blob: b1fd2025e59a1e0ea0a2f475b812987dac176e9f [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>
Hugh Dickinsba92a432008-07-25 01:45:43 -070031#include <linux/swap.h>
Neil Horman74aadce2007-10-16 23:26:35 -070032#include <linux/string.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070033#include <linux/init.h>
Hugh Dickinsca5b1722008-07-28 15:46:18 -070034#include <linux/pagemap.h>
Ingo Molnarcdd6c482009-09-21 12:02:48 +020035#include <linux/perf_event.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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070045#include <linux/mount.h>
46#include <linux/security.h>
47#include <linux/syscalls.h>
Jay Lan8f0ab512006-09-30 23:28:59 -070048#include <linux/tsacct_kern.h>
Matt Helsley9f460802005-11-07 00:59:16 -080049#include <linux/cn_proc.h>
Al Viro473ae302006-04-26 14:04:08 -040050#include <linux/audit.h>
Roland McGrath6341c392008-07-25 19:45:44 -070051#include <linux/tracehook.h>
Johannes Berg5f4123b2008-07-09 10:28:40 +020052#include <linux/kmod.h>
Eric Paris6110e3a2008-12-17 13:53:20 -050053#include <linux/fsnotify.h>
Al Viro5ad4e532009-03-29 19:50:06 -040054#include <linux/fs_struct.h>
Neil Horman61be2282009-09-23 15:56:58 -070055#include <linux/pipe_fs_i.h>
Ying Han3d5992d2010-10-26 14:21:23 -070056#include <linux/oom.h>
Oleg Nesterov0e028462011-03-06 18:02:54 +010057#include <linux/compat.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 Howells96f951e2012-03-28 18:30:03 +010062#include <asm/exec.h>
KAMEZAWA Hiroyuki43d2b112012-01-10 15:08:09 -080063
64#include <trace/events/task.h>
David Howellsa6f76f22008-11-14 10:39:24 +110065#include "internal.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070066
David Smith4ff16c22012-02-07 10:11:05 -060067#include <trace/events/sched.h>
68
Linus Torvalds1da177e2005-04-16 15:20:36 -070069int core_uses_pid;
Dan Aloni71ce92f2007-05-16 22:11:16 -070070char core_pattern[CORENAME_MAX_SIZE] = "core";
Neil Hormana2939802009-09-23 15:56:56 -070071unsigned int core_pipe_limit;
Alan Coxd6e71142005-06-23 00:09:43 -070072int suid_dumpable = 0;
73
Xiaotian Feng1b0d3002010-10-27 15:34:08 -070074struct core_name {
75 char *corename;
76 int used, size;
77};
78static atomic_t call_count = ATOMIC_INIT(1);
79
Linus Torvalds1da177e2005-04-16 15:20:36 -070080/* The maximal length of core_pattern is also specified in sysctl.c */
81
Alexey Dobriyane4dc1b12007-10-16 23:26:03 -070082static LIST_HEAD(formats);
Linus Torvalds1da177e2005-04-16 15:20:36 -070083static DEFINE_RWLOCK(binfmt_lock);
84
Al Viro8fc3dc52012-03-17 03:05:16 -040085void __register_binfmt(struct linux_binfmt * fmt, int insert)
Linus Torvalds1da177e2005-04-16 15:20:36 -070086{
Al Viro8fc3dc52012-03-17 03:05:16 -040087 BUG_ON(!fmt);
Linus Torvalds1da177e2005-04-16 15:20:36 -070088 write_lock(&binfmt_lock);
Ivan Kokshaysky74641f52009-04-30 15:08:49 -070089 insert ? list_add(&fmt->lh, &formats) :
90 list_add_tail(&fmt->lh, &formats);
Linus Torvalds1da177e2005-04-16 15:20:36 -070091 write_unlock(&binfmt_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -070092}
93
Ivan Kokshaysky74641f52009-04-30 15:08:49 -070094EXPORT_SYMBOL(__register_binfmt);
Linus Torvalds1da177e2005-04-16 15:20:36 -070095
Alexey Dobriyanf6b450d2007-10-16 23:26:04 -070096void unregister_binfmt(struct linux_binfmt * fmt)
Linus Torvalds1da177e2005-04-16 15:20:36 -070097{
Linus Torvalds1da177e2005-04-16 15:20:36 -070098 write_lock(&binfmt_lock);
Alexey Dobriyane4dc1b12007-10-16 23:26:03 -070099 list_del(&fmt->lh);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100 write_unlock(&binfmt_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101}
102
103EXPORT_SYMBOL(unregister_binfmt);
104
105static inline void put_binfmt(struct linux_binfmt * fmt)
106{
107 module_put(fmt->module);
108}
109
110/*
111 * Note that a shared library must be both readable and executable due to
112 * security reasons.
113 *
114 * Also note that we take the address to load from from the file itself.
115 */
Heiko Carstens1e7bfb22009-01-14 14:14:29 +0100116SYSCALL_DEFINE1(uselib, const char __user *, library)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117{
Al Viro964bd182008-07-26 03:33:14 -0400118 struct file *file;
Al Viro964bd182008-07-26 03:33:14 -0400119 char *tmp = getname(library);
120 int error = PTR_ERR(tmp);
Al Viro47c805d2011-02-23 17:44:09 -0500121 static const struct open_flags uselib_flags = {
122 .open_flag = O_LARGEFILE | O_RDONLY | __FMODE_EXEC,
123 .acc_mode = MAY_READ | MAY_EXEC | MAY_OPEN,
124 .intent = LOOKUP_OPEN
125 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700126
Al Viro6e8341a2009-04-06 11:16:22 -0400127 if (IS_ERR(tmp))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700128 goto out;
129
Al Viro47c805d2011-02-23 17:44:09 -0500130 file = do_filp_open(AT_FDCWD, tmp, &uselib_flags, LOOKUP_FOLLOW);
Al Viro6e8341a2009-04-06 11:16:22 -0400131 putname(tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700132 error = PTR_ERR(file);
133 if (IS_ERR(file))
134 goto out;
135
Al Viro6e8341a2009-04-06 11:16:22 -0400136 error = -EINVAL;
137 if (!S_ISREG(file->f_path.dentry->d_inode->i_mode))
138 goto exit;
139
140 error = -EACCES;
141 if (file->f_path.mnt->mnt_flags & MNT_NOEXEC)
142 goto exit;
143
Eric Paris2a12a9d2009-12-17 21:24:21 -0500144 fsnotify_open(file);
Eric Paris6110e3a2008-12-17 13:53:20 -0500145
Linus Torvalds1da177e2005-04-16 15:20:36 -0700146 error = -ENOEXEC;
147 if(file->f_op) {
148 struct linux_binfmt * fmt;
149
150 read_lock(&binfmt_lock);
Alexey Dobriyane4dc1b12007-10-16 23:26:03 -0700151 list_for_each_entry(fmt, &formats, lh) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152 if (!fmt->load_shlib)
153 continue;
154 if (!try_module_get(fmt->module))
155 continue;
156 read_unlock(&binfmt_lock);
157 error = fmt->load_shlib(file);
158 read_lock(&binfmt_lock);
159 put_binfmt(fmt);
160 if (error != -ENOEXEC)
161 break;
162 }
163 read_unlock(&binfmt_lock);
164 }
Al Viro6e8341a2009-04-06 11:16:22 -0400165exit:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166 fput(file);
167out:
168 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700169}
170
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700171#ifdef CONFIG_MMU
Oleg Nesterovae6b5852011-03-06 18:03:11 +0100172/*
173 * The nascent bprm->mm is not visible until exec_mmap() but it can
174 * use a lot of memory, account these pages in current->mm temporary
175 * for oom_badness()->get_mm_rss(). Once exec succeeds or fails, we
176 * change the counter back via acct_arg_size(0).
177 */
Oleg Nesterov0e028462011-03-06 18:02:54 +0100178static void acct_arg_size(struct linux_binprm *bprm, unsigned long pages)
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100179{
180 struct mm_struct *mm = current->mm;
181 long diff = (long)(pages - bprm->vma_pages);
182
183 if (!mm || !diff)
184 return;
185
186 bprm->vma_pages = pages;
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100187 add_mm_counter(mm, MM_ANONPAGES, diff);
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100188}
189
Oleg Nesterov0e028462011-03-06 18:02:54 +0100190static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos,
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700191 int write)
192{
193 struct page *page;
194 int ret;
195
196#ifdef CONFIG_STACK_GROWSUP
197 if (write) {
Michal Hockod05f3162011-05-24 17:11:44 -0700198 ret = expand_downwards(bprm->vma, pos);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700199 if (ret < 0)
200 return NULL;
201 }
202#endif
203 ret = get_user_pages(current, bprm->mm, pos,
204 1, write, 1, &page, NULL);
205 if (ret <= 0)
206 return NULL;
207
208 if (write) {
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700209 unsigned long size = bprm->vma->vm_end - bprm->vma->vm_start;
Linus Torvaldsa64e7152008-03-03 10:12:14 -0800210 struct rlimit *rlim;
211
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100212 acct_arg_size(bprm, size / PAGE_SIZE);
213
Linus Torvaldsa64e7152008-03-03 10:12:14 -0800214 /*
215 * We've historically supported up to 32 pages (ARG_MAX)
216 * of argument strings even with small stacks
217 */
218 if (size <= ARG_MAX)
219 return page;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700220
221 /*
222 * Limit to 1/4-th the stack size for the argv+env strings.
223 * This ensures that:
224 * - the remaining binfmt code will not run out of stack space,
225 * - the program will have a reasonable amount of stack left
226 * to work from.
227 */
Linus Torvaldsa64e7152008-03-03 10:12:14 -0800228 rlim = current->signal->rlim;
Jiri Slabyd554ed892010-03-05 13:42:42 -0800229 if (size > ACCESS_ONCE(rlim[RLIMIT_STACK].rlim_cur) / 4) {
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700230 put_page(page);
231 return NULL;
232 }
233 }
234
235 return page;
236}
237
238static void put_arg_page(struct page *page)
239{
240 put_page(page);
241}
242
243static void free_arg_page(struct linux_binprm *bprm, int i)
244{
245}
246
247static void free_arg_pages(struct linux_binprm *bprm)
248{
249}
250
251static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos,
252 struct page *page)
253{
254 flush_cache_page(bprm->vma, pos, page_to_pfn(page));
255}
256
257static int __bprm_mm_init(struct linux_binprm *bprm)
258{
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800259 int err;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700260 struct vm_area_struct *vma = NULL;
261 struct mm_struct *mm = bprm->mm;
262
263 bprm->vma = vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL);
264 if (!vma)
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800265 return -ENOMEM;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700266
267 down_write(&mm->mmap_sem);
268 vma->vm_mm = mm;
269
270 /*
271 * Place the stack at the largest stack address the architecture
272 * supports. Later, we'll move this to an appropriate place. We don't
273 * use STACK_TOP because that can depend on attributes which aren't
274 * configured yet.
275 */
Michal Hockoaacb3d12011-07-26 16:08:40 -0700276 BUILD_BUG_ON(VM_STACK_FLAGS & VM_STACK_INCOMPLETE_SETUP);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700277 vma->vm_end = STACK_TOP_MAX;
278 vma->vm_start = vma->vm_end - PAGE_SIZE;
Mel Gormana8bef8ff2010-05-24 14:32:24 -0700279 vma->vm_flags = VM_STACK_FLAGS | VM_STACK_INCOMPLETE_SETUP;
Coly Li3ed75eb2007-10-18 23:39:15 -0700280 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
Rik van Riel5beb4932010-03-05 13:42:07 -0800281 INIT_LIST_HEAD(&vma->anon_vma_chain);
Tavis Ormandy462e635e2010-12-09 15:29:42 +0100282
283 err = security_file_mmap(NULL, 0, 0, 0, vma->vm_start, 1);
284 if (err)
285 goto err;
286
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700287 err = insert_vm_struct(mm, vma);
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800288 if (err)
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700289 goto err;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700290
291 mm->stack_vm = mm->total_vm = 1;
292 up_write(&mm->mmap_sem);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700293 bprm->p = vma->vm_end - sizeof(void *);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700294 return 0;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700295err:
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800296 up_write(&mm->mmap_sem);
297 bprm->vma = NULL;
298 kmem_cache_free(vm_area_cachep, vma);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700299 return err;
300}
301
302static bool valid_arg_len(struct linux_binprm *bprm, long len)
303{
304 return len <= MAX_ARG_STRLEN;
305}
306
307#else
308
Oleg Nesterov0e028462011-03-06 18:02:54 +0100309static inline void acct_arg_size(struct linux_binprm *bprm, unsigned long pages)
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100310{
311}
312
Oleg Nesterov0e028462011-03-06 18:02:54 +0100313static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos,
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700314 int write)
315{
316 struct page *page;
317
318 page = bprm->page[pos / PAGE_SIZE];
319 if (!page && write) {
320 page = alloc_page(GFP_HIGHUSER|__GFP_ZERO);
321 if (!page)
322 return NULL;
323 bprm->page[pos / PAGE_SIZE] = page;
324 }
325
326 return page;
327}
328
329static void put_arg_page(struct page *page)
330{
331}
332
333static void free_arg_page(struct linux_binprm *bprm, int i)
334{
335 if (bprm->page[i]) {
336 __free_page(bprm->page[i]);
337 bprm->page[i] = NULL;
338 }
339}
340
341static void free_arg_pages(struct linux_binprm *bprm)
342{
343 int i;
344
345 for (i = 0; i < MAX_ARG_PAGES; i++)
346 free_arg_page(bprm, i);
347}
348
349static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos,
350 struct page *page)
351{
352}
353
354static int __bprm_mm_init(struct linux_binprm *bprm)
355{
356 bprm->p = PAGE_SIZE * MAX_ARG_PAGES - sizeof(void *);
357 return 0;
358}
359
360static bool valid_arg_len(struct linux_binprm *bprm, long len)
361{
362 return len <= bprm->p;
363}
364
365#endif /* CONFIG_MMU */
366
367/*
368 * Create a new mm_struct and populate it with a temporary stack
369 * vm_area_struct. We don't have enough context at this point to set the stack
370 * flags, permissions, and offset, so we use temporary values. We'll update
371 * them later in setup_arg_pages().
372 */
373int bprm_mm_init(struct linux_binprm *bprm)
374{
375 int err;
376 struct mm_struct *mm = NULL;
377
378 bprm->mm = mm = mm_alloc();
379 err = -ENOMEM;
380 if (!mm)
381 goto err;
382
383 err = init_new_context(current, mm);
384 if (err)
385 goto err;
386
387 err = __bprm_mm_init(bprm);
388 if (err)
389 goto err;
390
391 return 0;
392
393err:
394 if (mm) {
395 bprm->mm = NULL;
396 mmdrop(mm);
397 }
398
399 return err;
400}
401
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100402struct user_arg_ptr {
Oleg Nesterov0e028462011-03-06 18:02:54 +0100403#ifdef CONFIG_COMPAT
404 bool is_compat;
405#endif
406 union {
407 const char __user *const __user *native;
408#ifdef CONFIG_COMPAT
409 compat_uptr_t __user *compat;
410#endif
411 } ptr;
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100412};
413
414static const char __user *get_user_arg_ptr(struct user_arg_ptr argv, int nr)
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100415{
Oleg Nesterov0e028462011-03-06 18:02:54 +0100416 const char __user *native;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100417
Oleg Nesterov0e028462011-03-06 18:02:54 +0100418#ifdef CONFIG_COMPAT
419 if (unlikely(argv.is_compat)) {
420 compat_uptr_t compat;
421
422 if (get_user(compat, argv.ptr.compat + nr))
423 return ERR_PTR(-EFAULT);
424
425 return compat_ptr(compat);
426 }
427#endif
428
429 if (get_user(native, argv.ptr.native + nr))
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100430 return ERR_PTR(-EFAULT);
431
Oleg Nesterov0e028462011-03-06 18:02:54 +0100432 return native;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100433}
434
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435/*
436 * count() counts the number of strings in array ARGV.
437 */
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100438static int count(struct user_arg_ptr argv, int max)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700439{
440 int i = 0;
441
Oleg Nesterov0e028462011-03-06 18:02:54 +0100442 if (argv.ptr.native != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443 for (;;) {
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100444 const char __user *p = get_user_arg_ptr(argv, i);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445
Linus Torvalds1da177e2005-04-16 15:20:36 -0700446 if (!p)
447 break;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100448
449 if (IS_ERR(p))
450 return -EFAULT;
451
Jason Baron362e6662008-10-15 22:01:52 -0700452 if (i++ >= max)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453 return -E2BIG;
Roland McGrath9aea5a62010-09-07 19:37:06 -0700454
455 if (fatal_signal_pending(current))
456 return -ERESTARTNOHAND;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457 cond_resched();
458 }
459 }
460 return i;
461}
462
463/*
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700464 * 'copy_strings()' copies argument/environment strings from the old
465 * processes's memory to the new process's stack. The call to get_user_pages()
466 * ensures the destination page is created and not swapped out.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467 */
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100468static int copy_strings(int argc, struct user_arg_ptr argv,
Adrian Bunk75c96f82005-05-05 16:16:09 -0700469 struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470{
471 struct page *kmapped_page = NULL;
472 char *kaddr = NULL;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700473 unsigned long kpos = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474 int ret;
475
476 while (argc-- > 0) {
David Howellsd7627462010-08-17 23:52:56 +0100477 const char __user *str;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700478 int len;
479 unsigned long pos;
480
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100481 ret = -EFAULT;
482 str = get_user_arg_ptr(argv, argc);
483 if (IS_ERR(str))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100486 len = strnlen_user(str, MAX_ARG_STRLEN);
487 if (!len)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488 goto out;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100489
490 ret = -E2BIG;
491 if (!valid_arg_len(bprm, len))
492 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700494 /* We're going to work our way backwords. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495 pos = bprm->p;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700496 str += len;
497 bprm->p -= len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498
499 while (len > 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700500 int offset, bytes_to_copy;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501
Roland McGrath9aea5a62010-09-07 19:37:06 -0700502 if (fatal_signal_pending(current)) {
503 ret = -ERESTARTNOHAND;
504 goto out;
505 }
Roland McGrath7993bc12010-09-07 19:36:28 -0700506 cond_resched();
507
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508 offset = pos % PAGE_SIZE;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700509 if (offset == 0)
510 offset = PAGE_SIZE;
511
512 bytes_to_copy = offset;
513 if (bytes_to_copy > len)
514 bytes_to_copy = len;
515
516 offset -= bytes_to_copy;
517 pos -= bytes_to_copy;
518 str -= bytes_to_copy;
519 len -= bytes_to_copy;
520
521 if (!kmapped_page || kpos != (pos & PAGE_MASK)) {
522 struct page *page;
523
524 page = get_arg_page(bprm, pos, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525 if (!page) {
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700526 ret = -E2BIG;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527 goto out;
528 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700530 if (kmapped_page) {
531 flush_kernel_dcache_page(kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532 kunmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700533 put_arg_page(kmapped_page);
534 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535 kmapped_page = page;
536 kaddr = kmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700537 kpos = pos & PAGE_MASK;
538 flush_arg_page(bprm, kpos, kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539 }
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700540 if (copy_from_user(kaddr+offset, str, bytes_to_copy)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541 ret = -EFAULT;
542 goto out;
543 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544 }
545 }
546 ret = 0;
547out:
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700548 if (kmapped_page) {
549 flush_kernel_dcache_page(kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550 kunmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700551 put_arg_page(kmapped_page);
552 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553 return ret;
554}
555
556/*
557 * Like copy_strings, but get argv and its values from kernel memory.
558 */
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100559int copy_strings_kernel(int argc, const char *const *__argv,
David Howellsd7627462010-08-17 23:52:56 +0100560 struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700561{
562 int r;
563 mm_segment_t oldfs = get_fs();
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100564 struct user_arg_ptr argv = {
Oleg Nesterov0e028462011-03-06 18:02:54 +0100565 .ptr.native = (const char __user *const __user *)__argv,
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100566 };
567
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568 set_fs(KERNEL_DS);
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100569 r = copy_strings(argc, argv, bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700570 set_fs(oldfs);
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100571
Linus Torvalds1da177e2005-04-16 15:20:36 -0700572 return r;
573}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700574EXPORT_SYMBOL(copy_strings_kernel);
575
576#ifdef CONFIG_MMU
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700577
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578/*
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700579 * During bprm_mm_init(), we create a temporary stack at STACK_TOP_MAX. Once
580 * the binfmt code determines where the new stack should reside, we shift it to
581 * its final location. The process proceeds as follows:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582 *
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700583 * 1) Use shift to calculate the new vma endpoints.
584 * 2) Extend vma to cover both the old and new ranges. This ensures the
585 * arguments passed to subsequent functions are consistent.
586 * 3) Move vma's page tables to the new range.
587 * 4) Free up any cleared pgd range.
588 * 5) Shrink the vma to cover only the new range.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589 */
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700590static int shift_arg_pages(struct vm_area_struct *vma, unsigned long shift)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591{
592 struct mm_struct *mm = vma->vm_mm;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700593 unsigned long old_start = vma->vm_start;
594 unsigned long old_end = vma->vm_end;
595 unsigned long length = old_end - old_start;
596 unsigned long new_start = old_start - shift;
597 unsigned long new_end = old_end - shift;
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700598 struct mmu_gather tlb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700599
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700600 BUG_ON(new_start > new_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700601
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700602 /*
603 * ensure there are no vmas between where we want to go
604 * and where we are
605 */
606 if (vma != find_vma(mm, new_start))
607 return -EFAULT;
608
609 /*
610 * cover the whole range: [new_start, old_end)
611 */
Rik van Riel5beb4932010-03-05 13:42:07 -0800612 if (vma_adjust(vma, new_start, old_end, vma->vm_pgoff, NULL))
613 return -ENOMEM;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700614
615 /*
616 * move the page tables downwards, on failure we rely on
617 * process cleanup to remove whatever mess we made.
618 */
619 if (length != move_page_tables(vma, old_start,
620 vma, new_start, length))
621 return -ENOMEM;
622
623 lru_add_drain();
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700624 tlb_gather_mmu(&tlb, mm, 0);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700625 if (new_end > old_start) {
626 /*
627 * when the old and new regions overlap clear from new_end.
628 */
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700629 free_pgd_range(&tlb, new_end, old_end, new_end,
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700630 vma->vm_next ? vma->vm_next->vm_start : 0);
631 } else {
632 /*
633 * otherwise, clean from old_start; this is done to not touch
634 * the address space in [new_end, old_start) some architectures
635 * have constraints on va-space that make this illegal (IA64) -
636 * for the others its just a little faster.
637 */
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700638 free_pgd_range(&tlb, old_start, old_end, new_end,
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700639 vma->vm_next ? vma->vm_next->vm_start : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700640 }
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700641 tlb_finish_mmu(&tlb, new_end, old_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700642
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700643 /*
Rik van Riel5beb4932010-03-05 13:42:07 -0800644 * Shrink the vma to just the new range. Always succeeds.
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700645 */
646 vma_adjust(vma, new_start, new_end, vma->vm_pgoff, NULL);
647
648 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700649}
650
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700651/*
652 * Finalizes the stack vm_area_struct. The flags and permissions are updated,
653 * the stack is optionally relocated, and some extra space is added.
654 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655int setup_arg_pages(struct linux_binprm *bprm,
656 unsigned long stack_top,
657 int executable_stack)
658{
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700659 unsigned long ret;
660 unsigned long stack_shift;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661 struct mm_struct *mm = current->mm;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700662 struct vm_area_struct *vma = bprm->vma;
663 struct vm_area_struct *prev = NULL;
664 unsigned long vm_flags;
665 unsigned long stack_base;
Michael Neuling803bf5e2010-02-10 13:56:42 -0800666 unsigned long stack_size;
667 unsigned long stack_expand;
668 unsigned long rlim_stack;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700669
670#ifdef CONFIG_STACK_GROWSUP
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671 /* Limit stack size to 1GB */
Jiri Slabyd554ed892010-03-05 13:42:42 -0800672 stack_base = rlimit_max(RLIMIT_STACK);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700673 if (stack_base > (1 << 30))
674 stack_base = 1 << 30;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700675
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700676 /* Make sure we didn't let the argument array grow too large. */
677 if (vma->vm_end - vma->vm_start > stack_base)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700678 return -ENOMEM;
679
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700680 stack_base = PAGE_ALIGN(stack_top - stack_base);
681
682 stack_shift = vma->vm_start - stack_base;
683 mm->arg_start = bprm->p - stack_shift;
684 bprm->p = vma->vm_end - stack_shift;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700685#else
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700686 stack_top = arch_align_stack(stack_top);
687 stack_top = PAGE_ALIGN(stack_top);
Roland McGrath1b528182010-09-07 19:35:49 -0700688
689 if (unlikely(stack_top < mmap_min_addr) ||
690 unlikely(vma->vm_end - vma->vm_start >= stack_top - mmap_min_addr))
691 return -ENOMEM;
692
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700693 stack_shift = vma->vm_end - stack_top;
694
695 bprm->p -= stack_shift;
696 mm->arg_start = bprm->p;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700697#endif
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700698
699 if (bprm->loader)
700 bprm->loader -= stack_shift;
701 bprm->exec -= stack_shift;
702
703 down_write(&mm->mmap_sem);
Hugh Dickins96a8e132008-07-10 21:19:20 +0100704 vm_flags = VM_STACK_FLAGS;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700705
706 /*
707 * Adjust stack execute permissions; explicitly enable for
708 * EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X and leave alone
709 * (arch default) otherwise.
710 */
711 if (unlikely(executable_stack == EXSTACK_ENABLE_X))
712 vm_flags |= VM_EXEC;
713 else if (executable_stack == EXSTACK_DISABLE_X)
714 vm_flags &= ~VM_EXEC;
715 vm_flags |= mm->def_flags;
Mel Gormana8bef8ff2010-05-24 14:32:24 -0700716 vm_flags |= VM_STACK_INCOMPLETE_SETUP;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700717
718 ret = mprotect_fixup(vma, &prev, vma->vm_start, vma->vm_end,
719 vm_flags);
720 if (ret)
721 goto out_unlock;
722 BUG_ON(prev != vma);
723
724 /* Move stack pages down in memory. */
725 if (stack_shift) {
726 ret = shift_arg_pages(vma, stack_shift);
Anton Blanchardfc63cf22009-11-11 14:26:48 -0800727 if (ret)
728 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729 }
730
Mel Gormana8bef8ff2010-05-24 14:32:24 -0700731 /* mprotect_fixup is overkill to remove the temporary stack flags */
732 vma->vm_flags &= ~VM_STACK_INCOMPLETE_SETUP;
733
Michael Neuling5ef097d2010-03-05 13:42:57 -0800734 stack_expand = 131072UL; /* randomly 32*4k (or 2*64k) pages */
Michael Neuling803bf5e2010-02-10 13:56:42 -0800735 stack_size = vma->vm_end - vma->vm_start;
736 /*
737 * Align this down to a page boundary as expand_stack
738 * will align it up.
739 */
740 rlim_stack = rlimit(RLIMIT_STACK) & PAGE_MASK;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700741#ifdef CONFIG_STACK_GROWSUP
Michael Neuling803bf5e2010-02-10 13:56:42 -0800742 if (stack_size + stack_expand > rlim_stack)
743 stack_base = vma->vm_start + rlim_stack;
744 else
745 stack_base = vma->vm_end + stack_expand;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700746#else
Michael Neuling803bf5e2010-02-10 13:56:42 -0800747 if (stack_size + stack_expand > rlim_stack)
748 stack_base = vma->vm_end - rlim_stack;
749 else
750 stack_base = vma->vm_start - stack_expand;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700751#endif
Eric B Munson3af9e852010-05-18 15:30:49 +0100752 current->mm->start_stack = bprm->p;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700753 ret = expand_stack(vma, stack_base);
754 if (ret)
755 ret = -EFAULT;
756
757out_unlock:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758 up_write(&mm->mmap_sem);
Anton Blanchardfc63cf22009-11-11 14:26:48 -0800759 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700760}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761EXPORT_SYMBOL(setup_arg_pages);
762
Linus Torvalds1da177e2005-04-16 15:20:36 -0700763#endif /* CONFIG_MMU */
764
765struct file *open_exec(const char *name)
766{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767 struct file *file;
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200768 int err;
Al Viro47c805d2011-02-23 17:44:09 -0500769 static const struct open_flags open_exec_flags = {
770 .open_flag = O_LARGEFILE | O_RDONLY | __FMODE_EXEC,
771 .acc_mode = MAY_EXEC | MAY_OPEN,
772 .intent = LOOKUP_OPEN
773 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774
Al Viro47c805d2011-02-23 17:44:09 -0500775 file = do_filp_open(AT_FDCWD, name, &open_exec_flags, LOOKUP_FOLLOW);
Al Viro6e8341a2009-04-06 11:16:22 -0400776 if (IS_ERR(file))
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200777 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200779 err = -EACCES;
Al Viro6e8341a2009-04-06 11:16:22 -0400780 if (!S_ISREG(file->f_path.dentry->d_inode->i_mode))
781 goto exit;
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200782
Al Viro6e8341a2009-04-06 11:16:22 -0400783 if (file->f_path.mnt->mnt_flags & MNT_NOEXEC)
784 goto exit;
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200785
Eric Paris2a12a9d2009-12-17 21:24:21 -0500786 fsnotify_open(file);
Eric Paris6110e3a2008-12-17 13:53:20 -0500787
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200788 err = deny_write_access(file);
Al Viro6e8341a2009-04-06 11:16:22 -0400789 if (err)
790 goto exit;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700791
Al Viro6e8341a2009-04-06 11:16:22 -0400792out:
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200793 return file;
794
Al Viro6e8341a2009-04-06 11:16:22 -0400795exit:
796 fput(file);
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200797 return ERR_PTR(err);
798}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799EXPORT_SYMBOL(open_exec);
800
Mimi Zohar6777d772009-08-21 14:32:48 -0400801int kernel_read(struct file *file, loff_t offset,
802 char *addr, unsigned long count)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803{
804 mm_segment_t old_fs;
805 loff_t pos = offset;
806 int result;
807
808 old_fs = get_fs();
809 set_fs(get_ds());
810 /* The cast to a user pointer is valid due to the set_fs() */
811 result = vfs_read(file, (void __user *)addr, count, &pos);
812 set_fs(old_fs);
813 return result;
814}
815
816EXPORT_SYMBOL(kernel_read);
817
818static int exec_mmap(struct mm_struct *mm)
819{
820 struct task_struct *tsk;
821 struct mm_struct * old_mm, *active_mm;
822
823 /* Notify parent that we're no longer interested in the old VM */
824 tsk = current;
825 old_mm = current->mm;
David Rientjes05af2e12012-03-21 16:34:13 -0700826 sync_mm_rss(old_mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827 mm_release(tsk, old_mm);
828
829 if (old_mm) {
830 /*
831 * Make sure that if there is a core dump in progress
832 * for the old mm, we get out and die instead of going
833 * through with the exec. We must hold mmap_sem around
Oleg Nesterov999d9fc2008-07-25 01:47:41 -0700834 * checking core_state and changing tsk->mm.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835 */
836 down_read(&old_mm->mmap_sem);
Oleg Nesterov999d9fc2008-07-25 01:47:41 -0700837 if (unlikely(old_mm->core_state)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838 up_read(&old_mm->mmap_sem);
839 return -EINTR;
840 }
841 }
842 task_lock(tsk);
843 active_mm = tsk->active_mm;
844 tsk->mm = mm;
845 tsk->active_mm = mm;
846 activate_mm(active_mm, mm);
847 task_unlock(tsk);
848 arch_pick_mmap_layout(mm);
849 if (old_mm) {
850 up_read(&old_mm->mmap_sem);
Eric Sesterhenn7dddb122006-04-01 01:13:38 +0200851 BUG_ON(active_mm != old_mm);
Oleg Nesterov701085b2012-03-19 17:04:01 +0100852 setmax_mm_hiwater_rss(&tsk->signal->maxrss, old_mm);
Balbir Singh31a78f22008-09-28 23:09:31 +0100853 mm_update_next_owner(old_mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700854 mmput(old_mm);
855 return 0;
856 }
857 mmdrop(active_mm);
858 return 0;
859}
860
861/*
862 * This function makes sure the current process has its own signal table,
863 * so that flush_signal_handlers can later reset the handlers without
864 * disturbing other processes. (Other processes might share the signal
865 * table via the CLONE_SIGHAND option to clone().)
866 */
Arjan van de Ven858119e2006-01-14 13:20:43 -0800867static int de_thread(struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868{
869 struct signal_struct *sig = tsk->signal;
Oleg Nesterovb2c903b2007-10-16 23:27:22 -0700870 struct sighand_struct *oldsighand = tsk->sighand;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871 spinlock_t *lock = &oldsighand->siglock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700872
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700873 if (thread_group_empty(tsk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874 goto no_thread_group;
875
876 /*
877 * Kill all other threads in the thread group.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700879 spin_lock_irq(lock);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -0800880 if (signal_group_exit(sig)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700881 /*
882 * Another group action in progress, just
883 * return so that the signal is processed.
884 */
885 spin_unlock_irq(lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886 return -EAGAIN;
887 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888
Oleg Nesterovd3441932010-05-26 14:43:11 -0700889 sig->group_exit_task = tsk;
890 sig->notify_count = zap_other_threads(tsk);
891 if (!thread_group_leader(tsk))
892 sig->notify_count--;
893
894 while (sig->notify_count) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895 __set_current_state(TASK_UNINTERRUPTIBLE);
896 spin_unlock_irq(lock);
897 schedule();
898 spin_lock_irq(lock);
899 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700900 spin_unlock_irq(lock);
901
902 /*
903 * At this point all other threads have exited, all we have to
904 * do is to wait for the thread group leader to become inactive,
905 * and to assume its PID:
906 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700907 if (!thread_group_leader(tsk)) {
Oleg Nesterov81879262008-12-01 14:18:16 -0800908 struct task_struct *leader = tsk->group_leader;
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700909
Oleg Nesterov2800d8d2008-04-30 00:53:12 -0700910 sig->notify_count = -1; /* for exit_notify() */
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700911 for (;;) {
912 write_lock_irq(&tasklist_lock);
913 if (likely(leader->exit_state))
914 break;
915 __set_current_state(TASK_UNINTERRUPTIBLE);
916 write_unlock_irq(&tasklist_lock);
917 schedule();
918 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700919
Roland McGrathf5e90282006-04-10 22:54:16 -0700920 /*
921 * The only record we have of the real-time age of a
922 * process, regardless of execs it's done, is start_time.
923 * All the past CPU time is accumulated in signal_struct
924 * from sister threads now dead. But in this non-leader
925 * exec, nothing survives from the original leader thread,
926 * whose birth marks the true age of this process now.
927 * When we take on its identity by switching to its PID, we
928 * also take its birthdate (always earlier than our own).
929 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700930 tsk->start_time = leader->start_time;
Roland McGrathf5e90282006-04-10 22:54:16 -0700931
Pavel Emelyanovbac0abd2007-10-18 23:40:18 -0700932 BUG_ON(!same_thread_group(leader, tsk));
933 BUG_ON(has_group_leader_pid(tsk));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700934 /*
935 * An exec() starts a new thread group with the
936 * TGID of the previous thread group. Rehash the
937 * two threads with a switched PID, and release
938 * the former thread group leader:
939 */
Eric W. Biedermand73d6522006-03-28 16:11:03 -0800940
941 /* Become a process group leader with the old leader's pid.
Eric W. Biedermanc18258c2006-09-27 01:51:06 -0700942 * The old leader becomes a thread of the this thread group.
943 * Note: The old leader also uses this pid until release_task
Eric W. Biedermand73d6522006-03-28 16:11:03 -0800944 * is called. Odd but simple and correct.
945 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700946 detach_pid(tsk, PIDTYPE_PID);
947 tsk->pid = leader->pid;
Sukadev Bhattiprolu3743ca02007-10-18 23:39:51 -0700948 attach_pid(tsk, PIDTYPE_PID, task_pid(leader));
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700949 transfer_pid(leader, tsk, PIDTYPE_PGID);
950 transfer_pid(leader, tsk, PIDTYPE_SID);
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -0800951
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700952 list_replace_rcu(&leader->tasks, &tsk->tasks);
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -0800953 list_replace_init(&leader->sibling, &tsk->sibling);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700954
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700955 tsk->group_leader = tsk;
956 leader->group_leader = tsk;
Eric W. Biedermande12a782006-04-10 17:16:49 -0600957
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700958 tsk->exit_signal = SIGCHLD;
Oleg Nesterov087806b2011-06-22 23:10:26 +0200959 leader->exit_signal = -1;
Oleg Nesterov962b5642005-11-23 13:37:43 -0800960
961 BUG_ON(leader->exit_state != EXIT_ZOMBIE);
962 leader->exit_state = EXIT_DEAD;
Oleg Nesteroveac1b5e2011-07-21 20:00:43 +0200963
964 /*
965 * We are going to release_task()->ptrace_unlink() silently,
966 * the tracer can sleep in do_wait(). EXIT_DEAD guarantees
967 * the tracer wont't block again waiting for this thread.
968 */
969 if (unlikely(leader->ptrace))
970 __wake_up_parent(leader, leader->parent);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700971 write_unlock_irq(&tasklist_lock);
Oleg Nesterov81879262008-12-01 14:18:16 -0800972
973 release_task(leader);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -0800974 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700975
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700976 sig->group_exit_task = NULL;
977 sig->notify_count = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700978
979no_thread_group:
Oleg Nesterove6368252012-03-19 17:03:22 +0100980 /* we have changed execution domain */
981 tsk->exit_signal = SIGCHLD;
982
Linus Torvalds1da177e2005-04-16 15:20:36 -0700983 exit_itimers(sig);
Oleg Nesterovcbaffba2008-05-26 20:55:42 +0400984 flush_itimer_signals();
Oleg Nesterov329f7db2005-11-07 21:12:43 +0300985
Oleg Nesterovb2c903b2007-10-16 23:27:22 -0700986 if (atomic_read(&oldsighand->count) != 1) {
987 struct sighand_struct *newsighand;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700988 /*
Oleg Nesterovb2c903b2007-10-16 23:27:22 -0700989 * This ->sighand is shared with the CLONE_SIGHAND
990 * but not CLONE_THREAD task, switch to the new one.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700991 */
Oleg Nesterovb2c903b2007-10-16 23:27:22 -0700992 newsighand = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
993 if (!newsighand)
994 return -ENOMEM;
995
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996 atomic_set(&newsighand->count, 1);
997 memcpy(newsighand->action, oldsighand->action,
998 sizeof(newsighand->action));
999
1000 write_lock_irq(&tasklist_lock);
1001 spin_lock(&oldsighand->siglock);
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001002 rcu_assign_pointer(tsk->sighand, newsighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003 spin_unlock(&oldsighand->siglock);
1004 write_unlock_irq(&tasklist_lock);
1005
Davide Libenzifba2afa2007-05-10 22:23:13 -07001006 __cleanup_sighand(oldsighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007 }
1008
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001009 BUG_ON(!thread_group_leader(tsk));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010 return 0;
1011}
Oleg Nesterov0840a902007-10-16 23:27:22 -07001012
Linus Torvalds1da177e2005-04-16 15:20:36 -07001013/*
1014 * These functions flushes out all traces of the currently running executable
1015 * so that a new one can be started
1016 */
Arjan van de Ven858119e2006-01-14 13:20:43 -08001017static void flush_old_files(struct files_struct * files)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001018{
1019 long j = -1;
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001020 struct fdtable *fdt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021
1022 spin_lock(&files->file_lock);
1023 for (;;) {
1024 unsigned long set, i;
1025
1026 j++;
1027 i = j * __NFDBITS;
Dipankar Sarmabadf1662005-09-09 13:04:10 -07001028 fdt = files_fdtable(files);
Vadim Lobanovbbea9f62006-12-10 02:21:12 -08001029 if (i >= fdt->max_fds)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001030 break;
David Howells1fd36ad2012-02-16 17:49:54 +00001031 set = fdt->close_on_exec[j];
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032 if (!set)
1033 continue;
David Howells1fd36ad2012-02-16 17:49:54 +00001034 fdt->close_on_exec[j] = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035 spin_unlock(&files->file_lock);
1036 for ( ; set ; i++,set >>= 1) {
1037 if (set & 1) {
1038 sys_close(i);
1039 }
1040 }
1041 spin_lock(&files->file_lock);
1042
1043 }
1044 spin_unlock(&files->file_lock);
1045}
1046
Andrew Morton59714d62008-02-04 22:27:21 -08001047char *get_task_comm(char *buf, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001048{
1049 /* buf must be at least sizeof(tsk->comm) in size */
1050 task_lock(tsk);
1051 strncpy(buf, tsk->comm, sizeof(tsk->comm));
1052 task_unlock(tsk);
Andrew Morton59714d62008-02-04 22:27:21 -08001053 return buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001054}
J Freyensee7d74f492011-05-06 16:56:47 -07001055EXPORT_SYMBOL_GPL(get_task_comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001056
1057void set_task_comm(struct task_struct *tsk, char *buf)
1058{
1059 task_lock(tsk);
john stultz4614a696b2009-12-14 18:00:05 -08001060
KAMEZAWA Hiroyuki43d2b112012-01-10 15:08:09 -08001061 trace_task_rename(tsk, buf);
1062
john stultz4614a696b2009-12-14 18:00:05 -08001063 /*
1064 * Threads may access current->comm without holding
1065 * the task lock, so write the string carefully.
1066 * Readers without a lock may see incomplete new
1067 * names but are safe from non-terminating string reads.
1068 */
1069 memset(tsk->comm, 0, TASK_COMM_LEN);
1070 wmb();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071 strlcpy(tsk->comm, buf, sizeof(tsk->comm));
1072 task_unlock(tsk);
Ingo Molnarcdd6c482009-09-21 12:02:48 +02001073 perf_event_comm(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001074}
1075
Heiko Carstens96e02d12012-02-04 10:47:10 +01001076static void filename_to_taskname(char *tcomm, const char *fn, unsigned int len)
1077{
1078 int i, ch;
1079
1080 /* Copies the binary name from after last slash */
1081 for (i = 0; (ch = *(fn++)) != '\0';) {
1082 if (ch == '/')
1083 i = 0; /* overwrite what we wrote */
1084 else
1085 if (i < len - 1)
1086 tcomm[i++] = ch;
1087 }
1088 tcomm[i] = '\0';
1089}
1090
Linus Torvalds1da177e2005-04-16 15:20:36 -07001091int flush_old_exec(struct linux_binprm * bprm)
1092{
Linus Torvalds221af7f2010-01-28 22:14:42 -08001093 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001094
1095 /*
1096 * Make sure we have a private signal table and that
1097 * we are unassociated from the previous thread group.
1098 */
1099 retval = de_thread(current);
1100 if (retval)
1101 goto out;
1102
Matt Helsley925d1c42008-04-29 01:01:36 -07001103 set_mm_exe_file(bprm->mm, bprm->file);
1104
Heiko Carstens96e02d12012-02-04 10:47:10 +01001105 filename_to_taskname(bprm->tcomm, bprm->filename, sizeof(bprm->tcomm));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001107 * Release all of the old mmap stuff
1108 */
Oleg Nesterov3c77f842010-11-30 20:55:34 +01001109 acct_arg_size(bprm, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001110 retval = exec_mmap(bprm->mm);
1111 if (retval)
Al Virofd8328b2008-04-22 05:11:59 -04001112 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001113
1114 bprm->mm = NULL; /* We're using it now */
Linus Torvalds7ab02af2010-02-02 12:37:44 -08001115
Mathias Krausedac853a2011-06-09 20:05:18 +02001116 set_fs(USER_DS);
Al Viro19e51092012-02-23 22:29:17 -05001117 current->flags &= ~(PF_RANDOMIZE | PF_FORKNOEXEC | PF_KTHREAD);
Linus Torvalds7ab02af2010-02-02 12:37:44 -08001118 flush_thread();
1119 current->personality &= ~bprm->per_clear;
1120
Linus Torvalds221af7f2010-01-28 22:14:42 -08001121 return 0;
1122
1123out:
1124 return retval;
1125}
1126EXPORT_SYMBOL(flush_old_exec);
1127
Al Viro1b5d7832011-06-19 12:49:47 -04001128void would_dump(struct linux_binprm *bprm, struct file *file)
1129{
1130 if (inode_permission(file->f_path.dentry->d_inode, MAY_READ) < 0)
1131 bprm->interp_flags |= BINPRM_FLAGS_ENFORCE_NONDUMP;
1132}
1133EXPORT_SYMBOL(would_dump);
1134
Linus Torvalds221af7f2010-01-28 22:14:42 -08001135void setup_new_exec(struct linux_binprm * bprm)
1136{
Linus Torvalds221af7f2010-01-28 22:14:42 -08001137 arch_pick_mmap_layout(current->mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001138
1139 /* This is the point of no return */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001140 current->sas_ss_sp = current->sas_ss_size = 0;
1141
David Howellsda9592e2008-11-14 10:39:05 +11001142 if (current_euid() == current_uid() && current_egid() == current_gid())
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001143 set_dumpable(current->mm, 1);
Alan Coxd6e71142005-06-23 00:09:43 -07001144 else
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001145 set_dumpable(current->mm, suid_dumpable);
Alan Coxd6e71142005-06-23 00:09:43 -07001146
Heiko Carstens96e02d12012-02-04 10:47:10 +01001147 set_task_comm(current, bprm->tcomm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001148
Benjamin Herrenschmidt0551fbd2006-02-28 16:59:19 -08001149 /* Set the new mm task size. We have to do that late because it may
1150 * depend on TIF_32BIT which is only updated in flush_thread() on
1151 * some architectures like powerpc
1152 */
1153 current->mm->task_size = TASK_SIZE;
1154
David Howellsa6f76f22008-11-14 10:39:24 +11001155 /* install the new credentials */
1156 if (bprm->cred->uid != current_euid() ||
1157 bprm->cred->gid != current_egid()) {
Marcel Holtmannd2d56c52007-08-17 21:47:58 +02001158 current->pdeath_signal = 0;
Al Viro1b5d7832011-06-19 12:49:47 -04001159 } else {
1160 would_dump(bprm, bprm->file);
1161 if (bprm->interp_flags & BINPRM_FLAGS_ENFORCE_NONDUMP)
1162 set_dumpable(current->mm, suid_dumpable);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001163 }
1164
Ingo Molnarf65cb452008-12-16 13:40:44 +01001165 /*
1166 * Flush performance counters when crossing a
1167 * security domain:
1168 */
1169 if (!get_dumpable(current->mm))
Ingo Molnarcdd6c482009-09-21 12:02:48 +02001170 perf_event_exit_task(current);
Ingo Molnarf65cb452008-12-16 13:40:44 +01001171
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172 /* An exec changes our domain. We are no longer part of the thread
1173 group */
1174
1175 current->self_exec_id++;
1176
1177 flush_signal_handlers(current, 0);
1178 flush_old_files(current->files);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001179}
Linus Torvalds221af7f2010-01-28 22:14:42 -08001180EXPORT_SYMBOL(setup_new_exec);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001181
David Howellsa6f76f22008-11-14 10:39:24 +11001182/*
Oleg Nesterova2a84742009-09-05 11:17:13 -07001183 * Prepare credentials and lock ->cred_guard_mutex.
1184 * install_exec_creds() commits the new creds and drops the lock.
1185 * Or, if exec fails before, free_bprm() should release ->cred and
1186 * and unlock.
1187 */
1188int prepare_bprm_creds(struct linux_binprm *bprm)
1189{
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001190 if (mutex_lock_interruptible(&current->signal->cred_guard_mutex))
Oleg Nesterova2a84742009-09-05 11:17:13 -07001191 return -ERESTARTNOINTR;
1192
1193 bprm->cred = prepare_exec_creds();
1194 if (likely(bprm->cred))
1195 return 0;
1196
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001197 mutex_unlock(&current->signal->cred_guard_mutex);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001198 return -ENOMEM;
1199}
1200
1201void free_bprm(struct linux_binprm *bprm)
1202{
1203 free_arg_pages(bprm);
1204 if (bprm->cred) {
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001205 mutex_unlock(&current->signal->cred_guard_mutex);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001206 abort_creds(bprm->cred);
1207 }
1208 kfree(bprm);
1209}
1210
1211/*
David Howellsa6f76f22008-11-14 10:39:24 +11001212 * install the new credentials for this executable
1213 */
1214void install_exec_creds(struct linux_binprm *bprm)
1215{
1216 security_bprm_committing_creds(bprm);
1217
1218 commit_creds(bprm->cred);
1219 bprm->cred = NULL;
Oleg Nesterova2a84742009-09-05 11:17:13 -07001220 /*
1221 * cred_guard_mutex must be held at least to this point to prevent
David Howellsa6f76f22008-11-14 10:39:24 +11001222 * ptrace_attach() from altering our determination of the task's
Oleg Nesterova2a84742009-09-05 11:17:13 -07001223 * credentials; any time after this it may be unlocked.
1224 */
David Howellsa6f76f22008-11-14 10:39:24 +11001225 security_bprm_committed_creds(bprm);
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001226 mutex_unlock(&current->signal->cred_guard_mutex);
David Howellsa6f76f22008-11-14 10:39:24 +11001227}
1228EXPORT_SYMBOL(install_exec_creds);
1229
1230/*
1231 * determine how safe it is to execute the proposed program
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001232 * - the caller must hold ->cred_guard_mutex to protect against
David Howellsa6f76f22008-11-14 10:39:24 +11001233 * PTRACE_ATTACH
1234 */
Al Virof47ec3f2011-11-21 21:15:42 -05001235static int check_unsafe_exec(struct linux_binprm *bprm)
David Howellsa6f76f22008-11-14 10:39:24 +11001236{
David Howells0bf2f3a2009-02-06 11:45:46 +00001237 struct task_struct *p = current, *t;
Al Virof1191b52009-03-30 07:35:18 -04001238 unsigned n_fs;
Al Viro498052b2009-03-30 07:20:30 -04001239 int res = 0;
David Howellsa6f76f22008-11-14 10:39:24 +11001240
Tejun Heo4b9d33e2011-06-17 16:50:38 +02001241 if (p->ptrace) {
1242 if (p->ptrace & PT_PTRACE_CAP)
1243 bprm->unsafe |= LSM_UNSAFE_PTRACE_CAP;
1244 else
1245 bprm->unsafe |= LSM_UNSAFE_PTRACE;
1246 }
David Howellsa6f76f22008-11-14 10:39:24 +11001247
David Howells0bf2f3a2009-02-06 11:45:46 +00001248 n_fs = 1;
Nick Piggin2a4419b2010-08-18 04:37:33 +10001249 spin_lock(&p->fs->lock);
Oleg Nesterov437f7fd2009-04-24 01:02:45 +02001250 rcu_read_lock();
David Howells0bf2f3a2009-02-06 11:45:46 +00001251 for (t = next_thread(p); t != p; t = next_thread(t)) {
1252 if (t->fs == p->fs)
1253 n_fs++;
David Howells0bf2f3a2009-02-06 11:45:46 +00001254 }
Oleg Nesterov437f7fd2009-04-24 01:02:45 +02001255 rcu_read_unlock();
David Howells0bf2f3a2009-02-06 11:45:46 +00001256
Al Virof1191b52009-03-30 07:35:18 -04001257 if (p->fs->users > n_fs) {
David Howellsa6f76f22008-11-14 10:39:24 +11001258 bprm->unsafe |= LSM_UNSAFE_SHARE;
Al Viro498052b2009-03-30 07:20:30 -04001259 } else {
Oleg Nesterov8c652f92009-04-24 01:01:56 +02001260 res = -EAGAIN;
1261 if (!p->fs->in_exec) {
1262 p->fs->in_exec = 1;
1263 res = 1;
1264 }
Al Viro498052b2009-03-30 07:20:30 -04001265 }
Nick Piggin2a4419b2010-08-18 04:37:33 +10001266 spin_unlock(&p->fs->lock);
Al Viro498052b2009-03-30 07:20:30 -04001267
1268 return res;
David Howellsa6f76f22008-11-14 10:39:24 +11001269}
1270
Linus Torvalds1da177e2005-04-16 15:20:36 -07001271/*
1272 * Fill the binprm structure from the inode.
1273 * Check permissions, then read the first 128 (BINPRM_BUF_SIZE) bytes
David Howellsa6f76f22008-11-14 10:39:24 +11001274 *
1275 * This may be called multiple times for binary chains (scripts for example).
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276 */
1277int prepare_binprm(struct linux_binprm *bprm)
1278{
David Howellsa6f76f22008-11-14 10:39:24 +11001279 umode_t mode;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001280 struct inode * inode = bprm->file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001281 int retval;
1282
1283 mode = inode->i_mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001284 if (bprm->file->f_op == NULL)
1285 return -EACCES;
1286
David Howellsa6f76f22008-11-14 10:39:24 +11001287 /* clear any previous set[ug]id data from a previous binary */
1288 bprm->cred->euid = current_euid();
1289 bprm->cred->egid = current_egid();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001290
David Howellsa6f76f22008-11-14 10:39:24 +11001291 if (!(bprm->file->f_path.mnt->mnt_flags & MNT_NOSUID)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292 /* Set-uid? */
1293 if (mode & S_ISUID) {
David Howellsa6f76f22008-11-14 10:39:24 +11001294 bprm->per_clear |= PER_CLEAR_ON_SETID;
1295 bprm->cred->euid = inode->i_uid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001296 }
1297
1298 /* Set-gid? */
1299 /*
1300 * If setgid is set but no group execute bit then this
1301 * is a candidate for mandatory locking, not a setgid
1302 * executable.
1303 */
1304 if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) {
David Howellsa6f76f22008-11-14 10:39:24 +11001305 bprm->per_clear |= PER_CLEAR_ON_SETID;
1306 bprm->cred->egid = inode->i_gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001307 }
1308 }
1309
1310 /* fill in binprm security blob */
David Howellsa6f76f22008-11-14 10:39:24 +11001311 retval = security_bprm_set_creds(bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001312 if (retval)
1313 return retval;
David Howellsa6f76f22008-11-14 10:39:24 +11001314 bprm->cred_prepared = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001315
David Howellsa6f76f22008-11-14 10:39:24 +11001316 memset(bprm->buf, 0, BINPRM_BUF_SIZE);
1317 return kernel_read(bprm->file, 0, bprm->buf, BINPRM_BUF_SIZE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001318}
1319
1320EXPORT_SYMBOL(prepare_binprm);
1321
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001322/*
1323 * Arguments are '\0' separated strings found at the location bprm->p
1324 * points to; chop off the first by relocating brpm->p to right after
1325 * the first '\0' encountered.
1326 */
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001327int remove_arg_zero(struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001328{
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001329 int ret = 0;
1330 unsigned long offset;
1331 char *kaddr;
1332 struct page *page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001333
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001334 if (!bprm->argc)
1335 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001336
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001337 do {
1338 offset = bprm->p & ~PAGE_MASK;
1339 page = get_arg_page(bprm, bprm->p, 0);
1340 if (!page) {
1341 ret = -EFAULT;
1342 goto out;
1343 }
Cong Wange8e3c3d2011-11-25 23:14:27 +08001344 kaddr = kmap_atomic(page);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001345
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001346 for (; offset < PAGE_SIZE && kaddr[offset];
1347 offset++, bprm->p++)
1348 ;
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001349
Cong Wange8e3c3d2011-11-25 23:14:27 +08001350 kunmap_atomic(kaddr);
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001351 put_arg_page(page);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001352
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001353 if (offset == PAGE_SIZE)
1354 free_arg_page(bprm, (bprm->p >> PAGE_SHIFT) - 1);
1355 } while (offset == PAGE_SIZE);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001356
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001357 bprm->p++;
1358 bprm->argc--;
1359 ret = 0;
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001360
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001361out:
1362 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001363}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001364EXPORT_SYMBOL(remove_arg_zero);
1365
1366/*
1367 * cycle the list of binary formats handler, until one recognizes the image
1368 */
1369int search_binary_handler(struct linux_binprm *bprm,struct pt_regs *regs)
1370{
Roland McGrath85f33462008-12-09 19:36:38 -08001371 unsigned int depth = bprm->recursion_depth;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001372 int try,retval;
1373 struct linux_binfmt *fmt;
Oleg Nesterov63081912012-03-30 18:26:36 +02001374 pid_t old_pid, old_vpid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375
Linus Torvalds1da177e2005-04-16 15:20:36 -07001376 retval = security_bprm_check(bprm);
1377 if (retval)
1378 return retval;
1379
Al Viro473ae302006-04-26 14:04:08 -04001380 retval = audit_bprm(bprm);
1381 if (retval)
1382 return retval;
1383
Denys Vlasenkobb188d72011-06-29 04:13:39 +02001384 /* Need to fetch pid before load_binary changes it */
Oleg Nesterov63081912012-03-30 18:26:36 +02001385 old_pid = current->pid;
Denys Vlasenkobb188d72011-06-29 04:13:39 +02001386 rcu_read_lock();
Oleg Nesterov63081912012-03-30 18:26:36 +02001387 old_vpid = task_pid_nr_ns(current, task_active_pid_ns(current->parent));
Denys Vlasenkobb188d72011-06-29 04:13:39 +02001388 rcu_read_unlock();
1389
Linus Torvalds1da177e2005-04-16 15:20:36 -07001390 retval = -ENOENT;
1391 for (try=0; try<2; try++) {
1392 read_lock(&binfmt_lock);
Alexey Dobriyane4dc1b12007-10-16 23:26:03 -07001393 list_for_each_entry(fmt, &formats, lh) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001394 int (*fn)(struct linux_binprm *, struct pt_regs *) = fmt->load_binary;
1395 if (!fn)
1396 continue;
1397 if (!try_module_get(fmt->module))
1398 continue;
1399 read_unlock(&binfmt_lock);
1400 retval = fn(bprm, regs);
Roland McGrath85f33462008-12-09 19:36:38 -08001401 /*
1402 * Restore the depth counter to its starting value
1403 * in this call, so we don't have to rely on every
1404 * load_binary function to restore it on return.
1405 */
1406 bprm->recursion_depth = depth;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001407 if (retval >= 0) {
David Smith4ff16c22012-02-07 10:11:05 -06001408 if (depth == 0) {
1409 trace_sched_process_exec(current, old_pid, bprm);
Oleg Nesterov63081912012-03-30 18:26:36 +02001410 ptrace_event(PTRACE_EVENT_EXEC, old_vpid);
David Smith4ff16c22012-02-07 10:11:05 -06001411 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001412 put_binfmt(fmt);
1413 allow_write_access(bprm->file);
1414 if (bprm->file)
1415 fput(bprm->file);
1416 bprm->file = NULL;
1417 current->did_exec = 1;
Matt Helsley9f460802005-11-07 00:59:16 -08001418 proc_exec_connector(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001419 return retval;
1420 }
1421 read_lock(&binfmt_lock);
1422 put_binfmt(fmt);
1423 if (retval != -ENOEXEC || bprm->mm == NULL)
1424 break;
1425 if (!bprm->file) {
1426 read_unlock(&binfmt_lock);
1427 return retval;
1428 }
1429 }
1430 read_unlock(&binfmt_lock);
Tetsuo Handab4edf8b2011-07-26 16:08:42 -07001431#ifdef CONFIG_MODULES
Linus Torvalds1da177e2005-04-16 15:20:36 -07001432 if (retval != -ENOEXEC || bprm->mm == NULL) {
1433 break;
Johannes Berg5f4123b2008-07-09 10:28:40 +02001434 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001435#define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e))
1436 if (printable(bprm->buf[0]) &&
1437 printable(bprm->buf[1]) &&
1438 printable(bprm->buf[2]) &&
1439 printable(bprm->buf[3]))
1440 break; /* -ENOEXEC */
Tetsuo Handa91219352011-07-26 16:08:41 -07001441 if (try)
1442 break; /* -ENOEXEC */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001443 request_module("binfmt-%04x", *(unsigned short *)(&bprm->buf[2]));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001444 }
Tetsuo Handab4edf8b2011-07-26 16:08:42 -07001445#else
1446 break;
1447#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001448 }
1449 return retval;
1450}
1451
1452EXPORT_SYMBOL(search_binary_handler);
1453
1454/*
1455 * sys_execve() executes a new program.
1456 */
Oleg Nesterovba2d01622011-03-06 18:02:37 +01001457static int do_execve_common(const char *filename,
1458 struct user_arg_ptr argv,
1459 struct user_arg_ptr envp,
1460 struct pt_regs *regs)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001461{
1462 struct linux_binprm *bprm;
1463 struct file *file;
Al Viro3b125382008-04-22 05:31:30 -04001464 struct files_struct *displaced;
Oleg Nesterov8c652f92009-04-24 01:01:56 +02001465 bool clear_in_exec;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001466 int retval;
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +04001467 const struct cred *cred = current_cred();
1468
1469 /*
1470 * We move the actual failure in case of RLIMIT_NPROC excess from
1471 * set*uid() to execve() because too many poorly written programs
1472 * don't check setuid() return code. Here we additionally recheck
1473 * whether NPROC limit is still exceeded.
1474 */
1475 if ((current->flags & PF_NPROC_EXCEEDED) &&
1476 atomic_read(&cred->user->processes) > rlimit(RLIMIT_NPROC)) {
1477 retval = -EAGAIN;
1478 goto out_ret;
1479 }
1480
1481 /* We're below the limit (still or again), so we don't want to make
1482 * further execve() calls fail. */
1483 current->flags &= ~PF_NPROC_EXCEEDED;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001484
Al Viro3b125382008-04-22 05:31:30 -04001485 retval = unshare_files(&displaced);
Al Virofd8328b2008-04-22 05:11:59 -04001486 if (retval)
1487 goto out_ret;
1488
Linus Torvalds1da177e2005-04-16 15:20:36 -07001489 retval = -ENOMEM;
Oliver Neukum11b0b5a2006-03-25 03:08:13 -08001490 bprm = kzalloc(sizeof(*bprm), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001491 if (!bprm)
Al Virofd8328b2008-04-22 05:11:59 -04001492 goto out_files;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001493
Oleg Nesterova2a84742009-09-05 11:17:13 -07001494 retval = prepare_bprm_creds(bprm);
1495 if (retval)
David Howellsa6f76f22008-11-14 10:39:24 +11001496 goto out_free;
Al Viro498052b2009-03-30 07:20:30 -04001497
1498 retval = check_unsafe_exec(bprm);
Oleg Nesterov8c652f92009-04-24 01:01:56 +02001499 if (retval < 0)
Oleg Nesterova2a84742009-09-05 11:17:13 -07001500 goto out_free;
Oleg Nesterov8c652f92009-04-24 01:01:56 +02001501 clear_in_exec = retval;
Oleg Nesterova2a84742009-09-05 11:17:13 -07001502 current->in_execve = 1;
David Howellsa6f76f22008-11-14 10:39:24 +11001503
Linus Torvalds1da177e2005-04-16 15:20:36 -07001504 file = open_exec(filename);
1505 retval = PTR_ERR(file);
1506 if (IS_ERR(file))
Al Viro498052b2009-03-30 07:20:30 -04001507 goto out_unmark;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001508
1509 sched_exec();
1510
Linus Torvalds1da177e2005-04-16 15:20:36 -07001511 bprm->file = file;
1512 bprm->filename = filename;
1513 bprm->interp = filename;
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001514
1515 retval = bprm_mm_init(bprm);
1516 if (retval)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001517 goto out_file;
1518
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001519 bprm->argc = count(argv, MAX_ARG_STRINGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001520 if ((retval = bprm->argc) < 0)
David Howellsa6f76f22008-11-14 10:39:24 +11001521 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001522
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001523 bprm->envc = count(envp, MAX_ARG_STRINGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001524 if ((retval = bprm->envc) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001525 goto out;
1526
1527 retval = prepare_binprm(bprm);
1528 if (retval < 0)
1529 goto out;
1530
1531 retval = copy_strings_kernel(1, &bprm->filename, bprm);
1532 if (retval < 0)
1533 goto out;
1534
1535 bprm->exec = bprm->p;
1536 retval = copy_strings(bprm->envc, envp, bprm);
1537 if (retval < 0)
1538 goto out;
1539
1540 retval = copy_strings(bprm->argc, argv, bprm);
1541 if (retval < 0)
1542 goto out;
1543
1544 retval = search_binary_handler(bprm,regs);
David Howellsa6f76f22008-11-14 10:39:24 +11001545 if (retval < 0)
1546 goto out;
1547
1548 /* execve succeeded */
Al Viro498052b2009-03-30 07:20:30 -04001549 current->fs->in_exec = 0;
Kentaro Takedaf9ce1f12009-02-05 17:18:11 +09001550 current->in_execve = 0;
David Howellsa6f76f22008-11-14 10:39:24 +11001551 acct_update_integrals(current);
1552 free_bprm(bprm);
1553 if (displaced)
1554 put_files_struct(displaced);
1555 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001556
1557out:
Oleg Nesterov3c77f842010-11-30 20:55:34 +01001558 if (bprm->mm) {
1559 acct_arg_size(bprm, 0);
1560 mmput(bprm->mm);
1561 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001562
1563out_file:
1564 if (bprm->file) {
1565 allow_write_access(bprm->file);
1566 fput(bprm->file);
1567 }
David Howellsa6f76f22008-11-14 10:39:24 +11001568
Al Viro498052b2009-03-30 07:20:30 -04001569out_unmark:
Oleg Nesterov8c652f92009-04-24 01:01:56 +02001570 if (clear_in_exec)
1571 current->fs->in_exec = 0;
Kentaro Takedaf9ce1f12009-02-05 17:18:11 +09001572 current->in_execve = 0;
David Howellsa6f76f22008-11-14 10:39:24 +11001573
1574out_free:
Al Viro08a6fac2008-05-10 16:38:25 -04001575 free_bprm(bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001576
Al Virofd8328b2008-04-22 05:11:59 -04001577out_files:
Al Viro3b125382008-04-22 05:31:30 -04001578 if (displaced)
1579 reset_files_struct(displaced);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001580out_ret:
1581 return retval;
1582}
1583
Oleg Nesterovba2d01622011-03-06 18:02:37 +01001584int do_execve(const char *filename,
1585 const char __user *const __user *__argv,
1586 const char __user *const __user *__envp,
1587 struct pt_regs *regs)
1588{
Oleg Nesterov0e028462011-03-06 18:02:54 +01001589 struct user_arg_ptr argv = { .ptr.native = __argv };
1590 struct user_arg_ptr envp = { .ptr.native = __envp };
Oleg Nesterovba2d01622011-03-06 18:02:37 +01001591 return do_execve_common(filename, argv, envp, regs);
1592}
1593
Oleg Nesterov0e028462011-03-06 18:02:54 +01001594#ifdef CONFIG_COMPAT
1595int compat_do_execve(char *filename,
1596 compat_uptr_t __user *__argv,
1597 compat_uptr_t __user *__envp,
1598 struct pt_regs *regs)
1599{
1600 struct user_arg_ptr argv = {
1601 .is_compat = true,
1602 .ptr.compat = __argv,
1603 };
1604 struct user_arg_ptr envp = {
1605 .is_compat = true,
1606 .ptr.compat = __envp,
1607 };
1608 return do_execve_common(filename, argv, envp, regs);
1609}
1610#endif
1611
Oleg Nesterov964ee7d2009-09-23 15:56:59 -07001612void set_binfmt(struct linux_binfmt *new)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001613{
Hiroshi Shimamoto801460d2009-09-23 15:57:41 -07001614 struct mm_struct *mm = current->mm;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001615
Hiroshi Shimamoto801460d2009-09-23 15:57:41 -07001616 if (mm->binfmt)
1617 module_put(mm->binfmt->module);
1618
1619 mm->binfmt = new;
Oleg Nesterov964ee7d2009-09-23 15:56:59 -07001620 if (new)
1621 __module_get(new->module);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001622}
1623
1624EXPORT_SYMBOL(set_binfmt);
1625
Xiaotian Feng1b0d3002010-10-27 15:34:08 -07001626static int expand_corename(struct core_name *cn)
1627{
1628 char *old_corename = cn->corename;
1629
1630 cn->size = CORENAME_MAX_SIZE * atomic_inc_return(&call_count);
1631 cn->corename = krealloc(old_corename, cn->size, GFP_KERNEL);
1632
1633 if (!cn->corename) {
1634 kfree(old_corename);
1635 return -ENOMEM;
1636 }
1637
1638 return 0;
1639}
1640
1641static int cn_printf(struct core_name *cn, const char *fmt, ...)
1642{
1643 char *cur;
1644 int need;
1645 int ret;
1646 va_list arg;
1647
1648 va_start(arg, fmt);
1649 need = vsnprintf(NULL, 0, fmt, arg);
1650 va_end(arg);
1651
1652 if (likely(need < cn->size - cn->used - 1))
1653 goto out_printf;
1654
1655 ret = expand_corename(cn);
1656 if (ret)
1657 goto expand_fail;
1658
1659out_printf:
1660 cur = cn->corename + cn->used;
1661 va_start(arg, fmt);
1662 vsnprintf(cur, need + 1, fmt, arg);
1663 va_end(arg);
1664 cn->used += need;
1665 return 0;
1666
1667expand_fail:
1668 return ret;
1669}
1670
Jiri Slaby2c563732011-07-26 16:08:33 -07001671static void cn_escape(char *str)
1672{
1673 for (; *str; str++)
1674 if (*str == '/')
1675 *str = '!';
1676}
1677
Jiri Slaby57cc0832011-05-26 16:25:46 -07001678static int cn_print_exe_file(struct core_name *cn)
1679{
1680 struct file *exe_file;
Jiri Slaby2c563732011-07-26 16:08:33 -07001681 char *pathbuf, *path;
Jiri Slaby57cc0832011-05-26 16:25:46 -07001682 int ret;
1683
1684 exe_file = get_mm_exe_file(current->mm);
Jiri Slaby2c563732011-07-26 16:08:33 -07001685 if (!exe_file) {
1686 char *commstart = cn->corename + cn->used;
1687 ret = cn_printf(cn, "%s (path unknown)", current->comm);
1688 cn_escape(commstart);
1689 return ret;
1690 }
Jiri Slaby57cc0832011-05-26 16:25:46 -07001691
1692 pathbuf = kmalloc(PATH_MAX, GFP_TEMPORARY);
1693 if (!pathbuf) {
1694 ret = -ENOMEM;
1695 goto put_exe_file;
1696 }
1697
1698 path = d_path(&exe_file->f_path, pathbuf, PATH_MAX);
1699 if (IS_ERR(path)) {
1700 ret = PTR_ERR(path);
1701 goto free_buf;
1702 }
1703
Jiri Slaby2c563732011-07-26 16:08:33 -07001704 cn_escape(path);
Jiri Slaby57cc0832011-05-26 16:25:46 -07001705
1706 ret = cn_printf(cn, "%s", path);
1707
1708free_buf:
1709 kfree(pathbuf);
1710put_exe_file:
1711 fput(exe_file);
1712 return ret;
1713}
1714
Linus Torvalds1da177e2005-04-16 15:20:36 -07001715/* format_corename will inspect the pattern parameter, and output a
1716 * name into corename, which must have space for at least
1717 * CORENAME_MAX_SIZE bytes plus one byte for the zero terminator.
1718 */
Xiaotian Feng1b0d3002010-10-27 15:34:08 -07001719static int format_corename(struct core_name *cn, long signr)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001720{
David Howells86a264a2008-11-14 10:39:18 +11001721 const struct cred *cred = current_cred();
Oleg Nesterov565b9b12008-07-25 01:47:47 -07001722 const char *pat_ptr = core_pattern;
1723 int ispipe = (*pat_ptr == '|');
Linus Torvalds1da177e2005-04-16 15:20:36 -07001724 int pid_in_pattern = 0;
Xiaotian Feng1b0d3002010-10-27 15:34:08 -07001725 int err = 0;
1726
1727 cn->size = CORENAME_MAX_SIZE * atomic_read(&call_count);
1728 cn->corename = kmalloc(cn->size, GFP_KERNEL);
1729 cn->used = 0;
1730
1731 if (!cn->corename)
1732 return -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001733
1734 /* Repeat as long as we have more pattern to process and more output
1735 space */
1736 while (*pat_ptr) {
1737 if (*pat_ptr != '%') {
Xiaotian Feng1b0d3002010-10-27 15:34:08 -07001738 if (*pat_ptr == 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001739 goto out;
Xiaotian Feng1b0d3002010-10-27 15:34:08 -07001740 err = cn_printf(cn, "%c", *pat_ptr++);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001741 } else {
1742 switch (*++pat_ptr) {
Xiaotian Feng1b0d3002010-10-27 15:34:08 -07001743 /* single % at the end, drop that */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001744 case 0:
1745 goto out;
1746 /* Double percent, output one percent */
1747 case '%':
Xiaotian Feng1b0d3002010-10-27 15:34:08 -07001748 err = cn_printf(cn, "%c", '%');
Linus Torvalds1da177e2005-04-16 15:20:36 -07001749 break;
1750 /* pid */
1751 case 'p':
1752 pid_in_pattern = 1;
Xiaotian Feng1b0d3002010-10-27 15:34:08 -07001753 err = cn_printf(cn, "%d",
1754 task_tgid_vnr(current));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001755 break;
1756 /* uid */
1757 case 'u':
Xiaotian Feng1b0d3002010-10-27 15:34:08 -07001758 err = cn_printf(cn, "%d", cred->uid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001759 break;
1760 /* gid */
1761 case 'g':
Xiaotian Feng1b0d3002010-10-27 15:34:08 -07001762 err = cn_printf(cn, "%d", cred->gid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001763 break;
1764 /* signal that caused the coredump */
1765 case 's':
Xiaotian Feng1b0d3002010-10-27 15:34:08 -07001766 err = cn_printf(cn, "%ld", signr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001767 break;
1768 /* UNIX time of coredump */
1769 case 't': {
1770 struct timeval tv;
1771 do_gettimeofday(&tv);
Xiaotian Feng1b0d3002010-10-27 15:34:08 -07001772 err = cn_printf(cn, "%lu", tv.tv_sec);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001773 break;
1774 }
1775 /* hostname */
Jiri Slaby2c563732011-07-26 16:08:33 -07001776 case 'h': {
1777 char *namestart = cn->corename + cn->used;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001778 down_read(&uts_sem);
Xiaotian Feng1b0d3002010-10-27 15:34:08 -07001779 err = cn_printf(cn, "%s",
1780 utsname()->nodename);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001781 up_read(&uts_sem);
Jiri Slaby2c563732011-07-26 16:08:33 -07001782 cn_escape(namestart);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001783 break;
Jiri Slaby2c563732011-07-26 16:08:33 -07001784 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001785 /* executable */
Jiri Slaby2c563732011-07-26 16:08:33 -07001786 case 'e': {
1787 char *commstart = cn->corename + cn->used;
Xiaotian Feng1b0d3002010-10-27 15:34:08 -07001788 err = cn_printf(cn, "%s", current->comm);
Jiri Slaby2c563732011-07-26 16:08:33 -07001789 cn_escape(commstart);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001790 break;
Jiri Slaby2c563732011-07-26 16:08:33 -07001791 }
Jiri Slaby57cc0832011-05-26 16:25:46 -07001792 case 'E':
1793 err = cn_print_exe_file(cn);
1794 break;
Neil Horman74aadce2007-10-16 23:26:35 -07001795 /* core limit size */
1796 case 'c':
Xiaotian Feng1b0d3002010-10-27 15:34:08 -07001797 err = cn_printf(cn, "%lu",
1798 rlimit(RLIMIT_CORE));
Neil Horman74aadce2007-10-16 23:26:35 -07001799 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001800 default:
1801 break;
1802 }
1803 ++pat_ptr;
1804 }
Xiaotian Feng1b0d3002010-10-27 15:34:08 -07001805
1806 if (err)
1807 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001808 }
Xiaotian Feng1b0d3002010-10-27 15:34:08 -07001809
Linus Torvalds1da177e2005-04-16 15:20:36 -07001810 /* Backward compatibility with core_uses_pid:
1811 *
1812 * If core_pattern does not include a %p (as is the default)
1813 * and core_uses_pid is set, then .%pid will be appended to
Alan Coxc4bbafd2007-04-16 22:53:13 -07001814 * the filename. Do not do this for piped commands. */
Oleg Nesterov64093242008-10-18 20:28:22 -07001815 if (!ispipe && !pid_in_pattern && core_uses_pid) {
Xiaotian Feng1b0d3002010-10-27 15:34:08 -07001816 err = cn_printf(cn, ".%d", task_tgid_vnr(current));
1817 if (err)
1818 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001819 }
Alan Coxc4bbafd2007-04-16 22:53:13 -07001820out:
Alan Coxc4bbafd2007-04-16 22:53:13 -07001821 return ispipe;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001822}
1823
Oleg Nesterov5c99cbf2010-03-05 13:44:14 -08001824static int zap_process(struct task_struct *start, int exit_code)
Oleg Nesterovaceecc02006-06-26 00:26:05 -07001825{
1826 struct task_struct *t;
Oleg Nesterov8cd9c242008-07-25 01:47:42 -07001827 int nr = 0;
Oleg Nesterov281de332006-06-26 00:26:06 -07001828
Oleg Nesterovd5f70c02006-06-26 00:26:07 -07001829 start->signal->flags = SIGNAL_GROUP_EXIT;
Oleg Nesterov5c99cbf2010-03-05 13:44:14 -08001830 start->signal->group_exit_code = exit_code;
Oleg Nesterovd5f70c02006-06-26 00:26:07 -07001831 start->signal->group_stop_count = 0;
Oleg Nesterovaceecc02006-06-26 00:26:05 -07001832
1833 t = start;
1834 do {
Tejun Heo6dfca322011-06-02 11:14:00 +02001835 task_clear_jobctl_pending(t, JOBCTL_PENDING_MASK);
Oleg Nesterovaceecc02006-06-26 00:26:05 -07001836 if (t != current && t->mm) {
Oleg Nesterov281de332006-06-26 00:26:06 -07001837 sigaddset(&t->pending.signal, SIGKILL);
1838 signal_wake_up(t, 1);
Oleg Nesterov8cd9c242008-07-25 01:47:42 -07001839 nr++;
Oleg Nesterovaceecc02006-06-26 00:26:05 -07001840 }
Oleg Nesterove4901f92008-07-25 01:47:31 -07001841 } while_each_thread(start, t);
Oleg Nesterov8cd9c242008-07-25 01:47:42 -07001842
1843 return nr;
Oleg Nesterovaceecc02006-06-26 00:26:05 -07001844}
1845
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001846static inline int zap_threads(struct task_struct *tsk, struct mm_struct *mm,
Oleg Nesterov8cd9c242008-07-25 01:47:42 -07001847 struct core_state *core_state, int exit_code)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001848{
1849 struct task_struct *g, *p;
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001850 unsigned long flags;
Oleg Nesterov8cd9c242008-07-25 01:47:42 -07001851 int nr = -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001852
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001853 spin_lock_irq(&tsk->sighand->siglock);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -08001854 if (!signal_group_exit(tsk->signal)) {
Oleg Nesterov8cd9c242008-07-25 01:47:42 -07001855 mm->core_state = core_state;
Oleg Nesterov5c99cbf2010-03-05 13:44:14 -08001856 nr = zap_process(tsk, exit_code);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001857 }
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001858 spin_unlock_irq(&tsk->sighand->siglock);
Oleg Nesterov8cd9c242008-07-25 01:47:42 -07001859 if (unlikely(nr < 0))
1860 return nr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001861
Oleg Nesterov8cd9c242008-07-25 01:47:42 -07001862 if (atomic_read(&mm->mm_users) == nr + 1)
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001863 goto done;
Oleg Nesterove4901f92008-07-25 01:47:31 -07001864 /*
1865 * We should find and kill all tasks which use this mm, and we should
Oleg Nesterov999d9fc2008-07-25 01:47:41 -07001866 * count them correctly into ->nr_threads. We don't take tasklist
Oleg Nesterove4901f92008-07-25 01:47:31 -07001867 * lock, but this is safe wrt:
1868 *
1869 * fork:
1870 * None of sub-threads can fork after zap_process(leader). All
1871 * processes which were created before this point should be
1872 * visible to zap_threads() because copy_process() adds the new
1873 * process to the tail of init_task.tasks list, and lock/unlock
1874 * of ->siglock provides a memory barrier.
1875 *
1876 * do_exit:
1877 * The caller holds mm->mmap_sem. This means that the task which
1878 * uses this mm can't pass exit_mm(), so it can't exit or clear
1879 * its ->mm.
1880 *
1881 * de_thread:
1882 * It does list_replace_rcu(&leader->tasks, &current->tasks),
1883 * we must see either old or new leader, this does not matter.
1884 * However, it can change p->sighand, so lock_task_sighand(p)
1885 * must be used. Since p->mm != NULL and we hold ->mmap_sem
1886 * it can't fail.
1887 *
1888 * Note also that "g" can be the old leader with ->mm == NULL
1889 * and already unhashed and thus removed from ->thread_group.
1890 * This is OK, __unhash_process()->list_del_rcu() does not
1891 * clear the ->next pointer, we will find the new leader via
1892 * next_thread().
1893 */
Oleg Nesterov7b1c6152006-06-26 00:26:08 -07001894 rcu_read_lock();
Oleg Nesterovaceecc02006-06-26 00:26:05 -07001895 for_each_process(g) {
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001896 if (g == tsk->group_leader)
1897 continue;
Oleg Nesterov15b9f362008-07-25 01:47:39 -07001898 if (g->flags & PF_KTHREAD)
1899 continue;
Oleg Nesterovaceecc02006-06-26 00:26:05 -07001900 p = g;
1901 do {
1902 if (p->mm) {
Oleg Nesterov15b9f362008-07-25 01:47:39 -07001903 if (unlikely(p->mm == mm)) {
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001904 lock_task_sighand(p, &flags);
Oleg Nesterov5c99cbf2010-03-05 13:44:14 -08001905 nr += zap_process(p, exit_code);
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001906 unlock_task_sighand(p, &flags);
1907 }
Oleg Nesterovaceecc02006-06-26 00:26:05 -07001908 break;
1909 }
Oleg Nesterove4901f92008-07-25 01:47:31 -07001910 } while_each_thread(g, p);
Oleg Nesterovaceecc02006-06-26 00:26:05 -07001911 }
Oleg Nesterov7b1c6152006-06-26 00:26:08 -07001912 rcu_read_unlock();
Oleg Nesterov5debfa62006-06-26 00:26:09 -07001913done:
Oleg Nesterovc5f1cc82008-07-25 01:47:42 -07001914 atomic_set(&core_state->nr_threads, nr);
Oleg Nesterov8cd9c242008-07-25 01:47:42 -07001915 return nr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001916}
1917
Oleg Nesterov9d5b3272008-07-25 01:47:43 -07001918static int coredump_wait(int exit_code, struct core_state *core_state)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001919{
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001920 struct task_struct *tsk = current;
1921 struct mm_struct *mm = tsk->mm;
Oleg Nesterov269b0052010-05-26 14:43:08 -07001922 int core_waiters = -EBUSY;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001923
Oleg Nesterov9d5b3272008-07-25 01:47:43 -07001924 init_completion(&core_state->startup);
Oleg Nesterovb564daf2008-07-25 01:47:44 -07001925 core_state->dumper.task = tsk;
1926 core_state->dumper.next = NULL;
Oleg Nesterov269b0052010-05-26 14:43:08 -07001927
1928 down_write(&mm->mmap_sem);
1929 if (!mm->core_state)
1930 core_waiters = zap_threads(tsk, mm, core_state, exit_code);
Oleg Nesterov2384f552005-10-30 15:02:47 -08001931 up_write(&mm->mmap_sem);
1932
Oleg Nesterov57b59c42012-03-05 14:59:13 -08001933 if (core_waiters > 0)
Oleg Nesterov9d5b3272008-07-25 01:47:43 -07001934 wait_for_completion(&core_state->startup);
Oleg Nesterov57b59c42012-03-05 14:59:13 -08001935
Oleg Nesterovdcf560c2006-06-26 00:26:08 -07001936 return core_waiters;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001937}
1938
Oleg Nesterova94e2d42008-07-25 01:47:46 -07001939static void coredump_finish(struct mm_struct *mm)
1940{
1941 struct core_thread *curr, *next;
1942 struct task_struct *task;
1943
1944 next = mm->core_state->dumper.next;
1945 while ((curr = next) != NULL) {
1946 next = curr->next;
1947 task = curr->task;
1948 /*
1949 * see exit_mm(), curr->task must not see
1950 * ->task == NULL before we read ->next.
1951 */
1952 smp_mb();
1953 curr->task = NULL;
1954 wake_up_process(task);
1955 }
1956
1957 mm->core_state = NULL;
1958}
1959
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001960/*
1961 * set_dumpable converts traditional three-value dumpable to two flags and
1962 * stores them into mm->flags. It modifies lower two bits of mm->flags, but
1963 * these bits are not changed atomically. So get_dumpable can observe the
1964 * intermediate state. To avoid doing unexpected behavior, get get_dumpable
1965 * return either old dumpable or new one by paying attention to the order of
1966 * modifying the bits.
1967 *
1968 * dumpable | mm->flags (binary)
1969 * old new | initial interim final
1970 * ---------+-----------------------
1971 * 0 1 | 00 01 01
1972 * 0 2 | 00 10(*) 11
1973 * 1 0 | 01 00 00
1974 * 1 2 | 01 11 11
1975 * 2 0 | 11 10(*) 00
1976 * 2 1 | 11 11 01
1977 *
1978 * (*) get_dumpable regards interim value of 10 as 11.
1979 */
1980void set_dumpable(struct mm_struct *mm, int value)
1981{
1982 switch (value) {
1983 case 0:
1984 clear_bit(MMF_DUMPABLE, &mm->flags);
1985 smp_wmb();
1986 clear_bit(MMF_DUMP_SECURELY, &mm->flags);
1987 break;
1988 case 1:
1989 set_bit(MMF_DUMPABLE, &mm->flags);
1990 smp_wmb();
1991 clear_bit(MMF_DUMP_SECURELY, &mm->flags);
1992 break;
1993 case 2:
1994 set_bit(MMF_DUMP_SECURELY, &mm->flags);
1995 smp_wmb();
1996 set_bit(MMF_DUMPABLE, &mm->flags);
1997 break;
1998 }
1999}
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07002000
Masami Hiramatsu30736a42010-03-05 13:44:12 -08002001static int __get_dumpable(unsigned long mm_flags)
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07002002{
2003 int ret;
2004
Masami Hiramatsu30736a42010-03-05 13:44:12 -08002005 ret = mm_flags & MMF_DUMPABLE_MASK;
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07002006 return (ret >= 2) ? 2 : ret;
2007}
2008
Masami Hiramatsu30736a42010-03-05 13:44:12 -08002009int get_dumpable(struct mm_struct *mm)
2010{
2011 return __get_dumpable(mm->flags);
2012}
2013
Neil Horman61be2282009-09-23 15:56:58 -07002014static void wait_for_dump_helpers(struct file *file)
2015{
2016 struct pipe_inode_info *pipe;
2017
2018 pipe = file->f_path.dentry->d_inode->i_pipe;
2019
2020 pipe_lock(pipe);
2021 pipe->readers++;
2022 pipe->writers--;
2023
2024 while ((pipe->readers > 1) && (!signal_pending(current))) {
2025 wake_up_interruptible_sync(&pipe->wait);
2026 kill_fasync(&pipe->fasync_readers, SIGIO, POLL_IN);
2027 pipe_wait(pipe);
2028 }
2029
2030 pipe->readers--;
2031 pipe->writers++;
2032 pipe_unlock(pipe);
2033
2034}
2035
2036
Neil Horman898b3742010-05-26 14:42:59 -07002037/*
Holger Hans Peter Freyther1bef8292011-02-24 17:46:49 +01002038 * umh_pipe_setup
Neil Horman898b3742010-05-26 14:42:59 -07002039 * helper function to customize the process used
2040 * to collect the core in userspace. Specifically
2041 * it sets up a pipe and installs it as fd 0 (stdin)
2042 * for the process. Returns 0 on success, or
2043 * PTR_ERR on failure.
2044 * Note that it also sets the core limit to 1. This
2045 * is a special value that we use to trap recursive
2046 * core dumps
2047 */
David Howells87966992011-06-17 11:25:59 +01002048static int umh_pipe_setup(struct subprocess_info *info, struct cred *new)
Neil Horman898b3742010-05-26 14:42:59 -07002049{
2050 struct file *rp, *wp;
2051 struct fdtable *fdt;
2052 struct coredump_params *cp = (struct coredump_params *)info->data;
2053 struct files_struct *cf = current->files;
2054
2055 wp = create_write_pipe(0);
2056 if (IS_ERR(wp))
2057 return PTR_ERR(wp);
2058
2059 rp = create_read_pipe(wp, 0);
2060 if (IS_ERR(rp)) {
2061 free_write_pipe(wp);
2062 return PTR_ERR(rp);
2063 }
2064
2065 cp->file = wp;
2066
2067 sys_close(0);
2068 fd_install(0, rp);
2069 spin_lock(&cf->file_lock);
2070 fdt = files_fdtable(cf);
David Howells1dce27c2012-02-16 17:49:42 +00002071 __set_open_fd(0, fdt);
2072 __clear_close_on_exec(0, fdt);
Neil Horman898b3742010-05-26 14:42:59 -07002073 spin_unlock(&cf->file_lock);
2074
2075 /* and disallow core files too */
2076 current->signal->rlim[RLIMIT_CORE] = (struct rlimit){1, 1};
2077
2078 return 0;
2079}
2080
WANG Cong8cd3ac32009-01-06 14:42:48 -08002081void do_coredump(long signr, int exit_code, struct pt_regs *regs)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002082{
Oleg Nesterov9d5b3272008-07-25 01:47:43 -07002083 struct core_state core_state;
Xiaotian Feng1b0d3002010-10-27 15:34:08 -07002084 struct core_name cn;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002085 struct mm_struct *mm = current->mm;
2086 struct linux_binfmt * binfmt;
David Howellsd84f4f92008-11-14 10:39:23 +11002087 const struct cred *old_cred;
2088 struct cred *cred;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002089 int retval = 0;
Alan Coxd6e71142005-06-23 00:09:43 -07002090 int flag = 0;
Oleg Nesterovd5bf4c42010-05-26 14:43:06 -07002091 int ispipe;
Neil Hormana2939802009-09-23 15:56:56 -07002092 static atomic_t core_dump_count = ATOMIC_INIT(0);
Masami Hiramatsuf6151df2009-12-17 15:27:16 -08002093 struct coredump_params cprm = {
2094 .signr = signr,
2095 .regs = regs,
Jiri Slabyd554ed892010-03-05 13:42:42 -08002096 .limit = rlimit(RLIMIT_CORE),
Masami Hiramatsu30736a42010-03-05 13:44:12 -08002097 /*
2098 * We must use the same mm->flags while dumping core to avoid
2099 * inconsistency of bit flags, since this flag is not protected
2100 * by any locks.
2101 */
2102 .mm_flags = mm->flags,
Masami Hiramatsuf6151df2009-12-17 15:27:16 -08002103 };
Linus Torvalds1da177e2005-04-16 15:20:36 -07002104
Steve Grubb0a4ff8c2007-04-19 10:28:21 -04002105 audit_core_dumps(signr);
2106
Hiroshi Shimamoto801460d2009-09-23 15:57:41 -07002107 binfmt = mm->binfmt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002108 if (!binfmt || !binfmt->core_dump)
2109 goto fail;
Oleg Nesterov269b0052010-05-26 14:43:08 -07002110 if (!__get_dumpable(cprm.mm_flags))
2111 goto fail;
David Howellsd84f4f92008-11-14 10:39:23 +11002112
2113 cred = prepare_creds();
Oleg Nesterov5e43aef2010-05-26 14:43:07 -07002114 if (!cred)
David Howellsd84f4f92008-11-14 10:39:23 +11002115 goto fail;
Alan Coxd6e71142005-06-23 00:09:43 -07002116 /*
2117 * We cannot trust fsuid as being the "true" uid of the
2118 * process nor do we know its entire history. We only know it
2119 * was tainted so we dump it as root in mode 2.
2120 */
Masami Hiramatsu30736a42010-03-05 13:44:12 -08002121 if (__get_dumpable(cprm.mm_flags) == 2) {
2122 /* Setuid core dump mode */
Alan Coxd6e71142005-06-23 00:09:43 -07002123 flag = O_EXCL; /* Stop rewrite attacks */
David Howellsd84f4f92008-11-14 10:39:23 +11002124 cred->fsuid = 0; /* Dump root private */
Alan Coxd6e71142005-06-23 00:09:43 -07002125 }
Oleg Nesterov1291cf42005-10-30 15:02:54 -08002126
Oleg Nesterov9d5b3272008-07-25 01:47:43 -07002127 retval = coredump_wait(exit_code, &core_state);
Oleg Nesterov5e43aef2010-05-26 14:43:07 -07002128 if (retval < 0)
2129 goto fail_creds;
David Howellsd84f4f92008-11-14 10:39:23 +11002130
2131 old_cred = override_creds(cred);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002132
2133 /*
2134 * Clear any false indication of pending signals that might
2135 * be seen by the filesystem code called to write the core file.
2136 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002137 clear_thread_flag(TIF_SIGPENDING);
2138
Xiaotian Feng1b0d3002010-10-27 15:34:08 -07002139 ispipe = format_corename(&cn, signr);
2140
Alan Coxc4bbafd2007-04-16 22:53:13 -07002141 if (ispipe) {
Oleg Nesterovd5bf4c42010-05-26 14:43:06 -07002142 int dump_count;
2143 char **helper_argv;
2144
Oleg Nesterov99b64562011-07-26 16:08:34 -07002145 if (ispipe < 0) {
2146 printk(KERN_WARNING "format_corename failed\n");
2147 printk(KERN_WARNING "Aborting core\n");
2148 goto fail_corename;
2149 }
2150
Neil Horman898b3742010-05-26 14:42:59 -07002151 if (cprm.limit == 1) {
Neil Horman725eae32009-09-23 15:56:54 -07002152 /*
2153 * Normally core limits are irrelevant to pipes, since
2154 * we're not writing to the file system, but we use
Neil Horman898b3742010-05-26 14:42:59 -07002155 * cprm.limit of 1 here as a speacial value. Any
2156 * non-1 limit gets set to RLIM_INFINITY below, but
Neil Horman725eae32009-09-23 15:56:54 -07002157 * a limit of 0 skips the dump. This is a consistent
2158 * way to catch recursive crashes. We can still crash
Neil Horman898b3742010-05-26 14:42:59 -07002159 * if the core_pattern binary sets RLIM_CORE = !1
Neil Horman725eae32009-09-23 15:56:54 -07002160 * but it runs as root, and can do lots of stupid things
2161 * Note that we use task_tgid_vnr here to grab the pid
2162 * of the process group leader. That way we get the
2163 * right pid if a thread in a multi-threaded
2164 * core_pattern process dies.
2165 */
2166 printk(KERN_WARNING
Neil Horman898b3742010-05-26 14:42:59 -07002167 "Process %d(%s) has RLIMIT_CORE set to 1\n",
Neil Horman725eae32009-09-23 15:56:54 -07002168 task_tgid_vnr(current), current->comm);
2169 printk(KERN_WARNING "Aborting core\n");
2170 goto fail_unlock;
2171 }
Oleg Nesterovd5bf4c42010-05-26 14:43:06 -07002172 cprm.limit = RLIM_INFINITY;
Neil Horman725eae32009-09-23 15:56:54 -07002173
Neil Hormana2939802009-09-23 15:56:56 -07002174 dump_count = atomic_inc_return(&core_dump_count);
2175 if (core_pipe_limit && (core_pipe_limit < dump_count)) {
2176 printk(KERN_WARNING "Pid %d(%s) over core_pipe_limit\n",
2177 task_tgid_vnr(current), current->comm);
2178 printk(KERN_WARNING "Skipping core dump\n");
2179 goto fail_dropcount;
2180 }
2181
Xiaotian Feng1b0d3002010-10-27 15:34:08 -07002182 helper_argv = argv_split(GFP_KERNEL, cn.corename+1, NULL);
Tetsuo Handa350eaf72009-01-06 14:41:11 -08002183 if (!helper_argv) {
2184 printk(KERN_WARNING "%s failed to allocate memory\n",
2185 __func__);
Neil Hormana2939802009-09-23 15:56:56 -07002186 goto fail_dropcount;
Tetsuo Handa350eaf72009-01-06 14:41:11 -08002187 }
Neil Horman32321132007-10-16 23:26:36 -07002188
Oleg Nesterovd5bf4c42010-05-26 14:43:06 -07002189 retval = call_usermodehelper_fns(helper_argv[0], helper_argv,
2190 NULL, UMH_WAIT_EXEC, umh_pipe_setup,
2191 NULL, &cprm);
2192 argv_free(helper_argv);
2193 if (retval) {
Andi Kleend025c9d2006-09-30 23:29:28 -07002194 printk(KERN_INFO "Core dump to %s pipe failed\n",
Xiaotian Feng1b0d3002010-10-27 15:34:08 -07002195 cn.corename);
Oleg Nesterovd5bf4c42010-05-26 14:43:06 -07002196 goto close_fail;
Andi Kleend025c9d2006-09-30 23:29:28 -07002197 }
Oleg Nesterovc7135412010-05-26 14:43:05 -07002198 } else {
2199 struct inode *inode;
2200
2201 if (cprm.limit < binfmt->min_coredump)
2202 goto fail_unlock;
2203
Xiaotian Feng1b0d3002010-10-27 15:34:08 -07002204 cprm.file = filp_open(cn.corename,
Alexey Dobriyan6d4df672006-12-06 20:40:39 -08002205 O_CREAT | 2 | O_NOFOLLOW | O_LARGEFILE | flag,
2206 0600);
Oleg Nesterovc7135412010-05-26 14:43:05 -07002207 if (IS_ERR(cprm.file))
2208 goto fail_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002209
Oleg Nesterovc7135412010-05-26 14:43:05 -07002210 inode = cprm.file->f_path.dentry->d_inode;
2211 if (inode->i_nlink > 1)
2212 goto close_fail;
2213 if (d_unhashed(cprm.file->f_path.dentry))
2214 goto close_fail;
2215 /*
2216 * AK: actually i see no reason to not allow this for named
2217 * pipes etc, but keep the previous behaviour for now.
2218 */
2219 if (!S_ISREG(inode->i_mode))
2220 goto close_fail;
2221 /*
2222 * Dont allow local users get cute and trick others to coredump
2223 * into their pre-created files.
2224 */
2225 if (inode->i_uid != current_fsuid())
2226 goto close_fail;
2227 if (!cprm.file->f_op || !cprm.file->f_op->write)
2228 goto close_fail;
2229 if (do_truncate(cprm.file->f_path.dentry, 0, 0, cprm.file))
2230 goto close_fail;
2231 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002232
Masami Hiramatsuf6151df2009-12-17 15:27:16 -08002233 retval = binfmt->core_dump(&cprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002234 if (retval)
2235 current->signal->group_exit_code |= 0x80;
Oleg Nesterovd5bf4c42010-05-26 14:43:06 -07002236
Neil Horman61be2282009-09-23 15:56:58 -07002237 if (ispipe && core_pipe_limit)
Masami Hiramatsuf6151df2009-12-17 15:27:16 -08002238 wait_for_dump_helpers(cprm.file);
Oleg Nesterovd5bf4c42010-05-26 14:43:06 -07002239close_fail:
2240 if (cprm.file)
2241 filp_close(cprm.file, NULL);
Neil Hormana2939802009-09-23 15:56:56 -07002242fail_dropcount:
Oleg Nesterovd5bf4c42010-05-26 14:43:06 -07002243 if (ispipe)
Neil Hormana2939802009-09-23 15:56:56 -07002244 atomic_dec(&core_dump_count);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002245fail_unlock:
Xiaotian Feng1b0d3002010-10-27 15:34:08 -07002246 kfree(cn.corename);
2247fail_corename:
Oleg Nesterova94e2d42008-07-25 01:47:46 -07002248 coredump_finish(mm);
Oleg Nesterov5e43aef2010-05-26 14:43:07 -07002249 revert_creds(old_cred);
2250fail_creds:
2251 put_cred(cred);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002252fail:
WANG Cong8cd3ac32009-01-06 14:42:48 -08002253 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002254}
Linus Torvalds3aa0ce82010-10-14 14:32:06 -07002255
2256/*
2257 * Core dumping helper functions. These are the only things you should
2258 * do on a core-file: use only these functions to write out all the
2259 * necessary info.
2260 */
2261int dump_write(struct file *file, const void *addr, int nr)
2262{
2263 return access_ok(VERIFY_READ, addr, nr) && file->f_op->write(file, addr, nr, &file->f_pos) == nr;
2264}
Linus Torvalds8fd01d62010-10-14 19:15:28 -07002265EXPORT_SYMBOL(dump_write);
Linus Torvalds3aa0ce82010-10-14 14:32:06 -07002266
2267int dump_seek(struct file *file, loff_t off)
2268{
2269 int ret = 1;
2270
2271 if (file->f_op->llseek && file->f_op->llseek != no_llseek) {
2272 if (file->f_op->llseek(file, off, SEEK_CUR) < 0)
2273 return 0;
2274 } else {
2275 char *buf = (char *)get_zeroed_page(GFP_KERNEL);
2276
2277 if (!buf)
2278 return 0;
2279 while (off > 0) {
2280 unsigned long n = off;
2281
2282 if (n > PAGE_SIZE)
2283 n = PAGE_SIZE;
2284 if (!dump_write(file, buf, n)) {
2285 ret = 0;
2286 break;
2287 }
2288 off -= n;
2289 }
2290 free_page((unsigned long)buf);
2291 }
2292 return ret;
2293}
Linus Torvalds8fd01d62010-10-14 19:15:28 -07002294EXPORT_SYMBOL(dump_seek);