]>
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 | |
613cc2b6 | 22 | * formats. |
1da177e4 LT |
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> |
6e84f315 | 35 | #include <linux/sched/mm.h> |
f7ccbae4 | 36 | #include <linux/sched/coredump.h> |
3f07c014 | 37 | #include <linux/sched/signal.h> |
6a3827d7 | 38 | #include <linux/sched/numa_balancing.h> |
29930025 | 39 | #include <linux/sched/task.h> |
ca5b172b | 40 | #include <linux/pagemap.h> |
cdd6c482 | 41 | #include <linux/perf_event.h> |
1da177e4 LT |
42 | #include <linux/highmem.h> |
43 | #include <linux/spinlock.h> | |
44 | #include <linux/key.h> | |
45 | #include <linux/personality.h> | |
46 | #include <linux/binfmts.h> | |
1da177e4 | 47 | #include <linux/utsname.h> |
84d73786 | 48 | #include <linux/pid_namespace.h> |
1da177e4 LT |
49 | #include <linux/module.h> |
50 | #include <linux/namei.h> | |
1da177e4 LT |
51 | #include <linux/mount.h> |
52 | #include <linux/security.h> | |
53 | #include <linux/syscalls.h> | |
8f0ab514 | 54 | #include <linux/tsacct_kern.h> |
9f46080c | 55 | #include <linux/cn_proc.h> |
473ae30b | 56 | #include <linux/audit.h> |
6341c393 | 57 | #include <linux/tracehook.h> |
5f4123be | 58 | #include <linux/kmod.h> |
6110e3ab | 59 | #include <linux/fsnotify.h> |
5ad4e53b | 60 | #include <linux/fs_struct.h> |
61be228a | 61 | #include <linux/pipe_fs_i.h> |
3d5992d2 | 62 | #include <linux/oom.h> |
0e028465 | 63 | #include <linux/compat.h> |
b44a7dfc | 64 | #include <linux/vmalloc.h> |
1da177e4 | 65 | |
7c0f6ba6 | 66 | #include <linux/uaccess.h> |
1da177e4 | 67 | #include <asm/mmu_context.h> |
b6a2fea3 | 68 | #include <asm/tlb.h> |
43d2b113 KH |
69 | |
70 | #include <trace/events/task.h> | |
a6f76f23 | 71 | #include "internal.h" |
1da177e4 | 72 | |
4ff16c25 DS |
73 | #include <trace/events/sched.h> |
74 | ||
d6e71144 AC |
75 | int suid_dumpable = 0; |
76 | ||
e4dc1b14 | 77 | static LIST_HEAD(formats); |
1da177e4 LT |
78 | static DEFINE_RWLOCK(binfmt_lock); |
79 | ||
8fc3dc5a | 80 | void __register_binfmt(struct linux_binfmt * fmt, int insert) |
1da177e4 | 81 | { |
8fc3dc5a | 82 | BUG_ON(!fmt); |
92eaa565 ON |
83 | if (WARN_ON(!fmt->load_binary)) |
84 | return; | |
1da177e4 | 85 | write_lock(&binfmt_lock); |
74641f58 IK |
86 | insert ? list_add(&fmt->lh, &formats) : |
87 | list_add_tail(&fmt->lh, &formats); | |
1da177e4 | 88 | write_unlock(&binfmt_lock); |
1da177e4 LT |
89 | } |
90 | ||
74641f58 | 91 | EXPORT_SYMBOL(__register_binfmt); |
1da177e4 | 92 | |
f6b450d4 | 93 | void unregister_binfmt(struct linux_binfmt * fmt) |
1da177e4 | 94 | { |
1da177e4 | 95 | write_lock(&binfmt_lock); |
e4dc1b14 | 96 | list_del(&fmt->lh); |
1da177e4 | 97 | write_unlock(&binfmt_lock); |
1da177e4 LT |
98 | } |
99 | ||
100 | EXPORT_SYMBOL(unregister_binfmt); | |
101 | ||
102 | static inline void put_binfmt(struct linux_binfmt * fmt) | |
103 | { | |
104 | module_put(fmt->module); | |
105 | } | |
106 | ||
90f8572b EB |
107 | bool path_noexec(const struct path *path) |
108 | { | |
109 | return (path->mnt->mnt_flags & MNT_NOEXEC) || | |
110 | (path->mnt->mnt_sb->s_iflags & SB_I_NOEXEC); | |
111 | } | |
112 | ||
69369a70 | 113 | #ifdef CONFIG_USELIB |
1da177e4 LT |
114 | /* |
115 | * Note that a shared library must be both readable and executable due to | |
116 | * security reasons. | |
117 | * | |
118 | * Also note that we take the address to load from from the file itself. | |
119 | */ | |
1e7bfb21 | 120 | SYSCALL_DEFINE1(uselib, const char __user *, library) |
1da177e4 | 121 | { |
72c2d531 | 122 | struct linux_binfmt *fmt; |
964bd183 | 123 | struct file *file; |
91a27b2a | 124 | struct filename *tmp = getname(library); |
964bd183 | 125 | int error = PTR_ERR(tmp); |
47c805dc AV |
126 | static const struct open_flags uselib_flags = { |
127 | .open_flag = O_LARGEFILE | O_RDONLY | __FMODE_EXEC, | |
62fb4a15 | 128 | .acc_mode = MAY_READ | MAY_EXEC, |
f9652e10 AV |
129 | .intent = LOOKUP_OPEN, |
130 | .lookup_flags = LOOKUP_FOLLOW, | |
47c805dc | 131 | }; |
964bd183 | 132 | |
6e8341a1 AV |
133 | if (IS_ERR(tmp)) |
134 | goto out; | |
135 | ||
f9652e10 | 136 | file = do_filp_open(AT_FDCWD, tmp, &uselib_flags); |
6e8341a1 AV |
137 | putname(tmp); |
138 | error = PTR_ERR(file); | |
139 | if (IS_ERR(file)) | |
1da177e4 LT |
140 | goto out; |
141 | ||
142 | error = -EINVAL; | |
496ad9aa | 143 | if (!S_ISREG(file_inode(file)->i_mode)) |
1da177e4 LT |
144 | goto exit; |
145 | ||
30524472 | 146 | error = -EACCES; |
90f8572b | 147 | if (path_noexec(&file->f_path)) |
1da177e4 LT |
148 | goto exit; |
149 | ||
2a12a9d7 | 150 | fsnotify_open(file); |
6110e3ab | 151 | |
1da177e4 | 152 | error = -ENOEXEC; |
1da177e4 | 153 | |
72c2d531 AV |
154 | read_lock(&binfmt_lock); |
155 | list_for_each_entry(fmt, &formats, lh) { | |
156 | if (!fmt->load_shlib) | |
157 | continue; | |
158 | if (!try_module_get(fmt->module)) | |
159 | continue; | |
1da177e4 | 160 | read_unlock(&binfmt_lock); |
72c2d531 AV |
161 | error = fmt->load_shlib(file); |
162 | read_lock(&binfmt_lock); | |
163 | put_binfmt(fmt); | |
164 | if (error != -ENOEXEC) | |
165 | break; | |
1da177e4 | 166 | } |
72c2d531 | 167 | read_unlock(&binfmt_lock); |
6e8341a1 | 168 | exit: |
1da177e4 LT |
169 | fput(file); |
170 | out: | |
171 | return error; | |
1da177e4 | 172 | } |
69369a70 | 173 | #endif /* #ifdef CONFIG_USELIB */ |
1da177e4 | 174 | |
b6a2fea3 | 175 | #ifdef CONFIG_MMU |
ae6b585e ON |
176 | /* |
177 | * The nascent bprm->mm is not visible until exec_mmap() but it can | |
178 | * use a lot of memory, account these pages in current->mm temporary | |
179 | * for oom_badness()->get_mm_rss(). Once exec succeeds or fails, we | |
180 | * change the counter back via acct_arg_size(0). | |
181 | */ | |
0e028465 | 182 | static void acct_arg_size(struct linux_binprm *bprm, unsigned long pages) |
3c77f845 ON |
183 | { |
184 | struct mm_struct *mm = current->mm; | |
185 | long diff = (long)(pages - bprm->vma_pages); | |
186 | ||
187 | if (!mm || !diff) | |
188 | return; | |
189 | ||
190 | bprm->vma_pages = pages; | |
3c77f845 | 191 | add_mm_counter(mm, MM_ANONPAGES, diff); |
3c77f845 ON |
192 | } |
193 | ||
0e028465 | 194 | static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos, |
b6a2fea3 OW |
195 | int write) |
196 | { | |
197 | struct page *page; | |
198 | int ret; | |
9beae1ea | 199 | unsigned int gup_flags = FOLL_FORCE; |
b6a2fea3 OW |
200 | |
201 | #ifdef CONFIG_STACK_GROWSUP | |
202 | if (write) { | |
d05f3169 | 203 | ret = expand_downwards(bprm->vma, pos); |
b6a2fea3 OW |
204 | if (ret < 0) |
205 | return NULL; | |
206 | } | |
207 | #endif | |
9beae1ea LS |
208 | |
209 | if (write) | |
210 | gup_flags |= FOLL_WRITE; | |
211 | ||
1e987790 DH |
212 | /* |
213 | * We are doing an exec(). 'current' is the process | |
214 | * doing the exec and bprm->mm is the new process's mm. | |
215 | */ | |
9beae1ea | 216 | ret = get_user_pages_remote(current, bprm->mm, pos, 1, gup_flags, |
5b56d49f | 217 | &page, NULL, NULL); |
b6a2fea3 OW |
218 | if (ret <= 0) |
219 | return NULL; | |
220 | ||
221 | if (write) { | |
b6a2fea3 | 222 | unsigned long size = bprm->vma->vm_end - bprm->vma->vm_start; |
98da7d08 | 223 | unsigned long ptr_size; |
a64e715f LT |
224 | struct rlimit *rlim; |
225 | ||
98da7d08 KC |
226 | /* |
227 | * Since the stack will hold pointers to the strings, we | |
228 | * must account for them as well. | |
229 | * | |
230 | * The size calculation is the entire vma while each arg page is | |
231 | * built, so each time we get here it's calculating how far it | |
232 | * is currently (rather than each call being just the newly | |
233 | * added size from the arg page). As a result, we need to | |
234 | * always add the entire size of the pointers, so that on the | |
235 | * last call to get_arg_page() we'll actually have the entire | |
236 | * correct size. | |
237 | */ | |
238 | ptr_size = (bprm->argc + bprm->envc) * sizeof(void *); | |
239 | if (ptr_size > ULONG_MAX - size) | |
240 | goto fail; | |
241 | size += ptr_size; | |
242 | ||
3c77f845 ON |
243 | acct_arg_size(bprm, size / PAGE_SIZE); |
244 | ||
a64e715f LT |
245 | /* |
246 | * We've historically supported up to 32 pages (ARG_MAX) | |
247 | * of argument strings even with small stacks | |
248 | */ | |
249 | if (size <= ARG_MAX) | |
250 | return page; | |
b6a2fea3 OW |
251 | |
252 | /* | |
253 | * Limit to 1/4-th the stack size for the argv+env strings. | |
254 | * This ensures that: | |
255 | * - the remaining binfmt code will not run out of stack space, | |
256 | * - the program will have a reasonable amount of stack left | |
257 | * to work from. | |
258 | */ | |
a64e715f | 259 | rlim = current->signal->rlim; |
98da7d08 KC |
260 | if (size > READ_ONCE(rlim[RLIMIT_STACK].rlim_cur) / 4) |
261 | goto fail; | |
b6a2fea3 OW |
262 | } |
263 | ||
264 | return page; | |
98da7d08 KC |
265 | |
266 | fail: | |
267 | put_page(page); | |
268 | return NULL; | |
b6a2fea3 OW |
269 | } |
270 | ||
271 | static void put_arg_page(struct page *page) | |
272 | { | |
273 | put_page(page); | |
274 | } | |
275 | ||
b6a2fea3 OW |
276 | static void free_arg_pages(struct linux_binprm *bprm) |
277 | { | |
278 | } | |
279 | ||
280 | static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos, | |
281 | struct page *page) | |
282 | { | |
283 | flush_cache_page(bprm->vma, pos, page_to_pfn(page)); | |
284 | } | |
285 | ||
286 | static int __bprm_mm_init(struct linux_binprm *bprm) | |
287 | { | |
eaccbfa5 | 288 | int err; |
b6a2fea3 OW |
289 | struct vm_area_struct *vma = NULL; |
290 | struct mm_struct *mm = bprm->mm; | |
291 | ||
292 | bprm->vma = vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL); | |
293 | if (!vma) | |
eaccbfa5 | 294 | return -ENOMEM; |
b6a2fea3 | 295 | |
f268dfe9 MH |
296 | if (down_write_killable(&mm->mmap_sem)) { |
297 | err = -EINTR; | |
298 | goto err_free; | |
299 | } | |
b6a2fea3 OW |
300 | vma->vm_mm = mm; |
301 | ||
302 | /* | |
303 | * Place the stack at the largest stack address the architecture | |
304 | * supports. Later, we'll move this to an appropriate place. We don't | |
305 | * use STACK_TOP because that can depend on attributes which aren't | |
306 | * configured yet. | |
307 | */ | |
aacb3d17 | 308 | BUILD_BUG_ON(VM_STACK_FLAGS & VM_STACK_INCOMPLETE_SETUP); |
b6a2fea3 OW |
309 | vma->vm_end = STACK_TOP_MAX; |
310 | vma->vm_start = vma->vm_end - PAGE_SIZE; | |
d9104d1c | 311 | vma->vm_flags = VM_SOFTDIRTY | VM_STACK_FLAGS | VM_STACK_INCOMPLETE_SETUP; |
3ed75eb8 | 312 | vma->vm_page_prot = vm_get_page_prot(vma->vm_flags); |
5beb4930 | 313 | INIT_LIST_HEAD(&vma->anon_vma_chain); |
462e635e | 314 | |
b6a2fea3 | 315 | err = insert_vm_struct(mm, vma); |
eaccbfa5 | 316 | if (err) |
b6a2fea3 | 317 | goto err; |
b6a2fea3 OW |
318 | |
319 | mm->stack_vm = mm->total_vm = 1; | |
fe3d197f | 320 | arch_bprm_mm_init(mm, vma); |
b6a2fea3 | 321 | up_write(&mm->mmap_sem); |
b6a2fea3 | 322 | bprm->p = vma->vm_end - sizeof(void *); |
b6a2fea3 | 323 | return 0; |
b6a2fea3 | 324 | err: |
eaccbfa5 | 325 | up_write(&mm->mmap_sem); |
f268dfe9 | 326 | err_free: |
eaccbfa5 LFC |
327 | bprm->vma = NULL; |
328 | kmem_cache_free(vm_area_cachep, vma); | |
b6a2fea3 OW |
329 | return err; |
330 | } | |
331 | ||
332 | static bool valid_arg_len(struct linux_binprm *bprm, long len) | |
333 | { | |
334 | return len <= MAX_ARG_STRLEN; | |
335 | } | |
336 | ||
337 | #else | |
338 | ||
0e028465 | 339 | static inline void acct_arg_size(struct linux_binprm *bprm, unsigned long pages) |
3c77f845 ON |
340 | { |
341 | } | |
342 | ||
0e028465 | 343 | static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos, |
b6a2fea3 OW |
344 | int write) |
345 | { | |
346 | struct page *page; | |
347 | ||
348 | page = bprm->page[pos / PAGE_SIZE]; | |
349 | if (!page && write) { | |
350 | page = alloc_page(GFP_HIGHUSER|__GFP_ZERO); | |
351 | if (!page) | |
352 | return NULL; | |
353 | bprm->page[pos / PAGE_SIZE] = page; | |
354 | } | |
355 | ||
356 | return page; | |
357 | } | |
358 | ||
359 | static void put_arg_page(struct page *page) | |
360 | { | |
361 | } | |
362 | ||
363 | static void free_arg_page(struct linux_binprm *bprm, int i) | |
364 | { | |
365 | if (bprm->page[i]) { | |
366 | __free_page(bprm->page[i]); | |
367 | bprm->page[i] = NULL; | |
368 | } | |
369 | } | |
370 | ||
371 | static void free_arg_pages(struct linux_binprm *bprm) | |
372 | { | |
373 | int i; | |
374 | ||
375 | for (i = 0; i < MAX_ARG_PAGES; i++) | |
376 | free_arg_page(bprm, i); | |
377 | } | |
378 | ||
379 | static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos, | |
380 | struct page *page) | |
381 | { | |
382 | } | |
383 | ||
384 | static int __bprm_mm_init(struct linux_binprm *bprm) | |
385 | { | |
386 | bprm->p = PAGE_SIZE * MAX_ARG_PAGES - sizeof(void *); | |
387 | return 0; | |
388 | } | |
389 | ||
390 | static bool valid_arg_len(struct linux_binprm *bprm, long len) | |
391 | { | |
392 | return len <= bprm->p; | |
393 | } | |
394 | ||
395 | #endif /* CONFIG_MMU */ | |
396 | ||
397 | /* | |
398 | * Create a new mm_struct and populate it with a temporary stack | |
399 | * vm_area_struct. We don't have enough context at this point to set the stack | |
400 | * flags, permissions, and offset, so we use temporary values. We'll update | |
401 | * them later in setup_arg_pages(). | |
402 | */ | |
9cc64cea | 403 | static int bprm_mm_init(struct linux_binprm *bprm) |
b6a2fea3 OW |
404 | { |
405 | int err; | |
406 | struct mm_struct *mm = NULL; | |
407 | ||
408 | bprm->mm = mm = mm_alloc(); | |
409 | err = -ENOMEM; | |
410 | if (!mm) | |
411 | goto err; | |
412 | ||
b6a2fea3 OW |
413 | err = __bprm_mm_init(bprm); |
414 | if (err) | |
415 | goto err; | |
416 | ||
417 | return 0; | |
418 | ||
419 | err: | |
420 | if (mm) { | |
421 | bprm->mm = NULL; | |
422 | mmdrop(mm); | |
423 | } | |
424 | ||
425 | return err; | |
426 | } | |
427 | ||
ba2d0162 | 428 | struct user_arg_ptr { |
0e028465 ON |
429 | #ifdef CONFIG_COMPAT |
430 | bool is_compat; | |
431 | #endif | |
432 | union { | |
433 | const char __user *const __user *native; | |
434 | #ifdef CONFIG_COMPAT | |
38b983b3 | 435 | const compat_uptr_t __user *compat; |
0e028465 ON |
436 | #endif |
437 | } ptr; | |
ba2d0162 ON |
438 | }; |
439 | ||
440 | static const char __user *get_user_arg_ptr(struct user_arg_ptr argv, int nr) | |
1d1dbf81 | 441 | { |
0e028465 ON |
442 | const char __user *native; |
443 | ||
444 | #ifdef CONFIG_COMPAT | |
445 | if (unlikely(argv.is_compat)) { | |
446 | compat_uptr_t compat; | |
447 | ||
448 | if (get_user(compat, argv.ptr.compat + nr)) | |
449 | return ERR_PTR(-EFAULT); | |
1d1dbf81 | 450 | |
0e028465 ON |
451 | return compat_ptr(compat); |
452 | } | |
453 | #endif | |
454 | ||
455 | if (get_user(native, argv.ptr.native + nr)) | |
1d1dbf81 ON |
456 | return ERR_PTR(-EFAULT); |
457 | ||
0e028465 | 458 | return native; |
1d1dbf81 ON |
459 | } |
460 | ||
1da177e4 LT |
461 | /* |
462 | * count() counts the number of strings in array ARGV. | |
463 | */ | |
ba2d0162 | 464 | static int count(struct user_arg_ptr argv, int max) |
1da177e4 LT |
465 | { |
466 | int i = 0; | |
467 | ||
0e028465 | 468 | if (argv.ptr.native != NULL) { |
1da177e4 | 469 | for (;;) { |
1d1dbf81 | 470 | const char __user *p = get_user_arg_ptr(argv, i); |
1da177e4 | 471 | |
1da177e4 LT |
472 | if (!p) |
473 | break; | |
1d1dbf81 ON |
474 | |
475 | if (IS_ERR(p)) | |
476 | return -EFAULT; | |
477 | ||
6d92d4f6 | 478 | if (i >= max) |
1da177e4 | 479 | return -E2BIG; |
6d92d4f6 | 480 | ++i; |
9aea5a65 RM |
481 | |
482 | if (fatal_signal_pending(current)) | |
483 | return -ERESTARTNOHAND; | |
1da177e4 LT |
484 | cond_resched(); |
485 | } | |
486 | } | |
487 | return i; | |
488 | } | |
489 | ||
490 | /* | |
b6a2fea3 OW |
491 | * 'copy_strings()' copies argument/environment strings from the old |
492 | * processes's memory to the new process's stack. The call to get_user_pages() | |
493 | * ensures the destination page is created and not swapped out. | |
1da177e4 | 494 | */ |
ba2d0162 | 495 | static int copy_strings(int argc, struct user_arg_ptr argv, |
75c96f85 | 496 | struct linux_binprm *bprm) |
1da177e4 LT |
497 | { |
498 | struct page *kmapped_page = NULL; | |
499 | char *kaddr = NULL; | |
b6a2fea3 | 500 | unsigned long kpos = 0; |
1da177e4 LT |
501 | int ret; |
502 | ||
503 | while (argc-- > 0) { | |
d7627467 | 504 | const char __user *str; |
1da177e4 LT |
505 | int len; |
506 | unsigned long pos; | |
507 | ||
1d1dbf81 ON |
508 | ret = -EFAULT; |
509 | str = get_user_arg_ptr(argv, argc); | |
510 | if (IS_ERR(str)) | |
1da177e4 | 511 | goto out; |
1da177e4 | 512 | |
1d1dbf81 ON |
513 | len = strnlen_user(str, MAX_ARG_STRLEN); |
514 | if (!len) | |
515 | goto out; | |
516 | ||
517 | ret = -E2BIG; | |
518 | if (!valid_arg_len(bprm, len)) | |
1da177e4 | 519 | goto out; |
1da177e4 | 520 | |
b6a2fea3 | 521 | /* We're going to work our way backwords. */ |
1da177e4 | 522 | pos = bprm->p; |
b6a2fea3 OW |
523 | str += len; |
524 | bprm->p -= len; | |
1da177e4 LT |
525 | |
526 | while (len > 0) { | |
1da177e4 | 527 | int offset, bytes_to_copy; |
1da177e4 | 528 | |
9aea5a65 RM |
529 | if (fatal_signal_pending(current)) { |
530 | ret = -ERESTARTNOHAND; | |
531 | goto out; | |
532 | } | |
7993bc1f RM |
533 | cond_resched(); |
534 | ||
1da177e4 | 535 | offset = pos % PAGE_SIZE; |
b6a2fea3 OW |
536 | if (offset == 0) |
537 | offset = PAGE_SIZE; | |
538 | ||
539 | bytes_to_copy = offset; | |
540 | if (bytes_to_copy > len) | |
541 | bytes_to_copy = len; | |
542 | ||
543 | offset -= bytes_to_copy; | |
544 | pos -= bytes_to_copy; | |
545 | str -= bytes_to_copy; | |
546 | len -= bytes_to_copy; | |
547 | ||
548 | if (!kmapped_page || kpos != (pos & PAGE_MASK)) { | |
549 | struct page *page; | |
550 | ||
551 | page = get_arg_page(bprm, pos, 1); | |
1da177e4 | 552 | if (!page) { |
b6a2fea3 | 553 | ret = -E2BIG; |
1da177e4 LT |
554 | goto out; |
555 | } | |
1da177e4 | 556 | |
b6a2fea3 OW |
557 | if (kmapped_page) { |
558 | flush_kernel_dcache_page(kmapped_page); | |
1da177e4 | 559 | kunmap(kmapped_page); |
b6a2fea3 OW |
560 | put_arg_page(kmapped_page); |
561 | } | |
1da177e4 LT |
562 | kmapped_page = page; |
563 | kaddr = kmap(kmapped_page); | |
b6a2fea3 OW |
564 | kpos = pos & PAGE_MASK; |
565 | flush_arg_page(bprm, kpos, kmapped_page); | |
1da177e4 | 566 | } |
b6a2fea3 | 567 | if (copy_from_user(kaddr+offset, str, bytes_to_copy)) { |
1da177e4 LT |
568 | ret = -EFAULT; |
569 | goto out; | |
570 | } | |
1da177e4 LT |
571 | } |
572 | } | |
573 | ret = 0; | |
574 | out: | |
b6a2fea3 OW |
575 | if (kmapped_page) { |
576 | flush_kernel_dcache_page(kmapped_page); | |
1da177e4 | 577 | kunmap(kmapped_page); |
b6a2fea3 OW |
578 | put_arg_page(kmapped_page); |
579 | } | |
1da177e4 LT |
580 | return ret; |
581 | } | |
582 | ||
583 | /* | |
584 | * Like copy_strings, but get argv and its values from kernel memory. | |
585 | */ | |
ba2d0162 | 586 | int copy_strings_kernel(int argc, const char *const *__argv, |
d7627467 | 587 | struct linux_binprm *bprm) |
1da177e4 LT |
588 | { |
589 | int r; | |
590 | mm_segment_t oldfs = get_fs(); | |
ba2d0162 | 591 | struct user_arg_ptr argv = { |
0e028465 | 592 | .ptr.native = (const char __user *const __user *)__argv, |
ba2d0162 ON |
593 | }; |
594 | ||
1da177e4 | 595 | set_fs(KERNEL_DS); |
ba2d0162 | 596 | r = copy_strings(argc, argv, bprm); |
1da177e4 | 597 | set_fs(oldfs); |
ba2d0162 | 598 | |
1da177e4 LT |
599 | return r; |
600 | } | |
1da177e4 LT |
601 | EXPORT_SYMBOL(copy_strings_kernel); |
602 | ||
603 | #ifdef CONFIG_MMU | |
b6a2fea3 | 604 | |
1da177e4 | 605 | /* |
b6a2fea3 OW |
606 | * During bprm_mm_init(), we create a temporary stack at STACK_TOP_MAX. Once |
607 | * the binfmt code determines where the new stack should reside, we shift it to | |
608 | * its final location. The process proceeds as follows: | |
1da177e4 | 609 | * |
b6a2fea3 OW |
610 | * 1) Use shift to calculate the new vma endpoints. |
611 | * 2) Extend vma to cover both the old and new ranges. This ensures the | |
612 | * arguments passed to subsequent functions are consistent. | |
613 | * 3) Move vma's page tables to the new range. | |
614 | * 4) Free up any cleared pgd range. | |
615 | * 5) Shrink the vma to cover only the new range. | |
1da177e4 | 616 | */ |
b6a2fea3 | 617 | static int shift_arg_pages(struct vm_area_struct *vma, unsigned long shift) |
1da177e4 LT |
618 | { |
619 | struct mm_struct *mm = vma->vm_mm; | |
b6a2fea3 OW |
620 | unsigned long old_start = vma->vm_start; |
621 | unsigned long old_end = vma->vm_end; | |
622 | unsigned long length = old_end - old_start; | |
623 | unsigned long new_start = old_start - shift; | |
624 | unsigned long new_end = old_end - shift; | |
d16dfc55 | 625 | struct mmu_gather tlb; |
1da177e4 | 626 | |
b6a2fea3 | 627 | BUG_ON(new_start > new_end); |
1da177e4 | 628 | |
b6a2fea3 OW |
629 | /* |
630 | * ensure there are no vmas between where we want to go | |
631 | * and where we are | |
632 | */ | |
633 | if (vma != find_vma(mm, new_start)) | |
634 | return -EFAULT; | |
635 | ||
636 | /* | |
637 | * cover the whole range: [new_start, old_end) | |
638 | */ | |
5beb4930 RR |
639 | if (vma_adjust(vma, new_start, old_end, vma->vm_pgoff, NULL)) |
640 | return -ENOMEM; | |
b6a2fea3 OW |
641 | |
642 | /* | |
643 | * move the page tables downwards, on failure we rely on | |
644 | * process cleanup to remove whatever mess we made. | |
645 | */ | |
646 | if (length != move_page_tables(vma, old_start, | |
38a76013 | 647 | vma, new_start, length, false)) |
b6a2fea3 OW |
648 | return -ENOMEM; |
649 | ||
650 | lru_add_drain(); | |
2b047252 | 651 | tlb_gather_mmu(&tlb, mm, old_start, old_end); |
b6a2fea3 OW |
652 | if (new_end > old_start) { |
653 | /* | |
654 | * when the old and new regions overlap clear from new_end. | |
655 | */ | |
d16dfc55 | 656 | free_pgd_range(&tlb, new_end, old_end, new_end, |
6ee8630e | 657 | vma->vm_next ? vma->vm_next->vm_start : USER_PGTABLES_CEILING); |
b6a2fea3 OW |
658 | } else { |
659 | /* | |
660 | * otherwise, clean from old_start; this is done to not touch | |
661 | * the address space in [new_end, old_start) some architectures | |
662 | * have constraints on va-space that make this illegal (IA64) - | |
663 | * for the others its just a little faster. | |
664 | */ | |
d16dfc55 | 665 | free_pgd_range(&tlb, old_start, old_end, new_end, |
6ee8630e | 666 | vma->vm_next ? vma->vm_next->vm_start : USER_PGTABLES_CEILING); |
1da177e4 | 667 | } |
2b047252 | 668 | tlb_finish_mmu(&tlb, old_start, old_end); |
b6a2fea3 OW |
669 | |
670 | /* | |
5beb4930 | 671 | * Shrink the vma to just the new range. Always succeeds. |
b6a2fea3 OW |
672 | */ |
673 | vma_adjust(vma, new_start, new_end, vma->vm_pgoff, NULL); | |
674 | ||
675 | return 0; | |
1da177e4 LT |
676 | } |
677 | ||
b6a2fea3 OW |
678 | /* |
679 | * Finalizes the stack vm_area_struct. The flags and permissions are updated, | |
680 | * the stack is optionally relocated, and some extra space is added. | |
681 | */ | |
1da177e4 LT |
682 | int setup_arg_pages(struct linux_binprm *bprm, |
683 | unsigned long stack_top, | |
684 | int executable_stack) | |
685 | { | |
b6a2fea3 OW |
686 | unsigned long ret; |
687 | unsigned long stack_shift; | |
1da177e4 | 688 | struct mm_struct *mm = current->mm; |
b6a2fea3 OW |
689 | struct vm_area_struct *vma = bprm->vma; |
690 | struct vm_area_struct *prev = NULL; | |
691 | unsigned long vm_flags; | |
692 | unsigned long stack_base; | |
803bf5ec MN |
693 | unsigned long stack_size; |
694 | unsigned long stack_expand; | |
695 | unsigned long rlim_stack; | |
1da177e4 LT |
696 | |
697 | #ifdef CONFIG_STACK_GROWSUP | |
d71f290b | 698 | /* Limit stack size */ |
d554ed89 | 699 | stack_base = rlimit_max(RLIMIT_STACK); |
d71f290b JH |
700 | if (stack_base > STACK_SIZE_MAX) |
701 | stack_base = STACK_SIZE_MAX; | |
1da177e4 | 702 | |
d045c77c HD |
703 | /* Add space for stack randomization. */ |
704 | stack_base += (STACK_RND_MASK << PAGE_SHIFT); | |
705 | ||
b6a2fea3 OW |
706 | /* Make sure we didn't let the argument array grow too large. */ |
707 | if (vma->vm_end - vma->vm_start > stack_base) | |
708 | return -ENOMEM; | |
1da177e4 | 709 | |
b6a2fea3 | 710 | stack_base = PAGE_ALIGN(stack_top - stack_base); |
1da177e4 | 711 | |
b6a2fea3 OW |
712 | stack_shift = vma->vm_start - stack_base; |
713 | mm->arg_start = bprm->p - stack_shift; | |
714 | bprm->p = vma->vm_end - stack_shift; | |
1da177e4 | 715 | #else |
b6a2fea3 OW |
716 | stack_top = arch_align_stack(stack_top); |
717 | stack_top = PAGE_ALIGN(stack_top); | |
1b528181 RM |
718 | |
719 | if (unlikely(stack_top < mmap_min_addr) || | |
720 | unlikely(vma->vm_end - vma->vm_start >= stack_top - mmap_min_addr)) | |
721 | return -ENOMEM; | |
722 | ||
b6a2fea3 OW |
723 | stack_shift = vma->vm_end - stack_top; |
724 | ||
725 | bprm->p -= stack_shift; | |
1da177e4 | 726 | mm->arg_start = bprm->p; |
1da177e4 LT |
727 | #endif |
728 | ||
1da177e4 | 729 | if (bprm->loader) |
b6a2fea3 OW |
730 | bprm->loader -= stack_shift; |
731 | bprm->exec -= stack_shift; | |
1da177e4 | 732 | |
f268dfe9 MH |
733 | if (down_write_killable(&mm->mmap_sem)) |
734 | return -EINTR; | |
735 | ||
96a8e13e | 736 | vm_flags = VM_STACK_FLAGS; |
b6a2fea3 OW |
737 | |
738 | /* | |
739 | * Adjust stack execute permissions; explicitly enable for | |
740 | * EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X and leave alone | |
741 | * (arch default) otherwise. | |
742 | */ | |
743 | if (unlikely(executable_stack == EXSTACK_ENABLE_X)) | |
744 | vm_flags |= VM_EXEC; | |
745 | else if (executable_stack == EXSTACK_DISABLE_X) | |
746 | vm_flags &= ~VM_EXEC; | |
747 | vm_flags |= mm->def_flags; | |
a8bef8ff | 748 | vm_flags |= VM_STACK_INCOMPLETE_SETUP; |
b6a2fea3 OW |
749 | |
750 | ret = mprotect_fixup(vma, &prev, vma->vm_start, vma->vm_end, | |
751 | vm_flags); | |
752 | if (ret) | |
753 | goto out_unlock; | |
754 | BUG_ON(prev != vma); | |
755 | ||
756 | /* Move stack pages down in memory. */ | |
757 | if (stack_shift) { | |
758 | ret = shift_arg_pages(vma, stack_shift); | |
fc63cf23 AB |
759 | if (ret) |
760 | goto out_unlock; | |
1da177e4 LT |
761 | } |
762 | ||
a8bef8ff MG |
763 | /* mprotect_fixup is overkill to remove the temporary stack flags */ |
764 | vma->vm_flags &= ~VM_STACK_INCOMPLETE_SETUP; | |
765 | ||
5ef097dd | 766 | stack_expand = 131072UL; /* randomly 32*4k (or 2*64k) pages */ |
803bf5ec MN |
767 | stack_size = vma->vm_end - vma->vm_start; |
768 | /* | |
769 | * Align this down to a page boundary as expand_stack | |
770 | * will align it up. | |
771 | */ | |
772 | rlim_stack = rlimit(RLIMIT_STACK) & PAGE_MASK; | |
b6a2fea3 | 773 | #ifdef CONFIG_STACK_GROWSUP |
803bf5ec MN |
774 | if (stack_size + stack_expand > rlim_stack) |
775 | stack_base = vma->vm_start + rlim_stack; | |
776 | else | |
777 | stack_base = vma->vm_end + stack_expand; | |
b6a2fea3 | 778 | #else |
803bf5ec MN |
779 | if (stack_size + stack_expand > rlim_stack) |
780 | stack_base = vma->vm_end - rlim_stack; | |
781 | else | |
782 | stack_base = vma->vm_start - stack_expand; | |
b6a2fea3 | 783 | #endif |
3af9e859 | 784 | current->mm->start_stack = bprm->p; |
b6a2fea3 OW |
785 | ret = expand_stack(vma, stack_base); |
786 | if (ret) | |
787 | ret = -EFAULT; | |
788 | ||
789 | out_unlock: | |
1da177e4 | 790 | up_write(&mm->mmap_sem); |
fc63cf23 | 791 | return ret; |
1da177e4 | 792 | } |
1da177e4 LT |
793 | EXPORT_SYMBOL(setup_arg_pages); |
794 | ||
7e7ec6a9 NP |
795 | #else |
796 | ||
797 | /* | |
798 | * Transfer the program arguments and environment from the holding pages | |
799 | * onto the stack. The provided stack pointer is adjusted accordingly. | |
800 | */ | |
801 | int transfer_args_to_stack(struct linux_binprm *bprm, | |
802 | unsigned long *sp_location) | |
803 | { | |
804 | unsigned long index, stop, sp; | |
805 | int ret = 0; | |
806 | ||
807 | stop = bprm->p >> PAGE_SHIFT; | |
808 | sp = *sp_location; | |
809 | ||
810 | for (index = MAX_ARG_PAGES - 1; index >= stop; index--) { | |
811 | unsigned int offset = index == stop ? bprm->p & ~PAGE_MASK : 0; | |
812 | char *src = kmap(bprm->page[index]) + offset; | |
813 | sp -= PAGE_SIZE - offset; | |
814 | if (copy_to_user((void *) sp, src, PAGE_SIZE - offset) != 0) | |
815 | ret = -EFAULT; | |
816 | kunmap(bprm->page[index]); | |
817 | if (ret) | |
818 | goto out; | |
819 | } | |
820 | ||
821 | *sp_location = sp; | |
822 | ||
823 | out: | |
824 | return ret; | |
825 | } | |
826 | EXPORT_SYMBOL(transfer_args_to_stack); | |
827 | ||
1da177e4 LT |
828 | #endif /* CONFIG_MMU */ |
829 | ||
51f39a1f | 830 | static struct file *do_open_execat(int fd, struct filename *name, int flags) |
1da177e4 | 831 | { |
1da177e4 | 832 | struct file *file; |
e56b6a5d | 833 | int err; |
51f39a1f | 834 | struct open_flags open_exec_flags = { |
47c805dc | 835 | .open_flag = O_LARGEFILE | O_RDONLY | __FMODE_EXEC, |
62fb4a15 | 836 | .acc_mode = MAY_EXEC, |
f9652e10 AV |
837 | .intent = LOOKUP_OPEN, |
838 | .lookup_flags = LOOKUP_FOLLOW, | |
47c805dc | 839 | }; |
1da177e4 | 840 | |
51f39a1f DD |
841 | if ((flags & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0) |
842 | return ERR_PTR(-EINVAL); | |
843 | if (flags & AT_SYMLINK_NOFOLLOW) | |
844 | open_exec_flags.lookup_flags &= ~LOOKUP_FOLLOW; | |
845 | if (flags & AT_EMPTY_PATH) | |
846 | open_exec_flags.lookup_flags |= LOOKUP_EMPTY; | |
847 | ||
848 | file = do_filp_open(fd, name, &open_exec_flags); | |
6e8341a1 | 849 | if (IS_ERR(file)) |
e56b6a5d CH |
850 | goto out; |
851 | ||
852 | err = -EACCES; | |
496ad9aa | 853 | if (!S_ISREG(file_inode(file)->i_mode)) |
6e8341a1 | 854 | goto exit; |
e56b6a5d | 855 | |
90f8572b | 856 | if (path_noexec(&file->f_path)) |
6e8341a1 | 857 | goto exit; |
e56b6a5d CH |
858 | |
859 | err = deny_write_access(file); | |
6e8341a1 AV |
860 | if (err) |
861 | goto exit; | |
1da177e4 | 862 | |
51f39a1f DD |
863 | if (name->name[0] != '\0') |
864 | fsnotify_open(file); | |
865 | ||
6e8341a1 | 866 | out: |
e56b6a5d CH |
867 | return file; |
868 | ||
6e8341a1 AV |
869 | exit: |
870 | fput(file); | |
e56b6a5d CH |
871 | return ERR_PTR(err); |
872 | } | |
c4ad8f98 LT |
873 | |
874 | struct file *open_exec(const char *name) | |
875 | { | |
51689104 PM |
876 | struct filename *filename = getname_kernel(name); |
877 | struct file *f = ERR_CAST(filename); | |
878 | ||
879 | if (!IS_ERR(filename)) { | |
880 | f = do_open_execat(AT_FDCWD, filename, 0); | |
881 | putname(filename); | |
882 | } | |
883 | return f; | |
c4ad8f98 | 884 | } |
1da177e4 LT |
885 | EXPORT_SYMBOL(open_exec); |
886 | ||
6777d773 MZ |
887 | int kernel_read(struct file *file, loff_t offset, |
888 | char *addr, unsigned long count) | |
1da177e4 LT |
889 | { |
890 | mm_segment_t old_fs; | |
891 | loff_t pos = offset; | |
892 | int result; | |
893 | ||
894 | old_fs = get_fs(); | |
895 | set_fs(get_ds()); | |
896 | /* The cast to a user pointer is valid due to the set_fs() */ | |
897 | result = vfs_read(file, (void __user *)addr, count, &pos); | |
898 | set_fs(old_fs); | |
899 | return result; | |
900 | } | |
901 | ||
902 | EXPORT_SYMBOL(kernel_read); | |
903 | ||
b44a7dfc | 904 | int kernel_read_file(struct file *file, void **buf, loff_t *size, |
bc8ca5b9 | 905 | loff_t max_size, enum kernel_read_file_id id) |
b44a7dfc MZ |
906 | { |
907 | loff_t i_size, pos; | |
908 | ssize_t bytes = 0; | |
909 | int ret; | |
910 | ||
911 | if (!S_ISREG(file_inode(file)->i_mode) || max_size < 0) | |
912 | return -EINVAL; | |
913 | ||
39eeb4fb MZ |
914 | ret = security_kernel_read_file(file, id); |
915 | if (ret) | |
916 | return ret; | |
917 | ||
39d637af DK |
918 | ret = deny_write_access(file); |
919 | if (ret) | |
920 | return ret; | |
921 | ||
b44a7dfc | 922 | i_size = i_size_read(file_inode(file)); |
39d637af DK |
923 | if (max_size > 0 && i_size > max_size) { |
924 | ret = -EFBIG; | |
925 | goto out; | |
926 | } | |
927 | if (i_size <= 0) { | |
928 | ret = -EINVAL; | |
929 | goto out; | |
930 | } | |
b44a7dfc | 931 | |
a098ecd2 SB |
932 | if (id != READING_FIRMWARE_PREALLOC_BUFFER) |
933 | *buf = vmalloc(i_size); | |
39d637af DK |
934 | if (!*buf) { |
935 | ret = -ENOMEM; | |
936 | goto out; | |
937 | } | |
b44a7dfc MZ |
938 | |
939 | pos = 0; | |
940 | while (pos < i_size) { | |
941 | bytes = kernel_read(file, pos, (char *)(*buf) + pos, | |
942 | i_size - pos); | |
943 | if (bytes < 0) { | |
944 | ret = bytes; | |
945 | goto out; | |
946 | } | |
947 | ||
948 | if (bytes == 0) | |
949 | break; | |
950 | pos += bytes; | |
951 | } | |
952 | ||
953 | if (pos != i_size) { | |
954 | ret = -EIO; | |
39d637af | 955 | goto out_free; |
b44a7dfc MZ |
956 | } |
957 | ||
bc8ca5b9 | 958 | ret = security_kernel_post_read_file(file, *buf, i_size, id); |
b44a7dfc MZ |
959 | if (!ret) |
960 | *size = pos; | |
961 | ||
39d637af | 962 | out_free: |
b44a7dfc | 963 | if (ret < 0) { |
a098ecd2 SB |
964 | if (id != READING_FIRMWARE_PREALLOC_BUFFER) { |
965 | vfree(*buf); | |
966 | *buf = NULL; | |
967 | } | |
b44a7dfc | 968 | } |
39d637af DK |
969 | |
970 | out: | |
971 | allow_write_access(file); | |
b44a7dfc MZ |
972 | return ret; |
973 | } | |
974 | EXPORT_SYMBOL_GPL(kernel_read_file); | |
975 | ||
09596b94 MZ |
976 | int kernel_read_file_from_path(char *path, void **buf, loff_t *size, |
977 | loff_t max_size, enum kernel_read_file_id id) | |
978 | { | |
979 | struct file *file; | |
980 | int ret; | |
981 | ||
982 | if (!path || !*path) | |
983 | return -EINVAL; | |
984 | ||
985 | file = filp_open(path, O_RDONLY, 0); | |
986 | if (IS_ERR(file)) | |
987 | return PTR_ERR(file); | |
988 | ||
989 | ret = kernel_read_file(file, buf, size, max_size, id); | |
990 | fput(file); | |
991 | return ret; | |
992 | } | |
993 | EXPORT_SYMBOL_GPL(kernel_read_file_from_path); | |
994 | ||
b844f0ec MZ |
995 | int kernel_read_file_from_fd(int fd, void **buf, loff_t *size, loff_t max_size, |
996 | enum kernel_read_file_id id) | |
997 | { | |
998 | struct fd f = fdget(fd); | |
999 | int ret = -EBADF; | |
1000 | ||
1001 | if (!f.file) | |
1002 | goto out; | |
1003 | ||
1004 | ret = kernel_read_file(f.file, buf, size, max_size, id); | |
1005 | out: | |
1006 | fdput(f); | |
1007 | return ret; | |
1008 | } | |
1009 | EXPORT_SYMBOL_GPL(kernel_read_file_from_fd); | |
1010 | ||
3dc20cb2 AV |
1011 | ssize_t read_code(struct file *file, unsigned long addr, loff_t pos, size_t len) |
1012 | { | |
ec695579 | 1013 | ssize_t res = vfs_read(file, (void __user *)addr, len, &pos); |
3dc20cb2 AV |
1014 | if (res > 0) |
1015 | flush_icache_range(addr, addr + len); | |
1016 | return res; | |
1017 | } | |
1018 | EXPORT_SYMBOL(read_code); | |
1019 | ||
1da177e4 LT |
1020 | static int exec_mmap(struct mm_struct *mm) |
1021 | { | |
1022 | struct task_struct *tsk; | |
615d6e87 | 1023 | struct mm_struct *old_mm, *active_mm; |
1da177e4 LT |
1024 | |
1025 | /* Notify parent that we're no longer interested in the old VM */ | |
1026 | tsk = current; | |
1027 | old_mm = current->mm; | |
1028 | mm_release(tsk, old_mm); | |
1029 | ||
1030 | if (old_mm) { | |
4fe7efdb | 1031 | sync_mm_rss(old_mm); |
1da177e4 LT |
1032 | /* |
1033 | * Make sure that if there is a core dump in progress | |
1034 | * for the old mm, we get out and die instead of going | |
1035 | * through with the exec. We must hold mmap_sem around | |
999d9fc1 | 1036 | * checking core_state and changing tsk->mm. |
1da177e4 LT |
1037 | */ |
1038 | down_read(&old_mm->mmap_sem); | |
999d9fc1 | 1039 | if (unlikely(old_mm->core_state)) { |
1da177e4 LT |
1040 | up_read(&old_mm->mmap_sem); |
1041 | return -EINTR; | |
1042 | } | |
1043 | } | |
1044 | task_lock(tsk); | |
1045 | active_mm = tsk->active_mm; | |
1046 | tsk->mm = mm; | |
1047 | tsk->active_mm = mm; | |
1048 | activate_mm(active_mm, mm); | |
615d6e87 DB |
1049 | tsk->mm->vmacache_seqnum = 0; |
1050 | vmacache_flush(tsk); | |
1da177e4 | 1051 | task_unlock(tsk); |
1da177e4 LT |
1052 | if (old_mm) { |
1053 | up_read(&old_mm->mmap_sem); | |
7dddb12c | 1054 | BUG_ON(active_mm != old_mm); |
701085b2 | 1055 | setmax_mm_hiwater_rss(&tsk->signal->maxrss, old_mm); |
31a78f23 | 1056 | mm_update_next_owner(old_mm); |
1da177e4 LT |
1057 | mmput(old_mm); |
1058 | return 0; | |
1059 | } | |
1060 | mmdrop(active_mm); | |
1061 | return 0; | |
1062 | } | |
1063 | ||
1064 | /* | |
1065 | * This function makes sure the current process has its own signal table, | |
1066 | * so that flush_signal_handlers can later reset the handlers without | |
1067 | * disturbing other processes. (Other processes might share the signal | |
1068 | * table via the CLONE_SIGHAND option to clone().) | |
1069 | */ | |
858119e1 | 1070 | static int de_thread(struct task_struct *tsk) |
1da177e4 LT |
1071 | { |
1072 | struct signal_struct *sig = tsk->signal; | |
b2c903b8 | 1073 | struct sighand_struct *oldsighand = tsk->sighand; |
1da177e4 | 1074 | spinlock_t *lock = &oldsighand->siglock; |
1da177e4 | 1075 | |
aafe6c2a | 1076 | if (thread_group_empty(tsk)) |
1da177e4 LT |
1077 | goto no_thread_group; |
1078 | ||
1079 | /* | |
1080 | * Kill all other threads in the thread group. | |
1da177e4 | 1081 | */ |
1da177e4 | 1082 | spin_lock_irq(lock); |
ed5d2cac | 1083 | if (signal_group_exit(sig)) { |
1da177e4 LT |
1084 | /* |
1085 | * Another group action in progress, just | |
1086 | * return so that the signal is processed. | |
1087 | */ | |
1088 | spin_unlock_irq(lock); | |
1da177e4 LT |
1089 | return -EAGAIN; |
1090 | } | |
d344193a | 1091 | |
ed5d2cac | 1092 | sig->group_exit_task = tsk; |
d344193a ON |
1093 | sig->notify_count = zap_other_threads(tsk); |
1094 | if (!thread_group_leader(tsk)) | |
1095 | sig->notify_count--; | |
1da177e4 | 1096 | |
d344193a | 1097 | while (sig->notify_count) { |
d5bbd43d | 1098 | __set_current_state(TASK_KILLABLE); |
1da177e4 LT |
1099 | spin_unlock_irq(lock); |
1100 | schedule(); | |
d5bbd43d ON |
1101 | if (unlikely(__fatal_signal_pending(tsk))) |
1102 | goto killed; | |
1da177e4 LT |
1103 | spin_lock_irq(lock); |
1104 | } | |
1da177e4 LT |
1105 | spin_unlock_irq(lock); |
1106 | ||
1107 | /* | |
1108 | * At this point all other threads have exited, all we have to | |
1109 | * do is to wait for the thread group leader to become inactive, | |
1110 | * and to assume its PID: | |
1111 | */ | |
aafe6c2a | 1112 | if (!thread_group_leader(tsk)) { |
8187926b | 1113 | struct task_struct *leader = tsk->group_leader; |
6db840fa | 1114 | |
6db840fa | 1115 | for (;;) { |
780de9dd | 1116 | cgroup_threadgroup_change_begin(tsk); |
6db840fa | 1117 | write_lock_irq(&tasklist_lock); |
dfcce791 KT |
1118 | /* |
1119 | * Do this under tasklist_lock to ensure that | |
1120 | * exit_notify() can't miss ->group_exit_task | |
1121 | */ | |
1122 | sig->notify_count = -1; | |
6db840fa ON |
1123 | if (likely(leader->exit_state)) |
1124 | break; | |
d5bbd43d | 1125 | __set_current_state(TASK_KILLABLE); |
6db840fa | 1126 | write_unlock_irq(&tasklist_lock); |
780de9dd | 1127 | cgroup_threadgroup_change_end(tsk); |
6db840fa | 1128 | schedule(); |
d5bbd43d ON |
1129 | if (unlikely(__fatal_signal_pending(tsk))) |
1130 | goto killed; | |
6db840fa | 1131 | } |
1da177e4 | 1132 | |
f5e90281 RM |
1133 | /* |
1134 | * The only record we have of the real-time age of a | |
1135 | * process, regardless of execs it's done, is start_time. | |
1136 | * All the past CPU time is accumulated in signal_struct | |
1137 | * from sister threads now dead. But in this non-leader | |
1138 | * exec, nothing survives from the original leader thread, | |
1139 | * whose birth marks the true age of this process now. | |
1140 | * When we take on its identity by switching to its PID, we | |
1141 | * also take its birthdate (always earlier than our own). | |
1142 | */ | |
aafe6c2a | 1143 | tsk->start_time = leader->start_time; |
266b7a02 | 1144 | tsk->real_start_time = leader->real_start_time; |
f5e90281 | 1145 | |
bac0abd6 PE |
1146 | BUG_ON(!same_thread_group(leader, tsk)); |
1147 | BUG_ON(has_group_leader_pid(tsk)); | |
1da177e4 LT |
1148 | /* |
1149 | * An exec() starts a new thread group with the | |
1150 | * TGID of the previous thread group. Rehash the | |
1151 | * two threads with a switched PID, and release | |
1152 | * the former thread group leader: | |
1153 | */ | |
d73d6529 EB |
1154 | |
1155 | /* Become a process group leader with the old leader's pid. | |
c18258c6 EB |
1156 | * The old leader becomes a thread of the this thread group. |
1157 | * Note: The old leader also uses this pid until release_task | |
d73d6529 EB |
1158 | * is called. Odd but simple and correct. |
1159 | */ | |
aafe6c2a | 1160 | tsk->pid = leader->pid; |
3f418548 | 1161 | change_pid(tsk, PIDTYPE_PID, task_pid(leader)); |
aafe6c2a EB |
1162 | transfer_pid(leader, tsk, PIDTYPE_PGID); |
1163 | transfer_pid(leader, tsk, PIDTYPE_SID); | |
9cd80bbb | 1164 | |
aafe6c2a | 1165 | list_replace_rcu(&leader->tasks, &tsk->tasks); |
9cd80bbb | 1166 | list_replace_init(&leader->sibling, &tsk->sibling); |
1da177e4 | 1167 | |
aafe6c2a EB |
1168 | tsk->group_leader = tsk; |
1169 | leader->group_leader = tsk; | |
de12a787 | 1170 | |
aafe6c2a | 1171 | tsk->exit_signal = SIGCHLD; |
087806b1 | 1172 | leader->exit_signal = -1; |
962b564c ON |
1173 | |
1174 | BUG_ON(leader->exit_state != EXIT_ZOMBIE); | |
1175 | leader->exit_state = EXIT_DEAD; | |
eac1b5e5 ON |
1176 | |
1177 | /* | |
1178 | * We are going to release_task()->ptrace_unlink() silently, | |
1179 | * the tracer can sleep in do_wait(). EXIT_DEAD guarantees | |
1180 | * the tracer wont't block again waiting for this thread. | |
1181 | */ | |
1182 | if (unlikely(leader->ptrace)) | |
1183 | __wake_up_parent(leader, leader->parent); | |
1da177e4 | 1184 | write_unlock_irq(&tasklist_lock); |
780de9dd | 1185 | cgroup_threadgroup_change_end(tsk); |
8187926b ON |
1186 | |
1187 | release_task(leader); | |
ed5d2cac | 1188 | } |
1da177e4 | 1189 | |
6db840fa ON |
1190 | sig->group_exit_task = NULL; |
1191 | sig->notify_count = 0; | |
1da177e4 LT |
1192 | |
1193 | no_thread_group: | |
e6368253 ON |
1194 | /* we have changed execution domain */ |
1195 | tsk->exit_signal = SIGCHLD; | |
1196 | ||
baa73d9e | 1197 | #ifdef CONFIG_POSIX_TIMERS |
1da177e4 | 1198 | exit_itimers(sig); |
cbaffba1 | 1199 | flush_itimer_signals(); |
baa73d9e | 1200 | #endif |
329f7dba | 1201 | |
b2c903b8 ON |
1202 | if (atomic_read(&oldsighand->count) != 1) { |
1203 | struct sighand_struct *newsighand; | |
1da177e4 | 1204 | /* |
b2c903b8 ON |
1205 | * This ->sighand is shared with the CLONE_SIGHAND |
1206 | * but not CLONE_THREAD task, switch to the new one. | |
1da177e4 | 1207 | */ |
b2c903b8 ON |
1208 | newsighand = kmem_cache_alloc(sighand_cachep, GFP_KERNEL); |
1209 | if (!newsighand) | |
1210 | return -ENOMEM; | |
1211 | ||
1da177e4 LT |
1212 | atomic_set(&newsighand->count, 1); |
1213 | memcpy(newsighand->action, oldsighand->action, | |
1214 | sizeof(newsighand->action)); | |
1215 | ||
1216 | write_lock_irq(&tasklist_lock); | |
1217 | spin_lock(&oldsighand->siglock); | |
aafe6c2a | 1218 | rcu_assign_pointer(tsk->sighand, newsighand); |
1da177e4 LT |
1219 | spin_unlock(&oldsighand->siglock); |
1220 | write_unlock_irq(&tasklist_lock); | |
1221 | ||
fba2afaa | 1222 | __cleanup_sighand(oldsighand); |
1da177e4 LT |
1223 | } |
1224 | ||
aafe6c2a | 1225 | BUG_ON(!thread_group_leader(tsk)); |
1da177e4 | 1226 | return 0; |
d5bbd43d ON |
1227 | |
1228 | killed: | |
1229 | /* protects against exit_notify() and __exit_signal() */ | |
1230 | read_lock(&tasklist_lock); | |
1231 | sig->group_exit_task = NULL; | |
1232 | sig->notify_count = 0; | |
1233 | read_unlock(&tasklist_lock); | |
1234 | return -EAGAIN; | |
1da177e4 | 1235 | } |
0840a90d | 1236 | |
59714d65 | 1237 | char *get_task_comm(char *buf, struct task_struct *tsk) |
1da177e4 LT |
1238 | { |
1239 | /* buf must be at least sizeof(tsk->comm) in size */ | |
1240 | task_lock(tsk); | |
1241 | strncpy(buf, tsk->comm, sizeof(tsk->comm)); | |
1242 | task_unlock(tsk); | |
59714d65 | 1243 | return buf; |
1da177e4 | 1244 | } |
7d74f492 | 1245 | EXPORT_SYMBOL_GPL(get_task_comm); |
1da177e4 | 1246 | |
6a6d27de AV |
1247 | /* |
1248 | * These functions flushes out all traces of the currently running executable | |
1249 | * so that a new one can be started | |
1250 | */ | |
1251 | ||
82b89778 | 1252 | void __set_task_comm(struct task_struct *tsk, const char *buf, bool exec) |
1da177e4 LT |
1253 | { |
1254 | task_lock(tsk); | |
43d2b113 | 1255 | trace_task_rename(tsk, buf); |
1da177e4 LT |
1256 | strlcpy(tsk->comm, buf, sizeof(tsk->comm)); |
1257 | task_unlock(tsk); | |
82b89778 | 1258 | perf_event_comm(tsk, exec); |
1da177e4 LT |
1259 | } |
1260 | ||
1261 | int flush_old_exec(struct linux_binprm * bprm) | |
1262 | { | |
221af7f8 | 1263 | int retval; |
1da177e4 LT |
1264 | |
1265 | /* | |
1266 | * Make sure we have a private signal table and that | |
1267 | * we are unassociated from the previous thread group. | |
1268 | */ | |
1269 | retval = de_thread(current); | |
1270 | if (retval) | |
1271 | goto out; | |
1272 | ||
6e399cd1 DB |
1273 | /* |
1274 | * Must be called _before_ exec_mmap() as bprm->mm is | |
1275 | * not visibile until then. This also enables the update | |
1276 | * to be lockless. | |
1277 | */ | |
925d1c40 | 1278 | set_mm_exe_file(bprm->mm, bprm->file); |
6e399cd1 | 1279 | |
1da177e4 LT |
1280 | /* |
1281 | * Release all of the old mmap stuff | |
1282 | */ | |
3c77f845 | 1283 | acct_arg_size(bprm, 0); |
1da177e4 LT |
1284 | retval = exec_mmap(bprm->mm); |
1285 | if (retval) | |
fd8328be | 1286 | goto out; |
1da177e4 LT |
1287 | |
1288 | bprm->mm = NULL; /* We're using it now */ | |
7ab02af4 | 1289 | |
dac853ae | 1290 | set_fs(USER_DS); |
b88fae64 ZY |
1291 | current->flags &= ~(PF_RANDOMIZE | PF_FORKNOEXEC | PF_KTHREAD | |
1292 | PF_NOFREEZE | PF_NO_SETAFFINITY); | |
7ab02af4 LT |
1293 | flush_thread(); |
1294 | current->personality &= ~bprm->per_clear; | |
1295 | ||
613cc2b6 AS |
1296 | /* |
1297 | * We have to apply CLOEXEC before we change whether the process is | |
1298 | * dumpable (in setup_new_exec) to avoid a race with a process in userspace | |
1299 | * trying to access the should-be-closed file descriptors of a process | |
1300 | * undergoing exec(2). | |
1301 | */ | |
1302 | do_close_on_exec(current->files); | |
221af7f8 LT |
1303 | return 0; |
1304 | ||
1305 | out: | |
1306 | return retval; | |
1307 | } | |
1308 | EXPORT_SYMBOL(flush_old_exec); | |
1309 | ||
1b5d783c AV |
1310 | void would_dump(struct linux_binprm *bprm, struct file *file) |
1311 | { | |
f84df2a6 EB |
1312 | struct inode *inode = file_inode(file); |
1313 | if (inode_permission(inode, MAY_READ) < 0) { | |
1314 | struct user_namespace *old, *user_ns; | |
1b5d783c | 1315 | bprm->interp_flags |= BINPRM_FLAGS_ENFORCE_NONDUMP; |
f84df2a6 EB |
1316 | |
1317 | /* Ensure mm->user_ns contains the executable */ | |
1318 | user_ns = old = bprm->mm->user_ns; | |
1319 | while ((user_ns != &init_user_ns) && | |
1320 | !privileged_wrt_inode_uidgid(user_ns, inode)) | |
1321 | user_ns = user_ns->parent; | |
1322 | ||
1323 | if (old != user_ns) { | |
1324 | bprm->mm->user_ns = get_user_ns(user_ns); | |
1325 | put_user_ns(old); | |
1326 | } | |
1327 | } | |
1b5d783c AV |
1328 | } |
1329 | EXPORT_SYMBOL(would_dump); | |
1330 | ||
221af7f8 LT |
1331 | void setup_new_exec(struct linux_binprm * bprm) |
1332 | { | |
221af7f8 | 1333 | arch_pick_mmap_layout(current->mm); |
1da177e4 LT |
1334 | |
1335 | /* This is the point of no return */ | |
1da177e4 LT |
1336 | current->sas_ss_sp = current->sas_ss_size = 0; |
1337 | ||
8e96e3b7 | 1338 | if (uid_eq(current_euid(), current_uid()) && gid_eq(current_egid(), current_gid())) |
e579d2c2 | 1339 | set_dumpable(current->mm, SUID_DUMP_USER); |
d6e71144 | 1340 | else |
6c5d5238 | 1341 | set_dumpable(current->mm, suid_dumpable); |
d6e71144 | 1342 | |
e9ea1e7f | 1343 | arch_setup_new_exec(); |
e041e328 | 1344 | perf_event_exec(); |
82b89778 | 1345 | __set_task_comm(current, kbasename(bprm->filename), true); |
1da177e4 | 1346 | |
0551fbd2 BH |
1347 | /* Set the new mm task size. We have to do that late because it may |
1348 | * depend on TIF_32BIT which is only updated in flush_thread() on | |
1349 | * some architectures like powerpc | |
1350 | */ | |
1351 | current->mm->task_size = TASK_SIZE; | |
1352 | ||
a6f76f23 | 1353 | /* install the new credentials */ |
8e96e3b7 EB |
1354 | if (!uid_eq(bprm->cred->uid, current_euid()) || |
1355 | !gid_eq(bprm->cred->gid, current_egid())) { | |
d2d56c5f | 1356 | current->pdeath_signal = 0; |
1b5d783c | 1357 | } else { |
1b5d783c AV |
1358 | if (bprm->interp_flags & BINPRM_FLAGS_ENFORCE_NONDUMP) |
1359 | set_dumpable(current->mm, suid_dumpable); | |
1da177e4 LT |
1360 | } |
1361 | ||
1362 | /* An exec changes our domain. We are no longer part of the thread | |
1363 | group */ | |
1da177e4 | 1364 | current->self_exec_id++; |
1da177e4 | 1365 | flush_signal_handlers(current, 0); |
1da177e4 | 1366 | } |
221af7f8 | 1367 | EXPORT_SYMBOL(setup_new_exec); |
1da177e4 | 1368 | |
a2a8474c ON |
1369 | /* |
1370 | * Prepare credentials and lock ->cred_guard_mutex. | |
1371 | * install_exec_creds() commits the new creds and drops the lock. | |
1372 | * Or, if exec fails before, free_bprm() should release ->cred and | |
1373 | * and unlock. | |
1374 | */ | |
1375 | int prepare_bprm_creds(struct linux_binprm *bprm) | |
1376 | { | |
9b1bf12d | 1377 | if (mutex_lock_interruptible(¤t->signal->cred_guard_mutex)) |
a2a8474c ON |
1378 | return -ERESTARTNOINTR; |
1379 | ||
1380 | bprm->cred = prepare_exec_creds(); | |
1381 | if (likely(bprm->cred)) | |
1382 | return 0; | |
1383 | ||
9b1bf12d | 1384 | mutex_unlock(¤t->signal->cred_guard_mutex); |
a2a8474c ON |
1385 | return -ENOMEM; |
1386 | } | |
1387 | ||
c4ad8f98 | 1388 | static void free_bprm(struct linux_binprm *bprm) |
a2a8474c ON |
1389 | { |
1390 | free_arg_pages(bprm); | |
1391 | if (bprm->cred) { | |
9b1bf12d | 1392 | mutex_unlock(¤t->signal->cred_guard_mutex); |
a2a8474c ON |
1393 | abort_creds(bprm->cred); |
1394 | } | |
63e46b95 ON |
1395 | if (bprm->file) { |
1396 | allow_write_access(bprm->file); | |
1397 | fput(bprm->file); | |
1398 | } | |
b66c5984 KC |
1399 | /* If a binfmt changed the interp, free it. */ |
1400 | if (bprm->interp != bprm->filename) | |
1401 | kfree(bprm->interp); | |
a2a8474c ON |
1402 | kfree(bprm); |
1403 | } | |
1404 | ||
b66c5984 KC |
1405 | int bprm_change_interp(char *interp, struct linux_binprm *bprm) |
1406 | { | |
1407 | /* If a binfmt changed the interp, free it first. */ | |
1408 | if (bprm->interp != bprm->filename) | |
1409 | kfree(bprm->interp); | |
1410 | bprm->interp = kstrdup(interp, GFP_KERNEL); | |
1411 | if (!bprm->interp) | |
1412 | return -ENOMEM; | |
1413 | return 0; | |
1414 | } | |
1415 | EXPORT_SYMBOL(bprm_change_interp); | |
1416 | ||
a6f76f23 DH |
1417 | /* |
1418 | * install the new credentials for this executable | |
1419 | */ | |
1420 | void install_exec_creds(struct linux_binprm *bprm) | |
1421 | { | |
1422 | security_bprm_committing_creds(bprm); | |
1423 | ||
1424 | commit_creds(bprm->cred); | |
1425 | bprm->cred = NULL; | |
2976b10f SE |
1426 | |
1427 | /* | |
1428 | * Disable monitoring for regular users | |
1429 | * when executing setuid binaries. Must | |
1430 | * wait until new credentials are committed | |
1431 | * by commit_creds() above | |
1432 | */ | |
1433 | if (get_dumpable(current->mm) != SUID_DUMP_USER) | |
1434 | perf_event_exit_task(current); | |
a2a8474c ON |
1435 | /* |
1436 | * cred_guard_mutex must be held at least to this point to prevent | |
a6f76f23 | 1437 | * ptrace_attach() from altering our determination of the task's |
a2a8474c ON |
1438 | * credentials; any time after this it may be unlocked. |
1439 | */ | |
a6f76f23 | 1440 | security_bprm_committed_creds(bprm); |
9b1bf12d | 1441 | mutex_unlock(¤t->signal->cred_guard_mutex); |
a6f76f23 DH |
1442 | } |
1443 | EXPORT_SYMBOL(install_exec_creds); | |
1444 | ||
1445 | /* | |
1446 | * determine how safe it is to execute the proposed program | |
9b1bf12d | 1447 | * - the caller must hold ->cred_guard_mutex to protect against |
c2e1f2e3 | 1448 | * PTRACE_ATTACH or seccomp thread-sync |
a6f76f23 | 1449 | */ |
9e00cdb0 | 1450 | static void check_unsafe_exec(struct linux_binprm *bprm) |
a6f76f23 | 1451 | { |
0bf2f3ae | 1452 | struct task_struct *p = current, *t; |
f1191b50 | 1453 | unsigned n_fs; |
a6f76f23 | 1454 | |
9227dd2a EB |
1455 | if (p->ptrace) |
1456 | bprm->unsafe |= LSM_UNSAFE_PTRACE; | |
a6f76f23 | 1457 | |
259e5e6c AL |
1458 | /* |
1459 | * This isn't strictly necessary, but it makes it harder for LSMs to | |
1460 | * mess up. | |
1461 | */ | |
1d4457f9 | 1462 | if (task_no_new_privs(current)) |
259e5e6c AL |
1463 | bprm->unsafe |= LSM_UNSAFE_NO_NEW_PRIVS; |
1464 | ||
83f62a2e | 1465 | t = p; |
0bf2f3ae | 1466 | n_fs = 1; |
2a4419b5 | 1467 | spin_lock(&p->fs->lock); |
437f7fdb | 1468 | rcu_read_lock(); |
83f62a2e | 1469 | while_each_thread(p, t) { |
0bf2f3ae DH |
1470 | if (t->fs == p->fs) |
1471 | n_fs++; | |
0bf2f3ae | 1472 | } |
437f7fdb | 1473 | rcu_read_unlock(); |
0bf2f3ae | 1474 | |
9e00cdb0 | 1475 | if (p->fs->users > n_fs) |
a6f76f23 | 1476 | bprm->unsafe |= LSM_UNSAFE_SHARE; |
9e00cdb0 ON |
1477 | else |
1478 | p->fs->in_exec = 1; | |
2a4419b5 | 1479 | spin_unlock(&p->fs->lock); |
a6f76f23 DH |
1480 | } |
1481 | ||
8b01fc86 JH |
1482 | static void bprm_fill_uid(struct linux_binprm *bprm) |
1483 | { | |
1484 | struct inode *inode; | |
1485 | unsigned int mode; | |
1486 | kuid_t uid; | |
1487 | kgid_t gid; | |
1488 | ||
cb6fd68f KC |
1489 | /* |
1490 | * Since this can be called multiple times (via prepare_binprm), | |
1491 | * we must clear any previous work done when setting set[ug]id | |
1492 | * bits from any earlier bprm->file uses (for example when run | |
1493 | * first for a setuid script then again for its interpreter). | |
1494 | */ | |
8b01fc86 JH |
1495 | bprm->cred->euid = current_euid(); |
1496 | bprm->cred->egid = current_egid(); | |
1497 | ||
380cf5ba | 1498 | if (!mnt_may_suid(bprm->file->f_path.mnt)) |
8b01fc86 JH |
1499 | return; |
1500 | ||
1501 | if (task_no_new_privs(current)) | |
1502 | return; | |
1503 | ||
fea6d2a6 | 1504 | inode = bprm->file->f_path.dentry->d_inode; |
8b01fc86 JH |
1505 | mode = READ_ONCE(inode->i_mode); |
1506 | if (!(mode & (S_ISUID|S_ISGID))) | |
1507 | return; | |
1508 | ||
1509 | /* Be careful if suid/sgid is set */ | |
5955102c | 1510 | inode_lock(inode); |
8b01fc86 JH |
1511 | |
1512 | /* reload atomically mode/uid/gid now that lock held */ | |
1513 | mode = inode->i_mode; | |
1514 | uid = inode->i_uid; | |
1515 | gid = inode->i_gid; | |
5955102c | 1516 | inode_unlock(inode); |
8b01fc86 JH |
1517 | |
1518 | /* We ignore suid/sgid if there are no mappings for them in the ns */ | |
1519 | if (!kuid_has_mapping(bprm->cred->user_ns, uid) || | |
1520 | !kgid_has_mapping(bprm->cred->user_ns, gid)) | |
1521 | return; | |
1522 | ||
1523 | if (mode & S_ISUID) { | |
1524 | bprm->per_clear |= PER_CLEAR_ON_SETID; | |
1525 | bprm->cred->euid = uid; | |
1526 | } | |
1527 | ||
1528 | if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) { | |
1529 | bprm->per_clear |= PER_CLEAR_ON_SETID; | |
1530 | bprm->cred->egid = gid; | |
1531 | } | |
1532 | } | |
1533 | ||
9e00cdb0 ON |
1534 | /* |
1535 | * Fill the binprm structure from the inode. | |
1da177e4 | 1536 | * Check permissions, then read the first 128 (BINPRM_BUF_SIZE) bytes |
a6f76f23 DH |
1537 | * |
1538 | * This may be called multiple times for binary chains (scripts for example). | |
1da177e4 LT |
1539 | */ |
1540 | int prepare_binprm(struct linux_binprm *bprm) | |
1541 | { | |
1da177e4 LT |
1542 | int retval; |
1543 | ||
8b01fc86 | 1544 | bprm_fill_uid(bprm); |
1da177e4 LT |
1545 | |
1546 | /* fill in binprm security blob */ | |
a6f76f23 | 1547 | retval = security_bprm_set_creds(bprm); |
1da177e4 LT |
1548 | if (retval) |
1549 | return retval; | |
a6f76f23 | 1550 | bprm->cred_prepared = 1; |
1da177e4 | 1551 | |
a6f76f23 DH |
1552 | memset(bprm->buf, 0, BINPRM_BUF_SIZE); |
1553 | return kernel_read(bprm->file, 0, bprm->buf, BINPRM_BUF_SIZE); | |
1da177e4 LT |
1554 | } |
1555 | ||
1556 | EXPORT_SYMBOL(prepare_binprm); | |
1557 | ||
4fc75ff4 NP |
1558 | /* |
1559 | * Arguments are '\0' separated strings found at the location bprm->p | |
1560 | * points to; chop off the first by relocating brpm->p to right after | |
1561 | * the first '\0' encountered. | |
1562 | */ | |
b6a2fea3 | 1563 | int remove_arg_zero(struct linux_binprm *bprm) |
1da177e4 | 1564 | { |
b6a2fea3 OW |
1565 | int ret = 0; |
1566 | unsigned long offset; | |
1567 | char *kaddr; | |
1568 | struct page *page; | |
4fc75ff4 | 1569 | |
b6a2fea3 OW |
1570 | if (!bprm->argc) |
1571 | return 0; | |
1da177e4 | 1572 | |
b6a2fea3 OW |
1573 | do { |
1574 | offset = bprm->p & ~PAGE_MASK; | |
1575 | page = get_arg_page(bprm, bprm->p, 0); | |
1576 | if (!page) { | |
1577 | ret = -EFAULT; | |
1578 | goto out; | |
1579 | } | |
e8e3c3d6 | 1580 | kaddr = kmap_atomic(page); |
4fc75ff4 | 1581 | |
b6a2fea3 OW |
1582 | for (; offset < PAGE_SIZE && kaddr[offset]; |
1583 | offset++, bprm->p++) | |
1584 | ; | |
4fc75ff4 | 1585 | |
e8e3c3d6 | 1586 | kunmap_atomic(kaddr); |
b6a2fea3 | 1587 | put_arg_page(page); |
b6a2fea3 | 1588 | } while (offset == PAGE_SIZE); |
4fc75ff4 | 1589 | |
b6a2fea3 OW |
1590 | bprm->p++; |
1591 | bprm->argc--; | |
1592 | ret = 0; | |
4fc75ff4 | 1593 | |
b6a2fea3 OW |
1594 | out: |
1595 | return ret; | |
1da177e4 | 1596 | } |
1da177e4 LT |
1597 | EXPORT_SYMBOL(remove_arg_zero); |
1598 | ||
cb7b6b1c | 1599 | #define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e)) |
1da177e4 LT |
1600 | /* |
1601 | * cycle the list of binary formats handler, until one recognizes the image | |
1602 | */ | |
3c456bfc | 1603 | int search_binary_handler(struct linux_binprm *bprm) |
1da177e4 | 1604 | { |
cb7b6b1c | 1605 | bool need_retry = IS_ENABLED(CONFIG_MODULES); |
1da177e4 | 1606 | struct linux_binfmt *fmt; |
cb7b6b1c | 1607 | int retval; |
1da177e4 | 1608 | |
d7402698 | 1609 | /* This allows 4 levels of binfmt rewrites before failing hard. */ |
131b2f9f | 1610 | if (bprm->recursion_depth > 5) |
d7402698 KC |
1611 | return -ELOOP; |
1612 | ||
1da177e4 LT |
1613 | retval = security_bprm_check(bprm); |
1614 | if (retval) | |
1615 | return retval; | |
1616 | ||
1da177e4 | 1617 | retval = -ENOENT; |
cb7b6b1c ON |
1618 | retry: |
1619 | read_lock(&binfmt_lock); | |
1620 | list_for_each_entry(fmt, &formats, lh) { | |
1621 | if (!try_module_get(fmt->module)) | |
1622 | continue; | |
1623 | read_unlock(&binfmt_lock); | |
1624 | bprm->recursion_depth++; | |
1625 | retval = fmt->load_binary(bprm); | |
19d860a1 AV |
1626 | read_lock(&binfmt_lock); |
1627 | put_binfmt(fmt); | |
cb7b6b1c | 1628 | bprm->recursion_depth--; |
19d860a1 AV |
1629 | if (retval < 0 && !bprm->mm) { |
1630 | /* we got to flush_old_exec() and failed after it */ | |
1631 | read_unlock(&binfmt_lock); | |
1632 | force_sigsegv(SIGSEGV, current); | |
1633 | return retval; | |
1634 | } | |
1635 | if (retval != -ENOEXEC || !bprm->file) { | |
1636 | read_unlock(&binfmt_lock); | |
cb7b6b1c | 1637 | return retval; |
1da177e4 | 1638 | } |
1da177e4 | 1639 | } |
cb7b6b1c ON |
1640 | read_unlock(&binfmt_lock); |
1641 | ||
19d860a1 | 1642 | if (need_retry) { |
cb7b6b1c ON |
1643 | if (printable(bprm->buf[0]) && printable(bprm->buf[1]) && |
1644 | printable(bprm->buf[2]) && printable(bprm->buf[3])) | |
1645 | return retval; | |
4e0621a0 ON |
1646 | if (request_module("binfmt-%04x", *(ushort *)(bprm->buf + 2)) < 0) |
1647 | return retval; | |
cb7b6b1c ON |
1648 | need_retry = false; |
1649 | goto retry; | |
1650 | } | |
1651 | ||
1da177e4 LT |
1652 | return retval; |
1653 | } | |
1da177e4 LT |
1654 | EXPORT_SYMBOL(search_binary_handler); |
1655 | ||
5d1baf3b ON |
1656 | static int exec_binprm(struct linux_binprm *bprm) |
1657 | { | |
1658 | pid_t old_pid, old_vpid; | |
1659 | int ret; | |
1660 | ||
1661 | /* Need to fetch pid before load_binary changes it */ | |
1662 | old_pid = current->pid; | |
1663 | rcu_read_lock(); | |
1664 | old_vpid = task_pid_nr_ns(current, task_active_pid_ns(current->parent)); | |
1665 | rcu_read_unlock(); | |
1666 | ||
1667 | ret = search_binary_handler(bprm); | |
1668 | if (ret >= 0) { | |
3eaded86 | 1669 | audit_bprm(bprm); |
5d1baf3b ON |
1670 | trace_sched_process_exec(current, old_pid, bprm); |
1671 | ptrace_event(PTRACE_EVENT_EXEC, old_vpid); | |
9beb266f | 1672 | proc_exec_connector(current); |
5d1baf3b ON |
1673 | } |
1674 | ||
1675 | return ret; | |
1676 | } | |
1677 | ||
1da177e4 LT |
1678 | /* |
1679 | * sys_execve() executes a new program. | |
1680 | */ | |
51f39a1f DD |
1681 | static int do_execveat_common(int fd, struct filename *filename, |
1682 | struct user_arg_ptr argv, | |
1683 | struct user_arg_ptr envp, | |
1684 | int flags) | |
1da177e4 | 1685 | { |
51f39a1f | 1686 | char *pathbuf = NULL; |
1da177e4 LT |
1687 | struct linux_binprm *bprm; |
1688 | struct file *file; | |
3b125388 | 1689 | struct files_struct *displaced; |
1da177e4 | 1690 | int retval; |
72fa5997 | 1691 | |
c4ad8f98 LT |
1692 | if (IS_ERR(filename)) |
1693 | return PTR_ERR(filename); | |
1694 | ||
72fa5997 VK |
1695 | /* |
1696 | * We move the actual failure in case of RLIMIT_NPROC excess from | |
1697 | * set*uid() to execve() because too many poorly written programs | |
1698 | * don't check setuid() return code. Here we additionally recheck | |
1699 | * whether NPROC limit is still exceeded. | |
1700 | */ | |
1701 | if ((current->flags & PF_NPROC_EXCEEDED) && | |
bd9d43f4 | 1702 | atomic_read(¤t_user()->processes) > rlimit(RLIMIT_NPROC)) { |
72fa5997 VK |
1703 | retval = -EAGAIN; |
1704 | goto out_ret; | |
1705 | } | |
1706 | ||
1707 | /* We're below the limit (still or again), so we don't want to make | |
1708 | * further execve() calls fail. */ | |
1709 | current->flags &= ~PF_NPROC_EXCEEDED; | |
1da177e4 | 1710 | |
3b125388 | 1711 | retval = unshare_files(&displaced); |
fd8328be AV |
1712 | if (retval) |
1713 | goto out_ret; | |
1714 | ||
1da177e4 | 1715 | retval = -ENOMEM; |
11b0b5ab | 1716 | bprm = kzalloc(sizeof(*bprm), GFP_KERNEL); |
1da177e4 | 1717 | if (!bprm) |
fd8328be | 1718 | goto out_files; |
1da177e4 | 1719 | |
a2a8474c ON |
1720 | retval = prepare_bprm_creds(bprm); |
1721 | if (retval) | |
a6f76f23 | 1722 | goto out_free; |
498052bb | 1723 | |
9e00cdb0 | 1724 | check_unsafe_exec(bprm); |
a2a8474c | 1725 | current->in_execve = 1; |
a6f76f23 | 1726 | |
51f39a1f | 1727 | file = do_open_execat(fd, filename, flags); |
1da177e4 LT |
1728 | retval = PTR_ERR(file); |
1729 | if (IS_ERR(file)) | |
498052bb | 1730 | goto out_unmark; |
1da177e4 LT |
1731 | |
1732 | sched_exec(); | |
1733 | ||
1da177e4 | 1734 | bprm->file = file; |
51f39a1f DD |
1735 | if (fd == AT_FDCWD || filename->name[0] == '/') { |
1736 | bprm->filename = filename->name; | |
1737 | } else { | |
1738 | if (filename->name[0] == '\0') | |
1739 | pathbuf = kasprintf(GFP_TEMPORARY, "/dev/fd/%d", fd); | |
1740 | else | |
1741 | pathbuf = kasprintf(GFP_TEMPORARY, "/dev/fd/%d/%s", | |
1742 | fd, filename->name); | |
1743 | if (!pathbuf) { | |
1744 | retval = -ENOMEM; | |
1745 | goto out_unmark; | |
1746 | } | |
1747 | /* | |
1748 | * Record that a name derived from an O_CLOEXEC fd will be | |
1749 | * inaccessible after exec. Relies on having exclusive access to | |
1750 | * current->files (due to unshare_files above). | |
1751 | */ | |
1752 | if (close_on_exec(fd, rcu_dereference_raw(current->files->fdt))) | |
1753 | bprm->interp_flags |= BINPRM_FLAGS_PATH_INACCESSIBLE; | |
1754 | bprm->filename = pathbuf; | |
1755 | } | |
1756 | bprm->interp = bprm->filename; | |
1da177e4 | 1757 | |
b6a2fea3 OW |
1758 | retval = bprm_mm_init(bprm); |
1759 | if (retval) | |
63e46b95 | 1760 | goto out_unmark; |
1da177e4 | 1761 | |
b6a2fea3 | 1762 | bprm->argc = count(argv, MAX_ARG_STRINGS); |
1da177e4 | 1763 | if ((retval = bprm->argc) < 0) |
a6f76f23 | 1764 | goto out; |
1da177e4 | 1765 | |
b6a2fea3 | 1766 | bprm->envc = count(envp, MAX_ARG_STRINGS); |
1da177e4 | 1767 | if ((retval = bprm->envc) < 0) |
1da177e4 LT |
1768 | goto out; |
1769 | ||
1770 | retval = prepare_binprm(bprm); | |
1771 | if (retval < 0) | |
1772 | goto out; | |
1773 | ||
1774 | retval = copy_strings_kernel(1, &bprm->filename, bprm); | |
1775 | if (retval < 0) | |
1776 | goto out; | |
1777 | ||
1778 | bprm->exec = bprm->p; | |
1779 | retval = copy_strings(bprm->envc, envp, bprm); | |
1780 | if (retval < 0) | |
1781 | goto out; | |
1782 | ||
1783 | retval = copy_strings(bprm->argc, argv, bprm); | |
1784 | if (retval < 0) | |
1785 | goto out; | |
1786 | ||
f84df2a6 EB |
1787 | would_dump(bprm, bprm->file); |
1788 | ||
5d1baf3b | 1789 | retval = exec_binprm(bprm); |
a6f76f23 DH |
1790 | if (retval < 0) |
1791 | goto out; | |
1da177e4 | 1792 | |
a6f76f23 | 1793 | /* execve succeeded */ |
498052bb | 1794 | current->fs->in_exec = 0; |
f9ce1f1c | 1795 | current->in_execve = 0; |
a6f76f23 | 1796 | acct_update_integrals(current); |
82727018 | 1797 | task_numa_free(current); |
a6f76f23 | 1798 | free_bprm(bprm); |
51f39a1f | 1799 | kfree(pathbuf); |
c4ad8f98 | 1800 | putname(filename); |
a6f76f23 DH |
1801 | if (displaced) |
1802 | put_files_struct(displaced); | |
1803 | return retval; | |
1da177e4 | 1804 | |
a6f76f23 | 1805 | out: |
3c77f845 ON |
1806 | if (bprm->mm) { |
1807 | acct_arg_size(bprm, 0); | |
1808 | mmput(bprm->mm); | |
1809 | } | |
1da177e4 | 1810 | |
498052bb | 1811 | out_unmark: |
9e00cdb0 | 1812 | current->fs->in_exec = 0; |
f9ce1f1c | 1813 | current->in_execve = 0; |
a6f76f23 DH |
1814 | |
1815 | out_free: | |
08a6fac1 | 1816 | free_bprm(bprm); |
51f39a1f | 1817 | kfree(pathbuf); |
1da177e4 | 1818 | |
fd8328be | 1819 | out_files: |
3b125388 AV |
1820 | if (displaced) |
1821 | reset_files_struct(displaced); | |
1da177e4 | 1822 | out_ret: |
c4ad8f98 | 1823 | putname(filename); |
1da177e4 LT |
1824 | return retval; |
1825 | } | |
1826 | ||
c4ad8f98 | 1827 | int do_execve(struct filename *filename, |
ba2d0162 | 1828 | const char __user *const __user *__argv, |
da3d4c5f | 1829 | const char __user *const __user *__envp) |
ba2d0162 | 1830 | { |
0e028465 ON |
1831 | struct user_arg_ptr argv = { .ptr.native = __argv }; |
1832 | struct user_arg_ptr envp = { .ptr.native = __envp }; | |
51f39a1f DD |
1833 | return do_execveat_common(AT_FDCWD, filename, argv, envp, 0); |
1834 | } | |
1835 | ||
1836 | int do_execveat(int fd, struct filename *filename, | |
1837 | const char __user *const __user *__argv, | |
1838 | const char __user *const __user *__envp, | |
1839 | int flags) | |
1840 | { | |
1841 | struct user_arg_ptr argv = { .ptr.native = __argv }; | |
1842 | struct user_arg_ptr envp = { .ptr.native = __envp }; | |
1843 | ||
1844 | return do_execveat_common(fd, filename, argv, envp, flags); | |
0e028465 ON |
1845 | } |
1846 | ||
1847 | #ifdef CONFIG_COMPAT | |
c4ad8f98 | 1848 | static int compat_do_execve(struct filename *filename, |
38b983b3 | 1849 | const compat_uptr_t __user *__argv, |
d03d26e5 | 1850 | const compat_uptr_t __user *__envp) |
0e028465 ON |
1851 | { |
1852 | struct user_arg_ptr argv = { | |
1853 | .is_compat = true, | |
1854 | .ptr.compat = __argv, | |
1855 | }; | |
1856 | struct user_arg_ptr envp = { | |
1857 | .is_compat = true, | |
1858 | .ptr.compat = __envp, | |
1859 | }; | |
51f39a1f DD |
1860 | return do_execveat_common(AT_FDCWD, filename, argv, envp, 0); |
1861 | } | |
1862 | ||
1863 | static int compat_do_execveat(int fd, struct filename *filename, | |
1864 | const compat_uptr_t __user *__argv, | |
1865 | const compat_uptr_t __user *__envp, | |
1866 | int flags) | |
1867 | { | |
1868 | struct user_arg_ptr argv = { | |
1869 | .is_compat = true, | |
1870 | .ptr.compat = __argv, | |
1871 | }; | |
1872 | struct user_arg_ptr envp = { | |
1873 | .is_compat = true, | |
1874 | .ptr.compat = __envp, | |
1875 | }; | |
1876 | return do_execveat_common(fd, filename, argv, envp, flags); | |
ba2d0162 | 1877 | } |
0e028465 | 1878 | #endif |
ba2d0162 | 1879 | |
964ee7df | 1880 | void set_binfmt(struct linux_binfmt *new) |
1da177e4 | 1881 | { |
801460d0 HS |
1882 | struct mm_struct *mm = current->mm; |
1883 | ||
1884 | if (mm->binfmt) | |
1885 | module_put(mm->binfmt->module); | |
1da177e4 | 1886 | |
801460d0 | 1887 | mm->binfmt = new; |
964ee7df ON |
1888 | if (new) |
1889 | __module_get(new->module); | |
1da177e4 | 1890 | } |
1da177e4 LT |
1891 | EXPORT_SYMBOL(set_binfmt); |
1892 | ||
6c5d5238 | 1893 | /* |
7288e118 | 1894 | * set_dumpable stores three-value SUID_DUMP_* into mm->flags. |
6c5d5238 KH |
1895 | */ |
1896 | void set_dumpable(struct mm_struct *mm, int value) | |
1897 | { | |
abacd2fe ON |
1898 | unsigned long old, new; |
1899 | ||
7288e118 ON |
1900 | if (WARN_ON((unsigned)value > SUID_DUMP_ROOT)) |
1901 | return; | |
1902 | ||
abacd2fe ON |
1903 | do { |
1904 | old = ACCESS_ONCE(mm->flags); | |
7288e118 | 1905 | new = (old & ~MMF_DUMPABLE_MASK) | value; |
abacd2fe | 1906 | } while (cmpxchg(&mm->flags, old, new) != old); |
6c5d5238 | 1907 | } |
6c5d5238 | 1908 | |
38b983b3 AV |
1909 | SYSCALL_DEFINE3(execve, |
1910 | const char __user *, filename, | |
1911 | const char __user *const __user *, argv, | |
1912 | const char __user *const __user *, envp) | |
1913 | { | |
c4ad8f98 | 1914 | return do_execve(getname(filename), argv, envp); |
38b983b3 | 1915 | } |
51f39a1f DD |
1916 | |
1917 | SYSCALL_DEFINE5(execveat, | |
1918 | int, fd, const char __user *, filename, | |
1919 | const char __user *const __user *, argv, | |
1920 | const char __user *const __user *, envp, | |
1921 | int, flags) | |
1922 | { | |
1923 | int lookup_flags = (flags & AT_EMPTY_PATH) ? LOOKUP_EMPTY : 0; | |
1924 | ||
1925 | return do_execveat(fd, | |
1926 | getname_flags(filename, lookup_flags, NULL), | |
1927 | argv, envp, flags); | |
1928 | } | |
1929 | ||
38b983b3 | 1930 | #ifdef CONFIG_COMPAT |
625b1d7e HC |
1931 | COMPAT_SYSCALL_DEFINE3(execve, const char __user *, filename, |
1932 | const compat_uptr_t __user *, argv, | |
1933 | const compat_uptr_t __user *, envp) | |
38b983b3 | 1934 | { |
c4ad8f98 | 1935 | return compat_do_execve(getname(filename), argv, envp); |
38b983b3 | 1936 | } |
51f39a1f DD |
1937 | |
1938 | COMPAT_SYSCALL_DEFINE5(execveat, int, fd, | |
1939 | const char __user *, filename, | |
1940 | const compat_uptr_t __user *, argv, | |
1941 | const compat_uptr_t __user *, envp, | |
1942 | int, flags) | |
1943 | { | |
1944 | int lookup_flags = (flags & AT_EMPTY_PATH) ? LOOKUP_EMPTY : 0; | |
1945 | ||
1946 | return compat_do_execveat(fd, | |
1947 | getname_flags(filename, lookup_flags, NULL), | |
1948 | argv, envp, flags); | |
1949 | } | |
38b983b3 | 1950 | #endif |