blob: 1977c2a553aca711ba145d1670ea9a84fd45ea84 [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;
Dave Hansenfe3d1972014-11-14 07:18:29 -0800280 arch_bprm_mm_init(mm, vma);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700281 up_write(&mm->mmap_sem);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700282 bprm->p = vma->vm_end - sizeof(void *);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700283 return 0;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700284err:
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800285 up_write(&mm->mmap_sem);
286 bprm->vma = NULL;
287 kmem_cache_free(vm_area_cachep, vma);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700288 return err;
289}
290
291static bool valid_arg_len(struct linux_binprm *bprm, long len)
292{
293 return len <= MAX_ARG_STRLEN;
294}
295
296#else
297
Oleg Nesterov0e028462011-03-06 18:02:54 +0100298static inline void acct_arg_size(struct linux_binprm *bprm, unsigned long pages)
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100299{
300}
301
Oleg Nesterov0e028462011-03-06 18:02:54 +0100302static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos,
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700303 int write)
304{
305 struct page *page;
306
307 page = bprm->page[pos / PAGE_SIZE];
308 if (!page && write) {
309 page = alloc_page(GFP_HIGHUSER|__GFP_ZERO);
310 if (!page)
311 return NULL;
312 bprm->page[pos / PAGE_SIZE] = page;
313 }
314
315 return page;
316}
317
318static void put_arg_page(struct page *page)
319{
320}
321
322static void free_arg_page(struct linux_binprm *bprm, int i)
323{
324 if (bprm->page[i]) {
325 __free_page(bprm->page[i]);
326 bprm->page[i] = NULL;
327 }
328}
329
330static void free_arg_pages(struct linux_binprm *bprm)
331{
332 int i;
333
334 for (i = 0; i < MAX_ARG_PAGES; i++)
335 free_arg_page(bprm, i);
336}
337
338static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos,
339 struct page *page)
340{
341}
342
343static int __bprm_mm_init(struct linux_binprm *bprm)
344{
345 bprm->p = PAGE_SIZE * MAX_ARG_PAGES - sizeof(void *);
346 return 0;
347}
348
349static bool valid_arg_len(struct linux_binprm *bprm, long len)
350{
351 return len <= bprm->p;
352}
353
354#endif /* CONFIG_MMU */
355
356/*
357 * Create a new mm_struct and populate it with a temporary stack
358 * vm_area_struct. We don't have enough context at this point to set the stack
359 * flags, permissions, and offset, so we use temporary values. We'll update
360 * them later in setup_arg_pages().
361 */
Yuanhan Liu9cc64ce2013-02-20 13:16:01 +1100362static int bprm_mm_init(struct linux_binprm *bprm)
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700363{
364 int err;
365 struct mm_struct *mm = NULL;
366
367 bprm->mm = mm = mm_alloc();
368 err = -ENOMEM;
369 if (!mm)
370 goto err;
371
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700372 err = __bprm_mm_init(bprm);
373 if (err)
374 goto err;
375
376 return 0;
377
378err:
379 if (mm) {
380 bprm->mm = NULL;
381 mmdrop(mm);
382 }
383
384 return err;
385}
386
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100387struct user_arg_ptr {
Oleg Nesterov0e028462011-03-06 18:02:54 +0100388#ifdef CONFIG_COMPAT
389 bool is_compat;
390#endif
391 union {
392 const char __user *const __user *native;
393#ifdef CONFIG_COMPAT
Al Viro38b983b2012-09-30 13:38:55 -0400394 const compat_uptr_t __user *compat;
Oleg Nesterov0e028462011-03-06 18:02:54 +0100395#endif
396 } ptr;
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100397};
398
399static const char __user *get_user_arg_ptr(struct user_arg_ptr argv, int nr)
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100400{
Oleg Nesterov0e028462011-03-06 18:02:54 +0100401 const char __user *native;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100402
Oleg Nesterov0e028462011-03-06 18:02:54 +0100403#ifdef CONFIG_COMPAT
404 if (unlikely(argv.is_compat)) {
405 compat_uptr_t compat;
406
407 if (get_user(compat, argv.ptr.compat + nr))
408 return ERR_PTR(-EFAULT);
409
410 return compat_ptr(compat);
411 }
412#endif
413
414 if (get_user(native, argv.ptr.native + nr))
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100415 return ERR_PTR(-EFAULT);
416
Oleg Nesterov0e028462011-03-06 18:02:54 +0100417 return native;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100418}
419
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420/*
421 * count() counts the number of strings in array ARGV.
422 */
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100423static int count(struct user_arg_ptr argv, int max)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424{
425 int i = 0;
426
Oleg Nesterov0e028462011-03-06 18:02:54 +0100427 if (argv.ptr.native != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700428 for (;;) {
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100429 const char __user *p = get_user_arg_ptr(argv, i);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430
Linus Torvalds1da177e2005-04-16 15:20:36 -0700431 if (!p)
432 break;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100433
434 if (IS_ERR(p))
435 return -EFAULT;
436
Xi Wang6d92d4f2013-01-11 14:31:48 -0800437 if (i >= max)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700438 return -E2BIG;
Xi Wang6d92d4f2013-01-11 14:31:48 -0800439 ++i;
Roland McGrath9aea5a62010-09-07 19:37:06 -0700440
441 if (fatal_signal_pending(current))
442 return -ERESTARTNOHAND;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443 cond_resched();
444 }
445 }
446 return i;
447}
448
449/*
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700450 * 'copy_strings()' copies argument/environment strings from the old
451 * processes's memory to the new process's stack. The call to get_user_pages()
452 * ensures the destination page is created and not swapped out.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453 */
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100454static int copy_strings(int argc, struct user_arg_ptr argv,
Adrian Bunk75c96f82005-05-05 16:16:09 -0700455 struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456{
457 struct page *kmapped_page = NULL;
458 char *kaddr = NULL;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700459 unsigned long kpos = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460 int ret;
461
462 while (argc-- > 0) {
David Howellsd7627462010-08-17 23:52:56 +0100463 const char __user *str;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464 int len;
465 unsigned long pos;
466
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100467 ret = -EFAULT;
468 str = get_user_arg_ptr(argv, argc);
469 if (IS_ERR(str))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100472 len = strnlen_user(str, MAX_ARG_STRLEN);
473 if (!len)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474 goto out;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100475
476 ret = -E2BIG;
477 if (!valid_arg_len(bprm, len))
478 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700479
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700480 /* We're going to work our way backwords. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481 pos = bprm->p;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700482 str += len;
483 bprm->p -= len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484
485 while (len > 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486 int offset, bytes_to_copy;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487
Roland McGrath9aea5a62010-09-07 19:37:06 -0700488 if (fatal_signal_pending(current)) {
489 ret = -ERESTARTNOHAND;
490 goto out;
491 }
Roland McGrath7993bc12010-09-07 19:36:28 -0700492 cond_resched();
493
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494 offset = pos % PAGE_SIZE;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700495 if (offset == 0)
496 offset = PAGE_SIZE;
497
498 bytes_to_copy = offset;
499 if (bytes_to_copy > len)
500 bytes_to_copy = len;
501
502 offset -= bytes_to_copy;
503 pos -= bytes_to_copy;
504 str -= bytes_to_copy;
505 len -= bytes_to_copy;
506
507 if (!kmapped_page || kpos != (pos & PAGE_MASK)) {
508 struct page *page;
509
510 page = get_arg_page(bprm, pos, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511 if (!page) {
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700512 ret = -E2BIG;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700513 goto out;
514 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700516 if (kmapped_page) {
517 flush_kernel_dcache_page(kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518 kunmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700519 put_arg_page(kmapped_page);
520 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521 kmapped_page = page;
522 kaddr = kmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700523 kpos = pos & PAGE_MASK;
524 flush_arg_page(bprm, kpos, kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525 }
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700526 if (copy_from_user(kaddr+offset, str, bytes_to_copy)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527 ret = -EFAULT;
528 goto out;
529 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700530 }
531 }
532 ret = 0;
533out:
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700534 if (kmapped_page) {
535 flush_kernel_dcache_page(kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536 kunmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700537 put_arg_page(kmapped_page);
538 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539 return ret;
540}
541
542/*
543 * Like copy_strings, but get argv and its values from kernel memory.
544 */
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100545int copy_strings_kernel(int argc, const char *const *__argv,
David Howellsd7627462010-08-17 23:52:56 +0100546 struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547{
548 int r;
549 mm_segment_t oldfs = get_fs();
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100550 struct user_arg_ptr argv = {
Oleg Nesterov0e028462011-03-06 18:02:54 +0100551 .ptr.native = (const char __user *const __user *)__argv,
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100552 };
553
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554 set_fs(KERNEL_DS);
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100555 r = copy_strings(argc, argv, bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700556 set_fs(oldfs);
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100557
Linus Torvalds1da177e2005-04-16 15:20:36 -0700558 return r;
559}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560EXPORT_SYMBOL(copy_strings_kernel);
561
562#ifdef CONFIG_MMU
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700563
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564/*
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700565 * During bprm_mm_init(), we create a temporary stack at STACK_TOP_MAX. Once
566 * the binfmt code determines where the new stack should reside, we shift it to
567 * its final location. The process proceeds as follows:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568 *
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700569 * 1) Use shift to calculate the new vma endpoints.
570 * 2) Extend vma to cover both the old and new ranges. This ensures the
571 * arguments passed to subsequent functions are consistent.
572 * 3) Move vma's page tables to the new range.
573 * 4) Free up any cleared pgd range.
574 * 5) Shrink the vma to cover only the new range.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700575 */
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700576static int shift_arg_pages(struct vm_area_struct *vma, unsigned long shift)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700577{
578 struct mm_struct *mm = vma->vm_mm;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700579 unsigned long old_start = vma->vm_start;
580 unsigned long old_end = vma->vm_end;
581 unsigned long length = old_end - old_start;
582 unsigned long new_start = old_start - shift;
583 unsigned long new_end = old_end - shift;
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700584 struct mmu_gather tlb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700585
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700586 BUG_ON(new_start > new_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700587
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700588 /*
589 * ensure there are no vmas between where we want to go
590 * and where we are
591 */
592 if (vma != find_vma(mm, new_start))
593 return -EFAULT;
594
595 /*
596 * cover the whole range: [new_start, old_end)
597 */
Rik van Riel5beb4932010-03-05 13:42:07 -0800598 if (vma_adjust(vma, new_start, old_end, vma->vm_pgoff, NULL))
599 return -ENOMEM;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700600
601 /*
602 * move the page tables downwards, on failure we rely on
603 * process cleanup to remove whatever mess we made.
604 */
605 if (length != move_page_tables(vma, old_start,
Michel Lespinasse38a76012012-10-08 16:31:50 -0700606 vma, new_start, length, false))
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700607 return -ENOMEM;
608
609 lru_add_drain();
Linus Torvalds2b047252013-08-15 11:42:25 -0700610 tlb_gather_mmu(&tlb, mm, old_start, old_end);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700611 if (new_end > old_start) {
612 /*
613 * when the old and new regions overlap clear from new_end.
614 */
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700615 free_pgd_range(&tlb, new_end, old_end, new_end,
Hugh Dickins6ee86302013-04-29 15:07:44 -0700616 vma->vm_next ? vma->vm_next->vm_start : USER_PGTABLES_CEILING);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700617 } else {
618 /*
619 * otherwise, clean from old_start; this is done to not touch
620 * the address space in [new_end, old_start) some architectures
621 * have constraints on va-space that make this illegal (IA64) -
622 * for the others its just a little faster.
623 */
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700624 free_pgd_range(&tlb, old_start, old_end, new_end,
Hugh Dickins6ee86302013-04-29 15:07:44 -0700625 vma->vm_next ? vma->vm_next->vm_start : USER_PGTABLES_CEILING);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626 }
Linus Torvalds2b047252013-08-15 11:42:25 -0700627 tlb_finish_mmu(&tlb, old_start, old_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700628
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700629 /*
Rik van Riel5beb4932010-03-05 13:42:07 -0800630 * Shrink the vma to just the new range. Always succeeds.
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700631 */
632 vma_adjust(vma, new_start, new_end, vma->vm_pgoff, NULL);
633
634 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700635}
636
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700637/*
638 * Finalizes the stack vm_area_struct. The flags and permissions are updated,
639 * the stack is optionally relocated, and some extra space is added.
640 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700641int setup_arg_pages(struct linux_binprm *bprm,
642 unsigned long stack_top,
643 int executable_stack)
644{
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700645 unsigned long ret;
646 unsigned long stack_shift;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700647 struct mm_struct *mm = current->mm;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700648 struct vm_area_struct *vma = bprm->vma;
649 struct vm_area_struct *prev = NULL;
650 unsigned long vm_flags;
651 unsigned long stack_base;
Michael Neuling803bf5e2010-02-10 13:56:42 -0800652 unsigned long stack_size;
653 unsigned long stack_expand;
654 unsigned long rlim_stack;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655
656#ifdef CONFIG_STACK_GROWSUP
James Hogand71f2902014-05-13 23:58:24 +0100657 /* Limit stack size */
Jiri Slabyd554ed892010-03-05 13:42:42 -0800658 stack_base = rlimit_max(RLIMIT_STACK);
James Hogand71f2902014-05-13 23:58:24 +0100659 if (stack_base > STACK_SIZE_MAX)
660 stack_base = STACK_SIZE_MAX;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661
Helge Dellerd045c772015-05-11 22:01:27 +0200662 /* Add space for stack randomization. */
663 stack_base += (STACK_RND_MASK << PAGE_SHIFT);
664
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
David Drysdale51f39a12014-12-12 16:57:29 -0800754static struct file *do_open_execat(int fd, struct filename *name, int flags)
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;
David Drysdale51f39a12014-12-12 16:57:29 -0800758 struct open_flags open_exec_flags = {
Al Viro47c805d2011-02-23 17:44:09 -0500759 .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
David Drysdale51f39a12014-12-12 16:57:29 -0800765 if ((flags & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
766 return ERR_PTR(-EINVAL);
767 if (flags & AT_SYMLINK_NOFOLLOW)
768 open_exec_flags.lookup_flags &= ~LOOKUP_FOLLOW;
769 if (flags & AT_EMPTY_PATH)
770 open_exec_flags.lookup_flags |= LOOKUP_EMPTY;
771
772 file = do_filp_open(fd, name, &open_exec_flags);
Al Viro6e8341a2009-04-06 11:16:22 -0400773 if (IS_ERR(file))
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200774 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200776 err = -EACCES;
Al Viro496ad9a2013-01-23 17:07:38 -0500777 if (!S_ISREG(file_inode(file)->i_mode))
Al Viro6e8341a2009-04-06 11:16:22 -0400778 goto exit;
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200779
Al Viro6e8341a2009-04-06 11:16:22 -0400780 if (file->f_path.mnt->mnt_flags & MNT_NOEXEC)
781 goto exit;
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200782
783 err = deny_write_access(file);
Al Viro6e8341a2009-04-06 11:16:22 -0400784 if (err)
785 goto exit;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700786
David Drysdale51f39a12014-12-12 16:57:29 -0800787 if (name->name[0] != '\0')
788 fsnotify_open(file);
789
Al Viro6e8341a2009-04-06 11:16:22 -0400790out:
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200791 return file;
792
Al Viro6e8341a2009-04-06 11:16:22 -0400793exit:
794 fput(file);
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200795 return ERR_PTR(err);
796}
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -0800797
798struct file *open_exec(const char *name)
799{
Paul Moore51689102015-01-22 00:00:03 -0500800 struct filename *filename = getname_kernel(name);
801 struct file *f = ERR_CAST(filename);
802
803 if (!IS_ERR(filename)) {
804 f = do_open_execat(AT_FDCWD, filename, 0);
805 putname(filename);
806 }
807 return f;
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -0800808}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700809EXPORT_SYMBOL(open_exec);
810
Mimi Zohar6777d772009-08-21 14:32:48 -0400811int kernel_read(struct file *file, loff_t offset,
812 char *addr, unsigned long count)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813{
814 mm_segment_t old_fs;
815 loff_t pos = offset;
816 int result;
817
818 old_fs = get_fs();
819 set_fs(get_ds());
820 /* The cast to a user pointer is valid due to the set_fs() */
821 result = vfs_read(file, (void __user *)addr, count, &pos);
822 set_fs(old_fs);
823 return result;
824}
825
826EXPORT_SYMBOL(kernel_read);
827
Al Viro3dc20cb2013-04-13 20:31:37 -0400828ssize_t read_code(struct file *file, unsigned long addr, loff_t pos, size_t len)
829{
Al Viroec695572014-02-04 19:08:21 -0500830 ssize_t res = vfs_read(file, (void __user *)addr, len, &pos);
Al Viro3dc20cb2013-04-13 20:31:37 -0400831 if (res > 0)
832 flush_icache_range(addr, addr + len);
833 return res;
834}
835EXPORT_SYMBOL(read_code);
836
Linus Torvalds1da177e2005-04-16 15:20:36 -0700837static int exec_mmap(struct mm_struct *mm)
838{
839 struct task_struct *tsk;
Davidlohr Bueso615d6e82014-04-07 15:37:25 -0700840 struct mm_struct *old_mm, *active_mm;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841
842 /* Notify parent that we're no longer interested in the old VM */
843 tsk = current;
844 old_mm = current->mm;
845 mm_release(tsk, old_mm);
846
847 if (old_mm) {
Konstantin Khlebnikov4fe7efd2012-06-20 12:53:01 -0700848 sync_mm_rss(old_mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849 /*
850 * Make sure that if there is a core dump in progress
851 * for the old mm, we get out and die instead of going
852 * through with the exec. We must hold mmap_sem around
Oleg Nesterov999d9fc2008-07-25 01:47:41 -0700853 * checking core_state and changing tsk->mm.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700854 */
855 down_read(&old_mm->mmap_sem);
Oleg Nesterov999d9fc2008-07-25 01:47:41 -0700856 if (unlikely(old_mm->core_state)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700857 up_read(&old_mm->mmap_sem);
858 return -EINTR;
859 }
860 }
861 task_lock(tsk);
862 active_mm = tsk->active_mm;
863 tsk->mm = mm;
864 tsk->active_mm = mm;
865 activate_mm(active_mm, mm);
Davidlohr Bueso615d6e82014-04-07 15:37:25 -0700866 tsk->mm->vmacache_seqnum = 0;
867 vmacache_flush(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868 task_unlock(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869 if (old_mm) {
870 up_read(&old_mm->mmap_sem);
Eric Sesterhenn7dddb122006-04-01 01:13:38 +0200871 BUG_ON(active_mm != old_mm);
Oleg Nesterov701085b2012-03-19 17:04:01 +0100872 setmax_mm_hiwater_rss(&tsk->signal->maxrss, old_mm);
Balbir Singh31a78f22008-09-28 23:09:31 +0100873 mm_update_next_owner(old_mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874 mmput(old_mm);
875 return 0;
876 }
877 mmdrop(active_mm);
878 return 0;
879}
880
881/*
882 * This function makes sure the current process has its own signal table,
883 * so that flush_signal_handlers can later reset the handlers without
884 * disturbing other processes. (Other processes might share the signal
885 * table via the CLONE_SIGHAND option to clone().)
886 */
Arjan van de Ven858119e2006-01-14 13:20:43 -0800887static int de_thread(struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888{
889 struct signal_struct *sig = tsk->signal;
Oleg Nesterovb2c903b2007-10-16 23:27:22 -0700890 struct sighand_struct *oldsighand = tsk->sighand;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891 spinlock_t *lock = &oldsighand->siglock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700892
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700893 if (thread_group_empty(tsk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700894 goto no_thread_group;
895
896 /*
897 * Kill all other threads in the thread group.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700899 spin_lock_irq(lock);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -0800900 if (signal_group_exit(sig)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700901 /*
902 * Another group action in progress, just
903 * return so that the signal is processed.
904 */
905 spin_unlock_irq(lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700906 return -EAGAIN;
907 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700908
Oleg Nesterovd3441932010-05-26 14:43:11 -0700909 sig->group_exit_task = tsk;
910 sig->notify_count = zap_other_threads(tsk);
911 if (!thread_group_leader(tsk))
912 sig->notify_count--;
913
914 while (sig->notify_count) {
Oleg Nesterovd5bbd432012-10-08 19:13:01 +0200915 __set_current_state(TASK_KILLABLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916 spin_unlock_irq(lock);
917 schedule();
Oleg Nesterovd5bbd432012-10-08 19:13:01 +0200918 if (unlikely(__fatal_signal_pending(tsk)))
919 goto killed;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700920 spin_lock_irq(lock);
921 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922 spin_unlock_irq(lock);
923
924 /*
925 * At this point all other threads have exited, all we have to
926 * do is to wait for the thread group leader to become inactive,
927 * and to assume its PID:
928 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700929 if (!thread_group_leader(tsk)) {
Oleg Nesterov81879262008-12-01 14:18:16 -0800930 struct task_struct *leader = tsk->group_leader;
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700931
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700932 for (;;) {
Oleg Nesterove56fb282013-04-30 15:28:20 -0700933 threadgroup_change_begin(tsk);
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700934 write_lock_irq(&tasklist_lock);
Kirill Tkhaidfcce792015-04-16 12:48:01 -0700935 /*
936 * Do this under tasklist_lock to ensure that
937 * exit_notify() can't miss ->group_exit_task
938 */
939 sig->notify_count = -1;
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700940 if (likely(leader->exit_state))
941 break;
Oleg Nesterovd5bbd432012-10-08 19:13:01 +0200942 __set_current_state(TASK_KILLABLE);
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700943 write_unlock_irq(&tasklist_lock);
Oleg Nesterove56fb282013-04-30 15:28:20 -0700944 threadgroup_change_end(tsk);
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700945 schedule();
Oleg Nesterovd5bbd432012-10-08 19:13:01 +0200946 if (unlikely(__fatal_signal_pending(tsk)))
947 goto killed;
Oleg Nesterov6db840f2007-10-16 23:27:23 -0700948 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700949
Roland McGrathf5e90282006-04-10 22:54:16 -0700950 /*
951 * The only record we have of the real-time age of a
952 * process, regardless of execs it's done, is start_time.
953 * All the past CPU time is accumulated in signal_struct
954 * from sister threads now dead. But in this non-leader
955 * exec, nothing survives from the original leader thread,
956 * whose birth marks the true age of this process now.
957 * When we take on its identity by switching to its PID, we
958 * also take its birthdate (always earlier than our own).
959 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700960 tsk->start_time = leader->start_time;
Oleg Nesterov266b7a02013-07-03 15:08:35 -0700961 tsk->real_start_time = leader->real_start_time;
Roland McGrathf5e90282006-04-10 22:54:16 -0700962
Pavel Emelyanovbac0abd2007-10-18 23:40:18 -0700963 BUG_ON(!same_thread_group(leader, tsk));
964 BUG_ON(has_group_leader_pid(tsk));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965 /*
966 * An exec() starts a new thread group with the
967 * TGID of the previous thread group. Rehash the
968 * two threads with a switched PID, and release
969 * the former thread group leader:
970 */
Eric W. Biedermand73d6522006-03-28 16:11:03 -0800971
972 /* Become a process group leader with the old leader's pid.
Eric W. Biedermanc18258c2006-09-27 01:51:06 -0700973 * The old leader becomes a thread of the this thread group.
974 * Note: The old leader also uses this pid until release_task
Eric W. Biedermand73d6522006-03-28 16:11:03 -0800975 * is called. Odd but simple and correct.
976 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700977 tsk->pid = leader->pid;
Oleg Nesterov3f418542013-07-03 15:08:25 -0700978 change_pid(tsk, PIDTYPE_PID, task_pid(leader));
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700979 transfer_pid(leader, tsk, PIDTYPE_PGID);
980 transfer_pid(leader, tsk, PIDTYPE_SID);
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -0800981
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700982 list_replace_rcu(&leader->tasks, &tsk->tasks);
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -0800983 list_replace_init(&leader->sibling, &tsk->sibling);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700984
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700985 tsk->group_leader = tsk;
986 leader->group_leader = tsk;
Eric W. Biedermande12a782006-04-10 17:16:49 -0600987
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700988 tsk->exit_signal = SIGCHLD;
Oleg Nesterov087806b2011-06-22 23:10:26 +0200989 leader->exit_signal = -1;
Oleg Nesterov962b5642005-11-23 13:37:43 -0800990
991 BUG_ON(leader->exit_state != EXIT_ZOMBIE);
992 leader->exit_state = EXIT_DEAD;
Oleg Nesteroveac1b5e2011-07-21 20:00:43 +0200993
994 /*
995 * We are going to release_task()->ptrace_unlink() silently,
996 * the tracer can sleep in do_wait(). EXIT_DEAD guarantees
997 * the tracer wont't block again waiting for this thread.
998 */
999 if (unlikely(leader->ptrace))
1000 __wake_up_parent(leader, leader->parent);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001 write_unlock_irq(&tasklist_lock);
Oleg Nesterove56fb282013-04-30 15:28:20 -07001002 threadgroup_change_end(tsk);
Oleg Nesterov81879262008-12-01 14:18:16 -08001003
1004 release_task(leader);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -08001005 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001006
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001007 sig->group_exit_task = NULL;
1008 sig->notify_count = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001009
1010no_thread_group:
Oleg Nesterove6368252012-03-19 17:03:22 +01001011 /* we have changed execution domain */
1012 tsk->exit_signal = SIGCHLD;
1013
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014 exit_itimers(sig);
Oleg Nesterovcbaffba2008-05-26 20:55:42 +04001015 flush_itimer_signals();
Oleg Nesterov329f7db2005-11-07 21:12:43 +03001016
Oleg Nesterovb2c903b2007-10-16 23:27:22 -07001017 if (atomic_read(&oldsighand->count) != 1) {
1018 struct sighand_struct *newsighand;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019 /*
Oleg Nesterovb2c903b2007-10-16 23:27:22 -07001020 * This ->sighand is shared with the CLONE_SIGHAND
1021 * but not CLONE_THREAD task, switch to the new one.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022 */
Oleg Nesterovb2c903b2007-10-16 23:27:22 -07001023 newsighand = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
1024 if (!newsighand)
1025 return -ENOMEM;
1026
Linus Torvalds1da177e2005-04-16 15:20:36 -07001027 atomic_set(&newsighand->count, 1);
1028 memcpy(newsighand->action, oldsighand->action,
1029 sizeof(newsighand->action));
1030
1031 write_lock_irq(&tasklist_lock);
1032 spin_lock(&oldsighand->siglock);
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001033 rcu_assign_pointer(tsk->sighand, newsighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001034 spin_unlock(&oldsighand->siglock);
1035 write_unlock_irq(&tasklist_lock);
1036
Davide Libenzifba2afa2007-05-10 22:23:13 -07001037 __cleanup_sighand(oldsighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001038 }
1039
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001040 BUG_ON(!thread_group_leader(tsk));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001041 return 0;
Oleg Nesterovd5bbd432012-10-08 19:13:01 +02001042
1043killed:
1044 /* protects against exit_notify() and __exit_signal() */
1045 read_lock(&tasklist_lock);
1046 sig->group_exit_task = NULL;
1047 sig->notify_count = 0;
1048 read_unlock(&tasklist_lock);
1049 return -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001050}
Oleg Nesterov0840a902007-10-16 23:27:22 -07001051
Andrew Morton59714d62008-02-04 22:27:21 -08001052char *get_task_comm(char *buf, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001053{
1054 /* buf must be at least sizeof(tsk->comm) in size */
1055 task_lock(tsk);
1056 strncpy(buf, tsk->comm, sizeof(tsk->comm));
1057 task_unlock(tsk);
Andrew Morton59714d62008-02-04 22:27:21 -08001058 return buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001059}
J Freyensee7d74f492011-05-06 16:56:47 -07001060EXPORT_SYMBOL_GPL(get_task_comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061
Al Viro6a6d27d2012-08-21 09:56:33 -04001062/*
1063 * These functions flushes out all traces of the currently running executable
1064 * so that a new one can be started
1065 */
1066
Adrian Hunter82b89772014-05-28 11:45:04 +03001067void __set_task_comm(struct task_struct *tsk, const char *buf, bool exec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001068{
1069 task_lock(tsk);
KAMEZAWA Hiroyuki43d2b112012-01-10 15:08:09 -08001070 trace_task_rename(tsk, buf);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071 strlcpy(tsk->comm, buf, sizeof(tsk->comm));
1072 task_unlock(tsk);
Adrian Hunter82b89772014-05-28 11:45:04 +03001073 perf_event_comm(tsk, exec);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001074}
1075
1076int flush_old_exec(struct linux_binprm * bprm)
1077{
Linus Torvalds221af7f2010-01-28 22:14:42 -08001078 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001079
1080 /*
1081 * Make sure we have a private signal table and that
1082 * we are unassociated from the previous thread group.
1083 */
1084 retval = de_thread(current);
1085 if (retval)
1086 goto out;
1087
Davidlohr Bueso6e399cd2015-04-16 12:47:59 -07001088 /*
1089 * Must be called _before_ exec_mmap() as bprm->mm is
1090 * not visibile until then. This also enables the update
1091 * to be lockless.
1092 */
Matt Helsley925d1c42008-04-29 01:01:36 -07001093 set_mm_exe_file(bprm->mm, bprm->file);
Davidlohr Bueso6e399cd2015-04-16 12:47:59 -07001094
Linus Torvalds1da177e2005-04-16 15:20:36 -07001095 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096 * Release all of the old mmap stuff
1097 */
Oleg Nesterov3c77f842010-11-30 20:55:34 +01001098 acct_arg_size(bprm, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001099 retval = exec_mmap(bprm->mm);
1100 if (retval)
Al Virofd8328b2008-04-22 05:11:59 -04001101 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001102
1103 bprm->mm = NULL; /* We're using it now */
Linus Torvalds7ab02af2010-02-02 12:37:44 -08001104
Mathias Krausedac853a2011-06-09 20:05:18 +02001105 set_fs(USER_DS);
Zhang Yib88fae62014-01-23 15:55:57 -08001106 current->flags &= ~(PF_RANDOMIZE | PF_FORKNOEXEC | PF_KTHREAD |
1107 PF_NOFREEZE | PF_NO_SETAFFINITY);
Linus Torvalds7ab02af2010-02-02 12:37:44 -08001108 flush_thread();
1109 current->personality &= ~bprm->per_clear;
1110
Linus Torvalds221af7f2010-01-28 22:14:42 -08001111 return 0;
1112
1113out:
1114 return retval;
1115}
1116EXPORT_SYMBOL(flush_old_exec);
1117
Al Viro1b5d7832011-06-19 12:49:47 -04001118void would_dump(struct linux_binprm *bprm, struct file *file)
1119{
Al Viro496ad9a2013-01-23 17:07:38 -05001120 if (inode_permission(file_inode(file), MAY_READ) < 0)
Al Viro1b5d7832011-06-19 12:49:47 -04001121 bprm->interp_flags |= BINPRM_FLAGS_ENFORCE_NONDUMP;
1122}
1123EXPORT_SYMBOL(would_dump);
1124
Linus Torvalds221af7f2010-01-28 22:14:42 -08001125void setup_new_exec(struct linux_binprm * bprm)
1126{
Linus Torvalds221af7f2010-01-28 22:14:42 -08001127 arch_pick_mmap_layout(current->mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001128
1129 /* This is the point of no return */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001130 current->sas_ss_sp = current->sas_ss_size = 0;
1131
Eric W. Biederman8e96e3b2012-03-03 21:17:15 -08001132 if (uid_eq(current_euid(), current_uid()) && gid_eq(current_egid(), current_gid()))
Kees Cooke579d2c2013-02-27 17:03:15 -08001133 set_dumpable(current->mm, SUID_DUMP_USER);
Alan Coxd6e71142005-06-23 00:09:43 -07001134 else
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001135 set_dumpable(current->mm, suid_dumpable);
Alan Coxd6e71142005-06-23 00:09:43 -07001136
Peter Zijlstrae041e322014-05-21 17:32:19 +02001137 perf_event_exec();
Adrian Hunter82b89772014-05-28 11:45:04 +03001138 __set_task_comm(current, kbasename(bprm->filename), true);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001139
Benjamin Herrenschmidt0551fbd2006-02-28 16:59:19 -08001140 /* Set the new mm task size. We have to do that late because it may
1141 * depend on TIF_32BIT which is only updated in flush_thread() on
1142 * some architectures like powerpc
1143 */
1144 current->mm->task_size = TASK_SIZE;
1145
David Howellsa6f76f22008-11-14 10:39:24 +11001146 /* install the new credentials */
Eric W. Biederman8e96e3b2012-03-03 21:17:15 -08001147 if (!uid_eq(bprm->cred->uid, current_euid()) ||
1148 !gid_eq(bprm->cred->gid, current_egid())) {
Marcel Holtmannd2d56c52007-08-17 21:47:58 +02001149 current->pdeath_signal = 0;
Al Viro1b5d7832011-06-19 12:49:47 -04001150 } else {
1151 would_dump(bprm, bprm->file);
1152 if (bprm->interp_flags & BINPRM_FLAGS_ENFORCE_NONDUMP)
1153 set_dumpable(current->mm, suid_dumpable);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001154 }
1155
1156 /* An exec changes our domain. We are no longer part of the thread
1157 group */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001158 current->self_exec_id++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001159 flush_signal_handlers(current, 0);
Al Viro6a6d27d2012-08-21 09:56:33 -04001160 do_close_on_exec(current->files);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001161}
Linus Torvalds221af7f2010-01-28 22:14:42 -08001162EXPORT_SYMBOL(setup_new_exec);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001163
David Howellsa6f76f22008-11-14 10:39:24 +11001164/*
Oleg Nesterova2a84742009-09-05 11:17:13 -07001165 * Prepare credentials and lock ->cred_guard_mutex.
1166 * install_exec_creds() commits the new creds and drops the lock.
1167 * Or, if exec fails before, free_bprm() should release ->cred and
1168 * and unlock.
1169 */
1170int prepare_bprm_creds(struct linux_binprm *bprm)
1171{
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001172 if (mutex_lock_interruptible(&current->signal->cred_guard_mutex))
Oleg Nesterova2a84742009-09-05 11:17:13 -07001173 return -ERESTARTNOINTR;
1174
1175 bprm->cred = prepare_exec_creds();
1176 if (likely(bprm->cred))
1177 return 0;
1178
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001179 mutex_unlock(&current->signal->cred_guard_mutex);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001180 return -ENOMEM;
1181}
1182
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001183static void free_bprm(struct linux_binprm *bprm)
Oleg Nesterova2a84742009-09-05 11:17:13 -07001184{
1185 free_arg_pages(bprm);
1186 if (bprm->cred) {
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001187 mutex_unlock(&current->signal->cred_guard_mutex);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001188 abort_creds(bprm->cred);
1189 }
Oleg Nesterov63e46b92014-01-23 15:55:51 -08001190 if (bprm->file) {
1191 allow_write_access(bprm->file);
1192 fput(bprm->file);
1193 }
Kees Cookb66c5982012-12-20 15:05:16 -08001194 /* If a binfmt changed the interp, free it. */
1195 if (bprm->interp != bprm->filename)
1196 kfree(bprm->interp);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001197 kfree(bprm);
1198}
1199
Kees Cookb66c5982012-12-20 15:05:16 -08001200int bprm_change_interp(char *interp, struct linux_binprm *bprm)
1201{
1202 /* If a binfmt changed the interp, free it first. */
1203 if (bprm->interp != bprm->filename)
1204 kfree(bprm->interp);
1205 bprm->interp = kstrdup(interp, GFP_KERNEL);
1206 if (!bprm->interp)
1207 return -ENOMEM;
1208 return 0;
1209}
1210EXPORT_SYMBOL(bprm_change_interp);
1211
Oleg Nesterova2a84742009-09-05 11:17:13 -07001212/*
David Howellsa6f76f22008-11-14 10:39:24 +11001213 * install the new credentials for this executable
1214 */
1215void install_exec_creds(struct linux_binprm *bprm)
1216{
1217 security_bprm_committing_creds(bprm);
1218
1219 commit_creds(bprm->cred);
1220 bprm->cred = NULL;
Stephane Eranian2976b102013-06-20 11:36:28 +02001221
1222 /*
1223 * Disable monitoring for regular users
1224 * when executing setuid binaries. Must
1225 * wait until new credentials are committed
1226 * by commit_creds() above
1227 */
1228 if (get_dumpable(current->mm) != SUID_DUMP_USER)
1229 perf_event_exit_task(current);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001230 /*
1231 * cred_guard_mutex must be held at least to this point to prevent
David Howellsa6f76f22008-11-14 10:39:24 +11001232 * ptrace_attach() from altering our determination of the task's
Oleg Nesterova2a84742009-09-05 11:17:13 -07001233 * credentials; any time after this it may be unlocked.
1234 */
David Howellsa6f76f22008-11-14 10:39:24 +11001235 security_bprm_committed_creds(bprm);
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001236 mutex_unlock(&current->signal->cred_guard_mutex);
David Howellsa6f76f22008-11-14 10:39:24 +11001237}
1238EXPORT_SYMBOL(install_exec_creds);
1239
1240/*
1241 * determine how safe it is to execute the proposed program
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001242 * - the caller must hold ->cred_guard_mutex to protect against
Kees Cookc2e1f2e2014-06-05 00:23:17 -07001243 * PTRACE_ATTACH or seccomp thread-sync
David Howellsa6f76f22008-11-14 10:39:24 +11001244 */
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001245static void check_unsafe_exec(struct linux_binprm *bprm)
David Howellsa6f76f22008-11-14 10:39:24 +11001246{
David Howells0bf2f3a2009-02-06 11:45:46 +00001247 struct task_struct *p = current, *t;
Al Virof1191b52009-03-30 07:35:18 -04001248 unsigned n_fs;
David Howellsa6f76f22008-11-14 10:39:24 +11001249
Tejun Heo4b9d33e2011-06-17 16:50:38 +02001250 if (p->ptrace) {
1251 if (p->ptrace & PT_PTRACE_CAP)
1252 bprm->unsafe |= LSM_UNSAFE_PTRACE_CAP;
1253 else
1254 bprm->unsafe |= LSM_UNSAFE_PTRACE;
1255 }
David Howellsa6f76f22008-11-14 10:39:24 +11001256
Andy Lutomirski259e5e62012-04-12 16:47:50 -05001257 /*
1258 * This isn't strictly necessary, but it makes it harder for LSMs to
1259 * mess up.
1260 */
Kees Cook1d4457f2014-05-21 15:23:46 -07001261 if (task_no_new_privs(current))
Andy Lutomirski259e5e62012-04-12 16:47:50 -05001262 bprm->unsafe |= LSM_UNSAFE_NO_NEW_PRIVS;
1263
Oleg Nesterov83f62a22014-01-23 15:55:49 -08001264 t = p;
David Howells0bf2f3a2009-02-06 11:45:46 +00001265 n_fs = 1;
Nick Piggin2a4419b2010-08-18 04:37:33 +10001266 spin_lock(&p->fs->lock);
Oleg Nesterov437f7fd2009-04-24 01:02:45 +02001267 rcu_read_lock();
Oleg Nesterov83f62a22014-01-23 15:55:49 -08001268 while_each_thread(p, t) {
David Howells0bf2f3a2009-02-06 11:45:46 +00001269 if (t->fs == p->fs)
1270 n_fs++;
David Howells0bf2f3a2009-02-06 11:45:46 +00001271 }
Oleg Nesterov437f7fd2009-04-24 01:02:45 +02001272 rcu_read_unlock();
David Howells0bf2f3a2009-02-06 11:45:46 +00001273
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001274 if (p->fs->users > n_fs)
David Howellsa6f76f22008-11-14 10:39:24 +11001275 bprm->unsafe |= LSM_UNSAFE_SHARE;
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001276 else
1277 p->fs->in_exec = 1;
Nick Piggin2a4419b2010-08-18 04:37:33 +10001278 spin_unlock(&p->fs->lock);
David Howellsa6f76f22008-11-14 10:39:24 +11001279}
1280
Jann Horn8b01fc82015-04-19 02:48:39 +02001281static void bprm_fill_uid(struct linux_binprm *bprm)
1282{
1283 struct inode *inode;
1284 unsigned int mode;
1285 kuid_t uid;
1286 kgid_t gid;
1287
1288 /* clear any previous set[ug]id data from a previous binary */
1289 bprm->cred->euid = current_euid();
1290 bprm->cred->egid = current_egid();
1291
1292 if (bprm->file->f_path.mnt->mnt_flags & MNT_NOSUID)
1293 return;
1294
1295 if (task_no_new_privs(current))
1296 return;
1297
1298 inode = file_inode(bprm->file);
1299 mode = READ_ONCE(inode->i_mode);
1300 if (!(mode & (S_ISUID|S_ISGID)))
1301 return;
1302
1303 /* Be careful if suid/sgid is set */
1304 mutex_lock(&inode->i_mutex);
1305
1306 /* reload atomically mode/uid/gid now that lock held */
1307 mode = inode->i_mode;
1308 uid = inode->i_uid;
1309 gid = inode->i_gid;
1310 mutex_unlock(&inode->i_mutex);
1311
1312 /* We ignore suid/sgid if there are no mappings for them in the ns */
1313 if (!kuid_has_mapping(bprm->cred->user_ns, uid) ||
1314 !kgid_has_mapping(bprm->cred->user_ns, gid))
1315 return;
1316
1317 if (mode & S_ISUID) {
1318 bprm->per_clear |= PER_CLEAR_ON_SETID;
1319 bprm->cred->euid = uid;
1320 }
1321
1322 if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) {
1323 bprm->per_clear |= PER_CLEAR_ON_SETID;
1324 bprm->cred->egid = gid;
1325 }
1326}
1327
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001328/*
1329 * Fill the binprm structure from the inode.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001330 * Check permissions, then read the first 128 (BINPRM_BUF_SIZE) bytes
David Howellsa6f76f22008-11-14 10:39:24 +11001331 *
1332 * This may be called multiple times for binary chains (scripts for example).
Linus Torvalds1da177e2005-04-16 15:20:36 -07001333 */
1334int prepare_binprm(struct linux_binprm *bprm)
1335{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001336 int retval;
1337
Jann Horn8b01fc82015-04-19 02:48:39 +02001338 bprm_fill_uid(bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001339
1340 /* fill in binprm security blob */
David Howellsa6f76f22008-11-14 10:39:24 +11001341 retval = security_bprm_set_creds(bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001342 if (retval)
1343 return retval;
David Howellsa6f76f22008-11-14 10:39:24 +11001344 bprm->cred_prepared = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001345
David Howellsa6f76f22008-11-14 10:39:24 +11001346 memset(bprm->buf, 0, BINPRM_BUF_SIZE);
1347 return kernel_read(bprm->file, 0, bprm->buf, BINPRM_BUF_SIZE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001348}
1349
1350EXPORT_SYMBOL(prepare_binprm);
1351
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001352/*
1353 * Arguments are '\0' separated strings found at the location bprm->p
1354 * points to; chop off the first by relocating brpm->p to right after
1355 * the first '\0' encountered.
1356 */
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001357int remove_arg_zero(struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001358{
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001359 int ret = 0;
1360 unsigned long offset;
1361 char *kaddr;
1362 struct page *page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001363
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001364 if (!bprm->argc)
1365 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001366
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001367 do {
1368 offset = bprm->p & ~PAGE_MASK;
1369 page = get_arg_page(bprm, bprm->p, 0);
1370 if (!page) {
1371 ret = -EFAULT;
1372 goto out;
1373 }
Cong Wange8e3c3d2011-11-25 23:14:27 +08001374 kaddr = kmap_atomic(page);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001375
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001376 for (; offset < PAGE_SIZE && kaddr[offset];
1377 offset++, bprm->p++)
1378 ;
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001379
Cong Wange8e3c3d2011-11-25 23:14:27 +08001380 kunmap_atomic(kaddr);
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001381 put_arg_page(page);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001382
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001383 if (offset == PAGE_SIZE)
1384 free_arg_page(bprm, (bprm->p >> PAGE_SHIFT) - 1);
1385 } while (offset == PAGE_SIZE);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001386
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001387 bprm->p++;
1388 bprm->argc--;
1389 ret = 0;
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001390
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001391out:
1392 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001393}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001394EXPORT_SYMBOL(remove_arg_zero);
1395
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001396#define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001397/*
1398 * cycle the list of binary formats handler, until one recognizes the image
1399 */
Al Viro3c456bf2012-10-20 21:53:31 -04001400int search_binary_handler(struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001401{
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001402 bool need_retry = IS_ENABLED(CONFIG_MODULES);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001403 struct linux_binfmt *fmt;
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001404 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001405
Kees Cookd7402692012-12-17 16:03:20 -08001406 /* This allows 4 levels of binfmt rewrites before failing hard. */
Oleg Nesterov131b2f92013-09-11 14:24:39 -07001407 if (bprm->recursion_depth > 5)
Kees Cookd7402692012-12-17 16:03:20 -08001408 return -ELOOP;
1409
Linus Torvalds1da177e2005-04-16 15:20:36 -07001410 retval = security_bprm_check(bprm);
1411 if (retval)
1412 return retval;
1413
Linus Torvalds1da177e2005-04-16 15:20:36 -07001414 retval = -ENOENT;
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001415 retry:
1416 read_lock(&binfmt_lock);
1417 list_for_each_entry(fmt, &formats, lh) {
1418 if (!try_module_get(fmt->module))
1419 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420 read_unlock(&binfmt_lock);
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001421 bprm->recursion_depth++;
1422 retval = fmt->load_binary(bprm);
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001423 read_lock(&binfmt_lock);
1424 put_binfmt(fmt);
Al Viro19d860a2014-05-04 20:11:36 -04001425 bprm->recursion_depth--;
1426 if (retval < 0 && !bprm->mm) {
1427 /* we got to flush_old_exec() and failed after it */
1428 read_unlock(&binfmt_lock);
1429 force_sigsegv(SIGSEGV, current);
1430 return retval;
1431 }
1432 if (retval != -ENOEXEC || !bprm->file) {
1433 read_unlock(&binfmt_lock);
1434 return retval;
1435 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001436 }
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001437 read_unlock(&binfmt_lock);
1438
Al Viro19d860a2014-05-04 20:11:36 -04001439 if (need_retry) {
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001440 if (printable(bprm->buf[0]) && printable(bprm->buf[1]) &&
1441 printable(bprm->buf[2]) && printable(bprm->buf[3]))
1442 return retval;
Oleg Nesterov4e0621a2013-09-11 14:24:45 -07001443 if (request_module("binfmt-%04x", *(ushort *)(bprm->buf + 2)) < 0)
1444 return retval;
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001445 need_retry = false;
1446 goto retry;
1447 }
1448
Linus Torvalds1da177e2005-04-16 15:20:36 -07001449 return retval;
1450}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001451EXPORT_SYMBOL(search_binary_handler);
1452
Oleg Nesterov5d1baf32013-09-11 14:24:38 -07001453static int exec_binprm(struct linux_binprm *bprm)
1454{
1455 pid_t old_pid, old_vpid;
1456 int ret;
1457
1458 /* Need to fetch pid before load_binary changes it */
1459 old_pid = current->pid;
1460 rcu_read_lock();
1461 old_vpid = task_pid_nr_ns(current, task_active_pid_ns(current->parent));
1462 rcu_read_unlock();
1463
1464 ret = search_binary_handler(bprm);
1465 if (ret >= 0) {
Linus Torvalds3eaded862013-11-21 19:18:14 -08001466 audit_bprm(bprm);
Oleg Nesterov5d1baf32013-09-11 14:24:38 -07001467 trace_sched_process_exec(current, old_pid, bprm);
1468 ptrace_event(PTRACE_EVENT_EXEC, old_vpid);
Oleg Nesterov9beb2662013-09-11 14:24:40 -07001469 proc_exec_connector(current);
Oleg Nesterov5d1baf32013-09-11 14:24:38 -07001470 }
1471
1472 return ret;
1473}
1474
Linus Torvalds1da177e2005-04-16 15:20:36 -07001475/*
1476 * sys_execve() executes a new program.
1477 */
David Drysdale51f39a12014-12-12 16:57:29 -08001478static int do_execveat_common(int fd, struct filename *filename,
1479 struct user_arg_ptr argv,
1480 struct user_arg_ptr envp,
1481 int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001482{
David Drysdale51f39a12014-12-12 16:57:29 -08001483 char *pathbuf = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001484 struct linux_binprm *bprm;
1485 struct file *file;
Al Viro3b125382008-04-22 05:31:30 -04001486 struct files_struct *displaced;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001487 int retval;
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +04001488
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001489 if (IS_ERR(filename))
1490 return PTR_ERR(filename);
1491
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +04001492 /*
1493 * We move the actual failure in case of RLIMIT_NPROC excess from
1494 * set*uid() to execve() because too many poorly written programs
1495 * don't check setuid() return code. Here we additionally recheck
1496 * whether NPROC limit is still exceeded.
1497 */
1498 if ((current->flags & PF_NPROC_EXCEEDED) &&
Oleg Nesterovbd9d43f2013-07-03 15:08:34 -07001499 atomic_read(&current_user()->processes) > rlimit(RLIMIT_NPROC)) {
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +04001500 retval = -EAGAIN;
1501 goto out_ret;
1502 }
1503
1504 /* We're below the limit (still or again), so we don't want to make
1505 * further execve() calls fail. */
1506 current->flags &= ~PF_NPROC_EXCEEDED;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001507
Al Viro3b125382008-04-22 05:31:30 -04001508 retval = unshare_files(&displaced);
Al Virofd8328b2008-04-22 05:11:59 -04001509 if (retval)
1510 goto out_ret;
1511
Linus Torvalds1da177e2005-04-16 15:20:36 -07001512 retval = -ENOMEM;
Oliver Neukum11b0b5a2006-03-25 03:08:13 -08001513 bprm = kzalloc(sizeof(*bprm), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001514 if (!bprm)
Al Virofd8328b2008-04-22 05:11:59 -04001515 goto out_files;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001516
Oleg Nesterova2a84742009-09-05 11:17:13 -07001517 retval = prepare_bprm_creds(bprm);
1518 if (retval)
David Howellsa6f76f22008-11-14 10:39:24 +11001519 goto out_free;
Al Viro498052b2009-03-30 07:20:30 -04001520
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001521 check_unsafe_exec(bprm);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001522 current->in_execve = 1;
David Howellsa6f76f22008-11-14 10:39:24 +11001523
David Drysdale51f39a12014-12-12 16:57:29 -08001524 file = do_open_execat(fd, filename, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001525 retval = PTR_ERR(file);
1526 if (IS_ERR(file))
Al Viro498052b2009-03-30 07:20:30 -04001527 goto out_unmark;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001528
1529 sched_exec();
1530
Linus Torvalds1da177e2005-04-16 15:20:36 -07001531 bprm->file = file;
David Drysdale51f39a12014-12-12 16:57:29 -08001532 if (fd == AT_FDCWD || filename->name[0] == '/') {
1533 bprm->filename = filename->name;
1534 } else {
1535 if (filename->name[0] == '\0')
1536 pathbuf = kasprintf(GFP_TEMPORARY, "/dev/fd/%d", fd);
1537 else
1538 pathbuf = kasprintf(GFP_TEMPORARY, "/dev/fd/%d/%s",
1539 fd, filename->name);
1540 if (!pathbuf) {
1541 retval = -ENOMEM;
1542 goto out_unmark;
1543 }
1544 /*
1545 * Record that a name derived from an O_CLOEXEC fd will be
1546 * inaccessible after exec. Relies on having exclusive access to
1547 * current->files (due to unshare_files above).
1548 */
1549 if (close_on_exec(fd, rcu_dereference_raw(current->files->fdt)))
1550 bprm->interp_flags |= BINPRM_FLAGS_PATH_INACCESSIBLE;
1551 bprm->filename = pathbuf;
1552 }
1553 bprm->interp = bprm->filename;
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001554
1555 retval = bprm_mm_init(bprm);
1556 if (retval)
Oleg Nesterov63e46b92014-01-23 15:55:51 -08001557 goto out_unmark;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001558
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001559 bprm->argc = count(argv, MAX_ARG_STRINGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001560 if ((retval = bprm->argc) < 0)
David Howellsa6f76f22008-11-14 10:39:24 +11001561 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001562
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001563 bprm->envc = count(envp, MAX_ARG_STRINGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001564 if ((retval = bprm->envc) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001565 goto out;
1566
1567 retval = prepare_binprm(bprm);
1568 if (retval < 0)
1569 goto out;
1570
1571 retval = copy_strings_kernel(1, &bprm->filename, bprm);
1572 if (retval < 0)
1573 goto out;
1574
1575 bprm->exec = bprm->p;
1576 retval = copy_strings(bprm->envc, envp, bprm);
1577 if (retval < 0)
1578 goto out;
1579
1580 retval = copy_strings(bprm->argc, argv, bprm);
1581 if (retval < 0)
1582 goto out;
1583
Oleg Nesterov5d1baf32013-09-11 14:24:38 -07001584 retval = exec_binprm(bprm);
David Howellsa6f76f22008-11-14 10:39:24 +11001585 if (retval < 0)
1586 goto out;
1587
1588 /* execve succeeded */
Al Viro498052b2009-03-30 07:20:30 -04001589 current->fs->in_exec = 0;
Kentaro Takedaf9ce1f12009-02-05 17:18:11 +09001590 current->in_execve = 0;
David Howellsa6f76f22008-11-14 10:39:24 +11001591 acct_update_integrals(current);
Rik van Riel82727012013-10-07 11:29:28 +01001592 task_numa_free(current);
David Howellsa6f76f22008-11-14 10:39:24 +11001593 free_bprm(bprm);
David Drysdale51f39a12014-12-12 16:57:29 -08001594 kfree(pathbuf);
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001595 putname(filename);
David Howellsa6f76f22008-11-14 10:39:24 +11001596 if (displaced)
1597 put_files_struct(displaced);
1598 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001599
1600out:
Oleg Nesterov3c77f842010-11-30 20:55:34 +01001601 if (bprm->mm) {
1602 acct_arg_size(bprm, 0);
1603 mmput(bprm->mm);
1604 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001605
Al Viro498052b2009-03-30 07:20:30 -04001606out_unmark:
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001607 current->fs->in_exec = 0;
Kentaro Takedaf9ce1f12009-02-05 17:18:11 +09001608 current->in_execve = 0;
David Howellsa6f76f22008-11-14 10:39:24 +11001609
1610out_free:
Al Viro08a6fac2008-05-10 16:38:25 -04001611 free_bprm(bprm);
David Drysdale51f39a12014-12-12 16:57:29 -08001612 kfree(pathbuf);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001613
Al Virofd8328b2008-04-22 05:11:59 -04001614out_files:
Al Viro3b125382008-04-22 05:31:30 -04001615 if (displaced)
1616 reset_files_struct(displaced);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001617out_ret:
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001618 putname(filename);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001619 return retval;
1620}
1621
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001622int do_execve(struct filename *filename,
Oleg Nesterovba2d01622011-03-06 18:02:37 +01001623 const char __user *const __user *__argv,
Al Viroda3d4c52012-10-20 21:49:33 -04001624 const char __user *const __user *__envp)
Oleg Nesterovba2d01622011-03-06 18:02:37 +01001625{
Oleg Nesterov0e028462011-03-06 18:02:54 +01001626 struct user_arg_ptr argv = { .ptr.native = __argv };
1627 struct user_arg_ptr envp = { .ptr.native = __envp };
David Drysdale51f39a12014-12-12 16:57:29 -08001628 return do_execveat_common(AT_FDCWD, filename, argv, envp, 0);
1629}
1630
1631int do_execveat(int fd, struct filename *filename,
1632 const char __user *const __user *__argv,
1633 const char __user *const __user *__envp,
1634 int flags)
1635{
1636 struct user_arg_ptr argv = { .ptr.native = __argv };
1637 struct user_arg_ptr envp = { .ptr.native = __envp };
1638
1639 return do_execveat_common(fd, filename, argv, envp, flags);
Oleg Nesterovba2d01622011-03-06 18:02:37 +01001640}
1641
Oleg Nesterov0e028462011-03-06 18:02:54 +01001642#ifdef CONFIG_COMPAT
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001643static int compat_do_execve(struct filename *filename,
Al Viro38b983b2012-09-30 13:38:55 -04001644 const compat_uptr_t __user *__argv,
Al Virod03d26e2012-10-20 21:46:25 -04001645 const compat_uptr_t __user *__envp)
Oleg Nesterov0e028462011-03-06 18:02:54 +01001646{
1647 struct user_arg_ptr argv = {
1648 .is_compat = true,
1649 .ptr.compat = __argv,
1650 };
1651 struct user_arg_ptr envp = {
1652 .is_compat = true,
1653 .ptr.compat = __envp,
1654 };
David Drysdale51f39a12014-12-12 16:57:29 -08001655 return do_execveat_common(AT_FDCWD, filename, argv, envp, 0);
1656}
1657
1658static int compat_do_execveat(int fd, struct filename *filename,
1659 const compat_uptr_t __user *__argv,
1660 const compat_uptr_t __user *__envp,
1661 int flags)
1662{
1663 struct user_arg_ptr argv = {
1664 .is_compat = true,
1665 .ptr.compat = __argv,
1666 };
1667 struct user_arg_ptr envp = {
1668 .is_compat = true,
1669 .ptr.compat = __envp,
1670 };
1671 return do_execveat_common(fd, filename, argv, envp, flags);
Oleg Nesterov0e028462011-03-06 18:02:54 +01001672}
1673#endif
1674
Oleg Nesterov964ee7d2009-09-23 15:56:59 -07001675void set_binfmt(struct linux_binfmt *new)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001676{
Hiroshi Shimamoto801460d2009-09-23 15:57:41 -07001677 struct mm_struct *mm = current->mm;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001678
Hiroshi Shimamoto801460d2009-09-23 15:57:41 -07001679 if (mm->binfmt)
1680 module_put(mm->binfmt->module);
1681
1682 mm->binfmt = new;
Oleg Nesterov964ee7d2009-09-23 15:56:59 -07001683 if (new)
1684 __module_get(new->module);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001685}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001686EXPORT_SYMBOL(set_binfmt);
1687
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001688/*
Oleg Nesterov7288e112014-01-23 15:55:32 -08001689 * set_dumpable stores three-value SUID_DUMP_* into mm->flags.
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001690 */
1691void set_dumpable(struct mm_struct *mm, int value)
1692{
Oleg Nesterovabacd2f2014-01-23 15:55:31 -08001693 unsigned long old, new;
1694
Oleg Nesterov7288e112014-01-23 15:55:32 -08001695 if (WARN_ON((unsigned)value > SUID_DUMP_ROOT))
1696 return;
1697
Oleg Nesterovabacd2f2014-01-23 15:55:31 -08001698 do {
1699 old = ACCESS_ONCE(mm->flags);
Oleg Nesterov7288e112014-01-23 15:55:32 -08001700 new = (old & ~MMF_DUMPABLE_MASK) | value;
Oleg Nesterovabacd2f2014-01-23 15:55:31 -08001701 } while (cmpxchg(&mm->flags, old, new) != old);
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001702}
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001703
Al Viro38b983b2012-09-30 13:38:55 -04001704SYSCALL_DEFINE3(execve,
1705 const char __user *, filename,
1706 const char __user *const __user *, argv,
1707 const char __user *const __user *, envp)
1708{
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001709 return do_execve(getname(filename), argv, envp);
Al Viro38b983b2012-09-30 13:38:55 -04001710}
David Drysdale51f39a12014-12-12 16:57:29 -08001711
1712SYSCALL_DEFINE5(execveat,
1713 int, fd, const char __user *, filename,
1714 const char __user *const __user *, argv,
1715 const char __user *const __user *, envp,
1716 int, flags)
1717{
1718 int lookup_flags = (flags & AT_EMPTY_PATH) ? LOOKUP_EMPTY : 0;
1719
1720 return do_execveat(fd,
1721 getname_flags(filename, lookup_flags, NULL),
1722 argv, envp, flags);
1723}
1724
Al Viro38b983b2012-09-30 13:38:55 -04001725#ifdef CONFIG_COMPAT
Heiko Carstens625b1d72014-03-04 10:53:50 +01001726COMPAT_SYSCALL_DEFINE3(execve, const char __user *, filename,
1727 const compat_uptr_t __user *, argv,
1728 const compat_uptr_t __user *, envp)
Al Viro38b983b2012-09-30 13:38:55 -04001729{
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001730 return compat_do_execve(getname(filename), argv, envp);
Al Viro38b983b2012-09-30 13:38:55 -04001731}
David Drysdale51f39a12014-12-12 16:57:29 -08001732
1733COMPAT_SYSCALL_DEFINE5(execveat, int, fd,
1734 const char __user *, filename,
1735 const compat_uptr_t __user *, argv,
1736 const compat_uptr_t __user *, envp,
1737 int, flags)
1738{
1739 int lookup_flags = (flags & AT_EMPTY_PATH) ? LOOKUP_EMPTY : 0;
1740
1741 return compat_do_execveat(fd,
1742 getname_flags(filename, lookup_flags, NULL),
1743 argv, envp, flags);
1744}
Al Viro38b983b2012-09-30 13:38:55 -04001745#endif