blob: c4010b8207a144303ba498269427e42de1c72336 [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>
Mimi Zoharb44a7df2015-12-28 16:02:29 -050059#include <linux/vmalloc.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070060
61#include <asm/uaccess.h>
62#include <asm/mmu_context.h>
Ollie Wildb6a2fea2007-07-19 01:48:16 -070063#include <asm/tlb.h>
KAMEZAWA Hiroyuki43d2b112012-01-10 15:08:09 -080064
65#include <trace/events/task.h>
David Howellsa6f76f22008-11-14 10:39:24 +110066#include "internal.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070067
David Smith4ff16c22012-02-07 10:11:05 -060068#include <trace/events/sched.h>
69
Alan Coxd6e71142005-06-23 00:09:43 -070070int suid_dumpable = 0;
71
Alexey Dobriyane4dc1b12007-10-16 23:26:03 -070072static LIST_HEAD(formats);
Linus Torvalds1da177e2005-04-16 15:20:36 -070073static DEFINE_RWLOCK(binfmt_lock);
74
Al Viro8fc3dc52012-03-17 03:05:16 -040075void __register_binfmt(struct linux_binfmt * fmt, int insert)
Linus Torvalds1da177e2005-04-16 15:20:36 -070076{
Al Viro8fc3dc52012-03-17 03:05:16 -040077 BUG_ON(!fmt);
Oleg Nesterov92eaa562013-09-11 14:24:42 -070078 if (WARN_ON(!fmt->load_binary))
79 return;
Linus Torvalds1da177e2005-04-16 15:20:36 -070080 write_lock(&binfmt_lock);
Ivan Kokshaysky74641f52009-04-30 15:08:49 -070081 insert ? list_add(&fmt->lh, &formats) :
82 list_add_tail(&fmt->lh, &formats);
Linus Torvalds1da177e2005-04-16 15:20:36 -070083 write_unlock(&binfmt_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -070084}
85
Ivan Kokshaysky74641f52009-04-30 15:08:49 -070086EXPORT_SYMBOL(__register_binfmt);
Linus Torvalds1da177e2005-04-16 15:20:36 -070087
Alexey Dobriyanf6b450d2007-10-16 23:26:04 -070088void unregister_binfmt(struct linux_binfmt * fmt)
Linus Torvalds1da177e2005-04-16 15:20:36 -070089{
Linus Torvalds1da177e2005-04-16 15:20:36 -070090 write_lock(&binfmt_lock);
Alexey Dobriyane4dc1b12007-10-16 23:26:03 -070091 list_del(&fmt->lh);
Linus Torvalds1da177e2005-04-16 15:20:36 -070092 write_unlock(&binfmt_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -070093}
94
95EXPORT_SYMBOL(unregister_binfmt);
96
97static inline void put_binfmt(struct linux_binfmt * fmt)
98{
99 module_put(fmt->module);
100}
101
Eric W. Biederman90f85722015-06-29 14:42:03 -0500102bool path_noexec(const struct path *path)
103{
104 return (path->mnt->mnt_flags & MNT_NOEXEC) ||
105 (path->mnt->mnt_sb->s_iflags & SB_I_NOEXEC);
106}
107
Josh Triplett69369a72014-04-03 14:48:27 -0700108#ifdef CONFIG_USELIB
Linus Torvalds1da177e2005-04-16 15:20:36 -0700109/*
110 * Note that a shared library must be both readable and executable due to
111 * security reasons.
112 *
113 * Also note that we take the address to load from from the file itself.
114 */
Heiko Carstens1e7bfb22009-01-14 14:14:29 +0100115SYSCALL_DEFINE1(uselib, const char __user *, library)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116{
Al Viro72c2d532013-09-22 16:27:52 -0400117 struct linux_binfmt *fmt;
Al Viro964bd182008-07-26 03:33:14 -0400118 struct file *file;
Jeff Layton91a27b22012-10-10 15:25:28 -0400119 struct filename *tmp = getname(library);
Al Viro964bd182008-07-26 03:33:14 -0400120 int error = PTR_ERR(tmp);
Al Viro47c805d2011-02-23 17:44:09 -0500121 static const struct open_flags uselib_flags = {
122 .open_flag = O_LARGEFILE | O_RDONLY | __FMODE_EXEC,
Al Viro62fb4a12015-12-26 22:33:24 -0500123 .acc_mode = MAY_READ | MAY_EXEC,
Al Virof9652e12013-06-11 08:23:01 +0400124 .intent = LOOKUP_OPEN,
125 .lookup_flags = LOOKUP_FOLLOW,
Al Viro47c805d2011-02-23 17:44:09 -0500126 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700127
Al Viro6e8341a2009-04-06 11:16:22 -0400128 if (IS_ERR(tmp))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129 goto out;
130
Al Virof9652e12013-06-11 08:23:01 +0400131 file = do_filp_open(AT_FDCWD, tmp, &uselib_flags);
Al Viro6e8341a2009-04-06 11:16:22 -0400132 putname(tmp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700133 error = PTR_ERR(file);
134 if (IS_ERR(file))
135 goto out;
136
Al Viro6e8341a2009-04-06 11:16:22 -0400137 error = -EINVAL;
Al Viro496ad9a2013-01-23 17:07:38 -0500138 if (!S_ISREG(file_inode(file)->i_mode))
Al Viro6e8341a2009-04-06 11:16:22 -0400139 goto exit;
140
141 error = -EACCES;
Eric W. Biederman90f85722015-06-29 14:42:03 -0500142 if (path_noexec(&file->f_path))
Al Viro6e8341a2009-04-06 11:16:22 -0400143 goto exit;
144
Eric Paris2a12a9d2009-12-17 21:24:21 -0500145 fsnotify_open(file);
Eric Paris6110e3a2008-12-17 13:53:20 -0500146
Linus Torvalds1da177e2005-04-16 15:20:36 -0700147 error = -ENOEXEC;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700148
Al Viro72c2d532013-09-22 16:27:52 -0400149 read_lock(&binfmt_lock);
150 list_for_each_entry(fmt, &formats, lh) {
151 if (!fmt->load_shlib)
152 continue;
153 if (!try_module_get(fmt->module))
154 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155 read_unlock(&binfmt_lock);
Al Viro72c2d532013-09-22 16:27:52 -0400156 error = fmt->load_shlib(file);
157 read_lock(&binfmt_lock);
158 put_binfmt(fmt);
159 if (error != -ENOEXEC)
160 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700161 }
Al Viro72c2d532013-09-22 16:27:52 -0400162 read_unlock(&binfmt_lock);
Al Viro6e8341a2009-04-06 11:16:22 -0400163exit:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700164 fput(file);
165out:
166 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167}
Josh Triplett69369a72014-04-03 14:48:27 -0700168#endif /* #ifdef CONFIG_USELIB */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700169
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700170#ifdef CONFIG_MMU
Oleg Nesterovae6b5852011-03-06 18:03:11 +0100171/*
172 * The nascent bprm->mm is not visible until exec_mmap() but it can
173 * use a lot of memory, account these pages in current->mm temporary
174 * for oom_badness()->get_mm_rss(). Once exec succeeds or fails, we
175 * change the counter back via acct_arg_size(0).
176 */
Oleg Nesterov0e028462011-03-06 18:02:54 +0100177static void acct_arg_size(struct linux_binprm *bprm, unsigned long pages)
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100178{
179 struct mm_struct *mm = current->mm;
180 long diff = (long)(pages - bprm->vma_pages);
181
182 if (!mm || !diff)
183 return;
184
185 bprm->vma_pages = pages;
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100186 add_mm_counter(mm, MM_ANONPAGES, diff);
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100187}
188
Oleg Nesterov0e028462011-03-06 18:02:54 +0100189static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos,
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700190 int write)
191{
192 struct page *page;
193 int ret;
194
195#ifdef CONFIG_STACK_GROWSUP
196 if (write) {
Michal Hockod05f3162011-05-24 17:11:44 -0700197 ret = expand_downwards(bprm->vma, pos);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700198 if (ret < 0)
199 return NULL;
200 }
201#endif
Dave Hansen1e987792016-02-12 13:01:54 -0800202 /*
203 * We are doing an exec(). 'current' is the process
204 * doing the exec and bprm->mm is the new process's mm.
205 */
206 ret = get_user_pages_remote(current, bprm->mm, pos, 1, write,
207 1, &page, NULL);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700208 if (ret <= 0)
209 return NULL;
210
211 if (write) {
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700212 unsigned long size = bprm->vma->vm_end - bprm->vma->vm_start;
Linus Torvaldsa64e7152008-03-03 10:12:14 -0800213 struct rlimit *rlim;
214
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100215 acct_arg_size(bprm, size / PAGE_SIZE);
216
Linus Torvaldsa64e7152008-03-03 10:12:14 -0800217 /*
218 * We've historically supported up to 32 pages (ARG_MAX)
219 * of argument strings even with small stacks
220 */
221 if (size <= ARG_MAX)
222 return page;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700223
224 /*
225 * Limit to 1/4-th the stack size for the argv+env strings.
226 * This ensures that:
227 * - the remaining binfmt code will not run out of stack space,
228 * - the program will have a reasonable amount of stack left
229 * to work from.
230 */
Linus Torvaldsa64e7152008-03-03 10:12:14 -0800231 rlim = current->signal->rlim;
Jiri Slabyd554ed892010-03-05 13:42:42 -0800232 if (size > ACCESS_ONCE(rlim[RLIMIT_STACK].rlim_cur) / 4) {
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700233 put_page(page);
234 return NULL;
235 }
236 }
237
238 return page;
239}
240
241static void put_arg_page(struct page *page)
242{
243 put_page(page);
244}
245
246static void free_arg_page(struct linux_binprm *bprm, int i)
247{
248}
249
250static void free_arg_pages(struct linux_binprm *bprm)
251{
252}
253
254static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos,
255 struct page *page)
256{
257 flush_cache_page(bprm->vma, pos, page_to_pfn(page));
258}
259
260static int __bprm_mm_init(struct linux_binprm *bprm)
261{
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800262 int err;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700263 struct vm_area_struct *vma = NULL;
264 struct mm_struct *mm = bprm->mm;
265
266 bprm->vma = vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL);
267 if (!vma)
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800268 return -ENOMEM;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700269
270 down_write(&mm->mmap_sem);
271 vma->vm_mm = mm;
272
273 /*
274 * Place the stack at the largest stack address the architecture
275 * supports. Later, we'll move this to an appropriate place. We don't
276 * use STACK_TOP because that can depend on attributes which aren't
277 * configured yet.
278 */
Michal Hockoaacb3d12011-07-26 16:08:40 -0700279 BUILD_BUG_ON(VM_STACK_FLAGS & VM_STACK_INCOMPLETE_SETUP);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700280 vma->vm_end = STACK_TOP_MAX;
281 vma->vm_start = vma->vm_end - PAGE_SIZE;
Cyrill Gorcunovd9104d12013-09-11 14:22:24 -0700282 vma->vm_flags = VM_SOFTDIRTY | VM_STACK_FLAGS | VM_STACK_INCOMPLETE_SETUP;
Coly Li3ed75eb2007-10-18 23:39:15 -0700283 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
Rik van Riel5beb4932010-03-05 13:42:07 -0800284 INIT_LIST_HEAD(&vma->anon_vma_chain);
Tavis Ormandy462e635e2010-12-09 15:29:42 +0100285
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700286 err = insert_vm_struct(mm, vma);
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800287 if (err)
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700288 goto err;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700289
290 mm->stack_vm = mm->total_vm = 1;
Dave Hansenfe3d1972014-11-14 07:18:29 -0800291 arch_bprm_mm_init(mm, vma);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700292 up_write(&mm->mmap_sem);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700293 bprm->p = vma->vm_end - sizeof(void *);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700294 return 0;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700295err:
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800296 up_write(&mm->mmap_sem);
297 bprm->vma = NULL;
298 kmem_cache_free(vm_area_cachep, vma);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700299 return err;
300}
301
302static bool valid_arg_len(struct linux_binprm *bprm, long len)
303{
304 return len <= MAX_ARG_STRLEN;
305}
306
307#else
308
Oleg Nesterov0e028462011-03-06 18:02:54 +0100309static inline void acct_arg_size(struct linux_binprm *bprm, unsigned long pages)
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100310{
311}
312
Oleg Nesterov0e028462011-03-06 18:02:54 +0100313static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos,
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700314 int write)
315{
316 struct page *page;
317
318 page = bprm->page[pos / PAGE_SIZE];
319 if (!page && write) {
320 page = alloc_page(GFP_HIGHUSER|__GFP_ZERO);
321 if (!page)
322 return NULL;
323 bprm->page[pos / PAGE_SIZE] = page;
324 }
325
326 return page;
327}
328
329static void put_arg_page(struct page *page)
330{
331}
332
333static void free_arg_page(struct linux_binprm *bprm, int i)
334{
335 if (bprm->page[i]) {
336 __free_page(bprm->page[i]);
337 bprm->page[i] = NULL;
338 }
339}
340
341static void free_arg_pages(struct linux_binprm *bprm)
342{
343 int i;
344
345 for (i = 0; i < MAX_ARG_PAGES; i++)
346 free_arg_page(bprm, i);
347}
348
349static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos,
350 struct page *page)
351{
352}
353
354static int __bprm_mm_init(struct linux_binprm *bprm)
355{
356 bprm->p = PAGE_SIZE * MAX_ARG_PAGES - sizeof(void *);
357 return 0;
358}
359
360static bool valid_arg_len(struct linux_binprm *bprm, long len)
361{
362 return len <= bprm->p;
363}
364
365#endif /* CONFIG_MMU */
366
367/*
368 * Create a new mm_struct and populate it with a temporary stack
369 * vm_area_struct. We don't have enough context at this point to set the stack
370 * flags, permissions, and offset, so we use temporary values. We'll update
371 * them later in setup_arg_pages().
372 */
Yuanhan Liu9cc64ce2013-02-20 13:16:01 +1100373static int bprm_mm_init(struct linux_binprm *bprm)
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700374{
375 int err;
376 struct mm_struct *mm = NULL;
377
378 bprm->mm = mm = mm_alloc();
379 err = -ENOMEM;
380 if (!mm)
381 goto err;
382
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700383 err = __bprm_mm_init(bprm);
384 if (err)
385 goto err;
386
387 return 0;
388
389err:
390 if (mm) {
391 bprm->mm = NULL;
392 mmdrop(mm);
393 }
394
395 return err;
396}
397
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100398struct user_arg_ptr {
Oleg Nesterov0e028462011-03-06 18:02:54 +0100399#ifdef CONFIG_COMPAT
400 bool is_compat;
401#endif
402 union {
403 const char __user *const __user *native;
404#ifdef CONFIG_COMPAT
Al Viro38b983b2012-09-30 13:38:55 -0400405 const compat_uptr_t __user *compat;
Oleg Nesterov0e028462011-03-06 18:02:54 +0100406#endif
407 } ptr;
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100408};
409
410static const char __user *get_user_arg_ptr(struct user_arg_ptr argv, int nr)
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100411{
Oleg Nesterov0e028462011-03-06 18:02:54 +0100412 const char __user *native;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100413
Oleg Nesterov0e028462011-03-06 18:02:54 +0100414#ifdef CONFIG_COMPAT
415 if (unlikely(argv.is_compat)) {
416 compat_uptr_t compat;
417
418 if (get_user(compat, argv.ptr.compat + nr))
419 return ERR_PTR(-EFAULT);
420
421 return compat_ptr(compat);
422 }
423#endif
424
425 if (get_user(native, argv.ptr.native + nr))
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100426 return ERR_PTR(-EFAULT);
427
Oleg Nesterov0e028462011-03-06 18:02:54 +0100428 return native;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100429}
430
Linus Torvalds1da177e2005-04-16 15:20:36 -0700431/*
432 * count() counts the number of strings in array ARGV.
433 */
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100434static int count(struct user_arg_ptr argv, int max)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435{
436 int i = 0;
437
Oleg Nesterov0e028462011-03-06 18:02:54 +0100438 if (argv.ptr.native != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700439 for (;;) {
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100440 const char __user *p = get_user_arg_ptr(argv, i);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442 if (!p)
443 break;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100444
445 if (IS_ERR(p))
446 return -EFAULT;
447
Xi Wang6d92d4f2013-01-11 14:31:48 -0800448 if (i >= max)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700449 return -E2BIG;
Xi Wang6d92d4f2013-01-11 14:31:48 -0800450 ++i;
Roland McGrath9aea5a62010-09-07 19:37:06 -0700451
452 if (fatal_signal_pending(current))
453 return -ERESTARTNOHAND;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454 cond_resched();
455 }
456 }
457 return i;
458}
459
460/*
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700461 * 'copy_strings()' copies argument/environment strings from the old
462 * processes's memory to the new process's stack. The call to get_user_pages()
463 * ensures the destination page is created and not swapped out.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464 */
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100465static int copy_strings(int argc, struct user_arg_ptr argv,
Adrian Bunk75c96f82005-05-05 16:16:09 -0700466 struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467{
468 struct page *kmapped_page = NULL;
469 char *kaddr = NULL;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700470 unsigned long kpos = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471 int ret;
472
473 while (argc-- > 0) {
David Howellsd7627462010-08-17 23:52:56 +0100474 const char __user *str;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700475 int len;
476 unsigned long pos;
477
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100478 ret = -EFAULT;
479 str = get_user_arg_ptr(argv, argc);
480 if (IS_ERR(str))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100483 len = strnlen_user(str, MAX_ARG_STRLEN);
484 if (!len)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485 goto out;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100486
487 ret = -E2BIG;
488 if (!valid_arg_len(bprm, len))
489 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700491 /* We're going to work our way backwords. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492 pos = bprm->p;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700493 str += len;
494 bprm->p -= len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495
496 while (len > 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497 int offset, bytes_to_copy;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498
Roland McGrath9aea5a62010-09-07 19:37:06 -0700499 if (fatal_signal_pending(current)) {
500 ret = -ERESTARTNOHAND;
501 goto out;
502 }
Roland McGrath7993bc12010-09-07 19:36:28 -0700503 cond_resched();
504
Linus Torvalds1da177e2005-04-16 15:20:36 -0700505 offset = pos % PAGE_SIZE;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700506 if (offset == 0)
507 offset = PAGE_SIZE;
508
509 bytes_to_copy = offset;
510 if (bytes_to_copy > len)
511 bytes_to_copy = len;
512
513 offset -= bytes_to_copy;
514 pos -= bytes_to_copy;
515 str -= bytes_to_copy;
516 len -= bytes_to_copy;
517
518 if (!kmapped_page || kpos != (pos & PAGE_MASK)) {
519 struct page *page;
520
521 page = get_arg_page(bprm, pos, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522 if (!page) {
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700523 ret = -E2BIG;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700524 goto out;
525 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700526
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700527 if (kmapped_page) {
528 flush_kernel_dcache_page(kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529 kunmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700530 put_arg_page(kmapped_page);
531 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532 kmapped_page = page;
533 kaddr = kmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700534 kpos = pos & PAGE_MASK;
535 flush_arg_page(bprm, kpos, kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536 }
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700537 if (copy_from_user(kaddr+offset, str, bytes_to_copy)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538 ret = -EFAULT;
539 goto out;
540 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541 }
542 }
543 ret = 0;
544out:
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700545 if (kmapped_page) {
546 flush_kernel_dcache_page(kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547 kunmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700548 put_arg_page(kmapped_page);
549 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550 return ret;
551}
552
553/*
554 * Like copy_strings, but get argv and its values from kernel memory.
555 */
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100556int copy_strings_kernel(int argc, const char *const *__argv,
David Howellsd7627462010-08-17 23:52:56 +0100557 struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700558{
559 int r;
560 mm_segment_t oldfs = get_fs();
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100561 struct user_arg_ptr argv = {
Oleg Nesterov0e028462011-03-06 18:02:54 +0100562 .ptr.native = (const char __user *const __user *)__argv,
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100563 };
564
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565 set_fs(KERNEL_DS);
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100566 r = copy_strings(argc, argv, bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567 set_fs(oldfs);
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100568
Linus Torvalds1da177e2005-04-16 15:20:36 -0700569 return r;
570}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571EXPORT_SYMBOL(copy_strings_kernel);
572
573#ifdef CONFIG_MMU
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700574
Linus Torvalds1da177e2005-04-16 15:20:36 -0700575/*
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700576 * During bprm_mm_init(), we create a temporary stack at STACK_TOP_MAX. Once
577 * the binfmt code determines where the new stack should reside, we shift it to
578 * its final location. The process proceeds as follows:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700579 *
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700580 * 1) Use shift to calculate the new vma endpoints.
581 * 2) Extend vma to cover both the old and new ranges. This ensures the
582 * arguments passed to subsequent functions are consistent.
583 * 3) Move vma's page tables to the new range.
584 * 4) Free up any cleared pgd range.
585 * 5) Shrink the vma to cover only the new range.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700586 */
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700587static int shift_arg_pages(struct vm_area_struct *vma, unsigned long shift)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700588{
589 struct mm_struct *mm = vma->vm_mm;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700590 unsigned long old_start = vma->vm_start;
591 unsigned long old_end = vma->vm_end;
592 unsigned long length = old_end - old_start;
593 unsigned long new_start = old_start - shift;
594 unsigned long new_end = old_end - shift;
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700595 struct mmu_gather tlb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700597 BUG_ON(new_start > new_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700598
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700599 /*
600 * ensure there are no vmas between where we want to go
601 * and where we are
602 */
603 if (vma != find_vma(mm, new_start))
604 return -EFAULT;
605
606 /*
607 * cover the whole range: [new_start, old_end)
608 */
Rik van Riel5beb4932010-03-05 13:42:07 -0800609 if (vma_adjust(vma, new_start, old_end, vma->vm_pgoff, NULL))
610 return -ENOMEM;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700611
612 /*
613 * move the page tables downwards, on failure we rely on
614 * process cleanup to remove whatever mess we made.
615 */
616 if (length != move_page_tables(vma, old_start,
Michel Lespinasse38a76012012-10-08 16:31:50 -0700617 vma, new_start, length, false))
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700618 return -ENOMEM;
619
620 lru_add_drain();
Linus Torvalds2b047252013-08-15 11:42:25 -0700621 tlb_gather_mmu(&tlb, mm, old_start, old_end);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700622 if (new_end > old_start) {
623 /*
624 * when the old and new regions overlap clear from new_end.
625 */
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700626 free_pgd_range(&tlb, new_end, old_end, new_end,
Hugh Dickins6ee86302013-04-29 15:07:44 -0700627 vma->vm_next ? vma->vm_next->vm_start : USER_PGTABLES_CEILING);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700628 } else {
629 /*
630 * otherwise, clean from old_start; this is done to not touch
631 * the address space in [new_end, old_start) some architectures
632 * have constraints on va-space that make this illegal (IA64) -
633 * for the others its just a little faster.
634 */
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700635 free_pgd_range(&tlb, old_start, old_end, new_end,
Hugh Dickins6ee86302013-04-29 15:07:44 -0700636 vma->vm_next ? vma->vm_next->vm_start : USER_PGTABLES_CEILING);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700637 }
Linus Torvalds2b047252013-08-15 11:42:25 -0700638 tlb_finish_mmu(&tlb, old_start, old_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700639
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700640 /*
Rik van Riel5beb4932010-03-05 13:42:07 -0800641 * Shrink the vma to just the new range. Always succeeds.
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700642 */
643 vma_adjust(vma, new_start, new_end, vma->vm_pgoff, NULL);
644
645 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700646}
647
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700648/*
649 * Finalizes the stack vm_area_struct. The flags and permissions are updated,
650 * the stack is optionally relocated, and some extra space is added.
651 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700652int setup_arg_pages(struct linux_binprm *bprm,
653 unsigned long stack_top,
654 int executable_stack)
655{
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700656 unsigned long ret;
657 unsigned long stack_shift;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658 struct mm_struct *mm = current->mm;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700659 struct vm_area_struct *vma = bprm->vma;
660 struct vm_area_struct *prev = NULL;
661 unsigned long vm_flags;
662 unsigned long stack_base;
Michael Neuling803bf5e2010-02-10 13:56:42 -0800663 unsigned long stack_size;
664 unsigned long stack_expand;
665 unsigned long rlim_stack;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700666
667#ifdef CONFIG_STACK_GROWSUP
James Hogand71f2902014-05-13 23:58:24 +0100668 /* Limit stack size */
Jiri Slabyd554ed892010-03-05 13:42:42 -0800669 stack_base = rlimit_max(RLIMIT_STACK);
James Hogand71f2902014-05-13 23:58:24 +0100670 if (stack_base > STACK_SIZE_MAX)
671 stack_base = STACK_SIZE_MAX;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700672
Helge Dellerd045c772015-05-11 22:01:27 +0200673 /* Add space for stack randomization. */
674 stack_base += (STACK_RND_MASK << PAGE_SHIFT);
675
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700676 /* Make sure we didn't let the argument array grow too large. */
677 if (vma->vm_end - vma->vm_start > stack_base)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700678 return -ENOMEM;
679
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700680 stack_base = PAGE_ALIGN(stack_top - stack_base);
681
682 stack_shift = vma->vm_start - stack_base;
683 mm->arg_start = bprm->p - stack_shift;
684 bprm->p = vma->vm_end - stack_shift;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700685#else
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700686 stack_top = arch_align_stack(stack_top);
687 stack_top = PAGE_ALIGN(stack_top);
Roland McGrath1b528182010-09-07 19:35:49 -0700688
689 if (unlikely(stack_top < mmap_min_addr) ||
690 unlikely(vma->vm_end - vma->vm_start >= stack_top - mmap_min_addr))
691 return -ENOMEM;
692
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700693 stack_shift = vma->vm_end - stack_top;
694
695 bprm->p -= stack_shift;
696 mm->arg_start = bprm->p;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700697#endif
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700698
699 if (bprm->loader)
700 bprm->loader -= stack_shift;
701 bprm->exec -= stack_shift;
702
703 down_write(&mm->mmap_sem);
Hugh Dickins96a8e132008-07-10 21:19:20 +0100704 vm_flags = VM_STACK_FLAGS;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700705
706 /*
707 * Adjust stack execute permissions; explicitly enable for
708 * EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X and leave alone
709 * (arch default) otherwise.
710 */
711 if (unlikely(executable_stack == EXSTACK_ENABLE_X))
712 vm_flags |= VM_EXEC;
713 else if (executable_stack == EXSTACK_DISABLE_X)
714 vm_flags &= ~VM_EXEC;
715 vm_flags |= mm->def_flags;
Mel Gormana8bef8ff2010-05-24 14:32:24 -0700716 vm_flags |= VM_STACK_INCOMPLETE_SETUP;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700717
718 ret = mprotect_fixup(vma, &prev, vma->vm_start, vma->vm_end,
719 vm_flags);
720 if (ret)
721 goto out_unlock;
722 BUG_ON(prev != vma);
723
724 /* Move stack pages down in memory. */
725 if (stack_shift) {
726 ret = shift_arg_pages(vma, stack_shift);
Anton Blanchardfc63cf22009-11-11 14:26:48 -0800727 if (ret)
728 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729 }
730
Mel Gormana8bef8ff2010-05-24 14:32:24 -0700731 /* mprotect_fixup is overkill to remove the temporary stack flags */
732 vma->vm_flags &= ~VM_STACK_INCOMPLETE_SETUP;
733
Michael Neuling5ef097d2010-03-05 13:42:57 -0800734 stack_expand = 131072UL; /* randomly 32*4k (or 2*64k) pages */
Michael Neuling803bf5e2010-02-10 13:56:42 -0800735 stack_size = vma->vm_end - vma->vm_start;
736 /*
737 * Align this down to a page boundary as expand_stack
738 * will align it up.
739 */
740 rlim_stack = rlimit(RLIMIT_STACK) & PAGE_MASK;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700741#ifdef CONFIG_STACK_GROWSUP
Michael Neuling803bf5e2010-02-10 13:56:42 -0800742 if (stack_size + stack_expand > rlim_stack)
743 stack_base = vma->vm_start + rlim_stack;
744 else
745 stack_base = vma->vm_end + stack_expand;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700746#else
Michael Neuling803bf5e2010-02-10 13:56:42 -0800747 if (stack_size + stack_expand > rlim_stack)
748 stack_base = vma->vm_end - rlim_stack;
749 else
750 stack_base = vma->vm_start - stack_expand;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700751#endif
Eric B Munson3af9e852010-05-18 15:30:49 +0100752 current->mm->start_stack = bprm->p;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700753 ret = expand_stack(vma, stack_base);
754 if (ret)
755 ret = -EFAULT;
756
757out_unlock:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758 up_write(&mm->mmap_sem);
Anton Blanchardfc63cf22009-11-11 14:26:48 -0800759 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700760}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761EXPORT_SYMBOL(setup_arg_pages);
762
Linus Torvalds1da177e2005-04-16 15:20:36 -0700763#endif /* CONFIG_MMU */
764
David Drysdale51f39a12014-12-12 16:57:29 -0800765static struct file *do_open_execat(int fd, struct filename *name, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767 struct file *file;
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200768 int err;
David Drysdale51f39a12014-12-12 16:57:29 -0800769 struct open_flags open_exec_flags = {
Al Viro47c805d2011-02-23 17:44:09 -0500770 .open_flag = O_LARGEFILE | O_RDONLY | __FMODE_EXEC,
Al Viro62fb4a12015-12-26 22:33:24 -0500771 .acc_mode = MAY_EXEC,
Al Virof9652e12013-06-11 08:23:01 +0400772 .intent = LOOKUP_OPEN,
773 .lookup_flags = LOOKUP_FOLLOW,
Al Viro47c805d2011-02-23 17:44:09 -0500774 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775
David Drysdale51f39a12014-12-12 16:57:29 -0800776 if ((flags & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
777 return ERR_PTR(-EINVAL);
778 if (flags & AT_SYMLINK_NOFOLLOW)
779 open_exec_flags.lookup_flags &= ~LOOKUP_FOLLOW;
780 if (flags & AT_EMPTY_PATH)
781 open_exec_flags.lookup_flags |= LOOKUP_EMPTY;
782
783 file = do_filp_open(fd, name, &open_exec_flags);
Al Viro6e8341a2009-04-06 11:16:22 -0400784 if (IS_ERR(file))
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200785 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700786
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200787 err = -EACCES;
Al Viro496ad9a2013-01-23 17:07:38 -0500788 if (!S_ISREG(file_inode(file)->i_mode))
Al Viro6e8341a2009-04-06 11:16:22 -0400789 goto exit;
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200790
Eric W. Biederman90f85722015-06-29 14:42:03 -0500791 if (path_noexec(&file->f_path))
Al Viro6e8341a2009-04-06 11:16:22 -0400792 goto exit;
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200793
794 err = deny_write_access(file);
Al Viro6e8341a2009-04-06 11:16:22 -0400795 if (err)
796 goto exit;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700797
David Drysdale51f39a12014-12-12 16:57:29 -0800798 if (name->name[0] != '\0')
799 fsnotify_open(file);
800
Al Viro6e8341a2009-04-06 11:16:22 -0400801out:
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200802 return file;
803
Al Viro6e8341a2009-04-06 11:16:22 -0400804exit:
805 fput(file);
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200806 return ERR_PTR(err);
807}
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -0800808
809struct file *open_exec(const char *name)
810{
Paul Moore51689102015-01-22 00:00:03 -0500811 struct filename *filename = getname_kernel(name);
812 struct file *f = ERR_CAST(filename);
813
814 if (!IS_ERR(filename)) {
815 f = do_open_execat(AT_FDCWD, filename, 0);
816 putname(filename);
817 }
818 return f;
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -0800819}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700820EXPORT_SYMBOL(open_exec);
821
Mimi Zohar6777d772009-08-21 14:32:48 -0400822int kernel_read(struct file *file, loff_t offset,
823 char *addr, unsigned long count)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700824{
825 mm_segment_t old_fs;
826 loff_t pos = offset;
827 int result;
828
829 old_fs = get_fs();
830 set_fs(get_ds());
831 /* The cast to a user pointer is valid due to the set_fs() */
832 result = vfs_read(file, (void __user *)addr, count, &pos);
833 set_fs(old_fs);
834 return result;
835}
836
837EXPORT_SYMBOL(kernel_read);
838
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500839int kernel_read_file(struct file *file, void **buf, loff_t *size,
Mimi Zoharbc8ca5b2016-01-24 10:07:32 -0500840 loff_t max_size, enum kernel_read_file_id id)
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500841{
842 loff_t i_size, pos;
843 ssize_t bytes = 0;
844 int ret;
845
846 if (!S_ISREG(file_inode(file)->i_mode) || max_size < 0)
847 return -EINVAL;
848
Mimi Zohar39eeb4f2016-01-30 22:23:26 -0500849 ret = security_kernel_read_file(file, id);
850 if (ret)
851 return ret;
852
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500853 i_size = i_size_read(file_inode(file));
854 if (max_size > 0 && i_size > max_size)
855 return -EFBIG;
856 if (i_size <= 0)
857 return -EINVAL;
858
859 *buf = vmalloc(i_size);
860 if (!*buf)
861 return -ENOMEM;
862
863 pos = 0;
864 while (pos < i_size) {
865 bytes = kernel_read(file, pos, (char *)(*buf) + pos,
866 i_size - pos);
867 if (bytes < 0) {
868 ret = bytes;
869 goto out;
870 }
871
872 if (bytes == 0)
873 break;
874 pos += bytes;
875 }
876
877 if (pos != i_size) {
878 ret = -EIO;
879 goto out;
880 }
881
Mimi Zoharbc8ca5b2016-01-24 10:07:32 -0500882 ret = security_kernel_post_read_file(file, *buf, i_size, id);
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500883 if (!ret)
884 *size = pos;
885
886out:
887 if (ret < 0) {
888 vfree(*buf);
889 *buf = NULL;
890 }
891 return ret;
892}
893EXPORT_SYMBOL_GPL(kernel_read_file);
894
Mimi Zohar09596b92015-11-19 12:39:22 -0500895int kernel_read_file_from_path(char *path, void **buf, loff_t *size,
896 loff_t max_size, enum kernel_read_file_id id)
897{
898 struct file *file;
899 int ret;
900
901 if (!path || !*path)
902 return -EINVAL;
903
904 file = filp_open(path, O_RDONLY, 0);
905 if (IS_ERR(file))
906 return PTR_ERR(file);
907
908 ret = kernel_read_file(file, buf, size, max_size, id);
909 fput(file);
910 return ret;
911}
912EXPORT_SYMBOL_GPL(kernel_read_file_from_path);
913
Mimi Zoharb844f0e2016-02-01 08:36:21 -0500914int kernel_read_file_from_fd(int fd, void **buf, loff_t *size, loff_t max_size,
915 enum kernel_read_file_id id)
916{
917 struct fd f = fdget(fd);
918 int ret = -EBADF;
919
920 if (!f.file)
921 goto out;
922
923 ret = kernel_read_file(f.file, buf, size, max_size, id);
924out:
925 fdput(f);
926 return ret;
927}
928EXPORT_SYMBOL_GPL(kernel_read_file_from_fd);
929
Al Viro3dc20cb2013-04-13 20:31:37 -0400930ssize_t read_code(struct file *file, unsigned long addr, loff_t pos, size_t len)
931{
Al Viroec695572014-02-04 19:08:21 -0500932 ssize_t res = vfs_read(file, (void __user *)addr, len, &pos);
Al Viro3dc20cb2013-04-13 20:31:37 -0400933 if (res > 0)
934 flush_icache_range(addr, addr + len);
935 return res;
936}
937EXPORT_SYMBOL(read_code);
938
Linus Torvalds1da177e2005-04-16 15:20:36 -0700939static int exec_mmap(struct mm_struct *mm)
940{
941 struct task_struct *tsk;
Davidlohr Bueso615d6e82014-04-07 15:37:25 -0700942 struct mm_struct *old_mm, *active_mm;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700943
944 /* Notify parent that we're no longer interested in the old VM */
945 tsk = current;
946 old_mm = current->mm;
947 mm_release(tsk, old_mm);
948
949 if (old_mm) {
Konstantin Khlebnikov4fe7efd2012-06-20 12:53:01 -0700950 sync_mm_rss(old_mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700951 /*
952 * Make sure that if there is a core dump in progress
953 * for the old mm, we get out and die instead of going
954 * through with the exec. We must hold mmap_sem around
Oleg Nesterov999d9fc2008-07-25 01:47:41 -0700955 * checking core_state and changing tsk->mm.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700956 */
957 down_read(&old_mm->mmap_sem);
Oleg Nesterov999d9fc2008-07-25 01:47:41 -0700958 if (unlikely(old_mm->core_state)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700959 up_read(&old_mm->mmap_sem);
960 return -EINTR;
961 }
962 }
963 task_lock(tsk);
964 active_mm = tsk->active_mm;
965 tsk->mm = mm;
966 tsk->active_mm = mm;
967 activate_mm(active_mm, mm);
Davidlohr Bueso615d6e82014-04-07 15:37:25 -0700968 tsk->mm->vmacache_seqnum = 0;
969 vmacache_flush(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700970 task_unlock(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700971 if (old_mm) {
972 up_read(&old_mm->mmap_sem);
Eric Sesterhenn7dddb122006-04-01 01:13:38 +0200973 BUG_ON(active_mm != old_mm);
Oleg Nesterov701085b2012-03-19 17:04:01 +0100974 setmax_mm_hiwater_rss(&tsk->signal->maxrss, old_mm);
Balbir Singh31a78f22008-09-28 23:09:31 +0100975 mm_update_next_owner(old_mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700976 mmput(old_mm);
977 return 0;
978 }
979 mmdrop(active_mm);
980 return 0;
981}
982
983/*
984 * This function makes sure the current process has its own signal table,
985 * so that flush_signal_handlers can later reset the handlers without
986 * disturbing other processes. (Other processes might share the signal
987 * table via the CLONE_SIGHAND option to clone().)
988 */
Arjan van de Ven858119e2006-01-14 13:20:43 -0800989static int de_thread(struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990{
991 struct signal_struct *sig = tsk->signal;
Oleg Nesterovb2c903b2007-10-16 23:27:22 -0700992 struct sighand_struct *oldsighand = tsk->sighand;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700993 spinlock_t *lock = &oldsighand->siglock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -0700995 if (thread_group_empty(tsk))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996 goto no_thread_group;
997
998 /*
999 * Kill all other threads in the thread group.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001000 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001001 spin_lock_irq(lock);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -08001002 if (signal_group_exit(sig)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003 /*
1004 * Another group action in progress, just
1005 * return so that the signal is processed.
1006 */
1007 spin_unlock_irq(lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001008 return -EAGAIN;
1009 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010
Oleg Nesterovd3441932010-05-26 14:43:11 -07001011 sig->group_exit_task = tsk;
1012 sig->notify_count = zap_other_threads(tsk);
1013 if (!thread_group_leader(tsk))
1014 sig->notify_count--;
1015
1016 while (sig->notify_count) {
Oleg Nesterovd5bbd432012-10-08 19:13:01 +02001017 __set_current_state(TASK_KILLABLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001018 spin_unlock_irq(lock);
1019 schedule();
Oleg Nesterovd5bbd432012-10-08 19:13:01 +02001020 if (unlikely(__fatal_signal_pending(tsk)))
1021 goto killed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022 spin_lock_irq(lock);
1023 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001024 spin_unlock_irq(lock);
1025
1026 /*
1027 * At this point all other threads have exited, all we have to
1028 * do is to wait for the thread group leader to become inactive,
1029 * and to assume its PID:
1030 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001031 if (!thread_group_leader(tsk)) {
Oleg Nesterov81879262008-12-01 14:18:16 -08001032 struct task_struct *leader = tsk->group_leader;
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001033
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001034 for (;;) {
Oleg Nesterove56fb282013-04-30 15:28:20 -07001035 threadgroup_change_begin(tsk);
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001036 write_lock_irq(&tasklist_lock);
Kirill Tkhaidfcce792015-04-16 12:48:01 -07001037 /*
1038 * Do this under tasklist_lock to ensure that
1039 * exit_notify() can't miss ->group_exit_task
1040 */
1041 sig->notify_count = -1;
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001042 if (likely(leader->exit_state))
1043 break;
Oleg Nesterovd5bbd432012-10-08 19:13:01 +02001044 __set_current_state(TASK_KILLABLE);
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001045 write_unlock_irq(&tasklist_lock);
Oleg Nesterove56fb282013-04-30 15:28:20 -07001046 threadgroup_change_end(tsk);
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001047 schedule();
Oleg Nesterovd5bbd432012-10-08 19:13:01 +02001048 if (unlikely(__fatal_signal_pending(tsk)))
1049 goto killed;
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001050 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001051
Roland McGrathf5e90282006-04-10 22:54:16 -07001052 /*
1053 * The only record we have of the real-time age of a
1054 * process, regardless of execs it's done, is start_time.
1055 * All the past CPU time is accumulated in signal_struct
1056 * from sister threads now dead. But in this non-leader
1057 * exec, nothing survives from the original leader thread,
1058 * whose birth marks the true age of this process now.
1059 * When we take on its identity by switching to its PID, we
1060 * also take its birthdate (always earlier than our own).
1061 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001062 tsk->start_time = leader->start_time;
Oleg Nesterov266b7a02013-07-03 15:08:35 -07001063 tsk->real_start_time = leader->real_start_time;
Roland McGrathf5e90282006-04-10 22:54:16 -07001064
Pavel Emelyanovbac0abd2007-10-18 23:40:18 -07001065 BUG_ON(!same_thread_group(leader, tsk));
1066 BUG_ON(has_group_leader_pid(tsk));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001067 /*
1068 * An exec() starts a new thread group with the
1069 * TGID of the previous thread group. Rehash the
1070 * two threads with a switched PID, and release
1071 * the former thread group leader:
1072 */
Eric W. Biedermand73d6522006-03-28 16:11:03 -08001073
1074 /* Become a process group leader with the old leader's pid.
Eric W. Biedermanc18258c2006-09-27 01:51:06 -07001075 * The old leader becomes a thread of the this thread group.
1076 * Note: The old leader also uses this pid until release_task
Eric W. Biedermand73d6522006-03-28 16:11:03 -08001077 * is called. Odd but simple and correct.
1078 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001079 tsk->pid = leader->pid;
Oleg Nesterov3f418542013-07-03 15:08:25 -07001080 change_pid(tsk, PIDTYPE_PID, task_pid(leader));
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001081 transfer_pid(leader, tsk, PIDTYPE_PGID);
1082 transfer_pid(leader, tsk, PIDTYPE_SID);
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -08001083
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001084 list_replace_rcu(&leader->tasks, &tsk->tasks);
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -08001085 list_replace_init(&leader->sibling, &tsk->sibling);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001086
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001087 tsk->group_leader = tsk;
1088 leader->group_leader = tsk;
Eric W. Biedermande12a782006-04-10 17:16:49 -06001089
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001090 tsk->exit_signal = SIGCHLD;
Oleg Nesterov087806b2011-06-22 23:10:26 +02001091 leader->exit_signal = -1;
Oleg Nesterov962b5642005-11-23 13:37:43 -08001092
1093 BUG_ON(leader->exit_state != EXIT_ZOMBIE);
1094 leader->exit_state = EXIT_DEAD;
Oleg Nesteroveac1b5e2011-07-21 20:00:43 +02001095
1096 /*
1097 * We are going to release_task()->ptrace_unlink() silently,
1098 * the tracer can sleep in do_wait(). EXIT_DEAD guarantees
1099 * the tracer wont't block again waiting for this thread.
1100 */
1101 if (unlikely(leader->ptrace))
1102 __wake_up_parent(leader, leader->parent);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001103 write_unlock_irq(&tasklist_lock);
Oleg Nesterove56fb282013-04-30 15:28:20 -07001104 threadgroup_change_end(tsk);
Oleg Nesterov81879262008-12-01 14:18:16 -08001105
1106 release_task(leader);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -08001107 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001109 sig->group_exit_task = NULL;
1110 sig->notify_count = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001111
1112no_thread_group:
Oleg Nesterove6368252012-03-19 17:03:22 +01001113 /* we have changed execution domain */
1114 tsk->exit_signal = SIGCHLD;
1115
Linus Torvalds1da177e2005-04-16 15:20:36 -07001116 exit_itimers(sig);
Oleg Nesterovcbaffba2008-05-26 20:55:42 +04001117 flush_itimer_signals();
Oleg Nesterov329f7db2005-11-07 21:12:43 +03001118
Oleg Nesterovb2c903b2007-10-16 23:27:22 -07001119 if (atomic_read(&oldsighand->count) != 1) {
1120 struct sighand_struct *newsighand;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121 /*
Oleg Nesterovb2c903b2007-10-16 23:27:22 -07001122 * This ->sighand is shared with the CLONE_SIGHAND
1123 * but not CLONE_THREAD task, switch to the new one.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124 */
Oleg Nesterovb2c903b2007-10-16 23:27:22 -07001125 newsighand = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
1126 if (!newsighand)
1127 return -ENOMEM;
1128
Linus Torvalds1da177e2005-04-16 15:20:36 -07001129 atomic_set(&newsighand->count, 1);
1130 memcpy(newsighand->action, oldsighand->action,
1131 sizeof(newsighand->action));
1132
1133 write_lock_irq(&tasklist_lock);
1134 spin_lock(&oldsighand->siglock);
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001135 rcu_assign_pointer(tsk->sighand, newsighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001136 spin_unlock(&oldsighand->siglock);
1137 write_unlock_irq(&tasklist_lock);
1138
Davide Libenzifba2afa2007-05-10 22:23:13 -07001139 __cleanup_sighand(oldsighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001140 }
1141
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001142 BUG_ON(!thread_group_leader(tsk));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001143 return 0;
Oleg Nesterovd5bbd432012-10-08 19:13:01 +02001144
1145killed:
1146 /* protects against exit_notify() and __exit_signal() */
1147 read_lock(&tasklist_lock);
1148 sig->group_exit_task = NULL;
1149 sig->notify_count = 0;
1150 read_unlock(&tasklist_lock);
1151 return -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001152}
Oleg Nesterov0840a902007-10-16 23:27:22 -07001153
Andrew Morton59714d62008-02-04 22:27:21 -08001154char *get_task_comm(char *buf, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001155{
1156 /* buf must be at least sizeof(tsk->comm) in size */
1157 task_lock(tsk);
1158 strncpy(buf, tsk->comm, sizeof(tsk->comm));
1159 task_unlock(tsk);
Andrew Morton59714d62008-02-04 22:27:21 -08001160 return buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001161}
J Freyensee7d74f492011-05-06 16:56:47 -07001162EXPORT_SYMBOL_GPL(get_task_comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001163
Al Viro6a6d27d2012-08-21 09:56:33 -04001164/*
1165 * These functions flushes out all traces of the currently running executable
1166 * so that a new one can be started
1167 */
1168
Adrian Hunter82b89772014-05-28 11:45:04 +03001169void __set_task_comm(struct task_struct *tsk, const char *buf, bool exec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001170{
1171 task_lock(tsk);
KAMEZAWA Hiroyuki43d2b112012-01-10 15:08:09 -08001172 trace_task_rename(tsk, buf);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001173 strlcpy(tsk->comm, buf, sizeof(tsk->comm));
1174 task_unlock(tsk);
Adrian Hunter82b89772014-05-28 11:45:04 +03001175 perf_event_comm(tsk, exec);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001176}
1177
1178int flush_old_exec(struct linux_binprm * bprm)
1179{
Linus Torvalds221af7f2010-01-28 22:14:42 -08001180 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001181
1182 /*
1183 * Make sure we have a private signal table and that
1184 * we are unassociated from the previous thread group.
1185 */
1186 retval = de_thread(current);
1187 if (retval)
1188 goto out;
1189
Davidlohr Bueso6e399cd2015-04-16 12:47:59 -07001190 /*
1191 * Must be called _before_ exec_mmap() as bprm->mm is
1192 * not visibile until then. This also enables the update
1193 * to be lockless.
1194 */
Matt Helsley925d1c42008-04-29 01:01:36 -07001195 set_mm_exe_file(bprm->mm, bprm->file);
Davidlohr Bueso6e399cd2015-04-16 12:47:59 -07001196
Linus Torvalds1da177e2005-04-16 15:20:36 -07001197 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001198 * Release all of the old mmap stuff
1199 */
Oleg Nesterov3c77f842010-11-30 20:55:34 +01001200 acct_arg_size(bprm, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001201 retval = exec_mmap(bprm->mm);
1202 if (retval)
Al Virofd8328b2008-04-22 05:11:59 -04001203 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001204
1205 bprm->mm = NULL; /* We're using it now */
Linus Torvalds7ab02af2010-02-02 12:37:44 -08001206
Mathias Krausedac853a2011-06-09 20:05:18 +02001207 set_fs(USER_DS);
Zhang Yib88fae62014-01-23 15:55:57 -08001208 current->flags &= ~(PF_RANDOMIZE | PF_FORKNOEXEC | PF_KTHREAD |
1209 PF_NOFREEZE | PF_NO_SETAFFINITY);
Linus Torvalds7ab02af2010-02-02 12:37:44 -08001210 flush_thread();
1211 current->personality &= ~bprm->per_clear;
1212
Linus Torvalds221af7f2010-01-28 22:14:42 -08001213 return 0;
1214
1215out:
1216 return retval;
1217}
1218EXPORT_SYMBOL(flush_old_exec);
1219
Al Viro1b5d7832011-06-19 12:49:47 -04001220void would_dump(struct linux_binprm *bprm, struct file *file)
1221{
Al Viro496ad9a2013-01-23 17:07:38 -05001222 if (inode_permission(file_inode(file), MAY_READ) < 0)
Al Viro1b5d7832011-06-19 12:49:47 -04001223 bprm->interp_flags |= BINPRM_FLAGS_ENFORCE_NONDUMP;
1224}
1225EXPORT_SYMBOL(would_dump);
1226
Linus Torvalds221af7f2010-01-28 22:14:42 -08001227void setup_new_exec(struct linux_binprm * bprm)
1228{
Linus Torvalds221af7f2010-01-28 22:14:42 -08001229 arch_pick_mmap_layout(current->mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001230
1231 /* This is the point of no return */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001232 current->sas_ss_sp = current->sas_ss_size = 0;
1233
Eric W. Biederman8e96e3b2012-03-03 21:17:15 -08001234 if (uid_eq(current_euid(), current_uid()) && gid_eq(current_egid(), current_gid()))
Kees Cooke579d2c2013-02-27 17:03:15 -08001235 set_dumpable(current->mm, SUID_DUMP_USER);
Alan Coxd6e71142005-06-23 00:09:43 -07001236 else
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001237 set_dumpable(current->mm, suid_dumpable);
Alan Coxd6e71142005-06-23 00:09:43 -07001238
Peter Zijlstrae041e322014-05-21 17:32:19 +02001239 perf_event_exec();
Adrian Hunter82b89772014-05-28 11:45:04 +03001240 __set_task_comm(current, kbasename(bprm->filename), true);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001241
Benjamin Herrenschmidt0551fbd2006-02-28 16:59:19 -08001242 /* Set the new mm task size. We have to do that late because it may
1243 * depend on TIF_32BIT which is only updated in flush_thread() on
1244 * some architectures like powerpc
1245 */
1246 current->mm->task_size = TASK_SIZE;
1247
David Howellsa6f76f22008-11-14 10:39:24 +11001248 /* install the new credentials */
Eric W. Biederman8e96e3b2012-03-03 21:17:15 -08001249 if (!uid_eq(bprm->cred->uid, current_euid()) ||
1250 !gid_eq(bprm->cred->gid, current_egid())) {
Marcel Holtmannd2d56c52007-08-17 21:47:58 +02001251 current->pdeath_signal = 0;
Al Viro1b5d7832011-06-19 12:49:47 -04001252 } else {
1253 would_dump(bprm, bprm->file);
1254 if (bprm->interp_flags & BINPRM_FLAGS_ENFORCE_NONDUMP)
1255 set_dumpable(current->mm, suid_dumpable);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001256 }
1257
1258 /* An exec changes our domain. We are no longer part of the thread
1259 group */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001260 current->self_exec_id++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001261 flush_signal_handlers(current, 0);
Al Viro6a6d27d2012-08-21 09:56:33 -04001262 do_close_on_exec(current->files);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001263}
Linus Torvalds221af7f2010-01-28 22:14:42 -08001264EXPORT_SYMBOL(setup_new_exec);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001265
David Howellsa6f76f22008-11-14 10:39:24 +11001266/*
Oleg Nesterova2a84742009-09-05 11:17:13 -07001267 * Prepare credentials and lock ->cred_guard_mutex.
1268 * install_exec_creds() commits the new creds and drops the lock.
1269 * Or, if exec fails before, free_bprm() should release ->cred and
1270 * and unlock.
1271 */
1272int prepare_bprm_creds(struct linux_binprm *bprm)
1273{
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001274 if (mutex_lock_interruptible(&current->signal->cred_guard_mutex))
Oleg Nesterova2a84742009-09-05 11:17:13 -07001275 return -ERESTARTNOINTR;
1276
1277 bprm->cred = prepare_exec_creds();
1278 if (likely(bprm->cred))
1279 return 0;
1280
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001281 mutex_unlock(&current->signal->cred_guard_mutex);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001282 return -ENOMEM;
1283}
1284
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001285static void free_bprm(struct linux_binprm *bprm)
Oleg Nesterova2a84742009-09-05 11:17:13 -07001286{
1287 free_arg_pages(bprm);
1288 if (bprm->cred) {
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001289 mutex_unlock(&current->signal->cred_guard_mutex);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001290 abort_creds(bprm->cred);
1291 }
Oleg Nesterov63e46b92014-01-23 15:55:51 -08001292 if (bprm->file) {
1293 allow_write_access(bprm->file);
1294 fput(bprm->file);
1295 }
Kees Cookb66c5982012-12-20 15:05:16 -08001296 /* If a binfmt changed the interp, free it. */
1297 if (bprm->interp != bprm->filename)
1298 kfree(bprm->interp);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001299 kfree(bprm);
1300}
1301
Kees Cookb66c5982012-12-20 15:05:16 -08001302int bprm_change_interp(char *interp, struct linux_binprm *bprm)
1303{
1304 /* If a binfmt changed the interp, free it first. */
1305 if (bprm->interp != bprm->filename)
1306 kfree(bprm->interp);
1307 bprm->interp = kstrdup(interp, GFP_KERNEL);
1308 if (!bprm->interp)
1309 return -ENOMEM;
1310 return 0;
1311}
1312EXPORT_SYMBOL(bprm_change_interp);
1313
Oleg Nesterova2a84742009-09-05 11:17:13 -07001314/*
David Howellsa6f76f22008-11-14 10:39:24 +11001315 * install the new credentials for this executable
1316 */
1317void install_exec_creds(struct linux_binprm *bprm)
1318{
1319 security_bprm_committing_creds(bprm);
1320
1321 commit_creds(bprm->cred);
1322 bprm->cred = NULL;
Stephane Eranian2976b102013-06-20 11:36:28 +02001323
1324 /*
1325 * Disable monitoring for regular users
1326 * when executing setuid binaries. Must
1327 * wait until new credentials are committed
1328 * by commit_creds() above
1329 */
1330 if (get_dumpable(current->mm) != SUID_DUMP_USER)
1331 perf_event_exit_task(current);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001332 /*
1333 * cred_guard_mutex must be held at least to this point to prevent
David Howellsa6f76f22008-11-14 10:39:24 +11001334 * ptrace_attach() from altering our determination of the task's
Oleg Nesterova2a84742009-09-05 11:17:13 -07001335 * credentials; any time after this it may be unlocked.
1336 */
David Howellsa6f76f22008-11-14 10:39:24 +11001337 security_bprm_committed_creds(bprm);
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001338 mutex_unlock(&current->signal->cred_guard_mutex);
David Howellsa6f76f22008-11-14 10:39:24 +11001339}
1340EXPORT_SYMBOL(install_exec_creds);
1341
1342/*
1343 * determine how safe it is to execute the proposed program
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001344 * - the caller must hold ->cred_guard_mutex to protect against
Kees Cookc2e1f2e2014-06-05 00:23:17 -07001345 * PTRACE_ATTACH or seccomp thread-sync
David Howellsa6f76f22008-11-14 10:39:24 +11001346 */
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001347static void check_unsafe_exec(struct linux_binprm *bprm)
David Howellsa6f76f22008-11-14 10:39:24 +11001348{
David Howells0bf2f3a2009-02-06 11:45:46 +00001349 struct task_struct *p = current, *t;
Al Virof1191b52009-03-30 07:35:18 -04001350 unsigned n_fs;
David Howellsa6f76f22008-11-14 10:39:24 +11001351
Tejun Heo4b9d33e2011-06-17 16:50:38 +02001352 if (p->ptrace) {
1353 if (p->ptrace & PT_PTRACE_CAP)
1354 bprm->unsafe |= LSM_UNSAFE_PTRACE_CAP;
1355 else
1356 bprm->unsafe |= LSM_UNSAFE_PTRACE;
1357 }
David Howellsa6f76f22008-11-14 10:39:24 +11001358
Andy Lutomirski259e5e62012-04-12 16:47:50 -05001359 /*
1360 * This isn't strictly necessary, but it makes it harder for LSMs to
1361 * mess up.
1362 */
Kees Cook1d4457f2014-05-21 15:23:46 -07001363 if (task_no_new_privs(current))
Andy Lutomirski259e5e62012-04-12 16:47:50 -05001364 bprm->unsafe |= LSM_UNSAFE_NO_NEW_PRIVS;
1365
Oleg Nesterov83f62a22014-01-23 15:55:49 -08001366 t = p;
David Howells0bf2f3a2009-02-06 11:45:46 +00001367 n_fs = 1;
Nick Piggin2a4419b2010-08-18 04:37:33 +10001368 spin_lock(&p->fs->lock);
Oleg Nesterov437f7fd2009-04-24 01:02:45 +02001369 rcu_read_lock();
Oleg Nesterov83f62a22014-01-23 15:55:49 -08001370 while_each_thread(p, t) {
David Howells0bf2f3a2009-02-06 11:45:46 +00001371 if (t->fs == p->fs)
1372 n_fs++;
David Howells0bf2f3a2009-02-06 11:45:46 +00001373 }
Oleg Nesterov437f7fd2009-04-24 01:02:45 +02001374 rcu_read_unlock();
David Howells0bf2f3a2009-02-06 11:45:46 +00001375
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001376 if (p->fs->users > n_fs)
David Howellsa6f76f22008-11-14 10:39:24 +11001377 bprm->unsafe |= LSM_UNSAFE_SHARE;
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001378 else
1379 p->fs->in_exec = 1;
Nick Piggin2a4419b2010-08-18 04:37:33 +10001380 spin_unlock(&p->fs->lock);
David Howellsa6f76f22008-11-14 10:39:24 +11001381}
1382
Jann Horn8b01fc82015-04-19 02:48:39 +02001383static void bprm_fill_uid(struct linux_binprm *bprm)
1384{
1385 struct inode *inode;
1386 unsigned int mode;
1387 kuid_t uid;
1388 kgid_t gid;
1389
1390 /* clear any previous set[ug]id data from a previous binary */
1391 bprm->cred->euid = current_euid();
1392 bprm->cred->egid = current_egid();
1393
1394 if (bprm->file->f_path.mnt->mnt_flags & MNT_NOSUID)
1395 return;
1396
1397 if (task_no_new_privs(current))
1398 return;
1399
1400 inode = file_inode(bprm->file);
1401 mode = READ_ONCE(inode->i_mode);
1402 if (!(mode & (S_ISUID|S_ISGID)))
1403 return;
1404
1405 /* Be careful if suid/sgid is set */
Al Viro59551022016-01-22 15:40:57 -05001406 inode_lock(inode);
Jann Horn8b01fc82015-04-19 02:48:39 +02001407
1408 /* reload atomically mode/uid/gid now that lock held */
1409 mode = inode->i_mode;
1410 uid = inode->i_uid;
1411 gid = inode->i_gid;
Al Viro59551022016-01-22 15:40:57 -05001412 inode_unlock(inode);
Jann Horn8b01fc82015-04-19 02:48:39 +02001413
1414 /* We ignore suid/sgid if there are no mappings for them in the ns */
1415 if (!kuid_has_mapping(bprm->cred->user_ns, uid) ||
1416 !kgid_has_mapping(bprm->cred->user_ns, gid))
1417 return;
1418
1419 if (mode & S_ISUID) {
1420 bprm->per_clear |= PER_CLEAR_ON_SETID;
1421 bprm->cred->euid = uid;
1422 }
1423
1424 if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) {
1425 bprm->per_clear |= PER_CLEAR_ON_SETID;
1426 bprm->cred->egid = gid;
1427 }
1428}
1429
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001430/*
1431 * Fill the binprm structure from the inode.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001432 * Check permissions, then read the first 128 (BINPRM_BUF_SIZE) bytes
David Howellsa6f76f22008-11-14 10:39:24 +11001433 *
1434 * This may be called multiple times for binary chains (scripts for example).
Linus Torvalds1da177e2005-04-16 15:20:36 -07001435 */
1436int prepare_binprm(struct linux_binprm *bprm)
1437{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001438 int retval;
1439
Jann Horn8b01fc82015-04-19 02:48:39 +02001440 bprm_fill_uid(bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001441
1442 /* fill in binprm security blob */
David Howellsa6f76f22008-11-14 10:39:24 +11001443 retval = security_bprm_set_creds(bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001444 if (retval)
1445 return retval;
David Howellsa6f76f22008-11-14 10:39:24 +11001446 bprm->cred_prepared = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001447
David Howellsa6f76f22008-11-14 10:39:24 +11001448 memset(bprm->buf, 0, BINPRM_BUF_SIZE);
1449 return kernel_read(bprm->file, 0, bprm->buf, BINPRM_BUF_SIZE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001450}
1451
1452EXPORT_SYMBOL(prepare_binprm);
1453
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001454/*
1455 * Arguments are '\0' separated strings found at the location bprm->p
1456 * points to; chop off the first by relocating brpm->p to right after
1457 * the first '\0' encountered.
1458 */
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001459int remove_arg_zero(struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001460{
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001461 int ret = 0;
1462 unsigned long offset;
1463 char *kaddr;
1464 struct page *page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001465
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001466 if (!bprm->argc)
1467 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001468
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001469 do {
1470 offset = bprm->p & ~PAGE_MASK;
1471 page = get_arg_page(bprm, bprm->p, 0);
1472 if (!page) {
1473 ret = -EFAULT;
1474 goto out;
1475 }
Cong Wange8e3c3d2011-11-25 23:14:27 +08001476 kaddr = kmap_atomic(page);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001477
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001478 for (; offset < PAGE_SIZE && kaddr[offset];
1479 offset++, bprm->p++)
1480 ;
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001481
Cong Wange8e3c3d2011-11-25 23:14:27 +08001482 kunmap_atomic(kaddr);
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001483 put_arg_page(page);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001484
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001485 if (offset == PAGE_SIZE)
1486 free_arg_page(bprm, (bprm->p >> PAGE_SHIFT) - 1);
1487 } while (offset == PAGE_SIZE);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001488
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001489 bprm->p++;
1490 bprm->argc--;
1491 ret = 0;
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001492
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001493out:
1494 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001495}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001496EXPORT_SYMBOL(remove_arg_zero);
1497
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001498#define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001499/*
1500 * cycle the list of binary formats handler, until one recognizes the image
1501 */
Al Viro3c456bf2012-10-20 21:53:31 -04001502int search_binary_handler(struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001503{
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001504 bool need_retry = IS_ENABLED(CONFIG_MODULES);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001505 struct linux_binfmt *fmt;
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001506 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001507
Kees Cookd7402692012-12-17 16:03:20 -08001508 /* This allows 4 levels of binfmt rewrites before failing hard. */
Oleg Nesterov131b2f92013-09-11 14:24:39 -07001509 if (bprm->recursion_depth > 5)
Kees Cookd7402692012-12-17 16:03:20 -08001510 return -ELOOP;
1511
Linus Torvalds1da177e2005-04-16 15:20:36 -07001512 retval = security_bprm_check(bprm);
1513 if (retval)
1514 return retval;
1515
Linus Torvalds1da177e2005-04-16 15:20:36 -07001516 retval = -ENOENT;
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001517 retry:
1518 read_lock(&binfmt_lock);
1519 list_for_each_entry(fmt, &formats, lh) {
1520 if (!try_module_get(fmt->module))
1521 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001522 read_unlock(&binfmt_lock);
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001523 bprm->recursion_depth++;
1524 retval = fmt->load_binary(bprm);
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001525 read_lock(&binfmt_lock);
1526 put_binfmt(fmt);
Al Viro19d860a2014-05-04 20:11:36 -04001527 bprm->recursion_depth--;
1528 if (retval < 0 && !bprm->mm) {
1529 /* we got to flush_old_exec() and failed after it */
1530 read_unlock(&binfmt_lock);
1531 force_sigsegv(SIGSEGV, current);
1532 return retval;
1533 }
1534 if (retval != -ENOEXEC || !bprm->file) {
1535 read_unlock(&binfmt_lock);
1536 return retval;
1537 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001538 }
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001539 read_unlock(&binfmt_lock);
1540
Al Viro19d860a2014-05-04 20:11:36 -04001541 if (need_retry) {
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001542 if (printable(bprm->buf[0]) && printable(bprm->buf[1]) &&
1543 printable(bprm->buf[2]) && printable(bprm->buf[3]))
1544 return retval;
Oleg Nesterov4e0621a2013-09-11 14:24:45 -07001545 if (request_module("binfmt-%04x", *(ushort *)(bprm->buf + 2)) < 0)
1546 return retval;
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001547 need_retry = false;
1548 goto retry;
1549 }
1550
Linus Torvalds1da177e2005-04-16 15:20:36 -07001551 return retval;
1552}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001553EXPORT_SYMBOL(search_binary_handler);
1554
Oleg Nesterov5d1baf32013-09-11 14:24:38 -07001555static int exec_binprm(struct linux_binprm *bprm)
1556{
1557 pid_t old_pid, old_vpid;
1558 int ret;
1559
1560 /* Need to fetch pid before load_binary changes it */
1561 old_pid = current->pid;
1562 rcu_read_lock();
1563 old_vpid = task_pid_nr_ns(current, task_active_pid_ns(current->parent));
1564 rcu_read_unlock();
1565
1566 ret = search_binary_handler(bprm);
1567 if (ret >= 0) {
Linus Torvalds3eaded862013-11-21 19:18:14 -08001568 audit_bprm(bprm);
Oleg Nesterov5d1baf32013-09-11 14:24:38 -07001569 trace_sched_process_exec(current, old_pid, bprm);
1570 ptrace_event(PTRACE_EVENT_EXEC, old_vpid);
Oleg Nesterov9beb2662013-09-11 14:24:40 -07001571 proc_exec_connector(current);
Oleg Nesterov5d1baf32013-09-11 14:24:38 -07001572 }
1573
1574 return ret;
1575}
1576
Linus Torvalds1da177e2005-04-16 15:20:36 -07001577/*
1578 * sys_execve() executes a new program.
1579 */
David Drysdale51f39a12014-12-12 16:57:29 -08001580static int do_execveat_common(int fd, struct filename *filename,
1581 struct user_arg_ptr argv,
1582 struct user_arg_ptr envp,
1583 int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001584{
David Drysdale51f39a12014-12-12 16:57:29 -08001585 char *pathbuf = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001586 struct linux_binprm *bprm;
1587 struct file *file;
Al Viro3b125382008-04-22 05:31:30 -04001588 struct files_struct *displaced;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001589 int retval;
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +04001590
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001591 if (IS_ERR(filename))
1592 return PTR_ERR(filename);
1593
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +04001594 /*
1595 * We move the actual failure in case of RLIMIT_NPROC excess from
1596 * set*uid() to execve() because too many poorly written programs
1597 * don't check setuid() return code. Here we additionally recheck
1598 * whether NPROC limit is still exceeded.
1599 */
1600 if ((current->flags & PF_NPROC_EXCEEDED) &&
Oleg Nesterovbd9d43f2013-07-03 15:08:34 -07001601 atomic_read(&current_user()->processes) > rlimit(RLIMIT_NPROC)) {
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +04001602 retval = -EAGAIN;
1603 goto out_ret;
1604 }
1605
1606 /* We're below the limit (still or again), so we don't want to make
1607 * further execve() calls fail. */
1608 current->flags &= ~PF_NPROC_EXCEEDED;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001609
Al Viro3b125382008-04-22 05:31:30 -04001610 retval = unshare_files(&displaced);
Al Virofd8328b2008-04-22 05:11:59 -04001611 if (retval)
1612 goto out_ret;
1613
Linus Torvalds1da177e2005-04-16 15:20:36 -07001614 retval = -ENOMEM;
Oliver Neukum11b0b5a2006-03-25 03:08:13 -08001615 bprm = kzalloc(sizeof(*bprm), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001616 if (!bprm)
Al Virofd8328b2008-04-22 05:11:59 -04001617 goto out_files;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001618
Oleg Nesterova2a84742009-09-05 11:17:13 -07001619 retval = prepare_bprm_creds(bprm);
1620 if (retval)
David Howellsa6f76f22008-11-14 10:39:24 +11001621 goto out_free;
Al Viro498052b2009-03-30 07:20:30 -04001622
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001623 check_unsafe_exec(bprm);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001624 current->in_execve = 1;
David Howellsa6f76f22008-11-14 10:39:24 +11001625
David Drysdale51f39a12014-12-12 16:57:29 -08001626 file = do_open_execat(fd, filename, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001627 retval = PTR_ERR(file);
1628 if (IS_ERR(file))
Al Viro498052b2009-03-30 07:20:30 -04001629 goto out_unmark;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001630
1631 sched_exec();
1632
Linus Torvalds1da177e2005-04-16 15:20:36 -07001633 bprm->file = file;
David Drysdale51f39a12014-12-12 16:57:29 -08001634 if (fd == AT_FDCWD || filename->name[0] == '/') {
1635 bprm->filename = filename->name;
1636 } else {
1637 if (filename->name[0] == '\0')
1638 pathbuf = kasprintf(GFP_TEMPORARY, "/dev/fd/%d", fd);
1639 else
1640 pathbuf = kasprintf(GFP_TEMPORARY, "/dev/fd/%d/%s",
1641 fd, filename->name);
1642 if (!pathbuf) {
1643 retval = -ENOMEM;
1644 goto out_unmark;
1645 }
1646 /*
1647 * Record that a name derived from an O_CLOEXEC fd will be
1648 * inaccessible after exec. Relies on having exclusive access to
1649 * current->files (due to unshare_files above).
1650 */
1651 if (close_on_exec(fd, rcu_dereference_raw(current->files->fdt)))
1652 bprm->interp_flags |= BINPRM_FLAGS_PATH_INACCESSIBLE;
1653 bprm->filename = pathbuf;
1654 }
1655 bprm->interp = bprm->filename;
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001656
1657 retval = bprm_mm_init(bprm);
1658 if (retval)
Oleg Nesterov63e46b92014-01-23 15:55:51 -08001659 goto out_unmark;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001660
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001661 bprm->argc = count(argv, MAX_ARG_STRINGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001662 if ((retval = bprm->argc) < 0)
David Howellsa6f76f22008-11-14 10:39:24 +11001663 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001664
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001665 bprm->envc = count(envp, MAX_ARG_STRINGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001666 if ((retval = bprm->envc) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001667 goto out;
1668
1669 retval = prepare_binprm(bprm);
1670 if (retval < 0)
1671 goto out;
1672
1673 retval = copy_strings_kernel(1, &bprm->filename, bprm);
1674 if (retval < 0)
1675 goto out;
1676
1677 bprm->exec = bprm->p;
1678 retval = copy_strings(bprm->envc, envp, bprm);
1679 if (retval < 0)
1680 goto out;
1681
1682 retval = copy_strings(bprm->argc, argv, bprm);
1683 if (retval < 0)
1684 goto out;
1685
Oleg Nesterov5d1baf32013-09-11 14:24:38 -07001686 retval = exec_binprm(bprm);
David Howellsa6f76f22008-11-14 10:39:24 +11001687 if (retval < 0)
1688 goto out;
1689
1690 /* execve succeeded */
Al Viro498052b2009-03-30 07:20:30 -04001691 current->fs->in_exec = 0;
Kentaro Takedaf9ce1f12009-02-05 17:18:11 +09001692 current->in_execve = 0;
David Howellsa6f76f22008-11-14 10:39:24 +11001693 acct_update_integrals(current);
Rik van Riel82727012013-10-07 11:29:28 +01001694 task_numa_free(current);
David Howellsa6f76f22008-11-14 10:39:24 +11001695 free_bprm(bprm);
David Drysdale51f39a12014-12-12 16:57:29 -08001696 kfree(pathbuf);
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001697 putname(filename);
David Howellsa6f76f22008-11-14 10:39:24 +11001698 if (displaced)
1699 put_files_struct(displaced);
1700 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001701
1702out:
Oleg Nesterov3c77f842010-11-30 20:55:34 +01001703 if (bprm->mm) {
1704 acct_arg_size(bprm, 0);
1705 mmput(bprm->mm);
1706 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001707
Al Viro498052b2009-03-30 07:20:30 -04001708out_unmark:
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001709 current->fs->in_exec = 0;
Kentaro Takedaf9ce1f12009-02-05 17:18:11 +09001710 current->in_execve = 0;
David Howellsa6f76f22008-11-14 10:39:24 +11001711
1712out_free:
Al Viro08a6fac2008-05-10 16:38:25 -04001713 free_bprm(bprm);
David Drysdale51f39a12014-12-12 16:57:29 -08001714 kfree(pathbuf);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001715
Al Virofd8328b2008-04-22 05:11:59 -04001716out_files:
Al Viro3b125382008-04-22 05:31:30 -04001717 if (displaced)
1718 reset_files_struct(displaced);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001719out_ret:
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001720 putname(filename);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001721 return retval;
1722}
1723
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001724int do_execve(struct filename *filename,
Oleg Nesterovba2d01622011-03-06 18:02:37 +01001725 const char __user *const __user *__argv,
Al Viroda3d4c52012-10-20 21:49:33 -04001726 const char __user *const __user *__envp)
Oleg Nesterovba2d01622011-03-06 18:02:37 +01001727{
Oleg Nesterov0e028462011-03-06 18:02:54 +01001728 struct user_arg_ptr argv = { .ptr.native = __argv };
1729 struct user_arg_ptr envp = { .ptr.native = __envp };
David Drysdale51f39a12014-12-12 16:57:29 -08001730 return do_execveat_common(AT_FDCWD, filename, argv, envp, 0);
1731}
1732
1733int do_execveat(int fd, struct filename *filename,
1734 const char __user *const __user *__argv,
1735 const char __user *const __user *__envp,
1736 int flags)
1737{
1738 struct user_arg_ptr argv = { .ptr.native = __argv };
1739 struct user_arg_ptr envp = { .ptr.native = __envp };
1740
1741 return do_execveat_common(fd, filename, argv, envp, flags);
Oleg Nesterovba2d01622011-03-06 18:02:37 +01001742}
1743
Oleg Nesterov0e028462011-03-06 18:02:54 +01001744#ifdef CONFIG_COMPAT
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001745static int compat_do_execve(struct filename *filename,
Al Viro38b983b2012-09-30 13:38:55 -04001746 const compat_uptr_t __user *__argv,
Al Virod03d26e2012-10-20 21:46:25 -04001747 const compat_uptr_t __user *__envp)
Oleg Nesterov0e028462011-03-06 18:02:54 +01001748{
1749 struct user_arg_ptr argv = {
1750 .is_compat = true,
1751 .ptr.compat = __argv,
1752 };
1753 struct user_arg_ptr envp = {
1754 .is_compat = true,
1755 .ptr.compat = __envp,
1756 };
David Drysdale51f39a12014-12-12 16:57:29 -08001757 return do_execveat_common(AT_FDCWD, filename, argv, envp, 0);
1758}
1759
1760static int compat_do_execveat(int fd, struct filename *filename,
1761 const compat_uptr_t __user *__argv,
1762 const compat_uptr_t __user *__envp,
1763 int flags)
1764{
1765 struct user_arg_ptr argv = {
1766 .is_compat = true,
1767 .ptr.compat = __argv,
1768 };
1769 struct user_arg_ptr envp = {
1770 .is_compat = true,
1771 .ptr.compat = __envp,
1772 };
1773 return do_execveat_common(fd, filename, argv, envp, flags);
Oleg Nesterov0e028462011-03-06 18:02:54 +01001774}
1775#endif
1776
Oleg Nesterov964ee7d2009-09-23 15:56:59 -07001777void set_binfmt(struct linux_binfmt *new)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001778{
Hiroshi Shimamoto801460d2009-09-23 15:57:41 -07001779 struct mm_struct *mm = current->mm;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001780
Hiroshi Shimamoto801460d2009-09-23 15:57:41 -07001781 if (mm->binfmt)
1782 module_put(mm->binfmt->module);
1783
1784 mm->binfmt = new;
Oleg Nesterov964ee7d2009-09-23 15:56:59 -07001785 if (new)
1786 __module_get(new->module);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001787}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001788EXPORT_SYMBOL(set_binfmt);
1789
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001790/*
Oleg Nesterov7288e112014-01-23 15:55:32 -08001791 * set_dumpable stores three-value SUID_DUMP_* into mm->flags.
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001792 */
1793void set_dumpable(struct mm_struct *mm, int value)
1794{
Oleg Nesterovabacd2f2014-01-23 15:55:31 -08001795 unsigned long old, new;
1796
Oleg Nesterov7288e112014-01-23 15:55:32 -08001797 if (WARN_ON((unsigned)value > SUID_DUMP_ROOT))
1798 return;
1799
Oleg Nesterovabacd2f2014-01-23 15:55:31 -08001800 do {
1801 old = ACCESS_ONCE(mm->flags);
Oleg Nesterov7288e112014-01-23 15:55:32 -08001802 new = (old & ~MMF_DUMPABLE_MASK) | value;
Oleg Nesterovabacd2f2014-01-23 15:55:31 -08001803 } while (cmpxchg(&mm->flags, old, new) != old);
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001804}
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001805
Al Viro38b983b2012-09-30 13:38:55 -04001806SYSCALL_DEFINE3(execve,
1807 const char __user *, filename,
1808 const char __user *const __user *, argv,
1809 const char __user *const __user *, envp)
1810{
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001811 return do_execve(getname(filename), argv, envp);
Al Viro38b983b2012-09-30 13:38:55 -04001812}
David Drysdale51f39a12014-12-12 16:57:29 -08001813
1814SYSCALL_DEFINE5(execveat,
1815 int, fd, const char __user *, filename,
1816 const char __user *const __user *, argv,
1817 const char __user *const __user *, envp,
1818 int, flags)
1819{
1820 int lookup_flags = (flags & AT_EMPTY_PATH) ? LOOKUP_EMPTY : 0;
1821
1822 return do_execveat(fd,
1823 getname_flags(filename, lookup_flags, NULL),
1824 argv, envp, flags);
1825}
1826
Al Viro38b983b2012-09-30 13:38:55 -04001827#ifdef CONFIG_COMPAT
Heiko Carstens625b1d72014-03-04 10:53:50 +01001828COMPAT_SYSCALL_DEFINE3(execve, const char __user *, filename,
1829 const compat_uptr_t __user *, argv,
1830 const compat_uptr_t __user *, envp)
Al Viro38b983b2012-09-30 13:38:55 -04001831{
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001832 return compat_do_execve(getname(filename), argv, envp);
Al Viro38b983b2012-09-30 13:38:55 -04001833}
David Drysdale51f39a12014-12-12 16:57:29 -08001834
1835COMPAT_SYSCALL_DEFINE5(execveat, int, fd,
1836 const char __user *, filename,
1837 const compat_uptr_t __user *, argv,
1838 const compat_uptr_t __user *, envp,
1839 int, flags)
1840{
1841 int lookup_flags = (flags & AT_EMPTY_PATH) ? LOOKUP_EMPTY : 0;
1842
1843 return compat_do_execveat(fd,
1844 getname_flags(filename, lookup_flags, NULL),
1845 argv, envp, flags);
1846}
Al Viro38b983b2012-09-30 13:38:55 -04001847#endif