blob: 5c833c18d0d47bef760c9a4053f5188e34781105 [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>
Ingo Molnarcdd6c482009-09-21 12:02:48 +020036#include <linux/perf_event.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070037#include <linux/highmem.h>
38#include <linux/spinlock.h>
39#include <linux/key.h>
40#include <linux/personality.h>
41#include <linux/binfmts.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070042#include <linux/utsname.h>
Sukadev Bhattiprolu84d73782006-12-08 02:38:01 -080043#include <linux/pid_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070044#include <linux/module.h>
45#include <linux/namei.h>
46#include <linux/proc_fs.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070047#include <linux/mount.h>
48#include <linux/security.h>
Mimi Zohar6146f0d2009-02-04 09:06:57 -050049#include <linux/ima.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070050#include <linux/syscalls.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>
Roland McGrath6341c392008-07-25 19:45:44 -070054#include <linux/tracehook.h>
Johannes Berg5f4123b2008-07-09 10:28:40 +020055#include <linux/kmod.h>
Eric Paris6110e3a2008-12-17 13:53:20 -050056#include <linux/fsnotify.h>
Al Viro5ad4e532009-03-29 19:50:06 -040057#include <linux/fs_struct.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070058
59#include <asm/uaccess.h>
60#include <asm/mmu_context.h>
Ollie Wildb6a2fea2007-07-19 01:48:16 -070061#include <asm/tlb.h>
David Howellsa6f76f22008-11-14 10:39:24 +110062#include "internal.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070063
Linus Torvalds1da177e2005-04-16 15:20:36 -070064int core_uses_pid;
Dan Aloni71ce92f2007-05-16 22:11:16 -070065char core_pattern[CORENAME_MAX_SIZE] = "core";
Alan Coxd6e71142005-06-23 00:09:43 -070066int suid_dumpable = 0;
67
Linus Torvalds1da177e2005-04-16 15:20:36 -070068/* The maximal length of core_pattern is also specified in sysctl.c */
69
Alexey Dobriyane4dc1b12007-10-16 23:26:03 -070070static LIST_HEAD(formats);
Linus Torvalds1da177e2005-04-16 15:20:36 -070071static DEFINE_RWLOCK(binfmt_lock);
72
Ivan Kokshaysky74641f52009-04-30 15:08:49 -070073int __register_binfmt(struct linux_binfmt * fmt, int insert)
Linus Torvalds1da177e2005-04-16 15:20:36 -070074{
Linus Torvalds1da177e2005-04-16 15:20:36 -070075 if (!fmt)
76 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -070077 write_lock(&binfmt_lock);
Ivan Kokshaysky74641f52009-04-30 15:08:49 -070078 insert ? list_add(&fmt->lh, &formats) :
79 list_add_tail(&fmt->lh, &formats);
Linus Torvalds1da177e2005-04-16 15:20:36 -070080 write_unlock(&binfmt_lock);
81 return 0;
82}
83
Ivan Kokshaysky74641f52009-04-30 15:08:49 -070084EXPORT_SYMBOL(__register_binfmt);
Linus Torvalds1da177e2005-04-16 15:20:36 -070085
Alexey Dobriyanf6b450d2007-10-16 23:26:04 -070086void unregister_binfmt(struct linux_binfmt * fmt)
Linus Torvalds1da177e2005-04-16 15:20:36 -070087{
Linus Torvalds1da177e2005-04-16 15:20:36 -070088 write_lock(&binfmt_lock);
Alexey Dobriyane4dc1b12007-10-16 23:26:03 -070089 list_del(&fmt->lh);
Linus Torvalds1da177e2005-04-16 15:20:36 -070090 write_unlock(&binfmt_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -070091}
92
93EXPORT_SYMBOL(unregister_binfmt);
94
95static inline void put_binfmt(struct linux_binfmt * fmt)
96{
97 module_put(fmt->module);
98}
99
100/*
101 * Note that a shared library must be both readable and executable due to
102 * security reasons.
103 *
104 * Also note that we take the address to load from from the file itself.
105 */
Heiko Carstens1e7bfb22009-01-14 14:14:29 +0100106SYSCALL_DEFINE1(uselib, const char __user *, library)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700107{
Al Viro964bd182008-07-26 03:33:14 -0400108 struct file *file;
Al Viro964bd182008-07-26 03:33:14 -0400109 char *tmp = getname(library);
110 int error = PTR_ERR(tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700111
Al Viro6e8341a2009-04-06 11:16:22 -0400112 if (IS_ERR(tmp))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113 goto out;
114
Al Viro6e8341a2009-04-06 11:16:22 -0400115 file = do_filp_open(AT_FDCWD, tmp,
116 O_LARGEFILE | O_RDONLY | FMODE_EXEC, 0,
117 MAY_READ | MAY_EXEC | MAY_OPEN);
118 putname(tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700119 error = PTR_ERR(file);
120 if (IS_ERR(file))
121 goto out;
122
Al Viro6e8341a2009-04-06 11:16:22 -0400123 error = -EINVAL;
124 if (!S_ISREG(file->f_path.dentry->d_inode->i_mode))
125 goto exit;
126
127 error = -EACCES;
128 if (file->f_path.mnt->mnt_flags & MNT_NOEXEC)
129 goto exit;
130
Eric Paris6110e3a2008-12-17 13:53:20 -0500131 fsnotify_open(file->f_path.dentry);
132
Linus Torvalds1da177e2005-04-16 15:20:36 -0700133 error = -ENOEXEC;
134 if(file->f_op) {
135 struct linux_binfmt * fmt;
136
137 read_lock(&binfmt_lock);
Alexey Dobriyane4dc1b12007-10-16 23:26:03 -0700138 list_for_each_entry(fmt, &formats, lh) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700139 if (!fmt->load_shlib)
140 continue;
141 if (!try_module_get(fmt->module))
142 continue;
143 read_unlock(&binfmt_lock);
144 error = fmt->load_shlib(file);
145 read_lock(&binfmt_lock);
146 put_binfmt(fmt);
147 if (error != -ENOEXEC)
148 break;
149 }
150 read_unlock(&binfmt_lock);
151 }
Al Viro6e8341a2009-04-06 11:16:22 -0400152exit:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153 fput(file);
154out:
155 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700156}
157
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700158#ifdef CONFIG_MMU
159
160static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos,
161 int write)
162{
163 struct page *page;
164 int ret;
165
166#ifdef CONFIG_STACK_GROWSUP
167 if (write) {
168 ret = expand_stack_downwards(bprm->vma, pos);
169 if (ret < 0)
170 return NULL;
171 }
172#endif
173 ret = get_user_pages(current, bprm->mm, pos,
174 1, write, 1, &page, NULL);
175 if (ret <= 0)
176 return NULL;
177
178 if (write) {
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700179 unsigned long size = bprm->vma->vm_end - bprm->vma->vm_start;
Linus Torvaldsa64e7152008-03-03 10:12:14 -0800180 struct rlimit *rlim;
181
182 /*
183 * We've historically supported up to 32 pages (ARG_MAX)
184 * of argument strings even with small stacks
185 */
186 if (size <= ARG_MAX)
187 return page;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700188
189 /*
190 * Limit to 1/4-th the stack size for the argv+env strings.
191 * This ensures that:
192 * - the remaining binfmt code will not run out of stack space,
193 * - the program will have a reasonable amount of stack left
194 * to work from.
195 */
Linus Torvaldsa64e7152008-03-03 10:12:14 -0800196 rlim = current->signal->rlim;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700197 if (size > rlim[RLIMIT_STACK].rlim_cur / 4) {
198 put_page(page);
199 return NULL;
200 }
201 }
202
203 return page;
204}
205
206static void put_arg_page(struct page *page)
207{
208 put_page(page);
209}
210
211static void free_arg_page(struct linux_binprm *bprm, int i)
212{
213}
214
215static void free_arg_pages(struct linux_binprm *bprm)
216{
217}
218
219static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos,
220 struct page *page)
221{
222 flush_cache_page(bprm->vma, pos, page_to_pfn(page));
223}
224
225static int __bprm_mm_init(struct linux_binprm *bprm)
226{
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800227 int err;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700228 struct vm_area_struct *vma = NULL;
229 struct mm_struct *mm = bprm->mm;
230
231 bprm->vma = vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL);
232 if (!vma)
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800233 return -ENOMEM;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700234
235 down_write(&mm->mmap_sem);
236 vma->vm_mm = mm;
237
238 /*
239 * Place the stack at the largest stack address the architecture
240 * supports. Later, we'll move this to an appropriate place. We don't
241 * use STACK_TOP because that can depend on attributes which aren't
242 * configured yet.
243 */
244 vma->vm_end = STACK_TOP_MAX;
245 vma->vm_start = vma->vm_end - PAGE_SIZE;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700246 vma->vm_flags = VM_STACK_FLAGS;
Coly Li3ed75eb2007-10-18 23:39:15 -0700247 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700248 err = insert_vm_struct(mm, vma);
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800249 if (err)
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700250 goto err;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700251
252 mm->stack_vm = mm->total_vm = 1;
253 up_write(&mm->mmap_sem);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700254 bprm->p = vma->vm_end - sizeof(void *);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700255 return 0;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700256err:
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800257 up_write(&mm->mmap_sem);
258 bprm->vma = NULL;
259 kmem_cache_free(vm_area_cachep, vma);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700260 return err;
261}
262
263static bool valid_arg_len(struct linux_binprm *bprm, long len)
264{
265 return len <= MAX_ARG_STRLEN;
266}
267
268#else
269
270static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos,
271 int write)
272{
273 struct page *page;
274
275 page = bprm->page[pos / PAGE_SIZE];
276 if (!page && write) {
277 page = alloc_page(GFP_HIGHUSER|__GFP_ZERO);
278 if (!page)
279 return NULL;
280 bprm->page[pos / PAGE_SIZE] = page;
281 }
282
283 return page;
284}
285
286static void put_arg_page(struct page *page)
287{
288}
289
290static void free_arg_page(struct linux_binprm *bprm, int i)
291{
292 if (bprm->page[i]) {
293 __free_page(bprm->page[i]);
294 bprm->page[i] = NULL;
295 }
296}
297
298static void free_arg_pages(struct linux_binprm *bprm)
299{
300 int i;
301
302 for (i = 0; i < MAX_ARG_PAGES; i++)
303 free_arg_page(bprm, i);
304}
305
306static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos,
307 struct page *page)
308{
309}
310
311static int __bprm_mm_init(struct linux_binprm *bprm)
312{
313 bprm->p = PAGE_SIZE * MAX_ARG_PAGES - sizeof(void *);
314 return 0;
315}
316
317static bool valid_arg_len(struct linux_binprm *bprm, long len)
318{
319 return len <= bprm->p;
320}
321
322#endif /* CONFIG_MMU */
323
324/*
325 * Create a new mm_struct and populate it with a temporary stack
326 * vm_area_struct. We don't have enough context at this point to set the stack
327 * flags, permissions, and offset, so we use temporary values. We'll update
328 * them later in setup_arg_pages().
329 */
330int bprm_mm_init(struct linux_binprm *bprm)
331{
332 int err;
333 struct mm_struct *mm = NULL;
334
335 bprm->mm = mm = mm_alloc();
336 err = -ENOMEM;
337 if (!mm)
338 goto err;
339
340 err = init_new_context(current, mm);
341 if (err)
342 goto err;
343
344 err = __bprm_mm_init(bprm);
345 if (err)
346 goto err;
347
348 return 0;
349
350err:
351 if (mm) {
352 bprm->mm = NULL;
353 mmdrop(mm);
354 }
355
356 return err;
357}
358
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359/*
360 * count() counts the number of strings in array ARGV.
361 */
362static int count(char __user * __user * argv, int max)
363{
364 int i = 0;
365
366 if (argv != NULL) {
367 for (;;) {
368 char __user * p;
369
370 if (get_user(p, argv))
371 return -EFAULT;
372 if (!p)
373 break;
374 argv++;
Jason Baron362e6662008-10-15 22:01:52 -0700375 if (i++ >= max)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700376 return -E2BIG;
377 cond_resched();
378 }
379 }
380 return i;
381}
382
383/*
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700384 * 'copy_strings()' copies argument/environment strings from the old
385 * processes's memory to the new process's stack. The call to get_user_pages()
386 * ensures the destination page is created and not swapped out.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700387 */
Adrian Bunk75c96f82005-05-05 16:16:09 -0700388static int copy_strings(int argc, char __user * __user * argv,
389 struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700390{
391 struct page *kmapped_page = NULL;
392 char *kaddr = NULL;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700393 unsigned long kpos = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394 int ret;
395
396 while (argc-- > 0) {
397 char __user *str;
398 int len;
399 unsigned long pos;
400
401 if (get_user(str, argv+argc) ||
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700402 !(len = strnlen_user(str, MAX_ARG_STRLEN))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700403 ret = -EFAULT;
404 goto out;
405 }
406
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700407 if (!valid_arg_len(bprm, len)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408 ret = -E2BIG;
409 goto out;
410 }
411
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700412 /* We're going to work our way backwords. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700413 pos = bprm->p;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700414 str += len;
415 bprm->p -= len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416
417 while (len > 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700418 int offset, bytes_to_copy;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700419
420 offset = pos % PAGE_SIZE;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700421 if (offset == 0)
422 offset = PAGE_SIZE;
423
424 bytes_to_copy = offset;
425 if (bytes_to_copy > len)
426 bytes_to_copy = len;
427
428 offset -= bytes_to_copy;
429 pos -= bytes_to_copy;
430 str -= bytes_to_copy;
431 len -= bytes_to_copy;
432
433 if (!kmapped_page || kpos != (pos & PAGE_MASK)) {
434 struct page *page;
435
436 page = get_arg_page(bprm, pos, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437 if (!page) {
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700438 ret = -E2BIG;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700439 goto out;
440 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700442 if (kmapped_page) {
443 flush_kernel_dcache_page(kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444 kunmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700445 put_arg_page(kmapped_page);
446 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447 kmapped_page = page;
448 kaddr = kmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700449 kpos = pos & PAGE_MASK;
450 flush_arg_page(bprm, kpos, kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451 }
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700452 if (copy_from_user(kaddr+offset, str, bytes_to_copy)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453 ret = -EFAULT;
454 goto out;
455 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456 }
457 }
458 ret = 0;
459out:
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700460 if (kmapped_page) {
461 flush_kernel_dcache_page(kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700462 kunmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700463 put_arg_page(kmapped_page);
464 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465 return ret;
466}
467
468/*
469 * Like copy_strings, but get argv and its values from kernel memory.
470 */
471int copy_strings_kernel(int argc,char ** argv, struct linux_binprm *bprm)
472{
473 int r;
474 mm_segment_t oldfs = get_fs();
475 set_fs(KERNEL_DS);
476 r = copy_strings(argc, (char __user * __user *)argv, bprm);
477 set_fs(oldfs);
478 return r;
479}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480EXPORT_SYMBOL(copy_strings_kernel);
481
482#ifdef CONFIG_MMU
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700483
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484/*
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700485 * During bprm_mm_init(), we create a temporary stack at STACK_TOP_MAX. Once
486 * the binfmt code determines where the new stack should reside, we shift it to
487 * its final location. The process proceeds as follows:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488 *
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700489 * 1) Use shift to calculate the new vma endpoints.
490 * 2) Extend vma to cover both the old and new ranges. This ensures the
491 * arguments passed to subsequent functions are consistent.
492 * 3) Move vma's page tables to the new range.
493 * 4) Free up any cleared pgd range.
494 * 5) Shrink the vma to cover only the new range.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495 */
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700496static int shift_arg_pages(struct vm_area_struct *vma, unsigned long shift)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497{
498 struct mm_struct *mm = vma->vm_mm;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700499 unsigned long old_start = vma->vm_start;
500 unsigned long old_end = vma->vm_end;
501 unsigned long length = old_end - old_start;
502 unsigned long new_start = old_start - shift;
503 unsigned long new_end = old_end - shift;
504 struct mmu_gather *tlb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700505
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700506 BUG_ON(new_start > new_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700508 /*
509 * ensure there are no vmas between where we want to go
510 * and where we are
511 */
512 if (vma != find_vma(mm, new_start))
513 return -EFAULT;
514
515 /*
516 * cover the whole range: [new_start, old_end)
517 */
518 vma_adjust(vma, new_start, old_end, vma->vm_pgoff, NULL);
519
520 /*
521 * move the page tables downwards, on failure we rely on
522 * process cleanup to remove whatever mess we made.
523 */
524 if (length != move_page_tables(vma, old_start,
525 vma, new_start, length))
526 return -ENOMEM;
527
528 lru_add_drain();
529 tlb = tlb_gather_mmu(mm, 0);
530 if (new_end > old_start) {
531 /*
532 * when the old and new regions overlap clear from new_end.
533 */
Jan Beulich42b77722008-07-23 21:27:10 -0700534 free_pgd_range(tlb, new_end, old_end, new_end,
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700535 vma->vm_next ? vma->vm_next->vm_start : 0);
536 } else {
537 /*
538 * otherwise, clean from old_start; this is done to not touch
539 * the address space in [new_end, old_start) some architectures
540 * have constraints on va-space that make this illegal (IA64) -
541 * for the others its just a little faster.
542 */
Jan Beulich42b77722008-07-23 21:27:10 -0700543 free_pgd_range(tlb, old_start, old_end, new_end,
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700544 vma->vm_next ? vma->vm_next->vm_start : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545 }
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700546 tlb_finish_mmu(tlb, new_end, old_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700548 /*
549 * shrink the vma to just the new range.
550 */
551 vma_adjust(vma, new_start, new_end, vma->vm_pgoff, NULL);
552
553 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554}
555
556#define EXTRA_STACK_VM_PAGES 20 /* random */
557
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700558/*
559 * Finalizes the stack vm_area_struct. The flags and permissions are updated,
560 * the stack is optionally relocated, and some extra space is added.
561 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700562int setup_arg_pages(struct linux_binprm *bprm,
563 unsigned long stack_top,
564 int executable_stack)
565{
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700566 unsigned long ret;
567 unsigned long stack_shift;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568 struct mm_struct *mm = current->mm;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700569 struct vm_area_struct *vma = bprm->vma;
570 struct vm_area_struct *prev = NULL;
571 unsigned long vm_flags;
572 unsigned long stack_base;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573
574#ifdef CONFIG_STACK_GROWSUP
Linus Torvalds1da177e2005-04-16 15:20:36 -0700575 /* Limit stack size to 1GB */
576 stack_base = current->signal->rlim[RLIMIT_STACK].rlim_max;
577 if (stack_base > (1 << 30))
578 stack_base = 1 << 30;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700579
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700580 /* Make sure we didn't let the argument array grow too large. */
581 if (vma->vm_end - vma->vm_start > stack_base)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582 return -ENOMEM;
583
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700584 stack_base = PAGE_ALIGN(stack_top - stack_base);
585
586 stack_shift = vma->vm_start - stack_base;
587 mm->arg_start = bprm->p - stack_shift;
588 bprm->p = vma->vm_end - stack_shift;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589#else
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700590 stack_top = arch_align_stack(stack_top);
591 stack_top = PAGE_ALIGN(stack_top);
592 stack_shift = vma->vm_end - stack_top;
593
594 bprm->p -= stack_shift;
595 mm->arg_start = bprm->p;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596#endif
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700597
598 if (bprm->loader)
599 bprm->loader -= stack_shift;
600 bprm->exec -= stack_shift;
601
602 down_write(&mm->mmap_sem);
Hugh Dickins96a8e132008-07-10 21:19:20 +0100603 vm_flags = VM_STACK_FLAGS;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700604
605 /*
606 * Adjust stack execute permissions; explicitly enable for
607 * EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X and leave alone
608 * (arch default) otherwise.
609 */
610 if (unlikely(executable_stack == EXSTACK_ENABLE_X))
611 vm_flags |= VM_EXEC;
612 else if (executable_stack == EXSTACK_DISABLE_X)
613 vm_flags &= ~VM_EXEC;
614 vm_flags |= mm->def_flags;
615
616 ret = mprotect_fixup(vma, &prev, vma->vm_start, vma->vm_end,
617 vm_flags);
618 if (ret)
619 goto out_unlock;
620 BUG_ON(prev != vma);
621
622 /* Move stack pages down in memory. */
623 if (stack_shift) {
624 ret = shift_arg_pages(vma, stack_shift);
625 if (ret) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626 up_write(&mm->mmap_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627 return ret;
628 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629 }
630
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700631#ifdef CONFIG_STACK_GROWSUP
632 stack_base = vma->vm_end + EXTRA_STACK_VM_PAGES * PAGE_SIZE;
633#else
634 stack_base = vma->vm_start - EXTRA_STACK_VM_PAGES * PAGE_SIZE;
635#endif
636 ret = expand_stack(vma, stack_base);
637 if (ret)
638 ret = -EFAULT;
639
640out_unlock:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700641 up_write(&mm->mmap_sem);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700642 return 0;
643}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700644EXPORT_SYMBOL(setup_arg_pages);
645
Linus Torvalds1da177e2005-04-16 15:20:36 -0700646#endif /* CONFIG_MMU */
647
648struct file *open_exec(const char *name)
649{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650 struct file *file;
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200651 int err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700652
Al Viro6e8341a2009-04-06 11:16:22 -0400653 file = do_filp_open(AT_FDCWD, name,
654 O_LARGEFILE | O_RDONLY | FMODE_EXEC, 0,
655 MAY_EXEC | MAY_OPEN);
656 if (IS_ERR(file))
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200657 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200659 err = -EACCES;
Al Viro6e8341a2009-04-06 11:16:22 -0400660 if (!S_ISREG(file->f_path.dentry->d_inode->i_mode))
661 goto exit;
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200662
Al Viro6e8341a2009-04-06 11:16:22 -0400663 if (file->f_path.mnt->mnt_flags & MNT_NOEXEC)
664 goto exit;
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200665
Eric Paris6110e3a2008-12-17 13:53:20 -0500666 fsnotify_open(file->f_path.dentry);
667
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200668 err = deny_write_access(file);
Al Viro6e8341a2009-04-06 11:16:22 -0400669 if (err)
670 goto exit;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671
Al Viro6e8341a2009-04-06 11:16:22 -0400672out:
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200673 return file;
674
Al Viro6e8341a2009-04-06 11:16:22 -0400675exit:
676 fput(file);
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200677 return ERR_PTR(err);
678}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700679EXPORT_SYMBOL(open_exec);
680
Mimi Zohar6777d772009-08-21 14:32:48 -0400681int kernel_read(struct file *file, loff_t offset,
682 char *addr, unsigned long count)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700683{
684 mm_segment_t old_fs;
685 loff_t pos = offset;
686 int result;
687
688 old_fs = get_fs();
689 set_fs(get_ds());
690 /* The cast to a user pointer is valid due to the set_fs() */
691 result = vfs_read(file, (void __user *)addr, count, &pos);
692 set_fs(old_fs);
693 return result;
694}
695
696EXPORT_SYMBOL(kernel_read);
697
698static int exec_mmap(struct mm_struct *mm)
699{
700 struct task_struct *tsk;
701 struct mm_struct * old_mm, *active_mm;
702
703 /* Notify parent that we're no longer interested in the old VM */
704 tsk = current;
705 old_mm = current->mm;
706 mm_release(tsk, old_mm);
707
708 if (old_mm) {
709 /*
710 * Make sure that if there is a core dump in progress
711 * for the old mm, we get out and die instead of going
712 * through with the exec. We must hold mmap_sem around
Oleg Nesterov999d9fc2008-07-25 01:47:41 -0700713 * checking core_state and changing tsk->mm.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700714 */
715 down_read(&old_mm->mmap_sem);
Oleg Nesterov999d9fc2008-07-25 01:47:41 -0700716 if (unlikely(old_mm->core_state)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700717 up_read(&old_mm->mmap_sem);
718 return -EINTR;
719 }
720 }
721 task_lock(tsk);
722 active_mm = tsk->active_mm;
723 tsk->mm = mm;
724 tsk->active_mm = mm;
725 activate_mm(active_mm, mm);
726 task_unlock(tsk);
727 arch_pick_mmap_layout(mm);
728 if (old_mm) {
729 up_read(&old_mm->mmap_sem);
Eric Sesterhenn7dddb122006-04-01 01:13:38 +0200730 BUG_ON(active_mm != old_mm);
Balbir Singh31a78f22008-09-28 23:09:31 +0100731 mm_update_next_owner(old_mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700732 mmput(old_mm);
733 return 0;
734 }
735 mmdrop(active_mm);
736 return 0;
737}
738
739/*
740 * This function makes sure the current process has its own signal table,
741 * so that flush_signal_handlers can later reset the handlers without
742 * disturbing other processes. (Other processes might share the signal
743 * table via the CLONE_SIGHAND option to clone().)
744 */
Arjan van de Ven858119e2006-01-14 13:20:43 -0800745static int de_thread(struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746{
747 struct signal_struct *sig = tsk->signal;
Oleg Nesterovb2c903b2007-10-16 23:27:22 -0700748 struct sighand_struct *oldsighand = tsk->sighand;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749 spinlock_t *lock = &oldsighand->siglock;
750 int count;
751
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700752 if (thread_group_empty(tsk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753 goto no_thread_group;
754
755 /*
756 * Kill all other threads in the thread group.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700757 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758 spin_lock_irq(lock);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -0800759 if (signal_group_exit(sig)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700760 /*
761 * Another group action in progress, just
762 * return so that the signal is processed.
763 */
764 spin_unlock_irq(lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700765 return -EAGAIN;
766 }
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -0800767 sig->group_exit_task = tsk;
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700768 zap_other_threads(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769
Oleg Nesterovfea9d172008-02-08 04:19:19 -0800770 /* Account for the thread group leader hanging around: */
771 count = thread_group_leader(tsk) ? 1 : 2;
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700772 sig->notify_count = count;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773 while (atomic_read(&sig->count) > count) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774 __set_current_state(TASK_UNINTERRUPTIBLE);
775 spin_unlock_irq(lock);
776 schedule();
777 spin_lock_irq(lock);
778 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779 spin_unlock_irq(lock);
780
781 /*
782 * At this point all other threads have exited, all we have to
783 * do is to wait for the thread group leader to become inactive,
784 * and to assume its PID:
785 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700786 if (!thread_group_leader(tsk)) {
Oleg Nesterov81879262008-12-01 14:18:16 -0800787 struct task_struct *leader = tsk->group_leader;
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700788
Oleg Nesterov2800d8d2008-04-30 00:53:12 -0700789 sig->notify_count = -1; /* for exit_notify() */
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700790 for (;;) {
791 write_lock_irq(&tasklist_lock);
792 if (likely(leader->exit_state))
793 break;
794 __set_current_state(TASK_UNINTERRUPTIBLE);
795 write_unlock_irq(&tasklist_lock);
796 schedule();
797 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700798
Roland McGrathf5e90282006-04-10 22:54:16 -0700799 /*
800 * The only record we have of the real-time age of a
801 * process, regardless of execs it's done, is start_time.
802 * All the past CPU time is accumulated in signal_struct
803 * from sister threads now dead. But in this non-leader
804 * exec, nothing survives from the original leader thread,
805 * whose birth marks the true age of this process now.
806 * When we take on its identity by switching to its PID, we
807 * also take its birthdate (always earlier than our own).
808 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700809 tsk->start_time = leader->start_time;
Roland McGrathf5e90282006-04-10 22:54:16 -0700810
Pavel Emelyanovbac0abd2007-10-18 23:40:18 -0700811 BUG_ON(!same_thread_group(leader, tsk));
812 BUG_ON(has_group_leader_pid(tsk));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813 /*
814 * An exec() starts a new thread group with the
815 * TGID of the previous thread group. Rehash the
816 * two threads with a switched PID, and release
817 * the former thread group leader:
818 */
Eric W. Biedermand73d6522006-03-28 16:11:03 -0800819
820 /* Become a process group leader with the old leader's pid.
Eric W. Biedermanc18258c2006-09-27 01:51:06 -0700821 * The old leader becomes a thread of the this thread group.
822 * Note: The old leader also uses this pid until release_task
Eric W. Biedermand73d6522006-03-28 16:11:03 -0800823 * is called. Odd but simple and correct.
824 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700825 detach_pid(tsk, PIDTYPE_PID);
826 tsk->pid = leader->pid;
Sukadev Bhattiprolu3743ca02007-10-18 23:39:51 -0700827 attach_pid(tsk, PIDTYPE_PID, task_pid(leader));
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700828 transfer_pid(leader, tsk, PIDTYPE_PGID);
829 transfer_pid(leader, tsk, PIDTYPE_SID);
830 list_replace_rcu(&leader->tasks, &tsk->tasks);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700832 tsk->group_leader = tsk;
833 leader->group_leader = tsk;
Eric W. Biedermande12a782006-04-10 17:16:49 -0600834
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700835 tsk->exit_signal = SIGCHLD;
Oleg Nesterov962b5642005-11-23 13:37:43 -0800836
837 BUG_ON(leader->exit_state != EXIT_ZOMBIE);
838 leader->exit_state = EXIT_DEAD;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700839 write_unlock_irq(&tasklist_lock);
Oleg Nesterov81879262008-12-01 14:18:16 -0800840
841 release_task(leader);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -0800842 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700843
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700844 sig->group_exit_task = NULL;
845 sig->notify_count = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700846
847no_thread_group:
Jiri Pirko1f102062009-09-22 16:44:10 -0700848 if (current->mm)
849 setmax_mm_hiwater_rss(&sig->maxrss, current->mm);
850
Linus Torvalds1da177e2005-04-16 15:20:36 -0700851 exit_itimers(sig);
Oleg Nesterovcbaffba2008-05-26 20:55:42 +0400852 flush_itimer_signals();
Oleg Nesterov329f7db2005-11-07 21:12:43 +0300853
Oleg Nesterovb2c903b2007-10-16 23:27:22 -0700854 if (atomic_read(&oldsighand->count) != 1) {
855 struct sighand_struct *newsighand;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700856 /*
Oleg Nesterovb2c903b2007-10-16 23:27:22 -0700857 * This ->sighand is shared with the CLONE_SIGHAND
858 * but not CLONE_THREAD task, switch to the new one.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859 */
Oleg Nesterovb2c903b2007-10-16 23:27:22 -0700860 newsighand = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
861 if (!newsighand)
862 return -ENOMEM;
863
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864 atomic_set(&newsighand->count, 1);
865 memcpy(newsighand->action, oldsighand->action,
866 sizeof(newsighand->action));
867
868 write_lock_irq(&tasklist_lock);
869 spin_lock(&oldsighand->siglock);
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700870 rcu_assign_pointer(tsk->sighand, newsighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871 spin_unlock(&oldsighand->siglock);
872 write_unlock_irq(&tasklist_lock);
873
Davide Libenzifba2afa2007-05-10 22:23:13 -0700874 __cleanup_sighand(oldsighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875 }
876
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700877 BUG_ON(!thread_group_leader(tsk));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878 return 0;
879}
Oleg Nesterov0840a902007-10-16 23:27:22 -0700880
Linus Torvalds1da177e2005-04-16 15:20:36 -0700881/*
882 * These functions flushes out all traces of the currently running executable
883 * so that a new one can be started
884 */
Arjan van de Ven858119e2006-01-14 13:20:43 -0800885static void flush_old_files(struct files_struct * files)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886{
887 long j = -1;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700888 struct fdtable *fdt;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700889
890 spin_lock(&files->file_lock);
891 for (;;) {
892 unsigned long set, i;
893
894 j++;
895 i = j * __NFDBITS;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700896 fdt = files_fdtable(files);
Vadim Lobanovbbea9f62006-12-10 02:21:12 -0800897 if (i >= fdt->max_fds)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898 break;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700899 set = fdt->close_on_exec->fds_bits[j];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700900 if (!set)
901 continue;
Dipankar Sarmabadf1662005-09-09 13:04:10 -0700902 fdt->close_on_exec->fds_bits[j] = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903 spin_unlock(&files->file_lock);
904 for ( ; set ; i++,set >>= 1) {
905 if (set & 1) {
906 sys_close(i);
907 }
908 }
909 spin_lock(&files->file_lock);
910
911 }
912 spin_unlock(&files->file_lock);
913}
914
Andrew Morton59714d62008-02-04 22:27:21 -0800915char *get_task_comm(char *buf, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916{
917 /* buf must be at least sizeof(tsk->comm) in size */
918 task_lock(tsk);
919 strncpy(buf, tsk->comm, sizeof(tsk->comm));
920 task_unlock(tsk);
Andrew Morton59714d62008-02-04 22:27:21 -0800921 return buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922}
923
924void set_task_comm(struct task_struct *tsk, char *buf)
925{
926 task_lock(tsk);
927 strlcpy(tsk->comm, buf, sizeof(tsk->comm));
928 task_unlock(tsk);
Ingo Molnarcdd6c482009-09-21 12:02:48 +0200929 perf_event_comm(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930}
931
932int flush_old_exec(struct linux_binprm * bprm)
933{
934 char * name;
935 int i, ch, retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700936 char tcomm[sizeof(current->comm)];
937
938 /*
939 * Make sure we have a private signal table and that
940 * we are unassociated from the previous thread group.
941 */
942 retval = de_thread(current);
943 if (retval)
944 goto out;
945
Matt Helsley925d1c42008-04-29 01:01:36 -0700946 set_mm_exe_file(bprm->mm, bprm->file);
947
Linus Torvalds1da177e2005-04-16 15:20:36 -0700948 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700949 * Release all of the old mmap stuff
950 */
951 retval = exec_mmap(bprm->mm);
952 if (retval)
Al Virofd8328b2008-04-22 05:11:59 -0400953 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700954
955 bprm->mm = NULL; /* We're using it now */
956
957 /* This is the point of no return */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700958 current->sas_ss_sp = current->sas_ss_size = 0;
959
David Howellsda9592e2008-11-14 10:39:05 +1100960 if (current_euid() == current_uid() && current_egid() == current_gid())
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -0700961 set_dumpable(current->mm, 1);
Alan Coxd6e71142005-06-23 00:09:43 -0700962 else
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -0700963 set_dumpable(current->mm, suid_dumpable);
Alan Coxd6e71142005-06-23 00:09:43 -0700964
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965 name = bprm->filename;
Paolo 'Blaisorblade' Giarrusso36772092005-05-05 16:16:12 -0700966
967 /* Copies the binary name from after last slash */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968 for (i=0; (ch = *(name++)) != '\0';) {
969 if (ch == '/')
Paolo 'Blaisorblade' Giarrusso36772092005-05-05 16:16:12 -0700970 i = 0; /* overwrite what we wrote */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700971 else
972 if (i < (sizeof(tcomm) - 1))
973 tcomm[i++] = ch;
974 }
975 tcomm[i] = '\0';
976 set_task_comm(current, tcomm);
977
978 current->flags &= ~PF_RANDOMIZE;
979 flush_thread();
980
Benjamin Herrenschmidt0551fbd2006-02-28 16:59:19 -0800981 /* Set the new mm task size. We have to do that late because it may
982 * depend on TIF_32BIT which is only updated in flush_thread() on
983 * some architectures like powerpc
984 */
985 current->mm->task_size = TASK_SIZE;
986
David Howellsa6f76f22008-11-14 10:39:24 +1100987 /* install the new credentials */
988 if (bprm->cred->uid != current_euid() ||
989 bprm->cred->gid != current_egid()) {
Marcel Holtmannd2d56c52007-08-17 21:47:58 +0200990 current->pdeath_signal = 0;
991 } else if (file_permission(bprm->file, MAY_READ) ||
David Howellsa6f76f22008-11-14 10:39:24 +1100992 bprm->interp_flags & BINPRM_FLAGS_ENFORCE_NONDUMP) {
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -0700993 set_dumpable(current->mm, suid_dumpable);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994 }
995
David Howellsa6f76f22008-11-14 10:39:24 +1100996 current->personality &= ~bprm->per_clear;
997
Ingo Molnarf65cb452008-12-16 13:40:44 +0100998 /*
999 * Flush performance counters when crossing a
1000 * security domain:
1001 */
1002 if (!get_dumpable(current->mm))
Ingo Molnarcdd6c482009-09-21 12:02:48 +02001003 perf_event_exit_task(current);
Ingo Molnarf65cb452008-12-16 13:40:44 +01001004
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005 /* An exec changes our domain. We are no longer part of the thread
1006 group */
1007
1008 current->self_exec_id++;
1009
1010 flush_signal_handlers(current, 0);
1011 flush_old_files(current->files);
1012
1013 return 0;
1014
Linus Torvalds1da177e2005-04-16 15:20:36 -07001015out:
1016 return retval;
1017}
1018
1019EXPORT_SYMBOL(flush_old_exec);
1020
David Howellsa6f76f22008-11-14 10:39:24 +11001021/*
Oleg Nesterova2a84742009-09-05 11:17:13 -07001022 * Prepare credentials and lock ->cred_guard_mutex.
1023 * install_exec_creds() commits the new creds and drops the lock.
1024 * Or, if exec fails before, free_bprm() should release ->cred and
1025 * and unlock.
1026 */
1027int prepare_bprm_creds(struct linux_binprm *bprm)
1028{
1029 if (mutex_lock_interruptible(&current->cred_guard_mutex))
1030 return -ERESTARTNOINTR;
1031
1032 bprm->cred = prepare_exec_creds();
1033 if (likely(bprm->cred))
1034 return 0;
1035
1036 mutex_unlock(&current->cred_guard_mutex);
1037 return -ENOMEM;
1038}
1039
1040void free_bprm(struct linux_binprm *bprm)
1041{
1042 free_arg_pages(bprm);
1043 if (bprm->cred) {
1044 mutex_unlock(&current->cred_guard_mutex);
1045 abort_creds(bprm->cred);
1046 }
1047 kfree(bprm);
1048}
1049
1050/*
David Howellsa6f76f22008-11-14 10:39:24 +11001051 * install the new credentials for this executable
1052 */
1053void install_exec_creds(struct linux_binprm *bprm)
1054{
1055 security_bprm_committing_creds(bprm);
1056
1057 commit_creds(bprm->cred);
1058 bprm->cred = NULL;
Oleg Nesterova2a84742009-09-05 11:17:13 -07001059 /*
1060 * cred_guard_mutex must be held at least to this point to prevent
David Howellsa6f76f22008-11-14 10:39:24 +11001061 * ptrace_attach() from altering our determination of the task's
Oleg Nesterova2a84742009-09-05 11:17:13 -07001062 * credentials; any time after this it may be unlocked.
1063 */
David Howellsa6f76f22008-11-14 10:39:24 +11001064 security_bprm_committed_creds(bprm);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001065 mutex_unlock(&current->cred_guard_mutex);
David Howellsa6f76f22008-11-14 10:39:24 +11001066}
1067EXPORT_SYMBOL(install_exec_creds);
1068
1069/*
1070 * determine how safe it is to execute the proposed program
David Howells5e751e92009-05-08 13:55:22 +01001071 * - the caller must hold current->cred_guard_mutex to protect against
David Howellsa6f76f22008-11-14 10:39:24 +11001072 * PTRACE_ATTACH
1073 */
Al Viro498052b2009-03-30 07:20:30 -04001074int check_unsafe_exec(struct linux_binprm *bprm)
David Howellsa6f76f22008-11-14 10:39:24 +11001075{
David Howells0bf2f3a2009-02-06 11:45:46 +00001076 struct task_struct *p = current, *t;
Al Virof1191b52009-03-30 07:35:18 -04001077 unsigned n_fs;
Al Viro498052b2009-03-30 07:20:30 -04001078 int res = 0;
David Howellsa6f76f22008-11-14 10:39:24 +11001079
1080 bprm->unsafe = tracehook_unsafe_exec(p);
1081
David Howells0bf2f3a2009-02-06 11:45:46 +00001082 n_fs = 1;
Al Viro498052b2009-03-30 07:20:30 -04001083 write_lock(&p->fs->lock);
Oleg Nesterov437f7fd2009-04-24 01:02:45 +02001084 rcu_read_lock();
David Howells0bf2f3a2009-02-06 11:45:46 +00001085 for (t = next_thread(p); t != p; t = next_thread(t)) {
1086 if (t->fs == p->fs)
1087 n_fs++;
David Howells0bf2f3a2009-02-06 11:45:46 +00001088 }
Oleg Nesterov437f7fd2009-04-24 01:02:45 +02001089 rcu_read_unlock();
David Howells0bf2f3a2009-02-06 11:45:46 +00001090
Al Virof1191b52009-03-30 07:35:18 -04001091 if (p->fs->users > n_fs) {
David Howellsa6f76f22008-11-14 10:39:24 +11001092 bprm->unsafe |= LSM_UNSAFE_SHARE;
Al Viro498052b2009-03-30 07:20:30 -04001093 } else {
Oleg Nesterov8c652f92009-04-24 01:01:56 +02001094 res = -EAGAIN;
1095 if (!p->fs->in_exec) {
1096 p->fs->in_exec = 1;
1097 res = 1;
1098 }
Al Viro498052b2009-03-30 07:20:30 -04001099 }
Al Viro498052b2009-03-30 07:20:30 -04001100 write_unlock(&p->fs->lock);
1101
1102 return res;
David Howellsa6f76f22008-11-14 10:39:24 +11001103}
1104
Linus Torvalds1da177e2005-04-16 15:20:36 -07001105/*
1106 * Fill the binprm structure from the inode.
1107 * Check permissions, then read the first 128 (BINPRM_BUF_SIZE) bytes
David Howellsa6f76f22008-11-14 10:39:24 +11001108 *
1109 * This may be called multiple times for binary chains (scripts for example).
Linus Torvalds1da177e2005-04-16 15:20:36 -07001110 */
1111int prepare_binprm(struct linux_binprm *bprm)
1112{
David Howellsa6f76f22008-11-14 10:39:24 +11001113 umode_t mode;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001114 struct inode * inode = bprm->file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001115 int retval;
1116
1117 mode = inode->i_mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001118 if (bprm->file->f_op == NULL)
1119 return -EACCES;
1120
David Howellsa6f76f22008-11-14 10:39:24 +11001121 /* clear any previous set[ug]id data from a previous binary */
1122 bprm->cred->euid = current_euid();
1123 bprm->cred->egid = current_egid();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124
David Howellsa6f76f22008-11-14 10:39:24 +11001125 if (!(bprm->file->f_path.mnt->mnt_flags & MNT_NOSUID)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001126 /* Set-uid? */
1127 if (mode & S_ISUID) {
David Howellsa6f76f22008-11-14 10:39:24 +11001128 bprm->per_clear |= PER_CLEAR_ON_SETID;
1129 bprm->cred->euid = inode->i_uid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001130 }
1131
1132 /* Set-gid? */
1133 /*
1134 * If setgid is set but no group execute bit then this
1135 * is a candidate for mandatory locking, not a setgid
1136 * executable.
1137 */
1138 if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) {
David Howellsa6f76f22008-11-14 10:39:24 +11001139 bprm->per_clear |= PER_CLEAR_ON_SETID;
1140 bprm->cred->egid = inode->i_gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001141 }
1142 }
1143
1144 /* fill in binprm security blob */
David Howellsa6f76f22008-11-14 10:39:24 +11001145 retval = security_bprm_set_creds(bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001146 if (retval)
1147 return retval;
David Howellsa6f76f22008-11-14 10:39:24 +11001148 bprm->cred_prepared = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001149
David Howellsa6f76f22008-11-14 10:39:24 +11001150 memset(bprm->buf, 0, BINPRM_BUF_SIZE);
1151 return kernel_read(bprm->file, 0, bprm->buf, BINPRM_BUF_SIZE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001152}
1153
1154EXPORT_SYMBOL(prepare_binprm);
1155
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001156/*
1157 * Arguments are '\0' separated strings found at the location bprm->p
1158 * points to; chop off the first by relocating brpm->p to right after
1159 * the first '\0' encountered.
1160 */
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001161int remove_arg_zero(struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162{
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001163 int ret = 0;
1164 unsigned long offset;
1165 char *kaddr;
1166 struct page *page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001167
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001168 if (!bprm->argc)
1169 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001170
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001171 do {
1172 offset = bprm->p & ~PAGE_MASK;
1173 page = get_arg_page(bprm, bprm->p, 0);
1174 if (!page) {
1175 ret = -EFAULT;
1176 goto out;
1177 }
1178 kaddr = kmap_atomic(page, KM_USER0);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001179
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001180 for (; offset < PAGE_SIZE && kaddr[offset];
1181 offset++, bprm->p++)
1182 ;
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001183
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001184 kunmap_atomic(kaddr, KM_USER0);
1185 put_arg_page(page);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001186
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001187 if (offset == PAGE_SIZE)
1188 free_arg_page(bprm, (bprm->p >> PAGE_SHIFT) - 1);
1189 } while (offset == PAGE_SIZE);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001190
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001191 bprm->p++;
1192 bprm->argc--;
1193 ret = 0;
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001194
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001195out:
1196 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001197}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001198EXPORT_SYMBOL(remove_arg_zero);
1199
1200/*
1201 * cycle the list of binary formats handler, until one recognizes the image
1202 */
1203int search_binary_handler(struct linux_binprm *bprm,struct pt_regs *regs)
1204{
Roland McGrath85f33462008-12-09 19:36:38 -08001205 unsigned int depth = bprm->recursion_depth;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001206 int try,retval;
1207 struct linux_binfmt *fmt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001208
Linus Torvalds1da177e2005-04-16 15:20:36 -07001209 retval = security_bprm_check(bprm);
1210 if (retval)
1211 return retval;
Mimi Zohar6146f0d2009-02-04 09:06:57 -05001212 retval = ima_bprm_check(bprm);
1213 if (retval)
1214 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001215
1216 /* kernel module loader fixup */
1217 /* so we don't try to load run modprobe in kernel space. */
1218 set_fs(USER_DS);
Al Viro473ae302006-04-26 14:04:08 -04001219
1220 retval = audit_bprm(bprm);
1221 if (retval)
1222 return retval;
1223
Linus Torvalds1da177e2005-04-16 15:20:36 -07001224 retval = -ENOENT;
1225 for (try=0; try<2; try++) {
1226 read_lock(&binfmt_lock);
Alexey Dobriyane4dc1b12007-10-16 23:26:03 -07001227 list_for_each_entry(fmt, &formats, lh) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001228 int (*fn)(struct linux_binprm *, struct pt_regs *) = fmt->load_binary;
1229 if (!fn)
1230 continue;
1231 if (!try_module_get(fmt->module))
1232 continue;
1233 read_unlock(&binfmt_lock);
1234 retval = fn(bprm, regs);
Roland McGrath85f33462008-12-09 19:36:38 -08001235 /*
1236 * Restore the depth counter to its starting value
1237 * in this call, so we don't have to rely on every
1238 * load_binary function to restore it on return.
1239 */
1240 bprm->recursion_depth = depth;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001241 if (retval >= 0) {
Roland McGrath85f33462008-12-09 19:36:38 -08001242 if (depth == 0)
1243 tracehook_report_exec(fmt, bprm, regs);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001244 put_binfmt(fmt);
1245 allow_write_access(bprm->file);
1246 if (bprm->file)
1247 fput(bprm->file);
1248 bprm->file = NULL;
1249 current->did_exec = 1;
Matt Helsley9f460802005-11-07 00:59:16 -08001250 proc_exec_connector(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001251 return retval;
1252 }
1253 read_lock(&binfmt_lock);
1254 put_binfmt(fmt);
1255 if (retval != -ENOEXEC || bprm->mm == NULL)
1256 break;
1257 if (!bprm->file) {
1258 read_unlock(&binfmt_lock);
1259 return retval;
1260 }
1261 }
1262 read_unlock(&binfmt_lock);
1263 if (retval != -ENOEXEC || bprm->mm == NULL) {
1264 break;
Johannes Berg5f4123b2008-07-09 10:28:40 +02001265#ifdef CONFIG_MODULES
1266 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001267#define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e))
1268 if (printable(bprm->buf[0]) &&
1269 printable(bprm->buf[1]) &&
1270 printable(bprm->buf[2]) &&
1271 printable(bprm->buf[3]))
1272 break; /* -ENOEXEC */
1273 request_module("binfmt-%04x", *(unsigned short *)(&bprm->buf[2]));
1274#endif
1275 }
1276 }
1277 return retval;
1278}
1279
1280EXPORT_SYMBOL(search_binary_handler);
1281
1282/*
1283 * sys_execve() executes a new program.
1284 */
1285int do_execve(char * filename,
1286 char __user *__user *argv,
1287 char __user *__user *envp,
1288 struct pt_regs * regs)
1289{
1290 struct linux_binprm *bprm;
1291 struct file *file;
Al Viro3b125382008-04-22 05:31:30 -04001292 struct files_struct *displaced;
Oleg Nesterov8c652f92009-04-24 01:01:56 +02001293 bool clear_in_exec;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001294 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001295
Al Viro3b125382008-04-22 05:31:30 -04001296 retval = unshare_files(&displaced);
Al Virofd8328b2008-04-22 05:11:59 -04001297 if (retval)
1298 goto out_ret;
1299
Linus Torvalds1da177e2005-04-16 15:20:36 -07001300 retval = -ENOMEM;
Oliver Neukum11b0b5a2006-03-25 03:08:13 -08001301 bprm = kzalloc(sizeof(*bprm), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001302 if (!bprm)
Al Virofd8328b2008-04-22 05:11:59 -04001303 goto out_files;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001304
Oleg Nesterova2a84742009-09-05 11:17:13 -07001305 retval = prepare_bprm_creds(bprm);
1306 if (retval)
David Howellsa6f76f22008-11-14 10:39:24 +11001307 goto out_free;
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)
Oleg Nesterova2a84742009-09-05 11:17:13 -07001311 goto out_free;
Oleg Nesterov8c652f92009-04-24 01:01:56 +02001312 clear_in_exec = retval;
Oleg Nesterova2a84742009-09-05 11:17:13 -07001313 current->in_execve = 1;
David Howellsa6f76f22008-11-14 10:39:24 +11001314
Linus Torvalds1da177e2005-04-16 15:20:36 -07001315 file = open_exec(filename);
1316 retval = PTR_ERR(file);
1317 if (IS_ERR(file))
Al Viro498052b2009-03-30 07:20:30 -04001318 goto out_unmark;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001319
1320 sched_exec();
1321
Linus Torvalds1da177e2005-04-16 15:20:36 -07001322 bprm->file = file;
1323 bprm->filename = filename;
1324 bprm->interp = filename;
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001325
1326 retval = bprm_mm_init(bprm);
1327 if (retval)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001328 goto out_file;
1329
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001330 bprm->argc = count(argv, MAX_ARG_STRINGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001331 if ((retval = bprm->argc) < 0)
David Howellsa6f76f22008-11-14 10:39:24 +11001332 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001333
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001334 bprm->envc = count(envp, MAX_ARG_STRINGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001335 if ((retval = bprm->envc) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001336 goto out;
1337
1338 retval = prepare_binprm(bprm);
1339 if (retval < 0)
1340 goto out;
1341
1342 retval = copy_strings_kernel(1, &bprm->filename, bprm);
1343 if (retval < 0)
1344 goto out;
1345
1346 bprm->exec = bprm->p;
1347 retval = copy_strings(bprm->envc, envp, bprm);
1348 if (retval < 0)
1349 goto out;
1350
1351 retval = copy_strings(bprm->argc, argv, bprm);
1352 if (retval < 0)
1353 goto out;
1354
Oleg Nesterov7b34e422008-07-25 01:47:37 -07001355 current->flags &= ~PF_KTHREAD;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001356 retval = search_binary_handler(bprm,regs);
David Howellsa6f76f22008-11-14 10:39:24 +11001357 if (retval < 0)
1358 goto out;
1359
Stefani Seiboldd899bf72009-09-22 16:45:40 -07001360 current->stack_start = current->mm->start_stack;
1361
David Howellsa6f76f22008-11-14 10:39:24 +11001362 /* execve succeeded */
Al Viro498052b2009-03-30 07:20:30 -04001363 current->fs->in_exec = 0;
Kentaro Takedaf9ce1f12009-02-05 17:18:11 +09001364 current->in_execve = 0;
David Howellsa6f76f22008-11-14 10:39:24 +11001365 acct_update_integrals(current);
1366 free_bprm(bprm);
1367 if (displaced)
1368 put_files_struct(displaced);
1369 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001370
1371out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001372 if (bprm->mm)
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001373 mmput (bprm->mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001374
1375out_file:
1376 if (bprm->file) {
1377 allow_write_access(bprm->file);
1378 fput(bprm->file);
1379 }
David Howellsa6f76f22008-11-14 10:39:24 +11001380
Al Viro498052b2009-03-30 07:20:30 -04001381out_unmark:
Oleg Nesterov8c652f92009-04-24 01:01:56 +02001382 if (clear_in_exec)
1383 current->fs->in_exec = 0;
Kentaro Takedaf9ce1f12009-02-05 17:18:11 +09001384 current->in_execve = 0;
David Howellsa6f76f22008-11-14 10:39:24 +11001385
1386out_free:
Al Viro08a6fac2008-05-10 16:38:25 -04001387 free_bprm(bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001388
Al Virofd8328b2008-04-22 05:11:59 -04001389out_files:
Al Viro3b125382008-04-22 05:31:30 -04001390 if (displaced)
1391 reset_files_struct(displaced);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001392out_ret:
1393 return retval;
1394}
1395
1396int set_binfmt(struct linux_binfmt *new)
1397{
1398 struct linux_binfmt *old = current->binfmt;
1399
1400 if (new) {
1401 if (!try_module_get(new->module))
1402 return -1;
1403 }
1404 current->binfmt = new;
1405 if (old)
1406 module_put(old->module);
1407 return 0;
1408}
1409
1410EXPORT_SYMBOL(set_binfmt);
1411
Linus Torvalds1da177e2005-04-16 15:20:36 -07001412/* format_corename will inspect the pattern parameter, and output a
1413 * name into corename, which must have space for at least
1414 * CORENAME_MAX_SIZE bytes plus one byte for the zero terminator.
1415 */
Oleg Nesterov64093242008-10-18 20:28:22 -07001416static int format_corename(char *corename, long signr)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001417{
David Howells86a264a2008-11-14 10:39:18 +11001418 const struct cred *cred = current_cred();
Oleg Nesterov565b9b12008-07-25 01:47:47 -07001419 const char *pat_ptr = core_pattern;
1420 int ispipe = (*pat_ptr == '|');
Linus Torvalds1da177e2005-04-16 15:20:36 -07001421 char *out_ptr = corename;
1422 char *const out_end = corename + CORENAME_MAX_SIZE;
1423 int rc;
1424 int pid_in_pattern = 0;
1425
1426 /* Repeat as long as we have more pattern to process and more output
1427 space */
1428 while (*pat_ptr) {
1429 if (*pat_ptr != '%') {
1430 if (out_ptr == out_end)
1431 goto out;
1432 *out_ptr++ = *pat_ptr++;
1433 } else {
1434 switch (*++pat_ptr) {
1435 case 0:
1436 goto out;
1437 /* Double percent, output one percent */
1438 case '%':
1439 if (out_ptr == out_end)
1440 goto out;
1441 *out_ptr++ = '%';
1442 break;
1443 /* pid */
1444 case 'p':
1445 pid_in_pattern = 1;
1446 rc = snprintf(out_ptr, out_end - out_ptr,
Pavel Emelyanovb4888932007-10-18 23:40:14 -07001447 "%d", task_tgid_vnr(current));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001448 if (rc > out_end - out_ptr)
1449 goto out;
1450 out_ptr += rc;
1451 break;
1452 /* uid */
1453 case 'u':
1454 rc = snprintf(out_ptr, out_end - out_ptr,
David Howells86a264a2008-11-14 10:39:18 +11001455 "%d", cred->uid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001456 if (rc > out_end - out_ptr)
1457 goto out;
1458 out_ptr += rc;
1459 break;
1460 /* gid */
1461 case 'g':
1462 rc = snprintf(out_ptr, out_end - out_ptr,
David Howells86a264a2008-11-14 10:39:18 +11001463 "%d", cred->gid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001464 if (rc > out_end - out_ptr)
1465 goto out;
1466 out_ptr += rc;
1467 break;
1468 /* signal that caused the coredump */
1469 case 's':
1470 rc = snprintf(out_ptr, out_end - out_ptr,
1471 "%ld", signr);
1472 if (rc > out_end - out_ptr)
1473 goto out;
1474 out_ptr += rc;
1475 break;
1476 /* UNIX time of coredump */
1477 case 't': {
1478 struct timeval tv;
1479 do_gettimeofday(&tv);
1480 rc = snprintf(out_ptr, out_end - out_ptr,
1481 "%lu", tv.tv_sec);
1482 if (rc > out_end - out_ptr)
1483 goto out;
1484 out_ptr += rc;
1485 break;
1486 }
1487 /* hostname */
1488 case 'h':
1489 down_read(&uts_sem);
1490 rc = snprintf(out_ptr, out_end - out_ptr,
Serge E. Hallyne9ff3992006-10-02 02:18:11 -07001491 "%s", utsname()->nodename);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001492 up_read(&uts_sem);
1493 if (rc > out_end - out_ptr)
1494 goto out;
1495 out_ptr += rc;
1496 break;
1497 /* executable */
1498 case 'e':
1499 rc = snprintf(out_ptr, out_end - out_ptr,
1500 "%s", current->comm);
1501 if (rc > out_end - out_ptr)
1502 goto out;
1503 out_ptr += rc;
1504 break;
Neil Horman74aadce2007-10-16 23:26:35 -07001505 /* core limit size */
1506 case 'c':
1507 rc = snprintf(out_ptr, out_end - out_ptr,
1508 "%lu", current->signal->rlim[RLIMIT_CORE].rlim_cur);
1509 if (rc > out_end - out_ptr)
1510 goto out;
1511 out_ptr += rc;
1512 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001513 default:
1514 break;
1515 }
1516 ++pat_ptr;
1517 }
1518 }
1519 /* Backward compatibility with core_uses_pid:
1520 *
1521 * If core_pattern does not include a %p (as is the default)
1522 * and core_uses_pid is set, then .%pid will be appended to
Alan Coxc4bbafd2007-04-16 22:53:13 -07001523 * the filename. Do not do this for piped commands. */
Oleg Nesterov64093242008-10-18 20:28:22 -07001524 if (!ispipe && !pid_in_pattern && core_uses_pid) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001525 rc = snprintf(out_ptr, out_end - out_ptr,
Pavel Emelyanovb4888932007-10-18 23:40:14 -07001526 ".%d", task_tgid_vnr(current));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001527 if (rc > out_end - out_ptr)
1528 goto out;
1529 out_ptr += rc;
1530 }
Alan Coxc4bbafd2007-04-16 22:53:13 -07001531out:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001532 *out_ptr = 0;
Alan Coxc4bbafd2007-04-16 22:53:13 -07001533 return ispipe;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001534}
1535
Oleg Nesterov8cd9c242008-07-25 01:47:42 -07001536static int zap_process(struct task_struct *start)
Oleg Nesterovaceecc042006-06-26 00:26:05 -07001537{
1538 struct task_struct *t;
Oleg Nesterov8cd9c242008-07-25 01:47:42 -07001539 int nr = 0;
Oleg Nesterov281de332006-06-26 00:26:06 -07001540
Oleg Nesterovd5f70c02006-06-26 00:26:07 -07001541 start->signal->flags = SIGNAL_GROUP_EXIT;
1542 start->signal->group_stop_count = 0;
Oleg Nesterovaceecc042006-06-26 00:26:05 -07001543
1544 t = start;
1545 do {
1546 if (t != current && t->mm) {
Oleg Nesterov281de332006-06-26 00:26:06 -07001547 sigaddset(&t->pending.signal, SIGKILL);
1548 signal_wake_up(t, 1);
Oleg Nesterov8cd9c242008-07-25 01:47:42 -07001549 nr++;
Oleg Nesterovaceecc042006-06-26 00:26:05 -07001550 }
Oleg Nesterove4901f92008-07-25 01:47:31 -07001551 } while_each_thread(start, t);
Oleg Nesterov8cd9c242008-07-25 01:47:42 -07001552
1553 return nr;
Oleg Nesterovaceecc042006-06-26 00:26:05 -07001554}
1555
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001556static inline int zap_threads(struct task_struct *tsk, struct mm_struct *mm,
Oleg Nesterov8cd9c242008-07-25 01:47:42 -07001557 struct core_state *core_state, int exit_code)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001558{
1559 struct task_struct *g, *p;
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001560 unsigned long flags;
Oleg Nesterov8cd9c242008-07-25 01:47:42 -07001561 int nr = -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001562
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001563 spin_lock_irq(&tsk->sighand->siglock);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -08001564 if (!signal_group_exit(tsk->signal)) {
Oleg Nesterov8cd9c242008-07-25 01:47:42 -07001565 mm->core_state = core_state;
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001566 tsk->signal->group_exit_code = exit_code;
Oleg Nesterov8cd9c242008-07-25 01:47:42 -07001567 nr = zap_process(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001568 }
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001569 spin_unlock_irq(&tsk->sighand->siglock);
Oleg Nesterov8cd9c242008-07-25 01:47:42 -07001570 if (unlikely(nr < 0))
1571 return nr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001572
Oleg Nesterov8cd9c242008-07-25 01:47:42 -07001573 if (atomic_read(&mm->mm_users) == nr + 1)
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001574 goto done;
Oleg Nesterove4901f92008-07-25 01:47:31 -07001575 /*
1576 * We should find and kill all tasks which use this mm, and we should
Oleg Nesterov999d9fc2008-07-25 01:47:41 -07001577 * count them correctly into ->nr_threads. We don't take tasklist
Oleg Nesterove4901f92008-07-25 01:47:31 -07001578 * lock, but this is safe wrt:
1579 *
1580 * fork:
1581 * None of sub-threads can fork after zap_process(leader). All
1582 * processes which were created before this point should be
1583 * visible to zap_threads() because copy_process() adds the new
1584 * process to the tail of init_task.tasks list, and lock/unlock
1585 * of ->siglock provides a memory barrier.
1586 *
1587 * do_exit:
1588 * The caller holds mm->mmap_sem. This means that the task which
1589 * uses this mm can't pass exit_mm(), so it can't exit or clear
1590 * its ->mm.
1591 *
1592 * de_thread:
1593 * It does list_replace_rcu(&leader->tasks, &current->tasks),
1594 * we must see either old or new leader, this does not matter.
1595 * However, it can change p->sighand, so lock_task_sighand(p)
1596 * must be used. Since p->mm != NULL and we hold ->mmap_sem
1597 * it can't fail.
1598 *
1599 * Note also that "g" can be the old leader with ->mm == NULL
1600 * and already unhashed and thus removed from ->thread_group.
1601 * This is OK, __unhash_process()->list_del_rcu() does not
1602 * clear the ->next pointer, we will find the new leader via
1603 * next_thread().
1604 */
Oleg Nesterov7b1c6152006-06-26 00:26:08 -07001605 rcu_read_lock();
Oleg Nesterovaceecc042006-06-26 00:26:05 -07001606 for_each_process(g) {
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001607 if (g == tsk->group_leader)
1608 continue;
Oleg Nesterov15b9f362008-07-25 01:47:39 -07001609 if (g->flags & PF_KTHREAD)
1610 continue;
Oleg Nesterovaceecc042006-06-26 00:26:05 -07001611 p = g;
1612 do {
1613 if (p->mm) {
Oleg Nesterov15b9f362008-07-25 01:47:39 -07001614 if (unlikely(p->mm == mm)) {
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001615 lock_task_sighand(p, &flags);
Oleg Nesterov8cd9c242008-07-25 01:47:42 -07001616 nr += zap_process(p);
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001617 unlock_task_sighand(p, &flags);
1618 }
Oleg Nesterovaceecc042006-06-26 00:26:05 -07001619 break;
1620 }
Oleg Nesterove4901f92008-07-25 01:47:31 -07001621 } while_each_thread(g, p);
Oleg Nesterovaceecc042006-06-26 00:26:05 -07001622 }
Oleg Nesterov7b1c6152006-06-26 00:26:08 -07001623 rcu_read_unlock();
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001624done:
Oleg Nesterovc5f1cc82008-07-25 01:47:42 -07001625 atomic_set(&core_state->nr_threads, nr);
Oleg Nesterov8cd9c242008-07-25 01:47:42 -07001626 return nr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001627}
1628
Oleg Nesterov9d5b3272008-07-25 01:47:43 -07001629static int coredump_wait(int exit_code, struct core_state *core_state)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001630{
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001631 struct task_struct *tsk = current;
1632 struct mm_struct *mm = tsk->mm;
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001633 struct completion *vfork_done;
Oleg Nesterov2384f552005-10-30 15:02:47 -08001634 int core_waiters;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001635
Oleg Nesterov9d5b3272008-07-25 01:47:43 -07001636 init_completion(&core_state->startup);
Oleg Nesterovb564daf2008-07-25 01:47:44 -07001637 core_state->dumper.task = tsk;
1638 core_state->dumper.next = NULL;
Oleg Nesterov9d5b3272008-07-25 01:47:43 -07001639 core_waiters = zap_threads(tsk, mm, core_state, exit_code);
Oleg Nesterov2384f552005-10-30 15:02:47 -08001640 up_write(&mm->mmap_sem);
1641
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001642 if (unlikely(core_waiters < 0))
1643 goto fail;
1644
1645 /*
1646 * Make sure nobody is waiting for us to release the VM,
1647 * otherwise we can deadlock when we wait on each other
1648 */
1649 vfork_done = tsk->vfork_done;
1650 if (vfork_done) {
1651 tsk->vfork_done = NULL;
1652 complete(vfork_done);
1653 }
1654
Oleg Nesterov2384f552005-10-30 15:02:47 -08001655 if (core_waiters)
Oleg Nesterov9d5b3272008-07-25 01:47:43 -07001656 wait_for_completion(&core_state->startup);
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001657fail:
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001658 return core_waiters;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001659}
1660
Oleg Nesterova94e2d42008-07-25 01:47:46 -07001661static void coredump_finish(struct mm_struct *mm)
1662{
1663 struct core_thread *curr, *next;
1664 struct task_struct *task;
1665
1666 next = mm->core_state->dumper.next;
1667 while ((curr = next) != NULL) {
1668 next = curr->next;
1669 task = curr->task;
1670 /*
1671 * see exit_mm(), curr->task must not see
1672 * ->task == NULL before we read ->next.
1673 */
1674 smp_mb();
1675 curr->task = NULL;
1676 wake_up_process(task);
1677 }
1678
1679 mm->core_state = NULL;
1680}
1681
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001682/*
1683 * set_dumpable converts traditional three-value dumpable to two flags and
1684 * stores them into mm->flags. It modifies lower two bits of mm->flags, but
1685 * these bits are not changed atomically. So get_dumpable can observe the
1686 * intermediate state. To avoid doing unexpected behavior, get get_dumpable
1687 * return either old dumpable or new one by paying attention to the order of
1688 * modifying the bits.
1689 *
1690 * dumpable | mm->flags (binary)
1691 * old new | initial interim final
1692 * ---------+-----------------------
1693 * 0 1 | 00 01 01
1694 * 0 2 | 00 10(*) 11
1695 * 1 0 | 01 00 00
1696 * 1 2 | 01 11 11
1697 * 2 0 | 11 10(*) 00
1698 * 2 1 | 11 11 01
1699 *
1700 * (*) get_dumpable regards interim value of 10 as 11.
1701 */
1702void set_dumpable(struct mm_struct *mm, int value)
1703{
1704 switch (value) {
1705 case 0:
1706 clear_bit(MMF_DUMPABLE, &mm->flags);
1707 smp_wmb();
1708 clear_bit(MMF_DUMP_SECURELY, &mm->flags);
1709 break;
1710 case 1:
1711 set_bit(MMF_DUMPABLE, &mm->flags);
1712 smp_wmb();
1713 clear_bit(MMF_DUMP_SECURELY, &mm->flags);
1714 break;
1715 case 2:
1716 set_bit(MMF_DUMP_SECURELY, &mm->flags);
1717 smp_wmb();
1718 set_bit(MMF_DUMPABLE, &mm->flags);
1719 break;
1720 }
1721}
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001722
1723int get_dumpable(struct mm_struct *mm)
1724{
1725 int ret;
1726
1727 ret = mm->flags & 0x3;
1728 return (ret >= 2) ? 2 : ret;
1729}
1730
WANG Cong8cd3ac32009-01-06 14:42:48 -08001731void do_coredump(long signr, int exit_code, struct pt_regs *regs)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001732{
Oleg Nesterov9d5b3272008-07-25 01:47:43 -07001733 struct core_state core_state;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001734 char corename[CORENAME_MAX_SIZE + 1];
1735 struct mm_struct *mm = current->mm;
1736 struct linux_binfmt * binfmt;
1737 struct inode * inode;
1738 struct file * file;
David Howellsd84f4f92008-11-14 10:39:23 +11001739 const struct cred *old_cred;
1740 struct cred *cred;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001741 int retval = 0;
Alan Coxd6e71142005-06-23 00:09:43 -07001742 int flag = 0;
Andi Kleend025c9d2006-09-30 23:29:28 -07001743 int ispipe = 0;
Neil Horman7dc0b222007-10-16 23:26:34 -07001744 unsigned long core_limit = current->signal->rlim[RLIMIT_CORE].rlim_cur;
Neil Horman74aadce2007-10-16 23:26:35 -07001745 char **helper_argv = NULL;
1746 int helper_argc = 0;
1747 char *delimit;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001748
Steve Grubb0a4ff8c2007-04-19 10:28:21 -04001749 audit_core_dumps(signr);
1750
Linus Torvalds1da177e2005-04-16 15:20:36 -07001751 binfmt = current->binfmt;
1752 if (!binfmt || !binfmt->core_dump)
1753 goto fail;
David Howellsd84f4f92008-11-14 10:39:23 +11001754
1755 cred = prepare_creds();
1756 if (!cred) {
1757 retval = -ENOMEM;
1758 goto fail;
1759 }
1760
Linus Torvalds1da177e2005-04-16 15:20:36 -07001761 down_write(&mm->mmap_sem);
Roland McGrath00ec99d2007-11-11 19:13:43 -08001762 /*
1763 * If another thread got here first, or we are not dumpable, bail out.
1764 */
Oleg Nesterov999d9fc2008-07-25 01:47:41 -07001765 if (mm->core_state || !get_dumpable(mm)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001766 up_write(&mm->mmap_sem);
David Howellsd84f4f92008-11-14 10:39:23 +11001767 put_cred(cred);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001768 goto fail;
1769 }
Alan Coxd6e71142005-06-23 00:09:43 -07001770
1771 /*
1772 * We cannot trust fsuid as being the "true" uid of the
1773 * process nor do we know its entire history. We only know it
1774 * was tainted so we dump it as root in mode 2.
1775 */
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001776 if (get_dumpable(mm) == 2) { /* Setuid core dump mode */
Alan Coxd6e71142005-06-23 00:09:43 -07001777 flag = O_EXCL; /* Stop rewrite attacks */
David Howellsd84f4f92008-11-14 10:39:23 +11001778 cred->fsuid = 0; /* Dump root private */
Alan Coxd6e71142005-06-23 00:09:43 -07001779 }
Oleg Nesterov1291cf42005-10-30 15:02:54 -08001780
Oleg Nesterov9d5b3272008-07-25 01:47:43 -07001781 retval = coredump_wait(exit_code, &core_state);
David Howellsd84f4f92008-11-14 10:39:23 +11001782 if (retval < 0) {
1783 put_cred(cred);
Oleg Nesterov1291cf42005-10-30 15:02:54 -08001784 goto fail;
David Howellsd84f4f92008-11-14 10:39:23 +11001785 }
1786
1787 old_cred = override_creds(cred);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001788
1789 /*
1790 * Clear any false indication of pending signals that might
1791 * be seen by the filesystem code called to write the core file.
1792 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001793 clear_thread_flag(TIF_SIGPENDING);
1794
Linus Torvalds1da177e2005-04-16 15:20:36 -07001795 /*
1796 * lock_kernel() because format_corename() is controlled by sysctl, which
1797 * uses lock_kernel()
1798 */
1799 lock_kernel();
Oleg Nesterov64093242008-10-18 20:28:22 -07001800 ispipe = format_corename(corename, signr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001801 unlock_kernel();
Neil Horman7dc0b222007-10-16 23:26:34 -07001802 /*
1803 * Don't bother to check the RLIMIT_CORE value if core_pattern points
1804 * to a pipe. Since we're not writing directly to the filesystem
1805 * RLIMIT_CORE doesn't really apply, as no actual core file will be
1806 * created unless the pipe reader choses to write out the core file
1807 * at which point file size limits and permissions will be imposed
1808 * as it does with any other process
1809 */
Neil Horman74aadce2007-10-16 23:26:35 -07001810 if ((!ispipe) && (core_limit < binfmt->min_coredump))
Neil Horman7dc0b222007-10-16 23:26:34 -07001811 goto fail_unlock;
1812
Alan Coxc4bbafd2007-04-16 22:53:13 -07001813 if (ispipe) {
Neil Horman74aadce2007-10-16 23:26:35 -07001814 helper_argv = argv_split(GFP_KERNEL, corename+1, &helper_argc);
Tetsuo Handa350eaf72009-01-06 14:41:11 -08001815 if (!helper_argv) {
1816 printk(KERN_WARNING "%s failed to allocate memory\n",
1817 __func__);
1818 goto fail_unlock;
1819 }
Neil Horman74aadce2007-10-16 23:26:35 -07001820 /* Terminate the string before the first option */
1821 delimit = strchr(corename, ' ');
1822 if (delimit)
1823 *delimit = '\0';
Neil Horman32321132007-10-16 23:26:36 -07001824 delimit = strrchr(helper_argv[0], '/');
1825 if (delimit)
1826 delimit++;
1827 else
1828 delimit = helper_argv[0];
1829 if (!strcmp(delimit, current->comm)) {
1830 printk(KERN_NOTICE "Recursive core dump detected, "
1831 "aborting\n");
1832 goto fail_unlock;
1833 }
1834
1835 core_limit = RLIM_INFINITY;
1836
Andi Kleend025c9d2006-09-30 23:29:28 -07001837 /* SIGPIPE can happen, but it's just never processed */
Neil Horman32321132007-10-16 23:26:36 -07001838 if (call_usermodehelper_pipe(corename+1, helper_argv, NULL,
1839 &file)) {
Andi Kleend025c9d2006-09-30 23:29:28 -07001840 printk(KERN_INFO "Core dump to %s pipe failed\n",
1841 corename);
1842 goto fail_unlock;
1843 }
Andi Kleend025c9d2006-09-30 23:29:28 -07001844 } else
1845 file = filp_open(corename,
Alexey Dobriyan6d4df672006-12-06 20:40:39 -08001846 O_CREAT | 2 | O_NOFOLLOW | O_LARGEFILE | flag,
1847 0600);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001848 if (IS_ERR(file))
1849 goto fail_unlock;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001850 inode = file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001851 if (inode->i_nlink > 1)
1852 goto close_fail; /* multiple links - don't dump */
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001853 if (!ispipe && d_unhashed(file->f_path.dentry))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001854 goto close_fail;
1855
Andi Kleend025c9d2006-09-30 23:29:28 -07001856 /* AK: actually i see no reason to not allow this for named pipes etc.,
1857 but keep the previous behaviour for now. */
1858 if (!ispipe && !S_ISREG(inode->i_mode))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001859 goto close_fail;
Ingo Molnarc46f7392007-11-28 13:59:18 +01001860 /*
1861 * Dont allow local users get cute and trick others to coredump
1862 * into their pre-created files:
1863 */
David Howellsda9592e2008-11-14 10:39:05 +11001864 if (inode->i_uid != current_fsuid())
Ingo Molnarc46f7392007-11-28 13:59:18 +01001865 goto close_fail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001866 if (!file->f_op)
1867 goto close_fail;
1868 if (!file->f_op->write)
1869 goto close_fail;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001870 if (!ispipe && do_truncate(file->f_path.dentry, 0, 0, file) != 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001871 goto close_fail;
1872
Neil Horman7dc0b222007-10-16 23:26:34 -07001873 retval = binfmt->core_dump(signr, regs, file, core_limit);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001874
1875 if (retval)
1876 current->signal->group_exit_code |= 0x80;
1877close_fail:
1878 filp_close(file, NULL);
1879fail_unlock:
Neil Horman74aadce2007-10-16 23:26:35 -07001880 if (helper_argv)
1881 argv_free(helper_argv);
1882
David Howellsd84f4f92008-11-14 10:39:23 +11001883 revert_creds(old_cred);
1884 put_cred(cred);
Oleg Nesterova94e2d42008-07-25 01:47:46 -07001885 coredump_finish(mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001886fail:
WANG Cong8cd3ac32009-01-06 14:42:48 -08001887 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001888}