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> |
1da177e4 LT |
29 | #include <linux/stat.h> |
30 | #include <linux/fcntl.h> | |
ba92a43d | 31 | #include <linux/swap.h> |
74aadce9 | 32 | #include <linux/string.h> |
1da177e4 | 33 | #include <linux/init.h> |
ca5b172b | 34 | #include <linux/pagemap.h> |
cdd6c482 | 35 | #include <linux/perf_event.h> |
1da177e4 LT |
36 | #include <linux/highmem.h> |
37 | #include <linux/spinlock.h> | |
38 | #include <linux/key.h> | |
39 | #include <linux/personality.h> | |
40 | #include <linux/binfmts.h> | |
1da177e4 | 41 | #include <linux/utsname.h> |
84d73786 | 42 | #include <linux/pid_namespace.h> |
1da177e4 LT |
43 | #include <linux/module.h> |
44 | #include <linux/namei.h> | |
1da177e4 LT |
45 | #include <linux/mount.h> |
46 | #include <linux/security.h> | |
47 | #include <linux/syscalls.h> | |
8f0ab514 | 48 | #include <linux/tsacct_kern.h> |
9f46080c | 49 | #include <linux/cn_proc.h> |
473ae30b | 50 | #include <linux/audit.h> |
6341c393 | 51 | #include <linux/tracehook.h> |
5f4123be | 52 | #include <linux/kmod.h> |
6110e3ab | 53 | #include <linux/fsnotify.h> |
5ad4e53b | 54 | #include <linux/fs_struct.h> |
61be228a | 55 | #include <linux/pipe_fs_i.h> |
3d5992d2 | 56 | #include <linux/oom.h> |
0e028465 | 57 | #include <linux/compat.h> |
1da177e4 LT |
58 | |
59 | #include <asm/uaccess.h> | |
60 | #include <asm/mmu_context.h> | |
b6a2fea3 | 61 | #include <asm/tlb.h> |
43d2b113 KH |
62 | |
63 | #include <trace/events/task.h> | |
a6f76f23 | 64 | #include "internal.h" |
1da177e4 | 65 | |
4ff16c25 DS |
66 | #include <trace/events/sched.h> |
67 | ||
1da177e4 | 68 | int core_uses_pid; |
71ce92f3 | 69 | char core_pattern[CORENAME_MAX_SIZE] = "core"; |
a293980c | 70 | unsigned int core_pipe_limit; |
d6e71144 AC |
71 | int suid_dumpable = 0; |
72 | ||
1b0d300b XF |
73 | struct core_name { |
74 | char *corename; | |
75 | int used, size; | |
76 | }; | |
77 | static atomic_t call_count = ATOMIC_INIT(1); | |
78 | ||
1da177e4 LT |
79 | /* The maximal length of core_pattern is also specified in sysctl.c */ |
80 | ||
e4dc1b14 | 81 | static LIST_HEAD(formats); |
1da177e4 LT |
82 | static DEFINE_RWLOCK(binfmt_lock); |
83 | ||
8fc3dc5a | 84 | void __register_binfmt(struct linux_binfmt * fmt, int insert) |
1da177e4 | 85 | { |
8fc3dc5a | 86 | BUG_ON(!fmt); |
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 | ||
109 | /* | |
110 | * Note that a shared library must be both readable and executable due to | |
111 | * security reasons. | |
112 | * | |
113 | * Also note that we take the address to load from from the file itself. | |
114 | */ | |
1e7bfb21 | 115 | SYSCALL_DEFINE1(uselib, const char __user *, library) |
1da177e4 | 116 | { |
964bd183 | 117 | struct file *file; |
964bd183 AV |
118 | char *tmp = getname(library); |
119 | int error = PTR_ERR(tmp); | |
47c805dc AV |
120 | static const struct open_flags uselib_flags = { |
121 | .open_flag = O_LARGEFILE | O_RDONLY | __FMODE_EXEC, | |
122 | .acc_mode = MAY_READ | MAY_EXEC | MAY_OPEN, | |
123 | .intent = LOOKUP_OPEN | |
124 | }; | |
964bd183 | 125 | |
6e8341a1 AV |
126 | if (IS_ERR(tmp)) |
127 | goto out; | |
128 | ||
47c805dc | 129 | file = do_filp_open(AT_FDCWD, tmp, &uselib_flags, LOOKUP_FOLLOW); |
6e8341a1 AV |
130 | putname(tmp); |
131 | error = PTR_ERR(file); | |
132 | if (IS_ERR(file)) | |
1da177e4 LT |
133 | goto out; |
134 | ||
135 | error = -EINVAL; | |
6e8341a1 | 136 | if (!S_ISREG(file->f_path.dentry->d_inode->i_mode)) |
1da177e4 LT |
137 | goto exit; |
138 | ||
30524472 | 139 | error = -EACCES; |
6e8341a1 | 140 | if (file->f_path.mnt->mnt_flags & MNT_NOEXEC) |
1da177e4 LT |
141 | goto exit; |
142 | ||
2a12a9d7 | 143 | fsnotify_open(file); |
6110e3ab | 144 | |
1da177e4 LT |
145 | error = -ENOEXEC; |
146 | if(file->f_op) { | |
147 | struct linux_binfmt * fmt; | |
148 | ||
149 | read_lock(&binfmt_lock); | |
e4dc1b14 | 150 | list_for_each_entry(fmt, &formats, lh) { |
1da177e4 LT |
151 | if (!fmt->load_shlib) |
152 | continue; | |
153 | if (!try_module_get(fmt->module)) | |
154 | continue; | |
155 | read_unlock(&binfmt_lock); | |
156 | error = fmt->load_shlib(file); | |
157 | read_lock(&binfmt_lock); | |
158 | put_binfmt(fmt); | |
159 | if (error != -ENOEXEC) | |
160 | break; | |
161 | } | |
162 | read_unlock(&binfmt_lock); | |
163 | } | |
6e8341a1 | 164 | exit: |
1da177e4 LT |
165 | fput(file); |
166 | out: | |
167 | return error; | |
1da177e4 LT |
168 | } |
169 | ||
b6a2fea3 | 170 | #ifdef CONFIG_MMU |
ae6b585e ON |
171 | /* |
172 | * The nascent bprm->mm is not visible until exec_mmap() but it can | |
173 | * use a lot of memory, account these pages in current->mm temporary | |
174 | * for oom_badness()->get_mm_rss(). Once exec succeeds or fails, we | |
175 | * change the counter back via acct_arg_size(0). | |
176 | */ | |
0e028465 | 177 | static void acct_arg_size(struct linux_binprm *bprm, unsigned long pages) |
3c77f845 ON |
178 | { |
179 | struct mm_struct *mm = current->mm; | |
180 | long diff = (long)(pages - bprm->vma_pages); | |
181 | ||
182 | if (!mm || !diff) | |
183 | return; | |
184 | ||
185 | bprm->vma_pages = pages; | |
3c77f845 | 186 | add_mm_counter(mm, MM_ANONPAGES, diff); |
3c77f845 ON |
187 | } |
188 | ||
0e028465 | 189 | static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos, |
b6a2fea3 OW |
190 | int write) |
191 | { | |
192 | struct page *page; | |
193 | int ret; | |
194 | ||
195 | #ifdef CONFIG_STACK_GROWSUP | |
196 | if (write) { | |
d05f3169 | 197 | ret = expand_downwards(bprm->vma, pos); |
b6a2fea3 OW |
198 | if (ret < 0) |
199 | return NULL; | |
200 | } | |
201 | #endif | |
202 | ret = get_user_pages(current, bprm->mm, pos, | |
203 | 1, write, 1, &page, NULL); | |
204 | if (ret <= 0) | |
205 | return NULL; | |
206 | ||
207 | if (write) { | |
b6a2fea3 | 208 | unsigned long size = bprm->vma->vm_end - bprm->vma->vm_start; |
a64e715f LT |
209 | struct rlimit *rlim; |
210 | ||
3c77f845 ON |
211 | acct_arg_size(bprm, size / PAGE_SIZE); |
212 | ||
a64e715f LT |
213 | /* |
214 | * We've historically supported up to 32 pages (ARG_MAX) | |
215 | * of argument strings even with small stacks | |
216 | */ | |
217 | if (size <= ARG_MAX) | |
218 | return page; | |
b6a2fea3 OW |
219 | |
220 | /* | |
221 | * Limit to 1/4-th the stack size for the argv+env strings. | |
222 | * This ensures that: | |
223 | * - the remaining binfmt code will not run out of stack space, | |
224 | * - the program will have a reasonable amount of stack left | |
225 | * to work from. | |
226 | */ | |
a64e715f | 227 | rlim = current->signal->rlim; |
d554ed89 | 228 | if (size > ACCESS_ONCE(rlim[RLIMIT_STACK].rlim_cur) / 4) { |
b6a2fea3 OW |
229 | put_page(page); |
230 | return NULL; | |
231 | } | |
232 | } | |
233 | ||
234 | return page; | |
235 | } | |
236 | ||
237 | static void put_arg_page(struct page *page) | |
238 | { | |
239 | put_page(page); | |
240 | } | |
241 | ||
242 | static void free_arg_page(struct linux_binprm *bprm, int i) | |
243 | { | |
244 | } | |
245 | ||
246 | static void free_arg_pages(struct linux_binprm *bprm) | |
247 | { | |
248 | } | |
249 | ||
250 | static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos, | |
251 | struct page *page) | |
252 | { | |
253 | flush_cache_page(bprm->vma, pos, page_to_pfn(page)); | |
254 | } | |
255 | ||
256 | static int __bprm_mm_init(struct linux_binprm *bprm) | |
257 | { | |
eaccbfa5 | 258 | int err; |
b6a2fea3 OW |
259 | struct vm_area_struct *vma = NULL; |
260 | struct mm_struct *mm = bprm->mm; | |
261 | ||
262 | bprm->vma = vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL); | |
263 | if (!vma) | |
eaccbfa5 | 264 | return -ENOMEM; |
b6a2fea3 OW |
265 | |
266 | down_write(&mm->mmap_sem); | |
267 | vma->vm_mm = mm; | |
268 | ||
269 | /* | |
270 | * Place the stack at the largest stack address the architecture | |
271 | * supports. Later, we'll move this to an appropriate place. We don't | |
272 | * use STACK_TOP because that can depend on attributes which aren't | |
273 | * configured yet. | |
274 | */ | |
aacb3d17 | 275 | BUILD_BUG_ON(VM_STACK_FLAGS & VM_STACK_INCOMPLETE_SETUP); |
b6a2fea3 OW |
276 | vma->vm_end = STACK_TOP_MAX; |
277 | vma->vm_start = vma->vm_end - PAGE_SIZE; | |
a8bef8ff | 278 | vma->vm_flags = VM_STACK_FLAGS | VM_STACK_INCOMPLETE_SETUP; |
3ed75eb8 | 279 | vma->vm_page_prot = vm_get_page_prot(vma->vm_flags); |
5beb4930 | 280 | INIT_LIST_HEAD(&vma->anon_vma_chain); |
462e635e | 281 | |
b6a2fea3 | 282 | err = insert_vm_struct(mm, vma); |
eaccbfa5 | 283 | if (err) |
b6a2fea3 | 284 | goto err; |
b6a2fea3 OW |
285 | |
286 | mm->stack_vm = mm->total_vm = 1; | |
287 | up_write(&mm->mmap_sem); | |
b6a2fea3 | 288 | bprm->p = vma->vm_end - sizeof(void *); |
b6a2fea3 | 289 | return 0; |
b6a2fea3 | 290 | err: |
eaccbfa5 LFC |
291 | up_write(&mm->mmap_sem); |
292 | bprm->vma = NULL; | |
293 | kmem_cache_free(vm_area_cachep, vma); | |
b6a2fea3 OW |
294 | return err; |
295 | } | |
296 | ||
297 | static bool valid_arg_len(struct linux_binprm *bprm, long len) | |
298 | { | |
299 | return len <= MAX_ARG_STRLEN; | |
300 | } | |
301 | ||
302 | #else | |
303 | ||
0e028465 | 304 | static inline void acct_arg_size(struct linux_binprm *bprm, unsigned long pages) |
3c77f845 ON |
305 | { |
306 | } | |
307 | ||
0e028465 | 308 | static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos, |
b6a2fea3 OW |
309 | int write) |
310 | { | |
311 | struct page *page; | |
312 | ||
313 | page = bprm->page[pos / PAGE_SIZE]; | |
314 | if (!page && write) { | |
315 | page = alloc_page(GFP_HIGHUSER|__GFP_ZERO); | |
316 | if (!page) | |
317 | return NULL; | |
318 | bprm->page[pos / PAGE_SIZE] = page; | |
319 | } | |
320 | ||
321 | return page; | |
322 | } | |
323 | ||
324 | static void put_arg_page(struct page *page) | |
325 | { | |
326 | } | |
327 | ||
328 | static void free_arg_page(struct linux_binprm *bprm, int i) | |
329 | { | |
330 | if (bprm->page[i]) { | |
331 | __free_page(bprm->page[i]); | |
332 | bprm->page[i] = NULL; | |
333 | } | |
334 | } | |
335 | ||
336 | static void free_arg_pages(struct linux_binprm *bprm) | |
337 | { | |
338 | int i; | |
339 | ||
340 | for (i = 0; i < MAX_ARG_PAGES; i++) | |
341 | free_arg_page(bprm, i); | |
342 | } | |
343 | ||
344 | static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos, | |
345 | struct page *page) | |
346 | { | |
347 | } | |
348 | ||
349 | static int __bprm_mm_init(struct linux_binprm *bprm) | |
350 | { | |
351 | bprm->p = PAGE_SIZE * MAX_ARG_PAGES - sizeof(void *); | |
352 | return 0; | |
353 | } | |
354 | ||
355 | static bool valid_arg_len(struct linux_binprm *bprm, long len) | |
356 | { | |
357 | return len <= bprm->p; | |
358 | } | |
359 | ||
360 | #endif /* CONFIG_MMU */ | |
361 | ||
362 | /* | |
363 | * Create a new mm_struct and populate it with a temporary stack | |
364 | * vm_area_struct. We don't have enough context at this point to set the stack | |
365 | * flags, permissions, and offset, so we use temporary values. We'll update | |
366 | * them later in setup_arg_pages(). | |
367 | */ | |
368 | int bprm_mm_init(struct linux_binprm *bprm) | |
369 | { | |
370 | int err; | |
371 | struct mm_struct *mm = NULL; | |
372 | ||
373 | bprm->mm = mm = mm_alloc(); | |
374 | err = -ENOMEM; | |
375 | if (!mm) | |
376 | goto err; | |
377 | ||
378 | err = init_new_context(current, mm); | |
379 | if (err) | |
380 | goto err; | |
381 | ||
382 | err = __bprm_mm_init(bprm); | |
383 | if (err) | |
384 | goto err; | |
385 | ||
386 | return 0; | |
387 | ||
388 | err: | |
389 | if (mm) { | |
390 | bprm->mm = NULL; | |
391 | mmdrop(mm); | |
392 | } | |
393 | ||
394 | return err; | |
395 | } | |
396 | ||
ba2d0162 | 397 | struct user_arg_ptr { |
0e028465 ON |
398 | #ifdef CONFIG_COMPAT |
399 | bool is_compat; | |
400 | #endif | |
401 | union { | |
402 | const char __user *const __user *native; | |
403 | #ifdef CONFIG_COMPAT | |
404 | compat_uptr_t __user *compat; | |
405 | #endif | |
406 | } ptr; | |
ba2d0162 ON |
407 | }; |
408 | ||
409 | static const char __user *get_user_arg_ptr(struct user_arg_ptr argv, int nr) | |
1d1dbf81 | 410 | { |
0e028465 ON |
411 | const char __user *native; |
412 | ||
413 | #ifdef CONFIG_COMPAT | |
414 | if (unlikely(argv.is_compat)) { | |
415 | compat_uptr_t compat; | |
416 | ||
417 | if (get_user(compat, argv.ptr.compat + nr)) | |
418 | return ERR_PTR(-EFAULT); | |
1d1dbf81 | 419 | |
0e028465 ON |
420 | return compat_ptr(compat); |
421 | } | |
422 | #endif | |
423 | ||
424 | if (get_user(native, argv.ptr.native + nr)) | |
1d1dbf81 ON |
425 | return ERR_PTR(-EFAULT); |
426 | ||
0e028465 | 427 | return native; |
1d1dbf81 ON |
428 | } |
429 | ||
1da177e4 LT |
430 | /* |
431 | * count() counts the number of strings in array ARGV. | |
432 | */ | |
ba2d0162 | 433 | static int count(struct user_arg_ptr argv, int max) |
1da177e4 LT |
434 | { |
435 | int i = 0; | |
436 | ||
0e028465 | 437 | if (argv.ptr.native != NULL) { |
1da177e4 | 438 | for (;;) { |
1d1dbf81 | 439 | const char __user *p = get_user_arg_ptr(argv, i); |
1da177e4 | 440 | |
1da177e4 LT |
441 | if (!p) |
442 | break; | |
1d1dbf81 ON |
443 | |
444 | if (IS_ERR(p)) | |
445 | return -EFAULT; | |
446 | ||
362e6663 | 447 | if (i++ >= max) |
1da177e4 | 448 | return -E2BIG; |
9aea5a65 RM |
449 | |
450 | if (fatal_signal_pending(current)) | |
451 | return -ERESTARTNOHAND; | |
1da177e4 LT |
452 | cond_resched(); |
453 | } | |
454 | } | |
455 | return i; | |
456 | } | |
457 | ||
458 | /* | |
b6a2fea3 OW |
459 | * 'copy_strings()' copies argument/environment strings from the old |
460 | * processes's memory to the new process's stack. The call to get_user_pages() | |
461 | * ensures the destination page is created and not swapped out. | |
1da177e4 | 462 | */ |
ba2d0162 | 463 | static int copy_strings(int argc, struct user_arg_ptr argv, |
75c96f85 | 464 | struct linux_binprm *bprm) |
1da177e4 LT |
465 | { |
466 | struct page *kmapped_page = NULL; | |
467 | char *kaddr = NULL; | |
b6a2fea3 | 468 | unsigned long kpos = 0; |
1da177e4 LT |
469 | int ret; |
470 | ||
471 | while (argc-- > 0) { | |
d7627467 | 472 | const char __user *str; |
1da177e4 LT |
473 | int len; |
474 | unsigned long pos; | |
475 | ||
1d1dbf81 ON |
476 | ret = -EFAULT; |
477 | str = get_user_arg_ptr(argv, argc); | |
478 | if (IS_ERR(str)) | |
1da177e4 | 479 | goto out; |
1da177e4 | 480 | |
1d1dbf81 ON |
481 | len = strnlen_user(str, MAX_ARG_STRLEN); |
482 | if (!len) | |
483 | goto out; | |
484 | ||
485 | ret = -E2BIG; | |
486 | if (!valid_arg_len(bprm, len)) | |
1da177e4 | 487 | goto out; |
1da177e4 | 488 | |
b6a2fea3 | 489 | /* We're going to work our way backwords. */ |
1da177e4 | 490 | pos = bprm->p; |
b6a2fea3 OW |
491 | str += len; |
492 | bprm->p -= len; | |
1da177e4 LT |
493 | |
494 | while (len > 0) { | |
1da177e4 | 495 | int offset, bytes_to_copy; |
1da177e4 | 496 | |
9aea5a65 RM |
497 | if (fatal_signal_pending(current)) { |
498 | ret = -ERESTARTNOHAND; | |
499 | goto out; | |
500 | } | |
7993bc1f RM |
501 | cond_resched(); |
502 | ||
1da177e4 | 503 | offset = pos % PAGE_SIZE; |
b6a2fea3 OW |
504 | if (offset == 0) |
505 | offset = PAGE_SIZE; | |
506 | ||
507 | bytes_to_copy = offset; | |
508 | if (bytes_to_copy > len) | |
509 | bytes_to_copy = len; | |
510 | ||
511 | offset -= bytes_to_copy; | |
512 | pos -= bytes_to_copy; | |
513 | str -= bytes_to_copy; | |
514 | len -= bytes_to_copy; | |
515 | ||
516 | if (!kmapped_page || kpos != (pos & PAGE_MASK)) { | |
517 | struct page *page; | |
518 | ||
519 | page = get_arg_page(bprm, pos, 1); | |
1da177e4 | 520 | if (!page) { |
b6a2fea3 | 521 | ret = -E2BIG; |
1da177e4 LT |
522 | goto out; |
523 | } | |
1da177e4 | 524 | |
b6a2fea3 OW |
525 | if (kmapped_page) { |
526 | flush_kernel_dcache_page(kmapped_page); | |
1da177e4 | 527 | kunmap(kmapped_page); |
b6a2fea3 OW |
528 | put_arg_page(kmapped_page); |
529 | } | |
1da177e4 LT |
530 | kmapped_page = page; |
531 | kaddr = kmap(kmapped_page); | |
b6a2fea3 OW |
532 | kpos = pos & PAGE_MASK; |
533 | flush_arg_page(bprm, kpos, kmapped_page); | |
1da177e4 | 534 | } |
b6a2fea3 | 535 | if (copy_from_user(kaddr+offset, str, bytes_to_copy)) { |
1da177e4 LT |
536 | ret = -EFAULT; |
537 | goto out; | |
538 | } | |
1da177e4 LT |
539 | } |
540 | } | |
541 | ret = 0; | |
542 | out: | |
b6a2fea3 OW |
543 | if (kmapped_page) { |
544 | flush_kernel_dcache_page(kmapped_page); | |
1da177e4 | 545 | kunmap(kmapped_page); |
b6a2fea3 OW |
546 | put_arg_page(kmapped_page); |
547 | } | |
1da177e4 LT |
548 | return ret; |
549 | } | |
550 | ||
551 | /* | |
552 | * Like copy_strings, but get argv and its values from kernel memory. | |
553 | */ | |
ba2d0162 | 554 | int copy_strings_kernel(int argc, const char *const *__argv, |
d7627467 | 555 | struct linux_binprm *bprm) |
1da177e4 LT |
556 | { |
557 | int r; | |
558 | mm_segment_t oldfs = get_fs(); | |
ba2d0162 | 559 | struct user_arg_ptr argv = { |
0e028465 | 560 | .ptr.native = (const char __user *const __user *)__argv, |
ba2d0162 ON |
561 | }; |
562 | ||
1da177e4 | 563 | set_fs(KERNEL_DS); |
ba2d0162 | 564 | r = copy_strings(argc, argv, bprm); |
1da177e4 | 565 | set_fs(oldfs); |
ba2d0162 | 566 | |
1da177e4 LT |
567 | return r; |
568 | } | |
1da177e4 LT |
569 | EXPORT_SYMBOL(copy_strings_kernel); |
570 | ||
571 | #ifdef CONFIG_MMU | |
b6a2fea3 | 572 | |
1da177e4 | 573 | /* |
b6a2fea3 OW |
574 | * During bprm_mm_init(), we create a temporary stack at STACK_TOP_MAX. Once |
575 | * the binfmt code determines where the new stack should reside, we shift it to | |
576 | * its final location. The process proceeds as follows: | |
1da177e4 | 577 | * |
b6a2fea3 OW |
578 | * 1) Use shift to calculate the new vma endpoints. |
579 | * 2) Extend vma to cover both the old and new ranges. This ensures the | |
580 | * arguments passed to subsequent functions are consistent. | |
581 | * 3) Move vma's page tables to the new range. | |
582 | * 4) Free up any cleared pgd range. | |
583 | * 5) Shrink the vma to cover only the new range. | |
1da177e4 | 584 | */ |
b6a2fea3 | 585 | static int shift_arg_pages(struct vm_area_struct *vma, unsigned long shift) |
1da177e4 LT |
586 | { |
587 | struct mm_struct *mm = vma->vm_mm; | |
b6a2fea3 OW |
588 | unsigned long old_start = vma->vm_start; |
589 | unsigned long old_end = vma->vm_end; | |
590 | unsigned long length = old_end - old_start; | |
591 | unsigned long new_start = old_start - shift; | |
592 | unsigned long new_end = old_end - shift; | |
d16dfc55 | 593 | struct mmu_gather tlb; |
1da177e4 | 594 | |
b6a2fea3 | 595 | BUG_ON(new_start > new_end); |
1da177e4 | 596 | |
b6a2fea3 OW |
597 | /* |
598 | * ensure there are no vmas between where we want to go | |
599 | * and where we are | |
600 | */ | |
601 | if (vma != find_vma(mm, new_start)) | |
602 | return -EFAULT; | |
603 | ||
604 | /* | |
605 | * cover the whole range: [new_start, old_end) | |
606 | */ | |
5beb4930 RR |
607 | if (vma_adjust(vma, new_start, old_end, vma->vm_pgoff, NULL)) |
608 | return -ENOMEM; | |
b6a2fea3 OW |
609 | |
610 | /* | |
611 | * move the page tables downwards, on failure we rely on | |
612 | * process cleanup to remove whatever mess we made. | |
613 | */ | |
614 | if (length != move_page_tables(vma, old_start, | |
615 | vma, new_start, length)) | |
616 | return -ENOMEM; | |
617 | ||
618 | lru_add_drain(); | |
d16dfc55 | 619 | tlb_gather_mmu(&tlb, mm, 0); |
b6a2fea3 OW |
620 | if (new_end > old_start) { |
621 | /* | |
622 | * when the old and new regions overlap clear from new_end. | |
623 | */ | |
d16dfc55 | 624 | free_pgd_range(&tlb, new_end, old_end, new_end, |
b6a2fea3 OW |
625 | vma->vm_next ? vma->vm_next->vm_start : 0); |
626 | } else { | |
627 | /* | |
628 | * otherwise, clean from old_start; this is done to not touch | |
629 | * the address space in [new_end, old_start) some architectures | |
630 | * have constraints on va-space that make this illegal (IA64) - | |
631 | * for the others its just a little faster. | |
632 | */ | |
d16dfc55 | 633 | free_pgd_range(&tlb, old_start, old_end, new_end, |
b6a2fea3 | 634 | vma->vm_next ? vma->vm_next->vm_start : 0); |
1da177e4 | 635 | } |
d16dfc55 | 636 | tlb_finish_mmu(&tlb, new_end, old_end); |
b6a2fea3 OW |
637 | |
638 | /* | |
5beb4930 | 639 | * Shrink the vma to just the new range. Always succeeds. |
b6a2fea3 OW |
640 | */ |
641 | vma_adjust(vma, new_start, new_end, vma->vm_pgoff, NULL); | |
642 | ||
643 | return 0; | |
1da177e4 LT |
644 | } |
645 | ||
b6a2fea3 OW |
646 | /* |
647 | * Finalizes the stack vm_area_struct. The flags and permissions are updated, | |
648 | * the stack is optionally relocated, and some extra space is added. | |
649 | */ | |
1da177e4 LT |
650 | int setup_arg_pages(struct linux_binprm *bprm, |
651 | unsigned long stack_top, | |
652 | int executable_stack) | |
653 | { | |
b6a2fea3 OW |
654 | unsigned long ret; |
655 | unsigned long stack_shift; | |
1da177e4 | 656 | struct mm_struct *mm = current->mm; |
b6a2fea3 OW |
657 | struct vm_area_struct *vma = bprm->vma; |
658 | struct vm_area_struct *prev = NULL; | |
659 | unsigned long vm_flags; | |
660 | unsigned long stack_base; | |
803bf5ec MN |
661 | unsigned long stack_size; |
662 | unsigned long stack_expand; | |
663 | unsigned long rlim_stack; | |
1da177e4 LT |
664 | |
665 | #ifdef CONFIG_STACK_GROWSUP | |
1da177e4 | 666 | /* Limit stack size to 1GB */ |
d554ed89 | 667 | stack_base = rlimit_max(RLIMIT_STACK); |
1da177e4 LT |
668 | if (stack_base > (1 << 30)) |
669 | stack_base = 1 << 30; | |
1da177e4 | 670 | |
b6a2fea3 OW |
671 | /* Make sure we didn't let the argument array grow too large. */ |
672 | if (vma->vm_end - vma->vm_start > stack_base) | |
673 | return -ENOMEM; | |
1da177e4 | 674 | |
b6a2fea3 | 675 | stack_base = PAGE_ALIGN(stack_top - stack_base); |
1da177e4 | 676 | |
b6a2fea3 OW |
677 | stack_shift = vma->vm_start - stack_base; |
678 | mm->arg_start = bprm->p - stack_shift; | |
679 | bprm->p = vma->vm_end - stack_shift; | |
1da177e4 | 680 | #else |
b6a2fea3 OW |
681 | stack_top = arch_align_stack(stack_top); |
682 | stack_top = PAGE_ALIGN(stack_top); | |
1b528181 RM |
683 | |
684 | if (unlikely(stack_top < mmap_min_addr) || | |
685 | unlikely(vma->vm_end - vma->vm_start >= stack_top - mmap_min_addr)) | |
686 | return -ENOMEM; | |
687 | ||
b6a2fea3 OW |
688 | stack_shift = vma->vm_end - stack_top; |
689 | ||
690 | bprm->p -= stack_shift; | |
1da177e4 | 691 | mm->arg_start = bprm->p; |
1da177e4 LT |
692 | #endif |
693 | ||
1da177e4 | 694 | if (bprm->loader) |
b6a2fea3 OW |
695 | bprm->loader -= stack_shift; |
696 | bprm->exec -= stack_shift; | |
1da177e4 | 697 | |
1da177e4 | 698 | down_write(&mm->mmap_sem); |
96a8e13e | 699 | vm_flags = VM_STACK_FLAGS; |
b6a2fea3 OW |
700 | |
701 | /* | |
702 | * Adjust stack execute permissions; explicitly enable for | |
703 | * EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X and leave alone | |
704 | * (arch default) otherwise. | |
705 | */ | |
706 | if (unlikely(executable_stack == EXSTACK_ENABLE_X)) | |
707 | vm_flags |= VM_EXEC; | |
708 | else if (executable_stack == EXSTACK_DISABLE_X) | |
709 | vm_flags &= ~VM_EXEC; | |
710 | vm_flags |= mm->def_flags; | |
a8bef8ff | 711 | vm_flags |= VM_STACK_INCOMPLETE_SETUP; |
b6a2fea3 OW |
712 | |
713 | ret = mprotect_fixup(vma, &prev, vma->vm_start, vma->vm_end, | |
714 | vm_flags); | |
715 | if (ret) | |
716 | goto out_unlock; | |
717 | BUG_ON(prev != vma); | |
718 | ||
719 | /* Move stack pages down in memory. */ | |
720 | if (stack_shift) { | |
721 | ret = shift_arg_pages(vma, stack_shift); | |
fc63cf23 AB |
722 | if (ret) |
723 | goto out_unlock; | |
1da177e4 LT |
724 | } |
725 | ||
a8bef8ff MG |
726 | /* mprotect_fixup is overkill to remove the temporary stack flags */ |
727 | vma->vm_flags &= ~VM_STACK_INCOMPLETE_SETUP; | |
728 | ||
5ef097dd | 729 | stack_expand = 131072UL; /* randomly 32*4k (or 2*64k) pages */ |
803bf5ec MN |
730 | stack_size = vma->vm_end - vma->vm_start; |
731 | /* | |
732 | * Align this down to a page boundary as expand_stack | |
733 | * will align it up. | |
734 | */ | |
735 | rlim_stack = rlimit(RLIMIT_STACK) & PAGE_MASK; | |
b6a2fea3 | 736 | #ifdef CONFIG_STACK_GROWSUP |
803bf5ec MN |
737 | if (stack_size + stack_expand > rlim_stack) |
738 | stack_base = vma->vm_start + rlim_stack; | |
739 | else | |
740 | stack_base = vma->vm_end + stack_expand; | |
b6a2fea3 | 741 | #else |
803bf5ec MN |
742 | if (stack_size + stack_expand > rlim_stack) |
743 | stack_base = vma->vm_end - rlim_stack; | |
744 | else | |
745 | stack_base = vma->vm_start - stack_expand; | |
b6a2fea3 | 746 | #endif |
3af9e859 | 747 | current->mm->start_stack = bprm->p; |
b6a2fea3 OW |
748 | ret = expand_stack(vma, stack_base); |
749 | if (ret) | |
750 | ret = -EFAULT; | |
751 | ||
752 | out_unlock: | |
1da177e4 | 753 | up_write(&mm->mmap_sem); |
fc63cf23 | 754 | return ret; |
1da177e4 | 755 | } |
1da177e4 LT |
756 | EXPORT_SYMBOL(setup_arg_pages); |
757 | ||
1da177e4 LT |
758 | #endif /* CONFIG_MMU */ |
759 | ||
760 | struct file *open_exec(const char *name) | |
761 | { | |
1da177e4 | 762 | struct file *file; |
e56b6a5d | 763 | int err; |
47c805dc AV |
764 | static const struct open_flags open_exec_flags = { |
765 | .open_flag = O_LARGEFILE | O_RDONLY | __FMODE_EXEC, | |
766 | .acc_mode = MAY_EXEC | MAY_OPEN, | |
767 | .intent = LOOKUP_OPEN | |
768 | }; | |
1da177e4 | 769 | |
47c805dc | 770 | file = do_filp_open(AT_FDCWD, name, &open_exec_flags, LOOKUP_FOLLOW); |
6e8341a1 | 771 | if (IS_ERR(file)) |
e56b6a5d CH |
772 | goto out; |
773 | ||
774 | err = -EACCES; | |
6e8341a1 AV |
775 | if (!S_ISREG(file->f_path.dentry->d_inode->i_mode)) |
776 | goto exit; | |
e56b6a5d | 777 | |
6e8341a1 AV |
778 | if (file->f_path.mnt->mnt_flags & MNT_NOEXEC) |
779 | goto exit; | |
e56b6a5d | 780 | |
2a12a9d7 | 781 | fsnotify_open(file); |
6110e3ab | 782 | |
e56b6a5d | 783 | err = deny_write_access(file); |
6e8341a1 AV |
784 | if (err) |
785 | goto exit; | |
1da177e4 | 786 | |
6e8341a1 | 787 | out: |
e56b6a5d CH |
788 | return file; |
789 | ||
6e8341a1 AV |
790 | exit: |
791 | fput(file); | |
e56b6a5d CH |
792 | return ERR_PTR(err); |
793 | } | |
1da177e4 LT |
794 | EXPORT_SYMBOL(open_exec); |
795 | ||
6777d773 MZ |
796 | int kernel_read(struct file *file, loff_t offset, |
797 | char *addr, unsigned long count) | |
1da177e4 LT |
798 | { |
799 | mm_segment_t old_fs; | |
800 | loff_t pos = offset; | |
801 | int result; | |
802 | ||
803 | old_fs = get_fs(); | |
804 | set_fs(get_ds()); | |
805 | /* The cast to a user pointer is valid due to the set_fs() */ | |
806 | result = vfs_read(file, (void __user *)addr, count, &pos); | |
807 | set_fs(old_fs); | |
808 | return result; | |
809 | } | |
810 | ||
811 | EXPORT_SYMBOL(kernel_read); | |
812 | ||
813 | static int exec_mmap(struct mm_struct *mm) | |
814 | { | |
815 | struct task_struct *tsk; | |
816 | struct mm_struct * old_mm, *active_mm; | |
817 | ||
818 | /* Notify parent that we're no longer interested in the old VM */ | |
819 | tsk = current; | |
820 | old_mm = current->mm; | |
821 | mm_release(tsk, old_mm); | |
822 | ||
823 | if (old_mm) { | |
4fe7efdb | 824 | sync_mm_rss(old_mm); |
1da177e4 LT |
825 | /* |
826 | * Make sure that if there is a core dump in progress | |
827 | * for the old mm, we get out and die instead of going | |
828 | * through with the exec. We must hold mmap_sem around | |
999d9fc1 | 829 | * checking core_state and changing tsk->mm. |
1da177e4 LT |
830 | */ |
831 | down_read(&old_mm->mmap_sem); | |
999d9fc1 | 832 | if (unlikely(old_mm->core_state)) { |
1da177e4 LT |
833 | up_read(&old_mm->mmap_sem); |
834 | return -EINTR; | |
835 | } | |
836 | } | |
837 | task_lock(tsk); | |
838 | active_mm = tsk->active_mm; | |
839 | tsk->mm = mm; | |
840 | tsk->active_mm = mm; | |
841 | activate_mm(active_mm, mm); | |
842 | task_unlock(tsk); | |
843 | arch_pick_mmap_layout(mm); | |
844 | if (old_mm) { | |
845 | up_read(&old_mm->mmap_sem); | |
7dddb12c | 846 | BUG_ON(active_mm != old_mm); |
701085b2 | 847 | setmax_mm_hiwater_rss(&tsk->signal->maxrss, old_mm); |
31a78f23 | 848 | mm_update_next_owner(old_mm); |
1da177e4 LT |
849 | mmput(old_mm); |
850 | return 0; | |
851 | } | |
852 | mmdrop(active_mm); | |
853 | return 0; | |
854 | } | |
855 | ||
856 | /* | |
857 | * This function makes sure the current process has its own signal table, | |
858 | * so that flush_signal_handlers can later reset the handlers without | |
859 | * disturbing other processes. (Other processes might share the signal | |
860 | * table via the CLONE_SIGHAND option to clone().) | |
861 | */ | |
858119e1 | 862 | static int de_thread(struct task_struct *tsk) |
1da177e4 LT |
863 | { |
864 | struct signal_struct *sig = tsk->signal; | |
b2c903b8 | 865 | struct sighand_struct *oldsighand = tsk->sighand; |
1da177e4 | 866 | spinlock_t *lock = &oldsighand->siglock; |
1da177e4 | 867 | |
aafe6c2a | 868 | if (thread_group_empty(tsk)) |
1da177e4 LT |
869 | goto no_thread_group; |
870 | ||
871 | /* | |
872 | * Kill all other threads in the thread group. | |
1da177e4 | 873 | */ |
1da177e4 | 874 | spin_lock_irq(lock); |
ed5d2cac | 875 | if (signal_group_exit(sig)) { |
1da177e4 LT |
876 | /* |
877 | * Another group action in progress, just | |
878 | * return so that the signal is processed. | |
879 | */ | |
880 | spin_unlock_irq(lock); | |
1da177e4 LT |
881 | return -EAGAIN; |
882 | } | |
d344193a | 883 | |
ed5d2cac | 884 | sig->group_exit_task = tsk; |
d344193a ON |
885 | sig->notify_count = zap_other_threads(tsk); |
886 | if (!thread_group_leader(tsk)) | |
887 | sig->notify_count--; | |
1da177e4 | 888 | |
d344193a | 889 | while (sig->notify_count) { |
1da177e4 LT |
890 | __set_current_state(TASK_UNINTERRUPTIBLE); |
891 | spin_unlock_irq(lock); | |
892 | schedule(); | |
893 | spin_lock_irq(lock); | |
894 | } | |
1da177e4 LT |
895 | spin_unlock_irq(lock); |
896 | ||
897 | /* | |
898 | * At this point all other threads have exited, all we have to | |
899 | * do is to wait for the thread group leader to become inactive, | |
900 | * and to assume its PID: | |
901 | */ | |
aafe6c2a | 902 | if (!thread_group_leader(tsk)) { |
8187926b | 903 | struct task_struct *leader = tsk->group_leader; |
6db840fa | 904 | |
2800d8d1 | 905 | sig->notify_count = -1; /* for exit_notify() */ |
6db840fa ON |
906 | for (;;) { |
907 | write_lock_irq(&tasklist_lock); | |
908 | if (likely(leader->exit_state)) | |
909 | break; | |
910 | __set_current_state(TASK_UNINTERRUPTIBLE); | |
911 | write_unlock_irq(&tasklist_lock); | |
912 | schedule(); | |
913 | } | |
1da177e4 | 914 | |
f5e90281 RM |
915 | /* |
916 | * The only record we have of the real-time age of a | |
917 | * process, regardless of execs it's done, is start_time. | |
918 | * All the past CPU time is accumulated in signal_struct | |
919 | * from sister threads now dead. But in this non-leader | |
920 | * exec, nothing survives from the original leader thread, | |
921 | * whose birth marks the true age of this process now. | |
922 | * When we take on its identity by switching to its PID, we | |
923 | * also take its birthdate (always earlier than our own). | |
924 | */ | |
aafe6c2a | 925 | tsk->start_time = leader->start_time; |
f5e90281 | 926 | |
bac0abd6 PE |
927 | BUG_ON(!same_thread_group(leader, tsk)); |
928 | BUG_ON(has_group_leader_pid(tsk)); | |
1da177e4 LT |
929 | /* |
930 | * An exec() starts a new thread group with the | |
931 | * TGID of the previous thread group. Rehash the | |
932 | * two threads with a switched PID, and release | |
933 | * the former thread group leader: | |
934 | */ | |
d73d6529 EB |
935 | |
936 | /* Become a process group leader with the old leader's pid. | |
c18258c6 EB |
937 | * The old leader becomes a thread of the this thread group. |
938 | * Note: The old leader also uses this pid until release_task | |
d73d6529 EB |
939 | * is called. Odd but simple and correct. |
940 | */ | |
aafe6c2a EB |
941 | detach_pid(tsk, PIDTYPE_PID); |
942 | tsk->pid = leader->pid; | |
3743ca05 | 943 | attach_pid(tsk, PIDTYPE_PID, task_pid(leader)); |
aafe6c2a EB |
944 | transfer_pid(leader, tsk, PIDTYPE_PGID); |
945 | transfer_pid(leader, tsk, PIDTYPE_SID); | |
9cd80bbb | 946 | |
aafe6c2a | 947 | list_replace_rcu(&leader->tasks, &tsk->tasks); |
9cd80bbb | 948 | list_replace_init(&leader->sibling, &tsk->sibling); |
1da177e4 | 949 | |
aafe6c2a EB |
950 | tsk->group_leader = tsk; |
951 | leader->group_leader = tsk; | |
de12a787 | 952 | |
aafe6c2a | 953 | tsk->exit_signal = SIGCHLD; |
087806b1 | 954 | leader->exit_signal = -1; |
962b564c ON |
955 | |
956 | BUG_ON(leader->exit_state != EXIT_ZOMBIE); | |
957 | leader->exit_state = EXIT_DEAD; | |
eac1b5e5 ON |
958 | |
959 | /* | |
960 | * We are going to release_task()->ptrace_unlink() silently, | |
961 | * the tracer can sleep in do_wait(). EXIT_DEAD guarantees | |
962 | * the tracer wont't block again waiting for this thread. | |
963 | */ | |
964 | if (unlikely(leader->ptrace)) | |
965 | __wake_up_parent(leader, leader->parent); | |
1da177e4 | 966 | write_unlock_irq(&tasklist_lock); |
8187926b ON |
967 | |
968 | release_task(leader); | |
ed5d2cac | 969 | } |
1da177e4 | 970 | |
6db840fa ON |
971 | sig->group_exit_task = NULL; |
972 | sig->notify_count = 0; | |
1da177e4 LT |
973 | |
974 | no_thread_group: | |
e6368253 ON |
975 | /* we have changed execution domain */ |
976 | tsk->exit_signal = SIGCHLD; | |
977 | ||
1da177e4 | 978 | exit_itimers(sig); |
cbaffba1 | 979 | flush_itimer_signals(); |
329f7dba | 980 | |
b2c903b8 ON |
981 | if (atomic_read(&oldsighand->count) != 1) { |
982 | struct sighand_struct *newsighand; | |
1da177e4 | 983 | /* |
b2c903b8 ON |
984 | * This ->sighand is shared with the CLONE_SIGHAND |
985 | * but not CLONE_THREAD task, switch to the new one. | |
1da177e4 | 986 | */ |
b2c903b8 ON |
987 | newsighand = kmem_cache_alloc(sighand_cachep, GFP_KERNEL); |
988 | if (!newsighand) | |
989 | return -ENOMEM; | |
990 | ||
1da177e4 LT |
991 | atomic_set(&newsighand->count, 1); |
992 | memcpy(newsighand->action, oldsighand->action, | |
993 | sizeof(newsighand->action)); | |
994 | ||
995 | write_lock_irq(&tasklist_lock); | |
996 | spin_lock(&oldsighand->siglock); | |
aafe6c2a | 997 | rcu_assign_pointer(tsk->sighand, newsighand); |
1da177e4 LT |
998 | spin_unlock(&oldsighand->siglock); |
999 | write_unlock_irq(&tasklist_lock); | |
1000 | ||
fba2afaa | 1001 | __cleanup_sighand(oldsighand); |
1da177e4 LT |
1002 | } |
1003 | ||
aafe6c2a | 1004 | BUG_ON(!thread_group_leader(tsk)); |
1da177e4 LT |
1005 | return 0; |
1006 | } | |
0840a90d | 1007 | |
1da177e4 LT |
1008 | /* |
1009 | * These functions flushes out all traces of the currently running executable | |
1010 | * so that a new one can be started | |
1011 | */ | |
858119e1 | 1012 | static void flush_old_files(struct files_struct * files) |
1da177e4 LT |
1013 | { |
1014 | long j = -1; | |
badf1662 | 1015 | struct fdtable *fdt; |
1da177e4 LT |
1016 | |
1017 | spin_lock(&files->file_lock); | |
1018 | for (;;) { | |
1019 | unsigned long set, i; | |
1020 | ||
1021 | j++; | |
8ded2bbc | 1022 | i = j * BITS_PER_LONG; |
badf1662 | 1023 | fdt = files_fdtable(files); |
bbea9f69 | 1024 | if (i >= fdt->max_fds) |
1da177e4 | 1025 | break; |
1fd36adc | 1026 | set = fdt->close_on_exec[j]; |
1da177e4 LT |
1027 | if (!set) |
1028 | continue; | |
1fd36adc | 1029 | fdt->close_on_exec[j] = 0; |
1da177e4 LT |
1030 | spin_unlock(&files->file_lock); |
1031 | for ( ; set ; i++,set >>= 1) { | |
1032 | if (set & 1) { | |
1033 | sys_close(i); | |
1034 | } | |
1035 | } | |
1036 | spin_lock(&files->file_lock); | |
1037 | ||
1038 | } | |
1039 | spin_unlock(&files->file_lock); | |
1040 | } | |
1041 | ||
59714d65 | 1042 | char *get_task_comm(char *buf, struct task_struct *tsk) |
1da177e4 LT |
1043 | { |
1044 | /* buf must be at least sizeof(tsk->comm) in size */ | |
1045 | task_lock(tsk); | |
1046 | strncpy(buf, tsk->comm, sizeof(tsk->comm)); | |
1047 | task_unlock(tsk); | |
59714d65 | 1048 | return buf; |
1da177e4 | 1049 | } |
7d74f492 | 1050 | EXPORT_SYMBOL_GPL(get_task_comm); |
1da177e4 LT |
1051 | |
1052 | void set_task_comm(struct task_struct *tsk, char *buf) | |
1053 | { | |
1054 | task_lock(tsk); | |
4614a696 | 1055 | |
43d2b113 KH |
1056 | trace_task_rename(tsk, buf); |
1057 | ||
4614a696 | 1058 | /* |
1059 | * Threads may access current->comm without holding | |
1060 | * the task lock, so write the string carefully. | |
1061 | * Readers without a lock may see incomplete new | |
1062 | * names but are safe from non-terminating string reads. | |
1063 | */ | |
1064 | memset(tsk->comm, 0, TASK_COMM_LEN); | |
1065 | wmb(); | |
1da177e4 LT |
1066 | strlcpy(tsk->comm, buf, sizeof(tsk->comm)); |
1067 | task_unlock(tsk); | |
cdd6c482 | 1068 | perf_event_comm(tsk); |
1da177e4 LT |
1069 | } |
1070 | ||
96e02d15 HC |
1071 | static void filename_to_taskname(char *tcomm, const char *fn, unsigned int len) |
1072 | { | |
1073 | int i, ch; | |
1074 | ||
1075 | /* Copies the binary name from after last slash */ | |
1076 | for (i = 0; (ch = *(fn++)) != '\0';) { | |
1077 | if (ch == '/') | |
1078 | i = 0; /* overwrite what we wrote */ | |
1079 | else | |
1080 | if (i < len - 1) | |
1081 | tcomm[i++] = ch; | |
1082 | } | |
1083 | tcomm[i] = '\0'; | |
1084 | } | |
1085 | ||
1da177e4 LT |
1086 | int flush_old_exec(struct linux_binprm * bprm) |
1087 | { | |
221af7f8 | 1088 | int retval; |
1da177e4 LT |
1089 | |
1090 | /* | |
1091 | * Make sure we have a private signal table and that | |
1092 | * we are unassociated from the previous thread group. | |
1093 | */ | |
1094 | retval = de_thread(current); | |
1095 | if (retval) | |
1096 | goto out; | |
1097 | ||
925d1c40 MH |
1098 | set_mm_exe_file(bprm->mm, bprm->file); |
1099 | ||
96e02d15 | 1100 | filename_to_taskname(bprm->tcomm, bprm->filename, sizeof(bprm->tcomm)); |
1da177e4 LT |
1101 | /* |
1102 | * Release all of the old mmap stuff | |
1103 | */ | |
3c77f845 | 1104 | acct_arg_size(bprm, 0); |
1da177e4 LT |
1105 | retval = exec_mmap(bprm->mm); |
1106 | if (retval) | |
fd8328be | 1107 | goto out; |
1da177e4 LT |
1108 | |
1109 | bprm->mm = NULL; /* We're using it now */ | |
7ab02af4 | 1110 | |
dac853ae | 1111 | set_fs(USER_DS); |
19e5109f | 1112 | current->flags &= ~(PF_RANDOMIZE | PF_FORKNOEXEC | PF_KTHREAD); |
7ab02af4 LT |
1113 | flush_thread(); |
1114 | current->personality &= ~bprm->per_clear; | |
1115 | ||
221af7f8 LT |
1116 | return 0; |
1117 | ||
1118 | out: | |
1119 | return retval; | |
1120 | } | |
1121 | EXPORT_SYMBOL(flush_old_exec); | |
1122 | ||
1b5d783c AV |
1123 | void would_dump(struct linux_binprm *bprm, struct file *file) |
1124 | { | |
1125 | if (inode_permission(file->f_path.dentry->d_inode, MAY_READ) < 0) | |
1126 | bprm->interp_flags |= BINPRM_FLAGS_ENFORCE_NONDUMP; | |
1127 | } | |
1128 | EXPORT_SYMBOL(would_dump); | |
1129 | ||
221af7f8 LT |
1130 | void setup_new_exec(struct linux_binprm * bprm) |
1131 | { | |
221af7f8 | 1132 | arch_pick_mmap_layout(current->mm); |
1da177e4 LT |
1133 | |
1134 | /* This is the point of no return */ | |
1da177e4 LT |
1135 | current->sas_ss_sp = current->sas_ss_size = 0; |
1136 | ||
8e96e3b7 | 1137 | if (uid_eq(current_euid(), current_uid()) && gid_eq(current_egid(), current_gid())) |
6c5d5238 | 1138 | set_dumpable(current->mm, 1); |
d6e71144 | 1139 | else |
6c5d5238 | 1140 | set_dumpable(current->mm, suid_dumpable); |
d6e71144 | 1141 | |
96e02d15 | 1142 | set_task_comm(current, bprm->tcomm); |
1da177e4 | 1143 | |
0551fbd2 BH |
1144 | /* Set the new mm task size. We have to do that late because it may |
1145 | * depend on TIF_32BIT which is only updated in flush_thread() on | |
1146 | * some architectures like powerpc | |
1147 | */ | |
1148 | current->mm->task_size = TASK_SIZE; | |
1149 | ||
a6f76f23 | 1150 | /* install the new credentials */ |
8e96e3b7 EB |
1151 | if (!uid_eq(bprm->cred->uid, current_euid()) || |
1152 | !gid_eq(bprm->cred->gid, current_egid())) { | |
d2d56c5f | 1153 | current->pdeath_signal = 0; |
1b5d783c AV |
1154 | } else { |
1155 | would_dump(bprm, bprm->file); | |
1156 | if (bprm->interp_flags & BINPRM_FLAGS_ENFORCE_NONDUMP) | |
1157 | set_dumpable(current->mm, suid_dumpable); | |
1da177e4 LT |
1158 | } |
1159 | ||
f65cb45c IM |
1160 | /* |
1161 | * Flush performance counters when crossing a | |
1162 | * security domain: | |
1163 | */ | |
1164 | if (!get_dumpable(current->mm)) | |
cdd6c482 | 1165 | perf_event_exit_task(current); |
f65cb45c | 1166 | |
1da177e4 LT |
1167 | /* An exec changes our domain. We are no longer part of the thread |
1168 | group */ | |
1169 | ||
1170 | current->self_exec_id++; | |
1171 | ||
1172 | flush_signal_handlers(current, 0); | |
1173 | flush_old_files(current->files); | |
1da177e4 | 1174 | } |
221af7f8 | 1175 | EXPORT_SYMBOL(setup_new_exec); |
1da177e4 | 1176 | |
a2a8474c ON |
1177 | /* |
1178 | * Prepare credentials and lock ->cred_guard_mutex. | |
1179 | * install_exec_creds() commits the new creds and drops the lock. | |
1180 | * Or, if exec fails before, free_bprm() should release ->cred and | |
1181 | * and unlock. | |
1182 | */ | |
1183 | int prepare_bprm_creds(struct linux_binprm *bprm) | |
1184 | { | |
9b1bf12d | 1185 | if (mutex_lock_interruptible(¤t->signal->cred_guard_mutex)) |
a2a8474c ON |
1186 | return -ERESTARTNOINTR; |
1187 | ||
1188 | bprm->cred = prepare_exec_creds(); | |
1189 | if (likely(bprm->cred)) | |
1190 | return 0; | |
1191 | ||
9b1bf12d | 1192 | mutex_unlock(¤t->signal->cred_guard_mutex); |
a2a8474c ON |
1193 | return -ENOMEM; |
1194 | } | |
1195 | ||
1196 | void free_bprm(struct linux_binprm *bprm) | |
1197 | { | |
1198 | free_arg_pages(bprm); | |
1199 | if (bprm->cred) { | |
9b1bf12d | 1200 | mutex_unlock(¤t->signal->cred_guard_mutex); |
a2a8474c ON |
1201 | abort_creds(bprm->cred); |
1202 | } | |
1203 | kfree(bprm); | |
1204 | } | |
1205 | ||
a6f76f23 DH |
1206 | /* |
1207 | * install the new credentials for this executable | |
1208 | */ | |
1209 | void install_exec_creds(struct linux_binprm *bprm) | |
1210 | { | |
1211 | security_bprm_committing_creds(bprm); | |
1212 | ||
1213 | commit_creds(bprm->cred); | |
1214 | bprm->cred = NULL; | |
a2a8474c ON |
1215 | /* |
1216 | * cred_guard_mutex must be held at least to this point to prevent | |
a6f76f23 | 1217 | * ptrace_attach() from altering our determination of the task's |
a2a8474c ON |
1218 | * credentials; any time after this it may be unlocked. |
1219 | */ | |
a6f76f23 | 1220 | security_bprm_committed_creds(bprm); |
9b1bf12d | 1221 | mutex_unlock(¤t->signal->cred_guard_mutex); |
a6f76f23 DH |
1222 | } |
1223 | EXPORT_SYMBOL(install_exec_creds); | |
1224 | ||
1225 | /* | |
1226 | * determine how safe it is to execute the proposed program | |
9b1bf12d | 1227 | * - the caller must hold ->cred_guard_mutex to protect against |
a6f76f23 DH |
1228 | * PTRACE_ATTACH |
1229 | */ | |
f47ec3f2 | 1230 | static int check_unsafe_exec(struct linux_binprm *bprm) |
a6f76f23 | 1231 | { |
0bf2f3ae | 1232 | struct task_struct *p = current, *t; |
f1191b50 | 1233 | unsigned n_fs; |
498052bb | 1234 | int res = 0; |
a6f76f23 | 1235 | |
4b9d33e6 TH |
1236 | if (p->ptrace) { |
1237 | if (p->ptrace & PT_PTRACE_CAP) | |
1238 | bprm->unsafe |= LSM_UNSAFE_PTRACE_CAP; | |
1239 | else | |
1240 | bprm->unsafe |= LSM_UNSAFE_PTRACE; | |
1241 | } | |
a6f76f23 | 1242 | |
259e5e6c AL |
1243 | /* |
1244 | * This isn't strictly necessary, but it makes it harder for LSMs to | |
1245 | * mess up. | |
1246 | */ | |
1247 | if (current->no_new_privs) | |
1248 | bprm->unsafe |= LSM_UNSAFE_NO_NEW_PRIVS; | |
1249 | ||
0bf2f3ae | 1250 | n_fs = 1; |
2a4419b5 | 1251 | spin_lock(&p->fs->lock); |
437f7fdb | 1252 | rcu_read_lock(); |
0bf2f3ae DH |
1253 | for (t = next_thread(p); t != p; t = next_thread(t)) { |
1254 | if (t->fs == p->fs) | |
1255 | n_fs++; | |
0bf2f3ae | 1256 | } |
437f7fdb | 1257 | rcu_read_unlock(); |
0bf2f3ae | 1258 | |
f1191b50 | 1259 | if (p->fs->users > n_fs) { |
a6f76f23 | 1260 | bprm->unsafe |= LSM_UNSAFE_SHARE; |
498052bb | 1261 | } else { |
8c652f96 ON |
1262 | res = -EAGAIN; |
1263 | if (!p->fs->in_exec) { | |
1264 | p->fs->in_exec = 1; | |
1265 | res = 1; | |
1266 | } | |
498052bb | 1267 | } |
2a4419b5 | 1268 | spin_unlock(&p->fs->lock); |
498052bb AV |
1269 | |
1270 | return res; | |
a6f76f23 DH |
1271 | } |
1272 | ||
1da177e4 LT |
1273 | /* |
1274 | * Fill the binprm structure from the inode. | |
1275 | * Check permissions, then read the first 128 (BINPRM_BUF_SIZE) bytes | |
a6f76f23 DH |
1276 | * |
1277 | * This may be called multiple times for binary chains (scripts for example). | |
1da177e4 LT |
1278 | */ |
1279 | int prepare_binprm(struct linux_binprm *bprm) | |
1280 | { | |
a6f76f23 | 1281 | umode_t mode; |
0f7fc9e4 | 1282 | struct inode * inode = bprm->file->f_path.dentry->d_inode; |
1da177e4 LT |
1283 | int retval; |
1284 | ||
1285 | mode = inode->i_mode; | |
1da177e4 LT |
1286 | if (bprm->file->f_op == NULL) |
1287 | return -EACCES; | |
1288 | ||
a6f76f23 DH |
1289 | /* clear any previous set[ug]id data from a previous binary */ |
1290 | bprm->cred->euid = current_euid(); | |
1291 | bprm->cred->egid = current_egid(); | |
1da177e4 | 1292 | |
259e5e6c AL |
1293 | if (!(bprm->file->f_path.mnt->mnt_flags & MNT_NOSUID) && |
1294 | !current->no_new_privs) { | |
1da177e4 LT |
1295 | /* Set-uid? */ |
1296 | if (mode & S_ISUID) { | |
9e4a36ec EB |
1297 | if (!kuid_has_mapping(bprm->cred->user_ns, inode->i_uid)) |
1298 | return -EPERM; | |
a6f76f23 DH |
1299 | bprm->per_clear |= PER_CLEAR_ON_SETID; |
1300 | bprm->cred->euid = inode->i_uid; | |
9e4a36ec | 1301 | |
1da177e4 LT |
1302 | } |
1303 | ||
1304 | /* Set-gid? */ | |
1305 | /* | |
1306 | * If setgid is set but no group execute bit then this | |
1307 | * is a candidate for mandatory locking, not a setgid | |
1308 | * executable. | |
1309 | */ | |
1310 | if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) { | |
9e4a36ec EB |
1311 | if (!kgid_has_mapping(bprm->cred->user_ns, inode->i_gid)) |
1312 | return -EPERM; | |
a6f76f23 DH |
1313 | bprm->per_clear |= PER_CLEAR_ON_SETID; |
1314 | bprm->cred->egid = inode->i_gid; | |
1da177e4 LT |
1315 | } |
1316 | } | |
1317 | ||
1318 | /* fill in binprm security blob */ | |
a6f76f23 | 1319 | retval = security_bprm_set_creds(bprm); |
1da177e4 LT |
1320 | if (retval) |
1321 | return retval; | |
a6f76f23 | 1322 | bprm->cred_prepared = 1; |
1da177e4 | 1323 | |
a6f76f23 DH |
1324 | memset(bprm->buf, 0, BINPRM_BUF_SIZE); |
1325 | return kernel_read(bprm->file, 0, bprm->buf, BINPRM_BUF_SIZE); | |
1da177e4 LT |
1326 | } |
1327 | ||
1328 | EXPORT_SYMBOL(prepare_binprm); | |
1329 | ||
4fc75ff4 NP |
1330 | /* |
1331 | * Arguments are '\0' separated strings found at the location bprm->p | |
1332 | * points to; chop off the first by relocating brpm->p to right after | |
1333 | * the first '\0' encountered. | |
1334 | */ | |
b6a2fea3 | 1335 | int remove_arg_zero(struct linux_binprm *bprm) |
1da177e4 | 1336 | { |
b6a2fea3 OW |
1337 | int ret = 0; |
1338 | unsigned long offset; | |
1339 | char *kaddr; | |
1340 | struct page *page; | |
4fc75ff4 | 1341 | |
b6a2fea3 OW |
1342 | if (!bprm->argc) |
1343 | return 0; | |
1da177e4 | 1344 | |
b6a2fea3 OW |
1345 | do { |
1346 | offset = bprm->p & ~PAGE_MASK; | |
1347 | page = get_arg_page(bprm, bprm->p, 0); | |
1348 | if (!page) { | |
1349 | ret = -EFAULT; | |
1350 | goto out; | |
1351 | } | |
e8e3c3d6 | 1352 | kaddr = kmap_atomic(page); |
4fc75ff4 | 1353 | |
b6a2fea3 OW |
1354 | for (; offset < PAGE_SIZE && kaddr[offset]; |
1355 | offset++, bprm->p++) | |
1356 | ; | |
4fc75ff4 | 1357 | |
e8e3c3d6 | 1358 | kunmap_atomic(kaddr); |
b6a2fea3 | 1359 | put_arg_page(page); |
4fc75ff4 | 1360 | |
b6a2fea3 OW |
1361 | if (offset == PAGE_SIZE) |
1362 | free_arg_page(bprm, (bprm->p >> PAGE_SHIFT) - 1); | |
1363 | } while (offset == PAGE_SIZE); | |
4fc75ff4 | 1364 | |
b6a2fea3 OW |
1365 | bprm->p++; |
1366 | bprm->argc--; | |
1367 | ret = 0; | |
4fc75ff4 | 1368 | |
b6a2fea3 OW |
1369 | out: |
1370 | return ret; | |
1da177e4 | 1371 | } |
1da177e4 LT |
1372 | EXPORT_SYMBOL(remove_arg_zero); |
1373 | ||
1374 | /* | |
1375 | * cycle the list of binary formats handler, until one recognizes the image | |
1376 | */ | |
1377 | int search_binary_handler(struct linux_binprm *bprm,struct pt_regs *regs) | |
1378 | { | |
85f33466 | 1379 | unsigned int depth = bprm->recursion_depth; |
1da177e4 LT |
1380 | int try,retval; |
1381 | struct linux_binfmt *fmt; | |
6308191f | 1382 | pid_t old_pid, old_vpid; |
1da177e4 | 1383 | |
1da177e4 LT |
1384 | retval = security_bprm_check(bprm); |
1385 | if (retval) | |
1386 | return retval; | |
1387 | ||
473ae30b AV |
1388 | retval = audit_bprm(bprm); |
1389 | if (retval) | |
1390 | return retval; | |
1391 | ||
bb188d7e | 1392 | /* Need to fetch pid before load_binary changes it */ |
6308191f | 1393 | old_pid = current->pid; |
bb188d7e | 1394 | rcu_read_lock(); |
6308191f | 1395 | old_vpid = task_pid_nr_ns(current, task_active_pid_ns(current->parent)); |
bb188d7e DV |
1396 | rcu_read_unlock(); |
1397 | ||
1da177e4 LT |
1398 | retval = -ENOENT; |
1399 | for (try=0; try<2; try++) { | |
1400 | read_lock(&binfmt_lock); | |
e4dc1b14 | 1401 | list_for_each_entry(fmt, &formats, lh) { |
1da177e4 LT |
1402 | int (*fn)(struct linux_binprm *, struct pt_regs *) = fmt->load_binary; |
1403 | if (!fn) | |
1404 | continue; | |
1405 | if (!try_module_get(fmt->module)) | |
1406 | continue; | |
1407 | read_unlock(&binfmt_lock); | |
1408 | retval = fn(bprm, regs); | |
85f33466 RM |
1409 | /* |
1410 | * Restore the depth counter to its starting value | |
1411 | * in this call, so we don't have to rely on every | |
1412 | * load_binary function to restore it on return. | |
1413 | */ | |
1414 | bprm->recursion_depth = depth; | |
1da177e4 | 1415 | if (retval >= 0) { |
4ff16c25 DS |
1416 | if (depth == 0) { |
1417 | trace_sched_process_exec(current, old_pid, bprm); | |
6308191f | 1418 | ptrace_event(PTRACE_EVENT_EXEC, old_vpid); |
4ff16c25 | 1419 | } |
1da177e4 LT |
1420 | put_binfmt(fmt); |
1421 | allow_write_access(bprm->file); | |
1422 | if (bprm->file) | |
1423 | fput(bprm->file); | |
1424 | bprm->file = NULL; | |
1425 | current->did_exec = 1; | |
9f46080c | 1426 | proc_exec_connector(current); |
1da177e4 LT |
1427 | return retval; |
1428 | } | |
1429 | read_lock(&binfmt_lock); | |
1430 | put_binfmt(fmt); | |
1431 | if (retval != -ENOEXEC || bprm->mm == NULL) | |
1432 | break; | |
1433 | if (!bprm->file) { | |
1434 | read_unlock(&binfmt_lock); | |
1435 | return retval; | |
1436 | } | |
1437 | } | |
1438 | read_unlock(&binfmt_lock); | |
b4edf8bd | 1439 | #ifdef CONFIG_MODULES |
1da177e4 LT |
1440 | if (retval != -ENOEXEC || bprm->mm == NULL) { |
1441 | break; | |
5f4123be | 1442 | } else { |
1da177e4 LT |
1443 | #define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e)) |
1444 | if (printable(bprm->buf[0]) && | |
1445 | printable(bprm->buf[1]) && | |
1446 | printable(bprm->buf[2]) && | |
1447 | printable(bprm->buf[3])) | |
1448 | break; /* -ENOEXEC */ | |
91219352 TH |
1449 | if (try) |
1450 | break; /* -ENOEXEC */ | |
1da177e4 | 1451 | request_module("binfmt-%04x", *(unsigned short *)(&bprm->buf[2])); |
1da177e4 | 1452 | } |
b4edf8bd TH |
1453 | #else |
1454 | break; | |
1455 | #endif | |
1da177e4 LT |
1456 | } |
1457 | return retval; | |
1458 | } | |
1459 | ||
1460 | EXPORT_SYMBOL(search_binary_handler); | |
1461 | ||
1462 | /* | |
1463 | * sys_execve() executes a new program. | |
1464 | */ | |
ba2d0162 ON |
1465 | static int do_execve_common(const char *filename, |
1466 | struct user_arg_ptr argv, | |
1467 | struct user_arg_ptr envp, | |
1468 | struct pt_regs *regs) | |
1da177e4 LT |
1469 | { |
1470 | struct linux_binprm *bprm; | |
1471 | struct file *file; | |
3b125388 | 1472 | struct files_struct *displaced; |
8c652f96 | 1473 | bool clear_in_exec; |
1da177e4 | 1474 | int retval; |
72fa5997 VK |
1475 | const struct cred *cred = current_cred(); |
1476 | ||
1477 | /* | |
1478 | * We move the actual failure in case of RLIMIT_NPROC excess from | |
1479 | * set*uid() to execve() because too many poorly written programs | |
1480 | * don't check setuid() return code. Here we additionally recheck | |
1481 | * whether NPROC limit is still exceeded. | |
1482 | */ | |
1483 | if ((current->flags & PF_NPROC_EXCEEDED) && | |
1484 | atomic_read(&cred->user->processes) > rlimit(RLIMIT_NPROC)) { | |
1485 | retval = -EAGAIN; | |
1486 | goto out_ret; | |
1487 | } | |
1488 | ||
1489 | /* We're below the limit (still or again), so we don't want to make | |
1490 | * further execve() calls fail. */ | |
1491 | current->flags &= ~PF_NPROC_EXCEEDED; | |
1da177e4 | 1492 | |
3b125388 | 1493 | retval = unshare_files(&displaced); |
fd8328be AV |
1494 | if (retval) |
1495 | goto out_ret; | |
1496 | ||
1da177e4 | 1497 | retval = -ENOMEM; |
11b0b5ab | 1498 | bprm = kzalloc(sizeof(*bprm), GFP_KERNEL); |
1da177e4 | 1499 | if (!bprm) |
fd8328be | 1500 | goto out_files; |
1da177e4 | 1501 | |
a2a8474c ON |
1502 | retval = prepare_bprm_creds(bprm); |
1503 | if (retval) | |
a6f76f23 | 1504 | goto out_free; |
498052bb AV |
1505 | |
1506 | retval = check_unsafe_exec(bprm); | |
8c652f96 | 1507 | if (retval < 0) |
a2a8474c | 1508 | goto out_free; |
8c652f96 | 1509 | clear_in_exec = retval; |
a2a8474c | 1510 | current->in_execve = 1; |
a6f76f23 | 1511 | |
1da177e4 LT |
1512 | file = open_exec(filename); |
1513 | retval = PTR_ERR(file); | |
1514 | if (IS_ERR(file)) | |
498052bb | 1515 | goto out_unmark; |
1da177e4 LT |
1516 | |
1517 | sched_exec(); | |
1518 | ||
1da177e4 LT |
1519 | bprm->file = file; |
1520 | bprm->filename = filename; | |
1521 | bprm->interp = filename; | |
1da177e4 | 1522 | |
b6a2fea3 OW |
1523 | retval = bprm_mm_init(bprm); |
1524 | if (retval) | |
1525 | goto out_file; | |
1da177e4 | 1526 | |
b6a2fea3 | 1527 | bprm->argc = count(argv, MAX_ARG_STRINGS); |
1da177e4 | 1528 | if ((retval = bprm->argc) < 0) |
a6f76f23 | 1529 | goto out; |
1da177e4 | 1530 | |
b6a2fea3 | 1531 | bprm->envc = count(envp, MAX_ARG_STRINGS); |
1da177e4 | 1532 | if ((retval = bprm->envc) < 0) |
1da177e4 LT |
1533 | goto out; |
1534 | ||
1535 | retval = prepare_binprm(bprm); | |
1536 | if (retval < 0) | |
1537 | goto out; | |
1538 | ||
1539 | retval = copy_strings_kernel(1, &bprm->filename, bprm); | |
1540 | if (retval < 0) | |
1541 | goto out; | |
1542 | ||
1543 | bprm->exec = bprm->p; | |
1544 | retval = copy_strings(bprm->envc, envp, bprm); | |
1545 | if (retval < 0) | |
1546 | goto out; | |
1547 | ||
1548 | retval = copy_strings(bprm->argc, argv, bprm); | |
1549 | if (retval < 0) | |
1550 | goto out; | |
1551 | ||
1552 | retval = search_binary_handler(bprm,regs); | |
a6f76f23 DH |
1553 | if (retval < 0) |
1554 | goto out; | |
1da177e4 | 1555 | |
a6f76f23 | 1556 | /* execve succeeded */ |
498052bb | 1557 | current->fs->in_exec = 0; |
f9ce1f1c | 1558 | current->in_execve = 0; |
a6f76f23 DH |
1559 | acct_update_integrals(current); |
1560 | free_bprm(bprm); | |
1561 | if (displaced) | |
1562 | put_files_struct(displaced); | |
1563 | return retval; | |
1da177e4 | 1564 | |
a6f76f23 | 1565 | out: |
3c77f845 ON |
1566 | if (bprm->mm) { |
1567 | acct_arg_size(bprm, 0); | |
1568 | mmput(bprm->mm); | |
1569 | } | |
1da177e4 LT |
1570 | |
1571 | out_file: | |
1572 | if (bprm->file) { | |
1573 | allow_write_access(bprm->file); | |
1574 | fput(bprm->file); | |
1575 | } | |
a6f76f23 | 1576 | |
498052bb | 1577 | out_unmark: |
8c652f96 ON |
1578 | if (clear_in_exec) |
1579 | current->fs->in_exec = 0; | |
f9ce1f1c | 1580 | current->in_execve = 0; |
a6f76f23 DH |
1581 | |
1582 | out_free: | |
08a6fac1 | 1583 | free_bprm(bprm); |
1da177e4 | 1584 | |
fd8328be | 1585 | out_files: |
3b125388 AV |
1586 | if (displaced) |
1587 | reset_files_struct(displaced); | |
1da177e4 LT |
1588 | out_ret: |
1589 | return retval; | |
1590 | } | |
1591 | ||
ba2d0162 ON |
1592 | int do_execve(const char *filename, |
1593 | const char __user *const __user *__argv, | |
1594 | const char __user *const __user *__envp, | |
1595 | struct pt_regs *regs) | |
1596 | { | |
0e028465 ON |
1597 | struct user_arg_ptr argv = { .ptr.native = __argv }; |
1598 | struct user_arg_ptr envp = { .ptr.native = __envp }; | |
1599 | return do_execve_common(filename, argv, envp, regs); | |
1600 | } | |
1601 | ||
1602 | #ifdef CONFIG_COMPAT | |
1603 | int compat_do_execve(char *filename, | |
1604 | compat_uptr_t __user *__argv, | |
1605 | compat_uptr_t __user *__envp, | |
1606 | struct pt_regs *regs) | |
1607 | { | |
1608 | struct user_arg_ptr argv = { | |
1609 | .is_compat = true, | |
1610 | .ptr.compat = __argv, | |
1611 | }; | |
1612 | struct user_arg_ptr envp = { | |
1613 | .is_compat = true, | |
1614 | .ptr.compat = __envp, | |
1615 | }; | |
ba2d0162 ON |
1616 | return do_execve_common(filename, argv, envp, regs); |
1617 | } | |
0e028465 | 1618 | #endif |
ba2d0162 | 1619 | |
964ee7df | 1620 | void set_binfmt(struct linux_binfmt *new) |
1da177e4 | 1621 | { |
801460d0 HS |
1622 | struct mm_struct *mm = current->mm; |
1623 | ||
1624 | if (mm->binfmt) | |
1625 | module_put(mm->binfmt->module); | |
1da177e4 | 1626 | |
801460d0 | 1627 | mm->binfmt = new; |
964ee7df ON |
1628 | if (new) |
1629 | __module_get(new->module); | |
1da177e4 LT |
1630 | } |
1631 | ||
1632 | EXPORT_SYMBOL(set_binfmt); | |
1633 | ||
1b0d300b XF |
1634 | static int expand_corename(struct core_name *cn) |
1635 | { | |
1636 | char *old_corename = cn->corename; | |
1637 | ||
1638 | cn->size = CORENAME_MAX_SIZE * atomic_inc_return(&call_count); | |
1639 | cn->corename = krealloc(old_corename, cn->size, GFP_KERNEL); | |
1640 | ||
1641 | if (!cn->corename) { | |
1642 | kfree(old_corename); | |
1643 | return -ENOMEM; | |
1644 | } | |
1645 | ||
1646 | return 0; | |
1647 | } | |
1648 | ||
1649 | static int cn_printf(struct core_name *cn, const char *fmt, ...) | |
1650 | { | |
1651 | char *cur; | |
1652 | int need; | |
1653 | int ret; | |
1654 | va_list arg; | |
1655 | ||
1656 | va_start(arg, fmt); | |
1657 | need = vsnprintf(NULL, 0, fmt, arg); | |
1658 | va_end(arg); | |
1659 | ||
1660 | if (likely(need < cn->size - cn->used - 1)) | |
1661 | goto out_printf; | |
1662 | ||
1663 | ret = expand_corename(cn); | |
1664 | if (ret) | |
1665 | goto expand_fail; | |
1666 | ||
1667 | out_printf: | |
1668 | cur = cn->corename + cn->used; | |
1669 | va_start(arg, fmt); | |
1670 | vsnprintf(cur, need + 1, fmt, arg); | |
1671 | va_end(arg); | |
1672 | cn->used += need; | |
1673 | return 0; | |
1674 | ||
1675 | expand_fail: | |
1676 | return ret; | |
1677 | } | |
1678 | ||
2c563731 JS |
1679 | static void cn_escape(char *str) |
1680 | { | |
1681 | for (; *str; str++) | |
1682 | if (*str == '/') | |
1683 | *str = '!'; | |
1684 | } | |
1685 | ||
57cc083a JS |
1686 | static int cn_print_exe_file(struct core_name *cn) |
1687 | { | |
1688 | struct file *exe_file; | |
2c563731 | 1689 | char *pathbuf, *path; |
57cc083a JS |
1690 | int ret; |
1691 | ||
1692 | exe_file = get_mm_exe_file(current->mm); | |
2c563731 JS |
1693 | if (!exe_file) { |
1694 | char *commstart = cn->corename + cn->used; | |
1695 | ret = cn_printf(cn, "%s (path unknown)", current->comm); | |
1696 | cn_escape(commstart); | |
1697 | return ret; | |
1698 | } | |
57cc083a JS |
1699 | |
1700 | pathbuf = kmalloc(PATH_MAX, GFP_TEMPORARY); | |
1701 | if (!pathbuf) { | |
1702 | ret = -ENOMEM; | |
1703 | goto put_exe_file; | |
1704 | } | |
1705 | ||
1706 | path = d_path(&exe_file->f_path, pathbuf, PATH_MAX); | |
1707 | if (IS_ERR(path)) { | |
1708 | ret = PTR_ERR(path); | |
1709 | goto free_buf; | |
1710 | } | |
1711 | ||
2c563731 | 1712 | cn_escape(path); |
57cc083a JS |
1713 | |
1714 | ret = cn_printf(cn, "%s", path); | |
1715 | ||
1716 | free_buf: | |
1717 | kfree(pathbuf); | |
1718 | put_exe_file: | |
1719 | fput(exe_file); | |
1720 | return ret; | |
1721 | } | |
1722 | ||
1da177e4 LT |
1723 | /* format_corename will inspect the pattern parameter, and output a |
1724 | * name into corename, which must have space for at least | |
1725 | * CORENAME_MAX_SIZE bytes plus one byte for the zero terminator. | |
1726 | */ | |
1b0d300b | 1727 | static int format_corename(struct core_name *cn, long signr) |
1da177e4 | 1728 | { |
86a264ab | 1729 | const struct cred *cred = current_cred(); |
565b9b14 ON |
1730 | const char *pat_ptr = core_pattern; |
1731 | int ispipe = (*pat_ptr == '|'); | |
1da177e4 | 1732 | int pid_in_pattern = 0; |
1b0d300b XF |
1733 | int err = 0; |
1734 | ||
1735 | cn->size = CORENAME_MAX_SIZE * atomic_read(&call_count); | |
1736 | cn->corename = kmalloc(cn->size, GFP_KERNEL); | |
1737 | cn->used = 0; | |
1738 | ||
1739 | if (!cn->corename) | |
1740 | return -ENOMEM; | |
1da177e4 LT |
1741 | |
1742 | /* Repeat as long as we have more pattern to process and more output | |
1743 | space */ | |
1744 | while (*pat_ptr) { | |
1745 | if (*pat_ptr != '%') { | |
1b0d300b | 1746 | if (*pat_ptr == 0) |
1da177e4 | 1747 | goto out; |
1b0d300b | 1748 | err = cn_printf(cn, "%c", *pat_ptr++); |
1da177e4 LT |
1749 | } else { |
1750 | switch (*++pat_ptr) { | |
1b0d300b | 1751 | /* single % at the end, drop that */ |
1da177e4 LT |
1752 | case 0: |
1753 | goto out; | |
1754 | /* Double percent, output one percent */ | |
1755 | case '%': | |
1b0d300b | 1756 | err = cn_printf(cn, "%c", '%'); |
1da177e4 LT |
1757 | break; |
1758 | /* pid */ | |
1759 | case 'p': | |
1760 | pid_in_pattern = 1; | |
1b0d300b XF |
1761 | err = cn_printf(cn, "%d", |
1762 | task_tgid_vnr(current)); | |
1da177e4 LT |
1763 | break; |
1764 | /* uid */ | |
1765 | case 'u': | |
1b0d300b | 1766 | err = cn_printf(cn, "%d", cred->uid); |
1da177e4 LT |
1767 | break; |
1768 | /* gid */ | |
1769 | case 'g': | |
1b0d300b | 1770 | err = cn_printf(cn, "%d", cred->gid); |
1da177e4 LT |
1771 | break; |
1772 | /* signal that caused the coredump */ | |
1773 | case 's': | |
1b0d300b | 1774 | err = cn_printf(cn, "%ld", signr); |
1da177e4 LT |
1775 | break; |
1776 | /* UNIX time of coredump */ | |
1777 | case 't': { | |
1778 | struct timeval tv; | |
1779 | do_gettimeofday(&tv); | |
1b0d300b | 1780 | err = cn_printf(cn, "%lu", tv.tv_sec); |
1da177e4 LT |
1781 | break; |
1782 | } | |
1783 | /* hostname */ | |
2c563731 JS |
1784 | case 'h': { |
1785 | char *namestart = cn->corename + cn->used; | |
1da177e4 | 1786 | down_read(&uts_sem); |
1b0d300b XF |
1787 | err = cn_printf(cn, "%s", |
1788 | utsname()->nodename); | |
1da177e4 | 1789 | up_read(&uts_sem); |
2c563731 | 1790 | cn_escape(namestart); |
1da177e4 | 1791 | break; |
2c563731 | 1792 | } |
1da177e4 | 1793 | /* executable */ |
2c563731 JS |
1794 | case 'e': { |
1795 | char *commstart = cn->corename + cn->used; | |
1b0d300b | 1796 | err = cn_printf(cn, "%s", current->comm); |
2c563731 | 1797 | cn_escape(commstart); |
1da177e4 | 1798 | break; |
2c563731 | 1799 | } |
57cc083a JS |
1800 | case 'E': |
1801 | err = cn_print_exe_file(cn); | |
1802 | break; | |
74aadce9 NH |
1803 | /* core limit size */ |
1804 | case 'c': | |
1b0d300b XF |
1805 | err = cn_printf(cn, "%lu", |
1806 | rlimit(RLIMIT_CORE)); | |
74aadce9 | 1807 | break; |
1da177e4 LT |
1808 | default: |
1809 | break; | |
1810 | } | |
1811 | ++pat_ptr; | |
1812 | } | |
1b0d300b XF |
1813 | |
1814 | if (err) | |
1815 | return err; | |
1da177e4 | 1816 | } |
1b0d300b | 1817 | |
1da177e4 LT |
1818 | /* Backward compatibility with core_uses_pid: |
1819 | * | |
1820 | * If core_pattern does not include a %p (as is the default) | |
1821 | * and core_uses_pid is set, then .%pid will be appended to | |
c4bbafda | 1822 | * the filename. Do not do this for piped commands. */ |
6409324b | 1823 | if (!ispipe && !pid_in_pattern && core_uses_pid) { |
1b0d300b XF |
1824 | err = cn_printf(cn, ".%d", task_tgid_vnr(current)); |
1825 | if (err) | |
1826 | return err; | |
1da177e4 | 1827 | } |
c4bbafda | 1828 | out: |
c4bbafda | 1829 | return ispipe; |
1da177e4 LT |
1830 | } |
1831 | ||
5c99cbf4 | 1832 | static int zap_process(struct task_struct *start, int exit_code) |
aceecc04 ON |
1833 | { |
1834 | struct task_struct *t; | |
8cd9c249 | 1835 | int nr = 0; |
281de339 | 1836 | |
d5f70c00 | 1837 | start->signal->flags = SIGNAL_GROUP_EXIT; |
5c99cbf4 | 1838 | start->signal->group_exit_code = exit_code; |
d5f70c00 | 1839 | start->signal->group_stop_count = 0; |
aceecc04 ON |
1840 | |
1841 | t = start; | |
1842 | do { | |
6dfca329 | 1843 | task_clear_jobctl_pending(t, JOBCTL_PENDING_MASK); |
aceecc04 | 1844 | if (t != current && t->mm) { |
281de339 ON |
1845 | sigaddset(&t->pending.signal, SIGKILL); |
1846 | signal_wake_up(t, 1); | |
8cd9c249 | 1847 | nr++; |
aceecc04 | 1848 | } |
e4901f92 | 1849 | } while_each_thread(start, t); |
8cd9c249 ON |
1850 | |
1851 | return nr; | |
aceecc04 ON |
1852 | } |
1853 | ||
dcf560c5 | 1854 | static inline int zap_threads(struct task_struct *tsk, struct mm_struct *mm, |
8cd9c249 | 1855 | struct core_state *core_state, int exit_code) |
1da177e4 LT |
1856 | { |
1857 | struct task_struct *g, *p; | |
5debfa6d | 1858 | unsigned long flags; |
8cd9c249 | 1859 | int nr = -EAGAIN; |
dcf560c5 ON |
1860 | |
1861 | spin_lock_irq(&tsk->sighand->siglock); | |
ed5d2cac | 1862 | if (!signal_group_exit(tsk->signal)) { |
8cd9c249 | 1863 | mm->core_state = core_state; |
5c99cbf4 | 1864 | nr = zap_process(tsk, exit_code); |
1da177e4 | 1865 | } |
dcf560c5 | 1866 | spin_unlock_irq(&tsk->sighand->siglock); |
8cd9c249 ON |
1867 | if (unlikely(nr < 0)) |
1868 | return nr; | |
1da177e4 | 1869 | |
8cd9c249 | 1870 | if (atomic_read(&mm->mm_users) == nr + 1) |
5debfa6d | 1871 | goto done; |
e4901f92 ON |
1872 | /* |
1873 | * We should find and kill all tasks which use this mm, and we should | |
999d9fc1 | 1874 | * count them correctly into ->nr_threads. We don't take tasklist |
e4901f92 ON |
1875 | * lock, but this is safe wrt: |
1876 | * | |
1877 | * fork: | |
1878 | * None of sub-threads can fork after zap_process(leader). All | |
1879 | * processes which were created before this point should be | |
1880 | * visible to zap_threads() because copy_process() adds the new | |
1881 | * process to the tail of init_task.tasks list, and lock/unlock | |
1882 | * of ->siglock provides a memory barrier. | |
1883 | * | |
1884 | * do_exit: | |
1885 | * The caller holds mm->mmap_sem. This means that the task which | |
1886 | * uses this mm can't pass exit_mm(), so it can't exit or clear | |
1887 | * its ->mm. | |
1888 | * | |
1889 | * de_thread: | |
1890 | * It does list_replace_rcu(&leader->tasks, ¤t->tasks), | |
1891 | * we must see either old or new leader, this does not matter. | |
1892 | * However, it can change p->sighand, so lock_task_sighand(p) | |
1893 | * must be used. Since p->mm != NULL and we hold ->mmap_sem | |
1894 | * it can't fail. | |
1895 | * | |
1896 | * Note also that "g" can be the old leader with ->mm == NULL | |
1897 | * and already unhashed and thus removed from ->thread_group. | |
1898 | * This is OK, __unhash_process()->list_del_rcu() does not | |
1899 | * clear the ->next pointer, we will find the new leader via | |
1900 | * next_thread(). | |
1901 | */ | |
7b1c6154 | 1902 | rcu_read_lock(); |
aceecc04 | 1903 | for_each_process(g) { |
5debfa6d ON |
1904 | if (g == tsk->group_leader) |
1905 | continue; | |
15b9f360 ON |
1906 | if (g->flags & PF_KTHREAD) |
1907 | continue; | |
aceecc04 ON |
1908 | p = g; |
1909 | do { | |
1910 | if (p->mm) { | |
15b9f360 | 1911 | if (unlikely(p->mm == mm)) { |
5debfa6d | 1912 | lock_task_sighand(p, &flags); |
5c99cbf4 | 1913 | nr += zap_process(p, exit_code); |
5debfa6d ON |
1914 | unlock_task_sighand(p, &flags); |
1915 | } | |
aceecc04 ON |
1916 | break; |
1917 | } | |
e4901f92 | 1918 | } while_each_thread(g, p); |
aceecc04 | 1919 | } |
7b1c6154 | 1920 | rcu_read_unlock(); |
5debfa6d | 1921 | done: |
c5f1cc8c | 1922 | atomic_set(&core_state->nr_threads, nr); |
8cd9c249 | 1923 | return nr; |
1da177e4 LT |
1924 | } |
1925 | ||
9d5b327b | 1926 | static int coredump_wait(int exit_code, struct core_state *core_state) |
1da177e4 | 1927 | { |
dcf560c5 ON |
1928 | struct task_struct *tsk = current; |
1929 | struct mm_struct *mm = tsk->mm; | |
269b005a | 1930 | int core_waiters = -EBUSY; |
1da177e4 | 1931 | |
9d5b327b | 1932 | init_completion(&core_state->startup); |
b564daf8 ON |
1933 | core_state->dumper.task = tsk; |
1934 | core_state->dumper.next = NULL; | |
269b005a ON |
1935 | |
1936 | down_write(&mm->mmap_sem); | |
1937 | if (!mm->core_state) | |
1938 | core_waiters = zap_threads(tsk, mm, core_state, exit_code); | |
2384f55f ON |
1939 | up_write(&mm->mmap_sem); |
1940 | ||
11aeca0b SS |
1941 | if (core_waiters > 0) { |
1942 | struct core_thread *ptr; | |
1943 | ||
9d5b327b | 1944 | wait_for_completion(&core_state->startup); |
11aeca0b SS |
1945 | /* |
1946 | * Wait for all the threads to become inactive, so that | |
1947 | * all the thread context (extended register state, like | |
1948 | * fpu etc) gets copied to the memory. | |
1949 | */ | |
1950 | ptr = core_state->dumper.next; | |
1951 | while (ptr != NULL) { | |
1952 | wait_task_inactive(ptr->task, 0); | |
1953 | ptr = ptr->next; | |
1954 | } | |
1955 | } | |
57b59c4a | 1956 | |
dcf560c5 | 1957 | return core_waiters; |
1da177e4 LT |
1958 | } |
1959 | ||
a94e2d40 ON |
1960 | static void coredump_finish(struct mm_struct *mm) |
1961 | { | |
1962 | struct core_thread *curr, *next; | |
1963 | struct task_struct *task; | |
1964 | ||
1965 | next = mm->core_state->dumper.next; | |
1966 | while ((curr = next) != NULL) { | |
1967 | next = curr->next; | |
1968 | task = curr->task; | |
1969 | /* | |
1970 | * see exit_mm(), curr->task must not see | |
1971 | * ->task == NULL before we read ->next. | |
1972 | */ | |
1973 | smp_mb(); | |
1974 | curr->task = NULL; | |
1975 | wake_up_process(task); | |
1976 | } | |
1977 | ||
1978 | mm->core_state = NULL; | |
1979 | } | |
1980 | ||
6c5d5238 KH |
1981 | /* |
1982 | * set_dumpable converts traditional three-value dumpable to two flags and | |
1983 | * stores them into mm->flags. It modifies lower two bits of mm->flags, but | |
1984 | * these bits are not changed atomically. So get_dumpable can observe the | |
1985 | * intermediate state. To avoid doing unexpected behavior, get get_dumpable | |
1986 | * return either old dumpable or new one by paying attention to the order of | |
1987 | * modifying the bits. | |
1988 | * | |
1989 | * dumpable | mm->flags (binary) | |
1990 | * old new | initial interim final | |
1991 | * ---------+----------------------- | |
1992 | * 0 1 | 00 01 01 | |
1993 | * 0 2 | 00 10(*) 11 | |
1994 | * 1 0 | 01 00 00 | |
1995 | * 1 2 | 01 11 11 | |
1996 | * 2 0 | 11 10(*) 00 | |
1997 | * 2 1 | 11 11 01 | |
1998 | * | |
1999 | * (*) get_dumpable regards interim value of 10 as 11. | |
2000 | */ | |
2001 | void set_dumpable(struct mm_struct *mm, int value) | |
2002 | { | |
2003 | switch (value) { | |
54b50199 | 2004 | case SUID_DUMPABLE_DISABLED: |
6c5d5238 KH |
2005 | clear_bit(MMF_DUMPABLE, &mm->flags); |
2006 | smp_wmb(); | |
2007 | clear_bit(MMF_DUMP_SECURELY, &mm->flags); | |
2008 | break; | |
54b50199 | 2009 | case SUID_DUMPABLE_ENABLED: |
6c5d5238 KH |
2010 | set_bit(MMF_DUMPABLE, &mm->flags); |
2011 | smp_wmb(); | |
2012 | clear_bit(MMF_DUMP_SECURELY, &mm->flags); | |
2013 | break; | |
54b50199 | 2014 | case SUID_DUMPABLE_SAFE: |
6c5d5238 KH |
2015 | set_bit(MMF_DUMP_SECURELY, &mm->flags); |
2016 | smp_wmb(); | |
2017 | set_bit(MMF_DUMPABLE, &mm->flags); | |
2018 | break; | |
2019 | } | |
2020 | } | |
6c5d5238 | 2021 | |
30736a4d | 2022 | static int __get_dumpable(unsigned long mm_flags) |
6c5d5238 KH |
2023 | { |
2024 | int ret; | |
2025 | ||
30736a4d | 2026 | ret = mm_flags & MMF_DUMPABLE_MASK; |
54b50199 | 2027 | return (ret > SUID_DUMPABLE_ENABLED) ? SUID_DUMPABLE_SAFE : ret; |
6c5d5238 KH |
2028 | } |
2029 | ||
30736a4d MH |
2030 | int get_dumpable(struct mm_struct *mm) |
2031 | { | |
2032 | return __get_dumpable(mm->flags); | |
2033 | } | |
2034 | ||
61be228a NH |
2035 | static void wait_for_dump_helpers(struct file *file) |
2036 | { | |
2037 | struct pipe_inode_info *pipe; | |
2038 | ||
2039 | pipe = file->f_path.dentry->d_inode->i_pipe; | |
2040 | ||
2041 | pipe_lock(pipe); | |
2042 | pipe->readers++; | |
2043 | pipe->writers--; | |
2044 | ||
2045 | while ((pipe->readers > 1) && (!signal_pending(current))) { | |
2046 | wake_up_interruptible_sync(&pipe->wait); | |
2047 | kill_fasync(&pipe->fasync_readers, SIGIO, POLL_IN); | |
2048 | pipe_wait(pipe); | |
2049 | } | |
2050 | ||
2051 | pipe->readers--; | |
2052 | pipe->writers++; | |
2053 | pipe_unlock(pipe); | |
2054 | ||
2055 | } | |
2056 | ||
2057 | ||
898b374a | 2058 | /* |
1bef8291 | 2059 | * umh_pipe_setup |
898b374a NH |
2060 | * helper function to customize the process used |
2061 | * to collect the core in userspace. Specifically | |
2062 | * it sets up a pipe and installs it as fd 0 (stdin) | |
2063 | * for the process. Returns 0 on success, or | |
2064 | * PTR_ERR on failure. | |
2065 | * Note that it also sets the core limit to 1. This | |
2066 | * is a special value that we use to trap recursive | |
2067 | * core dumps | |
2068 | */ | |
87966996 | 2069 | static int umh_pipe_setup(struct subprocess_info *info, struct cred *new) |
898b374a | 2070 | { |
e4fad8e5 | 2071 | struct file *files[2]; |
898b374a NH |
2072 | struct fdtable *fdt; |
2073 | struct coredump_params *cp = (struct coredump_params *)info->data; | |
2074 | struct files_struct *cf = current->files; | |
e4fad8e5 AV |
2075 | int err = create_pipe_files(files, 0); |
2076 | if (err) | |
2077 | return err; | |
898b374a | 2078 | |
e4fad8e5 | 2079 | cp->file = files[1]; |
898b374a NH |
2080 | |
2081 | sys_close(0); | |
e4fad8e5 | 2082 | fd_install(0, files[0]); |
898b374a NH |
2083 | spin_lock(&cf->file_lock); |
2084 | fdt = files_fdtable(cf); | |
1dce27c5 DH |
2085 | __set_open_fd(0, fdt); |
2086 | __clear_close_on_exec(0, fdt); | |
898b374a NH |
2087 | spin_unlock(&cf->file_lock); |
2088 | ||
2089 | /* and disallow core files too */ | |
2090 | current->signal->rlim[RLIMIT_CORE] = (struct rlimit){1, 1}; | |
2091 | ||
2092 | return 0; | |
2093 | } | |
2094 | ||
8cd3ac3a | 2095 | void do_coredump(long signr, int exit_code, struct pt_regs *regs) |
1da177e4 | 2096 | { |
9d5b327b | 2097 | struct core_state core_state; |
1b0d300b | 2098 | struct core_name cn; |
1da177e4 LT |
2099 | struct mm_struct *mm = current->mm; |
2100 | struct linux_binfmt * binfmt; | |
d84f4f99 DH |
2101 | const struct cred *old_cred; |
2102 | struct cred *cred; | |
1da177e4 | 2103 | int retval = 0; |
d6e71144 | 2104 | int flag = 0; |
d5bf4c4f | 2105 | int ispipe; |
9520628e | 2106 | bool need_nonrelative = false; |
a293980c | 2107 | static atomic_t core_dump_count = ATOMIC_INIT(0); |
f6151dfe MH |
2108 | struct coredump_params cprm = { |
2109 | .signr = signr, | |
2110 | .regs = regs, | |
d554ed89 | 2111 | .limit = rlimit(RLIMIT_CORE), |
30736a4d MH |
2112 | /* |
2113 | * We must use the same mm->flags while dumping core to avoid | |
2114 | * inconsistency of bit flags, since this flag is not protected | |
2115 | * by any locks. | |
2116 | */ | |
2117 | .mm_flags = mm->flags, | |
f6151dfe | 2118 | }; |
1da177e4 | 2119 | |
0a4ff8c2 SG |
2120 | audit_core_dumps(signr); |
2121 | ||
801460d0 | 2122 | binfmt = mm->binfmt; |
1da177e4 LT |
2123 | if (!binfmt || !binfmt->core_dump) |
2124 | goto fail; | |
269b005a ON |
2125 | if (!__get_dumpable(cprm.mm_flags)) |
2126 | goto fail; | |
d84f4f99 DH |
2127 | |
2128 | cred = prepare_creds(); | |
5e43aef5 | 2129 | if (!cred) |
d84f4f99 | 2130 | goto fail; |
d6e71144 | 2131 | /* |
9520628e KC |
2132 | * We cannot trust fsuid as being the "true" uid of the process |
2133 | * nor do we know its entire history. We only know it was tainted | |
2134 | * so we dump it as root in mode 2, and only into a controlled | |
2135 | * environment (pipe handler or fully qualified path). | |
d6e71144 | 2136 | */ |
54b50199 | 2137 | if (__get_dumpable(cprm.mm_flags) == SUID_DUMPABLE_SAFE) { |
30736a4d | 2138 | /* Setuid core dump mode */ |
d6e71144 | 2139 | flag = O_EXCL; /* Stop rewrite attacks */ |
8e96e3b7 | 2140 | cred->fsuid = GLOBAL_ROOT_UID; /* Dump root private */ |
9520628e | 2141 | need_nonrelative = true; |
d6e71144 | 2142 | } |
1291cf41 | 2143 | |
9d5b327b | 2144 | retval = coredump_wait(exit_code, &core_state); |
5e43aef5 ON |
2145 | if (retval < 0) |
2146 | goto fail_creds; | |
d84f4f99 DH |
2147 | |
2148 | old_cred = override_creds(cred); | |
1da177e4 LT |
2149 | |
2150 | /* | |
2151 | * Clear any false indication of pending signals that might | |
2152 | * be seen by the filesystem code called to write the core file. | |
2153 | */ | |
1da177e4 LT |
2154 | clear_thread_flag(TIF_SIGPENDING); |
2155 | ||
1b0d300b XF |
2156 | ispipe = format_corename(&cn, signr); |
2157 | ||
c4bbafda | 2158 | if (ispipe) { |
d5bf4c4f ON |
2159 | int dump_count; |
2160 | char **helper_argv; | |
2161 | ||
99b64567 ON |
2162 | if (ispipe < 0) { |
2163 | printk(KERN_WARNING "format_corename failed\n"); | |
2164 | printk(KERN_WARNING "Aborting core\n"); | |
2165 | goto fail_corename; | |
2166 | } | |
2167 | ||
898b374a | 2168 | if (cprm.limit == 1) { |
108ceeb0 JZ |
2169 | /* See umh_pipe_setup() which sets RLIMIT_CORE = 1. |
2170 | * | |
725eae32 NH |
2171 | * Normally core limits are irrelevant to pipes, since |
2172 | * we're not writing to the file system, but we use | |
108ceeb0 JZ |
2173 | * cprm.limit of 1 here as a speacial value, this is a |
2174 | * consistent way to catch recursive crashes. | |
2175 | * We can still crash if the core_pattern binary sets | |
2176 | * RLIM_CORE = !1, but it runs as root, and can do | |
2177 | * lots of stupid things. | |
2178 | * | |
725eae32 NH |
2179 | * Note that we use task_tgid_vnr here to grab the pid |
2180 | * of the process group leader. That way we get the | |
2181 | * right pid if a thread in a multi-threaded | |
2182 | * core_pattern process dies. | |
2183 | */ | |
2184 | printk(KERN_WARNING | |
898b374a | 2185 | "Process %d(%s) has RLIMIT_CORE set to 1\n", |
725eae32 NH |
2186 | task_tgid_vnr(current), current->comm); |
2187 | printk(KERN_WARNING "Aborting core\n"); | |
2188 | goto fail_unlock; | |
2189 | } | |
d5bf4c4f | 2190 | cprm.limit = RLIM_INFINITY; |
725eae32 | 2191 | |
a293980c NH |
2192 | dump_count = atomic_inc_return(&core_dump_count); |
2193 | if (core_pipe_limit && (core_pipe_limit < dump_count)) { | |
2194 | printk(KERN_WARNING "Pid %d(%s) over core_pipe_limit\n", | |
2195 | task_tgid_vnr(current), current->comm); | |
2196 | printk(KERN_WARNING "Skipping core dump\n"); | |
2197 | goto fail_dropcount; | |
2198 | } | |
2199 | ||
1b0d300b | 2200 | helper_argv = argv_split(GFP_KERNEL, cn.corename+1, NULL); |
350eaf79 TH |
2201 | if (!helper_argv) { |
2202 | printk(KERN_WARNING "%s failed to allocate memory\n", | |
2203 | __func__); | |
a293980c | 2204 | goto fail_dropcount; |
350eaf79 | 2205 | } |
32321137 | 2206 | |
d5bf4c4f ON |
2207 | retval = call_usermodehelper_fns(helper_argv[0], helper_argv, |
2208 | NULL, UMH_WAIT_EXEC, umh_pipe_setup, | |
2209 | NULL, &cprm); | |
2210 | argv_free(helper_argv); | |
2211 | if (retval) { | |
d025c9db | 2212 | printk(KERN_INFO "Core dump to %s pipe failed\n", |
1b0d300b | 2213 | cn.corename); |
d5bf4c4f | 2214 | goto close_fail; |
d025c9db | 2215 | } |
c7135411 ON |
2216 | } else { |
2217 | struct inode *inode; | |
2218 | ||
2219 | if (cprm.limit < binfmt->min_coredump) | |
2220 | goto fail_unlock; | |
2221 | ||
9520628e KC |
2222 | if (need_nonrelative && cn.corename[0] != '/') { |
2223 | printk(KERN_WARNING "Pid %d(%s) can only dump core "\ | |
2224 | "to fully qualified path!\n", | |
2225 | task_tgid_vnr(current), current->comm); | |
2226 | printk(KERN_WARNING "Skipping core dump\n"); | |
2227 | goto fail_unlock; | |
2228 | } | |
2229 | ||
1b0d300b | 2230 | cprm.file = filp_open(cn.corename, |
6d4df677 AD |
2231 | O_CREAT | 2 | O_NOFOLLOW | O_LARGEFILE | flag, |
2232 | 0600); | |
c7135411 ON |
2233 | if (IS_ERR(cprm.file)) |
2234 | goto fail_unlock; | |
1da177e4 | 2235 | |
c7135411 ON |
2236 | inode = cprm.file->f_path.dentry->d_inode; |
2237 | if (inode->i_nlink > 1) | |
2238 | goto close_fail; | |
2239 | if (d_unhashed(cprm.file->f_path.dentry)) | |
2240 | goto close_fail; | |
2241 | /* | |
2242 | * AK: actually i see no reason to not allow this for named | |
2243 | * pipes etc, but keep the previous behaviour for now. | |
2244 | */ | |
2245 | if (!S_ISREG(inode->i_mode)) | |
2246 | goto close_fail; | |
2247 | /* | |
2248 | * Dont allow local users get cute and trick others to coredump | |
2249 | * into their pre-created files. | |
2250 | */ | |
8e96e3b7 | 2251 | if (!uid_eq(inode->i_uid, current_fsuid())) |
c7135411 ON |
2252 | goto close_fail; |
2253 | if (!cprm.file->f_op || !cprm.file->f_op->write) | |
2254 | goto close_fail; | |
2255 | if (do_truncate(cprm.file->f_path.dentry, 0, 0, cprm.file)) | |
2256 | goto close_fail; | |
2257 | } | |
1da177e4 | 2258 | |
c7135411 | 2259 | retval = binfmt->core_dump(&cprm); |
1da177e4 LT |
2260 | if (retval) |
2261 | current->signal->group_exit_code |= 0x80; | |
d5bf4c4f | 2262 | |
61be228a | 2263 | if (ispipe && core_pipe_limit) |
f6151dfe | 2264 | wait_for_dump_helpers(cprm.file); |
d5bf4c4f ON |
2265 | close_fail: |
2266 | if (cprm.file) | |
2267 | filp_close(cprm.file, NULL); | |
a293980c | 2268 | fail_dropcount: |
d5bf4c4f | 2269 | if (ispipe) |
a293980c | 2270 | atomic_dec(&core_dump_count); |
1da177e4 | 2271 | fail_unlock: |
1b0d300b XF |
2272 | kfree(cn.corename); |
2273 | fail_corename: | |
5e43aef5 | 2274 | coredump_finish(mm); |
d84f4f99 | 2275 | revert_creds(old_cred); |
5e43aef5 | 2276 | fail_creds: |
d84f4f99 | 2277 | put_cred(cred); |
1da177e4 | 2278 | fail: |
8cd3ac3a | 2279 | return; |
1da177e4 | 2280 | } |
3aa0ce82 LT |
2281 | |
2282 | /* | |
2283 | * Core dumping helper functions. These are the only things you should | |
2284 | * do on a core-file: use only these functions to write out all the | |
2285 | * necessary info. | |
2286 | */ | |
2287 | int dump_write(struct file *file, const void *addr, int nr) | |
2288 | { | |
2289 | return access_ok(VERIFY_READ, addr, nr) && file->f_op->write(file, addr, nr, &file->f_pos) == nr; | |
2290 | } | |
8fd01d6c | 2291 | EXPORT_SYMBOL(dump_write); |
3aa0ce82 LT |
2292 | |
2293 | int dump_seek(struct file *file, loff_t off) | |
2294 | { | |
2295 | int ret = 1; | |
2296 | ||
2297 | if (file->f_op->llseek && file->f_op->llseek != no_llseek) { | |
2298 | if (file->f_op->llseek(file, off, SEEK_CUR) < 0) | |
2299 | return 0; | |
2300 | } else { | |
2301 | char *buf = (char *)get_zeroed_page(GFP_KERNEL); | |
2302 | ||
2303 | if (!buf) | |
2304 | return 0; | |
2305 | while (off > 0) { | |
2306 | unsigned long n = off; | |
2307 | ||
2308 | if (n > PAGE_SIZE) | |
2309 | n = PAGE_SIZE; | |
2310 | if (!dump_write(file, buf, n)) { | |
2311 | ret = 0; | |
2312 | break; | |
2313 | } | |
2314 | off -= n; | |
2315 | } | |
2316 | free_page((unsigned long)buf); | |
2317 | } | |
2318 | return ret; | |
2319 | } | |
8fd01d6c | 2320 | EXPORT_SYMBOL(dump_seek); |
282124d1 AV |
2321 | |
2322 | #ifdef __ARCH_WANT_KERNEL_EXECVE | |
2323 | int kernel_execve(const char *filename, | |
2324 | const char *const argv[], | |
2325 | const char *const envp[]) | |
2326 | { | |
2327 | struct pt_regs *p = current_pt_regs(); | |
2328 | int ret; | |
2329 | ||
2330 | ret = do_execve(filename, | |
2331 | (const char __user *const __user *)argv, | |
2332 | (const char __user *const __user *)envp, p); | |
2333 | if (ret < 0) | |
2334 | return ret; | |
2335 | ||
2336 | /* | |
2337 | * We were successful. We won't be returning to our caller, but | |
2338 | * instead to user space by manipulating the kernel stack. | |
2339 | */ | |
2340 | ret_from_kernel_execve(p); | |
2341 | } | |
2342 | #endif |