blob: 828ec5f07de00b7a21d016c022afc68ee0962331 [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
Eric W. Biederman90f85722015-06-29 14:42:03 -0500101bool path_noexec(const struct path *path)
102{
103 return (path->mnt->mnt_flags & MNT_NOEXEC) ||
104 (path->mnt->mnt_sb->s_iflags & SB_I_NOEXEC);
105}
106
Josh Triplett69369a72014-04-03 14:48:27 -0700107#ifdef CONFIG_USELIB
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108/*
109 * Note that a shared library must be both readable and executable due to
110 * security reasons.
111 *
112 * Also note that we take the address to load from from the file itself.
113 */
Heiko Carstens1e7bfb22009-01-14 14:14:29 +0100114SYSCALL_DEFINE1(uselib, const char __user *, library)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115{
Al Viro72c2d532013-09-22 16:27:52 -0400116 struct linux_binfmt *fmt;
Al Viro964bd182008-07-26 03:33:14 -0400117 struct file *file;
Jeff Layton91a27b22012-10-10 15:25:28 -0400118 struct filename *tmp = getname(library);
Al Viro964bd182008-07-26 03:33:14 -0400119 int error = PTR_ERR(tmp);
Al Viro47c805d2011-02-23 17:44:09 -0500120 static const struct open_flags uselib_flags = {
121 .open_flag = O_LARGEFILE | O_RDONLY | __FMODE_EXEC,
Al Viro62fb4a12015-12-26 22:33:24 -0500122 .acc_mode = MAY_READ | MAY_EXEC,
Al Virof9652e12013-06-11 08:23:01 +0400123 .intent = LOOKUP_OPEN,
124 .lookup_flags = LOOKUP_FOLLOW,
Al Viro47c805d2011-02-23 17:44:09 -0500125 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700126
Al Viro6e8341a2009-04-06 11:16:22 -0400127 if (IS_ERR(tmp))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700128 goto out;
129
Al Virof9652e12013-06-11 08:23:01 +0400130 file = do_filp_open(AT_FDCWD, tmp, &uselib_flags);
Al Viro6e8341a2009-04-06 11:16:22 -0400131 putname(tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700132 error = PTR_ERR(file);
133 if (IS_ERR(file))
134 goto out;
135
Al Viro6e8341a2009-04-06 11:16:22 -0400136 error = -EINVAL;
Al Viro496ad9a2013-01-23 17:07:38 -0500137 if (!S_ISREG(file_inode(file)->i_mode))
Al Viro6e8341a2009-04-06 11:16:22 -0400138 goto exit;
139
140 error = -EACCES;
Eric W. Biederman90f85722015-06-29 14:42:03 -0500141 if (path_noexec(&file->f_path))
Al Viro6e8341a2009-04-06 11:16:22 -0400142 goto exit;
143
Eric Paris2a12a9d2009-12-17 21:24:21 -0500144 fsnotify_open(file);
Eric Paris6110e3a2008-12-17 13:53:20 -0500145
Linus Torvalds1da177e2005-04-16 15:20:36 -0700146 error = -ENOEXEC;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700147
Al Viro72c2d532013-09-22 16:27:52 -0400148 read_lock(&binfmt_lock);
149 list_for_each_entry(fmt, &formats, lh) {
150 if (!fmt->load_shlib)
151 continue;
152 if (!try_module_get(fmt->module))
153 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700154 read_unlock(&binfmt_lock);
Al Viro72c2d532013-09-22 16:27:52 -0400155 error = fmt->load_shlib(file);
156 read_lock(&binfmt_lock);
157 put_binfmt(fmt);
158 if (error != -ENOEXEC)
159 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700160 }
Al Viro72c2d532013-09-22 16:27:52 -0400161 read_unlock(&binfmt_lock);
Al Viro6e8341a2009-04-06 11:16:22 -0400162exit:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700163 fput(file);
164out:
165 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166}
Josh Triplett69369a72014-04-03 14:48:27 -0700167#endif /* #ifdef CONFIG_USELIB */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700168
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700169#ifdef CONFIG_MMU
Oleg Nesterovae6b5852011-03-06 18:03:11 +0100170/*
171 * The nascent bprm->mm is not visible until exec_mmap() but it can
172 * use a lot of memory, account these pages in current->mm temporary
173 * for oom_badness()->get_mm_rss(). Once exec succeeds or fails, we
174 * change the counter back via acct_arg_size(0).
175 */
Oleg Nesterov0e028462011-03-06 18:02:54 +0100176static void acct_arg_size(struct linux_binprm *bprm, unsigned long pages)
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100177{
178 struct mm_struct *mm = current->mm;
179 long diff = (long)(pages - bprm->vma_pages);
180
181 if (!mm || !diff)
182 return;
183
184 bprm->vma_pages = pages;
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100185 add_mm_counter(mm, MM_ANONPAGES, diff);
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100186}
187
Oleg Nesterov0e028462011-03-06 18:02:54 +0100188static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos,
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700189 int write)
190{
191 struct page *page;
192 int ret;
193
194#ifdef CONFIG_STACK_GROWSUP
195 if (write) {
Michal Hockod05f3162011-05-24 17:11:44 -0700196 ret = expand_downwards(bprm->vma, pos);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700197 if (ret < 0)
198 return NULL;
199 }
200#endif
201 ret = get_user_pages(current, bprm->mm, pos,
202 1, write, 1, &page, NULL);
203 if (ret <= 0)
204 return NULL;
205
206 if (write) {
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700207 unsigned long size = bprm->vma->vm_end - bprm->vma->vm_start;
Linus Torvaldsa64e7152008-03-03 10:12:14 -0800208 struct rlimit *rlim;
209
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100210 acct_arg_size(bprm, size / PAGE_SIZE);
211
Linus Torvaldsa64e7152008-03-03 10:12:14 -0800212 /*
213 * We've historically supported up to 32 pages (ARG_MAX)
214 * of argument strings even with small stacks
215 */
216 if (size <= ARG_MAX)
217 return page;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700218
219 /*
220 * Limit to 1/4-th the stack size for the argv+env strings.
221 * This ensures that:
222 * - the remaining binfmt code will not run out of stack space,
223 * - the program will have a reasonable amount of stack left
224 * to work from.
225 */
Linus Torvaldsa64e7152008-03-03 10:12:14 -0800226 rlim = current->signal->rlim;
Jiri Slabyd554ed892010-03-05 13:42:42 -0800227 if (size > ACCESS_ONCE(rlim[RLIMIT_STACK].rlim_cur) / 4) {
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700228 put_page(page);
229 return NULL;
230 }
231 }
232
233 return page;
234}
235
236static void put_arg_page(struct page *page)
237{
238 put_page(page);
239}
240
241static void free_arg_page(struct linux_binprm *bprm, int i)
242{
243}
244
245static void free_arg_pages(struct linux_binprm *bprm)
246{
247}
248
249static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos,
250 struct page *page)
251{
252 flush_cache_page(bprm->vma, pos, page_to_pfn(page));
253}
254
255static int __bprm_mm_init(struct linux_binprm *bprm)
256{
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800257 int err;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700258 struct vm_area_struct *vma = NULL;
259 struct mm_struct *mm = bprm->mm;
260
261 bprm->vma = vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL);
262 if (!vma)
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800263 return -ENOMEM;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700264
265 down_write(&mm->mmap_sem);
266 vma->vm_mm = mm;
267
268 /*
269 * Place the stack at the largest stack address the architecture
270 * supports. Later, we'll move this to an appropriate place. We don't
271 * use STACK_TOP because that can depend on attributes which aren't
272 * configured yet.
273 */
Michal Hockoaacb3d12011-07-26 16:08:40 -0700274 BUILD_BUG_ON(VM_STACK_FLAGS & VM_STACK_INCOMPLETE_SETUP);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700275 vma->vm_end = STACK_TOP_MAX;
276 vma->vm_start = vma->vm_end - PAGE_SIZE;
Cyrill Gorcunovd9104d12013-09-11 14:22:24 -0700277 vma->vm_flags = VM_SOFTDIRTY | VM_STACK_FLAGS | VM_STACK_INCOMPLETE_SETUP;
Coly Li3ed75eb2007-10-18 23:39:15 -0700278 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
Rik van Riel5beb4932010-03-05 13:42:07 -0800279 INIT_LIST_HEAD(&vma->anon_vma_chain);
Tavis Ormandy462e635e2010-12-09 15:29:42 +0100280
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700281 err = insert_vm_struct(mm, vma);
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800282 if (err)
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700283 goto err;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700284
285 mm->stack_vm = mm->total_vm = 1;
Dave Hansenfe3d1972014-11-14 07:18:29 -0800286 arch_bprm_mm_init(mm, vma);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700287 up_write(&mm->mmap_sem);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700288 bprm->p = vma->vm_end - sizeof(void *);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700289 return 0;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700290err:
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800291 up_write(&mm->mmap_sem);
292 bprm->vma = NULL;
293 kmem_cache_free(vm_area_cachep, vma);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700294 return err;
295}
296
297static bool valid_arg_len(struct linux_binprm *bprm, long len)
298{
299 return len <= MAX_ARG_STRLEN;
300}
301
302#else
303
Oleg Nesterov0e028462011-03-06 18:02:54 +0100304static inline void acct_arg_size(struct linux_binprm *bprm, unsigned long pages)
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100305{
306}
307
Oleg Nesterov0e028462011-03-06 18:02:54 +0100308static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos,
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700309 int write)
310{
311 struct page *page;
312
313 page = bprm->page[pos / PAGE_SIZE];
314 if (!page && write) {
315 page = alloc_page(GFP_HIGHUSER|__GFP_ZERO);
316 if (!page)
317 return NULL;
318 bprm->page[pos / PAGE_SIZE] = page;
319 }
320
321 return page;
322}
323
324static void put_arg_page(struct page *page)
325{
326}
327
328static void free_arg_page(struct linux_binprm *bprm, int i)
329{
330 if (bprm->page[i]) {
331 __free_page(bprm->page[i]);
332 bprm->page[i] = NULL;
333 }
334}
335
336static void free_arg_pages(struct linux_binprm *bprm)
337{
338 int i;
339
340 for (i = 0; i < MAX_ARG_PAGES; i++)
341 free_arg_page(bprm, i);
342}
343
344static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos,
345 struct page *page)
346{
347}
348
349static int __bprm_mm_init(struct linux_binprm *bprm)
350{
351 bprm->p = PAGE_SIZE * MAX_ARG_PAGES - sizeof(void *);
352 return 0;
353}
354
355static bool valid_arg_len(struct linux_binprm *bprm, long len)
356{
357 return len <= bprm->p;
358}
359
360#endif /* CONFIG_MMU */
361
362/*
363 * Create a new mm_struct and populate it with a temporary stack
364 * vm_area_struct. We don't have enough context at this point to set the stack
365 * flags, permissions, and offset, so we use temporary values. We'll update
366 * them later in setup_arg_pages().
367 */
Yuanhan Liu9cc64ce2013-02-20 13:16:01 +1100368static int bprm_mm_init(struct linux_binprm *bprm)
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700369{
370 int err;
371 struct mm_struct *mm = NULL;
372
373 bprm->mm = mm = mm_alloc();
374 err = -ENOMEM;
375 if (!mm)
376 goto err;
377
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700378 err = __bprm_mm_init(bprm);
379 if (err)
380 goto err;
381
382 return 0;
383
384err:
385 if (mm) {
386 bprm->mm = NULL;
387 mmdrop(mm);
388 }
389
390 return err;
391}
392
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100393struct user_arg_ptr {
Oleg Nesterov0e028462011-03-06 18:02:54 +0100394#ifdef CONFIG_COMPAT
395 bool is_compat;
396#endif
397 union {
398 const char __user *const __user *native;
399#ifdef CONFIG_COMPAT
Al Viro38b983b2012-09-30 13:38:55 -0400400 const compat_uptr_t __user *compat;
Oleg Nesterov0e028462011-03-06 18:02:54 +0100401#endif
402 } ptr;
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100403};
404
405static const char __user *get_user_arg_ptr(struct user_arg_ptr argv, int nr)
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100406{
Oleg Nesterov0e028462011-03-06 18:02:54 +0100407 const char __user *native;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100408
Oleg Nesterov0e028462011-03-06 18:02:54 +0100409#ifdef CONFIG_COMPAT
410 if (unlikely(argv.is_compat)) {
411 compat_uptr_t compat;
412
413 if (get_user(compat, argv.ptr.compat + nr))
414 return ERR_PTR(-EFAULT);
415
416 return compat_ptr(compat);
417 }
418#endif
419
420 if (get_user(native, argv.ptr.native + nr))
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100421 return ERR_PTR(-EFAULT);
422
Oleg Nesterov0e028462011-03-06 18:02:54 +0100423 return native;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100424}
425
Linus Torvalds1da177e2005-04-16 15:20:36 -0700426/*
427 * count() counts the number of strings in array ARGV.
428 */
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100429static int count(struct user_arg_ptr argv, int max)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430{
431 int i = 0;
432
Oleg Nesterov0e028462011-03-06 18:02:54 +0100433 if (argv.ptr.native != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434 for (;;) {
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100435 const char __user *p = get_user_arg_ptr(argv, i);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700436
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437 if (!p)
438 break;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100439
440 if (IS_ERR(p))
441 return -EFAULT;
442
Xi Wang6d92d4f2013-01-11 14:31:48 -0800443 if (i >= max)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444 return -E2BIG;
Xi Wang6d92d4f2013-01-11 14:31:48 -0800445 ++i;
Roland McGrath9aea5a62010-09-07 19:37:06 -0700446
447 if (fatal_signal_pending(current))
448 return -ERESTARTNOHAND;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700449 cond_resched();
450 }
451 }
452 return i;
453}
454
455/*
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700456 * 'copy_strings()' copies argument/environment strings from the old
457 * processes's memory to the new process's stack. The call to get_user_pages()
458 * ensures the destination page is created and not swapped out.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700459 */
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100460static int copy_strings(int argc, struct user_arg_ptr argv,
Adrian Bunk75c96f82005-05-05 16:16:09 -0700461 struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700462{
463 struct page *kmapped_page = NULL;
464 char *kaddr = NULL;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700465 unsigned long kpos = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466 int ret;
467
468 while (argc-- > 0) {
David Howellsd7627462010-08-17 23:52:56 +0100469 const char __user *str;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470 int len;
471 unsigned long pos;
472
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100473 ret = -EFAULT;
474 str = get_user_arg_ptr(argv, argc);
475 if (IS_ERR(str))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100478 len = strnlen_user(str, MAX_ARG_STRLEN);
479 if (!len)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480 goto out;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100481
482 ret = -E2BIG;
483 if (!valid_arg_len(bprm, len))
484 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700486 /* We're going to work our way backwords. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487 pos = bprm->p;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700488 str += len;
489 bprm->p -= len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490
491 while (len > 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492 int offset, bytes_to_copy;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493
Roland McGrath9aea5a62010-09-07 19:37:06 -0700494 if (fatal_signal_pending(current)) {
495 ret = -ERESTARTNOHAND;
496 goto out;
497 }
Roland McGrath7993bc12010-09-07 19:36:28 -0700498 cond_resched();
499
Linus Torvalds1da177e2005-04-16 15:20:36 -0700500 offset = pos % PAGE_SIZE;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700501 if (offset == 0)
502 offset = PAGE_SIZE;
503
504 bytes_to_copy = offset;
505 if (bytes_to_copy > len)
506 bytes_to_copy = len;
507
508 offset -= bytes_to_copy;
509 pos -= bytes_to_copy;
510 str -= bytes_to_copy;
511 len -= bytes_to_copy;
512
513 if (!kmapped_page || kpos != (pos & PAGE_MASK)) {
514 struct page *page;
515
516 page = get_arg_page(bprm, pos, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517 if (!page) {
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700518 ret = -E2BIG;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519 goto out;
520 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700522 if (kmapped_page) {
523 flush_kernel_dcache_page(kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700524 kunmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700525 put_arg_page(kmapped_page);
526 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527 kmapped_page = page;
528 kaddr = kmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700529 kpos = pos & PAGE_MASK;
530 flush_arg_page(bprm, kpos, kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531 }
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700532 if (copy_from_user(kaddr+offset, str, bytes_to_copy)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533 ret = -EFAULT;
534 goto out;
535 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536 }
537 }
538 ret = 0;
539out:
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700540 if (kmapped_page) {
541 flush_kernel_dcache_page(kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700542 kunmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700543 put_arg_page(kmapped_page);
544 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545 return ret;
546}
547
548/*
549 * Like copy_strings, but get argv and its values from kernel memory.
550 */
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100551int copy_strings_kernel(int argc, const char *const *__argv,
David Howellsd7627462010-08-17 23:52:56 +0100552 struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553{
554 int r;
555 mm_segment_t oldfs = get_fs();
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100556 struct user_arg_ptr argv = {
Oleg Nesterov0e028462011-03-06 18:02:54 +0100557 .ptr.native = (const char __user *const __user *)__argv,
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100558 };
559
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560 set_fs(KERNEL_DS);
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100561 r = copy_strings(argc, argv, bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700562 set_fs(oldfs);
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100563
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564 return r;
565}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566EXPORT_SYMBOL(copy_strings_kernel);
567
568#ifdef CONFIG_MMU
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700569
Linus Torvalds1da177e2005-04-16 15:20:36 -0700570/*
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700571 * During bprm_mm_init(), we create a temporary stack at STACK_TOP_MAX. Once
572 * the binfmt code determines where the new stack should reside, we shift it to
573 * its final location. The process proceeds as follows:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700574 *
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700575 * 1) Use shift to calculate the new vma endpoints.
576 * 2) Extend vma to cover both the old and new ranges. This ensures the
577 * arguments passed to subsequent functions are consistent.
578 * 3) Move vma's page tables to the new range.
579 * 4) Free up any cleared pgd range.
580 * 5) Shrink the vma to cover only the new range.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700581 */
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700582static int shift_arg_pages(struct vm_area_struct *vma, unsigned long shift)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583{
584 struct mm_struct *mm = vma->vm_mm;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700585 unsigned long old_start = vma->vm_start;
586 unsigned long old_end = vma->vm_end;
587 unsigned long length = old_end - old_start;
588 unsigned long new_start = old_start - shift;
589 unsigned long new_end = old_end - shift;
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700590 struct mmu_gather tlb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700592 BUG_ON(new_start > new_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700593
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700594 /*
595 * ensure there are no vmas between where we want to go
596 * and where we are
597 */
598 if (vma != find_vma(mm, new_start))
599 return -EFAULT;
600
601 /*
602 * cover the whole range: [new_start, old_end)
603 */
Rik van Riel5beb4932010-03-05 13:42:07 -0800604 if (vma_adjust(vma, new_start, old_end, vma->vm_pgoff, NULL))
605 return -ENOMEM;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700606
607 /*
608 * move the page tables downwards, on failure we rely on
609 * process cleanup to remove whatever mess we made.
610 */
611 if (length != move_page_tables(vma, old_start,
Michel Lespinasse38a76012012-10-08 16:31:50 -0700612 vma, new_start, length, false))
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700613 return -ENOMEM;
614
615 lru_add_drain();
Linus Torvalds2b047252013-08-15 11:42:25 -0700616 tlb_gather_mmu(&tlb, mm, old_start, old_end);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700617 if (new_end > old_start) {
618 /*
619 * when the old and new regions overlap clear from new_end.
620 */
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700621 free_pgd_range(&tlb, new_end, old_end, new_end,
Hugh Dickins6ee86302013-04-29 15:07:44 -0700622 vma->vm_next ? vma->vm_next->vm_start : USER_PGTABLES_CEILING);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700623 } else {
624 /*
625 * otherwise, clean from old_start; this is done to not touch
626 * the address space in [new_end, old_start) some architectures
627 * have constraints on va-space that make this illegal (IA64) -
628 * for the others its just a little faster.
629 */
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700630 free_pgd_range(&tlb, old_start, old_end, new_end,
Hugh Dickins6ee86302013-04-29 15:07:44 -0700631 vma->vm_next ? vma->vm_next->vm_start : USER_PGTABLES_CEILING);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700632 }
Linus Torvalds2b047252013-08-15 11:42:25 -0700633 tlb_finish_mmu(&tlb, old_start, old_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700635 /*
Rik van Riel5beb4932010-03-05 13:42:07 -0800636 * Shrink the vma to just the new range. Always succeeds.
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700637 */
638 vma_adjust(vma, new_start, new_end, vma->vm_pgoff, NULL);
639
640 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700641}
642
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700643/*
644 * Finalizes the stack vm_area_struct. The flags and permissions are updated,
645 * the stack is optionally relocated, and some extra space is added.
646 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700647int setup_arg_pages(struct linux_binprm *bprm,
648 unsigned long stack_top,
649 int executable_stack)
650{
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700651 unsigned long ret;
652 unsigned long stack_shift;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653 struct mm_struct *mm = current->mm;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700654 struct vm_area_struct *vma = bprm->vma;
655 struct vm_area_struct *prev = NULL;
656 unsigned long vm_flags;
657 unsigned long stack_base;
Michael Neuling803bf5e2010-02-10 13:56:42 -0800658 unsigned long stack_size;
659 unsigned long stack_expand;
660 unsigned long rlim_stack;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661
662#ifdef CONFIG_STACK_GROWSUP
James Hogand71f2902014-05-13 23:58:24 +0100663 /* Limit stack size */
Jiri Slabyd554ed892010-03-05 13:42:42 -0800664 stack_base = rlimit_max(RLIMIT_STACK);
James Hogand71f2902014-05-13 23:58:24 +0100665 if (stack_base > STACK_SIZE_MAX)
666 stack_base = STACK_SIZE_MAX;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700667
Helge Dellerd045c772015-05-11 22:01:27 +0200668 /* Add space for stack randomization. */
669 stack_base += (STACK_RND_MASK << PAGE_SHIFT);
670
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700671 /* Make sure we didn't let the argument array grow too large. */
672 if (vma->vm_end - vma->vm_start > stack_base)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700673 return -ENOMEM;
674
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700675 stack_base = PAGE_ALIGN(stack_top - stack_base);
676
677 stack_shift = vma->vm_start - stack_base;
678 mm->arg_start = bprm->p - stack_shift;
679 bprm->p = vma->vm_end - stack_shift;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680#else
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700681 stack_top = arch_align_stack(stack_top);
682 stack_top = PAGE_ALIGN(stack_top);
Roland McGrath1b528182010-09-07 19:35:49 -0700683
684 if (unlikely(stack_top < mmap_min_addr) ||
685 unlikely(vma->vm_end - vma->vm_start >= stack_top - mmap_min_addr))
686 return -ENOMEM;
687
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700688 stack_shift = vma->vm_end - stack_top;
689
690 bprm->p -= stack_shift;
691 mm->arg_start = bprm->p;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692#endif
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700693
694 if (bprm->loader)
695 bprm->loader -= stack_shift;
696 bprm->exec -= stack_shift;
697
698 down_write(&mm->mmap_sem);
Hugh Dickins96a8e132008-07-10 21:19:20 +0100699 vm_flags = VM_STACK_FLAGS;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700700
701 /*
702 * Adjust stack execute permissions; explicitly enable for
703 * EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X and leave alone
704 * (arch default) otherwise.
705 */
706 if (unlikely(executable_stack == EXSTACK_ENABLE_X))
707 vm_flags |= VM_EXEC;
708 else if (executable_stack == EXSTACK_DISABLE_X)
709 vm_flags &= ~VM_EXEC;
710 vm_flags |= mm->def_flags;
Mel Gormana8bef8ff2010-05-24 14:32:24 -0700711 vm_flags |= VM_STACK_INCOMPLETE_SETUP;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700712
713 ret = mprotect_fixup(vma, &prev, vma->vm_start, vma->vm_end,
714 vm_flags);
715 if (ret)
716 goto out_unlock;
717 BUG_ON(prev != vma);
718
719 /* Move stack pages down in memory. */
720 if (stack_shift) {
721 ret = shift_arg_pages(vma, stack_shift);
Anton Blanchardfc63cf22009-11-11 14:26:48 -0800722 if (ret)
723 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724 }
725
Mel Gormana8bef8ff2010-05-24 14:32:24 -0700726 /* mprotect_fixup is overkill to remove the temporary stack flags */
727 vma->vm_flags &= ~VM_STACK_INCOMPLETE_SETUP;
728
Michael Neuling5ef097d2010-03-05 13:42:57 -0800729 stack_expand = 131072UL; /* randomly 32*4k (or 2*64k) pages */
Michael Neuling803bf5e2010-02-10 13:56:42 -0800730 stack_size = vma->vm_end - vma->vm_start;
731 /*
732 * Align this down to a page boundary as expand_stack
733 * will align it up.
734 */
735 rlim_stack = rlimit(RLIMIT_STACK) & PAGE_MASK;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700736#ifdef CONFIG_STACK_GROWSUP
Michael Neuling803bf5e2010-02-10 13:56:42 -0800737 if (stack_size + stack_expand > rlim_stack)
738 stack_base = vma->vm_start + rlim_stack;
739 else
740 stack_base = vma->vm_end + stack_expand;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700741#else
Michael Neuling803bf5e2010-02-10 13:56:42 -0800742 if (stack_size + stack_expand > rlim_stack)
743 stack_base = vma->vm_end - rlim_stack;
744 else
745 stack_base = vma->vm_start - stack_expand;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700746#endif
Eric B Munson3af9e852010-05-18 15:30:49 +0100747 current->mm->start_stack = bprm->p;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700748 ret = expand_stack(vma, stack_base);
749 if (ret)
750 ret = -EFAULT;
751
752out_unlock:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753 up_write(&mm->mmap_sem);
Anton Blanchardfc63cf22009-11-11 14:26:48 -0800754 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756EXPORT_SYMBOL(setup_arg_pages);
757
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758#endif /* CONFIG_MMU */
759
David Drysdale51f39a12014-12-12 16:57:29 -0800760static struct file *do_open_execat(int fd, struct filename *name, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762 struct file *file;
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200763 int err;
David Drysdale51f39a12014-12-12 16:57:29 -0800764 struct open_flags open_exec_flags = {
Al Viro47c805d2011-02-23 17:44:09 -0500765 .open_flag = O_LARGEFILE | O_RDONLY | __FMODE_EXEC,
Al Viro62fb4a12015-12-26 22:33:24 -0500766 .acc_mode = MAY_EXEC,
Al Virof9652e12013-06-11 08:23:01 +0400767 .intent = LOOKUP_OPEN,
768 .lookup_flags = LOOKUP_FOLLOW,
Al Viro47c805d2011-02-23 17:44:09 -0500769 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700770
David Drysdale51f39a12014-12-12 16:57:29 -0800771 if ((flags & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
772 return ERR_PTR(-EINVAL);
773 if (flags & AT_SYMLINK_NOFOLLOW)
774 open_exec_flags.lookup_flags &= ~LOOKUP_FOLLOW;
775 if (flags & AT_EMPTY_PATH)
776 open_exec_flags.lookup_flags |= LOOKUP_EMPTY;
777
778 file = do_filp_open(fd, name, &open_exec_flags);
Al Viro6e8341a2009-04-06 11:16:22 -0400779 if (IS_ERR(file))
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200780 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200782 err = -EACCES;
Al Viro496ad9a2013-01-23 17:07:38 -0500783 if (!S_ISREG(file_inode(file)->i_mode))
Al Viro6e8341a2009-04-06 11:16:22 -0400784 goto exit;
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200785
Eric W. Biederman90f85722015-06-29 14:42:03 -0500786 if (path_noexec(&file->f_path))
Al Viro6e8341a2009-04-06 11:16:22 -0400787 goto exit;
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200788
789 err = deny_write_access(file);
Al Viro6e8341a2009-04-06 11:16:22 -0400790 if (err)
791 goto exit;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792
David Drysdale51f39a12014-12-12 16:57:29 -0800793 if (name->name[0] != '\0')
794 fsnotify_open(file);
795
Al Viro6e8341a2009-04-06 11:16:22 -0400796out:
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200797 return file;
798
Al Viro6e8341a2009-04-06 11:16:22 -0400799exit:
800 fput(file);
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200801 return ERR_PTR(err);
802}
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -0800803
804struct file *open_exec(const char *name)
805{
Paul Moore51689102015-01-22 00:00:03 -0500806 struct filename *filename = getname_kernel(name);
807 struct file *f = ERR_CAST(filename);
808
809 if (!IS_ERR(filename)) {
810 f = do_open_execat(AT_FDCWD, filename, 0);
811 putname(filename);
812 }
813 return f;
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -0800814}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700815EXPORT_SYMBOL(open_exec);
816
Mimi Zohar6777d772009-08-21 14:32:48 -0400817int kernel_read(struct file *file, loff_t offset,
818 char *addr, unsigned long count)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700819{
820 mm_segment_t old_fs;
821 loff_t pos = offset;
822 int result;
823
824 old_fs = get_fs();
825 set_fs(get_ds());
826 /* The cast to a user pointer is valid due to the set_fs() */
827 result = vfs_read(file, (void __user *)addr, count, &pos);
828 set_fs(old_fs);
829 return result;
830}
831
832EXPORT_SYMBOL(kernel_read);
833
Al Viro3dc20cb2013-04-13 20:31:37 -0400834ssize_t read_code(struct file *file, unsigned long addr, loff_t pos, size_t len)
835{
Al Viroec695572014-02-04 19:08:21 -0500836 ssize_t res = vfs_read(file, (void __user *)addr, len, &pos);
Al Viro3dc20cb2013-04-13 20:31:37 -0400837 if (res > 0)
838 flush_icache_range(addr, addr + len);
839 return res;
840}
841EXPORT_SYMBOL(read_code);
842
Linus Torvalds1da177e2005-04-16 15:20:36 -0700843static int exec_mmap(struct mm_struct *mm)
844{
845 struct task_struct *tsk;
Davidlohr Bueso615d6e82014-04-07 15:37:25 -0700846 struct mm_struct *old_mm, *active_mm;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847
848 /* Notify parent that we're no longer interested in the old VM */
849 tsk = current;
850 old_mm = current->mm;
851 mm_release(tsk, old_mm);
852
853 if (old_mm) {
Konstantin Khlebnikov4fe7efd2012-06-20 12:53:01 -0700854 sync_mm_rss(old_mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855 /*
856 * Make sure that if there is a core dump in progress
857 * for the old mm, we get out and die instead of going
858 * through with the exec. We must hold mmap_sem around
Oleg Nesterov999d9fc2008-07-25 01:47:41 -0700859 * checking core_state and changing tsk->mm.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860 */
861 down_read(&old_mm->mmap_sem);
Oleg Nesterov999d9fc2008-07-25 01:47:41 -0700862 if (unlikely(old_mm->core_state)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863 up_read(&old_mm->mmap_sem);
864 return -EINTR;
865 }
866 }
867 task_lock(tsk);
868 active_mm = tsk->active_mm;
869 tsk->mm = mm;
870 tsk->active_mm = mm;
871 activate_mm(active_mm, mm);
Davidlohr Bueso615d6e82014-04-07 15:37:25 -0700872 tsk->mm->vmacache_seqnum = 0;
873 vmacache_flush(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874 task_unlock(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875 if (old_mm) {
876 up_read(&old_mm->mmap_sem);
Eric Sesterhenn7dddb122006-04-01 01:13:38 +0200877 BUG_ON(active_mm != old_mm);
Oleg Nesterov701085b2012-03-19 17:04:01 +0100878 setmax_mm_hiwater_rss(&tsk->signal->maxrss, old_mm);
Balbir Singh31a78f22008-09-28 23:09:31 +0100879 mm_update_next_owner(old_mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700880 mmput(old_mm);
881 return 0;
882 }
883 mmdrop(active_mm);
884 return 0;
885}
886
887/*
888 * This function makes sure the current process has its own signal table,
889 * so that flush_signal_handlers can later reset the handlers without
890 * disturbing other processes. (Other processes might share the signal
891 * table via the CLONE_SIGHAND option to clone().)
892 */
Arjan van de Ven858119e2006-01-14 13:20:43 -0800893static int de_thread(struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700894{
895 struct signal_struct *sig = tsk->signal;
Oleg Nesterovb2c903b2007-10-16 23:27:22 -0700896 struct sighand_struct *oldsighand = tsk->sighand;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700897 spinlock_t *lock = &oldsighand->siglock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700899 if (thread_group_empty(tsk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700900 goto no_thread_group;
901
902 /*
903 * Kill all other threads in the thread group.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700905 spin_lock_irq(lock);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -0800906 if (signal_group_exit(sig)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700907 /*
908 * Another group action in progress, just
909 * return so that the signal is processed.
910 */
911 spin_unlock_irq(lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700912 return -EAGAIN;
913 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700914
Oleg Nesterovd3441932010-05-26 14:43:11 -0700915 sig->group_exit_task = tsk;
916 sig->notify_count = zap_other_threads(tsk);
917 if (!thread_group_leader(tsk))
918 sig->notify_count--;
919
920 while (sig->notify_count) {
Oleg Nesterovd5bbd432012-10-08 19:13:01 +0200921 __set_current_state(TASK_KILLABLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922 spin_unlock_irq(lock);
923 schedule();
Oleg Nesterovd5bbd432012-10-08 19:13:01 +0200924 if (unlikely(__fatal_signal_pending(tsk)))
925 goto killed;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700926 spin_lock_irq(lock);
927 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700928 spin_unlock_irq(lock);
929
930 /*
931 * At this point all other threads have exited, all we have to
932 * do is to wait for the thread group leader to become inactive,
933 * and to assume its PID:
934 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700935 if (!thread_group_leader(tsk)) {
Oleg Nesterov81879262008-12-01 14:18:16 -0800936 struct task_struct *leader = tsk->group_leader;
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700937
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700938 for (;;) {
Oleg Nesterove56fb282013-04-30 15:28:20 -0700939 threadgroup_change_begin(tsk);
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700940 write_lock_irq(&tasklist_lock);
Kirill Tkhaidfcce792015-04-16 12:48:01 -0700941 /*
942 * Do this under tasklist_lock to ensure that
943 * exit_notify() can't miss ->group_exit_task
944 */
945 sig->notify_count = -1;
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700946 if (likely(leader->exit_state))
947 break;
Oleg Nesterovd5bbd432012-10-08 19:13:01 +0200948 __set_current_state(TASK_KILLABLE);
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700949 write_unlock_irq(&tasklist_lock);
Oleg Nesterove56fb282013-04-30 15:28:20 -0700950 threadgroup_change_end(tsk);
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700951 schedule();
Oleg Nesterovd5bbd432012-10-08 19:13:01 +0200952 if (unlikely(__fatal_signal_pending(tsk)))
953 goto killed;
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700954 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700955
Roland McGrathf5e90282006-04-10 22:54:16 -0700956 /*
957 * The only record we have of the real-time age of a
958 * process, regardless of execs it's done, is start_time.
959 * All the past CPU time is accumulated in signal_struct
960 * from sister threads now dead. But in this non-leader
961 * exec, nothing survives from the original leader thread,
962 * whose birth marks the true age of this process now.
963 * When we take on its identity by switching to its PID, we
964 * also take its birthdate (always earlier than our own).
965 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700966 tsk->start_time = leader->start_time;
Oleg Nesterov266b7a02013-07-03 15:08:35 -0700967 tsk->real_start_time = leader->real_start_time;
Roland McGrathf5e90282006-04-10 22:54:16 -0700968
Pavel Emelyanovbac0abd2007-10-18 23:40:18 -0700969 BUG_ON(!same_thread_group(leader, tsk));
970 BUG_ON(has_group_leader_pid(tsk));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700971 /*
972 * An exec() starts a new thread group with the
973 * TGID of the previous thread group. Rehash the
974 * two threads with a switched PID, and release
975 * the former thread group leader:
976 */
Eric W. Biedermand73d6522006-03-28 16:11:03 -0800977
978 /* Become a process group leader with the old leader's pid.
Eric W. Biedermanc18258c2006-09-27 01:51:06 -0700979 * The old leader becomes a thread of the this thread group.
980 * Note: The old leader also uses this pid until release_task
Eric W. Biedermand73d6522006-03-28 16:11:03 -0800981 * is called. Odd but simple and correct.
982 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700983 tsk->pid = leader->pid;
Oleg Nesterov3f418542013-07-03 15:08:25 -0700984 change_pid(tsk, PIDTYPE_PID, task_pid(leader));
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700985 transfer_pid(leader, tsk, PIDTYPE_PGID);
986 transfer_pid(leader, tsk, PIDTYPE_SID);
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -0800987
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700988 list_replace_rcu(&leader->tasks, &tsk->tasks);
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -0800989 list_replace_init(&leader->sibling, &tsk->sibling);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700991 tsk->group_leader = tsk;
992 leader->group_leader = tsk;
Eric W. Biedermande12a782006-04-10 17:16:49 -0600993
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700994 tsk->exit_signal = SIGCHLD;
Oleg Nesterov087806b2011-06-22 23:10:26 +0200995 leader->exit_signal = -1;
Oleg Nesterov962b5642005-11-23 13:37:43 -0800996
997 BUG_ON(leader->exit_state != EXIT_ZOMBIE);
998 leader->exit_state = EXIT_DEAD;
Oleg Nesteroveac1b5e2011-07-21 20:00:43 +0200999
1000 /*
1001 * We are going to release_task()->ptrace_unlink() silently,
1002 * the tracer can sleep in do_wait(). EXIT_DEAD guarantees
1003 * the tracer wont't block again waiting for this thread.
1004 */
1005 if (unlikely(leader->ptrace))
1006 __wake_up_parent(leader, leader->parent);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007 write_unlock_irq(&tasklist_lock);
Oleg Nesterove56fb282013-04-30 15:28:20 -07001008 threadgroup_change_end(tsk);
Oleg Nesterov81879262008-12-01 14:18:16 -08001009
1010 release_task(leader);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -08001011 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001013 sig->group_exit_task = NULL;
1014 sig->notify_count = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001015
1016no_thread_group:
Oleg Nesterove6368252012-03-19 17:03:22 +01001017 /* we have changed execution domain */
1018 tsk->exit_signal = SIGCHLD;
1019
Linus Torvalds1da177e2005-04-16 15:20:36 -07001020 exit_itimers(sig);
Oleg Nesterovcbaffba2008-05-26 20:55:42 +04001021 flush_itimer_signals();
Oleg Nesterov329f7db2005-11-07 21:12:43 +03001022
Oleg Nesterovb2c903b2007-10-16 23:27:22 -07001023 if (atomic_read(&oldsighand->count) != 1) {
1024 struct sighand_struct *newsighand;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001025 /*
Oleg Nesterovb2c903b2007-10-16 23:27:22 -07001026 * This ->sighand is shared with the CLONE_SIGHAND
1027 * but not CLONE_THREAD task, switch to the new one.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001028 */
Oleg Nesterovb2c903b2007-10-16 23:27:22 -07001029 newsighand = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
1030 if (!newsighand)
1031 return -ENOMEM;
1032
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033 atomic_set(&newsighand->count, 1);
1034 memcpy(newsighand->action, oldsighand->action,
1035 sizeof(newsighand->action));
1036
1037 write_lock_irq(&tasklist_lock);
1038 spin_lock(&oldsighand->siglock);
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001039 rcu_assign_pointer(tsk->sighand, newsighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001040 spin_unlock(&oldsighand->siglock);
1041 write_unlock_irq(&tasklist_lock);
1042
Davide Libenzifba2afa2007-05-10 22:23:13 -07001043 __cleanup_sighand(oldsighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001044 }
1045
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001046 BUG_ON(!thread_group_leader(tsk));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001047 return 0;
Oleg Nesterovd5bbd432012-10-08 19:13:01 +02001048
1049killed:
1050 /* protects against exit_notify() and __exit_signal() */
1051 read_lock(&tasklist_lock);
1052 sig->group_exit_task = NULL;
1053 sig->notify_count = 0;
1054 read_unlock(&tasklist_lock);
1055 return -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001056}
Oleg Nesterov0840a902007-10-16 23:27:22 -07001057
Andrew Morton59714d62008-02-04 22:27:21 -08001058char *get_task_comm(char *buf, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001059{
1060 /* buf must be at least sizeof(tsk->comm) in size */
1061 task_lock(tsk);
1062 strncpy(buf, tsk->comm, sizeof(tsk->comm));
1063 task_unlock(tsk);
Andrew Morton59714d62008-02-04 22:27:21 -08001064 return buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001065}
J Freyensee7d74f492011-05-06 16:56:47 -07001066EXPORT_SYMBOL_GPL(get_task_comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001067
Al Viro6a6d27d2012-08-21 09:56:33 -04001068/*
1069 * These functions flushes out all traces of the currently running executable
1070 * so that a new one can be started
1071 */
1072
Adrian Hunter82b89772014-05-28 11:45:04 +03001073void __set_task_comm(struct task_struct *tsk, const char *buf, bool exec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001074{
1075 task_lock(tsk);
KAMEZAWA Hiroyuki43d2b112012-01-10 15:08:09 -08001076 trace_task_rename(tsk, buf);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001077 strlcpy(tsk->comm, buf, sizeof(tsk->comm));
1078 task_unlock(tsk);
Adrian Hunter82b89772014-05-28 11:45:04 +03001079 perf_event_comm(tsk, exec);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001080}
1081
1082int flush_old_exec(struct linux_binprm * bprm)
1083{
Linus Torvalds221af7f2010-01-28 22:14:42 -08001084 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001085
1086 /*
1087 * Make sure we have a private signal table and that
1088 * we are unassociated from the previous thread group.
1089 */
1090 retval = de_thread(current);
1091 if (retval)
1092 goto out;
1093
Davidlohr Bueso6e399cd2015-04-16 12:47:59 -07001094 /*
1095 * Must be called _before_ exec_mmap() as bprm->mm is
1096 * not visibile until then. This also enables the update
1097 * to be lockless.
1098 */
Matt Helsley925d1c42008-04-29 01:01:36 -07001099 set_mm_exe_file(bprm->mm, bprm->file);
Davidlohr Bueso6e399cd2015-04-16 12:47:59 -07001100
Linus Torvalds1da177e2005-04-16 15:20:36 -07001101 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001102 * Release all of the old mmap stuff
1103 */
Oleg Nesterov3c77f842010-11-30 20:55:34 +01001104 acct_arg_size(bprm, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001105 retval = exec_mmap(bprm->mm);
1106 if (retval)
Al Virofd8328b2008-04-22 05:11:59 -04001107 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108
1109 bprm->mm = NULL; /* We're using it now */
Linus Torvalds7ab02af2010-02-02 12:37:44 -08001110
Mathias Krausedac853a2011-06-09 20:05:18 +02001111 set_fs(USER_DS);
Zhang Yib88fae62014-01-23 15:55:57 -08001112 current->flags &= ~(PF_RANDOMIZE | PF_FORKNOEXEC | PF_KTHREAD |
1113 PF_NOFREEZE | PF_NO_SETAFFINITY);
Linus Torvalds7ab02af2010-02-02 12:37:44 -08001114 flush_thread();
1115 current->personality &= ~bprm->per_clear;
1116
Linus Torvalds221af7f2010-01-28 22:14:42 -08001117 return 0;
1118
1119out:
1120 return retval;
1121}
1122EXPORT_SYMBOL(flush_old_exec);
1123
Al Viro1b5d7832011-06-19 12:49:47 -04001124void would_dump(struct linux_binprm *bprm, struct file *file)
1125{
Al Viro496ad9a2013-01-23 17:07:38 -05001126 if (inode_permission(file_inode(file), MAY_READ) < 0)
Al Viro1b5d7832011-06-19 12:49:47 -04001127 bprm->interp_flags |= BINPRM_FLAGS_ENFORCE_NONDUMP;
1128}
1129EXPORT_SYMBOL(would_dump);
1130
Linus Torvalds221af7f2010-01-28 22:14:42 -08001131void setup_new_exec(struct linux_binprm * bprm)
1132{
Linus Torvalds221af7f2010-01-28 22:14:42 -08001133 arch_pick_mmap_layout(current->mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001134
1135 /* This is the point of no return */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001136 current->sas_ss_sp = current->sas_ss_size = 0;
1137
Eric W. Biederman8e96e3b2012-03-03 21:17:15 -08001138 if (uid_eq(current_euid(), current_uid()) && gid_eq(current_egid(), current_gid()))
Kees Cooke579d2c2013-02-27 17:03:15 -08001139 set_dumpable(current->mm, SUID_DUMP_USER);
Alan Coxd6e71142005-06-23 00:09:43 -07001140 else
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001141 set_dumpable(current->mm, suid_dumpable);
Alan Coxd6e71142005-06-23 00:09:43 -07001142
Peter Zijlstrae041e322014-05-21 17:32:19 +02001143 perf_event_exec();
Adrian Hunter82b89772014-05-28 11:45:04 +03001144 __set_task_comm(current, kbasename(bprm->filename), true);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001145
Benjamin Herrenschmidt0551fbd2006-02-28 16:59:19 -08001146 /* Set the new mm task size. We have to do that late because it may
1147 * depend on TIF_32BIT which is only updated in flush_thread() on
1148 * some architectures like powerpc
1149 */
1150 current->mm->task_size = TASK_SIZE;
1151
David Howellsa6f76f22008-11-14 10:39:24 +11001152 /* install the new credentials */
Eric W. Biederman8e96e3b2012-03-03 21:17:15 -08001153 if (!uid_eq(bprm->cred->uid, current_euid()) ||
1154 !gid_eq(bprm->cred->gid, current_egid())) {
Marcel Holtmannd2d56c52007-08-17 21:47:58 +02001155 current->pdeath_signal = 0;
Al Viro1b5d7832011-06-19 12:49:47 -04001156 } else {
1157 would_dump(bprm, bprm->file);
1158 if (bprm->interp_flags & BINPRM_FLAGS_ENFORCE_NONDUMP)
1159 set_dumpable(current->mm, suid_dumpable);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001160 }
1161
1162 /* An exec changes our domain. We are no longer part of the thread
1163 group */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001164 current->self_exec_id++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001165 flush_signal_handlers(current, 0);
Al Viro6a6d27d2012-08-21 09:56:33 -04001166 do_close_on_exec(current->files);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001167}
Linus Torvalds221af7f2010-01-28 22:14:42 -08001168EXPORT_SYMBOL(setup_new_exec);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001169
David Howellsa6f76f22008-11-14 10:39:24 +11001170/*
Oleg Nesterova2a84742009-09-05 11:17:13 -07001171 * Prepare credentials and lock ->cred_guard_mutex.
1172 * install_exec_creds() commits the new creds and drops the lock.
1173 * Or, if exec fails before, free_bprm() should release ->cred and
1174 * and unlock.
1175 */
1176int prepare_bprm_creds(struct linux_binprm *bprm)
1177{
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001178 if (mutex_lock_interruptible(&current->signal->cred_guard_mutex))
Oleg Nesterova2a84742009-09-05 11:17:13 -07001179 return -ERESTARTNOINTR;
1180
1181 bprm->cred = prepare_exec_creds();
1182 if (likely(bprm->cred))
1183 return 0;
1184
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001185 mutex_unlock(&current->signal->cred_guard_mutex);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001186 return -ENOMEM;
1187}
1188
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001189static void free_bprm(struct linux_binprm *bprm)
Oleg Nesterova2a84742009-09-05 11:17:13 -07001190{
1191 free_arg_pages(bprm);
1192 if (bprm->cred) {
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001193 mutex_unlock(&current->signal->cred_guard_mutex);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001194 abort_creds(bprm->cred);
1195 }
Oleg Nesterov63e46b92014-01-23 15:55:51 -08001196 if (bprm->file) {
1197 allow_write_access(bprm->file);
1198 fput(bprm->file);
1199 }
Kees Cookb66c5982012-12-20 15:05:16 -08001200 /* If a binfmt changed the interp, free it. */
1201 if (bprm->interp != bprm->filename)
1202 kfree(bprm->interp);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001203 kfree(bprm);
1204}
1205
Kees Cookb66c5982012-12-20 15:05:16 -08001206int bprm_change_interp(char *interp, struct linux_binprm *bprm)
1207{
1208 /* If a binfmt changed the interp, free it first. */
1209 if (bprm->interp != bprm->filename)
1210 kfree(bprm->interp);
1211 bprm->interp = kstrdup(interp, GFP_KERNEL);
1212 if (!bprm->interp)
1213 return -ENOMEM;
1214 return 0;
1215}
1216EXPORT_SYMBOL(bprm_change_interp);
1217
Oleg Nesterova2a84742009-09-05 11:17:13 -07001218/*
David Howellsa6f76f22008-11-14 10:39:24 +11001219 * install the new credentials for this executable
1220 */
1221void install_exec_creds(struct linux_binprm *bprm)
1222{
1223 security_bprm_committing_creds(bprm);
1224
1225 commit_creds(bprm->cred);
1226 bprm->cred = NULL;
Stephane Eranian2976b102013-06-20 11:36:28 +02001227
1228 /*
1229 * Disable monitoring for regular users
1230 * when executing setuid binaries. Must
1231 * wait until new credentials are committed
1232 * by commit_creds() above
1233 */
1234 if (get_dumpable(current->mm) != SUID_DUMP_USER)
1235 perf_event_exit_task(current);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001236 /*
1237 * cred_guard_mutex must be held at least to this point to prevent
David Howellsa6f76f22008-11-14 10:39:24 +11001238 * ptrace_attach() from altering our determination of the task's
Oleg Nesterova2a84742009-09-05 11:17:13 -07001239 * credentials; any time after this it may be unlocked.
1240 */
David Howellsa6f76f22008-11-14 10:39:24 +11001241 security_bprm_committed_creds(bprm);
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001242 mutex_unlock(&current->signal->cred_guard_mutex);
David Howellsa6f76f22008-11-14 10:39:24 +11001243}
1244EXPORT_SYMBOL(install_exec_creds);
1245
1246/*
1247 * determine how safe it is to execute the proposed program
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001248 * - the caller must hold ->cred_guard_mutex to protect against
Kees Cookc2e1f2e2014-06-05 00:23:17 -07001249 * PTRACE_ATTACH or seccomp thread-sync
David Howellsa6f76f22008-11-14 10:39:24 +11001250 */
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001251static void check_unsafe_exec(struct linux_binprm *bprm)
David Howellsa6f76f22008-11-14 10:39:24 +11001252{
David Howells0bf2f3a2009-02-06 11:45:46 +00001253 struct task_struct *p = current, *t;
Al Virof1191b52009-03-30 07:35:18 -04001254 unsigned n_fs;
David Howellsa6f76f22008-11-14 10:39:24 +11001255
Tejun Heo4b9d33e2011-06-17 16:50:38 +02001256 if (p->ptrace) {
1257 if (p->ptrace & PT_PTRACE_CAP)
1258 bprm->unsafe |= LSM_UNSAFE_PTRACE_CAP;
1259 else
1260 bprm->unsafe |= LSM_UNSAFE_PTRACE;
1261 }
David Howellsa6f76f22008-11-14 10:39:24 +11001262
Andy Lutomirski259e5e62012-04-12 16:47:50 -05001263 /*
1264 * This isn't strictly necessary, but it makes it harder for LSMs to
1265 * mess up.
1266 */
Kees Cook1d4457f2014-05-21 15:23:46 -07001267 if (task_no_new_privs(current))
Andy Lutomirski259e5e62012-04-12 16:47:50 -05001268 bprm->unsafe |= LSM_UNSAFE_NO_NEW_PRIVS;
1269
Oleg Nesterov83f62a22014-01-23 15:55:49 -08001270 t = p;
David Howells0bf2f3a2009-02-06 11:45:46 +00001271 n_fs = 1;
Nick Piggin2a4419b2010-08-18 04:37:33 +10001272 spin_lock(&p->fs->lock);
Oleg Nesterov437f7fd2009-04-24 01:02:45 +02001273 rcu_read_lock();
Oleg Nesterov83f62a22014-01-23 15:55:49 -08001274 while_each_thread(p, t) {
David Howells0bf2f3a2009-02-06 11:45:46 +00001275 if (t->fs == p->fs)
1276 n_fs++;
David Howells0bf2f3a2009-02-06 11:45:46 +00001277 }
Oleg Nesterov437f7fd2009-04-24 01:02:45 +02001278 rcu_read_unlock();
David Howells0bf2f3a2009-02-06 11:45:46 +00001279
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001280 if (p->fs->users > n_fs)
David Howellsa6f76f22008-11-14 10:39:24 +11001281 bprm->unsafe |= LSM_UNSAFE_SHARE;
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001282 else
1283 p->fs->in_exec = 1;
Nick Piggin2a4419b2010-08-18 04:37:33 +10001284 spin_unlock(&p->fs->lock);
David Howellsa6f76f22008-11-14 10:39:24 +11001285}
1286
Jann Horn8b01fc82015-04-19 02:48:39 +02001287static void bprm_fill_uid(struct linux_binprm *bprm)
1288{
1289 struct inode *inode;
1290 unsigned int mode;
1291 kuid_t uid;
1292 kgid_t gid;
1293
1294 /* clear any previous set[ug]id data from a previous binary */
1295 bprm->cred->euid = current_euid();
1296 bprm->cred->egid = current_egid();
1297
1298 if (bprm->file->f_path.mnt->mnt_flags & MNT_NOSUID)
1299 return;
1300
1301 if (task_no_new_privs(current))
1302 return;
1303
1304 inode = file_inode(bprm->file);
1305 mode = READ_ONCE(inode->i_mode);
1306 if (!(mode & (S_ISUID|S_ISGID)))
1307 return;
1308
1309 /* Be careful if suid/sgid is set */
1310 mutex_lock(&inode->i_mutex);
1311
1312 /* reload atomically mode/uid/gid now that lock held */
1313 mode = inode->i_mode;
1314 uid = inode->i_uid;
1315 gid = inode->i_gid;
1316 mutex_unlock(&inode->i_mutex);
1317
1318 /* We ignore suid/sgid if there are no mappings for them in the ns */
1319 if (!kuid_has_mapping(bprm->cred->user_ns, uid) ||
1320 !kgid_has_mapping(bprm->cred->user_ns, gid))
1321 return;
1322
1323 if (mode & S_ISUID) {
1324 bprm->per_clear |= PER_CLEAR_ON_SETID;
1325 bprm->cred->euid = uid;
1326 }
1327
1328 if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) {
1329 bprm->per_clear |= PER_CLEAR_ON_SETID;
1330 bprm->cred->egid = gid;
1331 }
1332}
1333
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001334/*
1335 * Fill the binprm structure from the inode.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001336 * Check permissions, then read the first 128 (BINPRM_BUF_SIZE) bytes
David Howellsa6f76f22008-11-14 10:39:24 +11001337 *
1338 * This may be called multiple times for binary chains (scripts for example).
Linus Torvalds1da177e2005-04-16 15:20:36 -07001339 */
1340int prepare_binprm(struct linux_binprm *bprm)
1341{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001342 int retval;
1343
Jann Horn8b01fc82015-04-19 02:48:39 +02001344 bprm_fill_uid(bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001345
1346 /* fill in binprm security blob */
David Howellsa6f76f22008-11-14 10:39:24 +11001347 retval = security_bprm_set_creds(bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001348 if (retval)
1349 return retval;
David Howellsa6f76f22008-11-14 10:39:24 +11001350 bprm->cred_prepared = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001351
David Howellsa6f76f22008-11-14 10:39:24 +11001352 memset(bprm->buf, 0, BINPRM_BUF_SIZE);
1353 return kernel_read(bprm->file, 0, bprm->buf, BINPRM_BUF_SIZE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001354}
1355
1356EXPORT_SYMBOL(prepare_binprm);
1357
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001358/*
1359 * Arguments are '\0' separated strings found at the location bprm->p
1360 * points to; chop off the first by relocating brpm->p to right after
1361 * the first '\0' encountered.
1362 */
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001363int remove_arg_zero(struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001364{
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001365 int ret = 0;
1366 unsigned long offset;
1367 char *kaddr;
1368 struct page *page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001369
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001370 if (!bprm->argc)
1371 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001372
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001373 do {
1374 offset = bprm->p & ~PAGE_MASK;
1375 page = get_arg_page(bprm, bprm->p, 0);
1376 if (!page) {
1377 ret = -EFAULT;
1378 goto out;
1379 }
Cong Wange8e3c3d2011-11-25 23:14:27 +08001380 kaddr = kmap_atomic(page);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001381
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001382 for (; offset < PAGE_SIZE && kaddr[offset];
1383 offset++, bprm->p++)
1384 ;
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001385
Cong Wange8e3c3d2011-11-25 23:14:27 +08001386 kunmap_atomic(kaddr);
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001387 put_arg_page(page);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001388
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001389 if (offset == PAGE_SIZE)
1390 free_arg_page(bprm, (bprm->p >> PAGE_SHIFT) - 1);
1391 } while (offset == PAGE_SIZE);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001392
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001393 bprm->p++;
1394 bprm->argc--;
1395 ret = 0;
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001396
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001397out:
1398 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001399}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001400EXPORT_SYMBOL(remove_arg_zero);
1401
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001402#define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001403/*
1404 * cycle the list of binary formats handler, until one recognizes the image
1405 */
Al Viro3c456bf2012-10-20 21:53:31 -04001406int search_binary_handler(struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001407{
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001408 bool need_retry = IS_ENABLED(CONFIG_MODULES);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001409 struct linux_binfmt *fmt;
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001410 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001411
Kees Cookd7402692012-12-17 16:03:20 -08001412 /* This allows 4 levels of binfmt rewrites before failing hard. */
Oleg Nesterov131b2f92013-09-11 14:24:39 -07001413 if (bprm->recursion_depth > 5)
Kees Cookd7402692012-12-17 16:03:20 -08001414 return -ELOOP;
1415
Linus Torvalds1da177e2005-04-16 15:20:36 -07001416 retval = security_bprm_check(bprm);
1417 if (retval)
1418 return retval;
1419
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420 retval = -ENOENT;
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001421 retry:
1422 read_lock(&binfmt_lock);
1423 list_for_each_entry(fmt, &formats, lh) {
1424 if (!try_module_get(fmt->module))
1425 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001426 read_unlock(&binfmt_lock);
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001427 bprm->recursion_depth++;
1428 retval = fmt->load_binary(bprm);
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001429 read_lock(&binfmt_lock);
1430 put_binfmt(fmt);
Al Viro19d860a2014-05-04 20:11:36 -04001431 bprm->recursion_depth--;
1432 if (retval < 0 && !bprm->mm) {
1433 /* we got to flush_old_exec() and failed after it */
1434 read_unlock(&binfmt_lock);
1435 force_sigsegv(SIGSEGV, current);
1436 return retval;
1437 }
1438 if (retval != -ENOEXEC || !bprm->file) {
1439 read_unlock(&binfmt_lock);
1440 return retval;
1441 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001442 }
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001443 read_unlock(&binfmt_lock);
1444
Al Viro19d860a2014-05-04 20:11:36 -04001445 if (need_retry) {
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001446 if (printable(bprm->buf[0]) && printable(bprm->buf[1]) &&
1447 printable(bprm->buf[2]) && printable(bprm->buf[3]))
1448 return retval;
Oleg Nesterov4e0621a2013-09-11 14:24:45 -07001449 if (request_module("binfmt-%04x", *(ushort *)(bprm->buf + 2)) < 0)
1450 return retval;
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001451 need_retry = false;
1452 goto retry;
1453 }
1454
Linus Torvalds1da177e2005-04-16 15:20:36 -07001455 return retval;
1456}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001457EXPORT_SYMBOL(search_binary_handler);
1458
Oleg Nesterov5d1baf32013-09-11 14:24:38 -07001459static int exec_binprm(struct linux_binprm *bprm)
1460{
1461 pid_t old_pid, old_vpid;
1462 int ret;
1463
1464 /* Need to fetch pid before load_binary changes it */
1465 old_pid = current->pid;
1466 rcu_read_lock();
1467 old_vpid = task_pid_nr_ns(current, task_active_pid_ns(current->parent));
1468 rcu_read_unlock();
1469
1470 ret = search_binary_handler(bprm);
1471 if (ret >= 0) {
Linus Torvalds3eaded862013-11-21 19:18:14 -08001472 audit_bprm(bprm);
Oleg Nesterov5d1baf32013-09-11 14:24:38 -07001473 trace_sched_process_exec(current, old_pid, bprm);
1474 ptrace_event(PTRACE_EVENT_EXEC, old_vpid);
Oleg Nesterov9beb2662013-09-11 14:24:40 -07001475 proc_exec_connector(current);
Oleg Nesterov5d1baf32013-09-11 14:24:38 -07001476 }
1477
1478 return ret;
1479}
1480
Linus Torvalds1da177e2005-04-16 15:20:36 -07001481/*
1482 * sys_execve() executes a new program.
1483 */
David Drysdale51f39a12014-12-12 16:57:29 -08001484static int do_execveat_common(int fd, struct filename *filename,
1485 struct user_arg_ptr argv,
1486 struct user_arg_ptr envp,
1487 int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001488{
David Drysdale51f39a12014-12-12 16:57:29 -08001489 char *pathbuf = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001490 struct linux_binprm *bprm;
1491 struct file *file;
Al Viro3b125382008-04-22 05:31:30 -04001492 struct files_struct *displaced;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001493 int retval;
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +04001494
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001495 if (IS_ERR(filename))
1496 return PTR_ERR(filename);
1497
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +04001498 /*
1499 * We move the actual failure in case of RLIMIT_NPROC excess from
1500 * set*uid() to execve() because too many poorly written programs
1501 * don't check setuid() return code. Here we additionally recheck
1502 * whether NPROC limit is still exceeded.
1503 */
1504 if ((current->flags & PF_NPROC_EXCEEDED) &&
Oleg Nesterovbd9d43f2013-07-03 15:08:34 -07001505 atomic_read(&current_user()->processes) > rlimit(RLIMIT_NPROC)) {
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +04001506 retval = -EAGAIN;
1507 goto out_ret;
1508 }
1509
1510 /* We're below the limit (still or again), so we don't want to make
1511 * further execve() calls fail. */
1512 current->flags &= ~PF_NPROC_EXCEEDED;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001513
Al Viro3b125382008-04-22 05:31:30 -04001514 retval = unshare_files(&displaced);
Al Virofd8328b2008-04-22 05:11:59 -04001515 if (retval)
1516 goto out_ret;
1517
Linus Torvalds1da177e2005-04-16 15:20:36 -07001518 retval = -ENOMEM;
Oliver Neukum11b0b5a2006-03-25 03:08:13 -08001519 bprm = kzalloc(sizeof(*bprm), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001520 if (!bprm)
Al Virofd8328b2008-04-22 05:11:59 -04001521 goto out_files;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001522
Oleg Nesterova2a84742009-09-05 11:17:13 -07001523 retval = prepare_bprm_creds(bprm);
1524 if (retval)
David Howellsa6f76f22008-11-14 10:39:24 +11001525 goto out_free;
Al Viro498052b2009-03-30 07:20:30 -04001526
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001527 check_unsafe_exec(bprm);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001528 current->in_execve = 1;
David Howellsa6f76f22008-11-14 10:39:24 +11001529
David Drysdale51f39a12014-12-12 16:57:29 -08001530 file = do_open_execat(fd, filename, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001531 retval = PTR_ERR(file);
1532 if (IS_ERR(file))
Al Viro498052b2009-03-30 07:20:30 -04001533 goto out_unmark;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001534
1535 sched_exec();
1536
Linus Torvalds1da177e2005-04-16 15:20:36 -07001537 bprm->file = file;
David Drysdale51f39a12014-12-12 16:57:29 -08001538 if (fd == AT_FDCWD || filename->name[0] == '/') {
1539 bprm->filename = filename->name;
1540 } else {
1541 if (filename->name[0] == '\0')
1542 pathbuf = kasprintf(GFP_TEMPORARY, "/dev/fd/%d", fd);
1543 else
1544 pathbuf = kasprintf(GFP_TEMPORARY, "/dev/fd/%d/%s",
1545 fd, filename->name);
1546 if (!pathbuf) {
1547 retval = -ENOMEM;
1548 goto out_unmark;
1549 }
1550 /*
1551 * Record that a name derived from an O_CLOEXEC fd will be
1552 * inaccessible after exec. Relies on having exclusive access to
1553 * current->files (due to unshare_files above).
1554 */
1555 if (close_on_exec(fd, rcu_dereference_raw(current->files->fdt)))
1556 bprm->interp_flags |= BINPRM_FLAGS_PATH_INACCESSIBLE;
1557 bprm->filename = pathbuf;
1558 }
1559 bprm->interp = bprm->filename;
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001560
1561 retval = bprm_mm_init(bprm);
1562 if (retval)
Oleg Nesterov63e46b92014-01-23 15:55:51 -08001563 goto out_unmark;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001564
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001565 bprm->argc = count(argv, MAX_ARG_STRINGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001566 if ((retval = bprm->argc) < 0)
David Howellsa6f76f22008-11-14 10:39:24 +11001567 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001568
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001569 bprm->envc = count(envp, MAX_ARG_STRINGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001570 if ((retval = bprm->envc) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001571 goto out;
1572
1573 retval = prepare_binprm(bprm);
1574 if (retval < 0)
1575 goto out;
1576
1577 retval = copy_strings_kernel(1, &bprm->filename, bprm);
1578 if (retval < 0)
1579 goto out;
1580
1581 bprm->exec = bprm->p;
1582 retval = copy_strings(bprm->envc, envp, bprm);
1583 if (retval < 0)
1584 goto out;
1585
1586 retval = copy_strings(bprm->argc, argv, bprm);
1587 if (retval < 0)
1588 goto out;
1589
Oleg Nesterov5d1baf32013-09-11 14:24:38 -07001590 retval = exec_binprm(bprm);
David Howellsa6f76f22008-11-14 10:39:24 +11001591 if (retval < 0)
1592 goto out;
1593
1594 /* execve succeeded */
Al Viro498052b2009-03-30 07:20:30 -04001595 current->fs->in_exec = 0;
Kentaro Takedaf9ce1f12009-02-05 17:18:11 +09001596 current->in_execve = 0;
David Howellsa6f76f22008-11-14 10:39:24 +11001597 acct_update_integrals(current);
Rik van Riel82727012013-10-07 11:29:28 +01001598 task_numa_free(current);
David Howellsa6f76f22008-11-14 10:39:24 +11001599 free_bprm(bprm);
David Drysdale51f39a12014-12-12 16:57:29 -08001600 kfree(pathbuf);
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001601 putname(filename);
David Howellsa6f76f22008-11-14 10:39:24 +11001602 if (displaced)
1603 put_files_struct(displaced);
1604 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001605
1606out:
Oleg Nesterov3c77f842010-11-30 20:55:34 +01001607 if (bprm->mm) {
1608 acct_arg_size(bprm, 0);
1609 mmput(bprm->mm);
1610 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001611
Al Viro498052b2009-03-30 07:20:30 -04001612out_unmark:
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001613 current->fs->in_exec = 0;
Kentaro Takedaf9ce1f12009-02-05 17:18:11 +09001614 current->in_execve = 0;
David Howellsa6f76f22008-11-14 10:39:24 +11001615
1616out_free:
Al Viro08a6fac2008-05-10 16:38:25 -04001617 free_bprm(bprm);
David Drysdale51f39a12014-12-12 16:57:29 -08001618 kfree(pathbuf);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001619
Al Virofd8328b2008-04-22 05:11:59 -04001620out_files:
Al Viro3b125382008-04-22 05:31:30 -04001621 if (displaced)
1622 reset_files_struct(displaced);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001623out_ret:
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001624 putname(filename);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001625 return retval;
1626}
1627
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001628int do_execve(struct filename *filename,
Oleg Nesterovba2d01622011-03-06 18:02:37 +01001629 const char __user *const __user *__argv,
Al Viroda3d4c52012-10-20 21:49:33 -04001630 const char __user *const __user *__envp)
Oleg Nesterovba2d01622011-03-06 18:02:37 +01001631{
Oleg Nesterov0e028462011-03-06 18:02:54 +01001632 struct user_arg_ptr argv = { .ptr.native = __argv };
1633 struct user_arg_ptr envp = { .ptr.native = __envp };
David Drysdale51f39a12014-12-12 16:57:29 -08001634 return do_execveat_common(AT_FDCWD, filename, argv, envp, 0);
1635}
1636
1637int do_execveat(int fd, struct filename *filename,
1638 const char __user *const __user *__argv,
1639 const char __user *const __user *__envp,
1640 int flags)
1641{
1642 struct user_arg_ptr argv = { .ptr.native = __argv };
1643 struct user_arg_ptr envp = { .ptr.native = __envp };
1644
1645 return do_execveat_common(fd, filename, argv, envp, flags);
Oleg Nesterovba2d01622011-03-06 18:02:37 +01001646}
1647
Oleg Nesterov0e028462011-03-06 18:02:54 +01001648#ifdef CONFIG_COMPAT
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001649static int compat_do_execve(struct filename *filename,
Al Viro38b983b2012-09-30 13:38:55 -04001650 const compat_uptr_t __user *__argv,
Al Virod03d26e2012-10-20 21:46:25 -04001651 const compat_uptr_t __user *__envp)
Oleg Nesterov0e028462011-03-06 18:02:54 +01001652{
1653 struct user_arg_ptr argv = {
1654 .is_compat = true,
1655 .ptr.compat = __argv,
1656 };
1657 struct user_arg_ptr envp = {
1658 .is_compat = true,
1659 .ptr.compat = __envp,
1660 };
David Drysdale51f39a12014-12-12 16:57:29 -08001661 return do_execveat_common(AT_FDCWD, filename, argv, envp, 0);
1662}
1663
1664static int compat_do_execveat(int fd, struct filename *filename,
1665 const compat_uptr_t __user *__argv,
1666 const compat_uptr_t __user *__envp,
1667 int flags)
1668{
1669 struct user_arg_ptr argv = {
1670 .is_compat = true,
1671 .ptr.compat = __argv,
1672 };
1673 struct user_arg_ptr envp = {
1674 .is_compat = true,
1675 .ptr.compat = __envp,
1676 };
1677 return do_execveat_common(fd, filename, argv, envp, flags);
Oleg Nesterov0e028462011-03-06 18:02:54 +01001678}
1679#endif
1680
Oleg Nesterov964ee7d2009-09-23 15:56:59 -07001681void set_binfmt(struct linux_binfmt *new)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001682{
Hiroshi Shimamoto801460d2009-09-23 15:57:41 -07001683 struct mm_struct *mm = current->mm;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001684
Hiroshi Shimamoto801460d2009-09-23 15:57:41 -07001685 if (mm->binfmt)
1686 module_put(mm->binfmt->module);
1687
1688 mm->binfmt = new;
Oleg Nesterov964ee7d2009-09-23 15:56:59 -07001689 if (new)
1690 __module_get(new->module);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001691}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001692EXPORT_SYMBOL(set_binfmt);
1693
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001694/*
Oleg Nesterov7288e112014-01-23 15:55:32 -08001695 * set_dumpable stores three-value SUID_DUMP_* into mm->flags.
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001696 */
1697void set_dumpable(struct mm_struct *mm, int value)
1698{
Oleg Nesterovabacd2f2014-01-23 15:55:31 -08001699 unsigned long old, new;
1700
Oleg Nesterov7288e112014-01-23 15:55:32 -08001701 if (WARN_ON((unsigned)value > SUID_DUMP_ROOT))
1702 return;
1703
Oleg Nesterovabacd2f2014-01-23 15:55:31 -08001704 do {
1705 old = ACCESS_ONCE(mm->flags);
Oleg Nesterov7288e112014-01-23 15:55:32 -08001706 new = (old & ~MMF_DUMPABLE_MASK) | value;
Oleg Nesterovabacd2f2014-01-23 15:55:31 -08001707 } while (cmpxchg(&mm->flags, old, new) != old);
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001708}
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001709
Al Viro38b983b2012-09-30 13:38:55 -04001710SYSCALL_DEFINE3(execve,
1711 const char __user *, filename,
1712 const char __user *const __user *, argv,
1713 const char __user *const __user *, envp)
1714{
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001715 return do_execve(getname(filename), argv, envp);
Al Viro38b983b2012-09-30 13:38:55 -04001716}
David Drysdale51f39a12014-12-12 16:57:29 -08001717
1718SYSCALL_DEFINE5(execveat,
1719 int, fd, const char __user *, filename,
1720 const char __user *const __user *, argv,
1721 const char __user *const __user *, envp,
1722 int, flags)
1723{
1724 int lookup_flags = (flags & AT_EMPTY_PATH) ? LOOKUP_EMPTY : 0;
1725
1726 return do_execveat(fd,
1727 getname_flags(filename, lookup_flags, NULL),
1728 argv, envp, flags);
1729}
1730
Al Viro38b983b2012-09-30 13:38:55 -04001731#ifdef CONFIG_COMPAT
Heiko Carstens625b1d72014-03-04 10:53:50 +01001732COMPAT_SYSCALL_DEFINE3(execve, const char __user *, filename,
1733 const compat_uptr_t __user *, argv,
1734 const compat_uptr_t __user *, envp)
Al Viro38b983b2012-09-30 13:38:55 -04001735{
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001736 return compat_do_execve(getname(filename), argv, envp);
Al Viro38b983b2012-09-30 13:38:55 -04001737}
David Drysdale51f39a12014-12-12 16:57:29 -08001738
1739COMPAT_SYSCALL_DEFINE5(execveat, int, fd,
1740 const char __user *, filename,
1741 const compat_uptr_t __user *, argv,
1742 const compat_uptr_t __user *, envp,
1743 int, flags)
1744{
1745 int lookup_flags = (flags & AT_EMPTY_PATH) ? LOOKUP_EMPTY : 0;
1746
1747 return compat_do_execveat(fd,
1748 getname_flags(filename, lookup_flags, NULL),
1749 argv, envp, flags);
1750}
Al Viro38b983b2012-09-30 13:38:55 -04001751#endif