blob: a44b142fb4607baee2d832b1056103b5d157e5a7 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
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
Linus Torvalds1da177e2005-04-16 15:20:36 -070025#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>
Neil Horman74aadce2007-10-16 23:26:35 -070032#include <linux/string.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070033#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>
Sukadev Bhattiprolu84d73782006-12-08 02:38:01 -080042#include <linux/pid_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070043#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>
Jay Lan8f0ab512006-09-30 23:28:59 -070051#include <linux/tsacct_kern.h>
Matt Helsley9f460802005-11-07 00:59:16 -080052#include <linux/cn_proc.h>
Al Viro473ae302006-04-26 14:04:08 -040053#include <linux/audit.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070054
55#include <asm/uaccess.h>
56#include <asm/mmu_context.h>
Ollie Wildb6a2fea2007-07-19 01:48:16 -070057#include <asm/tlb.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070058
59#ifdef CONFIG_KMOD
60#include <linux/kmod.h>
61#endif
62
63int core_uses_pid;
Dan Aloni71ce92f2007-05-16 22:11:16 -070064char core_pattern[CORENAME_MAX_SIZE] = "core";
Alan Coxd6e71142005-06-23 00:09:43 -070065int suid_dumpable = 0;
66
Linus Torvalds1da177e2005-04-16 15:20:36 -070067/* The maximal length of core_pattern is also specified in sysctl.c */
68
Alexey Dobriyane4dc1b12007-10-16 23:26:03 -070069static LIST_HEAD(formats);
Linus Torvalds1da177e2005-04-16 15:20:36 -070070static DEFINE_RWLOCK(binfmt_lock);
71
72int register_binfmt(struct linux_binfmt * fmt)
73{
Linus Torvalds1da177e2005-04-16 15:20:36 -070074 if (!fmt)
75 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -070076 write_lock(&binfmt_lock);
Alexey Dobriyane4dc1b12007-10-16 23:26:03 -070077 list_add(&fmt->lh, &formats);
Linus Torvalds1da177e2005-04-16 15:20:36 -070078 write_unlock(&binfmt_lock);
79 return 0;
80}
81
82EXPORT_SYMBOL(register_binfmt);
83
Alexey Dobriyanf6b450d2007-10-16 23:26:04 -070084void unregister_binfmt(struct linux_binfmt * fmt)
Linus Torvalds1da177e2005-04-16 15:20:36 -070085{
Linus Torvalds1da177e2005-04-16 15:20:36 -070086 write_lock(&binfmt_lock);
Alexey Dobriyane4dc1b12007-10-16 23:26:03 -070087 list_del(&fmt->lh);
Linus Torvalds1da177e2005-04-16 15:20:36 -070088 write_unlock(&binfmt_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -070089}
90
91EXPORT_SYMBOL(unregister_binfmt);
92
93static inline void put_binfmt(struct linux_binfmt * fmt)
94{
95 module_put(fmt->module);
96}
97
98/*
99 * Note that a shared library must be both readable and executable due to
100 * security reasons.
101 *
102 * Also note that we take the address to load from from the file itself.
103 */
104asmlinkage long sys_uselib(const char __user * library)
105{
106 struct file * file;
107 struct nameidata nd;
108 int error;
109
Oleg Drokinb5005312006-03-25 03:07:01 -0800110 error = __user_path_lookup_open(library, LOOKUP_FOLLOW, &nd, FMODE_READ|FMODE_EXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700111 if (error)
112 goto out;
113
114 error = -EINVAL;
Jan Blunck4ac91372008-02-14 19:34:32 -0800115 if (!S_ISREG(nd.path.dentry->d_inode->i_mode))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116 goto exit;
117
Christoph Hellwige4543ed2005-11-08 21:35:04 -0800118 error = vfs_permission(&nd, MAY_READ | MAY_EXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700119 if (error)
120 goto exit;
121
Andi Kleenabe8be32008-02-08 04:20:23 -0800122 file = nameidata_to_filp(&nd, O_RDONLY|O_LARGEFILE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700123 error = PTR_ERR(file);
124 if (IS_ERR(file))
125 goto out;
126
127 error = -ENOEXEC;
128 if(file->f_op) {
129 struct linux_binfmt * fmt;
130
131 read_lock(&binfmt_lock);
Alexey Dobriyane4dc1b12007-10-16 23:26:03 -0700132 list_for_each_entry(fmt, &formats, lh) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700133 if (!fmt->load_shlib)
134 continue;
135 if (!try_module_get(fmt->module))
136 continue;
137 read_unlock(&binfmt_lock);
138 error = fmt->load_shlib(file);
139 read_lock(&binfmt_lock);
140 put_binfmt(fmt);
141 if (error != -ENOEXEC)
142 break;
143 }
144 read_unlock(&binfmt_lock);
145 }
146 fput(file);
147out:
148 return error;
149exit:
Trond Myklebust834f2a42005-10-18 14:20:16 -0700150 release_open_intent(&nd);
Jan Blunck1d957f92008-02-14 19:34:35 -0800151 path_put(&nd.path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152 goto out;
153}
154
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700155#ifdef CONFIG_MMU
156
157static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos,
158 int write)
159{
160 struct page *page;
161 int ret;
162
163#ifdef CONFIG_STACK_GROWSUP
164 if (write) {
165 ret = expand_stack_downwards(bprm->vma, pos);
166 if (ret < 0)
167 return NULL;
168 }
169#endif
170 ret = get_user_pages(current, bprm->mm, pos,
171 1, write, 1, &page, NULL);
172 if (ret <= 0)
173 return NULL;
174
175 if (write) {
176 struct rlimit *rlim = current->signal->rlim;
177 unsigned long size = bprm->vma->vm_end - bprm->vma->vm_start;
178
179 /*
180 * Limit to 1/4-th the stack size for the argv+env strings.
181 * This ensures that:
182 * - the remaining binfmt code will not run out of stack space,
183 * - the program will have a reasonable amount of stack left
184 * to work from.
185 */
186 if (size > rlim[RLIMIT_STACK].rlim_cur / 4) {
187 put_page(page);
188 return NULL;
189 }
190 }
191
192 return page;
193}
194
195static void put_arg_page(struct page *page)
196{
197 put_page(page);
198}
199
200static void free_arg_page(struct linux_binprm *bprm, int i)
201{
202}
203
204static void free_arg_pages(struct linux_binprm *bprm)
205{
206}
207
208static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos,
209 struct page *page)
210{
211 flush_cache_page(bprm->vma, pos, page_to_pfn(page));
212}
213
214static int __bprm_mm_init(struct linux_binprm *bprm)
215{
216 int err = -ENOMEM;
217 struct vm_area_struct *vma = NULL;
218 struct mm_struct *mm = bprm->mm;
219
220 bprm->vma = vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL);
221 if (!vma)
222 goto err;
223
224 down_write(&mm->mmap_sem);
225 vma->vm_mm = mm;
226
227 /*
228 * Place the stack at the largest stack address the architecture
229 * supports. Later, we'll move this to an appropriate place. We don't
230 * use STACK_TOP because that can depend on attributes which aren't
231 * configured yet.
232 */
233 vma->vm_end = STACK_TOP_MAX;
234 vma->vm_start = vma->vm_end - PAGE_SIZE;
235
236 vma->vm_flags = VM_STACK_FLAGS;
Coly Li3ed75eb2007-10-18 23:39:15 -0700237 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700238 err = insert_vm_struct(mm, vma);
239 if (err) {
240 up_write(&mm->mmap_sem);
241 goto err;
242 }
243
244 mm->stack_vm = mm->total_vm = 1;
245 up_write(&mm->mmap_sem);
246
247 bprm->p = vma->vm_end - sizeof(void *);
248
249 return 0;
250
251err:
252 if (vma) {
253 bprm->vma = NULL;
254 kmem_cache_free(vm_area_cachep, vma);
255 }
256
257 return err;
258}
259
260static bool valid_arg_len(struct linux_binprm *bprm, long len)
261{
262 return len <= MAX_ARG_STRLEN;
263}
264
265#else
266
267static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos,
268 int write)
269{
270 struct page *page;
271
272 page = bprm->page[pos / PAGE_SIZE];
273 if (!page && write) {
274 page = alloc_page(GFP_HIGHUSER|__GFP_ZERO);
275 if (!page)
276 return NULL;
277 bprm->page[pos / PAGE_SIZE] = page;
278 }
279
280 return page;
281}
282
283static void put_arg_page(struct page *page)
284{
285}
286
287static void free_arg_page(struct linux_binprm *bprm, int i)
288{
289 if (bprm->page[i]) {
290 __free_page(bprm->page[i]);
291 bprm->page[i] = NULL;
292 }
293}
294
295static void free_arg_pages(struct linux_binprm *bprm)
296{
297 int i;
298
299 for (i = 0; i < MAX_ARG_PAGES; i++)
300 free_arg_page(bprm, i);
301}
302
303static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos,
304 struct page *page)
305{
306}
307
308static int __bprm_mm_init(struct linux_binprm *bprm)
309{
310 bprm->p = PAGE_SIZE * MAX_ARG_PAGES - sizeof(void *);
311 return 0;
312}
313
314static bool valid_arg_len(struct linux_binprm *bprm, long len)
315{
316 return len <= bprm->p;
317}
318
319#endif /* CONFIG_MMU */
320
321/*
322 * Create a new mm_struct and populate it with a temporary stack
323 * vm_area_struct. We don't have enough context at this point to set the stack
324 * flags, permissions, and offset, so we use temporary values. We'll update
325 * them later in setup_arg_pages().
326 */
327int bprm_mm_init(struct linux_binprm *bprm)
328{
329 int err;
330 struct mm_struct *mm = NULL;
331
332 bprm->mm = mm = mm_alloc();
333 err = -ENOMEM;
334 if (!mm)
335 goto err;
336
337 err = init_new_context(current, mm);
338 if (err)
339 goto err;
340
341 err = __bprm_mm_init(bprm);
342 if (err)
343 goto err;
344
345 return 0;
346
347err:
348 if (mm) {
349 bprm->mm = NULL;
350 mmdrop(mm);
351 }
352
353 return err;
354}
355
Linus Torvalds1da177e2005-04-16 15:20:36 -0700356/*
357 * count() counts the number of strings in array ARGV.
358 */
359static int count(char __user * __user * argv, int max)
360{
361 int i = 0;
362
363 if (argv != NULL) {
364 for (;;) {
365 char __user * p;
366
367 if (get_user(p, argv))
368 return -EFAULT;
369 if (!p)
370 break;
371 argv++;
372 if(++i > max)
373 return -E2BIG;
374 cond_resched();
375 }
376 }
377 return i;
378}
379
380/*
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700381 * 'copy_strings()' copies argument/environment strings from the old
382 * processes's memory to the new process's stack. The call to get_user_pages()
383 * ensures the destination page is created and not swapped out.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700384 */
Adrian Bunk75c96f82005-05-05 16:16:09 -0700385static int copy_strings(int argc, char __user * __user * argv,
386 struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700387{
388 struct page *kmapped_page = NULL;
389 char *kaddr = NULL;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700390 unsigned long kpos = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700391 int ret;
392
393 while (argc-- > 0) {
394 char __user *str;
395 int len;
396 unsigned long pos;
397
398 if (get_user(str, argv+argc) ||
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700399 !(len = strnlen_user(str, MAX_ARG_STRLEN))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400 ret = -EFAULT;
401 goto out;
402 }
403
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700404 if (!valid_arg_len(bprm, len)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405 ret = -E2BIG;
406 goto out;
407 }
408
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700409 /* We're going to work our way backwords. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410 pos = bprm->p;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700411 str += len;
412 bprm->p -= len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700413
414 while (len > 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415 int offset, bytes_to_copy;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416
417 offset = pos % PAGE_SIZE;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700418 if (offset == 0)
419 offset = PAGE_SIZE;
420
421 bytes_to_copy = offset;
422 if (bytes_to_copy > len)
423 bytes_to_copy = len;
424
425 offset -= bytes_to_copy;
426 pos -= bytes_to_copy;
427 str -= bytes_to_copy;
428 len -= bytes_to_copy;
429
430 if (!kmapped_page || kpos != (pos & PAGE_MASK)) {
431 struct page *page;
432
433 page = get_arg_page(bprm, pos, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434 if (!page) {
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700435 ret = -E2BIG;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700436 goto out;
437 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700438
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700439 if (kmapped_page) {
440 flush_kernel_dcache_page(kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441 kunmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700442 put_arg_page(kmapped_page);
443 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444 kmapped_page = page;
445 kaddr = kmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700446 kpos = pos & PAGE_MASK;
447 flush_arg_page(bprm, kpos, kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448 }
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700449 if (copy_from_user(kaddr+offset, str, bytes_to_copy)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700450 ret = -EFAULT;
451 goto out;
452 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453 }
454 }
455 ret = 0;
456out:
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700457 if (kmapped_page) {
458 flush_kernel_dcache_page(kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700459 kunmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700460 put_arg_page(kmapped_page);
461 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700462 return ret;
463}
464
465/*
466 * Like copy_strings, but get argv and its values from kernel memory.
467 */
468int copy_strings_kernel(int argc,char ** argv, struct linux_binprm *bprm)
469{
470 int r;
471 mm_segment_t oldfs = get_fs();
472 set_fs(KERNEL_DS);
473 r = copy_strings(argc, (char __user * __user *)argv, bprm);
474 set_fs(oldfs);
475 return r;
476}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477EXPORT_SYMBOL(copy_strings_kernel);
478
479#ifdef CONFIG_MMU
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700480
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481/*
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700482 * During bprm_mm_init(), we create a temporary stack at STACK_TOP_MAX. Once
483 * the binfmt code determines where the new stack should reside, we shift it to
484 * its final location. The process proceeds as follows:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485 *
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700486 * 1) Use shift to calculate the new vma endpoints.
487 * 2) Extend vma to cover both the old and new ranges. This ensures the
488 * arguments passed to subsequent functions are consistent.
489 * 3) Move vma's page tables to the new range.
490 * 4) Free up any cleared pgd range.
491 * 5) Shrink the vma to cover only the new range.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492 */
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700493static int shift_arg_pages(struct vm_area_struct *vma, unsigned long shift)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494{
495 struct mm_struct *mm = vma->vm_mm;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700496 unsigned long old_start = vma->vm_start;
497 unsigned long old_end = vma->vm_end;
498 unsigned long length = old_end - old_start;
499 unsigned long new_start = old_start - shift;
500 unsigned long new_end = old_end - shift;
501 struct mmu_gather *tlb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700503 BUG_ON(new_start > new_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700505 /*
506 * ensure there are no vmas between where we want to go
507 * and where we are
508 */
509 if (vma != find_vma(mm, new_start))
510 return -EFAULT;
511
512 /*
513 * cover the whole range: [new_start, old_end)
514 */
515 vma_adjust(vma, new_start, old_end, vma->vm_pgoff, NULL);
516
517 /*
518 * move the page tables downwards, on failure we rely on
519 * process cleanup to remove whatever mess we made.
520 */
521 if (length != move_page_tables(vma, old_start,
522 vma, new_start, length))
523 return -ENOMEM;
524
525 lru_add_drain();
526 tlb = tlb_gather_mmu(mm, 0);
527 if (new_end > old_start) {
528 /*
529 * when the old and new regions overlap clear from new_end.
530 */
531 free_pgd_range(&tlb, new_end, old_end, new_end,
532 vma->vm_next ? vma->vm_next->vm_start : 0);
533 } else {
534 /*
535 * otherwise, clean from old_start; this is done to not touch
536 * the address space in [new_end, old_start) some architectures
537 * have constraints on va-space that make this illegal (IA64) -
538 * for the others its just a little faster.
539 */
540 free_pgd_range(&tlb, old_start, old_end, new_end,
541 vma->vm_next ? vma->vm_next->vm_start : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700542 }
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700543 tlb_finish_mmu(tlb, new_end, old_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700545 /*
546 * shrink the vma to just the new range.
547 */
548 vma_adjust(vma, new_start, new_end, vma->vm_pgoff, NULL);
549
550 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551}
552
553#define EXTRA_STACK_VM_PAGES 20 /* random */
554
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700555/*
556 * Finalizes the stack vm_area_struct. The flags and permissions are updated,
557 * the stack is optionally relocated, and some extra space is added.
558 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700559int setup_arg_pages(struct linux_binprm *bprm,
560 unsigned long stack_top,
561 int executable_stack)
562{
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700563 unsigned long ret;
564 unsigned long stack_shift;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565 struct mm_struct *mm = current->mm;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700566 struct vm_area_struct *vma = bprm->vma;
567 struct vm_area_struct *prev = NULL;
568 unsigned long vm_flags;
569 unsigned long stack_base;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700570
571#ifdef CONFIG_STACK_GROWSUP
Linus Torvalds1da177e2005-04-16 15:20:36 -0700572 /* Limit stack size to 1GB */
573 stack_base = current->signal->rlim[RLIMIT_STACK].rlim_max;
574 if (stack_base > (1 << 30))
575 stack_base = 1 << 30;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700577 /* Make sure we didn't let the argument array grow too large. */
578 if (vma->vm_end - vma->vm_start > stack_base)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700579 return -ENOMEM;
580
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700581 stack_base = PAGE_ALIGN(stack_top - stack_base);
582
583 stack_shift = vma->vm_start - stack_base;
584 mm->arg_start = bprm->p - stack_shift;
585 bprm->p = vma->vm_end - stack_shift;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700586#else
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700587 stack_top = arch_align_stack(stack_top);
588 stack_top = PAGE_ALIGN(stack_top);
589 stack_shift = vma->vm_end - stack_top;
590
591 bprm->p -= stack_shift;
592 mm->arg_start = bprm->p;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700593#endif
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700594
595 if (bprm->loader)
596 bprm->loader -= stack_shift;
597 bprm->exec -= stack_shift;
598
599 down_write(&mm->mmap_sem);
600 vm_flags = vma->vm_flags;
601
602 /*
603 * Adjust stack execute permissions; explicitly enable for
604 * EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X and leave alone
605 * (arch default) otherwise.
606 */
607 if (unlikely(executable_stack == EXSTACK_ENABLE_X))
608 vm_flags |= VM_EXEC;
609 else if (executable_stack == EXSTACK_DISABLE_X)
610 vm_flags &= ~VM_EXEC;
611 vm_flags |= mm->def_flags;
612
613 ret = mprotect_fixup(vma, &prev, vma->vm_start, vma->vm_end,
614 vm_flags);
615 if (ret)
616 goto out_unlock;
617 BUG_ON(prev != vma);
618
619 /* Move stack pages down in memory. */
620 if (stack_shift) {
621 ret = shift_arg_pages(vma, stack_shift);
622 if (ret) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700623 up_write(&mm->mmap_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624 return ret;
625 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626 }
627
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700628#ifdef CONFIG_STACK_GROWSUP
629 stack_base = vma->vm_end + EXTRA_STACK_VM_PAGES * PAGE_SIZE;
630#else
631 stack_base = vma->vm_start - EXTRA_STACK_VM_PAGES * PAGE_SIZE;
632#endif
633 ret = expand_stack(vma, stack_base);
634 if (ret)
635 ret = -EFAULT;
636
637out_unlock:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700638 up_write(&mm->mmap_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700639 return 0;
640}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700641EXPORT_SYMBOL(setup_arg_pages);
642
Linus Torvalds1da177e2005-04-16 15:20:36 -0700643#endif /* CONFIG_MMU */
644
645struct file *open_exec(const char *name)
646{
647 struct nameidata nd;
648 int err;
649 struct file *file;
650
Oleg Drokinb5005312006-03-25 03:07:01 -0800651 err = path_lookup_open(AT_FDCWD, name, LOOKUP_FOLLOW, &nd, FMODE_READ|FMODE_EXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700652 file = ERR_PTR(err);
653
654 if (!err) {
Jan Blunck4ac91372008-02-14 19:34:32 -0800655 struct inode *inode = nd.path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700656 file = ERR_PTR(-EACCES);
Miklos Szeredi1a159dd2007-10-16 23:27:08 -0700657 if (S_ISREG(inode->i_mode)) {
Christoph Hellwige4543ed2005-11-08 21:35:04 -0800658 int err = vfs_permission(&nd, MAY_EXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659 file = ERR_PTR(err);
660 if (!err) {
Andi Kleenabe8be32008-02-08 04:20:23 -0800661 file = nameidata_to_filp(&nd,
662 O_RDONLY|O_LARGEFILE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663 if (!IS_ERR(file)) {
664 err = deny_write_access(file);
665 if (err) {
666 fput(file);
667 file = ERR_PTR(err);
668 }
669 }
670out:
671 return file;
672 }
673 }
Trond Myklebust834f2a42005-10-18 14:20:16 -0700674 release_open_intent(&nd);
Jan Blunck1d957f92008-02-14 19:34:35 -0800675 path_put(&nd.path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700676 }
677 goto out;
678}
679
680EXPORT_SYMBOL(open_exec);
681
682int 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
697EXPORT_SYMBOL(kernel_read);
698
699static 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);
Eric Sesterhenn7dddb122006-04-01 01:13:38 +0200733 BUG_ON(active_mm != old_mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734 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 */
Arjan van de Ven858119e2006-01-14 13:20:43 -0800747static int de_thread(struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700748{
749 struct signal_struct *sig = tsk->signal;
Oleg Nesterovb2c903b2007-10-16 23:27:22 -0700750 struct sighand_struct *oldsighand = tsk->sighand;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700751 spinlock_t *lock = &oldsighand->siglock;
Oleg Nesterov329f7db2005-11-07 21:12:43 +0300752 struct task_struct *leader = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753 int count;
754
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700755 if (thread_group_empty(tsk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756 goto no_thread_group;
757
758 /*
759 * Kill all other threads in the thread group.
760 * We must hold tasklist_lock to call zap_other_threads.
761 */
762 read_lock(&tasklist_lock);
763 spin_lock_irq(lock);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -0800764 if (signal_group_exit(sig)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700765 /*
766 * Another group action in progress, just
767 * return so that the signal is processed.
768 */
769 spin_unlock_irq(lock);
770 read_unlock(&tasklist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771 return -EAGAIN;
772 }
Oleg Nesterov14342612006-03-28 16:10:59 -0800773
774 /*
775 * child_reaper ignores SIGKILL, change it now.
776 * Reparenting needs write_lock on tasklist_lock,
777 * so it is safe to do it under read_lock.
778 */
Sukadev Bhattiprolu88f21d82007-10-18 23:39:50 -0700779 if (unlikely(tsk->group_leader == task_child_reaper(tsk)))
Sukadev Bhattiprolu2894d6502007-10-18 23:39:49 -0700780 task_active_pid_ns(tsk)->child_reaper = tsk;
Oleg Nesterov14342612006-03-28 16:10:59 -0800781
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -0800782 sig->group_exit_task = tsk;
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700783 zap_other_threads(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700784 read_unlock(&tasklist_lock);
785
Oleg Nesterovfea9d172008-02-08 04:19:19 -0800786 /* Account for the thread group leader hanging around: */
787 count = thread_group_leader(tsk) ? 1 : 2;
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700788 sig->notify_count = count;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789 while (atomic_read(&sig->count) > count) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700790 __set_current_state(TASK_UNINTERRUPTIBLE);
791 spin_unlock_irq(lock);
792 schedule();
793 spin_lock_irq(lock);
794 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795 spin_unlock_irq(lock);
796
797 /*
798 * At this point all other threads have exited, all we have to
799 * do is to wait for the thread group leader to become inactive,
800 * and to assume its PID:
801 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700802 if (!thread_group_leader(tsk)) {
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700803 leader = tsk->group_leader;
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700804
805 sig->notify_count = -1;
806 for (;;) {
807 write_lock_irq(&tasklist_lock);
808 if (likely(leader->exit_state))
809 break;
810 __set_current_state(TASK_UNINTERRUPTIBLE);
811 write_unlock_irq(&tasklist_lock);
812 schedule();
813 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814
Roland McGrathf5e90282006-04-10 22:54:16 -0700815 /*
816 * The only record we have of the real-time age of a
817 * process, regardless of execs it's done, is start_time.
818 * All the past CPU time is accumulated in signal_struct
819 * from sister threads now dead. But in this non-leader
820 * exec, nothing survives from the original leader thread,
821 * whose birth marks the true age of this process now.
822 * When we take on its identity by switching to its PID, we
823 * also take its birthdate (always earlier than our own).
824 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700825 tsk->start_time = leader->start_time;
Roland McGrathf5e90282006-04-10 22:54:16 -0700826
Pavel Emelyanovbac0abd2007-10-18 23:40:18 -0700827 BUG_ON(!same_thread_group(leader, tsk));
828 BUG_ON(has_group_leader_pid(tsk));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700829 /*
830 * An exec() starts a new thread group with the
831 * TGID of the previous thread group. Rehash the
832 * two threads with a switched PID, and release
833 * the former thread group leader:
834 */
Eric W. Biedermand73d6522006-03-28 16:11:03 -0800835
836 /* Become a process group leader with the old leader's pid.
Eric W. Biedermanc18258c2006-09-27 01:51:06 -0700837 * The old leader becomes a thread of the this thread group.
838 * Note: The old leader also uses this pid until release_task
Eric W. Biedermand73d6522006-03-28 16:11:03 -0800839 * is called. Odd but simple and correct.
840 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700841 detach_pid(tsk, PIDTYPE_PID);
842 tsk->pid = leader->pid;
Sukadev Bhattiprolu3743ca02007-10-18 23:39:51 -0700843 attach_pid(tsk, PIDTYPE_PID, task_pid(leader));
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700844 transfer_pid(leader, tsk, PIDTYPE_PGID);
845 transfer_pid(leader, tsk, PIDTYPE_SID);
846 list_replace_rcu(&leader->tasks, &tsk->tasks);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700848 tsk->group_leader = tsk;
849 leader->group_leader = tsk;
Eric W. Biedermande12a782006-04-10 17:16:49 -0600850
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700851 tsk->exit_signal = SIGCHLD;
Oleg Nesterov962b5642005-11-23 13:37:43 -0800852
853 BUG_ON(leader->exit_state != EXIT_ZOMBIE);
854 leader->exit_state = EXIT_DEAD;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855
856 write_unlock_irq(&tasklist_lock);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -0800857 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700859 sig->group_exit_task = NULL;
860 sig->notify_count = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700861
862no_thread_group:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863 exit_itimers(sig);
Oleg Nesterov329f7db2005-11-07 21:12:43 +0300864 if (leader)
865 release_task(leader);
866
Oleg Nesterovb2c903b2007-10-16 23:27:22 -0700867 if (atomic_read(&oldsighand->count) != 1) {
868 struct sighand_struct *newsighand;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869 /*
Oleg Nesterovb2c903b2007-10-16 23:27:22 -0700870 * This ->sighand is shared with the CLONE_SIGHAND
871 * but not CLONE_THREAD task, switch to the new one.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700872 */
Oleg Nesterovb2c903b2007-10-16 23:27:22 -0700873 newsighand = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
874 if (!newsighand)
875 return -ENOMEM;
876
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877 atomic_set(&newsighand->count, 1);
878 memcpy(newsighand->action, oldsighand->action,
879 sizeof(newsighand->action));
880
881 write_lock_irq(&tasklist_lock);
882 spin_lock(&oldsighand->siglock);
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700883 rcu_assign_pointer(tsk->sighand, newsighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700884 spin_unlock(&oldsighand->siglock);
885 write_unlock_irq(&tasklist_lock);
886
Davide Libenzifba2afa2007-05-10 22:23:13 -0700887 __cleanup_sighand(oldsighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888 }
889
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700890 BUG_ON(!thread_group_leader(tsk));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891 return 0;
892}
Oleg Nesterov0840a902007-10-16 23:27:22 -0700893
Linus Torvalds1da177e2005-04-16 15:20:36 -0700894/*
895 * These functions flushes out all traces of the currently running executable
896 * so that a new one can be started
897 */
Arjan van de Ven858119e2006-01-14 13:20:43 -0800898static void flush_old_files(struct files_struct * files)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700899{
900 long j = -1;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700901 struct fdtable *fdt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902
903 spin_lock(&files->file_lock);
904 for (;;) {
905 unsigned long set, i;
906
907 j++;
908 i = j * __NFDBITS;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700909 fdt = files_fdtable(files);
Vadim Lobanovbbea9f62006-12-10 02:21:12 -0800910 if (i >= fdt->max_fds)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700911 break;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700912 set = fdt->close_on_exec->fds_bits[j];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700913 if (!set)
914 continue;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700915 fdt->close_on_exec->fds_bits[j] = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916 spin_unlock(&files->file_lock);
917 for ( ; set ; i++,set >>= 1) {
918 if (set & 1) {
919 sys_close(i);
920 }
921 }
922 spin_lock(&files->file_lock);
923
924 }
925 spin_unlock(&files->file_lock);
926}
927
Andrew Morton59714d62008-02-04 22:27:21 -0800928char *get_task_comm(char *buf, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700929{
930 /* buf must be at least sizeof(tsk->comm) in size */
931 task_lock(tsk);
932 strncpy(buf, tsk->comm, sizeof(tsk->comm));
933 task_unlock(tsk);
Andrew Morton59714d62008-02-04 22:27:21 -0800934 return buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700935}
936
937void set_task_comm(struct task_struct *tsk, char *buf)
938{
939 task_lock(tsk);
940 strlcpy(tsk->comm, buf, sizeof(tsk->comm));
941 task_unlock(tsk);
942}
943
944int flush_old_exec(struct linux_binprm * bprm)
945{
946 char * name;
947 int i, ch, retval;
948 struct files_struct *files;
949 char tcomm[sizeof(current->comm)];
950
951 /*
952 * Make sure we have a private signal table and that
953 * we are unassociated from the previous thread group.
954 */
955 retval = de_thread(current);
956 if (retval)
957 goto out;
958
959 /*
960 * Make sure we have private file handles. Ask the
961 * fork helper to do the work for us and the exit
962 * helper to do the cleanup of the old one.
963 */
964 files = current->files; /* refcounted so safe to hold */
965 retval = unshare_files();
966 if (retval)
967 goto out;
968 /*
969 * Release all of the old mmap stuff
970 */
971 retval = exec_mmap(bprm->mm);
972 if (retval)
973 goto mmap_failed;
974
975 bprm->mm = NULL; /* We're using it now */
976
977 /* This is the point of no return */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700978 put_files_struct(files);
979
980 current->sas_ss_sp = current->sas_ss_size = 0;
981
982 if (current->euid == current->uid && current->egid == current->gid)
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -0700983 set_dumpable(current->mm, 1);
Alan Coxd6e71142005-06-23 00:09:43 -0700984 else
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -0700985 set_dumpable(current->mm, suid_dumpable);
Alan Coxd6e71142005-06-23 00:09:43 -0700986
Linus Torvalds1da177e2005-04-16 15:20:36 -0700987 name = bprm->filename;
Paolo 'Blaisorblade' Giarrusso36772092005-05-05 16:16:12 -0700988
989 /* Copies the binary name from after last slash */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990 for (i=0; (ch = *(name++)) != '\0';) {
991 if (ch == '/')
Paolo 'Blaisorblade' Giarrusso36772092005-05-05 16:16:12 -0700992 i = 0; /* overwrite what we wrote */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700993 else
994 if (i < (sizeof(tcomm) - 1))
995 tcomm[i++] = ch;
996 }
997 tcomm[i] = '\0';
998 set_task_comm(current, tcomm);
999
1000 current->flags &= ~PF_RANDOMIZE;
1001 flush_thread();
1002
Benjamin Herrenschmidt0551fbd2006-02-28 16:59:19 -08001003 /* Set the new mm task size. We have to do that late because it may
1004 * depend on TIF_32BIT which is only updated in flush_thread() on
1005 * some architectures like powerpc
1006 */
1007 current->mm->task_size = TASK_SIZE;
1008
Marcel Holtmannd2d56c52007-08-17 21:47:58 +02001009 if (bprm->e_uid != current->euid || bprm->e_gid != current->egid) {
1010 suid_keys(current);
1011 set_dumpable(current->mm, suid_dumpable);
1012 current->pdeath_signal = 0;
1013 } else if (file_permission(bprm->file, MAY_READ) ||
1014 (bprm->interp_flags & BINPRM_FLAGS_ENFORCE_NONDUMP)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001015 suid_keys(current);
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001016 set_dumpable(current->mm, suid_dumpable);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001017 }
1018
1019 /* An exec changes our domain. We are no longer part of the thread
1020 group */
1021
1022 current->self_exec_id++;
1023
1024 flush_signal_handlers(current, 0);
1025 flush_old_files(current->files);
1026
1027 return 0;
1028
1029mmap_failed:
Kirill Korotaev3b9b8ab2006-09-29 02:00:05 -07001030 reset_files_struct(current, files);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031out:
1032 return retval;
1033}
1034
1035EXPORT_SYMBOL(flush_old_exec);
1036
1037/*
1038 * Fill the binprm structure from the inode.
1039 * Check permissions, then read the first 128 (BINPRM_BUF_SIZE) bytes
1040 */
1041int prepare_binprm(struct linux_binprm *bprm)
1042{
1043 int mode;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001044 struct inode * inode = bprm->file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001045 int retval;
1046
1047 mode = inode->i_mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001048 if (bprm->file->f_op == NULL)
1049 return -EACCES;
1050
1051 bprm->e_uid = current->euid;
1052 bprm->e_gid = current->egid;
1053
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001054 if(!(bprm->file->f_path.mnt->mnt_flags & MNT_NOSUID)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055 /* Set-uid? */
1056 if (mode & S_ISUID) {
1057 current->personality &= ~PER_CLEAR_ON_SETID;
1058 bprm->e_uid = inode->i_uid;
1059 }
1060
1061 /* Set-gid? */
1062 /*
1063 * If setgid is set but no group execute bit then this
1064 * is a candidate for mandatory locking, not a setgid
1065 * executable.
1066 */
1067 if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) {
1068 current->personality &= ~PER_CLEAR_ON_SETID;
1069 bprm->e_gid = inode->i_gid;
1070 }
1071 }
1072
1073 /* fill in binprm security blob */
1074 retval = security_bprm_set(bprm);
1075 if (retval)
1076 return retval;
1077
1078 memset(bprm->buf,0,BINPRM_BUF_SIZE);
1079 return kernel_read(bprm->file,0,bprm->buf,BINPRM_BUF_SIZE);
1080}
1081
1082EXPORT_SYMBOL(prepare_binprm);
1083
Arjan van de Ven858119e2006-01-14 13:20:43 -08001084static int unsafe_exec(struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001085{
1086 int unsafe = 0;
1087 if (p->ptrace & PT_PTRACED) {
1088 if (p->ptrace & PT_PTRACE_CAP)
1089 unsafe |= LSM_UNSAFE_PTRACE_CAP;
1090 else
1091 unsafe |= LSM_UNSAFE_PTRACE;
1092 }
1093 if (atomic_read(&p->fs->count) > 1 ||
1094 atomic_read(&p->files->count) > 1 ||
1095 atomic_read(&p->sighand->count) > 1)
1096 unsafe |= LSM_UNSAFE_SHARE;
1097
1098 return unsafe;
1099}
1100
1101void compute_creds(struct linux_binprm *bprm)
1102{
1103 int unsafe;
1104
Marcel Holtmannd2d56c52007-08-17 21:47:58 +02001105 if (bprm->e_uid != current->uid) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106 suid_keys(current);
Marcel Holtmannd2d56c52007-08-17 21:47:58 +02001107 current->pdeath_signal = 0;
1108 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001109 exec_keys(current);
1110
1111 task_lock(current);
1112 unsafe = unsafe_exec(current);
1113 security_bprm_apply_creds(bprm, unsafe);
1114 task_unlock(current);
1115 security_bprm_post_apply_creds(bprm);
1116}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001117EXPORT_SYMBOL(compute_creds);
1118
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001119/*
1120 * Arguments are '\0' separated strings found at the location bprm->p
1121 * points to; chop off the first by relocating brpm->p to right after
1122 * the first '\0' encountered.
1123 */
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001124int remove_arg_zero(struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001125{
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001126 int ret = 0;
1127 unsigned long offset;
1128 char *kaddr;
1129 struct page *page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001130
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001131 if (!bprm->argc)
1132 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001133
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001134 do {
1135 offset = bprm->p & ~PAGE_MASK;
1136 page = get_arg_page(bprm, bprm->p, 0);
1137 if (!page) {
1138 ret = -EFAULT;
1139 goto out;
1140 }
1141 kaddr = kmap_atomic(page, KM_USER0);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001142
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001143 for (; offset < PAGE_SIZE && kaddr[offset];
1144 offset++, bprm->p++)
1145 ;
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001146
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001147 kunmap_atomic(kaddr, KM_USER0);
1148 put_arg_page(page);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001149
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001150 if (offset == PAGE_SIZE)
1151 free_arg_page(bprm, (bprm->p >> PAGE_SHIFT) - 1);
1152 } while (offset == PAGE_SIZE);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001153
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001154 bprm->p++;
1155 bprm->argc--;
1156 ret = 0;
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001157
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001158out:
1159 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001160}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001161EXPORT_SYMBOL(remove_arg_zero);
1162
1163/*
1164 * cycle the list of binary formats handler, until one recognizes the image
1165 */
1166int search_binary_handler(struct linux_binprm *bprm,struct pt_regs *regs)
1167{
1168 int try,retval;
1169 struct linux_binfmt *fmt;
David Howells7fa30312008-02-08 04:19:28 -08001170#if defined(__alpha__) && defined(CONFIG_ARCH_SUPPORTS_AOUT)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001171 /* handle /sbin/loader.. */
1172 {
1173 struct exec * eh = (struct exec *) bprm->buf;
1174
1175 if (!bprm->loader && eh->fh.f_magic == 0x183 &&
1176 (eh->fh.f_flags & 0x3000) == 0x3000)
1177 {
1178 struct file * file;
1179 unsigned long loader;
1180
1181 allow_write_access(bprm->file);
1182 fput(bprm->file);
1183 bprm->file = NULL;
1184
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001185 loader = bprm->vma->vm_end - sizeof(void *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001186
1187 file = open_exec("/sbin/loader");
1188 retval = PTR_ERR(file);
1189 if (IS_ERR(file))
1190 return retval;
1191
1192 /* Remember if the application is TASO. */
1193 bprm->sh_bang = eh->ah.entry < 0x100000000UL;
1194
1195 bprm->file = file;
1196 bprm->loader = loader;
1197 retval = prepare_binprm(bprm);
1198 if (retval<0)
1199 return retval;
1200 /* should call search_binary_handler recursively here,
1201 but it does not matter */
1202 }
1203 }
1204#endif
1205 retval = security_bprm_check(bprm);
1206 if (retval)
1207 return retval;
1208
1209 /* kernel module loader fixup */
1210 /* so we don't try to load run modprobe in kernel space. */
1211 set_fs(USER_DS);
Al Viro473ae302006-04-26 14:04:08 -04001212
1213 retval = audit_bprm(bprm);
1214 if (retval)
1215 return retval;
1216
Linus Torvalds1da177e2005-04-16 15:20:36 -07001217 retval = -ENOENT;
1218 for (try=0; try<2; try++) {
1219 read_lock(&binfmt_lock);
Alexey Dobriyane4dc1b12007-10-16 23:26:03 -07001220 list_for_each_entry(fmt, &formats, lh) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001221 int (*fn)(struct linux_binprm *, struct pt_regs *) = fmt->load_binary;
1222 if (!fn)
1223 continue;
1224 if (!try_module_get(fmt->module))
1225 continue;
1226 read_unlock(&binfmt_lock);
1227 retval = fn(bprm, regs);
1228 if (retval >= 0) {
1229 put_binfmt(fmt);
1230 allow_write_access(bprm->file);
1231 if (bprm->file)
1232 fput(bprm->file);
1233 bprm->file = NULL;
1234 current->did_exec = 1;
Matt Helsley9f460802005-11-07 00:59:16 -08001235 proc_exec_connector(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001236 return retval;
1237 }
1238 read_lock(&binfmt_lock);
1239 put_binfmt(fmt);
1240 if (retval != -ENOEXEC || bprm->mm == NULL)
1241 break;
1242 if (!bprm->file) {
1243 read_unlock(&binfmt_lock);
1244 return retval;
1245 }
1246 }
1247 read_unlock(&binfmt_lock);
1248 if (retval != -ENOEXEC || bprm->mm == NULL) {
1249 break;
1250#ifdef CONFIG_KMOD
1251 }else{
1252#define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e))
1253 if (printable(bprm->buf[0]) &&
1254 printable(bprm->buf[1]) &&
1255 printable(bprm->buf[2]) &&
1256 printable(bprm->buf[3]))
1257 break; /* -ENOEXEC */
1258 request_module("binfmt-%04x", *(unsigned short *)(&bprm->buf[2]));
1259#endif
1260 }
1261 }
1262 return retval;
1263}
1264
1265EXPORT_SYMBOL(search_binary_handler);
1266
1267/*
1268 * sys_execve() executes a new program.
1269 */
1270int do_execve(char * filename,
1271 char __user *__user *argv,
1272 char __user *__user *envp,
1273 struct pt_regs * regs)
1274{
1275 struct linux_binprm *bprm;
1276 struct file *file;
Peter Zijlstrabdf4c482007-07-19 01:48:15 -07001277 unsigned long env_p;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001278 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001279
1280 retval = -ENOMEM;
Oliver Neukum11b0b5a2006-03-25 03:08:13 -08001281 bprm = kzalloc(sizeof(*bprm), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001282 if (!bprm)
1283 goto out_ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001284
1285 file = open_exec(filename);
1286 retval = PTR_ERR(file);
1287 if (IS_ERR(file))
1288 goto out_kfree;
1289
1290 sched_exec();
1291
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292 bprm->file = file;
1293 bprm->filename = filename;
1294 bprm->interp = filename;
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001295
1296 retval = bprm_mm_init(bprm);
1297 if (retval)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001298 goto out_file;
1299
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001300 bprm->argc = count(argv, MAX_ARG_STRINGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001301 if ((retval = bprm->argc) < 0)
1302 goto out_mm;
1303
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001304 bprm->envc = count(envp, MAX_ARG_STRINGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001305 if ((retval = bprm->envc) < 0)
1306 goto out_mm;
1307
1308 retval = security_bprm_alloc(bprm);
1309 if (retval)
1310 goto out;
1311
1312 retval = prepare_binprm(bprm);
1313 if (retval < 0)
1314 goto out;
1315
1316 retval = copy_strings_kernel(1, &bprm->filename, bprm);
1317 if (retval < 0)
1318 goto out;
1319
1320 bprm->exec = bprm->p;
1321 retval = copy_strings(bprm->envc, envp, bprm);
1322 if (retval < 0)
1323 goto out;
1324
Peter Zijlstrabdf4c482007-07-19 01:48:15 -07001325 env_p = bprm->p;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001326 retval = copy_strings(bprm->argc, argv, bprm);
1327 if (retval < 0)
1328 goto out;
Peter Zijlstrabdf4c482007-07-19 01:48:15 -07001329 bprm->argv_len = env_p - bprm->p;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001330
1331 retval = search_binary_handler(bprm,regs);
1332 if (retval >= 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001333 /* execve success */
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001334 free_arg_pages(bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001335 security_bprm_free(bprm);
1336 acct_update_integrals(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001337 kfree(bprm);
1338 return retval;
1339 }
1340
1341out:
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001342 free_arg_pages(bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001343 if (bprm->security)
1344 security_bprm_free(bprm);
1345
1346out_mm:
1347 if (bprm->mm)
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001348 mmput (bprm->mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001349
1350out_file:
1351 if (bprm->file) {
1352 allow_write_access(bprm->file);
1353 fput(bprm->file);
1354 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001355out_kfree:
1356 kfree(bprm);
1357
1358out_ret:
1359 return retval;
1360}
1361
1362int set_binfmt(struct linux_binfmt *new)
1363{
1364 struct linux_binfmt *old = current->binfmt;
1365
1366 if (new) {
1367 if (!try_module_get(new->module))
1368 return -1;
1369 }
1370 current->binfmt = new;
1371 if (old)
1372 module_put(old->module);
1373 return 0;
1374}
1375
1376EXPORT_SYMBOL(set_binfmt);
1377
Linus Torvalds1da177e2005-04-16 15:20:36 -07001378/* format_corename will inspect the pattern parameter, and output a
1379 * name into corename, which must have space for at least
1380 * CORENAME_MAX_SIZE bytes plus one byte for the zero terminator.
1381 */
Alan Coxc4bbafd2007-04-16 22:53:13 -07001382static int format_corename(char *corename, const char *pattern, long signr)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383{
1384 const char *pat_ptr = pattern;
1385 char *out_ptr = corename;
1386 char *const out_end = corename + CORENAME_MAX_SIZE;
1387 int rc;
1388 int pid_in_pattern = 0;
Alan Coxc4bbafd2007-04-16 22:53:13 -07001389 int ispipe = 0;
1390
1391 if (*pattern == '|')
1392 ispipe = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001393
1394 /* Repeat as long as we have more pattern to process and more output
1395 space */
1396 while (*pat_ptr) {
1397 if (*pat_ptr != '%') {
1398 if (out_ptr == out_end)
1399 goto out;
1400 *out_ptr++ = *pat_ptr++;
1401 } else {
1402 switch (*++pat_ptr) {
1403 case 0:
1404 goto out;
1405 /* Double percent, output one percent */
1406 case '%':
1407 if (out_ptr == out_end)
1408 goto out;
1409 *out_ptr++ = '%';
1410 break;
1411 /* pid */
1412 case 'p':
1413 pid_in_pattern = 1;
1414 rc = snprintf(out_ptr, out_end - out_ptr,
Pavel Emelyanovb4888932007-10-18 23:40:14 -07001415 "%d", task_tgid_vnr(current));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001416 if (rc > out_end - out_ptr)
1417 goto out;
1418 out_ptr += rc;
1419 break;
1420 /* uid */
1421 case 'u':
1422 rc = snprintf(out_ptr, out_end - out_ptr,
1423 "%d", current->uid);
1424 if (rc > out_end - out_ptr)
1425 goto out;
1426 out_ptr += rc;
1427 break;
1428 /* gid */
1429 case 'g':
1430 rc = snprintf(out_ptr, out_end - out_ptr,
1431 "%d", current->gid);
1432 if (rc > out_end - out_ptr)
1433 goto out;
1434 out_ptr += rc;
1435 break;
1436 /* signal that caused the coredump */
1437 case 's':
1438 rc = snprintf(out_ptr, out_end - out_ptr,
1439 "%ld", signr);
1440 if (rc > out_end - out_ptr)
1441 goto out;
1442 out_ptr += rc;
1443 break;
1444 /* UNIX time of coredump */
1445 case 't': {
1446 struct timeval tv;
1447 do_gettimeofday(&tv);
1448 rc = snprintf(out_ptr, out_end - out_ptr,
1449 "%lu", tv.tv_sec);
1450 if (rc > out_end - out_ptr)
1451 goto out;
1452 out_ptr += rc;
1453 break;
1454 }
1455 /* hostname */
1456 case 'h':
1457 down_read(&uts_sem);
1458 rc = snprintf(out_ptr, out_end - out_ptr,
Serge E. Hallyne9ff3992006-10-02 02:18:11 -07001459 "%s", utsname()->nodename);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001460 up_read(&uts_sem);
1461 if (rc > out_end - out_ptr)
1462 goto out;
1463 out_ptr += rc;
1464 break;
1465 /* executable */
1466 case 'e':
1467 rc = snprintf(out_ptr, out_end - out_ptr,
1468 "%s", current->comm);
1469 if (rc > out_end - out_ptr)
1470 goto out;
1471 out_ptr += rc;
1472 break;
Neil Horman74aadce2007-10-16 23:26:35 -07001473 /* core limit size */
1474 case 'c':
1475 rc = snprintf(out_ptr, out_end - out_ptr,
1476 "%lu", current->signal->rlim[RLIMIT_CORE].rlim_cur);
1477 if (rc > out_end - out_ptr)
1478 goto out;
1479 out_ptr += rc;
1480 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001481 default:
1482 break;
1483 }
1484 ++pat_ptr;
1485 }
1486 }
1487 /* Backward compatibility with core_uses_pid:
1488 *
1489 * If core_pattern does not include a %p (as is the default)
1490 * and core_uses_pid is set, then .%pid will be appended to
Alan Coxc4bbafd2007-04-16 22:53:13 -07001491 * the filename. Do not do this for piped commands. */
1492 if (!ispipe && !pid_in_pattern
Linus Torvalds1da177e2005-04-16 15:20:36 -07001493 && (core_uses_pid || atomic_read(&current->mm->mm_users) != 1)) {
1494 rc = snprintf(out_ptr, out_end - out_ptr,
Pavel Emelyanovb4888932007-10-18 23:40:14 -07001495 ".%d", task_tgid_vnr(current));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001496 if (rc > out_end - out_ptr)
1497 goto out;
1498 out_ptr += rc;
1499 }
Alan Coxc4bbafd2007-04-16 22:53:13 -07001500out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001501 *out_ptr = 0;
Alan Coxc4bbafd2007-04-16 22:53:13 -07001502 return ispipe;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001503}
1504
Oleg Nesterovd5f70c02006-06-26 00:26:07 -07001505static void zap_process(struct task_struct *start)
Oleg Nesterovaceecc02006-06-26 00:26:05 -07001506{
1507 struct task_struct *t;
Oleg Nesterov281de332006-06-26 00:26:06 -07001508
Oleg Nesterovd5f70c02006-06-26 00:26:07 -07001509 start->signal->flags = SIGNAL_GROUP_EXIT;
1510 start->signal->group_stop_count = 0;
Oleg Nesterovaceecc02006-06-26 00:26:05 -07001511
1512 t = start;
1513 do {
1514 if (t != current && t->mm) {
1515 t->mm->core_waiters++;
Oleg Nesterov281de332006-06-26 00:26:06 -07001516 sigaddset(&t->pending.signal, SIGKILL);
1517 signal_wake_up(t, 1);
Oleg Nesterovaceecc02006-06-26 00:26:05 -07001518 }
1519 } while ((t = next_thread(t)) != start);
1520}
1521
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001522static inline int zap_threads(struct task_struct *tsk, struct mm_struct *mm,
1523 int exit_code)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001524{
1525 struct task_struct *g, *p;
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001526 unsigned long flags;
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001527 int err = -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001528
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001529 spin_lock_irq(&tsk->sighand->siglock);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -08001530 if (!signal_group_exit(tsk->signal)) {
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001531 tsk->signal->group_exit_code = exit_code;
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001532 zap_process(tsk);
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001533 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001534 }
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001535 spin_unlock_irq(&tsk->sighand->siglock);
1536 if (err)
1537 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001538
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001539 if (atomic_read(&mm->mm_users) == mm->core_waiters + 1)
1540 goto done;
1541
Oleg Nesterov7b1c6152006-06-26 00:26:08 -07001542 rcu_read_lock();
Oleg Nesterovaceecc02006-06-26 00:26:05 -07001543 for_each_process(g) {
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001544 if (g == tsk->group_leader)
1545 continue;
1546
Oleg Nesterovaceecc02006-06-26 00:26:05 -07001547 p = g;
1548 do {
1549 if (p->mm) {
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001550 if (p->mm == mm) {
1551 /*
1552 * p->sighand can't disappear, but
1553 * may be changed by de_thread()
1554 */
1555 lock_task_sighand(p, &flags);
Oleg Nesterovd5f70c02006-06-26 00:26:07 -07001556 zap_process(p);
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001557 unlock_task_sighand(p, &flags);
1558 }
Oleg Nesterovaceecc02006-06-26 00:26:05 -07001559 break;
1560 }
1561 } while ((p = next_thread(p)) != g);
1562 }
Oleg Nesterov7b1c6152006-06-26 00:26:08 -07001563 rcu_read_unlock();
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001564done:
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001565 return mm->core_waiters;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001566}
1567
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001568static int coredump_wait(int exit_code)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001569{
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001570 struct task_struct *tsk = current;
1571 struct mm_struct *mm = tsk->mm;
1572 struct completion startup_done;
1573 struct completion *vfork_done;
Oleg Nesterov2384f552005-10-30 15:02:47 -08001574 int core_waiters;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001575
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001576 init_completion(&mm->core_done);
1577 init_completion(&startup_done);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001578 mm->core_startup_done = &startup_done;
1579
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001580 core_waiters = zap_threads(tsk, mm, exit_code);
Oleg Nesterov2384f552005-10-30 15:02:47 -08001581 up_write(&mm->mmap_sem);
1582
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001583 if (unlikely(core_waiters < 0))
1584 goto fail;
1585
1586 /*
1587 * Make sure nobody is waiting for us to release the VM,
1588 * otherwise we can deadlock when we wait on each other
1589 */
1590 vfork_done = tsk->vfork_done;
1591 if (vfork_done) {
1592 tsk->vfork_done = NULL;
1593 complete(vfork_done);
1594 }
1595
Oleg Nesterov2384f552005-10-30 15:02:47 -08001596 if (core_waiters)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001597 wait_for_completion(&startup_done);
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001598fail:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001599 BUG_ON(mm->core_waiters);
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001600 return core_waiters;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001601}
1602
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001603/*
1604 * set_dumpable converts traditional three-value dumpable to two flags and
1605 * stores them into mm->flags. It modifies lower two bits of mm->flags, but
1606 * these bits are not changed atomically. So get_dumpable can observe the
1607 * intermediate state. To avoid doing unexpected behavior, get get_dumpable
1608 * return either old dumpable or new one by paying attention to the order of
1609 * modifying the bits.
1610 *
1611 * dumpable | mm->flags (binary)
1612 * old new | initial interim final
1613 * ---------+-----------------------
1614 * 0 1 | 00 01 01
1615 * 0 2 | 00 10(*) 11
1616 * 1 0 | 01 00 00
1617 * 1 2 | 01 11 11
1618 * 2 0 | 11 10(*) 00
1619 * 2 1 | 11 11 01
1620 *
1621 * (*) get_dumpable regards interim value of 10 as 11.
1622 */
1623void set_dumpable(struct mm_struct *mm, int value)
1624{
1625 switch (value) {
1626 case 0:
1627 clear_bit(MMF_DUMPABLE, &mm->flags);
1628 smp_wmb();
1629 clear_bit(MMF_DUMP_SECURELY, &mm->flags);
1630 break;
1631 case 1:
1632 set_bit(MMF_DUMPABLE, &mm->flags);
1633 smp_wmb();
1634 clear_bit(MMF_DUMP_SECURELY, &mm->flags);
1635 break;
1636 case 2:
1637 set_bit(MMF_DUMP_SECURELY, &mm->flags);
1638 smp_wmb();
1639 set_bit(MMF_DUMPABLE, &mm->flags);
1640 break;
1641 }
1642}
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001643
1644int get_dumpable(struct mm_struct *mm)
1645{
1646 int ret;
1647
1648 ret = mm->flags & 0x3;
1649 return (ret >= 2) ? 2 : ret;
1650}
1651
Linus Torvalds1da177e2005-04-16 15:20:36 -07001652int do_coredump(long signr, int exit_code, struct pt_regs * regs)
1653{
1654 char corename[CORENAME_MAX_SIZE + 1];
1655 struct mm_struct *mm = current->mm;
1656 struct linux_binfmt * binfmt;
1657 struct inode * inode;
1658 struct file * file;
1659 int retval = 0;
Alan Coxd6e71142005-06-23 00:09:43 -07001660 int fsuid = current->fsuid;
1661 int flag = 0;
Andi Kleend025c9d2006-09-30 23:29:28 -07001662 int ispipe = 0;
Neil Horman7dc0b222007-10-16 23:26:34 -07001663 unsigned long core_limit = current->signal->rlim[RLIMIT_CORE].rlim_cur;
Neil Horman74aadce2007-10-16 23:26:35 -07001664 char **helper_argv = NULL;
1665 int helper_argc = 0;
1666 char *delimit;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001667
Steve Grubb0a4ff8c2007-04-19 10:28:21 -04001668 audit_core_dumps(signr);
1669
Linus Torvalds1da177e2005-04-16 15:20:36 -07001670 binfmt = current->binfmt;
1671 if (!binfmt || !binfmt->core_dump)
1672 goto fail;
1673 down_write(&mm->mmap_sem);
Roland McGrath00ec99d2007-11-11 19:13:43 -08001674 /*
1675 * If another thread got here first, or we are not dumpable, bail out.
1676 */
1677 if (mm->core_waiters || !get_dumpable(mm)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001678 up_write(&mm->mmap_sem);
1679 goto fail;
1680 }
Alan Coxd6e71142005-06-23 00:09:43 -07001681
1682 /*
1683 * We cannot trust fsuid as being the "true" uid of the
1684 * process nor do we know its entire history. We only know it
1685 * was tainted so we dump it as root in mode 2.
1686 */
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001687 if (get_dumpable(mm) == 2) { /* Setuid core dump mode */
Alan Coxd6e71142005-06-23 00:09:43 -07001688 flag = O_EXCL; /* Stop rewrite attacks */
1689 current->fsuid = 0; /* Dump root private */
1690 }
Oleg Nesterov1291cf42005-10-30 15:02:54 -08001691
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001692 retval = coredump_wait(exit_code);
1693 if (retval < 0)
Oleg Nesterov1291cf42005-10-30 15:02:54 -08001694 goto fail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001695
1696 /*
1697 * Clear any false indication of pending signals that might
1698 * be seen by the filesystem code called to write the core file.
1699 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001700 clear_thread_flag(TIF_SIGPENDING);
1701
Linus Torvalds1da177e2005-04-16 15:20:36 -07001702 /*
1703 * lock_kernel() because format_corename() is controlled by sysctl, which
1704 * uses lock_kernel()
1705 */
1706 lock_kernel();
Alan Coxc4bbafd2007-04-16 22:53:13 -07001707 ispipe = format_corename(corename, core_pattern, signr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001708 unlock_kernel();
Neil Horman7dc0b222007-10-16 23:26:34 -07001709 /*
1710 * Don't bother to check the RLIMIT_CORE value if core_pattern points
1711 * to a pipe. Since we're not writing directly to the filesystem
1712 * RLIMIT_CORE doesn't really apply, as no actual core file will be
1713 * created unless the pipe reader choses to write out the core file
1714 * at which point file size limits and permissions will be imposed
1715 * as it does with any other process
1716 */
Neil Horman74aadce2007-10-16 23:26:35 -07001717 if ((!ispipe) && (core_limit < binfmt->min_coredump))
Neil Horman7dc0b222007-10-16 23:26:34 -07001718 goto fail_unlock;
1719
Alan Coxc4bbafd2007-04-16 22:53:13 -07001720 if (ispipe) {
Neil Horman74aadce2007-10-16 23:26:35 -07001721 helper_argv = argv_split(GFP_KERNEL, corename+1, &helper_argc);
1722 /* Terminate the string before the first option */
1723 delimit = strchr(corename, ' ');
1724 if (delimit)
1725 *delimit = '\0';
Neil Horman32321132007-10-16 23:26:36 -07001726 delimit = strrchr(helper_argv[0], '/');
1727 if (delimit)
1728 delimit++;
1729 else
1730 delimit = helper_argv[0];
1731 if (!strcmp(delimit, current->comm)) {
1732 printk(KERN_NOTICE "Recursive core dump detected, "
1733 "aborting\n");
1734 goto fail_unlock;
1735 }
1736
1737 core_limit = RLIM_INFINITY;
1738
Andi Kleend025c9d2006-09-30 23:29:28 -07001739 /* SIGPIPE can happen, but it's just never processed */
Neil Horman32321132007-10-16 23:26:36 -07001740 if (call_usermodehelper_pipe(corename+1, helper_argv, NULL,
1741 &file)) {
Andi Kleend025c9d2006-09-30 23:29:28 -07001742 printk(KERN_INFO "Core dump to %s pipe failed\n",
1743 corename);
1744 goto fail_unlock;
1745 }
Andi Kleend025c9d2006-09-30 23:29:28 -07001746 } else
1747 file = filp_open(corename,
Alexey Dobriyan6d4df672006-12-06 20:40:39 -08001748 O_CREAT | 2 | O_NOFOLLOW | O_LARGEFILE | flag,
1749 0600);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001750 if (IS_ERR(file))
1751 goto fail_unlock;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001752 inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001753 if (inode->i_nlink > 1)
1754 goto close_fail; /* multiple links - don't dump */
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001755 if (!ispipe && d_unhashed(file->f_path.dentry))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001756 goto close_fail;
1757
Andi Kleend025c9d2006-09-30 23:29:28 -07001758 /* AK: actually i see no reason to not allow this for named pipes etc.,
1759 but keep the previous behaviour for now. */
1760 if (!ispipe && !S_ISREG(inode->i_mode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001761 goto close_fail;
Ingo Molnarc46f7392007-11-28 13:59:18 +01001762 /*
1763 * Dont allow local users get cute and trick others to coredump
1764 * into their pre-created files:
1765 */
1766 if (inode->i_uid != current->fsuid)
1767 goto close_fail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001768 if (!file->f_op)
1769 goto close_fail;
1770 if (!file->f_op->write)
1771 goto close_fail;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001772 if (!ispipe && do_truncate(file->f_path.dentry, 0, 0, file) != 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001773 goto close_fail;
1774
Neil Horman7dc0b222007-10-16 23:26:34 -07001775 retval = binfmt->core_dump(signr, regs, file, core_limit);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001776
1777 if (retval)
1778 current->signal->group_exit_code |= 0x80;
1779close_fail:
1780 filp_close(file, NULL);
1781fail_unlock:
Neil Horman74aadce2007-10-16 23:26:35 -07001782 if (helper_argv)
1783 argv_free(helper_argv);
1784
Alan Coxd6e71142005-06-23 00:09:43 -07001785 current->fsuid = fsuid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001786 complete_all(&mm->core_done);
1787fail:
1788 return retval;
1789}