]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/fs/exec.c | |
3 | * | |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | |
5 | */ | |
6 | ||
7 | /* | |
8 | * #!-checking implemented by tytso. | |
9 | */ | |
10 | /* | |
11 | * Demand-loading implemented 01.12.91 - no need to read anything but | |
12 | * the header into memory. The inode of the executable is put into | |
13 | * "current->executable", and page faults do the actual loading. Clean. | |
14 | * | |
15 | * Once more I can proudly say that linux stood up to being changed: it | |
16 | * was less than 2 hours work to get demand-loading completely implemented. | |
17 | * | |
18 | * Demand loading changed July 1993 by Eric Youngdale. Use mmap instead, | |
19 | * current->executable is only used by the procfs. This allows a dispatch | |
20 | * table to check for several different types of binary formats. We keep | |
21 | * trying until we recognize the file or we run out of supported binary | |
22 | * formats. | |
23 | */ | |
24 | ||
1da177e4 LT |
25 | #include <linux/slab.h> |
26 | #include <linux/file.h> | |
9f3acc31 | 27 | #include <linux/fdtable.h> |
ba92a43d | 28 | #include <linux/mm.h> |
615d6e87 | 29 | #include <linux/vmacache.h> |
1da177e4 LT |
30 | #include <linux/stat.h> |
31 | #include <linux/fcntl.h> | |
ba92a43d | 32 | #include <linux/swap.h> |
74aadce9 | 33 | #include <linux/string.h> |
1da177e4 | 34 | #include <linux/init.h> |
ca5b172b | 35 | #include <linux/pagemap.h> |
cdd6c482 | 36 | #include <linux/perf_event.h> |
1da177e4 LT |
37 | #include <linux/highmem.h> |
38 | #include <linux/spinlock.h> | |
39 | #include <linux/key.h> | |
40 | #include <linux/personality.h> | |
41 | #include <linux/binfmts.h> | |
1da177e4 | 42 | #include <linux/utsname.h> |
84d73786 | 43 | #include <linux/pid_namespace.h> |
1da177e4 LT |
44 | #include <linux/module.h> |
45 | #include <linux/namei.h> | |
1da177e4 LT |
46 | #include <linux/mount.h> |
47 | #include <linux/security.h> | |
48 | #include <linux/syscalls.h> | |
8f0ab514 | 49 | #include <linux/tsacct_kern.h> |
9f46080c | 50 | #include <linux/cn_proc.h> |
473ae30b | 51 | #include <linux/audit.h> |
6341c393 | 52 | #include <linux/tracehook.h> |
5f4123be | 53 | #include <linux/kmod.h> |
6110e3ab | 54 | #include <linux/fsnotify.h> |
5ad4e53b | 55 | #include <linux/fs_struct.h> |
61be228a | 56 | #include <linux/pipe_fs_i.h> |
3d5992d2 | 57 | #include <linux/oom.h> |
0e028465 | 58 | #include <linux/compat.h> |
1da177e4 | 59 | |
ccae47c6 SJR |
60 | #include <trace/events/fs.h> |
61 | ||
1da177e4 LT |
62 | #include <asm/uaccess.h> |
63 | #include <asm/mmu_context.h> | |
b6a2fea3 | 64 | #include <asm/tlb.h> |
43d2b113 KH |
65 | |
66 | #include <trace/events/task.h> | |
a6f76f23 | 67 | #include "internal.h" |
1da177e4 | 68 | |
4ff16c25 DS |
69 | #include <trace/events/sched.h> |
70 | ||
d6e71144 AC |
71 | int suid_dumpable = 0; |
72 | ||
e4dc1b14 | 73 | static LIST_HEAD(formats); |
1da177e4 LT |
74 | static DEFINE_RWLOCK(binfmt_lock); |
75 | ||
8fc3dc5a | 76 | void __register_binfmt(struct linux_binfmt * fmt, int insert) |
1da177e4 | 77 | { |
8fc3dc5a | 78 | BUG_ON(!fmt); |
92eaa565 ON |
79 | if (WARN_ON(!fmt->load_binary)) |
80 | return; | |
1da177e4 | 81 | write_lock(&binfmt_lock); |
74641f58 IK |
82 | insert ? list_add(&fmt->lh, &formats) : |
83 | list_add_tail(&fmt->lh, &formats); | |
1da177e4 | 84 | write_unlock(&binfmt_lock); |
1da177e4 LT |
85 | } |
86 | ||
74641f58 | 87 | EXPORT_SYMBOL(__register_binfmt); |
1da177e4 | 88 | |
f6b450d4 | 89 | void unregister_binfmt(struct linux_binfmt * fmt) |
1da177e4 | 90 | { |
1da177e4 | 91 | write_lock(&binfmt_lock); |
e4dc1b14 | 92 | list_del(&fmt->lh); |
1da177e4 | 93 | write_unlock(&binfmt_lock); |
1da177e4 LT |
94 | } |
95 | ||
96 | EXPORT_SYMBOL(unregister_binfmt); | |
97 | ||
98 | static inline void put_binfmt(struct linux_binfmt * fmt) | |
99 | { | |
100 | module_put(fmt->module); | |
101 | } | |
102 | ||
90f8572b EB |
103 | bool path_noexec(const struct path *path) |
104 | { | |
105 | return (path->mnt->mnt_flags & MNT_NOEXEC) || | |
106 | (path->mnt->mnt_sb->s_iflags & SB_I_NOEXEC); | |
107 | } | |
dee5220c | 108 | EXPORT_SYMBOL(path_noexec); |
90f8572b | 109 | |
090bd5f0 SF |
110 | bool path_nosuid(const struct path *path) |
111 | { | |
112 | return !mnt_may_suid(path->mnt) || | |
113 | (path->mnt->mnt_sb->s_iflags & SB_I_NOSUID); | |
114 | } | |
115 | EXPORT_SYMBOL(path_nosuid); | |
116 | ||
69369a70 | 117 | #ifdef CONFIG_USELIB |
1da177e4 LT |
118 | /* |
119 | * Note that a shared library must be both readable and executable due to | |
120 | * security reasons. | |
121 | * | |
122 | * Also note that we take the address to load from from the file itself. | |
123 | */ | |
1e7bfb21 | 124 | SYSCALL_DEFINE1(uselib, const char __user *, library) |
1da177e4 | 125 | { |
72c2d531 | 126 | struct linux_binfmt *fmt; |
964bd183 | 127 | struct file *file; |
91a27b2a | 128 | struct filename *tmp = getname(library); |
964bd183 | 129 | int error = PTR_ERR(tmp); |
47c805dc AV |
130 | static const struct open_flags uselib_flags = { |
131 | .open_flag = O_LARGEFILE | O_RDONLY | __FMODE_EXEC, | |
132 | .acc_mode = MAY_READ | MAY_EXEC | MAY_OPEN, | |
f9652e10 AV |
133 | .intent = LOOKUP_OPEN, |
134 | .lookup_flags = LOOKUP_FOLLOW, | |
47c805dc | 135 | }; |
964bd183 | 136 | |
6e8341a1 AV |
137 | if (IS_ERR(tmp)) |
138 | goto out; | |
139 | ||
f9652e10 | 140 | file = do_filp_open(AT_FDCWD, tmp, &uselib_flags); |
6e8341a1 AV |
141 | putname(tmp); |
142 | error = PTR_ERR(file); | |
143 | if (IS_ERR(file)) | |
1da177e4 LT |
144 | goto out; |
145 | ||
146 | error = -EINVAL; | |
496ad9aa | 147 | if (!S_ISREG(file_inode(file)->i_mode)) |
1da177e4 LT |
148 | goto exit; |
149 | ||
30524472 | 150 | error = -EACCES; |
90f8572b | 151 | if (path_noexec(&file->f_path)) |
1da177e4 LT |
152 | goto exit; |
153 | ||
2a12a9d7 | 154 | fsnotify_open(file); |
6110e3ab | 155 | |
1da177e4 | 156 | error = -ENOEXEC; |
1da177e4 | 157 | |
72c2d531 AV |
158 | read_lock(&binfmt_lock); |
159 | list_for_each_entry(fmt, &formats, lh) { | |
160 | if (!fmt->load_shlib) | |
161 | continue; | |
162 | if (!try_module_get(fmt->module)) | |
163 | continue; | |
1da177e4 | 164 | read_unlock(&binfmt_lock); |
72c2d531 AV |
165 | error = fmt->load_shlib(file); |
166 | read_lock(&binfmt_lock); | |
167 | put_binfmt(fmt); | |
168 | if (error != -ENOEXEC) | |
169 | break; | |
1da177e4 | 170 | } |
72c2d531 | 171 | read_unlock(&binfmt_lock); |
6e8341a1 | 172 | exit: |
1da177e4 LT |
173 | fput(file); |
174 | out: | |
175 | return error; | |
1da177e4 | 176 | } |
69369a70 | 177 | #endif /* #ifdef CONFIG_USELIB */ |
1da177e4 | 178 | |
b6a2fea3 | 179 | #ifdef CONFIG_MMU |
ae6b585e ON |
180 | /* |
181 | * The nascent bprm->mm is not visible until exec_mmap() but it can | |
182 | * use a lot of memory, account these pages in current->mm temporary | |
183 | * for oom_badness()->get_mm_rss(). Once exec succeeds or fails, we | |
184 | * change the counter back via acct_arg_size(0). | |
185 | */ | |
0e028465 | 186 | static void acct_arg_size(struct linux_binprm *bprm, unsigned long pages) |
3c77f845 ON |
187 | { |
188 | struct mm_struct *mm = current->mm; | |
189 | long diff = (long)(pages - bprm->vma_pages); | |
190 | ||
191 | if (!mm || !diff) | |
192 | return; | |
193 | ||
194 | bprm->vma_pages = pages; | |
3c77f845 | 195 | add_mm_counter(mm, MM_ANONPAGES, diff); |
3c77f845 ON |
196 | } |
197 | ||
0e028465 | 198 | static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos, |
b6a2fea3 OW |
199 | int write) |
200 | { | |
201 | struct page *page; | |
202 | int ret; | |
203 | ||
204 | #ifdef CONFIG_STACK_GROWSUP | |
205 | if (write) { | |
d05f3169 | 206 | ret = expand_downwards(bprm->vma, pos); |
b6a2fea3 OW |
207 | if (ret < 0) |
208 | return NULL; | |
209 | } | |
210 | #endif | |
211 | ret = get_user_pages(current, bprm->mm, pos, | |
212 | 1, write, 1, &page, NULL); | |
213 | if (ret <= 0) | |
214 | return NULL; | |
215 | ||
216 | if (write) { | |
b6a2fea3 | 217 | unsigned long size = bprm->vma->vm_end - bprm->vma->vm_start; |
a64e715f LT |
218 | struct rlimit *rlim; |
219 | ||
3c77f845 ON |
220 | acct_arg_size(bprm, size / PAGE_SIZE); |
221 | ||
a64e715f LT |
222 | /* |
223 | * We've historically supported up to 32 pages (ARG_MAX) | |
224 | * of argument strings even with small stacks | |
225 | */ | |
226 | if (size <= ARG_MAX) | |
227 | return page; | |
b6a2fea3 OW |
228 | |
229 | /* | |
230 | * Limit to 1/4-th the stack size for the argv+env strings. | |
231 | * This ensures that: | |
232 | * - the remaining binfmt code will not run out of stack space, | |
233 | * - the program will have a reasonable amount of stack left | |
234 | * to work from. | |
235 | */ | |
a64e715f | 236 | rlim = current->signal->rlim; |
d554ed89 | 237 | if (size > ACCESS_ONCE(rlim[RLIMIT_STACK].rlim_cur) / 4) { |
b6a2fea3 OW |
238 | put_page(page); |
239 | return NULL; | |
240 | } | |
241 | } | |
242 | ||
243 | return page; | |
244 | } | |
245 | ||
246 | static void put_arg_page(struct page *page) | |
247 | { | |
248 | put_page(page); | |
249 | } | |
250 | ||
251 | static void free_arg_page(struct linux_binprm *bprm, int i) | |
252 | { | |
253 | } | |
254 | ||
255 | static void free_arg_pages(struct linux_binprm *bprm) | |
256 | { | |
257 | } | |
258 | ||
259 | static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos, | |
260 | struct page *page) | |
261 | { | |
262 | flush_cache_page(bprm->vma, pos, page_to_pfn(page)); | |
263 | } | |
264 | ||
265 | static int __bprm_mm_init(struct linux_binprm *bprm) | |
266 | { | |
eaccbfa5 | 267 | int err; |
b6a2fea3 OW |
268 | struct vm_area_struct *vma = NULL; |
269 | struct mm_struct *mm = bprm->mm; | |
270 | ||
271 | bprm->vma = vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL); | |
272 | if (!vma) | |
eaccbfa5 | 273 | return -ENOMEM; |
b6a2fea3 OW |
274 | |
275 | down_write(&mm->mmap_sem); | |
276 | vma->vm_mm = mm; | |
277 | ||
278 | /* | |
279 | * Place the stack at the largest stack address the architecture | |
280 | * supports. Later, we'll move this to an appropriate place. We don't | |
281 | * use STACK_TOP because that can depend on attributes which aren't | |
282 | * configured yet. | |
283 | */ | |
aacb3d17 | 284 | BUILD_BUG_ON(VM_STACK_FLAGS & VM_STACK_INCOMPLETE_SETUP); |
b6a2fea3 OW |
285 | vma->vm_end = STACK_TOP_MAX; |
286 | vma->vm_start = vma->vm_end - PAGE_SIZE; | |
d9104d1c | 287 | vma->vm_flags = VM_SOFTDIRTY | VM_STACK_FLAGS | VM_STACK_INCOMPLETE_SETUP; |
3ed75eb8 | 288 | vma->vm_page_prot = vm_get_page_prot(vma->vm_flags); |
5beb4930 | 289 | INIT_LIST_HEAD(&vma->anon_vma_chain); |
462e635e | 290 | |
b6a2fea3 | 291 | err = insert_vm_struct(mm, vma); |
eaccbfa5 | 292 | if (err) |
b6a2fea3 | 293 | goto err; |
b6a2fea3 OW |
294 | |
295 | mm->stack_vm = mm->total_vm = 1; | |
fe3d197f | 296 | arch_bprm_mm_init(mm, vma); |
b6a2fea3 | 297 | up_write(&mm->mmap_sem); |
b6a2fea3 | 298 | bprm->p = vma->vm_end - sizeof(void *); |
b6a2fea3 | 299 | return 0; |
b6a2fea3 | 300 | err: |
eaccbfa5 LFC |
301 | up_write(&mm->mmap_sem); |
302 | bprm->vma = NULL; | |
303 | kmem_cache_free(vm_area_cachep, vma); | |
b6a2fea3 OW |
304 | return err; |
305 | } | |
306 | ||
307 | static bool valid_arg_len(struct linux_binprm *bprm, long len) | |
308 | { | |
309 | return len <= MAX_ARG_STRLEN; | |
310 | } | |
311 | ||
312 | #else | |
313 | ||
0e028465 | 314 | static inline void acct_arg_size(struct linux_binprm *bprm, unsigned long pages) |
3c77f845 ON |
315 | { |
316 | } | |
317 | ||
0e028465 | 318 | static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos, |
b6a2fea3 OW |
319 | int write) |
320 | { | |
321 | struct page *page; | |
322 | ||
323 | page = bprm->page[pos / PAGE_SIZE]; | |
324 | if (!page && write) { | |
325 | page = alloc_page(GFP_HIGHUSER|__GFP_ZERO); | |
326 | if (!page) | |
327 | return NULL; | |
328 | bprm->page[pos / PAGE_SIZE] = page; | |
329 | } | |
330 | ||
331 | return page; | |
332 | } | |
333 | ||
334 | static void put_arg_page(struct page *page) | |
335 | { | |
336 | } | |
337 | ||
338 | static void free_arg_page(struct linux_binprm *bprm, int i) | |
339 | { | |
340 | if (bprm->page[i]) { | |
341 | __free_page(bprm->page[i]); | |
342 | bprm->page[i] = NULL; | |
343 | } | |
344 | } | |
345 | ||
346 | static void free_arg_pages(struct linux_binprm *bprm) | |
347 | { | |
348 | int i; | |
349 | ||
350 | for (i = 0; i < MAX_ARG_PAGES; i++) | |
351 | free_arg_page(bprm, i); | |
352 | } | |
353 | ||
354 | static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos, | |
355 | struct page *page) | |
356 | { | |
357 | } | |
358 | ||
359 | static int __bprm_mm_init(struct linux_binprm *bprm) | |
360 | { | |
361 | bprm->p = PAGE_SIZE * MAX_ARG_PAGES - sizeof(void *); | |
362 | return 0; | |
363 | } | |
364 | ||
365 | static bool valid_arg_len(struct linux_binprm *bprm, long len) | |
366 | { | |
367 | return len <= bprm->p; | |
368 | } | |
369 | ||
370 | #endif /* CONFIG_MMU */ | |
371 | ||
372 | /* | |
373 | * Create a new mm_struct and populate it with a temporary stack | |
374 | * vm_area_struct. We don't have enough context at this point to set the stack | |
375 | * flags, permissions, and offset, so we use temporary values. We'll update | |
376 | * them later in setup_arg_pages(). | |
377 | */ | |
9cc64cea | 378 | static int bprm_mm_init(struct linux_binprm *bprm) |
b6a2fea3 OW |
379 | { |
380 | int err; | |
381 | struct mm_struct *mm = NULL; | |
382 | ||
383 | bprm->mm = mm = mm_alloc(); | |
384 | err = -ENOMEM; | |
385 | if (!mm) | |
386 | goto err; | |
387 | ||
b6a2fea3 OW |
388 | err = __bprm_mm_init(bprm); |
389 | if (err) | |
390 | goto err; | |
391 | ||
392 | return 0; | |
393 | ||
394 | err: | |
395 | if (mm) { | |
396 | bprm->mm = NULL; | |
397 | mmdrop(mm); | |
398 | } | |
399 | ||
400 | return err; | |
401 | } | |
402 | ||
ba2d0162 | 403 | struct user_arg_ptr { |
0e028465 ON |
404 | #ifdef CONFIG_COMPAT |
405 | bool is_compat; | |
406 | #endif | |
407 | union { | |
408 | const char __user *const __user *native; | |
409 | #ifdef CONFIG_COMPAT | |
38b983b3 | 410 | const compat_uptr_t __user *compat; |
0e028465 ON |
411 | #endif |
412 | } ptr; | |
ba2d0162 ON |
413 | }; |
414 | ||
415 | static const char __user *get_user_arg_ptr(struct user_arg_ptr argv, int nr) | |
1d1dbf81 | 416 | { |
0e028465 ON |
417 | const char __user *native; |
418 | ||
419 | #ifdef CONFIG_COMPAT | |
420 | if (unlikely(argv.is_compat)) { | |
421 | compat_uptr_t compat; | |
422 | ||
423 | if (get_user(compat, argv.ptr.compat + nr)) | |
424 | return ERR_PTR(-EFAULT); | |
1d1dbf81 | 425 | |
0e028465 ON |
426 | return compat_ptr(compat); |
427 | } | |
428 | #endif | |
429 | ||
430 | if (get_user(native, argv.ptr.native + nr)) | |
1d1dbf81 ON |
431 | return ERR_PTR(-EFAULT); |
432 | ||
0e028465 | 433 | return native; |
1d1dbf81 ON |
434 | } |
435 | ||
1da177e4 LT |
436 | /* |
437 | * count() counts the number of strings in array ARGV. | |
438 | */ | |
ba2d0162 | 439 | static int count(struct user_arg_ptr argv, int max) |
1da177e4 LT |
440 | { |
441 | int i = 0; | |
442 | ||
0e028465 | 443 | if (argv.ptr.native != NULL) { |
1da177e4 | 444 | for (;;) { |
1d1dbf81 | 445 | const char __user *p = get_user_arg_ptr(argv, i); |
1da177e4 | 446 | |
1da177e4 LT |
447 | if (!p) |
448 | break; | |
1d1dbf81 ON |
449 | |
450 | if (IS_ERR(p)) | |
451 | return -EFAULT; | |
452 | ||
6d92d4f6 | 453 | if (i >= max) |
1da177e4 | 454 | return -E2BIG; |
6d92d4f6 | 455 | ++i; |
9aea5a65 RM |
456 | |
457 | if (fatal_signal_pending(current)) | |
458 | return -ERESTARTNOHAND; | |
1da177e4 LT |
459 | cond_resched(); |
460 | } | |
461 | } | |
462 | return i; | |
463 | } | |
464 | ||
465 | /* | |
b6a2fea3 OW |
466 | * 'copy_strings()' copies argument/environment strings from the old |
467 | * processes's memory to the new process's stack. The call to get_user_pages() | |
468 | * ensures the destination page is created and not swapped out. | |
1da177e4 | 469 | */ |
ba2d0162 | 470 | static int copy_strings(int argc, struct user_arg_ptr argv, |
75c96f85 | 471 | struct linux_binprm *bprm) |
1da177e4 LT |
472 | { |
473 | struct page *kmapped_page = NULL; | |
474 | char *kaddr = NULL; | |
b6a2fea3 | 475 | unsigned long kpos = 0; |
1da177e4 LT |
476 | int ret; |
477 | ||
478 | while (argc-- > 0) { | |
d7627467 | 479 | const char __user *str; |
1da177e4 LT |
480 | int len; |
481 | unsigned long pos; | |
482 | ||
1d1dbf81 ON |
483 | ret = -EFAULT; |
484 | str = get_user_arg_ptr(argv, argc); | |
485 | if (IS_ERR(str)) | |
1da177e4 | 486 | goto out; |
1da177e4 | 487 | |
1d1dbf81 ON |
488 | len = strnlen_user(str, MAX_ARG_STRLEN); |
489 | if (!len) | |
490 | goto out; | |
491 | ||
492 | ret = -E2BIG; | |
493 | if (!valid_arg_len(bprm, len)) | |
1da177e4 | 494 | goto out; |
1da177e4 | 495 | |
b6a2fea3 | 496 | /* We're going to work our way backwords. */ |
1da177e4 | 497 | pos = bprm->p; |
b6a2fea3 OW |
498 | str += len; |
499 | bprm->p -= len; | |
1da177e4 LT |
500 | |
501 | while (len > 0) { | |
1da177e4 | 502 | int offset, bytes_to_copy; |
1da177e4 | 503 | |
9aea5a65 RM |
504 | if (fatal_signal_pending(current)) { |
505 | ret = -ERESTARTNOHAND; | |
506 | goto out; | |
507 | } | |
7993bc1f RM |
508 | cond_resched(); |
509 | ||
1da177e4 | 510 | offset = pos % PAGE_SIZE; |
b6a2fea3 OW |
511 | if (offset == 0) |
512 | offset = PAGE_SIZE; | |
513 | ||
514 | bytes_to_copy = offset; | |
515 | if (bytes_to_copy > len) | |
516 | bytes_to_copy = len; | |
517 | ||
518 | offset -= bytes_to_copy; | |
519 | pos -= bytes_to_copy; | |
520 | str -= bytes_to_copy; | |
521 | len -= bytes_to_copy; | |
522 | ||
523 | if (!kmapped_page || kpos != (pos & PAGE_MASK)) { | |
524 | struct page *page; | |
525 | ||
526 | page = get_arg_page(bprm, pos, 1); | |
1da177e4 | 527 | if (!page) { |
b6a2fea3 | 528 | ret = -E2BIG; |
1da177e4 LT |
529 | goto out; |
530 | } | |
1da177e4 | 531 | |
b6a2fea3 OW |
532 | if (kmapped_page) { |
533 | flush_kernel_dcache_page(kmapped_page); | |
1da177e4 | 534 | kunmap(kmapped_page); |
b6a2fea3 OW |
535 | put_arg_page(kmapped_page); |
536 | } | |
1da177e4 LT |
537 | kmapped_page = page; |
538 | kaddr = kmap(kmapped_page); | |
b6a2fea3 OW |
539 | kpos = pos & PAGE_MASK; |
540 | flush_arg_page(bprm, kpos, kmapped_page); | |
1da177e4 | 541 | } |
b6a2fea3 | 542 | if (copy_from_user(kaddr+offset, str, bytes_to_copy)) { |
1da177e4 LT |
543 | ret = -EFAULT; |
544 | goto out; | |
545 | } | |
1da177e4 LT |
546 | } |
547 | } | |
548 | ret = 0; | |
549 | out: | |
b6a2fea3 OW |
550 | if (kmapped_page) { |
551 | flush_kernel_dcache_page(kmapped_page); | |
1da177e4 | 552 | kunmap(kmapped_page); |
b6a2fea3 OW |
553 | put_arg_page(kmapped_page); |
554 | } | |
1da177e4 LT |
555 | return ret; |
556 | } | |
557 | ||
558 | /* | |
559 | * Like copy_strings, but get argv and its values from kernel memory. | |
560 | */ | |
ba2d0162 | 561 | int copy_strings_kernel(int argc, const char *const *__argv, |
d7627467 | 562 | struct linux_binprm *bprm) |
1da177e4 LT |
563 | { |
564 | int r; | |
565 | mm_segment_t oldfs = get_fs(); | |
ba2d0162 | 566 | struct user_arg_ptr argv = { |
0e028465 | 567 | .ptr.native = (const char __user *const __user *)__argv, |
ba2d0162 ON |
568 | }; |
569 | ||
1da177e4 | 570 | set_fs(KERNEL_DS); |
ba2d0162 | 571 | r = copy_strings(argc, argv, bprm); |
1da177e4 | 572 | set_fs(oldfs); |
ba2d0162 | 573 | |
1da177e4 LT |
574 | return r; |
575 | } | |
1da177e4 LT |
576 | EXPORT_SYMBOL(copy_strings_kernel); |
577 | ||
578 | #ifdef CONFIG_MMU | |
b6a2fea3 | 579 | |
1da177e4 | 580 | /* |
b6a2fea3 OW |
581 | * During bprm_mm_init(), we create a temporary stack at STACK_TOP_MAX. Once |
582 | * the binfmt code determines where the new stack should reside, we shift it to | |
583 | * its final location. The process proceeds as follows: | |
1da177e4 | 584 | * |
b6a2fea3 OW |
585 | * 1) Use shift to calculate the new vma endpoints. |
586 | * 2) Extend vma to cover both the old and new ranges. This ensures the | |
587 | * arguments passed to subsequent functions are consistent. | |
588 | * 3) Move vma's page tables to the new range. | |
589 | * 4) Free up any cleared pgd range. | |
590 | * 5) Shrink the vma to cover only the new range. | |
1da177e4 | 591 | */ |
b6a2fea3 | 592 | static int shift_arg_pages(struct vm_area_struct *vma, unsigned long shift) |
1da177e4 LT |
593 | { |
594 | struct mm_struct *mm = vma->vm_mm; | |
b6a2fea3 OW |
595 | unsigned long old_start = vma->vm_start; |
596 | unsigned long old_end = vma->vm_end; | |
597 | unsigned long length = old_end - old_start; | |
598 | unsigned long new_start = old_start - shift; | |
599 | unsigned long new_end = old_end - shift; | |
d16dfc55 | 600 | struct mmu_gather tlb; |
1da177e4 | 601 | |
b6a2fea3 | 602 | BUG_ON(new_start > new_end); |
1da177e4 | 603 | |
b6a2fea3 OW |
604 | /* |
605 | * ensure there are no vmas between where we want to go | |
606 | * and where we are | |
607 | */ | |
608 | if (vma != find_vma(mm, new_start)) | |
609 | return -EFAULT; | |
610 | ||
611 | /* | |
612 | * cover the whole range: [new_start, old_end) | |
613 | */ | |
5beb4930 RR |
614 | if (vma_adjust(vma, new_start, old_end, vma->vm_pgoff, NULL)) |
615 | return -ENOMEM; | |
b6a2fea3 OW |
616 | |
617 | /* | |
618 | * move the page tables downwards, on failure we rely on | |
619 | * process cleanup to remove whatever mess we made. | |
620 | */ | |
621 | if (length != move_page_tables(vma, old_start, | |
38a76013 | 622 | vma, new_start, length, false)) |
b6a2fea3 OW |
623 | return -ENOMEM; |
624 | ||
625 | lru_add_drain(); | |
2b047252 | 626 | tlb_gather_mmu(&tlb, mm, old_start, old_end); |
b6a2fea3 OW |
627 | if (new_end > old_start) { |
628 | /* | |
629 | * when the old and new regions overlap clear from new_end. | |
630 | */ | |
d16dfc55 | 631 | free_pgd_range(&tlb, new_end, old_end, new_end, |
6ee8630e | 632 | vma->vm_next ? vma->vm_next->vm_start : USER_PGTABLES_CEILING); |
b6a2fea3 OW |
633 | } else { |
634 | /* | |
635 | * otherwise, clean from old_start; this is done to not touch | |
636 | * the address space in [new_end, old_start) some architectures | |
637 | * have constraints on va-space that make this illegal (IA64) - | |
638 | * for the others its just a little faster. | |
639 | */ | |
d16dfc55 | 640 | free_pgd_range(&tlb, old_start, old_end, new_end, |
6ee8630e | 641 | vma->vm_next ? vma->vm_next->vm_start : USER_PGTABLES_CEILING); |
1da177e4 | 642 | } |
2b047252 | 643 | tlb_finish_mmu(&tlb, old_start, old_end); |
b6a2fea3 OW |
644 | |
645 | /* | |
5beb4930 | 646 | * Shrink the vma to just the new range. Always succeeds. |
b6a2fea3 OW |
647 | */ |
648 | vma_adjust(vma, new_start, new_end, vma->vm_pgoff, NULL); | |
649 | ||
650 | return 0; | |
1da177e4 LT |
651 | } |
652 | ||
b6a2fea3 OW |
653 | /* |
654 | * Finalizes the stack vm_area_struct. The flags and permissions are updated, | |
655 | * the stack is optionally relocated, and some extra space is added. | |
656 | */ | |
1da177e4 LT |
657 | int setup_arg_pages(struct linux_binprm *bprm, |
658 | unsigned long stack_top, | |
659 | int executable_stack) | |
660 | { | |
b6a2fea3 OW |
661 | unsigned long ret; |
662 | unsigned long stack_shift; | |
1da177e4 | 663 | struct mm_struct *mm = current->mm; |
b6a2fea3 OW |
664 | struct vm_area_struct *vma = bprm->vma; |
665 | struct vm_area_struct *prev = NULL; | |
666 | unsigned long vm_flags; | |
667 | unsigned long stack_base; | |
803bf5ec MN |
668 | unsigned long stack_size; |
669 | unsigned long stack_expand; | |
670 | unsigned long rlim_stack; | |
1da177e4 LT |
671 | |
672 | #ifdef CONFIG_STACK_GROWSUP | |
d71f290b | 673 | /* Limit stack size */ |
d554ed89 | 674 | stack_base = rlimit_max(RLIMIT_STACK); |
d71f290b JH |
675 | if (stack_base > STACK_SIZE_MAX) |
676 | stack_base = STACK_SIZE_MAX; | |
1da177e4 | 677 | |
d045c77c HD |
678 | /* Add space for stack randomization. */ |
679 | stack_base += (STACK_RND_MASK << PAGE_SHIFT); | |
680 | ||
b6a2fea3 OW |
681 | /* Make sure we didn't let the argument array grow too large. */ |
682 | if (vma->vm_end - vma->vm_start > stack_base) | |
683 | return -ENOMEM; | |
1da177e4 | 684 | |
b6a2fea3 | 685 | stack_base = PAGE_ALIGN(stack_top - stack_base); |
1da177e4 | 686 | |
b6a2fea3 OW |
687 | stack_shift = vma->vm_start - stack_base; |
688 | mm->arg_start = bprm->p - stack_shift; | |
689 | bprm->p = vma->vm_end - stack_shift; | |
1da177e4 | 690 | #else |
b6a2fea3 OW |
691 | stack_top = arch_align_stack(stack_top); |
692 | stack_top = PAGE_ALIGN(stack_top); | |
1b528181 RM |
693 | |
694 | if (unlikely(stack_top < mmap_min_addr) || | |
695 | unlikely(vma->vm_end - vma->vm_start >= stack_top - mmap_min_addr)) | |
696 | return -ENOMEM; | |
697 | ||
b6a2fea3 OW |
698 | stack_shift = vma->vm_end - stack_top; |
699 | ||
700 | bprm->p -= stack_shift; | |
1da177e4 | 701 | mm->arg_start = bprm->p; |
1da177e4 LT |
702 | #endif |
703 | ||
1da177e4 | 704 | if (bprm->loader) |
b6a2fea3 OW |
705 | bprm->loader -= stack_shift; |
706 | bprm->exec -= stack_shift; | |
1da177e4 | 707 | |
1da177e4 | 708 | down_write(&mm->mmap_sem); |
96a8e13e | 709 | vm_flags = VM_STACK_FLAGS; |
b6a2fea3 OW |
710 | |
711 | /* | |
712 | * Adjust stack execute permissions; explicitly enable for | |
713 | * EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X and leave alone | |
714 | * (arch default) otherwise. | |
715 | */ | |
716 | if (unlikely(executable_stack == EXSTACK_ENABLE_X)) | |
717 | vm_flags |= VM_EXEC; | |
718 | else if (executable_stack == EXSTACK_DISABLE_X) | |
719 | vm_flags &= ~VM_EXEC; | |
720 | vm_flags |= mm->def_flags; | |
a8bef8ff | 721 | vm_flags |= VM_STACK_INCOMPLETE_SETUP; |
b6a2fea3 OW |
722 | |
723 | ret = mprotect_fixup(vma, &prev, vma->vm_start, vma->vm_end, | |
724 | vm_flags); | |
725 | if (ret) | |
726 | goto out_unlock; | |
727 | BUG_ON(prev != vma); | |
728 | ||
729 | /* Move stack pages down in memory. */ | |
730 | if (stack_shift) { | |
731 | ret = shift_arg_pages(vma, stack_shift); | |
fc63cf23 AB |
732 | if (ret) |
733 | goto out_unlock; | |
1da177e4 LT |
734 | } |
735 | ||
a8bef8ff MG |
736 | /* mprotect_fixup is overkill to remove the temporary stack flags */ |
737 | vma->vm_flags &= ~VM_STACK_INCOMPLETE_SETUP; | |
738 | ||
5ef097dd | 739 | stack_expand = 131072UL; /* randomly 32*4k (or 2*64k) pages */ |
803bf5ec MN |
740 | stack_size = vma->vm_end - vma->vm_start; |
741 | /* | |
742 | * Align this down to a page boundary as expand_stack | |
743 | * will align it up. | |
744 | */ | |
745 | rlim_stack = rlimit(RLIMIT_STACK) & PAGE_MASK; | |
b6a2fea3 | 746 | #ifdef CONFIG_STACK_GROWSUP |
803bf5ec MN |
747 | if (stack_size + stack_expand > rlim_stack) |
748 | stack_base = vma->vm_start + rlim_stack; | |
749 | else | |
750 | stack_base = vma->vm_end + stack_expand; | |
b6a2fea3 | 751 | #else |
803bf5ec MN |
752 | if (stack_size + stack_expand > rlim_stack) |
753 | stack_base = vma->vm_end - rlim_stack; | |
754 | else | |
755 | stack_base = vma->vm_start - stack_expand; | |
b6a2fea3 | 756 | #endif |
3af9e859 | 757 | current->mm->start_stack = bprm->p; |
b6a2fea3 OW |
758 | ret = expand_stack(vma, stack_base); |
759 | if (ret) | |
760 | ret = -EFAULT; | |
761 | ||
762 | out_unlock: | |
1da177e4 | 763 | up_write(&mm->mmap_sem); |
fc63cf23 | 764 | return ret; |
1da177e4 | 765 | } |
1da177e4 LT |
766 | EXPORT_SYMBOL(setup_arg_pages); |
767 | ||
1da177e4 LT |
768 | #endif /* CONFIG_MMU */ |
769 | ||
51f39a1f | 770 | static struct file *do_open_execat(int fd, struct filename *name, int flags) |
1da177e4 | 771 | { |
1da177e4 | 772 | struct file *file; |
e56b6a5d | 773 | int err; |
51f39a1f | 774 | struct open_flags open_exec_flags = { |
47c805dc AV |
775 | .open_flag = O_LARGEFILE | O_RDONLY | __FMODE_EXEC, |
776 | .acc_mode = MAY_EXEC | MAY_OPEN, | |
f9652e10 AV |
777 | .intent = LOOKUP_OPEN, |
778 | .lookup_flags = LOOKUP_FOLLOW, | |
47c805dc | 779 | }; |
1da177e4 | 780 | |
51f39a1f DD |
781 | if ((flags & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0) |
782 | return ERR_PTR(-EINVAL); | |
783 | if (flags & AT_SYMLINK_NOFOLLOW) | |
784 | open_exec_flags.lookup_flags &= ~LOOKUP_FOLLOW; | |
785 | if (flags & AT_EMPTY_PATH) | |
786 | open_exec_flags.lookup_flags |= LOOKUP_EMPTY; | |
787 | ||
788 | file = do_filp_open(fd, name, &open_exec_flags); | |
6e8341a1 | 789 | if (IS_ERR(file)) |
e56b6a5d CH |
790 | goto out; |
791 | ||
792 | err = -EACCES; | |
496ad9aa | 793 | if (!S_ISREG(file_inode(file)->i_mode)) |
6e8341a1 | 794 | goto exit; |
e56b6a5d | 795 | |
90f8572b | 796 | if (path_noexec(&file->f_path)) |
6e8341a1 | 797 | goto exit; |
e56b6a5d CH |
798 | |
799 | err = deny_write_access(file); | |
6e8341a1 AV |
800 | if (err) |
801 | goto exit; | |
1da177e4 | 802 | |
51f39a1f DD |
803 | if (name->name[0] != '\0') |
804 | fsnotify_open(file); | |
805 | ||
ccae47c6 SJR |
806 | trace_open_exec(name->name); |
807 | ||
6e8341a1 | 808 | out: |
e56b6a5d CH |
809 | return file; |
810 | ||
6e8341a1 AV |
811 | exit: |
812 | fput(file); | |
e56b6a5d CH |
813 | return ERR_PTR(err); |
814 | } | |
c4ad8f98 LT |
815 | |
816 | struct file *open_exec(const char *name) | |
817 | { | |
51689104 PM |
818 | struct filename *filename = getname_kernel(name); |
819 | struct file *f = ERR_CAST(filename); | |
820 | ||
821 | if (!IS_ERR(filename)) { | |
822 | f = do_open_execat(AT_FDCWD, filename, 0); | |
823 | putname(filename); | |
824 | } | |
825 | return f; | |
c4ad8f98 | 826 | } |
1da177e4 LT |
827 | EXPORT_SYMBOL(open_exec); |
828 | ||
6777d773 MZ |
829 | int kernel_read(struct file *file, loff_t offset, |
830 | char *addr, unsigned long count) | |
1da177e4 LT |
831 | { |
832 | mm_segment_t old_fs; | |
833 | loff_t pos = offset; | |
834 | int result; | |
835 | ||
836 | old_fs = get_fs(); | |
837 | set_fs(get_ds()); | |
838 | /* The cast to a user pointer is valid due to the set_fs() */ | |
839 | result = vfs_read(file, (void __user *)addr, count, &pos); | |
840 | set_fs(old_fs); | |
841 | return result; | |
842 | } | |
843 | ||
844 | EXPORT_SYMBOL(kernel_read); | |
845 | ||
3dc20cb2 AV |
846 | ssize_t read_code(struct file *file, unsigned long addr, loff_t pos, size_t len) |
847 | { | |
ec695579 | 848 | ssize_t res = vfs_read(file, (void __user *)addr, len, &pos); |
3dc20cb2 AV |
849 | if (res > 0) |
850 | flush_icache_range(addr, addr + len); | |
851 | return res; | |
852 | } | |
853 | EXPORT_SYMBOL(read_code); | |
854 | ||
1da177e4 LT |
855 | static int exec_mmap(struct mm_struct *mm) |
856 | { | |
857 | struct task_struct *tsk; | |
615d6e87 | 858 | struct mm_struct *old_mm, *active_mm; |
1da177e4 LT |
859 | |
860 | /* Notify parent that we're no longer interested in the old VM */ | |
861 | tsk = current; | |
862 | old_mm = current->mm; | |
863 | mm_release(tsk, old_mm); | |
864 | ||
865 | if (old_mm) { | |
4fe7efdb | 866 | sync_mm_rss(old_mm); |
1da177e4 LT |
867 | /* |
868 | * Make sure that if there is a core dump in progress | |
869 | * for the old mm, we get out and die instead of going | |
870 | * through with the exec. We must hold mmap_sem around | |
999d9fc1 | 871 | * checking core_state and changing tsk->mm. |
1da177e4 LT |
872 | */ |
873 | down_read(&old_mm->mmap_sem); | |
999d9fc1 | 874 | if (unlikely(old_mm->core_state)) { |
1da177e4 LT |
875 | up_read(&old_mm->mmap_sem); |
876 | return -EINTR; | |
877 | } | |
878 | } | |
879 | task_lock(tsk); | |
880 | active_mm = tsk->active_mm; | |
881 | tsk->mm = mm; | |
882 | tsk->active_mm = mm; | |
883 | activate_mm(active_mm, mm); | |
615d6e87 DB |
884 | tsk->mm->vmacache_seqnum = 0; |
885 | vmacache_flush(tsk); | |
1da177e4 | 886 | task_unlock(tsk); |
1da177e4 LT |
887 | if (old_mm) { |
888 | up_read(&old_mm->mmap_sem); | |
7dddb12c | 889 | BUG_ON(active_mm != old_mm); |
701085b2 | 890 | setmax_mm_hiwater_rss(&tsk->signal->maxrss, old_mm); |
31a78f23 | 891 | mm_update_next_owner(old_mm); |
1da177e4 LT |
892 | mmput(old_mm); |
893 | return 0; | |
894 | } | |
895 | mmdrop(active_mm); | |
896 | return 0; | |
897 | } | |
898 | ||
899 | /* | |
900 | * This function makes sure the current process has its own signal table, | |
901 | * so that flush_signal_handlers can later reset the handlers without | |
902 | * disturbing other processes. (Other processes might share the signal | |
903 | * table via the CLONE_SIGHAND option to clone().) | |
904 | */ | |
858119e1 | 905 | static int de_thread(struct task_struct *tsk) |
1da177e4 LT |
906 | { |
907 | struct signal_struct *sig = tsk->signal; | |
b2c903b8 | 908 | struct sighand_struct *oldsighand = tsk->sighand; |
1da177e4 | 909 | spinlock_t *lock = &oldsighand->siglock; |
1da177e4 | 910 | |
aafe6c2a | 911 | if (thread_group_empty(tsk)) |
1da177e4 LT |
912 | goto no_thread_group; |
913 | ||
914 | /* | |
915 | * Kill all other threads in the thread group. | |
1da177e4 | 916 | */ |
1da177e4 | 917 | spin_lock_irq(lock); |
ed5d2cac | 918 | if (signal_group_exit(sig)) { |
1da177e4 LT |
919 | /* |
920 | * Another group action in progress, just | |
921 | * return so that the signal is processed. | |
922 | */ | |
923 | spin_unlock_irq(lock); | |
1da177e4 LT |
924 | return -EAGAIN; |
925 | } | |
d344193a | 926 | |
ed5d2cac | 927 | sig->group_exit_task = tsk; |
d344193a ON |
928 | sig->notify_count = zap_other_threads(tsk); |
929 | if (!thread_group_leader(tsk)) | |
930 | sig->notify_count--; | |
1da177e4 | 931 | |
d344193a | 932 | while (sig->notify_count) { |
d5bbd43d | 933 | __set_current_state(TASK_KILLABLE); |
1da177e4 LT |
934 | spin_unlock_irq(lock); |
935 | schedule(); | |
d5bbd43d ON |
936 | if (unlikely(__fatal_signal_pending(tsk))) |
937 | goto killed; | |
1da177e4 LT |
938 | spin_lock_irq(lock); |
939 | } | |
1da177e4 LT |
940 | spin_unlock_irq(lock); |
941 | ||
942 | /* | |
943 | * At this point all other threads have exited, all we have to | |
944 | * do is to wait for the thread group leader to become inactive, | |
945 | * and to assume its PID: | |
946 | */ | |
aafe6c2a | 947 | if (!thread_group_leader(tsk)) { |
8187926b | 948 | struct task_struct *leader = tsk->group_leader; |
6db840fa | 949 | |
6db840fa | 950 | for (;;) { |
e56fb287 | 951 | threadgroup_change_begin(tsk); |
6db840fa | 952 | write_lock_irq(&tasklist_lock); |
dfcce791 KT |
953 | /* |
954 | * Do this under tasklist_lock to ensure that | |
955 | * exit_notify() can't miss ->group_exit_task | |
956 | */ | |
957 | sig->notify_count = -1; | |
6db840fa ON |
958 | if (likely(leader->exit_state)) |
959 | break; | |
d5bbd43d | 960 | __set_current_state(TASK_KILLABLE); |
6db840fa | 961 | write_unlock_irq(&tasklist_lock); |
e56fb287 | 962 | threadgroup_change_end(tsk); |
6db840fa | 963 | schedule(); |
d5bbd43d ON |
964 | if (unlikely(__fatal_signal_pending(tsk))) |
965 | goto killed; | |
6db840fa | 966 | } |
1da177e4 | 967 | |
f5e90281 RM |
968 | /* |
969 | * The only record we have of the real-time age of a | |
970 | * process, regardless of execs it's done, is start_time. | |
971 | * All the past CPU time is accumulated in signal_struct | |
972 | * from sister threads now dead. But in this non-leader | |
973 | * exec, nothing survives from the original leader thread, | |
974 | * whose birth marks the true age of this process now. | |
975 | * When we take on its identity by switching to its PID, we | |
976 | * also take its birthdate (always earlier than our own). | |
977 | */ | |
aafe6c2a | 978 | tsk->start_time = leader->start_time; |
266b7a02 | 979 | tsk->real_start_time = leader->real_start_time; |
f5e90281 | 980 | |
bac0abd6 PE |
981 | BUG_ON(!same_thread_group(leader, tsk)); |
982 | BUG_ON(has_group_leader_pid(tsk)); | |
1da177e4 LT |
983 | /* |
984 | * An exec() starts a new thread group with the | |
985 | * TGID of the previous thread group. Rehash the | |
986 | * two threads with a switched PID, and release | |
987 | * the former thread group leader: | |
988 | */ | |
d73d6529 EB |
989 | |
990 | /* Become a process group leader with the old leader's pid. | |
c18258c6 EB |
991 | * The old leader becomes a thread of the this thread group. |
992 | * Note: The old leader also uses this pid until release_task | |
d73d6529 EB |
993 | * is called. Odd but simple and correct. |
994 | */ | |
aafe6c2a | 995 | tsk->pid = leader->pid; |
3f418548 | 996 | change_pid(tsk, PIDTYPE_PID, task_pid(leader)); |
aafe6c2a EB |
997 | transfer_pid(leader, tsk, PIDTYPE_PGID); |
998 | transfer_pid(leader, tsk, PIDTYPE_SID); | |
9cd80bbb | 999 | |
aafe6c2a | 1000 | list_replace_rcu(&leader->tasks, &tsk->tasks); |
9cd80bbb | 1001 | list_replace_init(&leader->sibling, &tsk->sibling); |
1da177e4 | 1002 | |
aafe6c2a EB |
1003 | tsk->group_leader = tsk; |
1004 | leader->group_leader = tsk; | |
de12a787 | 1005 | |
aafe6c2a | 1006 | tsk->exit_signal = SIGCHLD; |
087806b1 | 1007 | leader->exit_signal = -1; |
962b564c ON |
1008 | |
1009 | BUG_ON(leader->exit_state != EXIT_ZOMBIE); | |
1010 | leader->exit_state = EXIT_DEAD; | |
eac1b5e5 ON |
1011 | |
1012 | /* | |
1013 | * We are going to release_task()->ptrace_unlink() silently, | |
1014 | * the tracer can sleep in do_wait(). EXIT_DEAD guarantees | |
1015 | * the tracer wont't block again waiting for this thread. | |
1016 | */ | |
1017 | if (unlikely(leader->ptrace)) | |
1018 | __wake_up_parent(leader, leader->parent); | |
1da177e4 | 1019 | write_unlock_irq(&tasklist_lock); |
e56fb287 | 1020 | threadgroup_change_end(tsk); |
8187926b ON |
1021 | |
1022 | release_task(leader); | |
ed5d2cac | 1023 | } |
1da177e4 | 1024 | |
6db840fa ON |
1025 | sig->group_exit_task = NULL; |
1026 | sig->notify_count = 0; | |
1da177e4 LT |
1027 | |
1028 | no_thread_group: | |
e6368253 ON |
1029 | /* we have changed execution domain */ |
1030 | tsk->exit_signal = SIGCHLD; | |
1031 | ||
1da177e4 | 1032 | exit_itimers(sig); |
cbaffba1 | 1033 | flush_itimer_signals(); |
329f7dba | 1034 | |
b2c903b8 ON |
1035 | if (atomic_read(&oldsighand->count) != 1) { |
1036 | struct sighand_struct *newsighand; | |
1da177e4 | 1037 | /* |
b2c903b8 ON |
1038 | * This ->sighand is shared with the CLONE_SIGHAND |
1039 | * but not CLONE_THREAD task, switch to the new one. | |
1da177e4 | 1040 | */ |
b2c903b8 ON |
1041 | newsighand = kmem_cache_alloc(sighand_cachep, GFP_KERNEL); |
1042 | if (!newsighand) | |
1043 | return -ENOMEM; | |
1044 | ||
1da177e4 LT |
1045 | atomic_set(&newsighand->count, 1); |
1046 | memcpy(newsighand->action, oldsighand->action, | |
1047 | sizeof(newsighand->action)); | |
1048 | ||
1049 | write_lock_irq(&tasklist_lock); | |
1050 | spin_lock(&oldsighand->siglock); | |
aafe6c2a | 1051 | rcu_assign_pointer(tsk->sighand, newsighand); |
1da177e4 LT |
1052 | spin_unlock(&oldsighand->siglock); |
1053 | write_unlock_irq(&tasklist_lock); | |
1054 | ||
fba2afaa | 1055 | __cleanup_sighand(oldsighand); |
1da177e4 LT |
1056 | } |
1057 | ||
aafe6c2a | 1058 | BUG_ON(!thread_group_leader(tsk)); |
1da177e4 | 1059 | return 0; |
d5bbd43d ON |
1060 | |
1061 | killed: | |
1062 | /* protects against exit_notify() and __exit_signal() */ | |
1063 | read_lock(&tasklist_lock); | |
1064 | sig->group_exit_task = NULL; | |
1065 | sig->notify_count = 0; | |
1066 | read_unlock(&tasklist_lock); | |
1067 | return -EAGAIN; | |
1da177e4 | 1068 | } |
0840a90d | 1069 | |
59714d65 | 1070 | char *get_task_comm(char *buf, struct task_struct *tsk) |
1da177e4 LT |
1071 | { |
1072 | /* buf must be at least sizeof(tsk->comm) in size */ | |
1073 | task_lock(tsk); | |
1074 | strncpy(buf, tsk->comm, sizeof(tsk->comm)); | |
1075 | task_unlock(tsk); | |
59714d65 | 1076 | return buf; |
1da177e4 | 1077 | } |
7d74f492 | 1078 | EXPORT_SYMBOL_GPL(get_task_comm); |
1da177e4 | 1079 | |
6a6d27de AV |
1080 | /* |
1081 | * These functions flushes out all traces of the currently running executable | |
1082 | * so that a new one can be started | |
1083 | */ | |
1084 | ||
82b89778 | 1085 | void __set_task_comm(struct task_struct *tsk, const char *buf, bool exec) |
1da177e4 LT |
1086 | { |
1087 | task_lock(tsk); | |
43d2b113 | 1088 | trace_task_rename(tsk, buf); |
1da177e4 LT |
1089 | strlcpy(tsk->comm, buf, sizeof(tsk->comm)); |
1090 | task_unlock(tsk); | |
82b89778 | 1091 | perf_event_comm(tsk, exec); |
1da177e4 LT |
1092 | } |
1093 | ||
1094 | int flush_old_exec(struct linux_binprm * bprm) | |
1095 | { | |
221af7f8 | 1096 | int retval; |
1da177e4 LT |
1097 | |
1098 | /* | |
1099 | * Make sure we have a private signal table and that | |
1100 | * we are unassociated from the previous thread group. | |
1101 | */ | |
1102 | retval = de_thread(current); | |
1103 | if (retval) | |
1104 | goto out; | |
1105 | ||
6e399cd1 DB |
1106 | /* |
1107 | * Must be called _before_ exec_mmap() as bprm->mm is | |
1108 | * not visibile until then. This also enables the update | |
1109 | * to be lockless. | |
1110 | */ | |
925d1c40 | 1111 | set_mm_exe_file(bprm->mm, bprm->file); |
6e399cd1 | 1112 | |
1da177e4 LT |
1113 | /* |
1114 | * Release all of the old mmap stuff | |
1115 | */ | |
3c77f845 | 1116 | acct_arg_size(bprm, 0); |
1da177e4 LT |
1117 | retval = exec_mmap(bprm->mm); |
1118 | if (retval) | |
fd8328be | 1119 | goto out; |
1da177e4 LT |
1120 | |
1121 | bprm->mm = NULL; /* We're using it now */ | |
7ab02af4 | 1122 | |
dac853ae | 1123 | set_fs(USER_DS); |
b88fae64 ZY |
1124 | current->flags &= ~(PF_RANDOMIZE | PF_FORKNOEXEC | PF_KTHREAD | |
1125 | PF_NOFREEZE | PF_NO_SETAFFINITY); | |
7ab02af4 LT |
1126 | flush_thread(); |
1127 | current->personality &= ~bprm->per_clear; | |
1128 | ||
221af7f8 LT |
1129 | return 0; |
1130 | ||
1131 | out: | |
1132 | return retval; | |
1133 | } | |
1134 | EXPORT_SYMBOL(flush_old_exec); | |
1135 | ||
1b5d783c AV |
1136 | void would_dump(struct linux_binprm *bprm, struct file *file) |
1137 | { | |
496ad9aa | 1138 | if (inode_permission(file_inode(file), MAY_READ) < 0) |
1b5d783c AV |
1139 | bprm->interp_flags |= BINPRM_FLAGS_ENFORCE_NONDUMP; |
1140 | } | |
1141 | EXPORT_SYMBOL(would_dump); | |
1142 | ||
221af7f8 LT |
1143 | void setup_new_exec(struct linux_binprm * bprm) |
1144 | { | |
221af7f8 | 1145 | arch_pick_mmap_layout(current->mm); |
1da177e4 LT |
1146 | |
1147 | /* This is the point of no return */ | |
1da177e4 LT |
1148 | current->sas_ss_sp = current->sas_ss_size = 0; |
1149 | ||
8e96e3b7 | 1150 | if (uid_eq(current_euid(), current_uid()) && gid_eq(current_egid(), current_gid())) |
e579d2c2 | 1151 | set_dumpable(current->mm, SUID_DUMP_USER); |
d6e71144 | 1152 | else |
6c5d5238 | 1153 | set_dumpable(current->mm, suid_dumpable); |
d6e71144 | 1154 | |
e041e328 | 1155 | perf_event_exec(); |
82b89778 | 1156 | __set_task_comm(current, kbasename(bprm->filename), true); |
1da177e4 | 1157 | |
0551fbd2 BH |
1158 | /* Set the new mm task size. We have to do that late because it may |
1159 | * depend on TIF_32BIT which is only updated in flush_thread() on | |
1160 | * some architectures like powerpc | |
1161 | */ | |
1162 | current->mm->task_size = TASK_SIZE; | |
1163 | ||
a6f76f23 | 1164 | /* install the new credentials */ |
8e96e3b7 EB |
1165 | if (!uid_eq(bprm->cred->uid, current_euid()) || |
1166 | !gid_eq(bprm->cred->gid, current_egid())) { | |
d2d56c5f | 1167 | current->pdeath_signal = 0; |
1b5d783c AV |
1168 | } else { |
1169 | would_dump(bprm, bprm->file); | |
1170 | if (bprm->interp_flags & BINPRM_FLAGS_ENFORCE_NONDUMP) | |
1171 | set_dumpable(current->mm, suid_dumpable); | |
1da177e4 LT |
1172 | } |
1173 | ||
1174 | /* An exec changes our domain. We are no longer part of the thread | |
1175 | group */ | |
1da177e4 | 1176 | current->self_exec_id++; |
1da177e4 | 1177 | flush_signal_handlers(current, 0); |
6a6d27de | 1178 | do_close_on_exec(current->files); |
1da177e4 | 1179 | } |
221af7f8 | 1180 | EXPORT_SYMBOL(setup_new_exec); |
1da177e4 | 1181 | |
a2a8474c ON |
1182 | /* |
1183 | * Prepare credentials and lock ->cred_guard_mutex. | |
1184 | * install_exec_creds() commits the new creds and drops the lock. | |
1185 | * Or, if exec fails before, free_bprm() should release ->cred and | |
1186 | * and unlock. | |
1187 | */ | |
1188 | int prepare_bprm_creds(struct linux_binprm *bprm) | |
1189 | { | |
9b1bf12d | 1190 | if (mutex_lock_interruptible(¤t->signal->cred_guard_mutex)) |
a2a8474c ON |
1191 | return -ERESTARTNOINTR; |
1192 | ||
1193 | bprm->cred = prepare_exec_creds(); | |
1194 | if (likely(bprm->cred)) | |
1195 | return 0; | |
1196 | ||
9b1bf12d | 1197 | mutex_unlock(¤t->signal->cred_guard_mutex); |
a2a8474c ON |
1198 | return -ENOMEM; |
1199 | } | |
1200 | ||
c4ad8f98 | 1201 | static void free_bprm(struct linux_binprm *bprm) |
a2a8474c ON |
1202 | { |
1203 | free_arg_pages(bprm); | |
1204 | if (bprm->cred) { | |
9b1bf12d | 1205 | mutex_unlock(¤t->signal->cred_guard_mutex); |
a2a8474c ON |
1206 | abort_creds(bprm->cred); |
1207 | } | |
63e46b95 ON |
1208 | if (bprm->file) { |
1209 | allow_write_access(bprm->file); | |
1210 | fput(bprm->file); | |
1211 | } | |
b66c5984 KC |
1212 | /* If a binfmt changed the interp, free it. */ |
1213 | if (bprm->interp != bprm->filename) | |
1214 | kfree(bprm->interp); | |
a2a8474c ON |
1215 | kfree(bprm); |
1216 | } | |
1217 | ||
b66c5984 KC |
1218 | int bprm_change_interp(char *interp, struct linux_binprm *bprm) |
1219 | { | |
1220 | /* If a binfmt changed the interp, free it first. */ | |
1221 | if (bprm->interp != bprm->filename) | |
1222 | kfree(bprm->interp); | |
1223 | bprm->interp = kstrdup(interp, GFP_KERNEL); | |
1224 | if (!bprm->interp) | |
1225 | return -ENOMEM; | |
1226 | return 0; | |
1227 | } | |
1228 | EXPORT_SYMBOL(bprm_change_interp); | |
1229 | ||
a6f76f23 DH |
1230 | /* |
1231 | * install the new credentials for this executable | |
1232 | */ | |
1233 | void install_exec_creds(struct linux_binprm *bprm) | |
1234 | { | |
1235 | security_bprm_committing_creds(bprm); | |
1236 | ||
1237 | commit_creds(bprm->cred); | |
1238 | bprm->cred = NULL; | |
2976b10f SE |
1239 | |
1240 | /* | |
1241 | * Disable monitoring for regular users | |
1242 | * when executing setuid binaries. Must | |
1243 | * wait until new credentials are committed | |
1244 | * by commit_creds() above | |
1245 | */ | |
1246 | if (get_dumpable(current->mm) != SUID_DUMP_USER) | |
1247 | perf_event_exit_task(current); | |
a2a8474c ON |
1248 | /* |
1249 | * cred_guard_mutex must be held at least to this point to prevent | |
a6f76f23 | 1250 | * ptrace_attach() from altering our determination of the task's |
a2a8474c ON |
1251 | * credentials; any time after this it may be unlocked. |
1252 | */ | |
a6f76f23 | 1253 | security_bprm_committed_creds(bprm); |
9b1bf12d | 1254 | mutex_unlock(¤t->signal->cred_guard_mutex); |
a6f76f23 DH |
1255 | } |
1256 | EXPORT_SYMBOL(install_exec_creds); | |
1257 | ||
1258 | /* | |
1259 | * determine how safe it is to execute the proposed program | |
9b1bf12d | 1260 | * - the caller must hold ->cred_guard_mutex to protect against |
c2e1f2e3 | 1261 | * PTRACE_ATTACH or seccomp thread-sync |
a6f76f23 | 1262 | */ |
9e00cdb0 | 1263 | static void check_unsafe_exec(struct linux_binprm *bprm) |
a6f76f23 | 1264 | { |
0bf2f3ae | 1265 | struct task_struct *p = current, *t; |
f1191b50 | 1266 | unsigned n_fs; |
a6f76f23 | 1267 | |
4b9d33e6 TH |
1268 | if (p->ptrace) { |
1269 | if (p->ptrace & PT_PTRACE_CAP) | |
1270 | bprm->unsafe |= LSM_UNSAFE_PTRACE_CAP; | |
1271 | else | |
1272 | bprm->unsafe |= LSM_UNSAFE_PTRACE; | |
1273 | } | |
a6f76f23 | 1274 | |
259e5e6c AL |
1275 | /* |
1276 | * This isn't strictly necessary, but it makes it harder for LSMs to | |
1277 | * mess up. | |
1278 | */ | |
1d4457f9 | 1279 | if (task_no_new_privs(current)) |
259e5e6c AL |
1280 | bprm->unsafe |= LSM_UNSAFE_NO_NEW_PRIVS; |
1281 | ||
83f62a2e | 1282 | t = p; |
0bf2f3ae | 1283 | n_fs = 1; |
2a4419b5 | 1284 | spin_lock(&p->fs->lock); |
437f7fdb | 1285 | rcu_read_lock(); |
83f62a2e | 1286 | while_each_thread(p, t) { |
0bf2f3ae DH |
1287 | if (t->fs == p->fs) |
1288 | n_fs++; | |
0bf2f3ae | 1289 | } |
437f7fdb | 1290 | rcu_read_unlock(); |
0bf2f3ae | 1291 | |
9e00cdb0 | 1292 | if (p->fs->users > n_fs) |
a6f76f23 | 1293 | bprm->unsafe |= LSM_UNSAFE_SHARE; |
9e00cdb0 ON |
1294 | else |
1295 | p->fs->in_exec = 1; | |
2a4419b5 | 1296 | spin_unlock(&p->fs->lock); |
a6f76f23 DH |
1297 | } |
1298 | ||
8b01fc86 JH |
1299 | static void bprm_fill_uid(struct linux_binprm *bprm) |
1300 | { | |
1301 | struct inode *inode; | |
1302 | unsigned int mode; | |
1303 | kuid_t uid; | |
1304 | kgid_t gid; | |
1305 | ||
1306 | /* clear any previous set[ug]id data from a previous binary */ | |
1307 | bprm->cred->euid = current_euid(); | |
1308 | bprm->cred->egid = current_egid(); | |
1309 | ||
090bd5f0 | 1310 | if (path_nosuid(&bprm->file->f_path)) |
8b01fc86 JH |
1311 | return; |
1312 | ||
1313 | if (task_no_new_privs(current)) | |
1314 | return; | |
1315 | ||
1316 | inode = file_inode(bprm->file); | |
1317 | mode = READ_ONCE(inode->i_mode); | |
1318 | if (!(mode & (S_ISUID|S_ISGID))) | |
1319 | return; | |
1320 | ||
1321 | /* Be careful if suid/sgid is set */ | |
1322 | mutex_lock(&inode->i_mutex); | |
1323 | ||
1324 | /* reload atomically mode/uid/gid now that lock held */ | |
1325 | mode = inode->i_mode; | |
1326 | uid = inode->i_uid; | |
1327 | gid = inode->i_gid; | |
1328 | mutex_unlock(&inode->i_mutex); | |
1329 | ||
1330 | /* We ignore suid/sgid if there are no mappings for them in the ns */ | |
1331 | if (!kuid_has_mapping(bprm->cred->user_ns, uid) || | |
1332 | !kgid_has_mapping(bprm->cred->user_ns, gid)) | |
1333 | return; | |
1334 | ||
1335 | if (mode & S_ISUID) { | |
1336 | bprm->per_clear |= PER_CLEAR_ON_SETID; | |
1337 | bprm->cred->euid = uid; | |
1338 | } | |
1339 | ||
1340 | if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) { | |
1341 | bprm->per_clear |= PER_CLEAR_ON_SETID; | |
1342 | bprm->cred->egid = gid; | |
1343 | } | |
1344 | } | |
1345 | ||
9e00cdb0 ON |
1346 | /* |
1347 | * Fill the binprm structure from the inode. | |
1da177e4 | 1348 | * Check permissions, then read the first 128 (BINPRM_BUF_SIZE) bytes |
a6f76f23 DH |
1349 | * |
1350 | * This may be called multiple times for binary chains (scripts for example). | |
1da177e4 LT |
1351 | */ |
1352 | int prepare_binprm(struct linux_binprm *bprm) | |
1353 | { | |
1da177e4 LT |
1354 | int retval; |
1355 | ||
8b01fc86 | 1356 | bprm_fill_uid(bprm); |
1da177e4 LT |
1357 | |
1358 | /* fill in binprm security blob */ | |
a6f76f23 | 1359 | retval = security_bprm_set_creds(bprm); |
1da177e4 LT |
1360 | if (retval) |
1361 | return retval; | |
a6f76f23 | 1362 | bprm->cred_prepared = 1; |
1da177e4 | 1363 | |
a6f76f23 DH |
1364 | memset(bprm->buf, 0, BINPRM_BUF_SIZE); |
1365 | return kernel_read(bprm->file, 0, bprm->buf, BINPRM_BUF_SIZE); | |
1da177e4 LT |
1366 | } |
1367 | ||
1368 | EXPORT_SYMBOL(prepare_binprm); | |
1369 | ||
4fc75ff4 NP |
1370 | /* |
1371 | * Arguments are '\0' separated strings found at the location bprm->p | |
1372 | * points to; chop off the first by relocating brpm->p to right after | |
1373 | * the first '\0' encountered. | |
1374 | */ | |
b6a2fea3 | 1375 | int remove_arg_zero(struct linux_binprm *bprm) |
1da177e4 | 1376 | { |
b6a2fea3 OW |
1377 | int ret = 0; |
1378 | unsigned long offset; | |
1379 | char *kaddr; | |
1380 | struct page *page; | |
4fc75ff4 | 1381 | |
b6a2fea3 OW |
1382 | if (!bprm->argc) |
1383 | return 0; | |
1da177e4 | 1384 | |
b6a2fea3 OW |
1385 | do { |
1386 | offset = bprm->p & ~PAGE_MASK; | |
1387 | page = get_arg_page(bprm, bprm->p, 0); | |
1388 | if (!page) { | |
1389 | ret = -EFAULT; | |
1390 | goto out; | |
1391 | } | |
e8e3c3d6 | 1392 | kaddr = kmap_atomic(page); |
4fc75ff4 | 1393 | |
b6a2fea3 OW |
1394 | for (; offset < PAGE_SIZE && kaddr[offset]; |
1395 | offset++, bprm->p++) | |
1396 | ; | |
4fc75ff4 | 1397 | |
e8e3c3d6 | 1398 | kunmap_atomic(kaddr); |
b6a2fea3 | 1399 | put_arg_page(page); |
4fc75ff4 | 1400 | |
b6a2fea3 OW |
1401 | if (offset == PAGE_SIZE) |
1402 | free_arg_page(bprm, (bprm->p >> PAGE_SHIFT) - 1); | |
1403 | } while (offset == PAGE_SIZE); | |
4fc75ff4 | 1404 | |
b6a2fea3 OW |
1405 | bprm->p++; |
1406 | bprm->argc--; | |
1407 | ret = 0; | |
4fc75ff4 | 1408 | |
b6a2fea3 OW |
1409 | out: |
1410 | return ret; | |
1da177e4 | 1411 | } |
1da177e4 LT |
1412 | EXPORT_SYMBOL(remove_arg_zero); |
1413 | ||
cb7b6b1c | 1414 | #define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e)) |
1da177e4 LT |
1415 | /* |
1416 | * cycle the list of binary formats handler, until one recognizes the image | |
1417 | */ | |
3c456bfc | 1418 | int search_binary_handler(struct linux_binprm *bprm) |
1da177e4 | 1419 | { |
cb7b6b1c | 1420 | bool need_retry = IS_ENABLED(CONFIG_MODULES); |
1da177e4 | 1421 | struct linux_binfmt *fmt; |
cb7b6b1c | 1422 | int retval; |
1da177e4 | 1423 | |
d7402698 | 1424 | /* This allows 4 levels of binfmt rewrites before failing hard. */ |
131b2f9f | 1425 | if (bprm->recursion_depth > 5) |
d7402698 KC |
1426 | return -ELOOP; |
1427 | ||
1da177e4 LT |
1428 | retval = security_bprm_check(bprm); |
1429 | if (retval) | |
1430 | return retval; | |
1431 | ||
1da177e4 | 1432 | retval = -ENOENT; |
cb7b6b1c ON |
1433 | retry: |
1434 | read_lock(&binfmt_lock); | |
1435 | list_for_each_entry(fmt, &formats, lh) { | |
1436 | if (!try_module_get(fmt->module)) | |
1437 | continue; | |
1438 | read_unlock(&binfmt_lock); | |
1439 | bprm->recursion_depth++; | |
1440 | retval = fmt->load_binary(bprm); | |
19d860a1 AV |
1441 | read_lock(&binfmt_lock); |
1442 | put_binfmt(fmt); | |
cb7b6b1c | 1443 | bprm->recursion_depth--; |
19d860a1 AV |
1444 | if (retval < 0 && !bprm->mm) { |
1445 | /* we got to flush_old_exec() and failed after it */ | |
1446 | read_unlock(&binfmt_lock); | |
1447 | force_sigsegv(SIGSEGV, current); | |
1448 | return retval; | |
1449 | } | |
1450 | if (retval != -ENOEXEC || !bprm->file) { | |
1451 | read_unlock(&binfmt_lock); | |
cb7b6b1c | 1452 | return retval; |
1da177e4 | 1453 | } |
1da177e4 | 1454 | } |
cb7b6b1c ON |
1455 | read_unlock(&binfmt_lock); |
1456 | ||
19d860a1 | 1457 | if (need_retry) { |
cb7b6b1c ON |
1458 | if (printable(bprm->buf[0]) && printable(bprm->buf[1]) && |
1459 | printable(bprm->buf[2]) && printable(bprm->buf[3])) | |
1460 | return retval; | |
4e0621a0 ON |
1461 | if (request_module("binfmt-%04x", *(ushort *)(bprm->buf + 2)) < 0) |
1462 | return retval; | |
cb7b6b1c ON |
1463 | need_retry = false; |
1464 | goto retry; | |
1465 | } | |
1466 | ||
1da177e4 LT |
1467 | return retval; |
1468 | } | |
1da177e4 LT |
1469 | EXPORT_SYMBOL(search_binary_handler); |
1470 | ||
5d1baf3b ON |
1471 | static int exec_binprm(struct linux_binprm *bprm) |
1472 | { | |
1473 | pid_t old_pid, old_vpid; | |
1474 | int ret; | |
1475 | ||
1476 | /* Need to fetch pid before load_binary changes it */ | |
1477 | old_pid = current->pid; | |
1478 | rcu_read_lock(); | |
1479 | old_vpid = task_pid_nr_ns(current, task_active_pid_ns(current->parent)); | |
1480 | rcu_read_unlock(); | |
1481 | ||
1482 | ret = search_binary_handler(bprm); | |
1483 | if (ret >= 0) { | |
3eaded86 | 1484 | audit_bprm(bprm); |
5d1baf3b ON |
1485 | trace_sched_process_exec(current, old_pid, bprm); |
1486 | ptrace_event(PTRACE_EVENT_EXEC, old_vpid); | |
9beb266f | 1487 | proc_exec_connector(current); |
5d1baf3b ON |
1488 | } |
1489 | ||
1490 | return ret; | |
1491 | } | |
1492 | ||
1da177e4 LT |
1493 | /* |
1494 | * sys_execve() executes a new program. | |
1495 | */ | |
51f39a1f DD |
1496 | static int do_execveat_common(int fd, struct filename *filename, |
1497 | struct user_arg_ptr argv, | |
1498 | struct user_arg_ptr envp, | |
1499 | int flags) | |
1da177e4 | 1500 | { |
51f39a1f | 1501 | char *pathbuf = NULL; |
1da177e4 LT |
1502 | struct linux_binprm *bprm; |
1503 | struct file *file; | |
3b125388 | 1504 | struct files_struct *displaced; |
1da177e4 | 1505 | int retval; |
72fa5997 | 1506 | |
c4ad8f98 LT |
1507 | if (IS_ERR(filename)) |
1508 | return PTR_ERR(filename); | |
1509 | ||
72fa5997 VK |
1510 | /* |
1511 | * We move the actual failure in case of RLIMIT_NPROC excess from | |
1512 | * set*uid() to execve() because too many poorly written programs | |
1513 | * don't check setuid() return code. Here we additionally recheck | |
1514 | * whether NPROC limit is still exceeded. | |
1515 | */ | |
1516 | if ((current->flags & PF_NPROC_EXCEEDED) && | |
bd9d43f4 | 1517 | atomic_read(¤t_user()->processes) > rlimit(RLIMIT_NPROC)) { |
72fa5997 VK |
1518 | retval = -EAGAIN; |
1519 | goto out_ret; | |
1520 | } | |
1521 | ||
1522 | /* We're below the limit (still or again), so we don't want to make | |
1523 | * further execve() calls fail. */ | |
1524 | current->flags &= ~PF_NPROC_EXCEEDED; | |
1da177e4 | 1525 | |
3b125388 | 1526 | retval = unshare_files(&displaced); |
fd8328be AV |
1527 | if (retval) |
1528 | goto out_ret; | |
1529 | ||
1da177e4 | 1530 | retval = -ENOMEM; |
11b0b5ab | 1531 | bprm = kzalloc(sizeof(*bprm), GFP_KERNEL); |
1da177e4 | 1532 | if (!bprm) |
fd8328be | 1533 | goto out_files; |
1da177e4 | 1534 | |
a2a8474c ON |
1535 | retval = prepare_bprm_creds(bprm); |
1536 | if (retval) | |
a6f76f23 | 1537 | goto out_free; |
498052bb | 1538 | |
9e00cdb0 | 1539 | check_unsafe_exec(bprm); |
a2a8474c | 1540 | current->in_execve = 1; |
a6f76f23 | 1541 | |
51f39a1f | 1542 | file = do_open_execat(fd, filename, flags); |
1da177e4 LT |
1543 | retval = PTR_ERR(file); |
1544 | if (IS_ERR(file)) | |
498052bb | 1545 | goto out_unmark; |
1da177e4 LT |
1546 | |
1547 | sched_exec(); | |
1548 | ||
1da177e4 | 1549 | bprm->file = file; |
51f39a1f DD |
1550 | if (fd == AT_FDCWD || filename->name[0] == '/') { |
1551 | bprm->filename = filename->name; | |
1552 | } else { | |
1553 | if (filename->name[0] == '\0') | |
1554 | pathbuf = kasprintf(GFP_TEMPORARY, "/dev/fd/%d", fd); | |
1555 | else | |
1556 | pathbuf = kasprintf(GFP_TEMPORARY, "/dev/fd/%d/%s", | |
1557 | fd, filename->name); | |
1558 | if (!pathbuf) { | |
1559 | retval = -ENOMEM; | |
1560 | goto out_unmark; | |
1561 | } | |
1562 | /* | |
1563 | * Record that a name derived from an O_CLOEXEC fd will be | |
1564 | * inaccessible after exec. Relies on having exclusive access to | |
1565 | * current->files (due to unshare_files above). | |
1566 | */ | |
1567 | if (close_on_exec(fd, rcu_dereference_raw(current->files->fdt))) | |
1568 | bprm->interp_flags |= BINPRM_FLAGS_PATH_INACCESSIBLE; | |
1569 | bprm->filename = pathbuf; | |
1570 | } | |
1571 | bprm->interp = bprm->filename; | |
1da177e4 | 1572 | |
b6a2fea3 OW |
1573 | retval = bprm_mm_init(bprm); |
1574 | if (retval) | |
63e46b95 | 1575 | goto out_unmark; |
1da177e4 | 1576 | |
b6a2fea3 | 1577 | bprm->argc = count(argv, MAX_ARG_STRINGS); |
1da177e4 | 1578 | if ((retval = bprm->argc) < 0) |
a6f76f23 | 1579 | goto out; |
1da177e4 | 1580 | |
b6a2fea3 | 1581 | bprm->envc = count(envp, MAX_ARG_STRINGS); |
1da177e4 | 1582 | if ((retval = bprm->envc) < 0) |
1da177e4 LT |
1583 | goto out; |
1584 | ||
1585 | retval = prepare_binprm(bprm); | |
1586 | if (retval < 0) | |
1587 | goto out; | |
1588 | ||
1589 | retval = copy_strings_kernel(1, &bprm->filename, bprm); | |
1590 | if (retval < 0) | |
1591 | goto out; | |
1592 | ||
1593 | bprm->exec = bprm->p; | |
1594 | retval = copy_strings(bprm->envc, envp, bprm); | |
1595 | if (retval < 0) | |
1596 | goto out; | |
1597 | ||
1598 | retval = copy_strings(bprm->argc, argv, bprm); | |
1599 | if (retval < 0) | |
1600 | goto out; | |
1601 | ||
5d1baf3b | 1602 | retval = exec_binprm(bprm); |
a6f76f23 DH |
1603 | if (retval < 0) |
1604 | goto out; | |
1da177e4 | 1605 | |
a6f76f23 | 1606 | /* execve succeeded */ |
498052bb | 1607 | current->fs->in_exec = 0; |
f9ce1f1c | 1608 | current->in_execve = 0; |
a6f76f23 | 1609 | acct_update_integrals(current); |
82727018 | 1610 | task_numa_free(current); |
a6f76f23 | 1611 | free_bprm(bprm); |
51f39a1f | 1612 | kfree(pathbuf); |
c4ad8f98 | 1613 | putname(filename); |
a6f76f23 DH |
1614 | if (displaced) |
1615 | put_files_struct(displaced); | |
1616 | return retval; | |
1da177e4 | 1617 | |
a6f76f23 | 1618 | out: |
3c77f845 ON |
1619 | if (bprm->mm) { |
1620 | acct_arg_size(bprm, 0); | |
1621 | mmput(bprm->mm); | |
1622 | } | |
1da177e4 | 1623 | |
498052bb | 1624 | out_unmark: |
9e00cdb0 | 1625 | current->fs->in_exec = 0; |
f9ce1f1c | 1626 | current->in_execve = 0; |
a6f76f23 DH |
1627 | |
1628 | out_free: | |
08a6fac1 | 1629 | free_bprm(bprm); |
51f39a1f | 1630 | kfree(pathbuf); |
1da177e4 | 1631 | |
fd8328be | 1632 | out_files: |
3b125388 AV |
1633 | if (displaced) |
1634 | reset_files_struct(displaced); | |
1da177e4 | 1635 | out_ret: |
c4ad8f98 | 1636 | putname(filename); |
1da177e4 LT |
1637 | return retval; |
1638 | } | |
1639 | ||
c4ad8f98 | 1640 | int do_execve(struct filename *filename, |
ba2d0162 | 1641 | const char __user *const __user *__argv, |
da3d4c5f | 1642 | const char __user *const __user *__envp) |
ba2d0162 | 1643 | { |
0e028465 ON |
1644 | struct user_arg_ptr argv = { .ptr.native = __argv }; |
1645 | struct user_arg_ptr envp = { .ptr.native = __envp }; | |
51f39a1f DD |
1646 | return do_execveat_common(AT_FDCWD, filename, argv, envp, 0); |
1647 | } | |
1648 | ||
1649 | int do_execveat(int fd, struct filename *filename, | |
1650 | const char __user *const __user *__argv, | |
1651 | const char __user *const __user *__envp, | |
1652 | int flags) | |
1653 | { | |
1654 | struct user_arg_ptr argv = { .ptr.native = __argv }; | |
1655 | struct user_arg_ptr envp = { .ptr.native = __envp }; | |
1656 | ||
1657 | return do_execveat_common(fd, filename, argv, envp, flags); | |
0e028465 ON |
1658 | } |
1659 | ||
1660 | #ifdef CONFIG_COMPAT | |
c4ad8f98 | 1661 | static int compat_do_execve(struct filename *filename, |
38b983b3 | 1662 | const compat_uptr_t __user *__argv, |
d03d26e5 | 1663 | const compat_uptr_t __user *__envp) |
0e028465 ON |
1664 | { |
1665 | struct user_arg_ptr argv = { | |
1666 | .is_compat = true, | |
1667 | .ptr.compat = __argv, | |
1668 | }; | |
1669 | struct user_arg_ptr envp = { | |
1670 | .is_compat = true, | |
1671 | .ptr.compat = __envp, | |
1672 | }; | |
51f39a1f DD |
1673 | return do_execveat_common(AT_FDCWD, filename, argv, envp, 0); |
1674 | } | |
1675 | ||
1676 | static int compat_do_execveat(int fd, struct filename *filename, | |
1677 | const compat_uptr_t __user *__argv, | |
1678 | const compat_uptr_t __user *__envp, | |
1679 | int flags) | |
1680 | { | |
1681 | struct user_arg_ptr argv = { | |
1682 | .is_compat = true, | |
1683 | .ptr.compat = __argv, | |
1684 | }; | |
1685 | struct user_arg_ptr envp = { | |
1686 | .is_compat = true, | |
1687 | .ptr.compat = __envp, | |
1688 | }; | |
1689 | return do_execveat_common(fd, filename, argv, envp, flags); | |
ba2d0162 | 1690 | } |
0e028465 | 1691 | #endif |
ba2d0162 | 1692 | |
964ee7df | 1693 | void set_binfmt(struct linux_binfmt *new) |
1da177e4 | 1694 | { |
801460d0 HS |
1695 | struct mm_struct *mm = current->mm; |
1696 | ||
1697 | if (mm->binfmt) | |
1698 | module_put(mm->binfmt->module); | |
1da177e4 | 1699 | |
801460d0 | 1700 | mm->binfmt = new; |
964ee7df ON |
1701 | if (new) |
1702 | __module_get(new->module); | |
1da177e4 | 1703 | } |
1da177e4 LT |
1704 | EXPORT_SYMBOL(set_binfmt); |
1705 | ||
6c5d5238 | 1706 | /* |
7288e118 | 1707 | * set_dumpable stores three-value SUID_DUMP_* into mm->flags. |
6c5d5238 KH |
1708 | */ |
1709 | void set_dumpable(struct mm_struct *mm, int value) | |
1710 | { | |
abacd2fe ON |
1711 | unsigned long old, new; |
1712 | ||
7288e118 ON |
1713 | if (WARN_ON((unsigned)value > SUID_DUMP_ROOT)) |
1714 | return; | |
1715 | ||
abacd2fe ON |
1716 | do { |
1717 | old = ACCESS_ONCE(mm->flags); | |
7288e118 | 1718 | new = (old & ~MMF_DUMPABLE_MASK) | value; |
abacd2fe | 1719 | } while (cmpxchg(&mm->flags, old, new) != old); |
6c5d5238 | 1720 | } |
6c5d5238 | 1721 | |
38b983b3 AV |
1722 | SYSCALL_DEFINE3(execve, |
1723 | const char __user *, filename, | |
1724 | const char __user *const __user *, argv, | |
1725 | const char __user *const __user *, envp) | |
1726 | { | |
c4ad8f98 | 1727 | return do_execve(getname(filename), argv, envp); |
38b983b3 | 1728 | } |
51f39a1f DD |
1729 | |
1730 | SYSCALL_DEFINE5(execveat, | |
1731 | int, fd, const char __user *, filename, | |
1732 | const char __user *const __user *, argv, | |
1733 | const char __user *const __user *, envp, | |
1734 | int, flags) | |
1735 | { | |
1736 | int lookup_flags = (flags & AT_EMPTY_PATH) ? LOOKUP_EMPTY : 0; | |
1737 | ||
1738 | return do_execveat(fd, | |
1739 | getname_flags(filename, lookup_flags, NULL), | |
1740 | argv, envp, flags); | |
1741 | } | |
1742 | ||
38b983b3 | 1743 | #ifdef CONFIG_COMPAT |
625b1d7e HC |
1744 | COMPAT_SYSCALL_DEFINE3(execve, const char __user *, filename, |
1745 | const compat_uptr_t __user *, argv, | |
1746 | const compat_uptr_t __user *, envp) | |
38b983b3 | 1747 | { |
c4ad8f98 | 1748 | return compat_do_execve(getname(filename), argv, envp); |
38b983b3 | 1749 | } |
51f39a1f DD |
1750 | |
1751 | COMPAT_SYSCALL_DEFINE5(execveat, int, fd, | |
1752 | const char __user *, filename, | |
1753 | const compat_uptr_t __user *, argv, | |
1754 | const compat_uptr_t __user *, envp, | |
1755 | int, flags) | |
1756 | { | |
1757 | int lookup_flags = (flags & AT_EMPTY_PATH) ? LOOKUP_EMPTY : 0; | |
1758 | ||
1759 | return compat_do_execveat(fd, | |
1760 | getname_flags(filename, lookup_flags, NULL), | |
1761 | argv, envp, flags); | |
1762 | } | |
38b983b3 | 1763 | #endif |