blob: a3a8ce83940f1dfdfeb2430571f6dc19359e253f [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>
Al Viro9f3acc32008-04-24 07:44:08 -040027#include <linux/fdtable.h>
Hugh Dickinsba92a432008-07-25 01:45:43 -070028#include <linux/mm.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070029#include <linux/stat.h>
30#include <linux/fcntl.h>
31#include <linux/smp_lock.h>
Hugh Dickinsba92a432008-07-25 01:45:43 -070032#include <linux/swap.h>
Neil Horman74aadce2007-10-16 23:26:35 -070033#include <linux/string.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070034#include <linux/init.h>
Hugh Dickinsca5b1722008-07-28 15:46:18 -070035#include <linux/pagemap.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070036#include <linux/highmem.h>
37#include <linux/spinlock.h>
38#include <linux/key.h>
39#include <linux/personality.h>
40#include <linux/binfmts.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070041#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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070046#include <linux/mount.h>
47#include <linux/security.h>
Mimi Zohar6146f0d2009-02-04 09:06:57 -050048#include <linux/ima.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070049#include <linux/syscalls.h>
Jay Lan8f0ab512006-09-30 23:28:59 -070050#include <linux/tsacct_kern.h>
Matt Helsley9f460802005-11-07 00:59:16 -080051#include <linux/cn_proc.h>
Al Viro473ae302006-04-26 14:04:08 -040052#include <linux/audit.h>
Roland McGrath6341c392008-07-25 19:45:44 -070053#include <linux/tracehook.h>
Johannes Berg5f4123b2008-07-09 10:28:40 +020054#include <linux/kmod.h>
Eric Paris6110e3a2008-12-17 13:53:20 -050055#include <linux/fsnotify.h>
Al Viro5ad4e532009-03-29 19:50:06 -040056#include <linux/fs_struct.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070057
58#include <asm/uaccess.h>
59#include <asm/mmu_context.h>
Ollie Wildb6a2fea2007-07-19 01:48:16 -070060#include <asm/tlb.h>
David Howellsa6f76f22008-11-14 10:39:24 +110061#include "internal.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070062
Linus Torvalds1da177e2005-04-16 15:20:36 -070063int 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 */
Heiko Carstens1e7bfb22009-01-14 14:14:29 +0100104SYSCALL_DEFINE1(uselib, const char __user *, library)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105{
Al Viro964bd182008-07-26 03:33:14 -0400106 struct file *file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700107 struct nameidata nd;
Al Viro964bd182008-07-26 03:33:14 -0400108 char *tmp = getname(library);
109 int error = PTR_ERR(tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700110
Al Viro964bd182008-07-26 03:33:14 -0400111 if (!IS_ERR(tmp)) {
112 error = path_lookup_open(AT_FDCWD, tmp,
113 LOOKUP_FOLLOW, &nd,
114 FMODE_READ|FMODE_EXEC);
115 putname(tmp);
116 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117 if (error)
118 goto out;
119
120 error = -EINVAL;
Jan Blunck4ac91372008-02-14 19:34:32 -0800121 if (!S_ISREG(nd.path.dentry->d_inode->i_mode))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122 goto exit;
123
Al Viro30524472008-07-22 00:02:33 -0400124 error = -EACCES;
125 if (nd.path.mnt->mnt_flags & MNT_NOEXEC)
126 goto exit;
127
Christoph Hellwigcb23beb2008-10-24 09:59:29 +0200128 error = inode_permission(nd.path.dentry->d_inode,
129 MAY_READ | MAY_EXEC | MAY_OPEN);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130 if (error)
131 goto exit;
Mimi Zohar6146f0d2009-02-04 09:06:57 -0500132 error = ima_path_check(&nd.path, MAY_READ | MAY_EXEC | MAY_OPEN);
133 if (error)
134 goto exit;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700135
Andi Kleenabe8be32008-02-08 04:20:23 -0800136 file = nameidata_to_filp(&nd, O_RDONLY|O_LARGEFILE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137 error = PTR_ERR(file);
138 if (IS_ERR(file))
139 goto out;
140
Eric Paris6110e3a2008-12-17 13:53:20 -0500141 fsnotify_open(file->f_path.dentry);
142
Linus Torvalds1da177e2005-04-16 15:20:36 -0700143 error = -ENOEXEC;
144 if(file->f_op) {
145 struct linux_binfmt * fmt;
146
147 read_lock(&binfmt_lock);
Alexey Dobriyane4dc1b12007-10-16 23:26:03 -0700148 list_for_each_entry(fmt, &formats, lh) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149 if (!fmt->load_shlib)
150 continue;
151 if (!try_module_get(fmt->module))
152 continue;
153 read_unlock(&binfmt_lock);
154 error = fmt->load_shlib(file);
155 read_lock(&binfmt_lock);
156 put_binfmt(fmt);
157 if (error != -ENOEXEC)
158 break;
159 }
160 read_unlock(&binfmt_lock);
161 }
162 fput(file);
163out:
164 return error;
165exit:
Trond Myklebust834f2a42005-10-18 14:20:16 -0700166 release_open_intent(&nd);
Jan Blunck1d957f92008-02-14 19:34:35 -0800167 path_put(&nd.path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700168 goto out;
169}
170
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700171#ifdef CONFIG_MMU
172
173static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos,
174 int write)
175{
176 struct page *page;
177 int ret;
178
179#ifdef CONFIG_STACK_GROWSUP
180 if (write) {
181 ret = expand_stack_downwards(bprm->vma, pos);
182 if (ret < 0)
183 return NULL;
184 }
185#endif
186 ret = get_user_pages(current, bprm->mm, pos,
187 1, write, 1, &page, NULL);
188 if (ret <= 0)
189 return NULL;
190
191 if (write) {
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700192 unsigned long size = bprm->vma->vm_end - bprm->vma->vm_start;
Linus Torvaldsa64e7152008-03-03 10:12:14 -0800193 struct rlimit *rlim;
194
195 /*
196 * We've historically supported up to 32 pages (ARG_MAX)
197 * of argument strings even with small stacks
198 */
199 if (size <= ARG_MAX)
200 return page;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700201
202 /*
203 * Limit to 1/4-th the stack size for the argv+env strings.
204 * This ensures that:
205 * - the remaining binfmt code will not run out of stack space,
206 * - the program will have a reasonable amount of stack left
207 * to work from.
208 */
Linus Torvaldsa64e7152008-03-03 10:12:14 -0800209 rlim = current->signal->rlim;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700210 if (size > rlim[RLIMIT_STACK].rlim_cur / 4) {
211 put_page(page);
212 return NULL;
213 }
214 }
215
216 return page;
217}
218
219static void put_arg_page(struct page *page)
220{
221 put_page(page);
222}
223
224static void free_arg_page(struct linux_binprm *bprm, int i)
225{
226}
227
228static void free_arg_pages(struct linux_binprm *bprm)
229{
230}
231
232static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos,
233 struct page *page)
234{
235 flush_cache_page(bprm->vma, pos, page_to_pfn(page));
236}
237
238static int __bprm_mm_init(struct linux_binprm *bprm)
239{
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800240 int err;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700241 struct vm_area_struct *vma = NULL;
242 struct mm_struct *mm = bprm->mm;
243
244 bprm->vma = vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL);
245 if (!vma)
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800246 return -ENOMEM;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700247
248 down_write(&mm->mmap_sem);
249 vma->vm_mm = mm;
250
251 /*
252 * Place the stack at the largest stack address the architecture
253 * supports. Later, we'll move this to an appropriate place. We don't
254 * use STACK_TOP because that can depend on attributes which aren't
255 * configured yet.
256 */
257 vma->vm_end = STACK_TOP_MAX;
258 vma->vm_start = vma->vm_end - PAGE_SIZE;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700259 vma->vm_flags = VM_STACK_FLAGS;
Coly Li3ed75eb2007-10-18 23:39:15 -0700260 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700261 err = insert_vm_struct(mm, vma);
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800262 if (err)
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700263 goto err;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700264
265 mm->stack_vm = mm->total_vm = 1;
266 up_write(&mm->mmap_sem);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700267 bprm->p = vma->vm_end - sizeof(void *);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700268 return 0;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700269err:
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800270 up_write(&mm->mmap_sem);
271 bprm->vma = NULL;
272 kmem_cache_free(vm_area_cachep, vma);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700273 return err;
274}
275
276static bool valid_arg_len(struct linux_binprm *bprm, long len)
277{
278 return len <= MAX_ARG_STRLEN;
279}
280
281#else
282
283static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos,
284 int write)
285{
286 struct page *page;
287
288 page = bprm->page[pos / PAGE_SIZE];
289 if (!page && write) {
290 page = alloc_page(GFP_HIGHUSER|__GFP_ZERO);
291 if (!page)
292 return NULL;
293 bprm->page[pos / PAGE_SIZE] = page;
294 }
295
296 return page;
297}
298
299static void put_arg_page(struct page *page)
300{
301}
302
303static void free_arg_page(struct linux_binprm *bprm, int i)
304{
305 if (bprm->page[i]) {
306 __free_page(bprm->page[i]);
307 bprm->page[i] = NULL;
308 }
309}
310
311static void free_arg_pages(struct linux_binprm *bprm)
312{
313 int i;
314
315 for (i = 0; i < MAX_ARG_PAGES; i++)
316 free_arg_page(bprm, i);
317}
318
319static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos,
320 struct page *page)
321{
322}
323
324static int __bprm_mm_init(struct linux_binprm *bprm)
325{
326 bprm->p = PAGE_SIZE * MAX_ARG_PAGES - sizeof(void *);
327 return 0;
328}
329
330static bool valid_arg_len(struct linux_binprm *bprm, long len)
331{
332 return len <= bprm->p;
333}
334
335#endif /* CONFIG_MMU */
336
337/*
338 * Create a new mm_struct and populate it with a temporary stack
339 * vm_area_struct. We don't have enough context at this point to set the stack
340 * flags, permissions, and offset, so we use temporary values. We'll update
341 * them later in setup_arg_pages().
342 */
343int bprm_mm_init(struct linux_binprm *bprm)
344{
345 int err;
346 struct mm_struct *mm = NULL;
347
348 bprm->mm = mm = mm_alloc();
349 err = -ENOMEM;
350 if (!mm)
351 goto err;
352
353 err = init_new_context(current, mm);
354 if (err)
355 goto err;
356
357 err = __bprm_mm_init(bprm);
358 if (err)
359 goto err;
360
361 return 0;
362
363err:
364 if (mm) {
365 bprm->mm = NULL;
366 mmdrop(mm);
367 }
368
369 return err;
370}
371
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372/*
373 * count() counts the number of strings in array ARGV.
374 */
375static int count(char __user * __user * argv, int max)
376{
377 int i = 0;
378
379 if (argv != NULL) {
380 for (;;) {
381 char __user * p;
382
383 if (get_user(p, argv))
384 return -EFAULT;
385 if (!p)
386 break;
387 argv++;
Jason Baron362e6662008-10-15 22:01:52 -0700388 if (i++ >= max)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700389 return -E2BIG;
390 cond_resched();
391 }
392 }
393 return i;
394}
395
396/*
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700397 * 'copy_strings()' copies argument/environment strings from the old
398 * processes's memory to the new process's stack. The call to get_user_pages()
399 * ensures the destination page is created and not swapped out.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400 */
Adrian Bunk75c96f82005-05-05 16:16:09 -0700401static int copy_strings(int argc, char __user * __user * argv,
402 struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700403{
404 struct page *kmapped_page = NULL;
405 char *kaddr = NULL;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700406 unsigned long kpos = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407 int ret;
408
409 while (argc-- > 0) {
410 char __user *str;
411 int len;
412 unsigned long pos;
413
414 if (get_user(str, argv+argc) ||
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700415 !(len = strnlen_user(str, MAX_ARG_STRLEN))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416 ret = -EFAULT;
417 goto out;
418 }
419
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700420 if (!valid_arg_len(bprm, len)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421 ret = -E2BIG;
422 goto out;
423 }
424
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700425 /* We're going to work our way backwords. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700426 pos = bprm->p;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700427 str += len;
428 bprm->p -= len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700429
430 while (len > 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700431 int offset, bytes_to_copy;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432
433 offset = pos % PAGE_SIZE;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700434 if (offset == 0)
435 offset = PAGE_SIZE;
436
437 bytes_to_copy = offset;
438 if (bytes_to_copy > len)
439 bytes_to_copy = len;
440
441 offset -= bytes_to_copy;
442 pos -= bytes_to_copy;
443 str -= bytes_to_copy;
444 len -= bytes_to_copy;
445
446 if (!kmapped_page || kpos != (pos & PAGE_MASK)) {
447 struct page *page;
448
449 page = get_arg_page(bprm, pos, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700450 if (!page) {
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700451 ret = -E2BIG;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452 goto out;
453 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700455 if (kmapped_page) {
456 flush_kernel_dcache_page(kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457 kunmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700458 put_arg_page(kmapped_page);
459 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460 kmapped_page = page;
461 kaddr = kmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700462 kpos = pos & PAGE_MASK;
463 flush_arg_page(bprm, kpos, kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464 }
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700465 if (copy_from_user(kaddr+offset, str, bytes_to_copy)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466 ret = -EFAULT;
467 goto out;
468 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469 }
470 }
471 ret = 0;
472out:
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700473 if (kmapped_page) {
474 flush_kernel_dcache_page(kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700475 kunmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700476 put_arg_page(kmapped_page);
477 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700478 return ret;
479}
480
481/*
482 * Like copy_strings, but get argv and its values from kernel memory.
483 */
484int copy_strings_kernel(int argc,char ** argv, struct linux_binprm *bprm)
485{
486 int r;
487 mm_segment_t oldfs = get_fs();
488 set_fs(KERNEL_DS);
489 r = copy_strings(argc, (char __user * __user *)argv, bprm);
490 set_fs(oldfs);
491 return r;
492}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493EXPORT_SYMBOL(copy_strings_kernel);
494
495#ifdef CONFIG_MMU
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700496
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497/*
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700498 * During bprm_mm_init(), we create a temporary stack at STACK_TOP_MAX. Once
499 * the binfmt code determines where the new stack should reside, we shift it to
500 * its final location. The process proceeds as follows:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501 *
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700502 * 1) Use shift to calculate the new vma endpoints.
503 * 2) Extend vma to cover both the old and new ranges. This ensures the
504 * arguments passed to subsequent functions are consistent.
505 * 3) Move vma's page tables to the new range.
506 * 4) Free up any cleared pgd range.
507 * 5) Shrink the vma to cover only the new range.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508 */
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700509static int shift_arg_pages(struct vm_area_struct *vma, unsigned long shift)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510{
511 struct mm_struct *mm = vma->vm_mm;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700512 unsigned long old_start = vma->vm_start;
513 unsigned long old_end = vma->vm_end;
514 unsigned long length = old_end - old_start;
515 unsigned long new_start = old_start - shift;
516 unsigned long new_end = old_end - shift;
517 struct mmu_gather *tlb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700519 BUG_ON(new_start > new_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700520
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700521 /*
522 * ensure there are no vmas between where we want to go
523 * and where we are
524 */
525 if (vma != find_vma(mm, new_start))
526 return -EFAULT;
527
528 /*
529 * cover the whole range: [new_start, old_end)
530 */
531 vma_adjust(vma, new_start, old_end, vma->vm_pgoff, NULL);
532
533 /*
534 * move the page tables downwards, on failure we rely on
535 * process cleanup to remove whatever mess we made.
536 */
537 if (length != move_page_tables(vma, old_start,
538 vma, new_start, length))
539 return -ENOMEM;
540
541 lru_add_drain();
542 tlb = tlb_gather_mmu(mm, 0);
543 if (new_end > old_start) {
544 /*
545 * when the old and new regions overlap clear from new_end.
546 */
Jan Beulich42b77722008-07-23 21:27:10 -0700547 free_pgd_range(tlb, new_end, old_end, new_end,
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700548 vma->vm_next ? vma->vm_next->vm_start : 0);
549 } else {
550 /*
551 * otherwise, clean from old_start; this is done to not touch
552 * the address space in [new_end, old_start) some architectures
553 * have constraints on va-space that make this illegal (IA64) -
554 * for the others its just a little faster.
555 */
Jan Beulich42b77722008-07-23 21:27:10 -0700556 free_pgd_range(tlb, old_start, old_end, new_end,
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700557 vma->vm_next ? vma->vm_next->vm_start : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700558 }
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700559 tlb_finish_mmu(tlb, new_end, old_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700561 /*
562 * shrink the vma to just the new range.
563 */
564 vma_adjust(vma, new_start, new_end, vma->vm_pgoff, NULL);
565
566 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567}
568
569#define EXTRA_STACK_VM_PAGES 20 /* random */
570
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700571/*
572 * Finalizes the stack vm_area_struct. The flags and permissions are updated,
573 * the stack is optionally relocated, and some extra space is added.
574 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700575int setup_arg_pages(struct linux_binprm *bprm,
576 unsigned long stack_top,
577 int executable_stack)
578{
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700579 unsigned long ret;
580 unsigned long stack_shift;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700581 struct mm_struct *mm = current->mm;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700582 struct vm_area_struct *vma = bprm->vma;
583 struct vm_area_struct *prev = NULL;
584 unsigned long vm_flags;
585 unsigned long stack_base;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700586
587#ifdef CONFIG_STACK_GROWSUP
Linus Torvalds1da177e2005-04-16 15:20:36 -0700588 /* Limit stack size to 1GB */
589 stack_base = current->signal->rlim[RLIMIT_STACK].rlim_max;
590 if (stack_base > (1 << 30))
591 stack_base = 1 << 30;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700592
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700593 /* Make sure we didn't let the argument array grow too large. */
594 if (vma->vm_end - vma->vm_start > stack_base)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700595 return -ENOMEM;
596
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700597 stack_base = PAGE_ALIGN(stack_top - stack_base);
598
599 stack_shift = vma->vm_start - stack_base;
600 mm->arg_start = bprm->p - stack_shift;
601 bprm->p = vma->vm_end - stack_shift;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602#else
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700603 stack_top = arch_align_stack(stack_top);
604 stack_top = PAGE_ALIGN(stack_top);
605 stack_shift = vma->vm_end - stack_top;
606
607 bprm->p -= stack_shift;
608 mm->arg_start = bprm->p;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700609#endif
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700610
611 if (bprm->loader)
612 bprm->loader -= stack_shift;
613 bprm->exec -= stack_shift;
614
615 down_write(&mm->mmap_sem);
Hugh Dickins96a8e132008-07-10 21:19:20 +0100616 vm_flags = VM_STACK_FLAGS;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700617
618 /*
619 * Adjust stack execute permissions; explicitly enable for
620 * EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X and leave alone
621 * (arch default) otherwise.
622 */
623 if (unlikely(executable_stack == EXSTACK_ENABLE_X))
624 vm_flags |= VM_EXEC;
625 else if (executable_stack == EXSTACK_DISABLE_X)
626 vm_flags &= ~VM_EXEC;
627 vm_flags |= mm->def_flags;
628
629 ret = mprotect_fixup(vma, &prev, vma->vm_start, vma->vm_end,
630 vm_flags);
631 if (ret)
632 goto out_unlock;
633 BUG_ON(prev != vma);
634
635 /* Move stack pages down in memory. */
636 if (stack_shift) {
637 ret = shift_arg_pages(vma, stack_shift);
638 if (ret) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700639 up_write(&mm->mmap_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700640 return ret;
641 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700642 }
643
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700644#ifdef CONFIG_STACK_GROWSUP
645 stack_base = vma->vm_end + EXTRA_STACK_VM_PAGES * PAGE_SIZE;
646#else
647 stack_base = vma->vm_start - EXTRA_STACK_VM_PAGES * PAGE_SIZE;
648#endif
649 ret = expand_stack(vma, stack_base);
650 if (ret)
651 ret = -EFAULT;
652
653out_unlock:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654 up_write(&mm->mmap_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655 return 0;
656}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700657EXPORT_SYMBOL(setup_arg_pages);
658
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659#endif /* CONFIG_MMU */
660
661struct file *open_exec(const char *name)
662{
663 struct nameidata nd;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664 struct file *file;
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200665 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700666
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200667 err = path_lookup_open(AT_FDCWD, name, LOOKUP_FOLLOW, &nd,
668 FMODE_READ|FMODE_EXEC);
669 if (err)
670 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200672 err = -EACCES;
673 if (!S_ISREG(nd.path.dentry->d_inode->i_mode))
674 goto out_path_put;
675
Al Viro30524472008-07-22 00:02:33 -0400676 if (nd.path.mnt->mnt_flags & MNT_NOEXEC)
677 goto out_path_put;
678
Christoph Hellwigcb23beb2008-10-24 09:59:29 +0200679 err = inode_permission(nd.path.dentry->d_inode, MAY_EXEC | MAY_OPEN);
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200680 if (err)
681 goto out_path_put;
Mimi Zohar6146f0d2009-02-04 09:06:57 -0500682 err = ima_path_check(&nd.path, MAY_EXEC | MAY_OPEN);
683 if (err)
684 goto out_path_put;
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200685
686 file = nameidata_to_filp(&nd, O_RDONLY|O_LARGEFILE);
687 if (IS_ERR(file))
688 return file;
689
Eric Paris6110e3a2008-12-17 13:53:20 -0500690 fsnotify_open(file->f_path.dentry);
691
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200692 err = deny_write_access(file);
693 if (err) {
694 fput(file);
695 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700696 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700697
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200698 return file;
699
700 out_path_put:
701 release_open_intent(&nd);
702 path_put(&nd.path);
703 out:
704 return ERR_PTR(err);
705}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706EXPORT_SYMBOL(open_exec);
707
708int kernel_read(struct file *file, unsigned long offset,
709 char *addr, unsigned long count)
710{
711 mm_segment_t old_fs;
712 loff_t pos = offset;
713 int result;
714
715 old_fs = get_fs();
716 set_fs(get_ds());
717 /* The cast to a user pointer is valid due to the set_fs() */
718 result = vfs_read(file, (void __user *)addr, count, &pos);
719 set_fs(old_fs);
720 return result;
721}
722
723EXPORT_SYMBOL(kernel_read);
724
725static int exec_mmap(struct mm_struct *mm)
726{
727 struct task_struct *tsk;
728 struct mm_struct * old_mm, *active_mm;
729
730 /* Notify parent that we're no longer interested in the old VM */
731 tsk = current;
732 old_mm = current->mm;
733 mm_release(tsk, old_mm);
734
735 if (old_mm) {
736 /*
737 * Make sure that if there is a core dump in progress
738 * for the old mm, we get out and die instead of going
739 * through with the exec. We must hold mmap_sem around
Oleg Nesterov999d9fc2008-07-25 01:47:41 -0700740 * checking core_state and changing tsk->mm.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700741 */
742 down_read(&old_mm->mmap_sem);
Oleg Nesterov999d9fc2008-07-25 01:47:41 -0700743 if (unlikely(old_mm->core_state)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744 up_read(&old_mm->mmap_sem);
745 return -EINTR;
746 }
747 }
748 task_lock(tsk);
749 active_mm = tsk->active_mm;
750 tsk->mm = mm;
751 tsk->active_mm = mm;
752 activate_mm(active_mm, mm);
753 task_unlock(tsk);
754 arch_pick_mmap_layout(mm);
755 if (old_mm) {
756 up_read(&old_mm->mmap_sem);
Eric Sesterhenn7dddb122006-04-01 01:13:38 +0200757 BUG_ON(active_mm != old_mm);
Balbir Singh31a78f22008-09-28 23:09:31 +0100758 mm_update_next_owner(old_mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759 mmput(old_mm);
760 return 0;
761 }
762 mmdrop(active_mm);
763 return 0;
764}
765
766/*
767 * This function makes sure the current process has its own signal table,
768 * so that flush_signal_handlers can later reset the handlers without
769 * disturbing other processes. (Other processes might share the signal
770 * table via the CLONE_SIGHAND option to clone().)
771 */
Arjan van de Ven858119e2006-01-14 13:20:43 -0800772static int de_thread(struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773{
774 struct signal_struct *sig = tsk->signal;
Oleg Nesterovb2c903b2007-10-16 23:27:22 -0700775 struct sighand_struct *oldsighand = tsk->sighand;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776 spinlock_t *lock = &oldsighand->siglock;
777 int count;
778
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700779 if (thread_group_empty(tsk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700780 goto no_thread_group;
781
782 /*
783 * Kill all other threads in the thread group.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700784 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700785 spin_lock_irq(lock);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -0800786 if (signal_group_exit(sig)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787 /*
788 * Another group action in progress, just
789 * return so that the signal is processed.
790 */
791 spin_unlock_irq(lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792 return -EAGAIN;
793 }
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -0800794 sig->group_exit_task = tsk;
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700795 zap_other_threads(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796
Oleg Nesterovfea9d172008-02-08 04:19:19 -0800797 /* Account for the thread group leader hanging around: */
798 count = thread_group_leader(tsk) ? 1 : 2;
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700799 sig->notify_count = count;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700800 while (atomic_read(&sig->count) > count) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700801 __set_current_state(TASK_UNINTERRUPTIBLE);
802 spin_unlock_irq(lock);
803 schedule();
804 spin_lock_irq(lock);
805 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700806 spin_unlock_irq(lock);
807
808 /*
809 * At this point all other threads have exited, all we have to
810 * do is to wait for the thread group leader to become inactive,
811 * and to assume its PID:
812 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700813 if (!thread_group_leader(tsk)) {
Oleg Nesterov81879262008-12-01 14:18:16 -0800814 struct task_struct *leader = tsk->group_leader;
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700815
Oleg Nesterov2800d8d2008-04-30 00:53:12 -0700816 sig->notify_count = -1; /* for exit_notify() */
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700817 for (;;) {
818 write_lock_irq(&tasklist_lock);
819 if (likely(leader->exit_state))
820 break;
821 __set_current_state(TASK_UNINTERRUPTIBLE);
822 write_unlock_irq(&tasklist_lock);
823 schedule();
824 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825
Roland McGrathf5e90282006-04-10 22:54:16 -0700826 /*
827 * The only record we have of the real-time age of a
828 * process, regardless of execs it's done, is start_time.
829 * All the past CPU time is accumulated in signal_struct
830 * from sister threads now dead. But in this non-leader
831 * exec, nothing survives from the original leader thread,
832 * whose birth marks the true age of this process now.
833 * When we take on its identity by switching to its PID, we
834 * also take its birthdate (always earlier than our own).
835 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700836 tsk->start_time = leader->start_time;
Roland McGrathf5e90282006-04-10 22:54:16 -0700837
Pavel Emelyanovbac0abd2007-10-18 23:40:18 -0700838 BUG_ON(!same_thread_group(leader, tsk));
839 BUG_ON(has_group_leader_pid(tsk));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700840 /*
841 * An exec() starts a new thread group with the
842 * TGID of the previous thread group. Rehash the
843 * two threads with a switched PID, and release
844 * the former thread group leader:
845 */
Eric W. Biedermand73d6522006-03-28 16:11:03 -0800846
847 /* Become a process group leader with the old leader's pid.
Eric W. Biedermanc18258c2006-09-27 01:51:06 -0700848 * The old leader becomes a thread of the this thread group.
849 * Note: The old leader also uses this pid until release_task
Eric W. Biedermand73d6522006-03-28 16:11:03 -0800850 * is called. Odd but simple and correct.
851 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700852 detach_pid(tsk, PIDTYPE_PID);
853 tsk->pid = leader->pid;
Sukadev Bhattiprolu3743ca02007-10-18 23:39:51 -0700854 attach_pid(tsk, PIDTYPE_PID, task_pid(leader));
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700855 transfer_pid(leader, tsk, PIDTYPE_PGID);
856 transfer_pid(leader, tsk, PIDTYPE_SID);
857 list_replace_rcu(&leader->tasks, &tsk->tasks);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700859 tsk->group_leader = tsk;
860 leader->group_leader = tsk;
Eric W. Biedermande12a782006-04-10 17:16:49 -0600861
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700862 tsk->exit_signal = SIGCHLD;
Oleg Nesterov962b5642005-11-23 13:37:43 -0800863
864 BUG_ON(leader->exit_state != EXIT_ZOMBIE);
865 leader->exit_state = EXIT_DEAD;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866 write_unlock_irq(&tasklist_lock);
Oleg Nesterov81879262008-12-01 14:18:16 -0800867
868 release_task(leader);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -0800869 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700870
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700871 sig->group_exit_task = NULL;
872 sig->notify_count = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700873
874no_thread_group:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875 exit_itimers(sig);
Oleg Nesterovcbaffba2008-05-26 20:55:42 +0400876 flush_itimer_signals();
Oleg Nesterov329f7db2005-11-07 21:12:43 +0300877
Oleg Nesterovb2c903b2007-10-16 23:27:22 -0700878 if (atomic_read(&oldsighand->count) != 1) {
879 struct sighand_struct *newsighand;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700880 /*
Oleg Nesterovb2c903b2007-10-16 23:27:22 -0700881 * This ->sighand is shared with the CLONE_SIGHAND
882 * but not CLONE_THREAD task, switch to the new one.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883 */
Oleg Nesterovb2c903b2007-10-16 23:27:22 -0700884 newsighand = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
885 if (!newsighand)
886 return -ENOMEM;
887
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888 atomic_set(&newsighand->count, 1);
889 memcpy(newsighand->action, oldsighand->action,
890 sizeof(newsighand->action));
891
892 write_lock_irq(&tasklist_lock);
893 spin_lock(&oldsighand->siglock);
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700894 rcu_assign_pointer(tsk->sighand, newsighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895 spin_unlock(&oldsighand->siglock);
896 write_unlock_irq(&tasklist_lock);
897
Davide Libenzifba2afa2007-05-10 22:23:13 -0700898 __cleanup_sighand(oldsighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700899 }
900
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700901 BUG_ON(!thread_group_leader(tsk));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902 return 0;
903}
Oleg Nesterov0840a902007-10-16 23:27:22 -0700904
Linus Torvalds1da177e2005-04-16 15:20:36 -0700905/*
906 * These functions flushes out all traces of the currently running executable
907 * so that a new one can be started
908 */
Arjan van de Ven858119e2006-01-14 13:20:43 -0800909static void flush_old_files(struct files_struct * files)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700910{
911 long j = -1;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700912 struct fdtable *fdt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700913
914 spin_lock(&files->file_lock);
915 for (;;) {
916 unsigned long set, i;
917
918 j++;
919 i = j * __NFDBITS;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700920 fdt = files_fdtable(files);
Vadim Lobanovbbea9f62006-12-10 02:21:12 -0800921 if (i >= fdt->max_fds)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922 break;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700923 set = fdt->close_on_exec->fds_bits[j];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924 if (!set)
925 continue;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700926 fdt->close_on_exec->fds_bits[j] = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927 spin_unlock(&files->file_lock);
928 for ( ; set ; i++,set >>= 1) {
929 if (set & 1) {
930 sys_close(i);
931 }
932 }
933 spin_lock(&files->file_lock);
934
935 }
936 spin_unlock(&files->file_lock);
937}
938
Andrew Morton59714d62008-02-04 22:27:21 -0800939char *get_task_comm(char *buf, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700940{
941 /* buf must be at least sizeof(tsk->comm) in size */
942 task_lock(tsk);
943 strncpy(buf, tsk->comm, sizeof(tsk->comm));
944 task_unlock(tsk);
Andrew Morton59714d62008-02-04 22:27:21 -0800945 return buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700946}
947
948void set_task_comm(struct task_struct *tsk, char *buf)
949{
950 task_lock(tsk);
951 strlcpy(tsk->comm, buf, sizeof(tsk->comm));
952 task_unlock(tsk);
953}
954
955int flush_old_exec(struct linux_binprm * bprm)
956{
957 char * name;
958 int i, ch, retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700959 char tcomm[sizeof(current->comm)];
960
961 /*
962 * Make sure we have a private signal table and that
963 * we are unassociated from the previous thread group.
964 */
965 retval = de_thread(current);
966 if (retval)
967 goto out;
968
Matt Helsley925d1c42008-04-29 01:01:36 -0700969 set_mm_exe_file(bprm->mm, bprm->file);
970
Linus Torvalds1da177e2005-04-16 15:20:36 -0700971 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700972 * Release all of the old mmap stuff
973 */
974 retval = exec_mmap(bprm->mm);
975 if (retval)
Al Virofd8328b2008-04-22 05:11:59 -0400976 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977
978 bprm->mm = NULL; /* We're using it now */
979
980 /* This is the point of no return */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700981 current->sas_ss_sp = current->sas_ss_size = 0;
982
David Howellsda9592e2008-11-14 10:39:05 +1100983 if (current_euid() == current_uid() && current_egid() == current_gid())
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -0700984 set_dumpable(current->mm, 1);
Alan Coxd6e71142005-06-23 00:09:43 -0700985 else
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -0700986 set_dumpable(current->mm, suid_dumpable);
Alan Coxd6e71142005-06-23 00:09:43 -0700987
Linus Torvalds1da177e2005-04-16 15:20:36 -0700988 name = bprm->filename;
Paolo 'Blaisorblade' Giarrusso36772092005-05-05 16:16:12 -0700989
990 /* Copies the binary name from after last slash */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700991 for (i=0; (ch = *(name++)) != '\0';) {
992 if (ch == '/')
Paolo 'Blaisorblade' Giarrusso36772092005-05-05 16:16:12 -0700993 i = 0; /* overwrite what we wrote */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994 else
995 if (i < (sizeof(tcomm) - 1))
996 tcomm[i++] = ch;
997 }
998 tcomm[i] = '\0';
999 set_task_comm(current, tcomm);
1000
1001 current->flags &= ~PF_RANDOMIZE;
1002 flush_thread();
1003
Benjamin Herrenschmidt0551fbd2006-02-28 16:59:19 -08001004 /* Set the new mm task size. We have to do that late because it may
1005 * depend on TIF_32BIT which is only updated in flush_thread() on
1006 * some architectures like powerpc
1007 */
1008 current->mm->task_size = TASK_SIZE;
1009
David Howellsa6f76f22008-11-14 10:39:24 +11001010 /* install the new credentials */
1011 if (bprm->cred->uid != current_euid() ||
1012 bprm->cred->gid != current_egid()) {
Marcel Holtmannd2d56c52007-08-17 21:47:58 +02001013 current->pdeath_signal = 0;
1014 } else if (file_permission(bprm->file, MAY_READ) ||
David Howellsa6f76f22008-11-14 10:39:24 +11001015 bprm->interp_flags & BINPRM_FLAGS_ENFORCE_NONDUMP) {
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001016 set_dumpable(current->mm, suid_dumpable);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001017 }
1018
David Howellsa6f76f22008-11-14 10:39:24 +11001019 current->personality &= ~bprm->per_clear;
1020
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021 /* An exec changes our domain. We are no longer part of the thread
1022 group */
1023
1024 current->self_exec_id++;
1025
1026 flush_signal_handlers(current, 0);
1027 flush_old_files(current->files);
1028
1029 return 0;
1030
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031out:
1032 return retval;
1033}
1034
1035EXPORT_SYMBOL(flush_old_exec);
1036
David Howellsa6f76f22008-11-14 10:39:24 +11001037/*
1038 * install the new credentials for this executable
1039 */
1040void install_exec_creds(struct linux_binprm *bprm)
1041{
1042 security_bprm_committing_creds(bprm);
1043
1044 commit_creds(bprm->cred);
1045 bprm->cred = NULL;
1046
1047 /* cred_exec_mutex must be held at least to this point to prevent
1048 * ptrace_attach() from altering our determination of the task's
1049 * credentials; any time after this it may be unlocked */
1050
1051 security_bprm_committed_creds(bprm);
1052}
1053EXPORT_SYMBOL(install_exec_creds);
1054
1055/*
1056 * determine how safe it is to execute the proposed program
1057 * - the caller must hold current->cred_exec_mutex to protect against
1058 * PTRACE_ATTACH
1059 */
Al Viro498052b2009-03-30 07:20:30 -04001060int check_unsafe_exec(struct linux_binprm *bprm)
David Howellsa6f76f22008-11-14 10:39:24 +11001061{
David Howells0bf2f3a2009-02-06 11:45:46 +00001062 struct task_struct *p = current, *t;
Al Virof1191b52009-03-30 07:35:18 -04001063 unsigned n_fs;
Al Viro498052b2009-03-30 07:20:30 -04001064 int res = 0;
David Howellsa6f76f22008-11-14 10:39:24 +11001065
1066 bprm->unsafe = tracehook_unsafe_exec(p);
1067
David Howells0bf2f3a2009-02-06 11:45:46 +00001068 n_fs = 1;
Al Viro498052b2009-03-30 07:20:30 -04001069 write_lock(&p->fs->lock);
Oleg Nesterov437f7fd2009-04-24 01:02:45 +02001070 rcu_read_lock();
David Howells0bf2f3a2009-02-06 11:45:46 +00001071 for (t = next_thread(p); t != p; t = next_thread(t)) {
1072 if (t->fs == p->fs)
1073 n_fs++;
David Howells0bf2f3a2009-02-06 11:45:46 +00001074 }
Oleg Nesterov437f7fd2009-04-24 01:02:45 +02001075 rcu_read_unlock();
David Howells0bf2f3a2009-02-06 11:45:46 +00001076
Al Virof1191b52009-03-30 07:35:18 -04001077 if (p->fs->users > n_fs) {
David Howellsa6f76f22008-11-14 10:39:24 +11001078 bprm->unsafe |= LSM_UNSAFE_SHARE;
Al Viro498052b2009-03-30 07:20:30 -04001079 } else {
Oleg Nesterov8c652f92009-04-24 01:01:56 +02001080 res = -EAGAIN;
1081 if (!p->fs->in_exec) {
1082 p->fs->in_exec = 1;
1083 res = 1;
1084 }
Al Viro498052b2009-03-30 07:20:30 -04001085 }
Al Viro498052b2009-03-30 07:20:30 -04001086 write_unlock(&p->fs->lock);
1087
1088 return res;
David Howellsa6f76f22008-11-14 10:39:24 +11001089}
1090
Linus Torvalds1da177e2005-04-16 15:20:36 -07001091/*
1092 * Fill the binprm structure from the inode.
1093 * Check permissions, then read the first 128 (BINPRM_BUF_SIZE) bytes
David Howellsa6f76f22008-11-14 10:39:24 +11001094 *
1095 * This may be called multiple times for binary chains (scripts for example).
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096 */
1097int prepare_binprm(struct linux_binprm *bprm)
1098{
David Howellsa6f76f22008-11-14 10:39:24 +11001099 umode_t mode;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001100 struct inode * inode = bprm->file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001101 int retval;
1102
1103 mode = inode->i_mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001104 if (bprm->file->f_op == NULL)
1105 return -EACCES;
1106
David Howellsa6f76f22008-11-14 10:39:24 +11001107 /* clear any previous set[ug]id data from a previous binary */
1108 bprm->cred->euid = current_euid();
1109 bprm->cred->egid = current_egid();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001110
David Howellsa6f76f22008-11-14 10:39:24 +11001111 if (!(bprm->file->f_path.mnt->mnt_flags & MNT_NOSUID)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112 /* Set-uid? */
1113 if (mode & S_ISUID) {
David Howellsa6f76f22008-11-14 10:39:24 +11001114 bprm->per_clear |= PER_CLEAR_ON_SETID;
1115 bprm->cred->euid = inode->i_uid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001116 }
1117
1118 /* Set-gid? */
1119 /*
1120 * If setgid is set but no group execute bit then this
1121 * is a candidate for mandatory locking, not a setgid
1122 * executable.
1123 */
1124 if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) {
David Howellsa6f76f22008-11-14 10:39:24 +11001125 bprm->per_clear |= PER_CLEAR_ON_SETID;
1126 bprm->cred->egid = inode->i_gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001127 }
1128 }
1129
1130 /* fill in binprm security blob */
David Howellsa6f76f22008-11-14 10:39:24 +11001131 retval = security_bprm_set_creds(bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001132 if (retval)
1133 return retval;
David Howellsa6f76f22008-11-14 10:39:24 +11001134 bprm->cred_prepared = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001135
David Howellsa6f76f22008-11-14 10:39:24 +11001136 memset(bprm->buf, 0, BINPRM_BUF_SIZE);
1137 return kernel_read(bprm->file, 0, bprm->buf, BINPRM_BUF_SIZE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001138}
1139
1140EXPORT_SYMBOL(prepare_binprm);
1141
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001142/*
1143 * Arguments are '\0' separated strings found at the location bprm->p
1144 * points to; chop off the first by relocating brpm->p to right after
1145 * the first '\0' encountered.
1146 */
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001147int remove_arg_zero(struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001148{
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001149 int ret = 0;
1150 unsigned long offset;
1151 char *kaddr;
1152 struct page *page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001153
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001154 if (!bprm->argc)
1155 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001156
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001157 do {
1158 offset = bprm->p & ~PAGE_MASK;
1159 page = get_arg_page(bprm, bprm->p, 0);
1160 if (!page) {
1161 ret = -EFAULT;
1162 goto out;
1163 }
1164 kaddr = kmap_atomic(page, KM_USER0);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001165
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001166 for (; offset < PAGE_SIZE && kaddr[offset];
1167 offset++, bprm->p++)
1168 ;
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001169
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001170 kunmap_atomic(kaddr, KM_USER0);
1171 put_arg_page(page);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001172
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001173 if (offset == PAGE_SIZE)
1174 free_arg_page(bprm, (bprm->p >> PAGE_SHIFT) - 1);
1175 } while (offset == PAGE_SIZE);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001176
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001177 bprm->p++;
1178 bprm->argc--;
1179 ret = 0;
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001180
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001181out:
1182 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001184EXPORT_SYMBOL(remove_arg_zero);
1185
1186/*
1187 * cycle the list of binary formats handler, until one recognizes the image
1188 */
1189int search_binary_handler(struct linux_binprm *bprm,struct pt_regs *regs)
1190{
Roland McGrath85f33462008-12-09 19:36:38 -08001191 unsigned int depth = bprm->recursion_depth;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001192 int try,retval;
1193 struct linux_binfmt *fmt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001194
Linus Torvalds1da177e2005-04-16 15:20:36 -07001195 retval = security_bprm_check(bprm);
1196 if (retval)
1197 return retval;
Mimi Zohar6146f0d2009-02-04 09:06:57 -05001198 retval = ima_bprm_check(bprm);
1199 if (retval)
1200 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001201
1202 /* kernel module loader fixup */
1203 /* so we don't try to load run modprobe in kernel space. */
1204 set_fs(USER_DS);
Al Viro473ae302006-04-26 14:04:08 -04001205
1206 retval = audit_bprm(bprm);
1207 if (retval)
1208 return retval;
1209
Linus Torvalds1da177e2005-04-16 15:20:36 -07001210 retval = -ENOENT;
1211 for (try=0; try<2; try++) {
1212 read_lock(&binfmt_lock);
Alexey Dobriyane4dc1b12007-10-16 23:26:03 -07001213 list_for_each_entry(fmt, &formats, lh) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001214 int (*fn)(struct linux_binprm *, struct pt_regs *) = fmt->load_binary;
1215 if (!fn)
1216 continue;
1217 if (!try_module_get(fmt->module))
1218 continue;
1219 read_unlock(&binfmt_lock);
1220 retval = fn(bprm, regs);
Roland McGrath85f33462008-12-09 19:36:38 -08001221 /*
1222 * Restore the depth counter to its starting value
1223 * in this call, so we don't have to rely on every
1224 * load_binary function to restore it on return.
1225 */
1226 bprm->recursion_depth = depth;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001227 if (retval >= 0) {
Roland McGrath85f33462008-12-09 19:36:38 -08001228 if (depth == 0)
1229 tracehook_report_exec(fmt, bprm, regs);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001230 put_binfmt(fmt);
1231 allow_write_access(bprm->file);
1232 if (bprm->file)
1233 fput(bprm->file);
1234 bprm->file = NULL;
1235 current->did_exec = 1;
Matt Helsley9f460802005-11-07 00:59:16 -08001236 proc_exec_connector(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001237 return retval;
1238 }
1239 read_lock(&binfmt_lock);
1240 put_binfmt(fmt);
1241 if (retval != -ENOEXEC || bprm->mm == NULL)
1242 break;
1243 if (!bprm->file) {
1244 read_unlock(&binfmt_lock);
1245 return retval;
1246 }
1247 }
1248 read_unlock(&binfmt_lock);
1249 if (retval != -ENOEXEC || bprm->mm == NULL) {
1250 break;
Johannes Berg5f4123b2008-07-09 10:28:40 +02001251#ifdef CONFIG_MODULES
1252 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001253#define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e))
1254 if (printable(bprm->buf[0]) &&
1255 printable(bprm->buf[1]) &&
1256 printable(bprm->buf[2]) &&
1257 printable(bprm->buf[3]))
1258 break; /* -ENOEXEC */
1259 request_module("binfmt-%04x", *(unsigned short *)(&bprm->buf[2]));
1260#endif
1261 }
1262 }
1263 return retval;
1264}
1265
1266EXPORT_SYMBOL(search_binary_handler);
1267
Al Viro08a6fac2008-05-10 16:38:25 -04001268void free_bprm(struct linux_binprm *bprm)
1269{
1270 free_arg_pages(bprm);
David Howellsa6f76f22008-11-14 10:39:24 +11001271 if (bprm->cred)
1272 abort_creds(bprm->cred);
Al Viro08a6fac2008-05-10 16:38:25 -04001273 kfree(bprm);
1274}
1275
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276/*
1277 * sys_execve() executes a new program.
1278 */
1279int do_execve(char * filename,
1280 char __user *__user *argv,
1281 char __user *__user *envp,
1282 struct pt_regs * regs)
1283{
1284 struct linux_binprm *bprm;
1285 struct file *file;
Al Viro3b125382008-04-22 05:31:30 -04001286 struct files_struct *displaced;
Oleg Nesterov8c652f92009-04-24 01:01:56 +02001287 bool clear_in_exec;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001288 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001289
Al Viro3b125382008-04-22 05:31:30 -04001290 retval = unshare_files(&displaced);
Al Virofd8328b2008-04-22 05:11:59 -04001291 if (retval)
1292 goto out_ret;
1293
Linus Torvalds1da177e2005-04-16 15:20:36 -07001294 retval = -ENOMEM;
Oliver Neukum11b0b5a2006-03-25 03:08:13 -08001295 bprm = kzalloc(sizeof(*bprm), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001296 if (!bprm)
Al Virofd8328b2008-04-22 05:11:59 -04001297 goto out_files;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001298
David Howellsa6f76f22008-11-14 10:39:24 +11001299 retval = mutex_lock_interruptible(&current->cred_exec_mutex);
1300 if (retval < 0)
1301 goto out_free;
Kentaro Takedaf9ce1f12009-02-05 17:18:11 +09001302 current->in_execve = 1;
David Howellsa6f76f22008-11-14 10:39:24 +11001303
1304 retval = -ENOMEM;
1305 bprm->cred = prepare_exec_creds();
1306 if (!bprm->cred)
1307 goto out_unlock;
Al Viro498052b2009-03-30 07:20:30 -04001308
1309 retval = check_unsafe_exec(bprm);
Oleg Nesterov8c652f92009-04-24 01:01:56 +02001310 if (retval < 0)
Al Viro498052b2009-03-30 07:20:30 -04001311 goto out_unlock;
Oleg Nesterov8c652f92009-04-24 01:01:56 +02001312 clear_in_exec = retval;
David Howellsa6f76f22008-11-14 10:39:24 +11001313
Linus Torvalds1da177e2005-04-16 15:20:36 -07001314 file = open_exec(filename);
1315 retval = PTR_ERR(file);
1316 if (IS_ERR(file))
Al Viro498052b2009-03-30 07:20:30 -04001317 goto out_unmark;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001318
1319 sched_exec();
1320
Linus Torvalds1da177e2005-04-16 15:20:36 -07001321 bprm->file = file;
1322 bprm->filename = filename;
1323 bprm->interp = filename;
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001324
1325 retval = bprm_mm_init(bprm);
1326 if (retval)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001327 goto out_file;
1328
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001329 bprm->argc = count(argv, MAX_ARG_STRINGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001330 if ((retval = bprm->argc) < 0)
David Howellsa6f76f22008-11-14 10:39:24 +11001331 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001332
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001333 bprm->envc = count(envp, MAX_ARG_STRINGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001334 if ((retval = bprm->envc) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001335 goto out;
1336
1337 retval = prepare_binprm(bprm);
1338 if (retval < 0)
1339 goto out;
1340
1341 retval = copy_strings_kernel(1, &bprm->filename, bprm);
1342 if (retval < 0)
1343 goto out;
1344
1345 bprm->exec = bprm->p;
1346 retval = copy_strings(bprm->envc, envp, bprm);
1347 if (retval < 0)
1348 goto out;
1349
1350 retval = copy_strings(bprm->argc, argv, bprm);
1351 if (retval < 0)
1352 goto out;
1353
Oleg Nesterov7b34e422008-07-25 01:47:37 -07001354 current->flags &= ~PF_KTHREAD;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001355 retval = search_binary_handler(bprm,regs);
David Howellsa6f76f22008-11-14 10:39:24 +11001356 if (retval < 0)
1357 goto out;
1358
1359 /* execve succeeded */
Al Viro498052b2009-03-30 07:20:30 -04001360 current->fs->in_exec = 0;
Kentaro Takedaf9ce1f12009-02-05 17:18:11 +09001361 current->in_execve = 0;
David Howellsa6f76f22008-11-14 10:39:24 +11001362 mutex_unlock(&current->cred_exec_mutex);
1363 acct_update_integrals(current);
1364 free_bprm(bprm);
1365 if (displaced)
1366 put_files_struct(displaced);
1367 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001368
1369out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001370 if (bprm->mm)
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001371 mmput (bprm->mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001372
1373out_file:
1374 if (bprm->file) {
1375 allow_write_access(bprm->file);
1376 fput(bprm->file);
1377 }
David Howellsa6f76f22008-11-14 10:39:24 +11001378
Al Viro498052b2009-03-30 07:20:30 -04001379out_unmark:
Oleg Nesterov8c652f92009-04-24 01:01:56 +02001380 if (clear_in_exec)
1381 current->fs->in_exec = 0;
Al Viro498052b2009-03-30 07:20:30 -04001382
David Howellsa6f76f22008-11-14 10:39:24 +11001383out_unlock:
Kentaro Takedaf9ce1f12009-02-05 17:18:11 +09001384 current->in_execve = 0;
David Howellsa6f76f22008-11-14 10:39:24 +11001385 mutex_unlock(&current->cred_exec_mutex);
1386
1387out_free:
Al Viro08a6fac2008-05-10 16:38:25 -04001388 free_bprm(bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001389
Al Virofd8328b2008-04-22 05:11:59 -04001390out_files:
Al Viro3b125382008-04-22 05:31:30 -04001391 if (displaced)
1392 reset_files_struct(displaced);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001393out_ret:
1394 return retval;
1395}
1396
1397int set_binfmt(struct linux_binfmt *new)
1398{
1399 struct linux_binfmt *old = current->binfmt;
1400
1401 if (new) {
1402 if (!try_module_get(new->module))
1403 return -1;
1404 }
1405 current->binfmt = new;
1406 if (old)
1407 module_put(old->module);
1408 return 0;
1409}
1410
1411EXPORT_SYMBOL(set_binfmt);
1412
Linus Torvalds1da177e2005-04-16 15:20:36 -07001413/* format_corename will inspect the pattern parameter, and output a
1414 * name into corename, which must have space for at least
1415 * CORENAME_MAX_SIZE bytes plus one byte for the zero terminator.
1416 */
Oleg Nesterov64093242008-10-18 20:28:22 -07001417static int format_corename(char *corename, long signr)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001418{
David Howells86a264a2008-11-14 10:39:18 +11001419 const struct cred *cred = current_cred();
Oleg Nesterov565b9b12008-07-25 01:47:47 -07001420 const char *pat_ptr = core_pattern;
1421 int ispipe = (*pat_ptr == '|');
Linus Torvalds1da177e2005-04-16 15:20:36 -07001422 char *out_ptr = corename;
1423 char *const out_end = corename + CORENAME_MAX_SIZE;
1424 int rc;
1425 int pid_in_pattern = 0;
1426
1427 /* Repeat as long as we have more pattern to process and more output
1428 space */
1429 while (*pat_ptr) {
1430 if (*pat_ptr != '%') {
1431 if (out_ptr == out_end)
1432 goto out;
1433 *out_ptr++ = *pat_ptr++;
1434 } else {
1435 switch (*++pat_ptr) {
1436 case 0:
1437 goto out;
1438 /* Double percent, output one percent */
1439 case '%':
1440 if (out_ptr == out_end)
1441 goto out;
1442 *out_ptr++ = '%';
1443 break;
1444 /* pid */
1445 case 'p':
1446 pid_in_pattern = 1;
1447 rc = snprintf(out_ptr, out_end - out_ptr,
Pavel Emelyanovb4888932007-10-18 23:40:14 -07001448 "%d", task_tgid_vnr(current));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001449 if (rc > out_end - out_ptr)
1450 goto out;
1451 out_ptr += rc;
1452 break;
1453 /* uid */
1454 case 'u':
1455 rc = snprintf(out_ptr, out_end - out_ptr,
David Howells86a264a2008-11-14 10:39:18 +11001456 "%d", cred->uid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001457 if (rc > out_end - out_ptr)
1458 goto out;
1459 out_ptr += rc;
1460 break;
1461 /* gid */
1462 case 'g':
1463 rc = snprintf(out_ptr, out_end - out_ptr,
David Howells86a264a2008-11-14 10:39:18 +11001464 "%d", cred->gid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001465 if (rc > out_end - out_ptr)
1466 goto out;
1467 out_ptr += rc;
1468 break;
1469 /* signal that caused the coredump */
1470 case 's':
1471 rc = snprintf(out_ptr, out_end - out_ptr,
1472 "%ld", signr);
1473 if (rc > out_end - out_ptr)
1474 goto out;
1475 out_ptr += rc;
1476 break;
1477 /* UNIX time of coredump */
1478 case 't': {
1479 struct timeval tv;
1480 do_gettimeofday(&tv);
1481 rc = snprintf(out_ptr, out_end - out_ptr,
1482 "%lu", tv.tv_sec);
1483 if (rc > out_end - out_ptr)
1484 goto out;
1485 out_ptr += rc;
1486 break;
1487 }
1488 /* hostname */
1489 case 'h':
1490 down_read(&uts_sem);
1491 rc = snprintf(out_ptr, out_end - out_ptr,
Serge E. Hallyne9ff3992006-10-02 02:18:11 -07001492 "%s", utsname()->nodename);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001493 up_read(&uts_sem);
1494 if (rc > out_end - out_ptr)
1495 goto out;
1496 out_ptr += rc;
1497 break;
1498 /* executable */
1499 case 'e':
1500 rc = snprintf(out_ptr, out_end - out_ptr,
1501 "%s", current->comm);
1502 if (rc > out_end - out_ptr)
1503 goto out;
1504 out_ptr += rc;
1505 break;
Neil Horman74aadce2007-10-16 23:26:35 -07001506 /* core limit size */
1507 case 'c':
1508 rc = snprintf(out_ptr, out_end - out_ptr,
1509 "%lu", current->signal->rlim[RLIMIT_CORE].rlim_cur);
1510 if (rc > out_end - out_ptr)
1511 goto out;
1512 out_ptr += rc;
1513 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001514 default:
1515 break;
1516 }
1517 ++pat_ptr;
1518 }
1519 }
1520 /* Backward compatibility with core_uses_pid:
1521 *
1522 * If core_pattern does not include a %p (as is the default)
1523 * and core_uses_pid is set, then .%pid will be appended to
Alan Coxc4bbafd2007-04-16 22:53:13 -07001524 * the filename. Do not do this for piped commands. */
Oleg Nesterov64093242008-10-18 20:28:22 -07001525 if (!ispipe && !pid_in_pattern && core_uses_pid) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001526 rc = snprintf(out_ptr, out_end - out_ptr,
Pavel Emelyanovb4888932007-10-18 23:40:14 -07001527 ".%d", task_tgid_vnr(current));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001528 if (rc > out_end - out_ptr)
1529 goto out;
1530 out_ptr += rc;
1531 }
Alan Coxc4bbafd2007-04-16 22:53:13 -07001532out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001533 *out_ptr = 0;
Alan Coxc4bbafd2007-04-16 22:53:13 -07001534 return ispipe;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001535}
1536
Oleg Nesterov8cd9c242008-07-25 01:47:42 -07001537static int zap_process(struct task_struct *start)
Oleg Nesterovaceecc042006-06-26 00:26:05 -07001538{
1539 struct task_struct *t;
Oleg Nesterov8cd9c242008-07-25 01:47:42 -07001540 int nr = 0;
Oleg Nesterov281de332006-06-26 00:26:06 -07001541
Oleg Nesterovd5f70c02006-06-26 00:26:07 -07001542 start->signal->flags = SIGNAL_GROUP_EXIT;
1543 start->signal->group_stop_count = 0;
Oleg Nesterovaceecc042006-06-26 00:26:05 -07001544
1545 t = start;
1546 do {
1547 if (t != current && t->mm) {
Oleg Nesterov281de332006-06-26 00:26:06 -07001548 sigaddset(&t->pending.signal, SIGKILL);
1549 signal_wake_up(t, 1);
Oleg Nesterov8cd9c242008-07-25 01:47:42 -07001550 nr++;
Oleg Nesterovaceecc042006-06-26 00:26:05 -07001551 }
Oleg Nesterove4901f92008-07-25 01:47:31 -07001552 } while_each_thread(start, t);
Oleg Nesterov8cd9c242008-07-25 01:47:42 -07001553
1554 return nr;
Oleg Nesterovaceecc042006-06-26 00:26:05 -07001555}
1556
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001557static inline int zap_threads(struct task_struct *tsk, struct mm_struct *mm,
Oleg Nesterov8cd9c242008-07-25 01:47:42 -07001558 struct core_state *core_state, int exit_code)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001559{
1560 struct task_struct *g, *p;
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001561 unsigned long flags;
Oleg Nesterov8cd9c242008-07-25 01:47:42 -07001562 int nr = -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001563
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001564 spin_lock_irq(&tsk->sighand->siglock);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -08001565 if (!signal_group_exit(tsk->signal)) {
Oleg Nesterov8cd9c242008-07-25 01:47:42 -07001566 mm->core_state = core_state;
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001567 tsk->signal->group_exit_code = exit_code;
Oleg Nesterov8cd9c242008-07-25 01:47:42 -07001568 nr = zap_process(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001569 }
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001570 spin_unlock_irq(&tsk->sighand->siglock);
Oleg Nesterov8cd9c242008-07-25 01:47:42 -07001571 if (unlikely(nr < 0))
1572 return nr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001573
Oleg Nesterov8cd9c242008-07-25 01:47:42 -07001574 if (atomic_read(&mm->mm_users) == nr + 1)
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001575 goto done;
Oleg Nesterove4901f92008-07-25 01:47:31 -07001576 /*
1577 * We should find and kill all tasks which use this mm, and we should
Oleg Nesterov999d9fc2008-07-25 01:47:41 -07001578 * count them correctly into ->nr_threads. We don't take tasklist
Oleg Nesterove4901f92008-07-25 01:47:31 -07001579 * lock, but this is safe wrt:
1580 *
1581 * fork:
1582 * None of sub-threads can fork after zap_process(leader). All
1583 * processes which were created before this point should be
1584 * visible to zap_threads() because copy_process() adds the new
1585 * process to the tail of init_task.tasks list, and lock/unlock
1586 * of ->siglock provides a memory barrier.
1587 *
1588 * do_exit:
1589 * The caller holds mm->mmap_sem. This means that the task which
1590 * uses this mm can't pass exit_mm(), so it can't exit or clear
1591 * its ->mm.
1592 *
1593 * de_thread:
1594 * It does list_replace_rcu(&leader->tasks, &current->tasks),
1595 * we must see either old or new leader, this does not matter.
1596 * However, it can change p->sighand, so lock_task_sighand(p)
1597 * must be used. Since p->mm != NULL and we hold ->mmap_sem
1598 * it can't fail.
1599 *
1600 * Note also that "g" can be the old leader with ->mm == NULL
1601 * and already unhashed and thus removed from ->thread_group.
1602 * This is OK, __unhash_process()->list_del_rcu() does not
1603 * clear the ->next pointer, we will find the new leader via
1604 * next_thread().
1605 */
Oleg Nesterov7b1c6152006-06-26 00:26:08 -07001606 rcu_read_lock();
Oleg Nesterovaceecc042006-06-26 00:26:05 -07001607 for_each_process(g) {
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001608 if (g == tsk->group_leader)
1609 continue;
Oleg Nesterov15b9f362008-07-25 01:47:39 -07001610 if (g->flags & PF_KTHREAD)
1611 continue;
Oleg Nesterovaceecc042006-06-26 00:26:05 -07001612 p = g;
1613 do {
1614 if (p->mm) {
Oleg Nesterov15b9f362008-07-25 01:47:39 -07001615 if (unlikely(p->mm == mm)) {
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001616 lock_task_sighand(p, &flags);
Oleg Nesterov8cd9c242008-07-25 01:47:42 -07001617 nr += zap_process(p);
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001618 unlock_task_sighand(p, &flags);
1619 }
Oleg Nesterovaceecc042006-06-26 00:26:05 -07001620 break;
1621 }
Oleg Nesterove4901f92008-07-25 01:47:31 -07001622 } while_each_thread(g, p);
Oleg Nesterovaceecc042006-06-26 00:26:05 -07001623 }
Oleg Nesterov7b1c6152006-06-26 00:26:08 -07001624 rcu_read_unlock();
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001625done:
Oleg Nesterovc5f1cc82008-07-25 01:47:42 -07001626 atomic_set(&core_state->nr_threads, nr);
Oleg Nesterov8cd9c242008-07-25 01:47:42 -07001627 return nr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001628}
1629
Oleg Nesterov9d5b3272008-07-25 01:47:43 -07001630static int coredump_wait(int exit_code, struct core_state *core_state)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001631{
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001632 struct task_struct *tsk = current;
1633 struct mm_struct *mm = tsk->mm;
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001634 struct completion *vfork_done;
Oleg Nesterov2384f552005-10-30 15:02:47 -08001635 int core_waiters;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001636
Oleg Nesterov9d5b3272008-07-25 01:47:43 -07001637 init_completion(&core_state->startup);
Oleg Nesterovb564daf2008-07-25 01:47:44 -07001638 core_state->dumper.task = tsk;
1639 core_state->dumper.next = NULL;
Oleg Nesterov9d5b3272008-07-25 01:47:43 -07001640 core_waiters = zap_threads(tsk, mm, core_state, exit_code);
Oleg Nesterov2384f552005-10-30 15:02:47 -08001641 up_write(&mm->mmap_sem);
1642
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001643 if (unlikely(core_waiters < 0))
1644 goto fail;
1645
1646 /*
1647 * Make sure nobody is waiting for us to release the VM,
1648 * otherwise we can deadlock when we wait on each other
1649 */
1650 vfork_done = tsk->vfork_done;
1651 if (vfork_done) {
1652 tsk->vfork_done = NULL;
1653 complete(vfork_done);
1654 }
1655
Oleg Nesterov2384f552005-10-30 15:02:47 -08001656 if (core_waiters)
Oleg Nesterov9d5b3272008-07-25 01:47:43 -07001657 wait_for_completion(&core_state->startup);
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001658fail:
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001659 return core_waiters;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001660}
1661
Oleg Nesterova94e2d42008-07-25 01:47:46 -07001662static void coredump_finish(struct mm_struct *mm)
1663{
1664 struct core_thread *curr, *next;
1665 struct task_struct *task;
1666
1667 next = mm->core_state->dumper.next;
1668 while ((curr = next) != NULL) {
1669 next = curr->next;
1670 task = curr->task;
1671 /*
1672 * see exit_mm(), curr->task must not see
1673 * ->task == NULL before we read ->next.
1674 */
1675 smp_mb();
1676 curr->task = NULL;
1677 wake_up_process(task);
1678 }
1679
1680 mm->core_state = NULL;
1681}
1682
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001683/*
1684 * set_dumpable converts traditional three-value dumpable to two flags and
1685 * stores them into mm->flags. It modifies lower two bits of mm->flags, but
1686 * these bits are not changed atomically. So get_dumpable can observe the
1687 * intermediate state. To avoid doing unexpected behavior, get get_dumpable
1688 * return either old dumpable or new one by paying attention to the order of
1689 * modifying the bits.
1690 *
1691 * dumpable | mm->flags (binary)
1692 * old new | initial interim final
1693 * ---------+-----------------------
1694 * 0 1 | 00 01 01
1695 * 0 2 | 00 10(*) 11
1696 * 1 0 | 01 00 00
1697 * 1 2 | 01 11 11
1698 * 2 0 | 11 10(*) 00
1699 * 2 1 | 11 11 01
1700 *
1701 * (*) get_dumpable regards interim value of 10 as 11.
1702 */
1703void set_dumpable(struct mm_struct *mm, int value)
1704{
1705 switch (value) {
1706 case 0:
1707 clear_bit(MMF_DUMPABLE, &mm->flags);
1708 smp_wmb();
1709 clear_bit(MMF_DUMP_SECURELY, &mm->flags);
1710 break;
1711 case 1:
1712 set_bit(MMF_DUMPABLE, &mm->flags);
1713 smp_wmb();
1714 clear_bit(MMF_DUMP_SECURELY, &mm->flags);
1715 break;
1716 case 2:
1717 set_bit(MMF_DUMP_SECURELY, &mm->flags);
1718 smp_wmb();
1719 set_bit(MMF_DUMPABLE, &mm->flags);
1720 break;
1721 }
1722}
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001723
1724int get_dumpable(struct mm_struct *mm)
1725{
1726 int ret;
1727
1728 ret = mm->flags & 0x3;
1729 return (ret >= 2) ? 2 : ret;
1730}
1731
WANG Cong8cd3ac32009-01-06 14:42:48 -08001732void do_coredump(long signr, int exit_code, struct pt_regs *regs)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001733{
Oleg Nesterov9d5b3272008-07-25 01:47:43 -07001734 struct core_state core_state;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001735 char corename[CORENAME_MAX_SIZE + 1];
1736 struct mm_struct *mm = current->mm;
1737 struct linux_binfmt * binfmt;
1738 struct inode * inode;
1739 struct file * file;
David Howellsd84f4f92008-11-14 10:39:23 +11001740 const struct cred *old_cred;
1741 struct cred *cred;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001742 int retval = 0;
Alan Coxd6e71142005-06-23 00:09:43 -07001743 int flag = 0;
Andi Kleend025c9d2006-09-30 23:29:28 -07001744 int ispipe = 0;
Neil Horman7dc0b222007-10-16 23:26:34 -07001745 unsigned long core_limit = current->signal->rlim[RLIMIT_CORE].rlim_cur;
Neil Horman74aadce2007-10-16 23:26:35 -07001746 char **helper_argv = NULL;
1747 int helper_argc = 0;
1748 char *delimit;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001749
Steve Grubb0a4ff8c2007-04-19 10:28:21 -04001750 audit_core_dumps(signr);
1751
Linus Torvalds1da177e2005-04-16 15:20:36 -07001752 binfmt = current->binfmt;
1753 if (!binfmt || !binfmt->core_dump)
1754 goto fail;
David Howellsd84f4f92008-11-14 10:39:23 +11001755
1756 cred = prepare_creds();
1757 if (!cred) {
1758 retval = -ENOMEM;
1759 goto fail;
1760 }
1761
Linus Torvalds1da177e2005-04-16 15:20:36 -07001762 down_write(&mm->mmap_sem);
Roland McGrath00ec99d2007-11-11 19:13:43 -08001763 /*
1764 * If another thread got here first, or we are not dumpable, bail out.
1765 */
Oleg Nesterov999d9fc2008-07-25 01:47:41 -07001766 if (mm->core_state || !get_dumpable(mm)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001767 up_write(&mm->mmap_sem);
David Howellsd84f4f92008-11-14 10:39:23 +11001768 put_cred(cred);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001769 goto fail;
1770 }
Alan Coxd6e71142005-06-23 00:09:43 -07001771
1772 /*
1773 * We cannot trust fsuid as being the "true" uid of the
1774 * process nor do we know its entire history. We only know it
1775 * was tainted so we dump it as root in mode 2.
1776 */
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001777 if (get_dumpable(mm) == 2) { /* Setuid core dump mode */
Alan Coxd6e71142005-06-23 00:09:43 -07001778 flag = O_EXCL; /* Stop rewrite attacks */
David Howellsd84f4f92008-11-14 10:39:23 +11001779 cred->fsuid = 0; /* Dump root private */
Alan Coxd6e71142005-06-23 00:09:43 -07001780 }
Oleg Nesterov1291cf42005-10-30 15:02:54 -08001781
Oleg Nesterov9d5b3272008-07-25 01:47:43 -07001782 retval = coredump_wait(exit_code, &core_state);
David Howellsd84f4f92008-11-14 10:39:23 +11001783 if (retval < 0) {
1784 put_cred(cred);
Oleg Nesterov1291cf42005-10-30 15:02:54 -08001785 goto fail;
David Howellsd84f4f92008-11-14 10:39:23 +11001786 }
1787
1788 old_cred = override_creds(cred);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001789
1790 /*
1791 * Clear any false indication of pending signals that might
1792 * be seen by the filesystem code called to write the core file.
1793 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001794 clear_thread_flag(TIF_SIGPENDING);
1795
Linus Torvalds1da177e2005-04-16 15:20:36 -07001796 /*
1797 * lock_kernel() because format_corename() is controlled by sysctl, which
1798 * uses lock_kernel()
1799 */
1800 lock_kernel();
Oleg Nesterov64093242008-10-18 20:28:22 -07001801 ispipe = format_corename(corename, signr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001802 unlock_kernel();
Neil Horman7dc0b222007-10-16 23:26:34 -07001803 /*
1804 * Don't bother to check the RLIMIT_CORE value if core_pattern points
1805 * to a pipe. Since we're not writing directly to the filesystem
1806 * RLIMIT_CORE doesn't really apply, as no actual core file will be
1807 * created unless the pipe reader choses to write out the core file
1808 * at which point file size limits and permissions will be imposed
1809 * as it does with any other process
1810 */
Neil Horman74aadce2007-10-16 23:26:35 -07001811 if ((!ispipe) && (core_limit < binfmt->min_coredump))
Neil Horman7dc0b222007-10-16 23:26:34 -07001812 goto fail_unlock;
1813
Alan Coxc4bbafd2007-04-16 22:53:13 -07001814 if (ispipe) {
Neil Horman74aadce2007-10-16 23:26:35 -07001815 helper_argv = argv_split(GFP_KERNEL, corename+1, &helper_argc);
Tetsuo Handa350eaf72009-01-06 14:41:11 -08001816 if (!helper_argv) {
1817 printk(KERN_WARNING "%s failed to allocate memory\n",
1818 __func__);
1819 goto fail_unlock;
1820 }
Neil Horman74aadce2007-10-16 23:26:35 -07001821 /* Terminate the string before the first option */
1822 delimit = strchr(corename, ' ');
1823 if (delimit)
1824 *delimit = '\0';
Neil Horman32321132007-10-16 23:26:36 -07001825 delimit = strrchr(helper_argv[0], '/');
1826 if (delimit)
1827 delimit++;
1828 else
1829 delimit = helper_argv[0];
1830 if (!strcmp(delimit, current->comm)) {
1831 printk(KERN_NOTICE "Recursive core dump detected, "
1832 "aborting\n");
1833 goto fail_unlock;
1834 }
1835
1836 core_limit = RLIM_INFINITY;
1837
Andi Kleend025c9d2006-09-30 23:29:28 -07001838 /* SIGPIPE can happen, but it's just never processed */
Neil Horman32321132007-10-16 23:26:36 -07001839 if (call_usermodehelper_pipe(corename+1, helper_argv, NULL,
1840 &file)) {
Andi Kleend025c9d2006-09-30 23:29:28 -07001841 printk(KERN_INFO "Core dump to %s pipe failed\n",
1842 corename);
1843 goto fail_unlock;
1844 }
Andi Kleend025c9d2006-09-30 23:29:28 -07001845 } else
1846 file = filp_open(corename,
Alexey Dobriyan6d4df672006-12-06 20:40:39 -08001847 O_CREAT | 2 | O_NOFOLLOW | O_LARGEFILE | flag,
1848 0600);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001849 if (IS_ERR(file))
1850 goto fail_unlock;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001851 inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001852 if (inode->i_nlink > 1)
1853 goto close_fail; /* multiple links - don't dump */
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001854 if (!ispipe && d_unhashed(file->f_path.dentry))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001855 goto close_fail;
1856
Andi Kleend025c9d2006-09-30 23:29:28 -07001857 /* AK: actually i see no reason to not allow this for named pipes etc.,
1858 but keep the previous behaviour for now. */
1859 if (!ispipe && !S_ISREG(inode->i_mode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001860 goto close_fail;
Ingo Molnarc46f7392007-11-28 13:59:18 +01001861 /*
1862 * Dont allow local users get cute and trick others to coredump
1863 * into their pre-created files:
1864 */
David Howellsda9592e2008-11-14 10:39:05 +11001865 if (inode->i_uid != current_fsuid())
Ingo Molnarc46f7392007-11-28 13:59:18 +01001866 goto close_fail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001867 if (!file->f_op)
1868 goto close_fail;
1869 if (!file->f_op->write)
1870 goto close_fail;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001871 if (!ispipe && do_truncate(file->f_path.dentry, 0, 0, file) != 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001872 goto close_fail;
1873
Neil Horman7dc0b222007-10-16 23:26:34 -07001874 retval = binfmt->core_dump(signr, regs, file, core_limit);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001875
1876 if (retval)
1877 current->signal->group_exit_code |= 0x80;
1878close_fail:
1879 filp_close(file, NULL);
1880fail_unlock:
Neil Horman74aadce2007-10-16 23:26:35 -07001881 if (helper_argv)
1882 argv_free(helper_argv);
1883
David Howellsd84f4f92008-11-14 10:39:23 +11001884 revert_creds(old_cred);
1885 put_cred(cred);
Oleg Nesterova94e2d42008-07-25 01:47:46 -07001886 coredump_finish(mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001887fail:
WANG Cong8cd3ac32009-01-06 14:42:48 -08001888 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001889}