blob: b8c43be24751fdedbdce5d8f2fdf9c444465a6d8 [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 Cookf31c4f62017-07-07 11:57:29 -0700218 unsigned long ptr_size, limit;
Linus Torvaldsa64e7152008-03-03 10:12:14 -0800219
Kees Cook3d6848e2017-06-23 15:08:57 -0700220 /*
221 * Since the stack will hold pointers to the strings, we
222 * must account for them as well.
223 *
224 * The size calculation is the entire vma while each arg page is
225 * built, so each time we get here it's calculating how far it
226 * is currently (rather than each call being just the newly
227 * added size from the arg page). As a result, we need to
228 * always add the entire size of the pointers, so that on the
229 * last call to get_arg_page() we'll actually have the entire
230 * correct size.
231 */
232 ptr_size = (bprm->argc + bprm->envc) * sizeof(void *);
233 if (ptr_size > ULONG_MAX - size)
234 goto fail;
235 size += ptr_size;
236
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100237 acct_arg_size(bprm, size / PAGE_SIZE);
238
Linus Torvaldsa64e7152008-03-03 10:12:14 -0800239 /*
240 * We've historically supported up to 32 pages (ARG_MAX)
241 * of argument strings even with small stacks
242 */
243 if (size <= ARG_MAX)
244 return page;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700245
246 /*
Kees Cookf31c4f62017-07-07 11:57:29 -0700247 * Limit to 1/4 of the max stack size or 3/4 of _STK_LIM
248 * (whichever is smaller) for the argv+env strings.
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700249 * 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 */
Kees Cookf31c4f62017-07-07 11:57:29 -0700254 limit = _STK_LIM / 4 * 3;
255 limit = min(limit, rlimit(RLIMIT_STACK) / 4);
256 if (size > limit)
Kees Cook3d6848e2017-06-23 15:08:57 -0700257 goto fail;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700258 }
259
260 return page;
Kees Cook3d6848e2017-06-23 15:08:57 -0700261
262fail:
263 put_page(page);
264 return NULL;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700265}
266
267static void put_arg_page(struct page *page)
268{
269 put_page(page);
270}
271
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700272static void free_arg_pages(struct linux_binprm *bprm)
273{
274}
275
276static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos,
277 struct page *page)
278{
279 flush_cache_page(bprm->vma, pos, page_to_pfn(page));
280}
281
282static int __bprm_mm_init(struct linux_binprm *bprm)
283{
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800284 int err;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700285 struct vm_area_struct *vma = NULL;
286 struct mm_struct *mm = bprm->mm;
287
288 bprm->vma = vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL);
289 if (!vma)
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800290 return -ENOMEM;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700291
Michal Hockof268dfe2016-05-23 16:26:02 -0700292 if (down_write_killable(&mm->mmap_sem)) {
293 err = -EINTR;
294 goto err_free;
295 }
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700296 vma->vm_mm = mm;
297
298 /*
299 * Place the stack at the largest stack address the architecture
300 * supports. Later, we'll move this to an appropriate place. We don't
301 * use STACK_TOP because that can depend on attributes which aren't
302 * configured yet.
303 */
Michal Hockoaacb3d12011-07-26 16:08:40 -0700304 BUILD_BUG_ON(VM_STACK_FLAGS & VM_STACK_INCOMPLETE_SETUP);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700305 vma->vm_end = STACK_TOP_MAX;
306 vma->vm_start = vma->vm_end - PAGE_SIZE;
Cyrill Gorcunovd9104d12013-09-11 14:22:24 -0700307 vma->vm_flags = VM_SOFTDIRTY | VM_STACK_FLAGS | VM_STACK_INCOMPLETE_SETUP;
Coly Li3ed75eb2007-10-18 23:39:15 -0700308 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
Rik van Riel5beb4932010-03-05 13:42:07 -0800309 INIT_LIST_HEAD(&vma->anon_vma_chain);
Tavis Ormandy462e635e2010-12-09 15:29:42 +0100310
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700311 err = insert_vm_struct(mm, vma);
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800312 if (err)
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700313 goto err;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700314
315 mm->stack_vm = mm->total_vm = 1;
Dave Hansenfe3d1972014-11-14 07:18:29 -0800316 arch_bprm_mm_init(mm, vma);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700317 up_write(&mm->mmap_sem);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700318 bprm->p = vma->vm_end - sizeof(void *);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700319 return 0;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700320err:
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800321 up_write(&mm->mmap_sem);
Michal Hockof268dfe2016-05-23 16:26:02 -0700322err_free:
Luiz Fernando N. Capitulinoeaccbfa2009-01-06 14:40:44 -0800323 bprm->vma = NULL;
324 kmem_cache_free(vm_area_cachep, vma);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700325 return err;
326}
327
328static bool valid_arg_len(struct linux_binprm *bprm, long len)
329{
330 return len <= MAX_ARG_STRLEN;
331}
332
333#else
334
Oleg Nesterov0e028462011-03-06 18:02:54 +0100335static inline void acct_arg_size(struct linux_binprm *bprm, unsigned long pages)
Oleg Nesterov3c77f842010-11-30 20:55:34 +0100336{
337}
338
Oleg Nesterov0e028462011-03-06 18:02:54 +0100339static struct page *get_arg_page(struct linux_binprm *bprm, unsigned long pos,
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700340 int write)
341{
342 struct page *page;
343
344 page = bprm->page[pos / PAGE_SIZE];
345 if (!page && write) {
346 page = alloc_page(GFP_HIGHUSER|__GFP_ZERO);
347 if (!page)
348 return NULL;
349 bprm->page[pos / PAGE_SIZE] = page;
350 }
351
352 return page;
353}
354
355static void put_arg_page(struct page *page)
356{
357}
358
359static void free_arg_page(struct linux_binprm *bprm, int i)
360{
361 if (bprm->page[i]) {
362 __free_page(bprm->page[i]);
363 bprm->page[i] = NULL;
364 }
365}
366
367static void free_arg_pages(struct linux_binprm *bprm)
368{
369 int i;
370
371 for (i = 0; i < MAX_ARG_PAGES; i++)
372 free_arg_page(bprm, i);
373}
374
375static void flush_arg_page(struct linux_binprm *bprm, unsigned long pos,
376 struct page *page)
377{
378}
379
380static int __bprm_mm_init(struct linux_binprm *bprm)
381{
382 bprm->p = PAGE_SIZE * MAX_ARG_PAGES - sizeof(void *);
383 return 0;
384}
385
386static bool valid_arg_len(struct linux_binprm *bprm, long len)
387{
388 return len <= bprm->p;
389}
390
391#endif /* CONFIG_MMU */
392
393/*
394 * Create a new mm_struct and populate it with a temporary stack
395 * vm_area_struct. We don't have enough context at this point to set the stack
396 * flags, permissions, and offset, so we use temporary values. We'll update
397 * them later in setup_arg_pages().
398 */
Yuanhan Liu9cc64ce2013-02-20 13:16:01 +1100399static int bprm_mm_init(struct linux_binprm *bprm)
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700400{
401 int err;
402 struct mm_struct *mm = NULL;
403
404 bprm->mm = mm = mm_alloc();
405 err = -ENOMEM;
406 if (!mm)
407 goto err;
408
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700409 err = __bprm_mm_init(bprm);
410 if (err)
411 goto err;
412
413 return 0;
414
415err:
416 if (mm) {
417 bprm->mm = NULL;
418 mmdrop(mm);
419 }
420
421 return err;
422}
423
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100424struct user_arg_ptr {
Oleg Nesterov0e028462011-03-06 18:02:54 +0100425#ifdef CONFIG_COMPAT
426 bool is_compat;
427#endif
428 union {
429 const char __user *const __user *native;
430#ifdef CONFIG_COMPAT
Al Viro38b983b2012-09-30 13:38:55 -0400431 const compat_uptr_t __user *compat;
Oleg Nesterov0e028462011-03-06 18:02:54 +0100432#endif
433 } ptr;
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100434};
435
436static const char __user *get_user_arg_ptr(struct user_arg_ptr argv, int nr)
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100437{
Oleg Nesterov0e028462011-03-06 18:02:54 +0100438 const char __user *native;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100439
Oleg Nesterov0e028462011-03-06 18:02:54 +0100440#ifdef CONFIG_COMPAT
441 if (unlikely(argv.is_compat)) {
442 compat_uptr_t compat;
443
444 if (get_user(compat, argv.ptr.compat + nr))
445 return ERR_PTR(-EFAULT);
446
447 return compat_ptr(compat);
448 }
449#endif
450
451 if (get_user(native, argv.ptr.native + nr))
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100452 return ERR_PTR(-EFAULT);
453
Oleg Nesterov0e028462011-03-06 18:02:54 +0100454 return native;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100455}
456
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457/*
458 * count() counts the number of strings in array ARGV.
459 */
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100460static int count(struct user_arg_ptr argv, int max)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700461{
462 int i = 0;
463
Oleg Nesterov0e028462011-03-06 18:02:54 +0100464 if (argv.ptr.native != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465 for (;;) {
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100466 const char __user *p = get_user_arg_ptr(argv, i);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468 if (!p)
469 break;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100470
471 if (IS_ERR(p))
472 return -EFAULT;
473
Xi Wang6d92d4f2013-01-11 14:31:48 -0800474 if (i >= max)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700475 return -E2BIG;
Xi Wang6d92d4f2013-01-11 14:31:48 -0800476 ++i;
Roland McGrath9aea5a62010-09-07 19:37:06 -0700477
478 if (fatal_signal_pending(current))
479 return -ERESTARTNOHAND;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480 cond_resched();
481 }
482 }
483 return i;
484}
485
486/*
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700487 * 'copy_strings()' copies argument/environment strings from the old
488 * processes's memory to the new process's stack. The call to get_user_pages()
489 * ensures the destination page is created and not swapped out.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490 */
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100491static int copy_strings(int argc, struct user_arg_ptr argv,
Adrian Bunk75c96f82005-05-05 16:16:09 -0700492 struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493{
494 struct page *kmapped_page = NULL;
495 char *kaddr = NULL;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700496 unsigned long kpos = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497 int ret;
498
499 while (argc-- > 0) {
David Howellsd7627462010-08-17 23:52:56 +0100500 const char __user *str;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501 int len;
502 unsigned long pos;
503
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100504 ret = -EFAULT;
505 str = get_user_arg_ptr(argv, argc);
506 if (IS_ERR(str))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100509 len = strnlen_user(str, MAX_ARG_STRLEN);
510 if (!len)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511 goto out;
Oleg Nesterov1d1dbf82011-03-06 18:02:21 +0100512
513 ret = -E2BIG;
514 if (!valid_arg_len(bprm, len))
515 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700517 /* We're going to work our way backwords. */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518 pos = bprm->p;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700519 str += len;
520 bprm->p -= len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521
522 while (len > 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523 int offset, bytes_to_copy;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700524
Roland McGrath9aea5a62010-09-07 19:37:06 -0700525 if (fatal_signal_pending(current)) {
526 ret = -ERESTARTNOHAND;
527 goto out;
528 }
Roland McGrath7993bc12010-09-07 19:36:28 -0700529 cond_resched();
530
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531 offset = pos % PAGE_SIZE;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700532 if (offset == 0)
533 offset = PAGE_SIZE;
534
535 bytes_to_copy = offset;
536 if (bytes_to_copy > len)
537 bytes_to_copy = len;
538
539 offset -= bytes_to_copy;
540 pos -= bytes_to_copy;
541 str -= bytes_to_copy;
542 len -= bytes_to_copy;
543
544 if (!kmapped_page || kpos != (pos & PAGE_MASK)) {
545 struct page *page;
546
547 page = get_arg_page(bprm, pos, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700548 if (!page) {
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700549 ret = -E2BIG;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550 goto out;
551 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700553 if (kmapped_page) {
554 flush_kernel_dcache_page(kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700555 kunmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700556 put_arg_page(kmapped_page);
557 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700558 kmapped_page = page;
559 kaddr = kmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700560 kpos = pos & PAGE_MASK;
561 flush_arg_page(bprm, kpos, kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700562 }
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700563 if (copy_from_user(kaddr+offset, str, bytes_to_copy)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564 ret = -EFAULT;
565 goto out;
566 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700567 }
568 }
569 ret = 0;
570out:
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700571 if (kmapped_page) {
572 flush_kernel_dcache_page(kmapped_page);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573 kunmap(kmapped_page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700574 put_arg_page(kmapped_page);
575 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576 return ret;
577}
578
579/*
580 * Like copy_strings, but get argv and its values from kernel memory.
581 */
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100582int copy_strings_kernel(int argc, const char *const *__argv,
David Howellsd7627462010-08-17 23:52:56 +0100583 struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584{
585 int r;
586 mm_segment_t oldfs = get_fs();
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100587 struct user_arg_ptr argv = {
Oleg Nesterov0e028462011-03-06 18:02:54 +0100588 .ptr.native = (const char __user *const __user *)__argv,
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100589 };
590
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591 set_fs(KERNEL_DS);
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100592 r = copy_strings(argc, argv, bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700593 set_fs(oldfs);
Oleg Nesterovba2d01622011-03-06 18:02:37 +0100594
Linus Torvalds1da177e2005-04-16 15:20:36 -0700595 return r;
596}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700597EXPORT_SYMBOL(copy_strings_kernel);
598
599#ifdef CONFIG_MMU
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700600
Linus Torvalds1da177e2005-04-16 15:20:36 -0700601/*
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700602 * During bprm_mm_init(), we create a temporary stack at STACK_TOP_MAX. Once
603 * the binfmt code determines where the new stack should reside, we shift it to
604 * its final location. The process proceeds as follows:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700605 *
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700606 * 1) Use shift to calculate the new vma endpoints.
607 * 2) Extend vma to cover both the old and new ranges. This ensures the
608 * arguments passed to subsequent functions are consistent.
609 * 3) Move vma's page tables to the new range.
610 * 4) Free up any cleared pgd range.
611 * 5) Shrink the vma to cover only the new range.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700612 */
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700613static int shift_arg_pages(struct vm_area_struct *vma, unsigned long shift)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700614{
615 struct mm_struct *mm = vma->vm_mm;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700616 unsigned long old_start = vma->vm_start;
617 unsigned long old_end = vma->vm_end;
618 unsigned long length = old_end - old_start;
619 unsigned long new_start = old_start - shift;
620 unsigned long new_end = old_end - shift;
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700621 struct mmu_gather tlb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700622
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700623 BUG_ON(new_start > new_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700625 /*
626 * ensure there are no vmas between where we want to go
627 * and where we are
628 */
629 if (vma != find_vma(mm, new_start))
630 return -EFAULT;
631
632 /*
633 * cover the whole range: [new_start, old_end)
634 */
Rik van Riel5beb4932010-03-05 13:42:07 -0800635 if (vma_adjust(vma, new_start, old_end, vma->vm_pgoff, NULL))
636 return -ENOMEM;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700637
638 /*
639 * move the page tables downwards, on failure we rely on
640 * process cleanup to remove whatever mess we made.
641 */
642 if (length != move_page_tables(vma, old_start,
Michel Lespinasse38a76012012-10-08 16:31:50 -0700643 vma, new_start, length, false))
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700644 return -ENOMEM;
645
646 lru_add_drain();
Linus Torvalds2b047252013-08-15 11:42:25 -0700647 tlb_gather_mmu(&tlb, mm, old_start, old_end);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700648 if (new_end > old_start) {
649 /*
650 * when the old and new regions overlap clear from new_end.
651 */
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700652 free_pgd_range(&tlb, new_end, old_end, new_end,
Hugh Dickins6ee86302013-04-29 15:07:44 -0700653 vma->vm_next ? vma->vm_next->vm_start : USER_PGTABLES_CEILING);
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700654 } else {
655 /*
656 * otherwise, clean from old_start; this is done to not touch
657 * the address space in [new_end, old_start) some architectures
658 * have constraints on va-space that make this illegal (IA64) -
659 * for the others its just a little faster.
660 */
Peter Zijlstrad16dfc52011-05-24 17:11:45 -0700661 free_pgd_range(&tlb, old_start, old_end, new_end,
Hugh Dickins6ee86302013-04-29 15:07:44 -0700662 vma->vm_next ? vma->vm_next->vm_start : USER_PGTABLES_CEILING);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663 }
Linus Torvalds2b047252013-08-15 11:42:25 -0700664 tlb_finish_mmu(&tlb, old_start, old_end);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700666 /*
Rik van Riel5beb4932010-03-05 13:42:07 -0800667 * Shrink the vma to just the new range. Always succeeds.
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700668 */
669 vma_adjust(vma, new_start, new_end, vma->vm_pgoff, NULL);
670
671 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700672}
673
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700674/*
675 * Finalizes the stack vm_area_struct. The flags and permissions are updated,
676 * the stack is optionally relocated, and some extra space is added.
677 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700678int setup_arg_pages(struct linux_binprm *bprm,
679 unsigned long stack_top,
680 int executable_stack)
681{
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700682 unsigned long ret;
683 unsigned long stack_shift;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700684 struct mm_struct *mm = current->mm;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700685 struct vm_area_struct *vma = bprm->vma;
686 struct vm_area_struct *prev = NULL;
687 unsigned long vm_flags;
688 unsigned long stack_base;
Michael Neuling803bf5e2010-02-10 13:56:42 -0800689 unsigned long stack_size;
690 unsigned long stack_expand;
691 unsigned long rlim_stack;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692
693#ifdef CONFIG_STACK_GROWSUP
James Hogand71f2902014-05-13 23:58:24 +0100694 /* Limit stack size */
Jiri Slabyd554ed892010-03-05 13:42:42 -0800695 stack_base = rlimit_max(RLIMIT_STACK);
James Hogand71f2902014-05-13 23:58:24 +0100696 if (stack_base > STACK_SIZE_MAX)
697 stack_base = STACK_SIZE_MAX;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700698
Helge Dellerd045c772015-05-11 22:01:27 +0200699 /* Add space for stack randomization. */
700 stack_base += (STACK_RND_MASK << PAGE_SHIFT);
701
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700702 /* Make sure we didn't let the argument array grow too large. */
703 if (vma->vm_end - vma->vm_start > stack_base)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700704 return -ENOMEM;
705
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700706 stack_base = PAGE_ALIGN(stack_top - stack_base);
707
708 stack_shift = vma->vm_start - stack_base;
709 mm->arg_start = bprm->p - stack_shift;
710 bprm->p = vma->vm_end - stack_shift;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711#else
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700712 stack_top = arch_align_stack(stack_top);
713 stack_top = PAGE_ALIGN(stack_top);
Roland McGrath1b528182010-09-07 19:35:49 -0700714
715 if (unlikely(stack_top < mmap_min_addr) ||
716 unlikely(vma->vm_end - vma->vm_start >= stack_top - mmap_min_addr))
717 return -ENOMEM;
718
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700719 stack_shift = vma->vm_end - stack_top;
720
721 bprm->p -= stack_shift;
722 mm->arg_start = bprm->p;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700723#endif
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700724
725 if (bprm->loader)
726 bprm->loader -= stack_shift;
727 bprm->exec -= stack_shift;
728
Michal Hockof268dfe2016-05-23 16:26:02 -0700729 if (down_write_killable(&mm->mmap_sem))
730 return -EINTR;
731
Hugh Dickins96a8e132008-07-10 21:19:20 +0100732 vm_flags = VM_STACK_FLAGS;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700733
734 /*
735 * Adjust stack execute permissions; explicitly enable for
736 * EXSTACK_ENABLE_X, disable for EXSTACK_DISABLE_X and leave alone
737 * (arch default) otherwise.
738 */
739 if (unlikely(executable_stack == EXSTACK_ENABLE_X))
740 vm_flags |= VM_EXEC;
741 else if (executable_stack == EXSTACK_DISABLE_X)
742 vm_flags &= ~VM_EXEC;
743 vm_flags |= mm->def_flags;
Mel Gormana8bef8ff2010-05-24 14:32:24 -0700744 vm_flags |= VM_STACK_INCOMPLETE_SETUP;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700745
746 ret = mprotect_fixup(vma, &prev, vma->vm_start, vma->vm_end,
747 vm_flags);
748 if (ret)
749 goto out_unlock;
750 BUG_ON(prev != vma);
751
752 /* Move stack pages down in memory. */
753 if (stack_shift) {
754 ret = shift_arg_pages(vma, stack_shift);
Anton Blanchardfc63cf22009-11-11 14:26:48 -0800755 if (ret)
756 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700757 }
758
Mel Gormana8bef8ff2010-05-24 14:32:24 -0700759 /* mprotect_fixup is overkill to remove the temporary stack flags */
760 vma->vm_flags &= ~VM_STACK_INCOMPLETE_SETUP;
761
Michael Neuling5ef097d2010-03-05 13:42:57 -0800762 stack_expand = 131072UL; /* randomly 32*4k (or 2*64k) pages */
Michael Neuling803bf5e2010-02-10 13:56:42 -0800763 stack_size = vma->vm_end - vma->vm_start;
764 /*
765 * Align this down to a page boundary as expand_stack
766 * will align it up.
767 */
768 rlim_stack = rlimit(RLIMIT_STACK) & PAGE_MASK;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700769#ifdef CONFIG_STACK_GROWSUP
Michael Neuling803bf5e2010-02-10 13:56:42 -0800770 if (stack_size + stack_expand > rlim_stack)
771 stack_base = vma->vm_start + rlim_stack;
772 else
773 stack_base = vma->vm_end + stack_expand;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700774#else
Michael Neuling803bf5e2010-02-10 13:56:42 -0800775 if (stack_size + stack_expand > rlim_stack)
776 stack_base = vma->vm_end - rlim_stack;
777 else
778 stack_base = vma->vm_start - stack_expand;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700779#endif
Eric B Munson3af9e852010-05-18 15:30:49 +0100780 current->mm->start_stack = bprm->p;
Ollie Wildb6a2fea2007-07-19 01:48:16 -0700781 ret = expand_stack(vma, stack_base);
782 if (ret)
783 ret = -EFAULT;
784
785out_unlock:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700786 up_write(&mm->mmap_sem);
Anton Blanchardfc63cf22009-11-11 14:26:48 -0800787 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789EXPORT_SYMBOL(setup_arg_pages);
790
Nicolas Pitre7e7ec6a2016-07-24 11:30:18 -0400791#else
792
793/*
794 * Transfer the program arguments and environment from the holding pages
795 * onto the stack. The provided stack pointer is adjusted accordingly.
796 */
797int transfer_args_to_stack(struct linux_binprm *bprm,
798 unsigned long *sp_location)
799{
800 unsigned long index, stop, sp;
801 int ret = 0;
802
803 stop = bprm->p >> PAGE_SHIFT;
804 sp = *sp_location;
805
806 for (index = MAX_ARG_PAGES - 1; index >= stop; index--) {
807 unsigned int offset = index == stop ? bprm->p & ~PAGE_MASK : 0;
808 char *src = kmap(bprm->page[index]) + offset;
809 sp -= PAGE_SIZE - offset;
810 if (copy_to_user((void *) sp, src, PAGE_SIZE - offset) != 0)
811 ret = -EFAULT;
812 kunmap(bprm->page[index]);
813 if (ret)
814 goto out;
815 }
816
817 *sp_location = sp;
818
819out:
820 return ret;
821}
822EXPORT_SYMBOL(transfer_args_to_stack);
823
Linus Torvalds1da177e2005-04-16 15:20:36 -0700824#endif /* CONFIG_MMU */
825
David Drysdale51f39a12014-12-12 16:57:29 -0800826static struct file *do_open_execat(int fd, struct filename *name, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828 struct file *file;
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200829 int err;
David Drysdale51f39a12014-12-12 16:57:29 -0800830 struct open_flags open_exec_flags = {
Al Viro47c805d2011-02-23 17:44:09 -0500831 .open_flag = O_LARGEFILE | O_RDONLY | __FMODE_EXEC,
Al Viro62fb4a12015-12-26 22:33:24 -0500832 .acc_mode = MAY_EXEC,
Al Virof9652e12013-06-11 08:23:01 +0400833 .intent = LOOKUP_OPEN,
834 .lookup_flags = LOOKUP_FOLLOW,
Al Viro47c805d2011-02-23 17:44:09 -0500835 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836
David Drysdale51f39a12014-12-12 16:57:29 -0800837 if ((flags & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
838 return ERR_PTR(-EINVAL);
839 if (flags & AT_SYMLINK_NOFOLLOW)
840 open_exec_flags.lookup_flags &= ~LOOKUP_FOLLOW;
841 if (flags & AT_EMPTY_PATH)
842 open_exec_flags.lookup_flags |= LOOKUP_EMPTY;
843
844 file = do_filp_open(fd, name, &open_exec_flags);
Al Viro6e8341a2009-04-06 11:16:22 -0400845 if (IS_ERR(file))
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200846 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200848 err = -EACCES;
Al Viro496ad9a2013-01-23 17:07:38 -0500849 if (!S_ISREG(file_inode(file)->i_mode))
Al Viro6e8341a2009-04-06 11:16:22 -0400850 goto exit;
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200851
Eric W. Biederman90f85722015-06-29 14:42:03 -0500852 if (path_noexec(&file->f_path))
Al Viro6e8341a2009-04-06 11:16:22 -0400853 goto exit;
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200854
855 err = deny_write_access(file);
Al Viro6e8341a2009-04-06 11:16:22 -0400856 if (err)
857 goto exit;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858
David Drysdale51f39a12014-12-12 16:57:29 -0800859 if (name->name[0] != '\0')
860 fsnotify_open(file);
861
Al Viro6e8341a2009-04-06 11:16:22 -0400862out:
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200863 return file;
864
Al Viro6e8341a2009-04-06 11:16:22 -0400865exit:
866 fput(file);
Christoph Hellwige56b6a52008-05-19 07:53:34 +0200867 return ERR_PTR(err);
868}
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -0800869
870struct file *open_exec(const char *name)
871{
Paul Moore51689102015-01-22 00:00:03 -0500872 struct filename *filename = getname_kernel(name);
873 struct file *f = ERR_CAST(filename);
874
875 if (!IS_ERR(filename)) {
876 f = do_open_execat(AT_FDCWD, filename, 0);
877 putname(filename);
878 }
879 return f;
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -0800880}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700881EXPORT_SYMBOL(open_exec);
882
Mimi Zohar6777d772009-08-21 14:32:48 -0400883int kernel_read(struct file *file, loff_t offset,
884 char *addr, unsigned long count)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700885{
886 mm_segment_t old_fs;
887 loff_t pos = offset;
888 int result;
889
890 old_fs = get_fs();
891 set_fs(get_ds());
892 /* The cast to a user pointer is valid due to the set_fs() */
893 result = vfs_read(file, (void __user *)addr, count, &pos);
894 set_fs(old_fs);
895 return result;
896}
897
898EXPORT_SYMBOL(kernel_read);
899
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500900int kernel_read_file(struct file *file, void **buf, loff_t *size,
Mimi Zoharbc8ca5b2016-01-24 10:07:32 -0500901 loff_t max_size, enum kernel_read_file_id id)
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500902{
903 loff_t i_size, pos;
904 ssize_t bytes = 0;
905 int ret;
906
907 if (!S_ISREG(file_inode(file)->i_mode) || max_size < 0)
908 return -EINVAL;
909
Mimi Zohar39eeb4f2016-01-30 22:23:26 -0500910 ret = security_kernel_read_file(file, id);
911 if (ret)
912 return ret;
913
Dmitry Kasatkin39d637a2014-10-26 12:42:07 +0200914 ret = deny_write_access(file);
915 if (ret)
916 return ret;
917
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500918 i_size = i_size_read(file_inode(file));
Dmitry Kasatkin39d637a2014-10-26 12:42:07 +0200919 if (max_size > 0 && i_size > max_size) {
920 ret = -EFBIG;
921 goto out;
922 }
923 if (i_size <= 0) {
924 ret = -EINVAL;
925 goto out;
926 }
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500927
Stephen Boyda098ecd2016-08-02 14:04:28 -0700928 if (id != READING_FIRMWARE_PREALLOC_BUFFER)
929 *buf = vmalloc(i_size);
Dmitry Kasatkin39d637a2014-10-26 12:42:07 +0200930 if (!*buf) {
931 ret = -ENOMEM;
932 goto out;
933 }
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500934
935 pos = 0;
936 while (pos < i_size) {
937 bytes = kernel_read(file, pos, (char *)(*buf) + pos,
938 i_size - pos);
939 if (bytes < 0) {
940 ret = bytes;
941 goto out;
942 }
943
944 if (bytes == 0)
945 break;
946 pos += bytes;
947 }
948
949 if (pos != i_size) {
950 ret = -EIO;
Dmitry Kasatkin39d637a2014-10-26 12:42:07 +0200951 goto out_free;
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500952 }
953
Mimi Zoharbc8ca5b2016-01-24 10:07:32 -0500954 ret = security_kernel_post_read_file(file, *buf, i_size, id);
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500955 if (!ret)
956 *size = pos;
957
Dmitry Kasatkin39d637a2014-10-26 12:42:07 +0200958out_free:
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500959 if (ret < 0) {
Stephen Boyda098ecd2016-08-02 14:04:28 -0700960 if (id != READING_FIRMWARE_PREALLOC_BUFFER) {
961 vfree(*buf);
962 *buf = NULL;
963 }
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500964 }
Dmitry Kasatkin39d637a2014-10-26 12:42:07 +0200965
966out:
967 allow_write_access(file);
Mimi Zoharb44a7df2015-12-28 16:02:29 -0500968 return ret;
969}
970EXPORT_SYMBOL_GPL(kernel_read_file);
971
Mimi Zohar09596b92015-11-19 12:39:22 -0500972int kernel_read_file_from_path(char *path, void **buf, loff_t *size,
973 loff_t max_size, enum kernel_read_file_id id)
974{
975 struct file *file;
976 int ret;
977
978 if (!path || !*path)
979 return -EINVAL;
980
981 file = filp_open(path, O_RDONLY, 0);
982 if (IS_ERR(file))
983 return PTR_ERR(file);
984
985 ret = kernel_read_file(file, buf, size, max_size, id);
986 fput(file);
987 return ret;
988}
989EXPORT_SYMBOL_GPL(kernel_read_file_from_path);
990
Mimi Zoharb844f0e2016-02-01 08:36:21 -0500991int kernel_read_file_from_fd(int fd, void **buf, loff_t *size, loff_t max_size,
992 enum kernel_read_file_id id)
993{
994 struct fd f = fdget(fd);
995 int ret = -EBADF;
996
997 if (!f.file)
998 goto out;
999
1000 ret = kernel_read_file(f.file, buf, size, max_size, id);
1001out:
1002 fdput(f);
1003 return ret;
1004}
1005EXPORT_SYMBOL_GPL(kernel_read_file_from_fd);
1006
Al Viro3dc20cb2013-04-13 20:31:37 -04001007ssize_t read_code(struct file *file, unsigned long addr, loff_t pos, size_t len)
1008{
Al Viroec695572014-02-04 19:08:21 -05001009 ssize_t res = vfs_read(file, (void __user *)addr, len, &pos);
Al Viro3dc20cb2013-04-13 20:31:37 -04001010 if (res > 0)
1011 flush_icache_range(addr, addr + len);
1012 return res;
1013}
1014EXPORT_SYMBOL(read_code);
1015
Linus Torvalds1da177e2005-04-16 15:20:36 -07001016static int exec_mmap(struct mm_struct *mm)
1017{
1018 struct task_struct *tsk;
Davidlohr Bueso615d6e82014-04-07 15:37:25 -07001019 struct mm_struct *old_mm, *active_mm;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001020
1021 /* Notify parent that we're no longer interested in the old VM */
1022 tsk = current;
1023 old_mm = current->mm;
1024 mm_release(tsk, old_mm);
1025
1026 if (old_mm) {
Konstantin Khlebnikov4fe7efd2012-06-20 12:53:01 -07001027 sync_mm_rss(old_mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001028 /*
1029 * Make sure that if there is a core dump in progress
1030 * for the old mm, we get out and die instead of going
1031 * through with the exec. We must hold mmap_sem around
Oleg Nesterov999d9fc2008-07-25 01:47:41 -07001032 * checking core_state and changing tsk->mm.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033 */
1034 down_read(&old_mm->mmap_sem);
Oleg Nesterov999d9fc2008-07-25 01:47:41 -07001035 if (unlikely(old_mm->core_state)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036 up_read(&old_mm->mmap_sem);
1037 return -EINTR;
1038 }
1039 }
1040 task_lock(tsk);
1041 active_mm = tsk->active_mm;
1042 tsk->mm = mm;
1043 tsk->active_mm = mm;
1044 activate_mm(active_mm, mm);
Davidlohr Bueso615d6e82014-04-07 15:37:25 -07001045 tsk->mm->vmacache_seqnum = 0;
1046 vmacache_flush(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001047 task_unlock(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001048 if (old_mm) {
1049 up_read(&old_mm->mmap_sem);
Eric Sesterhenn7dddb122006-04-01 01:13:38 +02001050 BUG_ON(active_mm != old_mm);
Oleg Nesterov701085b2012-03-19 17:04:01 +01001051 setmax_mm_hiwater_rss(&tsk->signal->maxrss, old_mm);
Balbir Singh31a78f22008-09-28 23:09:31 +01001052 mm_update_next_owner(old_mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001053 mmput(old_mm);
1054 return 0;
1055 }
1056 mmdrop(active_mm);
1057 return 0;
1058}
1059
1060/*
1061 * This function makes sure the current process has its own signal table,
1062 * so that flush_signal_handlers can later reset the handlers without
1063 * disturbing other processes. (Other processes might share the signal
1064 * table via the CLONE_SIGHAND option to clone().)
1065 */
Arjan van de Ven858119e2006-01-14 13:20:43 -08001066static int de_thread(struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001067{
1068 struct signal_struct *sig = tsk->signal;
Oleg Nesterovb2c903b2007-10-16 23:27:22 -07001069 struct sighand_struct *oldsighand = tsk->sighand;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001070 spinlock_t *lock = &oldsighand->siglock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001072 if (thread_group_empty(tsk))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001073 goto no_thread_group;
1074
1075 /*
1076 * Kill all other threads in the thread group.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001077 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078 spin_lock_irq(lock);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -08001079 if (signal_group_exit(sig)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001080 /*
1081 * Another group action in progress, just
1082 * return so that the signal is processed.
1083 */
1084 spin_unlock_irq(lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001085 return -EAGAIN;
1086 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087
Oleg Nesterovd3441932010-05-26 14:43:11 -07001088 sig->group_exit_task = tsk;
1089 sig->notify_count = zap_other_threads(tsk);
1090 if (!thread_group_leader(tsk))
1091 sig->notify_count--;
1092
1093 while (sig->notify_count) {
Oleg Nesterovd5bbd432012-10-08 19:13:01 +02001094 __set_current_state(TASK_KILLABLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001095 spin_unlock_irq(lock);
1096 schedule();
Oleg Nesterovd5bbd432012-10-08 19:13:01 +02001097 if (unlikely(__fatal_signal_pending(tsk)))
1098 goto killed;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001099 spin_lock_irq(lock);
1100 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001101 spin_unlock_irq(lock);
1102
1103 /*
1104 * At this point all other threads have exited, all we have to
1105 * do is to wait for the thread group leader to become inactive,
1106 * and to assume its PID:
1107 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001108 if (!thread_group_leader(tsk)) {
Oleg Nesterov81879262008-12-01 14:18:16 -08001109 struct task_struct *leader = tsk->group_leader;
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001110
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001111 for (;;) {
Oleg Nesterove56fb282013-04-30 15:28:20 -07001112 threadgroup_change_begin(tsk);
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001113 write_lock_irq(&tasklist_lock);
Kirill Tkhaidfcce792015-04-16 12:48:01 -07001114 /*
1115 * Do this under tasklist_lock to ensure that
1116 * exit_notify() can't miss ->group_exit_task
1117 */
1118 sig->notify_count = -1;
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001119 if (likely(leader->exit_state))
1120 break;
Oleg Nesterovd5bbd432012-10-08 19:13:01 +02001121 __set_current_state(TASK_KILLABLE);
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001122 write_unlock_irq(&tasklist_lock);
Oleg Nesterove56fb282013-04-30 15:28:20 -07001123 threadgroup_change_end(tsk);
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001124 schedule();
Oleg Nesterovd5bbd432012-10-08 19:13:01 +02001125 if (unlikely(__fatal_signal_pending(tsk)))
1126 goto killed;
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001127 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001128
Roland McGrathf5e90282006-04-10 22:54:16 -07001129 /*
1130 * The only record we have of the real-time age of a
1131 * process, regardless of execs it's done, is start_time.
1132 * All the past CPU time is accumulated in signal_struct
1133 * from sister threads now dead. But in this non-leader
1134 * exec, nothing survives from the original leader thread,
1135 * whose birth marks the true age of this process now.
1136 * When we take on its identity by switching to its PID, we
1137 * also take its birthdate (always earlier than our own).
1138 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001139 tsk->start_time = leader->start_time;
Oleg Nesterov266b7a02013-07-03 15:08:35 -07001140 tsk->real_start_time = leader->real_start_time;
Roland McGrathf5e90282006-04-10 22:54:16 -07001141
Pavel Emelyanovbac0abd2007-10-18 23:40:18 -07001142 BUG_ON(!same_thread_group(leader, tsk));
1143 BUG_ON(has_group_leader_pid(tsk));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001144 /*
1145 * An exec() starts a new thread group with the
1146 * TGID of the previous thread group. Rehash the
1147 * two threads with a switched PID, and release
1148 * the former thread group leader:
1149 */
Eric W. Biedermand73d6522006-03-28 16:11:03 -08001150
1151 /* Become a process group leader with the old leader's pid.
Eric W. Biedermanc18258c2006-09-27 01:51:06 -07001152 * The old leader becomes a thread of the this thread group.
1153 * Note: The old leader also uses this pid until release_task
Eric W. Biedermand73d6522006-03-28 16:11:03 -08001154 * is called. Odd but simple and correct.
1155 */
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001156 tsk->pid = leader->pid;
Oleg Nesterov3f418542013-07-03 15:08:25 -07001157 change_pid(tsk, PIDTYPE_PID, task_pid(leader));
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001158 transfer_pid(leader, tsk, PIDTYPE_PGID);
1159 transfer_pid(leader, tsk, PIDTYPE_SID);
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -08001160
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001161 list_replace_rcu(&leader->tasks, &tsk->tasks);
Oleg Nesterov9cd80bb2009-12-17 15:27:15 -08001162 list_replace_init(&leader->sibling, &tsk->sibling);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001163
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001164 tsk->group_leader = tsk;
1165 leader->group_leader = tsk;
Eric W. Biedermande12a782006-04-10 17:16:49 -06001166
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001167 tsk->exit_signal = SIGCHLD;
Oleg Nesterov087806b2011-06-22 23:10:26 +02001168 leader->exit_signal = -1;
Oleg Nesterov962b5642005-11-23 13:37:43 -08001169
1170 BUG_ON(leader->exit_state != EXIT_ZOMBIE);
1171 leader->exit_state = EXIT_DEAD;
Oleg Nesteroveac1b5e2011-07-21 20:00:43 +02001172
1173 /*
1174 * We are going to release_task()->ptrace_unlink() silently,
1175 * the tracer can sleep in do_wait(). EXIT_DEAD guarantees
1176 * the tracer wont't block again waiting for this thread.
1177 */
1178 if (unlikely(leader->ptrace))
1179 __wake_up_parent(leader, leader->parent);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001180 write_unlock_irq(&tasklist_lock);
Oleg Nesterove56fb282013-04-30 15:28:20 -07001181 threadgroup_change_end(tsk);
Oleg Nesterov81879262008-12-01 14:18:16 -08001182
1183 release_task(leader);
Oleg Nesteroved5d2ca2008-02-04 22:27:24 -08001184 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001185
Oleg Nesterov6db840f2007-10-16 23:27:23 -07001186 sig->group_exit_task = NULL;
1187 sig->notify_count = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001188
1189no_thread_group:
Oleg Nesterove6368252012-03-19 17:03:22 +01001190 /* we have changed execution domain */
1191 tsk->exit_signal = SIGCHLD;
1192
Linus Torvalds1da177e2005-04-16 15:20:36 -07001193 exit_itimers(sig);
Oleg Nesterovcbaffba2008-05-26 20:55:42 +04001194 flush_itimer_signals();
Oleg Nesterov329f7db2005-11-07 21:12:43 +03001195
Oleg Nesterovb2c903b2007-10-16 23:27:22 -07001196 if (atomic_read(&oldsighand->count) != 1) {
1197 struct sighand_struct *newsighand;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001198 /*
Oleg Nesterovb2c903b2007-10-16 23:27:22 -07001199 * This ->sighand is shared with the CLONE_SIGHAND
1200 * but not CLONE_THREAD task, switch to the new one.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001201 */
Oleg Nesterovb2c903b2007-10-16 23:27:22 -07001202 newsighand = kmem_cache_alloc(sighand_cachep, GFP_KERNEL);
1203 if (!newsighand)
1204 return -ENOMEM;
1205
Linus Torvalds1da177e2005-04-16 15:20:36 -07001206 atomic_set(&newsighand->count, 1);
1207 memcpy(newsighand->action, oldsighand->action,
1208 sizeof(newsighand->action));
1209
1210 write_lock_irq(&tasklist_lock);
1211 spin_lock(&oldsighand->siglock);
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001212 rcu_assign_pointer(tsk->sighand, newsighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001213 spin_unlock(&oldsighand->siglock);
1214 write_unlock_irq(&tasklist_lock);
1215
Davide Libenzifba2afa2007-05-10 22:23:13 -07001216 __cleanup_sighand(oldsighand);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001217 }
1218
Eric W. Biedermanaafe6c22006-09-27 01:51:13 -07001219 BUG_ON(!thread_group_leader(tsk));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001220 return 0;
Oleg Nesterovd5bbd432012-10-08 19:13:01 +02001221
1222killed:
1223 /* protects against exit_notify() and __exit_signal() */
1224 read_lock(&tasklist_lock);
1225 sig->group_exit_task = NULL;
1226 sig->notify_count = 0;
1227 read_unlock(&tasklist_lock);
1228 return -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001229}
Oleg Nesterov0840a902007-10-16 23:27:22 -07001230
Andrew Morton59714d62008-02-04 22:27:21 -08001231char *get_task_comm(char *buf, struct task_struct *tsk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001232{
1233 /* buf must be at least sizeof(tsk->comm) in size */
1234 task_lock(tsk);
1235 strncpy(buf, tsk->comm, sizeof(tsk->comm));
1236 task_unlock(tsk);
Andrew Morton59714d62008-02-04 22:27:21 -08001237 return buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001238}
J Freyensee7d74f492011-05-06 16:56:47 -07001239EXPORT_SYMBOL_GPL(get_task_comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001240
Al Viro6a6d27d2012-08-21 09:56:33 -04001241/*
1242 * These functions flushes out all traces of the currently running executable
1243 * so that a new one can be started
1244 */
1245
Adrian Hunter82b89772014-05-28 11:45:04 +03001246void __set_task_comm(struct task_struct *tsk, const char *buf, bool exec)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001247{
1248 task_lock(tsk);
KAMEZAWA Hiroyuki43d2b112012-01-10 15:08:09 -08001249 trace_task_rename(tsk, buf);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001250 strlcpy(tsk->comm, buf, sizeof(tsk->comm));
1251 task_unlock(tsk);
Adrian Hunter82b89772014-05-28 11:45:04 +03001252 perf_event_comm(tsk, exec);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001253}
1254
1255int flush_old_exec(struct linux_binprm * bprm)
1256{
Linus Torvalds221af7f2010-01-28 22:14:42 -08001257 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001258
1259 /*
1260 * Make sure we have a private signal table and that
1261 * we are unassociated from the previous thread group.
1262 */
1263 retval = de_thread(current);
1264 if (retval)
1265 goto out;
1266
Davidlohr Bueso6e399cd2015-04-16 12:47:59 -07001267 /*
1268 * Must be called _before_ exec_mmap() as bprm->mm is
1269 * not visibile until then. This also enables the update
1270 * to be lockless.
1271 */
Matt Helsley925d1c42008-04-29 01:01:36 -07001272 set_mm_exe_file(bprm->mm, bprm->file);
Davidlohr Bueso6e399cd2015-04-16 12:47:59 -07001273
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001275 * Release all of the old mmap stuff
1276 */
Oleg Nesterov3c77f842010-11-30 20:55:34 +01001277 acct_arg_size(bprm, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001278 retval = exec_mmap(bprm->mm);
1279 if (retval)
Al Virofd8328b2008-04-22 05:11:59 -04001280 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001281
1282 bprm->mm = NULL; /* We're using it now */
Linus Torvalds7ab02af2010-02-02 12:37:44 -08001283
Mathias Krausedac853a2011-06-09 20:05:18 +02001284 set_fs(USER_DS);
Zhang Yib88fae62014-01-23 15:55:57 -08001285 current->flags &= ~(PF_RANDOMIZE | PF_FORKNOEXEC | PF_KTHREAD |
1286 PF_NOFREEZE | PF_NO_SETAFFINITY);
Linus Torvalds7ab02af2010-02-02 12:37:44 -08001287 flush_thread();
1288 current->personality &= ~bprm->per_clear;
1289
Aleksa Saraic1df5a62016-12-21 16:26:24 +11001290 /*
1291 * We have to apply CLOEXEC before we change whether the process is
1292 * dumpable (in setup_new_exec) to avoid a race with a process in userspace
1293 * trying to access the should-be-closed file descriptors of a process
1294 * undergoing exec(2).
1295 */
1296 do_close_on_exec(current->files);
Linus Torvalds221af7f2010-01-28 22:14:42 -08001297 return 0;
1298
1299out:
1300 return retval;
1301}
1302EXPORT_SYMBOL(flush_old_exec);
1303
Al Viro1b5d7832011-06-19 12:49:47 -04001304void would_dump(struct linux_binprm *bprm, struct file *file)
1305{
Eric W. Biederman21245b82016-11-16 22:06:51 -06001306 struct inode *inode = file_inode(file);
1307 if (inode_permission(inode, MAY_READ) < 0) {
1308 struct user_namespace *old, *user_ns;
Al Viro1b5d7832011-06-19 12:49:47 -04001309 bprm->interp_flags |= BINPRM_FLAGS_ENFORCE_NONDUMP;
Eric W. Biederman21245b82016-11-16 22:06:51 -06001310
1311 /* Ensure mm->user_ns contains the executable */
1312 user_ns = old = bprm->mm->user_ns;
1313 while ((user_ns != &init_user_ns) &&
1314 !privileged_wrt_inode_uidgid(user_ns, inode))
1315 user_ns = user_ns->parent;
1316
1317 if (old != user_ns) {
1318 bprm->mm->user_ns = get_user_ns(user_ns);
1319 put_user_ns(old);
1320 }
1321 }
Al Viro1b5d7832011-06-19 12:49:47 -04001322}
1323EXPORT_SYMBOL(would_dump);
1324
Linus Torvalds221af7f2010-01-28 22:14:42 -08001325void setup_new_exec(struct linux_binprm * bprm)
1326{
Linus Torvalds221af7f2010-01-28 22:14:42 -08001327 arch_pick_mmap_layout(current->mm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001328
1329 /* This is the point of no return */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001330 current->sas_ss_sp = current->sas_ss_size = 0;
1331
Eric W. Biederman8e96e3b2012-03-03 21:17:15 -08001332 if (uid_eq(current_euid(), current_uid()) && gid_eq(current_egid(), current_gid()))
Kees Cooke579d2c2013-02-27 17:03:15 -08001333 set_dumpable(current->mm, SUID_DUMP_USER);
Alan Coxd6e71142005-06-23 00:09:43 -07001334 else
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001335 set_dumpable(current->mm, suid_dumpable);
Alan Coxd6e71142005-06-23 00:09:43 -07001336
Peter Zijlstrae041e322014-05-21 17:32:19 +02001337 perf_event_exec();
Adrian Hunter82b89772014-05-28 11:45:04 +03001338 __set_task_comm(current, kbasename(bprm->filename), true);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001339
Benjamin Herrenschmidt0551fbd2006-02-28 16:59:19 -08001340 /* Set the new mm task size. We have to do that late because it may
1341 * depend on TIF_32BIT which is only updated in flush_thread() on
1342 * some architectures like powerpc
1343 */
1344 current->mm->task_size = TASK_SIZE;
1345
David Howellsa6f76f22008-11-14 10:39:24 +11001346 /* install the new credentials */
Eric W. Biederman8e96e3b2012-03-03 21:17:15 -08001347 if (!uid_eq(bprm->cred->uid, current_euid()) ||
1348 !gid_eq(bprm->cred->gid, current_egid())) {
Marcel Holtmannd2d56c52007-08-17 21:47:58 +02001349 current->pdeath_signal = 0;
Al Viro1b5d7832011-06-19 12:49:47 -04001350 } else {
Al Viro1b5d7832011-06-19 12:49:47 -04001351 if (bprm->interp_flags & BINPRM_FLAGS_ENFORCE_NONDUMP)
1352 set_dumpable(current->mm, suid_dumpable);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001353 }
1354
1355 /* An exec changes our domain. We are no longer part of the thread
1356 group */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001357 current->self_exec_id++;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001358 flush_signal_handlers(current, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001359}
Linus Torvalds221af7f2010-01-28 22:14:42 -08001360EXPORT_SYMBOL(setup_new_exec);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001361
David Howellsa6f76f22008-11-14 10:39:24 +11001362/*
Oleg Nesterova2a84742009-09-05 11:17:13 -07001363 * Prepare credentials and lock ->cred_guard_mutex.
1364 * install_exec_creds() commits the new creds and drops the lock.
1365 * Or, if exec fails before, free_bprm() should release ->cred and
1366 * and unlock.
1367 */
1368int prepare_bprm_creds(struct linux_binprm *bprm)
1369{
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001370 if (mutex_lock_interruptible(&current->signal->cred_guard_mutex))
Oleg Nesterova2a84742009-09-05 11:17:13 -07001371 return -ERESTARTNOINTR;
1372
1373 bprm->cred = prepare_exec_creds();
1374 if (likely(bprm->cred))
1375 return 0;
1376
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001377 mutex_unlock(&current->signal->cred_guard_mutex);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001378 return -ENOMEM;
1379}
1380
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001381static void free_bprm(struct linux_binprm *bprm)
Oleg Nesterova2a84742009-09-05 11:17:13 -07001382{
1383 free_arg_pages(bprm);
1384 if (bprm->cred) {
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001385 mutex_unlock(&current->signal->cred_guard_mutex);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001386 abort_creds(bprm->cred);
1387 }
Oleg Nesterov63e46b92014-01-23 15:55:51 -08001388 if (bprm->file) {
1389 allow_write_access(bprm->file);
1390 fput(bprm->file);
1391 }
Kees Cookb66c5982012-12-20 15:05:16 -08001392 /* If a binfmt changed the interp, free it. */
1393 if (bprm->interp != bprm->filename)
1394 kfree(bprm->interp);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001395 kfree(bprm);
1396}
1397
Kees Cookb66c5982012-12-20 15:05:16 -08001398int bprm_change_interp(char *interp, struct linux_binprm *bprm)
1399{
1400 /* If a binfmt changed the interp, free it first. */
1401 if (bprm->interp != bprm->filename)
1402 kfree(bprm->interp);
1403 bprm->interp = kstrdup(interp, GFP_KERNEL);
1404 if (!bprm->interp)
1405 return -ENOMEM;
1406 return 0;
1407}
1408EXPORT_SYMBOL(bprm_change_interp);
1409
Oleg Nesterova2a84742009-09-05 11:17:13 -07001410/*
David Howellsa6f76f22008-11-14 10:39:24 +11001411 * install the new credentials for this executable
1412 */
1413void install_exec_creds(struct linux_binprm *bprm)
1414{
1415 security_bprm_committing_creds(bprm);
1416
1417 commit_creds(bprm->cred);
1418 bprm->cred = NULL;
Stephane Eranian2976b102013-06-20 11:36:28 +02001419
1420 /*
1421 * Disable monitoring for regular users
1422 * when executing setuid binaries. Must
1423 * wait until new credentials are committed
1424 * by commit_creds() above
1425 */
1426 if (get_dumpable(current->mm) != SUID_DUMP_USER)
1427 perf_event_exit_task(current);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001428 /*
1429 * cred_guard_mutex must be held at least to this point to prevent
David Howellsa6f76f22008-11-14 10:39:24 +11001430 * ptrace_attach() from altering our determination of the task's
Oleg Nesterova2a84742009-09-05 11:17:13 -07001431 * credentials; any time after this it may be unlocked.
1432 */
David Howellsa6f76f22008-11-14 10:39:24 +11001433 security_bprm_committed_creds(bprm);
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001434 mutex_unlock(&current->signal->cred_guard_mutex);
David Howellsa6f76f22008-11-14 10:39:24 +11001435}
1436EXPORT_SYMBOL(install_exec_creds);
1437
1438/*
1439 * determine how safe it is to execute the proposed program
KOSAKI Motohiro9b1bf122010-10-27 15:34:08 -07001440 * - the caller must hold ->cred_guard_mutex to protect against
Kees Cookc2e1f2e2014-06-05 00:23:17 -07001441 * PTRACE_ATTACH or seccomp thread-sync
David Howellsa6f76f22008-11-14 10:39:24 +11001442 */
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001443static void check_unsafe_exec(struct linux_binprm *bprm)
David Howellsa6f76f22008-11-14 10:39:24 +11001444{
David Howells0bf2f3a2009-02-06 11:45:46 +00001445 struct task_struct *p = current, *t;
Al Virof1191b52009-03-30 07:35:18 -04001446 unsigned n_fs;
David Howellsa6f76f22008-11-14 10:39:24 +11001447
Tejun Heo4b9d33e2011-06-17 16:50:38 +02001448 if (p->ptrace) {
Eric W. Biedermane747b4a2016-11-14 18:48:07 -06001449 if (ptracer_capable(p, current_user_ns()))
Tejun Heo4b9d33e2011-06-17 16:50:38 +02001450 bprm->unsafe |= LSM_UNSAFE_PTRACE_CAP;
1451 else
1452 bprm->unsafe |= LSM_UNSAFE_PTRACE;
1453 }
David Howellsa6f76f22008-11-14 10:39:24 +11001454
Andy Lutomirski259e5e62012-04-12 16:47:50 -05001455 /*
1456 * This isn't strictly necessary, but it makes it harder for LSMs to
1457 * mess up.
1458 */
Kees Cook1d4457f2014-05-21 15:23:46 -07001459 if (task_no_new_privs(current))
Andy Lutomirski259e5e62012-04-12 16:47:50 -05001460 bprm->unsafe |= LSM_UNSAFE_NO_NEW_PRIVS;
1461
Oleg Nesterov83f62a22014-01-23 15:55:49 -08001462 t = p;
David Howells0bf2f3a2009-02-06 11:45:46 +00001463 n_fs = 1;
Nick Piggin2a4419b2010-08-18 04:37:33 +10001464 spin_lock(&p->fs->lock);
Oleg Nesterov437f7fd2009-04-24 01:02:45 +02001465 rcu_read_lock();
Oleg Nesterov83f62a22014-01-23 15:55:49 -08001466 while_each_thread(p, t) {
David Howells0bf2f3a2009-02-06 11:45:46 +00001467 if (t->fs == p->fs)
1468 n_fs++;
David Howells0bf2f3a2009-02-06 11:45:46 +00001469 }
Oleg Nesterov437f7fd2009-04-24 01:02:45 +02001470 rcu_read_unlock();
David Howells0bf2f3a2009-02-06 11:45:46 +00001471
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001472 if (p->fs->users > n_fs)
David Howellsa6f76f22008-11-14 10:39:24 +11001473 bprm->unsafe |= LSM_UNSAFE_SHARE;
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001474 else
1475 p->fs->in_exec = 1;
Nick Piggin2a4419b2010-08-18 04:37:33 +10001476 spin_unlock(&p->fs->lock);
David Howellsa6f76f22008-11-14 10:39:24 +11001477}
1478
Jann Horn8b01fc82015-04-19 02:48:39 +02001479static void bprm_fill_uid(struct linux_binprm *bprm)
1480{
1481 struct inode *inode;
1482 unsigned int mode;
1483 kuid_t uid;
1484 kgid_t gid;
1485
Kees Cookcb6fd682016-05-17 12:14:39 -07001486 /*
1487 * Since this can be called multiple times (via prepare_binprm),
1488 * we must clear any previous work done when setting set[ug]id
1489 * bits from any earlier bprm->file uses (for example when run
1490 * first for a setuid script then again for its interpreter).
1491 */
Jann Horn8b01fc82015-04-19 02:48:39 +02001492 bprm->cred->euid = current_euid();
1493 bprm->cred->egid = current_egid();
1494
Andy Lutomirski380cf5b2016-06-23 16:41:05 -05001495 if (!mnt_may_suid(bprm->file->f_path.mnt))
Jann Horn8b01fc82015-04-19 02:48:39 +02001496 return;
1497
1498 if (task_no_new_privs(current))
1499 return;
1500
1501 inode = file_inode(bprm->file);
1502 mode = READ_ONCE(inode->i_mode);
1503 if (!(mode & (S_ISUID|S_ISGID)))
1504 return;
1505
1506 /* Be careful if suid/sgid is set */
Al Viro59551022016-01-22 15:40:57 -05001507 inode_lock(inode);
Jann Horn8b01fc82015-04-19 02:48:39 +02001508
1509 /* reload atomically mode/uid/gid now that lock held */
1510 mode = inode->i_mode;
1511 uid = inode->i_uid;
1512 gid = inode->i_gid;
Al Viro59551022016-01-22 15:40:57 -05001513 inode_unlock(inode);
Jann Horn8b01fc82015-04-19 02:48:39 +02001514
1515 /* We ignore suid/sgid if there are no mappings for them in the ns */
1516 if (!kuid_has_mapping(bprm->cred->user_ns, uid) ||
1517 !kgid_has_mapping(bprm->cred->user_ns, gid))
1518 return;
1519
1520 if (mode & S_ISUID) {
1521 bprm->per_clear |= PER_CLEAR_ON_SETID;
1522 bprm->cred->euid = uid;
1523 }
1524
1525 if ((mode & (S_ISGID | S_IXGRP)) == (S_ISGID | S_IXGRP)) {
1526 bprm->per_clear |= PER_CLEAR_ON_SETID;
1527 bprm->cred->egid = gid;
1528 }
1529}
1530
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001531/*
1532 * Fill the binprm structure from the inode.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001533 * Check permissions, then read the first 128 (BINPRM_BUF_SIZE) bytes
David Howellsa6f76f22008-11-14 10:39:24 +11001534 *
1535 * This may be called multiple times for binary chains (scripts for example).
Linus Torvalds1da177e2005-04-16 15:20:36 -07001536 */
1537int prepare_binprm(struct linux_binprm *bprm)
1538{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001539 int retval;
1540
Jann Horn8b01fc82015-04-19 02:48:39 +02001541 bprm_fill_uid(bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001542
1543 /* fill in binprm security blob */
David Howellsa6f76f22008-11-14 10:39:24 +11001544 retval = security_bprm_set_creds(bprm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001545 if (retval)
1546 return retval;
David Howellsa6f76f22008-11-14 10:39:24 +11001547 bprm->cred_prepared = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001548
David Howellsa6f76f22008-11-14 10:39:24 +11001549 memset(bprm->buf, 0, BINPRM_BUF_SIZE);
1550 return kernel_read(bprm->file, 0, bprm->buf, BINPRM_BUF_SIZE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001551}
1552
1553EXPORT_SYMBOL(prepare_binprm);
1554
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001555/*
1556 * Arguments are '\0' separated strings found at the location bprm->p
1557 * points to; chop off the first by relocating brpm->p to right after
1558 * the first '\0' encountered.
1559 */
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001560int remove_arg_zero(struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001561{
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001562 int ret = 0;
1563 unsigned long offset;
1564 char *kaddr;
1565 struct page *page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001566
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001567 if (!bprm->argc)
1568 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001569
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001570 do {
1571 offset = bprm->p & ~PAGE_MASK;
1572 page = get_arg_page(bprm, bprm->p, 0);
1573 if (!page) {
1574 ret = -EFAULT;
1575 goto out;
1576 }
Cong Wange8e3c3d2011-11-25 23:14:27 +08001577 kaddr = kmap_atomic(page);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001578
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001579 for (; offset < PAGE_SIZE && kaddr[offset];
1580 offset++, bprm->p++)
1581 ;
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001582
Cong Wange8e3c3d2011-11-25 23:14:27 +08001583 kunmap_atomic(kaddr);
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001584 put_arg_page(page);
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001585 } while (offset == PAGE_SIZE);
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001586
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001587 bprm->p++;
1588 bprm->argc--;
1589 ret = 0;
Nick Piggin4fc75ff2007-05-08 00:25:16 -07001590
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001591out:
1592 return ret;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001593}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001594EXPORT_SYMBOL(remove_arg_zero);
1595
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001596#define printable(c) (((c)=='\t') || ((c)=='\n') || (0x20<=(c) && (c)<=0x7e))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001597/*
1598 * cycle the list of binary formats handler, until one recognizes the image
1599 */
Al Viro3c456bf2012-10-20 21:53:31 -04001600int search_binary_handler(struct linux_binprm *bprm)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001601{
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001602 bool need_retry = IS_ENABLED(CONFIG_MODULES);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001603 struct linux_binfmt *fmt;
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001604 int retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001605
Kees Cookd7402692012-12-17 16:03:20 -08001606 /* This allows 4 levels of binfmt rewrites before failing hard. */
Oleg Nesterov131b2f92013-09-11 14:24:39 -07001607 if (bprm->recursion_depth > 5)
Kees Cookd7402692012-12-17 16:03:20 -08001608 return -ELOOP;
1609
Linus Torvalds1da177e2005-04-16 15:20:36 -07001610 retval = security_bprm_check(bprm);
1611 if (retval)
1612 return retval;
1613
Linus Torvalds1da177e2005-04-16 15:20:36 -07001614 retval = -ENOENT;
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001615 retry:
1616 read_lock(&binfmt_lock);
1617 list_for_each_entry(fmt, &formats, lh) {
1618 if (!try_module_get(fmt->module))
1619 continue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001620 read_unlock(&binfmt_lock);
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001621 bprm->recursion_depth++;
1622 retval = fmt->load_binary(bprm);
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001623 read_lock(&binfmt_lock);
1624 put_binfmt(fmt);
Al Viro19d860a2014-05-04 20:11:36 -04001625 bprm->recursion_depth--;
1626 if (retval < 0 && !bprm->mm) {
1627 /* we got to flush_old_exec() and failed after it */
1628 read_unlock(&binfmt_lock);
1629 force_sigsegv(SIGSEGV, current);
1630 return retval;
1631 }
1632 if (retval != -ENOEXEC || !bprm->file) {
1633 read_unlock(&binfmt_lock);
1634 return retval;
1635 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001636 }
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001637 read_unlock(&binfmt_lock);
1638
Al Viro19d860a2014-05-04 20:11:36 -04001639 if (need_retry) {
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001640 if (printable(bprm->buf[0]) && printable(bprm->buf[1]) &&
1641 printable(bprm->buf[2]) && printable(bprm->buf[3]))
1642 return retval;
Oleg Nesterov4e0621a2013-09-11 14:24:45 -07001643 if (request_module("binfmt-%04x", *(ushort *)(bprm->buf + 2)) < 0)
1644 return retval;
Oleg Nesterovcb7b6b12013-09-11 14:24:44 -07001645 need_retry = false;
1646 goto retry;
1647 }
1648
Linus Torvalds1da177e2005-04-16 15:20:36 -07001649 return retval;
1650}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001651EXPORT_SYMBOL(search_binary_handler);
1652
Oleg Nesterov5d1baf32013-09-11 14:24:38 -07001653static int exec_binprm(struct linux_binprm *bprm)
1654{
1655 pid_t old_pid, old_vpid;
1656 int ret;
1657
1658 /* Need to fetch pid before load_binary changes it */
1659 old_pid = current->pid;
1660 rcu_read_lock();
1661 old_vpid = task_pid_nr_ns(current, task_active_pid_ns(current->parent));
1662 rcu_read_unlock();
1663
1664 ret = search_binary_handler(bprm);
1665 if (ret >= 0) {
Linus Torvalds3eaded862013-11-21 19:18:14 -08001666 audit_bprm(bprm);
Oleg Nesterov5d1baf32013-09-11 14:24:38 -07001667 trace_sched_process_exec(current, old_pid, bprm);
1668 ptrace_event(PTRACE_EVENT_EXEC, old_vpid);
Oleg Nesterov9beb2662013-09-11 14:24:40 -07001669 proc_exec_connector(current);
Oleg Nesterov5d1baf32013-09-11 14:24:38 -07001670 }
1671
1672 return ret;
1673}
1674
Linus Torvalds1da177e2005-04-16 15:20:36 -07001675/*
1676 * sys_execve() executes a new program.
1677 */
David Drysdale51f39a12014-12-12 16:57:29 -08001678static int do_execveat_common(int fd, struct filename *filename,
1679 struct user_arg_ptr argv,
1680 struct user_arg_ptr envp,
1681 int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001682{
David Drysdale51f39a12014-12-12 16:57:29 -08001683 char *pathbuf = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001684 struct linux_binprm *bprm;
1685 struct file *file;
Al Viro3b125382008-04-22 05:31:30 -04001686 struct files_struct *displaced;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001687 int retval;
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +04001688
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001689 if (IS_ERR(filename))
1690 return PTR_ERR(filename);
1691
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +04001692 /*
1693 * We move the actual failure in case of RLIMIT_NPROC excess from
1694 * set*uid() to execve() because too many poorly written programs
1695 * don't check setuid() return code. Here we additionally recheck
1696 * whether NPROC limit is still exceeded.
1697 */
1698 if ((current->flags & PF_NPROC_EXCEEDED) &&
Oleg Nesterovbd9d43f2013-07-03 15:08:34 -07001699 atomic_read(&current_user()->processes) > rlimit(RLIMIT_NPROC)) {
Vasiliy Kulikov72fa5992011-08-08 19:02:04 +04001700 retval = -EAGAIN;
1701 goto out_ret;
1702 }
1703
1704 /* We're below the limit (still or again), so we don't want to make
1705 * further execve() calls fail. */
1706 current->flags &= ~PF_NPROC_EXCEEDED;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001707
Al Viro3b125382008-04-22 05:31:30 -04001708 retval = unshare_files(&displaced);
Al Virofd8328b2008-04-22 05:11:59 -04001709 if (retval)
1710 goto out_ret;
1711
Linus Torvalds1da177e2005-04-16 15:20:36 -07001712 retval = -ENOMEM;
Oliver Neukum11b0b5a2006-03-25 03:08:13 -08001713 bprm = kzalloc(sizeof(*bprm), GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001714 if (!bprm)
Al Virofd8328b2008-04-22 05:11:59 -04001715 goto out_files;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001716
Oleg Nesterova2a84742009-09-05 11:17:13 -07001717 retval = prepare_bprm_creds(bprm);
1718 if (retval)
David Howellsa6f76f22008-11-14 10:39:24 +11001719 goto out_free;
Al Viro498052b2009-03-30 07:20:30 -04001720
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001721 check_unsafe_exec(bprm);
Oleg Nesterova2a84742009-09-05 11:17:13 -07001722 current->in_execve = 1;
David Howellsa6f76f22008-11-14 10:39:24 +11001723
David Drysdale51f39a12014-12-12 16:57:29 -08001724 file = do_open_execat(fd, filename, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001725 retval = PTR_ERR(file);
1726 if (IS_ERR(file))
Al Viro498052b2009-03-30 07:20:30 -04001727 goto out_unmark;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001728
1729 sched_exec();
1730
Linus Torvalds1da177e2005-04-16 15:20:36 -07001731 bprm->file = file;
David Drysdale51f39a12014-12-12 16:57:29 -08001732 if (fd == AT_FDCWD || filename->name[0] == '/') {
1733 bprm->filename = filename->name;
1734 } else {
1735 if (filename->name[0] == '\0')
1736 pathbuf = kasprintf(GFP_TEMPORARY, "/dev/fd/%d", fd);
1737 else
1738 pathbuf = kasprintf(GFP_TEMPORARY, "/dev/fd/%d/%s",
1739 fd, filename->name);
1740 if (!pathbuf) {
1741 retval = -ENOMEM;
1742 goto out_unmark;
1743 }
1744 /*
1745 * Record that a name derived from an O_CLOEXEC fd will be
1746 * inaccessible after exec. Relies on having exclusive access to
1747 * current->files (due to unshare_files above).
1748 */
1749 if (close_on_exec(fd, rcu_dereference_raw(current->files->fdt)))
1750 bprm->interp_flags |= BINPRM_FLAGS_PATH_INACCESSIBLE;
1751 bprm->filename = pathbuf;
1752 }
1753 bprm->interp = bprm->filename;
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001754
1755 retval = bprm_mm_init(bprm);
1756 if (retval)
Oleg Nesterov63e46b92014-01-23 15:55:51 -08001757 goto out_unmark;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001758
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001759 bprm->argc = count(argv, MAX_ARG_STRINGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001760 if ((retval = bprm->argc) < 0)
David Howellsa6f76f22008-11-14 10:39:24 +11001761 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001762
Ollie Wildb6a2fea2007-07-19 01:48:16 -07001763 bprm->envc = count(envp, MAX_ARG_STRINGS);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001764 if ((retval = bprm->envc) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001765 goto out;
1766
1767 retval = prepare_binprm(bprm);
1768 if (retval < 0)
1769 goto out;
1770
1771 retval = copy_strings_kernel(1, &bprm->filename, bprm);
1772 if (retval < 0)
1773 goto out;
1774
1775 bprm->exec = bprm->p;
1776 retval = copy_strings(bprm->envc, envp, bprm);
1777 if (retval < 0)
1778 goto out;
1779
1780 retval = copy_strings(bprm->argc, argv, bprm);
1781 if (retval < 0)
1782 goto out;
1783
Eric W. Biederman21245b82016-11-16 22:06:51 -06001784 would_dump(bprm, bprm->file);
1785
Oleg Nesterov5d1baf32013-09-11 14:24:38 -07001786 retval = exec_binprm(bprm);
David Howellsa6f76f22008-11-14 10:39:24 +11001787 if (retval < 0)
1788 goto out;
1789
1790 /* execve succeeded */
Al Viro498052b2009-03-30 07:20:30 -04001791 current->fs->in_exec = 0;
Kentaro Takedaf9ce1f12009-02-05 17:18:11 +09001792 current->in_execve = 0;
David Howellsa6f76f22008-11-14 10:39:24 +11001793 acct_update_integrals(current);
Rik van Riel82727012013-10-07 11:29:28 +01001794 task_numa_free(current);
David Howellsa6f76f22008-11-14 10:39:24 +11001795 free_bprm(bprm);
David Drysdale51f39a12014-12-12 16:57:29 -08001796 kfree(pathbuf);
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001797 putname(filename);
David Howellsa6f76f22008-11-14 10:39:24 +11001798 if (displaced)
1799 put_files_struct(displaced);
1800 return retval;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001801
1802out:
Oleg Nesterov3c77f842010-11-30 20:55:34 +01001803 if (bprm->mm) {
1804 acct_arg_size(bprm, 0);
1805 mmput(bprm->mm);
1806 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001807
Al Viro498052b2009-03-30 07:20:30 -04001808out_unmark:
Oleg Nesterov9e00cdb2014-01-23 15:55:50 -08001809 current->fs->in_exec = 0;
Kentaro Takedaf9ce1f12009-02-05 17:18:11 +09001810 current->in_execve = 0;
David Howellsa6f76f22008-11-14 10:39:24 +11001811
1812out_free:
Al Viro08a6fac2008-05-10 16:38:25 -04001813 free_bprm(bprm);
David Drysdale51f39a12014-12-12 16:57:29 -08001814 kfree(pathbuf);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001815
Al Virofd8328b2008-04-22 05:11:59 -04001816out_files:
Al Viro3b125382008-04-22 05:31:30 -04001817 if (displaced)
1818 reset_files_struct(displaced);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001819out_ret:
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001820 putname(filename);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001821 return retval;
1822}
1823
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001824int do_execve(struct filename *filename,
Oleg Nesterovba2d01622011-03-06 18:02:37 +01001825 const char __user *const __user *__argv,
Al Viroda3d4c52012-10-20 21:49:33 -04001826 const char __user *const __user *__envp)
Oleg Nesterovba2d01622011-03-06 18:02:37 +01001827{
Oleg Nesterov0e028462011-03-06 18:02:54 +01001828 struct user_arg_ptr argv = { .ptr.native = __argv };
1829 struct user_arg_ptr envp = { .ptr.native = __envp };
David Drysdale51f39a12014-12-12 16:57:29 -08001830 return do_execveat_common(AT_FDCWD, filename, argv, envp, 0);
1831}
1832
1833int do_execveat(int fd, struct filename *filename,
1834 const char __user *const __user *__argv,
1835 const char __user *const __user *__envp,
1836 int flags)
1837{
1838 struct user_arg_ptr argv = { .ptr.native = __argv };
1839 struct user_arg_ptr envp = { .ptr.native = __envp };
1840
1841 return do_execveat_common(fd, filename, argv, envp, flags);
Oleg Nesterovba2d01622011-03-06 18:02:37 +01001842}
1843
Oleg Nesterov0e028462011-03-06 18:02:54 +01001844#ifdef CONFIG_COMPAT
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001845static int compat_do_execve(struct filename *filename,
Al Viro38b983b2012-09-30 13:38:55 -04001846 const compat_uptr_t __user *__argv,
Al Virod03d26e2012-10-20 21:46:25 -04001847 const compat_uptr_t __user *__envp)
Oleg Nesterov0e028462011-03-06 18:02:54 +01001848{
1849 struct user_arg_ptr argv = {
1850 .is_compat = true,
1851 .ptr.compat = __argv,
1852 };
1853 struct user_arg_ptr envp = {
1854 .is_compat = true,
1855 .ptr.compat = __envp,
1856 };
David Drysdale51f39a12014-12-12 16:57:29 -08001857 return do_execveat_common(AT_FDCWD, filename, argv, envp, 0);
1858}
1859
1860static int compat_do_execveat(int fd, struct filename *filename,
1861 const compat_uptr_t __user *__argv,
1862 const compat_uptr_t __user *__envp,
1863 int flags)
1864{
1865 struct user_arg_ptr argv = {
1866 .is_compat = true,
1867 .ptr.compat = __argv,
1868 };
1869 struct user_arg_ptr envp = {
1870 .is_compat = true,
1871 .ptr.compat = __envp,
1872 };
1873 return do_execveat_common(fd, filename, argv, envp, flags);
Oleg Nesterov0e028462011-03-06 18:02:54 +01001874}
1875#endif
1876
Oleg Nesterov964ee7d2009-09-23 15:56:59 -07001877void set_binfmt(struct linux_binfmt *new)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001878{
Hiroshi Shimamoto801460d2009-09-23 15:57:41 -07001879 struct mm_struct *mm = current->mm;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001880
Hiroshi Shimamoto801460d2009-09-23 15:57:41 -07001881 if (mm->binfmt)
1882 module_put(mm->binfmt->module);
1883
1884 mm->binfmt = new;
Oleg Nesterov964ee7d2009-09-23 15:56:59 -07001885 if (new)
1886 __module_get(new->module);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001887}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001888EXPORT_SYMBOL(set_binfmt);
1889
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001890/*
Oleg Nesterov7288e112014-01-23 15:55:32 -08001891 * set_dumpable stores three-value SUID_DUMP_* into mm->flags.
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001892 */
1893void set_dumpable(struct mm_struct *mm, int value)
1894{
Oleg Nesterovabacd2f2014-01-23 15:55:31 -08001895 unsigned long old, new;
1896
Oleg Nesterov7288e112014-01-23 15:55:32 -08001897 if (WARN_ON((unsigned)value > SUID_DUMP_ROOT))
1898 return;
1899
Oleg Nesterovabacd2f2014-01-23 15:55:31 -08001900 do {
1901 old = ACCESS_ONCE(mm->flags);
Oleg Nesterov7288e112014-01-23 15:55:32 -08001902 new = (old & ~MMF_DUMPABLE_MASK) | value;
Oleg Nesterovabacd2f2014-01-23 15:55:31 -08001903 } while (cmpxchg(&mm->flags, old, new) != old);
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001904}
Kawai, Hidehiro6c5d5232007-07-19 01:48:27 -07001905
Al Viro38b983b2012-09-30 13:38:55 -04001906SYSCALL_DEFINE3(execve,
1907 const char __user *, filename,
1908 const char __user *const __user *, argv,
1909 const char __user *const __user *, envp)
1910{
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001911 return do_execve(getname(filename), argv, envp);
Al Viro38b983b2012-09-30 13:38:55 -04001912}
David Drysdale51f39a12014-12-12 16:57:29 -08001913
1914SYSCALL_DEFINE5(execveat,
1915 int, fd, const char __user *, filename,
1916 const char __user *const __user *, argv,
1917 const char __user *const __user *, envp,
1918 int, flags)
1919{
1920 int lookup_flags = (flags & AT_EMPTY_PATH) ? LOOKUP_EMPTY : 0;
1921
1922 return do_execveat(fd,
1923 getname_flags(filename, lookup_flags, NULL),
1924 argv, envp, flags);
1925}
1926
Al Viro38b983b2012-09-30 13:38:55 -04001927#ifdef CONFIG_COMPAT
Heiko Carstens625b1d72014-03-04 10:53:50 +01001928COMPAT_SYSCALL_DEFINE3(execve, const char __user *, filename,
1929 const compat_uptr_t __user *, argv,
1930 const compat_uptr_t __user *, envp)
Al Viro38b983b2012-09-30 13:38:55 -04001931{
Linus Torvaldsc4ad8f92014-02-05 12:54:53 -08001932 return compat_do_execve(getname(filename), argv, envp);
Al Viro38b983b2012-09-30 13:38:55 -04001933}
David Drysdale51f39a12014-12-12 16:57:29 -08001934
1935COMPAT_SYSCALL_DEFINE5(execveat, int, fd,
1936 const char __user *, filename,
1937 const compat_uptr_t __user *, argv,
1938 const compat_uptr_t __user *, envp,
1939 int, flags)
1940{
1941 int lookup_flags = (flags & AT_EMPTY_PATH) ? LOOKUP_EMPTY : 0;
1942
1943 return compat_do_execveat(fd,
1944 getname_flags(filename, lookup_flags, NULL),
1945 argv, envp, flags);
1946}
Al Viro38b983b2012-09-30 13:38:55 -04001947#endif