = vma->vm_start - stack_base; |
670 | mm->arg_start = bprm->p - stack_shift; | |
671 | bprm->p = vma->vm_end - stack_shift; | |
1da177e4 | 672 | #else |
b6a2fea3 OW |
673 | stack_top = arch_align_stack(stack_top); |
674 | stack_top = PAGE_ALIGN(stack_top); | |
1b528181 RM |
675 | |
676 | if (unlikely(stack_top < mmap_min_addr) || | |
677 | unlikely(vma->vm_end - vma->vm_start >= stack_top - mmap_min_addr)) | |
678 | return -ENOMEM; | |
679 | ||
b6a2fea3 OW |
680 | stack_shift = vma->vm_end - stack_top; |
681 | ||
682 | bprm->p -= stack_shift; | |
1da177e4 | 683 | mm->arg_start = bprm->p; |
1da177e4 LT |
684 | #endif |
685 | ||
1da177e4 | 686 | if (bprm->loader) |
b6a2fea3 OW |
687 | bprm->loader -= stack_shift; |
688 | bprm->exec -= stack_shift; | |
1da177e4 | 689 | |
1da177e4 | 690 | down_write(&mm->mmap_sem); |
96a8e13e | 691 | vm_flags = VM_STACK_FLAGS; |
b6a2fea3 OW |
692 | |
693 | /* | |
694 | * Adjust stack execute permissions; explicitly enable for | |
695 | * EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X and leave alone | |
696 | * (arch default) otherwise. | |
697 | */ | |
698 | if (unlikely(executable_stack == EXSTACK_ENABLE_X)) | |
699 | vm_flags |= VM_EXEC; | |
700 | else if (executable_stack == EXSTACK_DISABLE_X) | |
701 | vm_flags &= ~VM_EXEC; | |
702 | vm_flags |= mm->def_flags; | |
a8bef8ff | 703 | vm_flags |= VM_STACK_INCOMPLETE_SETUP; |
b6a2fea3 OW |
704 | |
705 | ret = mprotect_fixup(vma, &prev, vma->vm_start, vma->vm_end, | |
706 | vm_flags); | |
707 | if (ret) | |
708 | goto out_unlock; | |
709 | BUG_ON(prev != vma); | |
710 | ||
711 | /* Move stack pages down in memory. */ | |
712 | if (stack_shift) { | |
713 | ret = shift_arg_pages(vma, stack_shift); | |
fc63cf23 AB |
714 | if (ret) |
715 | goto out_unlock; | |
1da177e4 LT |
716 | } |
717 | ||
a8bef8ff MG |
718 | /* mprotect_fixup is overkill to remove the temporary stack flags */ |
719 | vma->vm_flags &= ~VM_STACK_INCOMPLETE_SETUP; | |
720 | ||
5ef097dd | 721 | stack_expand = 131072UL; /* randomly 32*4k (or 2*64k) pages */ |
803bf5ec MN |
722 | stack_size = vma->vm_end - vma->vm_start; |
723 | /* | |
724 | * Align this down to a page boundary as expand_stack | |
725 | * will align it up. | |
726 | */ | |
727 | rlim_stack = rlimit(RLIMIT_STACK) & PAGE_MASK; | |
b6a2fea3 | 728 | #ifdef CONFIG_STACK_GROWSUP |
803bf5ec MN |
729 | if (stack_size + stack_expand > rlim_stack) |
730 | stack_base = vma->vm_start + rlim_stack; | |
731 | else | |
732 | stack_base = vma->vm_end + stack_expand; | |
b6a2fea3 | 733 | #else |
803bf5ec MN |
734 | if (stack_size + stack_expand > rlim_stack) |
735 | stack_base = vma->vm_end - rlim_stack; | |
736 | else | |
737 | stack_base = vma->vm_start - stack_expand; | |
b6a2fea3 | 738 | #endif |
3af9e859 | 739 | current->mm->start_stack = bprm->p; |
b6a2fea3 OW |
740 | ret = expand_stack(vma, stack_base); |
741 | if (ret) | |
742 | ret = -EFAULT; | |
743 | ||
744 | out_unlock: | |
1da177e4 | 745 | up_write(&mm->mmap_sem); |
fc63cf23 | 746 | return ret; |
1da177e4 | 747 | } |
1da177e4 LT |
748 | EXPORT_SYMBOL(setup_arg_pages); |
749 | ||
1da177e4 LT |
750 | #endif /* CONFIG_MMU */ |
751 | ||
c4ad8f98 | 752 | static struct file *do_open_exec(struct filename *name) |
1da177e4 | 753 | { |
1da177e4 | 754 | struct file *file; |
e56b6a5d | 755 | int err; |
47c805dc AV |
756 | static const struct open_flags open_exec_flags = { |
757 | .open_flag = O_LARGEFILE | O_RDONLY | __FMODE_EXEC, | |
758 | .acc_mode = MAY_EXEC | MAY_OPEN, | |
f9652e10 AV |
759 | .intent = LOOKUP_OPEN, |
760 | .lookup_flags = LOOKUP_FOLLOW, | |
47c805dc | 761 | }; |
1da177e4 | 762 | |
c4ad8f98 | 763 | file = do_filp_open(AT_FDCWD, name, &open_exec_flags); |
6e8341a1 | 764 | if (IS_ERR(file)) |
e56b6a5d CH |
765 | goto out; |
766 | ||
767 | err = -EACCES; | |
496ad9aa | 768 | if (!S_ISREG(file_inode(file)->i_mode)) |
6e8341a1 | 769 | goto exit; |
e56b6a5d | 770 | |
6e8341a1 AV |
771 | if (file->f_path.mnt->mnt_flags & MNT_NOEXEC) |
772 | goto exit; | |
e56b6a5d | 773 | |
2a12a9d7 | 774 | fsnotify_open(file); |
6110e3ab | 775 | |
e56b6a5d | 776 | err = deny_write_access(file); |
6e8341a1 AV |
777 | if (err) |
778 | goto exit; | |
1da177e4 | 779 | |
6e8341a1 | 780 | out: |
e56b6a5d CH |
781 | return file; |
782 | ||
6e8341a1 AV |
783 | exit: |
784 | fput(file); | |
e56b6a5d CH |
785 | return ERR_PTR(err); |
786 | } | |
c4ad8f98 LT |
787 | |
788 | struct file *open_exec(const char *name) | |
789 | { | |
790 | struct filename tmp = { .name = name }; | |
791 | return do_open_exec(&tmp); | |
792 | } | |
1da177e4 LT |
793 | EXPORT_SYMBOL(open_exec); |
794 | ||
6777d773 MZ |
795 | int kernel_read(struct file *file, loff_t offset, |
796 | char *addr, unsigned long count) | |
1da177e4 LT |
797 | { |
798 | mm_segment_t old_fs; | |
799 | loff_t pos = offset; | |
800 | int result; | |
801 | ||
802 | old_fs = get_fs(); | |
803 | set_fs(get_ds()); | |
804 | /* The cast to a user pointer is valid due to the set_fs() */ | |
805 | result = vfs_read(file, (void __user *)addr, count, &pos); | |
806 | set_fs(old_fs); | |
807 | return result; | |
808 | } | |
809 | ||
810 | EXPORT_SYMBOL(kernel_read); | |
811 | ||
3dc20cb2 AV |
812 | ssize_t read_code(struct file *file, unsigned long addr, loff_t pos, size_t len) |
813 | { | |
ec695579 | 814 | ssize_t res = vfs_read(file, (void __user *)addr, len, &pos); |
3dc20cb2 AV |
815 | if (res > 0) |
816 | flush_icache_range(addr, addr + len); | |
817 | return res; | |
818 | } | |
819 | EXPORT_SYMBOL(read_code); | |
820 | ||
1da177e4 LT |
821 | static int exec_mmap(struct mm_struct *mm) |
822 | { | |
823 | struct task_struct *tsk; | |
615d6e87 | 824 | struct mm_struct *old_mm, *active_mm; |
1da177e4 LT |
825 | |
826 | /* Notify parent that we're no longer interested in the old VM */ | |
827 | tsk = current; | |
828 | old_mm = current->mm; | |
829 | mm_release(tsk, old_mm); | |
830 | ||
831 | if (old_mm) { | |
4fe7efdb | 832 | sync_mm_rss(old_mm); |
1da177e4 LT |
833 | /* |
834 | * Make sure that if there is a core dump in progress | |
835 | * for the old mm, we get out and die instead of going | |
836 | * through with the exec. We must hold mmap_sem around | |
999d9fc1 | 837 | * checking core_state and changing tsk->mm. |
1da177e4 LT |
838 | */ |
839 | down_read(&old_mm->mmap_sem); | |
999d9fc1 | 840 | if (unlikely(old_mm->core_state)) { |
1da177e4 LT |
841 | up_read(&old_mm->mmap_sem); |
842 | return -EINTR; | |
843 | } | |
844 | } | |
845 | task_lock(tsk); | |
846 | active_mm = tsk->active_mm; | |
847 | tsk->mm = mm; | |
848 | tsk->active_mm = mm; | |
849 | activate_mm(active_mm, mm); | |
615d6e87 DB |
850 | tsk->mm->vmacache_seqnum = 0; |
851 | vmacache_flush(tsk); | |
1da177e4 | 852 | task_unlock(tsk); |
1da177e4 LT |
853 | if (old_mm) { |
854 | up_read(&old_mm->mmap_sem); | |
7dddb12c | 855 | BUG_ON(active_mm != old_mm); |
701085b2 | 856 | setmax_mm_hiwater_rss(&tsk->signal->maxrss, old_mm); |
31a78f23 | 857 | mm_update_next_owner(old_mm); |
1da177e4 LT |
858 | mmput(old_mm); |
859 | return 0; | |
860 | } | |
861 | mmdrop(active_mm); | |
862 | return 0; | |
863 | } | |
864 | ||
865 | /* | |
866 | * This function makes sure the current process has its own signal table, | |
867 | * so that flush_signal_handlers can later reset the handlers without | |
868 | * disturbing other processes. (Other processes might share the signal | |
869 | * table via the CLONE_SIGHAND option to clone().) | |
870 | */ | |
858119e1 | 871 | static int de_thread(struct task_struct *tsk) |
1da177e4 LT |
872 | { |
873 | struct signal_struct *sig = tsk->signal; | |
b2c903b8 | 874 | struct sighand_struct *oldsighand = tsk->sighand; |
1da177e4 | 875 | spinlock_t *lock = &oldsighand->siglock; |
1da177e4 | 876 | |
aafe6c2a | 877 | if (thread_group_empty(tsk)) |
1da177e4 LT |
878 | goto no_thread_group; |
879 | ||
880 | /* | |
881 | * Kill all other threads in the thread group. | |
1da177e4 | 882 | */ |
1da177e4 | 883 | spin_lock_irq(lock); |
ed5d2cac | 884 | if (signal_group_exit(sig)) { |
1da177e4 LT |
885 | /* |
886 | * Another group action in progress, just | |
887 | * return so that the signal is processed. | |
888 | */ | |
889 | spin_unlock_irq(lock); | |
1da177e4 LT |
890 | return -EAGAIN; |
891 | } | |
d344193a | 892 | |
ed5d2cac | 893 | sig->group_exit_task = tsk; |
d344193a ON |
894 | sig->notify_count = zap_other_threads(tsk); |
895 | if (!thread_group_leader(tsk)) | |
896 | sig->notify_count--; | |
1da177e4 | 897 | |
d344193a | 898 | while (sig->notify_count) { |
d5bbd43d | 899 | __set_current_state(TASK_KILLABLE); |
1da177e4 LT |
900 | spin_unlock_irq(lock); |
901 | schedule(); | |
d5bbd43d ON |
902 | if (unlikely(__fatal_signal_pending(tsk))) |
903 | goto killed; | |
1da177e4 LT |
904 | spin_lock_irq(lock); |
905 | } | |
1da177e4 LT |
906 | spin_unlock_irq(lock); |
907 | ||
908 | /* | |
909 | * At this point all other threads have exited, all we have to | |
910 | * do is to wait for the thread group leader to become inactive, | |
911 | * and to assume its PID: | |
912 | */ | |
aafe6c2a | 913 | if (!thread_group_leader(tsk)) { |
8187926b | 914 | struct task_struct *leader = tsk->group_leader; |
6db840fa | 915 | |
2800d8d1 | 916 | sig->notify_count = -1; /* for exit_notify() */ |
6db840fa | 917 | for (;;) { |
e56fb287 | 918 | threadgroup_change_begin(tsk); |
6db840fa ON |
919 | write_lock_irq(&tasklist_lock); |
920 | if (likely(leader->exit_state)) | |
921 | break; | |
d5bbd43d | 922 | __set_current_state(TASK_KILLABLE); |
6db840fa | 923 | write_unlock_irq(&tasklist_lock); |
e56fb287 | 924 | threadgroup_change_end(tsk); |
6db840fa | 925 | schedule(); |
d5bbd43d ON |
926 | if (unlikely(__fatal_signal_pending(tsk))) |
927 | goto killed; | |
6db840fa | 928 | } |
1da177e4 | 929 | |
f5e90281 RM |
930 | /* |
931 | * The only record we have of the real-time age of a | |
932 | * process, regardless of execs it's done, is start_time. | |
933 | * All the past CPU time is accumulated in signal_struct | |
934 | * from sister threads now dead. But in this non-leader | |
935 | * exec, nothing survives from the original leader thread, | |
936 | * whose birth marks the true age of this process now. | |
937 | * When we take on its identity by switching to its PID, we | |
938 | * also take its birthdate (always earlier than our own). | |
939 | */ | |
aafe6c2a | 940 | tsk->start_time = leader->start_time; |
266b7a02 | 941 | tsk->real_start_time = leader->real_start_time; |
f5e90281 | 942 | |
bac0abd6 PE |
943 | BUG_ON(!same_thread_group(leader, tsk)); |
944 | BUG_ON(has_group_leader_pid(tsk)); | |
1da177e4 LT |
945 | /* |
946 | * An exec() starts a new thread group with the | |
947 | * TGID of the previous thread group. Rehash the | |
948 | * two threads with a switched PID, and release | |
949 | * the former thread group leader: | |
950 | */ | |
d73d6529 EB |
951 | |
952 | /* Become a process group leader with the old leader's pid. | |
c18258c6 EB |
953 | * The old leader becomes a thread of the this thread group. |
954 | * Note: The old leader also uses this pid until release_task | |
d73d6529 EB |
955 | * is called. Odd but simple and correct. |
956 | */ | |
aafe6c2a | 957 | tsk->pid = leader->pid; |
3f418548 | 958 | change_pid(tsk, PIDTYPE_PID, task_pid(leader)); |
aafe6c2a EB |
959 | transfer_pid(leader, tsk, PIDTYPE_PGID); |
960 | transfer_pid(leader, tsk, PIDTYPE_SID); | |
9cd80bbb | 961 | |
aafe6c2a | 962 | list_replace_rcu(&leader->tasks, &tsk->tasks); |
9cd80bbb | 963 | list_replace_init(&leader->sibling, &tsk->sibling); |
1da177e4 | 964 | |
aafe6c2a EB |
965 | tsk->group_leader = tsk; |
966 | leader->group_leader = tsk; | |
de12a787 | 967 | |
aafe6c2a | 968 | tsk->exit_signal = SIGCHLD; |
087806b1 | 969 | leader->exit_signal = -1; |
962b564c ON |
970 | |
971 | BUG_ON(leader->exit_state != EXIT_ZOMBIE); | |
972 | leader->exit_state = EXIT_DEAD; | |
eac1b5e5 ON |
973 | |
974 | /* | |
975 | * We are going to release_task()->ptrace_unlink() silently, | |
976 | * the tracer can sleep in do_wait(). EXIT_DEAD guarantees | |
977 | * the tracer wont't block again waiting for this thread. | |
978 | */ | |
979 | if (unlikely(leader->ptrace)) | |
980 | __wake_up_parent(leader, leader->parent); | |
1da177e4 | 981 | write_unlock_irq(&tasklist_lock); |
e56fb287 | 982 | threadgroup_change_end(tsk); |
8187926b ON |
983 | |
984 | release_task(leader); | |
ed5d2cac | 985 | } |
1da177e4 | 986 | |
6db840fa ON |
987 | sig->group_exit_task = NULL; |
988 | sig->notify_count = 0; | |
1da177e4 LT |
989 | |
990 | no_thread_group: | |
e6368253 ON |
991 | /* we have changed execution domain */ |
992 | tsk->exit_signal = SIGCHLD; | |
993 | ||
1da177e4 | 994 | exit_itimers(sig); |
cbaffba1 | 995 | flush_itimer_signals(); |
329f7dba | 996 | |
b2c903b8 ON |
997 | if (atomic_read(&oldsighand->count) != 1) { |
998 | struct sighand_struct *newsighand; | |
1da177e4 | 999 | /* |
b2c903b8 ON |
1000 | * This ->sighand is shared with the CLONE_SIGHAND |
1001 | * but not CLONE_THREAD task, switch to the new one. | |
1da177e4 | 1002 | */ |
b2c903b8 ON |
1003 | newsighand = kmem_cache_alloc(sighand_cachep, GFP_KERNEL); |
1004 | if (!newsighand) | |
1005 | return -ENOMEM; | |
1006 | ||
1da177e4 LT |
1007 | atomic_set(&newsighand->count, 1); |
1008 | memcpy(newsighand->action, oldsighand->action, | |
1009 | sizeof(newsighand->action)); | |
1010 | ||
1011 | write_lock_irq(&tasklist_lock); | |
1012 | spin_lock(&oldsighand->siglock); | |
aafe6c2a | 1013 | rcu_assign_pointer(tsk->sighand, newsighand); |
1da177e4 LT |
1014 | spin_unlock(&oldsighand->siglock); |
1015 | write_unlock_irq(&tasklist_lock); | |
1016 | ||
fba2afaa | 1017 | __cleanup_sighand(oldsighand); |
1da177e4 LT |
1018 | } |
1019 | ||
aafe6c2a | 1020 | BUG_ON(!thread_group_leader(tsk)); |
1da177e4 | 1021 | return 0; |
d5bbd43d ON |
1022 | |
1023 | killed: | |
1024 | /* protects against exit_notify() and __exit_signal() */ | |
1025 | read_lock(&tasklist_lock); | |
1026 | sig->group_exit_task = NULL; | |
1027 | sig->notify_count = 0; | |
1028 | read_unlock(&tasklist_lock); | |
1029 | return -EAGAIN; | |
1da177e4 | 1030 | } |
0840a90d | 1031 | |
59714d65 | 1032 | char *get_task_comm(char *buf, struct task_struct *tsk) |
1da177e4 LT |
1033 | { |
1034 | /* buf must be at least sizeof(tsk->comm) in size */ | |
1035 | task_lock(tsk); | |
1036 | strncpy(buf, tsk->comm, sizeof(tsk->comm)); | |
1037 | task_unlock(tsk); | |
59714d65 | 1038 | return buf; |
1da177e4 | 1039 | } |
7d74f492 | 1040 | EXPORT_SYMBOL_GPL(get_task_comm); |
1da177e4 | 1041 | |
6a6d27de AV |
1042 | /* |
1043 | * These functions flushes out all traces of the currently running executable | |
1044 | * so that a new one can be started | |
1045 | */ | |
1046 | ||
82b89778 | 1047 | void __set_task_comm(struct task_struct *tsk, const char *buf, bool exec) |
1da177e4 LT |
1048 | { |
1049 | task_lock(tsk); | |
43d2b113 | 1050 | trace_task_rename(tsk, buf); |
1da177e4 LT |
1051 | strlcpy(tsk->comm, buf, sizeof(tsk->comm)); |
1052 | task_unlock(tsk); | |
82b89778 | 1053 | perf_event_comm(tsk, exec); |
1da177e4 LT |
1054 | } |
1055 | ||
1056 | int flush_old_exec(struct linux_binprm * bprm) | |
1057 | { | |
221af7f8 | 1058 | int retval; |
1da177e4 LT |
1059 | |
1060 | /* | |
1061 | * Make sure we have a private signal table and that | |
1062 | * we are unassociated from the previous thread group. | |
1063 | */ | |
1064 | retval = de_thread(current); | |
1065 | if (retval) | |
1066 | goto out; | |
1067 | ||
925d1c40 | 1068 | set_mm_exe_file(bprm->mm, bprm->file); |
1da177e4 LT |
1069 | /* |
1070 | * Release all of the old mmap stuff | |
1071 | */ | |
3c77f845 | 1072 | acct_arg_size(bprm, 0); |
1da177e4 LT |
1073 | retval = exec_mmap(bprm->mm); |
1074 | if (retval) | |
fd8328be | 1075 | goto out; |
1da177e4 LT |
1076 | |
1077 | bprm->mm = NULL; /* We're using it now */ | |
7ab02af4 | 1078 | |
dac853ae | 1079 | set_fs(USER_DS); |
b88fae64 ZY |
1080 | current->flags &= ~(PF_RANDOMIZE | PF_FORKNOEXEC | PF_KTHREAD | |
1081 | PF_NOFREEZE | PF_NO_SETAFFINITY); | |
7ab02af4 LT |
1082 | flush_thread(); |
1083 | current->personality &= ~bprm->per_clear; | |
1084 | ||
221af7f8 LT |
1085 | return 0; |
1086 | ||
1087 | out: | |
1088 | return retval; | |
1089 | } | |
1090 | EXPORT_SYMBOL(flush_old_exec); | |
1091 | ||
1b5d783c AV |
1092 | void would_dump(struct linux_binprm *bprm, struct file *file) |
1093 | { | |
496ad9aa | 1094 | if (inode_permission(file_inode(file), MAY_READ) < 0) |
1b5d783c AV |
1095 | bprm->interp_flags |= BINPRM_FLAGS_ENFORCE_NONDUMP; |
1096 | } | |
1097 | EXPORT_SYMBOL(would_dump); | |
1098 | ||
221af7f8 LT |
1099 | void setup_new_exec(struct linux_binprm * bprm) |
1100 | { | |
221af7f8 | 1101 | arch_pick_mmap_layout(current->mm); |
1da177e4 LT |
1102 | |
1103 | /* This is the point of no return */ | |
1da177e4 LT |
1104 | current->sas_ss_sp = current->sas_ss_size = 0; |
1105 | ||
8e96e3b7 | 1106 | if (uid_eq(current_euid(), current_uid()) && gid_eq(current_egid(), current_gid())) |
e579d2c2 | 1107 | set_dumpable(current->mm, SUID_DUMP_USER); |
d6e71144 | 1108 | else |
6c5d5238 | 1109 | set_dumpable(current->mm, suid_dumpable); |
d6e71144 | 1110 | |
e041e328 | 1111 | perf_event_exec(); |
82b89778 | 1112 | __set_task_comm(current, kbasename(bprm->filename), true); |
1da177e4 | 1113 | |
0551fbd2 BH |
1114 | /* Set the new mm task size. We have to do that late because it may |
1115 | * depend on TIF_32BIT which is only updated in flush_thread() on | |
1116 | * some architectures like powerpc | |
1117 | */ | |
1118 | current->mm->task_size = TASK_SIZE; | |
1119 | ||
a6f76f23 | 1120 | /* install the new credentials */ |
8e96e3b7 EB |
1121 | if (!uid_eq(bprm->cred->uid, current_euid()) || |
1122 | !gid_eq(bprm->cred->gid, current_egid())) { | |
d2d56c5f | 1123 | current->pdeath_signal = 0; |
1b5d783c AV |
1124 | } else { |
1125 | would_dump(bprm, bprm->file); | |
1126 | if (bprm->interp_flags & BINPRM_FLAGS_ENFORCE_NONDUMP) | |
1127 | set_dumpable(current->mm, suid_dumpable); | |
1da177e4 LT |
1128 | } |
1129 | ||
1130 | /* An exec changes our domain. We are no longer part of the thread | |
1131 | group */ | |
1da177e4 | 1132 | current->self_exec_id++; |
1da177e4 | 1133 | flush_signal_handlers(current, 0); |
6a6d27de | 1134 | do_close_on_exec(current->files); |
1da177e4 | 1135 | } |
221af7f8 | 1136 | EXPORT_SYMBOL(setup_new_exec); |
1da177e4 | 1137 | |
a2a8474c ON |
1138 | /* |
1139 | * Prepare credentials and lock ->cred_guard_mutex. | |
1140 | * install_exec_creds() commits the new creds and drops the lock. | |
1141 | * Or, if exec fails before, free_bprm() should release ->cred and | |
1142 | * and unlock. | |
1143 | */ | |
1144 | int prepare_bprm_creds(struct linux_binprm *bprm) | |
1145 | { | |
9b1bf12d | 1146 | if (mutex_lock_interruptible(¤t->signal->cred_guard_mutex)) |
a2a8474c ON |
1147 | return -ERESTARTNOINTR; |
1148 | ||
1149 | bprm->cred = prepare_exec_creds(); | |
1150 | if (likely(bprm->cred)) | |
1151 | return 0; | |
1152 | ||
9b1bf12d | 1153 | mutex_unlock(¤t->signal->cred_guard_mutex); |
a2a8474c ON |
1154 | return -ENOMEM; |
1155 | } | |
1156 | ||
c4ad8f98 | 1157 | static void free_bprm(struct linux_binprm *bprm) |
a2a8474c ON |
1158 | { |
1159 | free_arg_pages(bprm); | |
1160 | if (bprm->cred) { | |
9b1bf12d | 1161 | mutex_unlock(¤t->signal->cred_guard_mutex); |
a2a8474c ON |
1162 | abort_creds(bprm->cred); |
1163 | } | |
63e46b95 ON |
1164 | if (bprm->file) { |
1165 | allow_write_access(bprm->file); | |
1166 | fput(bprm->file); | |
1167 | } | |
b66c5984 KC |
1168 | /* If a binfmt changed the interp, free it. */ |
1169 | if (bprm->interp != bprm->filename) | |
1170 | kfree(bprm->interp); | |
a2a8474c ON |
1171 | kfree(bprm); |
1172 | } | |
1173 | ||
b66c5984 KC |
1174 | int bprm_change_interp(char *interp, struct linux_binprm *bprm) |
1175 | { | |
1176 | /* If a binfmt changed the interp, free it first. */ | |
1177 | if (bprm->interp != bprm->filename) | |
1178 | kfree(bprm->interp); | |
1179 | bprm->interp = kstrdup(interp, GFP_KERNEL); | |
1180 | if (!bprm->interp) | |
1181 | return -ENOMEM; | |
1182 | return 0; | |
1183 | } | |
1184 | EXPORT_SYMBOL(bprm_change_interp); | |
1185 | ||
a6f76f23 DH |
1186 | /* |
1187 | * install the new credentials for this executable | |
1188 | */ | |
1189 | void install_exec_creds(struct linux_binprm *bprm) | |
1190 | { | |
1191 | security_bprm_committing_creds(bprm); | |
1192 | ||
1193 | commit_creds(bprm->cred); | |
1194 | bprm->cred = NULL; | |
2976b10f SE |
1195 | |
1196 | /* | |
1197 | * Disable monitoring for regular users | |
1198 | * when executing setuid binaries. Must | |
1199 | * wait until new credentials are committed | |
1200 | * by commit_creds() above | |
1201 | */ | |
1202 | if (get_dumpable(current->mm) != SUID_DUMP_USER) | |
1203 | perf_event_exit_task(current); | |
a2a8474c ON |
1204 | /* |
1205 | * cred_guard_mutex must be held at least to this point to prevent | |
a6f76f23 | 1206 | * ptrace_attach() from altering our determination of the task's |
a2a8474c ON |
1207 | * credentials; any time after this it may be unlocked. |
1208 | */ | |
a6f76f23 | 1209 | security_bprm_committed_creds(bprm); |
9b1bf12d | 1210 | mutex_unlock(¤t->signal->cred_guard_mutex); |
a6f76f23 DH |
1211 | } |
1212 | EXPORT_SYMBOL(install_exec_creds); | |
1213 | ||
1214 | /* | |
1215 | * determine how safe it is to execute the proposed program | |
9b1bf12d | 1216 | * - the caller must hold ->cred_guard_mutex to protect against |
c2e1f2e3 | 1217 | * PTRACE_ATTACH or seccomp thread-sync |
a6f76f23 | 1218 | */ |
9e00cdb0 | 1219 | static void check_unsafe_exec(struct linux_binprm *bprm) |
a6f76f23 | 1220 | { |
0bf2f3ae | 1221 | struct task_struct *p = current, *t; |
f1191b50 | 1222 | unsigned n_fs; |
a6f76f23 | 1223 | |
4b9d33e6 TH |
1224 | if (p->ptrace) { |
1225 | if (p->ptrace & PT_PTRACE_CAP) | |
1226 | bprm->unsafe |= LSM_UNSAFE_PTRACE_CAP; | |
1227 | else | |
1228 | bprm->unsafe |= LSM_UNSAFE_PTRACE; | |
1229 | } | |
a6f76f23 | 1230 | |
259e5e6c AL |
1231 | /* |
1232 | * This isn't strictly necessary, but it makes it harder for LSMs to | |
1233 | * mess up. | |
1234 | */ | |
1d4457f9 | 1235 | if (task_no_new_privs(current)) |
259e5e6c AL |
1236 | bprm->unsafe |= LSM_UNSAFE_NO_NEW_PRIVS; |
1237 | ||
83f62a2e | 1238 | t = p; |
0bf2f3ae | 1239 | n_fs = 1; |
2a4419b5 | 1240 | spin_lock(&p->fs->lock); |
437f7fdb | 1241 | rcu_read_lock(); |
83f62a2e | 1242 | while_each_thread(p, t) { |
0bf2f3ae DH |
1243 | if (t->fs == p->fs) |
1244 | n_fs++; | |
0bf2f3ae | 1245 | } |
437f7fdb | 1246 | rcu_read_unlock(); |
0bf2f3ae | 1247 | |
9e00cdb0 | 1248 | if (p->fs->users > n_fs) |
a6f76f23 | 1249 | bprm->unsafe |= LSM_UNSAFE_SHARE; |
9e00cdb0 ON |
1250 | else |
1251 | p->fs->in_exec = 1; | |
2a4419b5 | 1252 | spin_unlock(&p->fs->lock); |
a6f76f23 DH |
1253 | } |
1254 | ||
9e00cdb0 ON |
1255 | /* |
1256 | * Fill the binprm structure from the inode. | |
1da177e4 | 1257 | * Check permissions, then read the first 128 (BINPRM_BUF_SIZE) bytes |
a6f76f23 DH |
1258 | * |
1259 | * This may be called multiple times for binary chains (scripts for example). | |
1da177e4 LT |
1260 | */ |
1261 | int prepare_binprm(struct linux_binprm *bprm) | |
1262 | { | |
72c2d531 AV |
1263 | struct inode *inode = file_inode(bprm->file); |
1264 | umode_t mode = inode->i_mode; | |
1da177e4 LT |
1265 | int retval; |
1266 | ||
1da177e4 | 1267 | |
a6f76f23 DH |
1268 | /* clear any previous set[ug]id data from a previous binary */ |
1269 | bprm->cred->euid = current_euid(); | |
1270 | bprm->cred->egid = current_egid(); | |
1da177e4 | 1271 | |
259e5e6c | 1272 | if (!(bprm->file->f_path.mnt->mnt_flags & MNT_NOSUID) && |
1d4457f9 | 1273 | !task_no_new_privs(current) && |
3cdf5b45 EB |
1274 | kuid_has_mapping(bprm->cred->user_ns, inode->i_uid) && |
1275 | kgid_has_mapping(bprm->cred->user_ns, inode->i_gid)) { | |
1da177e4 LT |
1276 | /* Set-uid? */ |
1277 | if (mode & S_ISUID) { | |
a6f76f23 DH |
1278 | bprm->per_clear |= PER_CLEAR_ON_SETID; |
1279 | bprm->cred->euid = inode->i_uid; | |
1da177e4 LT |
1280 | } |
1281 | ||
1282 | /* Set-gid? */ | |
1283 | /* | |
1284 | * If setgid is set but no group execute bit then this | |
1285 | * is a candidate for mandatory locking, not a setgid | |
1286 | * executable. | |
1287 | */ | |
1288 | if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) { | |
a6f76f23 DH |
1289 | bprm->per_clear |= PER_CLEAR_ON_SETID; |
1290 | bprm->cred->egid = inode->i_gid; | |
1da177e4 LT |
1291 | } |
1292 | } | |
1293 | ||
1294 | /* fill in binprm security blob */ | |
a6f76f23 | 1295 | retval = security_bprm_set_creds(bprm); |
1da177e4 LT |
1296 | if (retval) |
1297 | return retval; | |
a6f76f23 | 1298 | bprm->cred_prepared = 1; |
1da177e4 | 1299 | |
a6f76f23 DH |
1300 | memset(bprm->buf, 0, BINPRM_BUF_SIZE); |
1301 | return kernel_read(bprm->file, 0, bprm->buf, BINPRM_BUF_SIZE); | |
1da177e4 LT |
1302 | } |
1303 | ||
1304 | EXPORT_SYMBOL(prepare_binprm); | |
1305 | ||
4fc75ff4 NP |
1306 | /* |
1307 | * Arguments are '\0' separated strings found at the location bprm->p | |
1308 | * points to; chop off the first by relocating brpm->p to right after | |
1309 | * the first '\0' encountered. | |
1310 | */ | |
b6a2fea3 | 1311 | int remove_arg_zero(struct linux_binprm *bprm) |
1da177e4 | 1312 | { |
b6a2fea3 OW |
1313 | int ret = 0; |
1314 | unsigned long offset; | |
1315 | char *kaddr; | |
1316 | struct page *page; | |
4fc75ff4 | 1317 | |
b6a2fea3 OW |
1318 | if (!bprm->argc) |
1319 | return 0; | |
1da177e4 | 1320 | |
b6a2fea3 OW |
1321 | do { |
1322 | offset = bprm->p & ~PAGE_MASK; | |
1323 | page = get_arg_page(bprm, bprm->p, 0); | |
1324 | if (!page) { | |
1325 | ret = -EFAULT; | |
1326 | goto out; | |
1327 | } | |
e8e3c3d6 | 1328 | kaddr = kmap_atomic(page); |
4fc75ff4 | 1329 | |
b6a2fea3 OW |
1330 | for (; offset < PAGE_SIZE && kaddr[offset]; |
1331 | offset++, bprm->p++) | |
1332 | ; | |
4fc75ff4 | 1333 | |
e8e3c3d6 | 1334 | kunmap_atomic(kaddr); |
b6a2fea3 | 1335 | put_arg_page(page); |
4fc75ff4 | 1336 | |
b6a2fea3 OW |
1337 | if (offset == PAGE_SIZE) |
1338 | free_arg_page(bprm, (bprm->p >> PAGE_SHIFT) - 1); | |
1339 | } while (offset == PAGE_SIZE); | |
4fc75ff4 | 1340 | |
b6a2fea3 OW |
1341 | bprm->p++; |
1342 | bprm->argc--; | |
1343 | ret = 0; | |
4fc75ff4 | 1344 | |
b6a2fea3 OW |
1345 | out: |
1346 | return ret; | |
1da177e4 | 1347 | } |
1da177e4 LT |
1348 | EXPORT_SYMBOL(remove_arg_zero); |
1349 | ||
cb7b6b1c | 1350 | #define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e)) |
1da177e4 LT |
1351 | /* |
1352 | * cycle the list of binary formats handler, until one recognizes the image | |
1353 | */ | |
3c456bfc | 1354 | int search_binary_handler(struct linux_binprm *bprm) |
1da177e4 | 1355 | { |
cb7b6b1c | 1356 | bool need_retry = IS_ENABLED(CONFIG_MODULES); |
1da177e4 | 1357 | struct linux_binfmt *fmt; |
cb7b6b1c | 1358 | int retval; |
1da177e4 | 1359 | |
d7402698 | 1360 | /* This allows 4 levels of binfmt rewrites before failing hard. */ |
131b2f9f | 1361 | if (bprm->recursion_depth > 5) |
d7402698 KC |
1362 | return -ELOOP; |
1363 | ||
1da177e4 LT |
1364 | retval = security_bprm_check(bprm); |
1365 | if (retval) | |
1366 | return retval; | |
1367 | ||
1da177e4 | 1368 | retval = -ENOENT; |
cb7b6b1c ON |
1369 | retry: |
1370 | read_lock(&binfmt_lock); | |
1371 | list_for_each_entry(fmt, &formats, lh) { | |
1372 | if (!try_module_get(fmt->module)) | |
1373 | continue; | |
1374 | read_unlock(&binfmt_lock); | |
1375 | bprm->recursion_depth++; | |
1376 | retval = fmt->load_binary(bprm); | |
19d860a1 AV |
1377 | read_lock(&binfmt_lock); |
1378 | put_binfmt(fmt); | |
cb7b6b1c | 1379 | bprm->recursion_depth--; |
19d860a1 AV |
1380 | if (retval < 0 && !bprm->mm) { |
1381 | /* we got to flush_old_exec() and failed after it */ | |
1382 | read_unlock(&binfmt_lock); | |
1383 | force_sigsegv(SIGSEGV, current); | |
1384 | return retval; | |
1385 | } | |
1386 | if (retval != -ENOEXEC || !bprm->file) { | |
1387 | read_unlock(&binfmt_lock); | |
cb7b6b1c | 1388 | return retval; |
1da177e4 | 1389 | } |
1da177e4 | 1390 | } |
cb7b6b1c ON |
1391 | read_unlock(&binfmt_lock); |
1392 | ||
19d860a1 | 1393 | if (need_retry) { |
cb7b6b1c ON |
1394 | if (printable(bprm->buf[0]) && printable(bprm->buf[1]) && |
1395 | printable(bprm->buf[2]) && printable(bprm->buf[3])) | |
1396 | return retval; | |
4e0621a0 ON |
1397 | if (request_module("binfmt-%04x", *(ushort *)(bprm->buf + 2)) < 0) |
1398 | return retval; | |
cb7b6b1c ON |
1399 | need_retry = false; |
1400 | goto retry; | |
1401 | } | |
1402 | ||
1da177e4 LT |
1403 | return retval; |
1404 | } | |
1da177e4 LT |
1405 | EXPORT_SYMBOL(search_binary_handler); |
1406 | ||
5d1baf3b ON |
1407 | static int exec_binprm(struct linux_binprm *bprm) |
1408 | { | |
1409 | pid_t old_pid, old_vpid; | |
1410 | int ret; | |
1411 | ||
1412 | /* Need to fetch pid before load_binary changes it */ | |
1413 | old_pid = current->pid; | |
1414 | rcu_read_lock(); | |
1415 | old_vpid = task_pid_nr_ns(current, task_active_pid_ns(current->parent)); | |
1416 | rcu_read_unlock(); | |
1417 | ||
1418 | ret = search_binary_handler(bprm); | |
1419 | if (ret >= 0) { | |
3eaded86 | 1420 | audit_bprm(bprm); |
5d1baf3b ON |
1421 | trace_sched_process_exec(current, old_pid, bprm); |
1422 | ptrace_event(PTRACE_EVENT_EXEC, old_vpid); | |
9beb266f | 1423 | proc_exec_connector(current); |
5d1baf3b ON |
1424 | } |
1425 | ||
1426 | return ret; | |
1427 | } | |
1428 | ||
1da177e4 LT |
1429 | /* |
1430 | * sys_execve() executes a new program. | |
1431 | */ | |
c4ad8f98 | 1432 | static int do_execve_common(struct filename *filename, |
ba2d0162 | 1433 | struct user_arg_ptr argv, |
835ab32d | 1434 | struct user_arg_ptr envp) |
1da177e4 LT |
1435 | { |
1436 | struct linux_binprm *bprm; | |
1437 | struct file *file; | |
3b125388 | 1438 | struct files_struct *displaced; |
1da177e4 | 1439 | int retval; |
72fa5997 | 1440 | |
c4ad8f98 LT |
1441 | if (IS_ERR(filename)) |
1442 | return PTR_ERR(filename); | |
1443 | ||
72fa5997 VK |
1444 | /* |
1445 | * We move the actual failure in case of RLIMIT_NPROC excess from | |
1446 | * set*uid() to execve() because too many poorly written programs | |
1447 | * don't check setuid() return code. Here we additionally recheck | |
1448 | * whether NPROC limit is still exceeded. | |
1449 | */ | |
1450 | if ((current->flags & PF_NPROC_EXCEEDED) && | |
bd9d43f4 | 1451 | atomic_read(¤t_user()->processes) > rlimit(RLIMIT_NPROC)) { |
72fa5997 VK |
1452 | retval = -EAGAIN; |
1453 | goto out_ret; | |
1454 | } | |
1455 | ||
1456 | /* We're below the limit (still or again), so we don't want to make | |
1457 | * further execve() calls fail. */ | |
1458 | current->flags &= ~PF_NPROC_EXCEEDED; | |
1da177e4 | 1459 | |
3b125388 | 1460 | retval = unshare_files(&displaced); |
fd8328be AV |
1461 | if (retval) |
1462 | goto out_ret; | |
1463 | ||
1da177e4 | 1464 | retval = -ENOMEM; |
11b0b5ab | 1465 | bprm = kzalloc(sizeof(*bprm), GFP_KERNEL); |
1da177e4 | 1466 | if (!bprm) |
fd8328be | 1467 | goto out_files; |
1da177e4 | 1468 | |
a2a8474c ON |
1469 | retval = prepare_bprm_creds(bprm); |
1470 | if (retval) | |
a6f76f23 | 1471 | goto out_free; |
498052bb | 1472 | |
9e00cdb0 | 1473 | check_unsafe_exec(bprm); |
a2a8474c | 1474 | current->in_execve = 1; |
a6f76f23 | 1475 | |
c4ad8f98 | 1476 | file = do_open_exec(filename); |
1da177e4 LT |
1477 | retval = PTR_ERR(file); |
1478 | if (IS_ERR(file)) | |
498052bb | 1479 | goto out_unmark; |
1da177e4 LT |
1480 | |
1481 | sched_exec(); | |
1482 | ||
1da177e4 | 1483 | bprm->file = file; |
c4ad8f98 | 1484 | bprm->filename = bprm->interp = filename->name; |
1da177e4 | 1485 | |
b6a2fea3 OW |
1486 | retval = bprm_mm_init(bprm); |
1487 | if (retval) | |
63e46b95 | 1488 | goto out_unmark; |
1da177e4 | 1489 | |
b6a2fea3 | 1490 | bprm->argc = count(argv, MAX_ARG_STRINGS); |
1da177e4 | 1491 | if ((retval = bprm->argc) < 0) |
a6f76f23 | 1492 | goto out; |
1da177e4 | 1493 | |
b6a2fea3 | 1494 | bprm->envc = count(envp, MAX_ARG_STRINGS); |
1da177e4 | 1495 | if ((retval = bprm->envc) < 0) |
1da177e4 LT |
1496 | goto out; |
1497 | ||
1498 | retval = prepare_binprm(bprm); | |
1499 | if (retval < 0) | |
1500 | goto out; | |
1501 | ||
1502 | retval = copy_strings_kernel(1, &bprm->filename, bprm); | |
1503 | if (retval < 0) | |
1504 | goto out; | |
1505 | ||
1506 | bprm->exec = bprm->p; | |
1507 | retval = copy_strings(bprm->envc, envp, bprm); | |
1508 | if (retval < 0) | |
1509 | goto out; | |
1510 | ||
1511 | retval = copy_strings(bprm->argc, argv, bprm); | |
1512 | if (retval < 0) | |
1513 | goto out; | |
1514 | ||
5d1baf3b | 1515 | retval = exec_binprm(bprm); |
a6f76f23 DH |
1516 | if (retval < 0) |
1517 | goto out; | |
1da177e4 | 1518 | |
a6f76f23 | 1519 | /* execve succeeded */ |
498052bb | 1520 | current->fs->in_exec = 0; |
f9ce1f1c | 1521 | current->in_execve = 0; |
a6f76f23 | 1522 | acct_update_integrals(current); |
82727018 | 1523 | task_numa_free(current); |
a6f76f23 | 1524 | free_bprm(bprm); |
c4ad8f98 | 1525 | putname(filename); |
a6f76f23 DH |
1526 | if (displaced) |
1527 | put_files_struct(displaced); | |
1528 | return retval; | |
1da177e4 | 1529 | |
a6f76f23 | 1530 | out: |
3c77f845 ON |
1531 | if (bprm->mm) { |
1532 | acct_arg_size(bprm, 0); | |
1533 | mmput(bprm->mm); | |
1534 | } | |
1da177e4 | 1535 | |
498052bb | 1536 | out_unmark: |
9e00cdb0 | 1537 | current->fs->in_exec = 0; |
f9ce1f1c | 1538 | current->in_execve = 0; |
a6f76f23 DH |
1539 | |
1540 | out_free: | |
08a6fac1 | 1541 | free_bprm(bprm); |
1da177e4 | 1542 | |
fd8328be | 1543 | out_files: |
3b125388 AV |
1544 | if (displaced) |
1545 | reset_files_struct(displaced); | |
1da177e4 | 1546 | out_ret: |
c4ad8f98 | 1547 | putname(filename); |
1da177e4 LT |
1548 | return retval; |
1549 | } | |
1550 | ||
c4ad8f98 | 1551 | int do_execve(struct filename *filename, |
ba2d0162 | 1552 | const char __user *const __user *__argv, |
da3d4c5f | 1553 | const char __user *const __user *__envp) |
ba2d0162 | 1554 | { |
0e028465 ON |
1555 | struct user_arg_ptr argv = { .ptr.native = __argv }; |
1556 | struct user_arg_ptr envp = { .ptr.native = __envp }; | |
835ab32d | 1557 | return do_execve_common(filename, argv, envp); |
0e028465 ON |
1558 | } |
1559 | ||
1560 | #ifdef CONFIG_COMPAT | |
c4ad8f98 | 1561 | static int compat_do_execve(struct filename *filename, |
38b983b3 | 1562 | const compat_uptr_t __user *__argv, |
d03d26e5 | 1563 | const compat_uptr_t __user *__envp) |
0e028465 ON |
1564 | { |
1565 | struct user_arg_ptr argv = { | |
1566 | .is_compat = true, | |
1567 | .ptr.compat = __argv, | |
1568 | }; | |
1569 | struct user_arg_ptr envp = { | |
1570 | .is_compat = true, | |
1571 | .ptr.compat = __envp, | |
1572 | }; | |
835ab32d | 1573 | return do_execve_common(filename, argv, envp); |
ba2d0162 | 1574 | } |
0e028465 | 1575 | #endif |
ba2d0162 | 1576 | |
964ee7df | 1577 | void set_binfmt(struct linux_binfmt *new) |
1da177e4 | 1578 | { |
801460d0 HS |
1579 | struct mm_struct *mm = current->mm; |
1580 | ||
1581 | if (mm->binfmt) | |
1582 | module_put(mm->binfmt->module); | |
1da177e4 | 1583 | |
801460d0 | 1584 | mm->binfmt = new; |
964ee7df ON |
1585 | if (new) |
1586 | __module_get(new->module); | |
1da177e4 | 1587 | } |
1da177e4 LT |
1588 | EXPORT_SYMBOL(set_binfmt); |
1589 | ||
6c5d5238 | 1590 | /* |
7288e118 | 1591 | * set_dumpable stores three-value SUID_DUMP_* into mm->flags. |
6c5d5238 KH |
1592 | */ |
1593 | void set_dumpable(struct mm_struct *mm, int value) | |
1594 | { | |
abacd2fe ON |
1595 | unsigned long old, new; |
1596 | ||
7288e118 ON |
1597 | if (WARN_ON((unsigned)value > SUID_DUMP_ROOT)) |
1598 | return; | |
1599 | ||
abacd2fe ON |
1600 | do { |
1601 | old = ACCESS_ONCE(mm->flags); | |
7288e118 | 1602 | new = (old & ~MMF_DUMPABLE_MASK) | value; |
abacd2fe | 1603 | } while (cmpxchg(&mm->flags, old, new) != old); |
6c5d5238 | 1604 | } |
6c5d5238 | 1605 | |
38b983b3 AV |
1606 | SYSCALL_DEFINE3(execve, |
1607 | const char __user *, filename, | |
1608 | const char __user *const __user *, argv, | |
1609 | const char __user *const __user *, envp) | |
1610 | { | |
c4ad8f98 | 1611 | return do_execve(getname(filename), argv, envp); |
38b983b3 AV |
1612 | } |
1613 | #ifdef CONFIG_COMPAT | |
625b1d7e HC |
1614 | COMPAT_SYSCALL_DEFINE3(execve, const char __user *, filename, |
1615 | const compat_uptr_t __user *, argv, | |
1616 | const compat_uptr_t __user *, envp) | |
38b983b3 | 1617 | { |
c4ad8f98 | 1618 | return compat_do_execve(getname(filename), argv, envp); |
38b983b3 AV |
1619 | } |
1620 | #endif |