blob: 711bc45d789c54886d1836ab0845b10bd7884714 [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) {
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700176 unsigned long size = bprm->vma->vm_end - bprm->vma->vm_start;
Linus Torvaldsa64e7152008-03-03 10:12:14 -0800177 struct rlimit *rlim;
178
179 /*
180 * We've historically supported up to 32 pages (ARG_MAX)
181 * of argument strings even with small stacks
182 */
183 if (size <= ARG_MAX)
184 return page;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700185
186 /*
187 * Limit to 1/4-th the stack size for the argv+env strings.
188 * This ensures that:
189 * - the remaining binfmt code will not run out of stack space,
190 * - the program will have a reasonable amount of stack left
191 * to work from.
192 */
Linus Torvaldsa64e7152008-03-03 10:12:14 -0800193 rlim = current->signal->rlim;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700194 if (size > rlim[RLIMIT_STACK].rlim_cur / 4) {
195 put_page(page);
196 return NULL;
197 }
198 }
199
200 return page;
201}
202
203static void put_arg_page(struct page *page)
204{
205 put_page(page);
206}
207
208static void free_arg_page(struct linux_binprm *bprm, int i)
209{
210}
211
212static void free_arg_pages(struct linux_binprm *bprm)
213{
214}
215
216static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos,
217 struct page *page)
218{
219 flush_cache_page(bprm->vma, pos, page_to_pfn(page));
220}
221
222static int __bprm_mm_init(struct linux_binprm *bprm)
223{
224 int err = -ENOMEM;
225 struct vm_area_struct *vma = NULL;
226 struct mm_struct *mm = bprm->mm;
227
228 bprm->vma = vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL);
229 if (!vma)
230 goto err;
231
232 down_write(&mm->mmap_sem);
233 vma->vm_mm = mm;
234
235 /*
236 * Place the stack at the largest stack address the architecture
237 * supports. Later, we'll move this to an appropriate place. We don't
238 * use STACK_TOP because that can depend on attributes which aren't
239 * configured yet.
240 */
241 vma->vm_end = STACK_TOP_MAX;
242 vma->vm_start = vma->vm_end - PAGE_SIZE;
243
244 vma->vm_flags = VM_STACK_FLAGS;
Coly Li3ed75eb2007-10-18 23:39:15 -0700245 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700246 err = insert_vm_struct(mm, vma);
247 if (err) {
248 up_write(&mm->mmap_sem);
249 goto err;
250 }
251
252 mm->stack_vm = mm->total_vm = 1;
253 up_write(&mm->mmap_sem);
254
255 bprm->p = vma->vm_end - sizeof(void *);
256
257 return 0;
258
259err:
260 if (vma) {
261 bprm->vma = NULL;
262 kmem_cache_free(vm_area_cachep, vma);
263 }
264
265 return err;
266}
267
268static bool valid_arg_len(struct linux_binprm *bprm, long len)
269{
270 return len <= MAX_ARG_STRLEN;
271}
272
273#else
274
275static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos,
276 int write)
277{
278 struct page *page;
279
280 page = bprm->page[pos / PAGE_SIZE];
281 if (!page && write) {
282 page = alloc_page(GFP_HIGHUSER|__GFP_ZERO);
283 if (!page)
284 return NULL;
285 bprm->page[pos / PAGE_SIZE] = page;
286 }
287
288 return page;
289}
290
291static void put_arg_page(struct page *page)
292{
293}
294
295static void free_arg_page(struct linux_binprm *bprm, int i)
296{
297 if (bprm->page[i]) {
298 __free_page(bprm->page[i]);
299 bprm->page[i] = NULL;
300 }
301}
302
303static void free_arg_pages(struct linux_binprm *bprm)
304{
305 int i;
306
307 for (i = 0; i < MAX_ARG_PAGES; i++)
308 free_arg_page(bprm, i);
309}
310
311static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos,
312 struct page *page)
313{
314}
315
316static int __bprm_mm_init(struct linux_binprm *bprm)
317{
318 bprm->p = PAGE_SIZE * MAX_ARG_PAGES - sizeof(void *);
319 return 0;
320}
321
322static bool valid_arg_len(struct linux_binprm *bprm, long len)
323{
324 return len <= bprm->p;
325}
326
327#endif /* CONFIG_MMU */
328
329/*
330 * Create a new mm_struct and populate it with a temporary stack
331 * vm_area_struct. We don't have enough context at this point to set the stack
332 * flags, permissions, and offset, so we use temporary values. We'll update
333 * them later in setup_arg_pages().
334 */
335int bprm_mm_init(struct linux_binprm *bprm)
336{
337 int err;
338 struct mm_struct *mm = NULL;
339
340 bprm->mm = mm = mm_alloc();
341 err = -ENOMEM;
342 if (!mm)
343 goto err;
344
345 err = init_new_context(current, mm);
346 if (err)
347 goto err;
348
349 err = __bprm_mm_init(bprm);
350 if (err)
351 goto err;
352
353 return 0;
354
355err:
356 if (mm) {
357 bprm->mm = NULL;
358 mmdrop(mm);
359 }
360
361 return err;
362}
363
Linus Torvalds1da177e2005-04-16 15:20:36 -0700364/*
365 * count() counts the number of strings in array ARGV.
366 */
367static int count(char __user * __user * argv, int max)
368{
369 int i = 0;
370
371 if (argv != NULL) {
372 for (;;) {
373 char __user * p;
374
375 if (get_user(p, argv))
376 return -EFAULT;
377 if (!p)
378 break;
379 argv++;
380 if(++i > max)
381 return -E2BIG;
382 cond_resched();
383 }
384 }
385 return i;
386}
387
388/*
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700389 * 'copy_strings()' copies argument/environment strings from the old
390 * processes's memory to the new process's stack. The call to get_user_pages()
391 * ensures the destination page is created and not swapped out.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392 */
Adrian Bunk75c96f82005-05-05 16:16:09 -0700393static int copy_strings(int argc, char __user * __user * argv,
394 struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395{
396 struct page *kmapped_page = NULL;
397 char *kaddr = NULL;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700398 unsigned long kpos = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399 int ret;
400
401 while (argc-- > 0) {
402 char __user *str;
403 int len;
404 unsigned long pos;
405
406 if (get_user(str, argv+argc) ||
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700407 !(len = strnlen_user(str, MAX_ARG_STRLEN))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408 ret = -EFAULT;
409 goto out;
410 }
411
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700412 if (!valid_arg_len(bprm, len)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700413 ret = -E2BIG;
414 goto out;
415 }
416
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700417 /* We're going to work our way backwords. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700418 pos = bprm->p;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700419 str += len;
420 bprm->p -= len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421
422 while (len > 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423 int offset, bytes_to_copy;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424
425 offset = pos % PAGE_SIZE;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700426 if (offset == 0)
427 offset = PAGE_SIZE;
428
429 bytes_to_copy = offset;
430 if (bytes_to_copy > len)
431 bytes_to_copy = len;
432
433 offset -= bytes_to_copy;
434 pos -= bytes_to_copy;
435 str -= bytes_to_copy;
436 len -= bytes_to_copy;
437
438 if (!kmapped_page || kpos != (pos & PAGE_MASK)) {
439 struct page *page;
440
441 page = get_arg_page(bprm, pos, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442 if (!page) {
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700443 ret = -E2BIG;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444 goto out;
445 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700446
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700447 if (kmapped_page) {
448 flush_kernel_dcache_page(kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700449 kunmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700450 put_arg_page(kmapped_page);
451 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452 kmapped_page = page;
453 kaddr = kmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700454 kpos = pos & PAGE_MASK;
455 flush_arg_page(bprm, kpos, kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456 }
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700457 if (copy_from_user(kaddr+offset, str, bytes_to_copy)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700458 ret = -EFAULT;
459 goto out;
460 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700461 }
462 }
463 ret = 0;
464out:
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700465 if (kmapped_page) {
466 flush_kernel_dcache_page(kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467 kunmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700468 put_arg_page(kmapped_page);
469 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470 return ret;
471}
472
473/*
474 * Like copy_strings, but get argv and its values from kernel memory.
475 */
476int copy_strings_kernel(int argc,char ** argv, struct linux_binprm *bprm)
477{
478 int r;
479 mm_segment_t oldfs = get_fs();
480 set_fs(KERNEL_DS);
481 r = copy_strings(argc, (char __user * __user *)argv, bprm);
482 set_fs(oldfs);
483 return r;
484}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485EXPORT_SYMBOL(copy_strings_kernel);
486
487#ifdef CONFIG_MMU
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700488
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489/*
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700490 * During bprm_mm_init(), we create a temporary stack at STACK_TOP_MAX. Once
491 * the binfmt code determines where the new stack should reside, we shift it to
492 * its final location. The process proceeds as follows:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493 *
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700494 * 1) Use shift to calculate the new vma endpoints.
495 * 2) Extend vma to cover both the old and new ranges. This ensures the
496 * arguments passed to subsequent functions are consistent.
497 * 3) Move vma's page tables to the new range.
498 * 4) Free up any cleared pgd range.
499 * 5) Shrink the vma to cover only the new range.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700500 */
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700501static int shift_arg_pages(struct vm_area_struct *vma, unsigned long shift)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502{
503 struct mm_struct *mm = vma->vm_mm;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700504 unsigned long old_start = vma->vm_start;
505 unsigned long old_end = vma->vm_end;
506 unsigned long length = old_end - old_start;
507 unsigned long new_start = old_start - shift;
508 unsigned long new_end = old_end - shift;
509 struct mmu_gather *tlb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700511 BUG_ON(new_start > new_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700513 /*
514 * ensure there are no vmas between where we want to go
515 * and where we are
516 */
517 if (vma != find_vma(mm, new_start))
518 return -EFAULT;
519
520 /*
521 * cover the whole range: [new_start, old_end)
522 */
523 vma_adjust(vma, new_start, old_end, vma->vm_pgoff, NULL);
524
525 /*
526 * move the page tables downwards, on failure we rely on
527 * process cleanup to remove whatever mess we made.
528 */
529 if (length != move_page_tables(vma, old_start,
530 vma, new_start, length))
531 return -ENOMEM;
532
533 lru_add_drain();
534 tlb = tlb_gather_mmu(mm, 0);
535 if (new_end > old_start) {
536 /*
537 * when the old and new regions overlap clear from new_end.
538 */
539 free_pgd_range(&tlb, new_end, old_end, new_end,
540 vma->vm_next ? vma->vm_next->vm_start : 0);
541 } else {
542 /*
543 * otherwise, clean from old_start; this is done to not touch
544 * the address space in [new_end, old_start) some architectures
545 * have constraints on va-space that make this illegal (IA64) -
546 * for the others its just a little faster.
547 */
548 free_pgd_range(&tlb, old_start, old_end, new_end,
549 vma->vm_next ? vma->vm_next->vm_start : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550 }
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700551 tlb_finish_mmu(tlb, new_end, old_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700553 /*
554 * shrink the vma to just the new range.
555 */
556 vma_adjust(vma, new_start, new_end, vma->vm_pgoff, NULL);
557
558 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700559}
560
561#define EXTRA_STACK_VM_PAGES 20 /* random */
562
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700563/*
564 * Finalizes the stack vm_area_struct. The flags and permissions are updated,
565 * the stack is optionally relocated, and some extra space is added.
566 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567int setup_arg_pages(struct linux_binprm *bprm,
568 unsigned long stack_top,
569 int executable_stack)
570{
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700571 unsigned long ret;
572 unsigned long stack_shift;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573 struct mm_struct *mm = current->mm;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700574 struct vm_area_struct *vma = bprm->vma;
575 struct vm_area_struct *prev = NULL;
576 unsigned long vm_flags;
577 unsigned long stack_base;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578
579#ifdef CONFIG_STACK_GROWSUP
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580 /* Limit stack size to 1GB */
581 stack_base = current->signal->rlim[RLIMIT_STACK].rlim_max;
582 if (stack_base > (1 << 30))
583 stack_base = 1 << 30;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700585 /* Make sure we didn't let the argument array grow too large. */
586 if (vma->vm_end - vma->vm_start > stack_base)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700587 return -ENOMEM;
588
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700589 stack_base = PAGE_ALIGN(stack_top - stack_base);
590
591 stack_shift = vma->vm_start - stack_base;
592 mm->arg_start = bprm->p - stack_shift;
593 bprm->p = vma->vm_end - stack_shift;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700594#else
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700595 stack_top = arch_align_stack(stack_top);
596 stack_top = PAGE_ALIGN(stack_top);
597 stack_shift = vma->vm_end - stack_top;
598
599 bprm->p -= stack_shift;
600 mm->arg_start = bprm->p;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700601#endif
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700602
603 if (bprm->loader)
604 bprm->loader -= stack_shift;
605 bprm->exec -= stack_shift;
606
607 down_write(&mm->mmap_sem);
608 vm_flags = vma->vm_flags;
609
610 /*
611 * Adjust stack execute permissions; explicitly enable for
612 * EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X and leave alone
613 * (arch default) otherwise.
614 */
615 if (unlikely(executable_stack == EXSTACK_ENABLE_X))
616 vm_flags |= VM_EXEC;
617 else if (executable_stack == EXSTACK_DISABLE_X)
618 vm_flags &= ~VM_EXEC;
619 vm_flags |= mm->def_flags;
620
621 ret = mprotect_fixup(vma, &prev, vma->vm_start, vma->vm_end,
622 vm_flags);
623 if (ret)
624 goto out_unlock;
625 BUG_ON(prev != vma);
626
627 /* Move stack pages down in memory. */
628 if (stack_shift) {
629 ret = shift_arg_pages(vma, stack_shift);
630 if (ret) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700631 up_write(&mm->mmap_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700632 return ret;
633 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634 }
635
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700636#ifdef CONFIG_STACK_GROWSUP
637 stack_base = vma->vm_end + EXTRA_STACK_VM_PAGES * PAGE_SIZE;
638#else
639 stack_base = vma->vm_start - EXTRA_STACK_VM_PAGES * PAGE_SIZE;
640#endif
641 ret = expand_stack(vma, stack_base);
642 if (ret)
643 ret = -EFAULT;
644
645out_unlock:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700646 up_write(&mm->mmap_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700647 return 0;
648}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700649EXPORT_SYMBOL(setup_arg_pages);
650
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651#endif /* CONFIG_MMU */
652
653struct file *open_exec(const char *name)
654{
655 struct nameidata nd;
656 int err;
657 struct file *file;
658
Oleg Drokinb5005312006-03-25 03:07:01 -0800659 err = path_lookup_open(AT_FDCWD, name, LOOKUP_FOLLOW, &nd, FMODE_READ|FMODE_EXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700660 file = ERR_PTR(err);
661
662 if (!err) {
Jan Blunck4ac91372008-02-14 19:34:32 -0800663 struct inode *inode = nd.path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664 file = ERR_PTR(-EACCES);
Miklos Szeredi1a159dd2007-10-16 23:27:08 -0700665 if (S_ISREG(inode->i_mode)) {
Christoph Hellwige4543ed2005-11-08 21:35:04 -0800666 int err = vfs_permission(&nd, MAY_EXEC);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700667 file = ERR_PTR(err);
668 if (!err) {
Andi Kleenabe8be32008-02-08 04:20:23 -0800669 file = nameidata_to_filp(&nd,
670 O_RDONLY|O_LARGEFILE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671 if (!IS_ERR(file)) {
672 err = deny_write_access(file);
673 if (err) {
674 fput(file);
675 file = ERR_PTR(err);
676 }
677 }
678out:
679 return file;
680 }
681 }
Trond Myklebust834f2a42005-10-18 14:20:16 -0700682 release_open_intent(&nd);
Jan Blunck1d957f92008-02-14 19:34:35 -0800683 path_put(&nd.path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700684 }
685 goto out;
686}
687
688EXPORT_SYMBOL(open_exec);
689
690int kernel_read(struct file *file, unsigned long offset,
691 char *addr, unsigned long count)
692{
693 mm_segment_t old_fs;
694 loff_t pos = offset;
695 int result;
696
697 old_fs = get_fs();
698 set_fs(get_ds());
699 /* The cast to a user pointer is valid due to the set_fs() */
700 result = vfs_read(file, (void __user *)addr, count, &pos);
701 set_fs(old_fs);
702 return result;
703}
704
705EXPORT_SYMBOL(kernel_read);
706
707static int exec_mmap(struct mm_struct *mm)
708{
709 struct task_struct *tsk;
710 struct mm_struct * old_mm, *active_mm;
711
712 /* Notify parent that we're no longer interested in the old VM */
713 tsk = current;
714 old_mm = current->mm;
715 mm_release(tsk, old_mm);
716
717 if (old_mm) {
718 /*
719 * Make sure that if there is a core dump in progress
720 * for the old mm, we get out and die instead of going
721 * through with the exec. We must hold mmap_sem around
722 * checking core_waiters and changing tsk->mm. The
723 * core-inducing thread will increment core_waiters for
724 * each thread whose ->mm == old_mm.
725 */
726 down_read(&old_mm->mmap_sem);
727 if (unlikely(old_mm->core_waiters)) {
728 up_read(&old_mm->mmap_sem);
729 return -EINTR;
730 }
731 }
732 task_lock(tsk);
733 active_mm = tsk->active_mm;
734 tsk->mm = mm;
735 tsk->active_mm = mm;
736 activate_mm(active_mm, mm);
737 task_unlock(tsk);
Balbir Singhcf475ad2008-04-29 01:00:16 -0700738 mm_update_next_owner(mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700739 arch_pick_mmap_layout(mm);
740 if (old_mm) {
741 up_read(&old_mm->mmap_sem);
Eric Sesterhenn7dddb122006-04-01 01:13:38 +0200742 BUG_ON(active_mm != old_mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743 mmput(old_mm);
744 return 0;
745 }
746 mmdrop(active_mm);
747 return 0;
748}
749
750/*
751 * This function makes sure the current process has its own signal table,
752 * so that flush_signal_handlers can later reset the handlers without
753 * disturbing other processes. (Other processes might share the signal
754 * table via the CLONE_SIGHAND option to clone().)
755 */
Arjan van de Ven858119e2006-01-14 13:20:43 -0800756static int de_thread(struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700757{
758 struct signal_struct *sig = tsk->signal;
Oleg Nesterovb2c903b2007-10-16 23:27:22 -0700759 struct sighand_struct *oldsighand = tsk->sighand;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700760 spinlock_t *lock = &oldsighand->siglock;
Oleg Nesterov329f7db2005-11-07 21:12:43 +0300761 struct task_struct *leader = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762 int count;
763
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700764 if (thread_group_empty(tsk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700765 goto no_thread_group;
766
767 /*
768 * Kill all other threads in the thread group.
769 * We must hold tasklist_lock to call zap_other_threads.
770 */
771 read_lock(&tasklist_lock);
772 spin_lock_irq(lock);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -0800773 if (signal_group_exit(sig)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774 /*
775 * Another group action in progress, just
776 * return so that the signal is processed.
777 */
778 spin_unlock_irq(lock);
779 read_unlock(&tasklist_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700780 return -EAGAIN;
781 }
Oleg Nesterov14342612006-03-28 16:10:59 -0800782
783 /*
784 * child_reaper ignores SIGKILL, change it now.
785 * Reparenting needs write_lock on tasklist_lock,
786 * so it is safe to do it under read_lock.
787 */
Sukadev Bhattiprolu88f21d82007-10-18 23:39:50 -0700788 if (unlikely(tsk->group_leader == task_child_reaper(tsk)))
Sukadev Bhattiprolu2894d6502007-10-18 23:39:49 -0700789 task_active_pid_ns(tsk)->child_reaper = tsk;
Oleg Nesterov14342612006-03-28 16:10:59 -0800790
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -0800791 sig->group_exit_task = tsk;
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700792 zap_other_threads(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793 read_unlock(&tasklist_lock);
794
Oleg Nesterovfea9d172008-02-08 04:19:19 -0800795 /* Account for the thread group leader hanging around: */
796 count = thread_group_leader(tsk) ? 1 : 2;
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700797 sig->notify_count = count;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700798 while (atomic_read(&sig->count) > count) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799 __set_current_state(TASK_UNINTERRUPTIBLE);
800 spin_unlock_irq(lock);
801 schedule();
802 spin_lock_irq(lock);
803 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700804 spin_unlock_irq(lock);
805
806 /*
807 * At this point all other threads have exited, all we have to
808 * do is to wait for the thread group leader to become inactive,
809 * and to assume its PID:
810 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700811 if (!thread_group_leader(tsk)) {
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700812 leader = tsk->group_leader;
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700813
814 sig->notify_count = -1;
815 for (;;) {
816 write_lock_irq(&tasklist_lock);
817 if (likely(leader->exit_state))
818 break;
819 __set_current_state(TASK_UNINTERRUPTIBLE);
820 write_unlock_irq(&tasklist_lock);
821 schedule();
822 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823
Roland McGrathf5e90282006-04-10 22:54:16 -0700824 /*
825 * The only record we have of the real-time age of a
826 * process, regardless of execs it's done, is start_time.
827 * All the past CPU time is accumulated in signal_struct
828 * from sister threads now dead. But in this non-leader
829 * exec, nothing survives from the original leader thread,
830 * whose birth marks the true age of this process now.
831 * When we take on its identity by switching to its PID, we
832 * also take its birthdate (always earlier than our own).
833 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700834 tsk->start_time = leader->start_time;
Roland McGrathf5e90282006-04-10 22:54:16 -0700835
Pavel Emelyanovbac0abd2007-10-18 23:40:18 -0700836 BUG_ON(!same_thread_group(leader, tsk));
837 BUG_ON(has_group_leader_pid(tsk));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838 /*
839 * An exec() starts a new thread group with the
840 * TGID of the previous thread group. Rehash the
841 * two threads with a switched PID, and release
842 * the former thread group leader:
843 */
Eric W. Biedermand73d6522006-03-28 16:11:03 -0800844
845 /* Become a process group leader with the old leader's pid.
Eric W. Biedermanc18258c2006-09-27 01:51:06 -0700846 * The old leader becomes a thread of the this thread group.
847 * Note: The old leader also uses this pid until release_task
Eric W. Biedermand73d6522006-03-28 16:11:03 -0800848 * is called. Odd but simple and correct.
849 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700850 detach_pid(tsk, PIDTYPE_PID);
851 tsk->pid = leader->pid;
Sukadev Bhattiprolu3743ca02007-10-18 23:39:51 -0700852 attach_pid(tsk, PIDTYPE_PID, task_pid(leader));
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700853 transfer_pid(leader, tsk, PIDTYPE_PGID);
854 transfer_pid(leader, tsk, PIDTYPE_SID);
855 list_replace_rcu(&leader->tasks, &tsk->tasks);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700856
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700857 tsk->group_leader = tsk;
858 leader->group_leader = tsk;
Eric W. Biedermande12a782006-04-10 17:16:49 -0600859
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700860 tsk->exit_signal = SIGCHLD;
Oleg Nesterov962b5642005-11-23 13:37:43 -0800861
862 BUG_ON(leader->exit_state != EXIT_ZOMBIE);
863 leader->exit_state = EXIT_DEAD;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864
865 write_unlock_irq(&tasklist_lock);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -0800866 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700867
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700868 sig->group_exit_task = NULL;
869 sig->notify_count = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700870
871no_thread_group:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700872 exit_itimers(sig);
Oleg Nesterov329f7db2005-11-07 21:12:43 +0300873 if (leader)
874 release_task(leader);
875
Oleg Nesterovb2c903b2007-10-16 23:27:22 -0700876 if (atomic_read(&oldsighand->count) != 1) {
877 struct sighand_struct *newsighand;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878 /*
Oleg Nesterovb2c903b2007-10-16 23:27:22 -0700879 * This ->sighand is shared with the CLONE_SIGHAND
880 * but not CLONE_THREAD task, switch to the new one.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700881 */
Oleg Nesterovb2c903b2007-10-16 23:27:22 -0700882 newsighand = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
883 if (!newsighand)
884 return -ENOMEM;
885
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886 atomic_set(&newsighand->count, 1);
887 memcpy(newsighand->action, oldsighand->action,
888 sizeof(newsighand->action));
889
890 write_lock_irq(&tasklist_lock);
891 spin_lock(&oldsighand->siglock);
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700892 rcu_assign_pointer(tsk->sighand, newsighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893 spin_unlock(&oldsighand->siglock);
894 write_unlock_irq(&tasklist_lock);
895
Davide Libenzifba2afa2007-05-10 22:23:13 -0700896 __cleanup_sighand(oldsighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700897 }
898
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700899 BUG_ON(!thread_group_leader(tsk));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700900 return 0;
901}
Oleg Nesterov0840a902007-10-16 23:27:22 -0700902
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903/*
904 * These functions flushes out all traces of the currently running executable
905 * so that a new one can be started
906 */
Arjan van de Ven858119e2006-01-14 13:20:43 -0800907static void flush_old_files(struct files_struct * files)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700908{
909 long j = -1;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700910 struct fdtable *fdt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700911
912 spin_lock(&files->file_lock);
913 for (;;) {
914 unsigned long set, i;
915
916 j++;
917 i = j * __NFDBITS;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700918 fdt = files_fdtable(files);
Vadim Lobanovbbea9f62006-12-10 02:21:12 -0800919 if (i >= fdt->max_fds)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700920 break;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700921 set = fdt->close_on_exec->fds_bits[j];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922 if (!set)
923 continue;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700924 fdt->close_on_exec->fds_bits[j] = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700925 spin_unlock(&files->file_lock);
926 for ( ; set ; i++,set >>= 1) {
927 if (set & 1) {
928 sys_close(i);
929 }
930 }
931 spin_lock(&files->file_lock);
932
933 }
934 spin_unlock(&files->file_lock);
935}
936
Andrew Morton59714d62008-02-04 22:27:21 -0800937char *get_task_comm(char *buf, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700938{
939 /* buf must be at least sizeof(tsk->comm) in size */
940 task_lock(tsk);
941 strncpy(buf, tsk->comm, sizeof(tsk->comm));
942 task_unlock(tsk);
Andrew Morton59714d62008-02-04 22:27:21 -0800943 return buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700944}
945
946void set_task_comm(struct task_struct *tsk, char *buf)
947{
948 task_lock(tsk);
949 strlcpy(tsk->comm, buf, sizeof(tsk->comm));
950 task_unlock(tsk);
951}
952
953int flush_old_exec(struct linux_binprm * bprm)
954{
955 char * name;
956 int i, ch, retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700957 char tcomm[sizeof(current->comm)];
958
959 /*
960 * Make sure we have a private signal table and that
961 * we are unassociated from the previous thread group.
962 */
963 retval = de_thread(current);
964 if (retval)
965 goto out;
966
967 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968 * Release all of the old mmap stuff
969 */
970 retval = exec_mmap(bprm->mm);
971 if (retval)
Al Virofd8328b2008-04-22 05:11:59 -0400972 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700973
974 bprm->mm = NULL; /* We're using it now */
975
976 /* This is the point of no return */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977 current->sas_ss_sp = current->sas_ss_size = 0;
978
979 if (current->euid == current->uid && current->egid == current->gid)
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -0700980 set_dumpable(current->mm, 1);
Alan Coxd6e71142005-06-23 00:09:43 -0700981 else
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -0700982 set_dumpable(current->mm, suid_dumpable);
Alan Coxd6e71142005-06-23 00:09:43 -0700983
Linus Torvalds1da177e2005-04-16 15:20:36 -0700984 name = bprm->filename;
Paolo 'Blaisorblade' Giarrusso36772092005-05-05 16:16:12 -0700985
986 /* Copies the binary name from after last slash */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700987 for (i=0; (ch = *(name++)) != '\0';) {
988 if (ch == '/')
Paolo 'Blaisorblade' Giarrusso36772092005-05-05 16:16:12 -0700989 i = 0; /* overwrite what we wrote */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990 else
991 if (i < (sizeof(tcomm) - 1))
992 tcomm[i++] = ch;
993 }
994 tcomm[i] = '\0';
995 set_task_comm(current, tcomm);
996
997 current->flags &= ~PF_RANDOMIZE;
998 flush_thread();
999
Benjamin Herrenschmidt0551fbd2006-02-28 16:59:19 -08001000 /* Set the new mm task size. We have to do that late because it may
1001 * depend on TIF_32BIT which is only updated in flush_thread() on
1002 * some architectures like powerpc
1003 */
1004 current->mm->task_size = TASK_SIZE;
1005
Marcel Holtmannd2d56c52007-08-17 21:47:58 +02001006 if (bprm->e_uid != current->euid || bprm->e_gid != current->egid) {
1007 suid_keys(current);
1008 set_dumpable(current->mm, suid_dumpable);
1009 current->pdeath_signal = 0;
1010 } else if (file_permission(bprm->file, MAY_READ) ||
1011 (bprm->interp_flags & BINPRM_FLAGS_ENFORCE_NONDUMP)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012 suid_keys(current);
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001013 set_dumpable(current->mm, suid_dumpable);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014 }
1015
1016 /* An exec changes our domain. We are no longer part of the thread
1017 group */
1018
1019 current->self_exec_id++;
1020
1021 flush_signal_handlers(current, 0);
1022 flush_old_files(current->files);
1023
1024 return 0;
1025
Linus Torvalds1da177e2005-04-16 15:20:36 -07001026out:
1027 return retval;
1028}
1029
1030EXPORT_SYMBOL(flush_old_exec);
1031
1032/*
1033 * Fill the binprm structure from the inode.
1034 * Check permissions, then read the first 128 (BINPRM_BUF_SIZE) bytes
1035 */
1036int prepare_binprm(struct linux_binprm *bprm)
1037{
1038 int mode;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001039 struct inode * inode = bprm->file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001040 int retval;
1041
1042 mode = inode->i_mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001043 if (bprm->file->f_op == NULL)
1044 return -EACCES;
1045
1046 bprm->e_uid = current->euid;
1047 bprm->e_gid = current->egid;
1048
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001049 if(!(bprm->file->f_path.mnt->mnt_flags & MNT_NOSUID)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001050 /* Set-uid? */
1051 if (mode & S_ISUID) {
1052 current->personality &= ~PER_CLEAR_ON_SETID;
1053 bprm->e_uid = inode->i_uid;
1054 }
1055
1056 /* Set-gid? */
1057 /*
1058 * If setgid is set but no group execute bit then this
1059 * is a candidate for mandatory locking, not a setgid
1060 * executable.
1061 */
1062 if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) {
1063 current->personality &= ~PER_CLEAR_ON_SETID;
1064 bprm->e_gid = inode->i_gid;
1065 }
1066 }
1067
1068 /* fill in binprm security blob */
1069 retval = security_bprm_set(bprm);
1070 if (retval)
1071 return retval;
1072
1073 memset(bprm->buf,0,BINPRM_BUF_SIZE);
1074 return kernel_read(bprm->file,0,bprm->buf,BINPRM_BUF_SIZE);
1075}
1076
1077EXPORT_SYMBOL(prepare_binprm);
1078
Arjan van de Ven858119e2006-01-14 13:20:43 -08001079static int unsafe_exec(struct task_struct *p)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001080{
1081 int unsafe = 0;
1082 if (p->ptrace & PT_PTRACED) {
1083 if (p->ptrace & PT_PTRACE_CAP)
1084 unsafe |= LSM_UNSAFE_PTRACE_CAP;
1085 else
1086 unsafe |= LSM_UNSAFE_PTRACE;
1087 }
1088 if (atomic_read(&p->fs->count) > 1 ||
1089 atomic_read(&p->files->count) > 1 ||
1090 atomic_read(&p->sighand->count) > 1)
1091 unsafe |= LSM_UNSAFE_SHARE;
1092
1093 return unsafe;
1094}
1095
1096void compute_creds(struct linux_binprm *bprm)
1097{
1098 int unsafe;
1099
Marcel Holtmannd2d56c52007-08-17 21:47:58 +02001100 if (bprm->e_uid != current->uid) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001101 suid_keys(current);
Marcel Holtmannd2d56c52007-08-17 21:47:58 +02001102 current->pdeath_signal = 0;
1103 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001104 exec_keys(current);
1105
1106 task_lock(current);
1107 unsafe = unsafe_exec(current);
1108 security_bprm_apply_creds(bprm, unsafe);
1109 task_unlock(current);
1110 security_bprm_post_apply_creds(bprm);
1111}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112EXPORT_SYMBOL(compute_creds);
1113
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001114/*
1115 * Arguments are '\0' separated strings found at the location bprm->p
1116 * points to; chop off the first by relocating brpm->p to right after
1117 * the first '\0' encountered.
1118 */
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001119int remove_arg_zero(struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001120{
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001121 int ret = 0;
1122 unsigned long offset;
1123 char *kaddr;
1124 struct page *page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001125
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001126 if (!bprm->argc)
1127 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001128
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001129 do {
1130 offset = bprm->p & ~PAGE_MASK;
1131 page = get_arg_page(bprm, bprm->p, 0);
1132 if (!page) {
1133 ret = -EFAULT;
1134 goto out;
1135 }
1136 kaddr = kmap_atomic(page, KM_USER0);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001137
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001138 for (; offset < PAGE_SIZE && kaddr[offset];
1139 offset++, bprm->p++)
1140 ;
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001141
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001142 kunmap_atomic(kaddr, KM_USER0);
1143 put_arg_page(page);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001144
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001145 if (offset == PAGE_SIZE)
1146 free_arg_page(bprm, (bprm->p >> PAGE_SHIFT) - 1);
1147 } while (offset == PAGE_SIZE);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001148
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001149 bprm->p++;
1150 bprm->argc--;
1151 ret = 0;
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001152
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001153out:
1154 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001155}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001156EXPORT_SYMBOL(remove_arg_zero);
1157
1158/*
1159 * cycle the list of binary formats handler, until one recognizes the image
1160 */
1161int search_binary_handler(struct linux_binprm *bprm,struct pt_regs *regs)
1162{
1163 int try,retval;
1164 struct linux_binfmt *fmt;
David Howells7fa30312008-02-08 04:19:28 -08001165#if defined(__alpha__) && defined(CONFIG_ARCH_SUPPORTS_AOUT)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001166 /* handle /sbin/loader.. */
1167 {
1168 struct exec * eh = (struct exec *) bprm->buf;
1169
1170 if (!bprm->loader && eh->fh.f_magic == 0x183 &&
1171 (eh->fh.f_flags & 0x3000) == 0x3000)
1172 {
1173 struct file * file;
1174 unsigned long loader;
1175
1176 allow_write_access(bprm->file);
1177 fput(bprm->file);
1178 bprm->file = NULL;
1179
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001180 loader = bprm->vma->vm_end - sizeof(void *);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001181
1182 file = open_exec("/sbin/loader");
1183 retval = PTR_ERR(file);
1184 if (IS_ERR(file))
1185 return retval;
1186
1187 /* Remember if the application is TASO. */
1188 bprm->sh_bang = eh->ah.entry < 0x100000000UL;
1189
1190 bprm->file = file;
1191 bprm->loader = loader;
1192 retval = prepare_binprm(bprm);
1193 if (retval<0)
1194 return retval;
1195 /* should call search_binary_handler recursively here,
1196 but it does not matter */
1197 }
1198 }
1199#endif
1200 retval = security_bprm_check(bprm);
1201 if (retval)
1202 return retval;
1203
1204 /* kernel module loader fixup */
1205 /* so we don't try to load run modprobe in kernel space. */
1206 set_fs(USER_DS);
Al Viro473ae302006-04-26 14:04:08 -04001207
1208 retval = audit_bprm(bprm);
1209 if (retval)
1210 return retval;
1211
Linus Torvalds1da177e2005-04-16 15:20:36 -07001212 retval = -ENOENT;
1213 for (try=0; try<2; try++) {
1214 read_lock(&binfmt_lock);
Alexey Dobriyane4dc1b12007-10-16 23:26:03 -07001215 list_for_each_entry(fmt, &formats, lh) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001216 int (*fn)(struct linux_binprm *, struct pt_regs *) = fmt->load_binary;
1217 if (!fn)
1218 continue;
1219 if (!try_module_get(fmt->module))
1220 continue;
1221 read_unlock(&binfmt_lock);
1222 retval = fn(bprm, regs);
1223 if (retval >= 0) {
1224 put_binfmt(fmt);
1225 allow_write_access(bprm->file);
1226 if (bprm->file)
1227 fput(bprm->file);
1228 bprm->file = NULL;
1229 current->did_exec = 1;
Matt Helsley9f460802005-11-07 00:59:16 -08001230 proc_exec_connector(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001231 return retval;
1232 }
1233 read_lock(&binfmt_lock);
1234 put_binfmt(fmt);
1235 if (retval != -ENOEXEC || bprm->mm == NULL)
1236 break;
1237 if (!bprm->file) {
1238 read_unlock(&binfmt_lock);
1239 return retval;
1240 }
1241 }
1242 read_unlock(&binfmt_lock);
1243 if (retval != -ENOEXEC || bprm->mm == NULL) {
1244 break;
1245#ifdef CONFIG_KMOD
1246 }else{
1247#define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e))
1248 if (printable(bprm->buf[0]) &&
1249 printable(bprm->buf[1]) &&
1250 printable(bprm->buf[2]) &&
1251 printable(bprm->buf[3]))
1252 break; /* -ENOEXEC */
1253 request_module("binfmt-%04x", *(unsigned short *)(&bprm->buf[2]));
1254#endif
1255 }
1256 }
1257 return retval;
1258}
1259
1260EXPORT_SYMBOL(search_binary_handler);
1261
1262/*
1263 * sys_execve() executes a new program.
1264 */
1265int do_execve(char * filename,
1266 char __user *__user *argv,
1267 char __user *__user *envp,
1268 struct pt_regs * regs)
1269{
1270 struct linux_binprm *bprm;
1271 struct file *file;
Al Viro3b125382008-04-22 05:31:30 -04001272 struct files_struct *displaced;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001273 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274
Al Viro3b125382008-04-22 05:31:30 -04001275 retval = unshare_files(&displaced);
Al Virofd8328b2008-04-22 05:11:59 -04001276 if (retval)
1277 goto out_ret;
1278
Linus Torvalds1da177e2005-04-16 15:20:36 -07001279 retval = -ENOMEM;
Oliver Neukum11b0b5a2006-03-25 03:08:13 -08001280 bprm = kzalloc(sizeof(*bprm), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001281 if (!bprm)
Al Virofd8328b2008-04-22 05:11:59 -04001282 goto out_files;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001283
1284 file = open_exec(filename);
1285 retval = PTR_ERR(file);
1286 if (IS_ERR(file))
1287 goto out_kfree;
1288
1289 sched_exec();
1290
Linus Torvalds1da177e2005-04-16 15:20:36 -07001291 bprm->file = file;
1292 bprm->filename = filename;
1293 bprm->interp = filename;
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001294
1295 retval = bprm_mm_init(bprm);
1296 if (retval)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001297 goto out_file;
1298
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001299 bprm->argc = count(argv, MAX_ARG_STRINGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001300 if ((retval = bprm->argc) < 0)
1301 goto out_mm;
1302
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001303 bprm->envc = count(envp, MAX_ARG_STRINGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001304 if ((retval = bprm->envc) < 0)
1305 goto out_mm;
1306
1307 retval = security_bprm_alloc(bprm);
1308 if (retval)
1309 goto out;
1310
1311 retval = prepare_binprm(bprm);
1312 if (retval < 0)
1313 goto out;
1314
1315 retval = copy_strings_kernel(1, &bprm->filename, bprm);
1316 if (retval < 0)
1317 goto out;
1318
1319 bprm->exec = bprm->p;
1320 retval = copy_strings(bprm->envc, envp, bprm);
1321 if (retval < 0)
1322 goto out;
1323
1324 retval = copy_strings(bprm->argc, argv, bprm);
1325 if (retval < 0)
1326 goto out;
1327
1328 retval = search_binary_handler(bprm,regs);
1329 if (retval >= 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001330 /* execve success */
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001331 free_arg_pages(bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001332 security_bprm_free(bprm);
1333 acct_update_integrals(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001334 kfree(bprm);
Al Viro3b125382008-04-22 05:31:30 -04001335 if (displaced)
1336 put_files_struct(displaced);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001337 return retval;
1338 }
1339
1340out:
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001341 free_arg_pages(bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001342 if (bprm->security)
1343 security_bprm_free(bprm);
1344
1345out_mm:
1346 if (bprm->mm)
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001347 mmput (bprm->mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001348
1349out_file:
1350 if (bprm->file) {
1351 allow_write_access(bprm->file);
1352 fput(bprm->file);
1353 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001354out_kfree:
1355 kfree(bprm);
1356
Al Virofd8328b2008-04-22 05:11:59 -04001357out_files:
Al Viro3b125382008-04-22 05:31:30 -04001358 if (displaced)
1359 reset_files_struct(displaced);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001360out_ret:
1361 return retval;
1362}
1363
1364int set_binfmt(struct linux_binfmt *new)
1365{
1366 struct linux_binfmt *old = current->binfmt;
1367
1368 if (new) {
1369 if (!try_module_get(new->module))
1370 return -1;
1371 }
1372 current->binfmt = new;
1373 if (old)
1374 module_put(old->module);
1375 return 0;
1376}
1377
1378EXPORT_SYMBOL(set_binfmt);
1379
Linus Torvalds1da177e2005-04-16 15:20:36 -07001380/* format_corename will inspect the pattern parameter, and output a
1381 * name into corename, which must have space for at least
1382 * CORENAME_MAX_SIZE bytes plus one byte for the zero terminator.
1383 */
Alan Coxc4bbafd2007-04-16 22:53:13 -07001384static int format_corename(char *corename, const char *pattern, long signr)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001385{
1386 const char *pat_ptr = pattern;
1387 char *out_ptr = corename;
1388 char *const out_end = corename + CORENAME_MAX_SIZE;
1389 int rc;
1390 int pid_in_pattern = 0;
Alan Coxc4bbafd2007-04-16 22:53:13 -07001391 int ispipe = 0;
1392
1393 if (*pattern == '|')
1394 ispipe = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001395
1396 /* Repeat as long as we have more pattern to process and more output
1397 space */
1398 while (*pat_ptr) {
1399 if (*pat_ptr != '%') {
1400 if (out_ptr == out_end)
1401 goto out;
1402 *out_ptr++ = *pat_ptr++;
1403 } else {
1404 switch (*++pat_ptr) {
1405 case 0:
1406 goto out;
1407 /* Double percent, output one percent */
1408 case '%':
1409 if (out_ptr == out_end)
1410 goto out;
1411 *out_ptr++ = '%';
1412 break;
1413 /* pid */
1414 case 'p':
1415 pid_in_pattern = 1;
1416 rc = snprintf(out_ptr, out_end - out_ptr,
Pavel Emelyanovb4888932007-10-18 23:40:14 -07001417 "%d", task_tgid_vnr(current));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001418 if (rc > out_end - out_ptr)
1419 goto out;
1420 out_ptr += rc;
1421 break;
1422 /* uid */
1423 case 'u':
1424 rc = snprintf(out_ptr, out_end - out_ptr,
1425 "%d", current->uid);
1426 if (rc > out_end - out_ptr)
1427 goto out;
1428 out_ptr += rc;
1429 break;
1430 /* gid */
1431 case 'g':
1432 rc = snprintf(out_ptr, out_end - out_ptr,
1433 "%d", current->gid);
1434 if (rc > out_end - out_ptr)
1435 goto out;
1436 out_ptr += rc;
1437 break;
1438 /* signal that caused the coredump */
1439 case 's':
1440 rc = snprintf(out_ptr, out_end - out_ptr,
1441 "%ld", signr);
1442 if (rc > out_end - out_ptr)
1443 goto out;
1444 out_ptr += rc;
1445 break;
1446 /* UNIX time of coredump */
1447 case 't': {
1448 struct timeval tv;
1449 do_gettimeofday(&tv);
1450 rc = snprintf(out_ptr, out_end - out_ptr,
1451 "%lu", tv.tv_sec);
1452 if (rc > out_end - out_ptr)
1453 goto out;
1454 out_ptr += rc;
1455 break;
1456 }
1457 /* hostname */
1458 case 'h':
1459 down_read(&uts_sem);
1460 rc = snprintf(out_ptr, out_end - out_ptr,
Serge E. Hallyne9ff3992006-10-02 02:18:11 -07001461 "%s", utsname()->nodename);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001462 up_read(&uts_sem);
1463 if (rc > out_end - out_ptr)
1464 goto out;
1465 out_ptr += rc;
1466 break;
1467 /* executable */
1468 case 'e':
1469 rc = snprintf(out_ptr, out_end - out_ptr,
1470 "%s", current->comm);
1471 if (rc > out_end - out_ptr)
1472 goto out;
1473 out_ptr += rc;
1474 break;
Neil Horman74aadce2007-10-16 23:26:35 -07001475 /* core limit size */
1476 case 'c':
1477 rc = snprintf(out_ptr, out_end - out_ptr,
1478 "%lu", current->signal->rlim[RLIMIT_CORE].rlim_cur);
1479 if (rc > out_end - out_ptr)
1480 goto out;
1481 out_ptr += rc;
1482 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001483 default:
1484 break;
1485 }
1486 ++pat_ptr;
1487 }
1488 }
1489 /* Backward compatibility with core_uses_pid:
1490 *
1491 * If core_pattern does not include a %p (as is the default)
1492 * and core_uses_pid is set, then .%pid will be appended to
Alan Coxc4bbafd2007-04-16 22:53:13 -07001493 * the filename. Do not do this for piped commands. */
1494 if (!ispipe && !pid_in_pattern
Linus Torvalds1da177e2005-04-16 15:20:36 -07001495 && (core_uses_pid || atomic_read(&current->mm->mm_users) != 1)) {
1496 rc = snprintf(out_ptr, out_end - out_ptr,
Pavel Emelyanovb4888932007-10-18 23:40:14 -07001497 ".%d", task_tgid_vnr(current));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001498 if (rc > out_end - out_ptr)
1499 goto out;
1500 out_ptr += rc;
1501 }
Alan Coxc4bbafd2007-04-16 22:53:13 -07001502out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001503 *out_ptr = 0;
Alan Coxc4bbafd2007-04-16 22:53:13 -07001504 return ispipe;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001505}
1506
Oleg Nesterovd5f70c02006-06-26 00:26:07 -07001507static void zap_process(struct task_struct *start)
Oleg Nesterovaceecc042006-06-26 00:26:05 -07001508{
1509 struct task_struct *t;
Oleg Nesterov281de332006-06-26 00:26:06 -07001510
Oleg Nesterovd5f70c02006-06-26 00:26:07 -07001511 start->signal->flags = SIGNAL_GROUP_EXIT;
1512 start->signal->group_stop_count = 0;
Oleg Nesterovaceecc042006-06-26 00:26:05 -07001513
1514 t = start;
1515 do {
1516 if (t != current && t->mm) {
1517 t->mm->core_waiters++;
Oleg Nesterov281de332006-06-26 00:26:06 -07001518 sigaddset(&t->pending.signal, SIGKILL);
1519 signal_wake_up(t, 1);
Oleg Nesterovaceecc042006-06-26 00:26:05 -07001520 }
1521 } while ((t = next_thread(t)) != start);
1522}
1523
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001524static inline int zap_threads(struct task_struct *tsk, struct mm_struct *mm,
1525 int exit_code)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001526{
1527 struct task_struct *g, *p;
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001528 unsigned long flags;
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001529 int err = -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001530
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001531 spin_lock_irq(&tsk->sighand->siglock);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -08001532 if (!signal_group_exit(tsk->signal)) {
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001533 tsk->signal->group_exit_code = exit_code;
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001534 zap_process(tsk);
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001535 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001536 }
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001537 spin_unlock_irq(&tsk->sighand->siglock);
1538 if (err)
1539 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001540
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001541 if (atomic_read(&mm->mm_users) == mm->core_waiters + 1)
1542 goto done;
1543
Oleg Nesterov7b1c6152006-06-26 00:26:08 -07001544 rcu_read_lock();
Oleg Nesterovaceecc042006-06-26 00:26:05 -07001545 for_each_process(g) {
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001546 if (g == tsk->group_leader)
1547 continue;
1548
Oleg Nesterovaceecc042006-06-26 00:26:05 -07001549 p = g;
1550 do {
1551 if (p->mm) {
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001552 if (p->mm == mm) {
1553 /*
1554 * p->sighand can't disappear, but
1555 * may be changed by de_thread()
1556 */
1557 lock_task_sighand(p, &flags);
Oleg Nesterovd5f70c02006-06-26 00:26:07 -07001558 zap_process(p);
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001559 unlock_task_sighand(p, &flags);
1560 }
Oleg Nesterovaceecc042006-06-26 00:26:05 -07001561 break;
1562 }
1563 } while ((p = next_thread(p)) != g);
1564 }
Oleg Nesterov7b1c6152006-06-26 00:26:08 -07001565 rcu_read_unlock();
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001566done:
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001567 return mm->core_waiters;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001568}
1569
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001570static int coredump_wait(int exit_code)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001571{
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001572 struct task_struct *tsk = current;
1573 struct mm_struct *mm = tsk->mm;
1574 struct completion startup_done;
1575 struct completion *vfork_done;
Oleg Nesterov2384f552005-10-30 15:02:47 -08001576 int core_waiters;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001577
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001578 init_completion(&mm->core_done);
1579 init_completion(&startup_done);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001580 mm->core_startup_done = &startup_done;
1581
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001582 core_waiters = zap_threads(tsk, mm, exit_code);
Oleg Nesterov2384f552005-10-30 15:02:47 -08001583 up_write(&mm->mmap_sem);
1584
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001585 if (unlikely(core_waiters < 0))
1586 goto fail;
1587
1588 /*
1589 * Make sure nobody is waiting for us to release the VM,
1590 * otherwise we can deadlock when we wait on each other
1591 */
1592 vfork_done = tsk->vfork_done;
1593 if (vfork_done) {
1594 tsk->vfork_done = NULL;
1595 complete(vfork_done);
1596 }
1597
Oleg Nesterov2384f552005-10-30 15:02:47 -08001598 if (core_waiters)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001599 wait_for_completion(&startup_done);
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001600fail:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001601 BUG_ON(mm->core_waiters);
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001602 return core_waiters;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001603}
1604
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001605/*
1606 * set_dumpable converts traditional three-value dumpable to two flags and
1607 * stores them into mm->flags. It modifies lower two bits of mm->flags, but
1608 * these bits are not changed atomically. So get_dumpable can observe the
1609 * intermediate state. To avoid doing unexpected behavior, get get_dumpable
1610 * return either old dumpable or new one by paying attention to the order of
1611 * modifying the bits.
1612 *
1613 * dumpable | mm->flags (binary)
1614 * old new | initial interim final
1615 * ---------+-----------------------
1616 * 0 1 | 00 01 01
1617 * 0 2 | 00 10(*) 11
1618 * 1 0 | 01 00 00
1619 * 1 2 | 01 11 11
1620 * 2 0 | 11 10(*) 00
1621 * 2 1 | 11 11 01
1622 *
1623 * (*) get_dumpable regards interim value of 10 as 11.
1624 */
1625void set_dumpable(struct mm_struct *mm, int value)
1626{
1627 switch (value) {
1628 case 0:
1629 clear_bit(MMF_DUMPABLE, &mm->flags);
1630 smp_wmb();
1631 clear_bit(MMF_DUMP_SECURELY, &mm->flags);
1632 break;
1633 case 1:
1634 set_bit(MMF_DUMPABLE, &mm->flags);
1635 smp_wmb();
1636 clear_bit(MMF_DUMP_SECURELY, &mm->flags);
1637 break;
1638 case 2:
1639 set_bit(MMF_DUMP_SECURELY, &mm->flags);
1640 smp_wmb();
1641 set_bit(MMF_DUMPABLE, &mm->flags);
1642 break;
1643 }
1644}
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001645
1646int get_dumpable(struct mm_struct *mm)
1647{
1648 int ret;
1649
1650 ret = mm->flags & 0x3;
1651 return (ret >= 2) ? 2 : ret;
1652}
1653
Linus Torvalds1da177e2005-04-16 15:20:36 -07001654int do_coredump(long signr, int exit_code, struct pt_regs * regs)
1655{
1656 char corename[CORENAME_MAX_SIZE + 1];
1657 struct mm_struct *mm = current->mm;
1658 struct linux_binfmt * binfmt;
1659 struct inode * inode;
1660 struct file * file;
1661 int retval = 0;
Alan Coxd6e71142005-06-23 00:09:43 -07001662 int fsuid = current->fsuid;
1663 int flag = 0;
Andi Kleend025c9d2006-09-30 23:29:28 -07001664 int ispipe = 0;
Neil Horman7dc0b222007-10-16 23:26:34 -07001665 unsigned long core_limit = current->signal->rlim[RLIMIT_CORE].rlim_cur;
Neil Horman74aadce2007-10-16 23:26:35 -07001666 char **helper_argv = NULL;
1667 int helper_argc = 0;
1668 char *delimit;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001669
Steve Grubb0a4ff8c2007-04-19 10:28:21 -04001670 audit_core_dumps(signr);
1671
Linus Torvalds1da177e2005-04-16 15:20:36 -07001672 binfmt = current->binfmt;
1673 if (!binfmt || !binfmt->core_dump)
1674 goto fail;
1675 down_write(&mm->mmap_sem);
Roland McGrath00ec99d2007-11-11 19:13:43 -08001676 /*
1677 * If another thread got here first, or we are not dumpable, bail out.
1678 */
1679 if (mm->core_waiters || !get_dumpable(mm)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001680 up_write(&mm->mmap_sem);
1681 goto fail;
1682 }
Alan Coxd6e71142005-06-23 00:09:43 -07001683
1684 /*
1685 * We cannot trust fsuid as being the "true" uid of the
1686 * process nor do we know its entire history. We only know it
1687 * was tainted so we dump it as root in mode 2.
1688 */
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001689 if (get_dumpable(mm) == 2) { /* Setuid core dump mode */
Alan Coxd6e71142005-06-23 00:09:43 -07001690 flag = O_EXCL; /* Stop rewrite attacks */
1691 current->fsuid = 0; /* Dump root private */
1692 }
Oleg Nesterov1291cf42005-10-30 15:02:54 -08001693
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001694 retval = coredump_wait(exit_code);
1695 if (retval < 0)
Oleg Nesterov1291cf42005-10-30 15:02:54 -08001696 goto fail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001697
1698 /*
1699 * Clear any false indication of pending signals that might
1700 * be seen by the filesystem code called to write the core file.
1701 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001702 clear_thread_flag(TIF_SIGPENDING);
1703
Linus Torvalds1da177e2005-04-16 15:20:36 -07001704 /*
1705 * lock_kernel() because format_corename() is controlled by sysctl, which
1706 * uses lock_kernel()
1707 */
1708 lock_kernel();
Alan Coxc4bbafd2007-04-16 22:53:13 -07001709 ispipe = format_corename(corename, core_pattern, signr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001710 unlock_kernel();
Neil Horman7dc0b222007-10-16 23:26:34 -07001711 /*
1712 * Don't bother to check the RLIMIT_CORE value if core_pattern points
1713 * to a pipe. Since we're not writing directly to the filesystem
1714 * RLIMIT_CORE doesn't really apply, as no actual core file will be
1715 * created unless the pipe reader choses to write out the core file
1716 * at which point file size limits and permissions will be imposed
1717 * as it does with any other process
1718 */
Neil Horman74aadce2007-10-16 23:26:35 -07001719 if ((!ispipe) && (core_limit < binfmt->min_coredump))
Neil Horman7dc0b222007-10-16 23:26:34 -07001720 goto fail_unlock;
1721
Alan Coxc4bbafd2007-04-16 22:53:13 -07001722 if (ispipe) {
Neil Horman74aadce2007-10-16 23:26:35 -07001723 helper_argv = argv_split(GFP_KERNEL, corename+1, &helper_argc);
1724 /* Terminate the string before the first option */
1725 delimit = strchr(corename, ' ');
1726 if (delimit)
1727 *delimit = '\0';
Neil Horman32321132007-10-16 23:26:36 -07001728 delimit = strrchr(helper_argv[0], '/');
1729 if (delimit)
1730 delimit++;
1731 else
1732 delimit = helper_argv[0];
1733 if (!strcmp(delimit, current->comm)) {
1734 printk(KERN_NOTICE "Recursive core dump detected, "
1735 "aborting\n");
1736 goto fail_unlock;
1737 }
1738
1739 core_limit = RLIM_INFINITY;
1740
Andi Kleend025c9d2006-09-30 23:29:28 -07001741 /* SIGPIPE can happen, but it's just never processed */
Neil Horman32321132007-10-16 23:26:36 -07001742 if (call_usermodehelper_pipe(corename+1, helper_argv, NULL,
1743 &file)) {
Andi Kleend025c9d2006-09-30 23:29:28 -07001744 printk(KERN_INFO "Core dump to %s pipe failed\n",
1745 corename);
1746 goto fail_unlock;
1747 }
Andi Kleend025c9d2006-09-30 23:29:28 -07001748 } else
1749 file = filp_open(corename,
Alexey Dobriyan6d4df672006-12-06 20:40:39 -08001750 O_CREAT | 2 | O_NOFOLLOW | O_LARGEFILE | flag,
1751 0600);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001752 if (IS_ERR(file))
1753 goto fail_unlock;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001754 inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001755 if (inode->i_nlink > 1)
1756 goto close_fail; /* multiple links - don't dump */
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001757 if (!ispipe && d_unhashed(file->f_path.dentry))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001758 goto close_fail;
1759
Andi Kleend025c9d2006-09-30 23:29:28 -07001760 /* AK: actually i see no reason to not allow this for named pipes etc.,
1761 but keep the previous behaviour for now. */
1762 if (!ispipe && !S_ISREG(inode->i_mode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001763 goto close_fail;
Ingo Molnarc46f7392007-11-28 13:59:18 +01001764 /*
1765 * Dont allow local users get cute and trick others to coredump
1766 * into their pre-created files:
1767 */
1768 if (inode->i_uid != current->fsuid)
1769 goto close_fail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001770 if (!file->f_op)
1771 goto close_fail;
1772 if (!file->f_op->write)
1773 goto close_fail;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001774 if (!ispipe && do_truncate(file->f_path.dentry, 0, 0, file) != 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001775 goto close_fail;
1776
Neil Horman7dc0b222007-10-16 23:26:34 -07001777 retval = binfmt->core_dump(signr, regs, file, core_limit);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001778
1779 if (retval)
1780 current->signal->group_exit_code |= 0x80;
1781close_fail:
1782 filp_close(file, NULL);
1783fail_unlock:
Neil Horman74aadce2007-10-16 23:26:35 -07001784 if (helper_argv)
1785 argv_free(helper_argv);
1786
Alan Coxd6e71142005-06-23 00:09:43 -07001787 current->fsuid = fsuid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001788 complete_all(&mm->core_done);
1789fail:
1790 return retval;
1791}