blob: 91441402d706025971657e088c8ea3d937270431 [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
Aleksa Saraic1df5a62016-12-21 16:26:24 +110022 * formats.
Linus Torvalds1da177e2005-04-16 15:20:36 -070023 */
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;
Lorenzo Stoakes9beae1e2016-10-13 01:20:17 +0100194 unsigned int gup_flags = FOLL_FORCE;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700195
196#ifdef CONFIG_STACK_GROWSUP
197 if (write) {
Michal Hockod05f3162011-05-24 17:11:44 -0700198 ret = expand_downwards(bprm->vma, pos);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700199 if (ret < 0)
200 return NULL;
201 }
202#endif
Lorenzo Stoakes9beae1e2016-10-13 01:20:17 +0100203
204 if (write)
205 gup_flags |= FOLL_WRITE;
206
Dave Hansen1e987792016-02-12 13:01:54 -0800207 /*
208 * We are doing an exec(). 'current' is the process
209 * doing the exec and bprm->mm is the new process's mm.
210 */
Lorenzo Stoakes9beae1e2016-10-13 01:20:17 +0100211 ret = get_user_pages_remote(current, bprm->mm, pos, 1, gup_flags,
212 &page, NULL);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700213 if (ret <= 0)
214 return NULL;
215
216 if (write) {
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700217 unsigned long size = bprm->vma->vm_end - bprm->vma->vm_start;
Kees Cook3d6848e2017-06-23 15:08:57 -0700218 unsigned long ptr_size;
Linus Torvaldsa64e7152008-03-03 10:12:14 -0800219 struct rlimit *rlim;
220
Kees Cook3d6848e2017-06-23 15:08:57 -0700221 /*
222 * Since the stack will hold pointers to the strings, we
223 * must account for them as well.
224 *
225 * The size calculation is the entire vma while each arg page is
226 * built, so each time we get here it's calculating how far it
227 * is currently (rather than each call being just the newly
228 * added size from the arg page). As a result, we need to
229 * always add the entire size of the pointers, so that on the
230 * last call to get_arg_page() we'll actually have the entire
231 * correct size.
232 */
233 ptr_size = (bprm->argc + bprm->envc) * sizeof(void *);
234 if (ptr_size > ULONG_MAX - size)
235 goto fail;
236 size += ptr_size;
237
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100238 acct_arg_size(bprm, size / PAGE_SIZE);
239
Linus Torvaldsa64e7152008-03-03 10:12:14 -0800240 /*
241 * We've historically supported up to 32 pages (ARG_MAX)
242 * of argument strings even with small stacks
243 */
244 if (size <= ARG_MAX)
245 return page;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700246
247 /*
248 * Limit to 1/4-th the stack size for the argv+env strings.
249 * This ensures that:
250 * - the remaining binfmt code will not run out of stack space,
251 * - the program will have a reasonable amount of stack left
252 * to work from.
253 */
Linus Torvaldsa64e7152008-03-03 10:12:14 -0800254 rlim = current->signal->rlim;
Kees Cook3d6848e2017-06-23 15:08:57 -0700255 if (size > READ_ONCE(rlim[RLIMIT_STACK].rlim_cur) / 4)
256 goto fail;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700257 }
258
259 return page;
Kees Cook3d6848e2017-06-23 15:08:57 -0700260
261fail:
262 put_page(page);
263 return NULL;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700264}
265
266static void put_arg_page(struct page *page)
267{
268 put_page(page);
269}
270
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700271static void free_arg_pages(struct linux_binprm *bprm)
272{
273}
274
275static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos,
276 struct page *page)
277{
278 flush_cache_page(bprm->vma, pos, page_to_pfn(page));
279}
280
281static int __bprm_mm_init(struct linux_binprm *bprm)
282{
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800283 int err;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700284 struct vm_area_struct *vma = NULL;
285 struct mm_struct *mm = bprm->mm;
286
287 bprm->vma = vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL);
288 if (!vma)
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800289 return -ENOMEM;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700290
Michal Hockof268dfe2016-05-23 16:26:02 -0700291 if (down_write_killable(&mm->mmap_sem)) {
292 err = -EINTR;
293 goto err_free;
294 }
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700295 vma->vm_mm = mm;
296
297 /*
298 * Place the stack at the largest stack address the architecture
299 * supports. Later, we'll move this to an appropriate place. We don't
300 * use STACK_TOP because that can depend on attributes which aren't
301 * configured yet.
302 */
Michal Hockoaacb3d12011-07-26 16:08:40 -0700303 BUILD_BUG_ON(VM_STACK_FLAGS & VM_STACK_INCOMPLETE_SETUP);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700304 vma->vm_end = STACK_TOP_MAX;
305 vma->vm_start = vma->vm_end - PAGE_SIZE;
Cyrill Gorcunovd9104d12013-09-11 14:22:24 -0700306 vma->vm_flags = VM_SOFTDIRTY | VM_STACK_FLAGS | VM_STACK_INCOMPLETE_SETUP;
Coly Li3ed75eb2007-10-18 23:39:15 -0700307 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
Rik van Riel5beb4932010-03-05 13:42:07 -0800308 INIT_LIST_HEAD(&vma->anon_vma_chain);
Tavis Ormandy462e635e2010-12-09 15:29:42 +0100309
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700310 err = insert_vm_struct(mm, vma);
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800311 if (err)
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700312 goto err;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700313
314 mm->stack_vm = mm->total_vm = 1;
Dave Hansenfe3d1972014-11-14 07:18:29 -0800315 arch_bprm_mm_init(mm, vma);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700316 up_write(&mm->mmap_sem);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700317 bprm->p = vma->vm_end - sizeof(void *);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700318 return 0;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700319err:
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800320 up_write(&mm->mmap_sem);
Michal Hockof268dfe2016-05-23 16:26:02 -0700321err_free:
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800322 bprm->vma = NULL;
323 kmem_cache_free(vm_area_cachep, vma);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700324 return err;
325}
326
327static bool valid_arg_len(struct linux_binprm *bprm, long len)
328{
329 return len <= MAX_ARG_STRLEN;
330}
331
332#else
333
Oleg Nesterov0e028462011-03-06 18:02:54 +0100334static inline void acct_arg_size(struct linux_binprm *bprm, unsigned long pages)
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100335{
336}
337
Oleg Nesterov0e028462011-03-06 18:02:54 +0100338static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos,
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700339 int write)
340{
341 struct page *page;
342
343 page = bprm->page[pos / PAGE_SIZE];
344 if (!page && write) {
345 page = alloc_page(GFP_HIGHUSER|__GFP_ZERO);
346 if (!page)
347 return NULL;
348 bprm->page[pos / PAGE_SIZE] = page;
349 }
350
351 return page;
352}
353
354static void put_arg_page(struct page *page)
355{
356}
357
358static void free_arg_page(struct linux_binprm *bprm, int i)
359{
360 if (bprm->page[i]) {
361 __free_page(bprm->page[i]);
362 bprm->page[i] = NULL;
363 }
364}
365
366static void free_arg_pages(struct linux_binprm *bprm)
367{
368 int i;
369
370 for (i = 0; i < MAX_ARG_PAGES; i++)
371 free_arg_page(bprm, i);
372}
373
374static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos,
375 struct page *page)
376{
377}
378
379static int __bprm_mm_init(struct linux_binprm *bprm)
380{
381 bprm->p = PAGE_SIZE * MAX_ARG_PAGES - sizeof(void *);
382 return 0;
383}
384
385static bool valid_arg_len(struct linux_binprm *bprm, long len)
386{
387 return len <= bprm->p;
388}
389
390#endif /* CONFIG_MMU */
391
392/*
393 * Create a new mm_struct and populate it with a temporary stack
394 * vm_area_struct. We don't have enough context at this point to set the stack
395 * flags, permissions, and offset, so we use temporary values. We'll update
396 * them later in setup_arg_pages().
397 */
Yuanhan Liu9cc64ce2013-02-20 13:16:01 +1100398static int bprm_mm_init(struct linux_binprm *bprm)
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700399{
400 int err;
401 struct mm_struct *mm = NULL;
402
403 bprm->mm = mm = mm_alloc();
404 err = -ENOMEM;
405 if (!mm)
406 goto err;
407
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700408 err = __bprm_mm_init(bprm);
409 if (err)
410 goto err;
411
412 return 0;
413
414err:
415 if (mm) {
416 bprm->mm = NULL;
417 mmdrop(mm);
418 }
419
420 return err;
421}
422
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100423struct user_arg_ptr {
Oleg Nesterov0e028462011-03-06 18:02:54 +0100424#ifdef CONFIG_COMPAT
425 bool is_compat;
426#endif
427 union {
428 const char __user *const __user *native;
429#ifdef CONFIG_COMPAT
Al Viro38b983b2012-09-30 13:38:55 -0400430 const compat_uptr_t __user *compat;
Oleg Nesterov0e028462011-03-06 18:02:54 +0100431#endif
432 } ptr;
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100433};
434
435static const char __user *get_user_arg_ptr(struct user_arg_ptr argv, int nr)
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100436{
Oleg Nesterov0e028462011-03-06 18:02:54 +0100437 const char __user *native;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100438
Oleg Nesterov0e028462011-03-06 18:02:54 +0100439#ifdef CONFIG_COMPAT
440 if (unlikely(argv.is_compat)) {
441 compat_uptr_t compat;
442
443 if (get_user(compat, argv.ptr.compat + nr))
444 return ERR_PTR(-EFAULT);
445
446 return compat_ptr(compat);
447 }
448#endif
449
450 if (get_user(native, argv.ptr.native + nr))
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100451 return ERR_PTR(-EFAULT);
452
Oleg Nesterov0e028462011-03-06 18:02:54 +0100453 return native;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100454}
455
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456/*
457 * count() counts the number of strings in array ARGV.
458 */
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100459static int count(struct user_arg_ptr argv, int max)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460{
461 int i = 0;
462
Oleg Nesterov0e028462011-03-06 18:02:54 +0100463 if (argv.ptr.native != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464 for (;;) {
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100465 const char __user *p = get_user_arg_ptr(argv, i);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467 if (!p)
468 break;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100469
470 if (IS_ERR(p))
471 return -EFAULT;
472
Xi Wang6d92d4f2013-01-11 14:31:48 -0800473 if (i >= max)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474 return -E2BIG;
Xi Wang6d92d4f2013-01-11 14:31:48 -0800475 ++i;
Roland McGrath9aea5a62010-09-07 19:37:06 -0700476
477 if (fatal_signal_pending(current))
478 return -ERESTARTNOHAND;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700479 cond_resched();
480 }
481 }
482 return i;
483}
484
485/*
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700486 * 'copy_strings()' copies argument/environment strings from the old
487 * processes's memory to the new process's stack. The call to get_user_pages()
488 * ensures the destination page is created and not swapped out.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489 */
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100490static int copy_strings(int argc, struct user_arg_ptr argv,
Adrian Bunk75c96f82005-05-05 16:16:09 -0700491 struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492{
493 struct page *kmapped_page = NULL;
494 char *kaddr = NULL;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700495 unsigned long kpos = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496 int ret;
497
498 while (argc-- > 0) {
David Howellsd7627462010-08-17 23:52:56 +0100499 const char __user *str;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700500 int len;
501 unsigned long pos;
502
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100503 ret = -EFAULT;
504 str = get_user_arg_ptr(argv, argc);
505 if (IS_ERR(str))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100508 len = strnlen_user(str, MAX_ARG_STRLEN);
509 if (!len)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510 goto out;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100511
512 ret = -E2BIG;
513 if (!valid_arg_len(bprm, len))
514 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700516 /* We're going to work our way backwords. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517 pos = bprm->p;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700518 str += len;
519 bprm->p -= len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700520
521 while (len > 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522 int offset, bytes_to_copy;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523
Roland McGrath9aea5a62010-09-07 19:37:06 -0700524 if (fatal_signal_pending(current)) {
525 ret = -ERESTARTNOHAND;
526 goto out;
527 }
Roland McGrath7993bc12010-09-07 19:36:28 -0700528 cond_resched();
529
Linus Torvalds1da177e2005-04-16 15:20:36 -0700530 offset = pos % PAGE_SIZE;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700531 if (offset == 0)
532 offset = PAGE_SIZE;
533
534 bytes_to_copy = offset;
535 if (bytes_to_copy > len)
536 bytes_to_copy = len;
537
538 offset -= bytes_to_copy;
539 pos -= bytes_to_copy;
540 str -= bytes_to_copy;
541 len -= bytes_to_copy;
542
543 if (!kmapped_page || kpos != (pos & PAGE_MASK)) {
544 struct page *page;
545
546 page = get_arg_page(bprm, pos, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547 if (!page) {
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700548 ret = -E2BIG;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549 goto out;
550 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700552 if (kmapped_page) {
553 flush_kernel_dcache_page(kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554 kunmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700555 put_arg_page(kmapped_page);
556 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557 kmapped_page = page;
558 kaddr = kmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700559 kpos = pos & PAGE_MASK;
560 flush_arg_page(bprm, kpos, kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700561 }
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700562 if (copy_from_user(kaddr+offset, str, bytes_to_copy)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700563 ret = -EFAULT;
564 goto out;
565 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566 }
567 }
568 ret = 0;
569out:
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700570 if (kmapped_page) {
571 flush_kernel_dcache_page(kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700572 kunmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700573 put_arg_page(kmapped_page);
574 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700575 return ret;
576}
577
578/*
579 * Like copy_strings, but get argv and its values from kernel memory.
580 */
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100581int copy_strings_kernel(int argc, const char *const *__argv,
David Howellsd7627462010-08-17 23:52:56 +0100582 struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583{
584 int r;
585 mm_segment_t oldfs = get_fs();
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100586 struct user_arg_ptr argv = {
Oleg Nesterov0e028462011-03-06 18:02:54 +0100587 .ptr.native = (const char __user *const __user *)__argv,
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100588 };
589
Linus Torvalds1da177e2005-04-16 15:20:36 -0700590 set_fs(KERNEL_DS);
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100591 r = copy_strings(argc, argv, bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700592 set_fs(oldfs);
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100593
Linus Torvalds1da177e2005-04-16 15:20:36 -0700594 return r;
595}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596EXPORT_SYMBOL(copy_strings_kernel);
597
598#ifdef CONFIG_MMU
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700599
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600/*
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700601 * During bprm_mm_init(), we create a temporary stack at STACK_TOP_MAX. Once
602 * the binfmt code determines where the new stack should reside, we shift it to
603 * its final location. The process proceeds as follows:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604 *
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700605 * 1) Use shift to calculate the new vma endpoints.
606 * 2) Extend vma to cover both the old and new ranges. This ensures the
607 * arguments passed to subsequent functions are consistent.
608 * 3) Move vma's page tables to the new range.
609 * 4) Free up any cleared pgd range.
610 * 5) Shrink the vma to cover only the new range.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700611 */
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700612static int shift_arg_pages(struct vm_area_struct *vma, unsigned long shift)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700613{
614 struct mm_struct *mm = vma->vm_mm;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700615 unsigned long old_start = vma->vm_start;
616 unsigned long old_end = vma->vm_end;
617 unsigned long length = old_end - old_start;
618 unsigned long new_start = old_start - shift;
619 unsigned long new_end = old_end - shift;
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700620 struct mmu_gather tlb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700621
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700622 BUG_ON(new_start > new_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700623
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700624 /*
625 * ensure there are no vmas between where we want to go
626 * and where we are
627 */
628 if (vma != find_vma(mm, new_start))
629 return -EFAULT;
630
631 /*
632 * cover the whole range: [new_start, old_end)
633 */
Rik van Riel5beb4932010-03-05 13:42:07 -0800634 if (vma_adjust(vma, new_start, old_end, vma->vm_pgoff, NULL))
635 return -ENOMEM;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700636
637 /*
638 * move the page tables downwards, on failure we rely on
639 * process cleanup to remove whatever mess we made.
640 */
641 if (length != move_page_tables(vma, old_start,
Michel Lespinasse38a76012012-10-08 16:31:50 -0700642 vma, new_start, length, false))
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700643 return -ENOMEM;
644
645 lru_add_drain();
Linus Torvalds2b047252013-08-15 11:42:25 -0700646 tlb_gather_mmu(&tlb, mm, old_start, old_end);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700647 if (new_end > old_start) {
648 /*
649 * when the old and new regions overlap clear from new_end.
650 */
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700651 free_pgd_range(&tlb, new_end, old_end, new_end,
Hugh Dickins6ee86302013-04-29 15:07:44 -0700652 vma->vm_next ? vma->vm_next->vm_start : USER_PGTABLES_CEILING);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700653 } else {
654 /*
655 * otherwise, clean from old_start; this is done to not touch
656 * the address space in [new_end, old_start) some architectures
657 * have constraints on va-space that make this illegal (IA64) -
658 * for the others its just a little faster.
659 */
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700660 free_pgd_range(&tlb, old_start, old_end, new_end,
Hugh Dickins6ee86302013-04-29 15:07:44 -0700661 vma->vm_next ? vma->vm_next->vm_start : USER_PGTABLES_CEILING);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700662 }
Linus Torvalds2b047252013-08-15 11:42:25 -0700663 tlb_finish_mmu(&tlb, old_start, old_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700665 /*
Rik van Riel5beb4932010-03-05 13:42:07 -0800666 * Shrink the vma to just the new range. Always succeeds.
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700667 */
668 vma_adjust(vma, new_start, new_end, vma->vm_pgoff, NULL);
669
670 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671}
672
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700673/*
674 * Finalizes the stack vm_area_struct. The flags and permissions are updated,
675 * the stack is optionally relocated, and some extra space is added.
676 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700677int setup_arg_pages(struct linux_binprm *bprm,
678 unsigned long stack_top,
679 int executable_stack)
680{
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700681 unsigned long ret;
682 unsigned long stack_shift;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700683 struct mm_struct *mm = current->mm;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700684 struct vm_area_struct *vma = bprm->vma;
685 struct vm_area_struct *prev = NULL;
686 unsigned long vm_flags;
687 unsigned long stack_base;
Michael Neuling803bf5e2010-02-10 13:56:42 -0800688 unsigned long stack_size;
689 unsigned long stack_expand;
690 unsigned long rlim_stack;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700691
692#ifdef CONFIG_STACK_GROWSUP
James Hogand71f2902014-05-13 23:58:24 +0100693 /* Limit stack size */
Jiri Slabyd554ed892010-03-05 13:42:42 -0800694 stack_base = rlimit_max(RLIMIT_STACK);
James Hogand71f2902014-05-13 23:58:24 +0100695 if (stack_base > STACK_SIZE_MAX)
696 stack_base = STACK_SIZE_MAX;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700697
Helge Dellerd045c772015-05-11 22:01:27 +0200698 /* Add space for stack randomization. */
699 stack_base += (STACK_RND_MASK << PAGE_SHIFT);
700
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700701 /* Make sure we didn't let the argument array grow too large. */
702 if (vma->vm_end - vma->vm_start > stack_base)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700703 return -ENOMEM;
704
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700705 stack_base = PAGE_ALIGN(stack_top - stack_base);
706
707 stack_shift = vma->vm_start - stack_base;
708 mm->arg_start = bprm->p - stack_shift;
709 bprm->p = vma->vm_end - stack_shift;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710#else
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700711 stack_top = arch_align_stack(stack_top);
712 stack_top = PAGE_ALIGN(stack_top);
Roland McGrath1b528182010-09-07 19:35:49 -0700713
714 if (unlikely(stack_top < mmap_min_addr) ||
715 unlikely(vma->vm_end - vma->vm_start >= stack_top - mmap_min_addr))
716 return -ENOMEM;
717
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700718 stack_shift = vma->vm_end - stack_top;
719
720 bprm->p -= stack_shift;
721 mm->arg_start = bprm->p;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700722#endif
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700723
724 if (bprm->loader)
725 bprm->loader -= stack_shift;
726 bprm->exec -= stack_shift;
727
Michal Hockof268dfe2016-05-23 16:26:02 -0700728 if (down_write_killable(&mm->mmap_sem))
729 return -EINTR;
730
Hugh Dickins96a8e132008-07-10 21:19:20 +0100731 vm_flags = VM_STACK_FLAGS;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700732
733 /*
734 * Adjust stack execute permissions; explicitly enable for
735 * EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X and leave alone
736 * (arch default) otherwise.
737 */
738 if (unlikely(executable_stack == EXSTACK_ENABLE_X))
739 vm_flags |= VM_EXEC;
740 else if (executable_stack == EXSTACK_DISABLE_X)
741 vm_flags &= ~VM_EXEC;
742 vm_flags |= mm->def_flags;
Mel Gormana8bef8ff2010-05-24 14:32:24 -0700743 vm_flags |= VM_STACK_INCOMPLETE_SETUP;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700744
745 ret = mprotect_fixup(vma, &prev, vma->vm_start, vma->vm_end,
746 vm_flags);
747 if (ret)
748 goto out_unlock;
749 BUG_ON(prev != vma);
750
751 /* Move stack pages down in memory. */
752 if (stack_shift) {
753 ret = shift_arg_pages(vma, stack_shift);
Anton Blanchardfc63cf22009-11-11 14:26:48 -0800754 if (ret)
755 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756 }
757
Mel Gormana8bef8ff2010-05-24 14:32:24 -0700758 /* mprotect_fixup is overkill to remove the temporary stack flags */
759 vma->vm_flags &= ~VM_STACK_INCOMPLETE_SETUP;
760
Michael Neuling5ef097d2010-03-05 13:42:57 -0800761 stack_expand = 131072UL; /* randomly 32*4k (or 2*64k) pages */
Michael Neuling803bf5e2010-02-10 13:56:42 -0800762 stack_size = vma->vm_end - vma->vm_start;
763 /*
764 * Align this down to a page boundary as expand_stack
765 * will align it up.
766 */
767 rlim_stack = rlimit(RLIMIT_STACK) & PAGE_MASK;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700768#ifdef CONFIG_STACK_GROWSUP
Michael Neuling803bf5e2010-02-10 13:56:42 -0800769 if (stack_size + stack_expand > rlim_stack)
770 stack_base = vma->vm_start + rlim_stack;
771 else
772 stack_base = vma->vm_end + stack_expand;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700773#else
Michael Neuling803bf5e2010-02-10 13:56:42 -0800774 if (stack_size + stack_expand > rlim_stack)
775 stack_base = vma->vm_end - rlim_stack;
776 else
777 stack_base = vma->vm_start - stack_expand;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700778#endif
Eric B Munson3af9e852010-05-18 15:30:49 +0100779 current->mm->start_stack = bprm->p;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700780 ret = expand_stack(vma, stack_base);
781 if (ret)
782 ret = -EFAULT;
783
784out_unlock:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700785 up_write(&mm->mmap_sem);
Anton Blanchardfc63cf22009-11-11 14:26:48 -0800786 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788EXPORT_SYMBOL(setup_arg_pages);
789
Nicolas Pitre7e7ec6a2016-07-24 11:30:18 -0400790#else
791
792/*
793 * Transfer the program arguments and environment from the holding pages
794 * onto the stack. The provided stack pointer is adjusted accordingly.
795 */
796int transfer_args_to_stack(struct linux_binprm *bprm,
797 unsigned long *sp_location)
798{
799 unsigned long index, stop, sp;
800 int ret = 0;
801
802 stop = bprm->p >> PAGE_SHIFT;
803 sp = *sp_location;
804
805 for (index = MAX_ARG_PAGES - 1; index >= stop; index--) {
806 unsigned int offset = index == stop ? bprm->p & ~PAGE_MASK : 0;
807 char *src = kmap(bprm->page[index]) + offset;
808 sp -= PAGE_SIZE - offset;
809 if (copy_to_user((void *) sp, src, PAGE_SIZE - offset) != 0)
810 ret = -EFAULT;
811 kunmap(bprm->page[index]);
812 if (ret)
813 goto out;
814 }
815
816 *sp_location = sp;
817
818out:
819 return ret;
820}
821EXPORT_SYMBOL(transfer_args_to_stack);
822
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823#endif /* CONFIG_MMU */
824
David Drysdale51f39a12014-12-12 16:57:29 -0800825static struct file *do_open_execat(int fd, struct filename *name, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827 struct file *file;
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200828 int err;
David Drysdale51f39a12014-12-12 16:57:29 -0800829 struct open_flags open_exec_flags = {
Al Viro47c805d2011-02-23 17:44:09 -0500830 .open_flag = O_LARGEFILE | O_RDONLY | __FMODE_EXEC,
Al Viro62fb4a12015-12-26 22:33:24 -0500831 .acc_mode = MAY_EXEC,
Al Virof9652e12013-06-11 08:23:01 +0400832 .intent = LOOKUP_OPEN,
833 .lookup_flags = LOOKUP_FOLLOW,
Al Viro47c805d2011-02-23 17:44:09 -0500834 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835
David Drysdale51f39a12014-12-12 16:57:29 -0800836 if ((flags & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
837 return ERR_PTR(-EINVAL);
838 if (flags & AT_SYMLINK_NOFOLLOW)
839 open_exec_flags.lookup_flags &= ~LOOKUP_FOLLOW;
840 if (flags & AT_EMPTY_PATH)
841 open_exec_flags.lookup_flags |= LOOKUP_EMPTY;
842
843 file = do_filp_open(fd, name, &open_exec_flags);
Al Viro6e8341a2009-04-06 11:16:22 -0400844 if (IS_ERR(file))
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200845 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700846
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200847 err = -EACCES;
Al Viro496ad9a2013-01-23 17:07:38 -0500848 if (!S_ISREG(file_inode(file)->i_mode))
Al Viro6e8341a2009-04-06 11:16:22 -0400849 goto exit;
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200850
Eric W. Biederman90f85722015-06-29 14:42:03 -0500851 if (path_noexec(&file->f_path))
Al Viro6e8341a2009-04-06 11:16:22 -0400852 goto exit;
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200853
854 err = deny_write_access(file);
Al Viro6e8341a2009-04-06 11:16:22 -0400855 if (err)
856 goto exit;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700857
David Drysdale51f39a12014-12-12 16:57:29 -0800858 if (name->name[0] != '\0')
859 fsnotify_open(file);
860
Al Viro6e8341a2009-04-06 11:16:22 -0400861out:
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200862 return file;
863
Al Viro6e8341a2009-04-06 11:16:22 -0400864exit:
865 fput(file);
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200866 return ERR_PTR(err);
867}
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -0800868
869struct file *open_exec(const char *name)
870{
Paul Moore51689102015-01-22 00:00:03 -0500871 struct filename *filename = getname_kernel(name);
872 struct file *f = ERR_CAST(filename);
873
874 if (!IS_ERR(filename)) {
875 f = do_open_execat(AT_FDCWD, filename, 0);
876 putname(filename);
877 }
878 return f;
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -0800879}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700880EXPORT_SYMBOL(open_exec);
881
Mimi Zohar6777d772009-08-21 14:32:48 -0400882int kernel_read(struct file *file, loff_t offset,
883 char *addr, unsigned long count)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700884{
885 mm_segment_t old_fs;
886 loff_t pos = offset;
887 int result;
888
889 old_fs = get_fs();
890 set_fs(get_ds());
891 /* The cast to a user pointer is valid due to the set_fs() */
892 result = vfs_read(file, (void __user *)addr, count, &pos);
893 set_fs(old_fs);
894 return result;
895}
896
897EXPORT_SYMBOL(kernel_read);
898
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500899int kernel_read_file(struct file *file, void **buf, loff_t *size,
Mimi Zoharbc8ca5b2016-01-24 10:07:32 -0500900 loff_t max_size, enum kernel_read_file_id id)
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500901{
902 loff_t i_size, pos;
903 ssize_t bytes = 0;
904 int ret;
905
906 if (!S_ISREG(file_inode(file)->i_mode) || max_size < 0)
907 return -EINVAL;
908
Mimi Zohar39eeb4f2016-01-30 22:23:26 -0500909 ret = security_kernel_read_file(file, id);
910 if (ret)
911 return ret;
912
Dmitry Kasatkin39d637a2014-10-26 12:42:07 +0200913 ret = deny_write_access(file);
914 if (ret)
915 return ret;
916
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500917 i_size = i_size_read(file_inode(file));
Dmitry Kasatkin39d637a2014-10-26 12:42:07 +0200918 if (max_size > 0 && i_size > max_size) {
919 ret = -EFBIG;
920 goto out;
921 }
922 if (i_size <= 0) {
923 ret = -EINVAL;
924 goto out;
925 }
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500926
Stephen Boyda098ecd2016-08-02 14:04:28 -0700927 if (id != READING_FIRMWARE_PREALLOC_BUFFER)
928 *buf = vmalloc(i_size);
Dmitry Kasatkin39d637a2014-10-26 12:42:07 +0200929 if (!*buf) {
930 ret = -ENOMEM;
931 goto out;
932 }
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500933
934 pos = 0;
935 while (pos < i_size) {
936 bytes = kernel_read(file, pos, (char *)(*buf) + pos,
937 i_size - pos);
938 if (bytes < 0) {
939 ret = bytes;
940 goto out;
941 }
942
943 if (bytes == 0)
944 break;
945 pos += bytes;
946 }
947
948 if (pos != i_size) {
949 ret = -EIO;
Dmitry Kasatkin39d637a2014-10-26 12:42:07 +0200950 goto out_free;
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500951 }
952
Mimi Zoharbc8ca5b2016-01-24 10:07:32 -0500953 ret = security_kernel_post_read_file(file, *buf, i_size, id);
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500954 if (!ret)
955 *size = pos;
956
Dmitry Kasatkin39d637a2014-10-26 12:42:07 +0200957out_free:
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500958 if (ret < 0) {
Stephen Boyda098ecd2016-08-02 14:04:28 -0700959 if (id != READING_FIRMWARE_PREALLOC_BUFFER) {
960 vfree(*buf);
961 *buf = NULL;
962 }
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500963 }
Dmitry Kasatkin39d637a2014-10-26 12:42:07 +0200964
965out:
966 allow_write_access(file);
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500967 return ret;
968}
969EXPORT_SYMBOL_GPL(kernel_read_file);
970
Mimi Zohar09596b92015-11-19 12:39:22 -0500971int kernel_read_file_from_path(char *path, void **buf, loff_t *size,
972 loff_t max_size, enum kernel_read_file_id id)
973{
974 struct file *file;
975 int ret;
976
977 if (!path || !*path)
978 return -EINVAL;
979
980 file = filp_open(path, O_RDONLY, 0);
981 if (IS_ERR(file))
982 return PTR_ERR(file);
983
984 ret = kernel_read_file(file, buf, size, max_size, id);
985 fput(file);
986 return ret;
987}
988EXPORT_SYMBOL_GPL(kernel_read_file_from_path);
989
Mimi Zoharb844f0e2016-02-01 08:36:21 -0500990int kernel_read_file_from_fd(int fd, void **buf, loff_t *size, loff_t max_size,
991 enum kernel_read_file_id id)
992{
993 struct fd f = fdget(fd);
994 int ret = -EBADF;
995
996 if (!f.file)
997 goto out;
998
999 ret = kernel_read_file(f.file, buf, size, max_size, id);
1000out:
1001 fdput(f);
1002 return ret;
1003}
1004EXPORT_SYMBOL_GPL(kernel_read_file_from_fd);
1005
Al Viro3dc20cb2013-04-13 20:31:37 -04001006ssize_t read_code(struct file *file, unsigned long addr, loff_t pos, size_t len)
1007{
Al Viroec695572014-02-04 19:08:21 -05001008 ssize_t res = vfs_read(file, (void __user *)addr, len, &pos);
Al Viro3dc20cb2013-04-13 20:31:37 -04001009 if (res > 0)
1010 flush_icache_range(addr, addr + len);
1011 return res;
1012}
1013EXPORT_SYMBOL(read_code);
1014
Linus Torvalds1da177e2005-04-16 15:20:36 -07001015static int exec_mmap(struct mm_struct *mm)
1016{
1017 struct task_struct *tsk;
Davidlohr Bueso615d6e82014-04-07 15:37:25 -07001018 struct mm_struct *old_mm, *active_mm;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019
1020 /* Notify parent that we're no longer interested in the old VM */
1021 tsk = current;
1022 old_mm = current->mm;
1023 mm_release(tsk, old_mm);
1024
1025 if (old_mm) {
Konstantin Khlebnikov4fe7efd2012-06-20 12:53:01 -07001026 sync_mm_rss(old_mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001027 /*
1028 * Make sure that if there is a core dump in progress
1029 * for the old mm, we get out and die instead of going
1030 * through with the exec. We must hold mmap_sem around
Oleg Nesterov999d9fc2008-07-25 01:47:41 -07001031 * checking core_state and changing tsk->mm.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001032 */
1033 down_read(&old_mm->mmap_sem);
Oleg Nesterov999d9fc2008-07-25 01:47:41 -07001034 if (unlikely(old_mm->core_state)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035 up_read(&old_mm->mmap_sem);
1036 return -EINTR;
1037 }
1038 }
1039 task_lock(tsk);
1040 active_mm = tsk->active_mm;
1041 tsk->mm = mm;
1042 tsk->active_mm = mm;
1043 activate_mm(active_mm, mm);
Davidlohr Bueso615d6e82014-04-07 15:37:25 -07001044 tsk->mm->vmacache_seqnum = 0;
1045 vmacache_flush(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001046 task_unlock(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001047 if (old_mm) {
1048 up_read(&old_mm->mmap_sem);
Eric Sesterhenn7dddb122006-04-01 01:13:38 +02001049 BUG_ON(active_mm != old_mm);
Oleg Nesterov701085b2012-03-19 17:04:01 +01001050 setmax_mm_hiwater_rss(&tsk->signal->maxrss, old_mm);
Balbir Singh31a78f22008-09-28 23:09:31 +01001051 mm_update_next_owner(old_mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001052 mmput(old_mm);
1053 return 0;
1054 }
1055 mmdrop(active_mm);
1056 return 0;
1057}
1058
1059/*
1060 * This function makes sure the current process has its own signal table,
1061 * so that flush_signal_handlers can later reset the handlers without
1062 * disturbing other processes. (Other processes might share the signal
1063 * table via the CLONE_SIGHAND option to clone().)
1064 */
Arjan van de Ven858119e2006-01-14 13:20:43 -08001065static int de_thread(struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001066{
1067 struct signal_struct *sig = tsk->signal;
Oleg Nesterovb2c903b2007-10-16 23:27:22 -07001068 struct sighand_struct *oldsighand = tsk->sighand;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001069 spinlock_t *lock = &oldsighand->siglock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001070
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001071 if (thread_group_empty(tsk))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001072 goto no_thread_group;
1073
1074 /*
1075 * Kill all other threads in the thread group.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001076 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001077 spin_lock_irq(lock);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -08001078 if (signal_group_exit(sig)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001079 /*
1080 * Another group action in progress, just
1081 * return so that the signal is processed.
1082 */
1083 spin_unlock_irq(lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001084 return -EAGAIN;
1085 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001086
Oleg Nesterovd3441932010-05-26 14:43:11 -07001087 sig->group_exit_task = tsk;
1088 sig->notify_count = zap_other_threads(tsk);
1089 if (!thread_group_leader(tsk))
1090 sig->notify_count--;
1091
1092 while (sig->notify_count) {
Oleg Nesterovd5bbd432012-10-08 19:13:01 +02001093 __set_current_state(TASK_KILLABLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001094 spin_unlock_irq(lock);
1095 schedule();
Oleg Nesterovd5bbd432012-10-08 19:13:01 +02001096 if (unlikely(__fatal_signal_pending(tsk)))
1097 goto killed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001098 spin_lock_irq(lock);
1099 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001100 spin_unlock_irq(lock);
1101
1102 /*
1103 * At this point all other threads have exited, all we have to
1104 * do is to wait for the thread group leader to become inactive,
1105 * and to assume its PID:
1106 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001107 if (!thread_group_leader(tsk)) {
Oleg Nesterov81879262008-12-01 14:18:16 -08001108 struct task_struct *leader = tsk->group_leader;
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001109
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001110 for (;;) {
Oleg Nesterove56fb282013-04-30 15:28:20 -07001111 threadgroup_change_begin(tsk);
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001112 write_lock_irq(&tasklist_lock);
Kirill Tkhaidfcce792015-04-16 12:48:01 -07001113 /*
1114 * Do this under tasklist_lock to ensure that
1115 * exit_notify() can't miss ->group_exit_task
1116 */
1117 sig->notify_count = -1;
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001118 if (likely(leader->exit_state))
1119 break;
Oleg Nesterovd5bbd432012-10-08 19:13:01 +02001120 __set_current_state(TASK_KILLABLE);
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001121 write_unlock_irq(&tasklist_lock);
Oleg Nesterove56fb282013-04-30 15:28:20 -07001122 threadgroup_change_end(tsk);
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001123 schedule();
Oleg Nesterovd5bbd432012-10-08 19:13:01 +02001124 if (unlikely(__fatal_signal_pending(tsk)))
1125 goto killed;
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001126 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001127
Roland McGrathf5e90282006-04-10 22:54:16 -07001128 /*
1129 * The only record we have of the real-time age of a
1130 * process, regardless of execs it's done, is start_time.
1131 * All the past CPU time is accumulated in signal_struct
1132 * from sister threads now dead. But in this non-leader
1133 * exec, nothing survives from the original leader thread,
1134 * whose birth marks the true age of this process now.
1135 * When we take on its identity by switching to its PID, we
1136 * also take its birthdate (always earlier than our own).
1137 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001138 tsk->start_time = leader->start_time;
Oleg Nesterov266b7a02013-07-03 15:08:35 -07001139 tsk->real_start_time = leader->real_start_time;
Roland McGrathf5e90282006-04-10 22:54:16 -07001140
Pavel Emelyanovbac0abd2007-10-18 23:40:18 -07001141 BUG_ON(!same_thread_group(leader, tsk));
1142 BUG_ON(has_group_leader_pid(tsk));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001143 /*
1144 * An exec() starts a new thread group with the
1145 * TGID of the previous thread group. Rehash the
1146 * two threads with a switched PID, and release
1147 * the former thread group leader:
1148 */
Eric W. Biedermand73d6522006-03-28 16:11:03 -08001149
1150 /* Become a process group leader with the old leader's pid.
Eric W. Biedermanc18258c2006-09-27 01:51:06 -07001151 * The old leader becomes a thread of the this thread group.
1152 * Note: The old leader also uses this pid until release_task
Eric W. Biedermand73d6522006-03-28 16:11:03 -08001153 * is called. Odd but simple and correct.
1154 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001155 tsk->pid = leader->pid;
Oleg Nesterov3f418542013-07-03 15:08:25 -07001156 change_pid(tsk, PIDTYPE_PID, task_pid(leader));
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001157 transfer_pid(leader, tsk, PIDTYPE_PGID);
1158 transfer_pid(leader, tsk, PIDTYPE_SID);
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -08001159
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001160 list_replace_rcu(&leader->tasks, &tsk->tasks);
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -08001161 list_replace_init(&leader->sibling, &tsk->sibling);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001163 tsk->group_leader = tsk;
1164 leader->group_leader = tsk;
Eric W. Biedermande12a782006-04-10 17:16:49 -06001165
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001166 tsk->exit_signal = SIGCHLD;
Oleg Nesterov087806b2011-06-22 23:10:26 +02001167 leader->exit_signal = -1;
Oleg Nesterov962b5642005-11-23 13:37:43 -08001168
1169 BUG_ON(leader->exit_state != EXIT_ZOMBIE);
1170 leader->exit_state = EXIT_DEAD;
Oleg Nesteroveac1b5e2011-07-21 20:00:43 +02001171
1172 /*
1173 * We are going to release_task()->ptrace_unlink() silently,
1174 * the tracer can sleep in do_wait(). EXIT_DEAD guarantees
1175 * the tracer wont't block again waiting for this thread.
1176 */
1177 if (unlikely(leader->ptrace))
1178 __wake_up_parent(leader, leader->parent);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001179 write_unlock_irq(&tasklist_lock);
Oleg Nesterove56fb282013-04-30 15:28:20 -07001180 threadgroup_change_end(tsk);
Oleg Nesterov81879262008-12-01 14:18:16 -08001181
1182 release_task(leader);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -08001183 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001184
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001185 sig->group_exit_task = NULL;
1186 sig->notify_count = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001187
1188no_thread_group:
Oleg Nesterove6368252012-03-19 17:03:22 +01001189 /* we have changed execution domain */
1190 tsk->exit_signal = SIGCHLD;
1191
Linus Torvalds1da177e2005-04-16 15:20:36 -07001192 exit_itimers(sig);
Oleg Nesterovcbaffba2008-05-26 20:55:42 +04001193 flush_itimer_signals();
Oleg Nesterov329f7db2005-11-07 21:12:43 +03001194
Oleg Nesterovb2c903b2007-10-16 23:27:22 -07001195 if (atomic_read(&oldsighand->count) != 1) {
1196 struct sighand_struct *newsighand;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001197 /*
Oleg Nesterovb2c903b2007-10-16 23:27:22 -07001198 * This ->sighand is shared with the CLONE_SIGHAND
1199 * but not CLONE_THREAD task, switch to the new one.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001200 */
Oleg Nesterovb2c903b2007-10-16 23:27:22 -07001201 newsighand = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
1202 if (!newsighand)
1203 return -ENOMEM;
1204
Linus Torvalds1da177e2005-04-16 15:20:36 -07001205 atomic_set(&newsighand->count, 1);
1206 memcpy(newsighand->action, oldsighand->action,
1207 sizeof(newsighand->action));
1208
1209 write_lock_irq(&tasklist_lock);
1210 spin_lock(&oldsighand->siglock);
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001211 rcu_assign_pointer(tsk->sighand, newsighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001212 spin_unlock(&oldsighand->siglock);
1213 write_unlock_irq(&tasklist_lock);
1214
Davide Libenzifba2afa2007-05-10 22:23:13 -07001215 __cleanup_sighand(oldsighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001216 }
1217
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001218 BUG_ON(!thread_group_leader(tsk));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001219 return 0;
Oleg Nesterovd5bbd432012-10-08 19:13:01 +02001220
1221killed:
1222 /* protects against exit_notify() and __exit_signal() */
1223 read_lock(&tasklist_lock);
1224 sig->group_exit_task = NULL;
1225 sig->notify_count = 0;
1226 read_unlock(&tasklist_lock);
1227 return -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001228}
Oleg Nesterov0840a902007-10-16 23:27:22 -07001229
Andrew Morton59714d62008-02-04 22:27:21 -08001230char *get_task_comm(char *buf, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001231{
1232 /* buf must be at least sizeof(tsk->comm) in size */
1233 task_lock(tsk);
1234 strncpy(buf, tsk->comm, sizeof(tsk->comm));
1235 task_unlock(tsk);
Andrew Morton59714d62008-02-04 22:27:21 -08001236 return buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001237}
J Freyensee7d74f492011-05-06 16:56:47 -07001238EXPORT_SYMBOL_GPL(get_task_comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001239
Al Viro6a6d27d2012-08-21 09:56:33 -04001240/*
1241 * These functions flushes out all traces of the currently running executable
1242 * so that a new one can be started
1243 */
1244
Adrian Hunter82b89772014-05-28 11:45:04 +03001245void __set_task_comm(struct task_struct *tsk, const char *buf, bool exec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001246{
1247 task_lock(tsk);
KAMEZAWA Hiroyuki43d2b112012-01-10 15:08:09 -08001248 trace_task_rename(tsk, buf);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001249 strlcpy(tsk->comm, buf, sizeof(tsk->comm));
1250 task_unlock(tsk);
Adrian Hunter82b89772014-05-28 11:45:04 +03001251 perf_event_comm(tsk, exec);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001252}
1253
1254int flush_old_exec(struct linux_binprm * bprm)
1255{
Linus Torvalds221af7f2010-01-28 22:14:42 -08001256 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001257
1258 /*
1259 * Make sure we have a private signal table and that
1260 * we are unassociated from the previous thread group.
1261 */
1262 retval = de_thread(current);
1263 if (retval)
1264 goto out;
1265
Davidlohr Bueso6e399cd2015-04-16 12:47:59 -07001266 /*
1267 * Must be called _before_ exec_mmap() as bprm->mm is
1268 * not visibile until then. This also enables the update
1269 * to be lockless.
1270 */
Matt Helsley925d1c42008-04-29 01:01:36 -07001271 set_mm_exe_file(bprm->mm, bprm->file);
Davidlohr Bueso6e399cd2015-04-16 12:47:59 -07001272
Linus Torvalds1da177e2005-04-16 15:20:36 -07001273 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274 * Release all of the old mmap stuff
1275 */
Oleg Nesterov3c77f842010-11-30 20:55:34 +01001276 acct_arg_size(bprm, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001277 retval = exec_mmap(bprm->mm);
1278 if (retval)
Al Virofd8328b2008-04-22 05:11:59 -04001279 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001280
1281 bprm->mm = NULL; /* We're using it now */
Linus Torvalds7ab02af2010-02-02 12:37:44 -08001282
Mathias Krausedac853a2011-06-09 20:05:18 +02001283 set_fs(USER_DS);
Zhang Yib88fae62014-01-23 15:55:57 -08001284 current->flags &= ~(PF_RANDOMIZE | PF_FORKNOEXEC | PF_KTHREAD |
1285 PF_NOFREEZE | PF_NO_SETAFFINITY);
Linus Torvalds7ab02af2010-02-02 12:37:44 -08001286 flush_thread();
1287 current->personality &= ~bprm->per_clear;
1288
Aleksa Saraic1df5a62016-12-21 16:26:24 +11001289 /*
1290 * We have to apply CLOEXEC before we change whether the process is
1291 * dumpable (in setup_new_exec) to avoid a race with a process in userspace
1292 * trying to access the should-be-closed file descriptors of a process
1293 * undergoing exec(2).
1294 */
1295 do_close_on_exec(current->files);
Linus Torvalds221af7f2010-01-28 22:14:42 -08001296 return 0;
1297
1298out:
1299 return retval;
1300}
1301EXPORT_SYMBOL(flush_old_exec);
1302
Al Viro1b5d7832011-06-19 12:49:47 -04001303void would_dump(struct linux_binprm *bprm, struct file *file)
1304{
Eric W. Biederman21245b82016-11-16 22:06:51 -06001305 struct inode *inode = file_inode(file);
1306 if (inode_permission(inode, MAY_READ) < 0) {
1307 struct user_namespace *old, *user_ns;
Al Viro1b5d7832011-06-19 12:49:47 -04001308 bprm->interp_flags |= BINPRM_FLAGS_ENFORCE_NONDUMP;
Eric W. Biederman21245b82016-11-16 22:06:51 -06001309
1310 /* Ensure mm->user_ns contains the executable */
1311 user_ns = old = bprm->mm->user_ns;
1312 while ((user_ns != &init_user_ns) &&
1313 !privileged_wrt_inode_uidgid(user_ns, inode))
1314 user_ns = user_ns->parent;
1315
1316 if (old != user_ns) {
1317 bprm->mm->user_ns = get_user_ns(user_ns);
1318 put_user_ns(old);
1319 }
1320 }
Al Viro1b5d7832011-06-19 12:49:47 -04001321}
1322EXPORT_SYMBOL(would_dump);
1323
Linus Torvalds221af7f2010-01-28 22:14:42 -08001324void setup_new_exec(struct linux_binprm * bprm)
1325{
Linus Torvalds221af7f2010-01-28 22:14:42 -08001326 arch_pick_mmap_layout(current->mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001327
1328 /* This is the point of no return */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001329 current->sas_ss_sp = current->sas_ss_size = 0;
1330
Eric W. Biederman8e96e3b2012-03-03 21:17:15 -08001331 if (uid_eq(current_euid(), current_uid()) && gid_eq(current_egid(), current_gid()))
Kees Cooke579d2c2013-02-27 17:03:15 -08001332 set_dumpable(current->mm, SUID_DUMP_USER);
Alan Coxd6e71142005-06-23 00:09:43 -07001333 else
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001334 set_dumpable(current->mm, suid_dumpable);
Alan Coxd6e71142005-06-23 00:09:43 -07001335
Peter Zijlstrae041e322014-05-21 17:32:19 +02001336 perf_event_exec();
Adrian Hunter82b89772014-05-28 11:45:04 +03001337 __set_task_comm(current, kbasename(bprm->filename), true);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001338
Benjamin Herrenschmidt0551fbd2006-02-28 16:59:19 -08001339 /* Set the new mm task size. We have to do that late because it may
1340 * depend on TIF_32BIT which is only updated in flush_thread() on
1341 * some architectures like powerpc
1342 */
1343 current->mm->task_size = TASK_SIZE;
1344
David Howellsa6f76f22008-11-14 10:39:24 +11001345 /* install the new credentials */
Eric W. Biederman8e96e3b2012-03-03 21:17:15 -08001346 if (!uid_eq(bprm->cred->uid, current_euid()) ||
1347 !gid_eq(bprm->cred->gid, current_egid())) {
Marcel Holtmannd2d56c52007-08-17 21:47:58 +02001348 current->pdeath_signal = 0;
Al Viro1b5d7832011-06-19 12:49:47 -04001349 } else {
Al Viro1b5d7832011-06-19 12:49:47 -04001350 if (bprm->interp_flags & BINPRM_FLAGS_ENFORCE_NONDUMP)
1351 set_dumpable(current->mm, suid_dumpable);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352 }
1353
1354 /* An exec changes our domain. We are no longer part of the thread
1355 group */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001356 current->self_exec_id++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001357 flush_signal_handlers(current, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001358}
Linus Torvalds221af7f2010-01-28 22:14:42 -08001359EXPORT_SYMBOL(setup_new_exec);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001360
David Howellsa6f76f22008-11-14 10:39:24 +11001361/*
Oleg Nesterova2a84742009-09-05 11:17:13 -07001362 * Prepare credentials and lock ->cred_guard_mutex.
1363 * install_exec_creds() commits the new creds and drops the lock.
1364 * Or, if exec fails before, free_bprm() should release ->cred and
1365 * and unlock.
1366 */
1367int prepare_bprm_creds(struct linux_binprm *bprm)
1368{
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001369 if (mutex_lock_interruptible(&current->signal->cred_guard_mutex))
Oleg Nesterova2a84742009-09-05 11:17:13 -07001370 return -ERESTARTNOINTR;
1371
1372 bprm->cred = prepare_exec_creds();
1373 if (likely(bprm->cred))
1374 return 0;
1375
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001376 mutex_unlock(&current->signal->cred_guard_mutex);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001377 return -ENOMEM;
1378}
1379
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001380static void free_bprm(struct linux_binprm *bprm)
Oleg Nesterova2a84742009-09-05 11:17:13 -07001381{
1382 free_arg_pages(bprm);
1383 if (bprm->cred) {
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001384 mutex_unlock(&current->signal->cred_guard_mutex);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001385 abort_creds(bprm->cred);
1386 }
Oleg Nesterov63e46b92014-01-23 15:55:51 -08001387 if (bprm->file) {
1388 allow_write_access(bprm->file);
1389 fput(bprm->file);
1390 }
Kees Cookb66c5982012-12-20 15:05:16 -08001391 /* If a binfmt changed the interp, free it. */
1392 if (bprm->interp != bprm->filename)
1393 kfree(bprm->interp);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001394 kfree(bprm);
1395}
1396
Kees Cookb66c5982012-12-20 15:05:16 -08001397int bprm_change_interp(char *interp, struct linux_binprm *bprm)
1398{
1399 /* If a binfmt changed the interp, free it first. */
1400 if (bprm->interp != bprm->filename)
1401 kfree(bprm->interp);
1402 bprm->interp = kstrdup(interp, GFP_KERNEL);
1403 if (!bprm->interp)
1404 return -ENOMEM;
1405 return 0;
1406}
1407EXPORT_SYMBOL(bprm_change_interp);
1408
Oleg Nesterova2a84742009-09-05 11:17:13 -07001409/*
David Howellsa6f76f22008-11-14 10:39:24 +11001410 * install the new credentials for this executable
1411 */
1412void install_exec_creds(struct linux_binprm *bprm)
1413{
1414 security_bprm_committing_creds(bprm);
1415
1416 commit_creds(bprm->cred);
1417 bprm->cred = NULL;
Stephane Eranian2976b102013-06-20 11:36:28 +02001418
1419 /*
1420 * Disable monitoring for regular users
1421 * when executing setuid binaries. Must
1422 * wait until new credentials are committed
1423 * by commit_creds() above
1424 */
1425 if (get_dumpable(current->mm) != SUID_DUMP_USER)
1426 perf_event_exit_task(current);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001427 /*
1428 * cred_guard_mutex must be held at least to this point to prevent
David Howellsa6f76f22008-11-14 10:39:24 +11001429 * ptrace_attach() from altering our determination of the task's
Oleg Nesterova2a84742009-09-05 11:17:13 -07001430 * credentials; any time after this it may be unlocked.
1431 */
David Howellsa6f76f22008-11-14 10:39:24 +11001432 security_bprm_committed_creds(bprm);
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001433 mutex_unlock(&current->signal->cred_guard_mutex);
David Howellsa6f76f22008-11-14 10:39:24 +11001434}
1435EXPORT_SYMBOL(install_exec_creds);
1436
1437/*
1438 * determine how safe it is to execute the proposed program
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001439 * - the caller must hold ->cred_guard_mutex to protect against
Kees Cookc2e1f2e2014-06-05 00:23:17 -07001440 * PTRACE_ATTACH or seccomp thread-sync
David Howellsa6f76f22008-11-14 10:39:24 +11001441 */
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001442static void check_unsafe_exec(struct linux_binprm *bprm)
David Howellsa6f76f22008-11-14 10:39:24 +11001443{
David Howells0bf2f3a2009-02-06 11:45:46 +00001444 struct task_struct *p = current, *t;
Al Virof1191b52009-03-30 07:35:18 -04001445 unsigned n_fs;
David Howellsa6f76f22008-11-14 10:39:24 +11001446
Tejun Heo4b9d33e2011-06-17 16:50:38 +02001447 if (p->ptrace) {
Eric W. Biedermane747b4a2016-11-14 18:48:07 -06001448 if (ptracer_capable(p, current_user_ns()))
Tejun Heo4b9d33e2011-06-17 16:50:38 +02001449 bprm->unsafe |= LSM_UNSAFE_PTRACE_CAP;
1450 else
1451 bprm->unsafe |= LSM_UNSAFE_PTRACE;
1452 }
David Howellsa6f76f22008-11-14 10:39:24 +11001453
Andy Lutomirski259e5e62012-04-12 16:47:50 -05001454 /*
1455 * This isn't strictly necessary, but it makes it harder for LSMs to
1456 * mess up.
1457 */
Kees Cook1d4457f2014-05-21 15:23:46 -07001458 if (task_no_new_privs(current))
Andy Lutomirski259e5e62012-04-12 16:47:50 -05001459 bprm->unsafe |= LSM_UNSAFE_NO_NEW_PRIVS;
1460
Oleg Nesterov83f62a22014-01-23 15:55:49 -08001461 t = p;
David Howells0bf2f3a2009-02-06 11:45:46 +00001462 n_fs = 1;
Nick Piggin2a4419b2010-08-18 04:37:33 +10001463 spin_lock(&p->fs->lock);
Oleg Nesterov437f7fd2009-04-24 01:02:45 +02001464 rcu_read_lock();
Oleg Nesterov83f62a22014-01-23 15:55:49 -08001465 while_each_thread(p, t) {
David Howells0bf2f3a2009-02-06 11:45:46 +00001466 if (t->fs == p->fs)
1467 n_fs++;
David Howells0bf2f3a2009-02-06 11:45:46 +00001468 }
Oleg Nesterov437f7fd2009-04-24 01:02:45 +02001469 rcu_read_unlock();
David Howells0bf2f3a2009-02-06 11:45:46 +00001470
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001471 if (p->fs->users > n_fs)
David Howellsa6f76f22008-11-14 10:39:24 +11001472 bprm->unsafe |= LSM_UNSAFE_SHARE;
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001473 else
1474 p->fs->in_exec = 1;
Nick Piggin2a4419b2010-08-18 04:37:33 +10001475 spin_unlock(&p->fs->lock);
David Howellsa6f76f22008-11-14 10:39:24 +11001476}
1477
Jann Horn8b01fc82015-04-19 02:48:39 +02001478static void bprm_fill_uid(struct linux_binprm *bprm)
1479{
1480 struct inode *inode;
1481 unsigned int mode;
1482 kuid_t uid;
1483 kgid_t gid;
1484
Kees Cookcb6fd682016-05-17 12:14:39 -07001485 /*
1486 * Since this can be called multiple times (via prepare_binprm),
1487 * we must clear any previous work done when setting set[ug]id
1488 * bits from any earlier bprm->file uses (for example when run
1489 * first for a setuid script then again for its interpreter).
1490 */
Jann Horn8b01fc82015-04-19 02:48:39 +02001491 bprm->cred->euid = current_euid();
1492 bprm->cred->egid = current_egid();
1493
Andy Lutomirski380cf5b2016-06-23 16:41:05 -05001494 if (!mnt_may_suid(bprm->file->f_path.mnt))
Jann Horn8b01fc82015-04-19 02:48:39 +02001495 return;
1496
1497 if (task_no_new_privs(current))
1498 return;
1499
1500 inode = file_inode(bprm->file);
1501 mode = READ_ONCE(inode->i_mode);
1502 if (!(mode & (S_ISUID|S_ISGID)))
1503 return;
1504
1505 /* Be careful if suid/sgid is set */
Al Viro59551022016-01-22 15:40:57 -05001506 inode_lock(inode);
Jann Horn8b01fc82015-04-19 02:48:39 +02001507
1508 /* reload atomically mode/uid/gid now that lock held */
1509 mode = inode->i_mode;
1510 uid = inode->i_uid;
1511 gid = inode->i_gid;
Al Viro59551022016-01-22 15:40:57 -05001512 inode_unlock(inode);
Jann Horn8b01fc82015-04-19 02:48:39 +02001513
1514 /* We ignore suid/sgid if there are no mappings for them in the ns */
1515 if (!kuid_has_mapping(bprm->cred->user_ns, uid) ||
1516 !kgid_has_mapping(bprm->cred->user_ns, gid))
1517 return;
1518
1519 if (mode & S_ISUID) {
1520 bprm->per_clear |= PER_CLEAR_ON_SETID;
1521 bprm->cred->euid = uid;
1522 }
1523
1524 if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) {
1525 bprm->per_clear |= PER_CLEAR_ON_SETID;
1526 bprm->cred->egid = gid;
1527 }
1528}
1529
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001530/*
1531 * Fill the binprm structure from the inode.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001532 * Check permissions, then read the first 128 (BINPRM_BUF_SIZE) bytes
David Howellsa6f76f22008-11-14 10:39:24 +11001533 *
1534 * This may be called multiple times for binary chains (scripts for example).
Linus Torvalds1da177e2005-04-16 15:20:36 -07001535 */
1536int prepare_binprm(struct linux_binprm *bprm)
1537{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001538 int retval;
1539
Jann Horn8b01fc82015-04-19 02:48:39 +02001540 bprm_fill_uid(bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001541
1542 /* fill in binprm security blob */
David Howellsa6f76f22008-11-14 10:39:24 +11001543 retval = security_bprm_set_creds(bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001544 if (retval)
1545 return retval;
David Howellsa6f76f22008-11-14 10:39:24 +11001546 bprm->cred_prepared = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001547
David Howellsa6f76f22008-11-14 10:39:24 +11001548 memset(bprm->buf, 0, BINPRM_BUF_SIZE);
1549 return kernel_read(bprm->file, 0, bprm->buf, BINPRM_BUF_SIZE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001550}
1551
1552EXPORT_SYMBOL(prepare_binprm);
1553
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001554/*
1555 * Arguments are '\0' separated strings found at the location bprm->p
1556 * points to; chop off the first by relocating brpm->p to right after
1557 * the first '\0' encountered.
1558 */
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001559int remove_arg_zero(struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001560{
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001561 int ret = 0;
1562 unsigned long offset;
1563 char *kaddr;
1564 struct page *page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001565
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001566 if (!bprm->argc)
1567 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001568
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001569 do {
1570 offset = bprm->p & ~PAGE_MASK;
1571 page = get_arg_page(bprm, bprm->p, 0);
1572 if (!page) {
1573 ret = -EFAULT;
1574 goto out;
1575 }
Cong Wange8e3c3d2011-11-25 23:14:27 +08001576 kaddr = kmap_atomic(page);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001577
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001578 for (; offset < PAGE_SIZE && kaddr[offset];
1579 offset++, bprm->p++)
1580 ;
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001581
Cong Wange8e3c3d2011-11-25 23:14:27 +08001582 kunmap_atomic(kaddr);
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001583 put_arg_page(page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001584 } while (offset == PAGE_SIZE);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001585
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001586 bprm->p++;
1587 bprm->argc--;
1588 ret = 0;
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001589
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001590out:
1591 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001592}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001593EXPORT_SYMBOL(remove_arg_zero);
1594
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001595#define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001596/*
1597 * cycle the list of binary formats handler, until one recognizes the image
1598 */
Al Viro3c456bf2012-10-20 21:53:31 -04001599int search_binary_handler(struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001600{
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001601 bool need_retry = IS_ENABLED(CONFIG_MODULES);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001602 struct linux_binfmt *fmt;
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001603 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001604
Kees Cookd7402692012-12-17 16:03:20 -08001605 /* This allows 4 levels of binfmt rewrites before failing hard. */
Oleg Nesterov131b2f92013-09-11 14:24:39 -07001606 if (bprm->recursion_depth > 5)
Kees Cookd7402692012-12-17 16:03:20 -08001607 return -ELOOP;
1608
Linus Torvalds1da177e2005-04-16 15:20:36 -07001609 retval = security_bprm_check(bprm);
1610 if (retval)
1611 return retval;
1612
Linus Torvalds1da177e2005-04-16 15:20:36 -07001613 retval = -ENOENT;
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001614 retry:
1615 read_lock(&binfmt_lock);
1616 list_for_each_entry(fmt, &formats, lh) {
1617 if (!try_module_get(fmt->module))
1618 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001619 read_unlock(&binfmt_lock);
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001620 bprm->recursion_depth++;
1621 retval = fmt->load_binary(bprm);
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001622 read_lock(&binfmt_lock);
1623 put_binfmt(fmt);
Al Viro19d860a2014-05-04 20:11:36 -04001624 bprm->recursion_depth--;
1625 if (retval < 0 && !bprm->mm) {
1626 /* we got to flush_old_exec() and failed after it */
1627 read_unlock(&binfmt_lock);
1628 force_sigsegv(SIGSEGV, current);
1629 return retval;
1630 }
1631 if (retval != -ENOEXEC || !bprm->file) {
1632 read_unlock(&binfmt_lock);
1633 return retval;
1634 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001635 }
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001636 read_unlock(&binfmt_lock);
1637
Al Viro19d860a2014-05-04 20:11:36 -04001638 if (need_retry) {
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001639 if (printable(bprm->buf[0]) && printable(bprm->buf[1]) &&
1640 printable(bprm->buf[2]) && printable(bprm->buf[3]))
1641 return retval;
Oleg Nesterov4e0621a2013-09-11 14:24:45 -07001642 if (request_module("binfmt-%04x", *(ushort *)(bprm->buf + 2)) < 0)
1643 return retval;
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001644 need_retry = false;
1645 goto retry;
1646 }
1647
Linus Torvalds1da177e2005-04-16 15:20:36 -07001648 return retval;
1649}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001650EXPORT_SYMBOL(search_binary_handler);
1651
Oleg Nesterov5d1baf32013-09-11 14:24:38 -07001652static int exec_binprm(struct linux_binprm *bprm)
1653{
1654 pid_t old_pid, old_vpid;
1655 int ret;
1656
1657 /* Need to fetch pid before load_binary changes it */
1658 old_pid = current->pid;
1659 rcu_read_lock();
1660 old_vpid = task_pid_nr_ns(current, task_active_pid_ns(current->parent));
1661 rcu_read_unlock();
1662
1663 ret = search_binary_handler(bprm);
1664 if (ret >= 0) {
Linus Torvalds3eaded862013-11-21 19:18:14 -08001665 audit_bprm(bprm);
Oleg Nesterov5d1baf32013-09-11 14:24:38 -07001666 trace_sched_process_exec(current, old_pid, bprm);
1667 ptrace_event(PTRACE_EVENT_EXEC, old_vpid);
Oleg Nesterov9beb2662013-09-11 14:24:40 -07001668 proc_exec_connector(current);
Oleg Nesterov5d1baf32013-09-11 14:24:38 -07001669 }
1670
1671 return ret;
1672}
1673
Linus Torvalds1da177e2005-04-16 15:20:36 -07001674/*
1675 * sys_execve() executes a new program.
1676 */
David Drysdale51f39a12014-12-12 16:57:29 -08001677static int do_execveat_common(int fd, struct filename *filename,
1678 struct user_arg_ptr argv,
1679 struct user_arg_ptr envp,
1680 int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001681{
David Drysdale51f39a12014-12-12 16:57:29 -08001682 char *pathbuf = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001683 struct linux_binprm *bprm;
1684 struct file *file;
Al Viro3b125382008-04-22 05:31:30 -04001685 struct files_struct *displaced;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001686 int retval;
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +04001687
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001688 if (IS_ERR(filename))
1689 return PTR_ERR(filename);
1690
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +04001691 /*
1692 * We move the actual failure in case of RLIMIT_NPROC excess from
1693 * set*uid() to execve() because too many poorly written programs
1694 * don't check setuid() return code. Here we additionally recheck
1695 * whether NPROC limit is still exceeded.
1696 */
1697 if ((current->flags & PF_NPROC_EXCEEDED) &&
Oleg Nesterovbd9d43f2013-07-03 15:08:34 -07001698 atomic_read(&current_user()->processes) > rlimit(RLIMIT_NPROC)) {
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +04001699 retval = -EAGAIN;
1700 goto out_ret;
1701 }
1702
1703 /* We're below the limit (still or again), so we don't want to make
1704 * further execve() calls fail. */
1705 current->flags &= ~PF_NPROC_EXCEEDED;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001706
Al Viro3b125382008-04-22 05:31:30 -04001707 retval = unshare_files(&displaced);
Al Virofd8328b2008-04-22 05:11:59 -04001708 if (retval)
1709 goto out_ret;
1710
Linus Torvalds1da177e2005-04-16 15:20:36 -07001711 retval = -ENOMEM;
Oliver Neukum11b0b5a2006-03-25 03:08:13 -08001712 bprm = kzalloc(sizeof(*bprm), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001713 if (!bprm)
Al Virofd8328b2008-04-22 05:11:59 -04001714 goto out_files;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001715
Oleg Nesterova2a84742009-09-05 11:17:13 -07001716 retval = prepare_bprm_creds(bprm);
1717 if (retval)
David Howellsa6f76f22008-11-14 10:39:24 +11001718 goto out_free;
Al Viro498052b2009-03-30 07:20:30 -04001719
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001720 check_unsafe_exec(bprm);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001721 current->in_execve = 1;
David Howellsa6f76f22008-11-14 10:39:24 +11001722
David Drysdale51f39a12014-12-12 16:57:29 -08001723 file = do_open_execat(fd, filename, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001724 retval = PTR_ERR(file);
1725 if (IS_ERR(file))
Al Viro498052b2009-03-30 07:20:30 -04001726 goto out_unmark;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001727
1728 sched_exec();
1729
Linus Torvalds1da177e2005-04-16 15:20:36 -07001730 bprm->file = file;
David Drysdale51f39a12014-12-12 16:57:29 -08001731 if (fd == AT_FDCWD || filename->name[0] == '/') {
1732 bprm->filename = filename->name;
1733 } else {
1734 if (filename->name[0] == '\0')
1735 pathbuf = kasprintf(GFP_TEMPORARY, "/dev/fd/%d", fd);
1736 else
1737 pathbuf = kasprintf(GFP_TEMPORARY, "/dev/fd/%d/%s",
1738 fd, filename->name);
1739 if (!pathbuf) {
1740 retval = -ENOMEM;
1741 goto out_unmark;
1742 }
1743 /*
1744 * Record that a name derived from an O_CLOEXEC fd will be
1745 * inaccessible after exec. Relies on having exclusive access to
1746 * current->files (due to unshare_files above).
1747 */
1748 if (close_on_exec(fd, rcu_dereference_raw(current->files->fdt)))
1749 bprm->interp_flags |= BINPRM_FLAGS_PATH_INACCESSIBLE;
1750 bprm->filename = pathbuf;
1751 }
1752 bprm->interp = bprm->filename;
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001753
1754 retval = bprm_mm_init(bprm);
1755 if (retval)
Oleg Nesterov63e46b92014-01-23 15:55:51 -08001756 goto out_unmark;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001757
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001758 bprm->argc = count(argv, MAX_ARG_STRINGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001759 if ((retval = bprm->argc) < 0)
David Howellsa6f76f22008-11-14 10:39:24 +11001760 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001761
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001762 bprm->envc = count(envp, MAX_ARG_STRINGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001763 if ((retval = bprm->envc) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001764 goto out;
1765
1766 retval = prepare_binprm(bprm);
1767 if (retval < 0)
1768 goto out;
1769
1770 retval = copy_strings_kernel(1, &bprm->filename, bprm);
1771 if (retval < 0)
1772 goto out;
1773
1774 bprm->exec = bprm->p;
1775 retval = copy_strings(bprm->envc, envp, bprm);
1776 if (retval < 0)
1777 goto out;
1778
1779 retval = copy_strings(bprm->argc, argv, bprm);
1780 if (retval < 0)
1781 goto out;
1782
Eric W. Biederman21245b82016-11-16 22:06:51 -06001783 would_dump(bprm, bprm->file);
1784
Oleg Nesterov5d1baf32013-09-11 14:24:38 -07001785 retval = exec_binprm(bprm);
David Howellsa6f76f22008-11-14 10:39:24 +11001786 if (retval < 0)
1787 goto out;
1788
1789 /* execve succeeded */
Al Viro498052b2009-03-30 07:20:30 -04001790 current->fs->in_exec = 0;
Kentaro Takedaf9ce1f12009-02-05 17:18:11 +09001791 current->in_execve = 0;
David Howellsa6f76f22008-11-14 10:39:24 +11001792 acct_update_integrals(current);
Rik van Riel82727012013-10-07 11:29:28 +01001793 task_numa_free(current);
David Howellsa6f76f22008-11-14 10:39:24 +11001794 free_bprm(bprm);
David Drysdale51f39a12014-12-12 16:57:29 -08001795 kfree(pathbuf);
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001796 putname(filename);
David Howellsa6f76f22008-11-14 10:39:24 +11001797 if (displaced)
1798 put_files_struct(displaced);
1799 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001800
1801out:
Oleg Nesterov3c77f842010-11-30 20:55:34 +01001802 if (bprm->mm) {
1803 acct_arg_size(bprm, 0);
1804 mmput(bprm->mm);
1805 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001806
Al Viro498052b2009-03-30 07:20:30 -04001807out_unmark:
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001808 current->fs->in_exec = 0;
Kentaro Takedaf9ce1f12009-02-05 17:18:11 +09001809 current->in_execve = 0;
David Howellsa6f76f22008-11-14 10:39:24 +11001810
1811out_free:
Al Viro08a6fac2008-05-10 16:38:25 -04001812 free_bprm(bprm);
David Drysdale51f39a12014-12-12 16:57:29 -08001813 kfree(pathbuf);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001814
Al Virofd8328b2008-04-22 05:11:59 -04001815out_files:
Al Viro3b125382008-04-22 05:31:30 -04001816 if (displaced)
1817 reset_files_struct(displaced);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001818out_ret:
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001819 putname(filename);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001820 return retval;
1821}
1822
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001823int do_execve(struct filename *filename,
Oleg Nesterovba2d01622011-03-06 18:02:37 +01001824 const char __user *const __user *__argv,
Al Viroda3d4c52012-10-20 21:49:33 -04001825 const char __user *const __user *__envp)
Oleg Nesterovba2d01622011-03-06 18:02:37 +01001826{
Oleg Nesterov0e028462011-03-06 18:02:54 +01001827 struct user_arg_ptr argv = { .ptr.native = __argv };
1828 struct user_arg_ptr envp = { .ptr.native = __envp };
David Drysdale51f39a12014-12-12 16:57:29 -08001829 return do_execveat_common(AT_FDCWD, filename, argv, envp, 0);
1830}
1831
1832int do_execveat(int fd, struct filename *filename,
1833 const char __user *const __user *__argv,
1834 const char __user *const __user *__envp,
1835 int flags)
1836{
1837 struct user_arg_ptr argv = { .ptr.native = __argv };
1838 struct user_arg_ptr envp = { .ptr.native = __envp };
1839
1840 return do_execveat_common(fd, filename, argv, envp, flags);
Oleg Nesterovba2d01622011-03-06 18:02:37 +01001841}
1842
Oleg Nesterov0e028462011-03-06 18:02:54 +01001843#ifdef CONFIG_COMPAT
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001844static int compat_do_execve(struct filename *filename,
Al Viro38b983b2012-09-30 13:38:55 -04001845 const compat_uptr_t __user *__argv,
Al Virod03d26e2012-10-20 21:46:25 -04001846 const compat_uptr_t __user *__envp)
Oleg Nesterov0e028462011-03-06 18:02:54 +01001847{
1848 struct user_arg_ptr argv = {
1849 .is_compat = true,
1850 .ptr.compat = __argv,
1851 };
1852 struct user_arg_ptr envp = {
1853 .is_compat = true,
1854 .ptr.compat = __envp,
1855 };
David Drysdale51f39a12014-12-12 16:57:29 -08001856 return do_execveat_common(AT_FDCWD, filename, argv, envp, 0);
1857}
1858
1859static int compat_do_execveat(int fd, struct filename *filename,
1860 const compat_uptr_t __user *__argv,
1861 const compat_uptr_t __user *__envp,
1862 int flags)
1863{
1864 struct user_arg_ptr argv = {
1865 .is_compat = true,
1866 .ptr.compat = __argv,
1867 };
1868 struct user_arg_ptr envp = {
1869 .is_compat = true,
1870 .ptr.compat = __envp,
1871 };
1872 return do_execveat_common(fd, filename, argv, envp, flags);
Oleg Nesterov0e028462011-03-06 18:02:54 +01001873}
1874#endif
1875
Oleg Nesterov964ee7d2009-09-23 15:56:59 -07001876void set_binfmt(struct linux_binfmt *new)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001877{
Hiroshi Shimamoto801460d2009-09-23 15:57:41 -07001878 struct mm_struct *mm = current->mm;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001879
Hiroshi Shimamoto801460d2009-09-23 15:57:41 -07001880 if (mm->binfmt)
1881 module_put(mm->binfmt->module);
1882
1883 mm->binfmt = new;
Oleg Nesterov964ee7d2009-09-23 15:56:59 -07001884 if (new)
1885 __module_get(new->module);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001886}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001887EXPORT_SYMBOL(set_binfmt);
1888
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001889/*
Oleg Nesterov7288e112014-01-23 15:55:32 -08001890 * set_dumpable stores three-value SUID_DUMP_* into mm->flags.
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001891 */
1892void set_dumpable(struct mm_struct *mm, int value)
1893{
Oleg Nesterovabacd2f2014-01-23 15:55:31 -08001894 unsigned long old, new;
1895
Oleg Nesterov7288e112014-01-23 15:55:32 -08001896 if (WARN_ON((unsigned)value > SUID_DUMP_ROOT))
1897 return;
1898
Oleg Nesterovabacd2f2014-01-23 15:55:31 -08001899 do {
1900 old = ACCESS_ONCE(mm->flags);
Oleg Nesterov7288e112014-01-23 15:55:32 -08001901 new = (old & ~MMF_DUMPABLE_MASK) | value;
Oleg Nesterovabacd2f2014-01-23 15:55:31 -08001902 } while (cmpxchg(&mm->flags, old, new) != old);
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001903}
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001904
Al Viro38b983b2012-09-30 13:38:55 -04001905SYSCALL_DEFINE3(execve,
1906 const char __user *, filename,
1907 const char __user *const __user *, argv,
1908 const char __user *const __user *, envp)
1909{
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001910 return do_execve(getname(filename), argv, envp);
Al Viro38b983b2012-09-30 13:38:55 -04001911}
David Drysdale51f39a12014-12-12 16:57:29 -08001912
1913SYSCALL_DEFINE5(execveat,
1914 int, fd, const char __user *, filename,
1915 const char __user *const __user *, argv,
1916 const char __user *const __user *, envp,
1917 int, flags)
1918{
1919 int lookup_flags = (flags & AT_EMPTY_PATH) ? LOOKUP_EMPTY : 0;
1920
1921 return do_execveat(fd,
1922 getname_flags(filename, lookup_flags, NULL),
1923 argv, envp, flags);
1924}
1925
Al Viro38b983b2012-09-30 13:38:55 -04001926#ifdef CONFIG_COMPAT
Heiko Carstens625b1d72014-03-04 10:53:50 +01001927COMPAT_SYSCALL_DEFINE3(execve, const char __user *, filename,
1928 const compat_uptr_t __user *, argv,
1929 const compat_uptr_t __user *, envp)
Al Viro38b983b2012-09-30 13:38:55 -04001930{
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001931 return compat_do_execve(getname(filename), argv, envp);
Al Viro38b983b2012-09-30 13:38:55 -04001932}
David Drysdale51f39a12014-12-12 16:57:29 -08001933
1934COMPAT_SYSCALL_DEFINE5(execveat, int, fd,
1935 const char __user *, filename,
1936 const compat_uptr_t __user *, argv,
1937 const compat_uptr_t __user *, envp,
1938 int, flags)
1939{
1940 int lookup_flags = (flags & AT_EMPTY_PATH) ? LOOKUP_EMPTY : 0;
1941
1942 return compat_do_execveat(fd,
1943 getname_flags(filename, lookup_flags, NULL),
1944 argv, envp, flags);
1945}
Al Viro38b983b2012-09-30 13:38:55 -04001946#endif