blob: d8e1191cb112eb2d77154908a32699a06c4bda78 [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>
KAMEZAWA Hiroyuki43d2b112012-01-10 15:08:09 -080062
63#include <trace/events/task.h>
David Howellsa6f76f22008-11-14 10:39:24 +110064#include "internal.h"
Alex Kelly179899f2012-10-04 17:15:24 -070065#include "coredump.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070066
David Smith4ff16c22012-02-07 10:11:05 -060067#include <trace/events/sched.h>
68
Alan Coxd6e71142005-06-23 00:09:43 -070069int suid_dumpable = 0;
70
Alexey Dobriyane4dc1b12007-10-16 23:26:03 -070071static LIST_HEAD(formats);
Linus Torvalds1da177e2005-04-16 15:20:36 -070072static DEFINE_RWLOCK(binfmt_lock);
73
Al Viro8fc3dc52012-03-17 03:05:16 -040074void __register_binfmt(struct linux_binfmt * fmt, int insert)
Linus Torvalds1da177e2005-04-16 15:20:36 -070075{
Al Viro8fc3dc52012-03-17 03:05:16 -040076 BUG_ON(!fmt);
Linus Torvalds1da177e2005-04-16 15:20:36 -070077 write_lock(&binfmt_lock);
Ivan Kokshaysky74641f52009-04-30 15:08:49 -070078 insert ? list_add(&fmt->lh, &formats) :
79 list_add_tail(&fmt->lh, &formats);
Linus Torvalds1da177e2005-04-16 15:20:36 -070080 write_unlock(&binfmt_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -070081}
82
Ivan Kokshaysky74641f52009-04-30 15:08:49 -070083EXPORT_SYMBOL(__register_binfmt);
Linus Torvalds1da177e2005-04-16 15:20:36 -070084
Alexey Dobriyanf6b450d2007-10-16 23:26:04 -070085void unregister_binfmt(struct linux_binfmt * fmt)
Linus Torvalds1da177e2005-04-16 15:20:36 -070086{
Linus Torvalds1da177e2005-04-16 15:20:36 -070087 write_lock(&binfmt_lock);
Alexey Dobriyane4dc1b12007-10-16 23:26:03 -070088 list_del(&fmt->lh);
Linus Torvalds1da177e2005-04-16 15:20:36 -070089 write_unlock(&binfmt_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -070090}
91
92EXPORT_SYMBOL(unregister_binfmt);
93
94static inline void put_binfmt(struct linux_binfmt * fmt)
95{
96 module_put(fmt->module);
97}
98
99/*
100 * Note that a shared library must be both readable and executable due to
101 * security reasons.
102 *
103 * Also note that we take the address to load from from the file itself.
104 */
Heiko Carstens1e7bfb22009-01-14 14:14:29 +0100105SYSCALL_DEFINE1(uselib, const char __user *, library)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106{
Al Viro964bd182008-07-26 03:33:14 -0400107 struct file *file;
Jeff Layton91a27b22012-10-10 15:25:28 -0400108 struct filename *tmp = getname(library);
Al Viro964bd182008-07-26 03:33:14 -0400109 int error = PTR_ERR(tmp);
Al Viro47c805d2011-02-23 17:44:09 -0500110 static const struct open_flags uselib_flags = {
111 .open_flag = O_LARGEFILE | O_RDONLY | __FMODE_EXEC,
112 .acc_mode = MAY_READ | MAY_EXEC | MAY_OPEN,
113 .intent = LOOKUP_OPEN
114 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115
Al Viro6e8341a2009-04-06 11:16:22 -0400116 if (IS_ERR(tmp))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117 goto out;
118
Jeff Layton669abf42012-10-10 16:43:10 -0400119 file = do_filp_open(AT_FDCWD, tmp, &uselib_flags, LOOKUP_FOLLOW);
Al Viro6e8341a2009-04-06 11:16:22 -0400120 putname(tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700121 error = PTR_ERR(file);
122 if (IS_ERR(file))
123 goto out;
124
Al Viro6e8341a2009-04-06 11:16:22 -0400125 error = -EINVAL;
126 if (!S_ISREG(file->f_path.dentry->d_inode->i_mode))
127 goto exit;
128
129 error = -EACCES;
130 if (file->f_path.mnt->mnt_flags & MNT_NOEXEC)
131 goto exit;
132
Eric Paris2a12a9d2009-12-17 21:24:21 -0500133 fsnotify_open(file);
Eric Paris6110e3a2008-12-17 13:53:20 -0500134
Linus Torvalds1da177e2005-04-16 15:20:36 -0700135 error = -ENOEXEC;
136 if(file->f_op) {
137 struct linux_binfmt * fmt;
138
139 read_lock(&binfmt_lock);
Alexey Dobriyane4dc1b12007-10-16 23:26:03 -0700140 list_for_each_entry(fmt, &formats, lh) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700141 if (!fmt->load_shlib)
142 continue;
143 if (!try_module_get(fmt->module))
144 continue;
145 read_unlock(&binfmt_lock);
146 error = fmt->load_shlib(file);
147 read_lock(&binfmt_lock);
148 put_binfmt(fmt);
149 if (error != -ENOEXEC)
150 break;
151 }
152 read_unlock(&binfmt_lock);
153 }
Al Viro6e8341a2009-04-06 11:16:22 -0400154exit:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155 fput(file);
156out:
157 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158}
159
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700160#ifdef CONFIG_MMU
Oleg Nesterovae6b5852011-03-06 18:03:11 +0100161/*
162 * The nascent bprm->mm is not visible until exec_mmap() but it can
163 * use a lot of memory, account these pages in current->mm temporary
164 * for oom_badness()->get_mm_rss(). Once exec succeeds or fails, we
165 * change the counter back via acct_arg_size(0).
166 */
Oleg Nesterov0e028462011-03-06 18:02:54 +0100167static void acct_arg_size(struct linux_binprm *bprm, unsigned long pages)
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100168{
169 struct mm_struct *mm = current->mm;
170 long diff = (long)(pages - bprm->vma_pages);
171
172 if (!mm || !diff)
173 return;
174
175 bprm->vma_pages = pages;
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100176 add_mm_counter(mm, MM_ANONPAGES, diff);
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100177}
178
Oleg Nesterov0e028462011-03-06 18:02:54 +0100179static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos,
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700180 int write)
181{
182 struct page *page;
183 int ret;
184
185#ifdef CONFIG_STACK_GROWSUP
186 if (write) {
Michal Hockod05f3162011-05-24 17:11:44 -0700187 ret = expand_downwards(bprm->vma, pos);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700188 if (ret < 0)
189 return NULL;
190 }
191#endif
192 ret = get_user_pages(current, bprm->mm, pos,
193 1, write, 1, &page, NULL);
194 if (ret <= 0)
195 return NULL;
196
197 if (write) {
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700198 unsigned long size = bprm->vma->vm_end - bprm->vma->vm_start;
Linus Torvaldsa64e7152008-03-03 10:12:14 -0800199 struct rlimit *rlim;
200
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100201 acct_arg_size(bprm, size / PAGE_SIZE);
202
Linus Torvaldsa64e7152008-03-03 10:12:14 -0800203 /*
204 * We've historically supported up to 32 pages (ARG_MAX)
205 * of argument strings even with small stacks
206 */
207 if (size <= ARG_MAX)
208 return page;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700209
210 /*
211 * Limit to 1/4-th the stack size for the argv+env strings.
212 * This ensures that:
213 * - the remaining binfmt code will not run out of stack space,
214 * - the program will have a reasonable amount of stack left
215 * to work from.
216 */
Linus Torvaldsa64e7152008-03-03 10:12:14 -0800217 rlim = current->signal->rlim;
Jiri Slabyd554ed892010-03-05 13:42:42 -0800218 if (size > ACCESS_ONCE(rlim[RLIMIT_STACK].rlim_cur) / 4) {
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700219 put_page(page);
220 return NULL;
221 }
222 }
223
224 return page;
225}
226
227static void put_arg_page(struct page *page)
228{
229 put_page(page);
230}
231
232static void free_arg_page(struct linux_binprm *bprm, int i)
233{
234}
235
236static void free_arg_pages(struct linux_binprm *bprm)
237{
238}
239
240static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos,
241 struct page *page)
242{
243 flush_cache_page(bprm->vma, pos, page_to_pfn(page));
244}
245
246static int __bprm_mm_init(struct linux_binprm *bprm)
247{
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800248 int err;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700249 struct vm_area_struct *vma = NULL;
250 struct mm_struct *mm = bprm->mm;
251
252 bprm->vma = vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL);
253 if (!vma)
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800254 return -ENOMEM;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700255
256 down_write(&mm->mmap_sem);
257 vma->vm_mm = mm;
258
259 /*
260 * Place the stack at the largest stack address the architecture
261 * supports. Later, we'll move this to an appropriate place. We don't
262 * use STACK_TOP because that can depend on attributes which aren't
263 * configured yet.
264 */
Michal Hockoaacb3d12011-07-26 16:08:40 -0700265 BUILD_BUG_ON(VM_STACK_FLAGS & VM_STACK_INCOMPLETE_SETUP);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700266 vma->vm_end = STACK_TOP_MAX;
267 vma->vm_start = vma->vm_end - PAGE_SIZE;
Mel Gormana8bef8ff2010-05-24 14:32:24 -0700268 vma->vm_flags = VM_STACK_FLAGS | VM_STACK_INCOMPLETE_SETUP;
Coly Li3ed75eb2007-10-18 23:39:15 -0700269 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
Rik van Riel5beb4932010-03-05 13:42:07 -0800270 INIT_LIST_HEAD(&vma->anon_vma_chain);
Tavis Ormandy462e635e2010-12-09 15:29:42 +0100271
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700272 err = insert_vm_struct(mm, vma);
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800273 if (err)
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700274 goto err;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700275
276 mm->stack_vm = mm->total_vm = 1;
277 up_write(&mm->mmap_sem);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700278 bprm->p = vma->vm_end - sizeof(void *);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700279 return 0;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700280err:
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800281 up_write(&mm->mmap_sem);
282 bprm->vma = NULL;
283 kmem_cache_free(vm_area_cachep, vma);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700284 return err;
285}
286
287static bool valid_arg_len(struct linux_binprm *bprm, long len)
288{
289 return len <= MAX_ARG_STRLEN;
290}
291
292#else
293
Oleg Nesterov0e028462011-03-06 18:02:54 +0100294static inline void acct_arg_size(struct linux_binprm *bprm, unsigned long pages)
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100295{
296}
297
Oleg Nesterov0e028462011-03-06 18:02:54 +0100298static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos,
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700299 int write)
300{
301 struct page *page;
302
303 page = bprm->page[pos / PAGE_SIZE];
304 if (!page && write) {
305 page = alloc_page(GFP_HIGHUSER|__GFP_ZERO);
306 if (!page)
307 return NULL;
308 bprm->page[pos / PAGE_SIZE] = page;
309 }
310
311 return page;
312}
313
314static void put_arg_page(struct page *page)
315{
316}
317
318static void free_arg_page(struct linux_binprm *bprm, int i)
319{
320 if (bprm->page[i]) {
321 __free_page(bprm->page[i]);
322 bprm->page[i] = NULL;
323 }
324}
325
326static void free_arg_pages(struct linux_binprm *bprm)
327{
328 int i;
329
330 for (i = 0; i < MAX_ARG_PAGES; i++)
331 free_arg_page(bprm, i);
332}
333
334static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos,
335 struct page *page)
336{
337}
338
339static int __bprm_mm_init(struct linux_binprm *bprm)
340{
341 bprm->p = PAGE_SIZE * MAX_ARG_PAGES - sizeof(void *);
342 return 0;
343}
344
345static bool valid_arg_len(struct linux_binprm *bprm, long len)
346{
347 return len <= bprm->p;
348}
349
350#endif /* CONFIG_MMU */
351
352/*
353 * Create a new mm_struct and populate it with a temporary stack
354 * vm_area_struct. We don't have enough context at this point to set the stack
355 * flags, permissions, and offset, so we use temporary values. We'll update
356 * them later in setup_arg_pages().
357 */
358int bprm_mm_init(struct linux_binprm *bprm)
359{
360 int err;
361 struct mm_struct *mm = NULL;
362
363 bprm->mm = mm = mm_alloc();
364 err = -ENOMEM;
365 if (!mm)
366 goto err;
367
368 err = init_new_context(current, mm);
369 if (err)
370 goto err;
371
372 err = __bprm_mm_init(bprm);
373 if (err)
374 goto err;
375
376 return 0;
377
378err:
379 if (mm) {
380 bprm->mm = NULL;
381 mmdrop(mm);
382 }
383
384 return err;
385}
386
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100387struct user_arg_ptr {
Oleg Nesterov0e028462011-03-06 18:02:54 +0100388#ifdef CONFIG_COMPAT
389 bool is_compat;
390#endif
391 union {
392 const char __user *const __user *native;
393#ifdef CONFIG_COMPAT
Al Viro38b983b2012-09-30 13:38:55 -0400394 const compat_uptr_t __user *compat;
Oleg Nesterov0e028462011-03-06 18:02:54 +0100395#endif
396 } ptr;
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100397};
398
399static const char __user *get_user_arg_ptr(struct user_arg_ptr argv, int nr)
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100400{
Oleg Nesterov0e028462011-03-06 18:02:54 +0100401 const char __user *native;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100402
Oleg Nesterov0e028462011-03-06 18:02:54 +0100403#ifdef CONFIG_COMPAT
404 if (unlikely(argv.is_compat)) {
405 compat_uptr_t compat;
406
407 if (get_user(compat, argv.ptr.compat + nr))
408 return ERR_PTR(-EFAULT);
409
410 return compat_ptr(compat);
411 }
412#endif
413
414 if (get_user(native, argv.ptr.native + nr))
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100415 return ERR_PTR(-EFAULT);
416
Oleg Nesterov0e028462011-03-06 18:02:54 +0100417 return native;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100418}
419
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420/*
421 * count() counts the number of strings in array ARGV.
422 */
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100423static int count(struct user_arg_ptr argv, int max)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424{
425 int i = 0;
426
Oleg Nesterov0e028462011-03-06 18:02:54 +0100427 if (argv.ptr.native != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700428 for (;;) {
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100429 const char __user *p = get_user_arg_ptr(argv, i);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430
Linus Torvalds1da177e2005-04-16 15:20:36 -0700431 if (!p)
432 break;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100433
434 if (IS_ERR(p))
435 return -EFAULT;
436
Jason Baron362e6662008-10-15 22:01:52 -0700437 if (i++ >= max)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700438 return -E2BIG;
Roland McGrath9aea5a62010-09-07 19:37:06 -0700439
440 if (fatal_signal_pending(current))
441 return -ERESTARTNOHAND;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442 cond_resched();
443 }
444 }
445 return i;
446}
447
448/*
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700449 * 'copy_strings()' copies argument/environment strings from the old
450 * processes's memory to the new process's stack. The call to get_user_pages()
451 * ensures the destination page is created and not swapped out.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452 */
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100453static int copy_strings(int argc, struct user_arg_ptr argv,
Adrian Bunk75c96f82005-05-05 16:16:09 -0700454 struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700455{
456 struct page *kmapped_page = NULL;
457 char *kaddr = NULL;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700458 unsigned long kpos = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700459 int ret;
460
461 while (argc-- > 0) {
David Howellsd7627462010-08-17 23:52:56 +0100462 const char __user *str;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463 int len;
464 unsigned long pos;
465
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100466 ret = -EFAULT;
467 str = get_user_arg_ptr(argv, argc);
468 if (IS_ERR(str))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100471 len = strnlen_user(str, MAX_ARG_STRLEN);
472 if (!len)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473 goto out;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100474
475 ret = -E2BIG;
476 if (!valid_arg_len(bprm, len))
477 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700478
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700479 /* We're going to work our way backwords. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480 pos = bprm->p;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700481 str += len;
482 bprm->p -= len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483
484 while (len > 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485 int offset, bytes_to_copy;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486
Roland McGrath9aea5a62010-09-07 19:37:06 -0700487 if (fatal_signal_pending(current)) {
488 ret = -ERESTARTNOHAND;
489 goto out;
490 }
Roland McGrath7993bc12010-09-07 19:36:28 -0700491 cond_resched();
492
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493 offset = pos % PAGE_SIZE;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700494 if (offset == 0)
495 offset = PAGE_SIZE;
496
497 bytes_to_copy = offset;
498 if (bytes_to_copy > len)
499 bytes_to_copy = len;
500
501 offset -= bytes_to_copy;
502 pos -= bytes_to_copy;
503 str -= bytes_to_copy;
504 len -= bytes_to_copy;
505
506 if (!kmapped_page || kpos != (pos & PAGE_MASK)) {
507 struct page *page;
508
509 page = get_arg_page(bprm, pos, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510 if (!page) {
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700511 ret = -E2BIG;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512 goto out;
513 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700515 if (kmapped_page) {
516 flush_kernel_dcache_page(kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517 kunmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700518 put_arg_page(kmapped_page);
519 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700520 kmapped_page = page;
521 kaddr = kmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700522 kpos = pos & PAGE_MASK;
523 flush_arg_page(bprm, kpos, kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700524 }
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700525 if (copy_from_user(kaddr+offset, str, bytes_to_copy)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700526 ret = -EFAULT;
527 goto out;
528 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529 }
530 }
531 ret = 0;
532out:
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700533 if (kmapped_page) {
534 flush_kernel_dcache_page(kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535 kunmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700536 put_arg_page(kmapped_page);
537 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538 return ret;
539}
540
541/*
542 * Like copy_strings, but get argv and its values from kernel memory.
543 */
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100544int copy_strings_kernel(int argc, const char *const *__argv,
David Howellsd7627462010-08-17 23:52:56 +0100545 struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700546{
547 int r;
548 mm_segment_t oldfs = get_fs();
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100549 struct user_arg_ptr argv = {
Oleg Nesterov0e028462011-03-06 18:02:54 +0100550 .ptr.native = (const char __user *const __user *)__argv,
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100551 };
552
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553 set_fs(KERNEL_DS);
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100554 r = copy_strings(argc, argv, bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700555 set_fs(oldfs);
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100556
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557 return r;
558}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700559EXPORT_SYMBOL(copy_strings_kernel);
560
561#ifdef CONFIG_MMU
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700562
Linus Torvalds1da177e2005-04-16 15:20:36 -0700563/*
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700564 * During bprm_mm_init(), we create a temporary stack at STACK_TOP_MAX. Once
565 * the binfmt code determines where the new stack should reside, we shift it to
566 * its final location. The process proceeds as follows:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567 *
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700568 * 1) Use shift to calculate the new vma endpoints.
569 * 2) Extend vma to cover both the old and new ranges. This ensures the
570 * arguments passed to subsequent functions are consistent.
571 * 3) Move vma's page tables to the new range.
572 * 4) Free up any cleared pgd range.
573 * 5) Shrink the vma to cover only the new range.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700574 */
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700575static int shift_arg_pages(struct vm_area_struct *vma, unsigned long shift)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576{
577 struct mm_struct *mm = vma->vm_mm;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700578 unsigned long old_start = vma->vm_start;
579 unsigned long old_end = vma->vm_end;
580 unsigned long length = old_end - old_start;
581 unsigned long new_start = old_start - shift;
582 unsigned long new_end = old_end - shift;
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700583 struct mmu_gather tlb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700585 BUG_ON(new_start > new_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700586
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700587 /*
588 * ensure there are no vmas between where we want to go
589 * and where we are
590 */
591 if (vma != find_vma(mm, new_start))
592 return -EFAULT;
593
594 /*
595 * cover the whole range: [new_start, old_end)
596 */
Rik van Riel5beb4932010-03-05 13:42:07 -0800597 if (vma_adjust(vma, new_start, old_end, vma->vm_pgoff, NULL))
598 return -ENOMEM;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700599
600 /*
601 * move the page tables downwards, on failure we rely on
602 * process cleanup to remove whatever mess we made.
603 */
604 if (length != move_page_tables(vma, old_start,
Michel Lespinasse38a76012012-10-08 16:31:50 -0700605 vma, new_start, length, false))
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700606 return -ENOMEM;
607
608 lru_add_drain();
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700609 tlb_gather_mmu(&tlb, mm, 0);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700610 if (new_end > old_start) {
611 /*
612 * when the old and new regions overlap clear from new_end.
613 */
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700614 free_pgd_range(&tlb, new_end, old_end, new_end,
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700615 vma->vm_next ? vma->vm_next->vm_start : 0);
616 } else {
617 /*
618 * otherwise, clean from old_start; this is done to not touch
619 * the address space in [new_end, old_start) some architectures
620 * have constraints on va-space that make this illegal (IA64) -
621 * for the others its just a little faster.
622 */
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700623 free_pgd_range(&tlb, old_start, old_end, new_end,
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700624 vma->vm_next ? vma->vm_next->vm_start : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700625 }
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700626 tlb_finish_mmu(&tlb, new_end, old_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700628 /*
Rik van Riel5beb4932010-03-05 13:42:07 -0800629 * Shrink the vma to just the new range. Always succeeds.
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700630 */
631 vma_adjust(vma, new_start, new_end, vma->vm_pgoff, NULL);
632
633 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634}
635
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700636/*
637 * Finalizes the stack vm_area_struct. The flags and permissions are updated,
638 * the stack is optionally relocated, and some extra space is added.
639 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700640int setup_arg_pages(struct linux_binprm *bprm,
641 unsigned long stack_top,
642 int executable_stack)
643{
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700644 unsigned long ret;
645 unsigned long stack_shift;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700646 struct mm_struct *mm = current->mm;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700647 struct vm_area_struct *vma = bprm->vma;
648 struct vm_area_struct *prev = NULL;
649 unsigned long vm_flags;
650 unsigned long stack_base;
Michael Neuling803bf5e2010-02-10 13:56:42 -0800651 unsigned long stack_size;
652 unsigned long stack_expand;
653 unsigned long rlim_stack;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654
655#ifdef CONFIG_STACK_GROWSUP
Linus Torvalds1da177e2005-04-16 15:20:36 -0700656 /* Limit stack size to 1GB */
Jiri Slabyd554ed892010-03-05 13:42:42 -0800657 stack_base = rlimit_max(RLIMIT_STACK);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658 if (stack_base > (1 << 30))
659 stack_base = 1 << 30;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700660
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700661 /* Make sure we didn't let the argument array grow too large. */
662 if (vma->vm_end - vma->vm_start > stack_base)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663 return -ENOMEM;
664
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700665 stack_base = PAGE_ALIGN(stack_top - stack_base);
666
667 stack_shift = vma->vm_start - stack_base;
668 mm->arg_start = bprm->p - stack_shift;
669 bprm->p = vma->vm_end - stack_shift;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700670#else
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700671 stack_top = arch_align_stack(stack_top);
672 stack_top = PAGE_ALIGN(stack_top);
Roland McGrath1b528182010-09-07 19:35:49 -0700673
674 if (unlikely(stack_top < mmap_min_addr) ||
675 unlikely(vma->vm_end - vma->vm_start >= stack_top - mmap_min_addr))
676 return -ENOMEM;
677
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700678 stack_shift = vma->vm_end - stack_top;
679
680 bprm->p -= stack_shift;
681 mm->arg_start = bprm->p;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700682#endif
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700683
684 if (bprm->loader)
685 bprm->loader -= stack_shift;
686 bprm->exec -= stack_shift;
687
688 down_write(&mm->mmap_sem);
Hugh Dickins96a8e132008-07-10 21:19:20 +0100689 vm_flags = VM_STACK_FLAGS;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700690
691 /*
692 * Adjust stack execute permissions; explicitly enable for
693 * EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X and leave alone
694 * (arch default) otherwise.
695 */
696 if (unlikely(executable_stack == EXSTACK_ENABLE_X))
697 vm_flags |= VM_EXEC;
698 else if (executable_stack == EXSTACK_DISABLE_X)
699 vm_flags &= ~VM_EXEC;
700 vm_flags |= mm->def_flags;
Mel Gormana8bef8ff2010-05-24 14:32:24 -0700701 vm_flags |= VM_STACK_INCOMPLETE_SETUP;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700702
703 ret = mprotect_fixup(vma, &prev, vma->vm_start, vma->vm_end,
704 vm_flags);
705 if (ret)
706 goto out_unlock;
707 BUG_ON(prev != vma);
708
709 /* Move stack pages down in memory. */
710 if (stack_shift) {
711 ret = shift_arg_pages(vma, stack_shift);
Anton Blanchardfc63cf22009-11-11 14:26:48 -0800712 if (ret)
713 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700714 }
715
Mel Gormana8bef8ff2010-05-24 14:32:24 -0700716 /* mprotect_fixup is overkill to remove the temporary stack flags */
717 vma->vm_flags &= ~VM_STACK_INCOMPLETE_SETUP;
718
Michael Neuling5ef097d2010-03-05 13:42:57 -0800719 stack_expand = 131072UL; /* randomly 32*4k (or 2*64k) pages */
Michael Neuling803bf5e2010-02-10 13:56:42 -0800720 stack_size = vma->vm_end - vma->vm_start;
721 /*
722 * Align this down to a page boundary as expand_stack
723 * will align it up.
724 */
725 rlim_stack = rlimit(RLIMIT_STACK) & PAGE_MASK;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700726#ifdef CONFIG_STACK_GROWSUP
Michael Neuling803bf5e2010-02-10 13:56:42 -0800727 if (stack_size + stack_expand > rlim_stack)
728 stack_base = vma->vm_start + rlim_stack;
729 else
730 stack_base = vma->vm_end + stack_expand;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700731#else
Michael Neuling803bf5e2010-02-10 13:56:42 -0800732 if (stack_size + stack_expand > rlim_stack)
733 stack_base = vma->vm_end - rlim_stack;
734 else
735 stack_base = vma->vm_start - stack_expand;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700736#endif
Eric B Munson3af9e852010-05-18 15:30:49 +0100737 current->mm->start_stack = bprm->p;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700738 ret = expand_stack(vma, stack_base);
739 if (ret)
740 ret = -EFAULT;
741
742out_unlock:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743 up_write(&mm->mmap_sem);
Anton Blanchardfc63cf22009-11-11 14:26:48 -0800744 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746EXPORT_SYMBOL(setup_arg_pages);
747
Linus Torvalds1da177e2005-04-16 15:20:36 -0700748#endif /* CONFIG_MMU */
749
750struct file *open_exec(const char *name)
751{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752 struct file *file;
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200753 int err;
Jeff Layton669abf42012-10-10 16:43:10 -0400754 struct filename tmp = { .name = name };
Al Viro47c805d2011-02-23 17:44:09 -0500755 static const struct open_flags open_exec_flags = {
756 .open_flag = O_LARGEFILE | O_RDONLY | __FMODE_EXEC,
757 .acc_mode = MAY_EXEC | MAY_OPEN,
758 .intent = LOOKUP_OPEN
759 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700760
Jeff Layton669abf42012-10-10 16:43:10 -0400761 file = do_filp_open(AT_FDCWD, &tmp, &open_exec_flags, LOOKUP_FOLLOW);
Al Viro6e8341a2009-04-06 11:16:22 -0400762 if (IS_ERR(file))
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200763 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200765 err = -EACCES;
Al Viro6e8341a2009-04-06 11:16:22 -0400766 if (!S_ISREG(file->f_path.dentry->d_inode->i_mode))
767 goto exit;
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200768
Al Viro6e8341a2009-04-06 11:16:22 -0400769 if (file->f_path.mnt->mnt_flags & MNT_NOEXEC)
770 goto exit;
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200771
Eric Paris2a12a9d2009-12-17 21:24:21 -0500772 fsnotify_open(file);
Eric Paris6110e3a2008-12-17 13:53:20 -0500773
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200774 err = deny_write_access(file);
Al Viro6e8341a2009-04-06 11:16:22 -0400775 if (err)
776 goto exit;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700777
Al Viro6e8341a2009-04-06 11:16:22 -0400778out:
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200779 return file;
780
Al Viro6e8341a2009-04-06 11:16:22 -0400781exit:
782 fput(file);
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200783 return ERR_PTR(err);
784}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700785EXPORT_SYMBOL(open_exec);
786
Mimi Zohar6777d772009-08-21 14:32:48 -0400787int kernel_read(struct file *file, loff_t offset,
788 char *addr, unsigned long count)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789{
790 mm_segment_t old_fs;
791 loff_t pos = offset;
792 int result;
793
794 old_fs = get_fs();
795 set_fs(get_ds());
796 /* The cast to a user pointer is valid due to the set_fs() */
797 result = vfs_read(file, (void __user *)addr, count, &pos);
798 set_fs(old_fs);
799 return result;
800}
801
802EXPORT_SYMBOL(kernel_read);
803
804static int exec_mmap(struct mm_struct *mm)
805{
806 struct task_struct *tsk;
807 struct mm_struct * old_mm, *active_mm;
808
809 /* Notify parent that we're no longer interested in the old VM */
810 tsk = current;
811 old_mm = current->mm;
812 mm_release(tsk, old_mm);
813
814 if (old_mm) {
Konstantin Khlebnikov4fe7efd2012-06-20 12:53:01 -0700815 sync_mm_rss(old_mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700816 /*
817 * Make sure that if there is a core dump in progress
818 * for the old mm, we get out and die instead of going
819 * through with the exec. We must hold mmap_sem around
Oleg Nesterov999d9fc2008-07-25 01:47:41 -0700820 * checking core_state and changing tsk->mm.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700821 */
822 down_read(&old_mm->mmap_sem);
Oleg Nesterov999d9fc2008-07-25 01:47:41 -0700823 if (unlikely(old_mm->core_state)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700824 up_read(&old_mm->mmap_sem);
825 return -EINTR;
826 }
827 }
828 task_lock(tsk);
829 active_mm = tsk->active_mm;
830 tsk->mm = mm;
831 tsk->active_mm = mm;
832 activate_mm(active_mm, mm);
833 task_unlock(tsk);
834 arch_pick_mmap_layout(mm);
835 if (old_mm) {
836 up_read(&old_mm->mmap_sem);
Eric Sesterhenn7dddb122006-04-01 01:13:38 +0200837 BUG_ON(active_mm != old_mm);
Oleg Nesterov701085b2012-03-19 17:04:01 +0100838 setmax_mm_hiwater_rss(&tsk->signal->maxrss, old_mm);
Balbir Singh31a78f22008-09-28 23:09:31 +0100839 mm_update_next_owner(old_mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700840 mmput(old_mm);
841 return 0;
842 }
843 mmdrop(active_mm);
844 return 0;
845}
846
847/*
848 * This function makes sure the current process has its own signal table,
849 * so that flush_signal_handlers can later reset the handlers without
850 * disturbing other processes. (Other processes might share the signal
851 * table via the CLONE_SIGHAND option to clone().)
852 */
Arjan van de Ven858119e2006-01-14 13:20:43 -0800853static int de_thread(struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700854{
855 struct signal_struct *sig = tsk->signal;
Oleg Nesterovb2c903b2007-10-16 23:27:22 -0700856 struct sighand_struct *oldsighand = tsk->sighand;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700857 spinlock_t *lock = &oldsighand->siglock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700859 if (thread_group_empty(tsk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860 goto no_thread_group;
861
862 /*
863 * Kill all other threads in the thread group.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865 spin_lock_irq(lock);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -0800866 if (signal_group_exit(sig)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700867 /*
868 * Another group action in progress, just
869 * return so that the signal is processed.
870 */
871 spin_unlock_irq(lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700872 return -EAGAIN;
873 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874
Oleg Nesterovd3441932010-05-26 14:43:11 -0700875 sig->group_exit_task = tsk;
876 sig->notify_count = zap_other_threads(tsk);
877 if (!thread_group_leader(tsk))
878 sig->notify_count--;
879
880 while (sig->notify_count) {
Oleg Nesterovd5bbd432012-10-08 19:13:01 +0200881 __set_current_state(TASK_KILLABLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700882 spin_unlock_irq(lock);
883 schedule();
Oleg Nesterovd5bbd432012-10-08 19:13:01 +0200884 if (unlikely(__fatal_signal_pending(tsk)))
885 goto killed;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886 spin_lock_irq(lock);
887 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888 spin_unlock_irq(lock);
889
890 /*
891 * At this point all other threads have exited, all we have to
892 * do is to wait for the thread group leader to become inactive,
893 * and to assume its PID:
894 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700895 if (!thread_group_leader(tsk)) {
Oleg Nesterov81879262008-12-01 14:18:16 -0800896 struct task_struct *leader = tsk->group_leader;
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700897
Oleg Nesterov2800d8d2008-04-30 00:53:12 -0700898 sig->notify_count = -1; /* for exit_notify() */
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700899 for (;;) {
900 write_lock_irq(&tasklist_lock);
901 if (likely(leader->exit_state))
902 break;
Oleg Nesterovd5bbd432012-10-08 19:13:01 +0200903 __set_current_state(TASK_KILLABLE);
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700904 write_unlock_irq(&tasklist_lock);
905 schedule();
Oleg Nesterovd5bbd432012-10-08 19:13:01 +0200906 if (unlikely(__fatal_signal_pending(tsk)))
907 goto killed;
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700908 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700909
Roland McGrathf5e90282006-04-10 22:54:16 -0700910 /*
911 * The only record we have of the real-time age of a
912 * process, regardless of execs it's done, is start_time.
913 * All the past CPU time is accumulated in signal_struct
914 * from sister threads now dead. But in this non-leader
915 * exec, nothing survives from the original leader thread,
916 * whose birth marks the true age of this process now.
917 * When we take on its identity by switching to its PID, we
918 * also take its birthdate (always earlier than our own).
919 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700920 tsk->start_time = leader->start_time;
Roland McGrathf5e90282006-04-10 22:54:16 -0700921
Pavel Emelyanovbac0abd2007-10-18 23:40:18 -0700922 BUG_ON(!same_thread_group(leader, tsk));
923 BUG_ON(has_group_leader_pid(tsk));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924 /*
925 * An exec() starts a new thread group with the
926 * TGID of the previous thread group. Rehash the
927 * two threads with a switched PID, and release
928 * the former thread group leader:
929 */
Eric W. Biedermand73d6522006-03-28 16:11:03 -0800930
931 /* Become a process group leader with the old leader's pid.
Eric W. Biedermanc18258c2006-09-27 01:51:06 -0700932 * The old leader becomes a thread of the this thread group.
933 * Note: The old leader also uses this pid until release_task
Eric W. Biedermand73d6522006-03-28 16:11:03 -0800934 * is called. Odd but simple and correct.
935 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700936 detach_pid(tsk, PIDTYPE_PID);
937 tsk->pid = leader->pid;
Sukadev Bhattiprolu3743ca02007-10-18 23:39:51 -0700938 attach_pid(tsk, PIDTYPE_PID, task_pid(leader));
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700939 transfer_pid(leader, tsk, PIDTYPE_PGID);
940 transfer_pid(leader, tsk, PIDTYPE_SID);
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -0800941
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700942 list_replace_rcu(&leader->tasks, &tsk->tasks);
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -0800943 list_replace_init(&leader->sibling, &tsk->sibling);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700944
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700945 tsk->group_leader = tsk;
946 leader->group_leader = tsk;
Eric W. Biedermande12a782006-04-10 17:16:49 -0600947
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700948 tsk->exit_signal = SIGCHLD;
Oleg Nesterov087806b2011-06-22 23:10:26 +0200949 leader->exit_signal = -1;
Oleg Nesterov962b5642005-11-23 13:37:43 -0800950
951 BUG_ON(leader->exit_state != EXIT_ZOMBIE);
952 leader->exit_state = EXIT_DEAD;
Oleg Nesteroveac1b5e2011-07-21 20:00:43 +0200953
954 /*
955 * We are going to release_task()->ptrace_unlink() silently,
956 * the tracer can sleep in do_wait(). EXIT_DEAD guarantees
957 * the tracer wont't block again waiting for this thread.
958 */
959 if (unlikely(leader->ptrace))
960 __wake_up_parent(leader, leader->parent);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700961 write_unlock_irq(&tasklist_lock);
Oleg Nesterov81879262008-12-01 14:18:16 -0800962
963 release_task(leader);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -0800964 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700966 sig->group_exit_task = NULL;
967 sig->notify_count = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968
969no_thread_group:
Oleg Nesterove6368252012-03-19 17:03:22 +0100970 /* we have changed execution domain */
971 tsk->exit_signal = SIGCHLD;
972
Linus Torvalds1da177e2005-04-16 15:20:36 -0700973 exit_itimers(sig);
Oleg Nesterovcbaffba2008-05-26 20:55:42 +0400974 flush_itimer_signals();
Oleg Nesterov329f7db2005-11-07 21:12:43 +0300975
Oleg Nesterovb2c903b2007-10-16 23:27:22 -0700976 if (atomic_read(&oldsighand->count) != 1) {
977 struct sighand_struct *newsighand;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700978 /*
Oleg Nesterovb2c903b2007-10-16 23:27:22 -0700979 * This ->sighand is shared with the CLONE_SIGHAND
980 * but not CLONE_THREAD task, switch to the new one.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700981 */
Oleg Nesterovb2c903b2007-10-16 23:27:22 -0700982 newsighand = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
983 if (!newsighand)
984 return -ENOMEM;
985
Linus Torvalds1da177e2005-04-16 15:20:36 -0700986 atomic_set(&newsighand->count, 1);
987 memcpy(newsighand->action, oldsighand->action,
988 sizeof(newsighand->action));
989
990 write_lock_irq(&tasklist_lock);
991 spin_lock(&oldsighand->siglock);
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700992 rcu_assign_pointer(tsk->sighand, newsighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700993 spin_unlock(&oldsighand->siglock);
994 write_unlock_irq(&tasklist_lock);
995
Davide Libenzifba2afa2007-05-10 22:23:13 -0700996 __cleanup_sighand(oldsighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700997 }
998
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700999 BUG_ON(!thread_group_leader(tsk));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001000 return 0;
Oleg Nesterovd5bbd432012-10-08 19:13:01 +02001001
1002killed:
1003 /* protects against exit_notify() and __exit_signal() */
1004 read_lock(&tasklist_lock);
1005 sig->group_exit_task = NULL;
1006 sig->notify_count = 0;
1007 read_unlock(&tasklist_lock);
1008 return -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001009}
Oleg Nesterov0840a902007-10-16 23:27:22 -07001010
Andrew Morton59714d62008-02-04 22:27:21 -08001011char *get_task_comm(char *buf, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012{
1013 /* buf must be at least sizeof(tsk->comm) in size */
1014 task_lock(tsk);
1015 strncpy(buf, tsk->comm, sizeof(tsk->comm));
1016 task_unlock(tsk);
Andrew Morton59714d62008-02-04 22:27:21 -08001017 return buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001018}
J Freyensee7d74f492011-05-06 16:56:47 -07001019EXPORT_SYMBOL_GPL(get_task_comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001020
Al Viro6a6d27d2012-08-21 09:56:33 -04001021/*
1022 * These functions flushes out all traces of the currently running executable
1023 * so that a new one can be started
1024 */
1025
Linus Torvalds1da177e2005-04-16 15:20:36 -07001026void set_task_comm(struct task_struct *tsk, char *buf)
1027{
1028 task_lock(tsk);
john stultz4614a696b2009-12-14 18:00:05 -08001029
KAMEZAWA Hiroyuki43d2b112012-01-10 15:08:09 -08001030 trace_task_rename(tsk, buf);
1031
john stultz4614a696b2009-12-14 18:00:05 -08001032 /*
1033 * Threads may access current->comm without holding
1034 * the task lock, so write the string carefully.
1035 * Readers without a lock may see incomplete new
1036 * names but are safe from non-terminating string reads.
1037 */
1038 memset(tsk->comm, 0, TASK_COMM_LEN);
1039 wmb();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001040 strlcpy(tsk->comm, buf, sizeof(tsk->comm));
1041 task_unlock(tsk);
Ingo Molnarcdd6c482009-09-21 12:02:48 +02001042 perf_event_comm(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001043}
1044
Heiko Carstens96e02d12012-02-04 10:47:10 +01001045static void filename_to_taskname(char *tcomm, const char *fn, unsigned int len)
1046{
1047 int i, ch;
1048
1049 /* Copies the binary name from after last slash */
1050 for (i = 0; (ch = *(fn++)) != '\0';) {
1051 if (ch == '/')
1052 i = 0; /* overwrite what we wrote */
1053 else
1054 if (i < len - 1)
1055 tcomm[i++] = ch;
1056 }
1057 tcomm[i] = '\0';
1058}
1059
Linus Torvalds1da177e2005-04-16 15:20:36 -07001060int flush_old_exec(struct linux_binprm * bprm)
1061{
Linus Torvalds221af7f2010-01-28 22:14:42 -08001062 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001063
1064 /*
1065 * Make sure we have a private signal table and that
1066 * we are unassociated from the previous thread group.
1067 */
1068 retval = de_thread(current);
1069 if (retval)
1070 goto out;
1071
Matt Helsley925d1c42008-04-29 01:01:36 -07001072 set_mm_exe_file(bprm->mm, bprm->file);
1073
Heiko Carstens96e02d12012-02-04 10:47:10 +01001074 filename_to_taskname(bprm->tcomm, bprm->filename, sizeof(bprm->tcomm));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001076 * Release all of the old mmap stuff
1077 */
Oleg Nesterov3c77f842010-11-30 20:55:34 +01001078 acct_arg_size(bprm, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001079 retval = exec_mmap(bprm->mm);
1080 if (retval)
Al Virofd8328b2008-04-22 05:11:59 -04001081 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001082
1083 bprm->mm = NULL; /* We're using it now */
Linus Torvalds7ab02af2010-02-02 12:37:44 -08001084
Mathias Krausedac853a2011-06-09 20:05:18 +02001085 set_fs(USER_DS);
Oleg Nesterovb40a7952012-10-25 22:28:12 +02001086 current->flags &=
1087 ~(PF_RANDOMIZE | PF_FORKNOEXEC | PF_KTHREAD | PF_NOFREEZE);
Linus Torvalds7ab02af2010-02-02 12:37:44 -08001088 flush_thread();
1089 current->personality &= ~bprm->per_clear;
1090
Linus Torvalds221af7f2010-01-28 22:14:42 -08001091 return 0;
1092
1093out:
1094 return retval;
1095}
1096EXPORT_SYMBOL(flush_old_exec);
1097
Al Viro1b5d7832011-06-19 12:49:47 -04001098void would_dump(struct linux_binprm *bprm, struct file *file)
1099{
1100 if (inode_permission(file->f_path.dentry->d_inode, MAY_READ) < 0)
1101 bprm->interp_flags |= BINPRM_FLAGS_ENFORCE_NONDUMP;
1102}
1103EXPORT_SYMBOL(would_dump);
1104
Linus Torvalds221af7f2010-01-28 22:14:42 -08001105void setup_new_exec(struct linux_binprm * bprm)
1106{
Linus Torvalds221af7f2010-01-28 22:14:42 -08001107 arch_pick_mmap_layout(current->mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108
1109 /* This is the point of no return */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001110 current->sas_ss_sp = current->sas_ss_size = 0;
1111
Eric W. Biederman8e96e3b2012-03-03 21:17:15 -08001112 if (uid_eq(current_euid(), current_uid()) && gid_eq(current_egid(), current_gid()))
Oleg Nesterov0f4cfb22012-10-04 17:15:27 -07001113 set_dumpable(current->mm, SUID_DUMPABLE_ENABLED);
Alan Coxd6e71142005-06-23 00:09:43 -07001114 else
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001115 set_dumpable(current->mm, suid_dumpable);
Alan Coxd6e71142005-06-23 00:09:43 -07001116
Heiko Carstens96e02d12012-02-04 10:47:10 +01001117 set_task_comm(current, bprm->tcomm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001118
Benjamin Herrenschmidt0551fbd2006-02-28 16:59:19 -08001119 /* Set the new mm task size. We have to do that late because it may
1120 * depend on TIF_32BIT which is only updated in flush_thread() on
1121 * some architectures like powerpc
1122 */
1123 current->mm->task_size = TASK_SIZE;
1124
David Howellsa6f76f22008-11-14 10:39:24 +11001125 /* install the new credentials */
Eric W. Biederman8e96e3b2012-03-03 21:17:15 -08001126 if (!uid_eq(bprm->cred->uid, current_euid()) ||
1127 !gid_eq(bprm->cred->gid, current_egid())) {
Marcel Holtmannd2d56c52007-08-17 21:47:58 +02001128 current->pdeath_signal = 0;
Al Viro1b5d7832011-06-19 12:49:47 -04001129 } else {
1130 would_dump(bprm, bprm->file);
1131 if (bprm->interp_flags & BINPRM_FLAGS_ENFORCE_NONDUMP)
1132 set_dumpable(current->mm, suid_dumpable);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001133 }
1134
Ingo Molnarf65cb452008-12-16 13:40:44 +01001135 /*
1136 * Flush performance counters when crossing a
1137 * security domain:
1138 */
1139 if (!get_dumpable(current->mm))
Ingo Molnarcdd6c482009-09-21 12:02:48 +02001140 perf_event_exit_task(current);
Ingo Molnarf65cb452008-12-16 13:40:44 +01001141
Linus Torvalds1da177e2005-04-16 15:20:36 -07001142 /* An exec changes our domain. We are no longer part of the thread
1143 group */
1144
1145 current->self_exec_id++;
1146
1147 flush_signal_handlers(current, 0);
Al Viro6a6d27d2012-08-21 09:56:33 -04001148 do_close_on_exec(current->files);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001149}
Linus Torvalds221af7f2010-01-28 22:14:42 -08001150EXPORT_SYMBOL(setup_new_exec);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001151
David Howellsa6f76f22008-11-14 10:39:24 +11001152/*
Oleg Nesterova2a84742009-09-05 11:17:13 -07001153 * Prepare credentials and lock ->cred_guard_mutex.
1154 * install_exec_creds() commits the new creds and drops the lock.
1155 * Or, if exec fails before, free_bprm() should release ->cred and
1156 * and unlock.
1157 */
1158int prepare_bprm_creds(struct linux_binprm *bprm)
1159{
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001160 if (mutex_lock_interruptible(&current->signal->cred_guard_mutex))
Oleg Nesterova2a84742009-09-05 11:17:13 -07001161 return -ERESTARTNOINTR;
1162
1163 bprm->cred = prepare_exec_creds();
1164 if (likely(bprm->cred))
1165 return 0;
1166
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001167 mutex_unlock(&current->signal->cred_guard_mutex);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001168 return -ENOMEM;
1169}
1170
1171void free_bprm(struct linux_binprm *bprm)
1172{
1173 free_arg_pages(bprm);
1174 if (bprm->cred) {
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001175 mutex_unlock(&current->signal->cred_guard_mutex);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001176 abort_creds(bprm->cred);
1177 }
1178 kfree(bprm);
1179}
1180
1181/*
David Howellsa6f76f22008-11-14 10:39:24 +11001182 * install the new credentials for this executable
1183 */
1184void install_exec_creds(struct linux_binprm *bprm)
1185{
1186 security_bprm_committing_creds(bprm);
1187
1188 commit_creds(bprm->cred);
1189 bprm->cred = NULL;
Oleg Nesterova2a84742009-09-05 11:17:13 -07001190 /*
1191 * cred_guard_mutex must be held at least to this point to prevent
David Howellsa6f76f22008-11-14 10:39:24 +11001192 * ptrace_attach() from altering our determination of the task's
Oleg Nesterova2a84742009-09-05 11:17:13 -07001193 * credentials; any time after this it may be unlocked.
1194 */
David Howellsa6f76f22008-11-14 10:39:24 +11001195 security_bprm_committed_creds(bprm);
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001196 mutex_unlock(&current->signal->cred_guard_mutex);
David Howellsa6f76f22008-11-14 10:39:24 +11001197}
1198EXPORT_SYMBOL(install_exec_creds);
1199
1200/*
1201 * determine how safe it is to execute the proposed program
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001202 * - the caller must hold ->cred_guard_mutex to protect against
David Howellsa6f76f22008-11-14 10:39:24 +11001203 * PTRACE_ATTACH
1204 */
Al Virof47ec3f2011-11-21 21:15:42 -05001205static int check_unsafe_exec(struct linux_binprm *bprm)
David Howellsa6f76f22008-11-14 10:39:24 +11001206{
David Howells0bf2f3a2009-02-06 11:45:46 +00001207 struct task_struct *p = current, *t;
Al Virof1191b52009-03-30 07:35:18 -04001208 unsigned n_fs;
Al Viro498052b2009-03-30 07:20:30 -04001209 int res = 0;
David Howellsa6f76f22008-11-14 10:39:24 +11001210
Tejun Heo4b9d33e2011-06-17 16:50:38 +02001211 if (p->ptrace) {
1212 if (p->ptrace & PT_PTRACE_CAP)
1213 bprm->unsafe |= LSM_UNSAFE_PTRACE_CAP;
1214 else
1215 bprm->unsafe |= LSM_UNSAFE_PTRACE;
1216 }
David Howellsa6f76f22008-11-14 10:39:24 +11001217
Andy Lutomirski259e5e62012-04-12 16:47:50 -05001218 /*
1219 * This isn't strictly necessary, but it makes it harder for LSMs to
1220 * mess up.
1221 */
1222 if (current->no_new_privs)
1223 bprm->unsafe |= LSM_UNSAFE_NO_NEW_PRIVS;
1224
David Howells0bf2f3a2009-02-06 11:45:46 +00001225 n_fs = 1;
Nick Piggin2a4419b2010-08-18 04:37:33 +10001226 spin_lock(&p->fs->lock);
Oleg Nesterov437f7fd2009-04-24 01:02:45 +02001227 rcu_read_lock();
David Howells0bf2f3a2009-02-06 11:45:46 +00001228 for (t = next_thread(p); t != p; t = next_thread(t)) {
1229 if (t->fs == p->fs)
1230 n_fs++;
David Howells0bf2f3a2009-02-06 11:45:46 +00001231 }
Oleg Nesterov437f7fd2009-04-24 01:02:45 +02001232 rcu_read_unlock();
David Howells0bf2f3a2009-02-06 11:45:46 +00001233
Al Virof1191b52009-03-30 07:35:18 -04001234 if (p->fs->users > n_fs) {
David Howellsa6f76f22008-11-14 10:39:24 +11001235 bprm->unsafe |= LSM_UNSAFE_SHARE;
Al Viro498052b2009-03-30 07:20:30 -04001236 } else {
Oleg Nesterov8c652f92009-04-24 01:01:56 +02001237 res = -EAGAIN;
1238 if (!p->fs->in_exec) {
1239 p->fs->in_exec = 1;
1240 res = 1;
1241 }
Al Viro498052b2009-03-30 07:20:30 -04001242 }
Nick Piggin2a4419b2010-08-18 04:37:33 +10001243 spin_unlock(&p->fs->lock);
Al Viro498052b2009-03-30 07:20:30 -04001244
1245 return res;
David Howellsa6f76f22008-11-14 10:39:24 +11001246}
1247
Linus Torvalds1da177e2005-04-16 15:20:36 -07001248/*
1249 * Fill the binprm structure from the inode.
1250 * Check permissions, then read the first 128 (BINPRM_BUF_SIZE) bytes
David Howellsa6f76f22008-11-14 10:39:24 +11001251 *
1252 * This may be called multiple times for binary chains (scripts for example).
Linus Torvalds1da177e2005-04-16 15:20:36 -07001253 */
1254int prepare_binprm(struct linux_binprm *bprm)
1255{
David Howellsa6f76f22008-11-14 10:39:24 +11001256 umode_t mode;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001257 struct inode * inode = bprm->file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001258 int retval;
1259
1260 mode = inode->i_mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001261 if (bprm->file->f_op == NULL)
1262 return -EACCES;
1263
David Howellsa6f76f22008-11-14 10:39:24 +11001264 /* clear any previous set[ug]id data from a previous binary */
1265 bprm->cred->euid = current_euid();
1266 bprm->cred->egid = current_egid();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001267
Andy Lutomirski259e5e62012-04-12 16:47:50 -05001268 if (!(bprm->file->f_path.mnt->mnt_flags & MNT_NOSUID) &&
Eric W. Biederman3cdf5b42011-11-21 16:40:54 -08001269 !current->no_new_privs &&
1270 kuid_has_mapping(bprm->cred->user_ns, inode->i_uid) &&
1271 kgid_has_mapping(bprm->cred->user_ns, inode->i_gid)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001272 /* Set-uid? */
1273 if (mode & S_ISUID) {
David Howellsa6f76f22008-11-14 10:39:24 +11001274 bprm->per_clear |= PER_CLEAR_ON_SETID;
1275 bprm->cred->euid = inode->i_uid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276 }
1277
1278 /* Set-gid? */
1279 /*
1280 * If setgid is set but no group execute bit then this
1281 * is a candidate for mandatory locking, not a setgid
1282 * executable.
1283 */
1284 if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) {
David Howellsa6f76f22008-11-14 10:39:24 +11001285 bprm->per_clear |= PER_CLEAR_ON_SETID;
1286 bprm->cred->egid = inode->i_gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001287 }
1288 }
1289
1290 /* fill in binprm security blob */
David Howellsa6f76f22008-11-14 10:39:24 +11001291 retval = security_bprm_set_creds(bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292 if (retval)
1293 return retval;
David Howellsa6f76f22008-11-14 10:39:24 +11001294 bprm->cred_prepared = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001295
David Howellsa6f76f22008-11-14 10:39:24 +11001296 memset(bprm->buf, 0, BINPRM_BUF_SIZE);
1297 return kernel_read(bprm->file, 0, bprm->buf, BINPRM_BUF_SIZE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001298}
1299
1300EXPORT_SYMBOL(prepare_binprm);
1301
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001302/*
1303 * Arguments are '\0' separated strings found at the location bprm->p
1304 * points to; chop off the first by relocating brpm->p to right after
1305 * the first '\0' encountered.
1306 */
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001307int remove_arg_zero(struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001308{
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001309 int ret = 0;
1310 unsigned long offset;
1311 char *kaddr;
1312 struct page *page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001313
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001314 if (!bprm->argc)
1315 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001316
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001317 do {
1318 offset = bprm->p & ~PAGE_MASK;
1319 page = get_arg_page(bprm, bprm->p, 0);
1320 if (!page) {
1321 ret = -EFAULT;
1322 goto out;
1323 }
Cong Wange8e3c3d2011-11-25 23:14:27 +08001324 kaddr = kmap_atomic(page);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001325
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001326 for (; offset < PAGE_SIZE && kaddr[offset];
1327 offset++, bprm->p++)
1328 ;
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001329
Cong Wange8e3c3d2011-11-25 23:14:27 +08001330 kunmap_atomic(kaddr);
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001331 put_arg_page(page);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001332
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001333 if (offset == PAGE_SIZE)
1334 free_arg_page(bprm, (bprm->p >> PAGE_SHIFT) - 1);
1335 } while (offset == PAGE_SIZE);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001336
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001337 bprm->p++;
1338 bprm->argc--;
1339 ret = 0;
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001340
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001341out:
1342 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001343}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001344EXPORT_SYMBOL(remove_arg_zero);
1345
1346/*
1347 * cycle the list of binary formats handler, until one recognizes the image
1348 */
Al Viro3c456bf2012-10-20 21:53:31 -04001349int search_binary_handler(struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001350{
Roland McGrath85f33462008-12-09 19:36:38 -08001351 unsigned int depth = bprm->recursion_depth;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352 int try,retval;
1353 struct linux_binfmt *fmt;
Oleg Nesterov63081912012-03-30 18:26:36 +02001354 pid_t old_pid, old_vpid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001355
Kees Cookd7402692012-12-17 16:03:20 -08001356 /* This allows 4 levels of binfmt rewrites before failing hard. */
1357 if (depth > 5)
1358 return -ELOOP;
1359
Linus Torvalds1da177e2005-04-16 15:20:36 -07001360 retval = security_bprm_check(bprm);
1361 if (retval)
1362 return retval;
1363
Al Viro473ae302006-04-26 14:04:08 -04001364 retval = audit_bprm(bprm);
1365 if (retval)
1366 return retval;
1367
Denys Vlasenkobb188d72011-06-29 04:13:39 +02001368 /* Need to fetch pid before load_binary changes it */
Oleg Nesterov63081912012-03-30 18:26:36 +02001369 old_pid = current->pid;
Denys Vlasenkobb188d72011-06-29 04:13:39 +02001370 rcu_read_lock();
Oleg Nesterov63081912012-03-30 18:26:36 +02001371 old_vpid = task_pid_nr_ns(current, task_active_pid_ns(current->parent));
Denys Vlasenkobb188d72011-06-29 04:13:39 +02001372 rcu_read_unlock();
1373
Linus Torvalds1da177e2005-04-16 15:20:36 -07001374 retval = -ENOENT;
1375 for (try=0; try<2; try++) {
1376 read_lock(&binfmt_lock);
Alexey Dobriyane4dc1b12007-10-16 23:26:03 -07001377 list_for_each_entry(fmt, &formats, lh) {
Al Viro71613c32012-10-20 22:00:48 -04001378 int (*fn)(struct linux_binprm *) = fmt->load_binary;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001379 if (!fn)
1380 continue;
1381 if (!try_module_get(fmt->module))
1382 continue;
1383 read_unlock(&binfmt_lock);
Kees Cookd7402692012-12-17 16:03:20 -08001384 bprm->recursion_depth = depth + 1;
Al Viro71613c32012-10-20 22:00:48 -04001385 retval = fn(bprm);
Roland McGrath85f33462008-12-09 19:36:38 -08001386 bprm->recursion_depth = depth;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001387 if (retval >= 0) {
David Smith4ff16c22012-02-07 10:11:05 -06001388 if (depth == 0) {
1389 trace_sched_process_exec(current, old_pid, bprm);
Oleg Nesterov63081912012-03-30 18:26:36 +02001390 ptrace_event(PTRACE_EVENT_EXEC, old_vpid);
David Smith4ff16c22012-02-07 10:11:05 -06001391 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001392 put_binfmt(fmt);
1393 allow_write_access(bprm->file);
1394 if (bprm->file)
1395 fput(bprm->file);
1396 bprm->file = NULL;
1397 current->did_exec = 1;
Matt Helsley9f460802005-11-07 00:59:16 -08001398 proc_exec_connector(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001399 return retval;
1400 }
1401 read_lock(&binfmt_lock);
1402 put_binfmt(fmt);
1403 if (retval != -ENOEXEC || bprm->mm == NULL)
1404 break;
1405 if (!bprm->file) {
1406 read_unlock(&binfmt_lock);
1407 return retval;
1408 }
1409 }
1410 read_unlock(&binfmt_lock);
Tetsuo Handab4edf8b2011-07-26 16:08:42 -07001411#ifdef CONFIG_MODULES
Linus Torvalds1da177e2005-04-16 15:20:36 -07001412 if (retval != -ENOEXEC || bprm->mm == NULL) {
1413 break;
Johannes Berg5f4123b2008-07-09 10:28:40 +02001414 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001415#define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e))
1416 if (printable(bprm->buf[0]) &&
1417 printable(bprm->buf[1]) &&
1418 printable(bprm->buf[2]) &&
1419 printable(bprm->buf[3]))
1420 break; /* -ENOEXEC */
Tetsuo Handa91219352011-07-26 16:08:41 -07001421 if (try)
1422 break; /* -ENOEXEC */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001423 request_module("binfmt-%04x", *(unsigned short *)(&bprm->buf[2]));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001424 }
Tetsuo Handab4edf8b2011-07-26 16:08:42 -07001425#else
1426 break;
1427#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428 }
1429 return retval;
1430}
1431
1432EXPORT_SYMBOL(search_binary_handler);
1433
1434/*
1435 * sys_execve() executes a new program.
1436 */
Oleg Nesterovba2d01622011-03-06 18:02:37 +01001437static int do_execve_common(const char *filename,
1438 struct user_arg_ptr argv,
Al Viro835ab322012-10-20 21:50:59 -04001439 struct user_arg_ptr envp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001440{
1441 struct linux_binprm *bprm;
1442 struct file *file;
Al Viro3b125382008-04-22 05:31:30 -04001443 struct files_struct *displaced;
Oleg Nesterov8c652f92009-04-24 01:01:56 +02001444 bool clear_in_exec;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001445 int retval;
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +04001446 const struct cred *cred = current_cred();
1447
1448 /*
1449 * We move the actual failure in case of RLIMIT_NPROC excess from
1450 * set*uid() to execve() because too many poorly written programs
1451 * don't check setuid() return code. Here we additionally recheck
1452 * whether NPROC limit is still exceeded.
1453 */
1454 if ((current->flags & PF_NPROC_EXCEEDED) &&
1455 atomic_read(&cred->user->processes) > rlimit(RLIMIT_NPROC)) {
1456 retval = -EAGAIN;
1457 goto out_ret;
1458 }
1459
1460 /* We're below the limit (still or again), so we don't want to make
1461 * further execve() calls fail. */
1462 current->flags &= ~PF_NPROC_EXCEEDED;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001463
Al Viro3b125382008-04-22 05:31:30 -04001464 retval = unshare_files(&displaced);
Al Virofd8328b2008-04-22 05:11:59 -04001465 if (retval)
1466 goto out_ret;
1467
Linus Torvalds1da177e2005-04-16 15:20:36 -07001468 retval = -ENOMEM;
Oliver Neukum11b0b5a2006-03-25 03:08:13 -08001469 bprm = kzalloc(sizeof(*bprm), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001470 if (!bprm)
Al Virofd8328b2008-04-22 05:11:59 -04001471 goto out_files;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001472
Oleg Nesterova2a84742009-09-05 11:17:13 -07001473 retval = prepare_bprm_creds(bprm);
1474 if (retval)
David Howellsa6f76f22008-11-14 10:39:24 +11001475 goto out_free;
Al Viro498052b2009-03-30 07:20:30 -04001476
1477 retval = check_unsafe_exec(bprm);
Oleg Nesterov8c652f92009-04-24 01:01:56 +02001478 if (retval < 0)
Oleg Nesterova2a84742009-09-05 11:17:13 -07001479 goto out_free;
Oleg Nesterov8c652f92009-04-24 01:01:56 +02001480 clear_in_exec = retval;
Oleg Nesterova2a84742009-09-05 11:17:13 -07001481 current->in_execve = 1;
David Howellsa6f76f22008-11-14 10:39:24 +11001482
Linus Torvalds1da177e2005-04-16 15:20:36 -07001483 file = open_exec(filename);
1484 retval = PTR_ERR(file);
1485 if (IS_ERR(file))
Al Viro498052b2009-03-30 07:20:30 -04001486 goto out_unmark;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001487
1488 sched_exec();
1489
Linus Torvalds1da177e2005-04-16 15:20:36 -07001490 bprm->file = file;
1491 bprm->filename = filename;
1492 bprm->interp = filename;
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001493
1494 retval = bprm_mm_init(bprm);
1495 if (retval)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001496 goto out_file;
1497
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001498 bprm->argc = count(argv, MAX_ARG_STRINGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001499 if ((retval = bprm->argc) < 0)
David Howellsa6f76f22008-11-14 10:39:24 +11001500 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001501
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001502 bprm->envc = count(envp, MAX_ARG_STRINGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001503 if ((retval = bprm->envc) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001504 goto out;
1505
1506 retval = prepare_binprm(bprm);
1507 if (retval < 0)
1508 goto out;
1509
1510 retval = copy_strings_kernel(1, &bprm->filename, bprm);
1511 if (retval < 0)
1512 goto out;
1513
1514 bprm->exec = bprm->p;
1515 retval = copy_strings(bprm->envc, envp, bprm);
1516 if (retval < 0)
1517 goto out;
1518
1519 retval = copy_strings(bprm->argc, argv, bprm);
1520 if (retval < 0)
1521 goto out;
1522
Al Viro3c456bf2012-10-20 21:53:31 -04001523 retval = search_binary_handler(bprm);
David Howellsa6f76f22008-11-14 10:39:24 +11001524 if (retval < 0)
1525 goto out;
1526
1527 /* execve succeeded */
Al Viro498052b2009-03-30 07:20:30 -04001528 current->fs->in_exec = 0;
Kentaro Takedaf9ce1f12009-02-05 17:18:11 +09001529 current->in_execve = 0;
David Howellsa6f76f22008-11-14 10:39:24 +11001530 acct_update_integrals(current);
1531 free_bprm(bprm);
1532 if (displaced)
1533 put_files_struct(displaced);
1534 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001535
1536out:
Oleg Nesterov3c77f842010-11-30 20:55:34 +01001537 if (bprm->mm) {
1538 acct_arg_size(bprm, 0);
1539 mmput(bprm->mm);
1540 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001541
1542out_file:
1543 if (bprm->file) {
1544 allow_write_access(bprm->file);
1545 fput(bprm->file);
1546 }
David Howellsa6f76f22008-11-14 10:39:24 +11001547
Al Viro498052b2009-03-30 07:20:30 -04001548out_unmark:
Oleg Nesterov8c652f92009-04-24 01:01:56 +02001549 if (clear_in_exec)
1550 current->fs->in_exec = 0;
Kentaro Takedaf9ce1f12009-02-05 17:18:11 +09001551 current->in_execve = 0;
David Howellsa6f76f22008-11-14 10:39:24 +11001552
1553out_free:
Al Viro08a6fac2008-05-10 16:38:25 -04001554 free_bprm(bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001555
Al Virofd8328b2008-04-22 05:11:59 -04001556out_files:
Al Viro3b125382008-04-22 05:31:30 -04001557 if (displaced)
1558 reset_files_struct(displaced);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001559out_ret:
1560 return retval;
1561}
1562
Oleg Nesterovba2d01622011-03-06 18:02:37 +01001563int do_execve(const char *filename,
1564 const char __user *const __user *__argv,
Al Viroda3d4c52012-10-20 21:49:33 -04001565 const char __user *const __user *__envp)
Oleg Nesterovba2d01622011-03-06 18:02:37 +01001566{
Oleg Nesterov0e028462011-03-06 18:02:54 +01001567 struct user_arg_ptr argv = { .ptr.native = __argv };
1568 struct user_arg_ptr envp = { .ptr.native = __envp };
Al Viro835ab322012-10-20 21:50:59 -04001569 return do_execve_common(filename, argv, envp);
Oleg Nesterovba2d01622011-03-06 18:02:37 +01001570}
1571
Oleg Nesterov0e028462011-03-06 18:02:54 +01001572#ifdef CONFIG_COMPAT
Al Virod03d26e2012-10-20 21:46:25 -04001573static int compat_do_execve(const char *filename,
Al Viro38b983b2012-09-30 13:38:55 -04001574 const compat_uptr_t __user *__argv,
Al Virod03d26e2012-10-20 21:46:25 -04001575 const compat_uptr_t __user *__envp)
Oleg Nesterov0e028462011-03-06 18:02:54 +01001576{
1577 struct user_arg_ptr argv = {
1578 .is_compat = true,
1579 .ptr.compat = __argv,
1580 };
1581 struct user_arg_ptr envp = {
1582 .is_compat = true,
1583 .ptr.compat = __envp,
1584 };
Al Viro835ab322012-10-20 21:50:59 -04001585 return do_execve_common(filename, argv, envp);
Oleg Nesterov0e028462011-03-06 18:02:54 +01001586}
1587#endif
1588
Oleg Nesterov964ee7d2009-09-23 15:56:59 -07001589void set_binfmt(struct linux_binfmt *new)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001590{
Hiroshi Shimamoto801460d2009-09-23 15:57:41 -07001591 struct mm_struct *mm = current->mm;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001592
Hiroshi Shimamoto801460d2009-09-23 15:57:41 -07001593 if (mm->binfmt)
1594 module_put(mm->binfmt->module);
1595
1596 mm->binfmt = new;
Oleg Nesterov964ee7d2009-09-23 15:56:59 -07001597 if (new)
1598 __module_get(new->module);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001599}
1600
1601EXPORT_SYMBOL(set_binfmt);
1602
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001603/*
1604 * set_dumpable converts traditional three-value dumpable to two flags and
1605 * stores them into mm->flags. It modifies lower two bits of mm->flags, but
1606 * these bits are not changed atomically. So get_dumpable can observe the
1607 * intermediate state. To avoid doing unexpected behavior, get get_dumpable
1608 * return either old dumpable or new one by paying attention to the order of
1609 * modifying the bits.
1610 *
1611 * dumpable | mm->flags (binary)
1612 * old new | initial interim final
1613 * ---------+-----------------------
1614 * 0 1 | 00 01 01
1615 * 0 2 | 00 10(*) 11
1616 * 1 0 | 01 00 00
1617 * 1 2 | 01 11 11
1618 * 2 0 | 11 10(*) 00
1619 * 2 1 | 11 11 01
1620 *
1621 * (*) get_dumpable regards interim value of 10 as 11.
1622 */
1623void set_dumpable(struct mm_struct *mm, int value)
1624{
1625 switch (value) {
Kees Cook54b50192012-07-30 14:39:18 -07001626 case SUID_DUMPABLE_DISABLED:
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001627 clear_bit(MMF_DUMPABLE, &mm->flags);
1628 smp_wmb();
1629 clear_bit(MMF_DUMP_SECURELY, &mm->flags);
1630 break;
Kees Cook54b50192012-07-30 14:39:18 -07001631 case SUID_DUMPABLE_ENABLED:
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001632 set_bit(MMF_DUMPABLE, &mm->flags);
1633 smp_wmb();
1634 clear_bit(MMF_DUMP_SECURELY, &mm->flags);
1635 break;
Kees Cook54b50192012-07-30 14:39:18 -07001636 case SUID_DUMPABLE_SAFE:
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001637 set_bit(MMF_DUMP_SECURELY, &mm->flags);
1638 smp_wmb();
1639 set_bit(MMF_DUMPABLE, &mm->flags);
1640 break;
1641 }
1642}
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001643
Alex Kelly10c28d92012-09-26 21:52:08 -04001644int __get_dumpable(unsigned long mm_flags)
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001645{
1646 int ret;
1647
Masami Hiramatsu30736a42010-03-05 13:44:12 -08001648 ret = mm_flags & MMF_DUMPABLE_MASK;
Kees Cook54b50192012-07-30 14:39:18 -07001649 return (ret > SUID_DUMPABLE_ENABLED) ? SUID_DUMPABLE_SAFE : ret;
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001650}
1651
Masami Hiramatsu30736a42010-03-05 13:44:12 -08001652int get_dumpable(struct mm_struct *mm)
1653{
1654 return __get_dumpable(mm->flags);
1655}
1656
Al Viro38b983b2012-09-30 13:38:55 -04001657#ifdef __ARCH_WANT_SYS_EXECVE
1658SYSCALL_DEFINE3(execve,
1659 const char __user *, filename,
1660 const char __user *const __user *, argv,
1661 const char __user *const __user *, envp)
1662{
Jeff Layton91a27b22012-10-10 15:25:28 -04001663 struct filename *path = getname(filename);
Al Viro38b983b2012-09-30 13:38:55 -04001664 int error = PTR_ERR(path);
1665 if (!IS_ERR(path)) {
Al Viroda3d4c52012-10-20 21:49:33 -04001666 error = do_execve(path->name, argv, envp);
Al Viro38b983b2012-09-30 13:38:55 -04001667 putname(path);
1668 }
1669 return error;
1670}
1671#ifdef CONFIG_COMPAT
1672asmlinkage long compat_sys_execve(const char __user * filename,
1673 const compat_uptr_t __user * argv,
1674 const compat_uptr_t __user * envp)
1675{
Jeff Layton91a27b22012-10-10 15:25:28 -04001676 struct filename *path = getname(filename);
Al Viro38b983b2012-09-30 13:38:55 -04001677 int error = PTR_ERR(path);
1678 if (!IS_ERR(path)) {
Al Virod03d26e2012-10-20 21:46:25 -04001679 error = compat_do_execve(path->name, argv, envp);
Al Viro38b983b2012-09-30 13:38:55 -04001680 putname(path);
1681 }
1682 return error;
1683}
1684#endif
1685#endif
1686
Al Viro282124d2012-09-30 13:20:09 -04001687#ifdef __ARCH_WANT_KERNEL_EXECVE
1688int kernel_execve(const char *filename,
1689 const char *const argv[],
1690 const char *const envp[])
1691{
Al Viroda3d4c52012-10-20 21:49:33 -04001692 int ret = do_execve(filename,
Al Viro282124d2012-09-30 13:20:09 -04001693 (const char __user *const __user *)argv,
Al Viroda3d4c52012-10-20 21:49:33 -04001694 (const char __user *const __user *)envp);
Al Viro282124d2012-09-30 13:20:09 -04001695 if (ret < 0)
1696 return ret;
1697
1698 /*
1699 * We were successful. We won't be returning to our caller, but
1700 * instead to user space by manipulating the kernel stack.
1701 */
Al Viroda3d4c52012-10-20 21:49:33 -04001702 ret_from_kernel_execve(current_pt_regs());
Al Viro282124d2012-09-30 13:20:09 -04001703}
1704#endif