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> | |
27 | #include <linux/mman.h> | |
28 | #include <linux/a.out.h> | |
29 | #include <linux/stat.h> | |
30 | #include <linux/fcntl.h> | |
31 | #include <linux/smp_lock.h> | |
74aadce9 | 32 | #include <linux/string.h> |
1da177e4 LT |
33 | #include <linux/init.h> |
34 | #include <linux/pagemap.h> | |
35 | #include <linux/highmem.h> | |
36 | #include <linux/spinlock.h> | |
37 | #include <linux/key.h> | |
38 | #include <linux/personality.h> | |
39 | #include <linux/binfmts.h> | |
40 | #include <linux/swap.h> | |
41 | #include <linux/utsname.h> | |
84d73786 | 42 | #include <linux/pid_namespace.h> |
1da177e4 LT |
43 | #include <linux/module.h> |
44 | #include <linux/namei.h> | |
45 | #include <linux/proc_fs.h> | |
46 | #include <linux/ptrace.h> | |
47 | #include <linux/mount.h> | |
48 | #include <linux/security.h> | |
49 | #include <linux/syscalls.h> | |
50 | #include <linux/rmap.h> | |
8f0ab514 | 51 | #include <linux/tsacct_kern.h> |
9f46080c | 52 | #include <linux/cn_proc.h> |
473ae30b | 53 | #include <linux/audit.h> |
1da177e4 LT |
54 | |
55 | #include <asm/uaccess.h> | |
56 | #include <asm/mmu_context.h> | |
b6a2fea3 | 57 | #include <asm/tlb.h> |
1da177e4 LT |
58 | |
59 | #ifdef CONFIG_KMOD | |
60 | #include <linux/kmod.h> | |
61 | #endif | |
62 | ||
63 | int core_uses_pid; | |
71ce92f3 | 64 | char core_pattern[CORENAME_MAX_SIZE] = "core"; |
d6e71144 AC |
65 | int suid_dumpable = 0; |
66 | ||
67 | EXPORT_SYMBOL(suid_dumpable); | |
1da177e4 LT |
68 | /* The maximal length of core_pattern is also specified in sysctl.c */ |
69 | ||
e4dc1b14 | 70 | static LIST_HEAD(formats); |
1da177e4 LT |
71 | static DEFINE_RWLOCK(binfmt_lock); |
72 | ||
73 | int register_binfmt(struct linux_binfmt * fmt) | |
74 | { | |
1da177e4 LT |
75 | if (!fmt) |
76 | return -EINVAL; | |
1da177e4 | 77 | write_lock(&binfmt_lock); |
e4dc1b14 | 78 | list_add(&fmt->lh, &formats); |
1da177e4 LT |
79 | write_unlock(&binfmt_lock); |
80 | return 0; | |
81 | } | |
82 | ||
83 | EXPORT_SYMBOL(register_binfmt); | |
84 | ||
f6b450d4 | 85 | void unregister_binfmt(struct linux_binfmt * fmt) |
1da177e4 | 86 | { |
1da177e4 | 87 | write_lock(&binfmt_lock); |
e4dc1b14 | 88 | list_del(&fmt->lh); |
1da177e4 | 89 | write_unlock(&binfmt_lock); |
1da177e4 LT |
90 | } |
91 | ||
92 | EXPORT_SYMBOL(unregister_binfmt); | |
93 | ||
94 | static inline void put_binfmt(struct linux_binfmt * fmt) | |
95 | { | |
96 | module_put(fmt->module); | |
97 | } | |
98 | ||
99 | /* | |
100 | * Note that a shared library must be both readable and executable due to | |
101 | * security reasons. | |
102 | * | |
103 | * Also note that we take the address to load from from the file itself. | |
104 | */ | |
105 | asmlinkage long sys_uselib(const char __user * library) | |
106 | { | |
107 | struct file * file; | |
108 | struct nameidata nd; | |
109 | int error; | |
110 | ||
b500531e | 111 | error = __user_path_lookup_open(library, LOOKUP_FOLLOW, &nd, FMODE_READ|FMODE_EXEC); |
1da177e4 LT |
112 | if (error) |
113 | goto out; | |
114 | ||
115 | error = -EINVAL; | |
116 | if (!S_ISREG(nd.dentry->d_inode->i_mode)) | |
117 | goto exit; | |
118 | ||
e4543edd | 119 | error = vfs_permission(&nd, MAY_READ | MAY_EXEC); |
1da177e4 LT |
120 | if (error) |
121 | goto exit; | |
122 | ||
834f2a4a | 123 | file = nameidata_to_filp(&nd, O_RDONLY); |
1da177e4 LT |
124 | error = PTR_ERR(file); |
125 | if (IS_ERR(file)) | |
126 | goto out; | |
127 | ||
128 | error = -ENOEXEC; | |
129 | if(file->f_op) { | |
130 | struct linux_binfmt * fmt; | |
131 | ||
132 | read_lock(&binfmt_lock); | |
e4dc1b14 | 133 | list_for_each_entry(fmt, &formats, lh) { |
1da177e4 LT |
134 | if (!fmt->load_shlib) |
135 | continue; | |
136 | if (!try_module_get(fmt->module)) | |
137 | continue; | |
138 | read_unlock(&binfmt_lock); | |
139 | error = fmt->load_shlib(file); | |
140 | read_lock(&binfmt_lock); | |
141 | put_binfmt(fmt); | |
142 | if (error != -ENOEXEC) | |
143 | break; | |
144 | } | |
145 | read_unlock(&binfmt_lock); | |
146 | } | |
147 | fput(file); | |
148 | out: | |
149 | return error; | |
150 | exit: | |
834f2a4a | 151 | release_open_intent(&nd); |
1da177e4 LT |
152 | path_release(&nd); |
153 | goto out; | |
154 | } | |
155 | ||
b6a2fea3 OW |
156 | #ifdef CONFIG_MMU |
157 | ||
158 | static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos, | |
159 | int write) | |
160 | { | |
161 | struct page *page; | |
162 | int ret; | |
163 | ||
164 | #ifdef CONFIG_STACK_GROWSUP | |
165 | if (write) { | |
166 | ret = expand_stack_downwards(bprm->vma, pos); | |
167 | if (ret < 0) | |
168 | return NULL; | |
169 | } | |
170 | #endif | |
171 | ret = get_user_pages(current, bprm->mm, pos, | |
172 | 1, write, 1, &page, NULL); | |
173 | if (ret <= 0) | |
174 | return NULL; | |
175 | ||
176 | if (write) { | |
177 | struct rlimit *rlim = current->signal->rlim; | |
178 | unsigned long size = bprm->vma->vm_end - bprm->vma->vm_start; | |
179 | ||
180 | /* | |
181 | * Limit to 1/4-th the stack size for the argv+env strings. | |
182 | * This ensures that: | |
183 | * - the remaining binfmt code will not run out of stack space, | |
184 | * - the program will have a reasonable amount of stack left | |
185 | * to work from. | |
186 | */ | |
187 | if (size > rlim[RLIMIT_STACK].rlim_cur / 4) { | |
188 | put_page(page); | |
189 | return NULL; | |
190 | } | |
191 | } | |
192 | ||
193 | return page; | |
194 | } | |
195 | ||
196 | static void put_arg_page(struct page *page) | |
197 | { | |
198 | put_page(page); | |
199 | } | |
200 | ||
201 | static void free_arg_page(struct linux_binprm *bprm, int i) | |
202 | { | |
203 | } | |
204 | ||
205 | static void free_arg_pages(struct linux_binprm *bprm) | |
206 | { | |
207 | } | |
208 | ||
209 | static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos, | |
210 | struct page *page) | |
211 | { | |
212 | flush_cache_page(bprm->vma, pos, page_to_pfn(page)); | |
213 | } | |
214 | ||
215 | static int __bprm_mm_init(struct linux_binprm *bprm) | |
216 | { | |
217 | int err = -ENOMEM; | |
218 | struct vm_area_struct *vma = NULL; | |
219 | struct mm_struct *mm = bprm->mm; | |
220 | ||
221 | bprm->vma = vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL); | |
222 | if (!vma) | |
223 | goto err; | |
224 | ||
225 | down_write(&mm->mmap_sem); | |
226 | vma->vm_mm = mm; | |
227 | ||
228 | /* | |
229 | * Place the stack at the largest stack address the architecture | |
230 | * supports. Later, we'll move this to an appropriate place. We don't | |
231 | * use STACK_TOP because that can depend on attributes which aren't | |
232 | * configured yet. | |
233 | */ | |
234 | vma->vm_end = STACK_TOP_MAX; | |
235 | vma->vm_start = vma->vm_end - PAGE_SIZE; | |
236 | ||
237 | vma->vm_flags = VM_STACK_FLAGS; | |
238 | vma->vm_page_prot = protection_map[vma->vm_flags & 0x7]; | |
239 | err = insert_vm_struct(mm, vma); | |
240 | if (err) { | |
241 | up_write(&mm->mmap_sem); | |
242 | goto err; | |
243 | } | |
244 | ||
245 | mm->stack_vm = mm->total_vm = 1; | |
246 | up_write(&mm->mmap_sem); | |
247 | ||
248 | bprm->p = vma->vm_end - sizeof(void *); | |
249 | ||
250 | return 0; | |
251 | ||
252 | err: | |
253 | if (vma) { | |
254 | bprm->vma = NULL; | |
255 | kmem_cache_free(vm_area_cachep, vma); | |
256 | } | |
257 | ||
258 | return err; | |
259 | } | |
260 | ||
261 | static bool valid_arg_len(struct linux_binprm *bprm, long len) | |
262 | { | |
263 | return len <= MAX_ARG_STRLEN; | |
264 | } | |
265 | ||
266 | #else | |
267 | ||
268 | static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos, | |
269 | int write) | |
270 | { | |
271 | struct page *page; | |
272 | ||
273 | page = bprm->page[pos / PAGE_SIZE]; | |
274 | if (!page && write) { | |
275 | page = alloc_page(GFP_HIGHUSER|__GFP_ZERO); | |
276 | if (!page) | |
277 | return NULL; | |
278 | bprm->page[pos / PAGE_SIZE] = page; | |
279 | } | |
280 | ||
281 | return page; | |
282 | } | |
283 | ||
284 | static void put_arg_page(struct page *page) | |
285 | { | |
286 | } | |
287 | ||
288 | static void free_arg_page(struct linux_binprm *bprm, int i) | |
289 | { | |
290 | if (bprm->page[i]) { | |
291 | __free_page(bprm->page[i]); | |
292 | bprm->page[i] = NULL; | |
293 | } | |
294 | } | |
295 | ||
296 | static void free_arg_pages(struct linux_binprm *bprm) | |
297 | { | |
298 | int i; | |
299 | ||
300 | for (i = 0; i < MAX_ARG_PAGES; i++) | |
301 | free_arg_page(bprm, i); | |
302 | } | |
303 | ||
304 | static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos, | |
305 | struct page *page) | |
306 | { | |
307 | } | |
308 | ||
309 | static int __bprm_mm_init(struct linux_binprm *bprm) | |
310 | { | |
311 | bprm->p = PAGE_SIZE * MAX_ARG_PAGES - sizeof(void *); | |
312 | return 0; | |
313 | } | |
314 | ||
315 | static bool valid_arg_len(struct linux_binprm *bprm, long len) | |
316 | { | |
317 | return len <= bprm->p; | |
318 | } | |
319 | ||
320 | #endif /* CONFIG_MMU */ | |
321 | ||
322 | /* | |
323 | * Create a new mm_struct and populate it with a temporary stack | |
324 | * vm_area_struct. We don't have enough context at this point to set the stack | |
325 | * flags, permissions, and offset, so we use temporary values. We'll update | |
326 | * them later in setup_arg_pages(). | |
327 | */ | |
328 | int bprm_mm_init(struct linux_binprm *bprm) | |
329 | { | |
330 | int err; | |
331 | struct mm_struct *mm = NULL; | |
332 | ||
333 | bprm->mm = mm = mm_alloc(); | |
334 | err = -ENOMEM; | |
335 | if (!mm) | |
336 | goto err; | |
337 | ||
338 | err = init_new_context(current, mm); | |
339 | if (err) | |
340 | goto err; | |
341 | ||
342 | err = __bprm_mm_init(bprm); | |
343 | if (err) | |
344 | goto err; | |
345 | ||
346 | return 0; | |
347 | ||
348 | err: | |
349 | if (mm) { | |
350 | bprm->mm = NULL; | |
351 | mmdrop(mm); | |
352 | } | |
353 | ||
354 | return err; | |
355 | } | |
356 | ||
1da177e4 LT |
357 | /* |
358 | * count() counts the number of strings in array ARGV. | |
359 | */ | |
360 | static int count(char __user * __user * argv, int max) | |
361 | { | |
362 | int i = 0; | |
363 | ||
364 | if (argv != NULL) { | |
365 | for (;;) { | |
366 | char __user * p; | |
367 | ||
368 | if (get_user(p, argv)) | |
369 | return -EFAULT; | |
370 | if (!p) | |
371 | break; | |
372 | argv++; | |
373 | if(++i > max) | |
374 | return -E2BIG; | |
375 | cond_resched(); | |
376 | } | |
377 | } | |
378 | return i; | |
379 | } | |
380 | ||
381 | /* | |
b6a2fea3 OW |
382 | * 'copy_strings()' copies argument/environment strings from the old |
383 | * processes's memory to the new process's stack. The call to get_user_pages() | |
384 | * ensures the destination page is created and not swapped out. | |
1da177e4 | 385 | */ |
75c96f85 AB |
386 | static int copy_strings(int argc, char __user * __user * argv, |
387 | struct linux_binprm *bprm) | |
1da177e4 LT |
388 | { |
389 | struct page *kmapped_page = NULL; | |
390 | char *kaddr = NULL; | |
b6a2fea3 | 391 | unsigned long kpos = 0; |
1da177e4 LT |
392 | int ret; |
393 | ||
394 | while (argc-- > 0) { | |
395 | char __user *str; | |
396 | int len; | |
397 | unsigned long pos; | |
398 | ||
399 | if (get_user(str, argv+argc) || | |
b6a2fea3 | 400 | !(len = strnlen_user(str, MAX_ARG_STRLEN))) { |
1da177e4 LT |
401 | ret = -EFAULT; |
402 | goto out; | |
403 | } | |
404 | ||
b6a2fea3 | 405 | if (!valid_arg_len(bprm, len)) { |
1da177e4 LT |
406 | ret = -E2BIG; |
407 | goto out; | |
408 | } | |
409 | ||
b6a2fea3 | 410 | /* We're going to work our way backwords. */ |
1da177e4 | 411 | pos = bprm->p; |
b6a2fea3 OW |
412 | str += len; |
413 | bprm->p -= len; | |
1da177e4 LT |
414 | |
415 | while (len > 0) { | |
1da177e4 | 416 | int offset, bytes_to_copy; |
1da177e4 LT |
417 | |
418 | offset = pos % PAGE_SIZE; | |
b6a2fea3 OW |
419 | if (offset == 0) |
420 | offset = PAGE_SIZE; | |
421 | ||
422 | bytes_to_copy = offset; | |
423 | if (bytes_to_copy > len) | |
424 | bytes_to_copy = len; | |
425 | ||
426 | offset -= bytes_to_copy; | |
427 | pos -= bytes_to_copy; | |
428 | str -= bytes_to_copy; | |
429 | len -= bytes_to_copy; | |
430 | ||
431 | if (!kmapped_page || kpos != (pos & PAGE_MASK)) { | |
432 | struct page *page; | |
433 | ||
434 | page = get_arg_page(bprm, pos, 1); | |
1da177e4 | 435 | if (!page) { |
b6a2fea3 | 436 | ret = -E2BIG; |
1da177e4 LT |
437 | goto out; |
438 | } | |
1da177e4 | 439 | |
b6a2fea3 OW |
440 | if (kmapped_page) { |
441 | flush_kernel_dcache_page(kmapped_page); | |
1da177e4 | 442 | kunmap(kmapped_page); |
b6a2fea3 OW |
443 | put_arg_page(kmapped_page); |
444 | } | |
1da177e4 LT |
445 | kmapped_page = page; |
446 | kaddr = kmap(kmapped_page); | |
b6a2fea3 OW |
447 | kpos = pos & PAGE_MASK; |
448 | flush_arg_page(bprm, kpos, kmapped_page); | |
1da177e4 | 449 | } |
b6a2fea3 | 450 | if (copy_from_user(kaddr+offset, str, bytes_to_copy)) { |
1da177e4 LT |
451 | ret = -EFAULT; |
452 | goto out; | |
453 | } | |
1da177e4 LT |
454 | } |
455 | } | |
456 | ret = 0; | |
457 | out: | |
b6a2fea3 OW |
458 | if (kmapped_page) { |
459 | flush_kernel_dcache_page(kmapped_page); | |
1da177e4 | 460 | kunmap(kmapped_page); |
b6a2fea3 OW |
461 | put_arg_page(kmapped_page); |
462 | } | |
1da177e4 LT |
463 | return ret; |
464 | } | |
465 | ||
466 | /* | |
467 | * Like copy_strings, but get argv and its values from kernel memory. | |
468 | */ | |
469 | int copy_strings_kernel(int argc,char ** argv, struct linux_binprm *bprm) | |
470 | { | |
471 | int r; | |
472 | mm_segment_t oldfs = get_fs(); | |
473 | set_fs(KERNEL_DS); | |
474 | r = copy_strings(argc, (char __user * __user *)argv, bprm); | |
475 | set_fs(oldfs); | |
476 | return r; | |
477 | } | |
1da177e4 LT |
478 | EXPORT_SYMBOL(copy_strings_kernel); |
479 | ||
480 | #ifdef CONFIG_MMU | |
b6a2fea3 | 481 | |
1da177e4 | 482 | /* |
b6a2fea3 OW |
483 | * During bprm_mm_init(), we create a temporary stack at STACK_TOP_MAX. Once |
484 | * the binfmt code determines where the new stack should reside, we shift it to | |
485 | * its final location. The process proceeds as follows: | |
1da177e4 | 486 | * |
b6a2fea3 OW |
487 | * 1) Use shift to calculate the new vma endpoints. |
488 | * 2) Extend vma to cover both the old and new ranges. This ensures the | |
489 | * arguments passed to subsequent functions are consistent. | |
490 | * 3) Move vma's page tables to the new range. | |
491 | * 4) Free up any cleared pgd range. | |
492 | * 5) Shrink the vma to cover only the new range. | |
1da177e4 | 493 | */ |
b6a2fea3 | 494 | static int shift_arg_pages(struct vm_area_struct *vma, unsigned long shift) |
1da177e4 LT |
495 | { |
496 | struct mm_struct *mm = vma->vm_mm; | |
b6a2fea3 OW |
497 | unsigned long old_start = vma->vm_start; |
498 | unsigned long old_end = vma->vm_end; | |
499 | unsigned long length = old_end - old_start; | |
500 | unsigned long new_start = old_start - shift; | |
501 | unsigned long new_end = old_end - shift; | |
502 | struct mmu_gather *tlb; | |
1da177e4 | 503 | |
b6a2fea3 | 504 | BUG_ON(new_start > new_end); |
1da177e4 | 505 | |
b6a2fea3 OW |
506 | /* |
507 | * ensure there are no vmas between where we want to go | |
508 | * and where we are | |
509 | */ | |
510 | if (vma != find_vma(mm, new_start)) | |
511 | return -EFAULT; | |
512 | ||
513 | /* | |
514 | * cover the whole range: [new_start, old_end) | |
515 | */ | |
516 | vma_adjust(vma, new_start, old_end, vma->vm_pgoff, NULL); | |
517 | ||
518 | /* | |
519 | * move the page tables downwards, on failure we rely on | |
520 | * process cleanup to remove whatever mess we made. | |
521 | */ | |
522 | if (length != move_page_tables(vma, old_start, | |
523 | vma, new_start, length)) | |
524 | return -ENOMEM; | |
525 | ||
526 | lru_add_drain(); | |
527 | tlb = tlb_gather_mmu(mm, 0); | |
528 | if (new_end > old_start) { | |
529 | /* | |
530 | * when the old and new regions overlap clear from new_end. | |
531 | */ | |
532 | free_pgd_range(&tlb, new_end, old_end, new_end, | |
533 | vma->vm_next ? vma->vm_next->vm_start : 0); | |
534 | } else { | |
535 | /* | |
536 | * otherwise, clean from old_start; this is done to not touch | |
537 | * the address space in [new_end, old_start) some architectures | |
538 | * have constraints on va-space that make this illegal (IA64) - | |
539 | * for the others its just a little faster. | |
540 | */ | |
541 | free_pgd_range(&tlb, old_start, old_end, new_end, | |
542 | vma->vm_next ? vma->vm_next->vm_start : 0); | |
1da177e4 | 543 | } |
b6a2fea3 OW |
544 | tlb_finish_mmu(tlb, new_end, old_end); |
545 | ||
546 | /* | |
547 | * shrink the vma to just the new range. | |
548 | */ | |
549 | vma_adjust(vma, new_start, new_end, vma->vm_pgoff, NULL); | |
550 | ||
551 | return 0; | |
1da177e4 LT |
552 | } |
553 | ||
554 | #define EXTRA_STACK_VM_PAGES 20 /* random */ | |
555 | ||
b6a2fea3 OW |
556 | /* |
557 | * Finalizes the stack vm_area_struct. The flags and permissions are updated, | |
558 | * the stack is optionally relocated, and some extra space is added. | |
559 | */ | |
1da177e4 LT |
560 | int setup_arg_pages(struct linux_binprm *bprm, |
561 | unsigned long stack_top, | |
562 | int executable_stack) | |
563 | { | |
b6a2fea3 OW |
564 | unsigned long ret; |
565 | unsigned long stack_shift; | |
1da177e4 | 566 | struct mm_struct *mm = current->mm; |
b6a2fea3 OW |
567 | struct vm_area_struct *vma = bprm->vma; |
568 | struct vm_area_struct *prev = NULL; | |
569 | unsigned long vm_flags; | |
570 | unsigned long stack_base; | |
1da177e4 LT |
571 | |
572 | #ifdef CONFIG_STACK_GROWSUP | |
1da177e4 LT |
573 | /* Limit stack size to 1GB */ |
574 | stack_base = current->signal->rlim[RLIMIT_STACK].rlim_max; | |
575 | if (stack_base > (1 << 30)) | |
576 | stack_base = 1 << 30; | |
1da177e4 | 577 | |
b6a2fea3 OW |
578 | /* Make sure we didn't let the argument array grow too large. */ |
579 | if (vma->vm_end - vma->vm_start > stack_base) | |
580 | return -ENOMEM; | |
1da177e4 | 581 | |
b6a2fea3 | 582 | stack_base = PAGE_ALIGN(stack_top - stack_base); |
1da177e4 | 583 | |
b6a2fea3 OW |
584 | stack_shift = vma->vm_start - stack_base; |
585 | mm->arg_start = bprm->p - stack_shift; | |
586 | bprm->p = vma->vm_end - stack_shift; | |
1da177e4 | 587 | #else |
b6a2fea3 OW |
588 | stack_top = arch_align_stack(stack_top); |
589 | stack_top = PAGE_ALIGN(stack_top); | |
590 | stack_shift = vma->vm_end - stack_top; | |
591 | ||
592 | bprm->p -= stack_shift; | |
1da177e4 | 593 | mm->arg_start = bprm->p; |
1da177e4 LT |
594 | #endif |
595 | ||
1da177e4 | 596 | if (bprm->loader) |
b6a2fea3 OW |
597 | bprm->loader -= stack_shift; |
598 | bprm->exec -= stack_shift; | |
1da177e4 | 599 | |
1da177e4 | 600 | down_write(&mm->mmap_sem); |
b6a2fea3 OW |
601 | vm_flags = vma->vm_flags; |
602 | ||
603 | /* | |
604 | * Adjust stack execute permissions; explicitly enable for | |
605 | * EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X and leave alone | |
606 | * (arch default) otherwise. | |
607 | */ | |
608 | if (unlikely(executable_stack == EXSTACK_ENABLE_X)) | |
609 | vm_flags |= VM_EXEC; | |
610 | else if (executable_stack == EXSTACK_DISABLE_X) | |
611 | vm_flags &= ~VM_EXEC; | |
612 | vm_flags |= mm->def_flags; | |
613 | ||
614 | ret = mprotect_fixup(vma, &prev, vma->vm_start, vma->vm_end, | |
615 | vm_flags); | |
616 | if (ret) | |
617 | goto out_unlock; | |
618 | BUG_ON(prev != vma); | |
619 | ||
620 | /* Move stack pages down in memory. */ | |
621 | if (stack_shift) { | |
622 | ret = shift_arg_pages(vma, stack_shift); | |
623 | if (ret) { | |
1da177e4 | 624 | up_write(&mm->mmap_sem); |
1da177e4 LT |
625 | return ret; |
626 | } | |
1da177e4 LT |
627 | } |
628 | ||
b6a2fea3 OW |
629 | #ifdef CONFIG_STACK_GROWSUP |
630 | stack_base = vma->vm_end + EXTRA_STACK_VM_PAGES * PAGE_SIZE; | |
631 | #else | |
632 | stack_base = vma->vm_start - EXTRA_STACK_VM_PAGES * PAGE_SIZE; | |
633 | #endif | |
634 | ret = expand_stack(vma, stack_base); | |
635 | if (ret) | |
636 | ret = -EFAULT; | |
637 | ||
638 | out_unlock: | |
1da177e4 | 639 | up_write(&mm->mmap_sem); |
1da177e4 LT |
640 | return 0; |
641 | } | |
1da177e4 LT |
642 | EXPORT_SYMBOL(setup_arg_pages); |
643 | ||
1da177e4 LT |
644 | #endif /* CONFIG_MMU */ |
645 | ||
646 | struct file *open_exec(const char *name) | |
647 | { | |
648 | struct nameidata nd; | |
649 | int err; | |
650 | struct file *file; | |
651 | ||
b500531e | 652 | err = path_lookup_open(AT_FDCWD, name, LOOKUP_FOLLOW, &nd, FMODE_READ|FMODE_EXEC); |
1da177e4 LT |
653 | file = ERR_PTR(err); |
654 | ||
655 | if (!err) { | |
656 | struct inode *inode = nd.dentry->d_inode; | |
657 | file = ERR_PTR(-EACCES); | |
1a159dd2 | 658 | if (S_ISREG(inode->i_mode)) { |
e4543edd | 659 | int err = vfs_permission(&nd, MAY_EXEC); |
1da177e4 LT |
660 | file = ERR_PTR(err); |
661 | if (!err) { | |
834f2a4a | 662 | file = nameidata_to_filp(&nd, O_RDONLY); |
1da177e4 LT |
663 | if (!IS_ERR(file)) { |
664 | err = deny_write_access(file); | |
665 | if (err) { | |
666 | fput(file); | |
667 | file = ERR_PTR(err); | |
668 | } | |
669 | } | |
670 | out: | |
671 | return file; | |
672 | } | |
673 | } | |
834f2a4a | 674 | release_open_intent(&nd); |
1da177e4 LT |
675 | path_release(&nd); |
676 | } | |
677 | goto out; | |
678 | } | |
679 | ||
680 | EXPORT_SYMBOL(open_exec); | |
681 | ||
682 | int kernel_read(struct file *file, unsigned long offset, | |
683 | char *addr, unsigned long count) | |
684 | { | |
685 | mm_segment_t old_fs; | |
686 | loff_t pos = offset; | |
687 | int result; | |
688 | ||
689 | old_fs = get_fs(); | |
690 | set_fs(get_ds()); | |
691 | /* The cast to a user pointer is valid due to the set_fs() */ | |
692 | result = vfs_read(file, (void __user *)addr, count, &pos); | |
693 | set_fs(old_fs); | |
694 | return result; | |
695 | } | |
696 | ||
697 | EXPORT_SYMBOL(kernel_read); | |
698 | ||
699 | static int exec_mmap(struct mm_struct *mm) | |
700 | { | |
701 | struct task_struct *tsk; | |
702 | struct mm_struct * old_mm, *active_mm; | |
703 | ||
704 | /* Notify parent that we're no longer interested in the old VM */ | |
705 | tsk = current; | |
706 | old_mm = current->mm; | |
707 | mm_release(tsk, old_mm); | |
708 | ||
709 | if (old_mm) { | |
710 | /* | |
711 | * Make sure that if there is a core dump in progress | |
712 | * for the old mm, we get out and die instead of going | |
713 | * through with the exec. We must hold mmap_sem around | |
714 | * checking core_waiters and changing tsk->mm. The | |
715 | * core-inducing thread will increment core_waiters for | |
716 | * each thread whose ->mm == old_mm. | |
717 | */ | |
718 | down_read(&old_mm->mmap_sem); | |
719 | if (unlikely(old_mm->core_waiters)) { | |
720 | up_read(&old_mm->mmap_sem); | |
721 | return -EINTR; | |
722 | } | |
723 | } | |
724 | task_lock(tsk); | |
725 | active_mm = tsk->active_mm; | |
726 | tsk->mm = mm; | |
727 | tsk->active_mm = mm; | |
728 | activate_mm(active_mm, mm); | |
729 | task_unlock(tsk); | |
730 | arch_pick_mmap_layout(mm); | |
731 | if (old_mm) { | |
732 | up_read(&old_mm->mmap_sem); | |
7dddb12c | 733 | BUG_ON(active_mm != old_mm); |
1da177e4 LT |
734 | mmput(old_mm); |
735 | return 0; | |
736 | } | |
737 | mmdrop(active_mm); | |
738 | return 0; | |
739 | } | |
740 | ||
741 | /* | |
742 | * This function makes sure the current process has its own signal table, | |
743 | * so that flush_signal_handlers can later reset the handlers without | |
744 | * disturbing other processes. (Other processes might share the signal | |
745 | * table via the CLONE_SIGHAND option to clone().) | |
746 | */ | |
858119e1 | 747 | static int de_thread(struct task_struct *tsk) |
1da177e4 LT |
748 | { |
749 | struct signal_struct *sig = tsk->signal; | |
750 | struct sighand_struct *newsighand, *oldsighand = tsk->sighand; | |
751 | spinlock_t *lock = &oldsighand->siglock; | |
329f7dba | 752 | struct task_struct *leader = NULL; |
1da177e4 LT |
753 | int count; |
754 | ||
755 | /* | |
756 | * If we don't share sighandlers, then we aren't sharing anything | |
757 | * and we can just re-use it all. | |
758 | */ | |
759 | if (atomic_read(&oldsighand->count) <= 1) { | |
1da177e4 LT |
760 | exit_itimers(sig); |
761 | return 0; | |
762 | } | |
763 | ||
764 | newsighand = kmem_cache_alloc(sighand_cachep, GFP_KERNEL); | |
765 | if (!newsighand) | |
766 | return -ENOMEM; | |
767 | ||
aafe6c2a | 768 | if (thread_group_empty(tsk)) |
1da177e4 LT |
769 | goto no_thread_group; |
770 | ||
771 | /* | |
772 | * Kill all other threads in the thread group. | |
773 | * We must hold tasklist_lock to call zap_other_threads. | |
774 | */ | |
775 | read_lock(&tasklist_lock); | |
776 | spin_lock_irq(lock); | |
777 | if (sig->flags & SIGNAL_GROUP_EXIT) { | |
778 | /* | |
779 | * Another group action in progress, just | |
780 | * return so that the signal is processed. | |
781 | */ | |
782 | spin_unlock_irq(lock); | |
783 | read_unlock(&tasklist_lock); | |
784 | kmem_cache_free(sighand_cachep, newsighand); | |
785 | return -EAGAIN; | |
786 | } | |
1434261c ON |
787 | |
788 | /* | |
789 | * child_reaper ignores SIGKILL, change it now. | |
790 | * Reparenting needs write_lock on tasklist_lock, | |
791 | * so it is safe to do it under read_lock. | |
792 | */ | |
84d73786 SB |
793 | if (unlikely(tsk->group_leader == child_reaper(tsk))) |
794 | tsk->nsproxy->pid_ns->child_reaper = tsk; | |
1434261c | 795 | |
aafe6c2a | 796 | zap_other_threads(tsk); |
1da177e4 LT |
797 | read_unlock(&tasklist_lock); |
798 | ||
799 | /* | |
800 | * Account for the thread group leader hanging around: | |
801 | */ | |
9e4e23bc | 802 | count = 1; |
aafe6c2a | 803 | if (!thread_group_leader(tsk)) { |
9e4e23bc | 804 | count = 2; |
53231250 RM |
805 | /* |
806 | * The SIGALRM timer survives the exec, but needs to point | |
807 | * at us as the new group leader now. We have a race with | |
808 | * a timer firing now getting the old leader, so we need to | |
809 | * synchronize with any firing (by calling del_timer_sync) | |
810 | * before we can safely let the old group leader die. | |
811 | */ | |
aafe6c2a | 812 | sig->tsk = tsk; |
932aeafb | 813 | spin_unlock_irq(lock); |
2ff678b8 TG |
814 | if (hrtimer_cancel(&sig->real_timer)) |
815 | hrtimer_restart(&sig->real_timer); | |
932aeafb | 816 | spin_lock_irq(lock); |
53231250 | 817 | } |
1da177e4 | 818 | while (atomic_read(&sig->count) > count) { |
aafe6c2a | 819 | sig->group_exit_task = tsk; |
1da177e4 LT |
820 | sig->notify_count = count; |
821 | __set_current_state(TASK_UNINTERRUPTIBLE); | |
822 | spin_unlock_irq(lock); | |
823 | schedule(); | |
824 | spin_lock_irq(lock); | |
825 | } | |
826 | sig->group_exit_task = NULL; | |
827 | sig->notify_count = 0; | |
828 | spin_unlock_irq(lock); | |
829 | ||
830 | /* | |
831 | * At this point all other threads have exited, all we have to | |
832 | * do is to wait for the thread group leader to become inactive, | |
833 | * and to assume its PID: | |
834 | */ | |
aafe6c2a | 835 | if (!thread_group_leader(tsk)) { |
1da177e4 LT |
836 | /* |
837 | * Wait for the thread group leader to be a zombie. | |
838 | * It should already be zombie at this point, most | |
839 | * of the time. | |
840 | */ | |
aafe6c2a | 841 | leader = tsk->group_leader; |
1da177e4 LT |
842 | while (leader->exit_state != EXIT_ZOMBIE) |
843 | yield(); | |
844 | ||
f5e90281 RM |
845 | /* |
846 | * The only record we have of the real-time age of a | |
847 | * process, regardless of execs it's done, is start_time. | |
848 | * All the past CPU time is accumulated in signal_struct | |
849 | * from sister threads now dead. But in this non-leader | |
850 | * exec, nothing survives from the original leader thread, | |
851 | * whose birth marks the true age of this process now. | |
852 | * When we take on its identity by switching to its PID, we | |
853 | * also take its birthdate (always earlier than our own). | |
854 | */ | |
aafe6c2a | 855 | tsk->start_time = leader->start_time; |
f5e90281 | 856 | |
1da177e4 LT |
857 | write_lock_irq(&tasklist_lock); |
858 | ||
aafe6c2a EB |
859 | BUG_ON(leader->tgid != tsk->tgid); |
860 | BUG_ON(tsk->pid == tsk->tgid); | |
1da177e4 LT |
861 | /* |
862 | * An exec() starts a new thread group with the | |
863 | * TGID of the previous thread group. Rehash the | |
864 | * two threads with a switched PID, and release | |
865 | * the former thread group leader: | |
866 | */ | |
d73d6529 EB |
867 | |
868 | /* Become a process group leader with the old leader's pid. | |
c18258c6 EB |
869 | * The old leader becomes a thread of the this thread group. |
870 | * Note: The old leader also uses this pid until release_task | |
d73d6529 EB |
871 | * is called. Odd but simple and correct. |
872 | */ | |
aafe6c2a EB |
873 | detach_pid(tsk, PIDTYPE_PID); |
874 | tsk->pid = leader->pid; | |
e713d0da | 875 | attach_pid(tsk, PIDTYPE_PID, find_pid(tsk->pid)); |
aafe6c2a EB |
876 | transfer_pid(leader, tsk, PIDTYPE_PGID); |
877 | transfer_pid(leader, tsk, PIDTYPE_SID); | |
878 | list_replace_rcu(&leader->tasks, &tsk->tasks); | |
1da177e4 | 879 | |
aafe6c2a EB |
880 | tsk->group_leader = tsk; |
881 | leader->group_leader = tsk; | |
de12a787 | 882 | |
aafe6c2a | 883 | tsk->exit_signal = SIGCHLD; |
962b564c ON |
884 | |
885 | BUG_ON(leader->exit_state != EXIT_ZOMBIE); | |
886 | leader->exit_state = EXIT_DEAD; | |
1da177e4 LT |
887 | |
888 | write_unlock_irq(&tasklist_lock); | |
1da177e4 LT |
889 | } |
890 | ||
891 | /* | |
fb085cf1 AN |
892 | * There may be one thread left which is just exiting, |
893 | * but it's safe to stop telling the group to kill themselves. | |
1da177e4 LT |
894 | */ |
895 | sig->flags = 0; | |
896 | ||
897 | no_thread_group: | |
1da177e4 | 898 | exit_itimers(sig); |
329f7dba ON |
899 | if (leader) |
900 | release_task(leader); | |
901 | ||
1da177e4 LT |
902 | if (atomic_read(&oldsighand->count) == 1) { |
903 | /* | |
904 | * Now that we nuked the rest of the thread group, | |
905 | * it turns out we are not sharing sighand any more either. | |
906 | * So we can just keep it. | |
907 | */ | |
908 | kmem_cache_free(sighand_cachep, newsighand); | |
909 | } else { | |
910 | /* | |
911 | * Move our state over to newsighand and switch it in. | |
912 | */ | |
1da177e4 LT |
913 | atomic_set(&newsighand->count, 1); |
914 | memcpy(newsighand->action, oldsighand->action, | |
915 | sizeof(newsighand->action)); | |
916 | ||
917 | write_lock_irq(&tasklist_lock); | |
918 | spin_lock(&oldsighand->siglock); | |
513627d7 | 919 | spin_lock_nested(&newsighand->siglock, SINGLE_DEPTH_NESTING); |
1da177e4 | 920 | |
aafe6c2a | 921 | rcu_assign_pointer(tsk->sighand, newsighand); |
1da177e4 LT |
922 | recalc_sigpending(); |
923 | ||
924 | spin_unlock(&newsighand->siglock); | |
925 | spin_unlock(&oldsighand->siglock); | |
926 | write_unlock_irq(&tasklist_lock); | |
927 | ||
fba2afaa | 928 | __cleanup_sighand(oldsighand); |
1da177e4 LT |
929 | } |
930 | ||
aafe6c2a | 931 | BUG_ON(!thread_group_leader(tsk)); |
1da177e4 LT |
932 | return 0; |
933 | } | |
934 | ||
935 | /* | |
936 | * These functions flushes out all traces of the currently running executable | |
937 | * so that a new one can be started | |
938 | */ | |
939 | ||
858119e1 | 940 | static void flush_old_files(struct files_struct * files) |
1da177e4 LT |
941 | { |
942 | long j = -1; | |
badf1662 | 943 | struct fdtable *fdt; |
1da177e4 LT |
944 | |
945 | spin_lock(&files->file_lock); | |
946 | for (;;) { | |
947 | unsigned long set, i; | |
948 | ||
949 | j++; | |
950 | i = j * __NFDBITS; | |
badf1662 | 951 | fdt = files_fdtable(files); |
bbea9f69 | 952 | if (i >= fdt->max_fds) |
1da177e4 | 953 | break; |
badf1662 | 954 | set = fdt->close_on_exec->fds_bits[j]; |
1da177e4 LT |
955 | if (!set) |
956 | continue; | |
badf1662 | 957 | fdt->close_on_exec->fds_bits[j] = 0; |
1da177e4 LT |
958 | spin_unlock(&files->file_lock); |
959 | for ( ; set ; i++,set >>= 1) { | |
960 | if (set & 1) { | |
961 | sys_close(i); | |
962 | } | |
963 | } | |
964 | spin_lock(&files->file_lock); | |
965 | ||
966 | } | |
967 | spin_unlock(&files->file_lock); | |
968 | } | |
969 | ||
970 | void get_task_comm(char *buf, struct task_struct *tsk) | |
971 | { | |
972 | /* buf must be at least sizeof(tsk->comm) in size */ | |
973 | task_lock(tsk); | |
974 | strncpy(buf, tsk->comm, sizeof(tsk->comm)); | |
975 | task_unlock(tsk); | |
976 | } | |
977 | ||
978 | void set_task_comm(struct task_struct *tsk, char *buf) | |
979 | { | |
980 | task_lock(tsk); | |
981 | strlcpy(tsk->comm, buf, sizeof(tsk->comm)); | |
982 | task_unlock(tsk); | |
983 | } | |
984 | ||
985 | int flush_old_exec(struct linux_binprm * bprm) | |
986 | { | |
987 | char * name; | |
988 | int i, ch, retval; | |
989 | struct files_struct *files; | |
990 | char tcomm[sizeof(current->comm)]; | |
991 | ||
992 | /* | |
993 | * Make sure we have a private signal table and that | |
994 | * we are unassociated from the previous thread group. | |
995 | */ | |
996 | retval = de_thread(current); | |
997 | if (retval) | |
998 | goto out; | |
999 | ||
1000 | /* | |
1001 | * Make sure we have private file handles. Ask the | |
1002 | * fork helper to do the work for us and the exit | |
1003 | * helper to do the cleanup of the old one. | |
1004 | */ | |
1005 | files = current->files; /* refcounted so safe to hold */ | |
1006 | retval = unshare_files(); | |
1007 | if (retval) | |
1008 | goto out; | |
1009 | /* | |
1010 | * Release all of the old mmap stuff | |
1011 | */ | |
1012 | retval = exec_mmap(bprm->mm); | |
1013 | if (retval) | |
1014 | goto mmap_failed; | |
1015 | ||
1016 | bprm->mm = NULL; /* We're using it now */ | |
1017 | ||
1018 | /* This is the point of no return */ | |
1da177e4 LT |
1019 | put_files_struct(files); |
1020 | ||
1021 | current->sas_ss_sp = current->sas_ss_size = 0; | |
1022 | ||
1023 | if (current->euid == current->uid && current->egid == current->gid) | |
6c5d5238 | 1024 | set_dumpable(current->mm, 1); |
d6e71144 | 1025 | else |
6c5d5238 | 1026 | set_dumpable(current->mm, suid_dumpable); |
d6e71144 | 1027 | |
1da177e4 | 1028 | name = bprm->filename; |
36772092 PBG |
1029 | |
1030 | /* Copies the binary name from after last slash */ | |
1da177e4 LT |
1031 | for (i=0; (ch = *(name++)) != '\0';) { |
1032 | if (ch == '/') | |
36772092 | 1033 | i = 0; /* overwrite what we wrote */ |
1da177e4 LT |
1034 | else |
1035 | if (i < (sizeof(tcomm) - 1)) | |
1036 | tcomm[i++] = ch; | |
1037 | } | |
1038 | tcomm[i] = '\0'; | |
1039 | set_task_comm(current, tcomm); | |
1040 | ||
1041 | current->flags &= ~PF_RANDOMIZE; | |
1042 | flush_thread(); | |
1043 | ||
0551fbd2 BH |
1044 | /* Set the new mm task size. We have to do that late because it may |
1045 | * depend on TIF_32BIT which is only updated in flush_thread() on | |
1046 | * some architectures like powerpc | |
1047 | */ | |
1048 | current->mm->task_size = TASK_SIZE; | |
1049 | ||
d2d56c5f MH |
1050 | if (bprm->e_uid != current->euid || bprm->e_gid != current->egid) { |
1051 | suid_keys(current); | |
1052 | set_dumpable(current->mm, suid_dumpable); | |
1053 | current->pdeath_signal = 0; | |
1054 | } else if (file_permission(bprm->file, MAY_READ) || | |
1055 | (bprm->interp_flags & BINPRM_FLAGS_ENFORCE_NONDUMP)) { | |
1da177e4 | 1056 | suid_keys(current); |
6c5d5238 | 1057 | set_dumpable(current->mm, suid_dumpable); |
1da177e4 LT |
1058 | } |
1059 | ||
1060 | /* An exec changes our domain. We are no longer part of the thread | |
1061 | group */ | |
1062 | ||
1063 | current->self_exec_id++; | |
1064 | ||
1065 | flush_signal_handlers(current, 0); | |
1066 | flush_old_files(current->files); | |
1067 | ||
1068 | return 0; | |
1069 | ||
1070 | mmap_failed: | |
3b9b8ab6 | 1071 | reset_files_struct(current, files); |
1da177e4 LT |
1072 | out: |
1073 | return retval; | |
1074 | } | |
1075 | ||
1076 | EXPORT_SYMBOL(flush_old_exec); | |
1077 | ||
1078 | /* | |
1079 | * Fill the binprm structure from the inode. | |
1080 | * Check permissions, then read the first 128 (BINPRM_BUF_SIZE) bytes | |
1081 | */ | |
1082 | int prepare_binprm(struct linux_binprm *bprm) | |
1083 | { | |
1084 | int mode; | |
0f7fc9e4 | 1085 | struct inode * inode = bprm->file->f_path.dentry->d_inode; |
1da177e4 LT |
1086 | int retval; |
1087 | ||
1088 | mode = inode->i_mode; | |
1da177e4 LT |
1089 | if (bprm->file->f_op == NULL) |
1090 | return -EACCES; | |
1091 | ||
1092 | bprm->e_uid = current->euid; | |
1093 | bprm->e_gid = current->egid; | |
1094 | ||
0f7fc9e4 | 1095 | if(!(bprm->file->f_path.mnt->mnt_flags & MNT_NOSUID)) { |
1da177e4 LT |
1096 | /* Set-uid? */ |
1097 | if (mode & S_ISUID) { | |
1098 | current->personality &= ~PER_CLEAR_ON_SETID; | |
1099 | bprm->e_uid = inode->i_uid; | |
1100 | } | |
1101 | ||
1102 | /* Set-gid? */ | |
1103 | /* | |
1104 | * If setgid is set but no group execute bit then this | |
1105 | * is a candidate for mandatory locking, not a setgid | |
1106 | * executable. | |
1107 | */ | |
1108 | if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) { | |
1109 | current->personality &= ~PER_CLEAR_ON_SETID; | |
1110 | bprm->e_gid = inode->i_gid; | |
1111 | } | |
1112 | } | |
1113 | ||
1114 | /* fill in binprm security blob */ | |
1115 | retval = security_bprm_set(bprm); | |
1116 | if (retval) | |
1117 | return retval; | |
1118 | ||
1119 | memset(bprm->buf,0,BINPRM_BUF_SIZE); | |
1120 | return kernel_read(bprm->file,0,bprm->buf,BINPRM_BUF_SIZE); | |
1121 | } | |
1122 | ||
1123 | EXPORT_SYMBOL(prepare_binprm); | |
1124 | ||
858119e1 | 1125 | static int unsafe_exec(struct task_struct *p) |
1da177e4 LT |
1126 | { |
1127 | int unsafe = 0; | |
1128 | if (p->ptrace & PT_PTRACED) { | |
1129 | if (p->ptrace & PT_PTRACE_CAP) | |
1130 | unsafe |= LSM_UNSAFE_PTRACE_CAP; | |
1131 | else | |
1132 | unsafe |= LSM_UNSAFE_PTRACE; | |
1133 | } | |
1134 | if (atomic_read(&p->fs->count) > 1 || | |
1135 | atomic_read(&p->files->count) > 1 || | |
1136 | atomic_read(&p->sighand->count) > 1) | |
1137 | unsafe |= LSM_UNSAFE_SHARE; | |
1138 | ||
1139 | return unsafe; | |
1140 | } | |
1141 | ||
1142 | void compute_creds(struct linux_binprm *bprm) | |
1143 | { | |
1144 | int unsafe; | |
1145 | ||
d2d56c5f | 1146 | if (bprm->e_uid != current->uid) { |
1da177e4 | 1147 | suid_keys(current); |
d2d56c5f MH |
1148 | current->pdeath_signal = 0; |
1149 | } | |
1da177e4 LT |
1150 | exec_keys(current); |
1151 | ||
1152 | task_lock(current); | |
1153 | unsafe = unsafe_exec(current); | |
1154 | security_bprm_apply_creds(bprm, unsafe); | |
1155 | task_unlock(current); | |
1156 | security_bprm_post_apply_creds(bprm); | |
1157 | } | |
1da177e4 LT |
1158 | EXPORT_SYMBOL(compute_creds); |
1159 | ||
4fc75ff4 NP |
1160 | /* |
1161 | * Arguments are '\0' separated strings found at the location bprm->p | |
1162 | * points to; chop off the first by relocating brpm->p to right after | |
1163 | * the first '\0' encountered. | |
1164 | */ | |
b6a2fea3 | 1165 | int remove_arg_zero(struct linux_binprm *bprm) |
1da177e4 | 1166 | { |
b6a2fea3 OW |
1167 | int ret = 0; |
1168 | unsigned long offset; | |
1169 | char *kaddr; | |
1170 | struct page *page; | |
4fc75ff4 | 1171 | |
b6a2fea3 OW |
1172 | if (!bprm->argc) |
1173 | return 0; | |
1da177e4 | 1174 | |
b6a2fea3 OW |
1175 | do { |
1176 | offset = bprm->p & ~PAGE_MASK; | |
1177 | page = get_arg_page(bprm, bprm->p, 0); | |
1178 | if (!page) { | |
1179 | ret = -EFAULT; | |
1180 | goto out; | |
1181 | } | |
1182 | kaddr = kmap_atomic(page, KM_USER0); | |
4fc75ff4 | 1183 | |
b6a2fea3 OW |
1184 | for (; offset < PAGE_SIZE && kaddr[offset]; |
1185 | offset++, bprm->p++) | |
1186 | ; | |
4fc75ff4 | 1187 | |
b6a2fea3 OW |
1188 | kunmap_atomic(kaddr, KM_USER0); |
1189 | put_arg_page(page); | |
4fc75ff4 | 1190 | |
b6a2fea3 OW |
1191 | if (offset == PAGE_SIZE) |
1192 | free_arg_page(bprm, (bprm->p >> PAGE_SHIFT) - 1); | |
1193 | } while (offset == PAGE_SIZE); | |
4fc75ff4 | 1194 | |
b6a2fea3 OW |
1195 | bprm->p++; |
1196 | bprm->argc--; | |
1197 | ret = 0; | |
4fc75ff4 | 1198 | |
b6a2fea3 OW |
1199 | out: |
1200 | return ret; | |
1da177e4 | 1201 | } |
1da177e4 LT |
1202 | EXPORT_SYMBOL(remove_arg_zero); |
1203 | ||
1204 | /* | |
1205 | * cycle the list of binary formats handler, until one recognizes the image | |
1206 | */ | |
1207 | int search_binary_handler(struct linux_binprm *bprm,struct pt_regs *regs) | |
1208 | { | |
1209 | int try,retval; | |
1210 | struct linux_binfmt *fmt; | |
1211 | #ifdef __alpha__ | |
1212 | /* handle /sbin/loader.. */ | |
1213 | { | |
1214 | struct exec * eh = (struct exec *) bprm->buf; | |
1215 | ||
1216 | if (!bprm->loader && eh->fh.f_magic == 0x183 && | |
1217 | (eh->fh.f_flags & 0x3000) == 0x3000) | |
1218 | { | |
1219 | struct file * file; | |
1220 | unsigned long loader; | |
1221 | ||
1222 | allow_write_access(bprm->file); | |
1223 | fput(bprm->file); | |
1224 | bprm->file = NULL; | |
1225 | ||
b6a2fea3 | 1226 | loader = bprm->vma->vm_end - sizeof(void *); |
1da177e4 LT |
1227 | |
1228 | file = open_exec("/sbin/loader"); | |
1229 | retval = PTR_ERR(file); | |
1230 | if (IS_ERR(file)) | |
1231 | return retval; | |
1232 | ||
1233 | /* Remember if the application is TASO. */ | |
1234 | bprm->sh_bang = eh->ah.entry < 0x100000000UL; | |
1235 | ||
1236 | bprm->file = file; | |
1237 | bprm->loader = loader; | |
1238 | retval = prepare_binprm(bprm); | |
1239 | if (retval<0) | |
1240 | return retval; | |
1241 | /* should call search_binary_handler recursively here, | |
1242 | but it does not matter */ | |
1243 | } | |
1244 | } | |
1245 | #endif | |
1246 | retval = security_bprm_check(bprm); | |
1247 | if (retval) | |
1248 | return retval; | |
1249 | ||
1250 | /* kernel module loader fixup */ | |
1251 | /* so we don't try to load run modprobe in kernel space. */ | |
1252 | set_fs(USER_DS); | |
473ae30b AV |
1253 | |
1254 | retval = audit_bprm(bprm); | |
1255 | if (retval) | |
1256 | return retval; | |
1257 | ||
1da177e4 LT |
1258 | retval = -ENOENT; |
1259 | for (try=0; try<2; try++) { | |
1260 | read_lock(&binfmt_lock); | |
e4dc1b14 | 1261 | list_for_each_entry(fmt, &formats, lh) { |
1da177e4 LT |
1262 | int (*fn)(struct linux_binprm *, struct pt_regs *) = fmt->load_binary; |
1263 | if (!fn) | |
1264 | continue; | |
1265 | if (!try_module_get(fmt->module)) | |
1266 | continue; | |
1267 | read_unlock(&binfmt_lock); | |
1268 | retval = fn(bprm, regs); | |
1269 | if (retval >= 0) { | |
1270 | put_binfmt(fmt); | |
1271 | allow_write_access(bprm->file); | |
1272 | if (bprm->file) | |
1273 | fput(bprm->file); | |
1274 | bprm->file = NULL; | |
1275 | current->did_exec = 1; | |
9f46080c | 1276 | proc_exec_connector(current); |
1da177e4 LT |
1277 | return retval; |
1278 | } | |
1279 | read_lock(&binfmt_lock); | |
1280 | put_binfmt(fmt); | |
1281 | if (retval != -ENOEXEC || bprm->mm == NULL) | |
1282 | break; | |
1283 | if (!bprm->file) { | |
1284 | read_unlock(&binfmt_lock); | |
1285 | return retval; | |
1286 | } | |
1287 | } | |
1288 | read_unlock(&binfmt_lock); | |
1289 | if (retval != -ENOEXEC || bprm->mm == NULL) { | |
1290 | break; | |
1291 | #ifdef CONFIG_KMOD | |
1292 | }else{ | |
1293 | #define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e)) | |
1294 | if (printable(bprm->buf[0]) && | |
1295 | printable(bprm->buf[1]) && | |
1296 | printable(bprm->buf[2]) && | |
1297 | printable(bprm->buf[3])) | |
1298 | break; /* -ENOEXEC */ | |
1299 | request_module("binfmt-%04x", *(unsigned short *)(&bprm->buf[2])); | |
1300 | #endif | |
1301 | } | |
1302 | } | |
1303 | return retval; | |
1304 | } | |
1305 | ||
1306 | EXPORT_SYMBOL(search_binary_handler); | |
1307 | ||
1308 | /* | |
1309 | * sys_execve() executes a new program. | |
1310 | */ | |
1311 | int do_execve(char * filename, | |
1312 | char __user *__user *argv, | |
1313 | char __user *__user *envp, | |
1314 | struct pt_regs * regs) | |
1315 | { | |
1316 | struct linux_binprm *bprm; | |
1317 | struct file *file; | |
bdf4c48a | 1318 | unsigned long env_p; |
1da177e4 | 1319 | int retval; |
1da177e4 LT |
1320 | |
1321 | retval = -ENOMEM; | |
11b0b5ab | 1322 | bprm = kzalloc(sizeof(*bprm), GFP_KERNEL); |
1da177e4 LT |
1323 | if (!bprm) |
1324 | goto out_ret; | |
1da177e4 LT |
1325 | |
1326 | file = open_exec(filename); | |
1327 | retval = PTR_ERR(file); | |
1328 | if (IS_ERR(file)) | |
1329 | goto out_kfree; | |
1330 | ||
1331 | sched_exec(); | |
1332 | ||
1da177e4 LT |
1333 | bprm->file = file; |
1334 | bprm->filename = filename; | |
1335 | bprm->interp = filename; | |
1da177e4 | 1336 | |
b6a2fea3 OW |
1337 | retval = bprm_mm_init(bprm); |
1338 | if (retval) | |
1339 | goto out_file; | |
1da177e4 | 1340 | |
b6a2fea3 | 1341 | bprm->argc = count(argv, MAX_ARG_STRINGS); |
1da177e4 LT |
1342 | if ((retval = bprm->argc) < 0) |
1343 | goto out_mm; | |
1344 | ||
b6a2fea3 | 1345 | bprm->envc = count(envp, MAX_ARG_STRINGS); |
1da177e4 LT |
1346 | if ((retval = bprm->envc) < 0) |
1347 | goto out_mm; | |
1348 | ||
1349 | retval = security_bprm_alloc(bprm); | |
1350 | if (retval) | |
1351 | goto out; | |
1352 | ||
1353 | retval = prepare_binprm(bprm); | |
1354 | if (retval < 0) | |
1355 | goto out; | |
1356 | ||
1357 | retval = copy_strings_kernel(1, &bprm->filename, bprm); | |
1358 | if (retval < 0) | |
1359 | goto out; | |
1360 | ||
1361 | bprm->exec = bprm->p; | |
1362 | retval = copy_strings(bprm->envc, envp, bprm); | |
1363 | if (retval < 0) | |
1364 | goto out; | |
1365 | ||
bdf4c48a | 1366 | env_p = bprm->p; |
1da177e4 LT |
1367 | retval = copy_strings(bprm->argc, argv, bprm); |
1368 | if (retval < 0) | |
1369 | goto out; | |
bdf4c48a | 1370 | bprm->argv_len = env_p - bprm->p; |
1da177e4 LT |
1371 | |
1372 | retval = search_binary_handler(bprm,regs); | |
1373 | if (retval >= 0) { | |
1da177e4 | 1374 | /* execve success */ |
b6a2fea3 | 1375 | free_arg_pages(bprm); |
1da177e4 LT |
1376 | security_bprm_free(bprm); |
1377 | acct_update_integrals(current); | |
1da177e4 LT |
1378 | kfree(bprm); |
1379 | return retval; | |
1380 | } | |
1381 | ||
1382 | out: | |
b6a2fea3 | 1383 | free_arg_pages(bprm); |
1da177e4 LT |
1384 | if (bprm->security) |
1385 | security_bprm_free(bprm); | |
1386 | ||
1387 | out_mm: | |
1388 | if (bprm->mm) | |
b6a2fea3 | 1389 | mmput (bprm->mm); |
1da177e4 LT |
1390 | |
1391 | out_file: | |
1392 | if (bprm->file) { | |
1393 | allow_write_access(bprm->file); | |
1394 | fput(bprm->file); | |
1395 | } | |
1da177e4 LT |
1396 | out_kfree: |
1397 | kfree(bprm); | |
1398 | ||
1399 | out_ret: | |
1400 | return retval; | |
1401 | } | |
1402 | ||
1403 | int set_binfmt(struct linux_binfmt *new) | |
1404 | { | |
1405 | struct linux_binfmt *old = current->binfmt; | |
1406 | ||
1407 | if (new) { | |
1408 | if (!try_module_get(new->module)) | |
1409 | return -1; | |
1410 | } | |
1411 | current->binfmt = new; | |
1412 | if (old) | |
1413 | module_put(old->module); | |
1414 | return 0; | |
1415 | } | |
1416 | ||
1417 | EXPORT_SYMBOL(set_binfmt); | |
1418 | ||
1da177e4 LT |
1419 | /* format_corename will inspect the pattern parameter, and output a |
1420 | * name into corename, which must have space for at least | |
1421 | * CORENAME_MAX_SIZE bytes plus one byte for the zero terminator. | |
1422 | */ | |
c4bbafda | 1423 | static int format_corename(char *corename, const char *pattern, long signr) |
1da177e4 LT |
1424 | { |
1425 | const char *pat_ptr = pattern; | |
1426 | char *out_ptr = corename; | |
1427 | char *const out_end = corename + CORENAME_MAX_SIZE; | |
1428 | int rc; | |
1429 | int pid_in_pattern = 0; | |
c4bbafda AC |
1430 | int ispipe = 0; |
1431 | ||
1432 | if (*pattern == '|') | |
1433 | ispipe = 1; | |
1da177e4 LT |
1434 | |
1435 | /* Repeat as long as we have more pattern to process and more output | |
1436 | space */ | |
1437 | while (*pat_ptr) { | |
1438 | if (*pat_ptr != '%') { | |
1439 | if (out_ptr == out_end) | |
1440 | goto out; | |
1441 | *out_ptr++ = *pat_ptr++; | |
1442 | } else { | |
1443 | switch (*++pat_ptr) { | |
1444 | case 0: | |
1445 | goto out; | |
1446 | /* Double percent, output one percent */ | |
1447 | case '%': | |
1448 | if (out_ptr == out_end) | |
1449 | goto out; | |
1450 | *out_ptr++ = '%'; | |
1451 | break; | |
1452 | /* pid */ | |
1453 | case 'p': | |
1454 | pid_in_pattern = 1; | |
1455 | rc = snprintf(out_ptr, out_end - out_ptr, | |
1456 | "%d", current->tgid); | |
1457 | if (rc > out_end - out_ptr) | |
1458 | goto out; | |
1459 | out_ptr += rc; | |
1460 | break; | |
1461 | /* uid */ | |
1462 | case 'u': | |
1463 | rc = snprintf(out_ptr, out_end - out_ptr, | |
1464 | "%d", current->uid); | |
1465 | if (rc > out_end - out_ptr) | |
1466 | goto out; | |
1467 | out_ptr += rc; | |
1468 | break; | |
1469 | /* gid */ | |
1470 | case 'g': | |
1471 | rc = snprintf(out_ptr, out_end - out_ptr, | |
1472 | "%d", current->gid); | |
1473 | if (rc > out_end - out_ptr) | |
1474 | goto out; | |
1475 | out_ptr += rc; | |
1476 | break; | |
1477 | /* signal that caused the coredump */ | |
1478 | case 's': | |
1479 | rc = snprintf(out_ptr, out_end - out_ptr, | |
1480 | "%ld", signr); | |
1481 | if (rc > out_end - out_ptr) | |
1482 | goto out; | |
1483 | out_ptr += rc; | |
1484 | break; | |
1485 | /* UNIX time of coredump */ | |
1486 | case 't': { | |
1487 | struct timeval tv; | |
1488 | do_gettimeofday(&tv); | |
1489 | rc = snprintf(out_ptr, out_end - out_ptr, | |
1490 | "%lu", tv.tv_sec); | |
1491 | if (rc > out_end - out_ptr) | |
1492 | goto out; | |
1493 | out_ptr += rc; | |
1494 | break; | |
1495 | } | |
1496 | /* hostname */ | |
1497 | case 'h': | |
1498 | down_read(&uts_sem); | |
1499 | rc = snprintf(out_ptr, out_end - out_ptr, | |
e9ff3990 | 1500 | "%s", utsname()->nodename); |
1da177e4 LT |
1501 | up_read(&uts_sem); |
1502 | if (rc > out_end - out_ptr) | |
1503 | goto out; | |
1504 | out_ptr += rc; | |
1505 | break; | |
1506 | /* executable */ | |
1507 | case 'e': | |
1508 | rc = snprintf(out_ptr, out_end - out_ptr, | |
1509 | "%s", current->comm); | |
1510 | if (rc > out_end - out_ptr) | |
1511 | goto out; | |
1512 | out_ptr += rc; | |
1513 | break; | |
74aadce9 NH |
1514 | /* core limit size */ |
1515 | case 'c': | |
1516 | rc = snprintf(out_ptr, out_end - out_ptr, | |
1517 | "%lu", current->signal->rlim[RLIMIT_CORE].rlim_cur); | |
1518 | if (rc > out_end - out_ptr) | |
1519 | goto out; | |
1520 | out_ptr += rc; | |
1521 | break; | |
1da177e4 LT |
1522 | default: |
1523 | break; | |
1524 | } | |
1525 | ++pat_ptr; | |
1526 | } | |
1527 | } | |
1528 | /* Backward compatibility with core_uses_pid: | |
1529 | * | |
1530 | * If core_pattern does not include a %p (as is the default) | |
1531 | * and core_uses_pid is set, then .%pid will be appended to | |
c4bbafda AC |
1532 | * the filename. Do not do this for piped commands. */ |
1533 | if (!ispipe && !pid_in_pattern | |
1da177e4 LT |
1534 | && (core_uses_pid || atomic_read(¤t->mm->mm_users) != 1)) { |
1535 | rc = snprintf(out_ptr, out_end - out_ptr, | |
1536 | ".%d", current->tgid); | |
1537 | if (rc > out_end - out_ptr) | |
1538 | goto out; | |
1539 | out_ptr += rc; | |
1540 | } | |
c4bbafda | 1541 | out: |
1da177e4 | 1542 | *out_ptr = 0; |
c4bbafda | 1543 | return ispipe; |
1da177e4 LT |
1544 | } |
1545 | ||
d5f70c00 | 1546 | static void zap_process(struct task_struct *start) |
aceecc04 ON |
1547 | { |
1548 | struct task_struct *t; | |
281de339 | 1549 | |
d5f70c00 ON |
1550 | start->signal->flags = SIGNAL_GROUP_EXIT; |
1551 | start->signal->group_stop_count = 0; | |
aceecc04 ON |
1552 | |
1553 | t = start; | |
1554 | do { | |
1555 | if (t != current && t->mm) { | |
1556 | t->mm->core_waiters++; | |
281de339 ON |
1557 | sigaddset(&t->pending.signal, SIGKILL); |
1558 | signal_wake_up(t, 1); | |
aceecc04 ON |
1559 | } |
1560 | } while ((t = next_thread(t)) != start); | |
1561 | } | |
1562 | ||
dcf560c5 ON |
1563 | static inline int zap_threads(struct task_struct *tsk, struct mm_struct *mm, |
1564 | int exit_code) | |
1da177e4 LT |
1565 | { |
1566 | struct task_struct *g, *p; | |
5debfa6d | 1567 | unsigned long flags; |
dcf560c5 ON |
1568 | int err = -EAGAIN; |
1569 | ||
1570 | spin_lock_irq(&tsk->sighand->siglock); | |
1571 | if (!(tsk->signal->flags & SIGNAL_GROUP_EXIT)) { | |
dcf560c5 | 1572 | tsk->signal->group_exit_code = exit_code; |
5debfa6d | 1573 | zap_process(tsk); |
dcf560c5 | 1574 | err = 0; |
1da177e4 | 1575 | } |
dcf560c5 ON |
1576 | spin_unlock_irq(&tsk->sighand->siglock); |
1577 | if (err) | |
1578 | return err; | |
1da177e4 | 1579 | |
5debfa6d ON |
1580 | if (atomic_read(&mm->mm_users) == mm->core_waiters + 1) |
1581 | goto done; | |
1582 | ||
7b1c6154 | 1583 | rcu_read_lock(); |
aceecc04 | 1584 | for_each_process(g) { |
5debfa6d ON |
1585 | if (g == tsk->group_leader) |
1586 | continue; | |
1587 | ||
aceecc04 ON |
1588 | p = g; |
1589 | do { | |
1590 | if (p->mm) { | |
5debfa6d ON |
1591 | if (p->mm == mm) { |
1592 | /* | |
1593 | * p->sighand can't disappear, but | |
1594 | * may be changed by de_thread() | |
1595 | */ | |
1596 | lock_task_sighand(p, &flags); | |
d5f70c00 | 1597 | zap_process(p); |
5debfa6d ON |
1598 | unlock_task_sighand(p, &flags); |
1599 | } | |
aceecc04 ON |
1600 | break; |
1601 | } | |
1602 | } while ((p = next_thread(p)) != g); | |
1603 | } | |
7b1c6154 | 1604 | rcu_read_unlock(); |
5debfa6d | 1605 | done: |
dcf560c5 | 1606 | return mm->core_waiters; |
1da177e4 LT |
1607 | } |
1608 | ||
dcf560c5 | 1609 | static int coredump_wait(int exit_code) |
1da177e4 | 1610 | { |
dcf560c5 ON |
1611 | struct task_struct *tsk = current; |
1612 | struct mm_struct *mm = tsk->mm; | |
1613 | struct completion startup_done; | |
1614 | struct completion *vfork_done; | |
2384f55f | 1615 | int core_waiters; |
1da177e4 | 1616 | |
dcf560c5 ON |
1617 | init_completion(&mm->core_done); |
1618 | init_completion(&startup_done); | |
1da177e4 LT |
1619 | mm->core_startup_done = &startup_done; |
1620 | ||
dcf560c5 | 1621 | core_waiters = zap_threads(tsk, mm, exit_code); |
2384f55f ON |
1622 | up_write(&mm->mmap_sem); |
1623 | ||
dcf560c5 ON |
1624 | if (unlikely(core_waiters < 0)) |
1625 | goto fail; | |
1626 | ||
1627 | /* | |
1628 | * Make sure nobody is waiting for us to release the VM, | |
1629 | * otherwise we can deadlock when we wait on each other | |
1630 | */ | |
1631 | vfork_done = tsk->vfork_done; | |
1632 | if (vfork_done) { | |
1633 | tsk->vfork_done = NULL; | |
1634 | complete(vfork_done); | |
1635 | } | |
1636 | ||
2384f55f | 1637 | if (core_waiters) |
1da177e4 | 1638 | wait_for_completion(&startup_done); |
dcf560c5 | 1639 | fail: |
1da177e4 | 1640 | BUG_ON(mm->core_waiters); |
dcf560c5 | 1641 | return core_waiters; |
1da177e4 LT |
1642 | } |
1643 | ||
6c5d5238 KH |
1644 | /* |
1645 | * set_dumpable converts traditional three-value dumpable to two flags and | |
1646 | * stores them into mm->flags. It modifies lower two bits of mm->flags, but | |
1647 | * these bits are not changed atomically. So get_dumpable can observe the | |
1648 | * intermediate state. To avoid doing unexpected behavior, get get_dumpable | |
1649 | * return either old dumpable or new one by paying attention to the order of | |
1650 | * modifying the bits. | |
1651 | * | |
1652 | * dumpable | mm->flags (binary) | |
1653 | * old new | initial interim final | |
1654 | * ---------+----------------------- | |
1655 | * 0 1 | 00 01 01 | |
1656 | * 0 2 | 00 10(*) 11 | |
1657 | * 1 0 | 01 00 00 | |
1658 | * 1 2 | 01 11 11 | |
1659 | * 2 0 | 11 10(*) 00 | |
1660 | * 2 1 | 11 11 01 | |
1661 | * | |
1662 | * (*) get_dumpable regards interim value of 10 as 11. | |
1663 | */ | |
1664 | void set_dumpable(struct mm_struct *mm, int value) | |
1665 | { | |
1666 | switch (value) { | |
1667 | case 0: | |
1668 | clear_bit(MMF_DUMPABLE, &mm->flags); | |
1669 | smp_wmb(); | |
1670 | clear_bit(MMF_DUMP_SECURELY, &mm->flags); | |
1671 | break; | |
1672 | case 1: | |
1673 | set_bit(MMF_DUMPABLE, &mm->flags); | |
1674 | smp_wmb(); | |
1675 | clear_bit(MMF_DUMP_SECURELY, &mm->flags); | |
1676 | break; | |
1677 | case 2: | |
1678 | set_bit(MMF_DUMP_SECURELY, &mm->flags); | |
1679 | smp_wmb(); | |
1680 | set_bit(MMF_DUMPABLE, &mm->flags); | |
1681 | break; | |
1682 | } | |
1683 | } | |
1684 | EXPORT_SYMBOL_GPL(set_dumpable); | |
1685 | ||
1686 | int get_dumpable(struct mm_struct *mm) | |
1687 | { | |
1688 | int ret; | |
1689 | ||
1690 | ret = mm->flags & 0x3; | |
1691 | return (ret >= 2) ? 2 : ret; | |
1692 | } | |
1693 | ||
1da177e4 LT |
1694 | int do_coredump(long signr, int exit_code, struct pt_regs * regs) |
1695 | { | |
1696 | char corename[CORENAME_MAX_SIZE + 1]; | |
1697 | struct mm_struct *mm = current->mm; | |
1698 | struct linux_binfmt * binfmt; | |
1699 | struct inode * inode; | |
1700 | struct file * file; | |
1701 | int retval = 0; | |
d6e71144 AC |
1702 | int fsuid = current->fsuid; |
1703 | int flag = 0; | |
d025c9db | 1704 | int ispipe = 0; |
7dc0b22e | 1705 | unsigned long core_limit = current->signal->rlim[RLIMIT_CORE].rlim_cur; |
74aadce9 NH |
1706 | char **helper_argv = NULL; |
1707 | int helper_argc = 0; | |
1708 | char *delimit; | |
1da177e4 | 1709 | |
0a4ff8c2 SG |
1710 | audit_core_dumps(signr); |
1711 | ||
1da177e4 LT |
1712 | binfmt = current->binfmt; |
1713 | if (!binfmt || !binfmt->core_dump) | |
1714 | goto fail; | |
1715 | down_write(&mm->mmap_sem); | |
6c5d5238 | 1716 | if (!get_dumpable(mm)) { |
1da177e4 LT |
1717 | up_write(&mm->mmap_sem); |
1718 | goto fail; | |
1719 | } | |
d6e71144 AC |
1720 | |
1721 | /* | |
1722 | * We cannot trust fsuid as being the "true" uid of the | |
1723 | * process nor do we know its entire history. We only know it | |
1724 | * was tainted so we dump it as root in mode 2. | |
1725 | */ | |
6c5d5238 | 1726 | if (get_dumpable(mm) == 2) { /* Setuid core dump mode */ |
d6e71144 AC |
1727 | flag = O_EXCL; /* Stop rewrite attacks */ |
1728 | current->fsuid = 0; /* Dump root private */ | |
1729 | } | |
6c5d5238 | 1730 | set_dumpable(mm, 0); |
1291cf41 | 1731 | |
dcf560c5 ON |
1732 | retval = coredump_wait(exit_code); |
1733 | if (retval < 0) | |
1291cf41 | 1734 | goto fail; |
1da177e4 LT |
1735 | |
1736 | /* | |
1737 | * Clear any false indication of pending signals that might | |
1738 | * be seen by the filesystem code called to write the core file. | |
1739 | */ | |
1da177e4 LT |
1740 | clear_thread_flag(TIF_SIGPENDING); |
1741 | ||
1da177e4 LT |
1742 | /* |
1743 | * lock_kernel() because format_corename() is controlled by sysctl, which | |
1744 | * uses lock_kernel() | |
1745 | */ | |
1746 | lock_kernel(); | |
c4bbafda | 1747 | ispipe = format_corename(corename, core_pattern, signr); |
1da177e4 | 1748 | unlock_kernel(); |
7dc0b22e NH |
1749 | /* |
1750 | * Don't bother to check the RLIMIT_CORE value if core_pattern points | |
1751 | * to a pipe. Since we're not writing directly to the filesystem | |
1752 | * RLIMIT_CORE doesn't really apply, as no actual core file will be | |
1753 | * created unless the pipe reader choses to write out the core file | |
1754 | * at which point file size limits and permissions will be imposed | |
1755 | * as it does with any other process | |
1756 | */ | |
74aadce9 | 1757 | if ((!ispipe) && (core_limit < binfmt->min_coredump)) |
7dc0b22e NH |
1758 | goto fail_unlock; |
1759 | ||
c4bbafda | 1760 | if (ispipe) { |
74aadce9 NH |
1761 | helper_argv = argv_split(GFP_KERNEL, corename+1, &helper_argc); |
1762 | /* Terminate the string before the first option */ | |
1763 | delimit = strchr(corename, ' '); | |
1764 | if (delimit) | |
1765 | *delimit = '\0'; | |
32321137 NH |
1766 | delimit = strrchr(helper_argv[0], '/'); |
1767 | if (delimit) | |
1768 | delimit++; | |
1769 | else | |
1770 | delimit = helper_argv[0]; | |
1771 | if (!strcmp(delimit, current->comm)) { | |
1772 | printk(KERN_NOTICE "Recursive core dump detected, " | |
1773 | "aborting\n"); | |
1774 | goto fail_unlock; | |
1775 | } | |
1776 | ||
1777 | core_limit = RLIM_INFINITY; | |
1778 | ||
d025c9db | 1779 | /* SIGPIPE can happen, but it's just never processed */ |
32321137 NH |
1780 | if (call_usermodehelper_pipe(corename+1, helper_argv, NULL, |
1781 | &file)) { | |
d025c9db AK |
1782 | printk(KERN_INFO "Core dump to %s pipe failed\n", |
1783 | corename); | |
1784 | goto fail_unlock; | |
1785 | } | |
d025c9db AK |
1786 | } else |
1787 | file = filp_open(corename, | |
6d4df677 AD |
1788 | O_CREAT | 2 | O_NOFOLLOW | O_LARGEFILE | flag, |
1789 | 0600); | |
1da177e4 LT |
1790 | if (IS_ERR(file)) |
1791 | goto fail_unlock; | |
0f7fc9e4 | 1792 | inode = file->f_path.dentry->d_inode; |
1da177e4 LT |
1793 | if (inode->i_nlink > 1) |
1794 | goto close_fail; /* multiple links - don't dump */ | |
0f7fc9e4 | 1795 | if (!ispipe && d_unhashed(file->f_path.dentry)) |
1da177e4 LT |
1796 | goto close_fail; |
1797 | ||
d025c9db AK |
1798 | /* AK: actually i see no reason to not allow this for named pipes etc., |
1799 | but keep the previous behaviour for now. */ | |
1800 | if (!ispipe && !S_ISREG(inode->i_mode)) | |
1da177e4 LT |
1801 | goto close_fail; |
1802 | if (!file->f_op) | |
1803 | goto close_fail; | |
1804 | if (!file->f_op->write) | |
1805 | goto close_fail; | |
0f7fc9e4 | 1806 | if (!ispipe && do_truncate(file->f_path.dentry, 0, 0, file) != 0) |
1da177e4 LT |
1807 | goto close_fail; |
1808 | ||
7dc0b22e | 1809 | retval = binfmt->core_dump(signr, regs, file, core_limit); |
1da177e4 LT |
1810 | |
1811 | if (retval) | |
1812 | current->signal->group_exit_code |= 0x80; | |
1813 | close_fail: | |
1814 | filp_close(file, NULL); | |
1815 | fail_unlock: | |
74aadce9 NH |
1816 | if (helper_argv) |
1817 | argv_free(helper_argv); | |
1818 | ||
d6e71144 | 1819 | current->fsuid = fsuid; |
1da177e4 LT |
1820 | complete_all(&mm->core_done); |
1821 | fail: | |
1822 | return retval; | |
1823 | } |