]>
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 | |
7744c33f SJR |
66 | #include <trace/events/fs.h> |
67 | ||
7c0f6ba6 | 68 | #include <linux/uaccess.h> |
1da177e4 | 69 | #include <asm/mmu_context.h> |
b6a2fea3 | 70 | #include <asm/tlb.h> |
43d2b113 KH |
71 | |
72 | #include <trace/events/task.h> | |
a6f76f23 | 73 | #include "internal.h" |
1da177e4 | 74 | |
4ff16c25 DS |
75 | #include <trace/events/sched.h> |
76 | ||
d6e71144 AC |
77 | int suid_dumpable = 0; |
78 | ||
e4dc1b14 | 79 | static LIST_HEAD(formats); |
1da177e4 LT |
80 | static DEFINE_RWLOCK(binfmt_lock); |
81 | ||
8fc3dc5a | 82 | void __register_binfmt(struct linux_binfmt * fmt, int insert) |
1da177e4 | 83 | { |
8fc3dc5a | 84 | BUG_ON(!fmt); |
92eaa565 ON |
85 | if (WARN_ON(!fmt->load_binary)) |
86 | return; | |
1da177e4 | 87 | write_lock(&binfmt_lock); |
74641f58 IK |
88 | insert ? list_add(&fmt->lh, &formats) : |
89 | list_add_tail(&fmt->lh, &formats); | |
1da177e4 | 90 | write_unlock(&binfmt_lock); |
1da177e4 LT |
91 | } |
92 | ||
74641f58 | 93 | EXPORT_SYMBOL(__register_binfmt); |
1da177e4 | 94 | |
f6b450d4 | 95 | void unregister_binfmt(struct linux_binfmt * fmt) |
1da177e4 | 96 | { |
1da177e4 | 97 | write_lock(&binfmt_lock); |
e4dc1b14 | 98 | list_del(&fmt->lh); |
1da177e4 | 99 | write_unlock(&binfmt_lock); |
1da177e4 LT |
100 | } |
101 | ||
102 | EXPORT_SYMBOL(unregister_binfmt); | |
103 | ||
104 | static inline void put_binfmt(struct linux_binfmt * fmt) | |
105 | { | |
106 | module_put(fmt->module); | |
107 | } | |
108 | ||
90f8572b EB |
109 | bool path_noexec(const struct path *path) |
110 | { | |
111 | return (path->mnt->mnt_flags & MNT_NOEXEC) || | |
112 | (path->mnt->mnt_sb->s_iflags & SB_I_NOEXEC); | |
113 | } | |
c088e31d | 114 | EXPORT_SYMBOL_GPL(path_noexec); |
90f8572b | 115 | |
bf31f6b5 SF |
116 | bool path_nosuid(const struct path *path) |
117 | { | |
118 | return !mnt_may_suid(path->mnt) || | |
119 | (path->mnt->mnt_sb->s_iflags & SB_I_NOSUID); | |
120 | } | |
121 | EXPORT_SYMBOL(path_nosuid); | |
122 | ||
69369a70 | 123 | #ifdef CONFIG_USELIB |
1da177e4 LT |
124 | /* |
125 | * Note that a shared library must be both readable and executable due to | |
126 | * security reasons. | |
127 | * | |
128 | * Also note that we take the address to load from from the file itself. | |
129 | */ | |
1e7bfb21 | 130 | SYSCALL_DEFINE1(uselib, const char __user *, library) |
1da177e4 | 131 | { |
72c2d531 | 132 | struct linux_binfmt *fmt; |
964bd183 | 133 | struct file *file; |
91a27b2a | 134 | struct filename *tmp = getname(library); |
964bd183 | 135 | int error = PTR_ERR(tmp); |
47c805dc AV |
136 | static const struct open_flags uselib_flags = { |
137 | .open_flag = O_LARGEFILE | O_RDONLY | __FMODE_EXEC, | |
62fb4a15 | 138 | .acc_mode = MAY_READ | MAY_EXEC, |
f9652e10 AV |
139 | .intent = LOOKUP_OPEN, |
140 | .lookup_flags = LOOKUP_FOLLOW, | |
47c805dc | 141 | }; |
964bd183 | 142 | |
6e8341a1 AV |
143 | if (IS_ERR(tmp)) |
144 | goto out; | |
145 | ||
f9652e10 | 146 | file = do_filp_open(AT_FDCWD, tmp, &uselib_flags); |
6e8341a1 AV |
147 | putname(tmp); |
148 | error = PTR_ERR(file); | |
149 | if (IS_ERR(file)) | |
1da177e4 LT |
150 | goto out; |
151 | ||
152 | error = -EINVAL; | |
496ad9aa | 153 | if (!S_ISREG(file_inode(file)->i_mode)) |
1da177e4 LT |
154 | goto exit; |
155 | ||
30524472 | 156 | error = -EACCES; |
90f8572b | 157 | if (path_noexec(&file->f_path)) |
1da177e4 LT |
158 | goto exit; |
159 | ||
2a12a9d7 | 160 | fsnotify_open(file); |
6110e3ab | 161 | |
1da177e4 | 162 | error = -ENOEXEC; |
1da177e4 | 163 | |
72c2d531 AV |
164 | read_lock(&binfmt_lock); |
165 | list_for_each_entry(fmt, &formats, lh) { | |
166 | if (!fmt->load_shlib) | |
167 | continue; | |
168 | if (!try_module_get(fmt->module)) | |
169 | continue; | |
1da177e4 | 170 | read_unlock(&binfmt_lock); |
72c2d531 AV |
171 | error = fmt->load_shlib(file); |
172 | read_lock(&binfmt_lock); | |
173 | put_binfmt(fmt); | |
174 | if (error != -ENOEXEC) | |
175 | break; | |
1da177e4 | 176 | } |
72c2d531 | 177 | read_unlock(&binfmt_lock); |
6e8341a1 | 178 | exit: |
1da177e4 LT |
179 | fput(file); |
180 | out: | |
181 | return error; | |
1da177e4 | 182 | } |
69369a70 | 183 | #endif /* #ifdef CONFIG_USELIB */ |
1da177e4 | 184 | |
b6a2fea3 | 185 | #ifdef CONFIG_MMU |
ae6b585e ON |
186 | /* |
187 | * The nascent bprm->mm is not visible until exec_mmap() but it can | |
188 | * use a lot of memory, account these pages in current->mm temporary | |
189 | * for oom_badness()->get_mm_rss(). Once exec succeeds or fails, we | |
190 | * change the counter back via acct_arg_size(0). | |
191 | */ | |
0e028465 | 192 | static void acct_arg_size(struct linux_binprm *bprm, unsigned long pages) |
3c77f845 ON |
193 | { |
194 | struct mm_struct *mm = current->mm; | |
195 | long diff = (long)(pages - bprm->vma_pages); | |
196 | ||
197 | if (!mm || !diff) | |
198 | return; | |
199 | ||
200 | bprm->vma_pages = pages; | |
3c77f845 | 201 | add_mm_counter(mm, MM_ANONPAGES, diff); |
3c77f845 ON |
202 | } |
203 | ||
0e028465 | 204 | static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos, |
b6a2fea3 OW |
205 | int write) |
206 | { | |
207 | struct page *page; | |
208 | int ret; | |
9beae1ea | 209 | unsigned int gup_flags = FOLL_FORCE; |
b6a2fea3 OW |
210 | |
211 | #ifdef CONFIG_STACK_GROWSUP | |
212 | if (write) { | |
d05f3169 | 213 | ret = expand_downwards(bprm->vma, pos); |
b6a2fea3 OW |
214 | if (ret < 0) |
215 | return NULL; | |
216 | } | |
217 | #endif | |
9beae1ea LS |
218 | |
219 | if (write) | |
220 | gup_flags |= FOLL_WRITE; | |
221 | ||
1e987790 DH |
222 | /* |
223 | * We are doing an exec(). 'current' is the process | |
224 | * doing the exec and bprm->mm is the new process's mm. | |
225 | */ | |
9beae1ea | 226 | ret = get_user_pages_remote(current, bprm->mm, pos, 1, gup_flags, |
5b56d49f | 227 | &page, NULL, NULL); |
b6a2fea3 OW |
228 | if (ret <= 0) |
229 | return NULL; | |
230 | ||
231 | if (write) { | |
b6a2fea3 | 232 | unsigned long size = bprm->vma->vm_end - bprm->vma->vm_start; |
da029c11 | 233 | unsigned long ptr_size, limit; |
a64e715f | 234 | |
98da7d08 KC |
235 | /* |
236 | * Since the stack will hold pointers to the strings, we | |
237 | * must account for them as well. | |
238 | * | |
239 | * The size calculation is the entire vma while each arg page is | |
240 | * built, so each time we get here it's calculating how far it | |
241 | * is currently (rather than each call being just the newly | |
242 | * added size from the arg page). As a result, we need to | |
243 | * always add the entire size of the pointers, so that on the | |
244 | * last call to get_arg_page() we'll actually have the entire | |
245 | * correct size. | |
246 | */ | |
247 | ptr_size = (bprm->argc + bprm->envc) * sizeof(void *); | |
248 | if (ptr_size > ULONG_MAX - size) | |
249 | goto fail; | |
250 | size += ptr_size; | |
251 | ||
3c77f845 ON |
252 | acct_arg_size(bprm, size / PAGE_SIZE); |
253 | ||
a64e715f LT |
254 | /* |
255 | * We've historically supported up to 32 pages (ARG_MAX) | |
256 | * of argument strings even with small stacks | |
257 | */ | |
258 | if (size <= ARG_MAX) | |
259 | return page; | |
b6a2fea3 OW |
260 | |
261 | /* | |
da029c11 KC |
262 | * Limit to 1/4 of the max stack size or 3/4 of _STK_LIM |
263 | * (whichever is smaller) for the argv+env strings. | |
b6a2fea3 OW |
264 | * This ensures that: |
265 | * - the remaining binfmt code will not run out of stack space, | |
266 | * - the program will have a reasonable amount of stack left | |
267 | * to work from. | |
268 | */ | |
da029c11 KC |
269 | limit = _STK_LIM / 4 * 3; |
270 | limit = min(limit, rlimit(RLIMIT_STACK) / 4); | |
271 | if (size > limit) | |
98da7d08 | 272 | goto fail; |
b6a2fea3 OW |
273 | } |
274 | ||
275 | return page; | |
98da7d08 KC |
276 | |
277 | fail: | |
278 | put_page(page); | |
279 | return NULL; | |
b6a2fea3 OW |
280 | } |
281 | ||
282 | static void put_arg_page(struct page *page) | |
283 | { | |
284 | put_page(page); | |
285 | } | |
286 | ||
b6a2fea3 OW |
287 | static void free_arg_pages(struct linux_binprm *bprm) |
288 | { | |
289 | } | |
290 | ||
291 | static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos, | |
292 | struct page *page) | |
293 | { | |
294 | flush_cache_page(bprm->vma, pos, page_to_pfn(page)); | |
295 | } | |
296 | ||
297 | static int __bprm_mm_init(struct linux_binprm *bprm) | |
298 | { | |
eaccbfa5 | 299 | int err; |
b6a2fea3 OW |
300 | struct vm_area_struct *vma = NULL; |
301 | struct mm_struct *mm = bprm->mm; | |
302 | ||
303 | bprm->vma = vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL); | |
304 | if (!vma) | |
eaccbfa5 | 305 | return -ENOMEM; |
b6a2fea3 | 306 | |
f268dfe9 MH |
307 | if (down_write_killable(&mm->mmap_sem)) { |
308 | err = -EINTR; | |
309 | goto err_free; | |
310 | } | |
b6a2fea3 OW |
311 | vma->vm_mm = mm; |
312 | ||
313 | /* | |
314 | * Place the stack at the largest stack address the architecture | |
315 | * supports. Later, we'll move this to an appropriate place. We don't | |
316 | * use STACK_TOP because that can depend on attributes which aren't | |
317 | * configured yet. | |
318 | */ | |
aacb3d17 | 319 | BUILD_BUG_ON(VM_STACK_FLAGS & VM_STACK_INCOMPLETE_SETUP); |
b6a2fea3 OW |
320 | vma->vm_end = STACK_TOP_MAX; |
321 | vma->vm_start = vma->vm_end - PAGE_SIZE; | |
d9104d1c | 322 | vma->vm_flags = VM_SOFTDIRTY | VM_STACK_FLAGS | VM_STACK_INCOMPLETE_SETUP; |
3ed75eb8 | 323 | vma->vm_page_prot = vm_get_page_prot(vma->vm_flags); |
5beb4930 | 324 | INIT_LIST_HEAD(&vma->anon_vma_chain); |
462e635e | 325 | |
b6a2fea3 | 326 | err = insert_vm_struct(mm, vma); |
eaccbfa5 | 327 | if (err) |
b6a2fea3 | 328 | goto err; |
b6a2fea3 OW |
329 | |
330 | mm->stack_vm = mm->total_vm = 1; | |
fe3d197f | 331 | arch_bprm_mm_init(mm, vma); |
b6a2fea3 | 332 | up_write(&mm->mmap_sem); |
b6a2fea3 | 333 | bprm->p = vma->vm_end - sizeof(void *); |
b6a2fea3 | 334 | return 0; |
b6a2fea3 | 335 | err: |
eaccbfa5 | 336 | up_write(&mm->mmap_sem); |
f268dfe9 | 337 | err_free: |
eaccbfa5 LFC |
338 | bprm->vma = NULL; |
339 | kmem_cache_free(vm_area_cachep, vma); | |
b6a2fea3 OW |
340 | return err; |
341 | } | |
342 | ||
343 | static bool valid_arg_len(struct linux_binprm *bprm, long len) | |
344 | { | |
345 | return len <= MAX_ARG_STRLEN; | |
346 | } | |
347 | ||
348 | #else | |
349 | ||
0e028465 | 350 | static inline void acct_arg_size(struct linux_binprm *bprm, unsigned long pages) |
3c77f845 ON |
351 | { |
352 | } | |
353 | ||
0e028465 | 354 | static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos, |
b6a2fea3 OW |
355 | int write) |
356 | { | |
357 | struct page *page; | |
358 | ||
359 | page = bprm->page[pos / PAGE_SIZE]; | |
360 | if (!page && write) { | |
361 | page = alloc_page(GFP_HIGHUSER|__GFP_ZERO); | |
362 | if (!page) | |
363 | return NULL; | |
364 | bprm->page[pos / PAGE_SIZE] = page; | |
365 | } | |
366 | ||
367 | return page; | |
368 | } | |
369 | ||
370 | static void put_arg_page(struct page *page) | |
371 | { | |
372 | } | |
373 | ||
374 | static void free_arg_page(struct linux_binprm *bprm, int i) | |
375 | { | |
376 | if (bprm->page[i]) { | |
377 | __free_page(bprm->page[i]); | |
378 | bprm->page[i] = NULL; | |
379 | } | |
380 | } | |
381 | ||
382 | static void free_arg_pages(struct linux_binprm *bprm) | |
383 | { | |
384 | int i; | |
385 | ||
386 | for (i = 0; i < MAX_ARG_PAGES; i++) | |
387 | free_arg_page(bprm, i); | |
388 | } | |
389 | ||
390 | static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos, | |
391 | struct page *page) | |
392 | { | |
393 | } | |
394 | ||
395 | static int __bprm_mm_init(struct linux_binprm *bprm) | |
396 | { | |
397 | bprm->p = PAGE_SIZE * MAX_ARG_PAGES - sizeof(void *); | |
398 | return 0; | |
399 | } | |
400 | ||
401 | static bool valid_arg_len(struct linux_binprm *bprm, long len) | |
402 | { | |
403 | return len <= bprm->p; | |
404 | } | |
405 | ||
406 | #endif /* CONFIG_MMU */ | |
407 | ||
408 | /* | |
409 | * Create a new mm_struct and populate it with a temporary stack | |
410 | * vm_area_struct. We don't have enough context at this point to set the stack | |
411 | * flags, permissions, and offset, so we use temporary values. We'll update | |
412 | * them later in setup_arg_pages(). | |
413 | */ | |
9cc64cea | 414 | static int bprm_mm_init(struct linux_binprm *bprm) |
b6a2fea3 OW |
415 | { |
416 | int err; | |
417 | struct mm_struct *mm = NULL; | |
418 | ||
419 | bprm->mm = mm = mm_alloc(); | |
420 | err = -ENOMEM; | |
421 | if (!mm) | |
422 | goto err; | |
423 | ||
b6a2fea3 OW |
424 | err = __bprm_mm_init(bprm); |
425 | if (err) | |
426 | goto err; | |
427 | ||
428 | return 0; | |
429 | ||
430 | err: | |
431 | if (mm) { | |
432 | bprm->mm = NULL; | |
433 | mmdrop(mm); | |
434 | } | |
435 | ||
436 | return err; | |
437 | } | |
438 | ||
ba2d0162 | 439 | struct user_arg_ptr { |
0e028465 ON |
440 | #ifdef CONFIG_COMPAT |
441 | bool is_compat; | |
442 | #endif | |
443 | union { | |
444 | const char __user *const __user *native; | |
445 | #ifdef CONFIG_COMPAT | |
38b983b3 | 446 | const compat_uptr_t __user *compat; |
0e028465 ON |
447 | #endif |
448 | } ptr; | |
ba2d0162 ON |
449 | }; |
450 | ||
451 | static const char __user *get_user_arg_ptr(struct user_arg_ptr argv, int nr) | |
1d1dbf81 | 452 | { |
0e028465 ON |
453 | const char __user *native; |
454 | ||
455 | #ifdef CONFIG_COMPAT | |
456 | if (unlikely(argv.is_compat)) { | |
457 | compat_uptr_t compat; | |
458 | ||
459 | if (get_user(compat, argv.ptr.compat + nr)) | |
460 | return ERR_PTR(-EFAULT); | |
1d1dbf81 | 461 | |
0e028465 ON |
462 | return compat_ptr(compat); |
463 | } | |
464 | #endif | |
465 | ||
466 | if (get_user(native, argv.ptr.native + nr)) | |
1d1dbf81 ON |
467 | return ERR_PTR(-EFAULT); |
468 | ||
0e028465 | 469 | return native; |
1d1dbf81 ON |
470 | } |
471 | ||
1da177e4 LT |
472 | /* |
473 | * count() counts the number of strings in array ARGV. | |
474 | */ | |
ba2d0162 | 475 | static int count(struct user_arg_ptr argv, int max) |
1da177e4 LT |
476 | { |
477 | int i = 0; | |
478 | ||
0e028465 | 479 | if (argv.ptr.native != NULL) { |
1da177e4 | 480 | for (;;) { |
1d1dbf81 | 481 | const char __user *p = get_user_arg_ptr(argv, i); |
1da177e4 | 482 | |
1da177e4 LT |
483 | if (!p) |
484 | break; | |
1d1dbf81 ON |
485 | |
486 | if (IS_ERR(p)) | |
487 | return -EFAULT; | |
488 | ||
6d92d4f6 | 489 | if (i >= max) |
1da177e4 | 490 | return -E2BIG; |
6d92d4f6 | 491 | ++i; |
9aea5a65 RM |
492 | |
493 | if (fatal_signal_pending(current)) | |
494 | return -ERESTARTNOHAND; | |
1da177e4 LT |
495 | cond_resched(); |
496 | } | |
497 | } | |
498 | return i; | |
499 | } | |
500 | ||
501 | /* | |
b6a2fea3 OW |
502 | * 'copy_strings()' copies argument/environment strings from the old |
503 | * processes's memory to the new process's stack. The call to get_user_pages() | |
504 | * ensures the destination page is created and not swapped out. | |
1da177e4 | 505 | */ |
ba2d0162 | 506 | static int copy_strings(int argc, struct user_arg_ptr argv, |
75c96f85 | 507 | struct linux_binprm *bprm) |
1da177e4 LT |
508 | { |
509 | struct page *kmapped_page = NULL; | |
510 | char *kaddr = NULL; | |
b6a2fea3 | 511 | unsigned long kpos = 0; |
1da177e4 LT |
512 | int ret; |
513 | ||
514 | while (argc-- > 0) { | |
d7627467 | 515 | const char __user *str; |
1da177e4 LT |
516 | int len; |
517 | unsigned long pos; | |
518 | ||
1d1dbf81 ON |
519 | ret = -EFAULT; |
520 | str = get_user_arg_ptr(argv, argc); | |
521 | if (IS_ERR(str)) | |
1da177e4 | 522 | goto out; |
1da177e4 | 523 | |
1d1dbf81 ON |
524 | len = strnlen_user(str, MAX_ARG_STRLEN); |
525 | if (!len) | |
526 | goto out; | |
527 | ||
528 | ret = -E2BIG; | |
529 | if (!valid_arg_len(bprm, len)) | |
1da177e4 | 530 | goto out; |
1da177e4 | 531 | |
b6a2fea3 | 532 | /* We're going to work our way backwords. */ |
1da177e4 | 533 | pos = bprm->p; |
b6a2fea3 OW |
534 | str += len; |
535 | bprm->p -= len; | |
1da177e4 LT |
536 | |
537 | while (len > 0) { | |
1da177e4 | 538 | int offset, bytes_to_copy; |
1da177e4 | 539 | |
9aea5a65 RM |
540 | if (fatal_signal_pending(current)) { |
541 | ret = -ERESTARTNOHAND; | |
542 | goto out; | |
543 | } | |
7993bc1f RM |
544 | cond_resched(); |
545 | ||
1da177e4 | 546 | offset = pos % PAGE_SIZE; |
b6a2fea3 OW |
547 | if (offset == 0) |
548 | offset = PAGE_SIZE; | |
549 | ||
550 | bytes_to_copy = offset; | |
551 | if (bytes_to_copy > len) | |
552 | bytes_to_copy = len; | |
553 | ||
554 | offset -= bytes_to_copy; | |
555 | pos -= bytes_to_copy; | |
556 | str -= bytes_to_copy; | |
557 | len -= bytes_to_copy; | |
558 | ||
559 | if (!kmapped_page || kpos != (pos & PAGE_MASK)) { | |
560 | struct page *page; | |
561 | ||
562 | page = get_arg_page(bprm, pos, 1); | |
1da177e4 | 563 | if (!page) { |
b6a2fea3 | 564 | ret = -E2BIG; |
1da177e4 LT |
565 | goto out; |
566 | } | |
1da177e4 | 567 | |
b6a2fea3 OW |
568 | if (kmapped_page) { |
569 | flush_kernel_dcache_page(kmapped_page); | |
1da177e4 | 570 | kunmap(kmapped_page); |
b6a2fea3 OW |
571 | put_arg_page(kmapped_page); |
572 | } | |
1da177e4 LT |
573 | kmapped_page = page; |
574 | kaddr = kmap(kmapped_page); | |
b6a2fea3 OW |
575 | kpos = pos & PAGE_MASK; |
576 | flush_arg_page(bprm, kpos, kmapped_page); | |
1da177e4 | 577 | } |
b6a2fea3 | 578 | if (copy_from_user(kaddr+offset, str, bytes_to_copy)) { |
1da177e4 LT |
579 | ret = -EFAULT; |
580 | goto out; | |
581 | } | |
1da177e4 LT |
582 | } |
583 | } | |
584 | ret = 0; | |
585 | out: | |
b6a2fea3 OW |
586 | if (kmapped_page) { |
587 | flush_kernel_dcache_page(kmapped_page); | |
1da177e4 | 588 | kunmap(kmapped_page); |
b6a2fea3 OW |
589 | put_arg_page(kmapped_page); |
590 | } | |
1da177e4 LT |
591 | return ret; |
592 | } | |
593 | ||
594 | /* | |
595 | * Like copy_strings, but get argv and its values from kernel memory. | |
596 | */ | |
ba2d0162 | 597 | int copy_strings_kernel(int argc, const char *const *__argv, |
d7627467 | 598 | struct linux_binprm *bprm) |
1da177e4 LT |
599 | { |
600 | int r; | |
601 | mm_segment_t oldfs = get_fs(); | |
ba2d0162 | 602 | struct user_arg_ptr argv = { |
0e028465 | 603 | .ptr.native = (const char __user *const __user *)__argv, |
ba2d0162 ON |
604 | }; |
605 | ||
1da177e4 | 606 | set_fs(KERNEL_DS); |
ba2d0162 | 607 | r = copy_strings(argc, argv, bprm); |
1da177e4 | 608 | set_fs(oldfs); |
ba2d0162 | 609 | |
1da177e4 LT |
610 | return r; |
611 | } | |
1da177e4 LT |
612 | EXPORT_SYMBOL(copy_strings_kernel); |
613 | ||
614 | #ifdef CONFIG_MMU | |
b6a2fea3 | 615 | |
1da177e4 | 616 | /* |
b6a2fea3 OW |
617 | * During bprm_mm_init(), we create a temporary stack at STACK_TOP_MAX. Once |
618 | * the binfmt code determines where the new stack should reside, we shift it to | |
619 | * its final location. The process proceeds as follows: | |
1da177e4 | 620 | * |
b6a2fea3 OW |
621 | * 1) Use shift to calculate the new vma endpoints. |
622 | * 2) Extend vma to cover both the old and new ranges. This ensures the | |
623 | * arguments passed to subsequent functions are consistent. | |
624 | * 3) Move vma's page tables to the new range. | |
625 | * 4) Free up any cleared pgd range. | |
626 | * 5) Shrink the vma to cover only the new range. | |
1da177e4 | 627 | */ |
b6a2fea3 | 628 | static int shift_arg_pages(struct vm_area_struct *vma, unsigned long shift) |
1da177e4 LT |
629 | { |
630 | struct mm_struct *mm = vma->vm_mm; | |
b6a2fea3 OW |
631 | unsigned long old_start = vma->vm_start; |
632 | unsigned long old_end = vma->vm_end; | |
633 | unsigned long length = old_end - old_start; | |
634 | unsigned long new_start = old_start - shift; | |
635 | unsigned long new_end = old_end - shift; | |
d16dfc55 | 636 | struct mmu_gather tlb; |
1da177e4 | 637 | |
b6a2fea3 | 638 | BUG_ON(new_start > new_end); |
1da177e4 | 639 | |
b6a2fea3 OW |
640 | /* |
641 | * ensure there are no vmas between where we want to go | |
642 | * and where we are | |
643 | */ | |
644 | if (vma != find_vma(mm, new_start)) | |
645 | return -EFAULT; | |
646 | ||
647 | /* | |
648 | * cover the whole range: [new_start, old_end) | |
649 | */ | |
5beb4930 RR |
650 | if (vma_adjust(vma, new_start, old_end, vma->vm_pgoff, NULL)) |
651 | return -ENOMEM; | |
b6a2fea3 OW |
652 | |
653 | /* | |
654 | * move the page tables downwards, on failure we rely on | |
655 | * process cleanup to remove whatever mess we made. | |
656 | */ | |
657 | if (length != move_page_tables(vma, old_start, | |
38a76013 | 658 | vma, new_start, length, false)) |
b6a2fea3 OW |
659 | return -ENOMEM; |
660 | ||
661 | lru_add_drain(); | |
2b047252 | 662 | tlb_gather_mmu(&tlb, mm, old_start, old_end); |
b6a2fea3 OW |
663 | if (new_end > old_start) { |
664 | /* | |
665 | * when the old and new regions overlap clear from new_end. | |
666 | */ | |
d16dfc55 | 667 | free_pgd_range(&tlb, new_end, old_end, new_end, |
6ee8630e | 668 | vma->vm_next ? vma->vm_next->vm_start : USER_PGTABLES_CEILING); |
b6a2fea3 OW |
669 | } else { |
670 | /* | |
671 | * otherwise, clean from old_start; this is done to not touch | |
672 | * the address space in [new_end, old_start) some architectures | |
673 | * have constraints on va-space that make this illegal (IA64) - | |
674 | * for the others its just a little faster. | |
675 | */ | |
d16dfc55 | 676 | free_pgd_range(&tlb, old_start, old_end, new_end, |
6ee8630e | 677 | vma->vm_next ? vma->vm_next->vm_start : USER_PGTABLES_CEILING); |
1da177e4 | 678 | } |
2b047252 | 679 | tlb_finish_mmu(&tlb, old_start, old_end); |
b6a2fea3 OW |
680 | |
681 | /* | |
5beb4930 | 682 | * Shrink the vma to just the new range. Always succeeds. |
b6a2fea3 OW |
683 | */ |
684 | vma_adjust(vma, new_start, new_end, vma->vm_pgoff, NULL); | |
685 | ||
686 | return 0; | |
1da177e4 LT |
687 | } |
688 | ||
b6a2fea3 OW |
689 | /* |
690 | * Finalizes the stack vm_area_struct. The flags and permissions are updated, | |
691 | * the stack is optionally relocated, and some extra space is added. | |
692 | */ | |
1da177e4 LT |
693 | int setup_arg_pages(struct linux_binprm *bprm, |
694 | unsigned long stack_top, | |
695 | int executable_stack) | |
696 | { | |
b6a2fea3 OW |
697 | unsigned long ret; |
698 | unsigned long stack_shift; | |
1da177e4 | 699 | struct mm_struct *mm = current->mm; |
b6a2fea3 OW |
700 | struct vm_area_struct *vma = bprm->vma; |
701 | struct vm_area_struct *prev = NULL; | |
702 | unsigned long vm_flags; | |
703 | unsigned long stack_base; | |
803bf5ec MN |
704 | unsigned long stack_size; |
705 | unsigned long stack_expand; | |
706 | unsigned long rlim_stack; | |
1da177e4 LT |
707 | |
708 | #ifdef CONFIG_STACK_GROWSUP | |
d71f290b | 709 | /* Limit stack size */ |
d554ed89 | 710 | stack_base = rlimit_max(RLIMIT_STACK); |
d71f290b JH |
711 | if (stack_base > STACK_SIZE_MAX) |
712 | stack_base = STACK_SIZE_MAX; | |
1da177e4 | 713 | |
d045c77c HD |
714 | /* Add space for stack randomization. */ |
715 | stack_base += (STACK_RND_MASK << PAGE_SHIFT); | |
716 | ||
b6a2fea3 OW |
717 | /* Make sure we didn't let the argument array grow too large. */ |
718 | if (vma->vm_end - vma->vm_start > stack_base) | |
719 | return -ENOMEM; | |
1da177e4 | 720 | |
b6a2fea3 | 721 | stack_base = PAGE_ALIGN(stack_top - stack_base); |
1da177e4 | 722 | |
b6a2fea3 OW |
723 | stack_shift = vma->vm_start - stack_base; |
724 | mm->arg_start = bprm->p - stack_shift; | |
725 | bprm->p = vma->vm_end - stack_shift; | |
1da177e4 | 726 | #else |
b6a2fea3 OW |
727 | stack_top = arch_align_stack(stack_top); |
728 | stack_top = PAGE_ALIGN(stack_top); | |
1b528181 RM |
729 | |
730 | if (unlikely(stack_top < mmap_min_addr) || | |
731 | unlikely(vma->vm_end - vma->vm_start >= stack_top - mmap_min_addr)) | |
732 | return -ENOMEM; | |
733 | ||
b6a2fea3 OW |
734 | stack_shift = vma->vm_end - stack_top; |
735 | ||
736 | bprm->p -= stack_shift; | |
1da177e4 | 737 | mm->arg_start = bprm->p; |
1da177e4 LT |
738 | #endif |
739 | ||
1da177e4 | 740 | if (bprm->loader) |
b6a2fea3 OW |
741 | bprm->loader -= stack_shift; |
742 | bprm->exec -= stack_shift; | |
1da177e4 | 743 | |
f268dfe9 MH |
744 | if (down_write_killable(&mm->mmap_sem)) |
745 | return -EINTR; | |
746 | ||
96a8e13e | 747 | vm_flags = VM_STACK_FLAGS; |
b6a2fea3 OW |
748 | |
749 | /* | |
750 | * Adjust stack execute permissions; explicitly enable for | |
751 | * EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X and leave alone | |
752 | * (arch default) otherwise. | |
753 | */ | |
754 | if (unlikely(executable_stack == EXSTACK_ENABLE_X)) | |
755 | vm_flags |= VM_EXEC; | |
756 | else if (executable_stack == EXSTACK_DISABLE_X) | |
757 | vm_flags &= ~VM_EXEC; | |
758 | vm_flags |= mm->def_flags; | |
a8bef8ff | 759 | vm_flags |= VM_STACK_INCOMPLETE_SETUP; |
b6a2fea3 OW |
760 | |
761 | ret = mprotect_fixup(vma, &prev, vma->vm_start, vma->vm_end, | |
762 | vm_flags); | |
763 | if (ret) | |
764 | goto out_unlock; | |
765 | BUG_ON(prev != vma); | |
766 | ||
767 | /* Move stack pages down in memory. */ | |
768 | if (stack_shift) { | |
769 | ret = shift_arg_pages(vma, stack_shift); | |
fc63cf23 AB |
770 | if (ret) |
771 | goto out_unlock; | |
1da177e4 LT |
772 | } |
773 | ||
a8bef8ff MG |
774 | /* mprotect_fixup is overkill to remove the temporary stack flags */ |
775 | vma->vm_flags &= ~VM_STACK_INCOMPLETE_SETUP; | |
776 | ||
5ef097dd | 777 | stack_expand = 131072UL; /* randomly 32*4k (or 2*64k) pages */ |
803bf5ec MN |
778 | stack_size = vma->vm_end - vma->vm_start; |
779 | /* | |
780 | * Align this down to a page boundary as expand_stack | |
781 | * will align it up. | |
782 | */ | |
783 | rlim_stack = rlimit(RLIMIT_STACK) & PAGE_MASK; | |
b6a2fea3 | 784 | #ifdef CONFIG_STACK_GROWSUP |
803bf5ec MN |
785 | if (stack_size + stack_expand > rlim_stack) |
786 | stack_base = vma->vm_start + rlim_stack; | |
787 | else | |
788 | stack_base = vma->vm_end + stack_expand; | |
b6a2fea3 | 789 | #else |
803bf5ec MN |
790 | if (stack_size + stack_expand > rlim_stack) |
791 | stack_base = vma->vm_end - rlim_stack; | |
792 | else | |
793 | stack_base = vma->vm_start - stack_expand; | |
b6a2fea3 | 794 | #endif |
3af9e859 | 795 | current->mm->start_stack = bprm->p; |
b6a2fea3 OW |
796 | ret = expand_stack(vma, stack_base); |
797 | if (ret) | |
798 | ret = -EFAULT; | |
799 | ||
800 | out_unlock: | |
1da177e4 | 801 | up_write(&mm->mmap_sem); |
fc63cf23 | 802 | return ret; |
1da177e4 | 803 | } |
1da177e4 LT |
804 | EXPORT_SYMBOL(setup_arg_pages); |
805 | ||
7e7ec6a9 NP |
806 | #else |
807 | ||
808 | /* | |
809 | * Transfer the program arguments and environment from the holding pages | |
810 | * onto the stack. The provided stack pointer is adjusted accordingly. | |
811 | */ | |
812 | int transfer_args_to_stack(struct linux_binprm *bprm, | |
813 | unsigned long *sp_location) | |
814 | { | |
815 | unsigned long index, stop, sp; | |
816 | int ret = 0; | |
817 | ||
818 | stop = bprm->p >> PAGE_SHIFT; | |
819 | sp = *sp_location; | |
820 | ||
821 | for (index = MAX_ARG_PAGES - 1; index >= stop; index--) { | |
822 | unsigned int offset = index == stop ? bprm->p & ~PAGE_MASK : 0; | |
823 | char *src = kmap(bprm->page[index]) + offset; | |
824 | sp -= PAGE_SIZE - offset; | |
825 | if (copy_to_user((void *) sp, src, PAGE_SIZE - offset) != 0) | |
826 | ret = -EFAULT; | |
827 | kunmap(bprm->page[index]); | |
828 | if (ret) | |
829 | goto out; | |
830 | } | |
831 | ||
832 | *sp_location = sp; | |
833 | ||
834 | out: | |
835 | return ret; | |
836 | } | |
837 | EXPORT_SYMBOL(transfer_args_to_stack); | |
838 | ||
1da177e4 LT |
839 | #endif /* CONFIG_MMU */ |
840 | ||
51f39a1f | 841 | static struct file *do_open_execat(int fd, struct filename *name, int flags) |
1da177e4 | 842 | { |
1da177e4 | 843 | struct file *file; |
e56b6a5d | 844 | int err; |
51f39a1f | 845 | struct open_flags open_exec_flags = { |
47c805dc | 846 | .open_flag = O_LARGEFILE | O_RDONLY | __FMODE_EXEC, |
62fb4a15 | 847 | .acc_mode = MAY_EXEC, |
f9652e10 AV |
848 | .intent = LOOKUP_OPEN, |
849 | .lookup_flags = LOOKUP_FOLLOW, | |
47c805dc | 850 | }; |
1da177e4 | 851 | |
51f39a1f DD |
852 | if ((flags & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0) |
853 | return ERR_PTR(-EINVAL); | |
854 | if (flags & AT_SYMLINK_NOFOLLOW) | |
855 | open_exec_flags.lookup_flags &= ~LOOKUP_FOLLOW; | |
856 | if (flags & AT_EMPTY_PATH) | |
857 | open_exec_flags.lookup_flags |= LOOKUP_EMPTY; | |
858 | ||
859 | file = do_filp_open(fd, name, &open_exec_flags); | |
6e8341a1 | 860 | if (IS_ERR(file)) |
e56b6a5d CH |
861 | goto out; |
862 | ||
863 | err = -EACCES; | |
496ad9aa | 864 | if (!S_ISREG(file_inode(file)->i_mode)) |
6e8341a1 | 865 | goto exit; |
e56b6a5d | 866 | |
90f8572b | 867 | if (path_noexec(&file->f_path)) |
6e8341a1 | 868 | goto exit; |
e56b6a5d CH |
869 | |
870 | err = deny_write_access(file); | |
6e8341a1 AV |
871 | if (err) |
872 | goto exit; | |
1da177e4 | 873 | |
51f39a1f DD |
874 | if (name->name[0] != '\0') |
875 | fsnotify_open(file); | |
876 | ||
7744c33f SJR |
877 | trace_open_exec(name->name); |
878 | ||
6e8341a1 | 879 | out: |
e56b6a5d CH |
880 | return file; |
881 | ||
6e8341a1 AV |
882 | exit: |
883 | fput(file); | |
e56b6a5d CH |
884 | return ERR_PTR(err); |
885 | } | |
c4ad8f98 LT |
886 | |
887 | struct file *open_exec(const char *name) | |
888 | { | |
51689104 PM |
889 | struct filename *filename = getname_kernel(name); |
890 | struct file *f = ERR_CAST(filename); | |
891 | ||
892 | if (!IS_ERR(filename)) { | |
893 | f = do_open_execat(AT_FDCWD, filename, 0); | |
894 | putname(filename); | |
895 | } | |
896 | return f; | |
c4ad8f98 | 897 | } |
1da177e4 LT |
898 | EXPORT_SYMBOL(open_exec); |
899 | ||
b44a7dfc | 900 | int kernel_read_file(struct file *file, void **buf, loff_t *size, |
bc8ca5b9 | 901 | loff_t max_size, enum kernel_read_file_id id) |
b44a7dfc MZ |
902 | { |
903 | loff_t i_size, pos; | |
904 | ssize_t bytes = 0; | |
905 | int ret; | |
906 | ||
907 | if (!S_ISREG(file_inode(file)->i_mode) || max_size < 0) | |
908 | return -EINVAL; | |
909 | ||
39eeb4fb MZ |
910 | ret = security_kernel_read_file(file, id); |
911 | if (ret) | |
912 | return ret; | |
913 | ||
39d637af DK |
914 | ret = deny_write_access(file); |
915 | if (ret) | |
916 | return ret; | |
917 | ||
b44a7dfc | 918 | i_size = i_size_read(file_inode(file)); |
39d637af DK |
919 | if (max_size > 0 && i_size > max_size) { |
920 | ret = -EFBIG; | |
921 | goto out; | |
922 | } | |
923 | if (i_size <= 0) { | |
924 | ret = -EINVAL; | |
925 | goto out; | |
926 | } | |
b44a7dfc | 927 | |
a098ecd2 SB |
928 | if (id != READING_FIRMWARE_PREALLOC_BUFFER) |
929 | *buf = vmalloc(i_size); | |
39d637af DK |
930 | if (!*buf) { |
931 | ret = -ENOMEM; | |
932 | goto out; | |
933 | } | |
b44a7dfc MZ |
934 | |
935 | pos = 0; | |
936 | while (pos < i_size) { | |
bdd1d2d3 | 937 | bytes = kernel_read(file, *buf + pos, i_size - pos, &pos); |
b44a7dfc MZ |
938 | if (bytes < 0) { |
939 | ret = bytes; | |
940 | goto out; | |
941 | } | |
942 | ||
943 | if (bytes == 0) | |
944 | break; | |
b44a7dfc MZ |
945 | } |
946 | ||
947 | if (pos != i_size) { | |
948 | ret = -EIO; | |
39d637af | 949 | goto out_free; |
b44a7dfc MZ |
950 | } |
951 | ||
bc8ca5b9 | 952 | ret = security_kernel_post_read_file(file, *buf, i_size, id); |
b44a7dfc MZ |
953 | if (!ret) |
954 | *size = pos; | |
955 | ||
39d637af | 956 | out_free: |
b44a7dfc | 957 | if (ret < 0) { |
a098ecd2 SB |
958 | if (id != READING_FIRMWARE_PREALLOC_BUFFER) { |
959 | vfree(*buf); | |
960 | *buf = NULL; | |
961 | } | |
b44a7dfc | 962 | } |
39d637af DK |
963 | |
964 | out: | |
965 | allow_write_access(file); | |
b44a7dfc MZ |
966 | return ret; |
967 | } | |
968 | EXPORT_SYMBOL_GPL(kernel_read_file); | |
969 | ||
711aab1d | 970 | int kernel_read_file_from_path(const char *path, void **buf, loff_t *size, |
09596b94 MZ |
971 | loff_t max_size, enum kernel_read_file_id id) |
972 | { | |
973 | struct file *file; | |
974 | int ret; | |
975 | ||
976 | if (!path || !*path) | |
977 | return -EINVAL; | |
978 | ||
979 | file = filp_open(path, O_RDONLY, 0); | |
980 | if (IS_ERR(file)) | |
981 | return PTR_ERR(file); | |
982 | ||
983 | ret = kernel_read_file(file, buf, size, max_size, id); | |
984 | fput(file); | |
985 | return ret; | |
986 | } | |
987 | EXPORT_SYMBOL_GPL(kernel_read_file_from_path); | |
988 | ||
b844f0ec MZ |
989 | int kernel_read_file_from_fd(int fd, void **buf, loff_t *size, loff_t max_size, |
990 | enum kernel_read_file_id id) | |
991 | { | |
992 | struct fd f = fdget(fd); | |
993 | int ret = -EBADF; | |
994 | ||
995 | if (!f.file) | |
996 | goto out; | |
997 | ||
998 | ret = kernel_read_file(f.file, buf, size, max_size, id); | |
999 | out: | |
1000 | fdput(f); | |
1001 | return ret; | |
1002 | } | |
1003 | EXPORT_SYMBOL_GPL(kernel_read_file_from_fd); | |
1004 | ||
3dc20cb2 AV |
1005 | ssize_t read_code(struct file *file, unsigned long addr, loff_t pos, size_t len) |
1006 | { | |
ec695579 | 1007 | ssize_t res = vfs_read(file, (void __user *)addr, len, &pos); |
3dc20cb2 AV |
1008 | if (res > 0) |
1009 | flush_icache_range(addr, addr + len); | |
1010 | return res; | |
1011 | } | |
1012 | EXPORT_SYMBOL(read_code); | |
1013 | ||
1da177e4 LT |
1014 | static int exec_mmap(struct mm_struct *mm) |
1015 | { | |
1016 | struct task_struct *tsk; | |
615d6e87 | 1017 | struct mm_struct *old_mm, *active_mm; |
1da177e4 LT |
1018 | |
1019 | /* Notify parent that we're no longer interested in the old VM */ | |
1020 | tsk = current; | |
1021 | old_mm = current->mm; | |
1022 | mm_release(tsk, old_mm); | |
1023 | ||
1024 | if (old_mm) { | |
4fe7efdb | 1025 | sync_mm_rss(old_mm); |
1da177e4 LT |
1026 | /* |
1027 | * Make sure that if there is a core dump in progress | |
1028 | * for the old mm, we get out and die instead of going | |
1029 | * through with the exec. We must hold mmap_sem around | |
999d9fc1 | 1030 | * checking core_state and changing tsk->mm. |
1da177e4 LT |
1031 | */ |
1032 | down_read(&old_mm->mmap_sem); | |
999d9fc1 | 1033 | if (unlikely(old_mm->core_state)) { |
1da177e4 LT |
1034 | up_read(&old_mm->mmap_sem); |
1035 | return -EINTR; | |
1036 | } | |
1037 | } | |
1038 | task_lock(tsk); | |
1039 | active_mm = tsk->active_mm; | |
1040 | tsk->mm = mm; | |
1041 | tsk->active_mm = mm; | |
1042 | activate_mm(active_mm, mm); | |
615d6e87 DB |
1043 | tsk->mm->vmacache_seqnum = 0; |
1044 | vmacache_flush(tsk); | |
1da177e4 | 1045 | task_unlock(tsk); |
1da177e4 LT |
1046 | if (old_mm) { |
1047 | up_read(&old_mm->mmap_sem); | |
7dddb12c | 1048 | BUG_ON(active_mm != old_mm); |
701085b2 | 1049 | setmax_mm_hiwater_rss(&tsk->signal->maxrss, old_mm); |
31a78f23 | 1050 | mm_update_next_owner(old_mm); |
1da177e4 LT |
1051 | mmput(old_mm); |
1052 | return 0; | |
1053 | } | |
1054 | mmdrop(active_mm); | |
1055 | return 0; | |
1056 | } | |
1057 | ||
1058 | /* | |
1059 | * This function makes sure the current process has its own signal table, | |
1060 | * so that flush_signal_handlers can later reset the handlers without | |
1061 | * disturbing other processes. (Other processes might share the signal | |
1062 | * table via the CLONE_SIGHAND option to clone().) | |
1063 | */ | |
858119e1 | 1064 | static int de_thread(struct task_struct *tsk) |
1da177e4 LT |
1065 | { |
1066 | struct signal_struct *sig = tsk->signal; | |
b2c903b8 | 1067 | struct sighand_struct *oldsighand = tsk->sighand; |
1da177e4 | 1068 | spinlock_t *lock = &oldsighand->siglock; |
1da177e4 | 1069 | |
aafe6c2a | 1070 | if (thread_group_empty(tsk)) |
1da177e4 LT |
1071 | goto no_thread_group; |
1072 | ||
1073 | /* | |
1074 | * Kill all other threads in the thread group. | |
1da177e4 | 1075 | */ |
1da177e4 | 1076 | spin_lock_irq(lock); |
ed5d2cac | 1077 | if (signal_group_exit(sig)) { |
1da177e4 LT |
1078 | /* |
1079 | * Another group action in progress, just | |
1080 | * return so that the signal is processed. | |
1081 | */ | |
1082 | spin_unlock_irq(lock); | |
1da177e4 LT |
1083 | return -EAGAIN; |
1084 | } | |
d344193a | 1085 | |
ed5d2cac | 1086 | sig->group_exit_task = tsk; |
d344193a ON |
1087 | sig->notify_count = zap_other_threads(tsk); |
1088 | if (!thread_group_leader(tsk)) | |
1089 | sig->notify_count--; | |
1da177e4 | 1090 | |
d344193a | 1091 | while (sig->notify_count) { |
d5bbd43d | 1092 | __set_current_state(TASK_KILLABLE); |
1da177e4 LT |
1093 | spin_unlock_irq(lock); |
1094 | schedule(); | |
d5bbd43d ON |
1095 | if (unlikely(__fatal_signal_pending(tsk))) |
1096 | goto killed; | |
1da177e4 LT |
1097 | spin_lock_irq(lock); |
1098 | } | |
1da177e4 LT |
1099 | spin_unlock_irq(lock); |
1100 | ||
1101 | /* | |
1102 | * At this point all other threads have exited, all we have to | |
1103 | * do is to wait for the thread group leader to become inactive, | |
1104 | * and to assume its PID: | |
1105 | */ | |
aafe6c2a | 1106 | if (!thread_group_leader(tsk)) { |
8187926b | 1107 | struct task_struct *leader = tsk->group_leader; |
6db840fa | 1108 | |
6db840fa | 1109 | for (;;) { |
780de9dd | 1110 | cgroup_threadgroup_change_begin(tsk); |
6db840fa | 1111 | write_lock_irq(&tasklist_lock); |
dfcce791 KT |
1112 | /* |
1113 | * Do this under tasklist_lock to ensure that | |
1114 | * exit_notify() can't miss ->group_exit_task | |
1115 | */ | |
1116 | sig->notify_count = -1; | |
6db840fa ON |
1117 | if (likely(leader->exit_state)) |
1118 | break; | |
d5bbd43d | 1119 | __set_current_state(TASK_KILLABLE); |
6db840fa | 1120 | write_unlock_irq(&tasklist_lock); |
780de9dd | 1121 | cgroup_threadgroup_change_end(tsk); |
6db840fa | 1122 | schedule(); |
d5bbd43d ON |
1123 | if (unlikely(__fatal_signal_pending(tsk))) |
1124 | goto killed; | |
6db840fa | 1125 | } |
1da177e4 | 1126 | |
f5e90281 RM |
1127 | /* |
1128 | * The only record we have of the real-time age of a | |
1129 | * process, regardless of execs it's done, is start_time. | |
1130 | * All the past CPU time is accumulated in signal_struct | |
1131 | * from sister threads now dead. But in this non-leader | |
1132 | * exec, nothing survives from the original leader thread, | |
1133 | * whose birth marks the true age of this process now. | |
1134 | * When we take on its identity by switching to its PID, we | |
1135 | * also take its birthdate (always earlier than our own). | |
1136 | */ | |
aafe6c2a | 1137 | tsk->start_time = leader->start_time; |
266b7a02 | 1138 | tsk->real_start_time = leader->real_start_time; |
f5e90281 | 1139 | |
bac0abd6 PE |
1140 | BUG_ON(!same_thread_group(leader, tsk)); |
1141 | BUG_ON(has_group_leader_pid(tsk)); | |
1da177e4 LT |
1142 | /* |
1143 | * An exec() starts a new thread group with the | |
1144 | * TGID of the previous thread group. Rehash the | |
1145 | * two threads with a switched PID, and release | |
1146 | * the former thread group leader: | |
1147 | */ | |
d73d6529 EB |
1148 | |
1149 | /* Become a process group leader with the old leader's pid. | |
c18258c6 EB |
1150 | * The old leader becomes a thread of the this thread group. |
1151 | * Note: The old leader also uses this pid until release_task | |
d73d6529 EB |
1152 | * is called. Odd but simple and correct. |
1153 | */ | |
aafe6c2a | 1154 | tsk->pid = leader->pid; |
3f418548 | 1155 | change_pid(tsk, PIDTYPE_PID, task_pid(leader)); |
aafe6c2a EB |
1156 | transfer_pid(leader, tsk, PIDTYPE_PGID); |
1157 | transfer_pid(leader, tsk, PIDTYPE_SID); | |
9cd80bbb | 1158 | |
aafe6c2a | 1159 | list_replace_rcu(&leader->tasks, &tsk->tasks); |
9cd80bbb | 1160 | list_replace_init(&leader->sibling, &tsk->sibling); |
1da177e4 | 1161 | |
aafe6c2a EB |
1162 | tsk->group_leader = tsk; |
1163 | leader->group_leader = tsk; | |
de12a787 | 1164 | |
aafe6c2a | 1165 | tsk->exit_signal = SIGCHLD; |
087806b1 | 1166 | leader->exit_signal = -1; |
962b564c ON |
1167 | |
1168 | BUG_ON(leader->exit_state != EXIT_ZOMBIE); | |
1169 | leader->exit_state = EXIT_DEAD; | |
eac1b5e5 ON |
1170 | |
1171 | /* | |
1172 | * We are going to release_task()->ptrace_unlink() silently, | |
1173 | * the tracer can sleep in do_wait(). EXIT_DEAD guarantees | |
1174 | * the tracer wont't block again waiting for this thread. | |
1175 | */ | |
1176 | if (unlikely(leader->ptrace)) | |
1177 | __wake_up_parent(leader, leader->parent); | |
1da177e4 | 1178 | write_unlock_irq(&tasklist_lock); |
780de9dd | 1179 | cgroup_threadgroup_change_end(tsk); |
8187926b ON |
1180 | |
1181 | release_task(leader); | |
ed5d2cac | 1182 | } |
1da177e4 | 1183 | |
6db840fa ON |
1184 | sig->group_exit_task = NULL; |
1185 | sig->notify_count = 0; | |
1da177e4 LT |
1186 | |
1187 | no_thread_group: | |
e6368253 ON |
1188 | /* we have changed execution domain */ |
1189 | tsk->exit_signal = SIGCHLD; | |
1190 | ||
baa73d9e | 1191 | #ifdef CONFIG_POSIX_TIMERS |
1da177e4 | 1192 | exit_itimers(sig); |
cbaffba1 | 1193 | flush_itimer_signals(); |
baa73d9e | 1194 | #endif |
329f7dba | 1195 | |
b2c903b8 ON |
1196 | if (atomic_read(&oldsighand->count) != 1) { |
1197 | struct sighand_struct *newsighand; | |
1da177e4 | 1198 | /* |
b2c903b8 ON |
1199 | * This ->sighand is shared with the CLONE_SIGHAND |
1200 | * but not CLONE_THREAD task, switch to the new one. | |
1da177e4 | 1201 | */ |
b2c903b8 ON |
1202 | newsighand = kmem_cache_alloc(sighand_cachep, GFP_KERNEL); |
1203 | if (!newsighand) | |
1204 | return -ENOMEM; | |
1205 | ||
1da177e4 LT |
1206 | atomic_set(&newsighand->count, 1); |
1207 | memcpy(newsighand->action, oldsighand->action, | |
1208 | sizeof(newsighand->action)); | |
1209 | ||
1210 | write_lock_irq(&tasklist_lock); | |
1211 | spin_lock(&oldsighand->siglock); | |
aafe6c2a | 1212 | rcu_assign_pointer(tsk->sighand, newsighand); |
1da177e4 LT |
1213 | spin_unlock(&oldsighand->siglock); |
1214 | write_unlock_irq(&tasklist_lock); | |
1215 | ||
fba2afaa | 1216 | __cleanup_sighand(oldsighand); |
1da177e4 LT |
1217 | } |
1218 | ||
aafe6c2a | 1219 | BUG_ON(!thread_group_leader(tsk)); |
1da177e4 | 1220 | return 0; |
d5bbd43d ON |
1221 | |
1222 | killed: | |
1223 | /* protects against exit_notify() and __exit_signal() */ | |
1224 | read_lock(&tasklist_lock); | |
1225 | sig->group_exit_task = NULL; | |
1226 | sig->notify_count = 0; | |
1227 | read_unlock(&tasklist_lock); | |
1228 | return -EAGAIN; | |
1da177e4 | 1229 | } |
0840a90d | 1230 | |
3756f640 | 1231 | char *__get_task_comm(char *buf, size_t buf_size, struct task_struct *tsk) |
1da177e4 | 1232 | { |
1da177e4 | 1233 | task_lock(tsk); |
3756f640 | 1234 | strncpy(buf, tsk->comm, buf_size); |
1da177e4 | 1235 | task_unlock(tsk); |
59714d65 | 1236 | return buf; |
1da177e4 | 1237 | } |
3756f640 | 1238 | EXPORT_SYMBOL_GPL(__get_task_comm); |
1da177e4 | 1239 | |
6a6d27de AV |
1240 | /* |
1241 | * These functions flushes out all traces of the currently running executable | |
1242 | * so that a new one can be started | |
1243 | */ | |
1244 | ||
82b89778 | 1245 | void __set_task_comm(struct task_struct *tsk, const char *buf, bool exec) |
1da177e4 LT |
1246 | { |
1247 | task_lock(tsk); | |
43d2b113 | 1248 | trace_task_rename(tsk, buf); |
1da177e4 LT |
1249 | strlcpy(tsk->comm, buf, sizeof(tsk->comm)); |
1250 | task_unlock(tsk); | |
82b89778 | 1251 | perf_event_comm(tsk, exec); |
1da177e4 LT |
1252 | } |
1253 | ||
a9208e42 KC |
1254 | /* |
1255 | * Calling this is the point of no return. None of the failures will be | |
1256 | * seen by userspace since either the process is already taking a fatal | |
1257 | * signal (via de_thread() or coredump), or will have SEGV raised | |
1258 | * (after exec_mmap()) by search_binary_handlers (see below). | |
1259 | */ | |
1da177e4 LT |
1260 | int flush_old_exec(struct linux_binprm * bprm) |
1261 | { | |
221af7f8 | 1262 | int retval; |
1da177e4 LT |
1263 | |
1264 | /* | |
1265 | * Make sure we have a private signal table and that | |
1266 | * we are unassociated from the previous thread group. | |
1267 | */ | |
1268 | retval = de_thread(current); | |
1269 | if (retval) | |
1270 | goto out; | |
1271 | ||
6e399cd1 DB |
1272 | /* |
1273 | * Must be called _before_ exec_mmap() as bprm->mm is | |
1274 | * not visibile until then. This also enables the update | |
1275 | * to be lockless. | |
1276 | */ | |
925d1c40 | 1277 | set_mm_exe_file(bprm->mm, bprm->file); |
6e399cd1 | 1278 | |
1da177e4 LT |
1279 | /* |
1280 | * Release all of the old mmap stuff | |
1281 | */ | |
3c77f845 | 1282 | acct_arg_size(bprm, 0); |
1da177e4 LT |
1283 | retval = exec_mmap(bprm->mm); |
1284 | if (retval) | |
fd8328be | 1285 | goto out; |
1da177e4 | 1286 | |
a9208e42 KC |
1287 | /* |
1288 | * After clearing bprm->mm (to mark that current is using the | |
1289 | * prepared mm now), we have nothing left of the original | |
1290 | * process. If anything from here on returns an error, the check | |
1291 | * in search_binary_handler() will SEGV current. | |
1292 | */ | |
1293 | bprm->mm = NULL; | |
7ab02af4 | 1294 | |
dac853ae | 1295 | set_fs(USER_DS); |
b88fae64 ZY |
1296 | current->flags &= ~(PF_RANDOMIZE | PF_FORKNOEXEC | PF_KTHREAD | |
1297 | PF_NOFREEZE | PF_NO_SETAFFINITY); | |
7ab02af4 LT |
1298 | flush_thread(); |
1299 | current->personality &= ~bprm->per_clear; | |
1300 | ||
613cc2b6 AS |
1301 | /* |
1302 | * We have to apply CLOEXEC before we change whether the process is | |
1303 | * dumpable (in setup_new_exec) to avoid a race with a process in userspace | |
1304 | * trying to access the should-be-closed file descriptors of a process | |
1305 | * undergoing exec(2). | |
1306 | */ | |
1307 | do_close_on_exec(current->files); | |
221af7f8 LT |
1308 | return 0; |
1309 | ||
1310 | out: | |
1311 | return retval; | |
1312 | } | |
1313 | EXPORT_SYMBOL(flush_old_exec); | |
1314 | ||
1b5d783c AV |
1315 | void would_dump(struct linux_binprm *bprm, struct file *file) |
1316 | { | |
f84df2a6 EB |
1317 | struct inode *inode = file_inode(file); |
1318 | if (inode_permission(inode, MAY_READ) < 0) { | |
1319 | struct user_namespace *old, *user_ns; | |
1b5d783c | 1320 | bprm->interp_flags |= BINPRM_FLAGS_ENFORCE_NONDUMP; |
f84df2a6 EB |
1321 | |
1322 | /* Ensure mm->user_ns contains the executable */ | |
1323 | user_ns = old = bprm->mm->user_ns; | |
1324 | while ((user_ns != &init_user_ns) && | |
1325 | !privileged_wrt_inode_uidgid(user_ns, inode)) | |
1326 | user_ns = user_ns->parent; | |
1327 | ||
1328 | if (old != user_ns) { | |
1329 | bprm->mm->user_ns = get_user_ns(user_ns); | |
1330 | put_user_ns(old); | |
1331 | } | |
1332 | } | |
1b5d783c AV |
1333 | } |
1334 | EXPORT_SYMBOL(would_dump); | |
1335 | ||
221af7f8 LT |
1336 | void setup_new_exec(struct linux_binprm * bprm) |
1337 | { | |
46d98eb4 KC |
1338 | /* |
1339 | * Once here, prepare_binrpm() will not be called any more, so | |
1340 | * the final state of setuid/setgid/fscaps can be merged into the | |
1341 | * secureexec flag. | |
1342 | */ | |
1343 | bprm->secureexec |= bprm->cap_elevated; | |
1344 | ||
64701dee | 1345 | if (bprm->secureexec) { |
fe8993b3 KC |
1346 | /* Make sure parent cannot signal privileged process. */ |
1347 | current->pdeath_signal = 0; | |
1348 | ||
64701dee KC |
1349 | /* |
1350 | * For secureexec, reset the stack limit to sane default to | |
1351 | * avoid bad behavior from the prior rlimits. This has to | |
1352 | * happen before arch_pick_mmap_layout(), which examines | |
1353 | * RLIMIT_STACK, but after the point of no return to avoid | |
779f4e1c | 1354 | * needing to clean up the change on failure. |
64701dee KC |
1355 | */ |
1356 | if (current->signal->rlim[RLIMIT_STACK].rlim_cur > _STK_LIM) | |
1357 | current->signal->rlim[RLIMIT_STACK].rlim_cur = _STK_LIM; | |
1358 | } | |
1359 | ||
221af7f8 | 1360 | arch_pick_mmap_layout(current->mm); |
1da177e4 | 1361 | |
1da177e4 LT |
1362 | current->sas_ss_sp = current->sas_ss_size = 0; |
1363 | ||
e816c201 KC |
1364 | /* |
1365 | * Figure out dumpability. Note that this checking only of current | |
1366 | * is wrong, but userspace depends on it. This should be testing | |
1367 | * bprm->secureexec instead. | |
1368 | */ | |
473d8963 | 1369 | if (bprm->interp_flags & BINPRM_FLAGS_ENFORCE_NONDUMP || |
e816c201 KC |
1370 | !(uid_eq(current_euid(), current_uid()) && |
1371 | gid_eq(current_egid(), current_gid()))) | |
6c5d5238 | 1372 | set_dumpable(current->mm, suid_dumpable); |
473d8963 KC |
1373 | else |
1374 | set_dumpable(current->mm, SUID_DUMP_USER); | |
d6e71144 | 1375 | |
e9ea1e7f | 1376 | arch_setup_new_exec(); |
e041e328 | 1377 | perf_event_exec(); |
82b89778 | 1378 | __set_task_comm(current, kbasename(bprm->filename), true); |
1da177e4 | 1379 | |
0551fbd2 BH |
1380 | /* Set the new mm task size. We have to do that late because it may |
1381 | * depend on TIF_32BIT which is only updated in flush_thread() on | |
1382 | * some architectures like powerpc | |
1383 | */ | |
1384 | current->mm->task_size = TASK_SIZE; | |
1385 | ||
1da177e4 LT |
1386 | /* An exec changes our domain. We are no longer part of the thread |
1387 | group */ | |
1da177e4 | 1388 | current->self_exec_id++; |
1da177e4 | 1389 | flush_signal_handlers(current, 0); |
1da177e4 | 1390 | } |
221af7f8 | 1391 | EXPORT_SYMBOL(setup_new_exec); |
1da177e4 | 1392 | |
a2a8474c ON |
1393 | /* |
1394 | * Prepare credentials and lock ->cred_guard_mutex. | |
1395 | * install_exec_creds() commits the new creds and drops the lock. | |
1396 | * Or, if exec fails before, free_bprm() should release ->cred and | |
1397 | * and unlock. | |
1398 | */ | |
1399 | int prepare_bprm_creds(struct linux_binprm *bprm) | |
1400 | { | |
9b1bf12d | 1401 | if (mutex_lock_interruptible(¤t->signal->cred_guard_mutex)) |
a2a8474c ON |
1402 | return -ERESTARTNOINTR; |
1403 | ||
1404 | bprm->cred = prepare_exec_creds(); | |
1405 | if (likely(bprm->cred)) | |
1406 | return 0; | |
1407 | ||
9b1bf12d | 1408 | mutex_unlock(¤t->signal->cred_guard_mutex); |
a2a8474c ON |
1409 | return -ENOMEM; |
1410 | } | |
1411 | ||
c4ad8f98 | 1412 | static void free_bprm(struct linux_binprm *bprm) |
a2a8474c ON |
1413 | { |
1414 | free_arg_pages(bprm); | |
1415 | if (bprm->cred) { | |
9b1bf12d | 1416 | mutex_unlock(¤t->signal->cred_guard_mutex); |
a2a8474c ON |
1417 | abort_creds(bprm->cred); |
1418 | } | |
63e46b95 ON |
1419 | if (bprm->file) { |
1420 | allow_write_access(bprm->file); | |
1421 | fput(bprm->file); | |
1422 | } | |
b66c5984 KC |
1423 | /* If a binfmt changed the interp, free it. */ |
1424 | if (bprm->interp != bprm->filename) | |
1425 | kfree(bprm->interp); | |
a2a8474c ON |
1426 | kfree(bprm); |
1427 | } | |
1428 | ||
c2315c18 | 1429 | int bprm_change_interp(const char *interp, struct linux_binprm *bprm) |
b66c5984 KC |
1430 | { |
1431 | /* If a binfmt changed the interp, free it first. */ | |
1432 | if (bprm->interp != bprm->filename) | |
1433 | kfree(bprm->interp); | |
1434 | bprm->interp = kstrdup(interp, GFP_KERNEL); | |
1435 | if (!bprm->interp) | |
1436 | return -ENOMEM; | |
1437 | return 0; | |
1438 | } | |
1439 | EXPORT_SYMBOL(bprm_change_interp); | |
1440 | ||
a6f76f23 DH |
1441 | /* |
1442 | * install the new credentials for this executable | |
1443 | */ | |
1444 | void install_exec_creds(struct linux_binprm *bprm) | |
1445 | { | |
1446 | security_bprm_committing_creds(bprm); | |
1447 | ||
1448 | commit_creds(bprm->cred); | |
1449 | bprm->cred = NULL; | |
2976b10f SE |
1450 | |
1451 | /* | |
1452 | * Disable monitoring for regular users | |
1453 | * when executing setuid binaries. Must | |
1454 | * wait until new credentials are committed | |
1455 | * by commit_creds() above | |
1456 | */ | |
1457 | if (get_dumpable(current->mm) != SUID_DUMP_USER) | |
1458 | perf_event_exit_task(current); | |
a2a8474c ON |
1459 | /* |
1460 | * cred_guard_mutex must be held at least to this point to prevent | |
a6f76f23 | 1461 | * ptrace_attach() from altering our determination of the task's |
a2a8474c ON |
1462 | * credentials; any time after this it may be unlocked. |
1463 | */ | |
a6f76f23 | 1464 | security_bprm_committed_creds(bprm); |
9b1bf12d | 1465 | mutex_unlock(¤t->signal->cred_guard_mutex); |
a6f76f23 DH |
1466 | } |
1467 | EXPORT_SYMBOL(install_exec_creds); | |
1468 | ||
1469 | /* | |
1470 | * determine how safe it is to execute the proposed program | |
9b1bf12d | 1471 | * - the caller must hold ->cred_guard_mutex to protect against |
c2e1f2e3 | 1472 | * PTRACE_ATTACH or seccomp thread-sync |
a6f76f23 | 1473 | */ |
9e00cdb0 | 1474 | static void check_unsafe_exec(struct linux_binprm *bprm) |
a6f76f23 | 1475 | { |
0bf2f3ae | 1476 | struct task_struct *p = current, *t; |
f1191b50 | 1477 | unsigned n_fs; |
a6f76f23 | 1478 | |
9227dd2a EB |
1479 | if (p->ptrace) |
1480 | bprm->unsafe |= LSM_UNSAFE_PTRACE; | |
a6f76f23 | 1481 | |
259e5e6c AL |
1482 | /* |
1483 | * This isn't strictly necessary, but it makes it harder for LSMs to | |
1484 | * mess up. | |
1485 | */ | |
1d4457f9 | 1486 | if (task_no_new_privs(current)) |
259e5e6c AL |
1487 | bprm->unsafe |= LSM_UNSAFE_NO_NEW_PRIVS; |
1488 | ||
83f62a2e | 1489 | t = p; |
0bf2f3ae | 1490 | n_fs = 1; |
2a4419b5 | 1491 | spin_lock(&p->fs->lock); |
437f7fdb | 1492 | rcu_read_lock(); |
83f62a2e | 1493 | while_each_thread(p, t) { |
0bf2f3ae DH |
1494 | if (t->fs == p->fs) |
1495 | n_fs++; | |
0bf2f3ae | 1496 | } |
437f7fdb | 1497 | rcu_read_unlock(); |
0bf2f3ae | 1498 | |
9e00cdb0 | 1499 | if (p->fs->users > n_fs) |
a6f76f23 | 1500 | bprm->unsafe |= LSM_UNSAFE_SHARE; |
9e00cdb0 ON |
1501 | else |
1502 | p->fs->in_exec = 1; | |
2a4419b5 | 1503 | spin_unlock(&p->fs->lock); |
a6f76f23 DH |
1504 | } |
1505 | ||
8b01fc86 JH |
1506 | static void bprm_fill_uid(struct linux_binprm *bprm) |
1507 | { | |
1508 | struct inode *inode; | |
1509 | unsigned int mode; | |
1510 | kuid_t uid; | |
1511 | kgid_t gid; | |
1512 | ||
cb6fd68f KC |
1513 | /* |
1514 | * Since this can be called multiple times (via prepare_binprm), | |
1515 | * we must clear any previous work done when setting set[ug]id | |
1516 | * bits from any earlier bprm->file uses (for example when run | |
1517 | * first for a setuid script then again for its interpreter). | |
1518 | */ | |
8b01fc86 JH |
1519 | bprm->cred->euid = current_euid(); |
1520 | bprm->cred->egid = current_egid(); | |
1521 | ||
bf31f6b5 | 1522 | if (path_nosuid(&bprm->file->f_path)) |
8b01fc86 JH |
1523 | return; |
1524 | ||
1525 | if (task_no_new_privs(current)) | |
1526 | return; | |
1527 | ||
fea6d2a6 | 1528 | inode = bprm->file->f_path.dentry->d_inode; |
8b01fc86 JH |
1529 | mode = READ_ONCE(inode->i_mode); |
1530 | if (!(mode & (S_ISUID|S_ISGID))) | |
1531 | return; | |
1532 | ||
1533 | /* Be careful if suid/sgid is set */ | |
5955102c | 1534 | inode_lock(inode); |
8b01fc86 JH |
1535 | |
1536 | /* reload atomically mode/uid/gid now that lock held */ | |
1537 | mode = inode->i_mode; | |
1538 | uid = inode->i_uid; | |
1539 | gid = inode->i_gid; | |
5955102c | 1540 | inode_unlock(inode); |
8b01fc86 JH |
1541 | |
1542 | /* We ignore suid/sgid if there are no mappings for them in the ns */ | |
1543 | if (!kuid_has_mapping(bprm->cred->user_ns, uid) || | |
1544 | !kgid_has_mapping(bprm->cred->user_ns, gid)) | |
1545 | return; | |
1546 | ||
1547 | if (mode & S_ISUID) { | |
1548 | bprm->per_clear |= PER_CLEAR_ON_SETID; | |
1549 | bprm->cred->euid = uid; | |
1550 | } | |
1551 | ||
1552 | if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) { | |
1553 | bprm->per_clear |= PER_CLEAR_ON_SETID; | |
1554 | bprm->cred->egid = gid; | |
1555 | } | |
1556 | } | |
1557 | ||
9e00cdb0 ON |
1558 | /* |
1559 | * Fill the binprm structure from the inode. | |
1da177e4 | 1560 | * Check permissions, then read the first 128 (BINPRM_BUF_SIZE) bytes |
a6f76f23 DH |
1561 | * |
1562 | * This may be called multiple times for binary chains (scripts for example). | |
1da177e4 LT |
1563 | */ |
1564 | int prepare_binprm(struct linux_binprm *bprm) | |
1565 | { | |
1da177e4 | 1566 | int retval; |
bdd1d2d3 | 1567 | loff_t pos = 0; |
1da177e4 | 1568 | |
8b01fc86 | 1569 | bprm_fill_uid(bprm); |
1da177e4 LT |
1570 | |
1571 | /* fill in binprm security blob */ | |
a6f76f23 | 1572 | retval = security_bprm_set_creds(bprm); |
1da177e4 LT |
1573 | if (retval) |
1574 | return retval; | |
ddb4a144 | 1575 | bprm->called_set_creds = 1; |
1da177e4 | 1576 | |
a6f76f23 | 1577 | memset(bprm->buf, 0, BINPRM_BUF_SIZE); |
bdd1d2d3 | 1578 | return kernel_read(bprm->file, bprm->buf, BINPRM_BUF_SIZE, &pos); |
1da177e4 LT |
1579 | } |
1580 | ||
1581 | EXPORT_SYMBOL(prepare_binprm); | |
1582 | ||
4fc75ff4 NP |
1583 | /* |
1584 | * Arguments are '\0' separated strings found at the location bprm->p | |
1585 | * points to; chop off the first by relocating brpm->p to right after | |
1586 | * the first '\0' encountered. | |
1587 | */ | |
b6a2fea3 | 1588 | int remove_arg_zero(struct linux_binprm *bprm) |
1da177e4 | 1589 | { |
b6a2fea3 OW |
1590 | int ret = 0; |
1591 | unsigned long offset; | |
1592 | char *kaddr; | |
1593 | struct page *page; | |
4fc75ff4 | 1594 | |
b6a2fea3 OW |
1595 | if (!bprm->argc) |
1596 | return 0; | |
1da177e4 | 1597 | |
b6a2fea3 OW |
1598 | do { |
1599 | offset = bprm->p & ~PAGE_MASK; | |
1600 | page = get_arg_page(bprm, bprm->p, 0); | |
1601 | if (!page) { | |
1602 | ret = -EFAULT; | |
1603 | goto out; | |
1604 | } | |
e8e3c3d6 | 1605 | kaddr = kmap_atomic(page); |
4fc75ff4 | 1606 | |
b6a2fea3 OW |
1607 | for (; offset < PAGE_SIZE && kaddr[offset]; |
1608 | offset++, bprm->p++) | |
1609 | ; | |
4fc75ff4 | 1610 | |
e8e3c3d6 | 1611 | kunmap_atomic(kaddr); |
b6a2fea3 | 1612 | put_arg_page(page); |
b6a2fea3 | 1613 | } while (offset == PAGE_SIZE); |
4fc75ff4 | 1614 | |
b6a2fea3 OW |
1615 | bprm->p++; |
1616 | bprm->argc--; | |
1617 | ret = 0; | |
4fc75ff4 | 1618 | |
b6a2fea3 OW |
1619 | out: |
1620 | return ret; | |
1da177e4 | 1621 | } |
1da177e4 LT |
1622 | EXPORT_SYMBOL(remove_arg_zero); |
1623 | ||
cb7b6b1c | 1624 | #define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e)) |
1da177e4 LT |
1625 | /* |
1626 | * cycle the list of binary formats handler, until one recognizes the image | |
1627 | */ | |
3c456bfc | 1628 | int search_binary_handler(struct linux_binprm *bprm) |
1da177e4 | 1629 | { |
cb7b6b1c | 1630 | bool need_retry = IS_ENABLED(CONFIG_MODULES); |
1da177e4 | 1631 | struct linux_binfmt *fmt; |
cb7b6b1c | 1632 | int retval; |
1da177e4 | 1633 | |
d7402698 | 1634 | /* This allows 4 levels of binfmt rewrites before failing hard. */ |
131b2f9f | 1635 | if (bprm->recursion_depth > 5) |
d7402698 KC |
1636 | return -ELOOP; |
1637 | ||
1da177e4 LT |
1638 | retval = security_bprm_check(bprm); |
1639 | if (retval) | |
1640 | return retval; | |
1641 | ||
1da177e4 | 1642 | retval = -ENOENT; |
cb7b6b1c ON |
1643 | retry: |
1644 | read_lock(&binfmt_lock); | |
1645 | list_for_each_entry(fmt, &formats, lh) { | |
1646 | if (!try_module_get(fmt->module)) | |
1647 | continue; | |
1648 | read_unlock(&binfmt_lock); | |
1649 | bprm->recursion_depth++; | |
1650 | retval = fmt->load_binary(bprm); | |
19d860a1 AV |
1651 | read_lock(&binfmt_lock); |
1652 | put_binfmt(fmt); | |
cb7b6b1c | 1653 | bprm->recursion_depth--; |
19d860a1 AV |
1654 | if (retval < 0 && !bprm->mm) { |
1655 | /* we got to flush_old_exec() and failed after it */ | |
1656 | read_unlock(&binfmt_lock); | |
1657 | force_sigsegv(SIGSEGV, current); | |
1658 | return retval; | |
1659 | } | |
1660 | if (retval != -ENOEXEC || !bprm->file) { | |
1661 | read_unlock(&binfmt_lock); | |
cb7b6b1c | 1662 | return retval; |
1da177e4 | 1663 | } |
1da177e4 | 1664 | } |
cb7b6b1c ON |
1665 | read_unlock(&binfmt_lock); |
1666 | ||
19d860a1 | 1667 | if (need_retry) { |
cb7b6b1c ON |
1668 | if (printable(bprm->buf[0]) && printable(bprm->buf[1]) && |
1669 | printable(bprm->buf[2]) && printable(bprm->buf[3])) | |
1670 | return retval; | |
4e0621a0 ON |
1671 | if (request_module("binfmt-%04x", *(ushort *)(bprm->buf + 2)) < 0) |
1672 | return retval; | |
cb7b6b1c ON |
1673 | need_retry = false; |
1674 | goto retry; | |
1675 | } | |
1676 | ||
1da177e4 LT |
1677 | return retval; |
1678 | } | |
1da177e4 LT |
1679 | EXPORT_SYMBOL(search_binary_handler); |
1680 | ||
5d1baf3b ON |
1681 | static int exec_binprm(struct linux_binprm *bprm) |
1682 | { | |
1683 | pid_t old_pid, old_vpid; | |
1684 | int ret; | |
1685 | ||
1686 | /* Need to fetch pid before load_binary changes it */ | |
1687 | old_pid = current->pid; | |
1688 | rcu_read_lock(); | |
1689 | old_vpid = task_pid_nr_ns(current, task_active_pid_ns(current->parent)); | |
1690 | rcu_read_unlock(); | |
1691 | ||
1692 | ret = search_binary_handler(bprm); | |
1693 | if (ret >= 0) { | |
3eaded86 | 1694 | audit_bprm(bprm); |
5d1baf3b ON |
1695 | trace_sched_process_exec(current, old_pid, bprm); |
1696 | ptrace_event(PTRACE_EVENT_EXEC, old_vpid); | |
9beb266f | 1697 | proc_exec_connector(current); |
5d1baf3b ON |
1698 | } |
1699 | ||
1700 | return ret; | |
1701 | } | |
1702 | ||
1da177e4 LT |
1703 | /* |
1704 | * sys_execve() executes a new program. | |
1705 | */ | |
51f39a1f DD |
1706 | static int do_execveat_common(int fd, struct filename *filename, |
1707 | struct user_arg_ptr argv, | |
1708 | struct user_arg_ptr envp, | |
1709 | int flags) | |
1da177e4 | 1710 | { |
51f39a1f | 1711 | char *pathbuf = NULL; |
1da177e4 LT |
1712 | struct linux_binprm *bprm; |
1713 | struct file *file; | |
3b125388 | 1714 | struct files_struct *displaced; |
1da177e4 | 1715 | int retval; |
72fa5997 | 1716 | |
c4ad8f98 LT |
1717 | if (IS_ERR(filename)) |
1718 | return PTR_ERR(filename); | |
1719 | ||
72fa5997 VK |
1720 | /* |
1721 | * We move the actual failure in case of RLIMIT_NPROC excess from | |
1722 | * set*uid() to execve() because too many poorly written programs | |
1723 | * don't check setuid() return code. Here we additionally recheck | |
1724 | * whether NPROC limit is still exceeded. | |
1725 | */ | |
1726 | if ((current->flags & PF_NPROC_EXCEEDED) && | |
bd9d43f4 | 1727 | atomic_read(¤t_user()->processes) > rlimit(RLIMIT_NPROC)) { |
72fa5997 VK |
1728 | retval = -EAGAIN; |
1729 | goto out_ret; | |
1730 | } | |
1731 | ||
1732 | /* We're below the limit (still or again), so we don't want to make | |
1733 | * further execve() calls fail. */ | |
1734 | current->flags &= ~PF_NPROC_EXCEEDED; | |
1da177e4 | 1735 | |
3b125388 | 1736 | retval = unshare_files(&displaced); |
fd8328be AV |
1737 | if (retval) |
1738 | goto out_ret; | |
1739 | ||
1da177e4 | 1740 | retval = -ENOMEM; |
11b0b5ab | 1741 | bprm = kzalloc(sizeof(*bprm), GFP_KERNEL); |
1da177e4 | 1742 | if (!bprm) |
fd8328be | 1743 | goto out_files; |
1da177e4 | 1744 | |
a2a8474c ON |
1745 | retval = prepare_bprm_creds(bprm); |
1746 | if (retval) | |
a6f76f23 | 1747 | goto out_free; |
498052bb | 1748 | |
9e00cdb0 | 1749 | check_unsafe_exec(bprm); |
a2a8474c | 1750 | current->in_execve = 1; |
a6f76f23 | 1751 | |
51f39a1f | 1752 | file = do_open_execat(fd, filename, flags); |
1da177e4 LT |
1753 | retval = PTR_ERR(file); |
1754 | if (IS_ERR(file)) | |
498052bb | 1755 | goto out_unmark; |
1da177e4 LT |
1756 | |
1757 | sched_exec(); | |
1758 | ||
1da177e4 | 1759 | bprm->file = file; |
51f39a1f DD |
1760 | if (fd == AT_FDCWD || filename->name[0] == '/') { |
1761 | bprm->filename = filename->name; | |
1762 | } else { | |
1763 | if (filename->name[0] == '\0') | |
0ee931c4 | 1764 | pathbuf = kasprintf(GFP_KERNEL, "/dev/fd/%d", fd); |
51f39a1f | 1765 | else |
0ee931c4 | 1766 | pathbuf = kasprintf(GFP_KERNEL, "/dev/fd/%d/%s", |
51f39a1f DD |
1767 | fd, filename->name); |
1768 | if (!pathbuf) { | |
1769 | retval = -ENOMEM; | |
1770 | goto out_unmark; | |
1771 | } | |
1772 | /* | |
1773 | * Record that a name derived from an O_CLOEXEC fd will be | |
1774 | * inaccessible after exec. Relies on having exclusive access to | |
1775 | * current->files (due to unshare_files above). | |
1776 | */ | |
1777 | if (close_on_exec(fd, rcu_dereference_raw(current->files->fdt))) | |
1778 | bprm->interp_flags |= BINPRM_FLAGS_PATH_INACCESSIBLE; | |
1779 | bprm->filename = pathbuf; | |
1780 | } | |
1781 | bprm->interp = bprm->filename; | |
1da177e4 | 1782 | |
b6a2fea3 OW |
1783 | retval = bprm_mm_init(bprm); |
1784 | if (retval) | |
63e46b95 | 1785 | goto out_unmark; |
1da177e4 | 1786 | |
b6a2fea3 | 1787 | bprm->argc = count(argv, MAX_ARG_STRINGS); |
1da177e4 | 1788 | if ((retval = bprm->argc) < 0) |
a6f76f23 | 1789 | goto out; |
1da177e4 | 1790 | |
b6a2fea3 | 1791 | bprm->envc = count(envp, MAX_ARG_STRINGS); |
1da177e4 | 1792 | if ((retval = bprm->envc) < 0) |
1da177e4 LT |
1793 | goto out; |
1794 | ||
1795 | retval = prepare_binprm(bprm); | |
1796 | if (retval < 0) | |
1797 | goto out; | |
1798 | ||
1799 | retval = copy_strings_kernel(1, &bprm->filename, bprm); | |
1800 | if (retval < 0) | |
1801 | goto out; | |
1802 | ||
1803 | bprm->exec = bprm->p; | |
1804 | retval = copy_strings(bprm->envc, envp, bprm); | |
1805 | if (retval < 0) | |
1806 | goto out; | |
1807 | ||
1808 | retval = copy_strings(bprm->argc, argv, bprm); | |
1809 | if (retval < 0) | |
1810 | goto out; | |
1811 | ||
f84df2a6 EB |
1812 | would_dump(bprm, bprm->file); |
1813 | ||
5d1baf3b | 1814 | retval = exec_binprm(bprm); |
a6f76f23 DH |
1815 | if (retval < 0) |
1816 | goto out; | |
1da177e4 | 1817 | |
a6f76f23 | 1818 | /* execve succeeded */ |
498052bb | 1819 | current->fs->in_exec = 0; |
f9ce1f1c | 1820 | current->in_execve = 0; |
a961e409 | 1821 | membarrier_execve(current); |
a6f76f23 | 1822 | acct_update_integrals(current); |
82727018 | 1823 | task_numa_free(current); |
a6f76f23 | 1824 | free_bprm(bprm); |
51f39a1f | 1825 | kfree(pathbuf); |
c4ad8f98 | 1826 | putname(filename); |
a6f76f23 DH |
1827 | if (displaced) |
1828 | put_files_struct(displaced); | |
1829 | return retval; | |
1da177e4 | 1830 | |
a6f76f23 | 1831 | out: |
3c77f845 ON |
1832 | if (bprm->mm) { |
1833 | acct_arg_size(bprm, 0); | |
1834 | mmput(bprm->mm); | |
1835 | } | |
1da177e4 | 1836 | |
498052bb | 1837 | out_unmark: |
9e00cdb0 | 1838 | current->fs->in_exec = 0; |
f9ce1f1c | 1839 | current->in_execve = 0; |
a6f76f23 DH |
1840 | |
1841 | out_free: | |
08a6fac1 | 1842 | free_bprm(bprm); |
51f39a1f | 1843 | kfree(pathbuf); |
1da177e4 | 1844 | |
fd8328be | 1845 | out_files: |
3b125388 AV |
1846 | if (displaced) |
1847 | reset_files_struct(displaced); | |
1da177e4 | 1848 | out_ret: |
c4ad8f98 | 1849 | putname(filename); |
1da177e4 LT |
1850 | return retval; |
1851 | } | |
1852 | ||
c4ad8f98 | 1853 | int do_execve(struct filename *filename, |
ba2d0162 | 1854 | const char __user *const __user *__argv, |
da3d4c5f | 1855 | const char __user *const __user *__envp) |
ba2d0162 | 1856 | { |
0e028465 ON |
1857 | struct user_arg_ptr argv = { .ptr.native = __argv }; |
1858 | struct user_arg_ptr envp = { .ptr.native = __envp }; | |
51f39a1f DD |
1859 | return do_execveat_common(AT_FDCWD, filename, argv, envp, 0); |
1860 | } | |
1861 | ||
1862 | int do_execveat(int fd, struct filename *filename, | |
1863 | const char __user *const __user *__argv, | |
1864 | const char __user *const __user *__envp, | |
1865 | int flags) | |
1866 | { | |
1867 | struct user_arg_ptr argv = { .ptr.native = __argv }; | |
1868 | struct user_arg_ptr envp = { .ptr.native = __envp }; | |
1869 | ||
1870 | return do_execveat_common(fd, filename, argv, envp, flags); | |
0e028465 ON |
1871 | } |
1872 | ||
1873 | #ifdef CONFIG_COMPAT | |
c4ad8f98 | 1874 | static int compat_do_execve(struct filename *filename, |
38b983b3 | 1875 | const compat_uptr_t __user *__argv, |
d03d26e5 | 1876 | const compat_uptr_t __user *__envp) |
0e028465 ON |
1877 | { |
1878 | struct user_arg_ptr argv = { | |
1879 | .is_compat = true, | |
1880 | .ptr.compat = __argv, | |
1881 | }; | |
1882 | struct user_arg_ptr envp = { | |
1883 | .is_compat = true, | |
1884 | .ptr.compat = __envp, | |
1885 | }; | |
51f39a1f DD |
1886 | return do_execveat_common(AT_FDCWD, filename, argv, envp, 0); |
1887 | } | |
1888 | ||
1889 | static int compat_do_execveat(int fd, struct filename *filename, | |
1890 | const compat_uptr_t __user *__argv, | |
1891 | const compat_uptr_t __user *__envp, | |
1892 | int flags) | |
1893 | { | |
1894 | struct user_arg_ptr argv = { | |
1895 | .is_compat = true, | |
1896 | .ptr.compat = __argv, | |
1897 | }; | |
1898 | struct user_arg_ptr envp = { | |
1899 | .is_compat = true, | |
1900 | .ptr.compat = __envp, | |
1901 | }; | |
1902 | return do_execveat_common(fd, filename, argv, envp, flags); | |
ba2d0162 | 1903 | } |
0e028465 | 1904 | #endif |
ba2d0162 | 1905 | |
964ee7df | 1906 | void set_binfmt(struct linux_binfmt *new) |
1da177e4 | 1907 | { |
801460d0 HS |
1908 | struct mm_struct *mm = current->mm; |
1909 | ||
1910 | if (mm->binfmt) | |
1911 | module_put(mm->binfmt->module); | |
1da177e4 | 1912 | |
801460d0 | 1913 | mm->binfmt = new; |
964ee7df ON |
1914 | if (new) |
1915 | __module_get(new->module); | |
1da177e4 | 1916 | } |
1da177e4 LT |
1917 | EXPORT_SYMBOL(set_binfmt); |
1918 | ||
6c5d5238 | 1919 | /* |
7288e118 | 1920 | * set_dumpable stores three-value SUID_DUMP_* into mm->flags. |
6c5d5238 KH |
1921 | */ |
1922 | void set_dumpable(struct mm_struct *mm, int value) | |
1923 | { | |
abacd2fe ON |
1924 | unsigned long old, new; |
1925 | ||
7288e118 ON |
1926 | if (WARN_ON((unsigned)value > SUID_DUMP_ROOT)) |
1927 | return; | |
1928 | ||
abacd2fe | 1929 | do { |
6aa7de05 | 1930 | old = READ_ONCE(mm->flags); |
7288e118 | 1931 | new = (old & ~MMF_DUMPABLE_MASK) | value; |
abacd2fe | 1932 | } while (cmpxchg(&mm->flags, old, new) != old); |
6c5d5238 | 1933 | } |
6c5d5238 | 1934 | |
38b983b3 AV |
1935 | SYSCALL_DEFINE3(execve, |
1936 | const char __user *, filename, | |
1937 | const char __user *const __user *, argv, | |
1938 | const char __user *const __user *, envp) | |
1939 | { | |
c4ad8f98 | 1940 | return do_execve(getname(filename), argv, envp); |
38b983b3 | 1941 | } |
51f39a1f DD |
1942 | |
1943 | SYSCALL_DEFINE5(execveat, | |
1944 | int, fd, const char __user *, filename, | |
1945 | const char __user *const __user *, argv, | |
1946 | const char __user *const __user *, envp, | |
1947 | int, flags) | |
1948 | { | |
1949 | int lookup_flags = (flags & AT_EMPTY_PATH) ? LOOKUP_EMPTY : 0; | |
1950 | ||
1951 | return do_execveat(fd, | |
1952 | getname_flags(filename, lookup_flags, NULL), | |
1953 | argv, envp, flags); | |
1954 | } | |
1955 | ||
38b983b3 | 1956 | #ifdef CONFIG_COMPAT |
625b1d7e HC |
1957 | COMPAT_SYSCALL_DEFINE3(execve, const char __user *, filename, |
1958 | const compat_uptr_t __user *, argv, | |
1959 | const compat_uptr_t __user *, envp) | |
38b983b3 | 1960 | { |
c4ad8f98 | 1961 | return compat_do_execve(getname(filename), argv, envp); |
38b983b3 | 1962 | } |
51f39a1f DD |
1963 | |
1964 | COMPAT_SYSCALL_DEFINE5(execveat, int, fd, | |
1965 | const char __user *, filename, | |
1966 | const compat_uptr_t __user *, argv, | |
1967 | const compat_uptr_t __user *, envp, | |
1968 | int, flags) | |
1969 | { | |
1970 | int lookup_flags = (flags & AT_EMPTY_PATH) ? LOOKUP_EMPTY : 0; | |
1971 | ||
1972 | return compat_do_execveat(fd, | |
1973 | getname_flags(filename, lookup_flags, NULL), | |
1974 | argv, envp, flags); | |
1975 | } | |
38b983b3 | 1976 | #endif |