blob: 476f3ebf437ef40ddd7432200080825b7e9e992c [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>
Davidlohr Bueso615d6e82014-04-07 15:37:25 -070029#include <linux/vmacache.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070030#include <linux/stat.h>
31#include <linux/fcntl.h>
Hugh Dickinsba92a432008-07-25 01:45:43 -070032#include <linux/swap.h>
Neil Horman74aadce2007-10-16 23:26:35 -070033#include <linux/string.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070034#include <linux/init.h>
Hugh Dickinsca5b1722008-07-28 15:46:18 -070035#include <linux/pagemap.h>
Ingo Molnarcdd6c482009-09-21 12:02:48 +020036#include <linux/perf_event.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070037#include <linux/highmem.h>
38#include <linux/spinlock.h>
39#include <linux/key.h>
40#include <linux/personality.h>
41#include <linux/binfmts.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070042#include <linux/utsname.h>
Sukadev Bhattiprolu84d73782006-12-08 02:38:01 -080043#include <linux/pid_namespace.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070044#include <linux/module.h>
45#include <linux/namei.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070046#include <linux/mount.h>
47#include <linux/security.h>
48#include <linux/syscalls.h>
Jay Lan8f0ab512006-09-30 23:28:59 -070049#include <linux/tsacct_kern.h>
Matt Helsley9f460802005-11-07 00:59:16 -080050#include <linux/cn_proc.h>
Al Viro473ae302006-04-26 14:04:08 -040051#include <linux/audit.h>
Roland McGrath6341c392008-07-25 19:45:44 -070052#include <linux/tracehook.h>
Johannes Berg5f4123b2008-07-09 10:28:40 +020053#include <linux/kmod.h>
Eric Paris6110e3a2008-12-17 13:53:20 -050054#include <linux/fsnotify.h>
Al Viro5ad4e532009-03-29 19:50:06 -040055#include <linux/fs_struct.h>
Neil Horman61be2282009-09-23 15:56:58 -070056#include <linux/pipe_fs_i.h>
Ying Han3d5992d2010-10-26 14:21:23 -070057#include <linux/oom.h>
Oleg Nesterov0e028462011-03-06 18:02:54 +010058#include <linux/compat.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070059
60#include <asm/uaccess.h>
61#include <asm/mmu_context.h>
Ollie Wildb6a2fea2007-07-19 01:48:16 -070062#include <asm/tlb.h>
KAMEZAWA Hiroyuki43d2b112012-01-10 15:08:09 -080063
64#include <trace/events/task.h>
David Howellsa6f76f22008-11-14 10:39:24 +110065#include "internal.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070066
David Smith4ff16c22012-02-07 10:11:05 -060067#include <trace/events/sched.h>
68
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);
Oleg Nesterov92eaa562013-09-11 14:24:42 -070077 if (WARN_ON(!fmt->load_binary))
78 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -070079 write_lock(&binfmt_lock);
Ivan Kokshaysky74641f52009-04-30 15:08:49 -070080 insert ? list_add(&fmt->lh, &formats) :
81 list_add_tail(&fmt->lh, &formats);
Linus Torvalds1da177e2005-04-16 15:20:36 -070082 write_unlock(&binfmt_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -070083}
84
Ivan Kokshaysky74641f52009-04-30 15:08:49 -070085EXPORT_SYMBOL(__register_binfmt);
Linus Torvalds1da177e2005-04-16 15:20:36 -070086
Alexey Dobriyanf6b450d2007-10-16 23:26:04 -070087void unregister_binfmt(struct linux_binfmt * fmt)
Linus Torvalds1da177e2005-04-16 15:20:36 -070088{
Linus Torvalds1da177e2005-04-16 15:20:36 -070089 write_lock(&binfmt_lock);
Alexey Dobriyane4dc1b12007-10-16 23:26:03 -070090 list_del(&fmt->lh);
Linus Torvalds1da177e2005-04-16 15:20:36 -070091 write_unlock(&binfmt_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -070092}
93
94EXPORT_SYMBOL(unregister_binfmt);
95
96static inline void put_binfmt(struct linux_binfmt * fmt)
97{
98 module_put(fmt->module);
99}
100
Josh Triplett69369a72014-04-03 14:48:27 -0700101#ifdef CONFIG_USELIB
Linus Torvalds1da177e2005-04-16 15:20:36 -0700102/*
103 * Note that a shared library must be both readable and executable due to
104 * security reasons.
105 *
106 * Also note that we take the address to load from from the file itself.
107 */
Heiko Carstens1e7bfb22009-01-14 14:14:29 +0100108SYSCALL_DEFINE1(uselib, const char __user *, library)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700109{
Al Viro72c2d532013-09-22 16:27:52 -0400110 struct linux_binfmt *fmt;
Al Viro964bd182008-07-26 03:33:14 -0400111 struct file *file;
Jeff Layton91a27b22012-10-10 15:25:28 -0400112 struct filename *tmp = getname(library);
Al Viro964bd182008-07-26 03:33:14 -0400113 int error = PTR_ERR(tmp);
Al Viro47c805d2011-02-23 17:44:09 -0500114 static const struct open_flags uselib_flags = {
115 .open_flag = O_LARGEFILE | O_RDONLY | __FMODE_EXEC,
116 .acc_mode = MAY_READ | MAY_EXEC | MAY_OPEN,
Al Virof9652e12013-06-11 08:23:01 +0400117 .intent = LOOKUP_OPEN,
118 .lookup_flags = LOOKUP_FOLLOW,
Al Viro47c805d2011-02-23 17:44:09 -0500119 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700120
Al Viro6e8341a2009-04-06 11:16:22 -0400121 if (IS_ERR(tmp))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122 goto out;
123
Al Virof9652e12013-06-11 08:23:01 +0400124 file = do_filp_open(AT_FDCWD, tmp, &uselib_flags);
Al Viro6e8341a2009-04-06 11:16:22 -0400125 putname(tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700126 error = PTR_ERR(file);
127 if (IS_ERR(file))
128 goto out;
129
Al Viro6e8341a2009-04-06 11:16:22 -0400130 error = -EINVAL;
Al Viro496ad9a2013-01-23 17:07:38 -0500131 if (!S_ISREG(file_inode(file)->i_mode))
Al Viro6e8341a2009-04-06 11:16:22 -0400132 goto exit;
133
134 error = -EACCES;
135 if (file->f_path.mnt->mnt_flags & MNT_NOEXEC)
136 goto exit;
137
Eric Paris2a12a9d2009-12-17 21:24:21 -0500138 fsnotify_open(file);
Eric Paris6110e3a2008-12-17 13:53:20 -0500139
Linus Torvalds1da177e2005-04-16 15:20:36 -0700140 error = -ENOEXEC;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700141
Al Viro72c2d532013-09-22 16:27:52 -0400142 read_lock(&binfmt_lock);
143 list_for_each_entry(fmt, &formats, lh) {
144 if (!fmt->load_shlib)
145 continue;
146 if (!try_module_get(fmt->module))
147 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700148 read_unlock(&binfmt_lock);
Al Viro72c2d532013-09-22 16:27:52 -0400149 error = fmt->load_shlib(file);
150 read_lock(&binfmt_lock);
151 put_binfmt(fmt);
152 if (error != -ENOEXEC)
153 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700154 }
Al Viro72c2d532013-09-22 16:27:52 -0400155 read_unlock(&binfmt_lock);
Al Viro6e8341a2009-04-06 11:16:22 -0400156exit:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157 fput(file);
158out:
159 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700160}
Josh Triplett69369a72014-04-03 14:48:27 -0700161#endif /* #ifdef CONFIG_USELIB */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700163#ifdef CONFIG_MMU
Oleg Nesterovae6b5852011-03-06 18:03:11 +0100164/*
165 * The nascent bprm->mm is not visible until exec_mmap() but it can
166 * use a lot of memory, account these pages in current->mm temporary
167 * for oom_badness()->get_mm_rss(). Once exec succeeds or fails, we
168 * change the counter back via acct_arg_size(0).
169 */
Oleg Nesterov0e028462011-03-06 18:02:54 +0100170static void acct_arg_size(struct linux_binprm *bprm, unsigned long pages)
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100171{
172 struct mm_struct *mm = current->mm;
173 long diff = (long)(pages - bprm->vma_pages);
174
175 if (!mm || !diff)
176 return;
177
178 bprm->vma_pages = pages;
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100179 add_mm_counter(mm, MM_ANONPAGES, diff);
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100180}
181
Oleg Nesterov0e028462011-03-06 18:02:54 +0100182static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos,
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700183 int write)
184{
185 struct page *page;
186 int ret;
187
188#ifdef CONFIG_STACK_GROWSUP
189 if (write) {
Michal Hockod05f3162011-05-24 17:11:44 -0700190 ret = expand_downwards(bprm->vma, pos);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700191 if (ret < 0)
192 return NULL;
193 }
194#endif
195 ret = get_user_pages(current, bprm->mm, pos,
196 1, write, 1, &page, NULL);
197 if (ret <= 0)
198 return NULL;
199
200 if (write) {
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700201 unsigned long size = bprm->vma->vm_end - bprm->vma->vm_start;
Linus Torvaldsa64e7152008-03-03 10:12:14 -0800202 struct rlimit *rlim;
203
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100204 acct_arg_size(bprm, size / PAGE_SIZE);
205
Linus Torvaldsa64e7152008-03-03 10:12:14 -0800206 /*
207 * We've historically supported up to 32 pages (ARG_MAX)
208 * of argument strings even with small stacks
209 */
210 if (size <= ARG_MAX)
211 return page;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700212
213 /*
214 * Limit to 1/4-th the stack size for the argv+env strings.
215 * This ensures that:
216 * - the remaining binfmt code will not run out of stack space,
217 * - the program will have a reasonable amount of stack left
218 * to work from.
219 */
Linus Torvaldsa64e7152008-03-03 10:12:14 -0800220 rlim = current->signal->rlim;
Jiri Slabyd554ed892010-03-05 13:42:42 -0800221 if (size > ACCESS_ONCE(rlim[RLIMIT_STACK].rlim_cur) / 4) {
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700222 put_page(page);
223 return NULL;
224 }
225 }
226
227 return page;
228}
229
230static void put_arg_page(struct page *page)
231{
232 put_page(page);
233}
234
235static void free_arg_page(struct linux_binprm *bprm, int i)
236{
237}
238
239static void free_arg_pages(struct linux_binprm *bprm)
240{
241}
242
243static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos,
244 struct page *page)
245{
246 flush_cache_page(bprm->vma, pos, page_to_pfn(page));
247}
248
249static int __bprm_mm_init(struct linux_binprm *bprm)
250{
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800251 int err;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700252 struct vm_area_struct *vma = NULL;
253 struct mm_struct *mm = bprm->mm;
254
255 bprm->vma = vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL);
256 if (!vma)
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800257 return -ENOMEM;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700258
259 down_write(&mm->mmap_sem);
260 vma->vm_mm = mm;
261
262 /*
263 * Place the stack at the largest stack address the architecture
264 * supports. Later, we'll move this to an appropriate place. We don't
265 * use STACK_TOP because that can depend on attributes which aren't
266 * configured yet.
267 */
Michal Hockoaacb3d12011-07-26 16:08:40 -0700268 BUILD_BUG_ON(VM_STACK_FLAGS & VM_STACK_INCOMPLETE_SETUP);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700269 vma->vm_end = STACK_TOP_MAX;
270 vma->vm_start = vma->vm_end - PAGE_SIZE;
Cyrill Gorcunovd9104d12013-09-11 14:22:24 -0700271 vma->vm_flags = VM_SOFTDIRTY | VM_STACK_FLAGS | VM_STACK_INCOMPLETE_SETUP;
Coly Li3ed75eb2007-10-18 23:39:15 -0700272 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
Rik van Riel5beb4932010-03-05 13:42:07 -0800273 INIT_LIST_HEAD(&vma->anon_vma_chain);
Tavis Ormandy462e635e2010-12-09 15:29:42 +0100274
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700275 err = insert_vm_struct(mm, vma);
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800276 if (err)
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700277 goto err;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700278
279 mm->stack_vm = mm->total_vm = 1;
280 up_write(&mm->mmap_sem);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700281 bprm->p = vma->vm_end - sizeof(void *);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700282 return 0;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700283err:
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800284 up_write(&mm->mmap_sem);
285 bprm->vma = NULL;
286 kmem_cache_free(vm_area_cachep, vma);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700287 return err;
288}
289
290static bool valid_arg_len(struct linux_binprm *bprm, long len)
291{
292 return len <= MAX_ARG_STRLEN;
293}
294
295#else
296
Oleg Nesterov0e028462011-03-06 18:02:54 +0100297static inline void acct_arg_size(struct linux_binprm *bprm, unsigned long pages)
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100298{
299}
300
Oleg Nesterov0e028462011-03-06 18:02:54 +0100301static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos,
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700302 int write)
303{
304 struct page *page;
305
306 page = bprm->page[pos / PAGE_SIZE];
307 if (!page && write) {
308 page = alloc_page(GFP_HIGHUSER|__GFP_ZERO);
309 if (!page)
310 return NULL;
311 bprm->page[pos / PAGE_SIZE] = page;
312 }
313
314 return page;
315}
316
317static void put_arg_page(struct page *page)
318{
319}
320
321static void free_arg_page(struct linux_binprm *bprm, int i)
322{
323 if (bprm->page[i]) {
324 __free_page(bprm->page[i]);
325 bprm->page[i] = NULL;
326 }
327}
328
329static void free_arg_pages(struct linux_binprm *bprm)
330{
331 int i;
332
333 for (i = 0; i < MAX_ARG_PAGES; i++)
334 free_arg_page(bprm, i);
335}
336
337static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos,
338 struct page *page)
339{
340}
341
342static int __bprm_mm_init(struct linux_binprm *bprm)
343{
344 bprm->p = PAGE_SIZE * MAX_ARG_PAGES - sizeof(void *);
345 return 0;
346}
347
348static bool valid_arg_len(struct linux_binprm *bprm, long len)
349{
350 return len <= bprm->p;
351}
352
353#endif /* CONFIG_MMU */
354
355/*
356 * Create a new mm_struct and populate it with a temporary stack
357 * vm_area_struct. We don't have enough context at this point to set the stack
358 * flags, permissions, and offset, so we use temporary values. We'll update
359 * them later in setup_arg_pages().
360 */
Yuanhan Liu9cc64ce2013-02-20 13:16:01 +1100361static int bprm_mm_init(struct linux_binprm *bprm)
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700362{
363 int err;
364 struct mm_struct *mm = NULL;
365
366 bprm->mm = mm = mm_alloc();
367 err = -ENOMEM;
368 if (!mm)
369 goto err;
370
371 err = init_new_context(current, mm);
372 if (err)
373 goto err;
374
375 err = __bprm_mm_init(bprm);
376 if (err)
377 goto err;
378
379 return 0;
380
381err:
382 if (mm) {
383 bprm->mm = NULL;
384 mmdrop(mm);
385 }
386
387 return err;
388}
389
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100390struct user_arg_ptr {
Oleg Nesterov0e028462011-03-06 18:02:54 +0100391#ifdef CONFIG_COMPAT
392 bool is_compat;
393#endif
394 union {
395 const char __user *const __user *native;
396#ifdef CONFIG_COMPAT
Al Viro38b983b2012-09-30 13:38:55 -0400397 const compat_uptr_t __user *compat;
Oleg Nesterov0e028462011-03-06 18:02:54 +0100398#endif
399 } ptr;
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100400};
401
402static const char __user *get_user_arg_ptr(struct user_arg_ptr argv, int nr)
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100403{
Oleg Nesterov0e028462011-03-06 18:02:54 +0100404 const char __user *native;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100405
Oleg Nesterov0e028462011-03-06 18:02:54 +0100406#ifdef CONFIG_COMPAT
407 if (unlikely(argv.is_compat)) {
408 compat_uptr_t compat;
409
410 if (get_user(compat, argv.ptr.compat + nr))
411 return ERR_PTR(-EFAULT);
412
413 return compat_ptr(compat);
414 }
415#endif
416
417 if (get_user(native, argv.ptr.native + nr))
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100418 return ERR_PTR(-EFAULT);
419
Oleg Nesterov0e028462011-03-06 18:02:54 +0100420 return native;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100421}
422
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423/*
424 * count() counts the number of strings in array ARGV.
425 */
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100426static int count(struct user_arg_ptr argv, int max)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427{
428 int i = 0;
429
Oleg Nesterov0e028462011-03-06 18:02:54 +0100430 if (argv.ptr.native != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700431 for (;;) {
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100432 const char __user *p = get_user_arg_ptr(argv, i);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434 if (!p)
435 break;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100436
437 if (IS_ERR(p))
438 return -EFAULT;
439
Xi Wang6d92d4f2013-01-11 14:31:48 -0800440 if (i >= max)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441 return -E2BIG;
Xi Wang6d92d4f2013-01-11 14:31:48 -0800442 ++i;
Roland McGrath9aea5a62010-09-07 19:37:06 -0700443
444 if (fatal_signal_pending(current))
445 return -ERESTARTNOHAND;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700446 cond_resched();
447 }
448 }
449 return i;
450}
451
452/*
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700453 * 'copy_strings()' copies argument/environment strings from the old
454 * processes's memory to the new process's stack. The call to get_user_pages()
455 * ensures the destination page is created and not swapped out.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456 */
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100457static int copy_strings(int argc, struct user_arg_ptr argv,
Adrian Bunk75c96f82005-05-05 16:16:09 -0700458 struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700459{
460 struct page *kmapped_page = NULL;
461 char *kaddr = NULL;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700462 unsigned long kpos = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463 int ret;
464
465 while (argc-- > 0) {
David Howellsd7627462010-08-17 23:52:56 +0100466 const char __user *str;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467 int len;
468 unsigned long pos;
469
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100470 ret = -EFAULT;
471 str = get_user_arg_ptr(argv, argc);
472 if (IS_ERR(str))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100475 len = strnlen_user(str, MAX_ARG_STRLEN);
476 if (!len)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477 goto out;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100478
479 ret = -E2BIG;
480 if (!valid_arg_len(bprm, len))
481 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700483 /* We're going to work our way backwords. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484 pos = bprm->p;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700485 str += len;
486 bprm->p -= len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487
488 while (len > 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489 int offset, bytes_to_copy;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490
Roland McGrath9aea5a62010-09-07 19:37:06 -0700491 if (fatal_signal_pending(current)) {
492 ret = -ERESTARTNOHAND;
493 goto out;
494 }
Roland McGrath7993bc12010-09-07 19:36:28 -0700495 cond_resched();
496
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497 offset = pos % PAGE_SIZE;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700498 if (offset == 0)
499 offset = PAGE_SIZE;
500
501 bytes_to_copy = offset;
502 if (bytes_to_copy > len)
503 bytes_to_copy = len;
504
505 offset -= bytes_to_copy;
506 pos -= bytes_to_copy;
507 str -= bytes_to_copy;
508 len -= bytes_to_copy;
509
510 if (!kmapped_page || kpos != (pos & PAGE_MASK)) {
511 struct page *page;
512
513 page = get_arg_page(bprm, pos, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514 if (!page) {
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700515 ret = -E2BIG;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516 goto out;
517 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700519 if (kmapped_page) {
520 flush_kernel_dcache_page(kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521 kunmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700522 put_arg_page(kmapped_page);
523 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700524 kmapped_page = page;
525 kaddr = kmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700526 kpos = pos & PAGE_MASK;
527 flush_arg_page(bprm, kpos, kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528 }
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700529 if (copy_from_user(kaddr+offset, str, bytes_to_copy)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700530 ret = -EFAULT;
531 goto out;
532 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533 }
534 }
535 ret = 0;
536out:
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700537 if (kmapped_page) {
538 flush_kernel_dcache_page(kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539 kunmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700540 put_arg_page(kmapped_page);
541 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700542 return ret;
543}
544
545/*
546 * Like copy_strings, but get argv and its values from kernel memory.
547 */
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100548int copy_strings_kernel(int argc, const char *const *__argv,
David Howellsd7627462010-08-17 23:52:56 +0100549 struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550{
551 int r;
552 mm_segment_t oldfs = get_fs();
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100553 struct user_arg_ptr argv = {
Oleg Nesterov0e028462011-03-06 18:02:54 +0100554 .ptr.native = (const char __user *const __user *)__argv,
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100555 };
556
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557 set_fs(KERNEL_DS);
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100558 r = copy_strings(argc, argv, bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700559 set_fs(oldfs);
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100560
Linus Torvalds1da177e2005-04-16 15:20:36 -0700561 return r;
562}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700563EXPORT_SYMBOL(copy_strings_kernel);
564
565#ifdef CONFIG_MMU
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700566
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567/*
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700568 * During bprm_mm_init(), we create a temporary stack at STACK_TOP_MAX. Once
569 * the binfmt code determines where the new stack should reside, we shift it to
570 * its final location. The process proceeds as follows:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571 *
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700572 * 1) Use shift to calculate the new vma endpoints.
573 * 2) Extend vma to cover both the old and new ranges. This ensures the
574 * arguments passed to subsequent functions are consistent.
575 * 3) Move vma's page tables to the new range.
576 * 4) Free up any cleared pgd range.
577 * 5) Shrink the vma to cover only the new range.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578 */
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700579static int shift_arg_pages(struct vm_area_struct *vma, unsigned long shift)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580{
581 struct mm_struct *mm = vma->vm_mm;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700582 unsigned long old_start = vma->vm_start;
583 unsigned long old_end = vma->vm_end;
584 unsigned long length = old_end - old_start;
585 unsigned long new_start = old_start - shift;
586 unsigned long new_end = old_end - shift;
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700587 struct mmu_gather tlb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700588
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700589 BUG_ON(new_start > new_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700590
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700591 /*
592 * ensure there are no vmas between where we want to go
593 * and where we are
594 */
595 if (vma != find_vma(mm, new_start))
596 return -EFAULT;
597
598 /*
599 * cover the whole range: [new_start, old_end)
600 */
Rik van Riel5beb4932010-03-05 13:42:07 -0800601 if (vma_adjust(vma, new_start, old_end, vma->vm_pgoff, NULL))
602 return -ENOMEM;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700603
604 /*
605 * move the page tables downwards, on failure we rely on
606 * process cleanup to remove whatever mess we made.
607 */
608 if (length != move_page_tables(vma, old_start,
Michel Lespinasse38a76012012-10-08 16:31:50 -0700609 vma, new_start, length, false))
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700610 return -ENOMEM;
611
612 lru_add_drain();
Linus Torvalds2b047252013-08-15 11:42:25 -0700613 tlb_gather_mmu(&tlb, mm, old_start, old_end);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700614 if (new_end > old_start) {
615 /*
616 * when the old and new regions overlap clear from new_end.
617 */
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700618 free_pgd_range(&tlb, new_end, old_end, new_end,
Hugh Dickins6ee86302013-04-29 15:07:44 -0700619 vma->vm_next ? vma->vm_next->vm_start : USER_PGTABLES_CEILING);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700620 } else {
621 /*
622 * otherwise, clean from old_start; this is done to not touch
623 * the address space in [new_end, old_start) some architectures
624 * have constraints on va-space that make this illegal (IA64) -
625 * for the others its just a little faster.
626 */
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700627 free_pgd_range(&tlb, old_start, old_end, new_end,
Hugh Dickins6ee86302013-04-29 15:07:44 -0700628 vma->vm_next ? vma->vm_next->vm_start : USER_PGTABLES_CEILING);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629 }
Linus Torvalds2b047252013-08-15 11:42:25 -0700630 tlb_finish_mmu(&tlb, old_start, old_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700631
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700632 /*
Rik van Riel5beb4932010-03-05 13:42:07 -0800633 * Shrink the vma to just the new range. Always succeeds.
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700634 */
635 vma_adjust(vma, new_start, new_end, vma->vm_pgoff, NULL);
636
637 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700638}
639
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700640/*
641 * Finalizes the stack vm_area_struct. The flags and permissions are updated,
642 * the stack is optionally relocated, and some extra space is added.
643 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700644int setup_arg_pages(struct linux_binprm *bprm,
645 unsigned long stack_top,
646 int executable_stack)
647{
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700648 unsigned long ret;
649 unsigned long stack_shift;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650 struct mm_struct *mm = current->mm;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700651 struct vm_area_struct *vma = bprm->vma;
652 struct vm_area_struct *prev = NULL;
653 unsigned long vm_flags;
654 unsigned long stack_base;
Michael Neuling803bf5e2010-02-10 13:56:42 -0800655 unsigned long stack_size;
656 unsigned long stack_expand;
657 unsigned long rlim_stack;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658
659#ifdef CONFIG_STACK_GROWSUP
Linus Torvalds1da177e2005-04-16 15:20:36 -0700660 /* Limit stack size to 1GB */
Jiri Slabyd554ed892010-03-05 13:42:42 -0800661 stack_base = rlimit_max(RLIMIT_STACK);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700662 if (stack_base > (1 << 30))
663 stack_base = 1 << 30;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700665 /* Make sure we didn't let the argument array grow too large. */
666 if (vma->vm_end - vma->vm_start > stack_base)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700667 return -ENOMEM;
668
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700669 stack_base = PAGE_ALIGN(stack_top - stack_base);
670
671 stack_shift = vma->vm_start - stack_base;
672 mm->arg_start = bprm->p - stack_shift;
673 bprm->p = vma->vm_end - stack_shift;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674#else
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700675 stack_top = arch_align_stack(stack_top);
676 stack_top = PAGE_ALIGN(stack_top);
Roland McGrath1b528182010-09-07 19:35:49 -0700677
678 if (unlikely(stack_top < mmap_min_addr) ||
679 unlikely(vma->vm_end - vma->vm_start >= stack_top - mmap_min_addr))
680 return -ENOMEM;
681
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700682 stack_shift = vma->vm_end - stack_top;
683
684 bprm->p -= stack_shift;
685 mm->arg_start = bprm->p;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700686#endif
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700687
688 if (bprm->loader)
689 bprm->loader -= stack_shift;
690 bprm->exec -= stack_shift;
691
692 down_write(&mm->mmap_sem);
Hugh Dickins96a8e132008-07-10 21:19:20 +0100693 vm_flags = VM_STACK_FLAGS;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700694
695 /*
696 * Adjust stack execute permissions; explicitly enable for
697 * EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X and leave alone
698 * (arch default) otherwise.
699 */
700 if (unlikely(executable_stack == EXSTACK_ENABLE_X))
701 vm_flags |= VM_EXEC;
702 else if (executable_stack == EXSTACK_DISABLE_X)
703 vm_flags &= ~VM_EXEC;
704 vm_flags |= mm->def_flags;
Mel Gormana8bef8ff2010-05-24 14:32:24 -0700705 vm_flags |= VM_STACK_INCOMPLETE_SETUP;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700706
707 ret = mprotect_fixup(vma, &prev, vma->vm_start, vma->vm_end,
708 vm_flags);
709 if (ret)
710 goto out_unlock;
711 BUG_ON(prev != vma);
712
713 /* Move stack pages down in memory. */
714 if (stack_shift) {
715 ret = shift_arg_pages(vma, stack_shift);
Anton Blanchardfc63cf22009-11-11 14:26:48 -0800716 if (ret)
717 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718 }
719
Mel Gormana8bef8ff2010-05-24 14:32:24 -0700720 /* mprotect_fixup is overkill to remove the temporary stack flags */
721 vma->vm_flags &= ~VM_STACK_INCOMPLETE_SETUP;
722
Michael Neuling5ef097d2010-03-05 13:42:57 -0800723 stack_expand = 131072UL; /* randomly 32*4k (or 2*64k) pages */
Michael Neuling803bf5e2010-02-10 13:56:42 -0800724 stack_size = vma->vm_end - vma->vm_start;
725 /*
726 * Align this down to a page boundary as expand_stack
727 * will align it up.
728 */
729 rlim_stack = rlimit(RLIMIT_STACK) & PAGE_MASK;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700730#ifdef CONFIG_STACK_GROWSUP
Michael Neuling803bf5e2010-02-10 13:56:42 -0800731 if (stack_size + stack_expand > rlim_stack)
732 stack_base = vma->vm_start + rlim_stack;
733 else
734 stack_base = vma->vm_end + stack_expand;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700735#else
Michael Neuling803bf5e2010-02-10 13:56:42 -0800736 if (stack_size + stack_expand > rlim_stack)
737 stack_base = vma->vm_end - rlim_stack;
738 else
739 stack_base = vma->vm_start - stack_expand;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700740#endif
Eric B Munson3af9e852010-05-18 15:30:49 +0100741 current->mm->start_stack = bprm->p;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700742 ret = expand_stack(vma, stack_base);
743 if (ret)
744 ret = -EFAULT;
745
746out_unlock:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747 up_write(&mm->mmap_sem);
Anton Blanchardfc63cf22009-11-11 14:26:48 -0800748 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750EXPORT_SYMBOL(setup_arg_pages);
751
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752#endif /* CONFIG_MMU */
753
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -0800754static struct file *do_open_exec(struct filename *name)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756 struct file *file;
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200757 int err;
Al Viro47c805d2011-02-23 17:44:09 -0500758 static const struct open_flags open_exec_flags = {
759 .open_flag = O_LARGEFILE | O_RDONLY | __FMODE_EXEC,
760 .acc_mode = MAY_EXEC | MAY_OPEN,
Al Virof9652e12013-06-11 08:23:01 +0400761 .intent = LOOKUP_OPEN,
762 .lookup_flags = LOOKUP_FOLLOW,
Al Viro47c805d2011-02-23 17:44:09 -0500763 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -0800765 file = do_filp_open(AT_FDCWD, name, &open_exec_flags);
Al Viro6e8341a2009-04-06 11:16:22 -0400766 if (IS_ERR(file))
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200767 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200769 err = -EACCES;
Al Viro496ad9a2013-01-23 17:07:38 -0500770 if (!S_ISREG(file_inode(file)->i_mode))
Al Viro6e8341a2009-04-06 11:16:22 -0400771 goto exit;
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200772
Al Viro6e8341a2009-04-06 11:16:22 -0400773 if (file->f_path.mnt->mnt_flags & MNT_NOEXEC)
774 goto exit;
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200775
Eric Paris2a12a9d2009-12-17 21:24:21 -0500776 fsnotify_open(file);
Eric Paris6110e3a2008-12-17 13:53:20 -0500777
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200778 err = deny_write_access(file);
Al Viro6e8341a2009-04-06 11:16:22 -0400779 if (err)
780 goto exit;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781
Al Viro6e8341a2009-04-06 11:16:22 -0400782out:
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200783 return file;
784
Al Viro6e8341a2009-04-06 11:16:22 -0400785exit:
786 fput(file);
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200787 return ERR_PTR(err);
788}
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -0800789
790struct file *open_exec(const char *name)
791{
792 struct filename tmp = { .name = name };
793 return do_open_exec(&tmp);
794}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795EXPORT_SYMBOL(open_exec);
796
Mimi Zohar6777d772009-08-21 14:32:48 -0400797int kernel_read(struct file *file, loff_t offset,
798 char *addr, unsigned long count)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799{
800 mm_segment_t old_fs;
801 loff_t pos = offset;
802 int result;
803
804 old_fs = get_fs();
805 set_fs(get_ds());
806 /* The cast to a user pointer is valid due to the set_fs() */
807 result = vfs_read(file, (void __user *)addr, count, &pos);
808 set_fs(old_fs);
809 return result;
810}
811
812EXPORT_SYMBOL(kernel_read);
813
Al Viro3dc20cb2013-04-13 20:31:37 -0400814ssize_t read_code(struct file *file, unsigned long addr, loff_t pos, size_t len)
815{
Al Viroec695572014-02-04 19:08:21 -0500816 ssize_t res = vfs_read(file, (void __user *)addr, len, &pos);
Al Viro3dc20cb2013-04-13 20:31:37 -0400817 if (res > 0)
818 flush_icache_range(addr, addr + len);
819 return res;
820}
821EXPORT_SYMBOL(read_code);
822
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823static int exec_mmap(struct mm_struct *mm)
824{
825 struct task_struct *tsk;
Davidlohr Bueso615d6e82014-04-07 15:37:25 -0700826 struct mm_struct *old_mm, *active_mm;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827
828 /* Notify parent that we're no longer interested in the old VM */
829 tsk = current;
830 old_mm = current->mm;
831 mm_release(tsk, old_mm);
832
833 if (old_mm) {
Konstantin Khlebnikov4fe7efd2012-06-20 12:53:01 -0700834 sync_mm_rss(old_mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835 /*
836 * Make sure that if there is a core dump in progress
837 * for the old mm, we get out and die instead of going
838 * through with the exec. We must hold mmap_sem around
Oleg Nesterov999d9fc2008-07-25 01:47:41 -0700839 * checking core_state and changing tsk->mm.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700840 */
841 down_read(&old_mm->mmap_sem);
Oleg Nesterov999d9fc2008-07-25 01:47:41 -0700842 if (unlikely(old_mm->core_state)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700843 up_read(&old_mm->mmap_sem);
844 return -EINTR;
845 }
846 }
847 task_lock(tsk);
848 active_mm = tsk->active_mm;
849 tsk->mm = mm;
850 tsk->active_mm = mm;
851 activate_mm(active_mm, mm);
Davidlohr Bueso615d6e82014-04-07 15:37:25 -0700852 tsk->mm->vmacache_seqnum = 0;
853 vmacache_flush(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700854 task_unlock(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855 if (old_mm) {
856 up_read(&old_mm->mmap_sem);
Eric Sesterhenn7dddb122006-04-01 01:13:38 +0200857 BUG_ON(active_mm != old_mm);
Oleg Nesterov701085b2012-03-19 17:04:01 +0100858 setmax_mm_hiwater_rss(&tsk->signal->maxrss, old_mm);
Balbir Singh31a78f22008-09-28 23:09:31 +0100859 mm_update_next_owner(old_mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860 mmput(old_mm);
861 return 0;
862 }
863 mmdrop(active_mm);
864 return 0;
865}
866
867/*
868 * This function makes sure the current process has its own signal table,
869 * so that flush_signal_handlers can later reset the handlers without
870 * disturbing other processes. (Other processes might share the signal
871 * table via the CLONE_SIGHAND option to clone().)
872 */
Arjan van de Ven858119e2006-01-14 13:20:43 -0800873static int de_thread(struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874{
875 struct signal_struct *sig = tsk->signal;
Oleg Nesterovb2c903b2007-10-16 23:27:22 -0700876 struct sighand_struct *oldsighand = tsk->sighand;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877 spinlock_t *lock = &oldsighand->siglock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700879 if (thread_group_empty(tsk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700880 goto no_thread_group;
881
882 /*
883 * Kill all other threads in the thread group.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700884 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700885 spin_lock_irq(lock);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -0800886 if (signal_group_exit(sig)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887 /*
888 * Another group action in progress, just
889 * return so that the signal is processed.
890 */
891 spin_unlock_irq(lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700892 return -EAGAIN;
893 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700894
Oleg Nesterovd3441932010-05-26 14:43:11 -0700895 sig->group_exit_task = tsk;
896 sig->notify_count = zap_other_threads(tsk);
897 if (!thread_group_leader(tsk))
898 sig->notify_count--;
899
900 while (sig->notify_count) {
Oleg Nesterovd5bbd432012-10-08 19:13:01 +0200901 __set_current_state(TASK_KILLABLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902 spin_unlock_irq(lock);
903 schedule();
Oleg Nesterovd5bbd432012-10-08 19:13:01 +0200904 if (unlikely(__fatal_signal_pending(tsk)))
905 goto killed;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700906 spin_lock_irq(lock);
907 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700908 spin_unlock_irq(lock);
909
910 /*
911 * At this point all other threads have exited, all we have to
912 * do is to wait for the thread group leader to become inactive,
913 * and to assume its PID:
914 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700915 if (!thread_group_leader(tsk)) {
Oleg Nesterov81879262008-12-01 14:18:16 -0800916 struct task_struct *leader = tsk->group_leader;
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700917
Oleg Nesterov2800d8d2008-04-30 00:53:12 -0700918 sig->notify_count = -1; /* for exit_notify() */
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700919 for (;;) {
Oleg Nesterove56fb282013-04-30 15:28:20 -0700920 threadgroup_change_begin(tsk);
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700921 write_lock_irq(&tasklist_lock);
922 if (likely(leader->exit_state))
923 break;
Oleg Nesterovd5bbd432012-10-08 19:13:01 +0200924 __set_current_state(TASK_KILLABLE);
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700925 write_unlock_irq(&tasklist_lock);
Oleg Nesterove56fb282013-04-30 15:28:20 -0700926 threadgroup_change_end(tsk);
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700927 schedule();
Oleg Nesterovd5bbd432012-10-08 19:13:01 +0200928 if (unlikely(__fatal_signal_pending(tsk)))
929 goto killed;
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700930 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700931
Roland McGrathf5e90282006-04-10 22:54:16 -0700932 /*
933 * The only record we have of the real-time age of a
934 * process, regardless of execs it's done, is start_time.
935 * All the past CPU time is accumulated in signal_struct
936 * from sister threads now dead. But in this non-leader
937 * exec, nothing survives from the original leader thread,
938 * whose birth marks the true age of this process now.
939 * When we take on its identity by switching to its PID, we
940 * also take its birthdate (always earlier than our own).
941 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700942 tsk->start_time = leader->start_time;
Oleg Nesterov266b7a02013-07-03 15:08:35 -0700943 tsk->real_start_time = leader->real_start_time;
Roland McGrathf5e90282006-04-10 22:54:16 -0700944
Pavel Emelyanovbac0abd2007-10-18 23:40:18 -0700945 BUG_ON(!same_thread_group(leader, tsk));
946 BUG_ON(has_group_leader_pid(tsk));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700947 /*
948 * An exec() starts a new thread group with the
949 * TGID of the previous thread group. Rehash the
950 * two threads with a switched PID, and release
951 * the former thread group leader:
952 */
Eric W. Biedermand73d6522006-03-28 16:11:03 -0800953
954 /* Become a process group leader with the old leader's pid.
Eric W. Biedermanc18258c2006-09-27 01:51:06 -0700955 * The old leader becomes a thread of the this thread group.
956 * Note: The old leader also uses this pid until release_task
Eric W. Biedermand73d6522006-03-28 16:11:03 -0800957 * is called. Odd but simple and correct.
958 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700959 tsk->pid = leader->pid;
Oleg Nesterov3f418542013-07-03 15:08:25 -0700960 change_pid(tsk, PIDTYPE_PID, task_pid(leader));
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700961 transfer_pid(leader, tsk, PIDTYPE_PGID);
962 transfer_pid(leader, tsk, PIDTYPE_SID);
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -0800963
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700964 list_replace_rcu(&leader->tasks, &tsk->tasks);
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -0800965 list_replace_init(&leader->sibling, &tsk->sibling);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700966
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700967 tsk->group_leader = tsk;
968 leader->group_leader = tsk;
Eric W. Biedermande12a782006-04-10 17:16:49 -0600969
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700970 tsk->exit_signal = SIGCHLD;
Oleg Nesterov087806b2011-06-22 23:10:26 +0200971 leader->exit_signal = -1;
Oleg Nesterov962b5642005-11-23 13:37:43 -0800972
973 BUG_ON(leader->exit_state != EXIT_ZOMBIE);
974 leader->exit_state = EXIT_DEAD;
Oleg Nesteroveac1b5e2011-07-21 20:00:43 +0200975
976 /*
977 * We are going to release_task()->ptrace_unlink() silently,
978 * the tracer can sleep in do_wait(). EXIT_DEAD guarantees
979 * the tracer wont't block again waiting for this thread.
980 */
981 if (unlikely(leader->ptrace))
982 __wake_up_parent(leader, leader->parent);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700983 write_unlock_irq(&tasklist_lock);
Oleg Nesterove56fb282013-04-30 15:28:20 -0700984 threadgroup_change_end(tsk);
Oleg Nesterov81879262008-12-01 14:18:16 -0800985
986 release_task(leader);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -0800987 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700988
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700989 sig->group_exit_task = NULL;
990 sig->notify_count = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700991
992no_thread_group:
Oleg Nesterove6368252012-03-19 17:03:22 +0100993 /* we have changed execution domain */
994 tsk->exit_signal = SIGCHLD;
995
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996 exit_itimers(sig);
Oleg Nesterovcbaffba2008-05-26 20:55:42 +0400997 flush_itimer_signals();
Oleg Nesterov329f7db2005-11-07 21:12:43 +0300998
Oleg Nesterovb2c903b2007-10-16 23:27:22 -0700999 if (atomic_read(&oldsighand->count) != 1) {
1000 struct sighand_struct *newsighand;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001 /*
Oleg Nesterovb2c903b2007-10-16 23:27:22 -07001002 * This ->sighand is shared with the CLONE_SIGHAND
1003 * but not CLONE_THREAD task, switch to the new one.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001004 */
Oleg Nesterovb2c903b2007-10-16 23:27:22 -07001005 newsighand = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
1006 if (!newsighand)
1007 return -ENOMEM;
1008
Linus Torvalds1da177e2005-04-16 15:20:36 -07001009 atomic_set(&newsighand->count, 1);
1010 memcpy(newsighand->action, oldsighand->action,
1011 sizeof(newsighand->action));
1012
1013 write_lock_irq(&tasklist_lock);
1014 spin_lock(&oldsighand->siglock);
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001015 rcu_assign_pointer(tsk->sighand, newsighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001016 spin_unlock(&oldsighand->siglock);
1017 write_unlock_irq(&tasklist_lock);
1018
Davide Libenzifba2afa2007-05-10 22:23:13 -07001019 __cleanup_sighand(oldsighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001020 }
1021
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001022 BUG_ON(!thread_group_leader(tsk));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001023 return 0;
Oleg Nesterovd5bbd432012-10-08 19:13:01 +02001024
1025killed:
1026 /* protects against exit_notify() and __exit_signal() */
1027 read_lock(&tasklist_lock);
1028 sig->group_exit_task = NULL;
1029 sig->notify_count = 0;
1030 read_unlock(&tasklist_lock);
1031 return -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032}
Oleg Nesterov0840a902007-10-16 23:27:22 -07001033
Andrew Morton59714d62008-02-04 22:27:21 -08001034char *get_task_comm(char *buf, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035{
1036 /* buf must be at least sizeof(tsk->comm) in size */
1037 task_lock(tsk);
1038 strncpy(buf, tsk->comm, sizeof(tsk->comm));
1039 task_unlock(tsk);
Andrew Morton59714d62008-02-04 22:27:21 -08001040 return buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001041}
J Freyensee7d74f492011-05-06 16:56:47 -07001042EXPORT_SYMBOL_GPL(get_task_comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001043
Al Viro6a6d27d2012-08-21 09:56:33 -04001044/*
1045 * These functions flushes out all traces of the currently running executable
1046 * so that a new one can be started
1047 */
1048
Oleg Nesterov23aebe12014-04-07 15:38:39 -07001049void set_task_comm(struct task_struct *tsk, const char *buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001050{
1051 task_lock(tsk);
KAMEZAWA Hiroyuki43d2b112012-01-10 15:08:09 -08001052 trace_task_rename(tsk, buf);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001053 strlcpy(tsk->comm, buf, sizeof(tsk->comm));
1054 task_unlock(tsk);
Ingo Molnarcdd6c482009-09-21 12:02:48 +02001055 perf_event_comm(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001056}
1057
1058int flush_old_exec(struct linux_binprm * bprm)
1059{
Linus Torvalds221af7f2010-01-28 22:14:42 -08001060 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061
1062 /*
1063 * Make sure we have a private signal table and that
1064 * we are unassociated from the previous thread group.
1065 */
1066 retval = de_thread(current);
1067 if (retval)
1068 goto out;
1069
Matt Helsley925d1c42008-04-29 01:01:36 -07001070 set_mm_exe_file(bprm->mm, bprm->file);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001072 * Release all of the old mmap stuff
1073 */
Oleg Nesterov3c77f842010-11-30 20:55:34 +01001074 acct_arg_size(bprm, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075 retval = exec_mmap(bprm->mm);
1076 if (retval)
Al Virofd8328b2008-04-22 05:11:59 -04001077 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078
1079 bprm->mm = NULL; /* We're using it now */
Linus Torvalds7ab02af2010-02-02 12:37:44 -08001080
Mathias Krausedac853a2011-06-09 20:05:18 +02001081 set_fs(USER_DS);
Zhang Yib88fae62014-01-23 15:55:57 -08001082 current->flags &= ~(PF_RANDOMIZE | PF_FORKNOEXEC | PF_KTHREAD |
1083 PF_NOFREEZE | PF_NO_SETAFFINITY);
Linus Torvalds7ab02af2010-02-02 12:37:44 -08001084 flush_thread();
1085 current->personality &= ~bprm->per_clear;
1086
Linus Torvalds221af7f2010-01-28 22:14:42 -08001087 return 0;
1088
1089out:
1090 return retval;
1091}
1092EXPORT_SYMBOL(flush_old_exec);
1093
Al Viro1b5d7832011-06-19 12:49:47 -04001094void would_dump(struct linux_binprm *bprm, struct file *file)
1095{
Al Viro496ad9a2013-01-23 17:07:38 -05001096 if (inode_permission(file_inode(file), MAY_READ) < 0)
Al Viro1b5d7832011-06-19 12:49:47 -04001097 bprm->interp_flags |= BINPRM_FLAGS_ENFORCE_NONDUMP;
1098}
1099EXPORT_SYMBOL(would_dump);
1100
Linus Torvalds221af7f2010-01-28 22:14:42 -08001101void setup_new_exec(struct linux_binprm * bprm)
1102{
Linus Torvalds221af7f2010-01-28 22:14:42 -08001103 arch_pick_mmap_layout(current->mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001104
1105 /* This is the point of no return */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106 current->sas_ss_sp = current->sas_ss_size = 0;
1107
Eric W. Biederman8e96e3b2012-03-03 21:17:15 -08001108 if (uid_eq(current_euid(), current_uid()) && gid_eq(current_egid(), current_gid()))
Kees Cooke579d2c2013-02-27 17:03:15 -08001109 set_dumpable(current->mm, SUID_DUMP_USER);
Alan Coxd6e71142005-06-23 00:09:43 -07001110 else
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001111 set_dumpable(current->mm, suid_dumpable);
Alan Coxd6e71142005-06-23 00:09:43 -07001112
Oleg Nesterov23aebe12014-04-07 15:38:39 -07001113 set_task_comm(current, kbasename(bprm->filename));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001114
Benjamin Herrenschmidt0551fbd2006-02-28 16:59:19 -08001115 /* Set the new mm task size. We have to do that late because it may
1116 * depend on TIF_32BIT which is only updated in flush_thread() on
1117 * some architectures like powerpc
1118 */
1119 current->mm->task_size = TASK_SIZE;
1120
David Howellsa6f76f22008-11-14 10:39:24 +11001121 /* install the new credentials */
Eric W. Biederman8e96e3b2012-03-03 21:17:15 -08001122 if (!uid_eq(bprm->cred->uid, current_euid()) ||
1123 !gid_eq(bprm->cred->gid, current_egid())) {
Marcel Holtmannd2d56c52007-08-17 21:47:58 +02001124 current->pdeath_signal = 0;
Al Viro1b5d7832011-06-19 12:49:47 -04001125 } else {
1126 would_dump(bprm, bprm->file);
1127 if (bprm->interp_flags & BINPRM_FLAGS_ENFORCE_NONDUMP)
1128 set_dumpable(current->mm, suid_dumpable);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001129 }
1130
1131 /* An exec changes our domain. We are no longer part of the thread
1132 group */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001133 current->self_exec_id++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001134 flush_signal_handlers(current, 0);
Al Viro6a6d27d2012-08-21 09:56:33 -04001135 do_close_on_exec(current->files);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001136}
Linus Torvalds221af7f2010-01-28 22:14:42 -08001137EXPORT_SYMBOL(setup_new_exec);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001138
David Howellsa6f76f22008-11-14 10:39:24 +11001139/*
Oleg Nesterova2a84742009-09-05 11:17:13 -07001140 * Prepare credentials and lock ->cred_guard_mutex.
1141 * install_exec_creds() commits the new creds and drops the lock.
1142 * Or, if exec fails before, free_bprm() should release ->cred and
1143 * and unlock.
1144 */
1145int prepare_bprm_creds(struct linux_binprm *bprm)
1146{
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001147 if (mutex_lock_interruptible(&current->signal->cred_guard_mutex))
Oleg Nesterova2a84742009-09-05 11:17:13 -07001148 return -ERESTARTNOINTR;
1149
1150 bprm->cred = prepare_exec_creds();
1151 if (likely(bprm->cred))
1152 return 0;
1153
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001154 mutex_unlock(&current->signal->cred_guard_mutex);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001155 return -ENOMEM;
1156}
1157
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001158static void free_bprm(struct linux_binprm *bprm)
Oleg Nesterova2a84742009-09-05 11:17:13 -07001159{
1160 free_arg_pages(bprm);
1161 if (bprm->cred) {
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001162 mutex_unlock(&current->signal->cred_guard_mutex);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001163 abort_creds(bprm->cred);
1164 }
Oleg Nesterov63e46b92014-01-23 15:55:51 -08001165 if (bprm->file) {
1166 allow_write_access(bprm->file);
1167 fput(bprm->file);
1168 }
Kees Cookb66c5982012-12-20 15:05:16 -08001169 /* If a binfmt changed the interp, free it. */
1170 if (bprm->interp != bprm->filename)
1171 kfree(bprm->interp);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001172 kfree(bprm);
1173}
1174
Kees Cookb66c5982012-12-20 15:05:16 -08001175int bprm_change_interp(char *interp, struct linux_binprm *bprm)
1176{
1177 /* If a binfmt changed the interp, free it first. */
1178 if (bprm->interp != bprm->filename)
1179 kfree(bprm->interp);
1180 bprm->interp = kstrdup(interp, GFP_KERNEL);
1181 if (!bprm->interp)
1182 return -ENOMEM;
1183 return 0;
1184}
1185EXPORT_SYMBOL(bprm_change_interp);
1186
Oleg Nesterova2a84742009-09-05 11:17:13 -07001187/*
David Howellsa6f76f22008-11-14 10:39:24 +11001188 * install the new credentials for this executable
1189 */
1190void install_exec_creds(struct linux_binprm *bprm)
1191{
1192 security_bprm_committing_creds(bprm);
1193
1194 commit_creds(bprm->cred);
1195 bprm->cred = NULL;
Stephane Eranian2976b102013-06-20 11:36:28 +02001196
1197 /*
1198 * Disable monitoring for regular users
1199 * when executing setuid binaries. Must
1200 * wait until new credentials are committed
1201 * by commit_creds() above
1202 */
1203 if (get_dumpable(current->mm) != SUID_DUMP_USER)
1204 perf_event_exit_task(current);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001205 /*
1206 * cred_guard_mutex must be held at least to this point to prevent
David Howellsa6f76f22008-11-14 10:39:24 +11001207 * ptrace_attach() from altering our determination of the task's
Oleg Nesterova2a84742009-09-05 11:17:13 -07001208 * credentials; any time after this it may be unlocked.
1209 */
David Howellsa6f76f22008-11-14 10:39:24 +11001210 security_bprm_committed_creds(bprm);
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001211 mutex_unlock(&current->signal->cred_guard_mutex);
David Howellsa6f76f22008-11-14 10:39:24 +11001212}
1213EXPORT_SYMBOL(install_exec_creds);
1214
1215/*
1216 * determine how safe it is to execute the proposed program
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001217 * - the caller must hold ->cred_guard_mutex to protect against
David Howellsa6f76f22008-11-14 10:39:24 +11001218 * PTRACE_ATTACH
1219 */
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001220static void check_unsafe_exec(struct linux_binprm *bprm)
David Howellsa6f76f22008-11-14 10:39:24 +11001221{
David Howells0bf2f3a2009-02-06 11:45:46 +00001222 struct task_struct *p = current, *t;
Al Virof1191b52009-03-30 07:35:18 -04001223 unsigned n_fs;
David Howellsa6f76f22008-11-14 10:39:24 +11001224
Tejun Heo4b9d33e2011-06-17 16:50:38 +02001225 if (p->ptrace) {
1226 if (p->ptrace & PT_PTRACE_CAP)
1227 bprm->unsafe |= LSM_UNSAFE_PTRACE_CAP;
1228 else
1229 bprm->unsafe |= LSM_UNSAFE_PTRACE;
1230 }
David Howellsa6f76f22008-11-14 10:39:24 +11001231
Andy Lutomirski259e5e62012-04-12 16:47:50 -05001232 /*
1233 * This isn't strictly necessary, but it makes it harder for LSMs to
1234 * mess up.
1235 */
1236 if (current->no_new_privs)
1237 bprm->unsafe |= LSM_UNSAFE_NO_NEW_PRIVS;
1238
Oleg Nesterov83f62a22014-01-23 15:55:49 -08001239 t = p;
David Howells0bf2f3a2009-02-06 11:45:46 +00001240 n_fs = 1;
Nick Piggin2a4419b2010-08-18 04:37:33 +10001241 spin_lock(&p->fs->lock);
Oleg Nesterov437f7fd2009-04-24 01:02:45 +02001242 rcu_read_lock();
Oleg Nesterov83f62a22014-01-23 15:55:49 -08001243 while_each_thread(p, t) {
David Howells0bf2f3a2009-02-06 11:45:46 +00001244 if (t->fs == p->fs)
1245 n_fs++;
David Howells0bf2f3a2009-02-06 11:45:46 +00001246 }
Oleg Nesterov437f7fd2009-04-24 01:02:45 +02001247 rcu_read_unlock();
David Howells0bf2f3a2009-02-06 11:45:46 +00001248
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001249 if (p->fs->users > n_fs)
David Howellsa6f76f22008-11-14 10:39:24 +11001250 bprm->unsafe |= LSM_UNSAFE_SHARE;
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001251 else
1252 p->fs->in_exec = 1;
Nick Piggin2a4419b2010-08-18 04:37:33 +10001253 spin_unlock(&p->fs->lock);
David Howellsa6f76f22008-11-14 10:39:24 +11001254}
1255
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001256/*
1257 * Fill the binprm structure from the inode.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001258 * Check permissions, then read the first 128 (BINPRM_BUF_SIZE) bytes
David Howellsa6f76f22008-11-14 10:39:24 +11001259 *
1260 * This may be called multiple times for binary chains (scripts for example).
Linus Torvalds1da177e2005-04-16 15:20:36 -07001261 */
1262int prepare_binprm(struct linux_binprm *bprm)
1263{
Al Viro72c2d532013-09-22 16:27:52 -04001264 struct inode *inode = file_inode(bprm->file);
1265 umode_t mode = inode->i_mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001266 int retval;
1267
Linus Torvalds1da177e2005-04-16 15:20:36 -07001268
David Howellsa6f76f22008-11-14 10:39:24 +11001269 /* clear any previous set[ug]id data from a previous binary */
1270 bprm->cred->euid = current_euid();
1271 bprm->cred->egid = current_egid();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001272
Andy Lutomirski259e5e62012-04-12 16:47:50 -05001273 if (!(bprm->file->f_path.mnt->mnt_flags & MNT_NOSUID) &&
Eric W. Biederman3cdf5b42011-11-21 16:40:54 -08001274 !current->no_new_privs &&
1275 kuid_has_mapping(bprm->cred->user_ns, inode->i_uid) &&
1276 kgid_has_mapping(bprm->cred->user_ns, inode->i_gid)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001277 /* Set-uid? */
1278 if (mode & S_ISUID) {
David Howellsa6f76f22008-11-14 10:39:24 +11001279 bprm->per_clear |= PER_CLEAR_ON_SETID;
1280 bprm->cred->euid = inode->i_uid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001281 }
1282
1283 /* Set-gid? */
1284 /*
1285 * If setgid is set but no group execute bit then this
1286 * is a candidate for mandatory locking, not a setgid
1287 * executable.
1288 */
1289 if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) {
David Howellsa6f76f22008-11-14 10:39:24 +11001290 bprm->per_clear |= PER_CLEAR_ON_SETID;
1291 bprm->cred->egid = inode->i_gid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292 }
1293 }
1294
1295 /* fill in binprm security blob */
David Howellsa6f76f22008-11-14 10:39:24 +11001296 retval = security_bprm_set_creds(bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001297 if (retval)
1298 return retval;
David Howellsa6f76f22008-11-14 10:39:24 +11001299 bprm->cred_prepared = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001300
David Howellsa6f76f22008-11-14 10:39:24 +11001301 memset(bprm->buf, 0, BINPRM_BUF_SIZE);
1302 return kernel_read(bprm->file, 0, bprm->buf, BINPRM_BUF_SIZE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001303}
1304
1305EXPORT_SYMBOL(prepare_binprm);
1306
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001307/*
1308 * Arguments are '\0' separated strings found at the location bprm->p
1309 * points to; chop off the first by relocating brpm->p to right after
1310 * the first '\0' encountered.
1311 */
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001312int remove_arg_zero(struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001313{
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001314 int ret = 0;
1315 unsigned long offset;
1316 char *kaddr;
1317 struct page *page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001318
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001319 if (!bprm->argc)
1320 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001321
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001322 do {
1323 offset = bprm->p & ~PAGE_MASK;
1324 page = get_arg_page(bprm, bprm->p, 0);
1325 if (!page) {
1326 ret = -EFAULT;
1327 goto out;
1328 }
Cong Wange8e3c3d2011-11-25 23:14:27 +08001329 kaddr = kmap_atomic(page);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001330
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001331 for (; offset < PAGE_SIZE && kaddr[offset];
1332 offset++, bprm->p++)
1333 ;
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001334
Cong Wange8e3c3d2011-11-25 23:14:27 +08001335 kunmap_atomic(kaddr);
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001336 put_arg_page(page);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001337
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001338 if (offset == PAGE_SIZE)
1339 free_arg_page(bprm, (bprm->p >> PAGE_SHIFT) - 1);
1340 } while (offset == PAGE_SIZE);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001341
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001342 bprm->p++;
1343 bprm->argc--;
1344 ret = 0;
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001345
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001346out:
1347 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001348}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001349EXPORT_SYMBOL(remove_arg_zero);
1350
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001351#define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352/*
1353 * cycle the list of binary formats handler, until one recognizes the image
1354 */
Al Viro3c456bf2012-10-20 21:53:31 -04001355int search_binary_handler(struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001356{
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001357 bool need_retry = IS_ENABLED(CONFIG_MODULES);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001358 struct linux_binfmt *fmt;
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001359 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001360
Kees Cookd7402692012-12-17 16:03:20 -08001361 /* This allows 4 levels of binfmt rewrites before failing hard. */
Oleg Nesterov131b2f92013-09-11 14:24:39 -07001362 if (bprm->recursion_depth > 5)
Kees Cookd7402692012-12-17 16:03:20 -08001363 return -ELOOP;
1364
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365 retval = security_bprm_check(bprm);
1366 if (retval)
1367 return retval;
1368
Linus Torvalds1da177e2005-04-16 15:20:36 -07001369 retval = -ENOENT;
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001370 retry:
1371 read_lock(&binfmt_lock);
1372 list_for_each_entry(fmt, &formats, lh) {
1373 if (!try_module_get(fmt->module))
1374 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375 read_unlock(&binfmt_lock);
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001376 bprm->recursion_depth++;
1377 retval = fmt->load_binary(bprm);
1378 bprm->recursion_depth--;
Oleg Nesterov6b3c5382013-09-11 14:24:46 -07001379 if (retval >= 0 || retval != -ENOEXEC ||
1380 bprm->mm == NULL || bprm->file == NULL) {
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001381 put_binfmt(fmt);
1382 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383 }
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001384 read_lock(&binfmt_lock);
1385 put_binfmt(fmt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001386 }
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001387 read_unlock(&binfmt_lock);
1388
Oleg Nesterov6b3c5382013-09-11 14:24:46 -07001389 if (need_retry && retval == -ENOEXEC) {
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001390 if (printable(bprm->buf[0]) && printable(bprm->buf[1]) &&
1391 printable(bprm->buf[2]) && printable(bprm->buf[3]))
1392 return retval;
Oleg Nesterov4e0621a2013-09-11 14:24:45 -07001393 if (request_module("binfmt-%04x", *(ushort *)(bprm->buf + 2)) < 0)
1394 return retval;
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001395 need_retry = false;
1396 goto retry;
1397 }
1398
Linus Torvalds1da177e2005-04-16 15:20:36 -07001399 return retval;
1400}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001401EXPORT_SYMBOL(search_binary_handler);
1402
Oleg Nesterov5d1baf32013-09-11 14:24:38 -07001403static int exec_binprm(struct linux_binprm *bprm)
1404{
1405 pid_t old_pid, old_vpid;
1406 int ret;
1407
1408 /* Need to fetch pid before load_binary changes it */
1409 old_pid = current->pid;
1410 rcu_read_lock();
1411 old_vpid = task_pid_nr_ns(current, task_active_pid_ns(current->parent));
1412 rcu_read_unlock();
1413
1414 ret = search_binary_handler(bprm);
1415 if (ret >= 0) {
Linus Torvalds3eaded862013-11-21 19:18:14 -08001416 audit_bprm(bprm);
Oleg Nesterov5d1baf32013-09-11 14:24:38 -07001417 trace_sched_process_exec(current, old_pid, bprm);
1418 ptrace_event(PTRACE_EVENT_EXEC, old_vpid);
Oleg Nesterov9beb2662013-09-11 14:24:40 -07001419 proc_exec_connector(current);
Oleg Nesterov5d1baf32013-09-11 14:24:38 -07001420 }
1421
1422 return ret;
1423}
1424
Linus Torvalds1da177e2005-04-16 15:20:36 -07001425/*
1426 * sys_execve() executes a new program.
1427 */
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001428static int do_execve_common(struct filename *filename,
Oleg Nesterovba2d01622011-03-06 18:02:37 +01001429 struct user_arg_ptr argv,
Al Viro835ab322012-10-20 21:50:59 -04001430 struct user_arg_ptr envp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001431{
1432 struct linux_binprm *bprm;
1433 struct file *file;
Al Viro3b125382008-04-22 05:31:30 -04001434 struct files_struct *displaced;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001435 int retval;
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +04001436
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001437 if (IS_ERR(filename))
1438 return PTR_ERR(filename);
1439
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +04001440 /*
1441 * We move the actual failure in case of RLIMIT_NPROC excess from
1442 * set*uid() to execve() because too many poorly written programs
1443 * don't check setuid() return code. Here we additionally recheck
1444 * whether NPROC limit is still exceeded.
1445 */
1446 if ((current->flags & PF_NPROC_EXCEEDED) &&
Oleg Nesterovbd9d43f2013-07-03 15:08:34 -07001447 atomic_read(&current_user()->processes) > rlimit(RLIMIT_NPROC)) {
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +04001448 retval = -EAGAIN;
1449 goto out_ret;
1450 }
1451
1452 /* We're below the limit (still or again), so we don't want to make
1453 * further execve() calls fail. */
1454 current->flags &= ~PF_NPROC_EXCEEDED;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001455
Al Viro3b125382008-04-22 05:31:30 -04001456 retval = unshare_files(&displaced);
Al Virofd8328b2008-04-22 05:11:59 -04001457 if (retval)
1458 goto out_ret;
1459
Linus Torvalds1da177e2005-04-16 15:20:36 -07001460 retval = -ENOMEM;
Oliver Neukum11b0b5a2006-03-25 03:08:13 -08001461 bprm = kzalloc(sizeof(*bprm), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001462 if (!bprm)
Al Virofd8328b2008-04-22 05:11:59 -04001463 goto out_files;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001464
Oleg Nesterova2a84742009-09-05 11:17:13 -07001465 retval = prepare_bprm_creds(bprm);
1466 if (retval)
David Howellsa6f76f22008-11-14 10:39:24 +11001467 goto out_free;
Al Viro498052b2009-03-30 07:20:30 -04001468
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001469 check_unsafe_exec(bprm);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001470 current->in_execve = 1;
David Howellsa6f76f22008-11-14 10:39:24 +11001471
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001472 file = do_open_exec(filename);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001473 retval = PTR_ERR(file);
1474 if (IS_ERR(file))
Al Viro498052b2009-03-30 07:20:30 -04001475 goto out_unmark;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001476
1477 sched_exec();
1478
Linus Torvalds1da177e2005-04-16 15:20:36 -07001479 bprm->file = file;
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001480 bprm->filename = bprm->interp = filename->name;
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001481
1482 retval = bprm_mm_init(bprm);
1483 if (retval)
Oleg Nesterov63e46b92014-01-23 15:55:51 -08001484 goto out_unmark;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001485
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001486 bprm->argc = count(argv, MAX_ARG_STRINGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001487 if ((retval = bprm->argc) < 0)
David Howellsa6f76f22008-11-14 10:39:24 +11001488 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001489
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001490 bprm->envc = count(envp, MAX_ARG_STRINGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001491 if ((retval = bprm->envc) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001492 goto out;
1493
1494 retval = prepare_binprm(bprm);
1495 if (retval < 0)
1496 goto out;
1497
1498 retval = copy_strings_kernel(1, &bprm->filename, bprm);
1499 if (retval < 0)
1500 goto out;
1501
1502 bprm->exec = bprm->p;
1503 retval = copy_strings(bprm->envc, envp, bprm);
1504 if (retval < 0)
1505 goto out;
1506
1507 retval = copy_strings(bprm->argc, argv, bprm);
1508 if (retval < 0)
1509 goto out;
1510
Oleg Nesterov5d1baf32013-09-11 14:24:38 -07001511 retval = exec_binprm(bprm);
David Howellsa6f76f22008-11-14 10:39:24 +11001512 if (retval < 0)
1513 goto out;
1514
1515 /* execve succeeded */
Al Viro498052b2009-03-30 07:20:30 -04001516 current->fs->in_exec = 0;
Kentaro Takedaf9ce1f12009-02-05 17:18:11 +09001517 current->in_execve = 0;
David Howellsa6f76f22008-11-14 10:39:24 +11001518 acct_update_integrals(current);
Rik van Riel82727012013-10-07 11:29:28 +01001519 task_numa_free(current);
David Howellsa6f76f22008-11-14 10:39:24 +11001520 free_bprm(bprm);
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001521 putname(filename);
David Howellsa6f76f22008-11-14 10:39:24 +11001522 if (displaced)
1523 put_files_struct(displaced);
1524 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001525
1526out:
Oleg Nesterov3c77f842010-11-30 20:55:34 +01001527 if (bprm->mm) {
1528 acct_arg_size(bprm, 0);
1529 mmput(bprm->mm);
1530 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001531
Al Viro498052b2009-03-30 07:20:30 -04001532out_unmark:
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001533 current->fs->in_exec = 0;
Kentaro Takedaf9ce1f12009-02-05 17:18:11 +09001534 current->in_execve = 0;
David Howellsa6f76f22008-11-14 10:39:24 +11001535
1536out_free:
Al Viro08a6fac2008-05-10 16:38:25 -04001537 free_bprm(bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001538
Al Virofd8328b2008-04-22 05:11:59 -04001539out_files:
Al Viro3b125382008-04-22 05:31:30 -04001540 if (displaced)
1541 reset_files_struct(displaced);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001542out_ret:
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001543 putname(filename);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001544 return retval;
1545}
1546
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001547int do_execve(struct filename *filename,
Oleg Nesterovba2d01622011-03-06 18:02:37 +01001548 const char __user *const __user *__argv,
Al Viroda3d4c52012-10-20 21:49:33 -04001549 const char __user *const __user *__envp)
Oleg Nesterovba2d01622011-03-06 18:02:37 +01001550{
Oleg Nesterov0e028462011-03-06 18:02:54 +01001551 struct user_arg_ptr argv = { .ptr.native = __argv };
1552 struct user_arg_ptr envp = { .ptr.native = __envp };
Al Viro835ab322012-10-20 21:50:59 -04001553 return do_execve_common(filename, argv, envp);
Oleg Nesterovba2d01622011-03-06 18:02:37 +01001554}
1555
Oleg Nesterov0e028462011-03-06 18:02:54 +01001556#ifdef CONFIG_COMPAT
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001557static int compat_do_execve(struct filename *filename,
Al Viro38b983b2012-09-30 13:38:55 -04001558 const compat_uptr_t __user *__argv,
Al Virod03d26e2012-10-20 21:46:25 -04001559 const compat_uptr_t __user *__envp)
Oleg Nesterov0e028462011-03-06 18:02:54 +01001560{
1561 struct user_arg_ptr argv = {
1562 .is_compat = true,
1563 .ptr.compat = __argv,
1564 };
1565 struct user_arg_ptr envp = {
1566 .is_compat = true,
1567 .ptr.compat = __envp,
1568 };
Al Viro835ab322012-10-20 21:50:59 -04001569 return do_execve_common(filename, argv, envp);
Oleg Nesterov0e028462011-03-06 18:02:54 +01001570}
1571#endif
1572
Oleg Nesterov964ee7d2009-09-23 15:56:59 -07001573void set_binfmt(struct linux_binfmt *new)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001574{
Hiroshi Shimamoto801460d2009-09-23 15:57:41 -07001575 struct mm_struct *mm = current->mm;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001576
Hiroshi Shimamoto801460d2009-09-23 15:57:41 -07001577 if (mm->binfmt)
1578 module_put(mm->binfmt->module);
1579
1580 mm->binfmt = new;
Oleg Nesterov964ee7d2009-09-23 15:56:59 -07001581 if (new)
1582 __module_get(new->module);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001583}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001584EXPORT_SYMBOL(set_binfmt);
1585
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001586/*
Oleg Nesterov7288e112014-01-23 15:55:32 -08001587 * set_dumpable stores three-value SUID_DUMP_* into mm->flags.
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001588 */
1589void set_dumpable(struct mm_struct *mm, int value)
1590{
Oleg Nesterovabacd2f2014-01-23 15:55:31 -08001591 unsigned long old, new;
1592
Oleg Nesterov7288e112014-01-23 15:55:32 -08001593 if (WARN_ON((unsigned)value > SUID_DUMP_ROOT))
1594 return;
1595
Oleg Nesterovabacd2f2014-01-23 15:55:31 -08001596 do {
1597 old = ACCESS_ONCE(mm->flags);
Oleg Nesterov7288e112014-01-23 15:55:32 -08001598 new = (old & ~MMF_DUMPABLE_MASK) | value;
Oleg Nesterovabacd2f2014-01-23 15:55:31 -08001599 } while (cmpxchg(&mm->flags, old, new) != old);
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001600}
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001601
Al Viro38b983b2012-09-30 13:38:55 -04001602SYSCALL_DEFINE3(execve,
1603 const char __user *, filename,
1604 const char __user *const __user *, argv,
1605 const char __user *const __user *, envp)
1606{
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001607 return do_execve(getname(filename), argv, envp);
Al Viro38b983b2012-09-30 13:38:55 -04001608}
1609#ifdef CONFIG_COMPAT
Heiko Carstens625b1d72014-03-04 10:53:50 +01001610COMPAT_SYSCALL_DEFINE3(execve, const char __user *, filename,
1611 const compat_uptr_t __user *, argv,
1612 const compat_uptr_t __user *, envp)
Al Viro38b983b2012-09-30 13:38:55 -04001613{
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001614 return compat_do_execve(getname(filename), argv, envp);
Al Viro38b983b2012-09-30 13:38:55 -04001615}
1616#endif