blob: 4f2bebc276c52f3ca60b75fbff9eeea8ea21bbe5 [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"
Alex Kelly179899f2012-10-04 17:15:24 -070066#include "coredump.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070067
David Smith4ff16c22012-02-07 10:11:05 -060068#include <trace/events/sched.h>
69
Alan Coxd6e71142005-06-23 00:09:43 -070070int suid_dumpable = 0;
71
Alexey Dobriyane4dc1b12007-10-16 23:26:03 -070072static LIST_HEAD(formats);
Linus Torvalds1da177e2005-04-16 15:20:36 -070073static DEFINE_RWLOCK(binfmt_lock);
74
Al Viro8fc3dc52012-03-17 03:05:16 -040075void __register_binfmt(struct linux_binfmt * fmt, int insert)
Linus Torvalds1da177e2005-04-16 15:20:36 -070076{
Al Viro8fc3dc52012-03-17 03:05:16 -040077 BUG_ON(!fmt);
Linus Torvalds1da177e2005-04-16 15:20:36 -070078 write_lock(&binfmt_lock);
Ivan Kokshaysky74641f52009-04-30 15:08:49 -070079 insert ? list_add(&fmt->lh, &formats) :
80 list_add_tail(&fmt->lh, &formats);
Linus Torvalds1da177e2005-04-16 15:20:36 -070081 write_unlock(&binfmt_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -070082}
83
Ivan Kokshaysky74641f52009-04-30 15:08:49 -070084EXPORT_SYMBOL(__register_binfmt);
Linus Torvalds1da177e2005-04-16 15:20:36 -070085
Alexey Dobriyanf6b450d2007-10-16 23:26:04 -070086void unregister_binfmt(struct linux_binfmt * fmt)
Linus Torvalds1da177e2005-04-16 15:20:36 -070087{
Linus Torvalds1da177e2005-04-16 15:20:36 -070088 write_lock(&binfmt_lock);
Alexey Dobriyane4dc1b12007-10-16 23:26:03 -070089 list_del(&fmt->lh);
Linus Torvalds1da177e2005-04-16 15:20:36 -070090 write_unlock(&binfmt_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -070091}
92
93EXPORT_SYMBOL(unregister_binfmt);
94
95static inline void put_binfmt(struct linux_binfmt * fmt)
96{
97 module_put(fmt->module);
98}
99
100/*
101 * Note that a shared library must be both readable and executable due to
102 * security reasons.
103 *
104 * Also note that we take the address to load from from the file itself.
105 */
Heiko Carstens1e7bfb22009-01-14 14:14:29 +0100106SYSCALL_DEFINE1(uselib, const char __user *, library)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700107{
Al Viro964bd182008-07-26 03:33:14 -0400108 struct file *file;
Al Viro964bd182008-07-26 03:33:14 -0400109 char *tmp = getname(library);
110 int error = PTR_ERR(tmp);
Al Viro47c805d2011-02-23 17:44:09 -0500111 static const struct open_flags uselib_flags = {
112 .open_flag = O_LARGEFILE | O_RDONLY | __FMODE_EXEC,
113 .acc_mode = MAY_READ | MAY_EXEC | MAY_OPEN,
114 .intent = LOOKUP_OPEN
115 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116
Al Viro6e8341a2009-04-06 11:16:22 -0400117 if (IS_ERR(tmp))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700118 goto out;
119
Al Viro47c805d2011-02-23 17:44:09 -0500120 file = do_filp_open(AT_FDCWD, tmp, &uselib_flags, LOOKUP_FOLLOW);
Al Viro6e8341a2009-04-06 11:16:22 -0400121 putname(tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122 error = PTR_ERR(file);
123 if (IS_ERR(file))
124 goto out;
125
Al Viro6e8341a2009-04-06 11:16:22 -0400126 error = -EINVAL;
127 if (!S_ISREG(file->f_path.dentry->d_inode->i_mode))
128 goto exit;
129
130 error = -EACCES;
131 if (file->f_path.mnt->mnt_flags & MNT_NOEXEC)
132 goto exit;
133
Eric Paris2a12a9d2009-12-17 21:24:21 -0500134 fsnotify_open(file);
Eric Paris6110e3a2008-12-17 13:53:20 -0500135
Linus Torvalds1da177e2005-04-16 15:20:36 -0700136 error = -ENOEXEC;
137 if(file->f_op) {
138 struct linux_binfmt * fmt;
139
140 read_lock(&binfmt_lock);
Alexey Dobriyane4dc1b12007-10-16 23:26:03 -0700141 list_for_each_entry(fmt, &formats, lh) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700142 if (!fmt->load_shlib)
143 continue;
144 if (!try_module_get(fmt->module))
145 continue;
146 read_unlock(&binfmt_lock);
147 error = fmt->load_shlib(file);
148 read_lock(&binfmt_lock);
149 put_binfmt(fmt);
150 if (error != -ENOEXEC)
151 break;
152 }
153 read_unlock(&binfmt_lock);
154 }
Al Viro6e8341a2009-04-06 11:16:22 -0400155exit:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700156 fput(file);
157out:
158 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159}
160
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700161#ifdef CONFIG_MMU
Oleg Nesterovae6b5852011-03-06 18:03:11 +0100162/*
163 * The nascent bprm->mm is not visible until exec_mmap() but it can
164 * use a lot of memory, account these pages in current->mm temporary
165 * for oom_badness()->get_mm_rss(). Once exec succeeds or fails, we
166 * change the counter back via acct_arg_size(0).
167 */
Oleg Nesterov0e028462011-03-06 18:02:54 +0100168static void acct_arg_size(struct linux_binprm *bprm, unsigned long pages)
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100169{
170 struct mm_struct *mm = current->mm;
171 long diff = (long)(pages - bprm->vma_pages);
172
173 if (!mm || !diff)
174 return;
175
176 bprm->vma_pages = pages;
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100177 add_mm_counter(mm, MM_ANONPAGES, diff);
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100178}
179
Oleg Nesterov0e028462011-03-06 18:02:54 +0100180static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos,
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700181 int write)
182{
183 struct page *page;
184 int ret;
185
186#ifdef CONFIG_STACK_GROWSUP
187 if (write) {
Michal Hockod05f3162011-05-24 17:11:44 -0700188 ret = expand_downwards(bprm->vma, pos);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700189 if (ret < 0)
190 return NULL;
191 }
192#endif
193 ret = get_user_pages(current, bprm->mm, pos,
194 1, write, 1, &page, NULL);
195 if (ret <= 0)
196 return NULL;
197
198 if (write) {
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700199 unsigned long size = bprm->vma->vm_end - bprm->vma->vm_start;
Linus Torvaldsa64e7152008-03-03 10:12:14 -0800200 struct rlimit *rlim;
201
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100202 acct_arg_size(bprm, size / PAGE_SIZE);
203
Linus Torvaldsa64e7152008-03-03 10:12:14 -0800204 /*
205 * We've historically supported up to 32 pages (ARG_MAX)
206 * of argument strings even with small stacks
207 */
208 if (size <= ARG_MAX)
209 return page;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700210
211 /*
212 * Limit to 1/4-th the stack size for the argv+env strings.
213 * This ensures that:
214 * - the remaining binfmt code will not run out of stack space,
215 * - the program will have a reasonable amount of stack left
216 * to work from.
217 */
Linus Torvaldsa64e7152008-03-03 10:12:14 -0800218 rlim = current->signal->rlim;
Jiri Slabyd554ed892010-03-05 13:42:42 -0800219 if (size > ACCESS_ONCE(rlim[RLIMIT_STACK].rlim_cur) / 4) {
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700220 put_page(page);
221 return NULL;
222 }
223 }
224
225 return page;
226}
227
228static void put_arg_page(struct page *page)
229{
230 put_page(page);
231}
232
233static void free_arg_page(struct linux_binprm *bprm, int i)
234{
235}
236
237static void free_arg_pages(struct linux_binprm *bprm)
238{
239}
240
241static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos,
242 struct page *page)
243{
244 flush_cache_page(bprm->vma, pos, page_to_pfn(page));
245}
246
247static int __bprm_mm_init(struct linux_binprm *bprm)
248{
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800249 int err;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700250 struct vm_area_struct *vma = NULL;
251 struct mm_struct *mm = bprm->mm;
252
253 bprm->vma = vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL);
254 if (!vma)
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800255 return -ENOMEM;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700256
257 down_write(&mm->mmap_sem);
258 vma->vm_mm = mm;
259
260 /*
261 * Place the stack at the largest stack address the architecture
262 * supports. Later, we'll move this to an appropriate place. We don't
263 * use STACK_TOP because that can depend on attributes which aren't
264 * configured yet.
265 */
Michal Hockoaacb3d12011-07-26 16:08:40 -0700266 BUILD_BUG_ON(VM_STACK_FLAGS & VM_STACK_INCOMPLETE_SETUP);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700267 vma->vm_end = STACK_TOP_MAX;
268 vma->vm_start = vma->vm_end - PAGE_SIZE;
Mel Gormana8bef8ff2010-05-24 14:32:24 -0700269 vma->vm_flags = VM_STACK_FLAGS | VM_STACK_INCOMPLETE_SETUP;
Coly Li3ed75eb2007-10-18 23:39:15 -0700270 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
Rik van Riel5beb4932010-03-05 13:42:07 -0800271 INIT_LIST_HEAD(&vma->anon_vma_chain);
Tavis Ormandy462e635e2010-12-09 15:29:42 +0100272
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700273 err = insert_vm_struct(mm, vma);
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800274 if (err)
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700275 goto err;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700276
277 mm->stack_vm = mm->total_vm = 1;
278 up_write(&mm->mmap_sem);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700279 bprm->p = vma->vm_end - sizeof(void *);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700280 return 0;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700281err:
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800282 up_write(&mm->mmap_sem);
283 bprm->vma = NULL;
284 kmem_cache_free(vm_area_cachep, vma);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700285 return err;
286}
287
288static bool valid_arg_len(struct linux_binprm *bprm, long len)
289{
290 return len <= MAX_ARG_STRLEN;
291}
292
293#else
294
Oleg Nesterov0e028462011-03-06 18:02:54 +0100295static inline void acct_arg_size(struct linux_binprm *bprm, unsigned long pages)
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100296{
297}
298
Oleg Nesterov0e028462011-03-06 18:02:54 +0100299static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos,
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700300 int write)
301{
302 struct page *page;
303
304 page = bprm->page[pos / PAGE_SIZE];
305 if (!page && write) {
306 page = alloc_page(GFP_HIGHUSER|__GFP_ZERO);
307 if (!page)
308 return NULL;
309 bprm->page[pos / PAGE_SIZE] = page;
310 }
311
312 return page;
313}
314
315static void put_arg_page(struct page *page)
316{
317}
318
319static void free_arg_page(struct linux_binprm *bprm, int i)
320{
321 if (bprm->page[i]) {
322 __free_page(bprm->page[i]);
323 bprm->page[i] = NULL;
324 }
325}
326
327static void free_arg_pages(struct linux_binprm *bprm)
328{
329 int i;
330
331 for (i = 0; i < MAX_ARG_PAGES; i++)
332 free_arg_page(bprm, i);
333}
334
335static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos,
336 struct page *page)
337{
338}
339
340static int __bprm_mm_init(struct linux_binprm *bprm)
341{
342 bprm->p = PAGE_SIZE * MAX_ARG_PAGES - sizeof(void *);
343 return 0;
344}
345
346static bool valid_arg_len(struct linux_binprm *bprm, long len)
347{
348 return len <= bprm->p;
349}
350
351#endif /* CONFIG_MMU */
352
353/*
354 * Create a new mm_struct and populate it with a temporary stack
355 * vm_area_struct. We don't have enough context at this point to set the stack
356 * flags, permissions, and offset, so we use temporary values. We'll update
357 * them later in setup_arg_pages().
358 */
359int bprm_mm_init(struct linux_binprm *bprm)
360{
361 int err;
362 struct mm_struct *mm = NULL;
363
364 bprm->mm = mm = mm_alloc();
365 err = -ENOMEM;
366 if (!mm)
367 goto err;
368
369 err = init_new_context(current, mm);
370 if (err)
371 goto err;
372
373 err = __bprm_mm_init(bprm);
374 if (err)
375 goto err;
376
377 return 0;
378
379err:
380 if (mm) {
381 bprm->mm = NULL;
382 mmdrop(mm);
383 }
384
385 return err;
386}
387
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100388struct user_arg_ptr {
Oleg Nesterov0e028462011-03-06 18:02:54 +0100389#ifdef CONFIG_COMPAT
390 bool is_compat;
391#endif
392 union {
393 const char __user *const __user *native;
394#ifdef CONFIG_COMPAT
395 compat_uptr_t __user *compat;
396#endif
397 } ptr;
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100398};
399
400static const char __user *get_user_arg_ptr(struct user_arg_ptr argv, int nr)
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100401{
Oleg Nesterov0e028462011-03-06 18:02:54 +0100402 const char __user *native;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100403
Oleg Nesterov0e028462011-03-06 18:02:54 +0100404#ifdef CONFIG_COMPAT
405 if (unlikely(argv.is_compat)) {
406 compat_uptr_t compat;
407
408 if (get_user(compat, argv.ptr.compat + nr))
409 return ERR_PTR(-EFAULT);
410
411 return compat_ptr(compat);
412 }
413#endif
414
415 if (get_user(native, argv.ptr.native + nr))
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100416 return ERR_PTR(-EFAULT);
417
Oleg Nesterov0e028462011-03-06 18:02:54 +0100418 return native;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100419}
420
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421/*
422 * count() counts the number of strings in array ARGV.
423 */
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100424static int count(struct user_arg_ptr argv, int max)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425{
426 int i = 0;
427
Oleg Nesterov0e028462011-03-06 18:02:54 +0100428 if (argv.ptr.native != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700429 for (;;) {
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100430 const char __user *p = get_user_arg_ptr(argv, i);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700431
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432 if (!p)
433 break;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100434
435 if (IS_ERR(p))
436 return -EFAULT;
437
Jason Baron362e6662008-10-15 22:01:52 -0700438 if (i++ >= max)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700439 return -E2BIG;
Roland McGrath9aea5a62010-09-07 19:37:06 -0700440
441 if (fatal_signal_pending(current))
442 return -ERESTARTNOHAND;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443 cond_resched();
444 }
445 }
446 return i;
447}
448
449/*
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700450 * 'copy_strings()' copies argument/environment strings from the old
451 * processes's memory to the new process's stack. The call to get_user_pages()
452 * ensures the destination page is created and not swapped out.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453 */
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100454static int copy_strings(int argc, struct user_arg_ptr argv,
Adrian Bunk75c96f82005-05-05 16:16:09 -0700455 struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456{
457 struct page *kmapped_page = NULL;
458 char *kaddr = NULL;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700459 unsigned long kpos = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460 int ret;
461
462 while (argc-- > 0) {
David Howellsd7627462010-08-17 23:52:56 +0100463 const char __user *str;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464 int len;
465 unsigned long pos;
466
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100467 ret = -EFAULT;
468 str = get_user_arg_ptr(argv, argc);
469 if (IS_ERR(str))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100472 len = strnlen_user(str, MAX_ARG_STRLEN);
473 if (!len)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474 goto out;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100475
476 ret = -E2BIG;
477 if (!valid_arg_len(bprm, len))
478 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700479
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700480 /* We're going to work our way backwords. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481 pos = bprm->p;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700482 str += len;
483 bprm->p -= len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484
485 while (len > 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486 int offset, bytes_to_copy;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487
Roland McGrath9aea5a62010-09-07 19:37:06 -0700488 if (fatal_signal_pending(current)) {
489 ret = -ERESTARTNOHAND;
490 goto out;
491 }
Roland McGrath7993bc12010-09-07 19:36:28 -0700492 cond_resched();
493
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494 offset = pos % PAGE_SIZE;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700495 if (offset == 0)
496 offset = PAGE_SIZE;
497
498 bytes_to_copy = offset;
499 if (bytes_to_copy > len)
500 bytes_to_copy = len;
501
502 offset -= bytes_to_copy;
503 pos -= bytes_to_copy;
504 str -= bytes_to_copy;
505 len -= bytes_to_copy;
506
507 if (!kmapped_page || kpos != (pos & PAGE_MASK)) {
508 struct page *page;
509
510 page = get_arg_page(bprm, pos, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511 if (!page) {
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700512 ret = -E2BIG;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700513 goto out;
514 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700516 if (kmapped_page) {
517 flush_kernel_dcache_page(kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518 kunmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700519 put_arg_page(kmapped_page);
520 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521 kmapped_page = page;
522 kaddr = kmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700523 kpos = pos & PAGE_MASK;
524 flush_arg_page(bprm, kpos, kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525 }
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700526 if (copy_from_user(kaddr+offset, str, bytes_to_copy)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527 ret = -EFAULT;
528 goto out;
529 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700530 }
531 }
532 ret = 0;
533out:
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700534 if (kmapped_page) {
535 flush_kernel_dcache_page(kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536 kunmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700537 put_arg_page(kmapped_page);
538 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539 return ret;
540}
541
542/*
543 * Like copy_strings, but get argv and its values from kernel memory.
544 */
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100545int copy_strings_kernel(int argc, const char *const *__argv,
David Howellsd7627462010-08-17 23:52:56 +0100546 struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547{
548 int r;
549 mm_segment_t oldfs = get_fs();
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100550 struct user_arg_ptr argv = {
Oleg Nesterov0e028462011-03-06 18:02:54 +0100551 .ptr.native = (const char __user *const __user *)__argv,
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100552 };
553
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554 set_fs(KERNEL_DS);
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100555 r = copy_strings(argc, argv, bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700556 set_fs(oldfs);
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100557
Linus Torvalds1da177e2005-04-16 15:20:36 -0700558 return r;
559}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560EXPORT_SYMBOL(copy_strings_kernel);
561
562#ifdef CONFIG_MMU
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700563
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564/*
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700565 * During bprm_mm_init(), we create a temporary stack at STACK_TOP_MAX. Once
566 * the binfmt code determines where the new stack should reside, we shift it to
567 * its final location. The process proceeds as follows:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568 *
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700569 * 1) Use shift to calculate the new vma endpoints.
570 * 2) Extend vma to cover both the old and new ranges. This ensures the
571 * arguments passed to subsequent functions are consistent.
572 * 3) Move vma's page tables to the new range.
573 * 4) Free up any cleared pgd range.
574 * 5) Shrink the vma to cover only the new range.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700575 */
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700576static int shift_arg_pages(struct vm_area_struct *vma, unsigned long shift)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700577{
578 struct mm_struct *mm = vma->vm_mm;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700579 unsigned long old_start = vma->vm_start;
580 unsigned long old_end = vma->vm_end;
581 unsigned long length = old_end - old_start;
582 unsigned long new_start = old_start - shift;
583 unsigned long new_end = old_end - shift;
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700584 struct mmu_gather tlb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700585
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700586 BUG_ON(new_start > new_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700587
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700588 /*
589 * ensure there are no vmas between where we want to go
590 * and where we are
591 */
592 if (vma != find_vma(mm, new_start))
593 return -EFAULT;
594
595 /*
596 * cover the whole range: [new_start, old_end)
597 */
Rik van Riel5beb4932010-03-05 13:42:07 -0800598 if (vma_adjust(vma, new_start, old_end, vma->vm_pgoff, NULL))
599 return -ENOMEM;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700600
601 /*
602 * move the page tables downwards, on failure we rely on
603 * process cleanup to remove whatever mess we made.
604 */
605 if (length != move_page_tables(vma, old_start,
Michel Lespinasse38a76012012-10-08 16:31:50 -0700606 vma, new_start, length, false))
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700607 return -ENOMEM;
608
609 lru_add_drain();
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700610 tlb_gather_mmu(&tlb, mm, 0);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700611 if (new_end > old_start) {
612 /*
613 * when the old and new regions overlap clear from new_end.
614 */
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700615 free_pgd_range(&tlb, new_end, old_end, new_end,
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700616 vma->vm_next ? vma->vm_next->vm_start : 0);
617 } else {
618 /*
619 * otherwise, clean from old_start; this is done to not touch
620 * the address space in [new_end, old_start) some architectures
621 * have constraints on va-space that make this illegal (IA64) -
622 * for the others its just a little faster.
623 */
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700624 free_pgd_range(&tlb, old_start, old_end, new_end,
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700625 vma->vm_next ? vma->vm_next->vm_start : 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626 }
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700627 tlb_finish_mmu(&tlb, new_end, old_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700628
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700629 /*
Rik van Riel5beb4932010-03-05 13:42:07 -0800630 * Shrink the vma to just the new range. Always succeeds.
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700631 */
632 vma_adjust(vma, new_start, new_end, vma->vm_pgoff, NULL);
633
634 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700635}
636
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700637/*
638 * Finalizes the stack vm_area_struct. The flags and permissions are updated,
639 * the stack is optionally relocated, and some extra space is added.
640 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700641int setup_arg_pages(struct linux_binprm *bprm,
642 unsigned long stack_top,
643 int executable_stack)
644{
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700645 unsigned long ret;
646 unsigned long stack_shift;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700647 struct mm_struct *mm = current->mm;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700648 struct vm_area_struct *vma = bprm->vma;
649 struct vm_area_struct *prev = NULL;
650 unsigned long vm_flags;
651 unsigned long stack_base;
Michael Neuling803bf5e2010-02-10 13:56:42 -0800652 unsigned long stack_size;
653 unsigned long stack_expand;
654 unsigned long rlim_stack;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655
656#ifdef CONFIG_STACK_GROWSUP
Linus Torvalds1da177e2005-04-16 15:20:36 -0700657 /* Limit stack size to 1GB */
Jiri Slabyd554ed892010-03-05 13:42:42 -0800658 stack_base = rlimit_max(RLIMIT_STACK);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659 if (stack_base > (1 << 30))
660 stack_base = 1 << 30;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700662 /* Make sure we didn't let the argument array grow too large. */
663 if (vma->vm_end - vma->vm_start > stack_base)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664 return -ENOMEM;
665
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700666 stack_base = PAGE_ALIGN(stack_top - stack_base);
667
668 stack_shift = vma->vm_start - stack_base;
669 mm->arg_start = bprm->p - stack_shift;
670 bprm->p = vma->vm_end - stack_shift;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671#else
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700672 stack_top = arch_align_stack(stack_top);
673 stack_top = PAGE_ALIGN(stack_top);
Roland McGrath1b528182010-09-07 19:35:49 -0700674
675 if (unlikely(stack_top < mmap_min_addr) ||
676 unlikely(vma->vm_end - vma->vm_start >= stack_top - mmap_min_addr))
677 return -ENOMEM;
678
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700679 stack_shift = vma->vm_end - stack_top;
680
681 bprm->p -= stack_shift;
682 mm->arg_start = bprm->p;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700683#endif
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700684
685 if (bprm->loader)
686 bprm->loader -= stack_shift;
687 bprm->exec -= stack_shift;
688
689 down_write(&mm->mmap_sem);
Hugh Dickins96a8e132008-07-10 21:19:20 +0100690 vm_flags = VM_STACK_FLAGS;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700691
692 /*
693 * Adjust stack execute permissions; explicitly enable for
694 * EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X and leave alone
695 * (arch default) otherwise.
696 */
697 if (unlikely(executable_stack == EXSTACK_ENABLE_X))
698 vm_flags |= VM_EXEC;
699 else if (executable_stack == EXSTACK_DISABLE_X)
700 vm_flags &= ~VM_EXEC;
701 vm_flags |= mm->def_flags;
Mel Gormana8bef8ff2010-05-24 14:32:24 -0700702 vm_flags |= VM_STACK_INCOMPLETE_SETUP;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700703
704 ret = mprotect_fixup(vma, &prev, vma->vm_start, vma->vm_end,
705 vm_flags);
706 if (ret)
707 goto out_unlock;
708 BUG_ON(prev != vma);
709
710 /* Move stack pages down in memory. */
711 if (stack_shift) {
712 ret = shift_arg_pages(vma, stack_shift);
Anton Blanchardfc63cf22009-11-11 14:26:48 -0800713 if (ret)
714 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700715 }
716
Mel Gormana8bef8ff2010-05-24 14:32:24 -0700717 /* mprotect_fixup is overkill to remove the temporary stack flags */
718 vma->vm_flags &= ~VM_STACK_INCOMPLETE_SETUP;
719
Michael Neuling5ef097d2010-03-05 13:42:57 -0800720 stack_expand = 131072UL; /* randomly 32*4k (or 2*64k) pages */
Michael Neuling803bf5e2010-02-10 13:56:42 -0800721 stack_size = vma->vm_end - vma->vm_start;
722 /*
723 * Align this down to a page boundary as expand_stack
724 * will align it up.
725 */
726 rlim_stack = rlimit(RLIMIT_STACK) & PAGE_MASK;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700727#ifdef CONFIG_STACK_GROWSUP
Michael Neuling803bf5e2010-02-10 13:56:42 -0800728 if (stack_size + stack_expand > rlim_stack)
729 stack_base = vma->vm_start + rlim_stack;
730 else
731 stack_base = vma->vm_end + stack_expand;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700732#else
Michael Neuling803bf5e2010-02-10 13:56:42 -0800733 if (stack_size + stack_expand > rlim_stack)
734 stack_base = vma->vm_end - rlim_stack;
735 else
736 stack_base = vma->vm_start - stack_expand;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700737#endif
Eric B Munson3af9e852010-05-18 15:30:49 +0100738 current->mm->start_stack = bprm->p;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700739 ret = expand_stack(vma, stack_base);
740 if (ret)
741 ret = -EFAULT;
742
743out_unlock:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744 up_write(&mm->mmap_sem);
Anton Blanchardfc63cf22009-11-11 14:26:48 -0800745 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747EXPORT_SYMBOL(setup_arg_pages);
748
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749#endif /* CONFIG_MMU */
750
751struct file *open_exec(const char *name)
752{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753 struct file *file;
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200754 int err;
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
Al Viro47c805d2011-02-23 17:44:09 -0500761 file = do_filp_open(AT_FDCWD, name, &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);
Al Viro19e51092012-02-23 22:29:17 -05001086 current->flags &= ~(PF_RANDOMIZE | PF_FORKNOEXEC | PF_KTHREAD);
Linus Torvalds7ab02af2010-02-02 12:37:44 -08001087 flush_thread();
1088 current->personality &= ~bprm->per_clear;
1089
Linus Torvalds221af7f2010-01-28 22:14:42 -08001090 return 0;
1091
1092out:
1093 return retval;
1094}
1095EXPORT_SYMBOL(flush_old_exec);
1096
Al Viro1b5d7832011-06-19 12:49:47 -04001097void would_dump(struct linux_binprm *bprm, struct file *file)
1098{
1099 if (inode_permission(file->f_path.dentry->d_inode, MAY_READ) < 0)
1100 bprm->interp_flags |= BINPRM_FLAGS_ENFORCE_NONDUMP;
1101}
1102EXPORT_SYMBOL(would_dump);
1103
Linus Torvalds221af7f2010-01-28 22:14:42 -08001104void setup_new_exec(struct linux_binprm * bprm)
1105{
Linus Torvalds221af7f2010-01-28 22:14:42 -08001106 arch_pick_mmap_layout(current->mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001107
1108 /* This is the point of no return */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001109 current->sas_ss_sp = current->sas_ss_size = 0;
1110
Eric W. Biederman8e96e3b2012-03-03 21:17:15 -08001111 if (uid_eq(current_euid(), current_uid()) && gid_eq(current_egid(), current_gid()))
Oleg Nesterov0f4cfb22012-10-04 17:15:27 -07001112 set_dumpable(current->mm, SUID_DUMPABLE_ENABLED);
Alan Coxd6e71142005-06-23 00:09:43 -07001113 else
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001114 set_dumpable(current->mm, suid_dumpable);
Alan Coxd6e71142005-06-23 00:09:43 -07001115
Heiko Carstens96e02d12012-02-04 10:47:10 +01001116 set_task_comm(current, bprm->tcomm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001117
Benjamin Herrenschmidt0551fbd2006-02-28 16:59:19 -08001118 /* Set the new mm task size. We have to do that late because it may
1119 * depend on TIF_32BIT which is only updated in flush_thread() on
1120 * some architectures like powerpc
1121 */
1122 current->mm->task_size = TASK_SIZE;
1123
David Howellsa6f76f22008-11-14 10:39:24 +11001124 /* install the new credentials */
Eric W. Biederman8e96e3b2012-03-03 21:17:15 -08001125 if (!uid_eq(bprm->cred->uid, current_euid()) ||
1126 !gid_eq(bprm->cred->gid, current_egid())) {
Marcel Holtmannd2d56c52007-08-17 21:47:58 +02001127 current->pdeath_signal = 0;
Al Viro1b5d7832011-06-19 12:49:47 -04001128 } else {
1129 would_dump(bprm, bprm->file);
1130 if (bprm->interp_flags & BINPRM_FLAGS_ENFORCE_NONDUMP)
1131 set_dumpable(current->mm, suid_dumpable);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001132 }
1133
Ingo Molnarf65cb452008-12-16 13:40:44 +01001134 /*
1135 * Flush performance counters when crossing a
1136 * security domain:
1137 */
1138 if (!get_dumpable(current->mm))
Ingo Molnarcdd6c482009-09-21 12:02:48 +02001139 perf_event_exit_task(current);
Ingo Molnarf65cb452008-12-16 13:40:44 +01001140
Linus Torvalds1da177e2005-04-16 15:20:36 -07001141 /* An exec changes our domain. We are no longer part of the thread
1142 group */
1143
1144 current->self_exec_id++;
1145
1146 flush_signal_handlers(current, 0);
Al Viro6a6d27d2012-08-21 09:56:33 -04001147 do_close_on_exec(current->files);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001148}
Linus Torvalds221af7f2010-01-28 22:14:42 -08001149EXPORT_SYMBOL(setup_new_exec);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001150
David Howellsa6f76f22008-11-14 10:39:24 +11001151/*
Oleg Nesterova2a84742009-09-05 11:17:13 -07001152 * Prepare credentials and lock ->cred_guard_mutex.
1153 * install_exec_creds() commits the new creds and drops the lock.
1154 * Or, if exec fails before, free_bprm() should release ->cred and
1155 * and unlock.
1156 */
1157int prepare_bprm_creds(struct linux_binprm *bprm)
1158{
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001159 if (mutex_lock_interruptible(&current->signal->cred_guard_mutex))
Oleg Nesterova2a84742009-09-05 11:17:13 -07001160 return -ERESTARTNOINTR;
1161
1162 bprm->cred = prepare_exec_creds();
1163 if (likely(bprm->cred))
1164 return 0;
1165
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001166 mutex_unlock(&current->signal->cred_guard_mutex);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001167 return -ENOMEM;
1168}
1169
1170void free_bprm(struct linux_binprm *bprm)
1171{
1172 free_arg_pages(bprm);
1173 if (bprm->cred) {
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001174 mutex_unlock(&current->signal->cred_guard_mutex);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001175 abort_creds(bprm->cred);
1176 }
1177 kfree(bprm);
1178}
1179
1180/*
David Howellsa6f76f22008-11-14 10:39:24 +11001181 * install the new credentials for this executable
1182 */
1183void install_exec_creds(struct linux_binprm *bprm)
1184{
1185 security_bprm_committing_creds(bprm);
1186
1187 commit_creds(bprm->cred);
1188 bprm->cred = NULL;
Oleg Nesterova2a84742009-09-05 11:17:13 -07001189 /*
1190 * cred_guard_mutex must be held at least to this point to prevent
David Howellsa6f76f22008-11-14 10:39:24 +11001191 * ptrace_attach() from altering our determination of the task's
Oleg Nesterova2a84742009-09-05 11:17:13 -07001192 * credentials; any time after this it may be unlocked.
1193 */
David Howellsa6f76f22008-11-14 10:39:24 +11001194 security_bprm_committed_creds(bprm);
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001195 mutex_unlock(&current->signal->cred_guard_mutex);
David Howellsa6f76f22008-11-14 10:39:24 +11001196}
1197EXPORT_SYMBOL(install_exec_creds);
1198
1199/*
1200 * determine how safe it is to execute the proposed program
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001201 * - the caller must hold ->cred_guard_mutex to protect against
David Howellsa6f76f22008-11-14 10:39:24 +11001202 * PTRACE_ATTACH
1203 */
Al Virof47ec3f2011-11-21 21:15:42 -05001204static int check_unsafe_exec(struct linux_binprm *bprm)
David Howellsa6f76f22008-11-14 10:39:24 +11001205{
David Howells0bf2f3a2009-02-06 11:45:46 +00001206 struct task_struct *p = current, *t;
Al Virof1191b52009-03-30 07:35:18 -04001207 unsigned n_fs;
Al Viro498052b2009-03-30 07:20:30 -04001208 int res = 0;
David Howellsa6f76f22008-11-14 10:39:24 +11001209
Tejun Heo4b9d33e2011-06-17 16:50:38 +02001210 if (p->ptrace) {
1211 if (p->ptrace & PT_PTRACE_CAP)
1212 bprm->unsafe |= LSM_UNSAFE_PTRACE_CAP;
1213 else
1214 bprm->unsafe |= LSM_UNSAFE_PTRACE;
1215 }
David Howellsa6f76f22008-11-14 10:39:24 +11001216
Andy Lutomirski259e5e62012-04-12 16:47:50 -05001217 /*
1218 * This isn't strictly necessary, but it makes it harder for LSMs to
1219 * mess up.
1220 */
1221 if (current->no_new_privs)
1222 bprm->unsafe |= LSM_UNSAFE_NO_NEW_PRIVS;
1223
David Howells0bf2f3a2009-02-06 11:45:46 +00001224 n_fs = 1;
Nick Piggin2a4419b2010-08-18 04:37:33 +10001225 spin_lock(&p->fs->lock);
Oleg Nesterov437f7fd2009-04-24 01:02:45 +02001226 rcu_read_lock();
David Howells0bf2f3a2009-02-06 11:45:46 +00001227 for (t = next_thread(p); t != p; t = next_thread(t)) {
1228 if (t->fs == p->fs)
1229 n_fs++;
David Howells0bf2f3a2009-02-06 11:45:46 +00001230 }
Oleg Nesterov437f7fd2009-04-24 01:02:45 +02001231 rcu_read_unlock();
David Howells0bf2f3a2009-02-06 11:45:46 +00001232
Al Virof1191b52009-03-30 07:35:18 -04001233 if (p->fs->users > n_fs) {
David Howellsa6f76f22008-11-14 10:39:24 +11001234 bprm->unsafe |= LSM_UNSAFE_SHARE;
Al Viro498052b2009-03-30 07:20:30 -04001235 } else {
Oleg Nesterov8c652f92009-04-24 01:01:56 +02001236 res = -EAGAIN;
1237 if (!p->fs->in_exec) {
1238 p->fs->in_exec = 1;
1239 res = 1;
1240 }
Al Viro498052b2009-03-30 07:20:30 -04001241 }
Nick Piggin2a4419b2010-08-18 04:37:33 +10001242 spin_unlock(&p->fs->lock);
Al Viro498052b2009-03-30 07:20:30 -04001243
1244 return res;
David Howellsa6f76f22008-11-14 10:39:24 +11001245}
1246
Linus Torvalds1da177e2005-04-16 15:20:36 -07001247/*
1248 * Fill the binprm structure from the inode.
1249 * Check permissions, then read the first 128 (BINPRM_BUF_SIZE) bytes
David Howellsa6f76f22008-11-14 10:39:24 +11001250 *
1251 * This may be called multiple times for binary chains (scripts for example).
Linus Torvalds1da177e2005-04-16 15:20:36 -07001252 */
1253int prepare_binprm(struct linux_binprm *bprm)
1254{
David Howellsa6f76f22008-11-14 10:39:24 +11001255 umode_t mode;
Josef "Jeff" Sipek0f7fc9e2006-12-08 02:36:35 -08001256 struct inode * inode = bprm->file->f_path.dentry->d_inode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001257 int retval;
1258
1259 mode = inode->i_mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001260 if (bprm->file->f_op == NULL)
1261 return -EACCES;
1262
David Howellsa6f76f22008-11-14 10:39:24 +11001263 /* clear any previous set[ug]id data from a previous binary */
1264 bprm->cred->euid = current_euid();
1265 bprm->cred->egid = current_egid();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001266
Andy Lutomirski259e5e62012-04-12 16:47:50 -05001267 if (!(bprm->file->f_path.mnt->mnt_flags & MNT_NOSUID) &&
1268 !current->no_new_privs) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001269 /* Set-uid? */
1270 if (mode & S_ISUID) {
Eric W. Biederman9e4a36e2011-11-16 23:37:59 -08001271 if (!kuid_has_mapping(bprm->cred->user_ns, inode->i_uid))
1272 return -EPERM;
David Howellsa6f76f22008-11-14 10:39:24 +11001273 bprm->per_clear |= PER_CLEAR_ON_SETID;
1274 bprm->cred->euid = inode->i_uid;
Eric W. Biederman9e4a36e2011-11-16 23:37:59 -08001275
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)) {
Eric W. Biederman9e4a36e2011-11-16 23:37:59 -08001285 if (!kgid_has_mapping(bprm->cred->user_ns, inode->i_gid))
1286 return -EPERM;
David Howellsa6f76f22008-11-14 10:39:24 +11001287 bprm->per_clear |= PER_CLEAR_ON_SETID;
1288 bprm->cred->egid = inode->i_gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001289 }
1290 }
1291
1292 /* fill in binprm security blob */
David Howellsa6f76f22008-11-14 10:39:24 +11001293 retval = security_bprm_set_creds(bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001294 if (retval)
1295 return retval;
David Howellsa6f76f22008-11-14 10:39:24 +11001296 bprm->cred_prepared = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001297
David Howellsa6f76f22008-11-14 10:39:24 +11001298 memset(bprm->buf, 0, BINPRM_BUF_SIZE);
1299 return kernel_read(bprm->file, 0, bprm->buf, BINPRM_BUF_SIZE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001300}
1301
1302EXPORT_SYMBOL(prepare_binprm);
1303
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001304/*
1305 * Arguments are '\0' separated strings found at the location bprm->p
1306 * points to; chop off the first by relocating brpm->p to right after
1307 * the first '\0' encountered.
1308 */
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001309int remove_arg_zero(struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001310{
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001311 int ret = 0;
1312 unsigned long offset;
1313 char *kaddr;
1314 struct page *page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001315
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001316 if (!bprm->argc)
1317 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001318
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001319 do {
1320 offset = bprm->p & ~PAGE_MASK;
1321 page = get_arg_page(bprm, bprm->p, 0);
1322 if (!page) {
1323 ret = -EFAULT;
1324 goto out;
1325 }
Cong Wange8e3c3d2011-11-25 23:14:27 +08001326 kaddr = kmap_atomic(page);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001327
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001328 for (; offset < PAGE_SIZE && kaddr[offset];
1329 offset++, bprm->p++)
1330 ;
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001331
Cong Wange8e3c3d2011-11-25 23:14:27 +08001332 kunmap_atomic(kaddr);
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001333 put_arg_page(page);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001334
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001335 if (offset == PAGE_SIZE)
1336 free_arg_page(bprm, (bprm->p >> PAGE_SHIFT) - 1);
1337 } while (offset == PAGE_SIZE);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001338
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001339 bprm->p++;
1340 bprm->argc--;
1341 ret = 0;
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001342
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001343out:
1344 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001345}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001346EXPORT_SYMBOL(remove_arg_zero);
1347
1348/*
1349 * cycle the list of binary formats handler, until one recognizes the image
1350 */
1351int search_binary_handler(struct linux_binprm *bprm,struct pt_regs *regs)
1352{
Roland McGrath85f33462008-12-09 19:36:38 -08001353 unsigned int depth = bprm->recursion_depth;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001354 int try,retval;
1355 struct linux_binfmt *fmt;
Oleg Nesterov63081912012-03-30 18:26:36 +02001356 pid_t old_pid, old_vpid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001357
Linus Torvalds1da177e2005-04-16 15:20:36 -07001358 retval = security_bprm_check(bprm);
1359 if (retval)
1360 return retval;
1361
Al Viro473ae302006-04-26 14:04:08 -04001362 retval = audit_bprm(bprm);
1363 if (retval)
1364 return retval;
1365
Denys Vlasenkobb188d72011-06-29 04:13:39 +02001366 /* Need to fetch pid before load_binary changes it */
Oleg Nesterov63081912012-03-30 18:26:36 +02001367 old_pid = current->pid;
Denys Vlasenkobb188d72011-06-29 04:13:39 +02001368 rcu_read_lock();
Oleg Nesterov63081912012-03-30 18:26:36 +02001369 old_vpid = task_pid_nr_ns(current, task_active_pid_ns(current->parent));
Denys Vlasenkobb188d72011-06-29 04:13:39 +02001370 rcu_read_unlock();
1371
Linus Torvalds1da177e2005-04-16 15:20:36 -07001372 retval = -ENOENT;
1373 for (try=0; try<2; try++) {
1374 read_lock(&binfmt_lock);
Alexey Dobriyane4dc1b12007-10-16 23:26:03 -07001375 list_for_each_entry(fmt, &formats, lh) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001376 int (*fn)(struct linux_binprm *, struct pt_regs *) = fmt->load_binary;
1377 if (!fn)
1378 continue;
1379 if (!try_module_get(fmt->module))
1380 continue;
1381 read_unlock(&binfmt_lock);
1382 retval = fn(bprm, regs);
Roland McGrath85f33462008-12-09 19:36:38 -08001383 /*
1384 * Restore the depth counter to its starting value
1385 * in this call, so we don't have to rely on every
1386 * load_binary function to restore it on return.
1387 */
1388 bprm->recursion_depth = depth;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001389 if (retval >= 0) {
David Smith4ff16c22012-02-07 10:11:05 -06001390 if (depth == 0) {
1391 trace_sched_process_exec(current, old_pid, bprm);
Oleg Nesterov63081912012-03-30 18:26:36 +02001392 ptrace_event(PTRACE_EVENT_EXEC, old_vpid);
David Smith4ff16c22012-02-07 10:11:05 -06001393 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001394 put_binfmt(fmt);
1395 allow_write_access(bprm->file);
1396 if (bprm->file)
1397 fput(bprm->file);
1398 bprm->file = NULL;
1399 current->did_exec = 1;
Matt Helsley9f460802005-11-07 00:59:16 -08001400 proc_exec_connector(current);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001401 return retval;
1402 }
1403 read_lock(&binfmt_lock);
1404 put_binfmt(fmt);
1405 if (retval != -ENOEXEC || bprm->mm == NULL)
1406 break;
1407 if (!bprm->file) {
1408 read_unlock(&binfmt_lock);
1409 return retval;
1410 }
1411 }
1412 read_unlock(&binfmt_lock);
Tetsuo Handab4edf8b2011-07-26 16:08:42 -07001413#ifdef CONFIG_MODULES
Linus Torvalds1da177e2005-04-16 15:20:36 -07001414 if (retval != -ENOEXEC || bprm->mm == NULL) {
1415 break;
Johannes Berg5f4123b2008-07-09 10:28:40 +02001416 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001417#define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e))
1418 if (printable(bprm->buf[0]) &&
1419 printable(bprm->buf[1]) &&
1420 printable(bprm->buf[2]) &&
1421 printable(bprm->buf[3]))
1422 break; /* -ENOEXEC */
Tetsuo Handa91219352011-07-26 16:08:41 -07001423 if (try)
1424 break; /* -ENOEXEC */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001425 request_module("binfmt-%04x", *(unsigned short *)(&bprm->buf[2]));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001426 }
Tetsuo Handab4edf8b2011-07-26 16:08:42 -07001427#else
1428 break;
1429#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001430 }
1431 return retval;
1432}
1433
1434EXPORT_SYMBOL(search_binary_handler);
1435
1436/*
1437 * sys_execve() executes a new program.
1438 */
Oleg Nesterovba2d01622011-03-06 18:02:37 +01001439static int do_execve_common(const char *filename,
1440 struct user_arg_ptr argv,
1441 struct user_arg_ptr envp,
1442 struct pt_regs *regs)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001443{
1444 struct linux_binprm *bprm;
1445 struct file *file;
Al Viro3b125382008-04-22 05:31:30 -04001446 struct files_struct *displaced;
Oleg Nesterov8c652f92009-04-24 01:01:56 +02001447 bool clear_in_exec;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001448 int retval;
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +04001449 const struct cred *cred = current_cred();
1450
1451 /*
1452 * We move the actual failure in case of RLIMIT_NPROC excess from
1453 * set*uid() to execve() because too many poorly written programs
1454 * don't check setuid() return code. Here we additionally recheck
1455 * whether NPROC limit is still exceeded.
1456 */
1457 if ((current->flags & PF_NPROC_EXCEEDED) &&
1458 atomic_read(&cred->user->processes) > rlimit(RLIMIT_NPROC)) {
1459 retval = -EAGAIN;
1460 goto out_ret;
1461 }
1462
1463 /* We're below the limit (still or again), so we don't want to make
1464 * further execve() calls fail. */
1465 current->flags &= ~PF_NPROC_EXCEEDED;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001466
Al Viro3b125382008-04-22 05:31:30 -04001467 retval = unshare_files(&displaced);
Al Virofd8328b2008-04-22 05:11:59 -04001468 if (retval)
1469 goto out_ret;
1470
Linus Torvalds1da177e2005-04-16 15:20:36 -07001471 retval = -ENOMEM;
Oliver Neukum11b0b5a2006-03-25 03:08:13 -08001472 bprm = kzalloc(sizeof(*bprm), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001473 if (!bprm)
Al Virofd8328b2008-04-22 05:11:59 -04001474 goto out_files;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001475
Oleg Nesterova2a84742009-09-05 11:17:13 -07001476 retval = prepare_bprm_creds(bprm);
1477 if (retval)
David Howellsa6f76f22008-11-14 10:39:24 +11001478 goto out_free;
Al Viro498052b2009-03-30 07:20:30 -04001479
1480 retval = check_unsafe_exec(bprm);
Oleg Nesterov8c652f92009-04-24 01:01:56 +02001481 if (retval < 0)
Oleg Nesterova2a84742009-09-05 11:17:13 -07001482 goto out_free;
Oleg Nesterov8c652f92009-04-24 01:01:56 +02001483 clear_in_exec = retval;
Oleg Nesterova2a84742009-09-05 11:17:13 -07001484 current->in_execve = 1;
David Howellsa6f76f22008-11-14 10:39:24 +11001485
Linus Torvalds1da177e2005-04-16 15:20:36 -07001486 file = open_exec(filename);
1487 retval = PTR_ERR(file);
1488 if (IS_ERR(file))
Al Viro498052b2009-03-30 07:20:30 -04001489 goto out_unmark;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001490
1491 sched_exec();
1492
Linus Torvalds1da177e2005-04-16 15:20:36 -07001493 bprm->file = file;
1494 bprm->filename = filename;
1495 bprm->interp = filename;
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001496
1497 retval = bprm_mm_init(bprm);
1498 if (retval)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001499 goto out_file;
1500
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001501 bprm->argc = count(argv, MAX_ARG_STRINGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001502 if ((retval = bprm->argc) < 0)
David Howellsa6f76f22008-11-14 10:39:24 +11001503 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001504
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001505 bprm->envc = count(envp, MAX_ARG_STRINGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001506 if ((retval = bprm->envc) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001507 goto out;
1508
1509 retval = prepare_binprm(bprm);
1510 if (retval < 0)
1511 goto out;
1512
1513 retval = copy_strings_kernel(1, &bprm->filename, bprm);
1514 if (retval < 0)
1515 goto out;
1516
1517 bprm->exec = bprm->p;
1518 retval = copy_strings(bprm->envc, envp, bprm);
1519 if (retval < 0)
1520 goto out;
1521
1522 retval = copy_strings(bprm->argc, argv, bprm);
1523 if (retval < 0)
1524 goto out;
1525
1526 retval = search_binary_handler(bprm,regs);
David Howellsa6f76f22008-11-14 10:39:24 +11001527 if (retval < 0)
1528 goto out;
1529
1530 /* execve succeeded */
Al Viro498052b2009-03-30 07:20:30 -04001531 current->fs->in_exec = 0;
Kentaro Takedaf9ce1f12009-02-05 17:18:11 +09001532 current->in_execve = 0;
David Howellsa6f76f22008-11-14 10:39:24 +11001533 acct_update_integrals(current);
1534 free_bprm(bprm);
1535 if (displaced)
1536 put_files_struct(displaced);
1537 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001538
1539out:
Oleg Nesterov3c77f842010-11-30 20:55:34 +01001540 if (bprm->mm) {
1541 acct_arg_size(bprm, 0);
1542 mmput(bprm->mm);
1543 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001544
1545out_file:
1546 if (bprm->file) {
1547 allow_write_access(bprm->file);
1548 fput(bprm->file);
1549 }
David Howellsa6f76f22008-11-14 10:39:24 +11001550
Al Viro498052b2009-03-30 07:20:30 -04001551out_unmark:
Oleg Nesterov8c652f92009-04-24 01:01:56 +02001552 if (clear_in_exec)
1553 current->fs->in_exec = 0;
Kentaro Takedaf9ce1f12009-02-05 17:18:11 +09001554 current->in_execve = 0;
David Howellsa6f76f22008-11-14 10:39:24 +11001555
1556out_free:
Al Viro08a6fac2008-05-10 16:38:25 -04001557 free_bprm(bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001558
Al Virofd8328b2008-04-22 05:11:59 -04001559out_files:
Al Viro3b125382008-04-22 05:31:30 -04001560 if (displaced)
1561 reset_files_struct(displaced);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001562out_ret:
1563 return retval;
1564}
1565
Oleg Nesterovba2d01622011-03-06 18:02:37 +01001566int do_execve(const char *filename,
1567 const char __user *const __user *__argv,
1568 const char __user *const __user *__envp,
1569 struct pt_regs *regs)
1570{
Oleg Nesterov0e028462011-03-06 18:02:54 +01001571 struct user_arg_ptr argv = { .ptr.native = __argv };
1572 struct user_arg_ptr envp = { .ptr.native = __envp };
Oleg Nesterovba2d01622011-03-06 18:02:37 +01001573 return do_execve_common(filename, argv, envp, regs);
1574}
1575
Oleg Nesterov0e028462011-03-06 18:02:54 +01001576#ifdef CONFIG_COMPAT
1577int compat_do_execve(char *filename,
1578 compat_uptr_t __user *__argv,
1579 compat_uptr_t __user *__envp,
1580 struct pt_regs *regs)
1581{
1582 struct user_arg_ptr argv = {
1583 .is_compat = true,
1584 .ptr.compat = __argv,
1585 };
1586 struct user_arg_ptr envp = {
1587 .is_compat = true,
1588 .ptr.compat = __envp,
1589 };
1590 return do_execve_common(filename, argv, envp, regs);
1591}
1592#endif
1593
Oleg Nesterov964ee7d2009-09-23 15:56:59 -07001594void set_binfmt(struct linux_binfmt *new)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001595{
Hiroshi Shimamoto801460d2009-09-23 15:57:41 -07001596 struct mm_struct *mm = current->mm;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001597
Hiroshi Shimamoto801460d2009-09-23 15:57:41 -07001598 if (mm->binfmt)
1599 module_put(mm->binfmt->module);
1600
1601 mm->binfmt = new;
Oleg Nesterov964ee7d2009-09-23 15:56:59 -07001602 if (new)
1603 __module_get(new->module);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001604}
1605
1606EXPORT_SYMBOL(set_binfmt);
1607
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001608/*
1609 * set_dumpable converts traditional three-value dumpable to two flags and
1610 * stores them into mm->flags. It modifies lower two bits of mm->flags, but
1611 * these bits are not changed atomically. So get_dumpable can observe the
1612 * intermediate state. To avoid doing unexpected behavior, get get_dumpable
1613 * return either old dumpable or new one by paying attention to the order of
1614 * modifying the bits.
1615 *
1616 * dumpable | mm->flags (binary)
1617 * old new | initial interim final
1618 * ---------+-----------------------
1619 * 0 1 | 00 01 01
1620 * 0 2 | 00 10(*) 11
1621 * 1 0 | 01 00 00
1622 * 1 2 | 01 11 11
1623 * 2 0 | 11 10(*) 00
1624 * 2 1 | 11 11 01
1625 *
1626 * (*) get_dumpable regards interim value of 10 as 11.
1627 */
1628void set_dumpable(struct mm_struct *mm, int value)
1629{
1630 switch (value) {
Kees Cook54b50192012-07-30 14:39:18 -07001631 case SUID_DUMPABLE_DISABLED:
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001632 clear_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_ENABLED:
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001637 set_bit(MMF_DUMPABLE, &mm->flags);
1638 smp_wmb();
1639 clear_bit(MMF_DUMP_SECURELY, &mm->flags);
1640 break;
Kees Cook54b50192012-07-30 14:39:18 -07001641 case SUID_DUMPABLE_SAFE:
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001642 set_bit(MMF_DUMP_SECURELY, &mm->flags);
1643 smp_wmb();
1644 set_bit(MMF_DUMPABLE, &mm->flags);
1645 break;
1646 }
1647}
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001648
Alex Kelly10c28d92012-09-26 21:52:08 -04001649int __get_dumpable(unsigned long mm_flags)
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001650{
1651 int ret;
1652
Masami Hiramatsu30736a42010-03-05 13:44:12 -08001653 ret = mm_flags & MMF_DUMPABLE_MASK;
Kees Cook54b50192012-07-30 14:39:18 -07001654 return (ret > SUID_DUMPABLE_ENABLED) ? SUID_DUMPABLE_SAFE : ret;
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001655}
1656
Masami Hiramatsu30736a42010-03-05 13:44:12 -08001657int get_dumpable(struct mm_struct *mm)
1658{
1659 return __get_dumpable(mm->flags);
1660}